Misrepair in Context: TGFβ Regulation of DNA Repair

Repair of DNA damage protects genomic integrity, which is key to tissue functional integrity. In cancer, the type and fidelity of DNA damage response is the fundamental basis for clinical response to cytotoxic therapy. Here we consider the contribution of transforming growth factor-beta (TGFβ), a ubiquitous, pleotropic cytokine that is abundant in the tumor microenvironment, to therapeutic response. The action of TGFβ is best illustrated in head and neck squamous cell carcinoma (HNSCC). Survival of HNSCC patients with human papilloma virus (HPV) positive cancer is more than double compared to those with HPV-negative HNSCC. Notably, HPV infection profoundly impairs TGFβ signaling. HPV blockade of TGFβ signaling, or pharmaceutical TGFβ inhibition that phenocopies HPV infection, shifts cancer cells from error-free homologous-recombination DNA double-strand-break (DSB) repair to error-prone alternative end-joining (altEJ). Cells using altEJ are more sensitive to standard of care radiotherapy and cisplatin, and are sensitized to PARP inhibitors. Hence, HPV-positive HNSCC is an experiment of nature that provides a strong rationale for the use of TGFβ inhibitors for optimal therapeutic combinations that improve patient outcome.

INTRODUCTION DNA repair is executed by multiple pathways that must be coordinated to deal with different types of DNA damage, including oxidative damage, single strand breaks (SSB), and double strand breaks (DSB). Complex intracellular mechanisms have developed to ensure an appropriate DNA damage response (DDR). In cancer, gene mutations and altered cell signaling can give rise to dysregulated and aberrant DNA repair mechanisms that presumably contribute to genomic instability and mutational burden that are associated with cancer progression.
Cancer cells are actively involved in crosstalk with host cells of the tumor microenvironment (TME), which includes the vasculature, immune cells and stroma, constitutes a robust but skewed signaling network distinct from normal tissue. Though the TME is critical in shaping the biology of a tumor, the impact of context-related signaling on the tumor cell's DNA repair proficiency is poorly understood. This article reviews cell intrinsic execution of DSB repair proficiency and pathway competency to discuss DNA repair in the context of the TME.
We focus on transforming growth factor-beta (TGFβ), which is critically involved in extrinsic control of pathway competency in DSB repair. Recently, we determined that compromised TGFβ signaling caused by human papilloma virus (HPV) in head and neck squamous cell carcinoma (HNSCC) shifts DSB repair to error prone and inefficient alternative end-joining (altEJ). HPV is an experiment of nature that provides compelling evidence that signaling from TGFβ, and thus the TME, is critical for DNA repair execution and pathway choice. The insight gained from understanding of the mechanisms by which TGFβ signaling, DDR, and TME are functionally linked, paves the way to further exploit weakness in specific cancers and develop pertinent therapeutic strategies.

DSB REPAIR PATHWAYS
Tens of thousands of DNA lesions are produced in a cell's daily life as a result of endogenous metabolic activities such as DNA replication or reactive oxygen species (ROS), as well as exposure to exogenous agents like ultraviolet (UV) or ionizing radiation (1,2). Of the many types of DNA damage, DSBs are among the most dangerous. Failure to repair DSBs may lead to mutations, genomic, and chromosomal rearrangements, or cell death. In order to maintain genomic integrity, two predominant DSB repair pathways have developed to deal with different types of lesions: classical non-homologous end-joining (c-NHEJ) and homologous recombination repair (HRR) [ Figure 1; (3)]. In mammalian cells, c-NHEJ is the predominant DSB repair pathway that can efficiently rejoin most DSBs (4). Although c-NHEJ functions throughout the cell cycle, it is particularly critical in the G1 phase when the cell has yet to replicate its DNA (5,6). C-NHEJ is initiated by the binding of the Ku70/Ku80 heterodimer to the ends of the DSB, which is then recognized FIGURE 1 | DSB repair pathways. Once a DSB is induced by a DNA damaging agent, cancer cells will try to repair it by one of these three mechanisms. Critical repair proteins in each pathway are shown. with high affinity by the catalytic subunit of DNA-dependent protein kinase (DNA-PKcs), forming the DNA-PK complex (7). This enables the recruitment of nucleases, including Artemis, to trim any short overhangs that are present on the DSB ends (8), and polymerases, including Polµ, to fill in any gaps (9)(10)(11). The final ligation step involves a ligase complex comprised of DNA ligase IV, X-ray repair cross-complementing protein 4 (XRCC4), and XRCC4-like factor (XLF), which is responsible for bridging and ligating the two processed ends (12). Because DSB ends often require processing to remove damaged nucleotides to enable ligation, c-NHEJ is considered an error-prone form of repair that can lead to short insertions and/or deletions. However, growing evidence suggests the context under which c-NHEJ is used in DSB repair is a critical determinant of the repair outcome, i.e., error-prone or error-prevention (13,14). Although it awaits further evidence, it has been speculated that orchestrated repair with c-NHEJ in normal cells prevents chromosome instability, but in cancer cells with dysregulated repair pathways, inappropriate implementation of c-NHEJ-dependent end joining of non-contiguous ends can cause genomic alterations and lead to chromosome instability (13,14).
The repair of DSBs by HRR is a highly complex process that requires the generation of single-stranded 3 ′ overhangs at each end, after which the homologous sequence on the sister chromatid is used to accurately fill in the gaps and restore the original DNA sequence (15,16). In the first step of HRR, the DSB ends are recognized by the MRN complex, which is composed of meiotic recombination 11 homolog A (MRE11), RAD50, and Nijmegen breakage syndrome 1 (NBS1). Once bound to the DSB, the MRN complex activates ataxia telangiectasia mutated (ATM), a serine-threonine kinase that initiates the DDR by phosphorylating a plethora of substrates, thus facilitating the recognition of the DSB and the activation of downstream repair factors. In addition to its role in the MRN complex, MRE11 also associates with C-terminal binding protein interacting protein (CtIP), an endonuclease that allows for the removal of ∼100 nucleotides from the 3 ′ end of the DSB (17,18).
Following the initial processing step by MRE11 and CtIP, more extensive resection is subsequently mediated by either the EXO1 exonuclease or a combination of the RECQ helicase (BLM or RECQL4) and the DNA2 exonuclease (19). The single-stranded DNA formed by these resection steps is very quickly coated with and stabilized by replication protein A (RPA), preventing them from being degraded or forming DNA hairpins (20). For the final phase of HRR, RPA is replaced with RAD51 to facilitate homology search, strand invasion into the sister chromatid, and initiation of DNA synthesis at the region of the DSB (21). This process requires several proteins known as recombination mediators, e.g., the tumor suppressor breast cancer 1, early onset 2 (BRCA2) (22,23). Repair is then completed upon resolution of crossover junctions by resolvases (24). Because HRR requires a homologous template, it can only be used to repair DNA that has been replicated (i.e., during the S and G2 phases of the cell cycle).
A third mechanism for repair of DSBs is altEJ, which is also referred to as backup end joining, or microhomology-mediated end joining (25,26). A comparison of repair strategies is depicted in Figure 1. AltEJ uses poly [ADP-ribose] polymerase 1 (PARP1) to tether broken DNA ends (27), DNA polymerase theta (Pol θ) coded by the POLQ gene to initiate DNA replication at sites within two single-stranded 3 ′ overhangs (28), and DNA ligase I (LIG1) or DNA ligase III (LIG3) to join the DNA ends (29). AltEJ commonly occurs at sites containing short complementary sequences, known as microhomology, that are exposed after end resection; this requirement for resection and minimal homology means that altEJ has low fidelity and therefore frequently results in small deletions, insertions, and gross chromosomal rearrangements (30,31). Because its execution increases genomic instability, altEJ is believed to be more active in certain cancers (32).
Other DSB repair pathways, such as single-strand annealing (SSA), can result in large deletions during repair by annealing of longer (e.g., >100 nt) repeats following extensive end-resection. These are rarely used in mammalian cells and have been reviewed recently (24), and will not be discussed herein.

DSB REPAIR PATHWAY COMPETENCY IN CANCER
The mechanism by which DSB are repaired is determined by a variety of factors, although the outcome is ultimately determined by the presence or absence of end resection. The initial phase of c-NHEJ, i.e., binding of the Ku heterodimer to DSB ends, minimizes end resection to allow accurate end-joining. End processing and resection are therefore tightly regulated by Ku70/Ku80, along with WRN and 53BP1, which together protect DNA ends during the G1 phase when HRR cannot occur due to the absence of a sister chromatid. Resection is also normally limited to late S or G2 due to the cell-cycle dependent expression of CtIP and its activation by CDK1 or CDK2 (33,34). Importantly, resection requires the repositioning of 53BP1 on DSB ends by BRCA1, and therefore the loss of BRCA1 inhibits HRR, which was demonstrated by the fact that a deficiency in 53BP1 rescues the defect in HRR caused by the absence of BRCA1 (35). Noordermeer et al. recently demonstrated that 53BP1 effector complex, shieldin, localizes to DSB to prioritize c-NHEJ repair (36). In BRCA1-deficient cells, loss of shieldin or its subunits can restore HRR and resistance to PARP inhibition (37).
AltEJ was initially believed to be a backup pathway for c-NHEJ and HRR (26). The Ku heterodimer has much higher affinity for DSB ends relative to PARP1; thus, c-NHEJ is highly favored over altEJ in most circumstances (38). A higher frequency of altEJ-mediated repair was observed after the depletion of HRR factors such as RPA, BRCA1, and BRCA2 (39), suggesting HRR is used with priority in normal settings. In addition, because both HRR and altEJ require an initial resection step at DSB ends, both pathways are inhibited by c-NHEJ factors. Conversely, end resection is sufficient to block repair by c-NHEJ, as Ku70/Ku80 has very low affinity for single stranded DNA (40). Notably, accumulating evidence suggests that altEJ also competes with HRR for the repair of DSB (28,41). For example, by studying dysfunctional telomeres and accumulation of RAD51 at DSBs, Mateos-Gomez and co-authors found that the loss of a critical component in altEJ, Pol θ, increased HRR in mice (28). Similar findings have been reported in ovarian cancers: HRR was upregulated when Pol θ expression was inhibited, while Pol θ expression blocks RAD51-mediated HRR due to RAD51 binding motifs in Pol θ (41).
Cell cycle phase plays an important role in DSB repair pathway choice. In S and G2 phases, HRR is preferentially used to repair DSB due to the presence of CYREN, an inhibitor of c-NHEJ (42). AltEJ is largely inactive in normal cells, but in quickly dividing cancer cells, altEJ may be increased to handle the increased level of DNA damage and, as a result, generate more mutations as byproducts. Although the cell cycle dependency of altEJ is not clear, it is possible that HRR-deficient cells use altEJ mainly in S or G2 phases, while c-NHEJ defects may increase altEJ in G1 phase. In addition, host cell type, chromosomal location, and epigenetic modification are also important factors for pathway competence at DSBs, which has been reviewed elsewhere (43).

EXPLOITING DNA REPAIR DEFICITS IN CANCER CELLS
Many DNA repair genes are tumor suppressors, and are frequently mutated during tumor progression. Loss of functions in some DNA repair genes increases compensating mechanisms of repair that may expose a targetable "Achilles heel." Synthetic lethality, which is defined as a cytotoxic response to the loss or inhibition of a gene or pathway that only happens in the presence of another specific genetic deficit, can specifically target cancer cells containing a defect in a DNA repair gene. The clinical application of synthetic lethality is best exemplified using PARP inhibitors in cells that have germ line or somatic mutations in BRCA1 or BRCA2 (44,45). PARP inhibition results in the accumulation of single-strand breaks (SSB) that produce DSBs upon collision with a DNA replication fork, which require HRR to repair. This is one mechanism by which BRCA1/2-deficient cancers are highly sensitized to PARP inhibition, although other functions of PARP may contribute as well. For example, PARP is directly involved in detoxifying endogenous ROS (46,47), while BRCA1 down-regulates cellular levels of ROS (48). Thus, BRCA deficient cells may be more sensitive to increased cytotoxic ROS that result from PARP inhibition. Although the clinical benefits of PARP inhibition are clear in BRCA1/2-deficient cancers, tumors with HRR defects in the absence of BRCA gene mutations (i.e., BRCAness) are also responsive to PARP inhibition (49). Hence, mutation of genes that are directly or indirectly involved in HRR or the Fanconi anemia pathway, a DNA repair pathway that intersects HRR, can contribute to so-called BRCAness, i.e., tumors that behave as if BRCA1 or BRCA2 are mutated (49). Biomarkers to assess BRCAness, which include transcriptional signatures, genomic scars, or levels of Rad51 foci, are under intensive investigation (50)(51)(52).

SYNTHETIC LETHALITY BETWEEN ALTEJ AND OTHER DSB REPAIR PATHWAYS
Besides its key role in SSB repair, PARP1 is also a critical component of altEJ; thus, PARP inhibition can cause synthetic lethality in tumors that rely on altEJ (25). Other altEJ components are promising therapeutic targets for tumors that depend on this repair pathway. Indeed Pol θ -mediated end joining becomes critical when either HRR or c-NHEJ fails. A recent genetic screen reported that both Pol θ and another altEJ component, the structure-specific endonuclease FEN1, are synthetic lethal with BRCA2 (53). The potential synergy of HRR and altEJ is indicated by embryonic lethality of combined loss of Brca1/Fancd2 and Polq in mice (28,41). Depletion of Pol θ in human cancer cells deficient in HRR due to absence of BRCA1/2 increases chromosomal aberrations and impaired cell survival (28). Pol θ loss also increases HRR-impaired cells sensitivity to DNAdamage (41).
Moreover, there is growing evidence that defects in HRR can lead to increased dependence on altEJ for repair, particularly in the context of fork stalling during replication stress (26,54). In addition to its relationship with HRR, altEJ may also be synthetic lethal with c-NHEJ. Combined deletion of POLQ and Ku70, as well as POLQ and 53BP1, leads to markedly reduced cell proliferation and survival associated with excessive end resection and chromosomal aberrations (55). Notably, POLQ is highly expressed in a subset of cancer types (56,57) and its expression is associated with poor prognosis (58,59).

DNA REPAIR IN CONTEXT: THE TUMOR MICROENVIRONMENT
Cancer cells react to endogenous (e.g., replication stress) or exogenous (e.g., radiation, chemotherapy) DNA damage in the context of their TME. The TME plays an important role in determining cancer clinical behavior and progression, and can influence cancer cell response to therapy (60,61). Components of the TME include cellular constituents of bone-marrow derived cells, fibroblasts and vessels, insoluble extracellular matrix, and soluble cytokines and chemokines (62). These TME components closely collaborate with cancer cells for development of a neoplastic phenotype. In this "team, " frequent interactions and crosstalk in a complicated signaling network unite them as a whole. Better understanding of "tumor as a whole" could provide information about the optimal use of therapies, and improve the development of personalized therapy based on integrated features of a tumor derived from both cancer cells and TME composition.
The influence of TME conditions on DNA repair is complex. Although DNA repair is largely regulated through autonomous signaling cascades, tissue-wide stress responses from DNA damage may be networked among tumor cells, stromal cells, and other TME components. TME composition influences DNA repair efficiency by transmitting inter-and intra-cellular signals in a tissue-specific fashion. Many TME factors, which include cytokines, extracellular matrix, stromal cells, hypoxia, and inflammation, are known to modulate DNA repair efficiency (63)(64)(65)(66)(67)(68).
Here we focus on TGFβ, a highly pleiotropic cytokine and a canonical tumor suppressor. Because TGFβ suppresses epithelial cell cycle progression, all carcinomas must escape TGFβ growth regulation (69). Notably, high TGFβ expression and signaling is associated with poor prognosis in multiple cancer types (52,70) because TGFβ becomes a tumor promoter that is involved in tumor progression by modifying the TME, suppressing immune response, and promoting metastasis (71). Due to these critical functions and association with poor outcomes, TGFβ is an intriguing therapeutic target in clinical trials (72,73).

TGFβ BIOLOGY
Understanding the biology of TGFβ is rooted in understanding when and where it is active. There are three mammalian TGFβ isoforms, each encode a polypeptide that is cleaved intracellularly to form a roughly 24 kD TGFβ that is non-covalently associated with an 80 kD dimer of its pre-pro peptide, called latency associated peptide (LAP). This complex, TGFβ and LAP, is secreted as the latent TGFβ complex and is often sequestered in forms bound to extracellular matrix. TGFβ canonical signaling is initiated upon the release of TGFβ ligand from its latent complex, and subsequent binding to the type II TGFβ receptor (TβRII), which causes recruitment and phosphorylation of type I receptor (TβRI). Activated TβRI kinase phosphorylates the carboxyterminal serine residue of the mothers against DPP Homolog proteins SMAD2 or SMAD3, which induces oligomerization of SMAD2 or SMAD3 with SMAD4, and DNA binding of the complex to mediate transcriptional activation or repression of target genes. TGFβ also transduces signals through noncanonical signaling pathways, such as MAPK/ERK, PI3K/AKT, Rho/Rock. Activation of TGFβ is highly controlled in normal tissues. In cancers, TGFβ signaling is highly dysregulated (70,74,75). Defective TGFβ signaling, which can be caused by mutations in SMAD4 and TGFBR2, are frequent in certain types of cancer (52,73).
Both cancer cells and stromal cells produce TGFβ that may, when activated, elicit paracrine or autocrine signaling to stimulate fibroblasts, endothelial cells and immune cells that further alter the TME. Moreover, TGFβ is a potent immunosuppressive cytokine involved in shaping TME by inhibiting the activation and function of T cells (76,77), inducing immune suppressive myeloid cells (78), as well as by other multifaceted mechanisms (79)(80)(81). These complex interactions are just a few of TGFβ's roles in the TME.
An unexpected role for TGFβ, an intrinsically extracellular signal, is response to intracellular DNA damage. Glick et al. (82) was likely the first study to implicate TGFβ in the cellular response to DNA damage. They showed that Tgfb1 null murine keratinocytes were highly genomically unstable, independent of G1 arrest and p53 function (82). Consistent with a role of TGFβ in maintaining genomic stability, Maxwell et al. (83) demonstrated more centrosome aberrations and aneuploidy in irradiated Tgfb1-null compared to TGFβcompetent keratinocytes. Notably, this effect is phenocopied by TGFβ inhibition in human epithelial cells (83). In a subsequent study, Glick and colleagues revealed a DNA repair deficit due to hypermethylated O(6)-methylguanine DNA methyltransferase (MGMT) that affected the DDR of Tgfb1-null keratinocytes (84). Similarly, Kirshner et al. (85) showed that inhibiting TGFβ signaling attenuates DDR by compromising the function of ATM, the DDR kinase involved in DSB recognition (85). Consequently, pharmaceutical inhibition of TβRI kinase or knockout of Tgfb1 reduces ATM autophosphorylation and phosphorylation of its substrates, e.g., p53, Chk2, and Rad17, inhibits formation of radiation-induced γH2AX foci, and increases radiosensitivity (85). In support of this, Wiegman et al. (86) showed that exogenous TGFβ stimulates ATM and p53 phosphorylation in irradiated cells in a SMAD-independent fashion (86). Notably, TGFβ inhibition also reduces LIG4 expression, which is required in c-NHEJ (87).
Nucleotide excision repair (NER) is a versatile DNA repair pathway that eliminates a wide variety of helix-distorting base lesions induced by environmental carcinogenic sources. UVB radiation downregulates E-cadherin, a cell adhesion protein, in mouse skin and skin tumors whereas inhibiting the TGFβ pathway in these cells increases the NER of UV-induced DNA damage (88). E-cadherin inhibition in keratinocytes suppresses NER through activating the TGF-β pathway and increasing TGFβ1 mRNA levels. Interestingly, TGFβ is activated by ionizing radiation and in turn, promotes epithelial-mesenchymal transition characterized by loss of E-cadherin (89). Treatment of cells with exogenous TGFβ enhances NER of DNA damage formed by polycyclic aromatic hydrocarbons and UVC radiation independent of the cell cycle (90).
Consistent with role of TGFβ in DDR, the SMAD proteins, which are the critical transducers of TGFβ intracellular signaling, are involved in DSB repair. Both pSmad2 and Smad7 can colocalize with nuclear γH2AX foci at DSB, while pSmad2 foci formation is ATM dependent (91). Studies in Smad4 conditional knockout mice confirmed that TGFβ is critical in maintaining genomic stability through regulation of genes in the Fanconi anemia/BRCA DNA repair pathway (92). SMAD4 suppresses a micro RNA, miR182, which inhibits FOXO3, which is required for ATM kinase activity (32,52). MiR182 also suppresses BRCA1 expression (93,94). Hence, TGFβ signaling through SMAD4 promotes HRR in part by suppressing miR182 (52). C-NHEJ is also partially compromised because LIG4 expression and ATM activity are reduced once TGFβ signaling is blocked (85,87).
Although tumors must evade TGFβ growth control, at the time of clinical appearance, many tumors maintain signaling competency. Indeed, squamous cell carcinomas may take advantage of TGFβ signaling to maintain a sub-population of cells at a quiescent state for chemo-resistance (95). Moreover, the high TGFβ activity of TME could promote tumor-intrinsic resistance to cytotoxic agents due to its role in DNA damage recognition and repair. If so, pharmacological TGFβ blockade could sensitize certain tumors to radiation and other cytotoxic therapies. Exploration of brain, breast, and lung cancer preclinical models is consistent with this concept, since TGFβ inhibition radiosensitized 38 of 43 murine and human cancer cell lines in vitro (85,(96)(97)(98). Because TGFβ provides extrinsic control of several aspects of intracellular DNA repair pathway competency, one prediction is that tumors that are insensitive to TGFβ can be exploited by targeting their deficiency in DNA repair.

TGFβ SIGNALING REGULATES DNA REPAIR PATHWAY COMPETENCY
The contribution of TGFβ signaling as a barrier compromising therapeutic response to cytotoxic therapy is exemplified by HPV-positive (HPV+) HNSCC (52), which have much better (70%) survival at 5 years compared to HPV-negative (HPV-) cancers that attain only 30% survival, even when HNSCC location and stage are similar. The considerable difference in outcomes has stimulated significant interest based on the idea that the mechanism of sensitivity of HPV+ cancer to standardof-care chemoradiation therapy, which could provide insights that can be therapeutically exploited to achieve better response in HPV-cancer.
Consistent with a cell-intrinsic effect, HPV+ cancer cell lines exhibit altered expression of DNA repair proteins (99) and increased sensitivity to cytotoxic therapy (100,101). Most research has focused on oncogenic impairment of p53 and retinoblastoma (Rb) proteins by HPV E6 or E7, respectively. However, HPV proteins E5, E6, and E7 target both the type I and II TGFβ receptors and SMAD 2, 3, and 4, the transducers of ligand binding, for degradation (102,103). We examined the impact of HPV on TGFβ signaling in HNSCC at multiple levels. Functionally, TGFβ induced phosphorylation of SMAD2 (pSMAD), indicative of signaling competency, is significantly reduced in HPV+ cell line, patient derived xenografts, and primary tumor explants, compared to HPV− specimens (52). Notably, HPV+ specimens in a HNSCC tumor array with 130 HPV− and 65 HPV+ samples exhibit less pSMAD compared to HPV− specimens. In addition, TCGA HPV+ tumors are identified by low activity (243 HPV− vs. 36 HPV+ ones) of a TGFβ pathway signature, which contains 50 TGFβregulated genes that were derived from epithelial cells chronically stimulated or inhibited for the corresponding pathway.
Interestingly, Wang and colleagues engineered a conditional Smad4 deletion in oral mucosa that gave rise to spontaneous HNSCC accompanied by high rates of genomic instability (92). Subsequent studies by this group showed that Smad4 deletion leads to decreased Brca1 expression in mice and that loss of SMAD4 protein correlates with decreased BRCA1 and RAD51 proteins in human HNSCC. BRCA1 is crucial for HRR during Sphase/G2, acts upon the cell cycle machinery, and affects gene expression and cell fate decisions via chromatin remodeling and transcriptional activity (104). Wang and colleagues showed that BRCA1 is transcriptionally down-regulated by SMAD4dependent CtPB1 (105). A second, more direct route by which TGFβ controls of BRCA1 is via TGFβ suppression of miR-182, which targets BRCA1 message stability and translation in mouse and human cells. Thus, pharmacologically or genetically compromised TGFβ increase levels of miR-182 and consequently suppress BRCA1 (94).
Consistent with our earlier studies in breast, lung and brain cancer cells (85,(96)(97)(98), HPV+ HNSCC cell lines were more sensitive to radiation than HPV− cell lines (52). Indeed, the degree of pSMAD response to TGFβ and cellular radiosensitivity are highly correlated. Radiation sensitivity reflects the cumulative damage and inherent capacity to repair the damage based on the cells ability to recognize DNA damage, assemble the repair machinery, and execute repair; abrogation of any of these components decreases cell survival. As noted above, BRCA1 is critical for HRR-mediated DNA repair. HRR requires RAD51 binding to 3 ′ -single-stranded DNA overhangs from processed DSB and strand pairing; thus, the formation of RAD51 foci is evidence of HRR. Significantly, fewer RAD51 foci are formed in HPV-HNSCC cells and tumor specimens if TGFβ is pharmacologically blocked, which is not observed in HPV+ HNSCC cells.
A deficiency in HRR should increase the proportion of cells that are killed in response to PARPi. As predicted, TGFβunresponsive HPV+ cell lines are more sensitive to olaparib alone compared to TGFβ-responsive HPV-cancer cells, which were sensitized 4-fold by TGFβ inhibition (52).
Loss of effective HRR can activate altEJ, which competes with HRR for repair of DSBs in S-phase (41) and/or acts as backup repair when HRR or c-NHEJ are compromised (26). As mentioned earlier, altEJ requires PARP1 and Pol θ, the product of the POLQ gene (28). To evaluate altEJ, we established TGFβ-responsive cells with a reporter construct detecting altEJ events (106). As expected, a specific PARP1 inhibitor reduced altEJ events, while TGFβ inhibitors significantly increased altEJ events.
Thus, either pharmaceutical blockade of TGFβ signaling in HPV− cells or intrinsic-defects in TGFβ signaling in HPV+ cells shifts DSB repair to altEJ (Figure 2). This shift may result from decreased implementation of HRR and c-NHEJ, or may indicate an increase in altEJ competency. Notably, BRCA1 heterozygous cells exhibit preferential use of altEJ (107), suggesting that the decrease in BRCA1 that occurs upon loss of TGFβ signaling phenocopies the genetic BRCA1 loss. As c-NHEJ is also partially compromised because LIG4 expression and ATM activity are reduced (85,87), cells with deficient TGFβ signaling may increase altEJ to compensate for a deficiency in both HRR and c-NHEJ.
TGFβ-unresponsive cells that depend on altEJ are still capable of rejoining most DSBs, which hinders the maximal cytotoxic response to DNA damage. As mentioned above, recent studies demonstrate that Pol θ is required in altEJ (28,41). HRR-deficient ovarian and breast cancers exhibit increased POLQ expression, perhaps indicative of altEJ (41). Consistent with this, POLQ expression is increased in HPV+ vs. HPV− HNSCC TCGA (52). Notably, POLQ shRNA expressing HPV− cells treated with TGFβ inhibitors were more sensitive DNA damage, supporting the idea that Pol θ-mediated altEJ is increasingly used when TGFβ signaling is abrogated. These data suggest that combining altEJ inhibitors with pharmaceutical blockade of TGFβ signaling would induce synthetic lethality, thus creating a novel route that could boost the sensitivity of TGFβ-active tumors to therapies that involve DNA damage. Moreover, since altEJ is a backup mechanism that operates preferentially in cancer cells, inhibiting altEJ may pre-dominantly sensitize tumor cells and spare normal cells. Alternatively, one might select cancers that are defective in TGFβ to target drugs that interfere with altEJ, since these drugs will both improve the treatment effectiveness and the toxicity profile.
The importance of alt-EJ in TGFβ-deficient cells discovered in HNSCC raises several questions. We are following this lead to determine if the same mechanism is evident in other cancers in which TGFβ signaling is compromised. Almost all cervical cancers are HPV positive, which we would expect to show similar DDR choices as HNSCC. However, TGFβ signaling is compromised by various mechanisms in many cancers (108). Our analysis across the spectrum of cancers suggest that there is probably a generic mechanism (unpublished). We also find that the repair shift to altEJ in TGFβ-deficient cells occurs independent of miR-182. Further studies are necessary to decipher the underlying mechanisms.
Although most evidence supports TGFβ as a factor enforcing DNA-repair proficiency to protect against genomic instability, conflicting evidence about the role of TGFβ in DDR also exists (109)(110)(111)(112). For example, Pal et al. reported that TGFβ hinders DSB repair in cancer stem cells by reducing HRR gene expression, which was proposed to heighten genetic diversity and adaptability of cancer stem cells (112). This subpopulation may regulate DDR differently, but it is interesting that in another study, glioblastoma cancer stem cells make 5-fold more TGFβ than bulk cultures (97). Considering that TGFβ is a pleotropic cytokine, the extensive technical and conditional differences in these studies may lead to a different conclusion.
For example, TGFβ signaling induces squamous cell carcinoma cancer stem cells quiescence, which would be expected to affect repair pathway competency in a manner that contributes to chemo-resistance (95). However, many cancer cells have escaped TGFβ cell growth control, as stated above. For example, in our HNSCC study, the cell cycle was not significantly changed upon TGFβ stimulation or inhibition of HNSCC cell FIGURE 2 | Defective TGFβ signaling in HNSCC increases altEJ (52). (A) TGFβ signaling suppresses miR182, which suppresses expression of BRCA1 and FOXO3. DNA damage elicits activates ATM autophosphorylation and its phosphorylation of BRCA1, which gives rise to use of HRR in S-phase. (B) In contrast, deficient TGFβ signaling, which can be caused by HPV infection, TGFβ receptor kinase inhibitors (TBRi) or SMAD4 mutations (SMAD4 mut), leads to increased miR182 that suppresses BRCA1 and FOXO3. Loss of FOXO3 inhibits ATM auto-activation, which together with decreased BRCA1, impedes HRR. This is accompanied by increased altEJ.
lines (52), which indicates that TGFβ can regulate DNA repair pathway by mechanisms independent of cell cycle effects.

OUTLOOK
Intensive investigation of the TME has advanced our knowledge about the tumor as a whole, whereas in-depth analysis of DDR now provides mechanisms of DNA repair strategies and their implementation in cancer cells. How DDR is executed in different tumors is an area of growing complexity. The highly heterogeneous TME is a function of cancer cell genetics, epigenetics, host cell composition that result in complicated signaling networks. However, opportunities also lie in these challenges. DDR-related signaling pathways regulated by the TME components may contain biomarkers for cancer stratification, as well as molecular or cellular targets for drug development, which deserves significant investigation.