Renal Medullary Carcinomas Depend Upon SMARCB1 Loss and Are Sensitive to 2 Proteasome Inhibition 3 4 5 Andrew L

bioRxiv preprint doi: https://doi.org/10.1101/487579; this version posted December 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1 Renal medullary carcinomas depend upon SMARCB1 loss and are sensitive to 2 proteasome inhibition 3 4 5 Andrew L. Hong1,2,3, Yuen-Yi Tseng3, Jeremiah Wala3, Won Jun Kim2, Bryan D. Kynnap2,

6 Mihir B. Doshi3, Guillaume Kugener3, Gabriel J. Sandoval2,3, Thomas P. Howard2, Ji Li2,

7 Xiaoping Yang3, Michelle Tillgren2, Mahmoud Ghandi3, Abeer Sayeed3, Rebecca Deasy3,

8 Abigail Ward1,2, Brian McSteen4, Katherine M. Labella2, Paula Keskula3, Adam Tracy3,

9 Cora Connor5, Catherine M. Clinton1,2, Alanna J. Church1, Brian D. Crompton1,2,3,

10 Katherine A. Janeway1,2, Barbara Van Hare4, David Sandak4, Ole Gjoerup2,

11 Pratiti Bandopadhayay1,2,3, Paul A. Clemons3, Stuart L. Schreiber3, David E. Root3,

12 Prafulla C. Gokhale2, Susan N. Chi1,2, Elizabeth A. Mullen1,2, Charles W. M. Roberts6,

13 Cigall Kadoch2,3, Rameen Beroukhim2,3,7, Keith L. Ligon2,3,7, Jesse S. Boehm3,

14 William C. Hahn2,3,7*

15 Affiliations:

16 1Boston Children’s Hospital, 300 Longwood Avenue, Boston, Massachusetts, 02115, USA

17 2Dana-Farber Cancer Institute, 450 Brookline Avenue, Boston, Massachusetts, 02215 USA

18 3Broad Institute of Harvard and MIT, 415 Main Street, Cambridge, Massachusetts, 02142 USA

19 4Rare Cancer Research Foundation, 112 S. Duke Street, Suite 101, Durham, NC 27701 USA

20 5RMC Support, P.O. Box 73156, North Charleston, S.C. 29415 USA

21 6St. Jude Children’s Research Hospital, St. Jude Children’s Research Hospital, 262 Danny Thomas

22 Place, Memphis, Tennessee 38120, USA

23 7Brigham and Women’s Hospital, 75 Francis Street, Boston, Massachusetts, 02115 USA

25 *To whom correspondence should be addressed:

26 William C. Hahn, M.D., Ph.D.

27 Department of Medical Oncology

28 Dana-Farber Cancer Institute

29 450 Brookline Avenue, Dana 1538

30 Boston, MA 02215

31 617-632-2641 (phone)

32 [email protected]

36 Abstract:

37 Renal medullary carcinoma (RMC) is a rare and deadly kidney cancer in patients of African

38 descent with sickle cell trait. Through direct-to-patient outreach, we developed genomically

39 faithful patient-derived models of RMC. Using whole genome sequencing, we identified intronic

40 fusion events in one SMARCB1 allele with concurrent loss of the other allele, confirming that

41 SMARCB1 loss occurs in RMC. Biochemical and functional characterization of these RMC

42 models revealed that RMC depends on the loss of SMARCB1 for survival and functionally

43 resemble other cancers that harbor loss of SMARCB1, such as malignant rhabdoid tumors or

44 atypical teratoid rhabdoid tumors. We performed RNAi and CRISPR-Cas9 loss of function genetic

45 screens and a small-molecule screen and identified UBE2C as an essential gene in SMARCB1

46 deficient cancers. We found that the ubiquitin-proteasome pathway was essential for the survival

47 of SMARCB1 deficient cancers in vitro and in vivo. Genetic or pharmacologic inhibition of this

48 pathway leads to G2/M arrest due to constitutive accumulation of cyclin B1. These observations

49 identify a synthetic lethal relationship that may serve as a therapeutic approach for patients with

50 SMARCB1 deficient cancers.

53 Introduction

54 Renal medullary carcinoma (RMC) was first identified in 1995 and is described as the

55 seventh nephropathy of sickle cell disease (Jr et al., 1995). RMC is a rare cancer that occurs

56 primarily in patients of African descent that carry sickle cell trait and presents during adolescence

57 with symptoms of abdominal pain, hematuria, weight loss and widely metastatic disease. Due to

58 the aggressive behavior of this disease and the small numbers of patients, no standard of care

59 exists, and patients are generally treated with therapies including nephrectomy, chemotherapy and

60 radiation therapy. Despite, this aggressive regimen, the mean overall survival rate is only 6-8

61 months (Alvarez et al., 2015; Beckermann et al., 2017; Ezekian et al., 2017; Iacovelli et al., 2015).

62 Gene expression of tumor samples from patients with RMC indicates that this entity is distinct

63 from renal cell carcinomas and urothelial carcinomas (Swartz et al., 2002; Yang et al., 2004).

64 Recent studies have identified that SMARCB1 is lost at the protein level in RMC (Cheng

65 et al., 2008). SMARCB1 is a tumor suppressor and core member of the SWI/SNF complex

66 implicated in malignant rhabdoid tumors (MRT) and atypical teratoid rhabdoid tumors

67 (ATRT)(Roberts et al., 2002). MRTs and ATRTs harbor few somatic genetic alterations and occur

68 in young children (Gröbner et al., 2018; Lawrence et al., 2013; Lee et al., 2012; Ma et al., 2018).

69 In contrast to MRTs and ATRTs, where there is biallelic loss of SMARCB1 in the genome (Chun

70 et al., 2016; Torchia et al., 2016), in RMCs, fusion events in SMARCB1 have been identified in 4

71 of 5 patients (Calderaro et al., 2016) and 8 of 10 patients (Carlo et al., 2017) using whole exome

72 sequencing (WES), fluorescence in situ hybridization (FISH), array comparative genomic

73 hybridization (CGH), and RNA-sequencing. An unresolved question is whether these cancers

74 depend upon loss of this tumor suppressor, since deletion of SMARCB1 in mice does not lead to

75 renal cancers (Roberts et al., 2002).

76 To date, limited tissue from these patients has prevented full characterization of RMC.

77 Here, we characterized new patient-derived cell lines developed from samples following

78 neoadjuvant therapy and at relapse. These cell lines have allowed us to functionally demonstrate

79 dependence of RMC on loss of SMARCB1 for survival and to uncover the proteasome as a core

80 druggable vulnerability.

82 Results

83 Derivation and genomic characterization of RMC models

84 We developed models from two patients with a diagnosis of RMC (Methods). For the first

85 patient, we obtained the primary tissue from our local institution at the time of the initial

86 nephrectomy. We generated a short-term culture normal cell line, CLF_PEDS0005_N, and a tumor

87 cell line, CLF_PEDS0005_T1 (Supp Fig. 1A). In addition, we obtained fluid from a thoracentesis

88 performed when the patient relapsed 8 months into therapy. We isolated two cell lines that grew

89 either as an adherent monolayer, CLF_PEDS0005_T2A, or in suspension, CLF_PEDS0005_T2B.

90 Each of these tumor cell lines expressed the epithelial marker, CAM5.2, and lacked expression of

91 SMARCB1 similar to that observed in the primary tumor (Supp Fig. 1B). For the second patient,

92 we partnered with the Rare Cancer Research Foundation and obtained samples through a direct-

93 to-patient portal (www.pattern.org). The primary tumor tissue from the second patient was

94 obtained at the time of the initial nephrectomy. From this sample, we generated the tumor cell line,

95 CLF_PEDS9001_T1. Both patients received 4-8 weeks neoadjuvant chemotherapy prior to their

96 nephrectomy.

97 Prior studies have identified deletion of one allele of SMARCB1 along with fusion events

98 in SMARCB1 in RMC patients (Calderaro et al., 2016; Carlo et al., 2017). Specifically, one study

99 identified SMARCB1 fusion events in 8 of 10 patients utilizing FISH, while another study

100 identified fusion events in 4 of 5 patients utilizing a combination of FISH, RNA-sequencing and

101 Sanger sequencing. We performed WES (CLF_PEDS0005) or PCR-free whole genome

102 sequencing (WGS; CLF_PEDS9001) on the primary kidney tumor tissue. In both patients, we

103 found tumor purity was <20% and confirmed the presence of sickle cell trait (Supp Fig. 1C). This

104 low tumor purity is attributable to the stromal desmoplasia seen in RMC (Swartz et al., 2002).

105 Furthermore, we failed to identify homozygous or heterozygous deletions or mutations of

106 SMARCB1 due to the low tumor purity.

107 We performed WES on the normal cell line (CLF_PEDS0005_N) or whole blood

108 (CLF_PEDS9001) and compared it to the primary tumor cell lines (CLF_PEDS0005_T1 and

109 CLF_PEDS9001_T) and metastatic cell lines (CLF_PEDS0005_T2A and CLF_PEDS0005_T2B).

110 We used the Genome Analysis Toolkit (GATK) v4.0.4.0 for variant discovery and copy number

111 analyses (Methods) (McKenna et al., 2010) and found a low mutation frequency (1-3

112 mutations/mb; Fig. 1A) in the tumor cell lines similar to that of other pediatric cancers and cell

113 lines such as MRT, ATRT and Ewing sarcoma (Cibulskis et al., 2013; Johann et al., 2016; Wala

114 et al., 2018). We performed MuTect 2.0, a method to identify somatic point mutations, on these

115 samples and filtered mutations based on the Catalogue of Somatic Mutations in Cancer (COSMIC)

116 and found that only the metastatic cell lines harbored mutations in TP53 and TPR (Supp Data 1)

117 (Cibulskis et al., 2013). From our copy number analysis, we confirmed the heterozygous loss of

118 SMARCB1. In agreement with prior studies, we failed to find an identifiable mutation or deletion

119 to account for the loss of the second SMARCB1 allele with WES.

120 We then developed a dual-color break apart FISH using bacterial artificial chromosome

121 (BAC) probes surrounding SMARCB1 to ascertain if the loss of the second SMARCB1 allele was

122 from a fusion event as seen in prior studies (Methods) (Calderaro et al., 2016; Carlo et al., 2017).

123 83% of NA10851 B-lymphocyte control cells harbored 2 fused signals indicating the presence of

124 two SMARCB1 alleles. In contrast, 82% of cells from CLF_PEDS0005_T1 and 84% of

125 CLF_PEDS9001_T exhibited only one signal and a faint fused signal. These observations suggest

126 that the SMARCB1 loci in these patient derived cell lines harbor one rearranged copy of SMARCB1

127 resulting in one signal and a deletion of SMARCB1 in the other allele resulting in the loss of the

128 majority of the fused signal (Supp Data 2).

129 We subsequently performed PCR-free WGS to assess for structural variations that would

130 not be captured by WES to elucidate the breakpoint of the rearrangement in SMARCB1. We

131 achieved an average depth of coverage of 38x for the germline DNA and 69x for the tumor cell

132 line DNA. We performed Structural variation and indel Analysis By Assembly (SvABA) v0.2.1

133 on these samples (Wala et al., 2018). In both patient-derived cell lines, we found several small

134 indels in introns and translocations (Fig. 1B; Supp Data 3-4). For CLF_PEDS0005_T1, we found

135 a large deletion between BCR and MYH9 which is predicted to lead to loss of one allele of

136 SMARCB1 (Fig. 1C and D). We then identified a balanced translocation affecting SMARCB1 that

137 is predicted to disrupt SMARCB1 function. Specifically, we identified a fusion in intron 1 of

138 SMARCB1 to the intron region following the C-terminal end of C1orf116, yielding a non-

139 functional allele. For CLF_PEDS9001_T1, we identified a large deletion between TTC28 and

140 VPREB that would lead to loss of one allele of SMARCB1 (Fig. 1E and F). We then identified a

141 balanced translocation that leads to fusion of intron 10 of PLEKHA5 to intron 6 of SMARCB1.

142 These findings confirmed prior work that identified deletion of one allele and fusions involving

143 the other allele of SMARCB1 in RMC (Calderaro et al., 2016; Carlo et al., 2017).

144 We then performed qRT-PCR to confirm that SMARCB1 was lost at the introns identified

145 by WGS in our cell lines. Specifically, we interrogated each exon-exon junction of SMARCB1 by

146 developing primers that spanned each junction. We generated cDNA libraries from TC32, an

147 Ewing Sarcoma SMARCB1 wild-type cell line, and from our RMC models. As compared with

148 TC32, we failed to find exon-exon cDNA in CLF_PEDS0005_T1 suggesting that the translocation

149 in SMARCB1 occurred between exons 1-2 (Supp Fig. 1D). For CLF_PEDS9001_T, we found that

150 SMARCB1 expression was approximately 50% of that observed in the wild type SMARCB1-

151 expressing cell line TC32 for the first 5 exons, but that none of the exons were expressed after

152 exon 5. This observation confirmed that the translocation in SMARCB1 in CLF_PEDS9001_T

153 occurs in intron 6 of SMARCB1 (Supp Fig. 1D; Methods). We then assessed the genomic DNA

154 of the primary tumor tissues and designed primers surrounding the fusions identified by WGS. We

155 confirmed that the fusions identified in the cell lines could be identified in the genomic DNA of

156 the primary tumor tissues (Supp Fig. 1E-G; Methods). Taken together, we have developed

157 models from two patients with RMC which faithfully recapitulate known genomics of this disease.

158

159 Patient-derived models of RMC are similar to SMARCB1 deficient cancers

160 We performed RNA-sequencing and transcriptomic profiling to compare the RMC models

161 to other renal tumors or tumors that harbor loss of SMARCB1. Specifically, we compared the

162 Therapeutically Applicable Research to Generate Effective Treatments (TARGET) RNA-

163 sequencing data from pediatric renal tumors (e.g. Wilms Tumor, Clear Cell Sarcoma of the

164 Kidney, and Malignant Rhabdoid Tumor) or normal kidney tissues with the RMC models using t-

165 distributed stochastic neighbor embedding (tSNE) (Methods). The normal cell line,

166 CLF_PEDS0005_N, clustered with TARGET normal kidney tissues and RMC cell lines from both

167 patients clustered with the TARGET Rhabdoid Tumor samples (Fig. 2A). These observations

168 showed that these RMC cell lines share expression patterns with patients with MRTs.

169 To determine if the RMC cell lines clustered separately among other SMARCB1-deficient

170 cancers, we performed gene expression analysis of our RMC models and compared them to

171 publicly available datasets of MRT cell lines, patients with RMC, MRT or ATRT or synovial

172 sarcoma (a cancer driven by the fusion oncoprotein SSX-S18 which displaces SMARCB1;

173 Methods) (Barretina et al., 2012; Calderaro et al., 2016; Han et al., 2016; Johann et al., 2016;

174 Richer et al., 2017). Using tSNE, we found that the RMC cell lines closely mapped to a French

175 cohort of RMC, MRT and ATRT patients (Fig. 2B). These observations demonstrated that RMC

176 cell lines and SMARCB1 deficient patients express similar gene expression programs.

177 We then assessed the consequences of re-expressing SMARCB1. Specifically, we

178 generated doxycycline-inducible open reading frame (ORF) vectors harboring SMARCB1 and

179 stably infected our RMC models, G401 (MRT cell line) and HA1E (SMARCB1 wild-type

180 immortalized epithelial kidney cell line) (Hahn et al., 1999). We confirmed that the addition of

181 doxycycline used in our studies did not affect the proliferation of the parental cell lines (Methods

182 and Supp Fig. 2A-C).

183 We used these inducible cell lines to assess the biochemical stability of the SWI/SNF

184 complex by using 10-30% glycerol gradient sedimentation followed by SDS-PAGE (Fig. 2C-D

185 and Supp Fig. 2D). In HA1E SMARCB1 wild-type cells, the SWI/SNF complex members

186 SMARCB1, ARID1A and SMARCA4 are robustly expressed at higher molecular weights (e.g.

187 fractions 13-16). In G401 SMARCB1 deficient cells, the SWI/SNF complex is smaller and seen

188 at lower molecular weights (e.g. fractions 10-14). Furthermore, expression of SWI/SNF complex

189 members was modestly decreased in G401, consistent with our prior studies (Nakayama et al.,

190 2017; Wang et al., 2016). In our RMC cell lines, we found that the majority of ARID1A and

191 SMARCA4 was observed in fractions 11-13 similar to what we found in the MRT cell line G401.

192 In addition, we found increased expression and a shift of ARID1A and SMARCA4 to larger

193 fractions 13-15 upon re-expression of SMARCB1 in the RMC lines similar to what we observed

194 in the SMARCB1 wild-type HA1E cell line. We concluded that the composition of the SWI/SNF

195 complex is similar between RMC and other SMARCB1 deficient cancers.

196 We then used the inducible cell lines to measure the consequence of SMARCB1 re-

197 expression on the viability of the cells. We also generated cell lines with inducible expression of a

198 LacZ control to compare with re-expression of SMARCB1. In the HA1E SMARCB1 wild type

199 cells, we found no significant difference in viability between induction of SMARCB1 versus

200 induction of LacZ using direct counting of viable cells (Fig. 2E). In contrast, induction of

201 SMARCB1 in the MRT G401 SMARCB1 deficient cells decreased the number of viable cells by

202 41%. Similar to G401, we found that re-expression of SMARCB1 in each of the RMC models led

203 to significant decreases in cell viability (37-62%) (Fig. 2E). These findings identified that RMC

204 is dependent on the loss of SMARCB1 for survival.

205 Since MRT cell lines arrest and senesce when SMARCB1 is re-expressed(Betz et al.,

206 2002), we looked for evidence of senescence by staining for senescence-associated acidic β-

207 galactosidase in the RMC cells when SMARCB1 was re-expressed. Following 7 days of

208 SMARCB1 or LacZ re-expression, we stained the cells for β-galactosidase (Methods). We failed

209 to observe cells expressing β-galactosidase upon expression of LacZ in the RMC cells, but when

210 SMARCB1 was expressed, we found 44.6% (+/- 17%) of the RMC cells stained for β-

211 galactosidase (Fig. 2F and Supp Fig. 2E). These studies showed that re-expression of SMARCB1

212 in RMC cells leads to senescence.

213 We then assessed what genes were differentially expressed upon SMARCB1 re-expression

214 in the RMC and MRT cell lines as another way to assess the similarity between these two cancers.

215 We performed RNA-sequencing on the doxycycline-induced SMARCB1 RMC cell lines and

216 compared it to the uninduced cell lines or doxycycline-induced LacZ cell lines. We then re-

217 analyzed our previously published studies of MRT cells with SMARCB1 re-expression and

218 compared it to our RMC cells (Wang et al., 2016). We found 1,719 genes to be significantly

219 different (false discovery rate of <0.25) in the RMC cells and 2,735 genes in MRT cells. We

220 identified 527 genes that significantly overlapped between the RMC and MRT cell lines

221 (hypergeometric p-value less than 4.035e-63; Supp Data 6). We then confirmed that these changes

222 in gene expression led to alterations in protein expression by assessing SPARC, p21, c-MYC

223 expression by immunoblotting (Supp Fig. 2F-G) (Alpers et al., 2002; Bradshaw and Sage, 2001).

224 These findings confirm that changes in the transcriptome following SMARCB1 re-expression in

225 RMC cell lines are similar to other SMARCB1 deficient cancer cell lines. In sum, these

226 observations indicate that the RMC cell lines are functionally similar to those derived from other

227 SMARCB1 deficient cancers.

228

229 RNAi and CRISPR-Cas9 loss of function screens and small-molecule screens in RMC models

230 identify proteasome inhibition as a vulnerability

231 MRT, ATRT and RMC are aggressive and incurable cancers. We performed genetic (RNAi

232 and CRISPR-Cas9) and pharmacologic screens to identify druggable targets that would decrease

233 proliferation or survival specifically for these cancers. Specifically, we used the Druggable Cancer

234 Targets (DCT v1.0) libraries and focused on targets that were identified by suppression with RNAi,

235 gene deletion with CRISPR-Cas9 based genome editing, and perturbation by small molecules

236 (Hong et al., 2016; Seashore-Ludlow et al., 2015). We accounted for off-target effects in the RNAi

237 screens by using seed controls for each shRNA.

238 We performed these three orthogonal screens on both metastatic models of RMC,

239 CLF_PEDS0005_T2A and CLF_PEDS0005_T2B (Fig. 3A). We introduced the shRNA DCT v1.0

240 lentiviral library into these two cell lines and evaluated the abundance of the shRNAs after 26 days

241 using massively parallel sequencing (Methods). We confirmed depletion of known common

242 essential genes such as RPS6 (Supp Fig. 3A). We then analyzed the differential abundance

243 between the experimental and seed control shRNAs to collapse individual shRNAs to consensus

244 gene dependencies with RNAi Gene Enrichment Ranking (RIGER) (Luo et al., 2008). Of 444

245 evaluable genes, 72 genes scored with a RIGER p-value <0.05 in CLF_PEDS0005_T2A and 74

246 genes scored in CLF_PEDS0005_T2B.

247 In parallel, we introduced the CRISPR-Cas9 DCT v1.0 lentiviral library to determine the

248 differential representation of the CRISPR-Cas9 sgRNAs between 6 and 23 days to identify genes

249 depleted or enriched in this screen by massively parallel sequencing (Methods). We confirmed

250 that the distribution of sgRNAs among biological replicates was highly correlated (Supp Fig. 3B-

251 E). When compared to the controls, there was significant depletion of essential genes such as RPS6

252 (Supp Fig. 3F-G). We used RIGER to collapse the individual sgRNAs to consensus gene

253 dependencies and found 124 genes (of a total of 445 evaluable genes) and 136 genes (of a total of

254 445 evaluable genes) with a RIGER p-value <0.05 in CLF_PEDS0005_T2A and

255 CLF_PEDS0005_T2B cell lines, respectively.

256 We performed a small-molecule screen using a library of 440 compounds that have known

257 targets in the RMC cells (CLF_PEDS0005_T2A and CLF_PEDS0005_T2B) (Hong et al., 2016).

258 This library includes 72 FDA approved compounds, 100 compounds in clinical trials and 268

259 probes based on our prior studies (Hong et al., 2016). We calculated an area under the curve (AUC)

260 based on an 8-point concentration range and considered AUCs <0.5 as significant. Of the evaluable

261 compounds, 75 (18%) compounds significantly decreased cell viability in CLF_PEDS0005_T2A

262 and 82 (20%) compounds significantly decreased cell viability in CLF_PEDS0005_T2B.

263 We then looked for genes or targets of the small molecules that scored in all three of the

264 RNAi, CRISPR-Cas9 and small-molecule screens. We identified 21 genes in

265 CLF_PEDS0005_T2A and 27 genes in CLF_PEDS0005_T2B (Data File S7) of which 19 genes

266 scored in both screens (Fig. 3A). Among the 19 genes were components of the ubiquitin-

267 proteasome system (e.g. PSMB1, PSMB2, PSMB5, PSMD1, PSMD2, and CUL1), regulators of the

268 cell cycle (CDK1, CDK6, KIF11 and PLK1), genes involved in nuclear export (KPNB1 and XPO1)

269 and ARF1, ATP1A, BIRC5, BRD4, MTOR, RRM1 and TOP2A.

270 To eliminate small molecules and targets that affect normal renal tissue, we screened the

271 normal cell line (CLF_PEDS0005_N) with the small-molecule library. We calculated the robust

272 Z-scores for these screens in relationship to the Cancer Cell Line Encyclopedia (CCLE) to

273 normalize the responses to various compounds (Barretina et al., 2012; Rees et al., 2016; Seashore-

274 Ludlow et al., 2015). We then compared the results of this small-molecule screen with the RMC

275 cancer cell lines (CLF_PEDS0005_T2A and CLF_PEDS0005_T2B). We found that the tumor

276 cells were differentially sensitive (up to two standard deviations) upon treatment with proteasome

277 inhibitors, bortezomib and MLN2238, when compared to the normal cell line (Fig. 3B). These

278 findings suggest that the vulnerability to proteasome inhibition may be dependent on loss of

279 SMARCB1.

280

281 Validation of proteasome inhibition as a specific dependency in SMARCB1 deficient cancers

282 To validate the dependency of SMARCB1 deficient tumors to the ubiquitin-proteasome

283 system, we assessed the consequences of inhibiting proteasome function on survival of the primary

284 tumor cell line, CLF_PEDS0005_T1, by deleting components of the proteasome with CRISPR-

285 Cas9. We compared these findings with a model of undifferentiated sarcoma, CLF_PEDS015T,

286 that does not harbor mutations in SMARCB1 (Hong et al., 2016). We scaled the results based on

287 the non-targeting sgRNA negative controls and RPS6 as a common essential gene (Hart et al.,

288 2015). Compared to the control sgRNAs, there was an average decrease of 29% in viability in

289 CLF_PEDS015T while there was an average decrease of 74% in CLF_PEDS0005_T1 (Fig. 3C;

290 Methods). Although deletion of the proteasome members affected proliferation in all of the

291 models (two tailed t-test p = 1.1e-5 for CLF_PEDS015T and p = 5.4e-20 for CLF_PEDS0005_T1),

292 we found that suppression of proteasome components affected the RMC model

293 CLF_PEDS0005_T1 to a statistically greater degree (two tailed t-test p = 7.1e-8). We subsequently

294 validated that gene deletion by CRISPR-Cas9 of PSMB5, one of the primary targets of proteasome

295 inhibitors, in the CLF_PEDS0005_T2A and CLF_PEDS9001_T1 cell lines led to decreased

296 viability (Supp Fig. 3H-I).

297 We then determined whether this vulnerability to proteasome inhibition was specific to the

298 loss of SMARCB1. We treated the normal cell line, CLF_PEDS0005_N, and an early passage of

299 CLF_PEDS9001_T1 while it was a heterogenous population and retained SMARCB1 (Supp Fig.

300 3J) with bortezomib or the second-generation proteasome inhibitor, MLN2238. We observed

301 significantly decreased sensitivity to the proteasome inhibitors in the SMARCB1 retained isogenic

302 cell lines as compared to the SMARCB1 deficient cell lines (Fig. 3D-E). We then treated our

303 SMARCB1-inducible RMC and MRT cell lines with DMSO, bortezomib or MLN2238. Re-

304 expression of SMARCB1 led to a decrease in sensitivity to bortezomib or MLN2238 as compared

305 to the isogenic SMARCB1 deficient lines (Fig. 3F-G). The observed differential resistance to

306 SMARCB1 re-expression was between 2-3-fold with either bortezomib or MLN2238 (Supp Fig.

307 4A-B). We concluded that re-expression of SMARCB1 partially rescued the sensitivity of MRT

308 or RMC cell lines to proteasome inhibition.

309 We then compared the results of small-molecule screens performed in SMARCB1-

310 deficient cancer cell lines in CCLE to the rest of the CCLE cell lines (n=835). We found that

311 SMARCB1-deficient cell lines were significantly more sensitive (two-tailed t-test p-value = 0.011)

312 to treatment with MLN2238 than non-multiple myeloma CCLE cell lines (Fig. 4A) (Rees et al.,

313 2016; Seashore-Ludlow et al., 2015). The degree of sensitivity was similar to that of multiple

314 myeloma cell lines which are known to be sensitive to proteasome inhibition (Dimopoulos et al.,

315 2016). These findings confirm that SMARCB1 deficient cell lines are selectively vulnerable to

316 proteasome inhibition.

317 We then performed in vitro studies to confirm the findings from these high throughput

318 small molecule screens. We treated an additional 6 SMARCB1 deficient cell lines (4 MRT and 2

319 ATRT) with bortezomib. We compared these results to H2172, a lung cancer cell line that was not

320 sensitive to proteasome inhibition in CCLE small-molecule screens, and RPMI8226, an

321 established multiple myeloma cell line that is responsive to proteasome inhibition (Hideshima et

322 al., 2001). We found that our SMARCB1 deficient cell lines exhibited single digit nanomolar

323 sensitivity to proteasome inhibition similar to that observed in the multiple myeloma cell line

324 RPMI8226 (Supp Fig. 4C). In contrast, we found that the IC50 in H2172 was at least 3-fold higher.

325 Since there are no SMARCB1 wildtype pediatric kidney cancer cell lines in CCLE, we compared

326 the sensitivity to bortezomib or MLN2238 in our RMC models with wildtype SMARCB1 patient-

327 derived Wilms tumor cell line, CLF_PEDS1012_T (Fig. 4B-C and Supp Fig. 3J). We found that

328 CLF_PEDS1012_T was more resistant to proteasome inhibition as compared to our RMC models

329 and MRT cell line, G401. These findings suggest that SMARCB1-deficient cells are more sensitive

330 to proteasome inhibition.

331

332 Proteasome inhibition leads to cell cycle arrest in G2/M and subsequent cell death

333 We then studied how SMARCB1 loss leads to a dependency on the ubiquitin-proteasome

334 system. Since activation of c-MYC has been observed in SMARCB1-deficient cancers (Cheng et

335 al., 1999; Genovese et al., 2017), we assessed how c-MYC levels are altered upon proteasome

336 inhibition. Compared to RPMI8226, a multiple myeloma cell line that relies on c-MYC for survival

337 (Tagde et al., 2016), we failed to observe suppression of c-MYC protein levels following

338 bortezomib or MLN2238 treatment (Supp Fig. 4D). We then assessed c-MYC expression levels

339 by RNA-sequencing in the G401, CLF_PEDS9001T, and CLF_PEDS00005_T1 cell lines

340 following treatment with MLN2238 and found that c-MYC levels were increased after MLN2238

341 treatment (Supp Fig. 4E). In our models of RMC and MRT, these findings suggest proteasome

342 inhibition does not lead to suppression of c-MYC.

343 We then assessed if the SWI/SNF complex is altered upon treatment with a proteasome

344 inhibitor. We treated uninduced and induced SMARCB1 cells with DMSO or a proteasome

345 inhibitor. We failed to see a consistent significant change in total or nuclear protein when

346 immunoblotting for SWI/SNF complex members (BAF57, BAF60A, BAF155, BAF170, ARID1A

347 and SMARCA4) other than increases in SMARCB1 levels upon doxycycline treatment (Supp Fig.

348 5A-B). These findings suggest that inhibition of the proteasome in SMARCB1 deficient cancers

349 and its subsequent resistance upon SMARCB1 expression does not alter the total or nuclear levels

350 of the SWI/SNF complex.

351 ER stress has been implicated as a mechanism by which proteasome inhibitors act on

352 multiple myeloma cells (Obeng et al., 2006). We saw an increase in protein expression of markers

353 of ER stress, GRP78 and IRE1α, following treatment with MLN2238 (Supp Fig. 5C). Upon re-

354 expression of SMARCB1 and subsequent treatment with a proteasome inhibitor, we did not see

355 changes in either GRP78 or IRE1α protein levels (Supp Fig. 5C). These observations suggest that

356 although ER stress markers are elevated upon proteasome inhibition in SMARCB1-deficient cell

357 lines, they are not rescued by SMARCB1 re-expression.

358 We performed gene ontology (GO) based Gene Set Enrichment Analysis

359 (GSEA)(Subramanian et al., 2005) on the 527 significantly altered genes upon re-expression of

360 SMARCB1 (Supp Data 6) to identify classes of gene function enriched in this group of genes.

361 We identified numerous gene sets that involved the ubiquitin-proteasome system (Supp Data 8).

362 We then performed RNA-sequencing on G401 and the RMC cell lines treated with DMSO or

363 MLN2238. We identified 1,758 genes which were significantly (FDR<0.1) up- or down-regulated

364 upon treatment with MLN2238 (Supp Data 9). We compared the 527 differentially expressed

365 genes identified upon re-expression of SMARCB1 with the 1,758 differentially expressed genes

366 identified upon treatment with MLN2238 and identified 92 genes which overlapped. Of these

367 genes, we identified 63 genes which were differentially expressed with re-expression of

368 SMARCB1 and were differentially expressed with treatment with MLN2238 (Fig. 4D). From this

369 refined gene set, we performed GO-based GSEA (Subramanian et al., 2005) and found significant

370 enrichment (adjusted p-value ranging from 0 to 0.0061) in gene sets involving the cell cycle (Supp

371 Data 10).

372 We then assessed the cell cycle by DNA content with cells treated with DMSO or

373 MLN2238 for 24 hours as proteasome inhibitors have been found to cause a G2/M cell cycle arrest

374 in lymphomas, colorectal carcinomas, hepatocellular carcinomas, and glioblastoma multiforme

375 (Augello et al., 2018; Bavi et al., 2011; Gu et al., 2017; Yin et al., 2005). We observed a significant

376 shift in cells to G2/M (two-tailed t-test p-value 0.0005; Fig. 4E). Upon re-expression of

377 SMARCB1, we saw that this phenotype was rescued (Fig. 4F). By 48 hours, we saw a significant

378 increase in markers of programed cell death such as Annexin V and PI positive cells (Fig. 4G).

379 We found that treatment with a proteasome inhibitor led to increased cleaved caspase-3

380 levels in addition to changes to Annexin V and PI, suggesting that inhibition of the ubiquitin-

381 proteasome system leads to programmed cell death (Fig. 4H). We then asked whether restoration

382 of SMARCB1 expression inhibited cleaved-caspase 3 activation. We found that induction of

383 cleaved caspase-3 was less pronounced when SMARCB1 was re-expressed (Fig. 4H). These

384 observations suggest that proteasome inhibitors initially lead to a SMARCB1-dependent G2/M

385 cell cycle arrest and subsequent programmed cell death.

386

387 Proteasome inhibitor induced G2/M arrest is mediated in part by inappropriate cyclin B1

388 degradation driven by a dependency on UBE2C

389 We subsequently searched for genes related to the ubiquitin-proteasome system and

390 SMARCB1 function. We defined a set of 204 genes that were upregulated when comparing the

391 log2 fold change between SMARCB1 deficient cells to SMARCB1 re-expressed cells in RMC and

392 MRT cell lines (Supp Data 6). We then took this set of 204 genes and examined the Project

393 Achilles (genome scale CRISPR-Cas9 loss of function screens) DepMap Public 18Q3 dataset

394 (Meyers et al., 2017) to determine whether any SMARCB1 deficient cell lines required expression

395 of these genes for survival. This dataset included loss of function screens from 485 cancer cell

396 lines and included three ATRT SMARCB1-deficient cancer cell lines: COGAR359,

397 CHLA06ATRT and CHLA266. We found that SMARCB1 deficient cancer cell lines required

398 UBE2C, a ubiquitin-conjugating enzyme, for survival. We noted that these cell lines were in the

399 top 5% of cell lines (n=485) that required UBE2C for survival (empirical Bayes moderated t-test

400 p-value = 0.00016; Fig. 5A-B). These observations suggested that cancer cell lines that lack

401 SMARCB1 were also dependent on UBE2C.

402 Since the 3 cell lines profiled were ATRTs, we validated that the RMC cell lines were also

403 dependent on UBE2C for survival. We generated sgRNAs specific for UBE2C and assessed

404 viability by cell counting following gene deletion. We saw a significant decrease in cell viability

405 in SMARCB1 deficient cell lines as compared to urothelial carcinoma cell line, JMSU1 (SWI/SNF

406 wild type), or non-small cell lung cancer cell line, A549 (SMARCA4 mutant) (two-tailed t-test p-

407 values 4.5e-5 and 4.6e-5; Fig. 5C and Supp Fig. 6A).

408 UBE2C serves as the E2 enzyme which adds the first ubiquitin (Ub) to cyclin B1 for

409 degradation (Dimova et al., 2012; Grice et al., 2015). Cyclin B1 degradation is required in G2/M

410 at the end of metaphase to enter anaphase(Chang et al., 2003). Our integrated RNAi, CRISPR-

411 Cas9 and small molecule screens identified that our RMC models required expression of PLK1

412 and CDK1, genes involved in G2/M, for survival (Fig. 3A-B), and prior studies have identified

413 that inhibition of PLK1 in ATRT or MRT cells lead to arrest in G2/M (Alimova et al., 2017;

414 Morozov et al., 2007). Treatment of the RMC cell lines with MLN2238 led to accumulation of

415 cyclin B1 as compared to cyclin D1 suggesting that MLN2238 inhibits degradation of cyclin B1

416 (Fig. 5D). When we re-expressed SMARCB1, we found that cyclin B1 levels were unchanged

417 upon MLN2238 treatment. Although APC/C serves as the E3 ligase for cyclin B1, genetic deletion

418 of APC/C in Project Achilles showed that APC/C was an essential gene across all cancer cell lines.

419 These findings suggest SMARCB1 deficient cancer cells require UBE2C expression for survival,

420 in part by regulating cyclin B1 stability.

421

422 Effects of proteasome inhibition in vivo

423 These studies identify a lethal interaction between suppressing the ubiquitin-proteasome

424 system and SMARCB1-deficient cancers in vitro. The doses used in this study were based on in

425 vitro studies of multiple myeloma or lymphoma cell lines (Chauhan et al., 2011; Garcia et al.,

426 2016; Hideshima et al., 2003, 2001). For patients with primary or refractory multiple myeloma,

427 use of proteasome inhibitors has led to significant clinical responses (Jagannath et al., 2004;

428 Moreau et al., 2016; Richardson et al., 2005, 2003). We reasoned that if our SMARCB1 deficient

429 cancers were susceptible to proteasome inhibitors at similar in vitro dosing, we would also see

430 similar in vivo responses. We first determined whether these doses led to proteasome inhibition by

431 assessing the ability of these cell lines to cleave Suc-LLVY-aminoluciferin. We found that

432 treatment of the SMARCB1 deficient cell lines with either bortezomib or MLN2238 led to

433 inhibition of the proteasome to a similar extent observed when the multiple myeloma cell line

434 RPMI8226 was treated (Supp Fig. 6B; Methods). We also simulated the pharmacodynamics of

435 proteasome inhibitors in vivo by treating cells in vitro with a pulse dose of proteasome inhibitors

436 as has been performed in multiple myeloma and chronic myeloid leukemia cell lines (Crawford et

437 al., 2014; Kuhn et al., 2007; Shabaneh et al., 2013). We found that upon treatment with MLN2238,

438 SMARCB1 deficient cells arrested in G2/M, which led to cell death as measured by Annexin V/PI

439 staining and led to accumulation of cyclin B1 similar to treatment with a continuous dose of

440 MLN2238 (Methods; Supp Fig. 6C-F and 7A-C).

441 We subsequently performed in vivo studies to confirm the effect of monotherapy with

442 proteasome inhibition in tumor xenografts. We used the rhabdoid tumor cell line, G401, for our in

443 vivo studies because we noted that the primary tumor cell lines (CLF_PEDS0005_T1 and

444 CLF_PEDS9001_T) did not form subcutaneous xenograft tumors in immunodeficient mice (Supp

445 Fig. 7D-E). We allowed tumors to achieve an average volume of 148 mm3, treated mice with either

446 vehicle or MLN2238 (7 mg/kg twice a week) and measured tumors twice weekly. Treatment with

447 MLN2238 induced significant tumor stabilization (n=4) or regression (n=3) when assessed either

448 by relative growth (two-tailed t-test p-value 0.038; Supp Fig. 7F) or by absolute tumor volume

449 (two-tailed t-test p-value 0.012; Fig. 5E) and did not induce significant changes in body weight as

450 compared to vehicle controls (two-tailed t-test p-value 0.154; Supp Fig. 7G). Furthermore, there

451 was a significant survival for mice treated with MLN2238 (p-value 0.0489 by log-rank (Mantel-

452 Cox) test; Fig. 5F). Combined, these results demonstrate that MLN2238 induces tumor regression

453 in SMARCB1 deficient tumors in vivo.

454

455

456 Discussion

457 We have developed faithful patient-derived models of RMC which have been genomically

458 validated using PCR-free WGS, WES, RNA-sequencing and gene expression profiling. We have

459 shown that these models are dependent upon the loss of SMARCB1 for survival. Re-expression of

460 SMARCB1 in RMC leads to a significant decrease in cell counts and a senescence phenotype.

461 Biochemically, re-expression of SMARCB1 in RMC leads to stabilization of the SWI/SNF

462 complex in the same manner as for MRT. Diagnostically, patients with RMC are often

463 misdiagnosed with renal cell carcinoma (RCC) due to the rarity of RMC, the lack of access to

464 SMARCB1 histological stains and unknown sickle cell status (Beckermann et al., 2017). Although

465 SMARCB1 is currently included in targeted sequencing efforts nationwide(“AACR Project

466 GENIE: Powering Precision Medicine through an International Consortium,” 2017), our studies

467 along with prior studies (Calderaro et al., 2016; Carlo et al., 2017) suggest that conventional target

468 exome sequencing may fail to identify patients with RMC.

469 Patients with RMC and other SMARCB1 deficient cancers have a poor prognosis despite

470 aggressive multi-modal therapy. Using genetic and pharmacologic screens in these RMC models,

471 we identified the ubiquitin-proteasome system as a specific vulnerability in RMC. When we

472 looked more broadly at other SMARCB1 deficient cancers such as MRT and ATRT, we found

473 that these models were similarly sensitive to inhibition of the ubiquitin-proteasome system. Re-

474 expression of SMARCB1 partially rescued the sensitivity to proteasome inhibitors in RMC and

475 MRT models.

476 Prior studies have tested bortezomib in the G401 MRT cell line in an in vitro setting or

477 implicated MYC signaling and downstream activation of ER stress as a mechanism for sensitivity

478 to proteasome inhibitors in Kras/Tp53 mutant pancreatic cancers with Smarcb1 deficiency

479 (Genovese et al., 2017; Moreno and Kerl, 2016). However, the background of mutant KRAS may

480 be contributing to these findings as mutant HRAS or KRAS cancers are sensitive to enhanced

481 proteotoxic stress and ER stress (De Raedt et al., 2011; Denoyelle et al., 2006). Furthermore,

482 KRAS mutant cancers depend on several proteasome components in genome scale RNAi screens

483 (Aguirre and Hahn, 2018; Barbie et al., 2009; Luo et al., 2009). Our studies have identified that

484 the ubiquitin-proteasome system is a core vulnerability among a compendium of druggable targets

485 as tested by orthogonal methods of RNA interference, CRISPR-Cas9 gene deletion or small

486 molecule inhibition. We found that proteasome inhibition in SMARCB1-deficient cancer cell lines

487 result in cells accumulating in G2/M due to inappropriate degradation of cyclin B1.

488 Taken together, these results explain the sensitivity of SMARCB1 deficient lines to

489 proteasome inhibitors such as MLN2238 in vitro and in vivo. There have been case reports of one

490 adult and 2 children with RMC who exhibited extraordinary responses of 2-7 years following

491 diagnosis after empiric therapy with bortezomib has been used either as monotherapy or in

492 combination with chemotherapy (Carden et al., 2017; Ronnen et al., 2006). Our results suggest

493 that future clinical trials could incorporate oral proteasome inhibitor such as MLN2238 for patients

494 with RMC and potentially more broadly across SMARCB1 deficient cancers. Finally, our studies

495 show how interaction with patient advocacy foundations and online direct-to-patient consent

496 process may serve as a model to study other rare cancers and lead to advances in rare tumor

497 research (Sharifnia et al., 2017).

498

499 Materials and Methods

500 Derivation of RMC models: Patients assented or families consented to IRB approved protocols.

501 Patient PEDS0005 whole blood, adjacent normal kidney and tumor tissue were obtained within 6

502 hours as part of the nephrectomy following neoadjuvant chemotherapy. Upon relapse, pleural fluid

503 was obtained from a palliative thoracentesis. The tumor and adjacent normal kidney tissue was

504 minced into 2-3mm3 cubes. CLF_PEDS0005 primary tissue was then dissociated as previously

505 described and cultured in both RPMI media containing 10% FBS (Sigma) or DMEM/F-12 media

506 containing ROCK inhibitor, Y-27632, insulin, cholera toxin, 5% FBS, and penicillin/streptomycin

507 (Liu et al., 2012). Samples were gently passaged when cultures achieved 80-90% confluence. The

508 normal kidney cell line was named CLF_PEDS0005_N. The tumor kidney cell line was named

509 CLF_PEDS0005_T1. For the pleural fluid sample, samples were grown initially in conditioned

510 media as previously published (Liu et al., 2012). Adherent and suspension cells were continuously

511 passaged when cells reached confluence. By passage 5, cells were noted to be growing as adherent

512 and suspension cells, and these were sub-cultured to yield CLF_PEDS0005_T2A and

513 CLF_PEDS0005_T2B, respectively. Samples were then transitioned to DMEM/F-12 media or

514 RPMI as above at passage 13.

515 Patient PEDS9001 whole blood, adjacent normal kidney and tumor tissue were obtained

516 similarly to PEDS0005. Discarded tissue from the nephrectomy following neoadjuvant

517 chemotherapy was sent to our institution within 24 hours of resection. For the normal kidney and

518 tumor tissue, samples were minced to 2-3mm3 and plated onto 6 well plates (Corning, NY).

519 Following mincing, tumor samples were cultured without further digestion. Tumor samples were

520 grown in culture in the DMEM/F-12 media as described above to yield CLF_PEDS9001_T1, while

521 the normal samples yielded a cell culture that matched the tumor cell line.

522 Breakapart fluorescent in situ hybridization (FISH): We developed a custom dual-color

523 breakapart FISH probe, using BAC probes surrounding SMARCB1 at 22q11.23: RP11-662F7

524 (telomeric to SMARCB1, labeled in green) and RP11-1112A23 (centromeric to SMARCB1,

525 labeled in red) (Empire Genomics, Buffalo, NY). The probe set was hybridized to a normal control

526 to confirm chromosomal locations and to determine the frequency of expected fusion signals in

527 normal cells. 50 nuclei were scored by two independent observers (n=100 per cell line) in the

528 CLF_PEDS0005 and CLF_PEDS9001 models.

529 Whole Exome Sequencing: We performed whole exome sequencing (WES) from genomic DNA

530 extracted from whole blood, normal / tumor tissues, and our patient derived cell lines as noted in

531 the text. One ug of gDNA (as measured on a Nanodrop 1000 (Thermo Fisher Scientific)) was used

532 to perform standard (~60x mean target coverage for normal) or deep (~150x mean target coverage

533 for tumor and cell lines) WES. Illumina (Dedham, MA) chemistry used.

534 Whole Genome Sequencing: We performed PCR-free whole genome sequencing (WGS) from

535 gDNA extracted from whole blood, normal / tumor tissues, and our patient derived cell lines as

536 noted in the text. Two ug of gDNA was used to perform standard (for normal) or deep (for tumor

537 and cell lines) coverage. Illumina (Dedham, MA) HiSeq X Ten v2 chemistry was used. Average

538 coverage is indicated in the text.

539 RNA-sequencing: For Fig. 2A, samples were processed using Illumina TruSeq strand specific

540 sequencing. We performed poly-A selection of mRNA transcripts and obtained a sequencing depth

541 of at least 50 million aligned reads per sample. For SMARCB1 re-expression RNAseq experiments

542 and MLN2238 vs DMSO treated experiments, samples were collected as biological replicates or

543 triplicates. RNA was extracted using Qiagen RNeasy Plus Mini Kit (Qiagen, Hilden, Germany).

544 RNA was normalized using the Qubit RNA HS Assay (Thermo Fisher Scientific). Five hundred

545 ng of normalized RNA was subsequently used to create libraries with the Kapa Stranded mRNA-

546 seq kit (Kapa Biosystems, KK8420; Wilmington, MA). cDNA libraries were then quantitatively

547 and qualitatively assessed on a BioAnalyzer 2100 (Agilent, Santa Clara, CA) and by qRT-PCR

548 with Kapa Library Quantification Kit. Libraries were subsequently loaded on an Illumina HiSeq

549 2500 and achieved an average read depth of 10 million reads per replicate.

550 Genomic analyses: WGS - Samples were aligned to Hg19. SvABA v0.2.1 was used to identify

551 large deletions and structural variations (Forbes et al., 2015). WES – Samples were aligned to

552 Hg19. Samples were analyzed using GATK v4.0.4.0 for copy number variation (CNV), single

553 nucleotide polymorphism (SNP) and indel identification across our RMC samples simultaneously

554 using filtering parameters set by GATK (Broad Institute, Cambridge, MA) (McKenna et al., 2010).

555 MuTect2 was used to identify candidate somatic mutations and these were correlated with

556 COSMIC (Forbes et al., 2015). RNA – CLF_PEDS0005 and CLF_PEDS9001 samples and

557 TARGET Wilms and Rhabdoid tumor samples (dbGaP phs000218.v19.p7) were aligned or re-

558 aligned with STAR and transcript quantification performed with RSEM. The TARGET initiative

559 is managed by the NCI and information can be found at http://ocg.cancer.gov/programs/target.

560 These normalized samples were then analyzed with t-SNE (Maaten, 2014). The following

561 parameters were used in the t-SNE analyses: perplexity 10, theta 0, iterations 3000. RNA

562 sequencing samples in the SMARCB1 re-expression studies were subsequently aligned and

563 analyzed with the Tuxedo suite (e.g. aligned with TopHat 2.0.11, abundance estimation with

564 CuffLinks, differential analysis with CuffDiff and CummeRbund) (Trapnell et al., 2010). For the

565 comparison to previously published work (Wang et al., 2016), the published RNA sequencing

566 samples along with our samples were re-aligned with TopHat 2.0.11 and analyzed with the Tuxedo

567 suite. For samples treated with DMSO or MLN2238, samples were aligned as above and analyzed

568 with DESeq2 (Love et al., 2014).

569 Sanger sequencing confirmation of WGS findings: gDNA was extracted using QIAamp DNA

570 mini kit (Qiagen). We performed a mixing study of our RMC cell lines with gDNA isolated from

571 the G401 MRT cell line and then performed PCR amplification. We determined that the lower

572 limits of detection of these fusions with our methods were ~1% of tumor cell line gDNA with a

573 minimum 50ng of gDNA (Supp Fig. 1E). We subsequently performed the same PCR reactions

574 with 100ng of gDNA from the tumor tissue samples. Samples were gel purified and submitted for

575 Sanger sequencing (Eton Bio). We found that the sequences from the tumor tissue samples

576 matched those of the cell lines (Supp Fig. 1F and G), confirming that the genomic alterations that

577 we found in the cell lines reflect those found in the original tumor. Primers utilized were

578 CLF_PEDS0005 chr 1 forward (ATAAGACATAACTTGGCCGG), CLF_PEDS0005 SMARCB1

579 reverse (TTTTCCAAAAGGTTTACAAGGC), CLF_PEDS9001 chr12 forward

580 (AAAAGCATATGTATCCCTTGCT), CLF_PEDS9001 SMARCB1 reverse

581 (CCTCCAGAGCCAGCAGA).

582 Quantitative RT-PCR: RNA was extracted as above and normalized using the Nanodrop to 1ug.

583 One ug of RNA was then added to the High Capacity cDNA Reverse Transcription Kit (Thermo

584 Fisher Scientific) and PCR reactions were performed as per manufacturer’s recommendations.

585 cDNA was then diluted and added to primers (Supp Data 11) and Power SYBR Green PCR Master

586 Mix (Thermo Fisher Scientific). Samples were run on a BioRad CFX384 qPCR System in a

587 minimum of technical quadruplicates. Results shown are representative of at least 2 biological

588 replicates.

589 Gene-Expression Array analysis: We performed Affymetrix Human Genome U133 Plus 2.0 on

590 our RMC cell lines. We then combined the following GEO datasets using a GenePattern module

591 with robust multi-array (RMA) normalization GSE64019, GSE70421, GSE70678, GSE36133,

592 GSE94321 (Barretina et al., 2012; Calderaro et al., 2016; Johann et al., 2016; Richer et al., 2017;

593 Wang et al., 2016). We utilized COMBAT and then tSNE to account for batch effects and to

594 identify clusters of similarity (Chen et al., 2011; Johnson et al., 2007; Maaten, 2014).

595 Glycerol Gradients followed by SDS-PAGE: Nuclear extracts and gradients were performed as

596 previously published(Boulay et al., 2017). Briefly, 500ug of nuclear extract from approximately

597 30 million cells was resuspended in 0% glycerol HEMG buffer containing 1mM DTT, cOmplete

598 protease inhibitors and PhosStop (Roche). This was placed on a 10-30% glycerol gradient and

599 ultracentrifuged at 40k RPM for 16 hours at 4C. Following centrifugation, fractions of 550uL were

600 collected. Samples were then prepared with 1x LDS Sample Buffer (Thermo). Samples were run

601 on a 4-12% Bis-Tris gel and then transferred by immunoblotting in tris-glycine-SDS buffer with

602 methanol. Immunoblots were subsequently blocked with Licor Blocking Buffer (Lincoln, NE) and

603 then incubated with antibodies as noted in the methods section. Immunoblots shown are

604 representative of at least 2 biological replicates.

605 Cell lines: Primary cell lines were authenticated by Fluidigm or WES/WGS sequencing and tested

606 for mouse antibody production and mycoplasma. Established cell lines were authenticated by

607 Fluidigm SNP testing. Cell lines were refreshed after approximately 20 passages from the frozen

608 stock.

609 Small molecules: Bortezomib or MLN2238 was purchased from SelleckChem (Houston, TX) for

610 the in vitro studies. Compounds were resuspended in DMSO and frozen down in 20uL aliquots to

611 limit freeze-thaw cycles. Compounds were added as noted in the figure legends. In vitro studies

612 used 15nM for bortezomib and 100nM for MLN2238. For the pulse experiments, we used 2.5uM

613 of MLN2238.

614 SMARCB1 induction studies: pDONR223 SMARCB1 was Sanger sequenced (Eton Bio) and

615 aligned to the variant 2 of SMARCB1. SMARCB1 was subsequently cloned into the inducible

616 vector pLXI401 or pLXI403 (Genomics Perturbation Platform at the Broad Institute, Cambridge,

617 MA) by Gateway Cloning. LacZ was used as a control. Lentivirus was produced using tet-free

618 serum (Clontech, Mountain View, CA). Cell lines were infected with lentivirus to generate stable

619 cell lines. Cell lines were then confirmed to re-express LacZ or SMARCB1 by titrating levels of

620 doxycycline (Clontech). Parental cell lines were treated with increasing doses of doxycycline to

621 determine the toxicity to cells and measured by Cell-TiterGlo after 96 hours or by trypan blue

622 exclusion cell counting after 2 weeks (Supp Fig. 2A-B). Cells were then grown with or without

623 doxycycline in a 6 well plate. Cells were counted by Trypan blue exclusion on a ViCELL XR

624 (Beckman Coulter, Brea, CA) every 4-5 days. Results shown are the average of at least 3 biological

625 replicates.

626 Scenesence assays: Cells were plated in a 6 well dish and treated with or without doxycycline for

627 up to 7 days. Senescence was assessed with the Senescence β-Galactosidase Staining Kit without

628 modifications (Cell Signaling Technologies, Danvers, MA).

629 Druggable Cancer Targets (DCT) v1.0 shRNA/sgRNA libraries, pooled screens and small-

630 molecule profiling: These were performed as previously published (Hong et al., 2016). Briefly,

631 we utilized the DCT v1.0 shRNA (CP1050) and sgRNA (CP0026) libraries from the Broad

632 Institute Genetic Perturbation Platform (GPP) (http://www.broadinstitute.org/rnai/public/).

633 Viruses from both pools were generated as outlined at the GPP portal. As CLF_PEDS0005_T2A

634 and CLF_PEDS0005_T2B were expanded, we performed titrations with the libraries as outlined

635 at the GPP portal. For the sgRNA pool, both cell lines were first transduced with Cas9 expression

636 vector pXPR_BRD111. We screened the DCT v1.0 shRNA library in biological replicates and the

637 Cas9 expressing cell lines with the sgRNA pools at an early passage (<20) and at a multiplicity of

638 infection (MOI) <1, at a mean representation rate above 500 cells per sgRNA or shRNA. gDNA

639 was extracted and was submitted for sequencing of the barcodes. We achieved sequencing depths

640 of at least 500 reads per shRNA or sgRNA.

641 CRISPR-Cas9 validation studies: sgRNAs targeting the genes noted in the manuscript (e.g.

642 PSMB5, UBE2C and controls; Supp Data 12) were generated and introduced into the

643 pXPR_BRD003 backbone. These were then sequence confirmed by Sanger sequencing (Eton

644 Biosciences). Lentivirus was produced and used for infection to generate stable cell lines

645 expressing Cas9. Cells were counted or harvested for protein as noted in the text.

646 Immunoblots: After indicated treatments, cell lysates were harvested using RIPA buffer (Cell

647 Signaling Technologies) with protease inhibitors (cOmplete, Roche) and phosphatase inhibitors

648 (PhosSTOP, Roche). Antibodies used were as follows: ARID1A (Santa Cruz; sc-373784), α-

649 tubulin (Santa Cruz; sc-5286), β-Actin (C-4) (Santa Cruz; sc-47778), β-Actin (Cell Signaling;

650 8457), BAF155 (Santa Cruz; sc-10756), BAF170 (Santa Cruz; sc-166237), SMARCA4 (Santa

651 Cruz; 17798), BRM (Bethyl; A301-015A), CDT1 (Cell Signaling; 8064S), Cleaved Caspase-3

652 (Cell Signaling; 9664), c-MYC (Santa Cruz; sc-764) or c-MYC (Cell Signaling; 9402), cyclin B1

653 (Cell Signaling; 4135 and 4138), cyclin D1 (Santa Cruz; sc-718), Flag M2 (Sigma; F1804),

654 GAPDH (Cell Signaling; 2118S and 97166S), GRP78 (Rockland Antibodies, Limerick, PA; 200-

655 301-F36), IRE1-alpha (Cell Signaling; 3294), lamin A/C (Cell Signaling; 2032), PSMB5 (Abcam,

656 Cambridge, MA; ab3330), p21 Waf1/Cip1 (Cell Signaling; 2947), p53 (DO-1) (Santa Cruz; sc-

657 126), SMARCB1/SNF5 (Bethyl A301-087A), SPARC (Cell Signaling; 5420S), Ubiquitin (Cell

658 Signaling; 3936), UBE2C (Proteintech, Rosemont, IL; 66087-1). Results shown are representative

659 of at least 2 biological replicates.

660 Cell cycle and Annexin V/PI: Cells were treated using the conditions noted in the text. Two

661 million cells were spun down and resuspended in PBS. Cells were then subjected to FITC Annexin

662 V and PI staining as described (BD Pharmigen; 556547). Another set of cells were subjected to

663 PI/RNAse staining (BD Pharmingen; 550825 or Invitrogen F10797). Samples were analyzed

664 within 1 hour with the SA3800 Spectral Analyzer (Sony Biotechnology). Biological replicates

665 were performed. Data was analyzed with FlowJo v10 (FlowJo, Ashland, OR).

666 Proteasome function assay: We measured the cell’s ability to cleave Suc-LLVY-aminoluciferin

667 utilizing Proteasome-Glo (Promega) following a one-hour treatment with the noted proteasome

668 inhibitor and measured luminescence. Results shown are from at least 2 biological replicates.

669 In vivo tumor injections: This research project has been reviewed and approved by the Dana-

670 Farber Cancer Institute’s Animal Care and Use Committee (IACUC), in compliance with the

671 Animal Welfare Act and the Office of Laboratory Welfare (OLAW) of the National Institutes of

672 Health (NIH). Five million cells of G401 in 100µL of a 50% PBS / 50% Matrigel (BD Biosciences)

673 mixture were injected subcutaneously into flanks unilaterally in Taconic NCr-Nude (CrTac:NCr-

674 Foxn1nu) female mice at 7 weeks of age. When tumors reached approximately 150 mm3, mice

675 were randomized into various treatment groups: vehicle control (5% 2-hydroxypropyl-beta-

676 cyclodextrin (HPbCD)) or MLN2238 (7 mg/kg IV twice a week for 4 weeks). MLN2238 (diluted

677 in 5% 2-HPbCD) was purchased from MedChem Express. Randomizations to the treatment arm

678 occurred. Blinding was not performed. Statistical analysis was performed using the two-tailed t-

679 test or Mantel-Cox as noted in the text.

680

681

682 Supplementary Materials:

683 Supp Fig.1. Patient derived RMC models reflective of known biology.

684 Supp Fig.2. Cell lines tolerate treatment with doxycycline and re-express SMARCB1 upon

685 doxycycline induction.

686 Supp Fig.3. Functional genomic screens identify proteasome inhibition as a vulnerability in

687 RMC and other SMARCB1 deficient cancers.

688 Supp Fig.4. Sensitivity seen from treatment with proteasome inhibitors can be rescued with re-

689 expression of SMARCB1.

690 Supp Fig.5. Treatment with proteasome inhibition does not lead to changes that can be rescued

691 by SMARCB1 re-expression in the nuclear or cytoplasic levels of SWI/SNF complex members

692 or ER stress proteins.

693 Supp Fig.6. Pulse dosing of proteasome inhibitors leads to suppression of the proteasome, arrests

694 cells in G2/M and leads to programmed cell death.

695 Supp Fig.7. In vivo studies identify MLN2238 induces tumor regression.

696 Supp Data 1. Significant mutations identified by MuTect2

697 Supp Data 2. SMARCB1 Fluorescence In Situ Hybridization results

698 Supp Data 3. Structural changes identified by SvABA in CLF_PEDS0005_T

699 Supp Data 4. Structural changes identified by SvABA in CLF_PEDS9001_T

700 Supp Data 5. Fusion sequences identified by PCR-Free Whole Genome Sequencing

701 Supp Data 6. Average differential expression across inducible SMARCB1 RMC and MRT cell

702 lines following SMARCB1 re-expression

703 Supp Data 7. Overlap between RNAi, CRISPR-Cas9 and small-molecule screens

704 Supp Data 8. Gene Ontology Gene Set Enrichment Analysis from SMARCB1 re-expression

705 studies

706 Supp Data 9. Average differential expression across SMARCB1 RMC and MRT cell lines

707 following DMSO or MLN2238 treatment

708 Supp Data 10. Gene Ontology Gene Set Enrichment Analysis from cells treated with MLN2238

709 Supp Data 11. SMARCB1 exon-exon junction qRT-PCR primers

710 Supp Data 12. sgRNAs used in the CRISPR-Cas9 validation studies

711

712

713 References:

714 AACR Project GENIE: Powering Precision Medicine through an International Consortium. 715 2017. . Cancer Discovery 7:818–831. doi:10.1158/2159-8290.CD-17-0151 716 Aguirre AJ, Hahn WC. 2018. Synthetic Lethal Vulnerabilities in KRAS-Mutant Cancers. Cold 717 Spring Harb Perspect Med 8. doi:10.1101/cshperspect.a031518 718 Alimova I, Pierce AM, Harris P, Donson A, Birks DK, Prince E, Balakrishnan I, Foreman NK, 719 Kool M, Hoffman L, Venkataraman S, Vibhakar R. 2017. Targeting Polo-like kinase 1 in 720 SMARCB1 deleted atypical teratoid rhabdoid tumor. Oncotarget 8:97290–97303. 721 doi:10.18632/oncotarget.21932 722 Alpers CE, Hudkins KL, Segerer S, Sage EH, Pichler R, Couser WG, Johnson RJ, Bassuk JA. 723 2002. Localization of SPARC in developing, mature, and chronically injured human 724 allograft kidneys. Kidney International 62:2073–2086. doi:10.1046/j.1523- 725 1755.2002.00680.x 726 Alvarez O, Rodriguez MM, Jordan L, Sarnaik S. 2015. Renal medullary carcinoma and sickle 727 cell trait: A systematic review. Pediatric blood & cancer 62:1694–1699. 728 doi:10.1002/pbc.25592 [doi] 729 Augello G, Modica M, Azzolina A, Puleio R, Cassata G, Emma MR, Di Sano C, Cusimano A, 730 Montalto G, Cervello M. 2018. Preclinical evaluation of antitumor activity of the 731 proteasome inhibitor MLN2238 (ixazomib) in hepatocellular carcinoma cells. Cell Death 732 Dis 9:28. doi:10.1038/s41419-017-0195-0 733 Barbie DA, Tamayo P, Boehm JS, Kim SY, Moody SE, Dunn IF, Schinzel AC, Sandy P, Meylan 734 E, Scholl C, Fröhling S, Chan EM, Sos ML, Michel K, Mermel C, Silver SJ, Weir BA, 735 Reiling JH, Sheng Q, Gupta PB, Wadlow RC, Le H, Hoersch S, Wittner BS, 736 Ramaswamy S, Livingston DM, Sabatini DM, Meyerson M, Thomas RK, Lander ES, 737 Mesirov JP, Root DE, Gilliland DG, Jacks T, Hahn WC. 2009. Systematic RNA 738 interference reveals that oncogenic KRAS-driven cancers require TBK1. Nature 739 462:108–112. doi:10.1038/nature08460 740 Barretina J, Caponigro G, Stransky N, Venkatesan K, Margolin AA, Kim S, Wilson CJ, Lehar J, 741 Kryukov GV, Sonkin D, Reddy A, Liu M, Murray L, Berger MF, Monahan JE, Morais P, 742 Meltzer J, Korejwa A, Jane-Valbuena J, Mapa FA, Thibault J, Bric-Furlong E, Raman P, 743 Shipway A, Engels IH, Cheng J, Yu GK, Yu J, Jr PA, Silva M de, Jagtap K, Jones MD, 744 Wang L, Hatton C, Palescandolo E, Gupta S, Mahan S, Sougnez C, Onofrio RC, Liefeld 745 T, MacConaill L, Winckler W, Reich M, Li N, Mesirov JP, Gabriel SB, Getz G, Ardlie 746 K, Chan V, Myer VE, Weber BL, Porter J, Warmuth M, Finan P, Harris JL, Meyerson M, 747 Golub TR, Morrissey MP, Sellers WR, Schlegel R, Garraway LA. 2012. The Cancer Cell 748 Line Encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature 749 483:603–607. doi:10.1038/nature11003 [doi] 750 Bavi P, Uddin S, Ahmed M, Jehan Z, Bu R, Abubaker J, Sultana M, Al-Sanea N, Abduljabbar A, 751 Ashari LH, Alhomoud S, Al-Dayel F, Prabhakaran S, Hussain AR, Al-Kuraya KS. 2011. 752 Bortezomib Stabilizes Mitotic Cyclins and Prevents Cell Cycle Progression via Inhibition 753 of UBE2C in Colorectal Carcinoma. The American Journal of Pathology 178:2109–2120. 754 doi:10.1016/j.ajpath.2011.01.034 755 Beckermann KE, Sharma D, Chaturvedi S, Msaouel P, Abboud MR, Allory Y, Bourdeaut F, 756 Calderaro J, Cubas AA de, Derebail VK, Hong AL, Naik RP, Malouf GG, Mullen EA, 757 Reuter VE, Roberts CWM, Walker CL, Wood CG, DeBaun MR, Poppel HV, Tannir NM,

758 Rathmell WK. 2017. Renal Medullary Carcinoma: Establishing Standards in Practice. 759 Journal of Oncology Practice 13:414–421. doi:10.1200/JOP.2017.020909 760 Betz BL, Strobeck MW, Reisman DN, Knudsen ES, Weissman BE. 2002. Re-expression of 761 hSNF5/INI1/BAF47 in pediatric tumor cells leads to G1 arrest associated with induction 762 of p16ink4a and activation of RB. Oncogene 21:5193–5203. doi:10.1038/sj.onc.1205706 763 [doi] 764 Boulay G, Sandoval GJ, Riggi N, Iyer S, Buisson R, Naigles B, Awad ME, Rengarajan S, 765 Volorio A, McBride MJ, Broye LC, Zou L, Stamenkovic I, Kadoch C, Rivera MN. 2017. 766 Cancer-Specific Retargeting of BAF Complexes by a Prion-like Domain. Cell 767 171:178.e19. doi:10.1016/j.cell.2017.07.036 768 Bradshaw AD, Sage EH. 2001. SPARC, a matricellular protein that functions in cellular 769 differentiation and tissue response to injury. The Journal of Clinical Investigation 770 107:1049–1054. doi:10.1172/JCI12939 771 Calderaro J, Masliah-Planchon J, Richer W, Maillot L, Maille P, Mansuy L, Bastien C, Taille A 772 de la, Boussion H, Charpy C, Jourdain A, Blechet C, Pierron G, Gentien D, Choudat L, 773 Tournigand C, Delattre O, Allory Y, Bourdeaut F. 2016. Balanced Translocations 774 Disrupting SMARCB1 Are Hallmark Recurrent Genetic Alterations in Renal Medullary 775 Carcinomas. European urology 69:1055–1061. doi:10.1016/j.eururo.2015.09.027 [doi] 776 Carden MA, Smith S, Meany H, Yin H, Alazraki A, Rapkin LB. 2017. Platinum plus bortezomib 777 for the treatment of pediatric renal medullary carcinoma: Two cases. Pediatric blood & 778 cancer 64:10.1002/pbc.26402. Epub 2017 Jan 4. doi:10.1002/pbc.26402 [doi] 779 Carlo MI, Chaim J, Patil S, Kemel Y, Schram AM, Woo K, Coskey D, Nanjangud GJ, Voss MH, 780 Feldman DR, Hsieh JJ, Hakimi AA, Chen YB, Motzer RJ, Lee CH. 2017. Genomic 781 Characterization of Renal Medullary Carcinoma and Treatment Outcomes. Clinical 782 genitourinary cancer. doi:S1558-7673(17)30096-4 [pii] 783 Chang DC, Xu N, Luo KQ. 2003. Degradation of cyclin B is required for the onset of anaphase 784 in Mammalian cells. J Biol Chem 278:37865–37873. doi:10.1074/jbc.M306376200 785 Chauhan D, Tian Z, Zhou B, Kuhn D, Orlowski R, Raje N, Richardson P, Anderson KC. 2011. 786 In vitro and in vivo selective antitumor activity of a novel orally bioavailable proteasome 787 inhibitor MLN9708 against multiple myeloma cells. Clin Cancer Res 17:5311–5321. 788 doi:10.1158/1078-0432.CCR-11-0476 789 Chen C, Grennan K, Badner J, Zhang D, Gershon E, Jin L, Liu C. 2011. Removing Batch Effects 790 in Analysis of Expression Microarray Data: An Evaluation of Six Batch Adjustment 791 Methods. PLOS ONE 6:e17238. doi:10.1371/journal.pone.0017238 792 Cheng JX, Tretiakova M, Gong C, Mandal S, Krausz T, Taxy JB. 2008. Renal medullary 793 carcinoma: rhabdoid features and the absence of INI1 expression as markers of 794 aggressive behavior. Modern pathology : an official journal of the United States and 795 Canadian Academy of Pathology, Inc 21:647–652. doi:10.1038/modpathol.2008.44 [doi] 796 Cheng SW, Davies KP, Yung E, Beltran RJ, Yu J, Kalpana GV. 1999. c-MYC interacts with 797 INI1/hSNF5 and requires the SWI/SNF complex for transactivation function. Nature 798 genetics 22:102–105. doi:10.1038/8811 [doi] 799 Chun HJ, Lim EL, Heravi-Moussavi A, Saberi S, Mungall KL, Bilenky M, Carles A, Tse K, 800 Shlafman I, Zhu K, Qian JQ, Palmquist DL, He A, Long W, Goya R, Ng M, LeBlanc 801 VG, Pleasance E, Thiessen N, Wong T, Chuah E, Zhao YJ, Schein JE, Gerhard DS, 802 Taylor MD, Mungall AJ, Moore RA, Ma Y, Jones SJ, Perlman EJ, Hirst M, Marra MA. 803 2016. Genome-Wide Profiles of Extra-cranial Malignant Rhabdoid Tumors Reveal

804 Heterogeneity and Dysregulated Developmental Pathways. Cancer cell 29:394–406. 805 doi:10.1016/j.ccell.2016.02.009 [doi] 806 Cibulskis K, Lawrence MS, Carter SL, Sivachenko A, Jaffe D, Sougnez C, Gabriel S, Meyerson 807 M, Lander ES, Getz G. 2013. Sensitive detection of somatic point mutations in impure 808 and heterogeneous cancer samples. Nature Biotechnology 31:213–219. 809 doi:10.1038/nbt.2514 810 Crawford LJ, Chan ET, Aujay M, Holyoake TL, Melo JV, Jorgensen HG, Suresh S, Walker B, 811 Irvine AE. 2014. Synergistic effects of proteasome inhibitor carfilzomib in combination 812 with tyrosine kinase inhibitors in imatinib-sensitive and -resistant chronic myeloid 813 leukemia models. Oncogenesis 3:e90. doi:10.1038/oncsis.2014.3 814 De Raedt T, Walton Z, Yecies JL, Li D, Chen Y, Malone CF, Maertens O, Jeong SM, Bronson 815 RT, Lebleu V, Kalluri R, Normant E, Haigis MC, Manning BD, Wong K-K, Macleod 816 KF, Cichowski K. 2011. Exploiting cancer cell vulnerabilities to develop a combination 817 therapy for ras-driven tumors. Cancer Cell 20:400–413. doi:10.1016/j.ccr.2011.08.014 818 Denoyelle C, Abou-Rjaily G, Bezrookove V, Verhaegen M, Johnson TM, Fullen DR, Pointer JN, 819 Gruber SB, Su LD, Nikiforov MA, Kaufman RJ, Bastian BC, Soengas MS. 2006. Anti- 820 oncogenic role of the endoplasmic reticulum differentially activated by mutations in the 821 MAPK pathway. Nat Cell Biol 8:1053–1063. doi:10.1038/ncb1471 822 Dimopoulos MA, Sonneveld P, Leung N, Merlini G, Ludwig H, Kastritis E, Goldschmidt H, 823 Joshua D, Orlowski RZ, Powles R, Vesole DH, Garderet L, Einsele H, Palumbo A, Cavo 824 M, Richardson PG, Moreau P, Miguel JS, Rajkumar SV, Durie BG, Terpos E. 2016. 825 International Myeloma Working Group Recommendations for the Diagnosis and 826 Management of Myeloma-Related Renal Impairment. Journal of clinical oncology : 827 official journal of the American Society of Clinical Oncology 34:1544–1557. 828 doi:10.1200/JCO.2015.65.0044 [doi] 829 Dimova NV, Hathaway NA, Lee B-H, Kirkpatrick DS, Berkowitz ML, Gygi SP, Finley D, King 830 RW. 2012. APC/C-mediated multiple monoubiquitylation provides an alternative 831 degradation signal for cyclin B1. Nature Cell Biology 14:168–176. doi:10.1038/ncb2425 832 Ezekian B, Englum B, Gilmore BF, Nag UP, Kim J, Leraas HJ, Routh JC, Rice HE, Tracy ET. 833 2017. Renal medullary carcinoma: A national analysis of 159 patients. Pediatric blood & 834 cancer. doi:10.1002/pbc.26609 [doi] 835 Forbes SA, Beare D, Gunasekaran P, Leung K, Bindal N, Boutselakis H, Ding M, Bamford S, 836 Cole C, Ward S, Kok CY, Jia M, De T, Teague JW, Stratton MR, McDermott U, 837 Campbell PJ. 2015. COSMIC: exploring the world’s knowledge of somatic mutations in 838 human cancer. Nucleic Acids Research 43:805. doi:10.1093/nar/gku1075 839 Garcia JS, Huang M, Medeiros BC, Mitchell BS. 2016. Selective Toxicity of Investigational 840 Ixazomib for Human Leukemia Cells Expressing Mutant Cytoplasmic NPM1: Role of 841 Reactive Oxygen Species. Clin Cancer Res 22:1978–1988. doi:10.1158/1078-0432.CCR- 842 15-1440 843 Genovese G, Carugo A, Tepper J, Robinson FS, Li L, Svelto M, Nezi L, Corti D, Minelli R, 844 Pettazzoni P, Gutschner T, Wu C-C, Seth S, Akdemir KC, Leo E, Amin S, Molin MD, 845 Ying H, Kwong LN, Colla S, Takahashi K, Ghosh P, Giuliani V, Muller F, Dey P, Jiang 846 S, Garvey J, Liu C-G, Zhang J, Heffernan TP, Toniatti C, Fleming JB, Goggins MG, 847 Wood LD, Sgambato A, Agaimy A, Maitra A, Roberts CWM, Wang H, Viale A, 848 DePinho RA, Draetta GF, Chin L. 2017. Synthetic vulnerabilities of mesenchymal 849 subpopulations in pancreatic cancer. Nature 542:362–366. doi:10.1038/nature21064

850 Grice GL, Lobb IT, Weekes MP, Gygi SP, Antrobus R, Nathan JA. 2015. The Proteasome 851 Distinguishes between Heterotypic and Homotypic Lysine-11-Linked Polyubiquitin 852 Chains. Cell Reports 12:545–553. doi:10.1016/j.celrep.2015.06.061 853 Gröbner SN, Worst BC, Weischenfeldt J, Buchhalter I, Kleinheinz K, Rudneva VA, Johann PD, 854 Balasubramanian GP, Segura-Wang M, Brabetz S, Bender S, Hutter B, Sturm D, Pfaff E, 855 Hübschmann D, Zipprich G, Heinold M, Eils J, Lawerenz C, Erkek S, Lambo S, Waszak 856 S, Blattmann C, Borkhardt A, Kuhlen M, Eggert A, Fulda S, Gessler M, Wegert J, 857 Kappler R, Baumhoer D, Burdach S, Kirschner-Schwabe R, Kontny U, Kulozik AE, 858 Lohmann D, Hettmer S, Eckert C, Bielack S, Nathrath M, Niemeyer C, Richter GH, 859 Schulte J, Siebert R, Westermann F, Molenaar JJ, Vassal G, Witt H, Burkhardt B, Kratz 860 CP, Witt O, Tilburg CM van, Kramm CM, Fleischhack G, Dirksen U, Rutkowski S, 861 Frühwald M, Hoff K von, Wolf S, Klingebiel T, Koscielniak E, Landgraf P, Koster J, 862 Resnick AC, Zhang J, Liu Y, Zhou X, Waanders AJ, Zwijnenburg DA, Raman P, Brors 863 B, Weber UD, Northcott PA, Pajtler KW, Kool M, Piro RM, Korbel JO, Schlesner M, 864 Eils R, Jones DTW, Lichter P, Chavez L, Zapatka M, Pfister SM. 2018. The landscape of 865 genomic alterations across childhood cancers. Nature. doi:10.1038/nature25480 866 Gu JJ, Kaufman GP, Mavis C, Czuczman MS, Hernandez-Ilizaliturri FJ. 2017. Mitotic 867 catastrophe and cell cycle arrest are alternative cell death pathways executed by 868 bortezomib in rituximab resistant B-cell lymphoma cells. Oncotarget 8:12741–12753. 869 doi:10.18632/oncotarget.14405 870 Hahn WC, Counter CM, Lundberg AS, Beijersbergen RL, Brooks MW, Weinberg RA. 1999. 871 Creation of human tumour cells with defined genetic elements. Nature 400:464–468. 872 doi:10.1038/22780 873 Han Z-Y, Richer W, Fréneaux P, Chauvin C, Lucchesi C, Guillemot D, Grison C, Lequin D, 874 Pierron G, Masliah-Planchon J, Nicolas A, Ranchère-Vince D, Varlet P, Puget S, 875 Janoueix-Lerosey I, Ayrault O, Surdez D, Delattre O, Bourdeaut F. 2016. The occurrence 876 of intracranial rhabdoid tumours in mice depends on temporal control of Smarcb1 877 inactivation. Nature Communications 7:10421. doi:10.1038/ncomms10421 878 Hart T, Chandrashekhar M, Aregger M, Steinhart Z, Brown KR, MacLeod G, Mis M, 879 Zimmermann M, Fradet-Turcotte A, Sun S, Mero P, Dirks P, Sidhu S, Roth FP, Rissland 880 OS, Durocher D, Angers S, Moffat J. 2015. High-Resolution CRISPR Screens Reveal 881 Fitness Genes and Genotype-Specific Cancer Liabilities. Cell 163:1515–1526. 882 doi:10.1016/j.cell.2015.11.015 883 Hideshima T, Mitsiades C, Akiyama M, Hayashi T, Chauhan D, Richardson P, Schlossman R, 884 Podar K, Munshi NC, Mitsiades N, Anderson KC. 2003. Molecular mechanisms 885 mediating antimyeloma activity of proteasome inhibitor PS-341. Blood 101:1530–1534. 886 doi:10.1182/blood-2002-08-2543 887 Hideshima T, Richardson P, Chauhan D, Palombella VJ, Elliott PJ, Adams J, Anderson KC. 888 2001. The proteasome inhibitor PS-341 inhibits growth, induces apoptosis, and 889 overcomes drug resistance in human multiple myeloma cells. Cancer Research 61:3071– 890 3076. 891 Hong AL, Tseng YY, Cowley GS, Jonas O, Cheah JH, Kynnap BD, Doshi MB, Oh C, Meyer 892 SC, Church AJ, Gill S, Bielski CM, Keskula P, Imamovic A, Howell S, Kryukov GV, 893 Clemons PA, Tsherniak A, Vazquez F, Crompton BD, Shamji AF, Rodriguez-Galindo C, 894 Janeway KA, Roberts CW, Stegmaier K, Hummelen P van, Cima MJ, Langer RS, 895 Garraway LA, Schreiber SL, Root DE, Hahn WC, Boehm JS. 2016. Integrated genetic

896 and pharmacologic interrogation of rare cancers. Nature communications 7:11987. 897 doi:10.1038/ncomms11987 [doi] 898 Iacovelli R, Modica D, Palazzo A, Trenta P, Piesco G, Cortesi E. 2015. Clinical outcome and 899 prognostic factors in renal medullary carcinoma: A pooled analysis from 18 years of 900 medical literature. Canadian Urological Association journal = Journal de l’Association 901 des urologues du Canada 9:172. doi:10.5489/cuaj.2373 [doi] 902 Jagannath S, Barlogie B, Berenson J, Siegel D, Irwin D, Richardson PG, Niesvizky R, Alexanian 903 R, Limentani SA, Alsina M, Adams J, Kauffman M, Esseltine D-L, Schenkein DP, 904 Anderson KC. 2004. A phase 2 study of two doses of bortezomib in relapsed or 905 refractory myeloma. Br J Haematol 127:165–172. doi:10.1111/j.1365- 906 2141.2004.05188.x 907 Johann PD, Erkek S, Zapatka M, Kerl K, Buchhalter I, Hovestadt V, Jones DTW, Sturm D, 908 Hermann C, Wang MS, Korshunov A, Rhyzova M, Gröbner S, Brabetz S, Chavez L, 909 Bens S, Gröschel S, Kratochwil F, Wittmann A, Sieber L, Geörg C, Wolf S, Beck K, 910 Oyen F, Capper D, Sluis P van, Volckmann R, Koster J, Versteeg R, Deimling A von, 911 Milde T, Witt O, Kulozik AE, Ebinger M, Shalaby T, Grotzer M, Sumerauer D, 912 Zamecnik J, Mora J, Jabado N, Taylor MD, Huang A, Aronica E, Bertoni A, 913 Radlwimmer B, Pietsch T, Schüller U, Schneppenheim R, Northcott PA, Korbel JO, 914 Siebert R, Frühwald MC, Lichter P, Eils R, Gajjar A, Hasselblatt M, Pfister SM, Kool M. 915 2016. Atypical Teratoid/Rhabdoid Tumors Are Comprised of Three Epigenetic 916 Subgroups with Distinct Enhancer Landscapes. Cancer Cell 29:379–393. 917 doi:10.1016/j.ccell.2016.02.001 918 Johnson WE, Li C, Rabinovic A. 2007. Adjusting batch effects in microarray expression data 919 using empirical Bayes methods. Biostatistics 8:118–127. doi:10.1093/biostatistics/kxj037 920 Jr CJD, Mostofi FK, Sesterhenn IA. 1995. Renal medullary carcinoma. The seventh sickle cell 921 nephropathy. The American Journal of Surgical Pathology 19:1–11. 922 Kuhn DJ, Chen Q, Voorhees PM, Strader JS, Shenk KD, Sun CM, Demo SD, Bennett MK, van 923 Leeuwen FWB, Chanan-Khan AA, Orlowski RZ. 2007. Potent activity of carfilzomib, a 924 novel, irreversible inhibitor of the ubiquitin-proteasome pathway, against preclinical 925 models of multiple myeloma. Blood 110:3281–3290. doi:10.1182/blood-2007-01-065888 926 Lawrence MS, Stojanov P, Polak P, Kryukov GV, Cibulskis K, Sivachenko A, Carter SL, 927 Stewart C, Mermel CH, Roberts SA, Kiezun A, Hammerman PS, McKenna A, Drier Y, 928 Zou L, Ramos AH, Pugh TJ, Stransky N, Helman E, Kim J, Sougnez C, Ambrogio L, 929 Nickerson E, Shefler E, Cortes ML, Auclair D, Saksena G, Voet D, Noble M, DiCara D, 930 Lin P, Lichtenstein L, Heiman DI, Fennell T, Imielinski M, Hernandez B, Hodis E, Baca 931 S, Dulak AM, Lohr J, Landau DA, Wu CJ, Melendez-Zajgla J, Hidalgo-Miranda A, 932 Koren A, McCarroll SA, Mora J, Lee RS, Crompton B, Onofrio R, Parkin M, Winckler 933 W, Ardlie K, Gabriel SB, Roberts CW, Biegel JA, Stegmaier K, Bass AJ, Garraway LA, 934 Meyerson M, Golub TR, Gordenin DA, Sunyaev S, Lander ES, Getz G. 2013. Mutational 935 heterogeneity in cancer and the search for new cancer-associated genes. Nature 499:214– 936 218. doi:10.1038/nature12213 [doi] 937 Lee RS, Stewart C, Carter SL, Ambrogio L, Cibulskis K, Sougnez C, Lawrence MS, Auclair D, 938 Mora J, Golub TR, Biegel JA, Getz G, Roberts CWM. 2012. A remarkably simple 939 genome underlies highly malignant pediatric rhabdoid cancers. The Journal of Clinical 940 Investigation 122:2983–2988. doi:10.1172/JCI64400

941 Liu X, Ory V, Chapman S, Yuan H, Albanese C, Kallakury B, Timofeeva OA, Nealon C, Dakic 942 A, Simic V, Haddad BR, Rhim JS, Dritschilo A, Riegel A, McBride A, Schlegel R. 2012. 943 ROCK inhibitor and feeder cells induce the conditional reprogramming of epithelial cells. 944 The American journal of pathology 180:599–607. doi:10.1016/j.ajpath.2011.10.036; 945 10.1016/j.ajpath.2011.10.036 946 Love MI, Huber W, Anders S. 2014. Moderated estimation of fold change and dispersion for 947 RNA-seq data with DESeq2. Genome Biol 15:550. doi:10.1186/s13059-014-0550-8 948 Luo B, Cheung HW, Subramanian A, Sharifnia T, Okamoto M, Yang X, Hinkle G, Boehm JS, 949 Beroukhim R, Weir BA, Mermel C, Barbie DA, Awad T, Zhou X, Nguyen T, Piqani B, 950 Li C, Golub TR, Meyerson M, Hacohen N, Hahn WC, Lander ES, Sabatini DM, Root 951 DE. 2008. Highly parallel identification of essential genes in cancer cells. Proceedings of 952 the National Academy of Sciences of the United States of America 105:20380–20385. 953 doi:10.1073/pnas.0810485105; 10.1073/pnas.0810485105 954 Luo J, Emanuele MJ, Li D, Creighton CJ, Schlabach MR, Westbrook TF, Wong K-K, Elledge 955 SJ. 2009. A genome-wide RNAi screen identifies multiple synthetic lethal interactions 956 with the Ras oncogene. Cell 137:835–848. doi:10.1016/j.cell.2009.05.006 957 Ma X, Liu Yu, Liu Yanling, Alexandrov LB, Edmonson MN, Gawad C, Zhou X, Li Y, Rusch 958 MC, Easton J, Huether R, Gonzalez-Pena V, Wilkinson MR, Hermida LC, Davis S, 959 Sioson E, Pounds S, Cao X, Ries RE, Wang Z, Chen X, Dong L, Diskin SJ, Smith MA, 960 Auvil JMG, Meltzer PS, Lau CC, Perlman EJ, Maris JM, Meshinchi S, Hunger SP, 961 Gerhard DS, Zhang J. 2018. Pan-cancer genome and transcriptome analyses of 1,699 962 paediatric leukaemias and solid tumours. Nature. doi:10.1038/nature25795 963 Maaten L van der. 2014. Accelerating t-SNE using Tree-Based Algorithms. Journal of Machine 964 Learning Research 15:3221–3245. 965 McKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, Kernytsky A, Garimella K, 966 Altshuler D, Gabriel S, Daly M, DePristo MA. 2010. The Genome Analysis Toolkit: a 967 MapReduce framework for analyzing next-generation DNA sequencing data. Genome 968 Research 20:1297–1303. doi:10.1101/gr.107524.110 969 Meyers RM, Bryan JG, McFarland JM, Weir BA, Sizemore AE, Xu H, Dharia NV, Montgomery 970 PG, Cowley GS, Pantel S, Goodale A, Lee Y, Ali LD, Jiang G, Lubonja R, Harrington 971 WF, Strickland M, Wu T, Hawes DC, Zhivich VA, Wyatt MR, Kalani Z, Chang JJ, 972 Okamoto M, Stegmaier K, Golub TR, Boehm JS, Vazquez F, Root DE, Hahn WC, 973 Tsherniak A. 2017. Computational correction of copy number effect improves specificity 974 of CRISPR-Cas9 essentiality screens in cancer cells. Nature Genetics 49:1779–1784. 975 doi:10.1038/ng.3984 976 Moreau P, Masszi T, Grzasko N, Bahlis NJ, Hansson M, Pour L, Sandhu I, Ganly P, Baker BW, 977 Jackson SR, Stoppa AM, Simpson DR, Gimsing P, Palumbo A, Garderet L, Cavo M, 978 Kumar S, Touzeau C, Buadi FK, Laubach JP, Berg DT, Lin J, Bacco AD, Hui AM, Velde 979 H van de, Richardson PG, Group T-MS. 2016. Oral Ixazomib, Lenalidomide, and 980 Dexamethasone for Multiple Myeloma. The New England journal of medicine 374:1621– 981 1634. doi:10.1056/NEJMoa1516282 [doi] 982 Moreno N, Kerl K. 2016. Preclinical Evaluation of Combined Targeted Approaches in Malignant 983 Rhabdoid Tumors. Anticancer Res 36:3883–3887. 984 Morozov A, Lee SJ, Zhang Z-K, Cimica V, Zagzag D, Kalpana GV. 2007. INI1 induces 985 interferon signaling and spindle checkpoint in rhabdoid tumors. Clinical Cancer

986 Research: An Official Journal of the American Association for Cancer Research 987 13:4721–4730. doi:10.1158/1078-0432.CCR-07-0054 988 Nakayama RT, Pulice JL, Valencia AM, McBride MJ, McKenzie ZM, Gillespie MA, Ku WL, 989 Teng M, Cui K, Williams RT, Cassel SH, Qing H, Widmer CJ, Demetri GD, Irizarry RA, 990 Zhao K, Ranish JA, Kadoch C. 2017. SMARCB1 is required for widespread BAF 991 complex-mediated activation of enhancers and bivalent promoters. Nature Genetics. 992 doi:10.1038/ng.3958 993 Obeng EA, Carlson LM, Gutman DM, Jr WJH, Lee KP, Boise LH. 2006. Proteasome inhibitors 994 induce a terminal unfolded protein response in multiple myeloma cells. Blood 107:4907– 995 4916. doi:2005-08-3531 [pii] 996 Rees MG, Seashore-Ludlow B, Cheah JH, Adams DJ, Price EV, Gill S, Javaid S, Coletti ME, 997 Jones VL, Bodycombe NE, Soule CK, Alexander B, Li A, Montgomery P, Kotz JD, Hon 998 CS, Munoz B, Liefeld T, Dancik V, Haber DA, Clish CB, Bittker JA, Palmer M, Wagner 999 BK, Clemons PA, Shamji AF, Schreiber SL. 2016. Correlating chemical sensitivity and 1000 basal gene expression reveals mechanism of action. Nature chemical biology 12:109– 1001 116. doi:10.1038/nchembio.1986 [doi] 1002 Richardson PG, Barlogie B, Berenson J, Singhal S, Jagannath S, Irwin D, Rajkumar SV, 1003 Srkalovic G, Alsina M, Alexanian R, Siegel D, Orlowski RZ, Kuter D, Limentani SA, 1004 Lee S, Hideshima T, Esseltine D-L, Kauffman M, Adams J, Schenkein DP, Anderson 1005 KC. 2003. A phase 2 study of bortezomib in relapsed, refractory myeloma. N Engl J Med 1006 348:2609–2617. doi:10.1056/NEJMoa030288 1007 Richardson PG, Sonneveld P, Schuster MW, Irwin D, Stadtmauer EA, Facon T, Harousseau J-L, 1008 Ben-Yehuda D, Lonial S, Goldschmidt H, Reece D, San-Miguel JF, Bladé J, Boccadoro 1009 M, Cavenagh J, Dalton WS, Boral AL, Esseltine DL, Porter JB, Schenkein D, Anderson 1010 KC, Assessment of Proteasome Inhibition for Extending Remissions (APEX) 1011 Investigators. 2005. Bortezomib or high-dose dexamethasone for relapsed multiple 1012 myeloma. N Engl J Med 352:2487–2498. doi:10.1056/NEJMoa043445 1013 Richer W, Masliah-Planchon J, Clement N, Jimenez I, Maillot L, Gentien D, Albaud B, Chemlali 1014 W, Galant C, Larousserie F, Boudou-Rouquette P, Leruste A, Chauvin C, Han ZY, 1015 Coindre J-M, Varlet P, Freneaux P, Ranchère-Vince D, Delattre O, Bourdeaut F. 2017. 1016 Embryonic signature distinguishes pediatric and adult rhabdoid tumors from other 1017 SMARCB1-deficient cancers. Oncotarget 8:34245–34257. 1018 doi:10.18632/oncotarget.15939 1019 Roberts CW, Leroux MM, Fleming MD, Orkin SH. 2002. Highly penetrant, rapid tumorigenesis 1020 through conditional inversion of the tumor suppressor gene Snf5. Cancer cell 2:415–425. 1021 doi:S153561080200185X [pii] 1022 Ronnen EA, Kondagunta GV, Motzer RJ. 2006. Medullary renal cell carcinoma and response to 1023 therapy with bortezomib. Journal of clinical oncology : official journal of the American 1024 Society of Clinical Oncology 24:e14. doi:24/9/e14 [pii] 1025 Seashore-Ludlow B, Rees MG, Cheah JH, Cokol M, Price EV, Coletti ME, Jones V, Bodycombe 1026 NE, Soule CK, Gould J, Alexander B, Li A, Montgomery P, Wawer MJ, Kuru N, Kotz 1027 JD, Hon CS, Munoz B, Liefeld T, Dancik V, Bittker JA, Palmer M, Bradner JE, Shamji 1028 AF, Clemons PA, Schreiber SL. 2015. Harnessing Connectivity in a Large-Scale Small- 1029 Molecule Sensitivity Dataset. Cancer discovery. doi:2159-8290.CD-15-0235 [pii]

1030 Shabaneh TB, Downey SL, Goddard AL, Screen M, Lucas MM, Eastman A, Kisselev AF. 2013. 1031 Molecular basis of differential sensitivity of myeloma cells to clinically relevant bolus 1032 treatment with bortezomib. PLoS ONE 8:e56132. doi:10.1371/journal.pone.0056132 1033 Sharifnia T, Hong AL, Painter CA, Boehm JS. 2017. Emerging Opportunities for Target 1034 Discovery in Rare Cancers. Cell Chemical Biology 24:1075–1091. 1035 doi:10.1016/j.chembiol.2017.08.002 1036 Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, Paulovich A, 1037 Pomeroy SL, Golub TR, Lander ES, Mesirov JP. 2005. Gene set enrichment analysis: a 1038 knowledge-based approach for interpreting genome-wide expression profiles. 1039 Proceedings of the National Academy of Sciences of the United States of America 1040 102:15545–15550. doi:0506580102 [pii] 1041 Swartz MA, Karth J, Schneider DT, Rodriguez R, Beckwith JB, Perlman EJ. 2002. Renal 1042 medullary carcinoma: clinical, pathologic, immunohistochemical, and genetic analysis 1043 with pathogenetic implications. Urology 60:1083–1089. doi:S0090429502021544 [pii] 1044 Tagde A, Rajabi H, Bouillez A, Alam M, Gali R, Bailey S, Tai Y-T, Hideshima T, Anderson K, 1045 Avigan D, Kufe D. 2016. MUC1-C drives MYC in multiple myeloma. Blood 127:2587– 1046 2597. doi:10.1182/blood-2015-07-659151 1047 Torchia J, Golbourn B, Feng S, Ho KC, Sin-Chan P, Vasiljevic A, Norman JD, Guilhamon P, 1048 Garzia L, Agamez NR, Lu M, Chan TS, Picard D, Antonellis P de, Khuong-Quang D-A, 1049 Planello AC, Zeller C, Barsyte-Lovejoy D, Lafay-Cousin L, Letourneau L, Bourgey M, 1050 Yu M, Gendoo DMA, Dzamba M, Barszczyk M, Medina T, Riemenschneider AN, 1051 Morrissy AS, Ra Y-S, Ramaswamy V, Remke M, Dunham CP, Yip S, Ng H-K, Lu J-Q, 1052 Mehta V, Albrecht S, Pimentel J, Chan JA, Somers GR, Faria CC, Roque L, Fouladi M, 1053 Hoffman LM, Moore AS, Wang Y, Choi SA, Hansford JR, Catchpoole D, Birks DK, 1054 Foreman NK, Strother D, Klekner A, Bognár L, Garami M, Hauser P, Hortobágyi T, 1055 Wilson B, Hukin J, Carret A-S, Meter TEV, Hwang EI, Gajjar A, Chiou S-H, Nakamura 1056 H, Toledano H, Fried I, Fults D, Wataya T, Fryer C, Eisenstat DD, Scheinemann K, 1057 Fleming AJ, Johnston DL, Michaud J, Zelcer S, Hammond R, Afzal S, Ramsay DA, 1058 Sirachainan N, Hongeng S, Larbcharoensub N, Grundy RG, Lulla RR, Fangusaro JR, 1059 Druker H, Bartels U, Grant R, Malkin D, McGlade CJ, Nicolaides T, Tihan T, Phillips J, 1060 Majewski J, Montpetit A, Bourque G, Bader GD, Reddy AT, Gillespie GY, Warmuth- 1061 Metz M, Rutkowski S, Tabori U, Lupien M, Brudno M, Schüller U, Pietsch T, Judkins 1062 AR, Hawkins CE, Bouffet E, Kim S-K, Dirks PB, Taylor MD, Erdreich-Epstein A, 1063 Arrowsmith CH, Carvalho DDD, Rutka JT, Jabado N, Huang A. 2016. Integrated (epi)- 1064 Genomic Analyses Identify Subgroup-Specific Therapeutic Targets in CNS Rhabdoid 1065 Tumors. Cancer Cell 30:891–908. doi:10.1016/j.ccell.2016.11.003 1066 Trapnell C, Williams BA, Pertea G, Mortazavi A, Kwan G, Baren MJ van, Salzberg SL, Wold 1067 BJ, Pachter L. 2010. Transcript assembly and quantification by RNA-Seq reveals 1068 unannotated transcripts and isoform switching during cell differentiation. Nature 1069 biotechnology 28:511–515. doi:10.1038/nbt.1621 [doi] 1070 Wala JA, Bandopadhayay P, Greenwald NF, O’Rourke R, Sharpe T, Stewart C, Schumacher S, 1071 Li Y, Weischenfeldt J, Yao X, Nusbaum C, Campbell P, Getz G, Meyerson M, Zhang C- 1072 Z, Imielinski M, Beroukhim R. 2018. SvABA: genome-wide detection of structural 1073 variants and indels by local assembly. Genome Res 28:581–591. 1074 doi:10.1101/gr.221028.117

1075 Wang X, Lee RS, Alver BH, Haswell JR, Wang S, Mieczkowski J, Drier Y, Gillespie SM, 1076 Archer TC, Wu JN, Tzvetkov EP, Troisi EC, Pomeroy SL, Biegel JA, Tolstorukov MY, 1077 Bernstein BE, Park PJ, Roberts CW. 2016. SMARCB1-mediated SWI/SNF complex 1078 function is essential for enhancer regulation. Nature genetics. doi:10.1038/ng.3746 [doi] 1079 Yang XJ, Sugimura J, Tretiakova MS, Furge K, Zagaja G, Sokoloff M, Pins M, Bergan R, 1080 Grignon DJ, Stadler WM, Vogelzang NJ, Teh BT. 2004. Gene expression profiling of 1081 renal medullary carcinoma: potential clinical relevance. Cancer 100:976–985. 1082 doi:10.1002/cncr.20049 [doi] 1083 Yin D, Zhou H, Kumagai T, Liu G, Ong JM, Black KL, Koeffler HP. 2005. Proteasome inhibitor 1084 PS-341 causes cell growth arrest and apoptosis in human glioblastoma multiforme 1085 (GBM). Oncogene 24:344–354. doi:10.1038/sj.onc.1208225 1086

1087 1088 Acknowledgments: We thank our patients, the patient advocacy foundations and the RMC

1089 Alliance. We thank John Daley at the Dana-Farber Cancer Institute Flow Cytometry Core Facility,

1090 Anita Hawkins at the Brigham and Women’s Hospital CytoGenomics Core, Zach Herbert at the

1091 Molecular Biology Core Facilities, Tamara Mason at the Genomic Platform at the Broad Institute,

1092 and the Boston University Microarray Resource. We thank the Hahn lab, Boehm lab, Cichowski

1093 lab, Kim Stegmaier, Rosalind Segal, David Kwiatkowski, Seth Alper, Toni Choueiri, Carlos

1094 Rodriguez-Galindo, the COG Renal Tumors Committee for critical discussions and/or reading of

1095 the manuscript. This work was supported in part by the NCI Cancer Target Discovery and

1096 Development Network U01 CA176058 (WCH) and U01 CA217848 (SLS), Katie Moore

1097 Foundation (JSB), Merkin Family Foundation (JSB), T32 GM007753 (TPH), T32 GM007226

1098 (TPH), AACR-Conquer Cancer Foundation of ASCO Young Investigator Translational Cancer

1099 Research Award (ALH), CureSearch for Children’s Cancer (ALH), NCI T32 CA136432-09

1100 (ALH), NICHD K12 HD052896-08 (ALH), Pedals for Pediatrics (ALH), Friends for Dana Farber

1101 (ALH), Alex’s Lemonade Stand Young Investigator Grant (ALH), Boston Children’s Hospital

1102 OFD BTREC CDA (ALH), NCI P50CA101942 (ALH), Cure AT/RT (SNC/ALH), Team Path to

1103 Cure (ALH) and the Wong Family Award (ALH).

1104

1105 Author contributions: ALH and WCH designed the study. ALH and WCH wrote the manuscript.

1106 ALH, YYT, BDK, WJK, PK AS and RD created the cell line models of RMC. ALH, BDK and

1107 MBD performed the pooled RNAi and CRISPR-Cas9 screens. ALH and XY sequenced/analyzed

1108 the RNAi and CRISPR-Cas9 screens. YYT performed compound assays and YYT and ALH

1109 analyzed the results. CC serves as the patient advocacy foundation representative. ALH, GJS, BDK

1110 and WJK performed the glycerol gradients. ALH, JW, MG and GK performed the computational

1111 analyses. ALH, BDK, MBD, WJK, TH and JL performed the validation studies. ALH, MT, KL

1112 and PG performed the mouse studies. ALH and AT prepared the sequencing libraries and managed

1113 the sequencing effort. AC reviewed the histopathology and the immunohistochemistry. JSB, BDC,

1114 KAJ, KLL, BM, BVH and DS developed the IRB approved protocols. AW and CC managed the

1115 tissue biorepository. ALH, OG, PB, SNC, EAM, DR, SLS, PAC, CWMR, CK, RB, KLL, JSB and

1116 WCH supervised the studies. All authors discussed the results and implications and edited the

1117 manuscript.

1118

1119 Competing interests: The authors declare the following competing financial interests: P.B. and

1120 R.B. are consultants for Novartis (Cambridge, MA). C.K. is a Scientific Founder, member of the

1121 Board of Director, Scientific Advisory Board member, Shareholder, and Consultant for Foghorn

1122 Therapeutics, Inc. (Cambridge, MA). Disclosure information for C.K. is also found at:

1123 http://www.kadochlab.org. W.C.H. is a consultant for Thermo Fisher, Aju IB, MPM Capital and

1124 Paraxel. W.C.H. is a founder and shareholder and serves on the scientific advisory board of KSQ

1125 Therapeutics.

1126

1127 Data and materials availability: All primary data are available from the authors. Noted plasmids

1128 in the text are available through Addgene or the Genomics Perturbations Platform at the Broad

1129 Institute of Harvard and MIT. CLF_PEDS0005_T1, CLF_PEDS0005_T2B,

1130 CLF_PEDS0005_T2A and CLF_PEDS9001_T1 cell lines are available through the Cancer Cell

1131 Line Factory at the Broad Institute of Harvard and MIT.

1132

1133 Accession codes: Sequencing data reported in this paper (whole-genome sequencing and whole-

1134 exome sequencing) has been deposited in the database of Genotypes and Phenotypes (dbGaP)

1135 under study accession __pending____ and GEO GSE111787.

Figure 1 a Chromosome # Mut/

Chromosome12 # 345678910 141822mb

CLF_PEDS0005_NCLF−PEDS005N 1.3

CLF_PEDS0005_T1CLF−PEDS005T 2.4

CLF_PEDS0005_T2ACLF−PEDS005M Adherent 2.4

CLF_PEDS0005_T2BCLF−PEDS005M Suspension 2.4

CLF_PEDS9001_T1CLF−RCRF1009T 2.8 b CLF_PEDS0005_T1 CLF_PEDS9001_T1

Y Y

X X 1 1 22 22 21 21

20 20 2 2 19 19

18 18 17 * 3 17 3 16 16 15 4 15 * 4 14 14 5 5

13 13

6 6 12 12

7 7 11 11

8 8

10 10

9 9 c e CLF_PEDS0005_T1 CLF_PEDS9001_T1

1.0 1.0 0.6 0.6 SMARCB1 SMARCB1 0.2 0.2

206-08 20 30 (position in mb) 40 172023 30 (position in MB) 40 Tumor/Normal Ratio Chr 1 Chr 22 Tumor/Normal Ratio Chr 12 Chr 22 heterozygous deletion translocation heterozygous deletion translocation d CLF_PEDS0005_T1f CLF_PEDS9001_T1 Allele 1: heterozygous deletion Allele 1: heterozygous deletion chr22 chr22 chr22 chr22

23,525,723 36,748,243 22,697,439 28,474,405 BCR MYH9 TTC28 VPREB Allele 2: translocation Allele 2: translocation

chr22:24,133,357 chr 1:207,193,552 chr22:24,162,929 chr 12:19,425,709

, intron 10 intron ,

SMARCB1, intron 1 C1orf116, C-terminal SMARCB1, intron 6 PLEKHA5 , intron 10 intron ,

C1orf116, C-terminal SMARCB1, intron 1 PLEKHA5 SMARCB1, intron 6

chr1:207,193,551 chr22:24,133,358 chr12:19,425,710 chr22:24,162,930 bioRxiv preprint doi: https://doi.org/10.1101/487579; this version posted December 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1136 Figure legends: 1137 1138 Fig. 1: Patient derived models of RMC have quiet genomes and are driven by intronic

1139 translocations.

1140 (a) Copy number analysis of RMC models identifies heterozygous loss of alleles or low-level gains

1141 in primary and metastatic tumors. Rates of mutations per megabase are consistent with patients

1142 with RMC or other pediatric cancers such as rhabdoid tumor.

1143 (b) Circos plots of RMC cell lines. Red indicates a deletion. All deletions are located in the introns.

1144 Blue arcs indicate fusions identified with SvABA v0.2.1(Wala et al., 2018). Blue star indicates a

1145 SMARCB1 rearrangement.

1146 (c) Read counts from PCR-free Whole Genome Sequencing identify single copy deletion of

1147 SMARCB1 in CLF_PEDS0005_T1.

1148 (d) Second allele of SMARCB1 is lost by a balanced translocation occurring in intron 1 of

1149 SMARCB1 and fuses to the C-terminal end of C1orf116 in chromosome 1 in CLF_PEDS0005_T1.

1150 (e) Read counts from PCR-free Whole Genome Sequencing identify single copy deletion of

1151 SMARCB1 in CLF_PEDS9001_T1.

1152 (f) Second allele of SMARCB1 is lost by a balanced translocation occurring in intron 6 of

1153 SMARCB1 and fuses to the anti-sense intron 10 of PLEKHA5 in chromosome 12 in

1154 CLF_PEDS9001_T1.

1155

1156

1157

Figure 2 a b

RMC cell lines ATRT Matched normal PEDS0005_N MRT TARGET normalNormal MRT cell line TARGETMRT MRT PEDS0005_T2A RMC TARGETWT WT PEDS9001_T1 RMC cell line TARGETCCSK CCSK SS 2nd tSNE component 2nd tSNE component

1st tSNE component 1st tSNE component c ARID1A SMARCA4 1 5 10 15 20 1 5 10 15 20 G401 HA1E CLF_PEDS0005_T1 + SMARCB1 CLF_PEDS9001_T1 +SMARCB1

deFraction 14 LacZ SMARCB1 SMARCB1 - + (3) HA1E (3) (4) G401 (4) (5) CLF_PEDS0005_T1 (5) (4) G401 0005_T2A 0005_T2B 9001_T1 HA1E 0005_T2A 0005_T2B 9001_T1 CLF_PEDS0005_T2A doxycycline +++ (7) (5) ARID1A CLF_PEDS0005_T2B (8) (3) SMARCA4 CLF_PEDS9001_T1 (3) SMARCC1 SMARCC2 f No SMARCB1 SMARCB1 CLF_PEDS 0005_T2A bioRxiv preprint doi: https://doi.org/10.1101/487579; this version posted December 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1158 Fig. 2: Patient derived models of RMC are functionally similar to SMARCB1 deficient cancers.

1159 (a) Utilizing tSNE, gene expression by RNA-sequencing from patients with pediatric renal tumors

1160 profiled in TARGET such as clear cell sarcoma of the kidney (CCSK), malignant rhabdoid tumor

1161 (MRT) or Wilms tumor (WT) was compared with patient derived models of RMC. RMC cell lines

1162 (red) cluster with rhabdoid tumor samples (purple). The normal cell line (orange)

1163 (CLF_PEDS0005_N) clusters with other normal kidneys profiled in TARGET.

1164 (b) tSNE analysis of gene-expression array data shows RMC cell lines (red) clustering with RMC

1165 patients (blue) and these cluster with other MRT or ATRT samples.

1166 (c) Glycerol gradients (10-30%) followed by SDS PAGE analysis of rhabdoid tumor cell line,

1167 G401, as compared to SMARCB1 wild type cell line HA1E, shows ARID1A is seen in higher

1168 fractions when SMARCB1 is expressed. Gradients were performed on patient derived models of

1169 RMC with doxycycline-inducible SMARCB1. A similar rightward shift of ARID1A occurs upon

1170 re-expression of SMARCB1. These same shifts occur with SMARCA4. These experiments are

1171 representative of at least 2 biological replicates.

1172 (d) Fraction 14 of the glycerol gradients shows a modest increase in SWI/SNF complex members,

1173 SMARCC1, SMARCC2, SMARCA4 and ARID1A in HA1E, a SMARCB1 wild type cell line.

1174 When SMARCB1 is re-expressed in G401 and RMC lines, a similar pattern is seen. Images are

1175 representative of 2 biological replicates.

1176 (e) Using cell lines with stably transfected inducible SMARCB1, cell viability was assessed with

1177 or without expression of SMARCB1 over 8 days. There is no significant difference in SMARCB1

1178 wild type cell line, HA1E. Re-expression of SMARCB1 leads to significant decreases in cell

1179 viability as compared to LacZ control in SMARCB1 deficient cancer cell lines G401,

1180 CLF_PEDS9001_T, and CLF_PEDS0005. Error bars are standard deviations based on number of

1181 samples in parentheses.

1182 (f) CLF_PEDS0005_T2A cell line show signs of senescence following re-expression of

1183 SMARCB1. Images representative of 3 biological replicates.

1184

1185

Figure 3 a CLF_PEDS0005_T2A 19 overlap CLF_PEDS0005_T2B Ubiquitin- CUL1 shRNA CRISPR-Cas9 proteasome PSMB1 shRNA CRISPR-Cas9 (444) (445) system PSMB2 (444) (445) PSMB5 PSMD1 PSMD2 Cell Cycle CDK1 CDK6 72 genes 124 genes 74 genes 136 genes 21 KIF11 27 overlap PLK1 overlap Nuclear KPNB1 export XPO1

75 Other ARF1 82 compounds ATP1A compounds BIRC5 BRD4 MTOR Small molecules RRM1 Small molecules (417) TOP2A (410) b c CLF_PEDS015_T1 CLF_PEDS0005_T1 7.1e-87.1e 8 p valuevalue 1.1e-51.1e-5 5.4e-205.4e-20 robust Z-score CLF-PEDS0005_N

CLF-PEDS0005_T2 robust Z-score d e

f g

log [bortezomib, nM] log [MLN2238, nM] G401 CLF_PEDS9001_T1 bioRxiv preprint doi: https://doi.org/10.1101/487579; this version posted December 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1186 Fig. 3: Functional genomic screens identify inhibition of the proteasome as a vulnerability in

1187 RMC.

1188 (a) Left: RNAi suppression of 444 evaluable genes (red) identifies 72 genes that when suppressed

1189 caused a significant viability loss in CLF_PEDS0005_T2A. Genomic indels created by CRISPR-

1190 Cas9 in 445 evaluable genes (blue) identify 124 genes that cause a viability loss. RNAi and

1191 CRISPR-Cas9 screens were performed in biological replicates. Small-molecule screen (performed

1192 in technical replicates) with 417 evaluable compounds (green) identifies 75 compounds that lead

1193 to significant viability loss. Twenty one genes overlap across these three screens. Right: The same

1194 screens were performed with CLF_PEDS0005_T2B and 27 genes were found to be significantly

1195 depleted when suppressed by RNAi, genomically deleted by CRISPR-Cas9 or inhibited when

1196 treated with a small molecule. Center: List of 19 genes that overlapped between functional screens

1197 from CLF_PEDS0005_T2A and CLF_PEDS0005_T2B can be categorized into genes involving

1198 the ubiquitin-proteasome system, cell cycle and nuclear export.

1199 (b) Comparison of Z-score normalized small-molecule screens between CLF_PEDS0005_T2 and

1200 CLF_PEDS0005_N (normal isogenic cell line). Hits (blue) and proteasome inhibitors (red). Each

1201 dot is representative of the average of two technical replicates.

1202 (c) Relative log2 fold change in abundance from CRISPR-Cas9 screens between sgRNA controls

1203 (grey) and genes in the DCT v1.0 screen involving the proteasome (red). Data is taken at 23 days

1204 following selection and compared to an early time point. As compared to the undifferentiated

1205 sarcoma cell line CLF_PEDS015T, inhibition of the proteasome subunits leads to a more profound

1206 viability loss as compared with controls. Each dot is representative of a minimum of 2 biological

1207 replicates.

1208 (d) Short term cultures of the normal cell line (CLF_PEDS0005_N) or early passage of the

1209 heterogenous cell line (CLF_PEDS9001_early) were compared for assessment of viability to the

1210 primary tumor cell lines following treatment with bortezomib. Two-tailed t-test p-value = 0.008

1211 for PEDS0005 and two-tailed t-test p-value = 4.76e-5 for PEDS9001. Error bars represent standard

1212 deviations from two biological replicates.

1213 (e) Short term cultures of the normal cell line (CLF_PEDS0005_N) or early passage of the

1214 heterogenous cell line (CLF_PEDS9001_early) were compared for assessment of viability to the

1215 primary tumor cell lines following treatment with MLN2238. Error bars represent standard

1216 deviations from two biological replicates.

1217 (f) Re-expression of SMARCB1 in G401 leads to a rightward shift in the dose-response curve with

1218 bortezomib compared with uninduced cells. Error bars represent standard deviations from two

1219 biological replicates.

1220 (g) Re-expression of SMARCB1 in CLF_PEDS9001_T leads to a rightward shift in the dose-

1221 response curve with MLN2238 compared with uninduced cells. Error bars represent standard

1222 deviations from three biological replicates.

1223

1224

1225

Figure 4 a b c MLN2238 z-scores

CCLE Multiple SMARCB1 cell lines myeloma deficient cell lines cell lines d e f

● CLF_PEDS9001_T1 CLF_PEDS0005_T2B

1 ●● ●● ● ●● ● ● ●●●●● ●● ● 27.0±1.6 ● ● ●●●●●● ● ● ●● ● ●●●●● ● 53.2±0.3 DMSO DMSO ● ● ● ● 24.2±1.9 0 ● ● ●● ● ● ● ● ● ●● 16.4±0.3 MLN2238 ●● ● ●● ● ● ● 29.2±2.5 ● ● 8.3±0.7 ● ● ● −1 ●● ● ● ● ● ● ● ● ● 28.4±2.6 30.9±0.4 DMSO doxycycline −2 ● GO cell cycle ● ● 32.2±0.4 +SMARCB1 ● GO regulation of 41.4±1.4 ● MLN2238 proteasomal−ubiquitin LFC no doxycycline vs 24.6±2.3 MLN2238 −3 dependent protein 26.7±2.3 process +SMARCB1 LFC (No SMARCB1 - SMARCB1) −4 −2 0 2 4 0 50k 100k 150k 0 50k 100k 150k LFCLFC (MLN2238DMSO vs MLN2238- DMSO) PI: Area PI: Area ghDMSO MLN2238 6 6 10 10 SMARCB1 -+- + Q1 Q2 Q1 Q2 bortezomib -++ - 5 2.3±0.8 8.0±0.7 5 18.4±7.1 48.7±4.2 10 10 Cleaved 4 4 Caspase 3 10 10 3 3 GAPDH 10 10 2 2 10 10 SMARCB1 -+- + FITC: Annexin V FITC: 1 Annexin V FITC: 1 MLN2238 -++ - 10 Q4 Q3 10 Q4 Q3 Cleaved 0 89.3±1.1 0.5±0.3 0 22.4±7.0 10.8±4.4 10 10 Caspase 3 10 0 10 1 10 2 10 3 10 4 10 5 10 6 10 0 10 1 10 2 10 3 10 4 10 5 10 6 PI: Area PI: Area -tubulin bioRxiv preprint doi: https://doi.org/10.1101/487579; this version posted December 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1226 Fig. 4: Proteasome inhibitors such as MLN2238 are specific to SMARCB1 deficient cancers and

1227 lead to cell cycle arrest and programmed cell death.

1228 (a) Multiple myeloma cell lines and SMARCB1 deficient lines are similarly sensitive to

1229 proteasome inhibitor, MLN2238. Both are significantly different from other CCLE cell lines based

1230 on Z-score normalized sensitivity. * two-tailed t-test p-value < 0.05. n.d. no difference.

1231 (b) Early passage Wilms tumor (SMARCB1 wild type) cell line CLF_PEDS1012_T1 is not as

1232 sensitive to treatment with bortezomib compared with RMC and MRT cell lines. Error bars

1233 represent standard deviations following at least two biological replicates.

1234 (c) Early passage Wilms tumor (SMARCB1 wild type) cell line CLF_PEDS1012_T1 is not as

1235 sensitive to treatment with MLN2238 compared with RMC and MRT cell lines. Error bars

1236 represent standard deviations following at least two biological replicates.

1237 (d) Analysis of differential genes when SMARCB1 was re-expressed in SMARCB1 deficient

1238 cancers compared with differential genes when SMARCB1 deficient cancers were treated with

1239 MLN2238. Gene sets enriched based on GO based GSEA involved the cell cycle (blue) and

1240 regulation of the ubiquitin-proteasome system (black).

1241 (e) Treatment with proteasome inhibitor, MLN2238 for 24 hours leads to G2/M arrest in

1242 CLF_PEDS9001_T1. Values shown represent the percent of cells in G1 or G2/M. Error values

1243 shown are standard deviations from two biological replicates.

1244 (f) Treatment with MLN2238 for 24 hours leads to G2/M arrest in CLF_PEDS0005_T2B which

1245 can be prevented by re-expression of SMARCB1. Error values shown are standard deviations from

1246 two biological replicates.

1247 (g) Treatment of CLF_PEDS9001_T1 with MLN2238 for 48 hours leads to increased frequency

1248 of cells with Annexin V / PI staining and PI only staining. Error values shown are standard

1249 deviations from two biological replicates.

1250 (h) G401 cells stably infected with inducible SMARCB1 treated with either bortezomib or

1251 MLN2238 induce cleaved caspase 3 compared with DMSO controls after 24 hours. When

1252 SMARCB1 is re-expressed, cleaved caspase 3 levels are decreased compared to uninduced cell

1253 lines. Blots are representative of a minimum of 2 biological replicates.

1254

1255

1256

Figure 5 a d 12 PEDS0005 PEDS9001 _T1 _T1 9 SMARCB1 -+- + -+- + MLN2238 -++ - -++ - 6 cyclin B1

UBE2C SMARCB1 cyclin D1 3 −log10(p−value) -tubulin 0 −0.4−0.2 0.0 0.2 0.4 0.6 Effect size b e 1.00 ●

0.75 ● ● ●● ●● 0.50 ● ●● ●●● ● ● ● ●●●● ●●●●●●●●● ●●●●●●● ●●●●●●● ●●●● ● ●●●●●●● ●● ●● ●●●●● ●●●●●● 0.25 ●●●●●●●●●●●● ● ●●●●●●●●●● ● ●●●●●●●●●●●●● ●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●● UBE2C Scaled Rank ●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●● ● ● ●●●●●●●●●●●●●●●●●●●●●●● ● ●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●● 0.00 ● −1.00−0.75−0.50−0.25 0.00 UBE2C Dependency Score c f bioRxiv preprint doi: https://doi.org/10.1101/487579; this version posted December 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1257 Fig. 5: Cell cycle arrest secondary to proteasome inhibition acts via UBE2C/cyclin B1

1258 dysregulation and proteasome inhibition with MLN2238 is effective in vivo.

1259 (a) 204 genes were significantly upregulated when comparing the log2 fold change between

1260 SMARCB1 deficient cells to SMARCB1 re-expressed cells (Supp Data 6). We assessed how loss

1261 of these genes affected viability in 3 cell lines with loss of SMARCB1 as compared to the rest of

1262 the 482 cell lines in Project Achilles DepMap 18Q3, a genome-wide loss of function screen using

1263 CRISPR-Cas9 and calculating an effect size (e.g. differential of the 3 cell lines to 482 cell lines).

1264 A negative effect size identifies genes when deleted are required for cells for survival and the 204

1265 genes are identified in red. Deletion of UBE2C was significantly depleted. Deletion of SMARCB1

1266 serves as a positive control in these SMARCB1 deficient cancers as these cell lines have loss of

1267 SMARCB1.

1268 (b) SMARCB1 deficient lines are in the top 5% of cell lines ranked by how dependent they are on

1269 UBE2C based on Project Achilles DepMap 18Q3 dataset. Three ATRT cancer cell lines (red dots;

1270 CHLA266, CHLA06, COGAR359) were compared to 485 cell lines profiled in Project Achilles

1271 (CERES dataset 18Q3).

1272 (c) Gene deletion of UBE2C by CRISPR-Cas9 leads to significant viability defects in RMC and

1273 MRT cell lines as compared to either SWI/SNF wt cell line, JMSU1 (day 10), or SMARCA2 mutant

1274 cell line, A549 (day 6). Error bars shown are standard deviations from two biological replicates. *

1275 indicates a two-tailed t-test p-value < 0.05 and ** <0.005.

1276 (d) Treatment with proteasome inhibitor, MLN2238, leads to upregulation of cyclin B1 and this

1277 phenotype is rescued upon SMARCB1 re-expression in both CLF_PEDS0005_T1 and

1278 CLF_PEDS9001_T1. Cyclin D1 is included as a control to ensure that the effects of the proteasome

1279 are specific to cyclin B1. Blots are representative of two biological replicates.

1280 (e) G401 xenograft tumor growth over time shows that treatment effects from MLN2238 can be

1281 seen as early as 8 days from treatment initiation as compared to vehicle control. Dotted line reflects

1282 average volume of 148mm3. Error bars are reflective of standard error of the mean. * indicates

1283 Mantel-Cox p-value <0.05.

1284 (f) Kaplan-Meier curves from mice with G401 xenograft tumors treated with either vehicle or

1285 MLN2238.

1286

Supp Figure 1 ab CLF_ CLF_ CLF_ PEDS0005 CLF_ PEDS0005 PEDS0005 Tumor Tissue PEDS0005_T1 _T2A _T2B

H&E

CAM5.2

tumor

SMARCB1 cm 20 m c d CLF_PEDS0005 Ex1-2 1.0 Germline Ex2-3 0.8 Ex3-4 Ex4-5 0.6 CLF_ Ex5-6 PEDS0005_T1 0.4 Ex6-7 0.2 Ex7-8 Relative Quantity CLF_PEDS9001 0.0 Germline TC32 PEDS PEDS 0005_T1 9001_T1

CLF_ PEDS9001_T1

Hb e G401 0%100% 90% 100% g 0005_T1 CLF_PEDS0005 primary tumor 500bp- chr1 chr22 9001_T1 500bp-

% RMC cell line 100 60 40 10 0% 10 0.4 0.02 0% f CLF_PEDS9001 primary tumor chr12 chr22

500bp- G401 0005_T1 0005 primary G401 9001_T1 9001 primary bioRxiv preprint doi: https://doi.org/10.1101/487579; this version posted December 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1287 Supp Fig.1: Patient derived RMC models are reflective of known biology and are faithful

1288 models of the primary tumors.

1289 (a) Gross pathology of CLF_PEDS0005 nephrectomy. Circle highlights treated tumor.

1290 (b) Immunohistochemistry of CLF_PEDS0005 cells highlighting same staining patterns as seen in

1291 RMC.

1292 (c) Integrated Genomic Viewer (IGV) screen shot of the hemoglobin beta loci. At amino acid 7,

1293 there is a T/A heterozygous mutation (Glutamic acid/Valine) leading to sickle cell trait.

1294 (d) Quantitative reverse transcriptase PCR (qRT-PCR) of each exon-exon junction of SMARCB1

1295 confirms expression loss in RMC samples. Relative expression of these exon-exon junctions in the

1296 RMC models was compared to TC32, a Ewing Sarcoma cell line with wild-type SMARCB1.

1297 (e) Mixing studies identify that PCRs are specific to detect genomic fusions between 0.4-2% tumor

1298 purity. PCRs spanning the genomic fusion in SMARCB1 and partner loci were amplified and

1299 resolved on 2% agarose gels. Samples were mixed with G401, which has biallelic deletion of

1300 SMARCB1, to determine the tumor purity required to identify the fusion sequence.

1301 (f) PCR amplification of primary tumor samples which were <20% pure identifies SMARCB1

1302 fusions as identified on PCR-free WGS.

1303 (g) Sanger sequencing was performed from gel purified products in (F) and aligned with PCR-free

1304 WGS sequence.

1305

1306

1307

Supp Figure 2 aba

c 0005_ 0005_ T2A T2B G401 HA1E Doxycycline -+ -++ -+ -+ d SMARCB1 SMARCB1 -actin Fraction # 1 5 1015 20 0005 0005 0005 9001 G401 T1 T2A T2B T1 HA1E Doxycycline -+-+-+-+ SMARCB1 CLF_PEDS0005_T1 -actin + SMARCB1 e CLF_PEDS9001_T1 pLXI403 LacZ +SMARCB1 no dox dox

f 0005 0005_T2A 0005_T2B G401 _T1 Doxycycline -+-+-+-+ -+-+

0005_T2B LacZ -+---+------SMARCB1 ---+---+ -+-+ SMARCB1 TP21 SPARC

-actin

0005_T1 0005_T2A g 9001 0005_T1 0005_T2A 0005_T2B _T1 pLXI403 SMARCB1 no dox dox Doxycycline -+-+-+-+-+-+ -+ LacZ -+---+---+-- -- SMARCB1 ---+---+---+ -+ c-MYC

0005_T1 -actin 0005_T2B bioRxiv preprint doi: https://doi.org/10.1101/487579; this version posted December 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1308 Supp Fig.2: Re-expression of SMARCB1 in patient derived RMC models.

1309 (a) CellTiter-Glo of parental cell lines treated with increasing doses of doxycycline for 96 hours.

1310 Error bars shown are standard deviations from two biological replicates. Dotted line at doxycycline

1311 of 1ug/mL (or log 0) indicates dosage used in this study.

1312 (b) Effect of continuous doxycycline of 1ug/mL over 8 to 9 days. Cells were plated on day 0 and

1313 treated with 1ug/mL of doxycycline. Cells were counted at day 4 or 5 and then replated at the same

1314 cell number and treated with doxycycline. Error bars shown are standard deviations from two

1315 biological replicates.

1316 (c) SMARCB1 inducible vectors were stably transfected into cell lines. Administration of

1317 doxycycline caused similar induction of SMARCB1 across all cell lines. Blots are representative

1318 of at least two biological replicates.

1319 (d) SMARCB1 immunoblot from 10-30% glycerol gradients. SMARCB1 is prominently in

1320 fractions 13 and 14. Blots are representative of at least two biological replicates.

1321 (e) Representative images from three biological replicates of β-galactosidase staining following 7

1322 days of SMARCB1 or LacZ re-expression across CLF_PEDS0005 cell lines.

1323 (f) Utilizing stably infected inducible cell lines with either LacZ control or SMARCB1,

1324 immunoblots performed to confirm transcriptome findings. Following SMARCB1 expression,

1325 genes such as TP21 and SPARC were enriched. Blots are representative of at least two biological

1326 replicates.

1327 (g) Utilizing stably infected inducible cell lines with either LacZ control or SMARCB1, c-Myc

1328 protein expression is decreased following SMARCB1 expression. Blots are representative of at

1329 least two biological replicates.

1330

Supp Figure 3 aba Day 6 PEDS0005_T2A c Day 6 PEDS0005_T2B

d Day 23 PEDS0005_T2A e Day 23 PEDS0005_T2B f PEDS0005_T2A

g PEDS0005_T2B h i PEDS0005 PEDS9001 PEDS0005 PEDS9001 _T2A _T1 _T2A _T1 sgLuciferase sg PSMB5 _1 sg PSMB5 _2 sgLuciferase sg PSMB5 _1 sg PSMB5 _2 PSMB5 -tubulin j PEDS0005_N p4 PEDS0005_T1 p27 PEDS9001_T1 p5 (early) PEDS9001_T1 p31 PEDS9001_T1 p6 PEDS1012_T1 SMARCB1

-actin bioRxiv preprint doi: https://doi.org/10.1101/487579; this version posted December 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1331

1332 Supp Fig.3: Functional genomic screens identify proteasome inhibition as a vulnerability in RMC

1333 and other SMARCB1 deficient cancers.

1334 (a) Suppression of RPS6 in CLF_PEDS0005_T2A in RNAi screens. Following log2 normalized

1335 counts, shRNAs (black) and paired seed controls (grey) were assessed for off-target effects. Most

1336 pairs showed minimal off-target effects while Pair 4 showed significant off-target effects. Error

1337 bars represent standard deviation from at least 2 replicates. *** indicates a two-tailed t-test p-value

1338 <0.0005, ** <0.005.

1339 (b) and (c) Correlation of replicates from normalized counts of CRISPR-Cas9 screens in

1340 CLF_PEDS0005_T2A or CLF_PEDS0005_T2B at the early day 6 timepoint.

1341 (d) and (e) Correlation of replicates from normalized counts of CRISPR-Cas9 screens in

1342 CLF_PEDS0005_T2A or CLF_PEDS0005_T2B at the end of the screen (e.g. day 23) timepoint.

1343 (f) and (g) Log2 fold change in abundance of sgRNAs in CRISPR-Cas9 loss of function screens.

1344 Controls include sgControls (black) and common essential gene, RPS6. In comparison to these,

1345 genes involving the proteasome were similarly depleted like RPS6.

1346 (h) Gene deletion by CRISPR-Cas9 of PSMB5 leads to significant decrease in viable cells in RMC

1347 cell lines. Error bars represent standard deviation from at least 2 replicates.

1348 (i) Gene deletion by CRISPR-Cas9 of PSMB5 is confirmed by immunoblot.

1349 (j) Confirmation that the normal cell line, CLF_PEDS0005_N, early passaged tumor cell line,

1350 CLF_PEDS9001_T1 and Wilms tumor cell line, CLF_PEDS1012_T1, express SMARCB1 as

1351 compared to the primary RMC cancer cell lines.

1352

1353 1354

Supp Figure 4 ab

d 0005 0005 0005 9001 RPMI G401 T1 T2A T2B T18226 H2172 DMSO + + + + + + + bortezomib + + + + + + + MLN2238 + + + + + + +

c-MYC

-tubulin e DMSO MLN2238 c-MYC bioRxiv preprint doi: https://doi.org/10.1101/487579; this version posted December 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1355

1356 Supp Fig.4: SMARCB1 deficient cell lines are sensitive to proteasome inhibition.

1357 (a) Using the inducible SMARCB1 cell lines, SMARCB1 deficient lines have increased sensitivity

1358 to bortezomib when compared to cells with SMARCB1 re-expressed. Error bars represent standard

1359 deviation from at least 3 biological replicates. ** indicates a two-tailed t-test p-value < 0.005

1360 (b) Using the inducible SMARCB1 cell lines, SMARCB1 deficient lines have increased sensitivity

1361 to MLN2238 when compared to cells with SMARCB1 re-expressed. Error bars represent standard

1362 deviation from at least 3 biological replicates. * indicates a two-tailed t-test p-value < 0.05 and **

1363 < 0.005

1364 (c) Sensitivity to bortezomib as measured by IC50s. RMC cell lines (red), MRT cell lines (blue)

1365 and ATRT cell lines (yellow) are similarly sensitive to RPMI8226, multiple myeloma cell line.

1366 This is in comparison to H2172 which is significantly less sensitive to proteasome inhibition. Error

1367 bars represent standard deviation from at least 3 biological replicates. n.d. no difference. **

1368 indicates a two-tailed t-test p-value < 0.005.

1369 (d) c-MYC is not downregulated upon proteasome inhibition at the protein level. Cell lines were

1370 treated with DMSO, bortezomib or MLN2238 for 48 hours. c-MYC protein levels in RPMI8226,

1371 a multiple myeloma cell line, decrease following proteasome inhibition by immunoblot. However,

1372 this does not occur in the SMARCB1 deficient cancer cell lines. Immunoblots are representative

1373 of at least 2 biological replicates.

1374 (e) c-MYC is not downregulated upon proteasome inhibition in the transcriptome. Cell lines were

1375 treated with DMSO or MLN2238 for 48 hours. Samples from three biological replicates were

1376 subjected to RNA-sequencing and c-MYC levels were assessed. Across all the SMARCB1

1377 deficient cell lines, c-MYC transcript levels were increased. Error bars represent standard deviation

1378 from at least 3 biological replicates.

1379 1380

Supp Figure 5

a G401 Total Cytoplasmic Nuclear SMARCB1 -+- + -+- + -+- + MLN2238 -++ - -++ - -++ - ARID1A SMARCA4

SMARCC2

SMARCC1 SMARCD1 SMARCE1

SMARCB1

-tubulin

Lamin A/C

b CLF_PEDS9001_T1 Total Cytoplasmic Nuclear SMARCB1 -+- + -+- + -+- + MLN2238 -++ - -++ - -++ - ARID1A SMARCA4

SMARCC2

SMARCC1 SMARCD1 SMARCE1

SMARCB1

-tubulin

Lamin A/C

c G401 CLF_PEDS0005_T2A CLF_PEDS9001_T1 Constant dose Pulse dose Constant dose Pulse dose Constant dose Pulse dose SMARCB1 - - + + - - + + -+- + -+- + -+- + -+- + MLN2238 -++ - -++ - -++ - -++ - -++ - -++ - IRE1 GRP78 -tubulin bioRxiv preprint doi: https://doi.org/10.1101/487579; this version posted December 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1381 Supp Fig.5: Treatment with proteasome inhibitors does not alter total or nuclear protein levels of

1382 SWI/SNF complex members.

1383 (a) and (b) Immunoblots from total, cytoplasmic and nuclear protein fractions show no significant

1384 difference in SWI/SNF complex members upon treatment of proteasome inhibitors. Lamin A/C

1385 and alpha tubulin are loading controls. Blots representative of at least 2 biological repeats.

1386 (c) Treatment with proteasome inhibitor, MLN2238, leads to induction of IRE1α and GRP78.

1387 However, induction of these ER stress proteins is not rescued upon re-expression of SMARCB1.

1388

1389 1390

Supp Figure 6 a b JMSU1 A549 9001_T1 sgLuciferase + + + sgUBE2C guide 1 + + + sgUBE2C guide 2 + + + UBE2C c-MYC

-actin

A549 0005_T2A G401 sgLuciferase + + + sgUBE2C guide 1 + + + sgUBE2C guide 2 + + + UBE2C c-MYC -actin

c d e

** ** ** *** **

f DMSO MLN2238 6 6 10 10 Q1 Q2 Q1 Q2 2.1±0.1 9.3±1.4 5 5 26.7±0.2 46.4±4.3 10 10 4 4 10 10 3 3 10 10 2 2 10 10 FITC: Annexin V FITC: 1 Annexin V FITC: 1 10 Q4 Q3 10 QQ44 Q3Q3 0 88.3±1.4 0.3±0.1 0 18.7±1.4 8.5±2.8 10 10 10 0 10 1 10 2 10 3 10 4 10 5 10 6 10 0 10 1 10 2 10 3 10 4 10 5 10 6 PI: Area PI: Area bioRxiv preprint doi: https://doi.org/10.1101/487579; this version posted December 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1391 Supp Fig.6: Pulse dosing of proteasome inhibitors leads to suppression of the proteasome,

1392 arrests cells in G2/M and leads to programmed cell death.

1393 (a) Suppression of UBE2C protein by immunoblot when utilizing CRISPR-Cas9 guide RNAs

1394 targeting UBE2C. cMYC levels modestly decreased upon UBE2C deletion in SMARCB1 deficient

1395 cell lines. Blots representative of at least 2 biological repeats.

1396 (b) Proteasome inhibitors suppress proteasome activity following one hour of treatment by

1397 assessing the cell’s ability to cleave Suc-LLVY-aminoluciferin following a one-hour treatment

1398 with a proteasome inhibitor as indicated in the figure (Methods). Error bars represent standard

1399 deviation from at least 3 biological replicates.

1400 (c) Pulse treatment with MLN2238 leads to significant viability defects in SMARCB1 deficient

1401 cell lines similar to multiple myeloma cell line, RPMI8226, and contrasts to lung non-small cell

1402 lung cancer cell line, H2172. Error bars are standard deviations from a two biological replicates.

1403 (d) and (e) Cell cycle analysis of G401 (D) or CLF_PEDS9001_T1 (E) treated with a continuous

1404 dose or a pulse dose of MLN2238 shows an increase in cells arrested in G2/M. Error bars represent

1405 standard deviation from at least 2 biological replicates.

1406 (f) Pulse treatment with MLN2238 in CLF_PEDS9001_T1 leads to increased populations that are

1407 Annexin V/PI and PI positive. Figures representative of 3 biological replicates.

1408

Supp Figure 7 a b G401 PEDS0005_T1 PEDS0005_T2A PEDS9001_T1

Doxycycline -+ -++ -+ -+

SMARCB1 Ratio of Annexin V+/PI+ cells Ratio of between MLN2238 and DMSO c d e PEDS0005 PEDS9001 CLF_PEDS9001_T1 _T1 _T1 CLF_PEDS0005_T1 SMARCB1 -+- + -+- + MLN2238 -++ - -++ - cyclin B1

cyclin D1

-tubulin f g bioRxiv preprint doi: https://doi.org/10.1101/487579; this version posted December 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1409 Supp Fig.7: Pulse treatment with MLN2238 can be rescued with SMARCB1 re-expression and

1410 in vivo studies identify MLN2238 leading to tumor shrinkage

1411 (a) Pulse treatment with MLN2238 in CLF_PEDS9001_T1 in the setting of re-expression of

1412 SMARCB1 leads to a decreased fold change in double Annexin V+ / PI+ cells. Error bars represent

1413 standard deviation from at least 2 biological replicates. * two-tailed t-test p-value < 0.05.

1414 (b) Viability defects seen with pulse treatment with MLN2238 can be rescued with re-expression

1415 of SMARCB1 in SMARCB1 deficient cell lines. Error bars shown are standard deviations from

1416 two biological replicates. * indicates a two-tailed t-test p-value < 0.05, ** <0.005, ***<0.0005.

1417 (c) Pulse treatment with proteasome inhibitor, MLN2238, leads to upregulation of cyclin B1 and

1418 this phenotype is rescued upon SMARCB1 re-expression in both CLF_PEDS0005_T1 and

1419 CLF_PEDS9001_T1. Cyclin D1 is included as a control to ensure that the effects of the proteasome

1420 are specific to cyclin B1. Blots are representative of two biological replicates.

1421 (d) and (e) Primary tumor RMC cell lines (CLF_PEDS0005_T1 and CLF_PEDS9001_T) do not

1422 form tumors in vivo. 5 million cells were injected subcutaneously into Taconic immunodeficient

1423 mice and were monitored for tumor formation over 41 to 54 days.

1424 (f) Waterfall plot of relative tumor growth of xenograft G401 tumors at 26 days following

1425 treatment initiation with MLN2238. 7 mice accrued in each group were randomized to receive

1426 vehicle or MLN2238 once tumors reached an average volume of 148mm3. * two-sided t-test p-

1427 value <0.05.

1428 (g) % change in body weight of mice at day 26 as compared to day 1 following treatment with

1429 vehicle or with MLN2238. n.s. not significant based on a two-sided t-test p-value.

1430