Clinical Specimens
We collected 18 fresh NSCLC tissue samples and 2 pulmonary nodule tissue samples (from patients P01-P20) for metabolomic analyses, immunoblotting, and PDOT evaluations from The First Affiliated Hospital of Nanjing Medical University and The Affiliated Brain Hospital of Nanjing Medical University. The study received approval from the Research Ethics Committees of both institutions (2023-KL008-02, 2019-SR-450). Patient information for P1-P10 is listed in Supplementary Table.1, and for P11-P20 in Supplementary Table.13. Paired tumor tissues and non-cancerous adjacent tissues samples from 5 patients with wild-type EGFR (P01-P05) and 5 with EGFR-mutated NSCLC (P06-P10) were used for unbiased metabolomic analysis (data presented in Fig.1a) and ELISA assay (data presented in Fig.5c). Tumor and matched adjacent non-tumor samples from 3 wild-type EGFR patients (P01, P04, P05) and 3 EGFR-mutated patients (P06, P07, P08) underwent immunoblot analysis (data presented in Fig.1d, Fig.1j, Fig.5b, Extended Data Fig.2g). Two pulmonary nodule samples (from patients P11, P12) and samples from 8 tumor samples from EGFR-mutated NSCLC patients (P13-P20) were utilized to prepare PDOT models (data shown in Fig.5m).
Furthermore, we collected 39 paraffin-embedded NSCLC samples (from patients P21-P59) from The First Affiliated Hospital of Nanjing Medical University for clinical validation of the correlation between GFPT2 expression, EGFR activation, hyperglycosylation of tumors, and CD8+T cell infiltration. Patient information for P21-P59 is detailed in Supplementary Table.11. Through immunohistochemical analysis of GFPT2, these 39 patients were classified into a GFPT2-positive expression subgroup (n=19, P21-P39) and a GFPT2-negative expression subgroup (n=20, P40-P59), followed by pEGFR, CD8+T cell IHC, and PHA lectin immunofluorescence (IF) staining (data presented in Fig.5a; original histological imaging provided in Supplementary Fig.S2-S5). Notably, tumor samples and matched NAT from 4 patients (P21, P27, P33, and P36) underwent GFPT2 immunohistochemical analysis (data presented in Fig.1f, with original histological imaging shown in Supplementary Fig.S1)
In total, three publicly available clinical datasets were analyzed for the tissue distribution expression of HBP metabolic enzymes. (1) Proteomic data from 8 EGFR wild-type (EGFRWT) and 47 EGFR-mutated (EGFRMut) NSCLC patients, sourced from the Database of Genotypes and Phenotypes (ph-s001954.v1.p1)47, were utilized to investigate the induced expression of HBP metabolic enzymes under EGFR mutation conditions (data presented in Fig.1c). Patient information and related data are listed in Supplementary Table.3. (2) Two independent human proteomic databases, the tissue-based map of the human proteome (PRJNA183192)48 and the Human Proteome Map database (available at http://www.humanproteomemap.org/query.php), were utilized to analyze the expression of HBP metabolic enzymes across various major organs (data presented in Fig.1e and Extended Data Fig.1f). Patient information and related data are listed in Supplementary Table.4. (3) Proteomic profiles from matched tumor and adjacent non-tumor tissues of 66 NSCLC patients with EGFR mutations (sourced from the Database of Genotypes and Phenotypes, ph-s001954.v1.p1)47, along with those from an additional independent cohort of 51 EGFR-mutated NSCLC patients (IPX0001804000)49, were analyzed to determine whether GFPT2 is overexpressed in tumors due to EGFR mutation induction (data presented in Fig.1f). Patient information and related data are listed in Supplementary Table.4.
Isolation of CD8+T from human PBMC
Pre-coat a 12-well plate overnight at 4 °C using CD3 antibody (5 μg/ml, 317326, Biolegend), and before inoculation, rinse the wells three times with phosphate buffer saline (PBS) solution to remove any residual CD3. Collect fresh human peripheral blood into a 50 mL EP tube, add an equal volume of PBS, and gently mix by pipetting. Add 1.5 times the volume of the blood of Ficoll solution (00421, Serumwerk Bernburg AG, Germany) to another 50 mL EP tube, and gently layer the mixed blood along the side of the tube onto the Ficoll solution, avoiding mixing the two. Then centrifuge at room temperature at 800 g for 20 minutes, requiring slow acceleration and deceleration (acceleration 3, deceleration 0). After centrifugation, the solution is separated from top to bottom into plasma, peripheral blood mononuclear cells (PBMCs), lymphocyte separation liquid, and granulocytes. Discard the plasma layer, collect the PBMCs into a new 15 ml EP tube, and wash twice with 5 volumes of PBS, each time centrifuging at room temperature at 1500 rpm for 10 minutes. Discard the supernatant, resuspend the cells in 4 ml of red blood cell lysis solution (C3702, Beyotime), and let stand at room temperature for 5 minutes to lyse the remaining red blood cells. Then add 8 ml PBS to stop the lysis and centrifuge at room temperature at 1500 rpm for 10 minutes. Resuspend the cells in PBS and follow the procedure in the EasySepTM Human CD+8 T Cell Iso Kit (17953, STEMCELL) manual to extract CD8+T cells: add cocktail to the cell suspension at 50 μl/ml, invert several times, and incubate for 5 minutes at room temperature. Add magnetic beads to the cell suspension at 50 μl/ml, invert several times, place on a magnetic rack, and after 5 minutes, transfer the supernatant to a new 15 mL EP tube. Wash twice with 5 volumes of PBS, each time centrifuging at room temperature at 1500 rpm for 10 minutes. Then resuspend and count in 1640 complete medium containing CD28 (5 μg/ml, 302934, Biolegend) and IL-2 (10 ng/ml, 200-02, Pepertech), and finally inoculate at a density of 1 × 106 cells/ml onto the pre-coated CD3 12-well plate. After 48 hours, transfer and continue to culture in 1640 complete medium containing IL-2, stimulating weekly with CD3 and CD28 antibodies as described above50.
PDOT isolation and drug treatment
Fresh tumors from NSCLC patients were thoroughly washed in PBS and minced into a pulp under sterile and cool conditions. Digestive fluid was added at a ratio of 15 mL per 1 g of tumor (complete 1640 medium, 1 mg/ml type IV collagenase, 15 mmol/l HEPES) and digested on a shaker at 37 °C at 90 rpm for 20 minutes, after which an equal volume of complete 1640 medium was added to terminate the digestion. Spheroids ranging from 40-100 μm in diameter were collected by sieving twice. To lyse red blood cells, 5 ml of red blood cell lysis solution was added and the mixture was left to stand at room temperature for 5 minutes, followed by the addition of twice the volume of PBS to cease lysis, and centrifuged at 1500 rpm for 5 minutes at room temperature. The spheroids were resuspended in ECM gel (2.5 mg/ml type I collagen from mouse tails, 2 mM acetic acid, 100 mM HEPES, 1 g/l sodium bicarbonate) and seeded into a pre-cooled 48-well plate at a density of 100 clusters per 150 μl of ECM. After 24 hours, complete 1640 medium was added to continue culture36,51. The similarity in the proportion of immune cells within the spheroids to that within the tumor was validated by flow cytometry (FCM). The tumor classification and identification of EGFR mutations were confirmed by pathology experts.
PDOTs received treatment with 50 µM Arbutin, 200 µg/ml of the commercialized anti-PD-1 antibody (aPD-1), Tripolyimumab, or their combination. Four days post-treatment, PDOTs were enzymatically dissociated into single cells for FCM. Non-specific antibody binding was precluded with Human TruStain FcX™ (422301, Biolegend, USA). A suite of fluorochrome-conjugated monoclonal antibodies was then employed to identify surface antigens. Five minutes prior to FCM, propidium iodide (PI) staining was performed. The subset of CD45-PI+ cells was delineated to quantify the proportion of tumor cell death within the PDOTs. Gating strategies are presented in Supplementary Fig.6.
Ex vivo co-culture
5 × 104 tumor cells were plated in each well of a 48-well plate. These cells were then exposed to pre-activated T cells at a specific effector-to-target ratio (2:1). After a 48-hour co-culture period, cells were stained with 5 μg/ml PI (ST1569-10mg, Beyotime) and immediately analyzed using the CytoFLEX flow cytometer (Beckman Coulter, USA). Tumor cells were distinguished based on GFP positivity or CD45 negativity. Gating strategies are presented in Supplementary Fig.6. In order to verify whether GFPT2 in tumor cells may influence T cells in a non-cell-contact manner, the T cell cytotoxicity experiments described above were also conducted in Transwell inserts.
In the CD8+T cell chemotaxis assay, tumor cells were seeded at a density of 2 × 105 per well in the bottom chamber of 5-μm-pore Transwell inserts (3421, Costar). Pre-stained CD8+T cells were placed in the upper chamber at 1 × 105 cells/well and incubated for 48 h, either with or without 1.2 μg/ml CXCL10 neutralizing antibody (33016, R&D, USA). Following incubation, T cell migration to the lower chamber was quantified using a CytoFLEX flow cytometer (Beckman Coulter, USA).
To evaluate the impact of tumor GFPT2 expression on vascular cells, HUVEC were seeded in 24-well plates at a density of 2 × 105 cells/well and subjected to a wound healing assay. Once HUVEC reached 90-100 % confluence, a standardized wound was created in the cell monolayer using a pipette tip, with the well bottoms marked to guide the initial imaging of the wounded areas. Subsequently, tumor cells were introduced into the upper chamber of Transwell inserts and co-cultured with the indicated HUVECs for an additional 24 h. To neutralize the augmented VEGF secretion resulting from GFPT2 overexpression in tumor cell, 6 μg/ml Bevacizumab (Roche) was administered. At the 24-hour mark, wound gap was imaged using the Lionheart FXTM Intelligent Live Cell Imaging Analysis System (Bio-Tek Instruments, USA). The wound gaps were measured by Image J software, and migration rates were calculated using the formula: Migration rate = (wound gap (0 h) - wound gap (24 h))/wound gap (0 h), where wound gap=wound area/wound length.
Animal treatment and tumor challenges
Animal care and experimentation received approval from the Institutional Animal Care and Use Committee (IACUC) of China Pharmaceutical University and were conducted under the National Research Council's Guide for the care and use of laboratory animals and animal welfare. Female Balb/c Nude mice and C57BL/6N mice, both aged 4 - 5 weeks, were obtained from the Beijing Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China). Additionally, female immunodeficient NOD.Cg-PrkdscidIl2rgtm1Vst/Vs (NPG) mice, alongside their humanized counterpart model HSPC-NPG (created by transplanting human CD34+ hematopoietic stem and progenitor cells into NPG mice), were obtained from Beijing Vitalstar Biotechnology Co.,Ltd. (Beijing, China). Mice were housed in specific-pathogen-free facilities at the China Pharmaceutical University (animal authorization reference number: SYXK2021-0011), where they were maintained under a controlled environment of 22 – 24 °C, 50–60 % humidity and a 12 h light/dark cycle (08:00 to 20:00 light on) with access to standard laboratory food and water ad libitum. Pre-inoculation screenings ensured mice were free from pathogens and mycoplasma.
In the studies depicted in Fig.2, both NPG and HSPC-NPG mice were subcutaneously inoculated 1.0 × 107 H1975 or PC9/OR2 cells featuring stable GFPT2 knockdown (shGFPT2) on their right flank, and an equivalent number of control cells (shNC) on the left flank. In a separate set-up, C57BL/6N immunocompetent mice were introduced subcutaneously on the right hind flank with 5 × 105 EGFR-mutated LLC (EGFR T790M/L858R) cells, with or without GFPT2 interference. For GFPT2 interference subgroup, FTY720 was administrated (1 mg/kg, i.p. qod) to inhibit lymph node egress starting on day 6 post cell injection. Anti-CD8 group (250 μg/mouse, i.p, qwd) was employed to neutralize CD8+T cell activity starting on day 8 post cell injection. Upon the emergence of palpable tumors, tumor volume was periodically assessed every two days using the formula length × width2 × 0.5. Experimental endpoints were designated upon any group's average tumor size attaining the ethical limit of 1200 mm3 or if the animals showed any signs indicative of adverse health. CO2 inhalation was used to euthanize mice. Tumor weights were determined at the endpoint. Sample sizes were not predetermined through statistical methodologies; however, each experimental group comprised a minimum of five mice. Randomization of mice was executed prior to treatment according to the initial tumor volumes (not blinded to investigators).
In the investigation of GFPT2 pharmacological inhibition (Fig.5), the in vivo efficacy of Arbutin was initially assessed, an inhibitor specific to the GFPT2 isoform, previously identified and validated in vitro. Female C57BL/6N immunocompetent mice were introduced subcutaneously on the right hind flank with 5 × 105 EGFR-mutated LLC. The mice were randomized into two groups of eight mice on the basis of tumor size. Randomization and treatment initiated on day 1; the mean tumor volume at the start of dosing was 50 mm3. SPSS (0.9 % w/v saline, i.p.) or Arbution (50 mg/kg, i.p.) was administered daily. Tumor volume was periodically assessed every two days using the formula length × width2 × 0.5. After 21 days treatment, mice were sacrificed and tumor mass ascertained. Through evaluation of Arbutin's capacity to curb tumor progression, curtail CD8+T cell infiltration and activity within tumors, and modulate the HBP pathway in a tissue-specific manner, its potential as a therapeutic inhibitor was appraised. Subsequently, the combinatorial efficacy of Arbutin with PD-1 blockade was assayed. Female C57BL/6N immunocompetent mice harboring EGFR-mutated LLC tumors were assorted into distinct treatment regimens: (a) Control (0.9 % w/v saline, i.p., qd); (b) Arbutin (50 mg/kg, i.p. qd); (c) aPD-1 (200 μg/mouse, i.p., qw); (d) Arbutin (50 mg/kg, i.p. qd) + aPD-1 (200 μg/mouse, i.p., qw); (e) Bevacizumab (5 mg/kg, i.p., tiw) + aPD-1 (200 μg/mouse, i.p., qw). Tumor volumes were assessed every other day, until the survival end point was reached. Survival probability for each group was determined based on actual mortality observed within 77 days following xenograft implantation.
Single-cell RNA-seq alignment and analysis
To investigate the TIME diversity in GFPT2-knockdown tumors, we conducted single-cell RNA sequencing (scRNA-seq) on a pooled sample of 24,777 cells derived from three EGFR-mutated LLC C57BL/6N xenografts, each treated with either shNC or shGFPT2 (n=3 per group). Tumor tissues were washed in PBS, minced, enzymatically dissociated with 0.1 % trypsin for 8 minutes, followed by 0.8 mg/ml collagenase type I for 60 minutes at 37 °C with 5 % CO2, including intermittent vigorous shaking. The cell suspension was filtered through a 70-μm strainer, red blood cells were lysed, and the remaining cells were suspended in PBS for single-cell 3’-cDNA library construction using the 10X Genomics Chromium Single Cell 3’ protocol. The encapsulation, cDNA synthesis, and subsequent RNA-sequencing were conducted by Gene Denovo, Guangzhou, China.
Single-cell RNA-seq libraries were sequenced on Illumina NovaSeq using 150 nt paired-end reads. Data processing was performed with Cell Ranger 4.0.0, adhering to default and recommended parameters. Reads were aligned to the Mus musculus reference genome (Ensembl_release109) using STAR. After initial processing with the Seurat package into Seurat Objects, mitochondrial gene representation was assessed for quality control. Cells were excluded based on the following criteria: mitochondrial gene content over 10%, presence of more than 7,000 genes (nFeature > 7000), or a total count above 100,000 (nCount > 100,000). Post-filtering, the dataset was normalized and subjected to Principal Component Analysis for dimensionality reduction. Integration of different samples was performed with Harmony (Broad Institute), and clustering was conducted using the t-SNE algorithm. Cluster-defining marker genes were identified, with the top 10 markers delineating each cluster. CellChat was used for cell-cell communication analysis.
Analysis of tumor-infiltrating immune cells using flow cytometry
Tumor tissues were finely minced on ice using sterile forceps and scalpel. The resultant fragments underwent digestion with a mix of DNase I (10104159001, Sigma, USA) and Collagenase D (11088858001, Sigma, USA) at 37 °C for 45 min with agitation set at 220 rpm. Following digestion, the tissues suspensions were filtered through a 70 μm cell strainer and centrifuged at 1,000 rpm for 5 min. Cell pellets were resuspended with red blood cell lysis buffer (C3702, Beyotime, Shanghai, China) for 4 min on ice, followed by a subsequent 5 min centrifugation at 1000 rpm. Obtained single cells were then activated using a cell activation cocktail (423301, Biolegend) and cultured for 6 h at 37 °C, preparing for intracellular IFNγ and GZMB staining. Cell viability was assessed using the Zombie Aqua Fixable Viability Kit (423101, Biolegend, USA) and nonspecific binding was blocked with either Human TruStain FcX™ (422301, Biolegend, USA) or anti-mouse CD16/CD32 blocking antibody (101319, Biolegend, USA) 52. Both surface and intracellular antigens were detected using a panel of fluorochrome-conjugated monoclonal antibodies. Gating strategies employed to identify humanized or murine tumor-infiltrating immune cells are presented in Supplementary Fig.7 and 8.
For humanized tumor-infiltrating immune cells, cells were stained with the various combinations of fluorochrome-labelled antibodies in FACS buffer (PBS + 2 % FBS). Panel 1: GFP-labeling Tumor, APC-conjugated CD3 (1:100, 317317, Biolegend), APC/Cy7-conjugated CD8 (1:100, 344713, Biolegend), BV421-conjugated IFNγ (1:100, 502531, Biolegend). Panel 2: GFP-labeling Tumor, APC-conjugated CD3 (1:100, 317317, Biolegend), APC/Cy7-conjugated CD8 (1:100, 344713, Biolegend), BV421-conjugated GZMB (1:100, 396413, Biolegend). Panel 3: GFP-labeling Tumor, APC-conjugated CD3 (1:100, 317317, Biolegend), PerCP/Cy5.5-conjugated CD4 (1:100, 317428, Biolegend), APC/Cy7-conjugated CD25 (1:100, 302613, Biolegend), PE-conjugated Foxp3 (1:100, 320007, Biolegend). Cells were analyzed using a CytoFLEX II flow cytometer (Beckman Coulter) with FlowJo software.
For murine tumor-infiltrating immune cells, cells were stained with the various combinations of fluorochrome-labelled antibodies in FACS buffer (PBS + 2 % FBS). Panel 1: APC/750-conjugated CD45 (1:100, 147713, Biolegend), BV421-conjugated CD3 (1:100, 100335, Biolegend), PerCP/Cy5.5-conjugated CD8 (1:100, 100733, Biolegend), BV605-conjugated IFNγ (1:100, 505840, Biolegend), APC-conjugated GZMB (1:100, 372204, Biolegend); Panel 2: BV421-conjugated CD45 (1:100, 103133, Biolegend), PE/Cy7-conjugated CD3 (1:100, 100219, Biolegend), BV605-conjugated CD4 (1:100, 100547, Biolegend), Alexa 647-conjugated CD25 (1:100, 102019, Biolegend), PE-conjugated Foxp3 (1:100, 320007, Biolegend); Panel 3: BV421-conjugated CD45 (1:100, 103133, Biolegend), PE/Cy7 conjugated F4/80 (1:100, 123113, Biolegend), BV605-conjugated CD11b (1:100, 101237, Biolegend), PE-conjugated CD206 (1:100, 141705, Biolegend), APC-conjugated CD86 (1:100, 105011, Biolegend); Panel 4: APC/750-conjugated CD45 (1:100, 147713, Biolegend), BV421-conjugated CD3 (1:100, 100335, Biolegend), Pacific Blue-conjugated CD19 (1:100, 115526, Biolegend), PE-conjugated NK1.1 (1:100, 108707, Biolegend). Cells were analyzed using a CytoFLEX II flow cytometer (Beckman Coulter) with FlowJo software.
Public bioinformatics dataset analysis
The TCGA data set called Lung Adenocarcinoma (TCGA, PanCancer Atlas) was downloaded from the TCGA data portal (https://tcga-data.nci.nih.gov/tcga/). The correlations between the expression of HBP metabolites and survival in EGFR-mutated NSCLC were provided by Kaplan-Meier Plotter (http://kmplot.com/analysis/). The human proteome map (http://www.humanproteomemap.org/query.php) was used to describe the expressions of HBP metabolic enzymes. The nucleotide sequences in the GFPT2 promoter which Xbp1s binding to were predicted with JASPER 2022 (https://jaspar.genereg.net/). Gene ontology (GO) enrichment analysis was carried out with THE GENE ONTOLOGY RESOURCE (https://geneontology.org/). The poor prognosis of PD-1 inhibitors was obtained from timer 2.0 (http://timer.cistrome.org).
47 Chen, Y. J. et al. Proteogenomics of Non-smoking Lung Cancer in East Asia Delineates Molecular Signatures of Pathogenesis and Progression. Cell 182, 226-244 e217 (2020).
48 Uhlen, M. et al. Proteomics. Tissue-based map of the human proteome. Science 347, 1260419 (2015).
49 Xu, J. Y. et al. Integrative Proteomic Characterization of Human Lung Adenocarcinoma. Cell 182, 245-261 e217 (2020).
50 Puig-Saus, C. et al. Neoantigen-targeted CD8(+) T cell responses with PD-1 blockade therapy. Nature 615, 697-704 (2023).
51 Aref, A. R. et al. 3D microfluidic ex vivo culture of organotypic tumor spheroids to model immune checkpoint blockade. Lab Chip 18, 3129-3143 (2018).
52 Zeng, Q. et al. Aberrant hyperexpression of the RNA binding protein FMRP in tumors mediates immune evasion. Science378, eabl7207 (2022).