We first studied the effects of pericytes on the permeability of brain endothelial cell monolayers. The barrier function of brain endothelial cells occurs through a low level of transcytotic mechanisms (macropinocytosis and fenestrae), and an increase in paracellular mechanisms (intercellular tight junctions) . To some degree, the mechanisms that control and alter transcytotic events are different from those controlling paracellular permeability [62, 63]. Electrical resistance measured by TEER measures channel connections typically ascribed to paracellular permeation. Here, TEER was very low, consistent with little effect on paracellular channel formation. Albumin permeation, however, was also low, indicating that transcytotic mechanisms were likely to be low. Mature fenestrae would produce both a low TEER and high albumin permeation, so it is likely that fenestrae were also not present in large numbers in these cells. Co-culturing pericytes with the brain endothelial cells significantly improved TEER, suggesting improved tight junction function, but did not alter albumin permeation (Table 1). BMEC-pericyte co-cultures had significantly reduced permeation of I-HIV. Although the HIV-1 virus is much larger than the albumin molecule, the rate at which I-HIV crossed the brain endothelial cell monolayer was much faster than that of I-Alb regardless of whether the endothelial cells were co-cultured with pericytes. This is consistent with the vesicular process for HIV-1 transport being independent of nonspecific vesicular mechanisms such as macropinocytosis.
LPS produced an increase in I-HIV permeation in the monolayers co-cultured and not co-cultured with pericytes. LPS also decreased TEER (that is, it increased paracellular permeability); therefore, this increase in I-HIV could be mediated by transcytotic or paracellular mechanisms. However, the presence of pericytes magnified the LPS effect on HIV-1 permeability without affecting TEER. This suggests that the pericyte-dependent portion of LPS-enhanced I-HIV permeation is transcytotic rather than paracellular.
The above results show that pericytes facilitate the LPS-enhanced transport of HIV-1 across brain endothelial cells. We then investigated some of the mechanisms by which pericytes could mediate such facilitation. LPS crosses the in vivo BBB, even the disrupted BBB, poorly if at all . In vitro, LPS produces different patterns of cytokine release when added to the luminal versus the abluminal chamber, consistent with an inability to cross the in vitro BBB as well [37, 65]. We, therefore, postulated that LPS was likely acting at the luminal surface of the brain endothelial cell, rather than at the pericytes in the abluminal chamber, as the first step in a neuroimmune-based modulation of the crosstalk between pericytes and brain endothelial cells. More specifically, we postulated that LPS acts at the luminal surface of the brain endothelial cell to induce release of soluble factors from its abluminal surface. These soluble factors would then act on pericytes, inducing them to release soluble factors that would modulate brain endothelial cell transcytosis of I-HIV. To test this hypothesis, we examined the release of cytokines from BMECs and pericytes. It is known that both endothelial cells and pericytes release cytokines, that cytokines are important in communication between the cells of the neurovascular unit, that cytokines enhance HIV-1 transcytosis, and that LPS can act on one side of the BBB to affect the release of cytokines from the other side [37, 44, 66, 67].
We first exposed the luminal surface of BMEC monolayers to a short, 4 h exposure to LPS, then collected the abluminal culture media from these monolayers and abluminal culture media from BMEC monolayers not exposed to LPS was used as a control. We then assessed the ability of the culture media to release cytokines from pericytes. We found that culture media from LPS-exposed brain endothelial cells did indeed increase the release of two cytokines, KC and MCP-1, from pericytes. The level of these cytokines was much higher in the pericyte culture medias (40–140 pg/ml; see Figure 4) than in the brain endothelial cell culture medias (see Table 2), thus ruling out contamination as a source of these results. Likewise, it is unlikely that LPS from the endothelial cell culture media caused this stimulation as the LPS was added to the luminal chamber of the endothelial cells, whereas the media exposed to the pericytes was from the abluminal chamber. Furthermore, LPS contamination would have been expected to produce the cytokine profile observed in Figure 3 and not just increases in KC and MCP-1. Therefore, the most parsimonious explanation of our results fits our hypothetical model: LPS acting at the luminal surface of brain endothelial cells stimulates release of soluble factors from their abluminal surface which then modulates pericyte release of immune-active factors.
The levels of release of KC and MCP-1 are substantial. Given that the pericyte cultures contained about 50 μg of protein, we estimate MCP-1 and KC concentrations of 960 and 1440 pg/mg protein. This exceeds the levels produced in brain after in vivo administration of LPS .
Although we found the release of two cytokines from pericytes was specifically affected, it is likely that the release of many other immune active substances is similarly modulated. However, pericyte-secreted MCP-1 could be directly involved in the enhanced transcytosis of HIV-1. MCP-1 derived from astrocytes  and microglia  mediates cross talk with brain endothelial cells that increases the diapedesis across the BBB of HIV-1 infected macrophages and monocytes. MCP-1 also affects voltage-gated potassium channels in brain endothelial cells . Here, the evidence shows that pericytes are also a source of MCP-1 which is released in response to signals from brain endothelial cells. It could be that MCP-1 released from pericytes alters brain endothelial transcytotic processes that affect the permeation of free HIV-1.