Osteoblasts receive regulatory signals from hormones, growth factors, calcium, extracellular matrix, and other cells through a variety of receptors that utilize an array of signaling pathways and cytoplasmic messengers. This article addresses the nonuniform distribution of important signaling molecules (platelet-derived growth factor receptors [PDGFRs], nonreceptor tyrosine kinases, tyrosine kinase adaptor proteins, G proteins, and nitric oxide synthases [NOSs]) in the surface membranes of human and murine osteoblasts. We show that particular inner leaflet signaling molecules (e.g., heterotrimeric G proteins and Src family tyrosine kinases) are clustered and concentrated in Triton X-100-insoluble membranes that are enriched in caveolin, the major structural component of caveolae (50- to 100-nm flask-shaped invaginations of the plasma membrane that apparently are organized by oligomers of the protein caveolin). In addition, we show that a subset of highly ligand-responsive PDGFRs and mitogen-activated protein (MAP) kinase pathway effectors are present in the caveolin-enriched membrane fraction of osteoblasts.
OSTEOBLASTS ARE cells responsible for the production of bone matrix and are essential for bone development and for maintaining bone density.(1) As such, osteoblasts play critical roles in normal skeletal processes and in many pathological conditions including osteoporosis, osteopetrosis, and bone cancer. Osteoblasts, as central regulators of bone, receive, interpret, and coordinate a diverse set of extracellular stimuli. These signals include those originating from endocrine hormones, growth factors, cytokines, cell contact with bone matrix and neighboring cell types, soluble signals from osteoclasts, and biomechanical signals from strain and shear stress. Recently, we showed that osteoblasts have caveolae (detergent-resistant sphingolipid and cholesterol-rich 50- to 100-nm plasmalemmal vesicles) and have detergent-resistant membranes (DRMs) that are enriched in caveolin, a major structural protein of caveolae.(2) DRMs in general and caveolae specifically have been shown to have important roles in signal transduction and growth factor regulation. Here, we characterize the caveolin-enriched (CE) membrane signaling complexes of human and murine osteoblasts.
Caveolae and the more broadly defined DRM fractions, which include “flat” and vesicular structures, have important roles in cell signaling, which are beginning to be appreciated. (3–13) For instance, evidence suggests that the majority of signal transductions through the T cell receptor (TCR) actually are occurring in the DRMs of T cells, and that disruption of the lipid architecture of T cell microdomains adversely affects TCR signaling.(12,13) In monocytes, the major lipopolysaccharide (LPS) receptor CD14 resides in DRMs together with signaling molecules that regulate LPS-induced signaling.(11) With regard to caveolae, a highly specialized DRM, it has been shown in endothelial cells that these organelles contain abundant signal transduction molecules and are sites of enrichment for at least some of them.(8,14) Platelet-derived growth factor (PDGF)-induced signaling occurs in caveolae of endothelial cells,(7,8) and PDGF receptors (PDGFRs) are functional in isolated caveolae.(15) The response to hydrodynamic shear stress in blood vessels that results in increased endothelial nitric oxide synthase (eNOS) activity, tyrosine phosphorylation, and mitogen-activated protein (MAP) kinase regulation appears to occur principally within caveolae in vivo.(9,10) Moreover, it has been reported that caveolin, the major structural protein of caveolae,(16) binds and/or regulates the activities of Ras,(5,6) heterotrimeric G proteins,(3,6) nonreceptor tyrosine kinases,(6) and eNOS.(4) Based on the apparent signaling events occurring in DRMs (including caveolae) and the abundance of signaling molecules involved in multiple signaling pathways, it is inferred that DRMs are more than simple membrane regions in which cell surface signaling can occur, but are important loci for signal amplification and cross-talk between signaling pathways.
In this report we characterize the regions of osteoblast surface membranes that are defined by their insolubility in cold Triton X-100 (DRMs). In our companion article,(2) we showed that osteoblast DRMs are highly enriched in caveolin. Because not all DRMs contain caveolin (i.e., DRM from hemopoietic lineage cell types), we refer to the CE subset of DRMs as CEDRMs. Here, we show that the CEDRMs of osteoblasts are enriched in particular signal transduction molecules and tyrosine-phosphorylated proteins including the Src family nonreceptor tyrosine kinase, Fyn, and caveolin. Importantly, we also reveal the relationship between osteoblast PDGFRs and osteoblast CEDRMs: (i) PDGFRs are present in the CEDRMs of osteoblasts, (ii) PDGF causes the phosphorylation of PDGFRs in CEDRMs, (iii) after exposure to PDGF the fraction of PDGFRs in the CEDRMs is more heavily phosphorylated than the PDGFRs fraction in other compartments, and (iv) downstream effectors of PDGF signaling are present in osteoblast DRMs and are regulated dynamically in response to PDGF.
Our studies indicate that osteoblasts share the clustered organization of signal transduction molecules in their surface membranes with other cell types. This new knowledge of the spatial organization of signal transduction in osteoblasts should permit a more sophisticated analysis of their various physiologically important signaling pathways in response to mechanical shear, strain, soluble effector molecules, and matrix bound ligands.
MATERIALS AND METHODS
Human fetal osteoblasts(17) (hFOBs; generously provided by T. Spelsberg, Mayo Clinic, Rochester, MN, U.S.A.) were grown in Dulbecco's modified Eagle medium (DMEM) supplemented with 15% fetal bovine serum (FBS), L-glutamine, penicillin/streptomycin, and gentamycin. MC3T3-E1 (murine osteoblastic cells, referred to as MC3T3 cells(18) was grown in α-modified Eagle medium (α-MEM), 10% FBS, L-glutamine, and penicillin/streptomycin. Both cell types were cultured in 5% CO2 at 37°C and were verified to express the osteoblast-specific protein osteocalcin by radioimmunoassay of conditioned media. In some experiments, cells were treated with the phosphatase inhibitor pervanadate (1 mM Na3VO4 and 3 mM H2O2) for 20 minutes or with cytokines (see the following) in 5% CO2 at 37°C before analysis.
Cells were treated with 20 ng/ml PDGF-BB (Upstate Biotechnology, Lake Placid, NY, U.S.A.) in α-MEM, 10% FBS, L-glutamine, and penicillin/streptomycin for 20 minutes in 5% CO2 at 37°C. This time point was found to be optimal based on our kinetic analysis of PDGF-induced PDGFR activation.
MC3T3 cells were seeded at confluence in 100 mm × 20 mm petri dishes (Becton Dickinson, Franklin Lake, NJ, U.S.A.) in α-MEM as described previously for 12 h before cytokine treatment. Cells were then treated for 18 h with a mixture of 100 IU/ml interferon γ (IFN-γ; Gibco BRL, Rockville, MD, U.S.A.), 34 ng/ml tumor necrosis factor α (TNF-α; Gibco BRL), and 10 IU/ml interleukin-1β (IL-1β; Intergen, Purchase, NY, U.S.A.) after which cells were harvested for analysis and the conditioned media were used for nitrite measurement. This combination of cytokines was used to elicit the maximum amount of nitrite from osteoblasts as has been shown previously.(19)
Quantitative analysis of NO2− in the conditioned media of cytokine-treated cells and control MC3T3 cells was based on the Griess reaction.(20) In brief, 100 μl of conditioned media or nitrite standards (0-100 μM) were mixed with 100 μl of Griess reagent in microtiter plates, and absorbance at 530 nm was measured with a microplate reader (Molecular Devices, Sunnyvale, CA, U.S.A.).
The following monoclonal antibodies (mAb's) and polyclonal antibodies (pAb's) were used in these studies: antiphosphotyrosine pAb; anticaveolin mAb clone 2234; anticaveolin pAb; anti-Fyn mAb; anti-Shc pAb; anti-Nck mAb; antidynamin mAb; anti-Ras mAb; anti-Sos pAb; anti-neuronal NOS (nNOS) pAb; anti-eNOS pAb; anti-c-Raf mAb; anti-Erk2 (Transduction Labs, San Diego, CA, U.S.A.); anti-Gqα pAb, anti-Giα3 pAb, anti-Giα1 pAb, anti-Gsα pAb, and anti-Gβ pAb (Calbiochem, La Jolla, CA, U.S.A.); anti-Src pAb (Upstate Biotechnology, Lake Placid, NY, U.S.A.); anti-β-actin mAb (Sigma, St. Louis, MO, U.S.A.); anti-inducible NOS (iNOS) pAb; anti-Fyn pAb; anti-PDGFR pAb; anti-Raf-A; anti-Raf-B (Santa Cruz Biotech.); anti-ϵ-coatomer protein (COP) pAb (gift of Dr. Monty Kreiger, Whitehead Institute, Cambridge, MA, U.S.A.); and 4G10 antiphosphotyrosine mAb (gift of Dr. Brian Drukar, Oregon Health Sciences Center, Portland, OR, U.S.A.).
Successive detergent extraction method
Solubility analysis of membrane constituents was performed essentially as described.(21,22) In brief, cells were resuspended in buffer A (25 mM 2-[N-morpholino]ethanesulfonic acid [MES]; 150 mM NaCl, pH 6.5). To this, an equal volume of the same buffer with 2% Triton X-100, 2 mM Na3VO4, and 2 mM phenylmethylsulfonyl fluoride (PMSF) was added (1% final Triton X-100; buffer A complete), and the cells were incubated on ice for 30 minutes. Insoluble fractions were pelleted in a microcentrifuge (14,000g) for 20 minutes at 4°C. The supernatant was removed (“S” fraction, soluble) and the insoluble pellet was resuspended in buffer B (1% Triton X-100, 10 mM Tris, pH 7.6; 500 mM NaCl, 2 mM Na3VO4, 60 mM β-octylglucoside [Sigma], and 1 mM PMSF) for 30 minutes on ice. Debris was pelleted in a microcentrifuge (14,000g) for 20 minutes at 4°C, and the supernatant was collected. This fraction is referred to as “C,” for CEDRMs.
After treatments, various cell fractions and immunoprecipitates were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) using either 10% polyacrylamide gels or 5-15% gradient polyacrylamide gels, followed by electrotransfer to nitrocellulose (NC) membranes. NC membranes were stained with Ponceau S to verify uniform transfer and equal protein loading of gel lanes (where appropriate). The NC was incubated in blocking buffer (5% nonfat dry milk [wt/vol] in Tris-buffered saline [TBS], pH 7.6, 0.05% Tween 20 [vol/vol]) and then incubated with various monoclonal and polyclonal antisera in blocking buffer. After washing (2× in TBS; 0.05% Tween 20) the blots were incubated with the appropriate reporter antibody (donkey anti-mouse or sheep anti-rabbit immunoglobulin G [IgG]) conjugated to horseradish peroxidase (Amersham Corp.), washed (3× TBS; 0.05% Tween 20), developed by treatment with enhanced chemiluminescence reagents (Pierce, Rockford, IL, U.S.A.), and visualized by exposure to X-ray film. In many experiments, signals on immunoblots were quantified on a Phosphorimager (Molecular Dynamics, Sunnyvale, CA, U.S.A.) using ImageQuant software (Molecular Dynamics); the quantitative protein ratio in the soluble/CEDRM (S/C) fractions was calculated from the relative band intensities. In some experiments the NC was stripped (100 mM β-mercaptoethanol, 2% SDS [wt/vol], and 62.5 mM Tris, pH 6.7, at 60°C for 20 minutes), blocked, and immunoblotted with another antibody as described previously.
Cells were resuspended in buffer A-complete and homogenized at 4°C with a Teflon homogenizer. The homogenate was centrifuged at 100,000g using a 5-40% sucrose gradient. Successive fractions were collected from the top of the gradient and subjected to immunoblot analysis as described previously.
CE membranes isolated from sucrose gradients were resuspended in buffer A-complete and were immunoprecipitated as described in the following. Cell fractions generated by successive detergent extraction method (SDEM) were either directly immunoprecipitated or first boiled in 1% SDS for 10 minutes and then diluted 10-fold in 10 mM Tris, pH 7.5, 150 mM NaCl, and 1% Triton X-100 before immunoprecipitation. Membranes and cell fractions were immunoprecipitated with mAb's or pAb's previously bound to either goat anti-mouse conjugated superparamagnetic, polymer spheres (Dynal, Lake Success, NY, U.S.A.) using a magnetic particle concentrator (Dynal), or to protein A-Sepharose beads (Pharmacia, Piscataway, NJ, U.S.A.). Immunoprecipitates were washed four times in wash buffer (10 mM Tris, pH 7.5, 150 mM NaCl, and 1% Triton X-100), boiled in sample buffer (170 mM Tris, pH 6.8, 3% [wt/vol] SDS, 1.2% [vol/vol] β-mercaptoethanol, 2 M urea, and 3 mM EDTA), subjected to SDS-PAGE, and analyzed by immunoblot.
Protein concentrations in membrane fractions were quantified by MicroBCA assay (Pierce) using a bovine serum albumin (BSA) standard.
Distribution of G proteins, Src family tyrosine kinases, and tyrosine kinase adaptors in CEDRMs
Previously, we have shown that the DRMs of osteoblasts are highly enriched in caveolin, a major structural protein of caveolae(2,16); thus, the DRMs of osteoblasts are CEDRMs. To define the signaling molecules that localize to osteoblast CEDRMs, we subjected hFOB and MC3T3 cells to the SDEM and analyzed both the soluble (S) and CEDRM (C) fractions for the presence of Src family nonreceptor tyrosine kinases, tyrosine kinase “adaptor proteins,” various G proteins, and Sos (a guanosine triphosphate [GTP] exchange factor; Fig. 1).
The osteoblast CEDRMs were enriched in certain membrane inner-leaflet signaling proteins. Particular subunits of heterotrimeric G proteins (Giα1, Giα3, Gsα, and Gβ) were highly enriched in the osteoblast CEDRMs (C). In fact, at 10-40 μg of protein per gel lane, some of the Gα-subunits (Giα3 and Giα1) were either undetectable or barely detectable in the S fraction but were very abundant in the C fraction. In addition to the heterotrimeric G proteins, the Src family nonreceptor tyrosine kinase Fyn was found to be enriched 5:1 in the C fraction, while pp60 Src was enriched 2:1 in the C fraction. Other proteins distributed primarily to the soluble (S) cell fractions. The Gqα-heterotrimeric G protein subunit was enriched 2-fold in the S fraction. The large (100 kDa) G protein dynamin that is involved in vesicle transport(23) was found to distribute (9:1) to the S fraction, and Sos, a GTP exchange factor for the small G protein Ras, also was found almost exclusively in the S fraction.
We also examined the distribution of tyrosine kinase adaptor proteins (proteins that link tyrosine kinases with downstream signaling pathways; Fig. 1). Shc was found almost exclusively in the S fraction, while the Nck adaptor protein, although enriched in the S fraction, also was abundant in the osteoblast C fraction.
Expression and distribution of NOSs
As described by Damoulis and Hauschka,(19) osteoblasts release nitric oxide (NO) in response to cytokines, suggesting that osteoblasts must express one or more forms of NOS. To reveal the expression and distribution of NOS in osteoblasts, both cytokine-stimulated (producing 16 ± 1 μM nitrite) and cytokine-unstimulated MC3T3 cells (producing 1 ± 0.5 μM nitrite) were subjected to SDEM and immunoblot analysis using specific antibodies that recognize and distinguish nNOS, eNOS, and iNOS (Fig. 2).
We detected all three isoforms of NOS in osteoblasts: both eNOS (NOS 3) and nNOS (NOS 1) were expressed constitutively, whereas iNOS (NOS 2) was present only in the lysates of cells that had been stimulated with cytokines. Expression of eNOS and nNOS appeared unaffected by this treatment regimen (Fig. 2). Curiously, although iNOS was found exclusively in the S fraction, eNOS was almost entirely in the C fraction while nNOS was distributed between the S and C fractions, with greater amounts in C.
Tyrosine phosphorylation in the CEDRMs
Tyrosine phosphorylation is a critical protein modification occurring in many signaling pathways including those involving MAP kinase cascades (e.g., PDGFR signaling). Activation of PDGFRs, heterotrimeric G protein-coupled receptors, and other signaling molecules have been shown to increase tyrosine phosphorylation within CEDRMs.(8–10, 12, 24, 25) Because, as indicated in Fig. 1, the CEDRMs of osteoblasts contain tyrosine kinases along with various proteins that often are tyrosine-phosphorylated, we predicted that osteoblast CEDRMs would contain abundant tyrosine-phosphorylated proteins.
To begin to establish a role for osteoblast CEDRMs in signal transduction, we examined the C fraction for the presence of tyrosine-phosphorylated proteins. MC3T3 osteoblasts growing in media with 10% serum were analyzed at 80% confluence. These osteoblasts are subjected to a variety of signals from the surrounding extracellular matrix, autocrine growth factors, and serum growth factors. Cultures were treated with fresh serum-containing media with or without the potent phosphatase inhibitor pervanadate (1 mM Na3VO4 and 3 mM H2O2), fractionated by SDEM, and immunoblotted for phosphotyrosine content. Pervanadate preserved the level of tyrosine phosphorylation within the cells such that osteoblasts cultured (20 minutes) in the presence of pervanadate had >50 times the levels of tyrosine phosphorylation than osteoblasts cultured with media alone (Fig. 3). Many tyrosine-phosphorylated protein species from ≈15-250 kDa were present in the lysates of cells treated with pervanadate; only two species (≈200 kDa and 130 kDa) appeared to be slightly tyrosine-phosphorylated in control cells. The most prominent of the tyrosine-phosphorylated species present in the pervanadate-treated cell lysates migrated to approximately 20, 45, 60-65, 100-135, and 200 kDa (Fig. 3A). Although tyrosine phosphorylation was evident in both the Triton-soluble (S) and CEDRM fractions (C), certain phosphoprotein species were distributed preferentially. Phosphoprotein species of ≈60 kDa and 20-25 kDa (Fig. 3A, arrows) were distributed principally to C, whereas species of 28, 45, 7, and 135-160 kDa were found more abundantly in the S fraction. Other phosphoprotein species appeared to distribute relatively evenly between S and C. Thus, although pervanadate treatment resulted in the preservation of tyrosine phosphorylation in both S and C fractions, certain phosphoprotein species were confined to one domain or the other.
The 20- to 25-kDa phosphoprotein species that distributed to C was hypothesized to be caveolin, because caveolin can be tyrosine phosphorylated.(26,27) To test this possibility, we boiled the C and S fractions from pervanadate-treated osteoblasts in SDS and subjected the denatured proteins to immunoprecipitation with an antiphosphotyrosine mAb, followed by immunoblotting for caveolin (Fig. 3B, top panel). In the anticaveolin immunoblots, a 22-kDa protein species was detected in the antiphosphotyrosine but not in the control immunoprecipitation lane. Thus, it appears that caveolin was tyrosine-phosphorylated in the CEDRMs of osteoblasts. By a similar approach aimed at identifying the 60-kDa tyrosine-phosphorylated species in fraction C, we tested for the presence of tyrosine-phosphorylated Fyn. This Src family kinase was enriched in the C fraction (Fig. 1) and can itself be tyrosine phosphorylated. Immunoblotting the antiphosphotyrosine immunoprecipitate with an anti-Fyn pAb (Fig. 3B, bottom panel) indicated that Fyn phosphorylation was elevated in the pervanadate-treated osteoblasts and was selectively enriched in C. Thus, both caveolin and Fyn are distinct tyrosine-phosphorylated proteins found almost exclusively in the CEDRMs of osteoblasts.
To confirm the phosphorylation of caveolin and to determine if the phosphorylated caveolin interacted with other factors, we performed anticaveolin or negative control antibody immunoprecipitations of C from both pervanadate-treated cells and control cells and then subjected the immunoprecipitate to immunoblot analysis (Fig. 3C). Immunoblotting using an antiphosphotyrosine mAb revealed the presence of a 22-kDa protein, confirming that caveolin was phosphorylated in the C fraction of osteoblasts (Fig. 3C, right panel).
To detect interactions of phosphorylated caveolin with other factors, we immunoblotted the caveolin immunoprecipitates for various antigens that we hypothesized might interact with tyrosine-phosphorylated caveolin. This investigation resulted in the detection of the Nck adaptor protein in the caveolin immunoprecipitations from lysates of cells treated with pervanadate but not in control cells (Fig. 3C, left panel). This result is consistent with Nck being abundant in the CEDRMs of osteoblasts (Fig. 1) and suggests that phosphorylation of caveolin results in its de novo association with the Nck adaptor protein.
PDGFRs in osteoblast CEDRMs
PDGF is an important growth factor for osteoblasts(28) and also appears to have inhibitory effects on certain osteoblast functional activities.(28) In endothelial cells, PDGFRs have been shown to associate with caveolae,(8,15) require intact caveolae for maximal functional activity,(8) and retain function in isolated caveolae.(15) Because of the importance of PDGF in osteoblast mitogenesis and the association of PDGFRs with caveolae in endothelial cells, we examined the distribution of PDGFRs in the S and C fractions of osteoblasts. Initially, we examined MC3T3 cells that were either treated with PDGF or treated with media alone. Immunoblotting with an anti-PDGFR antibody that recognizes both PDGFR α- and β-chains indicated that a190-kDa PDGFR was present in both the S and the C fractions of the osteoblasts but was more abundant in the S fraction (Fig. 4); approximately 80-95% of PDGFRs were in S, and the remaining 5-20% of PDGFRs were distributed to C. Treatment of osteoblasts with 20 ng/ml PDGF did not significantly alter the distribution of PDGFRs.
However, tyrosine phosphorylation of PDGFRs was strongly enhanced in the C versus S fraction of cells treated with PDGF. PDGF caused a modest 2- to 5-fold increase in the tyrosine phosphorylation of the PDGFR in the S fraction (n = 3) and a striking 10- to 70-fold increase in the tyrosine phosphorylation of the PDGFR in the C fraction of the osteoblasts (n = 3; Fig. 4B, bottom panel). Other phosphorylated protein species also were present in the S and C fractions of both PDGF-treated and control osteoblasts, but they are not shown in Fig. 4 because this required overexposure of autoradiographs beyond the linear range for the film, thus prohibiting quantification of the phosphotyrosine content of PDGFRs. Although 20 minutes of PDGF treatment also resulted in an increase in the tyrosine phosphorylation of other protein species in the S fraction (data not shown), it did not appear to significantly increase the tyrosine phosphorylation of other proteins still present in the C fraction other than some >190-kDa species, which coimmunoprecipitated with the PDGFRs but did not detectably immunoblot with the anti-PDGFR antibody (Fig. 4C).
To rigorously confirm the results presented, the PDGFR was immunoprecipitated from the S and C fractions of PDGF-treated and mock-treated cells, followed by immunoblotting for the PDGFR (Fig. 4C, top panel) and pTyr (Fig. 4C, bottom panel). Immunoprecipitation using a negative control rabbit antiserum (Fig. 4C, left side) followed by anti-PDGFR immunoblotting resulted in the detection of nonspecific bands above 200 kDa (data not shown) and another species at approximately 100 kDa. Immunoprecipitation with anti-PDGFR followed by anti-PDGFR immunoblotting resulted in the detection of an additional, unique species at 190 kDa (presumably the PDGFR) that was not present in the negative control immunoprecipitates. The PDGFR was easily noted in the immunoprecipitates from the S fractions of PDGF-treated and mock-treated cells but required very long nonlinear exposure for detection in the immunoprecipitates from either of the C fractions (data not shown). After PDGFR detection, the NC was stripped and reimmunoblotted with antiphosphotyrosine mAb. Immunoprecipitates generated using the negative control antibody did not show any bands in antiphosphotyrosine antibody immunoblots (Fig. 4C, left side of bottom panel). In contrast, anti-PDGFR immunoprecipitates from both the S and the C fractions were readily detected by the antiphosphotyrosine antibody. The major species recognized by the antiphosphotyrosine antibody was identical in molecular weight to the band recognized by the anti-PDGFR antibody and was greater in intensity in the immunoprecipitates from cells treated with PDGF. There also were species greater than 190 kDa that reacted to the antiphosphotyrosine antibody that were present only in the C fraction from cells treated with PDGF. Interestingly, despite the fact that there is clearly a greater amount of PDGFR in the S fraction (Fig. 4C, top panel), the level of tyrosine phosphorylation of PDGFR from both S and C fractions is very similar (Fig. 4C, bottom panel), suggesting a significantly greater phosphotyrosine stoichiometry of the PDGFR in the C fraction.
MAP kinase pathway effectors involved in downstream PDGFR signaling are present in preassembled signaling complexes of osteoblasts
Ligand binding by the PDGFR initiates a cascade of signaling events involving the c-Raf-Mek-Erk MAP kinase pathway that culminates in the activation of Erk and its subsequent translocation to the nucleus.(29) Because phosphorylated PDGFR was detected in the CEDRMs of osteoblasts (Fig. 4), we asked whether MAP kinase pathway members also were present in these membranes. Src and Nck, two SH2 domains containing cytoplasmic signaling molecules that interact with the tyrosine-phosphorylated PDGFR, had already been identified within the CEDRMs of osteoblasts (Fig. 1), suggesting that this line of inquiry would be fruitful.
Initially, we examined CEDRMs isolated by density gradient fractionation (see Materials and Methods section), which we had previously shown to be highly enriched in caveolin but were markedly depleted of cytoskeletal and Golgi antigens.(2) In the CEDRM fractions derived from both human and murine osteoblasts we detected Ras, the small G protein that initiates the PDGF-induced MAP kinase pathway (Fig. 5A, bottom panel), c-Raf the MAP kinase-kinase-kinase (Fig. 5A, top panel, hFOB; Fig. 5C, top panel, MC3T3), and Erk2 the MAP kinase (Figs. 5A and 5C, middle panel). Although we were able to detect Ras, c-Raf, and Erk2 (Figs. 5A and 5C), we could detect only traces of Erk1 in either the S or the C fractions and no Mek in the CEDRM (data not shown). We also investigated the distribution of Raf-A and Raf-B, kinases that are related to c-Raf but are not involved in the PDGF-induced MAP kinase pathway. Although the antibodies employed did not produce results as clear as the anti-c-Raf antibody, both Raf-A and Raf-B appeared to be expressed in osteoblasts with C/S ratios similar to that for c-Raf (Fig. 5B).
Having established the presence of both PDGFRs and MAP kinase pathway effectors within the osteoblast CEDRMs, we next examined the effect of PDGF treatment on the membrane distribution of MAP kinase effectors. Immunoblot analysis of the CEDRMs of PDGF-treated and control MC3T3 cells using antibodies recognizing the c-Raf MAP KKK and Erk2 MAP K revealed their presence in the CEDRMs and indicated dynamic regulation of their distribution in response to PDGF (Fig. 5C, top two panels). This evidence suggests that in response to PDGF, both c-Raf and Erk2 move out of the CEDRMs to another cellular compartment (unidentified at this time). Note that the PDGF-dependent reduction of both the c-Raf and Erk2 signals is not caused by an artifact of gel loading because equal protein as determined by microBCA assay was loaded in the gel wells, Ponceau S staining verified equal protein transfer to NC, and immunoblotting showed equal caveolin signal in both lanes (Fig. 5C, bottom panel).
In this report we show that osteoblasts have lipid-organized signal transduction clusters that are responsive to ligand-induced growth factor receptor activation. The evidence supporting these findings includes the following: (i) that the CEDRMs of osteoblasts contain abundant signaling molecules including Src family tyrosine kinases, heterotrimeric G proteins, and NOS; (ii) that a subset of highly active PDGFRs is localized within the osteoblast CEDRMs; and (iii) that components of the PDGF-activated MAP kinase pathways are present in the osteoblast CEDRMs and, in response to PDGF, move from the CEDRMs to other cellular compartments.
Localization of signal transducing molecules to osteoblast CEDRMs
Certain receptors and cytoplasmic signaling molecules are distributed to preassembled DRMs in the plasma membrane, and, as a consequence, many of these proteins are imbued with detergent solubilities similar to those of the DRM itself.(30,31) This applies both to CEDRMs obtained from caveolin expressing cells and to DRMs obtained from cells that do not express caveolin. The cytoplasmic signaling molecules that have been found localized to CEDRMs and other DRMs are those that are modified by lipid attachment through myristoylation (e.g., pp60 Src and Gsα), palmitoylation (e.g., nNOS), or both myristoylation and palmitoylation (e.g., Giα and Goα).(3–6, 11, 14, 22, 32) Although lipidation of signaling proteins is important for their localization to both CEDRM and DRM, amino acids within a protein also influence its distribution. (33–35) Prenylation of proteins (e.g., farnesyl-Ras) may be an insufficient modification for their retention in DRM (CEDRM or other) isolated by methods employing detergents,(36) but nevertheless prenylation may be a signal that causes proteins to be targeted to caveolae (CEDRMs) and perhaps other similar microdomains (DRMs).(5) Proteins such as caveolin may distribute to DRMs independent of their lipid modifications,(34) and still others may find their way into CEDRMs via a direct interaction with caveolin. (3–6) Thus, some proteins may be in CEDRMs that are not found in other DRMs. In fact, interaction with caveolin may explain some of the differences in membrane distribution of proteins between cells that have caveolae (CEDRMs) and cells that do not. This situation may apply to pp60 Src, a protein that has been shown to interact directly with caveolin.(6) In peripheral blood T cells, which have DRMs but no caveolae, pp60 Src is primarily a Triton X-100-soluble molecule.(22) However, in cells with caveolae (endothelial cells(8) and the osteoblasts described here), pp60 Src is mostly resistant to Triton X-100 solubilization and is localized to CEDRMs. Whether covalent posttranslational lipid modifications occurring in osteoblasts serve to target pp60 Src and other proteins to CEDRMs and DRMs, and/or to stabilize the protein/membrane interaction, remains to be established.
Osteoblast CEDRMs are enriched in signaling molecules
In the experiments reported here we show that six various inner-leaflet signaling molecules of osteoblasts (pp60 Src, Fyn, Giα3, Giα1, Gsα, and Gβ) are largely or nearly completely resistant to solubilization by Triton X-100 but are solubilized by octylglucoside (Fig. 1). This evidence establishes that osteoblasts have preassembled CEDRMs containing signaling elements common to the DRMs of other cell types. As described previously, all but one of these proteins (Gβ) are subject to posttranslational lipid modifications that favor their localization to DRMs.(36) The presence of the single exception Gβ in the CEDRMs can be explained by its tight association with the prenylated Gγ-subunit and the possible influence of the acylated Gα-subunits.
Some membrane proteins were found to be predominately Triton X-100 soluble; Gqα was found at a 2:1 ratio in the soluble fraction, while the large G protein dynamin was >9:1 in the soluble fraction. Thus, in common with other cell types, osteoblasts have DRM fractions that are strongly enriched in a subset of signaling molecules but are not enriched in many others. Importantly, even though the osteoblast cell lines used in these studies had very different origins, the general distribution of the many proteins shown in Fig. 1 is consistent between human hFOB cells and murine MC3T3 cells. We infer that the partitioning of these proteins in CEDRMs is a general property of osteoblasts and is evolutionarily conserved.
eNOS and nNOS in osteoblast CEDRMs
Previous investigations of NOS expression in osteoblasts have relied on reverse-transcription polymerase chain reaction (RT-PCR), (37–40) Northern blot analysis,(38) or immunohistochemistry.(39,40) In our studies using immunoblot analysis, we were able to detect expression of three forms of NOS protein in the murine MC3T3 osteoblast cells (Fig. 2): eNOS and nNOS were constitutively present, while iNOS was expressed only after cytokine stimulation. MacPherson et al.(40) also reported constitutive expression of eNOS in osteoblasts (human primary osteoblasts [hOBs], and the osteosarcoma cell lines [TE85 and MG63]) and cytokine inducible expression of iNOS in hOB, but nNOS was not detected. Our new findings confirm and extend the observations made by those authors by providing protein data for eNOS and iNOS and by showing that nNOS also is expressed by osteoblasts.
Interestingly, the three isoforms of NOS expressed by osteoblasts have remarkably different solubilities. Although eNOS was almost entirely resistant to Triton X-100 solubilization and was primarily associated with the CEDRM, iNOS was completely Triton X-100 soluble, and nNOS was distributed unequally between the Triton-soluble and -insoluble (CEDRM) cell fractions. Both nNOS and eNOS have been described as having a direct interaction with caveolin,(4,41) suggesting one potential reason for their presence in CEDRMs. Additionally, eNOS can support dual acylation by myristic and palmitic acids,(42) which plays a role in directing its localization.(43,44) Both the interaction with caveolin and the attachment of lipids ultimately may determine the solubilities of the various forms of NOS present in osteoblasts. Differences in the expression and distribution of the three NOS isoforms suggest that they serve different functions in osteoblasts. This possibility is currently under investigation.
Tyrosine kinases and phosphoproteins in osteoblast CEDRMs
The organization of signaling components in detergent-resistant compartments (DRMs, CEDRMs, and caveolae) has been shown to be important for signal propagation through membrane receptors,(8, 12, 13, 15, 25) for shear stress-induced MAP kinase pathway activation,(9,10) and for activation of tyrosine kinases in cells of different lineages.(8, 12, 13, 15, 25)
We showed that both Src and Fyn nonreceptor tyrosine kinases are enriched in the CEDRMs of osteoblasts (Fig. 1). Thus, it is not surprising that tyrosine-phosphorylated proteins were found in CEDRMs of pervanadate-treated osteoblasts. Some phosphorylated species were common to both the soluble fraction and the CEDRM, while others apparently were distributed preferentially to a particular fraction (Fig. 4A). In the CEDRM fraction, the two prominent “unique” phosphoprotein species migrated to 20-25 kDa and 50-60 kDa (Fig. 4A, arrows). These likely represent caveolin and Src family tyrosine kinases (especially Fyn), respectively, which we showed were tyrosine phosphorylated in cells treated with pervanadate (see Fig. 4B). In the soluble fraction, phosphoprotein species of 35, 45, and 160 kDa were the most prominent unique species, but the identity of these phosphoproteins is not known. Confinement of signaling components to defined membrane domains may result in the concentration of effector targets to these regions. Thus, the localization of receptors and membrane proximal signaling components to specific membrane regions ultimately might affect and control the activation of downstream targets.
In analyzing the tyrosine phosphorylation of proteins within the osteoblast CEDRM, we detected an association of caveolin with the tyrosine kinase adaptor protein Nck. Although Nck has no known intrinsic enzymatic activity, it has been shown to be involved in tyrosine kinase-mediated activation events and contains three SH3 domains as well as a single C-terminal SH2 domain.(45) Association of Nck with tyrosine-phosphorylated caveolin suggests that caveolin phosphorylation may contribute to the recruitment of other proteins to caveolae, with resultant activation of downstream effector pathways.
PDGFRs and downstream members of the PDGF-induced MAP kinase pathway are found in osteoblast CEDRMs
We show the presence of PDGFRs in both S and CEDRM fractions of MC3T3 osteoblast cells. Although primarily in the S fraction, the subset of PDGFRs in the CEDRMs appeared to be much more active in ligand-dependent receptor phosphorylation than the S fraction of PDGFRs (Fig. 4). Although reasons for this are unknown, observations on Lck kinase phosphorylation and its membrane distribution may be instructive.(46) In those studies, Lck was shown to be hyperphosphorylated within the DRM of T cells compared with Lck in other membrane domains. This hyperphosphorylation is thought to be caused by the exclusion of the phosphatase CD45 from the T cell DRM. Thus, the increased tyrosine phosphorylation of PDGFRs in CEDRMs (Fig. 4) also may depend on the physical exclusion of phosphatases that would otherwise regulate its phosphorylation state and modulate signaling from the osteoblast CEDRMs. This possibility is currently under investigation.
We showed that osteoblast CEDRMs contain Ras, a G protein, which initiates activation of the PDGF-induced MAP kinase pathway; c-Raf, a MAP kinase-kinase-kinase; and Erk2, a MAP kinase. These effectors, acting downstream of the PDGFR, also have been found within endothelial cell CEDRMs(7, 8, 15) thus, their distribution within osteoblasts is not surprising. Interestingly, in the CEDRMs isolated from MC3T3 cells by sucrose gradient centrifugation, there was no detectable c-Raf, in contrast to the readily detectable c-Raf in the CEDRM fraction isolated by the SDEM. The c-Raf signal disparity in the CEDRM isolated by these two methods (Fig. 5A vs. Fig. 5C) is likely caused by dilution; CEDRM isolated on sucrose gradients is spread over five discrete fractions, whereas the CEDRM isolated by SDEM is contained in a single fraction. Thus, c-Raf may occur in all five sucrose density fractions, but at insufficient abundance in any single fraction to give a detectable signal.
In response to PDGF, there was a dramatic redistribution of both c-Raf and Erk2 away from the osteoblast CEDRM to another cellular compartment. Under the standard treatment conditions used in our studies, most of the PDGF-induced phosphorylation of PDGFRs was confined to the CEDRM of osteoblasts, suggesting the involvement of the osteoblast CEDRM in regulating the signaling induced by this growth factor. The observation that PDGF-responsive intracellular downstream signaling molecules (c-Raf and Erk2) also reside within osteoblast CEDRMs and redistribute in response to PDGF suggests that the cohort of PDGFRs within the osteoblast CEDRM are fully functional and further supports the hypothesis that osteoblast CEDRMs are important regulators of growth factor signaling.
The functionality of the CEDRM cohort of PDGFRs makes our observation that the osteoblast CEDRM had no detectable Sos quite curious. This GTP exchange factor is required for PDGFR-induced activation of Ras,(29) an event preceding c-Raf and Erk2 involvement, and its absence from the CEDRM suggests that Sos might be redistributed to the active receptor complexes only after ligand binding. Given that many of the downstream effectors of the PDGFR are within the CEDRM, the transient association of some effectors might be necessary to protect against inappropriate activation of the MAP kinase pathway.
Osteoblasts and CEDRMs
In this report we have shown for the first time that osteoblasts have CEDRMs that are enriched in signaling molecules including, heterotrimeric G proteins, nonreceptor tyrosine kinases, NOSs, and highly active PDGFRs. Additionally, we revealed that the osteoblast CEDRMs contain signaling elements of the PDGF-activated MAP kinase pathway that are responsive to PDGF. In the series of experiments described within we looked at the distribution of important signaling molecules, many of which were found to be enriched in the CEDRM of osteoblasts, which is to say that when the CEDRM fraction was compared with equal amounts of non-CEDRM cell fractions, the proteins in question were found in greater quantities in the CEDRM. The CEDRM fraction only contains approximately 10% of the total cell protein and 30% of the total membrane protein (Solomon, unpublished observation, 2000). Therefore, the conclusion that these proteins are enriched within the CEDRM is proven by our analysis, but the notion that greater numbers of each of these molecules reside in the CEDRM versus the membrane-at-large is not. The critical point is that the subset of signaling molecules within the CEDRM are contained in small well-defined membrane regions that make up an estimated 10% of the plasmalemmal membrane surface.(2) For instance, based on immunoblot and protein analysis, Src kinase is enriched 2:1 in the CEDRM fraction of osteoblasts. When normalized for protein content, the ratio of Src in CEDRM versus the rest of the cell is reduced to about 1:5. Given that Src is likely, for the most part, to be attached to the membrane, then for each unit area of the plasma membrane-at-large containing one Src molecule, the equivalent area of CEDRM would contain two Src molecules.
Because the enrichment of other examined proteins (e.g., heterotrimeric G proteins) within the CEDRM is even greater than that of Src, the CEDRMs are areas of the plasma membrane literally packed with signaling molecules as compared with the membrane-at-large. Our results regarding PDGFR signaling support the concept that the localization of signaling molecules into foci may have important implications with regard to signal transduction. These studies do not suggest that all of these critical signaling molecules are confined within the CEDRM, and they do not imply that they are functionally unimportant outside of the CEDRM. Indeed, there are PDGFRs in both the CEDRM and the membrane-at-large, and both populations are likely to be involved in regulating PDGF responses. That critical signaling molecules are enriched in the CEDRM, and that populations of CEDRM versus non-CEDRM receptors behave differently are novel principles for the analysis of osteoblastic responses to stimuli. The continued characterization of the CEDRM within osteoblasts and the definition of their spectrum of receptor signaling activities undoubtedly will have a strong influence on our understanding of osteoblast function and should provide a foundation for future research.
The authors thank J.R. Kasser for his support and T. Spelsberg (Mayo Clinic, Rochester, MN, U.S.A.) for providing the immortalized hFOB cell line. This work was supported by grants to P.V.H. from the National Institutes of Health (R01-AR44046) and to K.R.S. from the National Institutes of Health (National Research Service Award-HL09984).