|Home | About | Journals | Submit | Contact Us | Français|
Atherosclerosis begins as local inflammation of artery walls at sites of disturbed flow. The c-Jun NH2-terminal kinase (JNK) is thought to be one of the major regulators of flow-dependent inflammatory gene expression in endothelial cells in atherosclerosis. We now show that JNK activation by both onset of laminar flow and long-term oscillatory flow is matrix-specific, with enhanced activation on fibronectin compared to basement membrane protein or collagen. Flow-induced JNK activation on fibronectin requires new integrin ligation, and requires both the MAP kinase kinase MKK4 and p21-activated kinase (PAK). In vivo, JNK activation at sites of early atherogenesis correlates with the deposition of fibronectin. Inhibiting PAK reduces JNK activation in atheroprone regions of the vasculature in vivo. These results identify JNK as a matrix-specific, flow-activated inflammatory event. Together with other studies, these data elucidate a network of matrix-specific pathways that determine inflammatory events in response to fluid shear stress.
Atherosclerosis is an inflammatory disease in which endothelial cell (EC) activation leads to leukocyte recruitment into artery walls, followed by formation of plaques containing lipid-laden macrophages and smooth muscle cells, with necrotic cores 1. Plaques can occlude vessels and cause ischemia, or rupture to cause stroke or myocardial infarction. While systemic risk factors such as hypertension, smoking and obesity play important roles in atherogenesis, plaques show a predilection for vessel branch points and regions of high curvature where flow is low and shows a variety of complex patterns that are grouped together under the term disturbed flow 2. These areas show increased endothelial cell turnover, altered redox regulation and upregulation of pro-inflammatory genes that contribute to atherosclerotic progression 3,4. By contrast, areas of high laminar shear show down-regulation of pro-atherogenic genes and upregulation of atheroprotective genes, and are resistant to atherosclerosis 5.
In vitro, acute application of laminar flow to unstimulated cells transiently activates inflammatory events and is often used to investigate EC responses to flow. Interestingly, oscillatory flow activates most of the same events in a sustained manner, recapitulates features of athero-prone regions of arteries in vivo and has been used to model complex flow profiles found in vivo. Together, these methods have been widely used to study flow signaling associated with atherosclerosis.
Previous work in our lab showed that integrins are converted to a high affinity state downstream of a cell-cell junction mechanosensory complex consisting of PECAM-1, VE-cadherin and VEGF receptor 2 (VEGFR2) 6. In response to shear stress, this complex stimulates phosphoinositide-3-kinase (PI3K) which then leads to integrin activation. Subsequent binding of newly activated integrins to extracellular matrix (ECM) initiates downstream signals. Consistent with this model, several studies have shown that the subendothelial ECM modulates a subset of endothelial responses to flow 7,8.
ECs in the vasculature normally adhere to a basement membrane composed mainly of laminin, collagen (Coll) IV and entactin/nidogen. In vivo, areas of disturbed shear showing expression of atherogenic genes, such as the adhesion molecules ICAM-1 and VCAM-1, and increased deposition of fibronectin (FN). These changes occur at early times in Apolipoprotein E (ApoE) −/− mice and even in athero-resistant wild type mice 7. Fibrinogen (FG) is deposited at these sites at later stages of atherosclerosis. In general terms, adhesion to basement membranes or collagen is associated with a quiescent cell phenotype, while binding to FN or FG is associated with proliferation and migration in many cellular systems 9. Matrix remodeling may therefore promote an activated EC phenotype in these regions.
In vitro, integrin activation in response to flow has been linked to shear responses including activation of mitogen-activated protein kinases (MAPKs), Rho family GTPases and NF-κB 10–12. Our work has shown that in ECs plated on FN, but not on basement membrane or Coll, shear activates the inflammatory mediators NF-κB and PAK, suggesting that early changes in the subendothelial matrix in atheroprone regions of the vasculature may contribute to atherogenesis through promotion of an activated EC phenotype 7,13.
The c-Jun NH2-terminal kinases (JNKs) are MAPKs traditionally considered stress-activated protein kinases. This subfamily includes JNK1 and JNK2, which are ubiquitously expressed, and JNK3 which is expressed mainly in the heart, brain and testis 14. JNKs can regulate cell proliferation, apoptosis, migration, cytoskeletal rearrangements, inflammation, metabolic disease, neurodegenerative disease, oncogenesis and cancer progression 15. JNK is activated by inflammatory cytokines, and environmental stresses including UV irradiation, osmotic stress, redox stress and mechanical stress 16. JNK stimulates activation of the AP-1 transcription factor, resulting in the expression of inflammatory genes such as monocyte chemotactic protein-1 (MCP-1), IL-8 17,18 and VCAM-1 19.
JNK has been shown to be activated in response to onset of laminar shear, and its activation appears to involve G proteins, PI3K gamma, small GTPases, Src and the upstream kinase MEKK1 20–22. JNK has also been implicated in atherosclerosis since both feeding mice the JNK inhibitor SP600125 and genetic deletion of JNK2 decreased atherosclerotic plaque formation in ApoE−/− mice 23. These links between JNK, inflammation and atherosclerosis prompted us to investigate the activation of JNK in the context of fluid shear stress, ECM remodeling and atherosclerosis.
Bovine aortic endothelial cells (BAECs) were cultured as described 13. Laminar or oscillatory flow was applied to cells on coverslips using a parallel plate flow chamber 13. For Western blots, cells were lysed in SDS sample buffer and run on SDS polyacrylamide gels, transferred to PVDF membranes and probed with antibodies according to standard protocols. For immunoprecipitations, cells were lysed in buffer with 1% Triton X-100 and 0.5% NP-40 plus protease and phosphatase inhibitors, incubated with primary antibody and protein G-Sepharose, then beads were washed and eluted with SDS sample buffer. Eluates were analyzed by Western blotting as above. Immunohistochemistry was performed essentially as described 7. For details see online supplemental information.
JNK is typically activated by phosphorylation by the MAPK kinases (MAPKKs), MKK4 and MKK7, which phosphorylate T183 and Y185 on JNK within the activation motif Thr-Pro-Tyr 24. In bovine aortic endothelial cells (BAECs), both total and phospho-specific JNK antibodies recognized a major p54 and minor p46 band, which co-migrate with JNK2 and JNK1 (Fig. 1A). The p54 JNK2 band was 5.2-fold more intense than the p46 JNK1 band (Fig. 1B), thus, subsequent studies focused on the p54 band. However, the minor p46 JNK1 band behaved similarly in all of the assays.
Previous studies demonstrated that onset of shear stress stimulated JNK activity maximally at 30–60 min 21,25 (and our unpublished data, 2007). To determine whether the activation of JNK by shear stress depends on the subendothelial matrix, BAECs were plated on glass slides coated with either Coll I, matrigel (MG, a solubilized basement membrane preparation) or FN for 4 h, which is sufficient time to spread and form a confluent monolayer. Cells were then subjected to laminar shear stress (12 dynes/cm2) in a parallel plate flow chamber for 45 minutes. Cells on FN showed a marked increase in JNK phosphorylation in response to flow, whereas cells on MG or Coll showed little or no activation compared to static controls (Fig. 1C).
Atheroprone regions of arteries in vivo experience complex flow patterns that are often modeled in vitro by oscillatory flow. To determine whether the effect of ECM also applies to longer term stimulation by oscillatory flow, BAECs plated on different matrices were exposed to oscillatory shear for 18 hours. Activation of JNK was again highly matrix-dependent with a significant increase in JNK in response to flow seen only in cells on FN (Fig. 1D). Additionally, cells on FN exposed to 18 hours of laminar shear showed low JNK phosphorylation as compared to cells exposed to oscillatory flow (Fig. 1E).
Previous studies showed that a number of flow-dependent signals are initiated downstream of de novo integrin binding to the ECM beneath the endothelium, which rapidly occurs after their conversion to a high affinity state 6,10,26. Given the matrix-specificity of JNK activation by shear stress, we next asked whether new integrin ligation following flow mediates this response. To test this idea, BAECs that were stably adherent to FN were pre-incubated for 1 hour with either the anti-FN antibody 16G3, which blocks integrin binding sites on FN, or the control antibody 11E5, before being exposed to the acute onset of laminar shear stress. Treatment with anti-FN antibody under these conditions does not cause cell detachment or disruption of cytoskeletal organization but prevents formation of new integrin binding following integrin activation by shear 10,26,27. Blocking FN abrogated JNK phosphorylation in response to shear stress, indicating that new integrin ligation is required for activation of JNK by flow (Fig. 2).
We next tested whether these findings were relevant to activation of JNK in ECs at atheroprone sites in vivo. Previous work from our lab showed increased deposition of FN in atheroprone segments of arteries even in wild type (WT) mice that do not develop atherosclerosis; furthermore FN localization correlates with inflammatory gene expression 7. We therefore performed immunohistochemical analysis of phospho-JNK and FN in aortas from WT C57BL/6 mice. The greater curvature of the aortic arch is exposed to high laminar shear and is resistant to atherosclerosis, whereas the lesser curvature of the aortic arch is exposed to disturbed flow and is susceptible to atherosclerosis. Staining revealed a continuous FN stain beneath the endothelial layer of the atheroprone lesser curvature of the aortic arch, with phospho-JNK staining localized to nuclei of the endothelial cells in the same region (Fig. 3A). By contrast, the greater curvature showed very little FN or phospho-JNK staining. To quantify the correlation between FN and phospho-JNK, the aortic arches from 3 different mice were split into equal sized sections and scored for staining of FN and phospho-JNK in each section. FN co-localized with phospho-JNK with a frequency of 0.88, while for the random distribution (null hypothesis) co-localization occurred with a frequency of 0.54, p = 0.003.
To further investigate whether phospho-JNK is present in atherosclerotic lesions in vivo, hypercholesterolemic ApoE−/− mice that develop atherosclerotic plaque at regions of disturbed flow were examined 28. These mice were fed a high fat, Western diet for 10 weeks to accelerate plaque formation. Mice were then sacrificed and the carotid arteries stained for FN and phospho-JNK (Fig. 3B). The carotid sinus is a classic site of atherosclerosis and previous work showed staining for the ECM proteins FN and the inflammatory markers ICAM-1 and VCAM-1 in this region 7. We observed pronounced staining of FN in the carotid sinus of these mice which co-localized with phospho-JNK staining, whereas nearby regions of the vessel wall that did not experience disturbed flow showed little staining for any of these proteins.
We next investigated the upstream MKK in this pathway. To date, only MKK4 and MKK7 have been found to mediate JNK activation. MKK4, MKK7 or control siRNAs were used to test which of these MAPKKs mediate shear-induced JNK activation. Transient transfection of BAECs with MKK4 and MKK7 siRNAs decreased the level of those proteins by approximately 80–90% (Fig. 4A). While depletion of MKK4 inhibited shear-induced JNK phosphorylation, neither MKK7 nor control siRNAs had significant effects (Fig. 4A). To confirm this result, a rescue experiment was performed in which BAECs were co-transfected with MKK4 siRNA together with either empty vector, HA-tagged MKK4 or myc-tagged MKK7. Only the HA-MKK4 significantly rescued shear-induced JNK phosphorylation (Fig. 4B).
We also assessed the role of these MAPKKs by measuring their activation in response to flow. BAECs were plated on either Coll or FN exposed to short-term laminar shear. Activation was assessed by Western blotting with antibodies that recognize phosphorylated MKK4 or phosphorylated MKK7. Consistent with the matrix specificity of JNK activation, MKK4 phosphorylation increased in cells plated on FN, but decreased to slightly below baseline levels in cell on Coll (Fig. 5A). By contrast, MKK7 phosphorylation was unaffected by shear on either matrix (Fig. 5B). Finally, we examined cells in oscillatory shear. MKK4 phosphorylation was significantly elevated after 18h of oscillatory shear (Fig. 5C), whereas MKK7 showed no change (unpublished data, C. Hahn). Taken together, these data confirm the matrix-dependence for this pathway and identify MKK4 as the MAPKK responsible for ECM-dependent JNK activation in response to shear.
The p21-activated kinase (PAK) Ser/Thr kinases have also been implicated in activation of JNK in some systems 29. PAK is an interesting candidate since, like JNK, it is activated by flow in a matrix-specific manner, and is activated in atheroprone areas of arteries 8. In those experiments, PAK was found to mediate local breakdown of cell-cell junctions and increased vascular permeability. Inactive PAK is a homodimer in which its N-terminal autoinhibitory domain (PAK AID) binds and inhibits the C-terminal kinase domain in trans. Binding of active Rac or Cdc42 to PAK relieves this autoinhibition, allowing autophosphorylation of the protein on Ser141 and Ser423, which lead to sustained activation 30. Active PAK is recruited to target sites on the cell membrane via interaction of an N-terminal proline-rich sequence with Nck.
To investigate the involvement of PAK in JNK activation by flow, BAECs on FN were transfected with a vector encoding HA-tagged PAK AID or control vectors, and exposed to onset of laminar shear. Compared to empty vector and myc-tagged PAK2 controls, PAK AID completely blocked JNK activation in response to shear (Fig. 6A). To confirm these results, BAECs were treated with 20 µg/mL of either a control peptide or a peptide that contains the Nck-binding sequence of PAK fused to the transduction sequence from the HIV TAT protein; this peptide enters cells and blocks PAK function in several systems, both in vitro and in vivo 8,31–33. The blocking peptide efficiently inhibited the activation of JNK in response to onset of laminar shear. By contrast, a control peptide in which two critical prolines are mutated had no effect (Fig. 6B). Thus, PAK appears to be required for activation of JNK by flow in cells on FN.
Conversely, we tested the effect of activating PAK in cells on Coll where JNK activity is suppressed. Cells were transfected with either WT PAK as a control or a constitutively active PAK T423E mutant. We found that expression of even low levels of PAK-T423E increased JNK activity in the absence of flow (Fig. 6C). Applying flow to these cells had no further effect (not shown). This result shows that active PAK is sufficient to activate JNK in the absence of other stimuli.
PAK is activated in a sustained manner in response to oscillatory shear 13. To address whether PAK is required for activation of JNK under this condition, BAECs on FN were transfected with the HA-PAK AID construct and exposed to oscillatory shear for 18h. Inhibiting PAK strongly attenuated JNK activation compared to the empty vector control (Fig. 6D). Taken together, the data show that PAK is the critical determinant of matrix-specific JNK activation by both onset of laminar shear and by oscillatory shear.
Given these data and the activation of PAK and JNK at similar sites in arteries, we wondered whether PAK is a critical determinant of JNK activation at regions of early atherogenesis in vivo. To address this question, C57BL/6 mice were injected with either control or PAK-Nck blocking peptide once a day for three days, then carotid arteries were harvested and analyzed by immunohistochemistry for phospho-JNK. Mice treated with the control peptide had the expected phospho-JNK staining in the nuclei of the ECs in the carotid sinus, a well-described atheroprone region where flow is disturbed. By contrast, mice treated with the PAK-Nck blocking peptide showed an approximately 50% decrease in the number of cells that scored positive for JNK per length of vessel perimeter (Fig. 7A, B; p<0.05). These results confirm the importance of PAK in JNK activation in a region of disturbed flow in vivo.
These results show that JNK is activated in a matrix-specific manner in response to fluid shear stress. JNK is activated at early times by acute onset of laminar shear and at later times by oscillatory shear in cells on a FN matrix, but little or no activation occurred in cells on basement membrane protein or type I Coll. JNK activation is also seen in vivo at atheroprone regions of arteries coincident with FN in the subendothelial matrix. Matrix-specific activation of JNK by shear stress requires new integrin ligation, since blocking integrin binding to FN abrogated its activation. JNK activation by shear also requires MKK4 and PAK. Both kinases are activated by acute onset of laminar flow or by oscillatory shear stress in a matrix-specific manner, and inhibiting either kinase blocked JNK activation. Furthermore, a small peptide inhibitor of PAK decreased JNK activation in atheroprone areas of the vasculature in vivo. Taken together, these results suggest that JNK is activated by disturbed flow in ECs adherent to FN in vivo.
JNK is known to play a key role in expression of cytokines and adhesion molecules that mediate leukocyte recruitment and activation 18–21,34. Inhibiting or deleting JNK also reduced atherosclerosis in ApoE−/− mice 14. JNK can mediate apoptosis under stressful conditions 16, thus, could contribute to the elevated rates of apoptosis at atheroprone regions in vivo. Taken together, the data therefore suggest that ECM-dependent JNK activity is likely to play a significant role in atherogenesis.
Endothelial cells are normally adhered to a basement membrane in which the major components are LN, which binds integrins α6β1, α6β4 and α2β1, and Coll IV, which binds integrins α2β1 and α1β1 35,36. Injury, inflammation and angiogenesis that promote vascular remodeling lead to deposition of a provisional matrix containing proteins including FN and FG, which bind primarily α5β1 and αv integrins 37. Binding of these integrins to matrix is associated with an activated cell phenotype, with increased cell migration and proliferation 9,38, consistent with the EC phenotype in atheroprone regions 39,40 . This matrix remodeling most likely contributes to resolving the inflammation or injury in response to acute insults but under chronic stresses may have deleterious aspects.
Previous work in our lab showed that the altered subendothelial matrix at atheroprone sites influences activation of the transcription factor NF-κB, a key regulator of inflammatory gene expression during atherosclerotic progression 7. We also recently reported that matrix-specific NF-κB activation by flow is mediated by matrix-specific activation of PAK 13. That JNK, a third key inflammatory mediator, is activated in a matrix- and PAK-dependent manner therefore indicates, first, that matrix remodeling is a major determinant of endothelial activation in this system. It also reveals that the ECM regulates EC phenotype through a network of pathways in which PAK appears to be the central determinant, accounting for both JNK and NF-κB, as well as mediating effects on junctional integrity and permeability more directly.
The result that shear-induced JNK activation is among the events that depend on new integrin ligation also provides additional support for the junctional mechanotransducer model. In this model, integrin activation and ECM binding occur downstream of a junctional complex that stimulates integrin activation through PI3K (Fig 8). These data fit well with previous results describing a role for PI3K gamma in JNK activation by shear stress 22. Thus, a series of events beginning with rapid activation of signaling proteins in cell-cell junctions, followed by integrin activation and binding, leading to JNK activation and leukocyte recruitment appear to be components of a single pathway. Our results also show that MKK4 and PAK are upstream of matrix-specific JNK activation. Despite being identified as a potential activator of JNK some years ago 41, how PAK affects this pathway remains unknown. Upstream kinases such as MEKK1 or scaffold proteins such as JIPs are possible targets. Further work will be required to address this question.
Distinct from our results, it has been reported that laminar flow inhibits JNK activation in ECs by inflammatory cytokines such as tumor necrosis factor 42–44. This effect was proposed to be one of the means by which long term laminar shear is atheroprotective. Under these conditions, inhibition of JNK was mediated through MEK5 and ERK5/BMK1, which inhibited the MAPKKK ASK1 by inducing thioredoxin interacting protein 42,43,45. These results may appear to conflict with ours and those of other labs that report activation of JNK by flow 20–22,46,47. However, studies demonstrating suppression of JNK by laminar shear used cells plated on denatured collagen (gelatin). Thus, the matrix specificity of JNK activation may resolve the discrepancy.
In conclusion, our data provide evidence for JNK activation in atheroprone regions of the vasculature through a pathway that involves both disturbed flow and ECM remodeling. Together with published results, these data suggest that JNK contributes to atherogenesis in vivo. These results raise many questions for future work. Which factors determine local PAK activation ECM remodeling in areas of disturbed flow in vivoare major unsolved questions. Although roles for JNK in expression of inflammatory genes such as MCP-1, IL-8, VCAM-1 and prostaglandin D synthase 18,20,48, and in EC apoptosis and cell migration 15 have been identified, elucidation of the specific roles for JNK during atherogenesis will also be an important area for future research. Understanding the inflammatory pathways controlled by flow may provide novel therapeutic strategies for modulating atherosclerotic progression.
The authors are grateful to Melanie Cobb for MKK4 and MKK7 constructs. We thank Dr. Chong Wang for measurements of flow velocity from the peristaltic pump.
Sources of Funding
This work was supported by United States Public Health Service grants RO1 HL75092 to M.A.S. and training grant 5T32-HL007284-28, 29, 30 to C.H.