Src-null mice have higher bone mass because of decreased bone resorption and increased bone formation, whereas Abl-null mice are osteopenic, because of decreased bone formation. Compound I, a potent inhibitor of Src in an isolated enzyme assay (IC50 0.55 nM) and a Src-dependent cell growth assay, with lower activity on equivalent Abl-based assays, potently, but biphasically, accelerated differentiation of human mesenchymal stem cells to an osteoblast phenotype (1–10 nM). Compound I (≥0.1 nM) also activated osteoblasts and induced bone formation in isolated neonatal mouse calvariae. Compound I required higher concentrations (100 nM) to inhibit differentiation and activity of osteoclasts. Transcriptional profiling (TxP) of calvaria treated with 1 μM compound I revealed down-regulation of osteoclastic genes and up-regulation of matrix genes and genes associated with the osteoblast phenotype, confirming compound I's dual effects on bone resorption and formation. In addition, calvarial TxP implicated calcitonin-related polypeptide, β (β-CGRP) as a potential mediator of compound I's osteogenic effect. In vivo, compound I (1 mg/kg s.c.) increased vertebral trabecular bone volume 21% (microcomputed tomography) in intact female mice. Increased trabecular volume was also detected histologically in a separate bone, the femur, particularly in the secondary spongiosa (100% increase), which underwent a 171% increase in bone formation rate, a 73% increase in mineralizing surface, and a 59% increase in mineral apposition rate. Similar effects were observed in ovariectomized mice with established osteopenia. We conclude that the Src inhibitor compound I is osteogenic, presumably because of its potent stimulation of osteoblast differentiation and activation, possibly mediated by β-CGRP.
Src's importance in bone cell physiology became apparent after phenotyping of the Src knockout mouse, which has decreased bone resorption and increased bone formation (Soriano et al., 1991; Marzia et al., 2000). Src is highly expressed in osteoclasts (Horne et al., 1992) and is activated after stimulation of αvβ3 integrin (Sanjay et al., 2001), receptor activator of nuclear factor κ-B (RANK) (Armstrong et al., 2002), and/or c-Fms (Insogna et al., 1997). Src forms complexes with Pyk2 and c-Cbl in the osteoclast (Sanjay et al., 2001) and is essential for adhesion, motility, and the reorganization of the cytoskeleton that creates the osteoclast's specialized resorptive organelle, the ruffled border (Boyce et al., 1992). Src is also expressed in osteoblasts and osteoblast precursors (Id Boufker et al., 2010) and acts as a negative regulator of osteoblast differentiation by suppressing the activity of the Runx2 transcription factor, via Yes-associated protein (YAP) (Zaidi et al., 2004). However, Src's role in the osteoblast is not clear-cut, because Src has also been reported to be activated in osteoblasts by mechanical stimulation (Granet et al., 2001), fluoroaluminate (Jeschke et al., 1998), and parathyroid hormone (Izbicka et al., 1994) and to mediate the antiapoptotic effects of sex steroids on osteoblasts (Kousteni et al., 2001), all of which suggest a positive role for Src in bone formation. It is conceivable that Src plays several context-specific roles in the osteoblast that manifest themselves as a net increase in bone formation in the knockout.
Src's role in bone cells suggests that Src inhibitors would be useful in the treatment of osteoporosis and other conditions involving bone loss (Boyce et al., 2006). Several Src inhibitors have been shown to inhibit osteoclast formation and/or activity in in vitro assays (de Vries et al., 2009), and some have shown antiresorptive activity in animal models of bone loss, osteoporosis, or metastatic bone disease (Missbach et al., 1999; Shakespeare et al., 2003, 2008; Recchia et al., 2004; Koreckij et al., 2009). However, studies evaluating the effects of Src inhibitors on osteoblast biology are more limited: pyrazolopyrimidine 2 (PP2) was shown, in a preliminary report, to induce bone formation in isolated mouse calvariae, and dasatinib, a dual Src/Abl inhibitor, increased osteoblast differentiation in vitro (Id Boufker et al., 2010). However, there are currently no published animal studies with a Src inhibitor reporting an increase in bone formation.
In addition to Src, a number of other kinases have been implicated in bone cell physiology (Li et al., 2000; Buckbinder et al., 2007; Kim et al., 2009, 2010). Of particular interest to us was Abl, whose knockout had osteopenic bones due to a failure of osteoblast differentiation (Li et al., 2000), and which is often nonselectively inhibited by Src inhibitors. We mined the Wyeth corporate database for potent Src inhibitors, with weaker activity on Abl, reasoning that these compounds would have the antiresorptive and osteogenic activities of a Src inhibitor while avoiding the antiosteogenic activity of an Abl inhibitor. We identified compound I, a benzo(g)quinoline, and proceeded to test it in a series of in vitro and in vivo assays for osteogenic activity.
Materials and Methods
Synthesis of Compound I.
Compound I (Fig. 1) was synthesized at Wyeth (Pearl River, NY) (Berger et al., 2005). The material used for the in vivo studies was >99% pure. Stock solutions of compound I were dissolved in dimethyl sulfoxide (DMSO) and stored frozen.
The ability of compound I to inhibit Src and Abl was evaluated in time-resolved fluorescence resonance energy transfer assays using the LANCE (lanthanide chelate excite) platform (PerkinElmer Life and Analytical Sciences, Waltham, MA). For Src, compounds were incubated for 2 h at room temperature in black polystyrene plates (96-well, Corning Life Sciences, Lowell, MA; 384-well, MatriCal, Spokane, WA) with 10 ng/ml Src enzyme (Invitrogen, Carlsbad, CA), 400 nM Bio-EGPWLEEEEEAYGWMDF-NH2 substrate (Anaspec, Fremont, CA; custom synthesis), and 20 μM ATP in 20 mM HEPES, pH 7.4, 10 μg/ml bovine serum albumin (BSA), 10 mM MgCl2, 1 mM 2-β-mercaptoethanol, and 0.0025% Brij. After incubation for 2 h, kinase activity was stopped by adding EDTA to a final concentration of 11.25 mM and detected by using 1 nM antiphosphotyrosine-europium (PT66) (PerkinElmer Life and Analytical Sciences) and 5 μg/ml streptavidin-allophycocyanin (PerkinElmer Life and Analytical Sciences) for 30 min at room temperature before being read on an Envision or Victor II Plate reader (PerkinElmer Life and Analytical Sciences). The percentage of inhibition was calculated relative to maximum (no compound) and minimum (no enzyme) values, and IC50 values were generated by using either SAS Excel (SAS Institute, Cary, NC; Microsoft, Redmond, WA) or XLFit software (IDBS, Guildford, UK) Abl was assayed by incubating compounds for 0.5 h in 0.125 ng/ml c-Abl (Invitrogen), 2 nM biotin-NH-KEEEAIYAAPFAKKK-COOH substrate (Anaspec; custom synthesis), and 15 μM ATP in a buffer of 50 mM HEPES, pH 7.5, 20 μg/ml BSA, 10 mM MgCl2, and 0.001% Brij. After incubation, enzyme activity was quenched by using 30 mM EDTA, and kinase activity was detected by using 1 nM antiphosphotyrosine-europium (PT66) and 4 μg/ml streptavidin-allophycocyanin as described above.
Src Family Kinase Fibroblast Assay.
Anchorage-independent proliferation assays were performed with Rat2 cells transformed with novel activated Src, Fyn, or Lck with the wild-type human catalytic domains as described previously (Boschelli et al., 2001). In brief, Rat2 cells expressing a fusion construct consisting of the N terminus of v-Src up to the beginning of the catalytic domain fused to the catalytic domain of human Src, Fyn, or Lck followed by the C terminus of v-Src were used for proliferation assays in nonbinding 96-well plates.
32D Cell-Based Assay.
Our 32D cell-based selectivity assay for kinase inhibitors used a 32D murine myeloid cell line transduced with a green fluorescent protein-containing retrovirus Moloney leukemia virus promoter-driven Src family kinase (SFK) domain with v-Src fusion proteins made in the manner described above. 32D cells were infected with the SFK-containing viruses and selected for IL-3 independence. 32D cells are normally IL-3-dependent cells, but in the presence of an oncogenic version of an SFK these cells grow in an IL-3-independent and SFK-dependent manner. In addition to SFK-dependent cell lines, a 32D cell line was developed containing the kinase domain for Abl, using an identical strategy.
32D cells, stably transfected as above, were maintained by seeding twice per week at a density of 0.5 × 105 /ml in RPMI supplemented with 10% heat-inactivated fetal calf serum, 2 mM l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin in a 37°C incubator adjusted to 5% CO2. For assay, cells were plated in 96-well plates that already contained serial dilutions of test compounds, to a final cell density of 1 × 105 cells/well and a final DMSO concentration of 0.5%. After 2 days, plates were removed from the incubator, and viable cell number was determined by using a CellTiter-Glo luminescent cell viability assay (Promega, Madison, WI) and expressed as a percentage of the number of cells in wells containing no test compounds. Dose responses were then graphed by using model 205 in XLFit, and IC50 values were calculated.
Osteoblast Differentiation Assay.
Primary human mesenchymal stem cells (hMSC) (Cambrex, East Rutherford, NJ), which are mononuclear cells from bone marrow obtained from Ficoll gradient, were cultured in mesenchymal stem cell growth medium (Lonza Walkersville, Inc., Walkersville, MD) with 2 mM Glutamax-I (Invitrogen), and 100 U/ml penicillin/100 μg/ml streptomycin (Invitrogen).
For assay, cells were seeded at 4000/well in 96-well black plates with clear bottoms (Corning Life Sciences) and incubated at 37°C 5% CO2 for 3 days before the compound treatment. On the day of treatment, the culture medium was removed from the plate, and compound I was added to the plate as follows: compound I was first serially diluted in DMSO to make interim stocks of 0.1 μM, 1 μM, 10 μM, 100 μM, 1 mM, and 5 mM from a 60-mM DMSO stock. Another 1000× dilution was made when adding the interim stock into the corresponding treatment wells, to ensure that every well received the same amount of DMSO.
On day 4, the medium and the compound treatment were replenished. Plates were incubated for another 9 days, and on day 13 medium was completely removed from each well, and 50 μl of alkaline phosphatase substrate 4-methylumbelliferyl phosphate (4-MUP) (Sigma, St. Louis, MO) were added to each well and incubated for 15 min at 37°C in the dark. Plates were read immediately by using a Victor luminometer (excitation, 360 nm; emission, 440 nm; 1 s/well). Values were normalized to total cellular protein, determined as follows: after the alkaline phosphatase reading, an equal volume of 2× lysis buffer (0.2 M Tris-HCl, pH 9.8, 0.4% Triton X-100) was added to each well containing 4-MUP and frozen at −80°C. The next day, a 30-μl sample was removed and added to 200 μl of bicinchoninic acid substrate (Thermo Fisher Scientific, Waltham, MA), incubated 30 min at 37°C, and read on a plate reader. Standards were prepared ranging from 15 to 350 μg/ml in the same lysis buffer/4-MUP mixture.
In Vitro Neonatal Mouse Calvarial Bone Formation Assay.
Neonatal mouse calvariae were prepared from 4-day-old pups as described previously (Traianedes et al., 1998). In brief, calvariae were excised, cut in half along the sagittal suture, and incubated overnight in serum-free BGJ medium (Invitrogen) containing 0.1% BSA. Then each half-calvaria was placed with the concave surface downward on a stainless-steel grid (Small Parts Inc., Miramar, FL) in a 12-well tissue culture dish (BD Biosciences, San Jose, CA), each well containing 1 ml of BGJ medium with 1.0% fetal bovine serum. Calvariae were then incubated for 7 days with DMSO (control), compound I, or β-CGRP (Sigma) in a humidified atmosphere of 95% air and 5% CO2, with a medium change on day 4.
After organ culture, calvariae were fixed in 10% neutral phosphate-buffered formaldehyde at room temperature for at least 72 h, then decalcified for 6 h in 10% EDTA in phosphate-buffered saline. Calvariae in each group were embedded in parallel in the same paraffin block, and 4-μm sections were stained with hematoxylin-eosin. Consistent bone areas (200 μm away from frontal sutures) were selected for histomorphometric analysis. In brief, a 200-μm square grid was placed on each calvaria, and total bone area and the number of osteoblasts within the grid were determined by using the Osteomeasure System (Osteometrics Inc., Decatur, GA). All cells on the bone surface were counted as osteoblasts.
Transcriptional Profiling of Neonatal Mouse Calvariae.
To investigate the mechanisms by which incubation with compound I results in the stimulation of bone formation, we performed transcriptional profiling studies using RNA samples from 4-day-old neonatal mouse calvariae after incubation for 4 days with 1 μM compound I. To obtain sufficient RNA, we used whole (instead of half) calvariae. Calvariae were preincubated for 48 h in serum-free BGJ medium containing 0.1% BSA. Calvariae then were incubated for 4 days with DMSO (control) or 1 μM compound I in BGJ medium containing 1% fetal bovine serum. Total calvarial RNA was extracted from each calvaria sample by sonication in 1 ml of denaturing buffer (Promega) supplemented with 7 μl of β-mercaptoethanol, followed by two extractions with phenol/chloroform/isoamyl alcohol (50:49:1) and precipitation with an equal volume of isopropanol. The calvarial RNA precipitate was resuspended in 0.6 ml of RLT lysis buffer and further purified with the RNeasy kit (QIAGEN, Valencia, CA) following the manufacturer's protocol. A NuGEN (San Carlos, CA) Ovation kit was used for amplification and labeling of calvarial mRNA following the manufacturer's protocols. Each biotin-labeled mRNA sample (10 μg) was hybridized to MOE430_2 oligonucleotide arrays (Affymetrix, Santa Clara, CA) and subsequently washed, stained, and scanned on an Affymetrix 3000 Genechip Scanner following the manufacturer's protocols. MAS5.0 software (Affymetrix) was used to convert raw fluorescence data into mRNA signal intensity and present/absent absolute calls. In a parallel experiment, we confirmed by histomorphometric analysis that incubation with compound I for 4 days increased total bone area by more than 50% and activated osteoblasts without altering the number of osteoblasts (data not shown).
Gene Set Enrichment Analysis.
Gene set enrichment analysis (GSEA) was performed with GSEA version 2.0 (Broad Institute; http://www.broad.mit.edu/gsea) (Subramanian et al., 2005). Instead of selecting single differentially regulated genes, this method analyzed the entire transcriptome data to identify genes coordinately regulated in predefined gene sets from various biological pathways. For the pairwise comparison (compound I treated versus nontreated control), GSEA was performed by using the entire filtered and normalized list of 14,955 transcripts. In brief, all genes were ranked according to the pairwise comparison between the two groups (treated versus control), then the association between a given gene set and the group was measured by the nonparametric running sum statistic termed the enrichment score (ES), which was calculated by walking down the ranked list, increasing when encountering a gene in the given gene set and decreasing when encountering a gene not in the gene set. To estimate the statistical significance of the ES, a nominal p value was calculated by permuting the gene sets 1000 times. To adjust for multiple hypothesis testing, the maximum ES was normalized to account for the gene set size (normalized enrichment score) and the false discovery rate corresponding to each normalized enrichment score was calculated. The gene sets used are from curated version 2.5 Canonical Pathways and Gene Ontology Terms. All gene symbols are official National Center for Biotechnology Information Entrez Gene symbols.
Ingenuity Pathway Analysis.
The differentially regulated Affymetrix qualifiers along with normalized fold changes and p values were uploaded into Ingenuity Pathway Analysis Software v.9.0 (http://www.ingenuity.com). In the process of Ingenuity Pathway Analysis, each network, pathway, or biological function was assigned a p score [p score = −log10 (p value)] reflecting the probability of it being generated or enriched at random, whereby p value was calculated as the right-tailed sum of the hypergeometric distribution (Fisher's exact test).
Osteoclast Differentiation and Activity Assay.
Recombinant human osteoprotegerin (OPG) was purchased from PeproTech (Rocky Hill, NJ) and used at a final concentration of 100 ng/ml; salmon calcitonin was purchased from Bachem (Bubendorf, Switzerland) and used at a final concentration of 1 to 50 nM; bafilomycin A1 was purchased from Sigma and used at a concentration of 5 nM; recombinant human macrophage colony-stimulating factor (rhM-CSF) and recombinant human receptor activator of nuclear factor κ-B ligand (rhRANKL) were both provided with the Lonza Walkersville, Inc. Bullet Kit.
Poietics human osteoclast precursors were purchased from Lonza Walkersville, Inc., seeded at 10,000 cells per well in 100 μl and cultured for 10 days in osteoclast precursor (OCP) growth medium (Lonza Walkersville, Inc.) or OCP bullet kit medium (Lonza Walkersville, Inc.) containing rhM-CSF (33 ng/ml) and rhRANKL (2–8 ng/ml) on a layer of human bone particles (OsteoAssay human bone plate; Lonza Walkersville, Inc.), according to the manufacturer's instructions. Medium was changed on day 7, and fresh reagents were added. For differentiation assays, test and standard agents were added at day 0 of culture and supernatants were removed at day 10, after one media change with OCP on day 7, for measurement of calcium release using Calcifluor (Lonza Walkersville, Inc.); for activity assays, test and standard compounds were added on day 7, after differentiation of the osteoclast precursors and with the addition of fresh OCP medium. After an additional 3 days of culture, supernatants were removed for calcium measurement by using the Calcifluor assay kit. Calcium levels in the media were determined by using a standard curve according to the manufacturer's instructions. Data were expressed as a percentage of control (rhM-CSF and rhRANKL only) values.
Tartrate-resistant acid phosphatase (TRAP)-positive multinucleate and mononucleate cells were visualized in cultures by using Sigma's kit 386A according to the manufacturer's protocol.
In Vivo Studies in Mice.
Ovariectomized (OVX) C57BL6 and intact 129SV mice were purchased from Charles River Breeding Laboratories (Portage, MI) and maintained on a rodent diet containing 0.9% calcium and 0.7% phosphorus, housed in polycarbonate boxes with α-dri bedding, with a 14- to 10-h light/dark cycle with ad libitum access to water. After a 3-day acclimation period, intact animals were dosed subcutaneously each day with compound I suspended in Tween/methylcellulose. OVX mice were maintained for 30 days before dosing to permit the development of osteopenia. For each study, doses of 0.3, 1, 3, and 10 mg/kg were used. For dynamic histomorphometry studies, mice were injected with 15 mg/kg calcein 9 and 2 days before sacrifice. Mice were euthanized by exposure to CO2, and the femora and lumbar spines were recovered, dissected free of soft tissue, and stored in 70% ethanol until analyzed. All procedures were approved by Wyeth's Institutional Animal Care and Use committee.
High-resolution microcomputed tomography (microCT, μCT) was used to evaluate trabecular volume fraction and microarchitecture in the distal femur (μCT20; Scanco Medical AG, Basserdorf, Switzerland) and the fourth lumbar vertebrae (μCT40; Scanco Medical AG) (Rüegsegger et al., 1996). The femur was scanned at 35 keV with a slice increment of 9 μm. CT images were reconstructed with an isotropic voxel size of 9 μm, and the gray-scale images were segmented by using a constrained three-dimensional Gaussian filter (σ = 0.8, support = 1.0) to remove noise, and a fixed threshold (35% of maximal gray scale value) was used to extract the structure of mineralized tissue. Scanning was started approximately at the growth plate and extended proximally for 200 slices. Morphometric analysis was performed on 135 slices extending proximally beginning with the first slice in which the femoral condyles had fully merged. The entire fourth lumbar vertebra was scanned at 55 keV, with a slice increment of 12 μm, and CT images were reconstructed with an isotropic voxel size of 12 μm. The gray-scale images were segmented by using a constrained three-dimensional Gaussian filter (σ = 0.8, support = 1.0) to remove noise, and a fixed threshold (22% of maximal gray scale value) was used to extract the structure of mineralized tissue. The trabecular bone within the vertebral body (excluding regions near the endplates) was identified by using manually drawn contouring algorithms on approximately 200 CT slices per vertebrae (∼2.5 mm of vertebral height). Morphometric parameters computed for both skeletal sites included the bone volume fraction (BV/TV; %), trabecular thickness (Tb.Th; μm), trabecular number (Tb.N; mm−1), trabecular separation (Tb.Sp; μm), and connectivity density (Conn.Den; mm−3). Tb.Th, Tb.N, and Tb.Sp were computed by using algorithms that do not rely on assumptions about the underlying trabecular structure (Ruegsegger et al., 1996; Odgaard 1997).
The distal third of left femur was separated from the proximal femur and its posterior cortex was removed to expose the marrow with a diamond-blade band saw. The specimens were dehydrated in graded concentrations of ethanol, defatted in acetone, and embedded without decalcification in methyl methacrylate. An 8.0- and 10.0-μm frontal section was cut by using a microtome (Reichert Jung Polycut S; Cambridge Instruments, Heidelberg, Germany) for histomorphometric measurements as described previously (Seeherman et al., 2004). The 8-μm sections were stained with modified Von Kossa stain. The 10-μm section remained unstained. Static and dynamic histomorphometric parameters of distal femoral trabecular bone were measured and calculated by using a computerized digital microscopy histomorphometry analysis system (OsteoMeasure; OsteoMetrics, Inc., Decatur, GA). Total tissue area, trabecular bone area, and trabecular bone perimeter were measured from the Von Kossa-stained sections. Trabecular bone volume, trabecular number, trabecular thickness, and trabecular separation were calculated as described previously (Seeherman et al., 2004). Single-calcein-labeled perimeter, double-calcein-labeled perimeter, and interlabel width were measured on the unstained sections under fluorescent light. These data were used to calculate the percentage of labeled trabecular surface, mineral apposition rate, and bone formation rate surface as described previously (Seeherman et al., 2004).
All results are expressed as the mean ± S.E. Data were analyzed for significance by using analysis of variance and the method of least significant difference, or Dunnett's test, with or without transformation, as appropriate.
Characterization of Src and Abl Activities of Compound I.
Src inhibitors of the 4-anilino-3-quinolinecarbonitrile class are also potent Abl kinase inhibitors; however, the 4-anilino-7,8-dialkoxybenzo[g]quinoline-3-carbonitrile class tended to be less active in proliferation assays of chronic myelogenous leukemia cells (Berger et al., 2005), which are sensitive to Abl kinase inhibitors.
Several compounds of this class, including compound I (Fig. 1), were tested in enzyme assays to compare Src and Abl kinase inhibitory activities. Compound I was a potent (0.55 nM) inhibitor of Src in isolated enzyme assays with moderate selectivity over Abl (3.37 nM) (Table 1). Compound I did not inhibit less related kinases such as B-raf, glycogen synthase kinase 3 β, c-Jun N-terminal kinase 3, inhibitor of nuclear factor κ-B kinase subunit, and Akt when tested at concentrations up to 10 μM (data not shown).
Compound I inhibited growth in Src-based cellular assays, both the 32D cell-based assay (IC50 = 36 nM) and the fibroblast-based cell assay (IC50 = 46 nM). Unlike the nearly 6-fold selectivity observed for Src over Abl in the enzyme assay, there was only moderate selectivity (approximately 2-fold) for Src over Abl in the transformed 32D cells. Compound I was also potently active in cell-based assays whose growth depended on other Src family members, i.e., Lyn and Lck, but with apparently weaker activity on a Fyn-dependent assay (Table 1).
In Vitro Activity of Compound I on Osteoblast Differentiation and Bone Formation.
Compound I had biphasic effects on the differentiation of hMSC to an osteoblast phenotype (Fig. 2). At 1 to 10 nM, compound I induced an approximately 1.5-fold increase in alkaline phosphatase expression, whereas at 1 to 5 μM compound I inhibited differentiation up to 75%.
Compound I was potently osteogenic in the in vitro mouse calvarial bone formation assay (Fig. 3A). Compound I induced significant increase in total bone area at doses as low as 0.1 nM, with an efficacy of approximately 1.5-fold over control. In limited exposure experiments, exposure of calvariae to 0.1 μM compound I for only 1 day, at the start of the culture, was sufficient to induce increase in total bone area (Fig. 3B). At active concentrations, periosteal osteoblasts seemed plumper (Fig. 3, D–F), but despite an apparent accumulation of cells on the surface of the bone with compound I treatment (Fig. 3, D–F) there was no consistently significant increase in osteoblast number. Osteogenic activity was also noted in the calvarial assay with several other small-molecule Src inhibitors (data not shown).
Activity of Compound I in Osteoclast Differentiation and Activity Assays.
Standard osteoclast-active agents behaved as expected in this assay. Differentiation of osteoclasts induced by rhM-CSF and rhRANKL was inhibited by the RANKL antagonist OPG (Fig. 4A), and the activity of osteoclasts was inhibited by the proton pump inhibitor bafilomycin (Fig. 4B). In addition, in a single experiment, salmon calcitonin had no effect on osteoclast differentiation but inhibited osteoclast activity over 24 h but not 72 h, consistent with the well known “escape” phenomenon (data not shown). Morphologically, TRAP+ multinucleate cells were induced in large numbers by treatment with rhM-CSF and rhRANKL on the particle matrix, similar to those induced on plastic (Fig. 4, C and D).
Compound I inhibited osteoclast differentiation and activity but, in contrast to osteoblast-based assays, required a relatively high concentration of 100 nM (Fig. 4, A and B). At this concentration it reduced osteoclast differentiation by 82 to 92% and activity by 65 to 82% (Fig. 4, A and B). Lower concentrations of 10 and 1 nM were inactive. TRAP+ multinucleated cells were largely absent in cultures cotreated with 100 nM compound I according to the differentiation protocol (Fig. 4E), but were clearly present in cultures cotreated with 100 nM compound I according to the activity protocol (Fig. 4F), suggesting an effect on osteoclast differentiation rather than a toxic effect on mature osteoclasts.
Transcriptional Profiling of Neonatal Mouse Calvariae Treated with Compound I.
To identify transcriptional biomarkers of drug activity, mRNA recovered from calvariae treated 4 days with 1 μM compound I was globally profiled by using mouse MOE430_2 DNA arrays (Affymetrix). Of the tiled oligonucleotide probes, 16,250 individual probe sets were called present in all three samples of one or both of the control or drug-treated cohorts. Filtering the signal intensity for mean fold changes more than 2-fold (p < 0.05) revealed compound-related modulation in the expression of 436 genes (238 up-regulated, 198 down-regulated), of which a selection is shown in Table 2. At a 1.5-fold (p < 0.05) cutoff, there were 1085 genes modulated (642 up-regulated, 443 down-regulated) as shown in Supplemental Table 1. There was a clear down-regulation of the osteoclast-associated genes tartrate-resistant acid phosphatase (ACP5), the M-CSF receptor CSF1R, and the proton pump regulatory subunit ATP6V0D2. Several collagens were up-regulated as well as matrix proteins such as nephronectin, asporin (osteomodulin) (ASPN), and osteoglycin (OGN), proteins that either regulate, or are associated with, osteoblast differentiation and mineralization. An increase in Wnt antagonist expression (APC, DKK3, SFRP4) was also noted (Table 2). Several genes related to G protein-coupled receptor signaling were altered by compound I treatment, particularly those downstream of G proteins (Table 2). We also looked for indications that compound I may be acting by influencing Src-mediated YAP regulation of Runx2, as reported previously (Zaidi et al., 2004). Compound I induced up-regulation of YAP and also regulated the Runx2 target genes SOX9 and pleiotrophin (PTN) (Table 2). Two potentially osteogenic extracellular mediators, β-CGRP (Fig. 5A) and insulin-like growth factor 1, were up-regulated by compound I treatment (Table 2). Testing of β-CGRP confirmed its osteogenic properties in our calvarial assay system (Fig. 5B),
Activity of Compound I in In Vivo Mouse Bone Models.
In intact female mice dosed 30 days with 1 mg/kg compound I subcutaneously, μCT of the fourth lumbar vertebra revealed that the BV/TV was increased 21% and connectivity density increased 22%, associated with a thickening of trabeculae (increased 4.8%) and a reduction in the BS/BV (8.6%) (Table 3). Increased bone volume was also detected histologically in a different bone, the femur, the most pronounced effects being in the secondary spongiosa, where trabecular volume was increased 100%, associated with a 71% increase in trabecular number and a 17% increase in trabecular thickness (Figs. 6 and 7).
In OVX mice, the 1 mg/kg dose of compound I also increased BV/TV of the fourth lumbar vertebra by 32.9% and connectivity density by 24.5% (not significant), which was associated with a thickening of trabeculae (increased 13.9%) and a reduction in BS/BV (decreased 11.3%) (Table 4). Histology of the secondary spongiosa of the distal femur measured a 41% increase in trabecular volume (p < 0.06) (Fig. 7).
Time-course analysis of OVX mice indicated that the “increases” measured in the compound I-treated mice were approximately the same values as those found in the OVX mice at the start of treatment (Fig. 8), whereas values noted in the literature with the known anabolic agent PTH are larger (Bodine et al., 2007). This raised the possibility that the Src inhibitor compound I was merely acting to prevent bone loss rather than having a genuine anabolic effect. However, histomorphometric measurement of calcein-labeled bones clearly showed significant increases in mineral apposition rate (59.1% in intact mice, 74.8% in OVX mice), bone formation rate (171.3% in intact mice, 98.2% in OVX mice), and mineralizing surface (73.2% in intact mice) (Fig. 9), strongly suggesting that the increases in bone mass observed with compound I were caused by a genuine bone anabolic effect.
Although the bone anabolic effects of Src deletion have been known for many years (Marzia et al., 2000), this, to our knowledge, is the first report of an osteogenic effect of a small-molecule Src inhibitor. Compound I is a potent (1 mg/kg) compound on bone that increases bone formation rate by as much as 171% in intact mice. In vitro data from the calvarial assay and the hMSC differentiation assay revealed that this compound potently enhanced the differentiation and/or activation of osteoblasts, at doses of 0.1 and 1 nM, respectively. Although compound I inhibited osteoclast differentiation and activity, as would be expected of a Src inhibitor, this required a dose of 100 nM, much higher than those required for the stimulation of osteoblast differentiation and calvarial bone formation. The histomorphometric evidence of increased bone formation in vivo and the relatively less potent inhibition of bone resorption we observed in vitro suggests that the increase in bone we observed at 1 mg/kg compound I was most likely caused by a genuine osteogenic effect.
It should be noted, however, that this effect was seen at only a single dose, both lower and higher doses being without significant anabolic effect. This narrow therapeutic window, when first observed using peripheral quantitative computed tomography of the femurs of the intact SV129 model (data not shown), initially made us skeptical of the results and prompted us to examine bones from an additional site (the spine) and use additional detection methods (μCT and histology). Furthermore, we repeated the experiment in a model of established osteopenia, the OVX C57BL6 mouse, which had been allowed to lose bone several weeks after ovariectomy before dosing. In each case, similar results were obtained, suggesting that we were indeed observing a real and reproducible effect. A possible explanation for the loss of the effect with increasing dose could be the inhibition of other kinases such as Abl or Lyn, whose osteopenic knockout phenotypes (Li et al., 2000; Kim et al., 2009) would predict a negative effect on bone upon pharmacological inhibition. Furthermore, it is perhaps relevant that compound I had biphasic effects in the hMSC assay (Fig. 2), which if reproduced in vivo could explain the narrow therapeutic window.
A limitation of this study is that we cannot be sure that the osteogenic effect is caused solely by inhibition of Src. We know from knockout studies that the absence of the Src protein will result in a dual antiresorptive and osteogenic phenotype. Compound I is a potent inhibitor of Src in an enzymatic assay, and we have also shown that compound I is capable of inhibiting cell growth in two separate lines whose growth is driven by Src, confirming that compound I is capable of inhibiting Src within a cell. However, compound I is also a potent inhibitor of Lck and might be expected to inhibit, to some degree, other Src family kinases and perhaps other kinases whose role in osteoblast biology may be unknown. This is important because it has been shown that certain double/triple ablations of Src family kinase members (e.g., Hck−/− and Src−/−) result in an accentuated bone phenotype (Lowell and Soriano 1996). Moreover, other ATP-competitive kinase inhibitors such as the dual Src/Bcr-Abl inhibitor dasatinib and Bcr-Abl inhibitor imatinib have been found to exert significant effects on the skeleton via inhibition of one or more “off-target” kinases, including c-fms, platelet-derived growth factor receptor, and c-kit (Brownlow et al., 2009; Vandyke et al., 2010). A thorough investigation of the off-target effects of compound I on skeletal cell kinases, beyond the scope of the present report, would be required to definitively identify the kinases whose inhibition is responsible for the osteogenic effect of this compound.
We performed transcriptional profiling on neonatal mouse calvariae treated with compound I in an attempt to shed light on its mechanism of action. Results from individual gene chip studies can be variable, particularly at the level of the individual gene (Roman-Roman et al., 2003), so we have interpreted our data cautiously, highlighting only those processes where several contributing genes appear in our dataset. It should also be noted that we used a high concentration (1 μM) of compound I that probably inhibited several other enzymes, particularly other Src family kinases. Our data showed that several bone resorption-associated genes such as tartrate-resistant acid phosphatase (ACP5), CSF1R, and the regulatory subunit of the proton pump (ATP6V0D2) were down-regulated, and genes associated with the deposition of extracellular matrix were up-regulated, suggesting that compound I exerted a combined antiresorptive and osteogenic effect on the calvariae, as would be expected of a Src inhibitor from the knockout data.
Our data also revealed an increase in gene expression associated with osteoblasts or osteoblast differentiation. We saw induction of several Wnt antagonists (APC, DKK3, and SFRP4), which others have found to be characteristic of the late stages of osteoblast differentiation (Vaes et al., 2005). We also saw increases in the expression of genes such as IGFBP4, PTN, COL2A1, and ASPN that others have found to be associated with the phenotype of murine calvarial osteoblasts or osteocytes (COL5A1, COL22A1) (Paic et al., 2009). In addition, several genes were regulated by compound I (see Supplemental Table 1) in the opposite direction to that observed in microarray studies on osteoblasts subjected to microgravity (Pardo et al., 2005; Patel et al., 2007); ASPN, OGN, GADD45G, COL11A1, MMP16, ADAMTS5, DDR1, and ISLR all were up-regulated by compound I but down-regulated by microgravity; SEPHS2, COL4A1, COL4A2, ITGA6, and TUBA4 all were down-regulated by compound I but up-regulated by microgravity. We do not, however, have any explanation at this time why we did not detect up-regulation of some other typical osteoblastic genes, such as bone sialoprotein. Overall, however, our microarray data add support at the level of gene expression to our cell- and tissue-based studies, implying a stimulation of osteoblast differentiation and/or activation.
The only proposed molecular mechanism for Src's role in the osteoblast in the literature is phosphorylation of YAP, which thereby inhibits Runx2-mediated gene transcription (Zaidi et al., 2004). We found two Runx2 target genes to be regulated by compound I, and YAP itself was also regulated, which may be consistent with such a role for Src on this pathway, but is not of itself conclusive. We did note perturbation of several genes whose products modulate downstream GPCR signaling. Negative regulators of signaling such as β-arrestin 1 (ARRB1), RGS2, RGS5, CAMK2N, CAMKK2, DGKZ, and INPP5D all were down-regulated by compound I, suggesting that treatment results in a priming of GPCR pathways. β-CGRP, a known GPCR agonist, was up-regulated by compound I, and we were able to demonstrate its osteogenicity in our calvarial assay. However, further work is clearly necessary to elucidate the precise mechanism whereby compound I exerts its osteogenic effects.
It is worth noting that our selection of compound I for in vivo studies was based not only on its selectivity for Src over Abl, but also by its highly potent activity in the calvarial bone formation assay, where it was active at extremely low doses (0.1 nM), even lower than its IC50 in the enzymatic assay, and two orders of magnitude more potent than its activity in the 32D or fibroblast cell assays. It is more usual for IC50 values of ATP-competitive kinase inhibitors in cell- and tissue-based assays to be weaker than their IC50 values in isolated enzyme assays, in part because of the low concentration of ATP(10–20 μM) in typical isolated enzyme assays, relative to the high intracellular concentrations of ATP (1 mM or more) present in cells. The greater than expected potency of compound I in the calvarial assay is still unexplained but could conceivably be caused by tissue accumulation, leading to a higher local concentration of the compound in the bone.
The only currently approved anabolic therapy for osteoporosis is teriparatide (Forteo, Eli Lilly & Co., Indianapolis, IN) (PTH 1–34), and an identical peptide to teriparatide, hPTH(1–34), was used as a comparator in this study. In general, the effects of PTH were greater than those of compound I, perhaps reflecting the role of PTH as a master regulator of bone metabolism. However, the differential between PTH and compound I varied from site to site, being greatest in the distal femur (peripheral quantitative computed tomography studies; data not shown) but less so in the vertebra and secondary spongiosa, where compound I was comparable with PTH. This may reflect envelope-specific differences in the importance of PTH-induced and Src-related pathways.
In summary, we have identified a highly potent Src inhibitor with significant osteogenic activity in two mouse models. This not only supports the hypothesis that Src plays an inhibitory role in osteoblast biology, but also opens up the possibility of treating diseases involving bone loss with an osteogenic Src kinase inhibitor.
Participated in research design: Murrills, Fukayama, F. Boschelli, Patel, Carter, D. H. Boschelli, McKew, Li, Kharode, Follettie, Bex, Komm, and Bodine.
Conducted experiments: Fukayama, Matteo, Owens, Golas, Patel, Lane, Liu, Carter, Spaulding, Wang, D. H. Boschelli, Milligan, Kharode, and Diesl.
Contributed new reagents or analytic tools: Owens, Carter, Jussif, and Wang.
Performed data analysis: Murrills, Fukayama, Matteo, Owens, Carter, Spaulding, McKew, Li, Lockhead, Milligan, Kharode, Diesl, and Bai.
Wrote or contributed to the writing of the manuscript: Murrills, Fukayama, F. Boschelli, D. H. Boschelli, McKew, Li, Kharode, Bai, Bex, Komm, and Bodine.
We thank Wan-Jiang Wei for performing neonatal mouse calvarial experiments and Thomas R. Murrills for helping with the charts and graphics.
Article, publication date, and citation information can be found at http://jpet.aspetjournals.org.
- receptor activator of nuclear factor κ-B
- recombinant human RANK ligand
- dimethyl sulfoxide
- bovine serum albumin
- computed tomography
- recombinant human macrophage colony-stimulating factor
- human mesenchymal stem cells
- 4-methylumbelliferyl phosphate
- tartrate-resistant acid phosphatase
- analysis of variance
- parathyroid hormone
- calcitonin-related polypeptide, β
- Src family kinase
- gene set enrichment analysis
- enrichment score
- osteoclast precursor
- bone volume
- total volume
- bone surface
- trabecular thickness
- trabecular number
- trabecular separation
- connectivity density
- G protein-coupled receptor.
- Received July 5, 2011.
- Accepted December 6, 2011.
- Copyright © 2012 by The American Society for Pharmacology and Experimental Therapeutics