tissue engineering for in vitro analysis of matrix metalloproteinases in the pathogenesis of keloid...

Tissue Engineering for In Vitro Analysis of MatrixMetalloproteinases in the Pathogenesis of Keloid LesionsHanwei Li, PhD; Zayna Nahas, MD; Felicia Feng, BS; Jennifer H. Elisseeff, PhD; Kofi Boahene, MD

K eloids represent an extreme form of abnormal cutane-ous wound healing that can result in both function-ally restrictive and disfiguring scars. In normal wound

healing, a series of physiologic responses results in the forma-tion of a scar that is randomly disorganized and predomi-nantly composed of collagen deposition.1,2 Remodeling of thenewly formed tissue affects the type of scar that forms. Ke-loid lesions are a form of abnormal wound healing whose mo-lecular mechanism and pathogenesis are not well under-stood, thus making it a therapeutic challenge. In keloids,aberrant remodeling is thought to result in large quantities ofcollagen deposition after skin trauma in predisposed individu-

als. However, what distinguishes keloids is not only the ab-normal quantity of collagen formation but also the physical ex-tent of it. Clinically, keloids differ from hypertrophic scars inthat they extend beyond the boundaries of the initial injuryby invading surrounding healthy skin at the level of the der-mis and fail to regress over time.3 An increase in inflamma-tory markers, such as transforming growth factor β1, and el-evated levels of extracellular matrix (ECM) components, suchas fibronectin and certain proteoglycans, are also associatedwith the formation of keloid lesions.4-9

Currently available treatments for keloid lesions includecombinations of steroid injections, surgical removal, silicone

IMPORTANCE Keloid lesions form because of alterations in the mechanisms that governcutaneous wound healing. Although matrix metalloproteinases (MMPs) have been implicatedin keloid pathophysiology, many questions still remain about their involvement. Ourincomplete understanding of keloid pathophysiology has led to high recurrence rates incurrent treatments. No reliable animal model is available for studying keloids.

OBJECTIVE To gain a better understanding of the disease mechanisms involved in keloidlesions in the hopes of identifying therapeutic options.

DESIGN Fibroblasts derived from keloid tissue were incorporated in either Matrigel orpolyethylene glycol diacrylate mixed with type I collagen to create 3-dimensional models toinvestigate the role MMPs play in keloid formation. The MMP gene expressions were alsocompared between fibroblasts isolated from different sites within the same keloid lesion.

SETTING The Johns Hopkins School of Medicine, Baltimore, Maryland.

PARTICIPANTS Keloid fibroblasts were received from the Baylor College of Medicine, andadditional keloid fibroblasts were enzymatically isolated from the dermal layer of lesionsremoved from consenting patients at The Johns Hopkins Hospital.

RESULTS In the Matrigel system, MMP9 and MMP13 were observed to be significantlyupregulated in keloid fibroblasts. The addition of decorin resulted in a significant decrease oftype I collagen and MMP1, MMP9, and MMP13 gene expressions from keloid fibroblasts.Higher MMP gene expressions were observed in fibroblasts isolated from the margins of theoriginal keloid wound.

CONCLUSIONS AND RELEVANCE MMP9 and MMP13 are expressed significantly more inkeloid-derived cells, thus making them 2 potential targets for disease modification. Moleculesthat target organization of the lesion’s matrix can be beneficial in downregulating increasedmarkers during the disease. In addition, heterogeneity is observed with the varying MMPgene expressions from site-specific fibroblasts within the same keloid lesion.

JAMA Facial Plast Surg. doi:10.1001/jamafacial.2013.1211Published online September 19, 2013.

Author Affiliations: Department ofBiomedical Engineering, The JohnsHopkins University, Baltimore,Maryland (Li, Nahas, Feng, Elisseeff);Department of Otolaryngology–Headand Neck Surgery, The Johns HopkinsUniversity, Baltimore, Maryland(Boahene).

Corresponding Author: KofiBoahene, MD, Department ofOtolaryngology–Head and NeckSurgery, The Johns HopkinsUniversity, Baltimore, MD 21231([email protected]).

Research

Original Investigation

jamafacialplasticsurgery.com JAMA Facial Plastic Surgery Published online September 19, 2013 E1

Downloaded From: http://archfaci.jamanetwork.com/ by a University of Virginia User on 09/24/2013

gels meant to hydrate keratinocytes with the aim of alteringgrowth factor secretion, and radiation therapy.4,10 Althoughsurgical removal of keloids can offer temporary cosmetic im-provement, the surgical resection itself is often a trigger for theaberrant wound healing, thus risking more keloid formation,sometimes even extending into previously unaffected skin. Inaddition, steroid injections and radiation therapy are not freeof potential adverse effects and are often also associated withrecurrences.

Various hypotheses have been proposed to explain thepathogenesis of this disease, including abnormal regulationfrom the surrounding ECM proteins and irregularity in the en-zymes that monitor ECM degradation and cellular migration.4

Matrix metalloproteinases (MMPs) are composed of a familyof enzymes that are hypothesized to be involved in the keloiddisease state because of their abnormal expression and activ-ity, thus severely altering the remodeling process.11 Becauseof their important role in regulating connective tissue growthand wound healing, MMPs are considered potential targets inthe treatment of keloids. However, studies so far have drawnconflicting conclusions regarding the upregulation or down-regulation of these remodeling proteases in keloid forma-tion. Common to all of these studies is the use of a monolayercell culture. We believe that given the multidimensional na-ture of keloid physiology, a 3-dimensional cell culture systemwould better mimic the true environment of keloid forma-tion and that, in fact, the use of monolayer cell culture maybe contributing to this variation.11-13 Therefore, in this re-search, we applied tissue engineering strategies to create 3-di-mensional in vitro disease models to analyze the changes inMMP gene expressions between normal and keloid fibro-blasts. Two biomaterial systems were implemented for themodels, and through one of the models, the therapeutic ef-fects of an ECM modulator on the diseased fibroblast were alsoinvestigated, setting the precedence for future keloid studieswith different ECM modulators. Lastly, the differences in re-modeling markers from site-specific keloid fibroblasts were alsoanalyzed to address the heterogeneity observed in the phe-notypes of fibroblasts from keloid lesions.

MethodsKeloid FibroblastsKeloid fibroblasts were provided by Bindi Naik, MD, of BaylorCollege of Medicine and one of us (K. B.). Briefly, the isolationmethod involved rinsing keloid lesions thoroughly with phos-phate-buffered saline solution containing penicillin, 100 U/mL,and streptomycin, 100 μg/mL (PBS-PS), on receipt. The epi-dermal layer was removed, whereas the dermis layer wasminced before being placed in the collagenase digestion so-lution. The solution was composed of high-glucose Dulbeccomodified Eagle medium (Invitrogen) with 10% fetal bovine se-rum (Hyclone), with a final concentration of 5 mg/mL of type2 collagenase (Worthington) and 0.2 mg/mL of trypsin (Fluka).The minced explants were digested at 37°C with 5% carbon di-oxide for 6 hours on an orbital shaker. The collagenase filtratewas then filtered with a 40-μm strainer and rinsed with PBS-PS

3 times before being plated in T-175 flasks with fibroblast me-dia (FM). For encapsulation in biomaterials, all keloid fibro-blasts were cultured in monolayer for 1 passage until conflu-ency with FM. For lesion site–specific comparison, nomonolayer expansion of keloid fibroblasts was performed.

Normal Skin FibroblastsAs a control, normal human foreskin fibroblasts (HS27) wereobtained from American Type Culture Collection and cul-tured in monolayer between 1 and 2 passages until conflu-ency in FM before encapsulation in biomaterials. For lesionsite–specific comparison, normal fibroblasts were isolated froma female patient who underwent an elective upper eyelid skinexcision (blepharoplasty).

Medium ConditionThe FM consisted of high-glucose Dulbecco modified Eagle me-dium (Invitrogen) with 10% fetal bovine serum and penicil-lin, 100 U/mL, and streptomycin, 100 μg/mL.

PEG-Col1 In Vitro ExperimentThe first in vitro design used polyethylene glycol diacrylate(PEGDA) and type I bovine collagen (PEG-ColI) as the poly-mer to encapsulate the fibroblasts (Figure 1A). Type I colla-gen was neutralized with 0.1N sodium hydroxide and thenbrought up to 2 mg/mL with PBS-PS. A 20% PEGDA solutionin PBS-PS was mixed with the 2 mg/mL of type I collagen (BDBioscience) to create constructs with the final concentrationsof 10% PEGDA and 1 mg/mL of type I collagen. Fibroblasts,either keloid or normal (HS27), were mixed with the polymersolution at the concentration of 20 million cells/mL and thenmixed with the photoinitiation Irgacure 2959 (Ciba) for a finalconcentration of 0.05% (wt/vol). Using UV light, the con-structs of polymer-cell-photoinitiator were photopolymer-ized for 5 minutes at 3 mW/cm2 and a 365-nm wavelength incylindrical molds with a diameter of 5 mm. On encapsula-tion, the constructs were transferred to 24-well plates and cul-tured with 1.5 mL of FM for a total of 14 days. Constructs wereharvested at days 2, 7, and 14 for data analysis. Medium waschanged every 2 to 3 days until the time of harvest.

Matrigel In Vitro ExperimentThe second in vitro model involved encapsulating fibroblastsin Matrigel (BD Bioscience). Cell density was 2 million cells per100 μL of Matrigel, which was then incubated at 37°C for 30 min-utes. Once gelation was achieved, the constructs were cul-tured in 24-well plates with 1.5 mL of FM per construct for a totalof 21 days. Constructs were harvested at days 1, 7, and 21 for dataanalysis. In addition to the Matrigel constructs composed of ke-loid fibroblasts or HS27, a third group of constructs with encap-sulated keloid fibroblasts were cultured with the addition ofdecorin, 5 μg/mL (Sigma), which was administered with everymedium change, starting after 1 day of culture. Medium waschanged every 2 to 3 days until the time of harvest.

Biochemical AnalysesBiochemical data included DNA content, proteoglycan con-tent, and total collagen content. Harvested constructs were

Research Original Investigation Tissue Engineering for Keloid Lesions

E2 JAMA Facial Plastic Surgery Published online September 19, 2013 jamafacialplasticsurgery.com


lyophilized for 48 hours and then measured for their dryweights. In preparation for all 3 assays, lyophilized con-structs were homogenized in papain digestion buffer fromWorthington Biomedical and then incubated at 60°C for 16to 18 hours. The DNA quantity for each construct was deter-mined by measuring fluorescence with the low-assayHoechst 33258 dye from Molecular Probes. Briefly, the dyewas mixed with 10mM Tris, 1mM EDTA, and 0.2M sodiumchloride buffer at pH 7.4 to reach a final concentration of 0.1μg/mL. Varying concentrations of the calf thymus DNA (In-vitrogen) were mixed with the low-assay solution to gener-ate the standard curve. Total DNA quantities for each con-struct were measured from mixing the low-assay solutionwith 5 μL of the papain-digested samples. Fluorescenceintensities were measured on a fluorometer at 365-nm exci-tation and 458-nm emission.

The glycosaminoglycan (GAG) content was determinedthrough the dimethylmethylene blue assay, which consistedof the dimethylmethylene blue dye mixed with standard andexperimental samples. Varying concentrations of chondroi-tin sulfate C were used to generate the standard curve, whereas50 μL of papain-digested samples was used to calculate the totalGAG content per construct. The absorbance was measured at525 nm on an UV-Vis spectrophotometer.

The hydroxyproline assay was used to measure total col-lagen content. Papain-digested samples were hydrolyzed for18 hours at 115°C in 12N hydrochloride. After hydrolyzation,samples were mixed with methyl red, titrated with sodium hy-droxide and hydrochloride, and then diluted with deionizedwater to reach the volume of 1 mL. Trans-4 hydroxy-L-proline (Sigma-Aldrich) was dissolved in deionized water to

generate the standard curve. Both the standards and dilutedsamples were mixed with chloramine-tosylchloramide hy-drate and p-dimethylaminobenzaldehyde and then incu-bated in 60°C for 30 minutes. Absorbance values were mea-sured at 550 nm on a UV-Vis spectrophotometer, and the ratioof 1:10 hydroxyproline to collagen was used to calculate totalcollagen content.

RNA Extraction and RT-PCRHarvested constructs were homogenized in TRIzol Reagent (In-vitrogen) in preparation for total RNA extraction. The protocolaccompanying the reagent was followed for the extraction.Complementary DNA was then synthesized using the reversetranscriptase Superscript First-Strand Synthesis kit (Invitro-gen). Reverse transcription–polymerase chain reaction (RT-PCR) was performed with Taq recombinant polymerase or withSYBR Green PCR Master Mix (Applied Biosystems) and con-ducted on the ABI Prism 7700 Sequence Detection System (Per-kin Elmer/Applied Biosystems). Amplicons of RT-PCR weremixed with a loading dye and then run on 2% agarose gels inTris, acetic acid, and EDTA buffer within an electrophoresis boxwithanaccompanyingladderdye.Picturesofthegelsweretakenafter submerging them in diluted ethidium bromide and ex-posing them to UV light. The following genes were analyzed:type I collagen, MMP1, MMP2, MMP3, MMP9, MMP13, and MT1-MMP. All genes were normalized to the β-actin housekeepinggene. The MMP primers are listed in the article by Konttinen etal.14 TypeIcollagenprimersareF-TGACGAGACCAAGAACTGandR-CCATCCAAACCACTGAAACC, and β-actin primers areF-TGGCACCACACCTTCTACAATGAGC and R-GCACAGCTTCTC-CTTAATGTCACGC.

Figure 1. Type I Bovine Collagen In Vitro Model

HS27 KF

07 14

2.0

DN

A/D

ry W

eig

ht,

μg

/mg

Day

1.5

1.0

0.5

2

A B

C

07 14

2.0

GA

G/D

NA

, μ

g/μ

g

Day

1.5

1.0

0.5

2

D

07 14

20

Tota

l C

oll

agen

/DN

A,

μg

/μg

Day

15

10

5

2

E

PEGDA (UV)

Type 1Collagen

Fibrils

Fibroblasts

HS27 Keloid

Type I Collagen

β-Actin

2 7 14 2 7 14

Day

a

bc

d

dd

d

e

Letters indicate statistical significance from HS27 at corresponding time point(P < .05e and P< .001d); plus signs, statistical significance from day 2 of samecell type (P < .05c, P < .01a, and P < .001b). A, Schematic of type I bovine collagenin vitro model. Polyethylene glycol diacrylate (PEGDA) was mixed with type Icollagen fibrils before encapsulation of fibroblasts (HS27 and keloid fibroblasts)through UV photopolymerization. B, Reverse transcription–polymerase chainreaction demonstrated that keloid fibroblasts had increased type I collagen

gene expression compared with normal fibroblasts. C, DNA quantificationnormalized to dry weight indicated decreasing trend over time.D, Glycosaminoglycan (GAG) content by both fibroblast types normalized toDNA demonstrated increased GAG production from keloid fibroblasts. E, Totalcollagen content by both cell types normalized to DNA also demonstratedincreased collagen content from keloid fibroblasts.

Tissue Engineering for Keloid Lesions Original Investigation Research



Statistical AnalysisAll analyses were performed in triplicate and analyzed withthe t test for pairwise comparison. Statistical significance wasset at P <.05.

ResultsVerification of PEG-Col1 ModelConstructs were harvested at days 2, 7, and 14 for both nor-mal (HS27) and keloid fibroblasts, and RT-PCR demonstratedhigher expressions of type I collagen from keloid fibroblastsat all 3 time points (Figure 1B). This finding confirms that thealtered phenotype that is associated with keloid disease is re-tained in this 3-dimensional PEG-Col1 model.

Matrix Production From Fibroblasts in the PEG-Col1 ModelBoth HS27 and keloid fibroblasts demonstrated a decrease inDNA quantification during a span of 14 days in 3-dimensionalculture (Figure 1C). However, at all time points, the GAG andtotal collagen production was significantly higher in keloid fi-broblasts compared with normal fibroblasts (Figure 1, D andE). Specifically, there was a mean 2.02-fold increase in GAG perDNA and a 3.43-fold increase in total collagen per DNA in ke-loid fibroblasts.

MMP Gene Expression in Fibroblasts in the PEG-Col1 ModelThe RT-PCR for MMP1, MMP2, MMP3, MMP9, MMP13, andMT1-MMP indicated that there were no significant differ-ences between normal and keloid fibroblasts over time(Figure 2). MMP1, MMP2, MMP3, and MT1-MMP were consis-tently expressed in both fibroblast types at all time points.

MMP9 and MMP13 decreased in expression in both HS27 andkeloid fibroblasts as time increased.

Verification of the Matrigel ModelConstructs were harvested at days 1, 7, and 21, and as with thePEG-Col1 model, type I collagen was more strongly expressedby keloid fibroblasts than HS27 at all time points, indicatingthat the keloid disease phenotype is maintained in the Matri-gel model (Figure 3, A and B).

Matrix Production From Fibroblasts in the Matrigel ModelAs with the PEG-Col1 model, there was a decrease in DNAquantification as the time increased for both HS27 andkeloid fibroblasts (Figure 3C). Both GAG and total collagenwere produced in significantly greater amounts in keloidfibroblasts than in normal cells (Figure 3, D and E). Specifi-cally, keloid fibroblasts produced 1.37-fold, 1.56-fold, and2.08-fold more GAG per DNA than HS27 at days 1, 7, and 21,respectively. There was a time-dependent increase in GAGproduction in both normal and diseased fibroblasts; how-ever, the fold changes from keloid fibroblasts were signifi-cantly higher than HS27. The trend for total collagen perDNA was similar to the GAG data. Keloid fibroblasts pro-duced more matrix than HS27 at days 7 and 21, although thisfinding was not statistically significant. In addition, therewas markedly more collagen at day 21 than day 1 from thekeloid fibroblasts.

MMP Gene Expression in Fibroblasts in the Matrigel ModelMMP1, MMP2, MMP3, and MT1-MMP were consistently ex-pressed in both HS27 and keloid fibroblasts in the Matrigelmodel at all time points (Figure 4). HS27 expressed little to noMMP9 and MMP13, whereas keloid fibroblasts increased ex-pression of both MMP genes over time.

Effects of Decorin on Matrix Production in Keloid Fibroblastsin the Matrigel ModelDecorin was exogenously added into the medium of keloid fi-broblasts encapsulated in Matrigel from day 1 to day 21. At days7 and 21, constructs were harvested after treatment with deco-rin. There was a decrease in DNA quantification as time in-creased, whether decorin was added or not (Figure 5A). Di-methylmethylene blue assay demonstrated that, at day 21, thequantities of GAG per DNA were the same at day 7 and day 21when keloid fibroblasts were exposed to decorin (Figure 5B).This resulted in significantly less GAG per DNA at day 21 in ke-loid fibroblasts that were exposed to decorin compared withkeloid fibroblasts that were not exposed to decorin. Quantifi-cation of total collagen per DNA also demonstrated a similartrend as the GAG data, although the finding was not statisti-cally significant (Figure 5C).

Effects of Decorin on Gene ExpressionFrom Keloid FibroblastsType I collagen gene expression was significantly downregu-lated at day 21, when decorin was administered to keloid fi-broblasts (Figure 5D). In addition, MMP1 and MMP13 were alsosignificantly decreased at day 21 when keloid fibroblasts were

Figure 2. Reverse Transcription–Polymerase Chain Reactionof MMP Gene Expressions From Normal and Keloid Fibroblastsin the PEG-Col1 Model

HS27 Keloid

MMP1

MMP2

MMP3

MMP9

MMP13

MT1-MMP

ß-Actin

2 7 14 2 7 14

Day

No obvious differences in trends were observed between HS27 and keloidfibroblasts. β-Actin was the housekeeping gene.




treated with decorin. MMP9 did not change significantly ingene expression between day 7 and day 21 when keloid fibro-blasts were treated with decorin, and this differed from un-treated fibroblasts for which an increase in MMP9 was ob-served by day 21.

MMP Gene Expression From Specific Sites of Keloid LesionTwo keloid lesions were received from one patient; one le-sion was removed from the shoulder and the other from theear. Because of the small size of the auricle lesion, the enzy-matic isolation resulted in a mix of fibroblasts that were notseparated by tissue depth and thus were labeled as “mixed au-ricle.” The lesion from the shoulder was separated into 4 dif-ferent sites: superficial side fibroblasts, superficial center fi-broblasts, deep center fibroblasts, and keratinocytes(Figure 6A). The MMP gene expressions were analyzed amongthe different cell populations, both shoulder and auricle, andcompared with normal fibroblasts that were isolated from theskin of a patient who underwent eyelid skin excision.

Normal fibroblasts demonstrated little to no expression ofall MMP genes tested (Figure 6B). In comparing the fibro-blasts isolated from the shoulder, the superficial side and deepcenter had the highest expressions of MMP1, MMP2, MMP3,MMP9, and MT1-MMP. The superficial center fibroblasts hadlittle to no expression of those MMP genes. All fibroblasts fromthe shoulder had little to no expression of MMP13. The kera-tinocytes also expressed MMP2 and MMP3 with comparableintensities to superficial center fibroblasts and slightly higherexpressions of MMP1 and MMP9 than the superficial center fi-broblasts. The mixed auricle fibroblasts had high expressionsof all tested MMP genes, including MMP13.

DiscussionKeloids remain a clinical challenge and a source of significantpsychological distress to patients who can have grossly dis-figuring and painful lesions. Current treatment recommenda-

Figure 3. Matrigel In Vitro Model

HS27 Keloid fibroblasts

07 21

35

DN

A/D

ry W

eig

ht,

μg

/mg

Day

25

15

5

30

20

10

1

A B

C

07 21

GA

G/D

NA

, μ

g/μ

g

Day

1.5

1

0.5

1

D

07 21

1.2

Tota

l C

oll

agen

/DN

A,

μg

/μg

Day

0.8

0.4

1

0.6

0.2

1

E

Matrigel(37°C)

Fibroblasts HS27 Keloid

Type I Collagen

β-Actin

1 7 21 1 7 21

Day

a b

b

a

dac

de

bf

a

Letters indicate statistical significance from HS27 at corresponding time point(P < .05c, P < .01e, and P< .001b); plus signs, statistical significance from day 1 ofsame cell type (P < .05a, P < .01d, and P < .001f). A, Schematic of 3-dimensionalMatrigel constructs. B, Reverse transcription–polymerase chain reaction of typeI collagen from both HS27 and keloid fibroblasts after culture in Matrigeldemonstrates the expected higher gene expression from keloid fibroblasts.

C, DNA content decreased in both fibroblast types as time increased.D, Glycosaminoglycan (GAG) content was more significantly produced fromkeloid fibroblasts than HS27, resulting in a time-dependent increase. E, Totalcollagen content normalized to DNA also demonstrated more extracellularmatrix production from keloid fibroblasts than HS27.

Figure 4. Reverse Transcription–Polymerase Chain Reaction of MMPGene Expressions From Normal and Keloid Fibroblasts After In Vitro3-Dimensional Culture in Matrigel

HS27 Keloid

MMP1

MMP2

MMP3

MMP9

MMP13

MT1-MMP

ß-Actin

1 7 21 1 7 21

Day

Significant differences were observed in MMP9 and MMP13 gene expressionsbetween HS27 and keloid fibroblasts. β-Actin was the housekeeping gene.




tions reflect individual clinician experiences. The incom-plete understanding of keloid pathophysiology is in part dueto a lack of an animal model, preventing in vivo investigationof this disease. As stated previously, various hypotheses havebeen proposed to explain the pathogenesis of the disease. Oneof the hypotheses involves the state of MMP activity in keloiddisease, of which the literature has reported conflicting re-sults. For example, Seifert et al12 observed a downregulationof MMP3 from keloid fibroblasts when compared with nor-mal cells, whereas Fujiwara et al11 observed a several-fold in-crease in production of MMP1 and MMP2. Both of these stud-ies analyzed results after monolayer culture, which couldpotentially explain this discrepancy. Therefore, in the first partof this experiment, we studied keloid fibroblasts in their morenative 3-dimensional environment and observed how MMPgene expressions in keloid fibroblasts vary from normal fibro-blasts. Through the 3-dimensional models, we also evaluatedthe ECM produced by diseased and normal fibroblasts in vitroand observed the effects of an exogenously administered pro-teoglycan, decorin, known to have growth factor–regulatingproperties.

In the first model, fibroblasts were encapsulated in the hy-drogel polymer PEGDA. Because fibroblasts possess adher-ent properties and polyethylene glycol alone does not pro-mote cellular adhesion, type I collagen fibrils were incorporatedinto PEGDA to create a semi-interpenetrating network. In ad-dition, the presence of type I collagen fibrils better mimickedwhat the fibroblasts sense in their native environment within

connective tissue, in addition to promoting cellular adhe-sion. Although in vitro encapsulation of both normal and dis-eased fibroblasts resulted in type I collagen gene expressionat all 3 time points of harvest, keloid fibroblasts expressed sig-nificantly more type I collagen than HS27, thus indicating thatthe diseased fibroblasts in this 3-dimensional model main-tain their diseased phenotype in vitro. Biochemical analysisalso indicated that at all 3 time points diseased fibroblasts hadhigher rates of ECM production than normal fibroblasts. There-fore, keloid fibroblasts in this 3-dimensional model still ex-hibited their diseased characteristics not only in gene expres-sion but also in actual matrix synthesis.

Interestingly, not a lot of differences were found in MMPgene expressions between normal and keloid fibroblasts whencultured in the PEG-Col1 3-dimensional model. In general, mostof the analyzed MMP genes were expressed consistentlythrough all time points for both fibroblast types. However, bothMMP9 (gelatinase B) and MMP13 (collagenase 3) were stronglyexpressed at day 2 for both cell types and then downregu-lated as time increased. Although this finding suggested thatMMP9 and MMP13 are potential therapeutic targets, more re-search is needed to confirm these results.

In creating the next in vitro 3-dimensional model for fur-ther analysis of keloid disease, we focused on implementinga material that would better mimic the diseased environ-ment. Matrigel is a heterogeneous basement membrane pro-tein mixture secreted by cells from mouse sarcoma and com-mercialized by BD Bioscience. It resembles the ECM of various

Figure 5. Effect of Decorin on Keloid Fibroblasts in the Matrigel Model

Keloid fibroblasts

Keloid fibroblasts and Decorin

021

25

20

DN

A/D

ry W

eig

ht,

μg

/mg

Day

15

10

5

7

a

021

1.2

0.8

1.0

Tota

l C

oll

agen

/DN

A,

μg

/μg

Day

0.6

0.4

0.2

7

021

1.6

1.2

GA

G/D

NA

, μ

g/μ

g

Day

0.8

0.4

1.4

1.0

0.6

0.2

7

a

b

Keloid fibroblastsand decorinKeloid

Type I Collagen

MMP9

MMP13

MMP1

ß-Actin

7 21 7 21

Day

A B

C D

Letters indicate statistical significance from HS27 at corresponding time point(P < .01b); plus signs, statistical significance from day 7 of same cell type(P < .001a). A, DNA quantification normalized to respective dry weightsindicated decrease in both conditions as time increased. B, Glycosaminoglycan(GAG) content normalized to DNA demonstrated that the presence of decorinprevented an increase in the matrix component production. C, Total collagencontent normalized to DNA demonstrated a similar trend as the GAG data,

although not statistically significant. D, Reverse transcription–polymerase chainreaction of type I collagen and MMP genes that were affected by the presenceof decorin administered to keloid fibroblasts. Significant downregulations intype I collagen, MMP1, and MMP13 were observed at day 21, whereas MMP9retained the same expression as day 7 and was prevented from beingupregulated.




connective tissues, and it is composed of different matrix com-ponents, such as gelatin, fibronectin, laminin, type IV colla-gen, and various growth factors. Matrigel has been com-monly used for the study of cellular attachment anddifferentiation and especially the study of tumor cellinvasion.15-17 In addition, connective tissue diseases have beenassociated with keloids.18,19 Therefore, Matrigel was used inour second 3-dimensional in vitro model to encapsulate ke-loid fibroblasts for further characterization of the diseased cells.

Similar to the PEG-Col1 model, type I collagen was ex-pressed strongly by keloid fibroblasts and weakly by normalfibroblasts when the cells were cultured in Matrigel. Keloid fi-broblasts also produced more GAG and total collagen per DNAthan HS27, thus demonstrating that culturing the diseased fi-broblasts in Matrigel maintained their excessive synthetic ac-tivity state. Interestingly, more significant differences were ob-served with GAG production than collagen production, withhigher-fold changes in GAG per DNA from the diseased fibro-blasts as time increased. The literature has cited higher con-centrations of proteoglycans from healing wounds on animalmodels and different types of tumors removed from patientsand overexpression of various proteoglycans, such as chon-droitin sulfate, versican, and dermatan sulfate, throughmicroarray.20 Therefore, the increase in GAG production perDNA from the keloid fibroblasts in the Matrigel model is in ac-cord with the benign tumorlike characteristic of keloids. In ad-dition, the continuous increase in GAG production from thekeloid fibroblasts suggests that the diseased cells have an ab-normal phenotype that retains them in a perpetual healingstage, which could be the cause of the excessive fibroprolif-eration of the lesions.

The MMP gene analysis from the Matrigel model demon-strated that MMP1, MMP2, MMP3, and MT1-MMP were gen-erally consistently expressed from both HS27 and keloid fi-broblasts at all time points, similar to the trends observed withthe PEG-Col1 model. However, significant differences were ob-served in MMP9 and MMP13. Although there was little to noexpression of either MMP gene from the normal fibroblasts atany time point, keloid fibroblasts expressed more MMP9 andMMP13 as time increased. Therefore, the data from the Matri-gel model suggest that the gelatinase and collagenase are 2 mainplayers in the progression of keloid lesions through an upregu-lation of MMP gene expression that potentially stimulate moreremodeling of the wound and thus result in excessive tissueformation. In addition, because both MMP genes were upregu-lated in expression as time increased, this finding indicates thatthe 2 proteases potentially function in parallel within keloidlesions.

The difference observed in the MMP gene expressionchanges between the 2 different biomaterial models supportsthe hypothesis that the surrounding ECM has an important rolein the keloid pathogenesis. Another hallmark of the keloid ECMis the lack of structural organization in the scar formation.Therefore, we hypothesized that treating the diseased cells witha molecule that promotes orderly ECM arrangement could bea therapeutic treatment for the disease. Decorin is a form ofGAG that functions to maintain skin integrity and structuralorganization by binding to collagen fibrils. Yeo et al20 ob-

served that healing skin has significantly less decorin than nor-mal skin and that connective tissue from normal human breaststained strongly for decorin, whereas Reed and Iozzo21 ob-served irregular collagen fibrils in decorin knockout mice. Inour study, decorin was administered exogenously to keloid fi-broblasts with every medium change in an attempt to pro-duce a more organized ECM rather than the random arrange-ment that is characteristic of keloid ECM. After 21 days,although untreated keloid fibroblasts increased GAG produc-tion from day 7, those that were treated with decorin main-tained the same content of GAG per DNA, suggesting that thepresence of decorin prevented the diseased fibroblasts fromproducing more GAG. In addition, a similar trend was ob-served when quantifying total collagen per DNA at day 21 whenuntreated keloid fibroblasts produced more of the matrix than

Figure 6. Site-Specific Gene Expressions

MMP1

MMP2

MMP3

MMP9

MMP13

MT1-MMP

ß-Actin

A

B

Keloid

Shoulder

Side of excision

iv

ii

ii

iii

SuperficialSide

SuperficialCenter

DeepCenter

Keratino-cytes

MixedAuricle

NormalFibroblasts

Shoulder

A, Different sites from which keloid fibroblasts were isolated from the shoulderlesion for comparison of MMP gene expressions among the different sites:i, superficial side; ii, superficial center; iii, deep center; and iv, keratinocytes.Keratinocytes were also isolated from the shoulder lesion. B, Reversetranscription–polymerase chain reaction of MMP gene expressions fromprimary fibroblasts isolated from different sites of a keloid lesion and comparedwith the corresponding keratinocytes that lined the epidermis of the lesion, amixed fibroblast population from the ear, and normal fibroblasts isolated froman eyebrow lift. Fibroblasts from regions closest to the margins of the originalwound and the excision site (ie, superficial side and deep center) had higherexpressions of MMP1, MMP2, MMP3, and MMP9 compared with those farthestaway from the wound boundaries (superficial center). Keratinocytes alsoexpressed the same 4 MMP genes, although at lower intensities, whencompared with the superficial side and deep center. The mixed keloid fibroblasthad high expressions of MMP genes, whereas normal fibroblasts demonstratedlittle to no expression of the enzymes.




treated fibroblasts. We hypothesize that the presence of deco-rin bound to the collagen fibrils in the surrounding environ-ment of the keloid fibroblasts, which resulted in a more orga-nized arrangement of the ECM and thus downregulated ECMproduction.

In addition, type I collagen and MMP gene expressions wereaffected by the presence of decorin. Type I collagen, MMP1,and MMP13 were significantly downregulated in expression atday 21 after continuous treatment with decorin, whereas MMP9expression was maintained from day 7 to day 21 without theincrease that was observed from untreated fibroblasts. Thisfinding suggests that an increase in MMP expression is asso-ciated with the growth of a keloid lesion and that the pres-ence of a GAG that enhances structural organization coulddownregulate the expressions of the MMP genes. Further-more, we observed that an increased expression of MMP genescoincides with excessive ECM deposition, which also sup-ports the idea that targeting MMP genes can therapeuticallyregulate ECM production.

Heterogeneity is another characteristic of keloid diseasethat renders it a challenge to fully elucidate the mechanismsthat govern its pathogenesis. Specifically, differences are ob-served in cell behavior and phenotype, depending on the re-gion of keloids from which the cells were derived. Lu et al22

demonstrated that most fibroblasts derived from the periph-ery of keloid lesions were in a proliferative state, whereas fi-broblasts isolated from the central region of keloids were gen-erally in the quiescent phase. A study by Sayah et al23

demonstrated that the apoptotic indices were different be-tween peripheral fibroblasts and central fibroblasts. A 2008published work on monolayer culture by Seifert et al12 sug-gested that site specificity of fibroblast isolation can affect howgene expressions are observed. The heterogeneity demon-strated by these works on site specificity could also poten-tially explain the varying results that have been observed ondiffering MMP gene expressions, in addition to the mono-layer cultures. Therefore, to study the changes in MMP geneexpression due to site specificity, we isolated primary fibro-blasts from different regions of a keloid lesion and comparedthe MMP gene expressions to a mixed keloid cell populationand to a normal cell population. Fibroblasts were isolated from

the dermis of a shoulder lesion and separated into 3 regionsas depicted in Figure 6A. There were higher MMP gene ex-pressions from the regions of the lesion that were closer to themargins of the original wound (ie, superficial side) and the re-gion within the deep center of the keloid. The fibroblasts fromthe superficial periphery and the deep center strongly ex-pressed MMP1, MMP2, MMP3, and MMP9 when compared withthe superficial center fibroblasts. The cells isolated from theear consisted of a mixed fibroblast population and strongly ex-pressed all tested MMP genes, whereas fibroblasts from nor-mal skin generated very little to no expression. These data sug-gest that fibroblasts that are closer to the margins of the originalwound, where active cellular migration and reepithelializa-tion occur, have higher MMP gene expression, which couldhave a role in stimulating more tissue remodeling, thus result-ing in excessive growth. In addition, literature has cited thatinteractions between keratinocytes and fibroblasts are impor-tant in the process of wound healing, and the presence of ke-loid keratinocytes can increase fibroblast proliferation and thesecretion of soluble collagen types I and III.24,25 Therefore, wealso compared differences in the protease gene expression be-tween keloid fibroblasts and keloid keratinocytes derived fromthe shoulder lesion. However, the keratinocytes did not ex-press the MMP genes with as much intensity as some of thefibroblast population, thus indicating that the fibroblasts withinthe dermis of the keloid lesion are most likely the predomi-nant contributors in higher MMP gene expressions.

In conclusion, our in vitro data demonstrate that MMP9and MMP13 are 2 potential targets in therapeutically treatingkeloid lesions. Their concurrent expressions from the Matri-gel 3-dimensional disease model suggest that they enhance theexpressions of each other, thus promoting the growth of ke-loid lesions. Through manipulating the ECM, we observed thatmolecules that target organization of the lesion’s matrix canpossibly be beneficial in downregulating increased markersduring the disease. Site specificity of fibroblasts from keloidlesions have demonstrated different intensities of MMP geneexpressions, thus further supporting the heterogeneity that isobserved in keloid fibroblast phenotype. These studies set theprecedent for future tissue engineering studies to better elu-cidate the keloid pathogenesis.

ARTICLE INFORMATION

Accepted for Publication: January 27, 2013.

Published Online: September 19, 2013.doi:10.1001/jamafacial.2013.1211.

Author Contributions: Study concept and design:Elisseeff, Boahene, Li, Nahas.Acquisition of data: Li, Nahas, Feng.Analysis and interpretation of data: Elisseeff,Boahene, Li, Nahas.Drafting of the manuscript: Boahene, Feng, Li,Nahas.Critical revision of the manuscript for importantintellectual content: Elisseeff, Boahene, Li, Nahas.Obtained funding: Elisseeff.Study supervision: Elisseeff, Boahene.

Conflict of Interest Disclosures: None reported.

Additional Contributions: Bindi Naik, MD, ofBaylor College of Medicine, provided keloid

fibroblasts; Dr Boahene provided additional keloidfibroblasts enzymatically isolated from the dermallayer of lesions removed from consenting patientsat The Johns Hopkins Hospital.

REFERENCES

1. Clark RA, Ghosh K, Tonnesen MG. Tissueengineering for cutaneous wounds. J InvestDermatol. 2007;127(5):1018-1029.

2. Pollack SV. Wound healing: a review, I: thebiology of wound healing. J Dermatol Surg Oncol.1979;5(5):389-393.

3. Tredget EE, Nedelec B, Scott PG, Ghahary A.Hypertrophic scars, keloids, and contractures: thecellular and molecular basis for therapy. Surg ClinNorth Am. 1997;77(3):701-730.

4. Al-Attar A, Mess S, Thomassen JM, Kauffman CL,Davison SP. Keloid pathogenesis and treatment.Plast Reconstr Surg. 2006;117(1):286-300.

5. Younai S, Nichter LS, Wellisz T, Reinisch J, NimniME, Tuan TL. Modulation of collagen synthesis bytransforming growth factor-beta in keloid andhypertrophic scar fibroblasts. Ann Plast Surg.1994;33(2):148-151.

6. Bettinger DA, Yager DR, Diegelmann RF, CohenIK. The effect of TGF-beta on keloid fibroblastproliferation and collagen synthesis. Plast ReconstrSurg. 1996;98(5):827-833.

7. Kischer CW, Wagner HN Jr, Pindur J, et al.Increased fibronectin production by cell lines fromhypertrophic scar and keloid. Connect Tissue Res.1989;23(4):279-288.




8. Alaish SM, Yager DR, Diegelmann RF, Cohen IK.Hyaluronic acid metabolism in keloid fibroblasts.J Pediatr Surg. 1995;30(7):949-952.

9. Hunzelmann N, Anders S, Sollberg S, SchönherrE, Krieg T. Co-ordinate induction of collagen type Iand biglycan expression in keloids. Br J Dermatol.1996;135(3):394-399.

10. Chen MA, Davidson TM. Scar management:prevention and treatment strategies. Curr OpinOtolaryngol Head Neck Surg. 2005;13(4):242-247.

11. Fujiwara M, Muragaki Y, Ooshima A.Keloid-derived fibroblasts show increased secretionof factors involved in collagen turnover and dependon matrix metalloproteinase for migration. Br JDermatol. 2005;153(2):295-300.

12. Seifert O, Bayat A, Geffers R, et al. Identificationof unique gene expression patterns within differentlesional sites of keloids. Wound Repair Regen.2008;16(2):254-265.

13. Uchida G, Yoshimura K, Kitano Y, Okazaki M,Harii K. Tretinoin reverses upregulation of matrixmetalloproteinase-13 in human keloid-derivedfibroblasts. Exp Dermatol. 2003;12(suppl 2):35-42.

14. Konttinen YT, Ainola M, Valleala H, et al.Analysis of 16 different matrix metalloproteinases(MMP-1 to MMP-20) in the synovial membrane:different profiles in trauma and rheumatoidarthritis. Ann Rheum Dis. 1999;58(11):691-697.

15. Ma L, Teruya-Feldstein J, Weinberg RA. Tumourinvasion and metastasis initiated by microRNA-10bin breast cancer. Nature. 2007;449(7163):682-688.

16. Repesh LA. A new in vitro assay for quantitatingtumor cell invasion. Invasion Metastasis.1989;9(3):192-208.

17. Lal A, Glazer CA, Martinson HM, et al. Mutantepidermal growth factor receptor up-regulatesmolecular effectors of tumor invasion. Cancer Res.2002;62(12):3335-3339.

18. Igarashi A, Nashiro K, Kikuchi K, et al.Connective tissue growth factor gene expression intissue sections from localized scleroderma, keloid,and other fibrotic skin disorders. J Invest Dermatol.1996;106(4):729-733.

19. Leask A, Holmes A, Abraham DJ. Connectivetissue growth factor: a new and important player in

the pathogenesis of fibrosis. Curr Rheumatol Rep.2002;4(2):136-142.

20. Yeo TK, Brown L, Dvorak HF. Alterations inproteoglycan synthesis common to healing woundsand tumors. Am J Pathol. 1991;138(6):1437-1450.

21. Reed CC, Iozzo RV. The role of decorin incollagen fibrillogenesis and skin homeostasis.Glycoconj J. 2002;19(4-5):249-255.

22. Lu F, Gao J, Ogawa R, Hyakusoku H, Ou C.Biological differences between fibroblasts derivedfrom peripheral and central areas of keloid tissues.Plast Reconstr Surg. 2007;120(3):625-630.

23. Sayah DN, Soo C, Shaw WW, et al.Downregulation of apoptosis-related genes inkeloid tissues. J Surg Res. 1999;87(2):209-216.

24. Werner S, Krieg T, Smola H. Keratinocyte-fibroblast interactions in wound healing. J InvestDermatol. 2007;127(5):998-1008.

25. Lim IJ, Phan TT, Bay BH, et al. Fibroblastscocultured with keloid keratinocytes: normalfibroblasts secrete collagen in a keloidlike manner.Am J Physiol Cell Physiol. 2002;283(1):C212-C222.




tissue engineering for in vitro analysis of matrix metalloproteinases in the pathogenesis of keloid...

Documents