The organization of the human dystrophin gene into loop domains has been studied using two different experimental approaches: excision of DNA loops mediated by nuclear matrix-bound topoisomerase II and in situ hybridization of different probes with histone-depleted nuclei (nuclear halos). Our objective was to examine if the DNA loops mapped by this biochemical approach coincide with loops visualized by microscopy. The results obtained using both approaches were in good agreement. Eight loops separated by attachment regions of different length were mapped in the upstream part (up to exon 54) of the gene by topoisomerase II-mediated excision. One of these loops was then directly visualized by in situ hybridization of the corresponding bacmid clone with nuclear halos. This is the first direct demonstration that a DNA domain mapped as a loop using a biochemical approach corresponds to a loop visible on cytological preparations. The validity of this result and of the whole map of loop domains was confirmed by in situ hybridization using probes derived from other attachment regions or loops mapped by topoisomerase II-mediated cleavage; these probes hybridized on the core or halo region, respectively, of nuclear halos. Our results demonstrate that a single transcription unit may be organized into several loops and that DNA loop attachment regions may be fairly long. Three out of four replication origins mapped in this gene co-localize with loop attachment regions, and the major deletion hot spot is harbored in an attachment region. These results strongly suggest that partitioning of genomic DNA into specific loops attached to a skeletal structure is a characteristic feature of eukaryotic chromosome organization in interphase.