Research ArticleMechanotransduction

Cyclic GMP and Protein Kinase G Control a Src-Containing Mechanosome in Osteoblasts

See allHide authors and affiliations

Science Signaling  21 Dec 2010:
Vol. 3, Issue 153, pp. ra91
DOI: 10.1126/scisignal.2001423


Mechanical stimulation is crucial for bone growth and remodeling, and fluid shear stress promotes anabolic responses in osteoblasts through multiple second messengers, including nitric oxide (NO). NO triggers production of cyclic guanosine 3′,5′-monophosphate (cGMP), which in turn activates protein kinase G (PKG). We found that the NO-cGMP-PKG signaling pathway activates Src in mechanically stimulated osteoblasts to initiate a proliferative response. PKGII was necessary for Src activation, a process that also required the interaction of Src with β3 integrins and dephosphorylation of Src by a complex containing the phosphatases SHP-1 (Src homology 2 domain–containing tyrosine phosphatase 1) and SHP-2. PKGII directly phosphorylated and stimulated SHP-1 activity, and fluid shear stress triggered the recruitment of PKGII, Src, SHP-1, and SHP-2 to a mechanosome containing β3 integrins. PKGII-null mice showed defective Src and ERK (extracellular signal–regulated kinase) signaling in osteoblasts and decreased ERK-dependent gene expression in bone. Our findings reveal a convergence of NO-cGMP-PKG and integrin signaling and establish a previously unknown mechanism of Src activation. These results support the use of PKG-activating drugs to mimic the anabolic effects of mechanical stimulation of bone in the treatment of osteoporosis.


Mechanical stimulation induces bone growth and remodeling, which is critical for maintaining bone mass and strength (1). Compressive forces generated by weight bearing and locomotion induce small bone deformations and increase interstitial fluid flow. In response to these mechanical stimuli, osteoblasts lining endosteal and periosteal surfaces proliferate and differentiate, and osteocytes embedded in the canalicular bone network show enhanced survival (1, 2). Mechanoreceptors on osteoblasts and osteocytes include integrins associated with cytoskeletal proteins and mechanosensitive calcium channels (2). Mechanical stimulation rapidly and transiently increases intracellular calcium concentrations and production of nitric oxide (NO) and prostaglandin E2 (3, 4). These second messengers activate various signal transduction pathways, including the Raf–MEK [mitogen-activated protein (MAP) kinase or extracellular signal–regulated kinase (ERK) kinase]–ERK1/2 cascade, which leads to changes in gene expression (3).

NO synthesis in mechanically stimulated osteoblasts and osteocytes occurs in part through activation of endothelial NO synthase (eNOS) and increased abundance of inducible NOS (iNOS) (5, 6). NO is important for bone modeling and remodeling, as evidenced by in vitro and in vivo studies. Low doses of NO donors promote osteoblast proliferation and differentiation in vitro and increase bone formation in response to mechanical stimulation (7, 8). Furthermore, young eNOS-deficient mice have reduced bone mass due to defects in osteoblast number and maturation, and adults show impaired bone adaptation to interstitial fluid flow (9, 10). Moreover, iNOS-deficient mice and rodents treated with NOS inhibitors fail to increase bone formation after mechanical stimulation (6, 7, 11).

NO produced in mechanically stimulated osteoblasts and osteocytes activates soluble guanylate cyclase, which produces cyclic guanosine 3′,5′-monophosphate (cGMP), which in turn activates both soluble type I and membrane-bound type II protein kinase G (PKG); other cGMP targets include phosphodiesterases and cyclic nucleotide–gated ion channels (5, 12). PKGII-deficient mice are dwarfs because of a block in chondrocyte differentiation in bone growth plates, whereas PKGI-deficient mice have no obvious skeletal abnormalities (13). We have shown that NO and cGMP cooperate with a calcium-dependent pathway to activate ERK in response to fluid shear stress, but the mechanism (or mechanisms) whereby shear stress activates ERK are unknown (5). Here, we establish that the NO-cGMP-PKG pathway activates ERK through SHP-1 [Src homology 2 (SH2) domain–containing tyrosine phosphatase 1], SHP-2, and Src in shear-stressed osteoblasts and show defective signaling in PKGII-deficient mice. We uncover a link between the NO-cGMP-PKG pathway and β3 integrins, thereby explaining the requirement of both pathways for ERK activation in osteoblasts and osteocytes.


PKGII is necessary for fluid shear stress–induced osteoblast proliferation

Fluid shear stress stimulates osteoblast proliferation in an ERK-dependent fashion (14, 15), and this in vitro response correlates well with increased bone formation during mechanical stimulation in vivo, as shown by the differential osteogenic response to mechanical stress in different inbred strains of mice (16). Because we previously showed that the NO-cGMP-PKG signaling pathway is necessary for fluid shear stress–induced ERK activation (5), we examined the role of PKGII in early osteoblast proliferation with a small interfering RNA (siRNA) approach. We found that 15 min of laminar fluid shear stress (at 12 dynes/cm2) stimulated incorporation of bromodeoxyuridine (BrdU) in replicating DNA by a factor of ~3 in MC3T3-E1 murine osteoblast-like cells (referred to as MC3T3), a response that was prevented by siRNA-mediated depletion of PKGII (Fig. 1, A and B, and fig. S1A). Treatment with 8-(4-chlorophenylthio)-cGMP (8-CPT-cGMP; referred to as cGMP) mimicked the effects of fluid shear stress on cell proliferation in cells transfected with an siRNA targeting green fluorescent protein (GFP) but had no effect in PKGII siRNA–transfected cells (Fig. 1, A and B). Thus, PKGII activity is required for shear- and cGMP-induced early osteoblast proliferation. However, fluid shear responses may vary with the degree of osteoblast differentiation (1).

Fig. 1

Fluid shear stress– and cGMP-induced osteoblast proliferation and ERK activation. (A and B) MC3T3-E1–transformed murine osteoblast-like cells (MC3T3) were transfected with GFP or PKGII siRNAs and kept static, exposed to fluid shear stress (FSS; 12 dynes/cm2), or treated with 100 μM 8-CPT-cGMP (cGMP) for 10 min. BrdU incorporation into DNA was detected by immunofluorescence, with >300 cells analyzed per condition. (B) represents the mean of four experiments ± SEM; *P < 0.05. (C) Schema depicting Src activation. (D and E) hPOBs were sham-treated, subjected to FSS, or treated with 100 μM cGMP for the indicated times. Western blots were analyzed with phosphospecific antibodies against Src-pTyr418, Src-pTyr529, or ERK-pTyr204, and antibodies recognizing Src with nonphosphorylated Tyr529, total Src, or ERK (representative of two experiments). (F) MC3T3 cells were sham-treated, subjected to FSS, or treated with 100 μM cGMP for the times indicated. Changes in Src phosphorylation were expressed relative to the amount of pTyr418 or nonphosphorylated Tyr529 found in sham-treated cells. Mean ± SEM; n = 3. P < 0.05, sham compared to 2- or 5-min time points. (G) hPOBs were treated with 10 μM PP2 or PP3 for 1 hour and received 100 μM cGMP for 5 min (representative of two experiments). (H) MC3T3 cells were pretreated with PP2 or PP3 as in (G) before exposure to either FSS or 100 μM cGMP for 5 min. ERK phosphorylation was expressed relative to sham-treated cells. Mean ± SEM; n = 3. P < 0.05, PP2 compared to control and PP2 compared to PP3. (I and J) MC3T3 cells were transfected with siRNAs targeting GFP or two different sequences in Src and were exposed to FSS or treated with 100 μM cGMP for 5 min. (J) Mean ± SEM; n = 3. P < 0.05, siRNA targeting Src compared to siRNA targeting GFP.

Fluid shear stress– or cGMP-induced ERK activation requires Src

Src activity is regulated by phosphorylation. In unstimulated cells, the C-terminal Tyr529 is phosphorylated and interacts with the SH2 domain, keeping Src in a “closed” (inactive) conformation. Dephosphorylation of Tyr529 is a key event in Src activation because it changes the protein to an “open” conformation and enables autophosphorylation of Tyr418 in the kinase domain activation loop (Fig. 1C) (17). Exposing primary human osteoblasts (hPOBs), murine MC3T3 osteoblast-like cells, and MLO-Y4 osteocyte-like cells to fluid shear stress for up to 30 min rapidly induced Src dephosphorylation on Tyr529 and phosphorylation on Tyr418, indicating Src activation. Src activation peaked slightly before ERK activation, and treatment of cells with an NO donor or cGMP mimicked the effects of fluid shear stress (Fig. 1, D to F, and fig. S1, B to D). Results obtained with an antibody specific for Src phosphorylated at Tyr529 (Src-pTyr529) mirrored those obtained with antibody specific for Src that is not phosphorylated at Tyr529 (Src-nonpTyr529). The Src family kinase inhibitor PP2, but not the inactive analog PP3, prevented fluid shear stress–induced Src autophosphorylation and ERK activation; similarly, PP2, but not PP3, prevented cGMP-induced ERK activation (Fig. 1, G and H, and fig. S1E). siRNA directed against Src, but not a control siRNA, blocked fluid shear stress– and cGMP-induced activation of ERK1/2 (Fig. 1, I and J, and fig. S1F). Thus, Src is required for ERK activation in fluid shear stress– or cGMP-stimulated osteoblasts.

Membrane-localized PKG mediates Src activation

To determine whether fluid shear stress–induced Src activation was mediated by the NO-cGMP-PKG pathway, we used pharmacologic inhibitors of NOS [l-NG-nitroarginine methyl ester (l-NAME)], soluble guanylate cyclase (ODQ), or PKG [Rp-8-CPT-PET-cGMPS (Rp)] (Fig. 2A). Each of these agents almost completely prevented fluid shear stress–induced Src Tyr418 autophosphorylation and Tyr529 dephosphorylation in hPOBs and MC3T3 cells (Fig. 2, B and C). Thus, fluid shear stress–induced Src activation requires NO-cGMP-PKG signaling. To determine which PKG isoform mediated Src activation, we used siRNAs to selectively deplete cytosolic PKGI or membrane-bound PKGII (fig. S1A). siRNAs directed against PKGII, but not against PKGI, prevented fluid shear stress– and cGMP-induced Src activation (Fig. 2D).

Fig. 2

Src activation by membrane-bound PKG. (A) Schema of the NO-cGMP-PKG signaling pathway with inhibitors of NOS, soluble guanylate cyclase (sGC), and PKG. (B) hPOBs were sham-treated or exposed for 5 min to fluid shear stress (FSS; 12 dynes/cm2); some cells were pretreated with 4 mM l-NAME, 10 μM ODQ, or 100 μM Rp-8-CPT-PET-cGMPS (Rp) for 1 hour. Src phosphorylation was determined as in Fig. 1D (representative of two experiments). (C) MC3T3 cells were pretreated as in (B) and exposed to FSS for 5 min. Mean ± SEM; n = 3. P < 0.05, drug-treated cells exposed to fluid shear compared to FSS alone. (D) MC3T3 cells were transfected with siRNAs specific for GFP, PKGI, or PKGII; exposed to either FSS or 100 μM cGMP for 5 min; and analyzed as in Fig. 1F. Mean ± SEM; n = 3. P < 0.05, PKGII siRNA compared to GFP siRNA. (E to G) MC3T3 cells were transfected with PKGII (or GFP) siRNA and infected with adenoviral vectors encoding LacZ (control), siRNA-resistant wild-type (WT) or myristoylation-deficient PKGII (PKGII G2A), and WT or membrane-targeted PKGI (PKGI swap). The Western blot in (E) shows expression of PKGI and PKGII constructs in PKGII siRNA–transfected MC3T3 cells (whole-cell lysates); membrane association is indicated as determined by subcellular fractionation. In (F), cells were treated with 100 μM cGMP for 5 min or left untreated, and cells were analyzed as in Fig. 1E. (G) shows the mean ± SEM for three experiments. P < 0.05, PKGII WT or PKGI swap compared to LacZ virus in PKGII siRNA–transfected cells.

To assess whether the differing ability of PKGI and PKGII to regulate Src activity was due to differences in subcellular localization or substrate recognition, we reconstituted PKGII-depleted MC3T3 cells with siRNA-resistant PKG constructs. Expression of wild-type PKGII, but not a membrane binding–deficient PKGII-G2A mutant, in PKGII-deficient cells restored cGMP-induced Src activation (Fig. 2, E to G). [The mutant PKGII-G2A is cytosolic because it lacks the N-terminal myristoylation signal that targets PKGII to the plasma membrane (18).] A membrane-targeted PKGI/II chimera containing the N-terminal 39 amino acids of PKGII fused to the N terminus of PKGI (PKGI-swap) restored cGMP-induced Src activation in PKGII siRNA–treated MC3T3 cells, whereas similar amounts of wild-type PKGI did not (Fig. 2, E to G). All four PKG constructs produce similar total cellular PKG activities, but in different subcellular compartments (18). We conclude that membrane targeting of PKG activity is essential for cGMP-induced Src activation and that it involves a substrate recognized by both PKGI and PKGII.

Src is not directly activated by PKGII

Cyclic adenosine 3′,5′-monophosphate (cAMP)–dependent protein kinase (PKA) activates Src directly by phosphorylating Ser17, a site that could also be recognized by PKG (19). Purified PKGII did not affect Src phosphorylation or autophosphorylation and did not phosphorylate the PKA recognition site (or sites) in Src under conditions in which a known PKG and PKA substrate was efficiently phosphorylated (fig. S2, A to C). Some phosphorylation of the PKA recognition site (or sites) in Src was detectable in untreated MC3T3 cells; this phosphorylation was not increased in cGMP-treated cells, but was increased when PKA was activated with 8-Br-cAMP (fig. S2D). Thus, Src is not a PKGII substrate and is not activated by direct PKGII phosphorylation or interaction with PKGII.

Fluid shear stress– and cGMP-induced Src activation is mediated by SHP-1 and SHP-2

Rapid dephosphorylation of Src pTyr529 in fluid shear stress– or cGMP-treated osteoblasts (Fig. 1, D to F) implies a role for a protein tyrosine phosphatase (PTP) (or phosphatases). The PTP inhibitor vanadate did not affect Src phosphorylation at Tyr529 in unstimulated osteoblasts, suggesting low activity of PTPs that catalyze Src pTyr529 dephosphorylation. However, vanadate blocked fluid shear stress– and cGMP-induced Src pTyr529 dephosphorylation and prevented Src Tyr418 autophosphorylation (Fig. 3, A and B, and fig. S3A). These results suggest that PKG regulates Src through activation or recruitment of a PTP (or PTPs) that catalyzes Src pTyr529 dephosphorylation (or both).

Fig. 3

Fluid shear stress– and cGMP-induced Src activation mediated by SHP-1 and SHP-2. (A) MC3T3 cells were treated with 10 μM vanadate for 1 hour before a 5-min exposure to fluid shear stress (FSS; 12 dynes/cm2) or 100 μM 8-CPT-cGMP (cGMP). Src phosphorylation was analyzed as in Fig. 1F. Mean ± SEM; n = 3. P < 0.05, vanadate compared to control. (B) hPOBs were treated with vanadate and cGMP as in (A) (images are representative of two experiments). (C and D) MC3T3 cells were transfected with siRNAs targeting GFP, SHP-1, SHP-2, RPTP-α, or PTP-1B, and phosphatase abundance was quantified by Western blotting (whole-cell lysates for SHP-1, SHP-2, and PTP-1B and membrane lysates for RPTP-α). In (D), cells were exposed to FSS or treated with 100 μM cGMP for 5 min, and Src phosphorylation was analyzed as in Fig. 1F. Mean ± SEM; n = 3. P < 0.05, SHP-1 or SHP-2 siRNA compared to GFP siRNA. (E and F) MC3T3 cells were transfected with siRNAs targeting GFP, SHP-1 (E), or SHP-2 (F). Cells were infected with adenovirus encoding LacZ, siRNA-resistant human SHP-1 (E), or SHP-2 (F), and received 100 μM cGMP for 5 min (representative of three experiments).

Several PTPs are present in osteoblasts that can dephosphorylate Src Tyr529 in vitro, including the tandem SH2 domain–containing PTPs SHP-1 and SHP-2, PTP-1B, and the receptor-type phosphatase RPTP-α. All four proteins contain potential PKG phosphorylation sites and may activate Src in cells (20, 21). Depleting either SHP-1 or SHP-2 by siRNAs prevented fluid shear stress– and cGMP-induced dephosphorylation of Src pTyr529 in MC3T3 cells, whereas depleting PTP-1B or RPTP-α had no effect (Fig. 3, C and D, and fig. S3, B and C). Reconstitution of siRNA-resistant human SHP-1 or SHP-2 in SHP-1– or SHP-2–depleted MC3T3 cells restored cGMP-induced Src pTyr529 dephosphorylation and Tyr418 autophosphorylation, as well as ERK activation (Fig. 3, E and F). These results indicate that SHP-1 and SHP-2 are both required for fluid shear stress– and cGMP-induced Src activation. SHP-1 and SHP-2 also cooperate in epidermal growth factor–induced ERK activation, with SHP-2 acting as a scaffold that recruits SHP-1 to the receptor (22).

PKGII phosphorylates SHP-1 and stimulates PTP activity

The phosphatase activities of SHP-1 and SHP-2 are regulated by interactions between their SH2 and PTP domains and by phosphorylation of Tyr in the C termini, where the sequences of the two phosphatases diverge the most (21). We found that PKGII phosphorylated purified full-length SHP-1 wild-type and N-terminally truncated SHP-1 (ΔSH2, missing the two SH2 domains), but not the isolated catalytic domain (CAT); this localized the PKG phosphorylation site (or sites) to the C-terminal tail (Fig. 4, A and B). Under similar conditions, SHP-2 was not efficiently phosphorylated.

Fig. 4

PKGII phosphorylation of SHP-1 and SHP-2 and regulation of PTP activity. (A) SHP-1 constructs; AAA denotes alanine substitutions for Ser553, Ser556, and Ser557. (B) PKGII phosphorylation of bacterially expressed SHP-1 constructs in the presence of [γ-32PO4]ATP in vitro (representative of three experiments). (C) Effect of PKGII phosphorylation on SHP-1 PTPase activity in vitro (mean ± SEM, n = 3; *P < 0.05). (D and E) HEK 293T cells were cotransfected with Flag epitope–tagged SHP-1 constructs and empty vector, WT PKGII, or kinase-dead PKGII. Cells were labeled with 32PO4 for 4 hours and treated with 100 μM 8-CPT-cGMP (cGMP) for 10 min. Immunoprecipitates were analyzed by SDS-PAGE and autoradiography, and protein abundance was determined by Western blotting (representative of two experiments). (F) PTP activity of Flag-tagged SHP-1 constructs immunoprecipitated from HEK 293T cells transfected and treated as in (D) and (E). The activity of each construct was normalized to its activity in PKG-deficient cells (mean ± SEM, n = 4; *P < 0.05). (G) MC3T3 cells were transfected with SHP-1 siRNA and infected with adenovirus encoding LacZ or siRNA-resistant human SHP-1 WT, ΔCT, or AAA. Cells were exposed to 100 μM cGMP for 5 min, and Src phosphorylation or dephosphorylation was analyzed as in Fig. 1E (representative of three experiments). (H) Coimmunoprecipitation of SHP-1 and SHP-2 from MC3T3 cells treated with 100 μM cGMP for 5 min (representative of three experiments). (I) PTP activity in anti–SHP-2 immunoprecipitates (containing SHP-1 and -2) from control MC3T3 cells and from cells treated with 100 μM cGMP for 5 min. Mean ± SEM; n = 6. *P < 0.05.

We measured SHP-1 phosphatase activity with a phospho-Tyr529–containing Src peptide as a substrate. The specific activity of purified SHP-1ΔSH2 was similar to that of the isolated CAT domain and was higher than that of the full-length enzyme by a factor of ~5 (Fig. 4C), consistent with results reported for other substrates (21). PKGII phosphorylation of full-length SHP-1 or SHP-1ΔSH2 stimulated phosphatase activity, but PKGII did not stimulate SHP-1 CAT activity (Fig. 4C). Thus, PKGII phosphorylation of the SHP-1 C-terminal tail stimulates SHP-1 PTP activity toward Src phospho-Tyr529 in vitro.

Phosphorylation of Flag-tagged, full-length SHP-1 in human embryonic kidney (HEK) 293T cells was modestly enhanced by wild-type, but not kinase-dead PKGII, and SHP-1 phosphorylation occurred on a serine residue (or residues) (Fig. 4D and fig. S4A). An SHP-1 construct lacking the C-terminal 74 amino acids (SHP-1ΔCT) showed neither basal nor PKGII-stimulated phosphorylation, and a construct containing alanines substituted for Ser553, Ser556, and Ser557 in a potential PKG recognition sequence (SHP-1AAA) showed reduced basal phosphorylation, which was not increased in the presence of PKGII (Fig. 4E). Thus, PKGII targets at least one of these serines, which are conserved across species but are not present in SHP-2 (fig. S4B). Flag–SHP-1 isolated from HEK 293T cells expressing wild-type PKGII displayed about twice as much PTP activity as SHP-1 from cells cotransfected with empty vector or kinase-dead PKGII; in these experiments, wild-type PKGII expression did not alter the amount of Flag–SHP-1 protein in the cell (Fig. 4F). The PTP activities of SHP-1ΔCT and SHP-1AAA were not affected by PKGII expression, consistent with their lack of PKG phosphorylation. Wild-type SHP-1, but not SHP-1ΔCT or SHP-1AAA, restored cGMP-induced Src activation in SHP-1 siRNA–treated MC3T3 cells (Fig. 4G). Thus, SHP-1 phosphorylation and activation by PKGII is necessary for Src activation.

SHP-1–specific antibodies did not immunoprecipitate sufficient SHP-1 from MC3T3 cells to measure PTP activity; however, we found that SHP-1 coimmunoprecipitated with SHP-2 (Fig. 4H), which suggests the possibility of direct interaction between the two proteins (22). We used SHP-2 antibodies to isolate a complex of endogenous SHP-1 and SHP-2 from MC3T3 cells and found that cGMP stimulated PTP activity (Fig. 4I and fig. S4C). cGMP stimulation neither altered the proportion of SHP-1 and SHP-2 in the complex (Fig. 4H) nor affected Tyr phosphorylation of endogenous SHP-1 and SHP-2 (fig. S4D).

Src and ERK activation by cGMP and PKGII requires αvβ3 integrin ligation

Integrins link extracellular matrix proteins to the actin cytoskeleton, and anchoring of integrins to actin stress fibers is essential for fluid shear–induced c-fos induction (23). Osteoblasts express several integrins, including αvβ3 (2). ERK activation in fluid shear– or stretch-stimulated osteoblasts is prevented by β3 integrin–blocking antibodies and siRNAs directed against β3 integrin (24, 25). We found that siRNA depletion of β3 integrin prevented cGMP-induced Src and ERK activation in MC3T3 cells (Fig. 5, A and B) without affecting β1 integrin abundance or MC3T3 cell adhesion, the latter effect likely because cells adhere to secreted extracellular matrix proteins through multiple integrins. cGMP-induced Src and ERK activation was restored by expressing siRNA-resistant human β3 integrin in β3 integrin–depleted cells, but not by a β3ΔCT mutant [lacking the last three C-terminal amino acids (26)], which does not bind Src (Fig. 5, C and D). Thus, Src activation by PKGII requires Src interaction with the cytoplasmic tail of β3 integrin.

Fig. 5

Integrin dependence of Src activation by cGMP and PKGII. (A and B) MC3T3 cells were transfected with siRNAs targeting GFP or two different sequences in β3 integrin, and cell membranes were analyzed by Western blotting (A). Cells in (B) were exposed to 100 μM 8-CPT-cGMP (cGMP) for 5 min and analyzed as in Fig. 1E (representative of three experiments). (C) MC3T3 cells transfected with siRNAs targeting GFP or β3 integrin were infected with lentivirus encoding siRNA-resistant human β3WT or Src binding–deficient β3ΔCT. Cells were treated with 100 μM cGMP for 5 min and analyzed as in Fig. 1E. Abundance of human β3 integrin constructs was analyzed by Western blotting of whole-cell lysates; the amount of endogenous mouse β3 integrin is below detection limits. The bar graph shows the mean ± SEM of three experiments. P < 0.05, for β3WT compared to LacZ virus in β3 siRNA–transfected, cGMP-treated cells. (D) Coimmunoprecipitation of Src with human β3WT but not β3ΔCT from MC3T3 cells infected with β3-expressing lentivirus (representative of two experiments). Endogenous murine β3 is not efficiently immunoprecipitated. (E) hPOBs were kept in suspension (Susp.), allowed to attach to fibrinogen (FB)–coated dishes, or kept in suspension and stimulated with soluble FB (250 μg/ml) plus MnCl2 (2 mM) (Susp. + FB/Mn). After 1 hour, some cells received 100 μM cGMP for 5 min. Cells were analyzed as in Fig. 1E (representative of three experiments).

To determine whether cGMP-PKGII activation of Src and ERK requires ligand binding to β3 integrins, we either placed osteoblasts in suspension or allowed them to adhere to fibrinogen-coated plates for 1 hour (fibrinogen was used as a specific ligand for αvβ3 integrin, and waiting 1 hour ensured that ERK activation triggered by cell spreading had subsided) (27). cGMP activated Src and ERK in cells plated on fibrinogen but had no effect on cells in suspension, unless MnCl2 was added to directly activate αvβ3 integrin and enable binding to soluble fibrinogen (Fig. 5E and fig. S5). Similar results were obtained with fibronectin (fig. S5). Thus, cGMP-PKGII activation of Src and ERK in osteoblasts requires adhesive ligand (such as fibrinogen or fibronectin) binding to αvβ3 integrin.

Fluid shear stress triggers PKGII, SHP-2, and Src recruitment to β3 integrin–containing focal adhesions

Osteoblast stimulation by fluid shear stress induces assembly of actin stress fibers and β1 integrin–containing focal adhesions (23). We found that fluid shear stress promoted clustering of β3 integrin with SHP-2 in the periphery of osteoblasts (fig. S6) and increased coimmunoprecipitation of SHP-2 and β3 integrin (fig. S7A). Similarly, Kapur et al. reported that SHP-1 and SHP-2 association with β3 integrin is enhanced by fluid shear stress in human osteosarcoma cells (28). Fluid shear stress increased the colocalization of PKGII and Src with SHP-2 in the plasma membrane (fig. S6) and the amount of PKGII and vinculin associated with a detergent-insoluble fraction containing cytoskeletal and focal adhesion proteins (Fig. 6, A and B); there was no change in the amount of β-actin or caveolin-1 associated with the detergent-insoluble fraction. With vinculin serving as a focal adhesion marker, triple immunofluorescence staining showed that the percentage of focal adhesions containing PKGII and SHP-2, or Src and SHP-2, increased from ~20% in static cells to ~40% in shear-stressed osteoblasts (Fig. 6, C and D). We also assessed the association of PKGII with β3 integrin–containing integrins by bimolecular fluorescence complementation (BiFC). Two chimeras were cotransfected with human β3 integrin into MC3T3 cells: PKGII-VN, with the N-terminal half of the Venus fluorophore between the N-terminal membrane localization signal and the leucine zipper of PKGII, and αIIb-VC, with the C-terminal half of Venus fused to the C terminus of human αIIb integrin (Fig. 6E). We observed Venus fluorescence in the periphery of cells coexpressing αIIb-VC and β3 integrin with wild-type PKGII-VN, but not with mutant PKGII(G2A)-VN missing the N-terminal myristoylation signal, suggesting that membrane-bound PKGII directly or indirectly interacts with αIIb or β3 integrin (Fig. 6E and fig. S7B). We also demonstrated interactions among PKGII, β3 integrin, SHP-2, and Src in reciprocal coimmunoprecipitation experiments (fig. S7C). On the basis of these data and the fact that SHP-1 and SHP-2 coimmunoprecipitate (Fig. 4H), we conclude that PKGII, β3 integrin, Src, SHP-1, and SHP-2 may be present in a large complex and that fluid shear stress promotes assembly of this complex at focal adhesions. Because β3 integrin–containing focal adhesion complexes have been implicated as mechanosensors in osteoblasts and osteocytes and other cells (27, 29), we propose that this PKGII-, β3 integrin–, Src-, SHP-1–, and SHP-2–containing complex represents a mechanosome modulated by PKGII. The term “mechanosome” was previously coined to describe a hypothetical signaling complex composed of focal adhesion–associated proteins assembled and activated by mechanical stimulation (30).

Fig. 6

Characterization of a Src-containing mechanosensitive complex in osteoblasts. (A and B) Western blots of detergent-insoluble fractions isolated from static MC3T3 cells and from cells exposed for 5 min to orbital fluid shear stress (FSS; 120 rpm) (A). Bar graphs summarize four separate experiments. Mean ± SEM; n = 4. *P < 0.05, static compared to shear-stressed cells (B). (C and D) PKGII (white) and SHP-2 (red) (C) or Src (white) and SHP-2 (red) (D) colocalization with vinculin (green) was examined in static and shear-stressed (FSS) MC3T3 cells. Focal adhesions (FAs) were defined as vinculin-positive membrane complexes >0.5-μm size, and colocalization of PKGII and SHP-2 or Src and SHP-2 with FAs was scored as a percentage of total FAs. The graphs show the mean ± SE of proportion for three experiments, with ~35 cells evaluated per condition. *P < 0.05, static compared to shear-stressed cells. Scale bars, 5 μm. (E) Colocalization of αIIb and β3 integrins with PKGII monitored by bimolecular fluorescence complementation (BiFC) between αIIb-VC and WT PKGII-VN (VN-WT) or the membrane binding–deficient G2A mutant PKGII-VN (VN-G2A) in MC3T3 cells. Cells were transfected with human β3 and αIIb-VC. Representative confocal images of three separate experiments. Scale bars, 5 μm.

Src and ERK signaling is impaired in osteoblasts from PKGII-null mice

PKGII-deficient mice (Prkg2−/−) exhibit abnormalities in chondroblast differentiation, but osteoblasts from these mice were not previously examined (13). Primary osteoblasts from calvariae of Prkg2−/− mice and their wild-type littermates were morphologically similar; as expected, Prkg2−/− osteoblasts lacked messenger RNA (mRNA) encoding PKGII, but not PKGI (Fig. 7C). Src activation in response to fluid shear stress or cGMP was impaired in Prkg2−/− osteoblasts, whereas cells from wild-type littermates showed robust Src Tyr418 autophosphorylation and phospho-Tyr529 dephosphorylation (Fig. 7, A and B). A similar difference was seen in fluid shear stress– or cGMP-induced ERK phosphorylation. Because fos family genes are targets of PKGII and ERK (5), we assessed the effect of NO-cGMP-PKGII signaling in 1-week-old mice by measuring c-fos and fra-2 mRNA abundance in tibial diaphyses. The amount of c-fos and fra-2 transcripts was significantly lower in the tibial shafts of Prkg2−/− mice relative to those of wild-type mice (Fig. 7D). These results indicate that PKGII deficiency affects c-fos and fra-2 expression in bones of intact animals. Fos family transcription factors play a key role in skeletal development and maintenance (31).

Fig. 7

Signaling defect in PKGII-null osteoblasts. (A and B) WT and PKGII-null primary osteoblasts were stimulated with either fluid shear stress (FSS; 12 dynes/cm2) or 100 μM 8-CPT-cGMP (cGMP) for 5 min, and Src and ERK activation were analyzed as in Fig. 1E [representative of two (A) or three (B) experiments]. (C) Semiquantitative RT-PCR for PKGI and PKGII expression in primary calvarial osteoblasts isolated from 1-week-old Prkg2−/− mice and their WT littermates (representative of two experiments). (D) Tibial diaphyses were isolated from 1-week-old WT and Prkg2−/− mice, and c-fos, fra-2, and gapdh mRNA abundance was measured by quantitative RT-PCR. Mean ± SEM; n = 5. P < 0.05, WT compared to knockout mice. (E) Model of Src and ERK activation by FSS through NO-cGMP-PKGII, depicting the assembly of a mechanosensitive complex (shaded) containing PKGII, SHP-1, SHP-2, and Src bound to the cytoplasmic tail of β3 integrin (x = docking protein). Activation of the Ras-Raf-MEK-ERK cascade by Src occurs through Shc-dependent and -independent pathways (33).


We have identified a previously unknown mechanosensitive signaling system in osteoblasts, defining the events leading from shear stress activation of eNOS to Src activation. β3 Integrins function as mechanosensors (29), and the cytoplasmic tail of β3 integrin serves as a scaffold for a signaling complex that includes Src, SHP-1, and SHP-2 (26, 28, 32). We discovered that PKGII is recruited to the β3 integrin–SHP–Src complex in shear-stressed osteoblasts and that PKGII activated by cGMP phosphorylates and thereby activates SHP-1; the latter dephosphorylates and activates Src (Fig. 7E). Activated Src phosphorylates and recruits the adaptor Shc, which leads to activation of Ras and the Raf-MEK-ERK pathway and stimulation of cell growth (33, 34). This signaling system fills a gap in our understanding of how mechanical forces sensed by cell-matrix adhesions are translated into cellular responses, such as osteoblast proliferation and osteocyte survival (25). Other mechanisms of ERK activation in shear-stressed osteoblasts include calcium- and focal adhesion kinase–dependent pathways (5, 35).

PKGII and integrins converge on the Src-ERK pathway

Current models for mechanotransduction propose that cells sense and respond to forces at their points of attachment to the extracellular matrix, for example, at integrin-containing focal adhesion complexes (36). Osteocytes are tethered to the canalicular wall through αvβ3 integrin (29). In response to mechanical stimulation, integrins cluster and initiate “outside-in” signaling, but the exact nature of force-induced conformational changes in integrins and integrin-associated proteins remains unclear (2, 36).

Consistent with studies in other cell types (26, 37), we found that Src was associated with the cytoplasmic tail of β3 integrins, and this association was necessary for Src activation by cGMP and PKGII. Clustering of αvβ3 integrin in response to adhesive ligand binding (or shear stress, or both) may juxtapose Src molecules and allow partial activation through trans-autophosphorylation, but full Src activation downstream of integrins requires action of a protein tyrosine phosphatase, such as PTP-1B in platelets and RPTP-α in fibroblasts (20). We show that SHP-1 and SHP-2, as part of a larger complex, are necessary for fluid shear stress– and cGMP-induced Src activation. Consistent with our finding that Src and ERK activation by the cGMP-PKGII pathway required ligand binding to β3 integrins, others found that ERK activation by fluid shear stress is prevented by β3 integrin function–blocking antibodies (24). Some colocalization of SHP-2, Src, and PKGII with β3 integrin–containing complexes was apparent in the plasma membrane of static osteoblasts, but fluid shear stress increased SHP-2, Src, and PKGII recruitment to focal adhesions. In contrast, NO and cGMP activation of PKG induces disassembly of focal adhesions in chondroblasts and endothelial cells (38, 39).

PKGII selectively targets SHP-1 and SHP-2 to activate Src and ERK

Src binds to the SH2 domains of SHP-1 and SHP-2 in vitro and is a substrate for both phosphatases (34, 40). SHP-1 and SHP-2 serve nonredundant functions in Src and ERK activation, and SHP-2 appears to function as a scaffold, coupling SHP-1 or Src (or both) to growth factor receptors (22, 34). Fibroblasts from SHP-2–deficient mice show defective Src and ERK activation in response to integrin ligation and growth factor stimulation (41), and SHP-1–deficient mice develop osteopenia due to combined effects of decreased bone formation and increased resorption (42).

We found that SHP-1 and SHP-2 associate and that SHP-2 binding to β3 integrins is increased by fluid shear stress. The SHP-1–SHP-2 complex might bind to β3 integrin indirectly, through SHP-2 binding to the Tyr-phosphorylated adaptor protein DOK-1 (32). SHP-1 and SHP-2 are autoinhibited by interaction between their SH2 and PTP domains, but interaction of the SH2 domains with other Tyr-phosphorylated proteins results in partial phosphatase activation (21). We found that PKGII activation of SHP-1 is mediated by phosphorylation of Ser553, Ser556, or Ser557 in the C terminus of SHP-1; in contrast, protein kinase C phosphorylates SHP-1 on Ser591 and appears to inhibit phosphatase activity, although reports are conflicting (21). Tyr phosphorylation of SHP-1 and SHP-2 increases phosphatase activity (21), but we found no effect of PKGII activation by cGMP on the Tyr phosphorylation of SHP-1 and SHP-2.

Our results suggest that the main mechanism of PKGII activation of Src is through SHP-1– and SHP-2–mediated Src pTyr529 dephosphorylation, but additional mechanisms could exist, such as Src activation by interaction with other SH2 or SH3 domain–binding proteins or displacement of CSK (C-terminal Src kinase) (17). We did not find evidence for direct Src activation by PKGII, in contrast to Src activation by PKA (19). While this manuscript was in preparation, Leung et al. reported that Src activity in ovarian cancer cells depends on basal activity of PKGIα, with reciprocal phosphorylation between Src and PKGIα (43). We found no role of PKGI in Src activation in osteoblasts.

PKGII-deficient mice have skeletal defects

PKGII-deficient mice have no general metabolic disturbances and appear grossly normal at birth, but develop dwarfism because of abnormal endochondral ossification from defective chondroblast differentiation (13). In chondroblasts, PKGII promotes differentiation through regulation of Sox9 nuclear translocation and β-catenin stability (12) and antagonizes growth factor activation of the Raf-1–MEK–ERK cascade (44).

We found defective Src and ERK signaling in PKGII-deficient primary osteoblasts and decreased amounts of c-fos and fra-2 mRNA in tibial diaphyses of Prkg2−/− mice relative to wild-type littermates. Because c-fos and fra-2 are target genes of PKGII signaling, these data suggest that c-fos and fra-2 mRNA abundance is regulated by PKGII activity in vivo. c-fos is involved in osteoblast cell cycle regulation (45), and induction of c-fos expression in bone by mechanical loading correlates with new bone formation (7). Because the global PKGII knockout phenotype is dominated by impaired endochondral ossification leading to abnormal architecture of long bones and vertebrae, we plan to generate osteoblast- and osteocyte-specific PKGII knockout mice to evaluate the in vivo consequences of defective osteoblast mechanotransduction.

NO-cGMP-PKGII signaling is important for osteoblast mechanotransduction

Skeletal maintenance requires continuous mechanical input, as evidenced by bone mass loss in unloaded limbs due to decreased osteoblastic bone formation and increased osteoclastic resorption (1, 2). NO is crucial to the anabolic response to mechanical stimulation in vivo. In models of bone adaptation to loading, NOS inhibitors prevent loading-induced c-fos mRNA expression and impede osteogenesis (7, 11). In a hindlimb suspension model, mice deficient in eNOS (nos3−/−) or iNOS (nos1−/−) exhibit the same bone loss in unloaded limbs as wild-type mice. However, interstitial fluid flow induced by venous ligation protects wild-type but not nos3−/− mice from bone loss in the suspended limb, and reloading induces new bone formation in wild-type but not nos1−/− mice (6, 10).

We have identified a link between mechanical stimulation and activation of Src and ERK, establishing a central role of NO-cGMP-PKGII signaling in osteoblast mechanotransduction. Preclinical and clinical studies support osteogenic functions of NO, although optimal dosing of NO and the potential for NO-induced oxidative stress may be problematic (46). Our results suggest that agents that increase cGMP concentrations, such as NO-independent soluble guanylate cyclase activators and phosphodiesterase inhibitors, could have favorable “mechanomimetic” effects on bone.

Materials and Methods

Reagents and DNA constructs

Sources of reagents, including antibodies, siRNAs, and plasmid and viral vectors, are described in the Supplementary Materials.

Cell culture, transfections, and exposure to drugs and fluid shear stress

MC3T3-E1–transformed murine osteoblast-like cells (from the American Type Culture Collection, used at <12 passages) and hPOBs (established from surgical specimens according to an institutionally approved protocol) were cultured, transfected with Lipofectamine 2000 (Invitrogen), and characterized by histochemical staining for the presence of alkaline phosphatase activity and reverse transcription polymerase chain reaction (RT-PCR) quantification of vitamin D–induced osteocalcin mRNA expression as described (5). Cells were plated on etched glass slides, serum-deprived (in 0.1% fetal bovine serum) for 24 hours, and exposed to laminar fluid shear stress (12 dynes/cm2) in a parallel-plate flow chamber (Cytodyne Inc.) for 5 min unless noted otherwise. Sham-treated cells were grown under identical conditions and were placed in the flow chamber but not subjected to shear stress. Serum-deprived cells were preincubated with pharmacological inhibitors for 1 hour and treated with 100 μM 8-CPT-cGMP for 5 min unless noted otherwise.

Cell fractionation, immunoprecipitation, and immunoblotting

Detailed protocols for generation of membrane and cytosolic fractions, preparation of detergent-insoluble fractions containing focal adhesion proteins, immunoprecipitations, and Western blotting are provided in the Supplementary Materials.

Phosphorylation assays

Recombinant SHP-1 and SHP-2 were purified from bacteria as glutathione S-transferase (GST) fusion proteins bound to glutathione Sepharose; proteins were incubated for 15 min at 37°C in the presence of 10 μM [γ-32PO4]adenosine 5′-triphosphate (ATP), 10 μM cGMP, 10 mM MgCl2, and 50 ng of Flag-tagged PKGII (purified from HEK 293T cells). To examine SHP-1 phosphorylation in intact cells, we transfected 293T cells with Flag-tagged SHP-1 constructs and incubated them with 32PO4 (100 μCi/ml) for 4 hours, with 100 μM 8-CPT-cGMP added for the last 10 min. Cell lysates were subjected to immunoprecipitation with anti–Flag antibody, and immunoprecipitates were analyzed by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) and autoradiography.

PTP assays

Variable amounts of GST–SHP-1 constructs purified from bacteria, Flag–SHP-1 constructs immunoprecipitated from transfected HEK 293T cells, and anti–SHP-2 immunoprecipitates from MC3T3 cells were incubated for 15 to 60 min at 37°C with 100 μM C-terminal Src peptide (TSTEPQ-pY-QPGENL, from AnaSpec). Inorganic phosphate was measured by a colorimetric assay in a 96-well plate reader with malachite green and ammonium molybdate; the assay was linear with time and protein concentration.

Immunofluorescence staining, BrdU incorporation, and BiFC

Osteoblasts were either exposed to laminar fluid shear stress as described above or plated on glass coverslips in 24-well dishes and subjected to orbital fluid shear stress (120 rpm) for 5 min with similar results (Fig. 6 and fig. S6 show orbital fluid shear stress). A detailed description of BrdU treatment, transfection for BiFC, and sample preparation is provided in the Supplementary Materials. Cells were viewed with a confocal microscope (Olympus FV1000), a 40/1.3 oil-immersion objective, and 2× to 4× digital zoom. Images were analyzed with Fluoview (Olympus) and Photoshop (Adobe), and identical software settings were used for image acquisition of all samples in a given experiment.


RNA was extracted in TRI Reagent, and 1 μg of total RNA was subjected to reverse transcription and real-time PCR using previously described primers; gapd served as an internal reference with the 2−ΔΔCt method (5).

Prkg2−/− mice

Homozygous Prkg2−/− mice were generated as described (13); mice from each litter were genotyped by PCR, and tissues were harvested at the University of Bonn, Germany, with approval of the local Animal Welfare Committee. Osteoblast-like cells were extracted from the calvariae of 7-day-old mice by sequential collagenase digestion, and cells were characterized as described for hPOBs.

Statistical analyses

Pairwise comparisons were done by two-tailed Student’s t test and comparison of multiple groups by analysis of variance (ANOVA) with Dunnett’s posttest analysis to the control group; a P value of <0.05 was considered statistically significant. Where appropriate, such as for results of immunofluorescence staining and for normalized results, the Wilcoxon rank test was used for pairwise comparison and the Friedman test with Dunn’s posttest analysis for comparison of multiple groups.


Acknowledgments: We are grateful to T. Diep, A. Fridman, and J. Meerloo for technical support and to S. Ball, G. Firestein, and D. Boyle for providing operative bone specimens. Funding: This work was supported by NIH grants R01AR051300 (to R.B.P.), T32HL007261 (to N.M.), and P30NS047101 (UCSD Neuroscience Microscopy Shared Facility). Author contributions: H.R. performed BrdU uptake, fluid shear– and cGMP-induced Src/ERK phosphorylation, siRNA and viral reconstitution experiments, and with N.M. isolated hPOBs and performed RT-PCR. R.S. carried out the interaction and colocalization studies. S.Z. and D.E.C. performed site-directed mutagenesis, PTP activity assays, and phosphorylation studies. B.H., Y.C., and A.P. generated PKGII-deficient mice and isolated mPOBs. H.K. prepared β3 integrin virus. R.B.P., H.R., R.S., G.R.B., and S.S. designed the experiments, analyzed the data, and wrote the manuscript. Competing interests: The authors declare that they have no competing interests. According to university policy, a materials transfer agreement (MTA) is required from R.B.P. for SHP-1, PKGI, or PKGII constructs and from A.P. for any materials related to the Prkg2−/− mice.

Supplementary Materials

Materials and Methods

Fig. S1. siRNA knockdown of PKGI or PKGII; effects of fluid shear stress and cGMP on Src and ERK phosphorylation in MLO-Y4 and MC3T3 cells.

Fig. S2. PKGII does not directly phosphorylate Src.

Fig. S3. Effect of vanadate- and phosphatase-specific siRNAs on Src phosphorylation and dephosphorylation.

Fig. S4. Analysis of SHP-1 phosphorylation and function in MC3T3 cells.

Fig. S5. cGMP activation of Src requires ligation of β3 integrins.

Fig. S6. Colocalization of β3 integrins, PKGII, and Src with SHP-2–containing membrane complexes.

Fig. S7. Interactions among PKGII, SHP-2, Src, and β3 integrins.

Table S1. siRNA target sequences.


References and Notes

View Abstract

Stay Connected to Science Signaling

Navigate This Article