TGF-β Type II Receptor/MCP-5 Axis: at the Crossroad between Joint and Growth Plate Development

Aberrant release of MCP-5 within normal developing limbs alters joint formation, while blockade of CCR2 signaling is able to rescue the joint phenotype of Tgfbr2^Prx1KO mutants

In order to determine a role for the restricted levels of MCP-5 within the interzone, we used an ex-vivo approach in which hydrophobically-modified glycol chitosan (HGC)-nanoparticles loaded with MCP-5 protein were microinjected in the interzone of E14.5 WT dissected autopods (Figure 2A) to lead to an aberrant sustained release of MCP-5 within the developing joints. HGC-nanoparticles are biocompatible and non immunogenic and consist in hydrophilic shells and hydrophobic cores that bind proteins (Hirano, 1999; Li et al., 2010). We have developed a micro-injection unit that allows for precise injection localization and volume accuracy delivery (nanoliters) within distinct regions of developing autopods and reported that proteins are slowly released from HGC without diffusing (~90% in 48h), are biologically active and HGC do not interfere with skeletogenesis, including joint development (Li et al., 2010). Based on these pharmacokinetics studies, HGC-MCP-5 (delivering 1.5 ng of MCP-5 protein) was injected at the joint sites of E14.5 dissected autopods that were cultured for 2 days; un-injected joints from the same embryo were used as control (Figure 2A). IHC for MCP-5 showed that protein was sustainably delivered (after 48 h) within the interzone; Figure 2A shows that MCP-5 was detectable in the interphalangeal joint sections from the 2^nd digit implanted with HCG-MCP-5, while it was undetectable in the uninjected control joint interzone of the 5^th digit, in which endogenous MCP-5 was restricted to the adjacent growth plate chondrocytes. Compared to un-injected control, the aberrant sustained delivery of MCP-5 from HCG-MCP-5 implant within the developing interzone abolished Gdf5 expression and altered the interzone Collagen2 expression pattern. In fact, the interzone of implanted HCG-MCP-5 digits lost the characteristic chondrocyte segmentation and Collagen2 expressing cells appeared continuously distributed along the entire digit (Figure 2B). Increasing doses (0.75 to 1.5 ng) of implanted MCP-5 within the interzone resulted in gradual impairing of Gdf5 and Collagen2 patterns. Similar effects on Gdf5 and Collagen2 expressions were found among the 1.5 ng to the 6 ng doses, therefore 1.5 ng was used in all further experiments (data not shown). At least 3 embryos were implanted with each dose and at least 3 sections from each embryo were subjected to ISH analyses.

Open in a separate window

Figure 2

Aberrant expression of Gdf5 and Collagen2 in E14.5 WT embryos after HGC-MCP-5 limb implantation and ex-vivo culture

(A–B) Recombinant MCP-5 was loaded in HGC-nanoparticle and the HGC-MCP-5 complex (MCP-5, 1.5 ng) was implanted in the 2^nd, 3^rd and 4^th digit (highlighted by Methylene Blue) of dissected E14.5 WT. Injected and uninjected (5^th digit) limbs were cultured for 48h, embedded in paraffin and subjected to (A) IHC analyses using MCP-5 antibody and (B) ISH studies using Gdf5 and Collagen2 (Col2) probes (proximal side: right). Eight out of nine WT embryos implanted with HGC-MCP-5 showed an altered phenotype, as described in the Results; analyses were performed in at least 3 sections for each embryo and for each probe.

To determine whether the blockade of CCR2 (the sole MCP5 receptor) would restore interphalangeal joint formation in Tgfbr2^Prx1KO developing limbs, we used ex-vivo as wells as in-vivo approaches. We first evaluated CCR2 expression by IHC at different joint developmental time points (E12.5; E13.5, E14.5 and E18.5). CCR-2 expression was first detected within the developing digit at E13.5 and by E14.5 expression was found within the developing interzone and growth plates. This expression pattern was maintained up to E18.5 (Figure S2). To block the CCR2 signaling, we used the HCG-based ex-vivo autopod culture technique; in particular, we implanted HCG-MC-21 nanoparticles (delivering 1.5ng of MC-21 antibody) in the presumptive interzone of E14.5 Tgfbr2^Prx1KO mutants as well as NC littermates. The MC-21 is an anti-mouse CCR-2 monoclonal antibody that has been shown to block the murine CCR-2 with high affinity in vivo as well as in vitro (Bruhl et al., 2004; Mack et al., 2001). Histological and ISH studies demonstrated that blockade of MCP-5 signaling was able to restore the joint cavitation and Gdf5 expression abnormally missing in the mutants. In addition, while in mutant limbs Sox-9 and Collagen2 expressing cells were continuously distributed along the digit, in MC-21-implanted mutants they appeared to be restricted to the growth plate chondrocytes, adopting an expression pattern that was similar to controls (Figure 3A). For the in-vivo studies, pregnant females potentially carrying Tgfbr2^Prx1KO embryos were treated orally with 4mg/Kg/day of RS-504393 (a selective CCR2 inhibitor) starting at E11.5 and until either E14.5 or E18.5, when mice were sacrificed. The two time points allowed to determine the need of CCR-2 signaling in joint formation (E14.5) as well as its persistent effect at a later stage (E18.5). As shown in Figure 3B, blockade of MCP-5 signaling in Tgfbr2^Prx1KO embryos in-vivo was sufficient to restore Gdf5 expression and Collagen2 segmentation in E14.5 embryos. These effects were maintained in E18.5 Tgfbr2^Prx1KO embryos resulting in a well-defined joint space in between the growth plates (Figure 3C). Blockade of MCP-5 signaling in NC limbs, both ex-vivo and in-vivo, did not affect joint maturation and chondrocyte segmentation (Figure 3A–B). Our results indicate that a down-regulation of MCP-5 in developing interzone is needed for proper joint formation, Gdf5 and Collagen2 expressions and reveal the critical role for TGF-β signaling in controlling restrained levels of MCP-5 within the developing joints.

Open in a separate window

Figure 3

Rescue of Tgfbr2^Prx1KO joint phenotype by ex-vivo HGC-MC-21 limb implantation and RS-504393 in-vivo treatment

(A) MC-21 antibody was loaded into HGC-nanoparticle and the resulting HGC-MC-21 complex (MC-21, 1.5 ng) was implanted in the interzone of dissected E14.5 NC and Tgfbr2^Prx1KO digits. Limbs were cultured for 48h, embedded in paraffin and subjected either to hematoxylin staining or ISH analyses using Gdf-5, Col2 and Sox9 probes. (B–C) RS-504393 treatments were started at E11.5 on pregnant females potentially carrying Tgfbr2^Prx1KO embryos and continued until either (B) E14.5 (3 days treatment) or (C) E18.5 (7 days of treatment); embryos were then harvested and genotyped. NC and Tgfbr2^Prx1KO limbs were dissected, embedded in paraffin and subjected either to hematoxylin staining or ISH analyses using Gdf-5 and Col2 probes (proximal side: left). All the 4 Tgfbr2^Prx1KO embryos implanted with HGC-MC-21 showed the phenotype described in the Results; for the RS-504393 treatments, all the 4 Tgfbr2^Prx1KO embryos treated for 3 days and all the 9 Tgfbr2^Prx1KO embryos treated for 7 days showed the phenotype described in the Results; analyses were performed in at least 3 sections for each embryo and for each probe. The evaluation of CCR2 expression by IHC at different joint developmental time points (E12.5; E13.5, E14.5 and E18.5) is shown in Figure S2.

MCP-5 acts as a chondrogenic factor on isolated interzone cells. TGF-β inhibits MCP-5 expression on isolated TβRII⁺ cells

We have recently reported that within the interphalangeal joints TβRII expressing cells localize in specific niches such as the ventral and dorsal sites of the interzone (Figure S3) (T. Li, 2011). Reasoning that TβRII expressing cells control the developing environment of the interzone through paracrine and autocrine mechanisms, we relied on the microarray analysis as a source to identify additional genes that encode for membrane-associated proteins that are exclusively expressed in the interzone of NC but not in Tgfbr2^Prx1KO mice (i.e. Jagged1 and Pdgfrb), to isolate interzone cells from E14.5 autopod mesenchyme. An immunomagnetic selection system was used to isolate Jagged1, TβRII and PDGFRβ positive (Jag⁺/TβRII⁺/PDGFR⁺) cells that accounted for ~1% of the total (~30×10⁶ total cells from 30 embryos). mRNA was collected immediately after cell separation from the Jag⁺/TβRII⁺/PDGFR⁺ cell fraction and subjected to RT-PCR. As shown in Figure 4A, Jag⁺/TβRII⁺/PDGFR⁺ cells expressed key joint markers, such as Gdf5, Wnt9a, Sulf1 and Collagen1. With the purpose of understanding the effect of MCP-5 on interzone cells, we cultured the isolated Jag⁺/TβRII⁺/PDGFR⁺ fraction in micromass and treated them with or without MCP-5 (20 ng/ml) for 3 days. MCP-5 induced a chondrogenic-like phenotype in the cells that condensate in typical chondrogenic nodules (Figure 4B). In addition, MCP-5 treatment determined a marked decrease in the expressions of Gdf5, Wnt9a and Sulf1, while increasing Collagen2, compared with untreated cells (Figure 4C). These in-vitro data support our ex-vivo and in-vivo findings indicating that a down-regulation of MCP-5 is needed in interzone cells to allow a correct expression of key joint markers. The chondrogenic action of MCP-5 can be the consequence of several mechanisms ranging from proliferation to regulation of transcription of chondrogenic gene markers. Future investigations will be needed to dissect such mechanisms.

Open in a separate window

Figure 4

MCP-5 effect on joint interzone cells and relevance of TβRII signaling in regulating its expression in vitro

(A-B-C) Jag⁺/TβRII⁺/PDGFR⁺ cells were isolated from E14.5 WT autopods. (A) mRNA was collected after cell separation and subjected to RT-PCR for Collagen1 (Col1), Gdf5, Wnt9a and Sulf1. (B–C) Micromass cultures of isolated Jag⁺/TβRII⁺/PDGFR⁺ were treated ±MCP-5 (20ng/ml) for 3 days and analyzed for morphology (B) and gene expression by RT-PCR for Gdf5, Wnt9a, Sulf1 and Col2. (D) TβRII⁺ cells were FACS sorted from limb buds cells isolated from Tgfbr2-GFP-β-GEO-BAC digits using GFP as selection marker for TβRII expression. Micromass cultures of TβRII⁺ cells were treated ±TGF-β (20ng/ml) for 18h and analyzed for MCP-5 expression by qRT-PCR (n=3; mean ±SD; *p=0.0425 by one tailed Student’s t-test). For RT-PCR experiments, gel lanes correspond to different experiments that were repeated at least three times with similar results. The localization of TβRII expressing cells within the interphalangeal joints is visualized in Figure S3.

Hypothesizing that only the TGF-β signaling regulates MCP-5 expression we exclusively isolated TβRII expressing cells (TβRII⁺) from E14.5 limb mesenchyme of Tgfbr2-GFP-β-GEO-BAC transgenic mice. Tgfbr2-GFP-β-GEO-BAC mice contain both GFP and LacZ as imaging reporters for Tgfbr2 expression (Spagnoli et al., 2007). TβRII⁺ cells were FACS sorted using GFP as isolation marker and cultured in micromass for 18h with or without TGF-β (20ng/ml). qRT-PCR analyses showed that TGF-β significantly decreased MCP-5 expression compared to untreated controls (Figure 4D). These results support our hypothesis that TGF-β is needed to down-regulate MCP-5 expression.

TβRII/MCP-5 axis in joint development

Though no one disputes that interzone is needed for joint formation, there is limited information on the interzone joint-forming cell population and the role that these cells play in the process remains far from clear. The formation of the interzone is specified by an inductive signal in a restricted area of the chondrogenic blastema that leads to loss of Collagen2 expressing cells while joint marker (i.e. GDF-5) expressing cells emerge and segment the adjacent growth plates. It is unclear whether Collagen 2 expressing cells undergo to a selection process (cell death) or whether their transcriptome becomes programmed only toward joint marker expression (Khan et al., 2007). Combining LCM with gene-profiling analyses, we found that the progenitor joint-forming interzone cells develop in a setting that is distinct from the setting of the adjacent growth plate chondrocytes and characterized by a down-regulation of specific cytokines/chemokines, such as MCP-5. Our ex-vivo studies demonstrate that this differentially controlled expression of MCP-5 in the interzone versus the adjacent chondrocytes is essential for the proper process of interzone formation leading to expression of key joint markers (such as Gdf5), as well as the correct segmentation of adjacent growth plates. Consistently, the aberrant release of MCP-5 within the interzone interfered with its initiation and halted the segmentation process. Furthermore, in our in-vitro experiments we demonstrated a direct effect of MCP-5 on isolated interzone cells, characterized by a marked decrease in the expression of key joint markers (Gdf5, Wnt9a and Sulf1) accompanied with a stimulation of Collagen 2 expression and emergence of a chondrogenic phenotype. Our data support the hypothesis that interzone cells are capable to be committed to either joint forming cells or chondrocytes and MCP-5 plays a critical role in directing their differentiative program.

We have previously reported that the joints of Tgfbr2^Prx1KO mutants appear to be arrested in their development about the time when interzone starts to develop; differentiated chondrocytes extend with a continuous pattern across the phalanges lacking interzone cell condensation and Jagged-1, Noggin and Gdf-5 expressions (Spagnoli et al., 2007). In the present study, we found that MCP-5 down-regulation was completely impaired in the interzone of Tgfbr2^Prx1KO mice indicating that the TβRII signaling is needed to control MCP-5 expression and suggesting that the failure of the mutant’s joint formation may be due to the aberrant expression of MCP-5. Our ex-vivo and in-vivo studies sustained this hypothesis showing that in Tgfbr2^Prx1KO mice, the inhibition of MCP-5 activity by blocking its receptor CCR2 restored joint formation and in particular Gdf-5 expression and chondrocyte segmentation. Furthermore, a direct inhibitory action of TGF-β on MCP-5 expression was found in our in-vitro studies by treating TβRII⁺ cells (isolated from limb bud mesenchyme) with TGF-β. These findings classify MCP-5 as a key downstream mediator of TβRII signaling in its critical role as a joint-forming initiator. We postulate that interzone cells are programmed by the TβRII signaling to commit to joint forming cells by repressing MCP-5/CCR2 signaling while halting their differentiative program to Collagen 2 expressing chondrocytes.

MCP-5 its human homolog MCP-1 and its receptor CCR-2 have been largely studied in the arthritic process. The current hypothesis is that those MCPs are produced and act primarily through inflammatory cells (mostly monocytes) that are recruited within the arthritic joints (Haringman and Tak, 2004; Tak, 2006). Although there are substantial pharmacological data supporting a role for CCR2 antagonists in improving the course of experimental arthritis, reports with CCR2 knock-out and anti-CCR2 blocking antibody (MC-21) have provided contrasting results. In fact CCR2 knock-out mice are more susceptible to develop arthritis and in experimental arthritis MC-21 treatment improved the disease when given at an early stage while worsened when given later (Bruhl et al., 2004). Clinical studies using different interventions to block the MCP5/CCR-2 axis in patients with RA have showed disappointing results (Quinones et al., 2005). CCR2 is expressed in chondrocytes but its function is unknown (Borzi et al., 2000; Pulsatelli et al., 1999). Our data provide evidence for a role of the MCP-5/CCR-2 system that targets directly the joint cells. Future studies will be important to determine whether the TβRII signaling regulates MCP-5 expression to maintain articular cartilage integrity in adulthood. In this respect, the OA/RA process could be seen as the failure of an active cell joint population (TβRII cells) that is preserved throughout life to maintain a controlled cytokine environment rather than the result of passive joint cell degeneration due to the influx of inflammatory cells. The lack of CCR2 in the knock-out mice and the timing of the blockage by the MC-21 antibody in the OA studies may have perturbed the TβRII/MCP-5 equilibrium either in the wrong cells or at the wrong time. Consistent with this hypothesis is our observation that TβRII expressing cells localize in specific niches (T. Li, 2011).

TβRII/MCP-5 in endochondral cartilage development

It has been hypothesized that the developing joint acts as signaling center to control the adjacent endochondral template development (Archer et al., 2003). Our ex-vivo studies showed that release of MCP-5 within the interzone leads to an impairment of Ihh spatial organization and to a decrease of Collagen10 expression. Notably, this phenotype coincided with the one observed in the Tgfbr2^Prx1KO mutants suggesting a role for MCP-5 in the failure of the mutant’s chondrocyte progression to hypertrophy. This hypothesis was corroborated by our ex-vivo and in-vivo experiments: blockade of MCP-5 signaling in Tgfbr2^Prx1KO mice rescued Ihh, Collagen10 and Collagen1 expression patterns. Taken together, these results clearly indicate the need for TβRII signaling in maintaining low levels of MCP-5 in the interzone and therefore to signal, through a still undefined factor, the adjacent growth plate template maturation. Future studies are needed to identify such factor(s); potential candidates could be members of WNT family. The interzone expresses several members of the Wnt/β-catenin system that seem to be involved in joint formation (Hartmann et al., 2000; Hartmann and Tabin, 2001; Luyten et al., 2009; Spater et al., 2006). In addition, Wnt9a and β-catenin regulate Ihh expression (Spater et al., 2006). Since in isolated interzone cells we found an inhibitory effect of MCP-5 on Wnt9a expression and demonstrated a TβRII-dependent role for MCP-5 in regulating Ihh spatial organization in-vivo, we speculate Wnt9a as the missing ring linking MCP-5 and Ihh expressions and foresee future studies in this direction.

An outstanding question is what role, if any, does MCP-5 expressed by chondrocytes within the growth plate play in the endochondral mineralization process? Our data contribute to answer this question. We found that in normal limbs, the inhibition of MCP-5 activity within the growth plate resulted in an increase of Collagen10 by hypertrophic chondrocytes and, at later stages, Collagen1 in the ossification center. These findings indicate that, in addition to its indirect action from the joints, MCP-5 within the growth plate affects the growth plate maturation, inhibiting hypertrophy. Since the lack of TβRII signaling in the Tgfbr2^Prx1KO mutants does not seem to affect MCP-5 expression in the growth plate, it is likely that this effect of MCP-5 is independent of TβRII signaling. Interestingly, blockade of MCP-5 activity in Tgfbr2^Prx1KO mutants, did not lead to an increase in mineralization. One explanation could be that in normal mice, differently from Tgfbr2^Prx1KO mutants, blockade of the MCP-5-dependent inhibitory action on chondrocyte maturation could be additive with the MCP-TGF-β-dependent mechanism(s) promoting hypertrophy and ossification. As expected, in control mice no joint derangement was found as consequence of MCP-5 inhibition, being MCP-5 already undetectable in normal joints and indicating that MCP-5 from the growth plate has not effect on joint development.

LCM for RNA sampling and Microarray analysis

We obtained E14.5 autopods from two NC and two Tgfbr2^Prx1KO littermate embryos. Samples were frozen in OTC and cryosectioned (6µm), then processed with ethanol and xylene, and subjected to LCM. To minimize RNA degradation, the process from sectioning to shipment was performed in less than 20 minutes. We sequentially laser-captured cells in the interzone and the adjacent growth plates from NC and Tgfbr2^Prx1KO mutants (n=2). In the Tgfbr2^Prx1KO sections the presumptive interzone was identified as a bending between two growth plates (Figure 1A arrow). Captured specimens were frozen in SuperAmp Lysis Buffer (Miltenyi Biotec) and shipped to Miltenyi Biotec. mRNA was extracted and transcribed into cDNA according to Miltenyi Biotec’s undisclosed procedure (SuperAmp Miltenyi Biotec). The average length of cDNA products, measured and analyzed via the Agilent 2100 Bioanalyzer platform and software (Agilent Technologies), ranged from 200–1000 bp, indicating the integrity of the cDNAs. The purified PCR products (250 ng) were labeled with Cy3-dCTP and hybridized to an Agilent Whole Mouse Genome Oligo Microarrays 4×44K for 17 hours at 65° according to Miltenyi Biotec’s undisclosed protocol. Washing, staining and scanning of the microarrays were performed using the Agilent Technologies equipment. Signal intensities from the single-experiment data were normalized by dividing the intensity values by their median. NCBI GEO Accession number {"type":"entrez-geo","attrs":{"text":"GSE37287","term_id":"37287"}}GSE37287, http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc={"type":"entrez-geo","attrs":{"text":"GSE37287","term_id":"37287"}}GSE37287. Normalized data were expressed as fold changes of a candidate gene in the interphalangeal interzone sample compared to that obtained in chondrocytes using the Resolver (Rosetta Biosoftware). Pathway analysis was performed using the PANTHER Classification System (Celera) (Thomas et al., 2003). Lists of putative candidate genes with a fold change ± 2 and p-value 0.01 were compared to a reference list (Mus Musculus, NCBI) using a binomial test. Microarray data will be accessible through the Gene Expression Omnibus (GEO) with the final version of the manuscript.

Preparation of HGC-MCP-5/MC21 nanoparticles, implantation and ex-vivo limb cultures

HGC nanoparticles were prepared as previously reported (Li et al., 2010). Briefly, 25µg of either recombinant mouse MCP-5 (R&D Systems) or MC-21 antibody (anti-CCR2) (donated by M. Mack, Regensburg University) (Mack et al., 2001) were encapsulated into 250µg of HGC nanoparticles. Methylene Blue was added to the mixture to mark the injection site. HGC-protein implantation was performed using a microinjection unit (Picosprizer II) that allowed for precise localization of the injection and volume accuracy (0.0015µl/shot) as previously described (Li et al., 2010). HCG-protein implants (joint site) were performed in WT, NC or Tgfbr2^Prx1KO by delivering either HGC-MC-21 (1.5 ng, 5-shots, 0.3ng/shot) or increasing doses of HGC-MCP-5 (0.9ng, 1.5ng, 3ng and 6 ng). Limbs were cultured in BGJb medium for 48 hours. HGC-protein implantation was performed within the medial interphalangeal joint of 2^nd, 3^rd and 4^th digit and the 5^th uninjected digit of the same limb was used as controls. The forming or presumptive interphalangeal joint was identified as the bending between two adjacent growth plates.

RS-504393 in-vivo treatment

Pregnant NC females potentially generating Tgfbr2^Prx1KO mice, were treated with 4mg/Kg/day of RS-504393 (TOCRIS Bioscience), orally (in drinking water), starting at 11.5 days of gestation, for a total of either 3 (E14.5) or 7 days (E18.5). After euthanasia of the mother, limbs from either E14.5 or E18.5 embryos were dissected and prepared for histological studies. Genotyping was used to confirm the mutant phenotype, RS-504393-treated NC littermates and NC untreated mice were used as controls.

Isolation and culture of Jag⁺/TβRII⁺/PDGFR⁺ cells and TβRII⁺ cells

To obtain mesenchyme limb bud cells, the distal tips of E14.5 autopods from either WT or Tgfbr2-GFP-β-GEO-BAC embryos were dissected and digested with dispase (1U/ml), as previously reported (Spagnoli et al 2007). To obtain Jag⁺/TβRII⁺/PDGFR⁺ cells, WT mesenchyme limb bud cells are subjected to immunoselection using biotinylated TβRII, biotinylated PDGFRB (BAF532 and BAF1042, R&D Systems) and Jagged 1 (sc-6011, Santa Cruz Biothechnolgy) antibodies. Biotinylation of Jagged1 antibody was performed using Lightning-Link Biotin Conjugation Kit (Innova Biosciences) following manufacturer’s protocol. Immunoselection was performed in a MACS magnetic system (Miltenyi) as previously described (Longobardi et al 2005). Isolation of TβRII⁺ cells was performed by FACS sorting (MoFlo, Beckman) mesenchyme limb bud cells from Tgfbr2-GFP-β-GEO-BAC that contain GFP as gene reporters for Tgfbr2 (Spagnoli et al., 2007). Micromass cultures were performed as previously reported (Spagnoli et al., 2007). Jag⁺/TβRII⁺/PDGFR⁺ or TβRII⁺ cells were treated with ±20ng/ml MCP-5 (for 3 days) or with ±20ng/ml TGF-β (for 18hours), respectively.

Histology, IHC and ISH

Dissected limbs from E14.5 to E18.5 embryos were fixed overnight in 4%PFA at 4°C, paraffin embedded and sectioned at 5µm thickness. Hematoxylin staining was performed following standard protocols.

For IHC, the Vectastain Elite ABC Kit (DakoCytomation) was used. The following primary antibodies were used: anti-MCP-5 (G12) polyclonal (sc-33226, Santa Cruz) and anti-CCR-2 monoclonal (ab32144, Abcam). Histological sections were counterstained with Hematoxylin.

ISH studies were performed as previously reported using DIG-labeled probes (Spagnoli et al., 2007). The MCP-5 plasmid used to generate the MCP-5 probe was provided by Dr Karin E. Peterson (Louisiana State University). After hybridization, probes were detected by incubation with anti-DIG-AP antibody (Roche, Mannheim, Germany, dilution 1:5000).

Images were formatted without using any imaging enhancement.

Quantitative real-time PCR

mRNA from experiments using Jag⁺/TβRII⁺/PDGFR⁺ cells and TβRII⁺ cells was extracted and reversed transcribed into cDNA by using µMACS mRNA Isolation and cDNA Synthesis kit (Miltenyi Biotec). cDNAs (including from LCM) were analyzed by qRT-PCR (MyIQ) using SYBR Green followed by Gene Expression Analysis Software (Bio-Rad). Expression of genes of interest were normalized using either GAPDH or β-actin expressions.

We acknowledge the support of the Vanderbilt Mouse Pathology Core for Laser Capturing Microdissection and the UNC Flow Cytometry Core Facility for the FACS sorting. We are grateful to Dr. Matthias Mack (University of Regensburg, Regensburg, Germany) for providing the MC-21 antibody; to Dr Karin E. Peterson for the MCP-5 probe (Louisiana State University, Baton Rouge, LA); Dr. Froilán Granero-Moltó for his technical assistance and discussions. This work was supported by a NIH-NIAMS Grant 1R01AR057042-02 (to A.S.) and NC-TraCS Institute (NIH-CTSA) (to L.L.). We also acknowledge The Scientific and Technological Research Council of Turkey (to H.O.).