Fig. 5. AG binds to candidate target genes. (A) Binding sites identified by sequence analysis. `m' is a mutated version of the binding site in AP3, used as a negative control. Next to each numbered binding site, the corresponding gene and sequence are shown, with mismatches to the consensus AG binding site TT(A/T/G)CC(A/T)₆GG(A/T/C)AA (Shiraishi et al., 1993) marked in boldface. The white boxes represent sequences upstream of the start codon (except for AG, where the reference point is the 5' splice site of the second intron), with vertical lines indicating the position of the AG binding sites. The horizontal bars above some of the binding sites indicate the fragment amplified in the ChIP experiment (C). (B) Binding to AG in vitro, shown by electrophoretic mobility shift assays (EMSA). The probes contained the binding sites numbered in A. Each probe was incubated with extract from bacteria induced to express AG (+) or an empty expression vector (-). In all experiments, the same amount of labelled probe was used and a lane with probe 2 (not shown) was included to adjust the exposure to comparable levels. (C) Binding to AG in vivo, shown by ChIP. Numbers correspond to the binding sites shown in A; in each panel, PCR amplification (35 cycles) of sequences containing the binding site (black bars in A) is compared in immunoprecipitates obtained with antiserum against AG (+) or CLV3 (-). In the last panel on the left, the fourth exon of EIF4A1 was used as a negative control lacking AG binding sequences: with 35 cycles (not shown), no band was seen; with 40 cycles (panel), similar levels of contaminating template were amplified. The results shown were replicated in two fully independent ChIP experiments.