ER-associated degradation ligase HRD1 links ER stress to DNA damage repair by modulating the activity of DNA-PKcs

Significance

The co-occurrence of endoplasmic reticulum (ER) stress and DNA damage is common in cancer. However, the cross talk between ER–associated protein degradation (ERAD) and DNA damage response (DDR) remains underexplored. Here, we reveal a different mechanism by which ER-resident ERAD ligase HRD1 transduces ER stress signals to the nucleus by catalyzing the ubiquitination of HDAC1 in the cytoplasm and then preventing the HDAC1-mediated deacetylation of two nuclear proteins: KU70 and KU80. The acetylation sites in KU70/KU80 provide better binding for E3 ligase TRIM25. As such, increased HDAC1 degradation results in increased acetylation and TRIM25-mediated polyubiquitination and degradation of KU70/KU80. Inhibition of HDAC1 impairs KU70/KU80/DNA-dependent protein kinase catalytic subunit (DNA-PKcs)-regulated DDR and then improves ER stress inducers’ anticancer efficiency.

Abstract

Proteostasis and genomic integrity are respectively regulated by the endoplasmic reticulum–associated protein degradation (ERAD) and DNA damage repair signaling pathways, with both pathways essential for carcinogenesis and drug resistance. How these signaling pathways coordinate with each other remains unexplored. We found that ER stress specifically induces the DNA-PKcs-regulated nonhomologous end joining (NHEJ) pathway to amend DNA damage and impede cell death. Intriguingly, sustained ER stress rapidly decreased the activity of DNA-PKcs and DNA damage accumulated, facilitating a switch from adaptation to cell death. This DNA-PKcs inactivation was caused by increased KU70/KU80 protein degradation. Unexpectedly, the ERAD ligase HRD1 was found to efficiently destabilize the classic nuclear protein HDAC1 in the cytoplasm, by catalyzing HDAC1’s polyubiquitination at lysine 74, at a late stage of ER stress. By abolishing HDAC1-mediated KU70/KU80 deacetylation, HRD1 transmits ER signals to the nucleus. The resulting enhanced KU70/KU80 acetylation provides binding sites for the nuclear E3 ligase TRIM25, resulting in the promotion of polyubiquitination and the degradation of KU70/KU80 proteins. Both in vitro and in vivo cancer models showed that genetic or pharmacological inhibition of HADC1 or DNA-PKcs sensitizes colon cancer cells to ER stress inducers, including the Food and Drug Administration–approved drug celecoxib. The antitumor effects of the combined approach were also observed in patient-derived xenograft models. These findings identify a mechanistic link between ER stress (ERAD) in the cytoplasm and DNA damage (NHEJ) pathways in the nucleus, indicating that combined anticancer strategies may be developed that induce severe ER stress while simultaneously inhibiting KU70/KU80/DNA-PKcs-mediated NHEJ signaling.

Proteostasis and genomic integrity are both essential in maintaining cellular homeostasis (1, 2). The endoplasmic reticulum (ER) is an organelle that is involved in protein synthesis, folding, modification, trafficking, and secretion and thus plays an essential role in proteostasis (3, 4). Cancer cells have intrinsically high levels of ER stress caused by their dysregulated microenvironment, increased cell proliferation, metabolic reprogramming, and the activation of oncogenic signaling (5, 6). While unresolved ER stress leads to cell death, cells have evolved two key protein quality control pathways to deal with high levels of ER stress and restore homeostasis, called ER-associated protein degradation (ERAD) and the unfolded protein response (UPR) (7–9). The UPR signaling pathway comprises three major branches, each initiated by the activation of inositol-requiring protein 1 (IRE1), protein kinase RNA-like ER kinase (PERK), or activating transcription factor 6 (10). This allows the cell to transduce ER stress signals to the cytoplasm or nucleus by either transcriptional or nontranscriptional mechanisms. The UPR has been shown to play an important role in many biological processes and pathological conditions (10, 11); however, the importance of ERAD has only recently begun to be appreciated. ERAD is a multistep process that involves recognition, retrotranslocation, and degradation. It monitors and controls protein quality and quantity in the ER by removing misfolded proteins. An increasing number of ERAD ligases have been identified, of which HMG-CoA reductase degradation protein 1 (HRD1) is the most conserved and best characterized (12, 13). In addition to catalyzing the proteasome-mediated degradation of misfolded proteins in the ER (14), HRD1 can also act in a context-dependent manner by selectively catalyzing different substrates in the cytoplasm (15). As such, loss of HRD1 has been shown to result in a variety of biological outcomes, including embryonic lethality in mice (16, 17), impaired T cell priming and B cell immunity (18–20), and liver cirrhosis (21). However, it remains unknown whether HRD1 is involved in the regulation of the DNA damage response (DDR).

Genomic instability is a well-established hallmark of cancer cells (22). Oncogene-induced replication stress and metabolic reprogramming inevitably lead to increased levels of DNA damage (23–25). The subsequent activation of the DDR is coordinated with oncogenic signaling, leading to increased cell viability. As such, the DDR is a vulnerability of cancer cells that can be efficiently targeted and is an effective way to increase the sensitivity of cancer cells to genotoxic anticancer therapies. There are three major DNA damage sensor kinases, all of which are phosphoinositide 3 kinase–related protein kinases (PI3KKs): ataxia telangiectasia mutated (ATM), ataxia telangiectasia and Rad3-related (ATR), and DNA-dependent protein kinase (DNA-PK) (26). They lead to distinctive DDR pathways in response to different types of DNA damage. ATR is normally activated by single-strand DNA breaks (27), while ATM and DNA-PK are activated by DNA double-strand breaks (DSBs), the most toxic form of DNA damage (28, 29). When DSBs occur, KU70/KU80 or the MRN complex (consisting of Mre11, Rad50, and Nbs1) detect the damage and recruit DNA-PK or ATM to the damage sites, respectively, leading to the activation of nonhomologous end joining (NHEJ) or homologous recombination (HR) (30, 31). Activation of both pathways attenuates DNA damage–induced cell death, making them therapeutic targets in the context of cancer.

The co-occurrence of ER stress and DNA damage is common in many human diseases, including cancer (1). Intrinsic genomic instability or genotoxic drug-induced DNA damage in cancer cells can increase the number of misfolded proteins in the ER, leading to increased ER stress (32–34). Conversely, ER stress can promote the production of reactive oxygen species, a primary source of DNA damage (35, 36). It should be noted that although several lines of evidence point to a close relationship and complicated cross talk between the UPR and the DDR (1), the specific molecular relationship between ERAD and the DDR has not been investigated. In addition, the effects of both ER stress activators (an emerging cancer treatment) and DNA damage inducers (conventional cancer treatments) are affected by the development of drug resistance (37–40). Thus, understanding the cross talk between these two important biological processes may provide unique insights into combination therapy options that can improve treatment efficacy.

Here, we explore the dynamic changes in DDR signaling that take place after ER stress is triggered and describe hitherto unknown signaling from the ERAD enzyme HRD1 in the ER to histone deacetylase 1 (HDAC1) in the nucleus. HRD1-mediated HDAC1 degradation results in increased KU70/KU80 acetylation, which provides binding sites for TRIM25, leading to increased KU70/KU80 degradation and impaired DNA-PKcs activation. Our data demonstrate that this signaling from the ER to the nucleus plays an essential role in the resolution of cell fate in response to ER stress stimulation. Combined blockade of HDAC1 or the DNA-PKcs-regulated NHEJ pathway significantly improves the responses of cancer cells to a Food and Drug Administration (FDA)-approved inducer of ER stress.

Results

Selective Activation of the DNA-PKcs-NHEJ Pathway by ER Stress Controls Cell Fate.

We performed pathway enrichment analysis of the differentially expressed genes (DEGs) after thapsigargin (TG), brefeldin A, or celecoxib-induced ER stress in the GEO datasets. The ER stress–associated DEGs were plotted in a Venn diagram (Fig. 1A). The GO enrichment analysis results indicated that DEGs were significantly enriched in pathways such as DNA replication, response to ER stress, regulation of DNA replication, NHEJ, and DNA strand elongation involved in DNA replication (Fig. 1A). These data suggest that ER stress may cross talk with the DNA repair response pathways. To explore this, we treated cancer cells with ER stress inducers with the well-established DNA damage inducer doxorubicin used as a control. Activation of the UPR was confirmed by the increased expression of GRP78, p-IRE1α, p-PERK, and ATF4 (SI Appendix, Fig. S1A). Consistent with previous reports (41), cancer cells incurred DNA damage under ER stress conditions, as indicated by increased levels of γH2AX (SI Appendix, Fig. S1A). ER stress–induced DNA damage was not a cell type–specific event, because a similar phenomenon was observed in cell lines of different tissue origins (SI Appendix, Fig. S1A). We further analyzed colon cancer samples from TCGA. The results indicated that ER stress levels are elevated in cancer, and there is a positive correlation between response to ER stress and DNA repair (Fig. 1 B and C). In addition, TG, a representative ER stress inducer, was found to increase DNA damage in a time- and dose-dependent manner, as shown by western blot analysis, cell staining of γH2AX (Fig. 1 D and E and SI Appendix, Fig. S1 B and C), and also by the comet assay (Fig. 1F and SI Appendix, Fig. S1D).

Fig. 1.

Selective activation of the DNA-PKcs-NHEJ pathway controls cell fate under ER stress. (A) The ER stress–associated DEGs were plotted in a Venn diagram (Gene expression profiles were obtained from GEO datasets GSE200454, GSE1622 56, and GSE169584). The GO enrichment analysis results of DEGs were plotted in the bubble chart. (B) The level of ER stress in tumor and normal adjacent tissue of colon cancer was analyzed (data are obtained from TCGA datasets). (C) The correlation between the pathways response to ER stress pathway and DNA repair in colon cancer (data are obtained from TCGA datasets). (D) The expression levels of URP and DDR proteins were determined by western blots in HCT116 or LoVo cells after TG (1 μg/mL) treatment for the indicated time periods. (E) The γH2AX foci were determined by cell staining in HCT116 or LoVo cells after TG (1 μg/mL) treatment for the indicated time periods. (F) Comet assay used to detect DNA damage in HCT116/LoVo cells after TG (1 μg/mL) treatment for the indicated time periods (Up). The tail moment of comets was calculated using OpenComet software (Bottom). (G) HR or NHEJ efficiencies of ISce-I–induced DSBs in U2OS cells expression EJ5-GFP reporter (NHEJ, Left) or DR-GFP reporter (HR, Right) were determined by measuring GFP-positive cells by flow cytometry (FACS) after TG (1 μg/mL) treatment for the indicated time periods in the presence or absence of KU-57788 (1 μM, 8 h). (H) Representative crystal violet staining images of HCT116 and LoVo cells after TG (1 μg/mL) treatment for the indicated time periods in the presence or absence of KU-57788 (1 μM, 8 h), AZ20 (10 μM, 8 h), or KU-60019 (1 μM, 8 h) (Left). The images were quantified by ImageJ and the data were plotted in the graphs (Right). Data were derived from three independent experiments and represented as mean ± SEM in the bar graph. **P < 0.01, ##P < 0.01 (F–H).

We then investigated whether and which DDR pathway is activated under ER stress. To this end, the activities of the three key kinases in the DNA damage pathways, ATM, ATR, and DNA-PKcs, were measured. Interestingly, DNA-PKcs, but not ATM and ATR, was specifically and significantly activated 12 to 18 h after exposure to TG, as indicated by increased levels of phospho-DNA-PKcs at S2056 (p-DNA-PKcs) and increased expression of its downstream factors XRCC4 and Lig4 (Fig. 1D). Notably, although DNA damage was continuously increased after induction of ER stress, activation of the DNA-PKcs pathway was rapidly reduced after 24 h (Fig. 1D). DNA-PKcs is mainly involved in the regulation of the NHEJ pathway following DSBs (42). Consistently, ER stress activated the NHEJ pathway but not the HR pathway, as indicated by green fluorescent protein (GFP) signals obtained with the EJ5-GFP and DR-GFP systems, in a manner coordinated with the dynamic changes in DNA-PKcs activity (Fig. 1G). Inhibition of DNA-PKcs by the DNA-PK inhibitor KU-57788 completely abolished the ER stress–induced NHEJ pathway (Fig. 1G). These data suggest that ER stress promotes DNA-PKcs activity and subsequent activation of the NHEJ pathway.

Interestingly, our data showed that the activity of the DNA-PKcs-NHEJ pathway was rapidly decreased at a late stage of ER stress, which was accompanied by an increase in cell death (Fig. 1 G and H and SI Appendix, Fig. S2A). DNA-PK inhibitors or small interfering RNA (siRNA)-mediated genetic knockdown (KD) of DNA-PKcs significantly increased DNA damage levels (right lanes, SI Appendix, Fig. S2 B and C) and also accelerated ER stress–induced cell death (Fig. 1H and SI Appendix, Fig. S2A). In contrast, inhibition of ATM or ATR had no apparent effect on either DNA damage levels or cell death (Fig. 1H and SI Appendix, Fig. S2D). These data suggest that inactivation of the DNA-PKcs-NHEJ pathway controls the adaptation–death switch under ER stress.

Dynamic Changes in KU70/KU80 Expression Determine the Activity of DNA-PKcs during ER Stress.

Given the essential roles of DNA-PKcs signaling in controlling cell fate, we explored the mechanisms underlying DNA-PKcs inactivation with prolonged ER stress. It is known that KU70 and KU80 recognize damaged DNA and recruit DNA-PKcs to damage sites to trigger kinase cascades, leading to DNA repair by NHEJ. Consistent with this, levels of KU70/KU80 foci were increased at 12 to 18 h and declined after 24 h with prolonged ER stress (Fig. 2A), which coincided with the observed dynamic changes in DNA-PKcs activity and NHEJ activation. Western blot analysis demonstrated similar kinetics of KU70/KU80 expression (Fig. 2B). Recovery of the expression of KU70 and KU80 completely abrogated the effects of ER stress on DNA-PKcs activity (Fig. 2C) and cell death (Fig. 2D). However, such treatment did not have obvious effects on cell death in DNA-PK inhibitor–treated cells under ER stress (Fig. 2D). These data suggested that dynamic changes in KU70/KU80 protein expression determine DNA-PKcs activity and thus the sensitivity of cells to ER stress–induced cell death.

Fig. 2.

Dynamic changes of KU70/KU80 expression determine the activity of DNA-PKcs during ER stress. (A and B) KU70/KU80 foci (A) and their expression levels (B) were determined by cell staining and western blot, receptively, in HCT116 and LoVo cells treated with TG for different time periods. The images were quantified by ImageJ and the data were plotted in the bar graphs. (C) p-DNA-PKcs, KU70, and KU80 protein levels were determined by western blot in HCT116 and LoVo cells after KU70/KU80 OE and/or TG (1 μg/mL) treatment for the indicated time periods. GAPDH was used as a loading control. (D) Representative crystal violet staining images of HCT116 and LoVo cells with the indicated treatments (Up). The images were quantified by ImageJ and the data were plotted in the graphs (Bottom). Data are derived from three independent experiments and represented as mean ± SEM in the bar graph. **P < 0.01, ##P < 0.01 (A, B, and D).

ERAD Ligase HRD1 Regulates KU70/KU80 Protein Turnover by an Indirect Mechanism Involving HDAC1.

We then investigated how KU70/KU80 are repressed at a late stage of ER stress. Although their protein levels were altered, the mRNA levels of both the KU70 and KU80 genes were comparable during ER stress (SI Appendix, Fig. S3A). Treatment with the proteasome inhibitor MG132, but not chloroquine (CQ), increased KU70/KU80 protein levels and completely abolished long-term ER stress–induced KU70/KU80 downregulation (SI Appendix, Fig. S3B). These data suggest that KU70 and KU80 are repressed mainly by proteasome-mediated protein degradation under prolonged ER stress.

Because KU70 and KU80 expressions were changed with similar kinetics under ER stress, we applied bioinformatics tools to predict the E3 ligase that acts on both targets (SI Appendix, Fig. S3C). As such, nine E3 kinases were expected to possibly catalyze the ubiquitination of both KU70 and KU80 at high confidence levels. siRNA-mediated genetic KD of individual E3 ligase genes revealed that HRD1 and TRIM25 were the most likely candidates, as KD of either HRD1 or TRIM25 increased the levels of KU70/KU80, while KD of the other seven E3 ligases (SMURF1, NEDD4L, UBE4B, UBE4A, BRCC3, RNF8, and RNF40) had little effect (SI Appendix, Fig. S3D). We first focused our attention on the roles of HRD1 because it is a well-known ERAD ligase (19). Its effects on the expression levels of KU70/KU80 were validated by treating cells with another independent siRNA-HRD1 (Fig. 3A). HRD1 overexpression (OE) was found to repress KU70/KU80 expression (Fig. 3A). Importantly, ER stress–induced KU70/KU80 inhibition was completely abolished by HRD1 KD (Fig. 3B). We then reasoned that KU70/KU80 may be a substrate of HRD1; however, there was no detectable binding between HRD1 and KU70/KU80 (Fig. 3C). So this is unlikely. Thus, although HRD1 is required for ER stress–induced KU70/KU80 degradation, it acts in an indirect manner.

Fig. 3.

ERAD ligase HRD1 regulates KU70/KU80 protein turnover by an indirect mechanism involved HDAC1. (A and B) Representative western blot images of HRD1, KU70, and KU80 protein in HRD1 KD or OE cells (A) in the presence or absence of TG (1 μg/mL) for the indicated time durations (B). (C) The binding between KU70 or KU80 and HRD1 was determined by an in vitro IP assay using purified proteins. (D) Mass spectrum analysis of HRD1 binding proteins following proximity labeling of HRD1 under control and ER-stressed conditions, in the presence of MG132. (E and F) Representative western blot images of HRD1 and HDAC1 protein in HRD1 KD (E) or OE cells (F) in the presence or absence of TG (1 μg/mL) for the indicated time durations. (G) Representative western blot images of HRD1, HDAC1, KU70, and KU80 protein in HDAC1 KD cells or HDAC1 inhibitor SAHA (1 μM) treated cells in the presence or absence of TG (1 μg/mL) for the indicated time durations. β-actin was used as a loading control. (H) Schematic diagram of HRD1-mediated KU70/KU80 protein degradation through binding and inhibiting the protein levels of HDAC1.

To reveal how ERAD ligases are involved in nuclear protein KU70/KU80 degradation, we performed mass spectrometry analysis following proximity labeling of HRD1 under control and ER-stressed conditions, in the presence of MG132 (Fig. 3D). Interestingly, HDAC1 was a top hit of HRD1 binding partners under ER-stressed conditions but not under control conditions (Fig. 3D).

Because protein acetylation has been shown to either promote or inhibit protein stability, we wondered whether HRD1 regulates KU70/KU80 expression through its binding to and subsequent inhibition of HDAC1. Our data showed that HDAC1 was dramatically reduced 24 h after TG exposure (SI Appendix, Fig. S3E), and that recovery of HDAC1 expression abolished ER stress–induced KU70/KU80 degradation (SI Appendix, Fig. S3F). In addition, HRD1 KD rescued ER stress–induced HDAC1 inhibition while HRD1 OE promoted it (Fig. 3 E and F). Furthermore, in HDAC1 KD cells or cells treated with the HDAC1 inhibitor suberoylanilide hydroxamic acid (SAHA), HRD1 KD failed to rescue KU70/KU80 expression under ER stress (Fig. 3G). Moreover, KU70/KU80 OE abolished the effect of SAHA (SI Appendix, Fig. S3G). These data suggest that HRD1 inhibits KU70/KU80 expression by repressing the activity of HDAC1 (Fig. 3H).

HRD1 Catalyzes HDAC1 Ubiquitination and Degradation under ER Stress.

It should be noted that HRD1 is a well-known ER-resident E3 ligase, while HDAC1 is generally considered to be a nuclear protein. How does HRD1 transmit ER stress signals to the nucleus by inhibiting HDAC1? To explore this, we examined the localization and interaction of HRD1 and HDAC1 under different conditions. The time-course assay shows that although HRD1 and HDAC1 levels, respectively, increased 6 h and 12 h after induction of ER stress, their binding was increased at 24 h in the presence of MG132 (SI Appendix, Fig. S3E and S4A), which is consistent with the mass spectrometry screening results. This finding was further validated by the proximity ligation assay (Fig. 4A). In addition, we noted that the HRD1/HDAC1 proximity ligation assay signal was predominately located in the cytoplasm at perinuclear regions (Fig. 4A). Cell fractionation combined with an immunoprecipitation (IP) assay validated the above data, showing that HRD1 and HDAC1 mainly interacted with each other in the cytoplasm, but not in the nucleus (Fig. 4B). Despite known as ER resident protein, HRD1 was localized in both the nucleus and the cytoplasm (Fig. 4C and SI Appendix, Fig. S4B). The conventional nuclear protein HDAC1 accumulated in both the nucleus and the cytoplasm under ER stress in the presence of MG132 (Fig. 4C). These data suggested that although HRD1 can be present in both the nucleus and the cytoplasm, it binds with HDAC1 in the cytoplasm at a late stage of ER stress after proteasome degradation was blocked by treatment with MG132.

Fig. 4.

HRD1 catalyzes HDAC1 ubiquitination and degradation under ER stress. (A and B) The bindings between HRD1 and HDAC1 in living cells or fractioned cell lysates were determined by the PLA (A) and IP assay (B) in the presence or absence of TG (1 μg/mL, 24 h) and/or MG132 (20 μM, 6 h). (C) Analysis of the distribution of HDAC1 and HRD1 by the cell fractionation assay in the presence or absence of TG (1 μg/mL, 24 h) and/or MG132 (20 μM,6 h). (D) The HRD1/HDAC1 interaction and the binding mechanisms were determined by the NanoBiT proximity assay (Bottom). The diagram illustrated the domain structures of HRD1/HDAC1 and the principles of NanoBit proximity assay action (Up); FL, full-length; F, fragment. (E and F) Ub-HDAC1 was determined using IP/western blot assay in HCT116 cells with HRD1 KD (E), WT HRD1 OE or E3 ligase-defective mutant (MT, C291A/C309S/C329S) OE (F). (G) Ub-WT-HDAC1 or Ub-HDCA1 mutants at the predicted ubiquitination sites were determined by the IP/western blot assay in HRD1-V5 OE HCT116 cells. (H) The expression levels of WT HDAC1 or HDCA1 mutants at the predicted ubiquitination sites were determined by the western blot assay in the presence of TG (1 μg/mL, 24 h). (I) Schematic diagram showing that HRD1 binds with HDAC1 at its protein interaction domain (F1) through the C terminus in the cytoplasm and catalyzes HDAC1 polyubiquitination at residue K74 to induce HDAC1 protein degradation. Data are derived from three independent experiments and represented as mean ± SEM in the bar graph. **P < 0.01 (A and D).

We then used in vitro IP and Nanobit assays to investigate the HRD1/HDAC1 binding mechanisms. Both assays showed that HDAC1 and HRD1 directly interact with each other through the N terminus (Fragment 1, F1) of HDAC1 and C-terminal structures of HRD1 (F2, Fig. 4D and SI Appendix, Fig. S4 C and D). Polyubiquitination of HDAC1 was subsequently detected and was found to be mediated by K48-linked ubiquitination (SI Appendix, Fig. S4E). Inhibition of HRD1 expression decreased K48-linked HDAC1 polyubiquitination (Fig. 4E), while OE of wild-type (WT) HRD1, but not of kinase-dead HRD1, increased it (Fig. 4F). The deubiquitylating enzyme Usp2-cc abolished the effect of WT HRD1 (SI Appendix, Fig. S4F). We then utilized bioinformatics tools to predict the ubiquitination sites in HDAC1. The eight most likely lysine (K) ubiquitination sites were subsequently replaced with arginine (R) to generate ubiquitination-defective mutants. Our results showed that a single mutation at K74 completely abolished HDAC1 ubiquitination (Fig. 4G). WT HDAC1, but not HDAC1-K74R, was repressed under long-term exposure to TG or HRD1 OE (Fig. 4 G and H). The results indicate that the ERAD ligase HRD1 mainly binds with HDAC1 in the cytoplasm and catalyzes HDAC1 ubiquitination at residue K74, resulting in increased HDAC1 protein turnover (Fig. 4I).

HDAC1 Protects KU70/KU80 Protein from TRIM25-Mediated Protein Degradation.

We further investigated how HRD1/HDAC1 signaling leads to KU70/KU80 degradation under prolonged ER stress. It should be noted that in addition to HRD1, TRIM25 is another E3 ligase candidate that may catalyze KU70/KU80 protein degradation (SI Appendix, Fig. S3D). Whether TRIM25 has an inhibitory effect on KU70/KU80 was validated using an independent siRNA that specifically targets TRIM25 under both basal and ER-stressed conditions (Fig. 5A). Unlike HRD1, TRIM25 formed complexes with both KU70 and KU80 (Fig. 5B). In addition, the levels of the complexes were significantly increased 24 h after ER stress was triggered (Fig. 5B). Direct binding between KU70/TRIM25 and KU80/TRIM25 was confirmed in vitro using purified proteins (SI Appendix, Fig. S5 A and B). In vitro IP and Nanobit assays further demonstrated how they interact: TRIM25 binds with KU70 at its core domain and binds with KU80 at its vWA domain through its Ring domain (Fig. 5C and SI Appendix, Fig. S5 A and B). Ubiquitination levels of KU70/KU80 were consistently lower in TRIM25 KD cells than in control cells (SI Appendix, Fig. S5C). TRIM25 KD also abolished ER stress–induced KU70/KU80 downregulation and ubiquitination (SI Appendix, Fig. S5D).

Fig. 5.

HDAC1 protects KU70/U80 protein from TRIM25-mediated protein degradation. (A) Representative western blot images of KU70, KU80, and TRIM25 protein in the control or TRIM25 KD cells before and after TG (1 μg/mL) treatment in HCT116 cells. β-actin was used as a loading control. (B) The interactions between KU70 or KU80 and TRIM25 were analyzed by IP assays in HCT116 cells after TG (1 μg/mL) treatment for the indicated time durations. (C) The interaction between KU70 or KU80 and TRIM25 and their binding mechanisms were evaluated by the NanoBiT proximity assay. Data are derived from three independent experiments and represented as mean ± SEM in the bar graph. **P < 0.01. (D) Poly-Ub of WT and MT KU70/KU80 were determined by the IP/western blot assay in the control and/or TRIM25 OE HCT116 cells. (E and F) The expression levels of WT KU70/KU80 and ubiquitination defective mutant (K114/256R KU70 and K195/265/481R KU80) were determined by western blot in TRIM25 KD (E), or TG (1 μg/mL) treated (F) HCT116 cells. β-actin was used as a loading control. (G) Endogenous levels of KU70/KU80 and TRIM25 were determined by western blot in HCT116 cells in the presence or absence of SAHA (1 μM, 24 h) in TRIM25 KD cells. GAPDH was used as a loading control.

Changes in KU70 and KU80 ubiquitination levels have been reported previously in a mass spectrum analysis (43). Intriguingly, K114/256R-KU70 and K195/265/481R-KU80 mutations abolished the TRIM25-mediated ubiquitination of KU70 and KU80, respectively (Fig. 5D). WT KU70/KU80 levels were consistently higher in TRIM25 KD cells compared with control cells, while mutant levels were comparable regardless of TRIM25 expression levels (Fig. 5E). In addition, WT KU70/KU80 levels decreased when long-term ER stress was stimulated (24 h), an effect that was abrogated by the K114/256R-KU70 and K195/265/481R-KU80 mutations (Fig. 5F). We therefore conclude that TRIM25 catalyzes the ubiquitination of KU70 at K114/256 and of KU80 at K195/265/481, leading to their increased degradation under prolonged ER stress, and postulate that HDAC1 may regulate KU70/KU80 by inhibiting TRIM25. In support of this idea, HDAC1 inhibitor SAHA did not reduce KU70/KU80 levels in TRIM25 KD cells (Fig. 5G).

HDAC1-Mediated KU70/KU80 Deacetylation Impaired KU70/KU80’s Ability to Bind with TRIM25.

The remaining question is whether HDAC1 attenuates the degradation of KU70/KU80 by inhibiting the acetylation of TRIM25 or KU70/KU80 themselves. We showed that the acetylation levels of KU70/KU80, but not of TRIM25, were increased by treatment with the HDAC inhibitor SAHA (SI Appendix, Fig. S6 A and B). In addition, KU70/80 acetylation increased 24 h after ER stress was triggered, and this effect was completely abolished by the recovery of ER stress–repressed HDAC1 expression (Fig. 6 A, Upper). Cell fractionation combined with an IP assay confirmed that HDAC1 and KU70/KU80 were predominately localized to the nucleus, where acetylated-KU70KU80 expressed (SI Appendix, Fig. S6C). Consistent with the above data (Fig. 5G), TRIM25 KD completely abolished SAHA-induced KU70/KU80 ubiquitination and degradation (Fig. 6B), while it had no obvious effect on SAHA-regulated KU70/KU80 acetylation (lane 3 vs. 4, Fig. 6 C, Upper). Therefore, HDAC1 influences TRIM25-mediated KU70/KU80 degradation by modulating the acetylation of KU70/KU80, but not of TRIM25.

Fig. 6.

HDAC1-mediated KU70/KU80 deacetylation impaired KU70 /KU80’s binding ability with TRIM25. (A) The acetylation levels of KU70 (Up) and KU80 (Bottom) were determined by western blot of anti-acetylation (Ac) antibody following IP KU70 or KU80 in the control or HDAC1 OE HCT116 cells with TG (1 μg/mL) treatment for the indicated time periods. (B) Ub-KU70 and Ub-KU80 were determined by IP/western blot in HCT116 cells in the presence or absence of Trim 25 KD and/or SAHA (1 μM, 24 h). (C) The acetylation levels of KU70 (Up) and KU80 (Bottom) were determined by western blot of anti-acetylation (Ac) antibody following IP KU70 or KU80 in the control or Trim 25 KD (B) HCT116 cells with TG (1 μg/mL) treatment for the indicated time periods. (D and E) The acetylation (D) and ubiquitination (E) levels of WT KU70/KU80 and their acetylation defective mutants (K542/544R KU70, K7/126R KU80) were determined by western blot of anti-Ac and anti-Ub antibody, respectively, following IP KU70 or KU80 in HCT116 cells treated with TG (1 μg/mL) for the indicated time periods. (F) The bindings between TRIM25 and WT KU70/KU80, their acetylation defective mutants (K542/544R KU70 and K7/126R KU80) or their acetylation mimic mutants (K542/544Q KU70 and K7/126Q KU80) were compared by IP assays. (G) Proposed model of HDAC1-mediated KU70/KU80 stabilization by catalyzing deacetylation of KU70 at K542/544 and K7/126 at KU80, because the acetylation of KU70/KU80 at these residues is important for E3 ligase TRIM25 binding and the subsequent KU70/KU80 degradation mediated by TRIM25.

To clarify the mechanisms involved, we predicted potential acetylation sites in KU70 and KU80. The bioinformatics tool revealed that K317, K331, K338, K539, K542, K544, K553, and K556 in KU70, and K7, K126, K443, and K702 in KU80, were the most likely candidates (SI Appendix, Fig. S6D). We then generated a panel of mutations as indicated in SI Appendix, Fig. S6D, and expression levels of KU70/KU80 were measured after treatment with SAHA. The results show that SAHA inhibited the expression of WT KU70/KU80 as expected and most of the mutant proteins (SI Appendix, Fig. S6D). However, mutation of KU70 at K542 and K544 and mutation of KU80 at K7 and K126 mitigated the effects of SAHA-mediated repression (SI Appendix, Fig. S6D), while the double mutations KU70-K542/544R and KU80-K7/126R completely abrogated them (SI Appendix, Fig. S6E). These data suggest that KU70-K542/544 and KU80-K7/126 are the major sites of HDAC1-mediated deacetylation. Interestingly, the ubiquitination-defective mutants (K114/256R-KU70 and K195/265/481R-KU80) had no effect on the acetylation levels of KU70/KU80 (SI Appendix, Fig. S6F), whereas mutations at the acetylation sites (KU70-K542/544R and KU80-K7/126R) abolished KU70/KU80 acetylation, as expected, and also abrogated ER stress–induced KU70/KU80 ubiquitination (Fig. 6 D and E). Our data also showed that KU70/KU80 ubiquitination was increased by acetylation-mimicking mutations (KU70-K542/544Q and KU80-K7/126Q), and was decreased by acetylation-defective mutations (KU70-K542/544R and KU80-K7/126R) (SI Appendix, Fig. S6G). These data further suggest that HDAC1-regulated KU70/KU80 deacetylation mitigates the ability of TRIM25 to catalyze the ubiquitination of KU70/KU80.

Based on the above data, we hypothesized that the acetylation of KU70/KU80 may provide key binding sites for TRIM25. In support of this idea, TRIM25 failed to bind with the acetylation-defective mutant, while its binding with an acetylation-mimicking mutant was increased (Fig. 6F). Consistent with the binding domains identified in Fig. 5C, the key acetylation sites KU70-K542/544 and KU80-K7/126 were resident within the TRIM25 binding domains. We therefore concluded that HDAC1-mediated deacetylation of KU70 at K542/544 and KU80 at K7/126 attenuates their interactions with TRIM25, leading to reduced KU70/KU80 ubiquitination and subsequent protein degradation (Fig. 6G).

DNA-PK or HDAC1 Inhibition Improves the Sensitivity of ER Stress–Induced Cancer Cell Death.

Given our identification of the HRD1/HDAC1/KU70-KU80/DNA-PKcs axis as regulating adaptation–death switching during ER stress, we went on to pursue combined strategies to improve cancer treatments in the clinic. However, while TG is a well-established ER stress inducer, it is not approved for clinical use. Therefore, cells were treated with the cyclooxygenase-2 selective inhibitor celecoxib, an FDA-approved nonsteroidal anti-inflammatory drug (44), because it has been shown to induce ER stress in colon cancer cells, as this was confirmed in our experimental system (SI Appendix, Fig. S7A). As expected, celecoxib-induced ER stress was accompanied by increased DNA damage, as indicated by increased levels of γH2AX (SI Appendix, Fig. S7A). Despite the promotion of ER stress, colon cancer cells were relatively resistant to celecoxib in vitro, as demonstrated by crystal violate (Fig. 7A), soft agarose (SI Appendix, Fig. S7 B and C), and cleaved caspase 3/7 assays (SI Appendix, Fig. S7D). We then utilized genetic strategies to inhibit HDAC1. As shown, HDAC1 KD had mild or negligible effects on tumor growth, while the combination of HDAC1 KD and celecoxib treatment inhibited tumor growth in a synergistic manner (Fig. 7A and SI Appendix, Fig. S7 B and C). In a cell line–derived xenograft (CDX) assay, HDAC1 KD xenografts were found to be more sensitive to celecoxib than the scrambled shRNA (shnone) controls without influencing body weight (Fig. 7 B–E and SI Appendix, Fig. S7E), implying that the treatment had low toxicity. In addition, increased γH2AX, decreased Ki67, and increased cleaved caspase 3/7 levels were detected in celecoxib-treated shHDAC1-HCT116 CDXs (Fig. 7F and SI Appendix, Fig. S7F). Consistent with our in vitro data (SI Appendix, Fig. S7G), HDAC1 and the expression levels of conventional NHEJ components KU70, KU80, Lig4, and XRCC4 were reduced in dissected shHDAC1-HCT116 CDXs treated with celecoxib (Fig. 7G).

Fig. 7.

DNA-PK inhibitor or HDAC1 inhibitor improves the sensitivity of ER stress–induced cancer cell death. (A) Representative crystal violet staining images of the number of shHDAC1 and shnone HCT116 and LoVo cells treated with or without ER stress inducer celecoxib (Left). The images were quantified by ImageJ and the data were plotted in the graphs (Right). (B) HCT116 cells were inoculated into nude mice and then subjected to the celecoxib treatment. (C–E) Tumor volumes at the indicated time (C), tumor images (D), and tumor weight (E) of HCT116 xenografts were presented. (F) Representative immunohistochemistry staining for γH2AX and Ki67 was determined in the indicated xenografts (Left). Quantification of intensity was shown in the bar graph (Right). (Scale bar, 50 μm.) (G) Representative western blot images of HDAC1, KU70, KU80, Lig4, and XRCC4 of the indicated xenografts. Bar graphs represent the mean ± SEM from three independent assays (A and F). The average values of tumor volume and tumor weight are present in the graphs (means ± SD) (n = 5 for each group, C and E). *P < 0.05; **P < 0.01; N.S., not significant (A, C, E, and F).

We further sought to improve the sensitivity of celecoxib by pharmacologically inhibiting HDAC1 (SAHA) or DNA-PKcs (KU-57788) in vitro or in vivo in patient-derived xenografts. As shown, this treatment significantly improved the responses of cells to celecoxib (Fig. 8 A–F and SI Appendix, Figs. S8 A–D and S9 A–C). Combined treatments also reduced NHEJ components at the protein level (Figs. 8G and SI Appendix, Fig. S8E and S9D), while mice body weights did not change significantly (SI Appendix, Fig. S9E). These data suggest that inhibition of the HRD1/HDAC1/KU70-KU80/DNA-PKcs axis promotes the response of tumors to ER stress–inducing agents such as celecoxib.

Fig. 8.

DNA-PK/HDAC1 inhibitors improve the sensitivity of ER stress–induced cancer cell death. (A) Representative crystal violet staining images of the number of HCT116 and LoVo cells treated with celecoxib (10 μM, 24 h), SAHA (1 μM, 24 h), KU-57788 (1 μM, 8 h), or the combinations (Left). The images were quantified by ImageJ and the data were plotted in the graphs (Right). (B) Schematic diagram showed the PDX establishment pipeline with celecoxib, SAHA, KU-57788, or combined treatments. (C–E) PDX-tumor volumes at the indicated time (C), PDX-tumor images (D), and PDX-tumor weight (E) were presented. (F) Representative immunohistochemistry staining for γH2AX and Ki67 were determined in the indicated PDX-tumor (Bottom). Quantification of intensity was shown in the bar graph (Up). (Scale bar, 50 μm.) (G) Representative western blot images of HDAC1, KU70, KU80, Lig4, and XRCC4 of the indicated PDX-tumor. Bar graphs represent the mean ± SEM from three independent assays (A and F). The average values of tumor volume and tumor weight are present in the graphs (C and E) (means ± SD) (n = 5 for each group). **P < 0.01; N.S., not significant (A, C, E, and F).

KU70/KU80 Is Associated with HDAC1 In Vivo in Colon Cancer Tissues.

Increased levels of ER stress are characteristic of cancers. We sought to investigate the clinical relevance of the HRD1/HDAC1/KU70-KU80 axis in colon cancers. 41 paired samples were collected, and the expression levels of the above genes were assessed. Levels of HDAC1, KU70, and KU80 proteins were significantly increased in colon cancer tissues compared with normal controls (Fig. 9 A and B). Our data also showed that this increased expression was associated with advanced-stage tumors (Fig. 9C). In contrast, there were no obvious differences in the expression of HRD1 or TRIM25 between colon cancers and normal tissues, or between different stages of the disease (Fig. 9 A–C). In addition, although expression of HDAC1, KU70, and KU80 proteins in colon cancer was significantly increased, their mRNA levels were unchanged (Fig. 9D), suggesting that posttranscriptional mechanisms control the protein outputs of these genes. HDAC1 protein was also consistently positively associated with the expression levels of KU70 and KU80 proteins in colon cancers (Figs. 9E).

Fig. 9.

The analysis of HRD1/HDAC1/KU70/KU80 protein expressions in colon cancer tissues. (A and B) Representative WB (A) and quantification (B) of HRD1, HDAC1, TRIM25, KU70, and KU80 protein expression in 41 pairs of human colon cancer tissues (T) and their matched adjacent normal controls (N) are present. β-actin was used as a loading control. (C) The correlation between HRD1, HDAC1, TRIM25, KU70, and KU80 protein levels and tumor stages was shown in the box plot. (D) The mRNA levels of HDAC1, KU70, and KU80 were detected by qRT-PCR assays. (E) The scatter diagram showed the lineal correlation of HDAC1 protein/KU70 protein (P < 0.01) or HDAC1 protein/KU80 protein (P < 0.01) in colon cancer tissues. *P < 0.05; **P < 0.01; N.S., not significant (B–E).

Discussion

Here, we identify a mechanism linking ER stress to the DDR: Under the conditions of long-term ER stress, the ERAD ligase HRD1 induces HDAC1’s degradation, which abolishes HDAC1’s ability to protect KU70/KU80 from TRIM25-mediated protein degradation in the nucleus. In this way, the cell transmits ER stress signals to the nucleus to modulate the DDR (SI Appendix, Fig. S10). This is a report of a role of the ERAD machinery in regulating the DDR. Importantly, this mechanism was found to be critical for cell fate determination and may therefore be exploited in cancer therapy.

Accumulating data support a link between the UPR and the DDR. Conventional DDR genes have been shown to regulate ER stress modulators and vice versa. For example, DNA damage can disrupt the structure of the ER and induce ER stress by activating p53 (32). DNA damage can also activate the UPR sensor IRE1a through Abl-mediated phosphorylation, which subsequently catalyzes DDR genes through regulated IRE1α-dependent decay (RIDD) activity independently of XBP1 (45). Unexpectedly, we found that the ER-resident ERAD ligase HRD1 plays a role in the DDR by catalyzing the ubiquitination and degradation of a nuclear protein, HDAC1. Of note, HRD1/HDAC1 binding was evident in the cytoplasm when proteasome activity was blocked, but their interaction was barely detectable in the nucleus, although both proteins could be detected there. A similar mode of action has been reported for cytoplasmic Keap1-mediated degradation of the nuclear protein Nrf2 (46). This is likely to be an important way for cells to transmit cytoplasmic signals to the nucleus. It should also be noted that HRD1 level increased 6 h after ER stress being induced, and the binding between HRD1 and HDAC1 was significantly elevated at 24 h. Scaffolding or chaperone proteins have been shown to adapt the activities of HRD1 ligase to suit different classes of substrates (47–49). It is possible that unknown chaperone proteins are involved in modulating HRD1’s activity in HDAC1 degradation in the cytoplasm in a manner that is not solely depending on HRD1/HDAC1 protein levels. It will be interesting to screen for such chaperon proteins in the cytoplasm, because our data show that nuclear HRD1 does not bind with HDAC1, even at a late stage of ER stress. Involvement of posttranslational mechanisms for either HRD1 or HDAC1 also warrants further investigation.

The UPR leads to two opposing biological outcomes, adaptation and cell death, depending on the intensity and duration of the stress (10, 50). HRD1 knockout cells showed increased levels of misfolded proteins in the ER, suggesting that HRD1 is mainly involved in adaptation (51). However, these results are likely to contradict the previously reported protective role of HRD1. We found that HRD1 promotes the switch from adaptation to death by preventing NHEJ at a late stage of ER stress. In support of our data, a recent report has shown that HRD1 suppresses Nrf2-mediated cellular protection during liver cirrhosis (21). Therefore, ERAD-E3 ligands, like UPR pathways, may act as double-edged swords under ER stress. This finding raises the interesting question of how HRD1 activity is regulated to enable the production of opposing outcomes, which warrants further investigation.

HDAC1 has previously been reported to be involved in the DDR by catalyzing histone proteins such as H3, H4, and H2AX (52–54). Deacetylation of histone proteins alters chromatin structure and, subsequently, the expression of key DDR genes and cell cycle regulators (55). Our data reveal a distinct mechanism underlying the HDAC1-regulated DDR through the catalysis of the nonhistone proteins KU70/KU80. In particular, the identification of nonhistone substrates of HDACs has attracted increasing attention in recent years because HDAC inhibitors have been approved by the FDA (56). Our findings extend the broad mechanisms of HDACs in biological processes and expand the range of possible scenarios for the application of HDAC inhibitors. In addition, we have shown a mechanistic link between the acetylation and degradation of KU70/KU80. Deacetylation of KU70/KU80 reduces their ability to bind to the E3 ligase TRIM25, leading to delayed KU70/KU80 protein turnover. Both acetylation and ubiquitination are known to target lysine residues. Therefore, a competitive interaction was initially proposed. However, the acetylation of lysines in a given protein has been shown to create high-affinity binding sites for other proteins. Our data, together with other reports, suggest that lysine acetylation regulates protein stability mainly by modulating protein–protein interactions. It should also be noted that mutations in acetylation and ubiquitination sites are enriched in cancer cells. Gene-focused analysis using the ActiveDriver method has revealed significant co-occurrence of acetylation and ubiquitination posttranslational modifications (PTMs) and mutation hotspots in known oncoproteins (e.g., TP53, AKT1, and IDH1) (57). These data highlight candidate cancer driver genes with PTM-related mechanisms. PTM-specific mutations are consistently associated with poor survival of patients. Having identified key acetylation and ubiquitination sites in KU70/KU80, it will be interesting to further investigate the clinical relevance of KU70/KU80 lysine mutations in cancer.

An essential role of TRIM25 in catalyzing KU80 has recently been reported, which is consistent with our findings (58). Our data suggest that TRIM25 can also catalyze KU70, which enhances the effect of TRIM25 in influencing the NHEJ pathway. Therefore, increasing the activity of TRIM25 is a potential strategy to improve the responses of cancer cells to DNA damage or ER stress inducers. In particular, DNA damage predicts a better response to immunotherapy. Whether or not ER stress–induced DNA damage and subsequent inactivation of the DDR pathway can improve immunotherapy needs to be further investigated.

Taken together, ER stress is emerging as a powerful target for anticancer therapies. Our data not only reveal a mechanism underlying the link between ERAD and the DDR but also provide insights into how future cancer therapies may combine ER stress inducers with HDAC1 or DNA-PK inhibitors.

Materials and Methods

Human Colorectal Cancer Tissues.

Human colorectal cancer tissues and their corresponding adjacent normal controls were collected from Third Affiliated Hospital of Harbin Medical University in China. Samples were deidentified prior to transport to the laboratory. Specimens were collected and stored in liquid nitrogen immediately after surgery for further analysis. The study is approved by the Research Ethics Committee of Harbin Medical University, China. Moreover, informed consent was obtained from all participants, ensuring the legality, ethics, and protection of participants’ rights and interests throughout the study.

Other Methods.

Cell culture and transfection, animal experiments, immunohistochemistry assay, western blot, RNA extraction and quantitative reverse transcription (qRT) PCR, purification of recombinant protein, IP assay, crystal violet staining, immunofluorescence staining of cells for microscopy, proximity ligation assay (PLA), comet assay, NHEJ assay and HR assay, caspase 3/7 activity assay, ubiquitination assay, in vitro ubiquitination assay, cell counting Kit-8 (CCK-8) assay, proximity labeling–based methods coupled with mass spectrometry (MS), clone formation assay, Nanobit assay, nuclear and cytoplasmic fractionation, and bioinformatics analysis can be found in SI Appendix, Supplementary Materials and Methods.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix. Mass-spectrum data were deposited in the ProteomeXchange Consortium via the iProX partner repository (IPX0009510000) (59).