|Home | About | Journals | Submit | Contact Us | Français|
The precise regulation of cerebral blood flow is critical for normal brain function and its disruption underlies many neuropathologies. The extent to which smooth muscle-covered arterioles or pericyte-covered capillaries control vasomotion during neurovascular coupling remains controversial. We found that capillary pericytes in mice and humans do not express smooth muscle actin and are morphologically and functionally distinct from adjacent precapillary smooth muscle cells (SMCs). Using optical imaging we investigated blood flow regulation at various sites on the vascular tree in living mice. Optogenetic, whisker stimulation or cortical spreading depolarization caused microvascular diameter or flow changes in SMC but not pericyte-covered microvessels. During early stages of brain ischemia, transient SMC but not pericyte constrictions were a major cause of hypoperfusion leading to thrombosis and distal microvascular occlusions. Thus, capillary pericytes are not contractile and regulation of cerebral blood flow in physiological and pathological conditions is mediated by arteriolar smooth muscle cells.
Cerebral function consumes a large amount of energy; however, the precise locations and extent of neural activity within regions of the brain are constantly fluctuating. Consequently, the brain has evolved a specialized system to ensure coupling between energy demand and supply through precise spatial and temporal modulation of cerebral blood flow (CBF) (Hamel, 2006; Iadecola, 2004; Raichle and Mintun, 2006; Roy and Sherrington, 1890). While the mechanisms controlling neurovascular coupling are not well understood, previous work demonstrates the involvement of complex interactions between neurons, glia and vascular cells (Haydon and Carmignoto, 2006; Iadecola and Nedergaard, 2007; Kety and Schmidt, 1948; Roy and Sherrington, 1890). Neural activation induces transient local microvascular dilations, leading to increased blood flow and tissue oxygenation, a phenomenon that forms the basis for functional magnetic resonance imaging (fMRI) (Logothetis et al., 2001; Ogawa et al., 1990). The modulation of vessel diameter, vascular resistance, and blood flow is controlled by cells within the microvascular wall (mural cells), which have contractile properties. A precise understanding of the function of these cells in vivo is not only important for investigating the source of the signals obtained with fMRI blood-oxygen-level dependent (BOLD) techniques (Attwell and Iadecola, 2002; Logothetis and Wandell, 2004; Raichle and Mintun, 2006), but also for elucidating the pathophysiology of many diseases involving the brain microvasculature (Iadecola and Nedergaard, 2007; Puro, 2007; Winkler et al., 2014).
Mural cells on the cerebral vascular tree include arteriolar and venular smooth muscle cells (SMCs) and capillary pericytes (Rouget, 1874). These cells are thought to play important roles in microvascular development, angiogenesis (Armulik et al., 2010; Daneman et al., 2010), maintenance of the blood brain barrier (Bell et al., 2010), and are implicated in a variety of neuropathological conditions (Bell et al., 2010; Hall et al., 2014; Sagare et al., 2013; Yemisci et al., 2009). It is well known that microvascular smooth muscle regulates vessel diameter and blood flow (Brian et al., 1998; Devor et al., 2007; Fernández-Klett et al., 2010; Kornfield and Newman, 2014; Vanzetta et al., 2005). However, within brain micro-regions, the precise location where neural activity-induced vasomotion is modulated is less clear. Specifically, which segments of the vascular tree, especially at the transition between arterioles and terminal capillaries, are the primary sites of CBF regulation remains a topic of debate (Armulik et al., 2011; Hamilton et al., 2010; Iadecola and Nedergaard, 2007; Itoh and Suzuki, 2012; Krueger and Bechmann, 2010; Winkler et al., 2011).
Recent studies have suggested capillaries to be major sites for active CBF regulation (Chaigneau et al., 2003; Hall et al., 2014; Peppiatt et al., 2006). In fact, it was estimated that up to 84% of blood flow modulation may take place at the terminal capillary level, where pericytes are the predominant mural cell, suggesting that these cells have prominent contractile properties (Hall et al., 2014). However, separate studies in the cortex and retina suggest that while pericyte contractility does occur, the direct role of capillaries in flow control may not be very significant (Fernández-Klett et al., 2010; Kornfield and Newman, 2014). In addition to physiological control of microvascular flow, in cerebral ischemia, constriction of capillary pericytes that persists after their death has been proposed to prevent tissue reperfusion, leading to the “no-reflow” phenomenon (Hall et al., 2014; O’Farrell and Attwell, 2014; Yemisci et al., 2009). However, the difficulty in distinguishing the various mural cell types in vivo, especially at the transition between arterioles and capillaries, and the variability in experimental methodologies, have brought the concept of physiological and pathological capillary pericyte contractility into question (Vates et al., 2010). Resolution of these discrepancies is of great importance not only because of its implications for modeling physiological micro-regional blood flow and brain oxygenation, but also for understanding the role of the microvasculature and mural cells in neuropathological conditions.
We utilized a combination of genetically encoded microvascular mural cell labeling, high resolution imaging of vasomotor activity, functional calcium imaging, and optogenetic cell activation to precisely characterize the structural and functional properties of cerebral vascular mural cells in vivo. We demonstrate that capillaries are incapable of active vasomotor responses to a variety of stimuli such as direct pericyte optogenetic stimulation, physiological neural activation, and spreading depolarization. Pericytes in brain capillaries completely lack expression of smooth muscle actin (SMA), consistent with our finding that only microvessels enveloped by mural cells with a circumferential band-like morphology, typical of smooth muscle, display active vasomotor responses. Furthermore, we find that the diameter and branch order of precapillaries with circumferential smooth muscle is frequently indistinguishable from that of pericyte-covered capillaries. Thus, ambiguous distinction between these cell types could partly explain contradicting reports about capillary pericyte contractility. Finally, we demonstrate that during brain ischemia, SMC but not pericyte constriction leads to vessel occlusion due to secondary thrombus formation within distal microvessels. Our study provides novel functional and structural in vivo evidence defining the cellular components of the microvascular contractile apparatus, with important implications for future studies of the normal and pathological regulation of cerebral blood flow.
In order to perform a comprehensive study of the vasomotor properties of mural cells, we first set out to determine their precise distributions and morphological features at various locations within the vascular tree in the mouse cerebral cortex. A detailed morphological analysis was necessary in order to accurately define the various mural cells that could potentially control cerebral blood flow as inconsistencies in cell definition and identification likely contribute to contradicting reports in the literature. We first imaged mice that expressed membrane bound GFP (mGFP) specifically in cells with constitutive or inducible Cre recombinase driven by the NG2 promoter (NG2cre:mT/mG and NG2creER:mT/mG). NG2cre mice have Cre expression specifically in SMCs, pericytes, NG2 cells and oligodendrocytes, each of which can be clearly distinguished by their morphology and close association with vessels as described previously (Hill et al., 2014; Zhu et al., 2008). In constitutive NG2cre mice, Cre recombination efficiency was sufficient to observe near complete mGFP labeling of NG2 expressing mural cells, while single isolated cells could be visualized in inducible NG2creER mice after a single injection of tamoxifen. This approach revealed cells with distinct morphologies at each level of the vascular tree (Figure 1, Figure S1).
Cells with classical smooth muscle morphology, forming narrow circumferential bands that surround the entire vessel, were found on pial and penetrating arterioles with diameters ranging from 15–40 μm (Figure 1A, Figure S1). On more distal vessels, 3–15 μm in diameter, we also observed similar band-like cells with variable lengths all of which enveloped the entire vessel circumference (Figure 1B–G, Figure S1). Based on their circumferential band-like morphology, we classified these two mural cell populations as arteriole and pre-capillary SMCs respectively. Within the subset of smaller microvessels, with diameters ranging from 3–9 μm, we also observed morphologically distinct cells with processes that extended longitudinally for hundreds of microns along multiple capillary branch points (Figure 1B–H, Figure S1). In addition to their long processes, these cells also had short and thin processes that projected orthogonally across the vessel but rarely spanned its entire circumference (Figure S1). We classified this type of cell as capillary pericytes, due to their unique morphology and location. Immediately adjacent cells in post-capillary venules, displayed a band-like circumferential phenotype similar to arteriolar smooth muscle, but spanned larger distances and had a highly fenestrated patchwork-like appearance (Figure 1A–B, G).
To further distinguish the various mural cell types we crossed NG2cre mice with the brainbow-like Cre dependent confetti reporter (NG2cre:confetti). Consistent with mT/mG reporter mice, we could precisely identify morphologically distinct single cells (due to differential fluorescent protein labeling) with either SMC or pericyte morphology, further delineating the single cell territories and discrete morphologies of each cell type (Figure S2).
We then set out to determine the contractile properties of vessels at various levels of the vascular tree. We used in vivo two photon imaging to measure spontaneous single vessel diameter changes in the somatosensory cortex of awake head-fixed NG2cre:ZEG mice injected with an intravascular dye to visualize all vessels. In vivo time-lapse imaging in these mice revealed that 10–50μm diameter arterioles with a layer of cells with SMC morphology spontaneously dilated and contracted at relatively high frequencies while venules rarely displayed significant diameter changes. We focused our analysis on vessels smaller than 10μm in diameter as this is where we observed the transition from pre-capillary SMCs to capillary pericytes (Figure 1, Figure 2A–D). Consistent with the heterogeneous mural cell morphology that we observed, these vessels displayed highly variable spontaneous vasomotion that did not correlate well with their respective diameter (Figure 2A–D, Movie S1). To accurately describe this contractile behavior, we quantified the area under the curve of the percent changes in vessel diameter (Figure 2C, F–G), a measure that accounts for amplitude, frequency, and duration of diameter changes, and which we termed vasomotion index (see experimental procedures). This quantification revealed that the heterogeneous spontaneous diameter changes were clearly correlated to the type of mural cell covering these vessels (Figure 2G) (vasomotion indices: pre-capillary SMC-covered vessels 38.2 ± 6.3; capillary pericyte-covered vessels 1.8 ± 0.5, p <0.001 unpaired t-test).
To further characterize and correlate a functional intracellular signal in single mural cells with changes in vessel diameter we generated mice with a Cre dependent genetically encoded calcium indicator (GCaMP3) expressed specifically in NG2 positive mural cells (and some neurons, see below) (NG2cre:GCaMP3). Detailed analysis of changes in GCaMP3 fluorescence intensity showed SMC calcium transients preceding, coinciding or following vasomotion with a significant anti-correlation between increases in SMC calcium and vessel dilation (Figure 3A–B,F, Movie S1) (correlation, r = −0.49 ± 0.21 standard deviation, n = 24 vessel segments from 3 mice, significant differences from 0 correlation determined by a 99% confidence interval).While transient robust changes in GCaMP3 fluorescence were also detected in pericytes, no changes in vessel diameter were detected (Figure 3C–D, Movie S3) and thus there was no correlation between pericyte calcium fluctuations and changes in vessel diameter (Figure 3E–F) (correlation, r = 0.02 ± 0.07 standard deviation, n = 12 pericytes from 3 mice).
These morphological and functional data suggested that the presence or absence of the different mural cell types determined whether active vasomotion occurred in individual microvessels. To directly test this hypothesis we next looked for the presence of contractile proteins to correlate molecular and cellular identity with in vivo function.
To understand the reasons behind the heterogeneous contractility in smaller vessels we performed in vivo imaging of double transgenic mice expressing the fluorescent protein mCherry driven by the SMA promoter and endothelial GFP driven by the Tie2 promoter (SMA-mCherry:Tie2-GFP). In these mice, all vessels were labeled with GFP and arterioles were enveloped with a bright layer of SMA-mCherry labeled mural cells (Figure 4A–B). Artery specificity of mCherry was determined by the directionality of the blood flow and confirmed by labeling with intravenously injected Alexa Fluor 633 hydrazide (Figure S4) which binds to the elastin layer in arterioles (Shen et al., 2012). In addition to detecting bright SMA-mCherry on large penetrating arteries and arterioles, we often observed mCherry expression on smaller vessels with diameters less than 10μm (Figure 4C–G). Quantification revealed that the last SMA-mCherry expressing cells at the transition between arterioles and capillaries were found at both branch points and in the middle of single vessels with diameters ranging from 3.3 to 8.5 μm (Average of 5.4±1.3 μm) and branch orders ranging from 1st to 4th (0 being a penetrating arteriole) (n = 45 vascular tress from 4 mice) (Figure 4C–J). Unexpectedly, at the location of these terminal SMA-mCherry expressing cells, we frequently observed that the vessel diameter was smaller compared to the downstream capillary (Figure 4G), strongly suggesting that mural cells in capillaries are not capable of providing a baseline contractile tone like distal SMCs.
To verify that the observed lack of SMA expression in capillaries was not due to incomplete mosaic expression in our SMA-mCherry transgenic mice, we confirmed our observations with immunohistochemistry for SMA (Figure 4K) as well as in a separate transgenic line with mGFP expressed specifically in SMA+ cells (SMAcreER:mT/mG) (Figure S3). SMA antibody staining showed identical expression to SMA-mCherry mice and was not found in pericytes (Figure 4K, Figure S4). In addition to NG2 and SMA, expression of the beta receptor for platelet derived growth factor (PDGFRβ) is routinely used to identify pericytes. Importantly however, we found expression of PDGFRβ on both SMA+ SMCs and SMA-pericytes (Figure 4K, Figure S4).
In addition to seeing SMA-mCherry labeled cells on arteries and precapillary arterioles, due to bright fluorescence in these mice, we also observed weak mCherry expression on large venules (>50 μm) but not on smaller post-capillary vessels (Figure S3) suggesting that large venules have some limited contractile potential. Importantly, the SMA expression in arterioles and its absence from terminal capillaries was not unique to mice as we also observed it in human neocortex. (Figure 4J). Quantification of postmortem human tissues stained with antibodies against SMA revealed that 8.5±0.3% of the total length of human intraparenchymal cerebral vessels (n = 3 human samples) were enveloped with SMA-expressing mural cells. Similar to mice, the last SMA-expressing cells at the transition between arterioles and capillaries were found at both branch points and in the middle of single vessels, ranging from 4.63 to 15.5 μm in diameter (average diameter 9.1±0.4 μm, n = 41 vascular trees from 3 human samples).
These data provide strong evidence that regardless of vessel diameter or branch order, SMA is expressed in mural cells with circumferential SMC morphology but is absent from pericytes. This provides a likely explanation for the heterogeneity in spontaneous vasomotion that we observed at the precapillary versus capillary levels (Figure 2).
To directly assess the contractile properties of individual cells at various levels of the vascular tree, we generated double transgenic mice which express a light gated proton channel (channelrhodopsin 2, ChR2) fused to yellow fluorescent protein (YFP) specifically in NG2 expressing cells (NG2cre:ChR2-YFP) and crossed them with SMA-mCherry mice to generate NG2cre:ChR2-YFP:SMA-mCherry triple transgenic mice. These mice express ChR2-YFP in cortical pericytes, SMCs, oligodendrocyte lineage cells and a small number of cortical astrocytes as well as mCherry in all SMC (Figure 5A). The presence of voltage gated calcium channels in both SMCs and pericytes (Figure S5) (Borysova et al., 2013; Sakagami et al., 1999), assured that ChR2 stimulation would cause depolarization of both cell types.
Using these mice, we implemented a confocal laser line-scan technique to optimally activate ChR2 and cause targeted mural cell stimulation while simultaneously imaging vessel diameter (see experimental procedures). In arterioles larger than 10μm in diameter, which were covered by SMA expressing SMCs, light activation caused significant vessel constrictions (−19.41±1.90% change in diameter, n = 58 vessels from 8 mice) (Figure 5B–E). Similarly, light activation on vessels less than 10μm in diameter but still covered by SMA expressing pre-capillary SMCs, resulted in significant vessel constrictions (Figure 5B–E) (−11.79±2.06% constriction, n = 18 vessels from 8 mice). In contrast, vessels less than 10μm in diameter covered by pericytes but not SMA expressing SMCs showed no constriction after identical light activation of ChR2 (Figure 5B–E) (0.08±0.18% change in diameter, 39 vessels from 8 mice). To exclude possible artifacts by non-mural cell ChR2 stimulation in constitutive NG2cre:ChR2-YFP mice, we used low dose tamoxifen induction in NG2creER:ChR2-YFP, where we were able to label cells very sparsely to isolate individual SMCs and pericytes thus removing any concern about nearby expressing cells. ChR2 stimulation of isolated cells demonstrated the same results with SMC constriction and lack of pericyte contractility (Figure S5)
Given that small and potentially difficult to detect changes in vessel diameter can induce substantial changes in blood flow proportional to the 4th power of radius changes (as per Poiseuille equation of fluid dynamics), we directly tested the impact of targeted ChR2 activation on blood flow velocity. In order to avoid potential out of focal plane ChR2 activation by the confocal laser, which could prevent accurate local flow measurements due to proximal vessel constrictions, we used instead targeted two-photon stimulation of ChR2 within small regions of interest (ROI) along single vessels. Consistent with the observed constrictions, ChR2 activation on vessels covered by SMCs significantly reduced blood flow velocity (Figure 5F–I) (−44.62 ± 11.80% change, n = 10 vessels from 3 mice), while ChR2 activation on vessels that were not covered by SMCs did not significantly change blood flow (Figure 5H–I) (−2.22 ± 3.16% change, n = 30 vessels from 3 mice). These data strongly demonstrate that in the intact mouse neocortex, single cell depolarization of SMCs but not pericytes results in robust modulation of vessel diameter and CBF.
To determine which vessels respond to an increase in neuronal activity and thus regulate functional hyperemia, we imaged the vascular response to whisker stimulation in the awake mouse somatosensory cortex. To correlate neuronal activity with vascular changes, we again used the NG2cre:GCaMP3 transgenic mice. Unexpectedly, we found that in these mice both mural cells and neurons are brightly labeled (Figure 6A), likely due to ectopic neuronal Cre expression (Nishiyama et al., 2014). The dual labeling with GCaMP3 in neurons and NG2-expressing cells was very useful because it allowed us to identify and image vessels with SMCs and pericytes (Figure 6B) while concurrently monitoring neuronal activity (Figure 6C–D).
We performed in vivo time-lapse imaging of vessel diameter and neuronal and mural cell calcium changes during and after sensory stimulation in the awake mouse. Whisker stimulation (30 sec of 5Hz, 100ms air puffs) evoked robust neuronal calcium transients and significant vessel dilations (~10% change at peak dilation) in somatosensory cortex vessels covered with smooth muscle (Figure 6E–G) (significance determined by a 99% confidence interval, vessel diameter range 6.2–21μm, branch order 0–3, n = 57 vessels from 4 mice). This immediate active dilation was not observed in vessels that lacked a smooth muscle layer (Figure 6E–G) (vessel diameter range 3.3–8.1μm, branch order 1–4, n = 83 vessels from 4 mice). While, there was a slight (~1%) delayed dilation in vessels lacking smooth muscle, this occurred after peak dilation had occurred in vessels with SMCs, an effect that is most likely a passive result of upstream increases in blood flow (Figure 6G). Whisker stimulation also induced a robust decrease in calcium dependent GCaMP3 fluorescence in SMCs which correlated temporally with changes in vessel diameter (Figure 6G) (significance determined by a 99% confidence interval, n = 19 vessel segments from 3 mice). In contrast, changes in calcium dependent GCaMP3 fluorescence in pericytes were not correlated with whisker stimulation (Figure 6G) (n = 27 vessel segments from 3 mice). These data strongly suggest that in the awake mouse cortex, functional hyperemia is mediated by calcium-dependent arteriolar SMC contractility rather than direct capillary pericyte contractility.
Cortical spreading depolarization (CSD) occurs under numerous pathological situations and results in altered cerebral blood flow (Leao, 1944; Somjen, 2001). To determine the precise location within the vascular tree of CSD-induced vasomotor responses we loaded neural cells with the calcium sensitive dye Oregon Green Bapta 1AM (OGB) in SMA-mCherry transgenic mice that had been intravenously injected with a fluorescent dye to visualize the vessel lumen (Figure 7A–B).
To induce CSD we performed a single microinjection of 0.5M potassium chloride (KCl) 1–2 mm from the imaging site. As predicted, these injections resulted in prolonged increases in OGB fluorescence intensity in neurons and astrocytes consistent with the time course of CSD (Somjen, 2001) (Figure 7C–E). We detected significant diameter changes during CSD that occurred exclusively in vessels with a smooth muscle layer (~25% increase at peak dilation, ~10% decrease at peak constriction, significance determined by a 99% confidence interval, n = 10 vessels from 3 mice) but not in capillaries with pericytes that lack SMA (n = 10 vessels from 3 mice) (Figure 7C–E). Therefore, consistent with the optogenetic and sensory evoked experiments above, modulation of blood flow during CSD is mediated by smooth muscle covered vessels and not capillary pericytes.
Finally, we tested if cerebral ischemia could induce focal constrictions in microvessels that could prevent reestablishment of blood flow. We implemented a transient (90 minutes) filament occlusion of the middle cerebral artery (tMCAO) (see experimental procedures) and we performed in vivo imaging of cerebral MCA branches before, during, after and 4 hours after occlusion (Figure 8A, Figure S6A–B). Consistent with our previous experiments, we focused our analysis on the transition points between the last SMC on pre-capillary arterioles and pericytes on capillaries using SMA-mCherry and NG2creER:mT/mG transgenic mice.
In order to sample a large region during and after tMCAO, high resolution tiled images covering the entire cranial window were acquired (Figure S6). Confirmation of vessel occlusion during tMCAO was determined by the presence of sluggish and turbulent CBF and corresponding hypoxia induced vessel dilation evident in 20–30μm diameter pial arterioles (Figure S6C). Pre-capillary arteriolar to capillary transitions were analyzed in these vessels by noting and measuring changes in vessel perfusion and diameter. We observed multiple outcomes both during and after the tMCAO, particularly at the last SMC. These terminal SMCs often displayed focal constrictions that were transient and observed only during MCAO or induced during MCAO and maintained after reperfusion (Figure 8B–E, Figure S7, Figure S8A). These focal SMC constrictions resulted in RBC perfusion block to downstream capillaries which either maintained some plasma flow (Figure 8E, Figure S7, Movie S4–6) or displayed complete vessel occlusion (Figure 8B,C–D, Figure S7). RBC perfusion block was observed in 48 out of 59 SMC to pericyte transitions with 21 and 27 of those displaying reversible or maintained focal SMC constrictions respectively (Figure 8F–G). Focal constrictions in capillaries were not observed (Figure 8 and Figure S7–S8) even at the location of pericyte cell bodies, 4 and 24hrs after tMCAO, as previously reported (Hall et al., 2014) (Figure S8). Measurement of vessel diameter in RBC perfusion blocked SMC to pericyte transitions (see Figure 8F for measurement location diagram) revealed a significant average decrease in vessel diameter at terminal SMCs but not in downstream same branch or second branch pericyte covered capillaries (Figure 8H) (n = 51 terminal SMC covered pre-capillary arterioles, 51 same branch capillaries, and 27 second branch capillaries from 5 mice; significance determined by unpaired two-tailed t-test). Further characterization of all SMC to pericyte transitions revealed a variety of perfusion and vessel change outcomes during and after tMCAO (Figure 8I). These outcomes were characterized as maintained RBC perfusion (Figure S10E), only plasma perfusion (Figure 8 and Figure S7), complete block with thrombus (Figure 8 and Figure S7), or vessel collapse / no intravascular dye (Figure S7B,E) (defined as collapse or no i.v. dye labeling spanning greater than 20μm vessel length).
Overall, these data suggest that during cerebral ischemia and reperfusion, SMCs display focal constrictions that are transient or prolonged. In combination with numerous other cellular processes taking place during cerebral ischemia, these SMC constrictions appear to play a significant role as an early mechanism that prevents tissue reflow, leading to potentially irreversible microvascular occlusion.
We have precisely identified and functionally characterized in vivo the contractile perivascular mural cells that control neurovascular coupling and micro-regional CBF. Our results demonstrate that CBF regulation takes place exclusively at microvessels covered by SMCs that have a distinct circumferential morphology and express SMA (Figure S9). Several lines of evidence demonstrate that capillary pericytes, which do not express SMA and have a non-circumferential, longitudinal morphology spanning long distances, do not have the capacity for direct regulation of blood flow. First, in vivo spontaneous vasomotion in awake mice occurs in vessels covered by SMCs, regardless of their diameter or branch order, but does not occur in pericyte-covered vessels. Second, spontaneous calcium fluctuations in SMCs correlate with vessel diameter changes while calcium fluctuations in pericytes do not induce changes in capillary diameter. Third, single cell in vivo optogenetic activation of SMCs causes vessel constriction and flow reductions, while identical activation of pericytes causes no capillary changes. Fourth, sensory stimulation evokes a rapid and robust dilation of SMC-covered vessels but only a slow and nearly undetectable dilation of pericyte-covered capillaries which is likely to be passive. Fifth, cortical spreading depolarization induces active vasomotion in SMC-covered vessels but not in pericyte-covered capillaries. These multiple lines of independent in vivo evidence demonstrate unambiguously that capillary pericytes are not contractile and thus cannot directly regulate CBF.
While the capillary network comprises the majority of the brain vascular system, our data demonstrates that direct regulation of blood flow occurs at the level of arteries, arterioles, and pre-capillaries. Our findings describing mural cell contractility and control of CBF by arterioles greater than 10μm in diameter are consistent with a number of previous studies using tissue explant preparations (Borysova et al., 2013; Kornfield and Newman, 2014) and in vivo brain vessel imaging (Devor et al., 2007; Drew et al., 2011; Fernández-Klett et al., 2010; Iadecola et al., 1997; Kornfield and Newman, 2014; Vanzetta et al., 2005). Importantly, however, we have for the first time precisely established a direct relationship between microvascular mural cell molecular identity, single cell morphology, and cellular function at each level of the cerebral vascular tree in vivo. We unambiguously demonstrate that in vessels smaller than 10μm in diameter, at the transition between SMC and pericyte-covered regions, there is great heterogeneity of contractile properties.
It is critical to note that vascular diameter and/or vessel branch order are not reliable determinants of contractile capacity as microvessels smaller than 10μm in diameter, with branch orders ranging from 1st–4th, are enveloped by mural cells with heterogeneous SMA expression and contractile capacity. Previous studies have relied on vessel diameter, branch order, or both to distinguish between arterioles and terminal capillaries. This distinction cannot discriminate between SMA-expressing pre-capillary SMCs and SMA-deficient capillary pericytes, and likely contributes to seemingly contradictory conclusions from our study and others (Fernández-Klett et al., 2010; Hall et al., 2014). In addition to discrepancies from identifying terminal capillaries by their location and size, it should also be noted that all mural cells, including SMA-expressing SMCs, express the NG2 chondroitin sulfate proteoglycan and PDGFRβ (Armulik et al., 2011; Nakayama et al., 2013; Uemura et al., 2002). Therefore, NG2 and PDGFRβ expression cannot be used to unambiguously distinguish pericytes from SMCs. Thus far, the only reliable means to identify contractile vessels, particularly those with diameters less than 10μm, is by SMA expression and the distinction between circumferential band-like SMC versus longitudinal pericyte morphologies.
The morphology of capillary pericytes provides some clues about their functional role. In stark contrast with the robust circumferential arrangement of SMCs, pericytes have thin processes that extend longitudinally for distances of up to ~300μm, span multiple capillaries and rarely wrap around vessels (Figure S1). This arrangement does not appear to be structurally suited for exerting radial forces capable of executing vessel diameter changes. The novel observation that capillaries can have even larger diameters than pre-capillary arterioles (Figures 2A, ,4G)4G) is highly suggestive of pericytes not even providing a baseline contractile tone to the vessel. In addition to these observations, the logic for altering vessel diameter at the capillary level as a mechanism for flow regulation is not clear. It is well documented that RBCs need to undergo cell deformation in order to move through 3–5μm capillaries (Noguchi and Gompper, 2005; Pawlik et al., 1981; Skalak and Branemark, 1969). Thus, if contractility existed in capillaries, it is likely that even small changes in diameter would result in RBC perfusion block rather than a graded change in flow, posing a significant risk of focal microvascular thrombosis. We speculate that this phenomenon explains why small precapillary vessels with terminal SMCs are prone to irreversible distal microvascular occlusions after SMC constriction during cerebral ischemia.
Instead of a contractile role for pericytes, we speculate that the long thin pericyte processes could provide a network for sensing, transmitting, or coordinating information between the neural and vascular compartments. Gap junction coupling between pericytes and endothelial and SMCs has been demonstrated (Borysova et al., 2013; Cuevas et al., 1984; Oku et al., 2001). Together with astrocytes (Simard et al., 2003), pericytes may play a role in coordinating neurovascular coupling (Itoh and Suzuki, 2012) through SMC contractility. Recent findings showing sensory-evoked dilation in first order vessels prior to zero order penetrating arterioles (Hall et al., 2014) could be explained if the signal to induce vasomotion is transmitted through local astrocytes and pericytes up the vascular tree, reaching SMC covered pre-capillary arterioles prior to upstream arterioles. Intercellular communication between mural cells could also partly explain some of the variability we observed in our correlation experiments between SMC calcium and vasomotion. This variability is likely due to a number of complex interactions between proximal and distal SMCs and how upstream and downstream changes in diameter can alter blood flow and resistance which could then independently alter vessel diameter in the absence of active vasomotion at the location of our SMC calcium measurements. However, even with these complex flow dynamics and cellular interactions, a highly significant anti-correlation was present between vasomotion and calcium fluctuations in arteriolar SMCs but was absent from capillary pericytes.
In neuropathological states, such as ischemia, irreversible pericyte constriction and death have been suggested to impair capillary reflow (Hall et al., 2014; Yemisci et al., 2009). Instead, we found that during early stages of ischemia and reperfusion after MCA occlusion, focal constrictions occur exclusively in vessels surrounded by SMCs (Figure S10). These constrictions were either transient or maintained and appeared to lead to downstream microvascular occlusion. We believe that the contradictory observations of vascular contractility in ischemia could stem from differences in experimental approaches (Vates et al., 2010) and from the belief that pericytes express SMA in vivo (Yemisci et al., 2009) and that SMCs are not located on brain vessels other than 0 order penetrating arterioles (Hall et al., 2014) which leads to misidentification of SMCs as pericytes.
“No-reflow” after ischemia is likely a multifaceted, complex phenomenon (Figure S10). The variable SMC contractile behaviors and downstream flow outcomes in our data suggest that there are numerous multicellular processes occurring during both ischemia and reperfusion. SMC constrictions likely contribute to the process by temporarily slowing or blocking flow, and thus initiating lumen thrombotic mechanisms and impairing clot washout. However, additional intrinsic or extrinsic mechanisms affecting the microvasculature, including capillary plugging by leukocytes and platelets (del Zoppo et al., 1991), luminal entrapment of thromboemboli by endothelial lamellipodia (Grutzendler et al., 2014; Lam et al., 2010), and interstitial swelling (Rezkalla and Kloner, 2002) are likely to significantly contribute to microvascular occlusion, vessel collapse, and impaired reperfusion.
Our anatomical data of the percentage of vessels covered by SMA-expressing SMCs, supports a model in which CBF is directly regulated by approximately 8% (by length) of all intraparenchymal cerebral microvessels. Given the distribution of penetrating and pre-capillary arterioles (Blinder et al., 2013) and considering that only those covered by SMA can regulate vessel diameter, we speculate that the resolution of CBF regulation is on the order of a 75–150μm radius of tissue. Therefore, while the demand for CBF changes may be detected at single capillaries, regulation of functional hyperemia likely does not occur at this level, but instead at SMC-covered pre-capillaries that feed relatively larger brain volumes. In addition to advancing our understanding of local tissue oxygenation, this observation has significant implications for techniques that use changes in CBF as a correlate for neural activity. Given the spatial regulation of CBF we observed, BOLD fMRI and optical imaging of intrinsic signals (Grinvald et al., 1986), may have a fundamental hypothetical limit in spatial resolution, independent of their technical limitations for signal detection.
In addition to understanding normal CBF regulation, our findings have important implications for neuropathology. SMC malfunction on the already small proportion of vessels covered by SMCs could induce dramatic changes in neurovascular coupling and CBF regulation for the entire brain. For example, cerebral amyloid angiopathy, which occurs in Alzheimer’s disease due to arteriolar accumulation of amyloid (Farkas and Luiten, 2001; Iadecola, 2004), could cause SMC malfunction on the small proportion of proximal vessels, with major consequences on CBF regulation for an entire volume that we estimate to have a radius of 75–150μm. Likewise, the functional outcome after occlusion or loss of single microvessels during aging and a number of disease states (Armulik et al., 2011; Faraci, 2011; Harb et al., 2012; Iadecola, 2004) likely depends on whether or not these vessels are covered by SMCs or pericytes. Therefore, careful analysis of distinct mural cells on single vessels directly affected by pathology is necessary to fully understand cerebrovascular malfunction in multiple disease states.
In summary, our study demonstrates that under normal and pathological states, microvascular SMCs control CBF and neurovascular coupling. These cells are morphologically, molecularly, and functionally distinct from capillary pericytes, despite the fact that they can be found on vessels traditionally defined as capillaries, based on vessel diameter and/or branch order. Thus, true cerebral capillary pericytes are not contractile in vivo and do not independently regulate CBF but may have as of yet unidentified non-contractile CBF regulatory functions.
All animal experiments were approved by the Yale University Institutional Animal Care and Use Committee (IACUC). See also Supplemental Experimental Procedures. The following transgenic mice were used: NG2cre (Zhu et al., 2008) (Jackson Labs # 008533); NG2creER (Zhu et al., 2011) (Jackson Labs # 008538); floxed ChR2-YFP (Madisen et al., 2012) (Jackson Labs # 012569); SMA-mCherry (Armstrong et al., 2010); SMAcreER (Wendling et al., 2009); mT/mG (Muzumdar et al., 2007) (Jackson Labs # 007676); Tie2-GFP (Motoike et al., 2000) (Jackson Labs # 003658); Z/EG (Novak et al., 2000) (Jackson Labs # 003920); floxed GCaMP3 (Zariwala et al., 2012) (Jackson Labs # 014538); Confetti (Snippert et al., 2010) (Jackson labs # 013731). Male and female mice aged 4–16 weeks were used and all experiments were in awake, alert, head fixed mice or mice anesthetized via intraperitoneal injection of 100 mg/kg ketamine and 10 mg/kg xylazine (k/x) by body weight as indicated.
An acute cranial window preparation was used for all in vivo imaging experiments as described previously (Hill and Grutzendler, 2014; Schain et al., 2014). For two-photon imaging a mode locked MaiTai laser (Spectra Physics) was used on a two photon microscope (Prairie Technologies) with a 20× water immersion objective (Zeiss 1.0 NA). In some cases a confocal (Leica SP5) with a 20× water immersion objective (Leica 1.0 NA) was used for optimal fluorophore excitation and emission separation and for activation of channelrhodopsin (see below).
Texas Red dextran (70,000mw, Life Technologies cat# D-1864) or Cascade Blue dextran (10,000mw, Life Technologies cat# D-1976) were injected intravenously to label blood plasma and were used to determine changes in vessel diameter and flow velocity. Analysis of single cell morphology (in addition to SMA-mCherry expression) as described in Figures 1–4 was carried out from high resolution confocal or two photon Z stacks of single mural cells expressing YFP fused to ChR2 that were acquired using light wavelengths which were not optimal for stimulation of ChR2 but sufficient for YFP excitation and detection (561nm for confocal and 900nm for two photon) in order to avoid any chance of ChR2 stimulation due to imaging alone. To activate ChR2, a laser line scan or ROI based local activation were used. We used a 473/488nm confocal laser line scan activation protocol for optimal excitation of ChR2. In order to avoid potential artifacts from out of focus light activating upstream vessels and causing nonspecific changes in flow, we used two-photon laser stimulation (tuned to 800nm) (Duan et al., 2014) to determine the effect of single cell activation on blood flow. While both were effective at causing activation of ChR2, each method proved optimal for diameter and flow measurements respectively.
We used the mouse whisker sensory system to test the effects of external sensory input on diameter of vessels with and without smooth muscle expression in the awake mouse somatosensory cortex (Figure 6). After cranial window surgeries mice were allowed to recover from anesthesia before two-photon imaging and sensory stimulation. Dextran was injected intravenously to label blood vessels and was used to determine changes in vessel diameter described in data analysis. Whisker stimulation was done in awake head fixed mice using a small tube attached to a picospritzer (set to 30PSI) positioned 1–2 cm from the contralateral vibrissae. Stimulation consisted of 100ms air puffs at 5Hz for 2 or 30 seconds as indicated.
To image vasomotor responses during cortical spreading depolarization (CSD) (Figure 7) 0.8mM Oregon Green 488 BAPTA-1 AM (OGB) (Life Technologies cat# O-6807) was pressure injected through a pulled glass micropipette. To induce CSD, 0.5mM KCl was injected 1–2mm away from the imaging location. Time-lapse imaging was performed before, during, and after the CSD induction in order to detect the change in calcium in neurons and astrocytes with OGB and correlate these changes with smooth muscle-labeled and non-labeled vessel diameter changes with intravascular dye.
Focal ischemia was induced by transient (90 min) middle cerebral artery occlusion (tMCAO) using an intraluminal filament technique. The carotid artery was exposed in the neck region and a silicon rubber-coated 6-0 nylon monofilament with 2 mm tip (~ 230 μm diameter of the tip) was introduced into the common carotid artery and advanced 9 mm along the internal carotid artery (ICA) until occluding the origin of the MCA. After 90 min, the filament was withdrawn to establish reperfusion. Prior to MCAO, a cranial window was prepared and imaging was performed using a two photon or confocal microscope. Imaging was performed before, during, immediately after and 4 hours after tMCAO. Vessel perfusion at each stage was determined via detection of RBC flux using i.v. dyes.
We thank K. Hirschi (Yale University) for sharing SMA-mCherry transgenic mice; D. Greif (Yale University) for sharing SMAcreER mice; A. Nishiyama (University of Connecticut) for sharing NG2cre mice; Z.J. Zhou (Yale University) for sharing ChR2-YFP and GCaMP3 mice and M. Mesulam (Alzheimer Disease Center at Northwestern University) for providing postmortem human tissue. This work was supported by the following grants from the National Institutes of Health: R01-HL106815 and R01-NS089734 to J.G, and F32-NS090820 to R.A.H.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Author ContributionsR.A.H., L.T. and P.Y. designed and performed experiments and analyzed data. S.M. and S.G. performed experiments. R.A.H. and J.G. wrote the manuscript. J.G. designed and supervised the experiments.