Transcriptional regulation of protein complexes within and across species

Abstract

Yeast two-hybrid and coimmunoprecipitation experiments have defined large-scale protein-protein interaction networks for many model species. Separately, systematic chromatin immunoprecipitation experiments have enabled the assembly of large networks of transcriptional regulatory interactions. To investigate the functional interplay between these two interaction types, we combined both within a probabilistic framework that models the cell as a network of transcription factors regulating protein complexes. This framework identified 72 putative coregulated complexes in yeast and allowed the prediction of 120 previously uncharacterized transcriptional interactions. Several predictions were tested by new microarray profiles, yielding a confirmation rate (58%) comparable with that of direct immunoprecipitation experiments. Furthermore, we extended our framework to a cross-species setting, identifying 24 coregulated complexes that were conserved between yeast and fly. Analyses of these conserved complexes revealed different conservation levels of their regulators and provided suggestive evidence that protein-protein interaction networks may evolve more slowly than transcriptional interaction networks. Our results demonstrate how multiple molecular interaction types can be integrated toward a global wiring diagram of the cell, and they provide insights into the evolutionary dynamics of protein complex regulation.

Fig. 1.

Coregulated protein clusters in yeast. (a) A typical coregulated cluster and its scoring scheme. Orange ovals, protein cluster members; blue octagons, TFs; orange lines, PPI; blue arrows, TI. (b–i) Representative examples of coregulated protein clusters. Shown are enriched Gene Ontology (GO) biological processes (P < 0.05) of clusters: cell cycle (b); budding (c); cytoplasmic transport (d); cell shape and size regulation (e); mitochondrial membrane transport (f); histone deacetylation (g); hydrogen transport (h); and energy pathways (i).

Fig. 2.

Transcriptional interaction prediction in yeast. (a) Receiver operating characteristics curve of the logistic regression classifier. AUC, area under the curve; Sn, sensitivity; Sp, specificity. (b) An example of a predicted cluster regulated by Rpn4. Orange arrows, known Rpn4 TIs from Harbison et al. (28); purple, newly predicted Rpn4 TIs. Shades of red represent P values (≤0.05) for differential gene expression. (c) Fraction of differentially expressed genes in various gene sets. Green, genes bound by Rpn4 from the Harbison data; orange, genes in cluster models but not bound by Rpn4 based on the Harbison data.

Fig. 3.

Conserved coregulated protein clusters between yeast and fly. (a) A typical conserved coregulated cluster and its scoring scheme. Protein cluster members: orange ovals, yeast; green hexagons, fly. Blue octagons, transcription factors; orange/green lines, PPI; blue arrows, TI. Horizontal dotted links indicate cross-species sequence similarity between proteins (BLAST E value < 10⁻⁷). (b–d) Representative examples of conserved protein clusters across yeast and fly. Conserved GO biological processes of clusters (P < 0.05): protein catabolism (b); cell cycle (c); motor (d). Proteins are connected by direct (thick line) or indirect (connection via a common network neighbor; thin line) protein interactions.

Fig. 4.

Conservation levels of control mechanisms regulating conserved clusters. (a) Both the TF and its DNA-binding motif are conserved. (b) The TF is conserved as an ortholog but not its DNA-binding motif. (c) The TF regulating the conserved complex has likely changed.