Low Intensity Shear Stress Increases Endothelial ELR+ CXC Chemokine Production via a Focal Adhesion Kinase-p38β MAPK-NF-κB Pathway*

CXC chemokines with a glutamate-leucine-arginine (ELR) tripeptide motif (ELR+ CXC chemokines) play an important role in leukocyte trafficking into the tissues. For reasons that are not well elucidated, circulating leukocytes are recruited into the tissues mainly in small vessels such as capillaries and venules. Because ELR+ CXC chemokines are important mediators of endothelial-leukocyte interaction, we compared chemokine expression by microvascular and aortic endothelium to investigate whether differences in chemokine expression by various endothelial types could, at least partially, explain the microvascular localization of endothelial-leukocyte interaction. Both in vitro and in vivo models indicate that ELR+ CXC chemokine expression is higher in microvascular endothelium than in aortic endothelial cells. These differences can be explained on the basis of the preferential activation of endothelial chemokine production by low intensity shear stress. Low shear activated endothelial ELR+ CXC chemokine production via cell surface heparan sulfates, β3-integrins, focal adhesion kinase, the mitogen-activated protein kinase p38β, mitogen- and stress-associated protein kinase-1, and the transcription factor.

CXC chemokines with a glutamate-leucine-arginine (ELR) tripeptide motif (ELR ؉ CXC chemokines) play an important role in leukocyte trafficking into the tissues. For reasons that are not well elucidated, circulating leukocytes are recruited into the tissues mainly in small vessels such as capillaries and venules. Because ELR ؉ CXC chemokines are important mediators of endothelial-leukocyte interaction, we compared chemokine expression by microvascular and aortic endothelium to investigate whether differences in chemokine expression by various endothelial types could, at least partially, explain the microvascular localization of endothelial-leukocyte interaction. Both in vitro and in vivo models indicate that ELR ؉ CXC chemokine expression is higher in microvascular endothelium than in aortic endothelial cells. These differences can be explained on the basis of the preferential activation of endothelial chemokine production by low intensity shear stress. Low shear activated endothelial ELR ؉ CXC chemokine production via cell surface heparan sulfates, ␤ 3 -integrins, focal adhesion kinase, the mitogen-activated protein kinase p38␤, mitogen-and stress-associated protein kinase-1, and the transcription factor.
ELR ϩ CXC chemokines are critical regulators of leukocyte trafficking into inflamed tissues (4,5,7,10,11). For reasons that are not well elucidated, endothelial-leukocyte interaction and leukocyte emigration into the tissues is primarily concentrated in small vessels such as capillaries and post-capillary venules (8,9). This microvascular localization of endothelial-leukocyte interaction is only partially explained by the lower intensity of the hemodynamic shear forces in smaller vessels, which, in terms of possible mechanical interference, may be more conducive to leukocyte binding to the vessel wall than the higher shear intensities seen in the arterial trunks (7,12). The magnitude of shear stress in capillaries and post-capillary venules is Յ5 dynes/cm 2 (Յ0.5 pascal), which is significantly lower than the 10 -30 dynes/cm 2 (1-3 pascals) in arteries (13). Because ELR ϩ CXC chemokines are important mediators of endothelial-leukocyte interaction, we compared chemokine expression by microvascular and aortic endothelium to investigate whether differences in chemokine expression by various endothelial types could, at least partially, explain the microvascular localization of endothelial-leukocyte interaction. Hemodynamic shear is an important modulator of gene expression in endothelial cells (14 -17), but there are conflicting data on the effect of shear stress on IL-8/CXCL8 expression in endothelial cells (14,15,17,18). Therefore, we further investigated whether shear-induced chemokine expression in endothelial cells is related to the intensity of shear stress.
Using both in vitro and in vivo models, we show that ELR ϩ CXC chemokine expression is higher in microvascular endothelium than in aortic endothelial cells and that these differences can be explained on the basis of the preferential activation of endothelial chemokine production by low intensity shear stress. Low shear activation of endothelial ELR ϩ CXC chemokine production involves a signaling pathway comprised of cell surface heparan sulfates, ␤ 3 -integrins, focal adhesion kinase (FAK), p38␤ mitogen-activated protein kinase (MAPK), mitogen-and stress-associated protein kinase (MSK)-1, and NF-B.

EXPERIMENTAL PROCEDURES
Animals-All animal studies were approved by the local Institutional Animal Care and Use Committee. Three mixed breed pigs (10 Ϯ 4 kg) were anesthetized by intramuscular telazol (4.4 mg/kg), xylazine (4.4 mg/kg), and atropine (0.04 mg/kg) and inhaled isoflurane and were mechanically ventilated in a dorsally recumbent position. The right carotid artery and left jugular vein were cannulated, and blood samples were drawn from the cannulae. Capillary blood samples were collected using a glass capillary tube from a 1-inch long, 3-mm deep incision on the ventral surface of the right ear. Tissues (aorta, heart, jejunum, and kidney) were harvested after euthanasia.
Endothelial Cells-Human umbilical venous endothelial cells (HUVEC) were isolated by trypsin treatment of umbilical cord veins (19). Umbilical cords and cord blood were obtained from term placentae after approval by the local Institutional Review Board. Cells were seeded on flasks coated with a mixture of fibronectin (10 g/ml; Sigma) and gelatin (2% w/v, Sigma) and grown in the complete EGM-2 medium (Lonza Biosciences, Walkersville, MD). The cultures were purified by positive immunoselection using anti-CD31 ferromagnetic microbeads (Miltenyi, Auburn, CA) as per the manufacturer's instructions. The experiments were performed on culture passages 3 and 4. Human aortic endothelial cells (HAEC) and human microvascular endothelial cells (HMEC) (pulmonary, dermal) endothelial cells (Lonza) were used at passages 3-5.
Shear Stress-Endothelial cells were grown to confluence in 100 ϫ 20-or 35 ϫ 10-mm dishes and exposed to shear stress by using a cone-plate viscometer as described elsewhere (20). The cone-plate system is one of two devices commonly used to apply shear stress in vitro, the other being a parallel plate flow chamber (21). The cone-plate system provides a wider range of shear intensities and in greater uniformity of shear forces throughout the field as compared with the parallel plate flow chamber (21,22). Using a static viscosity of 0.8 centipoise for culture media (20), the shear stress applied with our cone at 6 rotations/s was computed as 20 dynes/cm 2 . Shear intensities of 4 and 10 dynes/cm 2 were attained by variation of rotational speed. We used 20 dynes/cm 2 to represent the wall shear stress in larger arteries (23), 4 dynes/cm 2 to simulate the physical conditions in capillaries and venules (24), and an intermediate intensity (10 dynes/cm 2 ). Cell viability (trypan blue exclusion) was unchanged after shear.
Neutrophil Chemotaxis-Neutrophil chemotaxis was measured in microchemotaxis chambers as described previously (details provided in the supplemental material) (5). We used supernatants from endothelial cultures in lower wells of microchemotaxis chambers (NeuroProbe, Gaithersburg, MD). The number of migrating cells was read off a standard curve generated from known numbers of labeled cells.
Real Time PCR-Total RNA was isolated using the TRIzol reagent (Ambion) as per the manufacturer's instructions. Primers were designed using Beacon Design software (Bio-Rad). Real time PCR protocol and the primer sequences are described in the supplemental material (25). The data were normalized for glyceraldehyde-3-phosphate dehydrogenase, and gene expression was compared between samples by using the 2 Ϫ⌬⌬CT method.
Immunocytochemistry-HUVEC grown on glass coverslips were exposed to shear for Յ6 h. The immunocytochemistry protocol is described in the supplemental material (26).
Electrophoretic Mobility Shift Assay-The electrophoretic mobility shift assay protocol has been described previously (28). We used the consensus binding sequence (underlined) for NF-B (5Ј-AGTTGAGGGGACTTTCCCAGG C-3Ј prelabeled with an infrared fluorophore (Licor, Lincoln, NE). For supershift assay, the nuclear extracts were incubated with the anti-p65 antibody (1 g, 30 min on ice). The gels were analyzed in an infrared scanner (Licor).
Luciferase Assays-Cell extracts after 8 h of shear stress were used in a luciferase assay system (Promega) as per the manufacturer's instructions.
Statistical Methods-Nonparametric tests were applied using SigmaStat 3.1.1 software (Systat, Point Richmond, CA). A p value of Ͻ0.05 was considered significant.

ELR ϩ CXC Chemokine Production by Microvascular and Aortic
Endothelial Cells-To determine whether ELR ϩ CXC chemokines are differentially expressed by microvascular and aortic endothelium, we first compared IL-8/CXCL8 and GRO-␣/CXCL1 production by low passage HMEC and HAEC in vitro. As seen in Fig. 1A, basal chemokine production in HMEC was higher than HAEC. HMEC chemokine production was also higher than HAEC following stimulation with 1-10 ng/ml tumor necrosis factor-␣ or Escherichia coli lipopolysaccharide (10 -500 ng/ml; data not depicted). These data were supported by subsequent in vivo measurements of porcine IL-8 in simultaneously drawn arterial, venous, and capillary blood samples from anesthetized pigs. Porcine IL-8 concentrations in capillary/venous blood were higher than in arterial samples (Fig. 1B), providing further indirect evidence for a microvascular origin for circulating ELR ϩ CXC chemokines. There was also strong immunoreactivity for IL-8 in microvascular but not in aortic endothelial cells (Fig. 1B, inset).
Low Intensity Laminar Flow Shear Increases ELR ϩ CXC Chemokine Expression in Endothelial Cells-To determine the mechanism(s) underlying the observed differences in chemokine expression by microvascular and aortic endothelium, we next investigated the hypothesis that low intensity shear stress in capillaries and veins (24) is a more efficient activator of chemokine expression in vascular endothelial cells than the higher shear intensities seen in arteries. We exposed HAEC, HMEC, and HUVEC to low (3 dynes/cm 2 ) and high (20 dynes/cm 2 ) laminar flow shear stress and measured IL-8/CXCL8 and GRO-␣/CXCL1 production over 18 h. Low intensity shear increased chemokine production in HAEC and HMEC, whereas high shear suppressed chemokine expression to below static levels ( Fig. 2A). To confirm the functional relevance of the chemokine levels that we measured in our culture supernatants, we next measured the ability of these supernatants to recruit neutrophils in vitro. As seen in Fig. 2B, low, but not high, shear increased the neutrophil chemotactic activity in endothelial culture supernatants.

Low Shear Induces Endothelial Chemokines via a FAK-p38-NFB Pathway
We also tested the cell lysates to measure intracellular and membrane-bound chemokines, which play an important role in endothelial-neutrophil interaction (32). Similar to changes in secreted chemokines, low shear up-regulated IL-8/CXCL8 and GRO-␣/CXCL1 concentrations, whereas higher shear intensities suppressed the expression of these two chemokines. Shear did not change GRO-␤/ CXCL2, GRO-␥/CXCL3, and NAP-2/CXCL7, whereas GCP-2/CXCL6 remained low near the lower limit of detection.
We next used a real time PCR microarray to investigate the effect of low and high shear intensities on the global cytokine/chemokine expression profile of endothelial cells. As seen in Fig. 2B, low shear increased the expression of a large number of chemokines and cytokines. In contrast, high shear suppressed the expression of all chemokines in the CXC, CC, C, and the CX 3 C families but differentially altered the expression of other cytokine subgroups. These data are consistent with our hypothesis that low intensity shear stress increases chemokine expression in microvascular endothelium, which may explain, at least partially, the microvascular localization of endothelial-leukocyte interaction.
Cell Surface Heparan Sulfates, ␤ 3 -Integrins, and FAK Transduce the Mechanical Signal of Low Intensity Shear Stress-To determine the signaling mechanism by which low shear up-regulates chemokine expression, we first investigated whether disruption of cell surface heparan sulfate proteoglycans (HSPGs), which are important endothelial mechanotransducers (33,34), can block low shear-induced chemokine production. We used a previously reported protocol of heparinase treatment (15 units/ml x 2 h) that has been shown to remove a substantial fraction of the HSPGs. As shown in Fig. 3A, heparinase treatment blocked low shear-induced IL-8 and GRO-␣ production.
We next investigated the role of FAK in low shear-induced chemokine production. As shown by Western blot and immunocytochemistry in Fig. 3C, low shear phosphorylated FAK in endothelial cells. Up-regulation of FAK by transient transfection increased low shear-induced chemokine production, whereas inhibition of FAK expression by overexpressing FRNK blocked chemokine production.
Shear-induced ELR ϩ CXC Chemokine Expression Requires p38␤ and MSK-1-An antibody array identified p38␤ and p38␦ as the two activated MAPKs in low shear-exposed cells (supplemental Fig. S3). Shear-induced p38 phosphorylation was confirmed by Western blotting and immunocytochemistry (Fig. 4A). However, low shear activation of endothelial chemokine production was completely blocked by SB202190, which inhibits p38␤, but not p38␦ (Fig. 4B). Furthermore, when individual p38 isoforms were overexpressed in endothelial cells through transient transfection, p38␤ was the most efficient p38 iso- shear-induced endothelial chemokine production. Overexpression of p38␤ increased shear-induced endothelial chemokine production. The statistical methods are as above. Transfection was confirmed by Western blotting for a FLAG sequence in the plasmid. D, knocking down p38␤, but not p38␦, by RNA interference inhibited low shear-mediated endothelial chemokine production. The bars show chemokine concentrations (means Ϯ standard errors) as measured by ELISA. The statistical methods are as above. Insets show Western blots to confirm that the siRNA reagents effectively down-regulated the expression of the specific isoform. The data in these figures are representative of at least three separate experiments, where each study group was in triplicate.  FEBRUARY 27, 2009 • VOLUME 284 • NUMBER 9 JOURNAL OF BIOLOGICAL CHEMISTRY 5951 form in up-regulating shear-activated chemokine expression (Fig. 4C). We further confirmed these data by using RNA interference to knock down p38␤ 2 (the major splice variant of p38␤) (35) and p38␦; down-regulation of p38␤ 2 , but not p38␦, blocked low shear-induced chemokine production (Fig. 4D).

Low Shear Induces Endothelial Chemokines via a FAK-p38-NFB Pathway
Among the downstream targets of p38, shear is a known activator of MSK-1 and ribosomal S6 kinase-2 (36). MSK-1 activation in endothelial cells is almost exclusively dependent on p38 signaling (36), and therefore, we investigated the role of MSK-1 in shear-induced chemokine production. As shown in Fig. 5A, low shear phosphorylated MSK-1. Inhibition of MSK-1 activity (Ro318220) partially blocked chemokine production (Fig. 5B), indicating that MSK-1 is an important, although possibly not the sole mediator of shear-induced chemokine production.
Low Shear-induced ELR ϩ CXC Chemokine Production Does Not Require rho-associated Protein Kinase or Phosphatidylinositol 3-Kinase-rho-associated protein kinase and phosphatidylinositol 3-kinase induce endothelial chemokine expression during inflammation (37). However, inhibition of these enzymes by using wortmannin and Y-27632, respectively, did not block low shear-induced chemokine production in our system (not depicted).
Shear-induced ELR ϩ CXC Chemokine Expression Occurs via NF-B Activation-We next investigated the role of NF-B, which is an important transcriptional regulator of ELR ϩ CXC chemokines and is also a major downstream target of MSK-1 (38). We first sought "shear-activated" NF-B genes in a PCR-based array. Low shear increased the transcription of NFKB2 (p52/p100), REL, and REL A/p65 (supplemental Fig.  S4). Because REL A/p65, the most highly up-regulated gene in this family following exposure to low intensity shear, is a well known substrate for MSK-1, we used p65 in subsequent experiments to investigate shear activation of the NF-B pathway.
Low shear caused the phosphorylation of NF-B p65 Ser 276 (Fig.  6A) and increased specific binding of NF-B proteins to DNA (Fig. 6B) and NF-B-mediated activation of IL-8 promoter (Fig. 6C). We next used signaling inhibitors to block various steps of the NF-B pathway (39): (a) inhibitors of I-B degradation including TPCK (a serine/cysteine proteases inhibitor) and MG-132 (a proteasomal inhibitor); (b) pyrrolidine dithiocarbamate (inhibits nuclear translocation of NF-B proteins); (c) parthenolide (inhibits I-B kinases); and (d) curcumin, which prevents NF-B activation upstream of the I-B kinases. However, all inhibitors except curcumin induced apoptotic changes in endothelial cells as early as 5-6 h. Curcumin inhibited low shearinduced chemokine mRNA and protein expression (supplemental Fig. S5) and established the quantitation of mRNA as an early measure of chemokine production. TPCK, MG-132, pyrrolidine dithiocarbamate, and parthenolide blocked shear-induced chemokine mRNA expression at 4 h, an early time point selected before the onset of apoptotic changes associated with NF-B inhibitors (Fig. 6D).
Shear-induced Intracellular Signaling Is Comprised of Discrete, Sequential Steps-To confirm the sequence of activation of FAK, p38, MSK-1, and NF-B, the major signaling mediators involved in low shear activation of chemokine production, we repeated Western blots to measure phosphorylation in the presence of specific inhibitors. As shown in Fig. 7A, down-regulation of FAK activity by FRNK inhibited the phosphorylation of p38, MSK-1, as well as NF-B p65. In contrast, p38 inhibition FIGURE 6. Low shear activates the NF-B pathway. A, a representative Western blot shows shear-induced phosphorylation of NF-B p65/rel A. The bar diagram (means Ϯ standard errors) shows average densitometric analysis of the blots. The measurements were compared by the Kruskal-Wallis H test/Dunn's post-test; * denotes p Ͻ 0.05. Inset, immunocytochemistry shows nuclear translocation and phosphorylation of NF-B p65/rel A. B, low shear increased specific p65 binding to the DNA, as detected in electrophoretic mobility shift assay. Lanes 1-3 represent nuclear extracts from static cultures, whereas lanes 4 -6 show extracts after exposure to low shear for 2 h. Each set includes the incubation of the nuclear extract with a probe specific for NF-B, a mutated (scrambled) probe, and with the specific probe and a specific antibody (supershift). C, low shearactivated IL-8/CXCL8 transcription is mediated via NF-B, as seen in luciferase reporter assays utilizing constructs with the wild-type and NF-B site-mutated IL-8 promoter sequence. The bar diagram (means Ϯ standard errors) shows the relative light unit measurements; statistical analysis was as above. D, low shear-induced chemokine mRNA expression was blocked by NF-B inhibitors including agents that inhibit I-B kinases (parthenolide), block I-B degradation (TPCK and MG-132), and inhibit the nuclear translocation of NF-B proteins (pyrrolidine dithiocarbamate (PDTC)). These agents were used for 4 h because of onset of apoptotic changes in endothelial cells with longer exposures. The bar diagram shows fold changes in chemokine mRNA above glyceraldehyde-3-phosphate dehydrogenase (means Ϯ standard errors) as a function of time; statistical analysis as above. The data in these figures represent at least three separate experiments.
did not affect FAK but prevented the activation of MSK-1 and NF-B. Finally, inhibiting MSK-1 activity did not affect the phosphorylation of FAK and p38 but blocked the phosphorylation of p65. To ensure that these results were not affected by possible nonspecific effects of the inhibitors on other signaling mediators, we further confirmed these data by using specific siRNA to knockdown p38␤ and MSK-1 (supplemental Fig. S6). The signaling pathway mediating low shear-induced chemokine expression is summarized in Fig. 7B.

DISCUSSION
We present a detailed analysis of the signaling mechanisms by which low intensity shear stress activates ELR ϩ CXC chemokine production in endothelial cells. Differences in the effects of low versus higher shear intensities can explain the differential expression of chemokines we observed in microvascular and aortic endothelial cultures. Our findings are in agreement with previous reports of Chen et al. (15) and Cheng et al. (17) but in contrast with the observations of Kato et al. (18) and Partridge et al. (40), who reported that shear stress suppressed endothelial IL-8 production. Kato et al. (18) and Partridge et al. (40) used relatively high shear intensities of 7 and 12 dynes/cm 2 , respectively, which can explain the observed differences.
We have shown that low intensity shear stress activates endothelial ELR ϩ CXC chemokine production via a HSPG/␤3-integrin/FAK/p38␤/MSK-1/NF-B pathway. The important role we observed for endothelial HSPGs is consistent with previous reports of HSPG effects in mechanotransduction (33,34). Endothelial glycocalyx is the thickest in the microvasculature (41), the site of chemokine production in our studies, and is consistent with our data that heparan sulfates (present in the glycocalyx) are involved in chemokine expression. Furthermore, HSPGs can amplify the effects of intracirculatory chemokine gradients by immobilizing the chemokines to the endothelial cell membrane, thereby facilitating detection by migrating leukocytes. Using immunoelectron microscopy, Middleton et al. (42) demonstrated that endothelial cells in postcapillary venules and small veins, but not in other blood vessels, display IL-8/CXCL8 on the luminal surface. HSPGs interact with fibronectin-binding integrins and FAK (33,34), and consistent with these reports, we detected that low shear activated ␤ 3 -and ␤ 1 -integrins. Although we did not investigate individual integrins, there is evidence that ␣ v ␤ 1 , ␣ v ␤ 3 , ␣ 3 ␤ 1 , and ␣ 3 ␤ 3 integrins identified in our PCR array are involved in focal adhesions and intercellular attachments (43,44).
Similar to our findings, FAK-induced p38 activation has been observed in cardiac myocytes, where mechanical stretch phosphorylates FAK to generate a binding site for grb2, a scaffolding protein that links FAK to the MAPK pathway (45). In endothelial cells, Li et al. (46) reported a transient activation of p38 after exposure to 12 dynes/cm 2 shear, but they did not measure the effect of lower intensities. Although the role of p38 in shearinduced ELR ϩ CXC chemokine expression has not been examined earlier, p38 is a known regulator of endothelial chemokine expression in response to environmental stress, ultraviolet light, heat, osmotic shock, microbial products, and inflammatory cytokines (47). We demonstrate the importance of the p38␤ isoform in shear effects, which is consistent with previous observations that endothelial cells express p38␤ as a major isoform and are relatively deficient in p38␦ (35). The activation of p38␤ is interesting because unlike p38␣, p38␤ has an anti-apoptotic role in endothelial cells, which is consistent with the cytoprotective effects observed with both shear stress as well as ELR ϩ CXC chemokines on endothelial cells (48,49). Further studies are needed to determine whether anti-apoptotic effects of shear in endothelial cells are mediated via ELR ϩ CXC chemokines.
In our study, p38 activated MSK-1, which, in turn, phosphorylated NF-B p65. MSK-1 is a key nuclear regulator of NF-B FIGURE 7. Low shear-induced endothelial signaling involves a series of discrete, sequential steps. A, inhibition of FAK by transient transfection with FRNK blocked low shear activation of p38, MSK-1, and NF-B. p38 inhibition blocked MSK-1 and NF-B activation but not that of FAK. Similarly, MSK-1 inhibition did not affect FAK and p38 but prevented NF-B activation. B, proposed signaling pathway involved in low shearmediated endothelial chemokine production. p38 MAPK is shown at the cusp of cytoplasm and nucleus because the site of its activation is not exactly known. These data are representative of at least three separate experiments.
p65-mediated transcription (50). The partial inhibition of low shear-mediated chemokine expression by Ro318220, which inhibits MSK-1, can be explained because other kinases including casein kinase II, Akt, I-B kinases, protein kinase A, and the p38 MAPKs can also phosphorylate p65 (38). p38 may also increase NF-B-mediated gene transcription by increasing I-B degradation and thereby promoting the translocation of the translocation of NF-B proteins into the nucleus (51,52). The involvement of the NF-B pathway in shear activation of endothelial chemokine production is consistent with previous reports of shear augmentation of the effects of inflammatory agents such as reactive oxygen species and tumor necrosis factor-␣ on NF-B activation (40,53,54).
The preferential activation of endothelial chemokine expression by low shear is consistent with the concentration of endothelial-leukocyte interaction in capillaries and post-capillary venules. Further work is needed to investigate the role of endothelial-derived ELR ϩ CXC chemokines in the normal noninflammatory circulation of leukocytes. These chemokines may provide a mechanism for the high neutrophil concentrations seen in the capillaries in the physiological state. During flow through narrow blood vessels, the blood cells are normally displaced toward the axis of the vessel to create a cell-depleted wall region, which facilitates flow by lowering the apparent viscosity (the Fahraeus-Lindqvist effect) (55). At branch points, this celldepleted wall region allows a progressively smaller fraction of cells to enter the daughter vessel than the parent vessel, causing a progressive reduction in hematocrit as blood courses its way from arterioles to capillaries (55). However, in contrast to RBCs, the neutrophil concentrations in capillaries are up to 40 -80 times higher than in larger blood vessels (56,57). Although some differences can be explained by the longer neutrophil transit times through capillaries, chemokine gradients may promote neutrophil circulation via narrow capillaries instead of the larger "thoroughfare" arteriolovenular shunts (58).