Supplementary information 3. Details of phylogenetic analyses.

As mentioned in the main text, MP and ME trees were constructed using PAUP ver. 4 (Swofford 1998), employing a heuristic search with random addition of sequences and TBR branch swapping procedure. In the MP analysis, cyt-b positions were weighted according to the relative numbers of mutations at different positions (4 for codon position 1, 11 for codon position 2, and 1 for codon position 3) found from inspecting all Asian elephant cyt-b sequences in GENBANK. Weights of 1 were used uniformly for tRNA-Thr, tRNA-Pro, and the D-loop. ME analysis consisted of constructing Neighbour Joining trees from HKY distances (Hasegawa et al. 1985) and a proportion of invariable sites (I = 0.935).

The HKY+I model of sequence evolution fit our data best, as judged by the Bayesian information criterion in the programme DT-ModSel (Minin et al. 2003). A partition homogeneity test showed no significant difference in evolution among the three subsets (cyt-b, t-RNA, D-loop) of data, so data were not partitioned for the Bayesian analyses (carried out using BEAST v.1.4.6; Drummond & Rambaut 2006). Input files were generated using BEAUti v.1.4.6 (BEAST package) with the HKY+I model, relaxed molecular clock (uncorrelated log-normal) assumption (Drummond et al. 2006), and normal priors for HKY kappa (the corrected transition/transversion ratio), I, and the coalescent population size parameter. Normal priors were also used for divergence times between the Loxodonta and Mammuthus-Elephas lineages (mean=7.6 Mya) and between Mammuthus and Elephas (mean=6.7 Mya) based on Rohland et al.’s (2007) findings, and between the a and b clades in the Asian elephant (mean=1.85 Mya, see below). These three calibration points allowed dating of other internal nodes. Optimized scale factors based on initial short runs were used in a Markov Chain Monte Carlo analysis of chain length 10,000,000, with 1,000,000 states discarded as burn-in. Parameters were sampled every 100 iterations. Three independent runs were carried out in BEAST and output analysed using Tracer v.1.3 (Rambaut and Drummond 2004). TreeAnnotator v.1.4.6 (BEAST package) was used to summarize information on posterior node heights.

References:

  1. Drummond, A. J. & Rambaut, A. 2006 BEAST v1.4.6. Available at http://beast.bio.ed.ac.uk/.
  2. Drummond, A. J., Ho, S. Y. W., Phillips, M. J. & Rambaut, A. 2006 Relaxed phylogenetics and dating with confidence. PLoS Biol. 4, 699-710.
  3. Hasegawa, M., Kishino, H. & Yano, T. 1985 Dating the human-ape splitting by a molecular clock of mitochondrial DNA. J. Mol. Evol. 22, 160-174.
  4. Minin, V., Abdo, Z., Joyce, P. & Sullivan, J. 2003 Performance-based selection of likelihood models for phylogeny estimation. Syst. Biol. 52, 674-683.
  5. Rambaut, A. & Drummond, A. J. 2004 Tracer v.1.3. Available at http://tree.bio.ed.ac.uk/software/tracer/.