Enumeration and phenotypic characterization of synovial fluid multipotential mesenchymal progenitor cells in inflammatory and degenerative arthritis




To evaluate synovial fluid (SF) for the presence of mesenchymal progenitor cells (MPCs), to compare SF MPCs with bone marrow (BM) MPCs, and to enumerate these cells in both inflammatory arthritis and osteoarthritis (OA).


SF from 100 patients with arthritis (53 rheumatoid arthritis [RA], 20 OA, and 27 other arthropathies) was evaluated. To establish multipotentiality, polyclonal and single cell–derived cultures of SF fibroblasts were examined by standard and quantitative differentiation assays. Their phenotype before and after expansion was determined by multiparameter flow cytometry. A colony-forming unit–fibroblast assay was used for SF MPC enumeration.


Regardless of the nature of the arthritis, both polyclonal and single cell–derived cultures of SF fibroblasts possessed trilineage mesenchymal differentiation potentials. The number of MPCs in a milliliter of SF was higher in OA (median 37) than in RA (median 2) (P < 0.00001). No significant differences in MPC numbers were found between early and established RA (median 3 and 2 cells/ml, respectively). Culture-expanded SF and BM MPCs had the same phenotype (negative for CD45 and positive for D7-FIB, CD13, CD105, CD55, and CD10). Rare, uncultured SF fibroblasts were CD45low and expressed low-affinity nerve growth factor receptor, similar to in vivo BM MPCs.


Our findings prove the presence of rare tripotential MPCs, at the single-cell level, in the SF of patients with arthritis. SF MPCs are clonogenic and multipotential fibroblasts that, despite the pathologic environment within a diseased joint, have a phenotype similar to that of uncultured BM MPCs. The higher prevalence of MPCs in OA SF suggests their likely origin from disrupted joint structures. These findings could determine the role of MPCs in the pathogenesis of inflammatory arthritis, together with their role in attempted joint regeneration in degenerative arthritis, which has yet to be established.

Mesenchymal progenitor cells (MPCs) possess high proliferative potential and can differentiate into several mesenchymal lineages, including bone, cartilage, fat, tendon, and stromal tissue (1–3). Initially identified in the bone marrow (BM), these cells have subsequently been found in trabecular bone, adipose tissue, and synovial tissue (4–9). Although there has been much speculation about the role of MPCs in the pathogenesis of inflammatory or degenerative arthritis, their basic biology, topography, and roles in joint physiology remain unknown (9–12). By definition, MPCs have great potential to repair damaged bone and cartilage and are likely to contribute to joint regeneration, which is a prominent feature of osteoarthritis (OA). A recent study described alterations in the activity of MPCs in the BM of OA patients (13), but their presence in OA synovial fluid (SF) has not yet been investigated. Furthermore, there have been several hypotheses concerning the potential role of MPCs in the pathogenesis of rheumatoid arthritis (RA), but these could not be tested in the absence of data on the phenotypic “fingerprint” of joint MPCs (10–12, 14).

If MPCs play a physiologic role in joint homeostasis and repair, then it is conceivable that they are present in SF, thus permitting direct access to superficial articular cartilage. Fibroblastic cells that expressed phenotypic markers of the mesenchymal lineage upon culture expansion (such as CD44, type I collagen, and vascular cell adhesion molecules) have been found in SF of arthritis patients (12). However, their multipotentiality has not been reported, and hence, it is still unclear whether cells initiating these cultures were true SF MPCs. Also, the inability to clearly define SF MPCs has hampered attempts to enumerate them in RA and OA and, hence, to explore their putative roles in the pathogenesis of these disorders.

The aim of this study was to apply knowledge obtained from our previous investigation of BM MPCs (15) to an evaluation of SF for the presence of MPCs and to compare their numbers in inflammatory arthritis and OA. Our results demonstrate that SF from arthritis patients contained clonogenic, highly proliferative MPCs and that these were similar in phenotype to BM MPCs, more numerous in patients with OA, and uncommon in early and established RA. These findings form the basis for an exploration of the role of SF MPCs in joint pathophysiology in RA and OA.



Approval for the study was obtained from the local ethics committees at the Calderdale and Huddersfield National Health Service (NHS) Trust and the West Yorkshire and Leeds Health Authority/Leeds Teaching Hospitals NHS Trust. Informed consent was obtained from all study subjects. BM was obtained from the posterior iliac crest of 8 donors and was separated using Lymphoprep (Nycomed, Oslo, Norway). SF was obtained from the knee joints of 100 patients with inflammatory or degenerative arthritis who had clinically discernible joint effusions. The diagnoses were RA in 53 patients, OA in 20, psoriatic arthritis (PsA) in 10, undifferentiated arthritis in 8, reactive arthritis (ReA) in 6, pseudogout in 2, and polymyalgia rheumatica in 1. The patients with RA fulfilled the American College of Rheumatology (formerly, the American Rheumatism Association) criteria for the disease (16). Seven of the patients with RA had early disease, defined as a median duration of <2 years, and the remainder were classified as having established RA, with a median duration of 9 years (range 5–20 years).

An assessment of both early and established RA was undertaken to ascertain whether MPCs were more common in early disease (possibly indicating a role in the pathogenesis of joint damage) or in late disease (possibly being related to joint damage or repair due to secondary OA in the later stages of disease). All of the patients with OA had had knee joint pain for >6 months (median duration 7 years) and, at the time of clinical examination, had cool effusions of the involved joints. There was no evidence of other arthropathies, as demonstrated by negative findings on analyses of SF for crystals. Chondrocalcinosis was not detected on radiographs. Some SF specimens were used for development work (n = 18); others were used for differentiation assays (n = 13) and phenotyping (n = 20). The majority of samples (n = 68) were used for MPC enumeration, and more detailed clinical information on these patients is presented in Table 1. Some specimens were used in more than one investigation.

Table 1. Clinical characteristics of the 68 patients whose synovial fluid was used for the mesenchymal progenitor cell enumeration experiments*
 Patient group
RA (n = 35)OA (n = 14)Other arthropathies (n = 19)
  • *

    The group of patients with other arthropathies had reactive arthritis (n = 4), psoriatic arthritis (n = 7), pseudogout (n = 2), and seronegative inflammatory arthritis (n = 6). RA = rheumatoid arthritis; OA = osteoarthritis; NA = not applicable; MTX = methotrexate; SSZ = sulfasalazine; NSAIDs = nonsteroidal antiinflammatory drugs.

  • Other treatments may have included cyclosporine, azathioprine, leflunomide, corticosteroids, or gold.

Disease duration   
Median8.5 years7 yearsNA
Range1 month to 20 years1–13 yearsNA
Treatment, % of patients   

SF was aspirated from knee joints either prior to intraarticular corticosteroid injection or for diagnostic assessment. The volume of fluid ranged between 2 and 75 ml. For isolating cells from the SF, freshly aspirated fluids were diluted 1:4 with phosphate buffered saline (Gibco, Paisley, UK) and centrifuged at 2,000 revolutions per minute for 10 minutes. In 50 experiments, SF cells were also subjected to differential density centrifugation using Lymphoprep. Three lines of normal human skin fibroblasts (American Type Culture Collection, Manassas, VA) and a line of primary skin fibroblasts (obtained from dermal explants) were cultured in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal calf serum (FCS) (both from Gibco).

Generation of polyclonal MPC cultures from BM and SF.

Cultured BM MPCs used as a positive control were generated from the BM plastic-adherent fraction according to standard methods and grown for 3–4 passages prior to differentiation (3). Similarly, cultures of fibroblastic cells from SF (putative SF MPCs) were established from the plastic-adherent fraction and grown in the same media for 3–4 passages prior to differentiation.

Standard and quantitative differentiation assays.

Osteogenic differentiation was induced by placing cells in standard osteoinductive conditions, as previously described (17). Alkaline phosphatase (AP) activity and calcium production were assessed at weeks 2 and 3 of culture using kits 82 and 85, respectively (Sigma, Poole, UK). Adipogenic differentiation was induced and assessed as previously described (15). For quantitative analysis of adipogenesis, duplicate cultures of 105 cells were seeded onto 19-mm coverslips (BDH Chemicals, Poole, UK) placed onto 12-well cluster plates (Corning Costar, Cambridge, MA). At week 3, coverslips were double-stained with Nile Red (final concentration 1 μg/ml) and 4′,6-diamidino-2-phenylindole (DAPI; final concentration 8 μg/ml) (both from Sigma) (18). Fluorescence was analyzed using an Axioskop fluorescence microscope (Zeiss, Wetzlar, Germany) and Capture VP software (version 1.4; Vysis Ltd., Richmond, UK). The number of lipid-laden (Nile Red–positive) cells was calculated as a percentage of total (DAPI-positive) cells. For chondrogenic differentiation, up to 5 replicate cultures of 2.5 × 105 cells were placed in Eppendorf tubes (BDH Chemicals) in 0.5 ml of chondroinductive media (19). Sulfated glycosaminoglycan (GAG) was visualized on frozen sections (5 μm thick) with 1% toluidine blue (Sigma). Production of sulfated GAG was measured in an Alcian blue binding assay (Immunodiagnostic Systems, Boldon, UK) following digestion in 100 μl of papain solution (13). Absorbance was read at 630 nm.

Colony-forming unit–fibroblast (CFU-F) assay and generation of single cell–derived clones from BM and SF.

The number of CFU-Fs in SF was assessed in a conventional CFU-F assay (20). The cell-seeding density depended on the number of cells available for the analysis and was within a range of 0.5–5 × 104/cm2. SF CFU-F numbers were established in relation to 106 total cells seeded and subsequently normalized to 1 ml of SF. A pilot study involving 10 patients with RA established an average CFU-F frequency of 1 cell/106 total SF cells (range 0.1–5.0), and this frequency estimate was used to obtain single cell–derived clones from the SF. For this, 5 × 106 cells were seeded into 24-well cluster plates, which corresponds to ∼1 prospective CFU-F cell per 5 wells. All the wells were initially carefully monitored for the presence of fibroblastic cells, and then for growing colonies, to ensure that no well that might have contained more than 1 colony was propagated any further. The wells containing single colonies were further expanded to the size of a confluent 75-cm2 flask (corresponding to ∼20–22 population doublings). The seeding density for the BM control sample was similarly calculated based on the previously established CFU-F frequency.

Immunophenotyping of cultured BM MPCs, SF MPCs, and skin fibroblasts.

For phenotyping of cultured MPCs, cells were used at passage zero (P0) in all experiments. Cells were treated with trypsin–EDTA (Gibco), and 105 cells were used per test. The following commercially available antibodies were used: low-affinity nerve growth factor receptor–phycoerythrin (LNGFR–PE), CD55–biotin (both from Becton Dickinson, Oxford, UK), and CD105–PE (Serotec, Kidlington, UK). Bone morphogenetic protein receptor type IA (BMPRIA) expression was detected using a biotinylated polyclonal goat antibody (R&D Systems, Abington, UK) at 500 ng/ml, followed by 1:200 dilution streptavidin–PE (Becton Dickinson). The following antibodies were kindly provided by Dr. Richard A. Jones (Haematological Malignancy Diagnostic Service, Leeds General Infirmary, Leeds, UK): CD10–PE, CD13–PE, CD45–PE, D7-FIB–PE, and isotype control mouse anti-human IgG1–PE and IgG2a–PE. The chosen panel of antibodies was similar to that used for phenotyping of BM MPCs (15). CD55 was included because it was shown to be expressed by synovial lining cells (21). Data were acquired using an XL/MCL flow cytometer (Beckman Coulter, Fullerton, CA) and analyzed with the WinMDI program, version 2.8 (Scripps Research Institute, La Jolla, CA).

Immunophenotyping of uncultured SF fibroblasts compared with BM MPCs.

BM MPCs or SF fibroblasts were initially enriched by overnight adherence to plastic (in DMEM/10% FCS) and harvested by trypsin–EDTA treatment. In SF, to set apart adherent fibroblasts and adherent monocyte/macrophages, the following antibodies were used: CD45–fluorescein isothiocyanate (FITC) (Dako, High Wycombe, UK), and CD45–PE/Cy5, CD13–FITC, and CD14–FITC (all from Dr. Richard A. Jones). Fibroblasts were identified as CD45low,CD13+,CD14− cells, and monocyte/macrophages were identified as CD45bright,CD13+,CD14+ cells. Having established detectability and the phenotype of SF fibroblasts following adherence, subsequent experiments used a dual-laser FACSort flow cytometer and D7-FIB–APC to directly detect fibroblasts in freshly obtained SF. To visualize rare fibroblasts (gated as live CD45low,D7-FIB+ cells) by 4-color cytometry, a minimum of 3 × 105 cells were collected and propidium iodide (Sigma) was used at 2 μg/ml.


Identification of SF fibroblasts and establishment of their culture.

We and other investigators have shown that multipotential BM MPCs appear as adherent fibroblast-like cells distinguished by their long projections and prominent nucleoli (15, 22). Adherent cells with similar basic morphology were found in the SF of patients with arthritis (Figure 1A). During the first week of culture, these cells underwent several rounds of cell division, with resulting fibroblastic colonies clearly detectable by day 7 (Figure 1B). This process of colony formation was similar to that of BM-derived MPCs. Following growth in culture, these colonies became bigger (Figure 1C) and subsequently merged, resulting in the formation of monolayers of typical spindle-shaped fibroblasts (Figure 1D).

Figure 1.

Morphologic appearance and colony formation by putative synovial fluid (SF) mesenchymal progenitor cells. A, Phase-contrast photomicrographs of SF fibroblast-like cells after overnight incubation in standard media (hematoxylin stained). B, Fibroblastic colony, on day 7 of culture. C, Part (approximately one-fourth) of a growing fibroblastic colony on day 14 of culture. D, Confluent layer of SF-derived fibroblasts. (Original magnification × 400 in A; × 50 in B and C; × 100 in D.)

Establishment of multilineage mesenchymal differentiation capacity of cultured SF-derived fibroblasts.

Cultures derived from 12 patients (5 RA, 2 ReA, 1 PsA, 3 OA, 1 seronegative inflammatory arthritis) were subjected to osteogenic, adipogenic, and chondrogenic differentiation. Cultured BM MPCs (n = 4) and skin fibroblasts (n = 4) were used as positive and negative controls of differentiation, respectively (Figure 2). In contrast to skin fibroblasts, all SF-derived and positive control BM cultures grown in osteogenic medium produced AP 2 weeks postinduction, consistent with successful progression toward an osteoblastic lineage (Figure 2, left panels). After 3 weeks of culture, SF-derived osteogenic cultures produced a mean ± SD of 78 ± 25 μg of calcium (n = 5), compared with 80 ± 37 μg of calcium produced by BM MPC-derived cultures (n = 4). This suggested that SF-derived fibroblasts had an osteogenic differentiation capacity similar to that of cultured BM MPCs. In media lacking osteogenic supplements, low levels of spontaneous AP were observed in some early-passage BM MPC– and SF-derived cultures (but not in skin fibroblasts), with no calcium produced by all cultures (results not shown).

Figure 2.

Differentiation of synovial fluid (SF)–derived cultures of fibroblasts into osteogenic, adipogenic, and chondrogenic lineages. Left panels, Osteogenic differentiation (alkaline-phosphatase stained); middle panels, adipogenic differentiation (oil red O–stained); right panels, chondrogenic differentiation (toluidine blue–stained; one-fourth of each pellet is shown). SF-derived fibroblasts were capable of trilineage mesenchymal differentiation. ReA = reactive arthritis; OA = osteoarthritis; RA = rheumatoid arthritis; PsA = psoriatic arthritis; BM = cultured bone marrow mesenchymal progenitor cells (positive control); FIB = cultured skin fibroblasts (negative control). (Original magnification × 100.)

Accumulation of lipid vacuoles upon adipogenic induction was observed in all BM- and SF-derived cultures, but not in cultured skin fibroblasts (Figure 2, middle panel). At week 3, a mean ± SD of 28 ± 9% of cells in adipogenic SF–derived cultures contained lipid vacuoli (n = 3), compared with 33 ± 15% of cells in BM-derived cultures (n = 3). Skin fibroblasts failed to undergo adipogenic progression (n = 4), and no spontaneous adipogenesis was seen in any type of culture grown in media lacking adipogenic supplements (results not shown).

In contrast to skin fibroblasts, BM- and SF-derived cultures grown in chondrogenic differentiation assay displayed prominent toluidine blue staining of extracellularly deposited proteoglycans, indicating chondrogenic differentiation (Figure 2, right panels). No sulfated GAG was produced in any pellets incubated in media lacking transforming growth factor β3 (results not shown). These experiments demonstrated that regardless of the nature of the arthritis, fibroblastic cultures derived from SF possessed all 3 mesenchymal differentiation potentials and could therefore be seen as an SF equivalent of cultured BM MPCs.

Multipotential nature of individual clonogenic fibroblasts present in SF.

As noted above, cultures of BM and SF MPCs used for the initial analysis were polyclonal (multicolony derived). To investigate the differentiation potential of single colony–derived cultures, we generated cultures from individual CFU-Fs. The aims of these experiments were to compare differentiation potentials of individual CFU-Fs from the same patient and between patients (4 patients, 3 with RA and 1 with ReA; n = 17 clones). Three clones derived from a single BM specimen (5-year-old donor) were used as positive controls of differentiation (Figure 3). For the purposes of adequate comparison, all culture-initiating CFU-Fs underwent ∼20–22 population doublings prior to differentiation (corresponding to ∼106-fold expansion in cell numbers). For both BM- and SF-derived clones, expansion lasted an average of 49 days (range 45–56 days), which corresponds to ∼1 cell division every 2 days. For all 4 patients analyzed, all fibroblastic colonies initially identified in 24-well plates were successfully expanded.

Figure 3.

Differentiation potentials of single cell–derived cultures grown from the synovial fluid (SF) of 4 patients with arthritis. Arrows indicate 3 clones from normal bone marrow (BM; positive control). A, Osteogenic differentiation. B, Adipogenic differentiation. C, Chondrogenic differentiation. All clones were expanded for 20–22 population doublings prior to differentiation, and each point represents the mean value for at least 3 replicate measurements. SF-derived clones from all 4 patients displayed levels of osteogenesis lower than those of the control BM clones, while clones from 3 of 4 patients were, on average, more adipogenic and chondrogenic than control BM clones. GAG = glycosaminoglycan.

As seen in Figure 3A, all SF-derived clones can undergo osteogenic differentiation, as assessed by deposition of extracellular Ca++ (ranging between 25 and 100 μg of Ca++ per 3 × 104 seeded cells). Levels of Ca++ production were very similar between the clones derived from the same patient, and were not particularly different among the 4 patients analyzed. BM-derived clones produced Ca++ at notably higher levels (>100 μg), consistent with previous demonstrations of high levels of osteogenesis in BM from young subjects (23).

All SF-derived clones also displayed successful adipogenic progression (Figure 3B). Proportions of generated adipocytes did not markedly differ among the 4 patients and were within the range obtained for polyclonal cultures. Control BM-derived clones showed relatively low levels of adipogenesis, consistent with the notion of reciprocally controlled adipogenesis and osteogenesis in BM MPCs (18). Similarly, all 17 SF-derived clones produced sulfated GAG upon chondrogenic induction, and the levels of production were very similar for the clones derived from the same patient (Figure 3C). Although negative control single cell–derived clones of skin fibroblasts were not available, all 4 polyclonal strains produced no sulfated GAG in all experiments (data not shown).

Since there was no single SF cell–derived clone that failed to differentiate into all 3 mesenchymal lineages, we concluded that single SF CFU-Fs could be equated to tripotential SF MPCs.

Enumeration of SF MPCs in RA, OA, and other arthropathies.

A CFU-F assay was next used to measure MPC numbers in the SF of a broad group of patients (n = 68) with different forms of arthritis. The aim of these experiments was to ascertain whether there were any significant differences in the numbers of MPCs in patients with different types of arthritis, and between inflammatory and degenerative diseases in particular.

The majority of MPCs in BM have been shown to copurify with mononuclear cells (MNCs) (20). In contrast to normal BM, where the ratios between the MNCs and neutrophil fractions remain fairly constant, the cellular composition of SF is highly variable, depending on the type of arthritis and the degree of inflammation. Based on this, we initially conducted parallel measurements of the colony-forming activity of MNC- and neutrophil-enriched fractions of SF from 50 patients (separated with Lymphoprep) and showed that although the majority of SF MPCs were copurified with MNCs, some were present in the neutrophil-enriched fraction. SF from patients with OA (n = 6) contained the highest number of MPCs in both fractions, with median frequencies of 2,200/106 MNCs (range 40–5,050) and 110/106 neutrophils (range 9.5–324). SF from RA patients (n = 28) contained an average of 2 MPCs/106 MNCs (range 0–1,015) and 0.2 MPCs/106 neutrophils (range 0–68).

Next, the number of MPCs per ml of SF was directly compared between RA and OA patients (n = 35 and n = 14, respectively). A group of patients with miscellaneous arthropathies was also included in the analysis (4 ReA, 7 PsA, 2 pseudogout, and 6 seronegative inflammatory arthritis) (Table 1). MPC numbers were significantly higher in OA patients (median frequency 37 cells/ml) compared with both the RA group (median frequency 2 cells/ml) and the group with other arthropathies (median frequency 4 cells/ml) (Figure 4A). The volumes of fluid collected were not significantly different for the OA and RA patient groups (median 8.8 ml and 12.5 ml, respectively), suggesting that lower MPC numbers in RA could not simply be reflective of an associated dilution effect on MPCs.

Figure 4.

Enumeration of synovial fluid mesenchymal progenitor cells (SF MPCs) in osteoarthritis (OA), rheumatoid arthritis (RA), and other arthropathies (reactive arthritis, psoriatic arthritis, pseudogout, and seronegative inflammatory arthritis). A, Comparison of SF MPC numbers between RA, OA, and miscellaneous (Misc) arthropathies patient groups. The median number of MPCs/ml of SF was 37 and 2 in OA and RA, respectively (* = P < 0.00001 by Mann-Whitney U test). B, Comparison of SF MPC numbers between RA and OA patient groups with similar levels of cellularity (0.2–2 × 106 total cells/ml). The median number was 38 and 3 cells/ml, respectively (* = P < 0.00001 by Mann-Whitney U test). C, Comparison of SF MPC numbers between patients with early RA (disease duration <2 years) and established (Establ) RA (disease duration >2 years). The median number was 3 and 2 cells/ml, respectively (P = 0.56). SF MPC numbers were shown to be independent of the overall cellularity and volume of aspirated fluid in all groups. NS = not significant.

The effect of the overall SF cellularity on MPC numbers was also investigated. SF from RA patients was, on average, 10-fold more cellular than that from OA patients (3.7 × 106 cells/ml versus 0.34 × 106 cells/ml, respectively). When we analyzed only patients with similar levels of inflammation (0.2–2 × 106 cells/ml; n = 9 patients in each group), the same >10-fold difference in MPC numbers between OA and RA was found (median frequency 38 cells/ml versus 3 cells/ml, respectively) (Figure 4B). Together, these data demonstrated that SF from OA patients had significantly more clonogenic tripotential MPCs, and this did not appear to be related to the influx of inflammatory cells into the joint. In patients with OA, a positive correlation was also found between MPC counts and disease duration, although it failed to reach statistical significance (R2 = 0.18).

Finally, we compared MPC numbers in SF from patients with early or established RA (n = 7 and n = 25, respectively). Only 1 patient in the early RA group had erosive disease of the hands or feet, as determined by plain radiography. In the group with established RA, 82% of the patients had erosive disease. No significant differences in MPC numbers between early and established RA were found, nor were there any differences in both SF volume and cellularity (Figure 4C). No correlations were found between MPC counts and the disease duration or the type of therapy for both groups of RA patients.

Immunophenotype of cultured polyclonal SF MPCs.

Having confirmed the presence of MPCs in SF, the phenotype of these cells following culture expansion was investigated. P0 cultures from 7 patients (3 RA, 2 OA, 1 PsA, and 1 pseudogout) were used in these experiments. Regardless of the nature of the arthritic disease, cultured SF MPCs expressed D7-FIB, CD13, CD105, and CD55 antigens similar to cultured BM MPCs or skin fibroblasts, and all types of cultures lacked CD45 and BMPRIA (Figure 5). Levels of expression of D7-FIB and CD13 did not markedly differ among all the cultures. SF-derived cultures showed characteristically low levels of CD10 expression (Figure 5, right panels). In contrast to skin fibroblasts, a proportion of cultured BM MPCs (∼10%, bottom middle panel) and SF MPCs (∼8%, bottom right panel) was LNGFR-positive. We and other investigators have previously shown that LNGFR is present on BM MPCs in vivo and that its expression gradually disappears in culture (15, 24). The presence of residual LNGFR on cultured SF MPCs from all 7 patients suggested their possible derivation from LNGFR-positive culture-initiating cells.

Figure 5.

Immunophenotype of cultured synovial fluid mesenchymal progenitor cells (MPCs) (right panels) compared with cultured bone marrow MPCs (middle panels) and skin fibroblasts (left panels). Because of homogeneous expression of CD45, D7-FIB, CD13, CD105, CD55, CD10, and bone morphogenetic protein receptor IA (BMPRIA) by all cells, the results for these markers are displayed as the mean fluorescence intensity on the FL2 channel. Because of heterogeneous expression levels of low-affinity nerve growth factor receptor (LNGFR), the results for this marker are displayed as the percentage of cells with fluorescence above control levels. Numbers in parentheses are the number of different donor-derived cultures tested for each type of antibody. Values are the mean ± SD.

Immunophenotype of SF fibroblasts prior to culture expansion compared with BM MPCs.

Direct comparisons of the phenotypes of rare SF fibroblasts and BM MPCs were based on our previous findings that BM MPCs have an essentially fibroblastic phenotype (CD45low,D7-FIB+,CD10+,CD13+,CD105+), with characteristic LNGFR expression that is absent on common skin fibroblasts (15). Our aims therefore were to identify fibroblasts in SF by flow cytometry and then to investigate their LNGFR expression. Although not every SF fibroblast proved to be clonogenic (i.e., equal to MPC), our estimate of their frequency was still very low (∼0.05% in OA and <0.0002% in RA). This suggested a requirement for preenrichment for their flow cytometry detection. Plastic adherence as a prospective enrichment method was first tested on BM and confirmed to result in ∼100-fold enrichment for MPCs, similar in magnitude to that of the previously described D7-FIB microbead-based method (15) (Figure 6A). Using BM, we also confirmed that binding for CD13, CD105, and other relevant antibodies was not affected by adherence and/or trypsinization. The expression of LNGFR was found to be slightly down-regulated (compare v with iv in Figure 6A).

Figure 6.

Immunophenotype of synovial fluid (SF) fibroblasts before culture expansion. A, Proof (using normal bone marrow [BM] cells) that rare MPCs (circled areas) can be successfully enriched by overnight adherence to plastic. MPCs were barely detectable when a standard 20,000 events were analyzed (i), whereas collecting 500,000 events revealed a clearly distinct population of CD45low,D7-FIB+ cells (ii). MPCs could be 100-fold enriched from BM using either a D7-FIB microbead-based method (iv) or overnight adherence to plastic (v). Note similar levels of enrichment by both methods and some reduction in the levels of LNGFR expression following plastic adherence. B, Detection of SF fibroblasts (circled areas) enriched by plastic adherence. Unenriched SF cells composed mainly of neutrophils (N) and lymphocytes (L) (i). The plastic-adherent fraction was enriched for cells with monocytic forward scatter (FSC)/ side scatter (SSC) characteristics (arrow) (ii). Immunophenotyping of cells within this area (iv and v) identified adherent monocyte/macrophages (M; CD45bright,CD14+) (iv) and a minor population of nonmonocytic cells (CD45low,CD14−) (iv), expressing a high density of CD13 (v), suggestive of fibroblasts. C, Phenotype of the 2 populations of adherent SF cells–fibroblasts (CD45low; circled area) and monocyte/macrophages (CD45high; M). SF from 2 patients with osteoarthritis (OA) is shown (one in ii–iv, the other in v). SF fibroblasts expressed CD105/endoglin/SH2 (ii), D7-FIB (iv), and LNGFR (v), as compared with negative staining of control antibody (i). D, Phenotype of fresh OA SF cells, analyzed by 4-color flow cytometry with 300,000 events collected. Ungated cells (i) and cells gated on a CD45low,D7-FIB+PI− population (circled areas) (ii–v) are shown. Gated cells expressed high levels of CD13 (iv) and LNGFR (v) as compared with negative staining of control antibody (ii). SF fibroblasts expressed LNGFR similar to BM MPCs. APC = allophycocyanin; FITC = fluorescein isothiocyanate; PE = phycoerythrin; PI = propidium iodide (see Figure 5 for other definitions).

As a rule, SF samples contained large numbers of nonadherent lymphocytes and neutrophils (Figure 6B, i). As expected, the plastic-adherent fraction was largely depleted of these cells (Figure 6B, ii) and consequently enriched for cells confined to a “monocytic” forward scatter/side scatter region. The majority (80–90%) of these cells were indeed CD45bright,CD13+,CD14+ monocyte/macrophages. However, a small population of CD45low cells could often be identified with the phenotype CD13+,CD14−, consistent with that of fibroblasts. This is consistent with a recent study showing histochemically that SF adherent cells consist of 2 distinct populations: monocyte/macrophages and fibroblasts (25). Subsequent phenotypic analysis of SF CD45low,CD13+ cells revealed homogeneous expression of CD105 and D7-FIB, further demonstrating their fibroblastic nature (Figure 6C, ii and iv). This population of adherent SF fibroblasts (CD45low,CD13+ cells) appeared with the highest frequency in SF from OA groups (4.4, 7.0, and 24.0% of adherent cells; n = 3) compared with other patient groups. In the majority of RA samples (n = 8), it was barely detectable except for 2 patients (1.3% and 0.25% of adherent cells), all correlating with the MPC counts obtained by the CFU-F assay. CD45low,CD13+ cells were also present in other arthropathies (pseudogout, 6% of adherent cells).

In all specimens tested, adherent SF fibroblasts (CD45low,CD13+ cells) uniformly expressed LNGFR (n = 3) (Figure 6C, v). Levels of LNGFR expression were similar to those of BM MPCs enriched by plastic adherence (Figures 6A and C, v). To further confirm LNGFR expression by SF fibroblasts in vivo, we conducted direct phenotyping of OA SF by 4-color flow cytometry (n = 2). SF CD45low,D7-FIB+ cells were CD13+ (Figure 6D, iv) and LNGFR+ (Figure 6D, v). This indicated that SF fibroblasts had the same in vivo phenotype as BM MPCs, including the expression of LNGFR.


It has long been known that SF from patients with arthritis contains fibroblastic cells capable of in vitro proliferation (26–28), but their multipotentiality was not investigated. This study demonstrates the presence of rare MPCs in the SF of subjects with arthritis and links them with clonogenic SF fibroblasts. This is also the first demonstration that SF fibroblasts are closely related in phenotype to BM MPCs, including the specific in vivo expression of LNGFR. Furthermore, we showed that individual MPCs can survive in the viscous, antiadhesive medium of the SF in vivo and, in spite of their pathologic environment, can maintain their multipotentiality in vitro. Finally, the fact that MPCs were found to be much more numerous in OA SF compared with other arthropathies suggests their possible role in the pathophysiology of arthritis, and OA in particular (10–12).

What are the tissue origins of SF MPCs? It is possible that these may be derived from disrupted cartilage, bone, synovium, periosteum, or BM itself. The fact that superficial cartilage contains chondroprogenitors has recently been demonstrated in the setting of neonatal animals (29). Moreover, a high degree of plasticity (more or less comparable to multipotential MPCs) has been documented for human articular chondrocytes dedifferentiated due to in vitro culture expansion (30, 31), with more recent studies demonstrating that this dedifferentiation was not required for the capacity of some cartilage-derived cells to differentiate to skeletal muscle in vivo (32). Besides cartilage, SF MPCs could originate from synovium, for instance, from type B synovial lining cells shed into the joint lumen or from vascular pericytes of the subsynovium, because the multipotential nature of pericytes was documented before (33, 34). Although it is well known that synovium contains multipotential MPCs (9, 35), their topography is still unknown.

Alternatively, MPCs could be released into the joint space from the infrapatellar fat pad (36) or could migrate directly from BM via small vascular channels (found in some animals, but not in humans) (14). Even in the absence of channels, cartilage and bone damage, particularly in OA, could allow direct access of BM MPCs into the joint cavity. This seems likely and might explain not only the high numbers of MPCs in OA SF, but also the observed phenotype similarity between multipotential cells in both localities. The significance of in vivo LNGFR expression in this context is unknown; its expression in early fetal subepithelial mesenchyme and restriction to pericytes in adults (37) point to some association with immaturity/multipotentiality, but this has not yet been tested and demonstrated experimentally. Interestingly, LNGFR was recently found to be preferentially expressed on keratinocyte stem cells (38).

It remains a possibility that MPCs are normally present in SF. Early studies described the presence of mesenchymal cells (such as synoviocytes and chondrocytes) in normal SF (39). A recent study of BM MPCs demonstrated that they are capable of adhesion-independent survival and expansion as floating single cells (40). Normal SF cells can indeed produce monolayers of fibroblastic cells in culture, albeit not yet tested for multipotentiality (Pascal E: personal communication). If MPCs are really present in normal SF, one can envision their physiologic role in the repair of slight damage to superficial cartilage by filling-in defects from the top down. Some degree of spontaneous repair in this manner was previously documented in a rabbit model (41). This situation would likely require some altered adhesive cell–matrix interactions at the site of damage, considering the antiadhesive properties of SF in general. Another recent study proposed that fibroblasts “floating” in SF (also suggested to originate from BM MPCs or “dropped out” of synovium) are involved in joint destruction in RA (25). These findings, however, do not preclude the possibility of SF MPCs having some role in physiologic joint repair, both in healthy individuals and those with OA, because regeneration may indeed have gone awry in RA (11). Direct proof of these hypotheses requires further testing.

In conclusion, this is the first study to show that SF from patients with arthritis contains a rare population of clonogenic tripotential MPCs that are greater in number in OA compared with RA and other arthropathies. The prevalence of such cells in OA suggests their local origin, more likely from disrupted cartilage, bone, or BM. Together with many recent studies, this strengthens the idea that joint tissues per se are rich in MPCs. What their physiologic role in healthy subjects is and how resident MPCs can be used for inducing self-repair in arthritis are the subjects of future studies. In addition, we presented a new methodology for the phenotypic analysis of rare MPCs using an adherence-based enrichment. Considering the difficulties of rare cell phenotyping, our study provides a basis for further work aimed at defining phenotypic and molecular signatures of human tissue–resident MPCs in vivo.


We thank the staff of the Geoffrey Giles Theatre of St. James's University Hospital for their help in the collection of bone marrow specimens. Many thanks to Drs. Colin Pease, Cathy Lawson, Ai Lyn Tan, Adam Greenstein, Ann Morgan, Alice Lorenzi, and Helena Marzo-Ortega, and Mrs. Julia Holdsworth for the collection of SF specimens. We also thank Dr. Alastair Mackay of Osiris Therapeutics for kindly providing preselected serum and Dr. Richard A. Jones for providing conjugated monoclonal antibodies and controls.