Background: In utero exposure to arsenic (As) is associated with adverse birth outcomes; this relationship may be mediated by epigenetic dysregulation. We evaluated the association between maternal prenatal As exposure, gestational age (GA), and birth weight (BW), and mediation by DNA methylation (DNAm).Methods: In a prospective birth cohort of pregnant women in Bangladesh exposed to As mainly through drinking water, maternal drinking water As was measured at ≤16 weeks GA, and toenail As was measured at <1 month postpartum. In a discovery phase, DNAm in cord blood was measured using the Infinium 450K array (N=44). In a validation sample of participants of the same cohort (N=570), cord blood DNAm was measured using bisulfite pyrosequencing for candidate CpG in miR124-3 (cg20277905), GNAL (cg06522054), and MCC (cg24937280) genes.Results: In the discovery phase, maternal water As was associated with DNAm of 396 CpGs (P<1.10X10-6; |β regression|>0.10). These CpGs were evaluated for associations with birth outcomes; 139 were associated with GA and one with BW (P<0.05). The main principal component of the top-ten CpGs fully mediated the inverse association between maternal prenatal water As and GA in a structural equation model (SEM) (P<0.001) but not for BW. In the validation phase, DNAm of miR124-3 and MCC mediated the association between maternal toenail As and GA (miR124-3 P=0.003; MCC P=0.006). In an adjusted SEM including miR124-3 and MCC, the association between As and GA was mediated by DMAm of miR124-3 but not of MCC (miR124-3 β=-0.04; P=0.02 MCC P=0.23).Conclusions: miR124-3 is a microRNA coding gene that may be involved in embryonic growth, and in mouse models prenatal As exposure has been associated with miR124-3 expression. Using a discovery and validation approach, we show that DNAm of miR124-3 may mediate the association between prenatal As exposure and GA.