Chemical genetics of rapamycin-insensitive TORC2 in S. cerevisiae

Abstract

Current approaches for identifying synergistic targets use cell culture models to see if the combined effect of clinically available drugs is better than predicted by their individual efficacy. New techniques are needed to systematically and rationally identify targets and pathways that may be synergistic targets. Here, we created a tool to screen and identify molecular targets that may synergize with new inhibitors of target of rapamycin (TOR), a conserved protein that is a major integrator of cell proliferation signals in the nutrient-signaling pathway. Although clinical results from TOR complex 1 (TORC1)-specific inhibition using rapamycin analogs have been disappointing, trials using inhibitors that also target TORC2 have been promising. To understand this increased therapeutic efficacy and to discover secondary targets for combination therapy, we engineered Tor2 in S. cerevisiae to accept an orthogonal inhibitor. We used this tool to create a chemical epistasis miniarray profile (ChE-MAP) by measuring interactions between the chemically inhibited Tor2 kinase and a diverse library of deletion mutants. The ChE-MAP identified known TOR components and distinguished between TORC1- and TORC2-dependent functions. The results showed a TORC2-specific interaction with the pentose phosphate pathway, a previously unappreciated TORC2 function that suggests a role for the complex in balancing the high energy demand required for ribosome biogenesis.

Figure 1. Modeling and Characterization of the as-TOR2 allele

(A) Homology model of mTOR based on the structure of PI3Kγ shown with the gatekeeper residue in gray. The known S. cerevisiae TOR1/TOR2 inhibitor QL-IX-55 (purple) and BEZ235 (magenta) are oriented based on a typical h-bonding interaction with the backbone carbonyl of valine in the active site at VanDer Waals distances away from other residues that form the ATP-binding pocket. The isoleucine gatekeeper clash with BEZ235 is exacerbated by mutation to leucine and alleviated by mutation to alanine. The smaller QL-IX-55 does not sense this residue. (B) Sequence alignment shows the gatekeeper residue (in purple) is isoleucine in mTOR and leucine in all other cases. The active site is highly conserved. (C) EC50 of as-TOR2 and wt-TOR2 growing in culture. as-TOR2 is significantly more sensitive to BEZ235 than wt-TOR2. (D) as-TOR2 has an identical growth rate to wt-TOR2 when grown on YPD. At higher doses (1μM BEZ235), growth of as-TOR2 is inhibited while wt-TOR2 is unaffected. Growth of wt-TOR2 begins to be affected at 2μM BEZ235. (E) In vivo phosphorylation of Ypk1 by TORC2 in wt-TOR2 and as-TOR2 containing cells. as-TOR2 is significantly more sensitive to BEZ235 than wt-TOR2. (F) IC50 values show BEZ235 does not inhibit TORC1, that as-TOR2 does not play a significant role in the catalytic function of TORC1, and that the compound selectively inhibits as-TOR2 in TORC2 over wt-TOR2. The in vitro values correspond well to in vivo results, which are typically less sensitive due to high concentrations of ATP and poor cell wall permeability of yeast.

Figure 2. Chemical Epistasis Mapping of TORC1 and TORC2

(A) TOR1 exists only as a member of TORC1. TOR2 may exist as a member of either TORC1 or TORC2. Rapamycin selectively inhibits TORC1. BEZ235 is selective for the as-TOR2 allele. Chemical-genetic interactions behave as traditional double deletion mutants. For interacting genes, a directional shift between the DMSO control and the [high] drug screen should occur. Dose-dependent positive interactions occur between genes in linear pathways, dose-dependent negative interactions occur between genes in parallel pathways. (B) TOR mutant strains were mated to a library of ~1000 non-essential single deletion mutants. The resulting double mutants were grown on plates containing DMSO or increasing concentrations of rapamycin or BEZ235. (C) ChE-MAP for rapamycin treated and BEZ235 treated datasets sorted according to ΔS-score. The strength of positive and negative chemical-genetic interactions (S-scores) are reported by yellow or blue squares respectively. Inset are top hits from each set. The as-TOR2 dataset is significantly smaller since many strains were very sick at the highest concentration of BEZ235 and were removed during quality filtering. (D) Experimental datasets (wt-TOR2+rapa, as-TOR2+BEZ) and control datasets (rr-TOR1+rapa, rr-TOR2+rapa, wt-TOR2+BEZ) are shown by percent of total interactions in the dataset above ΔS ≥ |2.6|. Positive interactions are in yellow, negative interactions are blue. Rapamycin and BEZ235 are selective for their intended targets and generate few off target interactions. (E) Scatterplot of ΔS-score vs S-score illustrates the specific effect of BEZ235 on as-TOR2. wt-TOR2 (red) is unaffected by the compound and cluster around 0. as-TOR2 (blue) is strongly affected and shows a direct relationship between ΔS-score and S-score at 1μM BEZ235. (F) Number of dose-dependent genetic interactions above a ΔS ≥ |2.6| in each set. 104 interactions recorded for rapamycin, 134 recorded for BEZ with overlap of 10. (G) Network illustrating genes that hit both TORC1 and as-TOR2 above specified threshold. Nodes and edges are colored yellow for positive or blue for negative interactions with the indicated complex.

Figure 3. Enrichment in Biological Processes

(A) The sphingolipid biosynthesis pathway shows consistent dose-dependent behavior across all members of the pathway that were included in the screen in good agreement with theoretical prediction. S-score is indicated on a color metric scale with blue as strongly negative and yellow as a strong positive interaction. (B) Genotyped and sequenced members of pulled tetrads grown on plates containing increasing concentrations of rapamycin or BEZ235 confirm phenotypes tested using the ChE-MAP. (C) Bar graph shows fraction of each functional biological category that was included in the E-MAP. (B) Enrichment in either of the two datasets above or below ΔS = 2.0 were calculated using a Fisher’s test to identify terms in the cellular process GO Slim that were significantly enriched. Significant p-values highlighted in yellow.

Figure 4. Effect of rapamycin and BEZ235 on metabolites in the PPP

(A) Network of ChE-MAP hits that have physical interactions with genes within the PPP gene ontology term. Rounded rectangles and blue edges indicate chemical-genetic interactions and are colored according to the ΔS-score for the indicated gene. Enrichment for PPP linked genes that positively interact with TORC2 is 2-fold higher than expected by random chance with p = 0.02. Black nodes indicate genes not found in our dataset (www.thebiogrid.org) or in the following citations: (Fan et al., 2008; Fasolo et al., 2011; Graille et al., 2005; Hesselberth et al., 2006; Krogan et al., 2006; Ptacek et al., 2005; Yu et al., 2008). Black edges indicate a physical interaction. (B) 6-phospho-D-gluconate levels quantified by LC/MS over a 60 minute time-course where cells are perturbed by nitrogen starvation, inhibited with rapamycin (wt-TOR2), or inhibited with BEZ235 or rapamycin (as-TOR2). (C) Ribose-5-phosphate levels quantified by LC/MS over a 60 minute time-course where cells are perturbed by nitrogen starvation, inhibited with rapamycin (wt-TOR2), or inhibited with BEZ235 or rapamycin (as-TOR2). (D) Isotope labeled 6-phospho-D-gluconate allows direct measurement of newly synthesized metabolite in as-TOR2 and wt-TOR2 cells and allows quantification of oxidative pentose phosphate pathway flux. Treatment with BEZ235 shows a significant change (**) after the short 30 minute time point.