Across the vascular bed, from arteriole to capillary to venule, the diameter of vessels first decreases then increases, and the morphology of their mural cells (smooth muscle cells or pericytes) abluminal to the endothelial tube changes. Penetrating arteries and arterioles are typically between 15 and 40 μm in diameter and are composed of an inner layer of endothelial cells with abundant caveolae. They contain an internal elastic lamina and therefore express elastin in the vessel wall, which can be detected from its binding to the dye Alexa 633 hydrazide (Shen et al., 2012). At the surface, they often have 2-3 outer layers of closely packed ring-shaped smooth muscle cells (SMCs), which are electrically coupled via gap junctions (Chow et al., 2020; Garcia and Longden, 2020) and strongly express smooth muscle alpha actin (αSMA). At the pial surface, the blood vessel is separated from the pial membrane by Virchow-Robin space, which narrows after the arteriole enters the brain through a “pial funnel” (Gao et al., 2015). Downstream from this region, SMCs form only one layer before transitioning to a pericyte morphology. Grant et al. (2019) report the first cells with “bump on a log” morphology as always occurring beyond the first branch from the arteriole, and most reports consider the arterioles to be covered with annular SMCs (e.g., Hamilton et al., 2010; Grant et al., 2019; Kisler et al., 2017). However, even near the pial surface, mural cells’ morphology changes from a simple annulus to gaining a distinct soma and processes. While these cells are often still termed smooth muscle cells, historically this transitional form would be called a pericyte (Zimmermann, 1923; Attwell et al., 2016).

At the first branch point from penetrating arterioles are precapillary sphincters: contractile mural cells encircling a narrow section of vessel between the penetrating arteriole and first order capillary (Chambers and Zweifach, 1946; Grubb et al., 2020). The first branches off the penetrating arteriole are often termed precapillary arterioles or first order capillaries (where penetrating arterioles are branch order 0). These vessels tend to have a diameter of between 5-15 μm, and are enwrapped with mural cells that express αSMA and have a protruding, distinct soma, and processes that cover the vessel to a large degree, recently termed “ensheathing pericytes” (Hartmann et al., 2015; Grant et al., 2019; but see Hill et al., 2015; Attwell et al., 2016). These first branches are often termed the precapillary arteriole or transitional zone, though these terms sometimes are used to mean only the first vessel branching off the penetrating arteriole (e.g., Thakore et al., 2021), and other times include the 2nd and 3rd branches with larger pericyte coverage and αSMA actin expression (Gonzales et al., 2020).

Downstream of these vessels, after 1-3 branches from the diving arteriole (Grant et al., 2019) pericytes become less dense and their processes cover less of the vessel as branch order increases and vessel diameter decreases (range 3-9 μm). These capillary pericytes have been reported to have low (Bandopadhyay et al., 2001) or absent (Nehls and Drenckhahn, 1991; Grant et al., 2019) expression of αSMA and have been termed mesh and thin strand pericytes (Hartmann et al., 2015). Latterly these have been combined into a single category (capillary pericytes), due to the difficulty in distinguishing a transition point between these two morphologies (Grant et al., 2019). These capillaries with low or absent αSMA expression and low pericyte coverage are often termed mid-capillaries.

In post-capillary venules, pericytes have a stellate morphology, becoming less extended with thicker, more radial processes (Joyce et al., 1985; Hartmann et al., 2015). These post-capillary venules branch to form large venules (> 50 μm), which have weak αSMA labeling, indicating limited contractile potential (Hill et al., 2015). On larger venules, SMCs are circumferential but are less compact and show a less complete coverage of the vasculature and a leaf-like instead of banded appearance compared to arteriolar SMCs (Hill et al., 2015).

Therefore, the anatomy of the vessels and mural cells of the cerebral vasculature, as commonly reported, is used to define the following distinct vascular segments: pial arterioles, penetrating or diving arterioles, precapillary sphincters, pre-capillaries, also termed transitional or low branch order capillaries, mid-capillaries, post-capillaries and venules. These have been recently condensed into 4 categories, arterioles, a transitional zone, capillaries and venules, with distinct mural cells on each (Hartmann et al., 2021a). In some ways this categorization is a helpful way to discuss features of different parts of the vascular bed, as the recent Hartmann et al. review does in a clear and informative manner. Our argument, however, is that it is important to recognize that these categories are overly simplistic because they are superimposed on more gradually changing anatomy and function across the vascular bed. For example, existing literature shows that mural cell morphologies transition gradually not abruptly along the network. This concept was first discussed by Zimmermann (1923) and is evident from more recent work which shows that there is a continuum of cell lengths seen from SMCs through ensheathing to mid-capillary pericytes (Grant et al., 2019). Furthermore, the shift in vessel coverage of pericyte processes argued to mark a boundary between ensheathing and mesh pericytes has a different distribution across branch orders than that of αSMA expression (Grant et al., 2019), i.e., αSMA transitions occur at largely different locations than do transitions in pericyte coverage of the vessel.

This heterogeneity of anatomy across the vascular tree presumably reflects the varying functional roles performed by different vessels. Anatomical transitions have been best mapped onto functional changes in terms of dilatory capacity and involvement in neurovascular coupling (i.e., dilation in response to increased neuronal activity). Disease affects different parts of the vasculature in specific ways, making it important to understand how and where such functional transitions occur to be able to intervene successfully to target pathological mechanisms.

Classically, neuronal activity causes local pial and penetrating arterioles to dilate, causing an increase in blood flow to active brain regions (Attwell et al., 2010). However, different components of the vasculature are now known to play distinct roles in generating this response. Mid-capillary regions, being closer on average than any other vessel segment to most neurons, likely first detect neuronal activity, either via the accompanying increase in extracellular potassium concentration (Longden et al., 2017) and/or via production of vasoactive signaling molecules such as prostaglandin E2 (Hall et al., 2014; Mishra et al., 2016) or EETs (Mishra et al., 2016; Zhang et al., 2021). This signal then spreads upstream from the capillary bed, causing dilation of upstream low branching order capillaries and arterioles (Hall et al., 2014; Rungta et al., 2018), via spread of hyperpolarizing currents carried by inward rectifying potassium channels (Zhang et al., 2011; Longden and Nelson, 2015). Whether mid-capillaries constrict and dilate in response to modulation of neuronal activity remains somewhat controversial. Some studies report that they do not (Hill et al., 2015), but a growing body of evidence supports their contractility (Hall et al., 2014; Kisler et al., 2017; Hartmann et al., 2021b; Shaw et al., 2021), though this may be via a different mechanism than in arterioles and lower branching order capillaries (Hartmann et al., 2021b).

The size, frequency and timing of responses to neuronal activity are often reported to vary across vascular compartments though not always in a consistent manner. Arteriole response magnitudes vary along their length, with surface sections of penetrating arterioles dilating more than deeper sections, probably as a result of the mechanical restriction imposed by brain tissue (Gao et al., 2015), though deeper arteriole sections have also been found to dilate more rapidly than shallower sections (Tian et al., 2010). In alpha-chloralose anaesthetized somatosensory cortex, Hall and colleagues found response frequency decreased as branching order increased (i.e., capillaries responded less frequently than arterioles; Hall et al., 2014), though first and second branch order capillaries showed larger and faster dilations than arterioles, a finding subsequently supported by other studies (Zhang et al., 2021). In the olfactory bulb, the region around the branch off the arteriole also dilated first, with slower responses up and downstream (Rungta et al., 2021). However, the same group found much more variable timings in neocortex of ketamine-medetomidine anaesthetized or awake mice, with capillaries or arterioles each being faster on some occasions (Rungta et al., 2021), and in both these reports, mid-capillary dilations were very small (though these data conflate responders and non-responders, unlike some other studies, Hall et al., 2014; Shaw et al., 2021). Precapillary sphincters, where studied, have often shown larger dilations, as a proportion of their diameter, than adjacent arterioles and capillaries (Grubb et al., 2020; Zambach et al., 2021).

In short, there are functional differences between neurovascular coupling responses along the vascular network, with mid-capillaries responding less frequently than upstream vessels to increases in local neuronal activity, and the fastest responses often, but not exclusively, observed in the first and second order capillaries. The physiological mechanisms underlying this heterogeneity in contractile responses are not wholly clear. Pericytes expressing αSMA seem likely to mediate more reliable, often larger and faster changes in vascular tone (Hill et al., 2015; Hartmann et al., 2021b), but as αSMA is commonly expressed in vessels up to and including the 3rd branch from the penetrating arteriole (Grant et al., 2019), variations in αSMA expression alone cannot explain the observed pattern of responses.

Indeed, several vasoactive pathways and mechanisms modulating the diameter of different vascular segments have been found to vary across the vascular network. Vessels beyond the penetrating arteriole were found to dilate via ATP-mediated increases in astrocytic calcium and prostaglandin action on EP4 receptors, while diving arterioles dilated via NMDA receptor mediated NO production (Mishra et al., 2016). Another functional transition point at the same location was also recently observed, as capillary endothelial cells were found to express MFSD2A, inhibiting caveolae formation, while arteriolar endothelial cells showed much lower MFSD2A expression, instead having abundant caveolae that were necessary for intact neurovascular coupling (Andreone et al., 2017). Because MFSD2A-mediated reduction in caveolae formation reduces transcytosis to sustain BBB integrity (Ben-Zvi et al., 2014), this suggests a possible transition in both BBB function and neurovascular coupling mechanisms at the same arteriole-to-capillary point.

It is not just the mechanism of vasodilation that is sensitive to the position in the microvasculature: The integration of vascular signals across the vascular network is an area of increasing focus, with many recent papers shedding new insights on how vasoactive signals are propagated upstream of their detection in the mid-capillary bed. Potassium currents (Longden et al., 2017), nitric oxide (Kovacs-Oller et al., 2020), calcium and ATP signals (Thakore et al., 2021) have all recently been shown to be important in the transmission of vasodilatory signals and, by their nature, these signals span vascular segments and may be influenced by functional variations across these sections. While many studies have not explicitly investigated how different vascular segments affect signal transmission, Thakore et al. (2021) recently showed that ATP application or TRPA1 channel activation in the capillary bed produces slowly-propagating calcium signals through capillary endothelial cells that depend on ATP release from Panx1 channels, but are converted, by IK and SK channels in a transitional segment, into rapidly-propagating electrical signals that dilate upstream arterioles. A similarly propagating signal has been shown to be initiated in the capillary bed by PGE2 (Rosehart et al., 2021; this issue). In these papers, the transitional zone was defined as the first branch off the penetrating arteriole, and did not include lower branching order vessels that would be expected to express αSMA, and it is not clear whether these higher branching order vessels also show rapid IK or SK-mediated propagation of dilation.

Perhaps unsurprisingly given the varying vasoactive properties of different vascular segments, their function can be differentially affected by disease. In Alzheimer’s disease, capillaries may mediate functionally significant decreases in cerebral blood flow, as capillaries but not arterioles were found to be constricted due to endogenous Aβ-mediated endothelin signaling to pericytes (Nortley et al., 2019), and to be plugged by neutrophils (Cruz Hernández et al., 2019). Similar effects may be seen after stroke, with capillaries constricting in brain slices during ischemia (Hall et al., 2014) and showing stalled flow and constrictions after reperfusion in vivo (Yemisci et al., 2009), though the largest constrictions may be seen in αSMA-expressing vessels rather than mid-capillary regions (Hill et al., 2015). Conversely, in mice carrying the main genetic risk factor for Alzheimer’s disease, APOE4, capillary function was unaffected, but the large pial vessels were dysfunctional, with vasomotion and dilation frequencies reduced compared to APOE3-expressing controls (Bonnar et al., 2021). However, none of these studies define the precise location of these transitions in disease susceptibility.

To summarize, the contractile properties of the vascular tree, in terms of both capacity and timing of dilations, mechanisms mediating neurovascular signaling and integration, and the sensitivity of these processes to disease states all vary depending on the position in the vascular network. In some cases (e.g., response frequency) function seems to gradually change along the network, while in others (e.g., mechanism of neurovascular coupling), there are more abrupt transitions of function, and in many cases the location of functional transitions remains unknown.

The major function of the brain’s blood supply is to provide it with nutrients. These include the energy substrates, oxygen and glucose, as well as nucleosides and amino acids needed for mRNA and protein synthesis. Classically, the dense capillary bed has been thought of as the site of nutrient exchange, but actually measuring where nutrient delivery occurs is not straightforward.

Improved two-photon phosphorescence lifetime imaging of oxygen probes has recently provided insights into the concentration gradients of oxygen across the microvasculature in mouse cortex, and therefore where most oxygen delivery likely takes place. In anaesthetized cortex, arterioles were found to be responsible for 50% of the extracted oxygen (Sakadžić et al., 2014), with capillaries (here defined as vessels two branches downstream of the penetrating arteriole) appearing to provide a reserve for oxygen delivery capacity, taking on a larger role during hypercapnia and, as suggested by modeling, when oxygen consumption rates increased. Indeed, in more recent studies in awake mice (therefore with a larger basal oxygen consumption rate), less oxygen was found to be delivered by the arterioles (34%), with the majority of the remainder expected to be delivered by the capillary bed (Li et al., 2019).

The factors driving oxygen delivery by these different vessels include red blood cell linear density, speed, local oxygen concentration gradients, and intravascular resistance to oxygen diffusion. These differ between cortical layers (Lyons et al., 2016; Schmid et al., 2017; Li et al., 2019) and are differently variable within these layers. For example, red blood cell (RBC) flow was lowest and more homogenous between capillaries and, correspondingly, oxygen extraction was largest in deep cortical layers, with faster flow and less oxygen extraction in superficial capillaries (Schmid et al., 2017; Li et al., 2019). This was in part because blood pressure dropped more in the diving arteriole when feeding deeper layers (Schmid et al., 2017). Furthermore, from arterial to venous capillaries, oxygenation becomes increasingly variable as an increasingly clear relationship between RBC flow and pO2 emerges [more O2 having been extracted from capillaries with slowly moving RBCs (Li et al., 2019)]. Thus, position in the vascular network in terms of both tissue depth and branching order affects how blood delivers oxygen to the tissue.

Unlike oxygen, which can diffuse through cell membranes, other nutrients are supplied to the brain via specialist transporters which allow their passage across the BBB. The conventional view is that arterioles and arteries control the blood supply to the tissue but do not participate in the exchange of nutrients in the brain, which is instead mediated by capillaries and post-capillary venules (Yuan and Rigor, 2010). The BBB strictly controls the influx and efflux of molecules across its endothelial layer largely by active transport. In addition to being vital for neuronal function, these proteins can be targeted for drug development to allow drugs to access the brain.

Little work has studied where in the vascular network these transporters are most active, though a recent RNA Seq study showed them to be predominantly expressed in capillary and venule mural cells, in contrast to transcription factors which were overrepresented at the arterial end of the vasculature (Vanlandewijck et al., 2018). A key question to be addressed in future, is whether this myriad of different transporters and transcription factors transition in similar or different locations across the vascular network.

The BBB is also a key site of regulation of immune access to the brain, which again varies across the vascular tree. Unlike arteriolar and capillary endothelial cells, post-capillary venular endothelial cells are connected by adherens rather than tight junctions, meaning they are more leaky than the capillaries which feed them (Rous and Smith, 1931; Pober and Sessa, 2014). The basement membrane that surrounds them also has a different structure, being lamellar instead of homogenous (Braverman, 1989). Leukocyte infiltration into the surrounding tissue primarily happens in this region, leukocytes migrating through gaps between pericytes and thus being regulated by pericyte morphology (Proebstl et al., 2012). Pericyte contractility may feed into this process, as pericyte relaxation (not contraction) widens the gaps between cells, facilitating leukocyte infiltration (Wang et al., 2012). Once in the parenchyma, leukocytes migrate along NG2-expressing pericytes, which release factors to support their migration (Stark et al., 2013). Thus, an important transition in vascular immune function appears to occur between NG2-positive pericytes on capillaries and the NG2-negative pericytes on veins, though whether this happens to the same extent in brain as in peripheral tissues remains unknown. In addition to regulation of immune cell entry, the vasculature can also transport cytokines into the brain, as well as respond to circulating factors with parenchymal production of cytokines and chemokines (Banks et al., 2009; Rustenhoven et al., 2017), but how this varies across the vascular network is also unknown.

More studies into the site of immune regulation in the brain would be valuable, however, as, perhaps unsurprisingly, these pathways of immune regulation by pericytes and endothelial cells are important for understanding and treating disease. Pericyte damage and increased permeability of the brain vasculature to plasma proteins and immune cells are hallmarks of Alzheimer’s disease, multiple sclerosis, and stroke (Daneman, 2012), where infiltrating immune cells have also been shown to induce vascular dysfunction, demyelination, axonal damage, and neurodegeneration (Hochmeister et al., 2006; Ryu et al., 2015). Understanding where in the network structural and functional contributions of the vasculature to the regulation of neural inflammation occur may thus enable us to understand which cells should be best targeted therapeutically.

As well as nutrients being delivered across the BBB, the vasculature also plays an important role in the removal of potentially neurotoxic substances from the surrounding brain tissue. Vasomotion, a phenomenon first reported in bat wing veins (Jones, 1853), is a spontaneous low frequency oscillation in blood vessel tone (typically centered near 0.1 Hz; Mayhew et al., 1996); and independent of heartbeat or respiration), which is present in vessels (particularly arteries) throughout the body (Aalkjær et al., 2011). This rhythmic pulsing of vessels is thought to contribute substantially to paravascular clearance (Aldea et al., 2019; van Veluw et al., 2020) and tissue oxygenation (Tsai and Intaglietta, 1989; Aalkjær et al., 2011; Thorn et al., 2011; Mateo et al., 2017). In arterioles, oscillations are generated within the vascular wall as a result of local SMC dilations and contractions, possibly via phospholipase C and phospholipase A₂ mediated cyclical release of calcium from IP3-sensitive stores (Haddock et al., 2002). Capillaries have not been reported to show these low frequency oscillations in diameter, though RBC flow does show similar fluctuations which may be due to upstream diameter fluctuations (Colantuoni et al., 1994; Biswal and Hudetz, 1996). However, a functional transition in vasomotion of vessel diameters has not yet been pinpointed.

Given the importance of vasomotion as a driving force for paravascular tissue clearance, it is not surprisingly affected by vascular-degenerating diseases. In a mouse model of Alzheimer’s disease, arterioles surrounded with Aβ showed impaired vasomotion when driven by neuronal activity, and showed reduced dextran clearance from the parenchyma (van Veluw et al., 2020). Mice carrying the main genetic risk factor for Alzheimer’s disease, APOE4, also showed a reduction in pial arteriole vasomotion compared to APOE3 controls (Bonnar et al., 2021). Thus, vasomotion is important for regulating supply and clearance of substances to the brain, and is affected by disease. However, we do not yet know where in the vascular tree the transition point falls in which the vessels cease to show vasomotion.

Vascular anatomy, contractility, nutrient delivery, immune regulation and clearance all differ across the vascular bed. To a large extent, functions of the vascular network have been assumed to transition broadly with branching anatomy, from vasomotion pial and diving arterioles that show reliable, large neurovascular coupling responses, through fast responding pre-capillaries/low branching order capillaries, to less reliably dilating mid-capillaries. These mid-capillaries are thought to have a stronger role than upstream vessels in BBB regulation and nutrient transport, while further downstream, post-capillary venules regulate immune cell infiltration. While this framework no doubt captures the broad changes that occur over the vascular network, it is by no means clear that the vasculature can be usefully divided into such discrete divisions. Indeed some of the research discussed above points to the existence of additional functional transitions within vascular segments, including the decreasing size of dilations in superficial and deeper diving arterioles (Gao et al., 2015), and pericyte coverage of vessels continuing to decrease into the capillary bed (Grant et al., 2019). However, most focus on localizing functional changes across the vasculature has fallen on contractile function or mural cell anatomy, so it is unclear whether other functions transition at the same, or different, locations. If different functions show the same transition points as contractile function, we could reasonably consider the vascular segments separated by these transition points as different vessel types. If, however, different vascular functions transition at different locations, it becomes somewhat meaningless to use these shifts in contractile function to define vascular segments.

To test the principle of whether there are clear locations within the vascular network at which transitions occur between multiple functions of the vasculature, we used immunohistochemistry to label different functional markers in brain slices expressing DsRed under the control of the NG2 promoter, as well as in vivo two-photon imaging of vascular responses to visual stimuli in awake mice to study the contractile characteristics of the vascular bed at the different functional transition points. There was no single point in the vasculature where the function transitioned from “arteriole-like” to “capillary-like.” Classic pericyte markers, NG2 and PDGFRβ were in fact expressed throughout the vascular network, while transitions in markers for contractility, arteriolar compliance, neurovascular coupling and angiogenic potential occurred at different points in the vascular bed. We observed no sudden transitions in mural cell morphology across the vascular network, or at termination points of functional markers, though branch points showed distinct changes in mural cell morphology throughout the vasculature studied (arteriole-capillary). Finally, we also observed transitions of neurovascular and vasomotion function down the depth of arterioles, highlighting that arterioles become more capillary-like further from the pial surface.

All experimental procedures were performed in accordance with the 1986 Animal (Scientific Procedures) Act and approved by the United Kingdom Home Office and the University of Sussex or University College London animal welfare ethical review boards. Experiments were performed on mice aged 1-8 months of both sexes (3-8 months for in vivo experiments). Mice were all on a C57BL/6J background which were either wild-type, expressed DsRed under the control of the NG2 promoter (Zhu et al., 2008), or for data collected from the 186 vessels recorded for the in vivo experiments expressed GCaMP6f under the control of the Thy1 promoter (Dana et al., 2014) or were SST-Cre crossed with floxed GCaMP6f. Food and water were available ad libitum, and mice were housed at 22°C in a 12 h light/dark cycle (which was reversed for the mice used in in vivo experiments).

Each immunohistochemistry experiment used tissue from a minimum of 3 animals, on at least 3 different experimental days. Methods for slicing, fixing and labelling were as described previously (Mishra et al., 2014; Boyd et al., 2021), as follows:

NG2, a chondroitin sulfate proteoglycan, is considered a pericyte marker and has been reported at low levels in smooth muscle cells (Kumar et al., 2017). However, NG2-controlled DsRed labeling is found throughout the microvascular tree, including in smooth muscle cells with a clear banded morphology (Figure 1). Immunohistochemical labeling for another pericyte marker, PDGFRβ (Winkler et al., 2010), also reveals protein expression throughout the microvascular tree, including in SMCs on pial vessels as well as penetrating arterioles and capillaries (Figure 1).

As discussed above, a much studied transition across the vascular bed has been that between vascular mural cells expressing high levels of αSMA and those expressing low or no levels of αSMA, which broadly occurs between ensheathing and mesh pericytes, though the distribution of αSMA termination positions is different from that of the change in vessel coverage (Grant et al., 2019). As discussed above, this switch in αSMA expression may coincide with the faster responses observed in low branching order capillaries (Rungta et al., 2018), though contractility of pericytes may extend beyond this transition in αSMA expression, as mid-capillary pericytes also dilate (Hall et al., 2014; Kisler et al., 2017; Hartmann et al., 2021b), presumably using different contractile machinery (Hartmann et al., 2021b). This latter study also found these mid-capillary pericytes to show slower constriction to optogenetic stimulation than did αSMA expressing-pericytes.

Vessels also show a transition in expression of elastin across the vascular bed, representing another flow regulation functional transition. In mammals, the number of layers of elastin expressed in the vessel wall scales with blood pressure and arterial size, being highest in the largest arteries (Cocciolone et al., 2018), but with only one layer, just abluminal to endothelial cells, for most intracerebral arterioles (Shinaoka et al., 2013). Elastin expression terminates before arterioles branch successively to become capillaries, presumably reflecting the lower blood pressures to which these smaller vessels are exposed. Elastin’s main role is to increase vessel distensibility, storing energy in the vessel wall during systole, and releasing it during diastole, thus smoothing blood flow across the heart beat and dampening pressure waves, protecting downstream vascular beds from large fluctuations in pressure (Shinaoka et al., 2013). In addition to this biomechanical role, it may also act as a signaling molecule, regulating and attracting smooth muscle and immune cells (Cocciolone et al., 2018).

Elastin can be readily labeled by i.v. injection or incubation of tissue with AlexaFluor 633 hydrazide (Shen et al., 2012), but its role in microvascular regulation remains understudied. Recently, however, it has been used to identify intracortical arterioles in which endothelial caveolae were found to be critical players in neurovascular coupling (Chow et al., 2020), and was found around precapillary sphincters branching off penetrating arterioles that regulate blood flow and pressure into the cortical capillary network (Grubb et al., 2020; Zambach et al., 2021).

To test whether the transitions between αSMA and elastin expression occur at the same location, thus defining two different vessel “types,” we used immunolabeling for αSMA combined with AlexaFluor 633 hydrazide labeling of elastin. We found that elastin terminated on vessels of large and small diameters (Figures 2B,D) but always upstream from the point of termination of αSMA (Figures 2A-D,F; 4/4 double-labeled branches showed non-overlapping termination points).

Nestin is an intermediate filament protein, commonly thought to be expressed in proliferating cells (Suzuki et al., 2010), and is associated with cardiovascular remodeling (Calderone, 2018). Recently it has been shown to be expressed in quiescent as well as proliferating endothelial cells and, indeed, may exert an inhibitory effect on endothelial cell proliferation (Dusart et al., 2018). Nestin expression by endothelial cells may therefore not necessarily reflect ongoing angiogenesis, but rather indicate the angiogenic potential of vessels. Double labeling of nestin and αSMA or elastin in NG2-DsRed mice revealed the capillary network to be nestin-positive (Figure 3). Nestin labeling extended from the capillary bed into lower branching order capillaries that were also αSMA positive (Figures 3B-D), but never overlapped with elastin-labeled vessels (Figures 3F-H).

To quantify the expression patterns of nestin, αSMA and elastin across the microvascular network, we next categorized the termination points of each label according to vessel lumen diameter, branching order from the penetrating arteriole (where the penetrating arteriole is 0th order; Figure 4A), and the distance between mural cells, or inter-soma distance (ISD). This latter measurement serves as an indicator of the morphology of these cells, as they transition from banded SMCs that are adjacent to each other, to pericytes with distinct soma and processes that become progressively longer along the vascular bed into the capillary network (Grant et al., 2019). Termination points of the different markers occurred at vessels of similar diameters (Figure 4B), but at different branch orders (Figure 4A) or ISD values (Figure 4C), indicative of different positions in the vascular tree. Specifically, elastin termination points were on vessels of significantly lower branch orders than termination points for αSMA or nestin, and mural cells at elastin and nestin termination points were significantly closer together than at termination points for αSMA. Furthermore, consistent with our observations from double labeling of elastin and αSMA (Figure 2), the ISD at which elastin terminated was always smaller than that where αSMA terminated (Figure 4D). Thus, each functional transition occurred at a different point in the vascular tree, as defined by branch order and/or ISD. This means that in addition to the functional transitions between elastin presence and absence, and αSMA presence and absence, there is an additional functional transition point, where nestin terminates, between these two positions.

The vasodilatory signaling molecule nitric oxide (NO) is released by sub-populations of interneurons during synaptic activation, and can mediate or modulate neurovascular coupling (Attwell et al., 2010). In neocortex, if not cerebellum, it seems predominantly involved in regulating arteriole but not capillary diameter, a transition appearing to occur between the diving arteriole and first capillary branch (Hall et al., 2014; Mishra et al., 2016). To investigate whether this reflects neuronal NO sources for vasodilation differing along the vascular tree, we immunohistochemically labeled brain slices for neuronal nitric oxide synthase (nNOS or NOS1) while labeling the vasculature with Alexa647-conjugated isolectin B4, which binds to the basement membrane (Peters and Goldstein, 1979). Consistent with previous studies (Vlasenko et al., 2007), some arterioles showed clear nNOS labeling around (but not within) vessels which extended into the capillary bed, as well as parenchymal signal (Figures 5A-F). To assess whether nNOS is preferentially expressed around particular elements of the vascular network, we measured the intensity of labeling immediately around vessels of different branching orders, and at increasing distances from these vessels (Figure 5G). Linear mixed modeling (with distance from vessel and branch order as fixed factors and vessel as a random factor) showed nNOS labeling to be significantly more intense at the vessel compared to the parenchyma (F = 7.20, d.f. = 3,30, p = 0.0009), but there was no difference in nNOS labeling around different branches (F = 0.87, d.f. = 3, 30, p = 0.47), nor did branch order affect the drop-off in signal away from the vessel (F = 0.25, d.f. = 9, 30, p = 0.98). When just the labeling at the vessel was considered, branch order was borderline-significant (F = 4.1, d.f. = 3,6, p = 0.068). Together these results suggest that nNOS interneurons do target blood vessels, but that the degree of association is similar between penetrating arterioles and the first few capillary branches from the arteriole. The tendency for marginally stronger labeling at arterioles may contribute to the different dependence of neurovascular coupling reported in capillaries and arterioles (Mishra et al., 2016), but is unlikely to explain it entirely. Interestingly, recent work has shown that NO can modulate the propagation of vasodilation through the vascular network (Kovacs-Oller et al., 2020). It is an open, and important, question if nNOS-derived NO could contribute to neurovascular coupling by modulating the integration of vascular signals across the network in addition to its direct effect on vasodilation, and how such NO release would interact with endothelial (eNOS)-derived NO, which is also an important modulator of neurovascular function (e.g., Toth et al., 2015; Chow et al., 2020). Studying the pattern of targets of NO signaling (e.g., soluble guanylyl cyclase, or cytochrome P450 ω hydroxylase (Sun et al., 2000) should be informative in identifying how different vessels can respond to this perivascular NO production.

Previous work suggests that mural cell morphology changes in some respects near to the point at which αSMA expression terminates, with αSMA-expressing ensheathing pericytes covering the underlying vessel to a greater degree than mesh pericytes immediately downstream of the αSMA termination point (though the distribution of changes in coverage is shifted to higher branch order vessels than the termination of αSMA; Grant et al., 2019). However, the degree of vessel coverage by processes is not the only morphological change that pericytes undergo across the vascular network, as they change cell length (indicated by ISD), as well as soma orientation and shape, number and orientation of processes and many other descriptors (Zimmermann, 1923). ISD is different across pericyte categories as defined by vessel coverage, with ensheathing pericytes having a lower ISD than mesh pericytes (Grant et al., 2019; Shaw et al., 2021), so we wondered whether any abrupt shifts in ISD would be observed at functional marker transitions, suggestive of a major change in vessel type at this point. We therefore plotted the ISD immediately before and after the termination points of αSMA and elastin, normalized to the ISD at the termination point (Figures 6A,B). There was no significant change in ISD at the termination point of either functional marker. We had measured more ISD values on the elastin vessels, so also calculated the ISD two further cells away from the termination point, which also showed no differences compared to the cells nearer the transition point, further emphasizing the lack of an abrupt change (Figure 6C).

Pericytes are often found at vessel bifurcations (Hartmann et al., 2015), but we also noticed that they appeared to have a different morphology at branch points, cells being clustered with a shorter ISD and greater vessel coverage (Figure 7A). Indeed, both vessel diameter and ISD were greater at branch points than expected based on the average of the upstream and downstream values of these parameters (Figure 7B). Furthermore, while the increase in diameter was larger at branch points on larger vessels (Figure 7C), the change in morphology, indicated by the relative change in ISD, occurred on all vessels irrespective of the vessel diameter or ISD at the branch point (across a range of vessels from 5-20 μm in diameter; Figures 7D,E). Because ISD generally increases as vessel diameter decreases into the vascular bed (Figure 7F), mural cells at bifurcations therefore, at least in terms of ISD, have a morphology more like larger upstream vessels. This suggests the vascular network shows functional specialization at bifurcations, in addition to classic arteriole-capillary transitions of function. Pericytes at branch points can generate different calcium signals and constriction of different downstream branches, but downstream branch points were less responsive to applied vasoactive agents than upstream vessels (Gonzales et al., 2020). Our data suggest that these downstream bifurcations may still be specialized in some manner, compared to adjacent non-branch capillaries. Pericytes are known to occur frequently at bifurcations (Hartmann et al., 2015). Our analysis reveals that these pericytes at or near branch points are also shorter than expected from the ISD of non-branch pericytes immediately up or downstream.

Functional markers and anatomical changes in mural cell density indicate, therefore, that vascular function changes gradually across the vascular network, with multiple functional transition points corresponding to changes in vascular distensibility, proliferative capacity and contractility. These changes are superimposed on gradually changing mural cell properties that do not show clear alterations (at least in some features) at these transition points, but which do show specializations at branch points that likely reflect functional changes. Furthermore, inspection of the images of DsRed-positive mural cells on arterioles (in Figures 1A, 2A, 3A, 7A and Supplementary Figure 1) show that, unlike commonly assumed, mural cells on arterioles often lose their annular shape as the arteriole dives into the cortex, forming intermediate pericyte morphologies with a distinct soma and processes in deeper regions.

We wanted to test, where possible, how these transitions reflected alterations in the physiological responses of different components of the vascular bed, so studied three properties of visual cortical microvasculature: frequency and size/timing of dilations in response to visual stimulation, and low frequency oscillations in the vasomotion range. We compared responses of the penetrating arteriole before and after smaller branches had come off the main vessel, as well as comparing responses of the arteriole to increasing branch orders of capillaries (see Supplementary Figure 2D for how different vessels’ branch order was defined in order to separate vascular responses along the length of the penetrating arteriole). All data is summarized in Table 1 (and data demographics [cortical depth, diameter] shown in Supplementary Figure 3).

Vessel responsiveness was assessed across the microvasculature by testing if visual stimulation led to an increase in vessel diameter (of > 0.5 standard deviations of the baseline). Sections of arterioles near the cortical surface dilated significantly more frequently than either downstream arteriole sections (≥ BO2) or downstream capillaries (Figure 8 and Supplementary Figures 4–6). Specifically, descending the penetrating arteriole, dilations occurred with a similar frequency, and responses when they occurred were of a similar size in the first two sections of the vessel (before the first capillary branch and between the 1st and 2nd branches off the penetrating vessel), but these superficial responses were more frequent and tended to be larger than responses of the deepest sections of the penetrating vessel (Figure 8).

Across the capillary bed (i.e., any vessel downstream of the penetrating arteriole, populated by high αSMA ensheathing pericytes or low αSMA mid-capillary pericytes), dilation responses were of a similar frequency and size, with response frequencies being similar to the deep sections of the penetrating vessel (i.e., PA2 + ; Figure 8). There were no differences between the speed of dilations across the different vascular segments (Supplementary Figures 4–6).

We next measured vasomotion of the vascular diameter of surface and deep sections of penetrating arterioles and different branches of the capillary bed, finding that vasomotion (power at 0.1Hz) was similar between arteriole sections and between capillaries (Figures 9A-F). When we compared arterioles near the cortical surface (PA0-1), to deeper arteriole sections (PA2 +) and capillaries (C1 +), there was some evidence that penetrating arterioles show more vasomotion near the pial surface, as power at 0.1Hz was highest in superficial penetrating arterioles (Figure 9G), and significantly more upstream arteriole sections (PA0-1: 33%) showed an increase in power in the vasomotion range compared to capillaries (C1 + : 13%) (0.5 SD above baseline; Figure 9G,H).

We found PDGFRβ and NG2 to be expressed throughout the arteriole-capillary vascular network. Both are classically considered to be pericyte markers (Winkler et al., 2010; Armulik et al., 2011), but our results are consistent with the described effects on smooth muscle cells of PDGFRβ gain-of-function mutations, which increase leukocyte accumulation in the aorta (He et al., 2015). Single cell RNA seq analyses also support the more widespread expression of both PDGFRβ and NG2, with mRNA transcripts found in smooth muscle cells as well as pericytes (Vanlandewijck et al., 2018), albeit at moderately lower levels.

The function of these two “marker” proteins may be different depending on their vascular location. NG2 is known to be important for neovascularization and stabilization of newly formed vessels, via the interaction of NG2 with integrins and growth factor receptors on the same and other cells (Stallcup, 2018). In capillary pericytes of the mature vasculature, however, it promotes the formation of new capillaries through angiogenesis, whereas in larger vessels it instead may promote arteriogenetic remodeling of vessel diameter (Rundek and Della-Morte, 2015). Such remodeling can occur after decreased tissue oxygen and, consistent with widespread NG2 expression, cells in surface and penetrating arterioles, and the capillary bed, have been found to proliferate after cerebral ischemia (Wei et al., 2001). PDGFRβ also functions differently in arteries compared to smaller vessels. In culture, pericytes but not smooth muscle cells shed PDGFRβ in response to stress (Sagare et al., 2015) and mutations that block PDGFRβ signaling reduce pericyte number and increase capillary leakiness, without affecting smooth muscle cells (Nikolakopoulou et al., 2017). Conversely, inhibition of PDGFRβ pharmacologically or using siRNA preserves cerebral arterial smooth muscle cells and arterial vascular tone after sub-arachnoid hemorrhage (Shiba et al., 2012; Wan et al., 2019), highlighting a potential pathophysiological contribution of PDGFRβ signaling to arterial smooth muscle cells.

Thus both NG2 and PDGFRβ are expressed throughout the cerebral microvasculature, but differ functionally depending on their location, presumably due to differences in expression levels of other proteins that are localized to different parts of the vascular network.

Given the involvement of nNOS-derived NO in arteriole but not capillary neurovascular coupling (Mishra et al., 2016), we expected to observe differential expression of nNOS along the vascular network. However, while nNOS was expressed at greater levels around vessels than in the parenchyma, this occurred to a similar degree for all vessels studied (up to 3rd branch order) and not just the diving arterioles. NO can control the electrical coupling of pericytes in the retina (Kovacs-Oller et al., 2020) as well as the production of other vasoactive molecules such as 20-HETE (Liu et al., 2008; Hall et al., 2014), so neuronally derived NO may be released onto all vessels but play a different role at arterioles than capillaries, generating a dilation in the former and modulating the response in the latter. A gradient of expression of the other constitutive NOS isoform, endothelial NOS, has not been reported, and RNA Seq data suggests it is expressed at similar levels in arterial, capillary and venous endothelial cells (Vanlandewijck et al., 2018).

As previously described, elastin and αSMA labeling in the vascular wall both label arterioles (Shen et al., 2012; Grant et al., 2019) with αSMA labeling extending into the capillary bed (Grant et al., 2019; Chow et al., 2020; Thakore et al., 2021). Here we show these termination points are distinct and non-overlapping, occurring at significantly different branch orders and with different pericyte morphologies as assessed by ISD. The location of elastin labeling quite well matches locations where we see transitions in physiological responses - the second branch off the diving arteriole - below which responses seem more like those in the capillary bed. Of our 12 elastin-labeled vessels, 8 terminated on the penetrating arteriole (75%), of which 5 (42% of the total) terminated before the first branch from the arteriole, i.e., upstream of where we observe a change in response frequencies. Thus, arteriole elastin labeling broadly, but not tightly, corresponds with the superficial part of the diving arteriole where neurovascular dilations and vasomotion were most frequently observed.

Though associated with contractile ability, the termination point of αSMA does not, however, correlate very well with the size or frequency of neurovascular coupling responses or vasomotion, we observed in vivo. αSMA universally terminated beyond the penetrating arteriole, but we found no differences in neurovascular response frequency, dilation size or vasomotion between different capillary branching orders (1-4 +), though many fewer of these smaller vessels express αSMA. Furthermore, responses were equally frequent in αSMA-expressing deep sections of the penetrating arteriole as in the capillary bed, and capillary dilations were actually larger than these deep arteriole dilations. This is at odds with previous findings, where dilations in ≥ 4th order vessels were substantially smaller than higher order vessels in whisker barrel cortex of awake mice (Rungta et al., 2021). The reasons for this are unclear. Firstly, we compare both response frequency and response sizes, whereas these two measures are conflated in Rungta et al’s paper. However, as we also saw similar response frequencies and sizes across the capillary bed, this cannot explain why we do not see smaller responses in ≥ 4th order vessels. The cortical area is different (visual vs. somatosensory), and the degree of neuronal stimulation might be different (whole field drifting gratings vs. a single whisker deflection), which could perhaps have an impact on neurovascular coupling. Another potential cause of these differences is the method of detecting vascular diameter. We used xy images from which we calculate the diameter perpendicular to every point of a small length of a vessel’s axis, thus averaging across space (12-109 pixels, or 2.4-22 microns), while Rungta et al. used line scans of vessels to measure vascular diameter at a single position. The spatial smoothing we used in this paper is likely to give us a higher sensitivity to small deflections in diameter.

Previous work in anaesthetized animals, has also found different vascular segments to show different time courses of dilation, with deeper sections of arterioles or first order capillaries responding faster than superficial arteriolar segments (Tian et al., 2010; Hall et al., 2014; Rungta et al., 2018). Our data from awake mice did not show this, with similar response kinetics between vascular segments.

The vascular segment between the penetrating arteriole, or end of elastin labeling, and the end of αSMA labeling has often been termed a “precapillary arteriole” or a “transitional segment” (e.g., Rungta et al., 2021), representing a region where vascular function transitions between arteriole and capillary. However, our in vivo data suggests that this segment, corresponding roughly to branch orders 1 to 3, is not (in our hands) where transitions of contractile behavior occur. Other transitions do occur in this zone: We found endothelial expression of the intermediate filament protein nestin extends out of the capillary bed to a position between the termination points of elastin and αSMA. In the retina, calponin, filamentous microtubules, αSMA, filamentous actin and myosin heavy chain were all also found to change expression levels across this section (Gonzales et al., 2020). However, these all transitioned at different positions, calponin terminating on the arteriole, microtubules on branch 1 and αSMA on branch 2, with filamentous actin and myosin heavy chain gradually decreasing in expression levels from branch orders 0 to 4 (beyond which vessels were not studied). These progressive changes in function across a number of markers fit with the gradual changes in mural cell morphology or ISD we observed before and after the termination points of elastin and αSMA, at roughly branch orders 0 and 3, respectively. Thus this “transitional zone” is not uniform, with a single type of vascular cell and, crucially, is not the only region where such transitions of function are occurring, as similar transitions in vasomotion, neurovascular coupling, and mural cell morphology also occur when descending the penetrating arteriole: Neurovascular coupling responses and vasomotion were observed more frequently in superficial segments of diving arterioles than in either downstream capillaries or deep sections of the arterioles, which were similar to each other in response characteristics. Correspondingly, mural cells lost their annular smooth muscle cell morphology to gain a distinct soma and processes at lower reaches of the diving arterioles. This suggests arteriolar mural cells can be pericytes, unlike has been argued (Hartmann et al., 2021a).

This data suggests that the division of the vascular network into four functional segments: arterioles with SMCs, a transitional zone with ensheathing pericytes, capillaries with capillary pericytes and venules with venular SMCs (Hartmann et al., 2021a), while helpful in discussing broad functional changes across the network, is overly simplistic and neglects the gradual transition in functions that occurs. Indeed, multiple other functional transitions occur at other positions in the vascular tree, including between the pial and penetrating arterioles: Penetrating arterioles exhibit higher contractile tone at low intravascular pressures than pial arterioles (Longden et al., 2016). Neurovascular response sizes are also smaller, possibly because of the lower external pressure on the surface vessel compared to penetrating arterioles which are surrounded by brain tissue (Gao et al., 2015).

Our data suggest that pericytes exist at a higher density at branch points than on surrounding vessel lengths, and this clustering occurs to a similar degree on small and large microvessels. This suggests some functional specialization at branch points. 90% of branch points in the transitional segment were previously found to have a pericyte at that location compared to only 45% of more distal branch points (Gonzales et al., 2020). This corresponds with the increase in pericyte ISD we report here, but our data suggest that even the distal branch points are functionally specialized, as they have a shorter ISD than surrounding non-branch point regions. The pericytes at proximal branch points had calcium sparks that corresponded with selective constriction of individual branches, suggesting branch points serve to direct blood flow to active neurons (Gonzales et al., 2020). Distal branch points (> 4th branch order) were not found to be contractile, and calcium changes in these pericytes were not reported. As our data suggest these distal capillaries do dilate to a similar degree as the proximal branches, it would be valuable to study whether these branch points’ pericyte calcium changes also correspond to changes in vascular diameter of the different downstream branches.

Our data and the wider literature currently support multiple transitions of function at different positions of the vascular bed, including down diving arterioles, and along increasingly branching capillaries. Contractile function and neurovascular coupling are the functions most widely studied, but transitions at different locations are also seen in oxygen supply as well as expression of transcription factors, transporters, regulators of angiogenesis and immune regulators. However, these studies are all limited by a low capacity to look at different functions. Rather than focusing on individual vascular functions, single cell RNA Seq has the potential to illuminate the whole range of gene expression differences that exist across the vascular network (He et al., 2018). Clustering of vascular endothelial and mural cells has revealed that endothelial gene expression changes gradually, suggesting that there are likely not simultaneous transitions in expression of many genes at the same point on the vessel. More abrupt transitions between smooth muscle cell and pericytes have been reported (Vanlandewijck et al., 2018), which may fit with the sharp transitions observed in some vascular features (e.g., the shift in vessel coverage from ensheathing to mesh pericytes). However, this study also shows large numbers of genes that are expressed in wider zones of the vasculature, e.g., pericytes and arteriolar but not arterial SMCs, supporting the existence of multiple transition points. Indeed, another RNA Seq study identifies multiple pericyte clusters suggestive of multiple functional transition points (Zeisel et al., 2018). Future studies could link single cell RNA seq data to specific spatial locations within the vascular network, to illuminate this issue further. Such data would allow us to better understand the vascular regions to target with pharmacological interventions, given Alzheimer’s disease and stroke are vasculature-degenerating conditions that differentially affect the various parts of the vascular network.