Recruitment of KMT2C/MLL3 to DNA Damage Sites Mediates DNA Damage Responses and Regulates PARP Inhibitor Sensitivity in Cancer

Chromatins are remodeled through posttranslational modifications of histones during transcription, replication, and DNA repair. Methylation of histone 3 (H3) at lysine 4 (K4) catalyzed by the lysine methyltransferase (KMT) family proteins is a mark of transcriptional activity (1). KMT2A and KMT2B catalyze H3K4 dimethylation (H3K4me2) and trimethylation (H3K4me3), while KMT2C and KMT2D mediate H3K4 monomethylation (H3K4me1). KMT2 proteins are enriched at enhancers and promoters, where they promote recruitment of transcriptional activators by opening chromatin through H3K4 methylation (2). KMT2C and KMT2D are also recruited to stalled replication forks to mediate H3K4me1 and chromatin opening required for MRE11 recruitment for replication restart (3–5). Mutations of all 4 KMT2 have been linked to cancer (1). KMT2A was found in gene fusions in childhood leukemias. Hepatitis B virus is inserted in KMT2B in hepatocellular carcinoma (HCC). KMT2C and KMT2D are frequently inactivated by deletions or mutations in solid tumors. Although knockout of the entire KMT2C gene is lethal in mice (6), deletion of the catalytic core of KMT2C SET domain induced cellular hyperproliferation and urothelial tumor formation (7), supporting a tumor-suppressing role of KMT2C. The mechanisms underlying the contribution of KMT2 alterations to cancer and the tumor-suppressing role of KMT2 are unknown or attributed to transcription.

DNA damage response (DDR) involves checkpoint activation and DNA repair upon DNA damage (8), which relies on assembly of DDR factors into foci at damaged sites in 2 steps called primary and secondary recruitment (9–12). Double-strand breaks (DSB) initially trigger primary recruitment in which lesions are recognized by MRE11–RAD50–NBS1 (MRN) complex, which recruits and activates ATM, resulting in phosphorylation of histone H2A.X (γH2A.X) in proximal regions of DSBs. MDC1 then binds to γH2A.X and triggers secondary recruitment by recruiting more MRN and activated ATM, facilitating spreading of γH2A.X to distal regions of DSBs. MDC1 also recruits RNF8/RNF168 that ubiquitylate chromatins, leading to recruitment of 53BP1, BRCA1, and others mediating DNA repair. Important to DDR focus assembly is chromatin relaxation via histone modifications at damaged sites, including H4K16 acetylation (13) and H4K20 methylation (14).

Recent studies identified an essential role of small noncoding RNAs (DNA damage response RNAs, DDRNAs) in DDR (15–17). Upon DNA damage, RNA polymerase II is recruited to DSBs by the MRN complex and synthesizes damage-induced long noncoding RNAs (dilicRNA) from sequences near DSBs, which are processed into DDRNAs by Drosha and Dicer (15, 16, 18). DDRNAs, and Drosha, Dicer, and Ago2, are essential for histone modifications, chromatin remodeling, and assembly of DDR foci at damage sites (17) and for secondary recruitment (19).

Prompted by the correlation between frequent KMT2 mutations with high mutation loads in non–small cell lung cancer (NSCLC), we identified an essential role of the KMT2 in DDR and DNA repair. KMT2C is directly recruited to DNA damage sites by DDRNAs and Ago2, where it mediates H3K4me1 and chromatin relaxation, allowing secondary recruitment of DDR factors and amplification of DDR signaling required for checkpoint activation and DNA repair. These studies have revealed a novel role of KMT2C and KMT2C-mediated H3K4me1 in DDR, besides their known functions in transcription. Furthermore, disruption of homologous recombination (HR)–mediated DNA repair by KMT2C/D inactivation sensitizes NSCLC cells and PDX to Poly ADP Ribose Polymerase inhibitors (PARPi), which are currently used to treat patients with breast, ovarian, and prostate cancer with BRCA1/2 mutations. Thus, KMT2C/D mutations are potentially novel biomarkers for the efficacy of PARPi therapies in NSCLC and other cancers with rare BRCA1/2 mutations.

Cell lines were obtained from ATCC (BJ, U2OS, NCI-H2342, NCI-H2135 NCI-H1869, NCI-H1975, NCI-H522, NCI-H226, HCC-827, and Calu-1), Life Cell Technology (HSAEC), Sigma (COR-L105), or RIKEN (LK-2, RERF-LC-AI, EBC-1 cells, and LC-1/sq-SF) and frozen after expanded for two passages. Cells were cultured for no more than 8 passages for the studies at 37°C in a humidified incubator containing 5% CO₂ in vendor-specified media described in Supplementary Methods and authenticated twice a year by expression profiling and by mycoplasma tests using Mycoplasma Detection Kit (R&D Systems).

Six-week-old female athymic nude mice were purchased from Charles River Laboratories and injected subcutaneously with 5 × 10⁵ COR-L105, 1 × 10⁶ NCI-H1975, 3 × 10⁶ NCI-H2135 or HCC-827, or 5 × 10⁶ NCI-H2342 or NCI-H522 cells, respectively, and treated intraperitoneal daily with 50 mg/kg of olaparib or vehicle starting on day 7 when tumors were 40 to 50 mm³ in size. Tumor volumes were measured every 4 days. Tumor tissues were harvested at the end of the studies. Apoptosis in tumor sections was analyzed by the DeadEnd Fluorometric TUNEL System (Promega) according to the manufacturer's protocol. More details are provided in Supplementary Methods.

Human lung cancer tissues were obtained from the Tumor Tissue and Pathology Shared Resource of Wake Forest Baptist Comprehensive Cancer Center (WFBCCC, Winston-Salem, NC). IHC for DDR markers was performed as described previously (20), with modifications of conditions for each antibody (Supplementary Table S1). Percent of tumor cells with positive nuclear staining was quantitated. More details are provided in Supplementary Methods.

Antibodies used in this study are provided in Supplementary Table S1.

Lentiviral vectors encoding short hairpin RNAs (shRNA) were designed based on the single-oligonucleotide RNA interference technology (Biosettia). DNA oligonucleotides for each shRNA (Supplementary Table S2) were cloned into the lentiviral pLV-H1-EF1α-puro (SORT-B19) or pLV-H1-EF1α-bsd (SORT-B22) vectors, following the manufacturer's protocol, and verified by DNA sequencing. Oligos of KMT2C sgRNAs (Supplementary Table S3) were designed using tools from CRISPOR.org and cloned into lentiCRISPRv2 (Zhang Lab GeCKO). Full-length human KMT2C coding sequence was chemically synthesized (Sango Biotech), cloned into pCMV-IRES2-eGFP (BD-Clontech) between XhoI and BamHI, and verified by DNA sequencing.

Recombinant retroviruses and lentiviruses were packaged and transduced into cells as described previously (21, 22). Transduced cells were purified with 80 (BJ) or 50 (HSAECs) μg/mL of hygromycin B, or 1.5 (BJ and HSAECs) or 2 (U2OS and NSCLC cells) μg/mL of puromycin.

Total RNA was isolated from cells using TRIzol reagent, transcribed to cDNA using iScript Reverse Transcription SuperMix (Bio-Rad), and quantified by qRT-PCR using SsoAdvanced SYBR Green SuperMix (Bio-Rad) in Bio-Rad CFX96 and primers in Supplementary Table S4. following the manufacturer's protocols. GAPDH was used as internal control to normalize the mRNA level for each gene.

Western blotting was performed as described before (22). More details are provided in Supplementary Methods.

DNA DSB resection assay was performed as reported previously (23, 24). U2OS-ER-AsiSI cells (provided by Dr. Legube) were treated with 500 nmol/L 4-OHT for 4 hours to induce DSBs. Genomic DNA was extracted using Genomic DNA Extraction Kit (FAVORGEN). Resection efficiency at selected sites was determined by qPCR using primers in Supplementary Table S5 and calculated with the equation: % ssDNA  =  1/[2⁁(ΔCT − 1) + 0.5] × 100. More details are provided in Supplementary Methods.

DDR factor enrichment at AsiSI/I-PpoI–induced DSBs was analyzed as reported previously (25–27). U2OS cells were treated with 500 nmol/L 4-OHT for 12 (I-PpoI) or 2 (AsiSI) hours, crosslinked with 0.67 mg/mL EGS for 20 minutes and 1% formaldehyde for 15 minutes, and subjected to chromatin immunoprecipitation (ChIP)-qPCR assay as described before (20, 28) using primers in Supplementary Tables S6 (I-PpoI) and S7 (AsiSI). ChIP-qPCR data were normalized to Input-qPCR data (29). More details are provided in Supplementary Methods.

U2OS-ER-AsiSI cells were treated with 500 nmol/L 4-OHT for 2 hours to induce DSBs, and analyzed by Chromatin Accessibility Assay Kit according to the manufacturer's protocol (EPIGENTEK), using primers around AsiSI-induced DSBs (Supplementary Table S7). Chromatin remodeling was determined by NaCl solubility assay as described previously (20) in bleomycin-treated cells. More details are provided in Supplementary Methods.

RNaseA treatment and DDRNA complementation were performed, and synthetic DDRNAs designed, as reported previously (16, 17). Complementary DDRNA oligonucleotides (Supplementary Table S8) were derived from a 150-bp region surrounding the I-PpoI site on Chromasome1, a 500-bp region surrounding AsiSI HR site-1 or HR site-2, or the IRF6 promoter on Chromosome 1 (negative control). More details are provided in Supplementary Methods.

Chromatin fractions were isolated as reported previously (30). BJ, U2OS, or HSAECs were treated with 10 μg/mL bleomycin for 6 hours, or irradiated with 2 Gy ionizing radiation and recovered for indicated time (time course). Chromatin fractions and whole-cell lysates were analyzed by Western blotting. More details are provided in Supplementary Methods.

Cells were treated with 10 μg/mL bleomycin for 6 hours, permeabilized with 2% Tween 20 in PBS for 10 minutes, treated with 1 mg/mL RNaseA for 25 minutes, and lysed in RIPA buffer for coimmunoprecipitation (Co-IP) assay as described previously (31).

BJ cells treated with bleomycin or γ-radiation were subjected to neutral comet assays to detect DSBs using CometAssay Kit (Trevigen). More details are provided in Supplementary Methods.

Immunostaining for DDR foci was performed as described previously (16, 20) in cells treated with bleomycin, olaparib, or γ-radiation, using antibodies in Supplementary Table S1. Focus number and size were quantified by ImageJ. More details are provided in Supplementary Methods.

Laser microirradiation was performed as reported previously (32). Cells were then fixed and permeated for immunostaining of DDR markers as described above. More details are provided in Supplementary Methods.

DNA repair was analyzed using GFP reporters for HR (33), nonhomologous end joining (NHEJ; ref. 34), and alt-NHEJ (MMEJ; ref. 25) provided by Drs. Wu (Scripps Research Institute, San Diego, CA) and Gorbunova (University of Rochester, Rochester, NY) following published protocols. More details are provided in Supplementary Methods.

BJ cells were seeded on cover glass in a 24-well plate, cultured for 24 to 48 hours, treated with 2 μg/mL bleomycin plus 10 μmol/L bromodeoxyuridine (BrdUrd) for 2 hours at 37°C, and stained with an anti-BrdUrd antibody (Abcam) and an Alexa Fluor-488 anti-Rat secondary antibody (Cell Signaling Technology) according to the manufacturer's protocol. BrdUrd-incorporating cells were analyzed by fluorescence microscope. Fifty cells/field and a total of 5 fields were counted. Each experiment was performed in triplicates.

Cells (2,000/well) were plated into 96-well plate overnight, treated with indicated concentrations of plaparib (Adooq Bioscience) or cisplatin (Sigma) for 6 days, and incubated with CCK8 substrate (Dojindo) for 1 to 2 hours at 37°C. Cell viability was determined by the absorbance at 450 nm. Cells without treatment and no-cell control were set as 100% and 0%, respectively.

Twenty-four to 48 hours after seeding (2 × 10⁶/10-cm dish), NCI-H2342 or LC-1/sq-SF cells (∼70% confluent) were transfected with 10 μg pCMV-IRES2-eGFP or 20 μg pCMV-KMT2C-IRES2-eGFP using Lipofectamine-2000 (Invitrogen), following the manufacturer's protocols. Cells were collected 48 hours later, resuspended in DPBS/2% FBS/Antibiotic-Antimycotic, live-sorted for eGFP positive cells by flow cytometry (BD FACSAria III), and immediately subjected to analyses.

NSCLC patient-derived xenografts (PDX) were analyzed by Lide Biotech Co on a fee-for-service basis. For in vitro analysis, after digestion by collagenases, tumor cells were isolated by centrifugation on Histopaque (Sigma) and treated with olaparib for 6 days. Cell viability was measured by CellTiter-Glo Reagent (Promega). Cells without treatment and no-cell control were set as 100% and 0%, respectively. For in vivo analysis, 6-week-old female nu/nu mice were transplanted subcutaneously with PDX, and treated intraperitoneal daily with 50 mg/kg olaparib or vehicle when tumors reached 100 to 200 mm³. Tumor volumes were calculated by length × width²/2. More details are provided in Supplementary Methods.

Genomic and clinical data of patients from The Cancer Genome Atlas (TCGA; refs. 35, 36) and MSKCCC (37) cohorts were downloaded from the latest version of cBioPortal (38, 39). Survival estimates were generated using the Kaplan–Meier method; curves compared with log-rank tests. Mutation data of cell lines were obtained from the Broad Institute Cancer Cell Line Encyclopedia (CCLE; refs. 40–42). Genes related to HR were obtained from the Gene Oncology (GO) term GO:0000724. The R deconstructSigs package (43) was adopted to determine the composition of mutational signatures (44) in tumors from TCGA. Tumor mutation loads, gene expression, and contributions of mutation signatures were compared between groups by Mann–Whitney tests.

Error bars in figures were derived from technical replicates in a single experiment. Each experiment was repeated at least 3 times. Although variations among biological repeats were usually large, the trends in each biological repeat were consistent, leading to same conclusions.

All animal studies were approved by the Institutional Animal Care and Use Committee. Human tumor samples were obtained from Tumor Tissue and Pathology Shared Resource of WFBCCC under an IRB-approved protocol.

Consistent with our published findings from patients at WFBCCC (45), analysis of the genomic data from lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC) patients in TCGA cohort revealed frequent mutations in chromatin remodeling (CR) factor genes including KMT2A, KMT2B, KMT2C, KMT2D, SMARCA4, ATRX, SETD2, and ARID1A. KMT2C and D had the highest mutation rates in both LUAD and LUSC (Fig. 1A and B). Although missense (MS) mutations were the most prevalent, high percentages of the KMT2C/D mutations were frameshift (FS) or nonsense (NS; Fig. 1A and B), suggesting that a large portion of these mutations are loss-of-function. No hotspot mutations were observed, although mutations seemed clustered in the N-terminus of KMT2C and D (Supplementary Fig. S1A–S1D). Mutations in KMT2, especially KMT2C/D, correlated with poor disease-free survival in LUAD (Fig. 1C and D), suggesting that they may drive NSCLC development.

Mutations in the CR genes were associated with high mutation loads, similar to those in DDR genes, in the TCGA cohort (Fig. 1E; Supplementary Fig. S1E) and WFBCCC POI cohort (45). Signature-4 is a COSMIC mutational signature associated with smoking and tobacco mutagens and abundant in NSCLC with a contribution score of 0.69 in LUAD and 0.45 in LUSC based on data from mSignatureDB (44). High contribution scores of Signature-4 significantly correlated with mutations in KMT2C or KMT2C/D combined in LUAD and in all KMT2 combined in both LUAD and LUSC (Fig. 1F). Mutations in KMT2C and KMT2C/D combined also had a trend of association with high contribution scores in LUSC, although not statistically significantly (P = 0.681 and 0.0850, respectively). High mutation loads are associated with response to immune checkpoint blockade (ICB) therapies in NSCLC (37, 46, 47). Consistently, in a cohort of patients with NSCLC treated with ICB (37), KMT2C mutations correlated with positive responses (Fig. 1G).

Correlation between KMT2 mutations and high mutation loads suggests that KMT2 may regulate DDRs. We focused on KMT2C with the highest mutation rate (Fig. 1A–B) and strongest survival association (Fig. 1D) in NSCLC. KMT2C knockdown reduced percentage of cells containing bleomycin- and γ-radiation–induced 53BP1-, BRCA1-, or Rad51-foci in primary BJ human fibroblast and human small airway epithelial (HSAEC) cells (Fig. 2A–C; Supplementary Fig. S2A–S2E), without altering the amount of bleomycin/γ-radiation–induced DNA damages (Supplementary Fig. S2F–S2G), suggesting that reduced focus formation was due to decreased recruitment of the DDR factors. KMT2C shRNAs also abrogated γ-radiation–induced ATM and γH2A.X phosphorylation without altering total protein levels in BJ cells (Supplementary Fig. S2J). Therefore, KMT2C is required for focus assembly at a step before 53BP1, BRCA1, and Rad51 and for DDR signaling.

γ-radiation and bleomycin reduced percentage of BrdUrd-incorporating BJ cells, indicating an S-phase checkpoint, which was attenuated by KMT2C knockdown (Fig. 2D and E; Supplementary Fig. S2H and S2I). Consistently, KMT2C shRNA partially abrogated γ-radiation- and bleomycin-induced proliferative arrest in BJ cells (Fig. 2F).

We investigated the role of KMT2C in 3 major DNA repair pathways, using GFP reporters for HR (33), NHEJ (34), and alt-NHEJ (Supplementary Fig. S3A–S3C; ref. 48). KMT2C shRNAs reduced HR, but not NHEJ or alt-NHEJ, to similar levels as BRCA1 shRNAs, without affecting expression of I-SceI used to induce DSBs (Fig. 2G–I; Supplementary Fig. S3D to S3F, S3L, and S3M). CRISPR/Cas9–mediated KMT2C knockout further reduced HR repair efficiency compared with shRNA-mediated knockdown (Supplementary Fig. S3G and S3H). KMT2A, B, or D knockdown also reduced HR (Supplementary Fig. S3I–S3K). KMT2C shRNAs disrupted 5′ strand resection at HR-repaired DSBs induced by restriction enzyme AsiSI (23, 24, 26) in U2OS cells (Fig. 2J; Supplementary Fig. S3N–S3P). Moreover, although γ-radiation and bleomycin induced similar amount of γH2A.X- and/or NBS1-foci in control and KMT2C shRNA-expressing BJ cells immediately after the treatments, these foci disappeared more slowly in KMT2C knockdown cells than the control cells, upon removal of treatments (Fig. 2K; Supplementary Fig. S4A–S4C), suggesting reduced overall DNA repair by KMT2C knockdown. Indeed, KMT2C shRNAs attenuated γ-radiation–induced DSB repair in BJ cells in Comet assays (Supplementary Fig. S4D).

Therefore, KMT2C is essential for DDR, including focus formation, checkpoint activation, and HR-mediated as well as the overall DNA repair.

We investigated whether KMT2C mediates DDRs by regulating DDR gene expression. KMT2C/D knockdown had no effect on the expression of ATM, 53BP1, BRCA1, NBS1, MDC1, RAD54B, and/or EXO1 in BJ and HSAEC cells and NSCLC cell lines (Supplementary Fig. S5A–S5F). KMT2C and/or D mutations did not correlate with reduced expression of DDR genes including ATM, ATR, BRCA1/2, EXO1, MDC1, NBS1, RAD54B, and 53BP1 in LUAD and LUSC in TCGA (Supplementary Fig. S5G and S5H). Among 61 genes involved in HR-mediated DSB repair (GO:0000724), KMT2C mutations correlated with reduced expression of only ZSWIM7 in LUAD and ERCC4 and RAD50 in LUSC in TCGA, whose expression was not reduced in KMT2C shRNA-expressing BJ and U2OS cells with disrupted DDR and HR (Supplementary Fig. S5I–S5L). Thus, KMT2C/D inactivation/mutations do not disrupt DDR/HR by abrogating DDR/HR gene transcription. Also, high mutation loads correlated with KMT2 mutations (Fig. 1E), but not KMT2C/D expression in either all patients or those with wild-type, KMT2C/D (Supplementary Fig. S5M and S5N), suggesting that mutations, not reduced expression, of KMT2C/D disrupt DDR and DNA repair in NSCLC.

Using an anti-KMT2C antibody with confirmed specificity (Supplementary Fig. S6A and S6B), KMT2C colocalized with γH2A.X-foci in bleomycin-treated BJ and U2OS cells and γ-radiation–treated BJ, HSAEC, and U2OS cells, and on microirradiation-induced damage tracks in U2OS and BJ cells (Fig. 3A–C). KMT2C was also enriched with H3K4me1, H3K4me2, phospho-ATM, ATM, γH2A.X, and 53BP1 on a site-specific DSB induced by I-PpoI on Chromosome 1 in U2OS cells (25), while localization of these protein on an enhancer containing no I-PpoI site on chromosome 1 was independent of I-PpoI induction (Fig. 3D; Supplementary Fig. S6C). KMT2C and H3K4me1 localization to other regions such as enhancers thus does not rely on DNA damage. I-PpoI induced KMT2C and H3K4me1 enrichment in a large region of at least 18 kb surrounding the DSB, and their enrichment as well as the peaks of their enrichment overlapped with those of γH2A.X and HR factors BRCA1 and RAD51, while NHEJ factors Ku86 and XRCC4 were enriched only in the proximal regions of the DSB (Fig. 3E). These proteins were not enriched at a control site without DSB (Supplementary Fig. S6E). These findings are consistent with the involvement of extended chromosomal regions in HR but not NHEJ and our observation that KMT2C is required for HR, but not NHEJ (Fig. 2G and H). These results were confirmed at DSBs induced by AsiSI (Fig. 3F; Supplementary Fig. S6D; refs. 26, 27, 49). KMT2C and H3K4me1 were also enriched with γH2A.X, BRCA1, and RAD51 in an extended, 19-kb chromosomal region surrounding the AsiSI-induced HR sites, and with Ku86 and XRCC4 in the proximal region at the NHEJ sites, but not in control region without DSBs (Fig. 3G; Supplementary Fig. S6F).

Thus, KMT2C is directly recruited to DNA damage sites.

In BJ and HSAEC cells, KMT2C shRNAs reduced sizes of γ-radiation- and bleomycin-induced γH2A.X- and NBS1-foci without altering percentage of cells containing these foci, but reduced both percentage of cells with and sizes of MDC1-foci (Fig. 4A; Supplementary Fig. S7A, S7C, and S7D), suggesting requirement of KMT2C in reported expansion of γH2A.X-foci from initial small sizes to large ones involving megabase domains around DSBs (50). KMT2C shRNAs also reduced percentage of cells with p-ATM-foci (Supplementary Fig. S7B), although the high immunofluorescence background prevented analysis of focus sizes. Thus, KMT2C is likely essential for secondary recruitment.

KMT2C knockdown reduced AsiSI- and I-PpoI–induced enrichment of H3K4me1, MDC1, and BRCA1 in both proximal and distal regions of DSBs, and that of γH2A.X, NBS1, and ATM only in distal, but not proximal regions (Fig. 4B; Supplementary Fig. S8A–S8D). NHEJ factor Ku-86 was only enriched in proximal regions, which was unaffected by KMT2C shRNAs (Supplementary Fig. S8A–S8D). We examined the time course of AsiSI-induced primary and secondary recruitment. Enrichment of KMT2C, γH2A.X, and NBS1 became detectable after 30 minutes and more evident at 60 minutes in proximal regions, and became detectable after 60 minutes and more evident at 120 minutes in distal regions, upon AsiSI induction; KMT2C shRNAs reduced both proximal and distal recruitment of KMT2C, and only distal, but not proximal enrichment of γH2A.X or NBS1 at any time point (Fig. 4C; Supplementary Fig. S8E). MDC1 recruitment was visible in proximal regions between 30 and 60 minutes and distal regions after 60 minutes, which were both reduced by KMT2C shRNAs (Supplementary Fig. S8E). These data demonstrate that KMT2C is essential for secondary, but not primary recruitment of NBS1 and γH2A.X and is essential for MDC1 recruitment that bridges primary and secondary recruitment. We assessed the effect of KMT2C on the time-dependent recruitment of DDR factors to γ-radiation–damaged chromatins using chromatin fractions of BJ cells (Fig. 4D; Supplementary Fig. S8F). Enrichment of NBS1 and p-ATM and that of ATM, γH2A.X, and KMT2C was initially detected at 2 and 10 minutes, respectively, after IR and reached higher levels at 30 minutes, which likely reflected primary and secondary recruitment, respectively; KMT2C shRNAs reduced the enhanced enrichment of NBS1, p-ATM, ATM, and γH2A.X at 30 to 60 minutes, but not their initial enrichment at 2 to 10 minutes. Enrichment of repair factors BRCA1, Ku86, and Ku70 was visible at 10 minutes and became substantial at 30 to 60 minutes. KMT2C knockdown reduced enrichment of BRCA1 but not Ku86 and Ku70. In total lysates, KMT2C shRNAs did not affect ATM and H2A.X protein levels (Supplementary Fig. S2J). Thus, KMT2C is essential for secondary, but not primary recruitment at DNA damage sites.

In an assay measuring chromatin relaxation based on nuclease accessibility, AsiSI induced chromatin relaxation in both proximal and distal regions of DSBs mainly repaired by HR, but not at control site without DSB, which was abrogated by KMT2C shRNAs (Fig. 4E). In NaCl extraction assay detecting chromatin relaxation based on histone release in 1 mol/L NaCl (17, 51), bleomycin-induced increase in γH2A.X, H2A.X, and H3 solubility was abrogated by KMT2C shRNAs in BJ and HSAEC (Fig. 4F; Supplementary Fig. S8G). Thus, KMT2C is required for chromatin relaxation at DNA damage sites.

Our findings therefore demonstrate that KMT2C is recruited to DNA damage sites where it mediates H3K4me1 and chromatin relaxation, allowing secondary recruitment of DDR factors.

The requirement of both KMT2C and DDRNAs for secondary recruitment prompted us examine whether DDRNAs and DDRNA machinery mediate KMT2C recruitment. Ago2 and Dicer shRNAs abrogated γ-radation- and bleomycin-induced KMT2C-foci in BJ cells and reduced sizes of, but not percentage of cells with, γH2A.X- and NBS1-foci (Supplementary Fig. S9A–S9D). Ago2 or Dicer knockdown reduced KMT2C and MDC1 enrichment in both proximal and distal regions of AsiSI-induced DSBs, and proximal, but not distal γH2A.X and NBS1 enrichment (Fig. 5A; Supplementary Fig. S9E and S9F), and abrogated KMT2C and H3K4me1 enrichment at I-PpoI-induced DSBs (Supplementary Fig. S9G and S9H), without affecting control regions without DSBs. Thus, the DDRNA machinery is required for KMT2C recruitment to DNA damages and secondary recruitment.

Treatment of permeabilized U2OS cells with DDRNA-degrading RNaseA reduced AsiSI-induced proximal and distal recruitment of KMT2C and MDC1 and distal, but not proximal NBS1 recruitment (Fig. 5B). Surprisingly, RNaseA did not reduce distal γH2A.X enrichment (Fig. 5B), likely because DDRNA removal by 25-minute RNaseA treatment was sufficient to disrupt KMT2C, NBS1, and MDC1 protein complexes held together by DDRNAs, but not to revert existing covalent modifications including γH2A.X. In rescue assays (16, 17), Ago2 and Dicer shRNA-mediated reduction in proximal and distal enrichment of KMT2C and distal enrichment of NBS1 at HR site1 and HR site 2 of AsiSI-induced DSBs were restored by synthetic DDRNAs derived from HR site 1 and 2, but not by those from HR site 2 or 1, respectively, or control RNAs from undamaged region (Fig. 5C and D). Moreover, RNaseA treatment abrogated KMT2C and H3K4me1 enrichment at I-PpoI–induced DSBs, which was restored by synthetic DDRNAs derived from the region surrounding this DSB, but not by control RNAs (Supplementary Fig. S9I and S9J). Besides site-specific DSBs, recruitment of KMT2C to bleomycin-damaged chromatins also requires DDRNAs, as RNaseA abrogated increased binding of KMT2C, along with Ago2 and other DDR factors such as ATM and NBS1, to damaged chromatins in HSAEC, BJ, and U2OS cells, without affecting total levels of these proteins in whole-cell lysates (Fig. 5E).

Thus, KMT2C recruitment to a DNA damage site requires specific DDRNAs generated from this same site and proteins mediating DDRNA biogenesis and functions.

Based on the role of Ago2 in miRNA functionality (52, 53), we reason that an Ago2-containing complex may bind to DNA damage sites through complementary sequences of DDRNAs and recruit KMT2C. Indeed, Ago2 was recruited to microirradiation-induced DNA damage tracks with KMT2C (Fig. 5F), and was enriched in both proximal and distal regions of AsiSI-induced DSBs (Supplementary Fig. S10A), at I-PpoI–induced DSBs (Supplementary Fig. S9J) and on bleomycin-damaged chromatins in a DDRNA-dependent manner (Fig. 4D; Supplementary Figs. S8F and S5E). These results confirm previously reported Ago2 accumulation at DSBs (54) and further demonstrate that Ago2 is recruited to DSBs via DDRNAs. Moreover, KMT2C shRNAs disrupted AsiSI-induced distal, but not proximal enrichment of Ago2 (Supplementary Fig. S10A). In BJ cells, Ago2 enrichment on damaged chromatins was detectable at 2 minutes after γ-radiation, along with NBS1 and p-ATM, and became more intensive at 30 to 60 minutes; KMT2C shRNAs reduced the enhanced Ago2 enrichment at 30 to 60 minutes, but not at 2 to 10 minutes (Fig. 4D; Supplementary Fig. S8F). These findings, along with abrogation of both distal and proximal recruitment of KMT2C by Ago2 shRNAs and RNaseA (Fig. 5A and B), indicates that although Ago2 is also a part of KMT2C-dependent secondary recruitment, it is required for KMT2C recruitment to proximal regions during primary recruitment.

As KMT2C is dispensable for proximal recruitment of NBS1 and ATM in primary recruitment (Fig. 4B; Supplementary Fig. S8C), we reciprocally tested the effects of NBS1 and ATM knockdown (Supplementary Fig. S10B) on KMT2C recruitment. At an AsiSI-induced HR site, NBS1 shRNAs reduced proximal and distal recruitment of KMT2C, NBS1, ATM, and 53BP1; ATM shRNAs reduced proximal and distal recruitment of KMT2C, ATM, and 53BP1, and only distal recruitment of NBS1; and 53BP1 shRNAs only reduced proximal and distal recruitment of itself but not that of KMT2C, NBS1, or ATM (Supplementary Fig. S10C). Thus, besides Ago2 and DDRNAs, primary recruitment of KMT2C at proximal regions of DSBs requires NBS1 and ATM.

Coimmunoprecipitation assays detected an interaction among Ago2, KMT2C, and NBS1 in a DNA damage–dependent and RNA-dependent manner. Immunoprecipitates of each protein from BJ and U2OS cells treated by bleomycin but not by RNaseA contained the other 2 (Fig. 5G–I; Supplementary Fig. S10D–S10F). NBS1 and MDC1, but not ATM, were also detected in the complex. These findings, together with colocalization of Ago2 and KMT2C at DNA damages (Fig. 4D; Supplementary Figs. S8F, S9J, S5E, and S5F), indicate that upon DNA damage, Ago2 recruited to DNA damages likely through DDRNAs with sequences complimentary to the damaged sites, in turn recruits KMT2C that mediates chromatin relaxation and secondary recruitment of DDR factors.

PARPi prevent repair of SSBs, which become DSBs, inducing cell death in tumors deficient in HR repair (55, 56). PARPi are used to treat patients with breast, prostate, ovarian, and pancreatic cancer with BRCA1/2 mutations, but not in NSCLC with rare BRCA1/2 mutations (Fig. 1A and B). Reduced HR by KMT2C/D inactivation suggests that KMT2C/D mutations may sensitize NSCLC to PARPi. Based on genomic data in Broad Institute Cancer Cell Line Encyclopedia (CCLE), we selected 7 lines (3 LUAD and 4 LUSC) with mutations in KMT2C/D, but not other CR or DDR genes associated with NSCLC (Fig. 1A and B), and 6 lines (3 LUAD and 3 LUSC) without any mutations in those CR or DDR genes (Fig. 6A; Supplementary Fig. S11A). The KMT2C/D-mutant lines were more sensitive to olaparib than the WT lines (Fig. 6A; Supplementary Fig. S11A and S11B). Olaparib inhibited the growth, and often caused regression or disappearance, of xenograft tumors from the mutant, but not the WT LUAD cell lines (Fig. 6B; Supplementary Fig. S11C). γ-Radiation induced less cells with BRCA1- and RAD51-foci and smaller γH2A.X-foci in KMT2C/D-mutant than the WT LUAD lines (Supplementary Fig. S11D), indicating DDR disruption by KMT2C/D mutations, likely at secondary recruitment, in NSCLC cells.

KMT2C/D shRNAs sensitized the WT LUAD lines to olaparib in vitro and the WT H1975 cells in xenografts (Fig. 6C and E; Supplementary Fig. S12A–S12D), as compared with shRNA control. Increased olaparib sensitivity in H1975 xenografts with KMT2C/D shRNAs was accompanied by increased apoptosis over control cells (Fig. 6F; Supplementary Fig. S12E). KMT2C shRNAs reduced percentage of HCC-827 cells with olaparib-induced 53BP1-, BRCA1-, or RAD51-foci and sizes of γH2A.X-foci (Supplementary Fig. S12F), demonstrating that KMT2C is required for olaparib-induced DDRs in NSCLC cells.

We transfected IRES-eGFP–coupled KMT2C into KMT2C-mutant NIH-H2342 and LC-1/sq-SF cells (Fig. 6A; Supplementary Fig. S11A). After sorting, eGFP-positive cells were enriched to 70% to 80%, and KMT2C overexpression was detected (Supplementary Fig. S13A–S13D). KMT2C-eGFP increased olaparib IC₅₀, size of γ-radiation–induced γH2A.X-foci and percentage of cells with BRCA1-foci (Supplementary Fig. S13E–S13H) in both cell lines, indicating that disrupted DDR and PARPi sensitivity are at least partly conferred by KMT2C mutations. KMT2C-mediated rescue in H2342 cells with both KMT2C (NS) and KMT2D (MS) mutations suggests that functions of KMT2C and KMT2D may be redundant and dose dependent. Alternatively, this particular MS KMT2D mutation may be a bystander.

BRCA1 shRNAs sensitized WT H1975 cells to olaparib to the same level as KMT2C shRNAs in vitro and in xenografts (Supplementary Fig. S14A–S14D). An inhibitor of ATM, a key mediator of γH2A.X spread and DNA repair, also sensitized WT NSCLC cell line to PARPi (Supplementary Fig. S14E), and further increased PARPi sensitivity in KMT2C/D-mutant lines likely by disrupting the remaining DNA repair in these cells (Supplementary Fig. S14F). Knockdown of Ago2, Dicer, and Drosha essential for KMT2C recruitment to damages also rendered the WT LUAD cell lines olaparib sensitive (Supplementary Fig. S15A–S15F).

The KMT2C/D-mutant cell lines and the WT lines with KMT2C/D, BRCA1, Ago2, Dicer, or Drosha shRNAs showed increased sensitivity to cisplatin as compared with respective controls (Supplementary Fig. S16A–S16I), consistent with disruption of DNA repair by inactivation of the DDRNA–KMT2C/D pathway or BRCA1. Thus, patients with NSCLC with KMT2C/D mutations are likely to respond better to PARPi, cisplatin, or combinations than those without KMT2C/D mutations.

Among human lung cancer (mostly NSCLC) samples from WFBCCC (45), those with KMT2C/D mutations contained less γH2A.X- or p-p53Ser15-positive tumor cells than the wild-type tumors (Fig. 7A–D; Supplementary Fig. S17A), confirming that KMT2C/D mutations abrogate DDR in patients with NSCLC.

Among a panel of NSCLC PDX models with genomic data (LIDE Biotech), we selected 6 containing mutations in KMT2C/D, but not other CR or DDR genes associated with NSCLC (Fig. 1A and B; Mutant 1–6), and 3 without any mutations in these genes (WT1–3; Supplementary Fig. S17B). Cells from the KMT2C/D-mutant PDX had lower olaparib IC₅₀ than those from the WT PDX in vitro (Fig. 7E). In vivo, olaparib suppressed growth of 2 of the 3 mutant PDXs we selected, accompanied by induction of apoptosis, while all 3 WT PDXs were resistant (Fig. 7F; Supplementary Fig. S17B–S17D). Olaparib-resistant Mutant-3 may contain other mutations that compromise KMT2C/D mutation–conferred PARPi sensitivity. Together, our findings indicate that KMT2C/D mutations are associated with DDR disruption and PARPi sensitivity in patients with NSCLC and are potential biomarkers for responses to therapies with PARPi and/or DNA-damaging agents.

This study identified a novel function of KMT2C in DDR, mediated by a new mechanism different from its transcriptional activity. KMT2C is recruited to DNA damages, where it mediates H3K4me1 and chromatin relaxation, allowing secondary recruitment of DDR factors and spreading of DDR signals essential for DDR (Fig. 7G). KMT2C knockdown greatly reduced the distal enrichment of γH2A.X, NBS1, and ATM, but only slightly reduced their proximal enrichment at time points consistent with secondary recruitment (Fig. 4B and C; Supplementary Fig. S8C–S8E), suggesting that KMT2C plays a more prominent role in distal spreading of DDR signals, likely by opening chromatin structures. Alternatively, enrichment of NBS1, ATM, and γH2AX during primary recruitment may be more extensive than MDC1-dependent secondary recruitment in proximal regions, and thus, the main consequence of secondary recruitment is amplification and distal spreading of DDR signals.

KMT2C did not mediate DDRs by promoting transcription of DDR/HR genes. This finding contradicts a study published during preparation of this manuscript, which suggested that KMT2C is required for DDR gene transcription in bladder cancer cells (57). The difference is likely due to tissue specificity or differential experimental conditions. The same study showed positive correlation between KMT2C and DDR gene expression in NSCLC in TCGA; however, this correlation did not account for the frequent loss-of-function KMT2C/D mutations in NSCLC. We believe that KMT2C/D mutations, not their differential expression, contribute to disrupted DNA repair and accumulation of mutations in NSCLC, because high tumor mutation loads correlated with KMT2C/D mutations, not expression (Fig. 1E; Supplementary Fig. S5M and S5N). Moreover, KMT2C/D mutations were not associated with reduced DDR/HR gene expression (Supplementary Fig. S5G and S5H), and KMT2C shRNAs did not decrease expression of the HR genes with reduced levels in KMT2C/D-mutant tumors (Supplementary Fig. S5I–S5L), suggesting that DDR defects in KMT2C/D-mutant NSCLC are unlikely due to disruption of KMT2C/D-mediated DDR/HR gene transcription. However, we cannot exclude possible contributions of KMT2-mediated transcription of genes with unidentified roles in DDR.

KMT2C knockdown abrogated recruitment of HR factors but not NHEJ factors, disrupted HR- but not alt-NHEJ–mediated repair, and even increased NHEJ repair likely due to compensation for reduced HR (Fig. 2G–I). Increased NHEJ appears to contradict the reduction in 53BP1 recruitment. 53BP1 promotes NHEJ by displacing BRCA1 from DSBs and inhibiting BRCA1-mediated resection and HR (58). Possibly, when BRCA1 recruitment and HR are impaired by KMT2C inactivation, the choice for NHEJ no longer relies on 53BP1. Regardless, increased NHEJ did not fully compensate the reduction in HR, because KMT2C knockdown attenuate overall DNA repair (Fig. 2K; Supplementary Fig. S4).

We identified KMT2C-mediated H3K4me1 as a novel, essential histone modification at DNA damage sites. Besides H3K4me1, H3K4me2 was also enriched at I-PpoI/AsiSi–induced DSBs, suggesting that KMT2A/B may also contribute to DDR. Interestingly, knockdown of each KMT2 disrupted HR (Fig. 2G; Supplementary Fig. S3I–S3K), suggesting that their functions are either nonredundant or dose dependent. Despite a recent report that H3K4me3 levels are decreased at DSBs through KDM5 (59), whether and how H3K4me1, 2, and 3 mediated by different KMT2 contribute to DDR warrants further investigation.

We demonstrated DDRNA-dependent Ago2 recruitment to DNA damages along with KMT2C. Ago2 forms a DNA damage/DDRNA–dependent complex with KMT2C (Fig. 5G–I; Supplementary Fig. S10D–S10F). Together with the requirement of Ago2 and DDRNAs for KMT2C recruitment (Fig. 5A–E, Supplementary Fig. S9A–S9D and S9G–S9J), these findings suggest that KMT2C is recruited to DNA damage sites by Ago2 and DDRNAs (Fig. 7G), although direct Ago2–KMT2C binding remains unknown. NBS1 and MDC1, but not ATM, were detected in the Ago2–KMT2C complex. Although we cannot rule out that KMT2C is recruited by direct binding with both NBS1 and Ago2, it is likely that NBS1 is in the complex to promote Ago2 recruitment, which in turn recruits KMT2C, as Ago2 and DDRNAs are required for secondary but not primary recruitment of NBS1, as demonstrated by us (Fig. 5A–D; Supplementary Fig. S9A–S9E) and others (19). Failure to detect ATM in the complex may be due to insufficient sensitivity of IP-Western assays for weak interactions, suboptimal experimental conditions or presence of ATM in a separate physical complex, although it functionally interacts with KMT2C, Ago2, and NBS1.

Correlation of KMT2 mutations with high mutation loads, reduced DDR, and poor survival in patients with NSCLC (Figs. 1C–E, and 7A–D) suggest that KMT2 mutations may drive NSCLC development by disrupting DDR. Moreover, the newly identified role of KMT2C/D in DDR and the mechanism underlying this role provide the mechanistic basis for the value of KMT2C/D mutations as the much-needed biomarkers for efficacy of PARPi treatments in NSCLC. Clinical trials on PARPi in NSCLC and other cancers with rare BRCA1/2 mutations have shown limited efficacy (60), likely due to lack of BRCA1/2-like biomarkers. The high KMT2C/D mutation rates in NSCLC and sensitization of NSCLC cells and PDX to PARPi by KMT2C/D knockdown and mutations support the need for future clinical trials evaluating BRCA1/2 mutations as biomarkers for therapeutic responses to PARPi in NSCLC.

Despite their large sizes, KMT2C and KMT2D are mutated more frequently in NSCLC (10%–25%) than other cancers including breast cancer (3%–9%). Likewise, the large BRCA1/2 genes are mutated more frequently in triple-negative (>15%) than in other subtypes of breast cancer (61) and NSCLC (4%–8%; Fig. 1A and B). Thus, there are cancer-specific and cancer subtype–specific selections for mutations in these genes. Many other factors besides gene size can affect mutation rates, including CpG contents, gene locus, transposons, and repetitive sequences (62). Although many KMT2C/D mutations in NSCLC are FS or NS, most of them are MS (Fig. 1A and B). Supporting the notion that at least some of these MS mutations are loss-of-function drivers, most DDR-defective NSCLC tissues contained only MS KMT2C/D mutations (Supplementary Figs. S17A and S6A–S6D), and NSCLC cells (NCI-H1869 and EBC-1) and PDX (Mutant-1, 2, 5, and 6) containing only MS KMT2C/D mutations showed PARPi sensitivity as those with FS and NS mutations (Supplementary Figs. S11B, S11C, S7D, S7G, and S17B). Future studies are needed to analyze consequences of these KMT2C/D mutations to determine the extent of loss-of-function of these genes in NSCLC. Rescued DDR defects and reduced PARPi sensitivity by wild-type KMT2C in KMT2C-mutant NSCLC cells (Supplementary Fig. S13) suggest that at least these particular KMT2C mutations are loss-of-function drivers of DDR defects and PARPi sensitivity.

The allele frequencies of KMT2C/D mutations (Fig. 6A; Supplementary Fig. S11A) indicate that a substantial proportion of these PARPi-sensitive NSCLC cell lines contain the mutations. Moreover, at least 2 KMT2C/D-mutant, PARPi-sensitive lines (LC-1/sq-SF and NCI-H1869) have no LOH for any KMT2C/D mutation, suggesting that a single loss-of-function mutation in KMT2C/D is sufficient to impair HR and confer PARPi sensitivity. This, together with KMT2C/D shRNA-mediated sensitization of NSCLC cells to PARPi (Fig. 6C–F; Supplementary Fig. S12A–S12D), suggests that partial loss of KMT2C/D is sufficient for PARPi sensitivity. The exact relationship between PARPi responses and KMT2C/D gene doses requires further investigation.

Although both mutations and knockdown of KMT2C/D conferred PARPi sensitization, effect of knockdown seemed weaker than the mutations (Fig. 6; Supplementary Figs. S11 and S12), likely because shRNAs reduced but did not eliminate KMT2C/D expression. Although shRNA-mediated reduction in KMT2C/D expression was obvious, the exact fold reduction was difficult to measure due to the nonlinear relationship between mRNA and protein levels, semiquantitative nature of Western blotting, and instability of shRNAs during cell culture and experiments. The mutant cell lines contain relatively stable KMT2C/D mutations, including FS and NS mutations highly likely to be loss-of-function. KMT2C/D and BRCA1 shRNAs reduced HR only by approximately 50% likely for the same reasons (Fig. 2G; Supplementary Fig. S3K and S3M), which is consistent with the reported effects of shRNAs for known key HR factors including BRCA1 and CtIP in similar reporter systems (63, 64). CRISPR/Cas9–mediated KMT2C knockout further reduced HR compared with KMT2C shRNAs (Supplementary Fig. S3G and S3H).

Our studies focusing on NSCLC can be expanded to other cancers (e.g., bladder cancer and melanoma) with high KMT2 mutation rates. Although we showed that DDR defects in KMT2C/D-mutant NSCLC cells at least partly contribute to PARPi sensitivity, impact of DDR-independent functions of KMT2C/D on PARPi sensitivity needs further analysis. How the known NSCLC driver mutations affect PARPi sensitivity and whether PARPis benefit patients with these driver mutations also warrant investigations. Studies are also needed to determine the underlying mechanisms for PARPi resistance in KMT2C/D-mutant tumors. Mutant-3 PDX carrying both KMT2C and KMT2A mutations were PARPi resistant (Fig. 7F and G). Because all KMT2 genes were essential for HR, PARPi resistance in Mutant-3 was unlikely due to the KMT2A mutation, but other genetic alterations that compromise PARPi sensitivity. In TCGA, 96 of 150 KMT2C-mutant patients with NSCLC have no other KMT2 mutations (P = 8e−4, binomial test). Therefore, most KMT2C-mutant patients will likely respond to PARPi, even if mutations in other KMT2 genes impair PARPi efficacy.

KMT2 mutations may serve as biomarkers for other NSCLC therapies. High tumor mutation loads, determined by genomic sequencing, are used to predict immunotherapy responses. Further investigation of the correlation of KMT2 mutations with high mutation loads and positive immunotherapy responses (Fig. 1E and G) may identify a set of quick and inexpensive immunotherapy biomarkers. KMT2C/D mutations and knockdown also sensitized NSCLC cells to cisplatin (Supplementary Fig. S16). Thus, KMT2C/D mutations may serve as biomarkers for the efficacy of PARPi therapies, immunotherapies, chemotherapies, and their combinations in NSCLC.

W.J. Petty reports personal fees from Jazz Pharmaceuticals outside the submitted work. W. Zhang reports personal fees from Astellas US LLC and Astellas Pharma US, Inc., FirstThought LLP, Decibio Consulting, and PHASE II International Ltd. outside the submitted work. No disclosures were reported by the other authors.

A. Chang: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–review and editing. L. Liu: Resources, data curation, formal analysis, investigation, writing–review and editing. J.M. Ashby: Data curation, formal analysis. D. Wu: Data curation, formal analysis. Y. Chen: Data curation, formal analysis. S.S. O'Neill: Resources, formal analysis, writing–review and editing. S. Huang: Data curation, formal analysis. J. Wang: Data curation, formal analysis. G. Wang: Data curation, formal analysis. D. Cheng: Data curation, formal analysis, project administration. X. Tan: Data curation, formal analysis. W.J. Petty: Resources, investigation, writing–review and editing. B.C. Pasche: Resources, validation, investigation. R. Xiang: Data curation, supervision, validation, investigation, project administration. W. Zhang: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, writing–review and editing. P. Sun: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, writing–original draft, project administration.

The authors thank Drs. Xiaohua Wu (The Scripps Research Institute), Vera Gorbunova (University of Rochester), Gaëlle Legube (Université de Toulouse, France) for reagents, and Tumor Tissue and Pathology, Biostatistics, Bioinformatics, Cell Engineering, and Cellular Imaging Shared Resources of WFBCCC. This study was supported by NIH/NCI grants CA131231 (P. Sun), CA172115 (P. Sun) and P30CA012197 (P. Sun and W. Zhang), and China National Science Foundation Young Scientists Award 81702994 (A. Chang). P. Sun is supported by the Anderson Oncology Research Professorship. W. Zhang is supported by the Hanes and Willis Family Professorship, a fellowship from the National Foundation for Cancer Research and a Stand Up to Cancer grant (AACR SU2C-Meg Vosburg T-Cell Lymphoma Dream Team Translational Research Grant).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.