Identification of transforming growth factor-beta-regulated microRNAs and the microRNA-targetomes in primary lung fibroblasts

Abstract

Background: Lung fibroblasts are involved in extracellular matrix homeostasis, which is mainly regulated by transforming growth factor-beta (TGF-β), and are therefore crucial in lung tissue repair and remodeling. Abnormal repair and remodeling has been observed in lung diseases like COPD. As miRNA levels can be influenced by TGF-β, we hypothesized that TGF-β influences miRNA expression in lung fibroblasts, thereby affecting their function.

Materials and methods: We investigated TGF-β1-induced miRNA expression changes in 9 control primary parenchymal lung fibroblasts using miRNA arrays. TGF-β1-induced miRNA expression changes were validated and replicated in an independent set of lung fibroblasts composted of 10 controls and 15 COPD patients using qRT-PCR. Ago2-immunoprecipitation followed by mRNA expression profiling was used to identify the miRNA-targetomes of unstimulated and TGF-β1-stimulated primary lung fibroblasts (n = 2). The genes affected by TGF-β1-modulated miRNAs were identified by comparing the miRNA targetomes of unstimulated and TGF-β1-stimulated fibroblasts.

Results: Twenty-nine miRNAs were significantly differentially expressed after TGF-β1 stimulation (FDR<0.05). The TGF-β1-induced miR-455-3p and miR-21-3p expression changes were validated and replicated, with in addition, lower miR-455-3p levels in COPD (p<0.05). We identified 964 and 945 genes in the miRNA-targetomes of unstimulated and TGF-β1-stimulated lung fibroblasts, respectively. The TGF-β and Wnt pathways were significantly enriched among the Ago2-IP enriched and predicted targets of miR-455-3p and miR-21-3p. The miR-455-3p target genes HN1, NGF, STRADB, DLD and ANO3 and the miR-21-3p target genes HHEX, CHORDC1 and ZBTB49 were consistently more enriched after TGF-β1 stimulation.

Conclusion: Two miRNAs, miR-455-3p and miR-21-3p, were induced by TGF-β1 in lung fibroblasts. The significant Ago2-IP enrichment of targets of these miRNAs related to the TGF-β and/or Wnt pathways (NGF, DLD, HHEX) in TGF-β1-stimulated fibroblasts suggest a role for these miRNAs in lung diseases by affecting lung fibroblast function.

Fig 1. Upregulation of ECM genes and α-SMA after TGF-β1 stimulation in primary parenchymal lung fibroblasts.

(A) Effective TGF-β1 stimulation of control fibroblasts in the discovery group was confirmed by the upregulation of FN1 (fibronectin 1), COL1A1 (collagen type I alpha I) and α-SMA (alpha-smooth muscle actin), genes that are well-known to be affected by TGF-β. (B) These genes were also upregulated in the control and COPD fibroblasts in the replication group after 2.5 ng/ml TGF-β1 and (C) after 7.5 ng/ml TGF-β1 stimulation. Data are presented as relative expression (2^-ΔCp). **p<0.01, ***p<0.001, ****p<0.0001.

Fig 2. Differentially expressed miRNAs after TGF-β1 stimulation in control lung fibroblasts in the discovery group.

(A) The miRNA expression in primary parenchymal lung fibroblasts of control subjects with and without TGF-β1 stimulation was determined by microarray. Unsupervised hierarchical clustering was used to generate the heatmap and pearson correlation was used as the distance metric. Twenty-nine miRNAs were differentially expressed after TGF-β1 stimulation (FDR<0.05). The heatmap shows the median-centered expression of the 29 miRNAs of which 8 miRNAs were downregulated and 21 miRNAs were upregulated after TGF-β1 stimulation. (B) Validation of differentially expressed miRNAs after TGF-β1 stimulation in the discovery group using qRT-PCR. Data are presented as relative expression (2^-ΔCp) normalized to RNU48. *p<0.05, **p<0.01.

Fig 3. Replication of differentially expressed miRNAs after TGF-β1 stimulation using qRT-PCR.

To replicate the TGF-β1-induced expression changes of the validated miRNAs, qRT-PCR was performed on the control and COPD fibroblasts in the replication group (A) stimulated with 2.5 ng/ml TGF-β1 and (B) stimulated with 7.5 ng/ml TGF-β1. Data are presented as relative expression (2^-ΔCp) normalized to RNU48. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Fig 4. Efficiency of Ago2-immunoprecipitation.

(A) Western blot of Ago2 protein in the Ago2- and IgG-immunoprecipitation. M marker; T Total fraction; FT Flow through fraction; IP Immunoprecipitation fraction. Arrow indicates the Ago2 protein. Ago2 protein can be detected in the Ago2-IP fraction, while it cannot be detected in the IgG₁-IP fraction. (B) To confirm miRNA enrichment in the IP fraction, qRT-PCR was performed for three randomly selected miRNAs expressed in lung fibroblasts, i.e. miR-31-5p, miR-455-3p and miR-199b. The levels of these miRNAs are strongly increased in the Ago2-IP fraction compared to the IgG₁-IP fraction.

Fig 5. Defining the miRNA-targetomes of unstimulated and TGF-β1-stimulated lung fibroblasts.

The overlap of the top 1,500 most IP-enriched probes in the (A) unstimulated and (B) TGF-β1-stimulated fibroblasts of the two control subjects are defined as the miRNA-targetome. The identified genes in the miRNA-targetomes are listed in S1 Tables.

Fig 6. Significant Ago2-IP-enrichment of miR-455-3p and miR-21-3p predicted targets.

For each miRNA, the percentages of predicted targets were calculated in all expressed genes and in the top 1,500 most IP-enriched genes in all four IP experiments. Chi-square test was used to determine whether the number of predicted targets in the Ago2-IP fraction for miR-455-3p and miR-21-3p in the top 1,500 most enriched genes was significant different from the expected based on the number of predicted targets in all expressed genes.