Phosphodiesterase type 5 inhibitors enhance chemotherapy in preclinical models of esophageal adenocarcinoma by targeting cancer-associated fibroblasts

Abstract

The chemotherapy resistance of esophageal adenocarcinomas (EACs) is underpinned by cancer cell extrinsic mechanisms of the tumor microenvironment (TME). We demonstrate that, by targeting the tumor-promoting functions of the predominant TME cell type, cancer-associated fibroblasts (CAFs) with phosphodiesterase type 5 inhibitors (PDE5i), we can enhance the efficacy of standard-of-care chemotherapy. In ex vivo conditions, PDE5i prevent the transdifferentiation of normal fibroblasts to CAF and abolish the tumor-promoting function of established EAC CAFs. Using shotgun proteomics and single-cell RNA-seq, we reveal PDE5i-specific regulation of pathways related to fibroblast activation and tumor promotion. Finally, we confirm the efficacy of PDE5i in combination with chemotherapy in close-to-patient and in vivo PDX-based model systems. These findings demonstrate that CAFs drive chemotherapy resistance in EACs and can be targeted by repurposing PDE5i, a safe and well-tolerated class of drug administered to millions of patients world-wide to treat erectile dysfunction.

Keywords: cancer-associated fibroblasts; chemotherapy; esophageal adenocarcinoma; phosphodiesterase type 5 inhibitors; preclinical models.

Graphical abstract

Figure 1

Expression of PDE5A in EAC patient samples compared with normal esophagus (A) Expression of PDE5A in the Wang et al. dataset.PDE5A is overexpressed in Barrett’s esophagus (BE, n = 20), esophageal adenocarcinoma (EAC, n = 21), and esophageal squamous cell carcinoma (ESCC, n = 9) compared with normal squamous epithelium (NSE, n = 19). (B) Expression of PDE5A in the Krause et al. dataset.PDE5A is overexpressed in BE (n = 5) and EAC (n = 48) compared with normal adjacent epithelium (NAE, n = 18). (C) Expression of PDE5A in RNA-sequencing data from the TCGA and GTEx projects analyzed through Xena.PDE5A is overexpressed in EAC (n = 85) from TCGA compared with NSE (n = 269) from GTEx. ∗∗∗∗p < 0.0001. (D) PDE5A expression is associated with worse overall survival in EAC patients (n = 52) in Peters et al. Log2-transformed counts data were extracted and comparative analysis was performed using a Welch’s t test. Box and whisker plots represent the IQR of log2 expression of PDE5A. Central line = median, outliers are individual points outside of whiskers.

Figure 2

Expression of PDE5 in the esophagus and manipulation of CAF differentiation in vitro by PDE5i (A) PDE5 expression analysed by IHC of normal esophagus and EAC. PDE5 expression is localized to both tumor cells and the surrounding stromal tissue. Scale bars, 200 μm. (B) Western blot and ICC for PDE5 expression in three cancer cell lines (MFD-1, FLO-1, and OE33) and five matched normal (NOF)/cancer (CAF) primary esophageal fibroblasts, all normalized to Hsc70. (C) NOFs treated with TGF-β1 ± 50 μM vardenafil for 72 h. TGF-β1-treated NOFs express higher α-SMA, and vardenafil pre-treatment abrogated the TGF-β1-induced expression of α-SMA. (D) CAFs treated with 50 μM vardenafil for 72 h reduced α-SMA expression 50% by western blot and ICC. One-way ANOVA, p values: ∗p < 0.05, ∗∗∗∗p < 0.0001. Results are representative of three independent experiments. See also Figure S1.

Figure 3

Inhibition of CAF-associated activity in vitro by PDE5i (A) Fibroblast contraction analyzed by NOFs treated with TGF-β1 ± 50 μM vardenafil embedded in collagen-1 gel. TGF-β1 treatment induced α-SMA expression and gel contraction in normal fibroblasts, and co-treatment with vardenafil inhibited the upregulation of α-SMA and reduced gel contraction by 50%. (B) CAF contraction was analyzed by collagen-1 gel contraction ± 50 μM vardenafil. Α-SMA expression and collagen-1 gel contraction is 2-fold greater in untreated CAFs compared with vardenafil- treated CAFs. (C and D) Cancer cell invasion was analyzed by transwell assays using conditioned medium from fibroblasts as the chemoattractant. (C) Normal fibroblasts ± TGF-β1 ± vardenafil and TGF-β1-treated NOFs promoted invasion of FLO-1 cells with a 5-fold induction; this induction was abrogated by vardenafil treatment. (D) CAFs treated with vardenafil reduced invasion of FLO-1 cells by 60%. (E) Organotypic co-culture of FLO-1 cells and CAFs also showed inhibition of invasion when treated with vardenafil. One-way ANOVA, p values: ∗p < 0.05, ∗∗p < 0.01. Results are representative of three independent experiments. See also Figure S1.

Figure 4

Effects of PDE5 Inhibition or siRNA on the proteomic profile of CAFs (A) Principal component analysis of all quantified proteins showed that vardenafil-treated CAFs clustered together with PDE5 siRNA-treated CAFs compared with vehicle-treated CAFs. (B) Comparison of DEPs in CAFs treated with vardenafil or PDE5 siRNA with a previously published dataset of DEPs in CAFs versus NOFs, identified 83 proteins down-regulated in CAFs versus NOFs but became up-regulated in CAFs following treatment with vardenafil or PDE5 siRNA. Conversely, 88 proteins were up-regulated in CAFs versus NOFs but became down-regulated in CAFs following treatment with vardenafil or PDE5 siRNA. (C) GO analysis using DAVID of the 171 proteins that reversed their trend of modulation following treatment with vardenafil or PDE5 siRNA compared with CAFs. (D) Proteins mapping to the respective GO terms.

Figure 5

Single-cell RNA sequencing of co-cultured EAC cell lines and CAFs treated with PDE5i (A) Cluster analysis of MFD-1 and primary CAFs based on differentially expressed genes from single-cell RNA sequencing analysis (n = 1,122 cells). First, cells were clustered as either cancer cells (MFD-1, blue) or fibroblasts (red). (B) Further cluster analysis showed CAFs treated with PDE5 inhibitor vardenafil (CAFp) clustered separately from vehicle-treated CAFs (CAFv). PDE5i-treated co-cultured CAFs clustered with all other CAF cultures, whereas vehicle-treated CAFs co-cultured with MFD-1 show a phenotypic shift and clustered separately. (C) Violin plots showing typical gene expression profiles for fibroblasts (THY1, DCN, ACTA2) or epithelial cells (KRT8, KRT18, EPCAM) show good separation between the clusters in A. (D) Heatmap shows the top differentially expressed genes between treatment groups. CAFs and MFD-1 cells have the most distinct gene expression profiles, and although PDE5i treatment and co-culture both affect gene expression of both cell types, CAFs were most affected under both conditions.

Figure 6

Modeling of patient response to chemotherapy with PDE5i using 3D-TGA. Sensitivity of close-to-patient cells was determined in 3D-TGA, with and without mesenchymal cell co-culture, after 4-day exposure to ECF and vardenafil (PDE5i) drug combinations (A) Viability curves were generated and IC₅₀ values determined for a cohort of EAC patients’ ECF-treated 3D-TGAs, with (+) and without (−) hMSC support and the addition of PDE5i (n = 15 patient samples from 8 patients). The patient cancer cell clusters were classified as sensitive (green), borderline (orange), or resistant (red) by comparison of IC₅₀ values with the mean peak serum concentrations achieved in patients at the doses used in UK clinical practice. This is marked with an asterisk, where the IC₅₀ drop is significant (CI > 95%, p < 0.05). Tumor regression score (TRG) denotes the chemotherapy response of the patient’s tumor clinically (TRG1-3, sensitive; 4–5, non-responsive). (B) Overall sensitivity of all the EAC patient samples co-cultured with hMSCs was determined for assays with and without the addition of PDE5i to ECF chemotherapy. Horizontal lines represent mean IC₅₀s. (C) Gene expression profiling of CAF phenotypes by bulk RNA-seq of 3D-TGA samples from three patients (Oes4R, Oes5R, and Oes7R) with and without hMSC support, and corresponding parent tumors. Heatmap showing expression of 50 well-recognized CAF marker genes. Whereas Oes4R and Oes7R upregulated CAF marker gene expression when hMSC support was added, Oes5R failed to do so (column outlined in blue). See also Figure S2.

Figure 7

Efficacy study of PDX mouse models treated with chemotherapy alongside PDE5i (A) Tumor volumes of PDX models treated with saline for injection (vehicle), CX alone, or CX + PDE5i vardenafil (Var) or tadalafil (Tad) (n = 15 per treatment arm). (B and C) Example images of Untreated, CX, CX + Vardenafil, and CX + Tadalafil treatment arms in mouse PDX models are presented: IHC stained with (B) anti-α-SMA and (C) anti-periostin antibodies. (D and E) CAF differentiation was quantified by thresholding of digital whole-slide images from PDX models to measure total area stained for α-SMA (D) and periostin (E). See also Figures S3 and S4 and Table S4. Box and whisker plots represent the IQR of log2 expression of PDE5A. Central line = median, outliers are individual points outside of whiskers.