Perivascular macrophages mediate the neurovascular and cognitive dysfunction associated with hypertension (original) (raw)
Characterization of PVMs. First, we used well-established criteria to identify brain PVMs (18). Brain PVMs were identified as myeloid cells closely associated with the abluminal side of cerebral blood vessels and (a) expressing the mannose receptor CD206 (Figure 1, B–D, and ref. 18); (b) expressing lymphatic vessel endothelial hyaluronan receptor-1 (LYVE1) (ref. 25 and Figure 1D); (c) expressing low levels of IBA1, compared with microglial cells (Figure 1C); and (d) endowed of phagocytic activity assessed by injection of dextran into the lateral ventricles (Figure 1, B–D, and ref. 26). PVMs were found most abundantly in association with pial and parenchymal vessels larger than 20 μm in diameter, particularly pial arterioles, a key site for flow regulation (7), the length of which was covered 70%–80% by PVMs (Supplemental Figure 1, A–E; supplemental material available online with this article; doi:10.1172/JCI86950DS1).
PVM identification. (A) Electron micrograph of a neocortical arteriole (diameter 10 μm). A PVM containing lysosomes (electron dense) is seen at the top right of the vessel (scale bar: 1 μm). PVMs can be identified as perivascular cells (A–D), positive for CD206 (B–D), able to phagocytose i.c.v.-injected FITC-dextran (B–D), weakly positive for the microglia marker IBA1 (C), and expressing the endothelial lymphatic vessel marker LYVE1 (D). Scale bars: 25 μm.
Slow pressor ANGII disrupts the BBB, leading to ANGII entry into the perivascular space. To investigate the role of PVMs in the cerebrovascular dysfunction induced by ANGII hypertension, we first sought to determine whether circulating ANGII crosses the BBB and reaches the perivascular space. Infusion of low concentrations of ANGII (600 ng/kg/min) induced a delayed increase in BP that was observed at day 7 and leveled off by day 14 (Figure 2A). At this time, ANGII concentration was increased in plasma (Figure 2B), but not in brain (Supplemental Figure 2, A–C). To determine whether plasma ANGII (MW ~1 kDa) is able to cross the BBB and gain access to the perivascular space, we examined the permeability of the BBB to sodium fluorescein (MW 0.3 kDa) and FITC-dextran (MW 3 kDa). We found that chronic ANGII infusion increased the permeability of the BBB to both markers at 14 days, whereas acute administration of ANGII (250 ng/kg/min) produced a comparable elevation in mean BP, but failed to do so (Figure 2C and Supplemental Figure 2E). The increased permeability to FITC-dextran was observed prior to 14 days, but the effect did not reach statistical significance (Supplemental Figure 2F). In contrast to FITC-dextran and sodium fluorescein, the BBB permeability to Evans blue, which binds to serum albumin (MW ~69 kDa), was not altered by chronic ANGII administration (Supplemental Figure 2E), indicating a previously unappreciated size restriction in the BBB dysfunction. The effect of ANGII on the BBB could not be attributed solely to the elevation in BP, since 2 weeks of phenylephrine infusion increased BP comparably to ANGII, but did not alter BBB permeability to FITC-dextran (Supplemental Figure 2, G and H).
Slow pressor ANGII disrupts the BBB, leading to ANGII entry into the perivascular space and PVMs. (A) Slow pressor ANGII gradually increases systolic blood pressure measured by tail-cuff plethysmography. *P < 0.05 vs. vehicle (Veh); n = 12–17 (2-way ANOVA and Bonferroni’s test). (B) ANGII plasma levels are increased after 2 weeks of ANGII administration. *P < 0.05 vs. Veh; n = 8–9 (Student’s t test). (C) Slow pressor hypertension but not acute i.v. ANGII administration increases BBB permeability to FITC-dextran (MW 3 kDa). *P < 0.05 vs. Veh; n = 5–7 (Student’s t test). (D) Confocal microscopy showing immunofluorescence labeling of biotinylated ANGII around cerebral blood vessels and in association with PVMs in mice treated with ANGII for 14 days but not in saline-treated mice (scale bar: 25 μm). (E) Orthogonal views (XY, XZ, YZ) illustrating colocalization of biotin-ANGII (green) and CD206 (blue). (F) Biotin-ANGII is associated with PVMs in mice treated with ANGII for 14 days but not in saline-treated mice or in mice acutely infused with ANGII. *P < 0.05 vs. Veh and acute ANGII; n = 3–4 per group; 86 ± 9 PVMs per animal (1-way ANOVA and Tukey’s test). (G) Electron micrographs of cortical arterioles showing immunoperoxidase labeling of biotinylated ANGII tracking along tight junctions (arrows) and reaching perivascular space (arrowheads) in mice treated with ANGII for 14 days (right) but not in saline-treated mice (left). Scale bar: 100 nm. PVM, perivascular macrophage; EC, endothelial cell; VSMC, vascular smooth muscle cells; BM, basement membrane.
To provide more direct evidence that blood-borne ANGII reaches the perivascular space in proximity to PVMs, we administered biotinylated ANGII i.v. on day 14 of the ANGII infusion, and used confocal and electron microscopy to detect the labeled peptide. As shown in Figure 2, D–F, nearly 60% of the PVMs were closely associated with biotinylated ANGII in mice treated with ANGII for 14 days (Figure 2, D–F). In contrast, little or no ANGII labeling was observed in mice implanted with pumps delivering saline or in which ANGII was administered acutely (Figure 2, D and F). Furthermore, by using electron microscopy, we demonstrated that ANGII-biotin, tracked along tight junctions, crossed the endothelium and reached the perivascular space (Figure 2G). Thus, in the “slow pressor” model circulating ANGII is able to cross the BBB and reach PVMs in the perivascular space.
PVM depletion by clodronate restores neurovascular function in ANGII slow pressor hypertension. Since ANGII reaches the perivascular space, we asked whether PVMs contribute to the neurovascular dysfunction induced by ANGII hypertension. To this end, we depleted PVMs using liposome-encapsulated clodronate (CLOD) (27). CLOD is phagocytosed by macrophages, leading to their demise by apoptosis. Thus, administration of CLOD into the cerebral ventricles (i.c.v.) is widely used to deplete PVMs in brain (28, 29). In mice treated with vehicle (liposome-encapsulated PBS), PVMs, as described in Figure 1, were observed wrapping around cerebral blood vessels (Figure 3A). CLOD treatment selectively reduced PVMs (CD206+CD45hiCD11b+), but not brain macrophages (CD206–CD45hiCD11b+) (Figure 3, A and B, and Supplemental Figure 3, A and B). CLOD did not deplete microglial cells (Supplemental Figure 3C) or blood leukocytes (Supplemental Figure 3D).
PVM depletion by CLOD restores neurovascular function in ANGII slow pressor hypertension. (A and B) Intracerebroventricular CLOD administration induces depletion of cells expressing the PVM marker CD206 in the somatosensory cortex assessed by histology. The vasculature is visualized by double label with the endothelial marker GLUT-1. The pial surface is at the top of the figure, and penetrating vessels can be seen diving into the neocortex. Scale bars: 50 μm. *P < 0.05 vs. PBS-Veh and PBS-ANGII; n = 5–10 per group (2-way ANOVA and Bonferroni’s test). (C) CLOD administration does not affect the increases in mean arterial pressure (MAP) induced by ANGII. *P < 0.05 vs. PBS-Veh and CLOD-Veh; n = 5–8 per group (2-way ANOVA and Bonferroni’s test). (D) CLOD does not alter the increase in CBF induced by whisker stimulation or cortical administration of ACh, but counteracts the attenuation of both responses induced by ANGII. *P < 0.05 vs. PBS-Veh and CLOD-Veh; #P < 0.05 vs. PBS-ANGII; n = 5–7 per group (2-way ANOVA and Bonferroni’s test).
Having established the effectiveness of CLOD, we then examined whether PVM depletion counteracts the cerebrovascular dysfunction induced by ANGII. CLOD or vehicle was administered i.c.v., and 7 days later mice were implanted with osmotic minipumps delivering ANGII or saline. Cerebrovascular function was assessed 2 weeks after minipump implantation. As before (10), slow pressor ANGII attenuated the CBF increase produced in the somatosensory cortex by mechanical stimulation of the whiskers (neurovascular coupling) and endothelium-dependent CBF responses to neocortical application of acetylcholine (ACh) (Figure 3D). Responses to adenosine were not affected (Supplemental Table 1), ruling out that the neurovascular dysfunction was due to failure of the vasomotor apparatus. CLOD depleted PVMs by 60%–65% in mice receiving vehicle or ANGII (Figure 3B), but did not alter the increase in BP (Figure 3C and Supplemental Figure 3E). Similarly, in vehicle-treated mice CLOD did not affect the CBF increases evoked by whisker stimulation, ACh, or adenosine (Figure 3D and Supplemental Table 1). However, in ANGII-treated mice CLOD counteracted the attenuation of the vascular responses (Figure 3D), implicating PVMs in the mechanisms of the dysfunction.
Acute i.v. administration of high concentrations of ANGII alters neurovascular function (11). Since ANGII needs to cross the BBB to reach the PVM, we hypothesized that the dysfunction induced by acute i.v. ANGII, which does not alter the BBB (Figure 2C), would be insensitive to PVM depletion. On the other hand, the neurovascular dysfunction caused by application of ANGII to the subarachnoid space underlying the cranial window (30), which bypasses the BBB and has direct access to the perivascular space, would be rescued by CLOD. Consistent with this prediction, CLOD counteracted the CBF dysfunction produced by subarachnoid application of ANGII, but not that induced by acute i.v. infusion of the peptide (Figure 4, A and B), attesting to the specificity of the effect of PVM depletion. These observations indicate that PVMs are able to induce neurovascular dysfunction when exposed to ANGII.
PVM depletion counteracts the neurovascular dysfunction induced by topical neocortical application but not by acute i.v. ANGII administration. (A) CLOD has no effect on the neurovascular dysfunction induced by acute i.v. administration of ANGII. *P < 0.05 vs. PBS-Veh and CLOD-Veh; n = 6 per group (2-way repeated-measures ANOVA and Bonferroni’s test). (B) CLOD-mediated PVM depletion rescues the neurovascular dysfunction induced by neocortical application of ANGII to the subarachnoid space (ANGII topical). *P < 0.05 vs. PBS-Veh and CLOD-Veh; n = 6–7 per group (2-way repeated-measures ANOVA and Bonferroni’s test).
PVM AT1Rs participate in the harmful cerebrovascular effects of ANGII. AT1Rs are involved in the cerebrovascular dysfunction induced by ANGII (10). Macrophages are well known to express AT1R and produce ROS in response to ANGII (23, 24). Therefore, we sought to determine whether AT1Rs on PVMs participate in the cerebrovascular effects of ANGII. To this end, we used BM chimeras to replace PVMs with PVMs lacking AT1Rs in WT mice (Atr1–/–) (20, 21). We then assessed the cerebrovascular effects of ANGII “slow pressor” hypertension in WT mice with transplanted Atr1–/– BM.
In preparation for these studies, we first determined the extent and rate of replacement of PVMs by BM-derived monocytes using WT mice with transplanted GFP+ BM. After BM transplant, GFP+CD206+ cells with the characteristic morphology of PVMs were observed along pial and penetrating vessels (Figure 5A). By 14 weeks after the BM transplant, 78% ± 2% of CD206+ cells were also GFP+, attesting to the high proportion of host PVMs replaced by donor PVMs (Figure 5B). Genomic DNA analysis confirmed that BM-derived cells were At1r–/– in WT mice engrafted with At1r–/– marrow (_At1r–/–_→WT), and At1r+/+ in WT mice engrafted with WT marrow (WT→WT) (Supplemental Figure 4A). The chimerism (_At1r–/–_→WT) was greater than 95% (Supplemental Figure 4A). Since ANGII can influence monocyte mobilization and trafficking (31), we sought to rule out that At1r deletion or ANGII administration alters monocyte migration and homing to the perivascular space. To this end, we counted perivascular CD206+ cells (PVMs) in the neocortex of BM chimeras with and without ANGII administration. No effect of At1r deletion or ANGII administration on PVM numbers was found (Supplemental Figure 4E).
PVM At1r deletion counteracts the harmful cerebrovascular effects of ANGII slow pressor hypertension. (A) Fourteen weeks after transplant of GFP+ bone marrow in WT mice, GFP+ and CD206+ PVM surrounds GLUT-1+ vessels. Scale bar: 50 μm. (B) By 14 weeks after the transplant, approximately 80% of CD206+ cells in the somatosensory cortex are replaced by GFP+ bone marrow cells. *P < 0.05 vs. 7 weeks group [χ2(1) = 7.78]; n = 5 per group. (C) ANGII increases mean arterial pressure equally in WT mice with transplanted WT bone marrow (WT→WT) or At1r–/– marrow (At1r–/– WT). *P < 0.05 vs. WT→WT–Veh and _At1r–/–_→WT–Veh; n = 7–10 per group (2-way ANOVA and Bonferroni’s test). (D) At1r deletion in PVMs does not affect the CBF responses to whisker stimulation or ACh, but counteracts the attenuation of both responses induced by ANGII. *P < 0.05 vs. WT→WT–Veh and _At1r–/–_→WT–Veh; #P < 0.05 vs. WT→WT–ANGII; n = 7–12 per group (2-way ANOVA and Bonferroni’s test).
CBF studies were performed 14 weeks after BM transplant, when the majority of PVMs had been replaced by the donor’s macrophages (see above). In BM chimeras, ANGII administration elevated BP equally in WT→WT and _At1r–/–_→WT chimeras (Figure 5C and Supplemental Figure 4C). WT mice engrafted with WT BM (WT→WT) exhibited normal cerebrovascular reactivity and attenuation by ANGII (Figure 5D), indicating that the irradiation and BM transplantation did not alter basic CBF responses and their susceptibility to ANGII. However, in _At1r–/–_→WT chimeras CBF responses to neural activation and ACh were markedly improved (Figure 5D). We then sought to assess the specificity of the rescue of the ANGII-induced cerebrovascular alterations in _At1r–/–_→WT chimeras. We found that _At1r–/–_→WT chimeras were protected from the cerebrovascular dysfunction induced by neocortical application of ANGII, but not of amyloid-β (Figure 6, A and B), a peptide that induces cerebrovascular dysfunction independently of AT1Rs (32). Therefore, AT1Rs in PVMs are required only for the neurovascular dysfunction induced by ANGII.
At1r deletion in PVMs rescues the neurovascular dysfunction induced by topical neocortical application of ANGII but not amyloid-β. (A) _At1r–/–_→WT chimeras are protected from the cerebrovascular effects induced by topical neocortical application of ANGII. *P < 0.05 vs. WT→WT–Veh and _At1r–/–_→WT–Veh; n = 5–7 per group (2-way repeated-measures ANOVA and Bonferroni’s test). (B) The attenuation of the CBF responses induced by topical application of amyloid-β (Aβ), an effect not dependent on AT1Rs, is not affected. This finding attests to the specificity of the effect of PVM At1r deletion on the cerebrovascular dysfunction induced by ANGII. *P < 0.05 vs. WT→WT–Veh and _At1r–/–_→WT–Veh (2-way repeated-measures ANOVA and Bonferroni’s test).
NOX2 in PVMs is required for the cerebrovascular effects of ANGII. Activation of AT1Rs leads to production of ROS derived from a NOX2-containing NADPH oxidase that has been implicated in the neurovascular dysfunction induced by ANGII slow pressor hypertension (10). Since macrophages are enriched in NOX2, in these experiments we sought to determine whether NOX2 in PVMs is responsible for the ANGII-induced neurovascular dysfunction. To this end, we generated _Nox2–/–_→WT chimeras. First, we confirmed, by genomic DNA analysis in BM-derived cells, that, in _Nox2–/–_→WT chimeras, PVMs are Nox2–/– (Supplemental Figure 4B). Fourteen weeks after BM transplant, ANGII minipumps were implanted, and cerebrovascular function was tested 2 weeks later. Nox2 deletion in BM cells (_Nox2–/–_→WT chimeras) did not affect slow pressor hypertension (Supplemental Figure 4D) or baseline cerebrovascular reactivity (Figure 7A), but ameliorated the cerebrovascular dysfunction induced by ANGII in (Figure 7A), implicating NOX2 in PVMs in its mechanisms. However, the rescue of the dysfunction was not complete, possibly implicating ROS sources other than NOX2 (33).
NOX2 in PVMs is required for the cerebrovascular effects of ANGII slow pressor hypertension. (A) Nox2 deletion in PVMs does not affect baseline cerebrovascular responses, but counteracts the cerebrovascular dysfunction induced by ANGII (_Nox2–/–_→WT chimeras). *P < 0.05 vs. WT→WT–Veh and _Nox2–/–_→WT–Veh; #P < 0.05 vs. WT→WT–ANGII; n = 5–10 per group (2-way ANOVA and Bonferroni’s test). (B) ANGII markedly increases ROS production in CD206+ cells (PVMs) assessed by dihydroethidine (DHE) microfluorography. Scale bar: 25 μm. (C and D) Deletion of At1r or Nox2 in PVMs attenuates the neurovascular oxidative stress induced by slow pressor ANGII in the neocortex. Scale bar: 50 μm. *P < 0.05 vs. WT→WT–Veh, _At1r–/–_→WT–Veh, and _Nox2–/–_→WT–Veh; #P < 0.05 vs. WT→WT–ANGII; n = 4–10 per group (2-way ANOVA and Bonferroni’s test).
Deletion of At1r or Nox2 in PVMs attenuates ANGII-induced vascular oxidative stress. The cerebrovascular dysfunction induced by ANGII slow pressor hypertension is mediated by ROS (10, 16, 17, 34). ANGII increases vascular ROS production, and the increase is reversed by ROS scavengers, and is not observed in NOX2-null mice, implicating a NOX2-containing NADPH oxidase as a source of ROS (10). Because of their close apposition to the outer vessel wall and large capacity for ROS production, PVMs can be a powerful source of vascular oxidative stress. As illustrated in Figure 7B, ANGII induces an increase in ROS signal, assessed by hydroethidine microfluorography (10, 35) in PVMs. To test whether deletion of At1r or Nox2 in PVMs suppresses oxidative stress induced by ANGII, we used _At1r–/–_→WT and _Nox2–/–_→WT chimeras and WT→WT chimeras as controls. In WT→WT mice, ANGII increases vascular ROS production (10, 35). In both _At1r–/–_→WT and _Nox2–/–_→WT chimeras, the increase in ROS induced by ANGII was attenuated (Figure 7, C and D), indicating that PVMs through AT1Rs and NOX2 are responsible for the vascular oxidative stress induced by ANGII in the somatosensory cortex.
PVM depletion ameliorates CBF dysfunction also in chronically hypertensive BPH/2J mice. A drawback of the slow pressor ANGII hypertension is that the elevation in BP is induced pharmacologically and is restricted to 2 weeks. To overcome these limitations, we also examined the role of PVMs in a model of spontaneous lifelong hypertension more representative of chronic hypertension in humans. To this end, we used BPH/2J mice, a mouse line developed from breeding of several strains of mice and selecting for the hypertension phenotype (36). BPH/2J mice exhibited elevated BP at 4 weeks of age (Supplemental Figure 5A), compared with normotensive controls, and the BP elevation remained stable until the mice were studied at 4–6 months of age. Because of their shortened lifespan (37), this age corresponds to middle age. At this time, levels of ANGII were elevated in plasma but not brain (Supplemental Figure 5, B and C). The BBB permeability to FITC-dextran was increased (Figure 8A). The increase in CBF induced by functional hyperemia or ACh was attenuated, whereas the CBF response to adenosine was normal (Figure 8B and Supplemental Table 1), attesting to the integrity of the vasorelaxing function of the smooth muscle. The CBF dysfunction was counteracted by neocortical application of the AT1R antagonist losartan or the ROS scavenger MnTBAP (Figure 8B), indicating that the neurovascular alterations are mediated by AT1Rs and ROS. CLOD administration (i.c.v.) in BPH/2J mice depleted PVMs in both cortex and hippocampus (Supplemental Figure 5, D and E), and rescued the response to both whisker stimulation and ACh (Figure 8D). These data implicate PVMs also in the cerebrovascular dysfunction occurring in a model of lifelong spontaneous hypertension.
PVMs mediate cerebrovascular dysfunction in chronically hypertensive BPH/2J mice. (A) Mean arterial pressure and BBB permeability to FITC-dextran are increased in BPH/2J mice. *P < 0.05 vs. control; #P < 0.05 vs. Veh; n = 5–9 (Student’s t test). (B) Antagonism of AT1Rs with losartan or ROS scavenging with MnTBAP counteracts the neurovascular dysfunction in BPH/2J mice. *P < 0.05 vs. control; n = 4–7 per group (1-way ANOVA and Tukey’s test). (C and D) CLOD has no effect on the mean arterial pressure but completely reverses the attenuation in CBF response to whisker stimulation and ACh in BPH/2J mice. *P < 0.05 vs. PBS-control and CLOD-control; #P < 0.05 vs. PBS–BPH/2J; n = 5 per group (2-way ANOVA and Bonferroni’s test).
PVM depletion improves cognitive function in BPH/2J mice with chronic hypertension. Finally, we sought to determine whether the rescue of cerebrovascular dysfunction afforded by PVM depletion leads to an improvement in the cognitive deficits associated with hypertension. In this study, we elected to use middle-aged BPH/2J mice, a model of spontaneous lifelong hypertension resulting in well-documented and reproducible cognitive deficits (38). We did not assess cognitive function in ANGII-treated mice because higher concentrations of ANGII and longer exposures are needed to observe cognitive deficits (4 weeks) (39, 40). Such longer administration times are not compatible with the time course of PVM depletion by CLOD, since PVMs begin to repopulate the perivascular space at this time (27). To examine the role of PVMs in the cognitive deficits in BPH/2J mice, we first used the novel object recognition task, a test of recognition memory that relies on the propensity of mice to explore novel objects more than familiar ones (41). Normotensive mice presented with an object to which they were previously exposed and a novel one spent more time exploring the novel object (Figure 9A). In contrast, BPH/2J mice spent equal time exploring the novel and the familiar object, reflecting a deficit in recognition memory (Figure 9A). The effect was observed both 1 and 24 hours after the familiarization trial (Supplemental Figure 5, F and G). CLOD treatment did not affect the performance of the normotensive controls and the total exploration time in both groups, but greatly ameliorated the memory deficit in BPH/2J mice (Figure 9A).
PVMs mediate cognitive dysfunction in chronically hypertensive BPH/2J mice. (A) PVM depletion by CLOD rescues the recognition memory deficits assessed by the novel object recognition test in BPH/2J mice. *P < 0.05 vs. pre-CLOD–control and post-CLOD–control; _n_ = 10–12 per group (2-way repeated-measures ANOVA and Bonferroni’s test). (**B**) PVM depletion does not affect the spatial memory assessed by Barnes maze test in control mice but rescues the deficits observed in BPH/2J mice. *_P_ < 0.05 vs. pre-CLOD–BPH/2J; _n_ = 15–20 per group (2-way repeated-measures ANOVA and Bonferroni’s test). (**C**) CLOD tends to reduce the distance traveled in BPH/2J mice, but the effect does not reach statistical significance. _P_ > 0.05 vs. pre-CLOD–BPH/2J; n = 15–20 per group (2-way repeated-measures ANOVA and Bonferroni’s test). (D) Representative tracks for control and BPH/2J mice on acquisition day 3 before and after vehicle or CLOD injection.
To provide additional evidence for the involvement of PVMs in the cognitive deficits induced by chronic hypertension, we then used the Barnes maze. This test is similar to the water maze navigation task in that it examines the ability of the mouse to learn and remember the location of a target zone (spatial learning and memory) (42). However, the Barnes maze has the advantage of avoiding the strong aversive stimulus of swimming that may affect performance (43). The mouse is placed on a circular platform with holes at its outer edge, 1 of which is the correct escape hole. Within a few trials a normal mouse will accurately identify the escape hole, avoiding the incorrect holes. The test is repeated daily for 3 days to test the ability of the mouse to retain the information, once the location of the escape hole was learned. Spatial memory is assessed by measurement of both the time spent to find the escape hole (primary latency) and the distance traveled to do so (44). Both BPH/2J and normotensive controls showed spatial learning over the 3-day test, as indicated by the reduced primary latency and distance traveled (Figure 9, B and C). However, both primary latency and distance traveled were significantly higher in BPH/2J mice compared with normotensive mice, reflecting spatial memory impairment (Figure 9, B and C). PVM depletion by CLOD did not affect the performance of the normotensive controls, but significantly improved the cognitive performance of the BPH/2J mice (Figure 9C).