Noncanonical UPR factor CREB3L2 drives immune evasion of triple-negative breast cancer through Hedgehog pathway modulation in T cells

Abstract

The unfolded protein response (UPR) pathway is crucial for tumorigenesis, mainly by regulating cancer cell stress responses and survival. However, whether UPR factors facilitate cell-cell communication between cancer cells and immune cells to drive cancer progression remains unclear. We found that adenosine 3′,5′-monophosphate response element–binding protein 3–like protein 2 (CREB3L2), a noncanonical UPR factor, is overexpressed and activated in triple-negative breast cancer, where its cleavage releases a C-terminal fragment that activates the Hedgehog pathway in neighboring CD8+ T cells. The enhanced Hedgehog pathway represses CD8+ T cell activation and inhibits its cytotoxic effects. Consequently, overexpression of CREB3L2 not only promotes tumor growth but also causes resistance to immune checkpoint blockade (ICB). Inhibition of the Hedgehog pathway impedes the growth of CREB3L2-overexpressed tumors and sensitizes them to ICB therapy. In summary, we identified a previously unidentified mechanism by which the UPR pathway dictates cross-talk between cancer cells and immune cells, providing important anticancer therapeutic opportunities.

Cancer cells secrete UPR factor CREB3L2 to repress cytotoxic T cells by activating the Hedgehog pathway.

INTRODUCTION

Cross-talk between cancer cells and immune cells plays a critical role in driving cancer progression (1). Immune cells, exposed to extracellular factors released by tumors and influenced by the metabolic conditions within the tumor microenvironment (TME), activate specific pathways that can impair their antitumor functions (2). At the same time, cancer cells are subjected to constant stress and must adapt to these challenging conditions. Despite this, the intricate relationship between stress-induced signaling in cancer cells and the regulation of antitumor immune responses remains largely unexplored.

During cancer progression, the transformed cells must cope with increased demands for protein and lipid production essential for their rapid proliferation while also adapting to stressful TMEs like hypoxia, nutrient deprivation, and acidosis (3, 4). To overcome these challenges, cancer cells activate and use a set of signaling pathways, notably the unfolded protein response (UPR), to ensure their survival (5, 6). The classical UPR involves three main branches that are initiated by three endoplasmic reticulum (ER)–anchored proteins, including IRE1 (inositol-requiring enzyme 1), PERK (protein kinase RNA-like ER kinase), and ATF6 (activating transcription factor 6). Activation of the UPR pathway facilitates proper protein folding, modification, transport, and the degradation of misfolded proteins, thereby alleviating ER stress in cancer cells (7–9). Beyond maintaining ER homeostasis, the UPR pathways are well established as crucial for the initiation and progression of cancer (10). These pathways also suppress apoptosis and promote angiogenesis (11, 12). However, the mechanisms by which cancer cells use the UPR to modulate the tumor immune microenvironment remain underexplored. It is currently known that both the IRE1 and ATF6 pathways control the expression of programmed death-ligand 1 (PD-L1), promoting immune evasion, while the PERK pathway facilitates immune evasion via regulation of immunologic cell death (13–16). In addition, the UPR pathways regulate the expression and secretion of pro-inflammatory cytokines, such as interleukin-6 (IL-6) and IL-8, which in turn modulate the function of intratumoral immune cells (17–19). Further studies are needed to better understand the cross-talk between cancer cells and immune cells mediated by the UPR pathways.

Beyond the three classical branches of the UPR, several other proteins are induced and activated upon ER stress. Notably, the adenosine 3′,5′-monophosphate response element–binding protein 3 (CREB3) family, comprising CREB3, CREB3-like protein 1 (CREB3L1), CREB3L2, CREB3L3, and CREB3L4, plays critical roles in maintaining cellular homeostasis under conditions of ER and Golgi stress (20, 21). These proteins constitute a major group of “noncanonical” UPR factors, and they are activated through regulated intramembrane proteolysis (RIP) (22), a mechanism they share with ATF6. Upon activation, these proteins are transported from the ER to the Golgi apparatus, where they undergo RIP mediated by Site-1 and Site-2 proteases (S1P and S2P, respectively). This process generates both an N-terminal fragment and a C-terminal fragment (23). The N terminus of the CREB3 family functions as a transcription factor, driving gene expression involved in stress response, extracellular matrix production, and cancer cell metastasis (24–26). The C-terminal fragment of CREB3L2 can also be secreted as an extracellular factor, directly influencing chondrocyte proliferation and axon growth (27–29). This represents a distinct mechanism through which the UPR pathway contributes to both cellular physiology and overall homeostasis.

In this study, we focus on CREB3L2 and its role in regulating cell-cell communication to promote immune evasion in human triple-negative breast cancers (TNBCs), a subtype that lacks effective targeted therapies. We found that CREB3L2 is not only overexpressed but also hyperactivated in TNBCs. Although CREB3L2 does not influence the proliferation of TNBC cells in vitro, it significantly promotes tumor growth in immunocompetent animals. The pro-tumorigenic effects of CREB3L2 are T cell dependent, as depletion of T cells completely abolishes its tumor-promoting functions. Mechanistically, activation of CREB3L2 in TNBC cells leads to the release of its C terminus, which subsequently activates the Hedgehog (Hh) pathway in CD8+ T cells. This activation represses CD8+ T cell activity, ultimately reducing their cytotoxic effects on TNBC cells. Collectively, our study proposes a previously unknown model for how a stress-responsive factor modulates the tumor immune microenvironment and highlights promising opportunities for targeted cancer therapies.

RESULTS

CREB3L2 is overexpressed and associated with CD8+ T cell status in TNBC

We first investigated the expression of CREB3L2 in human TNBC. Using 50 pairs of human TNBC samples and the matched adjacent mammary tissues, we found that the mRNA level of CREB3L2 was significantly increased in TNBCs compared to their normal counterparts (Fig. 1A). As an ER-anchored transmembrane protein, the full-length CREB3L2 (FL-CREB3L2) can be cleaved into an N-terminal fragment [N-terminal CREB3L2 (N-CREB3L2)] and a C-terminal fragment [C-terminal CREB3L2 (C-CREB3L2)] upon activation (Fig. 1B). By use of the protein lysate of TNBC samples, we found that CREB3L2 was not only increased but also activated in TNBCs, as gauged by the expression of the full-length and post-cleaved forms (N-terminal) of CREB3L2 (Fig. 1C). We next determined the upstream factors that can up-regulate CREB3L2 in TNBCs. While CREB3L2 is an ER stress–responsive protein, it is unknown which branch of the UPR facilitates the induction of CREB3L2 (30). It has been reported that the expression and activation of X-box binding protein 1 (XBP1) are increased in TNBCs (10). We found that XBP1 promoted the expression of CREB3L2 in TNBC cells (fig. S1, A and B). The expression of CREB3L2 was not dependent on ATF4 or ATF6, which is different from the regulation of CREB3L1, another member of the CREB3 family (26). TNBC is also well recognized as a mesenchymal-like subtype of breast cancer, which usually has a heightened epithelial-mesenchymal transition (EMT) signature (31, 32). We applied the human mammary luminal epithelial (HMLE) cells (33–35), a well-established EMT model in breast cancer, to interrogate whether EMT was also involved in the regulation of CREB3L2. We observed a notable increase in CREB3L2 expression in HMLE cells following EMT induction regardless of the induction strategy used (fig. S1C). Collectively, these data suggest that the expression of CREB3L2 is induced by both the UPR pathway and EMT program.

Fig. 1. Expression of CREB3L2 in TNBCs and its association with intratumoral CD8+ T cells.

(A) qPCR analysis showing the relative expression of CREB3L2 in 50 pairs of TNBC samples and their adjacent normal mammary tissues. (B) Schematic of the RIP during activation of CREB3L2. (C) Western blots showing the expression of the full-length (FL) and the cleaved N-terminal (N-ter) CREB3L2 in 10 pairs of TNBC samples and their adjacent normal mammary tissues. (D) Representative IHC staining of CD8 and CREB3L2 in TNBC tissues from four patients. Scale bar, 100 μm. (E) Quantification of CREB3L2 expression in 48 patients with TNBC stratified by CD8+ T cell infiltration. See Materials and Methods for details. (F) Kaplan-Meier survival analysis showing the relapse-free survival of patients with breast cancer stratified by the expression of CREB3L2 in tumors (adapted from KmPlot using the data from The Cancer Genome Atlas for patients exhibiting breast cancer with CD8+ T cell infiltration). Data are presented as means ± SEM. Paired, two-tailed Student’s t test was applied for (A). One-way ANOVA was applied for (E). Log-rank test was applied for (F). **P < 0.01.

We then examined whether CREB3L2 was associated with the malignant characteristics of TNBC. While overexpressed in TNBCs, the expression level of CREB3L2 in cancer cells did not tend to correlate with the tumor size or disease stage of TNBCs (fig. S1D), suggesting that CREB3L2 might not directly affect TNBC cell proliferation. We found an inverse correlation between CREB3L2 expression and the infiltration of CD8+ T cells in TNBCs (Fig. 1, D and E). While CREB3L2 was generally highly expressed in TNBCs, it was notable that the expression of CREB3L2 was relatively low in TNBC tissues that have high infiltration of CD8+ T cells (Fig. 1E). To further investigate the potential link between CREB3L2 expression and T cell status in human breast cancer, we analyzed data from The Cancer Genome Atlas (36) and found that, in CD8+ T cell–infiltrated breast tumors, the higher expression of CREB3L2 was significantly correlated with shorter relapse-free survival (Fig. 1F). Together, these data demonstrate that CREB3L2 is overexpressed in TNBC and associated with CD8+ T cell infiltration.

CREB3L2 promotes tumor growth in an immune-dependent manner via its C-terminal fragment

We next set out to investigate the role of CREB3L2 in TNBC. In a panel of murine breast cancer cell lines, we found that CREB3L2 was expressed in all the lines tested, with the highest expression in D2A1, an aggressive TNBC line (37, 38) (Fig. 2A). It was notable that the expression of CREB3L2 was associated with the expression of mesenchymal genes in these lines (fig. S2A). Two CREB3L2-targeted short hairpin RNAs (shRNAs) can effectively repress the expression of CREB3L2 in D2A1 (Fig. 2B). Unexpectedly, reduced CREB3L2 expression in D2A1 did not inhibit cell proliferation of the D2A1 cells (fig. S2B). Moreover, reduced CREB3L2 expression did not affect EMT progression in D2A1 cells either (fig. S2C). As a member of the noncanonical UPR pathway, we wondered whether CREB3L2 participates in responding to ER stress. Using thapsigargin as an ER stress inducer, we found that loss of CREB3L2 did not affect the response of D2A1 cells to ER stress (fig. S2, D and E). These results suggest that CREB3L2 might not affect the in vitro growth or mediate the UPR response of TNBC cells.

Fig. 2. C-CREB3L2 promotes tumor growth in an immune-dependent manner.

(A) Western blots showing the expression and activation of CREB3L2 in four murine breast cancer cell lines. FL-CREB3L2 and N-CREB3L2 were immunoblotted by an α-N-terminal CREB3L2 antibody (α-N-ter). Cleaved C-CREB3L2 (C-ter) was immunoblotted by an α-C-terminal CREB3L2 antibody (α-C-ter). (B) Western blots showing the expression of CREB3L2 in D2A1 cells transduced with a control shRNA (shScram) or two shRNAs targeting CREB3L2 (sh1 and sh2). (C) Quantification of in vivo tumor growth of D2A1 cells transduced with indicated shRNAs and implanted into BALB/c mice (n = 8 for each group of tumors). (D) Images of tumors (left) and weight measurement of tumors (right) collected from (C). (E) Quantification of in vivo tumor growth of D2A1 cells transduced with indicated shRNAs and implanted into BALB/c-nude mice (n = 10 for each group of tumors). (F) Schematic of the structure of full-length, N-terminal, and C-terminal fragments of the CREB3L2 proteins cloned into lentiviral constructs and used in gain-of-function studies. (G) Quantification of in vivo tumor growth (left) and weight of tumors (right) formed by pB2 cells transduced with a control vector or constructs that overexpress FL-CREB3L2, N-CREB3L2, or C-CREB3L2 (n = 16 for each group of tumors). (H) Western blots showing the expression of CREB3L2 in D2A1 cells transduced with a control shRNA (Ctrl), an shRNA targeting CREB3L2 (sh1), a construct overexpressing C-CREB3L2, or a combination of sh1 and C-CREB3L2. (I) Quantification of in vivo tumor growth (left) and weight of tumors (right) formed by D2A1 cells transduced with indicated constructs (n = 10 for each group of tumors). Data are presented as means ± SEM. Unpaired, two-tailed Student’s t test was applied for (D), (G), and (I) (right panel). Two-way ANOVA was applied for (C), (E), (G), and (I) (left panel). *P < 0.05 and **P < 0.01. ns, not significant.

We then tested the role of CREB3L2 in TNBC growth in vivo. In the immunocompetent BALB/c mice, reduced CREB3L2 expression markedly repressed tumor growth of the D2A1 cells in vivo (Fig. 2, C and D). In stark contrast, loss of CREB3L2 did not affect D2A1 tumor growth in nude mice, which lack functional T lymphocytes (Fig. 2E and fig. S3A). Consistently, loss of CREB3L2 also repressed the tumor growth of 4T1, another TNBC line, only in immunocompetent BALB/c mice (fig. S3, B to D). A similar dependency of CREB3L2’s function on T cells was also observed in B16, a melanoma tumor model (fig. S3, E to H). Therefore, we proposed that CREB3L2 affects tumor growth in a T cell–dependent manner.

To further validate these findings, we tested the in vivo role of CREB3L2 using gain-of-function strategies. CREB3L2 functions either through its N-terminal fragment (N-CREB3L2), which enters the nucleus and serves as a transcription factor to drive target gene expression, or through its C-terminal fragment (C-CREB3L2), which is secreted via the Golgi-initiated secretory vesical and serves as an extracellular factor (Fig. 1B). Accordingly, we transduced all three forms of CREB3L2—full-length, N-terminal, or C-terminal fragment—into pB2, a cell line with low CREB3L2 expression (Fig. 2F). The protein product of the FL-CREB3L2 construct undergoes cleavage at the Golgi apparatus, where it loses its transmembrane domain, generating an N-terminal fragment and a C-terminal fragment. To ensure the secretion of the truncated C-CREB3L2, the signal peptide of the binding immunoglobulin protein (BiP) was added into the C-CREB3L2 construct such that the protein product of this construct can be detected in the culture media (fig. S3I). In immunocompetent C57BL/6 mice, expression of both FL-CREB3L2 and C-CREB3L2 significantly promoted tumor growth, with C-CREB3L2 exhibiting a more potent effect than FL-CREB3L2 (Fig. 2G). In contrast, overexpression of N-CREB3L2 did not affect tumor growth, suggesting that the C terminus of CREB3L2 primarily mediates its tumor-promoting functions. Consistent with the findings in D2A1 cells, overexpression of CREB3L2 did not affect tumor growth of pB2 cells in nude mice (fig. S3J). To further confirm the role of C-CREB3L2, we re-expressed C-CREB3L2 in CREB3L2-depleted D2A1 cells to examine whether the C-terminal fragment could rescue the defects caused by CREB3L2 loss. As expected, re-expression of C-CREB3L2 effectively restored tumor formation of the CREB3L2-depleted D2A1 cells (Fig. 2, H and I, and fig. S3K). Together, these results demonstrate that CREB3L2 promotes tumor growth in an immune-dependent manner via its C terminus.

Cancer cell–derived CREB3L2 modulates CD8+ T cell infiltration to regulate tumor growth

Next, we pinpointed how CREB3L2 leverages the host immune microenvironment to facilitate tumor growth. Considering the notable difference between the immunocompetent and T cell defective mouse models, we analyzed the abundance of CD8+ and FOXP3+ T cells in D2A1 tumors. Using tumor sections collected from the D2A1 tumors, we found that the number of CD8+ T cells, but not regulatory T cells, was significantly increased upon CREB3L2 loss (Fig. 3A and fig. S4A). Consistent with the results of immunohistochemistry (IHC) staining, the percentage of CD8+ T cells was augmented in the CREB3L2-deficient D2A1 tumors compared to the controls, as gauged by flow cytometry analysis (Fig. 3B and fig. S4B). Similarly, the reduction of CREB3L2 expression in cancer cells also led to up-regulation of CD8+ T cells in the 4T1 and B16 tumors (Fig. 3C and fig. S4C). On the other hand, overexpression of CREB3L2, either the full-length protein or the C-terminal fragment, significantly reduced CD8+ T cell infiltration in the pB2 tumors (fig. S4D). Re-expression of C-CREB3L2 in the CREB3L2-deficient D2A1 cells reduced the elevated CD8+ T cells in the CREB3L2-deficient tumors (Fig. 3D and fig. S4E). These results clearly showed that expression and activation of CREB3L2 were strongly associated with the reduction of CD8+ T cells in tumors. To test whether CD8+ T cells were the key element mediating the regulatory function of CREB3L2 in tumors, we applied an α-CD8 antibody to deplete CD8+ T cells. We found that inhibition of tumor growth caused by loss of CREB3L2 was effectively rescued by CD8+ T cell depletion (Fig. 3, E and F). These findings indicate that expression and activation of CREB3L2 in tumor cells repress intratumoral infiltration of CD8+ T cells to promote tumor development.

Fig. 3. CREB3L2 modulates intratumoral infiltration of CD8+ T cells.

(A) IHC staining to quantify CD8+ T cells in tumors formed by D2A1 cells transduced with a control shRNA (shScram) or two shRNAs targeting CREB3L2 (sh1 and sh2). (Left) Representative images; (right) quantification of CD8+ T cell infiltration (n = 10 sections for each type of tumor). Scale bar, 100 μm. (B) Flow cytometry analysis to quantify CD8+ T cells in tumors of (A). (Left) Representative flow cytometry plots showing the distribution of CD4+ and CD8+ T cells in the lymphocyte subpopulation in the indicated tumors; (right) quantification of infiltrated CD8+ T cells in the tumor mass (n = 5 for each type of tumor). (C) Quantification of IHC staining (left) and flow cytometry analysis (right) to quantify CD8+ T cells in tumors formed by B16 cells transduced with indicated shRNAs (n = 10 sections for each type of tumor in IHC and n = 5 for each type of tumor analyzed by flow cytometry). (D) Quantification of IHC staining (left) and flow cytometry analysis (right) to quantify CD8+ T cells in tumors formed by D2A1 cells transduced with a control shRNA (shScram), an shRNA targeting CREB3L2 (sh1), a construct overexpressing C-CREB3L2, or a combination of sh1 and C-CREB3L2 (n = 10 sections for each type of tumor in IHC and n = 5 for each type of tumor analyzed by flow cytometry). (E) Quantification of in vivo tumor growth of the D2A1 cells transduced with a control shRNA (shScram) or an shRNA targeting CREB3L2 (sh1), treated with or without α-CD8 antibodies every 3 days (n = 10 for each group of tumors). (F) Quantification of the weight of tumors formed by D2A1 cells as in (E) (n = 10 for each group of tumors). Data are presented as means ± SEM. Unpaired, two-tailed Student’s t test was applied for (A) to (D) and (F). Two-way ANOVA was applied for (E). *P < 0.05 and **P < 0.01.

CREB3L2 directly represses the antitumor activity of CD8+ T cells via its C-terminal fragment

Increased intratumoral CD8+ T cell infiltration could be attributed to enhanced CD8+ T migration or enhanced CD8+ T activation, which can promote T cell proliferation and expansion. By using the conditioned media produced by D2A1 cells with altered CREB3L2 expression, we found that the expression of CREB3L2 in cancer cells did not affect the migration of primary CD8+ T cells (fig. S5A). We then examined whether CREB3L2 modulates CD8+ T cell activity. We applied an in vitro lymphocyte killing assay, in which cancer cells were cocultured with spleen-derived, primary CD8+ T cells, to assess the cytotoxic effects of T cells against cancer cells with varying levels of CREB3L2 expression. While α-CD3/CD28 antibody–activated CD8+ T cells exhibited a modest killing effect on the control D2A1 cells, the same number of CD8+ T cells nearly eradicated the D2A1 cells with CREB3L2 depletion (Fig. 4A). Re-expression of C-CREB3L2 mitigated the cell death caused by CD8+ T cells (Fig. 4, A and B). To confirm this phenotype, we labeled D2A1 cells with a blue fluorescence protein (BFP) and repeated the coculture assay to examine the percentage of apoptotic cells in the BFP-positive population. We found that loss of CREB3L2 resulted in a more pronounced cancer cell death triggered by CD8+ T cells (Fig. 4C and fig. S5B). A similar vulnerability to T cell–mediated killing was observed in 4T1 cells, confirming the role of CREB3L2 in TNBCs (fig. S5C). Consistently, the expression of CREB3L2 in B16 cells did not affect T cell migration but strongly repressed the cytotoxicity of CD8+ T cells (fig. S5, D to F). Moreover, addition of C-CREB3L2 protein promoted the survival of D2A1 cells when cocultured with CD8+ T cells (Fig. 4D and fig. S5, G and H), whereas clearance of C-CREB3L2 proteins by a neutralizing antibody (28) reversed this effect (Fig. 4E). Moreover, inhibiting the cleavage of CREB3L2 by AEBSF, an S1P inhibitor, also effectively promoted T cell killing against D2A1 cells (fig. S5I). Collectively, these data suggest that the cancer cell–secreted C terminus of CREB3L2 represses the antitumor activity of CD8+ T cells.

Fig. 4. C-CREB3L2 represses the cytotoxic effects of CD8+ T cells.

(A) D2A1 cells transduced with indicated constructs were cocultured with α-CD3/CD28 antibody–activated CD8+ T cells (Act CD8+ T) at different ratios for 3 days. The remaining live D2A1 cells were stained with crystal violet, and representative images were shown. Scale bar, 200 μm. (B) Quantification of remaining live D2A1 cells after coculture (10:1) with or without naïve or active CD8+ T cells (n = 3). (C) BFP-labeled D2A1 cells transduced with indicated constructs were cocultured with active CD8+ T cells for 3 days. Apoptosis of BFP-positive cells was measured by flow cytometry analysis (n = 3). (D) Quantification of remaining live D2A1 cells after coculture with or without CD8+ T cells in the presence of control bovine serum albumin (BSA) or C-CREB3L2 proteins for 3 days (n = 3). (E) Quantification of remaining live D2A1 cells after coculture with CD8+ T cells treated under indicated conditions (n = 3). (F) Quantification of T cells cocultured with D2A1 cells transduced with indicated constructs (n = 3). (G) Representative images (left) and quantifications (right) of the CD25+CD8+ population in the CD8+ T cells after coculture with D2A1 cells (n = 3). (H) qPCR results showing the expression of Gzmb in CD8+ T cells cocultured with indicated D2A1 cells (n = 5). The expression of Gzmb was normalized to GAPDH, and then the relative expression of Gzmb in each sample was normalized to the mean value of the control sample (naïve CD8+ T cocultured with D2A1-shScram). (I) qPCR results showing the expression of IFN-γ in CD8+ T cells cocultured with D2A1 cells (n = 5). The normalization strategy was the same as (H). (J) qPCR results showing the expression of Gzmb and IFN-γ in CD8+ T cells cocultured with D2A1 cells in the presence of indicated conditions (n = 3). Data are presented as means ± SEM. Samples are biological replicates. Unpaired, two-tailed Student’s t test was applied. *P < 0.05 and **P < 0.01.

We then further characterized the change of CD8+ T cells upon coculture with D2A1 cells with altered expression of CREB3L2. First, coculture with CREB2L2-depleted D2A1 cells modestly promoted the proliferation of CD8+ T cells without affecting the apoptosis of CD8+ T cells (Fig. 4F and fig. S6, A and B). Meanwhile, the percentage of CD25+CD8+ T cells was significantly increased upon coculture with CREB3L2-depleted cells, suggesting that T cell activation was repressed by CREB3L2 expression in cancer cells (Fig. 4G). The expression of Gzmb and IFN-γ, two marker genes of T cell activation, was markedly increased upon coculture with CREB3L2-depleted D2A1 cells, which was also reversed by re-expression of C-CREB3L2 (Fig. 4, H and I). Coculture with 4T1 cells caused a similar change of marker gene expression in CD8+ T cells (fig. S6C). Consistently, addition of the C-terminal fragment of the CREB3L2 protein effectively repressed the expression of these two marker genes, which could be reversed by a neutralizing antibody (Fig. 4J). Together, these findings strongly suggest that the C-terminal fragment of CREB3L2 from the TNBC cells represses CD8+ T cell activation.

CREB3L2 promotes Hh pathway activation in CD8+ T cells

Next, we sought to dissect how C-CREB3L2 affects CD8+ T cell activation. As a secreted factor, it has been reported that the C terminus of CREB3L2 can directly promote the interaction between Hh ligands and their receptor Patched-1 (Ptch1) to activate the Hh signaling in chondrocytes and axons (27, 29). Unexpectedly, we found that loss of CREB3L2 did not change the expression of Hh signaling marker genes in the D2A1 cells (fig. S7A), although these results were consistent with the findings that CREB3L2 did not regulate D2A1 cell proliferation (fig. S2B).

Therefore, we wondered whether the cancerous expression of CREB3L2 promotes the Hh signaling in T cells in a paracrine manner. By measuring the expression of Gli1, Ptch1, and FoxM1, marker genes of the Hh signaling, we found that the Hh pathway was inhibited during activation of primary CD8+ T cells, implying a negative role of the Hh pathway in CD8+ T cell activation (Fig. 5, A and B). Notably, coculture with D2A1 cells increased the expression of Hh genes in CD8+ T cells. In contrast, reduction of CREB3L2 expression in D2A1 cells significantly abolished the surge of Hh gene expression (Fig. 5C). Conversely, re-expression of the C-CREB3L2 fragment or addition of the C-CREB3L2 proteins effectively increased the expression of Hh genes in CD8+ T cells (Fig. 5, C and D, and fig. S7, B and C). Consistent with these in vitro results, we found that the expression of Hh genes in CD8+ T cells infiltrated in D2A1 tumors was reduced when the CREB3L2 expression in cancer cells was repressed by shRNAs (Fig. 5, E and F). Similarly, re-expression of C-CREB3L2 reversed the reduced expression of Hh genes in D2A1 tumor–derived CD8+ T cells (Fig. 5G). To further confirm whether exogenous CREB3L2 altered the Hh signaling in T cells, we transfected a Gli-responsive reporter in the EL4 T cell line before coculture with the D2A1 cells. As expected, coculture with control D2A1 cells markedly increased the transcriptional activity of the Gli reporter, while depletion or overexpression of CREB3L2 in D2A1 cells affected the reporter activity accordingly, indicating that the cleaved form of CREB3L2 directly modulated the Hh signaling in T cells (Fig. 5H). On the mechanistic level, it has been reported that C-CREB3L2 can bind to Hh ligands and promote the Hh ligand-Ptch1 interaction (27, 29). Consistent with these previous reports, we found that the C-CREB3L2 protein disrupted the interaction between Ptch1 and Smo in primary CD8+ T cells and EL4 T cells (Fig. 5I and fig. S7D), which could cause Smo activation and trigger the downstream Hh signaling. Together, these data demonstrate that the cancerous expression of CREB3L2 activates the Hh signaling in CD8+ T cells.

Fig. 5. CREB3L2 promotes the activation of the Hh pathway in CD8+ T cells.

(A) qPCR results showing the expression of three Hh pathway genes, Ptch1, Gli1, and FoxM1, in freshly isolated splenocytes, naïve CD8+ T cells, and α-CD3/CD28 antibody–activated CD8+ T cells (n = 3). (B) Western blots showing the expression of Ptch1 and Gli1 in cells from (A). (C) qPCR results showing the expression of Ptch1 and Gli1 in naïve CD8+ T cells, active CD8+ T cells, or active CD8+ T cells cocultured with D2A1 cells transduced with indicated constructs (n = 3). (D) qPCR results showing the expression of Ptch1 and Gli1 in CD8+ T cells cocultured with D2A1 cells in the presence of BSA, C-CREB3L2 proteins, or C-CREB3L2 proteins combined with α-CREB3L2 antibodies (n = 3). (E) qPCR results showing the expression of Ptch1 and Gli1 in CD8+ T cells isolated from tumors formed by D2A1 cells transduced with indicated constructs (n = 3 for each type of tumor). (F) Western blots showing the expression of Ptch1 and Gli1 in CD8+ T cells isolated from tumors formed by D2A1 cells as in (E) (n = 2 for each type of tumor). (G) qPCR results showing the expression of Ptch1 and Gli1 in CD8+ T cells isolated from tumors formed by D2A1 cells transduced with indicated constructs (n = 3 for each type of tumor). (H) The EL4 cells were transduced with a Gli-responsive luciferase reporter and a Renilla control reporter before coculture with or without indicated D2A1 cells. The Gli luciferase activity of EL4 cells was measured by dual-luciferase assays (n = 5). (I) CD8+ T cells were treated with BSA or C-CREB3L2 proteins. Using an α-Smo antibody, a co-IP assay was conducted to pull down the Smo protein, and the Smo-bound Ptch1 was measured by Western blotting. Data are presented as means ± SEM. Samples are biological replicates. Unpaired, two-tailed Student’s t test was applied. *P < 0.05 and **P < 0.01.

Pharmacological inhibition of the Hh pathway activates CD8+ T cells to repress tumor growth and sensitize tumors to α-PD-1 therapies

We next asked whether activation of the Hh pathway mediates the repressive role of CREB3L2 in regulating CD8+ T cells. We applied both an agonist [smoothened agonist (SAG)] and an antagonist (Gant61) of the Hh pathway to examine whether modulation of the Hh pathway could affect CD8+ T cell activation. By measuring the expression of Hh marker genes, we confirmed that treatment of SAG and Gant61 could activate or inhibit the Hh pathway, respectively (Fig. 6A). Activation of the Hh pathway reduced the percentage of CD25+ population in the CD8+ T cells, whereas inhibition of the Hh pathway increased the CD25+ population (Fig. 6B and fig. S8A). Consistently, the expression of Gzmb and IFN-γ was changed accordingly in the CD8+ T cells upon alteration of the Hh pathway, indicating that T cell activation was negatively regulated by the Hh pathway (Fig. 6C). Activating the Hh pathway by SAG weakened the cytotoxic effects of T cells against the CREB3L2-depleted D2A1 cells (Fig. 6D), while inhibiting the Hh pathway by Gant61 markedly sensitized D2A1 cells to T cell–mediated killing regardless of CREB3L2 overexpression (Fig. 6E). As a control, the same concentration of SAG (500 nM) and Gant61 (5 μM) only mildly affected D2A1 cell survival (fig. S8B). These results suggest that activation of the Hh pathway by CREB3L2 mediates its repressive effect on CD8+ T cell activation.

Fig. 6. The Hh pathway in CD8+ T cells mediates the pro-tumorigenic function of CREB3L2 in cancer cells.

(A) qPCR results showing the expression of Ptch1 and Gli1 in naïve CD8+ T cells, active CD8+ T cells treated with or without 500 nM SAG or 5 μM Gant61, or active CD8+ T cells cocultured with D2A1 cells in the presence of solvent control, SAG, or Gant61 (n = 3). The expression of Ptch1 and Gli1 was normalized to GAPDH, and then the relative gene expression was normalized to the control sample (naïve CD8+ T). (B) Flow cytometry analysis showing the percentage of the CD25+CD8+ population in CD8+ T cells treated with or without SAG or Gant61 (n = 4). (C) qPCR results showing the expression of Gzmb and IFN-γ in active CD8+ T cells cocultured with D2A1 cells in the presence of solvent control, SAG, or Gant61 (n = 3). (D and E) Quantification of D2A1 cell survival after a 3-day coculture with active CD8+ T cells in the presence of solvent control, SAG (D; n = 3), or Gant61 (E; n = 3). (F) Quantification of in vivo tumor growth of the indicated D2A1 cells treated with or without Gant61 (50 mg/kg) every 3 days in nude mice (left) or BALB/c mice (right) (n = 10 per group). (G) Index of tumor growth inhibition (TGI) of Gant61 in suppressing the formation of C-CREB3L2–overexpressing tumors in (F) (n = 10). (H) Quantification of in vivo tumor growth of the indicated D2A1 cells treated with or without α-PD-1 antibodies (5 mg/kg) every 3 days (n = 10 per group). (I) Quantification of in vivo tumor growth of the D2A1 cells treated with solvent control, Gant61 (50 mg/kg), α-PD-1 antibodies (5 mg/kg), or a combination of Gant61 and α-PD-1 antibodies every 3 days (n = 10 per group). Data are presented as means ± SEM. Samples are biological replicates. Unpaired, two-tailed Student’s t test was applied for (A) to (E). Two-way ANOVA was applied for (F) to (I). *P < 0.05 and **P < 0.01.

We then leveraged the link between CREB3L2-Hh signaling and T cell activation to suppress tumor growth. Consistent with a previous report (39), inhibition of the Hh pathway by Gant61 was able to repress tumor formation in nude mice, indicating that the cancerous Hh pathway was required for tumor growth in vivo (Fig. 6F, left panel). Notably, the administration of Gant61 repressed the expression of Hh pathway genes in intratumoral CD8+ T cells in the immunocompetent BALB/c mice, thereby increasing the infiltration of CD8+ T cells and promoting their activity (fig. S8, C to E). Therefore, the Gant61 treatment led to more pronounced inhibitory effects against tumor growth in BALB/c mice, especially for the C-CREB3L2–overexpressed D2A1 tumors (Fig. 6F, right panel). The tumor growth inhibition by Gant61 was increased from 40% in nude mice to nearly 70% in BALB/c mice, suggesting that blockage of the Hh pathway in T cells provided additional inhibitory effects against TNBCs with increased CREB3L2 expression (Fig. 6G).

Last, we aimed to test whether the CREB3L2-Hh signaling affects the response of TNBC cells to immune checkpoint blockade (ICB). TNBC is known as a type of tumor commonly resistant to ICB therapies. Consistent with previous reports, the D2A1 tumors were not responsive to α-PD-1 treatment. Notably, loss of CREB3L2 not only reduced tumor growth but also sensitized the D2A1 tumors to α-PD-1 therapies (Fig. 6H). Likewise, intervening the CREB3L2-Hh signaling by Gant61 also significantly enhanced the efficacy of ICB in treating the D2A1 tumors (Fig. 6I). Together, these findings indicate that inhibition of the Hh pathway activates CD8+ T cells, thereby repressing the growth of CREB3L2-overexpressed tumors and sensitizing the tumors to ICB therapies.

DISCUSSION

Our research identified an important mechanism of tumor immune evasion. During tumor development, cancer cells frequently experience stress because of adverse microenvironmental conditions, such as nutrient deprivation and hypoxia. Paradoxically, the signaling pathways activated by these stress conditions can promote tumor progression. The UPR pathway is a well-established example of this phenomenon (6). UPR activation not only helps cells maintain ER homeostasis but also directly influences tumor-related characteristics, including cell proliferation, survival, metabolic reprogramming, and tumor angiogenesis (40). Our study found that CREB3L2, a noncanonical UPR factor, can directly act on CD8+ T cells in the TME via paracrine signaling through its cleaved C-terminal fragment. This interaction suppresses the activation of CD8+ T cells through the Hh pathway, leading to tumor immune evasion. Previous research has suggested that the UPR pathway in tumor cells can modulate the antitumor activity of immune cells in the TME by up-regulating PD-L1 (13), stimulating the secretion of inflammatory factors (41), and governing tumor immunogenic cell death (15). In our study, we revealed that UPR factors themselves can function as effector molecules, directly regulating the activity of cytotoxic T cells. Thus, we have uncovered an undocumented mechanism by which tumor cells use intrinsic stress signaling pathways to modulate intercellular communication, thereby affecting tumor immunity and promoting tumor progression. These findings broaden our understanding of tumor immune regulation.

Our work offers an interesting perspective on the functions of the UPR pathway. The classical UPR consists of three branches: PERK, IRE1, and ATF6, which are activated when tumor cells experience ER stress or need to synthesize large amounts of proteins or lipids. In addition to these three branches, a series of other proteins, including the CREB3 family members, is also up-regulated and activated upon ER stress and thus is considered noncanonical UPR factors (21). Structurally, like ATF6, CREB3 family proteins are ER membrane bound and produce N-terminal and C-terminal fragments following RIP. Functionally, the classical role of these noncanonical UPR factors is to act as transcription factors, promoting the transcription of target genes through their N-terminal fragments. For example, the N terminus of CREB3 regulates the Golgi stress response by activating the transcription of Arf4 (42), and the N terminus of CREB3L1 can directly regulate the transcription of genes involved in the protein secretion pathway and collagen production, thereby modulating tumor metastasis (26, 43, 44). By revealing the new functions of C-CREB3L2, our findings highlight the “nonclassical,” yet equally important, roles of these noncanonical UPR factors in cancer. Previous studies have indicated that the C terminus of CREB3L2 can act as a secreted factor regulating the proliferation of various cells, including chondrocytes and glioma cells. We have found that C-CREB3L2 produced by TNBC cells does not affect the proliferation of cancer cells themselves but instead serves as a paracrine factor to modulate the activation and tumor-killing capacity of CD8+ T cells. Our work demonstrates that UPR proteins, traditionally considered intrinsic factors, can also function as extrinsic factors regulating cell-cell communication. This perspective opens avenues for exploring the potential biological functions of other RIP-generated secreted factors.

Previous research has indicated that the C-terminal fragment of CREB3L2 facilitates the ligand-receptor interaction of the Hh pathway, thereby promoting its activation (27). Our work suggests that the regulation of the Hh pathway by CREB3L2 may be cell type specific. In normal cells, such as chondrocytes and neurons, C-CREB3L2 binds with Indian Hh or Sonic Hh, promoting their interaction with Ptch1 and facilitating the release of the Smo protein to activate downstream signaling (27–29). Although a similar phenomenon was observed in glioma cells (45), we did not find that CREB3L2 affects the Hh pathway in TNBC or melanoma cells. This heterogeneity among different tumor types may be related to the distinct expression levels of Hh pathway ligands and receptors in various tumor cells. It is conceivable that, as a costimulatory factor, CREB3L2 may require an appropriate combination of ligand and receptor expression levels to function effectively.

We found that C-CREB3L2 produced by tumor cells can directly activate the Hh pathway in CD8+ T cells within the TME. Although it has been reported that the function of the immune synapse in cytotoxic T cells requires the Hh pathway (46), most studies indicated that the Hh pathway inhibits CD8+ T cell activity through various mechanisms. First, activation of the Hh pathway in CD8+ T cells suppresses T cell receptor signaling, thereby inhibiting T cell activation and proliferation (47–51). In addition, activation of the Hh pathway in CD4+ T cells inhibits CD8+ T cell activity by promoting the expression of IL-4 and other factors (52–54). Furthermore, Hh activation in macrophages can promote M2 polarization, which, in turn, suppresses CD8+ T cell activity (55). Activation of the Hh pathway is associated with an immunosuppressive microenvironment (56, 57). The inverse correlation between the Hh pathway activity and the CD8+ T cell presence aligns with the overexpression of Hh ligands, such as Sonic Hh, in various tumor types (58). While our findings suggest that the Hh pathway in CD8+ T cells can be directly activated by costimulatory factors secreted by tumor cells, it is also possible that the secreted C-terminal fragment of CREB3L2 may affect the Hh pathway in other immune cells, including CD4+ T cells and macrophages, thereby indirectly modulating CD8+ T cell activity. Further investigation into how tumor cells use CREB3L2 to orchestrate cross-talk between different immune cell types within the TME will be of great interest.

One limitation of our study is that we have not yet fully explored the potential interactions between CREB3L2 and other UPR factors in regulating cancer immunity. Although we have demonstrated that CREB3L2 plays a crucial role in modulating intratumoral CD8+ T cell function, other UPR pathways may also influence TME through diverse mechanisms. For example, activation of the IRE1-XBP1 axis in cancer cells has been linked to the establishment of an immunosuppressive TME. XBP1 drives cholesterol synthesis and secretion, which can be incorporated into extracellular vesicles and internalized by myeloid-derived suppressor cells via micropinocytosis, leading to myeloid-derived suppressor cell activation and resistance to immunotherapy (59). In addition, constitutive IRE1 ribonuclease activity has been shown to promote tumor vascularization and the secretion of immunoregulatory cytokines such as IL-6, IL-8, granulocyte-macrophage colony-stimulating factor, and transforming growth factor–β2, particularly in TNBC cells (17, 60). Given our finding that XBP1 may regulate the expression of CREB3L2 in TNBC, it is plausible that CREB3L2 could mediate, at least in part, the pro-tumorigenic and anti-inflammatory effects of the IRE1-XBP1 axis. Furthermore, CREB3L2 induction by XBP1 may work in concert with the IRE1-XBP1 pathway to orchestrate the development of an immunosuppressive TME. Future studies are needed to clarify these potential interactions and their implications for cancer immunotherapy.

Our study may offer a previously unidentified opportunity for cancer immunotherapy. The immune checkpoint mechanism, exemplified by PD-L1/PD-1, is a crucial strategy by which tumor cells evade immune surveillance and clearance. Drugs developed on the basis of this mechanism have achieved remarkable success in cancer treatment. According to our findings, CREB3L2 may be a stress-inducible checkpoint factor that mediates the regulation of tumor immunity through the UPR pathway. As a secreted factor, it is feasible to design neutralizing antibodies to target and eliminate this protein. Considering the unique activation mechanism of CREB3L2, we can focus on targeting key enzymes involved in its activation, such as S1P and S2P, for inhibition. Our previous findings revealed that using S1P or S2P inhibitors (e.g., PF-429242 and AEBSF) can inhibit tumor metastasis by blocking the activation of another UPR factor, CREB3L1 (26). Notably, certain anti-HIV drugs, such as the protease inhibitors Nelfinavir and Ritonavir, are known inhibitors of S1P and S2P. Thus, further exploration of these protease inhibitors, or the screening of more specific S1P/S2P inhibitors, could potentially target the “UPR-checkpoint” proteins like CREB3L2, thereby overcoming tumor immune evasion and inhibiting tumor progression.

MATERIALS AND METHODS

Cell culture

Murine cancer cells pB2, pB3, D2A1, and 4T1 and human HMLE cells (transformed mammary epithelial cells) were from the Weinberg lab at the Whitehead Institute. B16, MDA.MB.231, and 293T cells were obtained from American Type Culture Collection. All cancer cell lines were cultured in Dulbecco’s modified Eagle’s medium (DMEM) or RPMI 1640 medium supplemented with 10% fetal bovine serum (Hyclone) and penicillin/streptomycin (100 U/ml) at 37°C in a humidified atmosphere containing 5% CO₂. HMLE cells were maintained in a mixture (1:1) of DMEM and MEGM (Lonza). Primary immune cells were maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum, 1× GlutaMAX (Gibco, 35050061), penicillin/streptomycin (100 U/ml), and IL-2.

Reagents

Two antibodies were used to detect CREB3L2: (i) rabbit α-CREB3L2 (Merck, MABE1018), which recognizes the N-terminal fragment of CREB3L2, and (ii) rabbit α-CREB3L2 (Abcepta, AP9654b), which recognizes the C-terminal fragment of CREB3L2. Rabbit anti-FLAG (ab124783), mouse anti-GAPDH (Proteintech, 60004-1-Ig), rabbit anti-Ptch1 (Proteintech, 17520-1-AP), rabbit anti-Smo (Santa Cruz, sc-166685), rabbit anti-Gli1 (Cell Signaling Technology, 2534), Ultra-LEAF Purified anti-mouse CD3ε Antibody (Sigma-Aldrich, F1804), and InVivoMAb anti-mouse CD28 (Biocell, BE0015-5) were used in Western blot and T cell activation assays. InVivoMAb anti-mouse CD8 (Biocell, BE0061) was used in CD8+ T cell depletion. InVivoMAb anti-mouse PD-1 (Biocell, BE0146) was used in the α-PD-1 treatment experiment. The propidium iodide (PI) solution was from BioLegend (79997). 5-Bromo-2′-deoxyuridine solution was from Invitrogen (8811-6600). Cell Counting Kit-8 was from Yeasen (40203ES80). Recombinant murine IL-2 was from PEPROTECH (212-12). APC-AnnexinV Apoptosis Detection Kit with PI was from BioLegend (640932). The crystal violet solution was from Meilunbio (548-62-9). Gant61 was from MedChemExpress (HY-13901). SAG was from MedChemExpress (HY-12848B). The dual-luciferase reporter system was from Promega (E1910).

Clinical samples

Human TNBC tissues were provided by the First Central Hospital of Baoding. All patients provided written informed consent for sample collection and analysis. The use of human breast cancer tissue was approved by the Clinical Research Committee of the First Central Hospital of Baoding (no. 2024-11). The clinical information about these samples is shown in fig. S1D.

Plasmid construction and transfection

shRNAs targeting the mouse Creb3l2 and human XBP1, ATF4, and ATF6 genes were cloned into the pLKO.1 vector. The full-length [1 to 1566 base pairs (bp)], N-terminal (1 to 1131 bp), and C-terminal fragments (1291 to 1566 bp) of CREB3L2 were cloned into the pLVX-IRES-puro vector according to the design shown in Fig. 2F. A binding immunoglobulin protein signal peptide was added into the 5′ end of the C-terminal fragment to ensure its proper secretion. Lentivirus-packaging plasmids and core shRNA or overexpression plasmids were transfected into 293T cells. Virus-containing cell culture medium was harvested and filtered with a 0.45-μm filter (Millipore, SLHA033SS) and then used to infect cells. Stable cells were selected by selective drugs and used for further experiments. Gli-reporter was from Addgene (113712). All primers used for cloning are summarized in table S1.

RNA extraction and RT-qPCR

RNA preparation and the following reverse transcription quantitative polymerase chain reaction (RT-qPCR) were conducted as previously reported (61). RNA extraction was performed with the TRIzol Reagent (Ambion) or an RNeasy Kit (Qiagen) according to the manufacturer’s protocols. Complementary DNA was synthesized from purified RNA using the iScript cDNA Synthesis Kit (Bio-Rad, 1708891) according to the manufacturer’s instructions. The RT-qPCR system used 2× Hieff qPCR SYBR Green Master Mix (Yeasen, 11202ES03) and was used with the Bio-RadCFX96 Touch Real-Time PCR Detection System (Bio-Rad). The comparative C_t method was used for the data analysis, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) or actin was used as an internal control for normalization. Primer sequences are shown in table S1.

Co-IP experiment

Primary CD8+ or EL4 cells were lysed with coimmunoprecipitation (co-IP) lysis buffer [20 mM tris-HCl (pH 7.5), 100 mM KCl/NP-40, 0.1% EDTA, and 10% glycerol (1 mM)] with protease cocktail (Roche, 04693159001) and PhosSTOP (Roche, 04906837001). Protein concentration was measured by a bicinchoninic acid protein assay (Thermo Fisher Scientific, 23225). Protein A/G magnetic beads (Thermo Fisher Scientific, 88803) were washed and incubated with the designated primary antibodies at 4°C for 1 hour. An equal amount of protein from each sample was added and incubated at 4°C overnight with a rotator. After washing three times with co-IP lysis buffer, the beads were eluted with 1× SDS loading buffer before the eluted samples were analyzed by the Western blotting assay (61).

Western blotting

Cell lysates with equal amounts of protein were separated by 4 to 20% or 10% SDS–polyacrylamide gel electrophoresis gel (Bio-Rad), transferred to an Immobilon transfer membrane (GE Healthcare Life Sciences), and immunoblotted with specific antibodies as previously described (61). The expression of FL-CREB3L2 and N-CREB3L2 in the cell lysate was immunoblotted by an α-CREB3L2 antibody (Merck, MABE1018), which recognizes the N-terminal peptide sequence of CREB3L2. The expression of C-CREB3L2 was immunoblotted by an α-CREB3L2 antibody (Abcepta, AP9654b), which recognizes C-CREB3L2, or by an α-FLAG antibody. All immunoblots were visualized by enhanced chemiluminescence (Bio-Rad).

Expression and purification of exogenous C-CREB3L2 protein

293T cells were transfected with pLVX-IRES-FLAG-C-CREB3L2 and lysed with radioimmunoprecipitation assay buffer with protease cocktail and PhosSTOP 48 hours after transfection. Anti-DYKDDDDK (FLAG) was used to immunoprecipitate FLAG-C-CREB3L2 according to the manufacturer’s instructions. Then, these proteins were eluted from the affinity gels with 3× FLAG peptide. To remove the FLAG peptides, the eluted samples were added into the Amicon Ultra-4 Centrifugal Filter Unit (Millipore, UFC801008, 10 kDa cutoff) and centrifuged at 4000 rpm for 30 min. Then, 2 ml of phosphate-buffered saline (PBS) was added into the unit and centrifuged at 4000 rpm for 30 min. The step of PBS wash was repeated another two times, and ~50-μl concentrated proteins were obtained from the unit.

Xenograft assay

Indicated tumor cells were trypsinized, washed, resuspended in a complete medium, and subcutaneously injected into the flanks of mice (C57BL/6, BALB/c, or nude mice). To minimize individual variations, five to eight age- and sex-matched mice in each group were used. Tumor size and mouse body weight were measured every 3 or 4 days. Tumor volumes were measured by length (a) and width (b) using calipers and calculated using the equation V (mm³) = (a × b²)/2, where a is the largest diameter and b is the smallest diameter. The tumors were dissected and weighed at the endpoint of each experiment. Experimental animals were cared for in accordance with guidelines approved by the First Affiliated Hospital, Zhejiang University School of Medicine (no. 2023422).

Flow cytometry analysis for mouse tumors

Tumors collected from the xenograft assay were dissociated by a collagenase assay and stained with PI, Pacific Blue-α-CD8, and FITC-α-CD4 antibodies. Flow cytometry analysis was conducted using a Beckman flow cytometer. Dissociated tumor cells were initially gated to exclude debris, followed by gating for live cells by selecting the PI-negative population. Then, the singlet cell subfraction of the lymphocyte population was analyzed for CD8 and CD4 positivity.

IHC analysis

The breast cancer tissue was deparaffinized and hydrated, followed by antigen retrieval with tris/EDTA buffer (pH 9.0). The primary antibodies against CREB3L2, CD8, and FOXP3 were incubated with the slides in a humidified chamber at 4°C overnight. The IHC reaction was detected with Bond Polymer Refine Detection (Leica Biosystems, DS98000) and 3,3′-diaminobenzidine as a chromogen for visualization of the staining. Hematoxylin (Leica Biosystems) was used to counterstain. A similar process was applied for tumor tissues collected from xenografts. In the patient-derived tissues, the number of CD8+ T cells per image field was determined and counted by a pathologist. The “low” cohort refers to samples that contained 10 or fewer CD8+ cells per field, the “medium” cohort refers to samples that contained 10 to 50 CD8+ cells per field, and the “high” cohort refers to samples that contained 50 or more CD8+ cells per field.

Isolation and activation of primary T cells

Lymphocytes in spleens of the BALB/c or C57BL/6 mice were prepared from aseptically collected mouse spleens by crushing and passage through 70-μm cell strainers (Beyotime, FSTR070). Red blood cells were removed from the sample using ammonium chloride solution (Stemcell, 07800). Mouse CD8+ T cells were isolated from the primary lymphocyte population using the Dynabeads Untouched Mouse CD8 Cells kit (Invitrogen, 11417D) according to the manufacturer’s protocol. The lymphocytes were stimulated with an α-CD3 antibody and α-CD28 antibody in RPMI 1640 containing murine IL-2 (10 ng/ml), 10% fetal bovine serum, 1% penicillin/streptomycin solution, and 1× GlutaMAX for 48 hours.

Ex vivo T cell and tumor cell coculture assay

Tumor cells (2 × 10⁴) were seeded into six-well plates with DMEM complete medium. After 24 hours, the culture medium was changed to a complete RPMI 1640 medium containing mouse IL-2 (10 ng/ml). Naïve or activated CD8+ T cells were then added and cocultured with tumor cells at the ex vivo T cell–to–tumor cell ratio ranging between 1:1 and 10:1. After 48 hours, T cells in culture supernatants were harvested and stained with PI, an α-CD8 antibody, and an α-CD25 antibody for flow cytometry analysis using a Beckman flow cytometer. The remaining tumor cells were washed with 1× PBS and then stained with crystal violet for 5 min (1 ml per well) and washed with PBS before image analysis and quantification.

To analyze the apoptosis of tumor cells after coculture with CD8+ T cells, tumor cells labeled with BFP were seeded into six-well plates (2 × 10⁴ per well) and cocultured with naïve or activated CD8+ T cells at a set ex vivo T cell–to–tumor cell ratio (from 1:1 to 15:1) for 48 hours. Tumor cells were then collected by trypsinization and stained with the APC Annexin-V Apoptosis Detection Kit with PI according to the manufacturer’s instructions. Apoptotic blue fluorescence–positive tumor cells were analyzed by a Beckman flow cytometer. To measure proliferation of CD8+ T cells after coculture, 2 μM 5-bromo-2′-deoxyuridine solution (Invitrogen) was added into the coculture of cancer cells and T cells 24 hours before T cells were collected for flow cytometry analysis.

Statistical analysis

The measurement data in this study were statistically analyzed by Excel or GraphPad Prism 10.0 software and presented as means ± SEM, unless otherwise indicated. Image information was extracted using ImageJ software. Student’s t test was used for comparative analysis between two data groups. The analysis of variance (ANOVA) was used for comparative analysis between multiple data groups. P values <0.05 with a 95% confidence interval were considered significant. The number of biological replicates (n), the statistical tests performed, and the statistical significance (P) were specified in the legend of each figure, unless specified otherwise.

Supplementary Materials

This PDF file includes:

Fig. S1 to S8

Table S1