PRC2/EZH2β governs a unique repressive program compared to conventional PRC2/EZH2α. (A) Left: Genome-wide expression profiling was performed using epithelial cells transduced with EZH2α and EZH2β using Affymetrix Human Gene 1.0 ST Array. Significantly altered probes (P <0.05) were selected and visualized by cluster analysis. Right: A subset of known and well-characterized Polycomb targets were assessed by chromatin immunoprecipitation array using an antibody against HIS-epitope-tagged EZH2 isoforms in epithelial cells transduced with either the empty vector, HIS/EZH2β or HIS/EZH2α. Levels of binding were normalized to input controls for each of the three conditions and fold changes computed against empty vector control. Fold changes are presented according to percentile rank from 0 (unbound) to light blue (>90% percentile) of the isoform dataset. ChIP experiments were performed in duplicate with a representative dataset shown above. Targets identified as EZHβ- or EZH2α-specific from the whole genome experiment in Figure5A-left are boxed and labeled. Comparison with expression data generated by qPCR from the same samples reveals that the majority of the target bounds by each isoform are repressed. (B) Comparative quantification of the percentage of uniquely repressed and activated gene targets regulated by each isoform individually or both isoforms is indicated (P <0.05, log2 fold change >−2). The ontological classification of targets uniquely repressed by each isoform individually or in conjunction with the other isoform is also shown. (C) Comparative quantification of the percentage of uniquely repressed and activated gene targets against a subset of canonical EZH2 targets as determined by an independent ChIP-seq dataset that used an antibody predicted to cross-react with multiple EZH2 isoforms. ChIP: chromatin immunoprecipitation; EV: empty vector; EZH2: enhancer of zeste homologue 2; qPCR: quantitative polymerase chain reaction.