Taxonomy of the New World bee genus Agapostemon Guérin-Méneville – new names and synonymies (Hymenoptera: Halictidae)

Many early taxonomic works on North American bees were published by Europeans using specimens collected in the New World, some with type locations so imprecise that uncertainty on the nomenclatural status remains to this day. Two examples come from Fabricius (1745–1808) who described Andrena virescens Fabricius, 1775 and Apis viridula Fabricius, 1793 from “America” and “Boreal America”, respectively. The former species of Agapostemon Guérin-Méneville, 1844 occurs across most of the United States and southern Canada, the latter presumed an endemic to Cuba. The type materials of these two taxa have never been compared to each other, though a morphology-based phylogenetic analysis placed both in distinct species groups. Here we synonymize Apis viridula under Ag. virescens, thereby making Ag. femoralis (Guérin-Méneville, 1844) available as the name for the Cuban species. A lectotype for Ag. femoralis (the type species for the genus Agapostemon) is hereby designated to stabilize this taxonomy. We also synonymize Ag. obscuratus Cresson, 1869 under Ag. femoralis, suggesting that it represents a dark colour polymorphism. As Ag. cubensis Roberts, 1972 is a junior secondary homonym of Ag. cubensis (Spinola, 1851), we offer Ag. robertsi as a replacement name for the former.


Introduction
Much past confusion has surrounded the nomenclatural status of the North American sweat bee Agapostemon virescens (Fabricius, 1775) (Hymenoptera: Halictidae) largely due to the fact that, pre-1900, many taxonomists naming material from the New World were from Europe (e.g., Fabricius 1775Fabricius , 1793Lepeletier 1841;Smith 1853;Dalla Torre 1896) and vague or imprecise collection information was recorded for much of this material, including the type localities. Andrena virescens Fabricius, 1775 was named from material collected in "America" (Fabricius 1775); this species is widespread across most of the United States and southern Canada (Roberts 1972), but seemingly is not found in Mexico or Mesoamerica (Roberts 1972;Roberts & Brooks 1987;Ramírex-Freire et al. 2012); the few specimens identifi ed from Mexico, Honduras (det. by W.E. LaBerge), and Cuba (St. Clara) (see Ascher & Pickering 2020) may represent misidentifi cations. The latter is especially interesting as it likely represents a specimen of Ag. viridulus (Fabricius, 1793), a species now considered to be a Cuban endemic (Moure 1960;Roberts 1972;Moure et al. 2007). Cresson (1887) was the fi rst North American to apply a name currently associated with Ag. virescens to this taxon, recognizing Ag. nigricornis (Fabricius, 1793) as a valid species, likely following the classifi cation of Smith (1853); a more specifi c type locality of Georgia was provided for this taxon by Fabricius (1793), though Cresson (1887) also indicated that this species may be a synonym of Ag. sericeus (Forster, 1771). Soon after, several works on North American bees applied the name Ag. viridulus, and not the older Ag. virescens, to material collected in the United States and Canada, including Robertson (1895Robertson ( , 1897, Titus (1900), Crawford (1901, 1906(in Viereck et al. 1906), 1912, Vachal (1903), Cockerell (1902Cockerell ( , 1904Cockerell ( , 1909Cockerell ( , 1911, Shelford (1913Shelford ( , 1937, Stevens (1920), and Tucker (1909). Cockerell (1917a) later described Ag. tyleri Cockerell, 1917 from Mexico, but indicated that it was almost exactly like Ag. viridulus.
The name Ag. viridulus was likely applied to North American material because Robertson (1897) stated that "it is an absurd affection of authority to give this name [i.e., Ag. viridulus] to the Cuban species, before it is shown that Fabricius did not mean that his species came from the continent of North America, that he did not know where it came from, or that the description of Apis viridula does not apply to the North American species". In his study of Cuban bees, Baker (1906) expressed sentiments similar to those of Robertson (1897) on Ag. viridulus as a valid species in Cuba. Fabricius (1793) did not provide a type locality for Ag. viridulus in his original work, but in 1804 indicated that the species (as Megilla viridula) was from "Habitat in America boreali" (Fabricius 1804) which Robertson (1897) considered to mean North America. Robertson (1897) also indicated that Provancher (1882) likely misidentifi ed Ag. viridulus (= Ag. virescens) as Augochlora radiata (= Ag. sericeus (Forster, 1771)) based on the black metasoma (this misidentifi cation was confi rmed by Sheffi eld & Perron 2014), though in that same work he (i.e., Provancher 1882) recognized Ag. tricolor Lepeletier, 1841 (= Ag. virescens) as a distinct species. Crawford (1912) later used Ag. viridulus for specimens from Medicine Hat, Alberta, and the name Ag. virescens did not appear in his earlier revision of North American species (Crawford 1901), though again the morphology and geography fi t the current concept for this species (i.e., Roberts 1972). Though Ag. virescens is the oldest name for this taxon, Robertson (1897) and later authors did not consider Ag. viridulus a synonym of Ag. virescens presumably because Fabricius (1804) provided more precise type location information. However, Ag. virescens is now widely recognized as the only member of the genus occurring in eastern North America and all of Canada in which the female has the combination of a black metasoma (Sandhouse 1936;Roberts 1972;Packer et al. 2007) and dark-tipped clypeus. Ashmead (1900) was among the fi rst American entomologists to consider Ag. viridulus a Cuban species, likely based on the earlier opinion of Dalla Torre (1896), in contrast to the earlier opinion of Robertson (1897) quoted above. Dalla Torre's (1896) concept was subsequently followed by Friese (1902), with both considering Ag. femoralis Guérin-Méneville, 1844 a synonym of Ag. viridulus. The lectotype for Ag. viridulus was designated by Moure (1960) from two female specimens at the Sehested -Tønder Lund collection at the Zoologisk Museum in Copenhagen; Moure (1960) indicated that the species was apparently limited to Cuba, and provided a brief description but offered no diagnosis to separate this material from the more common and widespread North American species Ag. virescens despite the past controversy. Roberts (1972) accepted the treatment of Moure (1960), but indicated that due to the inadequacy of the original description, many authors believed that Ag. viridulus was a synonym of Ag. virescens. Almost 50 years later, Ag. viridulus is still considered an endemic to Cuba (e.g., Roberts 1972;Janjic & Packer 2003;Engel 2004;Genaro 2008). The male specimen of Ag. femoralis that Ashmead (1896, andsubsequently Krombein 1953) recorded from Eleuthera, Bahamas was presumably misidentifi ed; however, several other species are known from the Bahamas (Roberts 1972;Janjic & Packer 2003).
Another Cuban endemic, Ag. obscuratus Cresson, 1869 has also contributed to the confusion. At the time of describing Ag. obscuratus, Cresson (1869) fi rst thought it was likely only a variety of Ag. femoralis (= Ag. viridulus), but later indicated it was a distinct species, supporting the opinion of J.C. Gundlach (as cited in Cresson 1869); subsequently, Gundlach (1896) still considered it a distinct Cuban species. In fact, Cresson (1865) indicated that two of the 40 male specimens identifi ed as Ag. femoralis that he examined had the head and thorax dull black with very faint obscure blue colouration, matching what he would later (i.e., Cresson 1869) call Ag. obscuratus. Oddly, Cresson (1869) compared what he indicated as a female to the male of Ag. femoralis; Roberts (1972) correctly pointed out that the type materials of both taxa are male. Cockerell (1917b) seemingly made the same mistake for his Ag. obscuratus var. abjectus Cockerell, 1917, and also indicated that it was close to Ag. femoralis, but larger with the head and thorax coloured differently (i.e., obscurely dark green head and thorax in Ag. obscuratus abjectus) from Cresson's taxon that had the head and thorax black, with a more or less purple tinge. Since the last comprehensive revision (i.e., Roberts 1972), Ag. obscuratus has been considered a valid taxon.
Though the genus Agapostemon Guérin-Méneville, 1844 has received taxonomic treatment several times (e.g., Crawford 1901;Sandhouse 1936;Fischer 1950;Roberts 1972Roberts , 1973, only Roberts (1972) provided a comprehensive review of the genus throughout its entire range. Unfortunately, Roberts (1972) treated the species in a series of regional keys, and no comparisons were made that included both Ag. virescens and Ag. viridulus. However, Janjic & Packer (2003) produced a morphology-based phylogeny in which all species were included, including material from Cuba identifi ed as Ag. viridulus and Ag. obscuratus. Although Ag. virescens and Ag. viridulus are seemingly very similar (Roberts 1972), the results of Janjic & Packer (2003) clearly separated the two species, enough that they were placed in two different species groups, the latter belonging to a species group restricted to the Antilles which also contained Ag. obscuratus. Unfortunately, type material for all species was not compared and/ or used in that phylogenetic analysis, so they (i.e., Janjic & Packer 2003) used the taxonomic concepts of Roberts (1972; with some exceptions). Our purpose here is to clarify the taxonomy of Ag. virescens and Ag. viridulus.

Material and methods
Using the morphology-based matrix of Janjic & Packer (2003) as a guide, female specimens of Ag. virescens from throughout North America were examined for morphological consistency with the holotype (Fig. 1) and compared with the lectotype of Ag. viridulus (Fig. 2). The phylogeny of Janjic & Packer (2003) placed these two species into two morphologically unique species groups; those characters considered most useful (i.e., unique species specifi c character states) for distinguishing Ag. virescens from Ag. viridulus albeit chosen so not to damage Fabricius's types (i.e., not requiring dissection or relaxation of the specimens to permit movement of body parts) were compared, in addition to more general morphological features discussed below. The type material of Ag. femoralis (Fig. 3) and photos of the type material of Ag. semiviridis Cresson, 1865 (Fig. 4), Ag. obscuratus (Fig. 5), and Ag. obscuratus var. abjectus Cockerell, 1917(synonymized under Ag. obscuratus by Roberts 1972 (Fig. 6) were also examined; Ag. femoralis and Ag. semiviridis were previously synonymized under Ag. viridulus by Dalla Torre (1896) and Baker (1906), respectively, though the latter is the only taxon currently synonymized under Ag. viridulus that is from Cuba that has a female type specimen.
Morphological terminology generally follows that of Michener (2007), with some specifi c terms from Roberts (1972); surface sculpture terminology follows Harris (1979). Face length to width (L:W) ratios were calculated using the following: L = distance from lower edge of median ocellus to the basal edge of clypeus; W = greatest distance between the inner edges of the compound eyes, measured at apex of angulation.
DNA barcodes (Hebert et al. 2003) were also compared, using Cytochrome Oxidase I (COI) sequences previously published by Sheffi eld et al. (2009Sheffi eld et al. ( , 2017, and for material identifi ed as or sharing a Barcode Index Number (BIN, after Ratnasingham & Hebert 2013) with Ag. viridulus from Cuba in the Barcodes of Life Data (BOLD) system (Ratnasingham & Hebert 2007), this included a single specimen of Ag. obscuratus. Sequences were analyzed using the Barcode Gap Analysis tool on BOLD, using the Kimura 2 Parameter distance model with sequence alignment using MUSCLE. Using the same distance model and alignment parameters, nucleotides from sequences of a representative member of each BIN with a full DNA barcode were examined and compared.

Results
Morphologically, the lectotype of Ag. viridulus (Fig. 2) clearly matches Ag. virescens (in the splendens group of Moure & Hurd 1987) and not Cuban material in the the viridulus group seemingly confi ned to the Antilles (Moure & Hurd 1987;Janjic & Packer 2003). In fact, the females of Ag. virescens, including the type material of Ag. viridulus are quite different from material from Cuba, with specimens from the latter, including Ag. obscuratus, having a longer face (L:W = 0.78, versus 0.72 in Ag. virescens) (Fig. 7C, E; Roberts 1972: fi gs 33-34, 53-54), with prominent horizontal striations across most of the clypeus and supraclypeal area (Fig. 7C, E), while those of Ag. virescens and the type material of Ag. viridulus are coarsely and closely punctate, with striations limited to the medial area of supraclypeal area (Fig. 7A). Other morphological differences include the propodeal surface of Ag. virescens and the type material of Ag. viridulus, which is coarsely rugose, with the posterior surface of the propodeum encircled by a strong carina (Fig. 8A), while the surface is costulate to costate in Cuban material, including Ag. obscuratus, with the carina much less distinct, especially laterally (Fig. 8C).
Though males of Ag. virescens and Cuban material identifi ed as Ag. viridulus were also not compared or keyed by Roberts (1972), they are distinctive. Males of Ag. virescens have extensive yellow maculations, including a more pronounced triangular mark on the clypeus apically that extends basad medially (Fig. 7B), while the clypeal maculation on Cuban material (Fig. 7D), including Ag. obscuratus (Fig. 7F) is linear. Most of the legs in male Ag. virescens are extensively yellow, with dark markings restricted to the outer basal area of the fore and mid femur, an elongate basal mark basally on the inner front tibia, two elongate marks on the inner and outer basally on the mid tibia, and the hind leg with dark spot at the apex of femur, a small basal spot and an elongate spot in the outer basal half of the tibia; the legs of Cuban males, including Ag. obscuratus are primarily dark with yellow areas restricted to the inner surfaces, and the basal half of hind femur; Ag. virescens have the metasoma with a submedian yellow band on tergum 1, and extensive basal bands on terga 2-5 (Fig. 9A), while the metasoma of Cuban material is entirely dark (Fig. 9B-C). Structurally, the males of Ag. virescens also have a shorter face (L:W = 0.72) (Fig. 7B) than material from Cuba (L:W = 0.8) (Fig. 7D, F). In addition, the hind tibia and femur of Ag. virescens are relatively narrow, less than half as wide as long, while more robust in Cuban material including Ag. obscuratus, the femur less than twice as long as wide (Figs 3C, 5A, 6A); the hind tarsi of Ag. virescens are unmodifi ed (Fig. 10A) while the hind basitarsus of Cuban material has a prominent basal ridge and an apical groove ( Fig. 10B; Roberts 1972: fi gs 150, 158). The propodeum of Ag. virescens is coarsely rugose over its dorsal and lateral surfaces, with the posterior surface encircled by a strong carina (Fig. 8B), while males of Cuban material have a less distinctive carina (entirely lacking laterally), with the propodeal surfaces striate (Fig. 8D).
Based on our comparison of the type material of Ag. viridulus (Fig. 2) to female specimens of Ag. virescens from North America, including the holotype (Fig. 1), we conclude that the former is a synonym of the latter. As such, Ag. femoralis, with a type locality of Cuba (Fig. 3) would be the oldest name available for the Cuban species. We also rename the viridulus species group of Moure & Hurd (1987) the femoralis species group to refl ect this. We conclude that, though Robertson (1897) felt that Ag. viridulus was the priority name for the North American taxon, the type specimen of Ag. virescens matches North American material, and we assume that the type locality "America" referred to North America (the type material has previously even been labelled as such, see Fig. 1), as this specimen does not match other species outside of the continent (i.e., Central or South America, or the Antilles). This also stabilizes the taxonomy of a species that has been subject to many ecological and taxonomic studies (e.g., LaBerge & Ribble 1966;Roberts 1972;, 1981Eickwort 1981).
The BIN for Ag. virescens is AAB2708, and currently in BOLD there are 90 sequences from across Canada and the eastern United States; within this BIN the mean distance is 0.36%, with a maximum distance of 1.48%. There are two sequences (BOLD Process IDs BEECB084-07 and BEECB087-07) available for Ag. femoralis from Cuba (AAJ7225; the former of these identifi ed as Ag. obscuratus Cresson, 1869), differing from that of Ag. virescens by 12.1%, with 86 nucleotide differences, and supporting that though these species are considered morphologically very similar (i.e., Roberts 1972), there is enough genetic and morphological divergence to distinguish these species. Though only a single specimen of each taxon was sequenced, barcoded material of Ag. obscuratus, also from Cuba, is genetically almost identical to the sequence of Ag. femoralis, differing in only one nucleotide at position 494 (A to C, respectively). In addition, apart from colour, the morphology of both taxa are identical, Ag. obscuratus abjectus (synonymized under Ag. obscuratus by Roberts 1972) representing an intermediate coloured form.
Lastly, Ag. cubensis Roberts, 1972 became a junior secondary homonym of Ag. cubensis (Spinola, 1851) when Engel (2004) designated a lectotype for the latter taxon and recognized it as a synonym of Ag. viridulus [= Ag. femoralis]. Roberts' taxon is known only from the male holotype (Roberts 1972), and since it was described the name has only been used in a handful of other publications (i.e., Alayo 1973Alayo , 1976Janjic & Packer 2003). Therefore, we offer Ag. robertsi as a replacement name for Ag. cubensis Roberts, named for the late Radclyffe Burnand Roberts (1938Roberts ( -1988

Remarks
Under his discussion of Ag. virescens, Cockerell (1921) included a brief mention of Apis vitreus, originally named and described from the male (with a colour illustration) by Harris (1782: pl 39 fi g. 17); in the fi gure caption, Harris (1782) mentioned the brilliant green head and thorax, with yellow mouthparts, antennae, legs and abdomen, the latter with six black bands, which is consistent with most male Agapostemon in North America (e.g., Roberts 1972). Cockerell (1921) felt that this name should be considered valid, and presumably a synonym of Ag. virescens. As Harris's (1782) work was focused on the insects of England, Warncke (1973) considered it (as Apis vitreus Harris, 1776, from England) a synonym of Halictus tumulorum (Linnaeus, 1758). However, Ebmer (1974) considered Harris's taxon nomen dubium, indicating that the synonymy of Warncke (1973) was incorrect, and suggested that the large size of 11.5-12.5 mm (from Ebmer (1974), based on Harris's (1782) mention of "fi ve lines and a half") and description supported that it was likely a male of Agapostemon, partially supporting Cockerell's (1921) opinion. Ebmer (1974) also indicated that it was likely that at the time of collection (i.e., pre-1776), the New England states (where it was presumably collected) were regarded as belonging to the "motherland" (i.e., England), though by the time of publication of Harris's works (i.e., 1776-1782) the United States was independent. However, as four species of Agapostemon with males generally matching that described by Harris (1782) are found in the New England states, we follow Ebmer's (1974) recommendation and consider Harris's species nomen dubium until the type material can be located.
"America" was indicated as the type locality for Ag. virescens by Fabricius (1775), but the type specimen has seemingly been re-labelled after the fact (i.e., post Cresson 1887) with N[orth]. Amer[ica]. (Fig. 1); the specimen was previously examined by Cockerell (1921) and later by Charles D. Michener (as per Roberts 1972) who both considered it a valid representation of this taxon. Smith (1853) considered Apis virescens Fabricius 1793 (not 1775) a synonym of Ag. nigricornis, presumably not realizing that the species had been named 18 years previous (though he attributed it to the Banks Collection, as is the type material).

Notes
The lectotype ♂ of Halictus cubensis Spinola, 1851(designated by Engel 2004, from Cuba, Havana, Spinola Collection, MSNT, was not examined as the synonymy was not questioned.

Remarks
The name and description of Andrena femoralis is attributed to Guérin-Méneville (1844a), though the name (i.e., Andrena femoralis Guer.), type locality (i.e., Cuba) and fi rst illustration appear on plate 83, fi gure 1 (incorrectly recorded as plate 73 by Van der Vecht 1957) for Cuvier's (1836) work published in 1837 (Cuvier 1837). These images were later duplicated in Guérin-Méneville (1844b) as "Insectès, Pl[ate]. 73, Figure 1". Banks (1909)  indicating that the volume dedicated to insects (i.e., Guérin-Méneville 1844a) is dated 1829-1838 with many references to other dates within the text, including 1844, the year most researchers have used in citing this work (e.g., Roberts 1972;Michener 2000, 2007, though Moure (1960 indicated 1845). However, as Cuvier's original plates were published in 1837 (Cuvier 1837), Banks (1909) felt that those of the opinion "that a named fi gure is valid without [accompanying] text must credit such names" with the dates of the plate publication. However, Cowan (1971) more recently reviewed the issue of Guérin-Méneville's works and concluded that the date relevant to the insects is 1844, and this decision is followed here. Van der Vecht (1957) examined material that he assumed was part of the type material for Ag. femoralis at the Rijksmuseum van Natuurlijke Historie, Leiden (now Naturalis Biodiversity Center, Leiden), though he also indicated the possibility of syntypes existing at other institutions; other specimens with the same collection information exist at Naturalis Biodiversity Center. Thus, to stabilize Ag. femoralis as the valid name of this Cuban taxon, the specimen Van der Vecht (1957) provided details on (i.e., fi g. 3) is hereby selected as the lectotype. Incidentally, Ag. femoralis is the type species for the genus Agapostemon   Cresson, 1869; note, the apex of clypeus of female is damaged. (Sandhouse 1936(Sandhouse , 1943Michener 1997). Dalla Torre (1896) was the fi rst to treat Ag. femoralis as a synonym of Ag. viridulus (later followed by Friese 1902), though the assumption that the type material of Ag. viridulus was collected in Cuba was baseless.
The female type material of Halictus cubensis Spinola was not an Agapostemon, but rather Augochlora regina Smith, 1853 (Augochlorini) (Engel 2004), so Engel (2004) designated the male as a lectotype, and placed it into synonymy with Ag. femoralis. However, as a result of this designation, Ag. cubensis Roberts became a junior secondary homonym of Ag. cubensis (Spinola), thus requiring the replacement name provided below.
Though Roberts (1972) records the synonymy of Ag. semiviridis Cresson under Ag. viridulus as new, Baker (1906) had already treated it as a synonym. When described, Cresson (1865) indicated that it was potentially the female of Ag. viridulus.

Discussion
Though previous workers assumed that the type material of Ag. viridulus was from, and presumably endemic to Cuba (e.g., Dalla Torre 1896;Ashmead 1900;Cockerell 1910;Moure 1960;Roberts 1972), there is no evidence that this is true. In fact, in Fabricius's collected works there is only one bee species that was described with a type locality from Cuba, Coelioxys tridentata (Fabricius, 1775) [described as Apis tridentata] (Genaro 2006). As indicated by Genaro (2006), most early works on bees in the area are generally identifi ed as coming from "Americae Meridionalis Insulis" (includes Cuba and/or another Caribbean island). Thus, it is hard to believe that Fabricius would have provided a specifi c locality for a species from Cuba in an earlier work (Fabricius 1775), though not for a species described in 1793 for which no locality was recorded until done so in 1804 (Fabricius 1804). However, it is also likely that detailed information was lacking for many of the specimens that Fabricius received from the Western Hemisphere.
Within the BIN to which Ag. virescens belongs, there is little genetic variation, though members differ by over 12% from Ag. femoralis, suggesting that although these species were considered morphologically very similar by Roberts (1972), there is enough genetic and morphological divergence to distinguish the species. The COI sequence of Ag. obscuratus, also from Cuba, is genetically identical to material now recognized as Ag. femoralis. The phylogeny of Janjic & Packer (2003) recovered Ag. femoralis [as Ag. viridulus] and Ag. obscuratus as separate taxa, albeit within the same species group, suggesting that a new phylogenetic analysis is warranted, though this was likely due in large part to colour differences between the material examined. Based on genetic similarity, shared geography, and past opinions of Cresson (1869) and Roberts (1972), we consider the taxon Ag. obscuratus to represent a colour polymorphic form of Ag. femoralis, and place it into synonymy. As Roberts (1972) synonymized Cockerell's (1917b) obscurely dark green (i.e., intermediate in colour; Fig. 6) taxon under Cresson's (1869), we assume he was also of the opinion that colour polymorphisms existed in Agapostemon. Though Roberts (1972) also considered the specifi c rank of Ag. obscuratus questionable, he did not synonymize it under Ag. viridulus despite noting only differences in colour between the two taxa. Such variation within a metallic bee species, ranging from metallic green or blue to black is uncommon, but not unheard of in bees. For instance, Osmia (Melanosmia) tersula Cockerell, 1912 (Megachilidae) is typically metallic blue in most of its range (Sandhouse 1939;Mitchell 1962), though black specimens lacking metallic colouration do occur (Rightmyer et al. 2010). Many metallic halictid bees show extensive congeneric (e.g., Gibbs 2010Gibbs , 2011Engel 2013Engel , 2014aEngel , 2014b and conspecifi c (e.g., Ordway 1966) variation in body colouration; Ordway (1966) and Gibbs (2009) indicated that killing agents and other chemicals can alter metallic colouration to some extent. Combining molecular methods such as DNA barcoding to studies of bee taxonomy has proven very useful for recognizing colour variable species (e.g., Sheffi eld et al. 2011;Sheffi eld et al. 2020).
type materials from (ANSP). The Biodiversity Heritage Library (https://www.biodiversitylibrary.org/) has continuously proven invaluable as a resource for the study of older, highly relevant taxonomic literature, and was used extensively while completing this work.