HISTORICAL BIOGEOGRAPHY OF THE NEW WORLD SOLITAIRES (MYADESTES SPP.)

Gene products were amplified from total genomic DNA as described in previous studies from our group (Hunt et al. 2001). We used the following primers: ATP6&8-COIIGQL, COIIIHMH (Joseph et al. 2004); ND2-MetB (CGAAAATGAT GGTTTAACCCCTTCC), TrC (CGGACTTTAGCA GAAACTAAGAG), and ND2MYAD (sequencing only) (ACAGCCATAAAATTCCCACC) (designed in the Bermingham lab); COI-COIf and COIa (Palumbi 1996); cytochrome b-CBVL14828 (CCA CCCTCCACTCAGGCCTAATCAA) and H16064 (GGAGTCTTCAATCTTTGGTTTACAAGACC) with CB8(1) (GGCCAA ATATCATTTTGAGG) (designed in the Bermingham lab).

Mitochondrial DNA (mtDNA) is sometimes translocated to the nucleus, and nuclear copies of mtDNA genes (pseudogenes) have been amplified by mitochondrial gene primers (e.g., Sorenson and Quinn 1998). We took several precautions to ensure that our sequences were of mitochondrial origin. First, except for five Caribbean individuals, we extracted DNA from mitochondrion-rich muscle tissue rather than blood, thus decreasing the likelihood of amplifying nuclear copies of mtDNA genes. Second, we observed no insertions, deletions, or nonsense codons in the aligned sequences, nor did the electropherograms contain double peaks, which would have indicated co-amplification of nuclear and mitochondrial copies. Third, we detected no significant differences in the rate of evolution between gene regions (see below), which might have resulted from the translocation of part of the sequenced mitochondrial genome to the nucleus. Fourth, we sequenced genes that are widely spaced along the mtDNA molecule, and thus a consistent phylogenetic signal recovered from nuclear DNA would require a translocation of most of the molecule. Sequences were aligned using SEQUENCHER, version 4.0 (Gene Codes, Ann Arbor, Michigan).

For the five-gene data set, a partition homogeneity test (Farris et al. 1995) implemented in PAUP* (Swofford 2002) failed to reject homogeneous phylogenetic signal between the five genes (P = 0.31), so we combined the genes into a single concatenated sequence for further analysis. The ATP6&8 genes overlap by 10 bp including the last three codons of the ATP8 gene and the first three codons of the ATP6 gene. Thus, in the combined data set, we duplicated the overlap (e.g., Hunt et al. 2001) to maintain the first-position reading frame throughout the sequence. For the ATP6&8 data set, a hierarchical likelihood ratio test in MODELTEST, version 3.07 (Posada and Crandall 1998), identified the general time reversal with gamma-distributed rate heterogeneity (GTR+G) model of nucleotide evolution as the best fit to out data. We reconstructed the phylogeny using a Bayesian phylogenetic inference with Markov chain Monte Carlo simulations implemented in MRBAYES, version 3.1 (Ronquist and Huelsenbeck 2003). We used the following parameter settings: a 4 × 4 nucleotide substitution model with six parameters (i.e., general time reversal) with four gamma rate categories using the vertebrate mitochondrial molecular code. We ran MRBAYES for 10^sup 6^ generations, sampling every 100 generations. We graphed the -In likelihoods for the 10,001 sampled generations and found that the results reached stationarity after 3,000 sampled generations. Stationarity was confirmed using the "cumulative" function in the AWTY software package (Wilgenbusch et al. 2004). Therefore, we discarded the first 3,000 sampled generations as "burn-in" and created a consensus topology with the remainder.