Selecting DNA binding site motifs for our curated collection. (a) The in vitro motifs for TF Cin5 are very similar, but not equally enriched in the ChIP-chip data (see main text). (b) The Cst6 in vitro motif reported by Badis et al. [10] is a truncated version of the Cst6 motif reported in this study. The latter better matches TGACGTCA, the known site for the ATF/CREB family of bZIP TFs, of which Cst6 is a member. (c) For TF Ecm22 we selected the motif obtained in this study (which is different from the motif previously reported by Badis et al. [10]). The selected motif matches the sterol regulatory element TCGTATA, which had been reported to be bound by Ecm22 (and also its close paralog, Upc2). N/A, not available in Fordyce et al. [12].

Examples of TFs for which the in vitro and in vivo DNA binding site motifs are different. For both the in vitro and in vivo motifs of the three TFs we show their enrichment in the corresponding ChIP-chip data set, measured by the AUC and the associated P-value. We also show the in vitro motifs (from our curated collection) that are most similar to the in vitro and in vivo motifs of the three TFs of interest (the red lines indicate which parts of the motifs are similar). We notice that in all three cases the in vivo motifs are similar to the DNA binding site motifs of TFs from a different structural class. This suggests that in each of the three cases the in vivo motif (derived from ChIP-chip data) does not belong to the TF profiled by ChIP, but either to a co-regulatory TF (which binds a common set of targets as the profiled factor), or to a mediating TF (which binds DNA directly and mediates the interaction between the TF profiled by ChIP and the DNA - in this case we hypothesize that the TF tested by ChIP binds DNA indirectly thought the mediating TF). Motif sources from this study and Zhu et al. [11], Badis et al. [10], and MacIsaac et al. [20].

bZIP and VHR TFs. (a) Phylogeny and PBM-derived motifs for the eight bZIP and two VHR proteins analyzed in this study. The evolutionary tree was built from a ClustalW2 [59] multiple sequence alignment of the DBDs of the ten proteins, as annotated in UniProt [60]. Green and magenta backgrounds correspond to TFs that bind primarily to overlapping or adjacent half-sites, respectively. TFs that bind Yap-like half-sites are shown in red. TFs that bind Gcn4-like half-sites are shown in blue. All motif logos were generated using EnoLOGOS [58], based on motifs generated from PBM data in this study and Zhu et al. [11] using the Seed-and-Wobble algorithm [8,15]. The numbers next to the motif logos represent the E-scores of the 8-mer seeds used to construct the motifs [8]. For proteins that bind both overlapping and adjacent half-sites, the motif corresponding to the largest seed E-score (sometimes referred to as the primary motif) is shown in a black box. (b) ClustalW2 multiple sequence alignment of the basic regions of bZIP proteins against the DBDs of VHRs. The Vhr1 and Vhr2 regions shown are the ones that best align to the eight basic regions considered, and they correspond to the first putative VHR basic region (see (e)). The residues shown in red and blue are important for YAP-like versus Gcn4-like half-site specificity. The residues shown in green and magenta are important for overlapping versus adjacent half-site binding. (c) Recognition of Yap-like and Gcn4-like half-sites [30,61]. (d) Heat map of the DNA-binding preferences of Yap1 (as a representative of the Yap subfamily), Cst6, Sko1, Gcn4, Vhr1, and Vhr2. The rows correspond to 8-mers with an E-score ≥0.35 for any of the six TFs; the columns correspond to the TFs. The E-score scale is shown at the bottom. Black boxes indicate the 8-mers that correspond to various motifs (shown on the right). (e) Alignment of the full DBDs of Vhr1 and Vhr2. Residues that fold into alpha-helices (according to PSIPRED [62]) are shown in bold. Black boxes show the two putative basic domains in VHR proteins. (f) Alignment of the second putative VHR basic region to basic regions of the eight bZIPs analyzed in this study.

Position 17 in the basic region of bZIP TFs is important for specifying Yap versus Gcn4 half-sites. (a) ClustalW2 alignment of the basic regions of eight yeast bZIP proteins, and the corresponding DNA binding site motifs. (b) Sko1 (which contains a phenylalanine at position 17 in the basic region, similar to Yap proteins) has a stronger preference for adjacent Yap half-sites compared to Gcn4 and Cst6 (which contain serine and cysteine at position 17, respectively). (c) The trend observed in (b) is not simply due to the fact that Sko1 prefers adjacent half-sites to overlapping half-sites. If this were the case, we would expect Gcn4 to bind overlapping Yap half-sites with higher affinity than Sko1, but we do not observe such a trend.

DNA binding specificity of Hac1. (a) Alignment of the Hac1 basic region against the basic regions of the eight bZIP proteins shown in Figure 3. Hac1 is most similar to Gcn4 in terms of DNA-contacting residues. The font color scheme is the same as in Figure 3; residues indicated in gray are not part of the basic region (of either bZIP or bHLH proteins). (b) PBM-derived motifs for Gcn4, Hac1, and Cbf1. Despite the fact that Hac1 belongs to the bZIP structural class, the Hac1 motif barely matches the motif of the bZIP protein Gcn4. Instead, the Hac1 motif resembles E-box motifs bound by bHLH proteins (such as Cbf1). (c, d) In-depth comparison of the DNA binding specificities of Hac1 and Cbf1. The scatter plots show 8-mer E-scores.

Primary and secondary DNA binding site motifs. (a) Different categories of TFs with two distinct modes of binding DNA. (b) Scatter plot of GOMER scores for regions bound in vivo by Sko1, according to the primary versus the secondary Sko1 DNA binding site motifs. Regions that score highly according to the primary but not the secondary motifs are shown in red. Regions that score highly according to the secondary but not the primary motifs are shown in blue. (c) Gene Ontology (GO) categories enriched in the regions that score highly according to the primary but not the secondary Sko1 motif. (d) GO categories enriched in the regions that score highly according to the secondary but not the primary Sko1 motif. See main text and Additional file 1 for details.