Introduction

Pancreatic ductal adenocarcinoma (PDAC) is one of the most deadly cancer diseases with a high death rate worldwide [1], and the overall 5-year survival for PDAC patients was barely improved in the past decades [2]. The majority of PDAC patients after surgery suffer from recurrence and metastasis, which lacks efficient chemotherapy, radiotherapy, and immunotherapy [3]. The relatively low responses of immunotherapy on pancreatic cancer may be attributed to the exhaustion of T cells, which is induced by the complex immunosuppressive tumor microenvironment (TME) [4, 5]. Many studies focused on the interactions between the tumor cell and lymphocytes, and the potential mechanism beneath the immunodysfunction. Recent studies have shown that guanylate-binding proteins (GBPs) are a family of highly conserved GTPases, may play critical roles in manipulating the cross-talking between tumor cells and immune cells, and remodeling the antitumor immune responses [6,7,8,9].

GBPs are typical interferon-stimulated genes (ISGs), which are originally known to participate in the defense against intracellular pathogens [10, 11]. There are currently seven GBP family members in humans, including GBP1, GBP2, GBP3, GBP4, GBP5, GBP6, and GBP7 [12]. The expression level of GBPs increases as a response to interferon's (IFN's) stimulation to inhibit the proliferation of infected cells and to protect normal cells from inflammation-induced apoptosis [13]. However, more recently, GBPs have been identified on certain cancer cells and its expression influences their behavior. GBPs appear to display context-dependent effects in cancer. They displayed antitumor effect in several cancers such as colorectal cancer, skin cutaneous melanoma, and breast cancer [14,15,16], but they were strongly associated with tumor progression in pancreatic cancer, prostate cancer, ovarian cancer, and glioblastoma [17,18,19,20]. However, the expression patterns and functions of those GBP family members in PDAC were largely unknown, and their diagnostic value demands further investigation. Besides, GBPs were also indicated to be extensively regulated by epigenetic modification in several contexts [21, 22], and the exact patterns and effects of GBP epigenetics in pancreatic cancer are poorly understood.

Interestingly, several GBPs were also reported to be associated with immune cell infiltration and upregulation of immune checkpoints in cancer tissues [23, 24]. Previous study identified the positive correlations between GBP1-5 and the abundance of CD8+ T cells in hepatocellular carcinoma tissues [7]. GBP2 regulates tumor immune infiltration and promotes tumor immune escape through PD-L1 expression via STAT1 pathway [8]. Knockdown of GBP5 significantly inhibited the activity of IFN-γ/STAT1 and TNF-α/NF-κB signaling axes, and thereby repressed the mRNA expression of PD-L1 in triple-negative breast cancer cells [9]. However, the role of GBP4 in cancers, particularly in PDAC, was largely unknown, and its effects on T cell-tumor communication and T cell exhaustion are of potentially interesting.

Our study revealed that GBP4 was one of the most dominant GBP family members in PDAC, and the expression patterns and functions of GBP4 in PDAC were investigated in this manuscript. Here, we first assessed the impact of GBP4 expression on survival of PDAC patients. Then, we revealed the epigenetic regulation of GBP4 and its effect on GBP4 expression via the bisulfite genomic sequencing and the dCas9-SunTag-DNMAT3A-sgRNA-targeted methylation system, which was applied by several studies to achieved success on specific methylation of target genes [25, 26]. After that, we performed functional investigation and found that GBP4 expression enhanced the infiltration of T cells and the expression of immune checkpoints. In vitro studies showed that the high expression of GBP4 may be in relevant to a better response of anti-PD-1 treatment. Taken together, we revealed that, for the first time, the epigenetic regulatory mechanisms of GBP4 in pancreatic cancer and clarified the effects of GBP4 on antitumor immunology and T cell exhaustion in PDAC microenvironment. Our results may translate to the molecularly targeted therapy and immunotherapy for PDAC.

Materials and methods

Bioinformatics analyses

The transcripts per million (TPM) normalized RNA-sequencing (RNA-seq) data from the Cancer Genome Atlas Program (TCGA) and the Genotype-Tissue Expression (GTEx) were used to study mRNA levels of genes in PDAC samples (n = 179) and corresponding normal tissues (n = 171), respectively. Clinical proteomic tumor analysis consortium database (CPTAC) [27] was used to analyze the protein levels of genes in primary pancreatic cancer tissues (n = 137) and the corresponding normal tissues (n = 74). The tumor samples were divided into high expression group and low expression group according to the median expression levels of genes, then the survival of patients with PDAC were analyzed. DNA methylation level of GBP4 in TCGA project was analyzed using the Shiny Methylation Analysis Resource Tool (SMART) [28]. The abundances of 22 tumor-infiltrated immune cells in the PDAC samples from TCGA were estimated via the CIBERSORTx tool [29]. The gene signature scores of the progenitor/stem-like exhausted T cell and the terminally differentiated exhausted T cell in the TCGA cohort were calculated by single-sample gene set enrichment analysis (ssGSEA) method using the gene sets defined in a previous study (Table S1) [30].

Patients and samples

All surgical-resected PDAC specimens, including paraffin-embedded tissues and fresh tissues, and peripheral blood samples were obtained from Fudan University Shanghai Cancer Center and Shanghai Pancreatic Institute from 2022 to 2023. The study was approved by the ethics committee of Fudan University Shanghai Cancer Center and all subjects signed an informed consent to participate in this study. The studies were conducted in accordance with Helsinki Declaration.

Cell lines and culture conditions

The human pancreatic cancer cell lines (Capan-1, Mia PaCa-1, CFPAC-1, AsPC-1) and the immortalized normal pancreatic ductal epithelial cells (HPDE6-C7) were obtained from ATCC. Briefly, all cells used in this study were cultured in DMEM (Gibco, USA) containing 10% fetal bovine serum (Gibco, USA), 1% penicillin and streptomycin (Gibco, USA), and cultured in incubator at 37 °C temperature containing 5% CO2.

Tissue microarray (TMA) generation

A total of 111 primary pancreatic cancer patients who had received surgical resection with curative intent between 2015 and 2020 at the Department of Pancreatic Surgery, Fudan University Shanghai Cancer Center were selected for this study. The clinicopathological information, including age, gender, tumor size, histological grade, lymph node (negative or positive), and clinical stage (defined according to American Joint Committee on Cancer 7th edition) and follow-up information (5–10 years), were collected. This study was approved by the ethics committee of the Fudan University Shanghai Cancer Center, and all experiments were carried out in accordance with approved guidelines of Fudan University Shanghai Cancer Center. Written informed consent was signed by all of the participants. The tissues with hematoxylin–eosin staining were reviewed histologically by at least two pathologists. After re-identification of representative tumor and non-tumor areas, discs with a diameter of 1.5 mm were taken from each representative tumor tissues (tumor tissues and adjacent normal tissues) using a Tissue Arrayer and re-embedding them into a new paraffin block. Two TMAs were used in present study, one composed of 52 PDAC samples and 52 corresponding adjacent normal tissues, the other composed of 59 PDAC samples.

Immunohistochemical and fluorescent multiplex immunohistochemical staining

Immunohistochemical staining was used to examine the expression level of GBP4 protein in PDAC tissues and normal pancreatic tissues. TMA slides were subjected to the standard procedures including sample deparaffinizing and hydration, endogenous enzyme removal and antigen retrieval, blocking, antibody incubation (anti-human GBP4 antibody), and DAB staining. Three visual fields from different areas of each specimen were chosen randomly for the IHC evaluation. GBP4 expression was scored according to staining intensity and the percentage of positive cells as previously described. The percentage of positive cells was scored as follows: 0–5% (0), 5–25% (1), 25–50% (2), 50–75% (3), and 75–100% (4). The expression levels were classified as follows: negative (0–3, −); weakly positive (4, +); moderately positive (6, + +); and strongly positive (> 6, +  + +).

Fluorescent multiplex immunohistochemical (fluorescent mIHC) staining was used to evaluate the localization and abundance of T cell subsets in PDAC tissues. After deparaffinizing, hydration, several rounds of staining consisting of antigen retrieval, blocking, primary antibody (anti-human CD8 antibody, anti-human TCF7 antibody, and anti-human PD-1 antibody), HRP-conjugated goat secondary antibody, and tyramide signal amplification (TSA) steps were performed. The Opal™ 7-Color IHC Kit (NEL811001KT, Akoya Biosciences, USA) TSA dyes (Opal 520, Opal 620, and Opal 690) were used to visualize immunofluorescent signaling. The nuclear was counterstained by DAPI. Imaging of slides was performed on a Vectra Polaris Automated Quantitative Pathology Imaging System (PerkinElmer, Boston, Massachusetts, USA). More than three different high-power fields (HPFs) with the most active signals from a slide were analyzed. From these areas, TCF7+, PD-1+, TCF1+PD-1+ TILs (i.e., TILs expressing both TCF1 and PD-1), and TCF1PD-1+ TILs (i.e., TILs expressing PD-1 but not TCF1) were counted; the cell counts are presented as mean numbers per HPF.

The following primary antibodies were used: anti-human GBP4 antibody (TA368361S, ORIGENE, USA), anti-human CD8 antibody (ab178089, Abcam, USA), anti-human TCF7 antibody (14,464-1-AP, Proteintech, China), and anti-human PD-1 antibody (ab52587, Abcam, USA). The following secondary antibody was used: HRP Ms + Rb (ARH1001EA, Perkin Elmer, USA).

Bisulfite conversion

TIANGEN Genomic DNA Kit (DP304-02, TIANGEN, China) was used to extract genomic DNA from the collected pancreatic tumors tissues and adjacent normal tissues, and the SynergyH4 Hybrid Reader (Biotek) was used to determine the DNA concentration and purity. Subsequently, DNA bisulfite conversion kit (DP215-02, TIANGEN, China) was used for bisulfite conversion of extracted genomic DNA, according to the manufacturer’s protocol. Briefly, 1 µg of genome DNA was used for each reaction in 20 µL volume and mixed with 90 µL of bisulfite mixture and 10 µL of DNA protect buffer. Bisulfite conversion was performed using the following thermal profile: 95 °C for 10 min, 64 °C for 60 min. After conversion, the reaction mix was added with 600 μl of buffer PB and transferred to the spin column (place the column in a collection tube), followed by centrifugation at 12,000 rpm (~ 13,400 × g) for 30–60 s. Subsequently, the spin column was added with buffer PW (600 μL) and centrifuged at 12000 g for 30 s. Add 600 μl buffer DB to the spin column, incubate at room temperature (15–30 °C) for 15 min, centrifuge at 12,000 rpm (~ 13,400 × g) for 30–60 s. Then, the spin column was washed twice with buffer PW. Afterward, the spin column was placed in a new 1.5 mL collection tube and added with 20 μL of buffer EB to dissolve DNA, followed by centrifugation at 12000 g for 2 min.

Bisulfite genomic sequencing (BGS)

We used MethPrimer (http://www.urogene.org/cgi-bin/methprimer/methprimer.cgi) [31] to design bisulfite PCR primers that are used to amplify sequences from bisulfite-converted DNA. Bisulfite PCR was performed using methylation-specific PCR kit (EM101, TIANGEN, China). Each PCR reaction components include 1.6 µl dNTPs (2.5 mM), 1 µl primer forward (10 µM), 1 µl primer reverse (10 µM), 2 µl 10 × MSP PCR buffer, 1U MSP DNA Polymerase, ddH2O, and bisulfite-converted DNA. The PCR reaction was carried out according to the following conditions: 5 min denaturation at 95 °C was followed by 40 cycles of denaturation at 95 °C for 30 s, annealing at 58 °C for 30 s, elongation at 72 °C for 50 s, and the final 5 min extension at 72 °C. Bisulfite PCR products were run in 2% agarose electrophoresis, excised, and extracted using a gel extraction kit (DP304-02, TIANGEN, China). After purification, PCR products cloned into the pMD18-T vector (6011, TaKaRa, Japan) for sequencing. For each amplicon, at least 6 individual clones were sequenced using an ABI PRISM 3130xl Genetic Analyzer (Applied Biosystems). Finally, the sequence was compared with the original sequence, the methylation site and number were counted, and the degree of methylation was analyzed.

Construction and transfection of lentivirus

The GBP4, CXCL9, CXCL10, and CXCL11 constructs cloned from a human cDNA library were inserted into the vector pLVX-GFP vector to generate pLVX-GBP4-GFP, pLVX-CXCL9-GFP, pLVX-CXCL10-GFP, and pLVX-CXCL11-GFP, respectively. For shRNA construction, the shRNA sequence against GBP4, CXCL9, CXCL10, and CXCL11 were cloned into the pLVX-GFP vector. To establish gene silencing and overexpressing stable cell lines, the lentiviruses containing pLVX-GBP4-shRNA-GFP, pLVX-CXCL9-shRNA-GFP, pLVX-CXCL10-shRNA-GFP, pLVX-CXCL11-shRNA-GFP, pLVX-GBP4-GFP, pLVX-CXCL9-GFP, pLVX-CXCL10-GFP, pLVX-CXCL11-GFP and corresponding negative control vectors were produced using pMD2.G and psPAX2 together with Lipofection 2000 in 293 T cells, respectively. When grew in 70–80% confluent, the Capan-1 and AsPC-1 cells were transfected with the lentivirus, then the puromycin was used to select the infected cells. The overexpression or silencing efficiency of target genes was assessed by quantitative reverse transcription-polymerase chain reaction (qRT-PCR). The shRNA sequence against GBP4 used in this study is listed in Table S2.

Real‐time quantitative reverse transcription PCR (qRT–PCR)

This experiment was carried out as instructed by the reagent manufacturer. TRNzol reagent (DP424, TIANGEN, China) was used to isolate total RNA from the cells followed by complementary DNA (cDNA) synthesis from the total RNA using cDNA synthesis kit (KR107, TIANGEN, China). Then, the amplification was carried out using SYBR green quantitative reverse transcription‐polymerase chain reaction (QR0100, Sigma‐Aldrich, USA). 2 − ΔΔCt method was used to calculate the relative levels of RNAs. Primers utilized in the present study are listed in Table S3.

Generation of the targeted methylation system

DNMT3A encodes an enzyme that catalyzes the de novo methylation of cytosine at CpG sites. Several studies have confirmed that the CRISPY-Cas9 fused with DNMT3A is an effective targeted methylation tool. To verify the relationship between GBP4 expression and DNA methylation status, we constructed a dCas9-SunTag-DNMAT3A-sgRNA-targeted methylation system. This system consists of dCas9-SunTag-BFP vector, scFv-GFP-DNMT3A and sgRNA-mCherry vector. The dCas9-SunTag vector was purchased from Addgene plasmid depository, and then directly used for experiment. To construct the scFv-sfGFP-DNMT3A plasmid, we ligated the scFv-sfGFP-VP64 vector, which was previously digested with RsrII and SpeI, with DNMT3A flanked by same set of restriction enzyme recognition sites. The construction of single guide RNA (sgRNA) vector for GBP4 (sgGBP4) was achieved by inserting the sgRNA sequence into the pLentiGuid-mCherry vector. Deactivated Cas9 (dCas9) was fused to SunTag epitopes and the single-chain variable fragment (scFv) was fused to green fluorescent protein (GFP) and DNMT3A1. Multiple copies of scFvDNMT3A1 can be guided by single guide RNA (sgRNA) to recognize specific loci and methylate regions of interest. After that, lentivirus containing dCas9-SunTag vector, scFv-GFP-DNMT3A1 vector, and sgRNA vectors were produced using pMD2.G and psPAX2 together with Lipofection 2000 in 293 T cells, respectively. The primer sequences for DNMT3A cloning and sgRNA (sgGBP4) sequences used in this study are listed in Tables S4 and S5, respectively.

Viral particle transfection and triple-fluorescent-positive cell purification

Capan-1 and AsPC-1 cells were transfected with lentiviral particles of dCas9-SunTag-BFP, scFv-GFP-DNMT3A, and sgRNA-mCherry in the polybrene-containing media (4 μg/ml polybrene, 10% FBS, 1% Penicillin/Streptomycin, DMEM). 48 h after transfection, triple-fluorescence-positive cells were purified using flow cytometry (BD FASCAria II) and then routinely cultured in medium. After 30 days, triple-fluorescence-positive Capan-1 and AsPC-1 cells were repurified. Subsequently, methylation status and expression level of GBP4 gene in these cells were analyzed.

ELISA assay

The levels of cytokines including IFN-γ, IL-2, IL-6, and IL-10 in the supernatant of tumor cell–T cell coculture system were determined using ELISA kits (Abcam) according to the manufacturer’s instructions.

Immune cell chemotaxis assay

The peripheral blood mononuclear cells (PBMCs) were obtained from peripheral blood samples of PDAC patients using the Ficoll density gradient sedimentation. Subsequently, CD8+T cells were isolated from PBMCs using the human T cell isolation kit (STEMCELL Technologies, Vancouver, BC, Canada). CD8+T cells (1 × 106) were seeded into the upper chambers of transwell plate (3 μm, Corning, New York, NY, USA). 1 × 106 PDAC cells (Capan-1 and AsPC-1) with or without target gene (GBP4, CXCL9, CXCL10, or CXCL11) overexpression or silencing were placed in the low chambers. After 24 h of coculture, the migrated CD8+T cells in the medium of the lower chamber were collected and counted by flow cytometry.

Pancreatic cancer orthotopic xenograft mouse model

6–8-week-old female C57BL/6 mice were obtained from Fudan University Shanghai Cancer Center (Shanghai, China). All experiments were approved by Fudan University Shanghai Cancer Center (FUSCC) animal ethics committee. Treatment and care of the animals were in accordance with Institutional Guidelines and the Animal Welfare Assurance Act. Capan-1-Mock and Capan-1-GBP4-OE stable cell lines were used to establish the orthotopic xenograft pancreatic cancer model. Briefly, tumor cells were trypsinized and resuspended in DMEM at a density of 6 × 107 cells/mL. The mice were maintained with isoflurane anesthesia during surgery. Surgical aseptic technique was applied for the construction of the model. Each mouse was injected with a 50-μL cell suspension containing 3 × 106 pancreatic cancer cells into the pancreas. The wound was sutured with Vicryl 4–0 (Ethicon, Somerville, NJ). The sizes of tumors implanted orthotopically were measured on day 7, 10, 14, and 18 and the fresh tumor tissue samples were subjected to flow cytometry analysis.

TIL isolation and flow cytometric analysis

To obtain a single-cell suspension of TILs, 2–4 mm PDAC tumor fragments were digested with Solo Digestive Enzyme Kit (JZ-SC-58201, Sinotech Genomics, China) for 1 h at 37 °C. The mixture was filtered with a 70 μm strainer, centrifuged at RT, 500 g, 5 min, and resuspended in 40% Percoll (65,455-52-9, Cytiva, USA). The suspension was slowly added to the upper layer of 70% Percoll without disturbing the interface and then centrifuged at 4 °C, 500 g, 30 min (with acceleration of 5 and deceleration of 0). The middle mist layer (Presumable TILs) was carefully sucked and washed three times with PBS containing 2% FBS. Subsequently, 5 μl of fluorescein-labeled antibodies with different emission spectra and 105 cells were mixed into the suspension and incubated at room temperature for 15 min in the dark. Then, wash the cells with PBS three times and resuspend in 500 μl of PBS. A total of 10,000 lymphocytes were analyzed using the flow cytometry (BD FASCAria II). After that, the flowjo (V10.8) software was used to analyze the flow cytometry data. The following antibodies were used in the present study: FITC anti-human CD3 Antibody (317,326, Biolegend), PE/Cyanine7 anti-human CD8 Antibody (980,910, Biolegend, USA), PerCP/Cyanine5.5 anti-human PD-1 Antibody (329,937, Biolegend, USA), APC/Fire™ 750 anti-human CTLA-4 Antibody (369,627, Biolegend, USA), Alexa-Fluor 700 anti-LAG-3 Antibody (369,343, Biolegend, USA), and PE/Fire™ 810 anti-human TIM-3 Antibody (345,059, Biolegend, USA).

Generation of patient-derived organoids (PDOs)

Surgical-resected human PDAC tissues were immersed in cold PBS saline (B310kJ, Basal medium, USA) with 1% Penicillin/Streptomycin (S110JV, Basal medium, USA). After that, the tissues were minced and digested with Solo Digestive Enzyme Kit (JZ-SC-58201, Sinotech Genomics, China) for 1 to 2 h at 37 °C, and filtered by a 70 μm strainer, respectively, followed by centrifugation at 300 g for 5 min. Then, the pellet was embedded in GFR Matrigel (3533-005-02, R&D systems, USA) and cultured in a complete organoid expansion medium containing RSPO1 (120–38, Peprotech, USA), WNT3A (315–20, Peprotech, USA), NOGGIN (120-10C, Peprotech, USA), FGF10 (100–26, Peprotech, USA), EGF (AF-100–15, Peprotech, USA), and other supplements as previously described [27, 28]. The medium was refreshed every 4 days. Organoids were generated and passaged for 3–5 generations, and then cryopreserved separately.

Organoid-T cell coculture

To evaluate the potential effect of GBP4 on the sensitivity of tumor organoids to T cell-mediated killing, we first measured the mRNA expression of GBP4 in organoids using quantitative real-time PCR and then categorized the organoids into two groups based on the mRNA level of GBP4: GBP4 low organoids for the first quartile (0–25%) and GBP4 high organoids for the fourth quartile (75–100%). Prior to coculture, GBP4 high organoids and GBP4 low organoids were embedded in GFR Matrigel and cultured in a complete organoid expansion medium separately. 96-well flat-bottom plates were precoated with 5 μg/mL anti-CD28 (64–0289-42, eBioscience, USA) and kept overnight at 4℃.

For cytotoxicity assay, organoids (GBP4 high and GBP4 low organoids) and CD8 T cells were harvested and resuspended in 150 U/mL IL-2 (20002, Peprotech, USA) containing T cell medium, with or without 20 μg/mL anti-PD-1-blocking antibody (αPD-1; A2005, Selleck, USA), respectively, and then seeded in anti-CD28-coated plates under the effector-target ratio at 16:1, 8:1, 4:1, and 2:1, respectively, to co-culture for 48 h. Then, LDH (lactate dehydrogenase) detection reagent (C0017, Beyotime, China) was added to the plates and the optical density was recorded at 490 nm wavelength. The mean value of the three wells was substituted into the following formula to calculate the killing efficiency % = (the measured hole—effector cells natural release OD value—target organoids natural release OD value + blank control OD value)/(the maximum release hole OD value of the target cells—target organoids natural release OD value).

For live imaging, organoids (GBP4 high and GBP4 low organoids) were partially digested from Matrigel to retain the three-dimensional architecture. The remaining Matrigel was washed before coculture. After that, organoids were incubated with 10 μM DIO (3,3′-dioctadecyloxacarbocyanine perchlorate; sc-214168, ChemCruz, USA) for 30 min at 37 °C. Meanwhile, 5 × 104 CD8 T cells were stained with 5 μM DID (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindodicarbocyanine perchlorate; sc-213424, ChemCruz, USA) for 10 min at 37 °C. Then, DIO-labeled organoids and DID-labeled CD8 T cells were harvested and resuspended in 150 U/mL IL-2 containing T cell medium, with or without 20 μg/mL αPD-1, respectively, and then seeded in anti-CD28-coated plates, respectively. After 48 h of coculture, organoids and CD8 T cells were imaged under confocal laser scanning microscopy (Olympus FV3000) and quantified via Fiji (Image J) software.

After 48 h of coculture, T cells were harvested and washed twice with FACS buffer followed by cell staining with anti-CD8-BV421 (568,373, BD, USA), near-IR viability dye (Life technologies), and anti-IFNγ-APC (655,933, BD, USA) using the Cytofix/Cytoperm kit (554,714, BD, USA) at 4 °C for 30 min.

Statistical analysis

Spearman correlation analysis was used to examine relationships between study variables. The Kaplan–Meier method was used to draw survival curves, and log-rank test was used for inter group comparison. The ssGSEA analysis was conducted using “GSVA,” “limma,” and “GSEABase” R packages. Quantification of imaging data and flow cytometry data was performed with GraphPad Prism9 from at least three samples. Data are expressed as the mean with SD with statistical differences determined by one-way ANOVA or multiple unpaired t-test. A two-tailed p-value < 0.05 was considered statistically significant and asterisks indicated the significance level of the p-value: *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Results

The expression patterns and prognostic value of GBP4 in PDAC

To reveal the expression patterns of GBP4 in PDAC, we retrieved the gene expression data from public databases (TCGA, GTEx, and CPTAC) and found that both the mRNA and protein levels of GBP4 in PDAC tissues were significantly higher than those in normal pancreatic tissues (Fig. 1A and B). After that, we performed IHC against GBP4 on local PDAC samples, and validated that GBP4 was significantly overexpressed in tumor tissues compared to adjacent normal tissues (Fig. 1C and D).

Fig. 1
figure 1

Overexpression of GBP4 is an independent factor for unfavorable prognosis in PDAC patients. A The expression of GBP4 mRNA in PAAD tissues and normal pancreatic tissues from TCGA database. B The expression of GBP4 protein in PAAD tissues and normal pancreatic tissues from CPTAC database. C Representative immunohistochemical (IHC) staining images of GBP4 expression in tumor tissue and adjacent normal tissue of PDAC patient. D The expression of GBP4 between tumor and adjacent normal tissues from 52 patients by quantitative analysis of IHC score. E The overall survival curves of the high-GBP4 and low-GBP4 expression PDAC patients from TCGA database. F Representative IHC staining images of the high- and low-GBP4 expression in tumor tissues of PDAC patients. G The overall survival curves of the high-GBP4 and low-GBP4 expression PDAC patients from our cohort H Univariate Cox regression for screening the prognostic factors of PDAC. I Multivariate Cox regression for screening the prognostic factors of PDAC. ****P < 0.0001

To identify the correlation between GBP4 expression and the overall survival (OS) of PDAC patients, the survival curves were visualized by Kaplan–Meier analysis and compared by log-rank test. PDAC patients with higher GBP4 expression displayed shorter overall survival (OS) than those with lower GBP4 expression in TCGA project (Fig. 1E). Similar result was observed in our IHC cohort containing 111 tumor tissues of patients with PDAC (Fig. 1F and G). To determine the independence of GBP4 in predicting survival, we conducted univariate and multivariate Cox regression analysis. The p values of GBP4 expression, age, tumor stage, node stage, and histologic grade were less than 0.05 in the univariate Cox model (Fig. 1H) and these variables were then further put into multivariate Cox analysis. As indicated in results of multivariate Cox analysis, high expression of GBP4 was an independent factor for unfavorable prognosis in PDAC (Fig. 1I).

DNA demethylation of GBP4 contributes to its overexpression in PDAC

The regulation patterns of GBP4 expression are largely unknown. To characterize the epigenetic patterns, specifically, DNA methylation patterns in regulatory regions of GBP4 in PDAC, we performed DNA methylation analysis on PDAC samples via SMART, a web server for analyzing DNA methylation of TCGA project. The results characterized seven CpG sites whose methylation levels were significantly lower in PDAC tissues compared to those in the corresponding normal tissues (Fig. 2A). Spearman correlation analyses showed significant negative correlations between the methylation levels of these seven CpG sites and the mRNA expression of GBP4 (Fig. 2B). We then employed Methprimer [29] to screen for CpG islands in the region of GBP4 located between − 2000 to + 1000 bp relative to transcription start site (TSS) and identified one short CpG island (147 bp), which located in the intron 1 of GBP4, and partially overlapped with the seven identified CpG sites (Fig. 3A). After that, we performed bisulfite genomic sequencing to investigate the methylation status of CpGs located at chr1:89,198,504-89,199,814 GRCh38/hg38 in GBP4 from PDAC tissues and adjacent normal pancreatic tissues using the two pairs of bisulfite PCR primers (Fig. 3A). As presented in Fig. 3B, the methylation levels of seven CpG sites in GBP4 from PDAC tissues were significantly lower than those from adjacent normal tissues. We also observed that the methylation levels of these seven CpG sites in GBP4 from PDAC cells were lower than those from normal pancreatic cells (Fig. 3C and D). These results suggested that hypo-methylation of GBP4 may be associated with its increased expression in PDAC.

Fig. 2
figure 2

The upregulation of GBP4 in PDAC may be associated with the hypo-methylation of seven CpGs in its regulatory regions. A Comparison of the methylation status of CpG sites between tumor tissues from TCGA database and normal tissues from GTEx database. B Correlation analysis between the methylation status of CpG sites and GBP4 expression in PDAC from TCGA database. The data are shown as mean ± SD

Fig. 3
figure 3

GBP4 showed hypo-methylated status in PDAC. A Chromosomal distribution of the CpG sites in the regulatory regions of GBP4, which was visualized by MethPrime, and the primers for bisulfite sequencing. B Representative methylation profiles of CpG sites in the regulatory regions of GBP4 from adjacent normal tissues and PDAC tissues detected using bisulfite genomic sequencing. Open circles indicate unmethylated CpG sites and closed circles indicate methylated CpG sites. C Representative methylation profiles of CpG sites in the regulatory regions of GBP4 from normal pancreatic ductal epithelial cells (HPDE6-C7) and PDAC cells (Capan-1, CFPAC-1, AsPC-1, and Mia Paca-2) detected using bisulfite genomic sequencing. D Differentially methylated CpG sites of GBP4 between normal pancreatic ductal epithelial cells and PDAC cells. The data are shown as mean ± SD

Targeted DNA methylation repressed the expression of GBP4 in PDAC cells

To further validate the effect of DNA methylation on GBP4 expression, the dCas9-SunTag-DNMAT3A-sgGBP4-targeted methylation system was generated (Fig. 4A and B) and was used to transfect pancreatic cell lines. Bisulfite genomic sequencing demonstrated that both the Capan-1 and AsPC-1 cells treated with dCas9-SunTag-DNMAT3A-sgRNA-targeted methylation system exhibited increased methylation levels of seven CpG sites in the GBP4 locus, while methylation levels of these CpGs remained almost unchanged in the negative control (Fig. 4C and D). Correspondingly, the results of qRT-PCR showed the significantly decreased mRNA level of GBP4 in cells transfected with the targeted methylation system and the unchanged mRNA level of GBP4 in negative control (Fig. 4E and F). Among these, the dCas9-SunTag-DNMAT3A-sgGBP4-1, dCas9-SunTag-DNMAT3A-sgGBP4-4, dCas9-SunTag-DNMAT3A-sgGBP4-5, and dCas9-SunTag-DNMAT3A-sgGBP4-6 exhibited the most effective targeted methylation and transcriptional repression. These results demonstrated that the overexpression of GBP4 in PDAC cells was caused by its active demethylation and it can be reversed by the dCas9-SunTag-DNMAT3A-sgGBP4-targeted methylation system.

Fig. 4
figure 4

The expression of GBP4 in PDAC cells was downregulated by targeted methylation with dCas9-SunTag-DNMAT3A-sgGBP4 system. A Schematic of dCas9-SunTag-DNMT3A1-sgGBP4. Deactivated Cas9 (dCas9) was fused to SunTag epitopes and the single-chain variable fragment (scFv) was fused to green fluorescent protein (GFP) and DNMT3A1. Multiple copies of scFvDNMT3A1 can be guided by single guide RNA (sgRNA) to recognize specific loci and methylate regions of interest. B The recognition sites of sgRNAs used to guide the dCas9-SunTag-DNMT3A1 in the regulatory regions of human GBP4, with arrows (aligned to the magnified regions) indicating 20 bp binding sites of sgRNAs. Arrows point toward the PAM sequence. C, D The methylation status of CpG sites in the regulatory regions of GBP4 gene in Capan-1 cells and AsPC-1 cells were analyzed using bisulfite genomic sequencing after transfection using the dCas9-SunTag-DNMAT3A-sgGBP4 system, respectively. E The mRNA expression of GBP4 in Capan-1 cells and AsPC-1 cells were measured using quantitative real-time PCR after transfection using the dCas9-SunTag-DNMAT3A-sgGBP4 system, respectively. The data are shown as mean ± SD; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

GBP4 promoted T cell exhaustion and tumor progression in PDAC

To determine the potential influence of GBP4 on tumor immune microenvironment (TIME) of PDAC, we performed spearman correlation analysis between mRNA level of GBP4 and abundance of immune cells, which was estimated by CIBERSORTx tool, in the PDAC samples from TCGA database. The results indicated that the expression of GBP4 was positively correlated with the abundance of CD8+ T cells and trafficking level of CD8+ T cells (Fig. 5A). Trafficking of immune cells into tumor site is closely associated with chemokines and their receptors [30]. In consistent with this, our results observed the positive correlation between GBP4 expression and the expressions of most of chemokines and chemokine receptors, specially CXCL9, CXCL10, and CXCL11 (Figs. 5A and S1A and B), which were also negatively correlated with survival of PDAC patients (Fig. S1C–E). To further validate the relationship between GBP4 and chemokines in vitro, we generated pancreatic cancer cell lines with GBP4 overexpressed or silenced, in which the mRNA levels of CXCL9, CXCL10, and CXCL11 were significantly increased or decreased, respectively (Fig. 5B). Immune cell chemotaxis assay showed that GBP4-overexpressed and/or CXCL9/10/11-overexpressed PDAC cells attract more CD8+ T cells compared to control PDAC cells, respectively (Fig. 5C), whereas neither GBP4-silenced PDAC cells nor CXCL9/10/11-silenced PDAC cells increased the migration of CD8+ T cells significantly, demonstrating that GBP4 may promoted the infiltration of CD8+ T cells in PDAC tissues via CXCL9, CXCL10, and CXCL11. High infiltration of CD8+TILs was usually associated with better survival of tumor patients [31]; however, this phenomenon was not observed in patients with high-GBP4 expression. Correlation analysis of TCGA showed a significant positive relationship between GBP4 expression and expressions of immune checkpoint (IC) genes in T cell abundant samples (Fig. 6A). Meanwhile, we measured the frequency of PD-1, CTLA-4, LAG-3, and TIM-3 positive CD8+TILs in freshly resected PDAC tissues, respectively. The proportion of the IC positive CD8+TILs was significantly higher in the high-GBP4 expression group than those in the low-GBP4 expression group (Fig. 6B and C). However, no significant differences were observed in the mRNA levels of several ICs between GBP4-wild type, GBP4-knockdown, or GBP4-overexpressed PDAC cell lines (Fig. S2), indicating that expression of ICs may not be directly regulated by GBP4 in PDAC cells.

Fig. 5
figure 5

The expression of GBP4 on tumor cells improved the infiltration of CD8+ T cells in PDAC tissues. A Correlation analysis between mRNA expression of GBP4, abundance of CD8+T cells, trafficking level of CD8 + T cells, and chemokines (CXCL9, CXCL10, and CXCL11) in the PDAC samples from TCGA cohort (n = 178). B The effects of GBP4 overexpression or GBP4 silencing on the mRNA expression of chemokines (CXCL9, CXCL10, and CXCL11) were detected by qRT-PCR. C Chemotaxis assay analysis of CD8+ T cell migration ability in the supernatants of tumor cells with or without GBP4, CXCL9, CXCL10, or CXCL11 overexpression, and with or without GBP4, CXCL9, CXCL10, or CXCL11 silenced. The data are shown as mean ± SD; ****P < 0.0001; WT, wild type; OE, overexpression; KD, knock down

Fig. 6
figure 6

The expression of GBP4 on tumor cells improved the exhaustion of tumor infiltrating CD8 + T cells in PDAC tissues A Heatmap showing the relationship between GBP4, CD8+ T cells, and immune checkpoint inhibitors. B Representative frequencies of immune checkpoint (PD-1, CTLA-4, TIM-3, and LAG-3) positive CD8+ TILs from surgically resected PDAC tissue samples in a flow cytometric dot plot. C Comparison of the frequencies of PD-1+CD8+ TILs, CTLA-4+CD8+ TILs, TIM-3+ CD8+ TILs, and LAG-3+CD8+ TILs between GBP4 high and GBP4 low expressing PDAC tissue samples. D Comparison of the progenitor exhausted CD8+T cell (Tprog-ex) gene signature score between GBP4 high and GBP4 low expressing PDAC samples from TCGA cohort. E Comparison of the terminally exhausted CD8 + T cell (Tterm-ex) gene signature score between GBP4 high and GBP4 low expressing PDAC samples from TCGA cohort. F Representative images of mIHC staining of Tprog-ex (TCF7+PD-1+ CD8+T cells) and Tterm-ex (TCF7PD-1+ CD8+T cells) on GBP4 high and GBP4 low expressing PDAC samples from our cohort. G Comparison of the numbers of TCF7+PD-1+ CD8+T cells and TCF7PD-1+ CD8+T cells between GBP4 high and GBP4 low expressing PDAC samples from our cohort. *P < 0.05, ***P < 0.001, ****P < 0.0001

Recent studies identified two subsets of tumor infiltrating exhausted CD8+ T cells, including the progenitor/stem-like exhausted CD8+ T cell (Tprog-ex cell) and the terminally differentiated exhausted CD8+T cell (Tterm-ex cell) [32]. The Tprog-ex cells are characterized by relatively better antitumor functions, and may differentiate into Tterm-ex cell after persistent stimulation [33]. Gene signature analysis result showed that PDAC samples with high GBP4 exhibited significantly higher Tterm-ex cell score and lower Tprog-ex cell score than the samples with low-GBP4 expression (Fig. 6D and E) [34]. It was reported that TCF1+PD-1+CD8+ T cell and TCF1PD-1+CD8+ T cell represent progenitor/stem-like exhausted CD8+ T cell and terminally differentiated exhausted CD8+T cell, respectively [35]. In addition, our mIHC staining further validated that the number of TCF7+PD-1+CD8+ TILs was significantly higher in the PDAC tissues with low-GBP4 expression than that in the PDAC tissues with high-GBP4 expression (Fig. 6F, G, and H), and the number of TCF7PD-1+CD8+ TILs were significantly higher in the GBP4 high expression group compared with the GBP4 low expression group (Fig. 6F, G, and I). In addition, the effects of GBP4 on T cell function were evaluated using a coculture system consist of CD8+T cells and PDAC cells with GBP4-overexpressed or silenced. The mRNA expressions of T cell exhaustion markers, including CTLA-4, PD-1, TIM-3, TOX, and T-bet, were significantly increased in the GBP4-overexpression group and decreased in the GBP4-silenced group compared with control group (p < 0.05), while those of Ly108, CD69 and TCF7 in CD8+T cells were significantly decreased in the GBP4-overexpression group and increased in the GBP4-silenced group compared with control group (p < 0.05) (Fig. S3). In addition, the concentrations of IL-10 and IL-6 in the cultured supernatants were significantly higher in GBP4-overexpression group and lower in the GBP4-silenced group compared with control group (p < 0.05), and those of IL-2 and IFN-γ in the cultured supernatants were significantly lower in the GBP4-overexpression group and higher in the GBP4-silenced group compared with control group (p < 0.05) (Fig. S4). In consistent with these, analyze-based orthotopic xenograft mouse model showed that tumor growth and expression level of immune checkpoints, including PD-1 and TIM-3, were significantly increased in GBP4-overexpression group compared to control group (Fig. S5). These results further supported the protumor role of GBP4 via causing T cell exhaustion in PDAC.

GBP4 improved the sensitivity of patient-derived organoids to anti‐PD-1 therapy

Previous studies suggested that patients with immune-“hot” TME tend to be more responsive to anti-PDL-1/PD-1 therapy [36, 37], and high expression level of PD-L1 also contributed to a better outcome of anti-PDL-1/PD-1 therapy in many solid tumor types [38, 39]. As mentioned above, we found that GBP4 not only promoted the infiltration of CD8+T but also upregulated the expression of ICs, including PD-L1. Therefore, we hypothesized that GBP4 may be a potential biomarker to predict the efficacy of anti-PDL-1/PD-1 therapy. We first performed GSEA enrichment analysis on the samples from TCGA using the gene sets from the MSigDB database. The results showed that the gene set representing the effectiveness of anti-PD-1 therapy was significantly enriched in the group of high-GBP4 expression (Fig. 7A). After that, we performed in vitro experiment to directly test the potential effect of GBP4 on anti-PD-1 therapy. The CD8+T cells isolated from PBMC were cocultured with GBP4 high and GBP low patient-derived organoids separately (Fig. 7B), which were determined based on the expression level of GBP4 (Fig. 7C). Both cytotoxicity assay and organoid imaging analysis showed that GBP4 high organoids displayed a significantly increased cell death in the coculture system, compared to GBP4 low organoids, after treatment with anti-PD-1 antibody (Fig. 7D–F). Besides, the percentage of IFN-γ in CD8 + T cells significantly elevated in the coculture system with GBP4 high organoids compared to the coculture system with GBP4 low organoids after PD-1 blockade (Fig. 7Gand H). These findings indicated that GBP4 significantly improved the sensitivity of PDAC to PD-1 blockade therapy.

Fig. 7
figure 7

The expression of GBP4 on tumor cells improved the sensitivity of pancreatic cancer to PD-1 blockade therapy A GSEA analysis indicated that the gene set representing the effectiveness of anti-PD-1 therapy was significantly enriched in the PDAC samples with high-GBP4 expression based on data from the TCGA database. B A schematic representing the work flow for evaluating the impact of GBP4 on sensitivity of tumors to PD-1 blockade therapy using patient-derived pancreatic cancer organoid-T cell coculture system. B The mRNA expression of GBP4 in patient-derived pancreatic cancer organoids were measured using quantitative real-time PCR. Organoids were stratified into two groups based on the mRNA level of GBP4: GBP4 low organoids for the first quartile (n = 5) and GBP4 high organoids for the fourth quartile (n = 5). C Cytotoxicity assay using patient-derived pancreatic cancer organoids as targets. CD8 + T cells were cocultured with GBP4 high organoids or GBP4 low organoids in the absence or presence of anti-PD-1 antibody at the indicated effector-to-target (E/T) ratios for 48 h. D The immunofluorescence images showed the cytotoxicity of CD8+ T cells on the GBP4 high organoids or GBP4 low organoids in the absence or presence of anti-PD-1 antibody. Organoids (green) and CD8+ T cells (red) are shown. Scale bars, 50 μm. E The size of GBP4 high organoids or GBP4 low organoids treated with control or anti-PD-1 antibody in the coculture with CD8+ T cells. Optical images were analyzed using ImageJ to determine the organoid size. F The mean number of CD8+ T cells in per HPF in the coculture with GBP4 high organoids or GBP4 low organoids in the absence or presence of anti-PD-1 antibody. G Representative fow cytometry dot plots and statistical analysis of percentages of CD8+IFNγ+T cells in the coculture with GBP4 high organoids or GBP4 low organoids in the absence or presence of anti-PD-1 antibody. The data are shown as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; αPD-1, anti-PD-1 antibody

Discussion

The interaction between tumor cells and immune cells is critical in antitumor immunology and immunotherapy. Up-regulation of GBPs in several cancer types is associated with a poor prognosis [17,18,19,20], and further mediated immune cell infiltration and T cell exhaustion (23). These findings indicated that GBPs may represent the cross-talk between immune cells and tumor cells. In this article, GBP4 was observed to be highly expressed in PDAC and positively correlated with poor prognosis. We dissect the DNA methylation patterns of GBP4, and successfully manipulated the GBP4 expression via the dCas9-SunTag-DNMT3A-sgGBP4 system. We also observed that GBP4 was also positively correlated with the infiltration of immune cells, particularly CD8+ T cells. Trafficking of immune cells into tumor site is closely associated with chemokines and their receptors [40]. In consistent with this, we found that GBP4 regulates the expression of CXCL9/10/11 in pancreatic cancer cells and promotes the infiltration of infiltration of CD8+ T cells via these chemokines. We speculated that GBP4 may positively regulate the expressions of CXCL9/10/11 via lipopolysaccharide, since there exist direct interaction between lipopolysaccharide and GBPs [41, 42]. In this study, we also observed that the expression of several immune checkpoints on CD8+ T was significantly upregulated upon the expression of GBP4, the underlying mechanism may be related to interferon (IFN), since IFN-mediated upregulation of inhibitory immune checkpoints can facilitate immune escape [43, 44]. These findings indicated that GBP4 may be positively correlated with IC blockade response. We identified a higher abundance of terminally exhausted T cells and a lower proportion of progenitor exhausted T cells in GBP4 high PDAC tissues than in GBP4 high PDAC tissues. Progenitor exhausted T cells could differentiate into terminally exhausted T cells but exhibited relatively lower cytotoxicity [45]. Terminally exhausted T cells expressed more cytokines and exerted strong antitumor activity [46]. Moreover, previous study reported that anti-PD-1 treatment mainly induced the expansion of progenitor exhausted CD8+ TILs [47]. In consistent with these, our in vitro experiment showed that the cytotoxic activity of CD8+ T cells was higher in the coculture system with GBP4 high organoids than that in the coculture system with GBP4 low organoids after anti-PD-1 treatment. To this end, we speculated that anti-PD-1 treatment may re-invigorate exhausted CD8 T cells and induced the expansion of progenitor exhausted CD8+ T cells simultaneously, thereby leading to an increased antitumor immune response in GBP4 high PDAC patients.

To conclude, our studies revealed the unique expression patterns and DNA methylation regulatory mechanisms of GBP4 in PDAC, and proposed that GBP4 may serve as an indicator for better clinical outcomes of anti‐PD-1 immunotherapy in PDAC.