A Non-canonical Pathway with Potential for Safer Modulation of Transforming Growth Factor-β1 in Steroid-Resistant Airway Diseases.

Impaired therapeutic responses to anti-inflammatory glucocorticoids (GC) in chronic respiratory diseases are partly attributable to interleukins and transforming growth factor β1 (TGF-β1). However, previous efforts to prevent induction of GC insensitivity by targeting established canonical and non-canonical TGF-β1 pathways have been unsuccessful. Here we elucidate a TGF-β1 signaling pathway modulating GC activity that involves LIM domain kinase 2-mediated phosphorylation of cofilin1. Severe, steroid-resistant asthmatic airway epithelium showed increased levels of immunoreactive phospho-cofilin1. Phospho-cofilin1 was implicated in the activation of phospholipase D (PLD) to generate the effector(s) (lyso)phosphatidic acid, which mimics the TGF-β1-induced GC insensitivity. TGF-β1 induction of the nuclear hormone receptor corepressor, SMRT (NCOR2), was dependent on cofilin1 and PLD activities. Depletion of SMRT prevented GC insensitivity. This pathway for GC insensitivity offers several promising drug targets that potentially enable a safer approach to the modulation of TGF-β1 in chronic inflammatory diseases than is afforded by global TGF-β1 inhibition.

Introduction

Glucocorticoids (GCs) are widely used in the treatment of inflammatory and autoimmune diseases, such as pan class="Disease">asthma, chronic obstructive pulmonary disease (COPD), and rheumatoid arthritis (Rhen and Cidlowski, 2005). In most people with asthma, inflammation is well-controlled by inhaled GCs, whereas COPD-associated inflammation shows limited responsiveness to GCs. Moreover, approximately 10% of asthmatic patients remain symptomatic while using high doses of GCs (Chung et al., 2014). The pharmacodynamic manifestation of GC insensitivity includes a rightward shift in the concentration-response curve or a depression of the maximum response. Clinical GC insensitivity is defined as the presence of airway obstruction that is not relieved after 2 weeks of oral GC treatment, representing a reduction in the maximum GC response (Hew and Chung, 2010, Keenan et al., 2012).

GC activity in different cell types is dependent on the phenotype. Cell types not usually accorded a primary role in response to GCs, including epithelium, smooth muscle, and fibroblast, have well-documented capacity to add to the overall burden of inflammatory mediators (Keenan et al., 2012, Keenan et al., 2015). Forming at the interface between the host and the environment, airway epithelium detects environmental stimuli and secretes a range of cytokines and chemokines that contribute to airway inflammation and airway structural changes in asthma (Lambrecht and Hammad, 2012) and are modulated by GC treatment. Thus the airway or lung epithelium is the site of deposition and significant actions of inhaled GCs. A corollary of the importance of the airway epithelium in asthma pathogenesis and GC activity is that in patients with GC insensitivity this cell type also has the potential to contribute to clinical GC resistance.

GC anti-inflammatory activity can be impaired by pro-inflammatory cytokines, resulting in the insensitivity to GCs in inflammatory diseases (Dejager et al., 2014). Inflammatory stimuli, including tumor necrosis factor alpha, interleukin-1β (IL-1β), IL-17A, or viral infection, contribute to GC insensitivity in epithelium (Rider et al., 2015, Webster Marketon and Corry, 2013, Zijlstra et al., 2012). However, the insensitivity induced by transforming growth factor β1 (TGF-β1) demands most attention because of its extraordinary capacity at picomolar concentrations to ablate GC anti-inflammatory activity in airway and lung epithelial cells (Keenan et al., 2014, Salem et al., 2012). Moreover, we have recently identified that viral infection of human bronchial epithelial cells impairs GC anti-inflammatory activity through the induction of TGF-β1 activity (Xia et al., 2017).

Increased activity of TGF-β1 in epithelium during the progression of inflammatory airway diseases is well documented (Honkova et al., 2014, Torrego et al., 2007). TGF-β1 is established as a key driver of remodeling in chronic respiratory diseases (Lachapelle et al., 2018). However, elevated levels of TGF-β1 in asthmatic patients are not suppressed by the use of GCs (Chakir et al., 2003). Global inhibition of TGF-β1 results in serious adverse effects including autoimmune colitis and cardiac valve defects (Anderton et al., 2011), precluding the acceptability of agents that inhibit receptor activation for non-oncological or life-threatening indications (Lachapelle et al., 2018). We have previously excluded roles for Smad4-dependent nuclear Smad signaling and the well-known non-canonical kinase pathways in the signaling of TGF-β1-induced GC insensitivity (Keenan et al., 2014). Moreover, TGF-β1-induced impairment is not associated with impaired GRα expression, nuclear location, or altered glucocorticoid receptor alpha (GRα) phosphorylation in BEAS-2B cell line (Keenan et al., 2014). Therefore a hitherto undescribed pathway mediating GC insensitivity needs further investigation. The elucidation of this signaling operating in parallel to, but independently of, established canonical and known non-canonical pathways offers the enticing prospect of selective, clinically tractable approaches to TGF-β1 modulation.

Here, we show by unbiased transcriptomic RNA sequencing (RNA-seq) analyses that the GC insensitivity induced by TGF-β1 is broad, rather than being restricted to genes regulated by transactivation. Moreover, we identify phospho-cofilin1 as an essential link in the signaling chain subserving TGF-β1-induced GC insensitivity using a combination of proteomic, pharmacological, and genetic approaches. The upstream signaling revealed a pan class="Gene">LIMK2 pathway leading to the phosphorylation of cofilin1. PLD, a known downstream effector of phospho-cofilin1 (Han et al., 2007), was implicated in the induction of the nuclear hormone corepressor, SMRT, the distal effector of GC insensitivity. We show here for the first time that the products of PLD-mediated membrane phospholipid remodeling, lysophosphatidic acid (LPA) and phosphatidic acid (PA), induce GC insensitivity.

Results

TGF-β1 Has Extensive Modulatory Influence on GC Activity

To investigate the global impact of TGF-β1 on GC gene regulation, we conducted RNA-seq analyses of extracts of BEAS-2B cells (n = 3) incubated with TGF-β1 (40 pM), and then exposed to pan class="Chemical">dexamethasone (Dex, 30 nM), using concentrations previously established as maximally effective (Keenan et al., 2014, Salem et al., 2012, Xia et al., 2017). Genes influenced by TGF-β1 alone by more than 2-fold and those with smaller but significant changes (applying a Benjamini and Hochberg [BH] false discovery rate [FDR] < 0.05) were filtered from the analyzed set to avoid the confounding influence of direct TGF-β1 gene regulation. Based on these criteria, 569 genes were differentially affected by Dex (BH FDR<0.05). Of the remaining 245 genes suppressed by Dex, 181 were less suppressed in the presence of TGF-β1. Similarly, of the 324 genes induced by Dex, 213 genes were induced to a lesser extent by the combination of TGF-β1 and Dex (Figure 1A). However, 40 genes downregulated by Dex showed greater downregulation following TGF-β1 conditioning and 91 genes showed greater upregulation than induced by Dex alone (Figure 1A). Thus, TGF-β1 predominantly limited the magnitude of GC induction and repression of gene expression. The TGF-β1/Dex interaction could be ascribed to a functional antagonism, whereby the effects of Dex are simply opposed by the effects of TGF-β1 driving expression in the opposite direction, rather than being related to the signal-transduction-based interaction that we propose. The exclusion of genes significantly affected by TGF-β1 mitigates against, but does not preclude, this possibility. Thus we examined adding the TGF-β1 “arithmetic effect” to the significant Dex effect and compared this result to the actual impact of prior TGF-β1 conditioning on the subsequent Dex response for each of the 569 analyzed Dex changes in gene expression. The absolute mean difference in the theoretically additive gene expression and the combined gene expression effect is log2 0.55 ± 0.02 (p < 10−15), indicating that the interaction is not explained by any residual (non-significant) effects of TGF-β1 opposing those of Dex (Table S1). The molecular mechanism of GC-regulated gene expression is mainly through transactivation via binding to positive glucocorticoid response element (pGRE) DNA-binding sequences and direct repression via binding to negative GRE (nGRE) DNA-binding sequences and transrepression. We further analyzed whether the TGF-β1 impairment is confined to subsets of genes containing pGRE or nGRE or to those genes bearing neither of the consensus GRE motifs in their promoter regions. We annotated our genes using DNA sequences derived from the literature (Surjit et al., 2011). Although the nGRE-containing genes were under-represented in the set of genes that showed TGF-β1 impairment of Dex-induced gene expression, the impairment effect is observed with genes that express nGRE, pGRE, both, or neither (Table S2). Moreover, “FunRich” gene set enrichment analyses suggested that TGF-β1-modulated Dex-influenced genes across different biological pathways and processes (Table S3).

Figure 1

TGF-β1 Extensively Impairs GC Activity

(A) Analysis of transcriptomes of BEAS-2B cells incubated with Dex with or without TGF-β1 conditioning. Cells were pre-incubated with TGF-β1 (40 pM) for 24 h before stimulation with Dex (30 nM) or vehicle for 4 h. Data represent genes with a false discovery rate (FDR) <0.05 and not differentially affected by TGF-β1 alone (−1 0.05). Numbers of genes with specific patterns of GC-regulated expression are presented in the accompanying table together with graphical symbols.

(B) Quantitative real-time PCR (qRT-PCR) analysis of selected GC-inducible genes.

(C) Effects of TGF-β1 on pro-inflammatory cytokine levels regulated by GC. A549 cells were incubated with TGF-β1 (40 pM) for 12 h and incubated with Dex (10 nM) 30 min before incubation with IL-1α (1 ng/mL) for another 16 h. Levels of GM-CSF and IL-8 were measured by ELISA.

(D) Effects of TGF-β1 on the expression of GC-regulated proteins. BEAS-2B cells were incubated with TGF-β1 (40 pM) for 24 h and incubated with Dex (30 nM) for another 24 h.

Representative western blots are shown and aggregated data are presented as mean ± SEM. (A) n = 3, (B) n = 5–6, (C) n = 5, and (D) n = 6. N.D., not detectable. FC, fold control. Two-way ANOVA is used for (B); two-tailed Student's t test is used for (C, D). *p < 0.05, **p < 0.01, ***p < 0.001. (See also Tables S1, S2, and S3).

We investigated the underlying mechanism with a focus on the specific TGF-β1 impairment of GC-induced genes documented as effectors of GC anti-inflammatory activity and regulation of cytokine production. GC-induced mRNA encoding promyelocytic leukemia zinc finger (pan class="Gene">PLZF), FK506-binding protein 51 (FKBP5), and pyruvate dehydrogenase kinase 4 (PDK4) were suppressed by TGF-β1 (Figure 1B). GC induction of increased levels of the PLZF and FKBP5 proteins was also reduced by TGF-β1 (Figure 1D). TGF-β1 impairment of GC transrepressional mechanisms has been characterized in the A549 cell line (Salem et al., 2012), so we used the A549 cell line to investigate TGF-β1 impairment of GC transrepression. TGF-β1 attenuated the maximum inhibitory effect of Dex on IL-1α-induced IL-8 and granulocyte-macrophage colony-stimulating factor (GM-CSF) production in the A549 cell line (Figure 1C). The impact of these effects can be appreciated when considering that the levels of GM-CSF increase approximately 10-fold from 12 ± 5 pg/mL in the presence of IL1α/Dex to 106 ± 24 pg/mL, in cells additionally conditioned by TGF-β1. Thus our characterization indicates that TGF-β1 impairs GC-regulated gene expression changes in airway epithelial cells via mechanisms that transcend discrete transcriptional influences of GC, such as those dependent on nGRE, pGRE, or transcription factor repression.

Proteomic Analysis Reveals Phospho-cofilin1 as the Primary Candidate Mediator in TGF-β1-Induced GC Insensitivity

We had excluded roles for Smad4 and a wide range of non-canonical kinase signaling pathways in the TGF-β1 modulation of pan class="Chemical">steroid signaling, suggesting the involvement of a new pathway (Keenan et al., 2014). Candidate signaling proteins were assessed by an operational proteomic approach afforded by 2-dimensional differential in gel electrophoresis (2D-DIGE). We prioritized candidate protein expression changes by exploiting an unexpected difference in the activity of the different TGF-β gene products, whereby both TGF-β1 and TGF-β3 impaired GRE activity, but TGF-β2 was inactive in both BEAS-2B and A549 cells (Figure S1A). The bioactivity of TGF-β2 was validated not only by its induction of Smad7 (Figure S1B) but also by the induction of seven changes in 2D-DIGE protein spots (Figure 2A). The 2D-DIGE experiment in BEAS-2B cells treated with each of TGF-β1–3 revealed that of the 748 ± 35 (n = 5) protein spots detected, three spots were differentially expressed in extracts from TGF-β1- or TGF-β3-incubated cells, but not in those incubated with TGF-β2 (Figure 2A). Moreover, these changes in protein expression were insensitive to inhibition by a combination of mitogen-activated protein kinase (MAPK) and phosphatidylinositol 3-kinase (PI3K) pathway inhibitors that had been previously established to be individually ineffective on the TGF-β1-induced GC insensitivity (Keenan et al., 2014, Salem et al., 2012) and are now confirmed to be ineffective when used in combination (Figures S1C and S1D). These three spots were therefore prioritized for identification by picking and then subjected to tryptic digest before analysis by liquid chromatography-tandem mass spectrometry. The data were then analyzed by Mascot to identify the candidate proteins (listed in Table S4).

Figure 2

Proteomic Analyses Prioritize Phospho-cofilin1 As a Mediator in TGF-β1-Induced GC Insensitivity

(A) The 2D-DIGE comparative analysis between control and TGF-β isoforms. BEAS-2B cells were incubated with TGF-β isoforms (40 pM each) for 24 h, and lysates were subjected to 2D-DIGE. Overlay of Cy3 and Cy5 derived from a representative gel, highlighting differentially expressed protein spots: round for up-regulation; square for down-regulation. CyDye minimal labeling as follows: Control-Cy5 versus TGF-β1-Cy3; Control-Cy3 versus TGF-β2-Cy5; Control-Cy3 versus TGF-β3-Cy5. Venn diagram shows distribution of differentially expressed protein spots with TGF-β isoforms.

(B) Independent validation of TGF-β induction of phospho-cofilin1 by western blot.

(C) Western blot validation of successful knockdown of cofilin1 by cofilin1-targeted siRNA. (D) Effects of genetic ablation of cofilin1 on TGF-β1 impairment of GC-inducible gene expression. BEAS-2B cells were transfected with negative control or cofilin1-targeted siRNA, incubated with TGF-β1 (40pM) for 24h, and then incubated with Dex (30nM) for another 4h.

(E) Phospho-cofilin1 immunoreactivity in airway biopsies from subjects with increasing asthma severity scale bar, 30 μm.

Results are presented as mean ± SEM or mean with interquartile range. (A) n = 3–5, (B) n = 5, (C and D) n = 4–5, (E) n = 7–10. ns, not significant. Two-way ANOVA is used for (C, D); two-tailed paired Student's t test is used for (A, B); non-parametric Kruskal-Wallis is used for (E). *p < 0.05, **p < 0.01, ***p < 0.001. (See also Figures S1 and S2 and Table S4).

(A) The 2D-DIGE comparative analysis between control and TGF-β isoforms. BEAS-2B cells were incubated with TGF-β isoforms (40 pM each) for 24 h, and lysates were subjected to 2D-DIGE. Overlay of pan class="Chemical">Cy3 and Cy5 derived from a representative gel, highlighting differentially expressed protein spots: round for up-regulation; square for down-regulation. CyDye minimal labeling as follows: Control-Cy5 versus TGF-β1-Cy3; Control-Cy3 versus TGF-β2-Cy5; Control-Cy3 versus TGF-β3-Cy5. Venn diagram shows distribution of differentially expressed protein spots with TGF-β isoforms.

(C) Western blot validation of successful knockdown of cofilin1 by pan class="Gene">cofilin1-targeted siRNA. (D) Effects of genetic ablation of cofilin1 on TGF-β1 impairment of GC-inducible gene expression. BEAS-2B cells were transfected with negative control or cofilin1-targeted siRNA, incubated with TGF-β1 (40pM) for 24h, and then incubated with Dex (30nM) for another 4h.

(E) Phospho-cofilin1 immunoreactivity in airway biopsies from subjects with increasing pan class="Disease">asthma severity scale bar, 30 μm.

Phospho-cofilin1 was prioritized over the two other candidates because over-expression of cofilin1 had been reported to inhibit GR activity (Ruegg et al., 2004), and the cofilin1 levels of lymphocytes from steroid-insensitive asthmatic patients had been found to be inversely correlated with the ability of GC to suppress cytokine release (Vasavda et al., 2006). TGF-β1 and TGF-β3, but not TGF-β2, increased the phosphorylation of cofilin1 through MAPK-independent pathways (Figures 2B and S1E). Moreover, the elevation of phospho-cofilin1 levels by TGF-β1 was suppressed by 1 μM SB431542 (TβRI kinase inhibitor) (Figure S2A). Upon TGF-β1 treatment, phospho-cofilin1 protein levels increased, whereas Smad3 levels decreased (Figure S2B), consistent with the suggestion that TGF-β1 increased phospho-cofilin1 levels via a Smad3-independent, non-canonical kinase-dependent signaling pathway. However, further experiments are required to make this conclusion, given that the Smad3-targeted small interfering RNA (siRNA) did not further reduce the low Smad3 levels observed in the presence of TGF-β1 (Figure S2B). Incubation with cofilin1-targeted siRNA, which reduced cofilin1 protein levels to less than 20%, markedly diminished the level of TGF-β1-induced phospho-cofilin1 (Figure 2C). Cofilin1 siRNA completely prevented TGF-β1 inhibition of PLZF and FKBP5 gene expression (Figure 2D). The percentage of TGF-β1 suppression of Dex-induced ENaCα (epithelial sodium channel α) and PDK4 mRNA expression was reduced by cofilin1-targeting siRNA from 53% ± 5% to 37% ± 3% and from 54% ± 8% to 38% ± 5%, respectively. The relevance of phospho-cofilin1 was underlined by its elevated expression in the airway epithelium in biopsies from patients with severe, steroid-resistant asthma (see Table S5 for demographic data) compared with non-asthmatic subjects or those with mild-moderate asthma (Figure 2E).

TGF-β1-Induced Suppression of GC Activity Is due to the Activation of LIMK2

Cofilin1 phosphorylation is effected principally by the pan class="Gene">LIM domain kinase (LIMK). We therefore examined the effect of the putative upstream kinase LIMK using the small-molecule inhibitor, LIMKi3 (Scott et al., 2010). Cofilin1 phosphorylation was concentration-dependently reduced by LIMKi3 (1–10 μM). LIMKi3 (10 μM) completely prevented TGF-β1-induced elevation of phospho-cofilin1 levels (Figure 3A), but had no effect on Plasminogen activator inhibitor-1 (PAI-1) mRNA expression (Figure 3B), a Smad3-and Smad4-dependent TGF-β1 effect (Keenan et al., 2014). LIMKi3 prevented TGF-β1 suppression of GC induction of GILZ, PLZF, FKBP5, and PDK4 gene expression (Figure 3B) and partially restored ENaCα gene expression (Figure 3B) and GRE activity (Figures S3A and S3B). Reduction of LIMK2 expression by approximately 50% using siRNA attenuated TGF-β1 impairment of GC-regulated PLZF and FKBP5 expression (Figure 3C), whereas a 70% reduction of LIMK1 had no effect (Figure S3C). The differential influence of LIMK1 and LIMK2 on the TGF-β1-impaired GC activity is consistent with accumulating evidence demonstrating the potential association of LIMK2 level, but not LIMK1, with asthma severity in inflammatory cells (Aoki et al., 2009, Bigler et al., 2017, Persson et al., 2015).

Figure 3

TGF-β1-Induced Impairment of GC Activity Is Dependent on LIM Kinase 2 Activity

(A) Pharmacological inhibition of LIMK inhibits TGF-β1-induced increase in phosho-cofilin1 levels. BEAS-2B cells were incubated with LIMKi3 (1–10 μM) for 30 min before incubation with TGF-β1 (40 pM) for 24 h.

(B) Smad-dependent PAI-1 and GC-inducible gene expression targets were determined by qRT-PCR in BEAS-2B cells incubated with LIMKi3 (10 μM) for 30 min before TGF-β1 (40 pM) for 24 h followed by incubation with Dex (30 nM) for 4 h.

(C) BEAS-2B cells transfected with negative control or LIMK2-targeted siRNA were incubated with TGF-β1 (40 pM) for 24 h, followed by incubation with Dex (30 nM) for 4 h before mRNA measurement.

(D) Levels of GM-CSF and IL-8 were measured by ELISA and expressed as a percentage of the IL-1α response. A549 cells were incubated with LIMKi3 (10 μM) for 30 min before TGF-β1 (40 pM) for 12 h and incubated with Dex (10 nM) 30 min before incubation with IL-1α (1 ng/mL) for another 16 h.

(E) Effects of pharmacological inhibition of LIMK on TGF-β1 impairment of GC activity in human bronchial epithelial cells differentiated for 28 days at air-liquid interface (ALI). Cells were incubated with LIMKi3 (10 μM) 30 min before incubation with TGF-β1 (40 pM) for 24 h, then incubated with Dex (100 nM) or vehicle control for 4 h.

Data are presented as mean ± SEM from n = 4 individual donors. (A–D) n = 5–6 (E) = 4. ns, not significant. Two-way ANOVA is used for (A–D); two-tailed paired Student's t test is used for (E). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. (See also Figure S3).

(B) Smad-dependent PAI-1 and GC-inducible gene expression targets were determined by qRT-PCR in pan class="CellLine">BEAS-2B cells incubated with LIMKi3 (10 μM) for 30 min before TGF-β1 (40 pM) for 24 h followed by incubation with Dex (30 nM) for 4 h.

(C) BEAS-2B cells transfected with negative control or pan class="Gene">LIMK2-targeted siRNA were incubated with TGF-β1 (40 pM) for 24 h, followed by incubation with Dex (30 nM) for 4 h before mRNA measurement.

(D) Levels of GM-CSF and pan class="Gene">IL-8 were measured by ELISA and expressed as a percentage of the IL-1α response. A549 cells were incubated with LIMKi3 (10 μM) for 30 min before TGF-β1 (40 pM) for 12 h and incubated with Dex (10 nM) 30 min before incubation with IL-1α (1 ng/mL) for another 16 h.

In A549 cells, IL-1α-induced elevation of pan class="Gene">GM-CSF level was completely prevented by Dex. TGF-β1 enhanced IL-1α-induced increases in GM-CSF levels and compromised the Dex suppression of GM-CSF levels. Inhibition of LIMK using LIMKi3 (10 μM) restored much of the activity of Dex in suppressing both GM-CSF and IL-8 levels (Figure 3D).

We have previously established the TGF-β1 modulation of GC in air-liquid interface (ALI) differentiated primary human epithelial organotypic cultures (Keenan et al., 2014). This organotypic culture setting engendpan class="Disease">ers a differentiated cell population expressing functional activity, such as beating cilia and mucous production, and an intact epithelial barrier. In ALI cultures, TGF-β1 inhibition of Dex-induced GILZ and ENaCα expression was prevented by the inhibition of LIMK (Figure 3E).

There is a well-established connection between the TGF-β1 receptor and RhoA/B and GTP-bound Cdc42 activation (Kardassis et al., 2009). Moreover, LIMK activation is widely thought to result from phosphorylation by ROCK or Cdc42/PAK (Lee-Hoeflich et al., 2004, Sumi et al., 2000). These TGF-β1-activated pathways have been documented in a signaling chain in Swiss 3T3 fibroblasts (Vardouli et al., 2005), but are not yet evidenced in an epithelial cell phenotype. Hence, we investigated the effects of blocking ROCK (10 μM Y27632), Cdc42 (5 μM ML141), or both using well-validated inhibitors at concentrations that have been previously reported by other groups to effectively inhibit the respective enzymatic activities (Hong et al., 2013, Sumi et al., 2000). Neither the ROCK inhibitor or the Cdc42 inhibitor alone nor their combination was able to even partially mimic the effect of inhibiting LIMK (Figures 4A and S4A).

Figure 4

TGF-β1-Induced Impairment of GC Activity Is Not Due to the Activity of ROCK/Cdc42 or TGF-β1-Induced Actin Reorganization

(A) BEAS-2B cells were incubated with the ROCK inhibitor Y27632 (10 μM) or the Cdc42 inhibitor ML141 (5 μM) for 30 min before TGF-β1 (40 pM, 24 h incubation) followed by Dex (30 nM) incubation for 4 h before mRNA measurement.

(B) Fluorescence imaging of F-actin (green) and G-actin (red) in BEAS-2B cells incubated with LIMKi3 (10 μM), Y27632 (10 μM), and ML141 (5 μM) 30 min before TGF-β1 (40 pM) for 24 h; scale bar, 50 μm.

Data are presented as mean ± SEM. (A) n = 3–6, (B) n = 3–6. Two-way ANOVA is used throughout. *p < 0.05, **p < 0.01, ***p < 0.001. ns, not significant. (See also Figure S4).

(A) BEAS-2B cells were incubated with the ROCK inhibitor pan class="Chemical">Y27632 (10 μM) or the Cdc42 inhibitor ML141 (5 μM) for 30 min before TGF-β1 (40 pM, 24 h incubation) followed by Dex (30 nM) incubation for 4 h before mRNA measurement.

Cofilin1 phosphorylation at pan class="Chemical">serine 3 by LIM domain kinase switches off the actin filament severing activity of cofilin1 (Vardouli et al., 2005). However, neither LIMKi3 (10 μM) nor the inhibitor of Cdc42 ML141 (5 μM) affected the TGF-β1-induced re-distribution of actin between F and G forms. The ROCK inhibitor Y27632 (10 μM) prevented TGF-β1-induced increases in F-actin levels (Figure 4B), as did the F-actin-disrupting agent, cytochalasin B (Figure S4B), but neither Y27632 nor cytochalasin B restored GC-inducible gene expression (Figures 4A and S4C). These findings dissociate actin cytoskeletal changes induced by TGF-β1 from impaired GC responses.

TGF-β1 Impairment of GC Activity Is due to Phospho-cofilin1-Dependent Phospholipase D Activation

Although cofilin1 is widely regarded as the active form of the protein, phospho-pan class="Gene">cofilin1 has also been reported to activate phospholipase D1 (PLD1) (Han et al., 2007). We therefore examined the possible involvement of PLD in TGF-β1 signaling of GC insensitivity, using the selective inhibitors VU01550699 (7.5 μM) or FIPI (10 μM), at concentrations based on previous studies (Kang et al., 2013, Su et al., 2009) and as further validated in our initial experiments. Incubation of BEAS-2B cells with VU01550699 (7.5 μM) prevented TGF-β1 suppression of Dex-induced expression of GILZ, ENaCα, and FKBP5 and substantially enhanced Dex-induced GILZ mRNA expression (Figure 5A). Use of FIPI (10 μM) further confirmed the involvement of the PLD in TGF-β1 suppression of GC-regulated gene expression (Figure 5B). Moreover, both VU01550699 and FIPI reduced TGF-β1 impairment of Dex-induced GRE activity (Figure S5A). PLD was further implicated by the restoration of GC responsiveness through the use of PLD1/2-targeted siRNA, with combined knockdown of PLD1 and PLD2 attenuating the TGF-β1 inhibition of Dex-induced FKBP5 and PLZF mRNA expression (Figure 5C). Individual PLD1 and PLD2 knockdown each attenuated the inhibitory effects of TGF-β1 on GILZ and PLZF mRNA expression (Figure S5B). The TGF-β1 impairment of Dex suppression of pro-inflammatory cytokines GM-CSF and IL-8 levels was prevented by the inhibition of PLD by VU01550699 (7.5 μM) (Figure 5D), as was the TGF-β1-induced inhibition of GILZ and ENaCα expression in the ALI cultures (Figure 5E). These data strongly supported a role for PLD in the signaling downstream of phospho-cofilin1 to mediate TGF-β1 impairment of GC activity.

Figure 5

Inhibition of Phospholipase D (PLD) Prevents TGF-β1 Induction of GC Insensitivity

(A and B) Effects of PLD inhibitors (A) VU0155069 (VU01, 7.5 μM) or (B) FIPI (10 μM) on TGF-β1-induced suppression of GC-inducible gene expression was investigated.

(C) Effects of genetic depletion of PLD1/2 on TGF-β1-suppressed GC activity was investigated by qRT-PCR.

(D) Levels of GM-CSF and IL-8 were measured by ELISA and expressed as a percentage of the IL-1α response. A549 cells were incubated with VU0155069 (7.5 μM) for 30 min before TGF-β1 (40 pM) for 12 h and incubated with Dex (10 nM) 30 min before incubation with IL-1α (1 ng/mL) for another 16 h.

(E) Effects of pharmacological inhibition of PLD on TGF-β1 impairment of GC activity in human bronchial epithelial cells differentiated for 28 days at air-liquid interface (ALI).

(F and G) Effects of exogenous PA and LPA on Dex-inducible gene expression. BEAS-2B cells were exposed to (F) PA (10 or 30 μg/mL) or (G) LPA (10 or 30 μM) for 24 h, before incubation with Dex (30 nM) for 4 h.

Data are presented as mean ± SEM. (A–D) n = 4–5, (E–G) n = 4. ns, not significant. Two-way ANOVA is used for (A–D; F–G); two-tailed paired Student's t test is used for (E). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. (See also Figure S5).

Inhibition of Phospholipase D (pan class="Gene">PLD) Prevents TGF-β1 Induction of GC Insensitivity

(D) Levels of GM-CSF and pan class="Gene">IL-8 were measured by ELISA and expressed as a percentage of the IL-1α response. A549 cells were incubated with VU0155069 (7.5 μM) for 30 min before TGF-β1 (40 pM) for 12 h and incubated with Dex (10 nM) 30 min before incubation with IL-1α (1 ng/mL) for another 16 h.

Activation of PLD increases the hydrolysis of pan class="Chemical">phosphatidylcholine (PC) to choline and the second messenger lipid PA, which may be further metabolized to LPA and diacylglycerol (Csaki et al., 2013). Exogenous PLD products, PA and LPA, mimicked the effect of TGF-β1 in repressing the Dex-induced GRE activity (Figure S5C) and in decreasing GRE-regulated genes (Figures 5F and 5G), without cytotoxicity (Figure S5D). The mimicry of TGF-β1 by exogenous (lyso)phosphatidic acid could suggest that these PC metabolites were being mobilized as autocrine mediators of TGF-β1. This possibility was examined by the transfer of TGF-β1-conditioned supernatant to TGF-β1-naive BEAS-2B cells. BEAS-2B cells were incubated with TGF-β1 (40 pM) for 6 or 24 h to generate conditioned medium. Before transfer to serum-starved BEAS-2B cells, the ALK5 inhibitor SB431542 (10 μM) was added to the medium to block ALK5 signaling induced by the transferred TGF-β1. The conditioned medium did not contain a transferable GC suppressive factor (Figures S6A and S6B), leading us to conclude that the (lyso)phosphatidic acid induced by TGF-β1 acts within the intracellular compartment. Phospholipid remodeling could allow LPA access to LPA receptors, whereas blockade of LPA receptors (1–3) using Ki16425 (10 μM) (Ohta et al., 2003) had no effect on the impairment of GC-induced gene expression caused by either TGF-β1 or (lyso)phosphatidic acid (Figure S6C).

SMRT Induction Downstream of Phospho-cofilin1/PLD Mediates TGF-β1-Induced Repression of GR Target Gene Expression

The repressive activities of TGF-β1 and PA on peroxisome proliferator-activated receptor (PPAR) have been previously ascribed to stabilization of the pan class="Gene">PPAR interaction with the corepressor SMRT in macrophages and fibroblasts (Stockert et al., 2011, Tsukahara et al., 2010). We therefore analyzed whether TGF-β1 influences SMRT in BEAS-2B cells. TGF-β1 significantly increased SMRT levels, an effect that was dependent on the action of cofilin1 and PLD (Figures 6A and 6B). SMRT-targeted siRNA prevented the TGF-β1 induction of SMRT protein levels (Figure 6C), preventing TGF-β1 suppression of PLZF and FKBP5 mRNA, and attenuating suppression of ENaCα (percentage inhibition reduced from 67% ± 4% to 50% ± 5%) and PDK4 (percentage inhibition reduced from 65% ± 3% to 47% ± 4%) mRNA expression, respectively (Figures 6D, S7A, and S7B). Moreover, TGF-β1, PA, and LPA each completely prevented the transcriptional regulation of CYP27A1 (encoding cytochrome P450 family 27 subfamily A member 1) by the PPARγ agonist, rosiglitazone (Figure S8), consistent with a role for intracellular PA in mediating TGF-β1 inhibition of the nuclear hormone receptors, GR and PPARγ.

Figure 6

SMRT Induction Downstream of Phospho-cofilin1/PLD Activity Plays a Role in the TGF-β1-Induced Repression of GR Target Genes

(A and B) Effects of genetic deletion of (A) cofilin1 and (B) PLD1/2 on SMRT expression. Cell lysates were the same as those used in Figures 2D and 5C.

(C) Validation of successful knockdown of SMRT-targeted siRNA in BEAS-2B cells. (D) Effects of genetic ablation of SMRT on GC-inducible gene expression. BEAS-2B cells transfected with negative control or SMRT-targeted siRNA, were incubated with TGF-β1 (40pM) for 24h, and then incubated with Dex (30nM) for another 4h before mRNA measurement.

Data are presented as mean ± SEM. (A–D) n = 5. ns, not significant. Two-way ANOVA is used for (A–D). *p < 0.05, **p < 0.01, ***p < 0.001. (See also Figure S7).

SMRT Induction Downstream of Phospho-pan class="Gene">cofilin1/PLD Activity Plays a Role in the TGF-β1-Induced Repression of GR Target Genes

(C) Validation of successful knockdown of SMRT-targeted siRNA in pan class="CellLine">BEAS-2B cells. (D) Effects of genetic ablation of SMRT on GC-inducible gene expression. BEAS-2B cells transfected with negative control or SMRT-targeted siRNA, were incubated with TGF-β1 (40pM) for 24h, and then incubated with Dex (30nM) for another 4h before mRNA measurement.

Discussion

We identified a broad modulatory impact of TGF-β1 on GC activity, showing reduced maximal responses of both GC-induced and GC-repressed genes. A TGF-β1 signaling pathway was elucidated comprising LIMK2-mediated phosphorylation of pan class="Gene">cofilin1 that requires PLD1 and PLD2 enzyme activity and induction of the nuclear hormone corepressor SMRT (Figure 7). The increased expression of phospho-cofilin1 in the airway epithelium of severe, steroid-resistant asthmatic patients further supported the relevance of our findings, which we believe will facilitate drug discovery efforts focused on safely blocking the contribution of TGF-β1 to chronic inflammatory diseases (Lachapelle et al., 2018).

Figure 7

LIMK2-Mediated Cofilin1 Phosphorylation Signals TGF-β1-Induced GC Insensitivity

Phospho-cofilin1 is implicated as a key link in the chain of signalling behavior of TGF-β1-induced GC insensitivity. The activation of LIMK2 appears to be independent of known upstream regulators ROCK or Cdc42. Phospho-cofilin1-activated PLD mediates SMRT induction, which limits GR target gene regulation. The interventions including small-molecule inhibitors and siRNA are summarized in boxes.

LIMK2-Mediated pan class="Gene">Cofilin1 Phosphorylation Signals TGF-β1-Induced GC Insensitivity

Phospho-cofilin1 is implicated as a key link in the chain of signalling behavior of TGF-β1-induced GC insensitivity. The activation of pan class="Gene">LIMK2 appears to be independent of known upstream regulators ROCK or Cdc42. Phospho-cofilin1-activated PLD mediates SMRT induction, which limits GR target gene regulation. The interventions including small-molecule inhibitors and siRNA are summarized in boxes.

The relative importance of transactivation and transrepression to the anti-inflammatory activity of GC remains controversial (Keenan et al., 2015). GR dimer-dependent transactivation is increasingly regarded as essential for certain of the anti-inflammatory activities of GR (Vandevyver et al., 2013). As TGF-β1 suppressed GC-sensitive genes associated with pGRE, nGRE, transrepression, or other mechanisms of regulation with similar frequencies, the exogenous TGF-β1 effect is not limited in scope to a specific GC gene regulatory mechanism. We selected a panel of GRE-dependent GC-responsive genes, including GILZ, ENaCα, PLZF, PDK4, FKBP5, as well as the cytokines, GM-CSF and IL-8, to assess the impact of TGF-β1 on GC sensitivity. This selection was based on their importance in inflammation, as described briefly below, and having comparative data from prior studies (Keenan et al., 2014, Salem et al., 2012, Xia et al., 2017). GILZ mediates some of the anti-inflammatory effects of Dex in epithelial cells (Eddleston et al., 2007). Impaired epithelium sodium channel expression results in excess fluid and pulmonary edema (Kerem et al., 1999). FKBP5 is not only a biomarker of GC responsiveness but also is implicated in modulating the nuclear translocation and transcriptional activity of nuclear factor-κB (Erlejman et al., 2014). PLZF is a GC-inducible transcription factor that arrests cell cycle by inducing the expression of CDK inhibitors (Naito et al., 2015). PDK4 inactivates the pyruvate dehydrogenase complex (PDC), which catalyzes the conversion of pyruvate to acetyl-CoA. PDC metabolite flux is involved in the immunometabolic reprogramming of macrophages, a process that allows cellular defense mechanisms to accommodate the different microenvironments encountered in inflamed tissues (O'Neill and Pearce, 2016). Imparied GC induction of GILZ, ENaCα, FKBP5, PLZF, and PDK4 could amplify inflammation in the airway wall in asthma.

TGF-β1 impaired GC regulatory effects on the pro-inflammatory cytokines, IL-8 and pan class="Gene">GM-CSF. IL-8, a powerful chemoattractant and activator of neutrophils, is detected at increased levels in severe asthma (Shannon et al., 2008). Neutrophils are generally considered to be less directly sensitive to GC treatment. In GC-resistant asthma, inhibition of IL-8 release by GC is impaired (Nair et al., 2015, Wang et al., 2016). GM-CSF facilitates robust eosinophilic inflammation, long-lasting antigen-specific Th2 inflammation, mucous production, and airway hyperresponsiveness (Llop-Guevara et al., 2014). The compromised GC regulation of these inflammogens in epithelia conditioned by TGF-β1 is likely to contribute to the severity and limit the therapeutic responsiveness of chronic inflammation to GCs.

Global inhibition of TGF-β1 or ALK5 activity is known to increase the risk of pan class="Disease">autoimmune disorders and heart valve lesions (Anderton et al., 2011). Therefore further understanding of the TGF-β1 signaling cascades that interfere with GC activity may identify targets downstream of the receptors for selective restoration of GC responses, which avoid the detrimental effect of global TGF-β1 inhibition (Lachapelle et al., 2018). We used a hypothesis-free proteomic approach to examine the signaling of GC resistance by TGF-β1 because hypothesis-driven approaches, including the use of a wide panel of kinase inhibitors and effective Smad4 inhibition (Keenan et al., 2014, Salem et al., 2012), had been exhausted without success. Three candidate signaling proteins were prioritized by (1) inclusion of protein changes common to the active TGF-β forms, TGF-β1 and TGF-β3; (2) exclusion of changes induced by TGF-β2, which did not induce GC impairment; and (3) exclusion of changes that were prevented by the combination of MAPK and PI3K-targeting small-molecule inhibitors that failed to block TGF-β1 impairment of GC responses. Phospho-cofilin1 was further prioritized based on studies implicating cofilin1 in steroid sensitivity (Ruegg et al., 2004, Vasavda et al., 2006).

As one of the main activities of cofilin1 is actin severing, our attention turned to the possibility that TGF-β1 impairment of GC activity could be secondary to changes in the actin cytoskeleton, and possibly to influences on cellular mechanics (Krishnan et al., 2016). However, pharmacological inhibitors of pan class="Gene">LIMK and ROCK/Cdc42 showed differential effects on TGF-β1-induced phosphorylated-cofilin1, GC insensitivity, and actin polymerization. The ROCK inhibitor and cytochalasin B prevented TGF-β1 induction of F-actin levels without restoring GC activity, whereas the LIMK inhibitor prevented cofilin1 phosphorylation and restored GC activity, without effect on TGF-β1-induced F-actin. These observations do not support a role for actin cytoskeletal assembly in the regulation of GC sensitivity.

Although phospho-cofilin1 had been widely considered to be inactive, it was subsequently convincingly demonstrated to be an activator of pan class="Gene">PLD1 (Han et al., 2007, Han et al., 2011). Aberrant PLD activity has been identified in chronic inflammatory autoimmune disorders and in allergic asthma (Choi et al., 2015, Kang et al., 2013). Moreover, TGF-β1 increases PLD activity and increases the levels of PA and PA-related metabolites in epithelial cell lines (Zhou et al., 2000). We sought to establish whether there is a link between PLD and the TGF-β1 suppression of GC activity. TGF-β1 suppression of GC activity in epithelial cells was prevented by reducing the activity of PLD, consistent with a pivotal role for phospho-cofilin1-activated PLD in TGF-β1 impairment of GC activity (Figure 7). Importantly, exogenous PA and LPA mimicked the effects of TGF-β1 on a panel of Dex-inducible genes and GRE activity. This observation linking PA or LPA and GC sensitivity has implications beyond TGF-β1 to a broader relationship between PLD products and GC sensitivity. The levels of polyunsaturated LPA in human bronchoalveolar lavage (BAL) fluid increase 3-fold after segmental allergen challenge (to ∼1.5 μM) (Georas et al., 2007). When accounting for ∼100-fold dilution of airway surface fluid by the instilled lavage fluid, the epithelium may be exposed to LPA in situ of over 100 μM, which exceeds the concentrations of exogenous LPA and PA required to mimic GC resistance. LPA present in biological fluids, such as plasma and BAL, is thought to initiate effects in airway inflammation and remodeling by activation of cell surface G-protein coupled LPA1–7 receptors (Zhao and Natarajan, 2013), leading to the question of whether TGF-β1 mobilized LPA to exert intra- or extracellular effects. However, a series of careful supernatant transfer experiments indicated that the TGF-β1 impairment of GC activity was not accompanied by the release of transferable activity in the conditioned medium. Moreover, blockade of LPA receptors had no effect on the impairment of GC activity caused by either TGF-β1 or exogenous PA or LPA. Therefore extracellular activation of LPA receptors is less likely to account for the TGF-β1 effect than intracellular mobilization and action of PLD products.

Both PA and LPA are reported to bind to certain nuclear hormone receptors, including pan class="Gene">PPARγ (McIntyre et al., 2003). PA has been shown to inhibit PPARγ transactivation with high potency and high specificity, by stabilizing the interaction of PPARγ with the corepressor SMRT (Tsukahara et al., 2010). Moreover, PA has a biphasic influence on epidermal growth factor receptor (EGFR) expression, whereby lower concentrations of PA enable and higher concentrations impede PPARα/LXRα binding to the EGFR promoter (Henkels et al., 2016). We have now shown that the corepressor SMRT is a transcriptional target of TGF-β1 through a chain of signaling linking CK1δ/ɛ to phospho-cofilin1-activated PLD. The siRNA-mediated inhibition of SMRT leads to a strong derepression of TGF-β1-targeted GR gene expression, suggesting that SMRT signals GC resistance downstream of the activated PLD.

In summary, we establish that TGF-β1 potently induces insensitivity to GC activity. Pharmacological and genetic approaches implicated a chain of signaling through LIMK2/phospho-cofilin1/PLD, liberating PA and LPA to induce GC insensitivity through the induction of SMRT. Our findings identify new drug targets to modulate TGF-β1 that may be amenable to safe modulation by avoiding interference in essential growth and immunoregulatory TGF-β1 activities that are mediated by canonical and well-established non-canonical pathways operating upstream or independently of this pathway.

Limitations of the Study

We have provided compelling evidence that SMRT links TGF-β1 and GC activity, acting through a chain of signaling involving pan class="Gene">LIMK2 to phospho-cofilin1-activated PLD to induce GC insensitivity. However, the detailed mechanism underlying ALK5-mediated increases in the level of phospho-cofilin1 and PLD-mediated increases in the level of SMRT remain unsolved. We aim to address the translational significance of our findings in future studies designed to better understand the physiological functions of the identified novel non-canonical TGF-β1 signaling pathway in steroid-resistant airway diseases.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file