The Beauty and Complexity of Blood Vessel Patterning

Abstract

This review highlights new concepts in vascular patterning in the last 10 years, with emphasis on its beauty and complexity. Endothelial cell signaling pathways that respond to molecular or mechanical signals are described, and examples of vascular patterning that use these pathways in brain, skin, heart, and kidney are highlighted. The pathological consequences of patterning loss are discussed in the context of arteriovenous malformations (AVMs), and prospects for the next 10 years presented.

Blood vessel formation is a well-orchestrated developmental process that results in patterns at multiple scales. The entire vasculature patterns in the human body to efficiently transport blood to sites of oxygen (placenta and lungs) and nutrient (intestine) uptake, detoxification (liver), and to deliver oxygenated and nutrient-rich blood to organs. At the same time, each developing organ orchestrates patterning of arteries, veins, and capillaries to best serve the unique physiological functions of that organ. The resulting vascular patterns are both aesthetically pleasing and important for organ function.

Emerging blood vessels consist of endothelial cells and accessory pericytes. Simple capillaries maintain this streamlined structure as they mature, while larger vessels also recruit smooth muscle cells that form contractile layers. The largest arteries recruit fibroblasts, and some develop a vasculature to support the support cells (the vasa vasorum in the aorta). However, the lining of all normal vessels contains a layer of endothelial cells, and vessel patterning results from responses of endothelial cells to signals from the environment. These inputs include both signals that stimulate endothelial cell pathways via receptor engagement, and mechanical cues provided by blood flow and blood pressure that also engage endothelial cell signaling. Endothelial cell responses include collective migration, proliferation, anastomosis, and lumenization associated with sprouting angiogenesis, and the migration and cellular rearrangements associated with blood vessel remodeling.

One of us (V.L.B.) wrote a chapter for the first edition of Angiogenesis 10 years ago (Bautch 2012), and at that time VEGF-A was the main endothelial cell signaling pathway associated with blood vessel formation and patterning. In the ensuing years, the importance of other signals in vascular patterning has been revealed, as different pathways have been manipulated genetically and pharmacologically and as the vascular beds of different organs have been examined in more detail. For example, in developing zebrafish, Bmp signaling and not Vegfa is important for development of the caudal vein plexus (CVP) (Wiley et al. 2011), while in many situations other pathways integrate with VEGF-A to pattern blood vessels; VEGF-A also provides signaling important for endothelial cell survival and arterial differentiation.

In this review, we provide a brief overview of the major new advances and concepts in blood vessel patterning over the last 10 years, with a focus on signals that regulate endothelial cell behaviors. Several excellent reviews of signaling in endothelial cells and blood vessels exist for more in-depth analysis (Kiefer and Siekmann 2011; Claesson-Welsh 2016; Hiepen et al. 2020; Desroches-Castan et al. 2021; Fernández-Chacón et al. 2021). We then discuss several examples of vascular patterning that show cross talk with other vessels and tissues, along with aspects of the cellular and molecular mechanisms underlying the patterning. We examine the genesis of arteriovenous malformations (AVMs) that link defects in vessel patterning to vascular dysfunction and disease, and we conclude by highlighting some important areas of research and unanswered questions that are ripe for further investigation.

OVERVIEW

Concepts in Vessel Patterning

The concept of blood vessel patterning has been expanded and refined in the last 10 years. During organ development, endothelial cells coalesce into a honeycomb-shaped primary capillary plexus and invade avascular organs to form a primitive vascular bed. This new blood vessel sprouting from preexisting vascular beds is defined as sprouting angiogenesis, and organ-derived angioblasts also differentiate and contribute to vessels in some organs via a process called vasculogenesis. As blood vessels sprout, endothelial tip cells with filopodia guide the developing capillary sprout toward an angiogenic stimulus such as hypoxia-induced VEGF-A, while endothelial stalk cells proliferate as they follow behind the tip cells, causing the capillary sprout to elongate (Fig. 1A; Gerhardt et al. 2003; Pelton et al. 2014). The tip and stalk cells are endothelial cell phenotypes based on their position in the plexus, and they exchange position over time with stalk cells overtaking tip cells (Jakobsson et al. 2010; Arima et al. 2011; Bentley et al. 2014). Once a primitive organ vasculature forms, it often undergoes intensive vascular remodeling triggered by blood flow and develops into a branched, hierarchical network of large- and small-caliber vessels (Fig. 1A). A historical focus on the contribution of sprouting and network expansion to blood vessel patterning has been complemented by a new appreciation for the role that vessel remodeling plays in patterning many vascular beds once the primary vascular plexus has formed.

Figure 1.

Modes of blood vessel patterning. (A) (Upper panels) Whole-mount image of mouse embryo back skin at E13.5 (upper left panel) and E15.5 (upper right panel) stained with the pan-endothelial cell marker PECAM1 (green). Upper left shows primary plexus and upper right shows remodeled vascular network. Lower panels show changes over time. Sprouting angiogenesis is initiated by migrating tip cells followed by stalk cells (bottom left). The primary plexus undergoes blood flow–induced vascular remodeling to form a hierarchically organized vascular network (bottom right). (B) Whole-mount image of Tg(fli:LifeAct-GFP) zebrafish trunk vasculature at 28 hpf (left) and 36 hpf (right) expressing a tagged actin-binding protein in endothelial cells. Intersegmental vessels (ISVs) sprout dorsally from the dorsal aorta and connect with the dorsal longitudinal anastomotic vessel (DLAV) by 28 hpf; this pattern is essentially unchanged at 36 hpf. The caudal vein plexus (CVP) sprouts ventrally from the dorsal vein and undergoes extensive flow-mediated remodeling from 28 hpf to 36 hpf (lower higher magnification panels).

In some cases, organ vessel networks adopt a specific pattern as they form via sprouting angiogenesis, and they do not initially remodel extensively—this sequence is often associated with sprouting vessels that enter tissues comprised of defined structures that produce pro- and antiangiogenic cues. These cues allow sprouting vessels to invade certain areas and avoid others; the patterns presumably allow for further organ development as vessels invade. One example is the patterning of zebrafish intersegmental vessels (ISVs) that sprout from the dorsal aorta below somite borders and extend dorsally between the borders to form a connection with the dorsal longitudinal anastomotic vessel (DLAV) (Fig. 1B). Although some ISVs eventually remodel to form a venous return, the initial sprouting sets the overall embryonic pattern. Another example is neural tube vascularization in birds and mammals, where the initial pattern of ingressing vessels is regulated by signals from the neural tissue (James et al. 2009; James and Mukouyama 2011; Ruhrberg and Bautch 2013; Vieira et al. 2020), as discussed later in the article. In both cases, tissue structures, either the somites or regions of the developing neural tube, express antiangiogenic and/or repulsive cues that prevent vessels from forming random patterns in the tissue.

In contrast, the sprouting that forms the initial vessel plexus in many tissues is relatively unregulated, and the original pattern of the primitive plexus formed by sprouting angiogenesis is quickly remodeled, often in conjunction with increased blood flow. An example of this type of vascular patterning is the CVP that sprouts ventrally in the zebrafish to form a network that is then extensively remodeled via flow-mediated signals to form the ventral vein (Fig. 1B). This form of vessel patterning is also common in many mammalian vascular beds; for example, in the postnatal retina, a primitive plexus is remodeled behind the advancing vascular front via flow-dependent signals, and a similar process occurs in the embryonic back skin (Fig. 1A). Recent studies of vascular remodeling provide a framework to explain flow-induced vascular remodeling through directional migration of endothelial cells (Franco et al. 2015; Fonseca et al. 2020). Vessels that are exposed to low or no flow are more likely to be pruned, while conduits closer to developing arteries are exposed to higher blood flow that protects them from pruning. As increased levels of shear stress in vessels induce endothelial cell polarization, increased asymmetries in shear stress between juxtaposed vessel segments lead to endothelial cell migration away from low-/no-flow vessel segments toward high-flow vessels, resulting in pruning of low-/no-flow vessel segments. How endothelial cell migration is controlled during remodeling to form tissue-specific vascular patterns is poorly understood and an area of active investigation.

Soluble/Tethered Vascular Patterning Signals

Numerous signaling pathways regulate blood vessel patterning, and here we describe a set of core signals involved in vascular patterning in different organs at different times. Many signals influence endothelial cell behaviors and vessel patterns via ligand binding to membrane-localized receptors on endothelial cells; this instigates signaling through effector pathways. Many of the downstream effector cassettes are shared by different upstream signals, while others have dedicated (often called canonical) pathways within the endothelial cell that also regulate endothelial cell status and behaviors (Fig. 2). Within the VEGF signaling family, it has become more appreciated that other family members affect blood vessel formation and patterning. For example, the receptor VEGFR3, which becomes associated with lymphangiogenesis over time, initially contributes to blood endothelial cell expansion and patterning and maintains a complex relationship with VEGFR2 signaling (Dumont Daniel et al. 1998; Tammela et al. 2008; Benedito et al. 2012; Zarkada et al. 2015; Heinolainen et al. 2017). Another signaling pathway involved in vessel patterning is the Notch pathway (Fig. 2). Notch differs from most other vascular patterning pathways in that the ligand is tethered to one endothelial cell, and it binds the Notch receptor on a neighboring cell. The binding causes tension through the complex that initiates several cleavages, resulting in a soluble cytoplasmic peptide (Notch intracellular domain [NICD]) that translocates to the nucleus of the Notch-expressing cell, where it complexes with different transcription factors to alter gene expression. There is also evidence for a noncanonical Notch signaling pathway in endothelial cells that is operative under flow (Mack et al. 2017; Polacheck et al. 2017). Notch intersects with VEGF-A signaling, and this interaction was originally described as a feedback loop that sets up differences between tip cells and stalk cells (Jakobsson et al. 2009, and references therein). Tip cells are posited to be initially exposed to more VEGF-A ligand because of their position in the ligand gradient, which leads to CDC42-dependent filopodia formation, and these filopodia may further increase tip cell exposure to VEGF-A (Fantin et al. 2015; Zakirov et al. 2021). Elevated VEGF-A signaling by tip cells leads to up-regulation of the Notch ligand Dll4, which then elevates Notch signaling in neighboring stalk cells; elevated Notch in turn down-regulates VEGF-A signaling via effects on VEGF-A receptor expression (Jakobsson et al. 2009). This idea has been refined over the last 10–15 years (Siekmann et al. 2013), and current models indicate that Notch signaling heterogeneity requires regulated levels of VEGF-A signaling; elevated VEGF-A signaling “entrains” Notch signaling so that groups of endothelial cells cycle together between high- and low-Notch status, leading to blood vessel dysmorphogenesis (Bentley et al. 2008). VEGF-A and Notch signaling are also involved in arterial differentiation. VEGF-A signaling acts upstream of Notch to induce expression of Dll4 and Notch4 through MAPK (mitogen-activated protein kinase)-dependent ETS factors in arterial endothelium (Lawson et al. 2002; Wythe et al. 2013). Notch activation influences endothelial cell-cycle status and function, leading to expression of arterial identity genes (Fang et al. 2017).

Figure 2.

Endothelial cell signaling pathways. Diagram showing some major endothelial cell signaling pathways used at multiple times and places for vascular patterning. (NRP) Neuropilin, (ENG) endoglin, (MAPK) mitogen-activated protein kinase, (PI3K) phosphatidylinositol 3-kinase, (AKT) protein kinase B. (Created with BioRender.com.)

Notch also intersects with BMP signaling in blood vessel patterning, likely in several different ways. Canonical BMP signaling allows effectors to enter the nucleus to influence gene expression—SMAD1/5 cytoplasmic effectors are phosphorylated downstream of ligand binding to membrane-localized receptors; phosphorylation allows for binding to SMAD4, a co-SMAD that chaperones the complex into the nucleus (Fig. 2; David et al. 2009). There is evidence of nuclear BMP-Notch cooperativity, perhaps via DNA-binding complexes, to transcriptionally regulate targets such as Herp2 and Hey (Itoh et al. 2004; Larrivée et al. 2012). Notch also regulates expression of a negative regulator of BMP signaling, SMAD6, such that elevated Notch signaling increases SMAD6 expression and reduces sprouting in response to BMP ligand (Mouillesseaux et al. 2016). While BMP signaling patterns the extent of plexus branching in sprouting assays in vitro, it has been difficult to determine how and where this aspect of BMP signaling is operative in vivo, perhaps because in addition to a proangiogenic BMP signaling arm there exists an antiangiogenic or homeostatic BMP-signaling arm that may be dominant under most in vivo conditions (see below).

Chemokine signaling is required for proper vessel patterning, often to orient vessels relative to other structures in tissues (Fig. 2; Kiefer and Siekmann 2011). In vessel patterning, the most common ligand is CXCL12 (also known as SDF-1) that binds CXCR4, a G-protein-coupled receptor. Chemokine signaling releases the Gα subunit from the inhibitory βγ subunits, and Gα inhibits adenyl cyclase signaling while promoting PI3K and MAPK signaling, while βγ mobilizes intracellular calcium via PLC (Döring et al. 2014). CXCL12 binding to CXCR4 also recruits β-arrestin; this interaction leads to receptor endocytosis and desensitization but also increases MAPK signaling. A second chemokine receptor, CXCR7, is also involved in the pathway as a decoy that regulates CXCL12 availability (Luker et al. 2010). This pathway is used for pathfinding and appears to be instructive for directional migration. In vascular development, CXCL12 and CXCR4 are preferentially expressed by arteries or in tissue areas close to arteries and in endothelial tip cells at the vascular front (Kiefer and Siekmann 2011). In addition, endothelial CXCR4 expression is up-regulated in response to blood flow (Corti et al. 2011), suggesting a link to flow-directed endothelial cell migration. Functional studies reveal that CXCL12-CXCR4 signaling is important for organ- and tissue-specific blood vessel patterning (Tachibana et al. 1998; Li et al. 2013, 2021; Cavallero et al. 2015; Ivins et al. 2015).

Mechanical Cues in Vascular Patterning

In the last 10 years, the importance of mechanical cues in blood vessel patterning has become apparent. We have also come to better understand how these signals are translated into molecular signals in endothelial cells, and how signaling pathways are regulated downstream of mechanotransduction. Mechanical inputs often extensively mature and refine blood vessel patterns, although the exact force-derived signals that drive vascular remodeling in specific organs to form tissue-specific vascular branching networks are poorly understood. Collectively, the inputs use mechanotransduction to translate physical force vectors into signals that mold endothelial cell behaviors. The forces include those generated by blood flow—the outward pressure of blood against the vessel wall—and the shear stress vector generated by a relatively viscous liquid (blood) moving across the apical side of endothelial cells (Davies et al. 2010; Baeyens et al. 2016a). Forces imparted from the local microenvironment, such as substrate stiffness, also influence endothelial cell behaviors. The force most relevant to endothelial cell responses in blood vessel patterning is thought to be shear stress; molecular complexes that directly sense shear and translate that mechanical cue into a biochemical signal have been identified, and other pathways respond to these cues downstream of mechanotransduction complexes.

Examples of molecules and complexes that directly transduce shear stress forces in endothelial cells are Piezo and other channels, a complex of VE-cadherin/PECAM1/VEGFR2, Notch, and a PlexinD1/NRP1/VEGFR2 complex (Ando and Yamamoto 2013; Mehta et al. 2020). The endothelial glycocalyx consists of proteins on the luminal surface, primarily proteoglycans, that confer a net negative charge; the glycocalyx is also implicated in mechanotransduction (Tarbell and Pahakis 2006; Jiang et al. 2021), although the mechanisms are not well understood. The best understood molecular complex is VE-cadherin/PECAM1/VEGFR2 (Tzima et al. 2005), as tension sensor constructs revealed that laminar shear forces increase tension on PECAM1 while reducing tension on VE-cadherin molecules in junctions (Conway and Schwartz 2015). Notch mechanosensing is less well understood, but it is thought that initial transduction of mechanical signals through Notch does not use canonical signaling to the nucleus, while homeostatic flow-mediated Notch signaling may require the canonical pathway (Mack et al. 2017; Polacheck et al. 2017; Ruter et al. 2021). Once mechanical signals have been translated into biochemical signals in the endothelial cell, numerous pathways respond to flow by signaling alterations. For example, BMP signaling is thought to switch from a proangiogenic to an antiangiogenic or homeostatic signaling arm, presumably by using the endoglin (ENG) coreceptor to sensitize endothelial cells to a particular set of ligands, BMP9 and BMP10, which have high affinity for a BMP receptor subunit (ALK1) that transduces signals leading to junction stabilization and reduced proliferation (Baeyens et al. 2016b; Bautch 2019).

EXAMPLES OF ORGAN-SPECIFIC VASCULAR PATTERNING

CNS Vascular Patterning: Neuronal Control of CNS Vascularization via Sprouting Angiogenesis

Central nervous system (CNS) development begins around embryonic day (E)7.5 in the mouse, when the neural plate that will ultimately form the brain and spinal cord folds to form the neural tube. After neural tube closure, neuroepithelial cells generate progenitors that differentiate into neurons and support cells called glia. As the progenitors undergo active proliferation and differentiation, developing CNS tissues are vascularized in response to increasing oxygen and nutrient demands. CNS vascularization requires communication between neural tissues and vascular endothelial cells. Between E8.5 and E10.5, paraxial mesoderm-derived endothelial cell progenitors called angioblasts form a primitive vascular bed surrounding the neural tube, called the perineural vascular plexus (PNVP). Subsequently, these vessels sprout and invade the CNS parenchyma to form a rich capillary plexus (Fig. 3B,C; Hogan et al. 2004; Bautch and James 2009; Paredes et al. 2018). A periventricular vessel plexus (PVP) forms around the telencephalon (embryonic forebrain) and expands in a ventral-to-dorsal fashion to connect with vessel sprouts originating from the PNVP (Fig. 3A). The PVP originates from a prominent basal vessel located deep on the floor of the telencephalic vesicle. Ventral-to-dorsal vessel sprout distribution is regulated by homeobox transcription factors that are expressed regionally in both the vascular and neural compartments, such as Nkx2.1 and Dlx1/2 ventrally and Pax6 dorsally (Vasudevan et al. 2008; Paredes et al. 2018; Vieira et al. 2020).

Figure 3.

Central nervous system (CNS) vascularization through sprouting angiogenesis. (A) Coronal schematic illustration of sprouting angiogenesis in the developing mouse forebrain. By E10–E11, perineural vascular plexus (PNVP) surrounds the developing forebrain, then PNVP sprouts ingress radially from the pial surface toward the lateral ventricle (LV). From E10 to E12, a periventricular vessel plexus (PVP) derives from vessels in the basal telencephalic floor that progress tangentially from the ventral to the dorsal telencephalon (blue arrows). (B) Neural progenitor–derived VEGF-A and Wnt regulate CNS vascularization. VEGF-A controls PNVP formation and sprouting angiogenesis. Wnt7a/b affects sprouting angiogenesis and blood–brain barrier (BBB) development via canonical Wnt/β-catenin signaling. (Figure created from modified data in James and Mukouyama 2011.) (C) Schematic illustration of sprouting angiogenesis in the developing mouse spinal cord. Between E9.5 and 10.5, vessel sprouts from PNVP invade radially the primitive spinal cord. Motor neuron columns remain avascular until E12.5. (RP) Roof plate, (FP) floor plate.

Two major components of the neural tube-derived proangiogenic signal are VEGF-A and Wnt: VEGF-A controls PNVP formation and sprouting angiogenesis, whereas canonical Wnt/β-catenin signaling affects sprouting angiogenesis and blood–brain barrier (BBB) development (Figs. 2 and 3B). VEGF-A is up-regulated in the neural tube prior to PNVP formation (James and Mukouyama 2011; Tata et al. 2015; Paredes et al. 2018), and neuroepithelium-derived VEGF-A expression in response to hypoxia controls endothelial cell migration onto the outer (pial) neural tube surface as the PNVP develops (Hogan et al. 2004; Bautch and James 2009). Subsequently, neural tube–derived VEGF-A stimulates sprouting angiogenesis into the neural tissue from the PNVP (Gerhardt et al. 2003; Haigh et al. 2003; Raab et al. 2004; James et al. 2009; Rattner et al. 2019). Canonical Wnt/β-catenin signaling is not required for PNVP formation, but it is required for intraneural capillary plexus formation via sprouting angiogenesis. Neuroepithelium-derived Wnt7a/b (Stenman et al. 2008; Daneman et al. 2009) engages with endothelial G-coupled receptor GPR164 to induce canonical Wnt/β-catenin signaling and stimulate sprouting angiogenesis (Zhou and Nathans 2014; Posokhova et al. 2015; Vanhollebeke et al. 2015). RECK (reversion-inducing, cysteine-rich protein with Kazal motifs; a GPI-anchored membrane protein) is a coreceptor for Wnt7a/b expressed by CNS endothelial cells and important for canonical Wnt/β-catenin signaling during CNS vascular development (Ulrich et al. 2016; Cho et al. 2017). Despite the importance of both VEGF-A and canonical Wnt/β-catenin signaling for vessel ingression into the CNS, how these signals converge during CNS vascularization remains to be answered.

Transforming growth factor β (TGF-β) also regulates CNS vascularization in a brain region–specific manner. β8 integrin, expressed in neural progenitors, activates extracellular matrix–bound latent TGF-β, which then signals to endothelial cells (Arnold et al. 2014; Hirota et al. 2015; Ma et al. 2017). Indeed, neuronal deletion of β8 integrin or endothelial deletion of TGF-β type II receptor (Tgfbr2) leads to PVP vascular malformations and hemorrhage in the forebrain, and the vascular malformations result from abnormal sprouting angiogenesis and endothelial cell proliferation.

Blood vessels ingress and migrate within the CNS in a stereotypical pattern. This is especially visible in the neural tube (Fig. 3C), where the vessel ingression pattern is influenced by different neural progenitor domains along the dorsoventral axis (Paredes et al. 2018; Vieira et al. 2020). Between E9.5 and E10.5 in the mouse, sprouts invade the spinal cord ventrally between the floor plate and the motor neuron column, and along the lateroventral side in proximity to motor neurons. Subsequently, capillaries surround but do not initially invade the motor neuron column as they continue growing dorsally (Fig. 3C). As described above, VEGF-A is required for blood vessel ingression but is broadly expressed by the entire spinal cord, indicating that additional signals control the location and timing of PNVP blood vessel sprouting. These signals include soluble VEGF decoy receptor Flt1 (sFlt1), which helps control initial sprouting into the spinal cord in mouse (Himmels et al. 2017), chick (James et al. 2009; Takahashi et al. 2015), and zebrafish (Matsuoka et al. 2016, 2017; Wild et al. 2017). Motor neuron–expressed sFlt1 contributes to the avascularity of the motor neuron column (Himmels et al. 2017). The chemorepulsive molecule Sema3E is also expressed by motor neurons and suppresses VEGF-A-mediated sprouting angiogenesis (Takahashi et al. 2015).

Canonical Wnt/β-catenin signaling is also important for PNVP-derived vessel ingression patterning. Wnt7a/b is expressed by ventral and dorsal neural progenitors surrounding the ventricle at E10.5, and Wnt7a/b double mutants are devoid of blood vessel ingression ventrally, while mutants of the Wnt effector β-catenin lack vessel ingression in the entire spinal cord (Stenman et al. 2008; Daneman et al. 2009). Since Wnt7a/b double mutants show dorsal ingression, other Wnt ligands likely regulate dorsal spinal cord vascularization.

Although blood flow is likely involved in subsequent vessel remodeling in the CNS, how the primitive capillary plexus in the CNS remodels to form a hierarchically functional vascular network is poorly understood. One clue comes from the finding that sphingosine 1-phosphate receptor-1 (S1P1/S1PR1), a flow-regulated pathway in endothelial cells, suppresses angiogenic sprouting (Gaengel et al. 2012; Jung et al. 2012), thus regulating the transition from sprouting angiogenesis to vascular remodeling in a flow-dependent manner.

Skin Vascular Patterning: Coordinated Branching of Blood Vessels and Nerves

The skin has three layers: the surface epidermis provides a waterproof barrier; the underlying dermis has tough, supportive connective tissues that provide structure and support; and the deepest layer, the hypodermis, consists of fat and connective tissue. The dermis supports a rich layer of blood and lymphatic vessels, and it also contains numerous nerve endings, including a wide variety of cutaneous sensory nerve receptors. As skin develops, the dermis becomes vascularized via a primary capillary plexus that forms by sprouting angiogenesis around E13.5 in the mouse (Fig. 1). The primary capillary network undergoes intensive vascular remodeling and transforms into a hierarchical branching network by E15.5 (Fig. 1).

Recent studies show that branching patterns of blood vessels and nerves align with each other in the skin, suggesting that this alignment is functionally important. Dermal neurovascular alignment often results from cross talk such that the pattern of nerves determines the pattern of blood vessels or vice versa (Mukouyama et al. 2002; James and Mukouyama 2011), although vessel–nerve patterning is independently regulated by common guidance cues in some situations (Oh and Gu 2013; Andreone et al. 2015). Neurovascular cross talk is important in the skin of the developing limb, where a primary capillary plexus invades the dermis at E11–E12, and sensory nerves subsequently invade the skin. As the primary vascular plexus remodels at E14–E15, arterial branches follow the pattern of major sensory nerve fibers that is already set (Fig. 4). Alignment is established by sensory nerve–mediated vascular branching, because in mutants with misguided sensory nerves, arteriolar branches still align with misrouted nerves (Mukouyama et al. 2002). Vascular smooth muscle cells (VSMCs) associate with branched, remodeled blood vessels only after the vessels align with nerves and express arterial markers such as neuropilin-1 (NRP1), ephrinB2, and Connexin 40 (CX40). Thus, endothelial arterial differentiation and VSMC association is initiated in skin vessels only after they associate with nerves in a pattern of nerve-vessel alignment (Mukouyama et al. 2002). At the molecular level, sensory nerve–derived VEGF-A controls arterial differentiation through VEGF-A signaling via a VEGFR2/NRP1 high-affinity receptor complex, and nerve-derived CXCL12 directs vessel branching and alignment with sensory nerves through its receptor CXCR4 on endothelial cells (Mukouyama et al. 2002, 2005; Li et al. 2013).

Figure 4.

Coordinated branching of blood vessels and nerves in skin. At E11–E12, a capillary plexus forms in the skin via sprouting angiogenesis. Next, sensory nerves, composed of sensory axons and migrating Schwann cells, invade the skin. Between E14 and E15, plexus vessels remodel to form large-diameter vessels alongside sensory nerves that subsequently differentiate into vascular smooth muscle cell (VSMC)-covered arteries. Next, plexus vessels remodel to form venous branches alongside arteries. Between E15 and E17, sympathetic axons extend along arteries and innervate arteries. The panels also show important signaling for each step of the process. (EC) Endothelial cell. (Figure created from modified data in James and Mukouyama 2011.)

Venous branches follow arterial branches starting at E15–E17 in the mouse skin, after nerve–artery copatterning is established (Fig. 4). Arteriovenous alignment is established by artery endothelial cell–derived apelin, which controls venous branching through its receptor APJ in vein endothelial cells (Kidoya et al. 2015). Embryos lacking apelin or APJ have reduced arteriovenous alignment in the skin, and arteriovenous alignment is important for thermoregulation (Kidoya et al. 2015).

The next stage moves from sensory nerve–vessel interactions to sympathetic nerve–vessel interactions. Autonomic sympathetic axons extend along arteries and innervate skin targets such as hair follicles and glands (Glebova and Ginty 2005). In parallel, sympathetic axons innervate VSMC-covered arteries to control vascular tone and participate in blood pressure regulation (Fig. 4; Eichmann and Brunet 2014). In developing limb skin, sympathetic axons emerge in close proximity to arteries at E15 and follow arterial branches in response to arterial VSMC-derived NGF (YS Mukouyama, unpubl.), while a VSMC-derived axon guidance molecule, netrin-1, facilitates sympathetic innervation of arteries through its receptor DCC (Brunet et al. 2014). Thus, the branching pattern of arteries determines that of sympathetic axons via arterial VSMC-derived guidance cues.

Although sensory nerves determine the pattern of arterial branching and vessels determine sympathetic axon patterning in the skin, there are questions remaining. For example, how do sensory nerve–derived proangiogenic signals affect endothelial cell behaviors to selectively form arteries alongside nerves? A subset of endothelial cells in the primary capillary plexus express CXCR4, suggesting that CXCR4⁺ endothelial cells are prespecified to associate with nerves. During vascular remodeling of the primitive capillary plexus, nerve-derived CXCL12 recruits CXCR4⁺ endothelial cells to form large-diameter vessels to align with the nerves (Li et al. 2013). Then, nerve-derived VEGF-A promotes arterial differentiation (Mukouyama et al. 2005; Li et al. 2013). Indeed, CXCL12 exposure was not sufficient for arterial differentiation, but subsequent exposure to VEGF-A led to an arterial phenotype. It will be important to determine what controls CXCR4 expression in a subset of early endothelial cells, and how CXCL12-CXCR4 signaling affects endothelial cell behaviors to promote arterial alignment.

Coronary Vascular Patterning: Multiple Progenitors and Blood Flow Contributions

Complex developmental programs orchestrate mammalian heart development. Cardiac development begins at approximately E8 in mice and includes cardiac progenitor cell specification, morphogenesis of the linear heart tube, and cardiac looping. The tube is initially comprised of an outer myocardial layer and an inner endocardial layer. At around E10, the heart tube acquires the third and outermost cardiac layer, the epicardium. The coronary vasculature forms as the myocardium expands between E11 and E16.

The origins of the endothelial cells that contribute to the coronary vessels are complex. Three sources of coronary endothelial cells were identified by lineage tracing: the sinus venosus (SV), the endocardium, and the proepicardium; the first two sites are the major sources, and one source can compensate for deficiencies of the other source to assure heart vascularization (Sharma et al. 2017; Lupu et al. 2020). SV-derived endothelial cells form most of the primitive arterial and venous networks by sprouting, with subsequent contributions from some endocardium-derived endothelial cells (Red-Horse et al. 2010; Chen et al. 2014b; Zhang et al. 2016). The venous plexus initially forms superficially in the subepicardium (Fig. 5A), while the arterial plexus forms deep within the myocardium (Fig. 5B).

Figure 5.

Coronary vessel patterning through sprouting angiogenesis and vascular remodeling. (A) Venous plexus from the sinus venosus (SV) sprouts over the surface (subepicardium) of heart ventricle and then remodels to form a hierarchical venous network. (B) Arterial plexus from the SV forms deep within the myocardium. The plexus vessels attach to the aorta to initiate blood flow, triggering the vascular remodeling of the arterial plexus to form a hierarchical arterial network. (PT) Pulmonary trunk, (AO) aorta, (LA) left atrium, (RA) right atrium, (LV) left ventricle, (RV) right ventricle.

Each capillary plexus then is transformed into a hierarchical vascular network, but how this happens relative to vessel remodeling differs. The arterial plexus initially forms unconnected to a source of blood flow, then connects to the dorsal aorta to initiate blood flow and vascular remodeling. How does this connection to initiate blood flow occur? Plexus vessels near the aortic trunk grow into and fuse with the aorta to form coronary–aorta connections, called coronary artery stems. CXCL12 and VEGF-C secreted by cells of the aortic wall are essential for signaling to the plexus and anastomosis with the aorta (Chen et al. 2014a; Ivins et al. 2015). Formation of the coronary artery stem triggers the onset of blood flow to the plexus and vascular remodeling to form the hierarchical arterial network (Fig. 5B). Subsequent arterial remodeling is a similar flow-dependent process to that in the postnatal retinal vasculature; prespecified arterial endothelial cells are thought to migrate against the direction of blood flow toward the growing artery (Red-Horse and Siekmann 2019; Fonseca et al. 2020). DACH1, a transcription factor that regulates organ size in Drosophila, is crucial for recruitment of prespecified arterial cells through CXCL12-CXCR4 signaling; Dach1 deletion results in decreased CXCL12 expression, decreased endothelial cell polarization against flow, and small coronary arteries (Chang et al. 2017).

The coronary venous plexus also remodels to form a hierarchical venous network (Lavine et al. 2008; Red-Horse et al. 2010; Nam et al. 2013; Chen et al. 2014b). Although less is known about this process, the venous plexus is likely connected to the SV; thus, blood flow and remodeling occurs as the venous network expands. However, part of the venous plexus may develop initially without the SV connection, and one study concludes that Sema3D signaling is required for proper connection of coronary veins to the right atrium (Aghajanian et al. 2016). Thus, Sema3D is necessary for proper coronary venous connections but is dispensable for coronary arterial connections, while CXCL12-CXCR4 signaling is necessary for arterial connections but is dispensable for coronary venous connections.

What determines the stereotypic vessel patterning of coronary arteries and veins? One possible mechanism is nerve–vessel alignment, but the alignment that occurs appears to involve cross talk from vessels to nerves in the heart. Postganglionic sympathetic neurons that innervate the heart target the conduction system, cardiomyocytes and coronary vessels, and control heart rate, contractility, and myocardial blood flow, respectively. These sympathetic axons extend along the superior vena cava and SV, and venous endothelium-derived endothelin1 guides sympathetic axonal growth toward and along these coronary veins (Manousiouthakis et al. 2014; Poltavski et al. 2019). Sympathetic axons invade the heart ventricle after the hierarchical coronary network forms and VSMCs associate with remodeled arteries and veins, suggesting that vessel patterning regulates nerve branching in the heart. Moreover, within the subepicardium, coronary vein VSMC-derived NGF directs sympathetic axon growth along coronary veins; subsequently, myocardial coronary artery VSMCs express NGF, attracting axons toward the coronary arteries (Nam et al. 2013). These findings suggest that sympathetic axon branching is regulated via vessel–nerve cross talk. However, what determines coronary artery and vein patterning remains unclear.

Kidney Vascular Patterning: Building an Organ Whose Function Requires Vessel Patterning

The coordinated patterning of the vasculature during kidney organogenesis is complex and fascinating. In addition to providing oxygen and nutrients, vessels are central to kidney function via coordination of fluid and solute exchange at several locations, including the glomeruli and renal tubules. Murine kidney development begins at E10.5, when the ureteric bud stalk enters the metanephric mesenchyme, and capillaries associated with the bud enter the kidney rudiment as well (Munro et al. 2017b; Daniel et al. 2018). The nephrogenic mesenchyme is also thought to be a source of endothelial cells, along with the adjacent peri-Wolffian mesenchyme. A perfused primitive plexus that is connected to the common and caudal iliac artery forms. The nephrogenic zone then expands as the ureteric bud bifurcates to form new branches. The nephrogenic mesenchyme associated with the bud, called the cap mesenchyme, splits and maintains proximity to the bud, and inductive cross talk between these tissues is crucial to kidney development (Costantini and Kopan 2010). The endothelial plexus and the expanding ureteric bud/mesenchyme continue to copattern as the bud expands and supports the formation of the individual units that will mature into nephrons, but the vessels remain outside the bud–mesenchyme units.

As nephrons begin differentiation, the surrounding vascular plexus initially remains excluded from the structure. As nephrogenesis reaches a stage where an S-shaped body forms, a capillary loop is found next to maturing podocytes within the glomerulus; this primitive structure maintains connections to the surrounding plexus and eventually matures into several capillary loops in intimate proximity to the podocytes to form a glomerular capillary tuft within Bowman's capsule (Fig. 6; Herzlinger and Hurtado 2014). Mesangial cells subsequently migrate into the glomerulus. Podocytes extend structures called foot processes that form intimate contacts with endothelial cells, leading to endothelial cell fenestration that facilitates transfer of materials between endothelial cells and podocytes. Fluid and solutes are transferred from the blood to the renal side of the system in the glomeruli, and this fluid mix exits the kidney via a complex network of extensively patterned cortical renal tubules (Fig. 6). Glomerular filtrate first flows through the proximal convoluted renal tubule, then the loop of Henle, and finally the distal convoluted renal tubule to the collecting tubules. These tubules are very near to peritubular capillaries that expand from the plexus surrounding the nascent nephron, and the copatterning of kidney tubules and vessels allows for reabsorption of fluids and solutes into the blood to maintain the salt and water balance of the organism. The capillaries called the vasa recta follow the renal tubules into the medulla to bring oxygen and nutrients to the medullary zone of the kidney before ascending back to the cortex; the vasa recta also absorbs fluid back into the bloodstream, then drains into a venous system that exits the kidney via the renal vein.

Figure 6.

Structure and cellular components of kidney vasculature. Inductive interactions of the ureteric bud with nephrogenic mesenchyme initiate nephron development (not shown). The developing glomerulus recruits endothelial cells that form a complex with podocytes that filters fluid and solutes from blood (green arrows), and close apposition of kidney tubules (proximal and distal convoluted tubules, loop of Henle) with peritubular capillaries (vasa recta and others) allows for resorption of some of the glomerular filtrate before it exits the kidney (blue arrows). (Inset) VEGF-A from podocytes is important for formation and maintenance of the glomerular unit; podocyte-derived Sema3A prevents excessive glomerular vascularization. (Partially created with BioRender.com.)

Given the extensive copatterning of kidney nephron tubules and glomeruli with blood vessels and its functional significance, surprisingly little is known regarding vascular patterning signals during kidney development. VEGF-A signaling is crucial for kidney vessel patterning (Tufró 2000), and both glomerular podocytes (Eremina et al. 2003; Vaughan and Quaggin 2008) and renal tubules (Dimke et al. 2015) express VEGF-A that coordinates patterning. VEGF-A is also upstream of mesangial cell migration into the glomerulus, perhaps via regulation of PDGFB that is secreted by endothelial cells. Other vascular patterning cues implicated in kidney vascular–glomerular function include CXCL12-CXCR4 and Ang-Tie2 (Bartlett et al. 2016), but their potential roles in kidney vessel patterning are not well understood. The role of potential repulsive cues is also incompletely understood. Sema3A and Sema3F and their NRP receptors are expressed at appropriate times and places to be involved in kidney vascular patterning (Villegas and Tufro 2002), and Sema3a loss-of-function leads to excess glomerular vessels (Reidy et al. 2009; Reidy and Tufro 2011). However, direct evidence for Sema3F having a role in kidney vascular patterning in vivo is lacking, as genetic deletion of Sema3f or Nrp2 did not affect initial vessel patterning around the cap mesenchyme, although a later role in tubule patterning was not ruled out (Munro et al. 2017a). Thus, kidney vessel patterning is an area of active current research, and one future goal in the field is to develop kidney organoids that can replace dialysis in human patients.

VASCULAR PATTERNING GONE WRONG: ARTERIOVENOUS MALFORMATIONS

Arteriovenous malformations (AVMs) lack hierarchical vascular patterning and proper arteriovenous identity, leading to direct connections of dilated arteries and veins that bypass intervening capillaries. This shunt from the high-flow arteries to low-resistance veins results in flow disturbances and rupture-associated hemorrhage. Most AVMs arise spontaneously, although AVMs are observed in patients with genetic syndromes such as hereditary hemorrhagic telangiectasia (HHT), and this association has provided insight into crucial signaling pathways that govern vascular patterning (Atri et al. 2013; Lawton et al. 2015).

Mutations in several members of the BMP/TGF-β signaling pathway cause HHT: ENG (ENG, coreceptor, mutated in HHT1) (McAllister et al. 1994; Gallione et al. 1998), ALK1 (activin-like receptor kinase 1, type I receptor that interacts with ENG mutated in HHT2) (Johnson et al. 1996), SMAD4 (common BMP/TGF-β coeffector mutated in combined HHT/juvenile polyposis syndrome) (Gallione et al. 2004), and GDF2 (encodes BMP9 ligand that binds ALK1) (Wooderchak-Donahue et al. 2013). The ligands BMP9 and BMP10 bind ALK1 and ENG with high affinity and induce downstream SMAD signaling. Mouse studies show that postnatal conditional endothelial cell-specific deletion of Eng, Alk1, or Smad4 genes leads to retinal AVM development (Park et al. 2009; Mahmoud et al. 2010; Tual-Chalot et al. 2014; Ola et al. 2016, 2018; Jin et al. 2017; Alsina-Sanchís et al. 2018; Crist et al. 2018). These AVMs form in the remodeling capillary plexus where blood flow–induced shear stress is high, and they are accompanied by loss of arteriovenous identity (Tual-Chalot et al. 2014; Baeyens et al. 2016b; Ola et al. 2018). At the cellular level, deletion of Alk1 or Eng in capillaries and veins disrupts blood flow–mediated endothelial cell migration from low flow segments (capillaries and veins) toward high flow segments (arteries) during vascular remodeling, resulting in endothelial cell accumulation in capillaries that may cause capillary enlargement and AVM formation (Park et al. 2021). Consistent with the idea that endothelial cell mutations, blood flow, and an angiogenic stimulus are all required for AVM formation, BMP9 signaling through ALK1 is enhanced by blood flow–induced shear stress and suppresses endothelial PI3K (phosphatidylinositol 3-kinase)/AKT (protein kinase B) signaling, and endothelial cell–specific inactivation of Alk1 or Eng leads to enhanced VEGF-A and PI3K/AKT signaling (Han et al. 2014; Ola et al. 2016). Clinical improvements in symptoms are observed in HHT patients treated with VEGF signaling inhibitors (Dupuis-Girod et al. 2012), and pharmacological inhibition of VEGF-A or PI3K/AKT signaling reduces the severity and frequency of AVMs in mice lacking Alk1 or Eng function in endothelial cells (Ola et al. 2016; Jin et al. 2017).

Although some characteristics are shared by all AVMs, there are potentially important differences between HHT-associated AVMs and other AVMs, suggesting a different etiology. For example, AVMs that result from Notch perturbations are associated with failure of appropriate artery–vein fate specification. Notch receptors and ligands are preferentially expressed by arteries, and Notch signaling promotes arterial differentiation by enhancing arterial and suppressing venous properties (Gridley 2010). Ectopic expression of artery–vein markers characterizes Notch-induced AVMs, and both gain- and loss-of-function of Notch results in abnormal artery–vein specification and fusion of arteries and veins (Atri et al. 2013; Peacock et al. 2016). AVMs of mice with constitutively active Notch signaling more closely mimic human AVMs (Uyttendaele et al. 2001; Carlson et al. 2005; Kim et al. 2008; Murphy et al. 2009; Krebs et al. 2010). Notch activation promotes arteriovenous shunts that arise from enlarged capillary-like vessels, accompanied by increased endothelial cell size (Murphy et al. 2014). Interestingly, activation of Notch4 in endothelium of adult mice using a temporally regulatable system results in AVMs that are reversible upon loss of Notch4 transgene expression (Carlson et al. 2005; Murphy et al. 2008). This normalization leads to the conversion of large-caliber, high-flow AV shunts to capillary-like vessels and reexpression of the venous receptor EphB4, without changes in endothelial cell number, vascular damage, or hemorrhage (Murphy et al. 2012). Given that patient AVMs typically demonstrate Notch1 activation (ZhuGe et al. 2009; Murphy et al. 2012), targeting Notch may be a therapeutic strategy to induce AVM regression.

An intriguing question is whether Notch perturbations indirectly contribute to HHT (Peacock et al. 2016). As described in a previous section, activated SMADs downstream of ALK1 and ENG may physically interact with NICD (Itoh et al. 2004), SMAD1/5/8-binding sites are present in the regulatory region of many Notch downstream mediators (Morikawa et al. 2011), BMP9 induces Notch downstream mediators in culture (Larrivée et al. 2012), and Notch regulates expression of a BMP inhibitor in endothelial cells (Mouillesseaux et al. 2016; Ruter et al. 2021). Studies in mouse retinal vessels show that either ALK1 inhibition or Notch signaling loss results in a hyperfused and hypersprouting capillary plexus, and that Notch inhibition–induced hypersprouting is counteracted by the ALK1 ligand BMP9 (Larrivée et al. 2012). These studies suggest that Notch and BMP9-ALK1 signaling interact to promote normal vascular development. The involvement of Notch signaling in HHT-associated AVMs, however, remains unclear. AV shunts appear to form by different mechanisms in HHT- and Notch-associated AVMs. HHT-related AVMs arise from reduced blood flow–directed endothelial cell migration, followed by endothelial cell proliferation (Ola et al. 2016; Sugden et al. 2017). In contrast, Notch-associated AVMs display reversible endothelial cell hypertrophy (Murphy et al. 2014). Given that Eng deficiency also induces blood flow–mediated endothelial cell hypertrophy (Sugden et al. 2017), further work is needed to determine whether Notch signaling functions downstream of HHT in AVM formation.

THE NEXT 10 YEARS IN VASCULAR PATTERNING

From a simple set of molecular interactions with a few pathways, the last 10 years has shown us that blood vessel patterning is complex. The complexity occurs at several levels: numerous molecular pathways are used to pattern vessels, and some of these pathways integrate with each other and use both morphogenic and mechanical cues; different tissues produce both pro- and antiangiogenic cues at different times to pattern organ-specific vessels while allowing the organ domains to function properly; and, in many cases, intimate endothelial cell–organ cell interactions accompany patterning that aligns vessels to participate in organ function. This complexity also leads to great beauty, both physically in the vessel patterns, and intellectually, as our understanding of the intricate molecular cross talk needed to shape and maintain these patterns expands.

What do we envision for the next 10 years? First, we think it will be important to further understand organ-specific vascular patterning. While some pathways and processes are used in many situations, each organ has also evolved unique vascular patterning signals and unique temporal regulation of the signaling, to best serve the development and function of that organ. The explosion of -OMICs approaches and single-cell analysis will be important to harness in addressing patterning questions, and further refinement of tools such as organ-specific vascular Cre excision lines will also help to better define local “rules” of vascular patterning. Second, it will be important to visualize vascular remodeling in vivo to better understand vascular patterning and extend our current use of zebrafish live imaging to mammalian developmental and disease models. For example, it is possible to use windows to follow changes to vessels in mammals (Huang et al. 2020; Nia et al. 2020; Velasco et al. 2021), and it is also possible to do longitudinal studies of vessel changes (Chong et al. 2017). These assays, when refined and combined with mosaic labeling using Cre-activated reporters, will allow for single-cell-level analysis of vessel sprouting and remodeling in an appropriate physiological environment and contribute to our understanding of vascular patterning. Third, we predict that advances in single-cell analysis and in noninvasive imaging will allow us to dissect human AVMs so that we better understand whether different signals leading to AVM formation (i.e., ALK1/ENG vs. Notch) converge on a single cellular mechanism, or whether different endothelial cell misbehaviors can lead to AVMs. Finally, advances in microfluidics will expand our ability to manipulate vascular patterning outside the body going forward, as we learn how to combine morphogenetic and mechanical inputs to manipulate vascular network formation and remodeling. This will ultimately lead us to effectively vascularize tissues and organoids in vitro for replacement in vivo. We predict that the next 10 years will increase our knowledge of vascular patterning, and that this knowledge will dramatically expand our ability to manipulate the system while appreciating the inherent beauty of blood vessel patterns.

COMPETING INTEREST STATEMENT

The authors declare no competing financial interests.