Functional mesenchymal stem cell niches in adult mouse knee joint synovium in vivo

Authors


Abstract

Objective

We previously reported that human synovium contains cells that, after culture expansion, display properties of mesenchymal stem cells (MSCs). The objective of this study was to identify MSCs in native synovium in vivo.

Methods

To identify stem cells in the synovium in vivo, a double nucleoside analog cell-labeling scheme was used in a mouse model of joint-surface injury. For labeling of slow-cycling cells, mice received iododeoxyuridine (IdU) for 30 days, followed by a 40-day washout period. For labeling of cells that proliferate after injury, mice underwent knee surgery to produce an articular cartilage defect and received chlorodeoxyuridine (CIdU) for 4 days, starting at multiple time points after surgery. Unoperated and sham-operated joints served as controls. Knee joint paraffin sections were analyzed by double and triple immunostaining to detect nucleoside analogs, conventional MSC markers, and chondrocyte-lineage markers.

Results

Long-term–retaining, slow-cycling IdU-positive cells were detected in the synovium. At 4 days and 8 days after injury, there was marked proliferation of IdU-positive cells, which costained for CIdU. IdU-positive cells were nonhematopoietic, nonendothelial stromal cells, were distinct from pericytes, and stained positive for MSC markers. MSCs were phenotypically heterogeneous and located in topographically distinct niches in the lining layer and the subsynovial tissue. Twelve days after injury, double nucleoside–labeled cells within synovium were embedded in cartilage-specific metachromatic extracellular matrix and costained positive for the chondrocyte-lineage markers Sox9 and type II collagen.

Conclusion

Our findings provide the first evidence of the existence of resident MSCs in the knee joint synovium that undergo proliferation and chondrogenic differentiation following injury in vivo.

The synovial membrane is a tissue that lines the joint cavity of synovial joints and consists of a lining layer of macrophage-like (type A) and fibroblast-like (type B) synoviocytes and a loose sublining tissue. In the healthy joint, type A synoviocytes act in innate immunologic defense and support adaptive immunity, while type B synoviocytes function to regulate the release of nutrients and molecules, including hyaluronan, into the synovial fluid (1).

In response to injury of various types, including trauma, the synovial membrane rapidly becomes hyperplastic (2, 3). It is commonly believed that synovial hyperplasia is sustained mainly by stromal cells including type B (fibroblast-like) synoviocytes, also called synovial fibroblasts, with infiltration of blood-borne inflammatory and immune cells particularly in inflammatory joint diseases such as rheumatoid arthritis (4). The biologic function of synovial stromal cell proliferation is likely to depend on the nature of the injury, but it is believed to have a pivotal role in joint homeostasis and disease. Deregulation of this process is thought to contribute to the formation of pannus, which in rheumatoid arthritis causes destruction of cartilage and bone (4). Despite the high frequency and remarkable biologic and clinical relevance of synovial hyperplasia, very little is known about the identity of the synovial cells that proliferate following injury.

We provided the first evidence that the synovial membrane in adult humans contains cells that, after release and culture expansion, display features of multipotent mesenchymal/stromal stem cells (MSCs) (5–7). MSCs are defined as fibroblast-like cells that undergo sustained in vitro growth and have the capacity to form mesenchymal tissues such as cartilage and bone (8). After culture expansion, synovial membrane MSCs express the conventional markers of cultured MSCs including CD44, CD73, CD90, and CD105 but do not express hematopoietic or endothelial markers (9, 10). Despite extensive in vitro characterization, equivalent cells within their native synovial tissue in vivo have not been identified because of a lack of specific markers.

In adults, as reported for tissues and organs such as the hair follicle and the intestinal mucosa (11, 12), stem cells are quiescent slow-cycling cells that, through asymmetric cell division, generate transit-amplifying cells that, after a proliferative burst, differentiate to replace mature cells that are lost due to physiologic turnover. Following a stress stimulus such as an injury, this system rapidly intensifies, and the slow-cycling cells undergo intense proliferation. In this study, we took advantage of these functional features of adult stem cells and adapted an in vivo scheme of double nucleoside analog cell-labeling (13) to positively detect stem cells within the synovium of knee joints, using a validated mouse model of traumatic joint-surface injury (14). Here, we provide the first evidence of the existence in vivo, within postnatal knee joint synovium, of nonhematopoietic, nonendothelial stromal stem cells with a phenotype compatible with MSCs, distinct from pericytes, which proliferated following articular cartilage injury and, under specific experimental conditions, differentiated into chondrocytes in areas of cartilage metaplasia within the synovium.

MATERIALS AND METHODS

Animals and nucleoside analog administration.

All animal experiments were approved by the UK Home Office. Three-week-old male C57BL/6 mice received the artificial nucleoside iododeoxyuridine (IdU; Sigma) with the drinking water for 30 days, at a concentration of 1 mg/ml. IdU administration was then stopped for the next 40 days. Thereafter, mice underwent knee surgery to produce an articular cartilage injury in the patellar groove of the left femur, while the right leg was sham operated as detailed below. Mice received the artificial nucleoside chlorodeoxyuridine (CIdU; Sigma) either directly after surgery or on day 4 or day 8 after surgery, via subcutaneous injection of 200 μl at a concentration of 10 mg/ml in sterile phosphate buffered saline (PBS) followed by administration of CIdU at a concentration of 1 mg/ml in the drinking water for 4 days. At all time points, control mice received identical treatment with IdU and CIdU but underwent no surgical procedure. In another experiment, transgenic mice expressing the LacZ gene in pericytes and smooth muscle cells (15) received IdU as described above, without undergoing surgery. These mice were killed at time points similar to those at which treated mice were killed, but they did not receive CIdU.

Surgery.

Mice were subjected to surgery to produce a joint-surface injury, as previously described (14). Briefly, mice were anesthetized, and a medial parapatellar arthrotomy was performed under a dissection microscope. The joint was extended, and the patella was dislocated laterally. The joint was then fully flexed, and a longitudinal full-thickness injury was made in the patellar groove, using a custom-made device. The patellar dislocation was then reduced. The joint capsule and the skin were sutured in separate layers. The contralateral knee was subjected to arthrotomy and patellar dislocation without cartilage injury (sham-operated control).

Histologic assessment.

The mice were killed at the end of CIdU administration. The knee joints were dissected and fixed in 2% paraformaldehyde and 0.05% glutaraldehyde in PBS for 1 hour at room temperature. After decalcification for 2 weeks in 4% EDTA in PBS, knee joints were dehydrated and embedded in paraffin. Five-micrometer–thick sections were placed on Superfrost Plus slides (Menzer-Gläser). Sections were rehydrated and stained with hematoxylin and eosin or 1% toluidine blue, according to routine protocols.

Immunohistochemical analysis.

Paraffin sections were immunostained using protocols that were optimized for each antigen. After dewaxing and rehydration were performed, antigen retrieval was either enzyme based, using porcine pepsin (Sigma) in 0.2N HCl at concentrations from 0.5 mg to 3 mg for 45 minutes at 37°C, or was performed by heating sections in an EDTA-based buffer solution (pH 9) for 30 minutes at 100°C in an autostainer (Leica).

For immunohistochemical analysis, endogenous peroxidase was quenched for 10 minutes with 3% H2O2 in water, followed by blocking of endogenous avidin and biotin, using the Avidin/Biotin Blocking Kit (Vector) according to the manufacturer's instructions. Nonspecific binding was blocked with the immunoglobulin-blocking reagent and protein concentrate contained in the Mouse on Mouse (M.O.M.) Basic Kit (Vector). Primary antibodies for IdU or CIdU were applied for 1 hour at room temperature in 500 units of DNase (Sigma). Biotinylated anti-mouse IgG and anti-rat IgG antibodies were applied for 30 minutes. Detection of the signal was performed by incubation with the avidin−biotin−peroxidase reagent included in the Vectastain Elite ABC Kit (Vector) followed by development with 3,3′-diaminobenzidine peroxidase substrate including nickel solution to produce black staining. Following dehydration, sections were mounted in Depex mounting medium and analyzed with an Axioskop 40 microscope connected to an Axiocam (Zeiss).

For immunofluorescence analysis, autofluorescence was quenched with two 5-minute washes with 50 mM NH4CI, and nonspecific binding was blocked with 1% bovine serum albumin for 45 minutes. Sections were incubated with primary antibodies overnight at 4°C, and antigens were detected by incubation with Alexa Fluor secondary antibodies for 1 hour at room temperature. Table 1 lists the antibodies used in this study as well as the suppliers, the working dilutions, and the methods of antigen retrieval. After nuclear counterstaining using either TO-PRO-3 (Invitrogen) or Sytox green, sections were mounted with Mowiol and analyzed with a Zeiss 510 META Laser Scanning Confocal Microscope. Images were obtained with an Axiocam.

Table 1. Antibodies used in the study, suppliers, working dilutions, and methods of antigen retrieval*
AntibodyCodeSupplierOptimal dilutionAntigen retrieval method
  • *

    Includes primary and secondary antibodies used for immunohistochemical analysis and immunofluorescence staining, with optimal dilutions. IdU = iododeoxyuridine; CIdU = chlorodeoxyuridine; PDGFRα = platelet-derived growth factor receptor α; Sca-1 = stem cell antigen 1; vWF = von Willebrand factor; Santa Cruz = Santa Cruz Biotechnology.

  • Either boiling (B) for 30 minutes in a pH 9 buffer or incubation at 37°C for 45 minutes with pepsin (P).

Mouse anti-IdUab8955Abcam1:100B or P, 0.5–3 mg/ml
Rat anti-CIdUab6326Abcam1:100B or P, 0.5–3 mg/ml
Rabbit anti-CD73ab71822Abcam1:50B
Rabbit anti-Sox9ab59265Abcam1:50B
Rabbit anti–type II collagenab21291Abcam1:200P, 3 mg/ml
Rabbit anti-PDGFRαab61219Abcam1:50B
Rabbit anti-vimentinab45939Abcam1:200B
Rat anti-CD45550539BD PharMingen1:20B
Rat anti–Sca-1557403BD PharMingen1:200P, 0.5 mg/ml
Rabbit anti-vWFA-0082Dako1:200B or P, 3 mg/ml
Rabbit anti-p75AB1554Chemicon1:100B
Rabbit anti–β-galactosidaseAB1211Chemicon1:300P, 0.5 mg/ml
Goat anti-CD105AF1320R&D Systems1:20B
Goat anti-p75sc-6188Santa Cruz1:50B
Goat anti-CD146sc-18940Santa Cruz1:100B
Rat anti-CD44sc-18849Santa Cruz1:100B
Goat anti–c-Kitsc-1494Santa Cruz1:100P, 0.5 mg/ml
Biotinylated rabbit anti-ratBA-4001Vector1:200
Biotinylated anti-mouseMKB-2225Vector1:250
Alexa Fluor 488 goat anti-mouseA-11029Invitrogen1:500
Alexa Fluor 647 chicken anti-mouseA-21463Invitrogen1:200
Alexa Fluor 488 goat anti-ratA-11006Invitrogen1:500
Alexa Fluor 594 goat anti-ratA-11007Invitrogen1:500
Alexa Fluor 594 goat anti-rabbitA-11037Invitrogen1:500
Alexa Fluor 594 chicken anti-goatA-21468Invitrogen1:500
Alexa Fluor 488 donkey anti-goatA-11055Invitrogen1:100

Quantification of fluorescence intensity.

For each condition, 3 sections per joint from 3 mice underwent immunofluorescence staining for IdU. From every section, images of different areas of the synovium and articular cartilage were obtained with an Axioscope 40 microscope connected to an Axiocam, and the fluorescence intensity of IdU-positive cells was measured using ImageJ software (NIH Image, National Institutes of Health, Bethesda, MD; online at: http://rsbweb.nih.gov/ij/). Values were expressed as the percentage of IdU-positive synovial cells normalized to IdU-positive chondrocytes within each section, counting a minimum of 100 cells of each cell type per section (n = 3). Chondrocytes were not shown to proliferate; therefore, they were used as an intrasection reference for internal control to eliminate intersection variability.

Quantification of labeled cells.

For each condition, 3 sections per joint from 3 mice underwent immunofluorescence staining for IdU and CIdU. Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI). From every section, 3 images of different areas of the synovium were obtained. Cells were counted for single positivity of IdU or CIdU or for double positivity. Values were expressed as the percentage of positive cells compared with the total number of DAPI-counterstained cells, counting at least 1,000 cells per condition (n = 3).

For quantification of cells positive for IdU and MSC markers, 3 sections per joint from 3 uninjured control mice were double-stained for IdU and the corresponding MSC marker. From every section, images of different areas of the synovium were obtained. Cells were counted for single positivity of IdU or MSC markers or for double positivity. Values were expressed as the percentage of double-positive cells compared with the total number of IdU-positive cells or the total number of MSC marker–positive cells, counting at least 100 IdU-positive cells or MSC marker–positive cells per joint (n = 3).

RESULTS

Proliferation of slow-cycling cells in synovium in response to joint-surface injury.

In well-studied systems such as the hair follicle and the intestinal crypt, postnatal stem cells are quiescent, slow-cycling cells in resting conditions but undergo a burst of proliferation following injury (16). To investigate whether adult joint synovium has cells that display the functional behavior of stem cells, we adapted a previously reported double nucleoside cell-labeling scheme (13) in a standardized mouse model of joint-surface injury (14) (Figures 1A and C).

Figure 1.

Proliferation of slow-cycling cells following joint-surface injury. A, Double nucleoside analog cell-labeling scheme. B, Expected labels at different time points after injury: a, Iododeoxyuridine (IdU)–positive cells do not proliferate and retain IdU (green); b, IdU-positive cells proliferate and acquire chlorodeoxyuridine (CIdU) (red) while gradually losing IdU, hence appearing yellow, orange, or red; c, unlabeled IdU-negative cells (white) proliferate and acquire CIdU (red). C, Hematoxylin and eosin–stained sections of an uninjured control joint and a representative injured joint. F = femur; P = patella; S = synovium. Asterisk indicates articular cartilage injury. Bars = 200 μm. D, Left and middle, Higher-magnification views of boxed areas in C, showing single immunohistochemical staining for IdU and CIdU. c = uninjured control. Bars = 100 μm. Right, Double immunofluorescence staining for IdU and CIdU in sections corresponding to the boxed areas shown in the left column. Bars = 10 μm. E, Fluorescence intensity of IdU-positive synovial cells normalized to the fluorescence intensity of IdU-positive articular chondrocytes within each section in uninjured controls. F, Double immunofluorescence staining for IdU and CIdU in the subchondral bone marrow of the patellar groove region. G, Proportions of cells positive for IdU, CIdU, and both IdU and CIdU on days 4, 8, and 12. Values in E and G are the mean ± SEM results from 3 mice per group. 4dpi = 4 days after injury; 4dps = 4 days after sham operation.

We anticipated the following scenarios. During the initial labeling period with IdU, all dividing cells including stem cells and proliferative progenitors prior to differentiation would become labeled. During the 40-day washout period, the IdU label would be diluted and become undetectable in rapidly dividing cells while being retained by slow-cycling (stem) cells or by cells that had stopped dividing, e.g., as a consequence of differentiation. During the labeling period with CIdU, all dividing cells would incorporate CIdU, but only slow-cycling long-term–retaining IdU–positive cells would become double labeled. The IdU-labeled cells that did not divide after injury would not take up CIdU (Figure 1B).

In uninjured control mice, slow-cycling long-term–retaining IdU–positive cells were detected in the synovium of the knee joints, while CIdU-positive cells were infrequent and rarely costained for IdU (Figure 1D). On day 4 and day 8 after injury, there was marked accumulation of IdU and CIdU double-positive cells; this accumulation was still detectable 12 days after injury, although at a lower degree (Figure 1G). Of note, the intensity of fluorescence of IdU-positive cells within synovium became weaker over time (Figure 1E), presumably as an effect of repeated cell divisions following injury. These results demonstrate that the adult knee joint synovial membrane contains slow-cycling cells, and that following articular cartilage injury, at least a subset of these cells proliferate to generate a pool of transit-amplifying cells in vivo. Notably, under our experimental conditions, no difference in nucleoside labeling was detected in the subchondral bone marrow of the patellar groove region between uninjured controls and injured mice, at all time points tested (Figure 1F).

MSC-like phenotype of nucleoside-labeled cells.

We next investigated the phenotype of IdU-positive cells in the synovial tissue, using multiple immunofluorescence stainings in situ. In uninjured mice, IdU-positive cells were not of endothelial or hematopoietic origin, because they consistently stained negative for the endothelial marker von Willebrand factor (vWF) (Figures 2A and B), the panhematopoietic marker CD45 (Figures 2C and D), and the hematopoietic progenitor cell marker c-Kit (17) (Figures 2I and J). In injured mice, both CD45 and c-Kit were detected in an increasing number of cells, but these cells did not costain for IdU (results not shown). In contrast, most IdU-positive cells stained positive for vimentin, thus confirming the mesenchymal nature of the slow-cycling long-term–retaining cells (Figures 2E and F).

Figure 2.

Mesenchymal stem cell (MSC)–like phenotype of iododeoxyuridine (IdU)–positive cells in synovium. A–T, Representative images of immunofluorescence staining for IdU (green) and lineage markers (red and light blue) in uninjured joints. Nuclei are counterstained in dark blue. Enlarged images of the boxed areas are shown in the adjacent panels. IdU-positive cells (dashed arrows) were negative for the endothelial marker von Willebrand factor (vWF) (B) (arrowhead) and the panhematopoietic marker CD45 (D) (arrowheads), whereas they were positive for the mesenchymal/stromal cell marker vimentin (F) (arrows). A subset of stem cell antigen 1 (Sca-1)–positive cells costained for IdU (H) (arrow), while other Sca-1–positive cells were IdU negative (H) (arrowhead). Cells positive for both IdU and Sca-1 (J) (arrow) were negative for c-Kit (J) (arrowhead). IdU-positive cells costained for the MSC markers platelet-derived growth factor receptor α (PDGFRα) (L) (arrow), CD44 (N) (arrow), CD73 (P) (arrows), and p75 (R) (arrows). Arrowheads in L, N, P, and R indicate cells positive for MSC markers but negative for IdU. CD105 stained the endothelium and scattered cells (T) (arrowhead), which were negative for IdU. Bars = 10 μm. U and V, Proportions of cells positive for IdU and MSC markers (IdU+M+) in relation to total IdU-positive cells (U) or to total MSC marker–positive cells (V). Values are the mean ± SEM results from 3 mice per group.

Stem cell antigen 1 (Sca-1) is described as a positive selection marker for MSCs in fluorescence-activated cell sorting when using C57BL/6 mouse bone marrow (18). In uninjured knee joint synovium, we detected isolated Sca-1–positive cells that costained for IdU (Figures 2G and H). Because Sca-1 is also known to be expressed in conjunction with c-Kit by hematopoietic progenitors (19), we carried out triple immunofluorescence staining for Sca-1, c-Kit, and IdU. We detected cells positive for IdU and Sca-1 but negative for c-Kit (Figures 2I and J). These data confirmed that the slow-cycling long-term–retaining IdU–positive cells in synovium were nonhematopoietic in nature, with subsets having a Sca-1–positive, c-Kit–negative phenotype compatible with that of MSCs. IdU-positive cells also stained for platelet-derived growth factor receptor α (PDGFRα) (Figures 2K and L), a putative marker of murine bone marrow MSCs in vivo (20).

We then investigated the expression of markers that are known to be associated with human MSCs in culture (21). In uninjured mice, subsets of IdU-positive cells stained positive for established MSC markers such as the hyaluronan receptor CD44 (22) (Figures 2M and N), CD73 (23) (Figures 2O and P), and low-affinity nerve growth factor receptor p75, which is reported to be expressed by uncultured prospective MSCs in human bone marrow (24) (Figures 2Q and R). In contrast, CD105 was expressed by endothelium and scattered cells within synovium but did not colocalize with IdU-positive cells (Figures 2S and T). Taken together, these findings indicate that the synovium of mice contains slow-cycling mesenchymal cells that stain positive for known markers of MSCs.

We next determined the proportions of cells within the IdU-positive cell pool that costained for MSC markers and the proportions of cells within MSC marker–labeled cell populations that were positive for IdU. As shown in Figures 2U and V, none of the markers tested was specific for IdU-positive cells, because there were cells that were positive for MSC markers but negative for IdU. In addition, none of the individual MSC markers labeled all IdU-positive cells, thus suggesting a heterogeneous phenotype of the slow-cycling IdU-positive cells in synovium.

Topographic location and phenotypic heterogeneity of slow-cycling MSC-like cell populations within synovium of uninjured joints.

We next investigated the topographic distribution of IdU-positive cells in the synovium across the knee joint and also analyzed their phenotype by triple immunofluorescence staining, using multiple combinations of MSC markers in conjunction with IdU. IdU-positive cells were relatively evenly distributed in the synovium. They were scattered and often grouped in small clusters of 3–6 cells, with no specific topographic location. Notably, IdU-positive cells were observed in both the synovial lining and the subsynovial connective tissue. Most IdU-positive cells costained for vimentin and p75 in both synovial layers (Figure 3A). CD73 colocalized with IdU exclusively in the subsynovial tissue, where some of the IdU-positive cells that were positive for CD73 also stained positive for p75 (Figure 3B). IdU-positive cells that costained for CD44 were observed in the synovial lining, and some of these cells were also positive for p75 (Figure 3C). The topographic segregation of IdU-positive cell subsets within the synovium was further confirmed with triple immunofluorescence staining for IdU, CD44, and CD73. While cells that were positive for both IdU and CD73 were located exclusively in the sublining tissue, cells positive for IdU and CD44 were observed mainly in the lining layer or juxtaposed to it; however, these 2 MSC markers, CD44 and CD73, appeared to be mutually exclusive (Figure 3D). These results indicate the presence of topographically distinct and phenotypically heterogeneous IdU-positive MSC-like cell subsets in both layers of the synovium.

Figure 3.

Topography, relationship with pericytes, and phenotypic heterogeneity of MSC-like IdU-positive cells. Representative images of immunofluorescence staining for IdU (green) and lineage markers (red and blue) in uninjured joints are shown. A, A subset of IdU-positive cells costained for p75 and vimentin (arrows), while some cells were solely IdU positive (arrowheads). B, Some cells in the subsynovial tissue were positive for IdU, p75, and CD73 (arrows), and some were positive only for IdU and p75 (arrowheads). C, In the lining layer, cells positive for IdU, p75, and CD44 (arrows) were observed beside cells that were positive only for IdU and p75 (arrowheads). D, Cells positive for both IdU and CD44 were confirmed in the lining layer (arrows), whereas cells positive for IdU and CD73 were observed in subsynovial tissue (arrowheads). E, Triple immunofluorescence staining for IdU, CD146, and β-galactosidase (β-gal) in knee sections from transgenic mice expressing β-galactosidase in pericytes is shown. F, A higher-magnification view of the boxed area in E showed that CD146 is a pericyte marker (arrow) and confirmed that pericytes were IdU negative (dashed arrow). G, Triple immunofluorescence staining for IdU, CD146, and vWF is shown. H, A higher-magnification view of the boxed area in G showed that IdU-positive cells (dashed arrow) are negative for the pericyte marker CD146 (arrow) and the endothelial marker vWF. I, Triple immunofluorescence staining for IdU, CD146, and p75 is shown. J–L, Higher-magnification views of the boxed areas in I showed that pericytes are negative for IdU but positive for p75 (arrowheads), while cells positive for IdU and p75 that are located close to blood vessels (arrows) are negative for CD146. Merged images for IdU and p75 (J) and for IdU and CD146 (K), and a triple-channel merged image (L) are shown. Bars = 10 μm. See Figure 2 for other definitions.

Relationship between MSCs and pericytes in the knee synovium.

Pericytes were recently shown to display MSC-like properties in multiple tissues (25). In our study, subsets of IdU-positive cells were located close to blood vessels; hence, we investigated the relationship between perivascular IdU-positive cells and pericytes. To this end, we used a transgenic mouse in which the reporter β-galactosidase (β-gal) is expressed in pericytes and smooth muscle cells (15). CD146, a known pericyte marker (26, 27), colocalized with β-gal in cells with topography compatible with pericytes, thus validating this transgenic mouse postnatally (Figures 3E and F). IdU-positive cells were observed close to β-gal–positive pericytes in uninjured mice but did not colocalize with β-gal or CD146 (Figures 3E–L). Because CD146 stained a greater number of perivascular (β-gal–positive and β-gal–negative) cells, we also performed immunofluorescence staining for IdU, CD146, and vWF (Figures 3G and H). Under our experimental conditions, in all mice tested, IdU-positive cells were distinct from CD146-positive pericytes and vWF-positive endothelial cells. Similar results were obtained when using other pericyte markers such as NG2, PDGFRβ, and α-smooth muscle actin (results not shown).

Pericytes in tissues such as retina and liver are known to express p75 (27, 28). To investigate the relationship between cells positive for both IdU and p75 and pericytes positive for p75, we performed triple immunofluorescence staining to codetect IdU, p75, and CD146 in uninjured mice. In a subset of blood vessels, p75 was expressed by CD146-positive pericytes, all of which were negative for IdU. We detected cells that were double positive for IdU and p75 but were negative for CD146; these cells had a perivascular location that was not compatible with the topography of pericytes (Figures 3I–L). Of note, in the synovium of injured mice, at all 3 time points postinjury, we observed proliferation of p75-positive cells that costained for both IdU and CIdU, but we never detected proliferation of CD146-positive pericytes (results not shown). Taken together, our findings indicate that the IdU-positive slow-cycling MSC-like cells in the synovium are distinct from pericytes.

Chondrogenic fate of nucleoside-labeled cells in vivo.

After undergoing joint surgery, 5–10% of animals accidentally undergo dehiscence of the sutures. This results in patellar dislocation with secondary cartilage metaplasia within the synovium. Twelve days after injury (but not at earlier time points), in mice with patellar dislocation, toluidine blue staining revealed large areas of metachromasia within the synovium, mainly adjacent to the femoral and patellar articular cartilage (Figure 4K). Similar but minute areas of weak metachromasia were detected in injured mice without patellar dislocation at 12 days after injury (Figure 4J). The metachromasia in synovium shown by toluidine blue staining was observed neither in uninjured control mice (Figure 4A) nor in injured mice at earlier time points (results not shown).

Figure 4.

Detection of cells positive for both iododeoxyuridine (IdU) and chlorodeoxyuridine (CIdU) in the area of ectopic cartilage metaplasia within the synovium of injured mice. A, J, and K, Toluidine blue staining for metachromasia in the synovium was negative in control mice (A) but was positive 12 days after injury in nondislocated joints (J) (arrows) and dislocated joints (K). B and F, Higher-magnification images of the boxed area in A are shown. L and P, Higher-magnification images of the boxed area in K are shown. B–I, No Sox9 (B–E) or type II collagen (Col2) (F–I) was detectable in the synovium of uninjured joints. C–E, Merged images for IdU and Sox9 (C) and for CIdU and Sox9 (D), and a triple-channel merged image (E) are shown, all of which were enlarged from the boxed area in B. G–I, Merged images for IdU and type II collagen (G) and for CIdU and type II collagen (H), and a triple-channel merged image (I) are shown, all of which were enlarged from the boxed area in F. L–S, At 12 days postinjury, Sox9 (L–O) and type II collagen (P–S) were expressed by cells mostly positive for both IdU and CIdU (arrows) or for only IdU (dashed arrows). M–O, Merged images for IdU and Sox9 (M) and for CIdU and Sox9 (N), and a triple-channel merged image (O) are shown, all of which were enlarged from the boxed area in L. Q–S, Merged images for IdU and type II collagen (Q) and for CIdU and type II collagen (R), and a triple-channel merged image (S) are shown, all of which were enlarged from the boxed area in P. Bars = 10 μm.

We took advantage of the complication of patellar dislocation in injured mice to provide proof of concept that double nucleoside–labeled cells in synovium have another typical stem cell feature, i.e., the ability to differentiate into mature cells such as chondrocytes. To this end, sections were costained for IdU, CIdU, and the early chondrocyte lineage marker Sox9, a transcription factor required for cartilage formation (29). Although no Sox9 was detectable in the synovium of uninjured joint sections at all time points tested (Figures 4B–E), a number of cells expressed the marker of early chondrogenesis 12 days after injury (Figures 4L–O), with positive cells located in the area of metachromasia and most also being positive for IdU and CIdU. Notably, no Sox9-positive cells stained only for CIdU, while some Sox9-positive cells stained only for IdU, suggesting either that IdU-positive cells had differentiated without proliferation or that they had undergone very few cell cycles prior to day 8 postinjury.

Type II collagen, a chondrocyte-specific marker, was detected in consecutive sections mostly associated with cells showing an enlarged rounded nucleus (Figures 4P–S). As observed for Sox9, the majority of cells positive for type II collagen costained for both IdU and CIdU, but a number of cells costained only for IdU. Type II collagen was not detectable in the synovium at either 4 days or 8 days after injury (data not shown) or in knee joint sections obtained from uninjured mice (Figures 4F–I). Colocalization of IdU and CIdU with Sox9 and type II collagen was also observed 12 days after injury in the minute metachromatic areas of synovial tissue in nondislocated joints (data not shown). These data indicate that double nucleoside–labeled cells within metachromatic synovium display a chondrocyte-like phenotype.

DISCUSSION

Synovial hyperplasia is a very common phenomenon following joint injuries, but the identity of the underpinning proliferating stromal cells is not known. Similar to bone marrow and most connective tissues (30–33), the stroma of the synovial tissue contains cells that after release and culture expansion display features of MSCs (5). However, despite extensive in vitro characterization, the equivalent cells within their native tissue in vivo have not been identified because of a lack of specific markers.

In the current study, we adapted to our mouse model of traumatic knee joint–surface injury a double nucleoside analog–labeling scheme that was recently used to positively detect neural stem cells in vivo (13). We identified and characterized, in the mouse knee joint synovium, a population of quiescent, slow-cycling nonhematopoietic, nonendothelial, MSC-like stromal cells, present in both the lining layer and sublining tissue, that proliferated following injury to produce synovial hyperplasia (Figure 5).

Figure 5.

Schematic representations of cell phenotypes identified using the double nucleoside analog cell-labeling strategy. A, Schematic drawing of an uninjured control synovial joint. B, Details of the dashed box in A, showing cell populations in the synovium of uninjured joints. Iododeoxyuridine (IdU)–positive cells (green) were located in both the synovial lining (SL) and the subsynovial tissue (SST). Subsets of IdU-positive cells displayed a mesenchymal stem cell phenotype. IdU-negative cells (blue) included hematopoietic lineage cells (HC), endothelial cells (EC), pericytes (PC), and other cell types of unknown phenotype. C, Schematic drawing of a synovial joint 12 days after articular cartilage injury (arrowhead). D, Details of the dashed box in C, showing cell populations in the synovium. Proliferating cells were detected in both the synovial lining and the subsynovial tissue and were either double positive for IdU and chlorodeoxyuridine (CIdU; orange) or single positive for CIdU (red). Subsets of cells positive for IdU and CIdU and cells positive only for IdU (green) expressed chondrocyte lineage markers. The boxed areas in B and D show cell phenotypes. B = bone; C = cartilage; SM = synovial membrane; β-gal = β-galactosidase; SC = synovial cavity (see Figure 2 for other definitions).

Although IdU-positive cells expressed conventional MSC markers, none of the MSC markers tested was sufficient to label all IdU-positive cells, and all MSC markers also labeled cells that were negative for IdU. Moreover, IdU-positive cells were heterogeneous in their phenotype, possibly reflecting the coexistence of functionally distinct cell subsets.

IdU-positive cells were negative for CD105, an established marker of human MSCs in culture. We cannot exclude the possibility that CD105 would identify MSCs in humans but not in mice, because variations in MSC phenotype and biology across species are known (34). In addition, acquisition of CD105 might occur as a result of phenotypic rearrangements during ex vivo cell manipulations (35). Indeed, markers observed on cells in vitro cannot be directly extrapolated to the equivalent native cells in vivo. For this reason, we used an unbiased approach, based on nucleoside labeling, to detect stem cells within the synovium in vivo.

IdU-positive cells were detected in both the lining layer and the sublining tissue. In the lining layer, IdU-positive cells were negative for CD45, thus excluding their monocytic nature (36), but were positive for the MSC markers PDGFRα, p75, and CD44. However, CD44 is also known to be expressed by synovial fibroblasts (37). IdU-positive cells in the lining layer costained for cadherin 11 (results not shown), a known marker of synovial fibroblasts (38). The relationship between MSCs and fibroblasts in the synovium is currently being investigated in our laboratory.

Pericytes are adventitial cells located around small vessels of connective tissues, and data indicate that a subset of MSCs could be or could derive from pericytes (27, 39). In a recent study (25), pericytes in human tissues were reported to be positive for the MSC markers CD44, CD73, CD90, and CD105 and displayed mesenchymal multipotency in vitro (25). However, no evidence of differentiation of pericytes within their environments in vivo by lineage-tracking experiments was provided. In addition, joint tissues were not investigated, and therefore the paradigm of the pericyte origin of the cultured MSCs cannot be extrapolated to synovial tissue. Our findings indicate that the MSC population we identified in the synovium in vivo is phenotypically and functionally distinct from pericytes.

Following traumatic injury to the joint surface, we observed synovial hyperplasia that was sustained by proliferation of mesenchymal cells. Although we cannot exclude the contribution to synovial hyperplasia of incoming cells from bone marrow or from the circulation via the bloodstream, the observations that quiescent long-term–retaining IdU-positive cells were detected within the synovium of uninjured mice, that following injury, the number of IdU-positive and CIdU–double-positive cells increased in parallel with a decrease in the number of IdU-positive cells, and that the intensity of fluorescence of IdU-positive cells in synovium decreased over time following injury make a strong case in support of in situ proliferation of quiescent resident MSCs within synovium.

Twelve days after injury, cells within the synovial cartilage that were positive for both IdU and CIdU expressed the chondrocyte lineage markers Sox9 and type II collagen (Figures 5C and D). Although most cells with a chondrocyte phenotype were double positive for IdU and CIdU, a few scattered cells were positive only for IdU. This suggests that either proliferation was not a prerequisite for chondrogenic differentiation or that IdU-labeled cells had divided very few times before CIdU was administered 8 days after surgery, so that they retained IdU beyond the period of proliferation but were negative for CIdU. Although we cannot exclude the possibility that IdU-positive chondrocytes migrated out of the articular cartilage into the synovium, our study indicated an obvious contribution of synovial cells to the process of cartilage metaplasia in synovium, in keeping with previous observations that treatment of synovial tissue explants in vitro with chondrogenic growth factors induced cartilage formation (40, 41). Cartilage metaplasia was observed throughout the knee joint synovium but was much more pronounced at the transition zone where the synovium connects to the articular cartilage, mimicking the chondrophytes observed in osteoarthritis.

We did not observe any cartilage repair in the C57BL/6 mice used in this study. An identical experimental injury to the joint surface in young adult DBA/1 mice resulted in full healing of the lesion (13). We are currently investigating whether the resident MSCs in synovium are able to contribute to cartilage repair in a permissive mouse strain such as DBA/1. Knowledge of the cellular and molecular mechanisms underlying joint tissue homeostasis, remodeling, and repair in health and disease will help modulate MSC niches pharmacologically to achieve joint tissue regeneration and, in the broader picture, to influence outcomes of joint disorders and restore joint homeostasis.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. De Bari had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Kurth, Dell'Accio, Sharpe, De Bari.

Acquisition of data. Kurth, Crouch, De Bari.

Analysis and interpretation of data. Kurth, Dell'Accio, Augello, De Bari.

Ancillary