Tumor necrosis factor-alpha induced expression of matrix metalloproteinase-9 through p21-activated Kinase-1

First, we surveyed the expression pattern of gelatinases by biopsies from patients with non-healed chronic (open wounds for more than 30 days) or healed wounds, as well as normal skin tissue. As shown in Fig. 1A, massive amount of proMMP-9 (92-kDa) and mature form (82-kDa) were evident in non-healed wounds, while minimal level of the proteinase was found in healed and normal skin tissues. Conversely, low level of proMMP-2 was evenly expressed by all three skin samples, while mature MMP-2 was found in both healed and non-healed skin. Excessive expression of MMP-9 in non-healed skin is associated with persistent inflammation in chronic wounds. To confirm the notion, isolated human dermal fibroblasts at the early passages (p < 3) were cultured on plastic, or embedded in 3-D type-I collagen. Cultured on plastic, dermal fibroblasts expressed a minimal level of proMMP-9 in response to TNF-α; under the combination of TNF-α and TGF-β, the fibroblasts produced massive proMMP-9 (Fig. 1B). Only in 3D type-I collagen, the fibroblasts also generated mature MMP-9 in response to the combination of TNF-α and TGF-β. We also measured the steady state of MMP-9 mRNA level. Dermal fibroblasts cultured on plastic or in 3D collagen gel were exposed to TNF-α, TGF-β, and their combination for 16 hrs, and the mRNA level of MMP-9 was determined by real-time RT-PCR. As shown, the mRNA gives almost identical ranking of expression as to the protein levels, demonstrating that the regulation of MMP-9 by TNF-α and collagen is mostly at the mRNA level (Fig. 1C). Thus, multiple extracellular cues including combination of two drastic different cytokines, TNF-α and TGF-β, and ECM render the dermal fibroblasts to generate maximal level of MMP-9, implying a situation in inflaming tissues.

Primary human keratinocytes produced minimal level of proMMP-9 in response to TNF-α, and the expression was also additionally enhanced in concert with TGF-β (Fig. 1D). Of note are two clear distinguished features between the dermal fibroblasts and keratinocytes in terms of MMP-9 expression. First, ECM has a profound role to induce MMP-9 expression by dermal fibroblasts but not keratinocytes (data not shown). Second, MMP-2, a mesenchymal MMP is largely absent in keratinocytes. Taken together, both the keratinocytes and dermal fibroblasts of human skin may contribute to the massive expression of MMP-9 under the stimulation of inflammatory cytokines in inflammation.

To elucidate the intracellular signaling to govern MMP expression we transduced human dermal fibroblasts with lentivirus to express GFP, PAK1 wild type, PAK1 triple mutant (kinase inactive and absence of p21GTPase binding), and PAK1 K299R mutant (kinase inactive) respectively. The viral transduced dermal fibroblastic cell lines were then embedded in type-I collagen, followed by TNF-α stimulation at the indicated dose for 3 days. Gelatinases in conditioned medium were resolved by zymography. Strikingly, the expression of proMMP-9 induced by the synergistic action of TNF-α and collagen was totally suppressed by PAK1 variants, but not GFP which works as a negative control (Fig. 2A). The specific effects of PAK1 on MMP-9 expression were evident by the absence of regulation on the expression of proMMP-2 as well as the maturation of the zymogen (transition from 72-kDa to 62-kDa). The wild-type PAK1 at the situation of over expression partially ablated MMP-9 expression, which may due to blockage of signaling traffics under over-expression (data not shown). Thereafter, we measured the steady state of MMP-9 mRNA by the human fibroblasts expressing either GFP or PAK1 mutant (K299R) cultured on plastic or in 3D type-I collagen. After 16 hr treatment by either TNF-α or TGF-β, as well as their combination, the MMP-9 mRNA was determined by real-time RT-PCR. As shown in Fig. 2B, induction of MMP-9 mRNA at all these conditions was totally impaired by the kinase inactive PAK1. Finally, we measured the effect of PAK1 on the 5'-promoter activity of the MMP-9 gene. The 5' promoter activities of MMP-9 were measured by dual luciferase assay through transfection of dermal fibroblasts with a reporter plasmid expressing firefly luciferase driven by a 670-bp 5'-promoter of MMP-9, while the control was monitored by renilla luciferase driven by CMV promoter. As shown in Fig. 2C the cytokine-induced activation of MMP-9 promoter was impaired by the PAK1 mutant.

We then addressed the nature of TNF-α regulation of PAK1. First, we analyzed the cell lines constitutively expressing PAK1 driven the viral promoter. After 3 days of TNF-α exposure the protein level of the ectopically expressed PAK1 was unexpectedly elevated, indicating a possible mechanism by which TNF-α signaling results in stabilization of PAK1 protein (Fig. 2D). Importantly, the TNF-α induced accumulation of PAK1 protein is independent of the kinase activities and p21 GTPase binding, as measured by the similar elevation of the variants of K299R and the triple mutant in response to TNF-α. Measurement at fine time points demonstrated again the prompt mode of elevation of the ectopically expressed PAK1 after TNF-α treatment (Fig. 2E). Because the ectopically expressed PAK1 is driven by constitutively active viral promoter, the TNF-α induced elevation of the kinase is, therefore, likely mediated by stabilization of the protein. To strengthen the notion we treated the dermal fibroblasts with cycloheximide to block protein synthesis and to monitor the degradation of the kinase. Under 20 μg/ml of cycloheximide, which sufficiently suppresses many protein syntheses, the PAK1 protein was still elevated under TNF-α treatment, demonstrating that the cytokine-dependent elevation of PAK1 is mediated by stabilization (Fig. 2E).

Immortalized human keratinocytes (IKC) were also transduced by the same set of lentiviruses, and the cells were treated by cytokines for 3 days. As shown in Fig. 3A, proMMP-9 was abundantly induced by IKC transduced by the wild type PAK1 virus in response to TNF-α but not TGF-β stimulation. Conversely, the TNF-α induced expression of proMMP-9 was partially abolished by the cells expressing the kinase-negative PAK1 variants. Similar to the fibroblasts, the protein levels of the wild type PAK1 in the keratinocyte was clearly enhanced by TNF-α treatment (Fig. 3B). Taken together, we demonstrated a novel regulation of PAK1 by TNF-α, which is mediated by an unknown mechanism to stabilize the PAK1 protein.

The intriguing regulation of MMP-9 by PAK1 as shown in Fig. 2A prompted us to examine regulation of other MMPs. Dermal fibroblasts transduced by lentivirus expressing GFP or inactive PAK1 (K299R) were embedded in 3D type-I collagen. After 3 days of stimulation with cytokines, the secreted MMP-3 in conditioned medium was measured by Western blot analysis. As shown in Fig. 4A, TNF-α, but not TGF-β, induced MMP-3, while combination of the two cytokines promoted synergistic induction of MMP-3, in a manner very similar to MMP-9 expression. However, in a drastic contrast to MMP-9, the cytokine-induced MMP-3 expression is not affected by over expression of PAK1 variants. The mRNA levels of MMP-2, -3, -9, -14, and TIMP-1 were measured by real-time RT-PCR. As shown in Fig. 4B, after 16 hr stimulation the mRNA of MMP-3 was induced 6 folds by TNF-α and additionally enhanced by the combination with TGF-β, which is closely correlated to MMP-3 protein expression. The cytokine-mediated induction of these MMPs was not altered by forcedly expressed PAK1. In a similar fashion, the low level expression of the mRNA of MMP-2, -14, and TIMP-1 was also not altered by PAK1. Again, PAK1 variants thoroughly suppressed the cytokine-exerted MMP-9 mRNA elevation. These results clearly demonstrate a specific mode of PAK1 in regulation of MMP-9 expression.

JNK signaling has been well defined as a key regulator to initiate the transcription of many MMPs including MMP-9 [20, 21]. To probe the possible role of JNK in cytokine-induced MMP-9 expression, we treated human dermal fibroblasts in 3D type-I collagen or on plastic by inhibitors for JNK (SP600125), p38 MAP kinase (SB239063), and PI-3 kinase (wortmanin) at the effective concentration suggested by the manufacture. After 3 days of culture with TNF-α and/or TGF-β, the conditioned medium was examined for gelatinase activities. As shown in Fig. 5A, the cytokine-induced MMP-9 was suppressed thoroughly by the JNK inhibitor, although the inhibition efficacy was slightly lower by the cells cultured in 3D gel, which may be due to the adsorption or trap of the compound in the scaffold of ECM. Expression of proMMP-2 (72-kDa) and maturation to the active form (62-kDa) were clearly not altered by the JNK inhibitor, indicating specific inhibition and absence of general cellular toxicity. In contrast, the inhibitors for p38MAP kinase and PI3 kinase were found without effects on MMP-9 expression. Activation of JNK by fibroblasts ectopically expressing PAK1 variants was measured by antibody against p46 and p54 SAPK/JNK dually phosphorylated at Thr183 and Tyr185 (Fig. 5B). After 20 min treatment with cytokines JNK was phosphorylated in response to TNF-α by the controlled cells (GFP), and largely abrogated by the cells expressing inactive PAK1 (K299R) (Fig. 5C).

In addition to JNK and NF-κB pathways, TNF-α also activates p38MAP kinase, which prompted us to examine its role in regulation of MMP-9 expression. Human dermal fibroblasts transduced by lentivius expressing p38MAP kinase or the dominant negative variant; and the resultant cells were cultured either on plastic or in 3D gel followed by treatment with cytokines. As shown in Fig. 6A, cytokine-mediated proMMP-9 expression was not altered by p38MAP kinase, which is in line with the results of the inhibition experiment (Fig. 5A). In contrast, MMP-9, but not MMP-2 expression, was totally suppressed by inactive PAK1. Moreover, protein stability of the ectopically expressed p38MAP kinase was not regulated by TNF-α (Fig. 6B). Thus, JNK1 is likely to be a downstream effector of PAK1 as demonstrated previously by others [22], and we showed here such moiety in the TNF-α signaling to control MMP-9 expression.

Many studies have demonstrated the role of NF-κB in regulation of MMPs including MMP-9 by directly recruitment of NF-κB to the cis-elements in the 5'-promoter of MMP-9 [20, 21]. Of interest is to know if NF-κB signaling is also under control by PAK1. Activation of NF-κB was measured by degradation of its inhibitor, IkappaB-α by the fibroblasts under the context of GFP, PAK1 (K299R), p38MAP kinase and its dominant negative variant. After 10 min of stimulation with TNF-α, IkappaB-α was promptly up-shifted, presumably through hyper-phosphorylation (Fig. 7A). At 20 min after the challenge the IkappaB-α was totally diminished by all these cells. Thus, in contrast to JNK, NF-κB is independent of PAK1 signaling. To confirm the notion of requirement of NF-κB we treated the fibroblasts with inhibitor for IKK, and measured the mRNA and protein of MMP-9. As shown in Fig. 7B and 7C, the cytokine-induced expression of MMP-9 at both mRNA and protein levels was partially suppressed by the IKK inhibitor. Conversely, expression of MMP-2 was not affected by the inhibitor. Thus, NF-κB signaling is required, but not under control by PAK1, to induce MMP-9 expression.

Given the evidence of stabilization of the ectopically expressed PAK1 by TNF-a signaling, and regulation of MMP-9 expression by the kinase, an immediate question is, therefore, to what extend the endogenous PAK1 is actually activated and stabilized in response to TNF-α. As shown in Fig. 8A, the doublet bands of endogenous PAK1 were immediately elevated after 30 min treatment of keratinocytes by TNF-α, and the PAK1 protein was substantially accumulated at 120 min. A band at 25 kDa, also recognized by the antibodies for PAK1, a presumable degradation product at basal state, was gradually diminished, whereas the full length PAK1 accumulated after TNF-α treatment, indicating a possible mechanism that TNF-α suppresses the turnover of PAK1. As an internal control, two bands at about 45-kDa were constitutively expressed in a mode independent of TNF-α. As a general way to activate a protein kinase, a site within the pseudo-substrate loop in the catalytic domain is phosphorylated in response to stimulation [23]. Phosphorylation of PAK1 at threonine-423 has been demonstrated as an indicator for activation state of the kinase [24]. In response to TNF-α, phosphorylation of threonine-423, as measured by antibodies for the phosphopeptide, promptly appeared at 10 min, and peaked at 20 min (Fig. 8B). PAK1 can also be phosphorylated at threonine-212 presumably by Cdc2/Cdk5 [25]. As shown in Fig. 8C, threonine-212 phosphorylation occurred maximally at 10 min after TNF-α stimulation, prior to the threonine-423 phosphorylation.

We then fractioned cellular compartments, and found that most of PAK1 protein and the threonine-212 phosphorylated version being present in cytosolic fractions rather than in the membrane pools. Similarly, threonine-423 of PAK1 was phosphorylated in dermal fibroblasts in response to TNF-α stimulation but in the prompt manner as in the keratinocytes (Fig. 8E). Importantly, fibroblasts over expressing PAK1 variants failed to show threonine-423 phosphorylation, which sufficiently explains its role in the impairment of the signaling in MMP-9 expression (Fig. 8E and 2). Similar to keratinocytes, dermal fibroblasts expressing GFP showed prompt phosphorylation of threonine-212 in response to TNF-α (Fig. 8F). We were interested to know whether PAK1 can be activated by IL-1, which shares many signaling pathways with TNF-α. Our previous work showed that the hepatic stellate cells can vigorously produce both proMMP-9 and the mature proteinase in response to IL-1 stimulation only by the cells cultured in type-I collagen [7, 26]. As expected, in the rat hepatic stellate cells, PAK1 underwent phosphorylation at threorine-212 (10 min), followed by phosphorylation at threorine-423 (30 min) (Fig. 9). Taken together, through analysis of these three cell types we found that under TNF-α stimulation PAK1 is phosphorylated in a dynamics starting at threonine-212 followed by threonine-423 phophorylation, which confers the kinase to an active state. Given the fact of failed activation of PAK1 by the variants, and the evidence of their equal stabilization by these PAK1 variants, the TNF-α induced stabilization of PAK1 is, therefore, very likely independent of the kinase activation mechanism.

It still remains outstanding how ECM boosts the TNF-α signaling to induce MMPs by the mesenchymal cells in tissue environment, as shown in Fig. 1. First, we examined whether collagen can boost the JNK signaling, for the reason, in part, of its role in MMP-9 regulation, and under the PAK1 signaling. Our previous work also demonstrated additive stimulation of the minimal promoter of human MMP-9 by collagen and TNF-α, while the convergences between the TNF-α and collagen signals are not known [6]. Thus far, we examined the influence of collagen on the kinetics of JNK activation by human dermal fibroblasts. Cells were cultured on plastic or in type-I collagen to measure the strength and speed of the response. As shown in Fig. 10, phosphorylation of both JNK1 and JNK2 promptly occurred at 5 min after TNF-α challenge, reached to peaks at 15 min, and followed by gradual decline. Although collagen alone could not trigger JNK activation, it did magnify the amplitude of JNK1/2 phosphorylation. Densitometry scanning revealed a two-fold enhancement of phospho-JNK1/2 by the cells cultured in type-I collagen. Therefore, the synergistic effect of collagen to cooperate with TNF-α in induction of MMP-9 can be explained, at least in part, by the amplification of JNK activities.

Cytokines were purchased from R&D Systems. Antibodies against PAK1, p38 MAP kinase, I-kappaB, MMP-3 and MMP-9 were purchased from Santa Cruz Biotechnologies. Antibodies for JNK, phospho-JNK (Phospho-SAPK/JNK (Thr183/Tyr185), phospho-PAK1 (Thr423)/PAK2 (Thr402) were from Cell Signaling Technology. Antibodies for phosphor-212-threonine-PAK1 were from Sigma. Antibodies for GAPDH were from Chemicon. SuperSignal West Fermto Maximum Sensitivity Substrate was from PIERCE. The 5'-670 bp promoter of human MMP-9 was cloned into plasmid pGL2/firefly luciferase (pGL-5'670-MMP9). The Dual-Luciferase^® Reporter (DLR) Assay System was from Promega. Lentivirus encoding PAK1 variants including wild type, triple mutant (H83L, H86L and K299R) and K299R were kindly provided by G. Bokoch (The Scripps Research Institute, La Jolla, CA). Collagenase, inhibitors for IKK (#401486), JNK inhibitors (SP600125), and p38 MAP kinases (SB 239063) were from Calbiochem.

Clinical biopsies including normal skin and chronic wounds were collected according to the protocol approved by Internal Review Board at the University of Southern California and are consent by patients. The 6-mm punch biopsies were placed in 2-ml DMEM with antibiotics (200 U/ml penicillin G sodium, 200 U/ml streptomycin sulfate and 0.5 μg/ml amphotericin B). To accumulate secreted factors the biopsies were incubated in the medium for 6 hours with supply of 5% CO₂ at 37°C. The conditioned medium was cleared of debris by centrifugation at 5,000 g prior to zymography or Western blot analysis.

To isolate human dermal fibroblasts, full thickness skin was treated by 20 mM EDTA in DMEM at 37°C for 3 hrs. After removal of epithelial sheets the dermal tissue was incubated in collagenase (2 mg/ml) in DMEM at 37°C for 16 hrs. The resultant cells were seeded on dishes and cultured in DMEM with 10% FBS. The second passage of cells was used for experiments. Immortalized human keratinocytes were kindly provided by Dr. David Woodley at USC. Primary rat hepatic stellate cells were supplied from the USC Research Center for Alcoholic and Pancreatic Diseases, and cultured in DMEM with 10% FBS. Kratinocytes and human dermal fibroblasts were transduced by lentivirus prepared in 293T cells. To ensure the high transducing efficiency the cells were infected twice. Efficiency of transduction was measured by immunostaining with anti-PAK1 as well by expression of green fluorescent protein as an indicator. To create 3-dimensional culture, fibroblasts were embedded in type-I collagen as previously described [6]. To measure the promoter activities the dermal fibrobalsts were transfected by a reporter plasmid together with pGL/CMV-renilla luciferase as a reference. Luciferase activities were measured by the dual luciferase assay kit.

Cells on plastic or 3D ECM were treated with TNF-α (10 ng/ml), IL-1α (10 ng/ml), TGF-β1 (1 ng/ml) in DMEM with 1% FBS for 16 hrs. Total RNA was extracted using TRIZOL reagent according to the manufacturer's instructions (Invitrogen Life Technologies). First-strand cDNA was produced using First-Strand cDNA Synthesis by SuperScript II Reverse Transcriptase with random primers. Two micrograms of total RNA was used for each reverse transcription reaction mixture (20 μl). Real-time PCR was carried out using an ABI Prism 7900 HT (Applied Biosystems). 10 μl reactions were set up in 384-well PCR plate using the following final concentrations: 1 μmole each of forward and reverse primers, 1× SYBR Green master mix (qPCR Mastermix Plus for SYBR Green I, Eurogentec), and 5 ng of cDNA. For each condition three duplicates were used to minimize the variation. Cycling conditions were as follows: initial step (50°C for 2 min), hot activation (95°C for 10 min), amplification (95°C for 15 s, 60°C for 1 min) repeated 40 times, and quantification with a single fluorescence measurement. Data were analyzed using ABI Prism SDS 2.1 software. The relative gene expression was calculated by the ΔCt method. Briefly, the resultant mRNA was normalized to its own GAPDH. Final results were expressed as n-fold difference in gene expression relative to GAPDH mRNA and calibrator as follows: n-fold = 2^{-(ΔCt sample-ΔCt GAPDH)}, where ΔCt values of the sample and the calibrator were determined by subtracting the average Ct value of the transcript under investigation from the average Ct value of the GAPDH gene for each sample. For human MMP-9 the forward primer has the sequence GGG AGA CGC CCA TTT CG and the reverse primer is CGC GCC ATC TGC GTT T. For GAPDH, the forward primer, GAA GGT GAA GGT CGG AGT 3', and backward primer GAA GAT GGT GAT GGG ATT TC 3' 20 mer. For MMP-14, the primers are TGG AGG AGA CAC CCA CTT TGA, and GCC ACC AGG AAG ATG TCA TTT C. For MMP-2, the primers are GAG AAC CAA AGT CTG AAG AG, and GGA GTG AGA ATG CTG ATT AG. For MMP-3, the primers are GCT GCA AGG GGT GAG GAC AC, and GAT GCC AGG AAA GGT TCT GAA GTG. The primers for TIMP-1 are TCT GGC ATC CTC TTG TTG CTA T, and CCA CAG CGT CGA ATC CTT.