RESEARCH ARTICLE

Fetal articular cartilage regeneration versus adultfibrocartilaginous repair: secretome proteomics unravelsmolecular mechanisms in an ovine modelIris Ribitsch1,*, Rupert L. Mayer2,*, Monika Egerbacher3, Simone Gabner3, Maciej M. Kanduła4,5, Julie Rosser1,Eva Haltmayer1, Ulrike Auer6, Sinan Gultekin1, Johann Huber7, Andrea Bileck2, David P. Kreil4,‡,Christopher Gerner2,‡ and Florien Jenner1,‡,§

ABSTRACTOsteoarthritis (OA), a degenerative joint disease characterized byprogressive cartilage degeneration, is one of the leading causes ofdisability worldwide owing to the limited regenerative capacity ofadult articular cartilage. Currently, there are no disease-modifyingpharmacological or surgical therapies for OA. Fetal mammals, incontrast to adults, are capableof regenerating injured cartilage in the firsttwo trimesters of gestation. A deeper understanding of the propertiesintrinsic to the response of fetal tissue to injury would allow us tomodulate the way in which adult tissue responds to injury. In this study,we employed secretome proteomics to compare fetal and adult proteinregulation in response to cartilage injury using an ovine cartilage defectmodel. The most relevant events comprised proteins associated withthe immune response and inflammation, proteins specific for cartilagetissue and cartilage development, and proteins involved in cell growthand proliferation. Alarmins S100A8, S100A9 and S100A12 and coiled-coil domain containing 88A (CCDC88A), which are associated withinflammatory processes, were found to be significantly upregulatedfollowing injury in adult, but not in fetal animals. By contrast, cartilage-specific proteins like proteoglycan 4 were upregulated in response toinjury only in fetal sheep postinjury. Our results demonstrate the powerand relevance of the ovine fetal cartilage regeneration model presentedhere for the first time. The identification of previously unrecognizedmodulatory proteins that plausibly affect the healing process holds greatpromise for potential therapeutic interventions.

KEY WORDS: Articular cartilage, Osteoarthritis, Regeneration,Proteome, Fetus

INTRODUCTIONOsteoarthritis (OA), a degenerative joint disease characterized byprogressive articular cartilage degeneration, is one of the mostcommonly diagnosed diseases in general practice and one of theleading causes of disability worldwide (Barbour et al., 2013;Johnson and Hunter, 2014; Lopez and Murray, 1998; Woolf andPfleger, 2003). In addition to its significant medical, social andpsychological impact on quality of life, OA is associated withcommensurate socioeconomic costs (Rabenda et al., 2007; Yelinet al., 2007). As adult articular cartilage has little intrinsic repaircapacity and current treatment options are mostly palliative, thedisease prevalence and burden places a strong emphasis on the needfor new therapeutic strategies that could modify the structuralprogression of the disease and regenerate articular cartilage. Thedevelopment of disease-modifying anti-OA drugs has thus farproven to be challenging, owing to the complexity of the disease andthe pathophysiological pathways that drive OA progression. OA has amultifactorial aetiopathogenesis involving genetic, molecular andbiomechanical influences, as well as lifestyle and environmentalstress stimuli. Nevertheless, OA culminates in a consistent molecular,structural and clinical sequence of disease progression, characterizedby inflammation, gradual loss of proteoglycans, collagen type II(COL2) degradation, cartilage fibrillation, loss of maturational arrestand phenotypic stability of articular chondrocytes, as reviewedpreviously (Goldring et al., 2011; Pap and Korb-Pap, 2015).

Fetal mammals, by contrast to adults, are capable of regeneratinginjured tissues including skin, palate, tendon, bone and cartilage inthe first two trimesters of gestation (Al-Qattan et al., 1993; Degenand Gourdie, 2012; Kumahashi et al., 2004; Longaker et al., 1992,1990; Namba et al., 1998; Stone, 2000; Wagner et al., 2001; Walkeret al., 2000). Despite progress over the past decade, the mechanismsof scarless fetal healing and the embryogenetic mechanisms ofarticular chondrogenesis remain largely unknown (Decker et al.,2017; Jenner et al., 2014a,b; Lo et al., 2012).

Hence, in this study, we had three key aims: (1) to establisha standardized cartilage lesion model allowing comparison ofcartilage healing in adult and fetal sheep (Ovis aries); (2) toestablish the feasibility, repeatability and relevance of proteomicanalysis of minute fetal and adult cartilage samples; and (3) tocompare fetal and adult protein regulation in response tocartilage injury.

The proteomic analysis of the differential response of fetal andadult cartilage to injury will make a major contribution to ourunderstanding of cartilage biology and help us to elucidate themolecular mechanisms underlying OA and cartilage regeneration.Such insights could help us to identify and target factors that areReceived 21 November 2017; Accepted 18 May 2018

1VETERM, University Equine Hospital, University of Veterinary Medicine Vienna,Vienna 1210, Austria. 2Department of Analytical Chemistry, Faculty of Chemistry,University of Vienna, Vienna 1090, Austria. 3Histology & Embryology, Department ofPathobiology, University of Veterinary Medicine Vienna, Vienna 1210, Austria.4DepartmentofBiotechnology,BokuUniversityVienna,Vienna1180,Austria. 5Instituteof Bioinformatics, Johannes Kepler University, Linz 4040, Austria. 6Department ofCompanion Animals and Horses, University of Veterinary Medicine Vienna, Vienna1210, Austria. 7Teaching and Research Farm Kremesberg, Clinical Unit for HerdHealthManagement inRuminants,Department forFarmAnimalsandVeterinaryPublicHealth, University of Veterinary Medicine Vienna, Vienna 1210, Austria.

*These authors contributed equally to this work‡These authors contributed equally to this work

§Author for correspondence (florien.jenner@vetmeduni.ac.at)

F.J., 0000-0002-6977-1984

This is an Open Access article distributed under the terms of the Creative Commons AttributionLicense (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use,distribution and reproduction in any medium provided that the original work is properly attributed.

1

© 2018. Published by The Company of Biologists Ltd | Disease Models & Mechanisms (2018) 11, dmm033092. doi:10.1242/dmm.033092

Disea

seModels&Mechan

isms

crucial to promote a regenerative response and might allow thedevelopment of disease-modifying treatment strategies to inducecartilage regeneration in adult mammals. A major challenge for theproteomic characterization of the complex protein mixture incartilage extract is the wide dynamic range of protein abundance,making the detection of low-abundance proteins very difficult(Stenberg et al., 2013; Wilson and Bateman, 2008). Althoughtechnically demanding, however, studying the functional proteomegives a more comprehensive picture of disease aetiopathogenesisthan gene expression analysis alone, as its interpretation is notlimited by a possible disparity between gene and protein expressionlevels (Ritter et al., 2013).

RESULTSThe ovine model supports complex surgical manipulationsrequired for the investigation of cartilage regenerationEwes and fetal sheep tolerated the laparotomy, uterotomy and fetalmanipulation well. No postoperative complications or abortionswere encountered. Fetal sheep at 80 days of gestation (gd) hadage-appropriate crown-anus lengths within the reported range of10.1±1.3 cm (Mufti et al., 2000). The landmarks for standardizedinduction of medial femoral condylar cartilage lesions (Fig. 1) wereeasily identified and allowed placement of the lesion in the centreof the condyle.

Long-term evaluation confirmed fetal regenerative versusadult scarring cartilage repairIn the pilot study designed as a proof-of-principle of fetal regenerationat 28 days postoperatively versus adult scarring repair at 5 monthspostinjury, the defect was macroscopically undetectable in fetalsheep resulting in an OARSI (Osteoarthritis Research SocietyInternational) macroscopic score (Little et al., 2010) of 0 forcartilage, osteophytes and synovium. By contrast, in adult sheepthe defect was clearly evident and only partially filled withfibrocartilaginous tissue resulting in an OARSI macroscopic scoreof 5/16 for cartilage (surface roughening plus defect), 0/5 forosteophytes and 2/5 for synovium.Histologically (Fig. 2), no defect repair and only minimal

fibrocartilaginous regeneration adjacent to microfractures withoutintegration with the surrounding cartilage was achieved in adultsheep 5 months postoperatively. In fetal sheep, 28 days after surgery,the defect was filled with differentiating hyaline cartilage in about

80-90% of the repair tissue and 10-20% showed progressinghyalinization and full integration with the surrounding cartilage.

Injury- and repair-associated macroscopic and histologicalchanges in adult and fetal sheep 3 days postinjuryUpon harvest 3 days postoperatively, the OARSI macroscopic scorewas 4/16 for cartilage, because of the surgically induced defect inthe medial femoral condyle, and 0 for osteophytes (owing to theshort time since surgery, no OARSI score was assigned to themacroscopic appearance of the synovium). Within the first 3 daysafter injury, no histologically visible cartilage repair could bedetected. Therefore, none of the established repair scoring systemscould be applied. Instead, the description of the structural conditionswas based on the evaluation criteria of the International CartilageRepair Society (ICRS) assessment including the cartilage surfaceandmatrix, cell distribution, cell viability and subchondral bone, butwithout scores (Mainil-Varlet et al., 2003).

Adult control condyles showed healthy articular cartilage witha smooth surface, physiological matrix composition, and typicaldistribution of chondrocytes (Fig. 3A,B). COL2 staining washomogeneous and distinct throughout the whole articularcartilage (Fig. 3C).

Creation of the cartilage lesion in adults resulted in matrixdepletion at the site of injury as well as in the superficial zone(Fig. 3D,E). Next to the cartilage lesion an acellular area ofabout 100 μm thickness was found with either empty lacunae orhomogenous matrix lacking apparent lacunae. No cell clusteringwas observed. The microfractures penetrating the subchondralbone plate were visible (Fig. 3D,E). One sample showed a focalaccumulation of granulocytes in the bone marrow cavity below thecartilage lesion. In the immediate vicinity (∼10 μm) of the cartilagelesion, COL2 staining intensity was decreased (Fig. 3F).

Similar to the adults, fetal uninjured control samples showed asmooth cartilage surface, homogenous matrix composition, anddistinct COL2 staining throughout the whole condyles (Fig. 3G-I).

Although matrix depletion was also detected around the cartilagelesions in the fetal samples, it was less marked compared with adults(Fig. 3J,K). An almost acellular area of 50 μm surrounded thecartilage lesion. The lesion surface was partly covered with fibroustissue originating either from cartilage canals or connective tissueflanking the articular surface. The pattern of the COL2 stainingaround the fetal cartilage lesions was similar to that of adults with a10 μm zone of decreased staining intensity (Fig. 3L).

In adult control animals, no proliferating cells (demonstratedby expression of the proliferation antigen Ki67) were foundin the articular cartilage or the subchondral bone (Fig. 4A).Few Ki67-positive cells were detected in the bone marrowcavities. Chondrocytes in all cartilage zones expressed the matrixmetalloproteinases MMP9 and MMP13 with a stronger stainingintensity for MMP9 (Fig. 4B,C); however, no MMP-positiveosteocytes were observed.

Likewise, in the injured adult cartilage samples, no Ki67-positivecells were observed (Fig. 4D); however, in one sample, anaccumulation of Ki67-positive cells was found in a microfracturegap, which was filled with bone marrow. Both MMP9 and MMP13expression were reduced within and adjacent to the cartilagelesions (Fig. 4E,F), as compared with the intact cartilage of therespective sample.

In fetal healthy cartilage, evenly distributed Ki67-positive cells(Fig. 4G) and MMP-expressing cells (Fig. 4H,I) were detectedthroughout the whole cartilage. The staining pattern of MMP9appeared identical to that of MMP13.

Fig. 1. Diagram of the distal femur. The medial and lateral trochlea ridge andthe medial and lateral condyle are identified as landmarks. Cartilage lesionlocation and size is indicated in blue in adult and green in fetal sheep. The lesionwas centred 15 mm (adult) and 3 mm (fetus) distant from the medial trochlear-condylar junction on a line that virtually extended the medial trochlear ridge.

2

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033092. doi:10.1242/dmm.033092

Disea

seModels&Mechan

isms

Although in the fetal injured cartilage an almost cell-free zone of50 μm was found to surround the lesions, single Ki67-expressingcells as well as MMP-positive cells could still be detected in thisarea (Fig. 4J-L). More MMP-expressing cells were locatedadjacent to the cell-free zone, as well as in the cartilage canalsof the injured condyle.

The ovine model supports comprehensive molecularprofiling by high-resolution mass spectrometrySecretome analysis of control and injured (3 days postoperative)cartilage tissue samples derived from adult and fetal sheep, usinghigh-resolution mass spectrometry (MS), enabled the identificationof a total number of 2106 distinct proteins. Of these, 445 proteinswere found to be significantly regulated (q-value <0.05) in response

to cartilage injury in adult animals, compared with 74 proteins infetal animals (Fig. 5). Comparing protein baseline expression, 1288proteins were found to be significantly differentially regulatedbetween fetal and adult control animals. The injury response offetal and adult sheep was significantly differentially regulated in356 proteins. A comparison of protein regulation in adult andfetal animals (Fig. 5, Table 1, Tables S1,S2) revealed differences inthree key groups of proteins: (i) proteins associated with the immuneresponse and inflammation, (ii) proteins specific for cartilagetissue and cartilage development, and (iii) proteins involved incell growth and proliferation (Table 1, Tables S1,S2). Multiplewell-known factors in inflammatory processes, such as S100A8,S100A9, S100A12 and CCDC88A, were found to be significantlyupregulated following injury in adult (q<0.001) but not in

Fig. 2. Healing of adult (5 months postinjury) and fetal (28 days postinjury) cartilage defects: H&E-stained sections. (A,B) In adult sheep no repair isapparent except in areas of microfracture. Only minimal fibrocartilage is evident with partial hyalinization; there is no integration with surrounding cartilage (seeinsert). (C,D) In fetal sheep the defect is 80-90% filled with differentiating hyaline cartilage and the superficial 10-20% with repair tissue with progressinghyalinization; there is good integration with surrounding cartilage. *Processing artifact.

Fig. 3. Morphology and extracellular matrix composition of healthy and injured adult and fetal cartilage: H&E, Safranin O and COL2 staining.Condyles are shown from the four study groups: (A-C) uninjured adult controls; (D-F) injured adult cartilage; (G-I) uninjured fetal controls; and (J-L) injuredfetal cartilage. Arrows mark the transition from healthy cartilage to the lesion site; asterisks indicate sites of microfracture penetrating the subchondral bone plate(D-F). Fibrous tissue partly covering the surface of the lesion (arrows, J-L) was found in fetal injured condyles.

3

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033092. doi:10.1242/dmm.033092

Disea

seModels&Mechan

isms

fetal animals (Table 1). By contrast, several proteins with anti-inflammatory and growth-supporting effects, such as proteinphosphatase Mg2+/Mn2+-dependent 1A (PPM1A) and cell-divisioncycle 42 (CDC42), showed a significant increase in response to injuryin fetal sheep (q=0.005 and 0.006) compared with adults (Table 1).Cartilage-specific proteins, such as aggrecan (ACAN), cartilageoligomeric matrix protein (COMP), chondroadherin (CHAD) andproteoglycan 4 (PRG4), had a significantly higher baselineexpression in adults (q<0.001) and showed little injury responsein either age group with the exception of PRG4, which wassignificantly upregulated in fetal injured sheep (q=0.01). Otherproteins related to cell growth and proliferation, such as mitogen-activated protein kinase 3 (MAPK3/ERK1) and GA-bindingprotein transcription factor α subunit (GABPA), also displayeddifferences in abundance (q=0.02 and 0.04) as well as regulationbetween adult and fetal sheep (q=0.003 and 0.0001).Our results demonstrate the biological relevance and reproducibility

of our new ovine cartilage defect model and MS analysis (Fig. 6).Technical measurement reproducibility was excellent, with variationclearly lower than variation between biological replicates, indicatinga high sensitivity of the proteomics profiling workflow (Fig. 6).The robustness of our new cartilage defect model is reflected inthe variance across biological replicates being small in relation to theexamined biological effects, whether injury versus control ordifferences between adult and fetal samples (Fig. 6). For both adultand fetal samples, low variance across replicates indicates goodreproducibility of our experimental setup, confirming that biologicallymeaningful signals can sensitively be obtained from only a moderatesample size. Furthermore, our findings confirm good standardizationof articular cartilage defects between individuals of both the adult andfetal age group.

DISCUSSIONFetal scarless regeneration is a paradigm for ideal tissue repair.The mechanisms of this tightly regulated process, involving theinterplay of growth factors, cytokines, proteinases and cellularmediators combined with differences in cellular density, proliferationrate, inflammatory response, extracellular matrix (ECM) compositionand synthetic function, are poorly understood compared withadults (Cowin et al., 1998; Degen and Gourdie, 2012; Fergusonand O’Kane, 2004). Studies in dermal wound healing identifiedqualitative, quantitative and temporal differences in growth factorand cytokine expression between adult and fetal wounds (Cowinet al., 1998; Degen and Gourdie, 2012; Ferguson and O’Kane, 2004).

Our results illustrate the biological relevance, the technicalfeasibility and repeatability of a new ovine cartilage defect model(Fig. 1) and analytical approaches and confirm regeneration infetal sheep versus scarring repair in adult sheep (Fig. 2). The well-characterized ontogeny of the ovine immune and inflammatorysystem and bone marrow niche made the sheep an ideal model inwhich to examine fetal regeneration (Almeida-Porada et al., 2004;Jeanblanc et al., 2014). The sheep is also a well-accepted andvalidated model for musculoskeletal disorders and, in particular,cartilage degeneration (Dorotka et al., 2005; Entrican et al., 2015;Mrugala et al., 2008; van Turnhout et al., 2010a,b). Hence,sheep are commonly used to study therapies for OA and articularcartilage lesions (Lu et al., 2006; Munirah et al., 2007; Xue et al.,2013). In addition, fetal sheep share many important physiologicaland developmental characteristics with humans and have proventhemselves invaluable models for mammalian physiology(Almeida-Porada et al., 2004; Jeanblanc et al., 2014). Resultsobtained in the fetal lamb have been directly applicable to theunderstanding of human fetal growth and development and are

Fig. 4. Distribution of proliferation marker Ki67 and MMPs in healthy and injured adult and fetal cartilage. (A-C) Uninjured adult controls, (D-F) injuredadult cartilage, (G-I) uninjured fetal controls and (J-L) injured fetal cartilage. Inserts in fetal injured condyles indicate transition from healthy cartilage to thelesion site. Scale bars: 100 μm (adult samples); 200 μm (fetal samples); 20 μm (inserts).

4

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033092. doi:10.1242/dmm.033092

Disea

seModels&Mechan

isms

highly predictive of clinical outcome in a variety of applicationsincluding in utero stem-cell transplantation (Almeida-Poradaet al., 2004, 2007; Jeanblanc et al., 2014; Kim et al., 2013;Kuypers et al., 2012; Liechty et al., 2000; Porada et al., 2005).A number of specific characteristics make sheep particularlywell-suited for OA, regenerative medicine and fetal regeneration

research enabling results of high clinical relevance to be obtained:(1) large size facilitating repeated sampling from individual animalsand harvest of adequate sample sizes; (2) comparable body weight tohumans; (3) similar mechanical exertion to humans (Bruns et al.,2000; Russo et al., 2015); (4) relatively long life expectancy (lifespan8-12 years) allowing longitudinal analysis as well as evaluation of

Fig. 5. Venn diagram giving an overview of genes implicated by a range of differential screening tests (n=3 biological replicates/group, two technicalreplicates/biological replicate). Specifically, we examine the average total response to injury (magenta), the fetal response (blue), the adult response (red),and significant response differences (green). Separately assessing significances for each of the four tests improves sensitivity and specificity by avoiding anaccumulation of thresholding artifacts. Comparing cartilage on day 3 after injury with matching control tissues yielded 385 genes implicated in the totalresponse incorporating the average evidence from all sample types (7+9+0+35+261+2+56+15). Analogously, 74 genes were implicated in the fetal response(9+3+35+8+2+2+15), of which 13 were newly identified (3+8+2+0). Conversely, 445 genes were implicated in the adult response (45+261+64+2+56+2+15), ofwhich 111 were newly identified (45+64+2+0). Response differences are shown by 356 genes with an injury response in fetal samples that was significantlydifferent to that in adult samples (3+0+45+35+261+2+8+2), including three genes previously not implicated, eight genes so far only implicated in the fetalresponse, 45 genes so far only implicated in the adult response, two genes already implicated in both, 35 genes already implicated in the total and the fetalresponses, 261 genes already implicated in the total and the adult responses, and two genes implicated in all. These observations reflect the finding that responsestrength and direction of expression change can be affected differently in the injury response of fetal and adult samples.

Table 1. Selected proteins that are differentially expressed in response to cartilage injury

Name

Fetal control vs adultcontrol Fetal injury* vs control Adult injury* vs control

Fetal vs adult injuryresponse

logFC† q logFC† q logFC† q logFC† q

ACAN −10.74 5.54 E−11 1.77 1.00 E+00 −1.36 9.15 E−01 3.13 1.01 E−01CCDC88A 4.32 5.29 E−01 −10.45 5.62 E−03 10.05 8.04 E−04 −20.50 1.90 E−05CCDC42 −3.58 5.97 E−01 8.68 4.27 E−02 −3.45 9.39 E−01 12.13 5.47 E−03CHAD −10.94 2.14 E−09 1.48 1.00 E+00 −1.22 1.00 E+00 2.70 6.95 E−01COL2a1 −2.07 3.69 E−01 1.14 1.00 E+00 −1.01 1.00 E+00 2.15 1.00 E+00COMP −9.64 1.13 E−10 1.67 1.00 E+00 −0.05 1.00 E+00 1.72 1.00 E+00GABPA −3.01 3.50 E−02 8.23 1.52 E−05 −0.29 1.00 E+00 8.52 1.09 E−04MAPK3 −7.81 1.50 E−02 1.11 1.00 E+00 −13.73 1.23 E−05 14.84 2.55 E−03PPM1A −1.62 1.00 E+00 8.91 1.36 E−03 −0.21 1.00 E+00 9.12 4.69 E−03PRG4 −11.56 3.94 E−12 3.18 1.27 E−02 −1.43 5.29 E−01 4.61 1.63 E−03S100A12 −2.71 1.86 E−01 8.39 4.99 E−05 13.54 7.37 E−10 −5.15 7.15 E−02S100A8 −2.78 3.48 E−01 7.49 1.29 E−03 15.80 7.15 E−10 −8.32 3.53 E−03S100A9 0.08 1.00 E+00 6.34 4.25 E−01 15.45 9.32 E−08 −9.11 4.15 E−02

*Injury is at day 3. †logFC represents the fold change in a logarithmic scale to the basis 2 based on label-free quantification intensities.

5

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033092. doi:10.1242/dmm.033092

Disea

seModels&Mechan

isms

long-term efficacy and safety of treatments; (5) long gestationalperiod (150 days) provides sufficient temporal resolution to translatefindings obtained in sheep into human parameters (Jeanblanc et al.,2014); (6) extremely well-characterized immune developmentanalogous to humans (Almeida-Porada et al., 2004; Jeanblancet al., 2014;Maddox et al., 1987;Miyasaka and Trnka, 1986; Osburn,1981; Sawyer et al., 1978); (7) bone marrow ontogeny and nichedevelopment closely paralleling humans (Jeanblanc et al., 2014).Furthermore, for the sheep model, an acceptable annotation status

and representative subsets of identified proteins are available onsources such as the National Center for Biotechnology Information(NCBI; e.g. 30,584 genes and 69,889 proteins) (NCBI ResourceCoordinators, 2018), allowing good applicability and translationof the results.

In this study, we compared the adult and fetal response tocartilage injury 3 days after lesion induction as this time point isestablished to be within the time window of inflammation for bothadult and fetal individuals, one of the injury responses hypothesizedto crucially contribute to the difference between adult and fetalhealing. For the fetal injury response, it is only known that cartilageregeneration occurs within 4 weeks. This is in stark contrast tothe adult injury response, which comprises an inflammatory phaseof 3-5 days, a proliferative phase of 3-6 weeks and a remodelingphase of up to 1 year duration resulting in a fibrocartilaginous scar.As the timeline of the fetal injury response is not yet established,choosing a later date would have made data interpretation andcorrelation of adult and fetal data impossible. Three days is withinthe peak period of the adult inflammatory response, allows for

Fig. 6. Sample correlation structure. This figure compares pairwise sample correlations: Spearman rank correlation coefficients are given in the boxes belowthe diagonal and correlations are visualized above the diagonal (darker and narrower ellipses indicate higher correlations). Rows and columns show samplelabels, where A/F=adult/fetal, c/i=control/injured and #.# show biological and technical replicate numbers (n=3 biological replicates/group, two technicalreplicates/biological replicate).

6

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033092. doi:10.1242/dmm.033092

Disea

seModels&Mechan

isms

recruitment of monocytes/macrophages to the injury site and hasalso been shown to be associated with signs of inflammation in fetalinjuries in other tissues.The main factors identified within the secretome were ECM

proteins, growth factors and inflammatory mediators such ascytokines and chemokines. Considering the key chondrocytesignaling pathways regulating processes of inflammation, cellproliferation, differentiation and matrix remodeling, which includethe p38, JNK and ERK mitogen-activated protein (MAP) kinases,the phosphatidylinositol 3-kinase (PI3K)-AKT pathway, theJAK-STAT pathway, Rho GTPases and WNT–β-catenin andSMAD pathways (Beier and Loeser, 2010), our data provide aninitial indication of differences in the inflammatory response toinjury between adult and fetal cartilage and suggest the activeproduction of anti-inflammatory and growth factors, such asPPM1A and CDC42, in the fetal environment.PPM1A is a bona fide phosphatase, which is involved in the

regulation of many developmental processes such as skeletal andcardiovascular development. Through its role as phosphatase ofmany signaling molecules, such as p38 kinase, CDK2, PI3K, Axinand SMAD, upregulation of PPM1A abolishes, for example,transforming growth factor β (TGF-β)-induced antiproliferativeand transcriptional responses (Wang et al., 2014), as well as bonemorphogenetic protein (BMP) signaling (Duan et al., 2006).Furthermore, through the dephosphorylation of IκB kinase β,PPM1A terminates tumor necrosis factor α (TNFα)-induced NF-κBactivation and is thus involved in the regulation of inflammation, theimmune response and apoptosis (Sun et al., 2009).CDC42 belongs to the family of Rho GTPases and controls

a broad variety of signal transduction pathways regulating cellmigration, polarization, adhesion proliferation, differentiation andapoptosis in a variety of cell types (Sun et al., 2009). CDC42 isrequired in successive steps of chondrogenesis by promotingmesenchymal condensation through the BMP2/CDC42/PAK/p38/SMAD signaling cascade and chondrogenic differentiation throughthe TGF-β/CDC42/PAK/AKT/SOX9 signaling pathway (Wanget al., 2016). Another essential CDC42 function relevant to thecurrent study is its involvement in wound healing by attenuatingMMP1 expression (Rohani et al., 2014) and regulating spatiallydistinct aspects of the cytoskeleton machinery, especially actinmobilization toward the wound (Benink and Bement, 2005). Giventhe increase of actin-containing articular chondrocytes in responseto cartilage injury, these could also have a role in the healing ofcartilage defects (Wang et al., 2000).By contrast to the anti-inflammatory factors upregulated in fetal

sheep in response to injury, adult sheep displayed a significantincrease of inflammatory mediators such as alarmins S100A8,S100A9, S100A12 and CCDC88A. The alarmin S100 proteins aremarkers of destructive processes such as those occurring in OA(Liu-Bryan and Terkeltaub, 2015; Nefla et al., 2016; van den Boschet al., 2015). Accordingly, in OA articular S100A8 and S100A9protein secretion is increased, recruiting immune cells to inflamedsynovia, initiating the adaptive immune response, inducing canonicalWNT signaling and promoting cartilagematrix catabolism, osteophyteformation, angiogenesis and hypertrophic differentiation (Liu-Bryanand Terkeltaub, 2015; Nefla et al., 2016; van den Bosch et al., 2015).S100A8 and S100A9 upregulate markers characteristic for ECMdegradation [MMPs 1, 3, 9 and 13, interleukin 6 (IL-6) and IL-8] anddownregulate growth promotion markers (ACAN and COL2); thus,these alarmins have a destructive effect on chondrocytes, causingproteoglycan depletion and cartilage breakdown (Schelbergen et al.,2012). Also, S100A12 is upregulated in OA cartilage and has been

shown to increase the production of MMP13 and vascular endothelialgrowth factor (VEGF) in OA chondrocytes via the p38 MAPK andNF-κB pathways (Nakashima et al., 2012). Another relevant protein,which was significantly downregulated upon injury in fetal sheepbut significantly upregulated in injured adult sheep, is CCDC88A.CCDC88A is a multimodular signal transducer, which modulatesgrowth factor signaling during diverse biological and diseaseprocesses including cell migration, chemotaxis, development,self-renewal, apoptosis and autophagy. CCDC88A exerts its effecton these processes by integrating signals downstream of a varietyof growth factors, such as epidermal growth factor, insulin-likegrowth factor, VEGF, insulin, STAT3, platelet-derived growthfactor receptor and trimeric G-protein Gi (Dunkel et al., 2012;Ghosh et al., 2008). In addition, CCDC88A, which is expressedat high levels in immune cells of the lymphoid lineage, has animportant role in T-cell maturation, activation and cytokineproduction during pathological inflammation and its inhibitioncould help treat inflammatory conditions, as shown in vitro and inmouse studies (Kennedy et al., 2014). Furthermore, CCDC88A,through activation of Gαi, simultaneously enhances the profibrotic(PI3K-AKT-FOXO1 and TGF-β-SMAD) and inhibits the antifibrotic(cAMP-PKA-pCREB) pathways, shifting the fibrogenic signalingnetwork towards a profibrotic programme (Lopez-Sanchez et al.,2014). Interestingly, in the liver, sustained upregulation of CCDC88Aoccurs in all forms of chronic fibrogenic injuries, but not in acuteinjuries that heal without fibrosis. This observation indicates thatincreased expression of CCDC88A during acute injuries mightenhance progression to chronicity and fibrosis (Lopez-Sanchez et al.,2014). CCDC88A also regulates the PI3K-AKT pathway, whichexhibits pleiotropic functions in chondrogenesis, cartilage homeostasisand inflammation. This pathway might further induce an increase inMMP production by chondrocytes leading to subsequent cartilagedegeneration, through its multiple downstream target proteins (Chenet al., 2013; Fujita et al., 2004; Greene and Loeser, 2015; Kita et al.,2008; Litherland et al., 2008; Starkman et al., 2005; Xu et al., 2015).

Remarkably, in this study, the cartilage matrix proteins PRG4,ACAN, COMP and CHAD had a significantly higher baselineexpression in adult sheep and showed little injury response in either agegroup with the exception of PRG4, which was significantlyupregulated in fetal injured sheep. PRG4, in response to injury,increased 3.2-fold (q=0.01) in fetal sheep, which is a 4.6-fold higherincrease compared with adults (q=0.002), indicating a stronger andmore rapid cartilage matrix production. As PRG4-expressing cellsconstitute a cartilage progenitor cell population, the higher baselineexpression in adults is particularly surprising. This observation can beexplained, however, by the restriction of PRG4 to the most superficialcell layer in fetal joints compared with a distribution throughout theentire cartilage depth in older individuals (Kozhemyakina et al., 2015).

By contrast to the cartilage matrix glycoproteins, many growthfactors such as GABPA and MAPK3 showed, as expected,differential regulation following injury between adult and fetalsheep. GABPA, a member of the ETS protein family which isubiquitously expressed and has an essential role in cellular functionssuch as cell-cycle regulation, cellular growth, apoptosis anddifferentiation (Rosmarin, 2004), showed a further significantupregulation in fetal injury and no response to adult injury(q=0.0001). GABPA activates the transcriptional co-activator Yes-associated protein (YAP), which is essential for cellular and tissuedefences against oxidative stress, cell survival and proliferation andcan induce the expression of growth-promoting genes important fortissue regeneration after injury (Wen Chun Juan, 2016). The cellularimportance of GABPA is further highlighted by the observation that

7

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033092. doi:10.1242/dmm.033092

Disea

seModels&Mechan

isms

in Gabpa conditional knockout embryonic stem cells (ESCs),disruption of GABPA drastically repressed ESC proliferation andcells started to die within 2 days (Ueda et al., 2017).The growth regulator MAPK3 had a higher baseline expression

in adult sheep (logFC=7.8, q=0>02), but significantly decreased(13.7 logFC, q<0.0001) after injury; fetal MAPK3 remainedessentially unchanged. MAPK3 is an essential component of theMAP kinase signal transduction pathway and as such contributes tocell growth, adhesion, survival and differentiation through theregulation of transcription, translation and cytoskeletal rearrangements.MAPK3 also fulfils an essential role in the control of chondrogenesisand osteogenesis of mesenchymal stem cells (MSCs) under TGF-β ormechanical induction and positively regulates chondrogenesis ofMSCs (Arita et al., 2011; Pelaez et al., 2012; Bobick et al., 2010).Our results are consistent with previous studies in fetal skin

wounds, which have also shown a different and reduced inflammatoryresponse and decreased scar formation during dermal wound healingin fetal sheep and mice (Cowin et al., 1998). The chronic progressivearticular damage of OA is associated with similar levels of pro-inflammatory cytokines and chemokines throughout the disease(Ritter et al., 2013) and occurs through a complex programme involvingon-going local inflammation triggered by cytokines and endogenousactivation of innate immunity, complement and metabolic pathways(Pap and Korb-Pap, 2015). Therefore, the different inflammatoryresponse demonstrated here in the fetus might be a major contributorto fetal scarless cartilage healing. This is especially intriguing asfetal sheep have a normally functioning immune system by 75 gd(Almeida-Porada et al., 2004; Emmert et al., 2013). Leukocyteshave been shown to be present and increase rapidly at the endof the first trimester (Mackay et al., 1986; Maddox et al., 1987).Fetal sheep are able to form large amounts of specific antibodiesin response to antigen stimuli by 70 gd (Silverstein et al., 1963)and reject orthotopic skin grafts and stem cell xenotransplantsadministered after 75-77 gdwith the same competence and rapidity asadults (Silverstein et al., 1964). Furthermore, fetal sheep have aninflammatory response to injury before 80 gd (Kumta et al., 1994;Moss et al., 2008; Nitsos et al., 2006). The first evidence ofinflammation, the presence of TNF and IL-1, has even been shown asearly as 30-40 gd (Dziegielewska et al., 2000).In conclusion, our results demonstrate the power of a new ovine

fetal cartilage regeneration model and of our analytical approach.A comparison of fetal and adult protein regulation in response tocartilage injury found both positive and negative regulators ofinflammatory events to be differentially regulated. These findingshold promise for potential therapeutic interventions, as the presenceof a negative regulator is more easily mimicked than the absence of apositive regulator. Further studies employing this newly developedanimal model and using analytical techniques to identify proteinsinvolved in OA aetiology and pathogenesis, as well as potentialbiomarkers and therapeutic targets, are warranted. Such studies willenable us to work towards the goal of finding novel biomimeticsolutions, which might be exploited to favourably shift the adultcartilage healing milieu to a more fetal phenotype to induceregeneration by recapitulating cartilage ontogeny.

MATERIALS AND METHODSAnimal modelStandardized cartilage lesions were created in musculoskeletally mature(2-5 years, body weight 95±12 kg), female, healthy, non-gravid Merino-cross ewes (Ovis aries) without orthopaedic disease and fetal lambs (80 gd,term=∼150 days) with approval of the national (FederalMinistry of Science,BMWFW) and institutional animal welfare committee (approval numbers

68.205/0155-WF-/V/3b/2014 and 68.205/0028-II/3b/2014). For the fetalsubjects, only twin pregnancies were included to provide a twin lamb asuninjured control on a background of low genetic variation to allowdifferentiation between protein secretion of regular fetal development andfetal response to cartilage injury. Fetal hind limbs were exteriorized from theuterus via a standard ventral-midline laparotomy followed by an uterotomyover a randomly chosen uterine horn.

For the purpose of lesion induction, a minimally invasive medialparapatellar arthrotomy (Orth and Madry, 2013) was performed in bothknees in adult and fetal sheep. A bilateral full-thickness cartilage lesion witha diameter of 7 mm (adult sheep) and 1 mm (fetal lamb) was created in themedial femoral condyle region (Fig. 1) using a custom-made precisionsurgical instrument (trocar with sleeve) for adult sheep and for fetal lambs aVersi-handle (Ellis Instruments, Madison, NJ, USA) with adjustable depthcontrol, which was set at a lesion depth of 1 mm.

To ensure that differences in the healing response between fetal and adultarticular cartilage are not caused by differences in blood supply, which infetal sheep is starting at an articular cartilage depth of 400 μm (Namba et al.,1998), we created three microfractures through the subchondral bone platein adult cartilage defects to gain access to the bone marrow vasculature(Dorotka et al., 2005). The arthrotomy was closed in one dermal layer using6-0 monofilament nylon (Monosof, Covidien, Minneapolis, MN, USA) infetal lambs and in three layers using 2-0 monofilament absorbable polyester(Biosyn, Covidien) for the joint capsule and subcutaneous tissue and 2-0monofilament nylon for the skin in adult sheep. Adult animals were allowedfull weight-bearing immediately after surgery. All adult sheep were treatedwith antibiotics peri-operatively and received pain management. Painmanagement was provided with morphine to avoid anti-inflammatory drugs,which would influence the result of the study.

Pilot studyIn a pilot study, as a proof-of-principle of fetal regeneration versus adultfibrocartilaginous repair, three adult and three fetal injured sheep wereeuthanized at 5 months (adult) and 28 days (fetal) postoperatively formacroscopic and histological evaluation of the defect repair. At the timeof euthanasia, digital photographs were taken and the articular cartilagesurface and the cartilage defect healing response was macroscopicallyevaluated using the OARSI recommendations for macroscopic scoring ofcartilage damage in sheep, taking into account surface roughening,fibrillation, fissures and the presence and size of osteophytes and erosions(Little et al., 2010). For fetal sheep the OARSI macroscopic score wassize-adjusted by multiplying the adult lesion size cut-off values with3.4/36.4, the ratio of the reported tibia length of fetal versus adult sheep(Mufti et al., 2000; Salami et al., 2011).

Tissue harvestAt day 3 after injury, samples were collected from three biological replicatesper comparison group (adult injured, fetal injured, fetal uninjured twincontrol). In adult sheep, samples were also harvested from uninjuredcontrols (n=3).

After macroscopic scoring, the medial femoral condyles were surgicallyexcised and left and right knees were randomly assigned to MS andhistology. ForMS the (cartilage) tissue remnants contained in the defect areaand the cartilage rim surrounding the lesion (3 mm width in adults; 1 mmwidth in fetal sheep) were excised.

Histology and immunohistochemistryFor histological analysis the femoral condyles were fixed in 4% bufferedformalin. Condyles from adult sheep were decalcified in 8% neutral EDTA.After embedding in paraffin, 4 µm thick sections were cut and mounted on3-aminopropyltriethoxysilane (APES)-glutaraldehyde-coated slides (Sigma-Aldrich,Vienna,Austria).Consecutive sectionswere stainedwithHaematoxylinand Eosin (H&E) and Safranin O (Mulisch and Welsch, 2015).

For immunohistochemistry, sections were deparaffinized, rehydrated andendogenous peroxidase was blocked with 0.6% hydrogen peroxide inmethanol (15 min at room temperature). Nonspecific binding of antibodieswas prevented by incubation with 1.5% normal goat serum (DakoCytomation, Glostrup, Denmark) in phosphate-buffered saline (PBS;

8

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033092. doi:10.1242/dmm.033092

Disea

seModels&Mechan

isms

30 min at room temperature). Primary antibodies (anti-COL2, anti-Ki67,anti-MMP9 and anti-MMP13; Table 2) were incubated overnight at 4°C.An appropriate BrightVision Peroxidase system (Immunologic, Duiven,The Netherlands) was used and peroxidase activities were localized withdiaminobenzidine (DAB; Sigma-Aldrich). Cell nuclei were counterstainedwith Mayer’s Haematoxylin.

Tissue from adult sheep mammary glands and human placenta served aspositive controls. For negative controls, the primary antibody was omitted.

Mass spectrometryThe cartilage rim and the tissue remnants obtained from the lesion site werecultivated in serum-free RPMI medium (Gibco, Life Technologies, Austria)supplemented with 100 U/ml penicillin and 100 µg/ml streptomycin (ATCC,LGC Standards GmbH, Germany) for 6 h under standard cell-cultureconditions (37°C and 5% CO2). Afterwards, medium was sterile-filteredthrough a 0.2 µm filter and precipitated overnight with ice-cold ethanolat −20°C. After precipitation, proteins were dissolved in sample buffer[7.5 M urea, 1.5 M thiourea, 4% CHAPS, 0.05% SDS, 100 mMdithiothreitol (DDT)] and protein concentrations were determined usingBradford assay (Bio-Rad Laboratories, Munich, Germany).

A 20 μg sample of each protein was used for a filter-aided digestion, asdescribed previously (Aukland, 1991; Aukland et al., 1997a,b; Bileck et al.,2014; Wi�sniewski et al., 2009). Briefly, 3 kDa molecular weight cut-off filters(Pall Austria Filter GmbH) were conditioned with MS-grade water (MilliporeGmbH). Protein samples were concentrated on the pre-washed filter bycentrifugation at 15,000 g for 15 min. After reduction with DTT [5 mg/mldissolved in 8 M guanidinium hydrochloride in 50 mM ammoniumbicarbonate (ABC) buffer, pH 8] and alkylation with iodoacetamide (10 mg/ml in 8 M guanidinium hydrochloride in 50 mM ABC buffer), samples werewashed and 1 µg trypsin was added before incubation at 37°C for 18 h. Afterenzymatic digestion, peptide samples were cleaned with C18 spin columns(Pierce, ThermoScientific,Germany), dried and stored at−20°C until analysis.

For mass spectrometric analyses, dried samples were reconstituted in 5 µl30% formic acid (FA) containing 10 fmol each of four synthetic standardpeptides and diluted with 40 µl mobile phase A (H2O:ACN:FA=98:2:0.1).A 10 µl aliquot of the peptide solution was loaded onto a 2 cm×75 µm C18Pepmap100 precolumn (Thermo Fisher Scientific, Austria) at a flow rate of10 µl/min using mobile phase A. Afterwards, peptides were eluted from theprecolumn to a 50 cm×75 µm Pepmap100 analytical column (Thermo FisherScientific, Austria) at a flow rate of 300 nl/min and separation was achievedusing a gradient of 8% to 40% mobile phase B (ACN:H2O:FA=80:20:0.1)over 95 min. For mass spectrometric analyses, MS scans were performed inthe range of m/z 400-1400 at a resolution of 70,000 (at m/z=200). MS/MSscans of the eight most abundant ions were achieved through high-energycollisional dissociation fragmentation at 30% normalized collision energy andanalysed in the orbitrap at a resolution of 17,500 (at m/z=200). All sampleswere analysed in duplicate.

Data analysis and statistics of MS experimentsProtein identification and label-free quantitative (LFQ) data analysis wereperformed using the open source software MaxQuant 1.3.0.5 including theAndromeda search engine (Cox and Mann, 2008). Protein identificationwas achieved searching againstOvis aries in the Uniprot Database (version09/2014 with 26,864 entries) allowing a mass tolerance of 5 ppm for MSspectra and 20 ppm for MS/MS spectra, as well as a maximum of twomissed cleavages. In addition, carbamidomethylation on cysteine residueswas included as a fixed modification, whereas methionine oxidation andN-terminal protein acetylation were included as variable modifications.

Furthermore, search criteria included aminimumof two peptide identificationsper protein, at least one of them unique, and the false discovery rate (FDR)calculation based on q-values, performed for both peptide identification aswell as protein identification, less than 0.01. Prior to statistical analyses,proteins were filtered for reversed sequences, contaminants and a minimum ofthree independent identifications per protein.

Missing values were replaced by a global ε, set to the minimum intensityobserved in the entire data set divided by four. This sensitivity-basedpseudo-count reflects the prior belief of non-observed protein expression,maintaining a lower bound of a fourfold change for differences to proteinsnot observed in one sample group, thus maintaining sensitivity whileimproving specificity by mitigating the effects of random non-observations.The Spearman rank correlations between samples shown in Fig. 6 are notaffected by this transform. For the visualization of the sample correlationstructure in Fig. 6, ellipses were plotted as (x, y)=(cos(θ+d/2), cos(θ-d/2)),where θ ɛ [0,2π) and cos(d )=ρ, with ρ the Spearman rank correlationcoefficient (Murdoch and Chow, 1996).

Protein expression profile analysis was performed in the statisticalenvironment R (www.r-project.org). Differential expression contrasts werecomputed with Bioconductor libraries (www.bioconductor.org). Data werenormalized by a mean log-shift, standardizing mean expression levels persample. Linear models were fitted separately for each protein, computingsecond-level contrasts for a direct test of differences between fetal and adultresponses to injury. Conservative Benjamini-Yekutieli correctionwas used toadjust for multiple testing to give strong control of the FDR. We callsignificant features for q-values <0.05. Linear models were adjusted forthe nested correlation structure of technical and biological replicates(see Fig. 6). Significance was assessed by regularized t-tests. For these,group variances are shrunk by an Empirical Bayes procedure to mitigate thehigh uncertainty of variance estimates for the available sample sizes(Sun et al., 2009). The employed algorithms are implemented in the packagelimma (Smyth, 2005), which is available from Bioconductor.

AcknowledgementsWe would like to acknowledge Ms Manuela Miksic, Ms Manuela Hogn, Ms BrigitteMachac and Ms Claudia Hochsmann for their technical support.

Competing interestsThe authors declare no competing or financial interests.

Author contributionsConceptualization: I.R., F.J.; Methodology: I.R., D.P.K., C.G., F.J.; Software:D.P.K.; Validation: D.P.K., C.G., F.J.; Formal analysis: R.L.M., M.M.K., A.B., D.P.K.,C.G.; Investigation: I.R., R.L.M., M.E., S. Gabner, J.R., E.H., U.A., S. Gueltekin, J.H.,A.B., C.G., F.J.; Resources: M.E., D.P.K., C.G., F.J.; Data curation: D.P.K., C.G.;Writing - original draft: F.J.; Writing - review & editing: I.R., R.L.M., M.E., S. Gabner,M.M.K., J.R., E.H., U.A., S. Gueltekin, J.H., A.B., D.P.K., C.G., F.J.; Visualization:M.E., S. Gabner; Supervision: F.J.; Project administration: I.R., F.J.; Fundingacquisition: I.R., D.P.K., C.G., F.J.

FundingThis research was funded by the Austrian Research Promotion Agency ‘FFG’

(project number: 856858) and supported by the University of Veterinary MedicineVienna and the Faculty of Chemistry of the University of Vienna. Otherwise wereceived no further specific grant from any funding agency in the public, commercial,or not-for-profit sectors.

Supplementary informationSupplementary information available online athttp://dmm.biologists.org/lookup/doi/10.1242/dmm.033092.supplemental

Table 2. Sources, pre-treatments and dilutions of the antibodies used for histological analysis

Antibody Clone Concentration (v/v) Pre-treatment Source

COL2 2B1.5 1/100 0.04% Hyaluronidase* in PBS (4 h at 37°C) followed by1% protease* in PBS (30 min at 37°C)

Thermo Fisher Scientific,Waltham, MA, USA

Ki67 SP6 1/400 0.01 M Citrate buffer pH 6.0 (2 h at 85°C) Thermo Fisher ScientificMMP9 poly 1/100 0.01 M Citrate buffer pH 6.0 (30 min at 90-95°C) Abnova, Heidelberg, GermanyMMP13 poly 1/50 0.01 M Citrate buffer pH 6.0 (30 min at 90-95°C) Thermo Fisher Scientific

*Sigma Aldrich. PBS, phosphate-buffered saline.

9

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033092. doi:10.1242/dmm.033092

Disea

seModels&Mechan

isms

ReferencesAlmeida-Porada, G., Porada, C. and Zanjani, E. D. (2004). Plasticity of human stemcells in the fetal sheep model of human stem cell transplantation. Int. J. Hematol.79, 1-6.

Almeida-Porada, G., Porada, C., Gupta, N., Torabi, A., Thain, D. and Zanjani,E. D. (2007). The human–sheep chimeras as a model for human stem cellmobilization and evaluation of hematopoietic grafts’ potential. Exp. Hematol. 35,1594-1600.

Al-Qattan, M. M., Posnick, J. C., Lin, K. Y. and Thorner, P. (1993). Fetal tendonhealing: development of an experimental model. Plast. Reconstr. Surg. 92, 1155.

Arita, N., Pelaez, D. and Cheung, H. (2011). Activation of the extracellular signal-regulated kinases 1 and 2 (ERK1/2) is needed for the TGFβ-induced chondrogenicand osteogenic differentiation ofmesenchymal stem cells.Biochem.Biophys. Res.Commun. 405, 564-69.

Aukland, K. (1991). Distribution volumes and macromolecular mobility in rat tailtendon interstitium. Am. J. Physiol. 260, H409-H419.

Aukland, K., Reed, R. K. and Wiig, H. (1997a). The problem of gaining access tointerstitial fluid. An attempt to rationalize a wicked discussion on wicks.Lymphology 30, 111-115.

Aukland, K., Wiig, H., Tenstad, O. and Renkin, E. M. (1997b). Interstitial exclusionof macromolecules studied by graded centrifugation of rat tail tendon.Am. J. Physiol. 273, H2794-H2803.

Barbour, K. E., Helmick, C., Theis, K. A., Murphy, L., Hootman, J., Brady, T. J.and Cheng, Y. (2013). Prevalence of doctor-diagnosed arthritis and arthritis-attributable activity limitation—United States, 2010–2012. MMWR 62, 869-873.

Beier, F. and Loeser, R. F. (2010). Biology and pathology of Rho GTPase, PI-3kinase-Akt, and MAP kinase signaling pathways in chondrocytes. J. Cell.Biochem. 110, 573-580.

Benink, H. A. and Bement, W. M. (2005). Concentric zones of active RhoA andCdc42 around single cell wounds. J. Cell Biol. 168, 429-439.

Bileck, A., Kreutz, D., Muqaku, B., Slany, A. and Gerner, C. (2014).Comprehensive assessment of proteins regulated by dexamethasone revealsnovel effects in primary human peripheral blood mononuclear cells. J. ProteomeRes. 13, 5989-6000.

Bobick, B. E., Matsche, A. I., Chen, F. H. and Tuan, R. S. (2010). The ERK5 andERK1/2 signaling pathways play opposing regulatory roles during chondrogenesisof adult human bone marrow-derived multipotent progenitor cells. J. Cell. Physiol.224, 178-186.

Bruns, J., Kampen, J., Kahrs, J. and Plitz, W. (2000). Achilles tendon rupture:experimental results on spontaneous repair in a sheep-model. Knee Surg. SportsTraumatol. Arthrosc. 8, 364-369.

Chen, J., Crawford, R. andXiao, Y. (2013). Vertical inhibition of the PI3K/Akt/mTORpathway for the treatment of osteoarthritis. J. Cell. Biochem. 114, 245-249.

Cowin, A. J., Brosnan, M. P., Holmes, T. M. and Ferguson, M. W. (1998).Endogenous inflammatory response to dermal wound healing in the fetal andadult mouse. Dev. Dyn. 212, 385-393.

Cox, J. and Mann, M. (2008). MaxQuant enables high peptide identification rates,individualized p.p.b.-range mass accuracies and proteome-wide proteinquantification. Nat. Biotechnol. 26, 1367-1372.

Decker, R. S., Um, H.-B., Dyment, N. A., Cottingham, N., Usami, Y., Enomoto-Iwamoto,M., Kronenberg, M. S., Maye, P., Rowe, D.W., Koyama, E. et al. (2017).Cell origin, volume and arrangement are drivers of articular cartilage formation,morphogenesis and response to injury in mouse limbs. Dev. Biol. 426, 56-68.

Degen, K. E. and Gourdie, R. G. (2012). Embryonic wound healing: a primer forengineering novel therapies for tissue repair. Birth Defect Res. C 96, 258-270.

Dorotka, R., Bindreiter, U., Macfelda, K., Windberger, U. and Nehrer, S. (2005).Marrow stimulation and chondrocyte transplantation using a collagen matrix forcartilage repair. Osteoarthritis Cartilage 13, 655-664.

Duan, X., Liang, Y.-Y., Feng, X.-H. and Lin, X. (2006). Protein serine/threoninephosphatase PPM1A dephosphorylates Smad1 in the bone morphogeneticprotein signaling pathway. J. Biol. Chem. 281, 36526-36532.

Dunkel, Y., Ong, A., Notani, D., Mittal, Y., Lam, M., Mi, X. and Ghosh, P.(2012). STAT3 protein up-regulates Gα-interacting vesicle-associated protein(GIV)/Girdin expression, and GIV enhances STAT3 activation in a positivefeedback loop during wound healing and tumor invasion/metastasis. J. Biol.Chem. 287, 41667-41683.

Dziegielewska, K. M., Møller, J. E., Potter, A. M., Ek, J., Lane, M. A. andSaunders, N. R. (2000). Acute-phase cytokines IL-1beta and TNF-alpha in braindevelopment. Cell Tissue Res. 299, 335-345.

Emmert, M. Y., Weber, B., Wolint, P., Frauenfelder, T., Zeisberger, S. M.,Behr, L., Sammut, S., Scherman, J., Brokopp, C. E., Schwartlander, R. et al.(2013). Intramyocardial transplantation and tracking of humanmesenchymal stemcells in a novel intra-uterine pre-immune fetal sheep myocardial infarction model:a proof of concept study. PLoS ONE 8, e57759.

Entrican, G., Wattegedera, S. R. and Griffiths, D. J. (2015). Exploiting ovineimmunology to improve the relevance of biomedical models. Mol. Immunol.66, 68-77.

Ferguson, M. W. J. and O’Kane, S. (2004). Scar-free healing: from embryonicmechanisms to adult therapeutic intervention. Philos. Trans. R. Soc. B Biol. Sci.359, 839-850.

Fujita, T., Azuma, Y., Fukuyama, R., Hattori, Y., Yoshida, C., Koida, M., Ogita, K.and Komori, T. (2004). Runx2 induces osteoblast and chondrocyte differentiationand enhances their migration by coupling with PI3K-Akt signaling. J. Cell Biol.166, 85-95.

Ghosh, P., Garcia-Marcos, M., Bornheimer, S. J. and Farquhar, M. G. (2008).Activation of Galphai3 triggers cell migration via regulation of GIV. J. Cell Biol. 182,381-393.

Goldring, M. B., Otero, M., Plumb, D. A., Dragomir, C., Favero, M., ElHachem, K., Hashimoto, K., Roach, H. I., Olivotto, E., Borzı, R. M. et al.(2011). Roles of inflammatory and anabolic cytokines in cartilage metabolism:signals and multiple effectors converge upon MMP-13 regulation in osteoarthritis.Eur. Cell Mater. 21, 202-220.

Greene, M. A. and Loeser, R. F. (2015). Function of the chondrocyte PI-3 kinase-Aktsignaling pathway is stimulus dependent. Osteoarthritis Cartilage 23, 949-956.

Jeanblanc, C., Goodrich, A. D., Colletti, E., Mokhtari, S., Porada, C. D., Zanjani,E. D. and Almeida-Porada, G. (2014). Temporal definition of haematopoieticstem cell niches in a large animal model of in utero stem cell transplantation.Br. J. Haematol. 166, 268-278.

Jenner, F., Ijpma, A., Cleary, M., Heijsman, D., Narcisi, R., van der Spek, P. J.,Kremer, A., van Weeren, P. R., Brama, P. and van Osch, G. J. V. M. (2014a).Differential gene expression of the intermediate and outer interzone layers ofdeveloping articular cartilage in murine embryos. Stem Cells Dev. 23, 1883-1898.

Jenner, F., van Osch, G., Cleary, M., Ribitsch, I., Sauer, U., van Weeren, P. R.and Brama, P. (2014b) Laser capture microdissection of murine interzone cells:layer selection and prediction of RNA yield. J. Stem Cell Res. Ther. 4, 1000183.

Johnson, V. L. and Hunter, D. J. (2014). The epidemiology of osteoarthritis. BestPract. Res. Clin. Rheumatol. 28, 5-15.

Kennedy, J. M., Fodil, N., Torre, S., Bongfen, S. E., Olivier, J.-F., Leung, V.,Langlais, D., Meunier, C., Berghout, J., Langat, P. et al. (2014). CCDC88B is anovel regulator of maturation and effector functions of T cells during pathologicalinflammation. J. Exp. Med. 211, 2519-2535.

Kim, J., Zanjani, E. D., Jeanblanc, C. M., Goodrich, A. D. and Hematti, P. (2013).Generation of CD34+ cells from human embryonic stem cells using a clinicallyapplicable methodology and engraftment in the fetal sheep model. Exp. Hematol.41, 749-758.e5.

Kita, K., Kimura, T., Nakamura, N., Yoshikawa, H. and Nakano, T. (2008). PI3K/Akt signaling as a key regulatory pathway for chondrocyte terminal differentiation.Genes Cells 13, 839-850.

Kozhemyakina, E., Zhang, M., Ionescu, A., Ayturk, U. M., Ono, N., Kobayashi,A., Kronenberg, H., Warman, M. L. and Lassar, A. B. (2015). Identification of aPrg4-expressing articular cartilage progenitor cell population in mice. ArthritisRheumatol 67, 1261-1273.

Kumahashi, N., Ochi, M., Kataoka, H., Uchio, Y., Kakimaru, H., Sugawara, K.and Enomoto, K.-I. (2004). Involvement of ATP, increase of intracellular calciumand the early expression of c-fos in the repair of rat fetal articular cartilage. CellTissue Res. 317, 117-128.

Kumta, S., Ritz, M., Hurley, J. V., Crowe, D., Romeo, R. andO’Brien, B. M. (1994).Acute inflammation in foetal and adult sheep: the response to subcutaneousinjection of turpentine and carrageenan. Br. J. Plast. Surg. 47, 360-368.

Kuypers, E., Ophelders, D., Jellema, R. K., Kunzmann, S., Gavilanes, A. W. andKramer, B. W. (2012). White matter injury following fetal inflammatory responsesyndrome induced by chorioamnionitis and fetal sepsis: Lessons fromexperimental ovine models. Early Hum. Dev. 88, 931-936.

Liechty, K. W., Mackenzie, T. C., Shaaban, A. F., Radu, A., Moseley, A. M., Deans,R., Marshak, D. R. and Flake, A. W. (2000). Human mesenchymal stem cellsengraft and demonstrate site-specific differentiation after in utero transplantation insheep. Nat. Med. 6, 1282-1286.

Litherland, G. J., Dixon, C., Lakey, R. L., Robson, T., Jones, D., Young, D. A.,Cawston, T. E. and Rowan, A. D. (2008). Synergistic collagenase expressionand cartilage collagenolysis are phosphatidylinositol 3-kinase/Akt signaling-dependent. J. Biol. Chem. 283, 14221-14229.

Little, C. B., Smith, M. M., Cake, M. A., Read, R. A., Murphy, M. J. andBarry, F. P. (2010). The OARSI histopathology initiative-recommendations forhistological assessments of osteoarthritis in sheep and goats. OsteoarthritisCartilage 18 Suppl. 3, S80-S92.

Liu-Bryan, R. and Terkeltaub, R. (2015). Emerging regulators of the inflammatoryprocess in osteoarthritis. Nat. Rev. Rheumatol. 11, 35-44.

Lo, D. D., Zimmermann, A. S., Nauta, A., Longaker,M. T. andLorenz, H. P. (2012).Scarless fetal skin wound healing update. Birth Defect Res. C 96, 237-247.

Longaker, M. T., Whitby, D. J., Adzick, N. S., Crombleholme, T. M., Langer, J. C.,Duncan, B. W., Bradley, S. M., Stern, R., Ferguson, M. W. J. and Harrison,M. R. (1990). Studies in fetal wound healing VI. Second and early third trimesterfetal wounds demonstrate rapid collagen deposition without scar formation.J. Pediatr. Surg. 25, 63-69.

Longaker, M. T., Moelleken, B. R. W., Cheng, J. C., Jennings, R. W., Adzick,S. N., Mintorovich, J., Levinsohn, D. G., Gordon, L., Harrison, M. R. andSimmons, D. J. (1992). Fetal fracture healing in a lamb model. Plast. Reconstr.Surg. 90, 161-171.

Lopez, A. D. and Murray, C. C. (1998). The global burden of disease, 1990-2020.Nat. Med. 4, 1241-1243.

10

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033092. doi:10.1242/dmm.033092

Disea

seModels&Mechan

isms

Lopez-Sanchez, I., Dunkel, Y., Roh, Y.-S., Mittal, Y., De Minicis, S., Muranyi, A.,Singh, S., Shanmugam, K., Aroonsakool, N., Murray, F. et al. (2014). GIV/Girdin is a central hub for profibrogenic signalling networks during liver fibrosis.Nat. Commun. 5, 4451.

Lu, Y., Markel, M. D., Swain, C. and Kaplan, L. D. (2006). Development of partialthickness articular cartilage injury in an ovine model. J. Orthop. Res. 24, 1974-1982.

Mackay, C. R., Maddox, J. F. and Brandon, M. R. (1986). Thymocyte subpopulationsduring early fetal development in sheep. J. Immunol. 136, 1592-1599.

Maddox, J. F., Mackay, C. R. and Brandon, M. R. (1987). Ontogeny of ovinelymphocytes. An immunohistological study on the development of T lymphocytesin the sheep embryo and fetal thymus. Immunology 62, 97-118.

Mainil-Varlet, P., Aigner, T., Brittberg,M., Bullough, P., Hollander, A., Hunziker, E.,Kandel, R., Nehrer, S., Pritzker, K., Roberts, S. et al. (2003). Histologicalassessment of cartilage repair: a report by the histology endpoint committee of theinternational cartilage repair society (ICRS). J. Bone Joint Surg. Am. 85-A Suppl. 2,45-57.

Miyasaka, M. and Trnka, Z. (1986). Lymphocyte migration and differentiation in alarge-animal model: the sheep. Immunol. Rev. 91, 87-114.

Moss, T. J. M., Knox, C. L., Kallapur, S. G., Nitsos, I., Theodoropoulos, C.,Newnham, J. P., Ikegami, M. and Jobe, A. H. (2008). Experimental amniotic fluidinfection in sheep: Effects of Ureaplasma parvum serovars 3 and 6 on preterm orterm fetal sheep. Am. J. Obstet. Gynecol. 198, 122.e1-122.e8.

Mrugala, D., Bony, C., Neves, N., Caillot, L., Fabre, S., Moukoko, D., Jorgensen, C.and Noel, D. (2008). Phenotypic and functional characterisation of ovinemesenchymal stem cells: application to a cartilage defect model. Ann. Rheum. Dis.67, 288-295.

Mufti, A. M., Wani, G. M., Wani, N. A., Buchoo, B. A. and Khan, M. Z. (2000).Prenatal development of ovine fetus. Small Ruminant Res. 38, 87-89.

Mulisch, M. and Welsch, U. eds. (2015). Romeis-Mikroskopische Technik, 19thedn. Berlin, Heidelberg: Springer Berlin Heidelberg.

Munirah, S., Samsudin, O. C., Chen, H. C., Salmah, S. H. S., Aminuddin, B. S.and Ruszymah, B. H. I. (2007). Articular cartilage restoration in load-bearingosteochondral defects by implantation of autologous chondrocyte-fibrin constructs:an experimental study in sheep. J. Bone Joint Surg. Br. 89, 1099-1109.

Murdoch, D. J. and Chow, E. D. (1996). A graphical display of large correlationmatrices. The American Statistician 50, 178.

Nakashima, M., Sakai, T., Hiraiwa, H., Hamada, T., Omachi, T., Ono, Y., Inukai,N., Ishizuka, S., Matsukawa, T., Oda, T. et al. (2012). Role of S100A12 in thepathogenesis of osteoarthritis. Biochem. Biophys. Res. Commun. 422, 508-514.

Namba, R. S., Meuli, M., Sullivan, K. M., Le, A. X. and Adzick, N. S. (1998).Spontaneous repair of superficial defects in articular cartilage in a fetal lambmodel. J. Bone Joint Surg. Am. 80, 4-10.

NCBI Resource Coordinators. (2018). NCBI database resources of the nationalcenter for biotechnology information. Nucleic Acids Res. 46, D8-D13.

Nefla, M., Holzinger, D., Berenbaum, F. and Jacques, C. (2016). The danger fromwithin: alarmins in arthritis. Nat. Rev. Rheumatol. 12, 669-683.

Nitsos, I., Rees, S., Duncan, J., Kramer, B., Harding, R., Newnham, J. andMoss,T. (2006). Chronic exposure to intra-amniotic lipopolysaccharide affects the ovinefetal brain. Reprod. Sci. 13, 239-247.

Orth, P. and Madry, H. (2013). A low morbidity surgical approach to the sheepfemoral trochlea. BMC Musculoskelet. Disord. 14, 1511-1518.

Osburn, B. I. (1981). The ontogeny of the ruminant immune system and itssignificance in the understanding of maternal-fetal-neonatal relationships. Adv.Exp. Med. Biol. 137, 91-103.

Pelaez, D., Arita, N. and Cheung, H. (2012). Extracellular signal-regulated kinase(ERK) dictates osteogenic and/or chondrogenic lineage commitment ofmesenchymal stem cells under dynamic compression. Biochem. Biophys. Res.Commun. 417, 1286-1291.

Pap, T. and Korb-Pap, A. (2015). Cartilage damage in osteoarthritis andrheumatoid arthritis-two unequal siblings. Nat. Rev. Rheumatol. 11, 606-615.

Porada, C. D., Park, P. J., Almeida-Porada, G., Liu, W., Ozturk, F., Glimp, H. A. andZanjani, E. D. (2005). Gestational age of recipient determines pattern and level oftransgene expression following in utero retroviral gene transfer. Mol. Ther.11, 284-293.

Rabenda, V., Manette, C., Lemmens, R., Mariani, A.-M., Struvay, N. and Reginster,J.-Y. (2007). Prevalence and impact of osteoarthritis and osteoporosis on health-related quality of life among active subjects. Aging Clin. Exp. Res. 19, 55-60.

Ritter, S. Y., Subbaiah, R., Bebek, G., Crish, J., Scanzello, C. R., Krastins, B.,Sarracino, D., Lopez, M. F., Crow, M. K., Aigner, T. et al. (2013). Proteomicanalysis of synovial fluid from the osteoarthritic knee: comparison withtranscriptome analyses of joint tissues. Arthritis. Rheum. 65, 981-992.

Rohani, M. G., Pilcher, B. K., Chen, P. and Parks, W. C. (2014). Cdc42 inhibitsERK-mediated collagenase-1 (MMP-1) expression in collagen-activated humankeratinocytes. J. Invest. Dermatol. 134, 1230-1237.

Rosmarin, A. (2004). GA-binding protein transcription factor: a review of GABP asan integrator of intracellular signaling and protein–protein interactions. Blood CellsMol. Dis. 32, 143-154.

Russo, V.,Mauro, A., Martelli, A., DiGiacinto, O., DiMarcantonio, L., Nardinocchi,D., Berardinelli, P. and Barboni, B. (2015). Cellular and molecular maturation infetal and adult ovine calcaneal tendons. J. Anat. 226, 126-142.

Salami, S. O., Ibe, C. S., Umosen, A. D., Ajayi, I. E. and Maidawa, S. M. (2011).Comparative osteometric study of long bones in yankasa sheep and red sokotogoats. Int. J. Morphol. 29, 100-104.

Sawyer, M., Moe, J. and Osburn, B. I. (1978). Ontogeny of immunity andleukocytes in the ovine fetus and elevation of immunoglobulins related tocongenital infection. Am. J. Vet. Res. 39, 643-648.

Schelbergen, R. F. P., Blom, A. B., van den Bosch, M. H. J., Sloetjes, A.,Abdollahi-Roodsaz, S., Schreurs, B. W., Mort, J. S., Vogl, T., Roth, J., van denBerg, W. B. et al. (2012). Alarmins S100A8 and S100A9 elicit a catabolic effect inhuman osteoarthritic chondrocytes that is dependent on Toll-like receptor 4.Arthritis. Rheum. 64, 1477-1487.

Silverstein, A. M., Uhr, J. W., Kraner, K. and Lukes, R. (1963). Fetal responseto antigenic stimulus. II. Antibody production by the fetal lamb. J. Exp. Med.117, 799-812.

Silverstein, A. M., Prendergast, R. and Kraner, K. (1964). Fetal response toantigenic stimulus. IV. Rejection of skin homografts by the fetal lamb. J. Exp. Med.119, 955-964.

Smyth, G. K. (2005). limma: linear models for microarray data. In Bioinformatics andComputational Biology Solutions Using R and Bioconductor (ed. R. Gentleman, V.J.Carey, W. Huber, R. A. Irizarry, S. Dudoit), pp. 397-420. New York: Springer.

Starkman, B. G., Cravero, J. D., Delcarlo, M. and Loeser, R. F. (2005). IGF-Istimulation of proteoglycan synthesis by chondrocytes requires activation of the PI3-kinase pathway but not ERK MAPK. Biochem. J. 389, 723-729.

Stenberg, J., Ruetschi, U., Skioldebrand, E., Karrholm, J. and Lindahl, A.(2013). Quantitative proteomics reveals regulatory differences in the chondrocytesecretome from human medial and lateral femoral condyles in osteoarthriticpatients. Proteome Science 11, 43.

Stone, C. A. (2000). Unravelling the secrets of foetal wound healing: an insight intofracture repair in the mouse foetus and perspectives for clinical application.Br. J. Plast. Surg. 53, 337-341.

Sun, W., Yu, Y., Dotti, G., Shen, T., Tan, X., Savoldo, B., Pass, A., Chu, M.,Zhang, D. and Lu, X. (2009). PPM1A and PPM1B act as IKKβ phosphatases toterminate TNFα-induced IKKβ-NF-κB activation. Cell. Signal. 21, 95-102.

Ueda, A., Akagi, T. and Yokota, T. (2017). GA-binding protein alpha is involved inthe survival of mouse embryonic stem cells. Stem Cells 306, 391.

van den Bosch, M. H., Blom, A. B., Schelbergen, R. F. P., Vogl, T., Roth, J. P.,Sloetjes, A. W., van den Berg, W. B., van der Kraan, P. M. and van Lent, P. L. E.M. (2015). Induction of canonicalWnt signaling by the alarmins S100A8/A9 inmurineknee joints: implications for osteoarthritis. Arthritis Rheumatol. 68, 152-163.

van Turnhout, M., Schipper, H., Engel, B., Buist,W. andKranenbarg, S. (2010a).Postnatal development of collagen structure in ovine articular cartilage. BMCDev.Biol. 10, 62.

van Turnhout, M., Schipper, H., van Lagen, B., Zuilhof, H., Kranenbarg, S. andvan Leeuwen, J. L. (2010b). Postnatal development of depth-dependent collagendensity in ovine articular cartilage. BMC Dev. Biol. 10, 108.

Wagner, W., Reichl, J., Wehrmann, M. and Zenner, H. (2001). Neonatal ratcartilage has the capacity for tissue regeneration. Wound Repair. Regen. 9,531-536.

Walker, E. A., Verner, A., Flannery, C. R. and Archer, C. W. (2000). Cellularresponses of embryonic hyaline cartilage to experimental wounding in vitro.J. Orthop. Res. 18, 25-34.

Wang, Q., Breinan, H. A., Hsu, H.-P. and Spector, M. (2000). Healing of defects incanine articular cartilage: distribution of nonvascular alpha-smooth muscle actin-containing cells. Wound Repair. Regen. 8, 145-158.

Wang, J., Tucker-Kellogg, L., Ng, I. C., Jia, R., Thiagarajan, P. S., White, J. K.and Yu, H. (2014). The self-limiting dynamics of TGF-β signaling in silico and invitro, with negative feedback through PPM1A upregulation. PLoS Comput. Biol.10, e1003573.

Wang, J. R., Wang, C. J., Xu, C. Y., Wu, X. K., Hong, D., Shi, W., Gong, Y.,Chen, H. X., Long, F., Wu, X. et al. (2016). Signaling cascades governingCdc42-mediated chondrogenic differentiation and mensenchymal condensation.Genetics 202, 1055-1069.

Wen Chun Juan, W. H. (2016). Targeting the hippo signaling pathway for tissueregeneration and cancer therapy. Genes 7, 55.

Wilson, R. and Bateman, J. F. (2008). Cartilage proteomics: challenges, solutionsand recent advances. Prot. Clin. Appl. 2, 251-263.

Wi�sniewski, J. R., Zougman, A., Nagaraj, N. and Mann, M. (2009). Universalsample preparation method for proteome analysis. Nat. Meth. 6, 359-362.

Woolf, A. D. and Pfleger, B. (2003). Burden of major musculoskeletal conditions.Bull. World Health Organ. 81, 646-656.

Xu, J., Yi, Y., LI, L., Zhang, W. andWang, J. (2015). Osteopontin induces vascularendothelial growth factor expression in articular cartilage through PI3K/AKT andERK1/2 signaling. Mol. Med. Rep. 12, 4708-4712.

Xue, X., Zheng, Q., Wu, H., Zou, L. and Li, P. (2013). Different responses tomechanical injury in neonatal and adult ovine articular cartilage. Biomed. Eng.Online 12, 53.

Yelin, E., Murphy, L., Cisternas, M. G., Foreman, A. J., Pasta, D. J. and Helmick,C. G. (2007). Medical care expenditures and earnings losses among persons witharthritis and other rheumatic conditions in 2003, and comparisons with 1997.Arthritis. Rheum. 56, 1397-1407.

11

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm033092. doi:10.1242/dmm.033092

Disea

seModels&Mechan

isms