Mechanical Strain Differentially Regulates Endothelial Nitric-oxide Synthase and Receptor Activator of Nuclear κB Ligand Expression via ERK1/2 MAPK*

Exercise promotes positive bone remodeling through controlling cellular processes in bone. Nitric oxide (NO), generated from endothelial nitric-oxide synthase (eNOS), prevents resorption, whereas receptor activator of nuclear κB ligand (RANKL) promotes resorption through regulating osteoclast activity. Here we show that mechanical strain differentially regulates eNOS and RANKL expression from osteoprogenitor stromal cells in a magnitude-dependent fashion. Strain (0.25–2%) induction of eNOS expression was magnitude-dependent, reaching a plateau at 218 ± 36% of control eNOS. This was accompanied by increases in eNOS protein and a doubling of NO production. Concurrently, 0.25% strain inhibited RANKL expression with increasing response up to 1% strain (44 ± 3% of control RANKL). These differential responses to mechanical input were blocked when an ERK1/2 inhibitor was present during strain application. Inhibition of NO generation did not prevent strain-activated ERK1/2. To confirm the role of ERK1/2, cells were treated with an adenovirus encoding a constitutively activated MEK; Ad.caMEK significantly increased eNOS expression and NO production by more than 4-fold and decreased RANKL expression by half. In contrast, inhibition of strain-activated c-Jun kinase failed to prevent strain effects on either eNOS or RANKL. Our data suggest that physiologic levels of mechanical strain utilize ERK1/2 kinase to coordinately regulate eNOS and RANKL in a manner leading to positive bone remodeling.

The capacity of bone to remodel to meet functional structural demands was recognized by Wolff in 1892 as the "law of bone transformation" (1). Studies in humans (2,3) and animals (4,5) have shown that applied loads are associated with changes in bone density as well as skeletal macro-and microstructure (6,7). Removing load results in bone loss (8,9), whereas application of load causes bone apposition (10 -12), confirming that normal bone recognizes its loading environment and adapts to maintain an optimal functional structure. Only recently has understanding of the cellular processes underlying the ability of the skeleton to remodel (i.e. to serve Wolff's Law) made possible a search for the specific mechanisms by which mechanical force is translated into alterations in bone structure.
Studying cellular response to mechanical strain in vitro, we have previously shown that murine osteoprogenitor cells respond to the application of strain by decreasing the expression of receptor activator of NF-B ligand (RANKL) 1 mRNA (13). RANKL is the dominant molecule controlling osteoclastogenesis (14) and is up-regulated in response to hormones and factors that are known to promote bone resorption (15,16). We then showed that strain-induced reduction in RANKL expression required activation of ERK1/2 kinase, which was rapidly and sensitively activated by mechanical strain (17). This process in bone mirrors a signal cascade known to be pertinent to many of the responses of vascular tissue to shear forces (18 -21) as well as to strain (22,23). Since the diminished expression of RANKL by bone cells is inextricably linked to a repression of osteoclast formation, we wondered whether other "proformative" events might be associated with signals initiated by mechanical factors.
In this work, we show that endothelial nitric oxide synthase (eNOS) is regulated by strain in a divergent fashion; strain induces the expression of eNOS at the same time that this mechanical input decreases expression of RANKL. eNOS, which generates nitric oxide (NO), appears to promote an anabolic picture in bone (24,25). NO has been shown to have an inhibitory effect on both osteoclast formation and activation (26,27). There is also growing evidence that NO (and eNOS specifically) has roles in bone formation during growth (28,29) and in response to loading (30). We will show in this work that induction of eNOS expression, similarly to mechanical regulation of RANKL, also requires activation of ERK1/2 kinase. To confirm the role of ERK1/2 in these mechanically controlled events, we utilize an adenovirus causing constitutive ERK1/2 activity and show both that eNOS is up-regulated and that RANKL is down-regulated in the presence of an activated MAPK signaling system.

EXPERIMENTAL PROCEDURES
Materials and Reagents-Antibodies to total ERK1/2 were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA), and those to phosphorylated ERK1/2 were from New England Biolabs (Beverly, MA). ERK1/2 inhibitor PD98059 was obtained from Calbiochem, as was the JNK inhibitor (JNKi; SP600125). Fetal bovine serum was from Hyclone (Logan, UT). Other chemicals and supplies were purchased from Sigma.
Cell Culture-To generate primary stromal cell cultures, murine marrow cells collected from the tibiae and femurs of 3-5-week-old male C57BL/6 mice were plated in 6-well plates at 1.6 ϫ 10 6 cells/cm 2 as previously published (17). After 60 min, nonadherent cells containing the stromal elements were transferred to Bioflex collagen I-coated plates (Flexcell Corp., McKeesport, PA) in ␣-minimal essential medium plus 10% fetal bovine serum. The next day nonadherent cells were discarded; adherent stromal cells were cultured with 10 nM 1,25-dihydroxyvitamin D added to stimulate RANKL expression on day 4. Strain regimens were applied on day 6.
For experiments where lysates were made for Western analyses, inhibitors were added 30 min prior to strain, and the experiment was stopped as indicated. When the end point was to measure mRNA species, total mRNA was isolated 24 h after beginning strain induction.
Application of Mechanical Strain-Uniform equibiaxial mechanical strain was generated using a Flexcell Bioflex instrument (Flexcell Corp., McKeesport, PA) as previously described (13). Strain magnitudes were as noted from 0.25 to 2%, with strain frequency fixed at 10 cycles/min (0.17 Hz). Similar plates containing control cultures were kept in the same incubator but were not subjected to strain regimens.
Measurement of NO-A fluorometric assay was used to measure nitrite in samples using the reagent 2,3-diaminonaphthalene with comparison with a NaNO 2 standard curve (0 -10 M) as described previously (31,32). Briefly, 100 l of standards and samples were added to microtiter 96-well plates (DYNEX Technologies, Inc.) and mixed with 10 l of fresh 2,3-diaminonaphthalene (prepared in 0.62 M HCl) for 10 min at room temperature. The reactions were terminated with 5 l of 2.8 N NaOH. Formation of the 2,3-diaminonaphthotriazole end product was measured using an LB 50 plate reader (PerkinElmer Life Sciences) with excitation at 360 nm and emission at 440 nm. All standards and samples were measured in triplicate.
Western Blot Analysis-For NOS expression, proteins were extracted from stromal cells in boiling lysis buffer (10 mM Tris, pH 7.4, 1% SDS, and 1 mM sodium orthovanadate). Cell lysates were boiled for an additional 5 min and passed three times through a 26-gauge needle. After centrifugation at 16,000 ϫ g for 5 min to remove insoluble material, protein concentrations in the supernatant were determined using the Bio-Rad DC protein assay kit. Samples containing 200 g of total protein were electrophoresed through a 7.5% SDS-PAGE and transferred to a 0.45-m polyvinylidene difluoride membrane. Membranes were immersed in blocking buffer containing TBS with 0.1% Tween 20 (TBST) and 5% nonfat milk overnight at 4°C. The eNOS and inflammatory NOS (iNOS) isoforms were identified using respective polyclonal antibodies (1:1000; Transduction Laboratories, Lexington, KY); membranes were washed three times with TBST and then incubated with second antibody conjugated with horseradish peroxidase (1:1500). The proteins were detected by ECL plus chemiluminescence kit (Amersham Biosciences).
Real Time PCR to Assess mRNA Species-Analysis of eNOS, RANKL, and 18 S mRNA was performed as in Refs. 17 and 33 using the iCycler (Bio-Rad). Briefly, reverse transcription of 0.5 g of total RNA treated with DNase I was performed with random decamers (Ambion, Austin, TX) and superscript II reverse transcriptase (Invitrogen). For real time PCR, amplification reactions were performed in 25 l containing primers at 0.5 M and dNTPs (0.2 mM each) in PCR buffer and 0.03 units of Taq polymerase (Invitrogen) along with SYBR-green (Molecular Probes, Inc., Eugene, OR) at 1:150,000. Aliquots of cDNA were diluted 10 -10,000-fold for 18 S and 5-625-fold for eNOS, iNOS, and RANKL to generate relative standard curves with which sample cDNA was compared (34). For eNOS, forward and reverse primers were 5Ј-AAC CAG CGT CCT GCA AAC C-3Ј and 5Ј-CAA TGT GAG TCC GAA AAT GTC C-3Ј, respectively, creating an amplicon of 133 bp. For iNOS, forward primer 5Ј-GCA TGG ACC AGT ATA AGG CAA GCA-3Ј and reverse primer 5Ј-GCT TCT GGT CGA TGT CAT GAG CAA-3Ј amplified a 222-base pair amplicon (35). RANKL primers were reported in Ref. 17. Standards and samples were run in triplicate. Dilution curves showed that PCR efficiency was more than 90% for all species including 18 S.
Adenovirus Preparation-The adenovirus carrying the constitutively activated mitogen-activated protein kinase/extracellular signal-regulated kinase kinase (MEK) (Ad.caMEK) was generated in our laboratory beginning with the pMCL-MKK1 plasmid carrying the caMEK sequence generously provided by Dr. Natalie Ahn (University of Colorado). As published (36), the MEK was constitutively activated by removal of the coding sequence for amino acids 32-51 and alteration of the nucleotides encoding two serine regulatory sites at 218 and 221 to glutamic and aspartic acid, respectively. The caMEK sequence was removed with XbaI and HindIII and inserted into the pShuttle vector as per the Stratagene protocol (Stratagene, La Jolla, CA). The pShuttle now carrying the caMEK sequence was linearized with PmeI and cotransformed with pAdEasy in recA-proficient bacteria. Small kanamycin resistant colonies were selected and digested with PacI. After confirmation of appropriate band patterns, the viral DNA was used to transform recA-deficient bacteria. Colonies containing the desired gene as analyzed by PCR were selected, and the DNA was linearized with PacI and purified before calcium phosphate transfection of HEK293 cells. After transduction, HEK293 cell layers were overlaid with agarose and assessed for viral plaque formation at 10 days. Viruses were eluted from plaques and amplified in HEK293 cells; infected cells were subjected to three freeze/thaw cycles, and the supernatants were collected for use in experiments.
For virus applications, viral lysates were added with 10 g/ml Lipo-fectAMINE in serum-free medium at the indicated MOIs, with an equal volume of media containing 20% fetal bovine serum added at 5 h. Cellular responses were studied at 24 h.
Assessment of MAPKs-Western blotting for MAPK phosphorylation was performed as detailed in Ref. 17. Briefly, the supernatants of cell lysates were concentrated, and 10 g of protein was chromatographed and transferred to polyvinylidene difluoride membrane. Primary and secondary antibody incubation was followed by measurement of chemiluminescence (ECL-plus; Amersham Biosciences).
Assessment of JNK and ERK Activation-To visualize the ability of the Ad.caMEK to activate ERK1/2, we utilized Western analysis to see the 42-and 44-kDa phosphorylated bands as in Ref. 17. To assess c-Jun N-terminal protein kinase (JNK) activity, a c-Jun phosphorylation assay was utilized as previously; briefly, soluble cell lysates in cold Triton lysis buffer with proteinase inhibitors were incubated with 2.5 g of agarose-bound glutathione S-transferase-c-Jun-(1-79) (Calbiochem) in a total volume of 400 l of Triton lysis buffer overnight at 4°C on rotating platform (17,37). Washed agarose beads were mixed with 100 M ATP and 5 Ci of [␥-32 P]ATP, and reactions were carried out at 30°C for 30 min. Phosphoproteins were separated on a 15% SDS-PAGE gel, and phosphorylated glutathione S-transferase-c-Jun-(1-79) was visualized at ϳ37 kDa.
Statistical Analysis-Results are expressed as the mean Ϯ S.E. Statistical significance was evaluated by Dunnett or Bonferroni's post hoc one-way analysis of variance (GraphPad Prism).

Mechanical Strain Induces eNOS Expression in Murine
Stromal Cells-Bone cells express multiple forms of NOS, with eNOS being the most prominent isoform. iNOS and neuronal NOS (nNOS) have both been described in bone during growth and fracture healing but are clearly present in lower amounts than eNOS in bone cells themselves (38,39). Using Western blotting to identify species in C57BL/6 murine stromal cells, we were able to identify eNOS protein in samples where 200 g of lysate protein was loaded. To assess whether the NOS might be residing in the membrane component, we also separated membrane protein and found that similar loading of 200 g of membrane protein did not alter the amount of eNOS reacting with the antibody. In cells exposed to 2% magnitude mechanical strain overnight, eNOS increased, as shown in a representative blot (Fig. 1a). iNOS was not visible in Western blots from these lysates, despite successful blotting of an assay positive control for the iNOS species.
Because of sensitivity in assessing amounts of eNOS protein, we confirmed the strain inductive response using a sensitive real time PCR assay for eNOS mRNA. This assay was designed to recognize only eNOS, and not iNOS or nNOS, on the basis of primer specificity. In Fig. 1b, compiled from three separate experiments where cells were exposed to 24 h of 2% strain, eNOS mRNA significantly increased by more than 2-fold. Using primers specific for murine iNOS, we demonstrated a lack of response to strain of this NOS isoform; we performed real time PCR on control and cells strained for 24 h. Grouped real time PCR data from three experiments showed no significant difference in levels of iNOS mRNA between control (100 Ϯ 10%) and strained cells (90 Ϯ 25%). These data ruled out a contribution from iNOS to the strain-induced increase in nitric oxide described below.
Gene Expression Is Differentially Sensitive to Strain Magnitude-The response of eNOS was in the opposite direction to the strain-induced decrease in RANKL gene expression, which we had explored previously (17). To understand better the differences in mechanical regulation of these two genes, eNOS and RANKL, we performed a series of experiments where strain magnitude was varied. Our current strain instrumentation allowed application of uniform strain as low as 0.25%, and we studied four different strain regimens up to 2% magnitude. At least three separate experiments were performed at each strain magnitude and compiled in Fig. 2. For each sample, eNOS, RANKL, and 18 S were amplified from samples subjected to reverse transcription with random decamers as described under "Experimental Procedures." The data was expressed as percentage of target control mRNA compared with 18 S in the sample. eNOS mRNA responded to 0.25% strain, rising to 120%, and continued to increase, reaching a plateau at 161 Ϯ 28% compared with eNOS mRNA measured in unstrained cells (p Ͻ 0.001), a rise comparable with the previous series shown in Fig. 1b. RANKL also responded to application of 0.25% strain, reaching a nadir by 1% strain, as shown in Fig.  2. This result can be compared with previous results where strains less than or equal to 1% were not effective when dosed for only 6 h (17). In the experiments presented here, application of strain for 24 h allowed us to generate reproducible mechanical effects on both eNOS and RANKL gene expression with strains significantly lower than 1%. Results showing that strain caused a divergent response of these two genes at all magnitudes studied suggested that mechanical regulation of eNOS and RANKL share a proximal signaling pathway.
NO Release Is Stimulated by Strain and Reflects Increases in eNOS-We next measured nitric oxide production from murine cells during strain. NO did not rise immediately, in contrast to the immediate activation in response to fluid shear in cellular NOS of both osteocytes and osteoblasts (40,41). Even at 60 min, as measured by a more sensitive assay (31,32) than used in the latter studies, NO was not significantly different from that secreted into media by unstrained cell cultures (Fig. 3a). However, NO was significantly increased by 2-fold after 24 h of strain, as shown in Fig. 3b. Furthermore, the strain induction of NO required the activity of endogenous NOS, as the competitive inhibitor, L-NAME, blocked the effect of strain on NO production, shown in Fig. 3c. Because iNOS was not affected by the strain protocol (see Fig. 1c), we inferred that the increased NO resulted from increased gene and protein expression of the endothelial form of NOS.
Inhibition of ERK Activation Indicates That ERK Has Primary Effects on Both eNOS and RANKL-We had previously shown that straining cells at magnitudes less than 2% caused activation of both ERK1/2 and JNK (17). Here we set out to prove that strain induction of eNOS expression is blocked by inhibition of ERK1/2 activation. The effect of overnight strain in this series of five experiments was to increase eNOS expression to 195 Ϯ 26% of the unstrained control, as expected (Fig.  4a). Treatment with ERKi during the strain protocol prevented any strain-induced increase in eNOS as shown in the second gray bar in Fig. 4a, which shows an insignificant difference between the unstrained and strained cultures in the presence of ERKi. Exposure of unstrained controls overnight to an efficacious concentration of ERK inhibitor (ERKi) unexpectedly increased eNOS expression to 136 Ϯ 8% of unstrained control levels, suggesting that basal ERK1/2 activity operates some Total RNA was collected from three wells/sample, two samples/condition for real time assessment of eNOS, RANKL, and 18 S. eNOS, shown in gray bars, increased in response to strain in a magnitude-dependent fashion, becoming significant by 1% strain magnitude (#, p Ͻ 0.01; ##, p Ͻ 0.001). For strain-induced inhibition of RANKL mRNA expression, all strain magnitudes examined were significant (*, p Ͻ 0.001), and the 1% magnitude was significantly more inhibited than the 0.25% (**, p Ͻ 0.05).
regulatory control over eNOS expression in bone cells. However, when ERK1/2 was inhibited, application of strain failed to cause further increases in eNOS mRNA levels.
The effect of strain on control cells in this series of experiments was to diminish RANKL expression to 58 Ϯ 4% of control (Fig. 4b). The strain inhibition was ablated in cultures where ERK1/2 activation was blocked by treatment with ERKi; RANKL expression was not significantly different from that in unstrained cells treated with ERKi. However, as shown in Fig.   FIG. 3. Strain increases NO production in stromal cultures. a, stromal cells were subjected to strain regimen for the times designated, and medium was collected for NO measurement. There was no change from control values of nitrite (32.8 Ϯ 9 nM averaged over 10 experiments) in cultures strained for 60 min or less. b, strain application for 24 h doubles NO levels where strain was applied to stromal cells for 8 h followed by rest, for 4 h at the end of culture, or for a continuous 24 h. Medium was collected for nitrite assay at the 24-h point. Continuous strain caused a doubling of NO production, significant at p Ͻ 0.05 (*) by analysis of variance. Shown is a compilation of three experiments. c, strain increase in NO production requires the action of an endogenous nitric oxide synthase. The NOS inhibitor, L-NAME (300 M), was added to unstrained and strained cells for 24 h, and NO production was measured. The graph is a compilation of three experiments and shows that L-NAME inhibited the strain-induced NO. The only group significantly different from control was strain in the absence of L-NAME (p Ͻ 0.05).

FIG. 4. ERK inhibitor blocks strain effect for both RANKL and eNOS. a, ERKi was added (right set of bars)
to cultures one-half hour before application of strain (gray bar in each set; data compiled from six experiments). Strain was applied overnight (ϳ16 h), and total RNA was assayed. eNOS increased significantly in strained cultures as expected (a different from b, p Ͻ 0.0001). ERKi caused significant increases in eNOS compared with unstrained controls, but the addition of strain to ERKi-treated wells did not further increase eNOS (p ϭ 0.129). Using analysis of variance, all bars are different from control (i.e. a b c; p Ͻ 0.05). b, this graph shows RANKL mRNA in the same experimental design as a above. As shown in the graph, strain inhibited RANKL expression significantly (*, p ϭ 0.0038) but did not affect RANKL expression when ERKi was present in the culture (p ϭ 0.1065). Using analysis of variance, all bars are different from control (i.e. a b c; p Ͻ 0.05). c, inhibition of NO production does not prevent straininduced ERK activation. Cultures of stromal cells were treated with or without L-NAME at 300 M for 1 h prior to applying strain. After 10 min of strain, cell lysates were collected, and ERK1/2 was assessed. Phosphorylated ERK (ERKϳP) is shown in the top row and total ERK in the bottom. Strain increased, an effect that was not altered by the presence of L-NAME. 4, when primary stromal cultures were exposed to the ERKi for 16 h, RANKL expression increased significantly to 144 Ϯ 25% of that of control stromal cells. Compared with our previous experiments, ERKi applied for only 6 h during the straining protocol was shown to prevent strain-induced decrements in RANKL mRNA expression but did not affect basal RANKL expression (17). That a longer exposure to ERKi raised RANKL mRNA expression suggests that RANKL expression is affected by ERK1/2 even in the absence of mechanical input, indicating that ERK activation may tonically limit RANKL expression. In sum, inhibition of ERK activation blocks strain effects on both eNOS and RANKL.
We showed, in Fig. 3b, that NO did not rise until 24 h later, and this rise was preceded by an increase in eNOS expression and protein. Since nitric-oxide synthase is known to be activated acutely by strain (40), and Jessop et al. (42) have suggested that of the activation of ERK1/2 in osteoblast-like cells may be dependent on immediate NO production, we considered whether ERK1/2 activation required rapid NO generation during strain. Cells were cultured overnight with L-NAME and then subjected to strain regimen for 10 min. ERK1/2 activation was measured as shown in Fig. 4c. Strain activation of ERK1/2 was not altered in cells where endogenous NO production was prevented by the competitive inhibitor, L-NAME. This indicated that NO does not have an acute role in strain activation of ERK1/2.
Constitutive Activation of MEK Reproduces the Strain Effect on Both Genes-To further understand the relationship between ERK activation and its divergent regulation of eNOS and RANKL, we generated an adenovirus to deliver a constitutively activated MEK to cells. The use of adenovirus to transfer the gene was necessitated because these primary stromal cells are not very susceptible to infection by retro-and adenoassociated viruses as well as to liposome enhanced transient transfection (data not shown). We designed an infection protocol using a type 5 replication-deficient adenovirus encoding a green fluorescent protein driven by the strong cytomegalovirus promoter (Ad.GFP) (gift of Dr. Peter Thule, Emory University). Ad.GFP virus combined with LipofectAMINE infected more than 60% of the stromal cells in culture as assessed by percentage of green fluorescence-positive cells measured by fluorescence-activated cell sorting analysis (data not shown). To study the effect of activated ERK in culture, we thus selected adenovirus as the means of gene delivery.
The Ad.caMEK virus was used after amplification and titering in H293 cells by viral cytopathology. Plaque-purified empty adenovirus (Ad.empty) was used as a control. 24 h after infection with viruses, cell lysates were made to examine activation of ERK1/2. As shown in Fig. 5a, an MOI of 2 was sufficient to cause ERK phosphorylation. Increases in MOI above 4 did not further increase ERK activation. This may be due to other distal and unexplained effects of highly activated ERK systems, such as a balancing check on the MAPK system.
In compiled experiments where stromal cells were infected with Ad.caMEK virus and eNOS and RANKL mRNA was assessed 24 h after infection, the effect of mechanical strain was reproduced. As shown in Fig. 5b, Ad.caMEK stimulated eNOS expression by more than 4-fold at an MOI of 4 (423 Ϯ 117% of that expressed in cultures infected with Ad.empty).
Also, Ad.caMEK inhibited RANKL mRNA expression to a nadir of 56 Ϯ 12% as shown in Fig. 5c. Interestingly, the effect of this sustained ERK1/2 phosphorylation had a reproducibly equivalent effect to the plateau effect of strain on RANKL mRNA inhibition.
To further explore the effect of the Ad.caMEK on eNOS and to prove that NO production increased as a distal result, we also were able to show that NO was altered by ERK1/2 activation. As shown in Fig. 5d, Ad.caMEK caused a dose-dependent increase in NO release into the medium (measured as nitrite).
Compiling three experiments where Ad.empty was used as control (100%) for comparison with Ad.caMEK infection, we found that an MOI of 5 increased cell-generated nitrite to 397 Ϯ 87% of that of control cells, and an MOI of 10 caused increases to 374 Ϯ 66% of Ad.empty control (p Ͻ 0.01 for both conditions). Thus, caMEK delivery reproduced the effects of strain to cause divergent responses in eNOS and RANKL expression.
Inhibition of JNK Does Not Prevent the Strain Response-Strain also causes a sustained activation of JNK, specifically JNK2 kinase (17). With the current availability of a JNKi, we were able to examine the role that strain-induced JNK might play in the differential effects of mechanical factors on eNOS and RANKL. We first replicated our data showing that c-jun phosphorylation was increased after 15 min of strain, as shown in Fig. 6a. Also shown in this figure is a demonstration that 50 M concentration of the JNKi, SP600125, was able to inhibit the strain-induced increase in c-jun phosphorylation. Thus, we were able to inhibit JNK during the application of mechanical strain.
As shown in Fig. 6b, eNOS responded to overnight strain with a doubling of expression and continued to respond significantly to strain even in the presence of JNKi, although the effect was blunted. The question of whether this blunting was due to an already nearly maximal stimulation of eNOS expression or JNK represents a separate pathway by which strain up-regulates eNOS expression remains for future investigation. Similarly to effects on basal cultures caused by inhibition of ERK1/2, JNKi generated an increase in basal eNOS mRNA expression to 168 Ϯ 28% of basal levels.
JNKi also caused a significant increase in basal expression of RANKL as shown in Fig. 6c. In contrast to the effect of inhibition of the ERK signaling pathway, the strain effect prevailed in the presence of JNKi, as shown by the significant difference between control and strain cultures treated with the JNKi (right set of bars).

DISCUSSION
Complex interactions between signals controlling both osteoblast and osteoclast functions are ultimately responsible for the adaptive response of the skeleton to its loading environment. In data presented here, we have shown that application of mechanical strain to bone stromal cells induces a coordinate regulation through MAPK of two genes, eNOS and RANKL, which are predicted to work in synergy to promote positive bone remodeling. The effective levels of strain (ϳ0.25%) in bone stromal cells are far below those affecting vascular tissue, which continue to increase response until at least 20% strain (22), a level well above the fracture threshold of the hard skeleton (43).
Strain, through activation of ERK1/2, caused a doubling in the amount of eNOS expressed and NO generated by primary bone stromal cells. NO has long been implicated as one of the signals transducing mechanical stress in cells (25,44). In the vasculature, both shear (18) and strain (23,45,46) induce nitric oxide, and blocking NO synthesis prevents many of the downstream effects of biophysical inputs. Skeletal response to biophysical input also appears to involve NO metabolism. One of the earliest reports linking NO to bone remodeling showed that NOS inhibitors potentiated ovariectomized bone resorption in rats (27). Since then, studies have shown that NO can slow bone loss in animals (47) and even humans (48,49). Mechanical stimulation has been shown to initiate NO production: loading rat vertebrae and ulnae causes activation of NOS and NO release (50). Inhibition of NO production further appears to impair response to loading: in rats treated with an inhibitor of NOS, a regimen of four-point bending of tibiae induced less new bone formation than in controls (51). Further studies in rats showed that mechanically induced osteogenesis required and was enhanced by endogenous NO production, whereas inhibition of NO production decreased bone apposition (52). In vitro, studies of the effects of mechanical stimulation on osteoblast NO production have shown that both shear and deformation (strain) increase NO production (40,50,(53)(54)(55) and that nitric oxide causes proliferation of bone-forming cells (56). NO is likely to affect bone remodeling, therefore, through targeting the formation and function of both osteoclasts and osteoblasts.
NO in bone is largely generated through eNOS, which is the predominant isoform expressed in skeletal tissue (38,39,57). Transgenic mice lacking eNOS show decreased bone volume and formation until at least 12 weeks, when other adaptive processes must overcome the deficiencies caused by the absence of eNOS (29). eNOS null mice are furthermore unable to respond to estrogen with an anabolic response (58). In terms of the mechanical response, studies have implicated iNOS as also being involved in the reloading bone apposition after loss through hind limb raising; in the iNOS null transgenic, bone loss was similar to control during the unweighting protocol, but reloading failed to induce bone formation in the transgenic animal (5). In our studies, we did not see a response of iNOS to strain. The roles of eNOS and iNOS and their production by the many cells involved in bone remodeling will be important to explore in future studies.
Whereas eNOS is induced by strain, RANKL is coordinately repressed, as demonstrated by the divergence of response during the strain magnitude-dosing protocol (Fig. 3). RANKL inhibition by strain in vitro is discrete, rarely dropping below 50% of control levels, an effect faithfully replicated by activating ERK1/2 through delivery of an adenovirus encoding constitutively active MEK. In contrast, where strain induction of eNOS caused less than a doubling response, constitutive activation of MEK induced anywhere between 2-and 10-fold increases in eNOS expression, reflected in much greater increases in NO generation. This suggests that whereas eNOS response to strain appears to require similar activated MAPK pathways, a constitutively activated ERK1/2 might lead to additional pathways affecting eNOS expression.
Since strain also induces JNK activation, indeed prolonged beyond the transient activation of ERK1/2 (17), we tested the role of JNK in strain-induced increases in eNOS and decreases in RANKL. The JNKi used in these studies was shown to inhibit strain-induced JNK activation. Overnight application of JNKi had significant effects on basal levels of both eNOS and RANKL, suggesting that JNK is involved in regulating the expression of important genes in bone cells. However, JNKi failed to prevent either strain induction of eNOS or strain after infection with Ad.caMEK. Total ERK, show in the bottom row, does not change. b, stromal cells were infected with empty Ad (4 MOI, represented as "0" Ad.caMEK on the x-axis) or the indicated doses of Ad.caMEK, and total RNA was assayed for eNOS levels 24 h later. eNOS rose significantly in the presence of activated MEK (the asterisks show significance at p Ͻ 0.05). c, RANKL was examined in the same samples as for b showing a dose-dependent decrease in RANKL expression in the presence of the Ad.caMEK viral infection. An asterisk shows significance at p Ͻ 0.01. d, stromal cells infected with increasing MOI of Ad.caMEK showed a dose-dependent increase in NO production as assayed in the media using the sensitive nitrite assay. Empty adenovirus was used as control (there was no difference between nitrite in media of uninfected and Ad.empty-infected cells; data not shown). The asterisks show significant difference from Ad.empty at p Ͻ 0.05. inhibition of RANKL expression. While not ruling out subtle differences between the ability of JNK to regulate basal levels of either gene, our data indicate that strain's control of eNOS and RANKL expression operates almost exclusively through the ERK1/2 pathway.
Our results do not rule out modulation of the ERK1/2 pathway by other signaling pathways that are known to be activated by mechanical factors. Shear stress, another relevant mechanical factor in the skeleton, is also known to activate ERK1/2 (42,59) as well as nitric-oxide synthase (40,60). Differences between those signaling pathways activated by shear and stress are subtle; for instance, both shear and strain activate ERK in osteoblast-like cells but through putatively divergent upstream pathways (18,42). Recent papers underline the complexity of shear-induced signaling in osteoblasts in terms of proliferative and differentiative responses, including cyclooxygenase, G-proteins, and nitric-oxide synthases (61,62). This level of complexity is almost certain to be present in the cellular response to mechanical strain. It is possible that many signaling pathways will impact on both the upstream and downstream pathways involving ERK1/2 that regulate eNOS and RANKL expression.
The mechanisms by which ERK1/2 regulates eNOS and RANKL gene expression in bone cells are currently unknown. In the case of endothelial cell eNOS, mechanical shear is known to transiently increase gene transcription, a process dependent on the Src-Ras-MEK-ERK cascade, as well as to invoke a non-ERK-dependent increase in eNOS mRNA stability (18). We were not able to measure strain induction of eNOS in the presence of ERK inhibition, suggesting that strain, at least at magnitudes relevant in skeletal physiology, may be effected largely through increases in eNOS transcription. In terms of RANKL, mechanistic control of this important transcript has so far eluded analysis. Whereas endogenous RANKL expression is acutely sensitive to osteotropic factors (15,16,63), transient transfection of up to 7000 nucleotides of the promoter does not respond to agents known to effectively increase expression (64,65). Understanding RANKL regulation will require novel techniques to probe endogenous gene activity.
Thus, we have shown that strain has a magnitude-specific ability to control two genes in entirely divergent fashions, one inhibitory and one stimulatory. For regulation of eNOS and RANKL, this ability relies on activation of the ERK1/2 signaling cascade, suggesting that ERK represents a common distal pathway where multiple mechanical signals converge. The difference in effect suggests that the final regulatory result is determined by recruitment of specific co-regulators. In total, in vitro strain represents a paradigm for understanding how loading the skeleton results in positive bone remodeling. JNK is activated by strain, and JNKi prevents strain activation. b, two sets of bars are shown, representing variation from control levels of eNOS; the gray bars represent cells exposed to strain overnight. The right set of bars show cells where 30 M JNKi was added. In this set of six experiments, strain again increased eNOS (a unstrained control, p Ͻ 0.0001). In the presence of JNKi, the strained condition continues to be significantly different from its JNKi-treated control, reaching significance at b (p ϭ 0.0005). c, RANKL mRNA was assayed similarly to eNOS in b. Strain inhibited RANKL; a shows a significant decrease compared with unstrained control (p ϭ 0.0038). The strain effect prevailed in the presence of the JNKi; JNKi did not prevent strain-induced inhibition of RANKL mRNA expression; b is different from JNKitreated control (p ϭ 0.0071).