Background

Pancreatic ductal adenocarcinoma (PDAC) is a highly aggressive malignancy with a 5-year overall survival (OS) rate of only 13% and is mostly resistant to chemotherapy.1 Unfortunately, immune checkpoint inhibitor-targeted therapies2 also showed only limited efficacy in PDAC.3 4 This resistance is largely attributed to the desmoplastic and immunosuppressive tumor microenvironment (TME) that provides a barrier for T cell infiltration resulting in an immunologically cold neoplasm.5 Surprisingly, the microenvironment of human PDAC contains a relatively high number of T cells6 7 although these T cell populations have been selected for their protumor rather than cytotoxic properties8 during the evolution of the lesions.4 Also, cancer cell subpopulations that can evade elimination during immune editing are favored, resulting in the generation of clonal variants that can escape immune control. Several previous studies corroborate this view and show that cancer cell-intrinsic factors significantly contribute to the resistance of PDAC to T cell attack.9–12

Recently, we developed a three-dimensional (3D) co-culture platform of PDAC and T cells that enables the identification and validation of the immune resistance phenotype.13 From transcriptomic analyses of T cell-sensitive and T cell-resistant murine clonal PDAC cell lines generated from the KPC (KrasLSL-G12D/+; Trp53fl/fl; Ptf1aCre/+) mouse model we identified transglutaminase type 2 (TGM2) as a contributor to the T cell resistance phenotype in KPC cells. High TGM2 expression has been associated with worse survival of patients with PDAC14 and can contribute to the evasion from T cell interaction,14–17 although the underlying mechanisms remain elusive. Interestingly, gene expression analysis of the TCGA database also showed that TGM2 is highly expressed in PDAC in a T cell-enriched microenvironment and associated with poor patient survival.16 18 This suggests that additional TGM2-mediated regulation mechanisms contribute to T cell suppression beyond T cell exclusion. Here, we show in 3D co-cultures that TGM2-mediated T cell resistance of PDAC cells is mediated by microtubule (MT)-dependent vesicle trafficking and secretion of immunosuppressive cytokines, such as granulocyte colony-stimulating factor (G-CSF) and granulocyte-macrophage CSF (GM-CSF). Also, adding these immunosuppressive cytokines to TGM2-expressing PDAC cells led to T cell resistance by impairing T cell activation and consequently T cell-mediated antitumor cytotoxicity. The clinical relevance of these findings was corroborated in animal studies, as well as in PDAC patient-derived organoids (PDOs) where we sensitized TGM2high PDOs toward T cell-mediated killing after pretreatment with sublethal doses of MT-targeting chemotherapeutics. Collectively, our data demonstrate a yet less known connection between cytoskeletal modulation in cancer cells and T cell functionality that can be leveraged to enhance synergistic effects of MT-targeting chemotherapeutic agents, such as paclitaxel and vincristine, with immunotherapy.

MethodsCell lines

Generation, use, and culture of clonal cell lines were conducted as described by Lahusen et al.13 In brief, the T cell resistant (H5, A10) and sensitive (C8, C7) KPC cell lines were derived from KPC (KrasLSL-G12D/+; Trp53fl/fl; Ptf1aCre/+; Sifter et al 19) mouse models via limited dilution assay.

Stable knockdown of Tgm2 in KPC cells

Stable knockdown of TGM2 expression in resistant KPC PDAC cells was performed as described previously.13 Five plasmid constructs (Mission shRNA DNA pLKO.1-puro RRID: Addgene_8453, Sigma; TRCN0000009529 RefSeq: NM_009373, TRCN0000009530 RefSeq: NM_009373, TRCN0000009531 RefSeq: NM_009373, TRCN0000009532 RefSeq: NM_009373, TRCN0000009533 RefSeq: NM_009373), coding for different shRNA against murine Tgm2, were used. After puromycin selection (2 µg/mL), the knockdown cell line with the strongest reduction in TGM2 expression (determined by western blot, Clone ID TRCN0000009529; shTGM2) was selected. Cells stably expressing scrambled shRNA (shSCR) served as control. Recombinant lentiviral particles were handled under S2 safety conditions.

Animal studies and immunization with tumor cells

Male C57BL/6 wild-type mice, aged 8 weeks, were bred, kept, and treated as described by Lahusen et al.13 For T cell education by tumor antigens and in vivo studies, mice were injected with cancer cells.9 Mice were anesthetized via isoflurane inhalant anesthesia until lack of response to a toe pinch. All hair on both flanks and back was shaved, and surgical scrub was performed. One million KPC cells were suspended in 100 µL of serum-free cell culture medium and injected subcutaneously into the right and left flank of each mouse (syngeneic tumor allografts). 14 days after cancer cell injection, mice were euthanized using carbon dioxide-standard procedure followed by cervical dislocation. Longitudinal midline incision was performed to flip open the skin on both sides. Once the subcutaneous tumors with the adjacent inguinal and axillary lymph nodes were exposed, tumors as well as tumor-draining lymph nodes (TdLNs) were carefully removed, and TdLNs were subjected to T cell isolation.

Isolation of T cells from mice and T cell activation

T cell isolation from TdLNs was performed as described by Lin et al.9 Murine T cells were isolated via negative selection by using Dynabeads Untouched Mouse T Cells Kit (# 11 413D, Invitrogen) according to the manufacturers’ instructions. For ex vivo activation of naïve T cells isolated from mouse spleen (‘plate-activation’), cell culture plates were coated with an anti-CD3E antibody (human: #555329, murine: #567115, BD) at 5 µg/mL and an anti-CD28 antibody (human: #555725, murine: (#567110, BD) at 2.5 µg/mL and T cell incubated for 1 day prior to their isolation.

2D and 3D T cell co-culture with KPC PDAC spheroids or PDOs

These assays were carried out as described previously.9 13 20

Co-culture treatments with recombinant cytokines

Recombinant cytokines GM-CSF (human: #215-GM, mouse: #415-ML, Bio-Techne) and G-CSF (human: #214-CS, mouse: #414-CS, Bio-Techne) were added (in DPBS with 0.1% BSA; BSA from Sigma-Aldrich, #A9418) at the start of PDO T cell 3D co-culture (no matrix) or KPC PDAC T cell 2D co-culture at concentrations presenting the median concentration of the suggested range for use by the supplier (GM-CSF: 20 pg/mL; G-CSF: 40 pg/mL). For cytokine neutralization studies, G-CSF (40 pg/mL)/GM-CSF (20 pg/mL) recombinant protein mix was added either alone or in combination with the corresponding neutralizing antibody mix (nAbs; 1 µg/mL anti-human GM-CSF: #MAB215, biotechne; 5 µg/mL anti-human G-CSF: #MAB214, biotechne). In another condition, only nAbs mix was added to the PDO 3D T cell co-culture. Prior to the start of co-culture, reagents were preincubated for 15 min at 37°C, followed by another 15 min preincubation with T cells at 37°C. Reagents were then added at the start of co-culture with human T cells for 72 hours. 3D T cell co-culture of PDOs without treatment served as control.

Pretreatments with TGM2 inhibitors or nocodazole

The TGM2 inhibitors ZDON (#616467) and ERW1041E (#509522) (both Merck) were used for 24 hours pretreatment (50 µM or 100 µM in DMSO) of KPC spheroids prior to 3D culture. Additionally, ZDON (50 µM or 100 µM in DMSO) was used for 24 hours pretreatment21 22 of PDOs prior to start of T cell 3D co-culture. Nocodazole (#1228, Bio-Techne) was used at 50 µM and 100 µM23 for 3 hour as pretreatment of PDOs prior to start of T cell 3D co-culture.

Pretreatments with chemotherapeutics

Gemcitabine (Gem; #S1714, Selleckchem), paclitaxel (Pac; #S1150, Selleckchem), and vincristine (VCR; #5505911, TEVA) were used for either 4 hours pretreatment (Gem or Pac: 100 nM in DMSO) or 24 hours pretreatment (VCR: 100 nM injection solution with pH 3.5–5.5 and osmolarity of 600 mOsm/L) of PDOs at sublethal doses prior to 3D T cell co-culture (no matrix).

Conditioned medium transfer

Conditioned medium (CM) was collected after 72 hours 2D culture from KPC PDAC cells and added as a 50% mix with fresh medium to the respective other cell line prior to the start of 72 hours 2D co-culture with educated T cells.

Transwell migration assay

Migration of 500,000 human T cells in 0.5% FBS in the upper chamber toward the recombinant human cytokines G-CSF and GM-CSF in DMEM with 0.5% FBS (lower chamber) for 2 hours was measured. For cytokine pretreatment, 500,000 T cells were pretreated for 30 min at 37°C with G-CSF (40 pg/mL) or GM-CSF (20 pg/mL) prior to the migration assay. Endpoint readout was the measurement of T cell numbers in both upper and lower chambers using the CASY 0 (OLS) automatic cell counter. The percentage of migrated cells in the lower chamber compared with all cells in both chambers was calculated.

Mouse cytokine array

Cytokine array for detection of 62 mouse cytokines in CM of 3D cultured KPC cells was carried out as follows: After 72 hours spheroid formation, all medium was removed and 75 µl serum-free DMEM medium was added to the inside of the hydrogel chips. After 24 hours 3D culture, medium from 12 chips per condition was removed, pooled, and subjected to the cytokine array. The RayBio Mouse Cytokine Antibody Array C3 from RayBiotech (#AAM-CYT-3-2) was used according to the manufacturers’ instructions. Final dot blot results were automatically quantified by using the Protein Array Analyzer in Fiji.24 ,25 Results were blanked to background and normalized to positive control.

Flow cytometry

T cells (max. 1 mio. per condition) were stained with 1 µL of the respective antibodies (online supplemental table S1) per 100 µL sample in flow cytometry buffer (0.5% BSA and 2 mM EDTA in DBPS; EDTA: #A3145, AppliChem) for 10 min at 4°C in U-bottom 96-well plates. Samples were washed once with 200 µL buffer, centrifuged at 350 g for 10 min (4°C), and subjected to flow cytometry (Attune NxT, ThermoFisher) after resuspension in 250 µL buffer using the channel fitting to the respective fluorophore. In the case of overlapping fluorescence spectra, compensation was applied by using single marker stains. All gates were adjusted to unstained and isotype controls. For determining PDO cell numbers (frequency of total or total cell count) PDO-specific and T cell-specific FSC over SSC gates were created by using PDO-only and T cell-only controls.

Western blot

Western blots were performed as previously described.13 Primary antibodies were used in blocking buffer (1:10 000 anti-GAPDH, ThermoFisher, #MA1-16757; 1:1000 anti-TGM2, Cell Signaling, #3557) for 24 hours incubation.

Immunoprecipitation

KPC PDAC cell lysate protein extract (250 µg) was incubated with 2.5 µL antibody (α-tubulin, 1 µg/µL; #T6199, Sigma) overnight at 4°C. 15 µL dynabeads (#10 003D, Invitrogen) were equilibrated with lysis buffer (20 mM HEPES, 150 mM NaCl, 5% glycerol, 0.1% Triton-X, pH 7.5) and blocked in 1 mg/mL BSA overnight at 4°C. Thereafter, the beads were added to the protein/antibody solution and incubated for 1 hour at room temperature. After removal of the supernatant, beads were washed twice with 500 µL lysis buffer, and the protein was eluted with 50 µL SDS sample buffer (#BP-111R, Boston BioProducts). Binding of TGM2 to tubulin-coupled beads was analyzed by employing anti-TGM2-antibody (#3557, Cell Signaling) in a dilution of 1:1000 by western blotting. Beads without anti-α-tubulin antibody coating were used as control.

Immunocytochemistry

KPC PDAC cells (2×104) were seeded into chamber slides (#80841, Ibidi), and incubated overnight. Thereafter, the cells were fixed with 4% PFA/4% sucrose for 10 min at 37°C, permeabilized with 0.3% Triton-X-100/PBS for 5 min at RT, washed three times with 0.03% Triton-X-100/PBS, blocked with 2.5% BSA for 1 hour, and incubated with primary antibodies (anti-detyrosinated MTs clone AA12 isotype IgG2a: #ab254154, abcam; anti-β-tubulin: #PA1-16947, ThermoFisher) diluted 1:200 in PBS/0.03% Triton-X-100/1.25% BSA overnight at 4°C. After washing three times with 0.03% Triton-X-100/PBS, the samples were incubated for 1 hour at RT with secondary Alexa-fluor-coupled antibodies (anti-mouse AF488: #A-11001; anti-rabbit AF647: #A-21245, ThermoFisher) diluted 1:2000 in PBS/0.03% Triton-X-100/1.25% BSA. After washing the cells three times with 0.03% Triton-X-100/PBS, the samples were embedded and analyzed by confocal microscopy.

Immunofluorescence

Collagen patches with KPC spheroids and T cells were fixed in 4% PFA/4% sucrose for 1.5 hours at RT, then washed three times with PBS. Permeabilization was done using 0.3% Triton X-100/PBS for 15 min at RT, followed by three washes with 0.03% Triton X-100/PBS. Cells were blocked with 2.5% BSA in PBS for 30 min at RT. Primary antibody against mouse CK19 (#sc-376126 AF488 conjugated, Santa Cruz) was diluted in 1.25% BSA/PBS and incubated overnight at 4°C. Cells were washed three times with PBS and gently rocked for 5 min between each wash. After removing excess PBS, samples were mounted with Fluoromount-G (#00-4958-02, ThermoFisher). Slides were stored at 4°C and imaged within 2 weeks. Imaging per spheroid was done by fluorescence microscopy (BZ-X Series, Keyence) at 10X using the Cy5 channel for CellTracker Deep Red-stained T cells and the GFP channel for AF488 immunostained CK19 on KPC cells.

Immunohistochemistry

Immunohistochemistry staining for paraffin sections of allograft tumors dissected from mice 14 days after s.c. flank injection of KPC cells was performed as previously described.13 Primary antibodies against mouse (CD3: #ab16669, abcam; CD8: #ab209775, abcam; FOXP3: #ab215206, abcam; GZMB: #bs-1351R, Bioss; c-Caspase3/7: #9661, Cell Signaling) were incubated 1:200 overnight at 4°C.

Vesicle trafficking

Cells seeded on chamber slides (Ibidi) were transfected with 500 ng EGFP-BDNF (a gift from Matthias Kneussel) using the K2 Transfection System from Biontex according to the manufacturer’s instruction. After 16 hours of incubation, the cells were imaged every 5 s for 1 min using the Olympus IXplore Live fluorescence microscope. Vesicle trafficking speed (µm/s) was analyzed with Fiji plugin TrackMate V.7.13.2 (https://github.com/trackmate-sc/TrackMate) for 15 cells of each group.

IFNγ ELISA

Human/mouse IFNγ ELISA was applied with medium collected after 72 hours 2D T cell co-culture with KPC cells or 3D T cell co-culture with PDOs (no matrix). The assay was carried out with 190 µL medium by using the mouse IFN-γ DuoSet ELISA Kit (#DY485) or the human IFN-γ DuoSet ELISA Kit (#DY285B) with the DuoSet Ancillary Reagents Kit 2 (#DY008) from Bio-Techne according to the manufacturers’ instructions. Background control (no sample added) was subtracted and concentrations were determined via calculation of a standard curve for IFNγ.

MT plus-end tracking

Tracking of MT dynamics with mRuby-EB3 was carried out as previously described.26 27 Live-cell imaging was performed using a confocal laser scanning microscope TCS SP8-HCS (Leica) equipped with a Plan APO CS2 63×/1.3 NA water immersion objective. Images were acquired with a 90% open pinhole to compensate for thermal drift and active hardware autofocus. Cells were imaged in a K-Frame OKO-LABS stage-top incubator with 5% CO2 at 37°C. Video images were acquired at one frame per two seconds. For quantification of the MT velocity, 20 particles per cell in 8 cells per condition were tracked using the Fiji manual tracking tool.28 Additionally, mRuby-EB3 tracks were counted per frame for the first 10 frames (first 20 s) and the percentage of tracks in late frames only (frames 7–10, time 15–20 s) to all tracks counted over 10 frames was calculated.

RNA extraction and quantitative RT-PCR

RNA extraction and qRT-PCR were carried out as previously described.13 Primers were purchased from Qiagen (QuantiTect Primer Assay, #24990), and their sequences are provided in online supplemental table S2.

RNA sequencing and data analysis

For RNASeq, KPC PDAC clonal cell lines were grown for 72 hours as spheroids in hydrogel chips (3D culture) and were then embedded in collagen matrix for 48 hours. Subsequently, RNA was extracted from the matrix as described by Lin et al.9 Bulk RNA samples were sent to the Biomedical Sequencing Facility (BSF) of the CeMM Research Center for Molecular Medicine in Vienna, Austria. Primary analysis of BAM files was done by the BSF core facility by use of the DESeq2 Bioconductor package. DeSeq2 normalized counts per million were subjected to gene set enrichment analysis (GSEA) using the GSEA_4.3.2. software and hierarchical clustering/PCA using the ClustVis webtool.29 The GSEA analyzed mouse gene sets originated from the Gene Ontology (GO) database, Reactome database, or “Hallmarks” gene set collection (MSigDB). A GSEA map for GO biological pathways was generated by using Cytscape_3.10.1 with p values <0.025 and edge similarity (cut-off)=0.7 (unconnected single nodes were removed). For hierarchical clustering, rows/columns were clustered using correlation distance and average linkage. For PCA, rows were unit variance scaled and prediction ellipses at 95% confidence per phenotype cluster were calculated. Venn overlap of pairwise comparisons (upregulated genes filter: fold change >1, protein-coding) using DESeq2 adjusted p values and the Venn Diagram webtool.30

TCGA-PAAD RNAseq data analysis

Kaplan-Meier (KM) OS analysis of PDAC patients for high and low gene expression cohorts based on TCGA-PAAD data sets (https://portal.gdc.cancer.gov/projects/TCGA-PAAD,31) was carried out by using the KM-Plot webtool. Logrank statistics (p values and HRs) were calculated. The best expression cut-off was auto selected.29 Spearman correlation analyses were performed by using the TISIDB (for single genes) or GEPIA2 (for gene signatures) webtool32 for correlation with either single selected genes (GEPIA2), immune checkpoint markers (TISIDB), or TISDB immune subtype infiltration signatures. Spearman p values and r correlation coefficients (r) were calculated, respectively.

Statistical analysis

Statistical tests and graphing were carried out using GraphPad Prism V.10, Microsoft Excel, GSEA_4.3.2, ClustVis, Cytoscape, GEPIA2 and TISIDB. An unpaired t-test was used for dichotomous comparisons. Multiple unpaired t-tests or ordinary one-way analysis of variance were used for multiple comparisons. Logrank tests were applied for survival analyses and Spearman correlation for correlation analyses. Gene ranking for GSEA was performed via Signal2Noise, Diff_of_Class, or tTest. p<0.05 (dichotomous) or p-adjusted <0.05 (multiple testing) was set as the threshold for statistical significance.

ResultsIdentification of distinct T cell response phenotypes in clonal KPC PDAC cells

The heterogeneous parental cell line KPC7598 was used as a source for clonal lines with distinct T cell response phenotype using the co-culture platform established recently.13 For the co-culture studies, we applied primary murine T cells from two different sources: Tumor antigen-educated T cells were derived from TdLNs of C57BL/6 wild-type mice that were injected with clonal KPC cells, respectively. Naïve T cells were derived from mice which had not been exposed to tumors. Orthogonal readouts were used to establish and cross-validate distinct phenotypes: Cancer cell spheroid apoptosis mediated by tumor-educated T cell-mediated cancer spheroid cytotoxicity (figure 1A,B; online supplemental figure S1A–D), in vitro (figure 1C,D; online supplemental figure 1E–G) and in vivo (figure 1E,F) T cell infiltration, overall T cell activation (online supplemental figure S2A,B) and CD8+T cell abundance, as well as activated CD8+T cell subtypes expressing activation markers CD44 (figure 1G,H; online supplemental figure S2C) or CD122 (online supplemental figure S2D,E). Note that the percentage of CD8+CD44+ cells is significantly higher in the tumor antigen-educated compared with the naïve T cell co-culture group (online supplemental figure S2C). According to their primary in vitro T cell response phenotype, the most distinct clonal cancer cell lines were grouped into “T cell-sensitive (S)” (C7, C8) and “T cell-resistant (R)” (H5, A10). R-cells showed significantly reduced cancer cell apoptosis (figure 1A,B; online supplemental figure S1C,D), lower T cell infiltration rate in vitro (figure 1C,D) and in vivo (figure 1E,F), as well as diminished expression of T cell activation markers (figure 1G,H; online supplemental figure S2B–E) after co-culture with educated T cells compared with S-cells. Notably, the respective T cell response phenotype was confirmed by co-cultures with naïve T cells that were activated by exposure on tissue culture plates coated with anti-CD3/-CD28 (ie, “plate-activated” T cells) for T cell-mediated cytotoxicity in (online supplemental figure S1A,B), T cell infiltration (online supplemental figure S1E–G), and T cell activation (online supplemental figure S2A), as well as by standardized 2D co-culture assays for T cell-mediated cytotoxicity (online supplemental figure S1A–D), and T cell activation (figure 1G,H; online supplemental figure S1A, C–E). These data indicate robust T cell response phenotypes across different co-culture setups and T cell origins. The results also indicate that factors beyond antigen-specificity contribute to the differential T cell response. Thus, we characterized and validated distinct T cell response phenotypes of clonal KPC PDAC cell lines derived from the KPC mouse model.

Figure 1

Identification of distinct T cell response phenotypes in clonal KPC PDAC cells. (A) Representative fluorescence images (merged with brightfield) showing c-Casp3/7 positive (+) resistant (R; H5, A10) or sensitive (S; C8, C7) KPC spheroids after 72 hours 3D co-culture with tumor-educated T cells. Matrix with KPC spheroids only (no T cells added) was used as control. (B) Number of c-Casp3/7+cells/spheroid presented as fold change to KPC spheroids only control. Means were calculated for 10 spheroids/condition. N=4. (C) Representative fluorescence images showing educated T cell infiltration into the collagen matrix with R (H5, A10) or S (C8, C7) KPC spheroids after 72 hours 3D co-culture. Infiltrating CellTracker Deep Red-stained T cells are indicated by white arrows. Empty matrix with the addition of T cells only (no KPC spheroids; imaged at the z-axis level of KPC spheroids) was used as control. (D) Number of infiltrated T cells/spheroid. Means were calculated for 10 spheroids/condition. N=4. (E) Infiltration of CD3+T cells into subcutaneous (s.c.) syngeneic allograft tumors from resistant or sensitive clonal KPC cells. Percentage of infiltrating T cells in relation to all cells was detected by DAB staining immunohistochemistry and quantified in (F) for 10 areas/condition. N=3. (G) Representative flow cytometry plots for CD8+CD44+ T cells after 72 hours 2D educated T co-culture with resistant or sensitive clonal KPC cells. Cells were gated to CD44 (APC-Cy5) over CD8 (Pacific blue). The CD8+CD44+ T cell population gates and percentages (% of single cells) are indicated. (H) The percentage of CD8+CD44+ T cells for 72 hours 2D educated T cell co-culture was calculated. N=4. Bars are as indicated. *p≤0.05, **p≤0.01, ***p≤0.001, ****p≤0.0001. PDAC, pancreatic ductal adenocarcinoma.

TGM2 is a pivotal contributor to the T cell-resistant phenotype in KPC PDAC cells

These resistant and sensitive clonal KPC cell lines provided an opportunity to uncover molecular mechanisms underlying the distinct T cell response phenotypes. Toward this goal, we compared gene expression patterns of the cell lines (https://www.ncbi.nlm.nih.gov/sra/PRJNA1020447,33). Hierarchical clustering of the top 100 differentially expressed genes revealed the transcriptomic divergence between resistant and sensitive clones that also set them apart from the original heterogeneous KPC7598 cells (online supplemental figure S3A). GSEA uncovered distinct immune response phenotypes between resistant and sensitive cells (online supplemental figures S3B and S4). Notably, pathways associated with immune cell regulation were enriched in sensitive compared with resistant cells (online supplemental figures S3 and S4). In the overlap of upregulated genes, we identified 204 candidate genes (figure 2A) which were enough to separate T cell response phenotypes and the original clone in PCA (figure 2B). In the overlap of these candidate genes with the GO extracellular matrix pathway (selected due to the specific matrix-based embedded co-culture platform used), we identified 12 hit genes which were further investigated (figure 2C). Within these hit genes, the most pivotal gene whose expression positively correlated with reduced OS in PDAC patients (online supplemental figure S5A,B), as well as a regulatory T cell (Treg)-rich and a Programmed Cell Death Protein 1 (PD-1)-high expressing microenvironment, was TGM2 (online supplemental figure S5C,D). Accordingly, TGM2 upregulation was confirmed in resistant compared with sensitive cells on transcriptomic (figure 2D,E) and proteomic (figure 2F, online supplemental figure S6A,B) levels. There is a discordance between TGM2 mRNA and protein expression level. However, the TGM2 protein expression level is significantly higher in both resistant compared with sensitive cells (figure 2E,F; online supplemental figure S6A,B). As stated before, high TGM2 expression correlated with reduced OS in PDAC patients (online supplemental figure S5A,B). According to analysis based on data mining of the TCGA-PAAD RNAseq database by using the KM-Plot tool, patients with a high TGM2-expressing CD8+T cell-enriched PDAC environment have a lower OS, which comprise 43% of all analyzed PDAC (figure 2G). Taken together, we identified TGM2 as an essential gene contributing to the T cell-resistant phenotype in PDAC, with significant implications for immune modulation.

Figure 2

Transcriptomic analysis revealed TGM2 as a contributor to the T cell-resistant phenotype. (A) Venn overlap of significantly upregulated genes (RNAseq DESeq2) for resistant (R: H5, A10) compared with sensitive (S: C8, C7) clones in pairwise comparisons. The Tgm2-containing overlapping node is highlighted. Genes with ≥3 pairwise comparison overlaps were selected (204 candidates). N=3. (B) Principal component analysis (PCA) for the 204 candidates (RNAseq) with prediction ellipses (95% CI) for each T cell response phenotype cluster and the original cell line (KPC7598). N=3. (C) Venn overlap between the candidate genes from (A) and the GO extracellular matrix (ECM) gene set. The 12 overlapping genes including Tgm2 (highlighted) are indicated. (D) Volcano plot (RNAseq DESeq2) comparing gene expression for the most resistant (A10) and most sensitive (C8) KPC clone. Red points (including Tgm2, highlighted) are upregulated genes in A10 and blue points are upregulated genes in C8 (cut-off: log2(fold change) ≥1 or ≤−1 and -log10(adj. (P) ≥ 2). N=3. (E) Relative expression of Tgm2 (RNAseq) for each KPC clone (DESeq2 normalized counts). N=3. (F) Representative protein expression (western blot) of TGM2 from resistant or sensitive clonal KPC cells. For quantification, relative band densities were normalized to GAPDH. N=3. (G) Kaplan-Meier (KM) plots for overall survival (OS) of PDAC patients comparing TGM2 high and TGM2 low expression cohorts under CD8+T cell enriched (N=76) or deprived (N=101) microenvironments, respectively (logrank test). The analysis is based on data analysis of the TCGA-PAAD RNAseq database by using the KM-Plot tool. *p≤0.05, ****p≤0.0001. PDAC, pancreatic ductal adenocarcinoma.

TGM2 downregulation rendered resistant KPC cells more sensitive

We sought to validate the contribution of TGM2 to T cell resistance in our model and investigate additional TGM2-mediated resistance mechanisms. First, shRNA-mediated knockdown (kd) of TGM2 in the R-cell lines A10 and H5 was performed. Western blot analysis confirmed depletion of TGM2 by 88% and 68% in H5 and A10 cell lines relative to scrambled controls (figure 3A; online supplemental figure S6C,D). On TGM2 kd, T cell infiltration into 3D spheroids grown from TGM2-depleted cells increased 2–4 fold relative to the scrambled controls (figure 3B,C; online supplemental figure S6E,F). This effect in T cell infiltration was shown to be both antigen-specific after co-culture with educated T cells (figure 3B,C), as well as antigen-unspecific after co-culture with naïve plate-activated T cells (online supplemental figure S6E,F). Also, the expression of T cell activation markers (Gzmb, Tnfa, Ifng) in educated T cells co-cultured with both shTGM2 kd cell lines H5 and A10 increased by 4-fold over 10-fold compared with scrambled controls (shSCR) (figure 3D). Moreover, shTGM2 cells showed increased sensitivity to T cell-mediated cytotoxicity in Annexin V flow cytometry assays that revealed a lower percentage of viable shTGM2 cells compared with shSCR cells after educated T cell co-culture (figure 3E,F). The T cell activation/infiltration and cytotoxic phenotypes of shTGM2 and shSCR from both cell lines H5 and A10 cells were corroborated in animal studies, where we observed significantly decreased FOXP3+ (figure 3G,H), but increased CD8+ (figure 3G,H) and GZMB+ (online supplemental figure S6G,H) T cell infiltration, as well as increased c-Casp3/7+cells (online supplemental figure S6I,J) in subcutaneous (s.c.) allograft tumors of mice injected with shTGM2 compared with tumors from shSCR cells. These results indicate that downregulation of TGM2 sensitized resistant KPC PDAC clones toward T cell-mediated cytotoxicity and facilitates an inflamed tumor immune microenvironment.

Figure 3

TGM2 downregulation increased effector T cell functionality and sensitized resistant PDAC cells to T cell-mediated cytotoxicity. (A) Representative protein expression (western blot) of TGM2 from shSCR (control) and shTGM2 (TGM2 knockdown) KPC cells for H5 and A10 T cell resistant PDAC clonal cells. Band densities were normalized to GAPDH and are shown as shTGM2/shSCR fold change. N=4. (B) Representative fluorescence images showing educated T cell infiltration into collagen matrix with shSCR or shTGM2 (H5/A10 clones) KPC spheroids after 72 hours 3D co-culture. Infiltrating CellTracker Deep Red-stained T cells are indicated by white arrows. (C) Number of infiltrated T cells/spheroid. Means were calculated for ≥5 spheroids/condition. N≥4. (D) Relative expression (2−ΔCt, normalized to Gapdh) via RT-qPCR of T cell activation markers in educated T cells after 3D co-culture with shSCR or shTGM2 spheroids. N≥3. (E) Annexin V flow cytometry assay for the identification of viable cells in shSCR or shTGM2 cell lines grown for 72 hours without (KPC control) or with (2D co-culture) educated T cells. The viable cell population gate (DAPI/Annexin V-PE double negative single cells) is highlighted (bold) and the viable cell percentage is shown. (F) For quantification, the difference in percentage of viable KPC cells after 2D co-culture to viable KPC cells only (KPC control) was calculated. N=6. (G) Infiltration of CD8+ and FOXP3+T cells into subcutaneous (s.c.) syngeneic allograft tumors from shSCR- or shTGM2-transduced H5 or A10 cells. Percentage of infiltrating T cells in relation to all cells was detected by DAB staining immunohistochemistry and quantified in (H) for ≥8 areas/condition. N≥4. Bars are as indicated. *p≤0.05, **p≤0.01, ***p≤0.001, ****p≤0.0001. PDAC, pancreatic ductal adenocarcinoma.

TGM2 regulates MT network density and dynamics

Next, we aimed to reveal the underlying mechanism of action of TGM2. One notable fact was that the respective T cell response phenotype of shTGM2 and shSCR cells was comparable between co-culture with tumor antigen-educated and anti-CD3/-CD28-activated T cells (figure 3B,C; online supplemental figure S6E,F). Anti-CD3/-CD28-activated T cells were derived from tumor-naïve mice, that had not been tumor antigen educated. Therefore, our data indicate a partial KPC tumor antigen-unspecific effect and suggest paracrine signaling contributing to the modulation of a T cell response in our clones. Since paracrine signaling requires intact intracellular trafficking and secretion of soluble factors, these processes are largely dependent on cytoskeletal regulation. Here, in particular, MTs are essential for intracellular vesicle transport.23 Therefore, we also examined gene expression signatures of the main cytoskeletal components. We found MT-related gene sets, particularly key gene sets such as Microtubule Motor Activity and Microtubule Binding, which were significantly enriched in resistant compared with sensitive clones (figure 4A, online supplemental figure S7A). Among the four most differentially expressed tubulin proteins (TUBA1A, TUBA1B, TUBA1C, TUBB4B), expression of Tubulin Beta 4B Class IVb (TUBB4B) was most distinctly increased in resistant versus sensitive cells (online supplemental figure S7B). A correlation analysis based on TCGA public data generated with PDAC samples revealed significant negative correlations between TUBB4B expression and infiltration of different immune cell subtypes, including CD8+/CD4+T effector memory cells, activated B cells, NK cells, NKT cells, and neutrophils (figure 4B). Notably, expression of all four tubulin proteins combined as a gene signature showed a significant positive correlation with TGM2 (based on public TCGA-PAAD data analysis), suggesting involvement of TGM2 in tubulin regulation (figure 4C). Most strikingly, a tubulin β- immunoprecipitation assay confirmed a direct interaction between TGM2 and tubulin (figure 4D; online supplemental figure S7C,D). We examined the cellular effects of this interaction and found a marked reduction in cytoplasmic MT network density on the depletion of TGM2, indicated by a significantly lower mean fluorescence intensity of fluorescently stained ß-tubulin and detyrosinated (stabilized) MTs in shTGM2 compared with shSCR cells (figure 4E,F). Live-cell imaging of MT plus-end tracking showed a significant decrease in MT dynamics and on average shorter end binding protein 3 (EB3) track lengths per cell, reflected by lower velocities of MT growing ends (figure 4G,H) and decreased percentage of EB3 tracks in late frames (figure 4I) in TGM2-depleted versus scrambled control cells (figure 4G–I). Also, intracellular vesicle movement (velocity) was significantly decreased in shTGM2 compared with shSCR cells, indicating impaired vesicle trafficking on TGM2 depletion (figure 4J). These results demonstrate that TGM2 interacts with tubulin and modulates microtubular dynamics and vesicle trafficking.

Figure 4

TGM2 mediates microtubule network density and dynamics. (A) Gene set enrichment analysis comparing T cell resistant (R; H5, A10) to sensitive (S; C8, C7) KPC clones for the indicated GO molecular function (MF) gene sets. Nominal p value and normalized enrichment score (NES) were as indicated. N=3. (B) Pearson correlation of TUBB4B expression with the infiltration of different immune cell subtypes (as indicated), based on public TCGA-PAAD RNASeq data analysis using TISIDB. N=179. (C) Spearman correlation analysis of the tubulin expression signature with TGM2 based on TCGA-PAAD RNASeq data analysis using GEPIA2. N=179. (D) Representative immunoprecipitation assay (western blot) from p