Division of labor within the DNA damage tolerance system reveals non-epistatic and clinically actionable targets for precision cancer medicine

Abstract Crosslink repair depends on the Fanconi anemia pathway and translesion synthesis polymerases that replicate over unhooked crosslinks. Translesion synthesis is regulated via ubiquitination of PCNA, and independently via translesion synthesis polymerase REV1. The division of labor between PCNA-ubiquitination and REV1 in interstrand crosslink repair is unclear. Inhibition of either of these pathways has been proposed as a strategy to increase cytotoxicity of platinating agents in cancer treatment. Here, we defined the importance of PCNA-ubiquitination and REV1 for DNA in mammalian ICL repair. In mice, loss of PCNA-ubiquitination, but not REV1, resulted in germ cell defects and hypersensitivity to cisplatin. Loss of PCNA-ubiquitination, but not REV1 sensitized mammalian cancer cell lines to cisplatin. We identify polymerase Kappa as essential in tolerating DNA damage-induced lesions, in particular cisplatin lesions. Polk-deficient tumors were controlled by cisplatin treatment and it significantly delayed tumor outgrowth and increased overall survival of tumor bearing mice. Our results indicate that PCNA-ubiquitination and REV1 play distinct roles in DNA damage tolerance. Moreover, our results highlight POLK as a critical TLS polymerase in tolerating multiple genotoxic lesions, including cisplatin lesions. The relative frequent loss of Polk in cancers indicates an exploitable vulnerability for precision cancer medicine.


INTRODUCTION
Platinating agents, such as cisplatin, are mainstays in the treatment of a multitude of solid and hematological malignancies (1). Cisplatin induces highly toxic DNA interstrand crosslinks (ICLs), which block replication and transcription by preventing the separation of the DNA strands. These ICLs are normally repaired via the Fanconi anemia (FA) pathway (2). In humans, absence of the FA repair pathway results in the cancer predisposition FA syndrome that is characterized by bone marrow failure, depletion of hematopoietic stem cells (HSCs), reduced fertility, and increased incidence of cancer (3,4). Importantly, FA-patients are highly sensitive to drugs that induce ICLs, such as cisplatin (4).
FA-mediated resolution of ICLs requires that incisions are made on one strand in the DNA sugar-phosphate backbone adjacent to the crosslink, which enables its unhooking. The unhooked ICL lesion stalls high-fidelity replicative DNA polymerases (5), and requires specialized error-prone translesion synthesis (TLS) to continue replication opposite the lesion. In this way, TLS provides the template for subsequent homology directed repair that is required in ICL. In the absence of TLS, ICL repair is severely diminished (5), and cells are greatly sensitized to various types of ICLinducing agents (6).
A central pathway of TLS is promoted by site-specific mono-ubiquitination of the DNA sliding clamp Proliferating Cell Nuclear Antigen (PCNA-mUb) on lysine residue 164 (K164), mediated by the E2-E3 ubiquitin conjugase/ligase complex Rad6/Rad18 (7,8). Additionally, the TLS polymerase REV1 can bypass lesions independently of PCNA-mUb, and has a unique capacity to recruit other TLS polymerases via its specific C-terminal domain (9,10).
We have previously shown in vivo that injection of cisplatin resulted in complete bone marrow failure and hematopoietic stem cell depletion in Pcna K164R/K164R mice (from here on termed Pcna K164R ), in which PCNAubiquitination is abolished (6). Others have shown that Rev1-deficient mice developed progressive bone marrow failure in compound mutants, which besides Rev1 lack Xpc and therefore are also deficient for global nucleotide excision repair (NER) (11), probably due to endogenous lesions. However, the extent to which REV1 is required to tolerate cisplatin-induced lesions in vivo is not known. Moreover, both inhibition of PCNA-or REV1-dependent DNA damage tolerance (DDT) have been proposed to chemosensitize cancers to cisplatin (12,13), enhancing the efficacy of cancer treatment with platinating agents. However, it is unclear if, and to what extent, PCNA-ubiquitination or REV1 can compensate for the absence of each other in processing unhooked cisplatin ICLs. Moreover, because PCNAubiquitination and REV1 can recruit various distinct TLS polymerases, it is likely that loss of specific TLS polymerases sensitizes tumors to cisplatin. However, which TLS polymerase participates in ICL repair and whether its loss in tumors enhances the efficacy of cisplatin treatment remains to be examined.
Here, we studied the relevance of PCNA-ubiquitination and REV1 in the bypass of cisplatin lesions in vivo and in vitro. We generated Rev1-deficient mice, and compared them to Pcna K164R mice. We report that PCNAubiquitination, but strikingly not REV1, is essential in bypassing cisplatin lesions in vivo. Moreover, REV1 is dispensable for the tolerance of cisplatin lesions in various murine and human cancer cells in vitro. Interestingly, we identify TLS polymerase Kappa (POLK) to be essential for tolerance of cisplatin lesions in a murine lymphoma model. High-throughput compound screens revealed POLK to be essential to tolerate various classes of DNA damageinduced lesions. In vivo, cisplatin treatment significantly delayed the outgrowth of POLK-deficient lymphomas, and substantially increased the overall survival time of tumor bearing mice.

Reagents, strains, antibody-usage and PCR settings
See supplementary Tables S5-S12 for extensive details.

Histopathological analysis
For histopathological analysis, organs were collected and fixed in ethanol-acetic acid-formaldehyde (EAF) for 24 h after necropsy. Fixed tissues and organs were subsequently dehydrated, embedded in paraffin, and 2 m sections were prepared and stained with hematoxylin and eosin (H&E).

Generation of mouse embryonic fibroblast (MEFs) and PreB cells
PreB cells and MEFs were isolated from time-mated intercrosses of WT, Pcna K164R and Rev1-KO mice using standard procedures. For Rev1-Del mice, only MEFs were isolated. For the generation of PreB cell cultures, E14.5 embryos were isolated and single cell suspensions were generated from fetal livers and subsequently cultured on ST2 feeder cells in IL7-containing complete medium (Iscoves, 8% Fetal Calf Serum (FCS) and Penicillin and Streptomycin (Pen/Strep)) (15). MEFs were isolated as described previously (16). MEFs were immortalized using lentiviral expression of a p53-specific shRNA (17).

Somatic hypermutation analysis
DNA was extracted from germinal center B cells isolated from Peyer's patches using proteinase K treatment and ethanol precipitation. The JH4 flanking intronic sequence of endogenous rearrangements of VHJ558 family members were amplified during 40 cycles of PCR using PFU Ultra polymerase (Stratagene), as described previously (21). PCR products were purified using the QIAquick Gel Extraction kit (Qiagen) and cloned into the TOPO II blunt vector (Invitrogen Life Technologies) and sequenced on a 3730 DNA analyzer (Applied Biosystems). Sequence alignment was performed using Seqman software (DNAStar). Calculations exclude non-mutated sequences, insertions and deletions. Clonal sequences were counted only once.

In vitro class switch recombination assay
Spleens were isolated from WT and Rev1-KO mice. Single cell suspensions were generated and subjected to erythrocyte lysis. Naïve splenic B cells were enriched by performing a CD43 depletion using a biotinylated anti-CD43 antibody, followed by Streptavidin magnetic bead incubation. Cells were subsequently stimulated with Lipopolysaccharide (Escherichia coli, LPS, 5 g/ml, Sigma), or LPS and IL-4 (10 ng/ml, Reprotech) for four days, after which class switch recombination to IgG3 (LPS only stimulation) or IgG1 (LPS and IL-4 stimulation) was assessed using flow cytometry.

siRNA transfection
HCT116 and MCF-7 cells were grown in 6-well plates and transfected with SMARTpool siRNAs against human REV1 (Horizon-Discovery) using RNAImax according to the manufacturer's specifications. Non-targeting control SMARTpool siRNAs were used as control.

Survival assay, synergy experiments and high-throughput compound screen
For survival assays and synergy assays, cells were seeded at low confluency using the Multidrop Combi (Thermo Fisher Scientific), into 384-well plates (Greiner). After 24 h, the respective compounds were added (cisplatin, REV1i JH-RE-06, or both, or methyl methanesulfonate (MMS)) using a tecan d300e compound printer (HP), including 3-5 replicates per dosage. Positive (1 M phenylarsine oxide) and negative (0,1% DMSO) controls were added to each assay plate. Viability readout was performed after 72 h, using the CellTiter-Blue assay (G8081/2, Promega) following the protocol of the manufacturer using an EnVision multimode plate reader (PerkinElmer). The CTB data was normalized per plate using the normalized percentage inhibition (NPI) method. NPI sets the mean of the positive control value to 0 and mean of the negative control to 1. Synergy calculations were performed in R using the Chou-Talalay Combination index method (22). Similar protocols were used to perform the highthroughput drug screen: Using the Multidrop Combi (Thermo Fisher Scientific), untreated p53-KO lymphoma cells were seeded into 384 well plates at low confluency. After 24 h, the collection of compound libraries available at the NKI (Selleck GPCR, Kinase, Apoptosis, Phospatase, Epigenetic, LOPAC and NCI oncology) was added. This library was stored and handled as recommended by the manufacturer. Compounds from the master plate were diluted in daughter plates containing complete RPMI 1640 medium, using the microlab star liquid handling workstation (Hamilton). From the daughter plates, the diluted compounds were transferred into 384 well assay plates, in triplicate, with final concentrations of 1 M, and 5 M. In addition, Positive (1 M Phenylarsine oxide) and negative (0.1% DMSO) controls were added alternately to wells in column 2 and 23 of each assay plate. After 3 days, viability was measured using CellTiter-Blue assay (G8081/2, Promega) following the protocol of the manufacturer. The CTB data was normalized per plate using the normalized percentage inhibition (NPI) method. NPI sets the mean of the positive control value to 0 and mean of the negative control to 1. Using the replicate values of both conditions a two-sided t-test was performed. Afterwards the P-values were corrected for multiple testing using the Benjamini-Hochberg method. All calculations were done in R.

Colony formation assay
p53-KD MEFs were seeded in 10 cm dishes at various seeding densities in complete medium. One day later, the medium was removed and replaced with complete medium containing the indicated concentrations of cisplatin. After 6 days, medium was aspirated and the cells were washed with PBS and fixed in 5 ml 3:1 v/v Methanol: Acetic acid for 1 h. Following the fixation, colonies were stained by adding 3 ml of 0.3% Coomassie brilliant blue (Merck) solution prepared in H 2 O. After 1.5 h, the staining solution was removed, the dishes were washed with H 2 O and dried overnight. Colonies were counted using the Colcount (Oxford Optronix) and each condition was corrected for the plating efficiency before being normalized to the untreated condition. Data points represent the mean survival relative to the untreated control cells.

Incucyte ® assays
Viable p53-knockdown (KD) MEFs were seeded at the concentration of 300 cells per well in 96-well plates. The cells were allowed to grow for 8 days or 10 days in Incucyte (Sartorius) under standard tissue culture conditions in the presence or absence of cisplatin, mitomycin C. Images were taken every 4 h and results were analyzed in Excel (Microsoft) as percentage of confluence, that represents the percentage of the image area occupied by objects.

sgRNA design for murine Pcna, Polk and Rev1
Benchling was used to design guides targeting two exons or introns of each gene. These were cloned into pX330 (one guide per plasmid) or a modified form of pX330, named pX333-mCherry, which contained sites for two guides. The protocol from Ran et al. (23) was used to clone the guides in pX330. For pX333-mCherry, one guide was first cloned using BbsI. Plasmids were transformed into competent DH5a, sequenced using the U6 Fwd primer, and subsequently the second guide was cloned using BbsI. After transformation of DH5a with sequence-verified pX330 or pX333-mCherry, midipreps were performed using the HiPure Plasmid Midiprep kit (Invitrogen) according to the manufacturer's instructions.

Nucleofection
To obtain WT, Pcna K164R , Polk-KO and Rev1-KO lymphoma cell lines, 10 × 10 6 lymphoma cells were resuspended in 100 l additive-free RPMI 1640 and resuspended with 2 times 5g of pX330-puro each expressing a sgRNA against Pcna, or Polk and 1 g of pmx-eGFP construct (Amaxa, Lonza), or 10 g of pX333-mCherry featuring two guides against murine Rev1. Cells were nucleofected using a nucleofector II-b device (Amaxa, Lonza), using program A-30. Subsequently, cells were grown in regular medium for 2-3 days, and GFP or mCherry positive cells were sorted with MoFlo Astrios or FACSAria IIu (BD Biosciences) cell sorters and cultured for a few days. Clones were obtained via limiting dilution.

Isolation of genomic DNA, PCR and whole exome sequencing
To obtain DNA for mouse-genotyping, and to screen for the presence of Pcna-mutations, Rev1-KO and Polk-KO clones, cells were lysed in lysis buffer containing ProtK overnight at 55ºC, followed by a brief inactivation of ProtK at 85ºC for 5 min. The lysate was mixed with sterile BRAUN water, of which 2 l was used in a reaction containing MyTaq polymerase (1:100), MyTaq buffer (BioLine), and primers. DNA was amplified by PCR as indicated in Table 12.To further validate the Polk deletion, the genomic DNA of the clone was isolated using a genomic DNA isolation kit (Isolate II Genomic DNA kit, BioLine), followed by capture and exome sequencing using a HiSeq 2500 sequencing machine with 100 bp reads using paired end sequencing. The genome was aligned to the Mus musculus reference genome, and visualized using Integrated Genomics Viewer (IGV, version 2.8.6).

Western blot
To validate the genetic ablation of Rev1, lymphoma cells were lysed in RIPA buffer (25 mM Tris-HCl (pH 7.6), 150 mM NaCl, 1% NP-40, 1% sodium deoxycholate, and 0.1% SDS) containing 1× protease inhibitor cocktail (Roche) for 30 min on ice. Lysates were sonicated for 15 min using a BioRuptor (30 s on, 30 s off, maximum power, at 4ºC). Samples were spun at 20 000 g for 10 min, and the protein concentration was measured either via a Bradford assay, or via a BCA assay. Samples were run on a 4-12% Bis-Tris gel with MES running buffer on ice, at 150-200 V for 1.5-2 h, until the loading band ran off the gel. The gels were transferred using a BioRad Transblot Turbo semi-dry transfer system with the mixed-MW program. After staining with Ponceau-S, samples were blocked for 1 h using TBS-T containing 5% skim-milk powder, followed by incubation with primary antibodies overnight at 4ºC on a roller bank in TBS-T 1% milk. The next day, blots were washed 4 times for 5 min with TBS-T, followed by a 2 h incubation with the secondary antibodies at room temperature. The membrane was washed 3 times with TBS-T for 5 min, after which the membrane was imaged on an Oddyssey scanner (LiCor). For MEFs, similar procedures were followed, but here, the subcellular fractionation kit for cultured cells (ThermoFisher scientific) was used according to the manufacturer's protocol, and the unsonicated nuclear lysate was loaded.
In vivo transplantation assay 2 × 10 6 viable p53-KO lymphoma cells were transplanted into the flank of each mouse. Briefly, cells were washed 1 time with ice-cold PBS, resuspended in 100 l PBS, and mixed with 100 l Cultrex reduced growth factor basement membrane extract, type 2 (R&D systems) before transplantation for the Polk-KO tumor transplantation experiment. Tumor growth was measured biweekly using caliper measurements, and tumor volume was calculated using the following formula: volume = 0.5 × length × width. When the tumor reached 100 mm 3 , cisplatin was injected intravenously once every two weeks (6 mg/kg) for up to 5 times. Mice were sacrificed when the tumors reached 1500 mm 3 .

Neutral comet assay
Neutral comet assays were performed as described by Olive and Banath (24). Briefly, 8 × 10 3 cells were diluted in 0.4 ml of PBS and were added to 1.2 ml of 1% low-gellingtemperature agarose (Sigma). Subsequently, the cell suspension was transferred onto precoated slides (Menzel-Gläser). Cell lysis was performed in neutral lysis solution (2% sarkosyl, 0.5 M Na 2 EDTA, 0.5 mg/ml proteinase K) at pH 8.0 overnight at 37 • C. Slides were washed 3 times with neutral rinse and electrophoresis buffer (90 mM Tris, 90 mM boric acid, 2 mM Na2EDTA) at pH 8.5, and electrophoresis was performed in neutral rinse and electrophoresis buffer for 25 min at 20 V. Nuclei were stained with 2.5 mg/ml propidium iodide (Invitrogen) in distilled water for 20 min. Pictures of individual cells were taken with a Zeiss AxioObserver Z1 inverted microscope equipped with a cooled Hamamatsu ORCA AG black-and-white CCD camera, and were analyzed with CASP software (http://www.casplab.com). P-values were determined using an unpaired t-test with Welch's correction.

Reconstitution of Polk
HEK293T cells were transfected with the pMX-GFP-mPOLK or pMX-IRES-eGFP plasmids in combination with pCL-Eco as a packaging construct, using polyethylenimine as a transfection reagent. Virus was isolated after 48 h, and frozen at -80ºC until use. To transduce the lymphoma cells, 20 g/ml RetroNectin (Takara) was used to coat nontissue culture plates. Cells and retrovirus were mixed 1:1 (v/v), and cells were centrifuged at 2000 rpm for 90 min at 37ºC. After passaging, cells were sorted based on GFP expression using a FACSAria Illu to obtain transduced cells.

Statistical analysis
Statistical analyses to determine the hits from the highthroughput compound screen were done in R. All other statistical calculations were done in PRISM Graphpad (version 9.0.0).

Generation and validation of new Rev1-deficient mouse strains
We have previously shown the relevance of PCNA-Ub for tolerance of cisplatin (6). Pcna K164R mice were highly sensitive to cisplatin, which ablated the hematopoietic stem and progenitor compartment (6). Because REV1 can function independent of PCNA-Ub, we wanted to examine the relevance of REV1 for ICL tolerance. To this end, we generated genetically defined Rev1-deficient mice. We derived a Rev1-KO C57BL/6 strain using CRISPR/Cas9 that featured a large region of the Rev1 locus spanning exon 4 until exon 11, which encodes around two-thirds of the catalytic domain, and the majority of the BRCT domain (71 of 88 amino acid residues of the BRCT domain are deleted) ( Figure 1A), which could be detected via PCR (Supplementary Figure S1A). The deletion resulted in frameshift mutation that introduced multiple nonsense mutations (Supplementary Figure S1B, C). Moreover, while full length REV1 protein (139KDa) was clearly detectable in MEFs from WT and Pcna K164R mutant mice it was lacking in Rev1-KO mice ( Figure 1B). However, we did observe a protein species in Rev1-KO MEFs of a lower height, that was not clearly detectable in the WT MEFs. This potentially hypomorph protein migrated with a molecular weight of around 60 kDa. To exclude any residual function of a potential hypomorph REV1 in Rev1-KO mice, we performed functional analyses related to its catalytic activity (transferase domain) as well as the contribution of its C-terminus to immunoglobulin heavy chain class switch recombination (CSR). As REV1 substantially contributes to the formation of C/G > G/C transversions during somatic hypermutation (SHM) of immunoglobulin genes (25,26), we performed somatic hypermutation analysis ex-vivo on the JH4 intron region from germinal center B cells from Peyer's patches directly sorted from WT and Rev1-KO mice (Supplementary Figure S1D-H). In contrast to JH4 sequences from WT mice, C/G > G/C transversions were significantly reduced (Supplementary Figure S1E in G > T mutations, which we speculate is due to another SHM-associated protein that takes over for REV1 at guanines and incorporates thymidine rather than cytosines (Supplementary Figure S1E). Moreover, as Rev1-deficiency has been reported to reduce the efficiency of in vitro CSR (25,27), we examined CSR potential of WT and Rev1-KO naïve splenic B cells in vitro stimulated with lipopolysaccharide (LPS), or LPS and recombinant IL-4. Consistent with reported results, we observed a significant reduction in vitro CSR to IgG1 and IgG3 upon LPS, and IL-4/LPS exposure, respectively (Supplementary Figure S1F). Importantly, the CSR function of REV1 has been reported to depend on its recruitment of Ung via its C-terminal domain, suggesting that the C-terminal domain in Rev1-KO mice is non-functional. In addition to the Rev1-KO mouse, we made a Rev1-deficient mouse strain using the Cre-Lox system that included an even larger deletion, spanning from exon 4 until exon 13 ( Figure 1A). Western blots of Rev1-Del MEFs did not reveal the lower hypomorph band, indicating that in these mice, REV1 is no longer produced ( Figure 1B).
Taken together, our results indicate that the targeted deletion in Rev1 locus generated two new Rev1-deficient strains. The large deletions prevented translation of full length REV1 in both mouse strains, and hypomorph bands were not detected in Rev1-Del system. Consistent, at the functional level, the absence of REV1 in Rev1-KO mice resulted in a reduced CSR and expected alterations in the spectrum of SHM. The sum of these data indicates that the Rev1mutant mice are fully deficient for REV1.

Germ-cells depend on PCNA-ubiquitination but not REV1
During embryogenesis starting at E12.5, but preceding meiosis, germ cell development is initiated by the migration of primordial germ cells (PGCs) to the genital ridge. During this time, PGCs are highly dependent on ICL repair, and rapidly undergo apoptosis in its absence (28). Consistent with this notion, FA patients feature reduced fertility (3), and Fanca-KO mice are completely infertile (28). We reasoned, that if REV1 is essential for ICL repair, germ cell maintenance and fertility of Rev1-KO mice should be strongly reduced. Intercrosses of both heterozygous and homozygous Rev1-KO mice produced viable offspring with normal litter size, arguing that Rev1 is dispensable for germ cell development and maintenance ( Figure 1C, D). We were also able to breed homozygous Rev1-Del mice (data not shown). In line, no germ-cell defects were observed in these Rev1-KO and Rev1-Del mice ( Figure 1E). In contrast, as reported previously (14), Pcna K164R mice were completely in-fertile, as indicated by the absence of follicles in the ovaries and sperm and spermatogonia in the seminiferous tubules ( Figure 1E).
Apparently, although REV1 can function independently of PCNA-ubiquitination, REV1 cannot suppress the deleterious effects on germ cell maintenance that arise in the absence of PCNA-K164-Ub dependent DNA damage tolerance (DDT). Furthermore, the lack of fertility-defects in Rev1-deficient mice indicated that REV1 is not essential for ICL repair in germ cells.

In vivo tolerance to cisplatin lesions depends on PCNAubiquitination but not REV1
Cisplatin therapy affects highly proliferating tissues, such as the gastro-intestinal tract, and the hematopoietic system. We have previously reported that cisplatin treatment resulted in bone marrow failure and hematopoietic stem cell depletion in Pcna K164R mice (6). To examine whether Rev1deficiency resulted in similar sensitivity to cisplatin in vivo, we administered a single dose of cisplatin (0.8 mg/kg) intravenously into WT, Pcna K164R and Rev1-KO and Rev1-Del mice, and examined the bone marrow, sternum, and gastrointestinal tract 48 hours later by hematoxylin and eosin (H&E) staining ( Figure 1F-H). Compared to WT mice, cisplatin treatment reduced the levels of nucleated cells in the bone marrow ( Figure 1G) and sternum of Pcna K164R but not Rev1-KO and Rev1-Del mice (Supplementary Figure S2C), indicating that Pcna K164R nucleated cells in the bone marrow and sternum had undergone apoptosis upon cisplatin treatment. We then examined the small intestine and observed apoptotic cells in the crypts and villi of Pcna K164R but not Rev1-KO and Rev1-Del mice, which were recognizable by their pink hue, and increased nuclear condensation and fragmentation ( Figure 1H). This observation in Pcna K164R but not Rev1-KO and Rev1-Del mice revealed a central role for PCNA-ubiquitination in tolerating cisplatin lesions, and that PCNA-ubiquitination and REV1-dependent DDT are likely non-epistatic.
The pancytopenia observed in severe FA patients indicates attrition of HSCs, which suggests that HSCs are highly sensitive to ICLs. Moreover, replication stress and DNA damage directly contribute to HSC demise (6,29,30). To quantitatively determine the relevance of PCNA-ubiquitination for hematopoietic stem and progenitor cells, we analyzed the composition of the bone marrow by flow cytometry after cisplatin treatment (Figure 2A Figure S2B), such as LK cells, LSKs, and HSCs, in Pcna K164R mice, while they were only slightly, and non-significantly reduced in Rev1-KO and Rev1-Del mice (Figure 2A, B). Moreover, the reductions in various subsets, such as HSCs, MPP1s, and MPP2s, that were detected upon cisplatin treatment were nearly identical to those observed in WT mice, indicating that Rev1-KO and Rev1-Del HSCs are not more sensitive to cisplatin.
In conclusion, the absence of a strong reduction of nucleated cells of the bone marrow and sternum, and the absence of apoptotic cells in the gastro-intestinal tract of Rev1-KO and Rev1-Del mice upon cisplatin treatment indicates that Rev1-deficiency does not sensitize to ICLs. REV1 is most likely not required to tolerate cisplatin-induced ICLs in the hematopoietic system, as HSCs and early progenitor subsets of Rev1-KO and Rev1-Del mice are equally sensitive as those of WT mice. While we cannot entirely exclude that the non-significant -and in case of CMPs for Rev1-Del mice significant-reduction of certain hematopoietic subsets in Rev1-KO and Rev1-Del mice was the result of a very minor sensitivity to cisplatin, they may also be the result of increased replication stress in hematopoietic cells that rapidly proliferate to counteract cisplatin-induced apoptosis. Furthermore, our results indicate that in vivo, PCNA-Ub-facilitated DDT is the predominant pathway that tolerates cisplatin lesions and that REV1 cannot compensate for a deficiency in this pathway.

Transformed mammalian cells do not depend on REV1 for cellular fitness and tolerance to cisplatin lesions
REV1 inactivation using the small molecule inhibitor JH-RE-06 (referred from here on REV1i) has been proposed to chemosensitize cancer cells to cisplatin (12,31). To examine the relevance of REV1 during ICL repair in transformed cells in vitro, we determined the sensitivity of WT, Pcna K164R , and Rev1-KO p53-KD MEFs to interstrand crosslinking agents cisplatin and mitomycin C (MMC) by measuring confluency using Incucyte ® assays. Consistent with our in vivo data, Pcna K164R mutant but not Rev1-KO MEFs were hypersensitive to cisplatin and MMC ( Figure  3A). Similar findings were observed using colony survival assays, again highlighting that PCNA-Ub but not REV1 is required for tolerance of cisplatin lesions ( Figure 3B). To examine whether REV1 is required in tolerating cisplatin lesions in cancer cells, we genetically ablated Rev1 in murine lymphoma cells that were described previously (18) using the same CRISPR/Cas9 targeting strategy as used for the Rev1-KO mice. This lymphoma model was derived from p53-KO mice, and feature a complete deletion of p53. We validated the absence of REV1 protein in Rev1-KO lymphomas by western blot (Figure 3C), and analyzed the sensitivity of these isogenic WT, Pcna K164R and Rev1-KO lymphomas to cisplatin. Consistent with our in vivo data and in vitro MEF data, Pcna K164R lymphoma cells were very sensitive to cisplatin, while Rev1-KO lymphomas were indistinguishable from WT ( Figure 3D).
To exclude mouse to human variability with regards to REV1-sensitivity, we knocked down Rev1 with siRNAs in human colorectal and breast cancer cell lines HCT116 and MCF-7, respectively. Analysis by western blot indicated that REV1 levels were reduced by ∼70-90% compared to cells that had been targeted using a non-targeting control ( Figure 3E). We did not observe any significant difference in cisplatin sensitivity in HCT116 and MCF-7 cell lines in response to cisplatin ( Figure 3F). Because p53 regulates TLS (32,33), and HCT116 and MCF-7 feature WT p53 alleles, we analyzed isogenic cell lines where p53 was ablated using CRISPR/Cas9 (19,20). Regardless of the p53 status, a Rev1-KD in HCT116 and MCF-7 did not enhance their sensitivity to cisplatin compared to non-targeting controls ( Figure 3F).
To further exclude a function of REV1 in tolerating cisplatin-induced lesions in human cells, we performed synergy experiments using multiple doses of cisplatin and REV1i in the p53-KO murine lymphoma, human MCF-7 and HCT116, human prostate cancer cell lines 22RV1 and LNCaP, human ovarian cancer cell SKOV3, MEFs, and human melanoma cell line A375 (see Supplementary  table 3A for the concentrations). We did not observe synergy between REV1i and cisplatin in any cancer cell line tested ( Figure 3G), indicating that REV1 inhibition does not chemosensitize mammalian cells to ICLs. Moreover, as expected, we did not observe effects of the REV1i on WT or REV1 mutant MEFs and lymphomas ( Figure 3G, H) or PreBs (data not shown).
Besides a direct role of REV1 in tolerance of genotoxic lesions, recent reports also suggested that it is required for cancer cells to prevent replication stress (34). In this context, REV1 inhibition has been reported to be toxic to cancer cells in unperturbed conditions (34). We therefore examined whether REV1 was required for cell growth. In our Rev1-KO p53-KD MEFs ( Figure 3A) we did not observe any difference in cell growth compared to WT controls. In vivo, we noted a slight delay in the outgrowth of Rev1-KO lymphomas compared to WT controls ( Supplementary Figure S3B, C). However, ultimately all mice had to be sacrificed due to tumor growth twenty days after the tumor reached 100 mm 3 (Supplementary Figure S3C). Moreover, once established, Rev1-KO lymphomas did not grow slower than WT lymphomas, arguing against tumor growth defects (Supplementary Figure S3B).
To examine whether loss of p53 increased replication stress in the absence of REV1, we determined the number of DNA double strand breaks in WT and Rev1-KO primary and p53-KD MEFs using the neutral comet assay. Intriguingly, while in primary MEFs, Rev1-deficiency resulted in a significant increase of the olive tail moment, in p53-KD MEFs the tail moments between WT and Rev1-KO MEFs were significantly reduced (Supplementary Figure  S3D), indicating that in the absence of p53, Rev1-deficiency does not result in increased DNA strand breaks. Importantly, the increased strand breaks in p53-proficient Rev1deficient MEFs argue for a role of Rev1 in DDT, perhaps in the tolerance of non-ICL lesions from endogenous sources.
In conclusion, REV1 is not required to tolerate cisplatin lesions in transformed murine and human cell lines. Moreover, inactivation of p53 does not result in enhanced cytotoxicity of Rev1-deficient cells for cisplatin. Additionally, inactivation of REV1 alone in p53-deficient conditions is not sufficient to reduce cell fitness or slow down tumor growth.
Nucleic Acids Research, 2022, Vol. 50, No. 13 7429  Finally, in murine MEFs and mammalian tumors, PCNA-K164-ubiquitination is the predominant mode of DDT to tolerate genotoxic lesions. Our data do argue for a role of REV1 in DDT in counteracting DNA DSBs.

The TLS polymerase Kappa promotes tolerance of cisplatin lesions in vitro and in vivo
Because REV1 was not required to tolerate cisplatin lesions, we hypothesized that other Y-family polymerases were likely recruited by PCNA-mUb to enable effective TLS of cisplatin lesions. Previous reports have shown POLK to be required to tolerate cisplatin lesions in Xenopus egg extracts and human cells, but results in murine cells (35)(36)(37)(38)(39) were not consistent. We therefore investigated POLK as a candidate and deleted both alleles of Polk by CRISPR/Cas9 in our isogenic murine DDT lymphoma model ( Figure 4A). The Polk-KO was validated by genomic PCR-based analyses and subsequently validated by whole exome sequencing ( Figure 4B, Supplementary Figure S4A-C).
Having established an isogenic Polk-KO lymphoma subclone, we compared its cisplatin sensitivity to the WT parental clone. Importantly, we detected a striking increase in the sensitivity of Polk-deficient lymphoma cells to both cisplatin and the alkylating agent MMS (Figure 4C, D). Retroviral reconstitution of Polk desensitized the Polk-KO lymphoma, indicating that its hypersensitivity was solely dictated by the lack of POLK ( Figure 4C, D). Moreover, requirement for POLK to tolerate genotoxic lesions was further demonstrated by unbiased high-throughput compound screens ( Figure 4E-G, Supplementary Figure S4D-F), where Polk-KO cells were found to be highly sensitive to the ICL-inducing agent RITA (NSC 652287), the alkylating agent nitrogen mustard, and the toxic nucleotide analog Bromodeoxyuridine (BrdU) ( Figure 4E-G), to which Rev1-KO cells were sensitive, although to a less extent than Pcna K164R cells. The sensitivity of the cell lines to several identified compounds was subsequently validated by performing a dose-response curve. The dose-response curve frequently matched the sensitivity observed in the initial screen, further validating the sensitivity of the initial compound screen ( Figure 4H). However, in the dose response curve of RITA (NSC 652287), we did not detect enhanced sensitivity of Rev1-KO cells, arguing against the relevance of REV1 in tolerating RITA-induced lesions ( Figure 4H). Intriguingly, we picked up and validated a strong insensitivity of Pcna K164R cells, though not PolK-KO cells, to 5-Fluorouracil (5-FU), indicating that while POLK is required to tolerate ICLs, toxicity of 5-FU is caused by PCNA-ubiquitination independent of POLK ( Figure 4H). Intriguingly, we identified Pcna K164R , Rev1-KO and Polk-KO clones to be sensitive to various types of steroid hormones, such as 17a-hydroxytamoxifen and progesterone (Supplementary Figure S4G-I), which are known to induce DNA adducts that DDT-deficient DT40 cells are sensitive to (40). The steroid-hormone related hits suggest that specific TLS polymerases are required to tolerate specific adducts created by steroid hormones, and finally suggests the specific type of enigmatic endogenously produced lesions that REV1 and POLK are required to tolerate. Importantly, the data indicate the high relevance of POLK to tolerate DNA damage induced by many classical chemotherapeutic agents and suggests that POLK is a key TLS polymerase that tolerates a wide range of genotoxic lesions.
Because Polk-KO cells were highly sensitive to ICLinducing agents, we tested Polk-deficiency as a cancer specific vulnerability to cisplatin treatment. We subcutaneously transplanted isogenic WT and Polk-deficient lymphomas in nude mice to exclude any immune-mediated rejections. Tumor growth was followed using caliper measurements. Regarding vehicle controls, WT and Polk-KO lymphoma grew out rapidly and established large tumors within seven days. Consequently, all mice of the vehicle control groups had to be sacrificed within seven days after the start of treatment, after the tumors had reached a maximum allowed size of 1500 mm 3 ( Figure 4I, J).
Notably, while the WT lymphomas could not be controlled by a clinical dose of cisplatin, eight out of nine mice that had received Polk-deficient lymphomas survived five cycles of cisplatin treatment, and the tumor size reduced to nearly undetectable levels during the treatment phase ( Figure 4I, J). In seven out of nine Polk-KO tumor bearing mice, the tumors grew out within fifty days after cessation of treatment. In two out of nine mice however, tumors remained undetectable for up to two hundred days after the start of treatment. As homozygous inactivating deletions of Polk are relatively frequently observed in human cancers, such as prostate, ovarian, and pancreatic cancers as well as melanoma ( Figure 4K), our data suggest that these patients may benefit from treatments with platinating agents.
In conclusion, the genetic knockout of Polk in a p53-KO murine lymphoma strongly enhanced the cytotoxicity of cisplatin, resulting in marked increased in the overall survival time of Polk-KO tumor bearing mice that were treated with cisplatin. This preclinical mouse model reveals a unique opportunity for precision cancer medicine in predicting cisplatin responsiveness based on the Polk status.

DISCUSSION
Knowledge on ICL repair has expanded greatly in the past few years. The direct role of TLS in FA-mediated ICL repair highlights DDT as a critical arm of the DNA damage response network. Inhibition of DDT is being considered to enhance the treatment efficacy of platinating agents to treat cancer. These findings implicated that targeting of specific DDT components provides an attractive mode to chemosensitize tumors to platinating agents (12,13,18,31,34,41). PCNA-mUb and REV1 are considered as essential to recruit TLS polymerases (8)(9)(10)42). However, the division of labor between PCNA-Ub-dependent DDT and REV1-dependent DDT in mammalian cells has been unclear. Here, we directly compared the relevance of PCNAubiquitination-dependent and REV1-dependent DDT in mice as well as human and mouse cell lines. Our studies in mice and in mammalian cells strongly indicate PCNAubiquitination is the predominant mode of DDT, and that a genetic ablation of REV1 does not sensitize murine and human cancer cells to various classes of clinically approved DNA damaging agents. Along these lines, our analyses in the mammalian system exclude REV1 as being critical in coordinating replicative bypass of cisplatin-induced DNA lesions. Instead, our data identify the TLS polymerase POLK as an important DDT component that tolerates cisplatin lesions. Lack of POLK but not REV1 rendered human and mouse cell lines hypersensitive to cisplatin treatment. Loss of Polk strongly amplified the sensitivity to a cisplatin-monotherapy in an isogenic lymphoma transplantation model, resulting in robust longitudinal control of tumor outgrowth in all mice, as well as a total tumor clearance in two out of nine mice. Stable reconstitution of Polk rescued cisplatin sensitivity of Polk-KO lymphomas. According to the TCGA database, homozygous inactivation mutations in the Polk gene are relatively frequent in human prostate cancer, melanoma, and endometrial cancers. This notion in combination with our findings implies that the Polk status provides a novel parameter in predicting a tumor-specific hyper-responsiveness to cisplatin and screening of the Polk status in cancers likely provides an attractive avenue for personalized cancer medicine.
We previously reported that PCNA-ubiquitination is essential for the tolerance of ICLs, in vivo and in vitro (6,43). Indeed, Pcna K164R mice were hypersensitive to cisplatin, as indicated by a strong and significant reduction of nucleated cells in the bone marrow and sternum and the presence of elevated numbers of apoptotic cells in the gut. More-over, murine primary, immortalized, and transformed cells that carried the Pcna K164R mutation were hypersensitive to various genotoxic compounds in vitro, such as cisplatin and MMC. These data agree with previous reports in human and avian cell lines (9,44), indicating a central role of PCNA-Ub dependent DDT in higher organisms.
In contrast, this direct comparative study between Pcna K164R and Rev1-deficient systems revealed that REV1 is dispensable for the tolerance of ICLs in the mammalian system. Our results are somewhat puzzling given the many reports indicating the relevance of REV1 in ICL repair (not extensive: 12,31,34,45,46). However, we provide several lines of evidence in vivo, ex vivo and in vitro to corroborate this important conclusion. First, in our defined Rev1-KO mouse model, we were able to recapitulate the SHM nucleotide substitution pattern and reduced level of CSR that were established prior in independent Rev1-deficient mouse models (25)(26)(27). Second, we did not observe significant decreases of various hematopoietic subsets of cells in the bone marrow of Rev1-deficient mice as compared to WT mice after cisplatin treatment, while Pcna K164R mice had strong reductions of all subsets measured. Third, we noted that Pcna K164R mice, but not Rev1-mutant mice were infertile. Germ cells critically depend on ICL repair, as indicated by the absence or reduction of germ cells when ICL repair is impaired (28), and the high sensitivity of germ cell-derived tumors to cisplatin (47). Our findings further demonstrate that cells most sensitive to ICLs, i.e. stem cells, are not affected by loss of Rev1. The use of genetically defined mouse models and independent cell lines thereof exclude confounding effects that may result from genetic drifts. Fourth, the absence of sensitivity to ICLs in vitro of multiple Rev1-deficient murine and human transformed, and non-transformed cell lines argue against any supposed sensitivity being masked due to a difference in species, genetic makeup of the cell line, or whether cells are transformed or not. Finally, the multiple manners in which REV1 function was attenuated, either by CRISPR/Cas9mediated KO, siRNA-mediated KD, or via direct inhibition using small molecule inhibitor JH-RE-06, excludes that effects are masked by isolation of a particular resistant clone, and further indicated that REV1 is not essential for tolerating ICLs. Of note, while cisplatin and MMC are potent inducers of interstrand crosslinks, the majority of lesions introduced are in fact the much less toxic intrastrand crosslinks and adducts. While we cannot clearly distinguish the effects of interstrand and intrastrand crosslinks in our assays, our in vivo and in vitro results clearly demonstrate that REV1 is not important to tolerate ICLs. Our data indicate that inhibition of DDT to treat cancer would carry the risk of introducing potentially severe adverse effects, such as bone marrow failure, or enhanced levels of DNA damage (6). However, our data also imply that tumor-specific DDT defects provide unique opportunities for precision cancer medicine. Indeed, we observed a strong enhancement of cisplatin cytotoxicity in Polk-deficient lymphoma cells, which could be rescued by retroviral transduction of Polk. Our data agree with findings of Gautier et al., who demonstrated that POLK was essential for replication through unhooked ICLs (38). Interestingly, POLK appears to function in every phase of the cell cycle (38,41,48,49), increasing the range and efficacy of potential treatments targeting Polk-deficient tumors. Especially, the G1 activity would allow for the targeting of non-cycling dormant cancer stem cells, which would otherwise be protected from the cytotoxic effects of cisplatin and which could reoccur after cessation of treatment (50). In addition, as Polk-deficient cancer cells are sensitive to different classes of DNA damaging agents, recurrence due to acquired resistance of such cancer cells to one class of cytotoxic drugs could be mitigated by switching to a different class of DNA damaging agents. Importantly, we established the Polk-KO clone via limiting dilution, a process which takes roughly two months. During this time, and during culturing, Polk-KO cells did not become resistant to cisplatin, or developed other resistance mechanisms, which probably contributed to the extensive sensitivity of Polk-KO lymphomas to cisplatin in vivo. Intriguingly, because cisplatin introduces a plethora of different lesions in the DNA, perhaps the high sensitivity of Polk-deficient cells to cisplatin can also be explained by the requirement for POLK to tolerate a wide range of genotoxic lesions. For metastatic prostate cancer, current standard of treatment consists of a combination of androgen-receptor inhibitors and taxanes (51), which leaves platinating and alkylating agents open as an alternative treatment option, or as an adjuvant therapy.
Inhibition of DNA damage tolerance pathways is currently under intense investigation. Indeed, preclinical data shows promising results for the inhibition of RAD6B (13), or the RPA ubiquitination enzyme RFWD3 (52), both of which enhance the cytotoxicity of cisplatin. Other TLS polymerases besides POLK that could be required for ICL repair could be POLH, which is able to extend past ICLs (45,53,54) and whose knockout sensitizes cells to cisplatin, and REV3/REV7 (45,(55)(56)(57), as Rev3-KO cells are hypersensitive to DNA damage. These results also indicate the non-redundant requirement of multiple TLS polymerases and DDT pathways to tolerate cisplatin lesions. In line with this observation, we observed that Pcna K164R cells were more sensitive than Polk-deficient cells for cisplatin, arguing that distinct DDT-pathways or TLS polymerases act on cisplatin lesions. Perhaps the differential requirement of TLS polymerases to tolerate cisplatin lesions depend on the genomic context, cell cycle, or alternatively, distinct polymerases operate sequentially.
Platinating and alkylating agents are mainstays in current cancer treatments, and their synergistic action in combination with immune checkpoint inhibitors has resulted in renewed interest in the use of these compounds for the treatment of solid and hematological malignancies (58). Our data indicate that cancer-specific defects in the DDT system represent clinically actionable targets that can be exploited for personalized cancer medicine. In conclusion, we show that PCNA-ubiquitination, but not REV1, is required for tolerance of cisplatin induced ICLs in mammalian cells. Future studies should examine whether the loss of Polk in human cancers can be exploited in human cancer therapy and predict responsiveness to platinating agents, and whether cisplatin sensitivity of Polk-deficient cells is caused by sensitivity to cisplatin-induced intrastrand or interstrand crosslinks.

DATA AVAILABILITY
Flow Cytometry data has been deposited on the FlowRepository website (see https://flowrepository. org/id/FR-FCM-Z5K6). Whole genome sequencing data of a WT p53-KO lymphoma clone, and whole exome sequencing data of the PolK-KO clone can be found at: https://dataview.ncbi.nlm.nih.gov/object/PRJNA854140? reviewer=k50bkvh4lnqn09n1a5hp4pl2g1, using accession numbers SRR19901279 (for the WT lymphoma clone), and SRR19901280 (for the Polk-KO clone).