Low E2F2 Activity Is Associated with High Genomic Instability and Parpi Resistance

bioRxiv preprint doi: https://doi.org/10.1101/777870; this version posted October 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder.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 Low E2F2 activity is associated with high genomic instability and PARPi resistance

2 Jonathan P Rennhack1 and Eran R. Andrechek1,2

8 1 – Department of Physiology, Michigan State University, East Lansing, Michigan 9 10 2 – Corresponding author 11 Eran R. Andrechek 12 Department of Physiology 13 Michigan State University 14 2194 BPS Building 15 567 Wilson Road 16 East Lansing, MI 17 48824 18 [email protected] 19 20 (517)884-5042 Office 21 (517)884-5020 Lab 22 (517)355-5125 Fax 23 24 This work was supported with NIH R01CA160514 and Worldwide Cancer Research WCR - 14- 25 1153 to E.R.A as well as NIH 1F99CA212221–01 to J.P.R 26 27 Running Title: E2F2 activity correlates with DNA instability and PARPi resistance

28 29 30 Authors' contributions

31 JPR performed all experiments, JPR and ERA both contributed in the design of the experiments

32 and drafting the manuscript.

34 Abstract

35 The E2F family, classically known for a central role in cell cycle, has a number of emerging roles

36 in cancer including angiogenesis, metabolic reprogramming, metastasis and DNA repair. E2F1

37 specifically has been shown to be a critical mediator of DNA repair; however, little is known about

38 DNA repair and other E2F family members. Here we present an integrative bioinformatic and

39 high throughput drug screening study to define the role of E2F2 in maintaining genomic integrity

40 in breast cancer. We utilized in vitro E2F2 ChIP-chip and over expression data to identify

41 transcriptional targets of E2F2. This data was integrated with gene expression from E2F2

42 knockout tumors in an MMTV-Neu background. Finally, this data was compared to human

43 datasets to identify conserved roles of E2F2 in human breast cancer through the TCGA breast

44 cancer, Cancer Cell Line Encyclopedia, and CancerRx datasets. Here we have computationally

45 predicted that E2F2 transcriptionally regulates key mediators of DNA repair. Our gene expression

46 data supports this hypothesis and low E2F2 activity is associated with a highly unstable tumor. In

47 human breast cancer E2F2, status was also correlated with a patient’s response to PARP inhibition

48 therapy. Taken together this manuscript defines a novel role of E2F2 in cancer progression beyond

49 cell cycle and could be therapeutically relevant.

51 Author Summary

52 The E2F family of proteins have been known to regulate cell cycle and have recently been shown

53 to have a number of roles in tumor progression. Here we use a combination of computational

54 techniques and high-throughput drug screening data to establish a novel role of E2F2 in

55 maintaining genomic integrity. We have shown that a number of direct and indirect target genes

56 of E2F2 are involved in multiple classical DNA repair pathways. Importantly, this was shown to

57 be unique to E2F2 and not present with other activator E2Fs like E2F1. We have also shown that

58 E2F2 activity is positively correlated with PARP inhibitor sensitivity regardless of BRCA1/2

59 status. This is important due to the recent approval of PARP inhibitor therapy in the clinic. Based

60 on our work E2F2 activity could serve as a novel biomarker of response and may identify a new

61 cohort of patients which could benefit from PARPi therapy.

63 Keywords

64 Genomic stability, breast cancer, E2F transcription factors, mouse model

65 Introduction

66 Breast cancer remains the leading cause of cancer related deaths in women. This is largely

67 due to two factors, metastasis and the heterogeneity of breast cancer. While metastasis to distal

68 sites is responsible for mortality, the difficulty of treating a heterogeneous disease is one of the

69 primary factors allowing that progression to occur. The heterogeneity of breast cancer is evident

70 in several facets, including histological subtypes, progression and response to treatment.

71 Underlying this diversity are the unique genomic alterations, methylation patterns and the resulting

72 gene expression differences that are recognized in the PAM50 classification system (1, 2). Each

73 of the subtypes present (Luminal A/B, Basal, Claudin Low, HER2+ve and normal like) have a

74 unique transcriptomic profile, resulting in the dysregulation in key proteins in breast cancer,

75 including the alteration of the E2F family of transcription factors (3-7).

76 The E2F family of transcription factors is composed of nine unique family members (E2F1,

77 E2F2, E2F3a, E2F3b, E2F4, E2F5, E2F6, E2F7, and E2F8) (8-10). These family members bind a

78 conserved motif with gene specificity contributed by cofactors (11-13). As a result, the E2Fs have

79 been shown to directly regulate many downstream genes. Classically they have been divided into

80 family members that activate transcription (E2F1, E2F2, and E2F3a) and repressors of

81 transcription (E2F3b, E2F4, E2F5, E2F6, E2F7, and E2F8). These definitions have recently

82 become less clear, with each family member functioning in both activating and repressing roles

83 depending on the tissue and developmental context (11, 14).

84 The role of the E2F family has widely been described in cell cycle where the members

85 regulate the G1/S checkpoint in response to Cyclin D levels (15, 16). However, beyond the G1/S

86 checkpoint de-regulation E2Fs have a number of emerging roles in cancer (17). This includes

87 roles in other aspects of cancer progression including angiogenesis (18), metabolic reprograming

88 (19), and apoptosis (20, 21). Indeed, numerous accounts detail the role of the activators in

89 metastasis of human breast cancer as well as mouse models of the disease (3, 4, 6, 22-24).

90 An additional emerging role for the E2Fs has been in the regulation of genomic stability.

91 Specifically, the role of E2F1 has been well defined with both transcriptional and non-

92 transcriptional roles in DNA repair (25). In response to DNA damage, E2F1 undergoes post

93 translational phosphorylation by ATM (26), leading to protein stabilization and increased

94 expression of repair proteins. In addition to the transcriptional role in DNA repair, E2F1 is

95 physically recruited to sites of damage. During cases of double stranded breaks (27) or UV damage

96 (28) it was observed that E2F1 formed foci with other damage induced proteins at the site of DNA

97 damage. It has been shown the E2F1 is required for the efficient recruitment of other repair proteins

98 including XPA/XPC (28) and NBS1 (29). It has also been shown that E2F2 is transcriptionally

99 upregulated in response to DNA damage and has been shown to complex with Rad51 and sites of

100 DNA damage in neuronal cells (30).

101 The amplification of the centrosome within a cell leads to defects in cellular segregation

102 and DNA replication, which in turn leads to the single nucleotide variants, copy number

103 alterations, and translocations characteristics. Importantly, activator E2Fs have been shown to be

104 associated with centrosome amplification (31). It is through this amplification of the centrosome

105 that it is believed E2Fs contribute to the DNA instability associated with their misregulation.

106 However, the mechanism and specific E2Fs involved in this process remain unclear.

107 Together, there is an emerging role for the activator E2Fs role in maintaining genomic

108 integrity, but only the role of E2F1 has been well defined. Here we present a key role for E2F2 in

109 maintaining genomic integrity. Through the use of cell lines, mouse models, and human samples,

110 we have identified that low E2F2 activity level is associated with tumors containing high levels of

111 genomic instability. Furthermore, the levels of E2F2 have direct impact on therapeutic response

112 in clinical data. Indeed, tumors with high E2F2 activity have an increased sensitivity to cell cycle

113 targeted chemotherapy as well as targeted PARP inhibitors.

114

115 Results

116 Based on the published literature for E2Fs in non-cell cycle roles, we hypothesized that

117 E2F2 had key activities other than the traditional role in cell cycle. To test this hypothesis in large

118 transcriptomic datasets, we used principle components analysis on gene expression data from cells

119 infected with adenoviral delivered GFP compared with adenoviral delivered E2F2 (Figure 1A).

120 This analysis revealed a consistent gene expression profile associated with over expression of

121 E2F2 (Figure 1B). We used Significance Analysis of Microarrays (SAM) (32) analysis to identify

122 consistently overexpressed genes with the infection of Ad-E2F2 relative to GFP. E2F2 induction

123 allowed for the identification of overrepresented gene ontology groups through the use of

124 PANTHER analysis (Figure 1C). As expected, this uncovered over-representation of cell cycle

125 proteins. Interestingly we also identified a number of repair associated gene ontology groups,

126 including double-stranded break repair and non-recombinational repair.

127 To identify genes and pathways directly regulated by E2F2, we utilized publicly available

128 E2F2 ChIP-Chip data (Figure 1D) (33). This revealed numerous genes bound by E2F2 across the

129 genome. When target genes were analyzed with PANTHER and GATHER, the ontologies were

130 consistent with the E2F2 overexpression data. Indeed, cell cycle and DNA damage repair gene

131 ontology groups were overrepresented, including DNA repair and double-stranded break repair

132 (Figure 1E). Given the similar pathways it was surprising that when we compared predicted E2F2

133 targets obtained through ChIP-Chip and gene expression analysis that we identified little overlap

134 in the gene lists (Figure 1F). This may indicate that the pathways are regulated by both direct and

135 indirect / downstream targets of E2F2. To determine if there was bias towards one particular repair

136 pathway, we identified overlap between E2F2 regulated genes (gene expression and ChIP-chip)

137 and each of the major repair pathways including Non-Homologous End Joining, Homologous End

138 Joining, Base Excision repair, and nucleotide excision repair. This analysis illustrated that E2F2

139 regulates key proteins in each repair pathway (Figure 1G).

140 To determine if there is a role for E2F2 in DNA repair processes in the in vivo setting, we

141 utilized publicly available E2F2 knockout transcriptome data within the MMTV-Neu mouse model

142 mammary tumors (3, 34). Unsupervised clustering identified a consistent transcriptional profile

143 with E2F2 loss (Figure 2A). This revealed five major clusters, three primarily composed of

144 MMTV-Neu E2F2 knockout samples and two clustered primary populated with MMTV-Neu E2F

145 wildtype samples. This demonstrated a unique gene expression profile associated with the loss of

146 E2F2, which was unique relative to the MMTV-Neu E2F wildtype background. To explore the

147 enriched cellular processes in this data, we used Gene Set Enrichment Analysis (GSEA) comparing

148 MMTV-Neu tumors with and without E2F2. This revealed, similar to the in vitro data, that E2F2

149 null samples were enriched for both the instability gene set (Figure 2B) and the repair gene set

150 (Figure 2C). We do not believe this to be in conflict with the in vitro data with an upregulated

151 gene expression signature for repair not necessarily correlating with an increase in the repair

152 process.

153 To test whether misregulation of E2F2 associated repair and instability functions were

154 associated with a more unstable tumor, we predicted gene copy number alterations across the

155 genome. We utilized the Analysis of CNAs by Expression data (ACE) algorithm (35) to predict

156 copy number profiles of control FVB wildtype mammary glands (Figure 3A), MMTV-Neu E2F2

157 wildtype mammary tumors (Figure 3B), and MMTV-Neu E2F2 knockout mammary tumors

158 (Figure 3C). Examining a portion of chromosome 4 as a case study, we observed an increased

159 number of significant (p<.05) amplification and deletion events in the MMTV-Neu E2F2 knockout

160 samples relative to both wild FVB mammary controls and MMTV-Neu E2F2 wildtype tumors.

161 Expanding to the entire genome, this was a consistent across genotypes with the E2F2 knockout

162 samples being the most unstable (Figure 3D). Interestingly, when this was compared to the E2F1

163 knockout samples we also saw the E2F2 null samples to be increasingly unstable indicating a

164 specific role of E2F2 in genomic stability (Figure S1).

165 While we have identified a potential role for E2F2 associated instability in mouse tumors,

166 this role has not previously been examined in human breast cancer. To address a potential role for

167 E2F2 in breast cancer, an E2F2 activity signature was used to divide the TCGA breast cohort into

168 low / high E2F2 activity groups. As predicted from the mouse mammary tumor data, human breast

169 cancer with low E2F2 activity contained significantly more copy number variants than those with

170 high E2F2 activity (p<.05) (Figure 4A). Furthermore, tumors with low predicted E2F2 activity

171 were observed to have enrichment for genomic instability in gene set enrichment analysis.

172 Specifically, there was a significant enrichment of genes involved with the response to UV induced

173 damage (Figure 4B).

174 In order to determine if E2F2 preferentially regulated a specific repair pathway, we utilized

175 single sample gene set enrichment analysis (ssGSEA) (36). The resulting scores were low/high

176 normalized to return an activity score between 0 and 1 for the four major repair pathways. Low

177 E2F2 activity resulted in significantly lower activity in each repair pathway including: Base

178 Excision Repair, Nucleotide Excision Repair, Homologous End Joining, and Non-Homologous

179 End joining (Figure S2). Unsupervised hierarchical clustering of this data revealed that regardless

180 of the pathway, low E2F2 activity was associated with low ssGSEA pathway scores (Figure 4C).

181 Alterations in DNA repair pathways are manifested in response to therapy. Accordingly,

182 we sought to test whether E2F2 levels impacted therapeutic response using the CancerRx dataset.

183 After predicting the E2F2 activity level across all breast cancer datasets we identified differentially

184 lethal compounds between E2F2 low cell lines and E2F2 high cell lines (Table S2). This revealed

185 a number of interesting candidate compounds. For example. tumors with low E2F2 activity

186 responded poorly to cell cycle inhibiting compounds such as Cisplatin. However, they responded

187 well to PIK3 targeted therapy such as PI-103. In addition, we observed that tumors with high

188 E2F2 activity were sensitive to cell cycle compounds and resistant to other forms of therapy.

189 With a role for E2F2 in repair, we examined response to repair targeted therapy, including

190 PARP inhibitors. Surprisingly we identified that high E2F2 activity was associated with a

191 significantly higher response to common PARP inhibitors including Talazoparib (Figure 5A),

192 Olaparib (Figure 5B), and Rucaparib (Figure 5C) as identified by a significant decrease in the IC50

193 of each compound on each cell line. Importantly, this correlation was independent of the BRCA1/2

194 status of the cell line. To test if this correlation was a function of any E2F or specific to E2F2, we

195 performed a similar analysis for E2F1. In this analysis, we predicted E2F1 activity through the

196 use of a gene activity signature and identified the IC50 of each cell line. Importantly, in this

197 analysis we saw no difference in the IC50 or PARP targeted therapy in low or high E2F1 activity

198 (Figure S3). This indicates the identified differences in response to PARP targeted therapy are

199 specific for E2F2.

200

201 Discussion

202 Here we describe a role of E2F2 in repair and maintenance of genome integrity in both

203 MMTV-Neu mouse model mammary tumors as well as in human breast cancer patients. We have

204 identified this role using a combination of in silico, in vitro and in vivo datasets. Specifically, we

205 observed that E2F2 controls key members of many different repair pathways including HEJ,

206 NHEJ, BER, and NER. This involvement is associated with a genomically unstable tumors with

207 the loss of E2F2 activity in both the mouse model and the human disease.

208 Furthermore, this finding has direct clinical application. We have identified E2F2 activity

209 as a biomarker for response to cell cycle inhibition therapy. E2F2 activity, as determined by a

210 gene expression signature, correlates with PARP inhibition therapy. Interestingly, although

211 tumors with lowly active E2F2 are significantly more unstable than E2F2 tumors they are resistant

212 to PARP inhibitor therapy. We have identified that this effect is independent of BRCA1/2 status

213 and is correlated with E2F2 levels. This indicates that E2F2 has a role in PARP response and loss

214 of E2F2 may phenocopy or directly trigger other known causes of PARP inhibitor resistance.

215 This manuscript serves as a proof of concept study for an expanded role of the E2Fs in

216 DNA repair in breast cancer. The data presented here, combined with the previously established

217 literature, show key roles of E2F1 and E2F2 as contributing factors in the genomic instability seen

218 in breast cancer. Further research needs to be completed of the other E2F family members to

219 identify if this role is unique to E2F1 and E2F2 or if other E2Fs may play similar roles. Indeed,

220 while certain E2Fs have specific roles in mammary development (37, 38) and tumor biology (3, 4,

221 6), there is significant overlap and compensation amongst the E2Fs (38, 39). Taken

222 together, this data shows the key role that the E2Fs play in cancer progression and heterogeneity.

223 As central drivers of repair and due to the impact they have on the ability of a tumor to respond to

224 key therapies, there must be more research to understand the clinical application of E2F status and

225 the way that it should shape patient care.

226 Materials and Methods

227 Datasets Used

228 For the E2F2 overexpression data, Affymetrix cDNA microarray profiled gene expression data

229 was downloaded from previously published data (40). For E2F2 binding analysis we utilized

230 ChIP-Chip for data for E2F2 T lymphocytes isolated from 4-week-old C57B16:129SV mice (33).

231 For the E2F2 knockout data in the MMTV-Neu background we utilized the dataset GSE42533.

232 For the human patient analysis, we utilized the TCGA breast cancer cohort (41).

233

234 Overrepresentation analysis

235 All overrepresentation analysis experiments were performed through the use of GATHER (42)

236 and PANTHER bioinformatic analysis to identify overrepresented gene ontology groups.

237 Significant groups were noted filtered by a p value of less than .05 and a Bayes factor greater than

238 3.

239

240 ssGSEA and Clustering analysis

241 ssGSEA was performed on the Broad genepattern software with the designated genesets for each

242 repair pathway downloaded from msig-db. These scores were normalized between 0 (low) and 1

243 (high). Samples were clustered through the use of Morpheus

244 (https://software.broadinstitute.org/morpheus/). Clustering was performed using Complete linkage

245 unsupervised hierarchical clustering. Heatmaps were visualized using MATLAB.

246

247 ACE Analysis

248 ACE analysis was performed as previously described to infer copy number alterations from gene

249 expression data (43, 44). The analysis was performed with default parameters and used a

250 significance cutoff of q<.05.

251

252 E2F2 Activity Levels

253 E2F2 activity was assayed using a gene expression signature as previously described (37, 45, 46).

254 Briefly this method identifies differentially expressed genes between Ad-GFP and Ad-E2F2

255 infected cell lines and uses binary regression analysis to compare unknown samples and known

256 controls from each group to get a score between 0 (low E2F2 activity) and 1 (high E2F2 activity).

257

258 E2F2 Drug Sensitivity Data

259 Breast cancer gene expression data was downloaded from the cancer cell line encyclopedia (47).

260 From this data E2F2 activity was predicted as described above. For drug sensitivity data, we

261 downloaded the small molecule sensitivity dataset from CancerRx.org. Breast cancer cell lines

262 were divided into high and low E2F2 activity groups and significantly different compounds

263 between the two groups were identified by significantly different IC50’s as identified by a

264 student’s T-test.

265 Acknowledgements

266 We would like to acknowledge member of the Andrechek lab for their helpful comments and

267 critical reading of the manuscript.

268 Funding

269 This work was supported with NIH R01CA160514 and Worldwide Cancer Research WCR - 14-

270 1153 to E.R.A as well as NIH 1F99CA212221–01 to J.P.R

271 Availability of data and material

272 All datasets used in this publication can be accessed on Gene Expression Omnibus through their

273 appropriate GSE number as noted in the manuscript.

274 Competing interests

275 The authors declare no competing interests

276 Figure Legends

277 Figure 1: E2F2 target genes are enriched for DNA repair associated proteins

278 To understand genes regulated by E2F2 we utilized E2F2 overexpression data to compare the

279 expression profile of HMECs infected with Ad-GFP and Ad-E2F2 (A). A graphic representation

280 of the first three principle components reveals a consistent transcriptional response associated with

281 E2F2 overexpression through displaying MEFs plus Ad-GFP (blue) and MEFs plus Ad-E2F2 (red)

282 (B). Genes overexpressed with the addition of E2F2 as identified by SAM show a significant

283 (FDR<.05) overrepresentation in key gene groups including cell cycle and repair (C). ChIP-CHIP

284 of E2F2 binding genes show binding across the genome (D). The binding targets show a

285 significant (FDR <.05) overrepresentation in a cancer related gene groups (E). A Venn Diagram

286 showing genes predicted to be regulated by E2F2 from ChIP-Chip binding and over expression

287 analysis show a small overlap in genes (F). E2F2 overexpressed or bound genes (E2F2 regulated)

288 are shown to play a role in major repair pathways including Non-Homologous End Joining,

289 Homologous End Joining, Base Excision Repair, and Nucleotide Excision Repair (G).

290

291 Figure 2: Loss of E2F2 is associated with an enrichment of genomic instability markers

292 E2F2 loss is associated with consistent transcriptional responses as revealed by unsupervised

293 hierarchical clustering. MMTV-Neu or MMTV-Neu E2F2 knockout samples are arranged by

294 column and unique genes by row. Gene expression values are represented by color from low (blue)

295 to high (red) as indicated by the color bar. It is revealed that 2 main clusters contain an

296 overrepresentation of MMTV-Neu, E2F2 KO samples (blue) (A). Consistent gene expression sets

297 are enriched with the loss of E2F2 as identified by gene set enrichment analysis. This includes

298 instability gene sets (B) and repair gene sets (C).

299

300

301 Figure 3: E2F2 loss is associated with higher number of copy number alterations in the

302 MMTV-Neu mouse model

303 Copy number analysis obtained through the ACE algorithm for FVB tail DNA (A), MMTV-Neu

304 tumor (B), and MMTV-Neu E2F2 knockout tumor (C) shows increased copy number variants in

305 chromosome 4 in the E2F2 knockout sample compared to the normal DNA and MMTV-Neu

306 tumor. Neighborhood score is graphed along the y-axis and location along the x axis. Significant

307 (p<.05) amplification or deletion points are signified by the neighborhood score above or below

308 the red line. This is shown to be significant across the entire genome with the E2F2 null samples

309 having a significant increase in number of copy number event and the E2F1 null samples having

310 a decrease in the number of copy number alterations (*=p<.05, **=P<.01).

311

312 Figure 4: Low E2F2 activity is associated with decrease repair gene expression in the TCGA

313 breast cancer cohort

314 The TCGA cohort reveals that low E2F2 activity is associated with higher number of genes with

315 copy number alterations (A)

316

317

318

319 Figure 5: Low E2F2 activity is associated with resistance to PARP inhibitors across breast

320 cancer cell lines

321 Breast cancer cell lines divided in lowly active E2F2 and highly active E2F2 show a consistent

322 resistance to PAPRP targeted therapies associated with low E2F2 activity (p<.05) for Talazoparib

323 (A), Olaparib (B), and Rucaparib (C).

324

325

326 References

327 1. Bastien RRL, Rodriguez‐Lescure A, Ebbert MTW, Prat A, Munarriz B, Rowe L, et al. PAM50 Breast 328 Cancer Subtyping by RT‐qPCR and Concordance with Standard Clinical Molecular Markers. Bmc Medical 329 Genomics. 2012;5. 330 2. Perou CM, Sorlie T, Eisen MB, van de Rijn M, Jeffrey SS, Rees CA, et al. Molecular portraits of 331 human breast tumours. Nature. 2000;406(6797):747‐52. 332 3. Andrechek ER. HER2/Neu tumorigenesis and metastasis is regulated by E2F activator transcription 333 factors. Oncogene. 2015;34(2):217‐25. 334 4. Fujiwara K, Yuwanita I, Hollern DP, Andrechek ER. Prediction and Genetic Demonstration of a Role 335 for Activator E2Fs in Myc‐Induced Tumors. Cancer Res. 2011;71(5):1924‐32. 336 5. Ho GH, Calvano JE, Bisogna M, Van Zee KJ. Expression of E2F‐1 and E2F‐4 is reduced in primary 337 and metastatic breast carcinomas. Breast Cancer Res Treat. 2001;69(2):115‐22. 338 6. Hollern DP, Honeysett J, Cardiff RD, Andrechek ER. The E2F transcription factors regulate tumor 339 development and metastasis in a mouse model of metastatic breast cancer. Mol Cell Biol. 2014. 340 7. Worku D, Jouhra F, Jiang GW, Patani N, Newbold RF, Mokbel K. Evidence of a tumour suppressive 341 function of E2F1 gene in human breast cancer. Anticancer Res. 2008;28(4B):2135‐9. 342 8. Nevins JR. The Rb/E2F pathway and cancer. Hum Mol Genet. 2001;10(7):699‐703. 343 9. Trimarchi JM, Lees JA. Sibling rivalry in the E2F family. Nat Rev Mol Cell Biol. 2002;3(1):11‐20. 344 10. Van Den Heuvel S, Dyson NJ. Conserved functions of the pRB and E2F families. Nature reviews 345 Molecular cell biology. 2008;9(9):713‐24. 346 11. Freedman JA, Chang JT, Jakoi L, Nevins JR. A combinatorial mechanism for determining the 347 specificity of E2F activation and repression. Oncogene. 2009;28(32):2873‐81. 348 12. Rabinovich A, Jin VX, Rabinovich R, Xu X, Farnham PJ. E2F in vivo binding specificity: comparison 349 of consensus versus nonconsensus binding sites. Genome Res. 2008;18(11):1763‐77. 350 13. Zheng N, Fraenkel E, Pabo CO, Pavletich NP. Structural basis of DNA recognition by the 351 heterodimeric cell cycle transcription factor E2F‐DP. Genes Dev. 1999;13(6):666‐74. 352 14. Chong JL, Wenzel PL, Saenz‐Robles MT, Nair V, Ferrey A, Hagan JP, et al. E2f1‐3 switch from 353 activators in progenitor cells to repressors in differentiating cells. Nature. 2009;462(7275):930‐4. 354 15. Sears RC, Nevins JR. Signaling networks that link cell proliferation and cell fate. J Biol Chem. 355 2002;277(14):11617‐20. 356 16. Lee RJ, Albanese C, Fu M, D'Amico M, Lin B, Watanabe G, et al. Cyclin D1 is required for 357 transformation by activated Neu and is induced through an E2F‐dependent signaling pathway. Mol Cell 358 Biol. 2000;20(2):672‐83. 359 17. Chen HZ, Tsai SY, Leone G. Emerging roles of E2Fs in cancer: an exit from cell cycle control. Nat 360 Rev Cancer. 2009;9(11):785‐97. 361 18. Weijts B, Westendorp B, Hien BT, Martinez‐Lopez LM, Zijp M, Thurlings I, et al. Atypical E2Fs inhibit 362 tumor angiogenesis. Oncogene. 2018;37(2):271‐6. 363 19. Fajas L. Metabolic control in cancer cells. Ann Endocrinol (Paris). 2013;74(2):71‐3. 364 20. Phillips AC, Vousden KH. E2F‐1 induced apoptosis. Apoptosis. 2001;6(3):173‐82. 365 21. Hallstrom TC, Mori S, Nevins JR. An E2F1‐dependent gene expression program that determines 366 the balance between proliferation and cell death. Cancer Cell. 2008;13(1):11‐22. 367 22. Rennhack J, Andrechek E. Conserved E2F mediated metastasis in mouse models of breast cancer 368 and HER2 positive patients. Oncoscience. 2015;2(10):867‐71. 369 23. Yuwanita I, Barnes D, Monterey MD, O'Reilly S, Andrechek ER. Increased metastasis with loss of 370 E2F2 in Myc‐driven tumors. Oncotarget. 2015;6(35):38210‐24. 371 24. Wu L, de Bruin A, Wang H, Simmons T, Cleghorn W, Goldenberg LE, et al. Selective roles of E2Fs 372 for ErbB2‐ and Myc‐mediated mammary tumorigenesis. Oncogene. 2015;34(1):119‐28.

373 25. Biswas AK, Johnson DG. Transcriptional and nontranscriptional functions of E2F1 in response to 374 DNA damage. Cancer Res. 2012;72(1):13‐7. 375 26. Lin WC, Lin FT, Nevins JR. Selective induction of E2F1 in response to DNA damage, mediated by 376 ATM‐dependent phosphorylation. Genes Dev. 2001;15(14):1833‐44. 377 27. Liu K, Lin FT, Ruppert JM, Lin WC. Regulation of E2F1 by BRCT domain‐containing protein TopBP1. 378 Mol Cell Biol. 2003;23(9):3287‐304. 379 28. Guo R, Chen J, Zhu F, Biswas AK, Berton TR, Mitchell DL, et al. E2F1 localizes to sites of UV‐induced 380 DNA damage to enhance nucleotide excision repair. J Biol Chem. 2010;285(25):19308‐15. 381 29. Chen J, Zhu F, Weaks RL, Biswas AK, Guo R, Li Y, et al. E2F1 promotes the recruitment of DNA 382 repair factors to sites of DNA double‐strand breaks. Cell Cycle. 2011;10(8):1287‐94. 383 30. Castillo DS, Campalans A, Belluscio LM, Carcagno AL, Radicella JP, Canepa ET, et al. E2F1 and E2F2 384 induction in response to DNA damage preserves genomic stability in neuronal cells. Cell Cycle. 385 2015;14(8):1300‐14. 386 31. Lee MY, Moreno CS, Saavedra HI. E2F activators signal and maintain centrosome amplification in 387 breast cancer cells. Mol Cell Biol. 2014;34(14):2581‐99. 388 32. Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing 389 radiation response. Proc Natl Acad Sci U S A. 2001;98(9):5116‐21. 390 33. Laresgoiti U, Apraiz A, Olea M, Mitxelena J, Osinalde N, Rodriguez JA, et al. E2F2 and CREB 391 cooperatively regulate transcriptional activity of cell cycle genes. Nucleic Acids Res. 2013;41(22):10185‐ 392 98. 393 34. Guy CT, Webster MA, Schaller M, Parsons TJ, Cardiff RD, Muller WJ. Expression of the neu 394 protooncogene in the mammary epithelium of transgenic mice induces metastatic disease. Proc Natl Acad 395 Sci U S A. 1992;89(22):10578‐82. 396 35. Hu G, Chong RA, Yang Q, Wei Y, Blanco MA, Li F, et al. MTDH activation by 8q22 genomic gain 397 promotes chemoresistance and metastasis of poor‐prognosis breast cancer. Cancer Cell. 2009;15(1):9‐20. 398 36. Barbie DA, Tamayo P, Boehm JS, Kim SY, Moody SE, Dunn IF, et al. Systematic RNA interference 399 reveals that oncogenic KRAS‐driven cancers require TBK1. Nature. 2009;462(7269):108‐12. 400 37. Andrechek ER, Mori S, Rempel RE, Chang JT, Nevins JR. Patterns of cell signaling pathway 401 activation that characterize mammary development. Development. 2008;135(14):2403‐13. 402 38. To B, Andrechek ER. Transcription factor compensation during mammary gland development in 403 E2F knockout mice. PLoS One. 2018;13(4):e0194937. 404 39. Kong LJ, Chang JT, Bild AH, Nevins JR. Compensation and specificity of function within the E2F 405 family. Oncogene. 2007;26(3):321‐7. 406 40. Huang E, Ishida S, Pittman J, Dressman H, Bild A, Kloos M, et al. Gene expression phenotypic 407 models that predict the activity of oncogenic pathways. Nat Genet. 2003;34(2):226‐30. 408 41. Network TCGA. Comprehensive molecular portraits of human breast tumours. Nature. 409 2012;490(7418):61‐70. 410 42. Chang JT, Nevins JR. GATHER: a systems approach to interpreting genomic signatures. 411 Bioinformatics. 2006;22(23):2926‐33. 412 43. Hu G, Wei Y, Kang Y. The multifaceted role of MTDH/AEG‐1 in cancer progression. Clin Cancer Res. 413 2009;15(18):5615‐20. 414 44. Rennhack J, To B, Wermuth H, Andrechek ER. Mouse Models of Breast Cancer Share Amplification 415 and Deletion Events with Human Breast Cancer. J Mammary Gland Biol Neoplasia. 2017. 416 45. Gatza ML, Lucas JE, Barry WT, Kim JW, Wang Q, Crawford MD, et al. A pathway‐based classification 417 of human breast cancer. Proc Natl Acad Sci U S A. 2010;107(15):6994‐9. 418 46. Bild AH, Yao G, Chang JT, Wang Q, Potti A, Chasse D, et al. Oncogenic pathway signatures in human 419 cancers as a guide to targeted therapies. Nature. 2006;439(7074):353‐7.

420 47. Barretina J, Caponigro G, Stransky N, Venkatesan K, Margolin AA, Kim S, et al. The Cancer Cell Line 421 Encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature. 2012;483(7391):603‐7.

422

423

424 Supplemental Legends

425 Supplemental Figure 1

426 Copy number analysis obtained through the ACE algorithm for MMTV-Neu tumors, and MMTV-

427 Neu E2F1 and E2F2 knockout tumors shows increased copy number variants in the E2F2 knockout

428 sample compared to the normal DNA and MMTV-Neu tumors and decreased variants in the E2F1

429 knockout background. This is shown to be significant across the entire genome with the E2F2 null

430 samples having a significant increase in number of copy number event and the E2F1 null samples

431 having a decrease in the number of copy number alterations (*=p<.05, **=P<.01).

432

433 Supplemental Figure 2

434 Dotplot representation of figure 4, split by major repair pathway. ssGSEA shows low E2F2

435 activity is associated with lower gene expression enrichment in a number of different repair

436 pathways including Base Excision Repair (A), Nucleotide Excision Repair (B), and Homologous

437 End Joining (C), Non-Homologous End Joining (D).

438

439 Supplemental Figure 3

440 Breast cancer cell lines divided in lowly active E2F1 and highly active E2F1 show no consistent

441 resistance to PAPRP targeted therapies associated with low E2F1 activity for Olaparib (A),

442 Rucaparib (B) or Talazoparib (C).

443

444

445 Supplemental Table 1

446 Table One includes a list of of E2F2 bound or co-expressed genes included in each of the major

447 repair pathways.

448

449 Supplemental Table 2

450 Table Two includes the IC50 for various breast cancer cell lines ordered by E2F2 pathway

451 signature activity. The last row shows significance through a t-test for low (<0.5) and high (>0.5)

452 E2F2 predictions. Significant compounds (p<0.05) are shown in bold.

453

Rennhack - Figure 1

A D Ad-GFP Ad-E2F2

HMECs HMECs E2F2 E2F2 ChIP-chip Laresgoiti NAR 2013

B Wild Type E2F2 Overexpression E cell cycle DNA metabolic process

DNA repair

DNA-dependent DNA replication

double-strand break repair

Regulation of biosyn process

0 1 2 3 4 5 6 7 8 9 10 11 12 13 -Log(FDR) C F cell cycle E2F2 Regulated E2F2 Bound cell division sister chromatid segregation DNA unwinding 17 double-strand break repair 39817 760 antigen processing non-recombinational repair

0 1 2 3 4 5 6 7 8 9 10 11 12 13 -Log(FDR)

G NHEJ 7

E2F2 Regulated 3 0 0 1161 0 1 2 0 16 5 1 0 HEJ 2 0 1 0 2 0 0 1 0 0 0 3 4 0 0 NER 28 5 18

BER A Rennhack -FigureRennhack 2 B bioRxiv preprint

1132_Neu doi:

certified bypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. 1753_Neu https://doi.org/10.1101/777870 1357_Neu 2187_Neu 2182_Neu 1414_Neu 2184_Neu 2KO 2188_Neu 2191_Neu ; this versionpostedOctober3,2019. 1007_Neu 2KO 1960_Neu 2KO 2297_Neu C 1295_Neu 2KO 2491_Neu 2KO 2236_Neu 2KO 2366_Neu 2KO 1358_Neu 1363_Neu The copyrightholderforthispreprint(whichwasnot 1894_Neu 1906_Neu 2178_Neu 1183_Neu 2KO 2306_Neu 2KO 1131_Neu 2183_Neu 2KO 2241_Neu 2KO 2307_Neu 2KO 2186_Neu 2KO 2494_Neu 2KO 2501_Neu 2KO bioRxiv preprint doi: https://doi.org/10.1101/777870; this version posted October 3, 2019. The copyright holder for this preprint (which was not Rennhack - Figure 3 certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. A FVB-WT 0.7

0.5

0.3

0.1

-0.1

-0.3

-0.5 Neighborhood Score

-0.7

B MMTV-Neu E2F WT 0.7

0.5

0.3

0.1

-0.1

-0.3

-0.5 Neighborhood Score -0.7

C MMTV-Neu E2F2 KO 0.7

0.5

0.3

0.1

-0.1

-0.3

Neighborhood Score -0.5

-0.7

Chromosome 4 D **

65 * 60 55 50 45 40 35 30 25 20 15

Number of Events 10 5 0 E2F WT E2F1 KO E2F2 KO bioRxiv preprint doi: https://doi.org/10.1101/777870; this version posted October 3, 2019. 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. Rennhack - Figure 4

A B 5000 * 4000

3000

2000 Alterations 1000

Number of Number Copy 0 C Low E2F2 High E2F2

E2F2 Low / High REACTOME - Base Excision Repair GO - Nucleotide Excision Repair REACTOME - Homologous End Joining KEGG - Non Homologous End Joining

E2F2 Probability ssGSEA Enrichment Score

0.00 0.30 0.70 0.98 row min row max bioRxiv preprint doi: https://doi.org/10.1101/777870; this version posted October 3, 2019. 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.

Rennhack - Figure 5

A Talazoparib 6 * 5 4 3 2 1

log (IC50) 0 -1 Low E2F2 High E2F2 -2 -3

B Olaparib 8 7 * 6 5 4 3 log (IC50) 2 1 0 Low E2F2 High E2F2

C Rucaparib 7 6 * 5 4 3

log (IC50) 2 1 0 Low E2F2 High E2F2 bioRxiv preprint doi: https://doi.org/10.1101/777870; this version posted October 3, 2019. 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.

Rennhack - Figure S1 bioRxiv preprint doi: https://doi.org/10.1101/777870; this version posted October 3, 2019. 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.

Rennhack - Figure S2

A Base Excision Repair B Nucleotide Excision Repair 1.00 1.00 * * 0.75 0.75

0.50 0.50 Enrichment 0.25 Enrichment 0.25

0.00 0.00 Low E2F2 High E2F2 Low E2F2 High E2F2

C Homologous End Joining D Non-Homologous end Joining 1.00 1.00 * * 0.75 0.75

0.50 0.50 Enrichment 0.25 Enrichment 0.25

0.00 0.00 Low E2F2 High E2F2 Low E2F2 High E2F2 bioRxiv preprint doi: https://doi.org/10.1101/777870; this version posted October 3, 2019. The copyright holder for this preprint (which was not Rennhack - Figurecertified S3 by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

A Olaparib 600

500

400

300 IC50 200

100

0 Low E2F1 High E2F1

B Rucaparib 350 300 250 200

IC50 150 100 50 0 Low E2F1 High E2F1

C Talazoprib 350 300 250 200

IC50 150 100 50 0 Low E2F1 High E2F1