Several studies have investigated general properties of the genetic architecture of gene expression variation. Most of these used controlled crosses and it is unclear whether their findings extend to natural populations. Furthermore, systems biology has established that biological networks are buffered against large effect mutations, but there remains little data resolving this with natural variation of gene expression. Here we utilise RNA-Sequencing to assay gene expression in winter buds undergoing bud flush in a natural population of Populus tremula. We performed expression Quantitative Trait Locus (eQTL) mapping and identified 164,290 significant eQTLs associating 6,241 unique genes (eGenes) with 147,419 unique SNPs (eSNPs). We found approximately four times as many local as distant eQTLs, which had significantly higher effect size. eQTLs were primarily located in regulatory regions of genes (UTRs or flanking regions), regardless of whether they were local or distant. We used the gene expression data to infer a co-expression network and investigated to what degree eQTLs could explain the structure of the network: eGenes were present in the core of 28 of 38 network modules, however, eGenes were overall underrepresented in cores and overrepresented in the periphery of the network, with a negative correlation between eQTL effect size and network connectivity. We also observed a negative correlation between eQTL effect size and allele frequency and found that core genes have experienced stronger selective constraint. Our integrated genetics and genomics results suggest that prevalent purifying selection is the primary mechanism underlying the genetic architecture of natural variation in gene expression in P. tremula and that highly connected network hubs are buffered against deleterious effects as a result of regulation by numerous eSNPs, each of minor effect.