The cardiovascular system comprises the heart and a network of large elastic arteries, resistance arteries, arterioles, and terminal capillaries, connected to the venous return. Each arterial segment consists of three concentric layers: (i) the intima, lined with endothelial cells (ECs) that directly interface with the vessel lumen; (ii) the tunica media, which is made up of at least one layer of vascular smooth muscle cells (VSMCs); and (iii) the adventitia, which contains mesenchymal cells, pericytes, immune cells, and nerve terminals. In larger arteries, this outer layer includes a specialized vascular supply system termed “vasa vasorum”. In contrast to arteries and arterioles, capillaries do not possess a VSMC sheet. They are composed of continuous tubes of ECs, covered by a basement membrane and pericytes. These perivascular cells physically and functionally interact with the capillary endothelium through adhesion plaques, gap junctions, and peg-and-socket structures.

Compared to their diameter, arterioles contain a higher proportion of VSMCs and less elastin than large arteries, which accounts for their role as major regulators of vascular resistance and regional blood flow. The tiny capillaries deliver nutrients and oxygen to cells and remove waste products, while also contributing to blood flow modulation through the pericytes' ability to contract and relax. The endothelium acts as a master regulator of vasorelaxation, sensing and transducing chemical stimuli, such as acetylcholine (Ach) and bradykinin (BK), or mechanical forces, such as fluid shear stress from within the artery, into intracellular signalling. The process involves the secretion of diffusible transmitters by ECs, specifically nitric oxide (NO), prostacyclin (PGI₂), and the so-called hyperpolarizing factors, which interact to reduce cytosolic Ca²⁺ levels in VSMCs, allowing relaxation (1). Intercellular gap junctions facilitate bidirectional relaxation currents between small arterioles and upstream arteries, greatly enhancing tissue perfusion (2, 3).

The three endothelium-derived relaxing factors trigger intracellular Ca²⁺ changes using distinct second messengers. (i)

The endothelium-specific NO synthase (eNOS) is responsible for approximately 90% of NO generation from the L-arginine substrate. The signalling from NO involves the activation of soluble guanylate cyclase (sGC), which catalyzes the conversion of guanosine triphosphate (GTP) into inorganic pyrophosphate and the secondary messenger cyclic guanosine monophosphate (cGMP). CGMP activates protein kinase G, which reduces the cytosolic Ca²⁺ levels by promoting its uptake by sarcoplasmic reticulum calcium-ATPases (SERCA) (4, 5).

Over the past twenty years, the concept of vascular cell heterogeneity has revolutionised the field of vascular biology (16). Vascular cells exhibit differences among various organs and tissues, as indicated by variations in overall morphology, intracellular organisation, and gene expression profiles. Furthermore, significant differences in functional characteristics distinguish cells of large vessels from those of the microvasculature, likely due to epigenetic programming, as these differences persist even after isolation and in vitro culture (17, 18). The heterogeneities and the redundancy of vasorelaxant processes are essential for providing the necessary flexibility to perform various physiological duties. Classical research using isolated vascular models and pharmacological inhibitors has helped decipher regional variations in the expression and function of vasorelaxation-related mechanisms. Analyses using single-cell RNA transcriptomics have demonstrated angiotypic and organotypic parallels and diversities that could not be discriminated with classical pharmacological approaches (19, 20). It would be of paramount importance to apply the same approach to decipher vascular vasorelaxant transcriptomic profiles. Filling this knowledge gap will yield substantial clinical and therapeutic consequences. For instance, it may facilitate the development of novel pharmaceuticals that specifically target particular artery beds and organs, thereby reducing off-target effects.

This study conducted a qualitative transcriptomic analysis of key vasodilator pathways in vascular cell subtypes at both the organismal and regional levels, aiming to inform and establish a theoretical basis for future quantitative research. Our working hypothesis was that the expression of “sentinel” gene transcripts would provide a first insight into the angiotypic and organotypic heterogeneity of vasorelaxation. We extracted and analyzed single-cell RNA sequencing data on enzymes, receptors, and channels involved in vasorelaxant pathways from a publicly available Human Vascular Atlas (20). This database provides detailed molecular profiles of endothelial and mural cells, offering a framework for investigating physiological adaptations across various vascular beds in different organs and tissues (20). Results indicate previously unforeseen disparities between vascular cells from arterioles and capillaries or different organs, with further changes being imposed by aging.

The Human Vascular Atlas is a comprehensive collection of vascular cells from 19 organs and tissues. It integrates arterial, capillary, venous, spleen pulp (littoral), and lymphatic ECs, pericytes, and VSMCs, yielding approximately 67,000 cells from 62 male and female donors. As described in detail previously (20), datasets, other than Tabula Sapiens data, were mapped using CellRanger (version 3.0) and STARsolo (version 2.7.3a) to the human genome (GRCh38 version 3.0.0), with ambient RNA removed using CellBender (version 0.2.0). Pre-processed Tabula Sapiens data were used. Python (version 3), AnnData (versions 0.8.0 and 0.9.1) Pandas (versions 1.5.2 and 1.5.3), NumPy (versions 1.21.6 and 1.23.5), SciPy (versions 1.8.0 and 1.9.3), Matplotlib (version 3.5.2), Seaborn (version 0.12.2) and Scanpy (versions 1.8.2 and 1.9.1) were used for QC and downstream processing. Doublets were identified utilizing Scrublet (version 0.2.3), with cells exhibiting a Scrublet score ≥ 0.4 and an adjusted Benjamini–Hochberg-corrected P ≤ 0.7 classified as doublets and subsequently excluded. Cells of inferior quality were excluded based on the following criteria: minimum read count = 500, minimum gene count = 300, mitochondrial gene percentage ≤ 0.4, ribosomal gene percentage ≤ 0.3. The data were subsequently standardized to 104 and log-transformed [ln(x + 1)] utilizing the SCANPY workflow115. Untransformed reads were preserved in the layer designated as “counts”.

We employed the ShinyCell R package (v2.1.0) to interrogate the associated single-cell RNA-sequencing data to visualize cell metadata and gene expression. The fully interactive Shiny application enables the generation of low-dimensional uniform manifold approximation and projection (UMAP) embeddings, as well as the simultaneous investigation of expression patterns of multiple genes through bubble plots and heatmaps. Briefly, the “Vascular Cell Atlas” dataset (.h5ad, AnnData file) was downloaded in.h5ad format (AnnData file) from https://www.vascularcellatlas.org/ and converted into a Seurat object using the MuDataSeurat package (v 0.0.0.9000). Finally, using the ShinyCell R package, a ShinyCell application was generated from the Seurat object and run via RStudio for data analysis. ShinyCell generation was performed using R (version 4.4.3) and Seurat (version 5.2.1).

Considering the exploratory nature of our study, along with the limited number of cells and donors, we adopted a descriptive approach to reporting the results, without performing quantitative statistical comparisons.

This study examined single-cell transcriptomics of essential genes linked to vasorelaxation. Results indicate a considerable diversity at both the organismal and tissue levels, with age also contributing to variability.

Recent advancements in single-cell genomics and biological network analysis have significantly enhanced our understanding of the variety, flexibility, and modularity of vascular cells (26). These emerging technologies could also help to understand whether the transcriptional profile of vascular cells is tailored to specific organs and tissues or adheres to a more generalized angiotypic program. In a recent groundbreaking study, Barnett and colleagues reported that the global transcriptomic profile of a wide range of human vascular cells primarily corresponds to an angiotypic identity (20). These data contradict previous findings about mouse vascular endothelium, where an organotypic identity reportedly predominated (27). A targeted analysis of specific functions could help resolve the dispute. In this context, we examined the single-cell transcriptomics of key genes related to vasorelaxation in arterial and capillary ECs and arterial VSMCs at both the organismal- and tissue-specific levels. In addition, the expression of the same genes was analyzed in venous, lymphatic, and littoral ECs and pericytes.

Three separate yet interactive processes regulate vasorelaxation: the NO-cGMP pathway, the PGI₂-cAMP pathway, and the hyperpolarization-K+ channels pathway. Cellular heterogeneity and modular redundancy collaborate in enhancing the functional flexibility and adaptability of the vascular system from large arteries to arterioles and capillaries (28). Accordingly, the proportional expression of the three pathways is anticipated to differ along the vascular tree. NO, a crucial signaling molecule, is synthesized by three NOS isoforms, all of which employ L-arginine and molecular oxygen as substrates (29). NOS1 is pivotal in central and peripheral neurons. Its roles encompass the regulation of synaptic plasticity, central control of blood pressure, and vasodilation through peripheral nitrergic neurons. Inducible NOS2 can be expressed in several cell types in response to lipopolysaccharide, cytokines, or other stimuli. It produces substantial amounts of NO, exerting cytostatic effects on parasite target cells. NOS3 is predominantly present in the endothelium. It maintains vasodilation, regulates blood pressure, and possesses several additional vasoprotective effects. We verified the dominant expression of NOS3 in ECs identified by the co-expression of PECAM1 and EGFL7. All other isoforms were not expressed (less than 1% of the cell fraction) except for NOS1 in arterial ECs (7% positive cells) and NOS2 in lymphatic ECS (3% positive cells) (data not shown). Furthermore, the results reveal a remarkable angiotypic variability in NOS3 expression. Littoral ECs exhibited the most abundant expression of NOS3 relative to other EC subtypes. This finding aligns with the fundamental roles of littoral ECs, which filter for antigens, pathogens, and senescent red blood cells, utilizing NO as a permeabilizing and cytotoxic agent (30, 31). The observed abundance of NOS3 transcripts in arterial ECs compared with capillary ECs corresponds with prior findings indicating that this enzyme is predominantly confined to the endothelium of medium to large arterial blood arteries (21). A notable percentage of PECAM1- or EGFL7-positive ECs, especially those from the capillary region, were negative for NOS3. Additional research is required to ascertain the transcriptional repressor limiting NOS3 expression, and to understand whether NOS3-negative cells can be transcriptionally recruited to augment NO synthesis in response to environmental cues, along with gene expression upregulation in constitutively NOS3-positive ECs.

The diverse properties of ECs were corroborated in relation to PTGIS, which had the highest frequency in arterial and venous ECs, while exhibiting low abundance in capillary, lymphatic, and littoral ECs. PTGIS enables to differentiate between VSMCs, positive for this enzyme, and pericytes, which are negative. PTGIR was exclusively expressed in mural cells, exhibiting comparable frequencies in VSMCs and pericytes. By engaging its corresponding receptors, PGI₂ exerts cardiovascular actions, including vasodilation and the suppression of VSMC proliferation (32). Moreover, PGI₂ is said to influence pericytes to preserve vascular barrier integrity and capillary perfusion (33).

Regarding endothelial K+ channels, we observed a decrease in KCNN3 and KCNJ2 from arterial to capillary ECs. At the same time, the muscular K+ channels KCNMA1 and KCNA5 exhibited a greater abundance in VSMCs than in capillary pericytes. These data are compatible with the major role of muscular arterioles in regulating vascular resistance, with capillaries playing a subsidiary contribution through the contractile activity of pericytes.

Co-expression and clustering analyses further validated the heterogeneity of arterial and capillary ECs. The data indicated that, within each EC subgroup, the most prevalent genes—NOS3, PTGIS, and KCNN3—were infrequently co-expressed, with over 50% of the cells exhibiting negativity for any of the examined genes. In resistance arteries, elevated endothelial Ca²⁺ levels facilitate vasodilation by activating SKCa/IKCa channels in certain vascular beds and stimulating eNOS in others (34). An interaction between the two pathways can amplify induced vasorelaxation. For instance, endothelial SKCa and IKCa channel activators have been demonstrated to augment agonist-induced membrane hyperpolarization, Ca²⁺ transients, and the de novo NO generation (34). Based on our findings on limited co-expression, it is plausible to hypothesize that the modularity of vasorelaxant processes is governed by the interaction among various ECs rather than within a distinct group of ECs co-expressing all three genes.

Increased age correlates with endothelial dysfunction. The possibility that hyperpolarization-dependent vasodilation may compensate for reduced NO generation is intriguing. However, studies in experimental animal models suggest that a diminished expression and activity of K+ channels could also contribute to the dysfunction of older blood vessels (35).

In our study, NOS3 expression was consistently reduced in various EC types from the older cohort, while PTGIS frequency significantly escalated with age in arterial and venous ECs, and VSMCs. Similarly, KCNJ2, which encodes the Inward Rectifier Potassium Channel 2 and is regarded as a facilitator and maintainer of hyperpolarization, exhibited a significant increase in arterial ECs. The large-conductance KCNMA1 was almost twice as frequent in VSMCs of the older group compared to the younger group.

The transcriptome analysis revealed similarities and disparities across vascular cells from the brain, heart, and uterus. NOS3 exhibited a threefold greater frequency in coronary artery ECs than endometrium-derived ECs. Similarly, the brain and cardiac endothelium had a higher frequency of PTGIS, starkly contrasting ECs originating from the uterus. Likewise, significant variations in expression were noted with VSMCs. For instance, VSMCs from cerebral and cardiac arteries had high expression levels of PTGIS, whereas uterine VSMCs were devoid of this gene expression. The PTGIR gene was predominantly expressed in coronary and uterine VSMCs compared to cerebral VSMCs.

There is an increasing interest in the organotypic identity of K+ channels, mainly fuelled by the potential for targeted therapeutic applications. KCNN3 was a common component across the three studied organs, whereas KCNJ2 was found in cerebral and endometrium capillary ECs, albeit more extensively in the former. The big conductance BK channels, composed of pore-forming α subunits (BK-α, encoded by the KCNMA1 gene) and regulatory β1 subunits (BK-β1, encoded by the KCNMB1 gene), are prevalent in coronary arteries and modulate myocardial perfusion by linking intracellular Ca²⁺ homeostasis with excitation-contraction coupling (36–39). The voltage-dependent K+ (KV) subtype 5 channels, encoded by KCNA5, play a crucial role in regulating the precise coupling of coronary blood flow to myocardial metabolism (40). Our findings validated the elevated prevalence of KCNMA1 and KCNA5 gene transcripts in coronary artery VSMCs. This pattern was notably detected in brain VSMCs, but not in uterine VSMCs, which were abundant in KCNA5 but displayed a low frequency of KCNMA1. Evidence suggests that K+ channels play a crucial role in modulating uterine vascular tone and responding to pregnancy and hypoxia (41). Alterations in the expression of these channels have been documented during pregnancy and are regarded as potential therapeutic targets for pregnancy-related disorders (42). Elevated NO production and augmented endothelium-derived hyperpolarizing factor-mediated responses, chiefly via the activation of small and intermediate conductance K+ channels, significantly contribute to endothelium-dependent vasorelaxation of the uterine artery in nonpregnant rats, rising to approximately 70% in pregnant rats (43). Our observations regarding the abundant expression of KCNA5, which encodes the Voltage-Gated Potassium Channel Protein Kv1.5, suggest this channel may have a physiologic role in regulating uterine circulation. Seminal research in non-pregnant sheep showed that the administration of 4-aminopyridine (4-AP), a selective Kv channel blocker, did not affect the baseline uterine blood flow but diminished the estradiol-induced increase in uterine blood flow (44). Dysfunction of Kv channels may contribute to pregnancy complications, as indicated by findings showing hypoxia specifically diminished Kv currents assessed via the patch clamp technique in VSMCs from fetoplacental arteries and triggered vasoconstriction in placental cotyledons (44).

An additional insight favoring organotypic identity is provided by the large enrichment of several genes in ECs and VSMCs from coronary and brain arteries compared with organismal expression levels. For instance, we observed a striking enrichment in NOS3, PTGIS, and PTGIR in coronary artery vascular cells, accompanied by a higher abundance of K+ channels, specifically KCNN3, KCNN4, KCNMA1, and KCNJ2. These data are consistent with the potent vasorelaxant capacity of coronary arteries through multiple, redundant mechanisms. A similar, though lesser enrichment was observed in vascular cells from the brain circulation.

There is a great interest in the possibility that single-cell transcriptomics would provide a means to refine personalized treatments. Based on patients' specific requirements, one can opt for a generalized strategy addressing all primary vasorelaxant mechanisms or focus on certain vascular beds informed by identifying angiotypic and organotypic variability. Given its preliminary nature, we are far from claiming that this study can guide a personalized approach. However, in this section, we would like to present how a better understanding of vascular cells' heterogeneity may provide an alternative to complex multidrug treatments. The recent Lacunar Intervention Trial-2 (LACI-2) shows a global methodology (45). This interventional trial aimed to administer increasing dosages of isosorbide mononitrate, a NO donor, and Cilostazol, a phosphodiesterase type 3 inhibitor that enhances the prostacyclin-cAMP pathway in stroke patients. The findings demonstrate that the treatment diminished the risk of significant vascular incidents and improved functional and cognitive results relative to single-agent therapy, while maintaining a favorable safety profile (45). Rinvecalinase alpha (DM199) is a recombinant variant of the endogenous protein tissue kallikrein, presently undergoing trials in patients with strokes and complicated pregnancies (46). It is a bioactive kinin peptide source that concurrently activates the three principal vasorelaxant systems (47). On the other hand, K+ channel modulators offer a novel and appealing approach to target specific arterial beds selectively. Endothelial SKCa and IKCa activators may be appropriate for treating vascular problems in the brain, heart, and uteroplacental unit. Small-molecule activators of KCa2.x and 3.1 channels, including SKA-31, can rapidly suppress myogenic tone in isolated resistance arteries, promote significant vasodilation in intact vascular systems like the coronary circulation, and promptly reduce systemic blood pressure in vivo (48). At the same time, BKCa agonists may be effective for targeting the brain or heart specifically. Several clinical trials using BKCa openers for cardiovascular conditions, including hypertension, ischemic heart disease, stroke, and erectile dysfunction, have been commenced in recent decades. Nevertheless, the inadequate potency and lack of selectivity of these compounds led to the premature cessation of these trials (49). Only one medication candidate, andolast, targeting BKCa channels, remains in clinical development for the treatment of asthma (50). Voltage-sensitive Transcriptional regulators are also appealing. The nuclear factor erythroid 2-related factor 2 (Nrf2) has been shown to upregulate BKCa expression by directly binding to the antioxidant response element (ARE) motif of the KCNMA1 promoter, suggesting that treatment with FDA-approved Nrf2 activators may be an effective strategy for enhancing the treatment of coronary artery disease (38). Finally, activators of Kv channels could provide a means to induce extended vasorelaxation. We have demonstrated that VSMCs from coronary, cerebral, and uterine arteries express KCNA5, which encodes the Kv1.5 potassium channel. Recent studies have proposed Kv7 channels, encoded by KCNQ 1–5 genes, as promising new therapeutic targets to increase fetoplacental blood flow in preeclampsia (51). Nonetheless, this gene family was lowly expressed in VSMCs from our series, except KCNQ1, which was found to be expressed by more than 10% of brain vascular cells.

Our data reveal considerable heterogeneity in the expression of vasorelaxant pathway components among vascular cells throughout the arterial tree and different organs. K+ channels and the PGI₂-generating enzyme are posited to gain prominence in the ageing vasculature, perhaps offsetting the reduced expression of NOS3. K+ channels are a vital therapeutic target to improve the effectiveness of existing treatments utilizing NO donors and PGI₂ analogues.

The study has several limitations intrinsic to its explorative nature. We analyzed a few “sentinels” of relaxation, which may represent a simplification of the complex transcriptomic scenario. This initial attempt justifies additional research to discern the complexity through the inclusion of other vasorelaxant mediators. Similarly, it is premature to infer the physiological implications of these novel data within the enormous body of knowledge on human vascular tissue and bulk cell preparations. A comprehensive discussion on functional repercussions should be deferred until additional data on vascular cells from many human subjects under physiological and pathological settings are available. This will allow for the understanding of how the modification of gene expression through the recruitment or derecruitment of single cell populations for a specific gene correlates with functional changes. Additional restrictions arise from the source database's cell count, which impeded the incorporation of covariates. Rather than examining expression variations across age percentiles, we analyzed two age groups below and above 50. More ageing research will be attainable as the Atlas data is anticipated to be expanded with additional contributions. Another limitation of single-cell data is that the low overall depth of sequencing per cell means that several “drop-out” genes, which are expressed in cells albeit at low levels, are not apparent in the final data. Single-cell sequencing has lower read depth and capture efficiency compared to bulk RNA sequencing or conventional PCR, making it harder to detect lowly expressed genes (52). Deepening the search from general cell populations into cells from specific organs may rule out the possibility of missing relevant gene transcripts. This was the case of the Kv7 channels, a type of voltage-gated K+ channel that plays a crucial role in regulating neuronal excitability and vascular cells' control of blood flow. Their expression in brain vascular smooth muscle cells and endothelium contributes to regulating vessel tone and maintaining the blood-brain barrier (53, 54). Similarly, Kv7 channels have been detected in coronary arteries of animal models and human chorionic plate arteries (51, 55). An analysis of the general cell population indicated that Kv7 gene transcripts were lowly expressed. However, when looking at specific organs, we found an enrichment of the encoding gene transcripts in brain vascular cells at least for the KCNQ1 gene, while remaining low for KCNQ 2–5.

Finally, transcriptome data should be supplemented and validated by evaluating protein levels, potentially integrated with topographic analysis as shown in spatial transcriptomics and proteomics, and epigenetic mechanisms. Despite these limitations, our work lays the groundwork for formulating innovative hypotheses regarding vascular cell heterogeneity, which could potentially benefit patients. The most intriguing inquiries for future investigation involve the requisite number of additional vascular cells necessary to express the examined gene to observe alterations in vasorelaxation, and whether this objective can be therapeutically attained through transcriptional inducers, either independently or in conjunction with traditional activating ligands.