Cytokine Circuits in Cardiovascular Disease

Introduction

Atherosclerosis is a ubiquitous pathology in humans, being observed in ancient populations for at least the last 4000 years (Allam et al., 2009). Its slow natural progression amplifies during aging and can lead to acute myocardial infarction, typically beyond the 4^th decade of life. Anatomically and histologically, atherosclerosis is characterized by the development of a pronounced chronic inflammatory response in the intimal layer of artery walls (such as coronary arteries in human). The arterial intima is the layer between the arterial endothelium and the first band of elastic lamina in arteries, such that the intima is positioned on top of the smooth muscle cell rich medial layer and the outer arterial layer known as the adventitia. The progression of inflammation increases the size of the intima, forming an inflamed structure called plaque, which narrows the volume of space for blood flow through the vessel (stenosis). The inflammatory response can also result in sudden rupture of intimal plaque integrity that can trigger episodic occlusion of the vessel (often a coronary artery supplying the heart) giving rise to rapid ischemia and consequent myocardial infarction. As discussed in more detail below, atherosclerosis arises from two intersecting pathophysiological developments: (i) overwhelmed or defective cholesterol handling and (ii) low-level constitutive activation of the arterial vasculature due to oscillatory blood flow, such that vascular permeability appears to increase enough to allow cholesterol to access the artery wall in the first place. That is, despite low level inflammation being a natural feature of vessels characterized by oscillatory flow, atherosclerosis will typically not take hold if plasma cholesterol is low because it must begin to accumulate in the artery wall to advance disease. Because oscillatory blood flow is a feature of curved or branching arteries, plaques tend not to form continuously along the arterial intima, but rather at focal points around branches and curves of arteries.

Atherosclerosis has historically been the leading cause of cardiovascular disease. However, major advances in treatment, especially the use of statins, and improvements in diet and lifestyle over the past several decades have markedly reduced atherosclerosis as a cause of cardiovascular mortality, giving way to heart failure as the cardiovascular condition most starkly on the rise (Benjamin et al., 2018). Nonetheless, because of the ubiquitous tendency for atherosclerosis to develop in human subjects, there remains a need to find new ways to combat atherosclerosis. That is, some populations remain at high risk as relative non-responders to frontline lipid-lowering therapy (statins), while others may benefit from additional drugs that act in concert with standard-of-care approaches. In particular, as we discuss herein, autoimmune and chronic inflammatory diseases including lupus, rheumatoid arthritis, psoriasis, and, more recently, inflammatory bowel disease have been linked to increased cardiovascular comorbidity and potential premature mortality in these patients.

Besides drugs that target lipid management, there is now emerging evidence that targeting inflammation, particularly the cytokines that orchestrate inflammation, can further lower atherosclerosis. Therapeutically targeting soluble cytokines indeed has yielded dramatic benefits in a wide variety of inflammatory diseases, including in many of the autoimmune diseases listed above. However, translation of these powerful approaches into patients for the directed treatment of established cardiovascular disease is in its infancy. Here, we review the basic pathophysiology of cardiovascular inflammation, discuss the current status of antiinflammatory therapy in human atherosclerosis, followed by in-depth analysis of the underlying cytokine networks. We then focus on what is known and what remains unknown about the risk of cardiovascular disease in autoimmunity, with a view toward the possibility that treatments will emerge to combat not just the underlying autoimmunity but also its coincidence with cardiovascular disease.

Genesis, evolution and cellular composition of the atherosclerotic plaque

The Nobel Prize-winning, pioneering work of J. Goldstein and M. Brown on the cell biology of low density lipoprotein and its major receptor led to, in their own words, the “inescapable conclusion that the LDL pathway functions in man to protect against atherosclerosis”(Goldstein and Brown, 1977). Cholesterol is an essential molecule in animal cells that is required for appropriate membrane integrity and to facilitate signaling within membranes. It is also the precursor of various metabolites like vitamin D, bile acids, oxysterol intermediates that signal in immunity, and steroids, including sex hormones and hormones of the adrenal gland (Cyster et al., 2014). As such, cholesterol is synthesized in the liver and is distributed to cells in the body through secretion into plasma of esterified cholesterol packaged within the lipoprotein particles very low density lipoproteins (VLDL) and LDL, in which apolipoprotein B100 serves as the key amphipathic protein that assembles the lipoprotein. Alternatively, cholesterol is absorbed from the diet, packaged within intestinal epithelial cells along with triglycerides in particles called chylomicrons that rely on the alternatively spliced form of apolipoprotein B, apoB48, for assembly. These lipoprotein particles deliver cholesterol to tissues after passage through the endothelial barrier via transcytosis mechanisms involving a range of different receptors (Zhang et al., 2018), or via reduced junctional integrity associated with increased vascular permeability (Essler et al., 1999; Mundi et al., 2018).

Much of what we now know about cholesterol trafficking and metabolism, and associated atherosclerosis, has been learned using experimental models of the disease, especially from mouse models. The rate of LDL-cholesterol transcytosis across the endothelium is affected by conditions that impact cardiovascular susceptibility; for instance, estrogen, with an established protective role against atherosclerosis, limits transcytosis rates (Ghaffari et al., 2018; Sessa, 2018). Once LDL-cholesterol has entered the artery wall, it fails to broadcast deeply and does not readily pass out of tissues (Bremmelgaard et al., 1986; Stender and Hjelms, 1984). This entrapment of LDL in the subendothelial space is the basis for the widely held “response-to-retention” hypothesis for the genesis of atherosclerotic plaque (Williams and Tabas, 1995). The “response” of greatest importance is the ensuing inflammatory response, the subject of the present review.

The “response-to-retention” hypothesis was generated as a refinement of the “response-to-injury” hypothesis originally proposed by Russell Ross (Ross et al., 1977). The latter importantly underscores that the status of plaque endothelium is critical. While it is not necessarily overtly injured, the arterial endothelium at sites prone to plaque development (arterial branches and areas of nonlaminar blood flow) displays constitutive activation of NFkB and expression of adhesion molecules like Vascular Cell Adhesion Molecule 1 (VCAM1), which is rather normally observed in postcapillary venules activated by master proinflammatory cytokines TNFα or IL-1β. This state of activation in the endothelium is not sufficient to initiate plaque formation unless cholesterol levels in plasma rise (Hajra et al., 2000). That is, even C57BL/6 mice display activated arterial endothelium, but atherosclerosis does not develop in standard C57BL/6 mice. Mouse strains must be engineered to elevate plasma cholesterol, such as via loss of LDL receptor or VLDL receptor cofactors like apoE (Ishibashi et al., 1993; Zhang et al., 1992) following consumption of a cholesterol-enriched diet (Lichtman et al., 1999). Thus, the core elements of both the response-to-injury (initial endothelial activation) and response-to-retention (cholesterol accumulation) hypothesis appear to apply and are not mutually exclusive but rather complementary.

From an immunological perspective, it is fascinating to note that a specialized population of mononuclear phagocytes, sometimes referred to as vascular dendritic cells (Choi et al., 2009; Jongstra-Bilen et al., 2006; Millonig et al., 2001; Paulson et al., 2010), characterizes the normal arterial intima. So far as is known, these cells are present from birth onward in humans (Bobryshev and Lord, 1995, 1998; Waltner-Romen et al., 1998) and mice (Jongstra-Bilen et al., 2006; Liu et al., 2008), independent of the presence of atherosclerosis, possibly evolving to carry microbes out of the artery wall upon infection (Roufaiel et al., 2016). Cybulsky and colleagues have demonstrated that initiation of atherosclerosis is linked to the activity of these cells (Cybulsky and Jongstra-Bilen, 2010; Paulson et al., 2010) (Williams and Randolph, unpublished data). However, how they initiate the inflammatory cascades that promote the genesis of atherosclerotic plaque is not well understood. Paulson et al. make a case for these phagocytes being the first to acquire and accumulate cholesterol. Indeed, these cells may participate in the import and retention of cholesterol into the artery wall itself (Paulson et al., 2010), while simultaneously fostering the inflammatory triggers that subsequently recruit monocytes steadily throughout the course of the ensuing chronic disease (Swirski et al., 2006).

Cholesterol taken up by these vascular resident phagocytes and by monocytes that begin to swarm in, in concert with cholesterol retention in the artery wall, is readily esterified and stored in cytoplasmic lipid droplets. Such a phagocyte heavily burdened with lipid droplets is referred to as a foam cell. One long-held concept thought to link foam cells to the progression of atherosclerosis was the idea that esterified cholesterol could be not contained in esterified form efficiently enough, such that some excess level of free cholesterol would occur to drive signaling changes and cellular toxicity. In this model, the foam cell was considered the central driver of inflammation in plaques. However, in the last few years, a new concept emerged that, even early on in plaques, cholesterol crystals are formed and these, in turn, activate the inflammasome and lead to the local production of IL-1β (Duewell et al., 2010). This concept is currently one of the most compelling in the field and can be associated with recent clinical studies (discussed below) as well as much current research activity in the field. In concurrence with this concept, recent work illustrates that foam cells in plaques are not expressing inflammatory cytokine mRNA for IL-1β or other proinflammatory cytokines or chemokines, but instead such expression is observed in CCR2+ nonfoamy macrophages within plaques that are more akin to recently recruited monocytes (Kim et al., 2018) (Figure. 1). These data fit with observations that foam cell formation is not inherently pro-inflammatory, but rather antiinflammatory due to activation of the transcription factor Liver X Receptor (LXR) via binding cholesterol synthesis intermediate desmosterol as a ligand (Spann et al., 2012). Collectively, these new findings raise important questions for the future. Are the IL-1β+ inflammatory macrophages recovered from plaque an end-stage macrophage that has gone down a distinct and nonoverlapping differentiation pathway from that of foam cells (Figure 2)? Or are the proinflammatory CCR2+ cells monocyte-like cells part way through the process of differentiating into foam cells (Figure 2)? Since it’s known that newly recruited monocytes have limited ability to penetrate deeply within plaques (Williams et al., 2018), the presence of CCR2⁺ inflammatory macrophages are nearer the luminal surface, which may give these macrophages better access to immunomodulatory drugs.

Examples are readily found in the literature that illustrate that monocytes can go through a highly inflammatory transitional stage as they differentiate during inflammation. First, monocytes in the colon continuously differentiate to resident colonic macrophages that have antiinflammatory roles. However, in experimental colitis, they are arrested in the transitional phase of differentiation that coincides with a highly inflammatory state (Bain et al., 2013) (Figure 2). In a model of lung fibrosis, monocytes in transition to alveolar macrophages are inflammatory and pro-fibrotic, but ultimately differentiate to anti-inflammatory alveolar macrophages resembling resident macrophages observed in homeostasis (Misharin et al., 2017) (Figure 2). However, from single cell RNA sequencing (scRNAseq) data (Kim et al., 2018), an algorithm to predict differentiation pathways can be applied. The results suggest that there may be little or no cross-over in the pathways that lead to the IL-1β+ inflammatory signature versus that of the foam cell populations (unpublished observations). This issue is important to settle experimentally because if the inflammatory macrophage in plaque is on a continuum toward a non-inflammatory state, one scenario therapeutically might be to accelerate progression through the transient inflammatory period. However, if the inflammatory state is a distinct pathway that involves a common precursor with the foam cell (like the monocyte), but otherwise does not overlap with pathways leading to foam cells, therapeutics might be envisioned that would block the inflammatory trajectory altogether without affecting the cholesterol removal function of foam cells.

Besides monocyte recruitment expanding the pool of foam cells, plaque macrophages may also expand by proliferation (Robbins et al., 2013). The relative balance of macrophage proliferation versus recruitment from blood remains rather unclear. Beyond macrophages, there is growing appreciation for a diversity of cell types and phenotypes that also are found in plaques, including lymphocytes and dendritic cells (Choi et al., 2011; Frostegard et al., 1999; Macritchie et al., 2012; Sage et al., 2014; Yun et al., 2016). As discussed in detail elsewhere (Hansson and Hermansson, 2011; Wolf and Ley, 2019), antigen-specific T cells (primarily Th1/Th17) may promote plaque inflammation, whereas regulatory T cells may reign in plaque inflammation (Ait-Oufella et al., 2006; Subramanian et al., 2013). Recent advances using tetramers to identify regulatory T cells (Treg) recognizing a peptide in apoB, the major protein in LDL, supports the immune-modulatory potential of targeting these cells in mouse and man (Kimura et al., 2018) and the possibility that vaccination could be used to expand them (Kimura et al., 2015; Virmani et al., 2006). B lymphocytes are also implicated in local and systemic influences on disease progression, with pending clinical studies on the horizon (Sage et al., 2018). Innate B cell-derived antibodies have a disease-protective function, in part by clearing and modifying cellular responses to oxidized LDL (Sage et al., 2018). Furthermore, the arterial adventitia contains a diverse and rich immune cell milieu, including a variety of macrophages with a distinct life cycle and function from those in the intima (Ensan et al., 2016; Lim et al., 2018). As plaque advances in mice, especially aging mice, the adjacent adventitial compartment can develop a robust tertiary lymphoid compartment with disease-modulating activity (Grabner et al., 2009; Hu et al., 2015). It is beyond the scope of this review to discuss the biology and cell types of the adventitia, but it seems quite likely that the growing research on this topic will affect our understanding of plaque genesis and progression. We furthermore emphasize the technical problem that the adventitial network of cells creates for the study of intimal phagocytes: even in plaque-bearing arteries, single cell suspensions of the artery typically are comprised in majority of adventitial cells, such that studies of atherosclerosis by flow cytometry requires one to proceed with caution.

Finally, beyond hematopoietic cells, it has been long appreciated that smooth muscle cells comprise key components of advancing plaques (Schwartz et al., 1986; Schwartz et al., 2018). Indeed, very early work in the field highlighted the clonal proliferation of smooth muscle cells as a crucial feature of plaque (Benditt, 1974). It is now universally appreciated that the presence of smooth muscle cells and the matrix they produce near the plaque lumen (typically called the “fibrous cap”) serves to physically stabilize the plaque by separating the deeper areas in plaque, which in advanced stages are necrotic and highly prothrombotic, from the bloodstream at the arterial lumen (Figure 1). In humans, the thickness of the fibrous cap is inversely linked to the probability that a plaque will rupture to initiate myocardial infarction (Virmani et al., 2006). Smooth muscle cells have a complex origin, with intimal plaque smooth muscle cells invading from the medial and adventitial layers, but also sometimes arising during endothelial to mesenchymal transitions (Kovacic et al., 2012). Strikingly, smooth muscles also have the capacity to transdifferentiate into macrophage-like cells that contribute to the foam cells present in plaque (Cherepanova et al., 2016; Rong et al., 2003; Shankman et al., 2015). At least for foam cells that still bear enough features of smooth muscle cells to discern them as such (eg., cells expressing residual smooth muscle actin mRNA), these foam cells share with the more classical macrophage foam cells the feature of being relatively noninflammatory in gene expression profile (Kim et al., 2018) (Figure 1).

While we have emphasized that living foam cells are not the cells in plaque actively producing cytokines (Kim et al., 2018), this finding does not imply that foam cells do not strongly promote progression of plaque or inflammation indirectly (Figure 1). One major role they play in disease progression is linked to their death, either by necroptosis (Lin et al., 2013), which is directly connected to inflammation and release of danger-associated molecular patterns (DAMPs), or apoptosis (Feng et al., 2003). Pyroptosis that is, by definition, connected to inflammasome activation may occur in plaque, given the role of NLRP3 in disease progression (Duewell et al., 2010). However, based on the expression profile of plaque macrophages, this mode of death is more likely relevant to nonfoamy macrophages. In contrast to necroptosis and pyroptosis, apoptosis is not an inflammatory process, but if apoptotic cells are not readily removed through a process called efferocytosis, inflammation can nonetheless ensue. Efferocytosis is well documented to be defective in atherosclerosis (Thorp et al., 2008; Thorp and Tabas, 2009). Consequently, dead cell debris accumulates within plaques, with the accumulated material forming acellular pockets of debris referred to as necrotic core. Expansion of necrotic core is destabilizing and, in man, increases the probability of clinically significant plaques vulnerable to rupture. Recently, blockade of CD47, “the don’t eat me signal,” was shown to restore efferocytosis in experimental atherosclerosis (Kojima et al., 2016). That atherosclerosis progression was blunted strongly by anti-CD47 treatment indicates that restoring efferocytosis guards against the accumulation of necrotic core that drives plaque vulnerability, even as it also underscores the important role that failed efferocytosis has in driving plaque inflammation and progression. The study implicated the well-characterized proinflammatory cytokine tumor necrosis factor (TNF) in the blockade of efferocytosis (Kojima et al., 2016). It also seems possible that the lingering of necrotic debris derived from foam cells, with accumulated DAMPs and extracellular cholesterol crystals (Katz et al., 1976), is ultimately linked to the recruitment and activation of the nonfoamy CCR2⁺ IL-1β⁺ macrophages identified in plaques, tying the two macrophage types together in a positive feedback loop that promotes disease.

The cytokine IL-1β in cardiovascular disease – from bench to bedside

The aforementioned work suggests that targeting cytokines like TNF and IL-1 might comprise a viable therapeutic strategy to reduce disease burden. The success of statins as standard-of-care therapy naturally means that new therapeutics would most typically be add-ons to statin treatment. Thus, it is important to recognize that statin treatment is among the disease management changes in recent history that has significantly altered the clinical presentation of atherosclerosis. For instance, based on analysis of human endarterectomy specimens, plaque macrophage content has declined, while collagen content has risen (van Lammeren et al., 2014). Similarly showing a trend toward features less likely to prompt plaque rupture is the increase in calcified burden and decrease in noncalcified burden (which would include inflammatory cells and necrotic core) in response to statin therapy (Lee et al., 2018). This shift has led to a de-emphasis on the problem of plaque rupture, with an increasing recognition that erosion of the plaque surface rather than acute rupture is becoming more common (Libby and Pasterkamp, 2015). One might then wonder if mouse models of experimental atherosclerosis currently in use still mimic therapeutically relevant atherosclerosis in the modern era, as the features of mouse plaque (like high macrophage content) more closely resemble human disease as it appeared in the past than the present. However, considering the success of recent therapeutic approaches that arose from research in mouse models, it is reasonable to cautiously argue that the mouse models remain highly relevant and valuable. A major example of recent success in translating work from experimental models to the clinic is the work that led to the successful identification of IL-1β as a viable therapeutic target.

As its name suggests, interleukin 1 (originally called human leukocytic pyrogen) is a long-established mediator of inflammation first purified in 1977 (Dinarello et al., 1977) and cloned in 1984 (Auron et al., 1984). The cytokine was first pursued for its functional connection to fever and the acute phase response, and this work paved the way for the concept that there were, in fact, “hormones of the immune system”. It was also quickly recognized that IL-1, along with TNF, acts as a central regulator of endothelial cell activation. Through activation of NFkB that in turn programs upregulation of adhesion molecules and chemoattractants, IL-1 orchestrates inflammatory cell recruitment to sites of inflammation (Bussolino et al., 1988; Furie and McHugh, 1989; Huber et al., 1991; Moser et al., 1989; Pober et al., 1986), an activity especially relevant to atherosclerosis. It is now clear that IL-1 is part of a family of cytokines, with family members being broadly linked to many inflammatory and developmental diseases. In addition to the two majors of IL-1, IL-1α and IL1β, family members include IL-18, IL-33, IL-36α, IL-36β, IL-36γ, IL-37, and IL-38 (Dinarello, 2018).

IL-1β is generated as an inactive propeptide in the cell cytoplasm, and activation requires cleavage of the pro-form of the molecule. Such cleavage can occur either intra- or extra-cellularly. IL-1β is typically cleaved by caspase-1 (interleukin-1 converting enzyme), an intracellular cysteine protease that also requires its own cleavage for activation (Figure 3). IL-1β propeptide can be expressed and the inflammasome can be activated in a variety of cells, including endothelial cells, but is especially relevant in macrophages (Libby et al., 1986; Xiao et al., 2013). Expression can be stimulated downstream of any one of a multitude of stimuli, including recognition of microbes, binding of extracellular ATP, hypoxia, and phagocytosis of non-degradable particles. The NLRP3 (nucleotide oligomerization domain (NOD), leucine rich repeat (LRR), and pyrin domain (PYD) containing protein 3)-mediated inflammasome is broadly associated with a variety of disease and represents the best studied mechanism for IL-1β activation (Rajamaki et al., 2010) (Folco et al., 2014). The NLR family consists of approximately 2 dozen family members in human and in excess of 30 in mice, many of which share the NLR and Pyrin domains that are active components of the inflammasome complex (Guo et al., 2015; Ting et al., 2008).

Canonical activation of the inflammasome involves engagement of two independent signaling pathways (sometimes termed signal I and signal II) (Figure 3). In-depth analysis of the signaling pathways responsible for inflammasome activation were recently reviewed (Baldrighi et al., 2017; Grebe et al., 2018; He et al., 2016). In brief, signal I, or priming, is the step that induces expression of IL-1b and NLRP3 polypeptides. It occurs via activation of classic pattern recognition receptors, such as TLR2 or TLR4, or inflammation sensors like TNFR or IL1R (Franchi et al., 2009). These pathways converge on the common downstream transcription factor NFκB signaling pathway. NFκB activation prompts the expression of NLRP3 and IL-1β. Signal 2 then involves assembly of the NLRP3 inflammasome macromolecular complex, which in turn activates caspase 1 to cleave pro-IL1b to its active form. Crystals are often associated with the activation and progression of this second signal. In atherosclerosis, cholesterol crystals form in plaques (Small et al., 1984), including early plaques (Duewell et al., 2010), and even in cultures of cholesterol with macrophages (Kellner-Weibel et al., 1999), and these are thought to be vital to activation of NLRP3 and IL-1b in plaques (Figure 3). Mechanistically, crystal particulates can rupture the phagolysosome within macrophages, leading to spillage of cathepsin enzymes that in turn trigger NLPR3 activation and assembly. In late plaques, even during disease regression, extracellular crystals are common.

IL-1β and IL-18 lack N-terminal tags to promote their entry into the endoplasmic reticulum and subsequent export through conventional secretion pathways. As proteins generated in the cytoplasm, their secretion requires the formation of a specialized pore formed by the gasdermin d (Gsdmd) protein, which is cleaved downstream of active capase-11 (Kayagaki et al., 2015; Shi et al., 2015a). Following cleavage, Gsdmd oligomerizes at the cell membrane creating a pore for the release of cellular contents, including inflammatory molecules IL-1β and IL-18. This release also includes loss of membrane potential (K+), and calcium entry and, at least often, leads to the form of cell death called pyroptosis. However, whether this pathway inevitably leads to cellular death, or whether cells are able to release IL-1β and persist in the tissue is under great scrutiny, and this question is highly relevant to links between macrophage death, impaired efferocytosis, and inflammation in atherosclerotic plaque, as discussed above. Gsdmd-mediated pore formation was recently found to be controlled by the molecule called Ca2+-dependent recruitment of the endosomal sorting complexes required for transport (ESCRT), which functions to repair damaged membranes and thereby sparing death as a consequence of gasdermin pore formation (Evavold and Kagan, 2019; Ruhl et al., 2018).

Treatment of pigs chronically with exogenous IL-1β augments atherosclerosis progression (Shimokawa et al., 1996), whereas loss of IL-1β in apoE^−/− atherogenic mice decreases plaque burden and endothelial activation (Kirii et al., 2003). Fitting with the latter, atherosclerosis in mice depends on the MyD88 signaling adaptor relevant to IL-1R signaling (Michelsen et al., 2004) and NLRP3 inflammasome (Duewell et al., 2010). Subsequently, in humans, IL-1β and NLRP3 were associated with the risk of coronary artery disease (Afrasyab et al., 2016), and showed elevated expression in the aorta of patients with atherosclerotic plaque formation (Paramel Varghese et al., 2016; Shi et al., 2015b; Zheng et al., 2013).

The canakinumab anti-inflammatory thrombosis outcome study (CANTOS) trial was the first anti-inflammatory trial to treat atherosclerosis (Ridker et al., 2017), and the target chosen for evaluation was IL-1β. Specifically, the trial used a fully humanized monoclonal antibody, canakinumab, previously approved for the treatment of rheumatoid arthritis (Ruperto et al., 2012; Schett et al., 2016) and cryopyrin-associated periodic syndrome (Lachmann et al., 2009). This drug to neutralize IL-1β was tested in high-risk cardiovascular disease patients who had a history of myocardial infarction and elevated acute phase protein C-reactive protein (CRP). In addition to being randomized for receipt of canakinumab, patients were maintained on standard-of-care therapy, which typically included aggressive statin use. A dose of 150 mg every 3 months showed a significant benefit in the primary endpoint that surveyed the number of cardiovascular events occurring between 3–4 years later. This IL-1β-targeted therapy effectively reduced circulating cytokine IL-6 and CRP, independent of any changes in cholesterol. While not a primary outcome of the study, canakinumab was also associated with reduced cancer mortality(Ridker et al., 2017), which is consistent with previous studies (Apte et al., 2006).

CANTOS established that inflammation can be targeted in atherosclerosis to reduce clinically significant cardiovascular disease. However, targeting of IL-1β has its own inherit risks. Patients receiving canakinumab experienced elevated fatal infections, which led to an overall no change in “all-cause” mortality associated with the study (Ridker et al., 2017). The inclusion of only patients with elevated CRP levels of >2 mg/L, shows the potential for targeting the inflammatory arm of atherosclerosis, but it leaves open the question of whether this approach will be effective for the general population that may have local generation of IL-1β in plaques but not signs of systemic inflammation. Importantly, patients that showed the most dramatic reduction in serum inflammatory markers correlated with those having least occurrence of cardiovascular events. Thus, overall, this study represents an important cornerstone in the field as it suggests that targeting the inflammatory arm of atherosclerosis can provide clinical benefit in atherosclerosis, even in the context of statin therapy. It remains uncertain, however, exactly where and how canakinumab conferred protection. Did it make its way into plaque to bind and neutralize IL-1b inside of the plaque milieu? Or did it largely fail to access the dense and complex plaque interstitium but instead critically act at the endothelial surface of the affected artery walls by inhibiting endothelial activation and thus further recruitment of inflammatory cells? Or, given that responders had evidence of systemic inflammation that was reduced by the drug, was its main effect on a systemic target that may be anatomically distinct from plaque? If a systemic target was the most important target, what systemic target was pivotal? Finally, although its net effect was beneficial, were there detrimental effects of the blockade that suppressed an even greater benefit that might have otherwise occurred, given that IL-1b also has roles that are atheroprotective (Gomez et al., 2018)? Answering these questions that aim to better understand the mechanism of action might pave the way for a therapeutic approach to be devised in the future that has similar or greater reductions in atherosclerotic inflammation without heightened risk in host defense.

Systemic modifiers of atherosclerosis – lessons from autoimmunity

The possibility that therapeutics like canakinumab act at least in part systemically to quell atherosclerosis progression is not far-fetched, given the growing appreciation for systemic regulators of cardiovascular disease. So far, we have discussed atherosclerosis as a local disease, and indeed, lesions are local, chronic foci of inflammation. However, it is clear that there are many systemic modifiers of disease. Such modifiers include several well known risk factors like smoking, diet, and sedentary life style, as well as others like inhaled environment particulates (Bai and Sun, 2016) (Figure 4), with many of these common risk factors still not fully understood mechanistically. Dietary interventions like a shift to a “Mediterranean diet” strongly reduced cardiovascular events, even in a population of subjects also treated with statins (Estruch et al., 2018). Diet, of course, in turn affects the intestinal microbiome that, in the presence of the right substrates, can promote the generation of pro-atherogenic metabolites like trimethyl amine oxide (TMAO) from dietary intermediates (Tang and Hazen, 2014) (Figure 4). The mechanism by which TMAO promotes atherosclerosis is also not fully understood.

A growing list of systemic modifiers mechanistically act to affect the bone marrow’s production of myeloid cells. It has been documented that high neutrophil and monocyte counts in the blood are an independent risk factor for cardiovascular disease (Chapman et al., 2004; Coller, 2005; Danesh et al., 1998; Nasir et al., 2005), perhaps in part by increasing the supply of cells able to become recruited to atherosclerotic plaques. Mouse models of have begun to shed light on systemic modifiers that control myeloid cell output. Atherosclerosis is elevated after an event like myocardial infarction due to the elevation in β3-adrenergic signaling in a brain-bone marrow axis that sends more monocytes into the bloodstream (Dutta et al., 2012) (Figure 4), which in turn is linked to GM-CSF production (Anzai et al., 2017). The GM-CSF receptor also accounts for elevated myeloid cell output from the bone marrow driven by hypercholesterolemia (Wang et al., 2014). Atherosclerosis is also elevated by a disrupted sleep cycle that results in a reduction in the sleep-wake cycle mediator hypocretin. Hypocretin normally suppresses myelopoiesis through limiting production of colony stimulating factor 1 (CSF-1) in the bone marrow (McAlpine et al., 2019). Conversely, warm ambient temperature is associated with reduced cardiovascular risk and is linked in mice and man to a reduced output of monocytes from the marrow (Williams et al., 2017) (Fig. 4). Warm ambient temperature is often associated with increased sun exposure, potentially affecting vitamin D levels, which has been argued in observational studies to be linked to cardiovascular risk. However, a recent large clinical trial of vitamin D supplementation fail to indicate a key role for vitamin D in affecting cardiovascular outcomes (Manson et al., 2019), raising questions about the validity of the link (Fig. 4). Finally, in a phenomenon called clonal hematopoiesis, somatic mutations that occur in hematopoietic stem cells during aging can allow for selection and outgrowth of clones that express disease-adverse mutations. Some of these appear to negatively affect myelopoiesis or cardiovascular disease (Jaiswal et al., 2014) (Fig. 4). Thus, many lifestyle features, from diet to sleep to ambient temperature and aging, seem to act on atherosclerosis by regulating the bone marrow output of myeloid cells.

Another major class of systemic risk in cardiovascular disease may be unrelated to direct effects on myeloid cell output from the bone marrow – autoimmunity (Figure 4). It is striking that a number of diseases classified as autoimmune diseases carry substantially elevated risk of cardiovascular events and mortality. Chief among them is lupus with an elevated risk of atherosclerosis, where the presence autoantibodies in the disease increases the odds of cardiovascular events by more than 5 times (Ballocca et al., 2015). Well documented, increased risk is also observed in rheumatoid arthritis (Young et al., 2006) and psoriasis (Hu and Lan, 2017). More recently, data pointing to elevated cardiovascular risk has emerged in the inflammatory bowel disease, Crohn’s disease (Aniwan et al., 2018; Wu et al., 2017) and type I diabetes (Lee et al., 2015; Subramanian and Hirsch, 2018). If the reports in Crohn’s disease hold up in further investigation, the finding will be especially interesting, because Crohn’s disease is associated with reductions in plasma LDL-cholesterol (Aarestrup et al., 2018; Levy et al., 2000; Romanato et al., 2009), thus highlighting the possibility that increased cardiovascular disease can be sufficiently driven by inflammatory triggers like cytokines to overcome, or counterbalance, the protective effect that lowered LDL would confer.

Among all the diseases listed in the paragraphs above, only type I diabetes has a defined antigen discovered to date, with autoreactive T cells directed against specific insulin peptides, leading to T cell-mediated destruction of pancreatic islet β cells (Wan et al., 2018). One model related to the antigenic targets in autoimmunity that may be especially relevant in lupus is that human endogenous retroviruses, which are ubiquitous and active in the genome, elicit autoimmune T cells (Suntsova et al., 2015; Tokuyama et al., 2018). Recent data suggest that lupus patients overexpressed certain endogenous retroviral products due to loss of negative regulators that would otherwise suppressed expression; this was also true in mice (Treger et al., 2019). Perhaps with fundamental understanding of most autoimmune diseases still under investigation, it is not surprising that the mechanistic basis of heightened cardiovascular disease risk is scarcely understood. However, given that the consequence of elevated cardiovascular risk is decreased overall mortality, it is quite important that clinical care of autoimmune patients take elevated cardiovascular risk into account.

A reasonably straightforward concept that might account for heightened cardiovascular disease in autoimmunity is that circulating cytokines produced at the site of autoimmune reactions are made in enough abundance to spill into the circulation, and then act on distal arteries prone to atherosclerosis to promote vascular cell activation and dysfunction. We call this hypothesis cytokine spillover. To take this hypothesis into account, we will focus on recent efforts in the literature to understand the cardiovascular comorbidity in psoriasis, as now cardiovascular drivers in this autoimmune disease, while still primitive, are a bit better understood in psoriasis than the others. First, it is quite clear that there are systemic vascular changes in psoriasis. Recent transcriptional profiling shows evidence of venous endothelial activation to express adhesion molecules and chemokines, and expression of IL-1β mRNA and other cytokines by circulating leukocytes (Garshick et al., 2019). Circulating TNF and IL-17 have been documented in psoriasis (Hu and Lan, 2017). Drugs that neutralize TNF not only reduce the skin disease in some patients, but also effective at reducing cardiovascular events, as shown in numerous studies (Hu and Lan, 2017). One might assume that anti-TNF blockade acts by quelling the action of its target in the systemic circulation, which has a number of redundant effects with IL-1 on endothelium and other cellular targets.

However, there is some cause for questioning the cytokine spillover model. First, the extent of organ damage in lupus, a change that would be expected to correlate with the magnitude of cytokine release, is a lesser risk factor for cardiovascular events than the presence of autoantibodies (Ballocca et al., 2015) (Fig. 4). Furthermore, a particular observation arising in human subjects and mouse models of psoriasis related to the use of IL-23 targeted therapeutics is hard to explain in the context of the cytokine spillover concept. The cytokine IL-23 plays a role in the induction of Th17 immunity, although putative nonpathogenic Th17 cells can develop in the absence of IL-23 signaling (Meyer Zu Horste et al., 2016). Both IL-23 and IL-17 appear to have strong roles in psoriatic lesions in man (Nestle et al., 2009). This insight led to the more recent development of therapeutics targeting IL-23 and IL-17 after anti-TNF was already in clinical use. Therapeutics targeting IL-23 are effective at reducing skin lesion manifestations, and thus would certainly have been expected to reduce cytokine spillover. However, such treatment has not prevented enhanced cardiovascular disease in several studies (Hu and Lan, 2017), including a recent observational study with a prospective design that examined plaque burden and characteristics over a 1-year endpoint in psoriasis subjects using plaque imaging (Elnabawi et al., 2019).

One might argue that these studies suggest that the IL-23 – IL-17 circuit itself is not relevant to cardiovascular disease burden, in contrast to the effects of blocking broadly pro-inflammatory cytokines like IL-1 and TNF. However, in an exciting additional arm of the recent plaque imaging study, the impact of IL-17 blockade was able, in contrast to the therapeutic targeting IL-23, to not only prevent plaque expansion in the 1-year analysis, but disease burden actually declined (Elnabawi et al., 2019). Thus, it may be the case that IL-17, even if not IL-23, is a viable target to reduce cardiovascular risk in psoriasis. The mechanism of action remains unclear. In a mouse model of psoriasis initiated by the use of the TLR7 agonist imiquimod, we have recently reported that chronic experimental psoriasis leads to changes in the extracellular matrix that are dependent upon IL-17 but, like the clinical studies, not prevented by targeting IL-23 p40 (Huang et al., 2019). We argue that the changes in extracellular matrix have the capacity to promote lipoprotein retention and vascular stiffness, such that IL-17 might be linked to both enhanced atherosclerosis and hypertension. In other models of hypertension in mice, IL-17 has been established to be a critical player and may act by targeting the matrix-producing capacity of vascular smooth muscle cells (Wu et al., 2014). What remains rather unclear in these model systems is whether IL-17 is produced within the artery wall by T cells recruited therein (the models are dependent on T cells (Guzik et al., 2007; Huang et al., 2019) or if systemically circulating IL-17 was operative. Given the argument that multiple autoimmune diseases converge on common features like a role for nucleic acid sensors (Theofilopoulos et al., 2017), including TLR7, it seems possible that the data in the context of psoriasis will be relevant to other autoimmune diseases beyond psoriasis. IL-17 is elevated in rheumatoid arthritis (Kotake et al., 1999) (Figure 4) and lupus (Speeckaert et al., 2016), even though lupus is particularly associated with a strong type I interferon signature (Obermoser and Pascual, 2010).

Might neutralizing IL-17 might be an attractive approach to treating atherosclerosis in the general population, beyond individuals affected by autoimmunity? There is reason to consider that the answer may be indeed be yes. First, returning to the theme of myeloid cell supply from the bone marrow, the effector functions of IL-17 include regulation of the recruitment of neutrophils to some infections (Korn et al., 2009). Connected to this effector role, IL-17 promotes granulopoiesis (Forlow et al., 2001; Liu et al., 2010; Parsa et al., 2016; Stark et al., 2005; von Vietinghoff and Ley, 2009) and, thus, here we return to the theme of systemic mediators that affect output from the bone marrow. Second, most studies examining deficiency or blockade of IL-17 in mouse models of atherosclerosis are associated with reduced disease progression (Erbel et al., 2014; Erbel et al., 2009; Madhur et al., 2011; Smith et al., 2010), including models combining atherosclerosis with renal impairment (Ge et al., 2013; Nordlohne et al., 2018), which may be relevant to renal complications in lupus. On the other hand, the role of IL-17 in promoting collagen deposition not only might exacerbate disease by trapping lipoprotein (Huang et al., 2019), it may also promote the favorable deposition of collagen that creates the protective fibrous cap at the atherosclerotic lesion lumen (Gistera et al., 2013). These data collectively indicate that IL-17 may have some protective roles in atherosclerosis like favoring development of the fibrous cap, but overall it appears to be primarily linked to detrimental effects that drive disease progression.

We have implied that lessons from the links between autoimmunity and cardiovascular disease might yield insight into approaches to treat the inflammatory aspect of atherosclerosis in the general population. However, we would be remiss not to mention that there is reason for caution in this line of logic, related to similar logic being applied in the recent past to the use of methotrexate as a treatment to combat inflammation in atherosclerosis. Methotrexate has long been considered an anti-inflammatory intervention used clinically. Although the antiinflammatory impact of the drug methotrexate seems complex and not fully understood, it is thought to act through modulation of folate metabolism (Brown et al., 2016). Used widely and effectively in rheumatoid arthritis, methotrexate lowers accompanying cardiovascular risk (Roubille et al., 2015). It is also effective at reducing cardiovascular risk in psoriasis (Hu and Lan, 2017). These favorable outcomes served as part of the motivation for a second trial, the Cardiovascular Inflammation Reduction Trial (CIRT), that was developed side-by-side with the CANTOS trial discussed above. Unfortunately, methotrexate failed to reduce serum inflammatory markers IL-6, IL-1β, and CRP, and no significant changes in cardiovascular events during the study period were observed (Ridker et al., 2019). These findings raise strong caution about the possibility that effective targets to treat cardiovascular manifestations in autoimmunity can be extended to the general population. However, a major caveat with methotrexate is its nonclassical mechanism of action with respect to inflammation, as it may simply be unable to effect anti-inflammatory shifts in the general population.

Concluding remarks

In this review, we covered emerging concepts in the basic understanding of atherosclerosis and turned to consideration of the role that cytokines play in the disease and whether targeting cytokines has therapeutic potential in the general population or in patients with autoimmunity that carry heightened cardiovascular risk. We underscore that, even though the disease characteristics are evolving in humans as treatments improve, mouse models of atherosclerosis remain fundamental tools that have great utility in understanding the disease and evaluating targets for new therapy. A major point of emphasis is our recent understanding that there are two types of macrophages in plaques, foamy and nonfoamy macrophages, that we propose work in synergy to drive disease progression. This synergy likely revolves around the generation of crystalline cholesterol to spark IL-1β expression and activation in the nonfoamy population, with the foamy cells facilitating, likely through death in the plaque and impaired efferocytosis, the availability of extracellular cholesterol in the plaque. Fitting with this concept is the success of anti-IL1β as a treatment target. However, it remains unclear if the treatment was effective because it acted within plaques or acted systemically. Indeed, whether and how systemic cytokines distal to plaque might affect disease, even in conditions like autoimmunity, remain to be clearly delineated. What is very clear is that some modifiers of cardiovascular risk target production of myeloid cells in the bone marrow, cells that may go on to enter plaque and promote its progression. Indeed, a wide variety of life style modifiers act by affecting myelopoiesis in the bone marrow.

Novel therapeutics to neutralize cardiovascular risk are especially needed in populations at higher risk, such as those with autoimmunity, but more information on mechanisms underlying the cardiovascular comorbidity is needed. Overall, large reductions in risk can be achieved by changes in lifestyle (here we highlighted diet (Estruch et al., 2018), with mechanisms of action just now coming into focus (eg., (McAlpine et al., 2019)) and much more work needed. In a closing remark that converges science with modern society, the New York Times Magazine recently reported on how those in our society most at risk for cardiovascular disease are in the bottom earning bracket, working perhaps two full-time minimum wage jobs such that substantial risk factors accumulate like loss of sleep, depression and isolation, lack of time for exercise, consumption of inexpensive but unhealthy food, and the desire to smoke tobacco as a stimulus to get through it all (Dollars on the Margin, Matthew Desmond). Ongoing studies suggest that an increase in minimum wage reduces psychosocial stress and promotes healthier lifestyle related to most of these factors, including the decision and ability to quit smoking (Du and Leigh, 2015). Effective strategies to combat cardiovascular disease are, thus, to be found both inside and outside of the laboratory.