Here we have found that the Whi5 pathway is not the sole link between Cln3-Cdc28 and SBF activity. We have found several mutants that, like whi5, relieve the repression of SBF, and render its activity somewhat independent of Cln3-Cdc28. These mutants include chd1, hda2, pho23, sin3, rpd3, and stb1. Of these, pho23, sin3, stb1, and rpd3, are members of, or have been physically linked to, the Rpd3 histone deacetylase complex, a repressive histone deacetylase orthologous to mammalian HDAC1.
Although we do not know the exact relationship between these proteins and Whi5, we have found that the stb1 mutation is synergistic with whi5; that is, in the context of a bck2 mutation, the stb1 whi5 double mutant, unlike either single mutant, has little ability to respond to Cln3-Cdc28. Thus in some sense Stb1 identifies a pathway for regulating SBF that is separate from the Whi5 pathway.
While we have identified STB1
in a screen for repressors of SBF, others have previously identified STB1
as an activator of SBF or MBF 
. While paradoxical at first sight, it is quite common for transcription factors to have both positive and negative roles in transcription. An example is Fkh2, which collaborates with Mcm1 and Ndd1 transcription factors, and with Clb-Cdc28 kinase activity, to regulate mitotic genes. In this context, Fkh2 appears to be an activator in late G2 and mitosis, but a repressor at other times 
. Similarly, we imagine that Stb1 helps repress SBF in the absence of Cln3 and Bck2 (the situation in which we found it as a repressor), but helps activate SBF in the presence of Cln3 or Bck2 (the situation in which it was characterized as an activator). Consistent with this, cell cycle expression analysis of stb1
mutants shows that target genes are less repressed at troughs, and less induced at peaks; i.e., they are less regulated and more constitutive (e.g., Figure 3 in 
). The fact that CLN3
can induce expression of CLN2
even before Sin3, Rpd3, and Whi5 are lost from the promoter () is consistent with the idea that the initial expression of CLN2
depends on activation, perhaps via Stb1, rather than on loss of repression.
If Stb1 is both a repressor and an activator, then some of our assays may preferentially see one of these activities, and some may see the other. Presumably it is the lack of repression by Stb1 that allows the stb1 mutation to suppress the lethality of the cln3 bck2 mutant. But the cell size assay for responsiveness to CLN3 () may be more sensitive to Stb1 as an activator; in particular, the synergistic defect between whi5 and stb1 may be due to the lack of repression in the whi5 mutant, plus the lack of activation in the stb1 mutant. We note that the combinations of mutations that include stb1 tend to have relatively large cell sizes after induction of GAL-CLN3 (), perhaps showing that STB1 is needed for full induction of CLN2.
While Whi5 and Stb1 seem to define two pathways of regulation of SBF, it is still unclear how the Sin3-Rpd3 histone deacteylase complex is recruited to the CLN2
promoter. Previously, the Rpd3 complex has been linked to Stb1 
. More recently, Huang, Kaluarchchi and Andrews have found evidence for an association between Rpd3 and Whi5 (personal communication). Despite these associations, we found that even whi5 stb1
double mutants had at least some Sin3 (and so presumably Rpd3) at the CLN2
promoter, whereas swi6
mutants had little or no Sin3. Thus although one could imagine various relationships between these proteins, one model is that SBF has some ability to recruit each of Whi5, Stb1, and Sin3-Rpd3, but that these proteins in addition interact with each other (). Later, in a size- and growth-dependent fashion, Cln3-Cdc28 also joins the complex, and phosphorylates Whi5 and Stb1 and probably Swi6 and possibly Swi4. This causes the loss of the Rpd3 complex; a somewhat slower loss of Whi5 (); and perhaps allows phosphorylated Stb1 to help activate transcription (). That Swi6, along with Whi5 and Stb1, is probably a target of Cln3-Cdc28 phosphorylation is strongly suggested by the fact that over-expression of a mutant Whi5 lacking CDK phosphorylation sites is lethal in a mutant where Swi6 is likewise lacking CDK phosphorylation sites 
. The involvement of Swi6 as a likely target of Cln3-Cdc28, and as a recruiter of Sin3-Rpd3, may explain why even whi5 stb1
double mutants seem to have some slight residual Cln3-responsiveness (); that is, this residual responsiveness could be through direct phosphorylation of Swi6.
Model for regulation by Whi5, Stb1, and histone deacetylase.
Results reminiscent of ours with regard to Sin3 and Rpd3 were previously obtained by Veis et al. 
, who found that Sin3 and Rpd3 associate with the promoter of the CLB2
gene, which encodes a mitotic cyclin. Although CLB2
is most highly expressed in G2/M, the association of Sin3 and Rpd3 with the CLB2
promoter was lost in late G1, at about the same time we see loss of Sin3 and Rpd3 from the CLN2
promoter. Veis et al. interpreted their results in terms of the association between Sin3/Rpd3 and the Fkh2 (forkhead) transcription factor, and suggested that this association was sensitive to Start. However, we note that CLB2
, despite being most strongly up-regulated in G2/M, is a client of SBF as well as a client of Fkh2. The CLB2
promoter contains at least three clustered SBF/MBF binding sites, at least two of which are conserved in other species of yeast 
. In ChIP experiments, CLB2
is a target of SBF or MBF binding 
. Thus the loss of Sin3/Rpd3 from the CLB2
promoter in late G1 as seen by Veis and coworkers could involve SBF at the CLB2
promoter, and so could be related to the phenomenon we see at the CLN2
Another protein we find at the CLN2
promoter is Cln3. However, demonstrating this association was difficult, and required a special genetic background and over-expression of Cln3. Part of the difficulty in ChIPing Cln3 to the CLN2
promoter is presumably because Cln3 is a nonabundant protein, and does not bind DNA directly. But in addition, Cln3 may not be a stoichiometric member of the complex. Instead, it may bind weakly and transiently, phosphorylate its substrate(s), and leave. The two proteins we find to be essential for Cln3 responsiveness, Whi5 and Stb1, are both very likely substrates of Cln-Cdc28 
Cln3 is present at only about 100 molecules of protein per cell, and yet there are in the vicinity of 400 functional binding sites for SBF and the related factor MBF. The fact that Cln3 is sub-stoichiometric with respect to binding sites could provide a partial solution to the size control problem: Perhaps the amount of Cln3 in the cell, which is a function of cell size and growth rate, is titrated against the number of binding sites. And indeed we found that cells containing extra SCBs had to grow to a larger size to accomplish Start, and this effect could be compensated by one extra dose of CLN3. Extra SCBs did not enlarge a cln3 null mutant, and extra SCBs had no effect whatever on cln3 stb1 whi5 triple mutants. These findings are all supportive of the titration model.
Even though larger G1 cells contain more Cln3 molecules than smaller cells, the increase in Cln3 content with size is probably quite moderate, possibly only linearly correlated with cell size. Thus even at cell sizes adequate for Start, Cln3 may still be sub-stoichiometric with respect to binding sites. Thus we imagine that at any and all physiologically reasonable concentrations of Cln3, there will only be fractional occupancy of SBF sites, especially if Cln3 is a weak and transient binder. But as the amounts of Cln3 rise, and are titrated against a fixed number of SBF sites, that fractional occupancy will rise, until at some occupancy (i.e., at some critical cell size), CLN2 and other targets are expressed, and the cell passes through Start. The issue is, how to convert a relatively small change in total Cln3 into a large change in fractional occupancy, or, alternatively, how to convert a small change in occupancy into a large effect?
Although we do not know the answers to either of these questions, Ferrell and coworkers have described many mechanisms by which such “super-sensitivity” can occur 
. One mechanism would use the fact that SBF target genes have multiple SBF binding sites. Perhaps the binding of Cln3 to SBF is cooperative; or perhaps the Cln3 molecules, once bound, cooperate to do something else, such as phosphorylate a substrate. Cooperativity of any kind between multiple sites will give exponential sensitivity to Cln3 amounts, so this is one possible mechanism. A second mechanism is multisite phosphorylation. That is, perhaps the substrates of Cln3-Cdc28 have to be phosphorylated at multiple sites, and this can only happen when fractional occupancy of SBF sites by Cln3 is relatively high. Since phosphorylation is in a dynamic equilibrium with dephosphorylation, a requirement for multisite phosphorylation (at, say, five sites) imposes a super-sensitive threshold on the amount of kinase required 
. Multisite phosphorylation can give extreme sensitivity to the amounts of a protein kinase 
. A third possible mechanism is to consider the relationship between the complexes at the multiple SBF sites. There are three sites at CLN2
; if all three have repressive proteins, is enough Cln3 needed to fill all three sites simultaneously, even though occupancy of any one site is always transient? At any rate, although we do not know how supersensitivity works in this situation, there are lots of ways it could work in theory, as cited above.
There are remarkable parallels between the SBF/Cln3/Whi5,Stb1/Rpd3 regulatory module in yeast, and the E2F-Dp/Cyclin D1/Rb/HDAC1 regulatory module in mammalian cells. To begin with, the cluster of regulated genes is highly conserved: In S. cerevisiae 
, in the distantly related yeast S. pombe 
, in mammalian cells 
, and probably in most or all other eukaryotes, there is a highly conserved cluster of genes needed for DNA replication, and expressed around the G1/S transition. In both yeasts and mammals, the motifs regulating these genes contain a core “CGCG” element. In both yeasts and mammals, the transcription factors recognizing this element (SBF/MBF in the yeasts, E2F-Dp in mammals) contain a DNA binding domain with a “winged helix” fold 
. There is no apparent sequence homology between the yeast and mammalian DNA binding domains, but the domain is small, the evolutionary distance vast, and there are other examples where structure but not sequence has been preserved across time.
In E2F-Dp, the transactivator is repressed by binding of Rb and its family members. There are two mechanisms of repression 
. First, the transactivation domain is masked. Second, Rb family members (but possibly not Rb itself-
) recruit mSin3B and HDAC1 which deacetylate and otherwise modify chromatin so as to be inhospitable towards expression. Here, we likewise show that there are at least two pathways of regulation, one of them involving recruitment of a histone deacetylase. In mammals, the transactivation domain is unmasked when a cyclin-CDK complex such as cyclin D-CDK4 phosphorylates Rb and family members, disrupting binding to E2F-Dp, and allowing Sin3m and HDAC1 to leave the chromatin. Similarly, in yeast, Cln3-CDK phosphorylates Whi5 and probably Stb1. Whi5, Sin3, and Rpd3 all leave the chromatin. Interestingly, expression of the target gene Cln2 precedes the loss of the repressive proteins, consistent with a dominant activation, possibly due to Stb1. In any case, it is clear that there are deep, well-conserved parallels between the SBF/Cln3/Whi5,Stb1/Rpd3 regulatory module in yeast, and the E2F-Dp/Cyclin D1/Rb/HDAC1 regulatory module in mammals. It is possible that these modules have regulated the cluster of genes for DNA synthesis since early in eukaryotic evolution.