bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

1 The CUL5 ubiquitin ligase complex mediates resistance to CDK9 and MCL1

2 inhibitors in lung cancer cells

3

4 Author List & Affiliations

5 Shaheen Kabir1,2,3; Justin Cidado4; Courtney Andersen4; Cortni Dick4; Pei-Chun Lin1,2;

6 Therese Mitros1,2; Hong Ma1,2; Ron Baik1,2; Matthew Belmonte4; Lisa Drew4; Jacob

7 Corn1,2,5*.

8 1Innovative Genomics Institute, University of California, Berkeley, CA 94720

9 2Department of Molecular and Cell Biology, University of California, Berkeley, CA 94720

10 3Helen Diller Family Comprehensive Cancer Center, University of California, San

11 Francisco, CA 94158

12 4Bioscience Oncology, IMED Biotech Unit, AstraZeneca, Waltham, MA 02451

13 5Current address: Department of Biology, ETH Zurich, Switzerland

14 *correspondence: [email protected]

15

16

17

18

19

20

21

22

23

24

25

1 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

26 ABSTRACT

27 Overexpression of anti-apoptotic proteins MCL1 and Bcl-xL is a frequent event in blood

28 and solid cancers. Inhibitors targeting MCL1 are in clinical development, however many

29 cancer models are intrinsically resistant to this approach. To discover mechanisms

30 underlying resistance to MCL1 inhibition, we performed multiple flow-cytometry based

31 genome-wide CRISPR screens that interrogate two drugs directly or indirectly targeting

32 MCL1. Remarkably, both screens identified three components (CUL5, RNF7 and

33 UBE2F) of a cullin-RING ubiquitin ligase complex (CRL5) that resensitized cells to MCL1

34 inhibition. We find that levels of the BH3-only pro-apoptotic proteins Bim and Noxa are

35 proteasomally regulated by the CRL5 complex. Accumulation of Noxa caused by

36 depletion of CRL5 components particularly skewed the balance in favor of apoptosis

37 when cells were challenged with an MCL1 inhibitor. Discovery of a novel role of CRL5 in

38 apoptosis and resistance to MCL1 inhibitors exposes new drug targets and the potential

39 to improve combination treatments.

40

41

42 INTRODUCTION

43 Cancer cells frequently manipulate the intrinsic apoptotic pathway to evade cell death

44 and expand their proliferative capacity. Aberrant increases in levels of anti-apoptotic

45 proteins in the BCL-2 family, such as amplification of MCL1, have been widely implicated

46 in the transformation of cancer cells and the development of resistance to current

47 therapies (Kelly & Strasser, 2011). Members of the BCL-2 family are classified based on

48 the conservation of their BCL-2 homology (BH) domains: multi-domain proteins BAK,

49 BAX and BOK serve as apoptosis executors in the mitochondria; proteins containing

50 only the BH3 domain (BH3-only) promote BAK/BAX activation. Anti-apoptotic proteins

2 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

51 such as MCL1, Bcl-xL and BCL-2 inhibit apoptosis by antagonistic binding to BH3-only

52 pro-apoptotic proteins and BAK and BAX (Czabotar, Lessene, Strasser, & Adams,

53 2014).

54

55 High-resolution investigation of somatic copy number alterations has revealed gene

56 amplification of MCL1 and BCL2L1 (Bcl-xL) are key determinants of survival of breast

57 and non-small cell lung cancer (NSCLC), as depletion of MCL1 and Bcl-xL greatly

58 induces apoptosis in cancer cells (Goodwin, Rossanese, Olejniczak, & Fesik, 2015; Xiao

59 et al., 2015; Zhang et al., 2011). Various models of multiple myeloma, acute myeloid

60 leukemia (AML) and B-cell acute lymphoblastic leukemia (B-ALL) are also dependent on

61 MCL1 expression for survival (Koss et al., 2013). Amplification of MCL1 is also a

62 prognostic indicator for disease severity and progression (Campbell et al., 2018; Yin et

63 al., 2016), making it an attractive therapeutic target.

64

65 In an effort to restrict the action of anti-apoptotic proteins, numerous compounds have

66 been developed that mimic BH3-only proteins (BH3-mimetics). Unfortunately, the first

67 BH3-mimetics that specifically antagonized Bcl-xL were associated with significant

68 thrombocytopenia, thus complicating their therapeutic use (Lessene et al., 2013; Joel D

69 Leverson et al., 2015; Tao et al., 2014). Small-molecule inhibition of MCL1 has recently

70 gained significant attention (Figure 1A), and compounds that selectively target MCL1

71 are currently in clinical trials (Abulwerdi et al., 2014; Burke et al., 2015; Caenepeel et al.,

72 2018; Kotschy et al., 2016; J D Leverson et al., 2015; S64315, n.d.; Tron et al., 2018).

73 Promising reports of direct BH3-mimetic MCL1 inhibitors in preclinical hematological

74 malignancies show potent efficacy with low cytotoxicity (Kotschy et al., 2016; J D

75 Leverson et al., 2015). However, assessment of MCL1 inhibitors in solid breast tumors

3 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

76 showed little single agent activity unless combined with a chemotherapeutic agent

77 (Merino et al., 2017). Co-dosing Bcl-xL and MCL1 inhibitors to achieve effective

78 treatment may be complicated by severe accompanying side effects.

79

80 Beyond direct inhibitors of the BCL2 family of proteins, inhibitors of cyclin-dependent

81 kinase 9 (CDK9) can indirectly target MCL1. CDK9 inhibition restricts transcription

82 elongation thus exploiting the short-lived half-life of MCL1 mRNA and protein (Figure

83 1A) (Gregory et al., 2015; C.-H. Huang et al., 2014; Lemke et al., 2014). Selective CDK9

84 inhibitors have been developed and show improvement in murine models of arthritis and

85 increased survival in murine models of leukemia (Garcia-Cuellar et al., 2014; Hellvard et

86 al., 2016). Nonetheless, these compounds must be precisely and acutely dosed for

87 clinical applications due to their global effects on transcription. Moreover, although CDK9

88 inhibition suppresses MCL1 expression, it does not affect levels of some other anti-

89 apoptotic proteins such as Bcl-xL. This exposes a potential vulnerability in treatment

90 such that numerous cancer types may already have or develop a mode of resistance. In

91 order to truly harness the power of CDK9 inhibitors (CDK9i) or MCL1 inhibitors (MCL1i),

92 it is imperative to uncover additional targets that may sensitize cells to these treatments.

93

94 We reasoned that a genome wide search could uncover new targets that modulate the

95 therapeutic activity of CDK9i and MCL1i and suggest combination therapies to

96 resensitize tumor cells to these anti-cancer drugs. As lung cancer is the leading cause of

97 cancer mortality and most NSCLC patients develop resistance to first-line treatment, we

98 performed genome-wide CRISPR inhibition (CRISPRi) screens in a NSCLC line resistant

99 to both CDK9 and MCL1 inhibition (Siegel, Miller, & Jemal, 2016). We discovered that

100 disruption of the cullin 5-RING ubiquitin ligase (CRL5) complex markedly resensitized

4 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

101 cells to CDK9i or MCL1i. We show that the CRL5 complex causes resistance to loss of

102 MCL1 by targeting pro-apoptotic BH3-only proteins for proteasomal degradation.

103 Members of the CRL5 complex are thus attractive new targets for future combination

104 therapy to treat otherwise resistant NSCLC.

105

106

107 RESULTS

108 MCL1-amplified NSCLC lines resistant to CDK9i and MCL1i have increased Bcl-xL.

109 We assessed a panel of NSCLC lines for their sensitivity to CDK9i (AZD5576) or MCL1i

110 (AZD5991) currently in clinical trials (Figure 1B and Table S1). Several cell lines with

111 amplified MCL1 (copy number ≥ 3) rely on overexpression of MCL1 to escape apoptosis

112 and were highly sensitive to CDK9i and MCL1i. We also found several NSCLC lines that

113 remained resistant to CDK9i and MCL1i despite being MCL1-amplified. We examined

114 the ratios of MCL1 protein to Bcl-xL protein and found increased Bcl-xL expression

115 correlated with the MCL1-amplified lines that were resistant to CDK9i and MCL1i.

116 Conversely, MCL1-amplified cell lines with low levels of Bcl-xL were sensitive to CDK9i

117 and MCL1i.

118

119 Monitoring caspase activation in response to varying doses of MCL1i and CDK9i shows

120 that both drugs induce apoptosis and cell death in a sensitive cell line (H23) after just six

121 hours, with saturated cell death and caspase activation at 1μM. Conversely, a resistant

122 line (LK2) was about 100-fold more resistant to CDK9i and MCL1i at similar drug

123 concentrations (Figure 1C). Cell viability measurements illustrate the intrinsic resistance

124 of LK2s to MCL1i with a decrease in viability observed only at the highest dose tested of

125 30μM, in contrast to H23s that showed a large drop in cell viability at just 0.3μM of

5 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

126 compound. For CDK9i however, a significant decrease in cell viability is seen in both the

127 resistant and sensitive cell lines after extended treatment (24 hours) with 0.1μM

128 compound, highlighting the off-target toxicity of CDK9i at later time points presumably

129 stemming from global transcriptional arrest (Figure 1D). This emphasizes the

130 requirement of a tightly regulated acute treatment window for CDK9i to avoid non-

131 specific cell death at later time points.

132

133 Genome-wide CRISPRi screens identify factors that synergize with MCL1

134 inhibition.

135 Preclinical studies suggest dual inhibition of MCL1 and Bcl-xL may not be a clinically

136 viable option due to the highly toxic side effects. We therefore sought to identify alternate

137 pathways that could resensitize cancer cells to CDK9i or MCL1i. LK2 cells are resistant

138 to CDK9i and MCL1i, tolerate dense growth conditions, proliferate rapidly and transfect

139 and transduce easily, making them an optimal cell line for a functional genomics screen.

140 Because prolonged CDK9i treatment induces non-specific cell death, a typical growth-

141 based screen that measures cell abundance after 14 to 21 days of compound exposure

142 was inappropriate. Instead, we designed a positive selection FACS-based screen for

143 apoptosis that could finish within 6 hours (the optimal treatment window to maximize on-

144 target cell death and minimize non-specific cell death). We designed the screen to

145 expose cells to an acute treatment of CDK9i or MCL1i and then immediately sort

146 apoptosing cells using Cell Event staining, which fluoresces upon cellular caspase

147 activation (Figure 2A).

148

149 We generated a clonal LK2 CRISPRi cell line to ensure uniform expression of dCas9-

150 KRAB and functionally validated knockdown of a panel of genes by qRT-PCR (Figure

6 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

151 S1A). Biological duplicates of LK2 CRISPRi cells were transduced with a genome-wide

152 sgRNA library at a low multiplicity of infection to ensure delivery of one sgRNA per cell,

153 and at 500X coverage to promote equal guide representation. The sgRNA vectors also

154 encode a puromycin resistance cassette allowing for puromycin-selection of successfully

155 transduced cells. Library transduced cells were treated with either 3μM CDK9i for 6

156 hours or 1μM MCL1i for 12 hours and incubated with 0.5μM Cell Event for the duration

157 of the treatment. The concentrations and length of treatments were determined based on

158 where maximal differences in caspase activation were observed between resistant and

159 sensitive cell lines. To control for vehicle effects, we also performed two duplicate

160 genome-wide sorts with DMSO as a treatment condition. We harvested bulk untreated

161 cells on the same day as the FACS sort of DMSO- or drug-treated cells to provide an

162 accurate ‘background’ sampling of sgRNAs and eliminate confounding effects from

163 essential genes that had dropped out of the population. We measured quantitative

164 differences in sgRNA frequency by deep sequencing and integrated enriched or depleted

165 sgRNAs into gene-level hits by comparing sorted samples to the untreated control using

166 ScreenProcessing (Table S2-S4) and MAGeCK (Table S5-S8) analysis pipelines

167 (Horlbeck et al., 2016; Li et al., 2014).

168

169 We analyzed DMSO-treated apoptosing cells to determine if any gene knockdowns

170 promoted apoptosis upon exposure to vehicle. No significant hits in the DMSO control

171 were detected over the random distribution of non-targeting sgRNA controls (Figure

172 S1B), indicating that our gene calls in CDK9i- or MCL1i-treated samples were

173 specifically derived from drug synergy. We found multiple sgRNAs targeting Bcl-xL led to

174 re-sensitization of LK2 cells to CDK9i and MCL1i (Figure 2B-C), confirming reports that

175 depletion of both MCL1 and Bcl-xL initiates apoptosis (Goodwin et al., 2015; Xiao et al.,

7 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

176 2015; Zhang et al., 2011). Downregulation of the mitochondrial porin VDAC2 also

177 promoted apoptosis in drug-treated cells (Figure 2B-C), consistent with its proposed role

178 in BAK/BAX sequestration at the mitochondrial membrane (Cheng, Sheiko, Fisher,

179 Craigen, & Korsmeyer, 2003; Lauterwasser et al., 2016) (Chin et al., 2018). Interestingly,

180 EIF4G2 (DAP5) was identified as a hit in the CDK9i screen (Figure 2B) and has been

181 implicated in stimulating cap-independent translation of anti-apoptotic protein BCL-2,

182 potentially skewing the balance in favor of apoptosis (Marash et al., 2008).

183

184 While CDK9i and MCL1i reduce MCL1 activity through completely separate

185 mechanisms, the top resensitization hits were strikingly consistent between both

186 screens. These hits are furthermore part of two physical complexes, one potentially

187 involved in specialized translation and the other involved in protein degradation. First,

188 knockdown of multiple components of the eukaryotic translation initiation factor 3 (eIF-3)

189 complex (EIF3H and EIF3M) resensitize LK2 cells to CDK9i and MCL1i (Figure 2B-C

190 and Figure S1C-D). eIF-3 is reported to bind a highly specific program of messenger

191 RNAs involved in cell proliferation and apoptosis (Lee, Kranzusch, & Cate, 2015). One

192 might speculate that this complex could be involved in mediating translational activation

193 of certain anti-apoptotic proteins or repression of pro-apoptotic proteins. Regulation of

194 apoptosis by eIF-3 warrants further investigation, but we opted to focus on the second,

195 better-studied complex.

196

197 Knockdown of multiple members of a cullin-RING ubiquitin ligase complex (CRL)

198 including CUL5, UBE2F and RNF7 resensitizes LK2 cells to CDK9i and MCL1i (Figure

199 2B-C and Figure S1C-D). The cullin 5 (CUL5) scaffold was identified as a top

200 resensitizing hit in both screens. Repression of the NEDD8-conjugating activation

8 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

201 subunit UBE2F sensitizes LK2 cells to MCL1i (Figure 2C and Figure S1D). Additionally,

202 knockdown of Ring Finger Protein 7 (RNF7), a protein that binds to the CUL5 scaffold

203 and serves as a catalytic subunit, sensitizes cells to both CDK9i and MCL1i (Figure

204 S1C-D).

205

206 Depletion of the CUL5-RNF7-UBE2F ubiquitin ligase complex resensitizes cancer

207 cells to treatment with CDK9i or MCL1i.

208 Cullin-RING ligases are emerging as attractive cancer targets and a novel class of small

209 molecule neddylation inhibitors have recently been developed (Soucy et al., 2009). While

210 the cullin 3 complex (CRL3) is more typically associated with cancer phenotypes, there

211 is little data on a role for CRL5 in tumorigenesis or resistance. We therefore sought to

212 further examine the role of the CRL5 complex in resensitizing LK2 cells to CDK9 and

213 MCL1 inhibition.

214

215 Cullin-RING ubiquitin ligases (CRLs) comprise the largest known category of ubiquitin

216 ligases and are involved in regulating numerous dynamic cellular processes (Petroski &

217 Deshaies, 2005). They are multi-subunit complexes built upon the cullin scaffold

218 proteins, and CUL5 plays this role in the CRL5 complex (Figure 3A). CUL5 binds to a

219 RING protein (RNF7) that serves as a docking site for E2 ubiquitin ligases. CUL5 binds

220 exclusively to RNF7, indicating that RNF7 is indispensable for CRL5 activity (Kamura et

221 al., 2004). UBE2F, also known as the NEDD8-conjugating enzyme (NCE2) is an E2

222 ubiquitin ligase that neddylates and activates CRL5 (D. T. Huang et al., 2009). Elongin B

223 and C alongside a substrate adaptor bind to the N-terminus of CUL5 and dictate choice

224 of substrates that get polyubiquitinated by E3 ubiquitin ligases (Mahrour et al., 2008;

225 Okumura, Joo-Okumura, Nakatsukasa, & Kamura, 2016). Adaptor proteins add a layer

9 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

226 of regulation to CRL activity both by specifying substrates as well as defining the timing

227 and localization of ubiquitination, however the substrate adaptor regulating CRL5

228 resistance to MCL1 inhibition is not known.

229

230 Interestingly, another hit in the MCL1i screen was the BTB/POZ domain-containing

231 protein 3 (BTBD3; Figure S1D). While BTBD3 has no known function, members of the

232 BTB (also known as POZ) domain containing protein family can act as substrate

233 adaptors for cullin 3 (L. Xu et al., 2003). We speculated that BTBD3 might be a substrate

234 adaptor for CUL5 that specifically mediates resensitization to MCL1i and CDK9i, and

235 thus included it in later validation.

236

237 We made stable dCas9-KRAB LK2 cell lines that expressed an individual CRISPRi

238 sgRNA to knockdown CUL5, RNF7, UBE2F, as well as a non-targeting (NT) control. We

239 confirmed efficient knockdown of CUL5, RNF7 and UBE2F by western blotting (Figure

240 3B). CUL5 and RNF7 appeared to be dependent on each other for stability such that

241 when either protein was knocked down, levels of the corresponding protein were also

242 diminished (Figure 3C). Depletion of UBE2F resulted in the disappearance of a higher

243 molecular weight band corresponding to neddylated CUL5 (Figure 3B-C).

244

245 Using two independent stable LK2 CRISPRi lines targeting CUL5, RNF7, UBE2F or Bcl-

246 xL, we validated that individual knockdowns of these genes followed by treatment with

247 CDK9i or MCL1i induced extensive caspase activation (Figure 3D and Figure S2A).

248 Bcl-xL served as a positive control based on previous reports that dual inhibition of Bcl-

249 xL and MCL1 induces apoptosis (Goodwin et al., 2015; Xiao et al., 2015; Zhang et al.,

250 2011). As an orthogonal readout for apoptosis, we also assessed cleaved PARP (c-

10 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

251 PARP) levels following drug treatment (Figure 3E). No c-PARP was observed in

252 untreated stable knockdowns of CUL5, RNF7 or UBE2F, indicating that the CRL5

253 complex is not intrinsically required for LK2 cell survival. By contrast, knockdown of Bcl-

254 xL alone led to substantial apoptosis (Figure 3E). Depletion of CUL5, RNF7, UBE2F or

255 Bcl-xL when combined with CDK9i or MCL1i significantly induced PARP cleavage

256 (Figure 3E). Hence, loss of CRL5 activity alone does not induce apoptosis, suggesting

257 that it could be less toxic than Bcl-xL inhibition and better suited to co-administration with

258 MCL1 inhibitors.

259

260 To further validate the synergy observed between inhibition of CRL5 and MCL1, we

261 made isogenic knockouts of CUL5 and UBE2F. Purified Cas9 ribonucleoproteins (RNPs)

262 complexed with sgRNA pairs flanking coding exon 1 for CUL5 or UBE2F were

263 nucleofected into LK2s in order to completely disrupt coding potential (Lingeman, Jeans,

264 & Corn, 2017). Nucleofected pools were seeded at single cell densities by FACS sorting

265 and clones were initially screened for loss of protein by western blot and confirmed by

266 Sanger sequencing (Figure S2B-C). We did not attempt to knockout RNF7 because

267 cells did not tolerate extended culturing of CRISPRi-mediated stable knockdowns of

268 RNF7. Knockouts of CUL5 and UBE2F were viable and displayed high levels of

269 apoptosis when challenged with a single concentration of CDK9i or MCL1i (Figure S2D).

270 Dose response curves in the CUL5 and UBE2F knockouts showed significant caspase

271 activity at 1μM of CDK9i or MCL1i, whereas wild type cells showed no response even at

272 10μM doses (Figure 3F). Taken together, multiple lines of evidence indicate that

273 depletion of the CRL5 complex does not on its own induce high levels of apoptosis, but

274 resensitizes cells to CDK9 and MCL1 inhibition.

275

11 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

276 To further establish the mechanistic basis for CRL5 protection from CDK9i and MCL1i,

277 we asked whether depletion of Elongin B, Elongin C, or the potential BTBD3 adapter

278 also synergized with MCL1 inhibition. We performed individual knockdowns of each

279 gene by CRISPRi (Figure S3A). Reduction of Elongin B but not Elongin C induced

280 apoptosis at levels similar to that of CUL5 knockdown (Figure S3B). Elongin B and C

281 have often been described as functioning in a complex (Okumura, Matsuzaki,

282 Nakatsukasa, & Kamura, 2012), but our results suggest a potential separation of function

283 of the two proteins in sensitivity to MCL1 inhibitors. Knockdown of BTBD3 showed a

284 modest increase in apoptosis when cells were treated with MCL1i but did not entirely

285 phenocopy CUL5 or Elongin B knockdown (Figure S3B). This suggests that even if

286 BTBD3 does play a role in the CRL5 complex, it does not appear to be the primary

287 substrate adaptor directing ubiquitination of a substrate(s) that synergizes with MCL1

288 inhibition to induce apoptosis.

289

290 In an attempt to find the CRL5 substrate adaptor responsible for resensitization to MCL1

291 inhibitors, we used CRISPRi to individually knock down a large number of CUL5

292 substrate adaptors annotated in literature and tested them for synergy with MCL1i and

293 CDK9i (Okumura et al., 2016). Two different sgRNAs were tested per gene, but none of

294 the putative substrate adaptors analyzed induced apoptosis following CDK9i or MCL1i

295 treatment (Figure S3C and Table S9). We cannot rule out redundancy between

296 substrate adaptors such that one compensates for loss of another. The relevant CRL5

297 substrate adaptors in the context of resensitization to MCL1 inhibition remain to be

298 identified.

299

12 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

300 The CRL5 Complex Regulates Levels of the BH3-only Apoptotic Sensitizers Bim

301 and Noxa

302 We next searched for the substrate of CRL5 that potentiates resensitization to MCL1

303 inhibition. A literature search for apoptosis-related substrates of CUL5 implicated p53, a

304 master regulator of apoptosis. Several viral proteins hijack the CUL5 complex and serve

305 as substrate adaptors that bind p53 and mediate its degradation, which promotes viral

306 infection (Cai, Knight, Verma, Zald, & Robertson, 2006; Querido et al., 2001; Sato et al.,

307 2009). If the CRL5 complex degrades p53 in LK2 cells, inactivating CRL5 would result in

308 accumulation of p53 thereby promoting apoptosis. We Western blotted for p53 in drug

309 treated wild type and CUL5-knockdown cells but observed no difference in p53 levels

310 (Figure S3D). We also confirmed by western blotting that depletion of CUL5 had no

311 effect on MCL1 and Bcl-xL levels (Figure S3E).

312

313 Since p53 levels were not directly affected by loss of CRL5, we hypothesized that CRL5

314 could be targeting pro-apoptotic proteins for proteasomal degradation. In this case, loss

315 of CRL5 would result in accumulation of pro-apoptotic factors. This accumulation alone is

316 presumably not sufficient to induce apoptosis due to the overexpression of anti-apoptotic

317 proteins MCL1 and Bcl-xL. However, when combined with inhibition of MCL1, the

318 balance may then be skewed in favor of apoptosis. Indeed, Noxa was recently proposed

319 to be a substrate of CRL5, but this molecular activity was not linked to a cellular

320 phenotype (Zhou et al., 2017).

321

322 We used Western blotting to interrogate the levels of all eight BH3-only proteins in

323 individual CRISPRi knockdowns of CUL5, RNF7, UBE2F and a NT control. Strikingly, we

324 observed increased protein levels of only two BH3-only proteins, Noxa and Bim (Figure

13 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

325 4A). To determine whether treatment with CDK9i or MCL1i revealed any additional

326 differences in protein levels, we treated NT or CUL5 knockdown cells with DMSO, CDK9i

327 or MCL1i and immunoblotted for all BH3-only proteins. We found that levels of both

328 proapoptotic proteins Noxa and Bim were higher in cells lacking CUL5, while all other

329 BH3-only proteins remained unchanged (Figure 4B-C). Interestingly, CDK9i alone

330 reduced levels of Noxa and Bim, and this occurred through transcriptional

331 downregulation that is presumably related to the half-lives of these mRNAs (Figure 4D).

332 Knockdown of CUL5 together with CDK9i rescued protein levels of Noxa and Bim but not

333 mRNA levels, consistent with a model in which CRL5-mediated post-translational

334 degradation of Noxa and Bim is the mechanism of resistance to MCL1 inhibition.

335

336 To confirm that CRL5 targets Noxa and Bim for proteasomal degradation, we inhibited

337 proteaosmal and lysosomal function with epoxomicin and folimycin, respectively. No

338 increases in Bim or Noxa were detected following folimycin treatment, indicating that

339 these proteins are not lysosomally degraded (Figure 4E). We found that Bim

340 accumulates in epoxomicin-treated LK2 parental cells at levels equivalent to untreated

341 CUL5-KO cells. No change in Bim accumulation was observed in CUL5-KO cells treated

342 with epoxomicin. These results suggest that Bim is targeted for proteasomal degradation

343 exclusively by CUL5 in LK2 cells, as Bim levels are completely rescued by inhibition of

344 the proteasome in wild type cells but proteasomal inhibition has no effect on Bim levels

345 in CUL5-KO cells. Noxa accumulates in both parental and CUL5-KO cells following

346 epoxomicin treatment (Figure 4E), demonstrating that Noxa is also proteasomally

347 regulated. However, inhibition of the proteasome also rescues Noxa levels in CUL5-KO

348 cells, suggesting that Noxa can be targeted by additional ubiquitin ligase complexes. We

349 were unable to demonstrate direct targeting of Noxa and Bim by CRL5 using ubiquitin

14 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

350 co-immunoprecipitation (Figure S4), leaving open the possibility that these proteins are

351 indirect substrates of the CRL5 ligase complex.

352

353 Noxa is Required to Resensitize CUL5-deficient cells to MCL1 Inhibition

354 Based on the observation that removal of CUL5 alone does not induce apoptosis despite

355 increased levels of pro-apoptotic Bim and Noxa, we reasoned the high expression of

356 MCL1 in these cells was still sufficient to prevent cell death. However, when MCL1 is

357 inhibited in cells lacking CUL5, the accumulated Bim and Noxa may now be able to

358 initiate apoptosis.

359

360 To examine the dependency of Noxa and Bim to resensitize CUL5-deficient cells to

361 CDK9i or MCL1i, we used RNAi to knock down Noxa and/or Bim in the context of CUL5

362 CRISPRi. If induction of apoptosis in CUL5-deficient cells requires Noxa and/or Bim,

363 then removal of either or both factors should limit apoptosis and make the cells re-

364 resistant to inhibition of MCL1. To validate the experimental design, we co-depleted

365 CUL5 and BAK, the apoptosis effector required for cell death. In these conditions,

366 treatment with CDK9i or MCL1i no longer induced apoptosis, consistent with the

367 requirement of BAK in the initiation of apoptosis (Figure 5A-B and Figure S5A-B).

368 Induction of apoptosis remained high in MCL1i-treated cells lacking Noxa and CUL5 or

369 cells lacking Bim and CUL5. Similarly co-depletion of all three proteins Noxa, Bim and

370 CUL5 did not restrict apoptosis upon MCL1i treatment (Figure 5B and Figure S5B).

371 Thus Bim and Noxa are insufficient to induce cell death when CUL5-depleted cells are

372 treated with MCL1i.

373

15 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

374 However, CDK9i treatment of Noxa and CUL5 co-depleted cells led to almost no

375 apoptotic induction, indicating these cells had regained resistance to CDK9i (Figure 5A

376 and Figure S5A). Removal of both Noxa and Bim in CUL5-depleted cells did not lead to

377 an additive effect in the reacquisition of resistance to CDK9i, while cells lacking only Bim

378 and CUL5 were still sensitive to CDK9i. These data indicate that in CDK9i-treated cells

379 lacking CUL5, initiation of apoptosis is primarily dependent on Noxa.

380

381 We wondered whether the discrepancy between MCL1i and CDK9i treatments was

382 caused by the additional depletion of Noxa and Bim due to non-specific effects of CDK9i

383 that we previously observed (Figure 4B-C). However qRT-PCR after knockdown of

384 Noxa and Bim already showed extremely low levels of Noxa and Bim mRNA (Figure

385 S5C), making this an unlikely explanation. Instead, the difference in MCL1i and CDK9i

386 may arise from CDK9i targeting an additional protein involved in promoting apoptosis,

387 leading to overall abrogation of apoptosis. In conclusion, our work reveals that resistance

388 to CDK9i or MCL1i can be overcome by inactivation of the CRL5 complex and we

389 propose this is in part mediated by the accumulation of the CRL5 target Noxa (Figure

390 5C).

391

392

393 DISCUSSION

394 Cullin-RING complexes comprise the largest class of ubiquitin ligases and regulate a

395 diverse array of biological processes. While CUL3 is implicated in a wide array of cancer

396 biologies, relatively little is known about the physiological functions of many CUL5-

397 containing ubiquitin ligases. A few reports have implicated CUL5 in cancer progression.

398 Downregulation of CUL5 has been observed in numerous breast tumors (Fay et al.,

16 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

399 2003) and overexpression of CUL5 in T47D breast cancer cells has been shown to limit

400 cell growth (Burnatowska-Hledin et al., 2004; Lubbers et al., 2011). Additionally,

401 microRNAs 19a/b that negatively regulate CUL5 promote cell growth and invasion in

402 cervical cancer, which is abolished by complementation with miRNA-resistant CUL5 (X.-

403 M. Xu et al., 2012). While alterations in CUL5 are correlated with cancer phenotypes, no

404 clear mechanism has emerged to explain these findings. Our work reveals the CUL5-

405 RNF7-UBE2F ubiquitin ligase as a key modulator of apoptosis and suggests that CUL5

406 may restrict apoptosis by the constitutive degradation of Noxa and/or Bim. Indeed,

407 previous work has shown that the CUL5-RNF-UBE2F complex is capable of in vitro

408 ubiquitination of Noxa (Zhou et al., 2017). While several viral proteins hijack the CUL5

409 complex to degrade p53, a master regulator of apoptosis initiation (Cai et al., 2006;

410 Querido et al., 2001; Sato et al., 2009), we found no evidence however of differential p53

411 levels in DMSO-, CDK9i- or MCL1i-treated wild type and CUL5-KO cells. Thus, we

412 suggest that p53 signaling is not a major mechanism involved in resensitization of cells

413 to MCL1 inhibition.

414

415 Our discovery that inactivation of the CRL5 complex can resensitize cancers to MCL1

416 inhibition provides a therapeutic opportunity, not only in the identification of novel drug

417 targets, but also as a potential biomarker for patient stratification. Based on our data,

418 lung cancer patients that acquire loss of function mutations in either CUL5, RNF7,

419 UBE2F or Elongin B that coincide with high expression of MCL1/Bcl-xL would be

420 candidates for treatment with CDK9i or MCL1i because they are unlikely to have pre-

421 existing CRL5-mediated resistance or to acquire such resistance during treatment. In

422 fact we note that there is significant co-occurrence of MCL1 amplifications with CUL5

423 mutations and UBE2F deletions in NSCLC adenocarcinomas deposited in The Cancer

17 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

424 Genome Atlas (TCGA) (Cerami et al., 2012; Gao et al., 2013). Additionally, as MCL1 and

425 Bcl-xL focal amplifications co-occur frequently in breast cancer, it is possible that the

426 same stratification strategy may apply to a subset of breast cancer patients as well to

427 NSCLC.

428

429 Currently no small-molecule inhibitors exist to target specific components of the CRL5

430 complex. However, several potent inhibitors that broadly inhibit ubiquitin-proteasome

431 systems have been developed. MLN4924 inhibits the NEDD8 activating enzyme (NAE),

432 thereby preventing conjugation of NEDD8 onto all cullin scaffolds, and shows significant

433 anti-tumor activity in clinical trials (Soucy et al., 2009). Alternatively, there are several

434 clinically approved proteasome inhibitors such as bortezomib and carfilzomib currently

435 used in the treatment of multiple myeloma and lymphomas (Richardson et al., 2005)

436 (Kuhn et al., 2007). Although a therapeutic window exists for these compounds, patients

437 also experience adverse side effects due to the widespread effects of proteasome

438 inhibition, making it a complicated option for co-dosing strategies with CDK9i or MCL1i.

439

440 Recently more attention has focused on the development of compounds that directly

441 disrupt protein-protein interactions within specific Cullin-RING ligases, with promising

442 results. One example is the generation of a potent small molecule targeting the von

443 Hippel-Landau (VHL) substrate adaptor (Galdeano et al., 2014). VHL functions in

444 complex with CRL2 to degrade an essential transcription factor HIF-1α, which promotes

445 erythropoietin production. Other broad-spectrum drugs that prevent HIF-1α degradation

446 are in late-stage clinical trials for treatment of chronic anemia associated with kidney

447 disease and chemotherapy (Joharapurkar, Pandya, Patel, Desai, & Jain, 2018). Due to

448 the specialized composition of individual CRL complexes associated with distinct

18 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

449 biological functions, this approach promises a more specific and hopefully less toxic

450 inhibitor. Moreover, our observation that CUL5 and UBE2F knockouts are viable

451 suggests inhibition of these components may be well tolerated making them ideal co-

452 dosing targets. Identification of the substrate adaptor responsible for CRL5-mediated

453 regulation of Noxa/Bim could yield a promising target for MCL1-addicted tumors.

454

455 Innate and acquired resistance to MCL1 inhibition remains a major challenge for lung

456 (and breast) cancer patients, often due to the concomitant amplification of additional

457 anti-apoptotic factors such as Bcl-xL. Targeting multiple anti-apoptotic proteins may limit

458 the therapeutic window due to the accompanying cytotoxicity. Our results reveal an

459 alternate pathway that may be targeted in combination with MCL1 inhibition to combat

460 resistance and hopefully provide a safe and effective therapy.

461

462

463 Author Contributions

464 SK and JEC designed experiments. JC, CA, CD and MB assessed panel of cell lines for

465 MCL1/Bcl-xL status and sensitivity to CDK9i and MCL1i; they also performed all dose-

466 response and viability curves. PCL constructed and validated the initial CRISPRi clonal

467 LK2 cell line. SK performed all the screens and follow-up experiments. TM analyzed

468 screen results with input from SK, JC, JEC and LD. HM assisted with the validation

469 experiments. RB generated the CUL5 and UBE2F knockout clones. SK and JEC wrote

470 the manuscript with input from JC.

471

472

473

19 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

474 Funding Acknowledgements

475 This work was supported by a National Institutes of Health New Innovator Awards

476 (DP2 HL141006), the Li Ka Shing Foundation, the Heritage Medical Research

477 Institute, the California Institute of Regenerative Medicine (DISC1-08776), and

478 funded research support from AstraZeneca. This work used the Vincent J. Coates

479 Genomics Sequencing Laboratory at UC Berkeley, supported by the NIH S10

480 OD018174 Instrumentation Grant.

481

482

483 Conflict of Interest

484 JC, CA, CD and LD are employed by AstraZeneca, from whom funded research support

485 was received.

486

487

488 MATERIALS AND METHODS

489 Cell culture and reagents

490 The LK2 cell line was obtained from AstraZeneca and maintained in RPMI supplemented

491 with 10% (v/v) fetal bovine serum, 100 units/mL penicillin and streptomycin, GlutaMAX,

492 sodium pyruvate and non-essential amino acids. All cells tested negative for

493 Mycoplasma contamination using the MycoAlert Plus Mycoplasma detection kit from

494 Lonza. AstraZeneca licensed the CDK9 inhibitor (AZD5576) originally published as

495 PC585 (Garcia-Cuellar et al., 2014) and supplied it to us for this study. The MCL1

496 inhibitor (AZD5991) is synthesized by AstraZeneca (Tron et al., 2018).

497

498

20 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

499 Lentivirus production and transduction

500 Lentivirus was produced by transfecting HEK293T with standard packaging vectors

501 using the TransIT-LT1 Transfection Reagent (MIR 2306; Mirus Bio LLC). Viral

502 supernatant was collected at 48h and 72 h after transfection, filtered through a 0.45-

503 μm polyethersulfone syringe filter, snap-frozen and stored at −80 °C for future use. For

504 screens, viral titrations were performed by transducing LK2 cells at serial dilutions and

505 assessing BFP (present on the sgRNA-encoding plasmid) percentages 48h following

506 transduction. The viral dilution resulting in ~15% BFP-positive cells was used to

507 transduce cells for the screens.

508

509 Individual analysis of sgRNA phenotypes

510 sgRNA protospacers targeting CUL5, RNF7, UBE2F, Elongin B, Elongin C, Bcl-xL, Noxa

511 and a negative control non-targeting protospacer were individually cloned into BstXI/BlpI-

512 digested pCRISPRia-v2 (Addgene #8432) by ligating annealed complementary synthetic

513 oligonucleotide pairs. The sgRNA expression vectors were packaged into lentivirus as

514 previously described and successful transductants were selected with puromycin at a

515 final concentration of 2.5μg/mL. Sequences for protospacers are in Table S10.

516

517 Genome-scale CRISPRi screens

518 The human CRISPRi v2 library contains 5sgRNAs/annotated TSS of each gene

519 comprising a total of 104,535 sgRNAs. In order to maintain 500X coverage

520 (representation) of each guide at all times, ~50x106 cells need to be transduced with

521 one sgRNA. To ensure delivery of one sgRNA/cell a low MOI of ~0.3 is used, which

522 results in not every cell getting transduced, thus 5-6 fold more cells need to be

523 transduced in order to maintain 500X coverage. Replicate cultures of 3x108 cells were

21 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

524 plated in twenty 15cm dishes and transduced at an MOI of ~0.3 in the presence of

525 4μg/mL polybrene. Cells were split into 2.5μg/mL puromycin and selected for 4 days.

526 Cells were passaged into regular media and recovered for 2 days. 60x106 were

527 collected as the ‘background’ sample. 60x106 cells were treated with 3μM CDK9i and

528 0.5μM Cell Event for 6 hours. 60x106 cells were treated with 1μM MCL1i and 0.5μM

529 Cell Event for 12 hours. 60x106 cells were treated with DMSO (1:10,000X dilution) and

530 0.5μM Cell Event for 12 hours. After treatment, cells were trypsinized, washed and

531 fixed in 1% formaldehyde/PBS. The entirety of the cell population was then FACS-

532 sorted where GFP-positive (apoptosing) cells were isolated. As previously mentioned,

533 this was done in duplicate. Genomic DNA was purified from each cell population with

534 blood purification kits (Machery-Nagel, Nucleospin blood L or XL, depending on cell

535 number) and the sgRNA-encoding region was enriched, amplified, and processed for

536 sequencing on the Illumina Hiseq 2500. TruSeq index sequences unique to each cell

537 population were used to multiplex samples.

538

539 Pooled screen analysis

540 Data analysis was performed as described (Horlbeck et al., 2016; Li et al., 2014).

541 Briefly, sequence reads from the Illumina HiSeq 2500 were trimmed, aligned to

542 CRISPRi v2 sgRNA library, counted and normalized. For the Python-based

543 ScreenProcessing pipeline, sgRNA phenotypes and negative control gene phenotypes

544 were determined along with Mann-Whitney P values. The top three guide-level

545 phenotypes were collapsed to produce the gene-level phenotype score. For genes

546 with multiple annotated transcription start sites (TSSs), phenotypes were calculated

547 for each TSS and the TSS with the lowest Mann Whitney p-value was used to

548 represent the gene. For MAGeCK, all sgRNAs/gene (including both TSSs if there are

22 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

549 two) are ranked based on p-values from mean variance modeling. A robust ranking

550 aggregation (RRA) algorithm is then used to call genes that are significantly enriched or

551 depleted based on p-value and false discovery rate (FDR).

552

553 qRT-PCR

554 For qPCR, RNA was extracted with the RNeasy Mini Kits (Qiagen). cDNA was

555 produced from 1 μg of purified RNA using the iScript Reverse Transcription Supermix

556 for RT-qPCR (Bio-Rad Laboratories). qPCR reactions were performed with the

557 SsoAdvanced Universal SYBR Green Supermix (Bio-Rad Laboratories) in a total

558 volume of 10 μl with primers at final concentrations of 500 nM. Primer sequences are

559 included in Table S11. The thermocycler was set for 1 cycle of 95 °C for 30 sec, and

560 40 cycles of 95 °C for 5 s and 55 °C for 15 s, respectively. Fold enrichment of the

561 assayed genes over the control HPRT and/or GAPDH loci were calculated using the

562 2−ΔΔCT method as previously described (Livak & Schmittgen, 2001).

563

564 Cell Event apoptosis detection assay

565 Cell Event reagent was added to cells for the duration of the drug or vehicle treatment, at

566 a final concentration of 0.5μM. Cells were harvested by trypsinization and analyzed on a

567 flow cytometer, either the BD FACSARIA for the screen or an Attune Nxt cytometer

568 (ThermoFisher) for the follow-up experiments.

569

570 Caspase activation assay

571 Cells were plated at 5,000 cells/well of a 384-well white opaque plates in corresponding

572 cell growth media. Cells were treated with compounds at indicated concentrations for 6

573 hours (37 °C, 5% CO2) with a final DMSO concentration of 0.3%. Caspase-3/7 activation

23 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

574 was subsequently determined using a Caspase-Glo 3/7 Reagent (Promega Corporation)

575 as described in manufacturer's instructions. Dose-response curves were plotted and

576 analyzed (including EC50 determination) using GraphPad Prism. Percentage

577 of caspase activation was calculated against the maximum caspase activation value

578 (100%) obtained with a combination of MCL1, BCL2, and Bcl-xL inhibition.

579

580 Cell viability assay

581 Cells were plated at 5,000 cells/well of a 384-well white opaque plates in corresponding

582 cell growth media. Cells were treated with compounds at indicated concentrations for 24

583 hours (37 °C, 5% CO2) with a final DMSO concentration of 0.3% and assayed for

584 viability using the CellTiter-Glo Reagent (Promega Corporation) as described in

585 manufacturer's instructions. Results were normalized to the samples without treatment at

586 time 0. GI50values were calculated using nonlinear regression algorithms in GraphPad

587 Prism.

588

589 Immunoblotting

590 Primary antibodies against the following proteins were used: CUL5 (ab184177; Abcam);

591 RNF7 (ab181986; Abcam); UBE2F (ab185234; Abcam); Noxa (OP180; Millipore); Bim

592 (ab32158; Abcam); Bid (ab32060; Abcam); Puma (ab9643; Abcam); Bad (ab62465;

593 Abcam); Bik (ab52182; Abcam); Beclin (ab207612; Abcam); Bmf (ab9655; Abcam);

594 MCL1 (D35A5 #5453; CST); Bcl-xL (54H6; CST); c-PARP (19F4 #9546; CST) Ubiquitin

595 (P4D1 #3936; CST); HA (C29F4; CST); GAPDH (14C10 #2118; CST); p53 (DO-1 sc-

596 126; Santa Cruz Biotechnology); HSP90 (sc-69703; Santa Cruz Biotechnology). For

597 each protein antibody, manufacturer’s recommended dilutions were used. Mouse or

598 rabbit immunoglobulin G was visualized at a 1:10,000 dilution: donkey anti-mouse 680

24 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

599 (925-68022; LI-COR); donkey anti-rabbit 680 (925-68023; LI-COR); donkey anti-

600 mouse 800 (925-32212; LI-COR); donkey anti-rabbit 800 (925-32213; LI-COR). Blots

601 were imaged on an Odyssey CLx Imaging System (LI-COR).

602

603 Immunoprecipitation

604 LK2 wild type and CUL5-KO cells were transfected with equal amounts of His-Ubiquitin

605 and either HA-Noxa or HA-Bim. 48 hours after transfection, cells were treated with

606 100μM epoxomicin for 8 hours. Cells were harvested and lysed in either denaturing

607 buffer (8M Urea, 300mM NaCl, 0.5% NP-40, 50mM Na2HPO4, 50mM Tris pH8.0) or

608 lysis buffer from ThermoFisher (PierceTM HA-Tag Magnetic IP/CoIP Kit #88838).

609 Denatured extracts were bound to magnetic Ni-NTA beads (Qiagen #36111), while other

610 lysed extracts were bound to magnetic anti-HA beads. Beads were washed 5 times and

611 eluted in 1X Laemmli buffer by boiling for 5 minutes. Samples were loaded on a gel and

612 processed for immunoblotting.

613

614 siRNA treatment

615 Cells were reverse-transfected in 6-well plates using RNAiMAX (Thermo Fisher

616 Scientific) with 50nM siRNA. 48h following siRNA treatment cells were treated for the

617 Cell Event apoptosis assay as indicated and also harvested to verify knockdown by

618 qRT-PCR. BCL2L11 (Bim) siRNA SMARTpool was from Dharmacon (L-004383-00-

619 0005) and BAK siRNA was obtained from Ambion (Life Technologies 4457298).

25 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

620 REFERENCES 621 622 Abulwerdi, F., Liao, C., Liu, M., Azmi, A. S., Aboukameel, A., Mady, A. S. A., … 623 Nikolovska-Coleska, Z. (2014). A novel small-molecule inhibitor of mcl-1 blocks 624 pancreatic cancer growth in vitro and in vivo. Molecular Cancer Therapeutics, 625 13(3), 565–575. doi:10.1158/1535-7163.MCT-12-0767 626 Burke, J. P., Bian, Z., Shaw, S., Zhao, B., Goodwin, C. M., Belmar, J., … Fesik, S. W. 627 (2015). Discovery of tricyclic indoles that potently inhibit Mcl-1 using fragment- 628 based methods and structure-based design. Journal of Medicinal Chemistry, 629 58(9), 3794–3805. doi:10.1021/jm501984f 630 Burnatowska-Hledin, M. A., Kossoris, J. B., Van Dort, C. J., Shearer, R. L., Zhao, P., 631 Murrey, D. A., … Barney, C. C. (2004). T47D breast cancer cell growth is 632 inhibited by expression of VACM-1, a cul-5 gene. Biochemical and Biophysical 633 Research Communications, 319(3), 817–825. doi:10.1016/j.bbrc.2004.05.057 634 Caenepeel, S., Brown, S. P., Belmontes, B., Moody, G., Keegan, K. S., Chui, D., … 635 Hughes, P. E. (2018). AMG 176, a Selective MCL1 Inhibitor, is Effective in 636 Hematological Cancer Models Alone and in Combination with Established 637 Therapies. Cancer Discovery. 638 Cai, Q.-L., Knight, J. S., Verma, S. C., Zald, P., & Robertson, E. S. (2006). EC5S 639 ubiquitin complex is recruited by KSHV latent antigen LANA for degradation of 640 the VHL and p53 tumor suppressors. PLoS Pathogens, 2(10), e116. 641 doi:10.1371/journal.ppat.0020116 642 Campbell, K. J., Dhayade, S., Ferrari, N., Sims, A. H., Johnson, E., Mason, S. M., … 643 Blyth, K. (2018). MCL-1 is a prognostic indicator and drug target in breast 644 cancer. Cell Death & Disease, 9(2), 19. doi:10.1038/s41419-017-0035-2 645 Cerami, E., Gao, J., Dogrusoz, U., Gross, B. E., Sumer, S. O., Aksoy, B. A., … Schultz, 646 N. (2012). The cBio cancer genomics portal: an open platform for exploring 647 multidimensional cancer genomics data. Cancer Discovery, 2(5), 401–404. 648 doi:10.1158/2159-8290.CD-12-0095 649 Cheng, E. H. Y., Sheiko, T. V., Fisher, J. K., Craigen, W. J., & Korsmeyer, S. J. (2003). 650 VDAC2 inhibits BAK activation and mitochondrial apoptosis. Science, 651 301(5632), 513–517. doi:10.1126/science.1083995 652 Chin, H. S., Li, M. X., Tan, I. K. L., Ninnis, R. L., Reljic, B., Scicluna, K., … Dewson, 653 G. (2018). VDAC2 enables BAX to mediate apoptosis and limit tumor 654 development. Nature Communications, 9(1), 4976. doi:10.1038/s41467-018- 655 07309-4 656 Czabotar, P. E., Lessene, G., Strasser, A., & Adams, J. M. (2014). Control of apoptosis 657 by the BCL-2 protein family: implications for physiology and therapy. Nature 658 Reviews. Molecular Cell Biology, 15(1), 49–63. doi:10.1038/nrm3722 659 Fay, M. J., Longo, K. A., Karathanasis, G. A., Shope, D. M., Mandernach, C. J., Leong, 660 J. R., … Husain, A. (2003). Analysis of CUL-5 expression in breast epithelial 661 cells, breast cancer cell lines, normal tissues and tumor tissues. Molecular Cancer, 662 2, 40. doi:10.1186/1476-4598-2-40 663 Galdeano, C., Gadd, M. S., Soares, P., Scaffidi, S., Van Molle, I., Birced, I., … Ciulli, A. 664 (2014). Structure-guided design and optimization of small molecules targeting the 665 protein-protein interaction between the von Hippel-Lindau (VHL) E3 ubiquitin 666 ligase and the hypoxia inducible factor (HIF) alpha subunit with in vitro

26 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

667 nanomolar affinities. Journal of Medicinal Chemistry, 57(20), 8657–8663. 668 doi:10.1021/jm5011258 669 Gao, J., Aksoy, B. A., Dogrusoz, U., Dresdner, G., Gross, B., Sumer, S. O., … Schultz, 670 N. (2013). Integrative analysis of complex cancer genomics and clinical profiles 671 using the cBioPortal. Science Signaling, 6(269), pl1. 672 doi:10.1126/scisignal.2004088 673 Garcia-Cuellar, M. P., Füller, E., Mäthner, E., Breitinger, C., Hetzner, K., Zeitlmann, L., 674 … Slany, R. K. (2014). Efficacy of cyclin-dependent-kinase 9 inhibitors in a 675 murine model of mixed-lineage leukemia. Leukemia, 28(7), 1427–1435. 676 doi:10.1038/leu.2014.40 677 Goodwin, C. M., Rossanese, O. W., Olejniczak, E. T., & Fesik, S. W. (2015). Myeloid 678 cell leukemia-1 is an important apoptotic survival factor in triple-negative breast 679 cancer. Cell Death and Differentiation, 22(12), 2098–2106. 680 doi:10.1038/cdd.2015.73 681 Gregory, G. P., Hogg, S. J., Kats, L. M., Vidacs, E., Baker, A. J., Gilan, O., … Shortt, J. 682 (2015). CDK9 inhibition by dinaciclib potently suppresses Mcl-1 to induce 683 durable apoptotic responses in aggressive MYC-driven B-cell lymphoma in vivo. 684 Leukemia, 29(6), 1437–1441. doi:10.1038/leu.2015.10 685 Hellvard, A., Zeitlmann, L., Heiser, U., Kehlen, A., Niestroj, A., Demuth, H.-U., … 686 Mydel, P. (2016). Inhibition of CDK9 as a therapeutic strategy for inflammatory 687 arthritis. Scientific Reports, 6, 31441. doi:10.1038/srep31441 688 Horlbeck, M. A., Gilbert, L. A., Villalta, J. E., Adamson, B., Pak, R. A., Chen, Y., … 689 Weissman, J. S. (2016). Compact and highly active next-generation libraries for 690 CRISPR-mediated gene repression and activation. ELife, 5. 691 doi:10.7554/eLife.19760 692 Huang, C.-H., Lujambio, A., Zuber, J., Tschaharganeh, D. F., Doran, M. G., Evans, M. J., 693 … Lowe, S. W. (2014). CDK9-mediated transcription elongation is required for 694 MYC addiction in hepatocellular carcinoma. Genes & Development, 28(16), 695 1800–1814. doi:10.1101/gad.244368.114 696 Huang, D. T., Ayrault, O., Hunt, H. W., Taherbhoy, A. M., Duda, D. M., Scott, D. C., … 697 Schulman, B. A. (2009). E2-RING expansion of the NEDD8 cascade confers 698 specificity to cullin modification. Molecular Cell, 33(4), 483–495. 699 doi:10.1016/j.molcel.2009.01.011 700 Joharapurkar, A. A., Pandya, V. B., Patel, V. J., Desai, R. C., & Jain, M. R. (2018). Prolyl 701 Hydroxylase Inhibitors: A Breakthrough in the Therapy of Anemia Associated 702 with Chronic Diseases. Journal of Medicinal Chemistry, 61(16), 6964–6982. 703 doi:10.1021/acs.jmedchem.7b01686 704 Kamura, T., Maenaka, K., Kotoshiba, S., Matsumoto, M., Kohda, D., Conaway, R. C., … 705 Nakayama, K. I. (2004). VHL-box and SOCS-box domains determine binding 706 specificity for Cul2-Rbx1 and Cul5-Rbx2 modules of ubiquitin ligases. Genes & 707 Development, 18(24), 3055–3065. doi:10.1101/gad.1252404 708 Kelly, G. L., & Strasser, A. (2011). The essential role of evasion from cell death in 709 cancer. Advances in Cancer Research, 111, 39–96. doi:10.1016/B978-0-12- 710 385524-4.00002-7 711 Koss, B., Morrison, J., Perciavalle, R. M., Singh, H., Rehg, J. E., Williams, R. T., & 712 Opferman, J. T. (2013). Requirement for antiapoptotic MCL-1 in the survival of

27 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

713 BCR-ABL B-lineage acute lymphoblastic leukemia. Blood, 122(9), 1587–1598. 714 doi:10.1182/blood-2012-06-440230 715 Kotschy, A., Szlavik, Z., Murray, J., Davidson, J., Maragno, A. L., Le Toumelin-Braizat, 716 G., … Geneste, O. (2016). The MCL1 inhibitor S63845 is tolerable and effective 717 in diverse cancer models. Nature, 538(7626), 477–482. doi:10.1038/nature19830 718 Kuhn, D. J., Chen, Q., Voorhees, P. M., Strader, J. S., Shenk, K. D., Sun, C. M., … 719 Orlowski, R. Z. (2007). Potent activity of carfilzomib, a novel, irreversible 720 inhibitor of the ubiquitin-proteasome pathway, against preclinical models of 721 multiple myeloma. Blood, 110(9), 3281–3290. doi:10.1182/blood-2007-01- 722 065888 723 Lauterwasser, J., Todt, F., Zerbes, R. M., Nguyen, T. N., Craigen, W., Lazarou, M., … 724 Edlich, F. (2016). The porin VDAC2 is the mitochondrial platform for Bax 725 retrotranslocation. Scientific Reports, 6, 32994. doi:10.1038/srep32994 726 Lee, A. S. Y., Kranzusch, P. J., & Cate, J. H. D. (2015). eIF3 targets cell-proliferation 727 messenger RNAs for translational activation or repression. Nature, 522(7554), 728 111–114. doi:10.1038/nature14267 729 Lemke, J., von Karstedt, S., Abd El Hay, M., Conti, A., Arce, F., Montinaro, A., … 730 Walczak, H. (2014). Selective CDK9 inhibition overcomes TRAIL resistance by 731 concomitant suppression of cFlip and Mcl-1. Cell Death and Differentiation, 732 21(3), 491–502. doi:10.1038/cdd.2013.179 733 Lessene, G., Czabotar, P. E., Sleebs, B. E., Zobel, K., Lowes, K. N., Adams, J. M., … 734 Watson, K. G. (2013). Structure-guided design of a selective BCL-X(L) inhibitor. 735 Nature Chemical Biology, 9(6), 390–397. doi:10.1038/nchembio.1246 736 Leverson, J D, Zhang, H., Chen, J., Tahir, S. K., Phillips, D. C., Xue, J., … Souers, A. J. 737 (2015). Potent and selective small-molecule MCL-1 inhibitors demonstrate on- 738 target cancer cell killing activity as single agents and in combination with ABT- 739 263 (navitoclax). Cell Death & Disease, 6, e1590. doi:10.1038/cddis.2014.561 740 Leverson, Joel D, Phillips, D. C., Mitten, M. J., Boghaert, E. R., Diaz, D., Tahir, S. K., … 741 Souers, A. J. (2015). Exploiting selective BCL-2 family inhibitors to dissect cell 742 survival dependencies and define improved strategies for cancer therapy. Science 743 Translational Medicine, 7(279), 279ra40. doi:10.1126/scitranslmed.aaa4642 744 Li, W., Xu, H., Xiao, T., Cong, L., Love, M. I., Zhang, F., … Liu, X. S. (2014). 745 MAGeCK enables robust identification of essential genes from genome-scale 746 CRISPR/Cas9 knockout screens. Genome Biology, 15(12), 554. 747 doi:10.1186/s13059-014-0554-4 748 Lingeman, E., Jeans, C., & Corn, J. E. (2017). Production of purified casrnps for 749 efficacious genome editing. Current Protocols in Molecular Biology, 120, 750 31.10.1-31.10.19. doi:10.1002/cpmb.43 751 Livak, K. J., & Schmittgen, T. D. (2001). Analysis of relative gene expression data using 752 real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods, 25(4), 753 402–408. doi:10.1006/meth.2001.1262 754 Lubbers, J., Lewis, S., Harper, E., Hledin, M. P., Marquez, G. A., Johnson, A. E., … 755 Burnatowska-Hledin, M. A. (2011). Resveratrol enhances anti-proliferative effect 756 of VACM-1/cul5 in T47D cancer cells. Cell Biology and Toxicology, 27(2), 95– 757 105. doi:10.1007/s10565-010-9173-3 758 Mahrour, N., Redwine, W. B., Florens, L., Swanson, S. K., Martin-Brown, S., Bradford,

28 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

759 W. D., … Conaway, J. W. (2008). Characterization of Cullin-box sequences that 760 direct recruitment of Cul2-Rbx1 and Cul5-Rbx2 modules to Elongin BC-based 761 ubiquitin ligases. The Journal of Biological Chemistry, 283(12), 8005–8013. 762 doi:10.1074/jbc.M706987200 763 Marash, L., Liberman, N., Henis-Korenblit, S., Sivan, G., Reem, E., Elroy-Stein, O., & 764 Kimchi, A. (2008). DAP5 promotes cap-independent translation of Bcl-2 and 765 CDK1 to facilitate cell survival during mitosis. Molecular Cell, 30(4), 447–459. 766 doi:10.1016/j.molcel.2008.03.018 767 Merino, D., Whittle, J. R., Vaillant, F., Serrano, A., Gong, J.-N., Giner, G., … Lindeman, 768 G. J. (2017). Synergistic action of the MCL-1 inhibitor S63845 with current 769 therapies in preclinical models of triple-negative and HER2-amplified breast 770 cancer. Science Translational Medicine, 9(401). 771 doi:10.1126/scitranslmed.aam7049 772 Okumura, F., Joo-Okumura, A., Nakatsukasa, K., & Kamura, T. (2016). The role of 773 cullin 5-containing ubiquitin ligases. Cell Division, 11, 1. doi:10.1186/s13008- 774 016-0016-3 775 Okumura, F., Matsuzaki, M., Nakatsukasa, K., & Kamura, T. (2012). The Role of 776 Elongin BC-Containing Ubiquitin Ligases. Frontiers in Oncology, 2, 10. 777 doi:10.3389/fonc.2012.00010 778 Petroski, M. D., & Deshaies, R. J. (2005). Function and regulation of cullin-RING 779 ubiquitin ligases. Nature Reviews. Molecular Cell Biology, 6(1), 9–20. 780 doi:10.1038/nrm1547 781 Querido, E., Blanchette, P., Yan, Q., Kamura, T., Morrison, M., Boivin, D., … Branton, 782 P. E. (2001). Degradation of p53 by adenovirus E4orf6 and E1B55K proteins 783 occurs via a novel mechanism involving a Cullin-containing complex. Genes & 784 Development, 15(23), 3104–3117. doi:10.1101/gad.926401 785 Richardson, P. G., Sonneveld, P., Schuster, M. W., Irwin, D., Stadtmauer, E. A., Facon, 786 T., … Assessment of Proteasome Inhibition for Extending Remissions (APEX) 787 Investigators. (2005). Bortezomib or high-dose dexamethasone for relapsed 788 multiple myeloma. The New England Journal of Medicine, 352(24), 2487–2498. 789 doi:10.1056/NEJMoa043445 790 S64315. (n.d.). Phase I Study of S64315 Administred Intravenously in Patients With 791 Acute Myeloid Leukaemia or Myelodysplastic Syndrome - Full Text View - 792 ClinicalTrials.gov. Retrieved August 22, 2018, from 793 https://clinicaltrials.gov/ct2/show/NCT02979366 794 Sato, Y., Kamura, T., Shirata, N., Murata, T., Kudoh, A., Iwahori, S., … Tsurumi, T. 795 (2009). Degradation of phosphorylated p53 by viral protein-ECS E3 ligase 796 complex. PLoS Pathogens, 5(7), e1000530. doi:10.1371/journal.ppat.1000530 797 Siegel, R. L., Miller, K. D., & Jemal, A. (2016). Cancer statistics, 2016. CA: A Cancer 798 Journal for Clinicians, 66(1), 7–30. doi:10.3322/caac.21332 799 Soucy, T. A., Smith, P. G., Milhollen, M. A., Berger, A. J., Gavin, J. M., Adhikari, S., … 800 Langston, S. P. (2009). An inhibitor of NEDD8-activating enzyme as a new 801 approach to treat cancer. Nature, 458(7239), 732–736. doi:10.1038/nature07884 802 Tao, Z.-F., Hasvold, L., Wang, L., Wang, X., Petros, A. M., Park, C. H., … Souers, A. J. 803 (2014). Discovery of a Potent and Selective BCL-XL Inhibitor with in Vivo 804 Activity. ACS Medicinal Chemistry Letters, 5(10), 1088–1093.

29 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

805 doi:10.1021/ml5001867 806 Tron, A. E., Belmonte, M. A., Adam, A., Aquila, B. M., Boise, L. H., Chiarparin, E., … 807 Hird, A. W. (2018). Discovery of Mcl-1-specific inhibitor AZD5991 and 808 preclinical activity in multiple myeloma and acute myeloid leukemia. Nature 809 Communications, 9(1), 5341. doi:10.1038/s41467-018-07551-w 810 Xiao, Y., Nimmer, P., Sheppard, G. S., Bruncko, M., Hessler, P., Lu, X., … Phillips, D. 811 C. (2015). MCL-1 Is a Key Determinant of Breast Cancer Cell Survival: 812 Validation of MCL-1 Dependency Utilizing a Highly Selective Small Molecule 813 Inhibitor. Molecular Cancer Therapeutics, 14(8), 1837–1847. doi:10.1158/1535- 814 7163.MCT-14-0928 815 Xu, L., Wei, Y., Reboul, J., Vaglio, P., Shin, T.-H., Vidal, M., … Harper, J. W. (2003). 816 BTB proteins are substrate-specific adaptors in an SCF-like modular ubiquitin 817 ligase containing CUL-3. Nature, 425(6955), 316–321. doi:10.1038/nature01985 818 Xu, X.-M., Wang, X.-B., Chen, M.-M., Liu, T., Li, Y.-X., Jia, W.-H., … Tang, H. (2012). 819 MicroRNA-19a and -19b regulate cervical carcinoma cell proliferation and 820 invasion by targeting CUL5. Cancer Letters, 322(2), 148–158. 821 doi:10.1016/j.canlet.2012.02.038 822 Yin, J., Li, Y., Zhao, H., Qin, Q., Li, X., Huang, J., … Wei, S. (2016). Copy-number 823 variation of MCL1 predicts overall survival of non-small-cell lung cancer in a 824 Southern Chinese population. Cancer Medicine, 5(9), 2171–2179. 825 doi:10.1002/cam4.774 826 Zhang, H., Guttikonda, S., Roberts, L., Uziel, T., Semizarov, D., Elmore, S. W., … Lam, 827 L. T. (2011). Mcl-1 is critical for survival in a subgroup of non-small-cell lung 828 cancer cell lines. Oncogene, 30(16), 1963–1968. doi:10.1038/onc.2010.559 829 Zhou, W., Xu, J., Li, H., Xu, M., Chen, Z. J., Wei, W., … Sun, Y. (2017). Neddylation 830 E2 UBE2F promotes the survival of lung cancer cells by activating CRL5 to 831 degrade NOXA via the K11 linkage. Clinical Cancer Research, 23(4), 1104– 832 1116. doi:10.1158/1078-0432.CCR-16-1585 833 834

30 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

835 836 837 Figure 1. Several MCL1-amplified NSCLC lines are resistant to treatment with 838 CDK9i or MCL1i. 839 (A) Schematic illustrating the mechanism of action of CDK9 and MCL1 inhibitors. The 840 CDK9 inhibitor (CDK9i) inhibits transcription elongation, thus mRNAs with short half-lives 841 such as MCL1 are highly susceptible to acute CDK9 inhibition. The MCL1 inhibitor 842 (MCL1i) is a BH3-mimetic that binds directly to MCL1. 843 (B) Graphical representation of a panel of cell lines depicting their MCL1 copy number, 844 their ratio of MCL1:Bcl-xL protein and whether they are sensitive to the drug treatment 845 indicated. EC50 values plotted for a 6h CDK9i treatment (top graph) derived from 846 Caspase-Glo 3/7 assays. GI50 values plotted for a 24h MCL1i treatment (bottom graph) 847 using CellTiter-Glo. Maroon circles indicate cell lines resistant to drug despite being 848 MCL1-amplified. Highlighted in bright red is a resistant cell line (LK2) used for further 849 study in this report and a sensitive cell line (H23) is shown in grey. 850 (C) Dose response curves of LK2 and H23 treated with CDK9i (top) and MCL1i (bottom). 851 Caspase activation was measured at 6 hours post drug treatment at the indicated 852 concentrations by CaspaseGlo 3/7 and normalized to a positive control containing 853 inhibitors of MCL1, BCL2 and Bcl-xL. 854 (D) Cell viability curves of the resistant LK2 and sensitive H23 lines 24 hours following 855 drug treatment with CDK9i (top) or MCL1 (bottom) at increasing concentrations as 856 indicated. Viability was measured using the Cell Titer Glo assay normalized to a DMSO 857 control.

31 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

858 859 860 Figure 2. Genome-wide CRISPRi screens identify factors that resensitize lung 861 cancer cells to inhibition of MCL1. 862 (A) Schematic outlining the genome-wide CRISPRi screen in LK2 cells. Cells were 863 exposed to acute drug treatments, fixed and FACS-sorted using the fluorogenic 864 apoptotic detection reagent Cell Event. Enriched and depleted sgRNAs were identified 865 by next-generation sequencing. 866 (B + C) Volcano plots showing sgRNA-targeted genes significantly enriched or depleted 867 in the apoptosing cell population following treatment with CDK9i (B) or MCL1i (C). 868 Average of 2 independent experiments is graphed. Green highlighted points indicate 869 genes with a known role in apoptosis that had a significant fold change over background. 870 Magenta points highlight members of the CUL5-RNF7-UBE2F ubiquitin complex that 871 were significantly enriched. Beige highlighted points are members of the Eukaryotic 872 Translation Initiation Factor 3 (eIF3) complex. Blue point on CDK9i volcano plot 873 highlights EIF4G2, a gene that may be involved in cap-independent translation of Bcl-xL. 874

32 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

875 876 877 Figure 3. Depletion of the CUL5-RNF7-UBE2F ubiquitin complex induces 878 apoptosis upon treatment with CDK9i or MCL1i. 879 (A) Schematic depicting the CUL5 ubiquitin ligase scaffold and its interacting partners. 880 (B) Western blots confirming effective knockdown of CUL5, RNF7 and UBE2F by stable 881 lentiviral expression of dCas9-KRAB and corresponding sgRNAs. Asterisk indicates 882 neddylated CUL5. GAPDH serves as loading control. 883 (C) Quantification of western blots as in B. Error bars show standard deviations from 884 three independent biological replicates. 885 (D) Induction of apoptosis when cells depleted of CUL5, RNF7 or UBE2F (as in B) are 886 treated with CDK9i (3μM for 6h), MCL1i (1μM for 12h) or DMSO. Knockdown of Bcl-xL 887 serves as a positive control for induction of apoptosis. Percentage of apoptosis 888 determined by flow cytometry detection of the fluorogenic apoptotic detection reagent, 889 Cell Event. Error bars are standard deviations of three independent biological replicates. 890 (E) Western blotting for cleaved PARP (c-PARP) serves as orthogonal readout for 891 induction of apoptosis following knockdown of target genes and treatment with 3μM 892 CDK9i for 6 hours or 1μM MCL1i for 12 hours. Knockdown of Bcl-xL is included as a 893 positive control, however due to extreme toxicity of the combination treatment, cells were 894 harvested 3 hours after treatment with 3μM CDK9i or 0.5 hours after treatment with 1μM 895 MCL1i. GAPDH, loading control. 896 (F) Dose response curves of caspase induction showing resensitization of CUL5- 897 knockout (CUL5-KO c3) and UBE2F-knockout (UBE2F-KO c1) lines as compared to 898 parental LK2 cells when treated with CDK9i (top) and MCL1i (bottom). Caspase 899 activation was measured at 10 hours post drug treatment at the indicated concentrations 900 by CaspaseGlo 3/7 and normalized to a positive control containing inhibitors of MCL1, 901 BCL2 and Bcl-xL.

33 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

902 903 904 Figure 4. The CUL5-RNF7-UBE2F ubiquitin complex regulates levels of BH3-only 905 apoptotic sensitizers Bim and Noxa. 906 (A) Western blotting of all BH3-only proteins in cell lines with depleted CUL5, RNF7 or 907 UBE2F show increased levels of Bim and Noxa as compared to cells transduced with a 908 non-targeting (NT) sgRNA. GAPDH serves as the loading control. 909 (B) Western blot of all BH3-only proteins in NT and CUL5-depleted cells treated with 910 DMSO, CDK9i (3μM for 6h) or MCL1i (1μM for 12h). Tubulin serves as loading control. 911 (C) Quantification of western blots as in (B). Error bars derived from standard deviations 912 of three biological replicates consisting of independently treated and isolated protein 913 samples. Asterisk indicates p value <0.05 as determined by a paired student T-Test. 914 (D) qRT-PCR showing relative mRNA levels of Noxa, Bim and MCL1 in LK2 parental and 915 CUL5-KO c1 cells treated with DMSO, CDK9i (3μM for 6h) or MCL1i (3μM for 6h). Error 916 bars show standard deviations from three biological replicates on independently treated 917 and isolated RNA samples. 918 (E) Western blot of LK2 and CUL5-KO c1 cells treated for 6 hours with epoxomicin 919 (100μM), or folimycin (100μM) or both epoxomicin and folimycin (100μM each). GAPDH, 920 loading control. 921

34 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

922 923 924 Figure 5. CUL5-depleted cells regain resistance to CDK9i, but not MCL1i, when 925 Noxa is knocked down. 926 (A) CUL5 and Noxa were knocked down in dCas9-KRAB-expressing LK2 cells by 927 selecting for corresponding sgRNAs. Bim and Bak were knocked down by siRNA. 928 Apoptosis was measured using Cell Event detection by flow cytometry after treatment 929 with 3μM CDK9i for 6 hours. Error bars are standard deviations of 3 biological drug 930 treatment replicates. 931 (B) Genetic manipulations were performed and apoptosis was measured as in (A) 932 following MCL1i treatment (1μM for 12 hours). 933 (C) Schematic illustrating potential mechanisms of resensitization of resistant NSCLC 934 lines to inhibition of MCL1. 935

35 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

936 937 938 Figure S1. Performing the genome-wide CRISPRi screens and analysis of results. 939 (A) Bar graph showing knockdown of five genes by qRT-PCR to confirm functionality of 940 the dCas9-KRAB CRISPRi construct in LK2 cells. 941 (B) Volcano plots (as in Figure 2B) showing significantly enriched and depleted genes 942 when treated with CDK9i (left), MCL1 (middle) and DMSO (right). Black dots indicate 943 targeted genes. Grey dots indicate negative control pseudogenes with non-targeting 944 guides. Highlighted points of different colors show genes with significant fold changes 945 with annotated roles in apoptosis (green), factors of the CUL5-RNF7-UBE2F ubiquitin 946 complex (magenta), and members of the eIF3 complex (beige). 947 (C + D) Relative sgRNA enrichment was also determined by the MAGeCK analysis 948 pipeline. Top ten targeted genes following CDK9i and MCL1i treatment are shown in (C) 949 and (D) respectively. sgRNAs targeting members of the CUL5-RNF7-UBE2F complex 950 (outlined in magenta box) were once again identified as enriched and were amongst the 951 most significant hits.

36 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

952 953 954 Figure S2. Generation of CUL5- and UBE2F-knockouts also resensitize cells to 955 MCL1 inhibition. 956 (A) Induction of apoptosis when cells are depleted of CUL5, RNF7, UBE2F or Bcl-xL with 957 an additional set of sgRNAs than those used in Figure 3D and are treated with CDK9i 958 (3μM for 6h), MCL1i (1μM for 12h) or DMSO. Percentage of apoptosis determined by 959 flow cytometry for detection of the fluorogenic apoptotic detection reagent Cell Event. 960 Error bars are standard deviations of three biological replicates. 961 (B) Western blotting validating loss of CUL5 and UBE2F protein in the three CUL5-KO 962 clones and one UBE2F-KO clone respectively. As shown before, protein levels of RNF7 963 are also reduced when CUL5 is absent. GAPDH, loading control. 964 (C) Sanger sequencing traces of the alleles generated in the CUL5- and UBE2F 965 knockouts. Exon 1 size is indicated, and resulting deletions are shown. 966 (D) CUL5-KO and UBE2F-KO clones were treated with 3μM CDK9i for 6 hours or 1μM 967 MCL1i for 12 hours. Percent apoptosis determined using Cell Event detection by flow 968 cytometry. Error bars derived from standard deviations of three biological replicates.

37 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

969 970 971 Figure S3. Knockdown of Elongin B but not Elongin C or BTBD3 also resensitizes 972 cells to MCL1 inhibition. 973 (A) qRT-PCR to confirm efficient knockdown of Elongin B, Elongin C and BTBD3. 974 (B) Knockdown of Elongin B (EloB), but not Elongin C (EloC), also elicits similar levels of 975 apoptosis as compared to CUL5-knockdown when treated with CDK9i or MCL1i. 976 Knockdown of BTBD3 in conjunction with CDK9i and MCL1i has a modest increase in 977 apoptosis. Apoptosis measured by Cell Event, error bars are standard deviations of 978 three biological replicates. 979 (C) Plotted are percentages of apoptosis as determined by flow cytometry measuring 980 Cell Event fluorescence after 3μM drug treatment for 6 hours, as indicated on the X axis. 981 Each dot corresponds to an individual CRISPRi gene knockdown. Non-target (NT) 982 sgRNA serves as the negative control (white) and CUL5 knockdown serves as the 983 positive control (magenta). 984 (D) Western blot for p53 in NT and CUL5-depleted cells treated with DMSO, CDK9i (3μM 985 for 6h) or MCL1i (1μM for 12h). GAPDH serves as loading control. 986 (E) Western blot of MCL1 and Bcl-xL in LK2 and CUL5-KO c1 cells treated with DMSO, 987 CDK9i or MCL1i. HSP90 serves as loading control. 988

38 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

989 990 991 Figure S4. Indirect regulation of Bim and Noxa by the CUL5-RNF7-UBE2F 992 complex. 993 (A) LK2 and CUL5-KO c1 cells were transiently transfected with HA-Noxa and His- 994 Ubiquitin. 48 hours following transfection cells were incubated with 100μM epoxomicin 995 for 8 hours and subsequently harvested for immunoprecipitations (IP) with anti-HA or Ni- 996 NTA. Western blot show 1% input and 50% of IP loaded and probed with the indicated 997 antibodies. 998 (B) LK2 and CUL5-KO c1 cells were transiently transfected with HA-Bim and His- 999 Ubiquitin. 48 hours following transfection cells were incubated with 100μM epoxomicin 1000 for 8 hours and subsequently harvested for immunoprecipitations (IP) with anti-HA or Ni- 1001 NTA. Western blot show 1% input and 50% of IP loaded and probed with the indicated 1002 antibodies. 1003

39 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

1004 1005 1006 Figure S5. CUL5-kd cells regain resistance to CDK9i when Noxa is depleted. 1007 (A) As in Figure 5, CUL5 and Noxa were knocked down in dCas9-KRAB-expressing LK2 1008 cells by selecting for corresponding sgRNAs. Bim and Bak were knocked down by 1009 siRNA. Apoptosis was measured using Cell Event detection by flow cytometry after 1010 treatment with 3μM CDK9i for 6 hours. Bar graphs depict an independent set of 1011 transductions and siRNA transfections from Figure 5. Error bars are standard deviations 1012 of 3 biological drug treatment replicates. 1013 (B) Genetic manipulations were performed and apoptosis was measured as in (A) 1014 following MCL1i treatment. 1015 (C) Representative data set illustrating knockdown efficiency by qRT-PCR for genetic 1016 manipulations shown in (Figure 5A-B). Error bars are standard deviations of technical 1017 replicates. qRT-PCR was conducted to validate knockdown for each biological replicate 1018 of sgRNA transduction and siRNA transfection. nd, not determined. 1019

40 bioRxiv preprint doi: https://doi.org/10.1101/518555; this version posted January 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

1020 Table S1 (related to Figure 1). A subset of resistant and sensitive NSCLC lines. 1021 For each cell line, the table indicates MCL1 copy number, MCL1:Bcl-xL protein ratio and 1022 EC50 concentrations for both CDK9i and MCL1 treatments. 1023 MCL1 Copy Mcl1:BclxL MCL1i MCL1i CDK9i Number Protein GI50 (µM) Caspase Caspase Ratio EC50 (µM) EC50 (µM) SKLU1 1.5 0.41 10.000 30.000 30.000 HCC827 2.3 4.38 10.000 30.000 30.000 H460 3.2 16.58 10.000 30.000 30.000 H1734 3.6 2.63 6.150 20.056 30.000 LK2 4.7 3.71 7.040 30.000 30.000 H1395 6.6 3.79 10.000 15.404 30.000 H23 3.1 39.59 0.383 0.198 0.183 H2110 3.9 18.29 0.113 0.226 0.270 H1568 6.5 14.26 0.111 0.3 0.151 1024 1025 1026 1027 SUPPLIED AS ADDITIONAL SUPPORTING DATA: 1028 1029 Table S2 (related to Figure 2). ScreenProcessing: sgRNA counts. 1030 1031 Table S3 (related to Figure 2). ScreenProcessing: sgRNA phenotype scores. 1032 1033 Table S4 (related to Figure 2). ScreenProcessing: gene phenotype scores. 1034 1035 Table S5 (related to Figure S1). MAGeCK: sgRNA summary for CDK9i. 1036 1037 Table S6 (related to Figure S1). MAGeCK: gene summary for CDK9i. 1038 CDK9i treatment condition compared to ‘background’. Ordered by top enriched genes. 1039 1040 Table S7 (related to Figure S1). MAGeCK: sgRNA summary for MCL1i. 1041 1042 Table S8 (related to Figure S1). MAGeCK: sgRNA summary for MCL1i. 1043 MCL1i treatment condition compared to background. Ordered by top enriched genes. 1044 1045 Table S9 (related to Figure S4). Analysis of apoptosis following knockdown of 1046 putative CUL5 substrate adaptors challenged with CDK9i or MCL1i. 1047 1048 Table S10 (related to Figure 3 and S3). Protospacer sequences. 1049 1050 Table S11. Primer Sequences

41