Contractile determine the direction of flow at junctions

Albert L. Gonzalesa, Nicholas R. Kluga, Arash Moshkforoushb, Jane C. Leec, Frank K. Leec, Bo Shuic, Nikolaos M. Tsoukiasb,d, Michael I. Kotlikoffc, David Hill-Eubanksa, and Mark T. Nelsona,e,1

aDepartment of Pharmacology, University of Vermont, Burlington, VT 05404; bDepartment of Biomedical Engineering, Florida International University, Miami, FL 33199; cDepartment of Biomedical Sciences, College of Veterinary Medicine, Cornell University, Ithaca, NY 14853; dSchool of Chemical Engineering, National Technical University of Athens, 157 72 Zografou, Greece; and eDivision of Cardiovascular Sciences, University of Manchester, M13 9PL Manchester, United Kingdom

Contributed by Mark T. Nelson, September 2, 2020 (sent for review December 30, 2019; reviewed by William F. Jackson and Andy Shih) The essential function of the is to continuously Pericytes are ubiquitous in the capillary of all and efficiently supply the O2 and nutrients necessary to meet the vascular beds, reaching their highest densities within retinal and metabolic demands of every cell in the body, a function in which cerebral circulations (2). The defining morphological charac- vast capillary networks play a key role. Capillary networks serve teristics of a capillary , first depicted in meticulous hand an additional important function in the : drawings in the early 20th century and later described in detail in acting as a sensory network, they detect neuronal activity in the electron micrographs (3, 4), are a prominent outward protruding form of elevated extracellular K+ and initiate a retrograde, propa- nucleus that aligns with the vessel , extensions that span gating, hyperpolarizing signal that dilates upstream to the long axis, and projections that wrap around the endothelial rapidly increase local blood flow. Yet, little is known about how tube. This contrasts with all differentiated cells blood entering this network is distributed on a branch-to-branch described in the vasculature, which exhibit a morphology char- basis to reach specific in need. Here, we demonstrate that acterized by a long, fusiform cell body that adopts a ring-shaped capillary-enwrapping projections of junctional, contractile peri- structure in vivo that encircles the endothelial cell layer in ar- cytes within a postarteriole transitional region differentially con- teries and arterioles, with its long axis oriented perpendicular to strict to structurally and dynamically determine the morphology of the direction of blood flow.

capillary junctions and thereby regulate branch-specific blood The consumes a disproportionate share of the body’s PHYSIOLOGY flow. We further found that these contractile pericytes are capable energy resources and is highly sensitive to even brief disruptions of receiving propagating K+-induced hyperpolarizing signals prop- in blood flow. Because neurons lack the capacity to store sig- agating through the capillary network and dynamically channeling nificant energy reserves, the brain has evolved on-demand red blood cells toward the initiating signal. By controlling blood mechanisms for preferentially allocating blood flow to regions flow at junctions, contractile pericytes within a functionally dis- of higher neuronal activity. Such activity-dependent increases in tinct postarteriole transitional region maintain the efficiency and local blood flow (functional hyperemia) are mediated by an en- effectiveness of the capillary network, enabling optimal semble of mechanisms collectively termed neurovascular cou- of the brain. pling (NVC). The fact that NVC mechanisms exist suggests that simply oversupplying the entire brain is an inadequate evolution- functional hyperemia | cerebral blood flow | pericytes ary solution to the neuronal energy-resupply problem, implying that precision and efficiency are organizing principles that govern he fundamental purpose of the circulatory system is to pro- Tvide an uninterrupted supply of O2 and nutrients to all cells Significance of the body and to remove CO2 and other metabolic waste products. , which constitute the vast majority of the Capillaries—the most abundant vessels in the circulatory — vasculature in terms of length, are the sites of gas and nutrient system deliver O2 and nutrients to all cells of the body. In the exchange between the blood compartment—including O2-carrying brain and , capillaries also act as a sensory web that de- red blood cells (RBCs)—and the surrounding . Despite a tects neuronal activity. Here, we demonstrate that pericytes general appreciation of the relationship between the requirements localized at capillary junctions in a postarteriole transitional of tissues and the microvasculature that serves them, how RBCs region possess unique properties, notably including contrac- tility, that enable them to dynamically manipulate capillary and plasma are efficiently distributed throughout capillary net- branch diameters and exert fine control over the distribution of works so as to meet the needs of every cell remains poorly un- blood within the capillary network. In so doing, these con- derstood. Nowhere is our understanding of the mechanisms that tractile junctional pericytes fine tune the delivery of O2 and regulate the distribution of blood within capillary networks less nutrients and thus serve to meet the specific needs of neurons. complete than in the brain. Given these unique properties, pericytes represent a therapeutic The brain vasculature is composed of a network of inter- target for cardiovascular and neurodegenerative diseases. connected surface (pial) vessels that give rise to arterioles that penetrate orthogonally into the brain and feed a vast network of Author contributions: A.L.G. designed research; A.L.G. and N.R.K. performed research; capillaries. Arterioles are composed of an inner layer of endo- J.C.L., F.K.L., B.S., and M.I.K. contributed new reagents/analytic tools; A.L.G., A.M., and N.M.T. analyzed data; A.L.G., N.M.T., D.H.-E., and M.T.N. wrote the paper; A.M. wrote thelial cells (ECs), oriented in the direction of blood flow, sur- MatLab code for analysis; and M.T.N. was the principal investigator. rounded by a single layer of smooth muscle cells that wrap Reviewers: W.F.J., Michigan State University; and A.S., University of Washington. circumferentially around the endothelial cell layer and are sep- The authors declare no competing interest. arated from it by an (IEL) (1). Capillaries, Published under the PNAS license. on the other hand, are composed of endothelial cell tubes 1To whom correspondence may be addressed. Email: [email protected]. without a smooth layer or IEL; instead, much of their This article contains supporting information online at https://www.pnas.org/lookup/suppl/ surface is covered over with perivascular mural cells (pericytes), doi:10.1073/pnas.1922755117/-/DCSupplemental. which are embedded in the .

www.pnas.org/cgi/doi/10.1073/pnas.1922755117 PNAS Latest Articles | 1of12 Downloaded by guest on October 1, 2021 the operation of these mechanisms. In fact, such precision is a containing both diverging and converging junctions, with seg- fundamental assumption underpinning brain-activity mapping based ments ultimately reuniting to drain blood back into the venous on measurements of blood level-dependent functional circulation. In the retina, arterioles radiating from the optical magnetic resonance imaging (BOLD-fMRI). Our recent work has disk, visualized using hydrazide staining to preferentially stain established a mechanistic basis for the -to-microvascular the IEL (present only in and arterioles), are analogous signaling necessary for efficient communication of neuronal meta- to the penetrating arterioles found in the cerebral circulation bolic demands to the cerebral vasculature, showing that activation (18) (Fig. 1A). Thus, in this formulation, all vascular elements of capillary endothelial cell inwardly rectifying potassium (Kir) downstream of these radiating arterioles are considered capil- + channels by K , a byproduct of neuronal activity, induces a prop- laries. Within these so-defined capillary regions, all perivascular agating electrical (hyperpolarizing) signal that causes upstream ar- cells with a protruding nucleus and cell body located atop the teriolar dilation and increased blood flow into the capillary network vessel are considered to be pericytes. (5). However, although this NVC mechanism provides a means of In addition to their classic “bump on a log” appearance, peri- communicating the need for increased blood flow toward a meta- cytes possess cytoplasmic extensions that spread along, and pro- bolically active anatomical region, it leaves open the question of jections that wrap around, the capillary tube (8) (Fig. 1B). These how blood flow is regulated within the capillary network. Whereas pial vessels are interconnected, and thus have con- siderable capacity for redirecting blood flow, a single parenchy- mal and associated capillary networks feed blood to a distinct cylindrical cortical volume (diameter, ∼500 μm) (6). In the absence of control mechanisms downstream of the arteriole, this arrangement predicts a stochastic distribution of blood within a volume of nonuniform neural activity, and thus has implications for the precision of blood delivery. An alternative mechanism tested here is that contractile junctional pericytes dynamically control capillary branch diameters to exert fine control over the distribution of blood. Pericytes within the cerebral microcircula- tion are predominantly located at capillary junctions (7, 8); this is also the case for the microvasculature of the retina, which shares the same developmental origin as the brain vascula- ture (9). Our data support the conclusion that these junctional pericytes, specifically those in a specialized postarteriole transi- tional region, both structurally and dynamically determine the geometry of capillary junctions and regulate the directional dis- tribution of RBCs through the capillary network. Our findings further suggest that relaxation of proximal pericytes by K+-de- pendent hyperpolarizing signals plays a role in NVC mechanisms, directing blood flow toward the source of the signal. These mechanisms challenge the view that blood flow within the capillary network is essentially a passive process determined by the static architecture of the vasculature and suggest instead that pericytes provide dynamic control of blood perfusion in capillary networks to fine tune the delivery of O2 and nutrients to the tissue. Results Pericytes within CNS Capillary Networks Show Region-Specific Differences in Morphology and Distribution and Contribute to the Asymmetry of Capillary Branch Diameters. The nature and func- tion of pericytes has been a matter of controversy since these cells were first described by Rouget in 1873 (10). Speculation surrounding their potential role in regulating capillary blood flow, which began in earnest nearly a century ago (11, 12), has simmered to this day, with considerable bodies of literature supporting divergent points of view (for early reviews, see refs. 7, 13–17). There are also differences in opinion on the point at Fig. 1. The morphology of pericytes confers asymmetry on capillary junc- tions within retinal capillary networks. (A) Representative low-magnification which the arteriolar system ends and the capillary network — image identifying the sharp boundary (*) between arteriole and capillary begins differences that are intertwined with views on pericyte- vessels of the retinal vascular network stained with FITC-conjugated lectin related questions. Thus, any meaningful discussion of pericytes (green) and Texas Red-hydrazide (red). (Scale bar, 50 μm.) (B) Schematic in central nervous system (CNS) capillary networks must begin depiction (Left) and fluorescence images (Right) of the retinal vasculature of with a shared sense of what is meant by “pericytes” and “capil- NG2-dsRed mice stained with FITC-conjugated lectin showing arterioles and laries.” In terms of the brain microvaculature, we define vessels transitioning pericyte morphology from ensheathing to mesh and thin- furthest removed from surface pial arteries that are completely strand type. (Scale bars, 5 μm.) (C, Left) Representative low-magnification covered by a single continuous layer of concentric smooth muscle image of the retinal vascular network from an NG2-dsRed mouse stained μ cells as feeding arterioles. The transition from this last arteriole with FITC-conjugated lectin (green). (Scale bar, 50 m.) (Right) Summary data showing the percentage of capillary junctions containing cells positive for segment to the first segment of the capillary network is marked the pericyte marker NG2 and the ratio of daughter branch diameters, by a sharp boundary formed by the presence of an IEL on one expressed as DiameterSmall/DiameterLarge. The presence of NG2-positive cells side (arteriole) and its absence on the other (capillary). From the was determined from 526 junctions from nine confocal stacks. The ratio of single capillary segment emerging from the feeding arteriole, daughter branch diameters was determined from 102 junctions from 11 capillary networks branch out to form an interconnected web vascular trees (n = 3 to 4 mice).

2of12 | www.pnas.org/cgi/doi/10.1073/pnas.1922755117 Gonzales et al. Downloaded by guest on October 1, 2021 pericyte projections can take different forms, producing a con- transitional region modulate capillary diameters and contribute to tinuum of pericyte morphologies within the capillary network that the general structural asymmetry of perijunctional vessel diameters. can be broadly assigned to three subtypes. The most arteriole- proximate segments of the capillary network are occupied by The Arteriole-Proximate Region Constitutes a Functionally Distinct pericytes with short extensions and densely packed projections Domain of the CNS Capillary Bed. We next investigated potential that wrap around segments of the capillary (4), almost completely molecular correlates of region-dependent differences in pericyte encasing it (Fig. 1B). Superficially, this “dense,” or ensheathing, morphology and distribution within CNS capillary networks by morphology appears similar to that of smooth muscle cells in ar- analyzing the expression of a panel of structural and marker pro- teries and arterioles, but instead of individual vascular smooth teins. Using the mouse retinal preparation, we found that pericytes muscle cells, the dense banding pattern observed in these seg- throughout the capillary tree were positive for filamentous , ments reflects wrapping of bands of tightly packed projections although staining intensity was significantly lower in capillaries from single pericytes. This region is sometimes referred to as the compared with that in arterioles (Fig. 2A and SI Appendix,Fig.S2). “precapillary arteriole” (19), a term which gives an impression that Notably, only pericytes in the postarteriole transitional region the ensheathing pericytes in this region are smooth muscle cells, a showed immunostaining for α-actin (Acta2), the dynamically con- distinction addressed in greater detail below. Beyond this region tractile isoform of actin (Fig. 2B). Consistent with previous ob- of the capillary tree, dense pericytes transition to a second “mesh” servations (19, 22–24), quantification of immunostaining revealed type that shows considerable capillary coverage or longer exten- that α-actin immunofluorescence intensity exhibited a stepwise sions but lacks a prominent dense banding pattern, and later into a decrease in pericytes at first, second, and third junctions within the third, “loose” or “thin-strand” morphology (8) with long exten- capillary network compared with that in the feeding arteriole sions and fewer and minimal projections wrapping around the capillary tube (Fig. 1B). To further assess the regional and structural distribution of microvascular pericytes, we quantified pericyte coverage in retinal and brain capillary networks using NG2-DsRed-BAC transgenic mice, which express DsRed (Discosoma red fluorescent protein) under the control of the promoter for the Ng2 gene, encoding the putative pericyte marker neural/glial antigen 2 (NG2) (20). The retinal microvascular network was visualized by confocal imaging PHYSIOLOGY of flat-mounted retinal preparations stained with fluorescence- conjugated isolectin B4, which binds to D-galactose residues within the basement membrane (21). In line with previous reports (7, 8), we found that ∼52% of pericytes throughout the capillary network were located at capillary bifurcations (n = 581 pericytes from 10 confocal stacks). Further analysis revealed that ∼93% of junctions in the arteriole-proximate region of the capillary net- work corresponding roughly to the region occupied by pericytes with ensheathing or mesh morphologies (approximately first through third junctions), referred to hereafter as the “postarteriole transitional region,” contained NG2-positive cell bodies, whereas only ∼55% of junctions in more distal regions (fourth junction or beyond) were occupied by such cells (Fig. 1C). A transcranial two- photon laser-scanning microscopy (2PLSM) examination of the somatosensory cortex microvasculature of NG2-DsRed-BAC trans- genic mice, injected with fluorescein isothiocyanate (FITC)-dextran to illuminate the brain vasculature, revealed a similar distribution of pericytes at vascular junctions. In this case, ∼90% of junctions in the transitional region contained NG2-positive cells, whereas ∼45% of more distal capillary bifurcations were occupied by such cells (SI Appendix,Fig.S1). The common occurrence of pericytes at capillary junctions, particularly the extensive coverage at junctions in the post- arteriole transitional region, led us to consider the possibility that these cells might influence the static geometry of capillary branches. As a first step in determining the relationship between pericyte junctional coverage and branch symmetry, we assessed differences in the luminal diameters of downstream (daughter) branches, measured as the ratio of the smaller to larger branch Fig. 2. Cytoskeletal elements within a feeding retinal arteriole and capillary (DiaSmall/DiaLarge), ex vivo in the retinal preparation and in vivo networks. (A to E) Stitched together montage of high resolution fluores- in the brain. Interestingly, branch diameter symmetry increased cence images (Top) and whisker-box plot (Bottom) showing the retinal with increasing distance of junctions from the feeding arteriole in vasculature stained with FITC-conjugated lectin (gray) and labeled with both retinal (Fig. 1C)andbrain(SI Appendix,Fig.S1) capillary phalloidin, a pan-specific stain for filamentous actin (from 5 vascular trees) (A); immunostained for α-actin (n = 11 vascular trees) (B); immunostained for networks. In the retinal preparation, DiaSmall/DiaLarge values were ∼ calponin (n = 12 vascular trees) (C); stained with Tubulin Tracker, for fila- 0.65, 0.73, and 0.87 at first, second, and third junctions, respec- = > mentous microtubules (n 5 vascular trees) (D); and imaged in SMMHC tively, and plateaued at 0.9 at more distal (fourth and fifth) (Myh11)-tdTomato mice (n = 5 vascular trees) (E). Capillary junctional fluo- junctions. The fact that the symmetry of branch diameters in- rescence was calculated as described in Methods.(*P ≤ 0.05 vs. smooth creased in parallel with decreasing pericyte occupancy led us to muscle cells [SMC], #P ≤ 0.05 vs. first junction, $P ≤ 0.05 vs. second junction, Ϯ hypothesize that pericytes at junctions in the postarteriole and P ≤ 0.05 vs. third junction). (Scale bars, 10 μm.)

Gonzales et al. PNAS Latest Articles | 3of12 Downloaded by guest on October 1, 2021 (Fig. 2 B, Bottom), reaching undetectable levels at fourth junc- tions and beyond. Similar fluorescence patterns were observed in vivo and ex vivo using acta2-GCaMP-mCherry as well as acta2-GCaMP-mVermillion mice, in which fluorescence is driven by the Acta2 promoter (SI Appendix,Figs.S3,S4,andS6), which differs from findings of others (24). Perivascular cells within the retinal capillary bed also expressed a number of additional markers in common, including desmin (25), and shared a lack of calponin (Fig. 2C) (25, 26). Curiously, although these cells express tubulin monomers (26, 27), they lacked polymerized tubulin (i.e., microtubules) (Fig. 2D), unlike most cell types studied to date. Notably, many of these attributes, including the absence of large microtubule filaments, are strikingly different from those of smooth muscle cells in the upstream arteriole (summarized in SI Appendix, Table S1). Similar to the case for the retinal prepara- tion, we found that expression of α-actin in pericytes in the cere- bral circulation was limited to capillary branches proximal to the feeding arteriole (SI Appendix,Fig.S4A). Intriguingly, all peri- vascular cells within the retinal and cerebral capillary microcir- culation, including those deeper in the capillary bed that are universally recognized as pericytes, expressed the smooth muscle “specific” protein, myosin heavy chain (Myh11) (Fig. 2E and see SI Appendix,Fig.S4B) (26, 28). Although very low level Myh11 promotor activity can lead to a strong fluorescence signal, the fact that fluorescence intensity tended to decrease at successive junc- tions of capillaries suggests that the observed signals are a meaningful reflection of bona fide Myh11 expression. The fact that expression of the dynamically contractile α-isoform of actin is limited to the subpopulation of pericytes in the postarteriole transitional region (Fig. 1D) (23) predicts that only pericytes within this region would be capable of dynamically constricting or dilating branches of capillary bifurcations. To test this, we performed real-time measurements of changes in cap- illary branch diameter at proximal (first to third) and distal (fourth or beyond) junctions in fluorescence-conjugated isolectin B4-stained retinal preparations in response to two different stimuli: the G protein-coupled thromboxane A2 receptor ago- nist, U46619, and membrane depolarization with 60 mM K+. All pericytes at the first two to three (2.1 ± 0.2 junctions, 115 cap- illary trees from 17 confocal stacks) capillary junctions proximal to the feeding arterial rapidly contracted following administra- Fig. 3. U46619-induced dynamic contraction of pericytes in the post- tion of 100 nM U46619 or membrane potential depolarization arteriole transitional region. (A and B) Representative images (Left), time with 60 mM KCl (Fig. 3A), resulting in average decreases in di- course (Middle), and summary data (Right) showing capillary constriction ameter of ∼35% (U46619) and ∼15% (60 mM K+); they also following administration of U46619 (100 nM; red) or membrane depolar- dilated following exposure to a Ca2+-free extracellular solution ization with 60 mM KCl (blue), and immediate vessel relaxation to applica- 2+ 2+ = (Fig. 3 A, Right). Pericytes at capillary bifurcations more than four tion of zero extracellular Ca ([Ca ]o 0, green) mediated by pericytes in junctions distal to the feeding arteriole also responded to U46619, the postarteriole transitional region (proximal) (A) and at locations deeper in albeit much more slowly and to a lesser extent, gradually con- the capillary bed (distal) (B). Average starting baseline diameter indicated by dotted line (n = 30 junctions, n = 5 to 6 mice; *P ≤ 0.05 vs. U56619 and #P ≤ stricting by ∼5% over 10 to 15 min; however, they did not respond μ 2+ 0.05 vs. 60 mM KCl). (Scale bars, 5 m.) (C and D) Summary data showing to 60 mM KCl or Ca -free solution (Fig. 3B and SI Appendix, Fig. U46619 (100 nM)-induced pericyte-mediated capillary constriction in the S5). Notably, pretreatment with the actin-depolymerizing agents presence of cytocholasin D (5 μM) and latrunculin B (1 μM; orange) (n = 27 to cytocholasin D (5 μM) and latrunculin B (1 μM) or the myosin 38 junctions from n = 5 mice) (C) or the MLCK inhibitor ML-7 (5 μM, purple) light chain kinase (MLCK) inhibitor, ML-7 (5 μM), prevented the (n = 17 to 23 junctions from n = 4 mice) (D). (E) Summary data showing U46619-induced rapid contraction of junctional pericytes in the relaxation of arterioles and pericytes in the postarteriole transitional region transitional region and the slow contraction of distal pericytes (proximal) and more distal locations (distal), following microtubule depoly- merization with nocodazole (10 μM, light blue) (n = 8 vascular trees from n = (Fig. 3 C and D). Consistent with the absence of detectable mi- ≤ crotubule filaments in pericytes (Fig. 2D), microtubule depoly- 4 mice; *P 0.05 vs. nocodazole and U46619 treatment). merization with nocodazole (10 μM) had no effect on the relaxation of junctional pericytes following exposure to a Ca2+- Ca2+-Dependent Contraction of Pericytes in the Postarteriole free extracellular solution but did prevent the relaxation of arterial 2+ smooth muscle cells in feeding arterioles (Fig. 3E). Thus, in con- Transitional Region Is Regulated by IP3R and L-Type Ca Channel trast to observations in every muscle cell examined to date (29, Activity, but Not by Ryanodine Receptor Activity. Contraction, ini- 30), microtubules appear to play no role in the relaxation or tiated by myosin and actin cross-bridge cycling following phos- morphology of capillary pericytes in the transitional region. phorylation of the myosin regulatory light chain by MLCK, requires 2+ Collectively, these observations establish vascular segments in an increase in intracellular Ca concentration (31). U46619 acts α the postarteriole transitional region containing dynamically con- through the G q/11-coupled thromboxane receptor to activate tractile pericytes as a functionally distinct region of the CNS phospholipase C (PLC), which hydrolyzes phosphatidylinositol capillary network. 4,5-bisphosphate (PIP2) to diacylglycerol (DAG) and inositol

4of12 | www.pnas.org/cgi/doi/10.1073/pnas.1922755117 Gonzales et al. Downloaded by guest on October 1, 2021 1,4,5-trisphosphate (IP3), the latter of which engages IP3 receptors the postarteriole transitional region (Fig. 4 C and D). However, 2+ (IP3Rs) to promote release of Ca from intracellular stores (32), nimodipine treatment only partially inhibited U46619-induced whereas 60 mM K+ causes depolarization through activation of Ca2+ events and did not fully relax pericytes precontracted with voltage-dependent L-type Ca2+ channels, allowing influx of ex- U46619 (Fig. 4 E and F), suggesting that other sources of Ca2+ 2+ 2+ tracellular Ca . To obtain a better understanding of the Ca besides voltage-dependent Ca2+ channels (35) contribute to agonist- dynamics that lead to pericyte contraction, we examined the fre- induced contraction. To further test this, we examined pericyte- 2+ quency of agonist- and depolarization-induced Ca events pre- localized Ca2+ events following depletion of intracellular Ca2+ ceding contractions in three to six projections of pericytes in the stores or blocking endoplasmic reticulum (ER) Ca2+ release via first three segments of the postarteriole transitional region IP3Rs and ryanodine receptors (RyRs). Pretreatment with the (Fig. 4A). These experiments were performed in retinal prepara- sarco-endoplasmic reticulum Ca2+-ATPase (SERCA) inhibitor, tions from acta2-GCaMP-GR BAC transgenic mice, which express μ 2+ 2+ 2+ cyclopiazonic acid (CPA; 10 M), prevented U46619-induced Ca a modified GCaMP Ca sensor fused to the Ca -insensitive events and constriction (Fig. 4 G and H), implying that intracellular fluorescent protein mCherry under control of the Acta2 promoter Ca2+ stores also contribute to agonist-induced contraction of (33). We found that U46619 (100 nM) increased the average fre- pericytes. Consistent with this, administration of the membrane quency of Ca2+ events in projections of proximal pericytes from permeable IP3 analog Bt-IP3 (10 μM) increased the overall fre- 1.4 ± 0.2 (baseline) to 2.6 ± 0.2 events per 10 s (Fig. 4B). U46619 2+ 2+ quency of Ca events, and inhibition of IP3Rs by preincubation (100 nM) also stimulated Ca events in pericyte projections of μ distal pericytes, increasing their average frequency from 1.9 ± 0.2 with xestospongin C (1 M) blunted the effectiveness of sub- ± sequent U46619 treatment, reducing the frequency of U46619- (baseline) to 3.2 0.2 events per 10 s (SI Appendix,Fig.S6). Be- ∼ cause these distal pericytes do not express α-actin, we performed stimulated events by 50% (Fig. 4I). Notably, application of the 2+ RyR inhibitor tetracaine (100 nM) had no effect on baseline or these Ca imaging experiments using SMMHC-GCaMP6f trans- 2+ genic mice, which express a modified GCaMP Ca2+ sensor under U46619-stimulated Ca events in pericytes in the postarteriole Myh11 transitional region (Fig. 4J). The RyR activator caffeine (5 mM) control of the promoter of the gene encoding SMMHC 2+ (smooth muscle myosin heavy chain). Depolarization with 60 mM similarly had no effect on Ca in these pericytes but did increase 2+ K+ produced an effect similar to that of U46619 in proximal intracellular Ca in neighboring cells in pericytes, increasing the frequency of Ca2+ events from 1.7 ± 0.1 the feeding arteriole (Fig. 4K). These observations imply that (baseline) to 3.8 ± 0.2 events per 10 s (Fig. 4C). Consistent with RyRs, a defining feature of muscle cells, are functionally absent in previous reports of a functional role for L-type voltage-dependent pericytes within the transitional region, reinforcing the idea that, Ca2+ channels in pericytes (34), the selective L-type channel an- despite certain overt similarities, pericytes—including those in the PHYSIOLOGY tagonist nimodipine eliminated high K+-induced Ca2+ events and postarteriole transitional region—differ from smooth muscle cells completely relaxed precontracted (with 60 mM K+) pericytes in in key functional respects.

2+ Fig. 4. L-type Ca channels and IP3Rs, but not RyRs, contribute to the contractile dynamics of pericytes in the postarteriole transitional region. (A and B) Rep- resentative images and traces (A) and summary data (B) showing Ca2+ events in pericytes (P1, P2, and P3) following administration of 100 nM U46619 (n = 40 ROIs from 12 cells, n = 3 mice; *P ≤ 0.05 vs. baseline). (C) Summary data showing the frequency (freq.) of Ca2+ events at baseline, and following 60 mM K+- induced membrane depolarization in the absence and presence of 100 nM nimodipine (n = 36 ROIs from 12 cells, n = 4 mice; *P ≤ 0.05 vs. baseline and #P ≤ 0.05 vs. 60 mM KCl). (D) Summary data showing contraction of projections in response to K+-induced membrane depolarization (60 mM KCl) in the ab- sence and presence of 100 nM nimodipine (n = 9 cells, n = 3 mice. *P ≤ 0.05 vs. 60 mM KCl). (E) Summary data showing the frequency of Ca2+ events in the presence of nimodipine (n = 30 ROIs from 10 cells, n = 4 mice. *P ≤ 0.05 vs. nimodipine), without and with added U46619. (F) Summary data showing U46619-induced contractions in the absence and presence of nimodapine (100 nM) (n = 14 cells, 4 mice; *P ≤ 0.05 vs. U46619). (G) Summary data showing the frequency of Ca2+ events in the pres- ence of cyclopiazonic acid (CPA, 10 μM), without and with added U46619. (H) Summary data showing contraction of projections in the presence of CPA (10 μM), without and with added U46619 (n = 7 cells, n = 3 mice; *P ≤ 0.05 vs. U46619). (I) Summary data showing the contribution of ER Ca2+ to Ca2+ events. (Left)Ca2+ event frequency at baseline and follow-

ing administration of Bt-IP3 (10 μM) (n = 28 ROIs from 5 cells, n = 3 mice; *P ≤ 0.05 vs. baseline). (Right)Ca2+ event frequency in the presence of 1 μM xestospongin C (Xesto C), without and with added U46619 (n = 25 ROIs from 6 cells, n = 3 mice). (J) Summary data showing Ca2+ event frequency in the presence of 100 nM tetracaine, without and with added U46619 (n = 47 ROIs from 12 cells, n = 6 mice; *P ≤ 0.05 vs. baseline). (K) Representative image (Left) and trace (Right)ofCa2+ in smooth muscle (arteriole, red) and pericytes in the postarteriole transitional region (blue) following administration of caffeine (5 mM) and ionomycin (10 μM). The trace shows the average and SE of five arteriole/pericyte preparations (n = 5 vascular trees, n = 4 mice).

Gonzales et al. PNAS Latest Articles | 5of12 Downloaded by guest on October 1, 2021 Projections of Contractile Junctional Pericytes Are Capable of Acting branches (Fig. 5D and SI Appendix,Fig.S7). Localized Ca2+ ac- as Individual Functional Units to Differentially Contract Daughter tivity between ROIs within a projection or between projections Branches and Direct RBC Flux at Capillary Bifurcations. We next di- within the same capillary branch was more highly correlated than rectly assessed Ca2+ signaling in individual pericyte projections that between projections located on different capillary branches. in relation to contractile responses under basal (unstimulated) Thus, projections enwrapping different capillary branches are ca- conditions in the retinal preparation from acta2-GCaMP-GR trans- pable of exhibiting independent Ca2+ events, allowing them to act genic mice, utilizing Ca2+-insensitive mCherry fluorescence to con- as independent functional units to differentially constrict capillary firm that changes in fluorescence are attributable to changes in Ca2+ branches at bifurcation points. levels and not to contraction-associated movement artifacts. A 3D Next, using 2PLSM and 3D reconstructions of z-stack images reconstruction of a single pericyte revealed projections wrapping (Fig. 6A), we examined the relationship between Ca2+ signals around all three junctional branches (Fig. 5 and Movie S1). This and contraction of pericyte projections at junctions in the tran- morphology is typical of pericytes in the transitional region, which sitional region in an in vivo setting. Consistent with Ca2+-driven exhibited an average of 8.0 ± 0.4 projections per pericyte (n = 21 contractions, we found that the frequency of Ca2+ events in pericytes, n = 3 mice). Simultaneous tracking of increases in local pericyte projections was correlated with decreases in the lumen Ca2+ (Fig. 5, upward deflection) and capillary diameter (Fig. 5, diameter of the corresponding branch (Fig. 6B), such that peri- shaded downward deflection) in all projections of an individual cyte projections encircling smaller-diameter (constricted) capil- branch-covering junctional pericyte revealed that each projection lary segments exhibited a higher frequency of Ca2+ events than from a single pericyte exhibited distinct Ca2+ signaling and con- those encircling larger-diameter (dilated) segments. To further traction profiles (Movie S2), suggesting that individual projections determine how differential control of branch diameters ex vivo are capable of acting as independent functional units. Additional translates to differential control of blood flow in vivo, we mea- examples of Ca2+ and contractile events in individual projections sured the fractional distribution of RBCs at junctions—the pro- of proximal pericytes are depicted in Movie S3.UsingaPearson’s portion of cells flowing down each branch—following application correlation analysis, we examined whether local Ca2+ events of a contractile stimulus directly to individual junctional pericytes. within and between projections of an individual junctional pericyte The fractional distribution of RBCs was obtained by line-scan are correlated. Pairwise linear correlation coefficients between analysis, in which lines were positioned so as to span the cross- regions of interest (ROIs) were averaged over each projection section of both daughter branches (Fig. 6C) and scanned at a (Fig. 5C, diagonal elements) or pair of projections (Fig. 5C, frequency of 5 kHz. For these experiments, we used acta2-G- nondiagonal elements) to construct a correlation coefficient ma- CaMP-GR mice i.v. injected with tetramethylrhodamine (TRITC)- trix. Data from a single junctional pericyte revealed that ROIs dextran (red) or NG2-dsRed mice injected with FITC-dextran within the same projection were highly correlated, whereas cor- (green) to allow visualization of pericytes and capillaries. To de- relation coefficients between some projection pairs were low a focal stimulus, we inserted a micropipette containing (Fig. 5C). We next examined whether the location of projections fluorescently labeled dextran (to visualize the administration zone) influences the correlation of Ca2+ events by assessing correlations and U46619 (100 nM), or saline (control), through the cranial between projections enwrapping the same capillary branch and windowandmaneuveredittowithin5μm of the targeted junc- those enwrapping different branches. This analysis revealed that tional pericyte. After picospritzing U46619 (or saline) onto a Ca2+ events in ROIs within a projection or between projections junctional pericyte within the postarteriole transitional region, we enwrapping the same capillary branch were more highly correlated tracked the flux of RBCs in both downstream branches by 2PLSM. than were ROIs in projections enwrapping different capillary Local delivery of U46619 onto such pericytes decreased the

Fig. 5. Junctional pericytes are capable of inde- pendently controlling Ca2+ and contraction of capil- lary branches. (A) Representative images of a single proximal junctional pericyte from a retina isolated from an acta2-GCaMP-GR transgenic mouse showing projections wrapping around all capillary branches. (Scale bars, 5 μm.) (i) Schematic showing placement of ROIs for recording Ca2+ events (boxes) and changes in luminal diameter (dashed lines) for eight cellular projections from a single pericyte. (ii–iv) Representative image of fluorescence intensities 2+ relative to baseline (F/Fo) for Ca events restricted to one side (ii) or occurring on both sides (iii and iv)ofa pericyte projection wrapping around capillary branches. (B) Representative traces of Ca2+ fluores- cence (upward deflection) and contraction events (shaded downward deflection) from the eight ROIs depicted in i.(C) Pearson’s correlation coefficient (Corr.) matrix of the average correlation coefficients for all possible combinations of ROIs within (diago- nal elements) and between projections 1 and 8. Heat map depicts the degree of positive (red) and nega- tive (blue) correlation. (D) Whisker-box plot from junctional pericytes depicting average correlation coefficients for Ca2+ events within a projection or between projections located on the same or on dif- ferent capillary branches. Data from n = 5 pericytes (n = 36 projections; 114 ROIs) revealed a higher correlation for Ca2+ events in ROIs within a projection (137 ROI pairs in 36 projections; Corr. = 0.46 ± 0.04) or across projections within the same capillary branch (323 ROI pairs between 36 projection pairs; Corr. = 0.26 ± 0.03) compared to ROIs from different capillary branches (861 ROI pairs between 84 projection pairs; Corr. = 0.07 ± 0.02).

6of12 | www.pnas.org/cgi/doi/10.1073/pnas.1922755117 Gonzales et al. Downloaded by guest on October 1, 2021 PHYSIOLOGY

Fig. 6. Pericyte Ca2+ events determine capillary branch diameter in vivo. (A) Cerebral circulation in an anesthetized acta2-GCaMP-GR transgenic mouse injected with TRITC (red)-dextran to illuminate the vasculature, visualized through a cranial window using 2PLSM. (Scale bars, 50 μm.) (B, Left) Representative in vivo images showing the fluorescence intensity of GCaMP (Ca2+ events) and mCherry, and the geometry of capillary branches. (i) mCherry fluorescence; arrowheads indicate pericytes. (ii) GCaMP fluorescence intensity relative to baseline (F/Fo). (iii) Branch angle. (iv) Branch diameter. (Middle) Representative GCaMP traces; mCherry trace shown for comparison. (Right) Correlation between Ca2+ event frequency (event/s) and branch angle and diameter in the two

daughter branches (d1 and d2) of individual junctional pericytes. Different daughter branches are denoted by triangles and circles. (C–K) Disruption of blood flow following U46619-induced constriction of junctional pericytes in the postarteriole transitional region in vivo. (C) Representative images showing pipette placement and RBC flux before and after picospritzing U46619 (100 nM) onto the targeted proximal pericyte. (D and E) Representative traces and summary

data at 60 s showing the effects of U46619 on RBC flux (cells/s) down daughter branches d1 (purple) and d2 (orange). Flow of RBCs was completely, but transiently, halted in both branches (30%) (i), differed between the two branches (30%) (ii), or decreased in both branches (40%) (n = 10, n = 7 mice; *P ≤ 0.05 vs. baseline) (iii). (F–H) Blood flow remained symmetrical following application of aCSF (control) onto junctional pericytes in the postarteriole transitional

region in vivo. (F) Representative images showing pipette placement and RBC flux (line scans) through daughter branches d1 and d2 of a proximal pericyte, before and after picospritzing aCSF. (G) Representative traces showing the running average of RBC flux (cells/s) down each daughter branch, d1 (purple) and d2 (orange). RBC flux remained relatively constant (i and ii) throughout imaging; when changes in flux occurred, they were symmetrical between branches (iii). (H) Summary data showing RBC flux during baseline and 30 s after picospritzing aCSF (n = 10 pericyte junctions, n = 5 mice). (I–K) Blood flow is unaffected by stimulation of distal, noncontractile pericytes with U46619. (I) Representative images showing pipette placement and RBC flux (line scans) through the daughter branches of distal pericytes before and after picospritzing U46619. (J) Representative trace and summary data showing the effects of picospritzing aCSF onto distal pericytes (n = 19 pericyte junctions, n = 10 mice). (K) Representative trace and summary data showing the effects of picospritzing U46619 (100 nM) onto distal pericytes (n = 10 pericyte junctions, n = 6 mice). (Scale bars, 10 μm.)

Gonzales et al. PNAS Latest Articles | 7of12 Downloaded by guest on October 1, 2021 combined RBC flux through both daughter branches by an aver- transitional region can be engaged by retrograde hyperpolarizing age of 21 ± 5.3 cells per second (n = 10), but the degree of de- signals. Because individual pericyte projections are capable of acting crease between the two branches (d1 and d2) varied. For example, as separate functional units, the arriving vasodilatory signals can in some cases, RBC flux was completely halted (3 of 10 branches) preferentially dilate the daughter vessel supplying the stimulated or decreased (4 of 10 branches) to the same degree in each branch region. (symmetric). Notably, however, in a subset of cases (3 of 10), the Finally, to test the hypothesis that different projections of a decrease in RBC flux was greater in one branch than the other contractile pericyte are capable of receiving and isolating branch- (asymmetric) (Fig. 6 D and E), consistent with differential con- specific stimulation (hyperpolarization) in an in vivo setting, we traction of individual branches. By comparison, any changes in flux locally delivered 15 mM K+ onto capillary segments downstream following delivery of saline (control) were generally symmetrical (∼40 μm) of one branch of a junctional pericyte in the post- (Fig. 6 F–H). Transient activation of pericytes by focally picospritzing arteriole transitional region of NG2-DsRed-BAC transgenic U46619 had no effect on RBC flux at more distal capillary junctions mice and measured RBC flux in each daughter branch (Fig. 8A). (Fig. 6 I–K), consistent with the absence of the dynamic smooth As shown in Fig. 8 B and C, this led to an asymmetric shift in muscle α-actin in this pericyte subpopulation or slower dynamics that RBC flux, such that flux was robustly increased in the stimulated might not be captured with the transient U46619 delivery. branch (75% ± 21%) in association with a decrease (8 of 12 junctions) or no change (4 of 12 junctions) in flux in the other Junctional Pericytes Serve as Control Elements in K+-Mediated branch. By contrast, picospritzing 15 mM K+ onto a capillary Functional Hyperemia. We recently reported that retrograde hyper- segment downstream of one branch of a distal junctional pericyte polarizing signals mediated by K+-dependent activation of capil- caused a symmetrical increase in blood flow (30% ± 5%) in both lary endothelial Kir channels cause upstream arteriolar dilation and daughter branches (10 of 11 branches) (Fig. 8 D and E). This thereby increase blood flow into the capillary network (5). Peri- latter observation indicates that the increase in blood flow in- cytes and capillary endothelial cells are electrically coupled via gap duced upstream (i.e., at the arteriole level) by our previously junctions (36), implying that pericytes are capable of receiving reported K+-induced NVC mechanism is not subject to direc- electrical signals from the underlying . Accordingly, tional regulation by these noncontractile pericytes, which are we hypothesized that the retrograde hyperpolarizing signal induced incapable of actively facilitating blood flow in the direction of the by a K+ stimulus applied downstream of the transitional region stimulus. Collectively, these findings suggest that contractile per- would be conferred via gap junctions to overlying pericytes along icytes in the postarteriole transitional region are capable of re- the track of the propagating hyperpolarization, leading to branch- ceiving propagating K+-induced hyperpolarizing signals through specific changes in capillary diameter in the direction of the stim- the capillary network and dynamically modulating blood supply. ulus through hyperpolarization-dependent relaxation of pericyte projections (Fig. 7A). Discussion To test this hypothesis ex vivo, we used our recently developed Considerable research effort has been devoted to understanding pressurized retina preparation to examine changes in daughter how active neurons communicate their energy needs to the brain branch diameters at pericyte-occupied junctions of the post- microvasculature to increase the local delivery of blood-borne + arteriole transitional region following K stimulation of a down- nutrients and O2. The focus of much of this research has been stream capillary branch. In this retinal preparation, the distal on the arteriolar level of the vascular tree, especially the role of portion of the ophthalmic that feeds the retina is isolated in translating neuronal activity into parenchymal ar- and cannulated, and the retina tissue is pinned down en face, teriolar dilation. Recent work from our laboratory (5, 37, 38) allowing visualization of the entire retinal vasculature (Fig. 7B). and those of others (39–41) has shifted the focus downstream, After pressurizing the preparation, which caused the vascular tree showing that the expansive network of capillaries serves as a fed by the ophthalmic artery to develop myogenic tone (SI Ap- “sensory web” that detects neuronal activity and converts it to an pendix,Fig.S8), we locally delivered 15 mM K+ onto a capillary electrical (hyperpolarizing) signal that propagates upstream to segment downstream of a junctional pericyte located in the tran- cause arteriolar dilation. This process, initiated by neuronal sitional region and measured changes in the diameter of the activity-derived K+ and mediated by Kir2.1 channels in capillary stimulated branch and unstimulated branch. The retrograde ECs, provides an efficient mechanism for driving an increase in hyperpolarizing signal induced by K+ led to a change in pericyte blood flow toward active regions. But, like proposed / contractility, producing asymmetric responses in branch diameters arteriole-level signaling processes, this mechanism leaves open such that diameter increased preferentially in the branch imme- the question of how the distribution of blood flow and RBC flux diately upstream of the stimulus (Fig. 7 C–E). To further inves- within the capillary network is regulated. In the absence of such a tigate this, we created mice expressing a modified Ca2+ sensor regulatory process, blood distribution within the capillary bed with lower maximum brightness and lower Ca2+ affinity fused to would be governed solely by the static architecture of the mi- mVermilion (a modified mCherry protein with twofold higher crovascular network; this would decrease the efficiency of tar- fluorescence intensity) driven by an Acta2 promoter (Methods). geted blood delivery and invariably result in squandering of were prepared from acta2-GCaMP8.1-mVermilion trans- resources on relatively quiescent areas. Here, we demonstrate genic mice, enabling the simultaneous ratiometric measurement of that pericytes structurally alter the static symmetry of capillary Ca2+-dependent (green) and Ca2+-independent (red) fluorescence junctions within the microcirculation. We further show that signals in individual projections. Notably, application of K+ onto a contractile pericytes at capillary junctions in the transitional re- distal capillary segment asymmetrically impacted Ca2+ signaling in gion dynamically and differentially regulate daughter branch pericyte projections at upstream daughter branches, specifically diameters to control the distribution and directed perfusion of decreasing Ca2+ signals in pericyte projections surrounding the RBCs within brain tissue. Importantly, we provide strong experi- dilated branch immediately upstream of the stimulus (Fig. 7F). To mental evidence that pericytes play an integral role in our previ- further demonstrate branch-specific dilation, we subsequently moved ously established electrical-based NVC mechanism, showing that the picospritzing pipette to a capillary segment downstream of the contractile pericytes are capable of modulating junctional blood previously unstimulated branch, and applied a second 15-mM K+ flow by dilating in response to K+-dependent hyperpolarizing stimulus. Similar to our observations in the first experiment, this signals initiated at the site of neuronal activity, thereby channeling second experiment yielded a larger upstream dilation in the stimu- RBC flux in the direction of the stimulus. Because, in theory latedbranch(SI Appendix,Fig.S9). Collectively, these results are (supported by modeling; see SI Appendix,Fig.S10), the fraction of consistent with the idea that junctional pericytes in the postarteriole total RBCs decreases nonlinearly at each successive junction in

8of12 | www.pnas.org/cgi/doi/10.1073/pnas.1922755117 Gonzales et al. Downloaded by guest on October 1, 2021 limited criteria are not definitive and ignore a large body of evidence to the contrary. First, smooth muscle α-actin expression and con- tractile behavior are observed in cells that are agreed to be distinct from smooth muscle cells, including (granulation tissue ) (42), fibroblasts (43), myoepithelial cells (44), and developing embryonic cardiac myocytes (45, 46). The appear- ance of a smooth muscle-like banding pattern is an even less com- pelling criterion, given that this overtly similar morphology likely reflects a convergent phenotype driven by biophysical factors—an enwrapping morphology simply more efficiently imparts a contractile force on a tubular structure than any other morphology, regardless of the origin of the enwrapping cell(s). This banding pattern aside, the general morphology of pericytes is clearly distinct from that of smooth muscle cells, as described above. Moreover, there is a general consensus that perivascular cells deeper in the capillary bed are pericytes; yet these cells, like perivascular cells in the post- arteriole transitional region, express the “smooth muscle-specific” protein, Myh11. Perhaps most important, none of these peri- vascular cells express functional RyRs, a defining feature of all known smooth muscle cells; thus, the absence of RyR activity may in fact represent a definitive, pericyte-specific marker. Much re- mains to be learned about the ontogenetic relationship between pericytes in the CNS microvasculature and other cell types, as well as how pericytes come to populate the CNS capillary network. However, regardless of the nomenclature ultimately adopted, the unique features and functional significance of the critical cell net- work in the postcapillary transitional zone are clear.

Control of RBC Flux through Effects of Junctional Pericytes on the PHYSIOLOGY Static Symmetry of Capillary Bifurcations. Vascular network he- modynamics are governed by the applied pressure drop (ΔP; difference between inlet and outlet pressure) and the vascular geometry, which can dynamically change. Blood flow (Q) depends Fig. 7. Branch-specific dilation in response to retrograde hyperpolarization. on the radius (r) and length (l) of each vessel and on the effective + (A) Proposed mechanism by which K -induced retrograde hyperpolarization blood viscosity (neff), and can be approximated from Poiseuille’s + 4 increases cerebral blood flow (5). Increases in local K around capillaries law (Q = π ΔPr /8ηeffl). In the microcirculation, the presence of the activates KIR channels, generating local hyperpolarization that propagates endothelial surface layer (47) and the particulate nature of blood upstream to the feeding arteriole. Membrane hyperpolarization causes ar- introduce nonlinearities and a strong dependence of neff on he- terial and pericyte relaxation, promoting an increase in blood flow into the matocrit and vessel diameter (Fåhræus-Lindqvist effect) (48, 49), capillaries. (B) Schematic of a pressurized retina preparation. The ophthalmic ar- as well as uneven partitioning of RBCs at diverging bifurcations tery of an isolated mouse retina is cannulated and the retina is pinned down en face. (C, Left) Representative image of the feeding arteriole and capillary branch (phase separation) (50). Thus, RBCs entering a capillary bed will most proximate to the arteriole. Picospritzing pipette is targeted downstream of be nonuniformly distributed in a manner that reflects the arbori- one branch (Stim. branch). (C, Right)Ca2+ and branch diameters recorded over zation of the network and complex non-Newtonian fluid dynamics. time. (D) Representative traces showing average pericyte Ca2+ and diameter in The observed flow responses (RBC fluxes) in capillary networks stimulated and unstimulated branches following picospritzing 15 mM K+ and cannot be explained by these physical constraints, suggesting active TRITC-dextran tracer downstream of one branch of the monitored postarteriole regulation of blood perfusion by ensheathing pericytes. transitional segment. (E and F) Summary data showing the increase in branch In this context, we identified a role for junctional pericytes in 2+ + diameter and decrease in Ca in pericyte projections after picospritzing 15 mM K structurally modifying the resting geometry of capillary bifurca- = = ≤ downstream of one branch (n 18, n 5mice;*P 0.05 vs. Stim. branch). tions. Specifically, we showed that the symmetry of capillary junctions in wild-type mice, both in an ex vivo retinal preparation and in the mouse brain in vivo, is inversely related to pericyte the capillary network, the first few junctions make the bulk of the coverage such that junctions in the postarteriole transitional “ ” decisions impacting the distribution of blood flow throughout region, where pericytes are more prevalent, are less symmetrical the network. Thus, contractile junctional pericytes in this post- than more distal junctions. This asymmetry of pericyte-associated arteriole transitional region are well situated to exert an outsized junctions manifests most clearly as differences in diameters of influence on the distribution of RBCs in a local capillary network. daughter branches. Thus, by virtue of their potential influence on the structural symmetry of bifurcations, junctional pericytes are Role of Pericytes in Regulating Blood Flow in the Brain: The Definition uniquely positioned to exert effects on capillary diameter that Problem. Much of the controversy surrounding the functional role compensate for the default geometry of the system and the re- of capillary pericytes in NVC and regulation of blood flow is sistance imposed by physical factors. definitional, reflecting the fact that some consider the micro- vascular region containing α-actin-expressing, contractile peri- Dynamic Control of Daughter Branch Diameters and RBC Flux by vascular cells to be a precapillary arteriole and the cells themselves Junctional Pericytes in the Postarteriole Transitional Region. An to be atypical smooth muscle cells. This formulation simply de- analysis of spontaneous Ca2+ signaling in junctional pericytes in fines away any functional role for capillary pericytes. The persis- the transitional region of ex vivo retinal preparations from acta2- tent view that these contractile cells are smooth muscle cells also GCaMP-GR transgenic mice revealed different Ca2+-signaling finds support in the fact that some of these cells are capable of profiles in individual pericyte projections enwrapping branches of a exhibiting a smooth muscle cell-like banding pattern. However, these junction. Moreover, there were clear cases in which contractile

Gonzales et al. PNAS Latest Articles | 9of12 Downloaded by guest on October 1, 2021 Fig. 8. Branch-specific increases in capillary blood flow in response to retrograde hyperpolarization. (A) Representative images showing pipette place- ment and RBC flux (line scans) through daughter

branches d1 and d2 enwrapped by a junctional per- icyte in the postarteriole transitional segment, be- fore and after picospritzing 15 mM K+ downstream of one branch of the monitored junction. (B) Rep- resentative traces showing the running average of

RBC flux (cells/s) down each daughter branch, d1 (purple) and d2 (orange), following administration of 15 mM K+ downstream of one branch. (i–iii)K+- dependent increases in RBC flux through the stimu- lated branch reduced flux (67%; i and ii) or had no effect (33%; iii) on flow in the unstimulated branch. (C) Summary data showing RBC flux at baseline and 30 s after stimulation by picospritzing 15 mM K+ downstream of one branch of the monitored junc- tional pericyte in the postarteriole transitional seg- ment (n = 11 pericyte junctions, n = 6 mice; *P ≤ 0.05 vs. baseline). (D and E) Representative images, trace, and summary data showing the effects of picos- pritzing 15 mM KCl downstream of a distal pericyte (n = 12 pericyte junctions, n = 6 mice; *P ≤ 0.05 vs. baseline). (Scale bars, 10 μm.)

responses were restricted to individual projections in which Ca2+- which showed that stimulation with U46619 was capable of signaling events occurred. Thus, unlike all muscle cells, in which the causing asymmetric constriction of daughter branches. Importantly, functional working unit is the entire myocyte (51), an individual this differential control of branch diameters observed ex vivo pericyte is made up of multiple functional units corresponding to translated into differences in RBC flux in vivo. In these latter ex- separate capillary-enwrapping projections, each of which constitutes periments employing the mouse cranial window model and 2PLSM, a restricted Ca2+-signaling domain that is capable of contracting focal stimulation of a junctional pericyte in the postarteriole tran- independently of other projections. sitional region dynamically regulated RBC flux in downstream 2+ Because pericytes are able to compartmentalize Ca and branches, producing both symmetric and asymmetric changes in flux. contract individual projections, the prediction is that junctional pericytes should be able to independently regulate the diameter Contributions to NVC. Neurons are metabolically demanding cells of the daughter branches they enwrap. This prediction was borne and rely heavily on an on-demand process, termed neurovascular out by experiments performed on retinal capillary preparations, coupling, that translates neural activity into local increases in

10 of 12 | www.pnas.org/cgi/doi/10.1073/pnas.1922755117 Gonzales et al. Downloaded by guest on October 1, 2021 blood flow and thus delivery of needed nutrients. Recent work (rhodamine) band-pass filter. For contractility experiments, time-lapse im- from our laboratory has demonstrated the operation of a NVC ages were acquired at 30 frames/min and z-stack images were obtained at “ ” 0.2-μm increments with a 1.2-μm z-resolution. For Ca2+-imaging experi- mechanism, identifying capillaries as a wiring system that elec- × trically communicates neuronal demand upstream to promote ments, images were acquired at a resolution of 512 512 pixels at 30 frames/ s at 36 °C. For ex vivo pressurized-retina studies, the entire orbit containing arteriole dilation (5). Pericytes, which are electrically coupled to the eye, optic nerve, ophthalmic artery, and surrounding musculature was the underlying capillary endothelial cells (36), are in a position to 2+ isolated and placed in ice-cold, 95% O2/5% CO2-bubbled Ca -free and receive and potentially manipulate these electrical signals, either Mg2+-supplemented rPSS (SI Appendix, Methods). Outer muscles, nerves, as electrical sinks or boosters. Our data suggest that projections of and surrounding support tissues were removed using fine scissors, leaving contractile pericytes enwrapping one branch of a capillary junction the optic nerve, fine musculature, and ophthalmic artery intact. Branches in the postarteriole transitional region are able to dilate in re- from the ophthalmic artery that did not feed the central retinal artery sponse to K+-induced hyperpolarizing signals propagating along were ligated using a fine suture. The ophthalmic artery was cannulated that branch, dynamically manipulating junctional blood flow to using a fine glass cannula attached to a micromanipulator and filled with favor the flux of RBCs in the direction of the stimulus. This sur- filtered rPSS. The retina was slit at strategic positions along the edge to allow the spheroidal structure to lay flat on a custom silicon platform, with prising finding implies that membrane potential is not necessarily the ophthalmic artery and surrounding structures positioned under the uniform across all processes of an individual pericyte. This sug- imaging plane. The entire retina was continuously perfused with 37 °C gests that pericyte processes may be electrically compartmental- rPSS buffer at 5 mL/min. The cannula was pressurized via a gravity-fed line ized through an as-yet-unidentified mechanism, a behavior attached to a perfusion pressure monitor (PM-4; Living Systems) by moving reminiscent of neuronal processes and completely unlike that the pressure column to ∼60 mmHg (corresponding to pressure at the found in smooth muscle cells. With their ability to independently ophthalmic artery) and opening a two-way valve. Rapid removal of blood control the diameters of different branches, contractile pericytes cells from all retinal vessels confirmed successful cannulation and pres- are able to receive hyperpolarizing signals from downstream surization. Glass microinjection pipettes filled with 15 mM KCl were neurons, causing a branch-specific dilation that provides a path- guided to targeted pericytes and endothelial cells using a micromanipu- lator (Sutter Instruments), and a small volume was pressure injected using way that guides increased blood flow in the direction of active a Picospritzer III (Parker). neurons. By controlling the distribution of RBCs, contractile pericytes located at capillary bifurcations in the postarteriole In Vivo Cerebral Imaging. The cranial window model and in vivo imaging were transitional region provide a mechanism that ensures the prefer- performed as previously described (5), and further details can be found in ential delivery of RBCs to cells in need, thereby increasing the SI Appendix, Methods. FITC and TRITC were excited at 820 nm (∼2to189 efficiency of the system. mW of output power at the objective), GCaMP5 was excited at 940 nm

(∼5 to 55 mW of output power at the objective), and emitted fluores- PHYSIOLOGY Methods cence was separated through 500- to 550- and 570- to 610-nm bandpass Animals. All animals were used in accordance with protocols approved by the filters. Vessel diameters were determined using reconstructions of 2+ Institutional Animal Care and Use Committee of the University of Vermont. z-stack images (SI Appendix, Methods). Ca images were acquired at a × ∼ Acta2-GCaMP-GR mice were obtained from the CHROMus resource (33) and resolution of 512 100 pixels at 40 frames/s. RBC velocity and flux data Acta2-GCaMP8.1-mVermilion BAC transgenic mice were generated as part were collected by line scanning at 5 kHz. Glass microinjection pipettes of this study. Detailed descriptions of transgenic mice have been included in filled with U46619 or 15 mM KCl were guided to targeted pericytes and SI Appendix, Methods. Animals were housed on a 12-h light:dark cycle with endothelial cells, respectively, using a micromanipulator (Sutter Instru- free access to food, , and environmental enrichments. ments), and a small volume was pressure ejected using a Picospritzer III (Parker), as previously described (5). For in vivo tests of branch-specific responses, downstream capillary segments were chosen at random. Care Ex Vivo Imaging of Pericyte Contraction and Ca2+ in the Retina. Animals were was taken to ensure that ejection duration and pressure were calibrated deeply anesthetized with pentobarbital sodium (50 mg, i.p.) and killed by to obtain a small solution plume, and that placement of the pipette re- exsanguination and decapitation. Retinas were isolated, pinned down in an stricted agent delivery to the capillary under study and caused minimal en face (vitreal side up) orientation, and stored in ice-cold retinal physio- displacement of the surrounding tissue. Spatial coverage of the ejected logical saline solution (SI Appendix, Methods). For immunocytochemistry solution was monitored by including TRITC- or FITC-labeled dextran so- studies, en face retina preparations were mounted on a silicone block and lution (150 kDa; 0.2 mg/mL). washed in a magnesium-based physiological saline solution (Mg-PSS) (SI Appendix, Methods). Retinas were fixed by incubating with 4% parafor- ± maldehyde (Mg-PSS) for 15 min, permeabilized with 0.1% Triton X, and Modeling, Calculations, and Statistics. All data are presented as means SE. “ ” blocked with a blocking solution (Mg-PSS) containing 2% bovine serum al- Values of n refer to the number of projections, junctions, or junctional “ ” bumin (BSA) and 2% normal goat serum. Retinas were incubated overnight pericytes, and N refers to the number of animals. Relationships between 2+ at 4 °C in blocking solution containing cyanin (Cy)-conjugated mouse local Ca activity recorded in a single projection or in different projec- ’ monoclonal anti-α-actin (Sigma-Aldrich) or anti-calponin (Abcam, ab46794) tions of an individual pericyte were evaluated by performing a Pearson s antibodies. Filamentous actin and microtubules were stained ex vivo by in- correlation analysis using MATLAB R2020a (The MathWorks, Inc.). Detailed cubating isolated en face retinal preparations with Alexa Flour 488 Phal- description of modeling and statistics are provided in SI Appendix, Methods. loidin (1:200; Thermo Fisher Scientific) and Tubulin Tracker Green (1:50; Thermo Fisher Scientific), respectively. Blood vessels were counterstained Data Availability. All study data are included in the article and supporting with rhodamine (1:50; Vector Laboratories) or FITC-conjugated isolectin B4 information. (VWR) in Mg-PSS for 20 min at 37 °C. The junctional fluorescence minus background fluorescence was normalized to the smooth muscle fluorescence ACKNOWLEDGMENTS. We thank T. Keith and M. Gubrud for research ’ minus background at each junction and then multiplied by 100 to obtain the assistance; S. O Dwyer, M. Ross, G. Kopec-Belliveau, T. Wellman, Shaun Rein- ing, and D. Enders for technical assistance; Mark Rizzo and W. Gil Wier (De- percentage. A junctional pericyte was defined as a pericyte whose nucleus partment of Physiology, University of Maryland) for providing the mVermilion was within 8 μm (the size of ∼1 pericyte nucleus) of the center of the 2+ cDNA; and the University of California, Irvine (UCI) Transgenic Mouse Facility junction. For ex vivo pericyte Ca and contractility studies, en face retinal for embryo injections (P30-NCI-CA062203). Generation of the mouse strains, preparations were mounted on a silicone block, washed, and stained with acta2-GcaMP-GR and acta2-GCaMP8.1-mVermilion, was supported by rhodamine- or FITC-conjugated isolectin B4 (diluted 1:25 in rPSS bubbled CHROMus (Cornell Lung Blood Resource for Optogenetic Mouse Sig-

with 95% O2/5% CO2) by incubating for 20 min at 37 °C. Care was taken to naling). This study was supported by grants from the Totman Medical Research Trust, Fondation Leducq (Transatlantic Network of Excellence on the Patho- ensure that proper pH (bicarbonate based) and oxygenation (95% O2) were maintained throughout isolation, pinning, staining, and experimental pro- genesis of Small Vessel Disease of the Brain), the European Union (Horizon 2020 Research and Innovation Programme SVDs@target under the grant cedures. Pericytes were imaged using a Revolution confocal system (Andor agreement 666881), and the Henry M. Jackson Foundation for the Advance- × Technology) mounted on an upright Nikon microscope equipped with a 60 ment of Military Medicine (HU0001-18-2-0016) to M.T.N., and by grants from water-dipping objective (numerical aperture [NA] 1.0). Fluorescence was the NIH: T32-HL-007594 and K01-HL-138215 to A.L.G.; F32-HL152576 to excited with a 488-nm (FITC) or 560-nm (rhodamine) solid-state laser, and N.R.K.; R24-HL-120847 to M.I.K.; and R01-NS110656 and R35-HL140027 emitted fluorescence collected through a 527.5/49 nm (FITC) or 641.5/117 nm to M.T.N.

Gonzales et al. PNAS Latest Articles | 11 of 12 Downloaded by guest on October 1, 2021 1. O. Hassler, The windows of the internal elastic lamella of the cerebral arteries. 28. A. A. Berthiaume et al., Dynamic remodeling of pericytes In Vivo maintains capillary Virchows Arch. Pathol. Anat. Physiol. Klin. Med. 335, 127–132 (1962). coverage in the adult mouse brain. Cell Rep. 22,8–16 (2018). 2. R. G. Tilton, E. J. Miller, C. Kilo, J. R. Williamson, Pericyte form and distribution in rat 29. J. P. Kerr et al., Detyrosinated microtubules modulate mechanotransduction in heart retinal and uveal capillaries. Invest. Ophthalmol. Vis. Sci. 26,68–73 (1985). and . Nat. Commun. 6, 8526 (2015). 3. D. E. Sims, The pericyte–a review. Tissue Cell 18, 153–174 (1986). 30. H. A. T. Pritchard et al., Microtubule structures underlying the sarcoplasmic reticulum 4. R. Mazanet, C. Franzini-Armstrong, Scanning electron microscopy of pericytes in rat support peripheral coupling sites to regulate smooth muscle contractility. Sci. Signal. red muscle. Microvasc. Res. 23, 361–369 (1982). 10 , eaan2694 (2017). 5. T. A. Longden et al., Capillary K+-sensing initiates retrograde hyperpolarization to 31. R. Dabrowska, D. Aromatorio, J. M. Sherry, D. J. Hartshorne, Composition of the increase local cerebral blood flow. Nat. Neurosci. 20, 717–726 (2017). myosin light chain kinase from chicken gizzard. Biochem. Biophys. Res. Commun. 78, 6. P. Blinder et al., The cortical angiome: An interconnected vascular network with 1263–1272 (1977). noncolumnar patterns of blood flow. Nat. Neurosci. 16, 889–897 (2013). 32. R. Murray, G. A. FitzGerald, Regulation of thromboxane receptor activation in human 7. F. Fernández-Klett, N. Offenhauser, U. Dirnagl, J. Priller, U. Lindauer, Pericytes in platelets. Proc. Natl. Acad. Sci. U.S.A. 86, 124–128 (1989). capillaries are contractile in vivo, but arterioles mediate functional hyperemia in the 33. B. Shui, J. C. Lee, S. Reining, F. K. Lee, M. I. Kotlikoff, Optogenetic sensors and ef- mouse brain. Proc. Natl. Acad. Sci. U.S.A. 107, 22290–22295 (2010). fectors: CHROMus-the cornell heart lung blood institute resource for optogenetic 8. D. A. Hartmann et al., Pericyte structure and distribution in the cerebral cortex re- mouse signaling. Front. Physiol. 5, 428 (2014). vealed by high-resolution imaging of transgenic mice. Neurophotonics 2, 041402 34. L. Borysova, S. Wray, D. A. Eisner, T. Burdyga, How calcium signals in myocytes and (2015). pericytes are integrated across in situ microvascular networks and control microvas- 9. A. London, I. Benhar, M. Schwartz, The retina as a window to the brain-from eye cular tone. Cell Calcium 54, 163–174 (2013). research to CNS disorders. Nat. Rev. Neurol. 9,44–53 (2013). 35. B. Himpens, T. Kitazawa, A. P. Somlyo, Agonist-dependent modulation of Ca2+ sen- 10. C. Rouget, Memoire sur le development, la structure et les proprietes physiologiques sitivity in rabbit smooth muscle. Pflugers Arch. 417,21–28 (1990). des capillaires sanguins et lymphatiques. Arch. Physiol. Norm. Pathol. 5, 603–663 36. D. M. Wu, M. Minami, H. Kawamura, D. G. Puro, Electrotonic transmission within (1873). pericyte-containing retinal microvessels. Microcirculation 13, 353–363 (2006). 11. K.W. Z, Der Feinere Bau Der Blutcapillares (Anat. Entwicklungsgesch, 1923). 37. A. Moshkforoush et al., The capillary Kir channel as sensor and amplifier of neuronal 12. A. Krogh, The Anatomy and Physiology of Capillaries (Yale University Press, New signals: Modeling insights on K+-mediated neurovascular communication. Proc. Natl. Haven, 1924). Acad. Sci. U.S.A. 117, 16626–16637 (2020). 13. L. Díaz-Flores, R. Gutiérrez, H. Varela, N. Rancel, F. Valladares, Microvascular pericytes: 38. O. F. Harraz, T. A. Longden, F. Dabertrand, D. Hill-Eubanks, M. T. Nelson, Endothelial A review of their morphological and functional characteristics. Histol. Histopathol. 6, GqPCR activity controls capillary electrical signaling and brain blood flow through – 269 286 (1991). PIP2 depletion. Proc. Natl. Acad. Sci. U.S.A. 115, E3569–E3577 (2018). 14. R. A. Hill et al., Regional blood flow in the normal and ischemic brain is controlled by 39. C. Cai et al., Stimulation-induced increases in cerebral blood flow and local capillary arteriolar smooth muscle cell contractility and not by capillary pericytes. Neuron 87, depend on conducted vascular responses. Proc. Natl. Acad. Sci. U.S.A. 95–110 (2015). 115, E5796–E5804 (2018). 15. C. N. Hall et al., Capillary pericytes regulate cerebral blood flow in health and disease. 40. L. Khennouf et al., Active role of capillary pericytes during stimulation-induced ac- Nature 508,55–60 (2014). tivity and spreading depolarization. Brain 141, 2032–2046 (2018). 16. Y. Li et al., Pericytes impair capillary blood flow and motor function after chronic 41. R. L. Rungta, E. Chaigneau, B. F. Osmanski, S. Charpak, Vascular compartmentaliza- spinal cord . Nat. Med. 23, 733–741 (2017). tion of functional hyperemia from the synapse to the pia. Neuron 99, 362–375.e34 17. D. Kleinfeld, P. P. Mitra, F. Helmchen, W. Denk, Fluctuations and stimulus-induced (2018). changes in blood flow observed in individual capillaries in layers 2 through 4 of rat 42. J. J. Tomasek, G. Gabbiani, B. Hinz, C. Chaponnier, R. A. Brown, Myofibroblasts and neocortex. Proc. Natl. Acad. Sci. U.S.A. 95, 15741–15746 (1998). mechano-regulation of remodelling. Nat. Rev. Mol. Cell Biol. 3, 18. Z. Shen, Z. Lu, P. Y. Chhatbar, P. O’Herron, P. Kara, An artery-specific fluorescent dye 349–363 (2002). for studying neurovascular coupling. Nat. Methods 9 , 273–276 (2012). 43. H. Y. Zhang, M. Gharaee-Kermani, K. Zhang, S. Karmiol, S. H. Phan, Lung 19. R. I. Grant et al., Organizational hierarchy and structural diversity of microvascular alpha-smooth muscle actin expression and contractile phenotype in bleomycin- pericytes in adult mouse cortex. J. Cereb. Blood Flow Metab. 39, 411–425 (2019). induced pulmonary fibrosis. Am. J. Pathol. 148, 527–537 (1996). 20. X. Zhu, R. A. Hill, A. Nishiyama, NG2 cells generate oligodendrocytes and gray matter 44. P. D. Arora, C. A. McCulloch, Dependence of remodelling on alpha-smooth astrocytes in the spinal cord. Neuron Glia Biol. 4,19–26 (2008). muscle actin expression by fibroblasts. J. Cell. Physiol. 159, 161–175 (1994). 21. A. Mishra et al., Imaging pericytes and capillary diameter in brain slices and isolated 45. M. Kedinger et al., Smooth muscle actin expression during rat gut development and retinae. Nat. Protoc. 9, 323–336 (2014). induction in fetal fibroblastic cells associated with intestinal embryonic epithelium. 22. V. Nehls, D. Drenckhahn, Heterogeneity of microvascular pericytes for smooth muscle Differentiation 43,87–97 (1990). type alpha-actin. J. Cell Biol. 113, 147–154 (1991). 46. J. Woodcock-Mitchell et al., Alpha-smooth muscle actin is transiently expressed in 23. H. S. Wei et al., Erythrocytes are oxygen-sensing regulators of the cerebral microcir- embryonic rat cardiac and skeletal muscles. Differentiation 39, 161–166 (1988). culation. Neuron 91, 851–862 (2016). 47. A. R. Pries, T. W. Secomb, Microvascular blood viscosity in vivo and the endothelial 24. L. Alarcon-Martinez et al., Capillary pericytes express α-smooth muscle actin, which surface layer. Am. J. Physiol. Heart Circ. Physiol. 289, H2657–H2664 (2005). requires prevention of filamentous-actin depolymerization for detection. eLife 7, 48. R. Fåhræus, T. Lindqvist, The viscosity of the blood in narrow capillary tubes. Am. e34861 (2018). J. Physiol., 562–568 (1931). 25. S. Hughes, T. Chan-Ling, Characterization of smooth muscle cell and pericyte differ- 49. A. R. Pries, D. Neuhaus, P. Gaehtgens, Blood viscosity in tube flow: Dependence on entiation in the rat retina in vivo. Invest. Ophthalmol. Vis. Sci. 45, 2795–2806 (2004). diameter and hematocrit. Am. J. Physiol. 263, H1770–H1778 (1992). 26. M. Vanlandewijck et al., A molecular atlas of cell types and zonation in the brain 50. A. R. Pries, K. Ley, M. Claassen, P. Gaehtgens, Red cell distribution at microvascular vasculature. Nature 554, 475–480 (2018). bifurcations. Microvasc. Res. 38,81–101 (1989). 27. P. C. Stapor, W. L. Murfee, Identification of class III β-tubulin as a marker of angiogenic 51. F. S. Fay, C. M. Delise, Contraction of isolated smooth-muscle cells–structural changes. perivascular cells. Microvasc. Res. 83, 257–262 (2012). Proc. Natl. Acad. Sci. U.S.A. 70, 641–645 (1973).

12 of 12 | www.pnas.org/cgi/doi/10.1073/pnas.1922755117 Gonzales et al. Downloaded by guest on October 1, 2021