Replication has been poor for previously reported candidate genes involved in Major Depressive Disorder (MDD). One possible reason is phenotypic and genetic heterogeneity. The present study replicated genetic associations with MDD as defined in DSM-IV and with a more narrowly defined MDD subtype with a chronic and severe course. We first conducted a systematic review of genetic association studies on MDD published between September 2007 and June 2012 to identify all reported candidate genes. Genetic associations were then tested for all identified single nucleotide polymorphisms (SNPs) and the entire genes using data from the GAIN genome-wide association study (MDD: n = 1,352; chronic MDD subsample: n = 225; controls: n = 1,649). The 1,000 Genomes database was used as reference for imputation. From 157 studies identified inthe literature, 81 studies reported significant associations with MDD, involving 245 polymorphisms in 97 candidate genes, from which we were able to investigate 185 SNPs in 89 genes. We replicated nine candidate SNPs in eight genes for MDD and six in five genes for chronic MDD. However, these were not more than expected by chance. At gene level, we replicated 18 genes for MDD and 17 genes for chronic MDD, both significantly more than expected by chance. We showed that replication rates were improved for MDD compared to a previous, highly similar, replication study based on studies published before 2007. Effect sizes of the SNPs and replication rates of the candidate genes were improved in the chronic subsample compared to the full sample. Nonetheless, replication rates were still poor. © 2015 Wiley Periodicals, Inc.