使用遗传环境关联(GEA)准确检测局部适应的特征需要控制种群结构的中性模式以降低假阳性的风险。然而,气候梯度和中性种群结构之间的高度共线性会大大降低功率,在这种情况下,GEA方法的性能很少在实证研究中进行评估。在这项研究中,我们试图使用两种白松物种的宽范围样本,将局部适应和环境隔离(IBE)与远距离隔离(IBD)和冰川避难所(IBC)定植隔离(IBC)的影响分开。为此,来自168个基因的SNP,包括52个生长和物候的候选基因,在133和61个群体中进行了基因分型,分别。对于P.strobus和使用所有153个SNP,在冗余分析(RDA)中控制IBD和IBC时,气候(IBE)并未显着解释种群间差异。然而,在单基因座GEA分析(Bayenv2和LFMM)中,26个SNP与气候显著相关,表明局部适应发生在高基因流的存在下。对于P.Monticola,我们没有发现使用RDA的IBE的证据,也没有发现使用GEA和FST异常值测试的局部适应性特征较弱的证据,与通过表型可塑性的适应一致。在这两个物种中,大多数解释的种群间差异(69%至96%)无法在IBE的影响之间进行划分,IBD,IBCGEA方法可以不同地解释这种混淆的变化,这可以解释在Bayenv2和LFMM之间检测到的SNP的小重叠。我们的研究说明了考虑自然种群中性结构的内在困难,以及最大化气候变化的采样设计的重要性,同时最小化气候梯度和中性结构之间的共线性。
Accurately detecting signatures of local adaptation using genetic-environment associations (GEAs) requires controlling for neutral patterns of population structure to reduce the risk of false positives. However, a high degree of collinearity between climatic gradients and neutral population structure can greatly reduce power, and the performance of GEA methods in such
case is rarely evaluated in empirical studies. In this study, we attempted to disentangle the effects of local adaptation and isolation by environment (IBE) from those of isolation by distance (IBD) and isolation by colonization from glacial refugia (IBC) using range-wide samples in two white pine species. For this, SNPs from 168 genes, including 52 candidate genes for growth and phenology, were genotyped in 133 and 61 populations of Pinus strobus and P. monticola, respectively. For P. strobus and using all 153 SNPs, climate (IBE) did not significantly explained among-population variation when controlling for IBD and IBC in redundancy analyses (RDAs). However, 26 SNPs were significantly associated with climate in single-locus GEA analyses (Bayenv2 and LFMM), suggesting that local adaptation took place in the presence of high gene flow. For P. monticola, we found no evidence of IBE using RDAs and weaker signatures of local adaptation using GEA and FST outlier tests, consistent with adaptation via phenotypic plasticity. In both species, the majority of the explained among-population variation (69 to 96%) could not be partitioned between the effects of IBE, IBD, and IBC. GEA methods can account differently for this confounded variation, and this could explain the small overlap of SNPs detected between Bayenv2 and LFMM. Our study illustrates the inherent difficulty of taking into account neutral structure in natural populations and the importance of sampling designs that maximize climatic variation, while minimizing collinearity between climatic gradients and neutral structure.