Shear stress: devil's in the details

In this issue of Blood, Ni et al use an in vivo mouse model of disturbed flow that results in accelerated atherosclerosis to identify novel mechanosensitive genes.¹ 

Dating back to the turn of the previous century, pathologic and experimental observations led to the appreciation that the earliest lesions of atherosclerosis occur in a nonuniform manner throughout the arterial tree. This nonuniform distribution is believed to arise as a consequence of the endothelial dysfunction that occurs at areas of altered blood flow dynamics where arteries branch or turn sharply. Efforts to understand how mechanical forces can affect endothelial biology were catapulted forward in the early 1970s by the successful isolation and culture of endothelial cells (ECs) by Jaffe and colleagues.²  The ability to easily propagate and manipulate ECs ushered in the modern era of experimental vascular biology and catalyzed an enthusiastic and productive period of vascular study carried out largely in vitro.

The study of how alterations in shear stress affect the subsequent development of atherosclerosis has benefited from the use of cultured EC in devices created to mimic the conditions of laminar or disturbed flow. The first well-characterized shear stress apparatus was a cone-and-plate configuration introduced by Dewey and colleagues in 1981.³  Since then, various iterations have been developed that incorporate alterations in waveforms (pulsatile, oscillatory) and strain, and allow for the coculture of different vascular cell types. Using these devices, important insights have been gleaned regarding the endothelial cell's response to flowing blood.

Ni et al remind us that despite the use of increasingly sophisticated in vitro model systems, caution must still be used when estimating the accuracy to which such systems recapitulate physiology. As reported in a previous paper, Ni et al have adapted a partial ligation model of the carotid artery (introduced by Korshunov and Berk to study vascular remodeling⁴ ) to create a low-flow, oscillatory waveform that is seen in association with in vivo atherogenesis.⁵  Partial ligation is constituted by ligation of 3 branches off the distal left common carotid—the external carotid, internal carotid, and the occipital arteries—while leaving the superior thyroid artery patent. In ApoE knockout mice, endothelial dysfunction is apparent within 1 week, and atherosclerotic lesions develop within 2 weeks of surgery. In that report, the Jo laboratory presents microarray data derived from EC harvested from the ligated (left) and intact (right) carotid arteries at 12 and 48 hours after surgery. The strengths of this approach include the use of an in vivo model that creates disturbed flow waveforms very similar to those early in regions of “natural” atherosclerosis, an early timeframe for collection of the ECs, a simple, rapid, and specific method to harvest EC from a limited vascular region, and a species readily amenable to genetic manipulation.

Consistent with several previous publications, the current Ni study shows that KLF2, KLF4, eNOS, VCAM1, and BMP4 are regulated by altered shear stress; however, comparison of their in vivo data with microarray analyses (from their laboratory and others) derived from human umbilical vein EC (HUVEC) exposed to “atheroprone” flow⁶  also revealed striking differences. Perhaps the most important cautionary moral to this story was the observation that only 50% of the mechanosensitive genes identified in Ni et al's mouse model have been previously reported using in vitro HUVEC shear stress experiments. Furthermore, several genes identified in the current study had not been identified as flow-responsive in any previous publication. It is not unusual to find discrepancies in the microarray results of flow experiments from different laboratories; after all, different in vitro systems and waveforms are often used, cells are exposed to different culture conditions and derived from different species or from different vascular beds. A fascinating cause of endothelial functional heterogeneity is the anatomical origin of the cells (vascular bed, vessel type and size).⁷  There is also evidence that prolonged culture of EC can result in gene-specific loss of response to flow. Should these variables result in only a 50% identity of the genes shown to be flow-regulated? How do we determine experimental artifact from true positive/negative changes? Even the extracellular matrix can affect response to flow. The extracellular matrix below EC is normally composed primarily of collagen IV and laminin, but fibronectin is deposited in proatherogenic parts of vessels and could potentiate atherosclerosis.⁸  This is particularly noteworthy as fibronectin-coated dishes are commonly used in in vitro shear stress experiments.

Ni and colleagues also compared their results to previous in vivo efforts in the porcine model where gene expression in linear and curved segments of the pig aorta were compared. Surprisingly, of the 42 genes identified to be mechanosensitive in this study, only 2 (KLF4 and eNOS) were identified in the porcine tissue. This may reflect true differences in flow response between species. However, the dearth of overlap between the Jo laboratory result and the Davies' laboratory porcine results is likely to be the differential effect of acute exposure to disturbed flow (current Ni study) versus chronic exposure. Nevertheless, the identification of KLF4 and its target, eNOS,⁹  does speak to the importance of these 2 factors in the context of hemodynamics and atherogenesis.

Finally, we note that the differences in the mechanical cause of disturbed flow may make it somewhat difficult to find identity of results even between in vivo experiments performed in the same vascular bed of the same species. The Jo laboratory's model is substantially different in technique compared with a recent model that uses a perivascular (carotid) cuff to create altered flow. The perivascular cuff model creates low flow proximal to the cuff and oscillatory shear flow distal, but does not create a coincident low-flow/oscillatory waveform region.¹⁰  As nicely exemplified by the current study, the details of flow experiments matter and a return to studies conducted in vivo is likely requisite to answer one of the enduring questions facing vascular biologists: what are the core factors responsible for mechanotransduction and, thus, vascular homeostasis?