Epigenetic switch drives the conversion of fibroblasts into proinvasive cancer-associated fibroblasts

Abstract

Carcinoma-associated fibroblasts (CAF) mediate the onset of a proinvasive tumour microenvironment. The proinflammatory cytokine LIF reprograms fibroblasts into a proinvasive phenotype, which promotes extracellular matrix remodelling and collective invasion of cancer cells. Here we unveil that exposure to LIF initiates an epigenetic switch leading to the constitutive activation of JAK1/STAT3 signalling, which results in sustained proinvasive activity of CAF. Mechanistically, p300-histone acetyltransferase acetylates STAT3, which, in turn, upregulates and activates the DNMT3b DNA methyltransferase. DNMT3b methylates CpG sites of the SHP-1 phosphatase promoter, which abrogates SHP-1 expression, and results in constitutive phosphorylation of JAK1. Sustained JAK1/STAT3 signalling is maintained by DNA methyltransferase DNMT1. Consistently, in human lung and head and neck carcinomas, STAT3 acetylation and phosphorylation are inversely correlated with SHP-1 expression. Combined inhibition of DNMT activities and JAK signalling, in vitro and in vivo, results in long-term reversion of CAF-associated proinvasive activity and restoration of the wild-type fibroblast phenotype.

Figure 1. Epigenetic-dependent proinvasive fibroblasts phenotype.

(a) Schematic representation of experimental conditions for long-term maintenance of the proinvasive properties of LIF- and TGFβ1-activated fibroblasts. (b) Representative images of haematoxylin and eosin (H&E) coloration of paraffin-embedded sections of SCC12 in response to control hDF (Veh.) or hDF previously activated for 7 days with LIF or TGFβ1, in presence or absence of LIF blocking antibody (αLIF) after 15, 30 and 60 days of culture in 0.5% SVF media. Scale bar, 100 μm. I.I., invasion index (n=3; mean±s.d.; ***P<0.001). (c) Immunoblot of pJAK1, pSTAT3 and pSMAD2 in hDF cells stimulated by three CAF-CM, long-term LIF and TGFβ1-activated hDF-CM and short-term TGFβ-activated hDF-CM. Immunoblot of JAK1, STAT3, SMAD2 and tubulin as control. (d) Percentage of gel contraction by hDF stimulated by three CAF-CM, long-term LIF and TGFβ1-activated hDF-CM and short-term TGFβ-activated hDF-CM (n=3 in triplicate, mean+s.d., ***P<0.001). (e) Quantification of matrix remodelling by CAF in the presence of small molecules inhibitors for 6 days. (n=3 in triplicates).

Figure 2. P300 acetylates STAT3 for proinvasive CAF activity.

(a) Representative images of H&E coloration of paraffin-embedded sections of SCC12 in response to control hDF transfected with control siRNA (siLuc) or p300 (sip300) prior to control (veh.) or LIF stimulation. Scale bar, 100 μm (n=3; mean ±s.d.; ***P<0.001). (b) Immunoblot of p300, AcSTAT3 and pSTAT3 in hDF transfected with RNAi control (siLuc) or targeting p300 (siP300#1 and #2) before LIF stimulation. Immunoblot of STAT3 and tubulin as control. (c) Immunoblot of p300, AcSTAT3, pSTAT3 and STAT3 in hDF following immunoprecitation of STAT3 on 24 h LIF stimulation. Immunblot of p300 and tubulin in total lysate as control. (d) Immunoblot of pJAK1, AcSTAT3 and pSTAT3 in hDF in presence or absence of C646 inhibitor followed by long-term control (veh.) or LIF stimulation. Immunoblot of JAK1, STAT3 and tubulin as control. (e) Immunoblot of pJAK1, AcSTAT3 and pSTAT3 in control or long-term CTBP stimulated hDF. Immunoblot of JAK1, STAT3 and tubulin as control. (f) Representative images of H&E colouration of paraffin-embedded sections of SCC12 in response to control hDF (veh.) and hDF_LIF overexpressing a control (empty-GFP) or a wild-type STAT3 (STAT3-GFP) or acetylated-deficient STAT3 mutant (STAT3-K685R-GFP). Scale bar, 100 μm (I.I., invasion index; n=3; mean ±s.d.; ***P<0.001). (g) Immunoblot of pJAK1, pSTAT3 and AcSTAT3 in hDF overexpressing a control vector (empty-GFP), a wild-type (WT) STAT3 (STAT3-GFP) or acetylated-deficient STAT3 mutant (STAT3-K685R-GFP) in response to long-term LIF stimulation. Immunoblot of JAK1, STAT3, and tubulin as control.

Figure 3. DNMTs controls proinvasive fibroblasts' activity.

(a) Representative images of H&E colouration of paraffin-embedded sections of SCC12 in response to CAF in the absence (Veh.) or presence of 5′-Aza inhibitor (n=3, I.I., invasion index, mean±s.d., ***P<0.001). Scale bar 100 μm. (b) Percentage of gel contraction by CAF transfected with siRNA control (siLuc) or targeting DNMT1 (siDNMT1#1, #2, #3, #4) (n=3 in triplicate, mean+s.d., ***P<0.001). Bottom images show the contracted gels. (c) Representative images of H&E colouration of paraffin-embedded sections of SCC12 in response to CAF transfected with siRNA control (siLuc) or targeting DNMT1 (siDNMT#1, #2, #3, #4) (n=3, mean±s.d., ***P<0.001). Scale bar 100 μm. (d) Quantification of mRNA level of DNMT1, DNMT3a and DNMT3b in hDF following 48 h stimulation of LIF relative to control hDF (red dotted line) (n=3 in triplicate, mean+s.d., ***P<0.001). (e) Immunoblot of AcSTAT3, DNMT3b and STAT3 in hDF cells overexpressing a control or a wild type STAT3 or acetylated deficient STAT3 mutant following STAT3 immunoprecipitation upon LIF stimulation. Immunoblot of pSTAT3, STAT3 and tubulin on total lysates as control. (f) Representative images of H&E coloration of paraffin-embedded sections of SCC12 in response to hDF transfected with siRNA control (siLuc) or targeting DNMT3b (siDNMT3b#1 and #2) and subsequently long-term control- (Veh.) or LIF-activated hDF (n=3, mean±s.d., ***P<0.001). Scale bar 100 μm. (g) Immunoblot of DNMT3b in hDF activated or not by LIF transfected with siRNA control (siLuc) or targeting DNMT3b (siDNMT3b#1, #2, #3, #4). Immunoblot of tubulin as control.

Figure 4. Epigenetic-dependent sustained JAK1/STAT3 signalling in CAF.

(a) Immunoblot of pJAK1, pSTAT3 and AcSTAT3 in CAF after 7-days treatment in the presence or absence of 5′-Aza inhibitor at 10 and 3 μM final concentration. Immunoblot of JAK1, STAT3 and tubulin as control. (b) Representative images of H&E coloration of paraffin-embedded sections of SCC12 in response to CAF following 5′-Aza inhibitor alone or in combination with PTP inhibitor 1 (n=3, mean±s.d., ***P<0.001). (c) Immunoblot of DNMT1, pSHP-1, SHP-1, pJAK1, pSTAT3 and AcSTAT3 in CAF following 5′-Aza inhibitor alone or in combination with PTP inhibitor 1. Immunoblot of JAK1, STAT3 and tubulin as control. (d) Immunoblot of DNMT3b, SHP-1, pJAK1, pSTAT3 and AcSTAT3 in hDF activated or not by LIF transfected with siRNA control (siLuc) or siRNA targeting DNMT3b (siDNMT3b#1, #2). Immunoblot of JAK1, STAT3 and tubulin as control. (e) Schematic illustration of the PTPN6 gene structure and the regions analysed by bisulfite sequencing. The PTPN6 gene has two alternative promoters. We restricted our analysis to Promoter 1 since Promoter 2 is mostly active in the haematopoietic lineage. Vertical lines in Region 1 and Region 2 indicate the position of the CpG dinucleotides. (f) Bisulfite-sequencing results in two regions of the PTPN6 promoter (R1 and R2) in hDF transfected with control RNAi (siLuc) and RNAi targeting DNMT3b (siDNMT3b#1 and #2) and subsequently long-term control (−) or LIF-activated (+). Each line represents an individual sequence. Open and closed circles denote unmethylated and methylated CpG dinucleotides, respectively. Donut charts summarize the global proportion of methylated CpG with the percentage in the center (n=2, P<0.0001, two-tailed Fisher's exact test). (g) Representative images of H&E coloration of paraffin-embedded sections of SCC12 in response to CAF overexpressing control vector (GFP) or SHP-1 gene (SHP-1_GFP) (n=3, mean±s.d., ***P<0.001). (h) Immunoblot of SHP-1, pJAK1, pSTAT3 and AcSTAT3 in CAF overexpressing control vector (GFP) or SHP-1 gene (SHP-1_GFP). Immunoblot of JAK1, STAT3 and tubulin as control.

Figure 5. Phenotypic and molecular reversion of CAF.

(a) Schematic representation of the experimental conditions for long-term phenotypic and molecular reversion of the proinvasive property of activated fibroblasts. After inhibitors treatment for 7 or 21 days, cells are embedded in matrix for a 50-day gel contraction assay, assayed every day for a period of 5 days for molecular analysis, and assayed at day 5 after inhibitors removal for proinvasive ability. (See also Supplementary Fig. 5). (b) Percentage of gel contraction of hDF and CAF for a period of 50 days following a 7-day treatment of either Ruxolitinib or 5′-Aza or both inhibitors together. (c) Representative images of H&E coloration of paraffin-embedded sections of SCC12 in response to CAF, hDF_LIF and hDF_TGFβ1, 5 days after a period of 7-day treatment of Ruxolitinib or 5′-Aza or both inhibitors together (n=3, mean±s.d., ***P<0.001). Scale bar 100 μm. (d) Representative images of H&E colouration of paraffin-embedded sections of MDA-MB-468 human breast cancer cell line in response to breast CAF 5 days after a period of 7-day treatment of Ruxolitinib or 5′-Aza or both inhibitors together (n=3, mean±s.d., ***P<0.001).

Figure 6. Stromal DNMT and JAK favour tumour invasion in vivo.

(a) Schematic representation of the experimental conditions used for in vivo cell co-implantation into the mammary fat pad of 8-week-old BALB/c female mice. 67NR mouse breast carcinoma cells were co-injected together with LIF preactivated mDF and treated with inhibitors. (b) Graphic representation of tumour formation induced by 67NR cells alone or in the presence of LIF-preactivated mDF. Total numbers of mice bearing tumours after single or combined inhibitor treatment are shown. (c) H&E coloration of paraffin-embedded sections of primary tumours isolated from mice (left panels) showing 67NR cells invading from the primary tumour (T) into the adjacent tissue (aT) (black arrow). Picrosirius Red staining visualized by both parallel (middle panels) and orthogonal (right panels) light showing tumour ECM remodelling at the areas invaded by the tumoral 67NR cells. Scale bars, 200 and 100 μm for × 20 and × 40X magnifications, respectively. (d) Quantification of Picrosirius Red staining using ImageJ software. Percentage of threshold area is shown (mean±s.d.; *P<0.05).

Figure 7. Inverse correlation of AcSTAT3 and SHP-1 expression.

(a) AcSTAT3, pSTAT3 and SHP-1 immunohistological staining in human head and neck (n=50) and lung (n=50) carcinomas. Scale bar, 100 μm. (b) Quantification of mean Quick Score (QS) for SHP-1, both in the stroma and the tumour bulk, in lung carcinomas decorated with high- and low-acetylated STAT3 shown in a. (c) Quantification of mean Quick Score (QS) for pSTAT3, in the tumour bulk, in lung carcinomas decorated with high and low acetylated STAT3 shown in a. (d) Quantification of mean QS for SHP-1, both in the stroma and the tumour bulk, in head and neck carcinomas decorated with high- and low-acetylated STAT3 tumours shown in a. (e) Quantification of mean Quick Score (QS) for pSTAT3, in the tumour bulk, in head and neck carcinomas decorated with high- and low-acetylated STAT3 shown in a. (f) Plot of mean QS form AcSTAT3 (x axis) and SHP-1 (y axis) showing a negative correlation between Acetylated STAT3 and SHP-1 detection in lung (blue) and head and neck (red) carcinomas.