Altered environmental pH triggers a PSC phenotype switch in vitro. First, we aimed to illustrate the impact of environmental pH on the PSC phenotype in the healthy and cancerous pancreas (Figure 1A). We performed an unbiased screening using RNA-Seq of primary murine PSCs cultured at pHe6.6 and pHe7.4 (GSE223205; https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE223205). This pH range can plausibly occur locally in the pancreas stroma and along the course of PDAC (3). The replicates in each group were homogeneous according to principal component analysis (Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.170928DS1). We found a total of 1,769 genes differentially regulated between the 2 groups (log2FC > 0.58, P < 0.05), with 804 genes being upregulated at pHe7.4 and 965 genes being upregulated at pHe6.6 (Supplemental Figure 1B and Supplemental Table 1). A gene set enrichment analysis (GSEA) (Figure 1B and Supplemental Figure 1, C and D) revealed that cell proliferation, communication, adhesion, and cell cycle pathways are markedly upregulated in PSCs cultured at pHe7.4. In contrast, immune response–related pathways are upregulated at pHe6.6. Next, we compared our results with gene expression signatures of published PSC-derived CAF subpopulations (7, 22) (Figure 1C). Genes from iCAFs were highly expressed at pHe6.6, and PSCs cultured at pHe7.4 highly expressed myCAF marker genes (Supplemental Table 3). These findings suggest that PSCs isolated from the healthy pancreas with a physiological history of intermittent acidification have an immunomodulatory phenotype when kept in an acidic environment. However, they acquire a myofibroblastic phenotype upon removal of the extracellular acidity when kept at pHe 7.0 and 7.4. Therefore, we will designate them as iPSCs and myPSCs, respectively.

Environmental alkalization induces myofibroblastic PSC differentiation and proliferation. (A) Concept of the working hypothesis. In the healthy pancreas, the marked HCO3– secretion upon each meal results in a distinct stromal acidification, keeping the PSCs in a quiescent nonfibrotic phenotype. Upon malignant transformation in early PDAC (PanIN), ductal secretion decreases, resulting in a relief of the intermittent acidity ( = relative stromal alkalization), leading to a myofibroblastic PSC differentiation. Acidic → alkaline pHe is depicted by yellow → purple colors. (B) Hallmark gene set enrichment analysis (GSEA) of RNA-Seq data from PSCs cultured at pHe7.4 versus pHe6.6 shows the top 5 differentially regulated pathways (n/N = 3/3). (C) A heatmap of RNA-Seq expression mean Z scores computed for published signature genes of immunomodulatory CAFs (left) and myofibroblastic CAFs (right). The gene (rows) Z scores for pHe6.6 and pHe7.4 are color coded. Dark green indicates higher expression Z scores (n/N = 3/3). (D) Immunocytochemistry images of PSCs cultured at pHe6.6 or pHe7.4. The activation marker αSMA (yellow) and the general PSC marker vimentin (magenta) are labeled. Nuclei are stained with DAPI (cyan). Scale bar: 50 μm. (E) Cell areas multiplied by αSMA intensity is taken as a readout for the myofibroblastic PSC phenotype. It is plotted as a function of pHe. Mean values are shown as n ≥ 142 from N = 3 mice. Half-maximum (EC50) pHe-dependent PSC activation occurs at pHe7.0. Note the logarithmic scale of the ordinate. (F) Representative Western blot (top) of p53 and GAPDH of PSCs cultured at pHe6.6 or pHe7.4, with subsequent quantification (bottom) (n/N = 3/3). (G) Representative cell cycle histogram of PSCs cultured at pHe6.6 (red) and pHe7.4 (blue), assessed by flow cytometry with propidium iodide (PI) staining. Cell populations at different stages of the cell cycle are indicated by arrows. (H) The bar chart depicts the percentage of PSCs at the G0/G1 phase of the cell cycle when cultured at pHe 6.6 (red) and pHe7.4 (blue). Data points are npHe6.6 = 3 and npHe7.4 = 5 measurements from n = 3 individual mice. Statistical tests in F and H were performed with 2-tailed unpaired Student’s t tests. A was created with BioRender.com.

To confirm the pHe-dependent phenotype switch of PSCs, we cultured freshly isolated PSCs from WT mice in media with a pH range from pHe6.0 to pHe7.4 for 72 hours and then quantified αSMA expression as a derivative of the myofibroblastic phenotype. For a more robust readout of myofibroblastic PSC phenotype and αSMA quantity, we multiplied cell area with αSMA intensity. As expected from the RNA-Seq data, the myCAF phenotype heavily relied on the environmental pH (Figure 1, D and E). The highest expression of αSMA was found at pHe7.4 (pHe6.0: 1,166 ± 146 a.u., pHe6.6: 8,090 ± 1,137 a.u., pHe7.0: 140,365 ± 17,508 a.u., pHe7.4: 323,756 ± 30,015 a.u.; n ≥ 142 cells from N = 3 mice). Western blot experiments comparing αSMA normalized to vimentin protein levels in PSCs cultured at pHe6.6 versus pHe7.4 recapitulated the immunocytochemistry data (αSMA/vimentin protein expression) (pHe6.6: 0.11 ± 0.08 a.u., pHe7.4: 1.23 ± 0.18 a.u., P = 0.015, n = 4 samples from n = 4 mice) (Supplemental Figure 2, A and B). The percentage of viable cells in the population was reduced at pHe6.0 (12% ± 2%) and pHe6.3 (31% ± 5%), while it was hardly affected at pHe > 6.5 (pHe6.6: 83% ± 2%; pHe7.0: 96% ± 1%; pHe7.4: 94% ± 2%, cells from n = 5 fields of view of n = 5 mice, P < 0.0001) (Supplemental Figure 2, C and D). Increased cell proliferation of myPSCs was underlined by the fact that p53 protein was more expressed at pHe6.6 than at pHe7.4 (Figure 1F) (relative p53 expression to GAPDH: pHe6.6: 0,04 ± 0.01, pHe7.4: 0.01 ± 0.01;lysates from N = 3 mice each, P = 0.038). Cell cycle progression was also diminished at pHe6.6 compared with pHe7.4 (Figure 1, G and H; percentage of PSCs in G0/G1 after 72 hours: pHe6.6: 93 ± 1%; pHe7.4: 80 ± 2%; n ≥ 3 measurements from N = 3 mice, P = 0.017). Taken together, environmental pH was crucial in tuning the differentiation of PSCs. They acquired iPSC and myPSC phenotypes at pHe6.6 and pHe7.4, respectively.

NHE1 is a key regulator of PSC pH homeostasis. Since the myofibroblastic phenotype switch occurs in PSCs at pHe7.4 but not at pHe6.6, we hypothesized that pH sensory and regulatory ion transporters are involved in the process. Thus, we evaluated our RNA-Seq data regarding the respective gene ontology term (GO:0015075; https://www.ebi.ac.uk/QuickGO/term/GO:0015075) (Figure 2A). We found 43 transporter genes differentially upregulated at pHe7.4 and 44 at pHe6.6 as well as a number of genes that were highly expressed in both groups (Supplemental Table 3). From these, we selected and further validated the expression of 5 transporter genes (Figure 2B) and 10 ion channel genes (Supplemental Figure 3) and compared them with the housekeeper genes Ywhaz and 18s rRNA by means of quantitative PCR (qPCR). We used freshly isolated PSCs as well as PSCs cultivated at pHe7.4 or pHe6.6 for 5 days (Figure 2B; n = 6 from N = 3 mice). Notably, the Na+/H+ exchanger NHE1 (encoded by SLC9A1) has more mRNA expression in PSCs than most other pH-regulatory proteins under all investigated conditions. Because of its pathophysiological relevance in cancer (23) and good druggability with small molecules such as cariporide (24), we decided to further investigate this protein.

NHE1-mediated pH recovery is inhibited by cariporide in PSCs. (A) Volcano plot analysis of ion transporter genes (GO:0015075) derived from the RNA-Seq data of PSCs cultured at pHe7.4 and pHe6.6 (n/N = 3/3). Genes indicated by red (n = 43) and blue (n = 44) dots highlighted in rectangles are upregulated at pHe6.6 and pHe7.4, respectively. (B) Subsequent qPCR validation of ion transporter gene expression levels by the 2–ΔCT method compared with the housekeeping genes Ywhaz and 18s rRNA. Bar charts show mean expression of genes from freshly isolated quiescent PSCs (0 hours, white) and PSCs cultured at pHe7.4 (blue) or pHe6.6 (red) (n/N = 6/3). (C) Representative immunofluorescence images showing the cellular localization of NHE1 in freshly isolated quiescent PSCs (0 hours), PSCs cultured at pHe7.4 or pHe6.6 for 120 hours, or vehicle-treated KPfC-derived PSCs (PDAC-PSC) (NHE1: magenta; DAPI: blue). Scale bar: 10 μm. (D) NHE1 Western blots are shown, with the top bands at 100 kDa corresponding to the glycosylated NHE1, whereas bands with lower molecular weight (80 kDa) correspond to the unglycosylated NHE1. Lysates are from N = 3 mice each. (E) pHi recordings of WT PSCs cultured at pHe6.6 (left) and pHe7.4 (middle) or KPfC-derived CAFs (cultured at pHe7.4; right). pHi was acidified temporarily by applying the NH4+ prepulse (*) technique. NHE1-independent pHi recovery starts when pHi has reached its minimum in the presence of the Na+-free solution (“0 Na+”). NHE1-dependent pHi recovery can be observed in the last step (Ctrl) of the experiment when cariporide was added to the Na+-containing superfusion as indicated. Lines show mean pHi of npH6.6 = 35, npH6.6+CARI = 39, npH7.4 = 45, npH7.4+CARI = 68, nPDAC-CAF = 11, and nPDAC-CAF+CARI = 22 cells from N = 3 mice each. (F) Quantification of resting pHi of PSCs cultured at pHe6.6 (red) or pHe7.4 (blue) and CAFs (purple), respectively, derived from E (npH6.6 = 74, npH7.4 = 113, and nPDAC-CAF+CARI = 41 cells from N = 3 mice each). (G) Scatter plot depicts the rate of Na+-independent pHi recovery of PSCs cultured at pHe6.6 (red), pHe7.4 (blue), or CAFs (purple), derived from E (n/N see F). (H) Comparison of the rate of Na+-dependent pHi recovery of WT PSCs cultured at pHe6.6 (red) or pHe7.4 (blue), or CAFs (purple) as explained in E (n/N see E). Statistical tests in F–H were performed with 1-way ANOVA with Tukey’s post hoc test.

To explore the subcellular localization of NHE1 in vitro, we performed immunocytochemistry. Freshly isolated PSCs derived from healthy WT mice — i.e., quiescent PSCs — expressed NHE1 predominantly intracellularly (Figure 2C). However, NHE1 translocated to the plasma membrane of PSCs from healthy WT mice already after 24 hours of culturing in either pHe6.6 (iPSC) or pHe7.4 (myPSC). Similarly, CAFs freshly isolated from murine PDAC (KPfC) expressed NHE1 also mainly in their membrane.

Western blot analysis confirms our immunocytochemistry results. NHE1 was highly expressed in both freshly isolated and cultured PSCs as well as in PDAC-derived CAFs (Figure 2D). However, there were distinct differences: freshly isolated PSCs expressed the small (~80 kDa) intracellular, nonglycosylated NHE1 protein, whereas PSCs cultured for 120 hours and tumor-derived CAFs expressed the larger (~100 kDa) glycosylated, plasma membrane-residing NHE1.

We next focused on NHE1 functionality. We revealed its activity by measuring the pHi recovery following an NH4+ prepulse (25) (Figure 2E). When PSCs from WT mice are cultured in vitro for 72 hours, the pHi closely follows the pHe (Figure 2F). PSCs kept at pHe6.6 have a pHi of pHi 6.60 ± 0.01 (n cells/N mice = 74/3). In contrast, the pHi of PSCs cultured at pHe7.4 was pHi 7.10 ± 0.02 (n cells/N mice = 113/3). Similarly, the pHi of CAFs freshly isolated from KPfC-mice (cultured at pHe7.4) was pHi 7.11 ± 0.01 (n cells/N mice = 41/3). The NH4+ prepulse protocol revealed almost no Na+-independent pH recovery in any of these cells (Figure 2G) (pHe6.6: 0.040 ± 0.004 pH unit/min, n cells/N mice = 156/5; pHe7.4: 0.060 ± 0.003 pH unit/min, n cells/N mice = 135/5; PDAC-CAF: 0.015 ± 0.005 pH unit/min, n cells/N mice = 41/3). In contrast, pHi immediately recovered upon readdition of extracellular Na+. The rate of pHi recovery was faster in PSCs cultured at pHe7.4 than in PSCs cultured at pHe6.6 (Figure 2H) (pHe7.4: 0.38 ± 0.05 pH unit/min, n cells/N mice = 45/3; pHe6.6: 0.11 ± 0.01 pH unit/min, n cells/N mice = 35/3; P < 0.0001). The pH recovery was primarily due to the action of NHE1, as inhibition of NHE1 with cariporide (10μM) almost completely abolished pHi recovery (pHe6.6: 0.019 ± 0.003 pH unit/min, n cells/N mice = 39/3; pHe7.4: 0.061 ± 0.003 pH unit/min, n cells/N mice = 68/3). CAFs freshly isolated from tumor-bearing KPfC-mice also expressed a highly active NHE1, with the resting pHi and the rate of NHE1-dependent recovery being similar to PSCs cultured at pHe7.4 (vehicle-treated KPfC-mice: pH recovery: 0.44 ± 0.09 pH unit/min, n cells/N mice = 11/3). This also applies to CAFs isolated from another PDAC mouse model that is driven by κB-Ras1 deficiency (26) (Supplemental Figure 4, A and B, npH6.6 = 111, npH7.4 = 94 cells from n = 5 mice).

RNA-Seq and qPCR analyses indicate that the Na+- HCO3– cotransporter NBC1 (Slc4a4) and the monocarboxylate transporter 4 (Slc16a3) and numerous other transporters are also expressed in PSCs that could complement and/or compensate for blocked NHE1 activity. To rule out this possibility, we performed pHi measurements in CAFs derived from vehicle-treated KPfC-mice in a CO2/HCO3– buffered environment where HCO3– transporters are active (Supplemental Figure 4C). We found that Na+-dependent pH recovery was primarily due to the action of NHE1 (Supplemental Figure 4, D and E; pH recovery: 0.68 ± 0.04 pH unit/min, n cells/N mice = 30/3). In summary, NHE1 acted as the major acid extruder in PSCs and CAFs, and its inhibition with cariporide led to intracellular acidification of in vitro cultured PSCs and ex vivo PDAC-derived CAFs.

Lack of acidity facilitates YAP-1–mediated mechanotransduction and myofibroblastic PSC differentiation. Next, we aimed to get mechanistic insight into how cellular pH alkalinization results in the PSC phenotype switch. It is known that PSCs are quite susceptible to mechanical stimuli and express αSMA and, thus, exhibit a myofibroblastic phenotype predominantly on a rigid substrate (27, 28). We hypothesized that this major pathway of myPSC differentiation may be influenced by environmental pH. Therefore, we plated freshly isolated WT murine PSCs onto hydrogels with varying stiffness (Figure 3A) and cultured them at pHe7.4 or pHe6.6 for 72 hours. Cells were larger and had more αSMA (derived from cell area × αSMA intensity) when plated on the 1 GPa glass substrate than on the 11 kPa hydrogel surface (Figure 3B) (1 GPa: 335,493 ± 28,248 a.u., n cells/N mice = 170/3; 11 kPa: 88,304 ± 12,031 a.u., n cells/N mice = 65/3; P < 0.0001). This difference was not seen when culturing PSCs at pHe6.6 (1 GPa: 35,722 ± 7,685 a.u., n cells/N mice = 100/3; 11 kPa:39,577 ± 7,214 a.u., n cells/N mice = 51/3; P < 0.0001; Figure 3C). From this, we concluded that the myPSC but not the iPSC phenotype (PSCs cultured at pHe7.4 and pHe6.6, respectively) relied on substrate stiffness at least with respect to cell size and αSMA positivity.

PSC mechanotransduction mediated by YAP1 is inhibited at acidic pHe. (A) Immunocytochemistry images of PSCs cultured on substrates of different stiffnesses at pHe6.6 (top) and pHe7.4 (bottom). The activation marker αSMA (yellow) and the general stellate cell marker vimentin (magenta), as well as nuclei (cyan), are labeled. Scale bar: 50 μm. (B) Cell area multiplied with mean αSMA fluorescence intensity was taken as a readout of myofibroblastic PSC phenotype on hydrogels with 11 kPa stiffness or on glass (1 GPa). n11kPa = 65, n1GPa = 170 cells from N = 3 mice. (C) Representative immunofluorescence images of YAP1 in PSCs (green) under the indicated cell culture conditions. YAP1, when translocated from the cytosol (#) into the nucleus (*), initiates transcription. Scale bar: 50 μm. (D) The ratio of nuclear/cytosolic YAP1 fluorescence intensity was determined as a readout of YAP1-mediated signal transduction. n11kPa-pH6.6 = 68, n11kPa-pH7.4 = 51, n1GPa-pH6.6 = 43, n1GPa+pH7.4 = 63 cells from N = 3 mice each. Statistical tests in B and D were performed with 1-way ANOVA with Tukey’s post hoc test.

We then investigated whether pHe influenced mechanosignaling via YAP1, a well-characterized transcription factor in PSCs (29). Immunocytochemistry (Figure 3D) revealed that the nuclear/cytosolic ratio of YAP1 was higher in pHe7.4 than in pHe6.6, indicating increased YAP1-mediated transcription in myPSCs as compared with iPSCs (Figure 3D) (1 GPa, pHe7.4: 1.5 ± 0.1, n cells/N mice = 63/3; pHe6.6: 1.1 ± 0.1, n cells/N mice = 43/3; P = 0.018). Overall, these data point out that the YAP1-mediated mechanosignaling was pH dependent and not utilized in iPSCs kept in an acidic environment.

Acidic pHe partially alters myPSC phenotype only in the presence of cariporide. PSC phenotypes depend on the given stimulus (7, 30). Our results above indicate that environmental pH was crucial in PSC differentiation. However, in our initial experimental setting, we investigated a multitude of different pH-independent pathways by seeding cells from their native tissue environment onto plastic and into a cell culture medium. To test whether the sole removal of acidity was sufficient to induce differentiation of iPSC to myPSCs (e.g., through YAP-1–mediated mechanosignaling as shown above), we changed the medium pH of pHe6.6-cultured iPSCs, either to pHe6.6 (Resting) or to pHe7.4 (PanIN-like) after 72 hours (Figure 4A). Indeed, after just 3 days of culture at pHe7.4, PSCs differentiated from the iPSC to the myPSC phenotype as indicated by the drastic rise of αSMA expression (Figure 4B) (pHe6.6 → pHe6.6 [resting]: 35,722 ± 7685 a.u., n cells/N mice = 100/3; pHe6.6 → pHe7.4 [PanIN-like]: 1,414,669 ± 222,374 a.u., n cells/N mice = 64/3; P < 0.0001). Hence, we conclude that iPSCs cultured at pHe6.6 could further differentiate into myPSCs, with extracellular alkalinization (i.e., relief of extracellular acidity) being a sufficient stimulus.

The myofibroblastic phenotype of activated PSCs is partially reversed by cariporide but not by acidic pHe alone. (A) After culturing PSCs at pHe6.6 for 72 hours, medium was changed to pHe6.6 (Resting) or to pHe7.4 (PanIN-like) for another 72 hours. Lastly, the medium of pHe7.4-incubated cells was reacidified to pHe6.6 for another 72 hours (PDAC-like). Representative images of PSCs stained for αSMA (yellow), vimentin (magenta), and DAPI (cyan) are shown below each condition. Scale bar: 50 μm. (B) Scatter plot shows cell areas multiplied by αSMA fluorescence staining intensity (logarithmic scale) under conditions described in A. nResting = 100, nPanIN-like = 64, and nPDAC-like = 94; N = 3 mice. (C) Immunocytochemistry of PDAC-like PSCs in the absence (left) or presence (right) of cariporide (CARI). Scale bar: 50 μm. (D) Scatter plot shows cell areas multiplied by αSMA fluorescence intensity (logarithmic scale) under conditions described in C. nPDAC-like = 74, nPDAC-like+CARI = 74; N = 3 mice. (E) Intracellular pH measurements of PSCs where culture medium is reacidified after activation (pHe7.4 → pHe6.6, PDAC-like). Intracellular pH was acidified temporarily by applying the NH4+ prepulse technique, as shown in Figure 2. NHE1-dependent pH recovery can be observed when cells are superfused with Na+-containing solution (Ctrl) without (black) or with cariporide (red). Lines indicate the mean pHi of nPDAC-like = 35, nPDAC-like+CARI = 79 cells from N = 3 mice. (F) Comparison of the rate of Na+-dependent recovery of PSCs in the absence or presence of cariporide derived from E. (G) Illustration of extended working hypothesis. In manifest PDAC, acidic pHe fails to acidify pHi because of NHE1-mediated H+ extrusion (left). Therefore, PSCs and CAFs remain myofibroblastic, ultimately promoting tumor desmoplasia. However, upon NHE1 inhibition with cariporide, PSCs and CAFs fail to counterbalance the acid stress, which disrupts the myofibroblastic phenotype (right). Statistical comparison in B was performed with 1-way ANOVA with Tukey’s post hoc test, whereas in D and F, statistical comparisons were performed with 2-tailed unpaired Student’s t tests. G was created with BioRender.com.

We next tested whether PSCs (PanIN-like) can be reprogrammed from the myPSC to the iPSC phenotype by simply changing back the environmental pH to pHe6.6 (PDAC-like). This was not the case (Figure 4A). PSCs retained their myofibroblastic phenotype since αSMA positivity and cell size do not change (Figure 4B) (pHe7.4 → pHe6.6 [PDAC-like]: 1,421,502 ± 146,424 a.u., n cells/N mice = 94/3; P = 0.45). We reasoned that myofibroblastic PSCs retained their phenotype because of their pronounced NHE1 activity; reacidifying the environment alone would not alter pHi anymore because cells could get rid of excess H+ rapidly through NHE1. To test this hypothesis, we applied cariporide with environmental reacidification (Figure 4C). In this “PDAC-like” pH shift, application of cariporide indeed affected the myPSC phenotype (Figure 4D) (PDAC-like+CARI: 2,575,631 ± 282,664 a.u., n cells/N mice = 74/3; PDAC-like: 5,421,677 ± 557,953 a.u., n cells/N mice = 74/3; P < 0.0001). This interpretation was supported by pHi measurements (Figure 4E). NHE1-mediated pHi recovery also occurred at a high rate in PSCs whose culture medium was reacidified after activation (pHe7.4 → pHe6.6) (Figure 4F) (Na+-dependent pH recovery: control: 0.56 ± 0.07 pH unit/min, n cells/N mice = 35/3; cariporide: 0.14 ± 0.01 pH unit/min, n cells/N mice = 79/3; P < 0.0001). These findings indicate that the NHE1 function maintained the myofibroblastic PSC phenotype in a harsh acidic tumor environment, while NHE1 inhibition partially disrupted it (Figure 4G).

Adjuvant PDAC therapy with an NHE1 inhibitor reduces desmoplasia. If the inhibition of NHE1 decreased the myofibroblastic nature of PSCs, this would become evident by a reduced fibrosis in pancreatic cancer. We tested this idea in pancreatic tumor–bearing KPfC mice that harbored heterozygous loss of p53 and conditionally expressed mutant K-Ras (genotype Kraswt/LSL–G12D Tp53fl/+ PDX1-Cre+) (31, 32). Cariporide was given as an adjuvant drug complementing the standard therapy with gemcitabine (28, 29). To mimic the clinical situation of patients with PDAC who usually suffer from a late diagnosis and therapy initiation, we started to apply cariporide at week 20. At that time point, most KPfC mice have already developed manifest PDAC (33). We treated the mice with 3 mg/kg cariporide i.p. daily for 1 month and initiated chemotherapy with gemcitabine at the end of this period (Figure 5A).

NHE1 inhibitor treatment leads to reduced desmoplastic reaction in murine PDAC. (A) Schematic representation of the 4-week–long treatment protocol of KPfC mice. Treatment started at the age of week 20. Cariporide was applied daily (1/D), and gemcitabine (100 mg/kg i.p.) was coinjected with cariporide (3 mg/kg i.p.) on the days indicated by the arrows. (B) Total pancreas volume of KPfC mice after gemcitabine (GEM) or cariporide (CARI) monotherapy or gemcitabine + cariporide (GEM+CARI) combined chemotherapy. Inlet demonstrates that pancreas volume was measured via volume displacement. Data points depict individual pancreata; NVehicle = 11, NGEM = 9, NCARI = 11, NGEM+CARI = 11. (C) Relative tumor area in histological KPfC tissue sections was obtained by dividing total tumor area by total tissue area after H&E staining. Data points depict individual pancreata; NVehicle = 11, NGEM = 9, NCARI = 11, NGEM+CARI = 11. (D) Representative images of PDAC nodes (marked with *) after H&E and Sirius red stainings. The degree of fibrosis correlates with the area of Sirius red+ (marked in red, #) tissue neighboring the cancerous tissue. Scale bar: 100 μm. (E) Relative tumor fibrosis of each Sirius red–stained KPfC tissue section was determined by dividing the summed area of fibrosis within every tumor node (sum of thresholded black areas in every node in the inlet) by the summed area of the tumor nodes. Data points depict individual pancreata; NVehicle = 11, NGEM = 9, NCARI = 11, NGEM+CARI = 12. (F) To obtain the fibrosis per tumor node, the area of Sirius red+ fibrosis (black thresholded area in the inlet) was divided by the total area of the respective tumor node. Data points depict individual tumor nodes; nVehicle = 400, nGEM = 239, nCARI = 279, nGEM+CARI = 476. Data and statistical comparison in B, C, and F are shown as median ± 95% CI with Kruskal-Wallis statistical test with Dunn’s post hoc test, and in E as mean ± SEM with 1-way ANOVA with Tukey’s post hoc test. Inlets for A and B were created with BioRender.com.

The macroscopic total volume of the pancreata (Figure 5B) and microscopic relative area of tumor lesions normalized to total tissue area (Figure 5C) did not differ between the gemcitabine + cariporide double treatment group and the vehicle-treated group (volume: gemcitabine + cariporide: 0.4 mL, 95% CI, 0.3–0.7 mL; P = 0.84, vehicle: 0.4 mL, 95% CI, 0.3–0.7 mL; relative tumor area: gemcitabine + cariporide: 35%, 95% CI, 32–50%, n = 12 mice; vehicle: 41%, 95% CI, 34–54%, n = 10 mice; P = 0.78). However, the histochemical analysis of fibrosis by means of Sirius red staining revealed a clear difference (Figure 5D). The total area of fibrosis in tumors relative to total tumor area was reduced by ~ 30% in mice treated with gemcitabine + cariporide compared with vehicle treatment (Figure 5E; gemcitabine + cariporide: 31 ± 4%, n = 12 mice; vehicle: 45 ± 4%, n = 11 mice; P = 0.03). To ensure that statistics were not biased by a few individual tumor nodes that were very large in size, we investigated the extent of fibrosis in each individual tumor node in every tissue section (Figure 5F). We confirmed the antifibrotic effect of cariporide. Cariporide decreased fibrosis when compared with vehicle treatment (cariporide: 34%, 95% CI, 31%–37%, n individual tumor nodes/N mice = 279/11; vehicle: 40%, 95% CI, 37%–42%, n individual tumor nodes/N mice = 400/11, P = 0.0005). Moreover, gemcitabine + cariporide treatment decreased fibrosis more than gemcitabine alone (gemcitabine + cariporide: 31%, 95% CI, 27%–34%, n individual tumor nodes/N mice = 476/12; gemcitabine: 38%, 95% CI, 35%–41%, n individual tumor nodes/N mice = 239/9 P = 0.0001). These findings point out that NHE1 inhibition was an effective adjuvant therapeutic strategy to counter excess fibrosis in PDAC. The following sections delineate whether the decrease in fibrosis was due to a direct effect on PSCs or, rather, an indirect effect mediated by other cell types such as immune cells.

NHE1 orchestrates CAF activation in PDAC. We reasoned that the decreased fibrosis in the gemcitabine + cariporide treatment group was due to a decreased myCAF phenotype due to NHE1 inhibition. To ascertain whether PSCs express NHE1 in PDAC, we performed IHC (Figure 6A). We found that NHE1 was ubiquitously expressed in cell membranes in pancreatic cancer (Supplemental Figure 6). Indeed, it was expressed in αSMA+ CAFs (Figure 6B) but also CK19+ PDAC and ductal cells as well as in immune cells.

NHE1 orchestrates PDAC-derived CAF activation. (A) Representative H&E and IHC images of healthy and tumorous ducts. The colors of the IHC image indicate cell nuclei stained with NHE1 (magenta), αSMA+ PSCs and CAFs (yellow), DAPI (cyan), and CK19+ ductal or tumor cells (green). Scale bar: 40 μm. (B) αSMA+ cells (*) from A are depicted with higher magnification, which are also NHE1+. Scale bar: 20 μm (C) Immunocytochemistry of CAFs derived from KPfC mice after a 1-month treatment with vehicle (top) or gemcitabine + cariporide (bottom). Myofibroblast marker αSMA (yellow), the general mesenchymal marker vimentin (magenta), and nuclei (cyan) are labeled. Scale bar: 20 μm. (D) KPfC-derived CAF activation after therapy was assessed by multiplying cell area with the fluorescence intensity of αSMA. nVehicle = 61, nGEM = 59, nCARI = 63, nGEM+CARI = 70 from N ≥ 3 mice. Note the logarithmic scale of the ordinate. (E) Trajectories of migrating KPfC-derived CAFs are shown by individual black lines. The treatment of the respective mice is indicated. Trajectories of the treatment groups are always normalized to common starting points. The radii of the orange circles highlight the mean translocation of cells in each population. Scale bar: 20 μm. (F) Mean cell migration velocities of individual CAFs were calculated from the trajectories in C. nVehicle = 30, nGEM = 30, nCARI = 19, nGEM+CARI = 40 cells from N ≥ 3 mice. Statistical tests in D and F were performed with 1-way ANOVA with Tukey’s post hoc test.

To investigate whether the myCAF phenotype was affected during gemcitabine + cariporide treatment, we isolated CAFs from KPfC mice after the treatments outlined in Figure 5A and immediately evaluated their phenotype from αSMA staining fluorescence intensity and cell size (Figure 6C). To verify the fibroblastic nature of the isolated cells, we showed by Western blot that only KPfC-derived tumor cells but not CAFs expressd a high level of mutated KRas G12D (Supplemental Figure 7). We found that CAFs derived from KPfC mice treated with gemcitabine + cariporide were markedly less myofibroblastic than those isolated from mice receiving a gemcitabine monotherapy (Figure 6D) (gemcitabine + cariporide; 592,059 ± 165,706 a.u., n cells/N mice = 70/4; gemcitabine: 7,339,572 ± 1,609,547 a.u., n cells/N mice = 59/3; P = 0.0023). When assessing cell area and αSMA fluorescence intensity separately (Supplemental Figure 5, A and B), we observed a decrease in cell area in gemcitabine + cariporide–treated CAFs compared with gemcitabine treatment (vehicle: 3,822 ± 412 μm2, n cells/N mice = 61/3; cariporide: 2,602 ± 255 μm2, n cells/N mice = 63/3; P = 0.038; gemcitabine: 3,308 ± 329 μm2, n cells/N mice = 59/3; gemcitabine + cariporide; 2,006 ± 245 μm2, n cells/N mice = 70/4; P < 0.0001). This decrease in cell area could be attributed to NHE1 inhibition and may be a reflection of a reduced cell volume (34).

Finally, as a global readout of CAF function, we assessed cell migration after the treatments (Figure 6E). CAFs derived from KPfC mice treated with gemcitabine + cariporide migrated more slowly than CAFs isolated from gemcitabine-treated mice (Figure 6F) (gemcitabine + cariporide: 0.05 ± 0.01 μm/min, n cells/N mice = 40/4; gemcitabine: 0.1 ± 0.01 μm/min, n cells/N mice = 30/3; P = 0.0006). We also recapitulated this finding by exposing untreated κB-Ras–deficient PDAC-derived CAFs to cariporide (Supplemental Figure 5C). In these cells, migration was also inhibited in the presence of 10 μM cariporide (Supplemental Figure 5D) (cariporide: 0.08 μm/min ± 0.01, n cells/N mice = 30/3; control: 0.11 ± 0.01 μm/min, n cells/N mice = 43/5; P = 0.025). To summarize, inhibition of the Na+/H+ exchanger NHE1 in WT and pancreatic tumor–bearing mice shifted CAF differentiation toward a less myofibroblastic phenotype.

NHE1 inhibition enhances lymphocytic immune infiltration in KPfC mice. Above, we focused on NHE1 in PSCs and CAFs. However, it is known that a wide range of other cells, including PDAC cells, lymphocytes, and neutrophils, also express NHE1, as observable from Supplemental Figure 6 (18, 35–37). We, therefore, followed up the relevance of NHE1 in immune cell infiltration more closely after observing that the primarily periodic acid–Schiff+ (PAS+) infiltrates in vehicle- or gemcitabine-treated cohorts shifted to PAS– ones in the cariporide or gemcitabine + cariporide treatment groups (Figure 7A and Supplemental Figure 8). Furthermore, the immune cell infiltration in cariporide-treated animals was often accompanied by a disruption of the architecture of tumor foci, which may indicate a more effective immune response. These observations suggest that NHE1 inhibition shifts the immune cell infiltrate from a largely innate immune cell–rich one to a more lymphocytic infiltration.

Lymphocyte/neutrophil ratio increases upon NHE1 inhibition in tumor sections of KPfC mice. (A) H&E (left) and PAS-stained KPfC mouse tissue sections after vehicle and gemcitabine + cariporide (GEM+CARI) therapy. Cells of innate immunity, such as neutrophils (arrows), utilize glycogen and are thus PAS+ (purple), in contrast to, for example, lymphocytes. Scale bar: 50 μm. (B) Representative IHC images stained for Ly6G+ neutrophils (magenta, arrows on left image), CD3+ lymphocytes (yellow, arrows on the right image), and nuclei with DAPI (cyan). Scale bar: 50 μm. (C) CD3/Ly6G ratio was assessed by dividing the number of CD3+ cells by the number of Ly6G+ cells in every tumor node. Data points depict the mean CD3/Ly6G ratio derived from each tumor node in individual mice; NVehicle = 10, NGEM = 9, NCARI = 10, NGEM+CARI = 11 mice. (D) To obtain the CD3/Ly6G ratio per tumor node, the number of CD3+ cells was divided by the respective number of Ly6G+ cells in each tumor node. Data points depict individual tumor nodes; nVehicle = 386, nGEM = 276, nCARI = 301, nGEM+CARI = 398. Data and statistical comparison in D and E are represented as median ± 95% CI using Kruskal-Wallis statistical test with Dunn’s post hoc test.

To confirm this idea, we further characterized each PDAC sample quantitatively with CD3 and Ly6G IHC staining, labeling T lymphocytes and neutrophils, respectively (Figure 7B). We focused our evaluation on the CD3/Ly6G ratio in PDAC lesions, specifying that a high lymphocyte/neutrophil ratio has a favorable prognostic value for patient survival (38). Indeed, when comparing the number of CD3+ and Ly6G+ cells in each tumor section, we found that the CD3/Ly6G ratio (Figure 7C) had increased by approximately 4 times in the gemcitabine + cariporide group as compared with the vehicle-treated group (gemcitabine + cariporide: 8.5, 95% CI, 4.9–10.6, n = 11 mice; vehicle: 2.2, 95% CI, 0.9–5.2, n = 9 mice; P = 0.03). Upon assessing the CD3/Ly6G ratio of each individual tumor node (Figure 7D), we found it to be 2.5-fold higher in the gemcitabine + cariporide–treated than in the gemcitabine-treated group (gemcitabine + cariporide: 2.6, 95% CI, 2.1–3.1, n individual tumor nodes/N mice = 398/11; gemcitabine: 1.1, 95% CI, 0.9–1.5, n individual tumor nodes/N mice = 276/9; P < 0.0001). We also noted that T cells not only accumulated in the periphery of individual tumor foci but also penetrated into the depth of the cancer tissue when mice were treated with gemcitabine + cariporide (Figure 7, A and B).

Taken together, our results indicate that the small-molecule inhibitor of NHE1, cariporide, targeted the PDAC stroma in vivo on at least 2 vital fronts; it disrupted the vicious cycle leading to marked fibrosis, and it shifted tumor immune cell infiltrate to a more lymphocytic one, consistent with a more potent antitumor response.