xCT maintains cellular GSH store and prevents oxidative cell death in ABC-DLBCL cells.
We first examined the expression pattern of xCT in DLBCL cell lines. In a panel of 9 DLBCL cell lines examined, moderate-to-high levels of the xCT protein were detected in about half of those in the ABC category (4/9) but only 1 of 4 in the GCB subtype (Fig. 1A). Interestingly, 2 cell lines (Ly3-S and Ly10) express readily detectable xCT mRNA but have no xCT protein while the opposite is true for TMD8 (Supplementary Fig. 1A), suggesting roles played by post-transcriptional regulation. For additional experiments in this study, 3 ABC-DLBCL cell lines with readily detectable levels of xCT protein were studied, i.e. Riva, SuDHL2, and U2932. Detailed mechanistic analyses were performed in Riva and SuDHL2, both of which described as double-hit lymphomas8, 25.
To pharmacologically suppress xCT activity, we used Sulfasalazine (SASP), a specific inhibitor of xCT-mediated cystine uptake11. In each of the 3 cell lines tested, Riva, SuDHL2, and U2932, endogenous ROS was significantly elevated following 20 hrs of SASP treatment at individually determined IC50 concentrations (Fig. 1B, Supplementary Fig 1B). As expected, SASP markedly reduced the endogenous GSH pool although the kinetic response was different among the 3 cell lines. In Riva cells, SASP treatment triggered an initial rise in GSH level followed by a rapid return to the baseline at 8 hrs. In SuDHL2 cells, the GSH pool was completely depleted within 8 hrs. In U2932, a gradual decline was observed, resulting in only ~ 50% loss at 24 hrs (Fig. 1C) possibly due to a marked upregulation of the xCT protein following SASP treatment (Supplementary Fig. 1C). SASP-triggered ROS increase could be significantly attenuated by exogenous glutathione monoethyl ester (GEE) supplement (Fig. 1D), confirming that a diminished GSH pool was directly responsible for the ROS spike in SASP treated cells. We also examined cellular consequences of SASP treatment in Riva and SuDHL2. p-p38 signals were markedly increased at 12 and 24 hrs, indicating activation of the antioxidant response (Fig. 1E). Consistent with apoptotic cell death (PARP cleavage, Fig. 1E and data not shown), we detected reduction in pro-survival BCL2 family proteins (MCL1 and BCL2) and increase in pro-death BCL2 family members (Bim and Bax) following SASP treatment (Fig. 1E). Collectively, these results suggest that at steady state, xCT protects ABC-DLBCL cells from oxidative cell death by enabling GSH synthesis and maintaining the endogenous GSH store.
Dox triggers an early surge in cellular GSH as the result of an xCT-dependent, anti-oxidant defense response.
Dox is the cornerstone ingredient of chemotherapy regimens for DLBCL. In our previous study, we showed that in 4 out of 6 ABC-DLBCL cell lines tested, Dox-induced cytotoxicity was dependent on oxidative stress. We found that the STAT3-regulated SOD2 can antagonize this process.29 Other cellular redox regulators were not examined. As shown in Fig. 2A, Dox triggered a rapid decease in the GSH store within 4 hrs followed by a quick rebound at 8 hrs. By 24 hrs, the GSH pool was depleted in U2932 while remaining well-above the basal levels in Riva and SuDHL2 cells, in which there appeared to be a more robust antioxidant response. This Dox-triggered GSH resurgence in Riva and SuDHL2 could be eliminated if cells were pre-treated with SASP (Fig. 2B). SASP pre-treatment also augmented Dox-triggered ROS build-up significantly (Fig. 2C).
To validate the role of xCT using a genetic approach, we established 2 xCT stable knock-down (KD) Riva cell lines using lentivirus-mediated gene knock-down (shxCT#1 and shxCT#3, Fig. 2D). Interestingly, in control cells (Ctrli), 24 hr Dox treatment led to nearly 3-fold increase in xCT protein level possibly because xCT is a redox-sensitive gene14, 35. In comparison, both the basal and Dox-induced xCT levels were much lower in the two KD cell lines with the reduction being more pronounced in shxCT#1 (Fig. 2D), which correlated with an impaired ability of these cells to replenish their GSH store after Dox treatment (Fig. 2E). Specifically, Dox triggered a 4.8-fold GSH increase in control cells compared to only 1.9-fold in shxCT#3. In shxCT#1, the pre-treatment level was only 47% of the control cells, which crashed to a mere 10% after 24-hr Dox exposure (Fig. 2E). Taken together, these results demonstrate that xCT is an integral part of the protective, antioxidant response that ABC-DLBCL cells mount following Dox exposure.
xCT inhibition sensitizes ABC-DLBCL cell lines to Dox-induced cytotoxicity.
Next, we investigated whether inhibiting xCT could potentiate Dox-induced oxidative stress and therapeutic response in ABC-DLBCL cells. As shown in Fig. 3A & 3B, in 48-hr cell viability assays, both Riva and SuDHL2 were highly resistant to Dox when it was used as a single agent. At the highest dose of Dox tested, 2 ug, ~ 50% Riva cells and ~60% SuDHL2 cells remained viable. Yet, when the cells were pretreated with SASP at IC10 or IC25 dosages, the Dox-resistant fraction was markedly reduced to 15-17% for Riva and 22-26% for SuDHL2. We also note that in Riva cells, the maximum kill was reached at 500 nM Dox under all conditions while SuDHL2 cells showed a dose response from 100 nM to 2 uM with and without SASP combination. Corroborating the SASP effect, xCT KD by shRNA also improved Dox sensitivity (supplementary Fig. S2). To characterize the nature of cell death, we measured apoptosis based on SYTOX Green and Annexin V double staining. In all three cell lines tested, the combination of Dox and SASP increased the fraction of Annexin V+ cells relative to either Dox or SASP treatment only (Fig. 3C), confirming additive effects between these two drugs. Mechanistically, SASP synergized with Dox treatment to reduce MCL1 and increase BIM and BAX in Riva cells while BCL2 expression was largely unaffected. A similar trend was observed in SuDHL2 cells which express very little BCL2 and BIM (Fig. 3D). These changes in apoptosis regulators are consistent with more PARP cleavage being detected in cells treated with Dox/SASP combination. For c-Myc downregulation, a clear synergy between SASP and Dox was seen in Riva cells. In SuDHL2, while c-Myc was resistant to SASP, it was completely abolished by Dox and Dox/SASP combination. The pattern of p27/Kip1 alteration is unique. As single agents, both SASP and Dox reduced p27/Kip1, yet cells treated with Dox/SASP combination contained more of this protein than those in Dox treatment alone. The significance of this apparent anti-synergistic effect on p27/Kip1 is unclear (Fig. 3D). Taken together, these date suggest that xCT inhibition sensitized ABC-DLBCL cells to Dox-induced apoptosis through additive/synergistic effects on key cell death and cell cycle regulators.
Cell death triggered by Dox/SASP combination requires p38 activation.
To characterize the mechanism that connects oxidative stress and apoptosis in Dox-treated ABC-DLBCL cells, we examined the stress-activated protein kinases p38 and JNK in Riva and SuDHL2 cells 24 hours after single agent or combination treatment. As shown in Fig. 4A, untreated Riva and SuDHL2 cells did not contain activated p38, which was weakly activated by SASP alone, moderately activated by Dox alone and maximally activated by Dox/SASP combination. JNK has two isoforms, the p46 JNK1 and the p54 JNK2. In Riva cells, neither was activated by SASP but Dox activated them strongly and the Dox/SASP combination treatment activated them either further. A similar trend was seen in SuDHL2 cells except that p54 JNK2 already existed in an activated form in untreated and SASP-treated cells. Supporting the concept that a diminished GSH store is the mechanistic basis of p38 activation, exogenously supplied GSH efficiently antagonized p-p38 upregulation following Dox/SASP treatment (sFig. 4B). Next, we investigated the contribution of p38 activation to cell death with the use of SB203580, a specific p38 inhibitor. One-hour pretreatment with SB203580 was sufficient to inhibit p38 activity as evidenced by marked reduction of p-MAPKAPK-2, a downstream target protein of p38 in both Riva and SuDHL2 cells (Fig. 4B). In Riva cells, SB203580 substantially negated the SASP-associated sensitization effect on Dox cytotoxicity (Fig. 4C). In SuDHL2 cells, the SASP effect was nearly completely abolished by SB203580 pretreatment (Fig. 4D). Similar results were confirmed by direct measurement of apoptosis. Specifically, SB203580 reduced Dox/SASP-triggered apoptosis by 82% and 55%, in Riva and SuDHL2 cells, respectively (Fig. 4E). These observations indicate that p38 signaling is a major mediator of cytotoxicity triggered by Dox/SASP combination.
High xCT expression correlates with poor outcome in patients with non-GCB DLBCL but not GCB-DLBCL.
To examine the expression pattern and prognostic significance of xCT in DLBCL, we first detected xCT protein by IHC staining of diagnostic specimens from 87 DLBCL patients, all of which were treated with R-CHOP (Table 1). Representative examples of staining results are provided in Fig. 5A. Receiver operating characteristic curve analysis dichotomized samples into xCT high (n = 34) and low (n = 53) groups with a cutoff value of 75% positive staining (p = 0.046, supplementary Fig. S4). These two groups have comparable clinic-pathological features including age distribution, Ann Arbor stage, LDH level, IPI score, and COO status (Table 1). Cases with high xCT expression showed a tendency towards shorter overall survival (OS) and progression-free survival (PFS) although the difference did not reach statistical significance (OS, p = 0.273; PFS, p = 0.191, Fig. 5 B, bottom row). We then performed the same analysis after separating the cases into the two COO subtypes defined using the Han’s algorithm15. Among the 87 cases, 31 (35.6%) were classified as GCB-DLBCL while 56 (64.1%) fell into the non-GCB category. Survival analysis showed that within the non-GCB subtype, high xCT expression was significantly associated with inferior OS (5-years OS of 59.8% versus 87.0%, p = 0.038) and PFS (5-years PFS of 48.7% versus 78.4%, p = 0.043, Fig. 5B, top row). However, this prognostic association was absent among the GCB cases (5-years OS of 88.9% versus 78.0%, p = 0.299; 5-years PFS of 66.7% versus 68.4%, p = 0.543, Fig. 5B, middle row). These results suggest that the xCT-GSH axis is engaged in the therapeutic response of non-GCB DLBCLs, but not in GCB-DLBCLs.
xCT inhibition sensitizes ABC-DLBCL cell lines to Dox-induced cytotoxicity in ABC-DLBCL xenograft models.
Given our cell line-based mechanistic findings and the prognostic association of xCT among ABC-DLBCLs, we speculated that ABC (or IHC-defined non-GCB) cases with high xCT expression may have added capacity to replenish intracellular GSH stores when treated with R-CHOP, which is key to their poor treatment outcome. To test this hypothesis, we turned to mouse models. First, we evaluated the impact of xCT KD on in vivo tumor growth in NSG mice. Compared to the control Riva cell line (sh-Ctrli), the two xCT KD cell lines (sh-xCT#1 and #3) grew much more slowly (Fig 6A). In addition, at all 6 graft sites, sh-Ctrli tumors grew to > 400 mg, but only 3/6 and 1/8 tumors reached this size for the sh-xCT#1 and #3 groups (sFig 6A). Weight of tumors harvested at necropsy on Day 43 produced a similar result (sFig 6B), indicating a requirement for xCT for tumor initiation in vivo.
To generate proof-of-concept for therapeutic targeting of xCT, we next tested the CHOP/SASP combination treatment using the Riva-based NSG xenograft model. Tumor-bearing mice were randomly assigned to one of the 5 treatment arms when tumor size reached 250-300 mg, and the treatment began immediately. When used at 150 mg/kg, SASP alone did not show any activity against Riva tumors (Fig. 6B). When SASP was administered at the same time as the CHOP regimen, it did not provide any additional therapeutic benefit (Fig. 6B). Nevertheless, when SASP was used as a priming treatment and administered 3 days prior to the initiation of CHOP, a marked synergistic effect was observed after day 14 (Fig. 6B). On day 20, when compared to the CHOP response, the pre-SASP+CHOP treatment generated a tumor growth inhibition ratio (TGI) of 69.4%. These data indicate that xCT promotes in vivo growth of ABC-DLBCL cells and that targeting xCT can substantially sensitize CHOP-resistant ABC-DLBCLs to this first line regimen.