Results

Genetic System for Testing the Function of Genes in Joint Development
To generate a general system capable of specifically testing genes for functions in skeletal joint development, we engineered transgenic mice to express Cre recombinase in developing joints (Figure 1). Gdf5 is a gene strongly expressed in stripes across developing skeletal elements during embryonic joint formation. A bacterial artificial chromosome (BAC) containing the Gdf5 locus was modified by homologous recombination in bacteria to insert a cassette encoding Cre-internal ribosome entry site (IRES)-human placental alkaline phosphatase (hPLAP) into the translation start site of Gdf5 (Figure 1A). This modified BAC was then used to make lines of transgenic mice. The resulting Gdf5-Cre transgenic mice were tested for transgene expression and Cre recombinase activity by crossing them to R26R reporter mice that activate the expression of lacZ after Cre-mediated removal of transcriptional stop sequences (Soriano 1999). The resulting progeny were analyzed both for expression of the transgene by assaying HPLAP activity and for recombination of DNA by assaying LACZ activity. The progeny from all three lines showed strong LACZ expression primarily in joints, and in two of three lines HPLAP expression could also be seen in joint regions. Interestingly, HPLAP expression in the Gdf5-Cre transgenic GAC(A) line used for all subsequent breeding experiments was seen to precede LACZ expression during successive development of joints in the digits (Figure 1C) (unpublished data). These experiments clearly demonstrate that the Gdf5-Cre transgene expresses Cre recombinase and causes DNA recombination in developing joint regions.
Figure 1  A Genetic System to Drive Gene Recombination in Developing Joints
(A) A 140-kb BAC from the Gdf5 locus was modified by inserting Cre-IRES-hPLAP into the translation start site of Gdf5 and used to make transgenic mice. Not to scale. See Materials and Methods for details.
(B–E) Visualization of Gdf5-Cre driven recombination patterns based on activation of lacZ expression from the R26R Cre reporter allele. (B) LACZ activity is visible as blue staining in the ear (ea) and the joints of the shoulder (s), elbow (eb), wrist (w), knee (k), ankle (a), vertebra (vj), and phalanges (black arrowheads) of an E14.5 mouse embryo. (C) E14.5 hindlimb double-stained to show both HPLAP expression from the transgene (grey/purple staining) and LACZ expression from the rearranged R26R allele (blue staining). Note that both markers are visible in the oldest, proximal interphalangeal joint (black arrowhead), only HPLAP activity is visible in the more recently formed medial interphalangeal joint (black arrow), and neither HPLAP nor LACZ expression is visible in the youngest, most distal joint of the digit (white arrowhead). (D) Newborn (P0) forelimb with skin partially removed showing LACZ activity expressed in all phalangeal joints (red Salmon gal staining, black arrowheads) and regions of some tendons (asterisk). (E) Section through the most distal phalangeal joint of a P0 hindlimb stained with Alcian blue to mark cartilage showing LACZ expression (stained red) in all tissues of developing joints: articular cartilage (black arrowhead), precursors of ligaments and synovial membranes (black arrow), and cells where cavitation is occurring (asterisk). GAC(A) mice were crossed with lacZ ROSA26 Cre reporter strain (R26R) mice to analyze the pattern of Cre-mediated lacZ recombination throughout development. Joints in developing limbs begin forming in a proximal-distal pattern such that the shoulder joint forms prior to the elbow joint. In addition, three major stages of early joint development have been defined by histology as (1) interzone formation, (2) three-layer interzone formation, and (3) cavitation (Mitrovic 1978). Consistent with the proximal-distal pattern of joint development in the limbs, LACZ activity is seen at embryonic day 12.5 (E12.5) in the more proximal joints, including the shoulder and knee (unpublished data). By E14.5, LACZ expression is typically seen in all but the most distal joints of the limbs (Figure 1B and 1C), but with some variability in both strength and extent of expression from embryo to embryo. The strongest-staining embryos often have additional staining in fingertips (not seen in the E14.5 embryo in Figure 1C, but clearly detectable in the E13.5 embryo shown in Figure 2). Sections through developing joints show that LACZ is present in many cells at the interzone stage (unpublished data). However, expression of LACZ in nearly 100% of joint cells is not achieved until the three-layer interzone stage (for example, in the knee joint at E14.5 or in any of the phalangeal joints at E16.5 (unpublished data). Within the developing skeleton, Cre-mediated expression of LACZ remains strikingly specific to joints throughout development. Furthermore, it is seen in all the structures of postnatal synovial joints including the articular cartilage, joint capsule, and synovial membrane (Figure 1D and 1E) (unpublished data). These patterns are consistent with the well-established expression of Gdf5 in interzone regions during embryonic development (Storm and Kingsley 1996). Adult expression patterns of the Gdf5 gene are not as well characterized, but Gdf5 expression has previously been detected in adult articular cartilage using both RT-PCR and immunocytochemistry (Chang et al. 1994; Erlacher et al. 1998; Bobacz et al. 2002).
Figure 2  Bmpr1a Is Required for Webbing Regression and Apoptosis in Specific Regions of the Limb
(A and B) Control E14.5 forelimb (A) compared to a, E14.5 mutant forelimb (B) showing webbing between digits 1 and 2 (arrowheads) and extra tissue at the posterior of digit 5 (arrows).
(C) Gdf5-Cre induced lacZ expression from R26R in an E13.5 forelimb showing LACZ staining (blue) in metacarpal-phalangeal joints, between digits 1 and 2 (arrowhead), and in a region posterior to digit 5 (arrow).
(D and E) Sections of E14.5 hindlimbs showing apoptosis visualized by TUNEL staining (green) and proliferation visualized by staining for histone H3 phosphorylation (red). Controls show strong, uniform TUNEL staining between digits 1 and 2 (D, arrowhead) while mutants show patchy TUNEL staining interspersed with mitotic cells in similar regions (E). Scale bar = 200 μm.
(F) Quantitation of TUNEL staining and mitotic cells in the posterior region of the fifth digit shows apoptosis is reduced 30% while proliferation is increased 20% (asterisks indicate statistically significant difference).
(G and H) By E15.5, interdigital tissue has regressed in controls (G, arrowhead). In contrast, tissue remains in mutants at this location, primarily derived from cells that have undergone Gdf5-Cre-mediated recombination that inactivates Bmpr1a function and activates expression of LACZ (H). Scale bar = 75 μm. Other sites besides limb joints also have Cre-mediated lacZ expression. Starting at E13.5, LACZ activity is detected in an anterior and posterior domain of the limb bud (Figure 2C). At E14.5, LACZ activity is detectable in the developing ear pinnae, ribs, sternum, tissues in the face, and some regions of the brain and spinal cord (Figure 1B) (unpublished data). At birth, LACZ is also expressed in tendons running along the vertebral column, regions of tendons in the wrist and ankle, and some tendon insertions (Figure 1D) (unpublished data). By 5 wk of age, LACZ is also expressed in the hair follicles, ear cartilage, some cells in the growth plate of the long bones, and portions of the brain and spinal cord (unpublished data). Surprisingly, 23 of 63, or 37% of transgenic mice analyzed also show some degree of wider “ectopic” LACZ expression, which can extend throughout many different tissues in the animal. However, sustained expression of the transgene itself, as assayed by HPLAP activity, is still restricted primarily to joints in animals that show evidence of more generalized recombination based on LACZ expression (unpublished data). This suggests that in a fraction of animals, sporadic expression of Cre at some time early in development is sufficient to lead to both ectopic recombination and LACZ expression. While the fraction of animals with broader recombination patterns must be tracked and accounted for during experiments, these animals offer the potential benefit of revealing additional new functions of target genes that could be subsequently studied with additional site-specific Cre drivers.

Gdf5-Cre/Bmpr1afloxP Animals Survive to Adulthood with Ear, Webbing, and Joint Defects
We next used the Gdf5-Cre system to test the role of BMP signaling during normal joint development. Gdf5-Cre transgenic mice were bred to animals carrying a conditional floxed allele of the Bmpr1a locus (Mishina et al. 2002), usually in the presence of the R26R reporter allele to facilitate simultaneous visualization of Cre-mediated recombination patterns (see typical cross in Figure 3). PCR amplification confirmed that a key exon of the Bmpr1a gene was deleted in mice that also carried the Gdf5-Cre transgene (unpublished data). Previous studies have shown that the recombined Bmpr1afloxP allele mimics a null allele of the Bmpr1a locus when transmitted through the germline (Mishina et al. 2002). The Gdf5-Cre/Bmpr1afloxP conditional knockout mice were viable and survived to adulthood, showing that the Gdf5-Cre driver can bypass the early embryonic lethality previously reported in animals with a null mutation in the Bmpr1a locus (Mishina et al. 1995).
Figure 3  Gdf5-Cre-Mediated Deletion of Bmpr1a
(A) Breeding strategy simultaneously deletes Bmpr1afloxP and allows visualization of Gdf5-Cre-mediated recombination by lacZ expression from R26R.
(B–E) 5-week-old mutant and control mice stained with Alcian blue to mark cartilage and alizarin red to mark bone. (B) Ankle of control with strong blue staining lining each joint (arrowheads). (C) Ankle of mutant showing an absence of blue staining in most regions (arrowheads) and a joint fusion between the central (c) and second (2) tarsals (arrow). (D) Control and (E) mutant metatarsal/phalangeal joint which lacks blue staining in articular regions (arrowheads) but retains staining in the growth plate (asterisks).
(F) Control forelimb.
(G) Mutant forelimb with webbing between the first and second digit (black arrowhead). The viable Gdf5-Cre/Bmpr1afloxP mice showed several phenotypes. First, the conditional knockout mice had shorter ears that often lay flatter against their heads than controls (controls 13.1 ± 0.1 mm long, n = 38; mutants 11.8 ± 0.2 mm, n = 11; p < 0.0001). BMP signaling is known to be required for growth of the external ear of mice (Kingsley et al. 1992), and this phenotype likely reflects loss of Bmpr1a function in the fraction of ear cells that express the Gdf5-Cre transgene. Most mutant mice also showed soft tissue syndactyly or retention of webbing between the first and second digits of their feet, a phenotype that was more frequent and more severe in the forelimbs (201 of 220, or 91%, of forefeet and 109 of 220, or 50%, of hindfeet). Finally, mutant animals showed obvious skeletal changes in whole-mount skeletal preparations. At some sites in the ankles, joints seemed to be missing entirely, with fusion of bones that would normally be separate. For example, the second distal tarsal was fused to the central tarsal bone in every conditional knockout animal examined (18 of 18), a phenotype not observed in controls (zero of 18) (Figure 3B and 3C). At other locations, joints had clearly formed but showed dramatic loss of staining with the cartilage matrix marker Alcian blue (Figure 3B–3E) (unpublished data). Normal Alcian blue staining was seen in non-articular regions, such as the cartilaginous growth plate (Figure 3D and 3E, asterisk). These data suggest that Bmpr1a function is required for the formation of specific joints in the ankle region and for either generation or maintenance of articular cartilage in most other joints of the limb.

Developmental Origin of Webbing Phenotype
Interdigital mesenchyme is normally eliminated by apoptosis during embryonic development, a process that can be stimulated by BMP beads, inhibited by Noggin, or blocked by overexpression of dominant-negative BMP receptors (Garcia-Martinez et al. 1993; Yokouchi et al. 1996; Zou and Niswander 1996; Guha et al. 2002). Limbs of Gdf5-Cre/Bmpr1afloxP mutant embryos showed obvious retention of interdigital webbing between the first and second, but not other, digits of E14.5 forelimbs (Figure 2A and 2B), a pattern that corresponds to the presence or absence of webbing seen in the adult limb. They also showed excess tissue on the posterior margin of the fifth digit (Figure 2B, arrow). Analysis of LACZ expression in Gdf5-Cre/R26R reporter embryos showed that Cre-mediated recombination has occurred by E13.5 in the metacarpal-phalangeal joints, and in the interdigital region between the first and second, but not other, digits. In addition, a domain of recombination and expression of LACZ is also reproducibly seen in the posterior half of the fifth digit (Figure 2C). Terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) staining of interdigital mesenchyme between the first and second digits (Figure 2D and 2E) and the fifth digit flanking mesenchyme showed a decreased number of dying cells in the regions where excess tissue is retained in the mutant limbs. Numbers of phosphorylated histone H3-labeled proliferating cells were also elevated in these regions (Figure 2F). Most cells found in the webbed region between the first and second digits at E15.5 strongly expressed LACZ in Gdf5-Cre/Bmpr1afloxP mutant embryos (Figure 2H). These data suggest that regional loss of BMPR1A receptor signaling blocks programmed cell death in interdigital mesenchyme, and that the recombined cells survive and proliferate in the absence of BMPR1A signaling.

Failure of Early Joint Formation in Ankle Regions
The Bmpr1a gene is expressed in the interzone region of developing joints at E13.5 (Baur et al. 2000). In situ hybridization showed that the gene is also expressed in the interzones of ankle joints and prospective articular cartilage regions of digit joints at E15.5 (Figure 4). LACZ staining indicated that Cre-mediated recombination begins to occur in ankle joints around E14.5, and is extensive by E15.5 (Figure 4G and 4J) (unpublished data). In the ankle joint regions that were obviously fused in postnatal mutant animals, alterations in early joint marker expression could also be seen by E15.5. At this stage, the Gdf5 gene is normally expressed in stripes that mark the sites of joint formation (Figure 4F), and the gene for the major collagen protein of cartilage matrix (Col2a1) is down-regulated in the interzone region (Figure 4E). In contrast, Col2a1 staining extended completely through the joint region between the second and central tarsal of Gdf5-Cre/Bmpr1afloxP mutants (Figure 4H, black arrow), and Gdf5 expression was seen only as a small notch extending into where the joint should be forming (Figure 4I, bracket). These data suggest that the fusions seen between ankle bones in postnatal mutant skeletons are the result of incomplete segmentation of skeletal precursors during embryonic development, a defect confined to some locations in the ankle.
Figure 4  Bmpr1a Is Expressed in Joints and Is Required for Continued Joint Formation in the Ankle Region
(A) Diagram of ankle bones from a wild-type mouse; bones fusing in mutant are colored red. Roman numerals II–IV, metatarsals; 2, 3, and 4/5, distal row of tarsal bones; c, central tarsal bone; ta, talus; ca, calcaneus.
(B and C) In situ hybridization at E15.5 showing that Bmpr1a is expressed in ankle joint interzones (B, arrowheads) and in the forming articular regions of the phalangeal joints (C, arrowheads).
(D) Near adjacent section to (C) showing Gdf5-Cre induced LACZ expression from R26R in the forming joints of the digits (arrowheads).
(E–J) Marker gene expression and R26R LACZ staining patterns on near adjacent sections of control and mutant embryos. In control mice at E15.5 ankle joints are clearly delineated as regions that have down-regulated Col2 (E), express Gdf5 throughout (F), and express LACZ in most cells (G; white arrowheads and black arrows). In mutant embryos at the same stage, joint formation is incomplete. Faint Col2 expression can be seen connecting a medial region of tarsal 2 with metatarsal II (H, white arrowhead), and Gdf5 expression does not extend all the way across the joint at this location (I, white arrowhead). Between tarsals c and 2, mutants express Col2 across the normal joint-forming region (H, black arrow) and lack expression of Gdf5 at sites where skeletal fusions are observed (I, black arrow and bracket). (J) Scale bar = 100 μm.

Failure to Maintain Articular Cartilage in Other Joints
In most joints of Bmpr1a conditional knockout mice, embryonic segmentation of skeletal precursors occurred normally. Although Gdf5-Cre-mediated recombination was seen as early as E13.5 in digit interzone regions (see Figure 2C), no changes in cell death or cell proliferation could be seen in the metacarpal-phalangeal or metatarsal-phalangeal joints at E13.5 or E14.5 (unpublished data). Similarly, although clear LACZ expression was seen by E15.5 in interphalangeal joints and periarticular regions (Figure 4D), no difference in morphology or expression of Col2a1, Gdf5, or Bmpr1b was seen in the articular regions of the phalanges at these stages (unpublished data).
At birth, digit joints were generally indistinguishable from those in control animals; chondrocytes were abundant in articular regions and were surrounded by typical cartilage matrix with normal staining by Safranin O, a histological stain for proteoglycans (Figure 5). At this stage, both wild-type and mutant cells in articular regions also expressed high levels of Col2a1 and Aggrecan (Agg), the genes encoding the major structural proteins of cartilage matrix (Figure 5B and 5G) (unpublished data). No alterations in cellular apoptosis or proliferation were observed (unpublished data).
Figure 5  Bmpr1a Is Required to Maintain Expression of ECM Components in Articular Cartilage
In situ hybridization or LACZ staining on near adjacent sections of metacarpal-phalangeal joints (A–C and F–H) and the tarsal 2-metatarsal II joint (D–E and I–J) of P0 mice. At birth, articular cartilage of controls (A–E) and mutants (F–J) appears similar by Safranin O staining (A and F), and Col2 expression (B, G). Mat4 expression confirms that articular cartilage is initially specified in mutants (D andI, brackets). LACZ expression confirms Cre-mediated recombination has occurred in articular cartilage (C, H, E, and J). (K–T) Near adjacent sections of the metacarpal-phalangeal joints of P14 mice. Two weeks after birth, articular cartilage of controls stains with pericellular Safranin O (orange staining, K), and expresses Col2 (L), Agg (M), and SOX9 (N). In contrast, mutant articular cells are smaller and more densely packed, lack pericellular Safranin O staining (P), have reduced expression of Col2 (Q) and Agg (R), but retain normal levels of SOX9 protein (S, brackets; dashed line marks faint edges of articular surfaces). LACZ expression confirms Cre-mediated recombination has occurred in articular cells (O ansd T, brackets). (A and K) Scale bar = 75 μm. To determine whether articular cells were properly specified in mutants, we also analyzed expression of Matrilin-4 (Mat4), a gene expressed specifically in the periarticular and perichondral regions of developing joints (Klatt et al. 2001). In both control and mutant animals, transcription of Mat4 was clearly detectable in the articular cartilage layers of newborn joints (Figure 5D and 5I). In all experiments, expression of LACZ throughout articular regions indicated that Cre-mediated recombination had occurred throughout the articular regions (Figure 5C, 5H, 5E, and 5J). The normal histological appearance, staining properties, and marker gene expression patterns suggest that Bmpr1a is not required for the initial formation or specification of articular cartilage.
By 1 wk after birth, obvious differences began to be detected in the articular regions of mutant animals. The expression of Col2a1 was reduced throughout the articular surfaces of the carpals, metacarpals, and phalanges of the forefeet (unpublished data). Less severe reductions were also seen in articular cells of tarsals and metatarsals in the hindfeet (unpublished data). By 2 wk of age, Col2a1 expression was reduced in most cells of the articular region (Figure 5L and 5Q), accompanied by markedly reduced Safranin O staining (Figure 5K and 5P), and decreased expression of Agg and two genes normally expressed in more mature articular cartilage cells, Collagen 3 (Col3a1) and Collagen 10 (Col10a1) (Figure 5M and 5R) (unpublished data) (Eyre 2002). Inhibition of BMP signaling in cultured chondrocytes has previously been reported to induce Collagen 1 (Col1a1) expression, increase proliferation, and result in cells with flattened, fibroblast-like morphology (Enomoto-Iwamoto et al. 1998). However, we saw no increase in the expression of Col1a1 in mutant articular cartilage, and no proliferation was detected in articular cells of either mutant or control animals (unpublished data). While recombined LACZ marker expression was detected in most articular cartilage cells, it was also observed in scattered subarticular chondrocytes, growth plate chondrocytes, and osteoblasts (Figure 5O and 5T) (unpublished data). Although this implies that BMP signaling was defective in multiple cell types, the observed defects were confined to the articular cartilage. For example, Osteocalcin and Col1a1 expression appeared normal in osteoblasts (unpublished data). Together, these data suggest that BMPR1A activity is required in postnatal joint articular cartilage to maintain expression of many genes encoding structural components of cartilage matrix.
Previous studies have shown that Sox9 is required for normal cartilage differentiation, for expression of cartilage extracellular matrix (ECM) genes including Agg, and is a direct transcriptional regulator of the key cartilage matrix gene Col2a1 (Bell et al. 1997; Lefebvre et al. 1997; Bi et al. 1999; Sekiya et al. 2000). Notably, despite reduced expression of many cartilage matrix marker genes in Bmpr1a mutant mice, the SOX9 protein was present at normal levels in articular regions at all stages examined, including newborn, 2-wk-old, 7-wk-old, and 9-mo-old mice (Figure 5N and 5S) (unpublished data).

Synovial Hypertrophy, Cartilage Erosion, and Accelerated Cartilage Maturation
Conditional loss of Bmpr1a led to marked hypertrophy of the synovial membrane in the joint capsule of some joints, particularly in the ankle region. In the most severely affected joints, the expanded synovial membrane grew into the joint space and was associated with obvious loss or erosion of the articular cartilage (Figure 6A and 6B, asterisks, arrows). Accelerated cartilage maturation and increased expression of Col10a1 was frequently seen in the chondrocytes underlying the articular erosions (Figure 6C and 6D, brackets) (unpublished data). Interestingly, the regions of increased Col10a1 expression did not correspond to the regions that had undergone Cre-mediated recombination. Instead, increased expression of Col10a1 was seen in a zone of largely LACZ-negative cells stretching from the cartilage adjacent to the ossification front (where Col10a1 is normally expressed in maturing cartilage cells), toward the regions where surface articular cartilage was severely eroded or missing (Figure 6A and 6B, arrowheads). Previous studies suggest that parathyroid hormone-related protein, a diffusible signal made in the articular surface, may normally inhibit maturation of underlying cartilage (Vortkamp et al. 1996; Weir et al. 1996). Local loss of the articular surface could remove this inhibition and lead to a cell-nonautonomous acceleration of maturation in chondrocytes underlying points of articular erosion.
Figure 6  Synovial Membrane Expansion, Articular Surface Erosion, and Accelerated Maturation of Underlying Cartilage in Ankles of Bmpr1a Mutant Mice
Near adjacent sections from the tarsal 2-metatarsal II joint of 7-d-old mice. (A and B) LACZ staining (blue) shows Cre-mediated recombination is largely restricted to articular (arrowheads) and synovial cells (asterisks) in both controls and mutants. (C and D) In situ hybridization shows Col10 expression expands in mutants toward regions of synovial membrane expansion and articular surface erosion (brackets and arrows). This may be a cell nonautonomous effect of joint damage, since the LACZ expressing cells at the articular surface do not show upregulation of Col10 (arrowheads) and the region of expanded Col10 expression is largely made up of cells that have not undergone Cre-mediated recombination. Note the formation of a cartilaginous bridge along the joint capsule of the mutant where joint formation is disrupted at earlier stages (B, white arrowhead, and Figure 3, white arrowheads). (A) Scale bar = 75 μm. This synovial hypertrophy is associated with increased numbers of mononuclear cells resembling synoviocytes or macrophages, cell types that are difficult to distinguish even with surface markers at early postnatal stages. However, no neutrophils were observed, suggesting that there is little inflammation. At later stages synovial hypertrophy is reduced. Further work will be needed to determine whether synovial development is regulated by BMP signaling, or whether the synovium becomes enlarged as a response to nearby skeletal malformations (such as fusion of the second and central tarsals or defects in the articular cartilage).

Noninflammatory Degeneration of Articular Cartilage in Digit and Knee Joints
Outside of the ankle region, little or no evidence was seen for expansion of the synovial membrane. Instead, mutant mice showed histological signs of osteoarthritis, such as fibrillation of the articular surface (Figure 7). As previously seen in 1- and 2-wk-old animals, Safranin O staining and Agg and Col10 expression were all reduced in mutant articular regions of the forefeet and hindfeet by 7 wk of age, and the beginning signs of cartilage loss were observed (unpublished data). By 9 mo of age, many regions of articular cartilage were completely missing or extremely fibrillated, leaving regions of exposed bone on the surface (Figure 7A–7D). No alterations were seen in the expression of Osteocalcin, Col1a1, or matrix metalloprotease-13 at either 7 wk or 9 mo.
Figure 7  Loss of Bmpr1a Signaling Leads to Articular Cartilage Fibrillation and Degeneration in Digits and Knees of Aging Mice
(A–D) Near adjacent sections of metatarsal-phalangeal joints from 9 month old mice. Articular cartilage of controls is complete and stains strongly with Safranin O (A, orange stain). In contrast, articular cells of mutants are severely fibrillated or absent with much reduced staining of Safranin O (C, arrowheads). LACZ expression confirms Cre-mediated recombination has occurred in articular cells (B and D).
(E–P) Sagittal sections through knee joints of 7-wk- (E–J) or 9-mo-old animals (K–P); fe, femur; ti, tibia; gp, growth plate. Seven weeks after birth, the height of the tibial epiphysis is reduced in mutants (E and H, bars), and their articular layer stains poorly with Safranin O, is fibrillated, and is strikingly thinner (F and I, black arrowhead, and brackets). Near adjacent sections with LACZ staining confirm Cre-mediated recombination has occurred in articular cells (G and J). Note that in mutants, LACZ is absent in cells adjacent to those that do stain with Safranin O, suggesting Bmpr1a may act cell autonomously (I and J, white arrowheads). At 9 mo old, the mutant tibial epiphysis is extremely thin (K and N, bars), and the articular layer is completely absent, leaving bone to rub directly on bone (L and O, bracket). LACZ staining shows Cre-mediated recombination occurred in articular cells of controls (M) and in some remaining skeletal tissue of mutants (P). Also note aberrantly formed meniscal cartilage in mutants (E, H, K, and N, arrows), and increased sclerosis in mutant epiphyses (E, H, K, and N, asterisks).
(A and K) Scale bar = 50 μm; (I) scale bar = 300 μm. The major weight-bearing joint of the hindlimb, the knee, showed changes that closely paralleled that seen in the foot joints. All markers of cartilage matrix looked similar to controls at E16.5, suggesting that early stages of joint formation were not disrupted (unpublished data). By postnatal day 7, Safranin O staining and Col2a1 and Agg expression were clearly reduced in the mutant, despite continued expression of Sox9 (unpublished data). The overall shape of mutant knee skeletal elements appeared similar to controls, although the fibrocartilaginous meniscus that resides between the femur and tibia appeared much less dense in mutants at E16.5. Some cartilage formed in the meniscus region, but the size of these elements was greatly reduced and contained abundant cells with fibrous, noncartilaginous appearance (unpublished data). This reduction of the meniscus can also be seen in sections from 7-wk- and 9-mo-old animals (Figure 7E, 7H, 7K, and 7N, arrows).
At 7 wk of age the normally domed tibial epiphysis was flattened and depressed in the knees of mutant animals, markedly reducing the distance between the growth plate and articular surface (Figure 7E and 7H, vertical bar). Articular cartilage was also thinner than in control animals, showed nearly complete absence of Safranin O staining, and was either acellular or beginning to fibrillate in many regions (Figure 7F and 7I). The few large Safranin O-stained cells still apparent in mutant articular regions appeared to correspond in position to rare LACZ-negative cells in adjacent sections, suggesting that Bmpr1a is required cell-autonomously in articular cartilage (Figure 7I and 7J, white arrowheads). By 9 mo, large areas of mutant knees were devoid of articular cells, and the bones of the femur and tibia appeared to rub directly against each other. Furthermore, the epiphysis of the tibia was extremely depressed, to the point that growth plate cartilage was almost exposed through the surface of the bone (Figure 7K, 7L, 7N, and 7O). In addition, mutants at 7 wk and 9 mo showed subchondral sclerosis, especially in the epiphysis of the femur (Figure 7E, 7H, 7K, and 7N, asterisks). While subchondral sclerosis is commonly seen in cases of osteoarthritis, it is unclear in this case whether the sclerosis is mainly a response of bone formation to compensate for decreased articular cartilage, or whether it is the effect of loss of Bmpr1a signaling in some LACZ-positive cells that are also observed in these regions (unpublished data).
The histological signs of joint arthritis were accompanied by functional impairments in both grasping ability and range of motion in mutant animals. Gdf5-Cre/Bmpr1afloxP mutant animals showed a highly significantly reduced ability to grasp and remain suspended on a slender rod (mean suspension time: controls 38 ± 6 s, n = 39; mutants 6 ± 3 s, n = 11; p < 0.0001). Mutant mice also showed a clear decrease in the maximum range of mobility of two different joints in the digits, as assayed by passive manipulation (MT/P1 joint: controls 100 ± 0°, n = 26; mutants 82 ± 3°, n = 8; p < 0.0003; P1/P2 joint: controls 152 ± 1°, n = 23; mutants 140 ± 5°, n = 6; p < 0.05). The structural, histological, marker gene expression, and functional changes in mutant mice demonstrate that BMPR1A is required for normal postnatal maintenance of articular cartilage.