Inhibition of ALK5 Signaling Induces Physeal Dysplasia in Rats

Abstract

TGF-|β|, and its type 1 (ALK5) receptor, are critical to the pathogenesis of fibrosis. In toxicologic studies of 4 or more days in 10-week-old Sprague–Dawley rats, using an ALK5 inhibitor (GW788388), expansion of hypertrophic and proliferation zones of femoral physes were noted. Subphyseal hyperostosis, chondrocyte hypertrophy/hyperplasia, and increased matrix were present. Physeal zones were laser microdissected from ALK5 inhibitor-treated and control rats sacrificed after 3 days of treatment. Transcripts for TGF-|β|1, TGF-|β|2, ALK5, IHH, VEGF, BMP-7, IGF-1, bFGF, and PTHrP were amplified by real-time PCR. IGF and IHH increased in all physis zones with treatment, but were most prominent in prehypertrophic zones. TGF-|β|2, bFGF and BMP7 expression increased in proliferative, pre- and hypertrophic zones. PTHrP expression was elevated in proliferative zones but decreased in hypertrophic zones. VEGF expression was increased after treatment in pre- and hypertrophic zones. ALK5 expression was elevated in prehypertrophic zones. Zymography demonstrated gelatinolytic activity was reduced after treatment. Apoptotic markers (TUNEL and caspase-3) were decreased in hypertrophic zones. Proliferation assessed by Topoisomerase II and Ki67 was increased in multiple zones. Movat stains demonstrated that proteoglycan deposition was altered. Physeal changes occurred at doses well above those resulting in fibrosis. Interactions of factors is important in producing the physeal dysplasia phenotype.

Introduction

Transforming growth factor-beta (TGF-β) proteins and their receptors are important components of the intracellular signaling pathways involved in the pathogenesis of fibrosis. TGF-β plays a critical role in up-regulation of extracellular matrix associated proteins and collagen synthesis in diseases as diverse as chronic renal failure, hepatic cirrhosis, and scleroderma. (Border and Noble, 1994) TGF-β modulates matrix degradation and synthesis by a complex series of effects including down-regulation of several of the matrix metalloproteinases (MMP), up-regulation of tissue inhibitor of matrix metalloproteinase (TIMP), and profound up-regulation of plasminogen activator inhibitor (PAI-1) (Nakamura et al., 1992; Border and Noble, 1994). TGF-β is also involved in cell cycle regulation, embryo patterning and bone and cartilage morphogenesis.

There are multiple, closely related type I TGF-β receptors, the activin-like kinase (ALK) receptors—including ALK1, ALK4, ALK5, and ALK8—each with their own preferential ligands (Laping et al., 2002; Laping, 2003). ALK5 binds TGF-β1 with high affinity, but TGF-β signalling additionally requires co-binding of the type II receptor (Laping et al., 2002). Activated ALK5 then phosphorylates Smad proteins, which mediate intracellular signaling to the nucleus (Abdollah et al., 1997). Many drug companies have compounds targeting various points of this signaling cascade, including TGF-β activation/binding, receptor interactions and intracytoplasmic Smad signal transduction proteins. Using ALK5 inhibitors, recent studies have demonstrated that ALK5 activity and subsequent nuclear translocation of the signalling proteins Smad2 and Smad3 are directly responsible for activation of genes that participate in the accumulation of collagen and extracellular matrix proteins (Laping, 2003; Callahan et al., 2002). ALK5 inhibition has therefore gained interest as a therapeutic target in a wide variety of fibrotic diseases (Laping, 2003).

We recently reported on the antifibrotic effects of GW788388, a novel phenylpyridine pyrazole, that has been shown in vivo and in vitro to prevent overexpression of collagen IA1, a major component contributing to excessive extracellular matrix deposition in fibrotic disease (Gellibert et al., 2006). While performing preclinical rat toxicity studies with GW788388 and other inhibitors of the ALK5 type domain of the TGFβ receptor complex, we consistently noted expansion of the zones of hypertrophy and proliferation of the femoral physes. A characteristic bone lesion involving hypertrophy of the femoral physes was present in many of the animals tested at high doses (>300 mg/kg/day) but varied in intensity of phenotype among compounds.

Although the femur is referred to as the most marked change, other physes, including the tibia, were affected in longer-term studies; similar observations were noted in the mouse and dog (data not shown). We carefully assessed changes in TGFβ-associated osteoregulatory cytokine expression profiles in GW788388 treated rat physes by laser capture microdissection (LCM), Taqman real-time PCR, in parallel with in situ zymography and immunohistochemistry. These experiments were designed to explore the mechanism of physeal dysplasia and whether this involved dysregulation of cytokines and proteins that are modulated via the ALK5 (TGF-β receptor type I) pathway. This information on mechanism would then be utilized for risk assessment and safety evaluation and to put the lesion in proper perspective for potential human therapeutic clinical trials.

Materials and Methods

Test Compounds

GW788388 was synthesized by GlaxoSmithKline Pharmaceuticals, Les Ulis, France. Dosing suspensions were prepared in aqueous 0.5% methylcellulose at concentrations of 0.5 to 1 mg/ml.

In Vivo Studies

GW788388, were orally administered to 10-week or 9-month-old Sprague–Dawley rats for 4–10 days at doses of 100–1000 mg/kg/day. Mean AUC_t–24 was determined from blood samples prior to necropsy (data not shown). Necropsy was performed 24 hours after the final dose and tissues processed routinely for histopathology. In an additional time-course experiment, 10-week-old Sprague–Dawley rats were given 300 mg/kg/day GW788388 for 4 days. Six rats were necropsied on Day 3 prior to the third dose and 6 rats were necropsied on day 5, 24 hours after the last dose.

Compounds were also evaluated in the purine aminonucleoside nephrosis (PAN) rodent model (Grygielko et al., 2005). Briefly, Sprague–Dawley rats were pretreated by oral gavage with 1,3, or 10 mg/kg/day GW788388 once daily. The following day, purine aminonucleoside was injected at 15 mg/100 g. Treatment groups continued to receive GW788388 for 10 days and then were euthanized. Blood, urine, and kidneys were collected and and kidney mRNA and collagen quantitated exactly as previously described (Grygielko et al., 2005). An additional experiment was conducted in which 300 mg/kg of GW788388 or vehicle control were administered once daily for 4 days to 12 10-week-old Sprague–Dawley rats. Six rats were necropsied at 3 days just prior to the third dose, and 6 rats were necropsied on day 5, which was 24 hours after the last dose with equal numbers of controls at the same time points.

Histology

Femurs were collected at necropsy, fixed in 10% neutral-buffered formalin, trimmed, decalcified in Cal-Rite (Richard Allen Scientific, Kalamazoo, MI) and routinely processed through graded alchohols and xylenes. After paraffin embedding, 5-micron sections were placed onto glass slides for hematoxylin and eosin (H&E) staining. One femur was collected from rats given vehicle or 300 mg/kg/day GW788388, fixed in 70% alcohol, processed, embedded in methyl methacrylate, sectioned, and stained with modified Movat’s pentachrome according to the methods of Schenk (Schenk et al., 1984). In the 4-day time-course study, 1 femur was collected, embedded in OCT (Miles Inc., Elkhart, Ind.) and snap-frozen in liquid nitrogen. Five-μm cryosections were prepared on 4X adhesive-coated (Instrumedics, Hackensack, NJ) using a Leica CM3050 S cryostat equipped with a Cryo-Jane Tape Transfer system (Instrumedics), then immediately stored at −80°C. Cryosections were used for laser capture microdissection (LCM) and for assessment of growth plate mineralization on Von Kossa stained sections.

Laser Capture Microdissection, RNA Isolation, and cDNA Synthesis

Upon removal of stored cryosections from −80°C, sections were fixed in 70% ethanol, dehydrated in graded ethanol/xylene, dessicated, then microdissected onto CapSure Transfer Film Carriers (Arcturus, Mountain View, CA) using a PixCell IIe (Arcturus) LCM system. Several LCM caps (4 rats/sample) from each enriched zone—proliferation, prehypertrophy, hypertrophy, perichondrium—were pooled into separate tubes containing RNA lysis buffer (Stratagene, La Jolla, CA). Total RNA was extracted using the Absolutely RNA Microprep Kit (Stratagene) and quantitatively assessed using the QuantiT RiboGreen Kit (Molecular Probes, Eugene, OR). cDNA was synthesized using the iSCRIPT (Bio-Rad, Hercules, CA) cDNA synthesis kit.

Real-Time PCR (TaqMan):

Transcripts for TGF-β1, TGF-β2, ALK5, Indian hedgehog (IHH), vascular endothelial growth factor (VEGF), bone morphogenetic protein-7 (BMP-7/OP-1), insulin-like growth factor-1 (IGF-1), basic fibroblast growth factor (bFGF), and parathyroid hormone-related peptide (PTHrP) were amplified by real-time PCR. Primer express software (Applied Biosystems, Foster City, CA) was used to design primers and fluorogenic probes from published sequences for selected genes. The 25 μL Taqman reactions consisted of 1X TaqMan Universal Master Mix, 10 ng cDNA, and 400 nM each forward primer, reverse primer, and probe. Fifty cycles of real-time PCR were run and data was collected on the ABI Prism 7900HT Sequence Detection System (Applied Biosystems). Standard curves were constructed for each gene to verify primer and probe efficiency. Relative values for each gene from each zone were normalized to expression of 18s rRNA, then calibrated and expressed as Relative Fold (Figure 3) using the standard curve method described in Applied Biosystems User Bulletin #2. A student’s t-test (1-tailed, unequal variance) was performed and p values denoting a change in gene expression between control and treated regions/zones of less than or equal to 0.05 were considered statistically significant

In Situ Zymography

To characterize effects of drug treatment on gelatinase (MMP-2 and MMP-9 activity within physeal cartilage, 8-μm-thick cryosections of femorotibial joints were prepared and sectioned as described here, and in situ zymography was performed using previously described methods (Mook et al., 2003; Fredericks and Mook, 2004). Unfixed cryostat sections (8 μm thick) were air-dried for 10 minutes. FITC-labeled DQ-gelatin (EnzChek; Molecular Probes, Eugene, OR) was dissolved to a concentration of 1 mg/ml in water and then diluted 1:10 in 1% low gelling temperature agarose (Sigma) in PBS. Then, 50 μL of FITC DQ-gelatin mixture was put on top of the cryostat section and covered with a glass coverslip. After gelling at 4°C, sections were incubated at 37°C overnight. Fluorescence was visualized with a Nikon Eclipse E-800 photomicroscope (Nikon, Melville NY) equipped with fluorescent attachments and suitable excitation/barrier filter system (460–500 nm and 512–542 nm, respectively). Specificity of gelatinolysis due to MMP activity was determined by adding 20 nM EDTA to the FITC DG-gelatin/agarose mixture prior to incubation on tissue section.

Immunohistochemistry

Treatment-related effects on physeal chondrocyte proliferation and apoptosis were evaluated qualitatively by immunohistochemistry using monoclonal Ki-67 (MIB-5 clone) and Topoisomerase II (TopoII, DakoCytomation, Carpenteria, CA), as proliferation markers and antibodies to activated caspase-3 (DakoCytomation) for detection of apoptosis. TUNEL assay was included as a second in situ marker of apoptosis (APOTAG KIT). Formalin-fixed cryostat sections taken from rat femoral physes in the toxicologic study described here were stained for Ki-67, TopoII, TUNEL and caspase-3 as previously described using either Dakocytomation Autostainer or Ventana Discovery Staining Module (Hoivik et al., 2004; Olzinski et al., 2005). Appropriate controls included deletion of the primary antibody secondary antibodies or streptavidin-horseradish peroxidase complex, substitution of primary antibody with species isotype/concentration matched control immunoglobulin, and deletion of DNA polymerase (TUNEL).

Results

After 4 days of treatment with the ALK5 inhibitor GW788388 in 10-week-old rats, dose-dependent thickening of the femoral physis was noted microscopically. In rats given 300 mg/kg/day, changes were minimal, consisting of increased chondroid matrix, increased numbers of chondrocytes in the proliferative zone, and slight increases in the thickness of the hypertrophic layer (Figure 2). Severity of physeal changes increased with duration of dosing progressing from minimal to moderate in rats given 300 mg/kg/day for 10 days accompanied by subphyseal hyperostosis (Figure 1). In contrast, physeal lesions were not evident in 9-month-old rats treated similarly, and the femoral physes were similar to controls (where growth plates had not yet closed). The physeal lesion was not noted in any studies with 10-week-old rats at doses below 100 mg/kg, but higher doses corresponded with more severe physeal dysplasia, hypertrophy, and subphyseal hyperostosis. These observations suggest that physeal changes are dose-, duration- and age-dependent and progressive. Age dependency is likely linked to rates of longitudinal growth.

Hypertrophic physeal lesions have been found in previous 10-day toxicity studies with several other ALK5 inhibitors, but with variable expression dependent on individual compound potency and exposure achieved (data not shown). In acute rodent models of liver and renal disease, antifibrotic effects of GW788388 were noted at doses of 2–3 mg/kg/day, and included decreases in collagen mRNA (Gellibert et al., 2006). Exposures similar to those utilized in these experiments are projected to be clinically efficacious. Since physeal lesions have not been noted below 100 mg/kg, even in longer-term studies, the therapeutic index in regards to physeal dysplasia involves a relatively high multiple. Several other ALK5 inhibitors of different chemical series were also tested in the PAN model, and those with evidence of pharmacological activity in this model also demonstrated similar physeal lesions (data not shown).

Drug treatment resulted in zone-specific alterations in gene expression in physeal cartilage detected in LCM samples by real-time quantitative PCR (Figure 3). These included increases in IGF-1 (3.2-fold, p = 0.01) and IHH (5.8-fold, p = 0.001) in the perichondrium, increases in PTHrP (2.3-fold, p = 0.047), bFGF (3.8-fold, p = 0.0025), IGF-1 (2.4-fold, p = 0.04), IHH (2.6-fold, p = 0.001), BMP7 (2.7-fold, p = 0.03), and TGF-β2 (3.5-fold, p =0.001) in the resting/proliferating zone, increases in bFGF (8.8-fold, p = 0.001), IGF-1 (3.6-fold, p = 0.03), IHH (2.7-fold, p = 0.03), TGF-β2 (3.3-fold, p = 0.07, not statistically significant), and VEGF (3.2-fold, p = 0.01) in the prehypertrophic zone, and increases in IGF-1 (1.8-fold, p = 0.11, not statistically significant), BMP7 (6.6-fold, p = 0.0008), IHH (1.9-fold, p = 0.18, not statistically significant), and VEGF (2.7-fold, p = 0.049) in the hypertrophic zone (Figure 3). PTHrP was reduced to undetectable levels and TGF-β1 (7.0-fold, p = 0.00005) was significantly decreased in the hypertrophic zone. By day 5, (24 hours after the final dose of GW788388) gene expression patterns were generally similar to control values (data not shown), with the exception of both bFGF and ALK5, which remained up-regulated.

Von Kossa stained undecalcified cryosections were used to compare the degree of mineralization in the physis between ALK5 inhibitor treated rats and controls (Figure 4). Although the hypertrophic zone was expanded and unmineralized in drug-treated rats, the mineralization front was located in the distal hypertrophic zone in both control and treated rats in association with chondrocyte degeneration and matrix maturation. Subphyseal hyperostosis, morphologically consistent with retention of cartilage cores, contributed to the increase in mineralized matrix noted in rats treated with GW788388 (Figures 1 and 4).

There was no visible effect on mineralization of the cortical bone or metaphyseal trabeculae. Most noteable in Movat stained plastic sections were alterations in proteoglycan composition in the hypertrophic zone of drug-treated rats. The alcian blue component of the Movat procedure revealed abundant teal green staining within the thickened hypertrophic zone, in contrast to expected sea green staining representing mineralized cartilage in controls (Figure 5). Increased numbers of saffron (pink)-positive islands of matrix were also noted in physes from drug-treated rats. Staining patterns in other zones of the physeal cartilage were similar in controls and drug-treated rats.

A pronounced difference in gelatinase activity in physes from drug-treated and control rats was noted using in situ zymography. Gelatin-FITC is cleaved by gelatinases, yielding peptides whose fluorescence is representative of net proteolytic activity. Control physes demonstrated gelatinase activity corresponding to areas of chondrocytic lacunae that was abolished by 20 mM of EDTA. Reduced proteolytic activity was observed in the proliferative, prehypertrophic, and hypertrophic zones of treated rat physes when compared with controls (Figure 6). Gelatinase activity was unaffected in the proliferative zone.

Immunohistochemical stains for Ki67 (MIB5) and TopoII were similar and demonstrated a mild qualitative increase between in Ki67- and TopoII- positive nuclei in prehypertrophic zones in drug-treated rats; slight increases in numbers of positively labeled chondrocytes were also found in the hypertrophic and proliferative zones (Figure 7). In the hypertrophic and pre-hypertrophic zones in rats treated with GW788388, immunoreactivity for activated caspase-3 and positive TUNEL staining were decreased (Figure 8) consistent with reduced apoptosis and completion of final stages of endochondral ossification.

Discussion

In order to put the rodent physeal findings in proper context as far as clinical relevance, it was necessary to ascertain whether physeal dysplasia induced by the ALK5 inhibitors was due to class-wide pharmacology (at suprapharmacologic doses) or structurally based toxicology. Several lines of evidence support a pharmacologic mechanism:

1. Physeal dysplasia has been found in all toxicologic studies performed to date, with ALK5 inhibitors from different chemical series where adequate exposures have been achieved, except in one instance in which the compound was later found to lack pharmacologic activity in in vivo studies.

2. The degree of physeal dysplasia found at these toxicologic doses appears to be correlated with achieved AUC and potency among ALK5 inhibitors from different chemical series (data not shown).

3. Several previous studies have demonstrated that the TGF-β receptor is strongly expressed in the hypertrophic zone and zone of proliferation of the rat physis (Matsunaga et al., 1999) and TGF-β has well characterized effects on chondrocytic proliferation, differentiation, and maturation (Nasatzky et al., 2000; Alvarez et al., 2001; Yang et al., 2001).

4. TGF-β/TGF-βr knockout mice have physeal and limb abnormalities including thickened hypertrophic zone and disorganization of resting and proliferating zones (Serra et al., 1997), and other TGF-β receptor inhibitors have been associated with physeal expansion (Datto et al., 1999).

5. We demonstrated marked changes in the expression patterns of genes that are known to be regulated by TGF-β and that all play critical roles in physeal morphology and function.

Taken together, these results strongly suggest that the physeal lesion is due to a pharmacologically based mechanism involving TGF-β paracrine signaling pathways. It should be noted that the physeal effect occurred at suprapharmacologic doses far above those necessary to inhibit fibrosis in rats. Based on Von Kossa staining, there is no impairment of mineralization in drug-treated rats. Cotical and metaphyseal cancellous bone qualitatively are unaffected by GW788388 treatment, with effects limited to chondroid and osseous elements in and around the physis. Further, results of administration of GW788388 to 9-month-old rats where longitudinal growth is minimal suggests that the effects of treatment are significantly decreased in aged animals. Since the physeal lesion occurs at doses many-fold above the projected pharmacologically active dose in humans, and appears to be limited to actively remodeling physes with active endochondral ossification, the ALK5 inhibitor-induced physeal lesion should not pose a significant health or safety risk in adult human patients given appropriate clinical monitoring. Our data thus provides a framework for future clinical investigations of safety and tolerability of ALK5 inhibition therapy in human populations for indications such as chronic renal disease, cirrhosis, or cancer, where the target patient profile consists of adults in which physes are already closed.

LCM has only recently been utilized successfully in analyzing cytokine expression in growth plate cartilage from rodents (Landis et al., 2003; Jacquet et al., 2005). Isolation and quantitation of mRNA from laser-microdissected growth plates presents several technical challenges. There is a limited quantity of RNA available in growth plates due to low cellularity and preponderance of connective tissue matrix, and the specific cellular target—chondrocytes—are undergoing hypertrophy and death. Additional technical challenges involve preparation of quality undecalcified cryosections suitable and amenable to microdissection and other analytic techniques. Osteoregulatory genes of interest were of unknown abundance in the growth plate of either control or treated animals. We utilized the iSCRIPT cDNA Synthesis Kit (Biorad, Hercules, CA) using a mixture of oligo-dT and random primers (superior to gene specific reverse priming) to improve yields and allow smaller volume PCR reactions. The Standard Curve R² values were excellent, ranging between 0.997 and 1.000.

Chondrocyte maturation at the level of the physis is an ordered process, with the growth plate divided into discrete zones. These include the resting or reserve zone, proliferative, prehypertrophic and hypertrophic zones, and ending at the metaphysis or zone of mineralization. Growth factors secreted by chondrocytes or perichondrium coordinate the rate of long bone growth and modulate the process of endochondral ossification. These include TGF-β, PTHrP, bFGF, IGF1 and IHH (Tchetina et al., 2003; Van der Eeden et al., 2003). For chondrocytes to proliferate and enlarge, cleavage of collagens and remodeling of the extracellular matrix by metalloproteinases is necessary (Tchetina et al., 2003). MMP-9 is a key component of chondrocyte maturation and resulting vascular invasion of the growth plate, and MMP13 has also been implicated in chondrocyte hypertrophy and matrix remodeling in all zones of the physis (Wu et al., 2002; Tchetina et al., 2003).

One of the principle factors driving interstitial expansion of the growth plate is hyaluronan synthesis by hypertrophic chondrocytes, and this process is regulated by many of the factors listed here (Pavasant et al., 1996). It is interesting, therefore, that the hypertrophic layer of the physis in rats given 300 mg/kg/day GW788388 for 4 days showed increased deposition and altered proteoglycan staining. Presumably, these proteoglycan alterations are a consequence of dysregulation of TGF-β, PTHrP, IGF-1 and other interacting factors.

Multiple mRNAs and proteins were analyzed by LCM and Taqman, based on a history of a well-established relationship with TGF-β regulation and a prescribed effect on chondrocyte function in the physis. IGF-1, IHH, PTHrP, FGFs, BMPs, MMPs, and VEGF have all shown to be crucial regulators of chondrocyte proliferation and differentiation (Van Der Eerden et al., 2002). TGF-β1, TGF-β2, and ALK5 were assayed in order to evaluate basal levels of expression in the physis and how these might be altered when TGF-β signaling was disrupted. ALK5 expression was only up-regulated in the prehypertrophic zone. TGF-β1 and TGF-β2 were both expressed at relatively low copy numbers throughout the perichondrium and physis and showed minimal to mild change.

There was slightly decreased TGF-β1 expression in the hypertrophic zone after treatment, while TGF-β2 was minimally to mildly up-regulated in all zones of the physis. The increased TGF-β2 expression noted after treatment may have represented a compensatory physiologic response to decreased signaling as TGF-β2 also utilizes ALK5 receptor signaling. TGF-β1 and 2 have been shown to stimulate chondrocyte proliferation and synthesis of collagens I, III, and IV, and inhibit matrix maturation, hyaluronan deposition, collagen X expression and synthesis, and bone-specific alkaline phosphatase, as well as inhibit hypertrophic differentiation by stimulating PTHrP (Pavasant et al., 1996; Ferguson et al., 2000; Nasatzky et al., 2000; Pateder et al., 2000; Yang et al., 2001). When some or all of these functions are inhibited, major phenotypic changes in the physis would be expected.

PTHrP has several important functions. PTHrP expressed by prehypertrophic chondrocytes inhibits further differentiation delays hypertrophy, modulates matrix synthesis, inhibits apoptosis, and thus regulates subsequent mineralization (Loveys et al., 1993; Terkeltaub et al., 1998). It is also indirectly responsible for hypertrophic differentiation to osteoblasts and is required for TGF-β1 to inhibit hypertrophic differentiation (Serra et al., 1999). PTHrP has been localized to the perichondrium, reserve, and proliferative zones by several groups, but is not significantly expressed in the hypertrophic zone (Terkeltaub et al., 1998; Nazatzky et al., 2000; Tchetina et al., 2003). A similar pattern of expression was observed in our LCM studies. Following GW788388 treatment, further decreases in expression levels in the prehypertrophic and hypertrophic zones were noted, but there was increased expression in the perichondrium and reserve/proliferative zones. The decrease in PTHrP expression is likely due directly to decreased TGF-β signaling as PTHrP expression is directly stimulated by TGF-β (Pateder et al., 2000). It is interesting to note that physeal dysplasia has been previously reported in mice with targeted deletions of PTHrP (Amizuka et al., 1994; Karaplis et al., 1994; Serra et al., 1999).

BMP-7 (also called OP-1), a member of the TGF-β superfamily (but one that is not considered a natural ligand for the ALK5 receptor), regulates osteogenesis and chondrocyte differentiation (Grimsrud et al., 1998). BMP-7 also stimulates chondrocyte proliferation and up-regulates a host of chondroid extracellular matrix-associated molecules. In particular, BMP-7 enhances the accumulation of hyaluronan and proteoglycan. Up-regulation of BMP7 noted in the proliferative zone of this study after ALK5 inhibitor treatment may be due to the up-regulation of regional PTHrP noted in treated physes. PTHrP is a potent stimulator of BMP7 expression (Terkeltaub et al., 1998). The overall effect of BMP-7 upregulation after GW788388 treatment likely has a great effect on proteoglycan alterations noted in the hypertrophic zone with Movat’s pentachrome stain.

IGF-1 was up-regulated in all zones of the physis and the perichondrium after treatment with 300 mg/kg/day GW788388. IGF-1 induces activation of chondrocytic proliferation and stimulates stem cells from the reserve zone (Van der Eerden et al., 2003). Growth hormone acts on reserve zone chondrocytes to stimulate IGF-1 production, which in turn stimulates clonal expansion of proliferating chondroyctes (Van der Eerden et al., 2003). The reserve zone was not morphologically different in treated and control physes at the light microscopic level, but activation of proliferation may have had an important role in the pathogenesis of the lesion. Increased numbers of proliferating cells were noted with Ki67 and Topoisomerase II immunohistochemical staining. As has been noted with other growth factors, disruption of IGF-1 signaling has been shown to result in physeal abnormalities. The hypertrophic zone is decreased in thickness in IGF-1 null mice (Wang et al., 1999). The increased expression of IGF1 noted by LCM in concert with thickened hypertrophic zone are thus consistent with an opposing effect.

IHH belongs to the family of hedgehog proteins, which are morphogens involved in embryonic patterning. IHH regulates the pace of chondrocytic differentiation (Van der Eerden et al., 2003). IHH has been proposed to stimulate PTHrP expression in the perichondrium in a negative feedback loop, and has several important functions including stimulation of chondrocyte proliferation, regulation of chondrocytic hypertrophy, inhibition of differentiation and stimulation of physeal stem cells in the reserve zone (Alvarez et al., 2001, 2002). At least some of these functions are independent of PTHrP signaling. Increased IHH was noted in all zones of the physis of treated rats, and this response was likely from lack of negative feedback inhibition, as IHH stimulates TGF-β expression and high local TGF-β activity is associated with decreased IHH expression (Alvarez et al., 2001, 2002). In TGF-β and FGF null mice with physeal dysplasia, there is increased expression and synthesis of IHH within several zones of the physis, which may also mediate some of the morphologic changes (Serra et al., 1997; Liu et al., 2002).

bFGF inhibits chondrocyte proliferation by inhibiting IHH, therefore inhibition of bFGF signaling is likely to induce chondrocyte proliferation, as was suggested in our studies and those with an FGF receptor inhibitor (Baron et al., 1994; Brown et al., 2005). Chondrocyte differentiation is also modulated by bFGF, and both actions appear to antagonize effects of the MMPs (Baron et al., 1994; Van der Eerden et al., 2003). bFGF can induce PTHrP expression, and like PTHrP, tends to oppose effects of the BMPs. Physeal dysplasia reported in FGF null mice resembles the ALK5 inhibitor-induced physeal lesion.

In a recent study, administration of an FGF receptor kinase inhibitor to rats also resulted in cartilage dysplasia, physeal dysplasia, and chrondrocyte proliferation, but with marked thickening of the zone of ossification and generalized soft tissue mineralization (which were not features in the ALK5 inhibitor cases) (Brown et al., 2005). The FGF receptor kinase inhibitor used (PD176067) also had some VEGF receptor kinase inhibitory activity. Vascularization and endothelial cell proliferation, which are known to be major factors controlling matrix mineralization at the ossification zone of the metaphysis, are greatly enhanced by either bFGF or VEGF treatment (Cross and Claesson-Welsh, 2001; Brown et al., 2005).

Alterations in bFGF and VEGF and effects on vascularization may have contributed in part to the subphyseal hyperostosis with retention of primary spongiosa noted with ALK5 inhibition. Impaired osteoclast recruitment and/or differentiation likely contributed to this change, as it is affected directly or indirectly by TGF-β, VEGF, and bFGF (Gerber et al., 1999). Up-regulation of bFGF noted in rat physes after 3 days was likely a secondary response to elevated IHH, TGF-β2, and PTHrP rather than a direct effect of ALK5. VEGF is directly stimulated by TGF-β1, and it regulates vascularization and osteoblast differentiation near the metaphysis, and particularly in the hypertrophic zone (Gerber et al., 1999). A VEGF monoclonal antibody given to young cynomolgus monkeys resulted in physeal dysplasia (Ryan et al., 1999), and a VEGF receptor kinase inhibitor (ZD4190) also caused increased width of the physeal zone of hypertrophy in rats (Wedge et al., 2000).

Unlike ALK5 inhibition, the physeal dysplasia associated with the FGFr inhibitor PD176067 showed no differences in susceptibility between young and mature rats and was also associated with changes in serum calcium and phosphorus levels and generalized soft tissue mineralization. While local bFGF, and presumably VEGF, receptor signaling were inhibited by PD176067, bFGF and VEGF mRNA levels in physes treated with GW788388 were instead up-regulated and their signaling pathways should not have been impaired. Despite differences in morphology and cytokine pathways involved, the fact that bFGF, VEGF, ALK5, and other growth factor signaling inhibitors all have resulted in physeal dysplasia in animal models is intriguing. We propose that it is the interaction of these tightly regulated cytokines, rather than effects of single factors, that maintain normal physeal phenotype. Disruption of one or more growth factor signals may affect the entire cytokine cascade, resulting in similar morphologic changes, which include physeal dysplasia. Certainly, there are subtle differences between phenotypes, but the similarities are striking between lesions associated with disparate growth factors.

Given the rather promiscuous nature of cross-reactivity of kinase inhibitors, it could be argued that ALK5 inhibitors are affecting other receptor kinase domains such as VEGF or bFGF and therefore mimicking their pathology. Activity against multiple protein kinases were assayed with 10 μM GW788388, as well as other ALK5 inhibitors, in the presence of 0.1 mM ATP, and all kinase activities other than to ALKs were similar to the DMSO controls (data not shown). This lack of cross-reactivity serves to demonstrate that the physeal effects were a function of ALK inhibition at toxicologic, suprapharmacologic doses rather than off-target pharmacology. Closely related ALK5 inhibitors were previously assayed for effects on other ALK family members (Inman, 2002; Grygielko, 2005). These inhibitors were approximately 10–20-fold less potent as inhibitors of ALK4 and ALK7, and 100-fold less potent as inhibitors of ALK3 and ALK6. No inhibitory activity of ALK2 was observed. The inhibitory profile of GW788388 followed this profile (data not shown).

Since there is much less homology between ALK5 and ALK1 than between ALK5 and ALK2, GW788388 inhibition against ALK1 activity is presumed low. Given these data and less significant physeal expression of other ALKs, inhibitory effects on ALK5 may be more important for development of the phenotype. Based on the large difference in level of drug exposure between antifibrotic pharmacologic activity of ALK5 inhibitors and the onset of physeal changes, it may be necessary to completely or nearly completely block ALK signaling to effectively disrupt growth factor homeostasis in the growth plate enough to induce physeal dysplasia. A similar paradigm might occur with other physeal growth factors and explain why physeal dysplasia occurs in the knockout mice described here, but is harder to reproduce with anti-growth factor antibodies or small molecule inhibitors.

Results of the in situ zymography demonstrating decreased gelatinolytic activity in the physis after ALK5 inhibitory treatment at first seem paradoxical. In most tissues including skin, subcutis and kidney, TGF-β inhibits matrix metalloproteinase activity and stimulates secretion of metalloproteinase inhibitors. However, in some tissues including bone and cartilage, TGF-β has been shown to up-regulate MMP-9 and MMP-13 (Uria et al., 1998b; Wu et al., 2002; Tchetina et al., 2003). MMP-13 cleaves type II collagen in the chondroid matrix into fragments and MMP-9 further breaks down fragments into even smaller components.

In normal rodent physes, MMP-9 (gelatinase B) has demonstrated to be active in the hypertrophic zone and has been associated with terminal differentiation of chondrocytes at the time of mineral formation in the extracellular matrix (Tchetina et al., 2003). MMP-9 plays a crucial role in the remodeling and degradation of the ECM and is involved in initiation of physeal/metaphyseal angiogenesis (Ortega et al., 2003). MMP-9 and MMP-13 null mice have abnormally vascularized physes and disordered bone formation similar to some of the changes noted in the ALK5 inhibitor-treated phenotypes (Vu et al., 1998; Lee et al., 1999; Stickens et al., 2004). MMP-13 (collagenase 3), essential for normal chondrocytic hypertrophy, has been localized throughout the physis, but in higher amounts in the hypertrophic zone (Tchetina et al., 2003; Stickens et al., 2004). Degradation of type 2 collagen and proteoglycan occurs in the last stages of chondrocytic differentiation and appears to be related to chondrocytic apoptosis as well as ossification (Stickens et al., 2004). In addition to TGF-β, MMP activity in cartilage appears to also be modulated by bFGF and other growth factors (Borden et al., 1996; Uria et al., 1998a). It is probable that decreased MMP activity, found in our studies after GW788388 treatment, plays a major role in physeal dysplasias and may be a downstream link between the physeal lesions initiated by a variety of growth factor kinase inhibitors.

Using TUNEL and caspase-3 immunohistochemistry, a mild decrease in numbers of apoptotic chondrocytes was appreciated only in the hypertrophic zone of the physis. Apoptotic cells have been demonstrated in all zones of the human and rat physis, but with greater frequency in the hypertrophic zone where they are terminally differentiated (Chrysis et al., 2002; Cetin et al., 2004). Apoptosis is associated with metalloproteinase-mediated matrix degradation and VEGF-mediated vascularization, so that its coordinated timing is dependent on the interplay of the same growth factors and cytokines that are involved in chondrocyte differentiation, proliferation, maturation, and growth plate angiogenesis. Alteration in apoptosis likely represents a similar subversion of the normal tightly controlled cascade of cytokines and growth factors. Apoptosis itself may also be involved in the web of growth factor regulation of physeal functional morphology, as apoptotic chondrocytes have been shown to release TGF-β (Tchetina et al., 2003).

Conclusion

Our results indicate that toxicologic, suprapharmacologic doses of ALK5 inhibitors given to rats result in physeal dysplasia. These physeal effects are consistent with a pharmacologic mechanism in that they localize morphologically with TGF-β associated functional morphogenic activity, occur with multiple compounds of similar class and share many similarities with physeal effects noted in several growth factor knockout mice. The pathophysiologic mechanism appears to involve altered expression and asynchronous regulation of TGF-β-associated growth factors and metalloproteinases involved in physeal chondrocyte maturation, proliferation, apoptosis, and especially proteoglycan synthesis by hypertrophic chondrocytes.

Since inhibition of the ALK5 pathway results in physeal lesions similar to those associated with several other inhibitors of growth factors and their receptors involved in physis homeostasis, we propose that it is the disrupted interactions/coordination or temporal regulation of these tightly controlled factors on chondrocyte functions, rather than a singular effect, that are likely responsible for the common phenotype of physeal dysplasia. As the physeal lesion occurs at doses well above pharmacologically active doses, appears to be limited to actively remodeling physis, and is not associated with changes in adjacent cortical or trabecular bone, the ALK5 inhibitor-induced physeal lesion should not pose a significant clinical health or safety risk in adult human patients given appropriate clinical monitoring.