Suppression of 19S proteasome subunits marks emergence of an altered cell state in diverse cancers Peter Tsvetkova,1, Ethan Sokola,b, Dexter Jina,b, Zarina Brunea, Prathapan Thirua, Mahmoud Ghandic, Levi A. Garrawayc,d,e, Piyush B. Guptaa,b,f,g, Sandro Santagatah, Luke Whitesella, and Susan Lindquista,b,i,2 aWhitehead Institute for Biomedical Research, Cambridge, MA 02142; bDepartment of Biology, Massachusetts Institute of Technology, Cambridge, MA 02139; cBroad Institute, Cambridge, MA 02142; dDana-Farber Cancer Institute, Harvard Medical School, Boston, MA 02215; eDepartment of Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA 02115; fKoch Institute for Integrative Cancer Research, Cambridge, MA 02139; gHarvard Stem Cell Institute, Cambridge, MA 02138; hDepartment of Pathology, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA 02115; and iHoward Hughes Medical Institute, Cambridge, MA 02139 Edited by Carol Prives, Columbia University, New York, NY, and approved December 5, 2016 (received for review November 22, 2016) The use of proteasome inhibitors to target cancer’s dependence on proteasome inhibition. In this work we examine if such a mecha- altered protein homeostasis has been greatly limited by intrinsic nism is implemented by different cancers. and acquired resistance. Analyzing data from thousands of cancer lines and tumors, we find that those with suppressed expression Results of one or more 19S proteasome subunits show intrinsic protea- Sharply Reduced Expression of One 19S Proteasome Complex Subunit some inhibitor resistance. Moreover, such proteasome subunit sup- Occurs Frequently Across Diverse Cancer Cell Lines. We first char- pression is associated with poor outcome in myeloma patients, acterized relative mRNA expression levels for all genes that where proteasome inhibitors are a mainstay of treatment. Beyond encode proteasome subunits using the Genomics of Drug Sen- conferring resistance to proteasome inhibitors, proteasome subunit sitivity in Cancer (GDSC) database (20). As expected, across 789 suppression also serves as a sentinel of a more global remodeling of cancer cell lines, the average expression level of 20S subunits was highly correlated with the average expression level of 19S sub- the transcriptome. This remodeling produces a distinct gene signa- A ture and new vulnerabilities to the proapoptotic drug, ABT-263. This units (Fig. 1 ). However, in some cell lines, the expression of a frequent, naturally arising imbalance in 19S regulatory complex specific 19S subunit was much lower than the average expression of that subunit across all cell lines. To quantify this phenomenon composition is achieved through a variety of mechanisms, including MEDICAL SCIENCES for each cell line in the GDSC and the Cancer Cell Line Ency- DNA methylation, and marks the emergence of a heritably altered clopedia (CCLE) datasets, we calculated the SD from the mean and therapeutically relevant state in diverse cancers. of every 19S subunit in every cell line. Each cell line was then assigned a sigma score that is indicative of the greatest deviation drug resistance | epigenetic gene regulation | apoptosis | EMT | from the mean of any one 19S subunit in that cell line. Thus, a bortezomib sigma score of 3 indicates a cell line in which the mRNA ex- pression of at least one subunit of the 19S proteasome complex ells rely on the proteasome machinery to mediate protein was reduced by 3 SDs from the mean mRNA level of that par- Cturnover and maintain protein homeostasis (1–3). Upon ticular subunit across all cell lines. A sigma score of 2 indicates a oncogenic transformation, a myriad of proteotoxic stresses that reduction in the mRNA of 2 SDs, and so on. In 59 cancer cell lines, the mRNA level of at least one 19S tax the cellular machinery responsible for protein homeostasis σ are introduced (2, 4). These pressures cause cancer cells to rely subunit was 3 or more SDs (3 ) from the mean for that subunit heavily on enhanced proteasome function (2, 5–8). This de- pendency can be exploited using natural and synthetic com- Significance pounds that inhibit, with exquisite potency and selectivity, the catalytic function of the 20S proteasome (9, 10). Indeed, these In previous work, we used genome-wide screening to uncover a compounds are highly effective at inhibiting the growth of a wide counterintuitive mechanism by which cells can acquire resistance variety of cancer cell lines in culture. Unfortunately, their clinical to inhibitors of the proteasome’s catalytic core through experi- utility has been surprisingly limited, with therapeutic benefits mentally induced imbalances in the composition of its regulatory commonly observed for only a few types of cancer (9, 10). particle. However, in many cases, mechanisms uncovered in vitro The limited role for proteasome inhibitors as clinical chemo- for acquired resistance often do not translate to the context therapeutics can be attributed to biological processes that pro- of actual clinical cancers. Here, we show that this mechanism is mote intrinsic and acquired resistance. For example, cell culture actually deployed spontaneously and naturally in diverse human models of acquired resistance often accumulate mutations in the cancer lines and is associated not only with increased resistance catalytic subunits of the 20S proteasome (11, 12). In terms of to proteasome inhibitors both in vitro and in the clinic but clinical relevance, however, such mutations have yet to be de- also is symptomatic of a much more broadly altered state with tected in clinical samples (13). Additional mechanisms of re- a unique gene signature and drug targetable vulnerabilities. sistance include alteration of specific cellular pathways such as κ constitutive activation of NF- B (14), activation of the chaper- Author contributions: P. Tsvetkov designed research; P.B.G., L.W., and S.L. provided sci- one machinery (15, 16), or alterations in the EGFR/JAK1/ entific supervision; P. Tsvetkov and Z.B. performed research; M.G. and L.A.G. contributed STAT3 pathway (17). new reagents/analytic tools; P. Tsvetkov, E.S., D.J., P. Thiru, M.G., and P.B.G. analyzed To further explore and identify the cellular mechanisms of re- data; and P. Tsvetkov, S.S., L.W., and S.L. wrote the paper. sistance developed in the presence of distinct proteasome inhibi- Conflict of interest statement: L.A.G. is a consultant for Foundation Medicine, Novartis, tors, our group and others recently applied genome-wide genetic Boehringer Ingelheim, and Third Rock; an equity holder in Foundation Medicine; and a screens to various cell-line models (18, 19). In all models exam- member of the Scientific Advisory Board at Warp Drive. L.A.G. receives sponsored research ined, experimentally reducing the expression of one of the many support from Novartis, Astellas, BMS, and Merck. different subunits composing the 19S regulatory complex increases This article is a PNAS Direct Submission. resistance to inhibitors of the proteasome’s catalytic core (18, 19) 1To whom correspondence should be addressed. Email: [email protected]. and may show clinical relevance in multiple myeloma patients 2Deceased October 27, 2016. (18). These data suggested that transient, nonmutational mecha- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. nisms may play a key role in the ability of cancer cells to withstand 1073/pnas.1619067114/-/DCSupplemental. www.pnas.org/cgi/doi/10.1073/pnas.1619067114 PNAS Early Edition | 1of6 Downloaded by guest on September 30, 2021 A GDSC dataset B 19S regulatory complex subunits (GDSC dataset) 30 r=0.84 15 p<0.0001 20 10 10 0 (Z-score sum) 20S subunits -10 Expression -20 5 > 3 SD reduction -30 -40 -20 0 20 40 4 8 0 3 PSMC1PSMC2PSMC3PSMC4PSMC5PSMC6PSMD1PSMD2PSMD3PSMD PSMD5PSMD6PSMD7PSMD PSMD9 19S subunits PSMD1PSMD11PSMD12PSMD1PSMD14ADRM1 (Z-score sum) C D 19S regulatory complex subunits (CCLE dataset) TCGA dataset 15 120 90 10 60 30 Expression 5 # of events 0 > 3 SD reduction 0 0 4 PSMD5 PSMD1 PSMC6 PSMD6 PSMD2 PSMC2 PSMC3 PSMD7 PSMD3 PSMD9 PSMC5 PSMC4 PSMD4 PSMD8 PSMC1 ADRM1 PSMD10 PSMD12 PSMD14 PSMD11 PSMD13 PSMC2PSMC3PSMC4PSMC5PSMD1PSMD2PSMD3PSMD4PSMD5PSMD6PSMD7PSMD8PSMD9 PSMD1PSMD11PSMD12PSMD13PSMD1ADRM1 Fig. 1. Reduced expression of a single 19S proteasome complex subunit occurs naturally in many cancer cells. (A) The overall expression of all 26S proteasome subunits is tightly correlated. The correlation between the z-score sum of expression of 20S proteasome subunits versus the 19S proteasome subunits is plotted across 789 cancer cell lines taken from the GDSC dataset. Spearman correlation: r = 0.84 (P < 0.0001). (B and C) Representation of the expression levels for individual 19S subunits for each cancer cell line in the GDSC (B) and the CCLE datasets (C). Red dots indicate cells that exhibit reduced expression of a given subunit by more than 3 SDs from the mean (3σ). Each dot represents an individual cell line. (D) Analysis of primary tumor expression profiles taken from TCGA dataset. The value of each bar is the number of primary tumors exhibiting a 3σ drop in expression of at least one 19S subunit. across all 789 cell lines in the GDSC dataset. For ease of ref- paragangliomas, acute myeloid leukemias, renal cell carcinomas, erence, we designated these as “3σ lines” (Fig. 1B, red dots, and and cutaneous melanomas (Fig. S1D). Dataset S1). We also analyzed the CCLE dataset that includes Interestingly, as true for the GDSC and CCLE datasets, PSMD5 mRNA expression and DNA copy-number data for 990 cell lines. was the most commonly suppressed 19S subunit gene in human Of these, 6.3% were 3σ lines, a percentage similar to that in the tumor resection samples (Fig. 1D). In addition, other 19S subunit GDSC dataset (Fig. 1C, red dots, and Dataset S1). To determine genes that commonly showed changes in the GDSC and CCLE cell- PSMD10 PSMD1 PSMC6 PSMD6 whether this strong deviation from the mean of 19S subunits line datasets such as , , ,and were D could be explained by random variation in the expression of any suppressed in tumor resection samples as well (Fig.