GM-CSF mediates immune evasion via upregulation of PD-L1 expression in extranodal natural killer/T cell lymphoma

Background Granulocyte-macrophage colony stimulating factor (GM-CSF) is a cytokine that is used as an immunopotentiator for anti-tumor therapies in recent years. We found that some of the extranodal natural killer/T cell lymphoma (ENKTL) patients with the treatment of hGM-CSF rapidly experienced disease progression, but the underlying mechanisms remain to be elucidated. Here, we aimed to explore the mechanisms of disease progression triggered by GM-CSF in ENKTL. Methods The mouse models bearing EL4 cell tumors were established to investigate the effects of GM-CSF on tumor growth and T cell infiltration and function. Human ENKTL cell lines including NK-YS, SNK-6, and SNT-8 were used to explore the expression of programmed death-ligand 1 (PD-L1) induced by GM-CSF. To further study the mechanisms of disease progression of ENKTL in detail, the mutations and gene expression profile were examined by next-generation sequence (NGS) in the ENKTL patient’s tumor tissue samples. Results The mouse-bearing EL4 cell tumor exhibited a faster tumor growth rate and poorer survival in the treatment with GM-CSF alone than in treatment with IgG or the combination of GM-CSF and PD-1 antibody. The PD-L1 expression at mRNA and protein levels was significantly increased in ENKTL cells treated with GM-CSF. STAT5A high-frequency mutation including p.R131G, p.D475N, p.F706fs, p.V707E, and p.S710F was found in 12 ENKTL cases with baseline tissue samples. Importantly, STAT5A-V706fs mutation tumor cells exhibited increased activation of STAT5A pathway and PD-L1 overexpression in the presence of GM-CSF. Conclusions These findings demonstrate that GM-CSF potentially triggers the loss of tumor immune surveillance in ENKTL patients and promotes disease progression, which is associated with STAT5 mutations and JAK2 hyperphosphorylation and then upregulates the expression of PD-L1. These may provide new concepts for GM-CSF application and new strategies for the treatment of ENKTL. Supplementary Information The online version contains supplementary material available at 10.1186/s12943-021-01374-y.


Introduction
ENKTL is a rare type of mature T-cell lymphoma with high malignancy and low sensitivity to anthracyclinebased chemotherapy [1][2][3]. Asparaginase-based chemotherapy combined with radiotherapy is the main treatment for early-stage diseases, while the regimen of treatment for advanced diseases and relapsed/refractory diseases is high-dose chemotherapy followed by autologous hematopoietic stem cell transplant after standard asparaginase-based chemotherapy [4].
Granulocyte-macrophage colony stimulating factor (GM-CSF) is a cytokine that regulates the proliferation and differentiation of myeloid stem cells and mature granulocytes [5]. Recombinant human GM-CSF (hGM-CSF) was used to prevent and treat leukopenia, bone marrow hematopoietic dysfunction, myelodysplastic syndrome [6,7], and is used as an immunopotentiator for anti-tumor therapies in recent years [8][9][10]. The biological effects of GM-CSF primarily depend on its binding to GM-CSF receptor α (GM-CSFRα). Activation of the GM-CSF receptor is known to stimulate the janus kinase/signal transducer and activator of transcription (JAK/STAT) pathway, which has proved to be one of the pathways that regulate the expression of PD-L1 [11,12].
PD-1/CD279 is one of the co-inhibitory receptors that expresses on antigen-stimulated T-cells and interacts with two ligands, PD-L1 and PD-L2 [13]. Elevated expression of PD-L1 is related to poor prognosis and chemotherapy resistance, which has been confirmed in several types of tumors, including ENKTL [14,15]. Immunotherapies targeting PD1/PD-L1 have shown potential efficacy in a wide range of tumors, such as melanoma, non-small cell lung cancer (NSCLC), renal cell carcinoma, and Hodgkin's lymphoma [16][17][18][19]. Remarkably, it has been reported that PD-1 blockade was a potent strategy for NK/T-cell lymphomas when Lasparaginase regimens fail [20].
From clinical observation, we found that some of the ENKTL patients would develop disease progression after GM-CSF treatment. A recent study has reported that gastric cancer-derived GM-CSF activated neutrophils and induced neutrophil PD-L1 expression via JAK2/STAT3 signaling pathway. And the activated PD-L1 positive neutrophils effectively suppressed normal T-cell immunity in vitro and contributed to the growth and progression in gastric cancer in vivo [21]. Owing to the importance of JAK2/STAT pathway in hematological malignancies [22,23], and recurrent aberrations of JAK2 in several types of lymphoma, especially T-cell lymphoma [24], we hypothesized that GM-CSF might induce disease progression of ENKTL via JAK/STAT/PD-L1 axis.
In the current study, we found the disease progression caused by GM-CSF in T-cell lymphoma was associated with the decreases of tumor-infiltrating T cells and granzyme B release, which could be restored with PD-1 therapy. JAK2 hyperphosphorylation was related to poor survival and disease progression induced by GM-CSF. According to the next-generation sequencing (NGS) results, STAT5A was identified as one of the most frequently mutated genes in ENKTL patients. STAT5A mutations resulted in constitutive activation of STAT5 and PD-L1 expression. GM-CSF treatment induced higher PD-L1 expression in STAT5A mutated cells than in STAT5A wild-type cells. Together, these findings demonstrated that STAT5A mutations and JAK2 hyperphosphorylation might be the trigger for upregulation of PD-L1 induced by GM-CSF in ENKTL cells. It might explain the reason why ENKTL patients commonly underwent disease progression after GM-CSF treatment.

Western blot analysis
Cells were treated with the indicated concentrations as shown in the figures and washed twice with cold PBS. Whole-cell extracts were collected in RIPA lysis buffer (Santa Cruz Biotechnology, Germany), and protein concentration of the lysates was measured using a BCA Protein Assay Kit (Pierce Biotechnology, USA). The protein samples were electrophoresed through a 10% SDS-PAGE gel and transferred to a polyvinylidene difluoride (PVDF) membrane (Millipore, USA). After blocking, membranes were probed with primary antibodies (1:1000) followed by washing and incubation with a secondary antibody (1: 5000) conjugated to horseradish peroxidase (Amersham GE Healthcare, USA). Protein bands were visualized by applying a chemiluminescent reagent (Pierce ECL kit, Thermo Fisher Scientific, USA) [28]. The antibody against Phospho-Jak2 (Tyr1007/1008), JAK2, Phospho-Stat5 (Tyr694), STAT5, PD-L1 were purchased from Cell Signaling Technology (USA). The antibody against Phospho-STAT5A (Y694), STAT5A, Phospho-STAT5B (S731), STAT5B were purchased from Abcam (USA).

Transfection of shRNAs and plasmid DNAs
STAT5A and STAT5B shRNAs and an shRNA scramble control (Open Biosystems GE Healthcare Dharmacon Inc., USA) were transiently transfected along with a pSIH-H1-puro Lentivector Packaging Kit (System Biosciences, USA). Transfections were carried out in 293 T cells reaching ∼80% confluency in 10 cm dishes using Lipofectamine 2000 transfection reagent (Life Technologies, USA) and following the manufacturer's instructions. Transfection medium was replaced with fresh growth medium 5 h after transfection. At 48 h following the initial transfection, viral supernatant was collected and filtered through a 0.45 nm filter (System Biosciences, USA). SNK-6 and SNT-8 cells were infected and incubated with the viral particles overnight at 37°C. At 48 h after transfection, cells were placed under puromycin selection by supplementing the growth medium with puromycin (3 μg/ml for SNK-6, and 5μg/ml for SNT-8). Growth medium was replaced every 48 h for 2-3 weeks until isolated colonies (∼2 mm diameter) were apparent on the plate. At this point, individual clones were transferred to 12-well dishes and expanded in 1 μg/ ml puromycin for further analysis. Individual clones were verified by Western blot and RT-PCR [31].
In vivo mouse studies C57BL/6 mouse was purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd., and kept in a specific pathogen-free (SPF) barrier facility at the Animal Center of Sun Yat-sen University Cancer Center (n = 80). 8-12 weeks old female mice were used for all animal experiments. Experiments were approved by the institutional committee of Sun Yat-sen University Cancer Center (No. L102042018120I), and conducted following protocols approved by the Guangdong Provincial Animal Care and Use Committee.
EL4 cells or B16-F10 cells (5 × 10 5 cells in 200 μL growth medium) were subcutaneously injected into the right flank of immunocompetent C57BL/6 mouse. Tumor sizes were measured with calipers every 2 days and tumor volumes were calculated by applying the following formula: 1/2 (length×width 2 ). When tumor sizes reached approximately 100 mm 3 , mice were randomized into control or experimental groups. The terminal event was defined as tumors reaching a size of 2000 mm 3 in the control group, at which point animals were euthanized [32,33].

Patients and tissue specimens
Clinical data and/or samples of 109 ENKTL patients, 4 peripheral T-cell lymphoma, not otherwise specified (PTCL-NOS) patients, and 4 angioimmunoblastic T-cell lymphoma (AITL) from September 1999 to March 2019 were collected from Sun Yat-sen University Cancer Center (Figs. 6B, 7B), including medical records, tumor samples, peripheral blood, and buccal swabs samples. The clinical characteristics were summarized in Additional file: Table S2. Clinical data were collected from pathology reports and unprocessed medical files. The study was conducted with the permission of the Ethics Committee of the Sun Yat-sen University Cancer Institutional Board, and all patients involved provided informed written consent.

Tumor digestion
Tumors were extracted and finely minced. Tumor tissue was additionally blended with Collagenase type IV (Absin, China) for 2 h, 37°C. Tumors were extracted and re-suspension in PBS buffer containing 2% FBS for flow cytometric analysis.

Flow cytometry
Tumors were extracted and processed as described above before re-suspension in PBS buffer containing 2% FBS and PBS for flow cytometric analysis. Zombie NIR™ Fixable Viability Kit (Biolegend, USA) was applied to cells for 30 min on ice in the dark. Cells were washed and incubated with fluorochrome-conjugated antibody (anti-mouse CD45 Birliant violet 605, BioLegend; antimouse CD3 APC, BioLegend; anti-mouse CD8 FITC, BioLegend) at the manufacturer's recommended dilution for 30 min on ice in the dark. For samples requiring intracellular staining, cells were fixed with Fixation/Permeablization Diluent (eBioscience cat. 00-5223-56) for 30 min at room temperature, washed twice with Permeablization Buffer (eBioscience cat. 00-8333-56), and incubated with antibody (anti-mouse Granzyme B PE, BioLegend; anti-mouse Ki-67 Alexa Fluor 700, BioLegend) in permeabilization buffer for 30 min at room temperature in the dark. Following staining, cells were washed again with permeabilizaton buffer, subsequently washed with PBS, and re-suspended in PBS buffer for flow cytometric analysis on the CytoFLEX LX Flow Cytometer. 50,000-100,000 cells were analyzed per sample per mouse using Beckman CytExpert Software.

Next-generation sequencing
Formalin-fixed paraffin-embedded (FFPE) tumor samples and matched peripheral blood or buccal swabs samples of fourteen patients diagnosed with ENKTL were used to extract DNA and detect mutations by targeted next-generation sequencing (NGS) with a panel of the coding sequence of 102 ENKTL-relevant genes (Additional file: Table S1) (GeneseeqOne, Nanjing Geneseeq Technology Inc., China). Sequencing was performed on the Illumina HiSeq4000 platform followed by data analysis as previously described [41].
Generation and expression of STAT5A/B constructs Sanger sequencing was used to identify the STAT5A/B mutations in ENKTL cell lines, including NK-YS, SNK-6, and SNT-8. Next-generation sequencing (NGS) was used to identify the STAT5A/B mutations in ENKTL tumor samples. Wild-type STAT5A (STAT5A WT ) and wild-type STAT5B (STAT5B WT ) were amplified with Pri-meSTAR MAX DNA Polymerase (TaKaRa Bio Inc., code no. R450A) using NK-92 cell line cDNA as the template inserted into the pReceiver-M12 vector. STAT5A/B mutations were generated from STAT5A/B WT constructs and confirmed by Sanger sequencing. Eukaryotic expression vectors harboring STAT5A/B WT or STAT5A/B mutations were introduced into NK-92 cells.

Statistical analysis
Statistical analysis was carried out using IBM SPSS statistics software or GraphPad Prism using Student's t-test or one-way ANOVA or Dunnett's test. All experiments were repeated in triplicate. Data are expressed as mean ± standard deviation (SD). Statistical significance was defined as P < 0.05.

GM-CSF mediates immune evasion and results in tumor progression in ENKTL
To explore the mechanism of disease progression in vivo, we established the mouse models bearing EL4 cell tumors. The results showed that EL4 tumor-bearing mice treated with anti-PD-1 (n = 8) or the combined therapy (GM-CSF combined with PD-1 antibody, n = 8) had longer survival (Fig. 1A) and significant delay in tumor growth (Fig. 1B-D) compared with the mice in control group (n = 8) or mice treated with GM-CSF alone (n = 8). To determine the immunological effect of GM-CSF in vivo, we further examined the tumorinfiltrating lymphocytes (TILs) and its activation marker (granzyme B) in tumor tissues derived from mice. Tumors from mice treated with GM-CSF alone have significantly lower levels of CD3, CD8, and granzyme B than tumors from mice treated with IgG control, and tumors from mice received anti-PD-1 therapy or combined therapy have significantly higher levels of CD3 ( Fig. 2A, C-E). These results demonstrated that GM-CSF decreased TILs in vivo, which could be reversed by anti-PD-1 therapy. Furthermore, tumors from mice treated with combined therapy have significantly higher granzyme B levels ( Fig. 2A, E), indicated that anti-PD-1 therapy combined with GM-CSF might promote the antitumor effect of TILs.
We have also established the melanoma mouse model bearing B16-F10 tumors. The results showed that mice bearing B16-F10 tumors treated with combined regimen had longer survival than other groups (( Fig. 1E-H), and tumor samples of mice treated with PD-1 antibody or combined therapy had significantly higher levels of CD3, CD8, and granzyme B than IgG group (Fig. 2F-H). Tumor samples of mice treated by GM-CSF had significantly higher level of granzyme B than IgG group (Fig. 2H).
GM-CSF upregulates PD-L1 expression in ENKTL via JAK2/ STAT5 pathway To determine the molecular mechanism of GM-CSFinduced disease progression in ENKTL, NK-YS, SNK-6, and SNT-8 cells were treated with 100 ng/ml or 500 ng/ ml recombinant human GM-CSF in 0, 24, 48 and 72 h [42,43]. The cell viability had no statistic differentiation between control and GM-CSF treated groups among all three ENKTL cell lines (Fig. 3A-C), indicating that GM-CSF didn't affect the proliferation of ENKTL directly. But the mRNA and protein expression levels of PD-L1 strongly increased after GM-CSF (100 ng/ml and 500 ng/ ml) treatment in ENKTL cells (Fig. 3D, K). We found that the expressions of p-JAK2, p-STAT5, and PD-L1 were upregulated gradually in the treatment of GM-CSF (100 ng/ml) in a time-dependent manner (Fig. 3E-G), showing that GM-CSF regulated PD-L1 expression in ENKTL via JAK2/STAT5 pathway (Fig. 3H-J). The protein expression levels of PD-L1 in Jurkat, SU-DHL-6, and A-375 cells have no significant change after treated with GM-CSF (Fig. 3L).
To further confirm the role of JAK2/STAT5 in regulating PD-L1 expression in ENKTL cells, we used the JAK2 inhibitor fedratinib (3 nM) to treat the NK-YS, SNK-6, and SNT-8 cells. The protein levels of p-JAK2, p-STAT5, and PD-L1 were downregulated in these three cells, respectively (Fig. 4A-F). Moreover, the increase of PD-L1 triggered by GM-CSF was impaired with the treatment of fedratinib (Fig. 4A-F), which further proved that GM-CSF regulated the PD-L1 expression in ENKTL via JAK2/STAT5 pathway.

JAK2 hyperphosphorylation is correlated with poor prognosis in ENKTL patients treated with GM-CSF
According to clinical observation, some of the patients with ENKTL developed disease progression after receiving GM-CSF therapy for hematopoietic stem cell And relative mRNA expression of PD-L1 was measured by quantitative polymerase chain reaction (qPCR). g-j SNK-6 and SNT-8 cells expressing shSTAT5 or control were evaluated for STAT5 and PD-L1 protein expression by Western blot, and mRNA expression by qPCR. k The − 620 to − 500 nucleotide sequence of the 5′-flanking region of PD-L1 is shown. Underlined sequences are putative STAT5A and STAT5B transcription factor binding sites, as predicted by JASPAR database and PROMO. And PD-L1 promoter fragments cloned into pGL3-Basic vector. l Analysis of PD-L1 promoter fragment A constructs in 293 T cells transiently transfected with STAT5A or STAT5B for 48 h. Relative luciferase activity was determined as described. Error bars represent the SD of three independent experiments. * P < 0.05, ** P < 0.01, *** P < 0.001. Error bars represent the SD of three independent experiments mobilization. A total of 117 ENKTL patients were diagnosed in Sun Yat-sen University Cancer Center from September 1999 to April 2017, including 32 patients who have received GM-CSF treatment (1 patient was diagnosed in September 1999, and 31 patients were diagnosed from June 2006 to April 2017), were included in our research. The clinical characteristics were summarized in Additional file: Table S2 b Imaging data of a patient who had disease progression after treated by GM-CSF. The patient has received good partial remission after chemotherapy and was treated by GM-CSF for hematopoietic stem cell mobilization, who relapsed within 1 month. c Test results of serum EBV-DNA of five ENKTL patients who have been treated by GM-CSF. Patients who were treated by GM-CSF could induce elevated serum EBV-DNA copies, which could be reduced after chemotherapy of the markers that can monitor the condition of ENKT L [44]. From the 117 ENKTL patients, we have found 5 patients who have comparable imaging data and/or copies of plasma EBV-DNA results during the time of GM-CSF treatment from a retrospective study (Fig. 5B-C). The copies of plasma EBV-DNA of the 5 patients elevated after GM-CSF treatment and decreased after chemotherapy (Fig. 5C).
To determine the role of JAK2/STAT5/PD-L1 pathway in the prognosis of ENKTL patients treated by GM-CSF, we used immunostaining to detect the levels of these three proteins in 21 PTCL patients who received GM-CSF treatment from January 2011 to September 2015, including 14 ENKTL patients, 4 AITL patients and 4 PTCL-NOS patients (Fig. 6A-B). We investigated the correlation between the expression of p-JAK2, p-STAT5, and PD-L1 in ENKTL, and found that p-JAK2 had a clear positive correlation with p-STAT5 levels (p = 0.001, r = 0.801, Spearman rank correlation coefficient; Fig. 6C). Meanwhile, p-STAT5 had a clear positive correlation with PD-L1 levels (p = 0.037, r = 0.398, Spearman rank correlation coefficient (Fig. 6D). Disease progression that developed within a week after GM-CSF treatment was defined as short-term progression. We found the tumor samples from patients with short-term progression, which account for 28.6% (6/21) of all patients, expressed a higher level of p-JAK2 (p = 0.019) (Fig. 6E).
We calculated the optimal cutoff point according to ROC curves by comparing the sensitivity and specificity of progression-free survival (PFS) prediction and overall survival (OS). The cutoff expression value of PD-L1 was 5.0 for OS, and the cutoff expression value of p-JAK2 was 5.0 for PFS. Kaplan-Meier curves and log-rank tests were performed. We observed that Patients with low p-JAK2 levels had a longer PFS than those with high p-JAK2 levels (median PFS 25.3 days vs. 8.4 days; P = 0.03) (Fig. 6F). Patients with low PD-L1 levels had longer OS than those with high PD-L1 levels (median OS 46.2 days vs. 24.5 days; P = 0.005) (Fig. 6G). These data show that high expression of PD-L1 and p-JAK2 both predict poor prognosis of GM-CSF treatment. These data show that high expression of p-JAK2 and PD-L1 both predict inferior prognosis of GM-CSF treatment.
To explore the functional implications of these mutations, we generated expression constructs for wild-type, p.D475N, p.F706fs, p.V707E, p.S710F, and p.R131G variants of the STAT5A protein, as well as wild-type, p.T628S, p.S434L, and p.R423Q variants of the STAT5B protein. When expressed in NK-92 cells, increased autophosphorylation of STAT5A and PD-L1 has been observed in STAT5A mutations (p.D475N and p.V706fs) (Fig. 7D). NK-92 cells harboring p.T628S, p.S434L and p.R423Q vectors have no significant change in autophosphorylation of STAT5B and PD-L1 compared to cells harboring STAT5B WT vector (Fig. 7E). NK-92 cells harboring p.V706fs (c.2118_2119delTG) vector had more aggressive growth of p-STAT5A and PD-L1 than cells harboring STAT5A WT vectors after GM-CSF treatment (100 ng/ml) (Fig. 7F). The overview of the mechanism for GM-CSF-induced disease progression in ENKTL was shown in Fig. 8A.

Discussion
GM-CSF was regarded as an immunopotentiator in recent years [47][48][49], and showed significant and curative effects on several types of tumors when combined with immune checkpoint inhibitors [50,51], which has an opposite therapeutic effect in ENKTL. Previous studies found that recurrent numerical aberrations of JAK2 in ENKTL, which might be one of the main causes of disease progression induced by GM-CSF [24]. According to our results, the activation We proved that GM-CSF did not induce the proliferation of ENKTL directly and used different approaches to The PD-L1, p-STAT5, and total STAT5 protein levels in these cells were detected with Western blot. e NK-92 cells were transduced with empty vector, STAT5B WT , p. R423Q, p.T628S, p.S434L expression vectors. The PD-L1, p-STAT5, and total STAT5 protein levels in these cells were detected with Western blot. f NK-92 cells which were transduced with empty vector, STAT5A WT , p.R131G, p.D475N, p.F706fs, p.V707E, and p.S710F expression vectors were treated with GM-CSF (100 ng/ml) for 12 h. The PD-L1, p-STAT5, and total STAT5 protein levels in these cells were detected by Western blot. g STAT5A/B mutation rate of 144 melanoma patients reported by David Liu et al. [38]. h Scatter plots represent the CD274 mRNA of 144 melanoma patients (STAT5A mutant patients vs. STAT5A wildtype patients). I Scatter plots represents the CD274 mRNA of 144 melanoma patients (STAT5B mutant patients vs. STAT5B wildtype patients) demonstrate that GM-CSF could upregulate the PD-L1 expression in ENKTL. Furthermore, we found that GM-CSF could induce disease progression in vivo. Mouse treated with GM-CSF have decreased TILs (lower levels of CD3+ and CD8+ T cells than mouse treated with IgG control) and granzyme B from cytotoxic T cells, which could be reversed by anti-PD-1 therapy. The different phenomenon could be seen in melanoma cell line A-375 and B16-F10, which was proved to be a kind of GM-CSF sensitive tumor [50]. We found the number of TILs were decreased in the EL-4 cell bearing tumor mouse treated by GM-CSF, but did not observe the alteration of TIL number in the B16-F10 cell bearing tumor mouse treated by GM-CSF. JAK2 and STAT5 are highly expressed in EL-4 cells but are low expressed in B16-F10 cells. Furthermore, compared with the B16-F10 cells, the expression of PD-L1 was increased in EL-4 cells treated by GM-CSF. Previous researches have proved that high expression of PD-L1 could induce immunosuppression in vivo [52,53]. These results suggest the decrease of TILs induced by GM-CSF is associated with the expression levels of Jak2 and STAT5 in cancer cells. A previous study has reported that GM-CSF could induce resistance to imatinib and nilotinib in chronic myeloid leukemia (CML) via the activation of JAK2/STAT5, but didn't mention the Immunological role of JAK2/STAT5 pathway in anti-tumor therapy [54]. GM-CSF-induced PD-L1 expression of neutrophil infiltration in tumors, which could induce immunosuppression in gastric cancer (GC) [21]. We demonstrated that GM-CSF could induce PD-L1 expression in ENKTL cells. But the mechanism of disease progression induced by GM-CSF in ENKTL patients still needs to be further explored due to the lack of comparable tumor samples of patients after GM-CSF treatment and limitation of the small sample size.
A recent study has reported that alterations in JAK/ STAT pathway is highly prevalent in ENKTL, and STAT3 was one of the most frequently mutated in exons of 188 JAK/STAT pathway-related genes. They proved that the frequent mutations of STAT3 might be one of the reasons for high PD-L1 expression in ENKTL [55]. We used NGS with a different range of panels that targeted a total of 102 ENKTL-relevant genes. We also found that STAT5A/B mutations are highly prevalent in ENKTL, and STAT5A was one of the most frequently mutated genes in ENKTL patients. Early Research has reported that STAT5 mutations are rare in hematopoietic diseases [56], and remained few studies about STAT5 mutations in lymphoma so far. Therefore, exploration of PD-L1 expression regulated by STAT5 mutations, especially in ENKTL, is still meaningful in the future. Cells harboring p.D475N and p.V706fs (c.2118_2119delTG) vectors had increased auto-phosphorylation of STAT5A and PD-L1. More importantly, cells harboring p.V706fs (c.2118_2119delTG)

Availability of data and materials
The data generated and analyzed will be made from the corresponding author on reasonable request. The authenticity of this article has been validated by uploading the key raw data onto the Research Data Deposit public platform with an RDD number of RDDB2021000972 (www. researchdata.org.cn).

Declarations
Ethics approval and consent to participate All animal studies were performed with the permission of the institutional committee of Sun Yat-sen University Cancer Center, in compliance with protocols approved by the Guangdong Provincial Animal Care and Use Committee and experimental guidelines of the Animal Experimentation Ethics Committee of Sun Yat-sen University Cancer Center (L102042018120I and L102042020120N). For patient specimens, all patients were consented and enrolled to a Sun Yat-sen University Cancer Center IRB approved protocol, in accordance with ethical guidelines, allowing the collection and analysis of clinical data, archival, and paraffin specimens (B2021-110-01). This study conforms to the Declaration of Helsinki.

Consent for publication
Not required.