Glucocorticoids (GCs) are used widely in the treatment of various diseases, such as osteoarthritis, chronic and acute inflammation, and cancer, as well as in the case of organ transplantation, because they are very effective immunosuppressive drugs. However, high-dose GCs also have strong side effects, including bone loss. This effect on bone for a long time has been ascribed mainly to impairment of osteoblasts (OBs), the cells responsible for bone formation. Accordingly, the strong negative effects of GCs on bone-formation rates and OB survival are well documented,1, 2 and the latter effect also was extended to osteocytes.3 However, lack of bone formation and function of OB lineage cells cannot be responsible alone for some of the most important effects that GCs have on bone. These include rapid increase in fracture risk within the first 3 to 6 months of treatment,2, 4 which could be explained best by an effect of GCs on bone resorption itself. In addition, bone-resorption markers increase immediately in patients receiving intravenous GCs, and the maximum increase in serum cross-linked C-telopeptide (CTX) was already reached 24 hours after the very first infusion.5, 6 In bone explants the levels of specific markers of osteoclasts (OCs), the bone-resorbing cells, also increase significantly already after 3 hours of exposure to GCs.7 These observations clearly indicate an immediate effect of GCs on OCs. The consensus now is that GCs induce increased OC bone resorption at least at the beginning of treatment, in addition to prolonged OB impairment.2, 8
There is, however, confusing information in the literature about the mechanism of action of GCs on OC bone resorption, as assessed in vitro. Increased OC survival is stressed in several studies,9–11 but according to other studies, OC survival is decreased by GCs or not affected.12–14 Enhanced OC differentiation also has been proposed,13, 15, 16 but others found that OC differentiation was decreased by GCs or not affected.10, 17, 18 One would expect that besides survival and differentiation, the resorptive activity of existing OCs is also stimulated by GCs because of the very early increase in bone-resorption markers induced by GCs in patients.5, 6 However, we are not aware of any mechanistic OC culture study supporting this view. Instead, primary OCs isolated from rat bones eroded the surfaces of bone slices less extensively in the presence of high-dose GC.12, 14 So also did OCs generated from mouse bone marrow cells cultured in the presence of macrophage colony-stimulating factor (M-CSF) and receptor-activator of NF-κB ligand (RANKL),10 whereas human OCs generated from peripheral blood mononuclear cells in the presence of the same cytokines and seeded on bone slices did not respond to high-dose GCs.13, 15 This absence of direct stimulation of the resorptive activity of OCs in vitro would fit an indirect mechanism of action of GCs on OCs. A series of indirect mechanisms have been proposed,2, 8 including elevation in parathyroid hormone (PTH) levels resulting from GC-induced impaired calcium uptake in the gut and the kidney, GC-induced reduction of gonadal hormones, and mediation through OB lineage cells.7, 17 The latter was proposed because GCs affect the balance between RANKL and osteoprotegerin (OPG), two OB lineage cell products, respectively, that activate and inhibit OC differentiation/activity. On the other hand, studies based on transgenic mice, where GCs become inactivated specifically in OCs, elegantly show that GCs are able to affect OCs directly.9, 10
Thus in vitro experiments failed so far to explain the early bone fragilization and early increase in bone-resorption markers in GC-treated patients through a direct effect of GCs on OC bone resorption. In order to investigate this hypothesis further, we generated human OCs from purified blood CD14+ cells, seeded them on bone slices in the presence and absence of high-dose prednisolone similar to the one used in patients, and evaluated bone resorption, not limiting ourselves to the extent of bone-resorption areas as in earlier studies but also paying attention to the morphology of the resorption features. We found that GCs do not change total resorbed area, in accordance with earlier reports, but deeply affect shape, depth, and collagen leftovers of the resorption events, which have the appearance of deep trenches. Our study therefore indicates a strong influence of GCs on the way OCs exert their resorptive activity, thereby providing a mechanistic explanation for the observations in patients and also throwing light on how OC resorptive activity is controlled.
Materials and Methods
In vitro generation of human OCs
Human OC precursors were isolated from blood donations of healthy female volunteers (approved by the local ethics committee, 2007-0019), as described previously.19 In brief, CD14+ cells were isolated by positive selection using magnetic separation (R&D Systems, Abingdon, UK) after a Ficoll gradient (Amersham Biosciences GE Healthcare, Broendby, Denmark) separation. Cells were seeded in cell culture flasks (Greiner, Frickenhausen, Denmark) in α-MEM (Invitrogen, Taastrup, Denmark) containing 10% FCS (Cambrex/Invitrogen, Taastrup, Denmark) and 25 ng/mL of M-CSF (R&D Systems) for 2 days at 5% CO2 and 37°C in a humidified incubator. Subsequently, the medium was changed, and the cells were differentiated into OCs through the addition of 25 ng/mL of M-CSF and RANKL (R&D Systems) for the following 7 days with medium renewal twice. The matured OCs were reseeded on bone disks (IDS Nordic, Herlev, Denmark) in well plates for the indicated time in the presence of M-CSF and RANKL. About 20% of the bone surface was covered by OCs identified as multinucleated TRACP+ cells.
OCs were reseeded on bone disks and allowed to settle for 2 hours in the presence of M-CSF and RANKL. Subsequently, prednisolone (Sigma-Aldrich, St. Louis, MO, USA) was added to a concentration of 1.6 µM and OCs were incubated for the indicated time. According to the pharmacokinetics of orally taken prednisolone, this corresponds to the active concentration in the sera of patients receiving an oral daily dose of 41 mg of prednisolone (DAK tablets, Product Resumé, Danish Drug Agency). This theoretical estimate is supported by an earlier report measuring prednisolone serum concentrations in patients following oral medication.20 This dose was chosen because it is pharmacologic and relevant for the treatment of many medical conditions. To controls the same volume of solvent, DMSO (Sigma-Aldrich) was added to a final concentration of 0.1%. The resorption features (ie, cavitations and superficial demineralization patches) were stained with toluidine blue as described previously21 and analyzed microscopically. For determining the number of resorption events, a resorption feature with a continuous perimeter at the surface was counted as one. Maximal depth was calculated from a systematic random count of 100 to 150 resorption events per bone slice and was determined according to focus depth by using a microcator (VRZ 401, Heidenhain, Traunereut, Germany) fitted to an Olympus BX50 microscope (Olympus, Ballerup, Denmark). The resorbed bone surface was determined by analyzing the entire bone surface using a 36-point counting grid placed in the ocular and a ×10 objective of a Leica DM RXA 2(Leica, Ballerup, Denmark) or a Zeiss Axiovert 200 (Zeiss/Broch, Michelsen, Birkeroed, Denmark) microscope. The total resorbed bone (Zeiss) Broch, Michelsen, Birkeroed, Denmark) surface was presented as a percentage of the total bone area. Where indicated, the resorbed surface was subdivided: (1) Round excavations were termed pits, and (2) elongated excavations appearing as continuous grooves were termed trenches. The latter were at least twice as long as wide. The number of resorption events was determined by counting pits and trenches within a visual frame for the whole bone surface. The numbers were determined as the average number of events per square and subsequently as percentage of the control.
Metabolic activity was determined by using CellTiter-Blue (Promega, Nacka, Sweden) according to the instructions given by the supplier. The activity of tartrate-resistant acid phosphatase (TRACP), an OC activity marker, and the levels of C-terminal cross-linked telopeptide of type I collagen (CTX), a collagen degradation marker,22 were determined in the conditioned mediium as described previously19 and according to the instructions of the supplier (IDS, Herlev, Denmark), respectively.
Removal of organic matrix from bone slices and resorption lacunae
After careful analyses of bone resorption parameters, the same bone slices were treated with NaOCl to remove demineralized organic matrix. The procedure was as follows: Bone slices were covered in chloroform/isoamyl alcohol (24:1) for 5 minutes at room temperature, dried on paper, incubated for 15 minutes at room temperature while shaking gently in twofold diluted NaOCl stock (10% to 15% active chloride in stock; Sigma-Aldrich), dried on paper, stained with 0.2% toluidin as described previously,13, 21 and analyzed for resorption depth as aforementioned.
Scanning electron microscopic images
Bone slices were incubated with OCs for 72 hours as described earlier. They were washed in PBS and transferred to ddH2O. The OCs were gently removed with a cotton stick. The slices were washed in ddH2O and air dried. Subsequently, the slices were prepared for SEM and covered in gold according to a protocol described elsewhere.23
GCs act directly on OCs and alter the morphology of resorption lacunae
It is highly disputed whether GCs directly influence OC resorptive activity or not. To address this issue, we investigated whether GC affects OC resorption directly in a standard 3-day culture pit assay. At low resolution, toluidine staining did not show any remarkable differences in the extent of resorbed surface, in agreement with an earlier report that also used human OCs,13 but at higher magnification we found a change in the shape of the resorption pits (Fig. 1A). Most of the resorption events under control conditions appeared as discrete round pits, sometimes close to each other (see examples with orange marking), whereas those generated in the presence of GC tended to be more elongated, appearing as continuous grooves, from now on termed trenches (see examples with green marking).
Therefore, we thoroughly analyzed OC bone resorption in the presence of GC as compared with control conditions. GC treatment induced approximately a 30% decrease in pit numbers (Fig. 1B) and in release of TRACP into the conditioned medium (Fig. 1C), an enzyme considered to be a marker of OC activity. In contrast, GC exposure did not affect the extent of total resorbed surface (Fig. 1D, total bar). However, there appeared to be a qualitative difference in the resorbed surface because the presence of GC led to a larger proportion of resorption trenches (Fig. 1D, hatched subbar) compared with pits (non-hatched subbar). Actually, we found that about 45% of the resorbed surface consisted of resorption trenches compared with only 15% under control conditions. Furthermore, GCs induced a highly reproducible and significant increase in the maximum depth of resorption (Fig. 1E). As shown in Fig. 1F, there was a general shift in the frequency curve of resorption depths: More than half the pits were between 0.5 to 10 µm deep in controls, whereas more than half the pits were 5.5 to 30 µm deep under GC conditions. This shows that OCs are uniformly stimulated by GC to resorb deeper. GCs also induced a threefold increase in the levels of CTX in the conditioned medium, suggesting higher OC collagenolytic activity (Fig. 1G). Metabolic activity was unaffected when OCs were seeded on bone in the presence of 1.6 µM prednisolone but was negatively affected when seeded on plastic, indicating cell death (data not shown), which has been described previously.12, 14 Taken together, these observations strongly suggest that GCs directly affect OC resorptive behavior. Although the total number of resorptive events and TRACP activity after 72 hours was negatively affected by GCs, they were deeper and appeared more frequently as trenches, and OC collagenolytic activity was found to be elevated more.
Time dependency of GC-induced alterations of OC resorption
Since a 3-day culture in the presence of GCs leads to deeper and larger resorption lacunae (trenches) compared with controls, we investigated how soon these effects appear. After 24 hours of exposure to GC, resorption pits appeared very similar to control conditions, although after 72 hours of exposure their morphology changed dramatically to more trenches (green markings) than pits (orange markings; Fig. 2A). A quantitative analysis of controls showed that the number of resorption events more than doubled from 24 to 72 hours. Noteworthy, whereas a 72-hour exposure to GC decreased the number of resorptive events by one-third, there was almost no decrease at the 24-hour time point (Fig. 2B), suggesting that GCs do not greatly affect the proportion of active OCs at the beginning of the culture and thus probably do not interfere with the activation mechanism of the OCs. Similarly, while the surface of each resorptive event was increased significantly by GCs at 72 hours (trenches), there was no significant increase at 24 hours (Fig. 2D) compared with control. Thus indicates that the size of the bone area attacked by the OCs is unaffected by GCs when it starts excavating bone and that it transforms the resorption pit into a trench only at a later stage. In contrast, the GC-induced increase in resorption depth is the same at both time points (Fig. 2E), showing that GCs make the OCs resorb deeper from the beginning of the culture. Furthermore, GC did not affect the total resorbed area at either time point despite a 2.5-fold change in total resorbed area at 24 hours compared with 72 hours (Fig. 2C). Taking these observations together, it appears that GCs induce increased excavation depths before inducing trenches and that the latter features are likely to result from the extension of excavations initiated at the beginning of the culture. This extended excavation is in contrast with control conditions, where resorption is interrupted repeatedly by migration to new resorption sites in accordance with the normal OC resorption cycle.24
GCs affect the thickness of collagen leftovers at the bottom of the excavations
Although resorbing OCs solubilize both mineral and collagen, it is well established that, in the pit assay, OCs leave behind a fringe of undigested collagen at the bottom of the pit, rendering the cavitations more shallow. Since this study showed that GCs stimulate OC collagenolytic activity and render excavations deeper, we investigated whether GCs affected the abundance of collagen leftovers. SEM was used to assess the presence of the fringe at the bottom of both round pits and trenches obtained under both control (Fig. 3A) and GC conditions (Fig. 3B). We found that round pits under normal conditions had accumulated large amounts of demineralized collagen fibers but substantially less in the presence of GC (Fig. 3A, B, top). However, resorption trenches under control conditions showed even fewer demineralized fibers, and these fibers were totally missing in trenches generated in the presence of GC (Fig. 3A, B, bottom).
In order to evaluate the extent to which the thickness of the collagen fringe affects maximal depth, maximal depths were measured before and after removal of collagen fringes through NaOCl treatment. SEM analysis indicated that this treatment removes all collagen fringes (not shown). Figure 2F shows that pits generated under control conditions gained depth after the treatment up to the values reached in the presence of GC, whereas pits generated under the latter conditions did not. By subtracting the differences in resorption depth before and after NaOCl treatment, we estimated that the thickness of the demineralized organic matrix left in the pit in its deepest part is around 3 to 4 µm under control conditions, whereas under GC conditions it is between 0 and 1 µm, irrespective of the culture time (Fig. 2G). This indicates that control and GC-treated OCs demineralize bone to the same extent but that the latter show improved removal of demineralized collagen compared with controls.
A series of observations indicate that OC resorptive activity itself is involved in early GC effects on bone. However, the mechanism has remained largely unknown, and all attempts to show in vitro that GCs make OCs erode a larger bone surface area compared with controls have failed. This study documents once more the absence of effect on total resorbed area but also demonstrates that GCs lead to drastic changes in the morphology of the excavations, thereby reflecting that GCs directly affect the resorption mode of the OCs. Our observations therefore also threw light on a previously unrecognized control mechanism of OC resorption.
The most striking GC-induced change is the shift from a high proportion of round excavations to a high proportion of excavations with a trenchlike appearance. This change goes along with less numerous excavations and decreased levels of TRACP, an enzyme considered to be a marker of OC activity. Interestingly, these changes appear gradually because at the very beginning of the culture GCs do not greatly affect either the trenchlike appearance or the excavation number. Therefore, we conclude that GC exposure does not affect the trigger that determines whether an OC will initiate resorption or not, nor the mechanism controlling the initial size of the area that will undergo resorption underneath the OC. A scheme reflecting our observations is provided in Fig. 4. Under control conditions, OCs typically exert their resorptive activity by excavating bone down to a certain depth and then move away, leaving behind a round pit; then they make a new pit, possibly just next to the previous one; and so on. With time, this results in an increased number of resorption events, sometimes forming tracks of single pits (Fig. 2A, orange marking), and leads to increased area of total resorbed surface. This behavior is in line with the observations made by many others and led to the resorption-cycle concept.24, 25 In contrast, GC-treated OCs exert their resorptive activity by increasing the size of the excavations they initiated so that with time they appear like long, continuous trenches, contributing to increased area of total resorbed surface without an important increase in the number of resorption events (Fig. 4). Thus GC-treated OCs behave as if they were blocked in the resorption mode of their resorption cycle. Since the common belief is that TRACP is a marker of OC activity, it may be argued that a prolongation of the resorption phase should lead to an increased instead of the decreased release of TRACP shown in this study. However, there are data indicating that TRACP is rather involved at the end of the resorption phase of the resorption cycle,26, 27 and in this case it makes sense that a prolongation of the resorption phase results in the observations made in this study, that is, a decreased TRACP release, which goes along with the increased levels of the resorption product CTX. Noteworthy, the generation of a high proportion of trenches in the presence of GCs is also supported by pictures from another study performed with human OCs generated from peripheral blood mononucleated cells,13 but this report does not comment on the morphology of the excavations and mentions only the absence of effect of GCs on total resorption area. Our study is in accordance with this absence of effect on resorption area generated by human OCs,13, 15 but not with others, where less purified mouse or rat OCs were used and where GCs rather decreased total resorption areas.10, 12, 14 We speculate that this decrease could be due to use of OCs from other animal species or to the presence of other cell types.
The other striking GC effects shown in our study are that already after a 24 hours of culture, GC-treated OCs reach larger maximum excavation depths than controls and that already at this 24-hour time point GC-treated OCs remove collagen more thoroughly compared with controls (Fig. 4). Furthermore, the effect on maximum depth results only from this GC-induced enhancement of collagenolysis and cannot be ascribed to a deeper demineralization because hypochlorite treatment does not significantly affect depth measurements in excavations made in the presence of GCs. Interestingly, enhancement of collagenolysis is demonstrated not only by increased release of collagen from the pits but also by the higher levels of the collagen degradation product CTX in the conditioned medium. Our observations are also supported by GC-induced increase of cathepsin K,7 the main collagenolytic proteinase of OCs, and by the fact that cathepsin K overexpression leads to less collagen leftovers in resorption pits.28 We conclude that when the pit assay is performed under control conditions, collagen degradation is most often slower than demineralization. This results gradually in increased amounts of collagen at the bottom of the pit as it becomes deeper. In contrast, under GC exposure, collagenolysis is commensurate with demineralization rate, and the demineralized collagen is fully removed.
What could be the mechanism making GC-treated OCs increase the size of the pits they initiated rather than migrating away and generating additional pits at other sites of the bone surface? The answer is given by the link between the two main GC effects we just discussed: Excavations with a trenchlike appearance show almost no collagen leftovers compared with round pits. Interestingly, this includes the few trenches generated under control conditions, thus stressing the strength of the relation between the trenchlike appearance of the excavations and the degree of collagen removal. The importance of contact of the OC with mineral for initiation of resorption has been demonstrated previously.25, 29, 30 Our observations suggest that keeping contact with mineral is also important for continuation of resorption because prolonged resorption occurs only when degradation of demineralized collagen is complete. The reason why prolonged resorption proceeds laterally simply may be because of steric hindrance because the resorption area underneath an OC is much smaller than the spreading area of the OC itself. Lateral resorption is fully in line with the mechanism of dynamic sealing zone by which the OC can resorb and move simultaneously,31 as particularly well supported by video imaging.25 In contrast, incomplete removal of demineralized collagen leads to a shift in contact from mineral to collagen and arrest of resorption. This arrest is consistent with prevention of OC resorptive activity when bone surfaces are covered with collagen30 and with the inability of OCs to resorb poorly mineralized bone in rickets.32 Interestingly, this shift in cell-matrix contact also results in making OCs moving to a new resorption site, in accordance with the alternating resorption/migration episodes of the resorption cycle (Fig. 4). The generation of demineralized collagen in the resorption pit may well contribute in itself to making OCs moving away because OCs shifting from a mineralized surface to a pure collagen surface undergo a number of transformations reminiscent of an epithelial-mesenchymal transition29 and contributing to a migratory phenotype33 instead of a secretory phenotype. These transformations include the loss of extended ruffled border,32, 34 changes in expression of specific cell surface proteins,29, 35 the switch from cathepsin K to matrix metalloproteinases activity,33 and reorganization of the cytoskeleton with loss of actin ring and polarization and, instead, development of podosomes characteristic of migratory activity.24, 25, 29 The hypothesis that collagen signals to migrate away from the pit so that a new pit can be made nearby is also supported by the observation that OCs make rather continuous resorption tracks on pure mineral compared with distinct round pits on bone or dentine, where collagen is part of the matrix.36, 37 Similar observations were obtained on bone rendered anorganic.30 Continuous erosion over bone slices without leaving behind collagen remnants is reminiscent of continuous erosion over bone trabeculae and absence of collagen leftovers in vivo, where maybe factors from surrounding cells play the role of GCs in our experiments.38–40 One may speculate that the mechanism for stopping the continuation of erosion in physiology is via factors that reduce collagenolysis such as estrogen, which was reported to lower cathepsin K levels41 and to lead to more collagen in the pits.42 Thus the extent of erosion over bone surfaces may be regulated by the balance of factors either upregulating or downregulating cathepsin K. Therefore, in patients treated with high-dose GC, a cathepsin K inducer,7 continuous erosion may be strongly accentuated, whereas in postmenopausal patients, cathepsin K inhibitors decrease bone resorption.43
The increased risk of fractures conferred by exposure to GCs could not be explained just by lower bone mineral density,44 and GC treatment also leads to greater trabecular thinning.8, 45 These clinical observations remained unexplained but might reflect the alterations in OC behavior that are induced by a direct action of GCs on OCs in this in vitro study. As a support for this view, it is interesting that Dalle-Carbonare and colleagues reported increased length of resorption cavities in bone biopsies of GC-treated patients,45 which may correspond to the trenchlike appearance of the excavations in bone slices of this study. One may expect that deep trenchlike excavations affect bone strength differently than round excavations away from each other. Accordingly, novel skeletonization and meshing algorithms are being performed on 3D images of the excavations produced in the presence of GCs in this study, and so far, they predict lower stiffness and increased fracture risk despite unchanged total resorbed area.46 Furthermore, the serum levels of CTX are increased in GC-treated patients,5, 6 as are also the levels of CTX in the present pit assay in the presence of GCs. However, as stressed in the introduction, GC-induced osteoporosis is multifactorial2, 8 and also may result from increased OC survival and differentiation9–11, 13, 15, 16 and may be due both to direct effects and mediation by other cell types.2, 7, 8, 17 Therefore, the contribution of the effect of GC on resorptive activity revealed in this study should be considered in this general context.
In conclusion, GC exposure makes OCs erode bone surfaces over long distances without interruption by the episodes of pure migration that normally alternate with resorption. This change in behavior probably occurs through GC-induced increased collagenolysis, which ensures prolonged contact between the OC and mineral. Considerations on signaling pathways recently led Novack and Faccio to speculate that aberrant bone resorption in bone diseases is due not only to excessive resorptive activity of OCs but also to a combination of altered motion and resorption.47 This study provides clear biologic evidence supporting this view.
All the authors state that they have no conflicts of interest.
We wish to thank Birgit MacDonald and Vibeke Nielsen for excellent technical assistance and MD Flemming Brandt Sørensen from the Department of Pathology, Vejle Hospital, for letting us use his microscope to measure resorption depths. This study was financed in part by a grant from the Region of Southern Denmark (08/8932).