XLF-mediated NHEJ activity in hepatocellular carcinoma therapy resistance

The HCC cell lines PLC/PRF/5 (ATCC, CRL-8024), Huh7 (provided by Dr. H Nakabayashi, Hokkaido University, Japan), MHCC97 L (97 L) and MHCC97H (97H) (provided by Liver Cancer Institute of Fudan University, China) were cultured in DMEM containing 10% FBS. CD133-PE-labeled Huh7 cells were sorted using magnetic microbeads conjugated to an anti-PE antibody (Miltenyi Biotec. Germany). Sorted CD133+ and CD133− cells were cultured for further experiments. The chemotherapeutic drugs cisplatin (cis, Mayne Pharma, Melbourne, Australia), oxaliplatin (oxa, Jiangsu Hengrui Medicine Co., China), doxorubicin (dox, Main Luck Pharmaceuticals, China), and 5-fluorouacial (5FU, Sigma-Aldrich, St. Louis, MO) were applied to cells, and cell viability was determined by incubation with tetrazolium salt (Cell Counting Kit 8, Dojindo, Japan) and using colony-forming assays.

Small interfering RNAs (siRNAs) against human XLF and ERCC1 (Santa Cruz Biotechnology, Dallas, TX) were transfected into 97 L cells using Lipofectamine RNAiMAX (Life Technologies, Carlsbad, CA). Scrambled siRNA was used as a negative control. shRNA against human NHEJ1 and a scrambled control were constructed using a pEco-Lenti-H1-shRNA (GFP) kit (GenTarget Inc., San Diego, CA). Lentivirus particles were produced in 293 T cells using ViraPower Lentiviral Packaging Mix (Life Technologies) and concentrated by ultracentrifugation (20,000 g). 97 L cells were infected with shXLF or shCon lentiviral particles, followed by puromycin selection for 7–10 days.

A phospho-histone H2AX (Ser139) (γH2AX) antibody was obtained from Cell Signaling Technology (Beverly, MA). An antibody against Ligase IV was purchased from Santa Cruz Biotechnology (Santa Cruz Biotechnology, Dallas, Texas), and an ERCC1 antibody was purchased from Thermo Fisher Scientific (Waltham, MA). For western blotting (WB), PVDF membranes containing proteins electrophoretically separated from cell lysates were probed with relevant antibodies. The resultant immune complexes were visualized using enhanced chemiluminescence detection reagents (Bio-Rad, Hercules, CA). For immunofluorescence staining, cells were harvested and cyto-spun, followed by fixation with 4% paraformaldehyde. After blocking and permeabilization with 1%BSA/0.3%Triton-X100 in PBS for 1 h, the cells were incubated with anti-γH2AX antibody overnight at 4 °C. A FITC-conjugated secondary antibody (Life Technologies) was utilized to visualize the signal.

For the in vitro NHEJ assay, cellular nuclear protein fractions were isolated and applied to repair DNA DSBs in vitro as previously described [ 15]. Briefly, cellular nuclear protein was incubated with linearized plasmid DNA in T4 ligase buffer and 1 mM dNTPs for 2 h at 14 °C. After de-proteinization, the quantity of end-joined DNA products was measured by quantitative PCR with a pair of primers flanked primer-joining junction of the plasmid. Relative NHEJ activity was calculated as the ratio of end-joined products normalized to the loading control. For the in vivo NHEJ assay, the engineered construct pEGFP-PEM1-Ad2 [ 16] was transfected using X-tremeGENE HP reagent (Roche, Hong Kong). The starting construct was GFP-negative, and the successful repair of Hind III-induced DSBs by cellular NHEJ was identified by restored functionality of the GFP gene [ 16]. After transfection, the number of GFP-positive cells was measured by flow cytometry and normalized to the transfection efficiency for in vivo NHEJ activity.

97 L cells (1 × 10 ⁶) infected with lentiviruses harboring shRNA-XLF or shRNA-scramble were subcutaneously injected into nude mice to generate an HCC xenograft model. Two weeks after tumor cell injection, 4 cycles of oxaliplatin (4–10 mg/kg) were given via weekly intraperitoneal injection. Tumor volumes were measured at the endpoint. Nuclear proteins from the xenograft tumors were extracted to perform cell-free NHEJ assays. All animal experiments were approved by the Committee on the Use of Live Animals of The University of Hong Kong (CULATR 3091–13).

Tissue specimens were collected from 31 patients with HCC who received transarterial chemoembolization (TACE) as a first treatment followed by a hepatectomy. All patients were treated at Queen Mary Hospital. Tissue specimens were also collected from patients with HCC who underwent hepatectomy as a first treatment. This study was approved by the Institutional Review Board of the University of Hong Kong/Hospital Authority of Hong Kong (UW05–3597/I022). The need for informed consent was waived because the study was retrospective in design.

Data are presented as the mean ± SD. Paired and independent Student’s t tests were performed using SPSS 21 software (IBM Corp. Armonk, NY). Overall and disease-free survival rates for the included patients were analyzed using the Kaplan-Meier log-rank method.

To better understand how DNA damage repair contributes to HCC therapy resistance, we first investigated the correlation between the presence of DNA lesions induced by chemotherapeutic drugs and cellular sensitivity. To accomplish this, three types of drugs with different DNA-damaging mechanisms were applied. γH2AX nuclear foci staining was used to detect DSBs: within minutes of the occurrence of DNA damage, H2A.X becomes phosphorylated at Ser139 (γH2AX) at sites of damage [ 17]. Cisplatin and oxaliplatin (DNA crosslinking agents), doxorubicin (topoisomerase II inhibitor), and 5-FU (antimetabolites) can all induce DSBs both directly and indirectly (Fig. 1a). Following treatments with these drugs, 25% to 50% of 97 L and PLC cells showed positive staining for γH2AX in nuclear foci (Fig. 1b). The γH2AX-positive nuclear foci peaked at 9 h after drug treatment (Fig. 1c). Interestingly, at 48 h, the percentage of γH2AX-positive cells in the 97 L cell group was significantly reduced (Fig. 1c, right panel) compared to that in the PLC cell group, where the percentage remained high (Fig. 1c, left panel). This result indicates that these two cell lines harbor differences in their DSB repair capacity.

We further compared NHEJ activity in vitro and in vivo. Nuclear proteins were isolated from drug-treated cells to perform in vitro NHEJ assays using double-stranded plasmid DNA as a substrate. NHEJ activity in 97 L cells was 5- to 10-fold higher in cells submitted to drug treatment compared to untreated cells. In contrast, NHEJ activity in PLC cells did not change in response to drug treatment (Fig. 1d). Similarly, the drugs oxaliplatin and doxorubicin induced NHEJ activity in 97 L cells in live-cell assays but not in PLC cells (Fig. 1e). Thus, 97 L cells displayed significantly higher in vitro and in vivo NHEJ activity than PLC cells. Concordantly, we found that PLC cells were sensitive to the drugs, whereas 97 L cells were resistant to them (Fig. 1f). The half maximal inhibitory concentration (IC50) for oxaliplatin for inducing HCC cell death was consistently significantly higher in 97 L cells (4.8 μg/ml) than in PLC cells (2.6 μg/ml) (Fig. 1g). Thus, conventional chemotherapeutic drugs for HCC treatment can induce DSBs, and NHEJ responses following treatment with DNA-damaging drugs differ among different cells. HCC cells (97 L) with increased NHEJ activity are more resistant to drugs, whereas HCC cells (PLC) with no enhancement of NHEJ activity are more sensitive.

Cancer stem cells (CSCs) are highly resistant to chemotherapeutic drugs [ 18, 19]. We found that a correlation exists between relative chemotherapeutic response and NHEJ activity in HCC CD133+ (potential CSC population) and CD133− populations. In studying Huh7-CD133+ and Huh7-CD133− populations, we found that the former was more drug resistant than the latter, with 75–85% of CD133+ cells surviving after oxaliplatin or doxorubicin treatment (Fig. 2a). When investigating whether NHEJ mechanisms contribute to liver CSC resistance, we found that gene expression levels of the NHEJ factors Ku, DNA-PKcs, and XLF were significantly higher in CD133+ than in CD133− cells (Fig. 2b). We next transfected pEGFP-PEM1-Ad2 plasmids into both CD133+ and CD133− populations and then observed drug-induced NHEJ activity in vivo. The CD133+ cells contained a significantly higher GFP+ population than the CD133− cells (Fig. 2c), and this pattern was consistent with all three drug treatments (Fig. 2c). This result indicates that DSB repair capacity in CD133+ cells is higher than in CD133− cells. Statistical comparisons of the in vivo NHEJ assay results are summarized in Fig. 2d. The results indicate that significantly higher NHEJ activity is found in CD133+ cells. These results suggest that CSCs possess higher intrinsic NHEJ capacity.

XLF is a core member of the NHEJ protein complex. XLF interacts with XRCC4 to form an XRCC4/XLF dimer that bridges DNA ends in vitro; thus, XLF might stimulate ligation [ 20– 22]. XLF also accumulates at sites with DSBs [ 14]. Whether the known functions of XLF have a role in drug-induced DSB repair and cancer resistance is not known. Resistant 97 L HCC cells displayed high NHEJ activity both in vitro and in vivo (Fig. 1d, e). After knocking down XLF, the drug-induced in vitro NHEJ activity was significantly inhibited (Fig. 3a). In vivo, NHEJ activity was still induced by the drugs in shRNA-control-transfected cells, whereas it was not in cells transfected with shRNA-XLF (Fig. 3b). These results imply that down-regulating XLF significantly impairs NHEJ capacity. 97 L cells had higher DSB repair efficiency, and γH2AX foci were reduced after longer-term drug treatment (Fig. 1c). Following XLF knockdown, the number of γH2AX foci remained high, with 83.1% ± 6.1 of cells displaying >5 γH2AX foci (Fig. 3c), indicating the presence of unrepaired DSBs. As a result, 97 L cells became sensitized to the chemotherapeutic drugs tested (Fig. 3d).

We next explored the cross-regulation of XLF with other DNA repair genes, such as LIG4, as XLF stimulates the XRCC4/LIG4 complex [ 13]. We also investigated excision repair cross-complementing 1 (ERCC1) of the nucleotide excision repair (NER) pathway. Achieving SSL is ideal for producing therapeutic effects, and the HR and NHEJ pathways show high levels of SSL with the NER pathway [ 4]. Down-regulation of XLF was associated with reduced LIG4 mRNA (Fig. 3e, left panel) and protein expression levels (Fig. 3e, right panel), which is consistent with previous findings that XLF regulates LIG4 in cancer. Interestingly, in XLF-knockdown cells, ERCC1 mRNA and protein expression levels also decreased (Fig. 3e). We then investigated whether a synergistic effect exists between the chemotherapeutic drugs after two DNA repair genes are inhibited. In combination with ligase inhibitor (L189), the drugs inhibited the growth of shXLF cells more than the growth of shCon cells (Additional file 1: Fig. S1). Inhibiting XLF enhanced the chemotherapeutic sensitivity in resistant 97 L HCC cells (Fig. 3d). We then determined a synergistic effect compared to cisplatin alone, particularly when both XLF and ERCC1 were knocked down because ERCC1 positive-tumors predict cisplatin resistance in non-small-cell lung cancer and squamous cell carcinoma [ 23– 25]. The IC50 of cisplatin was 5.56 μg/ml for the siCon group, 3.15 μg/ml for the siXLF group, 2.51 μg/ml for the siERCC1 group, and 1.43 μg/ml for the siXLF + siERCC1 group (Fig. 3f), indicating a chemosensitization effect when there was knockdown of either XLF or ERCC1, where the most enhanced chemosensitivity was seen with knockdown of both XLF and ERCC1. Crosstalk between DNA repair pathways often results in the acquisition of resistance mechanisms in cancer [ 3]. This result suggests that targeting independent pathways may produce stronger synergy than targeting the same pathway at multiple points [ 2].

To establish a xenograft model, shCon- and shXLF-transfected 97 L cells were subcutaneously injected into the left and right flanks, respectively, of the same nude mouse. Using the drug administration protocol depicted in Fig. 4a, oxaliplatin treatment was started at the third week after cell injection and was continued for 4 weeks. At the endpoint, overall tumor volumes were significantly smaller for the shXLF-HCC xenografts than the shCon-HCC xenografts (Fig. 4b, c), indicating that XLF knockdown sensitized HCC xenografts to oxaliplatin. Cell-free NHEJ activity was measured in nuclear proteins extracted from xenograft tissue. A dramatic enhancement in NHEJ activity was observed in the shCon xenografts, whereas NHEJ activity was consistently inhibited in the shXLF xenografts (Fig. 4d). Thus, the shXLF-HCC xenograft tumors were more responsive to oxaliplatin as a result of shXLF-mediated reductions in NHEJ activity.

Cancer cells are frequently dysfunctional in one or more DDR and DNA damage repair pathways, which can be specifically inhibited to induce SSL. These dysfunctions can also lead to additional alterations in DDR pathways that may induce therapy resistance [ 1, 4]. Given that XLF-mediated induction of NHEJ activity in HCC cells is associated with chemotherapeutic drug sensitivity (Figs. 1 and 2), we investigated the incidence of XLF gene alterations using bioinformatic analysis. We also investigated genomic alterations of the core factors in the NHEJ pathway in HCC and other cancers. The NHEJ pathway includes 8 genes: NHEJ1 (XLF), XRCC4, XRCC5, XRCC6, LIG4, DNA-dependent protein kinase catalytic subunit (PRKDC), TP53BP1, and DNA cross-link repair 1C (DCLRE1C). Genomic alterations were identified using cBioPortal [http://www.cbioportal.org] [ 26, 27] based on 90 studies from 17 to 26 types of primary tumor and cancer lines assembled by The Cancer Genomics Atlas (TCGA) [ 28] and Asan Medical Center (AMC) [ 29]. The frequency of alterations, including mutation, deletion, amplification, and multiple alterations, for NHEJ pathway genes in HCC was found to be 17.6% by TCGA and 10% by AMC (Additional file 1: Fig. S2); in contrast, the frequency of alterations in the XLF gene was extremely low in HCC. Indeed, only one of 231 patients had a missense mutation (mutation rate: 0.4%)(Fig. 5a) in the AMC data collection, and no mutations were found in the 371 patients included in the TCGA data collection (Additional file 1: Fig. S3; Additional file 2: Table S1). Moreover, upregulation of XLF mRNA was found in only 5% of patients with HCC (Additional file 2: Table S1). Given the low genomic alteration rate of XLF in patients with HCC, we sought to measure XLF levels in primary HCC tissues from patients who did or did not receive TACE treatment before liver tumor resection to determine whether therapy drugs can impact XLF expression, because TACE delivers chemotherapy in combination with embolization to administer therapy directly to liver tumors. As shown in Fig. 5b, XLF expression was significantly increased in patients with HCC who underwent TACE compared to those who did not, indicating that the delivery of drugs by TACE may directly or indirectly stimulate XLF expression.

Given that XLF expression (Fig. 5b) and XLF-mediated alterations in NHEJ activity (Fig. 1d, e; Fig. 3a, b) can be induced by therapy drugs, we wanted to assess whether enhancing XLF expression would affect HCC clinical parameters. The patients in the TACE treatment group who showed higher levels of XLF expression displayed poor prognosis features such as larger tumor size, venous infiltration, and advanced-stage HCC disease (Table 1). More importantly, XLF expression was significantly associated with overall survival of HCC patients, with patients exhibiting higher levels of XLF showing shorter survival times (Fig. 5c). High XLF expression was also associated with a shorter disease-free survival time, although the trend did not reach statistical significance (Fig. 5d).