Osteocytic Expression of Constitutive NO Synthase Isoforms in the Femoral Neck Cortex: A Case-Control Study of Intracapsular Hip Fracture


  • The authors have no conflict of interest.


NO is an osteocytic signaling molecule that can inhibit osteoclasts. The NO synthases eNOS and nNOS were expressed by >50% of osteonal osteocytes in controls. Hip fracture cases showed +NOS osteocytes only in deep osteonal bone, and 25–35% reduced expression overall. These data are consistent with increased osteonal vulnerability to deep osteoclastic attack.

Introduction: Osteocytes may regulate the response to mechanical stimuli in bone through the production of local signaling molecules such as NO derived from the NO synthase eNOS. Because NO is inhibitory to osteoclastic resorption, it has been suggested that osteocytes expressing eNOS act as sentinels, confining resorption within single osteons. Recently, nNOS has been shown to be present in osteocytes of adult human bone.

Materials and Methods: Cross-sections of the femoral neck (eight female cases of intracapsular hip fracture and seven postmortem controls; age, 68–91 years) were analyzed by immunohistochemistry. The percentages of osteocytes expressing each of these two isoforms were calculated, and their distances to the nearest canal surface were measured.

Results: The percentage of +nNOS osteocytes was lower in the fracture cases than in the controls (cases: 43.12 ± 1.49, controls: 56.68 ± 1.45; p < 0.0001). Compared with nNOS, eNOS expression was further reduced (p = 0.009) in the cases but was not different in the controls (cases: 36.41 ± 1.53, controls: 56.47 ± 2.41; p < 0.0001). The minimum distance of +eNOS or +nNOS osteocytes to a canal surface was higher in the cases compared with controls (eNOS: controls; 44.4 ± 2.2 μm, cases: 61.7 ± 2.0 μm; p < 0.0001; nNOS: controls: 52.4 ± 1.7 μm, cases: 60.2 ± 2.1 μm; p = 0.0039). +eNOS osteocytes were closer to the canal surfaces than +nNOS osteocytes in the controls by 8.00 ± 4.0 μm (p = 0.0012).

Conclusion: The proportions of osteocytes expressing nNOS and eNOS were both reduced in the fracture cases, suggesting that the capacity to generate NO might be reduced. Furthermore, the reduction in NOS expression occurs in those osteocytes closest to the canal surface, suggesting that the ability of NO to minimize resorption depth might be impaired. Further studies are needed on the regulation of the expression and activity of these distinct NOS isoforms.


BY ADJUSTING THE remodeling process in response to mechanical strain, bone has the potential to protect itself against fracture. With aging and after menopause, the remodeling process becomes unbalanced, leading to excessive resorption and/or decreased formation, resulting in bone loss. Indeed, in the femoral neck cortex, where osteons undergoing resorption fuse,(1) there is an excessive number of resultant giant canals in the inferior region in osteoporotic hip fracture cases,(1) which increase porosity of the cortex. Moreover, an examination into the effect of disuse in animal models showed a rather similar pattern,(2) suggesting that one cause of the formation of giant canals is a decrease in physical activity.

It has been suggested that osteocytes may control the response to mechanical stimuli through the production of local signaling molecules such as NO. NO can modulate bone remodeling(3–5) and reverse bone loss caused by estrogen deficiency.(6, 7) It has also been reported that nitrate treatment for angina was associated with a reduction of hip fracture risk.(8) In a previous case-control study, we assessed the expression of the NO synthase eNOS in osteocytes from the femoral neck cortex. We showed not only that the expression of osteocytes expressing eNOS was lower in the femoral neck cortex of fracture cases compared with postmortem controls, but also that these were located further from the nearest canal surface in the fracture cases.(9) Because NO has been shown to inhibit osteoclastic resorption, it was suggested that osteocytes expressing eNOS might act as sentinels, confining resorption within single osteons,(9) and hence act to restrict the formation of these giant canals.

Recently, we have found that immunoreactivity for the neuronal isoform of NOS is also evident in adult human bone osteocytes.(10) The aim of this study was to determine, in a new series of subjects, whether the expression of nNOS in osteocytes was also reduced in cases of hip fracture and whether its distribution in cases and controls was similar to that of eNOS.


Subjects and biopsy preparation

Whole cross-section femoral neck biopsies were taken at hemiarthroplasty immediately distal to the fracture from eight female hip fracture cases (age, 72–91 years) and chilled.(11) Control biopsies were obtained postmortem from seven women (age, 68–89 years) with no history of disease known to affect bone metabolism. Written informed consent to participate in the study, as required by the district ethics committee, was obtained from all live subjects. Samples were sectioned (10 μm) in a whole body PMV cryomicrotome (Leica, Milton Keynes, UK) and transferred onto precooled microscope slides coated with a thin layer of pressure-sensitive glue(11) before storage at −80°C.


Immunohistochemistry was done as previously described.(10) Immunolabeling for each isoform was done in separate serial sections. After air drying, the sections were incubated in 0.3% hydrogen peroxide and in methanol to inhibit their endogenous peroxidase activity. Sections were subsequently washed in PBS and blocked in normal serum (Vector Laboratories Universal Kit; Vector Laboratories, Burlingame, CA, USA) for 1 h. Sections were incubated overnight at 4°C in 2 μg/ml rabbit polyclonal antibody raised against the carboxy terminus of eNOS of human origin (SC-654; Santa Cruz Biotechnology) or in 0.5 μg/ml mouse monoclonal antibody raised against the carboxy terminus of nNOS of human origin (N31020; Transduction Laboratories). Sections were washed three times in PBS for 5 minutes and incubated with a biotinylated antimouse/rabbit secondary antiserum (Vector Laboratories Universal Kit) for 30 minutes. After washing as above, the sections were incubated with a ABC reagent (PK-6200, Vector laboratories Universal kit) for 1 h. Immunoperoxidase activity was visualized by reaction in VIP substrate (SK-4600; Vector Laboratories), which gives a purple precipitate. Manufacturer's specifications stated that these antibodies do not cross-react with other NOS isoforms.

To determine osteocyte density, after immunostaining, the sections were incubated in methyl green (Vector Laboratories) for 1 minute at 37°C, dehydrated, and mounted in Clearmount media. The total number of osteocytes in a given field was determined by the sum of +NOS osteocytes and those stained with methyl green. The percentage of osteocytes positive for a particular NOS isoform is the number of positive cells divided by the total number of osteocytes.

Image analysis

For the quantification of the density of osteocytes exhibiting positive immunolabeling for the eNOS and nNOS isoforms, sections were examined using a Leica DMRB microscope, and TIFF images from three fields of cortical bone (0.1-1.1 mm2) from each biopsy were captured using an automated montage system, Montage Explorer v2 (Synoptics, Cambridge, UK). Osteocytes positive for eNOS (+eNOS) and nNOS (+nNOS), Haversian canals, and bone surfaces were marked on the image using PaintShop Pro v 7.00 (Jasc Software, Banbury, UK). The images were thresholded and analyzed automatically for the number and location of each +eNOS osteocyte and +nNOS osteocyte and the location and area of each Haversian canal using NIH Image v1.62 (Fig. 1). The distance between osteocytes and the nearest canal surface was measured as previously described.(9)

Figure Fig. 1..

(A) Representative image of nNOS staining in the inferior region of the biopsy from a control subject. (B) Osteocytes are marked on the image before the image is thresholded (red dot, +nNOS osteocyte; green dot, +nNOS osteocyte). Bar = 100 μm.

For osteonal bone analysis, osteocytes that were nearer to the edge of the field or an endosteal or periosteal surface than they were to a canal surface were censored (∼43% nNOS, 4806/11,099; 49% eNOS, 4505/9240). Osteocytes >150 μm from the canal surface were excluded on the basis that previous studies(1) had shown that >99% of the mean wall width in the Haversian canals in the elderly femoral neck cortex were below this thickness (∼9% nNOS, 582/6293; 7% eNOS, 315/4735). Haversian canals with less than five osteocytes per basic multicellular unit (BMU) were excluded from the analysis to allow comparison between osteocytes expressing a NOS isoform and osteocytes not expressing either isoform. This represented 21% and 24% of the osteocyte population in the field analysis for nNOS and eNOS, respectively.

In adult human bone, osteocyte lacunae are 20 μm long in the plane facing the bone surface (approximately the same plane as our cross-sections) and 10 μm wide, but only have a mean thickness of around 4 μm.(12) The osteocyte itself rarely fills its lacuna. Therefore, small errors will occur in the proportions of positive and negative osteocytes because the section is around 2.5 times the thickness of the typical osteocyte.


Differences between cases and controls in the distances of NOS isoforms from the nearest canal surface were analyzed using a linear mixed regression model in JMP statistical software, v3.1.6 (SAS Institute, Cary, NC, USA). The proportions of +eNOS and +nNOS staining cells were assumed to follow binomial distributions. Contrast analysis to test the regional differences between cases and controls in the percentages of osteocytes expressing the two NOS isoforms was carried out using Stata v8.2 statistical software (StataCorp).


Expression of eNOS and nNOS in cortical osteocytes

All bone:

Both cases and controls had a greater percentage of +nNOS osteocytes than +eNOS osteocytes (+15%, p = 0.017; Fig. 2), and this effect did not differ between the four regions of the femoral neck (p = 0.38, data not shown).

Figure Fig. 2..

Proportion of +NOS osteocytes (eNOS, stippled bars; nNOS, white bars) in the adult human femoral neck in cases of intracapsular hip fracture and postmortem controls. Data are shown as the mean ± SE.

Osteonal bone:

At BMU level, the percentage of osteocytes immunolabeled for nNOS was significantly lower in the fracture cases compared with the controls (cases: 43.13 ± 1.49, controls: 56.68 ± 1.45; p < 0.0001). The percentage of osteocytes immunolabeled for eNOS was also lower in the cases compared with controls (cases: 36.41 ± 1.53, controls: 56.47 ± 2.41; p < 0.0001).

The percentage of +nNOS osteocytes in fracture cases was, however, higher compared with the percentage of +eNOS osteocytes (d = 7.85, p = 0.0092). In contrast, there was no significant difference in the expression of these two NOS isoforms in control samples (Fig. 3).

Figure Fig. 3..

Proportion of +NOS osteocytes in the individual BMUs (osteonal bone) of the femoral neck. ENOS, stippled bars; nNOS, white bars. Data shown as the mean ± SE.

Regional analysis of the femoral neck showed that the proportion of osteocytes expressing nNOS was higher in the inferior region of the postmortem controls compared with the fracture cases (p < 0.0001; Table 1). Overall, the inferior and superior regions had a higher proportion of +nNOS osteocytes than the anterior and posterior regions (p = 0.0001; Table 1). In contrast, the proportion of osteocytes expressing eNOS was higher in the anterior and superior regions of the postmortem controls than in the fracture cases (p < 0.0001; Table 2).

Table Table 1. Proportions of +nNOS Staining Cells by Region in Cases and Control (Binomial Model)
original image
Table Table 2. Proportions of +eNOS Staining Cells by Region in Cases and Controls (Binomial Model)
original image

Distance of +NOS osteocytes from canal surfaces

The minimum distance from a +nNOS osteocyte to the canal surface was greater than that for an osteocyte lacking nNOS expression. Furthermore, the minimum distance from a +nNOS osteocyte to the nearest canal surface was greater in the cases than that for a +nNOS osteocyte in the control (cases: 60.23 ± 1.92, controls: 52.44 ± 1.87; p = 0.0039). The maximum distance from a +nNOS osteocyte to the nearest canal surface was not significantly different between cases and controls.

Similarly, the minimum distance from a +eNOS osteocyte to the nearest canal surface was also greater in the cases than in the controls (cases: 61.74 ± 1.86, controls: 44.42 ± 2.93; p < 0.0001). Unlike nNOS distribution, the maximum distance from a +eNOS osteocyte to the nearest canal surface was also greater in the cases than in the controls (cases: 100.31 ± 2.34, controls: 90.64 ± 3.68; p = 0.0275). Comparison between the distribution of +eNOS and +nNOS osteocytes showed that +eNOS osteocytes were also located nearer to the canal surface than +nNOS osteocytes (d = −12.2, p < 0.0001; Fig. 4).

Figure Fig. 4..

The minimum distance of a +NOS osteocyte from the nearest canal surface. Data are shown as the mean ± SE.

To address whether such changes in the relative number and position of NOS expression osteocytes exhibited differences around the circumference of the femoral neck, we examined changes in the defined regions. Our regional analysis showed that the minimum distance from the nearest canal surface of a +nNOS osteocyte was greater in all regions in the fracture cases compared with that in the controls (Fig. 5A). Similarly, the minimum distance from canal surfaces to the nearest +eNOS osteocyte was greater in anterior, inferior, and superior, but not posterior, regions in the fracture cases compared with controls (Fig. 5B)

Figure Fig. 5..

The minimum distance of (A) a +nNOS osteocyte and of (B) a +eNOS osteocyte from the nearest canal surface by region. Data are shown as mean ± SE.


To examine further the possibility that osteocytes regulate bone remodeling through the generation of NO in response to mechanical loading,(5) we studied the patterns of expression of eNOS and nNOS by osteocytes in the femoral neck. Osteocytic iNOS was not found in the material we examined. This may be because it is expressed transiently, only in response to a resumption of loading after a period of disuse as in rodents.(13) The study of Watanuki et al.(13) raised the interesting possibility that iNOS might be expressed to reinforce the loading response after a period of disuse when resumed loading becomes mismatched to residual bone strength. However, to clarify this possibility in humans would require an interventional design.

This study was performed in a distinct subset of our library of biopsies collected from fracture cases and postmortem controls and confirmed our previous findings that the proportion of osteocytes expressing eNOS was indeed decreased in the femoral neck of hip fracture cases compared with controls.(9) This study also confirms that the osteocytes expressing eNOS in the fracture cases were positioned further from the nearest canal surface than those osteocytes expressing eNOS in the postmortem controls. However, in this study, the effect was larger in the anterior and superior regions of the femoral neck than in the inferior region. The proportion of osteocytes expressing nNOS was also decreased in the fracture cases, with a similar distribution to those expressing eNOS. The expression and distribution of both eNOS and nNOS in the superior and inferior regions was reduced in the fracture cases.

The role of nNOS in bone is not clear. Although nNOS mRNA has been detected using RT-PCR in cultured rat osteoblasts and chick osteocytes(14) and cells from human fracture callus,(15) the majority of investigations have failed to find evidence of nNOS expression in young rodent or fetal human osteocytes or osteoblasts.(16–19) This discrepancy can not be explained on the basis of distinct antibody recognition patterns, because the antibodies used here and in previous studies are the same. It does remain possible, however, that nNOS is only expressed in osteocytes within adult bone tissue. This study is unique in using undecalcified frozen tissue and not decalcified paraffin embedded tissue, and such methodological difference in the preparations of the tissue is an alternative explanation.

We(9, 10) and others(14) have previously suggested that one role of osteocytic NO is the inhibition of osteoclastic bone resorption. In vitro studies on NO production by isolated embryonic chick osteocytes have shown that the amount of NO release in response to fluid shear stress or tensile mechanical strain is around 1–20 pM/h/cell.(20, 21) However, inhibition of osteoclast formation and activity in vitro requires NO concentrations of around 1–30 μM,(3, 22) and so the question as to whether osteocytes produce enough NO to inhibit osteoclast activity in vivo must be addressed. It may simply be that in vitro studies are relatively insensitive in determining both the rate of production and the responsiveness of cells. For example, the concentration of PTH required to stimulate bone resorption in vitro is several-fold higher than the circulating level in patients with hyperparathyroidism.(23, 24)

It has been shown that NO is a paracrine mediator, because its rate of aqueous diffusion is similar to that of oxygen, allowing it to reach cells at some distance from its origin. Lancaster modeled(25) a distribution of NO-generating cells, spaced at intervals (like osteocytes) of about 50 μm. In the absence of a nearby blood vessel, the average cell was exposed to about three times as much NO as it would have been exposed to solely through its own power of NO generation. Clustering of the NO producing cells doubled the NO concentration. In our study, osteocytes expressing either nNOS or eNOS were not evenly distributed (Fig. 1), so the possibility that certain areas of bone accumulate higher levels of NO must be considered. The direct applicability of Lancaster's results to bone tissue depends on whether the diffusion of NO through bone mineral is similar to its diffusion through soft tissues and whether mineralized bone has any scavenging properties; neither of these is known. If NO perfuses mineralized bone less well than soft tissues, NO might become relatively concentrated within osteocyte canaliculi.

Furthermore, it may be that the action of NO on osteoclasts is indirect, or in part, reflects the actions of other gene products. In other cells types, NO donors prevent hypoxia-induced accumulation of hypoxia-inducible factor-1α (HIF-1α),(26) and several HIF-1α target genes have the potential to induce cell apoptosis. Thus, NO production in osteocytes might stabilize osteocyte HIF-1α production and prevent apoptosis; as a consequence, bone resorption would be limited.(27) It has also been postulated that eNOS production of NO in stromal cells modulates bone resorption by regulating the expression of RANKL and osteoprotegerin (OPG).(28) If osteocytes behave in a similar manner to stromal cells, NO produced by osteocytes could modulate bone resorption by regulating RANKL and OPG expression.

In conclusion, this study confirms the presence of nNOS in osteocytes in human femoral neck cortex.(10) As with eNOS, osteocytes expressing nNOS are located toward the periphery of cortical BMUs (i.e., the cells expressing nNOS are the osteocytes formed in the early phase of bone formation). If NO production by either NOS isoform is sufficient to prevent osteoclastic resorption, NOS-expressing osteocytes may act to prevent excessively deep resorption. The reduction in the expression of both NOS isoforms in cases of hip fracture suggests that Haversian canals might more readily merge with each other to form the large pores that are a marked feature of both hip fracture bone(1) and disuse.(29) Whereas the mechanism by which NO exerts its action on remodeling requires further investigation, it is intriguing that nitrate donors prevent clinical bone loss.(30) Therefore, the role of NO generation by osteocytes is a potentially important component of pathways conserving bone mass in the human femur, justifying further investigation of both the physical and cellular mechanisms regulating NO concentrations beneath bone surfaces vulnerable to attack by osteoclasts.


This work was supported by MRC Programme Grant 9321536.