Specialized pulmonary vascular cells in development and disease

Abstract

Endothelial cells (ECs) develop organ-specific gene expression and function in response to signals from the surrounding tissue. In turn, ECs can affect organ development and morphogenesis and promote or hinder disease response. In the lung, ECs play an essential role in gas exchange with the external environment, requiring both a close physical connection and a strong axis of communication with alveolar epithelial cells. A complete picture of the composition of the pulmonary endothelium is therefore critical for a full understanding of development, maintenance, and repair of the gas exchange interface. Defining the factors that control lung-specific EC specification, establish EC heterogeneity within the lung, and promote the differing contributions of EC subtypes to development, health, and disease will facilitate the development of much-needed regenerative therapies. This includes targeting therapeutics directly to ECs, developing pluripotent or primary cell-derived ECs to replace damaged or diseased vasculature, and vascularizing engineered tissues for transplant.

1. INTRODUCTION

1.1. Inter- and Intra-Organ Endothelial Heterogeneity

Like other organs in the human body, the lung relies on the blood and lymphatic vasculature to maintain tissue homeostasis. In most organs, the blood vasculature transports gases, nutrients, signaling molecules, and cells to the surrounding tissue, while the lymphatic vasculature regulates fluid clearance and transports lipids and immune cells. Unlike other organs, however, the lung vasculature is specialized for its role in gas exchange with the cardiovascular system. To provide this critical function, endothelial cells (ECs) lining capillaries in the distal lung form a tight interface with alveolar epithelial type I (AT1) cells that permits the transfer of oxygen and carbon dioxide between them. This unique functional need drives the specialized development, structure, and regeneration of pulmonary ECs. The advent of single-cell RNA sequencing (scRNA-seq) has revolutionized our understanding of the exquisite heterogeneity of the lung endothelium and the contributions of endothelial plasticity to development and disease. However, the extent and function of EC heterogeneity in the lung remains an area of active study. Much remains to be discovered about the role of heterogeneous pulmonary ECs in building the functional unit of gas exchange, the lung alveolus; promoting tissue homeostasis during adult life; and repairing the gas exchange interface when it is damaged through injury or disease.

ECs in different organs share commonalities in function and embryonic origin but differ widely in morphology and gene expression depending on organ function (1, 2). In the brain, where barrier formation is paramount, ECs form a continuous endothelium linked by specialized tight and adherens junctions (3). In the kidney and small intestines, where filtration is essential, ECs form a fenestrated endothelium that allows less selective exchange of fluid and molecules (4, 5). In the liver, even greater permeability in sinusoidal ECs permits solute exchange (6). EC morphology and function is also influenced by the unique milieu of their organ-specific niche, including organ-specific growth factor signaling, mechanical forces, metabolic demands, oxygen tension, and interactions with supporting mural cells (1, 7). Lymphatic capillaries were assumed to be uniformly permeable for fluid uptake, but recent evidence suggests that lymphatic ECs also differ between organs (8). To some extent, function clearly dictates endothelial form, and distinct needs in different organs result in EC adaptation to organ-specific roles.

Organ-specific vascular beds are known to adapt to the needs of their specific environments to maintain homeostasis (9, 10), and recent advances in single-cell genomics have confirmed the essential paracrine signaling roles of organ-specific ECs during development, at homeostasis, and during regeneration (11-13). Genomic studies have also demonstrated intra-organ heterogeneity that is not limited to the distinction between veins, arteries, capillaries, and lymphatic vessels. ECs are heterogeneous within many healthy tissues in the mouse, including the heart, lung, fast- and slow-twitch muscle, large and small intestine, spleen, brain, testis, and kidney (13-18). EC heterogeneity has also been shown in human tissues such as the kidney, where many EC subtypes found in the mouse are also represented in humans (19). These technologies have expanded our knowledge of the composition of various organs and their variations during organ development and disease.

1.2. Angiocrine Signaling

Understanding EC heterogeneity is critical for understanding the ability of ECs to shape their niche through angiocrine signaling. EC angiocrine functions include promoting niche cell specification and differentiation, tissue morphogenesis and patterning, maintenance of homeostasis and metabolic function, and regeneration (12, 20). Organ-specific ECs express different cohorts of angiocrine factors that are specialized to the organ’s needs (11). Specialized EC signaling can aid in reconstitution of the bone marrow niche after myeloablation (21), coordinate organ-specific inflammatory responses (22), drive cancer metastasis and progression (23, 24), and promote functional regeneration or dysfunctional fibrosis (25, 26). This support is age-dependent: in the skeletal system, the vascular niche for hematopoietic stem cells (HSCs) declines with increasing age (27), and aged ECs induce an aging phenotype in young HSCs while young ECs partially rescue aged HSC phenotypes (28). In the lung, EC-epithelial co-culture experiments have identified EC thrombospondin signaling as essential for alveolar injury repair (29), and a platelet-EC signaling axis stimulates epithelial cell expansion to form new alveoli after pneumonectomy (30, 31). ECs are evidently a powerful driver of development and regenerative response within many organs; to harness this power towards new therapeutic approaches, it will be critical to understand both EC composition and function in development and injury response.

1.3. Endothelial Plasticity

ECs navigate development, maintenance of homeostasis, and tissue repair in part through their incredible plasticity. ECs undergo endothelial-to-hematopoietic transition (EHT) during early development in several organs (32, 33) or endothelial-to-mesenchymal transition (EndMT) during cardiac development (32, 34). Maintenance of differentiated EC fate within organs requires active input from the surrounding cells, often mediated through FGFR1 and VEGFR2 signaling (32). ECs transplanted from one organ into another will acquire the fate of the surrounding tissue, and ECs cultured outside of the organism assume a baseline proliferative fate (1). This plasticity is required for normal development, but can also contribute to pathology (32, 35). In the lung, maintenance of organ-specific EC fate is likely required for effective gas exchange, and reestablishing EC fates after injury may promote tissue repair.

This review discusses recent advances identifying the existence and function of heterogeneous pulmonary ECs in mouse and human and their role in development and disease. The power of ECs to adapt and respond to changes in their environment represents a strength that can be harnessed to promote regeneration, but it can also lead to dysfunctional adaptation. The importance of understanding pulmonary EC plasticity and heterogeneity is clear: pinpointing the cellular subtypes present will promote rational design of cell-based and targeted therapies to treat acute and chronic lung diseases and lung cancer, leading causes of death worldwide.

2. PULMONARY ENDOTHELIAL HETEROGENEITY IN MOUSE AND HUMAN

2.1. Identification of specialized pulmonary ECs in mouse development, homeostasis, and disease

The advent of scRNA-seq has provided unprecedented opportunities to examine the cellular composition of biological systems. Several groups have used this technology to investigate heterogeneity among lung cells, including ECs. This question is essential because ECs make up 40-50% of the lung (36) and may play an important role in building the alveolus during postnatal development (37) or healing the tissue during regeneration (38). On the epithelial side of the alveolus, alveolar type 2 (AT2) cells produce surfactant and serve as facultative progenitors while alveolar type 1 (AT1) cells perform gas exchange function (39, 40); whether EC subtypes could be specialized for proliferation or gas exchange remained unknown. This hypothesis has now been addressed in multiple contexts, providing the field with a better understanding of the composition of the lung vasculature and opening many more avenues of inquiry.

scRNA-seq in the mouse lung has repeatedly identified two distinct capillary EC subtypes in the lung. At mouse postnatal day (P)14, Aplnr+ and Plvap+ “Plvap ECs” and Apln+, Car4+ “Car4 ECs” can be identified in the distal lung (37). The two EC types differ in gene expression, location, ontogeny, and abundance (Figure 1A): Plvap ECs localize adjacent to arterial cells and express Mki67, a marker of proliferation, while Car4 ECs make up only 15% of all lung ECs, arise specifically at embryonic day (E)19, and are associated with alveolar “islands” between grooves that are regions of secondary septation (37). Monocle trajectory analysis suggests that Plvap ECs predate non-proliferative Car4 ECs and may be their precursors. Immunostaining for Car4 reveals a discrepancy between cell number and vessel contribution, suggesting that Car4 ECs may be larger than Plvap ECs; using sparse cell labeling to identify individual Car4 ECs confirms that these cells are highly branched, with contributions to 5-10 vessel segments (37) (Figure 1B). Differences in size and proliferative function suggest that these are truly independent EC types rather than transient states. This work also identified a higher level of Kdr (VEGFR2) expression in Car4 ECs, suggesting a specialized role in signaling with AT1 cells, the source of alveolar VEGFA (41). Indeed, deletion of Vegfa in the alveolar epithelium results in vessel defects and absence of Car4 ECs in scRNA-seq of mutant lungs at P7 and enlargement of alveolar airspaces at P21 (37). This suggests that Car4 ECs interact preferentially with AT1 cells (Figure 1A) and are essential for proper alveologenesis during postnatal lung development.

Figure 1. Identification of specialized pulmonary endothelial cells (ECs) in development, homeostasis, and disease.

(A) Capillary type 1 ECs (CAP1) and capillary type 2 ECs (CAP2) in the mouse and human lung differ in gene expression, abundance, and localization. CAP1s specifically express Plvap, Aplnr, and Gpihpb1. Mouse CAP2s specifically express Car4 and Apln, while mouse and human CAP2s express EDNRB. CAP2 ECs are located on the “thin side” of the alveolar wall in close contact with alveolar type 1 (AT1) epithelial cells. CAP1s are more abundant, while CAP2s make up approximately 15% of the capillary endothelium. (B) CAP2s also differ from CAP1s in size, spanning multiple vessel segments. (C) CAP1 ECs possess the ability to proliferate during development and regeneration, but rarely do so at homeostasis. Lineage tracing using CAP1-specific Cre mouse lines has shown that CAP1s can generate CAP2s. CAP2s arise during late embryonic and early postnatal life in the mouse, a process that requires VEGF signaling. AF1, alveolar fibroblast type 1; AF2, alveolar fibroblast type 2; AT1, alveolar type 1 epithelial cell; AT2, alveolar type 2 epithelial cell; CAP1, capillary endothelial type 1; CAP2, capillary endothelial type 2; RBC, red blood cell. Created with biorender.com.

scRNA-seq in the adult mouse identified the same two different capillary EC types, Gpihbp1-high miECs (analogous to Plvap ECs) and Car4-high miECs (analogous to Car4 ECs) (38). Immunostaining and RNAscope in control and influenza-injured lungs show that Gpihbp1 and Car4 are both highly expressed in alveolar capillary plexus endothelium. scRNA-seq after H1N1 influenza infection reveals that the same populations are present in the lung after injury (38). A significant increase in EC proliferation and the proportion of Car4-high miECs over the course of the regenerative process suggests that these cells may make essential contributions to lung regeneration. However, proliferating ECs after influenza infection are more transcriptionally similar to Gpihbp1-high miECs than to Car4-high miECs, suggesting that Car4-high ECs are not progenitor cells (38). Nevertheless, Car4-high ECs expressing proliferation markers can be identified by flow cytometry and immunostaining, indicating that proliferating ECs may upregulate CD34 and CAR4 as they move along a differentiation trajectory towards Car4-high EC fate (Figure 1C).

A third scRNA-seq study provided further evidence of mouse lung EC diversity. Gpihbp1-expressing “general capillaries,” or gCaps, express Edn1, Nos3, and Ptgis, suggesting they are a source of vasodilators and may be specialized to regulate vasomotor tone and interact with pericytes (39). gCap expression of MHC class II components suggests that they may also present antigens. Apln+, Car4+ ECs, termed “aerocytes” or aCaps in this study, express genes associated with leukocyte trafficking, sequestration, and adhesion (39). aCap expression of Kdr indicates that, as in alveologenesis, these cells may interact preferentially with AT1 cells in the adult (39). As gene expression differences only suggest functional differences, tamoxifen-inducible Cre mouse models that specifically label gCaps (Aplnr-CreERT2) and aCaps (Apln-CreERT2) were used to further probe their different roles in the alveolus. Lineage tracing and EdU incorporation studies demonstrate that gCaps proliferate during repair after elastase-induced injury but rarely at homeostasis, and gCaps can generate aCaps during repair or intermittently at homeostasis (39).

The current consensus in the field is that these and other groups have identified the same two EC types. Efforts to standardize nomenclature in the lung biology field have styled these cells CAP1 and CAP2 ECs (42) (Figure 1, Table 1). However, important differences between studies raise unanswered questions, identifying several topics for future investigation. The increase in the proportion of Car4-high CAP2s observed during regeneration after influenza and the observation of Car4-positive cells also expressing proliferation markers (38) contrasts with the designation of “aerocytes” as solely gas-exchanging cells (39). This discrepancy may be partially due to the difference between imaging modalities or between influenza and elastase injury methods, and it sheds light on how little functional evidence we have for the gas-exchanging properties of CAP2s beyond gene expression and location. Gpihbp1-high CAP1s and Car4-high CAP2s are both spread throughout the alveolar space (38) but appear to cover alveolar islands during secondary septation (37) and localize to “thin” alveolar walls in apposition to AT1 epithelial cells (39) (Figure 1A); this warrants further careful study in 3D and 4D. The extent of Car4 expression in the mouse lung vasculature is also unclear; one study reported Car4 expression in capillaries, veins, and interferon-activated ECs (16). The consequences of altered EC heterogeneity in lung function and disease also require further investigation (39), as these specialized pulmonary EC subtypes may make essential contributions to alveologenesis, gas exchange, and lung repair. These initial findings have since paved the way for examination of EC states in the human lung and in various lung disease states towards a greater understanding of the role of EC heterogeneity and plasticity in development and disease and the development of targeted therapeutics in the lung.

Table 1. Marker genes identifying EC subtypes in mouse and human lungs.

More detailed information about cellular subtypes in the lung is available from the Lung CellCards project through the Lung Molecular Atlas Program (LungMAP) Consortium, found at https://lungmap.net/cell-cards/.

Endothelial cell subtype	Marker genes in mouse	Marker genes in human
Pan-endothelial	Cdh5, Cldn5, Erg, Pecam1	CDH5, CLDN5, ERG, PECAM1
CAP1	Aplnr, Gpihbp1, Kit, Plvap	CD36, EDN1, FCN3, GPIHBP1, IL7R
CAP2	Apln, Car4, Ednrb, Tbx2	EDNRB, HPGD, SOSTDC1, TBX2
Arterial	Bmx, Cxcl12, Rbp7	BMX, CXCL12, DKK2, EFNB2, GJA5, SOX17
Venous	Slc6a2, Vegfc	ACKR1, NR2F2, SELP, VCAM1
Systemic venous	--	ACKR1, COL15A1, PLVAP, SPRY1
Lymphatic	Prox1, Thy1	FLT4, LYVE1, PDPN, PROX1

2.2. Heterogeneous endothelium in the human lung

Applying our increased knowledge of EC heterogeneity to understand and treat human disease is possible only if specialized EC subtypes in the mouse are conserved in the human lung. Mouse and human lungs differ in important ways, including the lack of respiratory bronchioles and respiratory airway secretory cells in mice (43). Early work suggested that CAP1 and CAP2 counterparts are present in humans (39, 44), and several scRNA-seq atlases in human lungs have since confirmed and extended these findings.

The human lung is highly vascularized and is composed of two circulations. The systemic circulation supplies oxygen-rich blood to the large airways and the pleura, while the pulmonary circulation brings oxygen-depleted blood into contact with the lung parenchyma for gas exchange (45). A meta-analysis of 15,142 vascular ECs from six human lung scRNA-seq datasets containing a total of 73 subjects highlights this diversity in EC function, identifying human pan-EC marker genes associated with the glycocalyx, focal adhesions and adherens junctions, transcription factors, signaling receptors, and regulators of EC migration, and creating a comprehensive list of human arterial, capillary, and venous markers (46). Comparison to a similar integrated analysis of 21,831 mouse ECs reveals that many of these genes and functions are shared between species (Table 1).

As in the mouse, the capillaries of the human lung contain two distinct EC populations (44, 46). However, unlike the mouse, CA4 (mouse Car4) is expressed by all human capillary ECs and is not an appropriate distinguishing marker. Human CAP2s specifically express HPGD, indicating that they might play a role in prostaglandin degradation, and the BMP receptor antagonist SOSTDC1 (46). Like mouse CAP2s, human CAP2s express the endothelin receptor EDNRB. Human CAP2s do not express genes essential for extravasation of leukocytes and maintenance of hemostasis associated with Weibel-Palade bodies, including VWF, SELP, and EDN (46). Human CAP1 ECs also possess a similar gene expression profile to mouse CAP1s, including expression of genes involved in lipid transcytosis, such as GPIHBP1 and CD36, and the innate immune response, such as FCN3 and CD14 (Table 1). Functional studies of CAP1 and CAP2 ECs in a human system will be challenging, but their similarity in gene expression to mouse capillary EC subtypes suggests a similar division of labor within the alveolus.

Several important differences between mouse and human lung ECs have also been identified. In the human lung, venous ECs expressing ACKR1 form two subpopulations: CPE⁺/DKK3⁺/PTGS1⁺ pulmonary venous ECs, and COL15A1⁺ systemic venous ECs (46). In the human lung, PLVAP expression is specific to systemic venous ECs, whereas in the mouse, Plvap is expressed by both CAP1 and large vessel ECs (37, 38, 46). This difference is underscored by further studies indicating that an increase in COL15A1-positive ECs occurs in patients with idiopathic pulmonary fibrosis (IPF) and nonspecific interstitial pneumonia (47, 48). This suggests that mouse and human differences in lung EC heterogeneity are relevant to vascular remodeling in human disease, highlighting the importance of studies in human tissue.

These subpopulations of human lung ECs are also observed by other meta-analyses, including the Human Lung Cell Atlas (HLCA) (49), which combines 49 individual datasets to form an integrated atlas of 486 human lungs containing 2.4 million cells (50). These scRNA-seq atlases serve as powerful references for studying rare cell types difficult to identify in single studies, allowing the assessment of changes in cellular composition and gene expression in disease states and differences in age, gender, and health status. These atlases have also driven advances in cell identification and nomenclature, including computational pipelines and web-based tools to improve cell identification in future scRNA-seq studies (51, 52). Although the use of lung cell atlases to study disease is reviewed elsewhere (53), it is essential to note that these tools have revolutionized our ability to identify cell types in transcriptomic space and provide a framework for future studies of disease states in mouse and human lung.

3. ENDOTHELIAL HETEROGENEITY IN DEVELOPMENT AND EARLY LUNG DYSFUNCTION

3.1. Endothelial heterogeneity during the transition to air breathing

The discovery that capillary EC heterogeneity arises after birth raises the important question of how it is specified. In the lung, embryonic development is thought to establish necessary cell types, while postnatal alveologenesis forms the structural unit of gas exchange after the transition to air breathing (54). However, much remains to be learned about the role of ECs in these processes and how EC dysfunction contributes to developmental disease. The transition from in utero to the external environment requires molecular, structural, and functional changes as the lungs move from a fluid-filled to a gas-filled state (55). In-depth analysis of mouse development using scRNA-seq emphasizes the dramatic transcriptomic changes occurring at birth (55, 56). At mouse embryonic day (E)12.5, ECs are divided between Aplnr⁺/Pecam1⁺ vascular progenitors and Prox1⁺/Ccl21a⁺ lymphatic progenitors (54). Venous and arterial ECs are specified from vascular progenitors before the transition to air breathing and are visible as clusters with distinct transcriptional identity by E15.5 (54). A population of proliferating ECs arises by E17.5, is maintained through birth to P7, and decreases after secondary septation, disappearing by P15 (54). True CAP1 and CAP2 fates are not established until late embryonic and early postnatal stages (54), again suggesting that these specialized EC subtypes are either necessary for, or perhaps defined by, lung maturation and air breathing.

Another scRNA-seq atlas following the trajectory of mouse lung development reported establishment of venous and arterial fates as early as E12, identifying Car4-expressing CAP2s at E18 and confirming that adult CAP1 and CAP2 markers increase in expression in postnatal lung (57). This work also identifies a proliferative subgroup of CAP1 ECs, indicating that they can self-renew; trajectory analysis suggests that a small population of CAP1s have an increased probability of assuming a CAP2 fate (57). scVelo, an mRNA-splicing-based quantification of transcriptional dynamics, shows that CAP2 ECs mature along the developmental trajectory as defined by their increased expression of Kdr (VEGFR2) (57). Expression of cognate ligand and receptor transcripts in CAP2 ECs and AT1 cells suggests that they interact to establish the alveolar basement membrane (57), though this requires further functional testing.

scRNA-seq of isolated mouse lung ECs has also identified dramatic changes in EC gene expression during the transition to air breathing and alveologenesis, with the greatest transcriptomic shift occurring in CAP1s (58, 59). CAP1s initially express Sparcl1, Mest, and Depp1 prenatally (E18.5) but shift to an early postnatal phenotype at P1 and P7 characterized by expression of genes involved in cytoskeletal organization, cell motility, proliferation, and angiogenesis (58). Early CAP1s also express high levels of Peg3, an imprinted gene that may be important for their self-renewal (58). By P21, CAP1s express fewer transcripts overall, which may represent a shift from development towards quiescence (58). Surprisingly, two populations of Gja5⁺/Bmx⁺ arterial ECs are also identified during development: an “immature” population expressing genes involved in vascular development and angiogenesis and a Dkk2⁺ population proposed to be more mature (58). The immature population shares gene expression similarities with microvascular ECs and is not found at P7 (58), suggesting that this transient population may contribute to artery development. Two arterial EC populations are also present in an scRNA-seq study of sorted CD31⁺/CXCL12⁺ ECs (59). One population expresses high levels of Gkn3 and genes involved in extracellular matrix and structural organization, while the other expresses high levels of Alox12 and genes involved in vascular development (59). Spatial analysis of these ECs in the developing arterial tree reveals that the Alox12⁺ population is located more distally, while the Gkn3⁺ population is found in proximal vessels and expresses genes suggesting a role in endothelial-mediated assembly of vessel walls (59). Loss of Cxcl12 expression causes transcriptional changes in the endothelium as well as in other lung cell compartments (59). Although these studies differ in their characterization of arterial EC subtypes, focusing on an axis of maturity (58) or on proximal-distal patterning (59), EC heterogeneity in the developing lung clearly extends past the capillaries and into larger vessels.

3.2. The endothelium in human lung development

Studying the developing human lung is challenging due in part to the difficulty of acquiring tissue samples. Nevertheless, recent work profiling the embryonic and fetal human lung from 5 to 22 weeks post-conception has provided key insights into early human lung development (60). This high-resolution cell atlas contains data from lung bud formation (5-6 weeks), mid-development (9-11 weeks), and later development (15-22 weeks) (60). At early stages, vascular ECs expressing THY1 and CD24 can already be differentiated from early lymphatic ECs expressing PROX1 and STAB1. In a striking difference from the mouse, arterial ECs expressing GRIA2 and GJA5 can also be distinguished at the stage of lung budding; no venous markers are expressed, indicating that arteries are specified before veins in the human lung (60). Trajectory analysis analyzing EC clusters from multiple stages predicts that middle- and late-stage capillary ECs may generate both artery and vein ECs (60). By 20 weeks, three different artery EC populations with temporal and spatial differences are evident. GJA5/OMD-expressing artery ECs are located more proximally, and GRIA2⁺ “immature” artery cells are still present alongside DKK2⁺ arterial ECs (60). Chromatin accessibility analysis using scATAC-seq demonstrates that the SOX17 binding motif is highly enriched in artery ECs. Later in development, two populations of lymphatic ECs are also distinguishable, with one SCG3-expressing population likely representing lymphatic valve ECs (60). Although venous EC markers are clearly expressed at 20 weeks, pulmonary and systemic venous ECs cannot be distinguished in fetal lungs (60). CAP2 or aCap ECs arise later in human development and are spatially localized to the developing air sacs, similar to the mouse (60). However, these cells are visible by 20-22 weeks post-conception (60), indicating that in humans, they do not arise during the transition to air breathing. As in the mouse, CAP1 expression of PEG3 is associated with the transition to air breathing, with high PEG3 expression at 1 day of life, moderate expression at 6 months, and decreased expression by early adulthood (58). Differences between human and mouse lung EC development highlight the need to develop in vitro or ex vivo human models to better study EC interactions with other lung cell types. Although a pioneering method has been developed to culture human fetal organoids, generating mechanistic data (61), these organoids do not yet contain ECs. Vascularized systems will be essential to study and manipulate cell-cell interactions in the lung and to develop and test novel treatments to restore lung function.

3.3. Endothelial contributions to early lung dysfunction

The discovery that pulmonary EC heterogeneity exists during development has led researchers to examine how these cells may contribute to developmental lung diseases or participate in lung repair. These studies have the potential to identify important developmental pathways that could be harnessed to promote regeneration and to identify EC subtypes with important roles in propagating or reversing disease states. Several initial findings have been extended to create nanoparticle-based or cell-based therapies for developmental disorders.

scRNA-seq studies of EC heterogeneity have identified c-Kit⁺ EC progenitors in the developing distal lung (62), both in neonatal mice (58) and in human neonates (63). In mice exposed to neonatal hyperoxia, c-Kit expression is decreased in alveolar ECs; this is regulated by the transcription factor FOXF1 (63). Mutations in the FOXF1 gene are associated with the human disease alveolar capillary dysplasia with misalignment of pulmonary veins (ACDMPV), which causes loss of alveolar capillaries, lung hypoplasia, and respiratory failure. This suggests that maintenance of c-Kit expression through FOXF1 may be critical for establishing the alveolar capillary network, and indeed, deletion of either Foxf1 or c-Kit decreases alveolar progenitors and causes alveolar simplification (63). The regenerative potential of c-Kit⁺ donor ECs has been tested through adoptive transfer into Foxf1-mutant mice or a hyperoxia-induced mouse model of bronchopulmonary dysplasia (BPD), a clinical diagnosis involving abnormal alveolarization and development of the pulmonary vasculature, often in preterm neonates. Unlike c-Kit⁻ ECs, c-Kit⁺ ECs integrate into the peripheral microvasculature and rescue alveolar simplification, suggesting an increase in neonatal lung alveologenesis (63). The c-Kit+ endothelial population is itself heterogeneous, with FOXF1⁺ and FOXF1- subpopulations; only c-KIT⁺/FOXF1⁺ cells engraft into the neonatal lung (64), indicating that the progenitor capacity of pulmonary ECs may be limited to a small EC subset. Despite their small numbers, EC progenitors are clearly essential for proper lung development. In a primate model of chorioamnionitis, loss of endothelial progenitor cells (EPCs) leads to a failure of differentiation along the CAP1-CAP2 trajectory (65). This is likely mediated by IL-1 and TNF signaling; a combined inflammatory blockade rescues changes to alveolar signaling (65), indicating that inflammation can dramatically alter EC behavior and differentiation. In the adult mouse and human lungs, c-Kit is primarily expressed by CAP1 ECs (39, 46), suggesting that at least a subset of these cells maintain their progenitor potential even in quiescence. The progenitor properties of these cells make them an attractive candidate for cell-based therapies for lung disease involving loss of ECs or failure of EC regeneration.

Although the genetic basis of ACDMPV is known, the mechanisms by which FOXF1 deficiency leads to lung defects were less clear, and mouse models recapitulate only some human disease phenotypes. A recent single-cell multiomics study profiled lung tissue from six infants with ACDMPV and compared it with published data from preterm neonatal and control infant lungs to determine changes in transcriptomes, chromatin accessibility, and cell-cell interactions and identify potential therapeutic targets (66). All expected EC subtypes are identified in the ACDMPV data: CAP1, CAP2, systemic vascular ECs, venous ECs, arterial ECs, and lymphatic ECs (66). The most severely affected ACDMPV lungs demonstrate absence of endothelial FOXF1 RNA, and dramatic changes are identified in almost all EC populations (66). CAP1 and CAP2 cells are greatly reduced or absent, and KIT expression is reduced or absent in the remaining CAP1 cells (66), suggesting the loss of EC progenitors. In contrast, the systemic EC population is greatly expanded (66). The proportion of CAP2 ECs in each subject inversely correlates with disease severity, and in the capillary ECs that remain, many signaling pathways are downregulated, including PTEN, ERK, WNT, and STAT3 in CAP1s and FAK and semaphorin in CAP2s. Further, CellChat analysis of scRNA-seq data shows that VEGFA-KDR signaling to CAP2s is absent in ACDMPV compared to normal lungs, while KDR signaling to the systemic vasculature is upregulated (66). These data highlight the critical role of FOXF1 in formation of the developing pulmonary capillary network and the importance of EC heterogeneity for the formation of intact lung structure.

Mice exposed to hyperoxia from birth to postnatal day 7 to model BPD demonstrate both loss of vessel structural complexity and EC transcriptomic differences (58). ECs with an “early” CAP1 fate are increased relative to “late” CAP1 ECs, with upregulation of p53 target genes (58). Loss of p53 during hyperoxia results in expansion of a “transitional” EC population expressing oxidative stress response and growth factor genes (67). After hypoxia, Apln, which is usually expressed in CAP2s (37, 39), becomes upregulated in CAP1s, while the CAP1 marker Aplnr is downregulated in proliferating ECs (58). This indicates that CAP2-CAP1 signaling may become dysregulated after injury. A population of proliferating venous ECs expressing Ddah1 arises after hyperoxia, indicating that EC activation also occurs in large vessels (58). Other studies profiling mouse EC transcriptomes post-hyperoxia have shown the expansion of CAP2 ECs, with upregulation of Apln or Car4 (68, 69). Decreased Icam2 expression and Vegf signaling in these CAP2s suggests that they are immature or undergoing vessel remodeling, potentially with increased CAP1-to-CAP2 transition (58, 68). Increases in Fgfr1 expression are also seen in CAP2 ECs after hyperoxia, and endothelial-specific deletion or inhibition of Fgfr1 protects against hyperoxia-induced lung injury through increased alveologenesis and improved respiratory function (70). Together, these studies confirm that ECs are significantly impacted by increased oxygen exposure and begin to identify subtype-specific mechanisms and increased heterogeneity that may contribute to endothelial dysfunction in BPD.

As BPD disproportionately affects male infants, several studies have examined sex-specific differences in hyperoxia mouse models. While female mice exposed to hyperoxia retain pulmonary angiogenesis, male mice have decreased angiogenesis (71, 72). Male and female mice exposed to hyperoxia also show differences in EndMT (73), suggesting that this mechanism may contribute to sex differences in BPD. Investigation of EC subtypes by scRNA-seq after hyperoxia reveals that female CAP2s have increased expression of oxidative phosphorylation genes but decreased expression of oxidative stress response and necroptosis genes (68, 69). Male CAP2s show decreased expression of Klf4, which mediates angiogenesis by regulating Notch signaling; female ECs show increased expression of Ddah1, which has been shown to enhance angiogenesis (69). As our understanding of the role of EC subtypes in developmental disease grows, potentially leading to new therapies, it will be essential to identify sex-specific mechanisms to enhance therapeutic opportunities for all.

4. ENDOTHELIAL HETEROGENEITY IN LUNG DISEASE

The initial identification of pulmonary EC subtypes demonstrated that these cells are altered with injury and suggested that they contribute differently to lung repair (38, 39). EC subtypes have since been profiled in a wide variety of animal disease models and human lung diseases, including viral infection, idiopathic pulmonary fibrosis (IPF), chronic obstructive pulmonary disease (COPD), pulmonary hypertension (PH), and lung cancer.

4.1. Endothelial changes induced by infection and inflammation

Respiratory infections result in a major inflammatory response in the lung. This is necessary for viral clearance but also damages the alveolus, impairs gas exchange, and contributes to acute respiratory distress syndrome (ARDS) (Figure 2). Efforts to target inflammation-induced lung injury and reverse ARDS require a greater knowledge of the mechanisms that contribute to functional or dysplastic lung repair. After viral infection, ECs proliferate to restore damaged vessels (38, 74, 75). In capillaries, progenitor potential is restricted to CAP1s (39); however, several pulmonary EC subtypes respond to inflammatory stimuli or are induced by viral damage. After intranasal infection of mice with PR8 H1N1 influenza A virus (IAV-PR8), one subgroup of CAP1 ECs expressing immune response genes arises and participates in inflammatory signaling, while another subgroup expresses genes associated with vascular development, suggesting a preferential contribution to regeneration (75). In the homeostatic mouse lung, expression of the transcription factor Atf3 defines a small subpopulation of CAP1s that contribute to proliferation and regeneration of the vasculature after IAV-PR8 infection (76). Atf3-expressing CAP1s increase in proportion after IAV-PR8 infection and upregulate genes involved in regulation of angiogenesis, vascular development, and endothelial cell migration (76). Deletion of Atf3 in mouse endothelial cells results in decreased EC proliferation, increased apoptosis, and altered EC signaling (76). This leads to defects in long-term alveolar regeneration, suggesting that regeneration of the vasculature is necessary for repair of alveolar structure after infection. Other EC transcription factors and signaling axes can also promote or impede lung regeneration after viral infection. COUP-TFII is required for EC proliferation and migration to promote lung repair but is suppressed by proinflammatory cytokines (74). TGF- βR2 interacts directly with VEGFA to promote autocrine signaling and recovery of CAP2 ECs (77). Sparcl1 expression promotes a pro-inflammatory macrophage state and increases the severity of lung injury (78). A longitudinal scRNA-seq atlas of lung regeneration after IAV-PR8 infection identifies an injury-induced capillary (iCAP) EC state that arises during the second week after infection and is derived from both CAP1 and CAP2 populations (79). This aberrant cell state persists in the lung up to a year after infection, localizing specifically to regions of densely injured tissue and expressing high levels of Sparcl1 and Ntrk2 (79). Both endothelial heterogeneity and signaling clearly contribute to lung regeneration after influenza infection in the mouse.

Figure 2. Effects of viral infection and inflammation on the pulmonary endothelium.

After viral infection of the healthy lung (A), an influx of immune cells creates inflammation that is necessary for viral clearance. (B) Inflammation is a “double-edged sword,” as it also causes tissue damage. ECs respond to inflammation and structural damage by proliferating, which is required to create new vessels but may also contribute to pathology. Different EC subtypes respond differently to infection, which impacts their gene expression and signaling. Created with biorender.com.

Viral infections have increased risk of mortality in the elderly, and aging alters various cellular compartments in the lung, including the endothelium (80). Comparison of scRNA-seq of young (16-18 weeks) and old (80-82 weeks) C57BL/6 mouse lungs infected with IAV-PR8 demonstrates a decreased wound healing score in aged CAP1 ECs (80). Although Car4⁺ CAP2s show no difference in wound healing score, their localization is altered: aged CAP2s no longer preferentially localize to areas of tissue damage as young CAP2s do (80). The transcriptome of aged CAP2s indicates increased inflammatory response and coagulation, perhaps related to the increased risk of thrombosis after viral infection in the elderly (80). Further studies in human tissue will be required to determine whether activation of these gene expression patterns occurs with increasing age in human respiratory infections such as COVID-19.

Since the start of the COVID-19 pandemic, the extent of vascular involvement in the disease has proven controversial. The current consensus indicates that replicative infection of ECs with SARS-CoV-2 is unlikely, but vascular dysfunction is widely accepted to be a secondary consequence (81). Severe endothelial injury and widespread thrombosis can be observed in COVID-19 patient lungs (82). Compared to influenza patient lungs, COVID-19 patient lungs reportedly have 2.7 times greater growth of new vessels (82). Several studies have examined EC changes with COVID-19, but a consensus has not yet been reached; differences may result from the broad spectrum of disease pathology. COVID-19 patients can exhibit either loss of ECs (83, 84) or an increase in ECs, with differences in EC gene expression compared to uninfected controls (85). Specific EC alterations include upregulation of VEGF and Notch-related genes and an increase in the proportion of capillary ECs, suggesting an increase in angiogenesis (86, 87). Other studies report an increase in the proportion of systemic venous ECs or an increase in expression of the systemic venous marker COL15A1 in capillaries (87). Like IAV-infected mice, human COVID-19 patients show altered EC gene signatures reflective of increased stress response, including increased expression of heat shock proteins, extracellular matrix (ECM) production, and matrix-receptor signaling (76, 87). COVID-19 ECs also demonstrate decreased expression of genes involved in cell-cell adhesion, junction stability, and quiescence, suggesting disrupted barrier maintenance and vessel wall integrity (87). Treatment of human pulmonary microvascular endothelial cells (HMPVECs) with plasma from COVID-19 patients induces similar decreases in EC barrier function (88). Although changes to EC numbers may depend on the extent and localization of viral-induced damage, the evidence clearly shows that the endothelium is severely affected in COVID. In some cases, the changes observed in human patients may be similar to those seen in influenza-infected mice (89). Extending this work to achieve a fuller understanding of how changes to endothelial function and signaling contribute to the progression of respiratory infections and repair of damaged tissue will be essential to determine whether targeting the endothelium represents a viable therapeutic strategy for ARDS.

4.2. Contributions of heterogeneous endothelial cells to pulmonary fibrosis

In pulmonary fibrosis (PF), normal lung alveolar structure is altered by excessive deposition of ECM, leading to loss of surface area for gas exchange and eventual respiratory failure. Fibrosis is characterized by alterations in multiple cell states, including the endothelium, both in mouse models of interstitial lung disease and in human patients. Although two drugs are FDA-approved to treat PF, these only slow disease progression and do not represent a cure. Understanding the complete picture of cellular alterations in PF will therefore be necessary to develop new therapies that can prevent or reverse the fibrotic state.

Mouse models of fibrosis have been widely used to assess cellular changes caused by PF. The most popular model for fibrosis, instillation of bleomycin into the trachea, induces PF-like phenotypes and can be used to evaluate cellular heterogeneity in disease. This includes studies of the role of mesenchymal cells (90-93), epithelial cell dysfunction (94-96), aberrant Krt8⁺ “transitional” epithelial cells (97, 98) and profibrotic macrophages (99). Lung fibrosis is clearly established by the interactions between multiple cell types, including the endothelium (100); this is unsurprising given that organ-specific ECs can drive fibrosis in several tissues, including the lung, heart, and liver (25, 101). One potential EC driver of increased fibrosis is increased vascular permeability, mediated by loss of endothelial S1PR1 signaling, which maintains endothelial junction integrity (102). Increased permeability may cause fibrosis through increases in tissue inflammation, and boosting S1PR1 signaling can partially rescue fibrotic phenotypes induced by bleomycin or viral infection (102, 103). However, how endothelial heterogeneity contributes to this process is not yet fully understood. In rats treated with bleomycin, several EC subtypes unique to injured lungs have been identified, including a population high in Ntrk2 and Cxcl12, while ECs expressing Nos3 and Cav1 are decreased (104). Gene ontology (GO) analysis shows that Ntrk2/Cxcl12-expressing ECs are enriched in genes involved in production of VEGF, regulation of angiogenesis, and ECM binding, suggesting a role for EC-mesenchymal and EC-immune crosstalk in fibrosis (104). A population of Ntrk2-expressing ECs has also been identified in a mouse bleomycin injury model; these cells accumulate with aging in the mouse and are associated with aberrant mechanical signaling through YAP/TAZ (105). ACKR1⁺ venous ECs are also associated with fibrosis and aging (105). Although endothelial signaling and heterogeneity clearly contribute to fibrosis, further functional studies will be required to fully understand the role of heterogeneous EC subtypes in PF, especially as a single dose of bleomycin is insufficient to induce the progressive fibrosis seen in human patients. The presence of Ntrk2-expressing ECs in mice from both single-dose bleomycin and influenza infection models suggests that induction of Ntrk2 expression may be a common characteristic of inflammatory injury (79, 105). Repetitive or chronic bleomycin injury induces greater fibrosis in the mouse and may better mimic the human disease, which is thought to develop over time through a cycle of repeated injury and incomplete repair (106). Cellular changes identified in the mouse must also be examined in a human context.

Several different groups have created transcriptomic atlases of human pulmonary fibrosis (48, 107-109). Many of these datasets are now compiled into the Idiopathic Pulmonary Fibrosis (IPF) Cell Atlas (110). Many different aberrant cell populations have been identified in studies of human IPF, including fibroblasts (111), abnormal “basaloid” epithelial cells (48, 107, 112), and profibrotic macrophages (109, 113). However, evidence continues to emerge indicating that the endothelium also plays an important role (114). An expansion of ECs expressing COL15A1, a gene associated with the systemic circulation in healthy human lungs, is seen in IPF (47, 48, 87). This expansion of COL15A1⁺ ECs is not common to all lung diseases, as it is not found in COPD, and in IPF lungs, COL15A1⁺ ECs localize specifically to areas of fibrosis (48). These cells are also a unique feature of human disease, as they have not yet been identified in mouse models of fibrosis, underscoring the importance of profiling cellular alterations in human tissue.

Precision cut lung slices (PCLS) present an opportunity for manipulation of human tissue ex vivo while preserving cell-cell interactions. Recent scRNA-seq analysis of PCLS treated with a “fibrotic cocktail” of cytokines reveals that several PF-like states can be recapitulated ex vivo (115). This includes a VWA1+/PLVAP+ EC state similar to the COL15A1⁺ EC state seen in vivo, albeit without high COL15A1 expression (115). Trajectory analysis indicates that the likely source of these aberrant ECs is a subpopulation of capillary endothelial cells (115). These data provide proof of principle for the use of PCLS in studying human pulmonary fibrosis ex vivo. Although we still have much to learn about the contributions of pulmonary EC subsets to IPF, we may be able to gain mechanistic insights from ex vivo preparations of human lung tissue.

4.3. Endothelial dysfunction in pulmonary hypertension

Pulmonary hypertension (PH) is an incurable condition with a high mortality rate that has been attributed to both artery and capillary EC dysfunction; emerging data indicates involvement of post-capillary pulmonary veins (116, 117). Smooth muscle cells involved in vasoactive responses and providing structural support have also been implicated in PH pathogenesis (118). Changes in EC-smooth muscle communication and immune signaling are also associated with PH, indicating that many factors contribute, and animal models may not capture all aspects of the human disease (116). Recent genomic data has shed light on novel mechanisms of EC and mural cell heterogeneity and signaling in PH pathology.

Particularly in the Pulmonary Arterial Hypertension (PAH) subgroup, many view PH as resulting from EC dysfunction, and PH ECs have an altered transcriptional state compared to non-PH ECs (117, 119-121). Transcriptional and functional changes to pulmonary artery ECs (PAECs) have been assessed after isolation of these cells from living PAH patients using right heart catheter balloon tips (117). Downregulated genes in PAH PAECs include genes associated with the Hop pathway in cardiac development, ALK in cardiac myocytes, leukocyte trans-endothelial migration, and the Fanconi anemia pathway (117). In addition, individual or small groups of genes relating to BMP and Wnt signaling, cancer pathogenesis, fatty acid oxidation, and glycolysis are differentially expressed (117). In several individual study participants, deregulation of anoikis renders PAH patient cells replication-competent in unfavorable culture conditions. If PAH PAECs are hyperproliferative or resistant to cell death, this may play a role in vascular remodeling and disease progression (117, 121). Several essential signaling pathways are dysregulated in PH ECs, including TNF and BMPR2 signaling (119, 120). Interestingly, gene expression signatures of cells from Group 1 (PAH) and Groups 2-5 (PH) patients overlap, suggesting that mechanisms of EC dysfunction may be common across subtypes (117). Changes in EC state, including changes to essential signaling pathways and cellular functions, are evidently one important characteristic of PH pathology.

Smooth muscle cells (SMCs) are also thought to drive vascular remodeling in PH, with high levels of ACTA2-expressing cells contributing to vascular wall thickening and reduction in vascular lumens (118). PCLS cultures reveal expansion of SMC coverage and an increase in immune cells in PAH lungs, while scRNA-seq of PAH and control donor lung samples reveals four distinct SMC populations: oxygen sensing, contractile, synthetic, and fibroblast-like (118). Pulmonary vascular remodeling in PAH patients is associated with a shift in proportion and localization of SMC subtypes, with an increase in VCAN⁺ synthetic SMCs found in the neointimal region (118). Trajectory analysis reveals a shift towards the synthetic fate in PAH patients. Further, PAH patient SMCs demonstrate dysregulated communication with immune cells, which likely disturbs vascular homeostasis (118). The shift from contractile towards synthetic phenotypes and the overall dysregulation in signaling suggests that altered SMC state can also contribute significantly to PH pathology; however, the effects of these changes on EC heterogeneity remain unknown.

In addition to cell-intrinsic changes to ECs and SMCs, changes in EC-SMC communication can contribute to the PH disease state. In lymphangioleiomyomatosis (LAM), a rare cystic lung disease caused by tuberous sclerosis complex (TSC) mutations activating mTORC1 in pulmonary mesenchymal cells, a subset of patients present with PH (122). The mTOR pathway has been implicated in vascular remodeling in PH, but as LAM patients have mesenchymal-specific mTOR activation, the triggers for EC dysfunction and PH in these patients remained unclear. In coculture studies, LAM fibroblasts induce increased proliferation and self-organization of LAM ECs, including formation of EC clusters in 2D and endothelial tubes in 3D (122). scRNA-seq of LAM lung tissue shows increased WNT2 expression in several mesenchymal cell subtypes, and treatment of control EC cultures with WNT activators recapitulates the LAM EC phenotype, suggesting a causative role for WNT2 in EC dysfunction (122). ECs from a mouse model for LAM with lung mesenchymal-specific deletion of Tsc2 demonstrate upregulation of genes involved in growth, metabolism, and angiogenesis, including several Wnt receptors (122). scRNA-seq of year-old control and LAM mice shows that arterial ECs are the most transcriptionally altered in the LAM model, and functional studies show significant vascular remodeling, with thicker medial vessel walls and increased right ventricular systolic pressure (122). Identification of a mesenchymal-endothelial signaling axis in LAM leading to PH in a mouse model suggests that altered mesenchymal-EC signaling may be a key player in PH pathology. Further work is needed to assess the roles of specific EC subtypes in human and mouse PH.

Several endothelial-targeted therapies have been suggested for treatment of PH. Decreased levels of Sirt7 correlate with increased PH pathology in the mouse, providing rationale for using endothelial-specific Sirt7 delivery as a possible therapeutic (123). Sirt7 therapy in mouse models ameliorates PH by suppressing vascular structural damage, senescence, and inflammation (123). Treatment with the NAD+ intermediate NR also improves PH phenotype in mouse models (123). The small GTPase RAB7, which maintains cellular health, is downregulated in patients with PH, suggesting another possible therapeutic target. Activating Rab7 expression with ML-098 in rats with chronic hypoxia reduces established PH or prevents its initiation (124). The effects of therapies targeting a single gene on EC and SMC heterogeneity and signaling remain unknown, and these therapies have not yet been translated into human patients. As PH patients have limited treatment options, our increased understanding of disease mechanisms will be invaluable in the quest to design novel therapeutics.

4.4. COPD as an endothelial disease

COPD is a chronic inflammatory pulmonary disease that presents with airflow obstruction and affects approximately 12% of the world population, with case and death rates increasing (125). Patients with COPD have significantly impaired endothelial function that correlates with the severity of airflow obstruction and pulmonary lesions (125, 126). COPD patients also demonstrate increased EC apoptosis in small vessels of the lungs (127). Notably, EC apoptosis has also been connected to emphysema in the mouse, where increased capillary EC apoptosis due to loss of endothelial Atf3 results in an emphysema-like phenotype (76).

EC dysfunction in COPD could represent a primary cause of disease or a secondary consequence; this distinction is essential to evaluate the potential of EC-targeted therapies (128). RNA-seq of whole-lung homogenates from patients with severe COPD and emphysema, mild COPD, smoker controls, and nonsmoker controls demonstrates a depletion of markers associated with healthy endothelium in COPD, including VEGFR2 and VEGFA, and upregulation of LRG1, which has been associated with pathogenic angiogenesis (128). There is a direct correlation between decreased EC marker expression and reduction in lung function measured by FEV1 and DLCO (128). In contrast, epithelial gene expression is more consistent in COPD and control tissues, suggesting that dysfunctional endothelium may drive disease. An elastase model of emphysema demonstrates EC transcriptional changes in pathways of blood vessel development, angiogenesis, TGF-B signaling, and extracellular matrix signaling at 7, 14, and 21 days post injury (128). Distinguishing early and late injury, elastase-treated mouse ECs possess a pro-apoptotic gene signature at day 7 and a pro-inflammatory signature at days 14 and 21. LRG1 knockout protects elastase-injured mice from severe parenchymal destruction, identifying a potential target for future therapies (128). Intravenous delivery of healthy ECs also reverses the emphysema phenotype, with significantly less parenchymal destruction and an increase in lung function, accompanied by an increase in EC proliferation (128). Although EC heterogeneity is not examined in this study, previous work using the elastase model indicates an increase in CAP1 EC proliferation as early as day 3 post-elastase, with proliferating CAP1 ECs proposed to generate CAP2 ECs during repair (39). This work suggests that future therapies for COPD may need to involve treatment of EC dysfunction.

4.5. Endothelial heterogeneity in lung cancer

Lung cancer is the second most prevalent type of cancer and causes the largest number of cancer deaths worldwide (129). ECs are one of many factors that play a role in tumor growth, cancer cell proliferation, and ultimately, metastasis. ECs form the tumor vasculature, an essential source of nutrients needed for tumor growth and survival (129). As tumors rapidly grow, oxygen depletion in the microenvironment leads to the activation of HIF-1 and VEGF to facilitate angiogenesis, a process often described as an “angiogenic switch” (130). New vessels form with gaps and pores that allow cancer cells to enter the blood flow and spread to other areas of the body (130). Although the role of the vasculature as a “highway” for dissemination of cancer cells is still being explored, there is some evidence that EC angiocrine signaling can drive this process to promote metastasis, and it is well-known that the tumor cell niche can either promote proliferation or suppress growth (131). The importance of angiogenesis in cancer led to the development of anti-angiogenic therapies, or VEGF-targeted therapies, but these therapies are hampered by the development of resistance (132, 133). Resistance can be intrinsic to cancer cells, resulting in re-emergence of the tumor vasculature when therapies are stopped, or acquired through a switch from sprouting to intussusceptive angiogenesis or a switch in angiogenic signaling from VEGF to another pathway, including EGF, bFGF, PIGF or G-CSF. Tumors can also co-opt existing vasculature if angiogenesis is prevented (133). Furthermore, the tumor vasculature is inherently heterogeneous, composed of ECs from existing vasculature, tumor cells, and transdifferentiated myeloid or mesenchymal cells, and this heterogeneity can underlie the development of resistance through dynamic changes that occur during therapy (129, 130). Single-cell transcriptomics approaches have examined the role of EC heterogeneity in lung cancer to determine whether this can be exploited to improve targeted therapies.

During the process of tumor angiogenesis, several studies show that cancer-associated or tumor ECs (TECs) are heterogeneous and have different functional roles (132). Tumor angiogenesis begins with a leading “tip” TEC, while proliferating “stalk” TECs lengthen vessels. scRNA-seq of TECs suggests that tip TECs express genes associated with EC migration and matrix remodeling, with VEGF signaling only a minority of their function, indicating that other viable therapeutic targets may exist (132). Tip TECs also show restricted expression of placental growth factor (132). One TEC subpopulation with similar characteristics to tip cells shows upregulated genes involved in the maturation of newly formed vessels, vessel barrier integrity and Notch signaling (132), indicating potential roles in tumor growth and malignancy.

Characterizing the response of existing heterogeneous ECs to metastasizing tumor cells shows that although CAP2s remain stable, CAP1 ECs upregulate metabolic and ribosomal genes, indicating that their interaction with cancer cells may “activate” them to induce enhanced biosynthesis (131). CAP1s also upregulate EC-derived cytokines, or “angiokines,” which may communicate with tumor cells to amplify tumor-derived signals (131). Depletion of EC-specific Wnt signaling in mice through EC-specific knockout of Wls increases metastasis in breast cancer and melanoma models; however, the Wnt signal does not differ across EC subtypes, and tumor cells do not preferentially interact with either CAP1 or CAP2 ECs (131). Also suggesting a role for Wnt signaling in TECs, EC-specific nanoparticle delivery of the Wnt receptor FZD4 to augment Wnt/β-catenin signaling inhibits disease progression in a mouse model for non-small cell lung cancer (134). Although much remains to uncover regarding endothelial-derived signals to tumor cells and the role of EC heterogeneity in this process, these studies indicate that changes to endothelial heterogeneity and signaling during cancer progression may play critical roles in metastasis. In the future, harnessing an increased understanding of EC heterogeneity will allow researchers to design better targeted therapeutics for patients affected by cancer.

5. TOWARDS A FUTURE OF ENDOTHELIAL-TARGETED AND CELL-BASED THERAPIES FOR LUNG REGENERATION

5.1. Endothelial-targeted nanomedicine

The necessity of improved therapies for patients with acute and chronic lung diseases, including ARDS, viral pneumonia, PH, COPD, and IPF, has led to many attempts to develop new treatment paradigms. Many drugs have been unsuccessful in reducing mortality in clinical trials, partially due to pulmonary physiology and the difficulty of delivery to the lungs (135). Advances in nanomedicine have the potential to address this problem, as nanocarriers can be modified to target specific areas and can accumulate in inflammatory lung sites, boosting drug potency and reducing side effects (135). Various nanocarriers can be used, including liposomes, polymer micelles, inorganic nanocarriers, and extracellular vesicles; the bioactive compounds composing the nanoparticle delivery system help to avoid enzyme degradation and do not target healthy organs (136). ECs represent a promising target for nanoparticle delivery as uptake can be controlled by blood flow, and unlike the airways, there is an absence of inhibitory mucus (137). Several recent studies have applied nanomedicine to target the endothelium in lung disease, including inflammatory injury and ARDS. The primary inducers of ARDS include pneumonia and sepsis, and its current treatments include medication and ventilation (135). However, although medications can decrease inflammatory symptoms, none currently treat the causes of ARDS; ventilators are problematic due to the possibility of ventilator-induced lung injury or shortages in the event of an epidemic. The advantages of nanomedicine to overcome these challenges include the ability to improve the bio-pharmacokinetics of drugs and to confer anti-inflammatory, antioxidant, or antiviral activity; however, challenges remain, including choosing the size of nanoparticles to deliver. Larger particles attach better to blood vessel walls and accumulate to a greater extent in the lungs (138), while bioimaging suggests that smaller particles have more intensity in the lungs (136). Although nanoparticle delivery of drug candidates specific for ARDS have entered clinical trials, none have reduced mortality (135), indicating that this avenue requires further optimization.

Localization of nanoparticles to the lungs requires avoiding their degradation by the liver, which can be achieved by using red blood cells (RBCs) as carriers; this method shows no induction of inflammatory markers (139). In human ex vivo models of the pulmonary artery, infusion of I-125 labeled nanocarriers (NCs) attached to RBCs increases nanocarrier delivery to the lungs compared to infusion of free I-131 NCs (139). In a mouse model of LPS-induced ARDS, nanoparticles delivered to naive mice are taken up by ECs, but those delivered to LPS-challenged mice are taken up by intravascular leukocytes (139). This propensity for immune-cell uptake during inflammatory disease can be harnessed to treat viral pneumonia through nanoparticle targeting of antibody-conjugated branched lipid nanoparticles (LNPs) to lung macrophages (140). Delivery of siRNA against TAK1, a target kinase that promotes inflammation, reduces inflammation caused by macrophage-released cytokines (140). However, it is also important to improve targeting of ECs over inflammatory cells during inflammatory injury. To this end, dual affinity to RBCs and target cells (DART) nanocarriers, in which nanocarriers are conjugated onto two affinity ligands, one that binds RBCs and one that binds the target cell, result in better selectivity of ECs in target organs using both anti-ICAM and anti-PECAM antibodies (141). In addition, a lung endothelial-specific LNP preparation to deliver Vegfa mRNA to lung ECs can rescue regeneration caused by TGF-βR2 deficiency (77). Although no studies have yet been performed that target nanocarriers specifically to EC subtypes or that manipulate EC plasticity, achieving these goals would allow nanoparticle targeting of ECs to become even more specific in targeting cells affected by disease.

Nanomedicine has also been used to treat PH associated with ACDMPV and the S52F mutation in FOXF1 (142). Patients with mutations in FOXF1 have decreased EC expression of STAT3, which is key for EC proliferation, and nanoparticle delivery of STAT3 cDNA protects Foxf1-mutant mice from right ventricular hypertrophy and pulmonary hypertension by activating EC proliferation and promoting lung angiogenesis (142). BMP9/ACVRL1/SMAD1 signaling is also implicated in ACDMPV; nanoparticles silencing ACVRL1 recapitulate the FOXF1 mutant phenotype, while IV treatment with BMP9 partially rescues alveolar simplification in Foxf1-mutant mice (143). To improve the clinical translation of this therapy, intravenous administration of PEI-PEG nanoparticles to deliver FOXF1 has also been tested in mice (144). Although this approach will require further testing prior to application, nanoparticle targeting could eventually be used for newborns and infants with FOXF1 mutations. In the future, specific delivery to different EC subtypes could aid in understanding the roles of EC states in disease and targeting them to improve patient outcomes.

5.2. Pluripotent stem cell modeling of lung-specific endothelium and cell-based therapies for lung disease

As ECs lining the vascular tree are essential for organ function, and it is clear that organ-specific ECs have distinct gene expression and function (1), it is of great interest to be able to generate organ-specific ECs in culture (145). Our recent increased knowledge of pulmonary EC heterogeneity indicates that it is essential to generate not just lung ECs, but cells representing the full spectrum of EC subtypes. Recent work co-differentiating mesoderm and endoderm from pluripotent stem cells (PSCs) to form vascularized lung and intestinal organoids shows enhanced EC heterogeneity, including the spontaneous emergence of lymphatic ECs (146). Introduction of angiogenic factors and transplantation into the mouse kidney capsule permits further maturation of these organoids, and at three months post-transplantation, FENDRR⁺ CAP2-like cells and KIT⁺ CAP1-like cells are observed (146). This provides initial evidence that lung EC subtypes can be generated from PSCs. PSC-derived epithelial cells are currently being used to pioneer cell-based therapies, including transplantation into airways and alveoli (147, 148). On the endothelial side of the gas exchange interface, lung EC subtypes derived from PSCs could be used for in vitro human disease modeling to complement mouse studies, cell-based therapies, or vascularizing engineered organs for transplant.

Cell-based therapies have the potential to improve lung repair, as many lung diseases demonstrate loss of cells or decreased cell function, and administration of functional cells may result in structural regeneration of damaged tissue (149). Further, as many complex diseases involve more than one cell type, the increased potency of progenitor cells could make them a better alternative to traditional therapies, particularly given the success of PSC co-differentiation (146). Primary EPCs have been tested for their therapeutic potential in PAH; transfection with endothelial nitric oxide synthase (eNOS) enhances EPC ability to reverse hemodynamic and remodeling abnormalities in a PAH mouse model, leading to a phase 1 clinical study known as the PHACeT trial (150). In human patients, EPCs have been selected through differential culture of mononuclear cells isolated by apheresis, followed by eNOS transfection and administration. The study reported modest but short-term improvements in pulmonary arterial pressure and TPR, accompanied by increases in inflammation (150). Intravenous delivery of primary lung ECs reverss the emphysema-like phenotype in an elastase injury mouse model (128), while transplantation of gene-edited EPCs into a mouse model of ACDMPV stimulates angiogenesis, improved oxygenation, and prevents alveolar simplification (64). Mouse embryonic stem cell-derived c-KIT⁺/FOXF1⁺ EPCs increase angiogenesis and prevent alveolar simplification in a hyperoxia-induced mouse model of BPD (151). Although primary and PSC-derived ECs can be used to enhance lung repair in mouse models, therefore, work remains to be done before these therapies can be successfully translated into humans. A significant challenge also remains in successfully capturing the range of EC heterogeneity present in the lung (145, 152). The advent of transcriptomic analysis of organ-specific ECs, ongoing studies of their subtypes and functionalities, and initial evidence for generation of PSC-derived CAP2s (146) all indicate that the promise of PSC-derived lung ECs as an ex vivo model and therapeutic avenue is closer than ever to being realized.

6. CONCLUSIONS AND PERSPECTIVES

The discovery of pulmonary endothelial cell heterogeneity both within capillaries and large vessels has revolutionized our understanding of the complex interplay between cell types in the lung during development, adult homeostasis, and disease. The many studies highlighted here emphasize the essential role of the pulmonary endothelium not only in gas exchange, but also in a wide variety of signaling interactions within the niche. Our understanding of the nuances in cellular state can only improve our understanding of these critical interactions to building the lung structure, maintaining it, and repairing it when it is damaged. As we continue to learn more about EC subtypes and their differing gene expression profiles, functional roles, and disease states, we will open the door for better disease modeling in mouse and human tissue- and cell-based systems, increased development of targeted therapies, and improved vascularization of ex vivo organoid systems and engineered tissues for transplant.