Fig. 7. Mad and Med bind directly to the bam silencer in vitro. (A) Sequence alignment of the bipartite bam and brk silencers, in which conserved base pairs are boxed. Sites A (red) and B (green) of the bam silencer are as previously described (Chen and McKearin, 2003a). (B) A Cy5 5'-modified oligonucleotide containing the bam silencer and unlabeled competitors in electrophoretic mobility shift assays (competitor A+B, the unlabeled silencer with sites A and B; competitor A, same flanking sequences with site A only; competitor B, same flanking sequences with site B only). (C) Immunoblot analysis of purified recombinant GST-tagged proteins with a mouse anti-GST antibody. Lane 1, 50 ng of GST; lane 2, 200 ng of GST-Mad; lane 3, 200 ng GST-Med. The 30 kDa bands in lane 3 are probably C-terminal degradation products of GST-Med protein. (D) Gel shift assay showing that Mad or Med bind to the bam silencer in vitro. Approximately 10 nmol of protein was used in each binding reaction. The double shift bands for Med may result from partially degraded proteins. The labeled probe without protein (lane 1) or with GST protein (lane 2) serve as negative controls. The labeled probe binds to Mad (lane 3) and Med (lane 7). The unlabeled competitors A+B (lane 4), A (lane 5) or B (lane 6) could effectively compete away Mad binding. The unlabeled competitors A+B (lane 8) or A (lane 9) could effectively compete for Med binding but B (lane 10) could only partially compete. (E) Current model for how Bmp niche signals control GSC identity by directly repressing bam transcription. Bmp signals from cap cells produce the highest levels of pMad, which associates with Med and directly occupies the bam silencer to repress its transcription in GSCs. As a cystoblast moves away from cap cells, levels of pMad are reduced to below the critical threshold level, bam transcription is then derepressed and activated by an unknown activator (indicated by `?').