Differentiation 84 (2012) 305–313

Contents lists available at SciVerse ScienceDirect

Differentiation

0301-46

Join the

http://d

Abbre

referenc

referenc

media in Corr

IKE/KEF

Tel.: þ4

E-m

journal homepage: www.elsevier.com/locate/diff

Interpreted gene expression of human dermal fibroblasts afteradipo-, chondro- and osteogenic phenotype shifts

Jonathan Rakar a,b,n, Susanna Lonnqvist a, Pehr Sommar a, Johan Junker a,c, Gunnar Kratz a,b,d

a Experimental Plastic Surgery, Department of Clinical and Experimental Medicine, Linkoping University, Swedenb Center for Integrative Regenerative Medicine (IGEN), Department of Clinical and Experimental Medicine, Faculty of Health Sciences, Linkoping University, Swedenc Division of Plastic Surgery, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, USAd Department of Plastic Surgery, County of Ostergotland, Linkoping, Sweden

a r t i c l e i n f o

Article history:

Received 10 February 2012

Received in revised form

30 July 2012

Accepted 19 August 2012Available online 27 September 2012

Keywords:

Fibroblasts

Phenotype

Plasticity

Media induction

Differentiation

81/$ - see front matter & 2012 International

International Society for Differentiation (ww

x.doi.org/10.1016/j.diff.2012.08.003

viations: A-FB, adipogenic media induced;

e group; C-FB, chondrogenic media induced;

e group; FB, human dermal fibroblasts (cont

nduced; FB OB, human osteoblast reference g

esponding author at: Department of Clinical

9, Faculty of Health, Linkoping University

6 (0)10 1034485; fax: þ46 (0)13 127465.

ail address: jonathan.rakar@liu.se (J. Rakar).

a b s t r a c t

Autologous cell-based therapies promise important developments for reconstructive surgery. In vitro

expansion as well as differentiation strategies could provide a substantial benefit to cellular therapies.

Human dermal fibroblasts, considered ubiquitous connective tissue cells, can be coaxed towards

different cellular fates, are readily available and may altogether be a suitable cell source for tissue

engineering strategies. Global gene expression analysis was performed to investigate the changes of the

fibroblast phenotype after four-week inductions toward adipocytic, osteoblastic and chondrocytic

lineages. Differential gene regulation, interpreted through Gene Set Enrichment Analysis, highlight

important similarities and differences of induced fibroblasts compared to control cultures of human

fibroblasts, adipocytes, osteoblasts and articular chondrocytes. Fibroblasts show an inherent degree of

phenotype plasticity that can be controlled to obtain cells supportive of multiple tissue types.

& 2012 International Society of Differentiation. Published by Elsevier B.V. All rights reserved.

1. Introduction

The requirement of cells, albeit alleviated by the use ofbiocompatible scaffolds, is often a limiting factor in tissuereconstruction. Research on adult stem cells has led to somesubstantiation of the hope of obtaining autologous tissue-specificcells through techniques involving cellular differentiation. Intheory, an autologous stem cell approach can provide the neces-sary cells required for any tissue reconstruction while avoidingthe problems associated with graft rejection and immunosup-pressive treatment. So far, such methods have only been appliedclinically in very limited fashion, involving only a few cell typesand with varying degree of success (Lazarus et al., 1995; Horwitzet al., 1999; Krampera et al., 2007). Among the impediments tousing adult stem cells are difficulties associated with their harvestand culture. More importantly, the validation of differentiatedphenotypes required for quality control is problematic. A better

Society of Differentiation. Publish

w.isdifferentiation.org)

FB AC, human adipocyte

FB CC, human chondrocyte

rol group); O-FB, osteogenic

roup

and Experimental Medicine,

, 58185 Linkoping, Sweden.

understanding of the inherent variations in phenotypic states ismotivated and we approach this by looking broadly at theexpression profiles of fibroblasts, three differentiated fibroblasttypes and the three wild-type targets.

Fibroblasts (FB) can be obtained through small skin biopsies,greatly expanded using standardized procedures, and are routi-nely used clinically. Dermal FB may be induced to differentiatetowards adipocytes (AC), osteoblasts (OB) and chondrocytes (CC)using defined induction media (Junker et al., 2010; Sommar et al.,2010; Toma et al., 2001; Bartsch et al., 2005; Lorenz et al., 2008).These features make FB appealing candidates for use in tissueengineering applications. Nonetheless, current biological classifi-cation of the FB cell type is still equivocal (Sorrell and Caplan,2009). Cellular phenotypes are indirectly defined by their parti-cular pattern of regulated gene expression, and differentiationinvolves switching from one pattern of gene expression toanother (Ben-Tabou de-Leon and Davidson, 2007). Certain tran-scription factors included in our analysis, key regulating tran-scription factors, are required for the development of specifictissues and control phenotype-defining expression (Lander et al.,2001; Pei, 2009). Microarrays were used to measure global geneexpression and the results further interpreted through gene-setenrichment analysis (GSEA). Gene Ontologies and pathways foundthrough GSEA identify putative biological attributes of the inducedphenotypes and we demonstrate a basis for a broad inherentphenotype plasticity of FB at the level of gene expression.

ed by Elsevier B.V. All rights reserved.

J. Rakar et al. / Differentiation 84 (2012) 305–313306

2. Materials and methods

2.1. Cell sources and reference cultures

Human dermal FB and preadipocytes were isolated from dis-carded tissue obtained from routine reduction abdomenoplasty.Skin was obtained from both men and women aged between 40and 65 years of age without any reported skin disease. Tissues wereprocessed within 2 hours of surgery and cell isolation was per-formed as previously described (Junker et al. 2010; Entenmann andHauner 1996). FB cultures were maintained in Dulbecco’s ModifiedEagle�s Medium (DMEM) with 10% fetal calf serum (FCS), 50 U/mlpenicillin and 50 mg/ml streptomycin (PEST). Preadipocytes weremaintained in DMEM/Ham’s F12 (1:1) with 10% FCS and PEST,until differentiated to adipocytes (AC) using an adipogenic medium(Pittenger et al., 1999).

Human OB were obtained from tissue discarded after total hiparthroplasty. Bone was minced and incubated for 45 min in 1 mg/ml collagenase (type 2, Invitrogen). Cells were expanded in theDMEM supplemented with 10% FCS, PEST, 1 mM dexamethasone,50 mM A2P and 10 mM b-glycerophosphate (BGP) (Jaiswal et al.,1997). Human articular CC were isolated from the discardedtissue following total knee arthroplasty as described byPettersson et al., 2009. Cultures were maintained in the DMEMsupplemented with 10% FCS, PEST, 10 mM 2-[4-(2-hydroxyethyl)-piperazin-1-yl]ethanesulfonic acid (HEPES), 0.1 mM MEM non-essential amino acid solution, 0.4 mM L-proline, and 0.2 mMascorbate-2-phosphate (A2P) (Johnstone et al., 1998).

Cells used in the described experiments were cultured in poly-styrene cell culture flasks at 37 1C, 5% CO2 and 95% humidity. Cultureexpansion at ratios between 1:3–1:5 were employed using anethylene-diamine-tetraacetic acid (EDTA) (0.01%)/Trypsin (0.125%)solution at 37 1C for approximately 15 min at 70–90% confluency.Cells between passages P4 and P8 were included in the experiments.Human material was handled in accordance with ethical standards atthe University Hospital of Linkoping, Sweden.

2.2. Induction of fibroblast differentiation

The isolated FB were passaged at least three times before theexperiments were started. This practice ensures that contaminatingkeratinocytes, microvascular endothelial cells and melanocytes arenegligible in the cultures. Replicate FB cultures were subjected toinduction media (Table 1) for four weeks with media changed every48–72 h. Control FB cultures were maintained in parallel throughoutthe experiments. Time points for staining, immunohistochemistry(IHC) and semi-quantitative real-time reverse-transcriptase PCR(qRT-PCR) were one (TP1), two (TP2), three (TP3) and four (TP4)weeks of induction. All experiments were repeated three to eighttimes with cells from different donors. Induced FB are referred to asA-FB, C-FB and O-FB (Table 1).

2.3. Histology and immunohistochemistry

Initial analyses of induced cultures were performed accordingto previously published methods (Junker et al., 2010). Briefly,

Table 1Induction media used for the differentiation of fibroblast cultures. PEST¼penicillin

methylxanthine, A2P¼ascorbate-2-phosphate, BGP¼beta-glycerophosphate, and TGF-b

Inductionmedium

Designation ofinduced group

Formulation

Adipogenic A-FB DMEM, 10% FCS, PEST,1 mM DEX, 0.5 m

Osteogenic O-FB DMEM, 10% FCS, PEST,0.1 mM DEX, 50

Chondrogenic C-FB DMEM, 1% FCS, PEST,1.125 mM insulin

A-FB cultures were stained for intracellular lipid accumulationusing Oil Red-O (#O0625, Sigma, Stockholm, Sweden)(Blanchette-Mackie et al., 1995) and analyzed using antibodiestowards perilipin A (dilution 1:200; #P1998, Sigma), lipoproteinlipase (LPL) (dilution 1:100; #SAB1401231, Sigma) and peroxi-some proliferator/activator gamma (PPARg) (dilution 1:50;#AV32880, Sigma). C-FB were stained for glucoseaminoglycanswith Alcian Blue stain (#A3157, Sigma) (Lev and Spicer, 1964;Mason, 1971) and analyzed using antibodies towards aggrecan(dilution 1:50; #AB3778, Abcam, UK), collagen II (dilution 1:100;#MAB8887, Abcam), and Sox9 (dilution 1:50; #HPA001758,Sigma). O-FB were stained for calcified matrix using von Kossastaining (Bills et al., 1974) and Alizarin Red (Dahl, 1952) (#A5533,Sigma) and analyzed using antibodies towards osteocalcin (dilu-tion 1:500; #AB1857, Chemicon, CA) (Beresford et al., 1984),osteonectin (Malaval et al., 1994) (dilution 1:1000; #AB1858,Chemicon) and Runx2 (dilution 1:50; #WH0000860M1, Sigma).Primary antibodies were detected using AlexaFluor 488 (green) or546 (red) conjugated secondary antibodies (dilution 1:500; Invi-trogen). Cell nuclei were visualized using 4’,6-diamidino-2-phenylindole (DAPI) either through sample mounting in ProLongGold DAPI (Invitrogen), or using a solution of 300 nM DAPI(Invitrogen). The antibodies used in this study have previouslybeen tested for cross-reactivity and specificity (Junker et al.,2010). Images were captured using BX41 (40X/0.75, 20X/0.50)or IX51 (20X/0.40) light/fluorescence microscopes (Olympus,Sweden) fitted with DP70 cameras and appropriate fluorescenceexcitation/emission filters. Controls included positive referencecultures and omission of primary antibodies.

2.4. RNA isolation

RNA was isolated from discrete replicate cultures, each samplecontaining between 0.5 and 8�106 cells at each TP. RNA wasisolated from AC after preadipocyte differentiation, OB and CCafter 14 days (TP2). Cells were detached by trypsin–EDTA treat-ment and washed in phosphate-buffered saline solution (PBS)(centrifuged at 200 g for 5 min). RNA was isolated using theNucleoSpin RNA II kit (Macherey-Nagel GmBH, Germany) accord-ing to manufacturer’s protocol (rev. 10, 2009). RNA was eluted in60 ml dH2O and checked spectrophotometrically in a NanoDrop2000 (Thermo Scientific, MA) to determine purity and yield.Initial samples were assessed by the Bioanalyzer 2100 (AgilentTechnologies Inc., CA) and all samples had RNA integrity numberabove 9.0. Samples were stored at -70 1C for up to threemonths and re-measured with the NanoDrop 2000 before use inexperiments.

2.5. Relative quantification with real-time PCR (qRT-PCR)

Reverse transcription was carried out using a High CapacityRNA-to-cDNA Synthesis Kit (Applied Biosystems, CA), accordingto manufacturer’s protocol but modified linearly for 50 ml reac-tions. 1 mg RNA was reverse transcribed in each sample. Hydro-xymethylbilane synthase (HMBS) was selected as endogenouscontrol based on candidate reference gene testing (TaqMan Express

(50 U/ml) and streptomycin (50 mg/ml), DEX¼dexamethasone, IBMX¼ isobutyl-

1¼transforming growth factor beta 1.

Reference

M IBMX, 10 mM insulin, 200 mM indomethacin (Pittenger et al., 1999)

mM A2P, 10 mM BGP (Jaiswal et al., 1997)

, 50 nM A2P and 10 ng/ml TGF-b1 (Johnstone et al., 1998)

J. Rakar et al. / Differentiation 84 (2012) 305–313 307

Human Endogenous Control Plate, Applied Biosystems). qRT-PCRwas performed using Fast Universal PCR MasterMix, fast 96-wellplates and the ABI 7900HT Fast Real-Time PCR system (AppliedBiosystems) with primer assays (TaqMan Gene Expression Assays20X, Applied Biosystems) containing buffered forward and reverseprimers with TaqMan minor groove binder (MGB) reporter-quencher probes. Exon-spanning primers were selected to avoidcontaminant genomic DNA amplification. For a list of primer–probe assays see Supplementary Table SI. Relative quantificationwas carried out using the DDC(T) method (Livak and Schmittgen,2001) and results are presented as fold changes relative to controlFB. Each sample was used in technical triplicates or quadruplicates,and at least six replicate cultures were analyzed at each time-point.Statistical analysis and graphing was done in Prism 5 (GraphPadSoftware, CA) and significance was tested by the Kruskal–Wallisrank-sum test with Dunn’s Multiple Comparison post-test. P valuesbelow 0.05 were considered statistically significant.

2.6. Microarray

Full gene expression was analyzed at TP4 using GeneChip HumanGene 1.0ST arrays (Affymetrix, UK). The procedure was carried outaccording to manufacturer’s protocol (Affymetrix GeneChip WT STLabeling Assay Manual rev 5, P/N 701880, with addendum rev 1,P/N 702577). A starting amount of 300 ng RNA from each samplewas used, and duplicate samples were used for each group. Aftersubsequent RNA-to-cDNA conversion steps each sample was hybri-dized to a separate array and the arrays were grouped according toculture medium. Raw data (.CEL-files) was imported into GeneSpringGX11 software (Agilent Technologies Inc., Santa Clara, US) andanalyzed using RMA16 (Irizarry et al., 2003) for the initial normal-ization. Probe sets with raw signals below the 20th percentile werefiltered out to minimize background as per manufacturer’s recom-mendation. Entity significance was tested using ANOVA with Tukey’sHonestly Significant Difference test and Benjamini–Hochberg correc-tions. Gene lists were further subjected to a fold-change cut-off of 2,relative to the FB group. Gene lists were constructed in GeneSpringand exported to the online tool WebGestalt (Zhang et al., 2005).WebGestalt is compatible with Affymetrix HuGene 1.0ST v1.0 probe-set IDs and was used to perform Gene-Set Enrichment Analysis(GSEA) on exported gene lists. Gene lists were automatically parsedby WebGestalt and only unique and unambiguous annotations wereretained. WebGestalt GSEA Toolbox settings were kept at default withthe significance threshold set to po0.05 for all analyses. GO termswere manually abbreviated to avoid higher-level terms that did notprovide additional information, and a sub-set of the terms are shownin the figures below. For additional results see supplementary file S2.Microarray dataset is available through Gene Expression Omnibus,with series accession number: GSE35545.

3. Results and discussion

3.1. Confirmation of induction

Typical phenotype markers were investigated to verify differ-entiation of fibroblasts upon media induction. Positivity of relevantstains and lineage markers (Fig. 1) verify induction at morphological,protein and gene levels on par with previous observations (Junkeret al., 2010; Sommar et al., 2010). The specific histological stainswere positive in all respective induced cultures, but negative forother specific stains (A-FB were positive for Oil Red-O, but negativefor Alizarin Red and Alcian Blue, and so on).The control FB cultures,regardless of length of culture (up to four weeks), were negative forall histological stains and lineage-specific antibodies. We broadenedour marker panel to include some verification of our microarray

findings as well as key regulating transcription factors PPARg, Sox9and Runx2-master regulatory genes for adipogenesis, chondrogen-esis and osteogenesis, respectively. Cultures from all donors behavedsimilarly and we found greater variation between discrete culturereplicates, regardless of donor origin, than between cultures fromdifferent donors (results not shown).

3.2. Differential gene expression overview

The effects of culturing are assumed to be similar in allreplicates and therefore comparisons between controls andinduced FB show differences attributable to culture environment,represented mainly by the differences in media composition.Refinement of these datasets can occur by including morereplicates under more conditions, something that is preferentiallydone in conjunction with further functional testing to directlycorrelate gene expression changes to functional phenotypicstates. Any quantification of differentiation is complicated byincomplete knowledge of the degree of phenotype variationinherent to, or allowed in, the definition of the reference celltypes as well as the starting population. Also, the inherited geneexpression background of reference cells that does not play a rolein their functional identity may be different to the tacit back-ground inherited by the induced fibroblasts. Given these limita-tions we opted to begin by looking at group correlations to gainan appreciation of the differences (Fig. 2A). Principal componentanalysis and hierarchical clustering (not shown) did not sub-stantiate our analyses further. Similarities of up-regulation eventsamong the references and among induced groups are shown inVenn diagrams of each grouping (Fig. 2B) compared to controlfibroblasts. Interpretations of the datasets using statistical meth-ods and GSEA in particular are discussed in the following sections.

Unfiltered baseline-to-median transformed array data was highlycorrelated between replicates of the same group (40.995;except AC: 0.981). The adipocyte (AC) group stood out withhigher intra-group variation as well as lower overall correlationwith other sample groups. The fibroblast controls were highlycorrelated with the induced fibroblast groups. Similarities wereobserved between A-FB and O-FB that could be explained byadipogenic and osteogenic media both being supplementedwith dexamethasone as well as reciprocal gene-regulatorycascades during adipogenesis and osteogenesis [reviewed in(Muruganandan et al., 2009)]. Expression changes in the C-FBgroup are largely events of down-regulation as evident whencomparing up-regulated genes in C-FB (64 genes) to down-regulated genes (251 genes, shown in Fig. 6A). These geneexpression tendencies may have been augmented by the lowerserum conditions (1% FCS) in chondrogenic cultures. Theemphasis on down-regulation events may be a further conse-quence of the culture environment, discussed further on. GSEAinterpretations of the datasets are detailed in the followingsections.

3.3. Adipogenic induction

Markers studied in A-FB indicate presence of functional lipidstorage regulating mechanisms and suggest adipogenesis in A-FB(Fig. 1). Fatty acid binding protein 4 (FABP4) and LPL are knowntargets of PPARg transcriptional activity (Tontonoz et al., 1994)and are up-regulated at both gene and protein levels. Fig. 3Ashows the number of genes up-regulated in A-FB and AC comparedto control FB, and their overlap. GSEA results (Fig. 3B) based on theoverlapping genes (Fig. 3C) represent numerous GO terms related tolipid storage and cellular lipid and fatty acid metabolic processes,supporting the observed protein expression data. Transcription factorForkhead box (FOX) A1 and FOXO4 targets are over-represented

Fig. 1. Marker expression of fibroblasts (FB) after media induced differentiation FB after adipogenic induction (A-FB); FB after osteogenic induction (O-FB); FB after

chondrogenic induction (C-FB); PCR: n¼6 (discrete samples, each based on four averaged per-plate technical replicates) showing geometric mean 795% CI. TP¼Time

point (weeks). *significant (po0.05) compared to fibroblasts. Top panels show light microscopy images of distinct features and stains. Middle panels show three

fluorescence images for each lineage (white arrowheads emphasize positive signals in the nuclei). Bottom panels show qRT-PCR gene expression results according to the

DDC(t) method, indicating fold-change compared to FB cultured in normal medium up to four weeks (TP1-4) and compared to wild-type controls (adipocytes (AC),

osteoblasts (OB), chondrocytes (CC)).

J. Rakar et al. / Differentiation 84 (2012) 305–313308

among the up-regulated genes. FOXA1 is connected to both CCAAT-enhancer/binding protein (CEBP) b and PPARg during adipogenesis(Fujimori and Amano, 2011), and was not itself found up-regulated.FOXO4 over-expression is known to induce glucose uptake andstimulate lipid droplet accumulation (Zhu et al., 2010), but it is notspecific to adipocytes. CEBPa, -b and -d target over-representation,together with pathways of adipogenesis, PPAR signaling, insulin

signaling and arachidonic acid metabolism, indicates that adipogeniccascades (Cao et al., 1991) are available in both A-FB and the ACreference group.

The 252 genes found in AC that were not up-regulated in A-FB,shed light on some differences between A-FB and AC. Additionalgenes in AC are involved in lipid turn-over and biosynthesis.There are also more genes related to insulin signaling and PPAR

signaling pathway as well as fatty acid, triglyceride and lipid

metabolism annotations. Glucose homeostasis and triglyceride

and cholesterol biosynthesis are annotations relevant for adipo-cytes that were not associated with the A-FB group. Similarly,glycerol and water channel activity and lipoprotein (LDL/VLDL)interactions represent genes that are not equally expressed inA-FB. Transcription factor GSEA did not yield significant resultsfor the 252 genes in AC.

Conversely, the genes up-regulated in A-FB but not AC repre-sent targets of numerous widely acting transcription factors suchas FOXO4, transcription factor 3 (TCF3), lymphoid enhancer-binding factor 1 (LEF1), nuclear factor of activated T-cells, cyto-plasmic, and calcineurin-dependent 1 (NFATC1) involved in amultitude of processes relating to differentiation, growth andimmunological actions. Interestingly, CEBPa and -b were alsorepresented, showing up-regulation of genes related to adipogen-esis that were not also up-regulated in AC. This indicates thatrelated but distinct regulatory cascades may be activated by theinduction treatment. Notable differences to AC include pathwaysof amino acid transport and metabolism, FGF signaling and cate-gories of platelet-attributed activity. Unique GO terms includewound healing, blood coagulation and IGFR signaling amongstothers, terms more typically associated to FB. The phenotypepresented in A-FB seems to have a more developed arsenal forlipid storage and a more pronounced state of energy regulationthan FB. While comparable to AC in this respect, A-FB seems lessdiverse in supporting metabolic pathways and secreted metabolicfactors, including adipocyte-related cascades. Similaritiesbetween A-FB and lipofibroblasts may exist in terms of lipidstorage mechanisms and should be investigated further by direct

Fig. 2. Overview of groups A: sample group correlations (averaged); B: Number

of genes (fold-change 42) higher than in fibroblasts. Comparisons are grouped

into induced fibroblast groups (A-, O-, C-FB¼Adipogenic, Osteogenic, Chondro-

genic media induced fibroblasts) and reference groups (AC¼adipocytes,

OB¼osteoblasts, CC¼chondrocytes).

J. Rakar et al. / Differentiation 84 (2012) 305–313 309

comparison to further understand the role and function of thedifferent cell types. If the homeostatic functions alluded to in ourresults are applicable to TE applications then A-FB may be apromising candidate for adipose tissue constructs.

3.4. Osteogenic induction

Expression of Runx2 as well as osteocalcin, osteonectin,osteopontin (SPP1) and osteoprotegerin (TNFRSF11B) was detectedat gene or protein levels in O-FB (Fig. 1). Master regulatory genesfound up-regulating below threshold in the microarray includedRUNX2 and activating transcription factor 4 (ATF4)-factors presentafter osteogenic lineage commitment of osteo–chondro progeni-tors (Rached et al., 2010). RUNX2 is reported to increase duringdifferentiation of osteoblasts and decreases with maturation(Komori 2010). The number of up-regulated genes in Fig. 4Ashows substantial overlap between O-FB and OB. GO termsassociated to the O-FB and OB gene overlap include ossification,regulation of bone mineralization, skeletal system development andresponsiveness to mechanical stimulus, which allude to activityimportant for osteoblasts (Fig. 4B). Many pathways seem relatedto signaling cascades involved in differentiation (JAK-STAT, AMPK,EGFR1), and others still are involved with the interleukin signalingand immunological modulation. The endochondral ossification

pathway complements the aforementioned GO terms. Four tran-scription factors were over-represented by the genes, includingtranscription factor Dp-1 (TFDP1) involved in cell cycle progres-sion (Ishida et al., 2005). The overlapping genes are listed in theheatmap (Fig. 4C).

The most prominent findings in OB that were not associated toO-FB relate to the GO terms cellular component organisation,chromatin assembly, nucleosome assembly and the molecular func-tions kinetochore, condensed chromosome, nuclear lumen, andnuclear matrix. Several genes coding for histone clusters (HIST)1, 2 and 4 were up-regulated. Pathways related to the cell cycle,E2F regulation and APC/C indicate strong growth regulatoryactivity in OB cultures. Important regulation of chromatin state,cell cycle and mitosis by Runx proteins and co-factors haveattributed important roles in osteoblast development (Benayahuet al., 2009). A more active state of genomic regulation than whatis found in FB or O-FB is suggested by these findings.

The annotations related to the O-FB, but not OB, genes mostlyimply ECM-related changes, including calcification. Because thesegenes are not found in OB the results suggest that the phenotypicstate of the O-FB is altered through different pathways to becomesupportive of an osteogenic environment. The O-FB phenotypeincludes functional features emulating osteogenic tissue, ratherthan completely differentiating into OB.

3.5. Chondrogenic induction

CC and C-FB cultures were maintained in static monolayersleading to a partial loss of the chondrocyte phenotype of CC andlimiting the C-FB phenotype induction. Mechanical stimulationand 3-dimensional culture are known to promote chondrocytedifferentiation, and SOX9 and ACAN up-regulation in particular(Thomopoulos et al., 2011), which we have previously observed in3-dimensional cultures of human articular CC as well as C-FBinduction in spinner flasks (Sommar et al., 2010). The relativesimplicity of monolayer culture, compared to 3-dimensionalculture with mechanical tension, is advantageous for routineculture expansion. Preliminary results indicated that we couldobtain sufficient data for expression analysis. C-FB in monolayersconsistently shows up-regulation of CC markers. However, adedifferentiation profile is evident from the relatively low expres-sion of COL2A1 and ACAN genes (Schnabel et al., 2002) in the CCcultures. Cartilage oligomeric protein (COMP), considered a nega-tive marker for dedifferentiation (Zaucke et al., 2001), is down-regulated in C-FB and expressed 5-fold lower in CC compared toFB. Chondrogenic induction of fibroblasts resulted in increasedexpression of aggrecan, collagen II and Sox9 proteins (Fig. 1)although the microarray findings show up-regulation below the2-fold cut-off for SOX9 and COL2A1.

Owing to the low number of overlapping genes and, moregenerally, the low number of up-regulated genes in C-FB (Fig. 5A)the use of GO terms and transcription factor GSEA providedlimited information about the C-FB phenotype (Fig. 5B). Besidesthe over-representation of targets of BTB and CNC homology 1(BACH1), a transcriptional repressor responsive to oxidative stressin cartilage (Ochiai et al., 2008), the information obtained fromGSEA largely relates to genes in adipo-related annotations. Similaradipo-related annotations were found for CC reference genesindicating relevance also for the CC phenotype. Several transcrip-tion factors (FOXO4, FOXA1, CEBPA, CEBPB etc) were found in CCindicating an increased metabolic activity and PPAR signalling

activity. Furthermore, several genes are involved in response to

chemical stimulation and carbohydrate and alcohol metabolic

processes.While the 14 genes shared between C-FB and CC provided

limited information (Fig. 5B), a few terms were found in C-FBalone. The GO terms extracellular region together with focal

adhesion pathways and the transcription factor zinc finger protein384 (ZNF384) activity suggests a change in matrix remodelingand cell–matrix interactions (Nakamoto et al., 2000). AMPKsignaling responds to changes in energy and indicates a switch

Fig. 3. Results of adipogenic induction and gene-set interpretation A: Venn diagram showing genes expressed higher in both adipogenic media induced fibroblasts

(A-FB) and adipocytes (AC) than in fibroblasts (FB); B: important gene-set enrichment analysis (GSEA) results for each set. Expanded lists are found in supplementary file S2.

J. Rakar et al. / Differentiation 84 (2012) 305–313310

towards catabolic pathways (Carling, 2004) condoning a predo-minantly down-regulating profile. FB are known for elaboratechanges in the ECM maintenance gene regulation. Matrix-relatedgenes fibromodulin (FMOD), biglycan (BGN) and fibronectin (FN) arehighly expressed in FB and not down-regulated in C-FB. Expressionof genes such as ADAM metallopeptidase with thrombospondintype 1 motif, 5 (ADAMTS5) and matrix metallopeptidase 13 (MMP13)furthermore suggest an active cartilage remodeling process.

3.6. Fibroblast gene expression

The FB phenotype is known to be variable over differentanatomical sites, during different physiological states and, moregenerally, in response to different environmental cues (Sorrell andCaplan, 2009). A major hurdle in the study of fibroblasts is thelack of uniquely distinguishing features between possible sub-types of fibroblastic cells, with a few notable exceptions. Exam-ples of established fibroblast types with distinguishing featuresinclude the myofibroblast and the lipofibroblast but even theseare not fully characterized fibroblastic cells. The analysis hereinsupports the view that FB are able to regulate important aspects oftheir functional identity and exist in a multitude of phenotypestates. This has been seen with the aforementioned myofibroblastphenotype in vivo but the natural existence of less distinctfibroblastic types has been more difficult to ascertain. The inducedFB cultures display a full range of cellular morphology fromfibroblastic to morphologies typical of the target phenotype. It isconceivable that the heterogeneity of the primary FB culturesresults in a broad spectrum of variably altered cells, leading toanalysis of a numerically average profile for each group. We havepreviously shown that flow-assisted cell sorting based on a panelof markers could identify differences between primary FB and

single-cell clones (notably increased cell-surface density of CD13,CD44 and CD105)(Junker et al., 2010), and different clonal popula-tions displayed somewhat diverse morphology. The resultinginduced phenotypes showed quantitatively equal differentiationmarker expression. Re-analysis of sorted sub-populations provedindistinguishable from primary culture populations (unpublisheddata) indicating that the cultures reverted to a balanced distribu-tion in vitro. These findings hint at a stable heterogeneity support-ing the notion of an inherent range of plasticity within FB cultures.Further experiments with single-cell clonal populations may deci-pher the phenotype distribution within cultures, and the range ofplasticity for a given cell. The practical meaning of this hetero-geneity must be evaluated in further functional testing and in vivo

studies.Fig. 6A shows overlaps of down-regulated genes in induced

groups, and Fig. 6B shows overlaps in the reference groups.Groups of genes that may be of particular importance to FB basedon their down-regulation in all other groups were investigatedfurther (Fig. 6C–E). The genes that were down-regulated in allreference cultures compared to FB (Fig. 6B) represent annotations(Fig. 6E), including BMP signaling pathway, TGF-b signaling path-

way, cell migration, cell adhesion, Wnt signaling, that are known tobe important in fibroblast biology. GO terms include cell surface

receptor linked signal transduction, which includes more than 25%of the genes, anatomical structure morphogenesis (20%) and cell

adhesion (15%). Genes in this set also represent annotations suchas blood vessel development, axonogenesis, erythrocyte differentia-

tion, skeletal system development. This shows substantial signalingreadiness and it is plausible that genes in these categories providethe background to the supportive role of fibroblasts in diversetissues. This may explain how FB are able to adapt to anotherniche—the required gene-regulating pathways are available.

Fig. 4. Results of osteogenic induction and gene-set interpretation A: Venn diagram showing genes expressed higher in both osteogenic media induced fibroblasts (O-

FB) and adipocytes (OB) than in fibroblasts (FB); B: important gene-set enrichment analysis (GSEA) results for each set. Expanded lists are found in supplementary file S2.

Fig. 5. Results of chondrogenic induction and gene-set interpretation A: Venn diagram showing genes expressed higher in both chondrogenic media induced

fibroblasts (C-FB) and adipocytes (CC) than in fibroblasts (FB); B: important gene-set enrichment analysis (GSEA) results for each set. Expanded lists are found in

supplementary file S2.

J. Rakar et al. / Differentiation 84 (2012) 305–313 311

The subset of genes regulated in all induced FB (Fig. 6A, C, and D)are involved in processes such as EGFR signaling, regulation of SMC

proliferation (includes MYOD1 targets) and cell differentiation. PPARAactivity and transcriptional regulators (ZIC2, TFAP2C, TCF12) indicatesprimed pathways important for multiple mesenchymal fates. Genesin this sub-set are targets of over 40 transcription factors, as found byGSEA, which alludes to an intricate network of regulation that canaccommodate numerous transcriptional cascades. An interesting

finding is the down-regulation of the a6-b4 integrin pathway

(WP244 and PC:1050) among all induced FB groups—a pathwayimportant for the interaction between dermal FB and the basal layerkeratinocytes. A typically dermal phenotype seems to be down-regulated in favor of phenotypes closer to the target tissue pheno-types. The wide inherent plasticity together with the supportingrole of the connective tissue shows that FB may be more importantthan is currently understood for the maintenance and repair of

Fig. 6. Gene-set comparisons from a fibroblast (FB) perspective. Overlaps of down-regulated genes in (A) induced groups and (B) reference groups compared to FB.

C: commonly regulated genes; D: annotations related to (C). E: annotations related to down-regulated genes in all reference cultures, but not in induced groups.

J. Rakar et al. / Differentiation 84 (2012) 305–313312

tissues, even to a degree that entails gross phenotype alteration. FBseem to be dynamic multi-lineage inducible cells and could thereforebe useful in the creation of complex autologous tissue engineeredconstructs.

3.7. Summary and conclusions

There is a diffuse heterogeneity of the isolated startingpopulation that makes mechanistic insights difficult to ascertain.The stability with which the experiments can be repeated showsthat the effects we see are entirely reproducible despite thepotentially confounding heterogeneity. The simplicity of the isola-tion, together with the high yield, and robustness of the cells add tothe value of FB as a cell source for TE. We have previously shown

that dermal FB appear to differentiate towards adipo-, chondro-, andosteogenic lineages given adequate stimuli (Junker et al., 2010).Repeating our experiments and analyzing at the depth of globalgene expression we corroborate our previous findings. The inducedFB show differential expression that coincide with that observed inreference cell types (AC, OB and CC), including functions indicatedby protein analysis as well as key regulating pathways. We notedifferences between induced FB and reference cultures and concludethat we have obtained FB phenotypes not previously described. Thedermal FB in vitro phenotype shows signs of activated transcriptionof genes important for multiple cellular lineages and loss of dermalspecification in response to induction stimulus. Their dynamicphenotype regulation and wide-spread availability make humandermal FB relevant for applications in clinical tissue engineering and

J. Rakar et al. / Differentiation 84 (2012) 305–313 313

regenerative medicine. The next steps to take include the construc-tion of 3D cultures and correlating our gene expression data tofunctional testing.

Acknowledgements

Financial support was received from the County of Ostergotlandand Vinnova without conflicts of interest, financial or otherwise.Authors wish to acknowledge the excellent assistance from AnitaLonn & Kristina Briheim, and convey thanks to I. Rundquist,J. Ungerback and M. Sigvardsson for meaningful discussions.

Appendix A. Supporting information

Supplementary data associated with this article can be found inthe online version at http://dx.doi.org/10.1016/j.diff.2012.08.003.

References

Bartsch, G., et al., 2005. Propagation, expansion, and multilineage differentiation ofhuman somatic stem cells from dermal progenitors. Stem Cells and Develop-ment 14, 337–348.

Ben-Tabou de-Leon, S., Davidson, E.H., 2007. Gene regulation: gene control net-work in development. Annual Review of Biophysics and Biomolecular Struc-ture 36, 191.

Benayahu, D., et al., 2009. Insights into the transcriptional and chromatinregulation of mesenchymal stem cells in musculo–skeletal tissues. Annals ofAnatomy—Anatomischer Anzeiger 191, 2–12.

Beresford, J.N., et al., 1984. Production of osteocalcin by human bone cells in vitro.Effects of 1,25(OH)2D3, 24,25(OH)2D3, parathyroid hormone, and glucocorti-coids. Metabolic Bone Disease and Related Research 5, 229–234.

Bills, C.E., et al., 1974. Complexes of organic acids with calcium phosphate: theVon Kossa stain as a clue to the composition of bone mineral. The JohnsHopkins Medical Journal 128, 194–207.

Blanchette-Mackie, E.J., et al., 1995. Perilipin is located on the surface layer ofintracellular lipid droplets in adipocytes. Journal of Lipid Research36, 1211–1226.

Cao, Z., et al., 1991. Regulated expression of three C/EBP isoforms during adiposeconversion of 3T3-L1 cells. Genes and Development 5, 1538–1552.

Carling, D., 2004. The AMP-activated protein kinase cascade–a unifying system forenergy control. Trends in Biochemical Sciences 29, 18–24.

Dahl, L.K., 1952. A simple and sensitive histochemical method for calcium.Proceedings of the Society for Experimental Biology and Medicine 80,474–479.

Entenmann, G., Hauner, H., 1996. Relationship between replication and differ-entiation in cultured human adipocyte precursor cells. The American Journalof Physiology 270, C1011–1016.

Fujimori, K., Amano, F., 2011. Forkhead transcription factor Foxa1 is a novel targetgene of C/EBPbeta and suppresses the early phase of adipogenesis. Gene 473,150–156.

Horwitz, E.M., et al., 1999. Transplantability and therapeutic effects of bonemarrow-derived mesenchymal cells in children with osteogenesis imperfecta.Nature Medicine 5, 309–313.

Irizarry, R.A., et al., 2003. Summaries of Affymetrix GeneChip probe level data.Nucleic Acids Research 31, e15.

Ishida, H., et al., 2005. Identification and characterization of novel isoforms ofhuman DP-1: DP-1{alpha} regulates the transcriptional activity of E2F1 as wellas cell cycle progression in a dominant-negative manner. The Journal ofBiological Chemistry 280, 24642–24648.

Jaiswal, N., et al., 1997. Osteogenic differentiation of purified, culture-expandedhuman mesenchymal stem cells in vitro. Journal of Cellular Biochemistry 64,295–312.

Johnstone, B., et al., 1998. In vitro chondrogenesis of bone marrow-derivedmesenchymal progenitor cells. Experimental Cell Research 238, 265–272.

Junker, J.P., et al., 2010. Adipogenic, chondrogenic and osteogenic differentiation ofclonally derived human dermal fibroblasts. Cells Tissues Organs 191, 105–118.

Komori, T., 2010. Regulation of bone development and extracellular matrix proteingenes by RUNX2. Cell and Tissue Research 339, 189–195.

Krampera, M., et al., 2007. Mesenchymal stem cells: from biology to clinical use.Blood Transfusion 5, 120–129.

Lander, E.S., et al., 2001. Initial sequencing and analysis of the human genome.Nature 409, 860–921.

Lazarus, H.M., et al., 1995. Ex vivo expansion and subsequent infusion of humanbone marrow-derived stromal progenitor cells (mesenchymal progenitorcells): implications for therapeutic use. Bone Marrow Transplantation16, 557–564.

Lev, R., Spicer, S.S., 1964. Specific staining of sulphate groups with Alcian Blue atlow pH. Journal of Histochemistry and Cytochemistry 12, 309.

Livak, K.J., Schmittgen, T.D., 2001. Analysis of relative gene expression data usingreal-time quantitative PCR and the 2(-delta delta C(T)) method. Methods 25,402–408.

Lorenz, K., et al., 2008. Multilineage differentiation potential of human dermalskin-derived fibroblasts. Experimental Dermatology 17, 925–932.

Malaval, L., et al., 1994. Cellular expression of bone-related proteins during in vitro

osteogenesis in rat bone marrow stromal cell cultures. Journal of CellularPhysiology 158, 555–572.

Mason, R.M., 1971. Observations on the glycosaminoglycans of aging bronchialcartilage studied with Alcian Blue. Histochemical Journal 3, 421–434.

Muruganandan, S., et al., 2009. Adipocyte differentiation of bone marrow-derivedmesenchymal stem cells: cross talk with the osteoblastogenic program.Cellular and Molecular Life Sciences 66, 236–253.

Nakamoto, T., et al., 2000. CIZ, a zinc finger protein that interacts with p130(cas)and activates the expression of matrix metalloproteinases. Molecular andCellular Biology 20, 1649–1658.

Ochiai, S., et al., 2008. Oxidative stress reaction in the meniscus of Bach 1 deficientmice: potential prevention of meniscal degeneration. Journal of OrthopaedicResearch 26, 894–898.

Pei, D., 2009. Regulation of pluripotency and reprogramming by transcriptionfactors. Journal of Biological Chemistry 284, 3365–3369.

Pettersson, S., et al., 2009. Human articular chondrocytes on macroporous gelatinmicrocarriers form structurally stable constructs with blood-derived biologicalglues in vitro. Journal of Tissue Engineering and Regenerative Medicine3, 450–460.

Pittenger, M.F., et al., 1999. Multilineage potential of adult human mesenchymalstem cells. Science 284, 143–147.

Rached, M.T., et al., 2010. FoxO1 is a positive regulator of bone formation byfavoring protein synthesis and resistance to oxidative stress in osteoblasts.Cell Metabolism 11, 147–160.

Schnabel, M., et al., 2002. Dedifferentiation-associated changes in morphology andgene expression in primary human articular chondrocytes in cell culture.Osteoarthritis and Cartilage 10, 62–70.

Sommar, P., et al., 2010. Engineering three-dimensional cartilage- and bone-liketissues using human dermal fibroblasts and macroporous gelatine microcar-riers. Journal of Plastic, Reconstructive and Aesthetic Surgery 63, 1036–1046.

Sorrell, J.M., Caplan, A.I., 2009. Fibroblasts-a diverse population at the center of itall. International Review of Cell and Molecular Biology 276, 161–214.

Thomopoulos, S., et al., 2011. Fibrocartilage tissue engineering: the role of thestress environment on cell morphology and matrix expression. Tissue Engi-neering Part A 17, 1039–1053.

Toma, J.G., et al., 2001. Isolation of multipotent adult stem cells from the dermis ofmammalian skin. Nature Cell Biology 3, 778–784.

Tontonoz, P., et al., 1994. Stimulation of adipogenesis in fibroblasts by PPARgamma 2, a lipid-activated transcription factor. Cell 79, 1147–1156.

Zaucke, F., et al., 2001. Cartilage oligomeric matrix protein (COMP) and collagen IXare sensitive markers for the differentiation state of articular primarychondrocytes. The Biochemical Journal 358, 17–24.

Zhang, B., et al., 2005. WebGestalt: an integrated system for exploring gene sets invarious biological contexts. Nucleic Acids Research 33, W741–748.

Zhu, J., et al., 2010. Effects of FoxO4 overexpression on cholesterol biosynthesis,triacylglycerol accumulation, and glucose uptake. Journal of Lipid Research 51,1312–1324.