Hox11 genes establish synovial joint organization and phylogenetic characteristics in developing mouse zeugopod skeletal elements

Synovial joints display distinct morphologies and organizations but the mechanisms by which they acquire such anatomical and functional characteristics are largely unknown (Archer et al., 2003; Pacifici et al., 2005; Pitsillides and Ashhurst, 2008). In the embryonic limb, joint formation initiates with the appearance of the so-called interzone that comprises flat and tightly packed mesenchymal cells and demarcates the boundary between adjacent cartilaginous skeletal anlaga (Holder, 1977; Mitrovic, 1978). The interzone is required for joint formation (Holder, 1977), but its specific roles have long remained unclear. To tackle this key issue, we carried out genetic cell tagging and tracking studies using Gdf5-Cre mice (Rountree et al., 2004) mated with RosaR26R mice (Soriano, 1999). We found that the lacZ-positive interzone cells gave rise to most if not all joint tissues, including articular cartilage layers, synovial lining and intra-joint ligaments, signifying that the interzone cells represent a specialized cohort of progenitor cells exclusively devoted to joint formation (Koyama et al., 2007a; Koyama et al., 2008). Other important studies showed that the Wnt/β-catenin signaling pathway maintains the function of interzone cells, cooperates with growth-plate-derived indian hedgehog and is essential for joint formation (Guo et al., 2004; Mak et al., 2006; Spater et al., 2006). Additional studies indicated that TGFβ signaling and TGFβ-versican interplays are needed for initiation of joint formation (Choocheep et al., 2010; Spagnoli et al., 2007) and that mechanical stimulation is required as well (Kahn et al., 2009; Pitsillides, 2006; Pitsillides et al., 1995). Despite these important advances on multiple fronts, other fundamental aspects of joint formation remain unclear and in particular, how joints acquire their diverse morphologies and organization.

Hox genes encode conserved homeodomain-containing nuclear proteins that regulate body plan, axis formation and organogenesis. Mammals have 39 Hox genes residing in four separate linkage clusters, termed A, B, C and D, that are subdivided into 13 paralogous groups each composed of 2-4 members. Hox9 through Hox13 group genes are specifically expressed during limb development, where they exert fundamental roles in patterning and growth of skeletal elements (Zakany and Duboule, 2007). In particularly striking examples, double-mutant mice lacking Hoxa11 and Hoxd11 were found to exhibit a significant retardation in radius and ulna development (Boulet and Capecchi, 2003; Davis et al., 1995), and triple mutants lacking all three Hox11 paralogous genes (Hoxa11, Hoxc11 and Hoxd11) displayed similar alterations in tibia and fibula development, whereas neighboring elements were largely normal (Wellik and Capecchi, 2003). Although seminal and pathbreaking in many respects, these and other studies did not examine whether or how Hox genes regulate synovial joint formation and what specific aspects of this process might be under Hox jurisdiction (Kmita et al., 2005; Yokouchi et al., 1991). The present study was carried out to tackle this key issue.

Mouse embryo mutants were generated by in vitro fertilization followed by transfer to foster mothers (Wellik and Capecchi, 2003), and all animal protocols were approved by the IACUC. Double Alizarin Red-Alcian Blue skeletal staining was performed according to established methods.

In situ hybridization was carried out as described (Koyama et al., 2007b). Paraffin sections pretreated with 10 μg/ml proteinase K for 10 minutes were post-fixed in 4% paraformaldehyde, washed with PBS containing 2 mg/ml glycine and treated with 0.25% acetic anhydride in triethanolamine buffer. Sections were hybridized with antisense ³⁵S-labeled riboprobes (approximately 1×10⁶ DPM/section) at 50°C for 16 hours, coated with Kodak NTB-3 emulsion diluted 1:1 with water and exposed for 10-14 days. Slides were developed with Kodak D-19 at 20°C and stained with Hematoxylin. Probes used were described previously (Koyama et al., 2007b). Dark- and bright-field images were captured using a digital camera and dark-field images were pseudo-colored using Adobe Photoshop software.

Triple mouse embryo mutants lacking all Hox11 genes and designated as Hox11aaccdd displayed short limbs and increased spinal curvature (see Fig. S1C in the supplementary material) and, interestingly, their limbs failed to undergo normal rotation (Coates et al., 2002) (see Fig. S1D in the supplementary material) compared with wild-type littermates (see Fig. S1A,B in the supplementary material). Alcian Blue and Alizarin Red staining showed that, as expected, embryonic day (E) 18.5 wild-type zeugopod elements exhibited typical morphologies and joint organization (Fig. 1A,D). The ulna (ul) was slightly wider than the radius (ra), and the normal tripartite elbow joint comprised direct articulation among the proximal portion of ulna containing a well-formed olecranon (ole) attached to the triceps brachii muscle tendon (Fig. 1C, asterisk), the proximal portion of radius and the distal portion of humerus (hu) (Fig. 1B,C). The tibia (ti) was far larger than the fibula (fi) (Fig. 1D); the normal bipartite knee joint involved direct articulation of the proximal tibia with distal femur (Fig. 1E) and was enclosed in a synovial cavity (Fig. 1F, opposing arrowheads), and the proximal portion of fibula articulated with the lateral condyle of the tibia (Fig. 1E, arrowhead).

Such stereotypic features were completely altered in E18.5 triple-Hox11-null embryos (4/4). Mutant zeugopod elements were hypomorphic and entirely cartilaginous (Fig. 1G-J, arrows), had lost their morphologic distinction in terms of length and width (Coates et al., 2002; Sears et al., 2007) and displayed elbow and knee joints with novel characteristics (Fig. 1H,K). In mutant elbows, the proximal articulating portions of ulna and radius resembled one another and the ulna lacked the olecranon (Fig. 1H) that, in addition to serving as the attachment site for the triceps brachii muscle tendon, is crucial for elbow extension and forward movement of forelimb and is absent or poorly developed in lower amniotes (Barnett and Lewis, 1958; Drapeau, 2004; Jenkins, 1973). A triceps brachii muscle tendon was still present (Fig. 1I, asterisk) but, strikingly, was connected to an ectopic skeletal element that closely resembled the ulnar patella characteristic of elbow joints of certain reptiles, amphibians and birds (Barnett and Lewis, 1958; Haines, 1969; Maisano, 2002; Vickaryous and Olson, 2007) (Fig. 1I, single arrowhead). The ectopic patella was attached to an apparently functional patellar tendon that reached the ulna metaphysis and resembled the tendon present in normal knees (Fig. 1I, double arrowheads). The ectopic patella could be observed as a distinct and sesamoid-like anatomical structure in whole-mount skeletal preparations (Fig. 2E, arrowhead) and displayed gene expression patterns typical of a functional articular skeletal element, including lubricin expression on its ventral side facing the joint (Fig. 2G, arrowhead) and abundant collagen IIB expression throughout its core (Fig. 2H). The changes in mutant knees were as encompassing. The proximal portions of both tibia and fibula fully articulated with an enlarged distal end of the femur that was able to accommodate both elements, thus resulting in a neo-tripartite joint (Fig. 1K). Indeed, the joint was enclosed in a single synovial cavity (Fig. 1L, opposing arrowheads) and exhibited an anterior ligament bridging femur to fibula (see Fig. S2E in the supplementary material, arrow). These anatomical features are also characteristic of knee joints in certain lower vertebrates (Dye, 1987; Haines, 1942).

Next, we asked whether the absence of the olecranon in mutant ulna was due to involution over developmental time or absence of embryonic specification. Thus, we first closely analyzed olecranon development in wild-type embryos (n=30). At E12.0, the cartilaginous and collagen-IIB-expressing humerus was still directly connected to the ulnar and radial anlaga and there was no overt sign of elbow joint formation (Fig. 3A,G, arrowheads). Approximately 12 hours later, the mesenchymal collagen-IIB-negative interzone had formed (Fig. 3B,H, arrowheads) and an olecranon primordium was already appreciable as a direct outgrowth and extension of the proximal ulnar epiphysis (Fig. 3B,H, curved arrow). By E13.0, the olecranon appeared well-consolidated and encircled the distal humeral epiphysis (Fig. 3C,I, curved arrow). Thus, we examined the crucial E12.5 stage in companion triple-Hox11-null embryos (5/5) and found that a collagen-IIB-negative interzone was forming but there was no obvious sign of olecranon formation (Fig. 3E,K, arrowhead). The same occurred at later stages (data not shown). To verify these observations, we determined the gene expression patterns of interzone marker Gdf5 and cartilage marker Sox9. In wild-type E12.5 embryos, the abundant Gdf5 transcripts demarcated the interzone that was flanked by Sox9-expressing olecranon primordium and the distal humeral end (Fig. 3D,J); the interzone was oriented at an ∼45° angle to fit between these two growing structures (Fig. 3D, white dashed line). In E12.5 triple-Hox11-null mutants, the interzone was equally rich in Gdf5 transcripts (Fig. 3F) but appeared to be oriented more perpendicularly (Fig. 3F, white dashed line), reflecting the altered morphologies of flanking Sox9-expressing mutant elements and the absence of an olecranon primordium (Fig. 3K,L). Thus, it appears that the olecranon is not specified in the mutants and olecranon development in wild types does not involve formation of a separate condensation (Barnett and Lewis, 1958).

To analyze the roles of individual Hox11 paralogous genes, we compared the elbow and knee joints in five-allele Hox11 mutants (4-5 for each genotype) that carry a wild-type copy (designated as Hox11Aaccdd, Hox11aaCcdd and Hox11aaccDd, respectively). Interestingly, elbow organization and ulnar olecranon development were largely restored in E18.5 Hox11Aaccdd and Hox11aaccDd (see Fig. S2B-C in the supplementary material) and knee organization was largely restored in Hox11aaCcdd and Hox11aaccDd mutants, but not Hox11Aaccdd mutants (see Fig. S2E-G in the supplementary material), suggesting that Hox11 gene function is not wholly redundant but includes forelimb- and hindlimb-specific joint formation roles.

Finally, we examined the superficial zone of articular tissue that establishes joint boundary and is essential for joint function and movement (Hunziker et al., 2007). In wild-type embryos, the superficial zone exhibited normal characteristics that included: flat and spindle-shaped cells oriented along the main axis of the articular perimeter; and low Safranin-O staining that reflects a physiologically low proteoglycan content compared with flanking epiphyseal cartilaginous tissue (Fig. 4A,B). In triple-Hox11-null mutants, the superficial layer was ill-defined. Its cells were round and their matrix stained as strongly with Safranin-O as the adjacent epiphyseal cartilaginous tissue (Fig. 4D,E). Changes in epiphyseal definition were further demonstrated by the relatively low expression of Ucma (Fig. 4F), a matrix gene product that is normally expressed very strongly in developing epiphyses (Tagariello et al., 2008) (Fig. 4C). We previously observed similar superficial zone changes in Hox11-null wrist and ankle joints (Koyama et al., 2010). In addition, some of the joints were partially fused and partial joint fusions were previously reported for more global Hox cluster deletions (Kmita et al., 2005). Thus, Hox11 genes are required for superficial zone formation in mammalian zeugopod elements, and other family members expressed in zeugopods, including Hox10 genes, cannot compensate for their function. Whether the changes in the superficial zone mirror native characteristics in lower vertebrate joints remains unclear.

Our data demonstrate that Hox11 genes co-regulate and coordinate zeugopod skeletal element development and the formation of adjacent synovial joints. In the absence of Hox11 function, the developing elbow and knee joints display novel anatomical characteristics and organization that include an ulnar patella in place of the olecranon and a single tripartite knee joint shared equally and fully by distal femur and proximal tibial and fibular ends. An essential feature of joints is that the opposing sides of each joint are reciprocally sculpted and interlocking and create functional biomechanical devices each perfectly suited for site-specific skeletal movement (Pacifici et al., 2005). Our data suggest that the opposing sides of mutant elbow and knee joints appear to undergo coordinated remodeling and produce adequate reciprocal morphologies most clearly seen in the enlarged distal femoral end that is able accommodate both tibial and fibular proximal ends. This suggests that joints are programmed to develop as single functional units and that the normal range of Hox11 action does not stop at the zeugopod-stylopod anatomical joint boundary.

Loss-of-function mutations in Hox genes often lead to homeotic transformations in which certain anatomical structures lose their identity and assume the identity of more anterior structures (McIntyre et al., 2007). A case in point is the anterior homeotic transformation in Hox10 triple mutants in which ectopic ribs form in the lumbar and sacral regions of their spine (Wellik and Capecchi, 2003). A similar process could have occurred in our Hox11 triple mutants such that the posterior ulna and fibula primordia would have undergone anterior homeotic transformation into the radius and tibia, respectively, resulting in morphologically similar structures. However, this interpretation does not account for the presence in the Hox11 triple-mutant joints of anatomical structures and organization typical of other and more ancestral species, including an ulnar patella and a tripartite knee joint (Fig. 5). What could these striking but unexpected findings signify and imply? One possibility is that the genetic blueprint and toolkit proteins (Carroll, 2008) needed for formation of those structures and morphological arrangements are still encoded in the mouse genome. These mechanisms, however, would normally be modified by species-specific Hox11 gene action or expression patterns to permit and sustain formation of mammalian joints. This interpretation fits well with, and provides additional experimental support for, current thinking on the genetic basis of morphological evolution and, in particular, the concept that new forms and anatomical organizations have arisen not necessarily from the evolution of new genes but, rather, changes in the function or expression of conserved genes (Carroll, 2008; Shubin et al., 2009). Thus, one could surmise that Hox11 expression in the elbow regions of species normally exhibiting an ulnar patella and a less-prominent or even absent olecranon would be much lower compared with that in mouse and/or could be restricted to the most-posterior margin of the limb. Another implication stems from the realization that the relative lengths, orientation, organization and joint structure of stylopod, zeugopod and autopod skeletal elements have undergone major and differential changes during limb evolution (Coates et al., 2002). Thus, it will be of great interest to clarify whether there are regulatory connections and interplays among substandard limb rotation, zeugopod skeletal element shortening and joint restructuring that concurrently occur in our Hox11 mutants.

We thank Drs John Fallon (University of Wisconsin) and Koji Tamura (Tohoku University) for comments and suggestions and Ms Christina Mundy for help with in situ hybridization. This study was supported by NIH grants AG025868 and AR046000 (to M.P.) and DK071929 and AR057018 (to D.W.). Deposited in PMC for release after 12 months.