Cancer-associated oxidoreductase ERO1-α promotes immune escape through up-regulation of PD-L1 in human breast cancer

Many human cancers have been reported to have enhanced expression of the immune checkpoint molecule programmed death-ligand 1 (PD-L1), which binds to programmed cell death-1 (PD-1) expressed on immune cells. PD-L1/PD-1 plays a role in inhibition of antitumor immunity by inducing T cell apoptosis and tolerance. Thus, it is crucial to elucidate mechanisms of PD-L1 expression on cancer cells. ERO1-α is an oxidase located in the endoplasmic reticulum. It is overexpressed in a variety of tumor types and it plays a role in disulfide bond formation in collaboration with PDI. Here, we investigated the influence of ERO1-α on expression of PD-L1 and immune escape. We demonstrated that ERO1-α augmented the expression of PD-L1 via facilitation of oxidative protein folding within PD-L1. In addition, we showed that overexpression of ERO1-α increased HIF-1α protein expression, resulting in an increase of PD-L1 mRNA as well as protein. In clinical cases, we observed that the expression of ERO1-α in triple negative breast cancer was related to the expression of PD-L1. Moreover, apoptosis of Jurkat leukemia T cells, which express PD-1, induced by tumor PD-L1 was inhibited when ERO1-α was depleted. The results suggest that targeting ERO1-α in tumor cells can be a novel approach for cancer immunotherapy. Therefore, the role of ERO1-α in tumor-mediated immunosuppression should be further explored.


INTRODUCTION
Breast cancer has been the most common cancer worldwide in women over the past decade [1]. Triple negative breast cancer (TNBC), which lacks expression of the estrogen receptor, progesterone receptor and HER2/ neu (HER2), accounts for 10% to 20% of breast cancer cases [2,3]. TNBC is generally characterized by a high rate of distant metastasis and poorer disease-specific survival than that of other breast cancer subtypes [4].
An immunosuppressive condition, particularly in a tumor-associated microenvironment, is one of major obstacles for the development of effective immunotherapy. Tumors have developed strategies to evade antitumor immunity, and the developing tumor often co-opts the host immune system for its own tumor-promoting purpose. Much attention has recently been paid to immune checkpoint molecules including programmed death-ligand 1 (PD-L1) as immunosuppressive molecules. Expression of PD-L1 is found in various types of cells and tissues www.impactjournals.com/oncotarget/ Oncotarget, 2017, Vol. 8, (No. 15), pp: 24706-24718

Research Paper
including the placenta, vascular endothelium, pancreatic islet cells, muscle, hepatocytes, and mesenchymal stem cells [5]. PD-L1 mRNA expression is elevated by interferon β [6], γ [7] and under a condition of hypoxia [8,9]. PD-L1 binds to programmed cell death-1 (PD-1), a CD28 family receptor that is expressed on activated T cells, activated B cells and myeloid cells [10], and PD-L1/PD-1 plays a role in inhibition of autoimmunity by inducing T cell apoptosis [11] and inducing tolerance [12]. Enhanced expression of PD-L1 has been found in human cancers including glioblastoma, melanoma, lung cancer, ovary cancer, colon cancer and breast cancer [10]. Importantly, the expression of PD-L1 in TNBC is enhanced compared with that in other types of breast cancer [4,13]. Some clinical studies using an anti-PD-L1 mAb have recently been performed [14,15]. Brahmer et al. reported that blocking PD-L1 by using an anti-PD-L1 mAb results in both durable tumor regression (objective response rate, 6% to 17%) and prolonged (≧24 weeks) disease stabilization in patients [15]. Therefore, it is important to clarify underlying mechanisms for PD-L1 expression and develop a strategy for the regulation of PD-L1 expression in cancer immunotherapy.
Endoplasmic reticulum oxidoreductase 1-α (ERO1-α) is an oxidoreductase that exists in the endoplasmic reticulum (ER). ERO1-α plays a central role in disulfide bond formation of secreted and cell surface molecules [16][17][18][19][20] in collaboration with protein disulfide isomerase (PDI). A disulfide bond is produced at the posttranslational level and is required for proper conformation and function of these molecules [20,21]. In fact, we have recently demonstrated that expression of G-CSF, CXCL1, CXCL2 and MHC class I is regulated by proper oxidative protein folding by ERO1-α [22,23]. We found that various types of tumors expressed high levels of ERO1-α and that the expression of ERO1-α in breast cancer is a poor prognosis factor [22][23][24]. However, the mechanisms by which ERO1-α expression affects the prognosis of breast cancer are still unclear. In this study, we found that ERO1-α up-regulates PD-L1 expression via not only oxidative protein folding but also unexpected up-regulation of PD-L1 mRNA expression through augmentation of hypoxia-inducible factor-1α (HIF-1α) protein expression in human TNBC cell lines. Knockdown of ERO1-α resulted in a significant attenuation of PD-L1-mediated T cell apoptosis, thus providing a lead for therapeutic modulation of hypoxia-mediated immunoresistance.

ERO1-α upregulates the expression of PD-L1 via both oxidative protein folding and increasing the mRNA of PD-L1
We have shown that the expression of ERO1-α in human breast cancer tissue and in cell lines was augmented compared with that in normal breast tissue [22,25]. In this study, we focused on how ERO1-α affects expression of the immune checkpoint molecule PD-L1 in TNBC because patients with TNBC generally have a poor prognosis. Therefore, we examined the role of ERO1-α in the expression of PD-L1 using TNBC line MDA-MB-231 cells. We established ERO1-α-overexpressed (OE: 9A3, 9C1 and 9C2) cells by introducing human cDNA of ERO1-α ( Figure 1A). When we compared mock cells with OE cells, we found by flow cytometry that the surface expression of PD-L1 in all OE cell lines was augmented ( Figure 1B, 1C). These results suggested that the expression of ERO1-α upregulated PD-L1 expression. We then compared PD-L1 mRNA expression levels in mock and OE cells by real-time RT-PCR. Unexpectedly, we found that PD-L1 mRNA expression levels in OE cells were significantly higher than those in mock cells ( Figure 1D). Since recent studies have demonstrated that PD-L1 mRNA expression level is regulated by HIF-1α [8,9], we examined the expression levels of HIF-1α. We found that the protein levels of HIF-1α in OE cells were higher than those in mock cells (Supplementary Figure 1A), indicating that HIF-1α may play a role in the upregulation of PD-L1 mRNA expression. To further examine the role of HIF-1α in OE cells, we silenced HIF-1α in 9C2 (9C2 siHIF-1α) cells using siRNA for HIF-1α (Supplementary Figure 1B). When we compared the PD-L1 mRNA expression level in 9C2 cells with the levels in control siRNA cells (9C2 siNC cells) and 9C2 siHIF-1α cells, we observed that the PD-L1 mRNA expression level in 9C2 siHIF-1α cells was decreased compared with that in 9C2 siNC cells (Supplementary Figure 1C). These results indicated that ERO1-α-mediated upregulation of HIF-1α augmented the mRNA expression of PD-L1. It has been shown that intracellular reactive oxygen species (ROS) increase HIF-1α protein levels by inhibiting HIFprolyl hydroxylase (PHD) activity [26,27]. In addition, ERO1-α has been shown to produce ROS during oxidation of a substrate protein [28]. Therefore, we compared the levels of ROS in mock cells and OE cells. Fluorescence microscopic analysis and flow cytometric analysis using CM-H 2 DCFDA showed that overexpression of ERO1-α in OE cells increased the levels of ROS compared with those in mock cells (Supplementary Figure 1D, 1E). Moreover, we found that OE cells had decreased levels of hydroxyl-HIF-1α, indicating that HIF-1α escaped from degradation by proteasome (Supplementary Figure 1F). These results suggested that expression of ERO1-α results in accumulation of HIF-1α protein via production of ROS, leading to the increased expression of PD-L1 mRNA. Next, we compared the protein levels of PD-L1 by Western blot analysis. Since ERO1-α acts as an oxidoreductase and PD-L1 protein has an intramolecular disulfide bond [29], we investigated the redox status of PD-L1 in mock cells and OE cells by Western blot analysis under a nonreducing condition using N-ethylmaleimide (NEM). As expected, we found that the ratio of the oxidized form (mature form) to the reduced form (immature form) of PD-L1 in OE cells was significantly higher than that in www.impactjournals.com/oncotarget mock cells ( Figure 1E, 1F). These results suggested that ERO1-α up-regulates the expression of PD-L1 via not only oxidative protein folding but also promoting HIF-1α-mediated mRNA expression.

Knockdown of ERO1-α downregulates PD-L1 expression via decreased oxidative protein folding
To further confirm the effect of ERO1-α on expression of PD-L1, we generated MDA-MB-231 cells with ERO1-α knockdown (KD: sh221 and sh222) using shRNA against ERO1-α ( Figure 2A). Scrambled control (SCR) cells were transfected with scrambled control shRNA. When we compared SCR cells and KD cells, we found by flow cytometry that the surface expression of PD-L1 in KD cells was significantly decreased ( Figure  2B, 2C). Next, we compared PD-L1 mRNA expression levels by real-time RT-PCR. The mRNA expression levels were not different in SCR cells and KD cells ( Figure 2D). We then compared the protein levels of PD-L1 by Western blot analysis. We found that the ratio of the oxidized form (mature form) to the reduced form (immature form) of PD-L1 in KD cells was significantly lower than that in SCR cells under a non-reducing condition ( Figure 2E, 2F). Thus, knockdown of ERO1-α resulted in decreased expression of PD-L1 through insufficient oxidative protein folding. However, since there was a discrepancy in the context of mRNA expression of PD-L1 between overexpression and knockdown of ERO1-α, we compared the levels of ROS (Supplementary Figure 2A) as well as HIF-1α protein levels (Supplementary Figure 2B) in SCR cells and KD cells. We found that both levels of ROS and HIF-1α protein levels were decreased in KD cells compared with the levels in SCR cells. We also examined whether HIF-1α regulates PD-L1 mRNA expression levels in in a steady state of MDA-MB-231 WT cells, because MDA-MB-231 WT cells expressed moderate levels of ROS and HIF-1α protein as shown in Supplementary  Figure 2A, 2B. Unexpectedly, knockdown of HIF-1α in WT cells using siRNA to HIF-1α did not affect the PD-L1 mRNA level (Supplementary Figure 2C, 2D). These results suggest that PD-L1 mRNA in a steady state of WT cells, SCR cells and KD cells is regulated by factors other than HIF-1α. In contrast, in MDA-MB-231 OE cells, PD-L1 mRNA expression levels were regulated by factors including HIF-1α. In fact, it has been shown that the PD-L1 mRNA expression level is regulated by NF-κB [30], interferon regulatory factor (IRF)-1 [31], STATs [32,33], MEK/ERK [32,34] and AP-1 [30].
To generalize these findings, we examined the effect of ERO1-α on the expression of PD-L1 using another

ERO1-α plays a critical role in IFN-γ-induced upregulation of PD-L1 expression in tumor cells
A previous study demonstrated that PD-L1 expression in cancer cells is enhanced by IFN-γ [35]. To investigate the role of ERO1-α in the expression of PD-L1 in cells treated with IFN-γ, we treated SCR cells and sh221 cells with IFN-γ (100 ng/ml) for 24 h and then compared the cell surface expression of PD-L1 in these cells using flow cytometry. We found that IFN-γ treatment increased the cell surface expression of PD-L1 in SCR cells compared with that in cells without IFN-γ treatment (Supplementary Figure 4A, 4B). In contrast, the expression of PD-L1 in sh221 cells treated with IFN-γ was not altered when compared with sh221 cells without IFN-γ treatment. We then compared the PD-L1 mRNA expression levels in these cells by using real-time PCR.
We found that IFN-γ treatment significantly and equally upregulated PD-L1 mRNA expression in SCR cells and sh221 cells (Supplementary Figure 4C, 4D). These results suggested that ERO1-α also plays a critical role in the IFN-γ-induced upregulation of PD-L1 expression.

Functional inhibition of ERO1-α by using the ERO1-α inhibitor EN460 leads to decreased expression of PD-L1 via inhibition of oxidative protein folding
Since the ERO1-α inhibitor EN460 is known to inhibit oxidative protein folding by reductive inactivation of ERO1-α [36], we treated MDA-MB-231 WT cells with EN460 to further examine the effect of ERO1-α on expression of PD-L1. We found by flow cytometric analysis that the surface expression of PD-L1 in MDA-MB-231 WT cells treated with EN460 was decreased compared with that in control cells ( Figure  3A, 3B). However, PD-L1 mRNA expression levels were not different in control cells and treated cells (data not shown). The results indicated that functional inhibition of ERO1-α did not influence the PD-L1 gene expression level. Next, we compared the protein levels by Western blot analysis under a non-reducing condition using methyl methanethiosulfonate (MMTS). We found that the ratio of the oxidized form (mature form) to the reduced form (immature form) of PD-L1 in treated cells was significantly decreased compared with that in control cells under a non-reducing condition ( Figure 3C). To confirm that the ERO1-α-PDI pathway oxidized PD-L1, we examined the redox status of ERO1-α, PDI, and PD-L1 by Western blotting. We found that the ERO1-α-PDI complex in treated cells was decreased compared with that in control cells under a non-reducing condition. Moreover, EN460 treatment resulted in conversion of the oxidized form of ERO1-α to the reduced form, indicating loss of function for reoxidation of PDI [36]. These results again suggested that the expression of PD-L1 is dependent on the oxidative folding activity of ERO1-α.

Knockdown of ERO1-α in MDA-MB-231 cells decreases the rate of apoptosis of Jurkat leukemia T cells
Jurkat leukemia T cells expressed PD-1 ( Figure 4A), which is known as a receptor for PD-L1. We observed that coculture of Jurkat cells with PD-L1-positive MDA-MB-231 SCR cells resulted in apoptotic cell death ( Figure 4B, 4C). To examine the effect of ERO1-α on the function of PD-L1, SCR and KD cells (sh221, sh222) were cocultured with Jurkat leukemia T cells. We found that the rate of apoptosis in Jurkat T cells cocultured with SCR cells was higher than that in Jurkat cells only ( Figure  4B, 4C). We also found that the rate of apoptosis in Jurkat  T cells cocultured with KD cells was lower than that in Jurkat T cells cocultured with SCR cells. On the other hand, the rate of apoptosis in Jurkat T cells cocultured with SCR cells was decreased by treatment with an anti-PD-L1 or anti-PD-1 mAb. These results suggested that the expression of ERO1-α in tumor cells upregulated PD-L1 expression, resulting in the induction of apoptosis of tumor-infiltrating T cells.

Expression of ERO1-α in a tumor is correlated with expression of PD-L1 in clinical cases
To examine the correlation between expression of ERO1-α and expression of PD-L1 in clinical cases, TNBC tissues were stained with an anti-ERO1-α mAb and anti-PD-L1 mAb, and they were classified into four groups based on the expression status ( Figure 5A). It was found the expression level of ERO1-α was significantly correlated with the expression of PD-L1 (RS=0.65, P < 0.001) ( Figure 5B). The ERO1-α score 0, 1+ group showed significantly reduced expression of PD-L1 compared with that in the ERO1-α score 2+, 3+ group ( Figure 5C). These results indicated that enhanced expression of ERO1-α in TNBC plays a critical role in the expression of PD-L1.

DISCUSSION
During tumor progression, tumors must develop strategies to evade antitumor immunity. Moreover, hypoxia, a hallmark of the tumor microenvironment, contributes to the acquisition of more aggressive properties in cancer cells such as drug resistance and metastatic ability. Among oxidoreductases, ERO1-α is unique, because it is upregulated within tumor cells [22][23][24][25] and under the condition of hypoxia [37,38]. Oxidative protein folding such as intramolecular disulfide bond formation is the most common post-transcriptional modification [20]. Proper disulfide bond formation is required for correct conformation and function of proteins [21]. Most of the proteins that are secreted from cells or expressed on the cell surface have intramolecular disulfide bonds. Therefore, oxidative protein folding is critical for the cells to function and survive. ERO1-α plays a central role in oxidative protein folding in collaboration with PDI [21,39]. We previously showed that ERO1-α is overexpressed in various types of cancer cell lines and cancer tissues including breast cancer tissues compared with levels in normal cells and normal tissues [22][23][24][25]. Moreover, we reported that overexpression of ERO1-α within the tumor was a poor prognostic factor in patients with breast cancer [24]. In this study, we found by using triple negative breast cancer MDA-MB-231 cells and MDA-MB-468 cells that overexpression of ERO1-α protein resulted in augmented expression of PD-L1. It is known that disulfide bond formation by ERO1-α is accompanied by the production of hydrogen peroxide (H 2 O 2 ), a potential source of ROS [28]. ROS increase the HIF-1α protein level via inhibition of enzymatic activity of a family of prolyl hydroxylases [26,27]. In fact, we showed that overexpression of ERO1-α increased levels of ROS (Supplementary Figure  1D, 1E), resulting in increased HIF-1α protein levels and both mRNA and protein levels of PD-L1. These results coincide with results of recent studies showing that HIF-1α regulates PD-L1 expression [8,9]. Moreover, since PD-L1 protein has an intramolecular disulfide bond [29], we examined the redox status of PD-L1. As expected, the expression of ERO1-α facilitated oxidative folding of PD-L1, resulting in an increase in the oxidized form of PD-L1. In contrast, knockdown or functional inhibition of ERO1-α resulted in decreased expression of PD-L1 only at post-transcriptional levels. It has been shown that the PD-L1 mRNA expression level is regulated by NF-κ [30], IRF-1 [31], STATs [32,33], MEK/ERK [32,34] and AP-1 [30]. Additionally, HIF-1α enhances PD-L1 mRNA expression under certain conditions including hypoxia [8,9]. Our results suggest that low or moderate levels of HIF-1α observed in WT cells, SCR cells, and KD cells had no effect on PD-L1 expression, indicating that PD-L1 expression is regulated by factors other than HIF-1α in these cells. However, high levels of HIF-1α detected in OE cells play a very important role in PD-L1 mRNA expression. The precise mechanism by which PD-L1 expression is regulated must be clarified in future studies. Thus, ERO1-α affected PD-L1 expression at both mRNA levels and posttranscriptional levels (Supplementary Figure 5).
Recently, it has been demonstrated that ERO1-αderived H 2 O 2 is cleared by the ER-resident peroxidases glutathione peroxidase (GPx) 7/8 and peroxiredoxin (Prx) IV [40,41]. However, we found high levels of ROS within OE cells. These results suggest that high levels of ROS produced by the action of ERO1-α in OE cells might be beyond the ROS-detoxifying ability of GPx7/8 and Prx IV, resulting in the accumulation of ROS and HIF-1α protein as well as increased expression of PD-L1 mRNA.
In this study using human TNBC cell lines, we showed that ERO1-α plays an important role in the expression of PD-L1 via facilitation of oxidative protein folding and in the upregulation of PD-L1 mRNA via transcriptional activity of HIF-1α. Importantly, downregulation of PD-L1 expression by knockdown of ERO1-α resulted in decreased PD-L1-induced apoptosis of T-cells. Therefore, it should be clarified whether targeting ERO1-α in tumor cells results in augmentation of T-cell-mediated tumor cell killing in vitro as well as in an immunocompetent mouse model of breast cancer.

Real-time PCR analysis
Total RNA was isolated from cultured cells and normal breast tissues using Trizol reagent (Life Technologies) and RNeasy Mini kits (QIAGEN, Valencia, CA) according to the instructions of the manufacturers. The cDNA mixture was synthesized from 1 μg total RNA by reverse transcription using Superscript III and oligo (dT) primer (Life Technologies) according to the manufacturer's protocol. PCR amplification was performed in 20 μl of PCR mixture containing 1 μl of cDNA mixture, 0.1 μl of Taq DNA polymerase (QIAGEN) and 6 pmol of primers. Real-time relative polymerase chain reaction (real-time PCR) was performed to determine the expression levels of ERO1-α, PD-L1 and β-actin. Expression values for each sample were normalized to β-actin, and fold levels of the indicated genes represent the mean (±SEM) of replicate reactions. Primer sequences were as follows: β-actin (ACTB), Hs0160665_g1; ERO1-α (ERO1L), Hs00205880_m1; and PD-L1, Hs01125301_m1 (Life Technologies). PCR cycles were performed on the StepOne Real-Time PCR System (Life Technologies) with the following cycle conditions: 2 min at 50 °C, 10 min at 95°C, 45 cycles of 15 s at 95°C and 1 min at 60°C. The delta-delta Ct method was used for data analysis.

Statistical analysis
Student's t-test or Welch's t-test was used for analysis of two unpaired samples. Dunnett's test was used for analysis of multiple samples. All analyses were carried out with STATMATE version 3.19 (ATMS Co., Ltd., Tokyo, Japan). A P-value of less than 0.05 was regarded as statistically significant. All statistical tests were two-sided.