Marmosa (Micoureus) perplexa Anthony, 1922

Voss, Robert S. & Giarla, Thomas C., 2021, A Revision of the Didelphid Marsupial Genus Marmosa Part 3. A New Species from Western Amazonia, with Redescriptions of M. perplexa Anthony, 1922, and M. germana Thomas, 1904, American Museum Novitates 2021 (3969), pp. 1-28 : 12-17

Marmosa (Micoureus) perplexa Anthony, 1922

Marmosa perplexa Anthony, 1922: 3 (original description).

Marmosa phaea: Tate, 1933: 84 (part, perplexa treated as synonym).

Marmosa (Marmosa) phaea: Cabrera, 1958: 23 (part, perplexa treated as synonym).

Micoureus regina: Gardner, 1993: 20 (part, perplexa treated as synonym).

Micoureus phaeus: Gardner, 2005: 13 (part, perplexa treated as synonym).

Marmosa (Micoureus) phaea: Voss and Jansa, 2009: 101 (part, perplexa treated as synonym).

TYPE MATERIAL AND TYPE LOCALITY: The holotype (by original designation, AMNH 47188 ) consists of the skin and skull of a young adult female collected on the trail from Zaruma to Loja near Punta Santa Ana, 3650 feet (1113 m) above sea level, in the western Andean foothills of Loja province, Ecuador, by H.E. Anthony on 21 December 1920 .

DISTRIBUTION AND SYMPATRY: Examined specimens of Marmosa perplexa are all from lowland, foothill, or lower-montane localities in southwestern Ecuador (Loja, El Oro) and northwestern Peru (Cajamarca, Tumbes). This species is not known to occur sympatrically with other congeners .

DESCRIPTION: The dorsal pelage of Marmosa perplexa is (as usual for the subgenus) dull grayish brown ( fig. 3 ); among Ridgway’s (1912) color swatches, the closest match seems to be Hair Brown, with remarkably little variation in side-by-side comparisons among five skins collected between 1920 and 2006 .5 Measured at midback, the dorsal fur varies from 11 to 13 mm long in the same series. The ventral pelage is almost entirely gray-based, with self-colored (pale buff) fur only on the chin and groin. The hind feet of some specimens (e.g., AMNH 61382 ) seem to be almost entirely covered with pale fur, but in others (e.g., USNM 513425 ) some of the metatarsal fur is dark; the metacarpal fur of the forefeet seems to be consistently brownish, but the manual phalanges are pale. The tail averages about 148% of head-and-body length, but with such extreme individual variation in this proportion (from 135% to 168%) as to cast doubt on the accuracy of some collectors’ measurements; excluding the highest and lowest computed fractions, however, the average is about the same (145%). Only 20 mm or less of the tail base is covered with short (ca. 5–6 mm) fur. The naked (scaly) caudal skin is completely dark from base to tip in some specimens (e.g., AMNH 61382 , USNM 513425 ), but the tail of the holotype has pale mottling near the tip, and the tail of one specimen ( UMMZ 176563 ) is almost half white .

In dorsal view, adult skulls of Marmosa perplexa ( fig. 4B ) are chiefly remarkable for their slender rostrums and for the absence of well-developed postorbital processes. In ventral view ( fig. 4E ), these skulls have long (4.8 ± 0.5 mm) maxillopalatine fenestrae that sometimes extend

5 But note that Anthony (1922) matched the dorsal pelage of the type with Saccardo’s Umber, a decidedly more saturated color.

from a point opposite P3 to a point opposite the M3 protocone on each side; one specimen ( MUSM 24484 ), however, has notably shorter maxillopalatine fenestrae. Another distinctive feature is the consistent presence of palatine fenestrae, either as one large hole or as multiple smaller perforations on each side. Lastly, several specimens (e.g., AMNH 47199 , MUSM 24484 , UMMZ 176563 ) have well-developed anteromedial bullar processes resembling those described above for M. jansae , and in one specimen ( MUSM 24484 ) the anteromedial process forms a unilaterally complete secondary foramen ovale. The second upper premolar (P2) has an incomplete lingual cingulum, and the molar toothrow is among the shortest known in the subgenus (LM = 7.7–8.2 mm). The postprotocrista of M3 is short (terminating at or near the base of the metacone), and m1–m3 lack posterior cingulids .

COMPARISONS: Tate (1933) synonymized Marmosa perplexa with M. phaea , and it is hard to fault his decision, because these taxa are very similar morphologically. However, M. perplexa and M. phaea are not closely related ( fig. 1 ), and they have highly divergent cytochrome b sequences (about 11% on average; Voss et al., 2020: table 4). Side-by-side comparisons of M. perplexa with typical examples of M. phaea (from the Colombian departments of Cauca, Huila, and Nariño) suggest that these species are externally indistinguishable, but skulls of M. phaea are slightly larger than those of M. perplexa , have better-developed postorbital processes, a complete lingual cingulum on P2, and long postprotocristae.

Morphological comparisons between M. perplexa and its allopatric sister species M. jansae were summarized in the preceding account.

REMARKS: In his original description of Marmosa perplexa, Anthony (1922: 3) emphasized the “peculiarly depressed frontal region” and the absence of postorbital processes in the type, a young adult, but Tate (1933) noted the absence of any frontal depression in a subsequently collected fully adult specimen (AMNH 61382) and suggested that the type skull was deformed. None of the other specimens that we assign to this species has a frontal depression resembling that of the type, but the absence of well-developed postorbital processes seems to be a valid diagnostic feature. Nevertheless, none of the material at hand is very old, and it is possible that specimens more mature than any we have yet seen might have better-developed processes.

Cytochrome b sequences that Voss et al. (2020) obtained from USNM 513425 (475 bp; GenBank accession MN 978645 ) and UMMZ 176563 (1149 bp; MN 978646 ) differ by about 5% (uncorrected), a noteworthy value, but maximum-likelihood analyses recovered these sequences as a strongly supported clade (with 100% bootstrap support; Voss et al., 2020: fig. 5 ) and we regard them as conspecific. Nevertheless , UMMZ 176563 is our only montane example of Marmosa perplexa , it is the only specimen we have seen with a half-white tail, and it is slightly larger than the other specimens we measured ( table 3). A less conservative interpretation of such observations might conclude that two species are represented, but it seems prudent to await additional material from the mountains of northwestern Peru to establish whether these are consistent differences between lowland and highland populations that merit taxonomic recognition .

HABITATS: The vegetation of southwestern Ecuador and northwestern Peru is a complex mosaic of humid (evergreen or semi-evergreen) forest, dry (deciduous) forest, and arid scrub. In general, the coastal lowlands are subject to the desiccating effect of the cold Humboldt Current, but a narrow strip of humid forest occurs along the foothills of the Andes, and condensation from nocturnal fogs can sustain patches of humid vegetation in otherwise arid landscapes ( Chapman, 1926). Site descriptions are therefore crucial for understanding the habitat affinities of species in this region.

According to Chapman (1926: 716), the AMNH expeditionary camp at Punta Santa Ana— the type locality of Marmosa perplexa —was in “a heavy stand of fine forest” in a well-watered headwater valley of the Río Tumbes , but “the surrounding district is arid and desolate.” A typed itinerary of Anthony’s 1921 fieldwork in Ecuador provides more information. 6

Punta Santa Ana is a high divide between two tributaries of the Rio Tumbez, and our camp was at an elevation of 3650 feet in a splendid body of large timber. This forest was extremely local, not more than three or four miles wide at the southern end and extending as a long tongue down a valley ... Probably the greater part of this region is in the humid tropical zone, reaching [upslope] into the subtropics, but immediately to the west, south, and east lie great areas of the arid tropical [zone] so that, in a sense, it is an oasis of life in a great barren section.

Anthony’s handwritten fieldnotes describe the local vegetation as including large trees (among which “figs,” presumably Ficus spp. , were very common), small tree ferns, and “cane” (probably dwarf bamboo, Chusquea sp. [ Poaceae ]). Mammals collected at this locality (in addition to M. perplexa ) included such humid-forest taxa as Alouatta palliata , Cebus albifrons , Potos flavus , Melanomys caliginosus , and Transandinomys talamancae .

Only the most cursory habitat descriptions are available for other sites at which Marmosa perplexa has been collected in Ecuador, but G.H.H. Tate’s fieldnotes for 1921 indicate that humid tropical forest (or remnants of such vegetation) were present at Piñas (in El Oro province) and Guainche (in Loja). However, the habitat near Portovelo (in El Oro) may have been less humid, because dry-forest rodent species such as Aegialomys xanthaeolus (e.g., USNM 513559, 513560) were collected in sympatry with M. perplexa at this locality.

The vegetation surrounding Quebrada Faical in the Parque Nacional Cerros de Amotepe of northwestern Peru has been described as “dry deciduous forest” ( Wiedenfeld et al., 1985: 305), “semi-deciduous forest” ( Parker et al., 1995: 202), or transitional between “Equatorial Dry Forest” and “Pacific Tropical Rainforest” ( Pacheco et al., 2007). According to Wiedenfeld et al. (1985), most of the trees had lost or were losing their leaves when researchers visited this site early in the dry season, but a photograph of the local habitat (presumably taken at the same time; Wiedenfeld et al., 1985: fig. 2 ) shows many still-leafy trees. The forest canopy is said to be about 20 m high, with such characteristic tree species as Triplaris cumingiana (Polygonaceae) , Cavanillesia platanifolia (Malvaceae) , Ficus jacobii (Moraceae) , Ceiba trichastandra (Malvaceae) , Muntingia calabura (Muntingiaceae) , Tessaria integrifolia (Asteraceae) , and Mimosa pellita (Fabaceae) ( Pacheco et al., 2007). Palms and giant herbs are conspicuously absent from photographs of the forest interior

6 These and other unpublished sources cited in the next paragraph are preserved in the AMNH Department of Mammalogy archives.

(e.g., fig. 7 ), additional indications of highly seasonal rainfall at this locality. Two specimens of Marmosa perplexa were captured in the kitchen of the biological station at El Faical, but two others were captured in the forest: one on the ground and another about 1 m above the ground on a tree branch; Transandinomys talamancae and Proechimys decumanus were taken in the same traplines that captured M. perplexa at this locality (V. Pacheco, personal commun.).

In summary, Marmosa perplexa appears to occur in a range of forest types in southwestern Ecuador and northwestern Peru. Based on the presence of tree ferns at Punta Santa Ana, the vegetation at the type locality was probably evergreen forest, and descriptions of similar sites in the immediate area (e.g., Robbins and Ridgley, 1990; Torres-Porras et al., 2017) suggest that this humid premontane vegetation is sustained by daily moisture from enveloping clouds. In the adjacent lowlands, M. perplexa seems to occur at drier sites where the forest is at least partially leafless in the dry season. Whether evergreen or semideciduous, however, the forests of southwestern Ecuador and northwestern Peru are among the most endangered habitats in the western hemisphere, and their endemic biotas are a focus of urgent conservation concern ( Parker et al., 1995; Linares-Palomino et al., 2010; Ferrer-Paris et al., 2018; Rivas et al., 2020).

SPECIMENS EXAMINED (N = 10): Ecuador — El Oro, Piñas (AMNH 61382, 61391); Loja, Guainche (AMNH 61390), 12 km E by road Portovelo (USNM 513425), Punta Santa Ana (AMNH 47188). Peru — Cajamarca, Cerro La Viuda (UMMZ 176563); Tumbes, Quebrada Faical (MUSM 24483–24486).

Figures

FIG. 3. Dorsal and ventral views of skins of Marmosa jansae (ROM 118880), M. perplexa (USNM 513425), and M. germana (FMNH 43179).

FIG. 4. Dorsal and ventral cranial views (×1.75) of Marmosa jansae (A, D; ROM 118880), M. perplexa (B, E; USNM 513425), and M. germana (C, F; TTU 101236).

FIG. 1. Maximum-likelihood (ML; left) and Bayesian (right) phylogenetic trees for putative species of Marmosa (Micoureus) inferred from concatenated sequences of six genes (after Voss et al., 2020: fig. 6). Numbers at nodes denote ML bootstrap support or Bayesian posterior probabilities. Outgroups are not shown. Dotted lines connect the same taxa across the two trees to highlight differences. Names of species treated in this report are in boldface.

FIG. 5. Rainforest (monte alto) habitat of Marmosa jansae at the Reserva Nacional Allpahuayo-Mishana, Loreto, Peru (photo by Christine L. Hice).

FIG. 2. Collection localities of examined specimens of Marmosa germana, M. jansae, and M. perplexa. Numbers are keyed to entries in our gazetteer (appendix 1).

FIG. 7. Semideciduous forest understory near Quebrada El Faical, Tumbes department, Peru, habitat of Marmosa perplexa (photo by Elizabeth Escobar).