A 3.3-kb KpnI repeat unit within the tandem repeat locus (D4Z4) and its upstream 2.5-kb HincII/KpnI fragment of the facioscapulohumeral muscular dystrophy (FSHD) gene region at 4q35-qter were sequenced and characterized. The 3.3-kb KpnI unit was 3303 bp in length and contained two homeodomain sequences, one LSau-like sequence, and several microsatellites. The GC content in the 3.3-kb unit was very high (73%), while it was only 35% in the nonrepeated region of the 2.5-kb fragment. We identified 99% homologous sequences (918 bp) in the 3' end of the 3.3- and 2.5-kb fragments. These findings suggested that the repeated sequences at the D4Z4 start from 918 bp upstream to the first KpnI site.