NFAT5 governs cellular plasticity-driven resistance to KRAS-targeted therapy in pancreatic cancer

Abstract

Resistance to KRAS therapy in pancreatic ductal adenocarcinoma (PDAC) involves cellular plasticity, particularly the epithelial-to-mesenchymal transition (EMT), which poses challenges for effective targeting. Chronic pancreatitis, a known risk factor for PDAC, elevates TGFβ levels in the tumor microenvironment (TME), promoting resistance to KRAS therapy. Mechanistically, TGFβ induces the formation of a novel protein complex composed of SMAD3, SMAD4, and the nuclear factor NFAT5, triggering EMT and resistance by activating key mediators such as S100A4. Inhibiting NFAT5 attenuates pancreatitis-induced resistance to KRAS inhibition and extends mouse survival. Additionally, TGFβ stimulates PDAC cells to secrete CCL2, recruiting macrophages that contribute to KRAS bypass through paracrine S100A4. Our findings elucidate the role of TGFβ signaling in EMT-associated KRAS therapy resistance and identify NFAT5 as a druggable target. Targeting NFAT5 could disrupt this regulatory network, offering a potential avenue for preventing resistance in PDAC.

Graphical abstract

Figure 1.

Pancreatitis drives KRAS* bypass. (A and B) GSEA analysis of RNA-seq data comparing KRAS* on and off tumors. PDAC cells from iKPC mice were orthotopically transplanted into C57BL/6 mice, followed by dox water administration to maintain KRAS* activation. After 1 wk, four mice continued receiving dox water (ON) while four were switched to normal water to deactivate KRAS* (OFF). 5 days later, tumors were collected for RNA-seq analysis. (A) Top deregulated gene sets in KRAS* OFF or ON tumors by GSEA analysis. (B) Enrichment plots of IFNα response and TGFβ signaling gene sets. (C) Experimental design for inducing chronic pancreatitis in spontaneous PDAC models using CAE at 100 μg/kg. (D) Kaplan–Meier survival analysis comparing mouse groups with KRAS* expression (on dox), KRAS* depletion (off dox), and KRAS* depletion plus CAE treatment (off dox + CAE). (E) Representative histological images illustrating the time-course analysis of malignant lesions and tumors during pancreatitis-induced tumor relapse. H&E, Mason’s Trichrome staining, and immunohistochemistry (IHC) were performed. (F) Quantification of IHC signal-positive cells using ImageJ. The percentage of relative area was calculated as 100*(positive cell area)/(total cell area). (G) Quantification of TGFβ IHC signal intensity using ImageJ. The optical density was calculated as log(max intensity/mean intensity), where the max intensity is 255 for 8-bit images. Log-rank (Mantel-Cox) test was used for D. Statistical analysis was performed using one-way ANOVA for F and G. The P values: ns, not significant; *, P < 0.05; **, P < 0.01, ***, P < 0.001; ****, P < 0.0001. Error bars represent the median ± SEM. All experimental data was verified in at least two independent experiments.

Figure 2.

Pancreatitis promotes KRASi resistance through TGFβ. (A) Experimental design for inducing pancreatitis in orthotopically transplanted PDAC models. (B) Bioluminescence imaging (BLI) monitoring of tumor formation across different treatment groups: vehicle control + saline + IgG (V), vehicle control + caerulein (CAE, 100 μg/kg) + IgG (C), vehicle control + saline + α-TGFβ neutralizing antibody (250 μg per dose, T), vehicle control + CAE + α-TGFβ neutralizing antibody (CT), MRTX1133 (10 mg/kg, BID) + saline + IgG (M), MRTX1133 + CAE + IgG (MC), MRTX1133 + saline + α-TGFβ neutralizing antibody (MT), and MRTX1133 + CAE + α-TGFβ neutralizing antibody (MCT). (C) Images of collected tumors at humane endpoints. (D) Comparison of tumor weights. (E) Comparison of tumor volumes. (F and G) Histological analysis using H&E (F) and IHC (G) staining to characterize tumor morphology and TAMs. (H) Quantification of IHC signal-positive cells from G using ImageJ. Statistical analysis was performed using one-way ANOVA for D, E, and H. The P values: ns, not significant; *, P < 0.05; **, P < 0.01, ***, P < 0.001; ****, P < 0.0001. Error bars represent the median ± SEM. All experimental data was verified in at least two independent experiments.

Figure S1.

TGFβ pathway activation is required for pancreatitis-induced KRASi resistance. (A) Experimental design for inducing pancreatitis in orthotopically transplanted PDAC models. (B) BLI monitoring tumor formation in comparison groups: MRTX1133 (10 mg/kg, BID) + saline + IgG (M), MRTX1133 + CAE (100 μg/kg) + IgG (MC), and MRTX1133 + CAE + α-TGFβ neutralizing antibody (250 μg per dose, MCT). Collected tumors are shown below. (C and D) Statistical comparison of tumor weight (C) and tumor volume (D) among the three experimental arms. (E) Western blot analysis of canonical TGFβ pathway activation status in PDAC tissues under different treatments. (F) H&E staining of mouse pancreas following various treatments. (G) IHC staining of TGFB1 in KRAS*-expressing tumors and KRAS*-depleted tumors for 5 days from iKPC mice. (H) Quantification of TGFβ signal intensity in G using ImageJ. OD, optical density. (I) Quantification of relative TGFβ-positive area in G using ImageJ. Statistical analysis for C, D, H, and I involved one-way ANOVA. The P values: ns, not significant; *, P < 0.05; **, P < 0.01, ***, P < 0.001; ****, P < 0.0001. Error bars represent the median ± SEM. All experimental data was verified in at least two independent experiments. Source data are available for this figure: SourceData FS1.

Figure 3.

TGFβ signaling determines PDAC sensitivity to KRAS* targeted therapy. (A) Genetic mutation rates of KRAS, TP53, and SMAD4 in PDAC subtypes, including well, moderately, and poorly differentiated (diff’d), and undifferentiated (undiff’d) subtypes. The QCMG PDAC dataset from cBioPortal was used for the study. (B) Comparison of cancer spheroid formation using three distinct iKPC PDAC cells. (C) The inability of TGFβ to induce KRAS*-independent cancer spheroid formation in some iKPC cell lines. (D) Assessment of Cdkn2a and Cdkn2b expression in various iKPC cell lines. (E) Determination of knockdown efficiency for Cdkn2a and Cdkn2b via qRT-PCR. (F and G) Comparative analysis of TGFβ-driven, KRAS*-independent cancer spheroid formation following the knockdown of Cdkn2a or Cdkn2b in two distinct KRAS* bypass-deficient iKPC cell lines. (H) Comparison of cancer spheroid formation from KPC PDAC cells upon treatment with G12Di MRTX1133 (0.3 μM), murine recombinant TGFβ (0.5 ng/ml), and TGFβRi SB505124 (3 μM). (I) Cancer spheroid formation from human PDAC MIA PaCa-2 cells upon treatment with KRAS^G12C inhibitor (G12Ci) ARS-1620 (7.5 μM), human recombinant TGFβ (0.5 ng/ml), and TGFβRi SB505124 (3 μM). (J) Cancer spheroid formation from human PDAC Panc 04.03 upon treatment with G12Di MRTX1133 (0.2 μM), human recombinant TGFβ (0.5 ng/ml), and TGFβRi SB505124 (3 μM). (K) Cancer spheroid formation from human PDAC AsPC-1 upon treatment with G12Di MRTX1133 (0.3 μM), human recombinant TGFβ (0.5 ng/ml), and TGFβRi SB505124 (1 μM). (L) Representative images of spheroids from KRAS*-independent escaper tumor cell lines under treatment with TGFβRi SB505124 (3 μM). The control E5 images in Fig. 3 L and Fig. 7 F were from the same experiment. Statistical analysis was performed using one-way ANOVA for B and E–K. The P values: ns, not significant; *, P < 0.05; **, P < 0.01, ***, P < 0.001; ****, P < 0.0001. Error bars represent the median ± SEM. All experimental data was verified in at least three independent experiments.

Figure 4.

NFAT5 interacts with SMAD3 and SMAD4. (A and B) Western blot analysis to determine the knockdown efficacy of Smad2, Smad3, and Smad4. (C) Examination of TGFβ-driven, KRAS*-independent iKPC cancer spheroid formation after SMADs knockdown compared to the scramble control. (D) Venn diagram illustrating the IP/MS results. Endogenous SMAD2, SMAD3, and SMAD4 were used as baits to pull down proteins in iKPC PDAC cells, with an IgG antibody serving as the negative control. Positive hits are defined as those with an abundance ratio >10 compared with IgG. (E and F) Validation of protein interactions through co-IP/western blot analysis in mouse iKPC PDAC cells using SMADs and NFAT5 as baits. (G–I) Validation of NFAT5–SMADs protein interactions through co-IP/western blot analysis in human MIA PaCa-2 PDAC cells. (J) Cell fractionation followed by pulldown of NFAT5 using α-IgG or α-SMAD4 antibody. WCL: whole cell lysate; Cyt: cytosol fraction; Nuc: nuclear fraction. (K) Analysis of NFAT5 and SMADs interaction under different treatments by co-IP/western blots. All experimental data was verified in at least two independent experiments. Source data are available for this figure: SourceData F4.

Figure 5.

NFAT5 is upregulated in PDAC. (A) Human tissue microarray (TMA) analysis of NFAT5 during pancreatic disease progression. (B) Quantification of histological scores in chronic pancreatitis (CP), PanIN, and PDAC. (C–E) Kaplan-Meier survival analysis of PDAC patients with high or low NFAT5 expression in the TCGA PAAD dataset, including overall survival (OS) analysis in the entire cohort (C), the SMAD4 wildtype (wt) cohort (D), and the SMAD4 mutation or deletion (mut/del) cohort (E). (F) IHC staining of NFAT5 in spontaneous tumors from iKPC mice. (G) Quantification of nuclear NFAT5 staining signal intensity in F using ImageJ. (H) IHC staining of NFAT5 in tumor tissues collected from iKPC mice during pancreatitis-induced tumor relapse. (I) Quantification of nuclear NFAT5 staining signal intensity in H using ImageJ. (J) IHC staining of NFAT5 in transplanted tumors under treatments with KRASi, CAE, and α-TGFβ neutralizing antibody. (K) Quantification of nuclear NFAT5 staining signal intensity in J using ImageJ. One-way ANOVA was used for statistical analysis for G, I, and K; the Chi-square test was performed for B; the Log-rank (Mantel-Cox) test was used for C–E. The P values: ns, not significant; *, P < 0.05; **, P < 0.01, ***, P < 0.001; ****, P < 0.0001. Error bars represent the median ± SEM. All experimental data was verified in at least two independent experiments.

Figure 6.

NFAT5 is essential for TGFβ-driven KRAS* bypass. (A) Nfat5 expression in primary KRAS*-expressing PDAC tumors (iKPC), KRAS*-reactivated escaper tumors (KRAS+ E), and KRAS*-independent escaper tumors (KRAS− E) from iKPC mice (left). The same dataset was reanalyzed to indicate Nfat5 expression in different subtypes of escaper tumors from iKPC mice (right), including classical, hybrid, and QM escapers. (B and C) Knockdown efficiency of Nfat5 in iKPC cells assessed by RT-PCR (B) and western blot (C) analysis. (D and E) Cancer spheroid formation assay comparing Nfat5 knockdown to the vehicle control in KRAS*-expressing iKPC PDAC cells and TGFβ-driven KRAS* bypass. Three different iKPC cell lines were used for the study. (F) Experimental design to assess the anti-tumor effect of Nfat5 knockdown in combination with G12Di in vivo. (G) Comparison of tumor growth between the scramble control and Nfat5 knockdown under treatment with G12Di MRTX1133 (10 mg/kg, QD) or vehicle control. Tumors were collected on day 21. (H) Tumor characterization by H&E staining. (I) Characterization of tumors from G by IHC staining. (J) Quantification of Ki67⁺ cell number per 10× view from I by ImageJ. One-way ANOVA was used for statistical analysis for A, B, D, E, and J; the unpaired, two-tailed t test was used for G at the time point of tumor collection. The P values: ns, not significant; *, P < 0.05; **, P < 0.01, ***, P < 0.001; ****, P < 0.0001. Error bars represent the median ± SEM. All experimental data was verified in at least two independent experiments. Source data are available for this figure: SourceData F6.

Figure 7.

NFAT5 inhibition mitigates KRAS* targeted therapy resistance. (A) Western blot analysis to determine the dose-dependent inhibition of NFAT5 expression by chemical compound KRN2. (B) Comparison of cancer spheroid formation under the treatment of different combinations of dox, TGFβ (0.5 ng/ml), and KRN2 (1 μM) in three distinct iKPC PDAC cell lines. (C) Cancer spheroid formation assay to assess the combination effect of G12Di MRTX1133 (0.03 μM) and KRN2 (1 μM) in KPC PDAC cells. (D) Cancer spheroid formation assay to determine the combination effect of G12Ci ARS-1620 (5 μM) and KRN2 (0.3 μM) in human PDAC MIA PaCa-2 cells. (E) Cancer spheroid formation assay to determine the combination effect of G12Di MRTX1133 (0.3 μM) and KRN2 (0.3 μM) in human PDAC AsPC-1 cells. (F) Comparison of cancer spheroid formation under treatment of DMSO control, TGFβRi SB505124 (3 μM), and KRN2 (1 μM) in three KRAS*-independent escaper tumor cell lines from iKPC mice without CAE treatment. The control E5 images in Fig. 3 L and Fig. 7 F were from the same experiment. (G) Comparison of cancer spheroid formation under treatment of DMSO control, TGFβRi SB505124 (3 μM), and KRN2 (1 μM) in three KRAS*-independent escaper tumor cell lines from CAE-treated iKPC mice. (H) Experimental design to evaluate the anti-tumor effect of KRN2 (3 mg/kg, QD) monotherapy and its combination with G12Di MRTX1133 (10 mg/kg, QD) in vivo. (I) BLI imaging to monitor tumor formation. (J) Kaplan–Meier survival analysis. OS, overall survival. (K) Measurement of mouse body weight along treatments. (L) Tumor characterization by H&E staining. (M) Schematic of the experimental design to assess the combined inhibition of KRAS and NFAT5 in the MIA PaCa-2 orthotopic xenograft model. Tumor-bearing mice were under treatment of vehicle control, G12Ci MRTX849 (100 mg/kg, QD), KRN2 (3 mg/kg, QD), and the combination (combo). (N) Measurement of mouse body weight along treatments. (O) Comparison of tumor weight and size on day 28. (P) Analysis of escaper tumor growth comparing treatment of vehicle control and KRN2 (3 mg/kg, QD). KRAS*-independent escaper tumor cells E725 were transplanted into nude mice subcutaneously. Tumors were collected and imaged on day 28. (Q) Comparison of tumor growth by BLI under treatments: MRTX1133 (10 mg/kg, BID) + saline + vehicle (M), MRTX1133 + CAE (100 μg/kg) + vehicle (MC), and MRTX1133 + CAE + KRN2 (3 mg/kg, QD, MCK). The KPC PDAC cells (1860) were orthotopically transplanted in immunocompetent mice. Tumors were collected on day 21. (R and S) Statistical comparison of tumor volume (R) and tumor weight (S) among the three experimental arms. One-way ANOVA was used for statistical analysis for B–G, R, and S; the unpaired, two-tailed t test was used for O and P at the time point of tumor collection. The P values: ns, not significant; *, P < 0.05; **, P < 0.01, ***, P < 0.001; ****, P < 0.0001. Error bars represent the median ± SEM. All experimental data was verified in at least two independent experiments. Source data are available for this figure: SourceData F7.

Figure 8.

S100A4 is a direct target of the NFAT5–SMADs complex. (A) Summary of RNA-seq analysis to identify candidate targets of the NFAT5–SMADs complex. (B) GSEA analysis to identify the loss of the EMT gene signature after the inhibition or knockdown of NFAT5. (C) Intersection of RNA-seq datasets to identify 99 candidate genes potentially activated by the NFAT5–SMADs complex. (D) Expression profile of gene candidates in primary and escaper PDAC tumor cells from iKPC mice. (E) Comparison of S100a4 expression in primary and escaper PDAC tumors based on KRAS reactivation status (left) and tumor subtypes (right). (F) Summary of ChIP-seq data revealing genes with proximal promoters bound by NFAT5 and SMADs. (G) Schematic representation of the NFAT5–SMADs interaction. (H) Overlapping genes between the 99 candidates from RNA-seq and 2,582 genes from ChIP-seq. (I) IHC staining of S100A4 in tumors during pancreatitis-driven KRAS* bypass and escaper tumors. (J) Quantification of relative S100A4 signal-positive area in I using ImageJ. (K) IHC staining of S100A4 in transplanted tumors under treatments with KRASi, CAE, and α-TGFβ neutralizing antibody. (L) Quantification of relative S100A4 signal-positive area in K using ImageJ. (M) Kaplan-Meier survival analysis of PDAC patients based on high or low S100A4 expression in TCGA PAAD dataset. (N) Expression changes of S100a4 after treatments with dox or TGFβ, following knockdown of Nfat5 or Smad2/3/4. (O) Binding of NFAT5 and SMADs at the S100a4 promoter. (P) NFAT5–SMADs binding comparison at the S100a4 promoter in Nfat5 wildtype and knockdown iKPC cells. (Q and R) Comparison of luciferase activity driven by full length (FL) of or truncated S100a4 promoter. (S) Comparison of S100a4 activation under treatment of NFAT5i KRN2 (1 μM) or TGFβRi SB505124 (3 μM) by luciferase reporter assay under the control of the S100a4 promoter (FL). (T) Western blot analysis of S100A4 expression regulated by TGFβ in iKPC spheroids. (U) Western blot analysis of EMT TF expression after Nfat5 or S100a4 knockdown. Statistical analysis for E, J, L, N, P, R, and S involved one-way ANOVA; the Log-rank (Mantel-Cox) test was used for M. The P values: ns, not significant; *, P < 0.05; **, P < 0.01, ***, P < 0.001; ****, P < 0.0001. Error bars represent the median ± SEM. All experimental data was verified in at least two independent experiments. Source data are available for this figure: SourceData F8.

Figure S2.

Discovery of S100a4 as a direct target of the NFAT5–SMADs complex. (A and B) GSEA to unveil deregulated gene sets by Smad3 and Nfat5 knockdown (A) and KRN2 (B). (C) GSEA to show the shared deregulated genes following Smad3 and Nfat5 knockdown. (D) Overview of DNA binding regions for NFAT5, SMAD2, SMAD3, and SMAD4 in the mouse genome as determined by ChIP-seq. (E) GSEA to indicate genes bound by the NFAT5–SMADs complex (left) and genes exclusively bound by SMADs (right).

Figure S3.

The NFAT5–SMADs complex regulates canonical TGFβ pathway targets. (A) Predicted DNA binding motifs for NFAT5 and SMADs, with predicted binding sites on the S100a4 (S100A4) gene region in the mouse (human) genome. (B) Expression levels of EMT TFs following Nfat5 knockdown or inhibition, and the binding of NFAT5 and SMADs to the DNA regions of EMT TFs. (C) The binding of NFAT5 and SMADs at the Nfat5 promoter.

Figure 9.

S100A4 is required for KRAS* bypass driven by the TGFβ–NFAT5 axis. (A) Knockdown efficiency of S100A4 in iKPC cells by western blot. (B) Examination of pathway activation after knockdown of S100a4 in iKPC cells by western blot. (C) TGFβ-driven, KRAS*-independent cancer spheroid formation comparison between S100a4 wildtype and knockdown in three distinct iKPC cell lines. (D) Tumor growth analysis of subcutaneously transplanted S100a4 wildtype and knockdown (KD) iKPC cells under treatment of vehicle control or G12Di MRTX1133 (10 mg/kg, QD). (E) Tumor characterization by H&E staining. (F and G) Rescue of TGFβ-driven, KRAS*-independent cancer spheroid formation by S100A4 after NFAT5 inhibition (KRN2, 1 μM) (F) and after knockdown of S100a4 and Nfat5 (G). Statistical analysis for C and F involved one-way ANOVA; the unpaired, two-tailed t test was used for G and for D at the time point of tumor collection. The P values: ns, not significant; *, P < 0.05; **, P < 0.01, ***, P < 0.001; ****, P < 0.0001. Error bars represent the median ± SEM. All experimental data was verified in at least two independent experiments. Source data are available for this figure: SourceData F9.

Figure 10.

Macrophages promote KRAS* bypass by providing paracrine S100A4. (A) Single-cell RNA-seq analysis to reveal S100a4 expression in tumors collected from KPC and iKPC mice, treated with MRTX1133 (Ki, 10 mg/kg, BID) or with KRAS off for 5 days. (B) Differential expression of Tgfb1 and S100a4 in mBMDMs compared to iKPC cells. (C) Assessment of Tgfb1 and S100a4 expression in mBMDMs post-treatment with M0 inducer (M-CSF), M1 inducer (LPS + IFNγ), M2 inducer (IL-4), tumor-conditioned medium collected from KPC cells (CM), and tumor-conditioned medium collected from KRAS-inhibited KPC cells (+KRASi CM). (D) S100a4 expression in mBMDMs under treatment with TGFβ (0.5 ng/ml), NFAT5i KRN2 (1 μM), or TGFβRi SB505124 (3 μM). (E) IHC staining of F4/80 and S100A4 in transplanted tumors with wildtype or Nfat5 knockdown after MRTX1133 treatment. (F) Quantification of relative F4/80 signal-positive area in E using ImageJ. (G) Quantification of S100A4 high stroma cell number in E using ImageJ. (H) IHC staining of F4/80 in transplanted tumors post MRTX1133 and KRN2 treatment. (I) Quantification of relative F4/80 signal-positive area in H using ImageJ. (J) TGFβ-driven, KRAS*-independent cancer spheroid formation with or without co-culture of mBMDMs (Mφs, 30,000 cells/well) after S100a4 knockout. (K) KRAS*-independent cancer spheroid formation in co-culture with mBMDMs. (L) KRAS*-independent, Nfat5-knockdown cancer spheroid formation in co-culture with mBMDMs. (M) KRAS*-independent cancer spheroid formation in co-culture with mBMDMs under treatment of S100A4 and TGFβ neutralizing antibodies. The concentrations for IgG isotype control, α-S100A4 antibody and α-TGFβ antibody were 10, 5, and 10 μg/ml, respectively. (N) Overlapping genes between RNA-seq datasets and secretome database. (O) Expression changes of Ccl2 in iKPC cells post TGFβ treatment, after Smad2/3/4 knockdown, and Nfat5 knockdown. (P) NFAT5 and SMADs binding at the Ccl2 promoter. (Q) Ccl2 expression in primary and escaper PDAC tumors based on KRAS reactivation status (left) and tumor subtypes (right). Statistical analysis for C, D, F, G, I, K–M, O, and Q involved one-way ANOVA; the unpaired, two-tailed t test was used for B and J. The P values: ns, not significant; *, P < 0.05; **, P < 0.01, ***, P < 0.001; ****, P < 0.0001. Error bars represent the median ± SEM. All experimental data was verified in at least two independent experiments.

Figure S4.

Schematic representation of intercellular crosstalk promoting KRAS* bypass.