We show that the time required to transcribe human genes larger than 800 kb spans more than one complete cell cycle, while their transcription speed equals that of smaller genes. Independently of their expression status, we find the long genes to replicate late. Regions of concomitant transcription and replication in late S phase exhibit DNA break hot spots known as common fragile sites (CFSs). This CFS instability depends on the expression of the underlying long genes. We show that RNA:DNA hybrids (R-loops) form at sites of transcription/replication collisions and that RNase H1 functions to suppress CFS instability. In summary, our results show that, on the longest human genes, collisions of the transcription machinery with a replication fork are inevitable, creating R-loops and consequent CFS formation. Functional replication machinery needs to be involved in the resolution of conflicts between transcription and replication machineries to ensure genomic stability.

microRNAs (miRNAs) are generated from long primary (pri-) RNA polymerase II (Pol II)-derived transcripts by two RNase III processing reactions: Drosha cleavage of nuclear pri-miRNAs and Dicer cleavage of cytoplasmic pre-miRNAs. Here we show that Drosha cleavage occurs during transcription acting on both independently transcribed and intron-encoded miRNAs. We also show that both 5'-3' and 3'-5' exonucleases associate with the sites where co-transcriptional Drosha cleavage occurs, promoting intron degradation before splicing. We finally demonstrate that miRNAs can also derive from 3' flanking transcripts of Pol II genes. Our results demonstrate that multiple miRNA-containing transcripts are co-transcriptionally cleaved during their synthesis and suggest that exonucleolytic degradation from Drosha cleavage sites in pre-mRNAs may influence the splicing and maturation of numerous mRNAs.