Brain capillary pericytes are metabolic sentinels that control blood flow through a K_ATP channel-dependent energy switch

SUMMARY

Despite the abundance of capillary thin-strand pericytes and their proximity to neurons and glia, little is known of the contributions of these cells to the control of brain hemodynamics. We demonstrate that the pharmacological activation of thin-strand pericyte K_ATP channels profoundly hyperpolarizes these cells, dilates upstream penetrating arterioles and arteriole-proximate capillaries, and increases capillary blood flow. Focal stimulation of pericytes with a K_ATP channel agonist is sufficient to evoke this response, mediated via K_IR2.1 channel-dependent retrograde propagation of hyperpolarizing signals, whereas genetic inactivation of pericyte K_ATP channels eliminates these effects. Critically, we show that decreasing extracellular glucose to less than 1 mM or inhibiting glucose uptake by blocking GLUT1 transporters in vivo flips a mechanistic energy switch driving rapid K_ATP-mediated pericyte hyperpolarization to increase local blood flow. Together, our findings recast capillary pericytes as metabolic sentinels that respond to local energy deficits by increasing blood flow to neurons to prevent energetic shortfalls.

Graphical Abstract

In brief

Hariharan et al. characterize a K_ATP channel-dependent energy switch mechanism that imbues brain capillary thin-strand pericytes with the ability to couple subtle changes in local glucose with profound elevations in brain blood flow.

INTRODUCTION

The precise control of cerebral hemodynamics is essential to meet the highly fluctuating moment-to-moment metabolic needs of neurons and glia. Blood travels over the cortical surface in pial arteries, branching and anastomosing extensively throughout their course.¹ Penetrating arterioles (PAs) split from these and dive into the parenchyma.² PAs in turn give rise to a tortuous capillary network that is covered by a diverse population of pericytes. At the point of branching from the PA, pre-capillary sphincters are found in approximately 70% of cases that exert dynamic control of blood flow into the capillary bed by virtue of their α-smooth muscle actin (α-SMA) expression.³ Downstream of this sphincter, the initial 3–4 branches of the capillary network are covered by contractile cells that also express α-SMA, making them capable of rapidly regulating the diameter of, and therefore blood flow through, the underlying vessel.^4–6 These branches are collectively referred to as the transitional segment,⁷ reflecting their positioning between the terminal PA and the deep capillary bed and their contractile cells have been referred to as smooth muscle cells (SMCs),^8,9 contractile pericytes⁵ and ensheathing pericytes.^4,10 Here, we use the term contractile pericytes to delineate these cells from upstream arteriolar SMCs while distinguishing them from their less contractile downstream counterparts. Deeper in the capillary bed (from approximately the fifth branch and above), the abluminal surface of the capillaries is adorned by the processes and cell bodies of thin-strand pericytes.^4,11 The latter extend long, narrow processes that stretch in some cases for hundreds of microns along the walls of local capillaries, coming into close apposition with neighboring pericytes of the same type.¹² These cells also form peg-socket junctions with the underlying endothelial cells (ECs), which are thought to be the sites of gap junction coupling, permitting the exchange of molecules and charge between these cells.^13–15

Pericytes contribute to multiple physiological processes including regulation of blood-brain barrier permeability and modulation of EC gene expression.^16,17 A growing body of evidence indicates that contractile pericytes of the transitional segment play a key role in controlling hemodynamics by rapidly regulating the diameter of the underlying capillaries.^3,5,18–22 However, physiological mechanisms for blood flow control by thin-strand pericytes have not been defined. Emerging evidence obtained using optogenetic actuators suggests that subtle contractile processes in these cells can regulate capillary diameter and, therefore, local capillary blood flow.⁶ Here, we show that the intrinsic electrical activity of thin-strand pericytes alone is sufficient for robust, remote blood flow control by these cells via communication through the underlying endothelium.

Adenosine triphosphate (ATP)-sensitive potassium (K⁺; K_ATP) channels are found in a range of tissues where they play a major role in coupling metabolism to membrane electrical activity.^23–26 In the brain, K_ATP channels are present in some neuronal populations,^27–30 astrocytes,³¹ SMCs, ECs, and pericytes^32,33 and their activation is known to exert hemodynamic effects.^32–38 Given that the vascular form of the K_ATP channel, composed of inward rectifier K⁺ (K_IR) 6.1 and sulfonylurea receptor 2 subunits is the most highly expressed ion channel subtype in thin-strand pericytes, accounting for almost one-half of their relative expression of all ion channel genes,^39–41 we hypothesized that K_ATP channels in these cells could link local metabolic substrate availability to membrane hyperpolarization and, ultimately, blood flow control. In support of this, we demonstrate that deep capillary pericyte K_ATP channels are the molecular cornerstone of an energy switch mechanism, wherein a decrease in local glucose availability belowa key threshold evokes a K_ATP channel-mediated membrane hyperpolarization. This hyperpolarizing electrical signal generated by pericyte K_ATP channels is fed into the underlying endothelium, likely by means of pericyte-EC gap junctions,¹⁵ where it then activates endothelial K_IR2.1 channels, promoting its regenerative transmission upstream to dilate the feeding PA by relaxing the overlying SMCs.^42,43 This process leads to a rapid increase in blood flow, thereby replenishing energy substrate delivery to neurons and glia. Our data thus recast thin-strand pericytes as metabolic sentinels that monitor energy availability and dynamically modulate blood flow through electrical signaling to ensure that the energy substrates required to support ongoing neuronal function are continually provided.

RESULTS

K_ATP channel activation increases arteriolar diameter and capillary blood flow in vivo

We began by visualizing the vascular network of a volume of cortex through a cranial window preparation in anesthetized mice (Figure 1A). We identified pial arteries on the brains surface and their arising PAs branching perpendicularly into the tissue by their morphology in relation to nearby veins, and imaged these PAs and their daughter capillaries down to at least the fifth branch of the capillary bed (Figure 1A). In vivo, these arteries constrict in response to intravascular pressure,⁴⁴ establishing a baseline of myogenic tone from which diameter can be modulated to adjust blood flow. Under our conditions, we found that PAs had 46.4 ± 2.8% tone (n = 8 arterioles from 5 mice), calculated by comparing baseline arteriolar diameter with passive diameter in the absence of extracellular calcium (Ca²⁺) and the presence of the voltage-dependent Ca²⁺ channel blocker diltiazem (200 μM) (Figure S1). The tone of the first- to fourth-order branches of the capillaries of the transitional segment collectively averaged 39.4 ± 2.2% tone at baseline, which was not different to the tone of PAs and did not differ by branch order (n = 35 capillaries from 5 mice) (Figure S1). To examine the influence of K_ATP channels on capillary and arteriole diameter and blood flow, we assessed the effects of pharmacologically modulating these channels through superfusion of agents over the brain surface. Strikingly, we found that the application of 10 μM pinacidil, a selective K_ATP channel opener, produced near-maximal dilation of both PAs and transitional segment capillaries (Figures 1B–1H), indicating that active K_ATP channels exert a strong influence on the vasculature. In turn, these substantial increases in diameter translated into profound elevation of capillary blood flow, measured as red blood cell (RBC) flux using a high-frequency line scanning approach (Figures 1I and 1J). To determine whether K_ATP channel activity contributes to basal blood flow, we applied the K_ATP channel blocker glibenclamide (10 μM) to the brain surface. We observed no change in the diameter of the PA or first- to fourth-order capillaries to this maneuver, and no change in capillary blood flow (Figure S2), suggesting that vascular K_ATP channel activity in the brain is minimal under resting conditions. To confirm that this concentration is sufficient to achieve K_ATP channel block in vivo, we also tested whether the pre-application of glibenclamide prevented pinacidil-evoked changes in vessel diameter. As expected, this maneuver eliminated pinacidil-evoked responses (Figure S3).

Figure 1. K_ATP channel activation increases arteriole diameter and capillary blood flow in vivo.

(A) In vivo imaging set-up. (Left) A cranial window was made over the somatosensory cortex and imaged using two-photon laser scanning microscopy. (Right) Imaging field of the vasculature containing fluorescein isothiocyanate (FITC)-dextran showing a pial vein and artery, a PA, and downstream capillaries. 2PLSM, 2-photon laser scanning microscopy.

(B) A PA with a pre-capillary sphincter and its downstream first- and second-order capillaries. (Top) Baseline diameter of the PA (red line). (Bottom) Dilation of the PA after the application of 10 μM pinacidil (red line: baseline diameter). The capillaries in view also dilated to this maneuver.

(C) A PA and its downstream first- to fourth-order capillaries. (Left) Baseline diameters of first- to fourth-order capillaries indicated by respective colored lines. (Right) The same first- to fourth-order capillaries, which dilated after application of 10 μM pinacidil.

(D–H) Summary data, analyzed using Student’s paired t-test, showing the application of 10 μM pinacidil produced a significant dilation of the (D) PA (n = 7 vessels, 7 mice, *p = 0.01, t₆ = 3.697), (E) first-order capillary (n = 9 vessels, 7 mice, *p = 0.017, t₈ = 2.987), (F) second-order capillary (n = 11 vessels, 6 mice, **p = 0.0029, t₁₀ = 3.921), (G) third-order capillary (n = 11 vessels, 6 mice, **p = 0.0011, t₁₀ = 4.513), and (H) fourth-order capillary (n = 4 vessels, 4 mice, *p = 0.0326, t₃ = 3.772).

(I) Line-scanning strategy used to measure blood flow in higher order capillaries. (Top right) Baseline kymograph displaying RBCs passing through the line-scanned capillary as dark shadows against the green fluorescence of FITC-containing plasma. (Bottom right) Kymograph of the same capillary post-pinacidil application showing an increase in RBC flux.

(J) Summary of RBC flux responses showing significant hyperemia to 10 μM pinacidil (n = 6 vessels, 3 mice, *p = 0.0188, t₅ = 3.424, Student’s paired t-test).

Pericytes transmit K_ATP channel-mediated electrical signals via the endothelium to exert remote control over the diameter of upstream arterioles

K_ATP channel expression is relatively low in SMCs and ECs of the brain compared with thin-strand pericytes,^32,39–41 and pharmacological maneuvers designed to activate these channels in isolated PAs do not lead to dilation.⁴⁵ Given that systems-level K_ATP channel activation, in contrast, evoked profound vasodilation and blood flow increases, we reasoned that the high expression of K_ATP channels in deep capillary thin-strand pericytes could be the primary source of hyperpolarizing signals that may then be relayed to upstream PAs to drive their vasodilation. To test this possibility, we maneuvered a pipette connected to a pressure-ejection system into the brain and positioned it next to a DsRed-positive thin-strand pericyte in Cspg4-DsRed mice (Figures 2A–2C). On average, targeted pericytes were 268.5 ± 25.9 μm from the upstream arteriole imaging site (n = 8 experiments, 8 mice). Consistent with our hypothesis, focal activation of K_ATP channels by pressure ejection of 10 μM pinacidil onto the pericyte cell body (Figure 2C) evoked a rapid and substantial upstream arteriolar dilation (Figures 2D, 2E, and 2G and Video S1), which was accompanied by an increase in underlying capillary blood flow (Figures 2F and 2H).

Figure 2. Capillary pericytes exert remote control over upstream PA diameter.

(A) Experimental strategy showing an ejection pipette positioned next to a pericyte.

(B) Z-projections of 3-dimensional volume acquisitions outlining the experimental strategy. (Left) Vasculature containing fluorescein isothiocyanate (FITC)-dextran and a pipette with TRITC-dextran positioned within the cortex. (Right) A PA and its downstream capillary network showing an ejection pipette containing TRITC-dextran with 10 μM pinacidil positioned next to a DsRed+ pericyte on an eighth-order capillary.

(C) Depiction of the evolution (left to right) of TRITC diffusion (red) after pressure-ejection of 10 μM pinacidil onto a DsRed+ pericyte. The brevity and low pressure of the ejection conditions (10 psi, 30 ms) ensured that the drug remained local.

(D) Focal stimulation of capillary pericytes with 10 μM pinacidil dilates the connected upstream PA. (Left) PA and first order capillary diameter at baseline indicated by magenta lines. (Right) Peak dilation of the same PA and first order capillary after pinacidil-ejection on the downstream pericyte.

(E) Time courses showing PA dilation to direct stimulation of a pericyte with pinacidil (orange, top), but no change in PA diameter when pinacidil was applied in the presence of the K_IR channel blocker Ba²⁺ (purple, middle) or when pinacidil was ejected onto a segment of capillary without a pericyte cell body (blue, bottom).

(F) 1-s kymograph segments showing raw RBC flux of a greater than fifth-order capillary at baseline, and hyperemia after pinacidil was ejected onto the overlying pericyte.

(G) Summary of PA diameter changes after pinacidil-ejection on a downstream pericyte (n = 14 paired measurements, 13 mice, ***p = 0.0007, t₁₃ = 4.439, Student’s paired t-test).

(H) Summary capillary RBC flux responses to pinacidil applied directly to a pericyte (n = 8 paired measurements, 4 mice, **p = 0.0045, t₇ = 4.108, Student’s paired t-test).

(I) Summary data showing PA diameter after pinacidil-stimulation of a pericyte in the presence of Ba²⁺ (n = 6 paired measurements, 6 mice, p = 0.6981, t₅ = 0.411, Student’s paired t-test).

(J) Summary blood flow data showing RBC flux before and after pinacidil stimulation of a pericyte in the presence of 100 μM Ba²⁺ (n = 5 paired measurements, 5 mice, p = 0.4613, t₄ = 0.814, Student’s paired t-test).

(K) Summary data showing PA diameter changes on stimulation of a capillary segment without a pericyte cell body with pinacidil (n = 6 paired measurements, 6 mice, p = 0.2162, t₅ = 1.415, Student’s paired t-test).

(L) Summary of RBC flux responses before and after stimulation of a capillary segment without pericytes with pinacidil (n = 5 paired measurements, 5 mice, p = 0.394, t₄ = 0.9543, Student’s paired t-test).

Pericytes connect with adjacent ECs via peg-socket contacts, which are thought to be the sites of gap junctions.^15,39,46 Accordingly, we reasoned that signals originating in pericytes may be transmitted upstream via connected underlying ECs to elicit PA dilation. Consistent with this possibility, we found that the diameter of a neighboring PA that was not connected to the target pericyte through its downstream capillaries was unchanged when pinacidil was ejected onto a target pericyte (Figure S4), indicating the necessity of an unbroken chain of vasculature between pericyte and PA for these responses. We previously identified an EC-mediated regenerative electrical signaling mechanism dependent on K_IR2.1, a strongly rectifying K_IR channel isoform, that transmits dilatory signals from deep capillaries to upstream PAs.⁴³ Consistent with the involvement of this mechanism, blocking K_IR2.1 channels through the application of 100 μM barium (Ba²⁺) to the cortical surface prior to pinacidil ejection onto a thin-strand pericyte abolished the increase in diameter of the connected upstream PA and eliminated changes in capillary blood flow (Figures 2E, 2I, and 2J). We note that neurons and glia may also express K_IR channels⁴⁷; however, we have previously observed that 100 μM Ba²⁺ applied to the brain surface does not significantly alter neuronal activity under these conditions.⁴³ Together, these data suggest that pericyte K_ATP channel-initiated hyperpolarization engages K_IR2.1-dependent electrical signaling through the capillary bed to produce its effects.

In further experiments seeking to assess whether pericytes are the locus of pinacidil-evoked vasodilatory drive, we found that the ejection of this agent onto a segment of capillary lacking a pericyte soma (distance between ejection site and nearest pericyte cell body: 64 ± 6.17 μm; n = 5 sites in 3 mice) (Figure S5) had no effect on arteriolar diameter or on capillary blood flow (Figures 2E, 2K, and 2L). This observation suggests that K_ATP channels in the target capillary ECs (cECs), in local neurons, or in nearby glial cells do not contribute to the hyperemic effects we observe to this maneuver. As expected, diameter and blood flow were also unchanged when pericytes were stimulated with vehicle (artificial cerebrospinal fluid containing 0.3 mg/mL TRITC-dextran) (Figure S6). In theory, K_ATP channels in arteriolar SMCs or ECs could also be inadvertently activated by the spread of pinacidil from these ejection experiments, although the data in Figure S4 argue against this. To test this possibility directly, we stimulated the PA with 10 μM pinacidil and found that this had no effect on arteriolar diameter (Figure S7), which aligns with previous observations of a lack of response of isolated PAs to K_ATP agonists⁴⁵ and thus suggests that arteriolar ECs and SMCs do not contribute to the effects observed. Collectively, therefore, our data are consistent with the conclusion that pericyte K_ATP channels exert remote control of upstream PA diameter by engaging cEC electrical signaling to transmit signals over long distances.

Expression of a dominant-negative mutant of the vascular K_ATP channel eliminates pericyte-mediated dilations and hyperemia

To further confirm the role of pericyte K_ATP channels in the control of blood flow and upstream PA diameter to pinacidil, we deployed mice that express a dominant-negative form of the K_IR6.1 subunit in which a Gly-Phe-Gly motif of the K⁺ selectivity filter is mutated to a non-functional alanine triplet (K_IR6.1^AAA), which in turn eliminates K_ATP conductance.^48,49 Expression of K_IR6.1^AAA was controlled by tamoxifen-inducible Cre-recombinase under the Cspg4 promoter to selectively suppress K_ATP channel activity in pericytes and SMCs. In this line, a floxed region containing the sequence for enhanced green fluorescent protein (EGFP) upstream of a stop codon is expressed under basal conditions, precluding expression of the downstream K_IR6.1^AAA sequence without Cre-recombinase activity. When recombination is induced, EGFP along with the stop codon are excised, permitting K_IR6.1^AAA expression (Figure 3A). Accordingly, induction of Cre activity in Cspg4-Cre-K_IR6.1^AAA mice by 4-hydroxy tamoxifen (4-OHT) eliminated EGFP expression in capillary pericytes, while EGFP expression was retained in adjacent ECs (Figures 3B and 3C), indicating successful cell type-selective expression of the K_IR6.1^AAA construct. To then reveal pericytes with inactive K_ATP channels, we applied NeuroTrace 500/525 (NT500/525)⁵⁰ to the brain surface, which specifically stained thin-strand pericytes (Figure 3C). Pressure-ejecting pinacidil onto thus identified EGFP-negative, NT500/525-positive pericytes did not produce an increase in upstream PA dilation or elevate local capillary blood flow in Cspg4-Cre-K_IR6.1^AAA mice (Figures 3C, 3D, 3I, and 3J), whereas Cre control (K_IR6.1^AAA mice given 4-OHT) and vehicle control (Cspg4-Cre-K_IR6.1^AAA mice given a 90:10% mixture of corn oil:ethanol) groups still demonstrated significant PA dilation (11%–13%) and capillary RBC flux still increased (32%–38%) to this maneuver (Figures 3D, 3E–3H). To control for the possibility that NT500/525 could itself disable pericyte K_ATP channel-mediated control of blood flow, we pressure-ejected pinacidil onto NT500/525-stained pericytes in control mice and found that upstream PA dilation and capillary hyperemia were preserved (Figure S8). Thus, we conclude that pericytes are the primary site of K_ATP-mediated upstream arteriolar dilation and local capillary hyperemia in vivo and focal stimulation of capillary pericytes is sufficient to produce these effects.

Figure 3. Capillary pericytes are a key locus of K_ATP channel-mediated control of blood flow.

(A) Pericyte K_ATP channels were genetically inactivated by crossing mice possessing a modified K_IR6.1 subunit (K_IR6.1^AAA) with Cspg4-Cre mice. Cre control mice (K_IR6.1^AAA+, Cre−) and K_IR6.1^AAA mice (K_IR6.1^AAA+, Cre+) were given 4-OHT, whereas vehicle control mice (K_IR6.1^AAA+, Cre+) were given vehicle.

(B) Inactivation of the K_IR6.1 subunit was evidenced by the elimination of EGFP signal in pericytes. (Left) Representative Z-projection from a vehicle control mouse. (Right) Representative Z-projection from a tamoxifen-induced Cspg4-Cre-K_IR6.1^AAA mouse showing fewer EGFP+ cells.

(C) Experimental strategy to identify and target inactivated K_ATP channels in pericytes. (Left) A Cspg4-Cre-K_IR6.1^AAA mouse, with EGFP+ ECs, and pericytes lacking EGFP, indicating successful K_IR6.1^AAA induction. (Right) The location of EGFP-negative pericytes was determined using the in vivo pericyte-specific dye Neurotrace (NT) 500/525. (Inset) A pipette containing fluorescein isothiocyanate (FITC) and 10 μM pinacidil positioned next to an EGFP-, NT 500/525+ pericyte.

(D) Traces of PA diameter showing dilation to downstream ejection of pinacidil onto a capillary pericyte in a Cre-control mouse (pink) and a lack of response in Cspg4-Cre-K_IR6.1^AAA mice (brown).

(E–J) Summary data of changes in PA diameter and blood flow to focal application of pinacidil onto a capillary pericyte across different experimental groups.

(E) PA diameter changes in Cre-control mice (n = 5 paired measurements, 5 mice, **p = 0.0026, t₄ = 6.684). (F) RBC flux changes in Cre-control mice (n = 5 paired measurements, 5 mice, ***p = 0.0006, t₄ = 9.908). (G) PA diameter changes in vehicle control mice (n = 3 paired measurements, 3 mice, *p = 0.0157, t₂ = 7.883).

(H) RBC flux changes in vehicle control mice (n = 3 paired measurements, 3 mice, **p = 0.0023, t₂ = 20.78). (I) PA diameter changes in Cspg4-Cre-K_IR6.1^AAA mice (n = 10 paired measurements, 10 mice, p = 0.3054, t₉ = 1.087). (J) RBC flux changes in Cspg4-Cre-K_IR6.1^AAA mice (n = 10 paired measurements, 10 mice, p = 0.7249, t₉ = 0.3631). All data were analyzed using Student’s paired t-test.

An energy switch couples decreases in local energy substrate availability to pericyte hyperpolarization via K_ATP channel activity

We next turned our attention to the mechanisms through which K_ATP channels may be engaged. In other tissues, K_ATP channels play a critical role in coupling metabolism to membrane electrical activity and are sensitive to the local level of glucose.⁵¹ We, thus, hypothesized that pericytes might sense fluctuations in energy substrate levels in the brain and respond to decreases in glucose availability with K_ATP channel-mediated electrical signals.

The glucose concentration in bulk cerebrospinal fluid is approximately 4 mM,⁵² whereas parenchymal glucose has been measured in the range of approximately 0.3–2.3 mM.^53–66 Accordingly, we wondered whether subtle changes in local glucose concentration through this physiological range would influence the degree of K_ATP channel activity and modulate pericyte membrane potential (Vm). To permit direct monitoring of pericyte Vm, we isolated capillaries from Cspg4-DsRed mice with intact thin-strand pericytes, and measured Vm using sharp microelectrode impalements (Figure 4A). Across all conditions of replete glucose (1 mM–4 mM), pericyte resting Vm averaged −36 mV (45 cells, 19 mice) (Figures 4B, 4E, 4H, and 4L). Under these conditions, the activation of K_ATP channels by bath application of 10 μM pinacidil hyperpolarized Vm by an average of approximately 23 mV (maximum: 40 mV), which was blocked by the co-application of 10 μM glibenclamide (Figures 4B–4E).

Figure 4. Decreasing glucose activates K_ATP channels to hyperpolarize pericytes.

(A) Experimental setup illustrating impalement of a DsRed+ capillary pericyte using a sharp microelectrode.

(B–D) Traces of Vm measurements at baseline (B), with 10 μM pinacidil (C), and with 10 μM pinacidil in the presence of 10 μM glibenclamide (D).

(E) Summary data showing that pinacidil hyperpolarizes pericyte Vm and that glibenclamide blocks this effect (baseline [n = 10 cells, 5 mice] vs. pinacidil [n = 9 cells, 4 mice]: ***p = 0.002, t₄₉ = 4.453; pinacidil vs. pinacidil + glibenclamide [n = 6 cells, 4 mice]: ****p < 0.0001, t₄₉ = 5.278, one-way ANOVA with Sidak’s multiple comparison test).

(F) Example trace from a continuous impalement recording where bath solution was switched from 2 mM glucose to 0.25 mM glucose resulting in pericyte hyperpolarization that was rapidly reversed by the application of 10 μM glibenclamide.

(G) Summary data from continuous recording impalements comparing the change in Vm in response to glucose depletion and the reversal of this by 10 μM glibenclamide (n = 5 recordings, 3 mice: **p = 0.0076, t₄ = 4.974, Student’s paired t-test).

(H–K) Example traces of Vm measurements with 2 mM bath glucose (H), with 0.75 mM bath glucose (I), with 0.25 mM glucose (J), and under 0.25 mM glucose conditions with the addition of 10 μM glibenclamide (K).

(L) Summary data showing that decreasing glucose to less than 1 mM hyperpolarizes pericyte Vm, and that the effects of 0.25 mM glucose were blocked by glibenclamide (4 mM glucose [n = 12 cells, 5 mice] vs. 2 mM glucose [n = 14 cells, 5 mice]: p > 0.9999, t₉₇ = 0.1204; 4 mM glucose vs. 1 mM glucose [n = 9 cells, 4 mice]: p = 0.8384, t₉₇ = 1.28; 4 mM glucose vs. 0.75 mM glucose [n = 20 cells, 4 mice]: ***p = 0.009, t₉₇ = 4.04; 4 mM glucose vs. 0.25 mM glucose [n = 16 cells, 4 mice]: ***p = 0.0005, t₉₇ = 4.193; 4 mM glucose vs. 0 glucose [n = 9 cells, 5 mice]: ***p = 0.0007, t₉₇ = 4.078; 0.25 mM glucose vs. 0.25 mM glucose +10 μM glibenclamide [n = 9 cells, 4 mice]: ****p < 0.0001, t₉₇ = 5.009; one-way ANOVA with Sidak’s multiple comparison test).

(M) Concentration-response curve showing pericyte Vm hyperpolarizes abruptly in response to decreasing glucose concentrations.

Strikingly, in continuous recordings we observed a rapid and profound hyperpolarization as the bath solution was switched from 2 mM glucose to 0.25 mM glucose, an effect that was then reversed by the application of glibenclamide (10 μM) in the continued presence of decreased glucose (Figures 4F and 4G). This rapid hyperpolarizing response to glucose depletion led us to characterize the concentration dependence of the relationship between glucose and pericyte Vm in detail. Thus, we impaled pericytes shortly after incubation with one of a series of physiological glucose concentrations. The Vm of pericytes in 1 mM glucose was no different to that of cells in 2 mM or 4 mM glucose (Figure S9). However, pericytes in 0.75 mM or 0.25 mM glucose (Figures 4I, 4J, and 4L) or in extracellular solution that lacked glucose (Figure S9) were all strongly hyperpolarized to approximately −50 mV, which was prevented by the inclusion of 10 μM glibenclamide in the bath (Figures 4K, 4L, and S9). These experiments suggest the presence of a glucose threshold around approximately 1 mM (half-maximal effective concentration of 934 μM) (Figure 4M), which we refer to as an energy switch, below which a dramatic increase in K_ATP channel activity occurs. Collectively, our data indicate that pericytes are acutely sensitive to fluctuations of glucose within the physiological range, and, if the concentration falls below a critical threshold, K_ATP channel activity is robustly increased to evoke substantial membrane hyperpolarization.

Glucose transporter 1 (GLUT1) blockade activates the pericyte energy switch in vivo and triggers profound arteriolar dilation to increase local blood flow

The endothelium plays a major role in glucose import into the brain, predominately via highly-expressed GLUT1 (Figure 5A), and pericytes also express the gene encoding GLUT1 (Slc2a1) and to a lesser extent the genes for GLUT3 (Slc2a3) and GLUT4 (Slc2a4).^40,41

Figure 5. Glucose levels control K_ATP channel activity and blood flow in vivo.

(A) Staining with an anti-GLUT1 antibody indicating the high density of this transporter in brain capillaries.

(B and C) Example traces of Vm measurements under 1 μM BAY-876 (B) and 1 μM BAY-876 in the presence of 10 μM glibenclamide (C).

(D) Summary data showing BAY-876 (n = 19 cells, 6 mice) hyperpolarizes pericyte Vm and this effect is blocked by glibenclamide (n = 10 cells, 6 mice: ***p = 0.0003, t₂₇ = 4.192, Student’s unpaired t-test).

(E) Effects of GLUT1 inhibitor BAY-876 (1 μM) on PA diameter. (Left) PA diameter indicated by white line at baseline. (Right) Dilation of the same PA after application of BAY-876 to the brain surface.

(F) BAY-876 also dilates first- to fourth-order capillaries. (Left) A Z-projection showing diameters of first- to fourth-order capillaries at baseline, indicated by colored lines. (Right) Dilation of the same capillaries after BAY-876 application.

(G–K) Summary data analyzed using Student’s paired t-test, showing dilation across all vessels with BAY-876. (G) PA diameter (n = 17 vessels, 5 mice, ****p < 0.0001, t₁₆ = 10.01). (H) First-order capillary diameter (n = 9 vessels, 5 mice, **p = 0.0029, t₈ = 4.22). (I) Second-order capillary diameter (n = 16 vessels, 5 mice, ****p < 0.0001, t₁₅ = 11.16). (J) Third-order capillary diameter (n = 16 vessels, 5 mice, ****p < 0.0001, t₁₅ = 8.399). (K) Fourth-order capillary diameter (n = 17 vessels, 5 mice, ****p < 0.0001, t₁₆ = 7.665).

(L–N) 1-s segments of raw kymographs demonstrating hyperemia to BAY-876 in control mice (L), and the elimination of hyperemic responses in the presence of glibenclamide (M) and in Cspg4-Cre-K_IR6.1^AAA mice (N). (Top) Baseline RBC flux. (Bottom) RBC flux measured in the same capillary after BAY-876 application.

(O) Summary RBC flux data from fifth and higher-order capillaries showing the elimination of the hyperemic response when BAY-876 was applied in the presence of glibenclamide or when BAY-876 was applied in Cspg4-Cre-K_IR6.1^AAA mice compared with controls (controls [n = 11 capillaries, 4 mice] vs. BAY-876 in the presence of glibenclamide [n = 24 capillaries, 5 mice]: **p = 0.0073, t₄₈ = 3.028; controls vs. Cspg4-Cre-K_IR6.1^AAA mice [n = 16 capillaries, 6 mice]: **p = 0.0013; t₄₈ = 3.637; one-way ANOVA with Dunnett’s multiple comparison test).

(P) Summary data showing significantly decreased dilatory responses to BAY-876 in the presence of glibenclamide and in Cspg4-Cre-K_IR6.1^AAA mice across all vessel orders (n = 5 mice per group [control and glibenclamide], n = 6 mice [Cspg4-Cre-K_IR6.1^AAA mice]; control vs. glibenclamide: PA: **p = 0.0013, t₂₁₃ = 3.9; first-order capillary: *p = 0.0243, t₂₁₃ = 3.066; second-order capillary: **p = 0.0094, t₂₁₃ = 3.355; third-order capillary: ***p = 0.0002, t₂₁₃ = 4.332; fourth-order capillary: ***p = 0.0002, t₂₁₃ = 4.324; control vs. Cspg4-Cre-K_IR6.1^AAA: PA: **p = 0.0013, t₂₁₃ = 3.905; first-order capillary: ***p = 0.0002, t₂₁₃ = 4.364; second-order capillary: ***p = 0.0005, t₂₁₃ = 4.15; third-order capillary: ****p < 0.0001, t₂₁₃ = 4.681; fourth-order capillary: ****p < 0.0001, t₂₁₃ = 5.571; one-way ANOVA with Sidak’s multiple comparison test).

Given its central role, we hypothesized that blocking GLUT1 would be sufficient to activate the pericyte energy switch and generate K_ATP channel activity to hyperpolarize pericyte Vm. This, in turn, should influence electrical signaling through the capillary network and drive an increase in arteriolar diameter and blood flow.

In line with the predictions of our hypothesis, blocking glucose entry using the selective GLUT1 inhibitor BAY-876 (1 μM) hyperpolarized the membrane of isolated pericytes to −51 mV, as seen with concentrations of glucose that are less than 1 mM (Figure 4L), and this effect was completely inhibited by glibenclamide (Figures 5B–5D). Based on the known Vm-diameter relationship of PAs, a 15-mV hyperpolarization is predicted to dilate these vessels by approximately 50% (see ref. 67). Accordingly, we tested the effect of 1 μM BAY-876 on PA and capillary diameter, and capillary blood flow when applied directly to the brain surface in vivo, reasoning that this could cause a dramatic decrease in parenchymal glucose and engage pericyte K_ATP channels. Strikingly, this maneuver produced a 48% increase in PA diameter (Figures 5E and 5G), consistent with our predictions, and profoundly dilated first- to fourth-order capillaries (Figures 5F and 5H–5K), while also almost doubling capillary blood flow (Figures 5L and 5O). Pre-incubation with glibenclamide (10 μM) or genetic inactivation of K_ATP channels in Cspg4-Cre-K_IR6.1^AAA mice eliminated the BAY-876-evoked increase in capillary RBC flux (Figures 5M–5O) and significantly decreased the dilatory effect of BAY-876 at the level of the PA (62%–68% decrease) and in first- and fourth-order capillaries (Figure 5P). Since astrocytes also robustly express GLUT1,⁶⁸ drugs targeting this transporter that are delivered to the brain surface could exert indirect effects through this avenue. Thus, we also assessed the effect of BAY-876 delivered to the blood stream (10 mg/kg), reasoning that the presence of the blood-brain barrier would enable us to preferentially target EC GLUT1 transporters. This approach produced similar effects to brain surface application, leading to robust dilation of the PA and first- to fourth-order capillaries, as well as a large increase in RBC flux (Figure S10). Together, these data suggest that a decrease in parenchymal glucose or blocking glucose import triggers pericyte K_ATP channel-mediated electrical signaling, which is transmitted upstream to the PA to drive dilation and an increase blood flow.

DISCUSSION

Our data recast pericytes as metabolic sentinels that form a brain-wide energy-sensing network, continually monitoring glucose concentrations and adjusting blood flow to protect neuronal health and function. We show that thin-strand pericytes exert profound remote control of arteriolar diameter and blood flow and implicate K_ATP channels as the molecular lynchpin of this ability. The cerebrovascular effects of activating and inhibiting K_ATP channels have been extensively studied.^38,69–71 Our data add to this by indicating that the metabolic sensitivity of this channel imbues pericytes with the ability to couple changes in local energy substrate levels to significant alterations of Vm, constituting an energy switch that, once triggered, exerts powerful control of local blood flow. These findings thus support a model in which pericyte K_ATP channels initiate robust hyperpolarization in response to a decrease in local glucose below a critical threshold, which can be transferred over long distances through engagement of capillary electrical signaling, eliciting relaxation of remote contractile pericytes and arteriolar SMCs, leading to vasodilation and an increase in blood flow into the deep capillary bed (Figure 6).

Figure 6. Pericyte K_ATP channel-mediated coupling of electrical activity with glucose availability.

When glucose is abundant, the pericyte ATP:ADP ratio is relatively high and keeps pericyte K_ATP channels closed. A decrease in GLUT1-mediated glucose import or a decrease in local glucose availability results in pericyte K_ATP channel activation, robustly hyperpolarizing pericyte Vm. This electrical signal is then fed into the underlying capillary endothelium via gap junctions to activate endothelial K_IR2.1 channels. The resulting hyperpolarization of ECs is rapidly transmitted upstream via a K_IR2.1 channel-dependent retrograde propagation of electrical signals. This remotely dilates PAs and increases blood flow, thereby replenishing local glucose levels and protecting ongoing neuronal metabolism and function.

Pericyte K_ATP channel-meditated electrical signals are transmitted through multiple branch orders to control blood flow

An elegant recent study deploying optogenetic tools in thin-strand pericytes has shown that these cells are capable of exerting slow constrictions of their underlying capillaries,⁶ yet it seems that these vessels do not dilate during functional hyperemia.^20,72 To rapidly control blood flow during neuronal activity, thin-strand pericytes could modulate ongoing electrical signaling through the underlying endothelium, which we have previously shown is transmitted over long distances to influence upstream arteriolar diameter.⁴³ Our data support this idea by demonstrating that focal activation of pericyte K_ATP channels is sufficient to evoke rapid dilation of remote PAs at distances up to at least 421 μm (the furthest site we stimulated in our experiments).

The selective application of pinacidil onto a capillary segment without a pericyte soma might be expected to still activate K_ATP channels in local thin-strand processes that reach for long distances along capillaries or in cECs.³² However, the observed lack of a response to this maneuver in our experiments suggests that the density of K_ATP channels is either too low in these processes and ECs, or the lower pericyte membrane area covered by the ejected drug results in the activation of too few channels to generate sufficient hyperpolarization to observe a functional effect upstream. The null result of these and other experiments, such as pinacidil-stimulation of pericytes in Cspg4-Cre-K_IR6.1^AAA mice or direct pinacidil-stimulation of PAs, serves to illustrate that the input of K_ATP channels recruited by stimulating the pericyte soma is necessary for these dilatory and blood flow responses and argue strongly against contributions from K_ATP channels found in yet other cells (e.g., local neurons or glia).

Collectively, then, our focal-ejection experiments show that pericytes are a key locus of K_ATP channel activity in the capillary bed, and, from this locus, hyperpolarization must be transmitted to upstream SMCs to evoke dilation at a distance. There are several possibilities as to how such long-range communication may be organized. The eponymous projections of thin-strand pericytes come into close contact with those of neighboring cells, and thus direct transfer of electrical signals between pericytes is one possibility. However, these processes do not seem to closely interdigitate and rather stay confined to their own territories.¹² Also, no evidence of direct pericyte-pericyte transfer of charge or chemical agents has been reported to our knowledge, with the exception of specialized interpericyte tunneling nanotube (IP-TNT) projections in the retina.⁷³ Thus, it presently seems unlikely that pericytes adhering to capillaries, without IP-TNTs, directly exchange electrical signals. Instead, mounting evidence indicates that thin-strand pericytes directly interface with cECs via gap junctions.¹⁴ Our prior work⁴³ revealed that electrical signaling through the brains capillary network to upstream arterioles is a major mechanism for blood flow control in the brain. This mechanism relies on cEC K_IR2.1 channels, which are activated by both external K⁺ and membrane hyperpolarization and transmit electrical signals upstream at a velocity of several millimeters per second.^24,43 Our data show that the K_IR2 channel blocker Ba²⁺ eliminates pinacidil-evoked remote dilation of PAs, suggesting that the activation of K_ATP channels in thin-strand pericytes generates membrane hyperpolarization that is then injected via peg-socket junctions into the underlying ECs to engage cEC electrical signaling and dilate upstream arterioles. We also recently reported that cEC Ca²⁺ signals control blood flow through a nitric oxide-dependent mechanism that relaxes contractile pericytes of the first- to fourth-order capillaries.²² These signals are strongly influenced by ongoing capillary electrical signaling, with hyperpolarization likely increasing the driving force for Ca²⁺ entry. Thus, it is possible that pericyte K_ATP-mediated electrical signals might also promote cEC Ca²⁺ signaling, which could be a contributory factor in the effects we observed.

A pericyte energy switch: Membrane potential is steeply influenced by local glucose concentration

The brain relies primarily on glucose and oxygen to fulfill its energy requirements. The central pathway for glucose entry is via the GLUT1 transporter, which is abundantly expressed in ECs.^41,74 Pericytes express several GLUT-encoding genes (Slc2a1 > Slc2a4 > Slc2a3, which translate to GLUTs 1, 4, and 3, respectively^40,41), suggesting that they may be capable of taking up glucose directly from their surroundings. Astrocytes also express GLUT1, while neurons primarily express GLUT3.⁷⁴ The intensive metabolic demands of neural activity exert a strong influence on parenchymal glucose concentration, such that it is lower than blood glucose.⁷⁴ Since pericytes are juxtaposed between neuroglia and the endothelium, they may be capable of responding to fluctuations in local glucose availability through metabolically evoked electrical signaling and blood flow modulation, by virtue of their robust K_ATP channel expression. To directly ascertain whether pericyte electrical behavior is influenced by local energy availability, we determined the relationship between external glucose concentration and pericyte Vm in granular detail, and specifically focused on the contribution of K_ATP channels in this context. By testing the effects of decreasing glucose from 4 mM (the concentration typically found in blood and bulk CSF) to 1 mM and lower (aligning with measurements of parenchymal glucose concentrations^{53,55,56,59–62,64,75}), we found that glucose concentrations of less than 1 mM precipitated a striking hyperpolarization mediated by K_ATP activation. Our data, thus, indicate that pericytes are steeply sensitive to local changes in this key energy substrate, and are consistent with the existence of a glucose concentration threshold around 1 mM, below which robust activation of K_ATP is elicited.

The near “all-or-none” effect of energy substrate abundance on membrane potential that we observed—reminiscent of flipping a switch—might be triggered by changes in the production of ATP in the pericyte, or accumulation of adenosine diphosphate (ADP), leading to a new set point for the intracellular ATP:ADP ratio. Accompanying this could be an amplification mechanism, such as the activation of capillary K_IR channels by membrane hyperpolarization,⁴² which in turn may boost K_ATP-initiated hyperpolarization. Together, these factors could translate a change in intracellular metabolism into a binary response, driving Vm toward E_K and facilitating potent hyperemic responses to small changes in external glucose availability. Detailed studies to determine these key intra- and intercellular mechanisms are currently in progress.

What might be the circumstances, physiological and pathological, that engage this energy-sensing mechanism? One possibility is that local fluctuations in glucose that occur during concerted neuronal activity^76,77 continually adjust the electrical input of pericytes to the capillary endothelium, resulting in the fine-grained regulation of blood flow to ensure that neuronal metabolism is protected on a moment-to-moment basis via continuous precision delivery of energy substrates. Given that K_ATP channels do not seem to contribute to functional hyperemia to a diffuse visual stimulus,⁶⁹ it may be that strong stimuli driving a robust network activity to rapidly ramp energy demands are required to engage this mechanism under physiological conditions. Notably, a recent study using whisker stimulation did note an input to functional hyperemia of Kcnj8-containing K_ATP channels,³³ thus suggesting that stimulus form and brain region could be key variables in the recruitment of K_ATP channels to control blood flow. It is also possible that the pericyte energy switch is reserved for pathological conditions such as hypoglycemia, where it might serve as an emergency failsafe that has evolved to protect brain energy supply by increasing blood flow.

Our experiments in which we block glucose import using BAY-876 provide further support for these ideas by demonstrating that a stimulus that likely dramatically decreases parenchymal glucose triggers the pericyte energy switch in vivo. The presence of residual, much diminished, dilations to BAY-876 after inhibiting or inactivating K_ATP channels is likely attributable to vascular K_ATP channel-independent mechanisms. This is perhaps unsurprising, given the vital importance of timely glucose and oxygen delivery for the ongoing maintenance of neuronal health and function—it is highly likely that energy-sensing responsibilities are distributed between multiple cell types to provide redundancy.

K_ATP channels may contribute to control of blood flow through multiple distinct mechanisms

cECs have recently been shown to also possess functional K_ATP channels, albeit at a lower current density than pericytes.³² Interestingly, both ECs and pericytes were shown to increase cerebral blood flow in response to adenosine elevation via engagement of adenosine A_2a receptors and consequent protein kinase A phosphorylation of K_ATP channels. These findings contrast with those here, in which we identify a glucose-mediated energy-sensing mechanism through which pericyte K_ATP channels are activated, thus raising the possibility that multiple mechanisms converge on the control of K_ATP channels in these two closely associated cell types. Adenosine is released from neurons during their activity,⁷⁸ and also generated from the hydrolysis of ATP and ADP by abundantly expressed ectonucleotidases.⁷⁹ Given that adenosine has a long-established role in increasing cerebral blood flow to neuronal activation,⁸⁰ these pathways likely also engage pericyte and EC K_ATP channels to hyperpolarize the membrane and evoke upstream arteriolar dilation similar to that which we have shown here. Further experiments are needed to delineate the relative contributions of these different pathways to capillary K_ATP activity under different circumstances and to determine whether yet other mechanisms recruit this channel to control blood flow.

Recasting pericytes as metabolic sentinels

Our observation that pericytes are sensitive to glucose naturally evokes the question of whether these cells detect the levels of other energy substrates and metabolites. It is clear that pericytes exist in, and are influenced by, a rich milieu of molecules and substrates and that, therefore, their activity is likely regulated by a complex mix of factors that fluctuate in concentration over widely varying timescales. One such factor, partnered with glucose to support brain metabolism, might be oxygen. Oxidative phosphorylation relies on local oxygen tension, which is a direct function of local blood flow.⁸¹ The oxidation of glucose provides vastly more ATP than glycolysis, and neuronal activity is primarily powered by oxidative phosphorylation.⁸² It is, thus, possible that the pericyte energy switch may also be activated by local transient decreases in oxygen,⁸³ which might lead to an abrupt fall in intracellular ATP production, influencing the ATP:ADP ratio and engaging K_ATP channels. Interestingly, stalling behavior (i.e., complete cessation of RBC flux) is relatively common in brain capillaries, with approximately 0.45% of capillaries estimated to be stalled at any one time.⁸⁴ The function of this phenomenon is unclear, but it seems possible that these events would lead to a localized decrease in oxygen tension because of the lack of transiting RBCs loaded with oxygen. This may in turn activate the pericyte energy switch, leading to signaling to relieve the stall before it damages neurons. Pericytes might also possess mechanisms to assess local carbon dioxide gradients,^85,86 which would reflect the degree of local metabolic activity,⁸⁷ and modulate blood flow in turn. As demonstrated recently, pericytes can also sense lactate generated during glycolysis in ECs,⁸⁸ which could also serve as an energy substrate that ultimately regulates pericyte K_ATP channel activity.

Pericytes play a range of roles in the brain, which include control of barrier function,¹⁶ and regulation of gene expression,¹⁷ vascular development and stability,^89,90 and blood flow.^19,91 Moreover, they seem to be particularly vulnerable in the context of dementias⁹² and other disorders impinging on brain function (e.g., diabetes, hypertension, and kidney dysfunction⁹³), and contractile pericytes have been noted to die in rigor, which is thought to contribute to a loss of brain blood flow control, precipitating neuronal dysfunction and decline.¹⁹ Our data reveal a key energy-sensing mode for pericyte contributions to hemodynamic control in health and disease, stemming from the sensitivity of thin-strand pericytes to subtle metabolic changes. Importantly, if glucose drops below a critical threshold, a robust electrical response is generated through the recruitment of pericyte K_ATP channels to increase local blood flow, thereby providing more glucose to replenish local levels and protect ongoing neuronal function. This mechanism may be critical for supporting the heavy energetic demands of neuronal computation,⁹⁴ and its disruption over long periods could contribute to the mismatch between energy supply and demand that contributes to cognitive decline and dementia.^95–97 Intriguingly, a recent vessel isolation and nuclei extraction for sequencing (VINE-seq) atlas of human vascular cells suggests that the molecular players that take center-stage in the energy switch we have elucidated here (Kcnj8, Kcnj2, and Slc2a1) are each profoundly downregulated in Alzheimer’s cerebrovasculature, which could potentially disable protective responses to local glucose decreases and imperil neuronal metabolism.⁹⁸ Further work is needed to assess whether the pericyte energy switch is disabled in Alzheimer’s disease, and ongoing experiments in our laboratory are now directly addressing these questions.

Summary and conclusion

Thin-strand pericytes express a rich complement of ion channels and G-protein-coupled receptors that equips these cells to sense and respond to a wide range of stimuli.³⁹ K_ATP channels are the most abundant ion channel expressed by thin-strand pericytes,³⁹ and we demonstrate that their activation in response to decreased local metabolic substrate availability produces a robust increase in blood flow. Our data thus reveal pericytes as metabolic sentinels that act as a brain-wide energy-sensing network, continually monitoring substrate availability, and adjusting blood flow as necessary to protect ongoing neuronal health and function. In conditions like sporadic Alzheimer’s disease, brain glucose levels and metabolism are profoundly dysregulated,^99–102 and, thus, determining the impact of this on pericyte energy sensing and accompanying blood flow control may yield potential targets for improving clinical outcomes in neurological diseases with a significant vascular and metabolic component.

Limitations of the study

Given the need for fine-grained changes in glucose concentrations during microelectrode impalements, we chose to characterize the glucose-pericyte Vm relationship by impaling cells shortly after exposure to a desired glucose concentration, thus avoiding the difficulties associated with maintaining an impalement through many solution changes. The variability resulting from this approach may introduce uncertainties in the exact glucose-Vm responses of pericytes, raising the need for further developments to enable long-term recordings for a deeper characterization of this relationship. It is also important to note the possibility for off-target effects of BAY-876 on the brain surface, given the widespread distribution of GLUT1 transporters. To address this, we also delivered BAY-876 directly to the blood stream to take advantage of the blood-brain barrier. However, the paucity of information on the blood-brain barrier permeability of BAY-876 may necessitate further characterization of the precise cellular target mediating the GLUT-1 effects observed here.

STAR★METHODS

Detailed methods are provided in the online version of this paper and include the following:

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Dr. Thomas A. Longden (thomas.longden@som.umaryland.edu).

Materials availability

This study did not generate unique new reagents.

Data and code availability

All data reported in this paper will be shared by the lead contact upon request. This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Adult (2–3 mo. old) male and female C57BL/6J mice (Jackson Laboratories, strain 000664), Cspg4-DsRed mice (C57BL/6J background; Jackson Laboratories, strain 008241), tamoxifen-inducible Cspg4-Cre recombinase mice (C57BL/6J background; Jackson Laboratories, strain 008538), and Cspg4-Cre-K_IR6.1^AAA mice (donated by W. J. Lederer) were group-housed on a 12-h light:dark cycle with environmental enrichment and free access to food and water in the animal facility at the University of Maryland Baltimore. Tamoxifen-inducible Cspg4-Cre-K_IR6.1^AAA mice were generated by crossing K_IR6.1^AAA mice expressing dominant-negative K_IR 6.1^AAA with Cspg4-Cre recombinase mice.^48,49 All animals were genotyped with allele-specific PCR reactions prior to experiments. Littermates were randomly assigned to experimental groups. All animal care was conducted in accordance with, and procedures received prior approval from, the University of Maryland Institutional Animal Care and Use Committee.

METHOD DETAILS

K_IR6.1^AAA induction

4-OHT (Catalog number H7904, Sigma), the active metabolite of tamoxifen, was dissolved in a corn oil:ethanol solution (90:10% v/v) at a concentration of 2 mg/ml.¹⁰³ Cspg4-Cre-K_IR6.1^AAA mice were given either 4-OHT (10 mg/kg, intraperitoneal injection; K_IR6.1^AAA induction) or vehicle (corn oil:ethanol; vehicle control), and control K_IR6.1^AAA mice were given 4-OHT (10mg/kg; Cre control) once a day for 5 consecutive days. 4 weeks after the last injection, mice were imaged in vivo as described below.

In vivo imaging

Cranial window preparation and in vivo imaging was performed as previously described.^22,43 Briefly, mice were anesthetized with isoflurane (5% induction, 1.5–2% maintenance). 150 μL of FITC-dextran (10mg/mL; Catalog number FD2000S, Sigma) or TRITC-dextran (40 mg/mL; Catalog number T1287, Sigma) dissolved in saline was injected retro-orbitally. A midline incision was made on the scalp to expose the skull, and a titanium head plate was affixed over the left hemisphere with a combination of dental cement and superglue. On securing the headplate in a holding frame, a circular cranial window (~2 mm diameter) was drilled in the skull over the somatosensory cortex. The skull piece was removed, and the dura was carefully resected. The brain surface was irrigated as necessary with saline. Upon conclusion of surgery, isoflurane anesthesia was replaced with α-chloralose (50 mg/kg; Catalog number 23120, Sigma) and urethane (750 mg/kg; Catalog number U2500, Sigma). Systolic and diastolic blood pressure was measured using tail cuff plethysmography (CODA High Throughput System, Kent Scientific) and heart rate was measured using a paw sensor (PhysioSuite, Kent Scientific). Body temperature was maintained at 37°C throughout the experiment using a rectal probe feedback-controlled electric heating pad (Harvard Apparatus, USA). Physiological variables (mean ± SEM) are presented in Table S1. While blood gases were not measured in the present study, our previous work using the same surgical procedure and anesthetic regimen found that these parameters were within normal ranges.⁴³ Oxygenated and warmed (35–36°C) aCSF (124 mM NaCl, 3 mM KCl, 2 mM CaCl₂, 2 mM MgCl₂, 1.25 mM NaH₂PO₄, 26 mM NaHCO₃, 4 mM glucose) was superfused over the exposed cortex for the duration of the experiment at a rate of ~1 mL/min and continuously monitored at the window with a temperature probe. Images were acquired through an Olympus 20x infinity-corrected Plan Fluorite 1.0 NA water-immersion objective mounted on a Scientifica Hyperscope (Scientifica, UK) coupled to a Coherent Chameleon Vision II Titanium-Sapphire pulsed fs laser (Coherent, USA). FITC- and TRITC-dextran or DsRed were excited at 820 nm or 920 nm, respectively, and emitted fluorescence was separated through 525/50 and 620/60 nm bandpass filters. Single-plane imaging data to examine the time course of vessel diameter changes was collected at 30 Hz using a resonant scanning mirror. 3D imaging data were typically gathered using standard galvo mirrors. To measure RBC flux, we performed line scans at 1 kHz. Line scans were oriented along the lumen parallel to the flow of blood to maximize flux signal. For pressure-ejection of agents in aCSF (vehicle) onto pericytes or ECs or SMCs, a pipette containing the agent of interest and FITC or TRITC-dextran (to enable visualization) was maneuvered into the cortex and positioned adjacent to the cell under study, after which the solution was ejected directly at 8–12 psi, for 30 ms. This approach restricted agent delivery to the target cell and caused minimal displacement of the surrounding tissue. For pharmacological and staining experiments, agents of interest were applied to the brain surface for a minimum of 20 min to allow penetration. All in vivo imaging experiments were routinely ended with the application of aCSF containing 0 Ca²⁺ supplemented with 5 mM EGTA and 200 μM diltiazem to elicit maximal relaxation of SMCs and contractile pericytes to enable the measurement of maximum vessel diameters.

Microelectrode impalement of pericytes on isolated microvessels

Membrane potential measurements were made by impaling pericytes on microvessels isolated from Cspg4-DsRed mice using a papain-based Neural Tissue Dissociation kit (Miltenyi Biotec, USA), as described previously.^32,41 Cortical tissue from one hemisphere was carefully dissected and minced into small pieces with microscalpels in an isolation solution containing 55 mM NaCl, 80 mM Naglutamate, 5.6 mM KCl, 2 mM MgCl₂, 10 mM HEPES and 4 mM glucose (pH 7.3). Minced tissue was incubated with enzyme P from the kit for 18 min at 37°C, followed by addition of enzyme A, homogenization by passing through a Pasteur pipette ~10 times and incubation for 15 min at 37°C. The homogenate was then passed through a 21 G needle 7 times and incubated for 12 min at 37°C. The cell suspension was filtered through a 62-μm nylon mesh and stored in ice-cold isolation solution. Cells were transferred to a silicone elastomer (SYLGARD 182)-lined perfusion chamber, and allowed to adhere for ~45 min. The chamber was perfused with bath solution consisting of 137 mM NaCl, 3 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, 10 mM HEPES and 0–4 mM glucose (pH 7.4). DsRed-positive pericytes on small capillary segments were identified using fluorescence and brightfield microscopy, and impaled with a sharp microelectrode (pulled to ~100–200 MΩ) filled with 0.5 M KCl. Only recordings fulfilling the following criteria were considered for analysis: stable baseline prior to impalement, sharp negative deflection of membrane potential upon impalement, immediate return to 0 mV upon withdrawal of the electrode. Membrane potential was recorded using an AxoClamp 900A digital amplifier and HS-2 headstage (Molecular Devices, USA). Signals were digitized and stored using Axon Digidata 1550B and pClamp 9 software (Molecular Devices, USA).

Immunohistochemistry

Brains were extracted from Cspg4-DsRed mice that underwent cardiac perfusion with 4% paraformaldehyde. Tissues were stored in 4% paraformaldehyde overnight at 2–8°C and dehydrated in 30% sucrose in 1x phosphate buffered saline (PBS). Immunostaining and optical clearing of brain samples were performed according to a modified CUBIC clearing method.^104,105 Briefly, fixed brains were immunostained by first blocking non-specific binding with normal goat serum (Vector Laboratories, USA). Blocked samples were incubated overnight at 2–8°C with rabbit anti-SLC2A1 polyclonal antibody (1:500 dilution, catalog number HPA031345, Atlas Antibodies, Stockholm, Sweden) and developed with Alexa Fluor 488 goat anti-rabbit IgG secondary antibody (1:1000, catalog number A-11034, ThermoFisher Scientific, USA). Samples were cleared by incubation in CUBIC R1 solution (see ref. 104) at 37°C with shaking for 2–3 weeks, and then incubated in RIMS (refractive index matching solution; 88% w/v Histodenz in 0.02 M PBS with 0.01% sodium azide) at 37°C until the samples were optically clear (~5 days) with solution being replaced every 24 hours. Cleared tissue was mounted in RIMS and imaged with a Nikon W1 spinning disk confocal microscope.

QUANTIFICATION AND STATISTICAL ANALYSIS

Diameter measurements were analyzed offline using ImageJ software. Vessel diameter was calculated as the average of three measurements per vessel type made from Z-stacks of 3D volume recordings using the full-width at half-maximum method. RBC flux data were binned at 1-s intervals and analyzed using SparkAn software (A. Bonev, University of Vermont). For pressure-ejection experiments, mean baseline diameter and flux were obtained by averaging the baseline for each measurement before ejection of pinacidil or aCSF, and peak diameter and RBC flux change was defined as the largest change from mean baseline. The distance from the site of pressure ejection to the feed arteriole or to the nearest pericyte cell body was measured using the Simple Neurite Tracer plugin for ImageJ software.¹⁰⁶ Statistical testing was performed using GraphPad Prism 7 software. Data are expressed as means ± s.e.m., and a p-value ≤0.05 was considered significant. Stars denote significant differences; ‘n.s.’ indicates comparisons that did not achieve statistical significance. Depending on the data set, two-sided paired or unpaired Student’s t-tests and one-way ANOVA with Dunnett’s or Sidak’s multiple comparison test were used. Statistical tests and sample sizes (denoted as ‘n’) are noted in figure legends. Statistical methods were not used to pre-determine sample sizes, and sample size was estimated based on similar experiments performed previously in our laboratory. Experiments were repeated to adequately reduce confidence intervals and avoid errors in statistical testing. Data collection and analysis were not performed blinded to the conditions of the experiments; no further randomization was performed. No data were excluded.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-SLC2A1 rabbit polyclonal antibody	Atlas Antibodies	Cat#HPA031345; RRID:AB_2673835
Alexa Fluor 488 goat anti-rabbit IgG secondary antibody	ThermoFisher Scientific	Cat#A-11034; RRID:AB_2576217
Chemicals, peptides, and recombinant proteins
4-hydroxy tamoxifen	Sigma	Cat#H7904
Corn oil	Sigma	Cat#C8267
Ethanol	Sigma	Cat#E7023
FITC-dextran	Sigma	Cat#FD2000S
TRITC-dextran	Sigma	Cat#T1287
α-chloralose	Sigma	Cat#23120
Urethane	Sigma	Cat#T1287
Sodium chloride	Sigma	Cat#S9888
Potassium chloride	Sigma	Cat#P5405
Calcium chloride dihydrate	Sigma	Cat#C5050
Magnesium chloride	RPI	Cat#M24000
D-Glucose	Sigma	Cat#G8270
Sodium bicarbonate	Sigma	Cat#S5761
Sodium phosphate monobasic	Sigma	Cat#S9638
EGTA	Sigma	Cat#E4378
Pinacidil hydrate	Cayman chemical	Cat#15416
Glibenclamide	Sigma	Cat#G0639
BAY-876	Tocris bioscience	Cat#6199
L-Glutamic Acid monosodium salt monohydrate	Sigma	Cat#49621
HEPES	Sigma	Cat#H3375
Paraformaldehyde	Sigma	Cat#158127
Sucrose	Sigma	Cat#S9378
Histodenz	Sigma	Cat#D2158
Sodium azide	Sigma	Cat#S2002
Critical commercial assays
Neural Tissue Dissociation kit (P)	Miltenyi biotech	Cat#130-092-628
Experimental models: Organisms/strains
Mice, C57BL/6J	Jackson Laboratories	JAX: 000664; RRID:IMSR_JAX:000664
Mice, Cspg4-DsRed	Jackson Laboratories	JAX: 008241; RRID:IMSR_JAX:008241
Mice, Cspg4-Cre recombinase	Jackson Laboratories	JAX: 008538; RRID:IMSR_JAX:008538
Mice, KIR6.1AAA	Donated by Dr. W J Lederer	N/A
Mice, Cspg4-Cre-KIR6.1AAA	This paper	N/A

Highlights.

• Focal activation of thin-strand pericyte K_ATP channels dilates the upstream arteriole

• Brain capillary thin-strand pericytes monitor local glucose levels

• A pericyte K_ATP channel energy switch couples glucose changes with hyperemia

• This pericyte electro-metabolic signaling may protect neuronal health and function