Vitamin D receptor-mediated stromal reprogramming suppresses pancreatitis and enhances pancreatic cancer therapy

Abstract

The poor clinical outcome in pancreatic ductal adenocarcinoma (PDA) is attributed to intrinsic chemoresistance and a growth-permissive tumor microenvironment. Conversion of quiescent to activated pancreatic stellate cells (PSCs) drives the severe stromal reaction that characterizes PDA. Here, we reveal that the vitamin D receptor (VDR) is expressed in stroma from human pancreatic tumors and that treatment with the VDR ligand calcipotriol markedly reduced markers of inflammation and fibrosis in pancreatitis and human tumor stroma. We show that VDR acts as a master transcriptional regulator of PSCs to reprise the quiescent state, resulting in induced stromal remodeling, increased intratumoral gemcitabine, reduced tumor volume, and a 57% increase in survival compared to chemotherapy alone. This work describes a molecular strategy through which transcriptional reprogramming of tumor stroma enables chemotherapeutic response and suggests vitamin D priming as an adjunct in PDA therapy. PAPERFLICK:

Figure 1

Activated and cancer-associated PSCs exhibit a pro-fibrotic, pro-inflammatory phenotype. (A) Heatmap representing selected genes from RNA-Seq analysis of primary mouse PSCs, demonstrating gene categories with altered expression during activation. Data are represented as log₂ fold change, activated (day 7) vs. pre-activated (day 3), n=3 per group. (B) Heatmap representing selected genes from RNA-Seq analysis of primary human PSCs, isolated from PDA patients (n=5) or cancer-free donors (n=4) and cultured for 15 days to achieve adequate yield and purity, expressed as log₂ fold change PDA vs. cancer-free. (C) Heatmap showing the relative abundance of negative (top) and positive (bottom) regulators of angiogenesis in pre-activated and activated primary mouse PSCs. (D) Vdr expression in mouse whole-pancreas homogenates and in isolated PSCs, cultured for 3 days to expand and purify, as measured by qRT-PCR. (E) Vdr expression in the indicated pancreatic populations by qRT-PCR (normalized to 36B4; n=5). Acini, ducts, and islets were isolated by laser capture microdissection (LCM); PSCs were isolated by density centrifugation (DC). (F) Vdr expression in pre-activated and activated mouse PSCs (left) and in human non-cancer associated and cancer-associated PSCs (right) determined by qRT-PCR (normalized to 36b4, n=3). Bars indicate the mean; error bars indicate SD.

Figure 2

A VDR-regulated transcriptional network opposes PSC activation. (A) Representative images of primary human CAPSCs treated with vehicle (DMSO) or 100nM calcipotriol (Cal) for 48h and stained with BODIPY 493/503 for detection of neutral lipids. Quantification of percent BODIPY-positive area per cell in 3 patient samples treated with DMSO or Cal appears below, plotted as the mean + SD. Statistical significance determined by Student’s unpaired t-test (*p<0.05). Scale bar = 20 μm. (B) Expression of ACTA2 in 27 primary human CAPSCs treated with vehicle or 100nM Cal for 48h. Values were plotted as DMSO/Cal and normalized to 36B4. (C) Heatmap representing selected genes from RNA-Seq analysis of primary mouse PSCs treated with DMSO (D) or Cal (C) and harvested on day 3 (pre-activated) or day 7 (activated) of culture after isolation (n=3). VDR target genes Cyp24a1 and Vdr are shown as controls. Related to Table S1. (D) Heatmap showing the relative abundance of negative (top) and positive (bottom) regulators of angiogenesis in activated primary mouse PSCs cultured in the presence of vehicle (DMSO) or Cal. (E) Expression levels of selected genes from the PSC activation or cancer signatures in CAPSCs treated with DMSO or 100nM Cal for 48h. Results are representative of 3 patient samples and are plotted as the mean + SD. qRT-PCR was performed in technical triplicate and values were normalized to 36B4. Statistical significance determined by Student’s unpaired t-test (*p < 0.05). (F) CAPSCs were transfected with siRNA pools against VDR (siVDR) or a non-targeting control (siNT). Cells were treated with DMSO or 100nM Cal for 48h and analyzed by qRT-PCR. Values were normalized to 36B4. Results are representative of 3 patient samples and are plotted as the mean + SD. Statistical significance determined by Student’s unpaired t-test (*p<0.05; n.s.=not significant).

Figure 3

VDR ligand modulates PSC activation in vivo. (A) Expression levels of selected genes in PSCs isolated from mice injected with cerulein (Cer) or cerulein + Cal for 12 weeks (n=10). Values were normalized to 36b4 and are plotted as the mean + SD. (B) Quantification of immunofluorescent staining for phospho-Stat3 (p-Stat3) on frozen sections from wild-type mice treated with cerulein or cerulein + Cal for 12 weeks (n=5). (C) Expression levels of selected genes in PSCs isolated from mice injected with Cer or Cer + Cal to induce acute pancreatitis (for details see Supplemental Experimental Procedures; n=5). Values were normalized to 36b4 and are plotted as the mean + SD. (D) Leukocyte recruitment, as measured by CD45-positive cells, in mice with acute pancreatitis (immunofluorescent staining of frozen sections, positive cells in 20X field, n=5). (E) Fibrosis, as measured by Sirius red staining, in mice with acute pancreatitis (per 20X field, n=5). (F) Expression levels of selected genes in PSCs isolated from Vdr^+/+ and Vdr^−/− mice injected with cerulein to induce acute pancreatitis (n=5). Means + SD are shown; values normalized to B2M. (G) Sirius red-positive area in Vdr^+/+ and Vdr^−/− mice with acute pancreatitis (per 20X field, n=5). Statistical significance determined by Student’s unpaired t-test (*p<0.05). (H) Expression levels of selected genes in PSCs isolated from Vdr^+/+ and Vdr^−/− mice after treatment with DMSO or 100 nM Cal for 48 h. Statistical significance determined by Student’s unpaired t-test (*p<0.05; n.s.=not significant).

Figure 4

Stromal VDR activation decreases pro-tumorigenic paracrine signaling. (A) Volcano plots representing gene expression changes detected by RNA-Seq in MIAPaCa-2 cells treated with 100nM Cal for 48h vs. media alone (left), with CAPSC conditioned media (CM) for 48h vs. media alone (middle), or with CM from Cal-treated CAPSC (100nM, 48h) for 48h vs. media alone. Blue indicates significant change; red indicates no significant change. (B) Heatmap representing selected genes from the RNA-Seq analyses described in (A), plotted as fold change vs. media alone (DMEM). (C–G) The indicated cell lines were incubated with Cal directly, or with CM from CAPSC with or without Cal treatment, as described above. Expression levels of candidate genes CXCL1, CSF2, and AURKB were determined by qRT-PCR. Values were normalized to 36B4; means + SD are shown. Statistical significance determined by Student’s unpaired t-test (*p < 0.05). Results are shown as replicates with 1 patient sample, and are representative of results from multiple patient samples (n=4), though sample-to-sample variability was noted. (H) Immunoblot for p-STAT3 from MIAPaCa-2 cells treated for 24h with 100nM Cal, CAPSC CM, Cal + CAPSC CM, or Cal + CAPSC (Cal-treated) CM. Actin served as a loading control. Values indicate densitometric ratios (p-STAT3/Actin). (I) Viability of MIAPaCa-2 cells, treated as described above, incubated with the indicated doses of gemcitabine for 48h. Results are representative of 3 CAPSC CM samples and are plotted as the mean ± SD. Statistical significance determined by Student’s unpaired t-test (*p<0.05). Asterisks designate statistically significant differences in viability between CM and CM (PSC+Cal) samples at the indicated dose of gemcitabine.

Figure 5

Stromal VDR activation shows efficacy against pancreatic carcinoma in vivo when combined with gemcitabine. KPC mice were treated for 9 days with gemcitabine (Gem), calcipotriol (Cal), or Gem + Cal (Gem: n=4; Cal: n=7; Gem+Cal: n=7 unless otherwise indicated). (A) Percent change in tumor volume at study endpoint, measured by high-resolution ultrasound. Plots indicate range, median, and quartiles. *p<0.02; Kruskal-Wallis and Dunn’s nonparametric comparison test. (B) Aniline blue-stained collagen fibers were quantified as positive pixels per 20X field. Plots indicate range, median, and quartiles. *p<0.05 by Mann-Whitney U test. (C) Gene expression in tumor homogenates was determined by qRT-PCR. Values were normalized to 36b4. Bars indicate mean + SD. *p<0.05 by Student’s unpaired t-test (compared to Gem alone). (D) Intratumoral concentrations of gemcitabine triphosphate (dFdCTP, measured by LC-MS/MS) in Gem- and Gem+Cal-treated mice 2h after the final dose of gemcitabine (n=4 and 7 respectively). Plots indicate range, median, and quartiles. *p<0.05 by Mann-Whitney U test. (E) IHC for cleaved caspase-3 (CC3) was quantified as %CC3-positive tumor cells per 20X field. Plots indicate range, median, and quartiles. *p<0.05 by Mann-Whitney U test.

Figure 6

VDR ligand enhances delivery and efficacy of gemcitabine. KPC mice were treated for 9 days with gemcitabine (Gem), calcipotriol (Cal), or Gem + Cal (Gem: n=4; Cal: n=7; Gem+Cal: n=7 unless otherwise indicated), or treated with Gem (n=12) or Gem+Cal (n=15) until moribund. (A) Dck, Cda and Slc29a1/Ent1 gene expression in tumor homogenates determined by qRT-PCR. Values were normalized to 36b4. Bars indicate mean + SD. (B) IHC for CD31 was quantified as CD31 (NovaRed)-positive area per 40X field. Plots indicate range, median, and quartiles. *p<0.05 by Mann-Whitney U test. (C) Representative CD31 IHC from Gem- and Gem+Cal-treated KPC tumors. Arrows indicate a collapsed vessel in a gemcitabine-treated tumor (top), and a vessel with an apparent lumen in a Gem + Cal-treated tumor (bottom). Scale bar = 50 μm. (D) Kaplan-Meier survival analysis for KPC mice treated with Gem or Gem+Cal. p=0.0186 by Mantel-Cox (log rank) test.

Figure 7

Model depicting a role for VDR in signal-dependent stromal remodeling, limiting pancreatic tumor-stroma crosstalk. PSCs progressively acquire tumor-supporting functions during activation, a process that is driven by pancreatic injury and tumor progression via secreted factors from the epithelial compartment (and possibly from immune/inflammatory cells). VDR activation drives reversion of PSCs to a more quiescent, less tumor-supportive state. As such, co-treatment of pancreatic tumors with gemcitabine to target the tumor cells and VDR ligand to deactivate PSCs leads to an overall decrease in the reciprocal tumor-stroma crosstalk that presents a major barrier to the delivery and efficacy of gemcitabine alone.