^{a}

^{a}

^{b}

^{d}

^{c}

^{a}

^{d}

^{a}

Relationships of the newly discovered dwarf crocodiles from Mauritania were inferred from mitochondrial 12S sequences. Specimens from 13 different

There are three extant species of crocodilian currently known to exist in Africa. These are comprised of the endemic and monotypic dwarf crocodile (

Nile crocodiles have disappeared from many parts of their historic range, particularly in northern Africa and in the Sahara desert

Presented here are the first mtDNA sequence data comparing the relict populations to other extant Nile crocodile populations. Initial results suggest significant genetic differentiation between relict and Eastern Nile crocodile populations. However, data from additional West African populations suggest an even more complex evolutionary history for

To examine the genetic variation within

DNA was extracted from liver or muscle tissue (either fresh, preserved in 98% ethanol, or dried for museum specimens) using QuiAmp tissue extraction kits (Qiagen). We used the primers 12SA-L (light chain; 5′ - AAA CTG GGA TTA GAT ACC CCA CTA T - 3′) and 12SB-H (heavy chain; 5′ - GAG GGT GAC GGG CGG TGT GT - 3′)

PCR products were purified using Qiaquick purification kits (Qiagen). Sequences were obtained using an automatic sequencer (ABI 377) and manually corrected using the computer program Sequence Navigator (Applied Biosystems).

Sequences were aligned using the computer program ClustalX

Prior to phylogenetic reconstructions, we tested for homogeneity of base frequencies among taxa using the χ^{2} test as implemented in PAUP*4.0b10 (which ignores correlation due to phylogenetic structure): (1) over all sites; (2) over parsimony-informative sites only; (3) without constant sites (parsimony-uninformative and constant sites will mislead the χ^{2}) test

Confidence in the phylogenetic signal for this molecular data set was assessed by calculating the skewness, or g1 statistic (implemented in PAUP*), which provides a measure of phylogenetic information content

A matrix of pairwise sequence differences for the 12S rRNA genes was calculated using the p-distance (

All Bayesian

To select the best-fit model of nucleotide substitution for our data set, the hierarchical likelihood-ratio test was carried out using MRMODELTEST 1.1b ^{6} generations each. Each chain consisted of one cold and three heated chains and the Markov chains were started from a random tree. The Markov chains were sampled every 100th generation, resulting in 10 000 sampled trees from each chain. The initial 1000 (10%) trees were disregarded as ‘burn-in’ (the portion of the chain that was sampled before stationarity was reached). Inferences, then, were based on the 9000 trees samples from each chain. The topologies were used to generate a strict-consensus tree, with the percentage of samples recovering any particular clade representing that clade’s posterior probability

The obtained 12S sequences (lengths referring to the aligned sequences including gaps) comprised 433 bp, with the exception of the tissue samples taken from populations from Senegal and Chad, which had strongly degenerated, making it impossible to get complete sequences from these samples. Nonetheless, we managed to get shorter fragments of the populations sequenced (195 bp for the Chad sample and 78 bp for the Senegal sample) (

In the data set a phylogenetic signal is clearly present (g1 = -1.8185, p = 0.01). When all characters were included, we found no significant deviation from the homogeneity of base frequencies among taxa (χ^{2} = 11.5660, p = 1.0000, df = 51). The same was true for the parsimony-informative sites only (χ^{2} = 46.8861, p = 0.6377, df = 51) and without constant sites (χ^{2} = 32.2954, p = 0.9739, df = 51).

Likelihood scores for models examined under MRMODELTEST showed that the GTR + G model _{(A)} = 0.3138, π_{(T)} = 0.2117, π_{(C)} = 0.2721, π_{(G)} = 0.2023], and a gamma distribution shape parameter (α = 0.2931).

Both the ML and the Bayesian approaches produced identical topologies.

Three major clades are evident. The first clade comprises the two included

On the genus level we find genetic differentiations (

Our results demonstrate that the African crocodiles are genetically much more diverse than previously thought. Both the ML and Bayesian analyses revealed the existence of two independent clades of

The inclusion of

The very low internal genetic variation of 0.0–0.7% found within the eastern

The level of distinctness indicated by these results suggests that taxonomic revision of

The geographic split between the lineages runs roughly through an area that was strongly influenced by the repeated extensions and regressions of the Sahara desert, particularly during the Pliocene and Pleistocene

A final aspect important to note is that both ML and Bayesian inference consistently placed

“

We thank Nicolà Lutzmann (ZFMK), Hans-Werner Hermann (Zoological Garden Cologne), Thomas Wilms (SMN, Karlsruhe), Rudolf Wicker (Zoological Garden Frankfurt), Frank Glaw (ZSM, Munich), Miguel Vences (University of Amsterdam), the “Ferme aux crocodiles” (Pierrelatte), Michael Klemens (Wildlife Conservation Society), The American Museum of Natural History (AMNH), and The Field Museum of Natural History (FMNH), who provided samples for this study. The second author is indebted to John P. Huelsenbeck, Fredrik Ronquist, Jonathan P. Bollback and Johan Nylander for their introduction to the MrBayes program.

Map of sample localities for specimens used in this study. Locality names are as follows: (1) Senegal, (2) Gambia, (3) Aioun el-Atrouss, Mauritania, (4) Liberia, (5) Edéa, Cameroon, (6) Lambaréné, Gabon, (7) Ennedi Mts., Chad, (8) Lake Nasser, Egypt, (9) Chor Melk en-Nasir, Sudan, (10) Kenya, (11) Kariba Dam, Zimbabwe, (12) Natal, South Africa, (13) Madagascar. Where the exact locality of the voucher is not known, we placed the corresponding number at a place where crocodiles are known to occur in the given country. The three desert localities are indicated by black numbers.

Fig. 2. Carte des localités ayant fourni des spécimens utilisés dans cette étude. Les noms des localités sont les suivantes : (1) Sénégal, (2) Gambie, (3) Aïoun el-Atrouss, Mauritanie, (4) Libéria, (5) Edéa, Cameroun, (6) Lambaréné, Gabon, (7) massif de l'Ennedi, Tchad, (8) Lac Nasser, (9) Chor Melk en-Nasir, Soudan, (10) Kenya, (11) Barrage de Kariba, Zimbabwe, (12) Natal, Afrique du Sud, (13) Madagascar. Lorsque la localité exacte de la pièce n’est pas connue, nous avons placé le numéro correspondant à l’endroit où les crocodiles sont connus dans la région. Les trois localités du désert sont indiquées par des numéros en noir.

Phylogram of the maximum likelihood tree (ML) and Bayesian analysis tree obtained from PAUP* and MrBayes searches using

Fig. 1. Phylogramme de l’arbre de probabilité maximale (ML) et arbre de l’analyse bayésienne obtenus avec PAUP* et avec les recherches de MrBayes en utilisant

List of voucher specimens for each species included in the present study, with their respective localities, collection numbers and accession numbers (12S)

Liste des spécimens de chaque espèce inclus dans la présente étude, avec leurs localités respectives, leurs numéros de collection et leurs numéros d’accession (12S)

Summary of the uncorrected p-distances (sequences excluded from the tree are marked)

Sommaire des distances p, avant correction (des séquences exclues de l'arbre sont indiquées)

Parameter estimates of the substitution model (GTR + G), sampled after the burn-in phase of the chain. The columns indicate the parameter, mean and 95% credible interval for the parameter. The parameters are TL, the tree length; _{ij}
_{GT} = 1); _{i}, base frequencies; and

Paramètre estimé du modèle de substitution (GTR+G), échantillonné après la phase de combustion de la chaîne. Les colonnes indiquent le paramètre, la moyenne et l’intervalle de 95% de crédibilité pour le paramètre. Les paramètres sont TL, la longueur de l’arbre ; _{ij}
_{GT} = 1) ; _{i}, fréquences des bases ;

_{GT}

_{CT}

_{CG}

_{AT}

_{AG}

_{AC}

_{A}

_{C}

_{G}

_{T}