Human papillomavirus E7 oncoprotein targetsRNF168 to hijack the host DNA damage responseJustine Sitza,b,c, Sophie Anne Blancheta,b,c, Steven F. Gameirod,e,f, Elise Biquanda,b,c, Tia M. Morgang, Maxime Galloya,b,c,Julien Dessapta,b,c, Elise G. Lavoiea, Andréanne Blondeaua, Brandon C. Smithh, Joe S. Mymrykd,e,f, Cary A. Moodyg,h,and Amélie Fradet-Turcottea,b,c,1

aOncology Division, Centre Hospitalier Universitaire (CHU) de Québec-Université Laval Research Center, Québec, QC, Canada G1R 1S3; bUniversité LavalCancer Research Center, Université Laval, Québec, QC, Canada G1V 0A6; cDepartment of Molecular Biology, Medical Biochemistry and Pathology, UniversitéLaval, Québec, QC, Canada G1V 0A6; dDepartment of Microbiology & Immunology, The University of Western Ontario, London, ON, Canada N6A 3K7;eDepartment of Otolaryngology - Head & Neck Surgery, The University of Western Ontario, London, ON, Canada N6A 3K7; fDepartment of Oncology, TheUniversity of Western Ontario, London, ON, Canada N6A 3K7; gDepartment of Microbiology and Immunology, University of North Carolina at Chapel Hill,Chapel Hill, NC 27599; and hLineberger Comprehensive Cancer Center, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599

Edited by Peter M. Howley, Harvard Medical School, Boston, MA, and approved August 14, 2019 (received for review April 9, 2019)

High-risk human papillomaviruses (HR-HPVs) promote cervical can-cer as well as a subset of anogenital and head and neck cancers. Dueto their limited coding capacity, HPVs hijack the host cell’s DNAreplication and repair machineries to replicate their own genomes.How this host–pathogen interaction contributes to genomic insta-bility is unknown. Here, we report that HPV-infected cancer cellsexpress high levels of RNF168, an E3 ubiquitin ligase that is criticalfor proper DNA repair following DNA double-strand breaks, andaccumulate high numbers of 53BP1 nuclear bodies, a marker ofgenomic instability induced by replication stress. We describe amechanism by which HPV E7 subverts the function of RNF168 atDNA double-strand breaks, providing a rationale for increasedhomology-directed recombination in E6/E7-expressing cervical can-cer cells. By targeting a new regulatory domain of RNF168, E7 bindsdirectly to the E3 ligase without affecting its enzymatic activity. AsRNF168 knockdown impairs viral genome amplification in differen-tiated keratinocytes, we propose that E7 hijacks the E3 ligase topromote the viral replicative cycle. This study reveals a mechanismby which tumor viruses reshape the cellular response to DNA dam-age by manipulating RNF168-dependent ubiquitin signaling. Impor-tantly, our findings reveal a pathway by which HPV may promotethe genomic instability that drives oncogenesis.

DNA double-strand break | high-risk human papillomavirus |53BP1 nuclear bodies | E7 protein | RNF168

Oncogenic viruses are estimated to contribute as much as 12%of all cancers worldwide (1, 2), and approximately 5% are

specifically caused by a small group of human papillomaviruses(HPVs) classified as high-risk HPVs (HR-HPVs). These subtypesof HPV belong to the alpha genus and infect basal cells in the oraland genital mucosal epithelia (3). Accordingly, HR-HPVs pro-mote the development of 99% of cervical cancers as well as asubset of anogenital and head and neck cancers, while low-riskHPVs of the same family induce the formation of benign genitalwarts (1, 4). In the past decade, the subset of oropharyngealsquamous cell carcinoma (a subtype of head and neck cancer) thatare HPV-positive has drastically increased, such that the virus isnow detected in 60 to 80% of cases (5–7). Although HPV is re-quired for the development of these cancers, only a small fractionof infected patients will progress to cancer following persistentinfection, indicating that additional mutations are required topromote carcinogenesis.The genome of HPV is a circular double-stranded DNA of

approximately 8 kb, which encodes 7 early proteins, E1, E2, E4,E5, E6, E7, and E8, as well as 2 structural proteins, L1 and L2,that are required for the formation of the capsid (8). In infectedcells, the replication of the viral genome as an extra chromosomalelement relies on the collaboration between the viral proteinsE1 and E2 as well as the DNA replication machinery of the hostcell. Intriguingly, an increasing amount of evidence now places the

host cell DNA repair machinery as an essential player during viralreplication (9, 10). For instance, proteins of the DNA double-strand break (DSB) signaling and repair pathways, such as ataxiatelangiectasia mutated (ATM) protein and its substrate H2AX(γ-H2AX), the MRE11-RAD50-NBS1 (MRN) complex, p53-binding protein 1 (53BP1), the breast cancer susceptibility gene 1(BRCA1), the replication protein A (RPA), and RAD51 (11–14),accumulate in nuclear structures termed “viral replication centers”(10). While the exact mechanism by which the DNA repair pro-teins contribute to viral replication is unknown, small moleculesthat inhibit the kinase activity of ATM, the exonuclease activity ofMRN, or the interaction of RAD51 with DNA abolish viral DNAamplification upon the differentiation of infected keratinocytes,highlighting their role during viral life cycle (12, 15, 16). Consistentwith this, an elegant study by Mehta and Laimins showed that theviral genome accumulates fewer DNA breaks than the host ge-nome during productive replication (17).DSB signaling and repair pathways are central to the main-

tenance of genome integrity and cell viability, as inappropriatelyrepaired DSBs lead to genomic instability and mutations, a hallmark

Significance

Human papillomaviruses (HPVs) cause 5% of all cancers, in-cluding cervical cancers and a growing number of oropharyn-geal cancers, which are reaching epidemic proportion. WhileHPV oncogenes promote the acquisition of cancer hallmarkssuch as sustained proliferation and resistance to cell death bytargeting the p53 and pRb pathways, the mechanism by whichthey induce genomic instability leading to cellular transfor-mation remains largely unknown. This paper describes how theHPV oncoprotein E7 directly impedes the cellular response toDNA double-strand breaks by interacting with a previouslyuncharacterized domain of the E3 ubiquitin ligase RNF168. Asthe function of RNF168 is essential for proper DNA repair, ourfindings reveal a mechanism by which HPV induces genomicinstability and fuels cancer progression.

Author contributions: J.S., C.A.M., and A.F.-T. designed research; J.S., S.A.B., E.B., T.M.M.,M.G., J.D., E.G.L., A.B., B.C.S., and A.F.-T. performed research; J.S., S.A.B., S.F.G., E.G.L.,A.B., J.S.M., and A.F.-T. contributed new reagents/analytic tools; J.S., S.A.B., S.F.G., E.B.,T.M.M., M.G., J.D., E.G.L., A.B., B.C.S., J.S.M., C.A.M., and A.F.-T. analyzed data; and A.F.-T.wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.1To whom correspondence may be addressed. Email: amelie.fradet-turcotte@crchudequebec.ulaval.ca.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1906102116/-/DCSupplemental.

First published September 9, 2019.

19552–19562 | PNAS | September 24, 2019 | vol. 116 | no. 39 www.pnas.org/cgi/doi/10.1073/pnas.1906102116

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

26, 2

020

of cancer (18). Furthermore, genomic instability can lead to theacquisition of many of the other hallmarks of cancer, including re-sistance to programmed cell death, proliferation, and dissemination.Ultimately, if cells are unable to repair DSBs, apoptosis or senes-cence is triggered (19). Consequently, cells have evolved a highlyorchestrated response that enables the detection, signaling, and re-pair of breaks. This process is regulated in a cell cycle-dependentmanner and activates 2 main types of DNA repair, nonhomologousend-joining (NHEJ) and homology-directed recombination repair(HDR) (20). In the G1 phase of the cell cycle, cells rely on canonicalNHEJ to religate DNA ends after limited end processing, whileHDR is activated in the S/G2 phases to promote repair that isguided by the template created through newly replicated DNA. Thelatter process is driven by extensive 5′ to 3′ end resection at thebreak. Following DSBs, recruitment of DNA repair proteins isinitiated locally by ATM-dependent phosphorylation and co-ordinated by ubiquitylation of chromatin surrounding the break(21, 22). The latter step is accomplished by the sequential re-cruitment of the E3 ligases RING finger protein 8 (RNF8) andRNF168, a process that culminates in the ubiquitylation of the N-terminal tail of H2A by RNF168 (23). This histone mark thenamplifies the recruitment of RNF168 and also drives the re-cruitment of downstream effectors such as 53BP1 and RNF169, aparalog of RNF168 that counteracts the accumulation of 53BP1 atDSBs (24–28). By counteracting the recruitment of BRCA1 atDSBs, the recruitment of 53BP1 and its downstream effectors(RIF1 and the shieldin complex) promote DNA repair pathway byNHEJ by inhibiting resection of DNA ends (20, 29). RNF168plays an essential role in the maintenance of genomic integrity, asknockout mice are predisposed to cancer development (30), andpatients who lack functional RNF168 suffer from RIDDLE syn-drome, a disease associated with immunodeficiency and radio-sensitivity (31).In HPV-positive (HPV+) primary tumors, RNF168 and 53BP1

accumulate in abnormally large foci, supporting the hypothesisthat one or more HPV-gene products subverts the DSB signalingand repair pathway in these cells (32). Here, we show that theexpression of RNF168 is specifically increased in HPV+ cancersand that HPV-transformed cell lines accumulate large foci of53BP1 in the G1 phase of the cell cycle (53BP1 nuclear bodies;53BP1-NBs), a marker of DNA damage that occurred in theprevious S-phase. To gain insight into the molecular events thatpromote genomic instability in HPV+ cells, we sought to deter-mine the impact of viral proteins on DSB signaling. Our resultsunexpectedly revealed that the E7 oncoprotein from HR-HPVdirectly targets a previously uncharacterized domain of RNF168(herein termed the “E7-Binding Domain”; E7BD). We showedthat the C-terminal domain of E7 mediates this interaction in-dependently of DSB signaling, and that RNF168 is essential topromote viral genome amplification in differentiating keratino-cytes. Importantly, we found that E7 hinders the function of theE3 ligase at DSBs, revealing a mechanism by which a viraloncoprotein induces genomic instability.

ResultsRNF168 Is Expressed at High Levels in HPV+ Cancers. As HPV+ tu-mors exhibit large nuclear foci of RNF168 and 53BP1, we firstsought to determine if the levels of these proteins are specificallyincreased in these tumors. Strikingly, analysis of RNF168 mRNAexpression in cancer samples from the cBioPortal database (33,34) revealed that cervical and head and neck are among the fewtypes of cancers in which the E3 ligase is overexpressed (SIAppendix, Fig. S1A). Next, we analyzed the Illumina HiSeqRNA-seq data from The Cancer Genome Atlas (TCGA) for297 cervical carcinoma (CESC; 278 HPV+ and 19 HPV−) and558 head and neck squamous cell carcinoma (HNSC; 73 HPV+,442 HPV−, and 43 normal-adjacent tissues) cohorts for expres-sion of RNF168 and 53BP1 (Fig. 1A). For comparison, the ex-

pression levels of 2 other ubiquitin ligases that also accumulateon chromatin following DSBs, RNF8 and RNF169, were ana-lyzed. Strikingly, both types of HPV-related cancers were foundto express high levels of RNF168 mRNA, as shown by the nor-malized RNA-seq absolute read count values. This increase ishighly significant when comparing HPV+ and HPV− samplesfrom the HNSC cohort (P ≤ 0.0001) and is observed indepen-dently of the HR-HPV type detected in these samples (Fig. 1Aand SI Appendix, Fig. S1B). In samples from the HNSC cohort,expression levels of 53BP1 were also significantly increased, al-though this was not the case in the CESC cohort. In contrast toRNF168 and 53BP1, the expression levels for RNF8 and RNF169were not increased in HPV+ samples. In fact, RNF8 expressionwas markedly reduced in HPV+ samples compared with HPV−

tumors, indicating that the increased expression of RNF168 doesnot reflect a general up-regulation of DNA repair genes in thesecancers. Consistent with these results, analysis of a panel of HPV+

cell lines isolated from cervical (SiHa, HeLa, CaSki) and head andneck carcinomas (UM-SCC-47; hereafter referred to as SCC47)revealed that the levels of RNF168 are markedly increased inHPV+ cell lines compared with HPV− CESC (C33-A) and HNSC(UM-SCC-1; referred to hereafter as SCC1) cell lines (Fig. 1B andSI Appendix, Fig. S1 C and D). The later observation suggestedthat viral proteins encoded by these integrated genomes lead toincreased expression of RNF168 in HPV+ cells.When analyzed by immunochemistry, parafilm-embedded HPV+

tumor tissues were found to accumulate large foci of RNF168 and53BP1 (32), a phenotype that is reminiscent of the formation of53BP1-NBs, a marker of genomic instability induced by replicationstress. These nuclear structures shield DNA ends from furtherprocessing until they reach the next late S phase, in which they willbe repaired by RAD52-mediated repair (35–38). To determinewhether the appearance of these foci is related to the HPV statusof these cells, we analyzed the localization of RNF168 and 53BP1in the panel of HPV+ and HPV− cells described here earlier byimmunofluorescence microscopy. In these experiments, we foundthat the number of large 53BP1 bodies was significantly increasedin HPV+ CESC and HNSC cell lines and directly correlated to thenumber of integrated viral genomes (Fig. 1 C and D and SI Ap-pendix, Fig. S1E). These foci were present only during the G1phase of the cell cycle (marked by the absence of cyclin A staining;Fig. 1E and SI Appendix, Fig. S1F) and were abolished by de-pletion of RNF168 (SI Appendix, Fig. S1 G–I), indicating that theycorrespond to 53BP1-NBs. Consistently, we found that BRCA1,but not the DNA repair factor RAD51, colocalizes with 53BP1 inthese bodies (Fig. 1E and SI Appendix, Fig. S1J). Taken together,these results indicate that increased expression of RNF168 andaccumulation of genomic instability are associated with the pres-ence of HPV in human cancer cells.

HPV E7 Proteins Directly Interact with RNF168. Many DSB signalingand repair proteins accumulate at viral replication centers duringthe viral replication cycle (10). Combined with the results describedhere earlier, this raises the possibility that RNF168-mediatedsignaling is directly subverted by one or more viral gene products.To explore this possibility, we used a LacO/LacR system that re-capitulates the recruitment of DSB repair factors downstream ofγ-H2AX in the absence of DNA damage (Fig. 2A) (24, 39).Specifically, we investigated if the viral replication proteins E1 andE2, E4, or the oncoproteins E6 and E7 from the prevalent HR-HPV31 could promote the recruitment of DSB-signaling proteinsat the LacO-array, visible as a single nuclear focus in this cell-based assay. E5 and L1/L2 proteins were not tested, as they are,respectively, a transmembrane protein and capsid structural pro-teins only expressed in terminally differentiated keratinocytes (4).In this assay, mCherry-LacR fused to the C-terminal DNAcleavage domain of the endonuclease FOKI was used as positivecontrol to monitor the recruitment of DSB-signaling factor at the

Sitz et al. PNAS | September 24, 2019 | vol. 116 | no. 39 | 19553

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

26, 2

020

LacO-array. Surprisingly, we found that mCherry-LacR-E6 and-E7 promoted a low but significant accumulation of 53BP1 at theLacO-array (Fig. 2 B and C and SI Appendix, Fig. S2 A and G).The absence of γ-H2AX at the array 6 h posttransfection showsthat E6 and E7 recruit 53BP1 independently of DSB signaling inthese conditions, unlike E2, which triggers the accumulation ofthe histone mark at the array, potentially through its ability tointeract with TOPBP1, a factor that promotes ATR activation (40)(Fig. 2D and SI Appendix, Fig. S2B). Consistent with an interactionbetween the viral proteins and RNF168, we found that FLAG-RNF168, but not FLAG-RNF8, colocalized with the viral onco-proteins at the array in as many as 80% of the cells (Fig. 2 E and Fand SI Appendix, Figs. S2 C, D, and H). Furthermore, while de-pletion of RNF8 drastically reduced the recruitment of 53BP1 toDSBs induced by targeting the nuclease FOKI at the LacO-array,it increased the amount of 53BP1 colocalizing with mCherry-LacR-E6 and -E7 (Fig. 2G and SI Appendix, Fig. S2 E and J). AsRNF168 is a limiting factor during DSB signaling (32), it is pos-

sible that the depletion of RNF8 increases the amount of RNF168available for recruitment to the array by E6 and E7.HPV31 E7 and E6 interact with histone modifiers of the his-

tone acetyl transferase (HAT) (41, 42) and deacetylase (HDAC)(43, 44) families. To exclude that E6 and E7 indirectly promotethe recruitment of RNF168 to the array by inducing local epige-netic changes in chromatin, we investigated if mCherry-LacR-RNF168 also promoted the accumulation of the viral oncoproteinsat the LacO-array. Under these conditions, only E7 was efficientlyrecruited to the array (Fig. 2H and SI Appendix, Fig. S2 F and I).Therefore, we decided to focus further analyses on the E7-mediated interaction. Using mCherry-LacR-E7, we also detectedthe accumulation of ubiquitin conjugates, as well as BRCA1, at theLacO-array (SI Appendix, Fig. S2 K–M), consistent with RNF168being recruited to the array by HPV31 E7. However, althoughRNF169 is a paralog of RNF168, it is not recruited directly to thearray by E7, as depletion of RNF168 prevents its recruitment (SIAppendix, Fig. S2N).

Fig. 1. Expression and localization of RNF168 and 53BP1 in HPV+ carcinoma. (A) Expression of the indicated genes in cervical cancers (CESC) and head andneck cancers (HNSC) stratified by human papillomavirus (HPV) status. Normalized RNA-seq data extracted from The Cancer Genome Atlas (TCGA) databaseare presented. For HNSC, normal-adjacent control tissues were used as an extra control (black). Statistical significance was assessed by a Mann–Whitney Utest. (B) Whole cell extracts (WCE) were analyzed by immunoblotting with the indicated antibodies. KAP1 and Tubulin were used as loading controls. (C)Representative images of 53BP1 and RNF168 foci observed in the indicated cell lines. Cells were fixed and processed for γ-H2AX, 53BP1, andRNF168 immunofluorescence. (Scale bar, 5 μm.) (D) Quantification of 53BP1 (Upper) and RNF168 (Lower) colocalizing with γ-H2AX. Data are presented as themean ± SEM (n = 3; *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001). ns, not significant. (E) Representative images of GFP-53BP1FFR and BRCA1 in CaSkicells treated or not with 1 Gy. Cyclin A staining was used to identify cells in S/G2 phase (positive) and G1 phase (negative).

19554 | www.pnas.org/cgi/doi/10.1073/pnas.1906102116 Sitz et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

26, 2

020

Finally, the interactions detected using the LacO-LacR teth-ering system could be indirect. To rule out this possibility, weshowed that recombinant GST-HPV16 E7 protein efficientlypulled down recombinant His-RNF168 in vitro (Fig. 2I and SIAppendix, Fig. S2O), indicating that the interaction is direct andindependent of posttranslational modifications. Taken together,these findings revealed that the HPV E7 protein specificallyinteracts with the E3 ligase RNF168 independently of the acti-vation of the cellular response to DNA damage.

RNF168 Interacts with E7 from HR-HPVs and Is Required for ViralReplication. Human papillomavirus E7 is essential for viral rep-lication and HPV-induced carcinogenic transformation. E7 is asmall protein of approximately 100 amino acids that is composedof 3 conserved regions (CR1–3; Fig. 3A). Mutations in any ofthese regions affect different biological activities of E7 (45). Togain insight into the role of the interaction between E7 andRNF168, we first investigated if this interaction is conservedamong the E7s from different types of virus and which domain ofE7 contributes to the interaction. According to their oncogenicpotential, HPVs of the alpha genus are subdivided as low- andhigh-risk (4). HPV6 and 11 are classified as low-risk, as they aremainly associated with the emergence of benign genital warts.HPV16, 18, and 31 are part of a group of 12 HPV types classifiedas high-risk because of their association with cancer development(4). Interestingly, we observed that only the HPV E7 proteinsfrom high-risk viruses promoted efficient recruitment of RNF168to the LacO-array (Fig. 3B and SI Appendix, Fig. S3 A and B).These findings show that the interaction of E7 with RNF168 is

conserved only among the alpha HR-HPV types. Surprisingly,we did not detect an interaction between RNF168 and E7 fromHPV5 and HPV8, 2 types of virus that have been associated withthe development of nonmelanoma skin cancers (46). In the LacO/LacR system, neither CR1 nor CR2, which comprises the pRb-binding motif (LXCXE), were sufficient to recapitulate the in-teraction with RNF168 (Fig. 3C and SI Appendix, Fig. S3 C andD). Accordingly, a point mutation or a small deletion that abol-ishes the interaction of E7 with the tumor suppressor pRb (C24Gand DLYC) did not severely impact the recruitment ofRNF168 by full-length E7 to the array (Fig. 3D and SI Appendix,Fig. S3 E and F). In contrast to CR1 and CR2, CR3 by itself wasable to recruit RNF168 to the array, similarly to full-length E7(Fig. 3C and SI Appendix, Fig. S3 C and D). This domain was alsorequired for the interaction of E7 with RNF168 in immunopre-cipitation experiments using cell extracts (Fig. 3E). CR3 is anatypical zinc finger-like domain that mediates the dimerization ofE7 and is needed for its transforming activity (45). Using a panelof well described HPV16 E7 CR3 mutants (45, 47), we found thatthe recruitment of RNF168 to mCherry-LacR-E7 foci was onlyslightly impaired by a mutant that abolishes the interaction ofE7 with HDACs (L67R; Fig. 3D and SI Appendix, Fig. S3 E and F).Interestingly, mutants of E7 that are transformation-defective andfail to interact with the cullin 2 ubiquitin ligase complex (CVQ,LEDLL) also significantly impair its interaction with RNF168.As E7 is required for viral replication, we next investigated

whether RNF168 is required for the replication of viral genomes.In CIN612 9E cells, which are derived from a biopsy of a low-grade cervical intraepithelial neoplasia (CIN1) and stably maintain

Fig. 2. E7 protein directly interacts with RNF168. (A) Schematic representation of the tethering LacO/LacR system inserted in U2OS 2-6-5. LacO, Lac operator;LacR, Lac repressor. (B) U2OS 2-6-5 cells transfected with plasmids expressing the indicated mCherry-LacR fusion protein or induced for the expression ofER-mCherry-LacR-FokI-DD were fixed and processed for 53BP1 immunofluorescence. (Scale bars, 5 μm.) (C–H) Quantification of the mCherry-LacR focicolocalizing with 53BP1 (C and G), γ-H2AX (D), and Flag (E and F) staining. In G, cells were depleted for RNF8 prior to transfection with mCherry-LacRconstructs. In H, HPV31 E6 or E7 fused to a GFP tag were used. All of the quantifications are represented as the mean ± SD (n = 3; *P ≤ 0.05; **P ≤ 0.01; ***P ≤0.001; ****P ≤ 0.0001). (I) Pull-down of His-RNF168 with GST-HPV16 E7. GST alone and MBP-His were used as negative controls. Inputs and pull-down wereanalyzed by immunoblotting with the indicated antibodies.

Sitz et al. PNAS | September 24, 2019 | vol. 116 | no. 39 | 19555

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

26, 2

020

multiple copies of the HPV31b genome in episomal form, de-pletion of RNF168 for 3 d had no impact on the amount of viralgenome that is maintained in undifferentiated cells (shRNF168 vs.shScramble at T0; Fig. 4 A and B). Interestingly, Southern blotanalysis of viral genome amplification, which occurs when CIN6129E cells are induced to differentiate in high-calcium medium for72 h, showed that RNF168 depletion substantially decreased thenumber of viral episomes compared with shScram (Fig. 4 A andB). In these experiments, the knockdown of RNF168 did not in-terfere with the ability of CIN612 9E cells to differentiate, as in-dicated by the increased levels of involucrin and keratin-10, 2specific markers for differentiation. Consistent with the increasedlevels of RNF168 detected in HPV+ tumors, we also detected

higher level of RNF168 in CIN612 9E compared with uninfectedkeratinocytes (Fig. 4C). Altogether, these results indicate that theexpression of RNF168 E3 ligase is increased in HPV-infected cellsand is required for productive viral replication upon differentia-tion. This phenotype is reminiscent of results obtained with knock-down or inhibition of ATM, MRE11, NBS1, RAD51, and BRCA1(11, 12, 15, 16).

E7 Favors DNA Repair by HDR by Limiting the Function of RNF168 atDSBs. Although the interaction of E7 with RNF168 can be easilydetected at the LacO-array, the pan-nuclear localization of theviral protein observed in untreated cells remained unchanged incells in which DNA lesions were induced by irradiation (IR),neocarzinostatin, or hydroxyurea, which induces lesions specifi-cally in S-phase by blocking replication fork progression (SI Ap-pendix, Fig. S4A). Similarly, E7 is not recruited to sites of intenseDNA damage induced by targeting the FOKI nuclease to theLacO-array (Fig. 2H and SI Appendix, Fig. S2F). However, HPV16E7 was recently reported to colocalize with 53BP1 using proximity-ligation assays, both in differentiated HPV+ keratinocytes and inCaSki (48), arguing that both proteins are near each other duringviral replication and cellular transformation. Consistently, we foundthat E7 and the CR3 domain alone are recruited to foci of RNF168

Fig. 3. E7 from HR-HPVs binds to RNF168 through the CR3 domain. (A)Schematic representation of HPV E7 protein. CR, conserved region; CXXC,zinc-finger motif; LXCXE, pRb-binding sites. An alignment of the CR3 do-main from different types of HPV (beta, yellow; alpha low-risk, green; alphahigh-risk, pink) is also presented. Conserved residues are indicated in bold,the zinc-binding domains are indicated in dark red, and the residues mu-tated used in this study are shown in light gray. (B–D) Quantification of themCherry-LacR-HPV E7 foci colocalizing with Flag-RNF168 were done as de-scribed in Fig. 2E. (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001.) (E) Immunoprecipi-tation of GFP-RNF168 from U2OS 2-6-5 cell extracts. The indicated mCherry-LacR fusion proteins were transfected in GFP- or GFP-RNF168–expressing cells.GFP fusion proteins were purified over GFP-binding proteins beads (GFP-Trap)and analyzed by immunoblotting. GAPDH was used as a loading control.

Fig. 4. RNF168 is required for viral replication. (A) CIN612 9E cells weretransiently transduced with lentiviral particles containing a control scram-bled shRNA (shScram) or one of the 2 RNF168-specific shRNAs for 72 h. DNAand proteins were harvested at T0 (undifferentiated) or following differ-entiation in high-calcium medium for 72 h. DNA was digested with BamHI(does not cut the viral genome) and HindIII (linearizes the viral genome) andanalyzed by Southern blotting using the HPV31 genome as a probe. WCEswere analyzed by immunoblotting with the indicated antibody. Involucrinand K-10 were used to demonstrate cellular differentiation. KAP1 andGAPDH were used as loading controls. (B) Quantification of the experimentpresent in A using densitometry of episomal bands with ImageJ software.Shown is the fold change normalized to T0 shScram, which is set to 1. Errorbars represent means ± SD (n = 3; *P ≤ 0.05). (C) RNF168 expression in HFKand CIN612 9E cells was analyzed 0, 48, and 96 h after differentiation inhigh-calcium media by immunoblotting with the indicated antibodies.GAPDH and KAP1 were used as loading controls.

19556 | www.pnas.org/cgi/doi/10.1073/pnas.1906102116 Sitz et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

26, 2

020

that have been described as cellular replication factories in un-perturbed conditions (Fig. 5A) (49). Thus, these results support amodel in which E7 interacts with a subpopulation of RNF168 thatlocalizes at replication forks.RNF168 is a limiting factor during DSB signaling, as it quickly

becomes saturated upon damage induction (32). This led us toinvestigate if E7 could reduce the amount of RNF168 at DSBs.Although the antibody against RNF168 detects high levels ofDNA damage that accumulate in 53BP1-NBs (Fig. 1 C and D),it is less efficient at detecting the recruitment of RNF168 atirradiation-induced foci (SI Appendix, Fig. S1E; only 30% ofγ-H2AX foci are RNF168-positive vs. 75% and 90% for conjugated-ubiquitin [FK2] and 53BP1 [Fig. 5B], respectively). Therefore, wequantified the formation of 53BP1 and FK2 foci as markers ofRNF168 activity. To do so, U2OS cells stably expressing GFP-HPV31E7 and E7 CR1-CR2 (referred to here as CR1/2) as a negativecontrol were exposed to increasing doses of IR (Fig. 5B and SIAppendix, Fig. S4 B–F). As the accumulation of 53BP1 to DSBsis regulated in a cell cycle manner (20), we performed the ex-periment in U2OS cells in which the impact of E7 on DSBsignaling can be uncoupled from its impact on cell cycle pro-gression (pRb degradation). Indeed, pRB is always phosphorylated(inactive) in U2OS as a result of the absence of pRB dephos-phorylation after mitosis (50). Interestingly, we observed that the

total number of radiation-induced 53BP1 and FK2 foci weresignificantly reduced in cells expressing the viral oncoproteincompared with cells expressing a negative control only (Fig. 5Band SI Appendix, Fig. S4 B, E, and F). Importantly, the levels ofγ-H2AX quantified in cells expressing HPV31 E7 were similar tothose in the control cell lines (Fig. 5B and SI Appendix, Fig. S4 C,E and F). The same results were obtained when GFP-HPV31 E7was expressed in human foreskin keratinocytes expressing thecatalytic component of telomerase (hTERT-HFK; Fig. 5C andSI Appendix, Fig. S4 G and H) (51). Inhibition of RNF168 or53BP1 leads to increased repair by HDR (20). Consistent withthe reduced accumulation of RNF168 substrates at DSBs in E7-expressing cells, we also found that stable expression of E6/E7 inC33-A cell (HPV− cervical carcinoma) increases DNA repair byHDR (Fig. 5 D and E and SI Appendix, Fig. S4 I–M). These ex-periments were performed by measuring gene targeting efficiencyof an mClover donor sequence at the LMNA locus in C33-A,C33-A stably expressing E6 and E7, and C33-A KO 53BP1 aspositive control. Altogether, these observations support a modelin which E7 directly hinders the function of RNF168 at DSBs.

E7 Interacts with RNF168 without Affecting Its E3 Ubiquitin LigaseActivity. RNF168 is an E3 ubiquitin ligase that is composed ofa RING domain (catalytic activity) and 2 ubiquitin recognition

Fig. 5. E7 hinders the function of RNF168 at DSBs. (A) Representative images of U2OS cells transfected with the indicated mCherry-LacR protein and Flag-RNF168. (B and C) Quantification of γ-H2AX and 53BP1 foci in U2OS cells (B) and hTERT-HFK (C) treated with the indicated amount of Gy. In B, FK2 foci werealso quantified. Expression of GFP-HPV31 E7 full-length and CR1/2 were induced 24 h before treatment. Results are represented as the mean ± SEM (n = 3;**P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001). (D) Schematic representation of the HDR assay based on gene editing of the LMNA locus. (E) C33-A stable celllines were transfected with Cas9, sgRNA against LMNA, and mClover donor vectors. Percentage of mClover-positive cells were analyzed by flow cytometry72 h posttransfection and normalized over percentage of mClover-positive C33-A cells in each replicate. Data are represented as the mean ± SEM (n = 3).

Sitz et al. PNAS | September 24, 2019 | vol. 116 | no. 39 | 19557

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

26, 2

020

modules, which enable the recognition of RNF8’s substrate(UDM1) as well as its own substrate (UDM2; Fig. 6A). The latteractivity is important for the autoamplification of RNF168 re-cruitment at DSBs. RNF168 also contains a PALB2-interactingdomain (PID), which promotes its interaction with the complexPALB2/BRCA2 (52, 53). In vitro, RNF168 promotes ubiquitylationof histone H2A (23). As E7 reduces the accumulation of conju-gated ubiquitin at DSBs (Fig. 5B), we investigated if E7 inhibitsthe E3 ligase activity of RNF168. Using in vitro ubiquitylationassay on mononucleosomes, we found that GST-HPV16 E7 doesnot interfere with H2A ubiquitylation by RNF168 (Fig. 6B). In thisassay, along with a ubiquitylation assay that was performed with-out mononucleosomes, we found that GST-HPV16 E7 is notubiquitylated (Fig. 6C and SI Appendix, Fig. S5A), suggesting thatthe viral protein is not a direct target of RNF168. We thus turnedour attention to the interaction between RNF168 and E7 to gaininsight into how the viral oncoprotein subverts the function of thisE3 ubiquitin ligase. We assessed the ability of various mCherry-LacR-RNF168 fusion proteins to promote the recruitment ofGFP-tagged E7 to the LacO-array. These experiments were donein RNF168-depleted cells to eliminate the endogenous protein asa potential confounding factor (24). Consistent with the resultsobtained in GST pull-down, we found that the catalytic activity of

RNF168 was dispensable for its interaction with E7 (Fig. 6 D andE and SI Appendix, Fig. S5 B–D), but necessary for the recruitmentof 53BP1 at the mCherry-LacR locus. Deletion of either one orboth motifs of RNF168 that bind ubiquitin (MIUs; Fig. 6A) hadlittle to no effect on the accumulation of GFP-HPV31 E7 at theLacO-array. Surprisingly, deletion of RNF168 amino acids 251 to383 abolished its interaction with E7 both at the LacO-arrayand at RNF168-positive foci observed in unperturbed cells, buthad no impact on its ability to recruit 53BP1 to the array (Fig. 6D and E and SI Appendix, Fig. S5E). Furthermore, when resi-dues 251 to 383 of RNF168 were fused to mCherry-LacR, thisregion was sufficient to recruit GFP-HPV31 E7 to the LacO-array. These results indicate that the ability of RNF168 to re-cruit 53BP1 and to interact with E7 resides in separate domainsof the protein that can be uncoupled genetically. The necessityof the residues 251 to 383 in the interaction betweenRNF168 and HPV E7 was further confirmed by immunopre-cipitation experiments in a U2OS RNF168 KO cell line (SIAppendix, Fig. S6 A–F). Altogether, our study has revealed apreviously uncharacterized domain of RNF168, the E7-binding domain (E7BD), that is targeted by the oncoproteinE7 to hinder its signaling function at DSBs.

Fig. 6. E7 interacts with a previously uncharacterized region of RNF168. (A) Schematic representation of RNF168 and its functional domains. LRM, LR-motif;PID, PALB2-interacting domain; R, Ring domain; UDM, Ub-dependent DSB recruitment module; UIM/MIU, motif-interacting with ubiquitin. (B) In vitroubiquitylation assay (UbRx) on mononucleosomes (MNs) isolated from HeLa cells with or without a gradient of GST or GST-HPV16 E7. A reaction withoutubiquitin was used as a negative control. (C) In vitro ubiquitylation assay on GST and GST-HPV16 E7 with MNs. A reaction without ubiquitin was used as anegative control. Asterisk indicates a nonspecific band. (D and E) RNF168-depleted U2OS 2-6-5 cells were transfected with plasmids expressing GFP-E7 and theindicated mCherry-LacR RNF168 construct and processed for 53BP1 immunofluorescence. Accumulation of GFP-HPV31 E7 (Left) and 53BP1 (Right) at themCherry-LacR RNF168 foci are represented as the mean ± SD (n = 3) in D (*P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001). Representative images obtained withthe indicated mutation/truncation of RNF168 are presented in E. (Scale bars, 5 μm.) (F) Proposed models for the inhibition of RNF168 function at DSBs by HR-HPV E7. P, phosphorylation; Ub, ubiquitin.

19558 | www.pnas.org/cgi/doi/10.1073/pnas.1906102116 Sitz et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

26, 2

020

DiscussionIn this study, we set out to understand how the HPV oncopro-teins contribute to the overexpression and accumulation ofRNF168 in large nuclear foci observed in HPV-associated tumors.Using a single-cell assay to determine the colocalization of DNArepair factors with mCherry-tagged viral proteins at an integratedLacO-array, we discovered a mechanism by which a host–patho-gen protein interaction promotes genomic instability in the hostcell. First, we discovered that a direct interaction between E7 fromoncogenic HPV types and RNF168 inhibits DSB signaling, raisingthe possibility that this viral–host interaction underlies the acqui-sition of genomic instability characteristic of HPV+ lesions thatprogress to cancer. Second, we found that, in contrast to manyother viruses, HPV relies on RNF168 to promote efficient viralreplication (as detailed later). Third, our work identified a domainof RNF168 bound by E7, termed E7BD, which lies between the2 highly conserved Ub-dependent DSB recruitment modules ofRNF168, and, when targeted by oncogenic HPVs, disrupts hostchromatin response to DNA breaks.Hijacking host ubiquitin machineries is emerging as a common

theme among viruses, and antiviral function of RNF168 has beendescribed for other viruses (reviewed in refs. 10, 54). In the lattercases, viruses have evolved different strategies to inhibit the ac-tivity of RNF168. First, the viral protein ICP0 from herpessimplex virus type 1 (HSV1) targets both RNF8 and RNF168 fordegradation and thus enables productive infection by inhibitingthe intrinsic antiviral response triggered upon infection (55–57).Second, RNF168 was recently found to counteract porcine re-productive and respiratory syndrome virus (PRRSV) infection,although how the virus inhibits RNF168 is still unknown (58).Both the Epstein–Barr virus (EBV) and Kaposi’s sarcoma her-pesvirus (KHSV), which cause Burkitt’s and Hodgkin’s lym-phoma, nasopharyngeal carcinoma, stomach adenocarcinoma,and Kaposi’s sarcoma (2), inhibit the recruitment of RNF168 atDSBs. These effects are attributed to the EBV tegument proteinBKRF4 and the KHSV LANA protein, which bind to chromatinand physically block the interaction of RNF168 with the nucle-osome (59, 60). The consequences of RNF168 inhibition on viralreplication and cancer progression induced by EBV and KHSVremain to be determined. In this study, we described a mecha-nism by which a viral oncoprotein hijacks the function of RNF168through a direct interaction. Specifically, we showed that E7 ac-complishes this task without affecting its enzymatic activity orinducing its degradation (SI Appendix, Fig. S6H). Importantly,and in contrast to previously described antiviral function, RNF168is required for the replication of the HPV genome.While our finding suggests that E7 titers RNF168 away from

DSBs, 2 mechanisms by which E7 subverts the function of thisE3 ubiquitin ligase could be envisioned from our findings. First,E7 may relocate RNF168 to safeguard the stability of viral rep-lication forks within the viral replication centers (Fig. 6F, model1), effectively sequestering this low-abundance protein. Second,the interaction of E7 with RNF168 may serve to rewire theubiquitylome of RNF168 to modify alternative substrates (Fig.6F, model 2). Both models are further discussed as follows.RNF168 was reported to promote efficient DNA replication

by stabilizing replication forks in S phase (49). Thus, E7 mayredirect this E3 ligase to viral replication centers to promote thereplication of its own genome (model 1). Consistent with thishypothesis, a lower number of DNA breaks are detected on theviral genome than on the host genome during productive viralreplication, indicating that DNA breaks are either less frequentin the viral genome or repaired more rapidly (17). Furthermore,RAD51 and BRCA1 are bound to HPV chromatin and are es-sential for productive viral replication, emphasizing the possi-bility that these HDR factors are involved in viral replicationeither through DNA repair by HDR or by protecting stalled

replication forks from degradation (15). In line with these find-ings, RNF168 was recently reported to interact with PALB2 andto promote HDR by facilitating the recruitment of RAD51 ontoresected DNA-ends (52, 53). From this perspective, the inter-action between E7 and RNF168 could facilitate the recruitmentof RAD51 into viral replication centers to stabilize ssDNA gen-erated at replication forks (Fig. 6F, model 1). Furthermore,53BP1 has been shown to enable HDR repair of heterochromatin-DSBs (HC-DSBs) specifically in G2 phase (61). This is particularlyintriguing given that 53BP1 localization to sites of HPV replica-tion increases upon differentiation and that productive viral rep-lication occurs in a G2-arrested environment (62). At HC-DSBs,53BP1 promotes ATM-dependent phosphorylation of the het-erochromatin building factor KAP1, leading to chromatin re-laxation that facilitates the recruitment of HDR factors (63).RNF168 recruitment of 53BP1 may thus serve a similar role onviral chromatin during productive replication. Additional studieswill be required to determine if E7/RNF168 can assemble incomplexes with PALB2/BRCA2 and/or whether the E7–RNF168interaction serves to preferentially recruit DNA repair factors toviral replication centers upon differentiation. Furthermore, char-acterization of the role of the newly identified domain in RNF168on its ability to safeguard genomic stability will be instrumental forour understanding of why it is targeted by HPV E7.It is likely that targeting of RNF168 by E7 plays additional

roles in the HPV life cycle other than those described in this study.We can imagine several reasons why HPV would hijack the functionof this E3 ubiquitin ligase. As suggested in model 2 (Fig. 6F), E7may redirect the activity of the E3 ligase to promote ubiquitylationand subsequent degradation of proteins such as pRb, a tumorsuppressor that is degraded by HR-HPV E7 (64). Consistent withthis hypothesis, RNF168 induces the degradation of other proteinssuch as the histone demethylase KDM4A/JMJD2A (65). Fur-thermore, the interaction of E7 with RNF168 is reminiscent of itsinteraction with cullin 2, an E3 ubiquitin ligase usurped by E7 topromote pRB degradation and stabilize APOBEC3A (66–68). AsE7 also uses distinct domains to interact with pRb and RNF168, itwill be of interest to investigate if all 3 proteins can be part of thesame complex.HPV-positive cancer cells exhibit genomic instability and

replication stress (69). In healthy cells, RNF168 expression ishighly controlled. Its overexpression leads to a massive spreadingof ubiquitin conjugates from the break point and concomitantrecruitment of downstream DNA repair factors such as 53BP1(32). When RNF168 is overexpressed, genomic instability emergesas a result of increased levels of mutagenic NHEJ, a phenotypethat is observed in S-phase when HDR is inhibited by increasedaccumulation of 53BP1 at DSBs (70). While the exact mechanismunderlying the overexpression of RNF168 in HPV+ tumors re-mains to be discovered, we propose that it results, at least in part,from a cellular adaptation to persistent hijacking and sequestra-tion of RNF168 by E7. Consistent with this model, we show thatHPV E7 reduces the numbers of 53BP1 and conjugate-ubiquitinfoci in conditions in which RNF168 is rate-limiting (Fig. 5B). Inthese conditions, any process that reduces the pool of RNF168available to promote DSB repair will have a direct impact ongenomic stability by interfering with the recruitment of 53BP1 andthus DNA repair pathway choice. Consequently, we found thatcervical cancer cells that express E6 and E7 exhibit higher levels ofDNA repair by HDR (Fig. 5E).In conclusion, our study has identified a mechanism by which

HR-HPV E7, through its interaction with RNF168, likely pro-motes genomic instability that drives cancer progression. E7appears to be the main viral driver of cancer progression (2, 45,64), as it is expressed in virtually all HPV-associated cancers andrarely mutated in these cancers compared with other viral geneproducts (71). Therefore, the DNA repair deficiency induced byE7 in HPV-associated cancers may present new synthetic lethal

Sitz et al. PNAS | September 24, 2019 | vol. 116 | no. 39 | 19559

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

26, 2

020

opportunities to target the ∼5% of all human cancers caused bythis important pathogen.

Materials and MethodsCell Culture and Transfections. Cell lines were maintained at 37 °C and 5%CO2. All culture media were supplemented with 10% fetal bovine serum(FBS). U-2-OS (U2OS) and U2OS 2-6-5, gift from Roger Greenberg, Universityof Pennsylvania, Philadelphia, PA (72), cell lines were cultured in McCoy’smedium (Life Technologies). C33-A, CaSki, SiHa, HeLa, SCC1, and SCC47 cellswere purchased from ATCC and cultured in Dulbecco’s modified Eagle’smedium (DMEM; Life Technologies). The media of SCC1 and SCC47 wasalso supplemented with 1% minimum essential media (MEM) nonessentialamino acid (Life Technologies). U2OS 2-6-5, U2OS, and U2OS RNF168 KOinducible cell lines were cultured in McCoy’s medium containing 40 mg/mLG418 and induced with 5 μg/mL doxycycline when indicated. U2OS Flp-inwere cultured in McCoy’s medium containing 200 μg/mL hygromycin and 5μg/mL blasticidin. hTERT-HFK cells were provided by Aloysius J. Klingelhutz(University of Iowa, Iowa City, IA) (51) and cultured in K-SFM media (LifeTechnologies) containing 0.16 ng/mL of epithelial growth factor and 25 μg/mLof bovine pituitary extract. Plasmid transfections were carried out usingLipofectamine 2000 (Invitrogen) according to the manufacturer’s protocol. Allcell lines were tested negative for mycoplasma contamination. CIN612 9E cellsand HFK-31 cells were grown in E medium supplemented with 5 ng/mLmouse epidermal growth factor (BD Biosciences) in the presence of mitomycinC-treated J2 3T3 fibroblast feeder cells (73). When necessary, J2 feeders wereremoved from HPV+ cells by incubation with 1 mM EDTA in PBS. For differ-entiation, subconfluent cells were harvested at T0, and the remaining plates ofcells were serum-starved overnight in basal keratinocyte growth medium(KGM; Lonza) with supplements for 16 h. Cells were then incubated in ker-atinocyte basal medium (KBM; Lonza) without supplements but with 1.8 mMCaCl2. Cells were allowed to differentiate for 48 to 96 h after addition of high-calcium medium. DNA and protein were harvested, and viral genome ampli-fication was measured by Southern blotting as for each experiment to ensureactivation of the productive phase of the viral life cycle.

Southern Blot Analysis. DNA was isolated and Southern blot analysis wasperformed as described previously (74). Briefly, cells were harvested in buffercontaining 400 mM NaCl, 10 mM Tris (pH 7.5), and 10 mM EDTA. Cells werelysed by the addition of 30 μL of 20% SDS and subsequently treated with15 μL of 10 mg/mL proteinase K overnight at 37 °C. DNA was then extractedwith phenol chloroform and precipitated using sodium acetate and ethanol.Five micrograms of resultant DNA were digested with BamHI (which doesnot cut the HPV31 genome) or HindIII (which linearizes the HPV31 genome),separated on 0.8% agarose gel for 15 h at 40 V, and subsequently trans-ferred to a positively charged nylon membrane (Immobilon-Ny+; Millipore).The 32P-labeled linearized HPV31 genome was used as a probe.

RNA Interference. Single siRNA duplexes targeting either RNF168 (D-007152-04)or a nontargeting sequence (D-001210-02-02) were purchased from Dharmacon.To knock down RNF8, a SMARTpools siRNA against RNF8 (D-006900-01-0020) waspurchased fromDharmacon. Unless stated otherwise, all siRNAs were transfectedin a forward transfection mode 24 h prior to cell processing using RNAimax(Invitrogen) according to the manufacturer’s protocol.

Generation of Lentiviruses. Lentiviruses were produced as previously de-scribed (15). Briefly, plasmids expressing shRNAs or the indicated GFP-taggedproteins were cotransfected with vesicular stomatitis virus G (VSV-G) plasmidDNA and Gag-Pol-Tet-Rev plasmid DNA into HEK-293T cells using poly-ethyleneimine (PEI). At 48 to 72 h posttransfection, supernatants containinglentivirus were harvested, and CIN612 9E, U2OS, or hTERT-HFKs were trans-duced in the presence of 8 μg/mL hexadimethrine bromide (Polybrene; Sigma).Three days posttransduction, CIN612 9E were either harvested at T0 (un-differentiated sample) or differentiated in high-calcium medium for 72 h asdescribed here earlier. Knockdown of RNF168 and the expression of the GFPconstructs was confirmed for each experiment by Western blotting.

CRISPR-Cas9 Genome Editing of 53BP1 and RNF168. C33-A and U2OS cells weretransiently transfected with sgRNAs targeting 53BP1 or RNF168, respectively(SI Appendix, Table S1), and expressed from the pX459 vector containingCas9 followed by the 2A-Puromycin cassette. The next day, cells weresubcloned to form single colonies. Clones were screened by immunoblot andimmunofluorescence to verify the loss of 53BP1 and RNF168. The genomicregions targeted by the CRISPR-Cas9 was subsequently characterized by PCRamplification, DNA sequencing, and TIDE analysis (75).

Immunofluorescence Microscopy. Immunofluorescence was done as previouslydescribed (24) with the indicated antibodies (SI Appendix, Table S2). Mi-crograph images were taken using a Zeiss LSM700 laser-scanning microscopeequipped with a 63× oil lens. In all micrographs, dashed lines indicated nucleusoutlines. Insets represent a 10-fold magnification of the indicated region.Quantification was done on 3 biological replicates, and at least 50 cells werecounted in each experiment. Unless stated otherwise, t test with Welch cor-rection was realized to assess statistical significance.

Clover Assay. C33-A stable cell lines were plated at 100,000 cells per well on6-well plates 24 h post transfection with pX330-LMNAgRNA1 andpCR2.1-CloverLMNAdonor plasmids as previously described (76). At72 h posttransfection, percentage of mClover-positive cells were analyzedby flow cytometry using an Accuri C6 (BD Biosciences).

Immunoprecipitation. Cells were transfected with mCherry-LacR, mCherryLacR-HPV31 E7 full length, or mCherryLacR-HPV31 E7 CR1-CR2–expressing plasmids.At 15 h posttransfection, GFP, GFP-RNF168, GFP-RNF168 WT, and ΔE7BD ex-pression were induced. Cells were lysed 9 h later in lysis buffer (50 mM Tris, pH7.5, 300 mM NaCl, 1 mM EDTA, 1% Triton X-100) followed by Micrococcalnuclease (Sigma) treatment for 30 min. Cleared cell lysate (1 mg) was sub-jected to immunoprecipitation using 20 μL GFP Trap beads for 3 h at 4 °C. After3 washing steps, beads were eluted in 2× Laemmli buffer for immunoblottinganalyses.

Purification of Recombinant Protein. GST-HPV16 E7, RNF168 WT, and ΔE7BDrecombinant proteins were purified with technical help from Marie-Christine Caron from the laboratory of Jean-Yves Masson (Université Laval,Québec, QC, Canada) as previously described (53). For MBP-His recombinantprotein purification, E. coli strain BL-21 was induced with 1 mM IPTG whenthey reached an OD600 of 0.6. Next, fusion proteins were purified on Ni-NTAagarose (Qiagen) according to the batch method described in the manu-facturer’s manual. All recombinant proteins were stored in 20 mM Trisacetate, pH 8.0, 200 mM potassium acetate, 10% glycerol, 1 mM EDTA,0.5 mM DTT.

Ubiquitylation Reaction. Mononucleosomes were isolated from the HeLa cellsand ubiquitylated as described previously (25). Briefly, 2.5 μg mono-nucleosomes were incubated with 30 nM Uba1, 1.5 μM UbcH5a, 4 μM HIS-RNF168, 22 μM ubiquitin (Boston Biochem), and 3.33 mM ATP in a buffercontaining 50 mM Tris·HCl, pH 7.5, 100 mM NaCl, 10 mM MgCl2, 1 μM ZnCl2,and 1 mM DTT at 30 °C for 4 h. His-GST (4.44 μM) or GST-E7 (2.2 to 4.4 μM)proteins were added to the mixture. The reaction was quenched by adding15 mM EDTA, pH 8.0. Denatured samples were separated on an SDS/PAGEand analyzed by Western blotting experiment.

GST Pull-Down. For GST pull-down, 1 μg of purified GST-tagged protein wasincubated with 4 μg of purified His-tagged protein in 1.5 mL of GSTB buffer(20 mM Hepes, pH 7.4, 150 mM KCl, 10% glycerol, 0.02% Triton X-100,0.2 mM EDTA, 0.002 μg/mL aprotinin, 0.001 μg/mL pepstatin A, 0.002 μg/mLleupeptin, 1 mM AEBSF, 1 mM DTT, 1 mg/mL bovine serum albumin [BSA]) for3 h at room temperature. Input was collected before the addition of gluta-thione Sepharose beads for 90 min at room temperature. Beads were thenwashed with GSTB buffer without BSA and directly eluted in 2× Laemmlibuffer for immunoblotting analyses.

RNA-Seq Analyses. Level 3 RNA-Seq by Expectation Maximization (RSEM)normalized Illumina HiSeq RNA expression data for the TCGA head and neckcancer (HNSC) and cervical carcinoma (CESC) cohorts were downloaded fromthe Broad Genome Data Analysis Center’s Firehose server (https://gdac.broadinstitute.org/). For all genes, the gene-level Firehose dataset was used.RSEM-normalized expression data were extracted, and the HPV status wasmanually curated based on published datasets as previously described(77). Primary patient samples with known HPV status were grouped asHPV-positive, HPV-negative, or normal-adjacent control tissue. For theboxplots, center lines show the medians, box limits indicate the 25th and75th percentiles as determined by GraphPad Prism, and whiskers extend1.5 times the interquartile range from the 25th and 75th percentiles. Pvalues were assigned using a 2-tailed nonparametric Mann–WhitneyU test.

ACKNOWLEDGMENTS. We are grateful to Jacques Archambault, JacquesCôté, Agnel Sfeir, Alexandre Orthwein, Sabine Elowe, Kavi Mehta, and mem-bers of the A.F.-T. laboratory for critical reading of the manuscript; and toDaniel Durocher, Jacques Archambault, Jean-Yves Masson, Graham Dellaire,

19560 | www.pnas.org/cgi/doi/10.1073/pnas.1906102116 Sitz et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

26, 2

020

Roger Greenberg, and Karl Munger for essential reagents such as plasmids,purified proteins, and cell lines. We thank Vincent Tremblay for technicalsupport to J.S., and Takashi Awaga for early work on establishing the LacO/LacR assay with HPV proteins. J.S. has received the doctoral fellowship “LucBélanger” from CHU de Québec–Université Laval Foundation.” S.A.B. and E.B.are supported by a master and a postdoctoral fellowship from the Fonds deRecherche du Québec - Santé (FRQS), respectively. This work was supported by

Canadian Institutes of Health Research Project Grants 152948 (to A.F.-T.) and142491 (to J.S.M.), and Scholarship for the Next Generation of Scientists19590 from Cancer Research Society (to A.F.-T.). A.F.-T. is a tier 2 Canada Re-search Chair in Molecular Virology and Genomic Instability and is supported bythe Foundation J.-Louis Lévesque. C.A.M. is supported by grants from theHealth and Human Services (HHS) NIH National Cancer Institute (CA181581and CA226523).

1. C. de Martel et al., Global burden of cancers attributable to infections in 2008: Areview and synthetic analysis. Lancet Oncol. 13, 607–615 (2012).

2. E. A. Mesri, M. A. Feitelson, K. Munger, Human viral oncogenesis: A cancer hallmarksanalysis. Cell Host Microbe 15, 266–282 (2014).

3. E. M. de Villiers, C. Fauquet, T. R. Broker, H. U. Bernard, H. zur Hausen, Classificationof papillomaviruses. Virology 324, 17–27 (2004).

4. J. Doorbar, N. Egawa, H. Griffin, C. Kranjec, I. Murakami, Human papillomavirusmolecular biology and disease association. Rev. Med. Virol. 25 (suppl. 1), 2–23(2015).

5. M. L. Gillison, A. K. Chaturvedi, W. F. Anderson, C. Fakhry, Epidemiology of humanpapillomavirus-positive head and neck squamous cell carcinoma. J. Clin. Oncol. 33,3235–3242 (2015).

6. Statistic Canada, Canadian cancer statistics 2016, special topic: HPV-associated cancers.http://www.cancer.ca/∼/media/cancer.ca/CW/cancer%20information/cancer%20101/Canadian%20cancer%20statistics/Canadian-Cancer-Statistics-2016-EN.pdf?la=en. Accessed27 March 2019.

7. A. C. Nichols et al., The epidemic of human papillomavirus and oropharyngeal cancerin a Canadian population. Curr. Oncol. 20, 212–219 (2013).

8. A. A. McBride, Mechanisms and strategies of papillomavirus replication. Biol. Chem.398, 919–927 (2017).

9. C. C. McKinney, K. L. Hussmann, A. A. McBride, The role of the DNA damage responsethroughout the papillomavirus life cycle. Viruses 7, 2450–2469 (2015).

10. M. D. Weitzman, A. Fradet-Turcotte, Virus DNA replication and the host DNA damageresponse. Annu. Rev. Virol. 5, 141–164 (2018).

11. K. A. Gillespie, K. P. Mehta, L. A. Laimins, C. A. Moody, Human papillomaviruses re-cruit cellular DNA repair and homologous recombination factors to viral replicationcenters. J. Virol. 86, 9520–9526 (2012).

12. C. A. Moody, L. A. Laimins, Human papillomaviruses activate the ATM DNA damagepathway for viral genome amplification upon differentiation. PLoS Pathog. 5,e1000605 (2009).

13. N. Sakakibara et al., Brd4 is displaced from HPV replication factories as they expandand amplify viral DNA. PLoS Pathog. 9, e1003777 (2013).

14. M. Kadaja, H. Isok-Paas, T. Laos, E. Ustav, M. Ustav, Mechanism of genomic instabilityin cells infected with the high-risk human papillomaviruses. PLoS Pathog. 5, e1000397(2009).

15. D. C. Anacker, D. Gautam, K. A. Gillespie, W. H. Chappell, C. A. Moody, Productivereplication of human papillomavirus 31 requires DNA repair factor Nbs1. J. Virol. 88,8528–8544 (2014).

16. W. H. Chappell, et al., Homologous recombination repair factors, Rad51 and BRCA1,are necessary for productive replication of human papillomavirus 31. J Virol. 90, 2639–2652 (2016).

17. K. Mehta, L. Laimins, Human papillomaviruses preferentially recruit DNA repair fac-tors to viral genomes for rapid repair and amplification. MBio 9, e00064-18 (2018).

18. D. Hanahan, R. A. Weinberg, Hallmarks of cancer: The next generation. Cell 144, 646–674 (2011).

19. L. L. Sandell, V. A. Zakian, Loss of a yeast telomere: Arrest, recovery, and chromosomeloss. Cell 75, 729–739 (1993).

20. N. Hustedt, D. Durocher, The control of DNA repair by the cell cycle. Nat. Cell Biol. 19,1–9 (2016).

21. N. P. Dantuma, H. van Attikum, Spatiotemporal regulation of posttranslationalmodifications in the DNA damage response. EMBO J. 35, 6–23 (2016).

22. M. D. Wilson, D. Durocher, Reading chromatin signatures after DNA double-strandbreaks. Philos. Trans. R. Soc. Lond. B Biol. Sci. 372, 20160280 (2017).

23. F. Mattiroli et al., RNF168 ubiquitinates K13-15 on H2A/H2AX to drive DNA damagesignaling. Cell 150, 1182–1195 (2012).

24. S. Panier et al., Tandem protein interaction modules organize the ubiquitin-dependent response to DNA double-strand breaks. Mol. Cell 47, 383–395 (2012).

25. A. Fradet-Turcotte et al., 53BP1 is a reader of the DNA-damage-induced H2A Lys15 ubiquitin mark. Nature 499, 50–54 (2013).

26. J. Kitevski-LeBlanc et al., The RNF168 paralog RNF169 defines a new class ofubiquitylated histone reader involved in the response to DNA damage. eLife 6,e23872 (2017).

27. Q. Hu, M. V. Botuyan, G. Cui, D. Zhao, G. Mer, Mechanisms of ubiquitin-nucleosomerecognition and regulation of 53BP1 chromatin recruitment by RNF168/169 andRAD18. Mol. cell 66, 473–487.e9 (2017).

28. M. D. Wilson et al., The structural basis of modified nucleosome recognition by 53BP1.Nature 536, 100–103 (2016).

29. D. Setiaputra, D. Durocher, Shieldin - the protector of DNA ends. EMBO Rep. 20,e47560 (2019).

30. T. Bohgaki et al., Genomic instability, defective spermatogenesis, immunodeficiency,and cancer in a mouse model of the RIDDLE syndrome. PLoS Genet. 7, e1001381(2011).

31. G. S. Stewart et al., The RIDDLE syndrome protein mediates a ubiquitin-dependentsignaling cascade at sites of DNA damage. Cell 136, 420–434 (2009).

32. T. Gudjonsson et al., TRIP12 and UBR5 suppress spreading of chromatin ubiquitylationat damaged chromosomes. Cell 150, 697–709 (2012).

33. E. Cerami et al., The cBio cancer genomics portal: An open platform for exploringmultidimensional cancer genomics data. Cancer Discov. 2, 401–404 (2012).

34. J. Gao et al., Integrative analysis of complex cancer genomics and clinical profilesusing the cBioPortal. Sci. Signal. 6, pl1 (2013).

35. A. Pombo et al., Regional and temporal specialization in the nucleus: Atranscriptionally-active nuclear domain rich in PTF, Oct1 and PIKA antigens associateswith specific chromosomes early in the cell cycle. EMBO J. 17, 1768–1778 (1998).

36. C. Lukas et al., 53BP1 nuclear bodies form around DNA lesions generated by mitotictransmission of chromosomes under replication stress. Nat. Cell Biol. 13, 243–253(2011).

37. J. A. Harrigan et al., Replication stress induces 53BP1-containing OPT domains inG1 cells. J. Cell Biol. 193, 97–108 (2011).

38. J. Spies et al., 53BP1 nuclear bodies enforce replication timing at under-replicatedDNA to limit heritable DNA damage. Nat. Cell Biol. 21, 487–497 (2019).

39. E. Soutoglou, T. Misteli, Activation of the cellular DNA damage response in the ab-sence of DNA lesions. Science 320, 1507–1510 (2008).

40. M. M. Donaldson et al., An interaction between human papillomavirus 16 E2 andTopBP1 is required for optimum viral DNA replication and episomal genome estab-lishment. J. Virol. 86, 12806–12815 (2012).

41. S. Jha et al., Destabilization of TIP60 by human papillomavirus E6 results in attenu-ation of TIP60-dependent transcriptional regulation and apoptotic pathway.Mol. Cell38, 700–711 (2010).

42. V. K. Subbaiah et al., E3 ligase EDD1/UBR5 is utilized by the HPV E6 oncogene todestabilize tumor suppressor TIP60. Oncogene 35, 2062–2074 (2016).

43. M. S. Longworth, L. A. Laimins, The binding of histone deacetylases and the integrityof zinc finger-like motifs of the E7 protein are essential for the life cycle of humanpapillomavirus type 31. J. Virol. 78, 3533–3541 (2004).

44. M. S. Longworth, R. Wilson, L. A. Laimins, HPV31 E7 facilitates replication by acti-vating E2F2 transcription through its interaction with HDACs. EMBO J. 24, 1821–1830(2005).

45. A. Roman, K. Munger, The papillomavirus E7 proteins. Virology 445, 138–168 (2013).46. V. Bouvard et al.; WHO International Agency for Research on Cancer Monograph

Working Group, A review of human carcinogens–Part B: Biological agents. LancetOncol. 10, 321–322 (2009).

47. B. Todorovic et al., Systematic analysis of the amino acid residues of human papil-lomavirus type 16 E7 conserved region 3 involved in dimerization and transformation.J. Virol. 85, 10048–10057 (2011).

48. D. F. Squarzanti et al., Human papillomavirus type 16 E6 and E7 oncoproteins interactwith the nuclear p53-binding protein 1 in an in vitro reconstructed 3D epithelium:New insights for the virus-induced DNA damage response. Virol. J. 15, 176 (2018).

49. J. A. Schmid et al., Histone ubiquitination by the DNA damage response is requiredfor efficient DNA replication in unperturbed S phase. Mol. cell 71, 897–910.e8 (2018).

50. C. Broceño, S. Wilkie, S. Mittnacht, RB activation defect in tumor cell lines. Proc. Natl.Acad. Sci. U.S.A. 99, 14200–14205 (2002).

51. T. Kiyono et al., Both Rb/p16INK4a inactivation and telomerase activity are requiredto immortalize human epithelial cells. Nature 396, 84–88 (1998).

52. D. Zong et al., BRCA1 haploinsufficiency is masked by RNF168-mediated chromatinubiquitylation. Mol. Cell 73, 1267–1281.e7 (2019).

53. M. S. Luijsterburg et al., A PALB2-interacting domain in RNF168 couples homologousrecombination to DNA break-induced chromatin ubiquitylation. eLife 6, e20922(2017).

54. J. M. Dybas, C. Herrmann, M. D. Weitzman, Ubiquitination at the interface of tumorviruses and DNA damage responses. Curr. Opin. Virol. 32, 40–47 (2018).

55. C. E. Lilley et al., A viral E3 ligase targets RNF8 and RNF168 to control histoneubiquitination and DNA damage responses. EMBO J. 29, 943–955 (2010).

56. C. E. Lilley, M. S. Chaurushiya, C. Boutell, R. D. Everett, M. D. Weitzman, The intrinsicantiviral defense to incoming HSV-1 genomes includes specific DNA repair proteinsand is counteracted by the viral protein ICP0. PLoS Pathog. 7, e1002084 (2011).

57. M. S. Chaurushiya et al., Viral E3 ubiquitin ligase-mediated degradation of a cellularE3: Viral mimicry of a cellular phosphorylation mark targets the RNF8 FHA domain.Mol. Cell 46, 79–90 (2012).

58. J. Bai et al., A high-throughput screen for genes essential for PRRSV infection using apiggyBac-based system. Virology 531, 19–30 (2019).

59. J. W. Leung et al., Nucleosome acidic patch promotes RNF168- and RING1B/BMI1-dependent H2AX and H2A ubiquitination and DNA damage signaling. PLoS Genet.10, e1004178 (2014).

60. T. H. Ho et al., A screen for epstein-barr virus proteins that inhibit the DNA damageresponse reveals a novel histone binding protein. J. Virol. 92, e00262-18 (2018).

61. A. Kakarougkas et al., Opposing roles for 53BP1 during homologous recombination.Nucleic Acids Res. 41, 9719–9731 (2013).

62. H. K. Wang, A. A. Duffy, T. R. Broker, L. T. Chow, Robust production and passaging ofinfectious HPV in squamous epithelium of primary human keratinocytes. Genes Dev.23, 181–194 (2009).

Sitz et al. PNAS | September 24, 2019 | vol. 116 | no. 39 | 19561

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

26, 2

020

63. A. T. Noon et al., 53BP1-dependent robust localized KAP-1 phosphorylation isessential for heterochromatic DNA double-strand break repair. Nat. Cell Biol. 12,177–184 (2010).

64. C. A. Moody, L. A. Laimins, Human papillomavirus oncoproteins: Pathways to trans-formation. Nat. Rev. Cancer 10, 550–560 (2010).

65. F. A. Mallette et al., RNF8- and RNF168-dependent degradation of KDM4A/JMJD2Atriggers 53BP1 recruitment to DNA damage sites. EMBO J. 31, 1865–1878 (2012).

66. J. A. Westrich et al., Human papillomavirus 16 E7 stabilizes APOBEC3A protein byinhibiting cullin 2-dependent protein degradation. J. Virol. 92, e01318-17 (2018).

67. K. Huh et al., Human papillomavirus type 16 E7 oncoprotein associates with the cullin2 ubiquitin ligase complex, which contributes to degradation of the retinoblastomatumor suppressor. J. Virol. 81, 9737–9747 (2007).

68. E. A. White et al., Systematic identification of interactions between host cell proteinsand E7 oncoproteins from diverse human papillomaviruses. Proc. Natl. Acad. Sci.U.S.A. 109, E260–E267 (2012).

69. C. A. Moody, Impact of replication stress in human papillomavirus pathogenesis.J. Virol. 93, e01012-17 (2019).

70. D. Zong et al., Ectopic expression of RNF168 and 53BP1 increases mutagenic but notphysiological non-homologous end joining. Nucleic Acids Res. 43, 4950–4961 (2015).

71. L. Mirabello et al., HPV16 E7 genetic conservation is critical to carcinogenesis. Cell 170,

1164–1174.e6 (2017).72. J. Tang et al., Acetylation limits 53BP1 association with damaged chromatin to pro-

mote homologous recombination. Nat. Struct. Mol. Biol. 20, 317–325 (2013).73. R. Wilson, L. A. Laimins, Differentiation of HPV-containing cells using organotypic

“raft” culture or methylcellulose. Methods Mol. Med. 119, 157–169 (2005).74. F. Fehrmann, D. J. Klumpp, L. A. Laimins, Human papillomavirus type 31 E5 protein

supports cell cycle progression and activates late viral functions upon epithelial dif-

ferentiation. J. Virol. 77, 2819–2831 (2003).75. E. K. Brinkman, T. Chen, M. Amendola, B. van Steensel, Easy quantitative assessment

of genome editing by sequence trace decomposition. Nucleic Acids Res. 42, e168

(2014).76. J. Pinder, J. Salsman, G. Dellaire, Nuclear domain ‘knock-in’ screen for the evaluation

and identification of small molecule enhancers of CRISPR-based genome editing.

Nucleic Acids Res. 43, 9379–9392 (2015).77. S. F. Gameiro et al., Analysis of class I major histocompatibility complex gene tran-

scription in human tumors caused by human papillomavirus infection. Viruses 9,

E252 (2017).

19562 | www.pnas.org/cgi/doi/10.1073/pnas.1906102116 Sitz et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

26, 2

020