Dual role for mammalian DNA polymerase ζ in PNAS PLUS maintaining stability and proliferative responses

Sabine S. Langea, Ella Bedforda, Shelley Reha, John P. Wittschiebenb, Steve Carbajalc, Donna F. Kusewitta,d, John DiGiovannic, and Richard D. Wooda,d,1

aDepartment of Molecular Carcinogenesis, The University of Texas MD Anderson Cancer Center Science Park, Smithville, TX 78957; bNovo Nordisk Foundation, 2900 Hellerup, Denmark; cDell Pediatric Research Institute, The University of Texas, Austin, TX 78723; and dGraduate School of Biomedical Sciences at Houston, Smithville, TX 78957

Edited by Philip C. Hanawalt, Stanford University, Stanford, CA, and approved January 11, 2013 (received for review October 5, 2012) DNA polymerase ζ (polζ) is critical for bypass of DNA damage and the including ribonucleotides accidentally incorporated during DNA associated mutagenesis, but also has unique functions in mammals. synthesis (8). It is required for embryonic development and for viability of hema- S. cerevisiae can survive without polζ, but the clearly has topoietic cells, but, paradoxically, skin epithelia appear to survive unique functions in mammals because Rev3L-knockout mouse polζ deletion. We wished to determine whether polζ functions in embryos die in midgestation (9–11). To circumvent this difficulty a tissue-specific manner and how polζ status influences skin tumor- and to investigate the function of polζ in vivo, mice containing igenesis. Mice were produced in which Rev3L (the catalytic subunit loxP recombination sites flanking essential Rev3L exons have been of polζ) was deleted in tissues expressing keratin 5. Efficient epider- used for conditional disruption of polζ function. In a previous study mal deletion of Rev3L was tolerated but led to skin and hair abnor- we used Cre controlled by the mouse mammary tu- malities, accompanied by evidence of DNA breaks. Unchallenged mor virus (MMTV) , which expresses in a mosaic fashion mice developed tumors in keratin 5-expressing tissues with age, in hematopoietic and epithelial cells. Normal hematopoietic cells consistent with the chromosomal instability accompanying a polζ did not survive MMTV-Cre–mediated Rev3L disruption, even in Rev3L defect. Unexpectedly, mice with the deletion were much more a p53-defective background (12). Similarly, inactivation of Rev3L sensitive to UVB radiation than mice defective in other DNA repair by CD21-Cre or CD19-Cre caused a loss of mouse B-cell viability ζ genes. Following irradiation, pol -defective mice failed to mount (13, 14). On the other hand, a mosaic of viable Rev3L-deleted and skin-regenerative responses and responded to stress by mobilizing nondeleted cells was present in skin epithelia of MMTV-Cre mice. melanocytes to the epidermis. However, they did not develop skin Cells that had deleted Rev3L were present in adult ear, skin, sali- tumors after chronic UVB irradiation. To determine the proliferative vary gland, and mammary gland (12). Therefore, deletion of Rev3L potential of polζ-deficient skin epithelia, keratinocytes were iso- in epidermally derived mammalian cells unexpectedly appeared lated and examined. These keratinocytes harbored chromosomal compatible with viability. gaps and breaks and exhibited a striking proliferation defect. These These observations raise several important questions. First, are results can be unified by a model in which slowly dividing cells ac- the requirements for polζ tissue specific, so that epidermal cells cumulate replication-associated DNA breaks but otherwise survive have a requirement for polζ that is fundamentally different from Rev3L deletion, but functional polζ is essential for responses requir- that of hematopoietic cells? Second, what is the importance of ing rapid proliferation, both in cell culture and in vivo. The results ζ reveal a biological role for mammalian polζ in tolerating DNA mammalian pol in defense against UV radiation in vivo? In- damage and enabling proliferative responses in vivo. vestigation of radiation sensitivity in mammalian cells has been limited previously to immortalized cells in culture. Third, if polζ DNA replication | double-strand breaks | UV radiation | carcinogenesis Significance ast, efficient genomic duplication requires that DNA be com- Fpletely intact and in the B form, because replicative DNA In mammalian cells DNA polymerase ζ (polζ) appears critical for polymerases cannot synthesize using a damaged DNA template bypass of DNA damage and was expected to be important for (1, 2). When such a template is encountered, replication halts, UV-induced skin carcinogenesis. To investigate the response to ζ and either a double-strand break (DSB) forms after replication UV radiation, we engineered mice lacking pol in the epidermis, fork collapse or the lesion is bypassed by translesion synthesis circumventing a requirement for embryonic development. These (TLS) polymerases or by template switching. After such damage mice were much more sensitive to UVB radiation than predicted, tolerance,theDNAcanberepaired.Therelativeimportanceof failed to mount skin-regenerative responses, and did not de- ζ fi each pathway for maintaining genomic stability and preventing velop UV-induced skin tumors. Even unirradiated pol -de cient keratinocytes had a marked proliferation defect and increased carcinogenesis is unknown. ζ TLS is mediated by specialized DNA polymerases (reviewed in chromosomal breaks. Thus in rapidly proliferating cells, pol ref. 3). DNA polymerase ζ (polζ, catalytic subunit REV3L) stands maintains levels of DNA breaks below a lethal threshold. out as the most important DNA polymerase for bypass of lesions in GENETICS ζ Author contributions: S.S.L., E.B., J.P.W., S.C., D.F.K., J.D., and R.D.W. designed research; template DNA. Pol plays a major role in the bypass of many types S.S.L., E.B., S.R., and S.C. performed research; S.C. and J.D. contributed new reagents/ of DNA damage, including pyrimidine(6-4)pyrimidone photo- analytic tools; S.S.L., E.B., S.R., J.P.W., D.F.K., and R.D.W. analyzed data; and S.S.L., J.P.W., products induced by UV radiation (4, 5) as well as lesions formed by and R.D.W. wrote the paper. chemical damaging agents such as cisplatin and benzo[a]pyrene (4). The authors declare no conflict of interest. Polζ also can be used for bypass of the frequent endogenously This article is a PNAS Direct Submission. formed abasic sites in DNA (4). In the yeast Saccharomyces cer- Freely available online through the PNAS open access option. evisiae, spontaneous and DNA damage-induced mutagenesis is 1To whom correspondence should be addressed. E-mail: [email protected]. ζ highly dependent on pol function (6), and the enzyme is involved This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. in the bypass of common impediments to DNA replication (7), 1073/pnas.1217425110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1217425110 PNAS | Published online February 5, 2013 | E687–E696 Downloaded by guest on October 2, 2021 is important for bypass of UV radiation damage, how does it Results modulate skin carcinogenesis in mammals? Mice Survive Epithelial Deletion of Rev3L but Have Skin Tissue Ab- Humans and mice defective in the DNA nucleotide excision normalities. Mice containing a floxed allele of Rev3L (12, 15) cou- repair (NER) pathway (NER) or in DNA polymerase η (polη) pled with either a wild-type allele or a knockout Rev3L allele (Fig. have a greatly elevated predisposition to UV radiation-induced skin 1A) were bred with transgenic mice expressing Cre recombinase cancer. A major contributor to thistumorigenesisistheaccumu- from a keratin 5 promoter (BK5.Cre mice). This promoter drives lation of point in key driver genes. Polζ-depleted cells in gene deletion at high efficiency (≥95%) in the basal epithelia culture are reported to have reduced UV-induced point mutations of skin epidermis, hair follicles, sebaceous glands, thymus, and (5, 6). It is possible that suppression of UV-induced mutagenesis mammary gland (17, 18), but not in hematopoietic tissues. The in key genes for skin carcinomas would diminish tumorigenesis in efficiency of Cre activity was monitored using the membrane- polζ-deleted epidermis. On the other hand, the observed increase Tomato/membrane-Green (mT/mG) Cre reporter gene (15, 19) in the frequency of chromosome rearrangements in Rev3L-deleted (Fig. 1B) and by PCR of genomic DNA (Fig. S1). Control BK5. cells (13, 15, 16) might accelerate cancer development. Mice effi- Cre;Rev3L+/lox and BK5.Cre;Rev3L+/+ mice retained one or two ciently deleting Rev3L in keratin 5-expressing epithelia were used functional alleles of Rev3L, respectively, while still expressing in this study, allowing us to ask definitively whether normal epi- keratin 5-driven Cre recombinase. − thelial tissues can develop and renew normally in vivo in the ab- Mice deleting polζ in basal epithelium (BK5.Cre;Rev3L /lox) sence of polζ.Wefind a major biological role of polζ in the defense were viable but were underrepresented at weaning (Fig. 1C). In against UV radiation-induced DNA damage in vivo. The results addition, they displayed defects in skin and hair growth. The BK5. − show that mammalian polζ is important not only for tolerating Cre;Rev3L /lox pups were recognizable by a thinner hair coat (Fig. DNA lesions but also for allowing cells to proliferate after damage. 1D), with hair becoming more normal in appearance by 8 wk of

AB

C

DE

Fig. 1. Abnormal skin epithelia and increased morbidity in aging Rev3L mice. (A) Schematic of (i) wild-type, (ii) knockout, (iii) floxed-intact, and (iv) floxed-deleted alleles of the murine Rev3L gene (12, 15). Vertical bars represent exons (red bars contain part of polymerase motif I and all of motif V); tri- angles are loxP sites. (B) mT/mG skin and salivary gland expressing RFP (no Cre activity) or GFP (Cre activity) in BK5.Cre transgenic mice. (C) Genotype percentages from a cross of male BK5.Cre;Rev3L+/lox − FG and female Rev3L+/ mice. The BK5.Cre and the floxed Rev3L allele are linked on chromo- some 10. (D) Representative photograph of 12-d-old − mouse pups with the genotypes BK5.Cre;Rev3L /lox − and BK5.Cre;Rev3L+/ .(E) H&E-stained skin sections from 7- to 10-wk-old mice. Arrows show the inter- follicular epidermis, and arrowheads point to hair follicles. (Scale bars: 100 μm.) (F) Quantification of total interfollicular epidermal cells per millimeter of skin from BK5.Cre;Rev3L−/lox, BK5.Cre;Rev3L+/lox,and BK5.Cre;Rev3L+/+ mice (n = 5). (G) Kaplan–Meier − morbidity-free survival curve of BK5.Cre;Rev3L /lox and BK5.Cre;Rev3L+/lox mice with increasing age (n = 19). *P < 0.05. Data points are mean ± SEM.

E688 | www.pnas.org/cgi/doi/10.1073/pnas.1217425110 Lange et al. Downloaded by guest on October 2, 2021 age. Rev3L-deleting mice did not maintain a synchronous hair keratin 5-expressing tissues and thus were deleting Rev3L. In each PNAS PLUS cycle and showed degenerative changes in hair follicles as they type of sebaceous gland, there was a continuum of lesions ranging aged (Figs. 1E and 2B). The interfollicular epidermis of the ex- from hyperplasia to frank neoplasia. Even in Rev3L-deleting mice perimental mice displayed significantly fewer cells per millimeter of that did not have tumors in these sebaceous glands, these organs skin (Fig. 1F) compared with control mice. Thus, deletion of Rev3L often had inflammatory and dysplastic changes not seen in control leads to a less cellular epidermis and to hair follicle abnormalities. mice. The body weights of Rev3L-deleting and control mice at the time of death were not significantly different (31.0 ± 1.3 g for BK5. Aged Rev3L-Deleting Mice Develop Squamous Cell Carcinomas. To − Cre;Rev3L /lox mice and 36.2 ± 1.7 g for BK5.Cre;Rev3L+/lox determine the long-term phenotypic effects of an epithelium- fi Rev3L Rev3L−/lox mice); however, the spleens of the experimental mice were four speci c knockout, experimental BK5.Cre; and ± control BK5.Cre;Rev3L+/lox mice were monitored as they aged. times larger than the spleens of control mice (3.4 0.52% of total Rev3L−/lox ± The experimental mice displayed a shortened morbidity-free sur- body weight for BK5.Cre; mice and 0.76 0.08% of total Rev3L+/lox P < vival (Fig. 1G) compared with the controls. The reason for this body weight for BK5.Cre; ; 0.001). The larger shorter survival was primarily the development of squamous cell spleens may be a response to increased inflammation in sebaceous carcinomas (SCCs) in specialized sebaceous glands (clitoral and glands and tumors. Therefore, the loss of epithelial Rev3L drives preputial glands, Meibomian glands of the eyelid, and Zymbal’s tumor development and promotes inflammation in specialized glands of the ear) and skin (Fig. 2 A and B and Fig. S2), which sebaceous glands. In addition, 3 of 22 experimental mice had fully developed in ∼90% of the mice. All these target organs contain developed lymphoma or plasmacytoma by age 390 d, whereas

A

B

Fig. 2. Epithelial carcinomas in skin deleting Rev3L and suppressed response to TPA treatment. (A) Pathological findings in BK5.Cre;Rev3L mice. One mouse with plasmacytoma was not examined for the other features. In a few mice from each group, some of the small sebaceous glands could not be located in histological sections and were not assessed for pa- CD E GENETICS thology. (B) Representative histologic sections from normal skin or Zymbal’sglandandfromtumorsin BK5.Cre;Rev3L−/lox mice. (Scale bars: 100 μm.) (C–E) − BK5.Cre;Rev3L /lox and BK5.Cre;Rev3L+/lox mice were treated with 0.85, 1.7, or 3.4 nmol TPA twice weekly for 2 wk (n =4).(C) Percent BrdU-positive cells, normalized to vehicle control. (D) Total interfollicular epidermal cell number per millimeter of skin, normal- ized to vehicle control. (E) Epidermal thickness after TPA treatment. *P ≤ 0.05; **P < 0.01. Data points are mean ± SEM.

Lange et al. PNAS | Published online February 5, 2013 | E689 Downloaded by guest on October 2, 2021 control mice >491 d of age had no fully developed instances (al- of responses (21), including protective thickening and replacement though 5 of 20 control mice had indications of possible early of damaged cells. We assessed the response of polζ-deleted epi- lymphoma by this time). thelia to UVB radiation. Unexpectedly, Rev3L-deleting skin is To test the response to moderate inflammation, the skin of extremely UV sensitive. Initial UV radiation-induced epidermal Rev3L-deleting and control mice was treated with the tumor-pro- damage was followed by massive epidermal disruption because the moting agent 12-O-tetradecanoylphorbol-13-acetate (TPA) (20). skin and hair follicles were unable to regenerate normally. − Although higher concentrations of TPA enhanced alopecia in the BK5.Cre;Rev3L /lox,BK5.Cre;Rev3L+/lox,andBK5.Cre;Rev3L+/+ − BK5.Cre;Rev3L /lox mice, it did not induce tumor formation after mice were irradiated with a single dose of 1,800 J/m2 UVB (ap- 40 wk of treatment. Treatment with TPA two times per week for proximately twice the minimal erythemal/edemal dose for the 2 wk stimulated epidermal DNA synthesis by fourfold in control FVB/N strain), and skin sections were examined in cohorts of mice mice but by less than 1.5-fold in Rev3L-deleting mice (Fig. 2C). after 1–10 d. For the first few days, both control and experimental There was a small but significant increase in epidermal cell number mice showed moderate erythema and edema, with skin in control in both control and experimental mice after treatment with the mice resuming normal appearance and thickness by day 8. Ery- highest concentration of TPA (Fig. 2D), but there was less epi- thema, edema, and inflammatory cell infiltration became pro- dermal thickening in the Rev3L-deleting skin (Fig. 2E). These data gressively more severe in experimental mice. Starting at 48 h there suggest a lack of sustained replicative potential in Rev3L-deleting was marked degeneration of the epidermis and hair follicles with skin in response to TPA. As in normal tissues, an inflammatory widespread ulceration in irradiated areas by 8–10 d (Fig. 3 A and stimulus alone does not suffice to induce tumorigenesis in Rev3L- B). Five days after UV irradiation, numerous dramatic changes deleting skin. were evident in polζ-deleting skin. In some areas there was epi- dermal thickening without an increase in cell number (described Rev3L-Deleting Mice Are Extremely Sensitive to UV Radiation. UV below). In other areas the epidermis was completely lost. There radiation of normal mouse skin causes a well-established cascade also were small foci of epidermal proliferation. Higher radiation

A

B CD

Fig. 3. Acute UV radiation severely affects Rev3L-deleting skin. (A) Representative H&E-stained skin sections from 7- to 10-wk-old BK5.Cre;Rev3L−/lox and BK5.Cre;Rev3L+/lox mice at 24, 48, 72, 120, or 192 h after 1,800 J/m2 UVB radiation. (Scale bars: 100 μm.) Arrows point to areas demonstrating inflammatory cell − influx. (B) Representative H&E-stained skin sections from 7- to 10-wk-old BK5.Cre;Rev3L /lox mice showing additional observed phenotypes at 120 h after 1,800 J/m2 UVB radiation. (Scale bars: 100 μm.) (C and D) Quantification of total interfollicular epidermal cells per millimeter of skin (C) or epidermal thickness − (D) from BK5.Cre;Rev3L /lox, BK5.Cre;Rev3L+/lox, or BK5.Cre;Rev3L+/+ mice that were untreated (No UV) or at various times after UVB radiation (n = 4). *P < 0.05; **P < 0.01. Data points are mean ± SEM.

E690 | www.pnas.org/cgi/doi/10.1073/pnas.1217425110 Lange et al. Downloaded by guest on October 2, 2021 doses produced even more rapid and severe epidermal loss in appeared. In marked contrast, hair follicles in control mice showed PNAS PLUS polζ-deleting mice. no morphologic evidence of damage (Fig. 3A). UV radiation-induced epidermal thickening in experimental mice during the first 24–120 h was caused by cell swelling and Chronic UV Radiation Is Poorly Tolerated in Rev3L-Deleting Skin and intercellular edema rather than by an increase in cell number Does Not Induce Tumors. Other mouse models that show an acute (Fig. 3 C and D). The DNA damage and stress markers γ-H2AX sensitivity to UV radiation include those with deficiencies in genes and p53 (measured by immunohistochemistry) increased in all responsible for NER or the TLS DNA polη. When chronically UV mice after UVB radiation (Fig. 4 A–C). Expression of these irradiated, such mice develop SCCs much more rapidly than their stress markers persisted longer at a higher level in polζ-deleting wild-type littermates (Table S1). To determine if a lack of Rev3L skin. Epidermal proliferation as measured by BrdU incorpora- leads to an increased incidence of UV radiation-induced tumor tion was enhanced 48 h after irradiation in control mice but not formation, chronic UV irradiation experiments were initiated fol- in experimental mice (Fig. 4E). Caspase-3 staining was seen in lowing standard protocols. We found that a low-dose exposure of the epidermis of control and experimental mice (Fig. 4D), in- Rev3L-deleting mice (50 J/m2 UVB radiation three times weekly) dicating apoptosis at frequencies comparable to those observed was barely tolerated. Even at this very low dose, Rev3L-deleting mice previously at similar doses of UVB radiation (21, 22). Much sometimes showed excessive reddening of the dorsal skin; these mice subsequent epidermal loss in Rev3L-deleting mice may be caused were removed from the study and returned when they had recovered. by nonapoptotic mechanisms. No experimental or control mice developed tumors after 52 wk The behavior of melanocytes after UV irradiation was examined of this low-dose regimen. Control mice were not notably affected − in pigmented mice of mixed C57BL6/J and FVB/N background. by the low dose, but BK5.Cre;Rev3L /lox mice were progressively − BK5.Cre;Rev3L /lox animals developed pronounced epidermal lost from the study because of the formation of nonhealing cuta- pigmentation. This tanning response, unusual for mice, was asso- neous ulcers (Fig. 5E). Irradiation of Rev3L-deleting mice with ciated with migration of melanocytes into the epidermis of irradi- higher doses (e.g., 100 J/m2 three times weekly) caused extensive ated Rev3L-deleting skin (Fig. S3). Melanocytes do not delete skin ulceration and was not tolerated for longer than 5 wk. As Rev3L with K5.Cre expression, and so the observed melanocyte a control for UV radiation-induced skin tumor formation, BK5. migration apparently occurs in response to signals initiated fol- Cre;Rev3L+/+ mice were irradiated with a higher-dose protocol of lowing damage to polζ-deleting keratinocytes. UVB for up to 50 wk, which induced SCCs in 7 of 16 mice. Pronounced disintegration of hair follicles occurred after irra- Patches of epidermal cells with stabilized, mutant p53 ex- diation of polζ-deleting skin. Increased DNA synthesis in hair pression commonly occur in response to UV radiation and are follicles was observed 24 h after UV radiation but declined rapidly believed to represent cell populations from which skin tumors thereafter (Fig. 5 A and B), whereas DNA-break signaling in- emerge (23). The p53 mutations seen in these patches are point creased concomitantly (Fig. 5 C and D). Hair follicles then dis- mutations detected with antibody PAb240, which recognizes

AB

C

D

Fig. 4. Acute UV radiation causes activation of epi- dermal p53 and γ-H2AX but does not increase pro- liferation in Rev3L-deleting mice. (A) Representative immunohistochemical sections of γ-H2AX–,p53-, GENETICS E caspase 3-, or BrdU-stained skin sections from 7- to − 10-wk-old BK5.Cre;Rev3L /lox and BK5.Cre;Rev3L+/lox mice 48 h after 1,800 J/m2 UVB radiation. Arrows point to representative positively stained cells. (Scale bars: 100 μm.) (B–E)Quantification of percent of γ-H2AX– (B), p53- (C), caspase 3- (D), or BrdU- (E) positive interfollicular epidermal cells in skin sections − from BK5.Cre;Rev3L /lox,BK5.Cre;Rev3L+/lox,orBK5. Cre;Rev3L+/+ mice that were untreated (No UV) or at various times after UVB radiation (n =4).*P < 0.05; **P < 0.01. Data points are mean ± SEM.

Lange et al. PNAS | Published online February 5, 2013 | E691 Downloaded by guest on October 2, 2021 A

B C

DE

Fig. 5. Acute UV radiation causes activation of fol- licular p53 and γ-H2AX, and chronic UV radiation causes skin loss. (A) Representative immunohisto- chemical sections of BrdU-stained skin from 7- to 10-wk-old BK5.Cre;Rev3L−/lox and BK5.Cre;Rev3L+/lox mice 24 h after a single dose of 1,800 J/m2 UVB ra- diation. Arrows point to representative positively stained cells. (Scale bars: 100 μm.) (B–D)Quantifica- tion of percent of BrdU- (B), γ-H2AX– (C), or p53- (D) positive hair follicle cells in skin sections from BK5. Cre;Rev3L−/lox,BK5.Cre;Rev3L+/lox, or BK5.Cre;Rev3L+/+ mice that were untreated (No UV) or at various times after UVB radiation (n =4).(E) Kaplan–Meier mor- F − bidity-free survival curve of BK5.Cre;Rev3L /lox,BK5. Cre;Rev3L+/lox, and BK5.Cre;Rev3L+/+ mice with in- creasing UV radiation dose (n =17).(F) p53 mutant patches in BK5.Cre;Rev3L mice. frequency is calculated as the length of p53 mutant focus per total skin length. *P < 0.05; **P < 0.01. Data points indicate mean ± SEM.

a commonly exposed epitope in a mutant conformation created epithelium to UV radiation, we isolated keratinocytes from the by a variety of substitutions. Complete inactivation skin of BK5.Cre;Rev3L mice. We hypothesized that they would of p53 is not detected by this approach. In the present studies, not be able to thrive following a stimulus that normally would no p53 mutant patches as detected by PAb240 staining were induce rapid proliferation. Within 5 d of isolation from skin, − observed in the skin of control mice at low cumulative doses there were far fewer BK5.Cre;Rev3L /lox keratinocytes than of UVB (4,000 J/m2), but many such patches were observed after control cells, and senescence was apparent (Fig. 6 A–C), with high-dose UV radiation (>150 kJ/m2). In Rev3L-deleting mice more cells harboring multiple nuclei (Fig. S4)andfewerun- some p53 mutant patches were observed after low doses of UV dergoing DNA synthesis (Fig. 6D). Potent proliferation signals radiation (Fig. 5F). Even though there is a decreased rate of therefore invoke a senescence barrier in polζ-defective kera- point mutagenesis in the absence of Rev3L, such p53-defective tinocytes, as they do in mouse embryonic fibroblasts (15). To − clones should have a significant selective advantage for pro- determine whether epithelial cells from BK5.Cre;Rev3L /lox mice liferation of Rev3L-deleted cells (15). However, UVB irradiation harbor chromosomal aberrations, metaphase spreads were ana- dramatically compromises epidermal integrity in Rev3L-deleting lyzed from keratinocytes of 3-d-old mice. There was an eightfold mice, suggesting progression of these patches to form a tumor is increase in the number of gaps and breaks (chromatid-type prevented by the high radiation sensitivity and severely impaired aberrations) per metaphase, compared with controls (Fig. 6 E proliferative potential of the polζ-deleting cells and tissues. and F). There was no significant difference in chromosomal- type aberrations (dicentrics and double minutes; Table S2)or Intrinsic Renewal Defect and Accumulation of Endogenous DNA in aneuploidy and polyploidy (Fig. S5). Consistent with the ob- Damage in Rev3L-Deleting Keratinocytes. To clarify the mecha- servation of gaps and breaks in the chromosomes, the BK5.Cre; − nisms underlying the extreme sensitivity of Rev3L-deleting skin Rev3L /lox keratinocytes displayed twice as many DNA DSBs

E692 | www.pnas.org/cgi/doi/10.1073/pnas.1217425110 Lange et al. Downloaded by guest on October 2, 2021 AB PNAS PLUS

CD

E F

GH

− Fig. 6. Growth defects and DNA breaks in Rev3L-deleting keratinocytes. (A) Representative phase-contrast photographs of BK5.Cre;Rev3L /lox and BK5.Cre; Rev3L+/lox keratinocytes 5 d after isolation. (B) Quantification of cells with senescent morphology in BK5.Cre;Rev3L−/lox and BK5.Cre;Rev3L+/lox keratinocytes 5 d after isolation (n = 7). (C) Quantification of cells per square millimeter of BK5.Cre;Rev3L−/lox and BK5.Cre;Rev3L+/lox keratinocytes 5 d after isolation (n =7). (D) BrdU incorporation, normalized to total cells and compared with incorporation from a single population doubling of a reference cell population, of BK5. − − Cre;Rev3L /lox and BK5.Cre;Rev3L+/lox keratinocytes at 1, 3, and 5 d after isolation (n = 4). (E) Chromosome spreads from BK5.Cre;Rev3L /lox and BK5.Cre; Rev3L+/lox keratinocytes 3 d after isolation. Arrows indicate chromosome breaks or gaps (n = 6). (F) Quantification of gaps and breaks per metaphase in BK5. − − Cre;Rev3L /lox and BK5.Cre;Rev3L+/lox keratinocytes (n = 6). (G) 53BP1 and γ-H2AX immunofluorescence of keratinocytes derived from BK5.Cre;Rev3L /lox and BK5.Cre;Rev3L+/lox mice 3 d after isolation (n = 7). (H) Quantification of BK5.Cre;Rev3L−/lox and BK5.Cre;Rev3L+/lox keratinocytes containing both 53BP1 and γ-H2AX foci 3 d after isolation (n = 7). *P < 0.05; **P < 0.01. Data points are mean ± SEM.

[cells containing foci of both p53-binding protein 1 (53BP1) and deleting cells. These breaks then activate cell-stress responses phosphorylated histone H2A.x (γ-H2AX)] as controls (Fig. 6 G (such as p53 stabilization) that eventually lead to cell death. and H). To determine whether epidermal stem cells are rapidly exhaus- GENETICS − Evidence of breaks and genomic stress also was apparent in the ted in unchallenged BK5.Cre;Rev3L /lox mice, 10-d-old mouse skin of unchallenged experimental and control mice. There was pups were injected with BrdU, and label-retaining cells in the stem elevated p53 staining in the interfollicular epidermis of BK5.Cre; cell niche (bulge region) of the hair follicle were quantified after 30, − Rev3L /lox mice compared with controls and possibly elevated 60, and 90 d. No significant change in the number of BrdU label- γ-H2AX staining (P = 0.053) (Fig. 7 A and B, and Fig. S6). retaining cells of any of the genotypes was observed (Fig. 7C and Therefore, similar to Rev3L-deleting cells in culture (15, 16, 24– Fig. S7). At 30 and 60 d there was a small but significant decrease 27), DSBs and chromosomal gaps and breaks are present in in hair follicles that stained with keratin 15 (K15, marking the Rev3L-deleting epidermal cells in vivo. The evidence suggests stem cell niche) in the experimental mice compared with controls that proliferation leads to an increase in DNA breaks in Rev3L- (Fig. 7D and Fig. S7). However, the epidermal and follicular

Lange et al. PNAS | Published online February 5, 2013 | E693 Downloaded by guest on October 2, 2021 AB

CD

E

Fig. 7. Increased DNA breaks in Rev3L-deleting mouse skin. (A and B) Percentage of (A) p53-positive cells and (B) γ-H2AX–positive cells in the interfollicular epidermis of 7- to 10-wk-old BK5.Cre;Rev3L−/lox and BK5.Cre;Rev3L+/lox mice (n = 5). (C) Quantification of BrdU label-retaining follicular cells at 30, 60, or 90 d after BrdU injection in BK5.Cre;Rev3L−/lox, BK5.Cre;Rev3L+/lox, or BK5.Cre;Rev3L+/+ mice (n = 6). (D) Quantification of the percent of follicles that stain for K15 − at 30, 60, or 90 d after BrdU injection in BK5.Cre;Rev3L /lox, BK5.Cre;Rev3L+/lox, or BK5.Cre;Rev3L+/+ mice (n = 6). *P < 0.05, **P < 0.01. Data points are mean ± SEM. (E) Model for thresholds of tolerance in Rev3L-deficient cells and mice. Two factors contribute to cytotoxicity in Rev3L-deleting cells: increasing DNA damage (x-axis) and increasing proliferation (y-axis). Together, these factors cause DNA breaks (dotted blue diagonal line) in cells lacking Rev3L. Any DNA damage or proliferation above a threshold level (dotted black horizontal and vertical lines) initiates cell-death programs (likely signaled by p53). When DNA damage and proliferation are below these threshold levels (i.e., in the lower left-hand quadrant of graph), the cells can survive. However, they still experience DNA breaks which cause genomic instability and can lead to cancer formation. The levels of proliferation and DNA damage caused by different stimuli are noted by arrows on the x- and y-axes.

− defects in BK5.Cre;Rev3L /lox mice do not measurably exhaust the These results provide an explanation for the extreme UV ra- skin stem cell compartment after 3 mo. This observation may indi- diation sensitivity. In the absence of polζ, not only are mecha- cate that Rev3L-defective stem cells do not respond well to stimuli nisms absent for bypass of UV-induced DNA lesions, but also for renewal. A similar defect in epidermal stem cell functionality the proliferative response of Rev3L-defective cells is impaired. − − is seen with third-generation telomerase-deficient (Terc / )mice (28, 29). Such mice have epidermal defects but no increased Discussion turnover of label-retaining stem cells in unchallenged skin. When Polζ Deletion Leads to Epithelial Abnormalities and Spontaneous − − Terc / mice are treated with TPA, telomerase-deficient skin has Tumor Formation in Aging Rev3L-Deleted Tissues. This study answers a higher number of label-retaining cells compared with controls several long-standing questions about the function of polζ in because these cells cannot proliferate in response to the stimulus. mammalian cells. First, it has seemed puzzling that hematopoietic

E694 | www.pnas.org/cgi/doi/10.1073/pnas.1217425110 Lange et al. Downloaded by guest on October 2, 2021 cells are largely inviable when polζ function is ablated, but epithelial tion-induced DNA lesions caused by repeated irradiation present PNAS PLUS cells can survive in tissues deleting Rev3L in a mosaic fashion (12). a continual challenge during DNA replication for Rev3L-deleted We show here that although epithelial tissues are viable when Rev3L cells, even with mutant p53. Further, a Rev3L defect is expected to is deleted efficiently, the epidermis is atypical, with reduced inter- reduce the frequency of UV radiation-induced point mutations (4, follicular cellularity and abnormalities in hair follicles. Unchallenged 6, 37). Base change mutagenesis of genes such as p53 drives tumor Rev3L-defective epidermis accumulates genomic damage leading to formation in mouse models of UV carcinogenesis (23). Rapid chromosome abnormalities. As a consequence, mice deleting Rev3L formation of UV radiation-induced p53 mutant patches is corre- in keratin 5-expressing tissues develop spontaneous SCCs. These lated with tumor incidence in Xpa-, Xpc-, and Ercc6 (Csb)-de- tumors arise predominantly in specialized sebaceous glands, all of ficient mice (23). In Rev3L-deleting skin, some p53 point mutant which harbor keratin 5-expressing epithelia. The incremental accu- clones indeed are detected after low-dose chronic irradiation, and mulation of chromosomal alterations over time eventually leads to these may have a proliferative advantage. Formation of skin car- changes necessary for tumor formation. As a tumor suppressor (12), cinomas did not take place, however. Instead, chronic irradiation Polζ falls into the category of gene products that limit spontaneous of polζ-defective mice with even a relatively low dose of UV ra- tumor incidence by controlling the level of DNA breaks and chro- diation leads to persistent skin ulceration. As a consequence, mosome rearrangements (30). tumor clones may be unable to proliferate in polζ-defective skin Frequent tumors in specialized sebaceous glands also are seen because of heavily damaged stromal elements and inadequate Rad51C Tp53 in mice predisposed for simultaneous loss of and . recruitment of stem cells to repopulate the epidermis after ra- ’ These mice acquire Zymbal s and preputial gland tumors after diation. We regularly observed slow or absent healing of ad- – 400 440 d (31). In contrast, 35% of NER-defective xeroderma ventitious skin wounds (nicks from shaving or wounds inflicted Xpa−/− pigmentosum group A ( ) mice (compared with 19% of by cage mates), consistent with a deficiency in tissue regeneration Xpa+/+ control mice) form spontaneous tumors after 2 y, but none after wounding. of these neoplasias arise in specialized sebaceous glands (32). Mice Melanocyte migration to the epidermis and donation of melanin δ e lacking the exonuclease activity of either DNA polymerase or to keratinocytes is typical of human skin but is not a normal re- have distinct mutator and cancer phenotypes, but neither group has sponse in UV-irradiated mice. However, it has been observed as − − an increased incidence of sebaceous gland tumors (33). The astressresponseinAtm / mice exposed to (38) − − Rad51C-p53 model (31) has an increased load of DNA breaks be- and in Terc / mice with transgenic expression of TERF2,which cause of a defect in DSB repair by . The Rev3L fi have aberrant telomeres (39). In these instances damaged melano- -de cient mice may phenocopy this defect, increasing the load cyte stem cells are believed to differentiate terminally to melano- of DSBs arising because of failure to resolve blocked forks and gaps cytes and undergo transit to the epidermis. Notably, in the keratin in the absence of polζ.REV3Lalsomighthaveamoredirectrolein 5-Cre conditional Rev3L model studied here, polζ deletion takes repair of some DSBs by homologous recombination (34, 35). place in the keratinocytes, not in the melanocytes. The melanocyte ζ Polζ Defect Confers Extreme UV Sensitivity. Inactivation of polζ migration in irradiated pol -deleting mouse epidermis therefore causes the epidermis to be exquisitely sensitive to UV radiation. As is triggered indirectly as a consequence of keratinocyte damage. discussed further below, this sensitivity is a consequence of the dual Thresholds of Tolerance to DNA Damage and Proliferation in Rev3L- role of polζ in bypassing lesions caused by UV radiation and in Deficient Cells. We propose that the phenotypes of Rev3L-deficient sustaining normal cell proliferation in UV radiation-damaged fi ζ mouse cells in vivo and in vitro can be uni ed by considering two epidermis. It is striking that pol -deleting skin appears to be the ζ most UV radiation sensitive of any known mouse mutant, including roles for pol : a function in maintaining genomic integrity in nor- prototypical models for UV radiation sensitivity. Table S1 sum- mally proliferating cells and a function in DNA damage tolerance following UV irradiation. These ideas are illustrated schematically marizes results of studies for mice defective in NER genes and E DNA polη. Haired NER-defective mice (shaved before irradia- in Fig. 7 . − − tion) tolerate 6,000 J·m 2·wk 1, whereas albino hairless NER-de- In all cells, endogenous DNA lesions and replication barriers − − ζ fective mice tolerate 400 J·m 2·wk 1or more. In the present study, arise continuously which require pol function. For example, these − − fi haired Rev3L-defective mice could tolerate only 150 J·m 2·wk 1. barriers may include dif cult-to-replicate DNA sequences (7) and The epithelial dysfunction and radiation sensitivity documented ribonucleotides occasionally incorporated during DNA synthesis ζ here depend strictly on Rev3L disruption, because all experiments (8). In the absence of pol , double-strand breaks and chromosome included isogenic mice retaining one allele of Rev3L and expressing abnormalities accumulate. Rapidly proliferating cells such as those BK5.Cre. It is unlikely that Cre recombinase contributes signifi- in the hematopoietic system cannot survive this genomic stress. In cantly to the increased sensitivity, because hairless mice constitu- cell culture, there is a persistent stimulus to replicate DNA and ζ tively deleting Ercc1 with the same BK5.Cre (18) show UV divide, and so pol is required for keratinocyte proliferation (this fi radiation sensitivity similar to that of other NER-knockout mice study) and proliferation of primary cultures of embryonic bro- (Table S1). Mice expressing K5.Cre for epithelial deletion of the blasts (15). Attenuation of damage-sensing checkpoints, for ex- − − Stat3 gene tolerate 3,600–14,000 J·m 2·wk 1 UV radiation and ample by immortalization with T antigen, allows proliferation respond with relatively normal epidermal hyperproliferation (36). of Rev3L-defective cells in culture at the expense of an increased After irradiation for 7–30 wk, all mice deficient in XPA, XPC, or load of chromosomal aberrations (15). the cyclobutane (CPD) bypass polymerase polη In unchallenged epithelia, longer cell-cycling times allow more develop SCCs (Table S1). Irradiation of hairless mice deleting opportunities for alternative pathways of DNA damage tolerance − − GENETICS Ercc1 with BK5.Cre (375 J·m 2·wk 1) led to tumors in all animals or repair, and the tissues can survive, although genomic damage by 20 wk (18). Similarly, mice defective for the genes responsible accumulates slowly with time and predisposes Rev3L-deleting tis- for Cockayne syndrome group A or B (Ercc8 and Ercc6,re- sues to tumor formation. However, if proliferation levels are too spectively) develop skin carcinomas by 40 wk. In marked contrast high (as with stimulation by placing cells into culture or when rapid to NER-defective models, deletion of Rev3L in the skin did not regeneration is signaled following UV radiation), an excessive increase the incidence of UV radiation-induced carcinomas. number of DNA breaks occur, and cell-death pathways are en- Several factors likely contribute to the absence of UV radiation- gaged. Acute UV radiation increases both the DNA damage load induced tumors in Rev3L-deleting skin. Irradiated Rev3L- and proliferative pressure, resulting in rapid loss of cell viability defective skin has increased cell death caused by chromosomal with catastrophic tissue damage that does not support formation damage and a reduced ability to regenerate normally. UV radia- of UV radiation-induced tumors.

Lange et al. PNAS | Published online February 5, 2013 | E695 Downloaded by guest on October 2, 2021 These observations may have implications for any therapies tissue fluorescence, treatment of mice with TPA, keratinocyte culture, designed to inactivate polζ to sensitize tumors to DNA-damaging chromosome analysis, immunofluorescence, BrdU incorporation assay, label agents (40, 41). Strategies that target polζ for short periods of time, retaining cell assay, treatment of mice with UV radiation, p53 mutant clone in conjunction with treatments that induce DNA damage and analysis, and statistical analysis. stimulate proliferation, likely would be highly effective. However, ζ ACKNOWLEDGMENTS. We thank David Trono, Maria Sandoval, and Rebecca although temporary inhibition of pol may be tolerated, caution must Gray for expert assistance, Lidza Kalifa for initiating the TPA experiments, be used with longer-term suppression, because normal tissues would and Fernando Benavides and the Genetics Services core for congenics be expected to experience significant cellular toxicity, genome in- analysis and advice. We thank the Research Animal Support Facility, the stability by chromosome breakage, and additional tumor formation. Core Facility, and the Integrated Imaging, Histology and Pathology Facility at Smithville for expert support. We appreciate discussion Materials and Methods with our laboratory colleagues and Patrick Moore for comments on the manuscript. This research was supported by National Institutes of Health Most experiments were completed with mice on a pure FVB/N strain (NIH) Grant CA132840 from the National Cancer Institute, by Grant background. Mice used for the aging and keratinocyte experiments were P30ES007784 from the National Institute of Environmental Health Sciences, genetically ∼75% FVB/N mixed with C57BL6/J and 129/OLA strains. De- by NIH Cancer Center Support Grant P30-CA016672 (University of Texas tailed experimental procedures are described in SI Materials and Methods, M. D. Anderson Cancer Center), and by the Grady F. Saunders, PhD Distinguished encompassing mouse breeding, animal monitoring, necropsy and histology, Professorship (to R.D.W.).

1. McCulloch SD, Kunkel TA (2008) The fidelity of DNA synthesis by eukaryotic 23. Rebel H, et al. (2005) Relationship between UV-induced mutant p53 patches and skin replicative and translesion synthesis polymerases. Cell Res 18(1):148–161. tumours, analysed by mutation spectra and by induction kinetics in various DNA- 2. Schmitt MW, Matsumoto Y, Loeb LA (2009) High fidelity and lesion bypass capability repair-deficient mice. Carcinogenesis 26(12):2123–2130. of human DNA polymerase delta. Biochimie 91(9):1163–1172. 24. Van Sloun PP, et al. (2002) Involvement of mouse Rev3 in tolerance of endogenous 3. Lange SS, Takata K, Wood RD (2011) DNA polymerases and cancer. Nat Rev Cancer and exogenous DNA damage. Mol Cell Biol 22(7):2159–2169. 11(2):96–110. 25. Hicks JK, et al. (2010) Differential roles for DNA polymerases eta, zeta, and REV1 in 4. Shachar S, et al. (2009) Two-polymerase mechanisms dictate error-free and error- lesion bypass of intrastrand versus interstrand DNA cross-links. Mol Cell Biol 30(5): – prone translesion DNA synthesis in mammals. EMBO J 28(4):383 393. 1217–1230. 5. Yoon JH, Prakash L, Prakash S (2010) Error-free replicative bypass of (6-4) photoproducts 26. Zander L, Bemark M (2004) Immortalized mouse cell lines that lack a functional Rev3 by DNA polymerase zeta in mouse and human cells. Genes Dev 24(2):123–128. gene are hypersensitive to UV irradiation and cisplatin treatment. DNA Repair (Amst) 6. Gan GN, Wittschieben JP, Wittschieben BØ, Wood RD (2008) DNA polymerase zeta 3(7):743–752. (pol zeta) in higher . Cell Res 18(1):174–183. 27. Sonoda E, et al. (2003) Multiple roles of Rev3, the catalytic subunit of polzeta in 7. Northam MR, Robinson HA, Kochenova OV, Shcherbakova PV (2010) Participation of maintaining genome stability in vertebrates. EMBO J 22(12):3188–3197. DNA polymerase zeta in replication of undamaged DNA in Saccharomyces cerevisiae. 28. Flores I, Blasco MA (2009) A p53-dependent response limits epidermal stem cell Genetics 184(1):27–42. functionality and organismal size in mice with short telomeres. PLoS ONE 4(3):e4934. 8. Lazzaro F, et al. (2012) RNase H and postreplication repair protect cells from 29. Flores I, Cayuela ML, Blasco MA (2005) Effects of telomerase and telomere length on ribonucleotides incorporated in DNA. Mol Cell 45(1):99–110. – 9. Esposito G, et al. (2000) Disruption of the Rev3l-encoded catalytic subunit of epidermal stem cell behavior. Science 309(5738):1253 1256. — polymerase zeta in mice results in early embryonic lethality. Curr Biol 10(19): 30. Negrini S, Gorgoulis VG, Halazonetis TD (2010) Genomic instability an evolving – 1221–1224. hallmark of cancer. Nat Rev Mol Cell Biol 11(3):220 228. 10. Bemark M, Khamlichi AA, Davies SL, Neuberger MS (2000) Disruption of mouse 31. Kuznetsov SG, Haines DC, Martin BK, Sharan SK (2009) Loss of Rad51c leads to polymerase zeta (Rev3) leads to embryonic lethality and impairs blastocyst embryonic lethality and modulation of Trp53-dependent tumorigenesis in mice. development in vitro. Curr Biol 10(19):1213–1216. Cancer Res 69(3):863–872. 11. Wittschieben J, et al. (2000) Disruption of the developmentally regulated Rev3l gene 32. Nakane H, et al. (2008) Impaired spermatogenesis and elevated spontaneous causes embryonic lethality. Curr Biol 10(19):1217–1220. tumorigenesis in xeroderma pigmentosum group A gene (Xpa)-deficient mice. DNA 12. Wittschieben JP, et al. (2010) Loss of DNA polymerase zeta enhances spontaneous Repair (Amst) 7(12):1938–1950. tumorigenesis. Cancer Res 70(7):2770–2778. 33. Albertson TM, et al. (2009) DNA polymerase epsilon and delta proofreading suppress 13. Schenten D, et al. (2009) Pol zeta ablation in B cells impairs the germinal center discrete mutator and cancer phenotypes in mice. Proc Natl Acad Sci USA 106(40): reaction, class switch recombination, DNA break repair, and genome stability. J Exp 17101–17104. Med 206(2):477–490. 34. Sharma S, et al. (2012) REV1 and polymerase ζ facilitate homologous recombination 14. Daly J, et al. (2012) Altered Ig hypermutation pattern and frequency in complementary repair. Nucleic Acids Res 40(2):682–691. ζ – mouse models of DNA polymerase activity. J Immunol 188(11):5528 5537. 35. Kane DP, Shusterman M, Rong Y, McVey M (2012) Competition between replicative 15. Lange SS, Wittschieben JP, Wood RD (2012) DNA polymerase zeta is required for and translesion polymerases during homologous recombination repair in Drosophila. proliferation of normal mammalian cells. Nucleic Acids Res 40(10):4473–4482. PLoS Genet 8(4):e1002659. 16. Wittschieben JP, Reshmi SC, Gollin SM, Wood RD (2006) Loss of DNA polymerase zeta 36. Kim DJ, Angel JM, Sano S, DiGiovanni J (2009) Constitutive activation and targeted causes chromosomal instability in mammalian cells. Cancer Res 66(1):134–142. disruption of signal transducer and activator of transcription 3 (Stat3) in mouse 17. Ramirez A, et al. (2004) A keratin K5Cre transgenic line appropriate for tissue-specific epidermis reveal its critical role in UVB-induced skin carcinogenesis. Oncogene 28(7): or generalized Cre-mediated recombination. Genesis 39(1):52–57. 950–960. 18. Doig J, et al. (2006) Mice with skin-specific DNA repair gene (Ercc1) inactivation are 37. Li Z, et al. (2002) hREV3 is essential for error-prone translesion synthesis past UV or hypersensitive to irradiation-induced skin cancer and show more rapid fi actinic progression. Oncogene 25(47):6229–6238. benzo[a]pyrene diol epoxide-induced DNA lesions in human broblasts. Mutat Res – 19. Muzumdar MD, Tasic B, Miyamichi K, Li L, Luo L (2007) A global double-fluorescent 510(1-2):71 80. Cre reporter mouse. Genesis 45(9):593–605. 38. Inomata K, et al. (2009) Genotoxic stress abrogates renewal of melanocyte stem cells 20. Abel EL, Angel JM, Kiguchi K, DiGiovanni J (2009) Multi-stage chemical carcinogenesis by triggering their differentiation. Cell 137(6):1088–1099. in mouse skin: Fundamentals and applications. Nat Protoc 4(9):1350–1362. 39. Rudolph KL, et al. (1999) Longevity, stress response, and cancer in aging telomerase- 21. Lu YP, et al. (1999) Time course for early adaptive responses to ultraviolet B light in deficient mice. Cell 96(5):701–712. the epidermis of SKH-1 mice. Cancer Res 59(18):4591–4602. 40. Xie K, Doles J, Hemann MT, Walker GC (2010) Error-prone translesion synthesis 22. van Oosten M, et al. (2000) Differential role of transcription-coupled repair in UVB- mediates acquired chemoresistance. Proc Natl Acad Sci USA 107(48):20792–20797. induced G2 arrest and apoptosis in mouse epidermis. Proc Natl Acad Sci USA 97(21): 41. Doles J, et al. (2010) Suppression of Rev3, the catalytic subunit of Polζ, sensitizes drug- 11268–11273. resistant lung tumors to chemotherapy. Proc Natl Acad Sci USA 107(48):20786–20791.

E696 | www.pnas.org/cgi/doi/10.1073/pnas.1217425110 Lange et al. Downloaded by guest on October 2, 2021