Vascular Pericytes Express Osteogenic Potential In Vitro and In Vivo


  • Mary Jo Doherty,

    1. The Wellcome Trust Center for Cell-Matrix Research, School of Biological Sciences, University of Manchester, Manchester, United Kingdom
    Search for more papers by this author
  • Brian A. Ashton,

    1. Department of Rheumatology, Institute of Orthopaedics, Robert Jones and Agnes Hunt Orthopaedic Hospital, Oswestry, Shropshire, United Kingdom
    Search for more papers by this author
  • Susan Walsh,

    1. School of Pharmacy and Pharmacology, University of Bath, Claverton Down, Trimbridge, Bath, United Kingdom
    Search for more papers by this author
  • Jon N. Beresford,

    1. School of Pharmacy and Pharmacology, University of Bath, Claverton Down, Trimbridge, Bath, United Kingdom
    Search for more papers by this author
  • Michael E. Grant,

    1. The Wellcome Trust Center for Cell-Matrix Research, School of Biological Sciences, University of Manchester, Manchester, United Kingdom
    Search for more papers by this author
  • Ann E. Canfield

    Corresponding author
    1. The Wellcome Trust Center for Cell-Matrix Research, Department of Medicine, University of Manchester, Manchester, United Kingdom
    • Address reprint requests to: Dr. A.E. Canfield, The Wellcome Trust Center for Cell-Matrix Research, Department of Medicine, University of Manchester, 2.205 Stopford Building, Oxford Road, Manchester M13 9PT, U.K.
    Search for more papers by this author


At postconfluence, cultured bovine pericytes isolated from retinal capillaries form three-dimensional nodule-like structures that mineralize. Using a combination of Northern and Southern blotting, in situ hybridization, and immunofluorescence we have demonstrated that this process is associated with the stage-specific expression of markers of primitive clonogenic marrow stromal cells (STRO-1) and markers of cells of the osteoblast lineage (bone sialoprotein, osteocalcin, osteonectin, and osteopontin). To demonstrate that the formation of nodules and the expression of these proteins were indicative of true osteogenic potential, vascular pericytes were also inoculated into diffusion chambers and implanted into athymic mice. When recovered from the host, chambers containing pericytes were found reproducibly to contain a tissue comprised of cartilage and bone, as well as soft fibrous connective tissue and cells resembling adipocytes. This is the first study to provide direct evidence of the osteogenic potential of microvascular pericytes in vivo. Our results are also consistent with the possibility that the pericyte population in situ serves as a reservoir of primitive precursor cells capable of giving rise to cells of multiple lineages including osteoblasts, chondrocytes, adipocytes, and fibroblasts.


BONE IS A DYNAMIC TISSUE that in adult life is subjected to continued cycles of remodeling to maintain its structural integrity, adapt to changes in mechanical loading, and to prevent accumulation of microdamage. An obvious corollary of this observation is that there must be continued replacement of the cells responsible for the resorption and formation of bone, osteoclasts, and osteoblasts, respectively. It has been established beyond reasonable doubt that osteoclasts ultimately derive from the hematopoetic stem cell.1 The precise origin of osteoblasts in the postnatal organism is less clear. According to one hypothesis, osteoblasts derive from primitive, multipotential precursors associated with the stromal tissue of bone and marrow.2–9 Inducible osteogenic precursors are also present in organs outside the skeleton (e.g., muscle). These cells only form bone in response to an inducing agent such as bone morphogenetic protein.10,11

In vivo, bone formation always occurs in a vascular environment.12 This observation has led to the suggestion that osteogenic precursors arrive with the invading capillaries and, more specifically, that pericytes can differentiate into functional osteoblasts.12–15 Consistent with this hypothesis are recent studies from this and other laboratories which have provided indirect evidence for the osteogenic potential of vascular pericytes and pericyte-like cells (termed calcifying vascular cells or CVCs) isolated from the aortic intima.13,14,16–22 Pericyte growth in vitro mimics that of osteoblasts in that at high cell density pericytes express alkaline phosphatase (ALP) and osteocalcin and form nodular structures composed of cells and an extracellular matrix rich in type I collagen.16–19,23 Minor amounts of type III collagen are also synthesized by pericytes in addition to a low Mr collagen (termed BRP collagen) which is distinct from, but related to, collagen type X.23,24 The extracellular matrix of these nodules then becomes heavily mineralized as crystals of hydroxyapatite are deposited.16–18 This process occurs in standard growth medium but can be accelerated by the addition of β-glycerophosphate.16 The presence of mineral deposits within the nodules was confirmed by electron microscopy, histochemical staining, and X-ray microprobe analysis.16–18

The principal aim of the present investigation was to define further the relationship between cultured pericytes and bone-forming cells by studying the temporal and spatial pattern of expression of markers of osteoblast differentiation (e.g., STRO-1, osteonectin, osteopontin, osteocalcin, and bone sialoprotein).25–31 Of these proteins, only osteocalcin has been detected in the culture medium from retinal pericytes.17 Pericyte-like cells isolated from the aorta (termed calcifying vascular cells) have been shown to synthesize osteopontin and osteonectin.19 To date, the pattern of expression of these proteins by retinal microvascular pericytes has not been investigated. Most importantly, the present study determined the osteogenic potential of cultured pericytes using the diffusion chamber assay. The results demonstrate for the first time that cultured pericytes are capable of forming bone, cartilage, and fibrous tissue in vivo, suggesting that pericytes can give rise to cells of multiple lineages.


Cells and culture conditions

Pericytes were isolated from adult bovine retinal microvessels as previously described.16,32,33 Cells were cultured initially in either 25-cm2 flasks or on 90-mm dishes (Costar, Cambridge, MA, U.S.A.) in Eagle's minimal essential medium (MEM) supplemented with 20% (v/v) fetal calf serum (FCS), 2 mM glutamine, 1 mM sodium pyruvate, 100 U/ml penicillin, streptomycin, nonessential amino acids (GIBCO BRL, Paisley, U.K.), and 50 μg/ml ascorbic acid (BDH Chemicals, Poole, Dorset, U.K.). Culture medium (20% FCS:MEM) was changed three times per week. Just before confluence, cells were subcultured at a 1:2 split ratio using trypsin-EDTA (0.05%; GIBCO). Cultures were incubated at 37°C in a humidified atmosphere consisting of 5% CO2 and 95% air.

Cells were characterized as pericytes by their morphology and growth pattern, presence of α-smooth muscle actin, immunoreactivity with monoclonal antibody (mAb) 3G5, and lack of immunoreactivity with von Willebrand factor (vWF).16,34 The results presented in this paper were obtained using 12 different preparations of bovine retinal pericytes at first or second passage.

Adult bovine aortic endothelial cells, isolated as previously described,32 fetal bovine tendon fibroblasts, and fetal aortic smooth muscle cells (kindly donated by Dr. Cay Kielty, University of Manchester, Manchester, U.K.) served as controls. These cells were cultured in 20% FCS:MEM as previously described.32


Confluent cultures of pericytes, endothelial cells, fibroblasts, and smooth muscle cells were harvested and seeded onto tissue culture slides (GIBCO) in 20% FCS:MEM. Pericytes were fixed in 1% (w/v) sucrose, 2% (v/v) formaldehyde in phosphate-buffered saline (PBS) at confluence (days 4–8 after plating) and when nodules were present (days 14–25 after plating); the other cell types were fixed at confluence. The cells were permeabilized and stained by indirect immunofluorescence as previously described.16 In some experiments, the DNA intercalating dye 4,6-diamidino-2-phenylindole (DAPI) (1 μg/ml) was also applied to each slide. The slides were washed extensively and finally mounted in Mowiol 4–88 (Harlow Chemical Co. Ltd., Harlow, England) containing 1 mg/ml p-phenylenediamine as an antifade agent.

The specific antisera used to characterize the cell types included: anti-α smooth muscle actin mAb (Sigma, Dorset, U.K.), rabbit anti-human vWF polyclonal antibody (Dako, Buckinghamshire, U.K.), and mAb 3G5 collected in medium from mouse hybridoma cells (CRL-1814, American Type Culture Collection, Rockville, MD, U.S.A.). This antibody was raised against fetal rat brain and reacts with a cell surface ganglioside present on bovine retinal capillary pericytes but shows no reactivity with endothelial cells, retinal pigmented cells, or bovine aortic smooth muscle cells in vitro.34

The other antisera used (kindly donated by Dr. L. Fisher, Bone Research Branch, NIDR, NIH, Bethesda, MD, U.S.A.) were LF-55 (bovine osteonectin), LF-101 (human bone sialoprotein), and LF-123 (human osteopontin), which have previously been shown to exhibit cross-reactivity with the bovine proteins.35,36 Fluorescein isothiocyanate–conjugated rabbit immunoglobulins to mouse immunoglobulins (Dako) or swine immunoglobulins to rabbit immunoglobulins were used as secondary antibodies. Controls included the substitution of the first antiserum by nonimmune serum.

Identification of STRO-1 expression in pericyte cultures

Bovine pericytes were grown on glass coverslips in 20% FCS:MEM until confluent (day 5 postplating) or until large mineralized nodules had formed (day 35 postplating). At both time points, cultures were fixed for 2 minutes in acetone prior to immunostaining with mAb STRO-1 (Developmental Studies Hybridoma Bank, University of Iowa) using APAAP as described previously.37 Briefly, the cell preparations were overlaid with STRO-1 (immunoglobulin M [IgM] subclass) hybridoma culture supernatant. Negative controls included culture supernatant containing IgM antibody directed against the antigen-presenting cell-specific antigen CD80 and tissue culture medium alone. After incubation in a moist chamber for 30 minutes at room temperature, the coverslips were rinsed with 0.05 M Tris-buffered saline (TBS, pH 7.6) and washed again using fresh buffer for a further 5 minutes. Excess buffer was removed and the coverslips were incubated for 30 minutes with biotinylated goat anti-mouse IgM antibodies (μ chain specific; Vector, Peterborough, U.K.) diluted to 1 μg/ml in TBS. The coverslips were washed as described above and incubated for 30 minutes with streptavidin-conjugated ALP (Dako) diluted 1:100 in TBS. The coverslips were then washed with fresh buffer. Freshly prepared substrate solution (0.5 mg/ml napthol AS-MX phosphate [initially dissolved in 25 μM dimethylformamide], 1 mg/ml Fast-Red TR salt, and 4 mM levamisole in 0.01 M Tris buffer, pH 8.2) was filtered directly onto the coverslips. The color reaction was allowed to develop for 10 minutes before immersion in TBS and tap water. The cultures were counterstained using Mayer's hematoxylin for 45 s, the excess stain removed with tap water, and the coverslips mounted in glycerol for microscopic examination.

In situ hybridization

Pericytes were grown on culture slides in 20% FCS:MEM until mineralized nodules had formed (day 19), at which point the cells were fixed for 1 h in 4% (w/v) paraformaldehyde in PBS and then washed in PBS. The prehybridization treatments used were as described previously.20 Briefly, the slides were immersed sequentially in 0.2 M HCl for 20 minutes, 2× SSC (1× SSC = 0.15 M sodium chloride and 0.015 M sodium citrate) for 10 minutes and PBS for 10 minutes. Control slides were treated with 10 mg/ml RNAse A (Boehringer Mannheim, Lewes, East Sussex, U.K.) in 0.5× SSC for 1 h at 37°C and rinsed in PBS. All slides were postfixed for 20 minutes in 0.4% (w/v) paraformaldehyde in PBS and immersed for 10 minutes in freshly prepared 0.25% (v/v) acetic anhydride in 0.1 M triethanolamine (pH 8.0). Slides were prehybridized for 1 h at 37°C in 50% formamide, 1 mg/ml of bovine serum albumin, 0.02% (w/v) Ficoll, 0.02% (w/v) polyvinyl pyrrolidone, 0.6 M NaCl, 0.2 mg/ml of sheared salmon sperm DNA, 10 mM Tris (pH 7.4), 0.5 mM EDTA, 10 mM dithiothreitol (DTT), and 10% (w/v) dextran sulfate. Hybridization with aliquots (50 μl) of heat-denatured [35S]-labeled probe (100 ng/ml in prehybridization mixture) was carried out overnight in prehybridization solution. Hybridization probes were prepared by random oligolabeling with [35S]α-dCTP (Amersham Int., Buckinghamshire, U.K.) to specific radioactivities of ∼1 × 108 cpm/μg of DNA. The cDNA probe used in this study encodes a 100-amino-acid human bone Gla protein precursor (kindly provided by Dr. P.J. Barr, Chiron Corp., Emeryville, CA, U.S.A.).38 After hybridization, the slides were washed with a series of high-stringency washes: twice for 5 minutes in 0.5× SSC containing 1 mM EDTA and 10 mM DTT; twice for 5 minutes in 0.5× SSC with 1 mM NaCl; 15 minutes in 50% formamide, 0.15 M NaCl, 5 mM Tris (pH 7.5), and 0.5 mM EDTA; four times 5 minutes in 0.5× SSC at 55°C; followed by 5 minutes at room temperature in 0.5× SSC. Slides were dehydrated in 70 and 95% ethanol with 0.3 M ammonium acetate and air dried. Autoradiography was performed with Ilford K5 emulsion (Ilford, Mobberly, Cheshire, U.K.), melted at 40°C, and diluted 1:1 with distilled water. The slides were exposed at 4°C for 14 days and developed in Kodak D-19 developer for 5 minutes, rinsed, fixed for 5 minutes, and counterstained with hematoxylin and eosin.

RNA isolation and analysis

RNA was isolated from cultured pericytes at confluence, when nodules had formed, and when these nodules had mineralized as previously described.20,23 In some experiments, RNA was also isolated from homogeneous preparations of nodules and mineralized nodules that were individually removed from the base of the flask using a sterile needle. Using this procedure, it was possible to obtain RNA from nodules and mineralized nodules that was minimally contaminated with RNA from cells in sparse, confluent, and multilayered areas present within these postconfluent cultures. RNA was shown to be intact if ethidium bromide staining revealed discrete 28 S and 18 S ribosomal RNA bands after electrophoresis of denatured RNA through 1% agarose gels containing 2.2 M formaldehyde.

cDNA probes

The following human cDNAs were used in this study: HOP, the 1493 bp insert containing the complete protein-encoding region of human osteopontin plus flanking regions on both the 3′ and 5′ ends35; hon-2, the 2500 bp insert containing the complete protein-encoding region of human osteonectin plus flanking regions on both the 3′ and 5′ ends39; and B6–5 g, the 1165 bp insert containing the complete protein-encoding region of human bone sialoprotein plus flanking regions on both the 3′ and 5′ ends.40 All the cDNAs were kindly provided by Drs. L. Fisher and M. Young (Bone Research Branch, NIDR, NIH, Bethesda, MD, U.S.A.).

Northern blotting

Samples of RNA (15 μg) from confluent cultures, from cultures containing nodules, and from cultures containing mineralized nodules, were separated in 1.2% agarose gels containing 2.2 M formaldehyde, transferred to nylon membranes (HybondN, Amersham Int.) by capillary action, and fixed by baking for 1 h at 80°C. Hybridization probes were prepared by random oligolabeling with [α-32P]dCTP using a Pharmacia (St. Albans, U.K.) kit to specific radioactivities of 1 × 109 cpm/μg of DNA. Filters were prehybridized and hybridized at 65°C in phosphate buffer containing 1% bovine serum albumin and 7% sodium dodecyl sulfate (SDS).20 Filters were washed in 2× SSC, 1% SDS, at 65°C for 15 minutes, and in several washes of 0.2× SSC, 0.1% SDS at 65°C for 1 h. Filters were then exposed to Fuji Medical X-ray film (Genetic Research Instrumentation Ltd., Essex, U.K.) at −70°C. This procedure was repeated using three different sets of pericyte RNA.

Amplification of cDNA

cDNA was amplified as described in detail elsewhere.41 Briefly, cDNA with an average length of 400 bases was generated on the liberated mRNA templates obtained from confluent pericytes and from purified nodules (mineralized and unmineralized) using reverse transcriptase and an oligo(dT)24 primer (Pharmacia). After addition of a homopolymer 3′ (dA) tail with terminal deoxynucleotide transferase, the resulting cDNA was amplified by PCR using a 3′ (dT)24-containing 60-base primer for 40 cycles (94°C for 1 minute, 42°C for 1 minute, 72°C for 2 minutes).41 Additional quantities of amplified cDNA were generated as required by reamplification using the same primer.41 Controls for the polymerase chain reaction (PCR) included the omission of cDNA from the reaction mix. It has previously been shown that the PCR applied to poly(A)+ RNA yields amplified cDNA that is several hundred bases in length and is representative of the extreme 3′ untranslated ends of the original transcripts.42 As a result of the length limitation, the abundance relationships in the original sample are preserved in the final amplified product.42

Southern blotting

Amplified cDNA (5–10 μl/lane) from confluent cells, nodules, and mineralized nodules was electrophoresed in 1.5% agarose in Tris-borate buffer, denatured in NaOH/NaCl, neutralized in Tris-HCl/NaCl, and transferred to HybondN nylon membranes as described.41 Tracks containing control samples for the PCR reaction did not stain with ethidium bromide. Probes labeled with32P were prepared by random priming (Oligolabeling Kit, Pharmacia) from templates that included the extreme 3′ untranslated sequences of the genes of interest. Hybridization and washing were carried out as described (Amersham's instructions).41 Southern blotting was repeated in triplicate with three separate groups of cDNA from each of the different stages of pericyte differentiation.

Implantation study

Diffusion chambers of volume 130 μl were assembled as described previously4 from commercially available components (Millipore U.K. Ltd., Harrow, U.K.). Chambers were placed in a Petri dish and sterilized by ultraviolet light for 1 h. Primary cultured bovine pericytes and freshly harvested rabbit marrow (used as a positive control) were inoculated into separate chambers (104–105 cells/chamber) which were then sealed. The diffusion chambers were implanted intraperitoneally in 8-week-old athymic mice (Harlan, Olac, Bicester, U.K.) under anesthesia. Five animals were each implanted with two diffusion chambers. In each animal, one diffusion chamber contained freshly harvested rabbit marrow4 and the other chamber contained pericytes. Two different preparations of pericytes were used in two separate experiments.

Diffusion chambers, harvested at 28 days (two animals) and 56 days (three animals) after implantation, were fixed in formalin and embedded in methylmethacrylate. Sections (7 μm) were stained with toluidine blue and von Kossa (using standard techniques). These time points encompass the range of incubation times found to be necessary for the formation of an osteogenic tissue by cells obtained from a variety of species.4,9


Pericyte growth and characterization

Pericytes appeared initially as large overlapping stellate cells often clumped together near a microvessel fragment. After reaching confluence, these cells proceeded to retract upon each other, forming nodules which eventually exhibited a brown–black coloration under phase contrast microscopy. These cells have previously been shown to be found within a mineralized matrix.16,17

Vascular pericytes were characterized by staining with antibodies to α-smooth muscle actin, mAb 3G5, and STRO-1 at each stage of their growth and differentiation. α-smooth muscle actin staining was observed in both confluent pericytes and in nodules as bundles of cytoplasmic filaments within the cells (Figs. 1A, 1B, 1C, 1D). The mAb 3G5 stained both confluent cells and nodules in pericyte cultures (Figs. 1E and 1F). The immunoreactivity appeared predominantly on the surface of the cell defining the ragged shape of the pericytes (Fig. 1E). Outlines of cells could also be detected on the surface of the nodule (Fig. 1F).

Figure FIG. 1.

Expression of α-smooth muscle actin and 3G5 antigen during pericyte differentiation in vitro. The presence of α-smooth muscle actin (B, D) and 3G5 antigen (E, F) in cultured pericytes was examined by indirect immunofluorescence. α-smooth muscle actin was present in confluent pericytes (B) and in the cells on the surface of mineralized nodules (D). (A) and (C) show DAPI staining of the fields in (B) and (D), respectively. Antibody 3G5 also stained confluent pericytes (E) and cells in mineralized nodules (F). (G) and (H) show a pericyte nodule stained with DAPI (G) and nonimmune serum (H). Bar = 150 μm.

Confluent pericytes showed no immunoreactivity with vWF (data not shown). Controls of bovine aortic endothelial cells were positive for vWF and were negative for α-smooth muscle actin and 3G5. Smooth muscle cells expressed α-smooth muscle actin but not vWF or 3G5 antigen (data not shown). Fibroblasts showed no immunoreactivity with mAb 3G5. In controls where the first antiserum was substituted by nonimmune serum, fluorescence was not observed (Figs. 1G and 1H).

To determine if pericytes expressed the antigen for STRO-1, a marker of primitive clonogenic marrow stromal cell precursors,25 cultures were fixed at confluence (day 5) and when mineralized nodules were present (day 35). Staining was not observed in sparse or confluent areas of these cultures (Fig. 2A). By comparison, positive staining was clearly apparent in multilayered areas (Fig. 2B) and in cells adjacent to pericyte nodules (Fig. 2C). Immunoreactivity was markedly reduced in nodules and mineralized nodules (Fig. 2C). Controls of IgM antibody directed against the irrelevant antigen CD80 and tissue culture medium alone were negative (Fig. 2D).

Figure FIG. 2.

Stage-specific expression of STRO-1 by cultured pericytes. The STRO-1 antigen was only expressed by cells in multilayered areas (B) and by cells adjacent to pericyte nodules (C). Confluent cells (A) and cells on the surface of mineralized nodules (C) were not immunoreactive. Staining did not occur when an antibody directed against the antigen-presenting cell-specific antigen CD80 was used (D). Bar = 100 μm.

Identification of bone matrix proteins in pericyte culture

RNA was extracted from confluent pericytes, postconfluent cultures containing nodules, and postconfluent cultures containing mineralized nodules. The presence of mRNA for osteopontin, osteonectin, and bone sialoprotein in each of these samples was then determined. Northern blotting studies demonstrated that mRNA for osteopontin (Fig. 3A), osteonectin (Fig. 3B), and bone sialoprotein (Fig. 3C) was present in all the samples analyzed. However, because of the inherent heterogeneity of pericyte cultures at postconfluence (i.e., sparse areas, confluent areas, multilayered areas, nodules, and mineralized nodules are all present within the same culture flask), this stategy did not permit us to determine at which stage of pericyte differentiation each gene was preferentially expressed. To address this issue, RNA was prepared from nodules and mineralized nodules which were individually picked from the base of the flask. The levels of mRNA for osteopontin and osteonectin in these samples was then examined by conventional Southern blot methods after the general amplification of all polyadenylated (poly(A)+) transcripts as described in the Materials and Methods. Previous studies have shown that by limiting the length of the amplified cDNA, the relative abundance of specific transcripts are preserved in the final product.42

Figure FIG. 3.

Expression of osteopontin, osteonectin, and bone sialoprotein mRNA by cultured pericytes. RNA prepared from confluent pericytes (track 1), postconfluent cultures containing nodules (track 2), and postconfluent cultures containing mineralized nodules (track 3) was subjected to Northern analysis as described in the Materials and Methods. (A) Hybridization with cDNA probe for osteopontin (OPN). (B) Hybridization with cDNA probe for osteonectin (ON). (C) Hybridization with cDNA probe for bone sialoprotein (BSP).

Amplified cDNA (Fig. 4) from confluent pericytes (track 1), purified nodules (track 2), and purified mineralized nodules (track 3) was electrophoresed on an agarose gel and subjected to Southern hybridization using probes for osteopontin (Fig. 4A) and osteonectin (Fig. 4B). The ethidium bromide–stained gel (Fig. 4C) shows the amount of cDNA loaded into each track of the agarose gel. These results demonstrate that osteopontin and osteonectin show different and distinct patterns of expression during pericyte differentiation. Osteopontin mRNA was present at low levels in confluent pericytes and was markedly up-regulated in pericyte nodules and when these nodules had mineralized (Fig. 4A, compare tracks 1–3). By comparison, osteonectin mRNA was most prevalent in confluent cultures. The level of osteonectin was markedly reduced in pericyte nodules and mineralized nodules (Fig. 4B, compare tracks 1–3).

Figure FIG. 4.

Expression of osteopontin and osteonectin during pericyte differentiation in vitro. Amplified cDNA samples from confluent pericytes (track 1), purified nodules (track 2), and purified mineralized nodules (track 3) were electrophoresed on an agarose gel and Southern blotted. (A) Hybridization with a radiolabeled osteopontin probe. (B) Hybridization with a radiolabeled osteonectin probe. (C) Ethidium bromide staining shows the amount of cDNA loaded onto each track of the gel.

In situ hybridization was then used to determine whether mRNA for osteocalcin was present in pericyte cultures. Osteocalcin mRNA was expressed by pericytes (Fig. 5A). Expression was restricted to a subset of cells at the periphery of nodules. No signal was detected in cells in sparse and confluent areas or in cells on the surface of the nodules. RNAse-treated controls were negative (Fig. 5B).

Figure FIG. 5.

Osteocalcin mRNA is expressed by a subpopulation of cultured pericytes. In situ hybridization of pericyte cultures using a cDNA probe for osteocalcin demonstrated that osteocalcin was only expressed by a subset of cells at the periphery of pericyte nodules (A). (B) RNAse-treated control. Bar = 200 μm.

Immunolocalization studies using specific antisera to bone sialoprotein, osteopontin, and osteonectin confirmed that all three proteins were expressed by pericytes at each stage of their growth and differentiation, although cells on the surfaces of nodules were stained particularly intensely (representative patterns of distribution are shown in Fig. 6). For each antibody, the intracellular pattern of staining was similar, with strong perinuclear staining in the cytoplasm and a weaker, more granular fluorescence in attachment processes (Figs. 6A and 6C). Controls, in which the first antiserum was substituted with normal rabbit serum, were negative (data not shown).

Figure FIG. 6.

Localization of bone sialoprotein and osteonectin in pericyte cultures. The presence of bone matrix proteins in pericyte cultures was determined by immunofluoresence. Bone sialoprotein (A, B) and osteonectin (C, D) were both expressed by sparse pericytes (A, C) and by cells in mineralized nodules (B, D). The same pattern of staining was observed for osteopontin (data not shown). Bar = 150 μm.

Implantation study

Diffusion chambers were inoculated with primary passaged pericytes or, as a positive control, freshly harvested rabbit marrow. After 28 days, pericytes had proliferated to form a fibrous tissue within the chamber (Fig. 7A). Cells adjacent to the filter had adopted a bipolar, spindle-shaped morphology. These cells were closely apposed to each other and were aligned parallel to the filter. Toward the center of the chamber, the cells had adopted a more spherical morphology and were well separated by matrix (Fig. 7A). There was no evidence of bone or cartilage formation at this stage. After 56 days, however, all of the diffusion chambers inoculated with pericytes contained a well-organized matrix comprising bone, cartilage, and fibrous tissue.

Figure FIG. 7.

Osteogenic differentiation of pericytes in vivo. Diffusion chambers loaded with pericytes were harvested after 28 days (A) and 56 days (B–D). Sections were stained with either toluidine blue (A, C) or toluidine blue and von Kossa (B, D). (A) A fibrous layer (f) was localized adjacent to the filter. (B) Fibrous tissue (f) and a preosteoblast layer (p) were observed aligned parallel to the filter. A layer of mineralized bone (b), which stains positively with von Kossa, was located further into the chamber. (C) In this section, adipocyte-like cells (arrows) and mineralized bone (arrowheads) were located next to the filter. Layers of cartilage (c) and fibrocartilage (fc) were also observed. (D) In some chambers, tissue spanned the gap between the two filters. Calcified cartilage (cc), bone (b), and a preosteoblast layer (p) were observed. Bar = 150 μm.

Different stages of tissue differentiation were noted within the same chamber (Fig. 7B, 7C, 7D). A layer of fibrous tissue was observed adjacent to the filter (Figs. 7B and 7D). In some areas, this fibrous layer had been replaced by bone (Fig. 7C, arrowheads). Further into the chamber, the cells appeared more spherical and were separated by matrix (Figs. 7B, 7C, 7D). The thickness of this “preosteoblast” layer was found to vary both within and between individual chambers (compare Figs. 7B, 7C, 7D). A layer of developing bone, which stained positively with von Kossa, was located adjacent to this fibrous tissue (Figs. 7B and 7D). The “preosteoblast” and bone layers formed by pericytes in this study are highly reminiscent of corresponding layers formed by marrow cell suspensions implanted in diffusion chambers (compare Fig. 7B this paper with Fig. 4d from Ashton et al.4). Mineral was found only in association with matrix, and there was no evidence of dystrophic calcification. Layers of cartilage (Figs. 7C and 7D), calcified cartilage (Fig. 7D), and fibrocartilage (Fig. 7C) were observed toward the center of the chamber. Small clusters of cells that morphologically resembled adipocytes were seen in all the diffusion chambers examined (compare Fig. 7C with previous papers6,43).

Control diffusion chambers inoculated with freshly isolated rabbit marrow contained a mixture of bone, cartilage, and fibrous tissue after 28 and 56 days (data not shown), as previously described.4


This study demonstrates that microvascular pericytes are capable of forming bone, cartilage, and fibrous tissue in a widely accepted model of in vivo osteogenesis—the diffusion chamber assay. In addition, we demonstrate that pericyte differentiation along the osteogenic pathway in vitro is associated with the stage-specific expression of markers of primitive clonogenic marrow stromal cells (STRO-1) and osteoblasts (osteopontin, osteonectin, bone sialoprotein, and osteocalcin). It is not clear under what circumstances pericytes express osteogenic potential in situ. However, the possibility that these cells are of functional importance in the growth, maintenance, and repair of the skeleton and in diseases involving ectopic ossification and calcification now merits serious consideration.

The mAb STRO-1 recognizes a surface antigen on a subset of human marrow stromal cells that includes clonogenic precursors (CFU-F) capable of giving rise to cells of multiple marrow stromal cell lineages, including osteoblasts.25,28 STRO-1 immunoreactivity was not detected in sparse and confluent regions of pericyte cultures, but its expression was detected in multilayered areas and in cells at the periphery of pericyte nodules (Fig. 2). This pattern of expression correlates with that observed previously for ALP.16 In addition, STRO-1 expression appeared to be reduced in cells toward the central region of the nodule, although we cannot exclude the possibility that this pattern of immunoreactivity results from the lack of penetration of the mineralized regions of the nodule by the STRO-1 antibody. At this time, mRNA for osteocalcin is expressed by the pericytes (Fig. 6). It is of interest, therefore, that similar findings have been observed during the differentiation of osteogenic precursors from human bone and marrow.31 Preliminary immunolocalization of the sites of STRO-1 reactivity in adult mammalian bone indicate that this antigen is expressed by microvascular pericytes in situ (S. Walsh, unpublished observations).

The in vitro osteogenic differentiation of pericytes was demonstrated in this study by monitoring the expression of genes that are characteristically expressed at high levels by cells of the osteoblast lineage, namely osteonectin, osteopontin, bone sialoprotein, and osteocalcin.26,27,29,30 Our results show that all four proteins are expressed by cultured pericytes (Figs. 3, 4, 5, 6). Furthermore, we demonstrate that osteonectin, osteopontin, and osteocalcin are expressed preferentially at specific stages of pericyte differentiation in vitro.

Osteonectin mRNA, as determined by polyA-PCR and Southern blotting, was preferentially expressed in confluent pericytes (Fig. 4), which suggests that this protein has a role in the early stages of pericyte growth and differentiation. Interestingly, high levels of osteonectin mRNA are present at sites of chondrogenesis and osteogenesis in mouse and human tissues.39,44 In this location, osteonectin may modulate the assembly and stabilization of fibrillar collagens and/or act as an inhibitor of mineralization by controlling the growth and size of hydroxyapatite crystals.26

Osteopontin, bone sialoprotein, and osteocalcin are secreted later in bone formation and are therefore believed to be involved in the process of remodeling.26,45,46 Osteopontin was detected at all three stages of pericyte differentiation; however, the mRNA for this protein was found to be markedly elevated in the mineralized nodules (Fig. 4). These results are consistent with previous studies which demonstrated that high levels of osteopontin are indicative of a mature osteoblast phenotype.26,29,47 Furthermore, osteopontin is secreted at the onset of mineralization, when it is thought to influence the rate of mineralization.26,47–50

Bone sialoprotein secretion is also associated with the initiation of mineralization,26,27,36,51 where it is thought to be involved in the nucleation of hydroxyapatite at the mineralization front of bone.49,50 Previously, bone sialoprotein has been detected in mature osteoblasts, hypertrophic chondrocytes, placenta, and breast carcinoma.26,29,51 This is the first report of the synthesis of this protein by vascular pericytes. Preliminary studies have shown that bone sialoprotein is also expressed at sites of calcification in atherosclerotic arteries (C.F. Farrington, I. Roberts, H. Nield, and A.E. Canfield, unpublished observations), where pericyte-like cells have also been identified.18,19,52,53 Therefore, the role of bone sialoprotein in the osteogenic differentiation of pericytes and in the calcification of arteries is currently under investigation.

In situ hybridization studies demonstrated that osteocalcin mRNA was expressed by cultured pericytes but that its expression was restricted to a population of cells at the periphery of mineralized nodules (Fig. 5). These results provide further evidence that pericytes differentiate into osteoblasts.29,30,54 Interestingly, recent studies using osteocalcin-deficient mice have suggested that although osteocalcin normally functions to limit bone formation, the absence of this protein had no effect on bone mineralization.55

The true osteogenic potential of vascular pericytes was demonstrated in this study using diffusion chambers (Fig. 7). This assay is probably the best available test of the osteogenicity of a given cell population, since diffusion chambers present an enclosed environment in which the implanted cells can proliferate and differentiate but from which host cells are excluded.7 Most importantly, cells from nonskeletal sites (spleen and thymus) consistently fail to form bone or cartilage in this assay4,6,10,56 unless an inducing agent such as decalcified bone matrix is also present.10 Furthermore, Villaneuva and Nimni57 have shown that endothelial cells derived from the rat liver or from bovine aorta do not form bone when implanted in diffusion chambers.

Pericytes implanted in diffusion chambers initially proliferated to form a layer of fibrous tissue adjacent to the filter (Fig. 7). Subsequently, a mixture of bone and cartilage were formed on the surface of the fibrous tissue (Fig. 7). Electron microscopic analysis of similar tissue formed by marrow cells implanted in diffusion chambers has demonstrated that this tissue resembles cartilage and bone both ultrastructurally and morphologically.5 The distribution of osteogenic tissue within the chambers also followed a distinctive pattern; bone was formed adjacent to the fibrous tissue and cartilage was located toward the center of the chamber. The temporal and spatial pattern of osteogenesis observed in this study mimics that observed following the implantation of freshly isolated marrow, cultured marrow cells, or bone-derived cells in diffusion chambers in vivo.3–5,9

An additional and important finding of this study is that microvascular pericyte cultures contain cells that appear to be capable of differentiating into mature cells of bone, cartilage, adipose, and fibrous connective tissues in vivo. In this regard, preliminary studies in this laboratory have demonstrated that, under defined conditions, adipogenic differentiation of vascular pericytes in vitro can also occur (H. Nield and A.E. Canfield, unpublished observations). These results suggest that the pericyte population in vivo may be a source of mesenchymal stem cells, or alternatively, that a pericyte is not a fully differentiated cell. These results differ from the studies of Decker et al.15 who recently proposed the pericyte as a “committed progenitor cell” of the osteogenic lineage. It must, however, be emphasized that our results do not permit a distinction to be made between the presence in the cultures of multiple, tissue-specific precursors or a single multipotential precursor.

In summary, these results demonstrate that pericytes have osteogenic potential in vitro and in vivo. These findings are consistent with the possibility that pericytes play a fundamental role in normal bone physiology and/or in its repair following injury. In addition, recent studies suggest that the calcification of arteries, a common complication of atherosclerosis, is mediated, at least in part, by pericyte-like cells present in these vessels.18,19 The environmental and molecular cues that trigger the osteogenic differentiation of pericytes are currently unknown and warrant further investigation.


We thank Dr. J. Hoyland for advice on in situ hybridization and Drs. L. Fisher and M. Young for kindly providing antibodies and cDNA probes used in these studies. The financial support of the Arthritis & Rheumatism Council and the Wellcome Trust is gratefully acknowledged.