Genetic analysis
Genomic DNA was extracted from ~0.2g silica-dried leaf tissue from all 192 genotypes using the Thermo Scientific MagJET Genomic DNA Kit (Thermo Scientific). Double-digest restriction-associated DNA (ddRAD) libraries were prepared using 2-5ng of DNA per sample in 20μL reactions following a modified Peterson et al. (2012) protocol. Briefly, restriction and ligation was carried out simultaneously in 20μL reactions using restriction enzymes MspI and EcoRI and universal adapter sequences for indexing PCR. Ligation products were amplified using 25 cycles of PCR. After indexing, products were checked on an agarose gel and purified. Samples were then pooled and size selected for fragments between 200 and 350bp using a Pippin Prep (Sage Science, Inc., Beverly, MA). The size-selected pool was quantified by qPCR and sequenced on an Illumina MiSeq Desktop Sequencer (Illumina, Inc. San Diego, CA) in 2x75 mode. Sequence reads were processed using a modified Stacks pipeline (Catchen et al. 2013; Andrews 2018). Potential chloroplast and mitochondrial sequences were filtered from the dataset by comparing them to other Populus sequences downloaded from GenBank. Specifically, we removed sequences that matched chloroplast sequences from P. fremontii and mitochondrial sequences from P. tremula x P. alba. Parameter values for clustering were based on tests following parameterization in Mastretta-Yanes et al. (2015). The minimum stack depth for each individual was three and the minimum number of individuals per locus cluster was three. All loci were used in the measure of FST were found to be in Hardy-Weinberg equilibrium. To calculate FST and a 95% confidence interval around FST, we bootstrapped population-level pairwise FST values 500 times using thedivPartCalc function in the R package diversity (Keenanet al. 2013).