Hemodynamic alterations as well as pro-inflammatory and exogenous stimuli can adversely affect vascular integrity and lead to the pathogenesis and/or progression of a number of major neurological disorders
[5, 29–31]. In vitro studies are set to support and facilitate our understanding of the mechanisms involved in the physiological and pathological modulation of cerebrovascular functions. This is of critical importance for the development of novel strategies aimed at reducing the burden of CNS disease associated with brain vascular impairments. Other in vitro systems have attempted to reproduce the physiological and functional characteristics of capillary or venular segments of the cerebrovascular system (e.g.,
[32, 33]). However, these models lack the ability to reproduce the environmental cues (e.g., hemodynamic stimuli) to which these vascular beds are exposed in vivo.
The use of hollow fiber technology allowed us and others to establish the first quasi-physiological in vitro BBB models
[34–39] which was then further humanized
 to reproduce not only healthy BBB properties but also properties of multiple drug resistance and leukocyte extravasation
[2, 24, 41]. This technology allows to develop the first artificial interlinked brain capillaries and venules segments, which retain their distinct vascular properties. As shown in Figure
1 the pumping mechanism generates arterial high velocity flow with a systolic blood pressure range (80 to 300 mmHg) comparable to in vivo. A dramatic drop of perfusion velocity and transmural pressure occurs when flow enters the capillary segment generating a vascular shear stress comparable to what has been reported for non-BBB vessels in vivo. After leaving the capillary module, the medium flow enters into the venule module were perivascular astrocytes were replaced by smooth muscle cells to more closely mimic the in situ cellular milieu of this vascular segment. At this level systolic transmural pressure is low (see Figure
2) and flow is characterized by low shear stress (≅ 3 dynes/cm2; see also Table
1). It is important to underscore again that direct measurements of shear levels in brain microvessels is lacking and that the values used in this manuscript are extrapolations from other vascular beds in vivo.
The DIV capillary-venule model not only mimics the rheological characteristics of the corresponding brain vascular segments but also their functional and physiological properties (e.g., Table
1). In agreement with previously published studies BBB capillaries In vivo are characterized by high TEER, lack of paracellular pathways, low permeability to polar molecules (e.g., paracellular markers) and a selective permeability that (in the absence of specific extrusion mechanisms) reflects the lipophilicity of the specific substance
. Our results have clearly shown that, similar to in vivo, the artificial capillary vascular bed was characterized by a high TEER (see Figure
3D); low permeability to the paracellular marker sucrose, and was capable of selective permeability reproducing the in vivo rank order of the tested substances (see Figure
. Overall, the relationship between lipophilicity and permeability found in the DIV-capillary module was similar to that reported by others in vivo.
Venules are characterized by a significantly more permissive vascular bed with a lower TEER and a reduced ability to provide a barrier to the passage of polar molecules than capillaries
. In this respect, the DIV-venules showed lower TEER, and a reduced selective permeability (see Figure
3D). Our findings also suggest that abluminal astrocytes and high shear stress are both necessary to establish a tight vascular bed
[4, 6, 10, 37, 38]. In fact, when the capillary modules were established under venular level of shear stress and vice versa, (venule modules established under capillary-like shear stress levels; see Figure
3B) no TEER increase suggesting the formation of stringent vascular bed was observed.
BBB opening following exposure to hyperosmotic mannitol is common clinical procedure used to enhance chemotherapeutic drug penetration into the CNS to treat patients with metastatic or primary brain tumors
. Hyperosmotic opening of the BBB is mediated by vasodilatation and shrinkage of cerebrovascular endothelial cells (and perhaps glia), with widening of the inter-endothelial tight junctions to an estimated radius of 200 Ǻ. This provides paracellular pathways previously lacking that can facilitate the passage of substances across the brain capillary endothelium
[21, 27, 28, 47]. Our results (see Figure
4) demonstrated that loss of vascular integrity of in the DIV capillary-venules model mimics the expected physiological vascular response to hyperosmotic agents (magnitude and duration) of the corresponding vascular segments in vivo. The tightness of brain capillary vascular beds is the most dependent upon the tight junctions’ (TJ) ability to seal the space between adjacent endothelial cells. The presence of tight junctions between endothelial cells in our capillary model was demonstrated by electron microscopy
 and confirmed by functional assays that measured for example permeability to K ions
 The mechanisms of osmotic blood–brain barrier disruption is believed to depend on endothelial cell shrinkage. This may be caused by the exposure to a hyperosmolar environment, efflux of water from the endothelial cell and subsequent cytoskeletal rearrangement. The latter may cause stretching of TJ and widening the interendothelial space. The temporary formation of paracellular routes of entry across the BBB as demonstrated by the monophasic TEER decrease.
The effect of mannitol was strikingly different in the venule compared to capillary modules. Recovery was delayed and overall disruption achieved was less notable at the venular level; the latter may be due to its more prominent paracellular pathway. There is no clear-cut explanation for the delayed recovery of venular TEER after osmotic challenge. However, we would like to underscore that when we performed experiments in vivo (rat, pig; intracarotid injections of mannitol) we found that at 30’-1 hr. intervals white matter venules retained blood–brain barrier disruption properties (measured by different tracer means) while gray matter small vessels (capillaries) were, at this time point, “intact”. The fact that we can reproduce this finding in vitro suggests that the persistence of disruption in larger vessels, and venules as seen in white matter, is due to intrinsic properties rather than vascular access issues, or parenchymal influences. It is also critical to address the issue of molecular mechanisms of osmotic disruption and recovery of TEER. For example, are TJ protein involved in a transcriptional vs. positional way? Are other mechanisms such as transcellular access relevant? We believe that our findings require additional studies, and that only a side-by-side comparison of our in vitro/in vivo will elucidate this. Finally, it is striking that a clinically relevant procedure such as osmotic BBBD has never been mechanistically explored to show what truly are the mechanisms underlying increased permeability after intrarterial mannitol.
Oxygen delivery to the CNS is another important function bestowed upon the brain vessels. Recent studies have shown that PO2 values increase from the post-capillary venules to the distal vessels of the cerebrovascular network and by contrast, measurements of the radial gradients are consistent with an increase oxygen loss
. These data support the hypothesis that venules may indeed play a critical role in oxygen delivery. These findings are consistent with our data related to the metabolic behavior of capillary and venule segments. In fact, the highly oxygenated blood from the arterial circulation reaching the brain microcapillaries allows the BBB endothelium to make use of the highly efficient aerobic-driven metabolic respiration to meet the cellular bioenergetic demand and to use less glucose in the process thus maximizing that delivered into the brain (see Figure
5A). On the other hand, if oxygen delivery to the brain primarily occurs at the venular level as suggested by recent studies
[48, 49] than a non-oxidative (anaerobic) metabolic behavior (see Figure
5C) would minimize vascular oxygen consumption leaving more oxygen available for brain delivery. Furthermore, recent studies have clearly shown that shear stress plays a key role in the modulation of bioenergetic metabolism of vascular endothelial cells favoring the utilization of the more energy rewarding aerobic pathway. This perhaps also suggests an additional link between shear stress/flow and vascular functions of the different cerebrovascular territories. However, additional and more specific studies will be necessary to validate this hypothesis and to understand the underlying mechanisms.
A limitation of our study concerns the lack of pericytes to the abluminal mixture of cells. Pericytes have been shown to control several aspects of BBB function in vivo[13, 50] including “barriergenesis” (via release of angiopoietin-1
) and protection against hypoxia-induced BBB disruption
. Although our preliminary experiments (data not shown) revealed no significant effects of pericytes on BBB tightness, these findings needs to be carefully reproduced under different combinatory approaches of cells (astrocytes/pericytes in different ratios). In addition, functional BBB modulatory effects specific for the endothelium or abluminal astrocytes (e.g., cell polarization) are still poorly understood and need to be further investigated. Another limitation of our study is that we did not attempt to isolate the differentiating effects of shear vs. abluminal cell type. Thus, how astrocytes or vascular smooth muscle influence endothelial cell differentiation under conditions of equal shear stress need to be further investigated. In particular.