| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Department of Biomedical Engineering, Duke University, Durham, North Carolina
Submitted 27 February 2007 ; accepted in final form 27 August 2007
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONREFERENCES |
|---|
hemodynamics; endothelium; in vitro; endothelial nitric oxide synthase
The localization of early lesions has been reviewed by Glagov et al. (6). Lesion predilection has been found to be elevated in regions in which flow simulations indicate that the local hemodynamic environments are complex. In the human carotid bifurcation, intimal thickening and plaque localization is greatest near the midpoint of the sinus on the nonflow divider wall, where flow separation, vortices, low shear stress, and directional oscillations are all observed in model geometries. In the human aorta, plaques are found primarily at the proximal inlets of ostia, whereas at the aortic bifurcation, intimal thickening tends to be higher on the lateral wall of the daughter arteries, a trait common to the major coronary arteries as well.
These susceptible regions are generally characterized by a relatively low wall shear stress. Additionally, the spatial shear stress gradient is elevated in these regions of complex flow, and this hemodynamic stress has also been implicated in the disease process. In the rabbit aorta, patterns of intimal monocyte deposition correlate positively with calculated wall shear stress gradient (2), although it is unclear whether the gradient modulates monocyte transport or endothelial function. We have recently demonstrated that a similar gradient parameter, maximum gradient stress, interacts with shear stress magnitude to predict transendothelial macromolecular permeability in the porcine iliac artery (9); the effects of high gradients are strongest where the shear stress magnitude is low.
Several in vitro studies have examined the response of endothelial cells to large spatial gradients in shear stress using flow chambers featuring a backward-facing step, behind which recirculation eddies form. Large spatial shear gradients occur near the stagnation point. Among the observations made in this type of system are increased rates of cellular division and cell loss very close to the stagnation point (4, 19), migration of cells away from the stagnation point into regions of high cell density (4, 19, 20), enhanced transendothelial permeability in the recirculation zone (15), and increased activation of transcription factors in the recirculation zone (12). However, it is difficult to attribute these responses to large spatial shear stress gradients per se, because of the presence of several important confounding fluid dynamic factors, including low shear stress magnitude, a stationary stagnation point in the vicinity, and elevated normal forces acting on the endothelium. McKinney et al. (11) have recently shown that increased protein expression of intercellular adhesion molecule-1 (ICAM-1) in human umbilical vein endothelial cells (HUVEC) behind a backward-facing step correlated with normal force and shear stress magnitude but not with gradient.
The goals of this study were to design and implement a new flow system with which to study the individual and combined effects of shear stress magnitude and spatial gradient on endothelial cells in vitro. Converging-width flow channels were used to expose cells to physiological ranges of both shear magnitude and gradient. Endothelial cells sampled from regions experiencing known shear environments were analyzed for expression levels of several genes implicated in early disease, and the effects of gradient and magnitude on these expression levels were characterized.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
|
1 mm2 in size, were identified within each channel that would experience shear stress magnitudes and gradients defined by the boxes in Fig. l. An additional gasket, also shown in Fig. 2 (bottom right), with the same overall dimensions as those described above, was cut such that the width of the flow channel dropped from 50 to 12.5 mm halfway along its length. CFD calculations confirmed that, by using two different flow rates, all four of the nominal shear stress magnitudes listed above could be produced with this design, gradient free, if regions not near the sudden convergence were selected. There were six different flow experiments, each defined by a certain gasket and a certain flow rate. Among these six unique flow experiments, 25 different sampling regions of known shear stress magnitude and gradient were defined. Figure 3 shows the locations of three sampling regions in one of the six flow experiments.
|
Flow loop and cells. The flow chamber was incorporated into a closed, sterile flow loop in which steady flow was achieved using a constant hydrostatic pressure head. The desired flow rate was maintained by a control system in which flow rate was continuously measured using a transit-time ultrasound flow probe (Transonic, Ithaca, NY), and a pinch-valve was adjusted as necessary using a software-driven motor.
The cells used were porcine aortic endothelial cells (PAOEC, Cell Applications, San Diego, CA) between passages two and eight. Cells were maintained in Medium 199 supplemented with 5% FBS, 2% porcine serum, and 1% antibiotic/antimycotic solution. The night before the flow experiment was initiated, this medium was replaced with low serum medium (the same as that above but lacking FBS), which was also circulated in the flow loops. This medium exchange was made on the first night in which no gaps larger than the size of a cell were observed in the monolayer. Cells were grown on glass slides (Electron Microscopy Sciences, Ft. Washington, PA) that fit the recess in the flow chamber base. The slides were coated with a cell-derived matrix that was produced by previously growing a monolayer of PAOEC on the slides, permeabilizing them with 1% Triton, and washing them thoroughly. Cells were exposed to flow for a period of 24 h.
Real-time quantitative PCR and data analysis.
After the flow period, the flow chambers were disassembled, and the slides were washed with PBS. Slides were placed on a mechanical stage attached to an inverted microscope with phase contrast optics. After a coordinate system was defined for the slide, the stage was used to position a narrow beam of visible light at each location from which cells were to be recovered, as determined by the CFD analysis. A sterile 21-gauge syringe needle was used to scrape cells at the site of the focused light spot, while the monolayer was viewed under the microscope using a x10 objective. A region 1 mm in diameter was scraped at each location. The needle was immediately immersed in 25 µl of lysis buffer on ice. cDNA was immediately synthesized from the lysate of each sample using the Cells-to-cDNA II Kit (Ambion, Austin, TX).
Expression levels of five genes, endothelial nitric oxide synthase (eNOS), c-jun, monocyte chemoattractant protein-1 (MCP-1), ICAM-1, and vascular cell adhesion molecule-1 (VCAM-1), were measured relative to 18S ribosomal RNA by real-time quantitative PCR (qPCR). The MyiQ Single-Color Real-Time PCR Detection System (Bio-Rad Labs, Hercules, CA) was implemented to detect PCR product using SYBR Green (Molecular Probes, Eugene, OR) with a two-step cycle. For each gene, custom primers were designed, based on published porcine cDNA sequences and synthesized by Integrated DNA Technologies (Coralville, IA). Primer sequences, product length, and experimentally determined melting temperatures of the PCR product for each gene are given in Table 1. To verify that measurements of relative transcript levels did not depend on the number of cells collected, serial dilutions of cell lysate, spanning the concentration range of this study, were reverse transcribed and subjected to qPCR. For each gene, logarithmic regression demonstrated that the difference in threshold cycles between 18S and the gene did not depend on dilution. Additionally, melt curve analysis was performed on the PCR product of every reaction in the study to verify that only one product was formed and negative reverse transcriptase controls confirmed that genomic DNA was not amplified.
|
CT method (10). The fold difference in relative gene expression for a sample, Yij, was analyzed as a function of shear stress magnitude and gradient using the following general linear regression model:
 | (1) |
0 is a constant, 1 and 2 are the main effects coefficients for the shear stress magnitude and gradient, respectively, 3 is the coefficient of the interaction term for these two variables, and 
ij is the random error of the sample. ANOVA was used to test the significance of the overall model, and multivariate regression was used to test the significance of each of the individual parameters in the model. For each gene, a total of 100 data points were included, four replicates of each of the 25 sampling regions. Adequacy of the multivariate model was established for each gene by confirming the normality and uniform variance of the fold differences. Additionally, residuals were plotted against each of the three regressors (magnitude, gradient, and their product) to confirm that no systematic deviations from linearity were apparent. At select constant levels of either shear stress magnitude or gradient, gene expression data were plotted against the other variable to confirm that the observed trends were independent of the particular flow experiment in which the data were collected. |
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
1, 2, and 3 terms, each of the five genes exhibits a different dependence on the shear variables: eNOS depends on shear magnitude only, c-jun depends on shear magnitude and gradient, MCP-1 depends on gradient and the interaction term, ICAM-1 depends on magnitude and the interaction term, and VCAM-1 shows no dependence on any of the terms.
|
|
The significance of the dependence of c-jun gene expression on shear stress magnitude and gradient is only marginal (P = 0.084 and P = 0.098, respectively), and, as shown in Fig. 4B, the sensitivity of gene expression to these variables is also weak. Expression of this gene increases slightly with shear stress magnitude and decreases slightly with increasing gradient.
The dependence of MCP-1 gene expression on the two variables is quite complex. As a main effect, shear stress magnitude does not predict expression level. However, there is a significant negative correlation between expression and gradient (P = 0.003), and the interaction term is significant (P = 0.001). As illustrated in Fig. 4C, the slope of MCP-1 gene expression vs. gradient changes sign as shear magnitude increases. Based on the equation describing the response surface, this occurs at a magnitude of 28 dyn/cm2. The interaction term also causes the dependence on magnitude, which is not significant in the absence of a gradient (P = 0.24), to become significant at larger gradients (e.g., P = 0.07 at 24 dyn/cm3).
ICAM-1 gene expression (Fig. 4D) is also influenced by an interaction between shear stress magnitude and gradient (P = 0.042). However, in contrast to MCP-1, there is a significant positive correlation between expression and shear stress magnitude (P < 0.001), whereas gradient has no main effect. There is a strong positive relation between transcript level and shear stress magnitude in the absence of a gradient (P < 0.001), which diminishes as gradient increases. An inverse relationship with gradient is strongest at the highest level of shear stress magnitude (P < 0.001 at 48 dyn/cm2).
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
To determine these sensitivities, a regression model that assumed simple linear relationships between the hemodynamic variables and responses was used. Although several cellular responses require a minimum threshold shear stress (5, 13), these thresholds are generally near or below the minimum shear stress of 6 dyn/cm2 used in this study. Additionally, no deviations from linearity were observed in plots of the residuals vs. the independent variables, indicating that the linear regression model was adequate.
We have previously found that the transendothelial permeability of the proximal porcine iliac artery decreases with increasing values of the parameter WSS/MGS0.41 over the range of 2–10 dyn0.59·cm–0.77 (9). As demonstrated in Fig. 1, the combinations of shear stress magnitude and gradient explored in this study span the normal ranges of these variables and that of the parameter WSS/MGS0.41. The latter parameter is minimized when shear stress magnitude is low and gradient is high. To the extent to which this situation could be analyzed in this study (moderate to high gradients were unattainable for shear stresses <6 dyn/cm2), it was generally associated with low levels of eNOS, ICAM-1, and MCP-1 transcripts (Fig. 4). As noted above, VCAM-1 expression was insensitive to both shear stress magnitude and gradient; thus the low shear-high gradient environment would not appear to favor monocyte recruitment. The reduced eNOS expression in this mileau may be playing a larger role and could be responsible for the elevated permeability seen earlier.
For the limited number of genes examined, we have demonstrated an atherogenic response, reduced eNOS expression, to low shear stress magnitude irrespective of shear gradient but no clear response to gradient alone. Naturally, there could be other genes or other cellular responses not explored in this study that are directly responsive to high gradients, but our results to date suggest that gradient exerts its negative effects mainly by modulating the response to shear stress magnitude. This latter concept is supported by our finding that gradient alone is a poor predictor of macromolecular permability in the porcine iliac artery, while exacerbating the response to low shear stress magnitude (9).
Interestingly, MCP-1 gene expression was greatest where the shear stress magnitude and gradient were both high, whereas ICAM-1 expression peaked when shear stress magnitude was high but the gradient was low. The opposite effects of gradient on the expression levels of these two molecules at high shear stress prohibits their simultaneous expression when shear is elevated. This could contribute to the well-documented (6) lower risk of atherosclerosis where shear stress magnitude is high, as could the gradient-independent shear enhancement of eNOS expression noted above.
For some genes, the short-term dose dependence of transcription on shear stress magnitude has been measured in vitro. The results presented here, although obtained after longer exposures, agree with several of these studies. The expression of eNOS mRNA has been shown to increase with shear level up to 25 dyn/cm2 in bovine aortic endothelial cells subjected to flow for 3 and 6 h (23, 26); we found similar results after 24 h in the absence of a gradient. Our finding that the MCP-1 transcript level is independent of shear magnitude in the absence of a gradient is consistent with previous work using HUVEC (17) in which similar levels of MCP-1 transcript were seen after 1.5-h exposure to shear stresses between 6 and 32 dyn/cm2. The magnitude dependence of ICAM-1 mRNA expression in the absence of a gradient, which we report here, is consistent with measurements of protein in HUVEC exposed to shear stresses of up to 20 dyn/cm2 for 4 h (21). McKinney et al. (11) recently correlated protein levels of ICAM-1 with various hemodynamic parameters using shear stress-preconditioned HUVEC exposed to flow for 8 h in a backward-facing step model. Consistent with the ICAM-1 gene levels seen in the present study, protein expression was positively correlated with shear stress magnitude and uncorrelated with gradient. Interestingly, the strongest predictor of ICAM-1 protein level was the normal stress associated with the reattachment point. This important variable may have had a significant influence on other responses that have been observed using the backward-facing step model (3, 4, 11, 12, 15, 19, 20). The converging-width flow chambers used in the current study benefit from the absence of the confounding effects of flow stagnation and reattachment.
The influence of combined physiological levels of shear stress magnitude and shear stress spatial gradient on endothelial biology has not hitherto been examined. The results of this study indicate that the expression level of each of the five genes studied depends differently on magnitude and gradient. As each gene has a unique set of promoter response elements and transcription factors with which they are known to interact, it is not clear where along the mechanotransduction and intracellular signaling pathways the responses of the various genes diverge. The sensitivity of some genes to one of the stimuli, but not the other, suggests that cells are capable of sensing and transducing shear stress magnitude and spatial gradient independently. The presence of an interaction term in some of the responses suggests that there may be feedback between these transduction pathways.
Notably, of the genes considered in this study, VCAM-1 is the only one whose promoter lacks the shear stress response element (SSRE). As it was the only gene that was unresponsive to both shear stress magnitude and gradient, the role of this regulatory element in the responses of the other genes merits exploration. Although the role of SSRE in the early regulation of gene expression in response to shear stress onset has been questioned (16), the role of transcriptional modulators, including SSRE, in the dependence of expression on shear magnitude has, to our knowledge, not been addressed.
There are a number of ways in which shear stress gradients could affect the biology and physiology of the arterial wall. Explanations of endothelial cell sensitivity to macroscopic gradients of shear stress based on direct interaction with the cell membrane or membrane-bound receptors are difficult to reconcile with calculations, based on realistic endothelial cell surface contours, that predict that subcellular spatial shear stress gradients arising from the nuclear bulge (1) are one to two orders of magnitude higher than the macroscopic gradients we calculated in the porcine iliac artery (9) and implemented in this study. The endothelial glycocalyx, which has emerged as a candidate mechanosensor (18), may diminish the shear gradient over the nuclear bulge. The thickness of the in vivo arterial endothelial glycocalyx has been reported to be as large as 3 µm (24), exceeding the thickness of the endothelium and its surface undulations arising from the nuclear bulge. The layer over the bulge may flatten under flow, reducing the local shear stress gradients. Furthermore, it has recently been reported that, in response to flow, heparan sulfate proteoglycans of the glycocalyx redistribute from the nuclear region to the neighborhood of the intercellular junctions, further flattening the surface contour of endothelial cells (27). Thus there is some uncertainty regarding the relative magnitude of the shear gradient arising from the macroscopic flow field and that is caused by cell surface topography. Even so, the factors mentioned above would have to reduce the topographical contribution by upwards of 90% for the macroscopic contribution to be controlling at the local level. This would continue to be the case even if, as Weinbaum et al. (25) have proposed, the glycocalyx transmits fluid shear stresses directly to the cortical cytoskeleton.
Macroscopic shear gradients could also act by creating differences in the shear forces that are experienced by adjacent cells. When a 50 dyn/cm3 gradient acts along the long dimension of an array of 10 x 30 µm endothelial cells, two successive cells experience a difference of 5 pN of shear force. This force differential can induce a gradient in cellular shear strain in the direction of the hemodynamic gradient, or it can act directly at the intercellular junctions. The shear strain gradient is easily calculated under the assumption that each cell deforms independently, and the calculated increase in intercellular gap distance is extremely small, of the order of 0.05 nm. Therefore, the macroscopic gradient certainly does not enhance transendothelial transport, as we found in vivo (9), through its effect on the available transport area.
The possibility remains that the macroscopic gradient acts through the force it induces at the junctions between the cells. The intercellular force estimated above is in the range capable of inducing changes in protein conformation and might be transduced by junctional mechanosensors or signaling molecules such as PECAM-1 (22) or connexin 43 (Cx43), which has been shown to be sensitive to shear stress gradients (3). Cx43 disruption by shear has been associated with increased activation of a variety of transcriptional factors such as NF-
B, Egr-1, c-jun, and c-fos (12). Other shear sensitive junctional proteins, such as claudin, occludin, or VE-cadherin, may be responsive to tensile mechanical force. Even so, we must note that the 5 pN of force calculated above is distributed along the entire 10 µm transverse length of the model cell.
We have demonstrated that, at the transcriptional level, endothelial cells are sensitive to both shear stress magnitude and gradient. Furthermore, the interactions of these stimuli seen in some of the responses suggest that some mechanotransduction pathways might be modulated by others. Interestingly, each of the five genes tested responded differently to the two shear-related variables, which might be expected to lead to considerable heterogeneity of expression in a real flow field. Both atheroprotective and atherogenic genes are responsive to the shear environment. Similarly, Passerini et al. (14) have found that a complex collection of mixed messages in the endothelial response to complex flow fields in vivo. Further studies to characterize the combined effects of shear stress magnitude and gradient may allow us to identify combinations of these variables that are particularly predisposing to arterial disease.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
CT method. Methods 25: 402–408, 2001.[CrossRef][Web of Science][Medline]This article has been cited by other articles:
|
|
 |
|
  A. J. Engler, M. Chan, D. Boettiger, and J. E. Schwarzbauer A novel mode of cell detachment from fibrillar fibronectin matrix under shear J. Cell Sci., May 15, 2009; 122(10): 1647 - 1653. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
2.2 cm); 12 dyn/cm2 and 12 dyn/cm3 for proximal band in convergence zone (x 