Noncanonical frizzled signaling regulates cell polarity of growth plate chondrocytes

The growth of long bones is controlled via regulation of chondrocyte maturation in the growth plate cartilage, which is composed of four morphologically distinct zones of cells (see Fig. 1A-C)(Kronenberg, 2003). Resting chondrocytes comprise a population of progenitor cells that reside at the ends of the bones and exhibit a low level of cell cycle activation. Maturation begins with the formation of discoid proliferative chondrocytes that display increased cell cycle activity. After many rounds of division, the cell cycle is downregulated and the cells change shape (prehypertrophic chondrocytes). Subsequently, prehypertrophic chondrocytes enlarge to form hypertrophic chondrocytes that secrete the extracellular matrix protein collagen X and undergo cell death. Thus, the rate of long-bone growth is regulated via changes in cell number and cell volume(Breur et al., 1991; Hunziker et al., 1987).

The acquisition of unique morphogenetic properties is one remarkable characteristic of the maturation of resting chondrocytes into proliferative chondrocytes (Dodds, 1930). During the transition, individual resting chondrocytes, which are round and dispersed in the cartilage matrix, give rise to large single columns of tightly packed discoid proliferative chondrocytes. Whether the precise arrangement of chondrocytes depends on physical or mechanical properties of the extracellular environment (Aszodi et al., 2003; Ham,1932), or is regulated by secreted signaling molecules(Abad et al., 2002), is unknown. Here we show that proliferative chondrocytes undergo oriented cell divisions that occur orthogonal to the direction of growth and then intercalate to form columns in a process that appears highly similar to convergent extension by cell intercalation(Keller et al., 1989; Keller et al., 2000). Both the division plane and cell orientation depend on noncanonical frizzled (Fzd)signaling. Disruption of Fzd signaling additionally alters polarized growth of the long bones, resulting in skeletal elements that are shorter and wider than wild-type bones. Together, these results suggest a model in which regulation of cell polarity by noncanonical Fzd signaling plays a crucial role in controlling the three-dimensional morphogenesis of skeletal elements.

Total RNA was isolated from stage-25 chick embryos or from the growth plate cartilage of E16 chick femora dissected free of the periosteum, using Trizol(Invitrogen) as per the manufacturer's instructions. RNA was treated with DNase I then reverse transcribed using Superscript III (Invitrogen). A control sample was prepared in parallel by omitting reverse transcriptase (-RT).

Retrovirus constructs were generated in RCAS(A)(Hughes, 2004) using the Slax21 shuttle vector and standard methods(Logan and Tabin, 1998). Fzd7 and dnFzd7 were subcloned as previously described(Hartmann and Tabin, 2000). Other sequences were amplified from stage-25 chick embryo cDNA by PCR(reaction details and primer sequences are available upon request). RCAS-Dkk, and RCAS-dnLef1 and RCAS-daCamKII were gifts from C. M. Chuong (University of Southern California) and C. J. Tabin(Harvard Medical School), respectively. RISAP and RCAS(A)-daβ-catenin were provided by C. L. Cepko (Harvard Medical School) and Matsuhiro Iwamoto(Thomas Jefferson University College of Medicine), respectively. Concentrated virus stocks (10⁸-10⁹ pfu/ml) were prepared by standard procedures (Logan and Tabin,1998), except for RISAP, RCAS(A)-D2, RCAS(A)-D2^KM and RCAS(A)-dnRock, which were pseudotyped with VSV-G and produced by transient expression to 10⁶-10⁷ pfu/ml(Chen et al., 1999).

Fertilized chicken eggs (Phil's Fresh Eggs, Forreston, IL, USA) were grown at 37.5°C and staged (HH) as described(Hamburger and Hamilton,1992). Concentrated retrovirus stocks were injected into the right forelimb at HH 19-21. For mosaic analysis, stocks were diluted to 10⁷ pfu/ml in DMEM+0.1% FBS. After incubation, the humerus and ulna were dissected and fixed in 4% paraformaldehyde in PBS (pH 7.2) at 4°C. Micromass cultures were prepared and analyzed as described(Cottrill et al., 1987; Tufan et al., 2002).

To analyze cell morphology, tissue was fixed overnight, dehydrated through an ethanol series, embedded in paraffin (Paraplast X-tra, Tyco-Healthcare) and sectioned. Hematoxylin and Eosin (H&E) staining was performed using a standard protocol. For quantification of bone dimensions, the length and width(halfway between the articular surface and the mineralized region) were determined from the center section of a cartilage element (n=4 for each condition). For lineage analysis, harvested tissue was fixed for 4 hours,washed in PBS, heat-inactivated at 65°C for 30 minutes, and stained for alkaline phosphatase activity as described(Chen et al., 1999).

To analyze cell proliferation, S-phase cells were labeled with the nucleotide analog bromodeoxyuridine (BrdU) for 4 hours prior to tissue harvest. Sections of BrdU-labeled tissue were digested in trypsin and depurinated in hydrochloric acid. Incorporated BrdU was detected using a mouse anti-BrdU monoclonal antibody (G3G4, S. J. Kaufman, Developmental Studies Hybridoma Bank, Iowa, USA) and Cy2-labeled donkey anti-mouse antibody (Jackson ImmunoResearch, Pennsylvania, USA). Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; Sigma-Aldrich). The BrdU labeling index was calculated as the percentage of total nuclei (DAPI-stained) that were BrdU-positive.

Transcripts of Fzd pathway components were detected by PCR (reaction details and primer sequences are available upon request). Whole-embryo RNA from pools of HH 16-23 chick embryos was used as a positive control (data not shown). Bands were visible on ethidium bromide-stained agarose gels after 27-29 cycles, whereas background bands in -RT samples were only observed after more than 35 cycles. For in situ hybridization analysis, tissue sections were hybridized to RNA probes labeled with digoxigenin (Roche) as described(http://genepath.med.harvard.edu/~cepko/protocol).

We adapted previously described methods to analyze cell division in sections of fixed tissue (Gong et al.,2004; Tibber et al.,2004). Dissected cartilage was fixed overnight, equilibrated in 30% sucrose in PBS, and frozen in O.C.T. Compound (Tissue-Tek). Tissue was cryosectioned at 20 μm, mounted on Superfrost Plus slides (VWR), and briefly air-dried. Sections were washed in PBS, permeabilized in 1% Triton X-100, and incubated with rhodamine-phalloidin (1:100; Molecular Probes) and DAPI (0.1 μg/ml; Sigma) to label the contractile ring and DNA,respectively. Three-dimensional images were generated from a z-series of optical sections collected on a Zeiss apotome deconvolution microscope. The orientation of cell division was determined by mathematical modeling in which a cube was drawn with daughter nuclei occupying diagonal corners on opposite faces (see Fig. 2D). In this model, line segment `b' is set parallel to the long axis of the cartilage(equivalent to the distance between the nuclei in the y-axis) and line segment `a' is the distance between nuclei in the x-axis. Line segments `a' and `b' were measured using Zeiss Axiovision software. Line segment `c' (the distance between daughter nuclei in the z-axis) was determined by counting the number of 1 μm optical sections spanning the centers of the daughter nuclei. The angle (θ) generated by the intersection of line segment `b' and a line segment connecting the daughter nuclei was determined using the formula shown in Fig. 2D. Calculated angles were categorized into bins of 10° and the percentage of the total cells in each bin was plotted as a bar chart. Cells in metaphase were imaged using anti-γ-tubulin (1:1500; Sigma), secondary anti-mouse Cy3 (1:500; Jackson ImmunoResearch) and DAPI (0.1 μg/ml). Spindle orientation at metaphase was calculated by applying the same mathematical model as above except that the cube was drawn with apposing centrosomes at opposite corners.

The orientation of the long axis of the cell was determined from section in situ hybridizations. θ_cell is defined by the intersection of a line drawn through the longitudinal axis of the cell profile and a line drawn parallel to the long axis of the cartilage.

Data for each manipulation were obtained from at least six limbs (a minimum of three limbs on each of two days) except where noted in the text and see Table S1 in the supplementary material. The Kolmogorov-Smirnov test was used to determine whether the distributions of θ were distinct for different experimental conditions (Herrick,1965; Ong and LeClare,1968). For this analysis, P>0.05 demonstrates similarity between compared distributions. Tests for arbitrary angles compared the observed values to a uniform distribution over θ=0-90°.

Columns of discoid proliferative chondrocytes are readily observed late in embryogenesis because thick extracellular matrix separates individual stacks of cells (Dodds, 1930). To determine whether the unique organization of proliferative chondrocytes depends on physical restriction by the extracellular matrix, we examined cell behavior in chick long bones at early developmental stages. At embryonic day 9(E9; HH 33-35), proliferative chondrocytes are elongated but not organized into obvious columns (Fig. 1C). However, progeny of individual chondroprogenitor cells labeled by infection with the replication-defective avian retrovirus, RISAP(Chen et al., 1999), displayed a zone-specific arrangement. Thus, resting chondrocytes gave rise to clones displaying radial symmetry (25/25) (Fig. 1D), whereas 80% of clones (39/48) in the proliferating zone formed single columns (Fig. 1E). A minority of clones (data not shown) appeared either as two adjacent columns (7/48) or as large groups of labeled cells (2/48). Thus,columnar organization is an intrinsic behavior of proliferative chondrocyte clones at all stages of development.

Although columns of proliferative chondrocytes grow by clonal expansion,cell division is biased against orientations that result in daughter cells aligned parallel to the column (Dodds,1930). We asked whether the division plane, like column organization, is regulated prior to the thickening of the extracellular matrix. In sections of E9 long bones (Fig. 2B), cells in telophase were identified by the presence of an actin-rich contractile ring between two future daughter cells(Fig. 2C,E). An angle of cell division, termed θ, was determined from three-dimensional images(Fig. 2D) (see Materials and methods), such that 0° defines a cleavage plane that generates daughter cells arranged in a column, and 90° divisions displace daughters laterally with respect to the forming column. Wild-type resting chondrocytes divided at arbitrary angles, whereas wild-type and GFP-expressing proliferative chondrocytes exhibited a strong bias for telophase θ=81-90°(Fig. 2F; see Fig. S1 in the supplementary material). Bias in telophase θ was observed in E6 (HH 27)(Fig. 2A,F) proliferative chondrocytes and in E9 prehypertrophic chondrocytes(Fig. 2F). Thus, collectively,and consistent with our lineage data, resting and proliferative chondrocytes display distinct behaviors that remain substantially unaffected by developmental changes in the microenvironment.

What, then, restricts the plane of division in proliferative chondrocytes?One possibility consistent with Hertwig's Rule(Wilson, 1900) is that the discoid cell shape constrains the mitotic apparatus to lateral orientations(Carreira-Barbosa et al.,2003; Toyoshima and Nishida,2007). If this is the case in proliferative chondrocytes,orientation of the mitotic spindle at metaphase(Fig. 2G) should be similar to telophase θ. Unexpectedly, metaphase θ(Fig. 2H) was significantly different from telophase θ (Fig. 2F) (P<0.0001), demonstrating that spindle formation is not restricted to a specific axis. Furthermore, metaphase θ was highly similar in the morphologically distinct resting and proliferative chondrocytes (Fig. 2H)(P=0.555), indicating that the position at which the spindle forms is not strongly influenced by chondrocyte shape.

Spindle rotation during asymmetric cell divisions in Drosophila(Adler and Taylor, 2001; Roegiers et al., 2001) and in Caenorhabditis elegans (Wu and Herman, 2006a) requires Fzd function. Although Fzd signaling is important for growth plate function(Hartmann and Tabin, 2000), a role for Fzd in regulating the division plane in chondrocytes has not been reported. To test this possibility, we used replication-competent avian retrovirus (RCAS) (Hughes,2004) to express a C-terminal truncated form of chicken frizzled 7(Fzd7^-C) (Hartmann and Tabin,2000) in the chick forelimb. Chicken Fzd7^-C is analogous to a molecule that inhibits Fzd signaling by a dominant-negative mechanism in zebrafish (Nasevicius et al.,1998). E9 humeri uniformly expressing high levels of Fzd7^-C are significantly shorter and wider than those of uninfected controls (Fig. 3A,C), but cell proliferation was not affected(Fig. 4). In these limbs,infected resting chondrocytes appeared normal(Fig. 4C), whereas proliferative chondrocytes displayed aberrant cell morphology and the long cell axes were not aligned (Fig. 4I). Nonetheless, infected cells flattened at a position consistent with the normal interface between resting chondrocytes and proliferative chondrocytes (data not shown), suggesting that downregulation of Fzd signaling neither interferes with cell morphology in general, nor disrupts global organization of the growth plate.

To test whether the observed abnormalities reflect failed chondrocyte maturation, we examined the expression of specific genes (see Fig. S2 in the supplementary material). Fzd7^-C-expressing chondrocytes appropriately express early chondrocyte collagens (Col2a1 and Col9a1) and do not express collagen I (Col1a1), which is produced by dedifferentiated chondrocytes, osteoblasts and other mesenchymal cells. In addition, infected cartilage contains domains of Indian hedgehog(Ihh)-expressing prehypertrophic chondrocytes that surround a central region of hypertrophic cells (Fig. 4; see Fig. S2 in the supplementary material). These domains of mature chondrocytes are greatly reduced in size compared with wild-type cartilage, as previously described (see Fig. S2 in the supplementary material)(Hartmann and Tabin, 2000). Although Ihh expression is decreased, Fzd7^-C-expressing resting and proliferative chondrocytes express parathyroid hormone related peptide (Pthlh) and patched(Ptch), respectively (Hartmann and Tabin, 2000). Pthlh and Ptch are two Ihh-responsive genes that are also components of a key feedback loop that regulates proliferative chondrocyte formation (reviewed by Kronenberg, 2003; Yang et al., 2003). Together,these data suggest that chondrocytes expressing Fzd7^-Cundergo maturation and participate in long-range signaling interactions between the resting and prehypertrophic chondrocytes that are crucial to maintain growth plate architecture and function.

We next examined the division plane in infected E9 chondrocytes. Expression of Fzd7^-C did not affect telophase θ of resting chondrocytes, whereas telophase θ for infected proliferative chondrocytes was uniformly distributed from 0-90° (P=0.133)(Fig. 3B). In agreement with previous studies that demonstrated similar phenotypes in limbs expressing Fzd7^-C and wild-type Fzd7(Hartmann and Tabin, 2000), we found that Fzd7 expression resulted in a telophase θindistinguishable from that of Fzd7^-C-expressing proliferative chondrocytes (Fig. 3B) (P=0.738). Collectively, our data suggest that chondrocyte polarity is highly sensitive to Fzd signaling levels.

Cartilage growth is a tightly controlled process that involves interactions between chondrocytes and the perichondrium/periosteum (PC/PO)(Alvarez et al., 2001; Hinoi et al., 2006), a tissue that lines the developing cartilage. Thus, effects of Fzd7/Fzd7^-C expression on chondrocyte morphogenesis might result from defects in PC/PO function. To test this possibility, we asked whether infection of the PC/PO is sufficient to produce the observed phenotype. This test is possible because we found that soft tissues of the limb, including the PC/PO, are more easily infected with replication-competent virus than cells of the cartilage. Thus, injection of lower titer virus stocks results in highly infected PC/PO and few infected chondrocytes(Fig. 3C; see Fig. S3 in the supplementary material). Moreover, these data demonstrate that in situ hybridization can be used to identify infected cells in tissue sections because the inserted sequence is expressed at higher levels than the endogenous locus. In this context, limbs infected with low titer virus are similar to wild-type limbs in cartilage length and chondrocyte morphology(Fig. 3C; see Fig. S3 in the supplementary material). These data strongly suggest that Fzd signaling in the PC/PO does not regulate growth plate architecture or the growth properties of cartilage elements.

We next asked whether the organization of the proliferative chondrocytes depends on cell-autonomous Fzd function. We created mosaic embryos by injecting diluted virus to generate small patches of infected chondrocytes surrounded by large numbers of wild-type cells. Cells within patches expressing Fzd7^-C displayed a range of morphological defects, but most cells had clearly defined long and short axes(Fig. 3E). Quantification of cell orientation (θ_cell; see Materials and methods) revealed that wild-type cells adjacent to infected patches displayed normal orientation(Fig. 3D,E,E′). By contrast, expression of Fzd7^-C randomized the cell orientation, even when infected chondrocytes were surrounded by wild-type cells that displayed normal morphology(Fig. 3D,E,E′). To confirm that altered morphology results from cell-autonomous defects in Fzd function, we asked whether a similar phenotype would result from downregulation of dishevelled (Dsh, also known as Dvl), an intracellular signaling mediator that directly interacts with Fzd via a PDZ domain(Wallingford and Habas, 2005). PDZ-domain deletions of Dsh [D2 construct in Rothbacher et al.(Rothbacher et al., 2000)]abrogate Fzd signaling in a dominant-negative manner(Axelrod et al., 1998). Cells expressing D2 had a measurable long axis that was misaligned(Fig. 3D,F,F′). Thus, Fzd function autonomously regulates cell polarity and cell arrangement in the proliferative chondrocytes.

Having demonstrated a role for Fzd signaling in chondrocyte polarity, we next asked which Fzd effector pathway is disrupted by Fzd7^-C expression. The canonical Fzd pathway acts via stabilization of the transcriptional regulator β-catenin(Papkoff et al., 1996). To test whether reduced β-catenin function accounts for the observed defects in polarity, we expressed dickkopf 1 (Dkk)(Suksaweang et al., 2004), an extracellular antagonist that blocks activation of theβ-catenin-dependent pathway, and a putative dominant-negative mutant form of Lef1 (dnLef1)(Kengaku et al., 1998), a key component of the β-catenin transcriptional complex(Fig. 5A-C). Consistent with their predicted activities, expression of these inhibitory molecules enhanced cartilage formation in micromass cultures of chick limb mesenchyme(Fig. 5D). In vivo,chondrocytes expressing dnLef1 and Dkk displayed minor morphological changes (Fig. 5B,C), but telophase θ was similar to that of wild-type proliferative chondrocytes (Fig. 5I). As a further test of the requirement to regulateβ-catenin activity, we expressed a truncated form of β-catenin that lacks the N-terminal domain that confers instability and is therefore constitutively activated (daβ-cat)(Rubinfeld et al., 1997). In mosaic cartilage, groups of morphologically normal daβ-cat-expressing cells were readily observed in both resting and proliferative chondrocytes(Fig. 5F,G). Collectively,these data suggest that proliferative chondrocyte morphogenesis is not sensitive to β-catenin levels. Consistent with this, proliferative chondrocytes display normal organization in β-catenin mutant mice(Hill et al., 2005; Mak et al., 2006).

Since interfering with the canonical pathway did not alter the division plane, we next addressed whether the polarity defects observed in Fzd7^-C-expressing chondrocytes could result from altered noncanonical signaling. Gene expression analysis revealed the presence of multiple isoforms of calcium-calmodulin dependent protein kinase II (CamKII),which are components of the noncanonical Fzd/Ca²⁺ pathway in chick growth plate cartilage (Kuhl et al.,2000). Because putative dominant-negative forms of CamKII are weak inhibitors of the pathway, we tested the role of Fzd/Ca²⁺ signaling by ectopically expressing a constitutively active form of CamKIIα(daCamKII) (Abzhanov et al.,2006; Kuhl et al.,2000). Expression of daCamKII altered cell morphology(Fig. 5H) and shifted the plane of cell division from telophase θ=81-90°(Fig. 5I). These data are consistent with a potential role for noncanonical Fzd signaling in orienting the plane of cell division, but do not fully account for the uniform telophaseθ in cells expressing Fzd7^-C.

Our gene expression analysis also showed that some components of the noncanonical planar cell polarity (PCP) pathway(Klein and Mlodzik, 2005) are expressed in the developing growth plate(Fig. 5J). The possibility that a PCP-like pathway functions in proliferative chondrocytes is further supported by the similar phenotypes in Fzd loss- and gain-of-function experiments (Fig. 3B)(Krasnow and Adler, 1994) and the fact that D2 has a stronger inhibitory effect on the PCP pathway than on the β-catenin-dependent pathway(Axelrod et al., 1998; Krasnow and Adler, 1994; Rothbacher et al., 2000). Since PCP signaling requires the DEP domain in Dsh(Rothbacher et al., 2000; Wallingford and Harland, 2002; Wallingford et al., 2000), we first tested whether the effect of D2 on cell orientation requires DEP function. We introduced a missense mutation into D2 (K441M; D2^KM), analogous to the dsh1 mutation in Drosophila that specifically abrogates DEP-dependent PCP signaling by dishevelled but permits normal activation of β-catenin-dependent signaling (Axelrod et al.,1998; Boutros et al.,1998). The K441M mutation blocked the ability of D2 to interfere with chondrocyte morphogenesis(Fig. 3D,G,G′). These data suggest that D2 function depends on interaction with effectors of PCP signaling.

In the Drosophila wing, localization of Van Gogh (Strabismus;vertebrate Vangl), a four-pass transmembrane regulator of PCP signaling that interacts with Dsh (Park and Moon,2002), to the plasma membrane on one face of the cell confers polarity that aligns hairs in the plane of the epithelium(Klein and Mlodzik, 2005). As such, both gain- and loss-of-function Vangl mutations disrupt PCP in Drosophila and vertebrates(Krasnow and Adler, 1994). We expressed chick Vangl2 bearing a C-terminal hemagglutinin epitope tag(Vangl2^HA) in the wing cartilage and confirmed, by immunofluorescence, that the protein is secreted to the membrane(Fig. 6B,B′). Like Fzd7^-C, expression of Vangl2 generated short,wide cartilage elements (Fig. 6A) containing disorganized proliferative chondrocytes(Fig. 4K) that displayed normal cell cycle characteristics (Fig. 4) but reduced zones of mature chondrocytes (see Fig. S2 in the supplementary material). Similar to Fzd7^-C, expression of Vangl2 resulted in a uniform telophase θ(Fig. 6D) (P=0.103). Expression of the related gene, Vangl1, produced results indistinguishable from those of Vangl2 by histology(Fig. 4J) and telophase θ(Fig. 6D) (P=0.75). In addition, Vangl2 expression randomized the orientation of proliferative chondrocytes in a cell-autonomous manner(Fig. 6C,C′,E). Expression of a truncated form lacking the C-terminal PDZ-binding motif(Vangl2^-C) that is required for interaction with the PCP pathway but not the canonical pathway(Park and Moon, 2002), did not affect cartilage growth (Fig. 6A) or the division plane (Fig. 6D), suggesting that, as with D2, the effect of Vangl2 expression is dependent on interaction with downstream effectors of the PCP pathway. Potential downstream effectors of this pathway include the Rho GTPases and Rho-associated kinase 2 (Rock2)(Kim and Han, 2005; Phillips et al., 2005). We determined the effect of altered Rho pathway function by expressing a putative dominant-negative fragment of chicken Rock2 (Rock2^-N)(Leung et al., 1996) in the developing wing. Expression of Rock2^-N had a marked effect on proliferative chondrocytes, resulting in changes in cell morphology(Fig. 4L) and in the division plane (Fig. 6D) that were highly similar to those of chondrocytes expressing Fzd7^-Cand Vangl2. Collectively, these data suggest the possibility that the regulation of Rho by noncanonical Fzd signaling controls the plane of cell division and cell polarity in proliferative chondrocytes.

One remarkable characteristic of proliferative chondrocytes is the precise alignment of the cell division plane. In this process, spindle orientation at metaphase is refined to telophase θ=81-90° in proliferative chondrocytes, but remains unrefined in resting chondrocytes. Collectively, our data suggest that noncanonical (β-catenin-independent) Fzd signaling regulates the refinement of spindle orientation in proliferative chondrocytes(Fig. 7). One possible mechanism is that Fzd signaling directly affects spindle alignment by regulating the localization of proteins that position the centrosomes. Alternatively, noncanonical Fzd signaling might regulate cell flattening during mitosis that in turn constrains the mitotic spindle in telophase. Both models are consistent with known functions of the noncanonical pathways, but the latter model is unlikely to act predominately because the division plane is largely unchanged in daCamKII-expressing chondrocytes that do not flatten. That the division plane and cell shape are not tightly linked is further supported by the fact that neighboring affected cells often display different axial polarities. Instead, this suggests that, as in the neural tube(Ciruna et al., 2006),noncanonical Fzd signaling additionally functions to re-establish polarity in chondrocytes following cell division.

Re-establishing polarity after telophase might be important for column formation. Here we show by lineage analysis that following cell division newly generated daughter cells incorporate into columns in a process that is highly similar to convergent extension by cell intercalation, a coordinated cell movement that leads to the simultaneous narrowing and elongation of the anterior-posterior axis in vertebrate embryos(Elul et al., 1997; Keller et al., 1989). Daughter cells with different polarities might be unable to undergo the coordinated cell movements necessary for intercalation. However, one important difference with other examples of convergent extension is that chondrocyte columns derive from clonal expansion and not from cell recruitment, which generates mixed lineage columns in terminal filaments (TFs) of the Drosophilaovariole (Godt and Laski,1995) and in the vertebrate notochord(Keller et al., 1989). In this context, noncanonical Fzd signaling might act by maintaining a functional interface between daughter cells or by promoting complementary shape changes in interacting cells.

A key question is how polarity is coordinated both locally and globally in the proliferative chondrocytes. Tissue polarity has been modeled by gradients of secreted factors or cell adhesion. A ligand-dependent signaling model is consistent with our data and the observation that Wnt5a functions in proliferative chondrocyte morphogenesis in mouse(Yang et al., 2003), in convergent extension movements in zebrafish(Topczewski et al., 2001; Zhu et al., 2006), and in the polarity of stereocilia of the inner ear(Qian et al., 2007). However,it is not known whether cells determine polarity by interpreting a Wnt ligand gradient or whether Wnt signaling is required for competence to respond to other graded signals (Witze et al.,2008), and a gradient of Wnt proteins in the growth plate has not been demonstrated. One test of this model could come from understanding how heparan sulfate proteoglycans of the glypican family regulate cell intercalation via noncanonical Fzd signaling(Ohkawara et al., 2003) and from determination of whether ligand-dependent mechanisms are responsible for abnormal cartilage architecture in Ext1(Koziel et al., 2004) or Gpc3 (Viviano et al.,2005) hypomorphic mice and knypek (gpc4)-null fish (Topczewski et al.,2001).

In cartilage, cell-cell contact is unlikely to globally coordinate cell polarity because cartilage structure is not altered in mature growth plates in which thick extracellular matrix separates individual columns(Dodds, 1930). Nonetheless,order within clones (columns) could result from high-fidelity adhesion-dependent local propagation of polarity initially conferred upon the mother row cell that established the column. For example, cell-cell contact through adherens junctions defines polarity in some epithelial stem cell populations (Song et al.,2002). Alternatively, cell-matrix interactions can orient matrix fibrils and determine polarized cell movement(Davidson et al., 2006; Goto and Keller, 2002; Marsden and DeSimone, 2003)via integrin receptor function (Yu et al.,2005; Yu et al.,2008; Zhou and Kramer,2005), or provide boundary functions that promote cell intercalation (Veeman et al.,2008). A potential role for cell-matrix adhesion is further supported by altered growth plate structure in mouse mutants of cell-matrix interaction molecules (see Table S2 in the supplementary material). However,of these mutants, only loss-of-function of integrin β1(Aszodi et al., 2003) and ColIX (Blumbach et al.,2008; Dreier et al.,2008) display substantial loss of proliferative chondrocyte structure. Despite similarities to the phenotypes we describe, differences such as decreased cell proliferation and hypocellularity in cell adhesion mutants suggest that the polarity defects described in the present studies might not be solely due to altered cell-matrix interactions. It remains to be determined whether Fzd signaling directly regulates integrin/matrix production or localization, or modulates integrin function by crosstalk via a Rock-dependent pathway (Fig. 7). Regardless, noncanonical Fzd signaling might help maintain cell polarity and clone identity by regulating the expression and/or localization of cell adhesion complexes on proliferative chondrocytes.

Three findings from our analysis above are consistent with the possibility that a PCP-like pathway regulates the polarity of proliferative chondrocytes. First, the similar phenotypes we observe in Fzd loss- and gain-of-function experiments are a hallmark of PCP signaling(Krasnow and Adler, 1994). Second, Vangl gain-of-function phenocopies Fzdloss-of-function, an effect that depends on the presence of the C-terminus of Vangl, which interacts with PCP proteins. Third, the cell-autonomous effects of Fzd are replicated by D2 only when DEP function is maintained. Further support is provided by microarray analysis that shows specific regulation of the core PCP gene Prickle1 during chondrocyte maturation (Belluoccio et al.,2008).

Although our data strongly point to a role for a PCP-like pathway in growth plate morphogenesis, direct genetic support for this model is lacking. If not the PCP pathway, which noncanonical effector pathway acts in proliferative chondrocytes? One thread that connects pathways that regulate cell polarity during convergent extension, PCP in the inner ear, and the columnar organization of proliferative chondrocytes with neurite extension(Carreira-Barbosa et al.,2003; Jessen et al.,2002; Nambiar et al.,2007) and with one specific asymmetric cell division in the Caenorhabditis elegans embryo (Wu and Herman, 2006b) is that each process is consistently and predictably affected by altered activity of PCP proteins. These underlying commonalities suggest the presence of a core polarity module(Lawrence et al., 2007) that regulates cytoskeletal structure via Rock activity and cell adhesion modules(Yu et al., 2005; Yu et al., 2008). Thus,context-dependent regulators might confer emergent properties on a core pathway to generate distinct mechanisms for the interpretation and propagation of cell polarity in diverse tissues.

One potential physiological role for polarized cell behaviors is the regulation of cartilage growth. In particular, cell divisions oriented to produce daughter cells aligned with the axis of growth are characteristic of growing tissues (Baena-Lopez et al.,2005; Fischer et al.,2006). Curiously, proliferative chondrocytes divide and flatten orthogonal to expectations, requiring the intercalation of daughter cells following cell division. Subsequent hypertrophy of the chondrocytes, not cell intercalation or cell proliferation, drives the elongation of cartilage(Breur et al., 1991) and of the frog notochord (Adams et al.,1990). This two-step process could ensure high-density packing of cells in the longitudinal axis to maximize the growth potential of hypertrophy while minimizing disruption of column structure during mitosis. In our studies, conditions that result in column loss produced bones that displayed decreased longitudinal growth and increased lateral expansion. Consistent with this observation, a strong correlation between long-bone dimensions and proliferative chondrocyte structure exists in the literature(Aszodi et al., 2003; Blumbach et al., 2008; Dreier et al., 2008; Viviano et al., 2005; Yang et al., 2003). Thus,chondrocyte columns appear to orient vectors of growth and, therefore,regulated cell polarity could provide vertebrates with a powerful system for sculpting the diverse morphologies of bone that are required to generate an articulated skeleton.