In vivo measurements of interindividual differences in PNAS PLUS DNA glycosylases and APE1 activities

Isaac A. Chaima,b, Zachary D. Nagela,b, Jennifer J. Jordana,b, Patrizia Mazzucatoa,b, Le P. Ngoa,b, and Leona D. Samsona,b,c,d,1

aDepartment of Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139; bCenter for Environmental Health Sciences, Massachusetts Institute of Technology, Cambridge, MA 02139; cDepartment of Biology, Massachusetts Institute of Technology, Cambridge, MA 02139; and dThe David H. Koch Institute for Integrative Cancer Research, Massachusetts Institute of Technology, Cambridge, MA 02139

Edited by Paul Modrich, Howard Hughes Medical Institute and Duke University Medical Center, Durham, NC, and approved October 20, 2017 (received for review July 6, 2017) The integrity of our DNA is challenged with at least 100,000 lesions with increased cancer risk and other diseases (8–10). However, per cell on a daily basis. Failure to repair DNA damage efficiently the lack of high-throughput assays that can reliably measure in- can lead to cancer, immunodeficiency, and neurodegenerative dis- terindividual differences in BER capacity have limited epidemio- ease. (BER) recognizes and repairs minimally logical studies linking BER capacity to disease. Moreover, BER helix-distorting DNA base lesions induced by both endogenous repairs DNA lesions induced by radiation and chemotherapy (3, 11), and exogenous DNA damaging agents. Levels of BER-initiating raising the possibility of personalized treatment strategies using DNA glycosylases can vary between individuals, suggesting that BER capacity in tumor tissue to predict which therapies are most quantitating and understanding interindividual differences in DNA likely to be effective, and using BER capacity in healthy tissue to repair capacity (DRC) may enable us to predict and prevent disease determine the dose individual patients can tolerate. in a personalized manner. However, population studies of BER Although methods have been developed to measure BER capacity have been limited because most methods used to mea- activity for every step of the pathway (12), each has drawbacks sure BER activity are cumbersome, time consuming and, for the that limit the potential for use in epidemiological studies and most part, only allow for the analysis of one DNA glycosylase at clinical translation. In vitro assays using cell lysates may not re-

a time. We have developed a fluorescence-based multiplex flow- capitulate physiological DNA repair conditions. With few ex- CELL BIOLOGY cytometric host cell reactivation assay wherein the activity of sev- ceptions (13, 14), the lack of multiplex assays for multiple DNA eral enzymes [four BER-initiating DNA glycosylases and the down- glycosylases and downstream BER steps has resulted in studies stream processing apurinic/apyrimidinic endonuclease 1 (APE1)] can focusing on a single-repair step or a single-repair pathway. A be tested simultaneously, at single-cell resolution, in vivo. Taking recently described fluorescence-based multiplex flow-cytometric advantage of the transcriptional properties of several DNA lesions, host cell reactivation (FM-HCR) overcomes these challenges, we have engineered specific fluorescent reporter plasmids for allowing for in vivo multiplexed DNA repair capacity (DRC) quantitative measurements of 8-oxoguanine DNA glycosylase, alkyl- measurements for multiple DNA repair pathways (15). Here we adenine DNA glycosylase, MutY DNA glycosylase, uracil DNA gly- report on the development of FM-HCR reporters for apurinic/ cosylase, and APE1 activity. We have used these reporters to measure apyrimidinic endonuclease 1 (APE1), and for four DNA gly- differences in BER capacity across a panel of cell lines collected from cosylase activities. We apply these reporters to measure inter- healthy individuals, and to generate mathematical models that pre- individual variations in BER in a panel of 24 cell lines derived dict cellular sensitivity to methylmethane sulfonate, H2O2,and5-FU from healthy individuals, and build mathematical models that from DRC. Moreover, we demonstrate the suitability of these re- predict cellular sensitivity to several agents that induce DNA porters to measure differences in DRC in multiple pathways using damage processed by the BER pathway. The models reveal the primary lymphocytes from two individuals. Significance DNA repair | transcriptional mutagenesis | DNA glycosylase | apurinic/apyrimidinic endonuclease 1 | base excision repair The DNA in each cell is damaged thousands of times daily. Con- sequently, a battery of DNA repair pathways exist that allow ur genomes are constantly challenged with damage that repair of this damage. Failure to repair can lead to devastating Oarises endogenously from products of cellular metabolic pro- diseases, including cancer and neurodegeneration. Each individ- cesses, as well as exogenously from environmental DNA damaging ual’s DNA repair capacity (DRC) is inherently different. Being able agents (1, 2). Failure to efficiently repair DNA damage can lead to measure an individual’s DRC could contribute to a personal- to cancer, degenerative disease, and premature aging (3–5). Thus, ized approach to prevent and treat disease. Here we present a variety of mechanisms to repair DNA damage have evolved that powerful tools for measuring in vivo base excision repair capacity serve to delay the onset of such diseases. for five distinct DNA lesions. We use these methods to predict the The base excision repair (BER) pathway recognizes and re- cellular responses to a variety of DNA damaging agents, and to pairs small, nonbulky DNA lesions, including alkylated, deaminated, monitor differences in DRC in primary human lymphocytes. Ad- and oxidized bases (5). BER involves five steps: (i) recognition and ditionally, we unveil previously unknown transcriptional muta- excision of a damaged base by a DNA glycosylase, forming an genesis induced by DNA lesions. abasic site; (ii) incision of the sugar-phosphate DNA backbone Author contributions: I.A.C. and L.D.S. designed research; I.A.C., Z.D.N., J.J.J., P.M., and at the abasic site; (iii) processing and removal of remaining sugar L.P.N. performed research; I.A.C. contributed new reagents/analytic tools; I.A.C. and J.J.J. moieties; (iv) gap-filling by a DNA polymerase; and (v) sealing of analyzed data; and I.A.C., Z.D.N., and L.D.S. wrote the paper. the remaining nick by a DNA ligase (2). The authors declare no conflict of interest. BER deficiencies have been linked with several diseases. This article is a PNAS Direct Submission. MutY DNA glycosylase (MUTYH) deficiency confers colorectal Published under the PNAS license. cancer predisposition (6), and uracil DNA glycosylase (UNG) 1To whom correspondence should be addressed. Email: [email protected]. deficiency leads to hyper IgM-syndrome (7). Furthermore, vari- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. ations in the activities of several BER enzymes are associated 1073/pnas.1712032114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1712032114 PNAS Early Edition | 1of10 Downloaded by guest on September 30, 2021 relative contribution of multiple repair pathways to cellular We first generated plasmids bearing a nonsynonymous site-specific sensitivity, and highlight the potential for FM-HCR assays to mutation that inactivates the chromophore for each of four fluo- guide clinicians in selecting therapy for individual patients. rescent proteins, namely mOrange, BFP, GFP, and mPlum. We Finally, we demonstrate the potential of our DRC reporters to further modified these plasmids by positioning one of four differ- inform epidemiological studies by measuring DRC in primary ent site-specific DNA lesions at each mutated site, each lesion human T lymphocytes. being a substrate for repair by one of the four DNA glycosylases. The mOrange reporter contained 8-oxoguanine (8oxoG) to Results measure OGG1 activity; the BFP reporter contained a uracil Validation of FM-HCR Assays for Measuring DNA Glycosylase Activity. (U) to measure UNG activity; the GFP reporter contained Building upon our previously described FM-HCR substrate that hypoxanthine (Hx) to measure AAG activity; and the mPlum reports 8-oxoguanine DNA glycosylase (OGG1) activity (15), we reporter contained adenine opposite 8oxoG to measure MUTYH took advantage of DNA-lesion–induced transcriptional mutagenesis activity. Following transfection, transcriptional mutagenesis induced to design FM-HCR substrates that report activity for UNG, alkyl- by the site-specific lesion leads to expression of fluorescent protein; adenine DNA glycosylase (AAG, also known as MPG), and MUTYH. detailed examples and the rationale for this approach are presented

AB C

Fig. 1. Rationale for transcriptional mutagenesis-based fluorescent reporters. The base pairs shown correspond to sites that code for a key of the chromophores of the fluorescent proteins. The transcribed strand is always shown on top. The base shown in mRNA corresponds to the ribonucleotide incorporated by RNAPII across from the specific base in the plasmid. OG, 8-oxoguanine. (A) WT reporter codes for a fluorescent protein, incorporation of the base shown at that specific site is necessary for fluorescence to occur. (B) Nonfluorescent reporter variants. (C) Base misincorporation opposite DNA lesions. Only transcriptional mutagenesis events (or incorporation of U opposite A for the A:OG ), caused by the presence of the lesion (or incorrect base for the A:OG base pair) will result in the expression of a fluorescent variant of the reporter. Incorporation of any other ribonucleotides results in nonfluorescent proteins. (1) To generate substrates, WT reporters were mutated through QuikChange site-directed mutagenesis (Materials and Methods) to produce nonfluorescent variants. (2) Through primer extension (Materials and Methods), base lesion is incorporated on the transcribed strand at the mutated site of the nonfluorescent variant (with the exception of the A:OG base pair for which OG is incorporated on the nontranscribed strand of the WT variant). (3) Repair by the cellular DNA repair machinery leads to removal of the lesion (or incorrect base for the A:OG base pair) and the source of transcriptional mutagenesis events, so that only nonfluorescent variants are expressed. (4) For the A:OG base pair in particular, a new round of repair is needed to remove the OG lesion.

2of10 | www.pnas.org/cgi/doi/10.1073/pnas.1712032114 Chaim et al. Downloaded by guest on September 30, 2021 A B repair reporter (GFP-C289T-Hx) that results in WT-GFP only PNAS PLUS upon incorporation of C opposite Hx. Upon repair, Hx is re- placed by A, leading to production of a nonfluorescent protein. The ability of these two plasmids to report on U and Hx repair was tested in MEFs deficient in the main UNG (plasmid com- binations #1 and #2 in Table 1), and in MEFs deficient in the only known Hx DNA glycosylase, AAG (plasmid combinations #3 and #4 in Table 1). As expected for both reporters, WT MEFs exhibited relatively low fluorescent reporter expression, − − − − whereas Ung / and Aag / MEFs exhibited ∼8- and 12-fold higher fluorescence, respectively (Fig. 2 B and C). MUTYH. During replication, DNA polymerases can incorporate A opposite an unrepaired 8oxoG 10–70% of the time, depending on the polymerase, leading to an A:8oxoG base pair (16). In this scenario, if the DNA repair machinery were to remove the 8oxoG lesion, a GC-to-TA transversion mutation would be fixed into DNA as a T would be incorporated opposite A during BER. Importantly, 8oxoG paired with A is an extremely poor substrate for OGG1. Instead, MUTYH removes the undamaged A, and C repair synthesis restores the C:8oxoG base pair that can then be D processed by OGG1. That OGG1 has a vast preference for ex- cising 8oxoG opposite C and rarely excises 8oxoG opposite A reduces the risk of mutation. For the MUTYH reporter, removal of A in the transcribed strand, positioned opposite an 8oxoG, is measured in the same way as for the previous three DNA glyco- sylase reporters described above. The MUTYH reporter plasmid, mPlum-T202WT-8oxoG (+), expresses WT-mPlum only upon CELL BIOLOGY incorporation of U opposite A; repair-dependent replacement of A with C leads to expression of nonfluorescent protein. Thus, as expected, MUTYH-deficient MEFs express higher levels of mPlum (sevenfold) than WT MEFs following reporter transfection (Fig. 2D and plasmid combinations #7 and #8 in Table 1). We further validated the MUTYH reporter using a panel of cell lines with varying levels of MUTYH activity generated as described in Materials and Methods. We measured activity using both FM-HCR (plasmid combinations #13 and #14 in Table 1) and the gold- standard radiolabeled oligonucleotide assay (SI Appendix, Fig. Fig. 2. FM-HCR assays for measuring DNA glycosylase activity on alkylated, S1A). MUTYH activity for the in vitro and the in vivo assays were deaminated, oxidized, and misincorporated bases. Proof-of-concept BER strongly correlated R2 = 0.90 (SI Appendix, Fig. S1B), confirming capacity measurements comparing WT MEFs and MEFs deficient in DNA the ability of FM-HCR assays to make accurate quantitative glycosylases known to repair the lesion of choice. (A) 8oxoG:C repair re- − − − − measurements of intermediate levels of BER activity, as has been porter in Ogg1 / cells. (B) Uracil repair reporter in Ung / cells. (C) Hx repair 6 − − − − demonstrated for other pathways, namely the O -methylguanine reporter in Aag / cells. (D) A:8oxoG repair reporter in Mutyh / cells. Error bars represent the SD calculated from at least three biological replicates. DNA methyltransferase (MGMT) (15) and the DNA mismatch repair (MMR) pathway (17).

in Fig. 1. Validation of the ability of each construct to report on a Simultaneous Measurement of Four DNA Glycosylase Activities. A specific DNA glycosylase activity is presented below. major advantage of FM-HCR assays is the ability to measure OGG1. We reproduced our previous results measuring 8oxoG multiple repair activities simultaneously; however, multiplexing repair in Ogg1-proficient and -deficient mouse embryonic fibro- requires that any given reporter plasmid does not alter the ability blasts (MEFs) (15). This plasmid (mOrange-A215C-8oxoG) re- of cells to repair damage in the other reporter plasmids and that ports on transcriptional misincorporation of A opposite the 8oxoG the fluorescent signals produced by each of the reporters do not − − lesion. As expected, Ogg1 / cells show approximately a 17-fold alter the ability to detect signals from the other reporters. To test higher level of mOrange than WT MEFs, indicating a much whether interference occurs between the reporter plasmids upon lower 8oxoG repair capacity in Ogg1-deficient vs. WT cells, cotransfection, we transfected a single BER reporter plasmid at which efficiently remove the source of the transcriptional errors a time (Fig. 3A and plasmid combinations #1 through #8 in (Fig. 2A and plasmid combinations #5 and #6 in Table 1). Since Table 1) or cotransfected four BER reporter plasmids simulta- this plasmid also reports on the activity of other DNA glycosylases neously (Fig. 3B and combinations #9 and #10 in Table 1). that can remove 8oxoG (i.e., NEIL1 and NEIL2), we infer that Nearly identical results were observed under both conditions, these enzymes play a minor role in 8oxoG repair in MEFs. That validating the ability of the FM-HCR assay to simultaneously the plasmid reports on 8oxoG repair by multiple DNA glycosylases assess the activity of at least four DNA glycosylases, thus de- is viewed as an advantage for ultimately measuring total DNA re- creasing the number of transfections needed to measure the four pair capacity for this lesion in epidemiological and clinical samples. activities from eight to two. It should be noted that only live cells UNG and AAG. We developed a U repair reporter (BFP-A191G-U) are considered for these analyses; this is achieved by incorpo- that produces WT-BFP transcripts only when RNA polymerase rating the dead-stain TO-PRO-3 (Materials and Methods)asa II (RNAPII) incorporates an A opposite U present in the tran- sixth simultaneously measured fluorescent signal (besides the scribed strand. Upon repair, U is replaced by C, leading to pro- four DNA glycosylase reporters and the transfection efficiency duction of a nonfluorescent protein. We also developed a Hx control plasmid).

Chaim et al. PNAS Early Edition | 3of10 Downloaded by guest on September 30, 2021 Table 1. Combinations of reporter plasmids and types of DNA damage used in each experiment Comb. AmCyan ng tagBFP ng EGFP ng mOrange ng mPlum ng Empty vector, ng

#1 WT 500 WT 30 1,970 #2 WT 500 A191G-U 90 1,910 #3 WT 500 WT 25 1,975 #4 WT 500 C289T-Hx 25 1,975 #5 WT 500 WT 30 1,970 #6 WT 500 A215C-8oxoG 120 1,880 #7 WT 500 WT 90 1,910 #8 WT 500 T202WT-8oxoG(+) 135 1,865 #9 WT 500 WT 30 WT 25 WT 30 WT 90 1,825 #10 WT 500 A191G-U 90 C289T-Hx 25 A215C-8oxoG 120 T202WT-8oxoG(+) 135 1,630 #11 WT 2,500 WT 100 WT 50 WT 75 WT 200 #12 WT 2,500 A191G-U 300 C289T-Hx 50 A215C-8oxoG 300 T202WT-8oxoG(+)300 #13 WT 30 WT 95 2,375 #14 WT 30 T202WT-8oxoG(+) 95 2,375 #15 WT 100 WT 50 #16 WT 100 617-THF 50 #17 WT 100 WT 50 #18 WT 100 A215C-THF 50

Headers correspond to the different plasmid reporters used. Nomenclature for plasmids can be found in Table 2. Comb., combination; ng, nanograms of plasmid used for each combination.

Development of FM-HCR Assays for Measuring APE1 Activity. DNA Substrates were developed using two complementary strategies, glycosylase activity determines the rate at which BER is initiated, namely measuring the ability of THF to induce transcriptional but the ability to measure downstream steps of this pathway can mutagenesis, or measuring the ability of THF to block transcrip- also contribute to understanding the BER status of a cell, tissue, tion (Fig. 4A). A transcriptional mutagenesis reporter analogous or individual. We engineered two different reporters for mea- to the mOrange-A215C-8oxoG reporter described above (mOrange- suring APE1 activity, the major AP-endonuclease that cleaves A215C-THF) was developed, where RNAPII-mediated incorpo- DNA at abasic sites generated by DNA glycosylases. We built ration of A opposite THF in the transcribed strand leads to expres- plasmids bearing a site-specific tetrahydrofuran (THF), an abasic sion of a transcript encoding WT mOrange, and upon THF repair site analog that is more chemically stable than a natural abasic G replaces the abasic site and no fluorescent protein is expressed. site and which is efficiently recognized and nicked by APE1 (18). A second reporter using a THF:C base pair (GFP-617-THF) was

ABSeparate Measurements Multiplexed Measurements

Fig. 3. Comparison of single and multiplexed measurements of four BER DNA glycosylase substrates. (A) FM-HCR analysis with four different BER reporter plasmids transfected separately into repair-proficient (WT) cells or MEFs deficient for one DNA glycosylase known to repair each one of the tested lesions at a time. (B) FM-HCR analysis with four different BER reporter plasmids cotransfected into the same cell lines. An AmCyan plasmid was cotransfected as a transfection efficiency control for all of the experiments. %R.E. in mutant MEFs was normalized to %R.E. in WT cells. Error bars represent the SD calculated from at least three biological replicates.

4of10 | www.pnas.org/cgi/doi/10.1073/pnas.1712032114 Chaim et al. Downloaded by guest on September 30, 2021 A PNAS PLUS

BC

Fig. 4. FM-HCR assays for measuring APE1 activity. (A, Upper) Steps 1a, 2a, and 3a, similar to Fig. 1 steps 1, 2, and 3. (A, Lower) WT reporter codes for a fluorescent protein, incorporation of the base shown site is necessary for fluorescence to occur. 1b. Base lesion is incorporated on the transcribed strand at the mutated site of the WT variant through primer extension (Materials and Methods). Upon transcription, the presence of the lesion results in nonfluorescent truncated transcripts. (2b) Fluorescence is restored only upon complete repair to G. These plasmids were used as proof-of-concept abasic site repair capacity measurements comparing Ape1-proficient (++Δ) and Ape1-deficient (ΔΔΔ) cells based on the different effects abasic sites can have on transcription. (B) Transcriptional misincorporation of A opposite THF in the mOrange-A215C-THF substrate. (C) Transcriptional blockage by THF in the GFP-617-THF substrate. Error bars represent the SD calculated from at least three biological replicates. ND, not detected.

generated to measure APE1 activity based on the transcription- Analysis of BER Capacity in a Panel of 24 Cell Lines Derived from

blocking properties of THF. To ensure that this substrate only Apparently Healthy Individuals. Given that small variations in the CELL BIOLOGY reports a repair event and does not report a bypass of the THF activities of some BER enzymes have been correlated with dis- lesion, GFP site C617 was mutated to the remaining three base ease risk (8–10), we sought to determine if our methods could combinations (C617A, C617G, and C617T), to test whether po- reliably measure small interindividual differences in BER capacity. tential transcriptional bypass events placing A, G, or U opposite We used the four DNA glycosylase transcriptional mutagenesis- THF would be able to generate fluorescent protein events; none of based reporters (for repair of Hx:T, 8oxoG:C, A:8xoG, and U:G), them did, and thus only repair to the WT sequence in the tran- as well as the transcription blockage reporter for APE1 to measure the BER capacity across a panel of 24 human B lymphoblastoid scribed strand (G), which leads to expression of WT transcripts (C cell lines derived from apparently healthy individuals of diverse at position 617), could produce fluorescent protein. ancestry (20) (plasmid combinations #11 and #12, and #15 and We validated the mOrange-A215C-THF transcriptional mu- #16 in Table 1). A range of BER activity was observed across the tagenesis reporter using isogenic mouse B cell lines that were ++Δ ΔΔΔ 24 cell lines (Fig. 5A and detailed data in SI Appendix, Table S2). APE1-proficient (Ape1 ) and APE1-deficient (Ape1 ) Moreover, each cell line displays a unique BER capacity fin- (19). Transfection of the mOrange-A215C-THF plasmid (plas- gerprint (Fig. 5B). mid combinations #17 and #18 in Table 1) into these cells resulted in no detectable fluorescent mOrange events for Ape1++Δ Regression Models Based on BER Capacity Can Predict Sensitivity to cells and ∼14% reporter expression for the deficient Ape1ΔΔΔ DNA Damaging Agents. We used mathematical modeling to test cells (Fig. 4B), indicating that WT cells efficiently remove the the potential for BER activity measurements to predict sensi- THF lesion before it can induce transcriptional mutagenesis de- tivity to three DNA damaging agents known to form DNA lesions tectable by FM-HCR. We also engineered a set of reporters to test processed by BER. Sensitivity to 0.4 mM methylmethane sulfonate μ μ whether transcriptional misincorporation of C, G, or U opposite (MMS; previously measured) (21), 25 MH2O2, and 100 M5- THF would be able to generate fluorescent protein events when Fluorouracil (5-FU) were measured for this cell panel. The 24 cell transfected into Ape1ΔΔΔ cells; they did not, suggesting that A is lines showed a broad and continuous range of sensitivities, spe- the major base misincorporated by RNAPII opposite THF in vivo. cifically, a 9.5-, 4.7-, and 2.7-fold range were observed for MMS, This result also reiterates that in the case of the GFP-617-THF H2O2,and5-FU,respectively(SI Appendix,Fig.S2;seedatainSI Appendix, Tables S4 and S5). While H2O2 and 5-FU sensitivity substrate, only removal of the transcription-blocking THF, and not = = misincorporation of C, can lead to GFP expression. were modestly correlated (R 0.47, P 0.02), MMS was not We validated the GFP-617-THF substrate, which reports significantly correlated to either H2O2 or 5-FU (SI Appendix,Table APE1 activity based on the ability of THF to block transcription, S6), indicating that sensitivity to one agent does not necessarily predict sensitivity to other agents (22). using the same APE1-proficient (Ape1++Δ) and APE1-deficient Multiple linear regression (MLR) seeks to find the best model Ape1ΔΔΔ ( ) cell lines. As expected, following transfection with describing the linear relationship between independent variables GFP-617-THF (Fig. 4C and plasmid combinations #15 and ++Δ (z-scored FM-HCR reporter data) (Fig. 5B) and a dependent #16 in Table 1), repair-proficient Ape1 cells show approx- variable (sensitivity to DNA damaging agents reported as per- imately a fivefold higher level of fluorescent reporter expression cent control growth). For this analysis, the BER plasmid reporter as a result of higher repair capacity compared with the repair- measurements reported here were combined with our previously deficient Ape1ΔΔΔ cells. For additional experiments, we chose published FM-HCR data for five other DNA repair pathways in to use the transcription-blocking GFP-617-THF substrate to report the same cell lines (15). Specifically, we included DRC measured APE1 activity because, unlike for the transcriptional mutagenesis for nucleotide excision repair (NER), MMR, direct reversal by reporter, this substrate yielded a nonzero readout for repair- MGMT, nonhomologous end-joining (NHEJ), and homologous proficient cells (Fig. 4). recombination (HR) (SI Appendix, Table S3).

Chaim et al. PNAS Early Edition | 5of10 Downloaded by guest on September 30, 2021 A1 A2 A3 B

A4 A5

Fig. 5. Interindividual variability in BER of a panel of 24 B lymphoblastoid cell lines. FM-HCR with BER substrates was assessed in all cell lines. (A1) Hx:T, (A2) U:G, (A3) A:8oxoG, (A4) 8oxoG:C, and (A5) THF:C. Cells are ranked from left to right based on increasing repair capacity. Error bars show SD of at least three biological replicates. (B) %R.E. was z-scored and organized in a heat map representation in ascending order of Hx excision activity (left-most column). Each row represents a cell line, and each column represents repair capacity for a single substrate indicated in the heading. Warm colors represent lower excision activity for all reporters but for THF. Cooler colors represent higher excision activity for all reporters but for THF. A color-coded scale at the right of the heat map provides the number of SDs above or below the mean corresponding to the colors in the heat map.

A variety of MLR models were built for each agent (SI Ap- are prioritized in the discussion of the implications of these pendix, Table S7). The models yielding the best correlation fit models below. 2 following leave-one-out cross-validation (LOOCV, R CV) and the lowest root mean square error (RMSE) were further con- Interindividual Variability in DRC of Primary Human Lymphocytes. sidered. Inclusion of all of the variables in the model results in Given that the ultimate goal of our methods is to contribute in the assessment, prevention, and treatment of human disease, we overfitting of the data and a low predictive power; as such, R2 CV sought to determine if they could be used in primary blood samples is prioritized over model R2 to minimize overfitting of the data. to reliably measure interindividual differences in DRC. We utilized The following models for predictions of percent control growth the same 10 DRC reporters that were used to measure DRC in the (%C.G.) were selected: lymphoblastoid cell lines (plasmid combinations in SI Appendix, Table S8). Following peripheral blood mononuclear cell (PBMC) = − × + × − × %C.G.MMS 13.2 8oxoG : C 8.7 U:G 8.3 THF isolation from the blood of two different donors, PHA stimulation + 8.0 × HR + 5.5 × NHEJ + 44.2; of T lymphocytes, and transfection with the multiplexed plasmids, [1] we were able to monitor significant differences in U, Hx, A:8oxoG, and NER repair. While donor “A” showed higher U, Hx, and A:8oxoG repair capacity, donor “B” showed higher NER capacity; %C.G. = −4.6 × 8oxoG : C + 3.8 × U:G − 3.1 × THF 5-FU no differences were observed for 8oxoG:C, THF:C, MMR, NHEJ, − 1. 9 × NHEJ + 1.8 × MGMT − 1.1 × A : 8oxoG and HR repair capacity (Fig. 6). MGMT activity could not be + 19.3; determined as the 30-fluorescent-events threshold was not met [2] (Materials and Methods), indicating that both individuals may have very high MGMT activity.

%C.G.H2O2 = −14.5 × 8oxoG : C + 14. 1 × U:G − 10. 5 × THF Discussion − 6.0 × HR + 4.8 × MGMT + 4.3 × NER + 52.2. The ability to simultaneously measure the in vivo repair activity of [3] multiple BER substrates extends our previously published FM- HCR substrate toolbox to include nearly all of the major DNA Special attention should be given to contributions by tran- repair pathways (15), namely BER (initiated by AAG, MUTYH, scriptional mutagenesis-based reporters (Hx:T, U:G, 8oxoG:C, OGG1, and UNG, and abasic site processing by APE1), plus HR, A:8oxoG, and MGMT) for which a negative sign indicates that MGMT, MMR, NER, and NHEJ. increased repair activity for that particular lesion correlates with Functional DRC assays, such as the FM-HCR presented here, hold the advantage that multiple regulatory levels—such as increased resistance. As shown in Eq. 4 (below) the “β” slopes epigenetic, transcriptional, translational, and posttranslational— (here, the numerical values that accompany each term) represent are integrated into the readout of the fluorescent reporter (12). the percentage point change in %C.G. that is associated with a 1 Moreover, assays performed in vivo report the activity of the repair “ ” SD increase in repair capacity in the x relevant pathway; for machinery in its physiological context. Therefore, the involvement example a 1 SD increase in 8oxoG:C repair activity is associated of redundant repair pathways is also integrated into the final repair with a 13.2% point increase in %C.G. for cells treated with score. For example, APE1 (23) and NEIL1 (24) have been shown MMS. MLR provides a confidence interval for each of the β to stimulate OGG1’s activity by increasing its turnover. Therefore, slopes; those that were significantly different from zero with at an in vitro assay that doesn’t recapitulate physiological conditions least 95% confidence have been underlined in the models and for OGG1, NEIL1, and APE1 could result in misrepresentation of

6of10 | www.pnas.org/cgi/doi/10.1073/pnas.1712032114 Chaim et al. Downloaded by guest on September 30, 2021 PNAS PLUS

Fig. 6. Interindividual variability in DRC of primary T lymphocytes. FM-HCR with 10 DRC substrates was assessed in PHA-stimulated PBMCs isolated from fresh whole blood from two different donors (empty bars correspond to donor “A” and filled bars to donor “B”). Error bars show SD of four replicates. Two-tailed unpaired t test *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001. No results are shown for MGMT as none of the samples met the 30-fluorescent-event threshold described in Materials and Methods.

OGG1’s in vivo activity. Moreover, our results agree with the notion of AP site-induced disruption of transcription is certainly not that OGG1 is the main 8oxoG repair enzyme while other DNA negligible, given that ∼20,000 abasic sites are formed on average glycosylases, such as NEIL1 and NEIL2, might also contribute to in a mammalian cell every day (41). Unrepaired abasic sites as the repair score observed (25, 26). Similarly, U can be repaired well as Hx, U, and 8oxoG may thus give rise to mutations during by UNG as shown here but also by SMUG1, MBD4, and TDG, as DNA replication, and may also result in truncated or mutated well as by the MMR machinery (27). Abasic sites are primarily mRNA transcripts that produce faulty proteins, having a direct recognized and incised by APE1, but APE2 can also act on the impact on cellular physiology (36). lesion (28). MUTYH mainly removes A opposite 8oxoG in the BER pathway, but MMR can play a role in repair at this base pair Interindividual Variation in BER Capacity. Most analyses of in- as well (29). Finally, the reporter for Hx repair has several unique terindividual variation in BER have been performed in vitro with characteristics; besides AAG, there are no other DNA glycosylases lysates of PBMCs isolated from the blood of healthy donors. This with Hx as a substrate, although the EndoV endonuclease may play complicates a direct comparison with the fold-range of activities a minor role in its removal (30, 31). Thus, in this particular case, observed using in vivo FM-HCR methods in immortalized human

Hx repair can be directly considered a proxy for AAG activity. B lymphoblastoid cell lines. Nevertheless, with the exception of CELL BIOLOGY Additionally, AAG’s wide lesion recognition and excision spectrum AAG, for which a 3.3- to 10-fold activity range has been reported (32–34) suggests that our results for Hx repair reliably represent in PBMCs (42–44) (compared with the 1.7-fold described here), the relative excision activity of the AAG DNA glycosylase, re- the degree of interindividual variation reported here agrees with gardless of its substrate. previously observed ranges for OGG1 in PBMCs [2- to 2.9-fold Based on the design principles of the constructs, we can make (45, 46), compared with 4.3-fold] and UNG in colon, stomach, and a clear distinction between step-specific and multistep BER re- liver [5.5-, 3.2-, and 3.1-fold (47), respectively, compared with 5.3- porters. Transcriptional mutagenesis-based constructs result in fold]. To our knowledge, only a single functional study of human fluorescence only while a specific DNA lesion is present. Re- MUTYH-associated variants has been reported (48), but an ex- moval of the lesion, regardless of the downstream events or in- tensive study on the functional variation of MUTYH across healthy termediates, results in lack of fluorescence, with the possible but individuals, like the one performed here (showing a 7.2-fold range), unlikely exception of abasic sites produced by DNA glycosylases. had not taken place so far. The low interindividual variation in As such, they measure DNA glycosylase or APE1 activity directly. AAG and APE1 activity across all 24 cell lines is particularly In contrast, transcription blockage constructs only result in fluo- striking. These phenomena may reflect a limitation for these reporters rescence upon completion of repair. to distinguish among cell lines in which the pathway is uniformly active or inactive throughout all samples. AAG activities for all Transcriptional Mutagenesis Insights. Hx preferentially pairs with C cell lines appear to be relatively low [high percent fluorescent during DNA replication (35), and while it is generally considered reporter expression (%R.E.) in Fig. 5A1, similar to the %R.E. by − − that the mispairing events that take place during replication can Aag / MEFs in Fig. 2C), potentially indicating that B lympho- also occur during transcription (36), it was only recently reported blastoid cells in general might not have particularly high AAG in Escherichia coli that Hx can mispair with C during transcrip- activity. In contrast, APE1 activity for all cell lines is relatively tion (37). Our results show that this is also the case, in vivo, in high (high %R.E. in Fig. 5A5, similar to the %R.E. by Ape1++Δ mammalian cells. Similarly, abasic sites follow the “A-rule” in Fig. 4C). Considering that deficiencies in APE1 are associated − − during replication, where A is preferentially incorporated across with disease (49, 50), and Ape1 / mice are embryonic-lethal, it is from the abasic site (36). In contrast, in vitro systems have shown not particularly striking that “healthy” cells stay within a small range conflicting results, indicating that mammalian RNAPII incor- of variation. However, a 4.9-fold range has been reported in a panel porates A, and to a lesser extent G (38), as well as C (39) op- ofa100humanPBMCsforaninvitroAPE1assay(51). posite THF. We observed that RNAPII incorporates A opposite THF in vivo, thus recapitulating the A-rule but for transcription. Mathematical Modeling. Chemo- and radiotherapy are widely used We further engineered a set of reporters to test if transcriptional to treat cancer, and both their effectiveness and the severity of mutagenesis directs incorporation of C, G, or U opposite the side effects depends, at least in part, upon DRC (21, 52–54). The abasic site but they did not result in any measurable fluorescent ability to determine ahead of time whether a particular treat- signal. It thus appears that under in vivo conditions, abasic sites ment would be effective could improve therapy and avoid putting cause a strong transcription blockage that can only be bypassed patients through unnecessary exposures. Recently, we utilized by the incorporation of A opposite the lesion. Moreover, we such an approach to build predictive models relating alkylating complemented the in vitro results (40), showing that transcrip- agent sensitivity to DRC for five repair pathways (21). Here, we tion is blocked by an abasic site in vivo. While it should be taken build upon these methods by incorporating BER capacity mea- into consideration that THF differs chemically from a true AP- surements to evaluate the contribution of 10 different functional site, its properties as a more chemically stable analog of an AP DRC scores toward predicting sensitivity of B lymphoblastoid site and a substrate for APE1 make it more suitable for quan- cells to three DNA damaging agents known to induce damage that titative assays such as FM-HCR. The physiological implications is repaired, in part, by the BER machinery.

Chaim et al. PNAS Early Edition | 7of10 Downloaded by guest on September 30, 2021 The fact that the strongest contributor in all three MLR models Overall, the methods presented here represent powerful tools (Eqs. 1–3) is repair of 8oxoG:C supports the idea that sensitivity for measuring cellular BER capacity in mammalian cells. Addi- for these three agents is driven, in part, by an oxidative stress tionally, we have unveiled previously unknown in vivo tran- response and in consequence, partially, by oxidative DNA damage. scriptional mutagenesis and blockage events that have important Importantly, the association between 8oxoG repair activity and consequences in protein homeostasis and cell physiology fol- resistance to MMS is consistent with the observation that HCT116 lowing DNA damage. Finally, we demonstrate that differences in cells overexpressing OGG1 acquire resistance to MMS-induced DRC between individuals can be measured in lymphocytes iso- cell death (55), as well as with the detection of fluorescence from lated from whole-blood samples. a reactive oxygen species reporter gene, sulfiredoxin-1 (SRXN1), in SRXN1-GFP mouse embryonic stem cells upon MMS treatment Materials and Methods (56). Moreover, MMS indirectly induces an increase in reactive Plasmids. As described previously (15), AmCyan, EGFP, mOrange, mPlum, and oxygen species levels in yeast (57). Oxidative damage to cytosine tagBFP reporter were subcloned into the pmaxCloning Vector (Lonza) can lead to DNA lesions that are substrates for the same glyco- between the KpnI and SacI restriction sites in the multiple cloning site. sylases that excise U (58); as with 5-FU, the seemingly counter- Plasmids were amplified in E. coli DH5α (Invitrogen) and purified using intuitive correlation between U excision activity and sensitivity to Qiagen endotoxin-free Maxi and Giga kits. H O could reflect BER imbalance. Furthermore, increased U 2 2 Generating Plasmids Containing Site-Specific DNA Damage Reporting via excision activity contributes toward sensitivity (rather than re- Transcriptional Mutagenesis or Transcriptional Blockage. Nonfluorescent var- sistance) for all three agents and might reflect the role that UNG iants of the reporter plasmids containing a point mutation at a site coding for and SMUG1 play in the excision of not only U but a variety of their respective chromophores were identified and generated via standard base lesions, such as 5-hydroxyuracil, isodialuric acid, alloxan, QuikChange site-directed mutagenesis (Agilent Technologies) (Table 2). and 5-hydroxymethyluracil (59, 60), all of which are formed, like To produce single-stranded plasmid DNA (ssDNA), previously described 8oxoG, in mammalian cells following oxidative challenge (61). methods were followed (15) with minor modifications. Reporter plasmids Although it remains unclear why BER imbalance is associated were nicked with either Nb.BtsI or Nt.BspQI (New England Biolabs), depending with elevated activity for some glycosylases and not others, we note on the lesion-containing strand (Table 2). The nicked strand was then that OGG1, for which higher activity does not appear to lead to digested with ExoIII (New England Biolabs), and the remaining ssDNA was imbalanced BER, is a bifunctional DNA glycosylase. In contrast, purified using a 1% agarose gel. The circular ssDNA was extracted from the the best-characterized example of BER imbalance (AAG) (3), as agarose gel using a Gel Extraction Kit (Qiagen). Fifteen picomoles of the respective phosphorylated lesion-containing oligonucleotide (Table 2) were well as the present case, wherein elevated uracil DNA glycosylase combined with 3.2 μg of the complementary single-stranded plasmid in 1× activity appears to be toxic to cells upon exposure to agents that Pfu polymerase AD buffer (Agilent Biotechnologies) in a final volume of produce UNG substrates, both involve monofunctional DNA gly- 46 μL (4:1 molar ratio). The mixture was heated to 85 °C in a thermal cycler cosylases. Thus, the differing demands placed on downstream re- for 6 min, and then allowed to anneal by cooling to 40 °C at 1 °C per minute. pair factors by the two DNA glycosylase mechanisms could To extend the primer, five units of Pfu polymerase AD (Agilent) and 0.4 μM potentially explain the opposite correlations between sensitivity to dNTP were added. The incubation parameters used for each ssDNA-plasmid/ 5-FU, H2O2, and OGG1 vs. UNG activity. oligonucleotide combination are shown in Table 2. Following extension, the re- Overall, incorporation of BER measurements into the MLR action mixture was cleaned up with a Qiagen PCR Purification kit column, eluted model for MMS sensitivity results in a substantial improvement in in EB buffer and subsequently combined with 1× Ligase Buffer (New England μ μ μ prediction capacity over our previous modeling, which lacked BER Biolabs), 0.4 MdNTP,1 M ATP, 10 ng/ L BSA, 1.5 units T4 DNA polymerase, and 80 units T4 DNA Ligase (both New England Biolabs), and incubated for an components, specifically regarding the correlation fit following 2 additional hour at 16 °C to yield closed circular plasmid. Finally, the product LOOCV (R CV) (21). This model results in a correlation fit of 0.40 2 was purified from a 1% agarose gel using a Qiagen gel extraction kit. for MMS sensitivity, compared with the previously reported R CV The presence of desired base lesions in the plasmid reporters was con- of 0.07 (21), highlighting the role of BER components in predicting firmed by analytical digests (SI Appendix, Figs. S3–S5). The names for the sensitivity to MMS. We acknowledge the limitations of the models, modified plasmids are listed in Table 2. namely that they are based on measurements on B lymphoblastoid samples, which might not represent the effects of DRC on DNA Cell Culture. The panel of 24 B lymphoblastoid cell lines was obtained from the 2 damage-based chemotherapy in tumors, and that the R CV could Coriell repository. This panel consists of a representative subset of a larger still be substantially improved. As such, further validation of our 450 cell line panel specifically chosen as representative samples of United DRC reporters and mathematical modeling, as well as incorpo- States residents with ancestry from all major regions of the world (20). These ration of other non-DRC variables associated with therapeutic and all other cell lines plus their respective culture conditions are detailed in response to DNA damaging agents, will need to take place in SI Appendix, Table S1. primary “healthy” and tumorous tissues. MUTYH Knockdown Cell Lines. MUTYH was knocked down in HCT116 + 3 cells, as previously described (64). shRNAs expressed from a Interindividual Variation in DRC Can Be Measured in Fresh Blood lentiviral plasmid (pGIPZ) were purchased from Open Biosystems to target Samples. Because blood can be accessed easily in a minimally MUTYH transcript (v1:#RHS4430-98904053 and v2:#RHS4430-99140608). invasive manner, it is commonly used as a surrogate for assessing Knockdown cells were compared with cells expressing a nontargeting shRNA the biological properties of other tissues. For example, a direct (#RHS4346). Virus was generated in 293T cells using packaging plasmids correlation between 8oxoG repair capacity in blood and lung (psPAX2 and pMD2.G; Addgene). Cells were infected with virus and stable tissue was previously established through in vitro assays (62). clones were selected using Puromycin. Furthermore, numerous epidemiological studies have demon- strated an association between cancer and DRC in lymphocytes MUTYH Nuclear Overexpression. Nuclear MUTYH isoform 4 cDNA was cloned (12, 63), highlighting the increased potential that the use of in into a retroviral plasmid (pBABE, C-terminal Flag-tagged). HCT116 + Chro- vivo DRC measurements in isolated blood can have on assessing mosome 3 cells overexpressing MUTYH were compared with the same cells an individual’s overall DRC and, more specifically, as a biomarker expressing empty vector. Virus was generated in 293T cells, and stable clones were selected using Puromycin. for disease risk. Even though it is still to be determined if these ’ kinds of measurements can help inform an individual s predispo- In Vitro MUTYH DNA Glycosylase Assays. Cells were sonicated in MUTYH DNA sition to disease or their potential tolerance to treatment with glycosylase dilution buffer (30 mM Tris pH 7.5, 1 mM EDTA, 1 mM β-mercap- DNA damaging agents (12), we not only demonstrate that our toethanol, 50 mM NaCl, and 30% glycerol) with protease inhibitor mixture. Pro- DRC reporters can be utilized in primary blood samples but also tein concentration was measured using micro BCA Kit (Pierce). DNA glycosylase that they can detect differences between individuals. assays were performed as previously published (65). A double-stranded

8of10 | www.pnas.org/cgi/doi/10.1073/pnas.1712032114 Chaim et al. Downloaded by guest on September 30, 2021 Table 2. Site-specific plasmids and extension conditions used for their production PNAS PLUS Extension Extension Plasmid name Nicking enzyme Sequence X Source temperature, °C time, h

BFP-A191G-U Nb.BtsI 5′ GGT CTT GCT GCC GXA GAG GAA GCT AGT AGC U IDT 64.0 3.5 GFP-C289T-dHx Nb.BtsI 5′ GAA GAA GAT GGT GCX CTC CTG GAC GTA GCC I EGT 61.0 2.0 GFP-617-THF Nb.BtsI 5′ GCT CAG GGC GXA CTG GGT GCT CAG GTA GTG THF IDT 66.0 1.0 mOrange-A215C-8oxoG Nb.BtsI 5′ GTA GGC CTT GGA GCC GXA GGT GAA CTG AGG 8oxoG EGT 64.0 2.5 mOrange-A215C-THF Nb.BtsI 5′ GCC TTG GAG CCG XAG GTG AAC TGA GGG GAC THF IDT 66.0 1.5 mPlum-T202WT-8oxoG (+) Nt.BspQI 5′ GTC CCC TCA GAT CAT GXA CGG CTC CAA GGC 8oxoG EGT 61.0 2.0

Numbers in the plasmid names indicate the position within the fluorescent reporter gene at which site-specific lesions were incorporated. The bases shown before and after the number correspond to the identity of the mutation in the nontranscribed strand, followed by the identity of the lesion present at the site. As an example, GFP-C289T-Hx corresponds to a nonfluorescent variant of GFP for which the base at position 289 was mutated from a C in the coding strand to a T,and a Hx was incorporated on the transcribed strand at the site. All base lesions (represented by an underlined X) were positioned on the transcribed strand unless otherwise shown by the symbol “(+).” GFP-617-THF corresponds to a nonmutated WT GFP variant for which site 617 was modified by incorporating a THF on the transcribed strand. EGT, Eurogentech; I, Inosine; IDT, Integrated DNA Techonologies; THF, tetrahydrofuran; U, Uracil.

oligonucleotide containing a [32P]γ-labeled strand (5′-TTGGGGAATGAG- MLR Models. Z-scored DRC values for each of 10 FM-HCR reporters used to TCAGGCCAC-3′) and a nonlabeled strand (5′-GGTGGCCTGAC8oxoGCATT- analyze DRC in the 24 Coriell lymphoblastoid cell lines, served as the (in- CCCCAA-3′) was incubated with an amount of extract determined to be in dependent) predictor variables [five BER reporters developed here (SI Ap- the linear range for activity at 37 °C for 60 min (base pair serving as target sub- pendix, Table S2); five previously published reporters for other pathways strate for MUTYH is underlined). The resulting AP sites were cleaved by incubation (15) (SI Appendix, Table S3)]. Sensitivity to three DNA damaging agents for with 0.1 M NaOH at 75 °C for 15 min, the mixture heat denatured and subjected the same 24 cell lines is reported in SI Appendix, Table S4 and expressed in – to 7 M urea 10% polyacrylamide gel electrophoresis. Phosphorimaging was terms of %C.G. MLR models were developed as described previously (21) and used to visualize and quantitate MUTYH DNA glycosylase activity. as detailed below, and take the following form:

DNA Repair Assay Transfections. = β + β + β + ... + β + ; Y 1x1 2x2 3x3 nxn b [4] 6

Electroporation. Cells growing in suspension (3 × 10 cells in 100 μL complete CELL BIOLOGY medium) were combined with a reporter plasmid mixture (Table 1) and electro- where Y represents the predicted response variable (sensitivity) reported porated using a 96-well Bio-Rad MXcell gene pulser, with an exponential wave- as %C.G.; xi, are the z-scored DNA repair capacity predictor variables for the

form at 260 V and 950 μF. Following electroporation, 100 μL complete medium 10 substrates reported as %R.E., βi are the substrate slopes along the corre- was added to each well in the electroporation plate and gently mixed. The sponding dimensions, and b is a constant that represents the y-axis intercept. electroporation mix was transferred to a 24-well cell culture plate, each well Z-scores for relative DNA repair capacity among the 24 cell lines were prefilled with 1.3 mL of complete medium and incubated at 37 °C and 5% generated for each pathway as follows, CO2 for 18 h; cells were then centrifuged for 5 min at 300 × g, resuspended in 250 μL of complete media containing TO-PRO-3 dead stain, and trans- xij − xj Zij = ; [5] ferred to a 96-well plate for flow cytometry. σj Lipofection. For lipofection, 150,000 adherent cells were plated in six-well plates 1 d before transfection to achieve 50–80% confluency on the day of transfection. where Zij is the z-score for a DNA repair pathway in cell line i in pathway j, xij Transfection with Lipofectamine LTX (Life Technologies) was performed accord- is the DRC for a given pathway j in cell line i, xj is the mean value of the DNA ing to the manufacturer’s instructions. Briefly, 2.5 μg of total plasmid DNA (Table repair capacity in pathway j over the 24 lymphoblastoid cell lines (i = 1–24), μ μ 1) mixed with 2.5 L Plus reagent and Opti-MEM, was further mixed with 6.2 L and σj is the SD of the DNA repair capacity in pathway j. MGMT scores were

lipofectamine LTX in Opti-MEM, and incubated at room temperature for 5 min transformed to logMGMT before z-scoring. μ before layering 200 L of the transfection reaction on top of the cells. Lipofection MLR models were generated by running MATLAB’s regress function. β-Scores conditions were scaled down for 12- and 24-well plates when necessary. Transfected (slopes) for each variable as well as the model fit to the data (R2) were calcu- cells were incubated for 18 h at 37 °C and 5% CO2, then trypsinized and resus- lated for each generated model. LOOCV was used to assess the prediction pended in a total of 250 μL of complete media containing TO-PRO-3 dead stain and power of each model. The correlation coefficient (R2 ) between the observed transferred to 75-mm Falcon tubes with Cell Strainer Caps (Fisher Scientific). CV cellular sensitivities and the predicted ones following LOOCV as well as their RMSE were calculated for each model. Initial models were generated by in- Flow Cytometry. Flow cytometric parameters and calculation of %R.E. have cluding all variables (model 0). To identify the optimal models (Eqs. 1–3), vari- been previously described (15, 34) and are detailed in SI Appendix, Supple- able selection was carried out by sequentially removing predictor variables with mentary Materials and Methods. the lowest contribution to the model (lowest β-slopes). In cases where two or more variables were tied for the smallest slope (β), each of the possible models Cell Sensitivity Assay. Chemicals. Hydrogen Peroxide (H O ) (Sigma) was prepared at working were tested wherein one of the tied variables was removed. The combination 2 2 2 concentrations in complete media. Next, 5-FU (Sigma) was prepared as a of variables that yielded the smallest RMSE and largest R CV was selected as the 100 mM stock in DMSO, and diluted to working concentrations in complete best model for each DNA damaging agent (SI Appendix,TableS7). media before addition to cells. XTT assay. For XTT assay, 25,000 lymphoblastoid cells were seeded in 100 μL PBMCs. Fresh whole blood from two different donors, collected in sodium in 96-well tissue culture plates and incubated with drugs (added in a volume heparin Vacutainer collection tubes, was purchased from Research Blood

of 50 μL complete media to a final concentration of 25 μMH2O2 and 100 μM Components. PBMCs were isolated using standard Ficoll-Paque (GE) density 5-FU). Each condition, including controls, was set up in three replicate gradient centrifugation. PBMC freezing and thawing procedures as well as wells. After 3 d, 2,3-bis(2-methoxy-4-nitro-5-sulfophenly)-5-[(phenylamino)- PHA stimulation of T lymphocytes and transfection conditions are described carbonyl]-2H-tetrazolium hydroxide (XTT) reagent (Cell Signaling Technol- in detail in SI Appendix, Supplementary Materials and Methods. ogy) was added to each well (including blank wells containing complete μ medium without cells), in a volume of 50 L. XTT reagent was activated im- ACKNOWLEDGMENTS. We thank Dr. Kefei Yu (Michigan State University) mediately before use by addition of an electron coupling reagent according to for the apurinic/apyrimidinic endonuclease 1-deficient cells, and Dr. Samuel the XTT kit protocol. After 24 h of additional incubation at 37 °C, absorption at Wilson (National Institute on Environmental Health Sciences) for the uracil 450 nm was measured. Results are shown as %C.G. representing the ratio of DNA glycosylase-deficient cells. This work was supported by a NIH Directors A450 in treated wells over A450 in untreated wells after subtraction of no-cell Pioneer Award DPI-ES022576, and NIH Grants R01-CA075576, R01-CA55042, controls. Each experiment was done at least in triplicate. MMS sensitivity was R01-CA149261, R01-ES022872, and P30-ES02109. L.D.S. is American Cancer measured previously using a Trypan blue exclusion assay (21). Society Research Professor Emeritus.

Chaim et al. PNAS Early Edition | 9of10 Downloaded by guest on September 30, 2021 1. Ciccia A, Elledge SJ (2010) The DNA damage response: Making it safe to play with 35. Nordmann PL, Makris JC, Reznikoff WS (1988) Inosine induced mutations. Mol Gen knives. Mol Cell 40:179–204. Genet 214:62–67. 2. Kim Y-J, Wilson DM, 3rd (2012) Overview of base excision repair biochemistry. Curr 36. Brégeon D, Doetsch PW (2011) Transcriptional mutagenesis: Causes and involvement Mol Pharmacol 5:3–13. in tumour development. Nat Rev Cancer 11:218–227. 3. Fu D, Calvo JA, Samson LD (2012) Balancing repair and tolerance of DNA damage 37. Morreall J, et al. (2015) Evidence for retromutagenesis as a mechanism for adaptive caused by alkylating agents. Nat Rev Cancer 12:104–120. mutation in Escherichia coli. PLoS Genet 11:e1005477. 4. Hoeijmakers JHJ (2001) Genome maintenance mechanisms for preventing cancer. 38. Damsma GE, Alt A, Brueckner F, Carell T, Cramer P (2007) Mechanism of transcrip- – Nature 411:366 374. tional stalling at cisplatin-damaged DNA. Nat Struct Mol Biol 14:1127–1133. 5. Wallace SS (2014) Base excision repair: A critical player in many games. DNA Repair 39. Kuraoka I, et al. (2003) Effects of endogenous DNA base lesions on transcription – (Amst) 19:14 26. elongation by mammalian RNA polymerase II. Implications for transcription-coupled 6. Jones S, et al. (2002) Biallelic germline mutations in MYH predispose to multiple co- DNA repair and transcriptional mutagenesis. J Biol Chem 278:7294–7299. lorectal adenoma and somatic G:C–>T:A mutations. Hum Mol Genet 11:2961–2967. 40. Tornaletti S, Maeda LS, Hanawalt PC (2006) Transcription arrest at an abasic site in the 7. Imai K, et al. (2003) Human uracil-DNA glycosylase deficiency associated with pro- transcribed strand of template DNA. Chem Res Toxicol 19:1215–1220. foundly impaired immunoglobulin class-switch recombination. Nat Immunol 4: 41. Brenerman BM, Illuzzi JL, Wilson DM, 3rd (2014) Base excision repair capacity in in- 1023–1028. forming healthspan. Carcinogenesis 35:2643–2652. 8. Shao C, et al. (2008) Altered 8-oxoguanine glycosylase in mild cognitive impairment 42. Leitner-Dagan Y, et al. (2014) Enzymatic MPG DNA repair assays for two different oxi- and late-stage Alzheimer’s disease brain. Free Radic Biol Med 45:813–819. dative DNA lesions reveal associations with increased lung cancer risk. Carcinogenesis 35: 9. Obtułowicz T, et al. (2010) Aberrant repair of etheno-DNA adducts in leukocytes and – colon tissue of colon cancer patients. Free Radic Biol Med 49:1064–1071. 2763 2770. 10. Sevilya Z, et al. (2014) Low integrated DNA repair score and lung cancer risk. Cancer 43. Calvo JA, et al. (2013) Aag DNA glycosylase promotes alkylation-induced tissue damage Prev Res (Phila) 7:398–406. mediated by Parp1. PLoS Genet 9:e1003413. 11. Matuo R, et al. (2010) DNA repair pathways involved in repair of lesions induced by 44. Crosbie PAJ, et al. (2012) Elevated N3-methylpurine-DNA glycosylase DNA repair ac- – 5-fluorouracil and its active metabolite FdUMP. Biochem Pharmacol 79:147–153. tivity is associated with lung cancer. Mutat Res 732:43 46. 12. Chaim IA, Nagel ZD (2017) Assessing BER capacity in the human population. The Base 45. Janssen K, et al. (2001) DNA repair activity of 8-oxoguanine DNA glycosylase 1 (OGG1) Excision Repair Pathway: Molecular Mechanisms and Role in Disease Development in human lymphocytes is not dependent on genetic polymorphism Ser326/Cys326. and Therapeutic Design (World Scientific, Singapore), pp 557–607. Mutat Res 486:207–216. 13. Guerniou V, et al. (2005) Repair of oxidative damage of thymine by HeLa whole-cell 46. Paz-Elizur T, et al. (2007) Development of an enzymatic DNA repair assay for mo- extracts: Simultaneous analysis using a microsupport and comparison with traditional lecular epidemiology studies: Distribution of OGG activity in healthy individuals. DNA PAGE analysis. Biochimie 87:151–159. Repair (Amst) 6:45–60. 14. Sauvaigo S, et al. (2004) An oligonucleotide microarray for the monitoring of repair 47. Myrnes B, Giercksky K-E, Krokan H (1983) Interindividual variation in the activity of enzyme activity toward different DNA base damage. Anal Biochem 333:182–192. O6-methyl guanine-DNA methyltransferase and uracil-DNA glycosylase in human 15. Nagel ZD, et al. (2014) Multiplexed DNA repair assays for multiple lesions and mul- organs. Carcinogenesis 4:1565–1568. tiple doses via transcription inhibition and transcriptional mutagenesis. Proc Natl 48. Raetz AG, et al. (2012) Cancer-associated variants and a common polymorphism of – Acad Sci USA 111:E1823 E1832. MUTYH exhibit reduced repair of oxidative DNA damage using a GFP-based assay in 16. van Loon B, Markkanen E, Hübscher U (2010) Oxygen as a friend and enemy: How to mammalian cells. Carcinogenesis 33:2301–2309. – combat the mutational potential of 8-oxo-guanine. DNA Repair (Amst) 9:604 616. 49. Vasko MR, Guo C, Kelley MR (2005) The multifunctional DNA repair/redox enzyme 17. McFaline-Figueroa JL, et al. (2015) Minor changes in expression of the mismatch re- Ape1/Ref-1 promotes survival of neurons after oxidative stress. DNA Repair (Amst) 4: pair protein MSH2 exert a major impact on glioblastoma response to temozolomide. 367–379. Cancer Res 75:3127–3138. 50. Fung H, Demple B (2005) A vital role for Ape1/Ref1 protein in repairing spontaneous 18. Wilson DM, 3rd, Takeshita M, Grollman AP, Demple B (1995) Incision activity of hu- DNA damage in human cells. Mol Cell 17:463–470. man apurinic endonuclease (Ape) at abasic site analogs in DNA. J Biol Chem 270: 51. Sevilya Z, et al. (2014) Low integrated DNA repair score and lung cancer risk. Cancer 16002–16007. Prev Res (Phila) 7:398–406. 19. Masani S, Han L, Yu K (2013) Apurinic/apyrimidinic endonuclease 1 is the essential 52. Jalal S, Earley JN, Turchi JJ (2011) DNA repair: From genome maintenance to bio- nuclease during immunoglobulin class switch recombination. Mol Cell Biol 33: – 1468–1473. marker and therapeutic target. Clin Cancer Res 17:6973 6984. 20. Collins FS, Brooks LD, Chakravarti A (1998) A DNA polymorphism discovery resource 53. Oeffinger KC, Bhatia S (2009) Second primary cancers in survivors of childhood cancer. – for research on human genetic variation. Genome Res 8:1229–1231. Lancet 374:1484 1485. 21. Nagel ZD, et al. (2017) DNA repair capacity in multiple pathways predicts chemo- 54. Wood ME, et al. (2012) Second malignant neoplasms: Assessment and strategies for – resistance in glioblastoma multiforme. Cancer Res 77:198–206. risk reduction. J Clin Oncol 30:3734 3745. 22. Fischer DS, Knobf MT, Durivage HJ, Beaulieu NJ (2003) The Cancer Chemotherapy 55. Lee M-R, et al. (2004) Transcription factors NF-YA regulate the induction of human Handbook (Mosby, Philadelphia). OGG1 following DNA-alkylating agent methylmethane sulfonate (MMS) treatment. 23. Hill JW, Hazra TK, Izumi T, Mitra S (2001) Stimulation of human 8-oxoguanine-DNA J Biol Chem 279:9857–9866. glycosylase by AP-endonuclease: Potential coordination of the initial steps in base 56. Hendriks G, et al. (2016) The extended ToxTracker assay discriminates between in- excision repair. Nucleic Acids Res 29:430–438. duction of DNA damage, oxidative stress, and protein misfolding. Toxicol Sci 150: 24. Mokkapati SK, Wiederhold L, Hazra TK, Mitra S (2004) Stimulation of DNA glycosylase 190–203. activity of OGG1 by NEIL1: Functional collaboration between two human DNA gly- 57. Griffiths LM, et al. (2009) Dynamic compartmentalization of base excision repair cosylases. Biochemistry 43:11596–11604. proteins in response to nuclear and mitochondrial oxidative stress. Mol Cell Biol 29: 25. Monden Y, et al. (1999) Human MMH (OGG1) type 1a protein is a major enzyme for 794–807. repair of 8-hydroxyguanine lesions in human cells. Biochem Biophys Res Commun 58. Kreutzer DA, Essigmann JM (1998) Oxidized, deaminated cytosines are a source of – 258:605 610. C–> T transitions in vivo. Proc Natl Acad Sci USA 95:3578–3582. 26. Dou H, Mitra S, Hazra TK (2003) Repair of oxidized bases in DNA bubble structures by 59. Kavli B, et al. (2002) hUNG2 is the major repair enzyme for removal of uracil from U:A – human DNA glycosylases NEIL1 and NEIL2. J Biol Chem 278:49679 49684. matches, U:G mismatches, and U in single-stranded DNA, with hSMUG1 as a broad η 27. Wilson TM, et al. (2005) MSH2-MSH6 stimulates DNA polymerase , suggesting a role specificity backup. J Biol Chem 277:39926–39936. – for A:T mutations in antibody genes. J Exp Med 201:637 645. 60. Dizdaroglu M, Karakaya A, Jaruga P, Slupphaug G, Krokan HE (1996) Novel activities 28. Hadi MZ, Wilson DM, 3rd (2000) Second human protein with homology to the Es- of human uracil DNA N-glycosylase for cytosine-derived products of oxidative DNA cherichia coli abasic endonuclease exonuclease III. Environ Mol Mutagen 36:312–324. damage. Nucleic Acids Res 24:418–422. 29. Fortini P, et al. (2003) 8-Oxoguanine DNA damage: At the crossroad of alternative 61. Dizdaroglu M, Nackerdien Z, Chao B-C, Gajewski E, Rao G (1991) Chemical nature of in repair pathways. Mutat Res 531:127–139. vivo DNA base damage in hydrogen peroxide-treated mammalian cells. Arch Biochem 30. Mi R, Alford-Zappala M, Kow YW, Cunningham RP, Cao W (2012) Human endonu- Biophys 285:388–390. clease V as a repair enzyme for DNA deamination. Mutat Res 735:12–18. 62. Paz-Elizur T, et al. (2003) DNA repair activity for oxidative damage and risk of lung 31. Moe A, et al. (2003) Incision at hypoxanthine residues in DNA by a mammalian ho- – mologue of the Escherichia coli antimutator enzyme endonuclease V. Nucleic Acids cancer. J Natl Cancer Inst 95:1312 1319. Res 31:3893–3900. 63. Nagel ZD, Chaim IA, Samson LD (2014) Inter-individual variation in DNA repair ca- 32. Ringvoll J, et al. (2008) AlkB homologue 2-mediated repair of ethenoadenine lesions pacity: A need for multi-pathway functional assays to promote translational DNA in mammalian DNA. Cancer Res 68:4142–4149. repair research. DNA Repair (Amst) 19:199–213. 33. Lee CYI, et al. (2009) Recognition and processing of a new repertoire of DNA sub- 64. Fry RC, et al. (2008) Genomic predictors of interindividual differences in response to strates by human 3-methyladenine DNA glycosylase (AAG). Biochemistry 48: DNA damaging agents. Genes Dev 22:2621–2626. 1850–1861. 65. Shinmura K, et al. (2000) Adenine excisional repair function of MYH protein on the 34. Chaim IA, et al. (2017) A novel role for transcription-coupled nucleotide excision re- adenine:8-hydroxyguanine base pair in double-stranded DNA. Nucleic Acids Res 28: pair for the in vivo repair of 3,N4-ethenocytosine. Nucleic Acids Res 45:3242–3252. 4912–4918.

10 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1712032114 Chaim et al. Downloaded by guest on September 30, 2021