Matrix metalloproteinases regulate morphogenesis, migration and remodeling of epithelium, tongue skeletal muscle and cartilage in the mandibular arch

Matrix metalloproteinases (MMPs) have been implicated in pathological and invasive extracellular remodeling of tissues in vivo and in invasive and migratory events in culture. However their roles in normal development and morphogenesis have been more elusive. Although these proteinases and their inhibitors are widely expressed during embryonic development, few experiments have addressed whether they are required for coordination of normal tissue growth and remodeling. The first branchial arch, which gives rise to the complex facial structure of the mandibular and maxillary processes, presents an intriguing challenge for elucidating events involving MMPs and extracellular matrix (ECM). The program of invasion of cranial neural crest (CNC), prechordal (paraxial) mesoderm and epithelium, and the morphogenesis and growth of skeletal muscle (tongue), cartilage, bone and teeth all involve ECM interactions. CNC cells migrate into the forming branchial arch and become determined for a number of developmental fates including chondrogenesis, osteogenesis and odontogenesis. The cranial paraxial mesoderm that colonizes the mesenchymal core of the first branchial arch contributes to the skeletal muscle of the tongue and other structures (Trainor and Tam, 1995; Trainor et al., 1994; Noden 1991). CNC-derived ectomesenchyme and sensory nerve cells infiltrate the first branchial arch from embryonic day (E; days post-coitum) 8.5 (3-6 somite embryos) through E10 (30-35 somites) in the developing mouse embryo (Lumsden, 1987).

Initial regulation for the timing and position of chondro-genesis, osteogenesis and tooth formation appears to controlled by inductive signals from mouse oral mandibular epithelia and the interaction with the underlying mesenchyme at E9-E11 (Imai et al., 1996; Hall, 1991; Hall and Ekanayake, 1991; Kollar and Mina, 1991). The ECM and epithelial cell basement membrane are potential sources of signals that guide migrating CNC to cellular condensations. Gelatinase B, collagenase-3 and tissue inhibitor of metalloproteinase (TIMP)-1 are highly enriched in bone (Reponen et al., 1994; Gack et al., 1995; Flenniken and Williams, 1990) and TIMP-2 is enriched in cartilage (Zafarullah et al., 1996), while TIMP-3 is widely expressed in epithelial tissues (Apte et al., 1994). Mesenchy-mal tissues frequently express stromelysins-1 and −3 and gelatinase A (Lefebvre et al., 1992; Reponen et al., 1992). How the ECM remodeling program is involved in the processes of CNC and mesoderm migration and in bone, cartilage, tooth and tongue formation during development has not been elucidated. In this study we have investigated the developmental expression of MMPs and TIMPs during mandibular differentiation and have obtained the first evidence of the morphogenetic processes in which they participate.

Timed pregnant Swiss Webster mice (Charles River Laboratories, Wilmington, MA.; Simonsen Laboratories, Gilroy, CA) were mated between 7 am and 10 am on day 0. Mice with vaginal plugs were killed on the day specified and the mandibles were dissected from the embryos (Slavkin et al., 1989) and processed for RNA isolation or explanted into culture.

Up to 23 mandibles from E9 through E15 embryos were pooled for each time point and immediately put into RNAzol B (Biotecx Laboratories, Inc., Houston, TX) on ice. RNA was prepared according to the manufacturer’s directions. RNA concentration was determined by absorbance at 260 nm or by comparison of ethidium bromide fluorescence staining with a known concentration of yeast tRNA (Maniatis et al., 1982).

RNA was reverse transcribed according to a modification of the method of Rappolee et al. (1989). The 20 μl reaction consisting of 1-2.5 μg RNA (heat-treated at 70°C for 5 minutes with 0.5 μg oligo dT; Collaborative Research/Becton Dickinson, Bedford, MA), 10 mM dithiothreitol, 0.5 mM each dATP, dCTP, dGTP, dTTP, 10 μg acetylated bovine serum albumin in reverse transcriptase buffer (50 mM Tris-HCl, pH 8.3, 75 mM KCl, 3 mM MgCl₂) and SuperScript RNase H reverse tran-scriptase (200 units, GIBCO BRL, Gaithersburg, MD) was incubated for 1 hour at 37°C. The RT reaction was stopped by heating to 93°C for 3 minutes. RNA was omitted from control reactions.

The resulting cDNA was amplified by the hot start method (D’Aquila et al., 1991). The reaction consisting of 200 μMeach dATP, dGTP, dCTP, dTTP, 1.8 μMeach 5′ and 3′ primers in PCR buffer (10 mM Tris, pH 8.3, 50 mM KCl and 2.5-4 mM MgCl₂, depending on the specific pair of primers) and 100 ng (or less) cDNA was overlaid with mineral oil and heated to 80°C in an Omnigene thermocycler (Hybaid, Middlesex, England). Amplitaq (1 unit, Perkin-Elmer Cetus, Emeryville, CA) was added (50 μl total reaction vol) below the oil. Control reactions did not contain cDNA. Samples were amplified for 35 cycles (1 cycle consisting of 1 minute at 95°C, 30 seconds at the specific annealing temperature and 45 seconds extension at 72°C) followed by a final extension of 7 minutes at 72°C, before electrophoresis in a 3% NuSieve GTG: 1% Seakem ME agarose gel (FMC, Rocklin, ME) in 0.04 M Tris-acetic acid, 0.002 M EDTA, pH 8, at 180 constant volts. Gels were stained in ethidium bromide. Sizes of gel bands were determined by comparison to 1 kb DNA ladder standard. Sequences for the oligonucleotide primers, annealing temperatures, final MgCl₂ concentration in the PCR reaction and expected amplimer length are as follows. Gelatinase A: 5′ primer TTC TTC TTC AAG GAC CGG TTC ATT TGG, bp 1409-1436; 3′ primer ATT TTC TTC TTC ACC TCA TTG TAT CTC CA, bp 1718-1747, 60°C, 4 mM MgCl₂, 339 bp amplimer (Collier et al., 1988). Gelatinase B: 5′ primer CGC TCA TGT ACC CGC TGT ATA GCT AC, (bp 1277-1302); 3′ primer TAG AGG CCT CAG AAG AGC CCG CA, bp 1575-1597), 60°C, 4 mM MgCl₂, 320 bp amplimer (Tanaka et al., 1993).Stromelysin-1: 5′ primer GAC AAA TTC TGG AGG TTT GAT GAG A, bp 1106-1130; 3′ primer ACC AGC TGT TGC TCT TCA ATA TGT

G, bp 1289-1313, 65°C, 4 mM MgCl₂, 207 bp amplimer (Sympson et al., 1994). Stromelysin-3: 5′ primer CCC TGC ACC ACT CTC CAA GC, bp 1140-1159; 3′ primer CCT ACG GGG CGA GGA AAG CG, bp 1423-1442, 60°C, 2.5 mM MgCl₂, 303 bp amplimer (Lefebvre et al., 1992). TIMP-1: 5′ primer ACC ACC TTA TAC CAG CGT TA, bp 207-226; 3′ primer AAA CAG GGA AAC ACT GTG CA, bp 501-520, 55°C, 2.5 mM MgCl₂, 314 bp amplimer (Gewert et al., 1987). TIMP-2: 5′ primer GGT CTC GCT GGA CAT TGG AGG AAA G, bp 587-611; 3′ primer GGG TCC TCG ATG TCC AGA AAC TCC TG, bp 924-949, 60°C, 2.5 mM MgCl₂, 363 bp amplimer (Boone et al., 1990).

The sources and preparation of mouse anti-sense probes were described previously (Alexander et al., 1996). For collagenase-3, a 730 bp cDNA insert cloned into PstI/SacII site of pBS was linearized with DraI and transcribed with T7 (Gack et al., 1995). In situ hybridization was performed on 7-8 μm sections from tissue fixed in 4% paraformaldehyde at 4°C overnight and paraffin embedded. Sections were hybridized with ³⁵S-labelled cRNA probes as described by Frohman et al. (1990), with some modifications (Alexander et al., 1996). Hybridization and washes were performed at 55°C and 65°C, respectively, for all probes except TIMP-1 and gelatinase A, which were hybridized overnight at 50°C and 47°C, respectively, and washed at 44°C. In situ studies with TIMP-3 probe were performed as described by Helms et al. (1994). Hybridization was done overnight at 55°C and washes at 64°C.

Mandibular arches from E10 (or E11 as stated in the Results section) embryos were cultured in BGJb medium (GIBCO-BRL) supplemented with 0.1 mg/ml ascorbic acid, 10 mM Hepes and penicillin-streptomycin, pH 7.4 (Slavkin et al., 1989; Shum et al., 1993). The conditioned medium (CM) was collected and changed at 2, 4, 6 and 8 days in culture and was designated as embryonic day of harvest of the explant plus days in culture (e.g. E10+6). For inhibitor experiments, stock solutions of 400 mM metalloproteinase inhibitor (MPI, 3-(N-hydroxycarbamoyl-2(R)-isobutyl propionyl-L-tryptophan methylamide) and 40 mMcontrol analogue, metalloproteinase inactive control compound (MIC, N-tert-butyloxycarbonyl)-L-leucine-L-tryptophan methylamide) (gifts from Dr Richard Galardy, Glycomed, Inc., Alameda, CA) were prepared in dimethylsulfoxide (DMSO). Thiorphan was prepared in 3% (vol/vol) ethanol; phosphoramidon, aprotinin, leupeptin and amiloride (Sigma, St. Louis, MO) were dissolved in medium. Working dilutions of all inhibitors and comparable DMSO and ethanol controls were made in culture medium. All experiments were done at least twice with 3-4 explants for each treatment.

For whole-mount staining of Meckel’s cartilage, cultured explants were processed and stained with Alcian Blue (Sigma) as described by Shum et al. (1993). For morphologic analysis, cultured explants (with adherent filter disc) were fixed, washed in distilled water, dehydrated in a graded series of ethanol, cleared in xylene or toluene and embedded in paraffin. For subsequent immunochemistry or in situ hybridization, sections were deparaffinized in xylene and hydrated in 100%, 95%, 75% ethanol and water.

Proteinase activity in the CM was analyzed by gel zymography (Adler et al., 1990; Chin et al., 1985) in a 10% or 12% Laemmli SDS-poly-acrylamide gel containing 1 mg/ml of protein substrate. The substrates used to detect gelatinases, caseinases and plasminogen activators were respectively, gelatin (type A, porcine skin; Sigma), α-casein (bovine milk; Sigma), and α-casein with 10 μg/ml beef plasminogen (Sigma). CM from equivalent numbers of cultured mandibles were compared. To detect TIMP inhibitory activity, 10% rabbit skin CM, activated by APMA (Unemori and Werb, 1988), was incorporated into a 12.5% SDS-polyacrylamide gel containing gelatin. CM from the mandible cultures was concentrated by quinine sulfate precipitation (Werb et al., 1986) prior to analysis.

To detect TIMP-1, stromelysin-1 and gelatinase A proteins in CM from mandible cultures, CM (200-1100 μl) was concentrated by quinine sulfate precipitation (Werb et al., 1986) or dehydration in a dialysis bag immersed in Aquacide (Calbiochem, La Jolla, CA), dissolved in sample buffer containing 5% β-mercaptoethanol (Laemmli, 1970), separated on 10% or 12.5% SDS-polyacrylamide gels and transferred onto Immobilon-P membranes (Millipore, Bedford, Mass.) in 25 mM Tris, pH 8.3, 192 mM glycine, 15% methanol, at 300-350 mA, 60-90 minutes. Blots were blocked overnight at 4°C or room temperature, in 5% bovine serum albumin (BSA) in PBST and 0.02% sodium azide and washed (0.5 mM Tris-HCl, 1.5 M NaCl, pH 7.65, followed by two changes of the same buffer containing 0.5% deoxycholate, 0.2% Triton X-100, and 0.1% SDS) prior to incubation with the primary antibody diluted in 1.5% BSA in PBS. The primary antibodies used were mouse monoclonal IgG anti-human stromelysin-1 (clone 188.2, 1:200 dilution, gift from Scott Wilhelm, Miles Research, New Haven, CT; Wilhelm et al., 1992), mouse monoclonal IgG anti-human gelatinase A, (1:200 dilution, Molecular Oncology, Gaithersburg, MD), and rabbit IgG anti-mouse TIMP-1 (1:200 dilution, made from rabbits immunized with a peptide of the 12 C-terminal amino acid residues of mouse TIMP-1 conjugated to keyhole limpet hemocyanin; C. Alexander and Z. Werb, unpublished data). Blots were incubated with the secondary antibody, biotinylated goat IgG anti-mouse or anti-rabbit IgG (1:4000 dilution in 1.5% BSA/PBS, Sigma), then reacted with horseradish per-oxidase-conjugated streptavidin (1:2000 in PBS, Amersham UK). Blots were washed between each incubation as described above. Bands were detected by enhanced chemiluminescence reagents on Hyperfilm-ECL (Amersham Life Science, Arlington Heights, IL).

Explants were fixed in 10% neutral buffered formalin at 4°C overnight, embedded in paraffin and 5-7 μm serial sections were cut. Sections were deparaffinized and hydrated according to the procedure above and endogenous peroxidase activity was quenched with 2% hydrogen peroxide in methanol for 30 minutes. Background activity was blocked sequentially with 0.5% casein and 10% goat serum in PBS. The serum was removed, but not washed off, prior to incubation with 1/5 dilution of mouse monoclonal anti-human desmin (clone II, ICN, Costa Mesa, CA) in 10% goat serum. Sections were washed with 0.1% Tween 20 in PBS (PBST) followed by PBS, incubated with 1/300 dilution of biotinylated goat anti-mouse IgG (Vector Labs, Burlingame, CA) in 5% BSA in PBST, washed again, reacted with the Vectorstain Elite ABC reagent (Vector Labs), washed, incubated with diaminobenzamidine reagent (Vector Labs), rinsed and finally counterstained in 20% Gill’s hematoxylin (Sigma). Fourteen control and nine MPI-treated mandibles were serially sectioned after 2, 4, 6, 8 and 9 days culture; representative sections are shown.

When total RNA from mandibles of E9-E15 embryos was analyzed by RT-PCR (Fig. 1A), we first observed gelatinase A mRNA at E9 that then increased significantly between E12-E14 after epithelial-mesenchymal interactions had taken place. Gelatinase B expression was also evident at E14. Stromelysin-1 mRNA was barely detected until E12, and then increased to

E15, while stromelysin-3 mRNA was expressed at E15. TIMP-1 and TIMP-2 mRNAs were expressed as early as E9 and attained maximal levels at E14. To determine which cells express these MMPs and TIMPs in the developing mandible in vivo, we used in situ hybridization analysis with ³⁵S-labelled antisense cRNA probes. Gelatinase A mRNA was initially diffuse throughout the mandible at E12, but showed a stronger, more restricted pattern of expression from E13 to E15 (Fig. 2A-F). Signal was intense in the perichondrium of Meckel’s cartilage (Fig. 2D), the osteogenic mesenchyme (Fig. 2C,D), the mesenchyme surrounding developing tooth buds (Fig. 2E,F), along the medial sulcus and base of the tongue, in a distinct layer underlying the outer mucosal epithelium of the tongue (the collagenous lamina propria; Strachan, 1994) (Fig. 2B) and in the mesenchyme below the epithelium of the oral sulcus (Fig. 2A,B). Little or no signal was detected in Meckel’s cartilage proper (Fig. 2C,D), genioglossus muscle (Fig. 2C) or in the epithelium of the oral surface, tongue and invaginating tooth buds (Fig. 2A,E,F).

Gelatinase B mRNA labelling was diffuse in E10-E12 mandibular arches, strong in the mesenchyme surrounding the incisors at E13 (data not shown), in osteoclasts in the developing bone surrounding Meckel’s cartilage (Fig. 2G,H), and the mesenchyme surrounding the molar tooth buds at E14, but was not present in incisor epithelium (Fig. 2G), oral epithelium, tongue epithelium or Meckel’s cartilage (data not shown).

Stromelysin-1 and stromelysin-2 mRNA were diffusely expressed in E9-E15 mandibles (data not shown). At E14, stromelysin-3 was localized to the inner muscle of the tongue, genioglossus muscle, and mesenchyme around the tooth buds and developing bone (Fig. 3A-E). Collagenase-3 mRNA gave only a diffuse signal in the mandible at E12, but was highly localized in a few osteoblasts along the edges of the osteoid at E14 (Fig. 3F). Interestingly, a much stronger signal for collagenase-3 mRNA was detected in the perichondrium and periphery of the hypertrophic cartilage zone of the axial skeleton and limbs (data not shown), than in the facial skeleton. TIMP-1 and TIMP-2 mRNA transcripts were localized to the ossifying areas along the outer edges of Meckel’s cartilage at E14-E14.5 (Fig. 4A-D). Some TIMP-1 mRNA was also seen in immature chondrocytes at the frontal (anterior) edge of Meckel’s cartilage (Fig. 4B, arrows). TIMP-2 mRNA transcripts were found along the perichondrium, in the mesenchyme around Meckel’s cartilage and the incisors, in the lamina propria layer of the tongue, tongue muscle and the genioglossus muscle, osteoid surrounding the incisors and in the dental papilla.

TIMP-3 mRNA was expressed intensely in the mesenchyme in the pre-osteoid region and along the medial axis of the tongue of E12 mandibles (Fig. 4E). In E13 (data not shown) and E14 mandibles, mRNA transcripts were observed in the mesenchyme surrounding Meckel’s cartilage, incisors and osteoid (Fig. 4F-H) and in the muscle of the tongue proper, the genioglossus muscle and the mesenchyme surrounding the perichondrium. No TIMP-3 mRNA was seen in Meckel’s cartilage, dental papilla or the oral, nasal and tongue epithelium.

We next determined whether the regulated pattern of MMP and TIMP expression was recapitulated during mandibular development in culture. E10 mandibles cultured for 8 days (E10+8) in a chemically defined, serumless medium increase their cell number 20-fold and protein content two-fold, and develop with the timing and pattern of chondrogenesis, osteogenesis, odontogenesis and glossogenesis in a manner similar to that seen in vivo (Slavkin et al., 1989; Slavkin, 1990). In addition, mandibular explants have the advantage of being accessible to experimental perturbation.

By RT-PCR, the relative expression of gelatinase A mRNA changed little, while that of gelatinase B mRNA, present at low levels, increased from E10+2 through E10+8 (Fig. 1B). Stromelysin-1 mRNA was not detectable until E10+4. Trace amounts of stromelysin-3 mRNA transcripts were observed at E10+2, increasing slightly to E10+8. TIMP-1 mRNA was also evident by E10+2 and increased significantly by E10+4. This pattern parallelled TIMP-1 expression in vivo between E9 and E10. TIMP-2 was abundant as early as E10+2, and increased with time in culture.

By in situ hybridization, E10+6 mandibles (Fig. 2I) had strong expression of gelatinase A mRNA, similar to E14-E15 mandibles in vivo (Fig. 2A-F). Signal was confined to the mesenchyme, particularly around the molar and incisor tooth buds, Meckel’s cartilage and osteoid region, in the oral sulcus and lamina propria of the tongue. These results indicate that the MMP and TIMP expression pattern during mandibular development is faithfully maintained during differentiation in culture.

The regulated expression of mRNA for MMPs and TIMPs was paralleled by the functional activity by gel zymography. Gelatinase A was the major MMP secreted in both the unactivated (72 kDa) and activated (60 kDa) forms, increasing through E10+8 (Fig. 5A,B). Gelatinase B, stromelysin-1 and urokinase-type plasminogen activator (uPA) activities were present by E10+4, increasing through E10+8 (Fig. 5A,C-E). Although caseinolytic activity of stromelysin-1 was low, its presence was detectable by immunoblotting (Fig. 5D).

TIMP-2 was the predominant inhibitor made by E10 explants (Fig. 6A). In contrast, very little TIMP-1 was detected by reverse zymograms and western blots (Fig. 6C). The low level of functional activity contrasts with the abundance of TIMP-1 mRNA in vivo and in culture. However, when E11 mandibular arches were explanted, TIMP-1 activity was expressed at higher levels (Fig. 6B).

We next determined the roles of MMPs in morphogenesis by perturbing function. A synthetic peptide hydroxamic acid analogue of the transition state of zinc metalloproteinases, MPI (Alexander et al., 1996, Grobelny et al., 1992), inhibited the activities of MMPs found in CM of E10+6 mandibles in zymograms (Fig. 7A). Interestingly, while the K_i for gelatinases in solution is 0.4 nM, inhibition of these gelatinases in zymograms was much less effective. Inhibition was evident at 0.04 μMincreasing to 99% efficacy at 1 μM.

Normally, the tongue bud originates as two lateral swellings along the midsection of the first branchial arch that grow larger and merge to form the medial sulcus, which then fuses. The epithelium at the outer edges of the tongue bud proliferates into the mesenchyme beneath and the cells along this U-shaped lamina degenerate, forming a groove or oral sulcus that frees the tongue from the floor of the mouth (Sperber, 1989). The rods of Meckel’s cartilage also elongate during development in vivo and in culture, and the angle between them becomes more acute, resulting in a V-shaped structure (Fig. 7B, dotted lines). E10 mandible explants cultured in 0.2-20μMMPI exhibited marked dose-dependent effects on morphogenesis, particularly inhibition of tongue morphogenesis (Fig. 7C). Effects were detected in explants cultured with 0.2 μM

MPI. In explants treated with ≥2μMMPI, the oral sulcus did not form, resulting in a primitive ‘tongue-tied’ tongue phenotype that was apparent by 4 days (Fig. 7D). The tongue bud persisted as swellings with an unfused medial sulcus. Although development was arrested or retarded at this stage, the explant size was generally comparable to controls, suggesting normal growth with defective or delayed differentiation. However, at 20 μMMPI, the mandibles were smaller and retained their original primitive rectangular shape (Fig. 7C). When we quantified the phenotypic morphology of the explants, no obvious gross defects were seen in control mandibles or mandibles cultured with 0.04 μM MPI. Abnormalities were detected by E10+3 with 0.4 μM MPI (Fig. 7E), and increased in a concentration- and time-dependent manner (data not shown). By E10+6, 72% of mandibles cultured with 0.2 μMMPI exhibited deficiency in morphogenesis, increasing to 90% at ≥2 μM.

In contrast, E10 mandibles cultured with inhibitors of trans-membrane zinc metalloendopeptidases (phosphoramidon and thiorphan, Fig. 8A), serine (aprotinin,10-100 μg/ml; ecotin, 5-50 μg/ml, data not shown) or cysteine proteinases (leupeptin, 1-10μM) or uPA (amiloride, 1-10 μg/ml) (Fig. 8B), did not exhibit any gross morphological changes. The control compound MIC, which is 10⁶-fold less active than MPI, had no detectable morphological effect.

To further explore the MPI effect on tongue and oral sulcus formation, we examined expression of desmin, an intermediate filament protein associated with early myogenic differentiation in myoblasts (Mayo et al., 1992; Babai et al., 1990). Both control and 2 μM MPI-treated explants at E10+2 showed migrating myoblasts at the lateral edges of the oral surface of the explant and at the base of the developing tongue bud (data not shown).

By E10+4, the tongue in control explants became more prominent with the development of the oral sulcus. Many desmin-positive cells were localized to the mesenchyme within the tongue bud and at the base of the tongue (Fig. 9A), with a trail of stained cells along the ventral region of the mandible, suggesting the migration and/or proliferation of myoblasts to the developing tongue. Myoblasts were also present on the sides of the mandibles. In contrast, MPI-treated mandibles retained their primitive phenotype at E10+4 (Fig. 9B). Myoblasts were seen near the oral surface and at the base and sides of the tongue, however, the prospective tongue remained as swellings.

By E10+6, the oral sulcus of the control mandibles deepened, freeing the now larger and defined tongue from the floor of the growing explant, and the medial sulcus was fused (Fig. 9C). Desmin was concentrated primarily within the tongue and the posterior regions of the mandibular processes. The MPI-treated mandibles also grew in size and volume, however, the medial sulcus persisted as a cleft (Fig. 9D). Clusters of myoblasts were evident on the oral side (top) of the explant, within the explant and along the ventral regions leading to the base of the lingual swellings and within the bilateral swellings themselves. This pattern of desmin staining at E10+6 was similar to that seen in the controls at E10+4. However, this delay in migration/proliferation was overcome by E10+9, even though the medial sulcus remained unfused, compromising further development of the tongue (data not shown).

In contrast to the effects of MPI on tongue formation, there were no discernible effects on tooth buds or bone (osteoid) formation (data not shown); however, Meckel’s cartilage development was compromised. This template for the future mandible is normally composed of three parts (Fig. 7B). The bilateral rods are the initial segments to develop (Chai et al., 1994). These rods elongate anteriorly and posteriorly to join the developing posterior segments that will ultimately make up two of the inner ear bones. The triangular frontal anterior segment forms last (Miyake et al., 1996). As development proceeds, the lateral rods straighten out, the anterior segment elongates further and fuses with the rods. In E10+6 explants cultured with 2 μMMPI, the anterior segment was either absent or only faintly stained by Alcian blue in >90% of the samples, indicative of reduced sulfated proteoglycan deposition (Fig. 10A,C). Staining of Meckel’s cartilage rods was discontinuous or faint, suggesting reduced cartilage development and/or inhibition of fusion of its segments (Fig. 10B,C). The effect of 20 μM MPI upon initial formation of the lateral rods was already obvious at E10+2 (data not shown). However, by E10+8, this defective phenotype was less apparent. Explants treated with amiloride, aprotinin and leupeptin for 5 days developed normally (Fig. 10B). These results suggest that MMP-mediated development of Meckel’s cartilage was only temporarily inhibited or retarded by MPI.

DISCUSSION

Our investigations show that an array of MMPs and TIMPs are expressed in the developing murine first branchial arch and that metalloproteinases contribute to migration and invasion of epithelium and cranial paraxial mesoderm, and the morphogenesis of the tongue and Meckel’s cartilage during development of the mandibular arch. Normal growth and development of the first branchial arch into a differentiated mandible that contains epithelium, tooth buds, condensing mesenchyme, tongue, cartilage and bone is governed by numerous processes including cell migration from the CNC and prechordal mesoderm, cell interactions with ECM components, epithelialmesenchymal interactions mediated by signals residing in the basal lamina and formation of cell condensations. The perichondrium and underlying cartilage of Meckel’s cartilage are ultimately degraded and replaced with bone by intramembraneous ossification (Frommer and Margolies, 1971). Tongue and tooth formation involve epithelial invagination into the underlying oral mesenchyme accompanied by mesenchymal cell condensations. Any changes in prechordal mesoderm and CNC migrations, cell numbers, expression of early patterning genes, growth factors and/or ECM molecules are likely to underlie craniofacial abnormalities including prognathic or micrognathic mandibular growth and cleft lip and palate. Our investigations suggest that these migratory and remodeling processes are mediated by MMPs and TIMPs in vivo and in cultured explants.

None of the MMPs and TIMPs that we examined were expressed in ectodermally derived epithelium. Stroma underlying epithelium expressed abundant gelatinase A by E9 in the first branchial arch and its mRNA increased during morphogenesis in vivo. Although its expression was initially diffuse, intense signal was localized to the mesenchyme surrounding the developing cartilage, teeth, bone, oral sulcus and in the tongue by E13 in vivo and in cultured mandibular explants, which secreted gelatinase A in an active form. Similar expression patterns have been reported by Reponen et al. (1992).

Stromelysin-3, gelatinase A and TIMP-2 mRNA transcripts were all expressed in the underlying lamina propria (the connective tissue layer that penetrates and connects the oral surface mucosa to the inner skeletal muscle of the tongue) of the developing tongue. There was a striking striped pattern of stromelysin-3 and TIMP-3 mRNA expression in the inner muscle fibers of the tongue proper and the genioglossus muscle which connects the rear of the tongue to the floor of the mouth. Expression of stromelysin-3 is frequently associated with apoptosis (Lefebvre et al., 1992). We have found apoptotic cells in the developing tongue (unpublished observations). It is possible that the complex stereotyped myotube structure is the result of selective death of myocytes (McClearn et al., 1995), triggered by degradation of ECM surrounding these cells.

The mature chondrocytes of Meckel’s cartilage did not express any of the MMPs or TIMPs studied. However, mRNA transcripts of gelatinase A and TIMP-2 were detected in the perichondrium, where regulated breakdown of the stroma of the fibroblastic sheath may be necessary for the increasing growth of the cartilaginous structure.

Bone undergoes extensive and rapid remodeling in the embryonic mandible. Gelatinase B, collagenase-3 and stromelysin-3 mRNA were all expressed in the osteoidal region, along with TIMP-1, −2 and −3. Expression of gelatinase B was concentrated mainly in the osteoclasts in the ossifying bone of the developing mandible and around the incisors and molar tooth germs (E13-E14), as noted previously (Reponen et al., 1994). We also saw gelatinase B expression in cultured mandibles.

Osteoblasts express collagenase-3 (MMP-13) in areas undergoing both intramembraneous (mandible, maxilla) and endochondral (axial skeleton, limbs) bone formation. Our study also found expression in osteoblasts in the osteoid of the mandible at E14 (Gack et al., 1995; Mattot et al., 1995). However our data point to a much higher expression during endochondral bone formation than intramembraneous bone formation. By contrast, localization and expression of TIMP-1 in the osteoblast population of ossifying bone is similar in both types of bone formation (Flenniken and Williams, 1990). Interestingly, only TIMP-2, but not TIMP-1 expression was observed in the perichondrium of Meckel’s cartilage, while both TIMP-1 and −2 were seen in the perichondrium of the ribs.

Whether these differences are due to the CNC origins of bone in mandible vs. mesoderm origins of long bones, or to intramembraneous vs. endochondral bone formation, remain to be determined.

MPI inhibits the activities of MMPs (Fini et al., 1991; Librach et al., 1991; Alexander et al., 1996). Morphogenesis of tongue and oral sulcus were specifically inhibited by MPI but not by inhibitors of serine and cysteine proteinases, suggesting that MMPs are involved, directly or indirectly, in these developmental events.

The normal fusion of the bilateral lingual processes was also inhibited by MPI, resulting in retention of the two epithelial edges of the medial sulcus. It is possible that gelatinase A, which localizes to the stroma surrounding this epithelial seam, is involved in degradation of the underlying basement membrane, resulting in apoptosis of epithelium (unpublished observations) and/or migration away from the medial edges, a phenomenon postulated for fusion of the palatal processes (Hudson and Shapiro, 1973; Sharpe and Ferguson, 1988). Inhibition of neutral endopeptidase and meprin with thiorphan and phosphoramidon causes vascular distension in rat embryonic brain that indirectly results in secondary mandibular dysmor-phogenesis (Spencer-Dene et al., 1994); however, in our culture system, they did not interfere with normal development of the mandibular arches.

Development of the tongue is initiated by migration and/or proliferation of myoblasts to the region of the prospective tongue. The skeletal muscle cells of the tongue arise from progenitor myoblasts that migrate from the occipital somites 2-5 (Deuchar, 1958; Hazelton, 1970; Noden, 1986, 1991; Wachtler and Jacob, 1986; Trainor and Tam, 1995). The cells in this migratory pathway are desmin-positive (Mayo et al., 1992) and increase between E9-E11. In E10 mandibles, only 10% of the somitic mesoderm cells are desmin-positive initially, but 49% are positive by E10+9, an increase attributable to proliferation of the myoblasts. Our desmin immunolocalization studies showed that both migration and number of myoblasts were perturbed by treatment with MPI. The distribution pattern of desminpositive cells in MPI-treated mandible explants resembled that of untreated controls 2 days earlier in culture. For example, at E10+6, patches of myoblasts were still present in the lateral processes of MPI-treated explants, similar to controls at E10+4 (Fig. 9D compared with 9A), whereas almost all of the myoblasts in the controls at E10+6 (Fig. 9C) were concentrated in the now recognizable tongue. These results support the premise that MPI-mediated inhibition of MMP-mediated degradation of ECM molecules delays the movement of myoblasts to their final destination.

Although bone formation is accompanied by high expression of MMPs, there was no effect of MPI on osteoid or tooth formation. However, we observed a cartilage maturation defect in MPI-treated cultures. During normal cartilage formation, mesenchymal condensations form and cartilage expands from the center of the dense mesenchymal condensations. Thus ECM remodeling may make room for the expanding compartment of chondrocytes and cartilage matrix. The MPI-mediated defect was manifested in delayed maturation and fusion of the component segments of Meckel’s cartilage. Perhaps the problem is the failure to remodel the perichondrium, a site of high gelatinase A expression, to permit the independent foci of chondrogenesis to expand and fuse to form a single entity.

The precise molecular targets for MPI in this system are not known. The ubiquitous presence of gelatinase A mRNA and its protein in active form, as well as its ability to degrade many ECM substrates (Birkedal-Hansen et al., 1993; Tournier et al., 1994), suggests a major role for this MMP in the remodeling of the differentiating mandible. Localization of the gelatinase A, but not gelatinase B, collagenase-3 and stromelysin-3 mRNA transcripts specifically to the oral sulcus region makes gelatinase A the prime candidate for degrading the ECM to form the oral sulcus that is necessary for subsequent tongue development. The MPI effect was already observed by E10+2 when gelatinase A activity was initially detected, when gelatinase B, stromelysin-1 and uPA expression is low. Because it was necessary to achieve >90% inhibition of MMP activity to see morphologic effects, our attempts to inhibit specific MMPs with anitsense oligonucleotides were unsuccessful (unpublished results). However other MMPs that were not investigated, such as MT1-MMP (Basbaum and Werb, 1996), or novel MMPs may be the real targets.

Our data indicate that MMPs play specific roles in two migration/invasion events and two mesenchymal differentiation events. The requirement for relatively high concentrations of the reversible MMP inhibitor suggests these processes are difficult to inhibit in their entirety in the pericellular environment (Basbaum and Werb, 1996). Indeed, the widespread expression of MMPs, but focal effects of their inhibition, indicate other regulatory events such as activation may control patterning through local proteolysis. Our study raises the issue of whether other migration events, such as those undertaken earlier in development by CNC also require proteolysis. It is now feasible to use approaches such as those described here to dissect these invasion/migration events. The altered developmental endpoints that we have developed in the mandibular arch will be useful for analyzing specific ECM remodeling events in whole animals and tissue explants from mice with mutant expression of MMPs and TIMPs.

We thank Dr Jill Helms and Dr Harold Slavkin for helpful discussions, Elizabeth Hansell and Dr Leif Lund for technical assistance with our in situ hybridization studies, Pablo Bringas, Jr. for instruction on the mandible explant culture system and Nilda Ubana for technical assistance with the desmin immunohistochemistry. This work was supported by funds from the National Institutes of Health (DE10306) and the US Department of Energy (DE-AC03-76-SF01012).