K_ATP channel–dependent electrical signaling links capillary pericytes to arterioles during neurovascular coupling

Significance

Precise blood flow control to meet neuronal metabolic needs is essential for brain health and it has recently emerged that capillaries play a vital role in this process. Thin-strand pericytes adorn the deepest reaches of the capillary bed, but their role in blood flow control is poorly understood. Isaacs et al. combine optogenetics and in vivo imaging to demonstrate that thin-strand pericytes play a major role in sensing neural activity and generating electrical signals that are transmitted upstream to regulate arteriole diameter and local blood flow, and therefore energy delivery to neurons.

Abstract

The brain has evolved mechanisms to dynamically modify blood flow, enabling the timely delivery of energy substrates in response to local metabolic demands. Several such neurovascular coupling (NVC) mechanisms have been identified, but the vascular signal transduction and transmission mechanisms that enable dilation of penetrating arterioles (PAs) remote from sites of increased neuronal activity are unclear. Given the exponential relationship between vessel diameter and blood flow, tight control of arteriole membrane potential and diameter is a crucial aspect of NVC. Recent evidence suggests that capillaries play a major role in sensing neural activity and transmitting signals to modify the diameter of upstream vessels. Thin-strand pericyte cell bodies and processes cover around 90% of the capillary bed, and here we show that these cells play a central role in sensing neural activity and generating and relaying electrical signals to arterioles. We identify a K_ATP channel–dependent neurovascular signaling pathway that is explained by the recruitment of thin-strand pericytes and we deploy vascular optogenetics to show that currents generated in individual thin-strand pericytes are sent over long distances to upstream arterioles to cause dilations in vivo. Genetic disruption of vascular K_ATP channels reduces the arteriole diameter response to neural activity and laser ablation of thin-strand pericytes eliminates the K_ATP-dependent component of NVC. Together, our findings indicate that thin-strand pericytes sense neural activity and transform this into K_ATP channel–dependent electrometabolic signals that inform upstream arterioles of local energy needs, promoting spatiotemporally precise energy distribution.

Due to the susceptibility of neurons to injury following metabolic deprivation, neurons and astrocytes communicate their fluctuating energy needs to nearby vascular structures to control local blood flow—a process known as neurovascular coupling (NVC). Several NVC mediators have been identified (1), but we lack consensus on the vascular transduction and transmission mechanisms through which neural signaling regulates the contractile cells that control local blood flow. These are primarily the smooth muscle cells (SMCs) of pial and penetrating arterioles (PAs) and the highly contractile ensheathing pericytes of the proximal branches of the capillary bed (1st- to ~4th-order capillaries) (2, 3). These cells modulate blood flow through either relaxation to permit vessel dilation or contraction to reduce lumen diameter. Recent work also indicates that thin-strand pericytes of the deep capillary bed (~5th-order capillaries and above) contract over longer time scales to modulate blood flow (4, 5).

We and others have suggested that capillaries play a major role in signaling to control blood flow (6–8). Capillary endothelial cell (EC) inward rectifier K⁺ (Kir) channels can generate membrane hyperpolarization to elevated external K⁺, which is then conducted upstream to the feeding PA SMCs (7). This dilates the PA and increases red blood cell (RBC) flux and velocity in capillaries. These findings redefined the capillary bed as a distributed sensor array embedded in the computing parenchyma, capable of relaying hyperemic signals to upstream SMCs and ensheathing pericytes to augment brain perfusion on a moment-to-moment basis. Thin-strand pericytes of the deep capillary bed are ideally positioned to participate in this process, and express a rich repertoire of membrane proteins that could enable them to fulfill an activity sensing and signal generation role (9).

Thin-strand pericytes can be distinguished by their minimal expression of alpha smooth muscle actin, and their strand-like processes that stretch along the capillary wall. These cells play a trophic role in maintaining blood–brain barrier integrity (10), but their role in blood flow control is much less understood. Experiments with depolarizing optogenetic actuators have shown that activating channelrhodopsin 2 (ChR2) in thin-strand pericytes can constrict capillaries and modify blood flow (4, 5). While these findings support an active role of thin-strand pericytes in blood flow modulation, not all studies have concluded that thin-strand pericytes are able to constrict their underlying capillary (11) and a number of studies failed to observe dilation of capillaries under thin-strand pericytes during neuronal activity (12). This raises questions as to whether, and how, thin-strand pericytes contribute to blood flow control in the physiological context.

Thin-strand pericyte architecture may enable control of blood flow via electrical coupling to the endothelium. Ultrastructural studies have identified “peg-socket” junctions—extensions of the pericyte or EC membrane that make direct contact with adjacent cells that are thought to be sites of gap junctions (GJs) (13). This suggests that electrical transmission from pericytes is, in principle, possible through the CNS vasculature, but a direct demonstration of this principle in vivo is lacking. Here, we show that 1) single thin-strand pericytes are electrically coupled into the brain vasculature and can remotely control PA diameter and blood flow, and; 2) the hyperpolarizing toolkit of thin-strand pericytes transduces neuronal activation into K_ATP channel-mediated electrical signals that rapidly evoke hyperemia. Thus, we reconceptualize thin-strand pericytes as regulators of flow throughout local microvascular networks via their ability to dilate PAs through long-range electrical signaling.

Results

A Mouse Model to Study Pericyte Electrical Signaling.

We crossed platelet-derived growth factor receptor β (PDGFRβ)-cre mice with mice harboring a STOP-floxed ArchT-enhanced green fluorescent protein (EGFP) fusion construct (Fig. 1A). ArchT is a mutant archaerhodopsin with improved light sensitivity (14) that pumps protons (H⁺) out of the cell in response to illumination, thereby generating membrane hyperpolarization. We reasoned that this could be precisely activated using a single-photon point-scanning approach in vivo and this, combined with multiphoton imaging, would allow us to control and measure PA diameter and capillary blood flow in the intact brain (Fig. 1B). In these mice, we observed ArchT-EGFP expression in SMCs and pericytes on all vascular branch orders (Fig. 1 C–E and Movie S1). To validate this approach for generating hyperpolarizing pericyte membrane currents, we isolated thin-strand pericytes and performed patch clamp recordings which confirmed ArchT-EGFP was functional. Exposure to a 530 ± 15 nm LED produced reversible outward currents of 7.6 ± 1.3 pA (n = 5 cells, 3 mice) that were time-locked to light stimulation. In contrast, whole-cell currents measured in pericytes expressing only a fluorescent protein (PDGFRβ-tdTomato) exhibited no response to light application (SI Appendix, Extended Data Fig. 1). Thus, we moved into intact animals with the goal of ascertaining whether electrical signals generated in individual pericytes could influence upstream PA SMC contractile status.

Fig. 1.

PDGFRβ-ArchT-EGFP mice. (A) Crossing mice heterozygous for cre-recombinase under the control of Pdgfrb with mice expressing a STOP-floxed ArchT-EGFP fusion protein allows expression of this hyperpolarizing opsin in mural cells. (B) PDGFRβ-ArchT-EGFP mice were studied using a point-scanning galvo–galvo confocal microscope fitted with a femtosecond laser and nondescanned detectors allowing for spatially precise optical activation interleaved with multiphoton imaging. (C) A central parenchymal arteriole (PA) giving rise to a capillary network (1st to 5th branch orders labeled). SMCs and pericytes express EGFP (green) surrounding Texas Red Dextran-loaded vasculature (red). (D) Deep capillaries (5th order and above) exhibit characteristic “bump-on-a-log” thin-strand pericytes. (E) Thin-strand pericyte processes (P) running along the length of the capillaries with a nanotube (NT) crossing the parenchyma between two capillaries.

Activation of ArchT in Pial Artery and PA SMCs Drives Dilations.

We initially focused on optically stimulating ArchT in SMCs in pial arteries. These experiments enabled us to validate whether hyperpolarization of pial SMCs dilated the vessel, and allowed us to observe spreading effects on untargeted vessel segments of branching arteries. We first identified bifurcating pial arteries on the brain surface giving rise to PAs diving into the parenchyma within several hundred microns of one another (SI Appendix, Extended Data Fig. 2A). To measure arterial diameter while activating ArchT, we interleaved 2-photon laser scanning microscopy (2PLSM) with 561 nm laser point-scanning over a region of interest (ROI). ArchT activation in SMCs on one pial branch caused a large dilation of the targeted segment compared to the connected, nontargeted pial branch (SI Appendix, Extended Data Fig. 2 B–D). In contrast, sham stimulation of the parenchyma evoked no dilatory responses (SI Appendix, Extended Data Fig. 2E). ArchT stimulation evoked a ~12% dilation of the target artery (n = 7 pial arteries from 3 mice; SI Appendix, Extended Data Fig. 2F). Interestingly, we noticed a clear emergence of dilations in distal artery segments also. These untargeted segments dilated by ~4% on average, which was larger than the ~1% dilation elicited by sham stimulation (SI Appendix, Extended Data Fig. 2G). These data are most easily explained by ArchT photocurrent being injected from the targeted SMCs into the endothelium via myoendothelial GJs (15) which then axially spreads along the length of the vessel to relax SMCs in the distal vessel segment. When we analyzed dilation magnitude as a function of distance along the vasculature, we observed only a minimal decrease over hundreds of microns (SI Appendix, Extended Data Fig. 3), which is also in good agreement with previous observations in retinal arterioles (16).

Moving deeper into the cortex (SI Appendix, Extended Data Fig. 4A), we found that PAs could also be reliably dilated in a stereotyped manner using our approach (SI Appendix, Extended Data Fig. 4B). PA dilation by SMC ArchT activation was “light dose” dependent, with the dilatory effect saturating by ~8 ArchT activation cycles that ranged from 2 to 3 s each (mean dilation after 8 cycles = 13.2 ± 2.7%, n = 12 PAs from 8 mice; SI Appendix, Extended Data Fig. 4C). By performing a single, long activation and following PA diameter over a longer period postirradiation (SI Appendix, Extended Data Fig. 4D), we ascertained that dilations were reset by homeostatic mechanisms within 30 s (average diameter at 30 s vs. baseline: −0.54 ± 2.34%, n = 4 arterioles, 3 mice; SI Appendix, Extended Data Fig. 4E). Because others have reported that light may energize proteins capable of driving arteriole vasomotion (17), we assessed whether 561 nm laser-scanning contributed to PA dilation using PDGFRβ-TdTomato mice (SI Appendix, Extended Data Fig. 4F). Importantly, no PA dilations occurred when applying an identical imaging and activation scheme in these experiments (SI Appendix, Extended Data Fig. 4G). Overall, laser control and sham irradiation had no effects on PA diameter, whereas the on-target PA SMC ArchT activations produced consistent dilations (SI Appendix, Extended Data Fig. 4H). Together, these experiments confirmed that optically generated H⁺ currents dilate arterioles and demonstrated precise spatial control using our 561 nm point-scanning approach.

Thin-Strand Pericyte Hyperpolarization Dilates Upstream Arterioles.

We next turned our attention to the deep capillary bed to ask whether hyperpolarizing currents generated in thin-strand pericytes could drive PA dilation (Fig. 2 A–C). We restricted these experiments to the upper ~100 µm of the cortex, given that 1) visible light is scattered by brain tissue, and 2) the “cone” of focused laser light could produce off-target effects in overlying cells if the focus was positioned too deeply (SI Appendix). As an additional stipulation, we also sought to target thin-strand pericytes whose soma resided on 5th- and 6th-order capillary branches downstream of the imaged PA, which are agreed to be distinct from ensheathing pericytes but were not so far away from the PA that additional branch points might overly “dilute” the possibly subtle input from a single cell in the capillary bed. We located 12 such pericytes across 7 mice and found that 561 nm point scanning to hyperpolarize individual pericytes dilated the upstream feeding PA (Fig. 2 D and E and SI Appendix, Extended Data Fig. 5, and Movie S2), resulting in a maximal diameter change of 4.98 ± 0.81% on average (n = 12 pericytes from 7 mice). In stark contrast, sham scanning the parenchyma using the same parameters dilated the PA by 0.96 ± 0.49% on average (n = 12 sham stimulations from 7 mice; Movie S3). In absolute units, pericyte stimulation dilated the upstream PA by 0.83 ± 0.12 µm on average, while sham stimulation dilated the upstream PA by 0.16 ± 0.10 µm (SI Appendix, Extended Data Fig. 6).

Fig. 2.

Activation of ArchT thin-strand pericytes increases arteriolar diameter. (A) A PA and downstream and capillary branches in a PDGFRβ-ArchT-EGFP mouse. (B) Single imaging plane from A. (C) Experimental approach. (D) 561 nm activation in a 5th-6th order thin-strand pericyte dilated the upstream PA whereas sham stimulation was without effect. (E) Mean diameter change after each ArchT-EGFP activation period for every experiment (n = 12 pericyte-sham pairs, 7 mice). (F) Maximal dilation for every pericyte/sham pair (n = 12 pericyte/sham pairs, 7 mice; paired t-test, t₁₁ = 4.350, P = 0.0012). (G) Experimental strategy to elicit PA dilation via ArchT activation in thin-strand pericytes and measure flux and velocity in the underlying capillary. Inset: PA dilation evoked by scanning an ArchT-EGFP+ thin-strand pericyte (orange; vs. sham, black). (H) An imaging field containing an arteriole with a downstream 5th order capillary and pericyte. The pericyte was stimulated by 561 nm scanning interleaved with blood flow data collection by line scanning the capillary. (I) RBC kymograph from a PDGFRβ-ArchT-EGFP capillary at baseline and after 5 561 nm ArchT activation cycles. (J) Increase in average RBC flux after 561 nm laser scanning in PDGFRβ-ArchT-EGFP animals (+4.0 RBC/s, n = 36 line scans, 15 mice; Wilcoxon rank-sum test, W = 355, P = 0.0029). (K) RBC flux vs. ArchT cycles – note that we excluded 5 (of 36 total) vessels which exhibited a stall. (L) Change in RBC flux relative to baseline after 4-5 ArchT activation periods (n = 15 ArchT-EGFP+ mice (31 capillaries), n = 3 control mice (17 capillaries); Welch’s t-test t_7.196 = 3.084, P = 0.0171). (M) Change in capillary RBC velocity relative to baseline after 4-5 ArchT activation periods (n = 15 ArchT-EGFP+ mice (31 capillaries), n = 3 control mice (17 capillaries); Welch’s t-test t_14.01 = 2.488, P = 0.0261).

As ArchT is a H⁺ pump, acidification resulting from external H⁺ accumulation might explain the upstream PA responses we observed. During NVC, the extent of parenchymal acidification is on the order of 0.1 pH units (18). Similarly, when completely inhibiting carbonic anhydrase in the parenchyma acidification of 0.05 - 0.2 pH units occurs (19). Thus, we chose to acidify the parenchyma around a pericyte by ~0.2 pH units to mimic this physiological maximum. While we cannot rule out effects of more extreme (and potentially pathological) pH changes on pericytes, these experiments suggested that physiological acidification around thin-strand pericytes does not remotely drive arteriole dilations. We determined this by simulating physiological H⁺ accumulation by ejecting acidified aCSF (pH 7.1) onto thin-strand pericytes (SI Appendix, Extended Data Fig. 7 A–C). This procedure caused fluctuations in arteriole diameter more than application of neutral aCSF (pH 7.3) but did not consistently cause dilations (SI Appendix, Extended Data Fig. 7D) suggesting that H⁺ accumulation around the pericyte under study does not explain the observed effects of ArchT stimulation on PA diameter (SI Appendix, Extended Data Fig. 7E).

These data thus support a model in which pericytes are capable of electrically controlling the contractile apparatus of upstream SMCs. According to our data, a single pericyte is able to maximally dilate an arteriole by about 4% of its resting diameter (i.e. the mean difference between paired pericyte vs. sham dilations). These findings prompted us to compare the magnitude of PA dilation evoked by thin-strand pericyte stimulation to our pial artery data that also indicated that hyperpolarization can be conducted along the artery (SI Appendix, Extended Data Fig. 8). In both cases, the conducted dilations were similar, suggesting they may be under the control of a common mechanism.

We previously demonstrated that Kir2.1 channels in capillary ECs play an important role in conducting electrical signals to upstream arterioles (7). Notably, these channels can be activated by hyperpolarization, which relieves voltage-dependent block of the channel and thereby permits K⁺ currents out of the cell (8). Thus, hyperpolarization originating in thin-strand pericytes could be passed into ECs to activate Kir2.1, which might then aid in transmitting the signal to PA SMCs. To test this possibility, we blocked Kir2.1 channels with 100 µM Ba²⁺, which almost eliminated PA dilations to thin-strand pericyte ArchT activation but had no effect of PA dilations to direct SMC hyperpolarization by ArchT (SI Appendix, Extended Data Fig. 9). These data thus support a role for capillary EC Kir2.1 channels in transmission of thin-strand pericyte-generated electrical signals.

ArchT Activation in SMCs and Thin-Strand Pericytes Controls Capillary Blood Flow.

Consistent with our observations of control of PA diameter by ArchT, capillary blood flow was also reliably increased by driving photocurrents directly in PA SMCs. We used line scanning to measure RBC flux and velocity (SI Appendix, Extended Data Fig. 4 I–K). In measurements taken from the 5th and 6th capillary branch orders, both velocity and flux increased during direct PA SMC hyperpolarization (SI Appendix, Extended Data Fig. 4 L–N). Intriguingly, all capillaries we studied exhibited robust increases in flux initially, but 4/5 of these initial flux increases waned with continued PA-SMC ArchT activation (SI Appendix, Extended Data Fig. 4 L and M). The reason for this curtailment is presently unclear, but one possibility is the existence of mechanisms to dampen blood flow to the capillary bed to match the energy demands of local neural tissue, which are expected to be lower in this paradigm as the neurons are not directly stimulated. We next targeted thin-strand pericytes and measured the expected blood flow response to upstream dilation (Fig. 2 G–I). Indeed, activating ArchT-EGFP in thin-strand pericytes increased both capillary RBC velocity and flux by ~16% and ~13%, respectively (Fig. 2 J–M). This effect was the direct result of activating ArchT in pericytes, as 561 nm laser scanning on PDGFRβ-tdTomato pericytes resulted in no deviation in RBC flux and only a ~2% increase in velocity (Fig. 2 L and M and SI Appendix, Extended Data Fig. 10). These data thus strongly support the conclusion that hyperpolarizing electrical signals generated in an individual pericyte result in local capillary hyperemia.

Given that activation of ChR2 evokes constriction of thin-strand pericytes and substantially reduces capillary blood flow (4, 5), we asked whether direct capillary dilation due to activation of ArchT could explain our blood flow data (SI Appendix, Extended Data Fig. 11). This analysis showed that capillary dilation could not readily explain the effects on blood flow that we typically observed after 4 to 5 ArchT activation periods playing out over ~20 to 30 s (SI Appendix, Extended Data Fig. 11B). However, we noted when applying our interleaved ArchT activation approach for ~2 min that smaller diameter capillaries reliably dilated by about ~130 nm (SI Appendix, Extended Data Fig. 11 C and D). Least squares regression showed that starting capillary diameter could predict the dilation response to ArchT extremely well, with smaller capillaries undergoing uniform dilations of ~5% whereas no change occurred in larger diameter capillaries (SI Appendix, Extended Data Fig. 11 D–F). As these subtle capillary diameter changes appeared long after the effects we observed on arteriole diameter and capillary blood flow, these are unlikely to contribute to the more immediate blood flow responses we observed to thin-strand pericyte ArchT activation (Fig. 2 G–M and SI Appendix, Extended Data Fig. 5). However, these data raise the intriguing possibility of hyperpolarization-induced thin-strand pericyte relaxations modifying capillary diameter and blood flow over longer periods.

Bringing these data together, the increase in RBC flux and velocity observed after ArchT activation in thin-strand pericytes is most easily explained by a conducted hyperpolarization that dilates the upstream PA. However, we wished to rule out off-target effects on the opsin which might result from inadvertent activation of ArchT in overlying PA SMCs by out-of-focus light. Thus, we sought an approach to limit expression of ArchT to individual thin-strand pericytes to better isolate their contributions to upstream arteriolar vasomotion.

Electrical Signals from Individual ArchT+ Thin-Strand Pericytes Control PA Diameter and Capillary Blood Flow.

The conical volume of 561 nm light used in our ArchT-EGFP stimulation approach makes it impossible to entirely exclude an influence of other ArchT-EGFP+ structures from our results in mice that widely express this opsin. Therefore, we chose to pursue a conditional expression approach to greatly constrain the expression of ArchT-EGFP and essentially eliminate off-target effects. This approach has the added benefit of ameliorating any developmental and systemic effects that may be associated with expressing ArchT-EGFP widely. To achieve this, we sparsely labeled pericytes using a tamoxifen-controlled cre-lox system. Crossing NG2-creER mice with floxed ArchT-EGFP mice installed a conditionally expressible ArchT-EGFP in NG2-positive cells (i.e. SMCs and pericytes). Thereafter, exposure to 10 mg/kg 4-hydroxytamoxifen (4-OHT) drove cre-dependent STOP-lox excision and expression of the ArchT-GFP fusion protein. By titrating both the administration time (4 d) and the time allowed for recombination (~2 to 3 wk), we were able to greatly constrain ArchT expression, and on average these mice had far less ArchT-EGFP in the entire visible vasculature (SI Appendix, Extended Data Fig. 12). In our initial preparation, we studied the only labeled pericyte we could find, with its soma on a 3rd order branch. As a provisional test, we found that activation of ArchT in this cell produced arteriole dilation ~160 µm upstream (along with dilation of the upstream 2nd order capillary, while downstream branches did not change diameter; SI Appendix, Extended Data Fig. 13), providing proof-of-principle that this approach would be viable for studying pericyte control of PA diameter. To then test our hypothesis that single thin-strand pericytes on deep capillaries are electrically coupled to the feeding PA, we sought vascular trees with just 1 to 2 ArchT-EGFP labeled thin-strand pericytes (Fig. 3 A–E and SI Appendix, Extended Data Fig. 14). We found that we could dilate arterioles by activating ArchT within these single pericytes, but not when scanning the PA or the parenchyma, confirming our ability to control arteriole diameter exclusively via pericyte-to-arteriole electrical communication (Fig. 3 F and G and Movie S4). We also assessed dilation kinetics evoked by individual pericyte stimulation via least squares regression of our PA diameter time series data. Single pericyte stimulation dilated arterioles by ~16 nm/s on average over the entire time course. Although these responses were subtle, multiple pericytes from the same capillary bed likely feed current to the arteriole in a physiological network, as opposed to the situation here where sustained inputs from a single distal pericyte evoke a modest to large dilation of the PA over longer periods. Intriguingly, we found that the average dilation kinetics were well explained by the vascular path length from the pericyte soma to the PA (R² = 0.668, P = 0.013; Fig. 3H). Importantly, sham dilation magnitude was not predicted by ROI distance from the PA under study, suggesting off-target effects were minimal and dilations occurring in sham experiments were likely to be the result of unrelated vasomotion (SI Appendix, Extended Data Fig. 14D).

Fig. 3.

Isolated thin-strand pericyte ArchT activation evokes PA dilation. (A) An isolated thin-strand ArchT-EGFP+ pericyte in an NG2creER-ArchT-EGFP mouse. Stimulation ROIs are depicted as hatched boxes. (B and C) Detail of this pericyte residing on a 5th-order capillary branch. (D) Experimental approach. (E) Mean intensity of EGFP in pericytes, PA-SMCs, sham ROIs, or a randomly placed ROI. (F) Single trial showing arteriole dilation in response to pericyte 561 nm scanning. Also shown is an experiment in the same mouse with a sham 561 nm scan ROI, or the ROI placed on the PA-SMC itself. Circles show data points used in the max dilation calculation for G. (G) Summary statistics for the maximum evoked PA diameter change during ArchT-EGFP activation (n = 5 PA-sham ROIs, 5 PA-pericyte ROIs, 4 PA-PA-SMC ROIs, 4 animals; one-way ANOVA, F(2, 11) = 5.492, P = 0.022, Dunnett’s multiple comparison P values on the graph). (H) The path length connecting the PA to the pericyte under study predicted the dilation kinetics (least squares linear regression R² = 0.668, P = 0.013). (I) To determine whether isolated pericytes could drive blood flow changes we took line scans on capillary segments with ArchT-EGFP+ pericytes and applied our interleaved imaging/activation approach. (J) In controls, we performed the same procedure on capillaries without a labeled pericyte. (K) Blood flow followed a similar trajectory to arteriole dilation. (L) Blood flow uniformly increased when 561 nm scan ROIs were on a pericyte, but not when the ROI contained only a capillary segment (Welch’s t test, t_12.81 = 6.752, P < 0.0001).

Finally, we found that RBC flux could be causally increased by hyperpolarizing ArchT-EGFP+ thin-strand pericytes in these mice (Fig. 3 I–L). This effect was due to ArchT-EGFP+ pericyte engagement, as scanning capillaries lacking an ArchT-EGFP+ pericyte did not evoke reliable flux increases. RBC velocity increases accompanied flux increases, where after 4 to 5 ArchT stimulation cycles a ~30% increase in velocity was observed which did not occur when pericytes lacking ArchT were targeted in line scans (SI Appendix, Extended Data Fig. 14 G–J). Altogether, these experiments argue strongly that individual pericytes can exert profound effects on capillary blood flow via dilation of the upstream PA.

Thin-Strand Pericyte K_ATP Channels are Necessary for NVC Initiated by Excitatory Neurons.

Given that our data indicated that individual thin-strand pericyte hyperpolarizations can control upstream PA diameter and blood flow, we wondered how such hyperpolarizations might be generated in the physiological context, and whether these could contribute to rapid hyperemia elicited by NVC. Recent work from our laboratory suggests that a key hyperpolarizing tool employed by thin-strand pericytes to control PA diameter and blood flow is the K_ATP channel (20). To confirm and extend these findings, we pressure-ejected the K_ATP channel activator pinacidil (10 µM) onto Neurotrace 500/525-labeled thin-strand pericytes (21) and observed Ca²⁺ dynamics and diameter in upstream PAs in mice expressing the Ca²⁺ biosensor GCaMP8.1 under the Acta2 promoter (Fig. 4 A and B). This resulted in a reduction of the GCaMP8.1 fluorescence of −19.8 ± 3.7% in SMCs of the upstream PA (n = 4 experiments, 4 mice; Fig. 4 C–F and SI Appendix, Extended Data Fig. 15) with an onset range of 5 to 10 s. This was followed by dilation of the PA by 0.95 ± 0.20 µm on average (Fig. 4G). We next wondered whether acute neural activation evokes hyperemia via a mechanism that engages thin-strand pericyte K_ATP channels. We therefore deployed mice expressing ChR2 fused to yellow fluorescent protein (YFP) under the Thy1 promoter to drive excitatory neuronal activity in the somatosensory cortex. To model NVC, we excited dendrites in mice anesthetized with urethane and alpha-chloralose by point scanning a 488 nm laser over a ~25,000 to 30,000 µm² region of the cortex (Fig. 4 H−O). Centering a ROI over a PA and delivering 10 iterative scans produced a highly stereotyped arteriolar dilation of 16.5 ± 2.3% (n = 8 arterioles, 3 mice) similar to the magnitude of dilatory responses evoked by whisker stimulation paradigms (22). We noticed two distinct phases in the PA dilation response to this approach. An initial prominent dilation occurred within 5 s of the ChR2 stimulus, and occasionally a later, smaller dilation phase appeared (~15 s after the stimulus offset; Fig. 4I). We sought to determine whether K_ATP channels might mediate some or all of this response. To do so, we included 20 µM glibenclamide in the agarose above the cranial window, reasoning that this saturating concentration would allow drug to leach into the cortex and maximally block local K_ATP channels, which have nanomolar affinity for glibenclamide. This procedure did not change the resting diameter of arterioles on average (Fig. 4 K and M) in line with our previous findings (20). However, glibenclamide dramatically suppressed the initial phase of the NVC response to ChR2 activation by ~40%, with no effect on the late phase (Fig. 4 L and O). As a result of the loss of this initial phase, the time to peak dilation was extended for PAs in preparations treated with glibenclamide (Fig. 4N). Vehicle [0.2% dimethylsulfoxide (DMSO) in artificial cerebrospinal fluid; aCSF]-containing agar had no effects (Fig. 4 K, N, and O). Combined, these data indicate that K_ATP channel activity is necessary for NVC initiated by excitatory pyramidal neurons and confirm that activation of this channel in capillary pericytes drives vasodilation of PAs in vivo by modulating membrane voltage and Ca²⁺ entry in SMCs. This led us to question whether thin-strand pericyte K_ATP channels are the locus of K_ATP channel contributions to NVC.

Fig. 4.

K_ATP channels contribute to NVC. (A) Experimental approach to activate K_ATP channels by ejecting 10 µM pinacidil onto Neurotrace 500/525-stained thin-strand pericytes in mice expressing GCaMP8.1 under the Acta2 promoter. (B) A pipette filled with 10 µM pinacidil next to a Neurotrace 500/525+ pericyte in an Acta2-GCaMP8.1 mouse. White line: shortest path to the upstream PA. (C) PA diameter before and after pinacidil or DMSO ejection onto a downstream thin-strand pericyte. White outline: baseline diameter. Arrow: Ca²⁺ decrease after K_ATP activation. (D) Time course showing the relative change in SMC GCaMP8.1 fluorescence and arteriole diameter occurring after application of pinacidil to a thin-strand pericyte. (E) Absolute change in GCaMP8.1 fluorescence before and after pinacidil (n = 4; paired ratio t-test, t₃ = 4.716, P = 0.0181). (F) Percent change in GCaMP8.1 fluorescence before and after application of pinacidil vs. baseline (n = 4). (G) Absolute PA diameter change following pinacidil application to a downstream pericyte (Student’s paired t-test, t₃ = 4.608, P = 0.0192). (H) Cartoon illustrating how pericyte K_ATP channels might contribute to activity dependent arteriole dilation. We tested this hypothesis by exciting neuronal ChR2 using 488 nm light and blocking K_ATP activity with glibenclamide. (I) PAs dilated following ChR2 excitation in Thy1-ChR2-EYFP+ mice (n = 8 PAs, 3 mice). Grey: population average. Yellow: a PA which dilates robustly in the initial phase; pink: a PA which has distinct early and late phases. (J) Comparison of the peak diameter change during the initial phase vs. the late phase of the NVC response (n = 8 PAs, 3 mice; paired Student’s t-test, t₇ = 3.56, P = 0.0092). (K) In a craniotomy sealed by cover glass preparation we included 20 µM glibenclamide in the agarose under the window, which had no effect on PA diameter (25 DMSO PAs, 4 mice, 33 glibenclamide PAs, 6 mice; Mann-Whitney U test, U = 373, P = 0.5432). (L) Time course of PA diameter changes before and after stimulation of neuronal ChR2 when either DMSO or glibenclamide was included in the agarose (n = 25 DMSO PAs, 4 mice, 33 glibenclamide PAs, 6 mice). (M) Example images showing raster scan areas for ChR2 activation in a DMSO-treated or glibenclamide-treated preparations. (N) The time to max dilation was significantly shorter for DMSO-treated PAs than glibenclamide treated PAs (n = 25 DMSO PAs, 4 mice, 33 glibenclamide PAs, 6 mice, Mann-Whitney U test, U = 261.5, P = 0.0171). (O) Glibenclamide blunted the amplitude of the initial phase but did not affect the late phase of the response (n = 25 DMSO PAs, 4 mice, 33 glibenclamide PAs, 6 mice; Two-way repeated measures ANOVA, interaction of NVC phase and treatment F(1, 56) = 41.13, P <0.0001), Šídák’s multiple comparisons test P values on graph).

Dominant-Negative K_ATP Channel Mutation in Mural Cells Impairs NVC.

To address this question directly, we sought to genetically disrupt K_ATP channels in thin-strand pericytes and assay NVC. We utilized a previously described mouse model of K_ATP channel loss-of-function in which an NG2-creER line was crossed with a Kir6.1 mutant mouse line (known as Kir6.1^AAA) in which an alanine triplet replaces key residues of the selectivity filter of the pore-forming subunit of the vascular K_ATP channel, rendering channels incorporating this subunit nonfunctional (20). This line has an EGFP sequence preceded by a STOP codon, both of which are flanked by LoxP sites. This is upstream of Kir6.1^AAA, meaning that EGFP is constitutively expressed by all cells until exposure to cre-recombinase induced by 4-OHT, which results in excision of the EGFP sequence and the STOP codon. This permits transcription of the dominant-negative mutant K_ATP channel subunit, causing loss of K_ATP channel function along with loss of EGFP signal (Fig. 5 A and B and SI Appendix, Extended Data Fig. 16). PDGFRβ-tdTomato mice acted as controls for pericyte density assessment as a function of capillary length, and we found that exposure to 4-OHT dissolved in corn oil (10 mg/kg i.p. once daily) for 5 d with 5 wk for recombination resulted in a 41% reduction in the number of pericyte soma per cortical capillary length relative to vehicle (Fig. 5C). Concentration and penetration issues with 4-OHT likely prevented more widespread recombination. To utilize our optogenetic NVC model in this context, we injected mice with AAV-CaMkiiα-ChR2-EYFP to express ChR2 in excitatory neurons in the cortex 1 wk ahead of the onset of 4-OHT injections. 5 to 6 wk after the last 4-OHT injection, we performed acute cranial window surgeries and imaging. Following ChR2 activation, we found that the initial PA dilation amplitude was reduced by ~25% in mice in which pericyte and SMC K_ATP channels were rendered nonfunctional, but dilation kinetics were no different (Fig. 5 D−F). Again, we noticed that the early aspects of the dilation seemed to be the most impacted (Fig. 5 D and F). However, our optogenetic approach is limited by the inability to image while we stimulate, and anesthesia is known to perturb the NVC response. Thus, to test our hypothesis in a more physiological context we moved into an awake model of NVC which would allow us to image concurrently with sensory stimulation, and study the stereotyped biphasic arteriole response reported by others (23). In these experiments, we induced dominant negative expression of the Kir6.1^AAA transgene under the NG2 promoter using the same scheme as above. One week after the last injection, we installed a window (Fig. 5G) and habituated mice to head fixation. Contralateral whiskers were stimulated via an air puff for a total of 8.5 s at a frequency of 1 Hz with a duty cycle of 50%. All awake mice exhibited dilations of PAs in response to whisker stimulation (Fig. 5H and Moive 5). The kinetics were unaffected by 4-OHT treatment (Fig. 5I). However, the arterioles in 4-OHT-treated mice exhibited a reduced initial phase of NVC (Fig. 5J). In this context, the late phase dilation (from 5 s to the end of the stimulus) was also reduced (SI Appendix, Extended Data Fig. 17A). 4-OHT treatment alone did not explain these effects (SI Appendix, Extended Data Fig. 17B).

Fig. 5.

K_ATP channels in NG2+ mural cells contribute to NVC. (A) An EGFP-Kir6.1^AAA transgenic mouse treated with 4-OHT, but lacking expression of cre recombinase under the NG2 promoter. EGFP (green) is expressed widely and can be seen throughout pericytes on capillaries (red). (B) A NG2creER-EGFP-Kir6.1^AAA mouse treated with 4-OHT. Successful recombination results in excision of the STOP-floxed EGFP construct leading to loss of EGFP signal and expression of Kir6.1^AAA, indicated by fewer EGFP+ pericytes (green) on capillaries (red). (C) Quantification of EGFP+ pericyte soma per 100 µm of ≥5th-order capillary path length (One-way ANOVA, n = 3 capillary beds, 3 PDGFRβ-tdTomato mice, 4 capillary beds, 4 vehicle treated (-4-OHT) NG2creER-EGFP-Kir6.1^AAA mice, and 9 capillary beds, 6 4-OHT-treated NG2creER-EGFP-Kir6.1^AAA mice; One-way ANOVA, F(2, 13) = 3.82, P = 0.0483, P values from Tukey’s multiple comparisons test on graph). (D) NVC response to ChR2 activation of CaMKiiα+ transfected excitatory neurons (n = 14 arterioles, 3 vehicle-treated NG2creER-EGFP-Kir6.1^AAA mice, 26 arterioles, 8 4-OHT–treated NG2creER-EGFP-Kir6.1^AAA mice). (E) Summary data. The kinetics of PA dilation to ChR2 stimulation were similar in mice receiving either vehicle or 4-OHT (n = 14 arterioles, 3 vehicle-treated NG2creER-EGFP-Kir6.1^AAA mice, n = 26 arterioles, 8 4-OHT–treated NG2creER-EGFP-Kir6.1^AAA mice; Mann–Whitney U test, U = 162.5, P = 0.5893). (F) The initial phase of dilation of PAs dilation to ChR2 was blunted by K_ATP inactivation in pericytes and SMCs (n = 14 arterioles, 3 vehicle-treated NG2creER-EGFP-Kir6.1^AAA mice, n = 26 arterioles, 8 4-OHT–treated NG2creER-EGFP-Kir6.1^AAA mice; Welch’s t test t_29.77 =2.178, P = 0.0375). (G) Top: Schematic for awake, head-restrained 2PLSM imaging. Bottom: Maximum intensity projection showing EGFP labeled pericytes and SMCs (green) on the vasculature (red) imaged through a chronic cranial window of an awake NG2creER-EGFP-Kir6.1^AAA mouse. (H) Trial averaged time course data of PA dilations in response to whisker stimulation in awake mice (n = 60 trials on 15 PAs, 4 vehicle-treated NG2creER-EGFP-Kir6.1^AAA mice, n = 116 trials on 28 PAs, 5 4-OHT-treated NG2creER-EGFP-Kir6.1^AAA mice). (I) Summary data. The kinetics of PA dilation to whisker stimulation were not affected by K_ATP channel inactivation in pericytes and SMCs (n = 15 arterioles, 4 vehicle-treated NG2creER-EGFP-Kir6.1^AAA mice, n = 28 arterioles, 5 4-OHT–treated NG2creER-EGFP-Kir6.1^AAA mice; Mann–Whitney U test, U = 184.5, P = 0.5213). (J) The initial phase of PA dilation to whisker stimulation was blunted by K_ATP channel inactivation in pericytes and SMCs (n = 15 arterioles, 4 vehicle-treated NG2creER-EGFP-Kir6.1^AAA mice, n = 28 arterioles, 5 4-OHT–treated NG2creER-EGFP-Kir6.1^AAA mice; Mann–Whitney U test, U = 109, P = 0.0093).

Together, these experiments confirmed that K_ATP contributes to NVC in the awake mouse somatosensory cortex and strongly indicated that K_ATP channels in NG2+ mural cells play a major role. Given that we and others have previously found that K_ATP channels in PA SMCs are incapable of driving changes in arteriole diameter and blood flow (20, 24) and that vascular smooth muscle in the CNS generally lacks the pore-forming subunit of K_ATP channels after development (25), the simplest interpretation of these data is that pericyte K_ATP channels are rapidly engaged by neural activity and contribute primarily to the amplitude of the rapid initial phase of the hemodynamic response to neural activity in the cortex. However, because these genetic manipulations were not strictly limited to thin-strand pericytes, we sought methods to focally disrupt thin-strand pericyte function while assaying NVC to determine their specific role in this process.

Ablation of Thin-Strand Pericytes Eliminates K_ATP Channel–Dependent Contributions to NVC.

To determine the precise contribution of thin-strand pericyte K_ATP channels to NVC, we adapted established protocols for photoablation (26) to selectively remove thin-strand pericytes labeled with the carbocyanine nuclear stain TO-PRO-3 iodide from capillaries in Thy1-ChR2-YFP mice (Fig. 6 A−C). To ablate pericytes we tuned our laser to 855 nm and scanned a small line segment (~3 µm) at 10 to 45 s intervals for 2 to 3 min using the laser’s full power (~50 to 60 mW at the objective) (Fig. 6D). We validated this approach in PDGFRβ-tdTomato animals by irradiating thin-strand pericytes and waiting ~60 min postablation to assess spontaneous recovery of fluorescence (n = 12 pericytes from 3 mice), which occurred in only 16% of pericytes targeted (SI Appendix, Extended Data Fig. 18 A, B, and H). We also noted structural microvascular changes associated with pericyte ablation (SI Appendix, Extended Data Figs. 18 D−G and 19 and 20). With this validation, we applied this approach in anesthetized Thy1-ChR2-EYFP animals, targeting as many thin-strand pericytes we could identify on capillary branches radiating from the PA from which dilation measurements were taken (Fig. 6 C−E). To prevent damage to the capillary endothelium, we targeted only pericytes with a soma protruding laterally from the capillary wall. A range of 3 to 9 pericytes were ablated in each experiment. By locating 2 PAs in a single animal, we were able to deliver “on target” ablations to the pericytes on capillaries emanating from one PA and sham ablations adjacent to pericytes on the capillaries of the other PA. In the latter case, we irradiated an equivalent area of parenchyma, providing a within-subject control for laser ablation and elapsed time. Intriguingly, the initial phase of PA dilation to neuronal ChR2 activation was reduced by pericyte ablation (Fig. 6 F and G). We observed PA dilation of ~17% to ChR2 activation in sham vascular trees and only ~7% dilation to ChR2 activation following thin-strand pericyte ablation from adjacent vascular trees in the same animal (Fig. 6H). We then performed a series of experiments in which we included 20 µM glibenclamide in the agar to preemptively block the K_ATP component of the initial phase of NVC and performed the same pericyte ablation procedures. Strikingly, the initial dilation amplitude was not reduced by ablation when K_ATP channels were first blocked (Fig. 6 I and J). When expressed as a ratio of the initial PA dilation prior to pericyte laser ablation, glibenclamide prevented the reduction of PA dilations resulting from pericyte ablation (Fig. 6K), indicating that the K_ATP contribution to NVC likely originates from thin-strand pericytes.

Fig. 6.

Ablation of thin-strand pericytes reduces arteriole dilation induced by optical excitation of neurons and this is prevented by K_ATP channel inhibition. (A) Experimental scheme depicting 2PLSM imaging and thin-strand pericyte ablation in anesthetized Thy1-ChR2-EYFP mice. (B) Imaging volume showing TO-PRO-3 iodide labeling of thin-strand pericytes (blue) on the vasculature (red). Neuronal ChR2 omitted for clarity. (C) Top: A PA with downstream thin-strand pericytes (blue, dashed circles) on capillaries, which were targeted for ablation. Bottom: Experimental workflow. After locating a suitable PA, baseline (“Pre”) dilations to optogenetically driven excitatory neural activity were measured, then ablation was performed on 4 to 9 cells (or sham regions) and then dilations to optogenetic neural activity were again measured (“Post”). (D) Schematic for pericyte ablation. (E) Example images showing a capillary with a TO-PRO-3-stained thin-strand pericyte before and after ablation. (F) Example time course showing the NVC response of an individual PA before and after ablation of 4 thin-strand pericytes. (G) Summary data. Thin-strand pericyte ablation reduced the initial phase of PA dilation (n = 13 PAs, 12 Thy1-ChR2-EYFP mice; paired Student’s t test, t₁₂ = 4.275, P = 0.0011). (H) Summary data of NVC responses within subject for sham-treated vascular networks vs. those in which pericytes were ablated (n = 5 Thy1-ChR2-EYFP mice, 2 PAs per mouse; paired Student’s t test, t₄ = 4.53, P = 0.0106). (I) Example NVC response of an individual PA before and after ablation of thin-strand pericytes in the presence of 20 µM glibenclamide in the agar, which reduced the initial dilation. No further effect was observed after ablation of thin-strand pericytes. (J) Summary data showing the magnitude of the initial phase of the dilation when glibenclamide was included in the agar. Thin-strand pericyte ablation did not reduce the initial (0 to 5 s) phase of PA dilation in the presence of glibenclamide (n = 6 PAs from 6 Thy1-ChR2-EYFP mice; paired Student’s t test, t₅ = 0.8447, P = 0.4368). (K) The ratio of the initial dilation before and after ablation in preparations with (+) and without (−) glibenclamide in the agar (Welch’s t test, t_7.797 = 3.635, P = 0.0069).

Discussion

The findings of this study cast brain thin-strand pericytes as neural activity sensors and hyperpolarizing signal generators, which send electrical information upstream to control the diameter of remote PAs. Our data support a model in which pericytes monitor neuronal activity and generate such signals via K_ATP channels to contribute to the initial phase of NVC and functional hyperemia on demand. This enables thin-strand pericytes to precisely adjust blood flow to satisfy ever-changing local neuronal metabolic needs on a moment-to-moment basis.

The contributions of pericytes to NVC have recently gained increasing attention, with many studies focused on the contributions of ensheathing pericytes to this process (11, 27, 28). Thin-strand pericytes of the deep capillary bed have received less attention, and studies of these cells have been focused primarily on their contractile capabilities (4). However, these cells are optimally poised to communicate with the endothelium via their peg-socket junctions (13), which formed the take-off point for the present work.

We developed and combined a range of genetic tools and optical approaches resulting in an optogenetic approach that allowed us to isolate the contributions of individual pericytes to electrical signaling throughout the brain vasculature in vivo. This single-photon optogenetic strategy enabled the local hyperpolarization of cells of interest to augment blood flow in the somatosensory cortex through evocation of dilatory responses of PAs. We discuss the technical considerations of these approaches and their context with other neurovascular optogenetic models in further detail in SI Appendix. This approach allowed us to dilate arterioles by direct hyperpolarization of SMCs, which presents experimental opportunities for future studies interrogating the dynamic contributions of PA and pial SMC membrane potential to brain blood flow control. Remarkably, PA dilations were also elicited by hyperpolarizing thin-strand pericytes deep within the capillary bed, hundreds of microns downstream of the arteriole. Based on thin-strand pericyte ion channel gene expression profile from a public dataset (29) and our prior studies implicating thin-strand pericyte K_ATP channels as energy substrate sensors in the brain (20), we chose to study the role of these channels during NVC. Others have found that global knockout of the pore-forming subunit of the vascular isoform of the K_ATP channel impaired whisker stimulation-evoked NVC responses, which was attributed to defective SMC development in these mice (25). However, we note that these findings are not universal, as in a similar global constitutive Kir6.1 knockout model basal blood flow was lower but NVC responses to visual stimulation were unaffected (30). By using a combination of acute pharmacological and optical approaches and short-term genetic manipulations in adult mice, we approached the question of K_ATP involvement in NVC from an alternative viewpoint and found that K_ATP channels are critical for the initial phase of NVC initiated by the optogenetic activation of excitatory neurons. Our NVC experiments agree with a previous study which found that K_ATP channel blockade in the somatosensory cortex of anesthetized mice blunted sensory stimulation induced arteriole dilation (31) and we implicate thin-strand pericytes as the locus of this response. This conclusion is also supported by evidence that pericytes in the retina are coupled to capillary ECs in intracellular dye transfer experiments (32). Functional evidence from ex vivo retinal preparations also supports the possibility of electrical signaling, in that directly hyperpolarizing one retinal pericyte with a microelectrode results in an attenuated hyperpolarization of another pericyte several hundred micrometers away (33). Here, activation of K_ATP channels in one retinal pericyte also resulted in a hyperpolarization in a distant pericyte that was, interestingly, largely unattenuated (33). Notably, we previously showed that a similar mechanism in the cortex depends on direct vascular connectivity between the capillaries under study and the upstream PA, where activation of thin-strand pericyte K_ATP channels in vivo resulted in rapid dilation of connected upstream arterioles and increased capillary blood flow, but no effects of ejected pinacidil could be seen on neighboring but unconnected arterioles (20). The present data from in vivo preparations in both anesthetized and awake mice suggest that this K_ATP-dependent pericyte electrical transmission system is recruited at the onset of cortical NVC to quickly direct blood flow to actively computing neuronal networks and satisfy the increased energy demands of this activity. Our results suggest this process is aided by capillary EC Kir2.1 channels, which may be activated not only by transmitted hyperpolarization but also by K⁺ released from adjacent thin-strand pericyte K_ATP channels. Indeed, the ability of Ba²⁺ to block ArchT-mediated dilations from distal pericytes argues that this signal is propagated via endothelial-Kir2.1 recruitment, although we note that direct action of Ba²⁺ on pericyte Kir channels (34) and Kir channels in other cells could also play a role in these results.

Our conclusion that thin-strand pericytes are the primary providers of K_ATP input to NVC is strengthened by previous observations of a lack of effect of K_ATP agonist-induced dilation of PAs in isolated myograph preparations and in vivo pressure ejection experiments performed in mice (20, 24). However, it is notable that other studies (35) have identified K_ATP currents in murine PA SMCs and capillary ECs, and thus these could conceivably be recruited in specific circumstances. Other studies have identified functional K_ATP channels in rabbit pial artery SMCs (36), which raises the possibility of species-specific differences in K_ATP expression and function throughout the brain vasculature. To this point, another study of rat cerebral arteries indicated that these lack functional K_ATP channels (37). Accordingly, it will be important to study K_ATP function in other models and directly in human vascular cells, which has important implications for the translational potential of work done in mammalian models. By strongly suggesting that thin-strand pericytes are electrically coupled to the feeding arteriole by forming a syncytium with the capillary network in intact brains, the present findings further imply that CNS pathology and neurodegeneration could result directly from loss of pericyte-to-arteriole electrical communication.

What neurogenic signal(s) might activate pericyte K_ATP channels during NVC? Other investigators have observed accumulation of a fluorescent glucose analog within neurons of the somatosensory cortex in a whisker pad electrical stimulation-dependent manner (38). It may be that optogenetic excitation of pyramidal neurons results in thin-strand pericyte K_ATP channel opening through a reduction in available parenchymal glucose, which we have previously shown can activate thin-strand pericyte K_ATP channels (20). This possibility is currently under direct experimental investigation in our laboratory. Earlier studies also found that neuronal stimulation initiated an uptake of glucose (preferentially over lactate) (39) and extracellular glutamate has also been shown to increase uptake of glucose into astrocytes (40). Accordingly, substrate depletion during activity could be a crucial stimulus that is sensed by pericytes to initiate NVC. Adenosine has also been implicated in NVC (41), which when applied to the cortex can activate K_ATP channels on pericytes via G_sPCR signaling to increase cerebral blood flow (35). Notably, this group also found adenosine can dilate arterioles when applied directly to downstream capillaries (35), potentially positioning adenosine sensing by pericytes as a plausible driver of NVC in the mouse somatosensory cortex.

In analogy to electrical signaling in neuronal dendritic trees, capillary pericytes may be akin to postsynaptic spines where currents are initially generated and passed into the underlying dendritic stalk to spread through the dendritic cable before summation at the soma. We propose that in the vasculature, the pericytes (spines) generate current and pass this to the capillary endothelium (dendrites) which then spreads to the upstream arteriole (soma). This arrangement forms a distributed long-range electrical communication system that enables actuation of the distant electromechanical blood flow control apparatus located in the PA SMCs and ensheathing pericytes from the deep capillary bed to control blood flow. In neurons, only particularly intense or temporally and spatially summed synaptic inputs can initiate an action potential. An intriguing possibility is that the arteriole and proximal capillary branches bearing contractile pericytes spatially and temporally sum voltage inputs from multiple distal pericytes, much like a neuronal soma sums synaptic input, to precisely match blood delivery to local network activity. The thin-strand pericyte electrical network might thus be viewed as a system evolved to stream data on neural activity to the electromechanical blood flow controllers (i.e. transitional zone capillaries and the feed arteriole). These contractile cells then integrate demand signals for blood flow from many pericytes simultaneously over time and continually output a blood flow response proportional to the computed demand.

Methods

A detailed description of animal husbandry, surgical procedures, electrophysiology, microscopy, and data analysis can be found in SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Movie S1.

In vivo Z-stack images taken through a cranial window in a PDGFRβ-ArchT-EGFP mouse. Labeling highlights slices corresponding to depth of 100-80 microns below the pial surface showing the lower region of the “targetable volume” for single-photon point scanning optogenetic experiments. The red channel is an i.v. dye (Texas Red Dextran) and the green channel is EGFP.

Movie S2.

In vivo single-photon point scanning optogenetic experiment targeting a 6^th order thin-strand pericyte while imaging the connected upstream arteriole diameter in a PDGFRβ-ArchT-EGFP mouse. The peak dilation is seen after 5 scan periods. This video was generated from maximum intensity projection images of every 10 frames of raw data. The red channel shows an i.v. dye (Texas Red Dextran) and the green channel is EGFP.

Movie S3.

In vivo single-photon point scanning optogenetic experiment targeting a sham region while monitoring arteriole diameter in a PDGFRβ-ArchT-EGFP mouse. This video was generated from maximum intensity projection images of every 10 frames of raw data. The red channel shows an i.v. dye (Texas Red Dextran) and the green channel is EGFP.

Movie S4.

In vivo single-photon point scanning optogenetic experiment targeting a thin-strand pericyte while showing dilation of the connected upstream arteriole (bottom center of the field) in a NG2creER-ArchT-EGFP mouse. This video was generated from maximum intensity projection images of every 10 frames of raw data. The red channel shows an i.v. dye (Texas Red Dextran) and the green channel is EGFP.

Movie S5.

In vivo imaging during awake whisker stimulation of a head-fixed mouse showing prominent dilation after vibrissae deflection. The red channel is an i.v. dye (TRITC Dextran) and the green channel is EGFP.

Data, Materials, and Software Availability

Source data are accessible online (42). All study data are included in the article and/or supporting information.