Atherosclerosis is one of the prominent pathological conditions that lead to myocardial infarction and stroke. A significant risk factor for its development is chronological age (1), accelerated by epigenetic aging (2). Aging is associated with declines in cellular homeostasis causing inflammation (3–6) and organ system dysfunction (7–9), thereby resulting in the distortion of the healthy balance between inflammation and tissue homeostasis (10). Aging affects this balance in multiple ways including decreased efficiency of cellular housekeeping and tissue repair functions (11–13), which leads to the accumulation of cellular debris (14), increased endoplasmic reticulum stress (15) defective tissue homeostasis (16, 17), declining endocrine functions (18) changes in RNA compositions (19) immune senescence (20, 21), and increased inflammation (22, 23). Understanding how interactions between these systems can affect the outcomes of pharmacological studies (24), offering opportunities to slow the effects of aging (25) and prevent the development of associated diseases (26–28).

The central actors in maintaining tissue homeostasis are macrophages (29, 30), dendritic cells (31, 32), stem cells (33), and vitamin D (34–38). However, our understanding of their interaction remains incomplete, as evidenced by multiple inconclusive clinical trials seeking to halt the progression of cardiovascular disease associated with vitamin D deficiency (39–43). In previous reports, we have reported that vitamin D deficiency affects the tipping points that control network–network interactions in over 500 diseases, including atherosclerosis (44).

This study provides deeper insights into the role of vitamin D in cardiovascular disease development. The use of a spectral clustering method for identifying vitamin D-regulated processes in 4,821 diseases provides a global effect positioning system that can be used to determine cause–effect relationships in atherosclerosis. Examination of vitamin D-associated genomic and non-genomic pharmacologies using this positioning system revealed origins of non-linear dose–response relationships observed in preclinical and clinical studies with vitamin D. In addition, it also identified functional relationships between multiple reciprocal feedback loops that sometimes regulate the opposing functions activated by nuclear and membrane-associated vitamin D receptors. Among the processes of relevance for the treatment of cardiovascular diseases are p53-mediated regulation of atherosclerotic plaque stability and macrophage autophagy, which are known to safeguard the plasticity of macrophages in tissue homeostasis. We have used an unsupervised machine learning approach for deciphering these complex cause–effect relationships and used protein network analysis for identifying interactions between protein networks that regulated the functions that are of relevance in atherosclerosis. The initial step was aimed at establishing a global effect positioning system (Table 1) using information theory-based methodology to allow for an unbiased analysis of vitamin D pharmacology in diseases. Information flow-based cause–effect analysis of 4,821 diseases led to the identification of the vitamin D interactome (Figure 2) comprising 363 proteins, where 64% of proteins were found to have highly statistically significant (p-value = 2.06 × 10⁻³³) disease associations. This observation establishes the value of our protein swarm-based framework for investigating cause–effect relationships in protein–protein interaction (PPI) networks and complements previous work reported by Silverbush and Sharan (161) who used drug response and cancer genomic data to orient the human PPI network.

Examination of the role of the vitamin D interactome in 4,821 diseases provided two key findings: (1) vitamin D-regulated processes are involved in over 2,000 diseases including cardiovascular disease, and (2) all functions regulated by vitamin D are subject to epigenetic regulation (Figure 3). Epigenetic regulation of atherosclerosis is well established (162), and our finding of the involvement of vitamin D signaling contributes to the development of diagnostics or therapeutic agents in atherosclerosis. Also depicted in Figure 3 is the observation that cardiovascular diseases including atherosclerosis are affected by PPI in swarms 580–611, 1,255–1,268, 1,990–1,995, 6,008–6,017, and 7,116–7,158.

The effects of vitamin D signaling in cardiovascular diseases were further examined as three separate categories: (1) genomic signaling mediated by PPI in swarms 580–611, 1,255–1,268, 1,990–1,995, 6,008–6,017, and 7,116–7,158 (Table 2), (2) mixture of genomic and non-genomic signaling mediated by PPI in swarms 726–788 (Table 3), and (3) non-genomic signaling mediated by PPI in swarms 3,620–3,633 (Table 4). Tables 2–4 show the results of evidence-based cause–effect analysis using network overlaps of proteins in the three swarm categories with PPI networks of atherosclerosis and the 10 biological functions affected by aging and involved in atherosclerosis. The significance of vitamin D's genomic, mixed genomic, and non-genomic signaling in atherosclerosis related to the 10 biological functions associated with atherosclerosis (Figures 4–6) was determined by measuring the extent of network overlap, ability to engage in physical interactions, and centrality of shared network nodes. Hence, the main inference from the results shown in Tables 2, 3 and Figures 4, 5 is that genomic and mixed genomic signaling by vitamin D modulates the 10 important biological functions associated with atherosclerosis. Additionally, examination of network overlaps between these 10 biological functions indicated that they are part of an integrated regulatory scheme that includes macrophage polarization and macrophage autophagy as central elements.

The effects of vitamin D's non-genomic signaling are summarized in Table 4 and Figure 6. The absence of overlap between proteins in swarm cluster 3,620–3,633 and macrophage autophagy in combination with the observation that these clusters are enriched with proteins involved in fibrosis suggests that non-genomic signaling by vitamin D plays an important role in tissue homeostasis.

The identification of protein network–network interactions described above allows the generation of a comprehensive view detailing how interactions between the 10 biological functions affect atherosclerosis.

A critical factor in atherosclerosis-associated mortality is plaque stability (163), where stable plaque forms carry a low risk of sudden mortality while rupture of unstable forms causes myocardial infarction and stroke. Our analysis (Tables 2, 3; Figures 4, 5) indicates that vitamin D signaling plays a key role in balancing plaque stability (164) by regulating macrophage autophagy (90) and macrophage polarization, a key determinant of the M1/M2 ratio, where macrophages with M1-like characteristics are associated with the inflammatory stages of atherosclerosis while macrophages with M2-like characteristics have anti-inflammatory characteristics and promote plaque regression (91, 165). Declining macrophage autophagy increases macrophage apoptosis, decreases macrophage efferocytosis, increases oxidative stress, and destabilizes plaque (132, 166).

A unique property of macrophages is that they can directly generate calcitriol, which is the most active form of vitamin D, in response to a variety of input signals (167). Calcitriol, in turn, activates macrophage autophagy and shifts the macrophage polarization toward anti-inflammatory M2-like macrophages (168, 169). Thus, by shifting macrophage polarization toward M2 (170), calcitriol can decrease inflammation (171) and increase the clearance of cellular debris by efferocytosis (172).

Additionally, of relevance to the development of atherosclerosis is aging. Aging shifts macrophage polarization toward M1 phenotypes (173), increases cellular senescence (174) and inflammation (175), and decreases macrophage functions (176). Calcitriol can counter aging-induced acceleration of atherosclerosis at the plaque initiation (177–179), progression (179, 180), rupture (181), regression (182–184), and healing stages of atherosclerosis (179, 185). Accordingly, shifting macrophage ratios from M1 to M2 controls plaque size and stability (170). For example, low shear stress or high endoplasmic reticulum negatively affects autophagy, and shifting macrophage polarization toward inflammatory M1 accelerates atherosclerosis (186–190). Vitamin D deficiency exacerbates this development by increasing M1 macrophage populations and decreasing autophagy (115, 191). Calcitriol administration can reverse this trend by shifting macrophage populations toward M2 phenotypes, which increases macrophage autophagy (192–195).

However, an important finding of our investigation is that the over-enrichment of M2 macrophage populations can result in excessive fibrosis. In other words, vitamin D's non-genomic signaling via PDIA3 at higher calcitriol concentrations promotes fibrosis, leading potentially to plaque destabilization (196). Maintenance of the delicate balance between M1 and M2 macrophage subpopulations is critical as it affects tissue homeostasis. Experimental data suggest that too much or too little on either side of the equation results in detrimental outcomes (197). This balance can be affected in multiple ways. For example, the involvement of immune responses in this delicate balancing act is implicated by observations that cathelicidins produced by vitamin D's genomic response to infections activate autophagy and shift macrophage populations toward M2 (198). In addition, the identification of regulatory schemes clearly establishes that vitamin D regulates multiple feedback loops (76, 199–203) involved in tissue homeostasis, which are subject to additional regulation by nitric oxide (204) and redox signaling (205). Evidence of these types of non-linear (U-shaped) cause–effect relationships/dose responses, expected due to opposing pharmacologies, is indeed observed in numerous preclinical and clinical studies with vitamin D [see references (156–160)].

The dynamics of protein–protein interactions impart functional diversity to proteomes and balance functions separating health and diseases. From a whole organism perspective, single-cell approaches do not provide the system-level insight that is needed to understand how interactions between proteomes at tissue levels affect the progression of a disease or the responses of organisms to treatments (206). To generate the required insights, integrated approaches are needed that can determine relationships between large heterogeneous, unstructured, and noisy data stored in different databases. By taking advantage of the modular designs of proteomes and principles governing information transfers in integrated systems (207, 208), protein swarm-based cause–effect analysis delivers unbiased insight into how interactions between tissue- and cell-associated proteomes regulate the physiological and pathological states of organisms.

“Swarm Intelligence” based analysis reveals cause-effect relationship between the many variables involved and provides understanding that is very difficult to ascertain with other means.

The reach of the current vitamin D-centered cause–effect analysis is limited by its focus on determining the system pharmacology of calcitriol. Therefore, our study neglects the potential contributions of calcidiol or the various vitamin D metabolites. Furthermore, technical constraints limit the resolution of network–network interactions by setting upper limits to the number of protein swarms used for sampling tissue proteomes. Lastly, the limiting capability of machine learning approaches in the biological cause–effect analysis is the scarcity of data on the concentration time dependency of dynamic cause–effect relationships at a system-wide scale.

Using a global framework for proteomics-based cause–effect analysis, our analysis identified a broad vitamin D-regulated scheme involved in tissue homeostasis of over 2,000 diseases. By providing a perspective on how interactions between vitamin D's genomic and non-genomic actions, nitric oxide and redox signaling, macrophage autophagy, macrophage polarization, and p53 signaling shape tissue homeostasis, our analysis addresses a long-standing enigma presented by observations that vitamin D supplementation, despite projected benefits, produces little benefits in clinical studies (209). Recognizing the importance of p53, nitric oxide, and redox signaling in vitamin D-regulated processes offers the opportunity to increase the efficacy of vitamin D supplementation in cardiovascular diseases. Toward this end, modalities integrating nitric oxide redox signaling and vitamin D-centered pharmacologies have already shown promising outcomes (210). Refinement of these approaches is expected to improve the treatment of cardiovascular diseases and increase the quality of life in aging (211).