An unexpected clade of South American ground beetles (Coleoptera, Carabidae, Bembidion)

Abstract Phylogenetic relationships of the Antiperyphanes Complex of the genus Bembidion are inferred using DNA sequences from seven genes (two nuclear ribosomal, four nuclear protein coding, and one mitochondrial protein coding). Redefined subgenera within the complex are each well-supported as monophyletic. Most striking was the discovery that a small set of morphologically and ecologically heterogeneous species formed a clade, here called subgenus Nothonepha. This unexpected result was corroborated by the discovery of deep pits in the lateral body wall (in the mesepisternum) of all Nothonepha, a trait unique within Bembidion. These pits are filled with a waxy substance in ethanol-preserved specimens. In one newly discovered species (Bembidion tetrapholeon sp. n., described here), these pits are so deep that their projections into the body cavity from the two sides touch each other internally. These structures in Bembidion (Nothonepha) are compared to very similar mesepisternal pits which have convergently evolved in two other groups of carabid beetles. The function of these thoracic pits is unknown. Most members of subgenus Nothonepha have in addition similar but smaller pits in the abdomen. A revised classification is proposed for the Antiperyphanes Complex.


Methods
Specimens examined and depositories. Specimens examined are from or will be deposited in the collections listed below. Each collection's listing begins with the coden used in the text. Most specimens were killed and preserved in Acer sawdust to which ethyl acetate was added. Specimens collected specifically for DNA sequencing were killed and stored in 95% or 100% ethanol, with best results obtained when the abdomen was slightly separated from the rest of the body to allow better penetration, or when the reproductive system was dissected out through the rear of the abdomen within a few minutes of the beetle's death in ethanol. Ethanol was decanted and refilled at least two times within the first few weeks after death. Storage was then at 4° or -20°C.
Taxon sampling for DNA studies. DNA was newly sequenced from 25 species of the Antiperyphanes Complex of Bembidion; to these sequences were added sequences of 20 more species of this complex acquired during previous studies (Table 1). Forty additional species of Bembidion were also included in the analyses (27 species of the Table 1. Sampling of Bembidion species of the Antiperyphanes Complex. The ID column indicates how the specimens were identified: if there is a number in square brackets, then the specimen was identified by me using one of the following references: 1: Jeannel (1962); 2: Toledano (2008); 3: Vigna Taglianti and Toledano (2008); 4: Moret and Toledano (2002); 5: Erwin (1982). If there is a single letter, then the specimen is the holotype of the species ("H"), or I compared the specimen to the primary type or syntypes ("T") of the species, in consultation with the relevant literature; the species marked "LT" was identified by Luca Toledano. Thus, all of these are "genseq-4" sequences (Chakrabarty et al. 2013), except for DNA2653 (genseq-1) and DNA2236 (genseq-3). In the third column, four-digit numbers in entries are D.R. Maddison voucher numbers for sequenced specimens; if a "*" appears in that column, the data from these specimens were obtained from GenBank from previous studies (Maddison 2012;Maddison and Ober 2011;Maddison and Toledano 2012;Maddison et al. 2013), and more data about the specimens is present in those papers. Locality details for newly sequenced specimens (those without a "*") are given in the Appendix 1. For those three species with a "*" in the "#" column but without a number, details of specimens sequenced is available in Maddison (2012). For each gene, GenBank accession numbers are listed.  Nine additional specimens of Bembidion tetrapholeon sp. n., were sequenced (Table 2) to examine variation. The 10 specimens sequenced in total included specimens from two localities in Chile and three localities in Argentina (Tables 1, 2), and included the typical uniformly black specimens, and others that have a large orange spot on their elytra. (Notholopha) sexfoveolatum C Open high-elevation grassland at Ecuador: Pichincha: Paso de la Virgen, 4060m, habitat of three species of subgenus Ecuadion: Bembidion chimborazonum, B. guamani, and B. humboldti D Leaf litter in cloud forest, Ecuador: Pichincha: Reserva Yanacocha, 0.1152°S, 78.5837°W, 3540m, habitat of Bembidion andersoni, B. georgeballi, and B. onorei. Table 2. Additional sampling of Bembidion tetrapholeon to examine DNA sequence variation. The DNA voucher number is listed in the "#" column. Color of elytra: BL: nearly black; OR: black with large orange spot. * indicates the holotype. Further details about the localities are provided under the description of B. tetrapholeon and in the Appendix 1. All of these are "genseq-2" sequences (Chakrabarty et al. 2013), except for DNA2356 (genseq-1) and DNA1752 (genseq-3). Fragments for these genes were amplified using the Polymerase Chain Reaction on an Eppendorf Mastercycler Thermal Cycler ProS, using TaKaRa Ex Taq and the basic protocols recommended by the manufacturer. Primers and details of the cycling reactions used are given in Maddison (2012). The amplified products were then cleaned, quantified, and sequenced at the University of Arizona's Genomic and Technology Core Facility using a 3730 XL Applied Biosystems automatic sequencer.

ID
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 2011a;Maddison and Maddison 2011b) with subsequent modifications by Chromaseq and manual inspection. Multiple peaks at a single position in multiple reads were coded using IUPAC ambiguity codes.
Sequences of COI for two species showed evidence of nuclear copies of this mitochondrial gene ("numts") (Thalmann et al. 2004). For B. georgeballi (voucher 2661) and B. walterrossii (voucher 2650), sequences obtained using the LC1490-HC2198 primer pair yielded different sequences from the B1490-Bcoi2R primer pair. The former had numerous double-peaks, suggesting that the sequences included numts (Maddison 2012); the LC1490-HC2198 sequence for B. walterrossii also had one stop codon near its 5' end. The reads from the B1490-Bcoi2R primers are much cleaner and show no double-peaks. Although all of these sequences have been submitted to GenBank, only the sequences from the B1490-Bcoi2R primers have been included in the analyses.
Alignment and data exclusion. The appropriate alignment was obvious for all protein-coding genes. There were no insertion or deletions (indels) evident in the sampled CAD, ArgK, Topo, or COI sequences. In wingless, there were two small, wellseparated indels, restricted to only three taxa: three inserted nucleotides in Bembidion (Zemetallina) parviceps Bates, and six inserted nucleotides in a different region in the two species of subgenus Omotaphus Netolitzky sampled. These inserted nucleotides were excluded from analyses.
The ribosomal genes showed a slightly more complex history of insertions and deletions. Both genes were first subjected to multiple sequence alignment in MAFFT version 7.130b (Katoh and Standley 2013), using the L-INS-i search option and otherwise default parameter values. Visual inspection suggested no needed improvements, and no ambiguously aligned regions that required exclusion.
Molecular phylogenetic analysis. Models of nucleotide evolution were chosen with the aid of jModelTest version 2.1.1 (Darriba et al. 2012;Guindon and Gascuel 2003) (for each gene) and PartitionFinder version 1.1.1 (Lanfear et al. 2012) (for parts of the partition chosen by PartitionFinder). Among the models supported by RAxML, the model chosen for all genes by the Bayesian Information Criterion was GTR+I+Γ.
Likelihood analyses of nucleotide data were conducted using RAxML version 7.2.6 (Stamatakis 2006). Analyses were conducted on each gene individually, as well as a matrix of seven genes concatenated together. Two different partitioning schemes were examined: (1) with seven parts, one for each gene; (2) as chosen using the Bayesian Information Criterion (BIC) using PartitionFinder (Lanfear et al. 2012). The partition chosen by BIC contained three parts: one part with third positions of COI; a second part with third positions of the nuclear protein-coding genes; a third part with all remaining sites. For bootstrap analyses 2000 replicates were conducted; maximum likelihood bootstrap (MLB) values are reported as percentages. In addition to these bootstrap analyses, searches for maximum likelihood trees were conducted using 1000 search replicates.
Most-parsimonious trees (MPTs) were sought using PAUP* (Swofford 2002). To search for most parsimonious trees, 2000 replicates were conducted, each beginning with a starting tree formed with the random addition sequence option, with each replicate saving no more than 25 trees. For parsimony bootstrap analyses in PAUP*, 1000 bootstrap replicates were examined, each of which used a heuristic search with four replicates, each beginning with a starting tree formed by the random addition sequence option, with TBR branch rearrangement, with each replicate saving no more than 25 trees; the estimated bootstrap values are reported as parsimony bootstrap percentages (PB).
Morphological methods. General methods of specimen preparation for morphological work, and terms used, are given in Maddison (1993;. After dissection from the body, genitalia were prepared by treatment in 10% KOH at 65 °C for 10 minutes followed by multi-hour baths of distilled water, 5% glacial acetic acid, distilled water, and then ethanol. Male genitalia, when studied, have been mounted in Euparal between two small coverslips attached to archival-quality heavyweight watercolor paper. Measurements for body length are from apex of the labrum to apex of the longer elytron.
An examination of external and internal features of the exoskeleton was conducted to search for possible apomorphies of one of the discovered clades, focusing on externally-visible pits in the mesothorax and anterior region of the abdomen. Internal skeletal elements were studied on specimens whose soft tissue was dissolved by placing the opened body in a 10% KOH solution at 65 °C for 10 minutes.
Studies of muscles and other internal soft tissue were conducted on specimens killed and preserved in 100% ethanol. This is not the ideal preservation medium, as it leaves muscles brittle and more difficult to trace. However, since the unexpected discovery of structures requiring internal examination, better-fixed specimens have not been available.
Photographs of body parts were taken with a Leica Z6 and JVC KY-F75U camera. For pronotal, elytral, and genitalic images, a stack of photographs at different focal planes was taken using Microvision's Cartograph software; these TIFF images were then merged using the PMax procedure in Zerene Systems's Zerene Stacker; the final images thus potentially have some artifacts caused by the merging algorithm.

Data resources
Sequences have been deposited in GenBank with accession numbers KJ653019 through KJ653220. GenBank numbers for the two apparent numts sequences are KJ653155 for DNA2661 and KJ653156 for DNA2650. Aligned data for each specimen as well as files containing inferred trees for each gene and concatenated matrices are available in Suppl. material 1 and 2, and have been deposited in the Dryad Digital Repository, http://doi.org/10.5061/dryad.47r16.

Molecular results
The inferred phylogeny is presented in Figs 7 and 8, with support values for notable clades given in Table 3. Maximum likelihood trees and maximum likelihood bootstrap trees for each gene and the concatenated matrices are illustrated in Suppl. material 3 and 4. They are also contained in the NEXUS file S1 in the Suppl. material 1.
The monophyly of each subgenus (indicated by color in Figs 7 and 8) is well supported by analyses of the concatenated matrix (MLB=100 and PB=100 for all but one subgenus) and at least four genes (Table 3), except for subgenus Antiperyphanes. Antiperyphanes is monophyletic in the maximum likelihood trees of four genes, but bootstrap support is low (Table 3).
The basal split in the Antiperyphanes Complex appears to be between clade b and clade d (Fig. 8). Clade b is strongly supported in seven-gene analyses (MLB=100, PB=100), and there is bootstrap support in individual analyses of four genes (Table 3). As a subgenus of Bembidion, this clade would take the name Nothonepha Jeannel, as it contains Bembidion lonae, the type species of Nothonepha. Clade d is also strongly supported in the seven-gene analyses (MLB=100, PB=99), and there is moderate to weak bootstrap support from four genes (Table 3).
Within the Antiperyphanes Complex, strongly supported relationships between subgenera include a clade containing Antiperyphanes and Chilioperyphus, and a sistergroup relationship between that clade and the subgenus Ecuadion (Fig. 8, Table 3). B.
(Antiperyphus) philippii appears as the sister group to subgenus Notholopha. Table 3. Support for and against various clades. The letter at the left corresponds to the circled letters of clades in Fig. 8 (although clade i, consisting of clades j and k, is absent in Fig. 8). ML: Maximum likelihood analysis; P: parsimony analysis. Numbers indicate the bootstrap support expressed as a percentage. Cells shaded in gray to black indicate that the clade is present with bootstrap support greater than 50 or present in the optimal (maximum likelihood or most parsimonious) trees but with bootstrap value below 50. Cells in white indicate that the clade has a bootstrap value less than 50 and is not present in the optimal tree; if the bootstrap value is listed in parentheses, it means that a contradictory clade was present in the optimal trees. For no analysis was there bootstrap support greater than 50 against any of these clades. Abbreviations: "exc." = "excluding". As a whole, the Antiperyphanes Complex is supported as monophyletic (clade a in Figs 7 and 8, Table 3), although not as strongly as in an earlier study with more limited sampling of the group (Maddison 2012). Individual gene support for the clade is only provided by CAD and to a lesser extent wingless (Table 3), but the concatenated analyses have MLB=88 and PB=86. Lack of monophyly of the complex in some analyses (e.g., individual gene analyses of 28S and 18S) is a result of Nothonepha falling in the Bembidion Series separate from the rest of the complex.
Different analytical methods yielded similar results for the concatenated, sevengene matrices. The two partition schemes examined (by gene and as chosen by Parti-tionFinder) resulted in maximum likelihood trees that differ only in the placement of Bembidion georgeballi within subgenus Ecuadion. In maximum likelihood bootstrap analyses, clades with MLB>90 were the same in both partition schemes. Parsimony analyses showed similar results to maximum likelihood (Table 3).
Within Bembidion tetrapholeon sp. n., the 10 specimens sequenced from five localities showed little variation in DNA, and the variation observed was not correlated  Table 3. with presence of an orange spot. There was no variation observed in CAD (n=10), ArgK (n=5), Topo (n=10), and 28S (n=10), over a total of more than 3090 bases. In the wingless gene (n=6) variation was observed at three third-position sites, all of which represent synonymous differences, and for each of which some other specimens were heterozygous for the variants. COI showed the most variation with variability at eight sites, seven of which represented synonymous differences and one a non-synonymous difference. At seven of these sites, nine of the ten specimens had the same nucleotide, with the tenth specimen being unique; the specimen that was unique varied from site to site. The most distinct specimen was DNA2236, from Chiloé, with three unique nucleotides in the more than 650 bases of COI sequenced.

Morphological results
With the unexpected discovery that B. tetrapholeon sp. n., belongs in a clade with an assortment of morphologically and ecologically diverse Bembidion, the search for synapomorphies for this clade became compelling.
Mesothoracic pits. The most striking derived feature observed was presence in all Nothonepha species of a pit in each lateral wall of the mesothorax. This mesepisternal pit (Fig. 9A) appears empty in many specimens killed in ethyl acetate, but in most specimens preserved in ethanol, a waxy substance is visible within it (Fig. 9A). When extracted and placed in glycerin on a microscope slide, this substance appears slightly yellowish-gray and contains no obvious substructure or particles (including no evident bacteria or fungal spores) when examined at 400× with transmitted, brightfield light (n=2, from B. tetrapholeon). In contrast, all other members of Bembidion examined to date lack such a pit (e.g., Fig. 9B).
In B. tetrapholeon these paired structures, one on either side, internally manifest as large intrusions which touch in the center of the body cavity (Figs 10A, C, E). Typical Bembidion have no such structures internally (Figs 10B, D, F). There is variation within Nothonepha in the size of the intrusions, with B. engelhardti having relatively small intrusions (and thus relatively shallow pits) (Fig. 11).
Examination of musculature in B. tetrapholeon (n=4) and B. tucumanum (n=1) revealed no muscles attached to the internal intrusions, although the course of some muscles appeared to be bent by the necessity to wrap around the structures. There were no evident large glands associated with the intrusions, although there were small patches of tissue on their internal surfaces.
Two other groups of carabids reported to have mesepisternal pits were also examined, members of subgenus Tachylopha of the genus Elaphropus (Bruneau de Miré 1966;Erwin 1970) and the genus Oodinus (Spence 1982). I have specimens of these preserved in 95% ethanol, and both have large pits in the mesepisternum in the same place as Bembidion (Nothonepha). In Elaphropus (Tachylopha) leleupi Basilewsky from South Africa the pits are filled with a waxy substance similar to that seen in Nothonepha (Fig. 12A). Internally these pits appear as two large intrusions that join in mid-thorax to form a tunnel (Fig. 12B); in the two specimens I have dissected there is no evidence of a septum at the point of joining, and the waxy substance fills the tunnel. I have seen ethanol-preserved specimens of E. (Tachylopha) basilewskyi Bruneau de Miré from Gabon (identified with Bruneau de Miré (1966)) and E. (Tachylopha) spenceri (Sloane) from Australia (identified with Baehr (1988)) that have similar pits also filled with a waxy substance. The apparently related subgenus Sphaerotachys also has mesepisternal pits, although they are much smaller than those seen in Tachylopha; a specimen I   have examined from Hans Merensky Nature Reserve, Republic of South Africa, has pits similar to those shown in Fig. 11. The single ethanol-preserved Oodinus alutaceus (Bates) (identified with Bousquet (1996)) that I have examined, from south Texas, also has mesepisternal pits, but internally the intrusions do not touch, and are more similar in structure to those of B. (Nothonepha) than E. (Tachylopha). In some specimens of Oodinus, the pits are also filled with a waxy substance (Spence 1982).
Abdominal pits. In additional to mesepisternal pits, B. tetrapholeon has a prominent pit on each side of the abdomen, ventrally, between abdominal segments II and III (Fig. 13A). In ethanol-preserved specimens, this pit is filled with a waxy substance similar to that in the mesepisternal pits. Almost all other species of subgenus Nothonepha have similar pits (e.g., B. (Nothonepha) sp. nr. lonae, Fig. 13C); they are lacking only in B. (Nothonepha) lonae (Fig. 13D), the sister to B. sp. nr. lonae.
Internally these abdominal pits are evident as knob-shaped intrusions (Fig. 14A, C). Consistent with the lack of externally visible pits, B. lonae lacks these intrusions, and has instead only a ridge between the abdominal segments (Fig. 14D), as is typical in Bembidion.
Outside of Nothonepha I have seen no species of Bembidion with as prominent abdominal pits, and most species lack them entirely. B. (Antiperyphanes) mandibulare, for example, lacks abdominal pits and has only a slight linear depression in that region (Fig.  13B), and internally a simple ridge is evident (Fig. 14B). There is much variation between members of the Antiperyphanes Complex in this feature, however, with some species having an evident pit/internal intrusion (e.g., B. (Antiperyphus) philippii, Fig. 14E, and B. (Antiperyphanes) zanettii), and others (e.g., B. (Ecuadion) chimborazonum, Fig. 14F) having a low, wide hump internally. Outside of the Antiperyphanes Complex all species examined either have a sinuate (Fig. 14G) or straight (Fig. 14H) ridge internally in this region.
Whatever internal structure is present between abdominal segments II and III, whether a ridge, or low mound, or knob-like intrusion, in the species examined this structure serves (at least in part) as an apodeme. In B. tetrapholeon (n=5), for example, a muscle bundle is attached to the apex of the internal intrusion, and extends forward to the small sclerotized patch (sp in Fig. 14C) in the membrane that serves as the front boundary above the ventral margin of the abdomen. Another muscle bundle extends from this sclerotized patch forward and laterally to the lateral wall of the body, where it attaches to an external rod-like sclerite that is connected to the posterior lateral corner of the metanotum; this rod-like sclerite extends posterior laterally from that point to near the metepimeron. I have examined B. (Ecuadion) chimborazonum (n=1) and it has a similar muscle attached to the low mound in the same region; B. (Bracteon) foveum Motschulsky (n=1), B. (Ocydromus Complex) nebraskense (n=2), and the bembidiine Lionepha casta (n=2) all have a similar muscle connecting the intersegmental ridge at the equivalent region to the small sclerotized patch.

Monophyly of Nothonepha.
DNA data strongly support Nothonepha as a clade. Four genes (CAD, wg, 28S, 18S) independently have bootstrap support for the clade (Table 3), and the concatenated seven-gene analysis has MLB=100 and PB=100. Combined with the striking synapomorphy of mesothoracic pits, evidence for this clade becomes convincing. As Darwin (1859) noted, "We may err in this respect in regard to single points of structure, but when several characters, let them be ever so trifling, occur together throughout a large group of beings having different habits, we may feel almost sure, on the theory of descent, that these characters have been inherited from a common ancestor." The unexpectedness of Nothonepha. The relationship between species here grouped into subgenus Nothonepha was so unexpected when first discovered from sequences of 28S that it was dismissed, and considered to be the result of DNA contamination or mislabeled extractions. This small clade, of only ten known species, includes some of the largest Bembidion in South America (B. germainianum, up to 6.4 mm in length, Fig. 5C), and some of the smallest (B. lonae, down to 2.4 mm in length, Fig.  5D). They range in habitat from cobbles shores of small, cold, clear rivers (Fig. 6D) to mixed shores of large rivers (Fig. 6A), and sand shores of desert rivers (Fig. 6C) to warm, exposed salt flats (Fig. 6B). In the field, beetles in this group give the appearance of rather different groups of carabids. In my first field encounter with live B. lonae, I mistook them for tachyines of the genus Elaphropus Motschulsky; B. germainianum is reminiscent of the Nearctic Bembidion perspicuum Casey, a member of the Ocydromus Series of Bembidion; B. eburneonigrum looks very much like a small member of the subgenus Notaphus as it scurries on the sandy shores of rivers.
There are many examples of clades throughout the tree of life that contain species of diverse forms living in diverse habitats. Why then is Nothonepha unexpected? The current classification, to the extent that it might be a predictor of relationships, would suggest that these taxa are not related. B. lonae is the only described species of those sampled that was placed in Nothonepha; the very similar but undescribed B. sp.nr. lonae would have been placed there as well. The other described species (B. germainianum, B. engelhardti, B. tucumanum, B. eburneonigrum, and B. tucumanum) had all been classified in subgenus Antiperyphus, along with B. philippii. When I first discovered B. tetrapholeon, I thought it represented a separate lineage requiring a new subgeneric name, as it is very different in form from any other South American Bembidion. This apparently added a third element to the diverse group. However, the current classification, constructed with limited data and without phylogenetic analyses, may not be the best predictor of phylogenetic relationships.
Nonetheless, Nothonepha is a morphologically heterogeneous group (Fig. 6). I have been studying Bembidion systematics for over three decades, and whatever predictive map my brain has developed from all the data accumulated over the years contained no hint that the species shown in Fig. 6 formed a clade.
However, it may not be the diversity of form within Nothonepha that is unusual, but rather the diversity given the lack of intermediate forms and small size of the group. Other clades in the South American fauna are also diverse in form and size (e.g., Ecuadion, Fig. 3), but they have many more species, some of which are intermediate between the more distinct forms. The lack of intermediate forms in Nothonepha may be a result of low speciation rates with high rates of morphological evolution, or it might be a result of extinction of intermediate forms; it is unlikely to be a lack of sampling, as enough collecting has been done in South America to suggest that there are not a large number of undescribed species of Nothonepha. A full investigation of patterns of morphological and molecular rates, times of divergence, and speciation and extinction rates relative to other clades is beyond the scope of this paper, but would be a worthwhile topic for future studies.
Function of mesepisternal pits. Exoskeletal invaginations are widespread in beetles (Grebennikov and Leschen 2010). These cavities occur on many different body parts, including in the lateral regions of the mesothorax (e.g., in the ptiliid tribe Discheramocephalini (Grebennikov 2008;2009)); their function probably varies from group to group. Many have been thought to be mycangia for storing fungal spores, but this is well documented in only two of the many independent origins of such cavities (Grebennikov and Leschen 2010). In a few groups of polyphagan beetles (e.g., the Staphylinoidea subfamily Scydmaeninae, the Cucujoid families Cyclaxyridae and Nitidulidae, the Tenebrionoidea family Zopheridae) there are species with pits containing a waxy substance (Grebennikov and Leschen 2010). Wax has been proposed to act as a defensive shield (Lawrence and Hlavac 1979), or as a medium for retaining fungal spores (Grebennikov and Leschen 2010).
The functions of mesepisternal pits in the two carabid groups in which they were previously described is not known, but functions have been hypothesized. Erwin (1970) proposes that pits of Elaphropus (Tachylopha) are insertion points for ant mandibles, allowing ants to carry the adults around. Bousquet (1996) calls the structures in Oodinus "apodemal pits", which implies a function as internal attachment points for muscles.
The similarities between the wax-filled mesepisternal pits observed in the bembidiine Bembidion (Nothonepha), the tachyines Elaphropus (Tachylopha) and Elaphropus (Sphaerotachys), and the oodine Oodinus are striking (e.g., Figs 9A and 12A in this paper, and Fig. 145 in Bousquet (1996)), enough so that a common function might be hypothesized. Although the shape and nature of the deepest pits and largest intrusions differs between the groups (e.g., there are no Nothonepha known with the merged intrusions of some Tachylopha), the less extreme forms are indistinguishable externally and internally. These features are surely convergent, as the clades are not closely related. Bembidion and Elaphropus are both members of the subfamily Trechinae, but they are each deeply nested within independent clades (Maddison and Ober 2011); Oodinus is a member of the tribe Oodini (Bousquet 1996;Spence 1982), which is nested within several clades in the superfamily Harpalinae, itself a well-supported clade (Maddison et al. 1999;Ober and Heider 2010). It may be that these pits serve different functions in these clades, but consistency of structure suggests they may be serving the beetles in similar ways in the three different groups.
Correlates in way of life might provide some hints about function. Members of the three carabid groups are all presumably generalist predators, as is typical in Carabidae (Thiele 1977). They are also all terrestrial, but associated with shorelines. Bembidion (Nothonepha) species occur at edges of bodies of water in southern South America. Elaphropus (Tachylopha) occurs in similar habitats. I have seen Elaphropus (Tachylopha) spenceri (Sloane) (identified using Baehr (1988)) from multiple localities in Queensland, Australia, labeled as being found at "water's edge" along creek shores. Bruneau de Miré (1966) reports numerous species along rivers, and states that they can be abundant in swamps and moist forest humus. Oodinus occurs at the edges of marshes and swamps (Bousquet 1996;Spence 1982). However, many other carabid groups lacking these pits are also found in these habitats, including many other Bembidion, Elaphropus, and Oodini.
It appears unlikely that mesepisternal pits are used as ant handles in Bembidion (Nothonepha) and Oodinus, and probably not in Elaphropus (Tachylopha). Erwin's (1970) hypothesis was based in part on the unusual elytral structure of Tachylopha, which is narrow above the mesepisternum and which possesses a notch into which the base of ant mandibles could fit. Although most Nothonepha have narrow-enough pronota to allow curved, sickle-shaped mandibles of a large ant access to the pits, that seems much less likely for Oodinus, which are wide-bodied, oval carabids. Oodinus have a distinct ledge along the lateral edge of their bodies; to fit a mandible tip into the pit underneath this ledge an ant would need exceptionally curved mandibles. In general, ground-nesting ants are relatively rare in wet, near-shore, seasonally inundated habitats, and one would not expect large ants (ground-nesting or arboreal) with sufficiently curved mandibles in these near-shore environments (Philip S. Ward, pers. comm.). I have observed ants only rarely in the near-water habitats of subgenus Nothonepha, and not at all in the case of the four localities at which I have collected B. tetrapholeon or the three localities at which I have found B. germainianum. Furthermore, there is no evidence for any association between ants and these three carabid groups.
There is also evidence against the mesepisternal pits functioning as apodemes. As noted above, Nothonepha and Tachylopha do not have muscles attached to the internal walls of the mesepisternal pits. Spence (1982) reports that there are no conspicuous muscles that attach to the internal walls of the pits in Oodinus.
It is possible that the function is as a reservoir for the wax, although where the wax is produced is not evident. Spence (1982) states that the pit walls are perforated by numerous channels in Oodinus; these might be ducts for glandular secretions. I have not detected any large glands internally near the intrusions, although there is a thin layer of tissue in places on the inner surface of the structures. It is also possible that the wax is not produced near the intrusion; it might be produced elsewhere on the beetle's body, and in fluid form flow into the pits. (As noted above, the wax is not evident in many ethyl acetate killed specimens, but rather in ethanol killed specimens, suggesting that the substance may precipitate in ethanol, but exist as a liquid otherwise.) More detailed histological work is needed to explore possible glandular sources.
Whatever the source of the wax, its function (if it has one) is unclear. It seems unlikely that it would be for retaining fungal spores (Grebennikov and Leschen 2010), as there is nothing known about these generalist predators that would suggest a benefit to the beetle to retain fungal spores. Wax as a defensive coating is plausible (Lawrence and Hlavac 1979), but more studies are needed to explore this and other possibilities.
Function of abdominal pits. In contrast to the mesepisternal pits, the abdominal pits in Bembidion (Nothonepha) evidently serve (in part) as apodemes, that is, as attachment points for muscles. However, the function of those muscles is unclear; in general the nature and function of muscles at the junction of the metathorax and abdomen in beetles is poorly known (Rolf Beutel, pers. comm. 2014). Even if the abdominal pits serve as apodemes, they may serve additional functions as well, perhaps related to the presence of wax that appears similar to that found in the mesepisternal pits.

Taxonomic treatment
Three clades of Bembidion comprise the South American fauna: subgenus Notaphus Dejean (32 species (Jeannel 1962;Toledano 2002;Toledano 2008)), subgenus Nothocys Jeannel (13 species (Toledano 2002;Toledano 2008)), and the Antiperyphanes Complex. Along with a number of lineages outside of South America, these all belong to one subclade of Bembidion, the Bembidion Series (Maddison 2012). Subgenus Notaphus is widespread and abundant throughout the New World, with over 50 species known from North America; in addition, seven species occur in the Old World (Lobl and Smetana 2003). Nothocys is restricted to southern South America, occurring in Chile, Argentina, Peru, and Bolivia (Toledano 2002;Toledano 2008). The largest of these clades is the Antiperyphanes Complex, with about 95 described species (Jeannel 1962;Maddison 2012;Toledano 2002;Toledano 2008), and many undescribed.

Antiperyphanes Complex
Remarks. Members of the Antiperyphanes Complex are diverse in form (Figs 1-3, 5). There are no recognized derived morphological characteristics of the group, although the clade is moderately well supported by the concatenated DNA sequence data (Fig. 8, Table 3). Within the South American fauna, most species can be recognized by the lack of an N sclerite in the internal sac of the male genitalia (Toledano 2008). However, some members of the complex, including some species in subgenus Nothonepha, have a small sclerite that could be homologous to the N sclerite (Toledano 2008).
The Antiperyphanes Complex, as here classified, consists of at least five subgenera: Antiperyphanes, Chilioperyphus, Antiperyphus, Notholopha, Ecuadion, and Nothonepha. Each of these subgenera is briefly discussed below, with notes about their composition. Two other poorly known subgenera, Pseudotrepanes Jeannel and Notoperyphus Bonniard de Saludo, are likely members of this complex, but specimens will need to be examined to confirm their membership.
Remarks. This group contains at least 19 described species (Toledano 2002;Toledano 2008), and is characterized by males having an aedeagus lacking a brush sclerite, and with a very long flagellum (Maddison et al. 2013;Toledano 2008). Included here are some species previously placed in subgenus Antiperyphus by Jeannel (1962): B. hirtipes (Jeannel), B. ringueleti (Jeannel), B. rufoplagiatum Germain, B. uniforme Csiki, and B. parvum (Jeannel). B. mandibulare Solier belongs to Antiperyphanes as well, and is nested well within it (Fig. 8); thus, subgenus Plocamoperyphus is a synonym of Antiperyphanes. There are no morphological characteristics of B. mandibulare that would suggest it is not a member of Antiperyphanes; it shares all apomorphies of the group. As first reviser, I choose Antiperyphanes as the valid name of the group.
Antiperyphanes has two distinct clades, each very well supported: the B. caoduroi group (clade j in Fig. 8; Fig. 1D), consisting (among the sampled species) of three large species from the northern Andes (Fig. 1D); (2) the remaining Antiperypanes (clade k in Fig. 8; Figs 1A-C, E). Each is supported by MLB=100 and PB=100 in the multi-gene analyses, and individually by support from five to seven genes (Table 3). As a whole, however, the monophyly of Antiperyphanes is only weakly supported by the combined analysis and analyses of four genes ( Table 3).
Members of this subgenus are found at the edges of bodies of water. For example, B. rufoplagiatum and B. zanettii are common on gravel and cobble river shores, B. ringueleti is found on gravel and sand shores of smaller creeks, and B. mandibulare on the sand beaches of the Pacific Ocean in Chile and sand beaches of rivers in Argentina. Jeannel, 1962Chilioperyphus Jeannel, 1962; type species Bembidium orregoi Germain, by original designation.

Subgenus Chilioperyphus
Remarks. This subgenus contains two described species (Jeannel 1962;Maddison et al. 2013) (Fig. 1F) and at least three undescribed species. Males are characterized by having a brush sclerite, and by having an extremely elongate flagellum, so long that it can only fit within the median lobe through folding (Maddison et al. 2013). Members of this subgenus occur on steep sand or clay banks of rivers and creeks.

Remarks. As noted by Toledano
In addition, Bembidion peterseni Jensen-Haarup (1910), from Mendoza, Argentina, can tentatively be placed here. I have examined a male syntype (in ZMUC), and it is similar in appearance to B. philippii, although with much deeper and longer elytral striae. It is not a member of Nothonepha (as it lacks mesepisternal pits), nor is it a member of Antiperyphanes (it has a brush sclerite, and does not have the long flagellum characteristic of Antiperyphanes). The internal sac of the male genitalia, although difficult to see because of the nature of the preparation, appears very similar to that of B. philippii.
B. philippii is common on sand shores of rivers in the provinces of Neuquén and Chubut in Argentina; it also occurs in Chile. Jeannel, 1962Notholopha Jeannel, 1962type species Bembidium punctigerum Solier, by original designation. Pacmophena Jeannel, 1962; type species Bembidium scitulum Erichson, by original designation.

Subgenus Notholopha
Remarks. Notholopha consists of 11 described species (Toledano 2002;Toledano 2008), and several undescribed (two of which are sequenced here). These are small beetles with large, protruding eyes (Fig. 2B,C), with small flagella in the internal sac of the male genitalia, and with brush sclerites. They have the general appearance when running of a member of the Holarctic subgenus Bembidion. Some frequent habitats similar to those of subgenus Bembidion, including dry habitats far from water (B. stricticolle), or upper banks of creeks (e.g., B. sp. "Nahuelbuta"). Others occur at high elevation near small rivulets in open, alpine areas (e.g., B. rugosellum and B. melanopodum), or at the edges of snowfields (B. sexfoveolatum).
Jeannel considered Pacmophena and Notholopha to be two subgenera within the genus Notholopha. As the characters that Jeannel used to distinguish the two are minor characters such as surface texture and antennal length, and as it appears that Pacmophena is paraphyletic with respect to the Notholopha (s. str.) (Fig. 7), I consider them synonymous, with Notholopha as the valid name.

Remarks.
Ecuadion is the largest subgenus in the Antiperyphanes Complex, with over 50 described species (Maddison and Toledano 2012;Moret and Toledano 2002;Toledano 2008;Vigna Taglianti and Toledano 2008) and likely many undescribed. It occurs from Costa Rica south to the mountains near Mendoza, Argentina. There are no known exoskeletal synapomorphies of the group, but it is well-supported by the molecular data, with bootstrap support in six of the seven genes examined (Table 3). Ecuadion falls into two distinct clades among the sampled species: (1) the B. chimborazonum group (clade m in Fig. 8), consisting mostly of larger, long-legged species (Figs 3A, B); (2) remaining Ecuadion (clade n in Fig. 8), consisting of mostly smaller species with shorter appendages (Figs 3C-F). The B. chimborazonum group is supported by all genes examined (Table 3); support for the complementary clade is not quite as strong, with the clearest evidence coming from ribosomal genes (Table 3).
Unlike most Bembidion, this subgenus has radiated in habitats away from water. Some occur in leaf litter in cloud forests (e.g., B. andersoni, B. georgeballi, B. onorei, B. sp. "Papallacta"; Fig. 4D), in habitats that would typically be occupied by the genus Trechus Clairville in North America. A number of species are found in open, high-elevation grasslands (e.g., B. chimborazonum, B. guamani, B. humboldti ; Fig. 4C). Some species occur on the upper banks of creek shores (e.g., B. sanctaemarthae, B. ricei); others are found on clay or silt cliffs (e.g., B. agonoides, B. walterrossii). Type locality. Argentina: Neuquén: Arroyo Queñi at Lago Queñi, 830m, 40.1575°S, 71.7210°W. The habitat at the type locality is a cobble, gravel, and coarse sand river shore (Fig. 6D); the river is cold and has crystal-clear water. In the same habitat members of the genus Bembidarenas Erwin are abundant, as is Bembidion (Antiperyphanes) rufoplagiatum.
Derivation of specific epithet. From the Greek "tetra", meaning "four", and "pholeon", meaning "pit", referring to the four prominent pits visible on the underside of adults.

Diagnosis.
A large, sleek, shiny Bembidion, with an unusual body form (Figs 5A,B) of narrow forebody and large elytra. With its shape, color, and smoothness it is one of the most distinctive Bembidion species in South America, and no other known species is likely to be confused with it; it is more reminiscent of some species in New Zealand, e.g., B. (Zeplataphus) dehiscens Broun (Lindroth 1976).
Length (4.7-5.7 mm, with most specimens above 5.0 mm). Color piceous (Fig.  5A), with legs and antennae in some specimens slightly paler, and with a few specimens having a large orange spot just in front of the elytral apices (Fig. 5B).
Head with shallow and parallel frontal furrows. Pronotum narrow, cordate, with hind angles flaring outward (Fig. 15). Very smooth, without punctures, and with a linear basolateral foveae; without distinct carina at hind angle. Lateral bead of pronotum not complete, not reaching front angle of prothorax and only in some specimens reaching the hind angle. One midlateral and one basolateral seta on each side.
Mesothorax with prominent pits in the mesepisternum (Fig. 9A), which appear internally as large intrusions that touch in the middle (Figs 10A, C, E). Smaller pits are present ventrally at the junction of abdominal segments II and III (Fig. 13A), which are evident internally as knob-like intrusions (Fig. 14A).
Hind wings full. Microsculpture absent from entire dorsal surface of the body except for the cervical region of the head, labrum, and faintly on the clypeus; the beetles are thus brilliantly shiny. Microsculpture absent from most of the ventral surface as well, with the most notable microsculpture being on the undersurface of the head.
Aedeagus with nearly straight ventral margin, tip of variable width (Fig. 17). Prominent brush sclerite, and with flagellum not clearly evident from the left side. There is no evident correlation between aedeagal structure and presence or absence of orange spots on the elytra (compare Figs 17A, B to Figs 17C, D).
Morphological variation. The most noted variation is in color of the elytra. Of the 257 specimens examined (including the six specimens identified by Luca Toledano), 245 have uniformly piceous elytra (Fig. 5A); the remaining 12 have a large, diffuse orange spot occupying most of the posterior third of the elytra (Fig. 5B). Ten of these orange-spotted specimens are from the three localities in Argentina, with at least one orange-spotted specimen from each locality, and two of the orange-spotted specimens are from Chile. In most of orange-spotted specimens, the posterior discal seta (ed5) is in the orange region, but immediately around the seta is a small dark spot. No other aspect of morphological or molecular variation was observed to be correlated with presence or absence of the orange spot.
DNA sequence variation. As noted above under Results, there was minor variation present in COI and the wingless gene, and no variation in the other genes studied.
Habitat and seasonality. At all four localities where habitat data were recorded, B. tetrapholeon specimens were found on cobble, gravel, and coarse sand shores of clear, fast-flowing rivers (Fig. 6D), from 55 m elevation (Rio Puntra, Isla Grande de Chiloé, Chile) to 830 m elevation (Arroyo Queñi, Neuquén, Argentina). These shorelines lack vascular plants. The beetles occur close to the water, most within 1 m. Specimens have been found in January and February.
Geographic distribution. In southern Argentina and Chile (Fig. 18). In Argentina this species has been found in Neuquén and Chubut, and in Chile from Regions X and XI.
Relationship to other Bembidion. B. tetrapholeon is a member of subgenus Nothonepha, as strongly supported by DNA sequences (Table 3) and the presence of shared, derived mesepisternal pits. B. tetrapholeon appears to be the sister of remaining Bembidion (Nothonepha) (Figs 7, 8). Four genes support this placement (CAD, wg, ArgK, and 18S; Table 3), although the support is weak or moderate in single-gene analyses.

Concluding remark
"when the same organ appears in several members of the same class, especially if in members having very different habits of life, we may attribute its presence to inheritance from a common ancestor." (Darwin 1859)  In groups such as beetles, in which a preponderance of lineages split without later reticulation, the evolutionary tree at the core of life's history yields hierarchical patterns in the distributions of characteristics. Any particular lineage in the tree, if separated long enough or with a high enough rate of evolution, will leave in the bodies of its descendants marks of its existence. The echoes from that deep historical well can reverberate down through later lineages in the form of signals scattered throughout the genomes. The repeated patterns of these branch markers both in the DNA and on the bodies of organisms are among the most compelling signs of the existence of the tree of life, and provide to us evidence about its shape. On occasion the clades thereby revealed are so unexpected that it is only with multiple independent markers, all showing the same pattern, that we can confidently accept the existence of the clade. Nothonepha is such a clade.
Many of us who study the diversity of life, and see the hierarchical patterns of organismal traits, are steeped in evidence of the existence of a genetic tree of life, so much so that we perhaps take it for granted (I often do). I think about the evidence about the tree's shape, but much less so evidence about its existence. In encountering the first evidence of Nothonepha, my belief in the tree-like structure of beetle genetic history was challenged, as the data made little sense in that light. As newly sequenced genes added to the evidence, my acceptance of the clade increased. The struggle was fully resolved when the mesothoracic pits came to light; this harmonizing of the morphological data with the molecular not only instilled a firm belief in this clade, but also a simple confirmation of the tree itself.

Acknowledgements
I am most grateful to Sergio Roig-Juñent, who arranged the collecting expedition that yielded most known specimens of Bembidion tetrapholeon, and most of the other Nothonepha studied; he was also a very welcome companion on the trip. I am very thankful to all those who helped with a collecting expedition to Ecuador during which most sequenced members of subgenus Ecuadion were collected. Mauricio Vega arranged many details of the trip, including relevant collecting and export permits; Wayne Maddison, Marco Reyes, and Mauricio Vega accompanied the author into the field and helped collect the specimens. I am also grateful to Reserva Yanacocha, Pichincha, Ecuador, for permission to collect on their lands. The Ecuadorian Ministry of the Environment and the Museum of Zoology of the Pontificia Universidad Católica de Ecuador assisted with permits.
I would also like to thank Elizabeth Arias and Kipling Will for organizing and providing excellent logistical support, including permit acquisition, for field work in Chile. I am also grateful for their companionship during my first experience with South American Bembidion.
Some key specimens were provided by others, and I am thankful for their contributions; without a network of collectors, a work such as this would not be possible. In particular, I would like to thank William D. Shepard (for the Rio Pullinque specimen of B. tetrapholeon), Wendy Moore (B. tucumanum), Karl M. Kjer (for South African Tachylopha), Geoff B. Monteith (for Australian Tachylopha), Caroline S. Chaboo (for South African Sphaerotachys), and James M. Pflug (for Oodinus from Texas).
I am also grateful to Luca Toledano, both for his review of the manuscript, and for the information he provided about specimens of B. tetrapholeon he has examined from Chile. My thanks to Wayne Maddison and an anonymous reviewer for their comments on the manuscript. Thanks as well to myrmecologists Alex Wild and Philip Ward for discussing the ants that might live in the same habitats as Nothonepha, Tachylopha, and Oodinus.
This project was made possible by funds provided by the University of Arizona, the Harold E. and Leona M. Rice Endowment Fund at Oregon State University, and National Sciences Foundation grant EF-0531754. National Sciences Foundation DEB-0445413, to Elizabeth Arias and Kipling Will, provided funds that helped support field work in Chile. The expedition to South Africa that yielded the Elaphropus (Sphaerotachys) I studied was supported by a Hellman Postdoctoral Fellowship (to C.S. Chaboo).