Increased presence of capillaries next to remodeling sites in adult human cancellous bone



Vascularization is a prerequisite for osteogenesis in a number of situations, including bone development, fracture healing, and cortical bone remodeling. It is unknown whether a similar link exists between cancellous bone remodeling and vascularization. Here, we show an association between remodeling sites, capillaries, proliferative cells, and putative osteoblast progenitors. Iliac crest biopsies from normal human individuals were subjected to histomorphometry and immunohistochemistry to identify the respective positions of bone remodeling sites, CD34-positive capillaries, smooth muscle actin (SMA)-positive putative osteoblast progenitors, including pericytes, Ki67-positive proliferative cells, and bone remodeling compartment (BRC) canopies. The BRC canopy is a recently described structure separating remodeling sites from the bone marrow, consisting of CD56-positive osteoblasts at an early differentiation stage. We found that bone remodeling sites were associated with a significantly increased presence of capillaries, putative osteoblast progenitors, and proliferative cells in a region within 50 µm of the bone or the canopy surface. The increases were the highest above eroded surfaces and at the level of the light-microscopically assessed contact of these three entities with the bone or canopy surfaces. Between 51 and 100 µm, their densities leveled to that found above quiescent surfaces. Electron microscopy asserted the close proximity between BRC canopies and capillaries lined by pericytes. Furthermore, the BRC canopy cells were found to express SMA. These ordered distributions support the existence of an osteogenic-vascular interface in adult human cancellous bone. The organization of this interface fits the current knowledge on the mode of action of vasculature on osteogenesis, and points to the BRC canopy as a central player in this mechanism. We propose a model where initiation of bone remodeling coincides with the induction of proximity of the vasculature to endosteal surfaces, thereby allowing capillary-BRC canopy interactions that activate marrow events, including recruitment of osteoblast progenitors to bone remodeling sites. © 2013 American Society for Bone and Mineral Research.


Bone remodeling is essential for maintaining healthy bones throughout adult life. It is a process where bone-forming osteoblasts (OBs) replace the bone matrix, which has been resorbed by osteoclasts (OCs). A failure in bone matrix restitution leads to osteoporosis and increased risk of fractures.

The clinical importance of bone formation has stimulated a lot of research aimed at understanding its mechanism. Much knowledge has been gained in recent years, especially in relation to the signaling pathways controlling OB differentiation.1 Furthermore, physiological systems such as the nervous and the vascular system have been found to be critical players. Regarding the latter, it has been stressed that vascularization is a prerequisite for bone formation in a number of situations, including bone development,2–4 fracture healing,3, 5 and distraction osteogenesis.3, 5, 6 Moreover, it was recently shown that the molecular interactions involved at the osteogenic-angiogenic interface can be manipulated to promote bone formation.3, 7

Vascularization may also play a key role in bone formation during bone remodeling. In line with this view, cortical bone remodeling sites are invariably associated with one or more capillaries.8–10 However, a systematic anatomic association has not been established in regard to cancellous bone remodeling. Still, the bone marrow is known to be rich in different types of microvessels, namely sinusoids and capillaries.8, 11–13 A possible role of these microvessels in cancellous bone remodeling is suggested by their decreased density found in disease situations such as osteoporosis.14 This role is also supported by the positive response of bone formation rate to blood flow.15, 16 Furthermore, studies on a rat model showed that bone formation rates correlate with capillary density in the bone marrow and that parathyroid hormone (PTH) induces redistribution of blood vessels closer to bone formation sites.17 It is worth noting that so-called paratrabecular sinusoids have also been mentioned in relation to human cancellous bone remodeling.14, 18 These structures are, however, likely to correspond to the bone remodeling compartments (BRCs) indentified by Hauge and colleagues.19 Hauge and colleagues demonstrated that the nature of the BRC is not vascular, because they found that the BRC canopy cells are of osteoblast lineage, not of endothelial lineage.19 Instead, Andersen and colleagues detected a dense network of capillaries at the bone marrow side of BRC canopies.20 It can be speculated that this dense capillary network at cancellous bone remodeling sites favors local OB generation through differentiation of OB progenitors recruited through several different mechanisms. Possible mechanisms include transportation of circulating osteoprogenitor cells,21 paracrine interactions between endothelial cells and OB-lineage BRC canopy cells,19, 22 and paracrine interactions between endothelial cells and bone marrow stromal cells (BMSCs) including the pericytes, which belong to the vascular wall itself.23–26 Furthermore, the capillaries serve as a conduit supplying nutrients and oxygen.13 Interestingly, oxygen tension has been related to bone matrix synthesis as well as increased proliferation of cells in the hematopoietic stem cell (HSC) niche.15, 27–29

An important question that still needs clarification is whether this dense capillary network is a unique property of bone remodeling sites, or whether the capillaries are evenly distributed along all cancellous bone surfaces. The current study addresses this issue by comparing the presence of capillaries above bone remodeling surfaces and quiescent surfaces, both close to and away from the surface. Furthermore, to assess a possible relation between capillary prevalence and OB recruitment, the capillary distribution is compared with the distribution of putative OB progenitors and cells in the proliferative status.

Materials and Methods

Patients and biopsies

Nine paraffin-embedded iliac crest biopsies with no apparent pathology in either bone or marrow were included for histomorphometry and immunohistochemistry. The biopsies were obtained at the Department of Haematology, Vejle Hospital, from patients under investigation for a hematological malignancy (Danish Ethical Committee approval, journal no. 20010082). Five of the patients were women, and four were men; their mean age was 61 (26 to 83) years. The biopsies were fixed in 4% neutral-buffered formaldehyde for at least 20 hours at room temperature immediately after obtention, decalcified in 10% buffered formic acid for 7 hours at room temperature, dehydrated in graded series of ethanol, and embedded in paraffin.

Three EPON-embedded transiliac bone biopsies from patients with primary hyperparathyroidism (PHPT) were included for electron microscopy (EM), because the increased activation frequency of bone remodeling in PHPT renders the detection of bone remodeling sites at the EM level more easily. The biopsies were obtained at the Department of Surgery, Aarhus University Hospital, immediately after surgical treatment for PHPT (Danish Ethical Committee approval, journal no. S-20070121).

Histochemistry and immunohistochemistry

Four 3.5-µm-thick adjacent sections were processed with the purpose of staining the first with Ki67, the second with a modified Masson trichrome staining,20 the third with CD34 class II and CD56, and the fourth with smooth muscle actin (SMA) and CD34 class II. This design is useful for analyzing the distribution of these stainings relative to quiescent surfaces, eroded surfaces, osteoid surfaces, and BRC canopies.

Before immunohistochemical staining, the sections were treated through the following steps: deparaffinization, blocking of endogenous peroxidase activity, rehydration, epitope retrieval, and blocking of unspecific adhesion. Then the sections were incubated with one of the following mouse antibodies: IgG1 against either Ki67 (clone MIB-1, DAKO, Glostrup, Denmark) or CD56 (clone 56C04, Lab Vision, Kalamazoo, MI, USA), or IgG2a against SMA (clone 1A4, DAKO). The IgG1 antibodies were detected with a horseradish-peroxidase polymer conjugated to goat anti-mouse IgG (Envision, DAKO). The IgG2a antibody was detected with a horseradish peroxidase-conjugated IgGa subtype-specific antibody (Jackson ImmunoResearch, West Grove, PA, USA). All primary antibodies were visualized with 3.3'-Diaminobenzidine (DAB +, DAKO). After visualization with DAB + , the Ki67 stained sections were counterstained with Mayer's hematoxylin and mounted. The CD56 and the SMA stained sections were blocked with mouse serum, and incubated with a fluorescein isothiocyanate-labeled (FITC) mouse anti-CD34 class II antibody (clone QBend10, DAKO), which was detected through alkaline-phosphatase-conjugated FAB anti-flourescein isothiocyanate (Roche, Basel, Switzerland), and visualized by liquid permanent red (LPR, DAKO). The sections were counterstained and mounted after LPR visualization. Negative controls were performed by omitting the addition of primary antibody. Positive controls were performed on four types of tissue: tonsil, kidney, liver, and appendix. CD34 and SMA stained all four types of tissue, CD56 stained tonsil and kidney tissue, and Ki67 stained tonsil and appendix tissue. All four tissues were negative for all four markers when the primary antibody was omitted (Nordic Immunohistochemical Quality Control,

Electron microscopy

The bone biopsies were transversely sliced into 1-mm-thick pieces and immediately fixed in 3% glutaraldehyde in 0.1 M cocadylate buffer (pH 7.4). The slices were then cleaved into pieces of 1 to 2 mm3, post-fixed in 1% OsO4 in 0.1 M cocadylate buffer (pH 7.4), dehydrated, cleared in propylene oxide, and embedded in Epon 812. Semi-thin sections (1 µm) were sectioned from the embedded pieces with a diamond knife on a Leica Ultracut UCT, stained with 1% toluidine blue, and examined for interesting areas. Finally, ultra-thin sections (70 nm) were sectioned from the regions of interest and stained with saturated aqueous uranyl acetate and lead citrate.

Microscopy and image analysis

Light microscopic analysis was performed with an upright DMRXAZ Leica microscope (Leica, Wetzler, Germany) with x2.5 to x100 objectives. Images were obtained using a DC500 CCD camera controlled by the IM500 (version 1.2) software (Leica Systems). Electron microscopic analysis was performed on a transmission electron microscope (Philips TEM 208, Eindhoven, The Netherlands), operated at 80 kV. The figures were assembled using the CorelDraw package (x4) (Fremont, CA, USA).


Pictures of the Masson trichrome-stained sections were obtained at x2.5 magnification and printed. The classical 2D perimeter measurements quiescent surface, eroded surface, and osteoid surface, and the extent of BRC canopy coverage, were assessed through light microscopy and marked on the print. All markings were verified by a second observer. The immunohistochemically stained sections were scanned with a MIRAX scanner (Zeiss, North Chesterfield, VA, USA), and countings were performed on a computer with a Merz grid attached to the screen using the MIRAX viewer software at x30 magnification. Orthogonal lines were drawn from the grid up to 100 µm into the bone marrow (Fig. 1). Four analyses were performed. In each, we quantified distinct hits along the orthogonal lines: In the first it was a CD56-stained BRC canopy cell, in the second a CD34-stained endothelial cell, in the third a smooth muscle actin (SMA)-stained cell, and in the fourth a Ki67-stained cell. The data are expressed as number of hits per orthogonal lines. All four analyses comprised several thousand randomized orthogonal lines. Discrimination between capillaries and sinusoids was based on differences in size, morphology, and staining intensity with the CD34 antibody. The capillaries were slender compared with the larger than 50-µm-wide, sac-like appearing sinusoids.30 Moreover, within the same section, the capillaries stained in general more intensively compared with the sinusoids.

Figure 1.

Illustration of the method used to quantify capillary hits next to the different types of bone surfaces. (A) The drawing shows the three types of bone surfaces: the BRC canopy (green) covering the surfaces undergoing remodeling, and the capillaries (red) in the neighboring bone marrow (blue). The semicircular line of the Merz grid intersects the bone surface at different points, and orthogonal lines are drawn at each intersection point, as shown. When the line hits a capillary, the respective distance at which this occurs is measured, taking as starting point the bone surface or the BRC canopy when it is present. These distances are indicated with double arrows, whereas the light-microscopically assessed contacts between capillaries and canopies are indicated with circles. The number of hits of these lines with capillaries, as well as the total number of orthogonal lines, are recorded. The data are expressed as number of hits per orthogonal lines. Note that the same orthogonal line can hit a capillary more than once. Note also that this approach is designed to quantify the presence of capillaries—not just the capillary number—thus taking also the effect of capillary orientation into account. The counting of SMA+ and Ki67+ cells has been done using the same approach. (B) A representative double-stained section showing CD34+ capillaries and CD56+ OB-lineage cells. The Merz grid and the orthogonal lines are indicated as in (A). The distances from the bone surface or the BRC canopy to the individual capillaries are indicated by purple lines and the light-microscopically assessed contacts by purple circles. Scale bar = 25 µm.

Statistical analysis

For each patient, the hits per orthogonal lines were grouped according to surface and extent of BRC canopy coverage. The quiescent surfaces, eroded surfaces, and osteoid surfaces from each patient were viewed as being related. Normality testing was done on between-group values, and an appropriate statistical test was performed, depending on whether data were parametric or nonparametric, the Repeated Measures Analysis of Variance (ANOVA) and the Friedman test, respectively. The results of a statistically significant post-test are shown in each graph with a letter ap < 0.05, bp < 0.01, cp < 0.001. The use of post-test depended on data parametricity; for parametric data Bonferroni's multiple comparison test was chosen, and for nonparametric, Dunns's multiple comparison test. All statistical analyses were performed with GraphPad Prism 5. In addition, a clustered logistic regression on the capillary data set was performed in STATA. The references were quiescent surface without a BRC canopy, and the exposure variables were eroded surface, osteoid surface, and presence of a BRC canopy. The regression analysis was repeated either two times (distances of 0 to 50 and 51 to 100 µm measured from the surface or the BRC canopy) or five times (contacts with the surface or the BRC canopy [ie, distance 0], and distances of 1 to 25, 26 to 50, 51 to 75, and 76 to 100 µm). Because of these repetitions, the significance levels were changed according to the Bonferroni correction.


Distribution of capillaries nearby bone surfaces and their relation to BRC canopies

The proportion of quiescent, eroded, and osteoid surface to total bone surface showed typical values for normal individuals (Fig. 2A). Consistent with previous findings,19, 20, 31 the BRC canopy covered a large proportion of the remodeling bone surface. In the current cohort, most of the eroded, half of the osteoid, and a few percent of the quiescent surface was covered with a canopy (Fig. 2B). This distribution pattern was similar for each of the patients, and accordingly, the effect of the type of bone surface on canopy coverage was statistically highly significant. Furthermore, the BRC canopies were positive for CD56 as previously reported.20 They shared this characteristic with the so-called reversal cells on the eroded surfaces but not with mature bone forming osteoblasts, which were weakly or not at all stained (Fig. 3A, D).

Figure 2.

Distribution of the three different types of bone surface (A) and extent of BRC canopy coverage over these surfaces (B). Each dot represents the proportion of orthogonal lines hitting a given bone surface (A) or a BRC canopy above a given surface in a given biopsy (B). The dots corresponding to the analysis of quiescent surfaces, eroded surfaces, and osteoid surfaces from the same biopsy are connected with a line. The horizontal bars show the mean of the nine biopsies. In B, the Repeated Measures ANOVA test shows a significant effect of the type of surface, p < 0.0001. Post-test for individual comparisons cp< 0.001.

Figure 3.

Relation between type of bone surface and prevalence of neighboring capillaries. (A) Representative double-stained sections showing CD34+ capillaries and CD56+ OB-lineage cells. (A, top) Frequent capillaries (arrowheads) are positioned next to a canopy-covered remodeling surface. The capillaries adjacent to the BRC canopy (large arrowheads) run parallel to it, thereby offering a large interface between the two entities. (A, bottom left) Enlargement of the framed area in the upper picture illustrating contact between a BRC canopy (arrows) and a capillary (arrowhead). (A, bottom right) A capillary (large arrowhead) runs perpendicularly to the quiescent bone surface, thereby offering only a small interface with the surface. (B, C) Prevalence of capillary hits over quiescent, eroded, and osteoid surfaces, at the level of their contact with the bone or the BRC canopy surface (B), at distances up to 50 µm into the bone marrow (C, left), and between 51 and 100 µm (C, right). The data are presented as in Fig. 2. (D) Discrimination between capillaries and sinusoids. The picture illustrates a difference in position and morphology between capillaries and sinusoids. The dashed black line shows the distance 100 µm away from the surface. S indicates a sinusoid. The predominant microvascular structure until 100 µm into the bone marrow is the capillary. The table shows the patient-specific number of capillary and noncapillary CD34 hits (mostly sinusoids) between 0 and 100 µm. Scale bars = 25 µm. N = 9. The horizontal bars in panel B show the median and in panel C the mean. Overall comparisons were analyzed by using the Friedman test (B, p < 0.0001) and the Repeated Measures ANOVA (C, left, p < 0.0001; C, right, p = 0.94). Post-tests for individual comparisons: bp < 0.01, cp < 0.001.

Our previous findings have shown numerous contacts between the BRC canopies and capillaries at the light microscopy level.20 In the current study, we investigated whether the presence of capillaries differs between quiescent surfaces and remodeling sites. Fig. 3A illustrates that capillaries can be detected both above remodeling sites and quiescent surfaces. A difference between these two regions, however, is that capillaries run parallel to BRC canopies, in line with our previous 3-dimensional assessments,20 whereas they run rather orthogonally to quiescent surfaces, which is an orientation that does not favor as much capillary-surface interaction. Accordingly, the quantifications of Fig. 3B show that the prevalence of the light-microscopically assessed contacts was about threefold higher over eroded surfaces and about twofold higher over osteoid surfaces compared with quiescent surfaces. When extending the hits from contact with the surface to 50 µm into the bone marrow, a distribution similar to that of contacts alone was obtained, with a significantly higher prevalence of hits over eroded and osteoid surfaces compared with quiescent surfaces (Fig. 3C, left). In contrast, the prevalence of capillary hits from 51 to 100 µm over eroded surfaces and osteoid surfaces equaled the level over quiescent surfaces (Fig. 3C, right). Furthermore, a clustered logistic regression was used to analyze the probability of capillary presence next to remodeling sites in greater detail. The distance was either split into two categories (0 to 50 and 51 to 100) or five categories (contacts, 1 to 25, 26 to 50, 51 to 75, and 76 to 100 µm), and we took into account the presence or absence of BRC canopy (as quantified in Fig. 2), as well as the proximity to erosion or osteoid. Restricting the analysis to two distance categories (not shown) gave a result similar to the one performed with the Repeated Measures ANOVA (Fig. 3C). When the regression analysis included all five distance categories, the probability of capillary presence in each of the first three categories (ie, up to 50 µm) was higher when the surface was covered with a BRC canopy and also higher above eroded surface compared with osteoid (Table 1). Furthermore, the highest level of capillary presence was reached in the case of light-microscopically assessed contact between a capillary and a BRC canopy-covered eroded surface.

Table 1. Clustered Logistic Regression for Analyzing the Increased Presence of Capillaries in Different Regions Next to Remodeling Sites
  • The regression analysis was performed on the different regions shown in the table. The indicated values represent odds ratios relative to the level next to the quiescent surface. The level of statistical significance after Bonferroni correction is shown. ns = not significant.

  • a

    p < 0.0002.

  • b

    p < 0.01.

1–251.22 (ns)1.79a1.10 (ns)1.62a
26–501.34 (ns)1.54b1.07 (ns)1.23 (ns)
51–751.09 (ns)0.97 (ns)1.08 (ns)0.96 (ns)
76–1001.12 (ns)1.29 (ns)1.01 (ns)1.16 (ns)

Because light microscopy suggests that contacts between capillaries and canopies are so important, the proximity of capillaries and canopies was analyzed at higher resolution. Electron microscopy (EM) reveals that capillaries and canopies are physically separated structures, although they show very close proximity (Fig. 4). This close proximity provides the basis for cell trafficking from the capillaries to the BRC (Fig. 4A, B) as would be required for cell recruitment onto the bone surface.

Figure 4.

Electron microscopic analysis of the bone marrow bone matrix interface in areas where capillaries are close to BRC canopies. (A) A marrow cell (MC) performs diapedesis (asterisks) through a BRC canopy (arrows) into the lumen of a BRC (BRC). A capillary (arrowheads) is situated close by. (B) Close proximity between an endothelial cell membrane (arrowheads) and a BRC canopy cell membrane (arrows) covering the BRC. E indicates an erythrocyte. (C) Close connection between a capillary (L, lumen), a BRC canopy (arrows), and an osteoclast (OC). The capillary comprises an endothelial cell (EC) and pericytes (P). (D) Enlargement of the framed area in C: Note the close apposition of the pericyte (P) and the BRC canopy (arrows). Scale bars = 5 µm (A, D), 1 µm (B), 10 µm (C).

It is also important to note that capillaries were the predominant type of microvessel within 100 µm of the surface in the present cohort (Fig. 3D). Only 2% of all microvessel hits were not capillaries (mostly sinuses). Farther away from the surface, the sinusoids became more numerous (not quantified).

Distribution of SMA-positive cells

A capillary consists of a layer of endothelial cells incompletely enveloped by pericytes. Interestingly, pericytes have osteogenic potential and share with OB-lineage cells expression of the contractile protein SMA.23–25, 32, 33 Therefore, the distribution of SMA was investigated. Expression of SMA was detected in the pericytes around the capillaries, in the osteoblastic BRC canopy cells, and in the osteoblastic reversal cells on the eroded surfaces but not in the mature cubic OBs (Fig. 5A, D). This SMA staining pattern is quite similar to that of CD56, except that CD56 was not detected in pericytes (Fig. 3A, D). The quantification of SMA-positive hits in the marrow showed a pattern of distribution similar to that of the capillaries, in accordance with the position of pericytes along capillaries. The prevalence of light-microscopically assessed surface contacts showed the largest values with about threefold and twofold more SMA-positive hits over eroded surfaces and osteoid surfaces, respectively, compared with quiescent surfaces (Fig. 5B). When extending the quantification of the hits up to 50 µm into the bone marrow, there were about twofold more SMA-positive hits over both eroded surfaces and osteoid surfaces compared with quiescent surfaces. In contrast, between 51 and 100 µm, the prevalence over eroded surfaces and osteoid surfaces leveled with that over quiescent surfaces. When analyzing the capillary-canopy contact areas at higher resolution by using EM, it becomes apparent that capillaries may indeed bring pericytes against BRC canopies (Fig. 4C). Furthermore, the frequency of SMA-positive hits within the BRC canopy cells themselves showed the highest frequency of SMA expression over eroded surfaces and the lowest over quiescent surfaces (Fig. 5D).

Figure 5.

Relation between type of bone surface and prevalence of neighboring SMA+ cells. (A) Representative double-stained sections showing SMA+ cells in the marrow, in the BRC canopies, on eroded surface, as well as CD34+ capillaries. The cuboidal OBs are negative for SMA, whereas the BRC canopy (arrows) is positive. The capillaries are partly surrounded by SMA+ pericytes (see framed area enlarged at the right). (B, C) Frequency of SMA+ cells over quiescent, eroded, and osteoid surfaces, at the level of their contact with the bone or the BRC canopy surface (B), at distances up to 50 µm into the bone marrow (C, left), and between 51 and 100 µm (C, right). The data are presented as in Fig. 2. (D) SMA expression in canopies. (D, left) Histological appearance of SMA expression in the BRC canopy (arrows) and in reversal cells lining eroded surfaces. Note the SMA staining in canopies is continuous over eroded surface (upper picture) and only partial over osteoid surfaces (lower picture). OCs (OC; upper picture) and OBs (lower picture) are negative. (D, right) Proportion of SMA-positive canopy cells over the different types of bone surfaces. The data are presented as in Fig. 2. Scale bars = 25 µm. N = 9 (B, C), N = 8 (D). The horizontal bars in panel B show the mean and in panels C and D the median. Overall comparisons were analyzed by using the Repeated Measures ANOVA (B, p = 0.0059) and the Friedman test (C, left, p < 0.0190; C, right, p = 0.9712; D, p < 0.0001). Post-tests for individual comparisons: bp < 0.01, cp < 0.001.

Distribution of Ki67-positive cells

A possible mechanism for providing sufficient bone-forming OBs is cell proliferation. Proliferative cells can be detected by using Ki67, a marker widely used in clinical assessment of cancer.34 Fig. 6A shows Ki67-positive cells in perivascular areas located above BRC canopies covering either an eroded surface (left) or an osteoid surface (right). It also shows a Ki67-positive cell in light-microscopically assessed contact with the BRC canopy (Fig. 6A, right). The number of Ki67-positive cells at the level of surface contact showed about a twofold higher prevalence over remodeling surfaces compared with quiescent surfaces (Fig. 6B). When extending the quantification of the Ki67-positive cells up to 50 µm into the bone marrow, the prevalence was approximately 1.5-fold higher over remodeling surfaces compared with quiescent surfaces, whereas the prevalence of Ki67-positive cells over eroded surfaces and osteoid surfaces from 51 to 100 µm leveled with quiescent surfaces (Fig. 6C). The prevalence of Ki67-positive cells, SMA-positive cells, and capillary hits thus all show a similar bone surface event-related distribution. Worth noting, however, the number of Ki67-positive cells was lower close to the surfaces compared with farther away (respective graphs of Fig. 6B and Fig. 6C), whereas the prevalence of capillary hits and SMA-positive cells was higher close to the surfaces compared with farther away (Fig. 3B, C and Fig. 5B, C).

Figure 6.

Relation between type of bone surface and frequency of neighboring proliferative cells. (A) Two examples showing Ki67+ proliferative cells above surfaces undergoing remodeling. Note capillaries (arrowheads) and BRC canopies (arrows). (A, right) A Ki67+ cell in light-microscopically assessed contact with a BRC canopy (arrowhead). (B, C) Frequency of Ki67+ cells over quiescent, eroded, and osteoid surfaces, at the level of contact with the bone or the BRC canopy surface (B), at distances up to 50 µm into the bone marrow (C, left), and between 51 and 100 µm (C, right). The data are presented as in Fig. 2. Scale bars = 25 µm. N = 9. The horizontal bars in panel B and panel C, left, show the median and in panel C, right, the mean. Overall comparisons were analyzed by using the Friedman test (B, p = 0.0307; C, left, p = 0.006) and the Repeated Measures ANOVA (C, right, p = 0.2462). Post-tests for individual comparisons: bp < 0.01.


Limited data exist on the involvement of vasculature in cancellous bone remodeling,17, 20, 35, 36 whereas the link between vasculature and cortical bone remodeling,8–10, 37 bone development,2–4 fracture healing,3, 5 and distraction osteogenesis3, 5, 6 is well established. To our knowledge, the current study is the first that demonstrates an increased presence of capillaries next to human cancellous bone remodeling sites. Furthermore, it shows that the prevalence of SMA-positive cells and proliferative cells in the bone marrow parallel the distribution of capillaries along the bone surfaces (Fig. 7).

Figure 7.

Model illustrating the link between activation of the bone remodeling cycle and activation of bone marrow events. Bone remodeling sites coincide with a higher prevalence of several events up to a distance of 50 µm into the bone marrow, such as presence of capillaries (CD34), cell proliferation (Ki67), and abundance of SMA+ cells (see text).

Interestingly, we find that the prevalence of capillary hits is linked to bone remodeling events only within 50 µm from the bone surface or the BRC canopy surface. Bourke and colleagues and Watchman and colleagues already stressed a higher vessel prevalence in a 50-µm-wide band along bone surfaces in three different types of human cancellous bone; however, they did not investigate vessel distribution in relation to bone surface events.38, 39 In a recent study on rats, Prisby and colleagues found an average distance of 67 µm between capillaries and osteoid surfaces in untreated animals, and redistribution of capillaries closer to osteoid (43 µm) when the animals were treated with parathyroid hormone (PTH).17 Compared with the findings of Prisby, Bourke, and Watchman,17, 38, 39 our quantifications further characterize the distribution of capillaries by indicating a systematic relation between bone remodeling events and prevalence of capillary hits. Interestingly, this distinct prevalence appears as soon as the remodeling cycle starts, because it is already seen above eroded surfaces, where it is also the highest. Our evidence is based on CD34 staining, which marks not only vascular endothelial cells but also hematopoietic stem and progenitor cells (HSPCs).40, 41 However, because the HSPCs only account for 1.5% of all bone marrow mononuclear cells41 and are small,38 they supposedly comprise a limited number of hits compared with the numerous, often elongated capillary profiles.

Regarding bone marrow vascularization, the emphasis in the literature has generally been on sinusoids, rather than on capillaries. Therefore, it should be especially stressed that the sinusoidal hits of the current study comprise only 2% of all vessel hits within a distance of 100 µm out in the bone marrow and that our quantifications concern capillaries only. This is in line with recently published studies showing that capillaries are most abundantly situated closest to cancellous bone.17, 36 Appreciating the differences between capillaries and vascular sinusoids is important because the dissimilarity in morphology implies distinct physiological functions. The capillaries are described as relatively thick-walled compared with the tenuous sinusoids.11 Moreover, the sinusoidal endothelium has a discontinuous basement membrane allowing maturing hematopoietic cells to pass through the endothelial wall,12 whereas the endothelium in the capillaries in the cortical osteons, and supposedly also in the capillaries in cancellous bone, is continuous.8 Overall, the capillaries seem to be mainly involved in producing regulatory factors42 and in supplying oxygen and nutrients to the bone,13 whereas the sinusoids seem to be mainly involved in the maturation and transendothelial migration of bone marrow hematopoietic cells to the systemic circulation.11, 43 Further discrimination between capillaries and sinusoids could be gained through the characterization of their respective extracellular matrix.

Regarding paratrabecular vasculature, it is also worth mentioning that in the past the BRC has been interpreted as a sinusoid.14 Later, Hauge and colleagues and Andersen and colleagues showed that the BRC canopy expresses osteoblast lineage and not endothelial lineage markers,19, 20 thereby demonstrating that the BRC is not a sinusoid. However, Andersen and colleagues showed that the BRC is instead covered by a network of capillaries, as particularly well stressed through 3D reconstructions.20 By the use of EM, we further show that the capillary and the BRC canopy are clearly separated structures that are not in direct physical continuity with each other. Still, our EM and light-microscopy quantifications stress the close proximity between capillaries and canopies, which is in line with the earlier finding of a fast communication between the vasculature and the BRC20 and which allows cell recruitment from the capillaries to the bone surface. In this respect, it is of interest to note a tangential orientation of capillaries along BRC canopies, observed both in the current study and by Andersen and colleagues.20 This tangential orientation favors a large communication surface between capillaries and canopies and is in contrast to a rather perpendicular direction of capillaries toward quiescent surfaces. This difference in capillary orientation is likely to be an important reason for the higher prevalence of capillary hits over remodeling sites compared with quiescent surface. The mechanism inducing changes in orientation is unknown. A possibility is that they result from mechanical forces related to the generation of BRCs. Pure redistribution of capillaries closer toward bone remodeling sites was also noted in response to PTH in a rat study.17 However, in many other situations including cortical bone remodeling, fracture healing, and bone development, the bone formation-associated increase in capillary density is dependent on angiogenesis.2–4, 7, 42 It remains to be investigated whether angiogenesis and capillary number may also to some extent contribute to increased capillary prevalence in cancellous bone remodeling.

In the context of a role of capillaries in bone remodeling, it is of interest that marrow Ki67- and SMA-positive cells show the same bone event-related distribution pattern as the capillaries. They always show a higher prevalence over remodeling surfaces compared with quiescent surfaces, despite a high variation of their expression levels amongst the patients. Just as for the capillaries, the distinct prevalence of these entities is especially marked at the level of their contact with the canopy surface, thereby pointing to a possible implication in bone surface events.

Ki67 is widely used as a cell proliferation marker because it is present at all phases of the proliferation cycle and absent from resting cells.34 The increased number of Ki67 hits around capillaries fits earlier reports showing that the perivascular space is a site with increased proliferation.28, 44, 45 Here, we show in addition that the sites of increased presence of capillaries and proliferative cells coincide with the areas of the bone marrow next to bone remodeling sites, ie, precisely where generation of OC and OB precursors is required. However, the current study does not identify the nature of the cells undergoing proliferation.

SMA is known to be expressed in smooth muscle cells, pericytes, and OB-lineage cells.23–25, 32, 33, 46, 47 Our immunostainings not only fit these localizations but also stress that SMA expression shows a similar bone event-related distribution pattern at three different levels. Firstly, this distribution is seen at the level of the bone marrow, where, as stated above, SMA follows the distribution of capillaries, which is exactly in accordance with its presence in pericytes and the localization of the pericytes around capillaries. Pericytes are known for their mesenchymal cell nature,23, 25, 48 including a capacity to proliferate,47 and an ability to differentiate into bone-forming OBs.23–25, 46, 47 Secondly, a similar bone event-related distribution of SMA-positive cells is found in the BRC canopy itself. Thus the highest SMA expression was always over eroded surfaces, despite big differences between the patients in the cohort. Interestingly, canopy cells have been speculated to be a reservoir of OB progenitors.19, 22, 37 In support of this view, we found that canopy cells express CD56 (NCAM) strongly, a factor that is transiently expressed during OB differentiation and is at its highest levels in the most immature OBs.33, 49 Thirdly, SMA is also found on the bone surfaces, at especially high levels in reversal cells. These cells are undergoing differentiation into mature bone-forming OBs, and accordingly, we found that they also express quite strongly CD56, whereas both CD56 and SMA expression is absent or very weak in mature OBs. Taking these observations together, one may thus envision an OB recruitment route, where capillaries would deliver pericytes to BRC canopies, which in turn would deliver them to remodeling surfaces. This situation is reminiscent of OB recruitment during cortical bone remodeling where proliferating OB progenitors associated with a capillary wall are delivered to reversal surfaces.10 SMA may support the migratory function of OB progenitors because it is a contractile protein involved in cell motility.47 This view is supported by our immunostainings that show a weakening of SMA expression at later stages when the cells settle as mature bone matrix secreting OBs or as bone lining cells. Thus, the capillary wall may contain a pool of OB progenitors that can be recruited to any given bone site when the capillary and the bone are in close proximity.

The association of vasculature with both proliferation and differentiation of cells in the bone marrow has already been repeatedly reported,28, 44, 45 but the current study highlights in addition that this association is found next to bone remodeling sites and close to canopies. The physical proximity to bone remodeling sites may explain earlier observations supporting a role of vasculature in cancellous bone maintenance, as for instance the relation between decreased skeletal blood flow and decreased BMD,15 decreased bone apposition rate,16 and gain of bone fat content.15 It also provides a histological basis for emerging data, supporting the association between atherosclerosis and osteoporosis.13, 50 Furthermore, this physical proximity fits the findings that oxygen pressure is critical for bone formation27 and that paracrine factors are the mediators of the interactions occurring between endothelial cells and OB-lineage cells in a series of other situations of osteogenesis, such as bone development, fracture healing, and others.42 These factors include growth factors, cytokines, chemokines, endothelins, prostanoids, angiotensins, nitric oxide, acidosis, and hypoxia. It remains to be identified which signals may be critical for capillary-bone communication during remodeling of adult human cancellous bone and whether capillary-related failures may compromise bone remodeling and contribute to bone diseases, such as osteoporosis. One may also have to take into account possible influences of other cells such as OCs and osteomac-like monocytic cells,51 which are present, respectively, at the bone side and at the bone marrow side of the BRC canopy (unpublished data).

A link between bone remodeling and hematopoiesis has previously been indicated by the involvement of OCs in the release of HSPCs.52 Furthermore, recent data suggest the existence of a joint periendosteal-perivascular hematopoietic stem cell (HSC) niche rather than two separate niches, associated with osteoblasts and vasculature, respectively, as had been proposed previously.38, 44, 45 In this respect, it is interesting that our data demonstrate the existence of regions of close proximity of vasculature to endosteal surfaces undergoing remodeling. We also find these regions to be associated with dividing and progenitor cells, and one may speculate that they could correspond with activated HSC niches.28, 44, 45 High oxygen tension is believed to induce the hematopoietic stem cells to proliferate, whereas hypoxia is believed to maintain them in a slow cycling state.28 Accordingly, the slow cycling, hypoxia-dependent HSC niche would be positioned at quiescent surfaces, whereas proliferation would be increased at the capillary-rich areas positioned next to bone remodeling sites. This implies that the activity of HSC niches would be determined by the occurrence of bone remodeling and that the position of active HSC niches would change according to the site of bone remodeling. In this respect, it is of interest to note the expression of CD56 in the BRC canopies as CD56 has been shown to support hematopoiesis.53 CD56 expression in the BRC canopy could thus play a role in the regulation of neighboring HSC niches. Furthermore, we speculate that the BRC canopy originates from the mesenchymal stromal cells surrounding the bone marrow space and lining the endosteal bone surfaces.37, 54

In conclusion, this study extends the physical association of vasculature with osteogenesis to human cancellous bone remodeling and identifies the osteoblastic BRC canopy as a possible structure mediating capillary-bone interaction. The regions of convergence between capillaries and canopies coincide with a higher prevalence of proliferative cells and of cells positive for SMA, which may reflect pericytes and canopy cells differentiating into OBs. Our observations thus highlight a link between activation of bone remodeling on the cancellous bone surfaces and activation of neighboring bone marrow events (Fig. 7). These observations lead to a model where initiation of bone remodeling coincides with the induction of proximity of the vasculature to endosteal surfaces, thereby allowing capillary-canopy interactions that support recruitment of OB progenitors to bone remodeling sites (Fig. 7). This model should be further evaluated on other cohorts, including larger patient cohorts showing deficient bone formation, such as in osteoporosis.


All authors state that they have no conflicts of interest.


We thank Birgit MacDonald, Kaja Rau Laursen, and Karin Trampedach for their excellent technical assistance, Kent Soee and Morten Moeller Nielsen for their editorial assistance, Birthe Oestergaard for her help in obtaining patient data, and Jens Randel Nyengaard for his advice concerning stereology.

This study was supported by the Danish southern region research grant no. 09/5367.

Authors' roles: The study was designed by HBK, JMD, and TLA. The study was conducted by HBK assisted by TLA, NM, and LR. The data were analyzed and interpreted by HBK, JMD, and TLA. The manuscript was drafted by HBK and JMD. The manuscript was revised by HBK, JMD, and TLA, where TLA mainly took part in revising the figures. The manuscript was approved in its final version by all authors. HBK takes responsibility for the integrity of the data analysis.

Note Added in Proof

At the time of publication, we became aware that we should have used a cycloid grid instead of a Merz grid, since iliac crest biopsies have a vertical section plane. We have therefore performed a new counting of the number of capillary contacts in six biopsies from the cohort, and compared these countings with those obtained previously with the Merz grid. On quiescent surface, the estimated mean amount of capillary contacts is 8.5% using a cycloid grid versus 7% using a Merz grid, on eroded surface 19% versus 18%, and on osteoid surface 14% versus 13%. In conclusion the estimated differences are both small and evenly distributed and therefore the overall message of the paper is maintained.