A preliminary characterization of Bembidion perspicuum LeConte , with a reclassification of related species ( Coleoptera , Carabidae ) north of México

Bembidion perspicuum LeConte is a species distinct from B. transversale Dejean, the names of which have been synonymized in the past. In the United States, B. perspicuum occurs from Oregon through Texas along the shores of rivers and creeks. We show that it is consistently diff erent at 40 bases in DNA sequences of cytochrome oxidase I and 4 bases in 28S ribosomal DNA, as well as in subtle morphological characters. In addition, we review the classifi cation of related species in the B. transversale and B. mexicanum species groups in America north of México. In the United States and Canada, the B. transversale group includes B. transversale, B. perspicuum, and B. sarpedon Casey; the B. mexicanum group includes B. mexicanum Dejean, B. lugubre LeConte, and B. pernotum Casey.


Introduction
Members of the Bembidion transversale group are large (6-8mm) Bembidion occurring along creek and river shores throughout much of western North America, east across Canada and adjacent United States to Nova Scotia and Newfoundland.A recent catalogue (1993) lists two species within the group, B. pernotum Casey (known from Colorado and New Mexico) and the widespread B. transversale Dejean.Th e related Bembidion mexicanum group contains very similar members whose combined range extends from South Dakota to Central America; in the United States this group is considered to consist of only the nominotypical species (Bousquet and Larochelle 1993).
In sequencing 28S rDNA (28S) and cytochrome oxidase I (COI) from specimens of the B. transversale group from several North American localities, and examination of morphological characters, it became evident that "B.transversale" is in reality several species.
Th is paper is the fi rst of several that will attempt to resolve the complexity of both the B. transversale and B. mexicanum groups.We herein report the discovery that the common form of the B. transversale group throughout much of the southwest United States is a separate species, B. perspicuum LeConte, distinct both morphologically and molecularly from the widespread and more northern B. transversale.We also briefl y document the other species in the two species groups, so that a revised classifi cation is available for future publications.A later paper (Maddison, in prep.) will revise the group in America north of México, including forms along the west coast of North America, and give a fuller treatment of all species.In advance of this more complete revision, however, we wish to document the distinctiveness of B. perspicuum, as this is one of the primary species studied in a forthcoming paper (Wild, Kanda, McKenna, Farrell, and Maddison, in prep.).

Methods
Members of the B. transversale and B. mexicanum groups were examined from or are deposited in the collections listed below.Each collection's listing begins with the coden used in the text.

DRM
David R. Methods for studying adult structures, and terms used, are given in Maddison (1993).Photographs of body parts were taken with a Leica Z6 and JVC KY-F75U camera using Microvision's Cartograph software for extended depth of fi eld (EDF) processing; the images thus potentially have some artifacts caused by the EDF algorithm.
In the section on classifi cation, we do not list which junior synonyms are so considered for the fi rst time.All names listed under B. transversale have previously been considered synonyms of B. transversale; the same applies to B. mexicanum.However, we have not conducted the historical analysis required to determine if any of the names herein considered junior synonyms of B. lugubre have ever been considered as such previously, although they and B. lugubre have been considered synonyms of B. mexicanum; the same applies to B. perspicuum and B. sarpedon (which have been considered synonyms of B. transversale, as have all of their synonyms).
Taxon sampling for DNA studies.We sequenced DNA from 28 specimens of the B. transversale group from Nova Scotia, Alberta, British Columbia, as well as 10 states in western U.S.A. (Table 1 DNA sequencing.Methods for obtaining DNA sequences are described in Maddison (2008).In brief, we obtained ca.1000 bases of sequence data in the D1 through D3 domains of 28S and between 600 and 1400 bases of COI.Fragments for these genes were amplifi ed using the Polymerase Chain Reaction on an Eppendorf Mastercycler Th ermal Cycler, using either Eppendorf Hotmaster Taq or TaKaRa Ex Taq and the basic protocols recommended by the manufacturers.Primers and details of the cycling reactions used are given in Maddison (2008).In particular, we used the primer pair LS58F and LS998R and the pair NLF184/21 and LS1041R to amplify and sequence 28S rDNA.For COI, two amplifi cation strategies were used: amplifi cation with B1490 and Pat, and sequencing with those primers plus Bcoi2R and Jerry (see Maddison, 2008), or amplifi cation and sequencing with the LCO1490 and HCO2198 primers (Hebert et al. 2003).Amplifi ed products were cleaned, quantifi ed, and sequenced at the University of Arizona's Genomic and Technology Core Facility using either a 3730 or 3730 XL Applied Biosystems automatic sequencer.
Assembly of multiple chromatograms for each gene fragment and initial base calls were made with Phred (Green and Ewing 2002) and Phrap (Green 1999) as orchestrated by Mesquite's Chromaseq package (Maddison and Maddison 2009a;Maddison and Maddison 2009b), with subsequent modifi cations by Chromaseq and manual inspection.Multiple peaks at a single position in both reads were coded using IUPAC ambiguity codes.Sequences have been deposited in GenBank with accession numbers GU454737 through GU454797.In addition, the 28S sequence for B. transversale voucher number DNA2157 is from GenBank, number EU677688.

Alignment and phylogenetic analysis
Alignment was not diffi cult for either gene, as there was no evidence of insertion or deletions in the history of these sequences.Aligned matrices are available at http:// bembidion.org/transversaleGroup.Th ree matrices were examined: (1) 28S rDNA; (2) COI with all data; (3) COI with only the fi rst ca.600 bases that were sequenced for all specimens, the so-called "barcode" piece (Hebert, Cywinska, Ball and DeWaard 2003).Th e second matrix (full-length COI) has the last half of COI missing for 10 of the 28 B. transversale group specimens as well as B. sejunctum and B. commotum; only 77.7% of the cells have data in them, as opposed to 98.6% for the shorter COI matrix and 99.6% for the 28S matrix.Phylogenetic analysis was conducted using parsimony, likelihood, and Bayesian methods.
Most-parsimonious trees were sought using PAUP* (Swoff ord 2002).For each search, 4000 replicates were conducted, each beginning with a starting tree formed with the random addition sequence option; each replicate was allowed to save no more than 25 trees.Th e number of most parsimonious trees found for each matrix ranged from 2 to 63,341; shortest trees were found in at least 2000 of the replicates.Bootstrap parsimony analyses were conducted using 1000 bootstrap replicates, with each replicate consisting of fi ve search replicates beginning with starting trees formed with the random addition sequence option, and with each search replicate allowed to save no more than 25 trees.
Models of nucleotide evolution where chosen with the aid of ModelTest (Posada 2005).Th e model chosen by the Akaike Information Criterion (AIC) for 28S rDNA was an HKY 2-parameter rate matrix with a proportion of sites being invariant (HKY+I).For COI, models were inferred treating each codon position separately as well as pooling all positions; phylogenetic analyses were conducted under both model structures.First positions are best fi t by a General Time Reversible (GTR) rate matrix with a proportion of sites being invariant (GTR+I); second positions a 1-parameter F81 rate matrix with a proportion of invariant sites (F81+I), and third positions with an HKY rate matrix and site-to-site rate variation following a gamma distribution (HKY+Γ).If instead all positions were pooled for COI, the model chosen was GTR+I.
Maximum Likelihood analyses were conducted with GARLI version 0.96 (Zwickl 2006).A GTR+I model was used for all COI sites.Th e maximum likelihood tree was sought using 500 search replicates.Maximum likelihood bootstrap analyses with 500 replicates were also conducted for all three matrices.
Bayesian analyses were conducted using MrBayes version 3.1.2(Huelsenbeck and Ronquist 2005).For each of the two COI matrices, the analyses were done both with separate models for each codon position, as well as GTR+I for all sites; thus, four Bayesian analyses were conducted for COI and one for 28S.A single MrBayes MCM-CMC search was conducted for each analysis, each with two runs of four chains each, for 10 million generations, with trees sampled every 1000 generations.All searches converged, as judged by an average standard deviation of split frequencies of less than 0.006 (Huelsenbeck and Ronquist 2005) and by stabilization of the likelihood scores and all parameter values as judged by visualization tools in Tracer (Rambaut and Drummond 2004), except for the full-length COI matrix with codon positions treated separately.Th at analysis did not converge after 100 million generations, and was discarded.A burn-in sample of the fi rst 7.5 million generations was excluded from the other analyses; the number of trees sampled for each analysis was thus 5,000.

Results of molecular analyses
Th e two forms, B. transversale and B perspicuum, consistently diff er by 4 bases in 28S, and 40 bases in COI (of which 14 fall within the fi rst 600 bases of the gene; Table 2).Th e base diff erences in COI yield predictions of two sites at which amino acids diff er in the protein.Th ese fi xed diff erences in both genes are also evident between the two specimens of B. transversale and two of B. perspicuum sampled microsympatrically from the same shore of the Arkansas River at Texas Creek, Colorado (38.4106°N 105.5844°W).
As a cautionary note, there is evidence (in the form of double peaks at sites in the chromatograms of both reads) for multiple copies of COI in one Bembidion perspicuum from Arizona (specimen 1120), one B. transversale from Montana (specimen 2163) and another from Utah (specimen 2183).Th is is likely to indicate nuclear copies of COI, and casts some doubt on the location of any particular sequenced copy.Nonetheless, the consistency within each species in the sequenced COI provides evidence of lack of gene fl ow even if some copies are nuclear.
Th e phylogenetic analyses (Fig. 1; Table 3) indicate clearly that the two forms are reciprocally monophyletic.Th is is supported in all analyses for both genes.Th e geographic distributions of sequenced individuals of these two forms are shown in Fig. 2.
We should note that while it appears from Fig. 1B that the California specimens of B. perspicuum are distinctive, there are no nucleotide sites in either gene at which there are consistent diff erences between California specimens and the remainder.

Morphological differences
Once the two forms became evident from the molecular analyses, specimens were separated and examined for morphological diff erences.B. transversale shows more contrast in the dorsal color pattern, and B. perspicuum tends to have a more orange hue (Fig. 3)."partitioned" indicates those analyses for which the gene was partitioned into codon positions with diff erent models used for fi rst, second, and third positions, as opposed to analyses in which all positions were pooled, and subject to a single model.

28S
Th e most consistent external diff erences are in the pronotum: B. transversale has a slightly more convex and cordate pronotum, with a smoother basal region, in contrast to the fl atter pronotum of B. perspicuum, which has less rounded sides, and with the basal region more evidently punctate (Fig. 4).In the localities from which DNA sequences were obtained (Fig. 2), specimens of B. transversale have in addition a very distinctive mentum, with much reduced lateral lobes (Fig. 5A), in contrast to the more normal mentum and typical lateral lobes of B. perspicuum (Fig. 5B).However, there are specimens in the west (western British Columbia south to California) that match B. transversale in all regards, including in DNA sequences and other morphological characters, and yet have a more typical mentum with large lateral lobes (Maddison, unpublished).In addition to these diff erences in external traits between B. transversale and B. perspicuum, the ostial fl ag of B. transversale extends further ventrally and has a more gentle curvature (arrow in Fig. 6A) than that of B. perspicuum (arrow in Fig. 6C).Th e diff erences observed in morphological characters, congruent with the consistent molecular diff erences between two unlinked genes, even where the two forms are microsympatric in Colorado, provides strong evidence that the two forms are not exchanging genes, and are two diff erent species.
As currently conceived, this species is a mix of diff erent forms.Th e typical form (including the holotype of B. transversale Dejean, and all specimens sequenced for this study) has an unusual mentum with much-reduced lateral lobes, which do not project anteriorly much beyond the mental tooth (Fig. 5A).Th is form occurs from Nova Scotia west to Alaska, south through eastern British Columbia, eastern Oregon, Idaho, Nevada, Utah, Colorado, and Wyoming.Western British Columbia, western Oregon, and California contain another form, with no consistent diff erences in COI or 28S from the typical form (Maddison, unpublished), with a normal B. transversale pronotum and aedeagus, but with full lateral lobes on the mentum.Th e mentum of this western form is thus similar to that of B. perspicuum, Fig. 5B, and in contrast to eastern B. transversale.Th e western form itself is rather variable morphologically, and is being examined in more detail (Maddison, in prep.).It may be a complex of multiple species, and thus the fi ve younger names mentioned above are only tentatively synonymous with B. transversale.In addition, while it is clear that the LeConte name and the two Casey names are not synonyms of B. perspicuum, the Motschulsky and Chaudoir names might be junior synonyms of B. perspicuum; examination of the types will be required to confi rm the synonymies.Morphological characters distinguishing these species from B. perspicuum are described above, under "Morphological Diff erences".
Th is species is known from northeastern Arizona, Utah, northern New Mexico, Colorado, and Wyoming.It is characterized by a shiny, slightly more cordate pronotum than other members of the group, without notable basal punctuation, relatively fl at elytral intervals, with small punctures in the elytral striae, and distinctive male genitalia.As fi rst revisers, we choose B. sarpedon Casey as the valid name of this species.
Th is species is known from South Dakota, Colorado, New Mexico, and Arizona, south at least to Guatemala.In the United States, specimens have unicolorous, dark elytra, a smoother pronotum, and tend to be larger than those of B. lugubre; some specimens from México and Central America have pale elytral apices.Th e two species can only be reliably distinguished by characters of the male genitalia, female spermatheca, or DNA sequences.Th e two species are microsympatric, occurring together on the same creek or river shore at numerous localities in Arizona and New Mexico.
Th is species is known from Texas, New Mexico, Arizona, Utah, and California south at least to Oaxaca, México.In the United States specimens of this species tend to be paler than those of B. mexicanum, with some specimens having a pale elytral apex, and with more notable punctures at the base of the pronotum.Male genitalia, female spermathecae, and DNA sequences are distinctive.
Th is species is known from northern New Mexico and southern Colorado.Individuals are colored similarly to B. transversale, but are brighter and more reddish, and are more convex and shinier than other members of the group, with larger punctures in the elytral striae, and more convex intervals.While their color pattern is similar to members of the B. transversale group, male genitalia and DNA sequences indicate a much closer relationship to B. mexicanum and B. lugubre.
).We have included one specimen each of B. mexicanum, B. commotum Casey, and B. sejunctum semiaureum Fall as outgroups to this study.Preliminary analyses of multiple genes across Bembidion (Maddison, unpublished) indicate that the B. mexicanum group and B. transversale group are closely related, forming a clade along with a few other members of Ocydromus (sens.lat.), including B. commotum and B. sejunctum.Vouchers are housed in the David Maddison voucher collection at Oregon State University.

Table 1 .
Specimens examined for 28S rDNA and COI.

Table 2 .
Number of fi xed diff erences between B. transversale and B. perspicuum for each gene.Only those specimens whose sequences were complete for the length indicated where used in the comparison.For example, there are 40 sites in the entire COI gene at which all 13 examined B. transversale have a diff erent nucleotide than all 5 examined B. perspicuum, and these nucleotide diff erences imply two consistent diff erences in amino acids.COI, fi rst ca.600 sites; scale bar 0.01 substitutions per site.Asterisks indicate specimens collected microsympatrically on a gravel bank along the shores of the Arkansas River at Texas Creek, Colorado (38.4106°N 105.5844°W).
Figure 2. Geographic distribution of specimens sampled for DNA sequences.Circles: Bembidion transversale; stars: Bembidion perspicuum.More complete data for these localities are available at http://bembidion.org/transversaleGroup/, as a KML fi le.

Table 3 .
Support values for monophyly of each of B. transversale and B. perspicuum.BPP: Bayesian Posterior Probability; ML Boot: likelihood bootstrap percentage; Pars Boot: parsimony bootstrap percentage.

Classification and species notes
Bousquet and Larochelle (1993)Bembidion transversale and B. mexicanum groups contain a combined six species in America north of México:Bembidium transversale Dejean, 1831:110.Type locality restricted to Nipigon, Ontario, by Lindroth(Lindroth 1963).Holotype female, in MNHN, examined by Kipling Will, who provided to us photographs, including of the mentum, allowing us to confi rm the identifi cation.Peryphus eros Motschulsky, 1850:10.Type locality California.Lectotype, designated byBousquet and Larochelle (1993), in ZMUM.Synonymy tentative.Ochthedromus mannerheimii LeConte, 1852:190.Type locality San Diego, California.Th ree specimens in the MCZ in the Bembidion transversale unit tray in the LeConte collection have gold discs (signifying California), and include syntypes for this name.Th e fi rst specimen is a female labeled as follows: [gold dot] / "O.mannerheimii Lec. S. D. transversale ‡ Man" [in LeConte's handwriting] / "transversale 27" [handwritten], and is certainly a syntype.Th e two following specimens, a male and female, are labeled with a gold dot in addition to a label of "transversale 28" and "transversale 29" respectively, and are quite possibly also syntypes.As the fi rst specimen is the only one that is surely a syntype, it is herein designated as the lectotype, MCZ type number 35571.All specimens externally match the western form (see below) of B. transversale.We have examined the aedeagus of the male, and it matches that of B. transversale.Bembidium haplogonum Chaudoir, 1868: 241.Type locality California.Location of types unknown.Synonymy tentative.
Type locality Tucson, Arizona.Holotype female, in USNM (type 36915), examined.Th is species is known from Texas, Kansas, Colorado, New Mexico, Arizona, Utah, Nevada, California, and Oregon.Its range in México has not been examined.Specimens from northern California and Oregon are dark, with only elytral apices being pale.Th ey are thus very similar in appearance to paler specimens of Bembidion lugubre, from which they are most readily distinguished by genitalic characters.