Influence of bone affinity on the skeletal distribution of fluorescently labeled bisphosphonates in vivo

Authors


Abstract

Bisphosphonates are widely used antiresorptive drugs that bind to calcium. It has become evident that these drugs have differing affinities for bone mineral; however, it is unclear whether such differences affect their distribution on mineral surfaces. In this study, fluorescent conjugates of risedronate, and its lower-affinity analogues deoxy-risedronate and 3-PEHPC, were used to compare the localization of compounds with differing mineral affinities in vivo. Binding to dentine in vitro confirmed differences in mineral binding between compounds, which was influenced predominantly by the characteristics of the parent compound but also by the choice of fluorescent tag. In growing rats, all compounds preferentially bound to forming endocortical as opposed to resorbing periosteal surfaces in cortical bone, 1 day after administration. At resorbing surfaces, lower-affinity compounds showed preferential binding to resorption lacunae, whereas the highest-affinity compound showed more uniform labeling. At forming surfaces, penetration into the mineralizing osteoid was found to inversely correlate with mineral affinity. These differences in distribution at resorbing and forming surfaces were not observed at quiescent surfaces. Lower-affinity compounds also showed a relatively higher degree of labeling of osteocyte lacunar walls and labeled lacunae deeper within cortical bone, indicating increased penetration of the osteocyte canalicular network. Similar differences in mineralizing surface and osteocyte network penetration between high- and low-affinity compounds were evident 7 days after administration, with fluorescent conjugates at forming surfaces buried under a new layer of bone. Fluorescent compounds were incorporated into these areas of newly formed bone, indicating that “recycling” had occurred, albeit at very low levels. Taken together, these findings indicate that the bone mineral affinity of bisphosphonates is likely to influence their distribution within the skeleton. © 2012 American Society for Bone and Mineral Research.

Introduction

Bisphosphonates (BPs) are widely used in the treatment of osteoporosis, Paget's disease, and cancer-associated bone disease, all of which involve excessive osteoclast activity. BPs target osteoclasts and inhibit bone resorption because of their strong affinity for bone mineral. The major mode of action of the nitrogen-containing BPs (N-BPs) is inhibition of the intracellular enzyme farnesyl pyrophosphate synthase.1–3 For a range of BPs, it was demonstrated that the potency for inhibiting this enzyme correlates with the antiresorptive potency in vivo,4 suggesting that inhibition of FPP synthase is the major determinant for in vivo potency. However, clinically used BPs also differ in their affinity for bone mineral, which has been suggested to influence clinical outcome by affecting the uptake and retention by the skeleton, the distribution within bone, and the potential recycling of these compounds.5, 6

The classical view is that the ability of BPs to bind to bone mineral is attributable to the two phosphonate groups, referred to as the “bone hook,” with added contribution from the hydroxyl side chain, which cooperates in chelating calcium ions.7 However, it has become clear that differences in bone mineral binding exist between the clinically relevant BPs that vary only in their second (R2) side chain, suggesting that this side chain also influences affinity for bone mineral.5 A variety of different methods and techniques have been employed in recent years to determine the rank order of mineral affinity between clinically used BPs and various analogues, including constant composition kinetic analysis of binding to hydroxyapatite8 or carbonated apatite,9 retention of BPs on hydroxyapatite columns,10 and competition binding assays and NMR analysis of binding of BPs to human bone.11, 12 Although not all techniques yielded identical rank orders of affinity, overall findings were similar, with alendronate and pamidronate consistently identified as having relatively high affinity, whereas risedronate (RIS) has intermediate to low affinity among the clinically used BPs.

It is still unclear how differences in affinity for bone mineral between BPs affect their microanatomical localization in bone and how this may impact on their clinical effects. Previous studies using radiolabeled alendronate or etidronate demonstrated preferential localization to areas of bone turnover, binding to newly formed mineral and mineral exposed by resorption.13–15 Furthermore, we recently demonstrated, using fluorescent analogues of RIS,16 that these compounds penetrate the osteocyte network and bind to osteocyte lacunar walls in vivo.17 However, whether differences exist between high- and low-affinity compounds in their preferential binding to formation and resorption sites, and their degree of penetration of the osteocyte network, is unclear.

As a proof-of-concept study, we have compared the microanatomical localization of a series of fluorescently labeled BP analogues that were predicted to differ widely in their affinity for bone mineral based on their parent compounds: RIS (high affinity), deoxy-RIS that lacks the hydroxyl side chain (medium affinity),18 and the phosphonocarboxylate analogue 3-PEHPC (low affinity).19

Materials and Methods

Synthesis and purification of fluorescent BP analogues

RIS and its analogues deoxy-RIS (dRIS) and 3-PEHPC were provided by Procter & Gamble Pharmaceuticals (now Warner Chilcott, Ireland). Fluorescent BP analogues were synthesized by stable conjugation of the respective fluorophore (FAM, carboxyfluorescein; ROX, carboxy-X-rhodamine; RRX, Rhodamine Red-X; AF647, Alexa Fluor 647) to the pyridine nitrogen of RIS via a linker strategy and purified by TLC and HPLC, as previously described.16 The fluorescent conjugates were dissolved in phosphate-buffered saline.

Binding to dentine in vitro

Dentine discs (5 mm diameter) from elephant tusk were incubated with 100 µL of 1- or 10-µM fluorescent BP analogues in Tris-buffered saline (TBS), pH 7.4, for 1 hour at room temperature. Solutions were then recovered and their fluorescence measured on a plate reader (FAM- and ROX-labeled compounds) or LI-COR Infrared Imager (LI-COR Biosciences, Lincoln, NE, USA; AF647-labeled compounds). The percentage of each compound bound to dentine was calculated from standard curves. To compare binding of compounds to resorbed and unresorbed dentine surfaces, mature osteoclasts isolated from newborn rabbits were seeded onto dentine and incubated for 48 hours, as previously described.20 Discs were then labeled with fluorescent BP analogues,21 fixed in 4% paraformaldehyde, mounted on glass slides in VectaShield (Vector Laboratories, UK), and analyzed by confocal microscopy. Z stacks, comprising images of 1-µm optical thickness, were captured from the dentine surface to the bottom of the resorption pit(s) in the field of view, then average fluorescence intensities from maximum-intensity projection images of both resorption pits and unresorbed surfaces (10 per treatment group) were determined using AIM software (Zeiss Ltd., UK).

Treatment of animals

Rat studies were conducted at the Procter & Gamble Health Sciences Center at Mason, OH, USA. They were performed in accordance with the guidelines of Procter & Gamble Pharmaceuticals' Institutional Animal Care and Use Committee, and meet the guidelines established by the Animal Welfare Act. Male Sprague-Dawley rats, approximately 9 weeks old, were obtained from Charles River Laboratories International, Inc. (Raleigh, NC, USA). They received equimolar amounts of fluorescent BP analogues (or vehicle) intravenously, as indicated in the relevant figure legends. Doses given were 0.35 mg/kg FAM-RIS, 0.34 mg/kg FAM-dRIS, 0.42 mg/kg ROX-RIS, 0.41 mg/kg ROX-3-PEHPC, 0.47 mg/kg RRX-3-PEHPC, 0.58 mg/kg AF647-RIS, and 0.56 mg/kg AF647-3-PEHPC (all 0.15 mg/kg RIS molar equivalent), unless stated otherwise. One or seven days later, animals were sacrificed and tibias were collected and fixed in 70% ethanol. Mouse studies were conducted at the University of Aberdeen. They were approved by the Home Office (UK) and conducted in accordance with the Animals (Scientific Procedures) Act 1986 and the Home Office Code of Practice. Three-month-old C57Bl/6 mice received equimolar amounts of fluorescent BP analogues (0.9 mg/kg AF647-RIS, 0.5 mg/kg FAM-3-PEHPC), xylenol orange (90 mg/kg), and/or calcein (15 mg/kg) subcutaneously, as indicated in the relevant figure legends. Mouse bones were fixed in 4% formaldehyde overnight, then stored in 70% ethanol before embedding.

Detection of AF647-labeled RIS and 3-PEHPC in whole tibias

The right tibias of rats intravenously injected with 0.29 mg/kg AF647-RIS or 0.28 mg/kg AF647-3-PEHPC were scanned on a LI-COR Odyssey Infrared Imager using the 680-nm laser. A tibia from a vehicle-treated rat served as background control. Fluorescence intensity was determined using LI-COR Odyssey software v. 2.1, and expressed relative to fluorescence of AF647-RIS at 1d.

Histological sample processing and analysis

Tibias were cut in half through the diaphysis, and a ∼4-mm section was cut from the proximal end and embedded in methylmethacrylate (MMA) for cross-sectional analysis. The remaining proximal end was embedded in MMA for longitudinal analysis. Polished block surfaces were analyzed using a Zeiss LSM510 META system and either AIM or ZEN software (Zeiss Ltd.). Images in xy were captured at a depth of ∼10 µm beneath the block surface. Instrument settings were optimized for each image to avoid saturation in any channel, unless otherwise stated. Forming bone surfaces were identified by the presence of unlabeled matrix on top of the fluorescent BP conjugates, as well as the appearance of osteoblasts in the process of becoming entombed in the matrix. Resorbing surfaces were identified by the irregular, scalloped appearance of the bone surface. Quiescent surfaces were identified by their smooth appearance and absence of matrix overlaying the fluorescent BP conjugates.

Quantification of bone surface labeling

Images (optical section thickness ∼5 µm) of different regions of the cortical bone in cross sections of tibias were obtained from each animal by confocal microscopy using a 20× objective (NA 0.75). Cortical bone surface labeling was quantified from at least three images of each animal that showed a forming endocortical and resorptive periosteal surface. Using Image J software, rectangular regions of equal size were drawn around each of the two bone surfaces, and total amount of fluorescence was determined within these areas by densitometry. Background fluorescence determined from nonlabeled areas was subtracted, and fluorescence at the endocortical surface was expressed relative to fluorescence at the periosteal surface.

Determination of relative mineral surface penetration

At least five images (optical section thickness 1.2 µm) of different regions of the endocortical surface in cross sections of tibias were obtained from each block using a 40× oil objective (NA 1.3). Using the profile tool in AIM software, six fluorescence profiles were obtained from each image at random points across the bone surface. Average distance of penetration into the mineral surface was determined for each of the compounds by calculating the total fluorescence (ie area under the curve), and determining the distance at which half of the cumulative fluorescence was reached. For each measurement, data were expressed relative to the compound with intermediate bone affinity, FAM-dRIS, or in the case of treatment with the three fluorescent RIS compounds, to FAM-RIS.

Quantification of labeling of osteocyte lacunar walls

Three images (optical section thickness 1.2 µm) of different regions throughout the cortical bone were obtained from cross sections of tibias using a 40× oil objective (NA 1.3), with detector gain settings optimized for detection of vascular channel wall labeling. Using AIM software, mean fluorescence intensity of each osteocyte lacunar wall and vascular channel wall was determined, and the distance of each osteocyte lacuna from the nearest labeled vascular channel or bone surface measured. At least 100 osteocytes were analyzed for each animal. Data from each image were corrected for background fluorescence, determined from nonlabeled areas, and fluorescence of the osteocyte lacunar walls was expressed relative to the average fluorescence of the vascular channel walls. For some animals, results were validated by analysis of a tibia from a vehicle-treated animal to correct for autofluorescence, which showed similar results (not shown). For determining the proportion of osteocyte lacunae that were labeled with a particular compound, a lacuna was considered to be labeled if the average fluorescence of the lacunar wall was greater than the mean +2 SD of the background fluorescence, determined from 10 background readings for each image.

Statistical analysis

Data were analyzed using GraphPad Prism software (GraphPad Software, Inc., La Jolla, CA, USA). Statistical significance was determined using one- or two-way repeated measures ANOVA and Bonferroni multiple comparisons post hoc test, unless stated otherwise. A p value <0.05 was considered statistically significant. Regression analysis was carried out using nonlinear (semilog) ordinary least squares fit, and goodness-of-fit (R2) was calculated.

Results

Binding of fluorescently labeled BP analogues to mineral surfaces

The phosphonocarboxylate 3-PEHPC consistently showed a significantly lower degree of binding to dentine than its parent compound RIS for all three fluorophore conjugates (Fig. 1A–C). In addition, at 10 µM, FAM-dRIS showed intermediate mineral binding compared with FAM-RIS and FAM-3-PEHPC. This confirms that these fluorescent conjugates exhibit similar differences in affinity to their parent compounds. However, the choice of fluorophore conjugated to RIS or 3-PEHPC appeared to influence binding to dentine, with ROX-labeled compounds showing a relatively greater binding compared with FAM- and AF647-labeled compounds. To confirm these results (obtained by analyzing the fluorescence intensity of the solution before and after incubation with dentine), confocal microscopy was used to quantify labeling of the dentine surface (Supplemental Fig. S1). This confirmed that FAM-3-PEHPC showed a significantly lower degree of binding to the dentine surface than FAM-RIS (p < 0.001; Supplemental Fig. S1B). To determine whether similar differences in bone mineral binding occur in vivo, tibias from rats treated with equimolar amounts of either AF647-RIS or AF647-3-PEHPC either 1 or 7 days before death were scanned on an Infrared Imager to detect AF647 fluorescence (Fig. 1D). For both compounds, fluorescence at day 7 was similar to day 1. However, AF647-3-PEHPC showed a significantly lower degree of binding to bone compared with AF647-RIS (p < 0.0001; combined data from the two time points), confirming its lower affinity for bone in vivo.

Figure 1.

Relative mineral affinities of fluorescent BP analogues. (A–C) Percentage of each of the fluorescent BP analogues that was bound to dentine discs in vitro after incubation with 1 or 10 µM at pH 7.4 for 1 hour, expressed as mean ± SD (n = 4). **p < 0.01; ***p < 0.001. (D) Fluorescence intensities of rat tibias 1 or 7 days after intravenous administration of either AF647-RIS (0.29 mg/kg) or AF647-3-PEHPC (0.28 mg/kg), determined using an Infrared Imager. Fluorescence intensity for each tibia was corrected for background fluorescence of the vehicle and expressed as mean ± SD (n = 2) relative to the AF647-RIS at 1-day group. ***p < 0.001 (combined data of 1- and 7-day animals), as analyzed by unpaired t test. Examples of scan results are shown on the right, with fluorescence shown in black (composite image).

Distribution of high- and low-affinity BP analogues on cortical bone surfaces in vivo

To assess potential differences in binding to resorption and formation surfaces, we analyzed cross sections through the tibias of the growing rats, focusing on specific areas that exhibited extensive endocortical bone formation and periosteal bone resorption and areas that were quiescent on both surfaces. At the doses used, all the compounds showed a significantly higher degree of binding to bone-forming endocortical surfaces compared with resorbing periosteal surfaces (Fig. 2A–F). These differences were not observed between quiescent endocortical and periosteal surfaces (Fig. 2G). Lower-affinity compounds showed relatively more intense labeling of bone surfaces compared with intracortical vascular channel walls (Fig. 2A–C).

Figure 2.

Distribution at forming endocortical and resorbing periosteal surfaces. Optical sections acquired by confocal microscopy of cortical bone surfaces in tibias of 9-week-old rats given a single dose of AF647-RIS, FAM-dRIS, and ROX-3-PEHPC (n = 3; A,D); ROX-RIS, FAM-dRIS, and AF647-3-PEHPC (n = 3; B,E,G); or ROX-RIS, FAM-RIS, and AF647-RIS (n = 3; C,F) 1 day before death. (A–C) Labeling of areas of cortical bone with a forming endocortical (left side of images) and resorbing periosteal surface (right side). For each image, fluorescence of the different labels (AF647, FAM, ROX) are shown separately for clarity. Scale bars = 100 µm. (D–F) Quantification of labeling of forming endocortical and resorbing periosteal surfaces by densitometry using ImageJ software (≥ 3 images/rat). Data were corrected for background fluorescence and expressed as mean ± SD (n = 3) of endocortical surface labeling intensity relative to periosteal surface labeling intensity (dotted lines) for each compound. *p < 0.05; **p < 0.01; ***p < 0.001, relative to periosteal surface. (G) Labeling of quiescent endocortical (left side) and periosteal surfaces (right side) in an area of cortical bone, showing a similar degree of labeling on both surfaces. Scale bars = 100 µm.

Distribution of high- and low-affinity BP analogues at resorbing surfaces

At resorbing periosteal surfaces, distinct patterns of labeling were observed between compounds (Fig. 3). This was most pronounced in animals that had received the compounds with the largest difference in affinity (ROX-RIS and AF647-3-PEHPC), which revealed a preferential binding of the lower-affinity compound to distinct areas that resembled Howship's lacunae (Fig. 3A, arrows), whereas the higher-affinity compound showed more uniform bone surface labeling (Fig. 3A). Comparison of the distribution of AF647-RIS and ROX-3-PEHPC, two compounds with a smaller difference in affinity for bone mineral based on in vitro dentine binding (Fig. 1), showed a higher degree of colocalization, with preferential binding to distinct lacunar areas for both compounds (Fig. 3B, arrows). These findings were confirmed in animals treated with ROX-RIS and AF647-RIS, in which AF647-RIS preferentially bound to distinct lacunar areas, whereas the higher-affinity ROX-RIS again labeled the periosteal surface more uniformly (Fig. 3C). No clear differences in distribution between the highest- and lowest-affinity compounds were observed at quiescent periosteal surfaces (Fig. 3D). Finally, FAM-RIS and FAM-3-PEHPC displayed a similar preferential labeling of resorbed surfaces on dentine discs in vitro (after resorption by rabbit osteoclasts; Supplemental Fig. S2).

Figure 3.

Distribution at resorbing and quiescent periosteal surfaces. Optical sections, acquired by confocal microcopy, of resorbing (A–C) or quiescent (D) periosteal surfaces from diaphysial cross sections of tibias of 9-week-old rats 1 day after administration of equimolar amounts of ROX-RIS (green), FAM-dRIS and AF647-3-PEHPC (red) (A,D); AF647-RIS (green), FAM-dRIS and ROX-3-PEHPC (red) (B); or ROX-RIS (green), FAM-RIS and AF647-RIS (red) (C). Only ROX- and AF647-labeled compounds are shown for clarity. Left panels show highest-affinity compounds and middle panels lowest-affinity compounds. Arrows indicate resorbed areas with more intense labeling of lower-affinity compounds. Scale bars = 50 µm. Inset in merged image of (A) shows a higher magnification image of a different resorbing site, demonstrating differential penetration (arrowheads) of the mineral by the two compounds (scale bar = 20 µm).

Mineral surface penetration of high- and low-affinity BP analogues

At bone-forming endocortical surfaces, a differential degree of penetration of the bone matrix was observed, with affinity for bone mineral negatively correlating with penetration into the mineralizing surface. In rats treated with ROX-RIS, FAM-dRIS, and AF647-3-PEHPC, farthest penetration into the mineralizing surface was observed with the lowest-affinity compound, AF647-3-PEHPC, whereas the high-affinity ROX-RIS showed the lowest degree of penetration (Fig. 4A,B). When the fluorescent tags between the high-affinity RIS and low-affinity 3-PEHPC were switched (ie, using compounds with a smaller difference in mineral affinity), the 3-PEHPC analogue still showed the highest degree of mineral surface penetration, although the differences between the three compounds were less pronounced (Fig. 4C,D). FAM-RIS, ROX-RIS, and AF647-RIS also showed differences in mineral surface penetration when administered together, with the ROX conjugate showing decreased surface penetration and the AF647 conjugate slightly increased penetration, relative to FAM-RIS (Fig. 4E,F). These findings are consistent with the differences in mineral affinity of these compounds in vitro, showing ROX-labeled compounds to have a higher mineral affinity than FAM- and AF647-labeled compounds (Fig. 1). Similar differences in surface penetration were observed in trabecular bone, with the low-affinity AF647-3-PEHPC again penetrating farthest into the bone matrix at sites of bone formation (Fig. 4G). Furthermore, analysis of the tibia of a mouse treated with AF647-RIS, FAM-3-PEHPC, and xylenol orange showed xylenol orange (even lower bone affinity than the 3-PEHPC conjugates) to penetrate farthest, whereas AF647-RIS showed the most superficial labeling (Fig. 4H). Moreover, when dentine discs from elephant tusk, a relatively poorly mineralized tissue, were labeled with fluorescent BP analogues in vitro, a similar pattern of mineral penetration was observed, with FAM-3-PEHPC penetrating further into the dentine surface compared to AF647-RIS, which in turn showed similar surface penetration to FAM-RIS (Supplemental Fig. S3). In contrast, no clear differences in penetration between high- and low-affinity compounds were observed at quiescent bone surfaces, and the seams of fluorescence at these areas were much thinner (Fig. 4I). Evidence of differential penetration was found at areas of resorption, although this was less pronounced than at forming surfaces (Fig. 3A).

Figure 4.

Penetration at mineral surfaces. Mineral surface penetration of fluorescent BP analogues was compared in tibias 1 day after administration. (A–G,I) Nine-week-old rats were treated with ROX-RIS (red), FAM-dRIS (green), and AF647-3-PEHPC (blue) (A,B,G,I); AF647-RIS expressed as mean ± SD (n = 2) (blue), FAM-dRIS (green), and ROX-3-PEHPC (red) (C,D); or AF647-RIS (blue), FAM-RIS (green), and ROX-RIS (red) (E,F). (H) A 3-month-old mouse was treated with AF647-RIS (blue), FAM-3-PEHPC (green), and xylenol orange (red). Optical sections show labeling at forming endocortical (A,C,E,H), trabecular (G), and quiescent endocortical bone surfaces (I). Asterisks indicate bone marrow. Fluorescence intensity profiles shown below each image were obtained along the white arrows. Scale bars = 20 µm. (B,D,F) Average depth of penetration of compounds into the bone matrix, relative to FAM-dRIS (B,D) or FAM-RIS (F). For each animal, ≥5 images and 6 fluorescence profiles per image were analyzed. Data are shown as mean ± SD (n = 3). *p < 0.05; **p < 0.01 (paired t test).

Penetration of the osteocyte canalicular network by high- and low-affinity BP analogues

We previously reported the binding of fluorescent RIS analogues, FAM-RIS and AF647-RIS, to the walls of osteocyte lacunae and within the canaliculi extending from the vascular channels or bone surface to the osteocyte lacunae.17 In the current study, it was consistently observed, both in rats (Fig. 5) and in mice (Supplemental Fig. S4), that the lower-affinity compounds (fluorescent analogues of dRIS and 3-PEHPC, and xylenol orange) showed increased labeling of osteocyte lacunar walls, as well as canaliculi extending from the bone surface or vascular channels, compared with the high-affinity fluorescent analogues of RIS. Quantification of osteocyte lacunar labeling relative to vascular channel labeling confirmed that the degree of penetration into the osteocyte network was significantly different between the fluorescent RIS, dRIS, and 3-PEHPC analogues (Fig. 5B,C; p < 0.05). The low-affinity AF647-3-PEHPC showed a significantly higher degree of labeling of the osteocyte network relative to the vascular channels compared with the high-affinity ROX-RIS (p < 0.05), with FAM-dRIS showing intermediate labeling (Fig. 5B). Similarly, AF647-RIS showed less binding to osteocyte lacunar walls than either FAM-dRIS and ROX-3-PEHPC (Fig. 5C). This difference reached statistical significance for FAM-dRIS (p < 0.05). Finally, coadministration of FAM-RIS, ROX-RIS, and AF647-RIS to rats resulted in relatively low osteocyte lacunar wall labeling for all three RIS conjugates (Fig. 5D), confirming that the main determinant for osteocyte network penetration is the affinity of the native compound.

Figure 5.

Penetration of the osteocyte network. (A) Xy maximum-intensity projection image generated from a z series (25 µm depth) of images of tibial cortical bone from a rat treated with FAM-RIS (green) and RRX-3-PEHPC (red), acquired using confocal microscopy. Yellow indicates colocalization. Scale bar = 20 µm. (B–E) Nine-week-old rats were given single doses of ROX-RIS, FAM-dRIS, and AF647-3-PEHPC (n = 3; B); AF647-RIS, FAM-dRIS, and ROX-3-PEHPC (n = 3; C,E); or AF647-RIS, FAM-RIS, and ROX-RIS (n = 3, D) 1 day before sacrifice, and optical sections were acquired from cross sections through tibial cortical bone using confocal microscopy with detector gain settings optimized to avoid saturation around osteocytes and vascular channels. For each rat, the mean fluorescence intensity of >100 osteocyte lacunar walls, and nearby vascular channel walls, was determined from three images and corrected for tissue autofluorescence. (B–D) Mean fluorescence intensity of osteocyte lacunar walls relative to vascular channel walls, expressed as mean ± SD (n = 3). *p < 0.05. (E) Average fluorescence intensities of osteocyte lacunar walls, expressed as a function of the distance to the nearest vascular channel or bone surface. Combined data from three animals are shown.

Accordingly, when analyzing the proportion of labeled osteocyte lacunae, the highest-affinity compound labeled fewer lacunae than the lower-affinity compounds (15 ± 8% for AF647-RIS versus 32 ± 3%, and 31 ± 3% for FAM-dRIS and ROX-3-PEHPC, respectively; mean ± SD, n = 3; p < 0.05). Similar results were found when the fluorophores were switched between the highest- and lowest-affinity compounds (14 ± 8% for ROX-RIS versus 26 ± 15% and 28 ± 20% for FAM-dRIS and AF647-3-PEHPC, respectively; mean ± SD, n = 3), although this was not statistically significant. However, the absolute values should be interpreted with caution because they depend strongly on the threshold used to define positivity. We used a conservative approach in which lacunae were considered positive if the fluorescence was greater than the mean plus 2 standard deviations of the background fluorescence, and the values are therefore likely to be an underestimation.

We previously showed that for fluorescent RIS, an inverse relationship exists between degree of osteocyte lacunar labeling and distance of the osteocyte lacuna from the nearest vascular channel or bone surface, ie, lacunae of osteocytes close to a vascular channel showed a higher degree of labeling compared with lacunae from more distant osteocytes.17 In the current study, similar inverse relationships between degree of labeling and distance from the nearest vascular channel or bone surface was found for the lower-affinity analogues FAM-dRIS and ROX-3-PEHPC (Fig. 5E). This relationship was found to be best described by a logarithmic function, with goodness-of-fit R2 FAM-dRIS 0.24; R2 ROX-3-PEHPC 0.23; and R2 AF647-RIS 0.25. For AF647-3-PEHPC, such a relationship was not apparent using the image analysis method employed (R2 0.01).

Localization of high- and low-affinity BP analogues 7 days after administration

To investigate localization of high- and low-affinity compounds 7 days after administration, we first treated mice with AF647-RIS and xylenol orange (ie, compounds with extreme differences in affinity) on day 1, followed by a single injection of calcein to label newly formed bone surfaces on day 7. Analysis of tibial cortical bone surfaces at day 8 showed that, at forming surfaces, both AF647-RIS and xylenol orange were buried under a layer of newly formed bone, the surface of which was clearly identified by labeling with calcein (Fig. 6A). Differences in penetration at mineralizing surfaces was also evident in the seams of AF647-RIS and xylenol orange running through the cortical bone, with the very low-affinity xylenol orange clearly buried deeper at the original bone surface compared with AF647-RIS (Fig. 6A). Similar results were seen at forming surfaces in tibias of rats, in which the difference in penetration of the high-affinity (ROX-RIS) and low-affinity compound (AF647-3-PEHPC) was still apparent 1 week after treatment (Fig. 6B,C; arrows). No differences in matrix penetration were found between high- and low-affinity compounds at quiescent surfaces (Fig. 6B,C; arrowheads), further confirming the observations made at 1 day (Fig. 4I). Further analysis of these forming cortical bone surfaces using high-detector gain settings revealed evidence of very small amounts of the compounds being incorporated into bone that had formed in the 7 days after administration of compounds. This “recycling” was somewhat more evident for the 3-PEHPC analogues compared with the RIS analogues (Fig. 6D,E). Small amounts of fluorescent BP conjugates were also found to be incorporated into newly formed bone matrix surrounding some vascular channels, ie, between the deeper seam of the original label and the new bone surface (Fig. 6F). Again, this was more evident for the 3-PEHPC analogues. Finally, differential labeling of osteocyte lacunar walls was still observed after 1 week, with lower-affinity 3-PEHPC analogues showing a higher degree of labeling of the osteocyte network than RIS analogues (Fig. 6F), similar to at 1 day (Fig. 5A).

Figure 6.

Distribution of fluorescent BP analogues 1 week after administration. (A) Optical section showing endocortical surface labeling in the cortical bone of a 3-month-old mouse 7 days after administration of AF647-RIS (blue) and xylenol orange (red) and 1 day after administration of calcein (green). Scale bar = 20 µm. (B,C) Optical sections of forming (arrows) and quiescent (arrowheads) endocortical and trabecular surfaces in the cortical bone of 9-week-old rats 7 days after administration of ROX-RIS (red), FAM-dRIS (green), and AF647-3-PEHPC (blue). Scale bars = 20 µm. (D,E) Optical sections of forming endocortical surfaces in cortical bone of 9-week-old rats 7 days after administration of (D) AF647-RIS (green), ROX-3-PEHPC (red), and FAM-dRIS or (E) ROX-RIS (green), AF647-3-PEHPC (red), and FAM-dRIS. Detector gain settings were optimized to detect compounds in newly formed bone (resulting in saturation of fluorescence at original site of binding). Only ROX- and AF647-labeled compounds are shown for clarity. Black arrows indicate bone surface. Fluorescence profiles shown on the right were obtained along the white arrows in the merged images. Scale bars = 20 µm. (F) Optical section of cortical bone of a 9-week-old rat 7 days after administration of FAM-RIS (green) and RRX-3-PEHPC (red) showing labeling around osteocyte lacunar walls. Scale bars = 20 µm.

Discussion

It has long been established that the two phosphonate groups of BPs confer the bone targeting property of these drugs because of their avid binding to calcium ions. The hydroxyl side chain, which is present in all the potent clinically used BPs, increases the affinity of this interaction by enabling tridentate binding to calcium ions.7 Although the other (R2) side chain was originally thought to influence only the potency of the drug for inhibition of FPP synthase, it has recently become evident that this side chain influences affinity for bone mineral, and consequently differences in affinity exist between the clinically relevant BPs that vary only in this side chain. A possible explanation is the existence of differences in electrical charge on the nitrogen atom; high-affinity alendronate is predominantly protonated at neutral pH, whereas the lower-affinity RIS is mostly deprotonated.8, 12 In addition, differences in the three-dimensional orientation have been implicated in the hydrogen-bonding characteristics of the nitrogen with hydroxyapatite.5, 6 Although it is unclear exactly how the R2 side chain influences mineral binding, Mukherjee and colleagues recently proposed a two-site binding model of BPs to hydroxyapatite, with either a weak binding involving one of the phosphonate groups or a stronger binding mode involving both phosphonate groups and the hydroxyl group, which is most favored by alendronate, followed by pamidronate, zoledronate, and RIS.22 Such differences in affinity for bone mineral may influence clinical outcome by affecting uptake and retention by the skeleton, distribution within bone, and the potential recycling of these compounds. We have investigated these possibilities by using novel fluorescent conjugates of RIS (high affinity) and two analogues with reduced affinity for bone mineral: dRIS (which lacks the geminal hydroxyl group; medium affinity) and 3-PEHPC (which lacks one of the phosphonate groups; low affinity).

Analysis of binding to dentine discs in vitro confirmed the expected differences in affinity between the fluorescent conjugates of RIS (high affinity), dRIS (medium affinity), and 3-PEHPC (low affinity), which correlated well with previous affinity data for the unlabeled compounds obtained by comparing elution times on hydroxyapatite columns,10, 18 although the fluorophore exerted an influence on mineral affinity (ROX increased affinity and AF647 decreased affinity).

After administration of AF647-RIS or AF647-3-PEHPC to growing rats, the amount of the fluorescent conjugate incorporated in tibias at 1 day was the same as at 7 days for both compounds. This is consistent with pharmacokinetic studies of BPs in humans, showing the vast majority of the BP is cleared from the circulation and either bound to bone or excreted via the urine within 24 hours.23 Our data further show that the lower-affinity AF647-3-PEHPC, once bound to bone, was efficiently retained in the skeleton, similar to AF647-RIS. However, the amount of AF647-3-PEHPC retained in bone was significantly less than the amount of AF647-RIS after in vivo administration of an equimolar dose. These results are consistent with a study in humans comparing urinary excretion of 14C-alendronate and 14C-risedronate as an inverse measure of skeletal binding, showing significantly higher urinary excretion of risedronate (55%) compared with the higher-affinity alendronate (48%; p < 0.02) within the first 72 hours after administration.24 An inverse relationship between mineral affinity and urinary excretion has also been seen with a range of BPs in rats.6

All compounds analyzed showed a higher degree of labeling at forming endocortical surfaces compared with resorbing periosteal surfaces, irrespective of affinity. By contrast, no difference in labeling was observed between quiescent endocortical and periosteal surfaces, suggesting that the differential labeling relates specifically to the type of surface (ie, forming or resorbing) rather than the anatomical site. The relatively high degree of labeling at forming surfaces is probably because of the large surface area provided by the newly formed mineral crystals that is available for binding. In a study using a far-red fluorescent pamidronate in rats, Kozloff and colleagues also found intense labeling at forming surfaces, although no quantitative comparison was made between labeling at forming and resorbing surfaces.25 In contrast, previous studies with radiolabeled BP have shown preferential binding to resorbing surfaces.13–15 This difference in labeling observed after administration of radiolabeled BP compared with fluorescently labeled BP could be a function of the specific activity of the labeled material; the purity of fluorescently labeled conjugates is greater than 95%, whereas only approximately 0.014% of the BP molecules are labeled with tritium. It is worth noting that a comparison of the localization of radiolabeled alendronate and etidronate in the growing rat showed that at equal doses of 1.3 mmol/kg, both BPs showed preferential binding to resorption surfaces compared with formation surfaces. However, when rats were treated with 0.12 mmol/kg radiolabeled alendronate and 73 mmol/kg radiolabeled etidronate (their respective pharmacologically effective doses), alendronate showed an even higher degree of preferential labeling of resorbing surfaces, whereas etidronate showed similar labeling of resorbing and forming surfaces, suggesting these differences are dose dependent.15

When analyzing the labeling at resorbing surfaces in more detail, fluorescent BP analogues showed preferential deposition at resorption lacunae, as opposed to adjacent nonresorbed areas. Similar results were obtained in vitro when dentine discs were labeled with fluorescent BP analogues after osteoclast resorption. This preferential binding to resorption sites is probably a result of the mineral becoming exposed during the resorption process, resulting in a greater mineral surface area available for binding. This is consistent with previous studies using radiolabeled BPs.13, 15 In our study, preferential binding at resorption sites was more evident with lower-affinity compounds than the higher-affinity compounds. However, as shown in Fig. 1, the absolute amount bound to bone is greater for the high-affinity compounds compared with the lower-affinity analogues. It is therefore possible that the high-affinity compounds became saturated at the resorption sites, which may have diminished the difference in labeling between resorption sites and the adjacent surfaces.

At forming surfaces, lower-affinity compounds were found to penetrate farther into the mineralizing matrix compared with compounds with higher affinity. Similar differences in penetration were observed at some resorption sites, although this was generally less pronounced than at forming surfaces. These differences are most likely a function of the degree of mineralization, with high-affinity compounds less able to penetrate as mineral content increases. One possible consequence of the deeper mineral penetration by lower-affinity compounds might be that a greater amount of these compounds binds to mineralizing surfaces than would be expected from their bone affinities. In contrast, at quiescent surfaces, the low- and high-affinity compounds colocalized in a narrow seam at the surface, indicating that none of the compounds can detectably penetrate fully mineralized bone.

We previously showed that low amounts of fluorescent RIS were bound to osteocyte lacunar and canalicular walls in bones of young adult mice.17 Binding to bone mineral surrounding osteocytes may be a clinically relevant phenomenon because very low concentrations of BPs have been reported to prevent apoptosis of osteocytes in vitro, and BP treatment was shown to prevent osteocyte apoptosis induced by glucocorticoid treatment or cyclic mechanical loading in vivo.26, 27 The anti-apoptotic effect of BPs is in stark contrast to their pro-apoptotic effects on osteoclasts and is mediated through distinct mechanisms,28, 29 involving opening of hexameric connexin-43 hemichannels in the plasma membrane, inducing calcium influx.30, 31 These potential interactions of BPs with osteocytes may contribute to their antifracture efficacy. It has been hypothesized that high-affinity compounds show less distribution to remote sites, including the vast osteocyte network, because these compounds would be trapped by the first mineral they encounter.5 In contrast, lower-affinity compounds may penetrate farther into the osteocyte network and show a higher degree of labeling of the mineral around osteocytes. The findings reported here support this notion, showing that the higher-affinity compounds more intensely label the mineral immediately surrounding blood vessels within the cortical bone relative to their binding to endocortical and periosteal surfaces, compared with the low-affinity compounds (Figs. 2 and 3). Conversely, the low-affinity compounds show a higher degree of labeling within the osteocyte canalicular network relative to their binding to intracortical vascular channel walls (Fig. 5). Penetration of compounds into the osteocyte network from the Haversian canals can be expected to occur according to a logarithmic function, reflecting the circular organization of the osteocyte network around the central canals. Indeed, an inverse logarithmic relationship was found between labeling of the osteocyte lacunar walls and distance of these lacunae from the nearest vascular channel or bone surface. The lowest-affinity compound, AF647-3-PEHPC, appears to be an exception to this, possibly because its bone affinity is so low that it penetrates the osteocyte canalicular network sufficiently freely so that osteocyte lacunar labeling no longer depends on distance from the nearest blood supply. It should be noted that our two- dimensional analysis does not take into account the three-dimensional structure of the bone tissue, or the structure of the intricate canalicular network, and therefore may not fully reflect the actual distance along the canalicular network between osteocytes and the nearest blood supply. In addition, although our studies highlight the importance of bone affinity in penetration of the osteocyte network, they do not quantify the amount of the different compounds that osteocytes may be exposed to, and therefore it is difficult to predict the pharmacological consequences.

The notion that BPs must recycle within bone is supported by published reports that BPs are detectable in body fluids long after they have been administered. The re-exposure of bone to BPs liberated by resorption has been proposed to explain the long duration of action of the drugs.5 Our data in rats that were analyzed 7 days after injection of fluorescent compounds provide evidence that there is some recycling back to bone surfaces, but this is difficult to detect compared with the much larger uptake after first exposure. These data are in accordance with a recent study looking at systemic recycling using radiolabeled BP, in which 14C-ibandronate injected intra-osseously into the femoral head of piglets showed very little redistribution into the contralateral femoral head in vivo up to 7 weeks after the initial injection.32 These observations therefore raise the question of whether recycling occurs to a sufficient degree to contribute substantially to the prolonged action of BPs in vivo, or whether other mechanisms, such as prolonged suppression of osteoclast generation from progenitors,33–35 are responsible.

In summary, these studies support the concept that BPs that vary in their bone mineral affinity behave differently in terms of distribution within bone in vivo. We have utilized three parent compounds with relatively large differences in bone mineral affinity to show differences in behavior in vivo in short-term experiments. Despite their more subtle differences in affinity, it is likely that clinically used BPs also exhibit differences in these pharmacokinetic characteristics, which could become more apparent after repeated administration, and may help to explain some of the clinical differences among BPs.

Disclosures

AJR, FPC, and CEM have received research grant funding from the Alliance for Better Bone Health (Procter & Gamble). MJR has received research grant funding from Warner Chilcott, the Alliance for Better Bone Health, Novartis, Merck, and Roche. CEM, RGGR, and MWL are consultants for Warner Chilcott. FHE is an employee of Warner Chilcott.

Acknowledgements

We are very grateful to Prof Alan Boyde (Queen Mary University of London) for his valuable advice and assistance in interpretation of the data. We also thank Gwyn Jeans at the Procter & Gamble Health Sciences Center and the staff at the Medical Research Facility in Aberdeen for expert assistance with the in vivo studies. In addition, the authors thank the Histology and Electron Microscopy core facility at the University of Aberdeen (especially Gillian Milne) for histological sample preparation, and the Microscopy and Imaging core facility (Kevin MacKenzie and Dr. Debbie Wilkinson) for confocal microscopy technical support. The authors are also grateful to Aysha Khalid for help with establishing image analysis methods, and Prof. Richard Aspden for advice on data analysis.

Authors' roles: Conception and design of the study: AJR, CEM, RGGR, MJR, MWL, FHE, and FPC. Design and synthesis of imaging probes: SS, KMB, BAK, and CEM. Acquisition and analysis of data: AJR, CS, and FPC. Interpretation of data: AJR, CS, RGGR, MWL, FHE, and FPC. Manuscript writing: AJR and FPC. Critical revision of manuscript content: RGGR, MWL, and FHE. Final approval of manuscript: All authors. AJR and FPC accept responsibility for the integrity of the data analysis.

Ancillary