Background Leucine-rich repeat kinase 2 (LRRK2) mutation 6055G→A (Gly2019Ser) accounts for roughly 1% of patients with Parkinson's disease in white populations, 13–30% in Ashkenazi Jewish populations, and 30–40% in North African Arab-Berber populations, although age of onset is variable. Some carriers have early-onset parkinsonism, whereas others remain asymptomatic despite advanced age. We aimed to use a genome-wide approach to identify genetic variability that directly affects LRRK2 Gly2019Ser penetrance. Methods Between 2006 and 2012, we recruited Arab-Berber patients with Parkinson's disease and their family members (aged 18 years or older) at the Mongi Ben Hamida National Institute of Neurology (Tunis, Tunisia). Patients with Parkinson's disease were diagnosed by movement disorder specialists in accordance with the UK Parkinson's Disease Society Brain Bank criteria, without exclusion of familial parkinsonism. LRRK2 carrier status was confirmed by Sanger sequencing or TaqMan SNP assays-on-demand. We did genome-wide linkage analysis using data from multi-incident Arab-Berber families with Parkinson's disease and LRRK2 Gly2019Ser (with both affected and unaffected family members). We assessed Parkinson's disease age of onset both as a categorical variable (dichotomised by median onset) and as a quantitative trait. We used data from another cohort of unrelated Tunisian LRRK2 Gly2019Ser carriers for subsequent locus-specific genotyping and association analyses. Whole-genome sequencing in a subset of 14 unrelated Arab-Berber individuals who were LRRK2 Gly2019Ser carriers (seven with early-onset disease and seven elderly unaffected individuals) subsequently informed imputation and haplotype analyses. We replicated the findings in separate series of LRRK2 Gly2019Ser carriers originating from Algeria, France, Norway, and North America. We also investigated associations between genotype, gene, and protein expression in human striatal tissues and murine LRRK2 Gly2019Ser cortical neurons. Findings Using data from 41 multi-incident Arab-Berber families with Parkinson's disease and LRRK2 Gly2019Ser (150 patients and 103 unaffected family members), we identified significant linkage on chromosome 1q23.3 to 1q24.3 (non-parametric logarithm of odds score 2·9, model-based logarithm of odds score 4·99, θ=0 at D1S2768). In a cohort of unrelated Arab-Berber LRRK2 Gly2019Ser carriers, subsequent association mapping within the linkage region suggested genetic variability within DNM3 as an age-of-onset modifier of disease (n=232; rs2421947; haplotype p=1·07 × 10−7). We found that DNM3 rs2421947 was a haplotype tag for which the median onset of LRRK2 parkinsonism in GG carriers was 12·5 years younger than that of CC carriers (Arab-Berber cohort, hazard ratio [HR] 1·89, 95% CI 1·20–2·98). Replication analyses in separate series from Algeria, France, Norway, and North America (n=263) supported this finding (meta-analysis HR 1·61, 95% CI 1·15–2·27, p=0·02). In human striatum, DNM3 expression varied as a function of rs2421947 genotype, and dynamin-3 localisation was perturbed in murine LRRK2 Gly2019Ser cortical neurons. Interpretation Genetic variability in DNM3 modifies age of onset for LRRK2 Gly2019Ser parkinsonism and informs disease-relevant translational neuroscience. Our results could be useful in genetic counselling for carriers of this mutation and in clinical trial design. Funding The Canada Excellence Research Chairs (CERC), Leading Edge Endowment Fund (LEEF), Don Rix BC Leadership Chair in Genetic Medicine, National Institute on Aging, National Institute of Neurological Disorders and Stroke, the Michael J Fox Foundation, Mayo Foundation, the Roger de Spoelberch Foundation, and GlaxoSmithKline.