|Home | About | Journals | Submit | Contact Us | Français|
We identified 38 genes as having a genetic interaction with a mutant form of the kinase specific chaperone, Cdc37, using a genome-wide synthetic screening approach. The identified genes included a sub-network of highly interacting genes enriched for functions in genome integrity and comprising multiple components of several discrete molecular machines. A network analysis approach related these machines to a small group of cell cycle checkpoint kinases.
Cdc37 is a component of the Hsp90 chaperone machine that helps to fold a wide variety of cellular proteins, including transcription factors and protein kinases. The folding function of these molecular chaperones is integrated into a larger process of quality control, where misfolded proteins are degraded via the ubiqutin/proteasome system.1 The intersection of folding and degradation pathways has been exploited for clinical use with Hsp90 inhibitors, such as geldanamycin, that promote rapid degradation of client proteins resulting in tumor cell apoptosis.2 However, it remains unclear which cellular processes are most affected by chaperone loss of function.
In S. cerevisae, kinome biogenesis depends on Cdc37, which protects the nascent chains from degradation during or shortly after synthesis. Although an essential gene, Cdc37 chaperone function can be compromised by mutation in one of two phosphorylation sites, Ser14 and Ser17.3 In the Ser14Ala mutation (cdc37S14A), there are reduced amounts of over 75% of kinases and reduced activity of other kinases.4 We determined how reduced kinase levels resulted in synthetic interactions with other genes, as a way of phenocopying the effects of geldanamycin, but restricting its effects to kinases. In this way, we avoided the problem of illuminating every process chaperoned by the Hsp90 machinery.5 The ultimate goal is to identify novel cellular processes that can be targeted in combination with geldanamycin treatment to increase its therapeutic effectiveness.
We isolated double mutants between cdc37S14A and each of the ~4700 knockout mutants in S. cerevisiae using synthetic gene arrays (SGA) (described previously Ref. 6). The strains were arrayed in 1536 density in duplicate and the experiment was performed twice (materials and methods in supplementary data) using a Singer robot to automate all pinning procedures. Double mutants that grew significantly more slowly (having synthetic fitness or synthetic lethality) than either single mutant (in 3 of 4 tests) were analyzed by random spore analysis. Thirty-eight genes made the final cutoff, which is close to the average number of synthetic interactions for yeast genes.6 The data are shown in Figure 1 which also shows how each gene interacts with others in the group based on data publicly available in The BioGrid.7
Genes having synthetic interactions with cdc37S14A were enriched for functions involved in cell cycle as determined by their gene ontology (16 out of 38 genes; 42.1% versus 6% for the genome as a whole, p = 5.04 × 10−8). Only two different protein kinases were present in the SGA hits. One of these is casein kinase II (CK2; three of four different subunits identified), which phosphorylates Cdc37 itself and therefore contributes to the chaperoning process,3 and the other was Cla4. Other chaperones that have synthetic interactions with Cdc37S14A when deleted include GIM4, HSC82, YDJ1 and STI1. These hits, with the exception of GIM4, validated the screen since they are known to promote kinase folding and have been characterized previously as having synthetic genetic interactions with CDC37.8-12
The most intriguing group of synthetic fitness mutants form a sub-group of highly interacting genes as determined from the BioGrid database (Fig 1). Each gene in this group forms a network with at least three others in the group, and is highly enriched for functions in genome integrity.13 Multiple components of Topoisomerase III and replication factor C were identified. Four subunits of the Hir chromatin assembly complex and the Rpd3S deacetylase complex were also present. Although these are not as highly interacting with other members of the genome integrity sub-group, they are functionally related. Notably, most of the synthetic interactions were not lethal but resulted in slow growth. The only synthetic lethal interactions were between cdc37S14A and ydj1Δ, hsc82Δ, sti1Δ and protein kinase cla4Δ.
Our findings suggest that Cdc37 buffers genome integrity. Our hypothesis was that protein kinases mediate this effect, but which ones? Our first approach was to overexpress individual kinases in an attempt to suppress the slow growth defects in the double mutants. We picked Cdc28, Rad53 and Ste20 as test kinases and overexpressed each of them in the double mutants of cdc37S14A and the 38 synthetic fitness/lethal mutants. None of these kinases, however, had any capability to suppress the slow growth phenotype of the double mutants (not shown); in fact, kinase overexpression further reduced growth of the double mutants (data not shown). One possibility to explain these results is that Cdc37 might chaperone proteins other than kinases, (as described previously in ref. 14). However, this would not diminish how Cdc37 affected kinome biogenesis, and the roles kinases play in regulating cell cycle and genome integrity. Indeed, it is possible that a decrease in multiple kinases might be responsible for generating the synthetic interactions. As an alternative, therefore, we performed a network analysis to identify kinases known to interact either physically or genetically with each of the 38 hits in our screen. This analysis was based on (previously published studies stored in the BioGrid, ref. 7, and MPact, ref. 15) interaction databases (see Methods in online Supplement). Kinases identified in this analysis were ranked according to how many interactions they had with different genes having synthetic interactions with the cdc37S14A mutant. The highest-ranked kinase was Rad53, which interacted with 17 of the SGA hits. Rad53 was followed by Bub1 (14 interactions), Cdc7 (11 interactions), Ssn8 (11 interactions), Dun1 (10 interactions), Clb2 (10 interactions) Clb5 (9 interactions), and Ssn3 (8 interactions). Note that three of these genes are cyclin components of protein kinases rather than encoding the kinase catalytic domain. These include Clb2 and Clb5 cyclins that interact directly with Cdc28 to regulate the cell cycle. Ssn8 is a cyclin component of Ssn3 which interacts with RNA polymerase II. The top 8 kinases and cyclins identified by this analysis were tabulated against the genes identified in the SGA analysis to determine the frequency with which they interacted with each other (Fig. 2). The results show that just three kinases: Rad53, Bub1 and Cdc7, interact with over 60% of the SGA hits. All of the genes in the highly interacting sub-network, with the exception of Lte1, interact with Rad53 and/or Bub1.
These combined results suggest that multiple kinases cooperate to ensure the fidelity by which genome integrity is monitored and regulated. When the level of these kinases is collectively lowered, as in the cdc37S14A mutant, the robustness of the genome integrity system is compromised. Interestingly, this hypothesis may explain the synergistic effects already observed between Hsp90 inhibitors and drugs that affect genome integrity.16
We are very grateful to Charlie Boone for providing the arrays of viable yeast gene-deletion strains and the SGA parent strain Y7092 for these studies. We also thank Huiming Ding for help with computed-based image analysis. This work was supported by National Institutes of Health grants GM70596 to A.J.C. and GM42728 to I.M.W. along with funds for the Singer RoToRHDA robot from the Department of Biochemistry, Albert Einstein College of Medicine and Dr. Steven Almo. A.M. thanks Ravi Iyengar for support from GM54508.
Supplemental material can be found at: www.landesbioscience.com/supplement/CaplanCC6-24-Sup.pdf