Inhibition of cyclooxygenase‐1 and ‐2 activity in keratinocytes inhibits PGE₂ formation and impairs vascular endothelial growth factor release and neovascularisation in skin wounds

Abstract

Inhibition of cyclooxygenase (Cox) enzymatic activity by non‐steroidal anti‐inflammatory drugs (NSAIDs) provides the molecular basis of analgesia following wounding or surgery. This study investigated the role of Cox activity in the regulation of vascular endothelial growth factor (VEGF) expression in keratinocytes and the formation of new blood vessels in acute wounds in mice. To this end, human HaCaT keratinocytes were stimulated with epidermal growth factor (EGF). EGF increased Cox‐1 mRNA in the presence of the constitutively expressed Cox‐1 protein in keratinocytes. EGF coinduced Cox‐2 and VEGF₁₆₅ mRNA and protein expression and an accumulation of prostaglandin E₂ (PGE₂) in cell culture supernatants. Inhibition of Cox isozyme activity by Cox‐1 and ‐2 siRNA or ibuprofen reduced PGE₂ and VEGF₁₆₅ release from keratinocytes. In a mouse model of excisional wound healing, Cox‐2 and VEGF₁₆₅ expression were colocalized in the granulation tissue of acute wounds. Oral treatment of mice with the Cox‐1 and ‐2 inhibitor diclofenac was associated with reduced levels of VEGF₁₆₅ protein and an impaired blood vessel formation in acute wound tissue. In summary, our data suggest that a reduction of PGE₂‐triggered VEGF₁₆₅ protein expression in wound keratinocytes is likely to contribute to the observed impairment of wound neovascularisation upon Cox inhibition.

Introduction

Wound healing represents a tightly controlled cascade of cellular movements that aim to restore impaired tissue upon injury. Haemorrhage‐mediated release of growth factors, cytokines and other mediators upon platelet degranulation kick‐starts the tissue regeneration process 1. This initial process is followed by acute wound inflammation that is central to the formation of new tissue during wound granulation 2. Extensive proliferation and directed migration of resident cells such as keratinocytes and fibroblasts coordinately result in the epithelial coverage of the wounded area and the formation of extracellular matrix (ECM). The differentiation of myofibroblasts within the wound tissue supports wound closure by contraction of the overall wound area 3. These processes are highly dependent on the formation of new blood vessels within the granulation tissue as the supply of oxygen and nutrients is essential for the formation of new tissue 4, 5. A series of studies have demonstrated that the pro‐angiogenic mediator vascular endothelial growth factor (VEGF) is pivotal to wound healing 6, 7. Besides macrophages, keratinocytes are also a well‐described source of wound‐derived VEGF 8, 9. Interestingly, both macrophages and keratinocytes have also been described to be the cellular basis of an observed differential cyclooxygenase (Cox)‐1 and ‐2 isoenzyme expression and PGE₂/D₂ formation at the wound site 10, 11.

Currently available non‐steroidal anti‐inflammatory drugs (NSAIDs) are potent inhibitors of Cox‐1 and Cox‐2 12. It has been known for decades that these drugs display both wanted and side effects by inhibiting prostaglandin synthesis 13. Thus, it is tempting to argue for a substantial interference of those Cox‐inhibitory drugs with physiological wound‐healing conditions. Expectedly, this assumption was conclusively true: inhibition of Cox‐1 or Cox‐2 isozyme activity indeed impaired tissue repair in different wound‐healing models in rats and mice 10, 14.

However, despite this knowledge, NSAIDs are commonly used to treat algetic conditions especially associated with trauma or injury. Nevertheless, the number of functional studies on the role of NSAIDs in acute cutaneous wound healing is very limited. Here, injury‐induced Cox‐2 was functionally related to wound impairments upon its selective inhibition, leading to disturbed re‐epithelialisation, ECM formation, angiogenesis and myofibroblast differentiation 14, 15, 16. These findings are complemented by additional studies that emphasise the essential role of the keratinocyte‐expressed, constitutive Cox‐1 isoform for skin repair 10, 11. However, the underlying mechanisms that control Cox‐mediated cellular responses in wounds still remain largely untouched. Thus, the present study focused on a potential functional connection of an induced Cox isozyme expression, PGE₂ formation and the subsequent formation of the angiogenic factor VEGF₁₆₅ in keratinocytes in vitro and during wound healing in vivo.

Materials and methods

Animals

Female C57Bl/6J mice were obtained from Charles River (Sulzfeld, Germany). At 12 weeks of age, the mice were caged individually using cages with an enriched environment. The mice were randomly assigned to different experimental groups, monitored for body weight and wounded as described below. The animal experiments were performed according to the guidelines and with the approval of the local ethics animal review board (Regierungspräsidium Darmstadt, D‐64278 Darmstadt, Germany). The approval number for this project was II25.3‐19c20/15‐F95/11.

Wounding of mice

Wounding of mice was performed as described previously 17, 18. The mice were briefly anaesthetised using ketamine (80 mg/kg) for analgesia and isoflurane (4% vol). Subsequently, six full‐thickness wounds (5 mm in diameter, 3–4 mm apart) were made on the backs of the mice by excising the skin and the underlying panniculus carnosus. The wounds were allowed to form a scab. Mice were sacrificed by cervical dislocation and subsequent bleeding. An area of 7–8 mm in diameter, which included the granulation tissue and the complete epithelial margins, was excised at the indicated time points for analysis. Back skin from non‐wounded mice served as a control. Wounds (n = 12) isolated from animals (n = 4) were used for RNA analysis. For protein analysis, wounds (n = 8) from individual mice (n = 4) were used.

Treatment of mice

Diclofenac (Sigma, Deisenhofen, Germany), SC‐560 (Witega, Berlin, Germany) or celecoxib (Celebrex^®, Pharmacia AG, Erlangen, Germany) were orally administered twice per day by gastrogavage (2·5 mg/kg/12 hours) in methylcellulose. Methylcellulose alone served as a control (mock).

Cell culture

Quiescent HaCaT keratinocytes 19 were stimulated with 20% (v/v) fetal calf serum (FCS), epidermal growth factor (EGF) (10 ng/ml) or a combination of cytokines (25 ng/ml IL‐1β, 50 ng/ml tumor necrosis factor (TNF)‐α) in the presence or absence of inhibitors. EGF and cytokines were obtained from Peprotech (Mannheim, Germany). Ibuprofen was obtained from Sigma (Deisenhofen, Germany). The HaCaT keratinocyte cell line was provided by P. Boukamp (DKFZ, Heidelberg, Germany).

Cell viability assay

Cell viability was assessed using the CytoTox96^® Non‐Radioactive Cytotoxicity Assay (Promega, Mannheim, Germany) according to the manufacturer's instructions.

RNA isolation

RNA isolation was performed as described previously 20. Every experimental time point depicts a total of 12 wounds (n = 12) isolated from four individual mice (n = 4) for analysis. Wound tissue had been separated into outer wound margin and inner granulation tissue compartments prior to RNA isolation.

RNase protection analysis

Ribonuclease (RNase) protection assays were performed as described previously 17, 20. All samples were quantified using phosphorimager photo‐stimulated luminescence (PSL) counts per 20 µg of total cellular RNA. The human cDNA probes were cloned using reverse transcriptase‐polymerase chain reactions. The probes corresponded to nucleotides (nt) 7–260 (for human Cox‐1, GenBank accession number U63846), nt 1819–2040 (for human Cox‐2, GenBank accession number M90100) and nt 961–1070 (for human glyceraldehyde‐3‐phosphate dehydrogenase (GAPDH), GenBank accession number M33197) of the published sequences.

Quantitative real‐time polymerase chain reaction (qRT‐PCR)

qRT‐PCR was performed to assess the expressions of Cox‐1, Cox‐2 and the VEGF₁₆₅ splice variant. TaqMan^® (Applied Biosystems, Darmstadt, Germany) Fast Advanced Master Mix (cat. No. 4444557) and the pre‐designed qRT‐PCR assays were purchased at Applied Biosystems (Darmstadt, Germany): human Cox‐1 (Hs00377726_m1), human Cox‐2 (Hs00153133:m1), human VEGF (Hs00173626_m1), mouse VEGF (Mm00437304_m1), mouse α‐smooth muscle actin (Mm00725412_s1), mouse EGF‐like module‐containing mucin‐like hormone receptor‐like 1 (Mm00802529_m1) or 4352339E for human GAPDH. qRT‐PCR was performed on 7500 Fast Real‐Time PCR system (Applied Biosystems) as follows: an initial denaturation step at 95°C (for 20 seconds) followed by 40 cycles at 95°C (for 3 seconds) and 60°C (for 30 seconds). Analyses of qRT‐PCR runs were performed by the sequence detector software.

Preparation of protein lysates

Wound tissue or cultured cells were homogenised in a lysis buffer (1% Triton X‐100, 20 mM Tris/HCl pH 8·0, 137mM NaCl, 10% glycerol, 1 mM DTT, 5 mM EDTA, 10 mM NaF, 2 mM Na₃VaO₄, 1 mM PMSF, 5 ng/ml aprotinin, 5 ng/ml leupeptin 50 nM okadaic acid). Extracts were cleared by centrifugation. Protein concentrations were determined using the BCA Protein Assay Kit (ThermoScientific, Köln, Germany).

Western blot analysis

Total protein (25–50 µg) were analysed by sodium dodecyl sulfate (SDS) gel electrophoresis. After transfer to a nitrocellulose membrane, specific proteins were detected using antibodies against Cox‐1 (sc‐1752), Cox‐2 (sc‐1747) from Santa Cruz (Heidelberg, Germany) or β‐actin (A5441) from Sigma (Deisenhofen, Germany). A secondary antibody coupled with horseradish peroxidase (Bio‐Rad, Munich, Germany) and the enhanced chemiluminescence (ECL) (Amersham, Freiburg, Germany) detection system was used to visualise the proteins.

Enzyme‐linked immunosorbent assay (ELISA)

Quantification of human VEGF₁₆₅ was performed using the murine Quantikine ELISA kit (R&D Systems, Wiesbaden, Germany) according to the instructions of the manufacturer.

Silencing of Cox‐1 and ‐2 expression by siRNA

HaCaT keratinocytes were cultured in 12‐well plates (2 × 10⁵ cells per well) to reach 40–60% confluency. Cells were subsequently transfected with siRNA (f.c. 10 nM) using Lipofectamine^® and OptiMEM (Life Technologies, Darmstadt, Germany) as described by the manufacturer. The Cox‐1‐ and Cox‐2‐specific siRNA was purchased from Ambion (Life Technologies, Darmstadt, Germany).

Immunohistochemistry

Wound biopsies were isolated from the back, fixed in zinc fixative solution (0·05% CaAc^.2H₂O, 0·5% ZnAc^.2H₂O, 0·5% ZnCl₂ in 0·1 M Tris‐Cl, pH 7·4) and embedded in paraffin. For immunostaining, sections were incubated overnight at 4°C with antibodies introduced against mouse CD31 (Chemicon/Merck, #CBL1337, Darmstadt, Germany), F4/80 (rat monoclonal, clone CI:A3‐1, #MCAP497 GA; Bio‐Rad AbD SeroTec, Puchheim, Germany) or α‐smooth muscle actin (mouse monoclonal alkaline phosphatase conjugate, clone 1A4, #A5691; Sigma Aldrich, München, Germany). Primary antibodies were detected using a biotinylated secondary antibody. The sections were subsequently stained with the avidin‐biotin‐peroxidase complex system (Santa Cruz, Heidelberg, Germany) using 3,3′‐diaminobenzidine‐tetra‐hydrochloride (Dako, Hamburg, Germany) or Permanent Red Substrate (Zytomed, Berlin, Germany) as chromogenic substrates. Finally, sections were counterstained with haematoxylin and mounted.

Blood vessel formation analysis

Total surface values of blood vessels in CD31‐stained histological sections were analysed and calculated using the BZ analyser in Hybrid Cell Count Image Analysis (Keyence, Osaka, Japan).

Determination of PGE₂ by LC/MS‐MS

Prostanoids were extracted from 200 µl cell culture samples and amounts were determined by liquid chromatography‐mass spectrometry (LC/MS‐MS) as described previously 10. In brief, for liquid–liquid extraction, samples were incubated twice with 600 µl ethylacetate, and the organic phase was removed at 45°C under nitrogen. For reconstitution, 50 µl of acetonitrile/water/formic acid (20:80:0·0025, v/v, pH 4·0) was used, and after centrifugation for 5 minutes at 18 000 g, samples were used for injection. Prostanoids were separated with a Synergi 150 × 2 mm Hydro‐RP column (Phenomenex, Aschaffenburg, Germany) and determined with an API 4000 triple quadrupole spectrometer (AB Sciex, Darmstadt, Germany). Standard PGE₂ was set in the range from 0·01 to 125 ng/ml.

Statistical analysis

Data are shown as means ± standard deviation (SD). Data analysis was carried out using the unpaired Student's t test or one‐way analysis of variance (ANOVA, Dunett's method or Bonferroni's Multiple Comparison test as indicated) on raw data and ‘Graph Pad Prism’ software (GraphPad Software, San Diego, USA), version 5.02. A P value < 0·05 was considered statistically significant.

Results

Cox gene expression by growth factors and inflammatory cytokines in keratinocytes

EGF is a well‐known mediator in the control of epithelial processes during wound healing 1, 21. In particular, EGF induces a robust expression and release of the VEGF₁₆₅ splice variant from cultured keratinocytes in vitro and also from wound keratinocytes in vivo 9, 17. The release of VEGF₁₆₅ protein might also be functionally linked to a simultaneous release of PGE₂ by EGF as keratinocytes are sensitive to this particular end product of the Cox‐mediated biosynthetic pathway 22. Therefore, we first stimulated cultured HaCaT keratinocytes 19 using wound‐related growth factors or cytokines. As given in Figure 1, keratinocytes expressed an inducible pattern of Cox‐1‐ and Cox‐2‐specific mRNA species upon growth factor stimulation. Notably, Cox‐1 showed a marked expressional induction at the mRNA level by FCS and EGF (Figure 1A). This induction, however, was not translated at the protein level as the keratinocytes constitutively expressed the Cox‐1 protein in the absence and presence of EGF (Figure 2A). Interestingly, Cox‐2 was also induced by serum and EGF (Figure 2B). In contrast to elevated Cox‐1 mRNA levels, the prominent EGF‐induced Cox‐2 mRNA expression led to increased Cox‐2 protein in keratinocytes (Figure 2B). However, the Cox‐1 expression was not responsive to inflammatory cytokines (IL‐1β, TNFα, IFNγ), whereas Cox‐2 showed a strong and prolonged expression upon cytokine stimulation (Figure 1C). As the next step, we assessed the release of PGE₂ from the cells to determine potential alterations in Cox isozyme activity upon EGF stimulation. As shown in Figure 2B, keratinocytes released significant amounts of PGE₂ upon EGF stimulation. Inhibition of Cox‐1 activity by the Cox‐1‐specific inhibitor SC‐560 did not reduce the overall EGF‐induced PGE₂ formation in relation to mock‐treated cells that had not been exposed to the inhibitor. However, enzymatic inhibition of both Cox isozymes by the Cox‐1/Cox‐2 inhibitor ibuprofen completely blunted the EGF‐induced PGE₂ release from the cells (Figure 2B).

Figure 1.

Induction of Cox‐1 and −2 mRNA expression in keratinocytes. Serum‐starved HaCaT keratinocytes were stimulated for the indicated time periods with 20% (v/v) fetal calf serum (FCS), EGF (10 ng/ml) or a combination of cytokines (2 nM IL‐1β, 20 ng/ml TNF‐α). Total cellular RNA was analysed by RNase protection assay for the presence of Cox‐1 (A) and Cox‐2 (B, C) mRNA as indicated. A quantification of Cox isoform mRNA (x‐fold of non‐stimulated control cells) is shown in the right panels. **P < 0·01; *P < 0·05 (Student's unpaired t test) as compared to control. Data represent means ± SD obtained from four independent cell culture experiments (n = 4).

Figure 2.

Cox protein expression and PGE₂ formation in keratinocytes. (A) Immunoblots for Cox‐1 and Cox‐2 protein in HaCaT keratinocytes in the presence or absence of epidermal growth factor (EGF) (10 ng/ml) as indicated. β‐actin was used to control equal loading. (B) PGE₂ concentrations (pg per ml supernatant) in keratinocyte cell culture supernatants as determined by LC/MS‐MS in the presence or absence of EGF (10 ng/ml), SC‐560 (200 ng/ml) or ibuprofen (40 ng/ml). **P < 0·01; ns, not significant (Student's unpaired t test) as compared to non‐EGF control. Data represent means ± SD obtained from four independent cell culture experiments (n = 4).

EGF‐induced VEGF ₁₆₅ expression is related to PGE₂ formation in keratinocytes

EGF is a potent expressional inducer of the VEGF₁₆₅ splice variant in keratinocytes 9, 17. Therefore, it was tempting to argue that the observed EGF‐mediated increase in the VEGF₁₆₅ mRNA expression (Figure 3A, left panel) and release (Figure 3A, right panel) might be functionally connected to the coinduced Cox‐mediated PGE₂ formation (Figure 2B) in keratinocytes. Our experiments argued for such a functional connection as the inhibition of Cox‐1 activity by SC‐560 (Figure 3B, left panel) as well as the inhibition of Cox‐2 activity by ibuprofen (Figure 3B, right panel) partially, but significantly, reduced VEGF₁₆₅ protein levels in supernatants from EGF‐treated keratinocytes without affecting the overall viability of the cells (Figure 3C).

Figure 3.

Inhibition of Cox activity attenuates epidermal growth factor (EGF)‐mediated vascular endothelial growth factor (VEGF) release from keratinocytes. (A) qRT‐PCR quantification of VEGF mRNA (left panel) or ELISA for VEGF₁₆₅ protein (right panel) in EGF (10 ng/ml)‐stimulated keratinocytes at indicated time points. Non‐stimulated keratinocytes served as control. **P < 0·01; *P < 0·05 (Student's unpaired t test) compared to control. Bars indicate the mean ± SD obtained from three independent cell culture experiments (n = 3). (B) VEGF₁₆₅ protein in supernatants from EGF (10 ng/ml) ‐stimulated keratinocytes in the presence or absence of SC‐560 (left panel) or ibuprofen (right panel) at indicated time points and concentrations. Non‐stimulated keratinocytes served as control. *P < 0·05; **P < 0·01 (Bonferroni's multiple comparison test) compared to EGF‐treated control cells. Bars indicate the mean ± SD obtained from seven independent cell culture experiments (n = 7). (C) Keratinocyte viability in EGF (10 ng/ml)‐stimulated keratinocytes in the presence or absence of SC‐560 (left panel) or ibuprofen (right panel) at indicated time points and concentrations.

To further complement our inhibitor‐based data on VEGF₁₆₅ expression and PGE₂ formation, we interfered with Cox isozyme activities using an siRNA approach. As given in Figure 4A, the Cox‐1‐ or Cox‐2‐specific siRNA species markedly reduced the expression of the constitutively expressed Cox‐1 as well as the EGF‐induced Cox‐2 protein, respectively. In the present study, we also added a mixture of both anti‐Cox‐1 and ‐2 siRNAs to cells to mimic the inhibitory effect of ibuprofen on Cox‐1 and ‐2 enzyme activity (Figure 2B) at the expressional level (Figure 4A). In line with the failure of SC‐560 to inhibit a Cox‐1‐driven PGE₂ release from keratinocytes (Figure 2B), the cells did not reduce the EGF‐induced PGE₂ formation upon loss of Cox‐1 protein by anti‐Cox‐1 siRNA treatment. Moreover, and as also seen for ibuprofen (Figure 2B), EGF‐induced PGE₂ levels were blunted only after simultaneous expressional reduction of both Cox isozymes (Figure 4B). Determination of VEGF₁₆₅ levels from the respective cell culture supernatants showed a moderate but significant decrease in EGF‐stimulated VEGF₁₆₅ release from HaCaT keratinocytes. This finding suggested that about 30% of EGF‐induced VEGF₁₆₅ biosynthesis might be functionally linked to Cox isozyme activity in keratinocytes.

Figure 4.

Attenuation of PGE₂ and vascular endothelial growth factor (VEGF₁₆₅) release by Cox‐specific siRNAs. (A) Immunoblot showing the attenuation of Cox‐1 or Cox‐2 protein expression in epidermal growth factor (EGF) (10 ng/ml) ‐stimulated keratinocytes by siRNAs. Cox‐1/2 represents simultaneous application of both Cox‐1 and −2 siRNAs. A scrambled siRNA was used as control (scr). Protein expression was analysed after 24 hours of treatment. β‐actin served as a loading control. PGE₂ concentrations as determined by LC/MS‐MS (B) or VEGF₁₆₅ protein as determined by enzyme‐linked immunosorbent assay (C) in keratinocyte cell culture supernatants in the presence or absence of EGF (10 ng/ml) and scrambled (scr) or Cox‐1‐ and Cox‐2‐specific siRNAs as indicated. *P < 0·05 (Student's unpaired t test) as compared to scrambled‐RNA‐treated control. Data represent means ± SD obtained from four independent cell culture experiments (n = 4).

Interaction of Cox isoenzymes and angiogenic processes in acute wounds

Tissue repair in acute wounds upon injury has been described to involve the expression of both Cox‐1 and ‐2 isozymes 10, 11, 14, 23. To investigate a potential link between the presence of Cox‐1 and −2 and VEGF₁₆₅ expression in acute wound tissue, we used an excisional wound‐healing model in mice 24. Here, isolated wound tissue was further separated into the outer wound margin and the inner granulation tissue area 10. In conformation of previous data, we observed a differential expression of the Cox‐1 and Cox‐2 isoforms in particular wound compartments. Cox‐1 mRNA levels were induced within the wound margins, whereas Cox‐2 expressions appeared elevated within the granulation tissue (Figure 5A and B). Interestingly, we found a strong spatial correlation of wound‐induced Cox‐2 expression and elevated levels of wound VEGF mRNA and VEGF₁₆₅ protein expression within the granulation tissue (Figure 5C and D). To strengthen a potential functional link between wound Cox isozyme activity and injury‐induced VEGF₁₆₅ expression at wounded sites, we treated wounded mice with an oral administration of Cox‐1 and ‐2 inhibitors. First, we observed a moderate (∼25%) decrease of VEGF₁₆₅ protein in acute wound tissue upon inhibition of both Cox isozymes by diclofenac (Figure 6A). However, the observed moderate reduction in VEGF₁₆₅ protein translated into a disturbed wound angiogenic phenotype. In accordance with the effects of ibuprofen on keratinocyte VEGF₁₆₅ release in vitro (Figure 2B), we observed a marked and significant decrease in blood vessel density and, thus, in overall blood vessel formation, which was restrictive upon simultaneous Cox‐1 and −2 inhibition by diclofenac treatment of mice (Figure 6B). Immunological stainings indicate a diclofenac‐mediated disturbance of the angiogenic process in acute wounds, which is primarily driven by the angiogenic growth factor VEGF₁₆₅ 1, 4, 6, 7 (Figure 6C).

Figure 5.

Cox‐2 and vascular endothelial growth factor (VEGF) expression colocalise in granulation tissue of acute wounds. qRT‐PCR quantification of Cox‐1 (A), Cox‐2 (B) and VEGF (C) mRNA or enzyme‐linked immunosorbent assay (ELISA) for VEGF₁₆₅ protein (D) expression in separated outer wound margin (wm) and inner wound granulation tissue (gt) compartments of acute murine wounds at the indicated time points after injury. Non‐wounded back skin served as a control (ctrl). **P < 0·01; *P < 0·05 (Student's unpaired t test) compared to wound margin (wm) tissue. Bars indicate the mean ± SD obtained from wounds (n = 12) isolated from four individual animals (n = 4). For ELISA, wounds (n = 8) from individual mice (n = 4) were used.

Figure 6.

Inhibition of Cox‐1 and Cox‐2 impairs blood vessel formation in acute wounds. (A) Enzyme‐linked immunosorbent assay (ELISA) for vascular endothelial growth factor (VEGF₁₆₅) protein in acute wound tissue at day 5, post‐wounding. Mice had been treated during healing by PBS or diclofenac (2·5 mg/kg/12 hour). Non‐wounded back skin served as a control (ctrl). (B) Analysis of blood vessel formation in acute 5‐day wounds of SC‐560‐, celecoxib‐ or diclofenac‐treated mice (2·5 mg/kg/12 hour). PBS‐treated mice served as control. The percent of total blood vessel‐covered area per wound section is given. **P < 0·05; *P < 0·01 (Student's unpaired t test) as indicated by the brackets. Bars indicate the mean ± SD obtained from wounds (n = 8) from individual mice (n = 4). (C) Paraffin‐fixed 5‐day wound sections of mock‐, SC‐560‐, celecoxib‐ or diclofenac‐treated mice (2·5 mg/kg/12 hour) were stained for the endothelial cell marker CD31. The squares indicate the localisations of the respective magnifications. Scale bars are given in the photographs.

Function of Cox isoenzymes on macrophage infiltration and myofibroblast formation in acute wounds

Wound inflammation and contraction contribute to normal skin repair upon injury 1, 3. For this reason, we finally assessed wound macrophage and myofibroblast numbers in the presence of a selective Cox‐1 and ‐2 inhibition in mice (Figure 7A and B). First, we determined the mRNA expression levels of EGF‐like module‐containing mucin‐like hormone receptor‐like 1 (Emr‐1) (Figure 7A). Emr‐1 represents the gene encoding the F4/80 epitope and can be utilised as a molecular marker to determine macrophage infiltration 2. However, the influx of macrophages into acute wounds was not altered by the inhibition of Cox isoenzymes (Figure 7A and C). In addition, we also analysed the mRNA levels of α‐smooth‐muscle actin (α‐SMA), which was used as a molecular marker for differentiating myofibroblasts 3. Notably, the formation of myofibroblasts in acute wounds appeared to be selectively sensitive to an inhibition of the Cox‐1 isoenzyme (Figure 7B and C).

Figure 7.

Inhibition of Cox‐1 impairs myofibroblast formation in acute wounds. qRT‐PCR quantification of Emr‐1 (A) or α‐SMA (B) mRNA expression in acute 5‐day wounds of SC‐560‐, celecoxib‐ or diclofenac‐treated mice (2·5 mg/kg/12 hour). PBS‐treated mice served as control. **P < 0·05; *P < 0·01 (Student's unpaired t test) compared to control. Bars indicate the mean ± SD obtained from wounds (n = 8) from individual mice (n = 4). (C) Paraffin‐fixed 5‐day wound sections of mock‐, SC‐560‐, celecoxib‐ or diclofenac‐treated mice (2·5 mg/kg/12 hour) were double stained for the F4/80 macrophage marker protein (brown staining) and the α‐SMA myofibroblast marker protein (red staining) as indicated. The squares indicate the localisations of the respective magnifications. Scale bars are given in the photographs.

Discussion

The cytosolic phospholipase A₂, the enzyme that provides the arachidonic acid substrate for Cox activity from lipid membrane constituents, is constitutively expressed in skin keratinocytes 25. In addition, Cox‐1, but not Cox‐2, was shown to be expressed in basal and suprabasal keratinocytes of the epidermis and in hair follicles 26. Therefore, it is reasonable to suggest that non‐wounded skin tissue drives a Cox‐1‐coupled prostaglandin biosynthesis as Cox‐2 is not expressed in normal skin 10. In line with that argument, the constitutive presence of PGE₂/PGD₂ in normal skin is completely blunted in Cox‐1‐deficient transgenic mice 10. Accordingly, a keratinocyte‐targeted overexpression of Cox‐2 protein (using a keratine 14 promoter‐directed Cox‐2 expression) triggered elevated skin PGE₂ levels. Interestingly, those elevated PGE2 levels were associated with a hyperplastic epidermis in these mice 27. Thus, the inducible Cox‐2 isoform now appears to function in the control of keratinocyte proliferation during dynamic cutaneous processes such as wound healing. Controlled keratinocyte proliferation is key to an undisturbed healing of acute skin wounds 1, 4, 5. Interestingly, PGE₂ is related to high proliferation rates in keratinocytes 22, 28. In contrast to the constitutive Cox‐1‐driven PGE₂ formation in normal skin keratinocytes 10, it appears that an induced keratinocyte proliferation is linked to the expression of Cox‐2 29. Indeed, Cox‐2 is induced in hyperplastic conditions of the skin 30, 31, and both Cox isoforms have been shown to inhibit the premature terminal differentiation of epidermal keratinocytes in a model of induced tumour formation 32. Thus, it is tempting to argue here that both the observed constitutive presence of Cox‐1 and an EGF‐induced Cox‐2 expression in keratinocytes might contribute to keratinocyte proliferation during wound healing. Therefore, it appears reasonable that EGF, a growth factor rapidly released upon wounding, was able to stimulate the observed expression of Cox‐1 (mRNA) and Cox‐2 (mRNA and protein) and a subsequent PGE₂ release from cultured keratinocytes.

In addition, the PGE₂ function might not be restricted only to wound keratinocyte proliferation as PGE₂ is able to directly stimulate intracellular signalling from the EGF‐receptor (EGF‐R) in primary murine keratinocytes. In this setting, PGE₂ led to an activation of both Erk‐1/2 and phosphatidyl‐inositol‐3‐kinase (PI3K) in cells. These kinases activated the binding of the transcription factors nuclear factor‐κB (NFκB), activator protein‐1 (AP‐1) and cyclic adenosine mono‐phosphate (cAMP) response element binding protein (CREB) to their respective promoter regions of the VEGF gene 22. Even more important, this study demonstrated the particular importance of the Cox‐2 isoform in control of the VEGF gene expression from keratinocytes. Transgenic mice containing a keratin‐14 promoter‐driven Cox‐2 expression in basal skin keratinocytes showed marked elevated VEGF mRNA expression in skin tissue 22. These findings underpin data from our study that have shown a dependency of Cox‐2‐associated PGE₂ production and VEGF expression in keratinocytes. We observed prominent inhibitory effects on EGF‐induced PGE₂ formation and VEGF₁₆₅ release from the cells when Cox‐2 activity was affected at both pharmacological and translational levels.

During wound healing, wound margin keratinocytes represent an important cellular source of the angiogenic growth factor VEGF 8, 9, 17. Thus, it is tempting to argue that the widely used analgetic NSAIDs, based on their potency to inhibit Cox‐1 and Cox‐2 12, might therefore also interfere with keratinocyte‐driven VEGF₁₆₅ formation as a consequence. Although human clinical trials to assess the effect of Cox inhibitors on cutaneous wound healing are not available, one might expect that the inhibition of total Cox isozyme activity by NSAIDs should interfere with wound angiogenic processes. This notion is indeed supported here as our study implicated a significant reduction in wound VEGF₁₆₅ protein content upon dual Cox isozyme inhibition by diclofenac. Additionally, the reduced amounts of wound‐expressed VEGF₁₆₅ translated into a marked disturbance in the formation of new vessels in acute wounds. In line with our finding, two of the most recent studies again confirm that notion. Celecoxib severely disturbed skin repair in two distinct rodent models of acute wound healing; celecoxib interfered with regeneration of murine excisional wounds 16 as well as with the effective integration of skin flaps in rats 15. Together with the presented data, both studies argue that wound VEGF expression and angiogenesis were dependent on wound Cox‐2 activity. It is again noteworthy to refer to our observation that both the restricted celecoxib‐mediated inhibition of Cox‐2 activity and the concurrent inactivation of both Cox1‐ and ‐2 isozymes by diclofenac caused impaired vessel formation in murine acute wounds. Although Cox‐2 appears to maintain the main function in stimulating VEGF₁₆₅ expression in keratinocytes, this function is also dependent on Cox‐1 activity.

Notably, the significant reduction of differentiated myofibroblasts upon selective inhibition of wound‐expressed Cox‐1, but not Cox‐2 activity, again argues for a particular role of wound margin keratinocyte‐derived prostaglandins in granulation tissue formation. This is true as essentially Cox‐1, and again not Cox‐2, was predominantly expressed in wound margin keratinocytes (this study; 10). The differential contribution of Cox‐1 and ‐2 also resembles regulatory conditions of keratinocyte terminal differentiation, where again both Cox isozymes must be actively involved 32.

The presented data shed more light on the actions of Cox‐inhibitory NSAIDs during acute wound healing. Although these drugs are widely used as analgetics to relieve pain upon trauma, wounding or surgery, there is limited information on the role of side effects of drug action in tissue repair. Our results strengthen the functional role of Cox‐2 in the regulation of VEGF₁₆₅ expression in keratinocytes. Oral administration of the prototypical NSAID diclofenac reduced VEGF₁₆₅ protein and disturbed blood vessel formation in a mouse model of acute wound healing. Thus, NSAIDs must be considered to interfere with the formation of new vessels during skin repair. The potential impairment of wound healing by NSAIDs should be considered in post‐injury analgesic medication in humans and might be overcome by topical addition of PGE derivatives.