Versican G1 Domain and V3 Isoform Overexpression Results in Increased Chondrogenesis in the Developing Chick Limb In Ovo

Authors


Abstract

Previous work has shown that versican proteoglycan is highly expressed in the extracellular matrix of precartilage limb mesenchyme. Although much of versican's role in chondrogenesis has been attributed to its glycosaminoglycan complement, N- and C-terminal G1 and G3 domains of versican have been shown to possess distinct functions when expressed ectopically. This study was undertaken to test the hypothesis that overexpression of the versican G1 domain and short V3 isoform, comprised of only G1 and G3, in the chick wing in ovo would result in increased chondrogenesis, suggesting function for discrete versican domains in limb skeletal development. Recombinant adenoviruses encoding G1 and V3 proteins were microinjected into proximal HH19–25 chick wing buds which resulted in significant enlargement of humeral primordia at HH35. Enhanced cartilage deposition appeared due to increased chondrogenic aggregation as a result of recombinant G1 or V3 overexpression, further implicating versican in early stages of limb development. Anat Rec 293:1669–1678, 2010. © 2010 Wiley-Liss, Inc.

INTRODUCTION

Limb chondrogenesis and patterning of the appendicular skeleton are multistep processes in which each phase involves a complex array of cellular interactions. The first stage of overt chondrogenesis involves formation of precartilage mesenchymal condensations, a pivotal step (Fell, 1925; Mackie et al., 1987; Daniels and Solursh, 1991; Hall and Miyake, 1995, 2000; Olsen et al., 2000). Mesenchymal condensations are regulated and formed by cell–cell and cell–extracellular matrix (ECM) interactions (Mackie et al., 1987; Knudson et al., 1996) coupled with cell movement and mitotic activity (DeLise et al., 2000; Hall and Miyake, 2000; Barna and Niswander, 2007). ECM molecules implicated in initiation of cellular aggregation include collagen type I (Dessau et al., 1980), hyaluronan (Knudson, 2003), fibronectin (Kulyk et al., 1989; Downie and Newman, 1995; Chimal-Monroy and Diaz, 1999), tenascin (Hall and Miyake, 2000), and syndecan proteoglycan (Shimizu et al., 2007). Cooperative action of ECM constituents in early stages of aggregation may lead to cross-bridging of cells via N-cadherin (Shum et al., 2003; Shimizu et al., 2007) and N-CAM (Chimal-Monroy and Diaz, 1999; Hall and Miyake, 2000; Shimizu et al., 2007) which triggers chondrogenesis.

Versican is a predominant chondroitin sulfate proteoglycan in the prechondrogenic limb that has also been shown important during early stages of chondrogenesis (Zhang et al., 2001; Williams et al., 2005; Kamiya et al., 2006; Shepard et al., 2008) and is later restricted to the perichondrium and epiphysis of developing cartilages (Shinomura et al., 1990; Yamamura et al., 1997; Shibata et al., 2003; Snow et al., 2005; Shepard et al., 2007). Chondrogenesis fails to occur in limb mesenchyme or in cartilage precursor cells that lack versican in vitro (Williams et al., 2005; Kamiya et al., 2006). Furthermore, versican knockdown resulted in targeted inhibition of chondrogenesis of the embryonic chick limb in ovo (Shepard et al., 2008).

As with other members of the hyalectin family, the core protein of versican is modular with amino- and carboxy-terminal G1 and G3 domains separated by chondroitin sulfate glycosaminoglycan attachment regions (GAG-α and –β). Alternative splicing of the GAG-α and -β domains yields four isoforms, termed V0-V3 that may regulate cellular behavior (Shinomura et al., 1993). V0 possesses both GAG attachment regions, while V1 contains only GAG-β, V2 contains GAG-α, and V3 has neither GAG attachment sequence (Zimmermann and Rouslahti, 1989; Zako et al., 1995). All four variants express the G1 hyaluronan-binding domain and G3 carboxy terminal domain consisting of a C-type lectin-like domain, two epidermal growth factor (EGF)-like repeats, and complement regulatory domain (Shinomura et al., 1993).

Versican has been implicated in developmental processes such as eliciting cell shape changes, regulation of migration, and proliferation (Wight, 2002) and loss of mature versican results in an embryonic lethal phenotype (hdf mutant) due to defects in cardiac morphogenesis (Mjaatvedt et al., 1998). In addition to the hyaluronan-binding G1 domain, the G3 domain interacts with fibronectin, fibrillin, fibulin, collagen type I, as well as tenascin and may also interact directly with EGF receptors (Yamagata et al., 1986; LeBaron et al., 1992; Aspberg et al., 1995; Zhang et al., 1998; Isogai et al., 2002). Versican's chondroitin sulfate chains also bind CD44 and L- and P-selectins (Kawashima et al., 2001; Kamiya et al., 2006). V0 and V1 isoforms are expressed in heart and blood vessels where cell shape is important (Landolt et al., 1995) and also in vitro during early stages of chondrogenesis when cell shape changes are occurring (Kamiya et al., 2006). On the other hand, V2 is expressed in the brain (Schmalfeldt et al., 1998) and at later stages of chondrogenesis where cell shape is more stable (Kamiya et al., 2006). Furthermore, V2 promotes cell elongation and subsequent inhibition of chondrogenesis in vitro, whereas V1 promotes rounding (Sheng et al., 2006). Ectopic expression of the V3 isoform in arterial smooth muscle increases cell adhesion resulting in decreased growth and migration (Lemire et al., 2002). V3 overexpression in the embryonic cardiac outflow tract also caused thickening of the myocardium as a result of increased adhesion among cardiomyocytes (Kern et al., 2007).

Interestingly, proteolytic cleavage of versican liberates domains that may regulate its function in some tissues (Sandy et al., 2001; Russell et al., 2003; Kern et al., 2006, 2007; Capehart, 2010). Indeed, increasing evidence has shown that individual domains of versican may function independently of the intact proteoglycan when expressed ectopically (Zhang et al., 1998, 2001; Ang et al., 1999; Yang et al., 1999; Kern et al., 2007). This is especially important when taken with the fact that individual versican domains have been reported to possess differing activities. Ectopic expression of the G3 domain in vitro promotes elongation of chondrocytes, inhibiting chondrogenesis through its EGF domain (Zhang et al., 2001). On the other hand, expression of the G1 domain facilitates cell rounding by reducing cell–substrate interactions by binding to hyaluronan (Daniels and Solursh, 1991; Kohda et al., 1996; Toole, 2001). Versican/hyaluronan complexes create a loose highly hydrated environment with a destabilizing effect on adhesion that may influence cell migration (Yamagata et al., 1986; Landolt et al., 1995; Lee et al., 1999; Yang et al., 1999). In other cases, however, the activity of different versican domains may be complementary. Ectopic G1 expression in vitro enhances cellular proliferation (Zhang et al., 1998, 1999) through reduction in cell adhesion via hyaluronan binding (Ang et al., 1999). The G3 domain also stimulates cellular proliferation via its EGF domains (Zhang et al., 1998) and has been shown to protect against apoptosis (Wu et al., 2002). Thus, in vitro data suggest that versican domains have diverse functions leaving open the question regarding effects that individual domains of versican may have during limb chondrogenesis in vivo. In the present study, this led us to ask if individual versican domains are able to function independently of the mature proteoglycan during chick limb morphogenesis in ovo. As G1 and G3 domains are found in all versican isoforms we reasoned that their use would provide valuable insight into versican function. This is also particularly interesting because little is known regarding the V3 isoform with respect to its expression or potential role in limb skeletogenesis. Because full length versican has previously been implicated in precartilage mesenchymal condensation in vitro (Williams et al., 2005; Kamiya et al., 2006) and deficits (Shepard et al., 2008) or delay (Choocheep et al., in press) in formation in limb cartilages have been reported through loss of versican in vivo, we hypothesized that expression of individual G1 domains and V3 isoform, containing only G1 and G3, would facilitate limb chondrogenesis in vivo. Recombinant adenoviruses encoding G1 and V3 versican were utilized in these studies as viral-mediated constitutive expression has provided a reliable and powerful tool to investigate protein function in several model organisms, including the developing chick (Yajima et al., 2002; Kern et al., 2007; Li et al., 2007). Our results demonstrate that expression of recombinant versican G1 domain and V3 variant in the chick wing in ovo led to additional chondrogenic aggregation resulting in enlargement of cartilage at sites along the humeral primordium.

MATERIALS AND METHODS

Recombinant Adenoviruses

Recombinant adenoviruses expressing a C-terminal hemaglutinin (HA)-tagged G1 domain and V3 isoform of murine versican were prepared using the Adeno-X vector system (Clontech) as described previously by Kern et al. (2007). Control viral constructs expressing β-galactosidase contained the same cytomegalovirus promoter and viral backbone. Recombinant adenoviruses were amplified in low passage HEK 293 cells with virions purified and titered using Adeno-X virus Purification and Rapid Titer kits (Clontech) according to manufacturer's instructions. Immunohistochemistry and Western analysis of lysates prepared from transfected HEK 293 cells using an anti-hemaglutinin tag antibody (HA-7; Sigma) were used to confirm expression of proteins with the predicted molecular weights of ∼40 and 70 kDa, respectively, for recombinant G1 and V3 versican proteins (Kern et al., 2007).

Microinjection of Recombinant Adenoviruses

Fertilized, viral-free White Leghorn chick eggs (SPF; Charles River) were incubated at 37.5°C in a humidified egg incubator until Hamburger and Hamilton (HH) stages 19–25 (Hamburger and Hamilton, 1951). Eggs were “windowed,” overlying membranes removed, and right wing buds of appropriately staged, viable embryos microinjected with recombinant G1, V3, LacZ adenoviruses, or with adenoviral vehicle alone. Viral microinjections targeting the proximal limb core were performed utilizing a pneumatic pump (Model 820, WPI). G1 and V3 adenoviruses were injected at a 1:1 ratio with LacZ-encoding virus to track and localize the area of viral infection (Kern et al., 2007). Injections were performed by inserting the microinjection capillary needle at a 45- to 50-degree angle along the proximodistal axis just proximal to the center of the wing bud at HH19-23 and midway between the trunk and presumptive elbow joint at HH24-25. Each microinjection delivered ∼1.5 × 106 ifu of G1 or V3 adenoviruses (G1/V3 titers, 2 × 109 ifu/mL) in a total volume of 1.5 μL (0.75 μL G1 or V3:0.75 μL LacZ). LacZ control viral injections (6 × 109 ifu/mL) were performed using the same total volume. Following microinjection, eggs were sealed with tape, returned to the incubator and allowed to develop up until HH35. Embryos assessed in this study received a nonlethal injection of viruses as defined by survival beyond 48 hr postinfection (Kern et al., 2007). Experimental and control embryos were subject to identical handling and treatment. Findings represented here are based on analysis of a minimum of 30 experimental embryos for each of the recombinant G1, V3, and LacZ adenoviral injection series.

Whole-mount Histochemical Staining

Alcian blue and alizarin red histochemistry followed a modification of Kuczuk and Scott (1984). Embryos were fixed overnight in 95% ethanol, followed by staining overnight in 0.02% alcian blue in acidified ethanol (pH 1.0) at room temperature. After washing in 95% ethanol embryos were macerated in 2% potassium hydroxide (KOH) until tissue was translucent and skeletal elements were clearly visible (∼3 hr). Overnight staining in 0.01% alizarin red in 1% KOH was followed by a series of 1 hr KOH:glycerol exchanges (80:20, 60:40, 40:60, 20:80%). Embryos were stored in 70% glycerol:30% phosphate-buffered saline (PBS).

β-galactosidase staining protocol was adapted from Kern et al. (2007). Briefly, embryos were fixed on ice in 4% paraformaldehyde for 30 min, rinsed with PBS, and washed in 0.02% sodium deoxycolate and 0.01% Tergitol-type NP-40 in PBS. Specimens were stained in 0.1% X-gal in 0.02% magnesium chloride, 0.1% potassium-ferrocyanide, 0.1% potassium-ferricyanide at 37°C, washed in PBS, post fixed in 4% paraformaldehyde, and stored at 4°C.

Statistical Analysis

Measurements of three humeral sites were taken (Fig. 2F): across the length of the ventral tuberculum (proximal), at the center of the humeral primordium (mid), and from the dorsal to ventral condyle (distal). Anatomical structures were identified according to Bellairs and Osmond (1998). Examination of data included calculation of means, standard deviations and P-values for proximal, mid and distal regions of injected limbs and uninjected contralateral controls using mean matched-pairs analysis and t-test (Triola, 2005). Mean value of differences between injected and uninjected contralateral control limbs of individual embryos at specific humeral sites was used to test the following null hypothesis: no differences between humeral elements of the injected and uninjected contralateral control. Statistical significance was set at P < 0.05 with analyses performed using STATDISK software (Triola, 2005).

Immuno- and Histochemical Staining of Tissue Sections

Mouse anti-hemaglutinin (HA-7, Sigma) was used at 3.3μg/mL to detect recombinant G1 and V3 proteins and rabbit antiphosphohistone H3 (Ser10) (Cell Signaling Technology) at 1:100 as recommended by the manufacturer to identify mitotic cells Rhodamine-conjugated peanut agglutinin (PNA; Vector Labs) was used at 30 μg/mL to detect precartilage mesenchymal aggregates (Zimmerman and Thies, 1984) and biotinlylated hyaluronic acid binding protein (HABP; Associates of Cape Cod) at 2.5 μg/mL to localize endogenous hyaluronan. Double labeling was routinely performed with the hemaglutinin tag antibody and other reagents when possible.

Paraformaldehyde-fixed specimens were paraffin embedded, sectioned at 7 μm, dewaxed, and subjected to antigen unmasking based on a high-temperature citric acid formula (Vector Labs) for 20 min (Snow et al., 2005). To remove potentially masking hyaluronan or chondroitin sulfate glycosaminoglycan residues sections were incubated with 0.1% testicular hyaluronidase (Sigma) in 30 mM sodium acetate, pH 5.2, 125 mM sodium chloride for 30 min at room temperature (Capehart et al., 1999). Sections receiving HABP were incubated in 0.25 U/mL chondroitinase ABC (Sigma) in 50 mM Tris, pH8.0, 60 mM sodium acetate with 0.1% bovine serum albumen (BSA; Sigma). Sections were blocked with 3% BSA, 1% goat serum in PBS 1 hour and incubated with primary immunoreagents overnight at 4°C. Sections were washed with PBS and incubated with FITC (2 μg/mL)- or rhodamine (6 μg/mL)-conjugated goat anti-mouse and -rabbit IgG secondary antibodies (Cappel) for 1–2 hr at room temperature. FITC-strepavidin (Vector Labs) was used at 5 μg/mL. Primary antibodies were omitted from control specimens. Sections were washed with PBS, postfixed in 80% and 50% ethanol, re-equilibrated in PBS, and mounted in DABCO anti-fading agent [10% PBS-90% glycerol containing 100 mg/mL 1,4-diazabicycle(2,2,2)octane; Sigma].

In Situ Detection of Apoptosis

Terminal transferase dUTP nick end labeling (TUNEL) assay for detection of apoptotic cells on paraffin embedded sections was performed using the ApopTag Peroxidase In Situ Kit (Chemicon) following manufacturer's instructions. Briefly, samples were postfixed and endogenous peroxidase quenched with 0.1% hydrogen peroxide. Sections were equilibrated in terminal transferase buffer before addition of reaction buffer containing digoxigenin-11-dUTP followed by detection using antidigoxigenin and diaminobenzidine substrate.

Detection of Endogenous Versican V3 Variant mRNA in Developing Limbs

Reverse transcriptase-polymerase chain reaction (RT-PCR) was performed to detect endogenous V3 transcripts between HH25–34. RNA was isolated using RNaqueous 4-PCR (Applied Biosystems, Foster City, CA) and cDNA prepared with Superscript III (Invitrogen, Carlsbad, CA) kits. Primers used to amplify chick V3 cDNA were: 1447F-5′-TACCTCAGACTGTCACAGATGG-3′ and 10067R-5′-ATACATGTGGCTCCATTGCGGCA-3′. Primer sequences were derived from the 3′ end of chick versican “Plus” and 5′ end of G3 domains (Zako et al., 1995) to yield a predicted amplicon of 313 base pairs.

In situ hybridization was performed to localize endogenous V3 transcripts following Cronin and Capehart (2007). The V3 primer set was used to prepare inserts from HH25 chick heart cDNA for ligation into pGEMT-Easy (Promega). Digoxygenin-labeled RNA anti-sense and sense probes generated with the DIG RNA Labeling kit (Roche) according to manufacturer's instructions.

RESULTS

Although much is known regarding versican expression, particularly of V0/V1 isoforms, during chick limb skeletogenesis (Kimata et al., 1986; Shinomura et al., 1990; Landolt et al., 1995; Shepard et al., 2007), little information is available on V3 (comprised of only G1 and G3 domains) expression during developmental stages relevant to the present study. As such, RT-PCR and in situ hybridization were utilized to detect endogenous V3 transcripts in the developing chick limb. RT-PCR showed expression of the predicted 313bp V3 amplicon in the wing between HH25-34 (Fig. 1A) and in situ hybridization localized low levels of V3 transcripts in chondrogenic areas of the limb core (Fig. 1B) that coincided with expression of endogenous full-length versican protein as described previously (Shinomura et al., 1990; Landolt et al., 1995; Shepard et al., 2007).

Figure 1.

Expression of endogenous versican V3 isoform transcripts in the chick limb. A: RT-PCR at Hamburger and Hamilton (HH) stages 25, 28, 34 shows the predicted 313 bp amplicon in the developing limb and heart. Heart cDNA was used to generate riboprobes for in situ hybridization. Position of the 298 base pair standard is marked by arrowhead and actin loading control for each lane at the bottom. B: In situ hybridization with antisense probe shows low level expression of V3 mRNA in the wing core at HH25 (asterisk). C: Sense control probe in an adjacent section shows lack of signal in the limb core. Note staining in the dorsal mesenchyme underlying the ectoderm represents nonspecific signal in panels (B) and (C).

Expression of recombinant versican G1 and V3 variant were selected for use in this study because all four versican isoforms contain G1 and G3 and differ in inclusion or deletion of the GAG-α and GAG-β chondroitin sulfate attachment regions. We did not attempt to express adenoviral constructs containing all or part of GAG-α or GAG-β domains as there was no certainty that chondroitin sulfate glycosylation would occur appropriately, potentially making experimental interpretation extremely difficult.

To investigate whether the versican G1 domain or short V3 isoform could function independently during limb morphogenesis and/or if one molecule was particularly critical for versican function, recombinant adenoviral constructs encoding these polypeptides were microinjected into the developing chick limb bud at HH19-25. Adenoviral titers of 2 × 103 ifu were reported to work well for functional studies of G1 and V3 effects on cardiac outflow tract development in which recombinant adenoviruses were microinjected into the anterior heart field (Kern et al., 2007), but in the present study we found this too low for efficient infection of limb mesenchyme due to rapid diffusion from the injection site. We systematically varied viral injection volume in order to determine the lowest titer needed to maintain embryonic viability and still obtain a consistent phenotype and thus used G1 and V3 adenoviruses at 1.5 × 106 ifu in combination with control LacZ encoding viruses in a total volume of 1.5 μL. In control specimens, we routinely injected the LacZ encoding adenovirus of higher titer and at the same volume without effect on the embryonic limb to rule out possible non-specific viral effects.

A minimum of 30 embryos from at least six separate experiments for each experimental and control group were evaluated for each treatment to validate the resulting phenotype (Table 1). Adenoviral-mediated G1 and V3 overexpression resulted in statistically significant peripheral expansion of humeral primordia at HH35 (Table 2) due to injections into the proximal limb bud at prechondrogenic and early chondrogenic stages. In some specimens in which the microinjection needle was withdrawn and reinserted into the limb during the same experiment, slight movement of injection site resulted in cartilage enlargement at different sites along the same humeral element. Two different phases of limb development were used for injection of recombinant adenoviruses, precartilage mesenchyme (HH19-21) and committed chondrogenic tissue (HH22-25). Injections during both intervals resulted in similar increases in skeletal anlagen thickness suggesting that overexpression of versican domains can have an effect on skeletal development even after mesenchymal cells have committed to chondrogenic fate.

Table 1. Viability and resulting phenotypes of embryos (HH35) injected with experimental G1, V3, and control LacZ adenoviruses
Adenovirus and injection StageNumber of embryos injected (n)Embryos with increased thickness of skeletal anlagen (n)aPercentage of embryos resulting in phenotype at HH35 (%)b
  • a

    Embryos included only if contributed to a statistically significant phenotype.

  • b

    Phenotype defined as increase in thickness of skeletal anlagen at HH35 of proximal, mid or distal humerus.

G1   
 HH19-2112433
 HH22-2522627
V3   
 HH19-2115426
 HH22-2515320
LacZ Control   
 HH19-211800
 HH22-251400
Table 2. Humeral measurements of limbs injected with G1, V3, or LacZ adenoviruses at HH19-25 and corresponding contralateral control (CLC) limbs (mean ± SD in mm)
AdenovirusHumeral area (n)InjectedCLCP-value
  1. Data represent mean of individual wings measured at each humeral area (n) exhibiting cartilage enlargement from 6 different experiments. Position of humeral measurements is shown in Figure 2F. Embryos (Table 1) in some instances exhibited humeral enlargement at 2 sites. In LacZ control experiments wings were measured at each of the humeral sites.

G1    
 Proximal (6)0.66 ± 0.140.45 ± 0.140.008
Mid (5)0.42 ± 0.100.35 ± 0.060.024
Distal (4)0.93 ± 0.330.83 ± 0.340.0302
V3    
 Proximal (6)0.86 ± 0.220.61 ± 0.190.002
Mid (4)0.49 ± 0.070.45 ± 0.070.018
Distal (3)0.86 ± 0.220.57 ± 0.19NS
LacZ    
 Proximal (16)0.61 ± 0.170.62 ± 0.16NS
Mid (16)0.48 ± 0.040.47 ± 0.04NS
Distal (16)0.62 ± 0.100.62 ± 0.10NS

Alcian blue and alizarin red whole mount histochemistry were used to evaluate gross morphological changes in the skeletal template. As shown in Fig. 2, at HH35 there was significant enlargement in the proximal humeral primordia of wings overexpressing the G1 domain or along its length, particularly toward the distal end (Fig. 2A and B; Table 2). Proximal and mid regions of limbs overexpressing the V3 isoform also exhibited areas of enlarged humeral growth (Fig. 2C and D; Table 2) with evidence of localized enlargement of cartilage first observed as early as HH31. Limbs injected with the β-galacotsidase-encoding control adenovirus were indistinguishable from respective uninjected contralateral controls (Fig. 2E). On average, V3-injected embryos had slightly larger increases in thickness of the skeletal anlagen but when taken in consideration relative to the contralateral control measurements this may be attributed to the variation of overall thickness of the skeletal template among individual embryos. Proximal humeral phenotypes ranged from moderate (G1-26%, V3-22%) to more dramatic expansion (G1-130%, V3-88%) as compared with contralateral control limbs and to LacZ-adenovirus injected limbs which showed an insignificant enlargement of 2%–8%. A similar pattern in extent of humeral expansion was consistent among sites. This variation in humeral enlargement is likely a result of slight variations in adenoviral uptake by limb mesenchyme in individual embryos which were handled in a similar manner.

Figure 2.

Whole-mount alcian blue and alizarin red staining of the HH35 wing demonstrating effects of recombinant versican G1 domain and V3 isoform expression in wings injected with recombinant adenoviruses at HH20 (C), HH21 (A, B, E) and HH22 (D). Targeted adenoviral injection resulted in enlargement of skeletal primordia in the proximal (A, C), distal (B) or mid humerus (D). There was little or no enlargement of the skeletal anlagen observed with the LacZ adenovirus control (E). Humeral sites from which measurements in Table 2 were taken are shown in panel (F). CLC indicates uninjected contralateral control limbs. Scale bar = 0.5 mm.

Ectopic expression of recombinant G1 and V3 proteins in areas where increases of thickness were observed was validated by immunolocalizing an antibody directed against the C-terminal hemaglutinin tag of both G1 and V3 constructs in HH35 limb sections (Fig. 3). Whole mount β-galactosidase staining was performed in order to facilitate localization of limb areas with viral infection prior to embedding/sectioning and subsequent antibody labeling (Fig. 3A and C). β-galactosidase staining has previously been demonstrated to provide reliable indication of adenoviral infection when co-injected with experimental constructs (Kern et al., 2007). In the present study, location of β-galactosidase reactivity and hemaglutinin tag-positive adenoviral expression consistently overlapped and LacZ staining also found to be a reliable marker of infection sites (Fig. 3). Expression of recombinant G1 and V3 proteins correlated with areas of increased cartilage deposition in mid-(G1, Fig. 3B) and proximal-(V3, Fig. 3D) humeral anlagen observed in whole mount alcian blue stained embryos.

Figure 3.

Whole mount β-galactosidase staining and hemaglutinin tag immunostaining shows localization of recombinant versican G1 domain and V3 isoform expression in enlarged humeral cartilages at HH35 in wings injected with recombinant adenoviruses at HH23 (C, D) and HH24 (A, B). Whole mount β-galactosidase reactivity localized sites of adenoviral infection (arrow, A; arrowhead, C). Immunostaining of hemaglutinin tagged G1 (B) and V3 (D) constructs correlated well with β-galactosidase reactivity in enlarged areas of the mid (mh; A,) and proximal (ph; C) humeral template. Inset in panel (A) shows higher magnification image of β-galactosidase positive cells indicated by the arrow. Scale bars = 250μm (A, B and C, D); 125μm, inset in A.

To suggest a possible mechanism for observed increases in chondrogenesis in areas expressing recombinant G1 and V3 proteins several parameters were evaluated, including increased mitosis, decreased apoptosis, as well as changes in hyaluronan acculmulation and/or cellular aggregation. Phosphohistone H3 antibody-labeling was used to visualize mitotic cells in sections from G1 or V3/LacZ adenoviral-infected limbs at HH21 to determine whether increased cell proliferation contributed to the observed thickening of skeletal primordia. In a series of 11 immunohistochemical experiments on 23 wing samples from 9 embryos as represented in Fig. 4, no correlation with changes in proliferation at sites of recombinant G1 or V3 adenoviral infection were observed at HH23-31 or later at HH35. Similar results were noted at HH35 for G1 and V3 adenovirus infected limbs at HH22-25 (not shown). Reduction in apoptosis during the postinfection incubation period as a result of recombinant G1 or V3 expression that could account for possible increases in chondrocyte number were also examined by TUNEL assay. As only very low levels of apoptosis have been observed previously in the wild-type chick wing at HH25 (Shepard et al., 2008), TUNEL was performed at HH30/31 when limb chondrogenesis had further progressed. Low numbers of scattered apoptotic cells were noted, but no correlation with changes in apoptotic cell number at sites of adenoviral infection were detected (not shown).

Figure 4.

Recombinant versican V3 isoform expression does not correlate with increased mitosis in humeral cross section of the HH31 wing following coinjection with recombinant V3/LacZ adenovirus at HH23. Change in overall number of antiphosphohistone H3 (ser10) positive cells (B) did not show correlation with V3/LacZ infected areas as indicated by β-galactosidase staining (boxed area in A) in either pre- and early chondrogenic stages HH23-31 or in later stage cartilaginous tissue (HH35). Scale bar = 100μm.

Because of its ability to bind hyaluronan, the effect of recombinant G1 expression on hyaluronan distribution in the developing chick limb was also assessed. Although less obvious at earlier stages, by HH35 definitive pericellular hyaluronan signal was noted surrounding individual recombinant G1-positive chondrocytes (Fig. 5A and B). G1-positive cells could also be observed in association with noninfected cells in areas of hyaluronan accumulation. In limbs coinjected with V3/LacZ adenoviruses, we were unable to ascertain localized changes in hyaluronan associated specifically with V3-positive chondrocytes at the stages examined.

Figure 5.

Cross section through the proximal humeral element of the chick wing coinjected with recombinant versican G1/LacZ at HH23 displays pericellular hyaluronan localization in peripheral limb cartilages at HH35. A: G1-hemaglutinin tag localization (arrows) associated with individual hyaluronan-positive cells or uninfected hyaluronan-positive chondrocytes (B). C: Area of infection shown by β-galactosidase reactivity. Asterisk, perichondrium. Scale bar = 25μm.

Rhodamine PNA labeling of the forming humerus at HH25 in wings injected at HH19-21 showed staining of cell aggregates within early chondrogenic primordia as expected, but in addition, PNA labeling displayed an overlapping distribution with recombinant G1 (Fig. 6A–C) and V3 (Fig. 6D–G) expression in several areas. Interestingly, multiple small PNA-positive cell clusters not observed in the contralateral control (Fig. 6H and I) were also G1- and V3-hemaglutinin tag-positive in addition to PNA/hemaglutinin tag staining in the limb core, suggesting that additional mesenchymal aggregates were undergoing incorporation into the cartilage template.

Figure 6.

Recombinant hemaglutinin-tagged versican G1 and V3 proteins codistribute with PNA-binding materials in sagittal sections through humeral cartilage condensations of the HH25 wing following injection with recombinant adenoviruses at HH19 (A–I). Double labeling with antihemaglutinin shows G1 (A) and V3 (D) overlapping with PNA in chondrogenic areas (B, E; overlay in C, F). Small clusters of cells colabeled with both antihemaglutinin tag and PNA are noted at the edges of chondrogenic regions in both G1 and V3 infected limbs (arrowheads). Enlarged view of area indicated by arrowheads in (F) is shown in (G). No hemaglutinin-tag immunoreactivity (H) is noted in the PNA-positive chondrogenic core (I) of noninjected contralateral control limbs (CLC). Scale bars = 100μm in C, 25μm in G.

DISCUSSION

It is well-documented that versican is expressed during early stages of limb development (Shinomura et al., 1990; Shibata et al., 2003; Snow et al., 2005). Endogenous versican is expressed in the ECM during formation of prechondrogenic condensations (Kimata et al., 1986) with expression reduced and replaced by aggrecan as mesenchymal cells differentiate into chondrocytes. As overt limb chondrogenesis continues, versican expression persists in perichondrial and joint interzone tissues (Shinomura et al., 1990; Yamamura et al., 1997; Shibata et al., 2003; Snow et al., 2005; Shepard et al., 2007). Although several studies have reported the importance of versican to early stages of limb skeletogenesis, the mechanism by which versican is involved is yet unclear.

We undertook the present study to determine whether N- and C-terminal versican domains could function independently of the mature proteoglycan when overexpressed during chick limb skeletogenesis and hypothesized that expression of individual G1 domains and V3 isoform would facilitate limb chondrogenesis in vivo. Our results showed that overexpression of G1 and V3 versican had growth promoting effects on humeral skeletal anlagen suggesting that individual versican domains may function during establishment of cartilages in early stages of wild type limb skeletogenesis. Replication-incompetent recombinant adenoviruses were used to express the versican G1 domain and V3 variant, comprised of only G1 and G3 domains, in the proximal limb at HH19-21 (mesenchymal cell stages) and HH22-25 (precartilage condensation and early chondrogenic stages). At both intervals adenoviral-mediated expression of either the G1 domain or V3 variant of versican resulted in enlargement of humeral cartilages and skeletal primordia due to localized increases in cartilage formation. Versican domains have been shown previously to function independently when expressed ectopically (Zhang et al., 1998, 2001; Ang et al., 1999; Yang et al., 1999; Kern et al., 2007) and increasing evidence suggests that versican may be proteolytically processed in different tissues which may enable individual versican domains to function apart from the intact proteoglycan (Sandy et al., 2001; Kern et al., 2006, 2007, Capehart, 2010). Indeed, versican proteolysis by ADAMTS family members liberates a fragment containing the G1 domain (Sandy et al., 2001).

In the present study, several potential mechanisms that could lead to increased cartilage size due to recombinant versican G1 and V3 overexpression were investigated. G1 and G3 domains have both been shown to enhance proliferation (Zhang et al., 1998, 1999; Ang et al., 1999; Wu et al., 2002) and localized increases in mitosis of precartilage mesenchyme and/or chondrocytes could possibly account for increased humeral size. Proliferative cells were found scattered throughout the proximal limb at several developmental stages but distribution of mitotic cells did not show significant correlation with areas of recombinant G1 and V3 expression as was also reported in the embryonic outflow tract (Kern et al., 2007). Reduction in number of apoptotic cells was also examined and though small numbers of scattered apoptotic cells were detected in the forming humerus, no obvious changes were noted in G1-, V3-, or LacZ- infected areas. This does not rule out possible reduction in apoptosis at other stages, but apoptotic level in relevant areas of the proximal limb appear low overall during the period spanned by the present study and thus is unlikely. In addition, by subjective assessment of chondrocyte-chondrocyte proximity in 10 fields from three separate experimental treatments only slight differences in accumulation of interstitial cartilage matrix between individual chondrocytes in areas of recombinant G1 and V3 expression at HH35 were apparent (see Fig. 3), suggesting that increased overall matrix deposition by individual chondrocytes alone was not responsible for increased local cartilage growth. The absence of data to support increased cell proliferation or decreased apoptosis in these studies points to a third mechanism in which versican G1 domain and V3 isoform facilitate aggregation of cartilage progenitors within the humeral primordium. As there is growing appreciation of the role of extracellular matrix in regulating gene expression and pattern formation, it is possible that expression of these recombinant matrix proteins by adenovirally infected cells in developing skeletal primordia could impact cellular behavior in the surrounding cell population.

Versican has been shown important for chondrogenesis in vitro, particularly with regard to formation of precartilage mesenchymal aggregates (Williams et al., 2005; Kamiya et al., 2006). In contrast, inhibition of versican in chick limb mesenchyme in vitro had no effect on mesenchymal aggregation (Zhang et al., 1998). Several factors can affect direct comparison of these studies, for example, different cell lines, plating densities, and use of protein knockdown techniques. The discrepancies and difficulty in interpreting the in vivo relevance of work in vitro emphasized the importance of attempting use of an in vivo model in which to evaluate versican function in the developing limb. One such study showed that targeted morpholino-mediated knockdown of versican in ovo inhibited precartilage mesenchymal condensation in the embryonic chick limb (Shepard et al., 2008).

PNA has been widely used as a marker for identifying aggregating precartilage limb mesenchyme and early chondrogenic foci (Zimmermann and Theis, 1984; Aulthouse and Solursh, 1987; Capehart et al., 1997). PNA labeling of the developing humerus at HH25 showed overlapping localization with recombinant G1- and V3 expression in the proximal limb core and also in “extra” small clusters of cells around the edges of the forming humerus. Colocalization of PNA-binding with G1 and V3-positive cells suggests that expression of recombinant versican domains led to localized increases in PNA-positive aggregates, resulting in additional cartilage formation that could account for chondrogenic expansion at specific humeral sites. Our observations are in agreement with the results of other studies in which virally mediated ectopic V3 isoform expression was reported to increase cell adhesion in vitro (Lemire et al., 2002; Kern et al., 2007) and thickness of the outflow tract myocardium in vivo (Kern et al., 2007). On the other hand, ectopic V3 expression reduced cartilage formation in a chondrogenic cell line in vitro, perhaps by competing with endogenous versican (Kamiya et al., 2006). In the present study, however, recombinant V3 overexpression continued well beyond the period when endogenous V0/V1 versican is found in proximal limb cartilage matrix in vivo, and so perhaps facilitated continued aggregation of cartilage-forming progenitors into the peripheral humeral primordium.

In the present study, obvious differences in hyaluronan were not readily observed in areas of recombinant versican domain expression until later chondrogenic stages where increased pericellular hyaluronan was noted about well-rounded G1-positive chondrocytes; thus it is uncertain whether recombinant versican G1 domain expression impacted significant hyaluronan accumulation at earlier stages. On the other hand, if hyaluronan was incrementally stabilized through increased binding of virally-expressed G1 domain, it is possible that additional mesenchyme were gradually incorporated into cellular aggregates that contributed to an overall increase in chondrocyte number in specific humeral locations. Indeed, versican/hyaluronan complexes in vitro have been shown to enhance recruitment of stromal cells during mammary tumor neovascularization (Koyama et al., 2007).

To our knowledge, this study is the first attempt to examine versican G1 domain and V3 isoform effects on limb skeletogenesis in vivo. It is interesting to note that although versican G1 and G3 domains have been reported to have differing activities in vitro, our data show that G1 and V3 both have gain-of-function effects on chondrogenesis in vivo and suggests that the mechanism by which endogenous versican is involved in limb development occurs by way of effecting chondrogenic condensations as shown in previous studies in vitro (Williams et al., 2005; Kamiya et al., 2006) and in vivo (Shepard et al., 2008). Alternatively, expression of recombinant versican domains could perhaps bind to other endogenous molecules within the limb resulting in their loss-of-function with similar overall effect. The observed enlargement of skeletal primordia resulting from overexpression of both recombinant G1 and V3 also suggests that versican domains are capable of acting independently of intact V0 and V1 isoforms during embryonic limb development.

Acknowledgements

The authors thank Dr. Said Said for assistance with statistical analysis.

Ancillary