PTENα functions as an immune suppressor and promotes immune resistance in PTEN-mutant cancer

PTEN is frequently mutated in human cancers and PTEN mutants promote tumor progression and metastasis. PTEN mutations have been implicated in immune regulation, however, the underlying mechanism is largely unknown. Here, we report that PTENα, the isoform of PTEN, remains active in cancer bearing stop-gained PTEN mutations. Through counteraction of CD8+ T cell-mediated cytotoxicity, PTENα leads to T cell dysfunction and accelerates immune-resistant cancer progression. Clinical analysis further uncovers that PTENα-active mutations suppress host immune responses and result in poor prognosis in cancer as relative to PTENα-inactive mutations. Furthermore, germline deletion of Ptenα in mice increases cell susceptibility to immune attack through augmenting stress granule formation and limiting synthesis of peroxidases, leading to massive oxidative cell death and severe inflammatory damage. We propose that PTENα protects tumor from T cell killing and thus PTENα is a potential target in antitumor immunotherapy. PTENα is an N-terminally extended isoform of PTEN, a gene frequently mutated in human cancers. Here the authors show that PTENα remains active in PTEN-mutant cancers and is associated with tumor immune escape by promoting tumor cell resistance to T cell cytotoxicity.

T he importance of phosphatase and tensin homolog deleted on chromosome ten (PTEN) in tumorigenesis is underscored by its frequent mutations, including missense, nonsense, frameshift, and large deletions 1,2 . The tumorsuppressive activity of PTEN largely depends on its lipid phosphatase activity, which antagonizes PI3K/AKT activation 1 . Accordingly, a significant fraction of PTEN missense mutations is identified in human cancers, which selectively abolish its phosphatase activity 3 . Additionally, more than half of PTEN mutations result in protein truncation and instability, which are functionally comparable to PTEN loss 1,4 . In spite of the dysregulation of PI3K-AKT signaling in cancers with PTEN truncated or missense mutations, the outcomes of tumors with different types of PTEN mutations are distinct 5,6 . Therefore, an in-depth understanding of the biological functions of various PTEN mutations in tumorigenesis is still in need, and crucial for cancer treatment.
Through restoring or enhancing the effector function of CD8 + tumor-infiltrating T cells, immunotherapy has revolutionized the treatment of patients with advanced-stage cancers 7 . However, this therapy is limited to a minority of patients with cancer, and the clinical efficacy is often compromised along with tumor development 7,8 . Previous studies have revealed that tumor cells can hijack strategies developed to limit immune responses, thereby affecting priming, activation, and recruitment of T cells, causing tumor immune escape 9 . Apart from these processes of T cells, effective anti-tumor immunity also requires cytotoxic T lymphocytes (CTLs)-mediated cell killing 10 . Researches have shown that the function of CTLs depends on the ROS production in the target cells, and recent studies further uncover that CTLs cause different types of cell death depending on the context of the target cells [11][12][13] . These phenomena lead to the hypothesis that, instead of affecting T cell function, tumor-intrinsic factors may directly affect the susceptibility of tumor cells to CTLs. To date, researches on tumor immune escape still mainly focus on the effects of tumor cells on T cells, while little is known about whether tumor cells can escape from CTLs-mediated killing through modulating their own metabolic state.
PTENα (also known as PTEN-L) is the first identified isoform of PTEN, which contains an additional disordered 173-amino acid N terminus 14,15 . This region confers functions of PTENα that are distinct from those of PTEN 14,15 . The role of PTENα in tumors is complicated, which limits tumor growth in glioblastoma cells, while promotes carcinogenesis in liver cancer 14,16 . Up to now, the effects of PTENα in anti-tumor immunity remain elusive. In this study, we find that PTENα acts as an immune suppressor that restricts CD8 + T cell response against PTEN mutant cancers, consequently resulting in poor outcome of the disease. Rather than directly affecting T cell priming or differentiation, PTENα promotes the cell resistance to T cell-mediated cytotoxicity and results in tumor immune escape. Reciprocally, loss of PTENα down-regulates translation of proteins associated with the oxidation-reduction process, and potentiates ferroptosis upon incubation with cytotoxic T cells, exacerbating inflammatory responses and promoting immune clearance. Our data thus demonstrate that PTENα acts as a suppressor of T cell cytotoxicity-induced oxidative cell death, and highlight the importance of targeting PTENα in antitumor immune therapy.

Results
PTENα remains active in PTEN-mutant cancer. Loss-offunction mutations in the PTEN gene are frequent in human cancer 3 . In addition to phosphatase-inactive mutations, abundant stop-gained mutations were detected in exons encoding the C-terminal domain of PTEN (Fig. 1a). Through induction of PTEN instability, stop-gained mutations restrict the function of phosphatase PTEN, whose effects are largely identical to those of phosphatase-inactive mutations 1 . However, through analyzing the survival of patients with tumor-bearing PTEN mutations, we found that the outcome of patients with stop-gained mutations was poorer than those with phosphatase-inactive mutations ( Fig. 1b and supplementary Fig. 1a). We next analyzed differentially expressed genes in cancers with these two-type mutations based on the matched diagnosis age (40-60 years), aneuploidy score (<3), and neoplasm histologic grade (G3). Utilizing gene set enrichment analysis (GSEA), we found that genes related to adaptive immune response were selectively enriched in cancers with phosphatase-inactive mutations as compared with those with stop-gained mutations (Fig. 1c). Similar results were also detected between the two-type mutations in a broader context (Fig. 1d). These results lead to the hypothesis that, aside from tumor-promoting effects, stop-gained mutations of PTEN restrict host immune response against cancer.
As an isoform of PTEN, PTENα is initially translated from a CUG codon upstream of and in-frame with the coding region of canonical PTEN 15 . As shown in Supplementary Fig. 1b, phosphatase-inactive mutations impaired the phosphatase activity of PTEN and its isoform PTENα. To assess the effects of stopgained mutations on PTEN isoforms, we analyzed the protein stability of the PTEN family with or without stop-gained mutations. As shown in Fig. 1e, f and Supplementary Fig. 1c, in contrast to the identical half-life between wild-type PTEN and PTENα, stop-gained mutations accelerated the degradation of PTEN rather than PTENα. To further confirm this result, we utilized the Jurkat and MOLT4 cells, which have been reported to carry stop-gained PTEN mutations 17 . In accordance with the bioinformatics data, we found that MOLT4 cells contain PTEN K267Rfs*9 mutation and Jurkat cells contain PTEN R234Afs*1 and PTEN L247* mutations by DNA-sequencing analysis (Supplementary Fig. 1d). We then employed an anti-PTEN antibody to assess the status of endogenous PTEN and PTENα in Jurkat and MOLT4 cells. As shown in Supplementary Fig. 1d, compared with wild-type PTEN and PTENα in HEK-293T, truncated PTENα rather than PTEN can be detected in both Jurkat and MOLT4 cells. Our data thus indicate that PTENα remains active in cancer with stop-gained PTEN mutations ( Supplementary Fig. 1b).
PTENα promotes tumor immune escape. To investigate the role of PTENα in tumorigenesis, we used CRISPR/Cas9 technology to knock out the endogenous PTEN in the B16-F10 cell line, which also abolished the expression of PTENα, and subsequently reenforced the expression of PTENα in Pten −/− cell line. As shown in Supplementary Fig. 2a-c, ectopic expression of PTENα elicited little effects on cell proliferation in vitro. In light of the impaired immune response in cancer with stop-gained mutations, we thus employed irradiated B16-F10 cell line as tumor vaccine to assess the role of PTENα in modulation of host immune response against cancer (Fig. 2a). As shown in Fig. 2b, in contrast to the identical volume of tumor with or without PTENα in unimmunized C57BL/6 mice, overexpression of PTENα weakened the effectiveness of cancer vaccination. Through analysis of tumorinfiltrating lymphocytes (TIL) by flow cytometry, less amounts of CD45 + immune cells were detected in PTENα-expressing tumors as relative to control tumors (Fig. 2c).
To study the status of immune cells in PTENα-expressing tumors, we applied scRNA-seq methods to investigate CD45 + immune cells isolated from PTENα-expressing or control tumors with two replicates. To remove batch-effect, data integration and batch-effect correction were applied to the data. Utilizing graphbased clustering to analyze the tumor-infiltrating immune cells, we identified 12 clusters of immune cells for 10× data. We then defined the clusters based on the exclusive expression of canonical marker genes. As shown in Supplementary Fig. 2d, e, and Supplementary Table 1, major immune cell types in cancers including CD4 + T, CD8 + T, Treg, NK, monocyte, macrophages, neutrophil, B, plasma B, plasmacytoid dendritic cells (pDCs), and conventional DCs (cDCs). Compared with more than 80% of lymphoid cells in the control tumor, the percentage of T lymphocytes and B lymphocytes were significantly reduced in PTENα-expressing tumors ( Supplementary Fig. 2f). Considering the importance of CD8 + T cells in host antitumor immunity, we performed unsupervised clustering and identified four clusters for CD8 + T cells, each with its unique signature genes ( Fig. 2d and Supplementary Fig. 3a). In contrast to the other three clusters, the fourth cluster of CD8 + T cells, CD8T-Ctla4, expressed high levels of exhaustion markers Clta4, Pdcd1, and Lag3, thus representing exhausted CD8 + T cells (Fig. 2e). As expected, the percentage of C4_CD8T-Ctla4 was remarkably increased in PTENα-expressing tumors as compared with control tumors (Fig. 2f, g). Consistent with scRNA-seq results, subsequent flow cytometry revealed that overexpression of PTENα limited the amounts of infiltrated lymphocytes, and enhanced PD-1 and TIM-3 expression on CD8 + tumor-infiltrating T cells (supplementary Fig. 3b-f). Furthermore, we also measured the production of IFNγ in the tumor-infiltrating CD8 + T cells in the treatment of PMA/ ionomycin. As shown in Supplementary Fig. 3g, the production of IFNγ was weakened in CD8 + T cells from PTENα-expressing tumors relative to those in control tumors.
To further confirm the immune regulatory role of PTENα in cancer, we performed the adoptive T-cell transfer assay. Mock or PTENα-expressing B16-Pten −/− cells were infected with LCMV-Cl13, and subcutaneously injected into nude mice. 10 days later, the activated CD8 + T cells were harvested from spleen in mice infected with LCMV-Cl13 on day-7 post infection, and subsequently transferred into the tumor-bearing nude mice. Similar procedures were repeated on 14-day and 18-day in B16 tumorbearing mice (Fig. 2h). As shown in Fig. 2i, viral infection hardly influenced tumor growth of Mock or PTENα-expressing B16 cells in nude mice, and ectopic expression of PTENα elicited no effects on cell proliferation, which is consistent to the data in C57BL/6 mice (Fig. 2i). Moreover, the transferred CD8 + T cells exhibited significantly antitumor effects on Mock rather than PTENαexpressing B16 cells in nude mice (Fig. 2i). Notably, through analysis of CD8 + T cells infiltrating in spleen or tumors, we found that the amount of adoptive transferred CD8 + T cells were identical between the nude mice with PTENα-expressing B16 cells and those with Mock B16 cells ( Supplementary Fig. 3h). Taken together, our data demonstrate that PTENα blocks T cellmediated cancer eradication and promotes cancer immune escape.
Furthermore, we also investigated the effects of PTENα bearing stop-gained mutations in cancer immune escape. Analog to the immunosuppressive effects of wild-type PTENα, overexpression of PTENα CTG-R233* also reduced the effectiveness of tumor vaccine and facilitated tumor growth ( Supplementary Fig. 3i). To ascertain whether the immunosuppressive role of PTENα is dependent on its secretion, we also employed the vector of PTENα CTG-Δ6A that has been reported to abolish secretion of PTENα 14 . Utilizing the cancer vaccine model, we found that overexpression of PTENα CTG-Δ6A also promoted cancer immune escape, which is identical to the effects of PTENα CTG-R233* , indicating that the secretion characteristics of PTENα hardly affects its immunoregulatory function ( Supplementary Fig. 3i). Considering that PTENα mutations also exhibit clinical relevance in colon cancers, we employed CT26 cells (murine colorectal cancer cell line) to assess the role of PTENα in tumor vaccination model. We first used shRNA to knock down the endogenous PTEN expression in CT26 cells, and subsequently stably transfected with the vector encoding Mock or PTENα (Supplementary Fig. 3j). In accordance with the data in B16 cells, ectopic expression of PTENα elicited immunosuppressive effects and promoted tumor development in immunized mice (supplementary Fig. 3k). Together, our data thus indicate that the presence of PTENα in cancers subverts host immune attack.
PTENα promotes T cell exhaustion in a T cell-extrinsic manner. Since the presence of PTENα in cancers drives CD8 + T cell dysfunction, we employed Ptenα −/− mice to further study the role of PTENα in the modulation of T cell fate. Although PTENα was expressed in both lymphoid and non-lymphoid organs (Supplementary Fig. 4a), Ptenα −/− mice developed normally, and no spontaneous tumor was detected within 2-year old. In order to determine whether loss of PTENα affects the proliferation, differentiation, and survival of T cells, we firstly employed a CFSEstaining assay to measure the T cell expansion during activation. As shown in Supplementary Fig. 4b, c, deletion of PTENα elicited little effects on T cell expansion during T cell priming. Besides, identical ratio of iTreg cell (CD4 + CD25 + Foxp3 + ), Th1 (CD4 + IFNγ + ), Th17 (CD4 + IL-17A + ) and Tr1 (CD4 + IL-10 + ) cells were induced under various T-cell polarizing conditions between wild-type and Ptenα −/− mice ( Supplementary Fig. 4d). Finally, we used Annexin V/7-AAD staining to assess cell death during T cell activation. As shown in Supplementary Fig. 4e, no significant difference was detected between wild-type and Ptenα −/− T cells.
We next utilized lymphocytic choriomeningitis virus clone 13 (LCMV-Cl13) to induce T cell exhaustion. As shown in Fig. 3a, b, high viral loads were found in the lungs in wild-type mice, whereas Ptenα −/− mice showed a much-reduced viral burden on day 30 post infection. Compared with wild-type mice, the frequency of virus-specific CD8 + T cells (GP 33-41 + ) was increased in the lungs from Ptenα −/− mice ( Fig. 3c and Supplementary  Fig. 5a). Consistent with published studies 18 , we observed that high levels of inhibitory receptors including PD-1, TIM-3, and LAG-3 were expressed on the surface of virus-specific wild-type CD8 + T cells rather than Ptenα −/− CD8 + T cells during chronic infection (Fig. 3d). Moreover, we stimulated CD8 + GP 33-41 + T cells with virus-specific peptides to induce the production of effector cytokines in wild-type and Ptenα −/− T cells. As shown in Phosphatase-inactive mutations, n = 51, *P = 0.0377). Clinical data were acquired from the TCGA database. c GSEA of RNA-seq data of UCEC patients with matched diagnosis age (40-60 years), aneuploidy score (<3), and neoplasm histologic grade (G3). The data were acquired from the TCGA database. ES is an abbreviation of Enrichment Score, and NES represents Normalized Enrichment Score. d Expression levels of immune surveillance-related genes in uterine corpus endometrial carcinoma (UCEC) patients with indicated PTEN mutations (Stop gained mutations, n = 117; Phosphatase-inactive mutations, n = 51, mean ± s.e.m., *P (PTPRC) = 0.0400, *P (CD4) = 0.0498, **P = 0.0022). e and f Immunoblot analysis of expression of GFP-tagged PTEN and PTENα carrying indicated mutations in HEK293T cells treated with 100 μg/ml CHX for indicated hours. An anti-GFP antibody was used. Gray values of PTEN and PTENα relative to GAPDH were determined, using for the line chart. Supplementary Fig. 5b, Ptenα −/− T cells produced greater amounts of effector cytokines than wild-type T cells upon exposure to viral peptide. Similar results were also observed in virus-specific T cells from spleens ( Supplementary Fig. 5c-e). These data thus indicate that deletion of PTENα blocks T cell dysfunction.
We next sought to determine whether T cell polarization is dependent on T cell-intrinsic PTENα. To this end, we used a bone marrow (BM) transplantation model in which Ptenα −/− or wildtype mice were lethally irradiated and transplanted with wild-type or Ptenα −/− BM and subsequently infected with LCMV-Cl13. As shown in Fig. 3e, on day 30 post-infection, the numbers of virus-specific T cells were identical between Ptenα −/− recipients of wildtype BM and those of Ptenα −/− BM. Of note, compared with wildtype donor mice, greater amounts of virus-specific T cells were detected in the two types of Ptenα −/− recipient mice (Fig. 3e). Conversely, LCMV-Cl13 infection resulted in lower amounts of virus-specific T cells in wild-type recipients, no matter which kind of BM is adoptive transferred (Fig. 3f). Consistent results were also detected by assessment of inhibitory receptor expression and effector cytokine production (Supplementary Fig. 6a-d). Combined with the data that PTENα is also expressed in nonlymphoid organs ( Supplementary Fig. 4a), our data thus demonstrate that T cell-extrinsic PTENα drives T cell exhaustion.
PTENα promotes cell resistance to T cell cytotoxicity. To investigate the mechanism by which PTENα promotes T cell exhaustion, we analyze the epithelial-immune cell interaction. As shown in Fig. 4a, loss of PTENα triggered stronger inflammatory response in the early stage of higher titers of LCMV-Cl13 infection, causing massive congestion and edema in the lungs of Ptenα −/− mice upon viral infection. Moreover, greater amounts of death cells were detected in Ptenα −/− lungs assessed by TUNEL staining (Fig. 4b). Besides, we performed mass spectrometry analysis of tissue interstitial fluid (TIF) in the lungs during inflammation, and found that a series of nuclear proteins were released in extracellular matrix of lungs from Ptenα −/− mice at the early stage of viral infection, which acts as danger-associated molecular patterns (DAMPs) and activates immune response (Fig. 4c). Consistently, loss of PTENα increased the amounts of HSP70, HSP90, HMGB1, and eATP (extracellular ATP) in TIF in the lung during viral infection (Figs. 4d, e). These results thus indicate that deletion of PTENα promotes immunogenic cell death upon LCMV infection.
Since LCMV-Cl13 is a non-cytolytic virus that hardly induces cell death in vitro 19 , we hypothesized that T cell-mediated cytotoxicity mainly contributed to the severe cell death. To this end, we performed in vitro killing assay to investigate whether PTENα affects target cell responses to CTLs. We pulsed Mock or PTENα expressing B16-Pten −/− cells with LCMV-GP [33][34][35][36][37][38][39][40][41] peptide and incubated with CD8 + T cells isolated from spleens of C57BL/6 mice untreated or infected with LCMV-Cl13. Notably, incubation of CD8 + T cells from the infected mice induced significantly lower cell lysis rates in PTENα-expressing B16 cells relative to those control cells (Fig. 4f). Moreover, compared with PTENα-expressing cells, the T cell-mediated cytotoxicity of Mock B16-Pten −/− cells induced high level of ATP release and intracellular ROS accumulation, which indicated that loss of PTENα resulted in non-apoptotic cell death under immune attack (Fig. 4g, h). Similar results were also detected in Ptenα +/+ and Ptenα −/− MEFs ( Supplementary Fig. 7a-c). Our data thus demonstrate that CTLs triggers immunogenic death of Ptenα −/− cells that exacerbates inflammatory damage upon exposure to pathogens. The existence of PTENα correspondingly limits antigen clearance, resulting in T cell exhaustion.
PTENα restricts oxidative cell death. Previous studies have revealed that ROS is essential for CTLs-mediated cytotoxicity 11 . In consideration of the higher ROS level in PTENα-deletion cells induced by CTLs, we treated Ptenα +/+ and Ptenα −/− MEFs with hydrogen peroxide (H 2 O 2 ) to mimic CTLs stimulation and investigated the nature of cell death caused by PTENα deficiency. As shown in Supplementary Fig. 8a, b, PTENα deficiency markedly increased cell susceptibility to ROS-mediated cell death. Moreover, using a series of cell death inhibitors, we found that the ferroptosis inhibitor ferrostatin-1 (Fer-1) was the only one completely abolishing the cell death induced by Reciprocally, supplementation of the ferroptosis activator erastin and RSL3 significantly promoted cell death compared with other cell death agonists did ( Supplementary Fig. 8f, g). To further confirm that loss of PTENα selectively induces ferroptosis, we measured the status of GPX4, which is a key regulator of ferroptosis. As shown in Supplementary Fig. 8h, the protein levels of GPX4 were reduced in lungs from Ptenα −/− mice compared with wild-type mice upon exposure to the virus.
In order to ascertain that suppression of ferroptosis mainly contributes to the oncogenic function of PTENα, we used α-TOC which is a radical scavenger and lipid peroxidation inhibitor, to suppress ferroptosis commitment. Considering only a portion of α-TOC dosage can reach the tumor by oral administration and its water-insoluble characteristic, we exploited a molecular-matched strategy to prepare α-TOC-loading TPGS nanoparticles (NP-VE) with extremely high drug loading levels (up to 10 mg/ml) 20 , which can effectively limit ROS production ( Supplementary  Fig. 8i, j). In vitro assays showed that α-TOC treatment weakened the difference of cell death between PTENα-expressing cells and controls cells exposure to cytotoxic T cells (Fig. 5d). Besides, we also used BODIPY-C11 dye to analyze the level of lipid ROS. Analog to the intracellular ROS level, the presence of PTENα limited the level of lipid ROS in B16 cells (Pten −/− ) under T cell attack (Fig. 5e). Moreover, treatment of NP-VE in the cancer vaccine model promoted the progression of Mock-expressing B16-Pten −/− cells to a similar level of PTENα expressing B16-Pten −/− cells (Fig. 5f). Flow cytometry analysis of the infiltrating Tumor volumes of C57BL/6 mice subjected to the cancer vaccination model were monitored overtime. The unimmunized group refers to mice without irradiated B16 cells immunization. The statistical significances between the immunized groups were indicated (PTENαunimmunized, n = 4 mice; other groups, n = 5 mice, mean ± s.e.m., *P (d16) = 0.0254, *P (d20) = 0.0339, *P (d22) = 0.0487, *P (d24) = 0.0450, **P = 0.0077). c The tumor-infiltrating lymphocytes (TILs) were isolated when the tumor volumes of the Mock-immunized group reach 200 mm 3 , and the cells were subjected to flow cytometry analysis. The cell counts of CD45 + cells were used for statistical analysis (n = 4 mice, mean ± SD, **P = 0.0044). Gating strategies were shown in Supplementary Fig. 3b. d-g TILs were isolated when the tumor volumes of Mock-immunized group reach 200 mm 3 , and CD45 + cells were sorted, using for 10× scRNA-seq with two replicates. 25,098 cells were identified as 12 clusters utilizing graph-based clustering. d immune cells revealed that injection of NP-VE restricted the amount of CD45 + immune cells infiltrating in tumors as compared with untreated Mock tumors ( Supplementary Fig. 9a). Besides, the presence of PTENα exhibited little effects on the recruitment of immune cells in tumors treated with NP-VE ( Supplementary Fig. 9a), suggesting that blockade of oxidative cell death is essential for the immunosuppressive effects of PTENα. Further flow cytometry analysis demonstrated that NP-VE treatment or PTENα presence restricted both T and B cell recruitment in tumors ( Supplementary Fig. 9b, c). In accordance with the suppressive effects of NP-VE treatment on T cell recruitment, increased expression of PD1 and reduced production of IFNγ were also detected in CD8 + T cells isolated from mock tumors treated with NP-VE ( Supplementary Fig. 9d, e). Taken together, our data demonstrate that PTENα increases cell resistance to CTLs-induced immunogenic cell death and consequently leads to cancer immune escape.
PTENα blocks stress granules formation during oxidative stress. To investigate the molecular mechanism by which PTENα modulates cell response to immune attack, we performed pulldown assays with FLAG-tagged proteins coupled with mass spectrometry under the condition of H 2 O 2 treatment. Interestingly, we noticed that majority of PTENα-associated proteins were involved in stress granules assembly (G3BP1, G3BP2, etc.) and formation (eIF2α, PCBP1, PCBP2, ZC3HAV1, etc.) (Fig. 6a). Subsequent co-immunoprecipitation and immunofluorescence assays further confirmed the association of PTENα with stress granules components (Fig. 6b, c). To determine whether PTEN also participates in this process, we co-transfected PTEN or PTENα with eIF2α in the presence or absence of H 2 O 2 treatment. To determine whether PTENα influences stress granules formation, we transfected mock or PTENα into Pten −/− B16-cells followed by stimulating with H 2 O 2 , which is also a known inducer of stress granules. In spite of extensive co-localization of PTENα with intrinsic G3BP1 detected by immunofluorescence assay, overexpression of PTENα significantly inhibited H 2 O 2 -induced cytoplasmic foci formation compared with cells transfected with empty vectors (Fig. 6f, g). Together, these results indicate that PTENα selectively interacts with stress granules-associated proteins and blocks stress granules formation under stress conditions.
PTENα maintains protein synthesis of peroxidases. The phosphorylation of eIF2α is crucial for stress granules formation 21 . Given that PTENα possesses the canonical phosphatase activity and interacts with eIF2α, we hypothesized that PTENα modulates eIF2α phosphorylation under multiple kinds of stress conditions. As anticipated, stimulated by H 2 O 2 , overexpression of PTENα in Pten −/− B16 cells inhibited eIF2α phosphorylation (Fig. 7a, lanes 4  and 6 versus lanes 3 and 5). Similar results were also observed in MEFs (Supplementary Fig. 10a, lanes 4 and 6 versus lanes 3 and 5).
Additionally, unlike PTENα, enforced expression of PTEN or the phosphatase-dead PTENα-C297S mutant did not significantly influence the eIF2α phosphorylation status in Pten −/− MEFs upon H 2 O 2 exposure ( Supplementary Fig. 10a, lanes 4 and 6 versus lanes 3 and 5). Besides, we also performed in vitro phosphatase assay and found that PTENα directly dephosphorylates eIF2α in vitro (Fig. 7b, lane 3 versus lane 2). Our data thus confirm that PTENα acts as a phosphatase on eIF2α.
Stress granules are known to be related with selective translation [22][23][24] . To determine which proteins are regulated by eIF2α phosphorylation and consequent stress granules assembly in Ptenα −/− MEFs, we performed a mass spectrometry analysis of Ptenα +/+ and Ptenα −/− MEFs under oxidative stress. As shown in Fig. 7f and Supplementary Fig. 10d, loss of PTENα selectively affected the synthesis of proteins associated with translation and oxidation-reduction processes. Ensued immunoblotting analysis confirmed that protein level of GPX4 rather than its mRNA level was significantly decreased in PTENα-null cells under oxidative stress condition (Fig. 7g, lane 4 Supplementary Fig. 10f). Accordingly, intracellular ROS levels in H 2 O 2 -treated PTENα-expressing B16 cells were diminished relative to control cells (Fig. 7h).
Collectively, we identify PTENα acts as phosphatase of eIF2α, which in turn promotes synthesis of peroxidases and blocks oxidative cell death, eventually ameliorating cancer immune evasion.

Discussion
Despite the wide application of immunotherapy in the treatment of cancers, only a subset of patients responds to the treatment, and the clinical efficacy is often compromised along with tumor development 10 . Notably, a recent study uncovers that PTEN mutations are frequently observed in immune-resistant cancers 25 . Although the combination of immunotherapies and PI3K-AKT pathway inhibitors elicits partially effective to reinvigorate tumorinfiltrating T cells 26 , the immunosuppressive tumor environment can hardly be transformed, which suggests that other signaling pathways are also triggered in addition to PTEN dysfunction. Here we identify that presence of PTENα in PTEN-mutant cancers counteracts immune attack and promotes tumor immune escape ( Supplementary Fig. 10 g). Rather than affecting T cell priming, tumor-intrinsic PTENα restricts T cell-mediated cytotoxicity through limiting ferroptosis, which attenuates immunogenic cell death and maintains immunosuppressive niche. Our data thus indicate that PTENα can be a potential target of immune therapy to regulate the susceptibility of tumor cells to immune responses.
As an N-terminally extended isoform of PTEN, PTENα translates through a CUG codon upstream of and in-frame with the coding region of canonical PTEN 15 . Accompanied with PTEN, the phosphatase-inactive mutations also impair the phosphatase activity of PTENα. However, we find that the stopgained mutations, which result in PTEN instability and degradation, hardly affect PTENα stabilization. Notably, the presence of stop-gained mutations of PTEN is correlated with less T-cell infiltration and worse patient outcome, suggesting the oncogenic role of PTENα in tumor development. Interestingly, the role of PTENα in tumors seems complicated due to the opposite effects of PTENα on tumor growth in glioblastoma and liver cancer 14,16 . Although clinical analysis reveals that sustained expression of PTENα in liver cancer tissue is correlated with accelerated tumor progression, both types of research use the model of nude mice bearing tumor xenograft and fail to investigate the role of PTENα in anti-tumor immunity 14,16 . Herein, we identify the immunosuppressive function of PTENα, which attenuates cancer cell susceptibility to T cell-mediated cytotoxicity through limiting oxidative cell death, thereby promoting cancer escape from immune eradication.
Among the inhibitory effects of PTENα on host antitumor immune response, we regard its negative role in oxidative cell death as the core mechanism in vivo. First, the tumor vaccination model reveals that tumor-intrinsic rather than T cell-intrinsic PTENα mainly contributes to cancer immune evasion. Second, both in vitro killing assay and T cell adoptive transfer model demonstrate that reduction of immunogenic cell death and ensued DAMPs release is the cause of limited recruitment of immune cell. Third, chronic viral infection and BMT models demonstrate that reduced cell death and consequently persistent antigenic stimulation induced by PTENα drive exhausted T cell formation. Above all, our results demonstrated that, through protecting tumor cells from T cell killing, PTENα limits the infiltration of leukocytes in tumors and impairs T cell effector function, eventually leading to tumor immune escape.
The physiological function of PTENα remains largely unknown. Here we demonstrate that, as a phosphatase of eIF2α, PTENα maintains host protein synthesis, and limits stress granule formation. Stress granules have a great impact on cells' adaption to environmental changes through selective modulation of host protein synthesis under various stress conditions 23,24 . We found that PTENα restricts stress granule formation and maintains translation of proteins associated with oxidation-reduction processes, which in turn limits oxidative cell death and ameliorates the inflammatory response by blocking DAMPs release. In accordance with viral infection, it has been reported that Pseudomonas aeruginosa infection triggers severe immunopathology in mice without PTEN-L (PTENα) 27 . Our work thus supports the notion that the presence of PTENα is critical for host protection from pathogen-induced tissue damage. In PTEN mutant cancer, this process is hijacked, thereby promoting the adaption of tumor cells to T cell cytotoxicity, resulting in tumor immune escape.
Ferroptosis has been implicated in a variety of pathological disorders, such as neurodegenerative events, ischemia/reperfusion injury, and bacterial infection 28 . Besides, a previous study has revealed that CTLs can kill cancer cells by inducing ferroptotic cell death 12 . A recent study also uncovers that the sensitivity of tumor cells to oxidative stress and ferroptosis limits their metastasis through blood 29 . In this study, our results identify the inhibitory role of PTENα in the modulation of ferroptosis. Through neutralizing intracellular ROS, PTENα enhanced cancer cell resistance against T cell killing. Moreover, the presence of PTENα promoted cancer metastasis through blood. Therefore, these results further confirm the importance of PTENα in cancer immune escape and highlight the potential of targeting PTENα to be a promising way in cancer treatment. Fig. 4 PTENα promotes cell resistance to T cell cytotoxicity. a-e Wild-type and Ptenα −/− mice were i.v. infected with 1 × 10 6 PFU LCMV-Cl13. On day 7 post-infection, lungs were harvested from the mice. a The lungs were subjected to hematoxylin and eosin (H&E) staining. The image is representative of four mice with similar results. Scale bar: 200 μm. b TUNEL staining of lungs from wild-type and Ptenα −/− mice on day 7 post-infection. The image is representative of four mice with similar results. The red arrow indicates TUNEL positive cells. c The tissue interstitial fluids were isolated from the lungs and subjected to MS analysis. The differential proteins were analyzed using GSEA with GO gene sets. ES is an abbreviation of enrichment score, and NES represents normalized enrichment score. KO refers to Ptenα −/− . Each sample is a mixture of fluids from three mice. d and e The interstitial fluids were isolated from the lungs, and used for immunoblot analysis with indicated antibodies (d), and ATP content assay (e) (n = 4 mixed samples from three mice, mean ± SD, ***P = 0.0002). For immunoblot analysis, equal amounts of loading proteins were showed by the Ponceau S staining. The relative ratios of Hsp70, Hsp90, and HMGB1 to Ponceau S-stained total proteins were indicated. UT, untreatment. f-h In vitro cytotoxicity assay of B16 cells. Mock or PTENα expressing Pten −/− B16 cells pulsed with GP 33-41 peptide were used as target cells. Effector cells and target cells were incubated at indicated ratios. Effector-uninfection refers to T cells from uninfected mice and acts as a negative control. f 20 h post-incubation, B16 cells were washed to remove lymphocytes and counted using CCK-8 (n = 3 cell cultures, mean ± SD, ***P (2) = 0.0006, ***P (25) = 0.0003, ****P < 0.0001). g Culture medium was collected 20 h post-incubation and centrifuged to remove cells and debris. The supernatant was subjected to ATP content assay (n = 3 cell cultures, mean ± SD, ns, not significant, **P = 0.0078, ****P < 0.0001). Untreatment control refers to pulsed B16 cells without effector incubation. h 12 h postincubation, B16 cells were harvested and stained with DCFDA, using for flow cytometry analysis. Cells with higher intensity of FSC and SSC were gated and considered as B16 cells. Mean fluorescence intensity (MFI) of DCFDA was used for statistical analysis (n = 3 cell cultures, mean ± SD, *P = 0.0255). Untreatment control refers to pulsed B16 cells without effector incubation. Statistical significance was assessed by two-tailed unpaired Student's t test (h) or one (e) or two (f and g)-way ANOVA followed by Tukey's multiple comparisons test. Data are representative of two (a, b, d-h) independent experiments. Source data are provided as a Source Data file. See also Supplementary Fig. 7.

Methods
Mice. Ptenα −/− mice (C57BL/6J background) were generated and reported in our previous study 30 . Balb/c and Balb/c nude mice were purchased from Charles River Laboratories. 6-8 weeks old male mice were used for the study. All animals were maintained under a specific pathogen-free condition at the Department of Laboratory Animal Science of Peking University Health Science Center and reared in standard conditions with controlled temperature (20-26°C), humidity (40-70%), Short hairpin RNA infection assay. pLKO.1-puro was used to knock down gene expression. Sequence for Gpx4 shRNA interference was 5′-CCGGGGAGCCCAT TCCTGAACCTTTCTCGAGAAAGGTTCAGGAATGGGCTCCTTTTTG-3′, and sequence for Pten shRNA interference was 5′-CCGGGGGTAAATACGTTCTT CATACCTCGAGGTATGAAGAACGTATTTACCCTTTTTG-3′.
Cell proliferation assay. Different numbers of cells were seeded into the 96-well plates by the gradient. Then the cells were subjected to a Cell counting Kit-8 (CCK-8, Dojindo) assay following the manufacturer's instructions to draw the stand curve. For assessment of cell proliferation, the B16 cells were seeded into a 96-well plate at the density of 1000 cells per well in 100 μl DMEM medium containing 10% (vol/vol) fetal bovine serum (FBS). The cells were cultured at 37°C and measured using CCK-8 daily. The cell numbers were determined by comparing them with the stand curve.
Colony formation assay. B16 cells were seeded in a six-well plate at the density of 100 cells per well and cultured at 37°C for 7 days. After fixing in methanol for 20 min and washing with PBS twice, the cells were stained with Crystal Violet Staining Solution (Beyotime) at room temperature for 20 min, using for photographing.
Melanoma mouse model. For the cancer vaccination model, 1 × 10 5 B16-F10 cells were irradiated (40 Gy), and cultured at 37°C overnight. Then the dying cells were collected and injected subcutaneously into the left flank of C57BL/6 mice. 10 days post-injection, 1 × 10 6 B16 cells were re-injected into the right flank of the mice through subcutaneous injection. The tumor volumes were monitored over time.
For adoptive T cell transfer assay, B16 cells were infected with 1 MOI LCMV-Cl13 for 24 h, and subcutaneous injected into the right flank of Balb/c nude mice. Meanwhile, C57BL/6 mice were intravenous (i.v.) infected with 1 × 10 6 LCMV-Cl13, and CD8 + T cells were isolated from the spleen of the mice on day 7 post infection. Then the CD8 + T cells were transferred into the tumor-bearing nude mice on day 10, 14, and 18 post tumor inoculation. Tumor volumes were monitored overtime. Quantitative real-time PCR (qRT-PCR). TRIzol reagents (invitrogen) were used to isolate total RNA from cells or tissues, followed by reverse transcribing with GoScriptTM Reverse Transcription System (Promega). TransStart Top Green qPCRSuperMix (TransGen Biotech) was used for real-time PCR. qRT-PCR was performed using Applied Biosystems 7500 Fast & 7500 Real-Time PCR System and 7500 Software v2.3 (Applied Biosystems) (all primers are listed in Supplementary  Table 2).
Bone marrow transplantation (BMT). To obtain bone marrow cells, donor mice were sacrificed, and the femurs and tibias were harvested. The bone marrow canals were washed out with RPMI1640 medium containing 1% (vol/vol) FBS, and the bone marrow cells were collected, followed by red blood cell (RBC) lysis. Recipient mice were lethally irradiated (896 cGy/mouse), and i.v. injected with 5 × 10 6 bone marrow cells. Experiments on transplanted mice were performed after a latency of 30 days to ensure bone marrow reconstitution.
In vitro cytotoxicity assay. C57BL/6 mice were untreated or i.v. infected with 1 × 10 6 PFU LCMV-Cl13. On day 7 post infection, lymphocytes were isolated from spleens of the infected mice, and the CD8 + T cells were isolated using Magni-Sort TM Mouse CD8 Positive Selection Kit (eBioscience). The CD8 + T cells were incubated with target cells pulsed with 2 μg/ml GP 33-41 peptide at indicated ratios. 8 h post incubation, target cells were collected and stained with DCFDA, using for flow cytometry analysis. For percent cell lysis rate, target cells were washed to remove lymphocytes, and live cells were measured using Cell counting Kit-8 (CCK-8, Dojindo) following the manufacturer's instructions.
Detection of ROS production. ROS-sensitive fluorescent probe 2′,7′-dichlorodihydrofluorescein diacetate (H2DCFDA) was used to detect generation of ROS in cultured cells. Briefly, cells were washed with PBS, followed by staining with 100 mM DCFDA for 30 min at 37°C. After washing with FACS (PBS + 1% (vol/ vol) FBS), stained cells were subjected to flow cytometry analysis.
For detection of lipid peroxidation, cells were collected, and stained with 2 μM BODIPY-C11. Lipid ROS level was determined by flow cytometry analysis with excitation at 488 nm. Mock or PTENα-expressing B16-Pten −/− cells were treated with 500 μM H 2 O 2 and the percent cytotoxicity was assessed using LDH release assay (n = 4 cell cultures, mean ± SD, ****P < 0.0001). UT untreatment. c Mock or PTENα-expressing B16-Pten −/− cells were treated with 500 μM H 2 O 2 in the absence or presence of 1 μM Fer1. Cells were stained with propidium iodide (PI), followed by flow cytometric analysis. All cells were gated. The death cell rates were used for statistical analysis (n = 3 cell cultures, mean ± SD, ****P < 0.0001). UT untreatment. d and e In vitro cytotoxicity assay for B16 cells. Mock or PTENα-expressing B16-Pten −/− cells pulsed with GP 33-41 peptide were used as target cells. Effector cells and target cells were incubated at indicated ratios. Effector-uninfection refers to non-antigen-specific T cells and acts as a negative control. d Before incubation, CD8 + T cells were stained with CFSE, then the effector cells and target cells were incubated in the presence or absence of 20 mM NP-VE. 20 h postincubation, the cells were harvested and subjected to PI staining. Death cells in the CFSE − cells were assessed by flow cytometric analysis (n = 3 cell cultures, mean ± SD, ns not significant, *P = 0.0188, ****P < 0.0001). Gating strategies were shown. e 12 h post-incubation, cells were harvested and stained with BODIPY-C11, followed by flow cytometric analysis. The gating strategy was identical to that in Fig. 4h. MFI of the target cells was used for statistical analysis (n = 3 cell cultures, mean ± SD, *P = 0.0207). f Mock or PTENα transfected Pten −/− B16 cells were subjected to the cancer vaccination model and subcutaneously injected in the presence or absence of 50 μl of 5 M NP-VE. Tumor volumes were monitored over time, and the statistical significances between the Mock groups were indicated (PTENα + NP-VE, n = 7 mice; other groups, n = 6 mice, mean ± s.e.m., *P = 0.0465, **P (d20) = 0.0015, **P (d20) = 0.0019). Statistical significance was assessed by two-tailed unpaired Student's t test (b, e, f) or one (d) or two (c)-way ANOVA followed by Tukey's multiple comparisons test. Data are representative of two (a-f) independent experiments. Source data are provided as a Source Data file. See also Supplementary Figs. 8 and 9.
TUNEL staining. For the detection of Terminal deoxynucleotidyl transferase (TdT)-mediated dUTP nick end labeling (TUNEL) positive cells, the in-situ cell death detection kit (Roche) was used according to the manufacturer's instructions.
Confocal microscopy. Cells were seeded on cover glasses. 12 h later, cells were fixed with 100% acetone, permeabilized with 0.5% (vol/vol) Triton X-100 and then blocked using 1% (wt/vol) BSA dissolved in PBS. Primary antibody was applied for at least 1 h, followed by incubation with fluorophore-conjugated secondary antibody for 1 h. DAPI (BioDee Biotechnology) was used to indicate nuclear, and cells were evaluated with fluorescence microscopy using NIS-Elements AR Analysis 4.20.00 64 bit software (Nikon). NIS Elements Viewer 4.20 was used to analyze the confocal microscopy data.
LDH release assay. The CytoTox 96R Non-Radioactive Cytotoxicity Assay (Promega) was used to detect lactate dehydrogenase (LDH) release and cell death rates. In brief, cells were seeded in 96-well plates and treated with indicated reagents. The experimental LDH release and maximum LDH release were determined according to the manufacturer's instructions, and the percentage of cytotoxicity was calculated by dividing experimental LDH release by maximum LDH release.
Interstitial fluid isolation. Tissue was harvested from mice. Bulk tissue was placed on a 40 μm cell strainer, followed by centrifuging at 45 × g for 5 min at 4°C to remove surface fluid. Then the tissue was centrifuged at 500 × g for 10 min at 4°C, and the interstitial fluid was collected. After spinning at 10,000 × g for 10 min at 4°C to remove insoluble particles, the supernatant was subjected to immunoblot analysis and ATP content assay.
Ponceau S staining. Protein samples were subjected to immunoblot analysis. After the electrophoresis and transmembrane, polyvinylidene fluoride membrane was stained by Ponceau S staining solution (Beyotime) for 30 min. The membrane was washed and used for scanning.
ATP content assay. ATP contents in the interstitial fluids were detected using ATP assay kit (Nanjing Jiancheng) according to the manufacturer's instructions.
Propidiumiodide (PI) staining. Cultured cells were digested and fixed with 100% Ethanol, followed by staining with 1.1 mg/ml PI containing 1% (vol/vol) Triton-X100. Stained cells were used for flow cytometry analysis. For cell cycle analysis, peaks of cells with diploid DNA content were analyzed. Determination of cell death was performed by analyzing the sub-G1 peak.
Preparation of lymphocytes. Lymphocytes in spleen were isolated by grinding the tissues in PBS containing 1% (vol/vol) FBS, followed by filtering through a 75 μm strainer. To isolate lymphocytes infiltrating in the lung, minced tissues were digested with 0.5 mg/ml Collagenase D (Roche) and 25 μg/ml DNase I (Sigma) at 37°C for 30 min. Then the tissues were grinded, and filtered through a 75 μm strainer. A 40% (10 ml)/70% (5 ml) Percoll gradient (GE Healthcare) was used to isolate the tissues by centrifuging 800 × g for 20 min. Cells at the inter-layer were collected and counted for further operation. For isolation of TIL, tumors were minced, grinded and filtered just like the lungs mentioned above. Then the cells were resuspended in 5 ml FACS buffer, adding to a 40% (5 ml)/80% (5 ml) Percoll gradient. The gradient was centrifuged at 400 × g for 45 min, and the cells at the inter-layer of 40%/80% Percoll gradient was collected for further operation.
CFSE staining. For CFSE staining, splenocytes were harvested, and washed twice with PBS, followed by staining with CFSE at dark for 10 min. Then the cells were washed twice with PBS containing 5% (w/w) FBS, and subjected to flow cytometry analysis.
Flow cytometry. For surface staining, cells were incubated with specific antibodies for 30 min at room temperature. The flow cytometry analyzer (BD Biosciences), FACSuite Software Bundle v1.0 (BD Biosciences), and FACSDiva Software v6.1 (BD Biosciences) were used for acquiring and analyzing the cells. The FACS data were analyzed with FlowJo v7.6.1 software. To perform intracellular cytokine staining, cells were stimulated with 100 ng/ml PMA and 500 ng/ml ionomycin, or 2 μg/ml GP 33-41 peptide together with GolgiPlug and GolgiStop (BD) for 5 h. Subsequently, cells were fixed and permeabilized, followed by staining with specific anti-cytokine antibodies.
Pulldown assay and mass spectrometry. H1299 cells were seeded and transfected with FLAG-tagged-PTENα. The whole-cell extracts were incubated with FLAG-beads (Sigma-Aldrich), followed by washing four times using PBS-N (PBS with 0.1% (vol/vol) NP40). Immunoprecipitated proteins were eluted using the 3× FLAG-peptide and subjected to SDS-PAGE. After staining with Coomassie Brilliant blue, the gel was excised and subjected to in-gel trypsin digestion and dried. 10 µl 0.1% formic acid was used to dissolve the peptides, and the peptides were sampled onto a 100 µm × 10 cm fused silica emitter packed with reversed-phase ReproSil-Pur C18-AQ resin (3 µm and 120 Å; Ammerbuch). Linear gradients of 5-32% acetonitrile in 0.1% formic acid were used to elute the sample at a flow rate of 300 nl/min for 50 min. An LTQ Orbitrap Elite mass spectrometer (Thermo-Fisher) equipped with a nanoelectrospray ion source (ProxeonBiosystems) was used to acquire the mass spectra data. Collision-induced dissociation (normalized collision energy, 35%; activation Q, 0.250; activation time, 10 ms) was used to perform fragmentation in the LTQ with a target value of 3000 ions. To search the raw files, the SEQUEST engine against a database from the Uniprot protein sequence database was used. Parameters were set as follows: protein modifications were set to carbamidomethylation (C) (fixed), oxidation (M) (variable), and phosphorylation (S, T, Y) (variable); the enzyme specificity was set to trypsin; a maximum missed cleavages were set to 2; the precursor ion mass tolerance was set to 10 ppm, and MS/MS tolerance was set to 0.5 Da.
Co-immunoprecipitation and immunoblot analysis. HEK293T cells were lysed by Co-IP lysis buffer containing 150 mM NaCl, 1 mM EDTA, 20 mM Tris-HCl pH 8.0, 10% (vol/vol) glycerol, 0.5% (vol/vol) NP40, and Protease inhibitor Cocktail (Roche). Cell lysates were subjected to SDS-PAGE and immunoblot analysis. For the co-immunoprecipitation assay, IgG or the appropriate antibodies were added into the cell lysates, followed by incubation with Protein A/G beads (EMD Fig. 6 PTENα blocks stress granules formation during oxidative stress. a MS analysis of PTENα associated proteins by FLAG pull-down assay in Pten −/− B16 cells treated with 100 μM H 2 O 2 . Proteins that interact with PTENα are indicated on the right (outlined text). b HEK293T cells were co-transfected with FLAG-tagged PTENα and GFP-tagged indicated vectors, followed by 100 μM H 2 O 2 treatment for 3 h. Cell lysates were immunoprecipitated with anti-FLAG antibodies or normal mouse IgG. The immunoprecipitated proteins were subjected to immunoblot analysis with anti-GFP antibodies. c HeLa cells were cotransfected with FLAG-tagged PTENα and GFP-tagged eIF2α vectors, and interaction between the proteins was shown by confocal fluorescence microscopy. DAPI was used to indicate nuclear. Scale bars, 10 μm. d HEK293T cells were co-transfected with FLAG-tagged eIF2α and GFP-tagged PTEN or PTENα vectors in the absence or presence of H 2 O 2 (100 μM). Cell lysates were immunoprecipitated with anti-FLAG antibody and analyzed by immunoblot with anti-GFP antibody. '*' refers to unspecific band. e HEK293T cells were transfected to express FLAG-tagged PTEN or PTENα, together with GFP-tagged G3BPs, followed by treatment of 100 μM H 2 O 2 for 3 h. Cell lysates were immunoprecipitated with anti-FLAG antibody and assessed by immunoblot analysis with anti-GFP antibody. f and g Pten −/− B16 cells were transfected with Mock or FLAG-tagged PTENα vectors, followed by confocal fluorescence microscopy analysis in the presence or absence of 100 μM H 2 O 2 with anti-G3BP1 and anti-PTEN (138G6) antibodies (f). Granules in the cells were counted, using for statistical analysis (g) (n = 20 cells, mean ± SD, ****P < 0.0001). Scale bars, 10 μm. UT, untreatment. Statistical significance was assessed by a two-tailed unpaired Student's t test (g). Data are representative of two (b-i) independent experiments. Source data are provided as a Source Data file.
Millipore) for at least 4 h. The precipitants were washed with PBS-N (PBS with 0.1% (vol/vol) NP40) four times, using for immune blotting analysis.
In vitro phosphatase assay. HEK293T cells were transfected to express FLAGtagged eIF2α and treated with 100 μM H 2 O 2 to induce phosphorylation of eIF2α. Then the phosphorylated eIF2α was precipitated with FLAG beads, and incubated with precipitated FLAG-tagged PTENα in phosphatase buffer (20 mM Hepes pH = 7.2, 1 mM DTT, 1 mM MgCl 2, and 1 mM EDTA) for one hour. The proteins were eluted with FLAG peptide and subjected to immunoblot analysis.
Single-cell RNA sequencing sample preparation and data analysis. For sample preparation, Mock or PTENα transfected Pten −/− B16 cells were subjected to the cancer vaccination model. The tumors were collected when the tumor volumes of Mock groups reached 200 mm 3 . The TILs were isolated, and CD45 + 7AAD − cells were sorted, using for scRNA-seq with two replicates. Each sample is a mix of cells from two mice. Library construction and 10× scRNA-seq were performed by Berry Genomics. Sequencing was performed in different sequencing lanes in Illumina NovaSeq 6000.
For raw data processing and quality control, Cell Ranger (version 6.0.1) was used to produce a raw unique molecular identifier (UMI) count matrix through processes including cellular barcodes demultiplex, mapping reads to the transcriptome and generating quantitative matrix. The matrix was converted into a Seurat object by the R package Seurat (version 4.0.1). Low-quality cells with UMI numbers >5000 or <300, or have over 10% mitochondrial-derived UMI counts were removed. 25,098 remained single cells were applied in downstream analyses.
Subsequently, the count matrix was normalized with the SCTransform function. Then we performed integration on the four datasets. In this process, potential Anchors were created with the FindIntegrationAnchors function using top 3000 variable genes, and IntegrateData function was used to integrate data and create a new matrix with 3000 features, in which potential batch effect was regressed out.
To reduce the dimensionality of the scRNA-Seq dataset, principal component analysis (PCA) was performed. Top 30 PCs were used to perform the downstream analysis with Elbowplot function. Subsequently, t-distributed stochastic neighbor embedding (tSNE) and uniform manifold approximation and projection (UMAP) were performed. The FindClusters function was used to divide all cells into 22 clusters with resolution set as 0.6, and the FindAllmarkers function was performed to identify preferentially expressed genes in clusters. Based on these genes, the clusters were identified as 12 cell types, and visualized with 2D tSNE plots, and the heatmap plot was made with the DoHeatmap function. The proportion of each cell type in the four samples were counted by the R package dplyr (1.0.6), and the bar graph were drawn using R package ggplot2 (3.3.3).
For further analysis of the subpopulation of Cd8T cells, all Cd8T cells were divided into four clusters with the FindClusters function, with resolution set as 0.08. The FindAllMarkers function was performed to identify preferentially expressed genes in the clusters. Based on these genes, four clusters were identified as four cell types, and visualized with 2D UMAP plot. Heatmap plot was made using the DoHeatmap function, and the proportion of each cell type in the Cd8T cells were counted by the R package dplyr (1.0.6). The bar graph was made using the R package ggplot2 (3.3.3). We also drawn the violin plots to show the expression of important genes in the Cd8T cells with the VlnPlot function.
Bioinformatics. Mutations, mRNA expression data and clinical data were from cBioportal 31,32 . Stop gained mutations includes all the stop-gained and frameshift mutations early terminated after H 185 (last amino acid of phosphatase tensin-type domain of PTEN). Loss of phosphatase activity mutations were from UniProt database.
For gene-set enrichment analysis (GSEA), differentially expressed genes or proteins were analyzed using applications from Broad Institute 21 (http:// www.broad.mit.edu/gsea/software/software_index.html) with default parameters. GO gene sets were downloaded from the 23 Broad Institute Molecular Signature Database (MSigDB).