Assimilating cell sheets and hybrid scaffolds for dermaltissue engineering

Kee Woei Ng,1 Wanrong Tham,2 Thiam Chye Lim,3 Dietmar Werner Hutmacher4,5

1Department of Surgery, National University of Singapore, 9 Engineering Drive 1, Singapore 1175762Department of Biological Sciences, National University of Singapore, 9 Engineering Drive 1, Singapore 1175763Department of Plastic Surgery, National University Hospital, 9 Engineering Drive 1, Singapore 1175764Division of Bioengineering, National University of Singapore, 9 Engineering Drive 1, Singapore 1175765Department of Orthopaedics Surgery, National University of Singapore, 5 Lower Kent Ridge Road, Singapore 119074

Received 6 March 2005; revised 7 April 2005; accepted 18 April 2005Published online 16 August 2005 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/jbm.a.30454

Abstract: Cell sheets can be used to produce neo-tissuewith mature extracellular matrix. However, extensive con-traction of cell sheets remains a problem. We devised atechnique to overcome this problem and applied it to tissueengineer a dermal construct. Human dermal fibroblastswere cultured with poly(lactic-co-glycolic acid)-collagenmeshes and collagen-hyaluronic acid foams. Resulting cellsheets were folded over the scaffolds to form dermal con-structs. Human keratinocytes were cultured on these dermalconstructs to assess their ability to support bilayered skinregeneration. Dermal constructs produced with collagen-hyaluronic acid foams showed minimal contraction, whilethose with poly(lactic-co-glycolic acid)-collagen meshescurled up. Cell proliferation and metabolic activity profileswere characterized with PicoGreen� and AlamarBlue™ as-says, respectively. Fluorescent labeling showed high cell

viability and F-actin expression within the constructs. Col-lagen deposition was detected by immunocytochemistryand electron microscopy. Transforming Growth Factor-�and �1, Keratinocyte Growth Factor and Vascular Endothe-lial Growth Factor were produced at various stages of cul-ture, measured by RT-PCR and ELISA. These results indi-cated that assimilating cell sheets with mechanically stablescaffolds could produce viable dermal-like constructs thatdo not contract. Repeated enzymatic treatment cycles for cellexpansion is unnecessary, while the issue of poor cell seed-ing efficiency in scaffolds is eliminated. © 2005 Wiley Peri-odicals, Inc. J Biomed Mater Res 75A: 425–438, 2005

Key words: skin; extracellular matrix; poly(lactic-co-glycolicacid); collagen; tissue engineering

INTRODUCTION

The clinical demand for tissue-engineered dermaland composite skin replacements continues to be high,despite there being a number of such products (Der-magraft�, Apligraf�) available today.1 Limitations ofthese products include poor mechanical properties,high costs and limitation of size. An attractive alter-native to current strategies of engineering the dermisis by the use of cell sheets. In recent years, tissueengineers have extended the use of cell sheets to pro-duce three-dimensional (3D) neo-tissues.2 Using cell

sheets has the advantage that an entirely natural neo-tissue assembled by the cells, with mature extracellu-lar matrix (ECM), can be produced, thereby enhancingthe regenerative capacity of the constructs. L’Heureuxet al.3 successfully built a tissue-engineered blood ves-sel by wrapping layers of fibroblasts and smooth mus-cle cells around a mandrel. This vessel exhibited welldefined three-layered organisation and had a burst-strength comparable to that of native human vessels.The group also stacked layers of fibroblast sheets toform a dermal equivalent, upon which keratinocyteswere cultured to form a skin equivalent.4 Advance-ment in the culture of cell sheets came with the devel-opment of culture dishes coated with temperature-responsive polymers. These culture surfaces becomehydrophilic at temperatures below a critical point, andhydrophobic at temperatures above the critical point.5

As a result, cells attach and proliferate on this surfaceabove the critical temperature, and easily detach whenthe temperature is reduced.6 It was shown that fourlayers of cardiomyocytes cultured using this tech-

Correspondence to: D.W. Hutmacher; e-mail: biedwh@tnus.edu.sg

Contract grant sponsor: Defence Science and TechnologyAgency; contract grant number: D20000-4530

Contract grant sponsor: National University of SingaporeYoung Investigator Award; contract grant number: R-397-000-601-712

© 2005 Wiley Periodicals, Inc.

nique formed electrically communicative and pulsatilemyocardial tissue both in vitro and in vivo.7

However, although cell sheets are strong enough toallow careful manipulation in a laboratory to producestacked or wrapped constructs, they contract exten-sively upon detachment from culture surfaces. Thus, itis a challenge to engineer large-size tissues of specificshape, using cell sheets alone. External supports suchas stainless steel rings3 and polymeric membranes8 aretherefore used. However, these supports are removedat the point of application, thereby still resulting inloss of size, shape, and even biological functions.

Our aim was to devise a novel technique to withstandthe contraction of cell sheets in vitro, and apply it fortissue engineering a dermal equivalent. We hypothe-sized that high density cultures of cells together withmechanically stable scaffolds will produce mature cellsheets with the scaffolds integrated within them. The cellsheets could then be folded over the scaffolds to form 3Ddermal-like constructs, which do not contract. The keybenefits of assimilating cell sheets with 3D scaffolds in-clude the possibility of producing a dermal constructswith mature ECM, which do not contract. In addition,the issue of cell seeding efficiency is abrogated, and theundesirable effects of repeated enzymatic treatment cy-cles are circumvented. Keratinocytes were further cul-tured on these dermal-like constructs to evaluate theirability to support the regeneration of a bilayered skinequivalent. In this study, we evaluated this techniqueusing two scaffolds with the potential to be used fordermal regeneration—a weft-knitted poly(lactic-co-gly-colic acid) mesh with collagen (PLGA-c), and a collagensponge crosslinked with hyaluronic acid (CHA). Ourprevious study has showed that a weft-knitted PLGA-based mesh was capable of supporting dermal fibroblastattachment and proliferation, without contraction.9

However, neo-tissue formed was immature and inho-mogeneously distributed.9 Using cell sheets in combina-tion with these meshes could result in the production ofmore mature ECM. CHA was chosen as a comparisonbecause it was shown to exhibit mechanical stability interms of greater resistance to enzymatic degradation anddecreased swelling.10 To our knowledge, this is the firststudy in which cell sheets are cultured together with 3Dscaffolds for dermal tissue engineering.

MATERIALS AND METHODS

Substrate preparation

Poly(lactic-co-glycolic acid) (PLGA 10:90) threads (ShanghaiTianchun Biomaterials Company Ltd., Shanghai, China) madeup of 12 continuous monofilaments of 20-�m diameter eachwere used. Two threads were combined to form a PLGA yarnand knitted into a mesh using a Silver Reed SK270 Knitting

Machine (Suzhou Zhenzuo Mechanical Instrument CompanyLtd., Suzhou, China). The meshes were folded and fused byheat along the edges to form two-layered meshes, and weresubsequently cut into 0.8 � 0.8 cm squares. The thickness of themeshes was 326 �20 �m.11 Extracted rat-tail collagen (1.3mg/mL in 0.05% acetic acid) was polymerized into the PLGAmeshes by neutralizing with 71.2 mg/mL sodium bicarbonate(volume ratio 100:9). The PLGA-c scaffolds were then frozen at�80°C overnight and lyophilized for at least 6 h to form thefinal scaffold (Fig. 1). The scaffolds were UV-sterilized prior touse. CHA sponges crosslinked with starch dialdehyde weregenerously supplied by Dr. D. Bakoo from Slovak TechnicalUniversity in sterile packages.10 The sponges were also cutinto 0.8 � 0.8-cm squares and used without further treat-ment (Fig. 1).

Cell isolation, culture, and seeding

Human dermal fibroblasts (HDFs) and human keratinocytes(HKs) were harvested from human skin samples obtainedfrom abdominoplasty, with informed consent and under aprotocol approved by the Institutional Review Board. Enzy-matic digestion of the skin sample was carried out as describedpreviously.9 Briefly, epidermis and dermis were separated via5 mg/mL dispase treatment for 16 h at 4°C. The dermis wasdisaggregated with 2 mg/mL collagenase type I (Gibco, Grandisland, NY), overnight at 37°C. Isolated HDFs were cultured inDMEM (Gibco) supplemented with 10% FBS (Hyclone, SouthLogan, UT) and 1% Penicillin-Streptomycin solution (Sigma-Aldrich, St. Louis, MO), with medium change every 2 days.The epidermis was digested with 0.1% trypsin for 10 min at37°C. Isolated HKs were cultured in Defined Keratinocyte Se-rum-free Medium (Gibco) supplemented with 1% Penicillin-Streptomycin solution (Sigma-Aldrich), with medium changeevery 3 days. HDFs at passage 4 and HKs at passage 2 wereused in the experiment.

The procedure for creating a 3D dermal equivalent isdivided into three stages as shown in Figure 2. At stage 1(S1), HDFs were trypsinized, counted, and plated at 50,000per cm2 into six-well plates (well diameter 35 mm). In ad-dition, 100,000 HDFs were seeded into either PLGA-cmeshes or CHA and placed centrally into each HDF-platedwell. The control group consisted of plated HDFs only(HDFsheet). A silicon-coated glass slide was used to weighdown each scaffold onto the bottom of the plate. The cul-tures were maintained in 4 mL DMEM supplemented asbefore, with addition of 50 �g/mL L-ascorbic acid, over 2weeks [Fig. 2(a) and (b)]. At stage 2 (S2), cell sheets werepeeled off the culture plates using forceps [Fig. 2(c)]. Thescaffolds were embedded and integrated within the cellsheets, which could support the scaffolds on their own [Fig.2(d)]. The cell sheets were then folded over the scaffolds(PLGA-c or CHA) from four sides to form a 3D dermalequivalent [Fig. 2(e)–(h)]. In the HDFsheet control group,each cell sheet was folded onto itself to form a stacked HDFconstruct. All constructs were cultured for 1 week. At stage3 (S3), 200,000 HKs per cm2 were seeded onto the dorsalsurfaces of each specimen and cocultured for 1 week [Fig.2(i)] in keratinocyte growth medium made up of low cal-cium (0.08 mM) DMEM-Ham’s F-12 medium (3:1), supple-

426 NG ET AL.

mented with: 10% FBS, 1% Penicillin-streptomycin, 10ng/mL EGF, 5 �g/mL Insulin, 0.4 �g/mL Hydrocortisone,10�10 M Cholera Toxin, 20�12 M Triiodothyronine, 0.18 mMAdenine, and 5 �g/mL Transferrin (all supplements pur-chased from Sigma-Aldrich, St. Louis, MO). Hereon, timepoints are denoted by the stage and day of culture; forexample, S1D14 denotes stage 1 day 14 and S3D7 denotesstage 3 day 7 of culture.

Cell viability and F-actin expression

Cell viability in the constructs was assessed qualitatively byfluorescent labeling, as described previously.9 Briefly, speci-mens were incubated for 15 min at 37°C with 2 �g/mL fluo-rescein diacetate (Molecular Probes Inc., Eugene, OR), and in20 �g/mL propidium iodide solution (Molecular Probes Inc.)

for 2 min at room temperature. Expression of F-actin wasanalysed to study cell alignment patterns within the scaffolds.Specimens were fixed in 3.7% formalin for 30 min at roomtemperature and then incubated with 200 �g/mL RNAse A(Sigma, St. Louis, MO) for 30 min at room temperature. Sub-sequently, incubation with 5 U/mL F-actin specific phalloidin(Alexa Fluo 488 phalloidin, Molecular Probes Inc., Eugene, OR)was carried out for 45 min at room temperature and counter-stained with 0.1 mg/mL propidium iodide solution for 2 minat room temperature. Specimens were viewed under a confocallaser microscope (Olympus IX70-HLSH100 Fluoview, Olym-pus, Tokyo; Japan).

PicoGreen� DNA quantification assay

Cell proliferation in the specimens (n � 3) was evaluatedusing the PicoGreen� DNA quantification assay (Molecular

Figure 1. Gross morphology of the two hybrid scaffolds used in the study: (a) knitted PLGA mesh with lyophilized collagen(PLGA-c), and (b) collagen-hyaluronic acid sponge (CHA). Scanning electron micrographs showed (c) homogenouslydistributed collagen network between PLGA yarns in PLGA-c specimens and (d) randomly arranged network of collagen andhyaluronic acid with inhomogeneous pore distribution in CHA specimens. Scale bars: (a,b) 0.25 cm; (c,d) 200 �m. [Colorfigure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

ASSIMILATING CELL SHEETS FOR DERMAL TISSUE ENGINEERING 427

Figure 2. A pictorial representation of the three-stage time-line protocol for culturing cell sheets with hybrid scaffolds toform dermal equivalents. Light microscopy images of (a) PLGA-c and (b) CHA specimens cultured with HDFs at S1D14,showing cells proliferating into the PLGA-c spaces and anchoring the CHA onto the culture well. (c) Cell sheets were peeledoff from culture plates at S1D14. (d) At this stage, scaffolds were integrated within the cell sheets as demonstrated. (e–h) Eachcell sheet was folded over a scaffold from four sides in sequence to obtain a dermal equivalent. (i) A schematic representationof the cross section of a bilayered skin equivalent after seeding of HKs onto the dermal equivalents at S2D7. [Color figure canbe viewed in the online issue, which is available at www.interscience.wiley.com.]

428 NG ET AL.

Probes Inc., Eugene, OR). Specimens were treated with 1-mLenzyme solution comprising of 0.25% trypsin (Hyclone,South Logan, UT), 0.1% collagenase I (Gibco, Grand Island,NY), and 0.1% hyalurodinase (Sigma, St. Louis, MO) for 12 hat 37°C to obtain a homogenous cell suspension. A volumeof 1 mL PicoGreen� working solution was added to 20-�Laliquots (1:50 dilution) of each sample in 1 mL DNase-freeTris-EDTA buffer, and incubated at room temperature for 5min. Fluorescence was then measured with a plate reader(Genios�, Tecan Group Ltd, Maennedorf, Switzerland) atexcitation and emission wavelengths of 485 and 535 nm,respectively, and corrected for fluorescence of reagentblanks. The amount of DNA was calculated by extrapolatinga standard curve obtained by running the assay with thegiven DNA standard.

AlamarBlue™ metabolic assay

The levels of specimen metabolic activity (n � 6) werequantified via the AlamarBlue™ assay (Biosource Int., Cam-arillo, CA). At each time point, culture medium was aspi-rated, and 2 mL complete medium with 10% AlamarBlue™reagent was added and incubated for 2 h at 37°C, 5% CO2.Absorbance was measured with a plate reader (Genios�) atwavelengths of 560 and 595 nm, together with blank controlsusing medium with reagent and medium only. Metabolicactivity was directly correlated to the percentage reductionof AlamarBlue™ reagent by the cells, calculated based onthe manufacturer’s protocol (www.biosource.com).

Transmission electron microscopy

Specimens for electron microscopy analysis were fixed in2.5 % gluteraldehyde for 4 h at 4°C, and postfixed in 1%osmium tetroxide for 2 h at room temperature. After dehy-dration through an ethanol series, specimens were embed-ded in Araldite in a flat embedding mold. Ultrathin sectionswere stained with 2.7% lead citrate and saturated uranylacetate and viewed on a JEOL 1010 transmission electronmicroscope at 100 kV acceleration voltage and spot size 1.

Histology and immunocytochemistry

In vitro and in vivo specimens were snap frozen in liquidnitrogen and stored at �80°C prior to cryo-sectioning toobtain 8 �m-thick sections. Masson’s Trichrome stainingwas carried using the manufacturer’s protocol (DakoCyto-mation, Glostrup, Denmark), to analyze cell and ECM dis-tribution. Expression of major dermal ECM proteins (colla-gen I and III), keratinocyte distribution (cytokeratin) andbasement membrane organization (laminin) were evaluatedthrough immunocytochemistry. Sections were air dried andstored at �20°C until use. Fixation with methanol at �20°Cfor 10 min was carried out before blocking with 10% goatserum (Gibco, Grand Island, NY) in phosphate-buffered sa-

line at room temperature for 30 min was performed. Mono-clonal primary antibodies used were: Mouse anti-HumanCollagen I (1:200, Chemicon Inc., Temecula, CA), Mouseanti-Human Collagen III (1:200, Chemicon Inc.), Mouse anti-Human Cytokeratin (1:200, DakoCytomation), Mouse anti-Human Laminin (1:200, DakoCytomation). Primary antibod-ies were incubated on the sections, in a wet chamber, for 16 hat 4°C. Secondary antibody staining was performed withhorseradish peroxidase-conjugated antimouse kit (DAKOEnvision�, DakoCytomation). Primary antibody specificitywas verified using native human skin sections, while nega-tive controls were done on specimen sections using the samestaining protocol, replacing the primary antibodies with an-tibody diluent. All sections were counterstained withhaematoxylin and mounted before viewing.

Reverse transcriptase-polymerase chain reaction(RT-PCR)

The mRNA expression of seven growth factors (KGF,bFGF, TGF-�, TGF-�1, VEGF, EGF, and IGF-1) was evalu-ated just prior to HK seeding (S2D7), to characterize theexpression profile of the dermal constructs, and at the end ofthe coculture period (S3D7) to characterize the expressionprofile of the bilayered skin constructs. Total RNA wasisolated from the specimens (n � 3) at those time points,using TRIzol (Gibco, Grand Island, NY) according to themanufacturer’s protocol and stored in 10 �L aliquots at�80°C. Two-step RT-PCR was done using cMaster RT Kitand cMaster 2.5� PCR MasterMix (Eppendorf, Hamburg;Germany), on Mastercycler Gradient 5331 (Eppendorf,Hamburg; Germany), at 60°C annealing temperature. ThecDNAs of �-actin, which was used as the housekeepinggene, were amplified using the primers 5-GATGAT-ATCGCCGCGCTCGTCGT-3 and 5-TGGGTCATCTTC-TCGCGGT-3 (356-kb product). KGF cDNAs were amplifiedusing the primers 5-AAGTAAAAGGGACCCAAGAGAT-GAAG-3 and 5-CAACAAACATTTCCCCTCCGTTG-3(247-kb product). Commercially available primer kits(Maxim Biotech. Inc., San Francisco, CA) were used foramplification of the other growth factor cDNAs. PCR prod-ucts were separated using 1.5% agarose gel electrophoresis.

Enzyme-linked immunosorbent assays (ELISA)

Culture media was collected from each sample (n �3)during routine media changes at various time points andstored in 2-mL aliquots at �80°C. ELISA kits specific forhuman growth factors were used according to manufactur-ers’ protocols and standards provided (bFGF, EGF, VEGF—Chemicon Inc.; KGF, TGF-�, TGF-1, IGF-1—R&D Systems,Minneapolis, MN).

Statistical analysis

All quantitative values are expressed as mean � standarddeviation. Statistical analyses were performed using analy-

ASSIMILATING CELL SHEETS FOR DERMAL TISSUE ENGINEERING 429

sis of variance (ANOVA) with Tukey’s method for multiplecomparisons (family error rate � 0.05), on Minitab� 14(Minitab Inc., State College, PA).

RESULTS

Gross morphological observations

Over the periods of S2 and S3, the greatest extent oflateral contraction was observed in HDFsheet con-trols, with size reductions of 80–90% observed. Con-traction of CHA specimens was minimal. PLGA-cspecimens appeared to have contracted as much asHDFsheet controls. However, instead of a lateral con-traction, PLGA-c specimens shrunk as a result of curl-ing of the meshes (Fig. 3).

Cell proliferation and metabolic profiles

Results from PicoGreen� assay showed that DNAamounts rose equally in all the specimen groups fromS1D7 to S1D14 [Fig. 4(a)], with no significant differ-ences seen between the groups. From S1D14 to S2D7,DNA amounts in PLGA-c and HDFsheet remainedsteady while that in CHA continued to increase, sig-nificantly exceeding the levels recorded in PLGA-cand HDFsheet (ANOVA, p � 0.01; Tukey’s method).

From S2D7 to S3D7, DNA amounts for CHA andHDFsheet dropped significantly while that in PLGA-cremained statistically unchanged (ANOVA, p � 0.05).At S3D7, specimen groups were ranked as CHA,PLGA-c, and HDFsheet in decreasing order of totalDNA amounts, from both HDFs and HKs (ANOVA,p � 0.01; Tukey’s method). DNA quantification resultscorresponded to cell viability assessment using confo-cal laser microscopy images. Viable cells were labeledgreen while nonviable cell nuclei and PLGA threadswere labeled red. High cell viability in all specimengroups was observed prior to cell sheet encapsulationat S1D14. Subsequently, a general increase in the num-ber of nonviable cells was observed at S3D7.

Results from the AlamarBlue™ assay showed sim-ilar metabolic activity profiles in all three specimengroups [Fig. 4(b)]. Metabolic activity peaked at S1D14just prior to detachment of cell sheet, and droppedsignificantly thereafter. Upon seeding of HKs, a rise inmetabolic activity was recorded. At S3D7, the meta-bolic activity of CHA specimens was significantlyhigher than the other two specimen groups (ANOVA,p � 0.01; Tukey’s method).

F-actin expression

Figure 5 shows images of specimens at S1D14, rep-resentative of the entire culture period. F-actin stress

Figure 3. Gross observation showing specimen shrinkage. Digital images are scaled to allow direct size comparisons. Thelargest extend of shrinkage due to lateral contraction was observed in HDFsheet specimens (80–90%) while contraction ofCHA specimens was minimal. PLGA-c specimens did not contract laterally but curled up instead. Scale bar: 1 cm. [Colorfigure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

430 NG ET AL.

fibers were labeled green while nuclei and PLGAthreads were labeled red. In PLGA-c specimens, HDFsexpressed defined stress fibers that were oriented to-wards the PLGA threads. Cells anchored onto thePLGA and stretched outwards to form a cellular net-work [Fig. 5(a)]. Stress fibers were also clearly ex-

pressed in CHA specimens and were oriented accord-ing to the alignment of HDFs [Fig. 5(b)]. Somerandomly aligned stress fibers were observed in HDF-sheet specimens detached from the culture wells [Fig.5(c)]. However, these were scattered and poorly ex-pressed due to the loss of mechanical support.

Figure 4. Analysis of cell proliferation and metabolic activity profiles. Data points indicate mean values � standard deviation.(a) Plot of DNA amount versus time points. Specimens were homogenized by enzymatic digestion, and aliquots of 20 �L used foranalysis using PicoGreen� DNA quantification assay. (*CHA reading was statistically higher than other groups, p � 0.01;**Differences between all readings are statistically significant, p � 0.01). Numbers in parentheses indicate approximate total cellnumbers (HDFs � HKs) in millions, at S3D7. The panel of fluorescent images shows viable cells labeled green, nonviable cell nucleiand PLGA threads labeled red, for all specimen groups at S1D14 and S3D7. Scale bar: 200 �m (b). Plot of percent AlamarBlue™reduction versus time points. (***CHA reading was statistically higher than other groups, p � 0.01). [Color figure can be viewedin the online issue, which is available at www.interscience.wiley.com.]

ASSIMILATING CELL SHEETS FOR DERMAL TISSUE ENGINEERING 431

Histology and immunocytochemistry

Folded cell sheets were evident in histological im-ages of all specimens at S3D7 (Fig. 6). The sheetsformed collagen-rich neo-tissue over the scaffolds,with dense HDF aggregates observed between foldedinterfaces. Neo-tissue agglomerates formed above thePLGA-c mesh, similar to that seen in HDFsheet, whilethinner neo-tissue layers concentrated on the periph-ery of the CHA scaffold. HDFs infiltrated both thePLGA-c and CHA scaffolds. However, the cellularnetwork within CHA was not homogenously distrib-uted. Thin keratin-rich layers were detected intermit-tently on the dorsal surfaces of the constructs. How-ever, distribution of keratin-rich layers was nothomogenous and did not exhibit the complete differ-entiated strata as in native epidermis.

Collagens I and III were expressed within the foldedcell sheets in all specimen groups but was not presentor only weakly expressed in the dense HDF aggre-gates at the folded interfaces, within the PLGA-c meshand within the loose cellular network of the CHAscaffold [Fig. 7(a)–(f)]. Cytokeratin-positive keratino-cyte layers of less than five cell layers thick wereevident on the dorsal surfaces of the specimens [Fig.7(g)–(i)]. Laminin expression was weak and notclearly defined between the keratinocyte layers andunderlying neo-tissue, suggesting only a prematurebasement membrane organisation [Fig. 7(j)–(l)].

Ultrastructures

Representative transmission electron microscopy mi-crographs presented showed that within the PLGA-cmeshes, HDFs were well attached onto the PLGA

threads [Fig. 8(a)]. Typically, a keratinocyte layer notmore than five cell layers thick was observed in the S3D7specimens. Squamous morphology similar to terminallydifferentiated keratinocytes was observed with desmo-somes along cell–cell junctions identified [Fig. 8(b)].Abundant collagen bundles with normal banding peri-odicity ( 67 nm), but of smaller bundle size compared tonative dermis, were observed within the folded cell sheetmass in all specimen groups [Fig. 8(c)]. No defined base-ment membranes were found.

Growth factor secretion

A representative gel picture for S2D7 is shown inFigure 9, while a summary of RT-PCR results is pre-sented in Table I. TGF-�1 cDNA was detected in allthe specimen groups at both the time points assessed(S2D7 and S3D7). KGF cDNA was similarly detectedin all the specimens except in HDFsheet at S2D7.bFGF, TGF-�, VEGF, EGF, and IGF-1 cDNAs were notdetected at all.

Histograms of mean growth factor secretion levelsdetected in the culture supernatants and basal levelsin the culture media are presented in Figure 10. Priorto the detachment of the cell sheets at S1, levels of KGFsecretion averaged between 75.5 � 9.3 pg/mL to142.6 � 6.6 pg/mL in all the specimen groups(ANOVA, p � 0.05). A sharp rise was detected atS2D7, with levels in all groups peaking between474.2 � 57.0 pg/mL to 594.7 � 106.2 pg/mL at S2D7(ANOVA, p � 0.05). This was followed by a sharpdrop, to a range of 14.4 � 4.5 pg/mL to 170.2 � 24.3pg/mL at S3D7. At S3D7, KGF secretion by CHAspecimens was significantly higher than that in the

Figure 5. Analysis of F-actin expression in S1D14 specimens. F-actin stress fibers were labeled green while nuclei and PLGAthreads were labeled red. (a) In PLGA-c specimens, cells anchored onto the PLGA threads stretched outwards to form acellular network, expressing stress fibers that were oriented towards the PLGA. (b) Stress fibers in CHA specimens wereoriented according to the alignment of the cells. (c) Within the detached cell sheets, stress fibers were scattered, and poorlyexpressed in most regions due to the loss of mechanical support. Scale bar: 20 �m. [Color figure can be viewed in the onlineissue, which is available at www.interscience.wiley.com.]

432 NG ET AL.

other two groups (Tukey’s method). KGF levels in theculture media were negligible.

TGF-� was detected only at S3D7, at 53.0 � 14.1ng/mL in CHA, 21.4 � 8.4 ng/mL in PLGA-c, and32.5 � 11.5 ng/mL in HDFsheet, with no significantdifference between the groups (ANOVA, p � 0.05).TGF-� levels in the culture media were negligible.

TGF-�1 secretion profiles were similar betweengroups, with recorded levels consistently higher thanin culture media. In S1, all specimen groups showedTGF-�1 secretion of between 0.94 � 0.04 ng/mL and

1.07 � 0.04 ng/mL, after deducting basal levels, withno significant difference (ANOVA, p � 0.05). At S2D7,TGF-�1 levels gradually dropped to a range of 0.52 �0.02 ng/mL to 0.69 � 0.07 ng/mL and then to below0.3 ng/mL at S3D7.

Secretion of VEGF by all specimen groups averagedaround 5 ng/mL during S1, with no significant changeor differences between the groups (ANOVA, p � 0.05).Mean secretion levels increased at S2D7 to 17.1 � 6.7ng/mL to 21.7 � 5.6 ng/mL. At S3D7, VEGF levels inPLGA-c and HDFsheet specimens did not change sig-

Figure 6. Histological analysis of S3D7 specimens, with Masson’s Trichrome stain (collagen—blue, cell nuclei—black, keratin—red). (a,b) Folded cell sheets formed collagen-rich neo-tissue over PLGA-c specimens, with dense cell aggregates between foldedinterfaces. Thin keratinocyte layers were observed on dorsal surfaces of the constructs. (c,d) Thinner neo-tissue layers were seenon the peripheral of CHA specimens, with sparse cellular network within. (e,f) Collagen-rich neo-tissue and keratinocyte layerswere observed in HDFsheet specimens, similar to that in PLGA-c. Scale bars: (a,c,e) 100 �m; (b,d,f) 50 �m. [Color figure can beviewed in the online issue, which is available at www.interscience.wiley.com.]

ASSIMILATING CELL SHEETS FOR DERMAL TISSUE ENGINEERING 433

nificantly but surged in CHA specimens to 31.7 � 10.1ng/mL at S3D7. VEGF levels in the culture mediawere negligible.

No appreciable amounts of bFGF, EGF, and IGF-1were detected in all specimen groups throughout alltime points.

DISCUSSION

To overcome the issue of extensive contraction as-sociated with the use of cell sheets, we propose a novelmethod of culturing cell sheets with scaffolds. This

Figure 7. Immunocytochemistry analysis of S3D7 specimens. (a–f) Collagens I and III were expressed within the folded cellsheets in all specimens. However, collagen was undetected or only weakly expressed in the dense cell aggregates at cell sheetinterfaces, and within PLGA-c and CHA interiors (*). (g–i) Cytokeratin-positive keratinocyte layers were evident on the dorsalsurfaces of the specimens. However, (j–l) only traces of laminin deposition was detected between the keratinocyte andunderlying neo-tissue. Scale bar: (a–f) 50 �m; (g–l) 20 �m. [Color figure can be viewed in the online issue, which is availableat www.interscience.wiley.com.]

434 NG ET AL.

method was demonstrated in this study to tissue en-gineer a dermal equivalent. The 3D dermal equiva-lents formed using PLGA-c and CHA were viable,metabolically active, expressed collagens I and III, andsecreted KGF, TGF-�1, and VEGF.

Conventionally, cell populations are expanded invitro through repeated cycles of cell harvesting viaprotease treatment, which have undesirable effects oncells.12,13 In our technique, cells were only subjected toone enzymatic treatment, that is, during cell isolation.Isolated cells were then cultured together with a suit-able scaffold for a period of time that is required toform a cell sheet. The resulting dermal equivalent isthus composed entirely of primary cells without anyother manipulation.

In the context of tissue engineering skin, the mate-rials used in this study have been well evaluated, withnatural and synthetic materials each having their ad-vantages and disadvantages.9 Natural materials suchas collagen and hyaluronic acid have been extensivelyused in a number of currently available tissue engi-neered skin products.14,15 To our knowledge, the onlytissue engineered skin products that uses a syntheticdermal scaffold are Dermagraft�16 and Invitra CSS™

(Invitrx, Inc. www.invitrx.com). Both products use acommercial PLGA (10:90) mesh, Vicryl�, as the dermalscaffold. To assimilate the advantages of different ma-

terials, hybrid scaffolds of various combinations havebeen used.17,18 Nonetheless, the search for an idealscaffold for skin tissue engineering goes on.1

In comparison to the Dermagraft� and InvitraCSS™, our approach has a few advantages. First, theabove two products used a warp-knitted mesh,whereas we have used a weft-knitted PLGA mesh,which has greater 3D volume19 and has been shown ina separate study to exhibit tensile strength (18.4 � 2.5MPa) and elongation at break (78.0 � 5.4%) that ischaracteristic of native skin.11 Second, we have freeze-dried collagen into the PLGA mesh to enhance scaf-fold biocompatibility11 and to create a microporousnetwork which bridged the large-pore spaces withinthe mesh to enhance cell distribution. Third, by cul-turing the scaffold in combination with a fibroblastsheet, we created a neo-tissue with more maturedECM, compared to simply culturing fibroblast withinthe scaffold, without cell sheets.9 However, in ourstudy, the PLGA-c meshes used curled up duringculture. This was in contrast to our earlier study inwhich PLGA-polycaprolactone (PCL) meshes did notcontract over culture, for two reasons.9 First, dense

TABLE ISummary of RT-PCR Results

TimePoint

Gene

PLGA-c CHAHDFsheet(control)

S2D7 S3D7 S2D7 S3D7 S2D7 S3D7

bFGF � � � � � �KGF � � � � � �TGF-� � � �TGF-�1 � � � � � �VEGF � � � � � �EGF � � � � � �IGF-1 � � � � � ��-actin � � � � � �

� positive expression.� negative expression.Shaded box: RT-PCR not performed.

Figure 8. Ultrastructural analysis of S3D7 specimens. (a) Fibroblasts (F) were well attached onto the PLGA threads (P). (b)Desmosomes were expressed along cell–cell junctions between adjacent keratinocytes (arrow). (c) Collagen bundles withnormal banding periodicity were observed in the dermal components of all specimen groups. Scale bars: (a) 0.5 �m; (b,c) 0.2�m.

Figure 9. Representative gel picture of S2D7 RT-PCR re-sults. TGF-�1 cDNA was consistently detected in all groupswhile KGF cDNA was detected in all groups except HDF-sheet. All other growth factors were not detected. Lanes 1, 2:PLGA-c; Lanes 3, 4: CHA; Lanes 5, 6: HDFsheet.

ASSIMILATING CELL SHEETS FOR DERMAL TISSUE ENGINEERING 435

ECM was not formed in the previous study in whichcell sheets were not used, resulting in relatively lowercontractive forces exerted on the scaffolds. Second, thePCL filaments enhanced the ability of the previousmeshes to resist contraction. In this study, PCL wasremoved to minimize the amount of material used,while maintaining adequate mechanical properties.11

As a result, the contraction of the dense cell sheets onone side of the meshes, coupled with the loss of me-chanical properties as a result of PLGA degradation,which begins after 2 weeks in aqueous environmentand completes after 3 months in vivo,20 caused them tocurl up as shown. Nonetheless, our recent studyshowed that this shortcoming could be overcome byusing larger PLGA meshes and modifying the cellsheet folding technique (unpublished data).

When cell sheets were detached from the culturevessels, they began to contract and exert contractileforces on the scaffolds.21 In a contracted scaffold, fi-broblasts were shown to undergo apoptosis.22 Thiscould explain why there was no further increase in cellproliferation in PLGA-c and HDFsheet specimens. Be-cause CHA scaffolds could withstand the contractionforces, they present the cells with larger contact sur-face areas, as well as more volume for cell prolifera-tion, resulting in the continuous rise in DNA amounts

registered. Studies have shown that wound fibroblastsunderwent apoptosis after completion of repair.23 Thiscould explain the overall drop in DNA amounts re-corded in all the specimens at S3D7, as neo-tissuematured, despite the addition of HKs. However, theeffect of tissue necrosis, due to pH changes as scaf-folds degrade, was also possible. The evaluation ofapoptosis and necrosis in such a construct could be ofinterest in future studies.

AlamarBlue™ is a nontoxic metabolic indicator,which acts as an oxygen substitute in the electrontransport chain.24 When used on 3D constructs such asin this study, the assay measures the gross metabolicactivity levels of the specimens, and do not correlate tothe number of viable cells.25 The detachment of cellsheets caused a drastic drop in metabolic activity ascells lost their anchorage and went into growth ar-rest.26 Only cells within the scaffolds maintained theiranchorage, as shown by F-actin expression (Fig. 5),and therefore remained metabolically active. With theintroduction of HKs in S3, metabolic activity increasedpossibly because dermal fibroblasts and keratinocytesstimulate each other through a series of paracrinesignaling pathways.27 The higher metabolic activitylevels registered in CHA over S3 was because more

Figure 10. Results of growth factor secretion detected via ELISA. Data points indicate mean values � standard deviation.The range of basal levels of growth factors detected in culture media with supplements only are indicated by dotted (. . ..HDFmedia) and dashed (—-HK-HDF media) lines. Basal levels were negligible when not indicated.

436 NG ET AL.

HKs were seeded, due to the larger surface areasavailable on CHA specimens at the point of seeding.

Our preliminary growth factor expression analysisis comparable to other tissue engineered dermal con-structs. Dermagraft� has been shown to secrete ap-proximately 1 ng/mL VEGF and 500 pg/mL TGF-�1after 3 and 5 days of recovery from cryopreservation,respectively.28 This translates into 0.41–1.03 ng/mLVEGF and 207–517 pg/mL TGF-�1 per 106 cells. Incomparison, our dermal equivalents produced 0.53–1.76 ng/mL VEGF and 378–515 pg/mL per 106 cellsafter 7 days in culture, taking mean values from allthree specimen groups. (Cell number approximationswere obtained using an estimate of 6.57 pg genomicDNA per diploid human cell � 3 � 109 base pairs perhaploid genome, MR � 660 per base pair29). Further-more, mRNA expression for KGF, EGF, TGF-�, andbFGF were marginal in Dermagraft�, quoted as lessthan three copies per cell.28 Apligraf�, a bilayered skinequivalent, has also been reported to express mRNAof all the growth factors targeted in this study.30 Wehypothesize that the mRNAs for TGF-� and VEGFwere transiently expressed and therefore not detectedby RT-PCR at the sampling time points, even thoughboth were detected by ELISA. In HDFsheet at S2D7,mRNA for KGF was not detected possibly due to anexperimental error. Low calcium media was used inthe coculture of HKs and HDFs in this study to en-courage HK proliferation over the dermal equivalent.Organotypic coculture at the air–liquid interfacewould need to be carried out in future studies toinduce keratinocyte differentiation to obtain the dif-ferentiated epidermal strata.

In summary, our approach of assimilating cellsheets with 3D scaffolds could overcome the issue ofextensive contraction when using cell sheets alone.This technique could be used to tissue engineer adermal-like tissue of specific shape and size. In addi-tion, the method abrogates the issue of cell seedingefficiency because those cells that fall to the bottom ofthe vessel will contribute towards the eventual forma-tion of the cell sheet, and therefore are not lost orwasted. The undesirable effects of repeated enzymatictreatment cycles carried out in conventional cell ex-pansion protocols will also be circumvented, as cellswill be allowed to proliferate and mature over a pro-longed period of time to form neo-tissue. From aclinical point of view, this technique might offer newavenues for autologous, scaffold-based skin tissue en-gineering.

The authors would like to thank Dr Michael Raghunathfor his advice on immunocytochemistry work; Dr. StephenFeinberg and Dr. Ricky R Lareu for reviewing the manu-script; Dr. Achuth HN, Dr. Lu Jia, and Dr Moochhala S fromthe Defence Medical and Environment Research Institute fortheir technical support; and Miss Deborah Loh from the

Electron Microscopy Unit for her technical assistance. Partsof this paper have been presented at the 6th Meeting of theTissue Engineering Society International, Orlando, FL, Dec10–13 2003 and the 7th World Biomaterials Congress, Syd-ney, Australia, May 17–21, 2004.

References

1. Hutmacher DW, Vanscheidt W. Matrices for tissue-engineeredskin. Drugs Today 2002;38:113–133.

2. Yamato M, Okano T. Cell sheet engineering. Mater Today2004;May:42–47.

3. L’Heureux N, Paquet S, Labbe R, Germain L, Auger F. Acompletely biological tissue-engineered human blood vessel.FASEB J 1998;12:47–56.

4. Michel M, L’Heureux N, Pouliot R, Xu W, Auger F, Germain L.Characterization of a new tissue-engineered human skinequivalent with hair. In Vitro Cell Dev Biol 1999;35:318–326.

5. Yamada N, Okano T, Sakai H, Karikusa F, Sawasaki Y, SakuraiY. Thermo-responsive polymeric surfaces; Control of attach-ment and detachment of cultured cells. Macromol Rapid Com-mun 1990;11:571–576.

6. Kwon OH, Kikuchi A, Yamato M, Sakurai Y, Okano T. Rapidcell sheet detachment from poly(N-isopropylacrylamide)-grafted porous cell culture membranes. J Biomed Mater Res2000;50:82–89.

7. Shimizu T, Yamato M, Kikuchi A, Okano T. Cell sheet engi-neering for myocardial tissue reconstruction. Biomaterials2003;24:2309–2316.

8. Yamato M, Utsumi M, Kushida A, Konno C, Kikuchi A, OkanoT. Thermo-responsive culture dishes allow the intact harvest ofmultilayered keratinocyte sheets without dispase by reducingtemperature. Tissue Eng 2001;7:473–480.

9. Ng KW, Khor HL, Hutmacher DW. In vitro characterization ofnatural and synthetic dermal matrices cultured with humandermal fibroblasts. Biomaterials 2004;25:2807–2818.

10. Rehakova M, Bakos D, Vizarova K, Soldan M, Jurickova M.Properties of collagen and hyaluronic acid composite materialsand their modification by chemical crosslinking. J BiomedMater Res 1996;30:369–372.

11. Ng KW, Louis J, Ho BST, Achuth HN, Lu J, Moochhala S, LimTC, Hutmacher DW. Characterization of a novel bioactivepoly(lactic-co-glycolic acid) and collagen hybrid matrix for der-mal regeneration. Polym Int. Forthcoming.

12. Okano T, Yamada N, Sakai H, Sakurai Y. A novel recoverysystem for cultured cells using plasma-treated polystyrenedishes grafted with poly(N-isopropylacrylamide). J BiomedMater Res 1993;27:1243–1251.

13. Schaefer BM, Reinartz J, Bechtel MJ, Inndorf S, Lang E,Kramer MD. Dispase-mediated basal detachment of cul-tured keratinocytes induces urokinase-type plasminogen ac-tivator (uPA) and its receptor (uPA-R, CD87). Exp Cell Res1996;228:246 –253.

14. Kirsner RS, Falanga V, Eaglstein WH. The development ofbioengineered skin. Trends Biotechnol 1998;16:246–249.

15. Jones I, Currie L, Martin R. A guide to biological skin substi-tutes. Br J Plast Surg 2002;55:185–193.

16. Naughton G, Mansbridge J, Gentzkow G. A metabolically ac-tive human dermal replacement for the treatment of diabeticfoot ulcers. Artif Organs 1997;21:1203–1210.

17. Tiwari A, Salacinski HJ, Punshon G, Hamilton G, Seifalian AM.Development of a hybrid cardiovascular graft using a tissueengineering approach. FASEB J 2002;16:791–796.

18. Tsuchiya K, Mori T, Chen G, Ushida T, Tateishi T, MatsunoT, Sakamoto M, Umezawa A. Custom-shaping system for

ASSIMILATING CELL SHEETS FOR DERMAL TISSUE ENGINEERING 437

bone regeneration by seeding marrow stromal cells onto aweb-like biodegradable hybrid sheet. Cell Tissue Res 2004;316:141–153.

19. Walker RP, Adanur S. Knitting. In: Adanur S, editor. Welling-ton Sears handbook of industrial textiles. Lancaster, PA: Tech-nomic Publishing Company Inc.; 1995. p 127–132.

20. Miller RA, Brady JM, Cutright DE. Degradation rates of oralresorbable implants (polylactates and polyglycolates): Ratemodification with changes in PLA/PGA copolymer ratios.J Biomed Mater Res 1977;11:711–719.

21. Majno G, Gabbiani G, Hirschel BJ, Ryan GB, Statkov PR. Con-traction of granulation tissue in vitro: Similarity to smoothmuscle. Science 1971;173:548–550.

22. Fluck J, Querfeld C, Cremer A, Niland S, Krieg T, Sollberg S.Normal human primary fibroblasts undergo apoptosis inthree-dimensional contractile collagen gels. J Invest Dermatol1998;110:153–157.

23. Desmouliere A, Redard M, Darby I, Gabbiani G. Apoptosismediates the decrease in cellularity during the transition be-tween granulation tissue and scar. Am J Pathol 1995;146:56–66.

24. Nociari MM, Shalev A, Benias P, Russo C. A novel one-step,highly sensitive fluorometric assay to evaluate cell-mediatedcytotoxicity. J Immunol Methods 1998;213:157–167.

25. Ng KW, Leong DTW, Hutmacher DW. The emerging challengeto measure cell proliferation in three dimensions. Tissue Eng2005;11:182–191.

26. Folkman J, Moscona A. Role of cell shape in growth control.Nature 1978;273:345–349.

27. Goulet F, Poitras A, Rouabhia M, Cusson D, Germain L, AugerFA. Stimulation of human keratinocyte proliferation throughgrowth factor exchanges with dermal fibroblasts in vitro.Burns 1996;22:107–112.

28. Mansbridge JN, Liu K, Pinney RE, Patch R, Ratcliffe A, Naugh-ton GK. Growth factors secreted by fibroblasts: Role in healingdiabetic foot ulcers. Diabetes Obes Metab 1999;1:265–279.

29. Morton NE. Parameters of the human genome. Proc Natl AcadSci USA 1991;88:7474–7476.

30. Hardin-Young J, Parenteau NL. Bilayered skin constructs. In:Atala A, Lanza RP, editors. Methods of tissue engineering. SanDiego, CA: Academic Press; 2002. p 1177–1188.

438 NG ET AL.