Genetic techniques and analyses
Up to 30 adult fish were captured at each coastal stream and lake tributary site via backpack electric fishing, anesthetized, and a portion of the caudal fin removed and preserved in 100% ethanol. Genomic DNA was isolated from fin clips following a modification of Casquet, Thebaud, and Gillespie (2012), using 300 µl 5% chelex 100 (Bio-Rad) in 1.5 ml tubes, 40 µg Proteinase K, and a final 90°C heating step for 10 min. Microsatellite genotyping was performed using 10 primer sets developed for Galaxias brevipinnis (Genbank Accession #MK783109-MK783118) and two loci developed for G. vulgaris(Waters, Esa, & Wallis, 1999) that cross-amplify G. brevipinnis . These 12 loci were genotyped using two methods: (1) 8–9% non-denaturing acrylamide gels stained with SYBR Green 1 (ThermoFisher) scored manually against a 10-bp ladder (ThermoFisher) and (2) in multiplex groups (Gbr 02, Gbr 10, Gbr 16 labelled with 6-FAM; Gvu 05, Gvu 07, Gbr 12 with VIC; Gbr 09,Gbr 13 Gbr 22 with NED; Gbr 07, Gbr 14 andGbr 21 with PET) following Schuelke (2000) with modifications described in Townsend, King, and Jamieson (2012). Genotypes were scored using GeneMapper V4.1 (ThermoFisher).
Genetic diversity estimates H o,H e, Ne andF IS were calculated for each sampled population using the GenAlEx excel add on (Peakall & Smouse, 2006). To test for Hardy-Weinberg equilibrium, a Fisher’s Exact test using GenePop (Rousset, 2008; Raymond & Rousset, 1995) was carried out with the default settings and Bonferroni corrections were applied (Sham & Purcell, 2014). Population genetic structuring was determined by pairwise F ST calculated in the GenAlEx excel add on (Peakall & Smouse, 2006). To evaluate potential isolation by distance, pairwise F ST was compared to pairwise geographic distance, measured by water way connection for each category of pairwise comparison (e.g. lake tributary-lake tributary, lake tributary-coastal stream, coastal stream-coastal stream). Significance of the distance patterns was determined using the Mantel test in GenAlEx where possible. Interpretation of genetic results was aided by Bayesian clustering data from the program STRUCTURE 2.3 (Pritchard, Stephens, &Donnelly, 2000). Twenty replicates of a model with a burn in period of 20,000 iterations and run time of 30,000 iterations were run for all numbers of populations K from 1–24. The CLUMPAK online service was used to aggregate STRUCTURE runs and determine the most probable number of populations (Kopelman et al., 2015).