QSOX1 facilitates dormant esophageal cancer stem cells to evade immune elimination via PD-L1 upregulation and CD8 T cell exclusion

Abstract

Dormant cancer stem cells (DCSCs) exhibit characteristics of chemotherapy resistance and immune escape, and they are a crucial source of tumor recurrence and metastasis. However, the underlying mechanisms remain unrevealed. We demonstrate that enriched Gzmk⁺ CD8⁺ T cells within the niche of esophageal DCSCs restrict the outgrowth of tumor mass. Nonetheless, DCSCs can escape immune elimination by enhancing PD-L1 signaling, thereby maintaining immune equilibrium. Quiescent fibroblast-derived quiescin sulfhydryl oxidase 1 (QSOX1) promotes the expression of PD-L1 and its own expression in DCSCs by elevating the level of reactive oxygen species. Additionally, high QSOX1 in the dormant tumor niche contributes to the exclusion of CD8⁺ T cells. Conversely, blocking QSOX1 with Ebselen in combination with anti-PD-1 and chemotherapy can effectively eradicate residual DCSCs by reducing PD-L1 expression and promoting CD8⁺ T cell infiltration. Clinically, high expression of QSOX1 predicts a poor response to anti-PD-1 treatment in patients with esophageal cancer. Thus, our findings reveal a mechanism whereby QSOX1 promotes PD-L1 upregulation and T cell exclusion, facilitating the immune escape of DCSCs, and QSOX1 inhibition, combined with immunotherapy and chemotherapy, represents a promising therapeutic approach for eliminating DCSCs and preventing recurrence.

Keywords: PD-1/PD-L1; T cell exclusion; esophageal cancer stem cell; reactive oxygen species; tumor dormancy.

Fig. 1.

Gzmk⁺ CD8⁺ T cells restrict dormant tumor outgrowth. (A) The incidence of progressive and dormant tumors in C57BL/6 mice 24 wk after injection of gradiently diluted mEC2 and mEC25 cells (n = 10 mice per group). (B) Hematoxylin-Eosin (H&E) staining of progressive and dormant tumors. (C) Immunofluorescence (IF) staining of progressive and dormant tumors with antibodies against fibronectin (fibroblasts), CD45 (immune cells), and pancytokeratin (tumor cells). (D) 10× Genomics single-cell RNA sequencing was performed to analyze progressive and dormant tumors, and 10 cell types were annotated. (E) The percentages of cell types in progressive tumor (PT) and dormant tumor (DT). (F) IF staining was used to confirm T cell (CD3⁺) infiltration in PT (n = 5) and DT (n = 4). (G) The counts of T cell clusters in PT and DT. (H) The level of Granzyme K⁺ CD8⁺ T cells in PT (n = 5) and DT (n = 4) was analyzed using IF staining. (I) Foci formation assay of mEC2 and mEC25 cells with treatment of recombinant Granzyme K. (J) Antibodies were used to eliminate CD8⁺ T cells (5 mg/kg, i.v.) or neutralize GZMK (5 mg/kg, i.v.) in an animal model of tumor dormancy (n = 4 per group). (K) IF staining showed that the ratio of proliferating to apoptotic tumor cells was increased by CD8⁺ T cell elimination or GZMK neutralization. In panel (E), the P-value was calculated using one-way ANOVA, and in panels (F), (H), (I), and (K), the unpaired two-tailed Student’s t-test with Welch’s correction was performed for comparisons between two groups. **P < 0.01, ***P < 0.001; ns, no significant difference. All data are presented as mean ± SD. C, Capsule; T, Tumor cell; N, Necrotic core.

Fig. 2.

Surviving DCSCs evade immune clearance via upregulation of PD-L1. (A) UMAP visualization reveals the subpopulations of tumor cells in PT (n = 5,255 cells) and DT (n = 525 cells) with the percentage of each subpopulation presented in a pie chart. (B) Pseudotime analysis of tumor cells using Monocle2. (C) IF staining confirms the high level of CD44⁺ esophageal CSCs in PT (n = 4) and DT (n = 4). (D) A sphere formation assay was performed to evaluate the stemness of tumor cells sorted from PT (n = 6) and DT (n = 6). (E) Pathway analysis was performed based on the differential genes’ expression between tumor cells in dormant and progressive tumors. (F) A volcanic plot displays the differential genes in tumor cells between DT and PT. (G) Cell communication analysis revealed that PD-1/PD-L1 signaling (encoded by Pdcd1 and Cd274, respectively) mediates the interaction between tumor cells and T cells in dormant tumors. (H) IHC staining confirms higher PD-L1 expression in DT compared to PT. (I) Dormant tumors were generated by s.c. injection of mEC2 cells (1 × 10⁶ cells per mouse). The mice were treated with anti-PD-1 (5 mg/kg, i.v.) 60 d after injection, and the volume of the dormant tumors was measured. (J) Double IF staining analyzed the proportion of surviving tumor cells and CD8⁺ T cells in dormant tumors posttreatment. (K) The levels of proliferative (Ki67⁺) or apoptotic (Cleaved Caspase-3⁺, Cl-Casp3⁺) tumor cells were assessed by multiple IF staining. (L) Schematic diagram illustrating the immune equilibrium between esophageal DCSC and Gzmk⁺ CD8⁺ T cell in dormant tumor. For all panels, P-value analyses were performed using the unpaired two-tailed Student’s t test with Welch’s correction. *P < 0.05, **P < 0.01, ***P < 0.001. All data are presented as mean ±SD. C, Capsule; T, Tumor cell; N, Necrotic core.

Fig. 3.

QSOX1 up-regulates PD-L1 by increasing ROS levels in dormant tumors. (A) Dot plot shows that both Qsox1 and Cd274 (encoding PD-L1) are highly expressed in tumor cells in DT compared to PT. (B) Double IF staining showed the coexpression of QSOX1 and PD-L1 in DT. (C) Multiple IF staining was used to analyze the level of proliferative fibroblasts (Ki67⁺ Fibronectin⁺) in PT and DT. (D) Multiple IF staining confirmed that QSOX1 is expressed in both tumor cells and fibroblasts. (E) Western blotting revealed that recombinant mouse QSOX1 (rmQSOX1) protein treatment (10 ng/mL, 24 h) up-regulated both itself and PD-L1 in mEC2 and mEC25 cells, while QSOX1 inhibitor Ebselen abrogated this upregulation (10 μM, 24 h). (F) After homogenizing dormant and progressive tumors, the concentration of H₂O₂ in the supernatants was measured using a commercial kit. (G) The level of hydrogen peroxide (H₂O₂) in the culture supernatant of mEC2 cells was analyzed 24 h after treatment with rmQSOX1 (10 ng/mL) or/and Ebselen (10 μM). (H) The expressions of phosphorylated and total Stat3, PD-L1, and QSOX1 in mEC2 and mEC25 cells were analyzed by western blotting under treatment with H₂O₂ at gradient concentrations. (I) ChIP-qPCR confirmed that Stat3 transcriptionally regulate the expression of Qsox1 and Cd274 in mEC2 cells with treatments with rmQSOX1 (10 ng/mL, 24 h) alone or in combination with Ebselen (10 μM, 24 h). (J) Double IF staining confirmed that phosphorylated Stat3 highly coexpressed with QSOX1 or PD-L1 in DT derived from mEC2 cells. In all panels, P-value analyses were performed using the unpaired two-tailed Student’s t-test with Welch’s correction. *P < 0.05, **P < 0.01, ***P < 0.001. Error bars indicate as mean ± SD. C, Capsule; T, Tumor cell; N, Necrotic core.

Fig. 4.

Abundance of QSOX1 inhibits T cell infiltration. (A) Multiple IF staining was performed to analyze the correlation between QSOX1 expression and CD8⁺ T cell infiltration level in ESCC tissues (n = 55 images from 11 samples). (B) The infiltration level of CD8⁺ T cells in tumors derived from Qsox1- or vector-transfected mEC2 cells was analyzed by IF staining (n = 5 tumors). (C and D) Schematic diagram of T cell chemotaxis. Qsox1- or vector-transfected mEC2 and mEC25 cells were treated with Ebselen (10 μM, 24 h), and their conditioned media were used for chemotactic T cells in the µ-Slide Chemotaxis (ibidi, #80326). Blood-derived CD8⁺ T cells were prelabeled with CFSE (5 μM, 10 min). The chemotactic ability of CD8⁺ T cells was evaluated by counting the number of migrating cells under the microscope. (E and F) Conditioned media from tumor cells treated with rmQSOX1 alone (10 ng/mL, 24 h) or together with Ebselen (10 μM, 24 h) were for chemotaxis of CD8⁺ T cells, and the number of migrating CD8⁺ T cells was counted. (G) The effect of H₂O₂ on the chemotactic ability of CD8⁺ T cells was analyzed by adding H₂O₂ to the conditioned medium (100 μM, 24 h). (H) CD8⁺ T cells were treated with rmQSOX1 (5 ng/mL, 24 h) or PBS, and RNA sequencing was performed to explore the key genes regulated by QSOX1 in CD8⁺ T cells. (I) GO analysis showed the biological processes regulated by rmQSOX1 treatment. (J) Heat map showed the genes that were associated with leukocyte migration and chemotaxis. (K) RT-qPCR analysis revealed that T cell migration-related genes Itgb2 and Cxcr2 were down-regulated with rmQSOX1, which was rescued by Ebselen. In (A), Pearson correlation analysis was used. For other panels, P-value analyses were performed using the unpaired two-tailed Student’s t-test with Welch’s correction. **P < 0.01, ***P < 0.001. Data are shown as mean ± SD.

Fig. 5.

Blocking both QSOX1 and PD-1 in combination with chemotherapy eliminates dormant tumors. (A) The protein level of QSOX1 in esophageal squamous cell carcinoma (ESCC) tissues from nonresponders (n = 13) or responders (n = 15) with anti-PD-1 treatment (Toripalimab) was analyzed by IHC staining. (B) Multiple IF staining confirmed the expression of QSOX1 in both fibroblasts and tumor cells in ESCC tissues from nonresponders. (C) The level of proliferative fibroblasts (pan-CK^- Fibronectin⁺ Ki67⁺) in ESCC tissues from nonresponders (n = 6) and responders (n = 6) were analyzed by multiple IF staining. (D) Dormant tumors were generated by s.c. injection of mEC2 cells (1 × 10⁶ cells per mouse). The mice were treated with Ebselen (10 mg/kg, i.p.) alone or in combination with anti-PD-1 (5 mg/kg, i.v.) or chemotherapy (paclitaxel, 6 mg/kg, i.p. + cisplatin, 6 mg/kg, i.p.) four times 56 d after the tumor cells were injected. Carboxymethylcellulose sodium (1% CMC-Na, i.p.) was used as a vehicle for Ebselen. Concurrently with the treatment, the volume of the dormant tumors was measured. (E) The levels of proliferative (Ki67⁺) or apoptotic (Cl-Casp3⁺) tumor cells were assessed by multiple IF staining. (F) Double IF staining analyzed the proportion of CD8⁺ T cells in dormant tumors posttreatment. For all panels, P-values were analyzed using the unpaired two-tailed Student’s t-test with Welch’s correction. *P < 0.05, **P < 0.01, ***P < 0.001; ns, no significant difference. Data are shown as mean ± SD. C, Capsule; T, Tumor cell; N, Necrotic core.

Fig. 6.

Inhibiting QSOX1 with Ebselen in combination with anti-PD-1 and chemotherapy eradicates residual DCSCs. Quiescent fibroblast-derived QSOX1 shapes the dormant tumor milieu into a highly oxidative state by producing H₂O₂, which up-regulates the expression of PD-L1 and itself in DCSCs and facilitates T cell exclusion, resulting in immune equilibrium between DCSCs and PD-1⁺ Gzmk⁺ CD8⁺ T cells. Inhibiting QSOX1 with Ebselen, in combination with anti-PD-1 and chemotherapy, enhances anti-tumor activity and promotes the infiltration of PD-1⁺ Gzmk⁺ CD8⁺ T cells, thereby eradicating residual DCSCs in dormant tumors and preventing relapse.