## Similar protocols

## Protocol publication

[…] We sampled multiple individuals from each of the major Izu Islands and the small island of Tadanae and from three “mainland” populations including Atami, Izu Peninsula, and Hatsushima, a continental island population off the east coast of the Izu Peninsula (; ). We also sampled two individuals of P. japonicus as outgroups as numerous phylogenetic studies have inferred it as the sister lineage to P. latiscutatus
–, .DNA was isolated from tissue using Qiagen DNeasy columns. We amplified two mitochondrial genes (cyt b and ND1) using primers CB1 and CB6THR for cyt b; newly developed primers ND1-LATF, 5′-CTC TCC CTA ATC ATG CAC CCA TTT TTC AC-3′ and ND1-LATR, 5′-TGA GCT CCT TAG TGC AGG TTC AGA TCC TG-3′ for ND1; and one nuclear gene (R35) using primers R35F and R35R using standard polymerase chain reaction (PCR) techniques (95°C for 2 min followed by 40 cycles of 95°C for 30 s, 55°C for 30 s and 72°C for 60 s). PCR products were cleaned using ExoSap-IT (USB Corp. Ohio, USA). Purified templates were dye-labeled using BigDye (Applied Biosystems, California, USA) and sequenced on an ABI 3077 automated DNA sequencer (Applied Biosystems) using the same primers. Nucleotide sequences were examined and aligned by eye and an open reading frame for these genes were verified using **MacClade** v4.08 . The sizes of the final data sets were 953 bp (cyt b), 957 bp (ND1), and 612 bp (R35) for the number of individuals listed in . R35 sequences with two or more polymorphic sites were phased into individual alleles using Bayesian inference with PHASE 2.1.1 , . [...] Reconstructing colonization history requires knowledge of a taxon’s phylogenetic history and age of lineage divergences. Bayesian phylogenetic analyses assuming a relaxed molecular clock permit the simultaneous estimation of phylogeny, divergence time , , and biogeographic history while also incorporating rate heterogeneity among lineages and phylogenetic uncertainty (and thus, estimates of error) in the tree estimation process. Moreover, these analyses estimate statistical support for phylogenetic and biogeographic reconstructions by calculating Bayesian posterior probabilities.We used **beast** v1.7.3 to estimate the phylogeny, divergence times, and biogeographic history using the combined mtDNA data set (cyt b and ND1) for the 155 sampled individuals of P. latiscutatus. Because assuming different nucleotide substitution models for individual data partitions improves both phylogenetic and divergence time estimation , , we calculated the best partitioning scheme and substitution models for the codon positions of each gene using **Partitionfinder** v1.0.1 . Partitionfinder recommended three total partitions: cyt b codon position one, ND1 two; cyt b two, ND1 three; cyt b three, ND1 one and the substitution models TrN + G for partitions one and three, and HKY + G for partition two.Estimating divergence times from molecular data requires some a priori estimate ages for at least one divergence. These are commonly estimated by incorporating fossil taxa as age constraints to “calibrate” the relaxed molecular clock. However, there are no known Plestiodon fossils that can be used as calibration age constraints. We therefore used the age distribution of the most recent common ancestor of P. japonicus and P. latiscutatus inferred by a multi-locus time-calibrated phylogenetic analysis of Plestiodon
as our age calibration constraint. Although secondary calibrations have been rightly criticized for potentially compounding date estimation error , we note that Bayesian age estimation permits the explicit incorporation of this error by permitting age calibration constraints (rather than point estimates) in the form of statistical distributions, thus eliminating at least one negative feature of secondary calibrations . Also, the age distribution is broad and thus likely not overly precise (see Results).Simultaneously with estimating phylogeny and divergence times, we inferred ancestral biogeographic area using the discrete traits model of Lemey et al. . We coded all P. latiscutatus individuals into groups representing each island or mainland peninsula populations.Each BEAST analysis was run for 107 generations and sampled every 2000 generations. We modeled the age of the root of the tree (P. japonicas+P. latiscutatus) as a normal distribution of ages with a mean = 6.3 Ma and standard deviation = 1.38 (95% CI = 3.6–9.0 Ma; ) and enforced a separate lognormal relaxed molecular clock for the cyt b and ND1 data. We otherwise used default priors except that we modeled the mean rate of the cyt b and ND1 molecular clocks at uniform distributions with bounds of 0.0 and 0.1 substitutions per site. We ran eight BEAST analyses assuming a birth-death tree prior. To determine convergence amongst each analysis, we constructed cumulative posterior probability plots for each clade using the cumulative function in **AWTY** . Stationarity was assumed when the cumulative posterior probabilities of all clades stabilized. If posterior probability estimates for clades were similar in the analyses, the results were combined. We interpret posterior probabilities ≥0.95 as suitably strong support for both phylogenetic reconstruction and estimation of ancestral biogeographic area . [...] We performed phylogenetic comparative analyses to estimate correlations amongst life history characteristics (correlations of phylogenetic independent contrasts) and then linked those life history traits to the type of predators on multiple Izu Islands (phylogenetic ANOVA). For statistical analysis, arithmetic means of the sampled life history and phenotype characters were converted to independent contrasts to remove non-independence due to phylogenetic history under different phylogenetic scenarios (see below). Correlations of independent contrasts of life history traits were performed using the CAPER package in R using scripts written by M.C.B. We conducted phylogenetic generalized least squares (PGLS) analysis of variance (phylogenetic ANOVA) analyses to determine how variation in life history traits correlates to the presence of predators on each island after accounting for phylogenetic relationships. There exist three major classes of skink predators on the Izu Islands including weasels, birds, and snakes, but we focus only on the potential influence of snake and bird predation on life history variation. Unlike weasels that are historically native to Oshima (they were subsequently introduced into Toshima in the 1930s, Hachijojima in the 1960s, Miyake and Aogashima in the 1980s), and predatory birds that inhabit all islands, snakes inhabit eight of the 11 islands sampled for this study and likely derive from independent colonizations . Therefore, snakes offer multiple independent opportunities to assess the affects on the presence of different predators on the life history traits of their prey, P. latiscutatus. Moreover, snake-less islands offer an opportunity to assess the effects of bird-only predation on P. latiscutatus life history and color evolution.Phylogenetic ANOVA (pANOVA hereafter) analyses were performed using the **GEIGER** package in R and scripts written by M.C.B. We regressed the mean values of both hatchling and maternal life history traits against a dummy variable coded 0 or 1 (0 = island has snakes; 1 = island does not have snakes). The overall correlation coefficient (R) represents deviations from the mean of the comparison group and was tested for significance using a t-test .Because both independent contrasts and phylogenetic ANOVA analyses use phylogenetic information to remove the effect of non-independence caused by the organisms’ shared evolutionary history, the results are therefore fundamentally reliant on the underlying phylogeny. Inspection of the mtDNA phylogeny () reveals multiple populations that are not monophyletic including the mainland, Miyake, Niijima, Shikine, and Toshima populations. This pattern could be indicative of multiple colonizations to these islands (i.e., the pattern represents the true colonization history), or it could result simply from a stochastic process where drift has not eliminated older mtDNA haplotypes such as incomplete lineage sorting (i.e., the pattern is an artifact of molecular evolution). Incomplete lineage sorting is a phenomenon where ancestral alleles (or haplotypes, in the case of mtDNA) that are present before a lineage splits (i.e., when two or more populations are reproductively isolated) are retained in its descendant lineages after the divergence . These ancestral alleles will be lost to genetic drift over time and replaced by new alleles unique to the new descendant lineages; however, during short time frames, there is a chance that these alleles will not be lost due to drift (e.g., –). The distinction between these two processes (multiple colonizations and incomplete lineage sorting) has significant effects on how we interpret life history evolution. We assume that our sampled life history data represents the frequency of these traits on each island (e.g., the mean hatchling SVL for the entire population of Aogashima is 30.4±1.3 mm; ). If two or more lineages colonized an island, then it would suggest that each lineage convergently evolved the same distributions of life history traits, thereby increasing the strength of correlations amongst traits in subsequent comparative analysis. This is not problematic if indeed the islands were colonized multiple times. However, if the non-monophyly of island populations is simply an artifact of molecular evolution, and that each island was indeed colonized only once, then assuming multiple colonizations would create false positive support for our trait correlation analyses.We therefore performed the independent contrast and pANOVA comparative analyses, assuming both the multiple colonization scenario supported by our inferred phylogeny, and single colonization scenarios that assume the presence of two distinct lineages on an island is an artifact of incomplete lineage sorting. For the multiple colonization scenario, we performed the comparative analyses using a phylogeny pruned to the minimum number of possible independent colonizations. Assuming a single colonization to each island poses a challenge because there are multiple possible resolutions of our phylogeny compatible with a single colonization scenario (e.g., Toshima may be sister to a mainland or Niijima + Shikine clade []). We therefore developed a novel method that accounts for the multiple possible resolutions whereby we sample a tree from the posterior distribution of trees estimated by the BEAST analysis, and prune the tree to include only one randomly chosen individual per island. We then use this tree for the IC and pANOVA analyses. We repeated this process for all 8000 trees in the posterior distribution thereby creating a distribution of p values from each test using R scripts written by M.C.B. Because there is no objective way to interpret distributions of p values, we subjectively interpret the results assuming that a distribution of p values that cluster closer to 0 are more suggestive of a statistical relationship between two variables than a distribution that includes mid- to high p values. We note that the tree topology of the pruned tree will depend on which of the multiple island lineages are retained (i.e., sampling any one of the Toshima lineages would change the topology of a single tree). However, because this process is random across the 8000 trees, we assume there is no systematic bias. Indeed, repeating the analysis with different random number seeds yields identical results (not shown). We also emphasize that this method also incorporates phylogenetic uncertainty as not all clades have a posterior probability of 1.0. Life history data, DNA data, and R scripts are available from the Dryad Digital Repository: doi:10.5061/dryad.v47s1. […]

## Pipeline specifications

Software tools | MacClade, BEAST, PartitionFinder, AWTY, GEIGER |
---|---|

Application | Phylogenetics |

Organisms | Mus musculus |

Diseases | Leukemia, Biphenotypic, Acute |