TGFβ1-driven SMAD2/3 phosphorylation and myofibroblast emergence are fully dependent on the TGFβ1 pre-activation of MAPKs and controlled by maternal leucine zipper kinase

Abstract

Following wounding, endogenously secreted TGFβs drive resident and bone marrow-derived cells to convert into α-smooth actin (SMA)-rich, contractile myofibroblasts. The TGFβ effect is initiated by the phosphorylation of SMADs 2 and 3 (SMAD2/3). This event has been referred to as the canonical response to TGFβ. TGFβ also elicits other responses viewed as parallel events not directly connected to the SMAD activation, and thus referred to as noncanonical. A recognized response is the phosphorylation of the -activated kinase (TAK1/MAP3K), an upstream component of the mitogen-activated protein kinase (MAPK) cascade. We have now examined the relationship between these two effects of TGFβ1 at their earliest stages. The bulk of the studies were carried out with primary fibroblasts derived from the human cornea. The results’ widespread relevance was confirmed in critical experiments with dermal-, and Tenon’s capsule-derived fibroblasts. Cells were treated with kinase inhibitors or targeting siRNAs followed by induction by 2ng/ml TGFβ1, and/or 10 ng/ml TNF-α. Cells were collected after 1 to 30 min for Western blot analysis and assayed for the accumulation of phosphorylated TAK1, ASK1, JNK1/2, p38, HPS27, MELK, SMAD2/3, and GAPDH. The effect of the kinase inhibitors on α-SMA expression and α-SMA stress fiber organization was also tested. For the immediate response to TGFβ1 we found that a) activation of the MAPK pathway was completed within 1 min after the addition of TGFβ1; b) phosphorylation of JNK1/2 was fully dependent on TAK1 and ASK1 activity, c) phosphorylation of MELK was fully dependent on JNK1/2 activity; d) phosphorylation of ASK1 depends on MELK activity, indicating the existence of an ASK1-MELK positive activation feedback loop; e) phosphorylation of SMAD2/3 started only after a 5 min period and reached a nadir after 10–15 min, f) the latter phosphorylation was fully blocked by inhibition of TAK1, ASK1, JNK1/2, and MELK, and siRNA-driven MELK downregulation; g) the inhibitors equally blocked the α-SMA protein expression, stress fiber development, and cell morphology changes at 72 h. These results demonstrate that the activation of the canonical pathway is fully subordinate to the activity of the MAPK pathway, challenging the concept of canonical and noncanonical TGFβ pathways and that SMAD2/3 activation is mediated by MELK, a kinase not previously associated with rapid pharmacological responses.

1. INTRODUCTION

In both corneal and cutaneous wounds, endogenously secreted TGFβ, leads to the conversion of local and bone-marrow-derived myofibroblasts.^1–10 Myofibroblasts are characterized by the expression and organization of α-smooth muscle actin (α-SMA), a protein that confers to the myofibroblast a mechanical strength for wound contraction. Additionally, myofibroblasts secrete enhanced amounts of extracellular matrix proteins. The myofibroblast conversion plays an essential role in the contractile force needed to achieve rapid wound closure to maintain barrier function. However, the persistence of myofibroblasts in a healing wound results in scar formation due to the continued presence of the cells and the disordered nature of the secreted ECM. Such an imperfect outcome is particularly morbid in the case of the cornea, where scarring critically interferes with normal vision.

The cellular-molecular events associated with the human corneal stromal keratocyte to myofibroblast conversion can be conveniently studied in cells derived from human cadaver corneas. The corneal stromal cells (keratocytes) are isolated and grown in culture in the presence of serum, promoting the differentiation into fibroblasts. However, if serum is withdrawn and the replacement medium is complemented with transforming growth factor beta 1 (TGFβ1), the cells convert into myofibroblasts within 48–72 h. Prominent myofibroblast biomarkers include α-SMA, organized as prominent intracellular fibers, two-dimensional cell area increase, and secretion of collagen type 1. This in-culture response recapitulates several features of in situ corneal wound healing.

In terms of signal transduction, TGFβ1 induces the formation of an activated TGFβR1- TGFβR2 receptor complex. Once complexed, the TGFβR2 kinase phosphorylates SMAD2 and SMAD3 (SMAD2/3), binding to SMAD4. The complex migrates to the nuclei, where it orchestrates the gene expression events that underpin the fibroblast to myofibroblast phenotype change. These events are commonly known as the TGFβ1 canonical response.^12–14 In addition, TGFβ is known to elicit other, noncanonical, cellular responses. The most recognizable such reaction is the phosphorylation of the TGFβ1-activated kinase (TAK1), an upstream component of the mitogen-activated protein kinase (MAPK) cascade. This activation can be expected to result in the phosphorylation of several downstream components of the cascades, including the kinases p38 and JNK. ¹⁵

This distinction between a canonical and noncanonical response to TGFβ reflects the established view that the initial signal transduction responses are initially independent of each other. In the present report, we show that, in human corneal-derived fibroblasts (HCFs), the phosphorylation of SMAD2/3 in response to TGFβ1 is characterized by rapid activation of the MAPK cascade through the upstream TAK1/MAP3K7 and ASK1/MAP3K5 kinases, which is a noncanonical response. Whereas the canonical response, phosphorylation of SMAD2/3 occurs only after a previously unrecognized five min initiation delay. Furthermore, our study shows that contrary to the concept of independence of pathways, phosphorylation of SMAD2/3 and the subsequent emergence of the myofibroblast phenotype is fully dependent on the activation of the noncanonical MAPK branch of the response, specifically, by Maternal embryonic leucine zipper kinase (MELK/murine MPK38).^16,17

2. MATERIALS and METHODS

2.1. Cell culture

Human corneas from unidentifiable cadaver donors were obtained from the Eye-bank for Sight Restoration, NY, NY. The Icahn School of Medicine Institutional Review Board has determined that as per section 45 CFR Part 46 of the U.S.A. Code of Federal Regulations unidentifiable cadaver tissue does not constitute research in Human subjects (see http://grants.nih.gov/grants/policy/hs/faqs_specimens.htm?Display=Graphics for further information). Minute biopsies (0.1–0.2 mm²) of Tenon’s capsule fibroblasts were obtained from patients undergoing open-angle glaucoma trabeculectomy under a protocol approved by the Icahn School of Medicine at Mount Sinai Institutional Review Board and informed consent from the tissue donor.

Corneal stromal and Tenon’s capsule fibroblasts (HCF) were isolated by explant culture. The corneal epithelium and endothelium were removed and the stroma was cut into approx. 1×1 mm sections. These sections and the Tenon’s capsule biopsies were cultured for 1 to 2 weeks in Dulbecco’s Modified Eagle’s Medium/Ham’s Nutrient Mixture F-12 (DMEM/F12; Gibco, Grand Island, NY) with 10 % fetal bovine serum (FBS; Gibco) and an antibiotic/antimycotic mix (Sigma, St. Louis, MO). The outgrown cells trypsinized and recultured for up to the 10^th passage. However, unless specified otherwise, experiments were conducted with the 2^nd to 4^th passage cells. Human adult dermal fibroblasts (HDF) were purchased from ATCC (Manassas, VA) and maintained in low serum growth medium (ATCC). Prior to experimentation, the medium was complemented with 5 % FBS for 48 h.

2.2. Experimental cell protocols and reagents

Cells were seeded and cultured for 48 h in supplemented serum-free medium (SSFM; DMEM/F12 plus 1X RPMI-1640 Vitamin Mix; 1X ITS Liquid medium supplement; 1 mg/ml glutathione; 2 mM L-glutamine; 1 mM sodium pyruvate; 0.1 mM MEM non-essential amino acids (Gibco) and ABAM (Sigma, St. Louis, MO). Unless indicated otherwise, experiments were initiated by the addition of the inhibitors described in Table 1A, followed 10 min later by 2 ng/ml TGF-β1 (R&D Systems, Minneapolis, MN). Cells were collected within 30 min after the addition of either TGFβ1 (Fisher, 240B002), TNFα (Fisher, NC1675619 ) or LPS (Fisher, 501122025), and the phosphorylation of, TAK1, ASK1, JNK1/2, p38, MELK, SMAD2/3, SMAD2 or SMAD3 was determined by Western blot using the antibodies described in Table 1 B. The accumulation of α-SMA was determined after 72 h incubation. In addition, MELK expression was downregulated using two different targeting siRNAs, a pooled set of siRNAs (Santa Cruz, Dallas, Tx; cat sc-610176) or an esiRNA preparation (Sigma, St. Louis, Mo; cat EHU030401 ). Trypsin-harvested cells were resuspended in 100 μl P3 electroporation solution (Lonza, Walkersville, MD) containing either 1 μM MELK siRNA, 1 μM scrambled siRNA, or 0.3 μM of the esiRNA, and subjected to electrophoresis in a 4D Nucleofector instrument (Lonza), using a pre-optimized setting. After electroporation, the cells were seeded in SSFM for 48 h prior to the addition of reagents.

Table 1.

A. Kinase inhibitors used in this studv. B. Antibodies used in this study.

	Target	Inhibitor	Supplier	IC50	Reference
A.
1	ΤAΚ1	5z-7oxoxeaenol	Tocris	8 nM	18
2	ΤAΚ1	TAKinib	Fisher	9 nM	19
3	ASK1	Selonsertib	Cayman	5 nM	20
4	ASK1	ASK1–110	Cayman	24 nM	21
5	JNK1/2	SP600125	Cayman	40–90 nM	22
6	JNK1/2	JNK inhib. VIII	Cayman	720/720 nM	23
7	JNK1/2	JNK inhib. XVI	Cayman	4.7/18.7 nM	23
8	MELK	OTSSP167	Cayman	0.42 nM	24
9	MELK	MELK8A	Medchemexpress	2 nM	25
10	p38	Birb796	Fisher	0.1 nM	26
11	p38	VX-745	Cayman	45–51 nM	27
12	p38	JX-401	Cayman	32 nM	28
13	p38	p38 Inhib-VIII	Cayman	39 nM	29
14	ERK1/2	FR180204	Cayman	310 nM	30
15	TFGR1	LY 364947	Cayman	59 nM	31
B.
1	p.TAK1	Ser412	Cell Sig./9339	Rabbit	1:500
2	p.TAK1	Thr184/187	Cell Sig. /4508	Rabbit	1:500
3	p.TAK1	Thr184/187	Fisher/PIMA515073	Rabbit	1:500
4	p-TAK1	Thr187	Cell Sig. /4536	Rabbit	1:500
5	p.Ask1	Ser967	Cell Sig./3764	Rabbit	1:700
6	p.Ask1	Ser83	Fisher/PIPA536618	Rabbit	1:500
7	p. Ask1	Thr838	Fisher/ PA5 64541	Rabbit	1:500
8	p.Ask1	Thr845	Cell Sig./3765	Rabbit	1:500
9	p. JNK1/2	Thr183/Tyrl85	Fisher/PIMA515175	Rabbit	1:700
10	p.p38	Thr180/Tyrl82	Cell Sig./4511	Rabbit	1:800
11	P-HSP27	Ser78	Cell Sig./2405 T	Rabbit	1:700
12	p.ERKl/2	Tyr 204	St Cruz/sc-7383	Rabbit	1:500
13	p.ERKl/2	Thre-202/Tyr204	Cell Sig./9101	Mouse	1:250
14	MELK	N/A	St. Cruz/sc- 517,201	Mouse	1:500
15	p.MELK	Thr167/Ser171	Abclonal/AP1175	Rabbit	1:300
16	P-SMAD2&3	S465/467&S423/425	Abclonal/AP0548	Rabbit	1:700
17	P-SMAD2	Ser465/467	Cell Sig./310e	Rabbit	1:700
18	P-SMAD2	Ser245/250/255	Cell Sig./3104	Rabbit	1:300
19	P-SMAD3	Ser423/425	Cell Sig./9520	Rabbit	1:500
20	α-SMA	N/A	Cell Sig./19.245	Rabbit	1:1000 μg/ml
21	GAPDII	N/A	Cell signaling	Rabbit	1:3000
22	HRP-Gt > Rb	N/A	Cell signaling	Rabbit	1:1000
23	HRP-Gt > Mo	N/A	Invitrogen	mouse	1:2000

2.3. Western Blot

Cells were washed twice in cold phosphate-buffered saline (PBS) and harvested in cell lysis buffer (20 mM Tris, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 2.5 mM sodium pyrophosphate, 1 mM β-glycerol phosphate, and 1 mM Na₃VO₄, pH 7.5, 1 mM PMSF, 1 mM benzamidine, 10 μg/ml leupeptin, and 10 μg/ml aprotinin) and fully disrupted by sonication. After centrifugation at 20,000 rfc for 15 min, supernatant protein concentration was determined using the micro-BCA protein assay kit (Pierce Biotechnology, Rockford, IL). Protein was denatured by boiling for 1 min in SDS sample buffer and 30–40 μg of protein was electrophoresed in 10% polyacrylamide mini-gels and electroblotted onto PVDF membranes (Bio-Rad, Hercules, CA). Runs incorporated a color-coded protein MW ruler (Pierce). The PVDF membrane was stained with Ponceau dye (Sigma) to ascertain that the cumulative electrophoresis and electro-transfer steps have preserved the identity of protein amounts between the samples. If the stain revealed a visible difference in transferred protein or local spatial defects, the blots were discarded. Membranes were blocked with 5% fat-free milk in 1 x PBS buffer-0.1% Tween-20 for 1 h, incubated overnight at 4°C with primary antibodies (Abs; all listed in Table 1 B), diluted in blocking medium, and for 45 min at room temperature with the proper HRP-conjugated secondary antibody. The procurement of the anti-p.MELK antibody requires special mention. An extensive literature search found only one article describing the application of such an antibody, generated against the combined thr167/ser171 region.³⁰ The article, though, does not address changes on p.MELK related to cellular status. While the supplier does not offer this antibody in its catalog, it preserved a stock that was facilitated to us. It can be purchased under the catalog number included in Table 1B. In some experiments, the initial reagents were stripped from the PVDF membrane and the membrane was re-probed with a second experimental Ab or an anti-GAPDH antibody. Bound HRP activity was visualized using different chemiluminescent reagents, as needed according to the strength of the signal. For targets generating a strong signal, we used Lumigen (Southfield, MI). For weak signals, we used Amersham ECL Select (Fisher). The chemiluminescent signal was captured by X-ray film exposure or using an ImageQuant 800 analyzer (Cytiva, Marlborough, MA), and quantized by densitometry. Bands were considered identical when the difference in measured intensity was under 10 %. Unless otherwise stated p.SMAD2, pSMAD3 refers to phosphorylation at the C-terminus.

2.4. Immunohistochemistry

Cells were seeded in SSFM on bovine collagen 1 (Advanced Biomatrix, Carlsbad, CA)-coated glass coverslips and treated with TGFβ1, alone or with the addition of inhibitors, or left untreated (control). After 72 h, the coverslips were washed with PBS, fixed with 10 % buffered formalin for 10 min, blocked with 1% bovine serum albumin, and incubated overnight with rabbit anti-human α-smooth muscle action (α-SMA). After washing, coverslips were incubated for 45 min with 0.1 μg/ml Alexa 568 conjugated goat-anti rabbit IgG and mounted in DAPI-complemented mounting media. Fluorescent images were captured in a Nikon Eclipse Ni microscope using a single exposure condition for all samples.

2.5. Data analysis

Results for Figure 8 are reported as means ± SEM using One-way ANOVA to assess statistical significance. Significance for before and after comparison was assessed using paired Students’ t-test. The different levels of significance are designated as follows: *p value < 0.05, **p value < 0.01, ***p value < 0.001.

Figure 8. α-SMA accumulation and cell morphology in naïve cells and cells treated for 72 h with TGFβ1 alone or in the presence of inhibitors for TAK1, ASK1, JNK1/2 or MELK. Top left micrographs.

All images were captured under identical conditions of illumination and image capture. Top right plot. Relative average fluorescent per cell, calculated as described in Methods. Bottom micrograph panel. Micrographs of the same experiments shown in the top frame were electronically manipulated to artificially increase the α-SMA signal (red fluorescence). Nuclei are stained by DAPI. A and J. Control cells, no TGFβ1. B and K. TGFβ1 only. C and L. 5z-7oxozeaenol; 10 μM. D and M. Selonsertib; 10 μM. E and N. SP600125; 10 μM. F and O. MELK8A; 10 μM. G and P. OTS167; 10 nM. H and Q. OTS167; 25 nM.

3. RESULTS

3.1. TGFβ1 rapidly activates TAK1, ASK1 and JNK1/2 but SMAD2/3 is activated after a substantial dormancy period.

Fig. 1A is a representative micrograph of the Ponceau staining of the protein electro-transferred to a nitrocellulose membrane, a preliminary measure of equal loading. The introduction of TGFβ1 to HCF cultures results in the rapid phosphorylation of the MAP3Ks, ³² TAK1, and ASK1 (Fig. 1B). Maximal signals were reached within 1 min and remained essentially unchanged for at least 15 min. TAK1 phosphorylation was identified only at Ser412.³³ Three antibodies (Table 1B) against the well-characterized TAK1 Thr184 and or Thr187 sites ³⁴ failed to elicit a signal. Within one min after TGFβ1 addition ASK1 became fully phosphorylated at three characterized sites, Ser967, Thr838, and Ser83 ^35,36. A validated antibody against the ASK1 Thr845 site, which has been associated with ASK1 activation by H₂O_2,³⁷ failed to generate a signal (not shown). The activation of TAK1 and ASK1 was matched by rapid, sustained phosphorylation of two downstream targets of these two MAP3Ks, the terminal kinases, JNK1/2/MAPK8/9, and p38/MAPK14. The third terminal kinase, ERK1/2/MAPK42/44, which displayed a substantial level of phosphorylation prior to TGFβ1 addition, was not activated.

Figure 1. Phosphorylation time course for TAK, ASK1, JNK1/2, p38, and SMAD2/3 after addition of TGFβ1.

A. Representative Ponceau staining of protein electro-transferred onto a nitrocellulose membrane. B. Representative Western blots. Four MAP kinases, TAK1, ASK1, JNK1/2 and p38 reached maximal activation levels within the first 1–2 min and remained at similar levels of activation for at least 15 min. In the case of ASK1 the phosphorylation occurred simultaneously in three locations. ERK1/2, probed with either the rabbit polyclonal (Rb) or the mouse monoclonal (Mo), showed no activation. The dual SMAD2 and SMAD3 antibody against the C-terminus (SMAD2/3;) produced a much stronger signal of the shorter SMAD3. In contrast to the rapid effects of TGFβ1 in all the MAP cascade kinases, the phosphorylation of SMAD2/3 does not start for the first five min. An antibody against the Ser245/250/255 present in the linker region of SMAD2 also yielded a visible signal at 15 min with no effect at the 2 min mark. The last line shows the result of stripping the p.SMAD2/3 blot and staining for GAPDH. C. Mean ± SD intensity of p.TAK1 S412 (n =4) and p.SMAD3 (n =7) over the first 15 min after the addition of TGFβ1. D. Differential effect of an inhibitor of TGFR kinase on TAK1 and SMAD2. Phosphorylation of TAK1 is not affected. E. Comparison of the intensities of p.SMAD2 and p.p38 15 min after the addition of TGFβ1 in cells from passage 7 or 10 when preincubated in SSFM for 24 or 48 h.

In stark contrast to the rapid effect on the MAPK branch of the response to TGFβ1, phosphorylation of SMAD2/3 at their C-term (S465/467 and S423/425), the hallmark of the canonical response, was not detectable for several min after the addition of TGFβ1. This silent period was followed by a rapid increase in p.SMAD2/3 which was observed at 15 min. In the initial experiments depicted in this figure, we used a combined anti-p.SMAD2/ anti-p.SMAD3 antibody cocktail (Table 1B). This cocktail generated a much stronger signal for p.SMAD3, the shorter isoform of p.SMAD2/3; this antibody-dependent result does not provide any information as to the actual relative concentration of both isoforms. Mean ± SD relative intensities, relative to the maximal intensity observed for both p.TAK1 (n = 3) and p.SMAD3 (n = 7) are depicted in Fig. 1C. To our knowledge, this is the first report of a dormancy period in the activation of SMAD2/3 and a marked temporal difference in the rapidity of the classical canonical and noncanonical responses to TGFβ1. Inhibition of the kinase activity by LYS 3649 abolished the phosphorylation of SMAD2 but had no effect on the level of p.TAK1 (Fig. 1D). Thus, the activations’ temporal difference may reflect the different mechanisms of activation of both pathways; activation of TAK1 occurs via TRAFs, independently of the kinase activity of the TGFR1-TGFR2 tetramer.³⁸

We also examined the phosphorylation of SMAD2 at its linker region, albeit with less detail. This area contains sites phosphorylated by serine kinases and has been associated with autonomous SMAD functions.³⁹ A commercially available antibody recognizing phosphorylation at the Ser245/250/255 sites (Table 1B) yielded a phosphorylation signal displaying a delayed response similar to the delay observed with the C-terminus sites (Fig. 1B). Finally, Figure 1E shows that the phosphorylation levels measured in this study were not substantially affected by either the cell passage or the length of the preincubation in SSFM.

3.2. Inhibition of TAK1, ASK1, or JNK1/2, but not of p38, prevents the phosphorylation of SMAD2/3

The time gap between the initiation of noncanonical and canonical responses to TGFβ1 opened the possibility that noncanonical signaling could affect the canonical SMAD2/3 activation. To test this possibility, we examine the effects of inhibitors of TAK1, ASK1, JNK1/2, and p38, on the accumulation p.SMAD2 and p.SMAD3. Inhibition of TAK1 by two unrelated inhibitors, 5z-7oxozeaenol and TAKinib, nearly abolished p.JNK1/2 (Fig. 2A). A comparison of the concentration of inhibitors causing similar reduction on the JNK1/2 signal indicated that 5z-7oxozeaenol was substantially more effective than the newer TAKinib reagent. Equivalent effects on JNK1/2 were obtained by the use of two ASK1 inhibitors, Selonsertib and ASK1 inhibitor 10. Thus, it would appear that the maintenance of a substantial level of p.JNK1/2 required the simultaneous input of both MAP3Ks.

Figure 2. Effect of MAPK inhibitors on p.JNK1/2, p.38, p.SMAD2 and p.SMAD3.

A. The phosphorylation of JNK1/2 is prevented by TAK1 inhibition by TAKinib or 5z-7oxozeaenol (5z-7ox) or inhibition of ASK1 by either, Selonsertib or ASK1 inhibitor 10 (ASKi10). All inhibitors were added at 10 μM, 10 min before the addition of TGFβ1. All samples were collected 15 min later or at the indicated times. B. The phosphorylation of SMAD2 or SMAD3 is blocked by the inhibition of the TAK1 or ASK1 with the inhibitors used in A to inhibit p.JNK1/2 accumulation. Additionally, the accumulation of p.SMAD2 or 3 is blocked by the inhibition of JNK1/2 by SP600125 (SP600), JNK inhibitor VIII (JNKiVIII), or JNK inhibitor XVI (JNKiXVI). C. Inhibition of TAK1 by TAKinib also reduces p.38 accumulation. D and E. In contrast to the inhibitory effect of inhibition of JNK1/2, inhibition of p38 or ERK1/2 activity does not affect the accumulation of p.SMAD2. Note that the effectiveness of all four JNK1/2 inhibitors is confirmed by their effect on HSP27, a p38 canonical target.

Each of the four inhibitors described above also blocked the accumulation of p.SMAD2 and p.SMAD3 (Fig. 2B). The phosphorylation of the SMADs was also inhibited by SP600125, a specific inhibitor of JNK1/2. Takinib also inhibited the accumulation of p38. These effects occur without affecting the p.TAK1 level (Fig. 2C). Finally, the application of four different p38 inhibitors or an inhibitor of ERK1/2 (Table 1A) revealed that the activity of these two terminal kinases has no effect on the phosphorylation of p.SMAD2 by TGFβ1 (Fig 2D and E). With respect to p38 this result was opposite to our previously proposed direct involvement of p38 in the early activation of SMAD2.⁴⁰ Hence, we sought independent confirmation of the effectiveness of all four inhibitors. Uniformly, the inhibitors caused a > 90 % decrease in the phosphorylation of p.HSP27, a well-established p38 target. It should be noted that our previous conclusion was based on the inhibitory effect of SB 203580, which has been shown to have non-specific effects.^41,42 Overall, the results presented in Figs. 1 and 2 indicate that, in response to TGFβ1, both TAK1 and ASK1 are rapidly phosphorylated and jointly induce the activation of JNK1/2, which in turn leads to phosphorylation of SMAD2/3.

3.3. MELK inhibition blocks and rapidly reverts SMAD2 phosphorylation

To identify mediators of the effect of the MAKP kinase inhibition on SMAD phosphorylation and/or contributors to the observed latency of the canonical response, we performed an extensive screen of inhibitors of signal transduction pathways. These studies revealed that the phosphorylation of SMAD2 was prevented by OTSSP167 (OTS167) or by MELK8A, two inhibitors of the maternal embryonic leucine zipper kinase (Fig. 3A). Very similar degrees of inhibition of about 75 % were obtained with 0.1μM OTS167 or 10 μM MEL8A, consistent with the reported difference in dissociation constants (Table 1) To confirm that the effect of these inhibitors was indeed due to their effect on MELK activity, we additionally examined the effect of two MELK mRNA interference agents, a cocktail of six classical siRNAs or a whole-length RNA antisense digest (esiRNA). The application of either antisense agent resulted in 67 and 92 % reduction of total MELK protein by the siRNA and esiRNA treatments, respectively, when compared to the signal intensities in the scrambled siRNA treated or siRNA omitted samples (Fig 3B). Equivalent inhibition was obtained with OTS167 for p.SMAD3 (not shown).

Figure 3. Effect of MELK inhibition or downregulation on the phosphorylation of p.SMAD2.

A. Effect of OTS167 or MELK8A on p.SMAD2. The inhibitors were added 10 min before the addition of TGFβ1, cells were harvested 15 min afterward. B. Effect of MELK protein reduction through siRNA electroporation. Top. Total MELK levels 48 h after electroporation. The addition of siRNA or esiRNA led to reductions of the MELK level observed when using scrambled siRNA (SsiRNA) equal to 67.6 % and 90.4 %, respectively. Bottom. Effect of scrambled siRNA, the siRNA cocktail, the siRNA, or the omission of RNA during the electroporation step (Untreat.) on p.SMAD2. There was no measurable difference between SsiRNA and the untreated samples. The siRNA cocktail and the esiRNA reduced the p.SMAD2 signal by 69.8 % and 77.8 %, respectively.

3.4. MELK becomes phosphorylated after the addition of TGFβ1 in a TAK1-, JNK1/2-, and ASK1dependent manner

Since MELK inhibition prevented the phosphorylation of SMAD2 and SMAD3, we sought to determine whether the effect of MELK depends on its phosphorylation. Fig. 4A shows that TGFβ1 caused a maximal phosphorylation response of MELK within the first two min after the addition of TGFβ1 that remained essentially unchanged for at least 15 min. This phosphorylation was majorly (> 90 %) prevented by the inhibition of either TAK1, JNK1/2 or ASK1 (Figure 4 B). It is pertinent to mention that the anti-p.MELK antibody generated a substantial nonspecific stain throughout the immunoblot. Nevertheless, multiple observations provided robust assurance of the bona fide status of the signal. Firstly, it was the only band absent in the column for the untreated control. Secondly, it was the only band that showed an increase upon the addition of TGFβ1. Thirdly, the p.MELK signal overlapped with the total MELK signal at approx. 71 kD.

Figure 4. Effect of TGFβ1 and inhibitors of TAK1 and JNK on the phosphorylation of MELK.

A. MELK phosphorylation as a function of time of activation by TGFβ1. B. Effect of inhibitors of TAK1 (Takinib and 5z-7oxozeanol or 5z-7 ox), JNK1/2 (SP600125, JNKiVIII, or JNKiXVi) or ASK1 (Selonsertib or ASKi10) on cells treated for 15 min with TGFβ1.

3.5. The sustained phosphorylation of ASK1 depends on TAK1, JNK1/2 and MELK activity.

A previous study has shown that MELK activates ASK1 by phosphorylating Thr838.⁴¹ We hypothesized that given that MELK activation is indirectly dependent on ASK1 activity, a reciprocal effect may establish a positive feedback loop. To test this hypothesis, we examined the effect of kinase inhibitors on TAK1, ASK1, JNK1/2, and MELK. OTS167 had no effect on the TGFβ1-induced phosphorylation of TAK1 (Fig. 5 A, n = 3) but brought the p.ASK1(S967) and p.MELK signals to their control levels (Fig. 5B). In fact, inhibiting TAK1, ASK1 or JNK1/2 led in all cases to sizeable decreases or complete blockade of the phosphorylation of these two entities (Fig. 5 B). The effect of OTS extended to the other two ASK1 phosphorylation sites, S83, and Trh838 (Fig. 5 C).

Figure 5. Effect of TAK1, ASK1, JNK1/2 and MELK inhibitors on the phosphorylation of TAK1, ASK1, MELK and SMAD2.

A-C. Inhibitors were added 10 min before the addition of TGFβ1. A. Lack of effect of OTS167 on the phosphorylation of Ser412 in TAK1. B. Effect of inhibitors for TAK1 (5z-7oxozeanol; 5z-7ox), MELK (OTS167),ASK1 (Selonsertib; Selons.) and JNK1/2 (SP600125) on the phosphorylation of Ser 967 residue of ASK1 and on MELK. C. Effect of OTS167 in all three ASK1 phosphorylation sites. D. The concentration of p.SMAD2 is stable in the 20–40 min following the addition of TGFβ1. The addition of 0.2 μM OTS167 results in the full reversal of p.SMAD2 accumulation between 5 and 10 min afterwards. E. Reversal of the accumulation of phosphorylated SMAD2, ASK1(967) and JNK1/2 upon addition of OTS167. E. Time course of decrease in the levels of p.SMAD2, p.ASK1(S967), and p. JNK1/2 after the addition of 1 μM OTS167.

3.6. At the steady state of [p.SMAD2], SMAD2, ASK1 and JNK1/2 are subjected to very rapid dephosphorylation

The concentration of p.SMAD2 did not change in the 20 to 40 min post-TGFβ1 activation interval (Fig. 5D). The same was true for p.ASK1(S967) and JNK1/2. This fact implies that during this period phosphorylation and dephosphorylation activities on these three proteins were balanced. We exploited this condition and the very low levels of OTS167 needed to attain full inhibition of p.SMAD2 accumulation to investigate dephosphorylations activities. The addition of 0.2 μM OTS167 at t= 30 min resulted in the elimination of the p.SMAD2 signal between 5 and 10 min (Fig. 5D). To get a higher temporary resolution of the effect of a sudden arrest of MELK activity, OTS167 was added after the 30 min of exposure at the oversaturating 1 μM concentration. The objective was to attain as fast a build-up of the OTS167 cytosolic concentration as possible, in order to examine the temporal effect of this addition on p. SMAD2, p.ASK1 (S967), and p.JNK1/2. All three kinases rapidly lost their phosphorylation (Fig.5 E). SMAD2 phosphorylation was reduced by more than 74 % within 1 min and was completely absent by the second minute (n = 2), p. ASK1 and p.JNK1/2 waned at a slower rate; after 3 min p.ASK1(S967) and p.JNK2 were indistinguishable from the untreated controls (n= 3) and where essentially nil after 5 min. These results suggest that the maintenance of the TGFβ1-activated state is subjected to a very rapid phosphorylation-dephosphorylation interplay for most participating kinases. The disappearance of p.SMAD2 between the 1–2 min interval represents a lower dephosphorylation rate limit, as it may reflect the time needed to achieve inhibitory OTS167 concentration in the cytosol.

3.7. Activation of MELK in the absence of TGFR activity does not result in p.SMAD2 accumulation at the C-terminus of linker regions.

The results above indicate that MELK activation depends on the prior activation of the TAK1-JNK1/2 cascade. JNK1/2 can be activated by multiple inflammation-related cytokine inputs. Hence, we examined the effect of two proinflammatory agents, TNF-α or LPS on the phosphorylation status of JNK1/2, MELK, and SMAD2. TNF-α caused a 4. 7 ± 0.5 (n =4) -fold increase in p.JNK1 and a 5.3 ± 0.98 increase in pJNK2 for JNK2 when compared to the levels induced by TGFβ1 (Fig. 6 A). LPS caused a very similar p.JNK1/2 increase. Additionally, TNFα activated ASK1 but only to the same level elicited by TGFβ1 (n = 2). Thus, the large increase in p.JNK1/2 seems to involve an additional activation path unrelated to the TGFβ1-TAK1-ASK1 path indicated by the results in Fig. 2. Consistent with the activation of JNK1/2, the two proinflammatory agents caused also the phosphorylation of MELK.

Figure 6. Effect of TGFβ1, TNF-α, or LPS on JNK1/2, MELK, or SMAD2/3 phosphorylation.

A. Effect of agonists added by separate, 15 min incubation. TNF-α and LPS cause a much larger increase in JNK1/2 than that induced by TGFβ1 but p.ASK1(S967) and p.MELK levels were very similar to those achieved with TGFβ1. TGFβ1 induced phosphorylation at both C-terminus and linker regions. Neither of these regions became phosphorylated in response to TNF-α B. Effect of inhibition of TGFR on the phosphorylation at the linker region. C. Time course for p.SMAD2 after combined addition of TGFβ1 and TNF-α. Adding TNF-α does not affect the delayed p.SMAD2 activation shown in Figure 1.

Using protein overexpression in transformed cells, mutation analysis, and tagged protein immunoprecipitation Jung et al ⁴³ demonstrated a physical interaction between MELK and SMAD2/3 in the absence of TGFβ1 and the ability of activated MELK to phosphorylate at Ser245 in the SMAD2 linker region. Hence, we examined whether a TGFR-independent activation of MELK is involved in the observed linker region phosphorylation shown in Fig. 1. We could not document such an effect in our cellular system. The activation of MELK by TNF-α did not lead to any significant linker region phosphorylation as detectable by the antibody we used. Additionally, the inhibition of TGFR by Lys3694, which preserves the activation of the MAPK cascade by TGFβ1 (Fig. 1B) fully abolished the linker zone phosphorylation (Fig. 6B). Finally, consistent with the fact that the large increase in p.JNK1/2 levels attained with TNF-α did not result in an increase in p.MELK above the level attained TGFβ1, complementation with TNF-α did not with shorten the length of the silent period preceding the start of p.SMAD2 accumulation (Fig 6C), or its magnitude at 15 min; the mean ± SEM signal ratio for p.SMAD2 between TGFβ1 alone and TGFβ1 +TNF-α was 0.93 ± 0.12 (n =4). The only difference we were able to document between TGFβ1 alone or when used in combination with TNFα was a moderate decrease in the effectiveness of inhibitors. TAKinib and Selonsertib reduced the TGFβ1-derived p.SMAD2 signal by 83 ± 8 % (n = 4) and 93 ± 4 (n = 3), respectively, while the corresponding figures for the TNF-α derived signal were 49 ± 8 %, and 27 ± 11 % (n = 3). When we used the more powerful TAK1 inhibitor 5z-7oxozeaenol at 3 μM, though, the inhibition in both conditions exceeded 90 %. Finally, it is noteworthy that the very high pJNK1/2 levels elicited by TNF-αI or LPS did not result in comparable increases in p.MELK levels, which remained unchanged from the level obtained with TGFβ1 (n = 2).

3.8. Inhibition of TAK1, ASK1, JNK1/2, or MELK inhibits αSMA synthesis

The contractile activity is essential to promote wound closure. The principal biochemical event underpinning the fibroblast conversion into myofibroblasts is the de novo expression of α-SMA and its organization into intracellular contractile fibrils. Figure 7 depicts the effect of inhibition of TAK1, JNK1/2, or MELK on the total α-SMA accumulation after 72 h incubation with TGFβ1. All the chemical inhibitors used caused a reduction of 70–80 % inhibition of α-SMA.

Figure 7. Effect of inhibitors of TAK1, MELK, and JNK and MELK siRNAs on the expression of α-SMA after 72 h of TGFβ1 induction.

Cells were cultivated for 48 h in SSFM prior to the addition of TGFβ1 and the inhibitors. For the siRNA experiments, cells were first treated by electroporation and then reseeded and cultivated for 48h in SSFM prior to the addition of reagents.

3.9. Inhibition of MAPKs prevents the development of a myofibroblast morphology

While the de novo expression of α-SMA is the most prominent event in the onset of the myofibroblast phenotype, many other changes occur, including an overt change in cellular morphology. Hence, we performed immunocytochemical analysis of the treated cells to assess the level of α-SMA fibrillar formation and cell shape. After the 72h incubation in TGFβ1, the images of cells stained for α-SMA were captured under identical exposure and image capture condition. Fig. 8A–H are representative areas of each experimental condition. The graph (Fig. 8 I) shows the average ± SEM signal intensity (n = 3). Except for the case of Selonsertib, all the inhibitor-treated samples had total fluorescence levels which were not statistically significantly different from the untreated sample. The micrographs in the bottom panel (Fig. 8J–Q) have been electronically enhanced in Photoshop (Microsoft, Seattle, WA), as needed to reveal the cell shape and the localization and structure of any α-SMA stain within the cells. Cells maintained in SSFM alone (control) were small and narrow and displayed a predominantly perinuclear α-SMA stain (Fig. 8 A–J). The TGFβ1- treated cells displayed moderate to high increases in cross-sectional area and in the density of α-SMA fibers, which now extend throughout the cell (Fig. 8A–K). Cells that were treated with either 10 nM OTS167 (Fig 8O) or with 10 μM MELK8A (Fig. 8P) showed an α-SMA stain and cell shape closely resembling the untreated control cells. The images for all three other inhibitors, 5z-7oxozeaenol, TAKinib, and Selonsertib clearly either lacked or showed lower expression of α-SMA than the cells treated solely with TGFβ1, but showed morphological differences with the control and OTS167 cells. Thus, even though these inhibitors had largely limited the de novo expression of α-SMA and myofibroblasts’ morphological development, they may have had cellular effects unrelated to the impact of TGFβ1.

3.10. Dermal and Tenon’s capsule fibroblasts display similar delayed activation of p.SMAD2 and sensitivity to OTS167

The cultured corneal fibroblasts studied in this report derive from the corneal keratocytes. These are highly specialized neural crest-derived cells that reside in a unique, avascular environment of high oncotic pressure and UV exposure. They express a unique keratan sulfate, and abnormally high levels of aldehyde dehydrogenases ^44–46 Thus, we performed a minimalistic assessment of the significance of the results obtained in other tissue fibroblasts. Figure 9 shows that both in fibroblasts obtained from human skin and in fibroblasts derived from the Tenon’s capsule, the highly elastic collagenous layer enveloping the corneal sclera, the accumulation of p.SMAD2 in response to TGFβ1 shows two of the features observed in the corneal fibroblasts, delayed activation of SMAD2 and high sensitivity to OTS167.

Figure 9. Delayed p.SMAD2 accumulation and effect of OTS167 in dermal and Tenon’s capsule derived fibroblasts.

A. Dermal fibroblasts. B. Tenon’s capsule fibroblasts.

4. DISCUSSION.

In this report, we demonstrate that in cultured fibroblasts derived from the human cornea, phosphorylation of SMAD2/3 in response to TGFβ1, commonly referred to as the immediate canonical response is, a) characterized by a previously unrecognized silent period of about 5 min; b) fully dependent on a much faster noncanonical response, namely the activation of two MAP3K1, TAK1 and ASK1 and their downstream MAPK, JNK1/2; and c) ultimately mediated by MELK which is activated by JNK1/2. These results bring into question the concept of independent canonical and noncanonical TGFβ1 effects.

MELK plays a dominant role in the overall process of SMAD2/3 activation through effects that may simultaneously and independently occur at two distinct nodes of the overall process. Firstly, its inhibition results in the dephosphorylation of ASK1 and JNK1/2 indicating that the process sustaining the activated state of all these three kinases involves a recurrent self-reinforcing cycle of interactions, where activated ASK1 is needed to attain the phosphorylation of MELK, and vice versa, activation of ASK1 requires the involvement of activated MELK. The limited temporal resolution of our studies does not allow us to ascertain whether an initial minimal activation of ASK1 by TGFβ1 leads to weak activation of JNK1/2 and MELK or whether MELK is first weakly activated through TAK1-JNK1/2 and subsequently reinforces its activation by causing the activation of ASK1. Nevertheless, the absolute dependence of ASK phosphorylation on TAK1 activity is consistent with the absence of a direct effect of TGFβ1 on its activation and favors the indirect activation through an initial MELK activation. This dependence on a self-reinforcing loop may be required because of the moderate degree of activation of the MAPK cascade by TGFβ1 so that the combined input of both MAP3Ks is needed to overcome the dephosphorylation activity on JNK1/2. Secondly, the extremely rapid effect of MELK inhibition by OTS167 when p.SMAD2 has achieved a steady state level of phosphorylation (Figs. 3 and 5) demonstrates the absolute dependence of the phosphorylated state of SMADs on active MELK. The exact nature of the mechanism of control of SMAD phosphorylation remains to be resolved. Our studies show that phosphorylation does not occur only on the C- terminus but also on the linker region. The phosphorylation at both sites displays the delayed response and independently These concepts are graphically depicted in Fig. 10. Our studies do not provide information on the source of the observed delayed nature of the SMAD2/3 phosphorylation. It may relate to the organization of the SMAD anchor for receptor activation (SARA) complex to the hetero-tetrameric TGFBR1-TGFβR2 receptor.⁴⁷

Figure 10. Depiction of the dependence of SMAD2/3 activation by TGFβ1 on the pre-activation of the MAPKs and MELK.

The model is based on the data indicating that the maintenance of the noncanonical response depends on a self-reinforcing ASK1-JNK1/2MELK loop. The inhibition of any of the four kinases led to rapid dephosphorylation of all the four kinases suggesting that each one of these kinases is simultaneously subjected to continuous dephosphorylation activities (not included in the graph). It is assumed that the activation of ASK1 via TAK1 or through activation by any other TGFβ1- derived signal is unlikely and thus, it proposes that ASK1 activation is secondary to an initial preactivation of MELK. The sketch incorporates several features not examined in this report, including the role of, TRAF, ubiquitination and TAB proteins in the activation of TAK1 and the role of the SARA complex in the attachment of the SMADs to the TGF receptor.

Several aspects of the phosphorylation of TAK1 and ASK1 and the dependence of JNK1/2 activation on these two kinases deserve attention. Firstly, activation of TAK1 phosphorylation is known to occur through an autophosphorylation process, involving one or more polypeptide sites, in response to cytokine-induced ubiquitination of a triad of TGF-beta-activated kinase 1 and MAP3K7-binding protein proteins.^48–50 In any given condition and cell type, different polypeptide locations may be phosphorylated depending on the nature of the stimuli, the expression of different TABs and other participating proteins, and existing post-translational modifications.⁵¹ In our studies we identified the phosphorylation of Ser412 as a site immediately modified by TGFβ1 and, though we have not exhausted the examination of putative phosphorylation sites in TAK1, we failed to observe any change at the Thr-178 and Thr-184 residues. The phosphorylation of these latter sites has been associated with TAK activation of NFκB and AP-1.⁵² Thus, the absence of such phosphorylation in our cells may relate to the absence of activation of these SMAD2/3-unrelated pathways. With regard to ASK1, the phosphorylation of Ser967 has been shown to be associated in multiple systems with an inhibitory binding of 14–3-3 proteins. ^53,54 However, the activity state and the implication of the S967 site phosphorylation when at least two other sites are simultaneously phosphorylated is difficult to evaluate. ASK phosphorylation on Thr845, a modification that has been associated with astrocyte apoptosis ⁵⁵ could not be identified in the corneal fibroblasts. Secondly, in spite of the large difference in p.JNK1/2, the phosphorylation level of MELK was the same whether the stimuli were TGFβ1, TNF-α, or LPS. It may be that the p.MELK level obtained with TGFβ1 is already at a peak, or that another factor, playing a rate-limiting role, is involved or mediates the JNK1/2-MELK interaction. Thirdly, this rapid loss of p.SMAD2 upon inhibition of MELK activity reveals that simultaneously with the phosphorylation activity, SMAD2 is subjected to vigorous dephosphorylation. PPM1A has already been identified as a potential candidate for this activity.⁵⁶ The presence of a very active dephosphorylation activity could be behind the seeming discrepancy between our results on Figure 6 and those obtained by Jung et al, namely TGFβ-independent phosphorylation at the linker region by MELK. It should be noted, though that in their study, they found that TGFβ enhanced the S245 phosphorylation. Thus, it is conceivable that in our system and conditions strong dephosphorylation activities cancel any phosphorylation effected by MELK alone and build-up of p.SMAD2 phosphorylation at Ser245 occurs only after the rate of phosphorylation is increased by TGFβ1. Fourthly, in spite of the absolute dependence of SMAD2/3 phosphorylation on MELK activity, p.SMAD2 was not generated when the phosphorylation of MELK was elicited through TNF-α or LPS, indicating that MELK activity is not sufficient condition SMAD2/3 activation. Fifthly, the effects of all the inhibitors on SMAD2/3 activation show a strong correlation with the inhibition of α-SMA synthesis at 72h and the change in cell morphology, hallmarks of the emergence of the myofibroblast phenotype, demonstrating a very strong correlation between the kinase activity in the early and late response to TGFβ1. Sixthly, the limited study of cultured fibroblasts derived from two other connective tissues suggests that these signal transduction patterns may be common to many mesenchyme-derived cell types.

5. CONCLUSIONS

The current finding leads to an alternative view of the current understanding of the signal transduction events induced by TGFβ1. Firstly, in fibroblasts, the phosphorylation of SMAD2/3 in response to TGFβ1 is characterized by a delayed response involving a 5-minute silent period. Secondly, rather than independent early canonical and noncanonical responses, the response to TGFβ1 consists of a single coordinated event where the activation of the MAPK cascade through TAK1 leads to the activation of MELK and a recurrent ASK1-JNK1/2-MELK positive feedback loop. Thirdly, activated MELK is absolutely essential for the phosphorylation of SMAD2/3 and the subsequent development of the myofibroblasts phenotype.

Highlights Wolosin 2023.

The current finding leads to an alternative view of the current understanding of the signal transduction events induced by TGFβ1. Firstly, in fibroblasts, the phosphorylation of SMAD2/3 in response to TGFβ1 is characterized by a delayed response involving a 5-minute silent period. Secondly, rather than independent early canonical and noncanonical responses, the response to TGFβ1 consists of a single coordinated event where the activation of the MAPK cascade through TAK1 leads to the activation of MELK and a recurrent ASK1-JNK1/2-MELK positive feedback loop. Thirdly, activated MELK is absolutely essential for the phosphorylation of SMAD2/3 and the subsequent development of the myofibroblasts phenotype.