Novtautesamerican MUSEUM PUBLISHED by the AMERICAN MUSEUM of NATURAL HISTORY CENTRAL PARK WEST at 79TH STREET, NEW YORK, N.Y

NovtautesAMERICAN MUSEUM PUBLISHED BY THE AMERICAN MUSEUM OF NATURAL HISTORY CENTRAL PARK WEST AT 79TH STREET, NEW YORK, N.Y. 10024 Number 3177, 34 pp., 12 figures August 23, 1996

Madagascan Poison Frogs (Mantella) and Their Skin Alkaloids

JOHN W. DALY,1 NIRINA RABE ANDRIAMAHARAVO,2 MARTA ANDRIANTSIFERANA,2 AND CHARLES W. MYERS3

CONTENTS Abstract ...... 2 Introduction ...... 2 Subfamily Mantellinae Laurent, 1946 ...... 3 Genus Mantella Boulenger, 1882 ...... 7 Mantella aurantiaca Mocquard, 1900 ...... 13 Mantella baroni Boulenger, 1888 ...... 17 Mantella betsileo (Grandidier, 1872) ...... 19 Mantella cowanii Boulenger, 1882 ...... 20 Mantella crocea Pintak and B6hme, 1990 ...... 21 Mantella expectata Busse and Bohme, 1992 ...... 21 Mantella laevigata Methuen and Hewitt, 1913 ...... 22 Mantella pulchra Parker, 1925 ...... 22 Mantella viridis Pintak and B6hme, 1988 ...... 24 Summary of Alkaloids ...... 24 Acknowledgments ...... 28 Appendix 1: Voucher Specimens ...... 28 Appendix 2: Alkaloids Identified in Mantella Skin ...... 30 References ...... 31

' Research Associate in Herpetology, American Museum of Natural History; Chief, Laboratory of Bioorganic Chemistry, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health. 2 Laboratoire de Chimie Organique "Product Naturals," Universite d'Antananarivo, Antananarivo, 1001, Mad- agascar. 3 Chairman and Curator, Department of Herpetology and Ichthyology, American Museum of Natural History.

ABSTRACT Mantella is demonstrably convergent with apo- Although the monophyly of Mantella is sup- sematic Neotropical poison frogs of the family ported by presence of lipophilic skin alkaloids, Dendrobatidae in presence of a variety of similar these compounds may be sequestered from ar- skin alkaloids, and variously convergent in a few thropod prey and, if so, are unlikely to be phy- other traits. Mantella cowanii is the best example logenetically informative within the genus. Inter- of color pattern convergence to a Dendrobates. specific and interpopulational differences in al- Inasmuch as resemblances are primarily to de- kaloid profiles may be a reflection of dietary dif- rived features of phylogenetically advanced den- ferences correlated with habitat diversity. Mantella drobatids, the old notion ofrelationship, although habitats include upland swamp forest (M. auran- intriguing, is not supported by current understand- tiaca, M. crocea), boggy areas within forest (M. ing. But the phylogenetic placement of Mantella pulchra), upland riverine forest (M. baroni), both is nonetheless unclear and even its relationship to disturbed and relatively undisturbed lowland for- other "mantellines" requires corroboration. est (M. betsileo), and semiarid streambeds (M. ex- Alkaloid profiles in skin extracts of 9 of 11 rec- pectata, M. viridis). Although most species are ter- ognized species of Mantella are reported. Levels restrial, M. laevigata is semiarboreal in forest and of alkaloids ranged from relatively low (M. bet- in bamboo groves. sileo, M. expectata, M. laevigata) to moderate (M. Excluding published conjectures based on frogs aurantiaca, M. crocea, M. pulchra) to high (M. without provenance (from commercial dealers), baroni (madagascariensis, auctorum), M. cowanii, assumptions of extraordinary intraspecific or in- M. viridis). However, there was variation in levels trapopulational variability in Mantella color pat- for different populations of the same species. The terns have proven unfounded (unlike the situation major alkaloids in most species and populations in Dendrobates), and most species are readily iden- were of the pumiliotoxin class, consisting mainly tified in life by coloration alone. The name Man- of pumiliotoxins and allopumiliotoxins, although tella baroni Boulenger, 1888, is resurrected from homopumiliotoxins and other pumiliotoxin-class synonymy for a widespread upland species cur- alkaloids also occurred. A decahydroquinoline and rently known as M. madagascariensis (Grandidier, related "dimeric" alkaloids were prominent in cer- 1872)-a nomen dubium not certainly applicable tain populations of M. betsileo, M. pulchra, and to any known species (but conceivably based on M. laevigata, but were absent in other populations M. pulchra Parker, 1925). Use of the name M. and in other species. "Izidine" alkaloids, namely betsileo (Grandidier, 1872) for a widespread low- disubstituted pyrrolizidines, indolizidines, and land species is accepted with reservation because quinolizidines, were prominent in all populations the type locality falls outside the known distri- ofM. baroni, but were usually relatively minor or bution and habitat. absent in other species.

INTRODUCTION All amphibians have cutaneous granular mary in Myers et al., 1995: 17), with over ("poison") glands, that in most species se- 200 lipophilic "dendrobatid alkaloids" hav- crete diverse pharmacologically active com- ing now been reported (Daly et al., 1987,1993: pounds-including lipophilic alkaloids, which 276). have a peculiar taxonomic distribution in Prior to 1984, most lipophilic frog alka- amphibian skin (Daly et al., 1978, 1987, 1993: loids were known in nature only from den- 276). drobatids (e.g., Daly et al., 1978) until the The first amphibian lipophilic alkaloids to discovery of some in the South American be chemically elucidated, the samandarines bufonid genus Melanophryniscus (Daly et al., from the European fire salamander (Sch6pf, 1984; Garraffo et al., 1993a), in Australian 1961), remain known only from the Old myobatrachids of the genus Pseudophryne World genus Salamandra (for comment on (Daly et al., 1984, 1990), and in presumptive a contrary report, see Daly et al., 1993: 198). ranids, rhacophorids, or mantellids of the Different lipophilic alkaloids were subse- Madagascan genus Mantella (Daly et al., quently found to characterize a monophyletic 1984; Garraffo et al., 1993b). Lipophilic al- group ofseveral genera ofpoison frogs within kaloids have not been detected in skin ex- the Neotropical family Dendrobatidae (sum- tracts from some 70 other genera ofamphib- 1 996 DALY ET AL.: MADAGASCAN POISON FROGS 3 ians. A few structurally distinct classes of al- Sequestering of lipophilic alkaloids pro- kaloids-batrachotoxins, epibatidines, ge- vided in the frogs' diet has recently been dem- phyrotoxins, and histrionicotoxins-are still onstrated for Mantella, which resembles den- not known to occur naturally in amphibians drobatids in that frogs reared in captivity on other than dendrobatids (but see comment Collembola, Drosophila, and Acheta do not under Mantella pulchra on histrionicotoxins contain detectable alkaloids (Daly et al., in in a pet-trade specimen). Homobatrachotox- manuscript). Although it is not known in, however, was recently identified in skin whether the other non-dendrobatid genera and feathers of a toxic bird from Papua New (Melanophryniscus and Pseudophryne) with Guinea (Dumbacher et al., 1992). skin alkaloids have similar sequestering sys- Although they represent four families, anu- tems, it is now conceivable that many, ifnot, rans with lipophilic skin alkaloids share a all frog alkaloids are "second-hand" chemi- number of attributes. All are small (1-5 cm cals that are not manufactured by the frogs SVL), primarily terrestrial frogs that lay ter- themselves. This helps explain some, albeit restrial eggs. The great majority are brilliantly not all, otherwise puzzling instances ofinter- colored (and often boldly patterned) and all and intraspecific variation in skin toxins, and except the Australian Pseudophryne are di- gives a clearer perspective on the taxonomic urnal. Many are toothless. usefulness of toxins. Myers et al. (1995), in Lipophilic skin alkaloids appear to func- a preliminary assessment of the systematic tion defensively and the ability to synthesize implications, pointed out that the underlying and/or uptake and sequester them seems to genetic mechanisms for alkaloid sequestering have evolved at least five times in the Am- provide useful synapomorphies, but that dif- phibia. The samandarine alkaloids are cur- ferences in alkaloid profiles ofrelated species rently thought to be synthesized by the sal- may for the most part reflect dietary differ- amanders (comment in Daly et al., 1993: 199), ences. and the possibility that dendrobatid frogs also Profiles of lipophilic skin alkaloids differ synthesize some of their diverse alkaloids remarkably in different species and popula- cannot be disproved at this time. tions of frogs containing them (Daly et al., Nonetheless, evidence is accumulating that 1984, 1987, 1990; Garraffo et al., 1993a, many "dendrobatid alkaloids" are seques- 1993b; Myers et al., 1995). If sequestering tered from some of the small arthropods on systems and diet are the prime determinants, which they feed. This discovery originated in then alkaloid profiles ought to show strong the observation that skin extracts ofcaptive- correlations with habitats and kinds of prey. raised dendrobatids contained no alkaloids With this in mind, the occurrence of skin (Daly et al., 1980, 1992). It was then dem- alkaloids in nine species ofMantella are pre- onstrated that alkaloids fed to captive-raised sented here in conjunction with general ob- frogs were effectively sequestered into skin servations on habitats. Specifics of the diet (Daly et al., 1994a). The ability to sequester ofMantella spp. and ofpossible interspecific alkaloids is absent in the dendrobatid Colos- differences in the genetic base of their se- tethus, which lacks lipophilic alkaloids in the questering systems remain open questions. wild. Finally, feeding leaf-litter insects to dendrobatids being raised in terraria led to SUBFAMILY MANTELLINAE LAURENT, 1946 establishment of significant levels of many, The correct familial placement ofthe Man- but not all, of the alkaloids present in wild- tellinae is uncertain, as discussed below un- caught frogs from the leaf-litter locality (Daly der Mantella. The content ofthe group varies et al., 1994b). What arthropods might be the somewhat according to author, but always source of alkaloids not detected in dendro- includes at least the genera Mantella and batids that were fed leaf-litter insects is un- Mantidactylus (s.l.). Noble (1931, and in No- known. Indeed, it is still possible that such ble and Parker, 1926: 4n) proposed a close alkaloids (pumiliotoxins, histrionicotoxins, relationship between these frogs on the basis decahydroquinolines) may be synthesized by ofcertain anatomical similarities and a Mad- the anuran in the wild, but not when raised agascan distribution. Laurent (1946) erected in captivity. a subfamily to include the Madagascan gen- 4 AMERICAN MUSEUM NOVITATES NO. 3177

M. viridis (9E)

M. laevigata (7F) M. betsileo (6C) M. betsileo (6D), M. Iaevigata (7DE), M. pulchra (9AB)

M. baroni (8D), M. pulchra (9C)

i(7AB), M. baroni (8B) 1. baroni (8C)

M. baroni (8E) M. baroni (8F)

M. expectata (7C)

Tolagnaro 0 100 Km -. L - - -

Fig. 1. Madagascar, showing collecting localities for Mantella spp. discussed in this paper. The numberAetter combinations within parentheses match the population designations in figures 6-9 (gas chromatograms) and in appendix 2. era Mantella, Mantidactylus, Gephyroman- 1 943b, 1946) makes no mention ofthe genus tis, Trachymantis (= Laurentomantis), and and one assumes that he accepted at face val- the Sri Lankan Pseudophilautus. ue Noble's proposal that it is related to Man- Except to designate Mantella as type genus tidactylus. Since then, a few authors (notably of the Mantellinae, Laurent's work (1 943a, Liem, 1970, and Guibe, 1978) have not rec- 1 996 DALY ET AL.: MADAGASCAN POISON FROGS 5 ognized the subfamily but neither have they suspect that the glands correlate with similar been concerned with the relationships of mating behavior at least in some cases. Fem- Mantella itself. Otherwise, except for doubts oral glands, shown as a derived character in concerning the inclusion of Pseudophilautus two places on her cladogram (her fig. 2, with (see Frost, 1985: 439), no one seems to have character 28 as part of nodes D and J), were explicitly questioned the naturalness of the considered to be a primitive feature in the Mantellinae, which has been uncritically ac- Mantellinae (Blommers-Schlosser, 1993: cepted as a unit for various phylogenetic anal- 212), although the supposedly correlated yses (see below). Half a century after Noble mating behavior was proposed as a synapo- first proposed the relationship, Blommers- morphy for the subfamily. Schlosser (1979: 65) asserted that: Femoral glands in Mantella have not been figured or described to our knowledge. Blom- The genus Mantella is very closely related to the mers-Schl6sser (1979: 65) said that the glands genus Mantidactylus (especially to the wittei, depres- siceps and pulcher groups, see Table IV for compar- are diffiuse and Blommers-Schl6sser and Blanc ison). The cytogenetic data agree with those of these (1991: 263) indicated that femoral glands are three species groups (cf. Blommers-Schl6sser, 1978). a male character in Mantella. Glaw and The tadpole resembles those ofthe wittei and depres- Vences (1994: 70) indicated that the glands siceps groups (cf. Arnoult, 1966). The lateral meta- tarsalia are entirely connected. The omosternum is are indistinct or absent in male Mantella and widely forked posteriorly and the stemum is slightly absent in females. Cursory examination of forked anteriorly in Mantella cowanji (cf. Guibe, 1978) some of our voucher material (appendix 1) and the femoral glands are diffuse. suggests that the purported glands are coex- istent with the patch of granular skin on the In habitus and general color pattern, some underside ofthe thigh; the granular patch oc- Mantella are also reminiscent of the several curs in both males and females, but sexual bicolored species of the Mantidactylus al- dimorphism seems evident only in M. baroni bofrenatus species group (see p. 160 and color and M. betsileo. In female baroni, the gran- photographs 111-1 13 in Glaw and Vences, ular thigh patch is pale brownish, whereas it 1994). In fact, it is not clear whether Man- is paler and has a more sharply defined pe- tidactylus (s.l.) possesses any synapomor- rimeter in males. In female betsileo, the patch phies that would exclude the monophyletic is black and consists of large angular "gran- Mantella. ules," whereas it is pale brownish and Nonetheless, the significance of any phe- smoother in the males. In these species, male notypic similarity remains to be elaborated. and female specimens could be separated by Convincing synapomorphies still have not the naked eye on the basis of differences in been presented to link Mantella with any oth- thigh granulation. There were no pigmenta- er genus and the monophyly of the Mantel- tion differences in a few M. crocea, although linae must simply be questioned as a matter granulation is perhaps stronger in the female, of course. and there were no obvious sexual differences The only explicit mantelline synapomor- in M. pulchra, M. expectata, M. /aevigata, phy proposed by Blommers-Schlosser (1993: and M. viridis (unskinned males were not 210, 212, fig. 2) is a loosely defined behav- available for M. aurantiaca or M. cowan/i). ioral trait: However, in addition to species differences, Abbreviated mating contact ... no real amplexus; the there is notable individual variation in the male covers the head and shoulders ofthe female with granular thigh area at least in M. expectata. his thighs (femoral glands) in arboreal species or em- Some male and female expectata have the braces the female very loosely, inguinally or axillary, ventral thigh patch appearing "normally" in ground dwelling species, which lasts very short, from seconds to a few minutes. This behavior is known granular, whereas others have deep parallel only from the Mantellinae. grooving in a superficially smoother patch. Strongly granular skin underneath the thigh But, as noted by Blommers-Schlosser is commonplace among frogs, but this area (1993: 210), femoral glands occur in at least seems exceptionally variable in Mantella. a half-dozen [petropedetine and ranine] gen- Femoral "glands" in Mantella, when they era other than mantellines, and one might occur (e.g., in betsileo and baroni as described 6 AMERICAN MUSEUM NOVITATES NO. 3177

A B

C D

Fig. 2. A, B. Mantella aurantiaca from the Andasibe region; orange and red variants from different demes (AMNH 136892, 2 23 mm SVL, and 136921, 9 22 mm SVL, respectively). C. Mantella baroni (M. madagascariensis, auctorum), from An'Ala (AMNH 140556, 8 24 mm SVL). D. Mantella betsileo from Ambavala (AMNH 140573, 2 23 mm SVL). above), bear little superficial resemblance to visible. Glaw and Vences (1994: 70) tabulat- the often strikingly well-defined glands in ed presence-absence and distribution by sex Mantidactylus (s.l.), some of which have a of the glands in the subgenera or species central pore,4 and the homology might there- groups of Mantidactylus, and histological fore be questioned. However, Blommers- comparisons with Mantella would be useful. Schlosser (1979: 35, 44, 65) indicated a sim- The possibility of seasonal variation should ilarity in stating that some Mantidactylus have also be investigated. the femoral glands "diffused" or not always Alkaloids were lacking in three species of Mantidactylus (Garraffo et al., 1 993b: 1017), Man- 4Blommers-Schlosser's (1979: 66) schematic draw- although present in the nine species of ings of femoral gland variation in Mantidactylus (s.l.) tella examined. The possibility that Manti- are variously misleading in either not explicitly distin- dactylus (s.l.) is paraphyletic with respect to guishing the glands from the area of coarsely granular Mantella is mentioned above, even though skin on the underside of the thigh or in implying the no one has made a convincing case for re- absence or restriction of such granular skin. lating these genera. We neither assume nor 1996 DALY ET AL.: MADAGASCAN POISON FROGS 7

A B

C F D

Fig. 3. A. Mantella cowanii from unknown locality (AMNH 140550, 2 26 mm SVL). B. Mantella crocea from 14-18 km N Andasibe (AMNH 136897, adult 8, 17.5 mm SVL). C. M. expectata from Massif Isalo (AMNH 136938, 9 26 mm SVL). D. Mantella laevigata from Ambavala (AMNH 140564, 9 24 mm SVL).

reject the monophyly of the Mantellinae in Genus Mantella Boulenger, 1882 this paper, which focuses on Mantella without further reference to Mantidactylus (s.l.) The three oldest species names assigned to or other putative mantelline genera. Man- this genus were first placed in Dendrobates tella probably is monophyletic as defined by (betsileo and madagascariensis of Grandi- Guibe (1978: 81-82) and Blommers-Schl6s- dier, 1872, and ebenaui Boettger, 1880). Bou- ser (1979: 61), and we here suggest that the lenger (1882: 141) erected the genus Mantella genetic mechanism leading to presence ofdi- in the family Dendrobatidae only for these verse lipophilic alkaloids in the skin (whether three nominal species, basing his concept pri- from sequestering or biosynthesis) is an ad- marily on available specimens ofM. betsileo ditional synapomorphy for the genus. But the (which, therefore, was appropriately desig- aposematic frogs ofthe Neotropical Dendro- nated as type species by Liem, 1970: 100). batidae are similarly defined (e.g., Myers et Boulenger (1882: 471-472) named Mantella al., 1995) and so the questions of relatedness cowanii as an addendum to his new genus. or convergence are revisited below. Although the placement within the Den- 8 AMERICAN MUSEUM NOVITATES NO. 3177 drobatidae was followed by a few workers synapomorphic compared with the more into the 20th century (e.g., Werner, 1901; widely distributed conditions in Mantella. Mocquard, 1909: 65; Hewitt, 1911), Man- Despite some striking similarities in apo- tella subsequently became aligned not with sematic colorations (see below), coloration New World dendrobatids but with Old World does not provide any obvious support for "tree frogs." Noble (1931: 524-526) allied hypothesizing relationship between dendro- Mantella with other Madagascan genera that batids and Mantella. The most plesiomorph- he placed in the redefined Polypedatidae (= ic aposematic dendrobatids are striped, Rhacophoridae), concluding that "Mantella whereas the least colorful species ofMantella may be considered a terrestrial tree frog, for are bicolor, with brown to green dorsa and its pads, although small, agree with those of darker brown or blackish sides. Polypedates." Mantella does approach a correlated con- Laurent (1946) broke away from this view dition ofthe tympanum andjaw musculature by assigning Mantella to the Ranidae, which that is considered synapomorphic for the he later (1951) expanded to comprise addi- Dendrobatidae. In dendrobatids always and tional subfamilies (including the Rhacophor- in Mantella frequently, the posterodorsal inae). Since then, Mantella (alone or as part (and, in dendrobatids, sometimes dorsal) part of the Mantellinae) has been shuffled about of the tympanum is externally concealed. In between the Ranidae (e.g., Guibe, 1964; Mantella, the partial tympanic concealment Blommers-Schlosser, 1979; Dubois, 1984) is externally correlated with a weak to strong and the Rhacophoridae (e.g., Liem, 1970; supratympanic fold, which largely follows the Guibe, 1978; Channing, 1989), and the Man- underlying anterior edge of the m. depressor tellinae also have been elevated to familial mandibulae. In dendrobatids, a similarly rank (Blommers-Schl6sser and Blanc, 1991). aligned supratympanic fold varies from More recent analyses include those of Blom- primitively present (e.g., in Aromobates) to a mers-Schl6sser (1993), who returned to Lau- vague swelling (e.g., in Epipedobates trivit- rent's (1951) subfamilial classification and tatus) to completely absent (e.g., in Dendro- explicitly treated the Mantellinae and Rha- bates pumilio). But whether the supratym- cophorinae as related subfamilies (sister panic fold is present or absent, the partial groups) within the Ranidae, and Ford (e.g., concealment ofthe tympanum ultimately re- 1993: fig. 3), who considered the Ranidae as flects the fact that the large external slip' of being nonmonophyletic and regarded the the m. depressor mandibulae partially over- mantellines as being nested within the family laps the tympanum in dendrobatids (Myers Rhacophoridae. To summarize, the mantel- et al., 1991: fig. 5), and, similarly overlaps or lines have been on a kind of taxonomic see- else skirts the edge of the tympanic ring in saw but have not fallen offdefinitively on one Mantella. The usual condition in Mantella side or the other. (i.e., in the nine species in appendix 1) is for One ofthe few things implicit in taxonomic discourse of the last six decades is that mantellines are not closely related to dendroba- s The m. depressor mandibulae is similarly configured tids, even though general resemblances be- in dendrobatids and Mantella. A large superficial slip tween Mantella and Dendrobates still receive originates broadly from the dorsal fascia, completely comment. Cited anatomical differences in- concealing a smaller, deeper slip originating from the clude external and internal structure of the long otic ramus of the squamosal bone. In Mantella, a digits (supradigital scutes in dendrobatids, and third slip, not concealed ventrally, originates from the intercalary cartilages and Y-shaped terminal posterior and posteroventral margin ofthe tympanic ring. phalanges in Mantella), and thigh muscula- In dendrobatids, the third slip is more variable and often poorly defined, sometimes virtually absent or not sep- ture (distal tendon of the m. semitendinous arable from the slip that originates mainly from the otic pierces that of the m. gracilis complex prior ramus (e.g., Myers et al., 1984: 8). Such divisions of the to insertion in dendrobatids but inserts deep- m. depressor mandibulae occur in certain other anurans, er than the m. gracilis in Mantella). The but the overlapping of the tympanum by the largest slip aforesaid dendrobatid character states are may be less common. 1996 DALY ET AL.: MADAGASCAN POISON FROGS 9

A .2

ffL. -x

Fig.4. A, B. Mantellapulchra from An'Ala (AMNH 136907,621 mm SVL,136908,922 mm SVL). C. Mantella viridis from Montagne des Fran9ais (AMNH 140580, 2 26 mm SVL). 10 AMERICAN MUSEUM NOVITATES NO. 3177

-io 0000,.

Fig. 5. Comparisons in dorsal and ventral views of Mantella baroni (AMNH 140556) on left side, and Mantella pulchra (AMNH 140552) on right. Both from 2 km SW An'Ala. These specimens (d left, Q right) are equivalent in size (24 mm SVL), but, taking sexual dimorphism into account, M. baroni is the larger species. 1 996 DALY ET AL.: MADAGASCAN POISON FROGS I1I the depressor muscle to slightly overlap the tion ofskin alkaloids, Mantella is convergent posterodorsal part of the tympanic ring, but with aposematic dendrobatids in a general there is no overlap at all in several skinned way. Thus, the unicolored Mantella auran- carcasses ofM. pulchra, in which any external tiaca (fig. 2A) is comparable to a few Phyl- concealment is due simply to muscle bulk lobates (although the ancestral patterns are raising the skin away from the tympanum. very different). Bright flash marks in con- Another, even more striking similarity be- cealed parts of the limbs of some Mantella tween Mantella and the "advanced" (apo- have their counterparts not only in the den- sematic) dendrobatids is ofcourse the subject drobatid genus Epipedobates but in a variety of this paper-the shared presence of lipo- of other frogs as well. As long recognized, philic alkaloids. This trait, however, also is some Mantella are convergent mainly with shared with Australian Pseudophryne and Dendrobates because the iridescent coloring South American Melanophryniscus, as first is displayed as rounded markings and even reported by Daly et al. in 1984. Maxson and limb bracelets. Mantella cowanii (fig. 3A), Myers (1985: 54) noted that dendrobatid an- which resembles some populations of Den- tisera showed no cross reactivity with Man- drobates histrionicus, is the best example of tella albumins or Pseudophryne antisera, a Dendrobates-like pattern. Species of Den- which was taken to imply that "any phylo- drobates, however, are highly variable in col- genetic association could be no more recent or pattern both within and among popula- than some 100-120 million years ago." tions. Such variation has not yet been clearly Clearly, much remains to be learned about demonstrated for any Mantella (see below). the comparative anatomy and phylogenetic Some species of Dendrobates (e.g., Pana- relationships of major anuran groups. None- manian populations of D. pumilio) do not theless, present evidence suggests that apo- uniformly occupy available habitat but occur sematic dendrobatids and Mantella are as dense populations ofwell-separated demes. monophyletic groups that are derived rela- It is an intriguing possibility that at least tive to other dendrobatids and other pur- Mantella baroni may exhibit similar popu- ported mantellines. All the similarities there- lation structure (see observation by John Ca- fore may be postulated as striking examples dle on page 18). If so, one might also expect ofconvergence beyond ranoid (or hyloid and similar territorial behavior (which might also ranoid) synapomorphies. More than 60 years correlate with lack of pronounced sexual di- ago, G. K. Noble penned the following note morphism in size). to himself: The number ofrecognized species ofMan- tella has about tripled in the last decade (since It is queer that loss of teeth often runs to iridescent Frost, 1985), in part due to discovery ofnew colors. Note Dendrobates and Mantella. (From Gen- species concurrent with increased fieldwork era Salientia [n.d.], a bound volume of typed literature excerpts and hand-written notes, in the AMNH in Madagascar, and in part due to resurrec- Dept. Herpetology library.) tion of valid species names from synonymy. Much of the taxonomic confusion has in- We can now suggest a few pieces to Noble's volved belief that at least one species is ex- puzzle (but in recognition of widespread traordinarily variable (like some Dendro- toothlessness, in dull toads for example, we bates)-a concept for which the name Man- should reverse his statement-iridescent col- tella "madagascariensis," has been most re- ors often run to loss ofteeth). The bright hues cently used, even though Guibe (1964) set are warning colors advertising the existence this name aside as being unattributable. of noxious skin alkaloids. Loss of teeth in Glaw and Vences (1992a, 1994) seem to these cases probably reflects specialization on have renewed the belief that extreme vari- very small prey, some ofwhich may provide ability in color pattern is expected in single the alkaloids sequestered and accumulated populations ofMantella, first (1 992a: 166) in by the frogs. It is an ecological niche entered their announcement of a "variable colour more than once. morph," and then by resurrection (in quo- In color patterns, as well as in accumula- tation marks) of the unused name Mantella 12 AMERICAN MUSEUM NOVITATES NO. 3177

M. aurantiaca (B)

3231B 3A 392 323C/ 434 235C 265F

(D)

211D I 247C | 251 D / 384A 384B

M. betsileo (E) ID

307G I Fig. 6. Gas chromatographic traces showing alkaloid profiles from Mantella spp. A. M. aurantiaca (10 skins, approx. 14-18 km N Andasibe, upland swamp forest, Nov. 1989). B. M. aurantiaca (5 skins, same region and habitat as preceding, Jan. 1993). C. M. betsileo (11 skins, Antanambaobe, inland clove tree forest, Dec. 1990). D. M. betsileo (8 skins, Ambavala, remnant forest around a bamboo grove, Jan. 1994). E. M. betsileo (12 skins, Farakaraina, disturbed coastal forest, Dec. 1993). F. M. cowanii (3 skins, from dealer, Dec. 1993). Boldface designations of alkaloids match those in appendix 2. The OV-1 column is programmed to 280°C at 10°C per minute from an initial temperature of 150°C; other conditions are as given in Daly 1996 DALY ET AL.: MADAGASCAN POISON FROGS 13 loppei (from Roux, 1935) for some highly All these have been investigated for skin diverse specimens (1994: 185, black and white alkaloids except Mantella bernhardi and M. figs. 335-337 and color photos 58-60). But haraldmeieri, which are both from south- most, ifnot all, ofthese seem to be frogs from eastern Madagascar. The nine species sam- the animal trade and there is no telling how pled are shown in figures 2-5 and the local- many populations or demes or species might ities sampled are mapped in figure 1. be represented-nor can the possibility of The following accounts summarize field captive-produced hybrids be ruled out a observations7 and results ofalkaloid analyses priori. for each of the nine species. Gas chromato- Commercial and other exportation of graphic profiles are shown in figures 6-9. The Mantella (see Glaw and Vences, 1994: 31- occurrence ofindividual alkaloids ofvarious 326), especially to Europe, has beneficially classes is presented in appendix 2 and the stimulated much of the recent work on these structures of representative alkaloids are frogs but also is introducing a new kind of shown in figures 10-12. confusion into the literature. Localities, if available at all, are sometimes based on sec- Mantella aurantiaca Mocquard, 1900 ond- or third-hand information that is au- Figure 2A, B tomatically suspect until verified in the field, and we especially urge authors to indicate This easily recognized unicolored frog ap- clearly any locality data provided by dealers pears to occupy a very small range in the (who understandably might be unwilling to uplands ofeast-central Madagascar, where its share their sources of income). distribution correlates with wet montane for- The number of species of Mantella will ests (Glaw and Vences, 1992a, 1994; Zim- increase with investigation of possibly un- merman et al., 1990). Specimens used for named species (e.g., see mention under M. alkaloid analysis were collected with the help betsileo and M. laevigata) and discovery of of local people in upland swamp-forest hab- others. But, excluding Mantella loppei (see itats in the region north of Andasibe. The above), the 11 named species of Mantella widely disjunct swamp forest localities (about that seem clearly recognizable at this time are 1000 m elev.) were separated mainly by cul- as follow: tivated areas comprising open fields and forest of pine or eucalyptus. The swamp forests Mantella aurantiaca Mocquard, 1900 themselves contained varying amounts of Mantella baroni Boulenger, 1888 (M. mad- standing water, grassy hillocks and a variety agascariensis, auctorum) of trees, including Pandanus. Frogs were ac- Mantella bernhardi Vences et al., 1994 tive throughout such habitat. The general area Mantella betsileo (Grandidier, 1872) is located in the region of the type locality, Mantella cowanii Boulenger, 1882 "une foret entre Beforona et Moramanga" Mantella crocea Pintak and Bohme, 1990 (Mocquard, 1900: 111). Mantella expectata Busse and Bohme, 1992 Demes from different swamp forests dif- Mantella haraldmeieri Busse, 1981 fered somewhat in coloration, varying from Mantella laevigata Methuen and Hewitt, 1913 yellow through orange to red, any given color Mantella pulchra Parker, 1925 being nearly uniform dorsally and ventrally Mantella viridis Pintak and B6hme, 1988 (including palms and soles), with a red calf spot in the concealed part of the shank that 6 We agree with these authors that controlled exportation and especially habitat management are probably better conservation solutions for Mantella than attempted "protection" through CITES listing, which often has 7Fieldwork was accomplished by the first two authors the unintentional effect of making research so difficult and local guides, with help at times from the third author as to be impractical. and A. Rabemanantsoa. et al. (1993). Emergent temperatures can differ somewhat with different columns and variations in flow rates. Some typical emergent temperatures for some common mantelline alkaloids are: 217B 1660, 251D 1720, 267C 1900, 307A 2160, 323A 2300. 14 AMERICAN MUSEUM NOVITATES NO. 3177

M. crocea (A) M. crocea (B)

307A

307A I 323B

M. expectata (C) M. Iaevigata (D)

195C>195A

307A 323D L193D 307G

M. laevigata (E) M. laevigata (F)

195C, 195A, 193D V 249A 30,17G

Fig. 7. Gas chromatographic traces showing alkaloid profiles from Mantella spp.; conditions as in figure 6. A. M. crocea (10 skins, approx. 14-18 km N Andasibe, upland swamp forest, Nov. 1989). B. M. crocea (35 skins, same region and habitat as preceding, Jan. 1993). C. M. expectata (20 skins, Massif Isalo, semiarid streambed, Jan. 1993). D. M. laevigata (6 skins, Ambodimanga, bamboo grove, Dec. 1990). E. M. laevigata (5 skins, Varary, bamboo grove, Jan. 1994). F. M. laevigata (6 skins, Nosy Mangabe, insular forest, Dec. 1993). . 1996 DALY ET AL.: MADAGASCAN POISON FROGS 15

307G

< 325A

D 217A>217B

309A

291 E -, 307F

Fig. 8. Gas chromatographic traces showing alkaloid profiles from Mantella baroni (M. madagascariensis, auctorum); conditions as in figure 6. A. 10 skins, approx. 14-18 km N Andasibe, stream-side forest, Nov. 1989. B. 6 skins, approx. 12 km SE Andasibe by road, stream-side forest, Dec. 1993 (see Garraffo et al., 1993b, fig. 2B, for a gas chromatogram of another sample of 10 skins taken at the same stream; compare samples 8B and 8B' in appendix 2). C. 10 skins, 30-35 km S Moramanga, stream-side forest, Nov. 1989. D. 3 skins, An'Ala, stream-side forest, Dec. 1993. E. 10 skins, Ranomafana, stream- side forest, Nov. 1989. F. 17 skins, Sahavondrona, disturbed stream-side forest, Jan. 1993. 16 AMERICAN MUSEUM NOVITATES NO. 3177

M. pulchra (A) M. pulchra (B)

I1I 267H 293B 382, 384A, 384B \ 384A 384B /

M. viridis (E)

Fig. 9. Gas chromatographic traces showing alkaloid profiles from Mantella spp.; conditions as in figure 6. A. M. pulchra, 6 skins, near Ambavala, boggy ridge forest, Dec. 1990. B. M. pulchra, 5 skins, same locality as preceding, Jan. 1994. C. M. pulchra, 5 skins, An'Ala, boggy stream-side forest, Jan. 1993. D. M. viridis, 30 skins, Montagne des Fran9ais, semiarid streambed forest and nearby drainage areas, Jan. 1994. E. M. viridis, 5 skins, approx. 13 km S Antsiranana, semiarid streambed forest, Jan. 1994. F. M. viridis, 1 skin, "region of Antsiranana," from dealer, Nov. 1989. 1 996 DALY ET AL.: MADAGASCAN POISON FROGS 17 is most evident on yellow frogs. We have seen tween 1882 and 1981 (cowanii, haraldmeieri, red-colored specimens of unknown prove- pulchra), leading to an erroneous belief in a nance with a black ear spot. The uniformly species of extraordinary variability-the bright coloration seems highly derived, but composite species having been known until the dorsum and limbs of metamorphs have now as M. cowanii (sensu Guibe, 1964, 1978) dark markings according to Arnoult (1966: and M. madagascariensis (sensu Busse, 1981; 938, fig. 4), possibly representative of an an- Blommers-Schl6sser and Blanc, 1991; Glaw cestral pattern. and Vences, 1992a, 1994). Skin extracts from seven samples of M. Guibe (1978: 84) had properly considered aurantiaca all contained mainly pumiliotox- Dendrobates madagascariensis Grandidier ins and allopumiliotoxins. Levels ranged from (1872) as an "espece douteuse," that is, as a very high to quite low (fig. 6A,B and data not species indeterminata or a nomen dubium- shown; see also fig. 6A of Daly et al., 1984). a species not identifiable from the original Homopumiliotoxins also occurred, but in publication or a name not certainly appli- minor amounts. Varying, but always minor cable to any known species of frog. He based amounts of the alkaloids 235C and 233F, this decision on the poor condition of the proposed to be dehydrohomopumiliotoxins syntypes and on the inadequate original de- (Garraffo et al., 1993b), were present. One scription, which reads as follows (Grandidier, trace alkaloid 323C appears to represent a 1872: 10-11): new subclass of pumiliotoxins and will be d D'un noir bleuatre uniforme. Abdomen seme de referred to as an "isopumiliotoxin" until its taches d'un bleu clair-, cuisses et face interne desjambes structure can be defined. No decahydroqui- d'un beau rouge. nolines were detected. "Izidine" alkaloids oc- 9 D'un beau noir mat avec une tache d'un vert clair curred rarely and only as minor or trace con- veloute a la naissance et sur I'avant de chacun des quatre membres. Abdomen seme de taches d'un bleu stituents. A 5,8-disubstituted indolizidine ciel. Face interne des jambes d'un beau rouge. 239C and a putative 1,4-disubstituted quin- La peau de ces Dendrobates est finement chagrinee. olizidine 265L were present in two samples Long. du corps, Om,022; des membres poster., obtained from commercial dealers. Two mi- Om,032. nor alkaloids 392 and an 0-acetyl derivative Habit.: Foret d'Ambalavatou, entre Mananzarine 434 represent a new class of alkaloids. Al- et Fianarantsoua. kaloid 392 has an empirical formula of Grandidier's emphasis on coloration ofthe C22H36N204 and contains the equivalent of male's thighs and, in both sexes, the con- six rings or double bonds. It affords a mass cealed part of the lower leg ("thighs and in- spectral base peak at m/z 252 (C14H22NO3). side surface ofthe legs a beautiful red") seems Ifthe skin alkaloids ofM. aurantiaca orig- more applicable to Mantella pulchra than to inate from dietary sources, then the swamp Mantella baroni (cf. colors in fig. 5), but, un- forest habitat must provide arthropod prey less new information is forthcoming, we sug- affording large amounts ofpumiliotoxins and gest that the name Dendrobates madagas- allopumiliotoxins, only small amounts ofthe cariensis Grandidier be left unapplied, as a "izidine" alkaloids, and none of the decahy- nomen dubium as intended by Guibe (1978).8 droquinolines. Mantella baroni Boulenger, 1888 8 After writing the above, we obtained the second edition of Glaw and Vences' admirable Fieldguide, which (Mantella madagascariensis, auctorum) was published in December 1994. These authors (p. 403) Figures 2C, 5 designated a lectotype ofDendrobates madagascariensis The name Mantella baroni is here resur- because "the paralectotype differs largely from the lec- rected for a species that, as currently recog- totype and probably belongs to a different species ... nized, occurs in upland forest over a long Both specimens are in a very bad state of conservation . . . Some colour patterns are present on the legs (which geographic range in eastern Madagascar. are separated from the body) [ofthe lectotype].... These Based primarily on studies ofpreserved spec- colour patterns partly indicate an attribution to the col- imens, the distinctive coloration and color our morph which in the past was attributed to Mantella pattern ofM. baroni has been confused with madagascariensis, and partly also are similar to ... the patterns ofother valid species named be- Mantellapulchra. No unequivocal attribution ofthe mad- 18 AMERICAN MUSEUM NOVITATES NO. 3177

Although the type locality of M. baroni is stream in second-growth forest near Saha- only "Madagascar," Boulenger's (1888: 106, vondrona. pl. 6, fig. 2) description and illustration seem John E. Cadle (in litt., August 6, 1995) con- clearly applicable to the present species. Our firmed the above observations on apparent samples, which span a north-south distance habitat preference of M. baroni but also em- ofnearly 300 km (figs. 2C, 5 left), show little phasized that the species is sometimes found variability in color pattern. We know baroni in unlikely places: as a black frog with a pale, sharply defined In the Ranomafana region (Ranomafana National canthal-supraocular line, lacking a labial line, Park and surrounding areas), Mantella baroni is found a large green lateral blotch that runs onto the in a variety ofhabitats. Although most specimens are arm, and a green inguinal blotch extending found near streams, actual stream-side habitats vary; the recorded elevation in this region is about 700- onto the posterior flank and continuous with 1200 m. I would characterize most situations as ri- an elongated green blotch covering most of parian primary montane rainforest; the streams usu- the upper thigh; in vivid contrast to all the ally have a rocky (as opposed to sandy) substrate, and green on black, the shank and foot are totally are generally fast-flowing white-water streams. An- other streamside habitat where they were abundant orange with black markings. The shank and in one area was dense grass (up to > 1 m tall). Mantella foot also are orange below, but all the rest of baroni was also found in swampy areas of little relief the ventral surfaces are black with blue mark- with meandering stream courses, and characterized ings (fig. 5, left). The green markings may run by arborescentPandanus species, many with stilt roots; to yellowish green, and there is said to be a these areas were often boggy and with sandy substrate. One specimen was collected in primary montane rain- morph with yellow markings but it was not forest away from streams. Finally, near the village of seen by us. Sahavondrona, M. baroni was seemingly common on Samples of skins were obtained at six lo- a hillside with a very dry aspect and chaparral-like calities (fig. 1), with an additional sample of shrubby vegetation (very likely disturbed); although there were some streams in this area, several man- unknown orgin from a commercial dealer. tellas were found out in the open away from streams, The habitat varied somewhat, although all and I was surprised to find them in this seemingly specimens collected were found along streams stressful environment (they were the only frogs found in upland forest in an elevational range of in that habitat). about 800-1000 m above sea level. Collec- One thing that impresses me overall about the distribution ofMantella baroni in the Ranomafana area tions were made along fairly rapidly flowing is the very local nature ofpopulations. They are abun- streams in relatively undisturbed forest at sites dant in small areas, then apparently absent over wide east ofAndasibe and near Ranomafana. Oth- stretches ofapparently suitable habitat. This is all the er collections were made along small mean- more surprising given the habitat breadth for the spe- dering streams in forest exhibiting varying cies in the general region. degrees of disturbance at the sites north of Skin extracts from all samples had mod- Andasibe, south of Moramanga, and near erate to high levels of alkaloids (fig. 8A-F; An'Ala. One population was found by a large see also fig. 2B in Garraffo et al., 1 993b). The alkaloid profiles varied among populations, but pumiliotoxins and/or allopumiliotoxins were always major alkaloids. "Izidine" al- agascariensis lectotype to any described form is possible kaloids were also prominent components. The [emphasis added]; we suggest to continue using the name 5,8-disubstituted indolizidine 217B and the madagascariensis for the morph figured on cp. 61. The 1 ,4-disubstituted quinolizidines 217A and synonyms Mantella baroni Boulenger, 1888, and Phry- 231A were present in all samples either as nomantis maculatus Thominot, 1889, belong to this major or minor alkaloids. Other indolizi- morph and we therefore continue considering these names dines and quinolizidines were also present. as synonyms of madagascariensis." Homopumiliotoxins often were present as Their own conclusions argue against the last action. minor components. Other alkaloids occurred Except in the rare case of objective synonyms (based on in one or more samples. Thus, the skin sam- the same type specimen), junior synonyms cannot be attributed to an older name of an unidentifiable species. ples from relatively undisturbed forest by a Because ofthe confusion associated with the name mad- meandering stream near An'Ala (fig. 8D) and agascariensis, nothing is to be gained by petitioning for by a cascading rocky stream reached by road its conservation. east of Andasibe (fig. 8B) contained a major 1 996 DALY ET AL.: MADAGASCAN POISON FROGS 19 alkaloid 281F (C17H31N02) that was pro- the southeastern plateau-west ofthe eastern posed to be a dihydropumiliotoxin (Garraffo forest zone on the escarpment and therefore et al., 1993b). The 8-deoxypumiliotoxins well inland from the eastern lowlands (maps represent a new subclass of alkaloids de- in Grandidier, 1893; Labatut and Rahari- scribed from dendrobatid frogs (Jain et al., narivonirina, 1969; Methuen and Hewitt, 1995). The sample from east of Andasibe 1913; Sibree, 1880)-which causes us con- contained an 8-deoxypumiliotoxin 291E siderable concern and leads us to question (C19H33NO) and a previously undetected al- whether the name M. betsileo is being cor- kaloid 251P (C16H29NO) of unknown struc- rectly applied. A search for existing Mantella ture. Alkaloid 251P had a mass spectral base habitat along Grandidier's route would seem peak at m/z 136 (CqH14N+ ) and a major frag- worthwhile. Grandidier (1893) provided a ment at m/z 122 (C8H12N+). The skin sam- detailed route map of his 1869-1870 expe- ple from near Sahavondrona (fig. 8F) condition, including travel in the Betsileo terri- tained the pyrrolizidine oximes 236 and 252A, tory. A contemporary account ofthe territory possibly of millipede origin, and an alkaloid and its limits was given by Shaw (1875). 281G (C17H31NO2) that is proposed to be an Two of the localities reported here are in- 8-deoxypumiliotoxin. land from the northeast coast (fig. 1), but still The skin sample of unknown origin (from at low elevations. Our collections were made a dealer) had high levels ofallopumiliotoxins in three quite different habitats, one from an 323B and 325A (gas chromatograph not inland clove tree forest (about 100-200 m shown). It was lacking the quinolizidines 217A elev.), one from a disturbed patch of inland and 231A found in all other M. baroni sam- forest and second growth around a small ples. It had a previously undetected 3,5-di- bamboo grove (about 100-200 m), and one substituted indolizidine 211E (C13H2_NO), a from disturbed, very open coastal forest (50 putative 1 ,4-disubstituted quinolizidine m). The inland collections were in the Man- 265L, and an alkaloid 392 (C22H36N204) of anara Reserve area near Sandrakatsy. M. bet- unknown structure. sileo was microsympatric with the semiar- An early report (Daly et al., 1984) on al- boreal M. laevigata at the bamboo grove. M. kaloids of M. madagascariensis and another betsileo was also seen in microsympatry with (Garraffo et al., 1993b) on M. sp., cf. mad- M. pulchra in nearby, relatively undisturbed agascariensis, both apply to Mantella pulchra forest. One specimen (AMNH 140574) of and are discussed under that name. Mantella from that forest seemed in the field somewhat intermediate between microsympatric M. betsileo and M. pulchra; the pos- Mantella betsileo sibility of hybridization was considered, in- (Grandidier, 1872) asmuch as the specimen had the dorsal col- Figure 2D oring of betsileo and the lateral body pattern The name Mantella betsileo is being used of pulchra. However, it had golden yellow for a species with an orangish (or yellowish) flash marks in the groin and in the calf and brown dorsum set off by black sides, with a lacked bright coloring on the rear ofthe thigh pale labial line extending to the arm; the spec- (compare hind leg colors in pulchra, fig. 5, imen shown (fig. 2D) was overall black ven- right). AMNH 140574 may represent an un- trally (including palms and soles) with a blue recognized species, but additional collections line around the lower lip and small, irregular are needed; color notes are provided at the pale blue markings on the venter and under end of appendix 1, under Mantella sp. Ex- the hind limbs. tracts of M. pulchra (fig. 9B), but not of M. As currently recognized, Grandidier's M. betsileo, were obtained from this site. betsileo appears to have a wide, nearly cir- All three skin extracts from M. betsileo had cumcoastal range (Kuchling, 1993, map), at relatively low levels of alkaloids (fig. 6C, D, elevations from sea level to about 500 m E), but alkaloid profiles differed significantly (Glaw and Vences, 1994: 177). However, the among the three populations. The inland type locality, "Pays des Betsileos" [land of populations (fig. 6C, D) contained a decahy- the Betsileo Tribe], is inland and upland on droquinoline (195A), accompanied by un- 20 AMERICAN MUSEUM NOVITATES NO. 3177 saturated analogs (193D, 189) and apparent ognition of M. cowanii, its taxonomic status dimers 384A and 384B. The "dimers" had has been greatly confused, mainly because, empirical formulas of C26H44N2. Pumili- until recently, workers have not seen living otoxins, allopumiliotoxins, and homopumi- specimens, but have attempted to include the liotoxins were minor components of the in- pattern of preserved specimens into pre- land populations. The inland sample from sumptive variation of other species. Busse the bamboo grove had a 5,8-disubstituted in- (1981) considered it a synonym of "mada- dolizidine 217B, and a 1,4-disubstituted gascariensis," whereas other recent authors quinolizidine 217A as minor components (fig. have used cowanii as a valid name (usually 6D). A related quinolizidine 219F was pres- spelled "cowani") but have misapplied it to ent as a minor component in the clove tree other species. forest population (fig. 6C). An alkaloid, 195C, Bohme et al. (1993) cleared up some ofthe was present as a major component in one nomenclatural confusion and provided pho- inland population (the clove tree forest), but tographs in dorsal and ventral view of a syn- was absent in the other (bamboo grove). The type of M. cowanii, concluding that it might structure of195C remains uncertain (see Gar- be distinct from "madagascariensis." The raffo et al., 1 993b). Thus, the two inland pop- type locality is East Betsileo. Specimens re- ulations were similar in containing the de- cently found their way into the animal trade, cahydroquinoline 195A as a major compo- reputedly from upland forest (about 1500 m nent along with "dimers" 384A and 384B, elev.) near Ambositra (Andrew Clark, per- and with pumiliotoxins, allopumiliotoxins, sonal commun.). And Vences et al. (1994) and "izidine" alkaloids as minor compo- have reported rediscovery ofthe species near nents. But the nature of the "izidine" alka- both Ambatolampy and Fianarantsoa, in the loids was quite different for the two inland same uplands as Ambositra. populations. Our specimens also were provided in 1993 In contrast, the coastal population had the by a commercial dealer, who gave an im- decahydroquinoline 195A as a minor com- probable (lowland) locality (Anosibe An'Ala) ponent and the "dimers" 384A and 384B were as source, for which reason we consider the not detected (fig. 6E). Pumiliotoxins, allo- specimens as lacking specific locality data. All pumiliotoxins, a homopumiliotoxin, pyrrol- the above localities are indicated on the map izidines, indolizidines, and a quinolizidine (fig. 1). were present in the coastal population. An The major skin alkaloid in M. cowanii was alkaloid 251Q (C16H29NO) of unknown pumiliotoxin 251D, which was present at very structure was present as a minor component high levels (fig. 6F). The only other Mantella in the coastal population and as a trace com- species that had pumiliotoxin 251D as a ma- ponent in the inland bamboo grove popula- jor alkaloid was M. expectata, a frog from tion. It afforded a mass spectral base peak at semiarid regions of western Madagascar. In m/z 122 (C8Hl2N+). M. cowanii other pumiliotoxins and allopu- Different sets ofsmall arthropods available miliotoxin 267A were present as minor com- as prey in such diverse habitats might be ex- ponents. All alkaloids appeared to be pum- pected to contribute to the marked differ- iliotoxins/allopumiliotoxins and deoxypu- ences in alkaloid profiles of M. betsileo. miliotoxins. No decahydroquinolines or "izidine" alkaloids were detected even as trace components. Mantella cowanii Boulenger, 1882 The predominance of pumiliotox- Figure 3A in/allopumiliotoxins in the skin of M. cowanii suggests that dietary prey available to or This very distinctive species of the east- targeted by this species at the presumed up- central uplands has vivid orange-red limb land forest habitat contain mainly pumili- markings and axillary/groin spots on a pure otoxin 251D and related alkaloids. It also black body; pale labial and canthal lines are suggests that arthropods containing decahy- lacking; ventral surfaces are black with a doz- droquinolines or "izidine" alkaloids are ab- en or so pale blue spots. Despite easy rec- sent from the microhabitat. 1 996 DALY ET AL.: MADAGASCAN POISON FROGS 21

Mantella crocea the nature of the swamp forests inhabited in Pintak and B6hme, 1990 this region by M. aurantiaca and M. crocea Figure 3B are needed. It is not known ifthey ever occur together. This endemic of northeastern Madagascar is, like M. aurantiaca, known only from upland swamp forests (about 1000 m elev.) north Mantella expectata ofAndasibe. The coloration seems fairly con- Busse and B6hme, 1992 stant and diagnostic (fig. 3B): A black face Figure 3C mask is sharply defined dorsolaterally until midbody, where it falls off obliquely to the This recently described species appears to venter; a pale labial line extends to the arm. occur widely in semiarid western Madagas- The dorsal surfaces above the black mask and car, where it probably has been confused with the posterior flanks are contrasting yellowish M. betsileo (see Busse and B6hme, 1992: 60). to orangish brown, ranging from nearly uni- The dorsum is greenish yellow to yellowish form to finely black-stippled, with or without brown, sometimes with a reddish brown suf- a blackish vertebral line and sometimes also fusion posteriorly, the dorsal color being with a weak, grayish hour-glass marking on sharply set off from uniformly black sides; a the back (AMNH 136896). The limbs are somewhat variable pale labial stripe extends colored much like the dorsum, often with to the arm. The limbs may be a vivid light faint dark bands on the thigh and shank. The blue (photos in Busse and Bohme, 1992; Glaw groin and rear of thigh are bright reddish or- and Vences, 1 992a), varying through grayish ange, this color also predominating under- blue (fig. 3C) to light or dark brown, with neath the shank and, variably, on the adja- blue coloring remaining on rear of thigh and cent ventral surfaces of thigh and foot. The concealed calfarea.9 The ventral surfaces are throat and venter are black with very pale overall heavily marbled or spotted pale blue bronzy yellow or whitish markings (ventral on black, with a tendency for a blue horseshoe coloration can be seen in Pintak and B6hme's shaped marking on the chin. 1990 description, and in Zimmerman et al., The type locality ofM. expectata is 20 km 1990 [as Mantella sp.]). SE Toliara; the frog has also been reported Four separate samples ofskins ofM. crocea from near Morondava (as Mantella sp., Meier, were obtained with the help of local collec- 1986, color photo), with other localities being tors. Pumiliotoxins and allopumiliotoxins added by Glaw and Vences (1994: 177). Our were major alkaloids as they had been in an- collections were made in southwestern Mad- other other swamp-forest Mantella (auran- agascar in the Massif Isalo near Ranohira. tiaca) from this region. Levels ranged from The frog occurred in streambeds (about 800 low to moderate (fig. 7A, B and data not m elev.) that were relatively dry with only shown). Furthermore, as in M. aurantiaca, standing pools ofwater. Vegetation along the the putative dehydrohomopumiliotoxins streambed consisted mainly of grassy hill- 235C and 233F were minor components and ocks. There were infrequent small trees. The "isopumiliotoxin" 323C and alkaloids 392 frogs were said to be abundant after infre- and 434 were detected. However, all samples quent heavy rains, but even calling males were ofM. crocea contained a putative 5,8-disub- difficult to locate under the dry conditions stituted indolizidine 295B (CI9H33NO2) as a that pertained when collections were made minor component. This alkaloid, which has in January 1993 and January 1994. Both a mass spectral base peak at m/z 154 samples were from the same streambed. This (C9HI6NO+ ), was not detected in any sample of M. aurantiaca. The marked similarity between alkaloid 9 A few brown-limbed specimens were seen in the sam- profiles in M. crocea and M. aurantiaca is not pled population and the limbs ofcaptives changed from unexpected if arthropod prey from similar blue to blackish brown. Busse and Bohme (1992: 60) swamp forest habitats are the source of their noted that this change does not affect the areas of blue skin alkaloids. However, further fieldwork on on the hidden parts of the hind limbs. 22 AMERICAN MUSEUM NOVITATES NO. 3177 drying bed led to a wet, spring-fed forested blue markings. Pale markings are weak or drainage area where, quite remarkably, there absent on the grayish throat and chest, which seemed to be relatively fewer frogs than in in preserved specimens are blackish brown the open streambed. to light brown, sometimes but not always in Skin extracts of M. expectata contained sharp contrast to the black belly. several pumiliotoxins including 251D, 237A Three samples of skins were obtained, two and 307A. Levels were relatively low (fig. 7C from similar inland bamboo groves (about and data not shown). One ofthe pumiliotox- 100 m elev.) in the Mananara reserve near ins (323D) was a previously undetected Sandrakatsy and the third from forest at about member of the class and its structure is un- 100 m on a small coastal island, Nosy Man- certain. Several "izidine" alkaloids were gabe. Levels ofalkaloids from all populations present including a 3,5-disubstituted indoli- were relatively low (fig. 7D, E, F). All three zidine 275C, 5,8-disubstituted indolizidines populations contained pumiliotoxin 307G as 207A and 219F, and a 1,4-disubstituted a minor alkaloid. The populations from the quinolizidine 207C. A tricyclic alkaloid 261C inland bamboo groves contained as minor (C18H31N) ofunknown structure was present. alkaloids the decahydroquinoline 195A and Another previously undetected alkaloid 197C related unsaturated congeners and trace (C12H23NO) may prove to be a hydroxylated amounts of "dimers" 384A and 384B (fig. 5,8-disubstituted indolizidine. It gave a base 7D, E). The island population did not have peak at m/z 154 (CgH16NO+) and a major the decahydroquinoline class of compounds fragment ion at m/z 96 (C6H1ON+). (fig. 7F). "Izidine" alkaloids were present, but The occurrence of pumiliotoxins/allopu- limited in number. The bicyclic alkaloid 195C miliotoxins as major alkaloids both in a frog ofunknown structure was present as a minor from a swamp forest (M. aurantiaca) and one component in extract from one bamboo grove from a semiarid streambed (M. expectata) population, while the 5,6,8-trisubstituted in- suggests that putative alkaloid-containing dolizidine 223A and the 1,4-disubstituted prey with such alkaloids can occupy a wide quinolizidine 223J were present as minor range of microhabitat niches. components in extract from the other bamboo grove population. A 3,5-disubstituted indolizidine 249A was a minor component in Mantella laevigata extract from the island population. Methuen and Hewitt, 1913 Mantella laevigata is an arboreal species Figure 3D that breeds in tree holes (Glaw and Vences, This species occurs in the northeastern 1992b). Most of the specimens seen by us coastal region and is the only Mantella treat- were 1 to 2 meters off the ground, but a few ed in this paper that is semiarboreal in habit, were on the ground. If its alkaloids derive the others being entirely terrestrial. The rel- from its food, then only the inland bamboo atively large finger discs (fig. 3D), emphasized grove sites appear to provide prey with de- in the original description, are correlated with cahydroquinoline and related alkaloids. Many its scansorial abilities. Glaw and Vences alkaloids were shared between the M. laevi- (1994: 180, color photo 48) indicated the ex- gata and M. betsileo collected together at and istence of a similarly colored, possibly un- around a bamboo grove, but the profiles were named, species (at Marojezy) that occurs different (compare fig. 6D and 7E). sympatrically with M. laevigata but lacks the enlarged finger discs. Mantella pulchra Parker, 1925 The anterior dorsal surfaces ofM. laevigata Figures 4A, B, 5 are green or yellowish green, in marked contrast to the rest of the black body and limbs, This species has a relatively wide distri- except the rear of the thigh, which was pale bution in eastern Madagascar, where it seems blue in the specimen illustrated (fig. 3D); pale to occur from an elevation of about 200 m labial and canthal lines are lacking. The ven- up to about 1000 m, becoming sympatric at ter and undersides of the limbs (including some upland sites with M. baroni. palms and soles) were overall black with pale Mantella pulchra is a bronzy brown or 1 996 DALY ET AL.: MADAGASCAN POISON FROGS 23 blackish frog with an ill-defined bronze can- abundant in stream-side areas that were thal-supraocular line, no labial line, a large somewhat boggy. M. pulchra was also found green lateral blotch running onto the upper about 50 m up the forested slopes in another arm, and a smaller green inguinal blotch con- boggy area. Another very localized popula- fluent with an elongated green blotch cov- tion of M. pulchra was sampled in a boggy ering most ofthe upper thigh; there is a vivid area ofridge forest at Ambavala, in the Man- red or orange flash mark'0 on the concealed anara Reserve near Sandrakatsy, about 300 part ofthe thigh and calf. There may be a few km northeast of An'Ala. Additional speci- dark red spots underneath the thigh, but oth- mens from a dealer were said to have come erwise the ventral surfaces (including palms from Anosibe An'Ala, some 70 km southwest and soles) are black overall, nearly uniform of An'Ala. or with blue markings overall, including blue Levels of alkaloids varied in three popu- edging around the lower lip. The coloration lation samples of skin ofM. pulchra (fig. 9A, ofM. pulchra therefore is similar to the larger B, C), being quite high in the An'Ala sample M. baroni, but the two are immediately dis- (fig. 9C) and relatively low in two samples tinguished by hind-limb pattern-a distinct from Ambavala (9A, B). Profiles also differed flash mark in pulchra vs. the uniform orange considerably in the three samples. The pop- and black coloring in baroni (fig. 5). There is ulation of M. pulchra from An'Ala shared a variant in the commercial trade having blue many alkaloids with the microsympatric M. rather than green blotches (Glaw and Vences, baroni, including pumiliotoxins and allopu- 1994: color fig. 63). miliotoxins, 1 ,4-disubstituted quinolizidines Busse (1981: 29) and Blommers-Schl6sser 217A and 231A, and the putative deoxypu- and Blanc (1991) considered pulchra a syn- miliotoxin 291E. It differed from microsym- onym of M. "madagascariensis." Glaw and patric baroni in having an "isopumiliotoxin" Vences (1 992a: 164-165) tentatively treated 267H, two 3,5-disubstituted pyrrolizidines it as a color morph of "madagascariensis," 223M and 2510, and an alkaloid 239N even though they described call differences (C17H21N) ofunknown structure. Skins ofM. in sympatry (at Andasibe). In the same year, pulchra from a swampy ridge forest near Am- Andreone (1992) confirmed call and other bavala contained decahydroquinoline 195A differences in sympatry at the same general and unsaturated analogs and "dimers" locality (Perinet forest) and recognized that 384A/384B, pumiliotoxin 267C and the 3,5- two taxa were involved-M. "madagascar- disubstituted pyrrolizidine 267H (fig. 9A, B). iensis" (baroni) and another taxon for which The higher molecular-weight pumiliotoxins he used the name Mantella cowani pulchra. present as major alkaloids in the An'Ala pop- M. pulchra is recognized as a species in sub- ulation were absent in the Ambavala popu- sequent literature. lations. The reputed "Anosibe An'Ala" sam- We here report microsympatry of M. pul- ple (from a dealer) had very low levels of chra and the larger M. baroni along a me- alkaloids (gas chromatograph not shown). The andering forest stream near An'Ala, at an el- two major alkaloids in that sample were the evation of about 900 m. Mantella baroni oc- 3,5-disubstituted pyrrolizidine 223M and the curred throughout the collection area along pumiliotoxin 251D. the stream, whereas M. pulchra was more Profiles in M. pulchra differ considerably, perhaps reflecting different habitats and diet. Based on limited observation, M. pulchra 10 There is a curious differential fading in the orange seems to prefer forest that is slightly swampy or red flash marks ofpreserved M. pulchra. For example, or boggy underfoot, in contrast to our sam- after 1 /2 years in preservative, the thigh/calf mark ofthe ples of M. baroni, which came from stream- specimen in figure 5 changed to yellowish white with vivid orange edging along the ventrolateral edge of the side forest. shank, and the dull reddish markings under the thighs Two previous reports on skin alkaloids in became a bright vivid orange. This explains Parker's M. pulchra were published under the names (1925) otherwise puzzling original description, in which Mantella madagascariensis (in Daly et al., the flash marking is described as a bicolor "brilliant 1984) and Mantella sp., cf. madagascariensis yellow" and "bright crimson." (in Garraffo et al., 1993b). The first report 24 AMERICAN MUSEUM NOVITATES NO. 3177 was based on an extract from a specimen of coast. Both streambeds were forested, but M. pulchra (then in the synonymy of mad- both, especially the more heavily forested in- agascariensis sensu Busse, 1981) obtained land site, were disturbed by clearings. from a commercial dealer in the early 1980s. High levels of pumiliotoxins 307A and Results from that single specimen (AMNH 323A were present in the Montagne des Fran- 114047) are not in accord with alkaloid pro- cais sample (fig. 9D). In addition, an isomer files from any wild population subsequently of pumiliotoxin 307A of unknown structure sampled, in that it contained a set ofhistrion- was present. There were trace amounts of an icotoxins (283A, 285A, 285C) and an appar- 8-deoxypumiliotoxin 289C and a homopu- ent acyclic alkaloid 241B-none ofwhich has miliotoxin 223G. The inland sample, in con- been detected even in trace amounts in any trast, contained a wide variety of alkaloids species ofMantella collected on four trips to (fig. 9E). Pumiliotoxins 307A and 323A, the Madagascar. Thus, we are cautious and be- former at high levels, were present along with lieve that the presence ofhistrionicotoxins or several other pumiliotoxins and high levels acyclic alkaloids in Mantella must be dis- ofhomopumiliotoxin 223G. There were also counted unless authenticated with data from several 5,8-disubstituted indolizidines, three field-collected specimens. It is conceivable ofwhich, namely 193E, 217B, and 2211, were that the specimen obtained from a commer- present at moderately high levels. A previ- cial dealer in the United States had been fed ously undetected alkaloid 207M (C14H25N) with New World arthropods containing his- ofunknown structure was present. It afforded trionicotoxins and other alkaloids. major mass spectral fragment ions at m/z 164 and 140. An earlier (1993) sample ofone skin Mantella viridis ofM. viridis, obtained from a dealer and stat- Pintak and B6hme, 1988 ed to be from Antsiranana, had very high Figure 4C levels ofpumiliotoxins 307A and 323A along This frog appears to be endemic to the with minor amounts of 5,8-disubstituted in- northern tip of Madagascar. Dorsal surfaces dolizidines 217B and 237H (fig. 9F). and flanks are greenish yellow, with or with- Prey that are the putative source of pum- out vague grayish lines. A black face mask is iliotoxins would appear to have been present sharply defined dorsolaterally to just past the at both 1994 sites of collection ofM. viridis, arm, where it falls offobliquely to the venter but prey containing "izidine" and other al- (fig. 4C); a pale labial line extends to the arm. kaloids would seem to be available only at The rear ofthe thigh is pale blue. The ventral the more forested, inland site. surfaces were overall black (including palms and soles) with pale blue markings, including SUMMARY OF ALKALOIDS a tendency for a horseshoe-shaped mark on the lowerjaw. In preserved specimens, at least, This survey ofalkaloids in skin extracts of the throat and chest are brown, sometimes 9 of 11 species currently recognized in Man- in obvious contrast to the black belly (as in tella reveals the presence of about 100 com- M. laevigata). pounds. The structures ofmany ofthese were Samples of skins of M. viridis were ob- known, having been previously isolated and tained in 1994 from two sites in the region structurally defined from New World den- of the type locality (south of Antsiranana). drobatid frogs (Daly et al., 1993). However, One was inland, from a nearly dry streambed about 30 alkaloids from Mantella represent 13 km south of Antsiranana. The other was new compounds of unknown or uncertain from a semiarid stream drainage area on structure. Characterization ofthese alkaloids Montagne des overlooking the will be presented elsewhere. The major class- Franqais,II es of alkaloids found in mantelline skin can 11 A report ofM. betsileo from Montagne des Francais be summarized as follows: (Blommers-Schlosser and Blanc, 1991: 371) could not 1. Pumiliotoxin class, which includes be confirmed and may be a case of misidentification. pumiliotoxins, allopumiliotoxins, homopu- Only M. viridis was found and was relatively common, miliotoxins; the structures ofmost ofthese are more than 40 individuals being seen in about four hours known. In addition, this class contains less well- of collecting. defined compounds, i.e., the putative dehydro- 1996 DALY ET AL.: MADAGASCAN POISON FROGS 25

Pumiliotoxins (PTX)

OH CH3 OH3 "5OH 267"OH

251D 267C

OH , C H 3OH3 ,H OH3 OH3 - ""|8tOH

307A 307F 307G

OH3 NOCHH3 OH 30'OH H 23 309A 323A

Allopumiliotoxins (aPTX)

Homopumiliotoxins (hPTX)

OH OH3 OH3 H"OH

223G 321 B

Fig. 10. Representative pumiliotoxin-class alkaloids from Mantella spp. homopumiliotoxins, the 8-deoxypumiliotoxins, 2. Decahydroquinolines and related the 9,10-dihydropumiliotoxins, and the "iso- compounds. These occur together in several pumiliotoxins," most ofwhich are as yet unique species and consist of a decahydroquinoline to Mantella. Compounds of the pumiliotoxin 195A, which is well known from dendrobatid class occur as major alkaloids in all nine species frogs, and, in addition, various unsaturated ofMantella sampled. congeners and the apparent "dimers" 384A 26 AMERICAN MUSEUM NOVITATES NO. 3177

Fig. 11. Representative dehydrohomopumiliotoxins, a 9,1 0-dihydropumiliotoxin and an 8-deoxypumiliotoxin from Mantella spp. The structures are tentative. and 384B ofunknown structure. This group- es, hitherto undetected during 30 years ofin- ing of related compounds has been found in vestigation of alkaloids from New World only three of the nine species studied (M. dendrobatid frogs. Conversely, certain "den- betsileo, M. pulchra, M. laevigata). drobatid alkaloids," in particular the ba- 3. "Izidine" alkaloids, which consist of trachotoxins, the histrionicotoxins, the ge- 3,5-disubstituted pyrrolizidines, 3,5-disub- phyrotoxins, and epibatidine, were not de- stituted indolizidines, 5,8-disubstituted in- tected in any of the recent mantelline ex- dolizidines and 1,4-disubstituted quinolizi- tracts. Three histrionicotoxins were present dines. These occur in varying amounts and as minor alkaloids in a single Mantella skin types in most populations ofthe nine species. obtained a decade ago from a dealer in the In M. baroni, indolizidines and quinolizi- United States (Daly et al., 1984). However, dines are major alkaloids often occurring at as concluded under M. pulchra, these alka- high levels. loid identifications-from a skin of a frog of In addition to these three general classes, unknown provenance and maintained for an various bicyclic and tricyclic alkaloids of un- unknown length oftime in the New World- known structure occur in Mantella. Some must be discounted unless verified from field- seem likely to represent new structural class- caught specimens. 1996 DALY ET AL.: MADAGASCAN POISON FROGS 27

Decahydroquinoline-Class (DHQ)

OH3 OH3

N N H H 193D cis-195A 3,5-Disubstituted Pyrrolizidines (Pyr)

267H'

3,5-Disubstituted Indolizidines (3,5-1)

249A 275C

5,8-Disubstituted Indolizidines (5,8-1)

OH3

NJ lNJ9 C9H160H (sec, Z) 217B 219F 243C 279D 1,4-Disubstituted Quinolizidines (Q) OH3

217A 231A

Fig. 12. Representative decahydroquinoline-class and "izidine"-class alkaloids from Mantella spp. The structure of the decahydroquinoline-class alkaloid 193D is tentative.

It is ofinterest that an extract ofM. baroni from small millipedes (Daly et al., 1 994b). from near Sahavondrona contained pyrroli- Other possible dietary sources for frog skin zidine oximes, that at least in dendrobatid alkaloids are ants, some of which do contain frogs are suspected to have a dietary origin 3,5-disubstituted pyrrolizidines and 3,5-di- 28 AMERICAN MUSEUM NOVITATES NO. 3177 substituted indolizidines (Jones and Blum, ACKNOWLEDGMENTS 1983) and beetles, some which contain tricyclic coccinellines (op. cit.). The only known We thank Drs. John E. Cadle, Darrel R. ant alkaloids that were detected in Mantella Frost, H. Martin Garraffo, and Richard G. were the pyrrolizidines 223H and 251K. No Zweifel for reading and commenting on the beetle alkaloids were detected. Thus, other manuscript, and Dr. Garraffo for help in many unknown dietary sources for most of the aspects of data analysis. Fieldwork in Mad- mantelline alkaloids must be considered. The agascar was supported by a grant from the present study demonstrating pumiliotoxin- Ministry of Education, Science and Culture class alkaloids as major components in al- ofJapan, and we are indebted to Drs. Terumi kaloid fractions from nine species of Man- Nakajima and Yoichiro Kuroda for allowing tella suggests that, if the diet is the source, J.W.D. to participate in a joint Madagascan- prey containing these alkaloids must be wide- Japanese program on biologically active nat- ly distributed in forest habitats ranging from ural products. semiarid lowland to swampy upland.

APPENDIX 1: VOUCHER SPECIMENS The following Madagascan specimens are na road ca. 12 km (airline) SE Andasibe, along vouchers only for specimens relevant to this sur- stream 0.5 km S ofroad, AMNH 133656-133664; vey of skin alkaloids in Mantella. Brief color de- 11 km WSW Ranomafana, 2 km SE Sahavon- scriptions are given in the text and representative drona, ca. 1000 m (21°16'S, 47°22'E), AMNH specimens were photographed at the American 136898-136900; Parc Ranomafana, ca. 10 km Museum (see figs. 2-5). All specimens are alcohol (airline) SW Ranomafana (town on Fianarantsoa- preserved in the Amphibian Collection of the Ifanadiana road), AMNH 133667. American Museum of Natural History. At least one to several specimens in each series are stan- Mantella betsileo dard museum specimens, the others being carcasses of some of the frogs skinned for alkaloid Antanambaobe, Mananara Avaratra National extraction. The carcasses permit easy access to an- Park, ca. 100-200 m (ca. 16°16'S, 49°40'E), AMNH atomical characters and some will be useful for 136940-136942; Farakaraina Forest Station, ca. cleared and stained preparations. 10 km E Maroantsetra (on coast), 30 m (15°26'S, 49°52'E), AMNH 140566-140570; ca. 7 km SE Sandrakatsy, near village Ambavala, AMNH Mantella aurantiaca 140572-140573; ca. 8 km SE Sandrakatsy, near village Ambavala, Mananara Avaratra National No specific data (from dealers), AMNH 114042- Park, ca. 200 m (ca. 16°23'S, 49°44'E), AMNH 114046, 123693-123695; 14-18 km N Andasibe, 140571. ca. 1000 m (18°45'S, 48°25'E), AMNH 136889- 136892; region ofAndasibe, AMNH 136921; near Andasibe (Perinet), AMNH 133611-133641. Mantella cowanii No specific data (dealer claimed "Anosibe Mantella baroni An'Ala"), AMNH 140546-140550. 14-18 km N Andasibe, ca. 1000 m (18°45'S, 48025'E), AMNH 136888; near Andasibe (Peri- Mantella crocea net), AMNH 133668-133681; 7 km SE Andasibe, Near Andasibe (Perinet), AMNH 133642-133655; 1 1 km (by rd) E Andasibe, ca. 1000 m (18°58'S, 14-18 km N Andasibe, ca. 1000 m (18°45'S, 48028'E), AMNH 136887; ca. 2 km SW An'Ala, 48025'E), AMNH 136893-136897. 7-8 km SEAndasibe, ca. 1000 m (18057'S, 48°27'E), AMNH 136901-136902; 2 km SW An'Ala, Mantella AMNH 140555-140556; 30-35 km (airline) S expectata Moramanga, near km 45 on Moramanga-Anosibe MassifIsalo, ca. 10 km (by rd) SW Ranohira, 500 An'Ala road, along stream ca. 2 km W of road, m south [of] road, ca. 800 m (22°38'S, 45°22'E), AMNH 133665-133666; Moramanga-Toamasi- AMNH 136922-136939, 140617. 1996 DALY ET AL.: MADAGASCAN POISON FROGS 29

Mantella laevigata 140581; Montagne des Fran9ais, 8 km (by rd) southeast from Diego Suarez (=Antsiranana) Ambodimanga, Mananara Avaratra National Park, (12°19'S, 49°20'E), AMNH 140575-140580. ca. 100 m (ca. 16°22'S, 49°47'E), AMNH 136945- 136946; Nosy Mangabe, >100 m (15°30'S, Mantella species 49046'E), AMNH 140557-140562; ca. 7 km SE Sandrakatsy, near village Ambavala, AMNH About 8 km SE Sandrakatsy, near village Amba- 140563-140565. vala, Mananara Avaratra National Park, ca. 200 m (ca. 16°23'S, 49°44'E), AMNH 140574. See text discussion under Mantella betsileo re- Mantella pulchra garding this specimen. It apparently is an imma- No specific data (from dealer), AMNH 114047- ture female, 21 mm SVL. Color snapshots taken 114048; 2 km SW An'Ala, AMNH 140551- before formalin fixation show an orange-brown dorsum and limbs, with green at the upper arm 140552; ca. 2 km SW An'Ala, 7-8 km SE Anda- insertion and a large green patch covering the an- sibe, ca. 1000 m (18°57'S, 48°27'E), AMNH terior and most ofthe dorsal thigh surface. Ventral 136903-136916; ca. 8 km SE Sandrakatsy, near village Ambavala, Mananara Avaratra National surfaces were black overall, with bright blue edging around the lower Park, ca. 200 m (ca. 16°23'S, 49°44'E), AMNH lip and small, irregularly shaped blue spots on throat, venter, and limbs. There 140553-140554; ca. 8 km SE Sandrakatsy, village bright was a small golden yellow spot in the groin and a Ambavala, Mananara Avaratra National Park, ca. golden yellow flash mark be- 100-200 m (16°23'S, 49°44'E), AMNH 136943- sharply demarcated 136944; region of Anosibe An'Ala (from dealer) low the knee in the concealed part of the shank, (ca. 19°26'S, 48°12'E), AMNH 136918. these markings being white in preservative; the inguinal mark is small and the calf marking large, occupying the proximal 60 percent of the shank. Mantella viridis In lateral view (not visible in the color photographs), a pale (presumably green) anterior lateral No specific data (from dealer), AMNH 133682- blotch is confluent with the pale limb insertions, 133683; "region of Antsiranana" (from dealer, and a similar blotch posteriorly on the flanks is AMNH 136919-136920; 13 km S Diego Suarez confluent with the large (green) anterodorsal thigh (=Antsiranana) (12°12'S, 4901 6'E), AMNH patch. 30 AMERICAN MUSEUM NOVITATES NO. 3177

APPENDIX 2: ALKALOIDS IDENTIFIED IN MANTELLA SKIN

Flame-ionization gas chromatographic profiles are shown in figures 6-9. Structures of relatively common Mantella alkaloids are shown in figures 10-12. Abbreviations ofalkaloid classes are as follows: Pumiliotoxins (PTX), allopumiliotoxins (aPTX), homopumiliotoxins (hPTX), dehydrohomopumiliotoxins (dehydrohPTX), 6,1 0-dihydropumiliotoxins (dihydroPTX), 8-deoxypumiliotoxins (deoxy- PTX), decahydroquinolines and related alkaloids (DHQ), 3,5-disubstituted pyrrolizidines (3,5-P), 3,5- disubstituted indolizidines (3,5-I), 5,8-disubstituted indolizidines (5,8-I), 5,6,8-trisubstituted indolizidines (5,6,8-I), 1,4-disubstituted quinolizidines (Q), 2,5-disubstituted pyrrolidines (Pyr), tricyclics (Tri), pyrrolizidine oximes (Oximes), unclassified (Unclass). Classification and structures for some alkaloids remains tentative. Some alkaloids exist as diastereomers, i.e., PTX 307F often occurs in Mantella as a mixture of diastereomeric 307F' and 307F" as do PTX 307G and 3,5-P 267H (see Garraffo et al., 1993b), but only the gross designation, i.e., 307F, 307G, 267H, followed by the number of diastereomers in parentheses, is given here. An asterisk (*) means that a given sample has been previously described (in Garraffo et al., 1993b). Number designations for each population (6A, 6B, etc.) refer to figures 6- 9 except for 8B'(see Garraffo et al., 1993b, fig. 2B for this sample).

Species Population Major alkaloids Date Minor alkaloids No. skins Trace alkaloids M. aurantiaca 6A. N Andasibe PTX 323A. Nov. 1989 PTX 267C, 307A, aPTX 323B; dehydrohPTX 233F, 235C. 10 skins* PTX 305B, 307B; dehydrohPTX 221F, 265F. 6B. N Andasibe PTX 323A. Jan. 1993 aPTX 323B; dehydrohPTX 265F; "isoPTX" 323C; Unclass 392, 434. 5 skins PTX 267C, 277B, 307G, 309A, hPTX 337; dehydrohPTX 233F, 235C; 3,5-I 249A; Q 217A. M. baroni 8A. N. Andasibe PTX 309A; Q 217A. Nov. 1989 PTX 307A, 307F(2), aPTX 325A; dihydroPTX 281F; 5,8-I 217B; 235B, 241F, 243C, 253B, 279D; Q 231A, 10 skins* 233A; Unclass 281F. PTX 251D, hPTX 249F; dehydrohPTX 233F, 235C; 5,8-I 243B, 245B. 8B. E Andasibe PTX 309A; Q 217A, 231A. Dec. 1993 PTX 307G, 307F(2); aPTX 325A; deoxyPTX 291E, 293D; dihydroPTX 281F; 5,8-I 217B; 10 skins Unclass 251P. PTX 237A, 251D; hPTX 323E, 337; 5,8-I 241F, 243C, 253B; Q 233A. 8B'. E Andasibe PTX 309A; dihydroPTX 281F; Q 217A, 231A. Nov. 1989 PTX 307A; aPTX 325A; 3,5-I 249A, 275C; 5,8-I 217B, 241F; Unclass 293B 10 skins* hPTX 235J; 5,8-I 243C, 279D; Q 273A. 8C. S Moramanga Q 217A, 231A. Nov. 1989 PTX 307A, 309A; aPTX 325A; 3,5-I 249A; 5,8-I 217B; 5,6,8-I 223A. 10 skins* PTX 251D, 307F; 3,5-I 275C; 5,8-I 205A, 243D, 245C; Unclass 251L, 293G. 8D. An'Ala PTX 309A; Q 217A, 231A. Dec. 1993 PTX 307F(2), 323A; deoxyPTX 291E; dihydroPTX 281F; 3,5-I 275C; 5,8-I 217B, 243C; Q 273A. 3 skins PTX 237A, 307A; aPTX 325A; hPTX 323E; deoxyPTX 293D; 3,5-I 249A; 5,84I 241F, 279D; Unclass 275D. 8E. Ramomnafana PTX 309A; Q 217A. Nov. 1989 PTX 237A, 307F(2); 5,8-I 217B; 243D, 245C; Q 231A, 233A; Tri 207J; Unclass 293B. 10 skins* PTX 251D; 5,8-I 203A; Unclass 205C.

8F. Sahavondrana PTX 309A. Jan. 1993 PTX 251D, 307F; aPTX 325A; hPTX 251R, 265N; dehydrohPTX 265F; deoxyPTX 281G; 17 skins 5,8-I 217B; Q 231A, 273A(2). PTX 237A, 307A; 3,5-I 271F, 275C; 5,8-I 241F; Oxime 236, 252A. M. betsileo 6C. Antanambaobe DHQ 384A, 384B; Unclass 195C. Dec. 1990 PTX 307G(2); aPTX 323B; DHQ 195A; 5,84I 219F. 11 skins* PTX 251D, 307A; aPTX 321C, hPTX 223G, 317; DHQ 189, 193D; 5,8-I 217B; Q 217A, 231A, 249C; Tri 207J, 235K; Unclass 211D. 6D. Ambavala DHQ 195A; Unclass 195C. Jan. 1994 PTX 251D; DHQ 193D, 384A, 384B; 3,5-I 247C; 5,8-I 217B; Q 217A; Unclass 211D 8 skins PIX 237A, 307G; aPTX 323B; hPTX 321B, 337; DHQ 382, 3,5-I 205A, 247C; 5,84I 251B, 253B; Q 249H, Tri 235M, 265M, Unclass 251Q, 269E, 271B. 1996 DALY ET AL.: MADAGASCAN POISON FROGS 31

6E. Farakaraina PTX 307G; 3,5-P 223H. Dec. 1993 PTX 251D; hPTX 223G; DHQ 195A; 3,5-P 251K; 5,84I 219F; Q 231A; Unclass 195C. 12 skins PTX 237A, 267C, 293E; aPTX 323B; hPTX 321B, 335; deoxyPTX 291E; dehydrohPTX 265F; DHQ 181D, 189, 193D, 382, 384A, 384B; 3,54I 247C; Tri 265M; Unclass 237K, 269E. M. cowanii 6F. PTX 251D. Dec. 1993 PTX 265G, 307G(2), aPTX 7-epi-267A. 3 skins PTX 267C(2), 323A.

M. crocea 7A. N. Andasibe PTX 267C, 307A, 323A, aPTX 323B. Nov. 1989 dehydrohPTX 233F, 235C; 5,8-I 295B; Unclass 392, 434. 10 skins* PTX 265G, 305B, 307B; DHQ 195A; 3,5-P 223H(2); Q 231A.

7B. N. Andasibe PTX 267C, 307A, 323A; aPTX 323B. Jan. 1993 PTX 265G; Unclass 392, 434. 35 skins PTX 277B, 305B, 309A; dehydrohPTX 233F, 235C; "isoPTX" 267N, 323C; 3,5-P 2390; Unclass 293G.

M. expectata 7C. Massif Isalo PTX 251D; 5,8-I 219F; Unclass 261C. Jan. 1993 PTX 307A; 5,8^1193E; Unclass 323D. 20 skins PTX 267C(2), 305B; aPTX 323B; 3,5-I 223AB, 275C; 5,8-I 197C, 217B, 219J, 241F; Unclass 195C, 233I, 2350.

M. laevigata 7D. Ambodimanga Unclass 195C. Dec. 1990 PTX 307G, DHQ 193D, 195A. 6 skins* hPTX 223G; DHQ 189, 384A, 384B; Q 207I; Unclass 161, 211C, 211D, 223A.

7E. Varary PTX 307G. Jan. 1994 DHQ 193D, 195A; 5,8-I 223J; 5,6,8-I 223A; Unclass 195C. 5 skins hPTX 321B; deoxyPTX 291E; DHQ 189, 382, 384A, 384B, Unclass 275D.

7F. Nosy Mangabe PTX 307G. Dec. 1993 3,5-1 249A. 6 skins PTX 237A, 305B; deoxyPTX 291E; 3,5-P 223H; 3,5-I 247C, 275C; 5,8-I 219F; Q 217A; Pyr 223N, 225H.

M. pulchra 9A. Ambavala 3,5-P 267H(2). Dec. 1990 PTX 267C; DHQ 384A, 384B; Unclass 161, 293B. 6 skins* PTX 265G; aPTX 325A; 3,5-P 239K(2), 265H(2); DHQ 189; 193D, 195A, Unclass 195C, 211D.

9B. Ambavala PTX 267C. Jan. 1994 PTX 265G; DHQ 189, 193D(2), 382, 384A, 384B; 3,5-P 267H. S skins PTX 307A; DHQ 195A; 3,5-P 223H; Unclass 211D, 271B, 341C, 392, 434.

9C. An'Ala PTX 309A; Q 217A. Jan. 1993 PTX 251D, 307A, 307F, 325B; aPTX 325A; dihydroPTX 281F; 3,5-P 223M; 3,5-I 249A, 5 skins 275C; 5,8-1279D; Q 231A(2). PTX 237A, 307G; aPTX 323B; hPTX 337; deoxyPTX 291E, 293D; "isoPTX" 267N; 3,5-P 2510; 5,8-I 245B, 249J; Unclass 239N.

M. viridis 9D. Montagne des 8PTX 307A. Franlais PTX 307D, 323A. Jan. 1994 PTX 277B, 289C, 305D, 323F. 30 skins

9E. S. Antsiranana PIX 307A; hPTX 223G. Jan. 1994 PTX 289C, 323A; 5,8-I 193E, 217B(2), 2211, 237J, Unclass 207M, 2230, 235L(2), 265K, 267M. 5 skins PTX 209F, 305D; hPTX 239M; "isoPTX" 339C; 3,5-I 223AB; 5,8-I 231C, 245F; Q 207I.

9F. Antsiranana PTX 307A. Nov. 1989 PTX 307G, 321A (artefact), 323A; 5,8-I 217B, 237H. 1 skin* PTX 265G, 309B, 307H; hPTX 223G; 3,5-I 223AB; Q 217A, 231A, 233A.

REFERENCES Andreone, Franco de Madagascar. Ecologie et developpe- 1992. Syntopy of Mantella cowani Boulenger ment de Mantella aurantiaca Moc- and Mantella madagascariensis (Gran- quard, 1900. Bull. Mus. Natl. Hist. Nat. didier) in central-eastern Madagascar, Paris, ser. 2, 37: 931-940. with notes on the colouration in the ge- Blommers-Schlosser, Rose M. A. nus Mantella (Anura: Mantellidae). Boll. 1978. Cytotaxonomy of the Ranidae, Rhaco- Mus. Regionale Sci. Nat. Torino 10(2): phoridae, Hyperoliidae (Anura) from 421-450. Madagascar with a note on the karyo- Arnoult, J. type of two amphibians of the Sey- 1966. Contribution a l'etude des batrachiens chelles. Genetica 48(1): 23-40. 32 AMERICAN MUSEUM NOVITATES NO. 3177

1979. Biosystematics ofthe Malagasy frogs. I. dae), Australia (Myobatrachidae) and Mantellinae (Ranidae). Beaufortia Madagascar (Mantellinae). Toxicon 29(352): 1-77. 22(6): 905-919. 1993. Systematic relationships ofthe Mantel- Daly, John W., Charles W. Myers, and Noel Whit- linae Laurent 1946 (Anura Ranoidea). taker Ethol. Ecol. Evol. 5: 199-218. 1987. Further classification of skin alkaloids Blommers-Schlosser, R. M. A., and Ch. P. Blanc from Neotropical poison frogs (Dendro- 1991. Amphibiens (premiere partie). Faune de batidae), with a general survey of tox- Madagascar 75(1): 379 pp. + 12 pls. ic/noxious substances in the Amphibia. Boettger, Oskar Toxicon 25(10): 1023-1095. 1880. Diagnoses reptilium et batrachiorum Daly, John W., H. Martin Garraffo, L. K. Pannell, novorum a Carolo Ebenau in insula Thomas F. Spande, C. Severini, and V. Nossi-Be Madagascariensi lectorum. Erspamer Zool. Anz. 3(57): 279-283. 1990. Alkaloids from Australian frogs (Myob- Bohme, Wolfgang, Klaus Busse, and Frank Glaw atrachidae): pseudophrynamines and 1993. The identity of Mantella cowani Bou- pumiliotoxins. J. Nat. Prod. 53(2): 407- lenger, 1882 and Mantella haraldmeieri 421. Busse, 1981 (Anura: Mantellinae). Am- Daly, John W., Sherrie I. Secunda, H. Martin Gar- phib.-Reptilia 14: 269-273. raffo, Thomas F. Spande, Anthony Wis- Boulenger, George Albert nieski, Charles Nishihira, and John F. 1882. Catalogue of the Batrachia Salientia s. Cover, Jr. Ecaudata in the collection ofthe British 1992. Variability in alkaloid profiles in Neo- Museum. London, British Museum, xvi tropical poison frogs (Dendrobatidae): + 503 pp. + 30 pls. genetic versus environmental determi- 1888. Descriptions of new reptiles and batra- nants. Toxicon 30: 887-898. chians from Madagascar. Ann. Mag. Daly, John W., H. Martin Garraffo, and Thomas Nat. Hist., ser. 6, 1(2): 101-107 + pls. F. Spande 5-6. 1993. Amphibian alkaloids. In G. A. Cordell Busse, Klaus (ed.), The alkaloids, vol. 43, pp. 185- 1981. Revision der Farmuster-Variabilitait in 288. San Diego, Ca.: Press. der madagassischen Gattung Mantella Academic (Salientia: Ranidae). Amphib.-Reptilia Daly, John W., Sherrie I. Secunda, H. Martin Gar- 2: 25-42. raffo, Thomas F. Spande, Anthony Wis- Busse, Klaus, and Wolfgang Bohme nieski, and John F. Cover, Jr. 1992. Two remarkable frog discoveries of the 1994a. An uptake system for dietary alkaloids genera Mantella (Ranidae: Mantellinae) in poison frogs (Dendrobatidae). Toxi- and Scaphiophryne (Microhylidae: Sca- con 32(6): 657-663. phiophryninae) from the west coast of Daly, John W., H. Martin Garraffo, Cesar Jar- Madagascar. Rev. Francaise d'Aquar- amillo, and A. Stanley Rand iol. 19(1/2): 57-64. 1994b. Dietary source for skin alkaloids ofpoi- Channing, A. son frogs (Dendrobatidae)? J. Chem. 1989. A re-evaluation ofthe phylogeny ofOld Ecol. 20(4): 943-955. World treefrogs. S. Afr. J. Zool. 24(2): Dubois, Alain 116-131. 1984. La nomenclature supragenerique des Daly, John W., George B. Brown, Monica Men- amphibiens anoures. Mem. Mus. Natl. sah-Dwumah, and Charles W. Myers d'Hist. Nat., ser. A, Zool., 131: 1-64. 1978. Classification of skin alkaloids from Dumbacher, John P., Bruce M. Beehler, Thomas Neotropical poison-dart frogs (Dendro- F. Spande, H. Martin Garraffo, and John batidae). Toxicon 16(10): 163-188. W. Daly Daly, John W., Charles W. Myers, Jordan E. War- 1992. Homobatrachotoxin in the genus Pito- nick, and Edson X. Albuquerque hui: chemical defense in birds. Science 1980. Levels ofbatrachotoxin and lack ofsen- 258: 799-801. sitivity to its action in poison-dart frogs Ford, Linda S. (Phyllobates). Science 208: 1383-1385. 1993. The phylogenetic position of the dart- Daly, John W., Robert J. Highet, and Charles W. poison frogs (Dendrobatidae) among Myers anurans: an examination of the com- 1984. Occurrence of skin alkaloids in non- peting hypotheses and their characters. dendrobatid frogs from Brazil (Bufoni- Ethol. Ecol. Evol. 5: 219-231. 1996 DALY ET AL.: MADAGASCAN POISON FROGS 33

Frost, Darrel R. deoxypumiliotoxin 251F. J. Nat. Prod. 1985. Amphibian species of the world. Law- 58(1): 100-104. rence, KS: Allen Press and Assoc. Syst. Jones, Tappey H., and Murray S. Blum Coll., vii + 732 pp. 1983. Arthropod alkaloids: distribution, func- Garraffo, H. Martin, Thomas F. Spande, John W. tions, and chemistry. In S. William Pel- Daly, A. Baldessari, and Eduardo G. letier (ed.), Alkaloids: chemical and bi- Gros ological perspectives. vol. 1, pp. 33-84. 1993a. Alkaloids from bufonid toads (Melan- New York: Wiley. ophryniscus): decahydroquinolines, Kuchling, Gerald pumiliotoxins and homopumiliotoxins, 1993. Zur Verbreitung und Fortpflanzung von indolizidines, pyrrolizidines, and quin- Mantella betsileo in Westmadagaskar. olizidines. J. Nat. Prod. 56(3): 357-373. Salamandra 29(3/4): 273-276. Garraffo, H. Martin, Janet Caceres, John W. Daly, Labatut, F., and R. Raharinarivonirina Thomas F. Spande, N.R. Andriama- 1969. Madagascar. Etud historique. Paris: haravo, and M. Andriantsiferana Editions Fernand Nathan, 224 pp. 1993b. Alkaloids in Madagascan frogs (Man- Laurent, Raymond F. tella): pumiliotoxins, indolizidines, 1943a. Sur la position systematique et l'osteo- quinolizidines, and pyrrolizidines. J. logie du genre Mantidactylus Boulenger. Nat. Prod. 56(7): 1016-1038. Bull. Mus. R. Hist. Nat. Belgique 19(5): Glaw, Frank, and Miguel Vences 8 pp. + 1 pl. 1992a. A fieldguide to the amphibians and rep- 1943b. Contribution a l'osteologie et a la sys- tiles ofMadagascar. Cologne, Germany: tematique des rhacophorides non Af- privately printed, 331 pp. + 16 color ricains. Ibid. 19(28): 16 pp. + 2 pls. pls. 1946. Mises au point dans la taxonomie des 1992b. Zur Biologie, Biometrie und Fiirbung bei ranides. Rev. Zool. Bot. Afr. 39(4): 336- Mantella laevigata Methuen and Hew- 338. itt, 1913. Sauria (Berlin) 14(4): 25-29. 1951. Sur la necessite de supprimer la famille 1994. A fieldguide to the amphibians and rep- des Rhacophoridae mais de creer celle tiles ofMadagascar. Second edition, in- des Hyperoliidae. Ibid. 45(1): 116-122. cluding mammals and freshwater fish. Liem, Sioe Sing [David S.] Cologne, Germany: privately printed, 1970. The morphology, systematics, and evo- 480 pp. + 48 color pls. lution of the Old World treefrogs (Rha- Grandidier, Alfred cophoridae and Hyperoliidae). Fieldi- 1872. Description de quelques reptiles nou- ana, Zool. 57: vii + 145 pp. veaux decouverts a Madagascar en 1870. Maxson, Linda R., and Charles W. Myers Ann. Sci. Nat. Zool. (Paris), ser. 5, 1985. Albumin evolution in tropical poison 15(20): 6-11. frogs (Dendrobatidae): a preliminary re- 1893. Les voyageurs francais a Madagascar port. Biotropica 17(1): 50-56. pendant les trente demieres annees. Bull. Meier, Harald Soc. Geogr., ser. 7, 14: 289-300 + 4 1986. Neues fur den Terrarianer iiber die Gat- foldout maps. tung Mantella auf Madagaskar. Her- Guibe, Jean petofauna 8(41): 9-14. 1964. Revison des especes du genre Mantella Methuen, Paul A., and John Hewitt (Amphibia, Ranidae). Senckenb. Biol. 1913. On a collection ofBatrachia from Mad- 45(3/5): 259-264. agascar made during the year 1911. Ann. 1978. Les batraciens de Madagascar. Bonner Transvaal Mus. 4(2): 49-64 + pls. 9- Zool. Monogr. 11: 144 pp. + 82 pls. 10. Hewitt, John Mocquard, Francois 1911. A comparative review ofthe amphibian 1900. Nouvelle contribution a la faune her- faunas ofSouth Africa and Madagascar, petologique de Madagascar. Bull. Soc. with some suggestions regarding their Philom. Paris, ser 9, 2(4): 93-111 + pl. former lines of dispersal. Ann. Trans- 2. vaal Mus. 3(1): 29-39. 1909. Synopsis des familles, genres et especes Jain, Poonan, H. Martin Garraffo, Thomas F. des reptiles ecailleux et des batraciens Spande, Herman J. C. Yeh, and John de Madagascar. Nouv. Archiv. Mus. W. Daly Hist. Nat., Paris, ser. 5, 1: 1-110. 1995. A new subclass ofalkaloids from a den- Myers, Charles W., John W. Daly, and Victor drobatid poison frog: a structure for Martinez 34 AMERICAN MUSEUM NOVITATES NO. 3177

1984. An arboreal poison frog (Dendrobates) Roux, Jean from western Panama. Am. Mus. Nov- 1935. Sur un nouveau batracien de Madagas- itates 2783: 20 pp. car (Mantella loppei n.sp.). Bull. Soc. Myers, Charles W., Alfredo Paolillo O., and John Zool. France 60: 441-443. W. Daly Schopf, Clemens 1991. Discovery of a defensively malodorous 1961. Die Konstitution der Salamander-Al- and nocturnal frog in the family Den- kaloide. Experientia (Basel) 17(7): 285- drobatidae: phylogenetic significance of 295. a new genus and species from the Ven- Shaw, George A. ezuelan Andes. Am. Mus. Novitates 1875. The Betsileo: country and people. An- 3002: 33 pp. tananarivo Ann. and Madagascar Mag. Myers, Charles W., John W. Daly, H. Martin Gar- 1(4): 334-347. raffo, Anthony Wisnieski, and John F. Sibree, James Cover, Jr. 1880. The great African island. Chapters on 1995. Discovery ofthe Costa Rican poison frog Madagascar. London: Triibner and Co., Dendrobates granuliferus in sympatry Ludgate Hill, xii + 372 pp. + frontisp., with Dendrobates pumilio, and com- 3 pls., 2 foldout maps. ments on taxonomic use of skin alka- Vences, Miguel, Frank Glaw, Andre Peyrieras, loids Am. Mus. Novitates 3144: 21 pp. Wolfgang Bohme, and Klaus Busse Noble, Gladwyn Kingsley 1994. Der Mantella-madagascariensis-Kom- 1931. The biology ofthe Amphibia. New York: plex: Wiederentdeckung von Mantella McGraw-Hill, xiii + 577 pp. cowani und Beschreibung von Mantella Noble, Gladwyn Kingsley, and H. W. Parker bernhardi n. sp. Datz Aquar. Terrar. 47: 1926. A synopsis of the brevicipitid toads of 390-393. Madagascar. Am. Mus. Novitates 232: Werner, Franz 21 pp. 1901. Beschreibung neuer Dendrobatiden. Mit Parker, H. W. einer Revision dieser Batrachier-Fam- 1925. New and rare reptiles and batrachians ilie. Verhandl. K. K. Zool.-Bot. Gesell., from Madagascar. Ann. Mag. Nat. Hist. Vienna, for 1901, pp. 627-634. ser. 9, 16: 390-394. Zimmermann, Helmut, Elke Zimmermann, and Pintak, Thomas, and Wolfgang Bohme Peter Zimmermann 1988. Mantella viridis sp. n. (Anura: Ranidae: 1990. Feldstudien im Biotop vom Gold- Mantellinae) aus Nord-Madagaskar. fr6schchen, Mantella aurantiaca, im Salamandra 24(2/3): 119-124. tropischen Regenwalds Ost-Madagas- 1990. Mantella crocea sp. n. (Anura: Ranidae: kars. Herpetofauna 12(64): 21-24. Mantellinae) aus dem mittleren Ost- Madagaskar. Ibid. 26(1): 58-62.

Recent issues of the Novitates may be purchased from the Museum. Lists of back issues of the Novitates, Bulletin, and Anthropological Papers published during the last five years are available free of charge. Address orders to: American Museum ofNatural History Library, Department D, Central Park West at 79th St., New York, N.Y. 10024. TEL: (212) 769-5545. FAX: (212) 769- 5009. E-MAIL: [email protected]

L-

® This paper meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper).