Arachidonate 15-lipoxygenase-mediated production of Resolvin D5n-3 DPA abrogates pancreatic stellate cell-induced cancer cell invasion

Abstract

Activation of pancreatic stellate cells (PSCs) to cancer-associated fibroblasts (CAFs) is responsible for the extensive desmoplastic reaction observed in PDAC stroma: a key driver of pancreatic ductal adenocarcinoma (PDAC) chemoresistance leading to poor prognosis. Specialized pro-resolving mediators (SPMs) are prime modulators of inflammation and its resolution, traditionally thought to be produced by immune cells. Using liquid chromatography-tandem mass spectrometry (LC-MS/MS)-based lipid mediator profiling PSCs as well as primary human CAFs express enzymes and receptors to produce and respond to SPMs. Human PSC/CAF SPM secretion profile can be modulated by rendering these cells activated [transforming growth factor beta (TGF-β)] or quiescent [all-trans retinoic acid (ATRA)]. ATRA-induced nuclear translocation of arachidonate-15-lipoxygenase (ALOX15) was linked to increased production of n-3 docosapentaenoic acid-derived Resolvin D5 (RvD5_{n-3 DPA}), among other SPMs. Inhibition of RvD5_{n-3 DPA} formation increases cancer cell invasion, whereas addback of this molecule reduced activated PSC-mediated cancer cell invasion. We also observed that circulating concentrations of RvD5_{n-3 DPA} levels were decreased in peripheral blood of metastatic PDAC patients when compared with those measured in plasma of non-metastatic PDAC patients. Together, these findings indicate that RvD5_{n-3 DPA} may regulate cancer-stroma cross-talk and invasion.

Keywords: ALOX15; CAF subtypes; all-trans retinoic acid; cancer-associated fibroblast; lipid mediator; pancreatic ductal adenocarcinoma; specialized pro-resolving mediator.

Figure 1

Pancreatic stellate cells’ (PSCs’) lipid mediator profiling dependent on activation status. (A) Schematic depiction of the experiment: cultured PSCs (PS1) were treated daily with ATRA 1 µM (quiescent) for 7 days, with TGF-β 5 ng/mL (strongly activated) for 3 days, or with vehicle control (Ctrl, ethanol 0.01%, mildly activated by plastic culturing). Cells were then lysed with 100% methanol containing deuterated internal standards along with the culture media and stored at −80°C to allow for protein precipitation and subsequent lipid mediator extraction and identification using LC-MS/MS-based lipid mediator profiling. (B) Partial least squares discriminant analysis (PLS-DA, supervised) scores plot depicting the three biological replicates from Ctrl-, ATRA-, and TGF-β-treated PS1 cells. (C) PLS-DA variable in projection (VIP) score plot highlighting all significant mediators (VIP > 1) contributing to the separation of the clusters of the three conditions (Ctrl, ATRA, and TGF-β). (D–G) Pathway analysis highlighting the mediators that were found differentially regulated in PLS-DA analysis and their synthetic pathways. Each mediator is shown as a mini-heatmap depicting their relative concentration among the different conditions in PS1 cells: ATRA (top rectangle), Ctrl (middle), and TGF-β (bottom). Same pattern connecting lines indicate a synthetic pathway. Grayed-out analytes were not detected. Dark border highlight indicates a VIP > 1. Figure is split according to lipid mediator metabolomes: (D) docosahexaenoic acid (DHA), € n-3-docosapentaenoic acid (n-3 DPA), (F) eicosapentaenoic acid (EPA), or (G) arachidonic acid (AA). n = 3 (biological repeats with internal technical repeats for all experiments).

Figure 2

PSC and cancer-associated-fibroblast (CAF) lipid mediator synthetic enzyme expression and subcellular localization in quiescent and activated state. PSC (PS1) and CAFs previously sub-categorized into four groups (A–D, according to Neuzillet 2019 (12) and herein ordered by prognostic relevance: C, B, A, and D, from better to worse prognosis) were cultured on coverslips, treated with ATRA 1 µM daily for 7 days. Representative immunocytochemistry images for ALOX15 (A), ALOX12 (B), and COX-2 (C) expression in ATRA-treated PSC/CAF cells (red) compared to vehicle controls (blue). Respective IgG (control) antibodies were used for background setting (Supplementary Figures 5–7). Red dashed tram lines mark nuclear envelope. The yellow dotted line depicts cell boundaries defined by concurrent αSMA co-staining (not shown for simplicity). ALOX15, ALOX12, and COX2 are shown in grayscale (white color) and DAPI is shown in blue. Scale bar = 20 µm. ns, not significant.

Figure 3

Total and subcellular lipid mediator biosynthetic enzyme expression in PSC and cancer-associated-fibroblasts (CAFs) in different activation states. (A–C) Quantification of total cellular expression (median intensity) by immunocytochemistry for ALOX15, ALOX12, and COX2 from three independent experiments (median ± IQR) in PSC (PS1) and CAF (subtypes C, B, A, D: ordered from better to worst patient prognosis). (D–F) Quantification of proportion of nuclear ALOX15, ALOX12, and COX2 expression as defined by ratio of nuclear to total cellular median intensity delineated by DAPI edge (median ± IQR). (G–I) Quantification of proportion of nuclear envelope ALOX15, ALOX12, and COX2 expression as defined by ratio of perinuclear space to total cell median intensity where perinuclear space area was determined by expanding and shrinking DAPI edge by 1 µm (2 µm in total, red dashed lines). Mann–Whitney against same cell-type vehicle control (*) or against PS1 vehicle control (^&). *, ^&p < 0.05, **, ^&&p < 0.01, ***, ^&&&p < 0.001, ****, ^&&&&p < 0.0001. AFU, arbitrary fluorescent units.

Figure 4

3D in vitro physio-mimetic (organotypic and spheroid) invasion assays after ALOX15 inhibition [chemical (PD146176) and genetic (shRNA)]. (A) Schematic of PSCs (PS1) and cancer (MIAPaCa-2) cells were co-cultured at a 2:1 ratio, respectively, in an organotypic 3D invasion assay and treated daily with ATRA 1 µM, vehicle control (Ctrl, ethanol 0.1%), PD146176 5µM, and/or RvD5_{n-3 DPA} 1/10 nM for 10 days while allowing invasion. (B) Representative organotypic 3D invasion assay H&E images for each of the conditions. The dotted line represents the boundary between ECM gel and the cell layer from where invasion (marked by arrows) is counted. Scale bar = 300 µm. (C) Relative (normalized to vehicle control) invaded cell quantifications with each data point reflects the median value of the number of invaded cells for five serial 30× fields (limited to the area of cellularity avoiding edge artifacts) per one gel. Different symbols indicate technical replicates (2–6) within biological replicates (n = 3). (D) Experimental schematic of PSCs pre-treated with either ATRA (7 days), PD146176 5 µM (3 days), and/or RvD5_{n-3 DPA} 1 or 10 nM (7 days) and then co-cultured with cancer cells to form spheres and subsequently embedded in Matrigel/collagen gels and allowed to invade for 2 days. (E) Representative 20× bright field microscopy images of hanging droplet spheroid invasion assay model for each of the conditions. Scale bar = 200 µm. (F) Relative (normalized to respective vehicle control) invaded cell quantifications (%invasion derived from invaded area divided by central spheroid area). Different symbols indicate technical replicates (2–6) within biological replicates (n = 3). (G) Representative 20× images of the hanging droplet spheroid invasion assay model. ALOX15 shRNA knockdown PSCs or respective controls were pre-treated with either ATRA (7 days) or RvD5_{n-3 DPA} 10nM (7 days) and then co-cultured with cancer cells to form spheres and subsequently embedded in Matrigel/collagen gels and allowed to invade for 2 days. Scale bar = 200 µm. (H) Representative 20× bright field microscopy images of hanging droplet spheroid invasion assay model for each of the conditions. (I) Relative (normalized to vehicle control) invaded cell quantifications (%invasion derived from invaded area divided by the central spheroid area). Different symbols reflect technical replicates (2–6) within biological replicates (n = 3). Kruskal–Wallis with Dunn’s multiple comparisons test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Figure 5

Plasma lipid mediator profiling from patients with PDAC and healthy volunteers. Summary data from LC-MS/MS-based lipid mediator profiling using plasma from a cohort of 20 PDAC patients and 22 matched healthy volunteers resolved by normalized values using multivariate partial least squares discriminant analysis (PLS-DA) analyses (A–F) to understand clustering and differentiation according to lipid mediators as well as by absolute concentration values (G–O). Multivariate analyses, top: scores plot (A, C, E), bottom: VIP scores (B, D, F), left (A, B) PDAC (n = 20) vs. healthy volunteers (n = 22), center (C, D) non-metastatic PDAC (stages I–III, n = 11) vs. metastatic PDAC (stage IV, n = 9) vs. healthy volunteers (n = 22), and right (E, F) non-metastatic (n = 11) vs. healthy volunteers (n = 22). (A–I) Graphs represent differences observed in significantly regulated mediators (FDR-corrected p-value *<0.05 Mann–Whitney U-test). Zero values represent values below the detection limit for LC-MS/MS.

Figure 6

Pathway analysis of SPM production in human plasma from healthy volunteers and patients with PDAC. Pathway analysis on non-metastatic PDAC patients against healthy volunteers highlighting (in color) the mediators that were found differentially regulated in PLS-DA VIP scores >1 and the synthetic pathways highlighted in dark. Dull dots represent mediator not detected. Each synthetic pathway is demonstrated with a distinct connecting line. Figure is split according to lipid mediator metabolomes: (A) docosahexaenoic acid (DHA), (B) n-3-docosapentaenoic acid (n-3 DPA), (C) eicosapentaenoic acid (EPA), or (D) arachidonic acid (AA). PDAC (n = 20), healthy volunteers (n = 22).