FGFR2b signaling regulates ex vivo submandibular gland epithelial cell proliferation and branching morphogenesis

Fibroblast growth factors (FGFs), a family of 23 growth factors, are crucial for many developmental processes, including branching morphogenesis(Martin, 1998; Ornitz, 2000). Multiple FGF and FGF receptor (FGFR) isoforms are expressed in the developing mouse submandibular gland (SMG) (Hoffman et al.,2002). Altered SMG phenotypes occur in some FGF/FGFR transgenic mice (Ornitz and Itoh, 2001). The FGF10-null mouse has an absence of salivary, thyroid and pituitary glands,with minor defects in the formation of teeth, kidneys, hair follicles and digestive organs (Min et al.,1998; Ohuchi et al.,2000; Sekine et al.,1999). The FGFR2b-null mouse(De Moerlooze et al., 2000) and the soluble-FGFR2 transgenic mouse (Celli et al., 1998) do not develop salivary glands. The FGF7-null mouse has defects in kidney and hair development(Guo et al., 1996; Qiao et al., 1999); however,no SMG phenotype was reported. Functional compensation by other FGFs may occur with a single-FGF knockout. Targeted overexpression of FGF7 with the K14 promoter caused smaller salivary glands, excessive salivation, and a delay in gland differentiation (Guo et al.,1993). FGF8-conditional mutant and FGF8-hypomorphic mice also have defects in SMG development and differentiation(Jaskoll et al., 2004). Taken together, these studies suggest a primary role for FGF10/FGFR2b signaling in the initiation of the gland, an essential role for FGF8 at later stages, and possibly a role for FGF7 during differentiation. However, little is known about the role of FGFs after SMG initiation during early branching morphogenesis.

Branching morphogenesis of ex vivo SMGs involves both FGFR signaling(Hoffman et al., 2002; Jaskoll et al., 2002; Morita and Nogawa, 1999) and EGFR signaling (Kashimata and Gresik,1997; Kashimata et al.,2000). Previously, we identified a role for FGFR1 and phosphatidylinositol 3-kinase (PI3K) in SMG branching morphogenesis(Larsen et al., 2003). Here,we investigate the role of FGF7 and FGF10 signaling through FGFR2b. The known ligands for FGFR2b are FGF1, FGF3, FGF7 and FGF10. FGF7 and FGF10 both bind with high affinity to FGFR2b (Yeh et al.,2003), although FGF10 also binds FGFR1b(Igarashi et al., 1998).

Matrix metalloproteinases (MMPs) regulate branching morphogenesis in many organ systems. During ureteric bud branching, growth factors and extracellular matrix (ECM) components modulate MMP expression(Pohl et al., 2000). In mammary epithelium, interplay between MMPs, growth factors and morphogens is necessary for branching, although the MMPs are not required for proliferation(Simian et al., 2001). In SMG organ culture, exogenous collagenase decreased cleft formation, and TIMP1 increased it, by regulating collagen III fibril formation at the cleft site(Hayakawa et al., 1992; Nakanishi et al., 1986). However, the role of MMPs in SMG morphogenesis may not only be to cleave the collagen matrix, but also to release FGF-binding heparan sulfates from proteoglycans, or to directly cleave FGFRs to regulate FGFR function(Powers et al., 2000). MMP2 cleaves the extracellular domain of FGFR1, resulting in a cleavage product that can still bind FGFs and may regulate FGF signaling(Levi et al., 1996).

A fundamental but complex question is how FGFs interact with multiple FGFRs and result in distinct downstream signaling that coordinates proliferation,migration, differentiation and, ultimately, morphogenesis. Here, we investigate the role of FGFR2b and its ligands in serum-free ex vivo submandibular gland organ culture. We used salivary epithelium, separated from mesenchyme, to define distinct morphogenetic roles for FGF7 and FGF10. The downstream signaling and gene expression is ERK1/2-dependent and involves a regulatory network of FGFR1b/FGF1/MMP2 expression. FGF7 and FGF10 cause localized cell proliferation and they require both FGF1 and MMP activity to regulate epithelial budding and duct elongation during branching morphogenesis.

Submandibular/sublingual salivary gland rudiments (referred to as SMGs)dissected from either E12 or E13 ICR mice were cultured on Whatman Nuclepore Track-etch filters (13 mm, 0.1 μm pore size; VWR, Buffalo Grove, IL) at the air/medium interface. The filters were floated on 200 μl of DMEM/F12 in 50-mm glass-bottom microwell dishes (MatTek, Ashland, MA). The medium was supplemented with 100 U/ml penicillin, 100 μg/ml streptomycin, 150 μg/ml vitamin C and 50 μg/ml transferrin. Five gland rudiments were cultured on each filter at 37°C in a humidified 5% CO₂/95% air atmosphere. Glands were photographed at 2, 20 and 44 hours, and the end buds were counted at each timepoint. Each experiment was repeated at least three times.

Mesenchyme-free epithelia were cultured as previously described(Morita and Nogawa, 1999),with some modifications. E13 glands were incubated in 1.6 U/ml of Dispase in Hanks' balanced salt solution (Roche Molecular Biochemicals, Indianapolis, IN)at 37°C for 20 minutes. Epithelia with up to five buds were separated from the mesenchyme with fine forceps in Hanks' solution containing 10% BSA. Most importantly, horse serum was omitted from the media. The epithelial rudiments were placed on a Nuclepore filter, covered with 15 μl of laminin-1 (1 mg/ml; Trevigen, Gaithersburg, MD) or growth factor-reduced Matrigel, diluted 1:1 in medium (5 mg/ml; BD Biosciences, San Jose, CA), and the filter was floated on top of 200 μl of medium as described above. FGFs were added either alone or in combination to the media, and the glands were cultured for up to 44 hours. FGF1 (1, 20, 100, 500 ng/ml), FGF2 (1,10, 100, 200 ng/ml),FGF3 (1, 10, 100, 1000 ng/ml), FGF7 (50, 100, 500, 1000 ng/ml), or FGF10 (200,500, 1000, 2000 ng/ml), alone or in combination, was added to the cultures(all growth factors were purchased from R&D Systems, Minneapolis, MN).

Antisense and sense oligonucleotides were minimally phosphorothioated as previously described (Uhlmann et al.,2000). Residues that were phosphorothioate-modified are shown in bold: KGFR sense (1372-1390),3′-caggccaaccagtctgcct-5′; KGFR antisense (1390-1372),3′-aggcagactggttggcctg-5′(Post et al., 1996); Bek antisense (first two residues modified for mouse),3′-tggcagaactgtcaac-5′; Bek sense, 3′-gttgacagttctgcca-5′;antisense to both KGFR/Bek,5′-gctgaccatggtcac-3′; sense to both KGFR/Bek, 5′gtgaccatggtcagc-3′(Post et al., 1996). Oligonucleotides (2 μM) were added to the media and replaced after 44 hours. RT-PCR was used to determine the decrease in gene expression. FITC-labeled oligonucleotide uptake was detected with confocal microscopy(data not shown) [similar to Hoffman et al.(Hoffman et al., 2002)].

E12 or E13 SMGs were cultured for 48 hours as described above with recombinant (r) FGFRs or neutralizing antibodies to FGFs. The human rFGFR1b and rFGFR1c, and mouse rFGFR2b, rFGFR2c and rFGFR3c, chimeras were all added alone (1-20 μg/ml) and in various combinations (R&D Systems,Minneapolis, MN). Neutralizing antibodies to FGF1 (AF232), FGF2 (AF233), FGF7(MAB251) and Mouse IgG (MAB002) (R&D Systems), an FGF10 antibody previously used for neutralization of FGF10 activity (sc7375-L)(Harada et al., 2002), and a goat IgG control (sc2028-L; Santa Cruz Biotechnology) were added to the culture medium at concentrations of 10, 25 and 50 μg/ml. Combinations of either two or three anti-FGF antibodies (25 μg/ml each) were also tested. Glands were cultured for 44 hours, and branching morphogenesis was quantitated as described above.

SMGs or epithelial rudiments in laminin-1 were prepared as described above and incubated in an environmental chamber fitted on a Zeiss SV25 inverted microscope. Images were acquired once every 20 minutes for 36 hours using MetaMorph Software (Universal Imaging, Dowingtown, PA) and assembled into movies.

Proliferation and apoptosis were detected as previously described(Hoffman et al., 2002), using a BrdU Labeling and Detection Kit and an In Situ Cell Death Detection Kit,TMR-red (Roche Molecular Biochemicals, Indianapolis, IN), as described in the manufacturer's instructions. Nuclei were stained with SYBR-green, and for immunostaining, mouse-on-mouse reagents were used (M.O.M. reagents; Dako,Carpenteria, CA). With the apoptosis kit, the epithelium was stained with FITC-peanut lectin (Vector Laboratories, Burlingame, CA). The BrdU and apoptosis staining were quantitated using MetaMorph Software. The fluorescent pixels from all optical sections were measured and expressed as a ratio of the total pixel area of the gland or BrdU:SYBR-green. Five glands/condition were used, and the experiments were repeated three times.

Epithelial glands or rudiments cultured in laminin were also fixed on the filters in 4% paraformaldehyde (PFA) for 1 hour, permeabilized with 0.1%Triton X-100 for 15 minutes, and blocked overnight with 10% donkey serum,M.O.M. blocking reagent, and 1% BSA. Primary antibodies were added in M.O.M. protein reagent for 3 hours at room temperature and secondary reagents were added in PBS-Tween 20 (0.1%) for 2 hours. Primary antibodies included Flg(sc-121), bek (sc-122) (Santa Cruz), E-cadherin Mab 36 (BD Biosciences,Pharmingen), MMP2-hinge and MMP9-hinge (Triple Point Diagnostics), alpha-6 integrin (GoH3, Chemicon), and perlecan (MAB1948, Chemicon). Secondary antibodies were all donkey F(ab)2 fragments labeled with Cy2, Cy3 and Cy5(Jackson Immunoresearch Laboratories, West Grove, PA). All images were obtained with a Zeiss LSM 510 confocal microscope.

E13 salivary gland rudiments were cultured with 1.6 μg/ml rFGFR2b, the concentration required to give ∼50% inhibition of branching (data not shown). All FGFs tested were added at the same time as the rFGFR2b at the beginning of the experiment. A range of doses was tested, but only a single dose is shown. FGF1 (20, 100, 500, 1000 ng/ml), FGF3 (100, 500, 1000 ng/ml),FGF7 (50, 100, 500, 1000 ng/ml), FGF10 (100, 500, 1000, 2500 ng/ml), FGF2 (10,100, 500, 1000 ng/ml), FGF4 (10, 100, 500, 1000 ng/ml) and BMP7 (10, 100, 500,1000 ng/ml) were added to the cultures.

Epithelial rudiments were dissected free of mesenchyme and cultured in laminin-1 gels. The growth factors, inhibitors and rFGFRs were added at the same time. DMSO or PBS with 0.1% BSA was used as a carrier control. Su5402 (1,2.5, 5, 10 μM) was a gift from Dr F. Unda (Universidad del Pais Vasco,Spain). LY294002 (5 and 10 μM), UO126 (5 and 10 μM) and GO6983 (1.5μM) were all from Calbiochem (La Jolla, CA). The rudiments were cultured for 44 hours and photographed, and the bud number and duct length were measured using MetaMorph Software.

The media were concentrated in centrifugal filters (Millipore Biomax 10K NMWL Membrane, 0.5 ml volume). Glands were lysed, and total protein quantitated (BCA Assay, Pierce). The volume of media concentrate was normalized to protein in the gland lysates and analyzed on a 10% gelatin zymogram gel developed using Invitrogen Zymogram Buffers (Invitrogen,Carlsbad, CA). HT1080-conditioned medium (Chemicon, Temecula, CA) was included as a positive control.

At least eight epithelial rudiments were cultured in 15 μl of laminin-1 as described above. DNase-free RNA was prepared using an RNAqueous-4 PCR kit and a DNA-free DNase removal reagent (Ambion, Austin, TX). TaqManTM reverse transcription reagents (Applied Biosystems, Foster City, CA) were used to make cDNA. Real-time PCR was performed using primers designed with similar properties using Beacon Designer Software (Biorad, Hercules, CA), SYBR-green PCR Master Mix (Applied Biosystems), and a Biorad MyIQ real-time PCR thermocycler. Each cDNA (5-10 ng) was amplified with an initial denaturation at 95°C for 10 minutes, then 95°C for 15 seconds and 68°C for 30 seconds, for 40 cycles. Gene expression was normalized to the housekeeping gene S29. Melt-curve analysis was routinely run, and the PCR reactions were also analyzed by gel electrophoresis to confirm that a single product of the expected size was amplified. The reactions were run in triplicate, repeated three times, and the results combined to generate the graphs.

We used two approaches to perturb FGFR2 function. First, FGFR2expression was downregulated in E12 SMGs with antisense oligonucleotides that have been previously shown to decrease FGFR levels(Post et al., 1996; Zang et al., 2003). Branching was decreased ∼66% when compared with either the control oligonucleotide or the KGFR/Bek sense oligonucleotide control(Fig. 1A,B). A ∼50%decrease in FGFR2 expression was measured by RT-PCR after 24 hours(Fig. 1C), when compared with control and sense oligonucleotide-treated glands. The glands appeared to cleft, yet normal bud expansion did not occur, which suggests FGFR2 is important during ex vivo branching morphogenesis.

Second, we perturbed FGFR function by adding soluble rFGFR isoforms to the media, to compete out the endogenous FGFs. We tested a range of concentrations of FGFR isoforms, but rFGFR2b had the most dramatic effect on branching morphogenesis (Fig. 2A). The epithelium was smaller than the control and lacked enlarged terminal buds, but formed elongated ducts. The phenotype of the rFGFR2b-treated glands reflected how many buds were present at the beginning of the experiment. Smaller glands,3-5 buds, were more strongly inhibited (see Fig. 2B and Movie 1 in supplementary material). Lumen formation occurred within the ducts and is apparent in the light microscope images in the main ducts(Fig. 2A,C). rFGFR1b- and rFGFR3c-treated glands had fewer buds overall, but the morphology was similar to that of the control. Combinations of rFGFR1b and rFGFR3c had additive effects (data not shown), but the inhibition was less than with rFGFR2b(Fig. 2B). Our results suggest that multiple FGFs are involved in branching, that the ligands for FGFR2b play an important role in bud formation and expansion, and that other FGFs may regulate the extent of branching.

Proliferation was measured by BrdU incorporation, and FGFR2 was immunolocalized to the terminal epithelial buds of control and rFGFR2b-treated glands (Fig. 3A,B). rFGFR2b treatment dramatically decreased epithelial cell proliferation, but had less effect on mesenchymal cell proliferation. Therefore, the removal of FGFs that bind FGFR2b decreases epithelial proliferation but does not affect FGFR2 localization.

Apoptosis was not detected in the epithelial buds of rFGFR2b-treated glands, but there was a twofold increase in mesenchyme cell apoptosis(Fig. 3D-F), suggesting that the ligands for FGFR2b either directly or indirectly promote survival of the mesenchyme.

Exogenous FGF7 and FGF10 specifically rescued SMGs inhibited with rFGFR2b(Fig. 4A), whereas FGF1, FGF2,FGF3 and BMP7 did not. Soluble rFGFR2b binds to multiple FGFs, therefore we added neutralizing FGF antibodies alone and in combinations to identify the FGFs produced by the endogenous mesenchyme that were required for branching(Fig. 4B). None of the antibodies tested alone could inhibit branching; however, combinations of neutralizing antibodies to FGF1, FGF7 and FGF10 did significantly inhibit branching. These data suggest that multiple FGFs are involved, and that FGF1,FGF7 and FGF10 are important during ex vivo branching morphogenesis.

We used isolated epithelial rudiments cultured in serum-free medium to compare the direct effects of FGFs on epithelial morphogenesis. The epithelial rudiments do not grow without exogenous growth factors and extracellular matrix (ECM). FGF7 promoted budding of the epithelial rudiment with short ducts in a dose-dependent manner up to 200 ng/ml(Fig. 5A). FGF10 promoted duct elongation without enlargement of the epithelial buds, and increased the length and number of branches in a dose-dependent manner up to 2000 ng/ml(Fig. 5A). FGF1 and FGF4 promoted slight duct elongation, but the epithelium did not continue to grow(Fig. 5A). FGF2 promoted epithelial bud formation in a similar manner to FGF7. Growth factors such as FGF8, FGF18, BMP4 and BMP7 did not support epithelial growth and branching(Fig. 5A).

We hypothesized that combinations of FGFs binding to both FGFR1b and FGFR2b may result in a more complex morphogenesis. Combinations of FGF2/FGF10,FGF1/FGF7, or FGF1/FGF4 (Fig. 5B) each resulted in phenotypes with more complex morphology than any growth factor alone. FGF1/FGF2/FGF10 resulted in a more complex morphology, similar to in vivo morphogenesis, than any of the dual combinations (Fig. 5B), with elongated ducts, multiple branch points and enlarged epithelial buds. Combinations of FGF10/BMP4 or FGF10/BMP7 did not grow, suggesting that BMPs antagonize FGF10-mediated growth. FGFs and BMPs can play opposing roles in developmental systems involving epithelial-mesenchymal interactions(Neubuser et al., 1997; Niswander and Martin, 1993; Weaver et al., 2000).

Proliferation was measured at 8 hours, prior to morphogenesis, and the ratio of BrdU:SYBR-green was quantitated(Fig. 6A). The FGFs all increased proliferation, BMP4 decreased proliferation, and proliferation in BMP7-treated samples was similar to control(Fig. 6A,B). Therefore,multiple FGFs can stimulate epithelial proliferation, which is independent of later morphogenesis. However, by 44 hours(Fig. 6B) there were distinct patterns of epithelial cell proliferation with different FGFs. Proliferation was concentrated at the tips of the ducts with FGF10 treatment, and occurred throughout the buds and along the ducts with FGF7 treatment. The data suggest that downstream gene expression may regulate factors that specify where localized proliferation occurs. Either the localization of the FGFRs or the expression of downstream mediators of morphogenesis could specify the sites of proliferation.

Whole-mount immunolocalization of FGFR1 and FGFR2 in both FGF7- and FGF10-treated epithelial rudiments reveals that both receptors are expressed throughout the epithelium (Fig. 7A,B). Staining with both antibodies was blocked by pre-incubation with the immunizing peptides to which the antibodies were raised. There was increased FGFR1 staining around the periphery of the buds with FGF7 treatment and at the tips of the buds with FGF10 treatment. E-cadherin staining highlights the cell-cell junctions and acts as a control for antibody penetration through the tissue, while SYBR-green stains the nucleus of the cells and is pseudo-colored blue (Fig. 7A,B). Thus, FGFR localization alone does not specify where FGF7-or FGF10-induced proliferation occurs; therefore, possible co-receptors or cofactors may be involved, and multiple signaling pathways may exist if co-receptors are present.

Chemical inhibitors and recombinant FGFRs were added to epithelial rudiments to identify the receptors and downstream pathways of FGF7 and FGF10(Fig. 8). Su5402 (2.5 μM),an inhibitor of FGFR phosphorylation, UO126 (10 μM), a MEK1/2 inhibitor,and rFGFR2b (1.6 μg/ml) all significantly inhibited FGF7- and FGF10-mediated morphogenesis. They all resulted in a decrease in bud number with FGF7 treatment and a reduction in the length of the ducts with FGF10 treatment (Fig. 8B). However,LY294002 (10 μM), a PI3K inhibitor, inhibited FGF7-mediated morphogenesis but not FGF10-mediated morphogenesis, and GO6983 (1.5 μM), a broad PKC inhibitor, did not inhibit morphogenesis(Fig. 8A,B). In addition,rFGFR1b did not inhibit growth, but it did alter the morphology of FGF7-treated epithelium, resulting in a slight, but not significant, increase in bud number with more prominent ducts. This may suggest that FGFR1b plays a role in FGF7-mediated morphogenesis, possibly downstream of FGF7/FGFR2b signaling. Taken together, these data suggest that epithelial budding requires both PI3K- and MEK1/2-dependent signaling, whereas duct elongation requires MEK1/2-dependent signaling. Therefore, downstream gene expression from FGF7 and FGF10 signaling must result in differential expression of the mediators of morphogenesis.

The downstream mediators of morphogenesis likely involve the interplay between MMPs and growth factors (Simian et al., 2001). We compared levels of MMP, FGF and FGFR gene expression after FGF7 or FGF10 treatment by real-time PCR. Expression was normalized to S29 and expressed as a fold change in expression with FGF7 treatment compared with FGF10 treatment(Fig. 9B). Exogenous FGF7 increased expression of MMP2 (3.4-fold), FGF1 (4.1-fold) and FGFR1b (1.8-fold) compared with the levels of expression with FGF10 treatment. FGF10 treatment increased expression of MMP9 (2.9-fold)and FGFR2b (1.5-fold) compared with FGF7 treatment, although MMP9 expression was much lower (∼500-fold less) than MMP2 expression based on the threshold values of the PCR reactions. MMP14 (also known as MT1-MMP) was expressed at similar levels with either FGF7 or FGF10 treatment.

The pattern of immunolocalization of MMP2 and MMP9 was similar in whole SMGs and epithelial rudiments. MMP2 was localized in the mesenchyme and the peripheral epithelial cells in the whole gland(Fig. 9A, part a), whereas MMP9 was only localized in the mesenchyme (Fig. 9A, part b). Perlecan staining in the mesenchyme and the basement membrane outlines the epithelium in confocal sections. In FGF7- or FGF10-treated epithelium, MMP2 was localized in the cells near the periphery of the rudiment (Fig. 9A, part c,d). MMP9 was localized in the few mesenchyme cells remaining after the mechanical separation of epithelium and mesenchyme, and was not present in the epithelium (Fig. 9A, part e,f). The MMP9 mesenchyme staining explains why MMP9 was much less abundant than MMP2 in the real-time PCR. The cells positive for MMP9 did not stain withα6 integrin (Fig. 9A,part e), an epithelial marker, but did stain with perlecan, a mesenchyme marker (data not shown).

We analyzed MMP activity in the conditioned media by gelatin zymography,after 24 and 44 hours of FGF treatment(Fig. 9C). FGF7 increased both the pro and active forms of MMP2 when compared with FGF10. FGF10 increased MMP9 levels when compared with FGF7, although MMP9 is barely detectable by zymograms, as a few remaining mesenchyme cells produced it. Therefore, the changes in MMP gene expression are consistent with the changes in MMP activity. Epithelial morphogenesis is dependent on MMP activity, since GM6001,a broad MMP inhibitor, inhibited the enlargement of the epithelial buds and elongation of the ducts (Fig. 9D). In addition, neutralizing antibodies to FGF1 were required to inhibit morphogenesis in the intact SMG, so we added them to FGF7- and FGF10-treated epithelium. Anti-FGF1 inhibited growth, suggesting the epithelial production of FGF1 was required for FGF7- and FGF10-mediated morphogenesis. These data also support the hypothesis that FGF1 signaling(which may be through FGFR1b) and FGFR2b are required for branching morphogenesis. Taken together, our data suggest that MMP and FGF1 are mediators of morphogenesis, and are required for both FGF7- and FGF10-dependent morphogenesis.

FGFR2b is important for the initiation of salivary gland development, as mice lacking FGFR2b do not develop glands(De Moerlooze et al., 2000). The mechanisms of FGFR2b and FGF action during branching morphogenesis, after gland initiation occurs, are unknown. Here, we report that FGF7- and FGF10-mediated FGFR2b signaling regulates submandibular gland epithelial morphogenesis ex vivo by inducing localized proliferation and increased expression of FGFR1b/FGF1 and MMP2, and that both FGF1 and MMP activity are required for FGF-mediated epithelial morphogenesis.

Multiple FGFs and FGFRs are present in SMGs(Hoffman et al., 2002), and defining which FGFs are important in SMG development is hampered by the fact that many FGFs have overlapping functions. FGF10-knockout mice have no salivary glands (Ohuchi et al.,2000), suggesting that gland initiation is dependent on FGF10, but our ex vivo data suggest that, once the gland has formed, inhibition of FGF10 with neutralizing antibodies or decreasing FGF10 expression by antisense treatment (M.P.H., unpublished) does not inhibit branching. The lack of a SMG phenotype in single FGF-null mice, such as the FGF2- and FGF7-null mice (Sun et al., 2002; Zhou et al., 1998), could be interpreted as meaning they are not important for SMG development; however,overlapping FGF function is likely to compensate during development. Our experiments requiring combinations of neutralizing FGF1, FGF7 and FGF10 antibodies to inhibit branching when the endogenous mesenchyme is present support this hypothesis.

Previous reports show that FGF7 induces SMG epithelial proliferation but that epithelial budding also requires EGFR ligation(Morita and Nogawa, 1999). However, media with 10% horse serum was used in these experiments, which contains mitogenic growth factors and makes comparison difficult with our serum-free conditions. The EGFR-null mouse(Jaskoll and Melnick, 1999)has macroscopically normal glands with a reduction of buds/area, suggesting that EGFR regulates SMG branch number in vivo but is not essential for SMG initiation or early branching. An exogenous EGFR ligand is not required for epithelial branching; however, the ECM may contain EGFR ligands or the epithelium may produce an endogenous EGFR ligand. MMP degradation of laminin-5 is reported to release an EGF-like fragment during mammary involution (Schenk et al., 2003).

We have identified striking morphological differences between FGF7- and FGF10-treated SMG epithelium (Fig. 5). The function of FGFs in SMG epithelium is different from other branching organs. In the lung, FGF10 plays an important role during morphogenesis and induces primary and secondary bud formation, FGF1 induces robust growth with elongated buds, and FGF7 induces a cyst-like structure(Bellusci et al., 1997; Cardoso et al., 1997; Izvolsky et al., 2003a; Weaver et al., 2000). During uteric bud outgrowth, FGF1 and FGF7 are at opposite ends of a morphogenic spectrum, with FGF10 and FGF2 in between(Qiao et al., 2001). FGF1 induces elongated uteric buds with proliferating tips, whereas FGF7 induces proliferation with an amorphous morphology.

FGF10 is involved in Harderian and lacrimal gland development, where it induces epithelial proliferation but not branching morphogenesis(Govindarajan et al., 2000; Makarenkova et al., 2000). FGF10 is important during tooth development(Kettunen et al., 2000), where it is a survival factor for stem cell populations(Harada et al., 2002). In SMG epithelium, FGF10 promotes duct elongation by localized proliferation at the tip of the duct through ERK1/2-dependent pathways. Nuclear localization of FGFRs has been reported (Wells and Marti,2002), and FGFR2 was recently found in the nucleus during FGF9-induced proliferation (Schmahl et al., 2004). FGFR1b and FGFR2b are also localized throughout the epithelium; therefore, a co-receptor or cofactor is likely to be required to allow FGF10 to specifically stimulate just the tip cells to proliferate. The most likely candidates are heparan sulfate-containing cofactors or receptors,on the cell surface or in the ECM. The mitogenic activity of FGF10 is stimulated by heparin, whereas FGF7 is inhibited(Igarashi et al., 1998),suggesting the effects of FGF7 and FGF10 may be regulated by heparan sulfate. In the lung bud, the localization of heparan sulfate isoforms regulates FGF function at a particular stage or localization during development(Izvolsky et al., 2003a; Izvolsky et al., 2003b),although other transmembrane regulators of FGF signaling such as sefand XFLRT3 could be involved(Tsang and Dawid, 2004). This hypothesis remains to be tested in the context of salivary gland morphogenesis.

Both FGF7 and FGF10 signal through FGFR2b in an ERK1/2-dependent manner,but FGF7-mediated morphogenesis is also PI3K-dependant. In corneal epithelial cells, FGF7 also stimulates PI3K activity but is not inhibited by MAPK inhibitors (Chandrasekher et al.,2001). In our studies, FGF7 also increased FGFR1b and FGF1 expression compared with FGF10. The PI3K-dependent pathway(Fig. 8) and FGF1 activity(Fig. 9D) are both necessary for epithelial budding. PI3K may be downstream from either FGF10/FGFR1b, or from FGF1 binding either FGFR1b or FGFR2b. Therefore the downstream signaling from FGF7 may act in concert with FGF1 signaling to increase widespread proliferation resulting in epithelial budding. FGF-mediated increases in FGF gene expression have been identified in other systems, such as the lung, where FGF7 also increases expression of FGF1(Lebeche et al., 1999). Many studies on FGFR signaling focus on the roles of FGF1 and FGF2 in either migration or proliferation, and suggest that cells use Src and p38MAPK for migration and ERKs for proliferation(Boilly et al., 2000). Our data show that ERKs are necessary for proliferation; however, the role of cell migration is clearly important in SMG morphogenesis (see Movies 1 and 2 in supplementary material), and identifying the mechanisms of cell migration will be important for understanding gland development.FIG10 

There is increasing evidence that MMPs, in concert with FGFs, regulate morphogenesis (Simian et al.,2001). In breast cancer and lung epithelium, FGFs increase MMP production (Liu et al., 2002; Miura and Shiota, 2002). MMP2 cleaves the extracellular FGFR1, releasing a fragment that binds FGFs(Levi et al., 1996). Our data show that FGF7 increases MMP2 expression and activation, and FGFR1bexpression, which may allow the expanding epithelium to remodel the ECM and release soluble FGFR1b to regulate FGF function in a similar manner to our rFGFR experiments (Figs 2, 8). Expression profiling of MMPs in the salivary epithelium and mesenchyme reveal that most MMPs are expressed in the mesenchyme, except MMP2 and the two membrane-bound MMPs,MMP14 and MMP15 (M.P.H., unpublished). The epithelial localization of MMP2 around the periphery of the expanding bud and elongating duct, the inhibition of morphogenesis with an MMP inhibitor, and the lack of MMP9 in the epithelium suggest that MMP2 activation is important for epithelial morphogenesis. Information about the function of MMPs from individual knockouts has not revealed much about their role in development. MMP2-null mice are viable,although smaller, and are fertile (Itoh et al., 1997), whereas the MMP14 mouse has skeletal defects due to a lack of collagen turnover (Holmbeck et al., 1999). However, recently the MMP2/MMP14 double-knockout mouse was reported (Oh et al.,2004); the mice die after birth with respiratory failure, abnormal blood vessels and immature muscle fibers, suggesting a major developmental phenotype. Neither the phenotype of the salivary glands nor detailed analysis of organogenesis was reported, but the increased severity of the phenotype suggests that MMP2 and MMP14 play important roles during development that are not revealed by the single knockout. This report highlights the difficulties of drawing conclusions about in vivo function from the analysis of single gene knockouts in large families of proteins with overlapping functions, such as the FGFs and MMPs.

In conclusion, branching morphogenesis of mouse submandibular glands ex vivo is regulated by multiple FGFs and FGFRs. Downstream of FGF7 or FGF10 binding to FGFR2b there is a regulatory network of FGFR1b/FGF1/MMP2 expression that influences morphogenesis. Localized cell proliferation results in either epithelial budding or duct elongation. Increased MMP2 activity may increase ECM remodeling and FGFR cleavage, allowing localized expansion of the epithelium into the ECM at sites where proliferation occurs. The increased expression of FGFR1 and FGF1 with FGF7 could reinforce epithelial proliferation in an autocrine manner, resulting in widespread proliferation and budding. On the contrary, FGF10 treatment results in elongation of the duct behind the proliferating tip, where co-receptors or cofactors define the site of FGF10/FGFR2b signaling. The difference between FGF7- and FGF10-mediated morphogenesis is likely to be due to the factors that localize FGF binding to specific cells expressing FGFRs, resulting in different amounts of FGFR2b signaling, proliferation, and downstream expression of autocrine FGF signaling loops and proteases. Therefore, identifying the co-receptors or cofactors that regulate localized FGF binding and FGFR activation in the developing submandibular gland is an important future goal.

The authors thank Harry Grant and Rouman Pankov for critical reading of the manuscript, and Hynda Kleinman for her outstanding mentoring and support.