The “Neuro-Glial-Vascular” Unit: The Role of Glia in Neurovascular Unit Formation and Dysfunction

1The brain and retina, which constitute the central nervous system (CNS), are highly complex tissues, requiring high levels of energy for function and tight control for health. To achieve this, they contain a specialized vasculature that controls parenchymal homeostasis, transport of metabolites, and confers, in part, so-called immune privilege. Most importantly, the bi-directional movement of molecules across the blood-tissue barriers is strictly controlled to maintain CNS health and brain function. For many decades the focus of this regulatory capacity lay at the endothelial cell (EC), the predominant cell of the blood-brain barrier (BBB), but more recently the wider neurovascular unit (NVU) has been recognized as providing functionality. The NVU is a complex multi-hetero-cellular structure of EC, neurons, glia, smooth muscle cells (SMCs), pericytes, and extracellular matrix (ECM). Together, these components regulate blood flow and metabolism, thus allowing the controlled exchange of nutrients and metabolic waste products. To meet the high metabolic demand of the CNS, particularly in response to an intensification of physical or mental activity, increased neuronal activity leads to subsequent changes in cerebral blood flow (functional hyperemia), in a process called neurovascular coupling. While neurons can directly regulate this system, glial cells are often in direct physical contact with the vasculature and neurons, thus are critically positioned to interface between these cellular components where they may contribute to the relay of information and act as a modulator of such crosstalk. Additionally, NVU components are crucial for brain protection and homeostasis as ECs form the BBB, and glia physically ensheathing the blood vessels are seen as a secondary barrier. With this close physical interaction between ECs and glia, nutrients required for CNS function are delivered from the blood vessels to neurons mainly via glia cells, while waste compounds are passed via glial cell to microglia or back into the bloodstream. Dysfunction of the NVU is characterized by dysregulation of neurovascular coupling, neuron death, gliosis, microglia activation, mural cell transmigration, and BBB breakdown. Dysfunction of the BBB is associated with increased vascular leakage, transcellular transport, immune cell infiltration, and reduction of intercellular junctions (Figure 1). Accordingly, glial cells and other NVU cells work closely together to maintain CNS function and maintenance.

2In this review, we discuss glia cell types and their role in the NVU, by examining glia specializations to support neurons, the vasculature, and neuro-vascular interactions in the NVU. Lastly, we will highlight how rapidly improving techniques and tools will help us answer pressing outstanding questions in the field. While this review is not meant to be an exhaustive list of the many types of glia within the CNS, we aim to highlight their importance for the development, function, and dysfunction of the NVU in the brain and retina.

3The complex interaction between NVU cells requires each cellular component to operate in a complex and coordinated manner to ensure homeostatic control of the BBB and blood retina barrier (BRB). Each component exhibits specialized features that are critical to the overall maintenance of NVU function (Figure 1). Briefly, ECs form a single-layer lining of tubular blood vessels, which are specialized depending on the vascular bed in which they are situated. At the BBB/BRB, ECs exhibit reduced pinocytosis/transcytosis, increased expression of tight junction molecules, such as claudins, occludin, or zonula occludens 1 (ZO-1; Figure 2A;; ), and exclude free transport of substances over 400kDa. Mural cells, which constitute pericytes and vascular smooth muscle cells (vSMCs), are positioned in the basement membrane shared with ECs, maintain vascular stability, provide structural support for blood vessels, govern vasodilation/constriction, as well as contribute to NVU function by BBB maintenance. Neurons in the NVU transduce signals and control local cerebral blood flow directly, such as via nitric oxide (NO), as well as indirectly via glia cells, such as via arachnoid acid or potassium (K+). Additionally, neuronal activity itself shapes vascular and BBB formation via levels of neurotransmitter release. Glial cells physically ensheath blood vessels with their endfeet, creating the semi-permeable glia limitans. Importantly, glia physically connect vessels to neurons, modulate neurotransmission, and impact neurogenesis. Microglia, macrophages, and perivascular macrophages (PVM) play roles in phagocytosis and the CNS inflammatory response, ensuring CNS maintenance and health. The vascular basement membrane, which encompasses blood vessels, acts as a passage for fluid transport, while the perivascular basal laminae and ECM molecules support the glio-vascular interface. Together, these components form a spatially and functionally integrated NVU with bidirectional communication, namely neuro-vascular-coupling and vascular-neuro-coupling. However, the precise mechanisms of the diverse roles of glial cells in NVU form, maintenance and function remain unclear. Answering these fundamental questions will be of particular importance as the NVUs’ main role is considered neurovascular coupling, but indeed various other aspects such as integration of signaling, metabolism, and homeostasis occur across NVU components and thus must also be considered.

4Glia were originally described as scaffolds providing structural support and maintaining biophysical integrity, making their role in supporting the structure and biophysical integrity of the CNS their most widely described function. However, glia are now being increasingly appreciated for their many other functional and regulatory roles, such as neurotransmission, BBB function, and controlling immunity (Table 1). To fulfill the variety of specialized functions required, glia cells are highly specialized according to each CNS region with respect to proteomic signatures, electrophysiology, Ca2+ signaling, morphology, and proximities to synapses. Moreover, even within morphological groups, such as astrocytes, there is heterogeneity of cell types and their pathological responses that may also reflect regional differences in the structure and function of the NVU.

5The most abundant and widespread glial cell type in the brain are astrocytes. Astrocytes are fivefold more numerous than neurons, with individual astrocytes contacting up to two million neuron synapses with elaborate morphologies. This high spatial correspondence between astrocytes and neurons is accompanied by astrocytes regulating neuron health by controlling neurotransmitters, such as glutamate or adenosine, as well as maintaining hydromineral brain homeostasis, such as Ca2+, Cl–, or water. In addition to contacting neurons, astrocytes also contact blood vessels, affecting local blood flow by regulating blood vessel diameters by vasoconstriction (e.g., by arachnoid acid) and vasodilation (e.g., by prostaglandins). Astrocytes also physically and functionally contribute to the BBB and its permeability for factors such as molecular traffic of glucose or proteins. Thus, astrocytes are central to the function of a healthy functioning NVU, and overall CNS function.

6In the retina, the principal glial cell type are Müller glia (MG), which contact blood vessels and neurons, fulfilling similar functions as astrocytes in the brain. The structure, morphology, and species-specific differences of retinal MGs are well described and MG exhibit at least five apico-basal domains that stretch from the apical stem around photoreceptors to the basal endfoot in the inner limiting membrane. However, MG are not just highly organized apico-basally, but also laterally to interact with other cells; this intercalation between cells is in a so-called tiled fashion, thereby contacting almost all cells within the retina. The retina is protected by two separate components of the BRB. The outer (oBRB) consists of retinal pigment epithelium and the inner (iBRB) is located at the level of the retinal capillaries. The latter is established by MGs and pericytes, with dysfunction being implicated in several retinal diseases such as diabetic retinopathy (DR;;;; ). In the mammalian retina, there are also astrocytes that contact blood vessels and are pivotal as a structural growth template during angiogenesis, mainly via vascular endothelial growth factor (VEGF) and Hypoxia-inducible factor (HIF) pathways, following ganglion cell templates. This suggests that MG and astrocytes may work together in the mammalian retina to influence the development of the NVU as well as to regulate its function. In humans, the retinal astrocytes are limited to the inner vascular plexus, while MG contact both plexi and are likely to induce the BRB functionality/maturity in the deep plexus, raising the likelihood for differential coupling between astrocytes and MG in the healthy and potentially diseased retina. However, MG are the only glia found in the fovea, which is free of astrocytes, microglia, and vascular EC, suggesting that in some regions of the retina MG are sufficient to solely meet the functional needs required for high acuity vision.

7Besides the principal glial cells in the brain and retina (i.e., astrocytes and MG), other glial cells exist to fulfill crucial functions in the CNS. During development, progenitors, so-called radial glia, will divide to generate neurons and glia, while also making contact with the vasculature where they contribute to the regulation of CNS angiogenesis. In certain CNS regions, radial glia persist into adulthood as stem cells, contributing to BBB maintenance via retinoic acid signaling. Microglia are the primary immune cells in the CNS, surveying their environment and responding to insult, fulfilling roles in phagocytosis and inflammation where they express both pro-inflammatory (e.g., IL-1β) and anti-inflammatory (e.g., IL-10) molecules, with subsequent upregulation of factors such as Glial Fibrillary Acidic Protein (GFAP). Additionally, microglia-to-astrocyte crosstalk in response to glutamate plays a role in neuro-immune-interactions, and microglia are required for normal neurogenesis, mostly by nerve growth factor (NGF) and tumor necrosis factor (TNF; ). Another type of glia, called ependymal cells, line the brain ventricles, producing cerebrospinal fluid, and subpopulations acting as progenitors for astrocytes and oligodendrocytes. Thus, there are several types of glia found within each CNS tissue, yet several pressing questions remain to understand the importance of glia in the NVU. These include what role do glial cells play in driving NVU formation during development, how are glial cells directed to contact and support multi-cellular units within the NVU, and what are the consequences of dysfunction of these glial components on NVU function.

8The fact that glia are a structurally integral part of the NVU physically linking the vasculature and neurons, emphasizes their functionality in BBB/BRB formation and CNS development.

9Glia cells provide structural support to neuronal tissues for anisotropic mechanical tension, and loss of MG in the retina results in tearing of the tissue due to the loss of their biophysical support. Another biophysical role of glia is that they can swell, which subsequently affects the NVU by spatial changes. The impact of glia volume changes on neurons is facilitated and relayed by the fact that glia ensheath pre- and post-synaptic terminals of neurons to form the “tripartite synapse”. Due to this physical proximity, changes in glia cell size can modulate the extracellular space and subsequently neuron excitability. This is achieved by the synergistic activity of Aquaporin 4 (Aqp4), a channel protein, which is needed for water transport and is enriched in astrocytic endfeet, as well as the transient receptor potential cation channel TRPV4 for Ca2+ influx. However, even though Aqp4 and TRPV4 are considered main factors, glia swelling is a complex process and, depending on the context, other factors were shown to play a role in glia volume changes. These include K+ ion transport via connexin 43, Kir 4.1, or Na+/K+-ATPase, and ion flux via Na+-K+-Cl– co-transporter (NKCC1), or glutamate movement via specialized transporters. These factors can also change in disease or upon injury, as exemplified by sulfonylurea receptor 1 – transient receptor potential melastatin 4 (SUR1-TRPM4) which is upregulated in CNS injury, but the impact of such changes on NVU function remain poorly defined.

10Besides this physical link, glia also functionally link NVU components, exemplified by their impact on EC junctions, transporters, and pathways (Figure 1;;; ). This functional link is in part achieved by factors such as glial cell line-derived neurotrophic factor (GDNF), VEGF, fibroblast growth factor 2 (FGF2), or angiopoietin-1 (Figures 2B,C;;;;; ). But glia cells are also key to supporting neurotransmission by removal of neurotransmitters to terminate synaptic transmission and reestablish neuronal excitability, thus avoiding toxic overstimulation, called excitotoxicity. Glia also directly regulate neuronal activity within the synapse and synchronize/modulate synaptic inputs on the level of signaling via gliotransmitters, such as gamma-aminobutyric acid (GABA), glutamate, or cytokines. The major neurotransmitters are the excitatory glutamate and the inhibitory GABA, working together to regulate CNS function. Following the removal of neurotransmitters from the synaptic cleft, glia cells transfer these neurotransmitters back to neurons in a process called the glutamate-glutamine cycle that requires ammonia (NH3) and ammonium ion (NH4+) derived from NVU blood vessels, thus reestablishing functional neuron neurotransmitter pools. This maintenance of synaptic potentials comes at a very high metabolic cost with the energy for this provided by astrocytes and blood vessels. While it was previously assumed that both glutamate and GABAergic neurons are under astrocytic control, a recent in vitro study challenges this, suggesting that GABAergic neurons establish functional synaptic transmission without glia. Another neuromodulator released from neurons or glia, impacting neuronal function, is adenosine derived from adenosine triphosphate (ATP) breakdown. Adenosine stimulates receptors that regulate the release of GABA, glutamate, acetylcholine, noradrenaline, 5-HT, and dopamine.

11Thus, glia are specialized morphologically, biophysically, and molecularly to support and regulate the NVU. However, the multitude of glial functions within the NVU makes it challenging to discern which mechanisms are necessary and sufficient for NVU form and function. To understand the role of glia cells, further studies are needed where glia cells are disrupted (i.e., lacking, inhibited, or overactive). Such studies will allow us to disentangle NVU interactions, establish the exact role of glia cells within it and the pathophysiological consequences.

12As glia directly contact and ensheath blood vessels, they directly influence EC structure and function rather than passively co-exist. Indeed, in the last decade, it has become clear that glial cells play an active role in facilitating vascular angiogenesis via expression of factors, such as VEGF or transforming growth factor 1β (TGF-1β) in radial glia cells. Moreover, radial glia were shown to stabilize murine nascent cortex vessels via inhibition of Wnt signaling and proliferation in EC in a contact- and age-dependent manner, a process which is potentially mediated by MMP-2. Glia cells, specifically CNS-specific macrophages (microglia), were also shown to play pivotal roles in the fusion of blood vessels, called anastomosis. This is achieved by Notch1-expressing macrophages that link path-seeking dll4-expressing vascular tip cells as well as by acting as physical chaperones which express TIE2 and NRP1, regulating anastomosis downstream of VEGF-mediated endothelial tip cell induction. After vessel- and NVU-formation, glia are also pivotal for BBB maturity, via factors such as retinoic acid supplied by radial glial cells, which increases BBB stability and the expression of BBB-specific genes such as p-glycoprotein (P-gp), occludin, and Glut-1 [see details for BBB transport systems and junctions]. Similarly, astrocytic Src suppressed C kinase substrate (SSeCKS) reduces VEGF and increases EC tight junctions, or astrocytic angiotensin-converting enzyme-1 (ACE-1), which produces angiotensin II that facilitates BBB maturation and junction protein stabilization. ECs and glia interact bidirectionally in NVU development. Firstly, radial glia support EC maturation toward decreased proliferation, reduced tip cell marker DLL4, and reduced vascular permeability, thus supporting BBB maturation. Subsequently, ECs increase GFAP-positivity in radial glia in a VEGF-A dependent manner, leading to astrocyte differentiation and NVU formation. In addition to these molecular impacts, the migration patterns of blood vessels, astrocytes, and neurons are closely associated with each other like scaffolds, with astrocytes providing VEGF for EC migration and, vice versa, ECs in turn provide oxygen for astrocyte differentiation. Critically, in pathology, lactate-stimulated MG express G-protein–coupled receptor 81 (GPR81), which triggers neovascularization via pathways such as Wnt or Norrin.

13Besides these roles in angiogenesis, glia are also pivotal in regulating blood flow via regulating NVU synaptic activity as well as by releasing factors such as calcium, NO, arachidonic acid, and prostaglandins. It was also shown that angiotensinogen-to-angiotensin II cleavage occurs in glia, with angiotensin I receptor (AT1-R) causing vasoconstriction, while AT2-R causing vasodilation. Moreover, glia contribute to vasodilation indirectly by interaction with other NVU components such as pericytes, which then, in turn, impacts the vasodilatory state. Here, one important molecule is calcium, with calcium signaling not only being coordinated between glia cells, but also ECs. It remains to be understood to which extent glia-EC signaling is indirectly (via neurons or pericytes) or directly coupled. Lastly, neuron-derived NO regulates glia-mediated vasodilation via prostanoids and epoxyeicosatrienoic acids. Together, glial cells therefore not only play a role in angiogenesis, anastomosis, EC maturation, and blood flow regulation, with glia dysfunction potentially leading to BBB breakdown, pathological vascularization, dysregulation of vasoregulation and failure to deliver sufficient oxygenation.

14As the CNS has high metabolic demands and lacks a carbohydrate storage system, the NVU is critical to serving the retina and brain metabolic needs as glucose has to be continuously supplied via the blood to meet the constant CNS energy demand. This is particularly crucial as neurons rely on oxidative metabolism, making them sensitive to changes in levels of oxygen and potentially ischemia; on the other hand, astrocytes rely on glycolytic metabolism, and glucose can be stored in them in the form of glycogen, rendering them central players in NVU and CNS metabolism. Crucially, pyruvate carboxylase, an enzyme that is key to synthesizing the neurotransmitter glutamate from glucose via the anaplerotic pathway, is almost exclusively expressed in astrocytes, positioning them as essential producers of the neurotransmitters GABA and glutamate. Recently it was shown that retinal MG also conduct anaplerosis, suggesting that MG support retinal NVU metabolism similar to astrocytes in the brain.

15While the majority of vascular-to-neuron glucose metabolism occurs directly via glia in the NVU, a minor proportion of the glucose flux happens directly between blood vessels and neurons. At the end of glycolysis, lactate and pyruvate are produced, but instead of being merely “end-products” they are utilized further to generate energy. Indeed, lactate is taken up by neurons and is metabolized in preference to glucose when both are available. Once transported into the neuron, lactate is converted to pyruvate and used for ATP production. Thus, glia not only provide compounds for neurons but also complement them in their metabolic requirements. Besides glucose metabolism, astrocytes are also crucial for fatty acid oxidation (FAO) to generate ATP, and catabolic ketogenesis to generate ketone bodies for neuronal metabolism; with ketone bodies or lactate being used by neurons as an energy substrate to produce ATP. Importantly, neurons produce toxic fatty acids that are endocytosed by NVU astrocytes for cytoprotection and CNS health. Metabolically, astrocytes also play a critical role in L-serine de novo synthesis, which is converted to D-serine in neurons, acting as a co-agonist of N-methyl-D-aspartate (NMDA) receptors. Together, these data demonstrate that glial cells play a key role in maintaining the homeostatic status of the metabolism of neurotransmitters, glucose, FAO, L-serine, as well as NO, essential for maintaining normal function of the NVU. Any disturbance of this fine-tuned balance, therefore, such as would occur in diseases as diverse as stroke and diabetes, will influence NVU function resulting in further failure to supply adequate essential substrates for normal neuronal function.

16In addition to the above metabolic coupling, neuron-to-vascular coupling is achieved by NO, a gaseous neurotransmitter acting as a vasodilator that is needed for neurovascular coupling and regulating the vasodilatory vascular response, called functional hyperemia. In hyperemia, glia cells are essential in relaying either vasodilation or vasoconstriction depending on the available NO concentration. Further to physiological NO, glia-mediated NO and gliosis-related production of reactive oxygen (ROS) or nitrogen species (RNS), play a role in nitro-oxidative stress such as in neuroinflammation and disease such as epileptogenesis. Indeed, Sharma et al. suggest that epilepsy is preceded by a cascade of reactive gliosis, ROS/RNS, inflammatory cytokines, and neurodegeneration. This neuro-vascular metabolic coupling of glucose, fatty acid, L-serine, and NO via glia is dependent on glia directly contacting ECs within the NVU. The required spatial connections between endfeet and ECs are partially achieved via connexins, which form intercellular gap junction channels and hemichannels that are expressed in ECs and astrocyte endfeet, as well as are associated with BBB maturation. Also, Pannexins (Panx) that resemble connexin-based hemichannels, are thought to play a role in vasculo-neuro coupling although there is still a debate on their function in CNS development and health.

17Beyond their roles in physiological conditions, glia and microglia contribute to neuroinflammation in response to injury, stroke, or other neurological diseases. They act as primary initiators of the inflammation cascade by increased reactivity and the secretion of factors such as chemokines. One ubiquitous biomarker of so-called glial activation is increased expression of GFAP. Despite the extensive list of definitions for different glia cell states, the field agrees that inactive and reactive glia cells display changes in molecular profiles, including alterations in the cytoskeleton, metabolism, chaperones, secreted proteins, signaling proteins, and transporters. These molecular changes are accompanied by morphological transformation in cellular phenotype, such as hypertrophy, which may impact the ability of glial cells to provide multiple support functions for each NVU component(s) and hence its uncoupling.

18Besides the physiological functions of glial cells in inflammation, their roles in pathological settings are of wider interest. For example, increasing evidence also shows glia as a link between vascular and neurological contributions in cognitive impairment, Alzheimer’s Disease (AD), and seizures. Neuropathologically, AD is characterized by intracellular neurofibrillary tangles and brain parenchymal amyloid β-peptide (Aβ) deposits. The latter form neuritic plaques and cerebral amyloid angiopathy, leading to angiopathy and NVU dysregulation. Astrocytes have been shown to degrade amyloid-beta in an apolipoprotein E (APOE)-dependent manner, a process that could be impaired in AD, leading to plaque formation. Neurodegenerative disorders are complex conditions with multiple underlying causes and the role of glia has not been fully elucidated. Nevertheless, several astrocyte/glia risk factor genes, such as APOE, particularly the E4 isoform or Clusterin (CLU) and FERM Domain Containing Kindlin 2 (FERMT2), have been identified for AD. Also, it is generally acknowledged that AD involves inflammatory responses, initiated, or mediated via microglia and astrocytes that lead to BBB breakdown. Throughout AD disease progression, different aspects were found to affect the NVU. In pre-senile AD, increased proliferation (Ki-67), gliosis (GFAP), and vascular changes, but not neurogenesis were shown. In late-onset AD, vascular dysregulation and BBB breakdown are considered as the earliest biomarker, with vascular dysregulation preceding changes in amyloid beta deposition, metabolism, function, structure, and memory. However, future work is needed to link genetic and mechanistic causes of AD and the spatiotemporal impacts on individual NVU components and NVU function.

19In the retina, understanding the NVU and glia contribution is of particular interest as NVU dysfunction can precede neural dysfunction in patient retinas, such as in patients with Type 1 Diabetes who exhibit DR. In DR and other retinal diseases, inflammation and accompanying side effects are critical contributors to disease progression and vision loss. Crucially, it has been shown that MG provide VEGF, which contributes to the upregulation of inflammatory markers, such as ICAM1 and TNFα. These pro-inflammatory compounds induce pathological vascular leakage and retinal neovascularization, making anti-VEGF therapies an important therapeutic strategy in DR. In DR, it was shown that the MGs that express VEGF display a distinct morphology, suggesting that they enter a reactive phenotype with altered function before expressing VEGF, which in turn leads to changes in NVU and vascular function. Besides morphological changes and increased VEGF expression, MG also show altered expression of trophic factors in DR, such as NGF, brain-derived neurotrophic factor (BDNF), neurotrophin-3 (NT-3), neurotrophin-4 (NT-4), ciliary neurotrophic factor (CNTF), or glial cell line-derived neurotrophic factor (GDNF), as well as inflammatory factors such as interleukin 1β (IL-1β), IL-6, IL-8, and TNF-α. Together, these factors contribute to disease progression and neuron survival. Indeed, in diabetes, it is likely that such changes result in early NVU dysfunction and subsequent failure of neurovascular coupling and hence an insufficient supply of nutrients, which then cause further damage to the NVU. Another retinal disease with clear implications of MG dysfunction is macular telangiectasia 2 which causes loss of central vision due to vascular defects (MacTel 2). Critically, it was shown that loss of MG matches the area of macular pigment depletion in MacTel 2 patients. Moreover, the loss of these MG impacts on the NVU as a secondary hallmark feature of MacTel 2 is the formation of dysfunctional telangiectatic vessels. One potentially crucial aspect when thinking about the regional restriction of MG loss and vascular defects is that macular MG rely more on serine biosynthesis than peripheral MG, with serine synthesis disruption resulting in mitochondria dysfunction and oxidative stress and ultimately retinal pathology. This retinal pathology is also associated with swelling-induced MG volume changes and penetration of astrocytes into deeper layers. Thus, rather than having a passive role in pathogenesis, the activation state of glia, together with changes in their morphology and molecular expression, is directly associated with disease progression of diseases such as DR and MacTel 2.

20Finally, it is important to note that retinal demand for oxygen is even greater than that of the brain. This is especially so during dark adaptation when metabolic activity and oxygen demand is high. The configuration of the vascular supply to the retina means that during high demand the functional reserve of oxygen is minimal, even when neurovascular coupling maximizes functional hyperemia. This makes the retina extremely vulnerable to hypoxic damage and so even small alterations to the function of any component of the NVU, including MG, is likely to result in compromised retinal function.

21As the NVU is a heterologous structure formed by different cell types, it is crucial to study the NVU, and the role of glial cells in it, in various models to understand how those cells integrate to form a functional NVU. Despite our increased understanding of NVU and BBB formation, our insight into the exact processes that glial cells regulate in NVU formation and function is still far from complete. This is emphasized by the fact that in vivo studies are still limited in experimental scope and flexibility.

22As an in vivo model, rodents are an invaluable pre-clinical asset to study glia and their role in the NVU. However, caution has to be paid when wanting to draw direct conclusions from pre-clinical models. For example rodent glial cells are smaller and less complex than human glial cells, with human astrocytes being 2.6-fold larger, 10-fold more GFAP-positive primary processes, and relaying signals faster, suggesting that NVU metabolism and signaling dynamics are different between rodents and human. Similarly, certain glial cells are primate-specific, such as interlaminar astrocytes, and can therefore not be studied in rodents. To complement NVU and glia studies in rodents, zebrafish have also been invaluable, particularly as NVU function and development can be studied in real-time in vivo and throughout the CNS. Zebrafish characteristics, such as ex utero development, genetic tractability, and embryonic transparency render them a crucial asset. Crucially, the site of the BBB is conserved in zebrafish and humans in capillary ECs. However, even though neurovascular coupling was shown to be conserved, it remains to be answered to which extent the NVU is truly conserved across species. Again, glia in zebrafish and mammals show differences, exemplified by the lack of typical stellate astrocytes in zebrafish as well as the fact that progenitor cells are widely maintained in the zebrafish adult CNS while being mostly transient in mammals. However, zebrafish are being increasingly used as an experimental model to contribute to the understanding of astrocyte development and function. Importantly, zebrafish radial glia are capable of adult neurogenesis and harbor a very high regenerative capacity, making zebrafish a suitable model to study de- and re-generation. In vitro, reductionist models have also been invaluable in providing insights to study glia and their role in the NVU. However, glial cultures, originally established by dissociation and plating of brain homogenates, lose their structural context, rendering the information obtained of limited relevance. Similarly, cultured MGs were previously shown to de-differentiate casting doubt on the translatability of the data to in vivo settings. Advancements in establishing 3D cell cultures, however, have allowed for more physiological insights into glia biology. This is especially the case when considering recent NVU studies in organoids and organ-on-a-chip models. Still, in comparison to in vivo models, in vitro studies are often considered to provide limited insight into their tissue context and developmental processes only recapitulating single-timepoint and -context information.

23Our morphological and functional understanding of the NVU, through examining the molecular composition and regional specifications of the different cell types in the CNS, has also advanced significantly. This has allowed us to begin to understand the molecular profiles of cellular components of the NVU as well as their specialization and conservation across species. One aspect of particular interest is to examine glia-to-EC contacts, whether this is impacted by regional specialization, and whether this and NVU function, is influenced by glia/EC identities. Understanding such specializations will provide novel insights into the integration of CNS barriers provided by the ECs, basal lamina, and glia limitans, as their specialization might cause them to respond differently to stimulation, injury, or disease and how it affects the NVU. Regional specializations and barrier properties might also elucidate new routes for drug delivery. Thus, to understand how NVU components function as a unit, it is also crucial to establish the contribution of the individual components and how different NVU cell types combine spatially and functionally together. While the latter can be achieved by careful observation, the former usually requires system changes, ideally cell-specific, to be introduced (e.g., removal, over-expression, inhibition of proteins or cells, etc.). This is exemplified by a recent study that used tamoxifen-inducible astrocyte ablation in mice, showing that astrocytes are key to maintaining the integrity of the BBB as loss of astrocytes coincided with vascular leakage and decreased EC ZO-1 expression and BRB, thus confirming early work undertaken over three decades ago. This BBB alteration could not be rescued by other cells. Moreover, these alterations were accompanied by non-proliferative astrogliosis which, over time, limited vascular leakage. In addition, functional imaging is allowing the unraveling of NVU functions such as neuronal activity using dynamic calcium imaging and cellular relationships following metabolic compound exchanges. Similarly, optokinetic response measurements are increasingly used to measure visual acuity. Linked to performing functional imaging, understanding metabolic fluxes, storage, and turnover is crucial in beginning to understand how NVU components communicate and support each other’s function and how they may be disturbed in disease. This understanding of metabolic pathways is particularly challenging as metabolites are difficult to visualize, particularly over long periods of time. Multi-modal studies using live dyes, in vitro designs, as well as computational modeling, will help in the understanding metabolic pathways, fluxes via glia, as well as direct neuron-to-vascular transport.

24To understand the role of glia and the NVU, the correct tools are needed to visualize components with sufficient resolution to resolve subcellular structures. For this, zebrafish are particularly well suited as transgenic reporter lines exist to label as well as manipulate (i.e., upregulation, downregulation, and loss of function) each component of the NVU in vivo. As such, multi-transgenic reporter lines can be generated to visualize complex cell structures and interactions in the developing retina. Further, the function of the NVU can be visualized in real-time in vivo using several physiological readouts such as blood flow (e.g., red blood cell movement) or neuronal activity (calcium reporters). Elaborate integrations of NVU studies are needed to answer questions on the spatiotemporal integration of angiogenesis, barrierogenesis, gliogenesis, and NVU function. This is exemplified by recent studies showing that neuronal activity during development impacts glial maturation at the synapse. However, it is unclear how these events may in turn influence glial support for neurons and ultimately NVU formation. Combining state-of-the-art visualization tools, examinations of NVU dynamics and functional studies in vivo will compliment high-resolution structural studies [e.g., electron microscopy] and provide further insights into the roles of glia in the form and function of the NVU.

25As glial morphology is tightly linked to their function, quantitative objective analysis of glia morphology can provide novel insights. Recent advances in NVU in vivo imaging, such as Two-photon microscopy (TPM), Intrinsic optical signal imaging (IOSI), Optical coherence tomography (OCT), or Laser speckle contrast imaging (LSCI) will greatly contribute to our understanding of NVU functionality and dynamics. This will enable imaging of functional readouts, such as blood flow, following neuronal stimuli including exposure to light flicker in the retina, to ascertain the impact of glial dysfunction on neurovascular coupling mediated events. This is complemented by ever more sophisticated tools in which to study the NVU in freely behaving animals as well as the ability to investigate NVU function during development. The advancement of imaging modalities and spatio-temporal resolution offers great benefits for understanding these developmental and functional relationships in the NVU (Figure 3). However, one drawback to having this increase in resolution is the exponential increase in the amount of data that is generated per experiment. With the vast amount of experimental data, data analysis and computational expertise has become a limiting factor for many laboratories. Trained specialists and specialized data analysis training are needed to extract meaningful data in a high-throughput, standardized, and objective manner. Dedicated specialized computational approaches can address this analytical bottleneck to analyze data, and push scientific boundaries by computationally modeling what is experimentally impossible. This could be by in silico multi-transgenics (i.e., artificially overlaying transgene expression on a reference tissue) to establish a virtual retina atlas, similar to the zebrafish brain atlas, allowing for further neuronal activity mapping and co-localization analysis in silico. Similarly, deep learning approaches open new avenues from feature mining, over retinal ganglion counting, to fundus and OCT analysis. Furthermore, this is accompanied by an increased assessment of feature and data connectivity and relationships, such as principal component analysis and Uniform Manifold Approximation and Projection. Additionally, carefully designed data analysis workflows will provide new insights into our data and the NVU, most likely in a way that is beyond current comprehension.

26The coupling of quantitative analysis of glia shape and NVU function will also be essential in understanding the kinetics of NVU degeneration and dysfunction in disease. This is exemplified by studies that link cell feature information to genomic profiles or even link feature analysis to image cytometry, which measures cellular protein and DNA in images. Once robust and repeatable ways to characterize glia morphology are established, it will be possible to directly link cell morphology to function and to their molecular profiles, which could then in turn be linked to NVU functionality studies. One prominent example used transcriptomics to identify and classify retinal bipolar cells by matching their molecular expression with cell morphology. These comprehensive experimental paradigms could be used for each component in the NVU, throughout its development, to identify the molecular mechanisms regulating cell shape, function, and connectivity. These insights into precise cellular mechanisms that control the development and function of the NVU may also inform the pathogenesis of disease in mature tissues. As such, identification of the molecular and morphological changes (e.g., glial hypertrophy), that potentially precede pathology in disease, would facilitate the diagnosis of NVU dysfunction and provide the opportunity for early treatment and better clinical outcomes.

27It is increasingly clear that glial cells are critical for NVU development, function, maintenance, and dysfunction in disease. However, uncovering the precise contribution of glial cells to the NVU has been challenging to date, particularly when trying to understand and integrate their dual contribution to neurons and the vasculature. With the advancement of imaging, computational and genetic tools, it will be possible to use multidimensional approaches (morphology, function, genetics, interactions, and dynamics) to clarify the exact role(s) of glial cells in the NVU. Examining the role of glia in the “neuro-glial-vascular unit” with such a holistic approach will enhance the understanding, diagnosis, and treatment of aging and disease in the nervous system.

28EK, JG, and RM contributed to the conception and design of the study. EK wrote the first draft of the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.

29The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

30All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.