Cytosine methylation in the genome of Drosophila melanogaster has been elusive and controversial. Despite reports of methylated cytosine peaking in abundance at embryonic Stage 5, its location and function have not been established; two recent whole-genome bisulphite sequencing projects did not find methylation, and concluded that there is none. We used methylcytosine immunoprecipitation and deep sequencing to detect and map genome methylation in Stage 5 Drosophila embryos. We found methylation in ~25,000 highly localized and strand-asymmetrical peaks, limited to regions covering ~1% of the genome, dynamic in early embryogenesis, and concentrated in specific 5-base sequence motifs. At all sites we discovered, methylation is present on only a subset of DNA strands, indicating that it is mosaic at Stage 5. Gene body methylation is associated with lower expression, and many genes containing methylated regions have developmental or transcriptional functions. DNMT2 is the only known DNA methyltransferase in Drosophila, but lines deficient for DNMT2 retain genomic methylation, implying the presence of a novel methyltransferase. Our findings establish that methylation is present at specific sites on the Drosophila genome, and they invalidate conclusions about its function drawn from the phenotype of DNMT2 mutants. The association of methylation with lower expression of specific developmental genes at Stage 5 raises the possibility that it participates in control of gene expression during the maternal-zygotic transition. The methylated sites we have identified are those at which methylation is present in a relatively high proportion of cells at Stage 5; we speculate that other sites are methylated in rare cells at this and other developmental stages.