Epigenetic and Metabolic Reprogramming of Fibroblasts in Crohn's Disease Strictures Reveals Histone Deacetylases as Therapeutic Targets

Abstract

Background and aims: No effective therapeutic intervention exists for intestinal fibrosis in Crohn's disease [CD]. We characterized fibroblast subtypes, epigenetic and metabolic changes, and signalling pathways in CD fibrosis to inform future therapeutic strategies.

Methods: We undertook immunohistochemistry, metabolic, signalling pathway and epigenetic [Transposase-Accessible Chromatin using sequencing] analyses associated with collagen production in CCD-18Co intestinal fibroblasts and primary fibroblasts isolated from stricturing [SCD] and non-stricturing [NSCD] CD small intestine. SCD/NSCD fibroblasts were cultured with TGFβ and valproic acid [VPA].

Results: Stricturing CD was characterized by distinct histone deacetylase [HDAC] expression profiles, particularly HDAC1, HDAC2, and HDAC7. As a proxy for HDAC activity, reduced numbers of H3K27ac+ cells were found in SCD compared to NSCD sections. Primary fibroblasts had increased extracellular lactate [increased glycolytic activity] and intracellular hydroxyproline [increased collagen production] in SCD compared to NSCD cultures. The metabolic effect of TGFβ stimulation was reversed by the HDAC inhibitor VPA. SCD fibroblasts appeared 'metabolically primed' and responded more strongly to both TGFβ and VPA. Treatment with VPA revealed TGFβ-dependent and TGFβ-independent Collagen-I production in CCD-18Co cells and primary fibroblasts. VPA altered the epigenetic landscape with reduced chromatin accessibility at the COL1A1 and COL1A2 promoters.

Conclusions: Increased HDAC expression profiles, H3K27ac hypoacetylation, a significant glycolytic phenotype and metabolic priming characterize SCD-derived as compared to NSCD fibroblasts. Our results reveal a novel epigenetic component to Collagen-I regulation and TGFβ-mediated CD fibrosis. HDAC inhibitor therapy may 'reset' the epigenetic changes associated with fibrosis.

Keywords: Collagen-I; Crohn’s disease; fibrosis; histone deacetylase; valproic acid.

Figure 1.

Reduction of H3K27ac levels in stricturing [SCD] and non-stricturing [NSCD] intestine. [A] Representative H3K27ac IHC images for mucosal [M-H3K27ac] and muscularis propria [MP-H3K27ac] [top and bottom panels respectively; 400× magnification] for NSCD (left) and SCD (right) tissue. Examples of positively and negatively stained stromal cells highlighted by red or yellow arrows, respectively. [B] Box and whisker plot showing percentage of H3K27ac-positive cells [H3K27ac+ve] in the mucosa [M] [n = 14] of non-inflamed (inactive) and inflamed (active) SCD and NSCD samples compared to control. [C] Comparison of H3K27ac-positive cells in the mucosa [M] and [D] muscularis propria (MP) of NSCD and SCD tissue. [E] Fibrosis scores and [F] ulcer scores in SCD formalin-fixed paraffin-embedded sections relative to patient-matched NSCD sections [n = 14 pairs]. Box plots show 25th to 75th percentiles, median [horizontal bar], and lowest and highest value [whiskers]. Differences between SCD and NSCD samples were determined by a paired t-test. Significant results are indicated by asterisks [*p < 0.05, **p< 0.01, ****p < 0.0001].

Figure 2.

Stricturing [SCD] and non-stricturing [NSCD] Crohn’s disease primary fibroblast cultures are glycolytically and metabolically distinct. [A–D] TGFβ induces metabolic changes in CCD-18Co cells. [A] Metabolic changes are partially reversed by VPA. Plot shows principal component [PC] scores 1 and 2 [rotated to show group discrimination most clearly]. Blue: TGFβ added; black: no TGFβ; empty symbols: VPA added; filled symbols: no VPA. Ellipses represent standard deviations and the mean is indicated by a cross. [B] Metabolic effect of TGFβ is greater than VPA for CCD-18Co cells, and largely affects different individual metabolites. GABA is affected by VPA through an HDAC-independent mechanism, and is a positive control; here, it was significantly changed by VPA, p = 4.3e-5; p values from two-way ANOVA. [C] Hydroxyproline is increased by TGFβ treatment in CCD-18Co cells, partially counteracted by VPA treatment [p = 0.0073, interaction term, two-way ANOVA]. [D] TGFβ increases glycolytic activity in CCD-18Co cells, as evidenced by increased extracellular lactate, which is partially reversed by VPA [p = 0.0012, interaction term, linear mixed effects model]. [E] SCD and NSCD primary fibroblast cultures are significantly differentiated from each other in terms of glycolytic activity, as evidenced by extracellular lactate production [p = 0.00028, ANOVA, with SCD/NSCD, TGFβ, and VPA as factors]. [F] TGFβ treatment decreases glucose utilization in SCD primary cell cultures, reversed by VPA [p = 0.00094, interaction term, linear mixed effects model]. [G] TGFβ [left] has a similar metabolic effect in SCD and NSCD cultures, but effects are greater for SCD. [H] The same is true for VPA [right], although metabolic changes are different from TGFβ. Solid lines indicate a p-value threshold of 0.001 and dashed lines a threshold of 0.01. [I, J] PC analysis of endometabolome data indicates that, for SCD [I] and NSCD [J] cultures, there is an overall metabolic effect of both TGFβ and VPA; these effects are largely independent. Blue: TGFβ added; red: no TGFβ; empty symbols: VPA added; filled symbols: no VPA. Ellipses represent standard deviations, and the mean is indicated by a cross. NB. Separation occurs within the first three components for SCD and first five components for NSCD cells. Components are rotated to show separation more clearly.

Figure 3.

VPA impacts SMAD signalling, its downstream target TGFβ1|1, and reduces collagen in TGFβ-stimulated intestinal fibroblasts. [A] Representative immunofluorescent [IF] images of H3K27ac, Collagen-I, and TGFβ1|1 in CCD-18Co cells treated with vehicle control [PBS] or TGFβ1 in combination with VPA. [B] Heatmap showing mRNA levels of key selected genes in VPA- and TGFβ-treated cells [n = 4]; *significant differences associated with VPA treatment, ^$significant differences associated with TGFβ treatment. [C] Representative western blots for SMAD expression in cells treated with VPA and TGFβ [n = 3; SMAD4, n = 4]. [D–F] Protein quantification bar graphs from IF images [n = 4]. [G] SMAD4 protein level normalized to β-Actin loading control. [H] Pro-collagen-1α1 in conditioned media measured by ELISA [n = 4]. [I] VPA inhibits gel contraction by CCD-18Co fibroblasts. 3D gel contraction assays used 15 000 CCD-18Co cells and a Caco2 epithelial cell barrier in the presence or absence of 5 mM VPA or vehicle control [PBS] for 48 h. [J, K] Gel size was quantified after 48 h and Pro-col 1α1 levels in conditioned media measured by ELISA. Differences between treatments were determined by a paired t-test. Significant are results indicated by * or ^$ symbols [* or ^$p < 0.05, ** or ^$$p < 0.01, *** or ^$$$p < 0.001 and **** or ^$$$$p < 0.0001].

Figure 4.

Anti-fibrotic actions of VPA in primary CD patient-derived intestinal fibroblasts. [A] Representative immunofluorescent [IF] images of H3K27ac, Collagen-I, and TGFβ1|1 in primary CD fibroblast cultures treated with vehicle control [PBS] or TGFβ1 in combination with VPA. [B] IF protein quantification bar graphs for cultures derived from NSCD [n = 7] or SCD [n = 9] intestinal segments and if stimulated with TGFβ1 or vehicle control [PBS]. [C] Heat map showing differences between treatments for fibroblast activation markers. *Significant differences associated with VPA treatment; ^$significant difference associated with TGFβ treatment. [D–F] TGFβ1|1, COL1A1, and COL1A2 mRNA levels in cultured CD mucosal biopsies treated with TGFβ and/or VPA for 48 h [n = 9]. B1, inflammatory phenotype; B2, SCD phenotype. [G] Pro-collagen-1α1 levels in conditioned media measured by ELISA [n = 8]. Differences between treatments were determined by a paired t-test. Significant results are indicated by * or ^$ symbols [* or ^$p < 0.05, ** or ^$$p < 0.01, *** or ^$$$p < 0.001 and **** or ^$$$$p < 0.0001].

Figure 5.

Significant epigenetic reprogramming of fibroblasts following VPA-mediated changes in chromatin accessibility at gene promoters. [A] sc-ATAC-seq of CCD-18Co cells treated with PBS [n = 1060] or VPA [n = 1283] as a 2D t-distributed stochastic neighbour embedding [TSNE] plot. [B] Pathway enrichment analysis of promoters with increased and reduced promoter accessibility in VPA-treated fibroblasts using the Molecular Signatures Database. [C] Overlap between changes in gene expression [illumina array] and promoter accessibility [sc-ATAC-seq] as a Venn diagram. Up-regulated genes and genes with increased promoter accessibility [top]; down-regulated genes and genes with reduced promoter accessibility [bottom]. Box indicates down-regulated genes with reduced promoter accessibility. [D, E] sc-ATAC-seq peak distribution charts showing the sizes of called peaks [grey rectangles] associated with the COL1A1 and COL1A2 promoters in CCD-18Co cells treated with PBS [n = 1060] or VPA [n = 1283]. Peak height is proportional to the number of cells within that cluster that had cut sites falling within the called peak. Cut site tracks are shown from cells treated with PBS [gold] and VPA [blue]; these are measures of average accessibility of a particular base for each cell in the cluster. Significant differences between VPA- and PBS-treated cells were calculated by comparing the sum of cut sites per cell [within peaks] which are close to one of the transcription start sites for each gene, respectively [promoter sums analysis]; log₂ fold-change [FC] in VPA-treated cells and p values [****p < 0.0001] are shown.