Induction of the activating transcription factor-4 in the intratumoral CD8+ T cells sustains their viability and anti-tumor activities

Abstract

Immune suppressive factors of the tumor microenvironment (TME) undermine viability and exhaust the activities of the intratumoral cytotoxic CD8+ T lymphocytes (CTL) thereby evading anti-tumor immunity and decreasing the benefits of immune therapies. To counteract this suppression and improve the efficacy of therapeutic regiments, it is important to identify and understand the critical regulators within CD8+ T cells that respond to TME stress and tumor-derived factors. Here we investigated the regulation and importance of activating transcription factor-4 (ATF4) in CTL using a novel Atf4^ΔCD8 mouse model lacking ATF4 specifically in CD8+ cells. Induction of ATF4 in CD8+ T cells occurred in response to antigenic stimulation and was further increased by exposure to tumor-derived factors and TME conditions. Under these conditions, ATF4 played a critical role in the maintenance of survival and activities of CD8+ T cells. Conversely, selective ablation of ATF4 in CD8+ T cells in mice rendered these Atf4^ΔCD8 hosts prone to accelerated growth of implanted tumors. Intratumoral ATF4-deficient CD8+ T cells were under-represented compared to wild type counterparts and exhibited impaired activation and increased apoptosis. These findings identify ATF4 as an important regulator of viability and activity of CD8+ T cells in the TME and argue for caution in using agents that could undermine these functions of ATF4 for anti-cancer therapies.

INTRODUCTION

An adaptive immune response to a growing tumor relies on cell-mediated immunity that involves presenting the tumor-specific antigens to CD8+ T cells, which then become activated and differentiate into the cytotoxic T lymphocytes (CTL) within lymphoid tissue. These CTL migrate to the source of antigen (i.e., tumor tissue) and attack malignant cells (reviewed in (1, 2)). Within solid tumors, CTL kill their targets upon recognition of tumor-specific antigens expressed by the malignant cells (3), but also become prone to death and exhaustion that is driven by repeated antigenic stimulation and activation of inhibitory immune checkpoints (4–6). Additional cellular and acellular tumor microenvironment (TME) factors including regulatory T cells, myeloid-derived suppressor cells, immunosuppressive cytokines, as well as metabolic constraints (accumulation of adenosine, lactic acid, etc) further restrict the tumoricidal function and ultimately the viability of CTL, thereby undermining the anti-tumor immune response and decreasing the efficacy of immune therapies (reviewed in (7)).

Besides reacting to antigenic stimulation and the immune suppressive milieu, CTL - together with other intratumoral cellular types - are also exposed to a harsh intratumoral TME characterized by a deficit of oxygen and nutrients and activation of the Integrated Stress Response (8–10). This pathway involves activation of diverse protein kinases including PKR-like ER kinase (PERK) and general control non-derepressible 2 (GCN2). A converging end result of this activation is specific phosphorylation of the eIF2α eukaryotic translation initiation factor, global translational shutdown and efficient non-canonical translation of activating transcription factor-4 (ATF4, (8–10)). In addition to translational control, ATF4 expression is also subject to transcriptional regulation (11, 12).

ATF4 is a basic leucine-zipper transcription factor that governs the expression of genes pivotal for proper cellular responses to stress and changes in the redox state and amino acid metabolism (reviewed in (13, 14)). In addition, ATF4 controls the expression of several critical microRNAs that control survival and function of stressed cells (15–18). In all, ATF4 acts as a central regulator of the Integrated Stress Response (19) and plays a key role in tumor growth and progression (20). A critical malignant cell-intrinsic role of ATF4 that supports growth and progression of primary tumors and stimulates the metastatic process has been well established (21–23). However, the importance of ATF4-driven regulation in benign intratumoral compartments is not well understood.

Our recent studies showed that ablation of ATF4 in the cells of the TME in general (and particularly in the fibroblasts) robustly suppressed primary tumor growth as well as metastasis (24). ATF4 was also shown to control the metabolic reprogramming in activated CD4+ T cells (25). However, the functions of ATF4 in CTL in general and specifically in intratumoral CTL remain to be understood.

Here we determined the importance of ATF4 in the function of CTL in the context of growing tumors. The data presented here demonstrate that ATF4 was induced in CD8+ T cells by antigenic activation, tumor-derived factors and stress stimuli. Ablation of ATF4 in CTL notably undermined survival and activity of the intratumoral CTL. Accordingly, expression of ATF4 in CTL played a key role in restricting growth of solid tumors. In all, these data suggest that activation of ATF4 in CTL is important for anti-tumor immunity.

MATERIALS AND METHODS

Animal studies.

All animal experiments were approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania and were carried out in accordance with the IACUC guidelines. All mice were in the C57BL/6 background. WT, OT-I (Jackson Laboratory, Strain #:003831) and Cd8a-Cre mice (Jackson Laboratory, Strain #:008766) were purchased from Jackson lab. All mice used in the experiments had one Cre allele or no Cre alleles. Atf4^f/f mice were previously described (23, 24). OT-I-Atf4^ΔCD8 mice were generated by intercrossing OT-I mice with Cd8a and Atf4^f/f mice. All animal experiments were performed using both male and female littermates of 6–8 weeks of age. All animals were kept in specific pathogen-free facility in accordance with American Association for Laboratory Animal Science guidelines. All mice had water ad libitum and were fed with regular food. Littermates from different cages were randomly assigned into different group and were exposed to the same environment.

Cell culture and tumor-conditioned medium (TCM) preparation.

WI38 normal human fibroblasts, mouse MC38 colon adenocarcinoma cells and human HCT116 colorectal cancer cells were purchased from ATCC. MC38OVA colon adenocarcinoma cells (a generous gift from Dr. Susan Ostrand-Rosenberg, University of Maryland) were engineered to express Renilla luciferase. MH6499c4 pancreatic ductal adenocarcinoma were kindly provided by Dr. Ben Stanger (University of Pennsylvania). These cell lines were cultured at 37°C with 5% CO₂ in DMEM (Gibco, cat#11965–084) including 10% heat-inactivated Fetal Bovine Serum (FBS, Hyclone, cat#SH30071.03), 100 U/ml penicillin-streptomycin (ThermoFisher, cat#15140122) and L-glutamine (Gibco, cat#25030081). Tumor-condition medium (TCM), WI38 fibroblast-conditioned media (FCM) and Serum-free media (SFM) was prepared as previously reported (26).

T cells isolation and quantitative real-time PCR.

Human CD8⁺ T cells were obtained from the Human Immunology Core at the University of Pennsylvania. Murine CD4⁺ and CD8⁺ T cells were isolated from spleen or tumor tissue using T cells isolation kits (Stem cell, cat#19853 and 19765A). Human and murine CD8⁺ T cells were stimulated with anti-mouse CD3 (BioLegend, cat#100340) and anti-CD28 antibodies (BioLegend, cat#102116) for 24 hr in the presence or absence of Thapsigargin (100nM, cat#67526–96-8) or treated with tumor-conditioned medium (TCM) or fibroblast-conditioned medium (FCM) for 48 hr. After treatment, total of RNA was extracted from CD8⁺ T cells using Mini RNA extraction kit (Qiagen, cat#74004). Concentration of RNA was measured by nanodrop2000 and the mRNA expression of Atf4 and Ddit3 were tested using real-time PCR and results were normalized per β-actin mRNA levels. Primers are as follows: mouse Atf4 Forward: CCTGAACAGCGAAGTGTTGG and Reverse: TGGAGAACCCATGAGGTTTCAA; mouse Ddit3 Forward: GGAGCTGGAAGCCTGGTATG and Reverse: GGATGTGCGTGTGACCTCTG; mouse Actb Forward AGAGGGAAATCGTGCGTGAC and Reverse: CAATAGTGATGACCTGGCCGT; Human ATF4 Forward: AAACCTCATGGGTTCTCCAG and Reverse: GGCATGGTTTCCAGGTCATC; Human DDIT3 Forward: AGCACCAAAGCAGCCAT and Reverse: ACTCAGCTGCCATCTCTG; Human ACTB Forward: AGCACAGAGCCTCGCCTT and Reverse: CATCATCCATGGTGAGCTGG.

Flow cytometry analysis.

For immune profiling of spleen and tumor, tissues were processed as previously described (27). Briefly, spleens were ground up and passed through 40μm cell strainer. Red cells were depleted using RBC buffer followed by PBS washes before antibody staining. Tumor tissues were dissected and digested with 1 mg/ml Collagenase D (Roche, cat#11088882001) with 100 μg/ml DNase I (Roche, cat#10104159001) in RPMI-1640 medium with 2% FBS for 1 hr with continuous agitation at 37°C. Digestion mixture was passed through 100 μm cell strainer to prepare single cell suspension and washed with PBS supplemented with 2mM EDTA and 1% FBS.

The following antibodies were used to stain the resulting single cell suspensions: anti-mouse CD8-APC antibody (BioLegend, cat#126614) and anti-human CD8-PE/CY7 antibody (BioLegend, cat#344750), anti-mouse ATF4 primary antibody (Cell signaling Technology, cat#11815s, 1:150), anti-human ATF4 primary antibody (Cell signaling technology, cat#11815s, 1:150), goat anti-rabbit secondary antibody (Invitrogen, cat#A32732, 1:500), anti-CD45-APC/Cy7 (BioLegend, cat#157204), anti-CD3-PE (BioLegend, cat#100206), anti-CD8-AF700 (BioLegend, cat#100730), anti- CD69-BV421 (BioLgend, cat#104545, anti-PD-1-PE/Cy7 (BioLegend, cat#109110) and Annexin V-APC (BioLegend, cat#640941), anti-LAG3-PE (BioLegend, cat#125207), anti-TIGIT-Alexa Fluor 700 (R&D system, cat#FAB72671N), anti-TIM3-BV421 (BioLegend, cat#134019).

For cell surface staining, cells were incubated with indicated antibodies on the ice for 30mins. Following staining, cells were washed with PBS for twice and subjected to flow cytometry analysis. For analysis of intracellular ATF4 levels, T cells were fixed with Foxp3 IC Fixation buffer (Invitrogen, cat#00–5523-00) for 30 min, washed and stained with anti-ATF4 antibodies for 30 min followed by a double PBS wash and flow cytometry analysis.

For intracellular cytokines analysis, Atf4^f/f and Atf4^ΔCD8 T cells were isolated from spleen of indicated mice and stimulated with anti-CD3 (10 μg/ml) and anti-CD28 (5 μg/ml) Abs for 48 hr. After being stimulated, T cells were fixed with Foxp3 IC Fixation buffer (Invitrogen, cat#00–5523-00) for 30 min, washed and stained with anti-IL-2-Alexa fluor 700 (BioLegend, cat#503818), anti-TNF-α-PE (BioLegend, cat#506104), anti-IFN-γ-APC/CY7(BioLegend, cat#505849) and anti-Ki67-PE/CY7 (BioLegend, cat#151217) antibodies followed by flow cytometry analysis

Immunoblotting analysis.

Overall, the immunoblotting was carried out as previously described (28). For analyzing the ATF4 expression before and after OVA peptide stimulate, T cells were isolated from spleens of WT OT-I mice and stimulated with Vehicle or OVA peptide (1 μg/ml) for 48 hr. After stimulation, protein was extracted from these cells and ATF4 protein level was analyzed by immunoblotting using Rabbit anti-mouse ATF4 antibody (Cell Signaling Technology, cat#11815, 1:1000) and goat anti-rabbit secondary antibody (Abcam, cat#ab6721, 1:5000).

Cytotoxicity assay.

The ability of OT-I CTL to kill target MC38OVA cells expressing luciferase was evaluated in a luciferase-based cytotoxicity assay as described in (29). Target cells were cocultured with CTL for 4 hr at a 10:1 E:T ratio in 96-well black plate at a total volume of 200 μL of RPMI-1640 complete media. Target cells alone were seeded in parallel at the same density to quantify the spontaneous death luciferase expression (relative luminescent units; spontaneous death RLU). Target cells lysed with water considered as the maximal killing (maximal killing RLU). Following co-culture, 100 μl of luciferase substrate (Bright-Glo; Promega) was added to the remaining supernatant and cells. Luminescence was measured after a 10 min incubation using the EnVision (PerkinElmer) plate reader. The percent cell lysis was obtained using the following calculation: % lysis=100x (spontaneous death RLU- test RLU)/(spontaneous death RLU- maximal killing RLU).

Tumorigenesis studies.

For syngeneic subcutaneous tumor model, MC38 (1×10⁶ in 100μl PBS) and MH6499c4 (1×10⁶ in 100μl PBS) were subcutaneously inoculated into right flank of mice and tumor size was measured every other day using caliper. Tumor volume was calculated as width x width x length x 0.5 and tumor tissue were harvested until tumor volume reached ~1500 mm³.

Atf4^ΔCD8 mice genetic characterization.

Peripheral blood was collected from Atf4^f/f and Atf4^ΔCD8 mice through retro-orbital bleeding and lymphocytes were isolated using Lymphoprep buffer (Stem Cell, cat#07851). Splenic CD4⁺ and CD8⁺ cells were obtained from Atf4^f/f and Atf4^ΔCD8 mice using T cell isolation kits as specified above. Atf4 mRNA levels were assessed by q-PCR analysis. In addition, blood and spleen derived single cells were washed with PBS for twice and used for immune profiling using the following antibodies: anti-CD3-PE/CY7 (BioLegend, 100220), anti-CD8-AF700 (BioLegnd, cat#100714), anti-CD4-APC/CY7 (BioLegend, cat#100414), anti-CD19-PE (BioLegend, cat#115507), anti-NK1.1-APC (BioLegend, cat#108709).

Human database analysis.

We used data described in Ref (30) and acquired the deposited raw counts file from GEO (RNA-Seq) and used DESeq2 to normalize the counts of mapped reads. Samples were classified as Naïve, TEM and CD103⁺CD39⁺ three subtypes. ATF4 level were analyzed by using R software to generate the dot-plot with statistical analysis.

Quantification and statistical analysis.

All experiment described here are the representative of at least three independent experiments (n≥5 mice for each group unless specifically indicated). For in vitro experiments, cells or tissues from each of these animals were processed (at least) in biological triplicates. All data here were shown as average ± S.E.M. Statistical analysis between two groups was conducted with 2-tailed Student t test and multiple comparisons were performed by using One-way ANOVA or two-way ANOVA analysis with Tukey’s multiple-comparison. Tumor growth curve analysis was conducted with Repeated-measure two-way ANOVA (mixed-model) with Tukey’s multiple-comparison. P values < 0.05 were considered significant.

RESULTS

ATF4 is induced in CTLs exposed to tumor-derived factors and antigen stimulation.

We sought to determine the expression of ATF4 in CD8⁺ T cells upon their activation or exposure to the stress stimuli and tumor-derived factors. Treatment of mouse splenic CD8+ T cells with thapsigargin, an inducer of the endoplasmic reticulum stress (31), significantly increased the expression of ATF4 mRNA (Figure 1a). Importantly, this expression was further induced in cells activated by CD3/CD28 agonist antibodies (Figure 1a). These results are consistent with previously reported observations (32) and indicate that stress or T cell receptor activation regulate ATF4 levels. The effects of T cell receptor activation were further tested in the OT-I CD8+ T cells, which recognize target cells expressing OVA antigen. Treatment of these cells with specific antigen (OVA peptide) notably increased the levels of ATF4 protein (Figure 1b–c). This increase was also seen in OT-I splenic T cells (Figure 1d). This result further demonstrates that CD8+ T cell activation upregulates ATF4.

Fig.1. Induction of ATF4 in activated CD8⁺ T cells.

A. qPCR analysis of mRNA levels of Atf4 in mouse CD8⁺ T cells treated or not with a stress inducer thapsigargin (TG, 100nM, 24 hr) in the presence or absence of CD3/CD28 agonists.

B. Representative flow cytometry analysis of ATF4 protein levels in OT-I CD8+ T cells stimulated with OVA peptide (2 μg/ml for 48 hr).

C. Quantification of mean fluorescence intensity (MFI) of ATF4 analysis from panel B (n=6 for each group).

D.Immunoblotting analysis of ATF4 protein level in splenic OT-I T cells stimulated with OVA peptide (1 μg/ml) for 48 hr.

Data are represented as mean±S.E.M; Statistical significance was performed using students’ t-test (A, C).

Inside tumors, CD8+ T cells encounter antigens and become activated. In addition, these CD8+ T cells also are exposed to the tumor-derived factors (such as cytokines, adenosine, extracellular vesicles, etc.) and stress factors of the TME (hypoxia, deficit of nutrients, etc.). Thus, we further determined how ATF4 levels responded to factors that might be released into the TME by tumor cells.

Mouse MC38 colon adenocarcinoma cell-conditioned media (TCM) significantly upregulated ATF4 mRNA (Figure 2a) and protein (Figure 2b) levels in mouse CD8+ T cells. Similarly, levels of ATF4 and its target DDIT3 gene, which encodes pro-apoptotic CHOP protein (31), were increased in human CD8+ T cells upon treatment with TCM HCT116 human colon adenocarcinoma (Figure 2c–d–e). Importantly, this increase was not elicited by media conditioned by fibroblasts (FCM) indicating that tumor-derived factors can upregulate ATF4. Similar to mouse cells (Figure 1), human CD8+ T cells also upregulated ATF4 and DDIT3 in response to TCR activators or thapsigargin (Figure 2c–d–e) further indicating that T cell activation, stress and tumor-derived factors each upregulate ATF4.

Fig.2. Induction of ATF4 in the intratumoral CD8⁺ T cells.

A. mRNA levels of Atf4 in mouse CD8⁺ T cells treated or not with the media conditioned by MC38 tumor cells (TCM) for 48 hr (n=5).

B. Representative flow cytometry analysis of ATF4 protein levels in mouse CD8+ T cells treated as in panel A. MFI quantification (n=4) is shown on the right.

C. qPCR analysis of mRNA levels of ATF4 in human CD8+ T cells treated or not with the media conditioned by HCT116 tumor cells (TCM) or by WI38 normal human fibroblasts (FCM), or serum-free media (SFM), or with TG (100nM), or with CD3/CD28 agonists for 48 hr.

D. qPCR analysis of mRNA levels of DDIT3 in human CD8+ T cells treated or not with the media conditioned by HCT116 tumor cells (TCM) or by WI38 normal human fibroblasts (FCM), or with TG (100nM), or with CD3/CD28 agonists for 48 hr.

E. Representative flow cytometry analysis of ATF4 protein levels from experiments described in panel C assessed by flow cytometry. MFI quantification (n=4) is shown on the right.

F. qPCR analysis of Atf4 mRNA levels in mouse CD8+ T cells isolated from spleen from naïve (Na-Sp) or MC38 tumor-bearing (TB-Sp) mice or from MC38 tumor tissues (n=5 in each group).

G. Representative flow cytometry analysis of ATF4 protein levels in mouse CD8⁺ T cells from experiment described in panel E. MFI quantification (n=5) is shown on the right.

Data are represented as mean±S.E.M; Statistical significance was performed using students’ t-test (A and B) or one-way ANOVA with Tukey’s multiple-comparison test (C, D, E, F and G).

These conclusions were corroborated in vivo by analyses of mouse CD8+ T cells isolated from spleen or MC38 tumor tissues. Levels of ATF4 mRNA (Figure 2f) and protein (Figure 2g) in the intratumoral CD8+ T cells were notably greater than cells from spleens of naïve or tumor-bearing mice. Collectively, these data suggest that activation, TME stress and tumor-derived factors may act in concert to upregulate ATF4 in intratumoral CD8+ T cells.

Ablation of ATF4 in CTL undermines their viability and activity

Previous studies involving ablation of ATF4 either in malignant cells (23) or in all non-malignant cells within the TME (24) strongly indicated that ATF4 acts to promote tumor growth. Thus, we posited that increased expression of ATF4 in the intratumoral CD8+ T cells should also have a pro-tumorigenic function. However, mining the gene expression profiling data from the CD8+ T cells from patients with colorectal cancers comparing naïve and effector memory subsets with a highly activated CD103+CD39+ tumor-reactive subset, which expressed elevated levels of IFNγ, granzymes and perforin (30), we noted that ATF4 levels were greater in the latter subset (Figure 3a). These findings are inconsistent with a hypothetical pro-tumorigenic role of ATF4 in the CTL.

Fig. 3. Expression of ATF4 in human CD8+ T cell subsets and generation of the Atf4^ΔCD8 mice.

A. Relative ATF4 expression in three CD8+ T cells subtypes isolated from colorectal cancer samples including naïve T cells, effector memory T cells (T_em) and CD103⁺CD39⁺ highly active T cells using data from Ref (30). (n=7 for each group).

B. qPCR analysis of Atf4 mRNA levels in mouse CD4⁺ T cells isolated from the spleens of Atf4^+/+ (Atf4^+/+; Cd8a-Cre) or Atf4^ΔCD8 (Atf4^f/f; Cd8a-Cre) mice and treated as indicated.

C. qPCR analysis of Atf4 mRNA levels in mouse CD8⁺ T cells isolated from the spleens of Atf4^+/+ (Atf4^+/+; Cd8a-Cre) or Atf4^ΔCD8 (Atf4^f/f; Cd8a-Cre) mice and treated as indicated.

D. Flow cytometry analysis of ATF4 in Atf4^+/+ or Atf4^ΔCD8cells treated with vehicle or CD3/CD28 antibodies.

E. qPCR analysis of Ddit3 mRNA levels in mouse CD8⁺ T cells isolated from the indicated mice and treated as indicated.

Data are represented as mean±S.E.M; Statistical significance was performed using one-way ANOVA with Tukey’s multiple-comparison test (A, B, C, D and E).

This dichotomy prompted us to examine the functional role of ATF4 in the intratumoral CTL using genetically engineered mouse model that lacked ATF4 specifically in CD8+ cells (Figures S1a-b). These mice developed normally and did not exhibit any overt phenotypes (such as alterations in fertility, body weight, age associated mortality, etc). Comparative analysis of spleen size and subsets of immune cells in blood and spleen between control mice (Atf4^f/f) and Atf4^ΔCD8 mice also did not revealed any significant differences (Figure S2a-b). No notable variations in the frequencies of splenic CD8+ T cells that could be associated with ablation of ATF4 were uncovered in comparison to Cd8a-Cre mice (Figure S2c-d). Furthermore, activated ATF4-null CTLs did not dramatically differ from ATF4-competent cells in levels of IFN-γ, TNF-α, interleukin-2 or proliferation marker Ki-67 (Figure S2e).

Analysis of ATF4 mRNA levels revealed no changes in splenic CD4 T cells (Figure 3b) but nearly complete ablation of ATF4 mRNA (Figure 3Cc) and protein (Figure 3d) in CD8 T cells from Atf4^ΔCD8 mice. Additional functional validation of these mice was performed by analyzing the expression of ATF4 target gene Ddit3. Treatment of CD8+ T cells isolated from Atf4^ΔCD8 mice with thapsigargin did not induce Ddit3 (Figure 3e) indicating that ATF4 was properly inactivated in these cells.

We next tested the functional roles of ATF4 in CTL activities in vitro. To this end, we used OT-I mice in crossings that generated OT-I/Atf4^+/+ or OT-I/Atf4^ΔCD8 mice. Intriguingly, ATF4-deficient OT-I CD8+ T cells exhibited a notably lower expression of the CD69 activation marker upon antigenic stimulation in vitro (Figure 4a) suggesting that the status of ATF4 may play an important role in the regulation of activity of these cells. This possibility was further indirectly corroborated by analysis of PD-1 expression, which was increased by antigen stimulation to a significantly greater extent in the Atf4^ΔCD8 CD8+ T cells (Figure 4b). These data suggest that ATF4 function regulates activation of CD8+ T cells.

Fig 4. ATF4 regulates activities and viability of CD8⁺ T cells.

OT-I mice were used in crossings to generate OT-I; Atf4^+/+ or OT-I; Atf4^ΔCD8 mice.

A. Frequency of CD69⁺ cells among CD8⁺ T splenic cells of indicated genotype treated or not with OVA peptide (2 μg/ml for 48 hr).

B. Frequency of PD-1⁺ cells among CD8⁺ T splenic cells of indicated genotype treated or not with OVA peptide (2 μg/ml for 48 hr).

C.Frequency of apoptotic (Annexin V⁺) cells among CD8⁺ T splenic cells of indicated genotype treated or not with OVA peptide (2 μg/ml for 48 hr).

D. Percent of lysis of MC38OVA-luc malignant cells co-incubated with OVA-activated CD8⁺ T cells isolated from spleens of OT-I; Atf4^+/+ or OT-I; Atf4^ΔCD8 mice.

Data are represented as mean±S.E.M; Statistical significance was performed using students’ t-test (D) or one-way ANOVA with Tukey’s multiple-comparison test (A, B and C).

Given the paramount role of ATF4 in survival of stressed cells (14, 16, 31), we next analyzed the extent of apoptosis under these conditions. Naïve Atf4^+/+ and Atf4^ΔCD8 CD8+ T populations did not differ in the number of annexin V-positive cells (Figure 4c). However, upon antigenic stimulation, a greater number of apoptotic cells was found in the Atf4^ΔCD8 CD8+ population (Figure 4c) indicating that ATF4 may act to support viability of CD8+ T cells that have encountered the antigen.

To further corroborate the role of ATF4 in viability and function of CD8+ T cells, we compared the tumoricidal activities of Atf4^ΔCD8 and Atf4^+/+ CD8+ OT-I T cells upon their incubation with MC38 malignant cells expressing OVA and also luciferase to enable precise quantification of their lysis. These studies revealed that ablation of ATF4 decreased the cytolytic effect of OT-I cells on MC38-OVA-luciferase target cells (Figure 4d). Given that, at least to some extent, this phenotype could be attributed to impaired viability of the ATF4-deficient CTLs, these in vitro results collectively suggest that ATF4 plays an important role in the maintenance of viability and effector activity of CD8+ T cells.

Ablation of ATF4 in CTL promotes tumor growth

To test the effects of ATF4-deletion in CD8 T cells on antitumor responses in vivo, we assessed growth of MC38 colon adenocarcinoma tumors in Atf4^ΔCD8 mice. Notably, MC38 tumor growth was significantly accelerated in Atf4^ΔCD8 mice in comparison to Atf4^f/f (Figure 5a–c) or to Atf4^+/+; Cd8a-Cre control mice (Figure S3a-c). Similarly, ablation of ATF4 in CD8+ T cells supported greater growth of tumors formed by MH6499c4 pancreatic ductal adenocarcinoma cells (Figure 5d-f). These results suggest that ATF4 in CD8+ T cells contributes to their anti-tumor functions.

Fig 5. ATF4 in CD8⁺ T cells inhibits tumor growth in syngeneic mice.

A. Volumes of s.c. inoculated (10⁶/mouse) MC38 colon adenocarcinoma tumors grown in Atf4^f/f or Atf4^ΔCD8 mice (n=7)

B. Pictures of MC38 tumors harvested from indicated mice on Day 19

C. Mass of MC38 tumors from panel B.

D. Volumes of s.c. inoculated (10⁶/mouse) MH6499c4 tumors grown in Atf4^+/+ (Atf4^+/+; Cd8a-Cre) or Atf4^ΔCD8 (Atf4^f/f; Cd8a-Cre) mice (n=5)

E. Pictures of MH6499c4 tumors harvested from indicated mice on Day 19.

F. Mass of MH6499c4 tumors from panel B.

Data are represented as mean±S.E.M; Statistical significance was performed using one-way ANOVA with Tukey’s multiple-comparison test (A and D) and students’ t-test (C and F).

We next analyzed the status of splenic and intratumoral CD8+ T cells in these tumor-bearing mice. Comparison of the numbers of CD8+ T cells in the spleens did not reveal any significant differences between Atf4^+/+; Cd8a-Cre and Atf4^ΔCD8 mice bearing either MC38 or MH6499c4 tumors (Figure S4). However, notably lesser frequencies and numbers of CTL were found in both MC38 (Figure 6a–b, S6a and S5a-b) and MH6499c4 (Figure 6c–d) tumors grown in Atf4^ΔCD8 mice.

Fig 6. ATF4 in the intratumoral CD8+ T cells supports their activity and viability.

A. Frequency of CD3⁺CD8⁺ infiltrating T cells in MC38 colon adenocarcinoma tumors from Atf4^+/+ or Atf4^ΔCD8 mice (n=7).

B. Quantification of CD3⁺CD8⁺ infiltrating T cells from panel A.

C. Frequency of CD3⁺CD8⁺ infiltrating T cells in MH6499c4 tumors from Atf4^+/+ or Atf4^ΔCD8 mice (n=5).

D. Quantification of CD3⁺CD8⁺ infiltrating T cells from panel C.

E. Frequency of 69⁺ T cells among CD8⁺ T cells in MH6499c4 tumors from Atf4^+/+ or Atf4^ΔCD8 mice (n=5).

F. Frequency of PD-1⁺LAG3⁺TIM3⁺TIGIT⁺ T cells among CD8+ T cells in MH6499c4 tumors from Atf4^+/+ or Atf4^ΔCD8 mice (n=5).

G. Frequency of Annexin V⁺ T cells among CD8⁺ T cells in MH6499c4 tumors from Atf4^+/+ or Atf4^ΔCD8 mice (n=5)

Data are represented as mean±S.E.M; Statistical significance was performed using students’ t-test (B, D, E, F and G).

Furthermore, ATF4-deficient CD8+ T cells isolated from MH6499c4 tumors exhibited lower levels of CD69 activation marker (Figure 6e), greater levels of PD-1, which is associated with exhaustion of intratumoral T cells (Figure 6f), and increased frequencies of cells undergoing apoptosis as manifested by Annexin V staining (Figure 6g). Similar data were obtained in the tumor-infiltrating CD8+ T cells and CD8+CD3- cells (likely representing CD8a-expressing dendritic cells) from mice bearing MC38 tumors (Figure S5c-e). Importantly, such differences were not seen in CD8+ T cells isolated from the spleens of these mice (Figures S6b-d) suggesting that induction of ATF4 in the intratumoral CD8 T cells is important for their optimal function and viability.

DISCUSSION

Our data demonstrate that upregulation of ATF4 occurs in the intratumoral CTL (Figure 2). In vitro experiments show that this upregulation can be driven by antigenic activation, stress stimuli and tumor-derived factors (Figure 1–3). ATF4 supports viability of intratumoral CTL and protects these cells from apoptosis (Figure 4). In addition, loss of ATF4 impedes cytotoxic effects of CTLs. Accordingly, in the context of CTL, ATF4 function supports both survival and activity of CTLs and contributes to the anti-tumor immunity and restricts tumor growth (Figure 5–6). Importantly, additional important roles of ATF4 in other aspects of CTL biology (such as effect of ATF4 on CTL priming) cannot be ruled out.

These results were largely unexpected given that global inactivation of ATF4 in the TME notably suppressed tumor growth and progression (24). However, this report highlighted the critical importance of ATF4 expression specifically in fibroblasts and demonstrated that severity of the anti-tumor phenotype was comparable in mice that lacked ATF4 only in fibroblasts and in the entire TME. Furthermore, analysis of gene expression in tumors growing in ATF4-null mice at single cell levels revealed that the phenotype could not be explained by changes seen in the immune compartment (24). All these considerations point to a plausible explanation that a potent pro-tumorigenic role of ATF4 in fibroblasts and fibroblast-dependent changes in angiogenesis (24) mask a significant yet modest phenotype that we observe in mice specifically lacking ATF4 in CD8+ T cells.

Importantly, given that a subset of dendritic cells can also express the CD8a gene, we cannot rule out that some of phenotypes observed in Atf4^ΔCD8 mice can be attributed to the role of ATF4 in these dendritic cells. Indeed, we noticed that ablation of ATF4 decreases the number of CD3-CD8+ cells (Figure S5E). Thus, it is plausible that ATF4 may also play a role in the processes of antigen presentation mediated by the CD8a-expressing dendritic cells and that these changes further contribute to the phenotypes of tumor growth reported in this study.

A selective increase in the translation of ATF4 mRNA in response to the Integrated Stress Response and phosphorylation of eIF2α represents the key mechanism for ATF4 activation (8–10). Our studies show that levels of ATF4 mRNA are also increased in CD8+ T cells in response to antigen-specific or non-specific activation of the T cell receptor, stress or tumor-derived factors. These observations are consistent with previous reports that, depending on the nature of a stimulus, in addition to translational control, ATF4 expression is also subject to transcriptional regulation (11, 12). This additional mechanism for upregulation of ATF4 may perhaps reflect the importance of this transcription factor in CTL viability and activity.

The detailed mechanisms and downstream mediators that confer the pro-survival and activating function of ATF4 in the intratumoral CTL remain to be elucidated. It is probably safe to rule out the importance of CHOP (encoded by DDIT3 gene), which is induced in CD8+ T cells by T cell receptor stimulation and stress in a ATF4-dependent manner (Figure 2–3). However, CHOP has been implicated in driving cell death (reviewed in (31)). Moreover, genetic ablation of Ddit3 in T cells was shown to stimulate the anti-tumor CD8⁺ T cell immunity and to increase the efficacy of T cell-based immunotherapies (32).

It is more plausible that pro-survival microRNAs that are known to be regulated by ATF4 (16) can contribute to the role of ATF4 in anti-tumor immunity. For example, stress-induced ATF4-dependent miR-211 was shown to be recruited to the proximal promoter of DDIT3, where this recruitment led to an increased histone methylation, inhibition of CHOP expression and increased cell survival (17). Future studies aimed specifically at determining the role of miR-211 in CTL and anti-tumor immunity and, generally, at the understanding of putative mechanisms that regulate ATF4 function in CTL are warranted.

In addition to gaining the theoretical insight on understanding the role of ATF4 in regulating viability and activity of intratumoral CTL, our studies argue for caution to be exercised in the context of therapeutic efforts to inhibit ATF4 for treatment of solid tumors. Recent exciting progress in this area emerged with development and demonstration of anti-tumor activity for inhibitors of PERK (33, 34) as well as with development of ISRIB, a small molecule agent that binds to the nucleotide exchange factor eIF2B and prevents translational changes and activation of ATF4 downstream of the Integrated Stress Response (35–37).

Robust anti-tumorigenic effects of ISRIB were well documented in pre-clinical studies (22, 38–41). Importantly, ISRIB elicited attenuated T cell-mediated responses in the murine brain (42). Given these results along with our observations on the role of ATF4 in anti-tumorigenic function of CTL, clinical trials focused on safety and efficacy of ISRIB in human cancer patients should include immune profiling of tumors and extensive analyses of T cell function.

Data Availability:

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.