OUP user menu

New genus and two new species of Pleistocene oryzomyines (Cricetidae: Sigmodontinae) from Bonaire, Netherlands Antilles

Jelle S. Zijlstra, Paulina A. Madern, Lars W. van den Hoek Ostende
DOI: http://dx.doi.org/10.1644/09-MAMM-A-208.1 860-873 First published online: 16 August 2010

Abstract

A collection of about 500 molars from 5 Pleistocene localities on the island of Bonaire (off the coast of Venezuela), previously identified as Thomasomys sp., is reidentified as representing a new genus and 2 new species of oryzomyine rodents based on comparative examination and phylogenetic analysis of morphological characters. The material from 1 of the 5 localities is distinguished by its smaller size and several discrete characters; a larger species is represented in the other 4 localities. In addition, a single edentulous dentary represents a different species that is described as an indeterminate genus and species of Sigmodontinae.

Key words
  • biogeography
  • Bonaire
  • endemics
  • new genus
  • Oryzomyini
  • Pleistocene

Sigmodontine rodents (subfamily Sigmodontinae Wagner, 1843) rate among the largest mammalian clades recognized at the subfamilial rank. The most recent classification (Musser and Carleton 2005) included 377 species in 74 genera arranged in 8 tribes, with 10 genera left unassigned. These species inhabit many different habitats in their range, which includes all of South and Central America and parts of North America. Their diversity is the result of an impressive adaptive radiation, which encompasses such distinctive forms as the aquatic ichthyomyines, the gerbil-like Eligmodontia, and the fossorial Geoxus. The enormous extant diversity of the subfamily contrasts markedly with its scarce fossil record (Pardiñas et al. 2002); many genera, and even some tribes, have no fossils assigned to them, and the fossil record is of very little help in resolving the relations among extant clades. Despite a massive taxonomic literature (Musser and Carleton 2005), the taxonomy of the Sigmodontinae is still in a fluid state, with recent major revisions at the tribal (D'Elía et al. 2007), generic (Weksler 2006), and specific (Percequillo et al. 2008) levels. Although a classification of the subfamily's numerous genera into tribes has existed for decades, many taxonomic questions have been resolved only recently, and some others remain to be solved, as reflected by the number of genera that Musser and Carleton (2005) left incertae sedis because they could not confidently place them within any 1 tribe.

Of the 9 tribes currently recognized (D'Elía et al. 2007), the largest in terms of number of species and genera is Oryzomyini, which includes 121 species in 28 genera (Weksler et al. 2006). It also is the most widely distributed tribe, occurring from the eastern United States through all of Central and South America, including many oceanic islands, south to Tierra de Fuego in southernmost Argentina and Chile (Musser and Carleton 2005). This tribe recently has been the subject of a detailed phylogenetic study (Weksler 2006), which has resulted in a much improved genus-level classification (Weksler et al. 2006) and an enhanced understanding of the tribe's evolution. The Oryzomyini is notable for including a large number of endemic island forms, which are rare among other sigmodontine tribes. These insular endemics have been recorded from, among other localities, the Galápagos Islands (Nesoryzomys spp. and Aegialomys galapagoensis), the Brazilian island of Fernando de Noronha (Noronhomys vespuccii), and the Lesser Antilles (Megalomys spp., Oligoryzomys victus, and several unnamed species of uncertain generic affinities). With the exception of several Galápagos species, all of these have become extinct in the last few centuries (Musser and Carleton 2005). The fossil record of the tribe is especially sparse. The oldest representative is Carletonomys cailoi from the early or middle Pleistocene (Ensenadan) of the Argentinean pampas (Pardiñas 2008), and only 2 other exclusively pre-Holocene species have been described. One of these species, Holochilus primigenus Steppan (1996) from the Ensenadan of Bolivia, belongs to the same highly derived clade as Carletonomys (Pardiñas 2008). The other species, Megalomys curazensis Hooijer (1959) from the Pleistocene of Curaçao, Netherlands Antilles, is among the few insular sigmodontines which did not go extinct because of anthropogenic causes. It did not survive into the Holocene, perhaps because of competition with Oryzomys gorgasi, which is known from Holocene deposits on Curaçao (McFarlane and Debrot 2001; Voss and Weksler 2009).

In addition to describing M. curazensis as a new species, Hooijer (1959) also identified some fossil rodent teeth from the neighboring island of Bonaire as Thomasomys sp., which he considered most similar to the Brazilian species now placed in the separate genus Delomys (Musser and Carleton 2005). This identity is biogeographically implausible, because Delomys is restricted to the forests of southern Brazil and adjacent Argentina. The record from Bonaire was not noted in the revision of Delomys by Voss (1993). Although the genus Thomasomys, which is now used in a more restricted sense than by Hooijer (1959), does occur in northwestern Venezuela, it is an unlikely candidate to be found on Bonaire. This genus is found only in montane habitats in the Andes above 1,500 m in elevation and is not known from any offshore islands (Musser and Carleton 2005). The Bonaire Thomasomys remains the only record of the tribe Thomasomyini from an oceanic island (Musser and Carleton 2005) and is 1 of 2 fossil occurrences reported for the tribe. Undetermined species of Thomasomys are known from the late Pleistocene and Holocene of Ecuador (Pardiñas et al. 2002). The record of Hooijer (1959) has been mentioned intermittently in the literature (De Buisonjé 1974; McFarlane and Debrot 2001; Steadman and Ray 1982), but the material has never been restudied to confirm its identification.

We reexamined all Pleistocene rodent material from Bonaire examined by Hooijer (1959) and De Buisonjé (1974) to reconsider its taxonomic relations. Based on a phylogenetic analysis and on comparisons with other sigmodontine taxa, we conclude that the material represents a new genus, including 2 species that are described and compared herein.

Materials and Methods

The material studied is housed in the collections of the Netherlands Center for Biodiversity Naturalis in Leiden, The Netherlands (RGM 257819–257823 and 444400–444890; and numerous unnumbered specimens). This material was excavated from 5 different localities (Fontein, Porto Spanjo, Barcadera-Karpata, 80 m above sea level, and Seroe Grandi; Fig. 1) on the island of Bonaire, Netherlands Antilles. All localities are fissure fillings in raised Pleistocene limestone plateaus. See De Buisonjé (1974) for additional geological information. All molars were measured using a Leica Ortholux measuring microscope supplied by Leica Microsystems, Heerbrugg, Switzerland. Comparative material of Megalomys curazensis (RGM 433000–433203) and Oryzomys gorgasi (RGM 593395–593400) was examined in the Naturalis collections. In addition, illustrations and descriptions in Hershkovitz (1944), McFarlane and Debrot (2001), Pacheco (2003), and Weksler (2006) were used for comparative purposes. We follow Reig (1977) for molar terminology and refer to upper and lower molars as M1, M2, and M3 and m1, m2, and m3, respectively.

Fig. 1

Map of Bonaire showing the capital of Kralendijk and the localities of De Buisonjé (1974): Seroe Grandi, Porto Spanjo, Fontein, and Barcadera-Karpata. The location “80 m above sea level” could not be located. The inset map indicates the location of Bonaire off the coast of Venezuela.

We follow Weksler (2006), Musser and Carleton (2005), and Weksler et al. (2006) for oryzomyine taxonomy and phylogeny. To determine the phylogenetic relations of our Bonaire material within the Sigmodontinae, we scored both species described below for applicable characters in Pacheco (2003) and added them to his data matrix. Characters scored are 69 (Ml anteromedian flexus), 70 (size of anterolabial and anterolingual conule), 71 (M1–2 metaloph), 72 (M1–2 mesoloph), 73 (M3 mesoloph), 74 (upper molars labial cingulum), 75 (M2 protoflexus), 77 (interpenetration of flexi), 78 (m1 cusp arrangement), 79 (ridge on m1 anterolabial cingulum), 81 (m1 anteromedian fossettid), 82 (m3 posteroflexid), 83 (mesolophid), 84 (M3 metacone), 85 (M1 accessory root), 86 (M1–2 lingual roots), and 87 (m2 roots). The resultant matrix was analyzed in PAST (Hammer et al. 2001) to determine the phylogenetic relationships of our new species using unweighted parsimony analysis. Polymorphic data were treated as unordered. No bootstrapping was carried out, because almost all resolution in the phylogenetic trees would be lost due to the large amount of missing data in the new taxa. To further determine species relationships within the Oryzomyini, we scored them for applicable characters from Weksler (2006) and added them to his data matrix. Applicable characters included 29 (position of zygomatic plate), 31 (length of incisive foramen), 44 (orientation of mental foramen), 45 (capsular process), 46 (condition of masseteric ridges), 47 (extent of masseteric ridges), 49 (accessory roots of Ml), 50 (accessory roots of m1), 51 (roots of m2), 54 (hypsodonty), 55 (presence of labial cingulum), 57 (interpenetration of flexi), 58 (condition of Ml anterocone), 59 (M1 anteroloph), 60 (M1 protostyle), 61 (attachment of M1 paracone), 62 (M1–2 mesoloph), 63 (attachment of M1 median mure), 64 (M2 protoflexus), 65 (accessory loph of M2 paracone), 66 (division of M2 mesoflexus), 67 (M3 mesoloph), 68 (M3 posteroloph), 69 (M3 hypoflexus), 70 (condition of m1 anteroconid), 71 (m1 anterolabial cingulum), 72 (m1 ectolophid and ectostylid), 73 (m1–2 mesolophid), 74 (m2 anterolabial cingulum), 75 (m2–3 anterolophid), 76 (m3 anterolabial cingulum), and 77 (m3 posteroflexid). The same procedures as in the 1st phylogenetic analysis were used to analyze this matrix.

Results

Systematic Paleontology

Family Cricetidae Fischer, 1817

Subfamily Sigmodontinae Wagner, 1843

Tribe Oryzomyini Vorontsov, 1959

Agathaeromys, new genus

Type species.Agathaeromys donovani, new species.

Referred species.Agathaeromys praeuniversitatis, new species.

Distribution and age.—Known exclusively from 5 localities on the island of Bonaire, Netherlands Antilles, Kingdom of the Netherlands. Age is probably Pleistocene, approximately 900 to 230 thousand years ago (kya).

Etymology.—Named after the island of Bonaire, to which this genus is endemic, from Ancient Greek ἀμαθóς or agathos meaning good, ἀηρ or aêr meaning air, and μυς or mus meaning mouse, interpreting Bonaire as meaning good air in French. The reference to “good air” also may be taken metaphorically to refer to the fresh air in oryzomyine systematics created by the rigorous phylogenetic analysis of the tribe by Weksler (2006) and the subsequent elimination of Oryzomys as a polyphyletic wastebasket taxon by Weksler et al. (2006).

Diagnosis.—A member of Sigmodontinae distinguished from all other members of the subfamily by the following combination of characters: mesolophs present on M1, M2, and M3; paracone on M1 and M2 connected to anterior or middle portion of protocone; flexi of M1 and M2 meet at midline; protostyle on M1 absent; median mure of M1 connected to protocone; accessory labial roots on M1 present; accessory loph posterior to M2 paracone absent; M3 posteroloph present; mesolophids present on m1 and m2; entolophid on m1 and m2 oriented to murid; m1 anterolabial cingulum present; ectolophid and ectostylid on m1 absent; anterolophid on m2 and m3 absent; anterolabial cingulum on m3 absent.

Comparisons

Agathaeromys differs from all other oryzomyine genera by at least 2 characters (Table 1). We specifically compared Agathaeromys to the 2 extinct oryzomyines known from the nearby island of Curaçao (Oryzomys gorgasi and Megalomys curazensis) and to the phenetically similar taxa Nectomys and Sigmodontomys alfari. In all comparisons conditions in the taxa compared to are shown within parentheses.

View this table:
Table 1

Summary comparisons of Agathaeromys with other oryzomyine genera based on applicable characters in Weksler (2006). Characters with different states for the 2 Agathaeromys species were omitted, as were characters that distinguish only a single other genus from Agathaeromys (characters 60 from Sooretamys; 54 and 63 from Holochilus; 65 from Oecomys; and 74 from Sigmodontomys [polymorphic]). “Handleyomys-sl”refers to the Oryzomys species provisionally transferred to Handleyomys by Weksler et al. (2006). Polymorphisms are indicated by slashes.

To Nectomys (Hershkovitz 1944).—Cusps in general much broader and flexi shorter, molars less lophodont (cusps narrower and flexi longer, more lophodont); M1 anteroloph distinct from anteromedian fossette if present (indistinct); M1 and M2 paracone connected to middle portion of protocone (connected to posterior protocone); m1 anteroflexid usually absent, but confined to lingual side of procingulum if present (present as a long fossettid in the procingulum, perpendicular to the main axis of the tooth); median mures and murids nearly straight (noticeably curved at connection to metacone or entoconid); distinct posterostylid present on m2 (absent); no internal fossettid in m2 and m3 metaconid (internal fossettid in metaconid present); m3 mesolophid usually absent (present).

To Sigmodontomys alfari (Hershkovitz 1944).—Cusps in general much broader and flexi shorter, molars less lophodont (cusps narrower and flexi longer, more lophodont); M1 and M2 paracone connected to middle portion of protocone (connected to posterior portion of the protocone); m1 anteroflexid usually absent, but confined to lingual side of procingulum if present (present as a long fossettid in the procingulum, perpendicular to the main axis of the tooth); m2 and m3 anterolabial cingulum present (absent); m3 mesolophid usually absent (present).

To Oryzomys gorgasi (McFarlane and Debrot 2001).—M1 mesoloph complete (incomplete, not reaching labial margin); M3 labial cingulum very strong (absent); m1 anteroflexid absent (present, distinct from anteromedian fossettid and metaflexid); m1 and m2 entoconid usually connected to hypoconid or to mesolophid or both (connected to mesolo-phid-murid angle); m1 and m2 lingual cingulum usually weak (stronger); m3 mesolophid usually absent (present).

To Megalomys curazensis (specimens in RGM).—Cusps in general much broader and flexi shorter, molars less lophodont (cusps narrow, flexi long, molars more nearly lophodont); M1 anteromedian fossettid smaller (larger); M1 paracone connected to middle of protocone (connected to posterior portion of protocone or to median mure, posterior to the protocone); M1 and M2 paracone usually connected to mesoloph (usually not connected); median mures and murids nearly straight (noticeably curved at connection to metacone or entoconid); M2 posteroloph not connected to metacone (connected, creating an internal fossette); M3 reduced, with flexi absent or small (much less reduced); M3 labial cingulum strong (absent); M3 hypoflexus weak or absent (strong); m1 anteromedian fossettid present as a circular fossettid in the middle of the procingulum (present as a long, narrow fossettid in the procingulum perpendicular to the main axis of the tooth); m1 metaconid connected to procingulum or to protoconid or both (usually without any connections); m1 entoconid connected to middle of mesolophid (not connected, or connected at lingual cingulum); m2 metaconid about equal in size to protoconid (metaconid smaller); m2 posteroflexid broad (narrower); m3 mesolophid usually absent (present); m3 posterostylid absent (usually present).

Agathaeromys donovani, new species

Thomasomys sp.: Hooijer (1959); De Buisonjé (1974).

Holotype.— RGM 444670, an isolated left Ml (Fig. 2a).

Fig. 2

Type series of Agathaeromys donovani from Fontein, Bonaire, a) Left M1 (RGM 444670, holotype of A. donovani); b) left M2 (RGM 444731); c) left M2 (RGM 444747); d) left M3 (RGM 444823); e) left m1–3 (RGM 444633).

Type locality.—Fontein, Bonaire, Netherlands Antilles.

Paratypes.—Total 258 specimens, all from the type locality: 45 Ml (RGM 444637–444669, 444671–44682); 42 M2 (RGM 444713–444754); 47 M3 (RGM 444796–444842); 28 m1 (RGM 444683–444712); 41 m2 (RGM 444755–444795); 48 m3 (RGM 444843–444890); 1 maxillary with M1–2 (RGM 444634); 1 complete mandibular toothrow (RGM 444633); 2 mandibles with m1–2 (RGM 444635, 444636).

Distribution and age.—Known from Pleistocene fossils from 4 localities (Fontein, Porto Spanjo, Barcadera-Karpata, and “80 m above sea level”) on the island of Bonaire. Age probably about 900 to 540 kya.

Referred specimens.—Barcadera-Karpata (total 54): 7 M1 (RGM 444436–444442); 13 M2 (RGM 444454–444466); 6 M3 (RGM 444473–444478); 12 m1 (RGM 444435,444443– 444453); 6 m2 (RGM 444467–444472); 10 m3 (RGM 444479–RGM 444488).

Porto Spanjo (total 148): 19 Ml (RGM 444496–444514); 20 M2 (RGM 444542–444560, 444586); 20 M3 (RGM 444587–444606); 27 m1 (RGM 444515–444541); 25 m2 (RGM 444561–444585); 26 m3 (RGM 444607–444632); 1 maxillary with M1–2 (RGM 444494); 1 broken mandible with m1 (RGM 444595); 9 isolated edentulous dentaries.

80 m above sea level (total 5): 2 M3 (RGM 444490, 444491); 1 m1 (RGM 444489); 2 m3 (RGM 444492, 444493).

Etymology.—Named after Prof. Dr. Stephan K. Donovan in honor of his contributions to the knowledge of the geology and zoology of the Caribbean region.

Measurements.—Holotype (M1): length 2.63 mm, width 1.76 mm. See Table 4 and Fig. 3 for further measurements.

Fig. 3

Scatterplots of measurements (mm) of molars of Agathaeromy s by locality. a) m1; b) m2; c) m3; d) M1; e) M2; and f) M3. Legend: stars, Seroe Grandi; circles, Barcadera-Karpata; crosses, Porto Spanjo; triangles, pointing up, 80 m above sea level; and triangles, pointing down, Fontein.

Diagnosis.—See generic diagnosis for distinction from other genera. Distinguished from the other new species described herein by the following characters (the condition for the other new species is within parentheses): large, Ml length > 2.03 mm (small, M1 length < 1.94 mm); anteromedian flexus on M1 usually absent, but anteromedian fossette usually present (flexus usually present, fossette always absent); M1 metacone usually connected to base of posteroloph (to hypocone); Ml labial cingulum usually present, but mesoflexus open (cingulum absent); m1 anteromedian fossettid usually present (absent); m2 with 2 anterior roots (one anterior root).

Description of Type Series

M1 (n = 48).—Anteromedian flexus usually absent, superficial when present. Anteromedian fossette present. Anteroloph usually present. All procingulum structures disappear with little wear. Protocone connected to anterior and median mure. Paracone large, connected to middle part of protocone. Additional connection to mesoloph usually present. Mesoloph long, reaching labial cingulum. Protostyle and enterostyle absent. Hypoflexus broad. Metacone smaller than hypocone. Posteroloph present, but posteroflexus narrow and shallow. Metacone usually connected to base of posteroloph, sometimes additionally to hypocone or hypocone-mesocone connection. Labial cingulum present, but mesoflexus open. Flexi meet at midline. One anterior root, 2 posterior (1 lingual, 1 labial), 1 accessory medial labial root (roots observed in 27 specimens).

M2 (n = 43).—Procingulum largely absent, but anteroloph present and large. Paracone large, connected to mesoloph, middle of protocone, and sometimes also to anteroloph, connection to mesoloph absent in 1 case. Protocone large. Protoflexus present. Mesocone present, connected to protocone, mesoloph, and hypocone. Mesoloph complete, reaching labial margin. Hypocone large. Posteroloph present, usually enclosing a circular fossette behind the metacone. Metacone connected to hypocone at posterior margin. Labial cingulum present, but mesoflexus open. Flexi meet at midline. 2 labial roots (anterior and posterior), 1 lingual (roots observed in 24 specimens).

M3 (n = 47).—Triangular shape. Procingulum and anterolabial and anterolingual cingula absent. Paracone large, usually containing a large internal fosette. Sinus between protocone and paracone absent. Protocone large. Hypoflexus present. A large fossette is present in the middle of the molar, sometimes split into an anterior and posterior fossette by an additional connection between protocone and paracone. Mesoloph large, reaching the labial cingulum and connected to paracone and hypocone or protocone-hypocone connection. Metacone present, connected to hypocone and labial cingulum. Hypocone present, sometimes fused with metacone. Posteroloph present, sometimes absent. Labial cingulum complete, closing posteroflexus, metaflexus, and mesoflexus. Two anterior roots (1 labial, 1 lingual), 1 posterior (roots observed in 13 specimens).

m1 (n = 31).—Cusps arranged oppositely. Anteromedian flexid usually absent, anteroconid also sometimes superficially divided. Anteromedian fossettid present, sometimes connected to protoflexid but never to lingual margin. Anterior murid connecting protoconid and anterolabial conulid. Metaconid connected to anterior murid or connected to anterolingual conulid angle at lingual margin of anteromedian fossettid, usually doubly, sometimes connected to both. Mesoflexid long and broad. Anterolabial cingulum present, but not wholly closing protoflexid. Protoconid triangular, connected to mesoconid via median murid. Ectolophid absent. Mesolophid complete, reaching lingual margin, connected to mesoconid and entoconid, fused with entoconid in worn molars. Entoconid large, also connected to middle or anterior portion of hypoconid, connection to hypoconid sometimes absent. Posterolophid long and broad. Lingual cingulum absent. One anterior root, 1 posterior. Accessory labial and lingual rootlets usually present, particularly lingual rootlet sometimes absent (roots observed in 14 specimens).

m2 (n = 44).—Procingulum absent. Metaconid connected to protoconid at anterior margin. Anterolophid absent. Anterolabial cingulum present. Protoconid connected to mesoconid via median murid. Mesolophid complete, reaching lingual margin. Entoconid connected to anterior portion of hypoconid or to hypoconid-mesoconid connection, usually not connected to mesolophid; the connection to the hypoconid may also be absent, but the connection to mesolophid invariably present in that case. Hypoconid large, connected to mesoconid. Posterolophid broad to very broad. Labial cingulum incomplete, but entoflexid sometimes closed. Two anterior roots, 1 posterior; anterior roots sometimes fused (roots observed in 20 specimens).

m3 (n = 50).—Procingulum and anterolabial and anterolingual cingula absent. Metaconid small, connected to protoconid at anterior margin. Anterolophid absent. Mesoflexid and hypoflexid broad. Entoconid small, located before the hypoconid, usually connected to metaconid and hypoconid via median murid, but connection to median murid sometimes absent. Mesolophid absent, sometimes present. Hypoconid large. Posterolophid present, very broad. Lingual cingulum usually complete, but mesoflexid often open. Two anterior roots (usually almost completely fused), 1 posterior (roots observed in 17 specimens).

Description of Referred Material

In the following descriptions only traits that differ from those of the type series are mentioned.

80 m above sea level (total n = 5)

M3 (n=2).—Mesoloph present, but small in 1 molar. Metacone also connected to paracone. Roots as in type series (observed in 2 specimens).

m1 (n = 1).—Metaconid small, connected doubly to lingual side of procingulum. Entoconid much smaller than hypoconid, about equal to mesolophid, connected doubly to mesolophid. Two roots: 1 anterior, 1 posterior. Whether an accessory labial rootlet is present cannot be determined, because the tooth is still in a partial mandible.

m3 (n = 2).—Protoconid small, fused with metaconid. Metaconid large. Posteroflexid open. Roots as in type series (observed in 2 specimens).

Barcadera-Karpata (total n = 53)

M1 (n = 7).—Metacone about equal in size to hypocone. Roots as in type series (observed in 1 specimen).

M2 (n = 13).—Paracone small, connected to anteroloph. Anteroflexus usually closed. Three or 4 roots: 1 anterolabial, 1 posterolabial, and either 1 large lingual root or 2 fused anterolingual and posterolingual roots (observed in 9 specimens).

M3 (n = 6).—Paracone small. No specimens with roots are preserved.

m1 (n = 11).—Metaconid sometimes connected to mesolophid at the labial cingulum. Lingual cingulum sometimes complete, but all flexi sometimes open. Roots as in type series, but accessory lingual rootlet absent (observed in 5 specimens).

m2 (n = 6).—Labial cingulum complete. Roots as in type series (observed in 2 specimens).

m3 (n = 10).—Metaconid large. Roots as in type series (observed in 3 specimens).

Porto Spanjo (total n = 142)

Ml (n = 21).—As in type series (roots observed in 8 specimens).

M2 (n = 21).—Posteroloph usually absent. Labial cingulum absent. Roots as in type series (observed in 8 specimens).

M3 (n = 21).—As in type series (roots observed in 9 specimens).

m1 (n = 27).—Anteromedian flexid present, although often very short. Lingual cingulum usually complete, sometimes incomplete, entoflexid in particular often open. Roots as in type series, but accessory lingual rootlet absent (observed in 8 specimens).

m2 (n = 25).—Mesolophid fused with entoconid. Labial cingulum usually complete, but entoflexid often open. Roots as in type series (observed in 9 specimens).

m3 (n = 27).—1 anterior root, 1 posterior (observed in 7 specimens).

Description of Maxilla and Mandible

The hypodigm of A. donovani includes several mandibular and maxillary fragments from Porto Spanjo and Fontein. The incisive foramen terminates at the front of the alveoli of M1 in 2 specimens from Porto Spanjo and 3 from Fontein. The posterior margin of the zygomatic plate is situated approximately even with or slighty anterior to the front of M1 in 2 specimens from Porto Spanjo and 2 from Fontein. The Porto Spanjo collection includes 9 reasonably complete dentaries of A. donovani. The mental foramen opens laterally. The capsular process is well developed, reaching above the sigmoid notch, but this is preserved in 4 dentaries only. The anterior portions of the masseteric ridges are conjoined as a single crest, which reaches to about the front of the anterior m1 alveolus. A 10th dentary appears to represent a different species and is discussed below as an indeterminate member of Sigmodontinae.

Agathaeromys praeuniversitatis, new species

Oryzomys sp.: De Buisonjé (1974).

Holotype.— RGM 444401, an isolated right M1 (Fig. 4a).

Fig. 4

Type series of Agathaeromys praeuniversitatis from Seroe Grandi, West Bonaire, a) Right M1 (RGM 444401, inverse; holotype of A. praeuniversitatis); b) left M2 (RGM 444418); c) right M3 (RGM 444427, inverse); d) left m1 (RGM 444413); e) left m2 (RGM 444496); f) right m3 (RGM 444434, inverse).

Type locality.—Seroe Grandi, West Bonaire, Netherlands Antilles.

Paratypes.—Total 34 molars, all from the type locality: 9 Ml (RGM 444400, 444402–444409); 3 M2 (RGM 444417– 444419); 4 M3 (RGM 444427–444430); 7 m1 (RGM 444410– 444416); 7 m2 (RGM 444420–444426); 4 m3 (RGM 444431– 444434); 4 edentulous dentaries (unnumbered).

Distribution and age.—Known only from the type locality; age probably about 540–230 kya.

Etymology.—Named after Leiden University's Pre-University College program, which enabled the 1st author (JSZ) to participate in the research project leading to the identification of Agathaeromys and of this species.

Measurements.—Holotype (M1): length 1.81 mm, width 1.22 mm. See Table 4 and Fig. 3 for further measurements.

Diagnosis.—See generic diagnosis (above) for distinction from other genera and diagnosis of A. donovani (above) for distinction from that species.

Description of Type Series

M1 (n = 10).—Anteromedian flexus usually present. Anterolingual conule usually smaller, about half the size of the anterolabial conule, but about equal in size in some specimens. Anteroloph present, oriented perpendicular to anterior mure. Additional connection usually present between anterolabial conule and anteroloph, creating an anteromedian fossette separate from the anteroflexus. Protocone connected to anterior and median mure. Paracone large, connected to middle of protocone. Additional connection to mesoloph present. Mesoloph long, reaching labial margin. Protostyle and enterostyle absent. Hypoflexus broad. Metacone much smaller than hypocone. Posteroflexus present, but shallow, disappearing with little wear, and closed by the connection between metacone and posteroloph. Metacone sometimes connected to posteroloph instead of hypocone. Labial cingulum absent. Flexi meet at midline. Two anterior roots, 1 posterior, 1 very small accessory medial labial root (roots observed in 6 specimens).

M2 (n = 3).—Procingulum largely absent, but anteroloph present and large. Paracone small, connected to mesoloph and middle of protocone. Protocone large. Protoflexus present. Mesocone present, connected to protocone, mesoloph, and hypocone. Mesoloph large, reaching labial margin. Hypocone large. Metacone smaller, connected to hypocone. Posteroloph present. Labial cingulum absent. Mesoflexus, hypoflexus, and paraflexus broad. Flexi meet at midline. No specimens with roots are preserved.

M3 (n = 4).—Variable. Triangular shape. Procingulum and anterolabial and anterolingual cingula absent. Paracone large, with accessory loph posterior to it enclosing a large internal fossette. Protolophule sometimes present. A sinus or, in 1 case, an internal fossette is present between paracone and protocone. Protocone large, fused with hypocone. Hypocone sometimes containing a large internal fossette. Hypoflexus small or absent. A large fossette is present in the middle of the molar. Mesoloph small but complete, reaching the labial cingulum and connected to paracone and hypocone. Metacone present, connected to hypocone and labial cingulum, sometimes fused to hypocone. Posteroloph present. Labial cingulum usually complete, closing the posteroflexus, metaflexus, and mesoflexus. No specimens with roots are preserved.

m1 (n = 7).—Cusps arranged alternately. Anteromedian flexid absent, but superficial division between anterolabial and anterolingual conulids sometimes present. Protostylid sometimes present. Anterolabial cingulum long, but not wholly closing protoflexid. Protoconid large, connected to lingual side of procingulum via anterior murid. Metaconid small, connected to anterolingual conulid and to anterior murid-protoconid angle. Mesolophid narrow or fused with entoconid. Ectolophid usually absent, present in 1 specimen. Entoconid narrow but long, connected to mesoconid and hypoconid. Posterolophid present and large. Posteroflexid broad. One anterior root, 1 posterior. Accessory labial rootlet present in 1 of 4 specimens.

m2 (n = 7).—Procingulum largely absent. Metaconid and protoconid large. Anterolophid absent. Anterolabial cingulum present before protoconid. Protoconid connected to mesoconid. Mesolophid narrow but long, reaching labial margin. Mesoflexid sometimes closed by a connection between mesolophid and metaconid at the lingual cingulum. Entoflexid open. Entoconid broad, connected to mesoconid-hypoconid connection, sometimes directly to hypoconid, sometimes also to mesolophid. Hypoconid large, connected to mesoconid. Mesolophid, entoconid, and hypoconid fused in 1 case, probably as a result of wear. Posterolophid present. Posteroflexid open. Labial cingulum absent. One anterior root, 1 posterior (roots observed in 4 specimens).

m3 (n = 4).—Procingulum and anterolabial and anterolingual cingula largely absent. Protoconid small, connected with metaconid. Anterolophid absent. In 1 case, an internal fossettid is present between protoconid and metaconid. Metaconid large. Mesoflexid and hypoflexid broad. Entoconid small, located anterolabially of hypoconid, connected to metaconid and hypoconid. Mesolophid absent. Hypoconid large. Posterolophid present, broad. Lingual cingulum incomplete, closing posteroflexid and part of mesoflexid. One anterior root, 1 posterior (roots observed in 1 specimen).

Dentaries (n = 4).—The mental foramen opens laterally in 2 dentaries and opens dorsally, at the diastema, in 1; the condition could not be assessed in the 4th dentary. The capsular process is well developed, as in A. donovani. The anterior portions of the masseteric ridges are conjoined as a single crest, which reaches to about the front of the anterior m1 alveolus.

Sigmodontinae, gen. et sp. indet.

One edentulous dentary (Fig. 5; unnumbered, collection RGM) from Porto Spanjo appears to represent a genus and species distinct from Agathaeromys donovani (Fig. 5). This mandible is more slender than that of A. donovani. The diastema appears shorter, and the anterior tip of the incisor is only slightly lingual to the axis of the toothrow; in A. donovani it is placed much more lingually. The mental foramen opens dorsally, in the diastema, not laterally as in A. donovani. The posterior part of the dentary is mostly missing. The posterior part of the incisor is visible slightly posterolabially to the m2; it appears that m3 is shifted lingually relative to m1 and m2 in association with the position of the incisor, but the alveoli for m3 are almost completely missing. In contrast, in A. donovani the posterior tip of the incisor emerges posterolabially to the m3, at the capsular process, and m3 is not shifted relative to m1 and m2. Judging by the well-preserved alveoli, the m1 has a single large anterior root, 1 lingual rootlet, 2 labial rootlets, and a large posterior root. The m2 has anterolabial and anterolingual roots, an accessory anterolabial rootlet separated from the main anterolabial root by a thin wall, a large posterior root, and a smaller labial root anterolabial to the main posterior root. Because of the last trait, the axis of the posterior root is not perpendicular to the axis of the toothrow, as in A. donovani, but at an angle of about 60°. Nothing can be said about the alveoli of m3. We measured the alveolar length of m1 (from front of anterior alveolus to back of posterior alveolus) and the width of the posterior alveolus of m1 for this specimen and the associated edentulous dentaries of A. donovani from Porto Spanjo. Alveolar m1 length for this specimen is 2.78 mm and width of posterior alveolus of m1 is 1.22 mm. For the 9 dentaries of A. donovani, alveolar m1 length is 2.49–3.08 mm (average 2.70 mm; SD = 0.20 mm) and posterior alveolar m1 width is 1.25–1.48 mm (average 1.35 mm; SD = 0.09 mm). Thus, the enigmatic dentary has alveolar m1 length slightly above average, but posterior alveolar m1 width slightly smaller than the smallest A. donovani, but not significantly outside the variation of the A. donovani population (Z-test: P — 0.074).

Fig. 5

Edentulous dentaries from Porto Spanjo (collection RGM, unnumbered; not to scale), a) Right dentary of Agathaeromys donovani; b) right dentary of Sigmodontinae, gen. et sp. indet.

We identify this dentary as a member of Sigmodontinae because premolars seem to be absent. It does not represent a species of Agathaeromys because of the many differences enumerated above, and we are unable to provide a positive identification of this enigmatic dentary. Its identity may only be determined through the discovery of additional material.

Phylogenetic Analysis

Traits scored in both species of Agathaeromys for cladistic analysis are shown in Tables 2 and 3. The results of our phylogenetic analysis of the Sigmodontinae based on Pacheco (2003) are shown in Fig. 6. The resultant tree is broadly similar to that of Pacheco (2003), particularly in the deeper nodes, but the arrangement of the thomasomyine-oryzomyine clade is markedly different. However, the Oryzomyini and a core thomasomyine group are recovered as monophyletic in both trees. In our analysis Agathaeromys appears as sister to the derived oryzomyine Holochilus; this clade is successively joined by the other oryzomyines included.

Fig. 6

One of 3 most-parsimonious trees (length = 728) resulting from a reanalysis of morphological data in Pacheco (2003) with the addition of 2 new species of Agathaeromys. The other most-parsimonious trees differ in the arrangement of the clade containing Aepeomys lugens and 4 species of Thomasomys; 1 also switches Oecorny s and Handleyomys. Polymorphic entries were treated as transformation series.

View this table:
Table 2

Dental character states (Pacheco 2003) of 2 new species used for phylogenetic analysis of relationships within Sigmodontinae.

View this table:
Table 3

Applicable character states (Weksler 2006) of 2 new species used for phylogenetic analysis of relationships within Sigmodontinae. Character state A is polymorphic for 0 and 1, and C is polymorphic for 1 and 2.

To determine more fully the relations of Agathaeromys within the Oryzomyini we conducted another phylogenetic analysis based on data from Weksler (2006), which resulted in 2 equally parsimonious trees (Fig. 7). The analysis of Weksler (2006) has produced a division of the tribe Oryzomyini into 4 major clades, namely clade A (Scolomys and Zygodontomys), clade B (Handleyomys, Oecomys, Euryoryzomys, Hylaeamys, Nephelomys, Transandinomys, and Mindomys), clade C (Oligoryzomys, Neacomys, Microryzomys, and Oreoryzomys), and clade D (Oryzomys, Eremoryzomys, Cerradomys, Sooretamys, Aegialomys, Nesoryzomys, Holochilus, Lundomys, Pseudoryzomys, Melanomys, Nectomys, Amphinectomys, and Sigmodontomys—nomenclature follows Weksler et al. [2006]). He used the names “clade B*” and “clade D*” for slightly different clades that resulted from the unstable placement of the taxa Amphinectomys and Mindomys. The resultant trees are similar in many respects to the equivalent tree in Weksler (2006) but differ in several details: clade B* is paraphyletic with respect to all other oryzomyines in our analysis; the species of Oligoryzomys and Oecomys are arranged differently in all 3 trees; Amphinectomys appears as sister group of the Nectomys-Melanomys-Sigmodontomys clade within clade D or as sister to Nectomys alone in our trees instead of being sister to Handleyomys intectus within clade B*; and the arrangement of several taxa in clade D is markedly different. Most conflicting nodes, however, are poorly supported in Weksler's (2006) tree.

Fig. 7

Two most-parsimonious trees (length = 552) resulting from a reanalysis of the phylogenetic analysis of morphological data by Weksler (2006), with 2 species of Agathaeromys added. Polymorphic entries were treated as transformation series.

In both most-parsimonious trees the 2 species of Agathaeromys form a clade and appear within clade D of Oryzomyini. In addition, both trees support 2 main groups within clade D, namely a Holochilus group also including Lundomys, Pseudoryzomys, and Oryzomys and a Nectomys group also including Amphinectomys, Melanomys, Sigmodontomys, Aegialomys, and Nesoryzomys. These 2 groups, together with the basal genera Agathaeromys, Eremoryzomys, Cerradomys, and Sooretamys, form clade D in both trees. In 1 tree, however, Agathaeromys is recovered as sister to the Holochilus group, but in the other it is sister to a clade of Eremoryzomys and the Nectomys group. As discussed above, the topology of other parts of the tree contains some disparities with Weksler (2006), but a detailed discussion of oryzomyines outside clade D is beyond the scope of this paper.

Discussion

Taxonomy.—The Bonaire material is morphologically distinct from all other oryzomyine genera (Table 3) and forms a separate clade from other genera (Fig. 7). Accordingly, its placement as a separate genus, here named Agathaeromys, is justified. The molars from 1 locality, Seroe Grandi, representing A. praeuniversitatis, differ strikingly from all other material (A. donovani) in size (Fig. 3; Table 4), and in addition, we detected several discrete morphological differences between the 2 species (see diagnosis of A. donovani). The character states of A. praeuniversitatis usually also occur as rare variants in specimens of A. donovani. The latter is variable in morphology and size (Fig. 3) among the 4 localities from which it has been recorded, but the morphological variations are mostly differences of frequency or otherwise trivial and the size differences are small. Thus, we are confident that only 1 species is represented within the material we refer to as A. donovani.

View this table:
Table 4

Dental measurements of Agathaeromys donovani and A. praeuniversitatis. For each molar, length (L) and width (W) are included. In addition, P was computed using Student's t-test (HS: highly significant, P < 0.01).

Dental characters and cladistic reconstructions.—We assessed the phylogenetic relationships of Agathaeromys by including both species in previous phylogenetic analyses of sigmodontine rodents. Although only dental characters were available for Agathaeromys, the results provided robust support for its placement in Oryzomyini (Fig. 6) and for its status as a separate genus and allocation to clade D (Fig. 7). Few previous studies have used cladistic analysis to assess the relationships of sigmodontines known only from dental material (Steppan and Pardiñas 1998), and ours appears to be the 1st to use large samples of all molars of the fossil taxa under consideration. Despite the inherent limitations of using only dental material, our results suggest that cladistic analysis is a powerful tool to identify phylogenetic relationships within sigmodontines, even if some taxa are known from limited material. Future cladistic analyses may be able to resolve the relationships of other oryzomyines known only from limited material, such as several other extinct Antillean taxa.

Age.— The exact age of the deposits where Agathaeromys has been found is unknown. However, some inferences can be drawn from the geomorphological context and from similar absolutely dated deposits from Curaçao (De Buisonjé 1974; McFarlane and Lundberg 2002; Stienstra 1983). The material of A. praeuniversitatis from Seroe Grandi dates from the period between the depositions of the Lower and Middle Terraces (specifically, Middle Terrace II—De Buisonjé 1974). In contrast, A. donovani dates from the period between the depositions of the Middle and Higher Terraces (De Buisonjé 1974). Stienstra (1983) provides approximate dates for the depositions of the Terraces in Aruba, Curaçao, and Bonaire derived from correlation with sea-level fluctuations; these dates are supported by the absolute dating of a Middle Terrace II deposit from Curaçao at ∼400 kya by McFarlane and Lundberg (2002). If dates from Stienstra (1983) are accepted, the age of A. donovani is limited to the period between the end of the deposition of the Higher Terrace at ∼900 kya and the deposition of Middle Terrace II at ∼540 kya, and that of A. praeuniversitatis is limited to the period between the deposition of Middle Terrace II at ∼540 kya and the beginning of the deposition of the Lower Terrace at ∼230 kya. This would place the range of both species in the 2nd half of the Pleistocene.

Insular oryzomyines.Agathaeromys is 1 of several oryzomyine genera restricted to offshore islands, the others being Megalomys of the Lesser Antilles, Noronhomys of Fernando de Noronha, and Nesoryzomys of the Galápagos Islands, only the last of which is still extant. The oryzomyine tribe also encompasses many species restricted to such islands, including Oryzomys antillarum, Oryzomys nelsoni, Aegialomys galapagoensis, and Oligoryzomys victus (Musser and Carleton 2005; Weksler et al. 2006). It is noteworthy that the endemic oryzomyines seem to be restricted almost wholly to a single monophyletic part of the oryzomyine radiation, designated clade D by Weksler (2006). This clade is recovered in both morphological and molecular analyses (with slightly varying contents), but it is much more strongly supported by molecular (interphotoreceptor retinoid-binding protein) data (Weksler 2006). Clade D includes the genera Eremoryzomys, Oryzomys, Pseudoryzomys, Lundomys, Holochilus, Sooretamys, Cerradomys, Nesoryzomys, Aegialomys, Amphinectomys, Nectomys, Melanomys, and Sigmodontomys in Weksler's combined-data analyses (Weksler 2006). Our reanalysis of morphological data from Weksler (2006), including both species of Agathaeromys, indicates that this genus is likewise referable to clade D (Fig. 7). Among insular endemics clade D includes Nesoryzomys, Agathaeromys, and the presumed sister groups of Noronhomys, Aegialomys galapagoensis, Oryzomys nelsoni, and Oryzomys antillarum. Weksler et al. (2006) also suggest that Megalomys may be a member of this clade. If the association of all of these taxa with clade D is corroborated by further research, clade D includes all currently described insular endemic oryzomyines but 1 (Table 5). As shown by Weksler (2006), clade D also includes several semiaquatic taxa and taxa occurring in nonforest habitats. In contrast, no members of the 3 other clades are semiaquatic, and nonforest taxa are absent from clade B and restricted to single genera in clades A and C. Apparently, the semiaquatic, nonsylvan members of clade D are more likely to colonize offshore islands than their terrestrial, sylvan relatives.

View this table:
Table 5

Endemic insular oryzomyine species (Musser and Carleton 2005; Voss and Weksler 2009; Weksler et al. 2006; this paper).

Although clade D is consistently recovered, the internal topology of the clade is still far from clear, with several analyses from Weksler (2006) and our tree all showing different hypotheses of interrelationships among the clade's member genera. Two tight clusters of genera exist within clade D, the Holochilus and Nectomys groups; both clades are well supported in the combined-data analyses of Weksler (2006). Beyond these clades, however, relations are generally poorly supported, and topologies differ widely.

To resolve the interrelationships among the members of clade D, further research should focus on improving the taxon and character sampling of Weksler (2006), including more species of extant genera, in addition to such extinct representatives as Noronhomys vespuccii, Carletonomys cailoi, Megalomys spp., and Holochilus primigenus. We believe that a robust phylogenetic framework for clade D will be instrumental in addressing the relationships of the other oryzomyines of the Lesser Antillean region, most of which are still undescribed and known from limited material. Answering phylogenetic questions about the affinities of species such as the members of Megalomys, a genus diagnosed solely by its giant size, could shed light on the historical biogeography and biodiversity of the West Indian islands. Voss and Weksler (2009) recently revised the taxonomic status of the supposed Curaçao endemic Oryzomys curasoae and concluded that the species is undistinguishable from the mainland O. gorgasi. Their revision is 1 example of how such studies can improve our understanding of oryzomyine relationships.

Biogeography.Agathaeromys is currently the only non-volant mammal described from Bonaire, although 7 species of bats are known, all of which also occur in northern Venezuela (Simmons 2005). According to De Buisonjé (1974), the only other mammal known from Bonaire is a rare and unidentified armadillo. In contrast, De Buisonjá reported the large extinct rice rat Megalomys curazensis, the smaller rice rat Oryzomys sp., the field mouse Calomys laucha (= hummelincki), the capybara Hydrochoerus hydrochaeris (= Hydrochoeris isth-mius?), and the extinct ground sloth Paulocnus petrifactus from Curaçao, and Megalomys curazensis, Oryzomys sp., Calomys laucha, 2 other sigmodontines (Sigmodon cf. hispidus and Zygodontomys cf. brevidauda), and the opossum Didelphis marsupialis from Aruba. Since then, another extinct oryzomyine, Oryzomys curasoae, has been described from Curaçao (McFarlane and Debrot 2001), which later was realized to be identical to Oryzomys gorgasi from mainland Colombia and Venezuela (Voss and Weksler 2009), but the relationship of that taxon to the other oryzomyines from Curaçao remains to be determined. Biogeographically, Bonaire (area of 280 km2, distance from mainland of 80 km), a relatively small island far from the mainland, would be expected to have relatively low immigration and high extinction rates, resulting in relatively low diversity, but Aruba's (area of 180 km2, distance from mainland of 30 km) small distance to the mainland and Curaçao's (area of 450 km2, distance from mainland of 70 km) large size would be expected to result in higher diversity through higher immigration and lower extinction rates than on Bonaire. However, the taxonomy of most species from all 3 islands is still far from definitive, and further discoveries could result in a different pattern of diversity.

As discussed above, Agathaeromys appears in the fossil record during the 2nd half of the Pleistocene, and it is related most closely to a clade of extant oryzomyines occurring across South America. Because Bonaire is separated by a deep basin (1,700 m) from the mainland and was presumably never connected to it (De Buisonjé 1974), it seems likely that the ancestor of Agathaeromys migrated from the Venezuelan mainland at some time during the Pleistocene. Because of the scant fossil record of the Oryzomyini, this hypothetical ancestor cannot be connected to any fossil species from the mainland.

Acknowledgments

We thank M. Brugman, Pre-University College, Leiden University, for valuable assistance in setting up this project; M. Bosch and A. Gill for valuable comments; V. Pacheco for permission to use his unpublished data for the phylogenetic analysis in Fig. 6; and 2 anonymous reviewers for helpful comments on the manuscript.

Footnotes

  • Associate Editor was Burton K. Lim.

Literature Cited

View Abstract