To survive in fluctuating environments, organisms must respond to changes in their surroundings. One of the primary means of response at the cellular level is the adjustment of levels of gene transcription1
. In some cases, new transcriptional programs are established by the binding of regulatory proteins to new gene targets under a given environmental condition2
. These regulatory factors then act to either activate or repress transcription. Four straightforward mechanisms for the regulation of transcription-factor targeting have been demonstrated. These include altering the regulatory factor's concentration in the nucleus3
, altering binding affinity by post-translational modification4
or through an allosteric cofactor5
, and altering binding properties by expression of a protein cofactor6
. Here, we present experiments that provide evidence for a new mechanism of dynamic transcription factor target specification. In the proposed mechanism, the genome-wide distribution of DNA sequence motifs for which a factor has intermediate affinity is carefully coordinated with chromatin-mediated regulation of accessibility to those sites. In this way, rather than through use of a specific cofactor or post-translational modification, the genome itself is remodeled to change the targeting and biological outcome of an unaltered transcription factor.
rotein 1) directs a transcriptional program that is central to yeast metabolism, activating the transcription of genes encoding the ribosomal protein subunits and glycolytic enzymes7-9
. Nearly 40% of the mRNA initiation events in mitotically growing yeast are activated by Rap1, yet Rap1 is also required for the repression of these same genes in response to nutrient depletion10
. Because of its central role in the cellular economy, we asked whether Rap1 was targeted to new genomic loci upon a change in the nutrient environment. The genome-wide localization of Rap1 was monitored as yeast consumed and ultimately depleted their carbon sources (). We found that despite a decrease in Rap1 protein level (Figure S1
) and no evidence of posttranslational modification11,12
, Rap1 bound to an expanded target set. Fifty-two targets specific to low-glucose growth conditions were identified (hereafter “low-glucose targets”), while no targets specific to high-glucose growth were found. A representative subset of low-glucose targets was verified by ChIP-PCR (Figures S2 and S3
Rap1 binds to new targets in the absence of glucose
A transcription factor's target set could be expanded despite decreasing protein concentration if most potential binding sites were blocked during growth in glucose, but then in the absence of glucose, a subset of those sites were made available. In the case of Rap1, we reasoned that proteins involved directly in such a mechanism would have a genomic distribution in the presence of glucose that closely matched the Rap1 targets bound only in the absence of glucose. We identified eight such proteins using a computational approach (Methods and Figure S4
), four of which (Sut1, Mig1, Nrg1, and Sko1) had been previously characterized to interact physically with the Tup1-Ssn6 repressor complex13,14
. Identification of Tup1 itself through our screen was not possible because the genome-wide localization of Tup1 had not been reported. Tup1 is a well-characterized repressor that does not bind DNA directly15
, but is instead recruited to specific loci by other sequence-specific factors. Mig1, Nrg1, and Sko1 recruit Tup1 to the promoters of glucose-repressed genes, starch degrading genes, and osmotic stress inducible genes respectively16-18
. We hypothesized that Tup1 and its recruiters negatively regulate Rap1 binding to low-glucose targets.
This predicts that in cells lacking Tup1 or its recruiters, Rap1 would bind inappropriately to low-glucose targets, even in the presence of glucose. We determined the genomic distribution of Rap1 in strains grown in high-glucose media, but lacking the genes encoding Tup1 or its recruiters. Compared to wild-type cells, all deletion strains exhibited a specific and significant increase in Rap1 occupancy at low-glucose targets, with deletions of the TUP1 gene itself showing the strongest effect (). Accordingly, many low-glucose targets were considered “bound” by Rap1 in the deletion strains (p-value cutoff of 0.005, ). Recapitulation of a low-glucose Rap1 binding pattern through mutations in genes encoding Tup1 and its recruiters shows that these proteins are normally required to prevent the binding of Rap1 to low-glucose targets in a high-glucose environment.
In the absence of Tup1 or proteins that recruit Tup1, Rap1 binds specifically and inappropriately to low-nutrient targets
To further characterize the role of Tup1 in restricting the binding distribution of Rap1, we determined the genomic localization of Tup1 during exponential growth in the presence of glucose. We found that the majority of the low-glucose Rap1 sites that were inappropriately bound in tup1Δ strains were occupied by Tup1 in a wildtype strain. This provides evidence that inappropriate Rap1 binding in tup1Δ strains was caused directly by the absence of Tup1 at the affected loci (). If Tup1 blocked Rap1 binding directly, Tup1 would be predicted to vacate low-glucose Rap1 targets upon glucose depletion. To test this, we determined the genomic distribution of Tup1 after glucose depletion. Contrary to the prediction, low-glucose Rap1 targets bound by Tup1 in the presence of glucose remained bound after glucose was depleted. In fact, more low-glucose Rap1 targets were bound by Tup1 in low-glucose conditions (). Therefore, Tup1 itself does not block Rap1 binding directly.
Tup1 restricts Rap1 binding through chromatin-modifying co-factors
Tup1 is known to alter local chromatin structure by interacting with chromatin-modifying proteins19-21
, including Hda122
. Hda1 is a class 2 histone deacetylase required for repression of a subset of Tup1-repressed genes23
. To determine whether recruitment of Hda1 is required for Tup1's role in restricting Rap1 targets, we tested Rap1 localization in an hda1
Δ strain. Rap1 occupancy of low-glucose targets was significantly higher in the absence of hda1
, mimicking a low-glucose binding response (). Tup1 also interacts with Isw2, a member of the imitation-switch (ISWI) class of ATP-dependent chromatin remodeling complexes24-27
. Rap1 occupancy at low-glucose targets was also increased in isw2Δ
strains, but to a lesser degree than was observed for hda1
To corroborate these results, we used existing data23,28
to ask if the mRNA expression of low-glucose targets was affected in tup1
Δ, or hda1
Δ strains. Expression of low-glucose targets was increased in tup1
Δ and hda1
Δ, but not in isw
2Δ strains, whereas the expression of static targets was unaffected by these mutations (). These results are concordant with the observed changes in Rap1 binding in the respective mutants, and demonstrate a downstream consequence on gene regulation for mis-targeting of Rap1. These results also indicate that at many loci, Hda1 is required for Tup1's ability to block Rap1 binding and repress transcription in the presence of glucose.
The data predict that motifs for Rap1 and for recruiters of Tup1 occur in the promoters of low-glucose Rap1 targets. Sko1, Sut1, and Mig1 motifs are indeed specific (p < 0.05) for low-glucose targets. Furthermore, the Sko1, Sut1, Mig1, Nrg1 and Rap1 motifs all provide information that distinguishes low-glucose targets from all other genomic loci (). More intriguing were the differences in the types of Rap1 motif found in static versus low-glucose Rap1 targets. For the static targets, a strongly stereotypic consensus motif that matched the previously characterized Rap1 binding site9
was discovered (“Rap1 strong”, ). In contrast, 90% of the low-glucose targets contained a degenerate Rap1 motif of lower predicted affinity (“Rap1 weak”, ). We confirmed the lower affinity of Rap1 for conditional promoters by analysis of existing protein binding microarray data29
(). Therefore, low-glucose targets are distinguished from constitutive targets by harboring binding sites for recruiters of Tup1 and intermediate-affinity Rap1 binding sites.
Low-glucose Rap1 targets contain Sko1, Sut1, or Mig1 binding sites and a weak Rap1 consensus motif
The results above demonstrate the role of the Tup1 in restricting Rap1 binding and show that the histone deacetylase Hda1 is required to maintain a high-fidelity Rap1 binding pattern. We hypothesized that Tup1 and Hda1 could block Rap1 binding at low-glucose targets by stabilizing one or more nucleosomes in the presence of glucose. We therefore determined genome-wide nucleosome occupancy in the presence of glucose and after glucose had been depleted from the media. In glucose, nucleosome occupancy is higher at low-glucose targets than it is at loci constitutively bound by Rap1 (). However, upon glucose depletion, the promoters of low-glucose targets exhibit a sharp nucleosome loss, which results in lower nucleosome occupancy than even the static targets (). We next explored the relationship between Rap1 binding and nucleosome occupancy. If nucleosome occupancy facilitates Rap1 binding to intermediate affinity sites, we would expect Rap1 binding to increase as a function of nucleosome loss at those sites. We found that Rap1 occupancy increases as nucleosomes are lost (). More telling, the loss is predicted to occur specifically at a subset of intermediate-affinity sites, as opposed to other high-affinity or low-affinity sites throughout the genome. Even though nucleosome loss does occur at other sites throughout the genome, the intermediate-affinity sites are most sensitive to these changes in terms of Rap1 binding (). These results strongly suggest a causal role for nucleosome loss in facilitating Rap1 binding to intermediate-affinity loci in low-glucose conditions.
Loss of nucleosomes allows Rap1 to bind to intermediate-affinity targets
We propose a mechanism in which transcription factor binding is blocked at condition-specific targets by stabilizing nucleosomes at intermediate-affinity transcription factor binding sites (). In our experimental system, we have identified several key players and steps in this mechanism. Sequence-specific DNA binding proteins recruit Tup1 to specific loci, which in turn serves to recruit the activities of the histone deacetylase Hda1 and the ATP-dependent chromatin remodeler Isw2 to low-glucose targets. Our model proposes that in the presence of glucose, the activities of Hda1 and perhaps Isw2 stabilize a nucleosome to block an intermediate-affinity Rap1 binding site. In the absence of glucose, repression is relieved as a consequence of nucleosome release at the conditional promoters. Rap1 binds and remains bound to static sites through a combination of inherently low nucleosome occupancy and high Rap1 in vitro binding affinity, while selected intermediate-affinity sites are situated on a nucleosomal hair-trigger that allows conditional binding through elevated sensitivity to changes in nucleosome occupancy. This mechanism and the experiments performed here account for about half of the new targets bound by Rap1 upon glucose depletion, so it is likely that other mechanisms that influence environment-dependent target selection in this system await discovery.
A “nucleosomal hair-trigger” model for condition-dependent transcription factor binding through coordinated interplay between local chromatin structure and DNA-binding affinity