Previous analysis of the Hog1 dependent stress response led to a coarse-grained model of Hog1 function where the kinase regulates gene expression through three entirely independent paths: activation of Msn2/4; activation of Hot1; and de-repression of Sko1, with Sko1 and Hot1 acting at only 12 genes15,28
. Since the transcription factors Msn2/4 are activated in diverse stress conditions and regulate >100 genes, this model led to the view that the osmotic stress response is largely nonspecific29
. This network structure, and previous data comparing the gene expression program in salt and sorbitol, also suggested that the Hog1 dependent transcriptional response is the same in all osmolytes10
Using the mutant cycle approach, we have converted the previously incomplete and qualitative description of Hog1 dependent gene activation into a quantitative and nearly complete network model ( and values in Table S3
). Our model shows that the signal from Hog1 is spread out to multiple transcription factors and then recombined in different ways at different promoters (). This network architecture allows stress signals transmitted through Hog1 to not only influence the general stress program via
Msn2/4 but to supplement and tune it as well ( and ). The osmotic stress response is therefore highly specific as Hog1 acts at least partially independently of Msn2/4 at many genes (112 in total; ). It is likely that these genes – which are involved in a wide-range of processes including glycerol synthesis, free radical breakdown, ion transport, general metabolism and signaling (Table S2
) – play a central role in adapting to osmotic stress. In addition, we find that in KCl stress, signals are transmitted through both the Hog1 and general stress (Msn2/4) pathways and then integrated at the signaling and promoter level (). By comparing the transcriptional response in glucose to that in KCl we show that this network architecture allows budding yeast to respond to different osmolytes in different ways (as described in detail below); that is, the transcriptional program activated by Hog1 is context dependent.
What is the functional significance of the Hog1 network structure and the signal integration we have uncovered? A recent study of Hog1 signaling dynamics demonstrates that the Hog1 dependent transcriptional response in high salt stress functions to prepare cells for future changes in osmolarity while the immediate response to osmotic stress depends on more rapid post-translational mechanisms30
. We find that this transcriptional response includes the 200-gene general stress response (through Msn2/4) as well as 70 additional genes activated by Hog1 alone (through Sko1/Hot1 and at least one unknown factor; ). This broad program likely prepares the cell for both the damage caused by salt (due to disruption of protein-protein and protein-DNA interactions31
) and the osmotic imbalance induced in these harsh conditions. By contrast, when the osmotic stress is created by glucose, cells activate the 70 genes controlled by Hog1 alone, but do not expend the energy needed to activate the full 200 gene general (Msn2/4 dependent) stress program. This makes sense, as cell damage is likely to be limited under such conditions and Msn2/4 activation leads to energy conservation and slow growth32
, a process that is likely to be disadvantageous in a high glucose environment such as fruit. Instead, only a subset of the Msn2/4 dependent genes are activated in high glucose – those where Sko1/Hot1 and Msn2/4 cooperate to induce expression (). Interestingly, these genes are regulated in two distinct ways by the Hog1 network. At genes where Sko1/Hot1 and Msn2/4 cooperate with SUM gate logic, the expression levels are boosted above that created by the general stress response (Msn2/4) whenever Hog1 is activated. This form of regulation is found at several genes involved in converting glucose into the osmolyte glycerol (HXT1
), suggesting that additional capacity (beyond that created by Msn2/4 alone) through this pathway is beneficial in all osmotic stress conditions. By contrast, Sko1/Hot1 activity only alters expression at genes with OR gate logic when Msn2/4 activity is low (e.g. in high glucose). The genes regulated in this manner play more generic roles in stress recovery such as neutralizing free radicals and cell wall/cell membrane repair (e.g. CTT1
) and appear to be required at some minimum level after osmotic stress.
Overall, our model of the Hog1 network provides insight into the way a signal can create a context dependent gene expression program using a limited number of transcription factors. Because Hog1 acts through the general stress regulators Msn2/4, the response to osmotic stress depends on the combined action of multiple pathways (those regulating Msn2/4) and thus the overall state of the cell. However, by acting in parallel through the osmotic stress specific transcription factors Sko1 and Hot1, this generic stress response is adapted so that it is specific to, and presumably appropriate for, osmotic stress in at least two different stress conditions. We therefore anticipate that other stress signals will be transmitted through networks with a similar overlapping structure.
Beyond establishing the structure and function of the Hog1 transcriptional network, our results demonstrate the utility of double mutant analysis, and the overall strategy taken here, for dissecting gene regulatory systems. We have shown that, starting with two or more known/putative network components, it is possible to build a quantitative genome-wide network model and to identify the genes regulated by missing components. By performing a screen for the factors that act on these genes (using bioinformatics, microarrays, or reporter strains), it is possible to identify the missing components and integrate them into the network model. This approach has immediate application to studying conditionally activated pathways (and drug-pathway interactions) using gene KOs, and can be extended to other systems through the use of RNAi and chemical inhibitors.