Bioenergetic metabolites regulate base excision repair dependent cell death in response to DNA damage

Abstract

Base excision repair (BER) protein expression is important for resistance to DNA damage-induced cytotoxicity. Conversely, BER imbalance (Polß deficiency or repair inhibition) enhances cytotoxicity of radiation and chemotherapeutic DNA-damaging agents. Whereas inhibition of critical steps in the BER pathway result in the accumulation of cytotoxic DNA double-strand breaks, we report that DNA damage-induced cytotoxicity due to deficiency in the BER protein Polß triggers cell death dependent on PARP activation yet independent of poly(ADP-ribose) (PAR)-mediated AIF nuclear translocation or PARG, suggesting that cytotoxicity is not from PAR or PAR-catabolite signaling. Cell death is rescued by the NAD⁺ metabolite NMN and is synergistic with inhibition of NAD⁺ biosynthesis, demonstrating that DNA damage-induced cytotoxicity mediated via BER inhibition is primarily dependent on cellular metabolite bioavailability. We offer a mechanistic justification for the elevated alkylation-induced cytotoxicity of Polß deficient cells, suggesting a linkage between DNA repair, cell survival and cellular bioenergetics.

Introduction

Efficacy of chemotherapy or radiation treatment is intimately dependent on DNA repair capacity (1). Robust repair of therapeutically induced DNA damage can provide significant resistance whereas tumor-specific defects in DNA repair or inhibition of specific DNA repair proteins can provide therapeutic advantage (2). In particular, inhibiting base excision repair (BER) can be an effective means to improve response to temozolomide (TMZ), radiation, bleomycin and cisplatin, among other treatments (3-10). As with most DNA repair pathways, BER is a multi-step mechanism comprised of greater than 20 proteins, depending on the initial base lesion (3). However, inhibiting each step in the BER pathway will have different outcomes. DNA glycosylase inhibition or loss blocks BER initiation, leading to the accumulation of both cytotoxic (4) and mutagenic base lesions (5), the latter contributing to cellular dysfunction. In this regard, the preferred option is the inhibition of BER after repair initiation, promoting the accumulation of cytotoxic BER intermediates such as abasic sites and DNA single-strand breaks by inhibiting abasic site repair with methoxyamine, inhibiting the BER enzyme poly(ADP-ribose)polymerase-1 (PARP1) or by loss or inhibition of DNA polymerase ß (Pol ß) (2, 6, 7). We refer to inhibition of the intermediate steps in BER as the induction of “BER failure”, since repair is initiated yet is unable to be completed.

Importantly, understanding the mechanisms that are responsible for the increase in cell death due to BER inhibition or BER failure is critical in tailoring treatment, as well as, designing rational adjuvant or combination treatments that may further increase overall response. For example, inhibiting PARP1 has proven effective in improving TMZ induced cell death (8). Inhibition of PARP1 results in the accumulation of replication-mediated DNA double-strand breaks (DSBs) and the onset of apoptosis (9, 10). This detailed understanding of the mechanism of cell death induced by combining a DNA damaging agent (TMZ) and a PARP1 inhibitor suggests that PARP1 inhibition would be effective against many tumors but may be ineffective against tumors that are resistant to apoptosis (11). Further, cell death induced by PARP1 inhibition suggested a requirement for homologous recombination (HR) in the cellular response to the accumulated DSBs, prompting pre-clinical and clinical trials of PARP1 inhibitors in the treatment of HR defective tumors (2).

There are several BER proteins essential for the repair of TMZ-induced DNA lesions. Using a mouse embryonic fibroblast (MEF) cell model, we have shown that loss of Pol ß can significantly improve the cytotoxic effect of TMZ (12), suggesting that inhibition of Pol ß may improve response to TMZ in human tumor cells. TMZ is currently used in the treatment of glioblastoma (13) and it is therefore critical to evaluate the role of Pol ß in glioma cell response to TMZ treatment. No previous studies have investigated the role of Pol ß in the response of human glioma tumor cells to TMZ. Further, there is no mechanistic explanation for the increase in alkylation-induced cell death observed in cells that are deficient in Pol ß beyond the evidence that cell death in mouse cells is the result of accumulation of un-repaired BER intermediates (7, 12).

Acute alkylation damage has been suggested to induce cell death by multiple mechanisms, including necrosis (14), caspase-3 and caspase-9 activation and the onset of apoptosis (15), apoptosis inducing factor (AIF) translocation from the mitochondria to the nucleus (16-18), ADP-ribose induced activation of the Ca²⁺ channel TRPM2 or AMP-mediated inhibition of ATP transport (19-21). In most, if not all cases, cell death has been attributed to the direct action of either poly(ADP-ribose) (PAR) formed by PARP1 activation or PAR catabolites that accumulate after PAR degradation by the catabolic enzyme poly(ADP-ribose) glycohydrolase (PARG). Pol ß deficient mouse cells are hypersensitive to the cell killing effects of alkylating agents due to failure to repair the 5’dRP BER intermediate (22). However, the exact downstream signaling events and mechanism of cytotoxicity specifically induced by the un-repaired 5’dRP lesion remains unclear. Previous studies in mouse cells have not been conclusive. One report suggested that the absence of Pol ß led to damage-induced cell death via apoptosis (23) whereas a later study proposed a necrotic form of cell death for both wild-type and Pol ß deficient cells (24), similar to what has been proposed as a general mechanism of alkylation-induced cell death in mouse fibroblasts (14). However, this latter study required the use of apoptosis-deficient cells to observe necrotic cell death (14). None of these previous studies have identified a mechanism of cell death specific to Pol ß deficiency and BER failure or a failure to repair the cytotoxic BER intermediate 5’dRP.

The studies described herein were designed to specifically define the mechanism of cell death in human tumor cells resulting from failure to repair the BER intermediate 5’dRP due to ‘inhibition of’ or a ‘deficiency in’ Pol ß (BER failure). We have hypothesized that PARP1 functions in BER as both a complex coordinator and as a molecular repair sensor. As a BER molecular sensor, we suggest that PARP1 facilitates cell death in response to incomplete BER or BER failure. In support of this hypothesis, we show that a specific BER intermediate, a single-strand DNA strand break containing a 3’OH and 5’dRP, is an in vivo substrate in human cells that activates PARP1 in the context of BER and that elevated cytotoxicity observed in Pol ß deficient human cells is controlled by the activation of PARP1. Further, we provide clear evidence that following “BER failure” human cells die independent of RIP1 activation or AIF translocation, thus ruling out PAR as the cell death signal that is initiated upon BER failure. Further, we show that the observed cell death in Pol ß deficient cells is un-related to the accumulation of PAR catabolites such as ADP-ribose or AMP yet is dependent on NAD⁺ metabolite bioavailability or the bioenergetic capacity of the cell.

This study provides mechanistic insight into why Pol ß deficiency leads to cell death, defines the mode of death and offers a mechanistic link between BER failure and energy metabolism - the novel finding that DNA damage-induced cytotoxicity mediated via BER inhibition is primarily dependent on cellular metabolite bioavailability. Finally, we offer a mechanistic justification for the elevated alkylation-induced cytotoxicity of Pol ß deficient cells, suggesting a linkage between DNA repair, cell survival and cellular bioenergetics.

Results

Hyperactivation of PARP due to Pol ß deficiency and failure to repair the base excision repair intermediate 5’deoxyribose phosphate

BER is a finely tuned process that requires balanced expression of several proteins to avoid accumulation of mutatgenic or cytotoxic repair intermediates (3). To understand how alterations in BER enzyme activity in human tumor cells leads to DNA damage-induced cell sensitivity, we developed human glioma (LN428) cell lines with a functional deficiency in Pol ß by increasing expression of MPG and depleting the cell of Pol ß by stable, lentiviral-mediated expression of shRNA. As we have reported, human cells with elevated expression of MPG are sensitive to alkylation damage due to a deficiency in Pol ß (25), a phenotype that is enhanced by Pol ß knockdown (Pol ß-KD). Conversely, re-expression of Pol ß eliminated the alkylation hypersensitive phenotype (Figure S1 & S2; supplemental material). These cells (LN428/MPG and LN428/MPG/Polß-KD cells) are therefore functionally deficient in Pol ß and were utilized to determine the mechanism that mediates the enhanced DNA damage-induced cell death resulting from Pol ß deficiency.

The DNA binding and signaling molecules PARP1 and PARP2 have each been implicated in BER (3). PARP1 facilitates BER complex formation and it has been postulated that local, strand-break induced activation of PARP1 and the resultant synthesis of PAR mediates recruitment of the BER proteins XRCC1 and Pol ß to stimulate DNA repair (26). We therefore have hypothesized that in cells that fail to complete BER (e.g., when 5’dRP lesions are not repaired; herein referred to as ‘BER Failure’), PARP1 is hyper-activated and functions as a DNA damage signaling protein that triggers cell death. To determine whether PARP is activated by the BER intermediate (5’dRP) in vivo, we exposed the control (LN428) and corresponding BER defective cells (Pol ß deficient LN428/MPG and LN428/MPG/Polß-KD cells) to TMZ for up to 90 minutes. Whole cell extracts were probed by immunoblot for PAR accumulation following TMZ exposure (Figure 1A). The level of PAR accumulation was shown to correlate with the extent of the BER defect. PARP activation was elevated in the LN428/MPG cells (an intermediate level of sensitivity), with the highest level of PAR observed 30 minutes following exposure to TMZ whereas essentially no PARP activation was observed in the LN428 cells (Figure 1A). In the more sensitive cell line (LN428/MPG/Polß-KD), PARP activation was more robust and rapid as compared to that of the LN428/MPG cell line (Figure 1A), as PAR reached its highest level at 15 minutes after exposure to TMZ. Comparable results were also observed in a Pol ß defective breast cancer cell line where elevated TMZ-induced PARP activation is restricted to the cells with Pol ß deficiency (Figure S2B, C; supplemental material). Conversely, exposure to etoposide resulted in a low level of PARP activation at all time points for all three cell lines LN428, LN428/MPG and LN428/MPG/Polß-KD (Figure S2D; supplemental material). Thus, PARP activation is elevated in BER defective (Pol ß deficient) cells following alkylation damage.

Figure 1. PARP activation due to BER failure.

(A) Immunoblot of PAR to determine activation of PARP in the cell lines indicated before and after exposure to TMZ (1.5mM) for the time indicated. PARP1 and PCNA protein expression levels are also shown.

(B) (Upper panel) TMZ induced cytotoxicity in LN428, LN428/MPG and two clones of LN428/MPG cells complemented with Flag-Pol ß-WT. After treatment (48 hrs), viable cells were determined using a modified MTT assay. Plots show the % viable cells as compared to untreated (control) cells. Means are calculated from quadruplicate values in each experiment. Results indicate the mean ± S.E. of three independent experiments. [LN428, open square; LN428/MPG, open circle; LN428/MPG/Flag-Polß-WT, clones 1 and 6, filled triangle] (Lower panel) Immunoblot of PAR to determine activation of PARP1 after exposure to TMZ (1.5mM) for the time indicated. PCNA is shown as a loading control.

(C) (Upper panel) TMZ induced cytotoxicity in LN428, LN428/MPG and two clones of LN428/MPG cells complemented with Flag-Pol ß-K72A, measured as in Figure 1B. [LN428, open square; LN428/MPG, open circle; LN428/MPG/Flag-Polß-K72A, clones 5 and 16, filled triangle] (Lower panel) Immunoblot of PAR to determine activation of PARP1 after exposure to TMZ (1.5mM) for the time indicated. PCNA is shown as a loading control.

Since the combination of alkylating agent treatment and Pol ß deficiency triggers PARP activation, we next validated the significance and specificity of this finding by re-expression of Pol ß in the LN428/MPG and LN428/MPG/Polß-KD cells. We find that the BER deficient phenotype (increased cellular sensitivity to alkylating agents) observed in both the LN428/MPG and LN428/MPG/ Polß-KD cells was reversed by complementation (expression) of FLAG-Pol ß (WT) (Figure 1B, upper panel and Figure S1E; supplemental material) but not the 5’dRP lyase deficient (K72A) mutant of Pol ß (Figure 1C, upper panel). Similarly, we find that complementation with FLAG-Pol ß (WT) but not with the Pol ß 5’dRP lyase mutant eliminated the TMZ-induced activation of PARP observed in BER defective cells (Figure 1B & C; lower panel). These data therefore suggests that the Pol ß specific BER intermediate (5’dRP lesion) triggers rapid and robust PARP1 activation in vivo, triggering the onset of cytotoxicity.

The correlation between PARP activation and alkylation sensitivity prompted us to determine if inhibition of PARP reverses the cellular hypersensitivity of Pol ß deficient human tumor cells. We inhibited activation of PARP by pre- and co-treatment with the PARP1/PARP2 inhibitors PJ34 or DR2313. Inhibition of PARP by PJ34 significantly reduced the level of TMZ-induced PARP activation in the Pol ß deficient cells (LN428/MPG) (Figure 2A; lanes 3,4 & 7,8). We next assayed if PARP inhibition can rescue the alkylation-sensitive phenotype of LN428/MPG cells, as determined by an MTS assay 48 hours after TMZ exposure. Most importantly, we find that PARP inhibition by either PJ34 or DR2313 treatment converted the LN428/MPG cells from a sensitive phenotype to a resistant phenotype (Figure 2B and Figure S3A; supplemental material). Rescue by PARP inhibition was also observed in Pol ß deficient MDA-MB-231 cells (Figure S3B; supplemental material). It remains to be determined if this resistant phenotype is long-lived. In photoreceptor, PC12, SH-SY5Y and HeLa cells, PARP inhibition has been shown to be cyto-protective, as we observe herein (27-29). Further studies will determine if this resistant phenotype reported here persists or if the BER failure-induced single-strand breaks lead to the formation of DNA double-strand breaks and the onset of apoptosis after several rounds of replication (9). Regardless, these studies support our hypothesis that PARP hyperactivation mediates the alkylation sensitive phenotype of Pol ß deficient cells.

Figure 2. PARP activation mediates cellular hypersensitivity in BER defective cells.

(A) TMZ induced PARP activation in LN428 and LN428/MPG cells in the presence or absence of the PARP1/PARP2 inhibitor PJ34. Cells were pre-treated with PJ34 (4 μM) or vehicle control for 30 min before exposure to TMZ (1.5 mM) plus PJ34 (2 μM) for another 30 min. PCNA was used as a loading control.

(B) TMZ induced cytotoxicity (LN428 and LN428/MPG cell lines) in the presence (solid lines) or absence (dashed lines) of the PARP1/PARP2 inhibitor PJ34. Viable cells were measured 48 hours after exposure as in Figure 1B. [LN428, open and filled square; LN428/MPG, open and filled circle]

Un-repaired base excision repair intermediates (5’dRP lesions) trigger cell death via energy depletion in the absence of PAR or PAR-catabolite mediated signaling

A number of different mechanisms have been attributed to PARP1 activation-induced cell death. We first evaluated the involvement of caspase-dependent cell death in control cells as compared to the corresponding Pol ß deficient cells following TMZ treatment. These experiments (Figure S4A,B; supplemental material) rule out a caspase-dependent response due to BER failure, in-line with our previous report (25). Although it has been demonstrated that an autophagic-response contributes to TMZ-induced cell death in some cells (30), TMZ hypersensitivity of Pol ß deficient cells is not affected by the autophagy inhibitor 3-MA (Figure S4C; supplemental material). In support of this observation, we did not observe increased LC3 puncta in BER defective cells following TMZ exposure (25).

A major mechanism that has been attributed to PARP-activation induced cell death is direct PAR signaling to the mitochondria where PAR mediates translocation of apoptosis inducing factor (AIF) from the mitochondria to the nucleus to induce caspase-independent cell death (16-18) via a mechanism that requires receptor (TNFRSF)-interacting serine-threonine kinase 1 (RIP1) activation (31) (see model; Figure 3A). RIP1 can be inhibited by necrostatins, small molecule inhibitors shown to inhibit cell death (32, 33). Therefore, we investigated the role of RIP1 in the PARP-mediated cell death we observed by inhibiting RIP1 with Necrostatin-1 (33) and evaluating the impact of RIP1 inhibition on DNA damage-induced cell survival in both control and Pol ß deficient cells. However, inhibition of RIP1 did not prevent cell death in either the parental or Pol ß deficient cells (Figure S5; supplemental material), suggesting but not proving that AIF translocation may not be related to the observed cell death.

Figure 3. Absence of PAR or PAR-catabolite mediated cell death following BER failure.

(A) Model depicting the nexus of BER, the synthesis of PAR and the generation of PAR catabolites in response to BER failure-induced PARP1/PARP2 hyperactivation.

(B) Absence of mitochondria to nucleus translocation of AIF due to BER failure as determined by confocal microscopy. BER deficient cells (LN428/MPG) were treated with media (left panel) or 1.5 mM MMS (right panel) for 1 hour and then washed and allowed to recover in media for 5 hours prior to fixation and staining for AIF (green), actin (red) and nucleus (blue).

(C) PARG KD prevented degradation of DNA damage-induced PAR. (Left Panel) Immunoblot of PAR to determine the degradation of PAR in LN428/MPG/PARG-KD cells following treatment with 1.5 mM TMZ. PCNA protein expression level was shown as a loading control. (Right Panel) Preventing generation of PAR catabolites from degradation of PAR via PARG KD enhances TMZ-induced cytotoxicity. LN428 and LN428/MPG cells with (black solid bars) or without (white empty bars) PARG KD were exposed to TMZ (1 mM) or vehicle control (DMSO) for 48 hours. Viable cells were determined as in Figure 1B and reported as percentage relative to vehicle control treated cells (% control). Results indicate the mean ± S.E. of three independent experiments.

(D) HMGB1 released into the cell culture media, as demonstrated by immunoblot. LN428 and LN428/MPG cells were pre-treated with PJ34 (4 μM) or vehicle control for 30 min and then exposed to TMZ (1.5 mM) with or without PJ34 (2 μM) for 12 hours. HMGB1 was then captured using immobilized heparin and analyzed by immunoblot, as described in the Materials and Methods section.

We therefore next evaluated the sub-cellular localization of AIF in control and Pol ß deficient cells following exposure to the alkylating agents MMS or TMZ as compared to vehicle (media) by immunoflourescent staining and confocal microscopy (Figure 3B) or by sub-cellular fractionation and imunoblot analysis (Figure S6; supplemental material). In-line with the RIP1 inhibition data above, alkylating agent treatment of Pol ß deficient cells did not alter the sub-cellular localization of AIF (Figure 3B & S6; supplemental material). All the detectable AIF was localized to the mitochondria in both cell lines regardless of agent or time of exposure (up to 12 hours), thus ruling out PAR as a cell death signal upon BER failure.

In the absence of a PAR-mediated cell death process (AIF translocation), it is possible that cell death is initiated via the rapid breakdown of PAR (see Figure 1A) by the degradative enzyme PARG and the accumulation of the PAR catabolites ADP-ribose, ribose-5-phosphate and/or AMP (see model; Figure 3A) (34). ADP-ribose acts as a second messenger to activate the cation channel TRPM2 to trigger Ca²⁺ influx, resulting in cell death (19, 20) or inhibits ABC transporters (35) whereas elevated AMP can block ATP transport, leading to ATP depletion and cell death (21). To investigate the possibility that PAR catabolites contribute to PARP-mediated cell death in Pol ß deficient cells, we first blocked Ca²⁺ influx with BAPTA-AM, shown recently by Boothman and colleagues to abrogate PARP1-activation induced cell death (36, 37). Unlike that observed following DNA damage from reactive oxygen species (ROS) or oxidative stress, BAPTA-AM did not prevent the elevated damage-induced cell death in Pol ß deficient cells (Figure S7; supplemental material). However, as there may be multiple mechanisms of PAR-catabolite-induced cell death, we next knocked-down expression of PARG by stable-transduction of both cell lines with a lentivirus expressing shRNA specific to PARG. Expression of PARG mRNA is reduced to 35% as compared to the GFP-control cells when determined by qRT-PCR (not shown). Importantly, we found no evidence for PAR degrading activity in the cells with stable depletion of PARG (Figure 3C; left panel). When exposed to an alkylating agent, BER deficient PARG-KD cells accumulate significant levels of PAR with no evidence for PAR degradation (Figure 3C, left panel lanes 2-4). This is in contrast to the presence of PARG, when the PAR molecule is degraded within 60-90 minutes (Figure 1, lanes 7-12). These data demonstrate that these PARG-KD cells do not degrade PAR and hence do not accumulate PAR-catabolites, providing an opportunity to determine if PAR catabolites contribute to cell death in these cells. As shown in Figure 3C (right panel), PARG-KD did not rescue or reverse the enhanced damage-induced cell death phenotype of Pol ß deficient (LN428/MPG) cells. In fact, PARG-KD cells (black bars) were more sensitive to the cell killing effect of the alkylating agent TMZ as compared to the PARG expressing cells (open bars)(Figure 3C, right panel). The inability of necrostatins to abrogate the response and the lack of PAR-mediated AIF translocation strongly suggests that PAR is not acting as a signaling molecule to induce cell death, as has been suggested (38, 39). Further, the inability of BAPTA-AM and most importantly, PARGKD, to reverse the alkylation-sensitive phenotype of Pol ß deficient cells also suggests that the observed cell death is un-related to the accumulation of PAR catabolites such as ADP-ribose or AMP. Finally, one of the hallmarks of caspase-independent cell death is secretion of HMGB1 into the extracellular space (40, 41). A significant level of HMGB1 was secreted into the culture media following exposure of the BER defective cells (LN428/MPG) to TMZ as compared to that of the control LN428 cells (Figure 3D). HMGB1 release was mediated through PARP activation, likely due to PARP1 modification (41), as PARP inhibition greatly reduced the release of HMGB1 (Figure 3D). It is unclear how or if HMGB1 release due to failed BER is related to the recently reported role of HMGB1 in BER (42).

An alternate process of cell death due to PARP1 activation was originally proposed by Berger to involve energy (NAD⁺ and ATP) depletion (43, 44), in support of an earlier observation by Jacobson and colleagues demonstrating a decrease in NAD⁺ concurrent with an increase in PAR synthesis (45). We therefore measured NAD⁺ and ATP levels in the control (LN428) and Pol ß deficient (LN428/MPG and LN428/MPG/PolßKD) cells before and after exposure to MMS or TMZ. In line with the cytotoxicity and PARP1 activation results described above, exposure of Pol ß deficient cells to MMS or TMZ led to a rapid and drastic depletion of both NAD⁺ and ATP whereas the NAD⁺ and ATP levels in the control cells were not affected (Figure 4A). We next measured the impact of alkylation damage on the corresponding cells depleted of PARG (PARG-KD). If PAR catabolites trigger cell death, we would expect that NAD⁺ and ATP loss would be attenuated in PARG-KD cells. However, exposure of Pol ß deficient PARG-KD cells to TMZ led to enhanced depletion of both NAD⁺ and ATP (Figure 4B). The absence of PAR or PAR-catabolite mediated cell death together with the specific loss of NAD⁺ and ATP (Figure 4A) even when the formation of PAR-catabolites are prevented (Figure 4B), suggests that the BER failure response is linked to the cellular bioenergetic capacity of the cell.

Figure 4. BER failure induced cell death depends on NAD⁺ availability.

(A) Alkylation damage promotes NAD⁺ and ATP depletion in BER defective cells. NAD⁺ content (left panel): Cells were treated with media (white bars) or 0.5 mM MMS (black bars) for 1 hour prior to collection for NAD⁺ content analysis via enzymatic assay as described in the Materials and Methods section. ATP content (right panel): Cells were treated with media (white bars), 0.5 mM MMS (grey bars) or 1.5 mM MMS (black bars) for one hour. ATP content was measured after 1-hour recovery in normal media via the luminescence ATP assay described in the Materials and Methods section. NAD⁺ levels or ATP levels shown are the average of three independent experiments and are reported as percent control of the untreated control cell line.

(B) PARG-KD does not rescue alkylation damage induced NAD⁺ and ATP depletion in BER defective cells. NAD⁺ content (left panel): PARG-KD cell lines were treated with media (white bars) or 0.5 TMZ (black bars) for one hour prior to collection for NAD⁺ content analysis as described in the Materials and Methods section. ATP content (right panel): PARG-KD cells were treated with media (white bars), 0.5 mM TMZ (dotted bars), 1.0 mM TMZ (grey bars) or 1.5 mM TMZ (black bars) for one hour. ATP content was measured after 1-hour recovery in normal media via the luminescence ATP assay described in the Materials and Methods section. NAD⁺ levels or ATP levels shown are the average of three independent experiments and are reported as percent control of the untreated control cell line.

(C) Bioenergetic metabolites rescue Pol ß deficient cells from DNA damage-induced cell death. LN428 and LN428/MPG cells were pre-treated with NMN, NA or vehicle control (media) for 24 hours and were then exposed to TMZ (1 mM) in the presence or absence of NMN or NA for 48 hours. Viable cells were measured as in Figure 1B and were reported as percentage relative to vehicle control treated cells (% control). Results indicate the mean ± S.E. of three independent experiments.

(D) NAD⁺ biosynthesis inhibition augments BER failure-induced cell death. Cells were pretreated for 24 hours with a non-toxic 10nM dose of FK-866 (black bars) or DMSO (white bars). Cells were then exposed to media control or MMS (0.5 mM) for 1 hour. Viable cells were determined as in Figure 1B and results indicate the mean ± S.E. of three independent experiments.

For this paradigm to hold, we hypothesized that the availability of bioenergetic metabolites would impact the survival of Pol ß deficient cells exposed to an alkylating agent. In-line with this hypothesis, we find that supplementation of the cells with either ß-nicotinamide mononucleotide (NMN) (46) or nicotinic acid (NA) reversed the DNA damage-induced phenotype, rendering the Pol ß deficient cells (black bars) completely (NMN) or 80% (NA) resistant to the cell killing effects of the alkylating agent, as compared to the BER proficient cells (open bars) (Figure 4C). Conversely, we anticipated that the hypersensitive phenotype of Pol ß deficient cells would be exacerbated by a reduction in the cellular level of NAD⁺ and related bioenergetic metabolites. We therefore evaluated the impact of transient NAD⁺ depletion on the observed “BER Failure” response by pre-treating cells with FK-866, a highly specific non-competitive small molecule inhibitor of nicotinamide phosphoribosyltransferase (NAMPT), a critical enzyme in the NAD⁺ biosynthetic salvage pathway that catalyzes the synthesis of NMN (47). Most importantly, the sensitivity of control cells to alkylation damage was not altered by FK-866 treatment. However, the BER deficient cells are 9-fold more sensitive to MMS following a non-toxic (10 nM) treatment with FK-866, as compared to the untreated cells (Figure 4D) even though PAR synthesis after the combined FK-866 + MMS treatment is attenuated (Figure S8; supplemental material). These results support our overall hypothesis that the BER failure phenotype of Pol ß deficient cells is mediated by BER intermediate (5’dRP) induced PARP1 activation and induction of caspase-independent cell death that is uniquely dependent on the availability of bioenergetic metabolites such as NMN and NAD⁺.

Discussion

The requirement for BER in general and Pol ß more specifically in the repair of genomic DNA base damage, particularly DNA damage induced by alkylating agents such as the chemotherapeutic TMZ and the SN1 & SN2 alkylating agents MNNG and MMS, respectively (7, 12), elevates the significance of characterizing the mechanism responsible for Pol ß deficiency-induced cell death [e.g., a failure to complete repair of the BER intermediate 5’dRP in the absence of Pol ß]. As evidenced recently by the development of clinically significant PARP1 inhibitors, identifying BER proteins critical for response to DNA damaging agents (e.g., chemotherapy) can have broad human health implications. Equally important is a clear understanding of the mechanism(s) that contribute to the enhanced cell death observed upon DNA repair inhibition. For example, PARP1 inhibition triggers apoptosis via the accumulation of DSBs (9, 10) and a requirement for homologous recombination proteins such as BRCA1 and BRCA2 (2). To this end, we have developed a unique series of genetically modified human tumor cell lines as models of Pol ß deficiency that accumulate the cytotoxic BER intermediate 5’dRP following exposure to alkylating agents (TMZ, MMS and MNNG). By directly comparing BER (Pol ß deficient) defective and BER competent isogenic human cell lines, the cellular, biochemical and signaling responses to DNA base damage can be defined as either global (non-specific) or BER (Pol ß) specific effects, the latter resulting from a cellular response to the inability to complete BER, referred to herein as “BER Failure”. We have then utilized this system to define the mechanism of cell death resulting from Pol ß loss/inhibition or BER failure and propose and test paradigms to enhance the cell death response.

From these studies, we find that the un-repaired BER intermediates that accumulate upon DNA damaging agent exposure when Pol ß is deficient will activate PARP1, leading to a rapid onset of PARP1-dependent, caspase-independent cell death with little or no role for a caspase-dependent or autophagy-dependent process in the response. It remains to be determined if the BER failure-induced cell death observed herein is dependent on ERK1/2–mediated PARP1 phosphorylation (48), SIRT1-regulated deacetylation of PARP1 (49) or if the observed PARP1-induced cell death requires BAX, Calpain and JNK activation (50). Coincident with damage-induced necrosis in Pol ß deficient cells is PARP1-dependent HMGB1 secretion (41), a hallmark of caspase-independent cell death and inflammation signaling. HMGB1 functions in the extra-cellular space as a robust RAGE ligand and inflammatory cytokine or damage-associated molecular pattern molecule (40), suggesting that BER failure and the resulting PARP1 activation may trigger an inflammatory response in tissues with a BER imbalance such as ulcerative colitis (51).

There are multiple PARP1-activation induced cell death mechanisms, as outlined in the diagram shown in Figure 3A. In one, it is suggested that PAR, the product of PARP1 activation, is a cell death molecule. In this process, PAR initiates the translocation of AIF from the mitochondria to the nucleus by a RIP1-dependent mechanism (16-18, 31) (Figure 3A). Uniquely, PAR generated due to BER failure does not appear to trigger cell death via RIP1 activation nor does PAR function as a signal to initiate AIF translocation. PARP1 is involved in many DNA repair processes including homologous recombination (HR) and non-homologous end joining (NHEJ) in response to DSBs and has a role in telomere maintenance (52, 53). The question remains if PAR generated via BER failure is of a unique chemical make-up as compared to PAR generated from DSB-induced PARP1 activation. One possible explanation for the absence of a role for AIF in this study is the concentration of DNA damaging agents used. In this report, we have used TMZ or MMS at a maximum concentration of 1.5 mM or MNNG at a concentration of 5 μM, resulting in 90-95% cell death in the BER deficient cells with little or no cell death in the control cells (Figure S2A; supplemental material). Many reports of PAR-induced AIF translocation include MNNG concentrations of 100 and 500 μM (27, 50, 54). Such high concentrations of DNA damaging gents (e.g., MNNG at 20x and 100x that used herein) have the potential to directly induce DNA DSBs, create overwhelming levels of both nuclear and mitochondrial genome damage (55) as well as the possibility of direct protein alkylation. Regardless, it is clear that the cell death initiated by BER failure is independent of RIP1 activation and AIF translocation, thus ruling out PAR as the cell death signal that is initiated upon BER failure.

One explanation for the absence of PAR-mediated cell death is the rapid catabolism of PAR by PARG (34). In this study, we find that PAR synthesized due to PARP activation is degraded within 90 minutes (Figure 1). As summarized in Figure 3A, the breakdown products of PAR (PAR catabolites) are also likely mediators of cell death, including ADP-ribose (activator of the Ca²⁺ channel TRPM2) and AMP (inhibitor of ATP transport) (19-21). However, PARG knockdown did not reverse the DNA damage-sensitive phenotype of Pol ß deficient cells (Figure 3C), suggesting that damage-dependent cell death in Pol ß deficient cells is not initiated by PAR catabolites. Conversely, the PAR catabolite AMP may provide a protective phenotype by activation of AMPK, induction of autophagy and enhanced ATP synthesis, as recently reported following ROS-induced DNA damage and PARP1 activation (56). Although loss of AMPK activation and induction of autophagy upon PARG-KD could explain, in part, the enhanced cell death observed in the PARG-KD cells (Figure 3C), we suggest this is unlikely, since in this study, autophagy is not involved (Figure S4C) and the activation of AMPK, if any, does not appear to overcome the damage-induced cell death phenotype resulting from BER failure in the PARG proficient cells. Regardless, it is interesting to speculate that PARG may regulate AMPK activation in response to ROS-induced PARP1 activation (56). In all, these studies imply that the alkylation-sensitive phenotype of Pol ß deficient cells is un-related to the accumulation of PAR catabolites such as ADP-ribose or AMP and is likely wholly dependent on the metabolite bioavailability or bioenergetic capacity of the cell.

The over-riding response to the loss of Pol ß and an inability to complete BER (BER failure) is energy failure or depletion of bioenergetic metabolites with no evidence for cell death triggered by PAR or the PAR catabolites ADP-ribose or AMP. The energy collapse or depletion of NAD⁺ and ATP due to BER failure is offset by elevated levels of NMN (46) and is negatively affected by NAD⁺ biosynthesis inhibition (FK-866), suggesting that (i) FK-866 (APO866) and related clinically useful NAD⁺ biosynthesis inhibitors might be combined with TMZ and BER inhibitors to improve TMZ response and (ii) any stress on or defects in the NAD⁺ biosynthesis pathway such as over-activation of SIRT1 (57) or attenuating defects in NAMPT, NMNAT1 or related NAD biosynthetic enzymes (58) may have significant effects on cell survival following BER failure.

Similar phenotypes (stress-induced PARP1 activation and cell survival dependent on NAD⁺ metabolites) have been observed in diverse human cell types and mammalian organ systems, stressing the significance of these findings. PARP1 activation and the resulting “NAD⁺ depletion”-mediated or ATP-depletion mediated cell death plays a critical role in tissue injury from cerebral and myocardial ischemia (59-62). Analogous to the studies described herein, cellular protection from cerebral ischemia is provided by NAD⁺ metabolite supplementation (63, 64). Similarly, streptozotocin-induced diabetes results from PARP1 activation, energy imbalance and cell death dependent on the BER enzyme MPG (65-68). Most importantly, cellular NAD⁺ metabolism plays an essential role in pancreatic ß-cell viability and insulin secretion (69). With the observation that BER failure triggers NAD⁺ depletion, it is interesting to speculate if overall BER capacity controls susceptibility to ischemia or streptozotocin-induced and age-related diabetes onset via neuronal or ß-cell death from loss of bioenergetic metabolites subsequent to BER failure. The onset of these physiologically significant outcomes (stroke, neurodegeneration, ischemia, diabetes) involves PARP1 activation, NAD⁺ depletion and cell death, similar to that reported here. Although a portion of the environmental and endogenous stressors that induce these phenotypes via PARP1 activation will directly induce DNA single-strand breaks, it is reasonable to presume that a significant proportion of cell death related to stroke, retinal degeneration, ischemia and diabetes may initiate from genomic DNA base damage, requiring repair by the BER machinery. As such, the failure to repair the DNA damage and the resulting accumulation of DNA repair intermediates (BER failure) may be the trigger of PARP1 activation and cell death.

In summary, these studies suggest that PARP1 functions as a BER molecular sensor protein to induce caspase-independent cell death following BER failure and provides mechanistic insight into why Pol ß deficiency leads to cell death. Further, we show that the observed DNA damage dependent cell death in Pol ß deficient cells is un-related to the accumulation of PAR catabolites such as ADP-ribose or AMP yet is dependent on NAD⁺ metabolite bioavailability or bioenergetic capacity of the cell, suggesting a linkage between DNA repair capacity, cell survival and cellular bioenergetic metabolites. Finally, these studies have potentially important implications for therapeutic development as it relates to a chemotherapy-induced synthetic lethality approach to cancer therapy involving the combination of a chemotherapeutic DNA damaging agent, a DNA repair inhibitor and a regulator or inhibitor of NAD⁺ biosynthesis.

Materials and Methods

Cell culture and cell line development

The cell line LN428 is an established glioblastoma-derived cell line with mutations in p53, deletions in p14^ARF and p16 and is WT for PTEN (70, 71). LN428 cells were kindly provided by Ian Pollack (University of Pittsburgh) and were cultured in Alpha EMEM supplemented with 10% heat inactivated FBS, glutamine, antibiotic/antimycotic and Gentamycin. MDA-MB-231 cells and derivatives were described previously (25). Human MPG (WT), human MPG (N169D), Flag Pol ß (WT) and Flag Pol ß (K72A) expressing cell lines were developed by transfection using FuGene 6 Transfection Reagent (Roche) according to the manufacturer's protocol. Transfected cell lines were cultured in G418 and/or Puromycin for 2 weeks and individual clones stably expressing human MPG or Pol ß were selected. It was recently suggested that p14^ARF deficiency results in proteosome-mediated degradation of Pol ß (72). Although LN428 cells are deficient in p14^ARF (71), we note that the expression levels of Pol ß are stable. Lentiviral particles were generated by co-transfection of plasmid pCDF1-MCS1-EF1-copGFP (control) or pSIF-H1-hPOLB1-copGFP (Polß shRNA) together with pFIV-34N and pVSV-G into 293-FT cells (73) using FuGene 6 Transfection Reagent (Roche). Forty-eight hours after transfection, lentivirus-containing supernatant was collected and passed through 0.45 μM filters to isolate the viral particles. Lentiviral transduction was performed as described earlier (25). Briefly, 6.0 × 10⁴ cells were seeded into 6-well plate 24 hours before transduction. Cells were transduced for 18 hours at 32°C and cultured for 72 hours at 37°C. Cells expressing copGFP only or both copGFP and Pol ß specific shRNA were isolated by fluorescence activated cell sorting (FACS) and Polß-KD was confirmed by immunoblot analysis. All the cell lines developed and used in this study are described in Table S1 in the supplemental material.

Chemicals and reagents

Alpha EMEM was from Mediatech. RPMI 1640 and DMEM were from Cambrex Bioscience Group and Biowhittaker, respectively. Fetal bovine serum (FBS), heat inactivated FBS, Pen/Strep/Ampho, glutamine and antibiotic/antimycotic were from InVitrogen. Temozolomide (NSC# 362856; IUPAC name: 3-methyl-2-oxo-1, 3, 4, 5, 8-pentazabicyclo[4.3.0]nona-4,6,8-triene-7-carbooxamide; CAS number: 856622-93-1) (74) was obtained from the National Cancer Institute Developmental Therapeutics Program. A temozolomide (TMZ) stock solution was prepared in DMSO at 100 mM. MMS was from Sigma-Aldrich. Puromycin, Gentamicin and Neomycin were purchased from Clontech Laboratories, Irvine Scientific and InVitrogen, respectively. PJ34 was purchased from Calbiochem. FK-866 [(E)-[4-(1-Benzoylpiperidin-4-yl)butyl]-3-pyridin-3-yl)acrylamide] was obtained from the National Institute of Mental Health Chemical Synthesis and Drug Supply Program (NIMH# F-901) (47). ß-nicotinamide mononucleotide (NMN) was obtained from Sigma (cat# N3501) and Nicotinic Acid (NA) was obtained from Fisher (cat# AC12829).

Plasmid expression and RNAi vectors

Human MPG (WT) was expressed using the plasmid pRS1422, as described previously (25). The MPG expression plasmid (pRS1422) was then mutated at residue N169 using the Quickchange XL Site-Directed Mutagenesis Kit (Stratagene) to yield pIRES-Neo-MPG-N169D. The expression plasmid for FLAG-tagged WT human Pol ß was generated by PCR amplification of the human Pol ß cDNA using a FLAG-containing forward oligonucleotide and cloned into pENTR/D-TOPO as we described previously (25). pENTER/Flag-Pol ß(WT) was then mutated at residue K72 as described above to yield pENTER/Flag-Pol ß(K72A). Flag-Pol ß(WT) and Flag-Pol ß(K72A) was subsequently cloned into a Gateway-modifed pIRES-Puro vector by TOPO cloning, as we have described previously (25). The FIV-based lentiviral shRNA expression vector system specific for human Pol ß was as described previously (25) but was modified for copGFP expression (pSIF-H1-hPOLB1-copGFP). Lentiviral particles for co-expression of PARG shRNA and TurboGFP was prepared by transfection of four plasmids (the control plasmid pLK0.1-Puro-tGFP or the human PARG-specific shRNA plasmid pLK0.1-Puro-PARGshRNA4, plus pMD2.g(VSVG), pRSV-REV and pMDLg/pRRE) into 293-FT cells (73, 75) using FuGene 6 transfection reagent (Roche Diagnostic Corp, Indianapolis, IN). Culture media from transfected cells was collected 48 hours after transfection to isolate the viral particles, passed through 0.45 μm filters, used immediately or stored at -80°C in single-use aliquots. Transduction of LN428 cells with control lentivirus (GFP expression only) and human PARG specific shRNA lentivirus was completed as follows: Briefly, 6.0 × 10⁴ cells were seeded into 6-well plates and incubated for 24-30 hours at 10% CO₂ at 37°C. Cells were transduced for 18 hours with virus at 32°C and cultured for 72 hours at 37°C before isolation of the GFP-expressing population by fluorescence-activated cell sorting (FACS) using the UPCI Flow Cytometry Facility. Cells were then cultured to expand the population and analyzed for expression of PARG by qRT-PCR.

Quantitative RT-PCR Analysis

Expression of PARG and Pol ß mRNA was measured by quantitative RT-PCR using an Applied Biosystems StepOnePlus system. Briefly, 80,000 cells were lysed and reverse transcribed using the Applied Biosystems Taqman® Gene Expression Cells-to-CT™ Kit. Each sample was analyzed in triplicate and the results are an average of all three analyses. Analysis of mRNA expression was conducted as per the manufacturer (ΔΔC_T method) using Applied Biosystems TaqMan® Gene Expression Assays (human Pol ß: Hs00160263_m1; human PARG: Hs00608254_m1) and normalized to the expression of human ß-actin (part #4333762T).

Cell extract preparation and Western blot

Nuclear extracts were prepared and protein concentration was determined as we described previously (12). Twenty microgram of protein was loaded on a pre-cast 4-20% NuPAGE Tris-Glycine gel (InVitrogen). For whole cell extracts used in poly(ADP-ribose) (PAR) formation assays, 3 × 10⁶ cells were seeded into a 100 mm cell culture dish 24 hours before drug treatment. Cells were either treated with TMZ only or pre-exposed to a PARP inhibitor (PJ34 or DR2313) followed by PARP inhibitor plus TMZ treatment. After treatment, cells were washed twice with cold PBS, collected and lysed with 400 μL of 2 × Laemmli buffer (2% SDS, 20% Glycerol, 62.5 mM Tris-HCl pH6.8, 0.01% Bromophenol Blue). Samples were boiled for 8 min and extract from approximately 1.5 × 10⁵ cells were loaded each lane on a 4-12% pre-cast NuPAGE Tris-Glycine gel (InVitrogen) for immunoblot assay.

The following primary antibodies were used in immunoblot assays: anti-human MPG (Mab; clone 506-3D) (25); anti-Pol ß (Mab clone 61; Thermo Fisher Scientific); anti-APE1 (EMD Biosciences); anti-PCNA (Santa Cruz); anti-Flag (M2 Mab; Sigma-Aldrich); anti-poly(ADP-ribse) (PAR) (Clone 10H, kindly provided by M. Ziegler); anti-poly(ADP-ribose) polymerase-1 (PARP1) (BD Pharmingen) and anti-human HMGB1 (R&D Systems).

Cell cytotoxicity assay

TMZ induced cytotoxicity was determined by an MTS assay, a modified MTT assay as described previously (12). Results were calculated from the average of three or four separate experiments and are reported as the percentage of treated cells relative to the cells without treatment (% Control). For PJ34, cells were pre-exposed to the inhibitor for 30 minutes and were then treated with TMZ in the presence of the inhibitor for 48 hours. For NA and NMN, cells were pre-exposed to each for 24 hours (concentrations as indicated in the legend) and were then treated with TMZ (1.0 mM) in the presence of NA or NMN for 48 hours. The impact on cell growth and survival was determined by an MTS assay, as described previously (12).

HMGB1 release assay

Cells were pre-treated with media alone or with PARP inhibitor (PJ34) for 30 min before co-treatment with PJ34 (2 μM) and TMZ (1.5 mM) for 12 hours. Cell culture media was then collected and passed through 0.45 μM filters. 100 μL of immoblized Heparin (Thermo Fisher Scientific) slurry and 1 mL of media was mixed and rotated at 4°C for 2 hours before centrifugation at 8,000g to pull-down HMGB1 bound to immobilized Heparin (76). Pellets were boiled with 100 μL of 2 × Laemmli buffer and supernatants were used for immunoblot assay after brief centrifugation.

PAR assay

Cells (1.5 × 10⁶) where seeded in 100mm dishes 24 hours before treatment. For the FK-866 experiments, cells were then incubated in the presence of FK-866 (10 nM) or DMSO for an additional 24 hours. Media was then removed and replaced with fresh media or media supplemented with PJ34 (2 μM). After 30 minutes, cells were lysed immediately (0 time point) or media was replaced with TMZ for the times indicated in the figure legends. Extracts were prepared by washing the cells with PBS and preparing cell extract with 400 μl of 2X Laemmli Buffer. 20 μl of the cell extract was analyzed by immunoblot with a 4000-fold dilution of an anti-PAR primary antibody (Clone 10H) followed by a 5000-fold dilution of the horseradish peroxidase (HRP)-conjugated secondary goat anti-mouse Ab.

Immunofluorescence and confocal microscopy

Cells were cultured on glass coverslips for 24 hours prior to treatment with MMS or media control. One-hour post treatment cells were washed and allowed to recover in media for 5 hours. Cells were then fixed with 4% paraformaldehyde for 20 minutes, permeabilized with 0.5% Triton X-100 for 15 minutes, and blocked with 2% BSA for 1 hour, all at room temperature. AIF was detected by incubating 1 hour at room temperature with an anti-AIF antibody (Santa Cruz) at 1:100 dilution, followed by goat anti-mouse Alexa 488 (Molecular Probes) at 1:500, Alexa 647 phalloidin actin stain at 1:250 (Molecular Probes), and 5 μM DRAQ5 nuclear stain for 1 hour at room temperature. Slides were mounted and imaged on the Olympus fluoview 500 confocal microscope.

NAD⁺ and ATP measurements

Cells were seeded 24 hours prior to treatment with MMS or media control. NAD⁺: One-hour post treatment cells were trypsinized, counted and 1×10⁵ cells were pelleted. NAD⁺ lysates were prepared and NAD⁺ measurements obtained using the Enzychrom™ NAD⁺/NADH assay kit (BioAssay Systems). ATP: One-hour post treatment cells were washed and allowed to recover in normal media for one hour. Cells were then lysed and ATP content measured by luminescent output using the ATP-lite assay kit (PerkinElmer).

FK-866 cytotoxicity assay

Cells were seeded in 96 well plates 24 hours prior to treatment. Cells were pre-treated with 10nM FK-866 or DMSO control for 24 hours and then exposed to MMS for 1 hour. The cells were then washed with media and allowed to recover for 48 hours prior to assaying for cytotoxicity by an MTS assay previously described (12, 25). Results shown are the average of three independent experiments and reported as percent survival of MMS treated cells compared to control wells.