Depression is a risk factor for the later development of Alzheimer's disease (AD), but evidence for their genetic relationship is mixed. Assessing depression symptom specific genetic associations may better clarify this relationship. Using data from the UK Biobank, the GLAD Study and PROTECT, we performed the largest genome-wide meta-analyses (GWAS) of the nine depression symptom items, plus their sum score, on the Patient Health Questionnaire (PHQ-9) (GWAS equivalentN: 224,535—308,421). We then assessed global and local genetic correlations and statistical colocalisation between depression/depression symptoms and AD across six AD GWAS with varying proportions of clinical and proxy (family history) case ascertainment. We assessed bi-directional causal associations using Mendelian randomisation (MR) and tested the predictive utility of depression/depression symptom polygenic risk scores (PRS) for AD case/control status in three clinical AD cohorts. Our GWAS meta-analyses identified 37 genomic risk loci across the ten depression symptom phenotypes. Of the 72 global genetic correlation tests conducted between depression/depression symptoms and AD, 20 were significant atpFDR≤ 0.05. Only one significant genetic correlation was identified with AD GWAS containing clinical-only cases. Local genetic correlations were detected at 14 loci. Statistical colocalisation was not identified at these loci but was identified in the region of transmembrane protein 106B (TMEM106B) between multiple depression phenotypes and both clinical-only and clinical+proxy AD. MR and PRS analyses did not yield significant results after multiple testing correction. Our findings do not demonstrate a causal role of depression/depression symptoms on AD and suggest that previous evidence of genetic overlap between depression and AD may be driven by the inclusion of family-history-based proxy cases/controls. However, the identification of colocalisation atTMEM106Bwarrants further investigation.