Identification of novel strategies to expand the use of PARP inhibitors beyond BRCA deficiency is of great interest in personalized medicine. Here, we investigated the unannotated role of Kub5-HeraRPRD1B (K-H) in homologous recombination (HR) repair and its potential clinical significance in targeted cancer therapy.
Functional characterization of K-H alterations on HR repair of double-strand breaks (DSB) were assessed by targeted gene silencing, plasmid reporter assays, immunofluorescence, and Western blots. Cell survival with PARP inhibitors was evaluated through colony-forming assays and statistically analyzed for correlation with K-H expression in various BRCA1/2 nonmutated breast cancers. Gene expression microarray/qPCR analyses, chromatin immunoprecipitation, and rescue experiments were used to investigate molecular mechanisms of action.
K-H expression loss correlates with rucaparib LD50 values in a panel of BRCA1/2 nonmutated breast cancers. Mechanistically, K-H depletion promotes BRCAness, where extensive upregulation of PARP1 activity was required for the survival of breast cancer cells. PARP inhibition in these cells led to synthetic lethality that was rescued by wild-type K-H reexpression, but not by a mutant K-H (p.R106A) that weakly binds RNAPII. K-H mediates HR by facilitating recruitment of RNAPII to the promoter region of a critical DNA damage response and repair effector, cyclin-dependent kinase 1 (CDK1).
Cancer cells with low K-H expression may have exploitable BRCAness properties that greatly expand the use of PARP inhibitors beyond BRCA mutations. Our results suggest that aberrant K-H alterations may have vital translational implications in cellular responses/survival to DNA damage, carcinogenesis, and personalized medicine.
Improved understanding of novel factors that promote DNA repair and survival in cancer cells is vital toward the development of novel “druggable” targets and innovative strategies for personalized medicine. Here, we demonstrate that Kub5-HeraRPRD1B (K-H) is a novel predictive biomarker of response to PARP inhibitors in a subset of BRCA-proficient breast cancers. Aberrant loss of K-H in specific cancer cells compromises homologous recombination (HR) by concomitantly decreasing CDK1 expression and BRCA1 phosphorylation. The subsequent HR deficiency is compensated by enhanced PARP activity for cancer cell survival. Indeed, ablation of PARP activity in our cell models leads to synthetic lethality, which is reversed by reexpression of the wild-type, but not the mutant form of K-H with reduced RNAPII binding. Collectively, the mechanistic insights gained from our study reveal how K-H functions and whose expression level may be translationally exploited as a prognostic biomarker for targeted therapy.
The lack of tumor specificity and drug resistance remain a significant barrier to effective cancer therapy. In the presence of standard DNA-damaging chemotherapeutic agents, normal cells often depend on a full spectrum of functioning DNA damage checkpoints and repair processes to maintain genomic integrity and survival (1). In contrast, cancer cells frequently develop an increased addiction to specialized processes to evade growth arrest, senescence, and cell death, resulting in genomic instability and progressive genetic aberrations, furthering carcinogenesis and genetic heterogeneity (1). During this process, cancer cells may emerge with aberrant dependencies on specific pathways that represent potential druggable “cancer vulnerable” targets for selective destruction of malignant cells. For example, it is now established that cancer cells can survive the toxic effects of DNA-damaging agents by modulating their DNA repair potential (2). Thus, a comprehensive understanding of the mechanisms of DNA repair and survival in cancer cells following cancer therapy is the key toward the development of novel drugs and innovative strategies in personalized medicine.
Recently, we discovered that Kub5-Hera (K-H, aka, RPRD1B, CREPT, C20orf77; refs. 3, 4), the human homolog of yeast RTT103 transcription termination factor, plays critical roles in transcription and DNA repair via its ability to bind RNAPII and Ku70 in separate complexes (5). We previously showed that K-H depletion caused accumulation of endogenous persistent RNA:DNA hybrids (R-loops) during transcription, resulting in deleterious DNA double-strand breaks (DSB) potentially due to replication fork collision (3, 6) or involvement of transcription-coupled nucleotide excision repair (NER; ref. 7). K-H loss resulted in elevated persistent R-loops and DSBs that were not efficiently repaired due to the simultaneous role of K-H in stabilizing Artemis (4). Further studies showed that K-H depletion created a heightened state of genetic instability where caspase-dependent MLH1-PMS2 heterodimer degradation generated an indirect DNA mismatch repair–deficient phenotype (5). Moreover, evidence is emerging that K-H plays simultaneous roles in resolving persistent R-loops as well as playing essential roles in DNA repair (3, 5). In fact, K-H depletion in normal cells results in mutational and chromosomal instabilities (3, 5). Indeed, the role(s) of persistent R-loops caused by depletion of specific transcription termination factors (e.g., K-H) have emerged as newly discovered, potent carcinogenic DNA lesions that have yet to be fully elucidated.
Unexpectedly, however, we recently noted that K-H heterozygous (K-H+/−) mice were haploinsufficient, with females developing a repertoire of cancers commonly associated with a “BRCA-deficiency,” including mammary, ovarian, and cervical cancers, but only after 1 Gy ionizing radiation (IR) exposure (8). K-H knockout (K-H−/−) mice were embryonic lethal (4), similar to mice with deleted homologous recombination (HR) factors (e.g., BRCA1−/−, RAD51−/−, or CDK1−/−; refs. 9–11). These observations suggested a potential role of K-H in regulating HR, and thus K-H–deficient cells would have a hypersensitivity to PARP inhibitors.
PARP1 is a major DNA damage and repair (DDR) sensor that detects DNA lesions arising from endogenous or exogenous stresses and stimulates DNA repair (12). Suppressing PARP1 activity can cause elevated DNA single-strand breaks (SSB) that initiate replication fork collapse and formation of DSBs after collision with DNA replication machineries (13). DSBs created during DNA synthesis are preferentially repaired by HR to maintain genomic integrity (14). BRCA1 (breast cancer associated gene 1) is a major HR component whose phosphorylation by specific kinases [e.g., cyclin-dependent kinase 1 (CDK1)] allow its recruitment to DNA damage sites to facilitate simultaneous repair and checkpoint activation (15).
Failure to properly activate HR can result in a profound addiction to PARP1-dependent alternative nonhomolgous end joining (NHEJ) for repair of otherwise lethal complex DSBs (16). Deficiencies in BRCA1/2 genes, for example, impair HR-mediated repair of DSBs, conferring PARP inhibitor hypersensitivity by synthetic lethality (16). The rarity of adult tumors with BRCA deficiencies, however, currently restricts the therapeutic utility of PARP inhibitor monotherapy (17). This is further complicated by several reports of potential molecular mechanisms of resistance to PARP inhibition (PARP1i) in BRCA-mutant cancers (18). Interestingly, approximately 24% of high-grade triple-negative breast cancers (TNBC) without BRCA mutations positively responded to PARP1 inhibitors in phase II clinical trials (17). Thus, the search is on for suitable predictive markers and novel therapeutic targets that can potentiate the DNA-damaging effects of PARP inhibitors beyond BRCA deficiencies.
Here, we report the mechanistic basis for exploiting a subset of BRCA-proficient human breast tumors with K-H copy number and expression losses as biomarkers for hypersensitivity to PARP inhibitors. Our studies revealed K-H as a novel regulator of CDK1 expression and mediator of HR in certain BRCA-proficient breast cancers, including a triple-negative breast cancer model, which generally shows the worst prognosis among all subgroups of patients with breast cancer (17). Mechanistically, K-H loss promotes a “BRCAness” phenotype in BRCA-proficient cancer cells by compromising the transcription and activity of CDK1, a crucial effector of HR that phosphorylates BRCA1 (pS1497) for efficient recruitment of RAD51 (15). Cells deficient in K-H displayed a similar HR-deficient (HRD) gene signature as noted in cells with loss of critical HR factors (e.g., BRCA1, RAD51, or BRIT1; ref. 19). K-H is enriched at the CDK1 promoter to efficiently facilitate CDK1 transcription by interacting with the C-terminal domain of RNAPII (RNAPII-CTD). Our data suggest that K-H may be a potential therapeutic target or a predictive biomarker for synthetic lethality with PARP inhibitors in BRCA1/2 nonmutated breast as well as other cancers.
Materials and Methods
Cell lines and reagents
Human breast cancer cell lines were obtained from the ATCC and from the cell repository of the Hamon Center for Therapeutic Oncology Research at University of Texas Southwestern Medical Center (Dallas, TX). Cell lines were Mouse Antibody Production (MAP) tested and fingerprinted (20). MDA-MB-231 (231) cells knocked down for K-H or PARP1 expression were generated (3, 20, 21) using specific lentiviral shRNAs and maintained in RPMI media supplemented with 5%–10% FBS at 37°C in a humidified CO2 (5%) incubator. AG014361, AG014699 (rucaparib), and RO3066 (CDK1 inhibitor) were purchased from Selleck Chemicals at the highest purity. All cell lines were routinely checked and found free of Mycoplasma infection. Western blots, transfections, immunofluorescence, coimmunoprecipitations, terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL), and colony-forming assays (CFA) were performed using standard protocols. Specific experimental details are included in the Supplementary Information.
PARP activity assays
PARP activities were measured using immunoblot assays in vitro that monitored the formation of PAR levels per normalized total cellular protein lysate. Dot densities were quantified by NIH ImageJ and plotted using GraphPad PRISM after normalization to controls (no NAD+ addition).
HR activity assays
HR activity in DR-GFP–integrated cells was assessed by flow cytometry as described previously (22). GFP expression was quantified by flow cytometry after 72 hours using a BD LSRFortessa II Cell Analyzer. HR activities were normalized to levels found in siSCR-treated control cells.
Array-based comparative genomic hybridization analyses
Gene expression analyses
For qRT-PCR analyses of gene expression, total cellular RNA was collected using the RNAeasy Kit (Qiagen) according to the manufacturer's instructions. RNA quality and concentration were determined using a nanodrop reader. Total RNA (2 μg) was converted to cDNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems) following the manufacturer's instructions. Gene expression analyses were performed on a ABI-HT7900 (Applied Biosystems) real-time qPCR instrument. Multiplex reactions were performed using TaqMan Gene Expression Master Mix (Applied Biosystems) containing either RPRD1B (Hs01082395_m1), RPRD1B (Hs00221889_m1), or CDK1 (Hs00938777_m1) FAM-labeled TaqMan probes (Applied Biosystems), as well as an endogenous, normalization control VIC-GAPDH probe (Applied Biosystems; 4326317E-1006036). Triplicate results from multiple experiments were analyzed using SDS 2.3 software (Applied Biosystems). All samples were normalized to GAPDH expression.
Chromatin extracts were prepared as described previously (24) prior to chromatin immunoprecipitation (ChIP) using a standard procedure from Millipore ChIP Assay Kit. qPCR analyses were performed in triplicate, and values were normalized to each input and expressed relative to IgG control. Primer sequences are listed in Supplementary Table S5.
Luciferase reporter assays
Luciferase activities were assessed with the Dual-Glo Luciferase Assay System (Promega) with slight modifications from the manufacturer's experimental protocol using a Perkin Elmer Victor X3 Multimode Plate Reader. Firefly luciferase activities were normalized to Renilla luciferase levels as internal controls to compensate for variability in transfection efficiencies. Plasmid pWTCDK1-LUC containing 3 kb of wild-type human CDK1 promoters controlling the luciferase reporter gene were prepared as described previously (25). Specific experimental details are included in Supplementary Information. Reported promoter activities were relative to the siSCR or empty vector–treated controls.
Gene (mRNA) signature studies
After K-H knockdown in 231 cells, significantly altered mRNAs (P ≤ 0.01, with ≥1.5-fold change in gene expression relative to control; Supplementary Table S1) from microarray data were used for down streaming Ingenuity Pathway Analysis (http://www.ingenuity.com). We performed core analysis using the Ingenuity knowledge base as the reference set. Fisher exact test identified top molecular and cellular functions that were significantly enriched among these K-H–deficient signature genes.
Statistical analyses were conducted using the GraphPad PRISM program. Two-sided Student t tests or ANOVA were used when appropriate. Results were expressed as mean ± SEM. P ≤ 0.05 was considered statistically significant.
K-H depletion compromises HR-mediated DSB repair
The formation of BRCA-associated mammary, ovarian, and cervical cancers in mice bearing one K-H allele loss after IR treatment (1 Gy; ref. 8) implies a potentially unannotated role for K-H in mediating homologous recombination (HR). Because K-H associates with RNAPII, we first sought to determine whether overall mRNA expression in cells depleted for K-H using siRNA knockdown would resemble the transcriptome profiles obtained for cells with BRCA1, RAD51, or BRIT1 deficiencies, collectively termed the “HRD” signature (19). Ingenuity Pathway Analyses (IPA) of altered mRNA expression in BRCA-proficient MDA-MB-231 (hereafter 231) cells depleted for K-H expression revealed enrichment in top five molecular and cellular pathways (Fig. 1A) analogous to the top signature pathways found for HRD in cancer cells (19). Approximately 50 HRD genes (19) overlap with altered genes found in K-H–depleted 231 cells (Supplementary Table S1A–S1D). This suggests that K-H deficiency may be regarded as one of the HRD gene signatures. Consistent with the theory that K-H depletion affects HR function, we noted that in 231 cells depleted for K-H expression, the gene expression of CDK1, a major cell cycle and DNA damage response and repair (DDR) effector, was concurrently repressed (>2-fold, Supplementary Table S1A). This was intriguing because CDK1 mRNA/protein expression remains putatively constant in cells (26, 27). Loss of CDK1 expression could lead to downstream impairment of HR-mediated repair of DSBs by compromising BRCA1 phosphorylation for efficient function (28). By immunoblot analyses, we observed a concomitant reduction in CDK1 protein expression following K-H knockdown using four different siRNAs against K-H (Supplementary Fig. S1A). Further analysis using siK-H#4 showed decreased CDK1, but not CDK2, protein levels in several cell lines (Fig. 1B–D; Supplementary Fig. S1B–S1D) following K-H knockdown, consistent with our microarray data. We also noted a slight reduction in cyclin E expression with dose-response K-H knockdown (Fig. 1B), which is consistent with prior reports (4). Interestingly, CDK1 and K-H protein levels positively correlated with each other in a dose- and time-dependent manner as noted in transient siK-H knockdowns for up to 72 hours (Fig. 1B; Supplementary Fig. S1B and S1C). Importantly, CDK1 regulation as a result of K-H knockdown was specific, because siRNA depletion of PARP1 or XRN2 (an RNA transcription termination factor known to associate with K-H; ref. 3) showed no apparent CDK1 protein reduction (Supplementary Fig. S1B and S1E); importantly, knockdown of p15RS, a close homolog and known binding factor of K-H (29), also did not affect steady-state protein levels of CDK1 or K-H (Supplementary Fig. S1E). A transient S–G2–M cell-cycle defect was also noted in shK-H–depleted asynchronous cells (Supplementary Fig. S1F and S1G) consistent with K-H as one of the top hits in prior genome-wide RNAi screening of cell division in HeLa cells (30). A long-term effect on basal cell-cycle distribution in K-H–depleted cells was not noted. This is likely due to the ability of CDK2 to readily compensate for loss of CDK1 expression in controlling this cell-cycle checkpoint (28). Nevertheless, it is now known that CDK1 plays more exclusive functional roles in transcription (31) and HR-mediated DSB repair by phosphorylating BRCA1 for proper function (28). Indeed, we noted impairment of activated BRCA1 phosphorylation (pS1497) in siK-H–depleted 231 cells, despite having higher apparent basal levels of DSBs (as indicated by increased surrogate marker, γH2AX) compared with untreated or siSCR-treated controls (Fig. 1B; Supplementary Fig. S1B). Reduced pS1497 BRCA1 formation was also observed in K-H–depleted cells after exposure to rucaparib or IR (Supplementary Fig. S1H–S1J). Furthermore, exogenous insults did not cause CDK1 protein reduction (Supplementary Fig. S1H–S1J). Interestingly, p53-independent activation of p21 and phosphorylation of H2AX (pSer139) were also noted only in K-H–depleted cells, even in the absence of exogenous insult (Fig. 1B; Supplementary Fig. S1B).
Downregulation of CDK1 protein levels in cells depleted for K-H expression prompted us to further investigate a possible downstream impairment of HR-dependent DSB repair. Using the DR-GFP HR repair reporter system (32), we found that depletion of K-H reduced the rate of HR, comparable with known effects in cells depleted for RAD51 or CDK1 using siRNA knockdown (Fig. 1E; ref. 28). Importantly, reexpression of CDK1 in K-H–deficient cells rescued the HR defect. Analogous results were found in MCF-7 breast cancer cells, where reexpression of siRNA-resistant K-H cDNA in siK-H–depleted cells also partially rescued HR capacity (Supplementary Fig. S1K), suggesting that the functional defect in HR was specific in K-H–depleted cells, and was not an off-target effect. Moreover, we found that knockdown of 53BP1 in K-H–depleted cells could rescue HR deficiency (Supplementary Fig. S1L), which is a phenotype consistent with BRCA-deficient cells (33). Collectively, these results further validate a role for K-H in mediating HR.
K-H promotes RNAPII occupancy to the CDK1 promoter
Because K-H depletion resulted in a concomitant reduction in CDK1 protein level in various cell lines (Fig. 1B–D; Supplementary Fig. S1B–S1D), we further examined whether the mechanism underlying K-H regulation of CDK1 occurred at the transcriptional level. We measured CDK1 mRNA expression in 231 (Fig. 2A) and HCC1569 (Fig. 2B) cells after transient knockdown of K-H via siRNA directed to its 3′-UTR. CDK1 mRNA levels were reduced, suggesting involvement of K-H in CDK1 gene transcription. Similar results were found in shK-H 231 cells, where knockdown was achieved by shRNA specific to the K-H coding region (Fig. 2C). Prior studies suggested that K-H interacted with RNAPII in vitro (4, 34), but the mechanistic basis for this interaction has yet to be firmly established. We, therefore, examined the ability of K-H to mediate localization of RNAPII to a specific promoter region of CDK1 compared with the actin promoter region using ChIP assays. Depletion of K-H reduced RNAPII occupancy at the CDK1 promoter region (Fig. 2D), whereas RNAPII localization at the actin promoter was virtually unaffected. These results suggested that K-H may not universally affect gene transcription, but promotes transcription of specific genes (e.g., CDK1), possibly by enhancing the recruitment of RNAPII to specific gene promoter regions.
K-H associates with RNAPII to promote CDK1 transactivation
To examine the functional role of K-H binding to RNAPII in regulating CDK1 transcription, we initially used homology modeling (35) from crystallized RTT103 (the yeast homolog of K-H) to predict crucial K-H and RNAPII residues essential for their interaction. These analyses revealed that Arginine 106 (R106) could be a critical K-H residue with specific and strong affinity for the phospho-Ser 2 (pS2)–containing C-terminal domain (CTD) of RNAPII via hydrogen bonding and electrostatic interactions (Fig. 3A). Using coimmunoprecipitation (co-IP) experiments, we confirmed that endogenous K-H (Fig. 3A) or myc-tagged wild-type K-H pulled down endogenous RNAPII (Fig. 3B). In contrast, mutating the R106 residue of K-H to alanine (p.R106A) abrogated RNAPII co-IP (Fig. 3C), likely due to loss of crucial interactions. Rescue experiments using K-H–depleted 231 cells showed that reexpression of wild-type K-H cDNA partially rescued CDK1 protein expression and BRCA1 phosphorylation (Fig. 3D; Supplementary Fig. S2A). In contrast, reexpression of mutant K-H R106A cDNA or empty pCMV vector failed to rescue CDK1 protein loss and BRCA1 activation (Fig. 3D; Supplementary Fig. S2A).
We then examined the mechanism for how K-H, and its interaction with RNAPII, may regulate CDK1 transcription using a 3-kb portion of the human CDK1 gene promoter driving firefly luciferase, with CMV-Renilla expression used to control for transfection variations (25, 36). CDK1 promoter activity was significantly reduced after siK-H 3′-UTR knockdown versus siSCR treatment in 231 cells (Fig. 3E), consistent with our quantitative mRNA data (Fig. 2A–C). Importantly, overexpression of mutant R106A K-H cDNA in parental 231 failed to stimulate CDK1 promoter activity (relative to the overexpression of empty vector control, P > 0.05), while overexpression of wild-type K-H cDNA enhanced CDK1 promoter activity (∼3-fold). Similar experiments using HCC1569 cells (Fig. 3F) yielded comparable results suggesting that K-H binding to pSer2-CTD of RNAPII was crucial to efficiently promote CDK1 promoter activity and overall transcription. Indeed, ChIP analyses further showed that K-H was enriched at the promoter region of CDK1 likely to facilitate efficient transcription by RNAPII (Supplementary Fig. S2B).
K-H loss heightens PARP1 activity leading to PARP1 inhibitor synthetic lethality
A state of “BRCAness” induced by stable shK-H depletion may render cells hypersensitive to PARP1 inhibition, because synthetic lethality may result from two dysfunctional DNA repair mechanisms (37), for example, a phenotype analogous to BRCA-mutant cells that are hypersensitive to PARP inhibitors (16). To test this, we generated stable K-H or PARP1 shRNA knockdown clones of BRCA-proficient 231 TNBC cells (Supplementary Fig. S3A). We first determined whether depletion of K-H can stimulate PARP1 activity in a manner similar to BRCA-deficient cancers, which underlies their hypersensitivity to PARP inhibitors (38). Indeed, a significant elevation of PARP1 activity was observed in stable shK-H–depleted 231 cells (∼3-fold higher compared with shSCR, Fig. 4A; Supplementary Fig. S3B). In contrast, PARP1 activity was attenuated (∼3-fold) in stable shPARP1 versus shScr cells (Fig. 4A). Elevated PARP1 activity may minimize the toxic effects of intrinsic DSBs formed in shK-H cancer cells by driving PARP1-mediated alt-NHEJ repair for survival (38).
We then tested whether pharmacologic ablation of PARP activity induced synthetic lethality in BRCA-proficient, shK-H–deficient 231 cells. Using CFAs, we found that shK-H cells were hypersensitive to the PARP1 inhibitor, AG014361, compared with resistant parental or shSCR 231 cells (Supplementary Fig. S3C and S3D). Importantly, treatment of PARP1-deficient (shPARP1) 231 cells with AG014361 did not induce additional lethality attributed by the lack of target specificity (i.e., other than PARP1), as shown by the dose–response curve coinciding with parental or shSCR cells (Supplementary Fig. S3D).
We then evaluated the effects of rucaparib/AG014699 (Fig. 4B), a newer and more potent version of AG014361 currently in clinical trials (39). Treatment of shK-H 231 cells with rucaparib showed approximately 6-fold greater lethality than AG014361 exposures (LD50 = 0.66 vs. 3.7 μmol/L, respectively; Fig. 4B; Supplementary Fig. S3D). This increase in sensitivity was consistent with prior preclinical results indicating rucaparib was more potent than AG014361 in killing BRCA-deficient breast cancers. In contrast, comparable LD50 values (∼20 μmol/L) were obtained in parental, shSCR, or shPARP1 231 cells for either PARP1 inhibitor (Fig. 4B; Supplementary Fig. S3D). Further analyses of cell death in rucaparib-treated shK-H 231 cells showed a dose-dependent increase in apoptosis (Fig. 4C), confirmed by caspase-3 activation (cleavage) with staurosporine-exposed 231 cells used as a positive control (Supplementary Fig. S3E). For comparison, we confirmed that the lethality of rucaparib-treated BRCA-deficient HCC1937 was abolished after reconstitution with BRCA1 cDNA (Supplementary Fig. S3F–S3K). Importantly, depletion of K-H expression using siK-H or shK-H treatment of BRCA1-corrected HCC1937 cells induced lethality after long-term exposure to rucaparib, comparable with genetically matched BRCA1-deficient cells expressing normal K-H protein levels via CFA (Supplementary Fig. S3J and S3K). Overall, our results suggest a synthetic lethal relationship exists between PARP1 inhibition and K-H loss in BRCA-proficient breast cancer cells.
We next examined whether genetic suppression of PARP1 activity in shK-H–depleted cells might have similar synthetic lethal effects as PARP inhibitors (vide supra) for target validation. Indeed, dual depletion of PARP1 via siRNA-mediated knockdown (siPARP1) in shK-H 231 cells (Fig. 4D) or depletion of K-H using transient siK-H expression in stable shPARP1 231 knockdown cells reduced cell survival (Fig. 4E). Knocking down K-H in shPARP1 231 cells (Fig. 4F) induced apoptosis in a time-dependent manner (Fig. 4G). In contrast, siK-H or siPARP1 alone showed little or no lethality relative to untreated control cells (Fig. 4D-G). Collectively, these findings confirm that K-H deficiency in cancer cells could provide a vulnerable dependency on PARP activity, rendering K-H–depleted cells synthetic lethal upon PARP inhibition.
Sensitivity of K-H–depleted cells to PARP inhibitors is mediated by persistent DSBs
To elucidate the mechanism by which K-H depletion sensitizes cells to PARP inhibition, we treated shSCR versus shK-H 231 cells with vehicle (0.1% DMSO) or rucaparib (5 μmol/L) for 24 hours and quantified DSB formation at various times posttreatment using γH2AX and/or 53BP1 foci/nucleus formation as surrogate DSB markers (refs. 40–42; representative images, Supplementary Fig. S4A and S4B). Indeed, the levels of colocalized γH2AX and 53BP1 foci (Fig. 5A), indicative of persistent DSBs, significantly increased after rucaparib treatment, as exposed shK-H–depleted cells try to amplify the signal to recruit essential repair factors to DSB sites for survival (43). In contrast, shSCR 231 cells with wild-type K-H expression were resistant to PARP inhibitor treatment, and exposed cells contained significantly lower levels of DSBs after rucaparib exposure, as the rate of DNA repair was not grossly compromised. We further confirmed that K-H depletion compromised RAD51 foci formation after rucaparib treatment (Fig. 5B and C), consistent with impaired downstream HR function caused by loss of CDK1 that phosphorylates and activates BRCA1 (e.g., S1497) for efficient RAD51 recruitment. Moreover, overexpression of wild-type K-H, but not mutant K-H (p.R106A), in K-H–depleted cells rescued DSB formation/repair (Fig. 5D and E; Supplementary Fig. S3D and S3E) and the G2–M cell-cycle defect (Supplementary Fig. S4D–S4G) associated with K-H loss. Thus, hypersensitivity of shK-H–deficient cells to PARP inhibition may be due to the creation and inefficient repair of toxic DSBs generated by persistent R-loops resulting from K-H loss (3).
We then explored whether hypersensitivity to a PARP inhibitor (rucaparib) could be rescued by reexpression of siRNA-resistant wild-type K-H, CDK1, or mutant R106A K-H cDNAs in stable shK-H–depleted 231 BRCA-proficient cells (Fig. 5F). Indeed, depletion of K-H resulted in rucaparib hypersensitivity (Fig. 5F) compared with shSCR control. Reexpression of wild-type K-H in these cells rescued the lethality of rucaparib-treated shK-H 231 cells. Interestingly, reexpression of CDK1 cDNA in shK-H–deficient 231 cells also rescued lethality, suggesting that the ability of K-H to promote CDK1 expression was crucial for resistance to PARP inhibition. In contrast, reexpression of siRNA-resistant mutant K-H (p.R106A) cDNA, which ineffectively rescued CDK1 promoter activity (Fig. 3E and F) and protein expression (Fig. 3D), failed to rescue the lethality noted in rucaparib-exposed shK-H knockdown cells. Reexpression of empty vector in shK-H cells did not alter synthetic lethality caused by rucaparib exposure. These findings suggested that K-H may serve as a novel functional biomarker to identify BRCA-proficient tumors that are hypersensitive to PARP1 inhibitors.
K-H as potential predictive biomarker for PARP1 inhibitor synthetic lethality in BRCA-proficient breast cancers
We analyzed K-H CNAs in a panel of breast cancer cell lines (Fig. 6A) by aCGH (23). Approximately 55% of BRCA-proficient cancer cell lines showed variable gains in K-H copy number, whereas approximately 19% show variable K-H copy number loss. These results closely mimic K-H CNA and mRNA expression data obtained from The Cancer Genome Atlas gathered from patients with invasive breast cancer (Supplementary Fig. S5A). We noted that BRCA mutations exclusively occur whether K-H is amplified or deleted in breast cancer patient samples (Supplementary Fig. S5A). Further analysis of K-H mRNA levels in these cell lines showed significant correlation with their corresponding aCGH values (P = 0.0089; Fig. 6B) and K-H protein levels (P = 0.006; Fig. 6C). Again, we observed the same correlation in patients with breast cancer (Supplementary Fig. S5B and S5C). Importantly, K-H protein levels generally correlated with CDK1 expression by immunoblot analyses (Fig. 6D) and TCGA patient data (Supplementary Fig. S5D), consistent with K-H regulation of CDK1 expression. A subset of BRCA-proficient cancer cells with varying K-H/CDK1 levels were then treated with various rucaparib doses and survival was assessed by CFAs (Supplementary Table S2). Cells with lower K-H/CDK1 levels were significantly more hypersensitive to rucaparib (lower LC50 values) compared with cells with higher K-H/CDK1 levels. In fact, a significant positive correlation was obtained when examining rucaparib LC50 values versus K-H (P = 0.008; Fig. 6E) or CDK1 (P = 0.038; Fig. 6F) protein levels. These results suggest that endogenous K-H loss in certain BRCA-proficient cancers is accompanied by concomitant loss of CDK1 protein expression, resulting in inefficient BRCA1 phosphorylation and HR function. Monitoring K-H and CDK1 levels in patient tumors may, therefore, be a functional biomarker for PARP inhibitor synthetic lethal responses in otherwise BRCA-proficient tumors.
The search to identify novel molecular mechanisms of impaired HR in BRCA1/2 nonmutated breast cancers has been an area of great interest to expand the use of PARP inhibitors beyond treating cancers with BRCA deficiencies. Here, our study sheds new insight into the molecular mechanism by which K-H facilitates HR-mediated DSB repair by regulating the expression of CDK1—a crucial factor involved in a myriad of diverse biological processes ranging from cell-cycle progression to activation of HR factors during DSB repair (15, 44). Our mechanistic studies suggest that K-H is potentially involved in a putative complex residing at the promoter region of CDK1 to efficiently promote transcriptional activation with RNAPII. We are further exploring this complex for a comprehensive understanding of how CDK1 is regulated by K-H. Nevertheless, we noted a direct correlation between K-H and CDK1 protein levels in a panel of breast cancer cell lines (Fig. 6D) and patient samples (Supplementary Fig. S5D). CDK1 gene transcription responded synchronously to variations in K-H protein expression. For instance, aberrant K-H overexpression greatly enhanced CDK1 transcription, whereas K-H knockdown in 231 or HCC1569 breast cancer cell lines concomitantly reduced CDK1 transcription and protein expression (Figs. 1B–D and 2). Mechanistically, our study suggests that K-H may play a major role in enhancing CDK1 promoter activity by binding to RNAPII (Fig. 3A–F) for recruitment at the CDK1 promoter. On the basis of a recent structural study (29) and the results of our studies here, we speculate that K-H predominantly forms a homodimer, especially in K-H–overexpressing cancer cells. Accordingly, this dimer may bind two nearby phosphorylated sites in RNAPII CTD heptapeptide repeats to provide a docking scaffold for other factors, such as RPAP2 (RNA Polymerase II Associated Protein 2; ref. 29). This interaction could then promote efficient dephosphorylation of Ser-5 (pS5) in the RNAPII CTD tail to shift from initiation to elongation of CDK1 transcription by RNAPII. Indeed, several reports suggest that some pS2 sites in the RNAPII CTD must be present before removal of pS5 during transcription (29, 45). Consistent with this, we found that a K-H mutant (p.R106A) that weakly binds RNAPII pS2-CTD (Fig. 3C) failed to rescue the DSBs and PARP inhibitor synthetic lethality in K-H–depleted 231 cells (Fig. 5A–F; Supplementary Fig. S4), potentially due to a compromised BRCA1 phosphorylation.
One of the enabling mechanisms of tumorigenesis is genomic instability. Our studies here expand beyond the implicated roles of K-H in transcription termination predicted by its yeast homolog, RTT103 (3). We previously showed that loss of K-H resulted in increased formation of persistent R-loops (3), where upon collision with the replisome or transcription-coupled nucleotide excision factors could spontaneously collapse to R-loop–driven DSBs (7, 46). A recent study suggested the critical involvement of BRCA1 in a DNA repair mechanism that resolves R-loop–associated genomic instability (6). Thus, the combination of compromised BRCA1 phosphorylation and R-loop formation due to faulty transcription termination caused by aberrant K-H deficiency could explain increases in persistent R-loops (3) and DSBs (Fig. 1B; Supplementary Fig. S1B). Indeed, DSBs formed as a result of replication fork collapse were predominantly repaired by HR (47). Here, we showed that HR was defective in K-H–depleted cells, leading to elevated PARP1 enzymatic activity required for survival presumably via highly error-prone alt-NHEJ repair of DSBs. Indeed, the synthetic lethality caused by ablating PARP1 activity (by genetic silencing or use of PARP inhibitors, e.g., rucaparib) is consistent with a defect in HR and potential dependency of K-H–depleted cells on PARP1-dependent alt-NHEJ.
A notable translational implication gleaned from our study is the potential for K-H expression as a predictive functional biomarker for genomic instability due to “BRCAness” (Fig. 1; Supplementary Figs. S1 and S5E) and for overall PARP inhibitor sensitivity (Fig. 6C) in certain BRCA-proficient cancers. Examination of breast cancer TCGA data (http://cancergenome.nih.gov/) suggest that a subset (∼18%) of BRCA-proficient cancer patients with aberrant K-H loss may benefit from PARP inhibitor–based therapies. On the basis of our studies using a panel of breast cancer cell lines, K-HLow cells showed synthetic lethality to PARP inhibitors due to an HR deficiency resulting from the concomitant loss of CDK1 expression and activity, which is absolutely required to phosphorylate BRCA1 (15) and other HR factors (48) for proper DSB repair (Supplementary Fig. S6). Indeed, PARP inhibitor lethality directly correlated with K-H expression in several BRCA-proficient breast cancer cell lines (Fig. 6E). Importantly, breast cancer cells that overexpressed K-H were extremely resistant to PARP inhibitors, while K-HLow breast cancer cell lines were hypersensitive to PARP inhibitors due to persistent DSBs created by loss of HR function and synthetic lethal effects of inhibition of PARP1-mediated alt-DSB repair, consistent with the elevated PARP1 enzymatic activities present in cells depleted of K-H expression (Fig. 4A). Our prior studies showed that Artemis expression was also lost due to K-H's ability to mediate protein stability of this complex DSB repair protein (3). Because overexpression of CDK1 partially corrected PARP1 inhibition–related synthetic lethality in K-H–depleted cells, these data suggest that Artemis is dispensable for synthetic lethality with PARP inhibitors (e.g., rucaparib). This observation is corroborated by synthetic lethal screening studies in cancer cell lines suggesting that genetic loss of Artemis showed no hypersensitivity to PARP inhibition (49). Thus, our studies provide “proof-of-principle” suggesting that aberrant K-H deficiency in BRCA-proficient breast cancers may be used as a clinically relevant functional biomarker to predict synthetic lethality with PARP inhibitor–based personalized therapy in breast, as well as other human cancers. Future studies will determine whether expression levels of K-H that mediate HR repair through regulation of CDK1 can serve as determinants of therapeutic strategy and of clinical outcome for patients with BRCA-proficient tumors. Furthermore, HR loss in K-H–depleted cells may further contribute to its mutator phenotype, where slow-phase, Artemis-directed c-NHEJ and DNA mismatch repair defects were previously noted (3, 5). Overall, the mechanistic and biological insights on the aberrant function(s) of K-H alterations/mutations in certain cancers should aid (i) the prediction of the functional impact of somatic K-H mutations that arise in the clinic and (ii) the development of innovative strategies and novel chemical agents that target specific K-H functions to further expand the use of PARP inhibitors or selectively potentiate other DNA-damaging agents (e.g., IR).
Disclosure of Potential Conflicts of Interest
P. Patidar has ownership interest (including stock, patents, etc.) in Kub5/hera as a determinant of sensitivity to DNA damage (coinventor). No potential conflicts of interest were disclosed by the other authors.
Conception and design: E.A. Motea, F.J. Fattah, L. Xiao, A. Rommel, J.C. Morales, P. Patidar, D.A. Boothman
Development of methodology: E.A. Motea, F.J. Fattah, L. Xiao, A. Rommel, P. Patidar, Y. Xie, J.D. Minna, D.A. Boothman
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.A. Motea, F.J. Fattah, L. Xiao, A. Rommel, P. Patidar
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.A. Motea, F.J. Fattah, L. Xiao, L. Girard, A. Rommel, J.C. Morales, P. Patidar, Y. Zhou, Y. Xie, D.A. Boothman
Writing, review, and/or revision of the manuscript: E.A. Motea, F.J. Fattah, J.C. Morales, P. Patidar, A. Porter, J.D. Minna, D.A. Boothman
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Xiao, J.C. Morales
Study supervision: E.A. Motea, Y. Xie, D.A. Boothman
Others (provided plasmids and guided in their usage): A. Porter
We are grateful to Drs. Cheng-Ming Chiang and Shwu-Yuan Wu for their helpful suggestions with ChIP and dot-blot assays. We also thank Drs. Adi Gazdar, Yu-An Zhang, and Jaideep Chaudhary for their assistance with Bioinformatics. This work was supported by the NIH [R01 CA210489, to D.A. Boothman; Minority Supplement Award CA139217-05S1 and T32CA124334-06 (principal investigator: Dr. Jerry Shay; to E.A. Motea); and NIH/NCI CCSG 5P30CA142543 to the Simmons Comprehensive Cancer Center] and by the Medical Research Council (to A.C. Porter).
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.