Regulation of MRE11A by UBQLN4 leads to cisplatin resistance in patients with esophageal squamous cell carcinoma

Resistance to standard cisplatin‐based chemotherapies leads to worse survival outcomes for patients with esophageal squamous cell carcinoma (ESCC). Therefore, there is an urgent need to understand the aberrant mechanisms driving resistance in ESCC tumors. We hypothesized that ubiquilin‐4 (UBQLN4), a protein that targets ubiquitinated proteins to the proteasome, regulates the expression of Meiotic Recombination 11 Homolog A (MRE11A), a critical component of the MRN complex and DNA damage repair pathways. Initially, immunohistochemistry analysis was conducted in specimens from patients with ESCC (n = 120). In endoscopic core ESCC biopsies taken from 61 patients who underwent neoadjuvant chemotherapy (NAC) (5‐fluorouracil and cisplatin), low MRE11A and high UBQLN4 protein levels were associated with reduced pathological response to NAC (P < 0.001 and P < 0.001, respectively). Multivariable analysis of surgically resected ESCC tissues from 59 patients revealed low MRE11A and high UBLQN4 expression as independent factors that can predict shorter overall survival [P = 0.01, hazard ratio (HR) = 5.11, 95% confidence interval (CI), 1.45–18.03; P = 0.02, HR = 3.74, 95% CI, 1.19–11.76, respectively]. Suppression of MRE11A expression was associated with cisplatin resistance in ESCC cell lines. Additionally, MRE11A was found to be ubiquitinated after cisplatin treatment. We observed an amplification of UBQLN4 gene copy numbers and an increase in UBQLN4 protein levels in ESCC tissues. Binding of UBQLN4 to ubiquitinated‐MRE11A increased MRE11A degradation, thereby regulating MRE11A protein levels following DNA damage and promoting cisplatin resistance. In summary, MRE11A and UBQLN4 protein levels can serve as predictors for NAC response and as prognostic markers in ESCC patients.

Resistance to standard cisplatin-based chemotherapies leads to worse survival outcomes for patients with esophageal squamous cell carcinoma (ESCC). Therefore, there is an urgent need to understand the aberrant mechanisms driving resistance in ESCC tumors. We hypothesized that ubiquilin-4 (UBQLN4), a protein that targets ubiquitinated proteins to the proteasome, regulates the expression of Meiotic Recombination 11 Homolog A (MRE11A), a critical component of the MRN complex and DNA damage repair pathways. Initially, immunohistochemistry analysis was conducted in specimens from patients with ESCC (n = 120). In endoscopic core ESCC biopsies taken from 61 patients who underwent neoadjuvant chemotherapy (NAC) (5-fluorouracil and cisplatin), low MRE11A and high UBQLN4 protein levels were associated with reduced pathological response to NAC (P < 0.001 and P < 0.001, respectively). Multivariable analysis of surgically resected ESCC tissues from 59 patients revealed low MRE11A and high UBLQN4 expression as independent factors that can predict shorter overall survival [P = 0.01, hazard ratio (HR) = 5.11, 95% confidence interval (CI), 1.45-18.03; P = 0.02, HR = 3.74, 95% CI, 1. 19-11.76, respectively]. Suppression of MRE11A expression was associated with cisplatin resistance in ESCC cell lines. Additionally, MRE11A was found to be ubiquitinated after cisplatin treatment. We observed an amplification of UBQLN4 gene copy numbers and an increase in UBQLN4 protein levels in ESCC tissues. Binding of UBQLN4 to ubiquitinated-MRE11A increased MRE11A degradation, thereby regulating MRE11A protein levels following DNA damage and promoting cisplatin resistance. In summary, MRE11A and UBQLN4 protein levels can serve as predictors for NAC response and as prognostic markers in ESCC patients.

Introduction
Esophageal cancer represents the 7th cause of cancer death globally [1]. The two major histological types that account for over 90% are esophageal squamous cell carcinoma (ESCC) and esophageal adenocarcinoma (EAC) [2]. ESCC is the most common histology worldwide (~90%), with a particularly high incidence across Asia and southeastern Africa while being equally frequent to EAC (~50%) in northern America and Europe [3][4][5]. In locally advanced ESCCs, neoadjuvant chemotherapy (NAC) improves overall survival (OS) when compared to surgery alone [6] or postoperative adjuvant treatment [7], and platinum-based drugs stand as key agents in treatment regimens combined with fluorouracil and taxanes [7][8][9]. Despite the clinical advances in ESCC chemotherapy, the response rates remain as low as 40% [10]. Up to 80% of ESCC patients experience recurrence leading to early mortality [11] and the 5-year OS remains~20% [12]. Importantly, patients who have a better pathological response to NAC have a better OS and disease-free survival [13][14][15]. However, only a few biomarkers have been used to predict response to NAC, with their clinical relevance still being underestimated [16]. Therefore, in order to improve patient outcomes, it is critical to understand the molecular basis of the tumor progression in ESCC tumors and identify aberrant DNAdamage response (DDR) mechanisms that drive resistance to current drugs used in NAC.
The MRE11-RAD50-NBS1 (MRN) complex has an important role in DDR and chemotherapy treatment. MRN complex sensors and orchestrates DDR in double-strand breaks (DSB), stalled replication forks, dysfunctional telomeres, and immune responses [17,18]. Alterations of MRN complex have been associated with chemotherapy and radiotherapy resistance [18][19][20][21]. There is an assumption that the upregulation of the MRN complex is associated with the resistance to DNA damaging therapy due to the higher efficiency of DNA damage repair [17,[19][20][21][22]. However, there is still controversy whether this assumption can be generalized to all tumor types as some differences may arise from different tumor origin. In regards, it was previously shown that certain cancer types are less susceptible to DNA damaging therapies when they have lower expression of the MRN complex components [23][24][25]. The mechanisms controlling MRN complex, in particular Meiotic Recombination 11 Homolog A (MRE11A) as related to ESCC chemoresistance, are not fully understood.
Ubiquilin family proteins function as adaptors between ubiquitinated proteins and the proteasome [26][27][28][29]. Five ubiquilin genes (UBQLN1-4 and UBQLNL) have been identified in the human genome [28]. Ubiquilins contain two conserved domains: Ubiquitin-like domain (Ubl) and Ubiquitin-associated domain (UBA). The Ubl domain has homology to the ubiquitin domain and interacts with the proteasome regulatory component s5a, while the UBA domain binds to mono-and poly-ubiquitin [30]. UBQLN4 was recently shown to have a role to shuttle nuclear proteins to the cytosol for degradation through the nuclear pore [31]. Studies have shown that UBQLN4 targets misassembled ER-localized proteins to the proteasome, thus reducing proteotoxic cell stress and harboring a protective role [26]. However, there is no evidence showing whether UBQLN4 has a role in reducing genotoxic stress in ESCC tumor cells during chemotherapy treatment and how that affects the tumor sensitivity to chemotherapy.
Although hypomorphic mutations on MRE11A have been identified in different solid tumors [17], mutations in MRE11A gene are rare in ESCC tumors. UBQLN4 has a role in controlling nuclear protein levels by targeting ubiquitinated proteins to the proteasome [29,31]. In ESCC, we proposed the hypothesis that during DNA damage MRE11A is degraded by the ubiquitin-proteasome system facilitated by UBQLN4. The aims of this study were as follows: (a) to unravel the regulatory mechanism affecting MRE11A; (b) determine the role of MRE11A in controlling the response to cisplatinbased therapy in ESCC patients; and (c) analyze the associations between MRE11A expression and OS in patients with ESCC. obtained from clinical stage II/III [American Joint Committee on Cancer (AJCC) 8th ed.] ESCC patients who underwent NAC followed by surgery at Keio University Hospital, Department of Surgery, Tokyo, Japan, between 2011 and 2016. These patients did not undergo adjuvant treatment. Specimens met the following criteria: (a) histologically proven ESCC, (b) no previous history of chemotherapy or chemoradiotherapy for any malignancies, and (c) no microscopic residual tumor (R0). For NAC, cisplatin plus 5-fluorouracil was repeated twice every 3 weeks (day 1: cisplatin by 80 mgÁm À2 ; days 1-5: 5-fluorouracil by 800 mgÁm À2 ) [7]. Tumor regression grade was evaluated by pathologists at Keio University Hospital. Responders are defined as marked changes in twothirds or more in the tumor area. Nonresponders were defined as marked changes in less than two-third. Written informed consent was obtained from all subjects. Immunohistochemistry (IHC) for MRE11A and UBQLN4 was performed for this cohort and IHC was quantified to obtain H-score values. Associations between their H-scores and treatment response to NAC were then analyzed.
Formalin-fixed paraffin-embedded tissues from 59 surgically resected ESCC primary tumors were obtained from patients who underwent up-front surgery at Keio University Hospital from 1997 to 2002 [32]. IHC for MRE11A and UBQLN4 was also performed and IHC was quantified to obtain the H-score values. Univariate and multivariate analysis was performed to determine associations between H-scores and clinical variables.
All of the experiments followed World Medical Association Declaration of Helsinki and the NIH Belmont Report principles. Tissue specimens were coded according to HIPAA recommendation. This study was approved by Keio University School of Medicine Ethics Committee with a registration number of 2016-0233 and 2016-241. All human samples and clinical information for this study were obtained according to the protocol guidelines approved by the Joint Institutional Review Board of Providence Saint John's Health Center/ Saint John's Cancer Institute and the Western Institutional Review Board (USA).

Cell lines
Established human ESCC cell lines TE-4 (CVCL_3337), TE-8 (CVCL_1766), and TE-10 (CVCL_1760) were obtained from Keio University. Cell lines were cultured in Dulbecco's modified Eagle's medium or RPMI supplemented with 10% FBS at 37°C with 5% CO 2 . All human cell lines have been authenticated using short tandem repeat profiling within the last 3 years. All experiments were performed with mycoplasma-free cell lines.

Plasmids and ESCC cisplatin-resistant cell lines
MRE11A EX-q-0491-M35 and the empty control vector EX-EGFP-M35 were from Genecopoeia, Rockville, MD, USA; ubiquitin B pCMV6 vector and the empty vector (EV) were from Origene, Rockville, MD, USA; UBQLN4 was cloned in pLNCX2 plasmid as described previously [29]. For stable clones, UBQLN4 or EVs were packaged into lentivirus particles using HEK-293T cell lines. Purified lentiviral particles were transduced into TE-8 and TE-10 cell lines. Stable clones were selected with G418 100 lgÁmL À1 .

Immunohistochemistry (IHC)
Endoscopic biopsy specimens and surgically resected ESCC tumors were stained with UBQLN4 monoclonal antibody (mAb; #sc-136145; Santa Cruz Biotechnology, Santa Cruz, CA, USA) and/or MRE11A polyclonal antibody (pAb; #4895S; Cell Signaling Technology, Danvers, MA, USA) as previously described [33]. Images were taken by the BX43 upright microscope (Olympus, Tokyo, Japan) with a magnification of 209 using the MANTRA SNAP Software 1.03 (Perkin Elmer, Waltham, MA, USA). H-scores were calculated using the INFORM 2.4 software (Perkin Elmer) and following manual instructions. Briefly, the tumor area was segmented from the stromal area and nuclei/cytoplasm compartments were distinguished by detecting the intensity of hematoxylin and/or 3,3'-diaminobenzidine (DAB) staining. The optical signal threshold to classify the score into 4-bins was set to 0.05, 0.12, and 0.2. Five slides per case were analyzed and the average H-score was then used as the final value. The cutoff values for UBQLN4 and MRE11A were determined by considering H-score values for the mean value observed in normal adjacent epithelia esophagus tissues plus 10 SD.

Co-Immunoprecipitation assays
Cell lines were transfected with MRE11A and UBB plasmids as indicated in each experiment. After 24 h, cell lines were treated with 5 lM MG-132 (#S2619; Selleck chemicals) and/or 5 lM cisplatin for 12 h. After incubation, cell lines were washed and lysed in Co-Immunoprecitation (Co-IP) buffer (150 mM NaCl, 100 mM Tris/HCl pH 8, 1% NP-40, protease, and phosphatase inhibitors) by gently pipetting. The whole-cell (WC) lysates were quantified by bicinchoninic acid assay, and 250 lg of WC lysate (final concentration 1.25 lgÁlL À1 ) was incubated with UBQLN4 mAb (#sc-136145; Santa Cruz Biotechnology) and mouse (G3A1) mAb control (#5415; Cell Signaling Technology); or UBQLN4-Ps318 pAb (#A300-L20645, Bethyl) and mouse (G3A1) mAb control (#5415; Cell Signaling Technology); or MRE11A pAb (#4895S; Cell Signaling Technology) and normal rabbit pAb control (#2729S; Cell Signaling Technology); or DDDDK tag pAb(#ab1162; Abcam, Cambridge, MA, USA) and normal rabbit pAb control (#2729S; Cell Signaling Technology) in Co-IP buffer. Immunocomplex was further incubated with Dynabeads TM Protein A (#10002D; ThermoFisher Scientific, Waltham, MA, USA) for 1 h at 4°C. Only for the experiment showed in Fig. S3A, we incubated the beads from all the conditions with buffer or 50 nM of recombinant human USP2 catalytic domain protein (Bio-Techne Corporation, Minneapolis, MN, USA) for 30 min at 37°C, as indicated. Beads were washed three times with Co-IP buffer, recovered, and boiled in Fluorescent Master Mix for 5 min at 95°C in a dry bath. After centrifugation, the supernatants (SN) were recovered and protein concentration was adjusted to 1 lgÁlL À1 with 19 MM reagent. Samples were analyzed by automated western blot system [33,34].

On-chip cell sorting to purify UBQLN4-GFP cell lines
The microfluidic chip-based cell sorter, On-chip Sort (On-chip Biotechnologies, Tokyo, Japan), equipped with blue (488 nm), violet (405 nm), and red (637 nm) lasers, was used according to the manufacturer's protocol. TE-8 and TE-10 UBQLN4-OV cell lines were harvested, centrifuged, and diluted to 1 9 10 7 /mL. The sorting microfluidic chip was precoated with On-chip T-buffer before loading the sample. After priming the chip, excessive T-buffer was removed from all reservoirs, and replaced with fresh culture medium. Then, the prepared samples were loaded. Gates were set and the samples were sorted according to the cell size and GFP positivity of interest. The sorted samples were retrieved, and the sorting process was repeated three times to increase the purity of the sorted cell lines. After sorting, cell lines were immediately cultured until 80% confluence and then frozen.

Immunofluorescence (IF)
Cell lines (1-1.5 9 10 4 ) were seeded in 8-well Falcon TM chambered culture slides (Fisher Scientific) and incubated for 48 h, followed by treatment with 5 lM cisplatin for 12 h. After that, cell lines were fixed and the protocol was followed as described in [34] except that primary Ab dilutions in 5% BSA are as shown in Table S1. The photographs were acquired using a Nikon Eclipse Ti microscope and NIS elements software. Quantification of fluorescent signal was performed on IMAGEJ software (ImageJ; NIH, Bethesda, MD, USA).

Cell viability and colony formation assays
ESCC cell lines (1-2 9 10 3 ) were seeded in a 96-well plate. The number of viable cells was assessed using a Cell Titer-Glo Luminescent (Promega, Madison, WI, USA) according to the manufacturers' instructions. For colony formation, ESCC cell lines (2 9 10 3 ) were seeded in a 6-well plate and after 12 days of incubation processed as previously described [34].

Nuclear extraction
Nuclear and cytoplasmic fractions were isolated from the TE-10 and TE-8 cell line with the Nuclear Extract Kit (Active Motif, Carlsbad, CA, USA). Cells (8.8 9 10 6 cells/dish) were cultured in 100-mm dishes and harvested with 3 mL cold PBS/phosphatase inhibitor buffer. Cells were centrifuged and the WC pellet was gently suspended in 500 µL 19 hypotonic buffer and incubated for 15 min on ice. Then, 25 µL of detergent was added to induce cell lysis. After cell lysis, the cytoplasmic fraction (supernatant) was separated from the nuclear fraction (pellet) by centrifugation (for 30 s at 14 000 g). Then, the nuclear fraction (pellet) was resuspended in 50 µL of the complete lysis buffer and incubated with 2.5 µL detergent for 30 min on ice. The nuclear lysates were centrifuged for 10 min at 14 000 g. The nuclear fraction (supernatant) was then collected. Both nuclear and cytoplasmic fractions were analyzed by automated western blot.

Western blot
Automated western blot was performed according to the manufacturer's protocol (Protein Simple, San Jose, CA, USA; https://www.proteinsimple.com/wes. html [36,37]). Briefly, the proteins were extracted with lysis buffer (150 mM NaCl, 100 mM Tris/HCl pH 8, 1% NP-40, and protease inhibitors) and the protein concentration was adjusted to 0.5-1 lgÁlL À1 . The results were analyzed using the Compass Software (Protein Simple) and IMAGEJ software (ImageJ; NIH). The primary Ab dilutions are shown in Table S1. All uncropped western blot images are shown in Fig. S7,S8.

Biostatistical and bioinformatics analysis
Normal distribution of data was tested using the Kolmogorov-Smirnov test. The parametric Student's t-test was used for normally distributed data, and the nonparametric Wilcoxon rank-sum test was used for nonnormally distributed data. One-way ANOVA test and Bonferroni post hoc test was applied in the quantification of IF assays. Two-way ANOVA test and the Bonferroni post hoc test were applied for cell proliferation and drug sensitivity analysis. Correlation analyses were performed using Pearson's or Spearman's correlation coefficient (r) according to the data distribution. OS was compared using the Kaplan-Meier method and log-rank test. Multivariable analysis was performed by the Cox proportional regression model to analyze the independent predictor for OS. Significant factors considering the covariate factors and age were included for the multivariable analysis. A two-sided P < 0.05 was considered statistically significant: *P < 0.05; **P < 0.01; ***P < 0.001 and NS, not significant. All analyses were performed with GRAPHPAD PRISM 5 (GraphPad software Inc., La Jolla, CA, USA) or R version 3.5.0 (R Core Team, 2018 [38]), and the figures were made using CorelDraw graphics suite 89 (Corel Corporation, Ottawa, ON, Canada).
Then, MRE11A expression was assessed by IHC on 61 endoscopic biopsy FFPE tissues procured from patients with AJCC Stage II/III ESCC, who underwent cisplatin-based NAC followed by surgery. The biopsies were collected before NAC treatment. Then, the patients were stratified based on pathological response to NAC into responder and nonresponders. H-score values demonstrated that MRE11A expression was significantly increased in responder (n = 8) compared to nonresponder ESCC patients who underwent NAC (n = 53; H-score = 293.3 AE 3.4 versus 162.8 AE 11.8, P < 0.001, Fig. 1G,H).

MRE11A expression determines cisplatin resistance in ESCC cell lines
To determine the role of MRE11A in controlling cisplatin sensitivity, we performed MRE11A knockdown and overexpression (OV) in ESCC cell lines and analyzed them for cisplatin sensitivity. MRE11A knockdown promoted cisplatin resistance in ESCC cell lines ( Fig. 2A-D). Conversely, MRE11A-OV enhanced cisplatin sensitivity in ESCC cell lines (Fig. 2E-H).
In order to determine whether MRE11A was associated with cisplatin resistance, we established two cisplatin-resistant ESCC cell lines. Cisplatin resistance in ESCC cell lines was confirmed by drug sensitivity assays (Fig. 2I,J). In agreement with our previous observation, western blot analysis revealed that (H) Comparison of H-scores for MRE11A protein levels in core biopsy tissues from nonresponders (n = 53) or responders (n = 8) ESCC patients to NAC (***P < 0.001). Error bars represent the mean AE SD. Statistical differences were tested using Mann-Whitney test (C), an unpaired two-tailed t-test with Welch's correction (E and H) and log-rank test (F).
MRE11A expression was significantly lower in the cisplatin-resistant compared to respective parental cell lines (Fig. 2K-N). To further validate this observation, cisplatin-resistant cell lines were knockdown for MRE11A (Fig. S2A,B) and assessed for the sensitivity to cisplatin. The results showed that cisplatin-resistant cell lines with MRE11A knockdown were even more resistant to cisplatin compared to the respective parental cell lines (Fig. S2C,D). Conversely, cisplatin-resistant cell lines with MRE11A-OV (Fig. S2E,F) were tested for cisplatin sensitivity. MRE11A-OV in cisplatin-resistant cell lines restored the sensitivity to cisplatin (Fig. S2G,H). These results suggested that MRE11A expression controls cisplatin resistance in ESCC cell lines.

UBQLN4 promotes MRE11A degradation
We first examined the expression of MRE11A and UBQLN4 protein levels in three ESCC cell lines (TE-4, TE-8, and TE-10). Western blot analysis showed that when ESCC cell lines have increased UBQLN4, they consistently have low MRE11A protein levels (Fig. 3A). Ectopic OV of GFP-tagged UBQLN4 (UBQLN4-OV) or an EV was performed for TE-8 and TE-10 ESCC cell lines. After lentiviral transduction, positive clones were selected, then enriched for GFPpositive clones using the On-chip cell sorting system, and finally confirmed for UBQLN4-OV by western blot (Fig. 3B).
Our hypothesis is that during DNA damage ubiquitinated-MRE11A levels increased in the DNA damage sites and UBQLN4 may help in promoting the proteasome-mediated degradation. Thus, we examined the interaction between MRE11A and UBQLN4 using Co-IP assays. For Co-IPs, the ESCC cell lines were treated with cisplatin (to enhance the levels of ubiquitinated-MRE11A and promote the interaction with UBQLN4) and with MG-132 inhibitor (to block proteasomal degradation). As shown in Fig. 3C,D, UBQLN4 co-immunoprecipitated with MRE11A in ESCC cell lines. The reciprocal Co-IP was also performed.
Then, we performed experiments to assess the changes in ubiquitinated-MRE11A after cisplatin-induced DNA damage. To do this, we transfected TE-10 UBQLN4-OV cells with UBB-DDK and used anti-DDK antibody to immunoprecipitate endogenous tagged ubiquitinated-MRE11A in cisplatin-treated ESCC cell lines in the presence or absence of MG-132. To demonstrate that ubiquitinated-MRE11A was bound to the beads, we pretreated the immunocomplexes with buffer or the recombinant catalytic domain of the de-ubiquitinase USP2. If the interaction would be mediated by ubiquitination, USP2 would reduce the binding of MRE11A in the immunocomplexes. Ubiquitinated-MRE11A levels were increased in ESCC cell lines treated with cisplatin in the presence of MG-132 compared to absence of MG-132 (Fig. S3A). Moreover, USP2 decreased the amount of ubiquitinated-MRE11A (Fig. S3A). These results confirmed that cisplatin promotes MRE11A ubiquitination.
In order to analyze MRE11A protein stability in the presence of high UBQLN4 levels, cycloheximide chase assays were performed in UBQLN4-OV cell lines and compared to respective control EV cell lines. A significant decrease was observed for MRE11A protein levels in UBQLN4-OV compared to the control EV cell lines (Fig. 3G,H). In IF assays, UBQLN4-OV cell lines exhibited a reduction in MRE11A expression under basal conditions, but also under cisplatin treatment ( Fig. S4A-F). These observations were further validated using western blot in nuclear fractions obtained from TE-8 and TE-10 ESCC cell lines untreated or treated with cisplatin. Endogenous UBQLN4 protein levels increased in the nuclear fractions after cisplatin treatment, while the amount of endogenous MRE11A consistently decreased (Fig. S4G,H). To determine whether MRE11A degradation is mediated by the proteasome, we performed cycloheximide chasing assays in UBQLN4-OV cell lines in the presence or absence of the proteasome inhibitor MG-132. We observed that blockage of proteasomal degradation increased the levels of MRE11A (Fig. 3I,J), consistently with the observations as shown Fig. S3A. We concluded that UBQLN4-OV in ESCC leads to an accelerated MRE11A degradation by targeting ubiquitinated-MRE11A for proteasomal degradation.

High UBQLN4 leads to worse postoperative survival in ESCCs
The Cancer Genome Atlas database analysis revealed that UBQLN4 was highly expressed (mRNA z-score ≥ 1.5) in about 24% (23 of 95) and decreased in only 2% (mRNA z-score ≤ 1.5) of ESCC patients (Fig. 4A). In addition, ESCC tumors showed significantly higher UBQLN4 mRNA expression levels compared to normal adjacent esophageal epithelia (Fig. 4B). UBQLN4 is localized on chromosome 1q23.3, a region commonly amplified in other solid squamous tumor types, and we therefore analyzed the copy number variation (CNV) in this region in primary ESCC tumors using the TCGA ESCA dataset. Forty-four (48.4%) of the primary ESCC tumors showed a gain in the copy number for the UBQLN4 gene (Fig. 4C), and also, a significant positive correlation between linear CNV and UBQLN4 mRNA expression levels was observed (Fig. 4D).
Overall  14) versus high UBQLN4 and low MRE11A (n = 10) protein levels (P < 0.001). (J) Representative images of core biopsy tissues from nonresponders (Patients 1 and 2) or responders (Patients 1 and 2) ESCC patients to NAC that were stained for UBQLN4 using IHC. Scale bars = 50 µm. Right top insets on each picture show a magnification of MRE11A staining Scale bars = 10 µm. (K) Comparison of H-scores for UBQLN4 protein levels in core biopsy tissues from nonresponders (n = 53) and responders (n = 8) ESCC patients to NAC (***P < 0.001). Error bars represent the mean AE SD. Statistical differences were tested using Mann-Whitney test (B), an unpaired two-tailed t-test with Welch's correction (F and K), ordinary one-way ANOVA test and Bonferroni post hoc test (C and G), spearman correlation (D), and log-rank test (H and I).
To further elucidate the clinical significance of UBQLN4, IHC and H-score quantification were performed in surgically resected ESCC specimens, in which IHC for MRE11A was also performed ( Table 1). Since both nuclear and cytosolic staining for UBQLN4 was positively correlated (Fig. S5A,B), all further analyses were referred to nuclear UBQLN4 staining. UBQLN4 expression in primary ESCC tumors (n = 59) was significantly higher than normal adjacent esophageal epithelia (n = 10, Hscore = 82.0 AE 78.9 versus 12.2 AE 10.2, P < 0.0001, Fig. 4E,F). In addition, UBQLN4 expression was significantly higher in recurrent compared to nonrecurrent primary ESCC tumors (Fig. 4G). A cutoff Hscore value > 110 was considered as high for UBQLN4 in primary ESCC tumors (Fig. S5A). The survival curve indicated that patients with high UBQLN4 (median 5 years OS = 35 months, n = 20) had a significantly worse prognosis compared to those with low UBQLN4 (median 5 years OS = 72 months, n = 39, P < 0.01; Fig. 4H). Moreover, high UBQLN4 expression determined by IHC staining was an independent factor to predict poor OS in multivariable analysis (P = 0.02, HR = 3.74, 95% CI, 1.19-11.76, Table 2).
Finally, a comparison of patients with low UBQLN4/ high MRE11A (n = 14) and high UBQLN4/low MRE11A (n = 10) showed that patients with high UBQLN4/low MRE11A have a worse prognosis (P < 0.0001, Fig. 4I). These results confirm the role of UBQLN4 in ESCC and reinforce our conclusions on the clinical importance of high UBQLN4/low MRE11A in predicting OS.
IHC analysis for UBQLN4 was also performed for endoscopic biopsy FFPE prior to NAC, which are the same tissues used to assess MRE11A expression. The quantified H-score values demonstrated that UBQLN4 expression was significantly increased in the nonresponder group (n = 53) compared to the responder group (n = 8) (H-score = 131.8 AE 5.2 versus 60.4 AE 15.7, P < 0.001, Fig. 4J,K).

UBQLN4 expression determines the sensitivity to cisplatin in ESCC cell lines
Based on the observation that UBQLN4 expression was increased in patients who did not respond to cisplatin-based chemotherapy, we hypothesized that UBQLN4 expression may control resistance to cisplatin. Considering a relatively higher UBQLN4 expression, TE-4 cell lines were selected to perform knockdown experiments. UBQLN4 expression was efficiently reduced in si-UBQLN4 (pool siRNA) compared to si-Ctrl-treated cell lines (Fig. 5A). UBQLN4 knockdown had a suppressive effect in cell proliferation as evaluated by cell viability and colony formation assays (Fig. 5B,C). To test our hypothesis, the effect of UBQLN4 knockdown was examined in drug sensitivity assays by comparing si-UBQLN4 and si-Ctrl cell lines. UBQLN4 knockdown cell lines had significantly higher sensitivity to cisplatin (Fig. 5D). Next, UBQLN4-OV and control EV cell lines were assessed for cell proliferation and cisplatin resistance. UBQLN4-OV increased cell proliferation (Fig. 5E,F).
In drug sensitivity assays, UBQLN4-OV cell lines presented cisplatin resistance compared to control EV cell lines (Fig. 5G,H). To summarize, UBQLN4 expression level determined the response to cisplatin in ESCC cell lines.

UBQLN4-OV alleviated DNA damage induced by cisplatin in ESCC cell lines
Since UBQLN4 expression determines resistance to cisplatin in ESCC cell lines, we assumed that UBQLN4 may reduce the DNA damage after cisplatin treatment. To determine the DNA damage, two markers (53BP1 and c-H2AX) were analyzed by IF in TE-10 and TE-8 EV and UBQLN4-OV cell lines cisplatintreated or untreated. As anticipated, ESCC cell lines with UBQLN4-OV presented a reduction in 53BP1 (Fig. 6A-F) and c-H2AX foci formation (Fig. 6G-L) compared to control EV cell lines in both cisplatintreated and nontreated conditions. Then, c-H2AX was analyzed by IF in si-Ctrl or si-UBQLN4 TE-4 cell lines cisplatin-treated or untreated. Consistently, TE-4 cell lines transfected with si-UBQLN4 showed an enhancement in c-H2AX activation (Fig. S6A-C) compared to si-Ctrl in both cisplatin-treated and nontreated condition. Finally, we determined whether knockdown of MRE11A also modified the levels of DNA damage induced by cisplatin. Knockdown of MRE11A significantly decreased the levels of c-H2AX in cisplatin-treated conditions (Fig. S6D-F). In order to demonstrate that UBQLN4 localized to DNA damage areas, c-H2AX and UBQLN4 were analyzed by IF in si-UBQLN4 and si-Ctrl TE-4 cell lines cisplatintreated or untreated. We observed that UBQLN4 localized to DNA damage areas in ESCC cell lines (Fig. S6G) and that c-H2AX and UBQLN4 levels positively correlated in TE-4 cell lines treated with cisplatin (Fig. S6H). In summary, our results reveal that enhanced UBQLN4 expression alleviates the genotoxic stress induced by cisplatin and promotes cisplatin resistance in ESCC cell lines.

Discussion
This study showed that low MRE11A expression in tumor biopsies taken prior to NAC was associated with cisplatin resistance in ESCC, which was further validated in vitro using two cisplatin-resistant ESCC cell lines. In addition, low MRE11A expression in upfront surgically resected ESCC tissues was related to significantly worse OS. It is accepted that the loss of MRN complex function results in DNA repair deficiency and increase sensitivity to DNA damaging therapies [17,18,22]. However, in certain cancer types, loss of MRE11A and other components of the MRN complex promote a resistant phenotype in response to DNA damage [23][24][25]. Recently, new and more complex functions have been assigned to MRE11A in relation to the nuclease activity and the MRN complex [17].
In ESCC patients, UBQLN4 expression was increased and associated with increased gene copy number. Moreover, UBQLN4 protein levels were found to be increased in core biopsies from ESCC patients who have poor response to NAC. Multivariable analysis showed that UBQLN4 was an independent prognostic factor to predict OS. These results suggest a potential role for UBQLN4 to predict cisplatin resistance in ESCC tumors. Additionally, our in vitro results demonstrated that UBQLN4 OV alleviated DNA damage induced by cisplatin, while UBQLN4 knockdown caused the opposite effects.
In order to identify molecular mechanisms regulating MRE11A expression at the DSB sites, we interrogated its relationship with UBQLN4. We demonstrated that DNA damage, induced by cisplatin, promotes UBQLN4 increase at DSB in ESCC cell lines. Furthermore, ubiquitinated-MRE11A interacted with UBQLN4 during DNA damage induced by cisplatin, to be efficiently targeted to the proteasome for degradation. These results validated our previous observations demonstrating that UBQLN4 and ubiquitinated-MRE11A accumulate at the DSB after inducing DNA damage in other solid tumors [29]. In addition, we previously demonstrated that high UBQLN4 protein levels activate alternative DDR pathways to overcome MRE11A loss and DNA damage induced by cisplatin and other DNA damage drugs; however, further studies are needed to address this and identify those oncogenic pathways reducing DNA damage in ESCC. UBQLN2 and UBQLN4 have a conserved role as regulators of endoplasmic reticulum stress, autophagy, and mTOR signaling [40]. In regards, UBQLN4 targets nuclear and cytosolic proteins for proteasomal degradation and controls the proteotoxic cell stress [26,31,[40][41][42][43]. Our studies demonstrated that UBQLN4 levels increased in the nuclear fractions after cisplatin treatment and that also localized at the DSB. Further, we demonstrated that UBQLN4 targets ubiquitinated-MRE11A for degradation and reduced the genotoxic stress. However, UBLQN4 could potentially regulate the expression of other proteins associated with endoplasmic reticulum stress, autophagy, and mTOR signaling to promote cisplatin resistance in ESCC tumors. These protein interactions and specific pathways activation may also decrease the proteotoxic stress induced by cisplatin. Unraveling the pathways controlled by UBQLN4 in relation to proteotoxic stress may offer new potential targets driving cisplatin resistance and may have synergistic effects with the role of UBQLN4 in controlling DDR.
Various attempts have been made to predict prognosis for patients with locally advanced ESCC after NAC, including diagnostic imaging [44], tumor markers [45], inflammation scores [46], and other clinicopathological factors. In addition, recent molecular studies have shown the role of specific molecules activated during the DDR pathway that confer cisplatin resistance in ESCC [47,48]. However, the appropriate surrogate marker for treatment response before NAC is still unknown, and none of these markers are currently used in the clinic for early therapeutic decisions. Our study has demonstrated the potential application of MRE11A and UBQLN4 as prognostic markers and potential NAC response markers for patients with ESCC. Future studies will validate the usage of these two biomarkers in prospective NAC clinical trials for ESCC and other solid tumors receiving platinum-based therapies.

Conclusion
Decreased MRE11A expression is associated with cisplatin resistance in primary ESCC tumors and cell lines. On contrary, increased copy number for UBQLN4 gene upregulates UBQLN4 expression in ESCC. Mechanistically, cisplatin treatment increased ubiquitinated-MRE11A, which is targeted to the proteasome by UBQLN4 and leads to cisplatin resistance in vitro. As both MRE11A and UBQLN4 expressions were associated with clinical outcomes, they could serve as predictors for cisplatin-based NAC response and survival outcomes in patients with primary ESCC tumors.
staining for endogenous UBQLN4 and c-H2AX were performed in cisplatin-treated (5 lM, 12 h) or untreated TE-4 cell lines. UBQLN4 (red), c-H2AX (green), DAPI (blue), and the merged images are shown. Scale bars = 10 µm. H. Correlation between UBQLN4 and c-H2AX levels in cisplatin-treated TE-4 cell lines (Pearson r = 0.74, P < 0.001). Error bars represent the mean AE SD from n = 3 replicates. Statistical differences were tested using ordinary one-way ANOVA test and Bonferroni post hoc test (C and F). Fig. S7. Western blot uncropped images. Fig. S8. Western blot uncropped images. Table S1. List of antibodies (Ab) and dilutions utilized in this study.