Abstract

Due in part to scarcity of material, no published study has yet cladistically addressed the systematics of living and fossil Tenrecidae (Mammalia, Afrotheria). Using a noninvasive technique for sampling nuclear DNA from museum specimens, we investigate the evolution of the Tenrecidae and assess the extent to which tenrecids fit patterns of relationships proposed for other terrestrial mammals on Madagascar. Application of several tree-reconstruction techniques on sequences of the nuclear growth hormone receptor gene and morphological data for all recognized tenrecid genera supports monophyly of Malagasy tenrecids to the exclusion of the two living African genera. However, both parsimony and Bayesian methods favor a close relationship between fossil African tenrecs and the Malagasy Geogale, supporting the hypothesis of island paraphyly, but not polyphyly. More generally, the noninvasive extraction technique can be applied with minimal risk to rare/unique specimens and, by better utilizing museum collections for genetic work, can greatly mitigate field expenses and disturbance of natural populations.

With the exception of a single genus of shrew (Suncus), insectivoran-grade mammals from Madagascar are members of the family Tenrecidae (Eisenberg and Gould, 1970; Olson and Goodman, 2003). This group of placental mammals consists of eight genera endemic to Madagascar and two from equatorial Africa and is remarkably diverse, occupying terrestrial, semiarboreal, fossorial, and semiaquatic niches. Other Malagasy groups are similarly diverse; previous morphological investigations of its primates (Cartmill, 1975; Yoder, 1992), carnivorans (Veron, 1995), and rodents (Ellerman, 1940, 1941), as well as its tenrecs (Butler, 1984; Asher, 1999, 2000), have indicated multiple sister-group relationships with mainland taxa within each group.

Given the absence of modern taxa from the Malagasy Cretaceous and the isolation of Madagascar from other landmasses over the past ca. 80 to 90 million years (Krause, 2003), dispersal has become the primary hypothesis for explaining the arrival of many of Madagascar's inhabitants (cf. Raxworthy et al., 2002; Zakharov et al., 2004). Phylogeny can further illustrate the biogeographic history of a given group. Monophyly (Fig. 1A) and paraphyly (Fig. 1B) of island taxa are compatible with a single dispersal event leading to island colonization, whereas polyphyly (Fig. 1C) implies multiple colonization events.

Figure 1

Biogeographic implications of (A) monophyly, consistent with a single dispersal event to colonize Madagascar (cf. Eisenberg, 1975; Olson and Goodman, 2003); (B) paraphyly, consistent with a single dispersal event coupled with limited back-migration from Madagascar to Africa (cf. Butler, 1985); and (C) polyphyly, consistent with multiple dispersal events between Africa and Madagascar (cf. Asher 2000: fig. R1-12). Dotted lines in B and C indicate uncertainty in the positions of Erythrozootes and Protenrec.

Figure 1

Biogeographic implications of (A) monophyly, consistent with a single dispersal event to colonize Madagascar (cf. Eisenberg, 1975; Olson and Goodman, 2003); (B) paraphyly, consistent with a single dispersal event coupled with limited back-migration from Madagascar to Africa (cf. Butler, 1985); and (C) polyphyly, consistent with multiple dispersal events between Africa and Madagascar (cf. Asher 2000: fig. R1-12). Dotted lines in B and C indicate uncertainty in the positions of Erythrozootes and Protenrec.

The aforementioned morphological studies noting the diversity of Malagasy mammalian groups have to varying degrees implied island polyphyly (Fig. 1C); i.e., that each of the modern groups has undergone multiple dispersal events across water barriers in order to colonize the island. In contrast, recent molecular phylogenies of terrestrial Malagasy mammals have supported island monophyly (Fig. 1A) for living primates (Yoder et al., 1996), carnivorans (Flynn et al., 2005), tenrecs (Olson and Goodman, 2003), and possibly rodents (Jansa and Weksler, 2004; Steppan et al., 2004; see discussion below).

Many tenrecid species are rare and/or endangered (Vogel, 1983; Benstead and Olson, 2003) and are difficult to obtain for research purposes. For example, the semiaquatic Limnogale mergulus is known from barely over a dozen museum specimens in Europe and North America. Destructive sampling of such material (e.g., for DNA sequencing) is generally not possible. Because it is so difficult to obtain tissues, most molecular studies sampling this group (e.g., Emerson et al., 1999; Mouchaty et al., 2000; Douady et al., 2002; Malia et al., 2002) have included between one and five of the over two dozen species. Olson and Goodman (2003) described a much better sample and were the first to publish a study with representatives of all living tenrecid genera, including sequences from one nuclear (vWF) and three mitochondrial (12S, tRNA-Valine, ND2) genes. However, as of this writing (August 2005), their DNA sequences and alignments are unavailable from public sources (e.g., GenBank). No published study has yet cladistically analyzed the three recognized fossil tenrecids, Erythrozootes, Protenrec, and Parageogale (Butler, 1984; McKenna and Bell, 1997). Jacobs et al. (1987) named a fourth fossil genus, Ndamathaia. However, we follow Morales et al. (2000) in regarding this taxon as a non-tenrecid.

In this article, we provide new DNA sequence data from the nuclear growth hormone receptor (GHR) gene using a noninvasive procedure applied to museum specimens. We also include a morphological data set, enabling us to sample all recognized living and extinct tenrecid genera. To reconstruct phylogenetic trees, we apply both maximum parsimony (MP) and a Markov k (Mk) model (Lewis, 2001) in a Bayesian framework (Nylander et al., 2004). Using these data we estimate the fit of living and fossil tenrecs to phylogenetic and biogeographic patterns proposed for other Malagasy groups.

Materials and Methods

The Noninvasive Extraction Method

We obtained between 756 and 855 base pairs from exon 10 of the growth hormone receptor (GHR) gene from crania accessioned at the Zoologisches Museum Berlin (ZMB), Harvard Museum of Comparative Zoology (MCZ), and the Department of Ecology and Evolution, University of Lausanne (IZEA). Specifically, we used skulls of Hemicentetes semispinosus (ZMB 71599), Limnogale mergulus (ZMB 35258; Fig. 2), Potamogale velox (ZMB 46588), Setifer setosus (ZMB 44586; Fig. 2), Geogale aurita (MCZ 45044), and Micropotamogale lamottei (IZEA 4975; Fig. 2). New sequences were aligned with previously published GHR sequences (Malia et al., 2002; Adkins et al., 2001; Pantel et al., 2000; van Garderen et al., 1999; Zogopoulos et al., 1999; Wang et al., 1995; Adams et al., 1990; Baumbach et al., 1989; Smith et al., 1989; Leung et al., 1987). Table 1 shows GenBank accession numbers for extant taxa, including DQ202287 to DQ202292, for our new sequences.

Figure 2

Lateral view of crania in Micropotamogale (top, IZEA 4975), Limnogale (middle, ZMB 35258), and Setifer (bottom, ZMB 44586), used for noninvasive extraction of nuclear GHR sequences. Images were taken after DNA extraction. Boxes highlight patent lacrimal foramen in Setifer, and absence thereof in Micropotamogale and Limnogale.

Figure 2

Lateral view of crania in Micropotamogale (top, IZEA 4975), Limnogale (middle, ZMB 35258), and Setifer (bottom, ZMB 44586), used for noninvasive extraction of nuclear GHR sequences. Images were taken after DNA extraction. Boxes highlight patent lacrimal foramen in Setifer, and absence thereof in Micropotamogale and Limnogale.

Table 1

Taxon sample and accession numbers of taxa used in our sample of nuclear GHR sequences. Boldface indicates new GHR sequences; daggers indicate extinct taxa. For nomenclature we follow Nowak (1999) and Asher (2005).

High-level clade Supra-generic clade Genus Accession number 
Didelphimorphia Didelphidae Monodelphis AF238491 
Artiodactyla Bovidae Bos X70041 
 Capridae Ovis M82912 
 Suidae Sus X54429 
Carnivora Canidae Canis AF133835 
 Ursidae Ursus AF392879 
Chiroptera Phyllostomidae Artibeus AF392895 
 Pteropodidae Pteropus AF392893 
 Vespertilionidae Myotis AF392894 
Hyracoidea Procaviidae Procavia AF392896 
Lipotyphla Erinaceidae Echinosorex AF392887 
  Erinaceus AF392882 
 Soricidae Blarina AF392880 
  Crocidura AF392884 
  Sorex AF392881 
  Suncus AF392888 
 Talpidae Parascalops AF392883 
Macroscelidea Macroscelididae Elephantulus AF392876 
Perissodactyla Equidae Equus AF392878 
Primates Cercopithecidae Macaca U84589 
  Papio AF150751 
 Hominidae Homo X06562 
Proboscidea Elephantidae Elephas AF332013 
  Loxodonta AF332012 
Rodentia Geomyidae Geomys AF332028 
 Muridae Mus M33324 
  Rattus X16726 
Scandentia Tupaiidae Tupaia AF540643 
Sirenia Trichechidae Trichechus AF392891 
Tenrecoidea Chrysochloridae Chrysospalax AF392877 
 Geogalinae †Parageogale — 
  Geogale DQ202287 
 Oryzorictinae Limnogale DQ202289 
  Microgale AF392885 
  Oryzorictes AF392886 
 Potamogalinae Micropotamogale DQ202290 
  Potamogale DQ202291 
 Protenrecinae †Erythrozootes — 
  †Protenrec — 
 Tenrecinae Echinops AF392889 
  Hemicentetes DQ202288 
  Setifer DQ202292 
  Tenrec AF392890 
Tubulidentata Orycteropodidae Orycteropus AF392892 
Xenarthra Myrmecophagidae Myrmecophaga AF392875 
High-level clade Supra-generic clade Genus Accession number 
Didelphimorphia Didelphidae Monodelphis AF238491 
Artiodactyla Bovidae Bos X70041 
 Capridae Ovis M82912 
 Suidae Sus X54429 
Carnivora Canidae Canis AF133835 
 Ursidae Ursus AF392879 
Chiroptera Phyllostomidae Artibeus AF392895 
 Pteropodidae Pteropus AF392893 
 Vespertilionidae Myotis AF392894 
Hyracoidea Procaviidae Procavia AF392896 
Lipotyphla Erinaceidae Echinosorex AF392887 
  Erinaceus AF392882 
 Soricidae Blarina AF392880 
  Crocidura AF392884 
  Sorex AF392881 
  Suncus AF392888 
 Talpidae Parascalops AF392883 
Macroscelidea Macroscelididae Elephantulus AF392876 
Perissodactyla Equidae Equus AF392878 
Primates Cercopithecidae Macaca U84589 
  Papio AF150751 
 Hominidae Homo X06562 
Proboscidea Elephantidae Elephas AF332013 
  Loxodonta AF332012 
Rodentia Geomyidae Geomys AF332028 
 Muridae Mus M33324 
  Rattus X16726 
Scandentia Tupaiidae Tupaia AF540643 
Sirenia Trichechidae Trichechus AF392891 
Tenrecoidea Chrysochloridae Chrysospalax AF392877 
 Geogalinae †Parageogale — 
  Geogale DQ202287 
 Oryzorictinae Limnogale DQ202289 
  Microgale AF392885 
  Oryzorictes AF392886 
 Potamogalinae Micropotamogale DQ202290 
  Potamogale DQ202291 
 Protenrecinae †Erythrozootes — 
  †Protenrec — 
 Tenrecinae Echinops AF392889 
  Hemicentetes DQ202288 
  Setifer DQ202292 
  Tenrec AF392890 
Tubulidentata Orycteropodidae Orycteropus AF392892 
Xenarthra Myrmecophagidae Myrmecophaga AF392875 

Expanding upon the method of Rohland et al. (2004) for mitochondrial DNA, we obtained nuclear GHR sequences from museum crania, leaving the treated specimens completely intact. We incubated either lower jaws or rostra in 20 mL of a buffer containing 5 M guanidinium isothiocyanate, 50 mM Tris, pH 8.0, 25 mM NaCl, 1.3% Triton-X, 20 mM EDTA, and 50 mM DTT. To minimize the possibility of damage, we incubated the specimens at room temperature and rotated them in near-vertical tubes that permitted flow of the buffer but kept specimens stationary. DNA was then eluted from the buffer and the specimens washed and dried as described in Rohland et al. (2004). The DNA was eluted in a final volume of 200 μL 1×TE. PCR amplification was done using 2 units of Taq Gold and 60 cycles under the conditions described in Hofreiter et al. (2002). Depending on the taxon, we used seven to nine primer pairs to amplify GHR sequences (Tables 2, 3). When possible, we designed at least one primer per primer pair that selected against human GHR sequence to avoid amplification of contaminating human DNA, ubiquitous in the environment (Hofreiter et al., 2001). Amplification of human sequences occurred regularly when it was not possible to select against human DNA, showing that not only mitochondrial but also nuclear human DNA is an abundant contaminant. Due to the variability of the GHR sequences, different primer pairs were used for the different species for some of the amplified fragments (Table 3). Amplification products were cloned using the TOPO TA cloning kit (Invitrogen, The Netherlands) and multiple clones sequenced.

Table 2

Primer pairs for the amplification of the seven fragments used to determine GHR sequences in the six tenrecid species. Primer sequences are listed in Table 3. The length of the products is given in base pairs, including primers. n.p.: no product obtained.

 
Potamogale velox n.p. F2a/R2g 201 F3/R3 200 F4.1g/R4.1s 175 F4gapG/R4gap1 73 F5g/R5g 223 F6g/R6g 205 F7gap/R7shorta 195 
Limnogale mergulus F1g/R1g 181 F2/R2 199 F2a/R2g 201 F3g/R3g 225 F4/R4 217 F5/R5gap 264 F6g/R6g 205 F6/R6 206 F7/R7 189 
Geogale aurita F1g/R1g 181 F2/R2 199 F2a/R2g 201 F3g/R3g 225 F4/R4 217 F5/R5gap 264 F6g/R6g 205 F7gap/R7 195 
Micropotamogale lamottei n.p. F2g/R2 209 F3g/R3g 225 F4.1g/R4.1s 175 F4gapG/R4gap1 73 F5g/R5g 223 F6g/R6g 205 F7gap/R7 195 
Hemicentetes semispinosus F1g/R1gap 147 F2/R2 199 F3/R3 200 F4/R4 217 F5/R5gap 264 F5g/R5g 223 F6gap/R6gap 150 F7/R7 189 
Setifer setosus F1g/R1gap 147 F2/R2 199 F3/R3 200 F4/R4 217 F5/R5gap 264 F5g/R5g 223 F6gap/R6gap 150 F7/R7 189 
 
Potamogale velox n.p. F2a/R2g 201 F3/R3 200 F4.1g/R4.1s 175 F4gapG/R4gap1 73 F5g/R5g 223 F6g/R6g 205 F7gap/R7shorta 195 
Limnogale mergulus F1g/R1g 181 F2/R2 199 F2a/R2g 201 F3g/R3g 225 F4/R4 217 F5/R5gap 264 F6g/R6g 205 F6/R6 206 F7/R7 189 
Geogale aurita F1g/R1g 181 F2/R2 199 F2a/R2g 201 F3g/R3g 225 F4/R4 217 F5/R5gap 264 F6g/R6g 205 F7gap/R7 195 
Micropotamogale lamottei n.p. F2g/R2 209 F3g/R3g 225 F4.1g/R4.1s 175 F4gapG/R4gap1 73 F5g/R5g 223 F6g/R6g 205 F7gap/R7 195 
Hemicentetes semispinosus F1g/R1gap 147 F2/R2 199 F3/R3 200 F4/R4 217 F5/R5gap 264 F5g/R5g 223 F6gap/R6gap 150 F7/R7 189 
Setifer setosus F1g/R1gap 147 F2/R2 199 F3/R3 200 F4/R4 217 F5/R5gap 264 F5g/R5g 223 F6gap/R6gap 150 F7/R7 189 
Table 3

Primers used to obtain tenrec GHR sequences.

Primer fragment Sequence 
F1g 
G AAT TCAACAATGATGACT CTT GG
 
R1g 
GAATGT CAG GTT CATAAC AAC TGG TAC
 
R1gap 
AT CAT CAT CCT TTG CCC CA
 
F2 
GCT TCTAAS CAT TGA CCT GC
 
F2a 
CTAAGC ATT GAC TYD CAAAAATCA CT
 
R2 
TG GTC AAG GCA CAA GAG ATC TA
 
F2g 
G ATC RGA CAC AGA CAG RCT TCTAA
 
R2g 
TTGATT CTT CTG GTC AAG GCA C
 
R2.2s right 
GGT CAA GGC ACAAGA GAT CA
 
F3 
GTGACATGT GTGATG GTACCT CAG AGG TG
 
R3 
A KGA GCT GAC TCA GAY CCA
 
F3g 
ACA RAG GTT RAAAGG GGAAG
 
R3g 
TG GGC ATAAAA GTC GAT GTT TG
 
F4 
TAG CTTACT GTC TMYWGAYRC TG
 
R4 
CGG GGAAAG GAC CAC ACT C
 
R4gap1 
GG GAC ATC CCT GCT TTAAG
 
F4gapg 
CAG GTAAGC GAGATTACA CCA G
 
F4.1g 
CAATAC CAC TTC TTAATG GTG GAT C
 
R4.1s right 
CA CTG GAATAT CCC TGC TTTAAG
 
F5 
GAC TTT TAT GCC CAG GTAAGC GAG
 
F5g 
GC AGG GAG TGT GGT CCT TTC
 
R5g 
CT GTG GTG ATG TAA CTG TCT TCC TG
 
R5gap 
TGG TAA GGC TTT CTG TGG TGA
 
F6 
A CCT GGC CAA GCC AAC TTCA
 
R6 
GC ATC TCG GAG CTV GGK GCT
 
F6g 
AC TTC TGT GAG GCA GAT GCC A
 
R6g 
TGGACT ATATGGATG GAG GTATAG TCT G
 
F6gap 
CAG ATG CCAAAAAGT GCATTG
 
R6gap 
AGC TCG GGG CTC CTT CTG
 
F7 
AGAAAG CCT TAC CAC TAC TGC TGT
 
R7 
T GTT CAG TTG GTC TGT GCT CAC
 
F7gap 
CA CCA CAG AAA GCC TTACCA CTA
 
F7gap 
CA CCA CAG AAA GCC TTACCA CTA
 
R7 short a 
T GTT CAG TTG GTC TGT GCT C
 
Primer fragment Sequence 
F1g 
G AAT TCAACAATGATGACT CTT GG
 
R1g 
GAATGT CAG GTT CATAAC AAC TGG TAC
 
R1gap 
AT CAT CAT CCT TTG CCC CA
 
F2 
GCT TCTAAS CAT TGA CCT GC
 
F2a 
CTAAGC ATT GAC TYD CAAAAATCA CT
 
R2 
TG GTC AAG GCA CAA GAG ATC TA
 
F2g 
G ATC RGA CAC AGA CAG RCT TCTAA
 
R2g 
TTGATT CTT CTG GTC AAG GCA C
 
R2.2s right 
GGT CAA GGC ACAAGA GAT CA
 
F3 
GTGACATGT GTGATG GTACCT CAG AGG TG
 
R3 
A KGA GCT GAC TCA GAY CCA
 
F3g 
ACA RAG GTT RAAAGG GGAAG
 
R3g 
TG GGC ATAAAA GTC GAT GTT TG
 
F4 
TAG CTTACT GTC TMYWGAYRC TG
 
R4 
CGG GGAAAG GAC CAC ACT C
 
R4gap1 
GG GAC ATC CCT GCT TTAAG
 
F4gapg 
CAG GTAAGC GAGATTACA CCA G
 
F4.1g 
CAATAC CAC TTC TTAATG GTG GAT C
 
R4.1s right 
CA CTG GAATAT CCC TGC TTTAAG
 
F5 
GAC TTT TAT GCC CAG GTAAGC GAG
 
F5g 
GC AGG GAG TGT GGT CCT TTC
 
R5g 
CT GTG GTG ATG TAA CTG TCT TCC TG
 
R5gap 
TGG TAA GGC TTT CTG TGG TGA
 
F6 
A CCT GGC CAA GCC AAC TTCA
 
R6 
GC ATC TCG GAG CTV GGK GCT
 
F6g 
AC TTC TGT GAG GCA GAT GCC A
 
R6g 
TGGACT ATATGGATG GAG GTATAG TCT G
 
F6gap 
CAG ATG CCAAAAAGT GCATTG
 
R6gap 
AGC TCG GGG CTC CTT CTG
 
F7 
AGAAAG CCT TAC CAC TAC TGC TGT
 
R7 
T GTT CAG TTG GTC TGT GCT CAC
 
F7gap 
CA CCA CAG AAA GCC TTACCA CTA
 
F7gap 
CA CCA CAG AAA GCC TTACCA CTA
 
R7 short a 
T GTT CAG TTG GTC TGT GCT C
 

The challenges confronting ancient DNA studies (Hofreiter et al., 2001; Olson and Hassanin, 2003) are relevant to this work, as we used museum specimens collected nearly 100 years ago. Hence, we used appropriate laboratory techniques at a dedicated ancient DNA facility at the Max Planck Institut for Evolutionary Anthropology, Leipzig.

Figure 3 shows sequence overlap between adjacent fragments for all taxa. Except for Potamogale, which has a 6-bp gap between fragments 3 and 4, all species have continuous sequences when the amplified fragments are concatenated after trimming the primers. Table S1 details sequence overlap across our amplified fragments, all of which are identical within species, both between different fragments and alternative amplicons that span homologous regions. Moreover, each amplified fragment excluding primers is unique in our data set, making it highly unlikely that our sequences are chimaeric (see Olson and Hassanin, 2003). When compared to the available sequences in GenBank by Blast searches (Table S2), all fragments show closest matches to members of the Afrotheria, and 38 out of 47 fragments are closest to published GHR sequences of the Tenrecidae. Given the occasionally short length of the fragments, slightly closer matches to other members of Afrotheria are not surprising. Finally, except for two fragments from Setifer setosus (which match corresponding sequences from Echinops AF392889), all others differ slightly from previously published sequences available in GenBank.

Figure 3

Schematic view of PCR fragments used to reconstruct GHR sequences (see also Tables 2, 3). Dashed lines for Potamogale and Micropotamogale indicate missing sequences due to nonamplification of the first PCR fragment. A 6-bp gap between fragments 3 and 4 in the sequence data for Potamogale is indicated by the box. Deletions are shown as gaps in the sequence but do not represent missing data. The length of overlap among fragments is shown to scale.

Figure 3

Schematic view of PCR fragments used to reconstruct GHR sequences (see also Tables 2, 3). Dashed lines for Potamogale and Micropotamogale indicate missing sequences due to nonamplification of the first PCR fragment. A 6-bp gap between fragments 3 and 4 in the sequence data for Potamogale is indicated by the box. Deletions are shown as gaps in the sequence but do not represent missing data. The length of overlap among fragments is shown to scale.

Sequence Alignment

Using MacClade 4.07 (Maddison and Maddison, 2000) and Clustal X (Thompson et al., 1997), we concatenated GenBank files, added new sequences, and constructed alignments that preserved reading frames and contained few indels. GHR shows several conserved regions that facilitate a priori homology assessment. Nevertheless, some ambiguity remains regarding the positions of certain indels and adjacent nucleotides. Exploration of alignment ambiguity has occasionally (e.g., Messenger and McGuire, 1998), but not always (e.g., Douady et al., 2003), led to revised phylogenetic interpretations. For this reason, we explore a limited number of alternative alignments, differences across which are summarized in Table S3. Confidence indices mentioned in the text, as well as statistical comparisons of alternative topologies, are based on the first alignment (with the addition of the morphological partition, as indicated below), unless stated otherwise. Topological results were not significantly altered by using the other three alignments.

Each series of internal (i.e., not leading or trailing), contiguous gap characters was assumed to represent a single insertion and/or deletion event (indel). For all of our analyses including sequence data, we coded indels for each alignment, adding them as binary characters following the aligned nucleotides. Actual gap characters interspersed among the aligned nucleotides were treated as missing data. In Bayesian analyses, indels were treated using the binary (restriction site) model without assuming that all presence/absence characters have been observed (MrBayes command “LSET CODING = VARIABLE”). Sequence alignments and other supplementary data are available online at http://systematicbiology.org.

Morphological Data Collection

We used an anatomical dataset consisting of 126 characters, 20 of which are from the soft-tissues of the rostrum and cranial arterial supply, 46 from the cranium, 30 from the jaw and dentition, and 30 from the postcranial skeleton. Morphological characters were based on Asher (2000) and coded using specimens noted in Appendix 1. A nexus file with the morphological data is available at www.treebase.org (accession S1460).

Olson and Goodman (2003) questioned two coding decisions made by Asher (1999), (2000): occurrence of the fenestrate basioccipital in Microgale and morphology of the nasolacrimal duct (also known as the “lacrimal canal”) in Limnogale. Olson and Goodman stated that Asher (1999) coded both as absent, whereas they noted that a fenestrate basioccipital occurs in some species of Microgale (Asher [1999] sampled only M. talazaci) and stated that Limnogale possesses a nasolacrimal duct. As of this writing, M. talazaci remains the only species of Microgale with nuclear DNA sequences available to us (Malia et al., 2002). Hence, we still have a limited sample of this genus, but accept Olson and Goodman's (2003) observation and code the genus Microgale as polymorphic for the fenestrate basioccipital (character no. 35) in this study.

Concerning the presence of a nasolacrimal duct in Limnogale, this was in fact not the character cited as a potential semiaquatic tenrec synapomorphy by Asher (1999), (2000). Rather, absence of an external lacrimal foramen (character no. 53) was coded in both studies, as depicted here in Figure 2. There is a clear osteological difference in the expression of a single, conspicuous lacrimal foramen at the anterior margin of the orbit, dorsal to the infraorbital canal, in most tenrecs (e.g., Setifer, Fig. 2). This region is smooth and without a major foramen in both potamogalines and Limnogale (Fig. 2). Hence, we retain the coding of Asher (1999), (2000) for the present study (see also Sánchez-Villagra and Asher, 2002). Expression of a nasolacrimal duct was coded separately from the lacrimal foramen in Asher (2000) based on observations of soft tissue anatomy in histologically prepared anatomical sections (Asher, 2001). To our knowledge, no histological preparation of Limnogale has ever been made, so we cannot compare the patent, partly soft tissue nasolacrimal duct in most tenrecs and other mammals with any such structure in Limnogale. Hence, we code this character (no. 15) “missing” for Limnogale in our morphological data matrix. Coding these two characters (duct, foramen) independently is justified by the variable expression of the lacrimal foramen in taxa with a patent nasolacrimal duct (e.g., Frahnert, 1999).

Phylogenetic Inference

The search strategies described below were applied to each of the four alignments summarized in Table S3 using a 43-taxon data set sampling only GHR and a 23-taxon data set sampling GHR plus morphology.

MP analyses were undertaken with PAUP 4.0b10 (Swofford, 2002); Bayesian algorithms were applied with MrBayes 3.1 (Huelsenbeck and Ronquist, 2001). For the full GHR taxon sample, our MP analyses searched heuristically with at least 100 random addition replicates with TBR branch swapping, multiple states treated as polymorphic, and branches with a zero length under any optimization collapsed. For our Bayesian and likelihood bootstrap analyses, we used the HKY+I+G model for the full GHR taxon sample (Fig. 4) and GTR+I+G as the optimal model for the smaller GHR sample (Fig. 5), as indicated by the AIC in MrModeltest 2.1 (Nylander, 2004). We used the default PAUP commands given in MrModeltest (“DSet distance = JC objective = ME base = equal rates = equal pinv = 0 subst = all negbrlen = setzero; NJ showtree = no breakties = random”) to obtain an initial tree, used by PAUP to estimate maximum likelihood (ML) parameters for the likelihood bootstrap analysis of the 43-taxon GHR data set (which resulted in the values “Lset Base = [0.2547 0.3078 0.2293] Nst = 2 TRatio = 2.1473 Rates = gamma Shape = 1.7732 Pinvar = 0.1990”). In addition, the ML bootstrap analysis excluded indels, used 143 pseudoreplicates of an “as-is” addition sequence with TBR branch swapping, and obtained starting trees with stepwise addition.

Figure 4

Phylogenetic trees based on GHR sequences. Bayesian tree (left) is a majority rule consensus of 9,850 trees (1,000,000 generations sampling every 100), excluding the first 150 as burn-in, from alignment 1 (Table S3) using the HKY+I+G substitution model. MP tree (right) is strict consensus of six trees, each with 2389 steps. Numbers above and below nodes at left indicate, respectively, Bayesian posterior probability values and ML bootstrap support values. (The latter exclude gap characters.) Numbers below nodes at right indicate MP bootstrap support values. Malagasy tenrecs are shown in boldface. Branch lengths do not represent divergences.

Figure 4

Phylogenetic trees based on GHR sequences. Bayesian tree (left) is a majority rule consensus of 9,850 trees (1,000,000 generations sampling every 100), excluding the first 150 as burn-in, from alignment 1 (Table S3) using the HKY+I+G substitution model. MP tree (right) is strict consensus of six trees, each with 2389 steps. Numbers above and below nodes at left indicate, respectively, Bayesian posterior probability values and ML bootstrap support values. (The latter exclude gap characters.) Numbers below nodes at right indicate MP bootstrap support values. Malagasy tenrecs are shown in boldface. Branch lengths do not represent divergences.

Figure 5

Single topology supported by both Bayesian and MP algorithms using combined GHR and morphological data. Bayesian tree is a majority rule consensus as described in Figure 4, using the GTR+I+G substitution model for GHR and Mk (Lewis, 2001) for morphology. This tree was generated without gamma-distributed rate variation (Mk+G) and uses LSET CODING = ALL for the morphological partition; optimal trees from additional runs with Mk+G and LSET CODING = VARIABLE were compatible. MP with all character changes equally weighted supports a single best tree with 1550 steps. Numbers above nodes indicate Bayesian posterior probabilities; numbers below nodes indicate MP bootstrap support values. Malagasy tenrecs are shown in boldface, fossils with a dagger.

Figure 5

Single topology supported by both Bayesian and MP algorithms using combined GHR and morphological data. Bayesian tree is a majority rule consensus as described in Figure 4, using the GTR+I+G substitution model for GHR and Mk (Lewis, 2001) for morphology. This tree was generated without gamma-distributed rate variation (Mk+G) and uses LSET CODING = ALL for the morphological partition; optimal trees from additional runs with Mk+G and LSET CODING = VARIABLE were compatible. MP with all character changes equally weighted supports a single best tree with 1550 steps. Numbers above nodes indicate Bayesian posterior probabilities; numbers below nodes indicate MP bootstrap support values. Malagasy tenrecs are shown in boldface, fossils with a dagger.

Bayesian analyses of the larger GHR dataset were based on at least four independent runs, each using a random starting tree and 1,000,000 generations with one cold and three heated chains, sampling trees every 100 generations. Bayesian runs of the 43-taxon GHR and the 23-taxon combined morphology-GHR data sets both reached stationarity between approximately 10,000 to 14,000 generations, as determined by visually inspecting asymptotic graphs of likelihood scores across generations. Our phylogenetic conclusions are based on only succeeding generations, starting at 15,000, discarding the first 14,900 (sampling every 100th generation) as “burn-in.” Each run of 1,000,000 generations converged on a single, consistent result.

The taxon sample for the smaller, combined GHR-morphology data set was chosen based on availability of GHR sequences as well as osteological and soft tissue data from Asher (2000, 2001; see also Appendix 1). This sample included all genera of tenrecs, including fossils, plus Orycteropus afer, Procavia capensis, Elephantulus brachyrhynchus, a composite golden mole (Chrysospalax trevelyani GHR, Chrysochloris stuhlmanni and C. asiatica morphology), Erinaceus europaeus, three soricids, and Canis latrans, and was rooted with a composite didelphid (Monodelphis domestica GHR, Didelphis sp. morphology). Most soft tissue characters remain missing for two of the extant tenrecs: Limnogale and Oryzorictes.

Combined and GHR-only MP analysis of the smaller dataset used the same search parameters as described above, leaving DNA entries missing for the three fossil taxa. MP bootstrap support values were generated with 1,000 pseudoreplicates, each with 10 random addition replicates and TBR branch swapping. Bayesian search parameters for the smaller GHR dataset were as described above. In the combined Bayesian analysis of morphology and sequences we used different models for each partition: the GTR+I+G for sequences following the AIC in MrModelTest (Nylander 2004), the binary (restriction site) model for indels (with LSET CODING = VARIABLE), and Mk for morphology following Lewis (2001) and Nylander et al. (2004). We undertook multiple runs using both LSET CODING = VARIABLE and LSET CODING = ALL commands for the morphological partition. We also ran our morphological partition with (Mk+G) and without (Mk) gamma-shaped rate variation (LSET RATES = GAMMA).

Results

Affinities of Living Tenrecs

Parsimony, likelihood, and Bayesian methods applied to our GHR data consistently supported the monophyly of Malagasy tenrecs to the exclusion of the two living African genera with high support indices (Fig. 4). In each case, regardless of the alignment (Table S3) or algorithm used, and in agreement with Olson and Goodman (2003), Limnogale was closely related to Microgale, and potamogalines were reconstructed as the sister group to other living tenrecs. The extant Malagasy tenrec clade consisted of two radiations: spiny tenrecs (Tenrecinae) and soft tenrecs (Oryzorictinae plus Geogale). Less clear were the positions of Geogale and Oryzorictes within the soft-tenrec clade, and of Hemicentetes and Tenrec within the spiny tenrec clade.

Bayesian analysis of sequence data alone favors Oryzorictes at the base of a soft-tenrec clade, contradicting oryzorictine monophyly (Fig. 4). However, Bayesian support for a soft-tenrec clade excluding Oryzorictes ranged from 54 to 58 across the four alignments; and trees produced by MP for each of the four alignments left Oryzorictes and Geogale unresolved at the base of this clade (Fig. 4). Furthermore, we cannot statistically reject a monophyletic Oryzorictinae with Geogale as its sister taxon (Table 4). In contrast, statistical comparisons based on the GHR-only and combined datasets reject any sister-group relation between the semiaquatic Malagasy Limnogale and African potamogalines (Table 4).

Table 4

Tests of alternative topologies.

(A) One-tailed Shimodaira-Hasegawa test using 1,000 RELL bootstrap replicates, applied to the 43-taxon GHR dataset (Fig. 4), using HKY+I+G likelihood model in PAUP. The completely bifurcating tree with the highest likelihood score from alignment 1, run 1 (generation no. 478,200, see supporting data) was used for comparisons. Alternatives are the same except as indicated. Abbreviations are as follows: Ec, Echinops; Er, Erythrozootes, FT, fossil tenrecs; Ge, Geogale; He, Hemicentetes; Li, Limnogale; Mi, Microgale; MT, Malagasy tenrecs; On, Oryzorictinae; Or, Oryzorictes; Pa, Parageogale; Pn, Potamogalinae; Pr, Protenrec; Se, Setifer; Te, Tenrec; Tn, Tenrecinae. 

 
Tree   −ln L  Diff −ln L  SH-test P 
best Bayesian tree, align1, run1   12,034.29976  —  — 
(Ge(Or(Li,Mi)))   12,043.75843  9.45867  0.484 
(He,Te)   12,046.09929  11.79953  0.394 
((Li,Pn)(Tn,(Or(Ge,Mi))))   12,163.30139  129.00163  0.000* 
(Tn(Or(Ge(Mi(Li,Pn)))))   12,141.31100  107.01123  0.000* 

 
(B)Templeton and winning sites tests applied to 23-taxon combined data set (Fig. 5) using MP in PAUP. The completely bifurcating tree from Figure 5 was used for comparisons; alternatives are the same except as indicated. 

 
Tree Length Rank Sums* N z Templeton P Counts Winning Sites P 

 
As in Fig. 5 1550 (best) — — — — — 
((Or(Li,Mi))((Er,Pr)(Pa,Ge))) 1551 16.0 −0.3780 0.7055  
  −12.0    −3 1.0000 
(Pn(Er,Pr)(Pa(MT))) 1557 25.0 −1.9332 0.0532 0.1250 
On paraphyletic  −3.0    −1  
(Pn(Er,Pr)(Pa(MT))) 1557 77.0 14 −1.6977 0.0896 10 0.1796 
On monophyletic  −28.0    −4  
(Pn((Er,Pr)Pa)(MT)) 1556 21.0 −2.4495 0.0143* 0.0313* 
On paraphyletic  −0.0     
(Pn(Pa((Er,Pr)(MT)))) 1558 21.0 −2.2711 0.0231* 0.0313* 
On paraphyletic  −0.0     
((Li,Pn)(Tn,(Or((Ge,FT),Mi)))) 1592 1638.0 62 −5.3340 < 0.0001* 52 < 0.0001* 
  −315.0    −10  
(Tn(Or((Ge,FT)(Mi(Li,Pn))))) 1582 1254.0 56 −4.2762 < 0.0001* 44 < 0.0001* 
  −342.0    −12  
(Ge((Er,Pr)Pa)) 1552 3.0 −1.4142 0.1573 0.5000 
  −0.0     
(A) One-tailed Shimodaira-Hasegawa test using 1,000 RELL bootstrap replicates, applied to the 43-taxon GHR dataset (Fig. 4), using HKY+I+G likelihood model in PAUP. The completely bifurcating tree with the highest likelihood score from alignment 1, run 1 (generation no. 478,200, see supporting data) was used for comparisons. Alternatives are the same except as indicated. Abbreviations are as follows: Ec, Echinops; Er, Erythrozootes, FT, fossil tenrecs; Ge, Geogale; He, Hemicentetes; Li, Limnogale; Mi, Microgale; MT, Malagasy tenrecs; On, Oryzorictinae; Or, Oryzorictes; Pa, Parageogale; Pn, Potamogalinae; Pr, Protenrec; Se, Setifer; Te, Tenrec; Tn, Tenrecinae. 

 
Tree   −ln L  Diff −ln L  SH-test P 
best Bayesian tree, align1, run1   12,034.29976  —  — 
(Ge(Or(Li,Mi)))   12,043.75843  9.45867  0.484 
(He,Te)   12,046.09929  11.79953  0.394 
((Li,Pn)(Tn,(Or(Ge,Mi))))   12,163.30139  129.00163  0.000* 
(Tn(Or(Ge(Mi(Li,Pn)))))   12,141.31100  107.01123  0.000* 

 
(B)Templeton and winning sites tests applied to 23-taxon combined data set (Fig. 5) using MP in PAUP. The completely bifurcating tree from Figure 5 was used for comparisons; alternatives are the same except as indicated. 

 
Tree Length Rank Sums* N z Templeton P Counts Winning Sites P 

 
As in Fig. 5 1550 (best) — — — — — 
((Or(Li,Mi))((Er,Pr)(Pa,Ge))) 1551 16.0 −0.3780 0.7055  
  −12.0    −3 1.0000 
(Pn(Er,Pr)(Pa(MT))) 1557 25.0 −1.9332 0.0532 0.1250 
On paraphyletic  −3.0    −1  
(Pn(Er,Pr)(Pa(MT))) 1557 77.0 14 −1.6977 0.0896 10 0.1796 
On monophyletic  −28.0    −4  
(Pn((Er,Pr)Pa)(MT)) 1556 21.0 −2.4495 0.0143* 0.0313* 
On paraphyletic  −0.0     
(Pn(Pa((Er,Pr)(MT)))) 1558 21.0 −2.2711 0.0231* 0.0313* 
On paraphyletic  −0.0     
((Li,Pn)(Tn,(Or((Ge,FT),Mi)))) 1592 1638.0 62 −5.3340 < 0.0001* 52 < 0.0001* 
  −315.0    −10  
(Tn(Or((Ge,FT)(Mi(Li,Pn))))) 1582 1254.0 56 −4.2762 < 0.0001* 44 < 0.0001* 
  −342.0    −12  
(Ge((Er,Pr)Pa)) 1552 3.0 −1.4142 0.1573 0.5000 
  −0.0     

Application of MP to the morphological dataset yields optimal trees similar in some regards to those generated by sequences alone, such as monophyly of tenrecids and potamogalines, support for a spiny tenrec clade, and a sister taxon relationship between Echinops and Setifer. However, in contrast to the GHR signal, morphological data support the position of African potamogalines near Limnogale (Fig. 6). This relationship appears in most of the optimal trees in the combined MP analysis, but is unresolved in the strict consensus. Nevertheless, a Limnogale-potamogaline clade is supported by MP applied to the living taxa alone with a bootstrap value of 71 (not figured; see also Asher, 1999), and by Mk applied to the morphological dataset including fossils (Fig. 6). Using the morphological dataset alone, the alternative topologies summarized in Table 4, including variants that preserve monophyly of Malagasy tenrecs and a Limnogale-Microgale clade, are rejected by Templeton and winning sites tests.

Figure 6

Phylogenetic trees based on morphological data. Bayesian tree (left) is a majority rule consensus as described in Figure 4, using Mk (Lewis, 2001) and LSET CODING = VARIABLE. MP tree (right) is a strict consensus of six trees, 382 steps, all morphological character changes treated equally. Numbers above nodes indicate Bayesian posterior probabilities (left) or MP bootstrap support values (right). Malagasy tenrecs are shown in boldface, fossils with a dagger.

Figure 6

Phylogenetic trees based on morphological data. Bayesian tree (left) is a majority rule consensus as described in Figure 4, using Mk (Lewis, 2001) and LSET CODING = VARIABLE. MP tree (right) is a strict consensus of six trees, 382 steps, all morphological character changes treated equally. Numbers above nodes indicate Bayesian posterior probabilities (left) or MP bootstrap support values (right). Malagasy tenrecs are shown in boldface, fossils with a dagger.

Affinities of Extinct Tenrecs

Application of MP to the combined dataset including the three fossil tenrecs, regardless of alignment (Table S3) or analysis parameters, supports a Parageogale-Geogale clade with relatively high confidence, with MP-bootstrap support values (89) comparable to that for potamogalines (87; see Fig. 5). Bayesian analyses of the combined dataset also supported this clade, but with a posterior probability (80) weaker than that for potamogalines (98). Erythrozootes and Protenrec are also reconstructed together, in turn adjacent to Geogale-Parageogale, regardless of alignment or tree-building technique. However, support indices for this clade are much lower (posterior probability 67, MP bootstrap below 50), as are the supports for a clade joining the three fossil taxa with Geogale (posterior probability 76, MP bootstrap below 50; see Fig. 5).

Several morphological characters support a Parageogale-Geogale clade, which, following Butler (1984) may be referred to the Geogalinae. First, the reduction of its upper molar metacone (character no. 73, state 2), protocone (no. 74, state 1), and of the lower molar talonid (no. 85, state 1) favor its placement with other dentally zalambdodont taxa (i.e., in this sample, tenrecs and golden moles; see Asher and Sánchez-Villagra [2005] for a definition of anatomical zalambdodonty). Parageogale and Geogale share a highly reduced maxillary process of the zygoma (no. 59, state 1; also present in soricids). Geogalines also possess a broad gap between the anterior central incisors (character no. 126, state 1), a condition also seen in Erinaceus and in some specimens of Setifer (here coded as polymorphic). They also have two premaxillary teeth (no. 67, state 2; also present in some tenrecines and Erythrozootes). In contrast to the other nine tenrecid genera, fossil tenrecs plus Geogale possess a relatively long infraorbital canal (no. 60, state 0).

Templeton and winning sites tests based on MP reject alternative hypotheses placing all three fossils either outside of living Tenrecidae or together as the sister-clade to African potamogalines (Table 4). However, another alternative, placing Parageogale as the sister-taxon to a (potamogaline (Protenrec Erythrozootes)) clade, again with all African tenrecs outside of the Malagasy radiation (Fig. 1A), cannot be rejected.

Additional Tests of Fossil Tenrec Phylogeny

All three fossil tenrec genera were first described from the Kenyan Miocene (Butler and Hopwood, 1957) and remain known only from a few craniodental fragments (Butler, 1984). Published reviews including these taxa have generally supported their affinity to modern tenrecids (Butler, 1969, 1978, 1984, 1985; McKenna and Bell, 1997; Mein and Pickford, 2003; but see Poduschka and Poduschka, 1985). As is the case for other fossils over 1 million years in age, sequence data cannot be obtained from these specimens (Hofreiter et al., 2001). Of the 126 characters sampled in our morphological matrix, Erythrozootes and Protenrec are 24% complete and Parageogale is ca. 18% complete. Nevertheless, the most poorly known taxon in this study, Parageogale, shows a relatively well-supported position, consistent with the hypothesis originally presented by Butler and Hopwood (1957) that it is the sister-taxon to the living Geogale aurita, and contradicting the monophyly of the Malagasy radiation (Fig. 1A).

To test the hypothesis that the 22 characters sampled for Parageogale can accurately reconstruct its phylogeny, we used these same characters to reconstruct the phylogeny of other tenrecs in our study. That is, for each of the 10 living tenrecid genera, we replaced GHR data and all morphological characters, except for the 22 known for Parageogale, with missing entries and ran the modified morphology+GHR dataset using MP, as described above in Materials and Methods. Stated differently, if a living tenrecid genus had gone extinct in the early Miocene, and were known only from cranial fragments similar to those of Parageogale, would we be able to accurately (as defined by the full-data sample depicted in Fig. 5) reconstruct its phylogenetic position? If the respective extant taxon sampled only for the 22 Parageogale characters appears in a different part of the tree relative to its position in the full analysis, we would have less confidence in the placement of Parageogale.

In fact, the reduced dataset did not greatly change the position of any extant tenrec (Fig. 7). Out of the 10 modified datasets (1 for each living tenrecid genus), 2 (Echinops and Setifer) yielded the same tree as the full sample, and 6 of the remaining 8 yielded varying degrees of nonresolution in multiple shortest trees, consensuses of which (Fig. 7) were still compatible with the topology supported by the full dataset. Only two cases (Tenrec and Potamogale) yielded optimal trees with a slightly different topology. The former altered relations within spiny tenrecs (supporting Tenrec-Setifer rather than Tenrec-Hemicentetes), and the latter reconstructed Oryzorictes adjacent to Microgale-Limnogale to the exclusion of Geogale, preserving oryzorictine monophyly. However, Tenrec bootstrap resampling still supports a spiny-tenrec clade with a value of 79; and the Tenrec-Setifer clade has an MP bootstrap support value under 50. Similarly, for the run using a reduced sample for Potamogale, oryzorictine monophyly is supported with an MP bootstrap of just 57, and the unmodified, combined-data sample cannot reject this hypothesis (Table 4). In these and other cases, bootstrap resampling generally yielded lower support values compared to the full sample (cf. Fig. 5 versus Fig. 7), but in no case did a clade produced by a reduced-sample analysis contradict a well-supported clade in the full sample.

Figure 7

Results of reduced-character MP analyses of each of the 10 living tenrecid genera (as identified in boldface), with all GHR sequences and morphological characters, except for the 22 known for Parageogale, coded as missing. Matrices with either Echinops or Setifer coded in this fashion yield the same topology (top left), also identical to the combined-data topology depicted in Figure 5. Each tree represents either a strict consensus or a single, most-parsimonious result, as follows: Echinops 1 tree 1,539 steps, Geogale 3 trees 1,491 steps, Hemicentetes 4 trees 1,517 steps, Limnogale 3 trees 1,521 steps, Microgale 4 trees 1,521 steps, Micropotamogale 1 tree 1,520 steps, Oryzorictes 7 trees 1,515 steps, Potamogale 1 tree 1,511 steps, Setifer 1 tree 1,543 steps, Tenrec 2 trees 1,522 steps. Numbers adjacent to nodes represent MP bootstrap support values (100 pseudoreplicates of a simple addition sequence). Bootstrap values in tree at top left for Echinops are listed above nodes, Setifer below.

Figure 7

Results of reduced-character MP analyses of each of the 10 living tenrecid genera (as identified in boldface), with all GHR sequences and morphological characters, except for the 22 known for Parageogale, coded as missing. Matrices with either Echinops or Setifer coded in this fashion yield the same topology (top left), also identical to the combined-data topology depicted in Figure 5. Each tree represents either a strict consensus or a single, most-parsimonious result, as follows: Echinops 1 tree 1,539 steps, Geogale 3 trees 1,491 steps, Hemicentetes 4 trees 1,517 steps, Limnogale 3 trees 1,521 steps, Microgale 4 trees 1,521 steps, Micropotamogale 1 tree 1,520 steps, Oryzorictes 7 trees 1,515 steps, Potamogale 1 tree 1,511 steps, Setifer 1 tree 1,543 steps, Tenrec 2 trees 1,522 steps. Numbers adjacent to nodes represent MP bootstrap support values (100 pseudoreplicates of a simple addition sequence). Bootstrap values in tree at top left for Echinops are listed above nodes, Setifer below.

Discussion

Data Combination and Tenrec Phylogeny

A previous morphology-based investigation of tenrecid phylogeny published by one of us (Asher, 1999) argued for a clade of semiaquatic tenrecs, placing Malagasy Limnogale as the sister-taxon to continental African potamogalines. Character support for this clade was primarily from the skull, including a fenestrate basioccipital (no. 35 in this study), a shortened frontal bone (no. 61), and a reduced lacrimal foramen (no. 53). Importantly, none of these character states are consistently found in nontenrecid, semiaquatic, faunivorous, small mammals (Sánchez-Villagra and Asher, 2002), a factor that had previously led Asher to view the “semiaquatic” tenrec clade with increased confidence.

As discussed above, morphological data analyzed alone still yield some support for a semiaquatic clade, although recoding fenestration in the basioccipital to account for polymorphism in Microgale (as recommended by Olson and Goodman, 2003) has eliminated this character from optimizing unambiguously as a Limnogale-potamogaline synapomorphy. Furthermore, compared to the study of Asher (1999), the larger number of characters and sampled tenrecs in this study yields reduced support for a semiaquatic tenrec clade (Fig. 6).

However, the key reason for the nonrecovery of a semiaquatic clade in this study is the very strong sequence-based signal favoring a Limnogale-Microgale clade. Indeed, with their sample of different loci for multiple species of Microgale, Olson and Goodman (2003) found that Limnogale actually nests within that genus, comprising the sister-taxon to an M. dobsoni–M. talazaci clade to the exclusion of other Microgale species. The strength of the signal supporting a Microgale-Limnogale clade in our study (100 MP bootstrap, 100 ML bootstrap, and 100 Bayesian posterior probability in the GHR-only analysis [Fig. 4]; 95 MP bootstrap and 94 Bayesian posterior probability in the combined analysis [Fig. 5]) has convinced both of us that the previous interpretation of the morphological signal as indicative of a semiaquatic tenrec clade (Asher, 1999) is incorrect. Due to this unambiguous support from GHR sequences, which is considerably stronger than that from morphology alone for a semiaquatic clade and which prevails in the combined analysis, the cranial characters supporting the “semiaquatic” clade cited above must be reinterpreted as homoplastic.

If the morphological data used here are misleading regarding a semiaquatic tenrec clade, why do we then combine them with our GHR data? The most important reason for retaining morphology in our dataset is one of principle: most individual datasets are not in their entirety either “true” or “false”; but are themselves mosaics of variable character-data that may provide resolution at different levels in any given tree (Gatesy et al., 2003). Combined data sets enable recognition of phylogenetic signals that would remain obscure with the analysis of subdivisions thereof (Gatesy et al., 1999, 2005). Furthermore, including morphological data in the combined analysis remains the best means to sample fossil tenrecs. We cannot be completely sure that the morphology known for these fossils enables us to accurately understand their phylogenetic history. However, as discussed above, when used in simulations to replace the complete morphology-GHR dataset for each of the 10 living tenrecid genera, the morphological characters known for the most incomplete of the fossils (Parageogale) yield results that are largely congruent with the combined-data topology.

Character Assessment and Hindlimb Function in Potamogale

One recent study of hindlimb characters (Salton and Szalay, 2004) has also argued for the inclusion of Limnogale within the Malagasy radiation. By assessing characters of the tarsal complex in an “ecological and evolutionary framework,” Salton and Szalay proposed to identify phylogenetically informative characters: “traits with clear species-specific adaptations are a potential interference in cladistic analyses and cannot be meaningfully used without ecology-based character assessment” (Salton and Szalay, 2004:73). In regards to the “semiaquatic” clade, their procedure resulted in the identification of anatomical differences (e.g., astragalar neck-head transition) and similarities (e.g., medially directed tibial-fibular malleoli) between Limnogale and Potamogale (they did not include Micropotamogale in their analysis). In their opinion, the former comprise phylogenetic data in support of the “family level distinction” of Potamogale from other tenrecs, and the latter are interpreted as homoplastic.

However, we are concerned that Salton and Szalay (2004) did not identify a replicable optimality criterion (e.g., MP, ML) by which they reached their conclusions on homology. Furthermore, we believe that Salton and Szalay have not fully appreciated the function of the hindlimb in Potamogale. Regarding its locomotion, Salton and Szalay refer to its “heavy foot thrusts” (p. 90), and note that “heavy loading in the UAJ [upper ankle joint]… and UAJ stabilization plays an important role… in the aquatic locomotion of Potamogale” (p. 86). In regards to calcaneal morphology, Salton and Szalay state that “Potamogale has an extremely long and narrow calcaneus with a long tuber, appropriate for strong, dorsolateral aquatic propulsion” (p. 93). In fact, these inferences of locomotion run counter to published descriptions of locomotor behavior in Potamogale (e.g., DuChaillu, 1860; Kingdon, 1974), which indicate that it uses its massive tail, not its feet, for aquatic propulsion. As in the other two potamogaline species (Micropotamogale lamottei and M. ruwenzorii), digits II and III of the hindfoot in Potamogale are syndactyl, and their use in grooming has been documented (Nicoll, 1985; Kingdon, 1997). As summarized by Nowak (1999), the relatively small, nonwebbed pes of Potamogale is tucked under its pelvic region during swimming and is not used for propulsion. Dobson (1883:97–98) infers from its anatomy that during locomotion, “the sole [of the foot] lies so evenly against the [pelvic ventrum] as to present the least possible projection and interfere in the least degree with the rapid passage of the body through the water, propelled by the powerful tail.… [The tail] is doubtless the sole organ of propulsion.” Based on field observations, Kingdon observed that in the water, “the animal is propelled entirely by lateral movements of the back and tail” (Kingdon, 1974:15). In contrast, Limnogale (the “web-footed” tenrec), has been observed to use its hindlimbs for semiaquatic propulsion (Benstead and Olson, 2003:1272). Despite this, and without presenting new behavioral data for either taxon, Salton and Szalay (2004:100) propose the opposite: “… the tarsal complex indicates that [Limnogale's] hind limbs are less important for propulsion than those of Potamogale.”

Hence, we remain skeptical about Salton and Szalay's method for distilling phylogenetically informative data from their morphological observations. Although we agree with them that Limnogale is not more closely related to Potamogale than to other Malagasy tenrecs, contra Asher (1999), we do not believe they presented in their paper a basis for reaching this conclusion, independent of the sequence data analyzed by Olson and Goodman (2003), and confirmed with additional data in this article.

Tenrec Biogeography

Considering the living radiation alone, Malagasy tenrecs show substantial morphological diversity, yet are recognized as a single radiation by sequence data, as observed for primates (cf. Yoder, 1992, versus Yoder et al., 1996) and carnivorans (cf. Veron, 1995, versus Flynn et al., 2005). Similarly, our results support a cohesive Malagasy radiation and argue against Malagasy tenrec polyphyly. However, the living tenrecid radiation is not a complete picture of this group's diversity. Although its paleontological record is meager, fossil African tenrecids appear to have a close relationship with living Geogale. This relationship makes the Malagasy tenrec radiation paraphyletic (Figs. 1B, 5).

A similar phylogenetic scenario was presented by Jansa et al. (1999) for Malagasy nesomyine rodents. Based on cytochrome b sequences for multiple representatives of all genera of this group, Jansa et al. (1999) disputed previous interpretations of polyphyly (Ellerman, 1940, 1941), but argued that two mainland African genera (Steatomys and Tachyoryctes) nested within the Malagasy radiation. They suggested that this phylogenetic pattern would be consistent with colonization of Madagascar by nesomyines via a single founder event, followed by dispersal to Africa from Madagascar. The inclusion of mainland African muroids within the Malagasy radiation has subsequently been questioned (Steppan et al., 2004; Jansa and Weksler, 2004); and nesomyine monophyly remains possible. A definitive conclusion must await a study that synthesizes the taxon and character samples discussed by Jansa et al. (1999), Jansa and Weksler (2004), and Steppan et al. (2004).

Monophyly of Malagasy tenrecs is also possible. We have at present no way of knowing how the missing GHR nucleotides for Parageogale, or characters from its still unknown skeleton, would affect our estimate of its relationships. Some uncertainty regarding our results supporting paraphyly (Fig. 1B) is reflected in the nonrejection of at least one alternative topology that preserves Malagasy tenrec monophyly (Table 4); and we eagerly anticipate how this result is affected by future discoveries of better-preserved fossil tenrecid material. Nevertheless, the current hypothesis of a Parageogale-Geogale clade has support from both MP and Bayesian methods (Fig. 5). Furthermore, as discussed above, the limited morphological sample available for Parageogale appears to perform fairly well when these same characters are used to reconstruct the phylogeny of each of the ten living tenrecid genera, a result that slightly increases our confidence in the placement of this fossil.

As stated in the introduction, the absence of modern mammalian orders from Madagascar (and elsewhere) during the Mesozoic (Krause, 2003), during which time land connections existed with mainland Africa (until the late Jurassic) and India (until the early Late Cretaceous), has led many to favor dispersal as the prime mechanism by which modern mammals colonized Madagascar (e.g., Olson and Goodman, 2003; Yoder et al., 2003). Repeated monophyly of Madagascar's endemic radiations is consistent with dispersal, as individual colonization events are hypothesized to be rare, and a previously unpopulated island may have open adaptive zones into which a founder can radiate into a diverse clade.

Nonmonophyly is also compatible with dispersal, but requires more (potentially unparsimonious) crossings of a geographic barrier, in this case the Mozambique channel. No one will ever know exactly how or why the tenrec crossed the channel; but based on our phylogeny we can estimate how often such an event took place. Given the combined-data tenrec phylogeny presented in Figure 5, we hypothesize that a single founder event of Madagascar by the common ancestor of Malagasy tenrecs took place at some point after the Maastrichtian, during which time a diverse vertebrate fauna shows no sign of Madagascar's modern inhabitants (Krause, 2003). Prior to the Miocene, when fossil tenrecs were present in east (Butler, 1984) and southwest (Mein and Pickford, 2003) Africa, an additional dispersal of an animal related to the geogaline common ancestor took place from Madagascar to continental Africa. The position of Erythrozootes and Protenrec (in the Protenrecinae of Butler 1984) as sister taxa to Geogale-Parageogale implies that a protenrecine relative made this back-migration yet again.

However, we note that a Protenrec-Erythrozootes clade to the exclusion of geogalines has an unimpressive MP bootstrap value below 50 and a Bayesian posterior probability of 67 (Fig. 5). An alternative hypothesis, placing protenrecines as the sister clade to Parageogale within Geogalinae, would require just a single Madagascar-Africa dispersal event postdating the initial Madagascar colonization. This alternative is just two steps longer in MP analyses and cannot be statistically rejected (Table 4). As stated above, we are cognizant of yet another alternative that preserves Malagasy tenrec monophyly (Fig. 1A), also statistically unrejected in Table 4. This scenario would require only a single colonization event of Madagascar by tenrecs, with no back-migration, again at some point after the Late Cretaceous.

Nevertheless, the optimal explanation of the data presented in this article supports paraphyly of Malagasy tenrecs relative to their mainland relatives (Figs. 1A, 5), not monophyly (Fig. 1A) or polyphyly (Fig. 1C). DNA sequence data for extinct, pre-Pleistocene tenrecs will probably never be available; and even for certain living taxa, in particular Geogale and Limnogale, such data are very difficult to obtain. Our technique for sequencing nuclear DNA from museum specimens without damaging them eases this constraint, can be applied to other groups, and greatly reduces fieldwork expense and disturbance of living populations otherwise necessary for obtaining research material. This highlights yet further the value of museum collections for basic science (Suárez and Tsutsui, 2004).

Acknowledgments

We thank Peter Vogel for access to material of Micropotamogale, Judith Chupasko and the Museum of Comparative Zoology, Harvard University, for access to Geogale, and Roland Pfeiffer and Nadin Rohland for technical assistance. We thank Rod Page, Ron DeBry, Scott Steppan, and three anonymous reviewers for helpful comments that improved the manuscript. We are also grateful to Nils Hoff for help with Figure 1 and Knut Finstermeier for preparing Figure 3. RJA acknowledges support from the Deutsche Forschungsgemeinschaft (grant AS 245/2-1), the European Commission's Research Infrastructure Action via the SYNTHESYS Project (GB-TAF 218), and the National Science Foundation USA (DEB 9800908). MH was supported by the Max Planck Society.

NOTE: We wish to acknowledge the recent study of Poux et al. (2005), published after the completion of this paper, on the colonization of Madagascar by terrestrial mammals. Poux et al. sampled most Recent genera of tenrecids (except Potamogale and Geogale), plus a large sample of other endemic Malagasy genera, and report a tenrecid phylogeny congruent with that discussed here and by Olson and Goodman (2003), for example in supporting a Limnogale-Microgale clade. However, they did not address the phylogeny of fossil taxa.

References

Adams
T. E.
Baker
L.
Fiddes
R. J.
Brandon
M. R.
The sheep growth hormone receptor: Molecular cloning and ontogeny of mRNA expression in the liver
Mol. Cell. Endocrinol.
 , 
1990
, vol. 
73
 (pg. 
135
-
145
)
Adkins
R. M.
Gelke
E. L.
Rowe
D.
Honeycutt
R. L.
Molecular phylogeny and divergence time estimates for major rodent groups: Evidence from multiple genes
Mol. Biol. Evol.
 , 
2001
, vol. 
18
 (pg. 
777
-
791
)
Asher
R. J.
A morphological basis for assessing the phylogeny of the “Tenrecoidea” (Mammalia, Lipotyphla)
Cladistics
 , 
1999
, vol. 
15
 (pg. 
231
-
252
)
Asher
R. J.
Phylogenetic history of Tenrecs and other insectivoran mammals
 , 
2000
Stony Brook
State University of New York
 
Ph.D. Thesis
Asher
R. J.
Cranial anatomy in tenrecid insectivorans: Character evolution across competing phylogenies
American Museum Novitates
 , 
2001
, vol. 
3352
 (pg. 
1
-
54
)
Asher
R. J.
Rose
K. D.
Archibald
J. D.
Insectivoran-grade placental mammals: Character evolution and fossil history
The rise of placental mammals: Origin and relationships of the major clades
 , 
2005
Baltimore
Johns Hopkins University Press
(pg. 
Pages 50
-
70
)
Asher
R. J.
Sánchez-Villagra
M. R.
Locking yourself out: Diversity among dentally zalambdodont therian mammals
J. Mamm. Evol.
 , 
2005
, vol. 
12
 (pg. 
265
-
282
)
Baumbach
W. R.
Horner
D. L.
Logan
J. S.
The growth hormone-binding protein in rat serum is an alternatively spliced form of the rat growth hormone receptor
Genes Dev.
 , 
1989
, vol. 
3
 (pg. 
1199
-
1205
)
Benstead
J. P.
Olson
L. E.
Goodman
S. M.
Benstead
J. P.
Limnogale mergulus, web-footed tenrec or aquatic tenrec
The natural history of Madagascar
 , 
2003
Chicago
University of Chicago Press
(pg. 
Pages 1267
-
1273
)
Butler
P. M.
Leakey
L. S. B.
Insectivores and bats from the Miocene of East Africa: New material
Fossil vertebrates of Africa
 , 
1969
, vol. 
volume 1
 
London
Academic Press
(pg. 
Pages 1
-
37
)
Butler
P. M.
Maglio
V. J.
Cooke
H. B. S.
Insectivora and Chiroptera
Evolution of African mammals
 , 
1978
Cambridge, Massachusetts
Harvard University Press
(pg. 
Pages 56
-
68
)
Butler
P. M.
Macroscelidea, Insectivora, and Chiroptera from the Miocene of East Africa
Palaeovertebrata
 , 
1984
, vol. 
14
 (pg. 
117
-
200
)
Butler
P. M.
The history of African insectivores
Acta Zool. Fenn.
 , 
1985
, vol. 
173
 (pg. 
215
-
217
)
Butler
P. M.
Hopwood
A. T.
Insectivora and chiroptera from the Miocene rocks of Kenya colony
BMNH Fossil Mamm. Afr.
 , 
1957
, vol. 
13
 (pg. 
1
-
35
)
Cartmill
M.
Luckett
W. P.
Szalay
F. S.
Strepsirhine basicranial structures and the affinities of the Cheirogaleidae
Phylogeny of the primates: A multidisciplinary approach
 , 
1975
New York
Plenum
(pg. 
Pages 313
-
354
)
Dobson
G. E.
A monograph of the Insectivora, systematic and anatomical, Part II
Including the families Potamogalidae, Chrysochloridae, and Talpidae
 , 
1883
London
John Van Voorst
Douady
C. J.
Catzeflis
F.
Kao
D. J.
Springer
M. S.
Stanhope
M. J.
Molecular evidence for the monophyly of Tenrecidae (Mammalia) and the timing of the colonization of Madagascar by Malagasy tenrecs
Mol. Phylogenet. Evol.
 , 
2002
, vol. 
22
 (pg. 
357
-
363
)
Douady
C. J.
Catzeflis
F.
Springer
M. S.
Stanhope
M. J.
Molecular evidence for the monophyly of Tenrecidae: A reply to Asher
Mol. Phylogenet. Evol.
 , 
2003
, vol. 
26
 (pg. 
331
-
332
)
DuChaillu
P. B.
Descriptions of five new species of mammals discovered in western equatorial Africa
Proc. Boston Soc. Nat. Hist.
 , 
1860
, vol. 
7
 (pg. 
358
-
367
)
Eisenberg
J. F.
Luckett
W. P.
Szalay
F. S.
Phylogeny, behavior, and ecology in the mammalia
Phylogeny of the primates: A multidisciplinary approach
 , 
1975
New York
Plenum
(pg. 
Pages 47
-
68
)
Eisenberg
J. F.
Gould
E.
The tenrecs: A study in mammalian behavior and evolution
Smithsonian Contrib. Zool.
 , 
1970
, vol. 
27
 (pg. 
1
-
137
)
Ellerman
J. R.
The families and genera of living rodents
Rodents other than Muridae
 , 
1940
, vol. 
Volume 1
 
London
British Museum
Ellerman
J. R.
The families and genera of living rodents
Family Muridae
 , 
1941
, vol. 
Volume 2
 
London
British Museum
Emerson
G. L.
Kilpatrick
C. W.
McNiff
B. E.
Ottenwalder
J.
Allard
M. W.
Phylogenetic relationships of the order Insectivora based on complete 12s rRNA sequences from mitochondria
Cladistics
 , 
1999
, vol. 
15
 (pg. 
221
-
230
)
Flynn
J. J.
Finarelli
J. A.
Zehr
S.
Hsu
J.
Nedbal
M. A.
Molecular phylogeny of the Carnivora (Mammalia): Assessing the impact of increased sampling on resolving enigmatic relationships
Syst. Biol.
 , 
2005
, vol. 
54
 (pg. 
317
-
337
)
Frahnert
S.
Morphology and evolution of the Glires rostral cranium
Mitteilungen aus dem Museum für Naturkunde in Berlin Zoologische Reihe
 , 
1999
, vol. 
75
 (pg. 
229
-
246
)
Gatesy
J.
Amato
G.
Norell
M.
DeSalle
R.
Hayashi
C.
Combined support for wholesale taxic Atavism in Gavialine Crocodylians
Syst. Biol.
 , 
2003
, vol. 
52
 (pg. 
403
-
422
)
Gatesy
J.
Baker
R. H.
Hidden likelihood support in genomic data: Can forty-five wrongs make a right? Syst
Biol.
 , 
2005
, vol. 
54
 (pg. 
483
-
492
)
Gatesy
J.
O'Grady
P.
Baker
R. H.
Corroboration among data sets in simultaneous analysis: Hidden support for phylogenetic relationships among higher level artiodactyl taxa
Cladistics
 , 
1999
, vol. 
15
 (pg. 
271
-
313
)
Hofreiter
M.
Capelli
C.
Krings
M.
Waits
L.
Conard
N.
Münzel
S.
Rabeder
G.
Nagel
D.
Paunovic
M.
Jambresic
G.
Meyer
S.
Weiss
G.
Pääbo
S.
Ancient DNA analyses reveal high mitochondrial DNA sequence diversity and parallel evolution of Late Pleistocene cave bears
Mol. Biol. Evol.
 , 
2002
, vol. 
19
 (pg. 
1244
-
1250
)
Hofreiter
M.
Serre
D.
Poinar
H. N.
Kuch
M.
Paabo
S.
Ancient DNA
Nat Rev Genet.
 , 
2001
, vol. 
5
 (pg. 
353
-
359
)
Huelsenbeck
J.
Ronquist
F.
MrBayes: Bayesian inference of phylogenetic trees
Bioinformatics
 , 
2001
, vol. 
17
 (pg. 
754
-
755
)
Jacobs
L.
Anyonge
W.
Barry
J. C.
A giant tenrecid from the Miocene of Kenya
J. Mammal.
 , 
1987
, vol. 
68
 (pg. 
10
-
16
)
Jansa
S. A.
Goodman
S. M.
Tucker
P. K.
Molecular phylogeny and biogeography of the native rodents of Madagascar (Muridae: Nesomyinae): A test of the single origin hypothesis
Cladistics
 , 
1999
, vol. 
15
 (pg. 
253
-
270
)
Jansa
S. A.
Weksler
M.
Phylogeny of muroid rodents: Relationships within and among major lineages as determined by IRBP gene sequences
Mol. Phylogenet. Evol.
 , 
2004
, vol. 
31
 (pg. 
256
-
76
)
Kingdon
J.
East African mammals: An atlas of evolution in Africa
 , 
1974
, vol. 
Volume IIA
 
Chicago
The University of Chicago Press
Kingdon
J.
The Kingdon field guide to African mammals
 , 
1997
San Diego
Academic Press
Krause
D. W.
Goodman
S. M.
Benstead
J. P.
Late Cretaceous vertebrates of Madagascar: A window into Gondwanan biogeography at the end of the age of dinosaurs
The natural history of Madagascar
 , 
2003
Chicago
University of Chicago Press
(pg. 
Pages 40
-
74
)
Leung
D. W.
Spencer
S. A.
Cachianes
G.
Hammonds
R. G.
Collins
C.
Henzel
W. J.
Barnard
R.
Waters
M. J.
Wood
W. I.
Growth hormone receptor and serum binding protein: Purification, cloning and expression
Nature
 , 
1987
, vol. 
330
 (pg. 
537
-
543
)
Lewis
P. O.
A likelihood approach to estimating phylogeny from discrete morphological character data
Syst. Biol.
 , 
2001
, vol. 
50
 (pg. 
913
-
925
)
Maddison
D.
Maddison
W.
MacClade 4: Analysis of phylogeny and character evolution
 , 
2000
Sunderland, Massachusetts
Sinauer Associates
Malia
M. J.
Adkins
R. M.
Allard
M. W.
Molecular support for Afrotheria and the polyphyly of Lipotyphla based on analyses of the growth hormone receptor gene
Mol. Phylogenet. Evol.
 , 
2002
, vol. 
24
 (pg. 
91
-
101
)
McKenna
M. C.
Bell
S. K.
Classification of mammals above the species level
 , 
1997
New York
Columbia University Press
Mein
P.
Pickford
M.
Insectivora from Arrisdrift, a basal Middle Miocene locality in southern Namibia
Memoir Geol. Surv. Namibia
 , 
2003
, vol. 
19
 (pg. 
143
-
146
)
Messenger
S. L.
McGuire
J. A.
Morphology, molecules, and the phylogenetics of cetaceans
Syst. Biol.
 , 
1998
, vol. 
47
 (pg. 
90
-
124
)
Morales
J.
Pickford
M.
Salesa
M.
Soria
D.
The systematic status of Kelba Savage 1965: Kenyalutra Schmidt-Kittler 1987: and Ndamathaia Jacobs et al. 1987 (Viverridae, Mammalia) and a review of Early Miocene mongoose-like carnivores of Africa
Ann. Paléontol.
 , 
2000
, vol. 
86
 (pg. 
243
-
251
)
Mouchaty
S. K.
Gullberg
A.
Janke
A.
Arnason
U.
Phylogenetic position of the tenrecs (Mammalia: Tenrecidae) of Madagascar based on analysis of the complete mitochondrial genome sequence of Echinops telfairi
Zool. Scripta
 , 
2000
, vol. 
29
 (pg. 
307
-
317
)
Nicoll
M.
The biology of the giant otter shrew Potamogale velox
Nat. Geogr. Soc. Res. Rep.
 , 
1985
, vol. 
21
 (pg. 
331
-
337
)
Nowak
R. M.
Walker's mammals of the world
 , 
1999
6th edition
Baltimore
Johns Hopkins University Press
Nylander
J. A. A.
MrModelTest 2.1. Program distributed by the author
 , 
2004
Evolutionary Biology Centre, Uppsala University
Nylander
J. A.
Ronquist
F.
Huelsenbeck
J. P.
Nieves-Aldrey
J.-L.
Bayesian analysis of combined data
Syst. Biol.
 , 
2004
, vol. 
53
 (pg. 
47
-
67
)
Olson
L. E.
Goodman
S. M.
Goodman
S. M.
Benstead
J. P.
Phylogeny and biogeography of tenrecs
The natural history of Madagascar
 , 
2003
Chicago
University of Chicago Press
(pg. 
Pages 1235
-
1242
)
Olson
L. E.
Hassanin
A.
Contamination and chimerism are perpetuating the legend of the snake-eating cow with twisted horns (Pseudonovibos spiralis). A case study of the pitfalls of ancient DNA
Mol. Phylogenet. Evol.
 , 
2003
, vol. 
27
 (pg. 
545
-
548
)
Pantel
J.
Machinis
K.
Sobrier
M. L.
Duquesnoy
P.
Goossens
M.
Amselem
S.
Species-specific alternative splice mimicry at the growth hormone receptor locus revealed by the lineage of retroelements during primate evolution
J. Biol. Chem
 , 
2000
, vol. 
275
 (pg. 
18664
-
18669
)
Poduschka
W.
Poduschka
C.
Zur Frage des Gattungsnamens von “Geogale” aletris Butler und Hopwood, 1957 (Mammalia, Insectivora) aus dem Miozän Ostafrikas
Zeitschrift für Säugetierkunde
 , 
1985
, vol. 
50
 (pg. 
129
-
140
)
Poux
C.
Madsen
O.
Marquard
E.
Vieites
D. R.
de Jong
W. W.
Vences
M.
Asynchronous Colonization of Madagascar by the Four Endemic Clades of Primates, Tenrecs, Carnivores, and Rodents as Inferred from Nuclear Genes
Syst. Biol.
 , 
2005
, vol. 
54
 
5
(pg. 
719
-
730
)
Raxworthy
C. J.
Forstner
M. R.
Nussbaum
R. A.
Chameleon radiation by oceanic dispersal
Nature
 , 
2002
, vol. 
415
 (pg. 
784
-
787
)
Rohland
N.
Siedel
H.
Hofreiter
M.
Non-destructive DNA extraction method for mitochondrial DNA analyses of museum specimens
Biotechniques
 , 
2004
, vol. 
36
 (pg. 
814
-
821
)
Salton
J.
Szalay
F. S.
The tarsal complex of Afro-Malagasy Tenrecoidea: A search for phylogenetically meaningful characters
J. Mamm. Evol.
 , 
2004
, vol. 
11
 (pg. 
73
-
104
)
Sánchez-Villagra
M. R.
Asher
R. J.
Cranio-sensory adaptations in small, faunivorous mammals, with special reference to olfaction and the trigeminal system
Mammalia
 , 
2002
, vol. 
66
 (pg. 
93
-
109
)
Smith
W. C.
Kuniyoshi
J.
Talamantes
F.
Mouse serum growth hormone (GH) binding protein has GH receptor extracellular and substituted transmembrane domains
Mol. Endocrinol.
 , 
1989
, vol. 
3
 (pg. 
984
-
990
)
Steppan
S. J.
Adkins
R. M.
Anderson
J.
Phylogeny and divergence—date estimates of rapid radiations in muroid rodents based on multiple nuclear genes
Syst. Biol.
 , 
2004
, vol. 
53
 (pg. 
533
-
553
)
Suárez
A. V.
Tsutsui
N. D.
The value of museum collections for research and society
BioScience
 , 
2004
, vol. 
54
 (pg. 
66
-
74
)
Swofford
D. L.
PAUP, version 4.0b1.0
 , 
2002
Sunderland, Massachusetts
Sinauer Associates
Thompson
J. D.
Gibson
T. J.
Plewniak
F.
Jeanmougin
F.
Higgins
D. G.
The CLUSTAL-X windows interface: Flexible strategies for multiple sequence alignment aided by quality analysis tools
Nucleic Acids Res.
 , 
1997
, vol. 
25
 (pg. 
4876
-
4882
)
van Garderen
E.
van der Poel
H. J.
Swennenhuis
J. F.
Wissink
E. H.
Rutteman
G. R.
Hellmen
E.
Mol
J. A.
Schalken
J. A.
Expression and molecular characterization of the growth hormone receptor in canine mammary tissue and mammary tumors
Endocrinology
 , 
1999
, vol. 
140
 (pg. 
5907
-
5914
)
Veron
G.
La position systématique de Cryptoprocta ferox (Carnivora). Analyse cladistique des caractéres morphologiques de carnivores Aeluroidea actuels et fossiles
Mammalia
 , 
1995
, vol. 
59
 (pg. 
551
-
582
)
Vogel
P.
Contribution a l'ecologie et a la zoogéographie de Micropotamogale lamottei (Mammalia, Tenrecidae)
Revue d'Écologie (La Terre et la Vie)
 , 
1983
, vol. 
38
 (pg. 
37
-
48
)
Wang
X.
Souza
S. C.
Kelder
B.
Cioffi
J. A.
Kopchick
J. J.
A 40-amino acid segment of the growth hormone receptor cytoplasmic domain is essential for GH-induced tyrosine phosphorylated cytosolic proteins
J. Biol. Chem.
 , 
1995
, vol. 
270
 (pg. 
6261
-
6266
)
Yoder
A. D.
The applications and limitations of ontogenetic comparisons for phylogeny reconstruction: the case of the strepsirhine internal carotid artery
J. Hum. Evol.
 , 
1992
, vol. 
23
 (pg. 
183
-
195
)
Yoder
A. D.
Burns
M. D.
Zehr
S.
Delefosse
T.
Veron
G.
Goodman
S. M.
Flynn
J. J.
Single origin of Malagasy Carnivora from an African ancestor
Nature
 , 
2003
, vol. 
421
 (pg. 
734
-
737
)
Yoder
A. D.
Cartmill
M.
Ruvolo
M.
Smith
K.
Vilgalys
R.
Ancient single origin for Malagasy primates
Proc. Nat. Acad. Sci. USA
 , 
1996
, vol. 
93
 (pg. 
5122
-
5126
)
Zakharov
Evgueni V.
Smith
C. R.
Lees
D. C.
Cameron
A.
Vane-Wright
R. I.
Sperling
F. A. H.
Independent gene phylogenies and morphology demonstrate a Malagasy origin for a wide-ranging group of swallowtail butterflies
Evolution
 , 
2004
, vol. 
58
 (pg. 
2763
-
2782
)
Zogopoulos
G.
Nathanielsz
P.
Hendy
G. N.
Goodyer
C. G.
The baboon: A model for the study of primate growth hormone receptor gene expression during development
J. Mol. Endocrinol.
 , 
1999
, vol. 
23
 (pg. 
67
-
75
)

Appendix 1

List of osteological specimens examined. The geographic provenance of specimens is listed in parentheses following the taxon name. Crosses denote extinct taxa; asterisks denote taxa sampled for soft tissue characters using an uncataloged collection of histological specimens (see table 2 of Asher, 2001). Institutional abbreviations are as follows:

  • AMNH

    American Museum of Natural History, New York

  • BMNH

    The Natural History Museum, London

  • FMNH

    Field Museum of Natural History, Chicago

  • IZEA

    Institut de Zoologie et d'Ecologie Animale, Lausanne

  • MCZ

    Museum of Comparative Zoology, Harvard University

  • MNHN

    Museum Nationale d'Histoire Naturelle, Paris

  • USBA

    University at Stony Brook, Department of Anatomical Sciences

  • USNM

    United States National Museum, Washington

  • ZIUT

    Zoologisches Institut, Universität Tübingen

  • ZMB

    Zoologisches Museum Berlin

  1. *Didelphis sp. (Mexico, Nicaragua, USA): AMNH 28408, 28962, 29255, 70082, 145630, 146551, 148959, 201327; USBA MMr1, MMr4, MMr5

  2. *Blarina brevicauda (USA): AMNH 95256, 95297, 98912, 144485, 144486, 144487, 144488, 207017, 207018, 207019, 207020, 207754, 207755, 212504; FMNH 108390, 121194

  3. *Canis latrans (USA): AMNH 5392, 99653, 131833, 131865, 208363, 208367, 208371; USBA MCn28

  4. *Chrysochloris stuhlmani (Zaire, Uganda, South Africa, Burundi): AMNH 82399, 167615, 167963, 180909, 180911, 180912, 180913, 236000; USNM 49896; FMNH 26352, 26353, 26355, 127361, 148200, 148201, 148917

  5. *Crocidura olivieri (Zaire, Malawi, Burundi, Cameroon, Ghana): AMNH 48490, 48491, 48497, 161848, 236229, 239320, 239321, 239326; FMNH 137591, 137592

  6. *Echinops telfairi (Madagascar): AMNH 31270, 100751, 100753, 100760, 100767, 100808, 170602, 170605, 170606, 170607, 170608, 170599, 170609, 170610, 170611, 207717, 207718, 212918, 212919; USNM 464980

  7. *Elephantulus branchyrhynchus (Kenya): AMNH 86554, 86555, 86556, 86557, 86577, 86578, 86580, 86581, 86582, 86583, 86584; ZIUT 3860, 3861

  8. *Erinaceus europaeus (France, Italy, England, Germany): AMNH 3770, 42561, 42563, 57219, 70613, 140469, 140470, 160470, 201230, 215299; USNM 251763, 251764, ZIUT M140

  9. Erythrozootes chamerpes (Kenya): BMNH M14314, M21831; Butler, 1969, 1984

  10. *Geogale aurita (Madagascar): MCZ 37807, 45496, 46274; FMNH 151947, 156551, 15652, 156553, 159732, 159733; MNHN 1912-110A, 1912-110B, 1962-2518, 1962-2519, 1981-1374, 1987-110, 1991-1450

  11. *Hemicentetes semispinosus (Madagascar): AMNH 90421, 100777, 100780, 100781, 100783, 206755, 207711, 207712, 207713, 207714, 212921, 212922, 212935, 212938, 212940; USNM 83658; ZMB 71599

  12. Limnogale mergulus (Madagascar): AMNH 100688, 100689, MCZ 45050, 45054, 45055; BMNH 35.1.8.255, 35.1.8.256, 48.89, 48.90, 97.9.1.161; MNHN 1962-2511, 1962-2513, 1984-521; ZMB 35258

  13. *Microgale talazaci (Madagascar): AMNH 100708, 100709, 100714, 100799, 119216, 207003, 207077; USNM 520881; FMNH 154582, 154583, 154584, 154585, 154586, 154587, 154588, 154589; MNHN 1961-198, 1977-42, 1983-898, 1984-856, 1984-857

  14. *Micropotamogale sp. (Congo, Ivory Coast, Liberia): IZEA 4942, 4975; BMNH 67.213, 73.170; MNHN 1970-514, 1976-397, 1980-52, 1980-53, 1980-57

  15. Oryzorictes sp. (Madagascar): AMNH 31243, 31257; USNM 578789; FMNH 5637, 5639, 5640, 5641, 156226; MNHN 1897-520, 1912-111A, 1912-112A, 1941-34, 1962-2501, 1984-523, 1984-525, 1987-108

  16. *Orycteropus afer (Zaire): AMNH 34866, 51370, 51372, 51374, 51905, 51906, 51907, 51908, 51909, 51010, 70036, 70189

  17. Parageogale aletris (Kenya): BMNH M33046, Butler, 1984

  18. *Potamogale velox (Zaire, Cameroon, Congo, Gabon): AMNH 51161, 51162, 51164, 51165, 51319, 51322, 51324, 51334, 51344, 51348, 51368, 55203, 55204, 120250, 240968; USNM 266897, MCZ 35321, 35322; FMNH 72831, 25973; BMNH 26.11.1.62; MNHN 1892-2064, 1892-2065, 1898-1576, 1947-864, 1947-865, 1947-866, 1962-2520; ZMB 46588

  19. Protenrec sp. (Kenya, Uganda) BMNH M34149, M34150, M33036, M34153, M43551, M43552

  20. *Procavia capensis (Kenya, Zaire, Central African Republic, South Africa): AMNH 53777, 53781, 53784, 53785, 83411, 83412, 80997, 80998, 80999, 88418; USBA Mhy1, Mhy4, Mhy5

  21. *Setifer setosus (Madagascar): AMNH 100749, 100750, 100762, 170532, 170533, 170534, 170535, 170537, 170540, 170547, 170548, 170579, 170581, 170582, 170612, 207005, 207076; USNM 578790; BMNH 93.12.6.8; ZMB 44586

  22. *Sorex sp. (Canada, Finland, England, Sweden): AMNH 126007, 126990, 126991, 141626, 148521, 115593, 115594, 115595, 115597

  23. *Tenrec ecaudatus (Madagascar): AMNH 100729, 100732, 100733, 100735, 100738, 100809, 170502, 170511, 212913; USNM 19361, 577051; BMNH 70.3.10.5