The capacity for expression divergence (or flexibility) has been linked to several characteristics of gene promoters. The simplest association is with the number of binding sites for transcriptional regulators: promoters of flexible genes are characterized by a relatively large number of binding sites [36
]. This is perhaps not surprising, since the expression of genes with many regulators (and binding sites) can be affected by mutations in any one of these regulators (or promoter binding sites), thus increasing their mutational target size - that is, the number of possible mutations that would affect the expression of these genes.
One particular promoter binding site stands out for its large influence on expression divergence: promoters that contain a TATA box show a remarkable increase in expression divergence, as well as in responsiveness and in noise [7
]. The distinction between genes with promoters containing a TATA box and those without stands when the number of transcriptional regulators or of promoter binding sites is controlled; it is also consistent among genes from different functional classes - for example, those encoding membrane proteins, genes encoding metabolic proteins, and genes encoding ribosomal proteins (although these different groups also differ widely in the proportion of genes with promoters containing TATA boxes) [7
]. Strikingly, increased expression divergence of TATA-containing genes has been observed in species ranging from yeast to mammals, including also mutation-accumulation lines of yeasts, flies and worms [7
], suggesting that it reflects a general phenomenon. Interestingly, the promoters of TATA-containing genes are not associated with more mutations but only with increased expression divergence [7
]. Thus, we propose that promoters carrying a TATA box are inherently more sensitive to genetic perturbations than TATA-less promoters. This is also consistent with the distinction between constitutive and inducible genes and with previous studies that demonstrated that a canonical TATA box is important for dynamic regulation of gene expression whereas other sequence elements are important for maintaining constitutive expression levels [35
The TATA box is a ubiquitous core promoter element that is bound by the transcription pre-initiation complex (PIC). What could cause increased expression divergence of TATA promoters? Transcription can be considered as a two-step process: first the PIC is recruited by transcription factors and assembles at the core promoter together with RNA polymerase; and second, the polymerase is released from the PIC and transcribes the gene. The second step can be repeated multiple times (re-initiation) if the PIC remains bound to the core promoter, and this is believed to be facilitated by the TATA box [42
]. Thus, a TATA box could increase the extent of re-initiation, thereby amplifying gene expression. Notably, the binding of the PIC to the TATA box and the binding of transcription factors to other sites could be cooperative [44
]. This would make the effect of the TATA box on gene expression nonlinear, as any amplification of transcription factor binding would stabilize PIC binding and cause a further increase in re-initiation. In this way, TATA-containing genes could be more sensitive to regulatory mutations than TATA-less genes.
Importantly, TATA-containing promoters differ from other promoters not only in their expression flexibility but also in other properties [45
], and so it is possible that these secondary characteristics underlie their increased expression flexibility. Perhaps the most notable feature of TATA promoters is their atypical chromatin structure [46
]. At most yeast promoters, the region directly upstream of the transcription start site contains transcription factor binding sites and is nucleosome-free, increasing the accessibility of the binding sites to transcriptional regulators [49
] (Figure ). By contrast, at promoters with high expression flexibility, and at those containing a TATA-box, this region tends to be more occupied by nucleosomes (Figure ). We and others have proposed that because nucleo somes are thought to interfere with the binding of regulatory proteins, the regulation of nucleosome states might fine tune the expression of these genes [46
]. Such increased dependence on the regulation of chromatin structure is indeed observed: promoters that are relatively more occupied by nucleosomes show relatively large changes in expression when genes encoding chromatin regulators are mutated or deleted [48
]. As with the effect of the number of transcription factors, an increased dependence on chromatin regulators increases the mutational target size, affecting expression of these genes. Any mutation in a gene encoding a relevant chromatin regulator, or an upstream gene regulating the activity of the chromatin regulator, could affect transcription of the downstream target gene.
Figure 1 Promoter architecture associated with expression flexibility [46-48]. Top: the architecture of a typical promoter in which nucleosomes are regularly positioned but are excluded from a particular region upstream of the transcription start site. This nucleosome-free (more ...)