Contributions to Zoology, 77 (1) – 2008Ross Barnett; Nobuyuki Yamaguchi; Beth Shapiro; Richard Sabin: Ancient DNA analysis indicates the first English lions originated from North Africa

To refer to this article use this url:

Materials and methods

Sampling and laboratory procedures

next section

Small pieces of cortical bone (c. 5 × 5 × 2 mm) were sampled from the mandibles of the two Tower lion skulls (registration numbers NHM1952.10.20.15 and NHM1952.10.20.16) at the Natural History Museum, London, UK. Laboratory procedures were carried out as described in Barnett et al. (2006a) at the Henry Wellcome Ancient Biomolecules Centre (ABC), Oxford University, which is geographically isolated from modern molecular biology work and DNA amplification by polymerase chain reaction (PCR).

Data authenticity and analysis

Extraction of specimens NHM1952.10.20.15 and NHM 1952.10.20.16 took place in the ABC and was performed along with negative extraction controls. PCR amplification of a small hypervariable fragment of the mitochondrial control region was performed twice for each sample, each time incorporating negative amplification controls. The four resulting amplification products were then cloned using the TOPO TA system (Invitrogen, Carlsbad CA USA), sequenced on ABI377 automated sequencers (Perkin-Elmer, Wellesley MA USA), and aligned with previously published lion sequences (Barnett et al., 2006a, 2006b). A summary of the cloning results is presented in the Appendix. A total of 12 clones were sequenced from sample NHM 1952.10.20.15, and 10 from NHM1952. 10.20.16. Of these, only three sequences show evidence of postmortem DNA damage (E4, E6, and F5 in Table S1) and, in each instance, the damage occurs at nucleotide sites that are not known to be polymorphic in lions. A median-joining network was constructed from the resulting sequences using Network v4.1.0.3 (Bandelt et al., 1999).

Skull measurements

To investigate the origins of the Tower lions independently of the molecular results, morphological investigation was undertaken using Asiatic and North African Barbary lion skulls kept in natural history collections in the UK and Europe. A skull was classified as subadult if cemento-enamel junctions of all canines were already visible above the alveoli of the cleaned skull and yet the basioccipital-basisphenoid suture, and/or frontal suture, was still open. If those sutures were closed, a skull was classified adult. Seventy five craniometric measurements were taken of the cranium and mandible, modified from Yamaguchi et al. (2004), using a metal caliper to the nearest 0.02 mm, except for those of 10 larger variables that were measured to the nearest 0.05 mm using a larger metal calliper (for details see Appendix). To test the measurement errors, five skulls were randomly selected and each measurement was taken three times on each skull. The coefficient of variation for each of the 75 variables was calculated, and the variables with average coefficient of variations of more than 2% were excluded from the analysis by accepting the arbitrary cut-off line for reliability and consistency in measurements used in Yamaguchi et al. (2004) (see Appendix). We have measured all Asiatic and North African Barbary lion skulls kept, and available for measurement, in major natural history collections in the UK and Europe. However, as not every skull was intact, some variables were excluded from the analysis for maximising both Asiatic and North African Barbary lion specimens to be included into the analysis whilst retaining as many variables as possible. We retained 57 variables with 23 individuals (Table 1).

Statistics for morphological analysis

All statistical analyses were carried out using the SPSS statistics package (version 13: SPSS Inc., Chicago, USA). A principal component analysis (PCA) was carried out to reduce the numbers of variables for the subsequent analyses, which were based on extracted principal components whose eigenvalues were larger than 1 (Tabachnick and Fidell, 2007). Then, a discriminant analyses (DA) was carried out to investigate if Asiatic and North African Barbary lion skulls could be distinguished, with the prior probabilities computed from the group sizes. A DA is designed to develop classification functions to classify each specimen best by following a priori groupings, so that it will usually result in a fairly good discrimination between the groups (Tabachnick and Fidell, 2007). Therefore, a cross-validation test was also carried out to check which group each case would be classified into if it was classified by the functions derived from all cases other than itself. Then, we tested if the Tower lions would be classified as either Asiatic or North African Barbary, or both.


Table 1. Lion skulls used for the morphological analysis. Museums are Natural History Museum London, Muséum National d’Histoire Naturelle Paris, Museum für Naturkunde der Humboldt-Universität, Berlin, Natural History Museum, University of Oxford, For-schungsinstitut und Naturmuseum Senckenberg, Frankfurt, Musée Zoologique, Strasbourg, and female is indicated by (f), male (m), adult (a), and subadult (sa). Individual ID numbers (e.g. P1, L1 or T1) are corresponding to those in Fig 2.

In addition to sexual size dimorphism that is common in the Felidae, it has been suggested that the skull morphological characteristics of captive lions differ from those of wild animals (Hollister, 1917). While a preliminary morphometric analysis suggested that both Tower lions were males based on the greatest length of skull and canine size (Gittleman and Van Valkenburgh, 1997), specimens did not have their sex recorded. Additionally, the specimens had spent at least some time in captivity. We therefore included into the analysis both male and female, and both captive and wild, individuals in our comparative data set (see Table 1). We deliberately did so to find out if a discriminant analysis (DA) would be able to separate the North African Barbary lion from the Asiatic lion regardless of sex and whether an animal was captive or wild.


Table 2. Classification results obtained by a discriminant analysis. The only one misclassified case in the cross-validation test was P6 in Table 1.