PD‐1/PD‐L1 blockade rescue exhausted CD8+ T cells in gastrointestinal stromal tumours via the PI3K/Akt/mTOR signalling pathway

Abstract Objectives Although targeted therapy has revolutionized the treatment of gastrointestinal stromal tumours (GIST), it is almost never curative in GIST, and resistance commonly develops. One potential strategy is to combine targeted therapy with immunotherapy. Materials and methods We first studied Programmed cell death 1 ligand 1 (PD‐L1) expression and tumour‐infiltrating T cells (TILs) in GIST. IFN‐γ was used to induce the upregulation of PD‐L1 expression in GIST‐882 cells, a well‐known GIST cell line. CD8+ T‐cell apoptosis was determined by flow cytometry. The PI3K/Akt/mTOR levels in CD8+ T cells were examined by Western blotting. Results PD‐L1 expression was an independent factor of poor prognosis in GIST and resulted in exhausted T cells in the TILs population or the blood. Then, we found that PD‐L1 blockade alone could not increase tumour cell apoptosis in GIST. The apoptosis rate of CD8+ T cells was higher when T cells were cultured with PD‐L1+ GIST‐882 cells (GIST‐882 cells with high PD‐L1 expression) than when T cells were cultured with control GIST‐882 cells. However, when the PD‐L1 blockade was used, the apoptosis rates of the CD8+ T cells in the two groups became similar. Then, Western blotting showed the PI3K/Akt/mTOR levels of the CD8+ T cells rescued by the PD‐1/PD‐L1 blockade were higher than those of the CD8+ T cells not treated with the PD‐1/PD‐L1 blockade. Conclusions PD‐L1 expression was an independent poor prognosis factor in GIST. PD‐1/PD‐L1 blockade rescued exhausted CD8+ T cells in GIST via the PI3K/Akt/mTOR signalling pathway. In GIST, PD‐1/PD‐L1 not only function as predictive biomarkers but also improve current therapies as treatment targets.


| INTRODUC TI ON
Gastrointestinal stromal tumour (GIST) is the most common gastrointestinal soft tissue malignancy. 1 Approximately 85% of GISTs contain an activating mutation in the KIT proto-oncogene, whereas 5%-10% have a mutation in the gene encoding PDGFRA. 1,2 Targeted molecular therapy has revolutionized the treatment of GIST and significantly improved the prognosis of GIST patients.
In 2001, imatinib was first shown to treat metastatic gastrointestinal stromal tumours effectively, and it has improved the median overall survival of GIST patients from 9 months to over 5 years. [3][4][5][6][7] However, imatinib is almost never curative in GIST, and resistance commonly develops at a median time of 18 months, mostly due to a secondary KIT or PDGFRA mutation. 6,8 Even though sunitinib and other new targeted drugs can sometimes be effective in recurrent GIST, clinical progression and drug resistance, such as insensitivity to sunitinib, subsequently evolve within 1 year. 9,10 Another potential strategy to increase the efficacy of imatinib is to combine imatinib with immunotherapy.
Many studies have confirmed that T cells, especially CD8+ T cells, a crucial component of the cellular immune response, are critical for the anti-tumour effects of imatinib in GIST. T cells not only control a variety of bacterial and viral infections but also represent a major arm of the cell-mediated anti-tumour immune response. 11 CD8+ T cells have been shown to play an important role in host defence and exhibit cytotoxicity against malignancies. 12,13 However, in cancer, CD8+ T cells upregulate the expression of inhibitory receptors, resulting in dysfunction and apoptosis in CD8+ T cells, which are then described as exhausted CD8+ T cells. [15][16][17][18] This process of exhaustion results in insufficient numbers of CD8+ T cells capable of killing tumour cells and leads to rapid tumour progression, including proliferation, invasion and metastasis. 19 Programmed cell death protein 1 (PD-1) has been shown to be expressed on exhausted T cells and to be a major mechanism of immune escape that malignancies take advantage of to evade destruction. 20,21 PD-1 is a 288 amino acid protein that is expressed in activated mature T cells to regulate the balance between activating and inhibitory signals. 22 Programmed cell death 1 ligand 1 (PD-L1), the main ligand for Programmed cell death 1 ligand 1 (PD-L1), is expressed on tumours and can lead to impaired T-cell proliferation and effector functions, leading to apoptosis of tumour-specific T cells. 22,23 In multiple solid malignancies, PD-L1 is typically expressed on the surface of the tumour cells and appears to be upregulated, which helps tumour cells evade the cytotoxicity of T cells. 24,25 Thus, PD-1/PD-L1-targeted therapies can enhance T-cell responses and play a critical role in rescuing exhausted T cells by regulating costimulatory molecules. 26,27 A better understanding of the mechanisms of T-cell exhaustion can provide novel therapeutic targets for the treatment of different tumours. Here, we have known that the PD-1/PD-L1 axis is a critical pathway leading to T-cell exhaustion, with the expression of PD-1 on CD8+ T cells correlating with a severely exhausted T-cell response. 28 However, the understanding of PD-1/PD-L1 therapies is still limited in GIST. 29,30 Overall, CD8+ T-cell exhaustion mechanisms regulated by PD-1/PD-L1 in GIST remain largely undefined. In our study, we analysed the expression of PD-L1 associated with tumour-infiltrating T cells (TILs) and tumour biological characteristics in GIST. The frequency and functional characteristics of exhausted CD8+ T cells, which were identified based on their PD-1 expression, were evaluated. To determine the effects of the PD-1/PD-L1 axis on CD8+ T cells in GIST, the correlation of exhausted CD8+ T cells with the expression of PD-L1 was also addressed. Furthermore, we tested the combination of imatinib with PD-1/PD-L1 blockade on GIST cells and CD8+ T cells in vitro.

| Quantitative real-time RT-PCR was used to detect the expression of PD-L1 mRNA
The RNA samples were quantified using an ultraviolet spectrophotom-

| Immunofluorescent staining analysis for the quantification of TILs
Immunohistochemistry (IHC) staining for the PD-L1, CD4+ and CD8+ proteins was performed using the streptavidin-peroxidase method (SP method). The paraffin-fixed slides were dewaxed in xylene and rehydrated with 95% alcohol. PBS containing 10% normal goat non-immune serum was used to block the sections for 1 hour at room temperature. Then, the sections were incubated with the primary antibodies, including a rabbit anti-PD-L1 polyclonal antibody (1:200, ab153991; Abcam, Cambridge, UK), rabbit anti-human CD4 antibody (1:50) and mouse anti-human CD8 antibody (1:50), at 25°C for 2 hours. After that, they were treated with 0.2% Triton X-100 and incubated at 25°C for 1 hour. Then, they were incubated with a secondary antibody (Alexa Fluor 488-labelled goat anti-rabbit IgG, Alexa Fluor 647-labelled goat anti-mouse IgG, or Cy3-labelled goat anti-rat IgG) at 25°C for 1 hour. The slides were counterstained with haematoxylin and mounted under coverslips. The staining was quantified by manual counting and imagej software.

| Cell lines
The GIST-882 cell line (KIT exon 13 mutant; provided by Jonathan

| Flow cytometry and cytokine detection
Human-specific antibodies were purchased from Thermo Fisher A 488 nm excitation wavelength was selected, FITC fluorescence was detected by a 515 nm wavelength filter, and PI was detected by a 560 nm wavelength filter. The results were analysed with flowjo software (version 7.6) (Ashland, OR, USA).

| Western blot and cytokine array
Cells were harvested and suspended in RIPA lysis buffer containing 1 mmol/L phenylmethylsulfonyl fluoride (PMSF). Antibodies for p-AKT (PS473), p-PI3Kp85 (PY607), p-s6 (Ser240/244)) and βactin were purchased from Cell Signaling Technology. Total protein was isolated from frozen tumour tissue samples and tested for cytokine/chemokine expression using a Proteome Profiler Array (R&D Systems). Densitometry was conducted on blots using imagej. Signal quantitation was calculated using Quantity One software (Bio-Rad Laboratories, Hercules, CA, USA), and signals were normalized to the β-actin signal.

| Ethics statement
The study protocol was approved by the ethics committee of West China Hospital, Sichuan University. Written informed consent was obtained from the patients before beginning the study. Patient records/information were anonymized and deidentified prior to analysis, and the methods were adopted in accordance with the approved guidelines.

| Statistical analyses
All statistical analyses were performed using spss version 20.0 (for Window; IBM, AMONG, NY, USA). We used the chi-square test to compare categorical data and a t test or ANOVA to compare continuous data. The experimental data are expressed as the mean ± standard deviation (SD). The survival data were compared using the Kaplan-Meier method and the log-rank test to detect differences in the survival curves of the various groups. All P values were twotailed, and P values <0.05 were considered significant.

| PD-L1 expression in human GIST and tumour biological characteristics
To assess the expression levels of PD-L1 associated with tumour biological characteristics in human GIST, we evaluated 127 human specimens from patients who underwent surgeries for GIST in our hospital between January 2013 and December 2015.
Of the 127 studied patients, 89 patients (70%) were classified as high risk, and 14 patients (11%) were intermediate risk. Only nine patients (7%) were diagnosed as low/very low risk. There were 15 patients (12%) who had recurrent GIST. Among them, 11 patients had received adjuvant imatinib over 1 year after the first surgery, and four patients did not undergo adjuvant imatinib.
To identify PD-L1 expression that was relevant in GIST, we performed quantitative real-time PCR (qPCR) on freshly isolated tumour samples, which was a more accurate means of measuring the expression of PD-L1 within the tumours than immunohistochemistry ( Figure 1A). 31,32 The expression levels of PD-L1 were variable and closely related to the modified NIH risk classification ( Figure 1A was similar to that of the 15 patients who relapsed before surgery and higher than that of the high-risk GIST patients with low PD-L1 expression, which showed that PD-L1 expression affected clinical outcome in high-risk GIST patients ( Figure 1E). The rate of 5-year relapse-free survival (RFS) for patients with low PD-L1 expression was 94.44% compared to 56.25% for patients with high PD-L1 expression, and this difference achieved statistical significance (χ 2 = 7.28, P = 0.007, Figure 1F). The 5-year RFS in the intermediate PD-L1 expression group was also higher than that of the high PD-L1 expression group (83.33% vs 56.25%, respectively, χ 2 = 5.83, P = 0.016, Figure 1F). However, the difference between the low PD-L1 expression and intermediate PD-L1 expression groups was non-significant (χ 2 = 1.45, P = 0.23, Figure 1F). In a univariate analysis, PD-L1 expression, tumour size, tumour mitotic index and risk classification were associated with the risk of relapse, whereas patient age, anatomical site and tumour mutations were not. In a multivariate analysis, PD-L1 expression remained significant (P = 0.002), similar to the results of a previous study. 33 The results illustrated that PD-L1 expression was an independent poor prognosis factor in GIST.

| PD-L1 expression on tumour-infiltrating/blood T cells in GIST
To investigate the response of tumour-infiltrating/blood T cells to PD-L1 on the surface of GIST tumours, we performed IHC and flow cytometry to identify tumour-infiltrating/blood T cells from 127 GIST samples (Figure 2A). The median PD-L1 expression was used as a cut-off value. The percentages of CD4+ and CD8+ T cells were strongly and inversely correlated with PD-L1 expression in human GIST (F = 9.90, P = 0.0021, F = 5.25, P = 0.024, Figure 2B,C).  Figure 2D). 34 Interestingly, the percentage of CD8+ T cells in GISTs treated with imatinib was slightly higher than that in untreated GISTs, illustrating that imatinib might induce a dramatic increase in the number of CD8+ T cells (P = 0.041, Figure 2E). Moreover, in the blood, the percentages of CD4+ or CD8+ T cells were also inversely associated with the PD-L1 expression level, which is the same as what was observed in the tumours.
These results suggest that exhausted T cells exist in GIST and are related to the PD-L1 expression of the tumour.

| PD-L1 blockade in GIST-882 cells and IFN-γ-induced upregulation of PD-L1 expression on GIST-882 cells
Previous studies have reported that IFN-γ can induce PD-L1 expression in GIST. 35  flow cytometry, there was no obvious difference in the rates of apoptosis of the groups with or without PD-L1 blockade ( Figure 3D,E).
The PD-L1 blockade alone could not increase tumour cell apoptosis due to the high PD-L1 expression in human GIST cells in vitro, and the same result was obtained in a clinical trial. 30

| PD-L1 blockade for exhausted CD8+ T cells cultured with GIST-882 cells in vitro
To explore the effect of PD-L1 blockade on CD8+ T cells cultured  Figure 4A). Then, we tested the CD8+ T cells from the different groups. Western blotting confirmed that the p-PI3K, p-Akt and p-6s proteins were expressed at higher levels in Group D than in Group B, but there was no significant difference between Group A and Group C, even though the levels in Group C were slightly higher ( Figure 4B). The cell proliferation-related signal transduction pathway PI3K/ Akt/mTOR is involved in the regulation of a variety of cell proliferation and apoptosis functions; thus, aberrantly elevated mTOR activity is frequently observed in human malignancies. 52 However, mTOR signalling also plays a crucial role in both the innate and adaptive immune systems. Activated by the phosphorylation of PI3K/ Akt/mTOR, mTOR can affect the cell cycle and regulate apoptosis by mediating the important downstream signal, which is of great significance in the exploration of the signalling pathway. 53 Notably, AKTindependent metabolic responses have been identified in CD8+ T cells. 54,55 Previous studies have shown that the activation of the PI3K/Akt/mTOR pathway can increase nutrient uptake and energy production in CD8+ T cells. 56,57 Increasing evidence suggests that mTOR plays a central role in regulating the biological outcomes of immune cell stimuli. [58][59][60] In our study, we found that the PD-1/PD-L1 blockade could decrease apoptosis in the CD8+ T cells in vitro via the PI3K/Akt/mTOR signalling pathway, which was first reported in GIST. This mechanism in GIST could help us to discover more accurate therapies and new combination regimens to combat treatment side effects and resistant GIST, which constantly interfere with the treatment of GIST. Immunotherapy using PD-1/PD-L1 blockade is now widely used for the treatment of many malignant tumours, but only a subset of patients respond to the therapy. 39,40,61 The mechanism by which PD-1/PD-L1 regulates CD8+ T cells found in our study may resolve the problem of non-responding patients and allow the combined treatment to achieve effective clinical effects in GIST as soon as possible. However, our results are limited due to a lack of data from pre-clinical mouse models. In the future, we will establish a mouse model to explore the PI3K/Akt/mTOR pathway in vivo.

| D ISCUSS I ON
Taken together, our study elucidated that PD-1/PD-L1 block-

ACK N OWLED G EM ENTS
The authors thank the nursing and support staff at West China Hospital, Sichuan University. The study was supported by the two Natural Science Foundations of China (0040205301919 and 0040205401430) and the Sichuan Provincial Science and Technology Support Project (2016SZ0047).

CO N FLI C T O F I NTE R E S T
All the authors declare that they have no relevant conflicts of interest to disclose.