ORIGINAL ARTICLESee related Commentary on page 985

The E¡ects of Cyclooxygenase Isozyme Inhibition onIncisional Wound Healing in Mouse Skin

Karin Mˇller-Decker,Wolfgang Hirschner, Friedrich Marks, and Gerhard FˇrstenbergerDeutsches Krebsforschungszentrum, Research ProgramTumor Cell Regulation, Im Neuenheimer Feld 280, D-69120 Heidelberg, Germany

In addition to their proin£ammatory activities, prosta-glandins recently have been shown to be bene¢cialin the resolution of tissue injury and in£ammation.Thus, inhibition of cyclooxygenase-2, the predominantprostaglandin endoperoxide synthase under theseconditions, may not only result in attenuatingthe in£ammatory response but also in delaying tissueregeneration and repair. To this end, we investigatedcyclooxygenase isozyme expression and the e¡ectsof cyclooxygenase inhibitors on wound healing uponfull-thickness incisions in mouse skin. Immunohisto-chemical analysis revealed prominent expression ofcyclooxygenase isozymes in keratinocytes of the hyper-plastic epithelium, with cyclooxygenase-1 immuno-signals predominating in the suprabasal compartmentand cyclooxygenase-2 immunosignals spread through-out the whole epidermis. Moreover, dendritic cells,

resembling Langerhans cells, as well as endothelial cellsand macrophages in the vicinity of or within thegranulation tissue were found to express both isozymes.Inhibition of prostaglandin E2 synthesis by oraladministration of the cyclooxygenase-1-selective inhi-bitor SC-560 or the cyclooxygenase-2-selective inhibi-tor valdecoxib did not retard wound healing in mouseskin macroscopically. Except for a slight transient retar-dation of epithelialization early after wounding wound-induced neoangiogenesis, collagen deposition, and therestoration of tensile strength were not delayed by theseagents. Likewise, the nonselective inhibitor indometha-cin had no e¡ect on the tensile strength of incisionalskin wounds. Key words: angiogenesis/cyclooxygenase/epidermis/SC-560/valdecoxib. J Invest Dermatol 119:1189 ^1195,2002

Nonsteroidal anti-in£ammatory drugs (NSAID) areamong the most commonly used pharmaceuti-cals worldwide. The anti-in£ammatory e¡ect ofNSAID is thought to be mainly due to the inhi-bition of cyclooxygenases (COX) (Vane, 1998).

COX are bifunctional enzymes catalyzing the ¢rst two steps inprostanoid biosynthesis (Smith et al, 2000). Using arachidonicacid as a substrate, the resulting endoperoxide prostaglandin(PG)H2 is further converted by prostanoid synthases into the bio-logically active prostanoids such as the prostaglandins PGE2,PGF2a, PGD2, PGI2, and thromboxane A2 in a cell-type-speci¢cmanner.There are two COX isozymes, i.e., a constitutive COX-1and an inducible COX-2 the expression of which is stimulated bya wide variety of growth factors, cytokines, and hormones, inparticular in the course of in£ammatory processes (Williamset al, 1999). A transient upregulation of COX-2 protein, but notof COX-1, has been shown to occur upon super¢cial skinwounding (Scholz et al, 1995). A permanent COX-2 overexpres-sion has been found to be causally related to tumorigenesis (Neu-fang et al, 2001). The anti-in£ammatory e¡ects of NSAID mayrelate to an inhibition of COX-2 rather than of COX-1 (Smithet al, 1998). Resolution of the X-ray crystal structure of both COX

isozymes (Picot et al, 1994; Luong et al, 1996) led to the develop-ment of isozyme-selective inhibitors (Talley, 1999).A major adverse e¡ect of conventional NSAID inhibiting both

COX-1 and COX-2 is injury to the gastric mucosa (Soll et al,1991). In contrast, selective COX-2 inhibitors have been foundto be devoid of such side-e¡ects both in animals and humans(Masferrer et al, 1994; Seibert et al, 1994; Chan et al, 1995; Laineet al, 1999; Simon et al, 1999). There is nevertheless some concernas to adverse e¡ects of COX-2 inhibitors on epithelial woundhealing as tissue damage is a particular strong inducer of COX-2expression (Scholz et al, 1995; Kishimoto et al, 1998; Tessner et al,1998).To investigate the role of selective COX isozyme inhibition in

wound repair we have used the established model of acute inci-sional wounding in the skin of mice. Repair of full-thicknessincisions was followed up for 28 d and re-epithelialization,neovascularization of the wound bed, and collagen deposition aswell as the regain of tensile strength have been used as determi-nants of the healing process. As shown here, neither COXisozyme-selective inhibitors nor a nonselective COX inhibitorexhibited any signi¢cant e¡ect on skin wound healing.

MATERIALS ANDMETHODS

Materials Control diet 5010 and diet supplemented with valdecoxib andSC-560 were kindly supplied by Searle/Pharmacia (Skokie, IL) and storedat 41C until use. Indomethacin was from Sigma (Deisenhofen, Germany).Goat polyclonal anti-COX-1 (SC1754) and anti-COX-2 (SC1745) anti-bodies, peroxidase-conjugated goat anti-rabbit IgG, goat anti-rat IgG, anddonkey anti-goat IgGwere purchased from Santa Cruz (Heidelberg, Germany),

Reprint requests to: Karin Mˇller-Decker, Deutsches Krebsforschungs-zentrum, Research ProgramTumor Cell Regulation, Im Neuenheimer Feld280, D-69120 Heidelberg, Germany. Email: K.Mueller-Decker@dkfz.deAbbreviations: COX, cyclooxygenase; NSAID, nonsteroidal anti-in-

£ammatory drug; PG, Prosta-glandin.

Manuscript received December 17, 2001; April 18, 2002; accepted forpublication July 29, 2002

0022-202X/02/$15.00 � Copyrightr 2002 by The Society for Investigative Dermatology, Inc.

1189

puri¢ed monoclonal rat anti-mouse CD31 (PECAM-1) antibody fromBecton Dickinson (Heidelberg, Germany), and rabbit anti-mouse Ki-67from Dianova (Hamburg, Germany). Enzyme-linked immunosorbentassay-bovine serum albumin, goat serum, and Masson’s trichrome stainingkit were purchased from Sigma, PGE2 enzyme immunoassay fromCayman (Ann Arbor, MI). Rompun and Ketanest were purchased fromBayer Vital (Leverkusen, Germany) and from Park Davis (Berlin,Germany), respectively, nylon sutures were from Resorba (Nˇrnberg,Germany).

Animals Twelve week old female NMRI mice (an outbred strain fromRCC, Fˇllinsdorf, Switzerland) were used. The animals were kept underan arti¢cial day^night rhythm and were fed diet (see below) and water adlibitum. All animal experiments were approved by the Regierungs-prLsidium Karlsruhe, Germany (¢le number 059/2000).

Feeding of animals Groups of animals were fed rodent diet 5010,diet 5010 supplemented with 500 p.p.m. valdecoxib or 64 p.p.m. SC-560,altromin diet (Lage, Germany), or altromin diet supplemented with 16p.p.m. indomethacin. The drug-supplemented diets were given 7 d beforesurgery and during the course of the study.

Analysis of blood plasma for COX-inhibitor levels Plasma samples(0.3 ml) were collected 7 d after beginning the treatment with valdecoxibdiet and SC-560 diet or control diet. To this end, plasma was collectedfrom ¢ve animals per group, frozen immediately in liquid nitrogen, andstored at �701C. Analysis of blood plasma levels was carried out asdescribed (Reddy et al, 2000).

Skin incisional wounding For histologic, immunohistochemical, andprostaglandin analysis three to six animals per group and for thedetermination of tensile strength and plasma levels of inhibitors ¢veanimals per group were anesthetized by consecutive intraperitonealinjections of Rompun and Ketanest (20 mg per kg and 100 mg per kgbody weight, respectively). The hair of the back skin was shaved withelectric clippers and skin was wiped once with 70% ethanol. In theshaved area a single transverse full-thickness incision, 2 cm in length, wasmade extending through epidermis and dermis to the panniculus carnosus.Incisions were sealed with three interrupted nylon sutures spaced 0.5 cmeach. After wounding, mice were separated into individual cages for 3 d.The wounds and general health of the animals were monitored daily forup to 4 wk postsurgery. No complications were encountered. One biopsycomprising the total wound area and 3 mm of nonwounded skin on eachside was collected from each animal, bisected in the center, and preparedfor histochemistry and immunohistochemistry at days 1, 2, 3, 7, 14, and 28after injury. In matched controls similar areas of skin were excised from thenonwounded back skin of each animal. For prostaglandin analysis woundtissue was frozen immediately in liquid nitrogen and stored at ^801C untilextraction.

Histochemistry and immunohistochemistry For hematoxylin/eosinand Masson’s trichrome stainings, skin was ¢xed in phosphate-bu¡eredsaline-bu¡ered 4% paraformaldehyde for 16 h before embedding intopara⁄n and 5 mm sections were then processed according to theinstructions of the supplier. For immunohistochemical stainingcryosections, 5 mm thick, were used to detect various antigens. COXisozymes were immunodetected as described (Mˇller-Decker et al, 1998a;Neufang et al, 2001). Ki-67, and CD31 immunostainings were performedaccording to the same protocol, except that 1% enzyme-linked immuno-sorbent assay^bovine serum albumin in phosphate-bu¡ered saline wasused as blocking solution and sections were ¢xed in acetone for 10 min.Primary antibodies were diluted 1 : 50 (Ki-67), 1 : 100 (COX-1 and -2), or1 : 200 (CD31), and peroxidase-conjugated secondary antibodies 1 : 100in blocking solution. Development with diaminobenzidine was used forvisualization. Sections were counterstained with hematoxylin (Merck,Darmstadt, Germany), dehydrated, and mounted with Eukitt (O. Kindler,Freiburg, Germany). Specimens were examined by an Axioskop 2microscope connected to an Axiocam camera using Axiovision software,release 2.05 (Zeiss, Goettingen, Germany).

Quantitation of vessel density Vessel density in the wound bed wasdetermined by scanning the section at low magni¢cation (�100), onevisual ¢eld each from the center to the left and right periphery(equivalent to 3 mm2). Microvessel counts in this area were performedat�160 magni¢cation. According to a published procedure (Dellas et al,1997), any endothelial cell cluster positive for CD31 and clearly separatefrom an adjacent cluster was considered as a single vessel. At least sixwound beds (12 visual ¢elds) from three animals were counted. Resultsare expressed as mean number of vessels7SD.

Determination of prostaglandin concentrations Immediately afterexcision, biopsies were frozen in liquid nitrogen and then homogenizedwith a dismembrator (Braun, Melsungen, Germany) in liquid nitrogenand suspended in 2 ml of ice-cold ethanol. After thorough mixing for30 s, the proteins were pelleted by centrifugation (10 min at 3000� g and41C). Prostaglandins were extracted from the supernatant by solid phaseextraction. The pellet was dissolved in 8 M urea and used for themeasurement of protein concentration. Prostaglandin levels werequantitated by means of a commercially available enzyme immunoassayas previously described (Mˇller-Decker et al, 1995).

Measurement of tensile strength Seven or 15 d after wounding ¢veanimals per group were killed and skin samples were immediatelyprepared for the measurement of tensile strength. Analysis was performedas previously described (Gorodetsky et al, 1988) with minor modi¢cations.Brie£y, a rectangular piece of skin (2� 5 cm) with the wound in the centerwas excised. The samples were spread on ¢lter papers and cut per-pendicularly across the wound into ¢ve 3 mm strips. These were kept(up to 10 min at the most) in a humidi¢ed chamber until measurement.For the determination of tensile strength the ¢lter paper was removedfrom the central 4 cm of the strip. The individual strip was mounted onto an Instron tensiometer model 5542 (Instron Corporation, Canton,MA), leaving the central 2 cm of skin as the bridge between the upperand lower jaws. Strain (g per area of skin)/stress (elongation) curves andthe breaking point of the samples were recorded. The area under the peakof each record represents tensile strength, i.e., the total energy needed tobreak the skin sample. The tensile strength of ¢ve strips per wound of¢ve animals each (i.e., 25 records) was determined.

Statistics Data represent mean7SD. Using Student’s t test pr0.05 wasregarded as signi¢cant.

RESULTS

The incisional wound model mimics authentic surgical proce-dures (Davidson, 1998). This model system was employed tostudy COX isozyme expression in response to wounding and toexamine the e¡ects of the COX-2 selective inhibitor valdecoxib(Talley et al, 2000) and the COX-1 selective inhibitor SC-560(Smith et al, 1998) on re-epithelialization, angiogenesis, and col-lagen deposition. The tensile strength of the healing woundswas determined in the presence of the COX isozyme-selectiveinhibitors and the nonselective COX inhibitor indomethacin(Ferreira et al, 1971). Full-thickness 2 cm transverse incisions ex-tending through epidermis and dermis to the panniculus carno-sus were made on the back of mice at the age of 13 wk, i.e., whenthe majority of hair follicles were in the resting phase. Animalswere fed with control diet or experimental diet containing 500p.p.m. valdecoxib, 64 p.p.m. SC-560, or 16 p.p.m. indomethacin.These doses corresponded to the anti-in£ammatory doses of SC-560 (Smith et al, 1998), valdecoxib (Talley et al, 2000), and indo-methacin (Masferrer et al, 1994). Inhibitor treatment was started 7d before wounding resulting in steady-state plasma levels7SD of91.0742.0 ng valdecoxib per ml and 18.579.2 ng SC-560 per mlat the time point of wounding. Feeding of diets was continueduntil collection of skin samples. The PGE2 content in thewounded skin area was found to be strongly elevated at days 2and 7 after surgery as compared with unwounded matched con-trol tissue (Table I). Upon treatment with valdecoxib, SC-560, orindomethacin wound-induced PGE2 levels were reduced by75%, 50%, and 72% at day 2 and by 52%, 95%, and 93% atday 7 after surgery indicating that both COX-2 and COX-1 con-tributed to wound-induced prostaglandin synthesis albeit withvarying ratios in the course of healing. The e¡ects of the COXisozyme-selective inhibitors on the gross appearance and progressof wound healing were followed up by visual inspection and his-tologic and immunohistochemical analysis on days 1, 2, 3, 7, 14,and 28 after surgery.

Gross appearance and histology of incisional skin woundswere not a¡ected upon treatment of mice with COXisozyme-selective inhibitors Wounding-induced bleedingtime was similar in all groups. Irrespective of the treatment the

1190 MPLLER-DECKER ETAL THE JOURNAL OF INVESTIGATIVE DERMATOLOGY

tissue adjacent to the wound edges was swollen to a similar extentat day 1 after wounding and sealing of the wounds by adehydrated crust became visible at day 2. The thin scab wasprogressively lost and not retained longer than 7 d, except forone wound of three in the group treated with valdecoxib.Closure of the wounds at day 7 (Fig 1) and hair regrowthcovering the wounds at day 14, were similar in mice of alltreatment groups. Four weeks after surgery the woundsappeared to be fully healed in all mice at least upon visualinspection (Fig 1).For histologic analysis wound tissues were collected from three

to six animals per group. In both inhibitor-treated and controlanimals epidermal hyperplasia at the wound site was apparent1 d after injury and persisted for about 2 wk. By day 3 allwounds had re-epithelialized, but were still covered with a scab,which had dropped o¡ in all but one wound at day 7 (Fig 4). Atthat time the newly formed epidermis of control mice lookedmature, i.e., had developed a horny layer, whereas in wounds ofvaldecoxib- and SC-560-treated animals this maturation processseemed to be delayed (Fig 2D,F; see also Fig 4A^C). By day 14,however, new epidermis of inhibitor-treated animals wasindistinguishable from controls (Fig 5A^C). At day 28 afterinjury epidermis had re-established its normal thickness of onebasal cell layer and up to three suprabasal cell layers, includingstratum corneum independent of inhibitor treatment (Fig 3B,F).Determination of the proliferation index by countinginterfollicular Ki-67-immunopositive keratinocytes in thewound area, i.e., the wound edges plus immediately adjacent

¢elds (Fig 2G^L), was carried out 1, 2, 3, 7, 14, and 28 d afterwounding (Fig 2M). A representative immuno-histochemicalstaining for Ki-67 is shown for day 2, i.e., the time point ofmaximal proliferation (Fig 2G^L). Although slight di¡erencesof the proliferative indices of wound epithelium betweencontrol and inhibitor-treated animals were observed at this timepoint (and in the COX-1 inhibitor group at day 3) the overallkinetics of Ki-67 expression was similar in all treatment groups(Fig 2M).

Valdecoxib and SC-560 did not alter the wound-inducedCOX isozyme expression Con¢rming earlier observations(Mˇller-Decker et al, 1998b), COX-1 protein was found to beexpressed in individual keratinocytes as well as in dendritic cellsthat morphologically resemble Langerhans cells of normalepidermis. In control wounds COX-1 protein was located pre-dominantly in the suprabasal compartment of the hyperplasticepithelium as well as dendritic cells (Fig 3A). This pattern ofexpression persisted during the ¢rst 2 wk of follow up. At day28 after wounding the pattern of COX-1 expression was againsimilar as in unwounded skin (Fig 3B).COX-2 protein was barely detectable in keratinocytes of

normal interfollicular mouse epidermis, but prominent in outerroot sheath cells of anagen hair follicles, whereas outer rootsheath cells of telogen hair follicles were COX-2-negative (datanot shown). In wounded epidermis, however, COX-2 pro-tein was found in keratinocytes throughout the hyperplasticepithelium and in dendritic cells (Fig 3E). This pattern persistedduring the whole follow up, although expression decreasedtowards day 28 after surgery when the morphology ofepidermis looked normal again. In addition, COX-2 protein wasfound in the outer root sheath cells of anagen hair follicles and inbasal cells of associated sebaceous glands (Fig 3F). Furthermore,both COX isozymes were found in endothelial cells of largevessels as well as in macrophages in the vicinity of or within thewound bed (Fig 3C^H). No changes in COX isozymeexpression were found in the presence of either the COX-1 orthe COX-2 inhibitor (data not shown).

COX isozyme-selective inhibition did not impair wound-induced neovascularization Sproating of new capillariestowards the wound space was examined by immuno-histochemical staining of endothelial cells with CD31 antibody,representatively shown in Fig 4(A^Cn) for specimens collected7 d after wounding. In wounds of control mice blood vesseldensity as assessed by counting CD31-microvessel clusters, wasmaximal at days 7^14 and back to basal values at day 28 aftersurgery (Fig 4D). Similar values were recorded for inhibitor-treated mice, although vessel density was slightly increased inwounds of valdecoxib-treated animals at day 7 (Student’s t test,po0.05).

Collagen deposition was not a¡ected by COX isozyme-selective inhibition In a ¢rst step, collagen deposition withinthe wound bed was investigated by the Masson’s trichromeprocedure, which stains collagen ¢bers blue, cytoplasm red, andnuclei dark blue. Staining of wound specimens revealed tiny¢bers of collagen, the majority of which were arranged in aparallel fashion throughout the wound bed 7 d after wounding(data not shown). One week later collagen density had increasedand ¢bers were reorganized towards a net-like structure (Fig 5A^C,An^Cn) but were still di¡erent from the basket-weaveorganization of dermal collagen bundles in unwounded dermis.According to microscopic inspection, collagen deposition andcell density within the granulation tissue were not a¡ected byCOX inhibitor treatment (Fig 5An^Cn).Restoration of the tensile strength was not impeded by

COX isozyme-selective and nonselective inhibitors. Tensilestrength of wounded skin re£ecting regain of epidermal strength,¢brillar collagen deposition, as well as remodeling during thecourse of wound healing was measured by means of a tensiometer.

Table I. Inhibition of PGE2 levels in incisional wounds byvaldecoxib, SC-560, and indomethacin treatment 2 d and 7 d

postwounding

PGE2 (pg per mg protein)7 SEM

Treatment 2 d 7 d

Control diet (unwounded) 18778 81715Control diet (wounded) 2085079385 3997135Valdecoxib (wounded) 54757118 192727SC-560 (wounded) 113837774 2176Indomethacin (wounded)a 9787122 149789

aTwo and 7 d after wounding control values were 35487546 and 20987955,respectively, for the indomethacin experiment.Wounded and unwounded control tissue (15mm2) from two to four animals

per group was snap-frozen, pulverized, and PGE2 levels were determined by anenzyme immunoassay as described previously (Mˇller-Decker et al, 1995).

Figure1. Gross appearance and progress of wound healing on backskin of mice fed control diet or diet supplemented with 500 p.p.m.valdecoxib or 64 p.p.m. SC-560. Photographs were taken 7 and 28 dafter full thickness skin incisions were made.

COX ANDWOUND HEALING IN SKIN 1191VOL. 119, NO. 5 NOVEMBER 2002

Samples were collected 7 and 15 d after surgery. Tensile strengthof unwounded skin was found to be 2.7^3.5 N. In animals fed thecontrol diet the wounds had regained about 15% and 50% oforiginal tensile strength within 7 and 15 d, respectively. Similarvalues were recorded for valdecoxib- and indomethacin-treatedanimals, whereas a slight transient increase of tensile strengthwas measured upon SC-560 treatment (Fig 5D).

DISCUSSION

Using an established model (Davidson, 1998) we investigated thee¡ect of acute incisional wounding on COX isozyme expression

and the e¡ect of COX inhibitors on the wound response inmouse skin.COX-1 was constitutively expressed to about the same extent

in normal and wounded epidermis but its cellular localizationwas slightly di¡erent in the latter: whereas being detectable inindividual keratinocytes and dendritic cells throughout thenormal epithelium (Mˇller-Decker et al, 1998b), COX-1 proteinlocated predominantly to suprabasal keratinocytes of thehyperplastic wound epithelium. This alteration persisted for atleast 2 wk. COX-2 being barely detectable in interfollicular epi-dermis of unwounded skin (Mˇller-Decker et al, 1998b) wasfound to be rapidly induced upon wounding and becamestrongly expressed in keratinocytes and dendritic cells

Figure 2. Kinetics of re-epithelialization and proliferation in full-thickness incisional skin wounds.Mice were fed control diet (A,B,G,H) or dietcontaining 500 p.p.m.valdecoxib (C,D,I,J) or 64 p.p.m.SC-560 (E,F,K,L) and analyzed 2 d (G^L), 3 d (A,C,E), and 7 d (B,D,F) after surgery. Specimens werestained with anti-COX-2 antibodies (A^F) or anti-Ki-67 antibodies (G^L). The wound edges are marked with arrows (G,I,K). Epidermis (e), granulationtissue (g), and eschar (es). Original magni¢cation: � 630 (A^F, insert H), � 160 (G^L). (M) Kinetics of proliferation in interfollicular wound epidermis.Ki-67-immunopositive cells were counted in the wound center and the immediate neighboring ¢elds.Values represent mean7SD (nnp o 0.005, nnnp o0.0005) according to Student’s t test, n¼ 6 sections or 24 microscopic ¢elds (original magni¢cation�160)/group and time point.

1192 MPLLER-DECKER ETAL THE JOURNAL OF INVESTIGATIVE DERMATOLOGY

throughout the hyperplastic epithelium, in endothelial cells oflarge vessels, and in macrophages within the wound bed. SomeCOX-2 expressing keratinocytes were still detectable 28 d afterwounding, whereas the COX-1 expression pattern and epidermalmorphology looked normal at this time-point.Similar observations have been reported for other epithelia.Thus,

in gastric epithelium of mice injured by acetic acid (Mizuno et al,1997) and in trinitrobenzene sulfonic acid-induced colitis in rats(Reuter et al, 1996) COX-1 expression remained essentially unaf-fected, but disappeared in the mucosa adjacent to the ulcer craterin chronic gastric ulceration of rat stomach (Schmassmann et al,1998). On the other hand, COX-2 was rapidly induced uponinjury in the stomach of mice and rats (Mizuno et al, 1997; Gretzeret al, 1998) or upon induction of colitis in rat colon (Reuter et al,1996). Cellular localization of COX-2 in gastric mucosa and

colon, however, was di¡erent from that found in skin wounds:epithelial cells were the major COX-2-positive cell type in skin,whereas in injured gastric mucosa (Schmassmann et al, 1998) andcolonic epithelium (Reuter et al, 1996), COX-2 expression waspredominant in nonepithelial cells, including macrophages, en-dothelial cells, and ¢broblasts. Thus, the induction of COX-2protein appears to be a common feature of acute and chronicepithelial wounds with the expression persisting during thehealing process.

Figure 3. Localization of COX isozymes in incisional wounds byimmunohistochemistry with goat anti-COX-1 (A^D) and goatanti-COX-2 (E^H) antibodies. Wound specimens were collected fromcontrol animals at various time points: 24 h (A,E,C,G), 7 d (D,H), and 28d (B,F). Note COX-1 expression in suprabasal cells of interfollicular epider-mis (A), dendritic cells (A, insert), endothelial cells (C), and macrophages(D). Note COX-2 expression in basal as well as suprabasal interfollicularkeratinocytes (E), the outer root sheath of hair follicles and sebaceous glandepithelium (F), dendritic cells (E, insert), endothelial cells (G), as well asmacrophages (H). Original magni¢cation:� 630.

Figure 4. Localization of CD31-positive vessels in incisionalwounds from animals fed control diet or diet supplemented withvaldecoxib or SC-560.Wound specimens were collected 7 d postsurgeryfrom animals fed control diet (A,An), valdecoxib diet (B,Bn), or SC-560diet (C,Cn), and processed for immunohistochemistry using rat anti-CD31 antibody. Original magni¢cation: �100 (A^C), � 630 (An^Cn).(D) Vessel density as function of time after incision. Microvessel densitywas determined in the wound centers as shown under A^C by countingCD31-immunopositive endothelial clusters. Values represent mean 7 SD(nnpo 0.05), according to Student’s t test, n ¼ 6 sections or 12 microscopic¢elds/group and time point.

COX ANDWOUND HEALING IN SKIN 1193VOL. 119, NO. 5 NOVEMBER 2002

Wounding led to a strong increase of the prostaglandin levelin skin. This response was strongly inhibited by bothCOX-isozyme-selective inhibitors and the nonselective inhibitorindomethacin, indicating that both isozymes contributed towound-induced prostaglandin synthesis albeit with varying ratiosduring the course of healing. A similar suppression of prostaglan-din synthesis has also been reported for colonic and gastric injuryupon treatment with nonselective COX-inhibitors, such as indo-methacin or diclofenac, the selective COX-1 inhibitor SC-560, orCOX-2 inhibitors, including NS398, celecoxib, and L745337(Reuter et al, 1996; Lesch et al, 1998; Gretzer et al, 2001).

Notwithstanding the similarities of epithelial COX isozymeexpression and activity in the various wound models opposite ef-fects of COX-2 inhibitors on healing were observed. Regardingdermal aspects of the healing process in this study the kinetics ofcollagen deposition and the restoration of the mean tensilestrength were similar in incisional skin wounds of normal andvaldecoxib-treated animals. Wound-induced neovascularizationwas transiently increased rather than inhibited by valdecoxib.This result suggests a di¡erent sensitivity of wound- and tumor-associated angiogenesis towards COX-2 inhibitors, which havebeen shown to suppress the latter (Masferrer et al, 2000; Leahyet al, 2002). Concerning epidermis a slightly decreased prolifera-tive index early after wounding and a somewhat delayed matura-tion indicate a transient impairment of wound epithelializationby COX-2 inhibition.Similarly, the COX-2-selective inhibitors PD138387 and cele-

coxib did not impair the healing of acute acetic acid-induced gas-tric ulcer in rats when applied in anti-in£ammatory doses (Leschet al, 1998). Moreover, neither celecoxib nor valdecoxib did impairwound healing as monitored by collagen deposition into subcu-taneously implanted sponges in rats (Muscara et al, 2000). In con-trast, two other COX-2-selective inhibitors, i.e., NS398 (Mizunoet al, 1997) and L745337 (Reuter et al, 1996; Schmassmann et al,1998) were shown to delay ulcer healing in acute and chronic gas-tric mucosa and colonic epithelium of rats when given in anti-in£ammatory doses. In chronic gastric ulcers this delay correlatedwith a suppression of PGE2 levels, decreased epithelial cell prolif-eration in the ulcer margin, and reduced microvessel density inthe ulcer bed (Schmassmann et al, 1998). PGE2 levels and angio-genesis in carrageenin-induced granulation tissue have also beenfound to be suppressed by NS398 (Ghosh et al, 2000). These re-sults gave rise to the hypothesis that the apparently contradictorye¡ects of NS398 and L745337 on wound healing may be relatedto other e¡ects than inhibition of COX-2 activity (Lesch et al,1998).On the other hand, selective inhibition of COX-1 has been

shown to exacerbate injury in acid-challenged gastric mucosa(Gretzer et al, 2001) but did not a¡ect (except for a slight impair-ment of wound epithelialization) neovascularization, collagen de-position, and the breaking strength of acute incisional wounds inmouse skin as shown here. The simultaneous inhibition of bothCOX isozymes has also been shown to aggravate gastric injury inrat mucosa (Wallace et al, 2000; Gretzer et al, 2001) and to impairhealing of ischaemic wounds in rat skin (Quirinia and Viidik,1997). Again such an e¡ect was not observed with acute incisionalwounds in rat (Haws et al, 1996) and not in mouse skin as shownhere. Provided that these animal data can be extrapolated to hu-mans, postoperative administration of COX inhibitors are notexpected to impair healing of acute incisional wounds that arethe major lesion generated in surgery. Whether or not selectiveCOX inhibitors a¡ect chronic venous leg ulcers where macro-phages and endothelial cells are the main cellular sources of bothCOX-1 and COX-2 remains to be clari¢ed (Abd-El-Aleem et al,2001).

The authors thank Dagmar Kucher, Brigitte Steinbauer, Andrea Pohl-Arnold, SandraPfrang, and Ulrike Beckhaus for excellent assistance with the animals experiments andthe Instron-Wolpert GmbH, Darmstadt, Germany for technical support with the de-termination of the tensile strength. Financial support by Searle/Pharmacia, St Louis,MO is gratefully acknowledged.

REFERENCES

Abd-El-Aleem SA, Ferguson MWJ, Appleton I, Bhowmick A, McCollum CN,Ireland GW: Expression of cyclooxygenase isoforms in normal human skinand chronic venous ulcers. J Pathol 195:616^623, 2001

Chan CC, Boyce S, Brideau C, et al: Pharmacology of a selective cyclooxygenase-2inhibitor, L-745,337: a novel anti-in£ammatory agent with an ulcerogenicsparing e¡ect in rat and nonhuman primate stomach. J Pharmacol Exp Ther274:1531^1537, 1995

Figure 5. Collagen deposition in incisional wounds and tensilestrength after treatment with COX inhibitors.Wound specimens col-lected 14 d postsurgery from animals fed control diet (A,An), valdecoxibdiet (B,Bn), or SC-560 diet (C,Cn) were stained using the Masson’strichrome procedure. Original magni¢cation:�100 (A^C),�630 (An^Cn).Measurement of tensile strength of wounded skin was done by means ofa tensiometer (D). Values represent mean7SD (n¼ 25 stripes with ¢veanimals per group and time point). Indo, indomethacin.

1194 MPLLER-DECKER ETAL THE JOURNAL OF INVESTIGATIVE DERMATOLOGY

Davidson JM: Animal models of wound repair. Arch Dermatol Res 290(Suppl.): S1^S11, 1998

Dellas A, Moch H, Schultheiss E, Feichter G, Almendral AC, Gudat F, Torhost J:Angiogenesis in cervical neoplasia. Microvessel quantitation in precancerouslesions and invasive carcinomas with clinicopathological correlations. GynecolOncol 67:27^33, 1997

Ferreira SH, Moncada S, Vane JR: Indomethacin and aspirin abolish prostaglandinrelease from the spleen. Nat New Biol 231:237^239, 1971

Ghosh AK, Hirasawa N, Niki H, Ohuchi K: Cyclooxygenase-2-mediated angio-genesis in carrageenin-induced granulation tissue in rats. Pharmacol Exp Ther295:802^809, 2000

Gorodetsky R, McBrideWH,Withers HR: Assay of radiation e¡ects in mouse skinas expressed in wound healing. Radiat Res 116:135^144, 1988

Gretzer B, Ehrlich K, Maricic N, Lambrecht N, Respondek M, Peskar BM: Selectivecyclooxygenase-2 inhibitors and their in£uence on the protective e¡ect of amild irritant in the rat stomach. Br J Pharmacol 123:927^935, 1998

Gretzer B, Maricic N, Respondek M, Schuligoi R, Peskar BM: E¡ects of speci¢cinhibition of cyclo-oxygenase-1 and cyclo-oxygenase-2 in rat stomach withnormal mucosa and after acid challenge. Br J Pharmacol 132:1561^1573, 2001

Haws MJ, Kucan JO, Roth AC, Suchy H, Brown RE: The e¡ects of chronicketorolac tromethamine (toradol) on wound healing. Ann Plast Surg 37:147^151,1996

KishimotoY,Wada K, Nakamoto K, Kawasaki H, Hasegawa J: Levels of cyclooxy-genase-1 and ^2 mRNA expression at various stages of acute gastric injuryinduced by ischemia-reperfusion in rats. Arch Biochem Biophys 352:154^157, 1998

Laine L, Harper S, Simon T, et al: A randomized trial comparing the e¡ect of rofe-coxib, a cyclooxygenase 2-speci¢c inhibitor, with that of ibuprofen on the gas-troduodenal mucosa of patients wih osteoarthritis. Gastroenterology 117:776^683,1999

Leahy KM, Ornberg RL,Wang A, Koki AT, Masferrer JL: Cyclooxygenase inhibi-tion by celecoxib reduces proliferation and induces apoptosis in angiogenicendothelial cells in vivo. Cancer Res 62:625^631, 2002

Lesch CA, Gilbertsen RB, Song Y, Dyer RD, Sanchez B, Guglietta A: E¡ects ofnovel anti-in£ammatory compounds on healing of acetic acid-induced gastriculcer in rats. J Pharmacol ExpTher 287:301^306, 1998

Luong C, Miller A, Barnett J, Chow J, Ramesha C, Browner MF: Flexibility of theNSAID binding site in the structure of human cyclooxygenase-2. Nat StructBiol 3:927^933, 1996

Masferrer JL, Zweifel BS, Manning PT, et al: Selective inhibition of inducible cy-clooxygenase 2 in vivo is anti-in£ammatory and nonulcerogenic. Proc Natl AcadSci USA 91:3228^3232, 1994

Masferrer JL, Leahy KM, Koki AT, et al: Antiangiogenic and antitumor activities ofcyclooxygenase-2 inhibitors. Cancer Res 60:1306^1311, 2000

Mizuno H, Sakamoto C, Matsuda K, et al: Induction of cyclooxygenase 2 in gastricmucosal lesions and its inhibition by the speci¢c antagonist delays healing inmice. Gastroenterology 112:387^397, 1997

Mˇller-Decker K, Scholz K, Marks F, Fˇrstenberger G: Di¡erential expression ofprostaglandin H synthase isoenzymes during multistage carcinogenesis inmouse epidermis. Molec Carcinogenesis 12:31^41, 1995

Mˇller-Decker K, Kopp-Schneider A, Marks F, Seibert K, Fˇrstenberger G: Locali-zation of prostaglandin H synthase isozymes in murine epidermal tumors:suppression of skin tumor promotion by PGHS-2 inhibition. Mol Carcinogen-esis 23:36^44, 1998a

Mˇller-Decker K, Scholz K, Neufang G, Marks F, Fˇrstenberger G: Localization ofprostaglandin-H synthase-1 and -2 in mouse skin: implications for cutaneousfunction. Exp Cell Res 242:84^91, 1998b

Muscara MN, McKnightW, Asfaha S,Wallace JL:Wound collagen deposition in rats.E¡ects of an NO-NSAID and a selective COX-2 inhibitor. Br J Pharmacol129:681^686, 2000

Neufang G, Fˇrstenberger G, Heidt M, Marks F, Mˇller-Decker K: Abnormal di¡er-entiation of epidermis in transgenic mice constitutively expressing cycloooxy-genase-2 in skin. Proc Natl Acad Sci USA 98:7629^7634, 2001

Picot D, Loll PJ, Garavito RM:The X-ray crystal structure of the membrane proteinprostaglandin H2 synthase-1. Nature 367:243^249, 1994

Quirinia A,Viidik A: Diclofenac and indomethacin in£uence the healing of normaland ischaemic incisional wounds in skin. Scand J Plast Reconstr Hand Surg31:213^219, 1997

Reddy BS, Hirose Y, Lubet R, et al: Chemoprevention of colon cancer by speci¢ccyclooxygenase-2 inhibitor, celecoxib administered during di¡erent stages ofcarcinogenesis. Cancer Res 60:293^297, 2000

Reuter BK, Asfaha S, Buret A, Sharkey KA,Wallace JL: Exacerbation of in£amma-tion-associated colonic injury in rat through inhibition of cyclooxygenase-2.J Clin Invest 98:2076^2085, 1996

Schmassmann A, Peskar BM, Stettler C, Netzer P, Stro¡ T, Flogerzi B, Halter F:E¡ects of inhibition of prostaglandin endoperoxide synthase-2 in chronic gas-tro-intestinal ulcer models in rats. Br J Pharmacol 123:795^804, 1998

Scholz K, Fˇrstenberger G, Mˇller-Decker K, Marks F: Di¡erential expression orprostaglandin H-synthase isoenzymes in normal and activated keratinocytesin vivo and in vitro. Biochem J 306:263^269, 1995

Seibert K, ZhangY, Leahy K, et al: Pharmacological and biochemical demonstrationof the role of cyclooxygenase 2 in in£ammation and pain. Proc Natl Acad SciUSA 91:12013^12017, 1994

Simon LS,Weaver AL, Graham DY, et al: Anti-in£ammatory and upper gastrointest-inal e¡ects of celecoxib in rheumatoid arthritis A randomized controlled trial.JAMA 282:1921^1928, 1999

Smith CJ, ZhangY, Koboldt CM, et al: Pharmacological analysis of cyclooxygenase-1 in in£ammation. Proc Natl Acad Sci USA 95:13313^13318, 1998

Smith WL, DeWitt DL, Garavito RM: Cyclooxygenases: structural, cellular, andmolecular biology. Annu Rev Biochem 69:145^182, 2000

Soll AH, Weinstein WM, Kurata J, McCarthy D: Nonsteroidal anti-in£ammatorydrugs and peptic ulcer disease. Ann Intern Med 114:307^319, 1991

Talley JJ: Selective inhibitors of cyclooxygenase-2 (COX-). Prog Med Chem 36:201^234, 1999

Talley JJ, Brown DL, Carter JS, et al: 4 -[5-Methyl-3-phenylisoxazol-4 -yl]-benzene-sulfonamide, valdecoxib a potent and selective inhibitor of COX-2. J MedChem 43:775^777, 2000

Tessner TG, Cohn SM, Schloemann S, Stenson WF: Prostaglandins prevent de-creased epithelial cell proliferation associated with dextran sodium sulfateinjury in mice. Gastroenterology 155:874^882, 1998

Vane JR: Anti-in£ammatory drugs and their mechanism of action. In£amm Res47:78^S87, 1998

Wallace JL, McKnightW, Reuter BK,Vergnolle N: NSAID-induced gastric damagein rats: requirement of inhibition of both cyclooxygenase 1 and 2. Gastroenterology119:706^714, 2000

Williams CS, Mann M, DuBois RN: The role of cyclooxygenases in in£ammation,cancer, and development. Oncogene 18:7908^7916, 1999

COX ANDWOUND HEALING IN SKIN 1195VOL. 119, NO. 5 NOVEMBER 2002