The Role of Efferocytosis in Atherosclerosis

What is efferocytosis?

Even in health, the human body turns over more than one million cells per second through a process known as programmed cell death (PrCD), or apoptosis^{1, 2}. This process occurs for a variety of physiological reasons (e.g. negative T cell selection), but is largely driven by the regular homeostatic turnover of aged and senescent cells³. To remove these superfluous cells, the body engages in an evolutionarily conserved process known as programmed cell removal (PrCR), or efferocytosis^{4, 5}. Efferocytosis is a term derived from the Greek (meaning to carry the dead to the grave), and refers to the phagocytic engulfment of a cellular corpse^{4, 6, 7}. Both professional phagocytes (e.g. macrophages) and non-professional (e.g. neighboring) cells participate in efferocytosis.

Because the body must carefully ensure that only apoptotic cells are removed, and that there is no off-target clearance of healthy tissue, efferocytosis is a highly regulated process. Dozens of molecules have now been implicated in the cross-talk which occurs between a dying cell and a potential phagocyte^{4, 8, 9}. These include: 1. Chemoattractant ‘find me’ ligands which recruit phagocytes to the area of cell death; 2. ‘Bridging molecules’ that play an opsonin-like role while linking phagocytes to their target cells; and 3. Cell-surface ‘eat me’ ligands which physically engage and transactivate engulfment receptors on the phagocyte to initiate PrCR. Importantly, these molecules are counterbalanced by so-called ‘don’t eat me’ ligands that are ubiquitously expressed on viable cells, but are rapidly downregulated during PrCD¹⁰. While these signaling pathways remain an area of intense research, it is clear that highly specific regulation of these apparently redundant ‘eat me’ and ‘don’t eat me’ molecules ultimately determines whether a cell is viewed as ‘inedible’, and thus ignored by a phagocytic cell, or if it will be marked for engulfment and clearance from the body¹.

Evidence that efferocytosis is impaired in atherosclerosis

An interesting aspect of efferocytosis is the speed and efficiency with which the process occurs^{9, 11}. It has been estimated that under almost all conditions, cells undergoing apoptosis are recognized and targeted for clearance in a matter of minutes. Indeed, these cells are cleared so efficiently that it is nearly impossible to identify TUNEL-positive apoptotic bodies (A.B.’s) in any healthy tissue bed¹², including organs with very high rates of cellular turnover such as the thymus, tonsil and bone marrow. In those rare instances when a pathologist is able to document an A.B. in healthy tissue, it is commonly in close proximity to a phagocytic cell. Such co-localization suggests that these apoptotic cells are in the process of being identified for clearance, and likely would not have been observed had the tissue been sampled a few minutes later.

One of the few exceptions to this rule is within the atherosclerotic plaque. Unlike the rest of the body, ‘free’ apoptotic bodies (which are not in close proximity to macrophages or other professional phagocytes) are commonly observed in the growing lesion. Careful quantification of the numbers of ‘free’ vs ‘associated’ A.B.’s has led Martinet and colleagues to conclude that the capacity for efferocytosis is reduced nearly 20-fold in plaque compared to elsewhere in the body¹³. Atheromas also contain far fewer phagocytes that have ingested multiple - as opposed to single - cellular corpses, again suggesting an impairment in efferocytosis¹³. While apoptosis is known to be increased during plaque formation, the capacity of the healthy phagocytes is normally so high that Tabas and colleagues have concluded in their review that the problem is not the consequence of too much cell death, but rather too little cell clearance⁶.

How impaired efferocytosis might promote vascular disease

The accumulation of apoptotic and necrotic debris has long been associated with atherogenesis and plaque vulnerability^14–17. At its most simplistic level, expansion of the necrotic core can be viewed as contributing to plaque expansion, which in turn impinges on luminal flow, thus reducing coronary perfusion. However, impaired efferocytosis likely induces a number of other maladaptive changes beyond this mechanical impediment to blood flow. Many of these may be directly causal for atherosclerosis and plaque vulnerability¹³.

First, it is important to note that every time a phagocyte engulfs a dying cell it effectively doubles its intracellular content¹⁸. Because mammalian cells have difficulty metabolizing cholesterol, efferocytes must initiate reverse cholesterol transport machinery to avoid the intracellular accumulation of the membrane-derived lipids they have just ingested. Normally, externalized phosphatidylserine (PS) present on the surface of apoptotic cells upregulates ABCA1 in macrophages, which promotes the efflux of cholesterol to ApoA1^{18, 19}. Interestingly, necrotic cells (which also express PS) do not elicit a similar response¹⁸, and defective efferocytosis signaling has now been shown to suppress critical reverse cholesterol transport pathways in vascular cells¹⁹. The end result is the formation of foam cells, which are widely recognized as key drivers of atherosclerosis.

In addition to promoting salutary survival mechanisms that prevent the accumulation of potentially toxic cellular contents^{20, 21}, physiologic efferocytosis is also known to directly suppress inflammation. When a macrophage has successfully cleared a dying cell, beneficial factors such as IL-10 and TGF-β are released, presumably to signal that no additional inflammatory cells need to be recruited to the site of injury²². The opposite occurs when a phagocyte is unable to clear an A.B., however, and nonresolving inflammation or autoimmunity may ensue^{19, 23, 24}. Thus, failed efferocytosis appears to trigger a molecular switch in phagocytic cells, converting them from champions of inflammation resolution to drivers of vascular inflammation.

A final point to consider is that the reason apoptosis is classically considered an ‘immunologically silent’ form of cell death may be because dying cells are normally cleared before they have had the opportunity to undergo secondary necrosis^{4, 22}. When efferocytosis is impaired, however, apoptotic bodies very rapidly experience breakdown of their cell membranes, which leads to the release of previously-sequestered intracellular contents into the interstitium¹⁶. These factors - which include plaque destabilizing proteases¹⁶, cytokines which promote plaque angiogenesis, and thrombogenic tissue factors²⁵ – each could accelerate atherosclerosis and promote lesion vulnerability.

Taken together, defects in efferocytosis could theoretically effect much more than the physical growth of the interlesional necrotic core. Rather, they may also stimulate plaque inflammation (via release of chemokinetic cytokines), foam cell accumulation (via impaired reverse cholesterol transport) and plaque vulnerability (via atherothrombotic changes within the ECM). What has classically been viewed as a simple ‘waste management’ issue may actually play a larger role in the vascular biology of atherosclerosis (Figure 1).

Why is efferocytosis impaired in atherosclerosis?

It is not yet understood why efferocytosis is dysfunctional within the atherosclerotic plaque, when it appears to be so highly preserved elsewhere throughout the healthy human body¹³. Conceptually, the accumulation of apoptotic debris within the plaque must be due to at least one of the following problems, as previously outlined by Thorp and Tabas⁶: 1. Overwhelming apoptosis within the plaque; 2. An impairment in the phagocytic capacity of lesional macrophages; and/or 3. Reduced ‘edibility’ of apoptotic vascular cells.

While programmed cell death is certainly known to be increased during atherosclerosis²⁶, experimental data suggest that the efferocytic system should have sufficient capacity to deal with this increase in apoptosis, at least initially^{2, 8}. For example, studies in which macrophage apoptosis is induced in the beginning stages of lesion formation have found that triggered cell death actually prevents plaque expansion and promotes lesion stabilization^27–29. Similarly, forced SMC apoptosis (accomplished by Bennett and colleagues via cell-specific expression of the human diphtheria toxin receptor), has no deleterious effect in normal vessels, and only becomes harmful when induced in the late stages of plaque development³⁰. Thus, investigators have concluded that overwhelming apoptosis is not the reason uncleared cells accumulate in the diseased vessel wall, at least in the early stages of plaque development.

Conversely, several lines of evidence suggest that phagocytes may become less effective during atherogenesis. For example, macrophage ‘skewing’ toward the proinflammatory ‘M1’ phenotype in the plaque may reduce the number of beneficial ‘M2’ macrophages, thus inducing a relative deficit of those macrophages known to have a high phagocytic capacity^{31, 32}. Additionally, reactive oxygen species such as peroxynitrite present in the plaque may impair the ability of macrophages to phagocytose apoptotic bodies¹³ and dendritic cells are known to also have reduced efferocytic capacity as they mature^{33, 34}. While lipid-laden foam cells have been shown to retain their phagocytic ability at least to some extent^{13, 35, 36}, vascular SMCs, which are known to possess potent ‘non-professional’ efferocytic capabilities, lose this functionality when exposed to oxidized lipids³⁷. Thus, it is highly likely that the number or ratio of ‘competent’ efferocytes is reduced in or near the necrotic core, either through direct loss (e.g. apoptosis) or via the expansion of less effective phagocyte subpopulations (e.g. M1 macrophages and foamy SMCs).

Perhaps most importantly, diseased and dying vascular cells appear to become ‘poor substrates’ which are less ‘appealing’ to phagocytic cells than apoptotic tissue elsewhere in the body. This likely occurs for a number of reasons, including: 1. Genetic changes in subjects predisposed to atherosclerotic disease; 2. Inflammation-dependent modifications to efferocytosis signaling molecules within the plaque; and 3. The downstream sequelae of intralesional oxidized LDL.

First, studies investigating the heritable component of coronary artery disease (CAD) have found that carriers of the top genome-wide association study (GWAS) risk allele^38–40 have reduced intraplaque expression of a key ‘eat me’ ligand known as Calreticulin¹⁹. As a result, these subjects develop larger lesions replete with diseased vascular cells that are presumably resistant to phagocytic clearance. The Calreticulin axis is discussed in further detail below, but these data suggest that impaired efferocytosis may be the mechanism underlying the most important commonly inherited locus for atherosclerosis (the 9p21 locus), and may explain how these variants promote risk independently of all known traditional risk factors (e.g. hypertension, dyslipidemia, diabetes and smoking)⁴¹.

Second, emerging evidence has now linked inflammatory signaling directly to intra-arterial defects in efferocytosis-related ligand expression (Figure 2). For example, TNF-α, a cytokine known to be causal for- and upregulated in- atherosclerosis, triggers expression of a key ‘don’t eat me’ molecule known as CD47 on the surface of apoptotic vascular SMCs⁴². Similarly, inflammation related to toll-like receptor signaling may suppress the expression of pro-efferocytic ligands such as MFGE8⁴³, and inactivating post-translational modifications to other key ‘eat me’ ligands such as MerTK and LRP1 are induced under inflammatory conditions^{6, 8, 44, 45}, as discussed below. Because failed efferocytosis induces the secretion of TNF-α¹⁶, and TNF-α renders cells even more resistant to clearance, it is highly likely that an endless loop is active in the plaque, and that this may explain why efferocytosis is so impaired during atherogenesis. These data directly support the ‘inflammatory hypothesis’ of atherosclerosis⁴⁶.

Finally, the oxidized LDL present within the atherosclerotic plaque may also directly render apoptotic cells inedible¹³. OxLDL induces the generation of autoantibodies which decorate and presumably mask oxidized ‘eat me’ ligands on the surface of dying cells in the lesion^{47, 48}. OxLDL is also known to directly compete with apoptotic bodies for scavenger receptors, making them less likely to be cleared through competitive inhibition²³.

Taken together, we are still learning why PrCR becomes defective during atherogenesis¹³ and with ageing⁴⁹. However, the problem appears to involve multiple pathways, including processes that impair the capacity of the phagocytes present in the plaque and factors that render diseased vascular cells less likely to be cleared. As with most other aspects of vascular biology, genetic and inflammatory factors appear to play central roles in the pathobiology of lesional efferocytosis, and will likely emerge as areas of focus for future translational studies.

Pathways already specifically linked to efferocytosis in vascular disease

Several factors involved in PrCR have now been experimentally linked to apoptotic debris retention or necrotic core expansion in the growing atherosclerotic plaque (Table). These include the bridging molecule, complement C1q⁶², the phagocyte receptor, transglutaminase 2 (TG2)⁶³, and potentially other ‘eat me’ ligands such as lysoPC and Fas^{65, 67}. However, it is likely that some phagocytosis-related molecules have greater relevance to vascular disease than others, and emerging studies have identified a handful of efferocytosis pathways that appear to have a disproportionate impact on atherogenesis.

Opportunities for translation

Efferocytosis has tremendous potential as both a diagnostic and therapeutic target. For example, CD47 is not only upregulated in patients with plaque compared to no plaque, but also in subjects with symptomatic carotid disease (e.g. those with TIA or stroke) compared to those with an asymptomatic stenosis⁴². Given the data that genetic variants may correlate with alterations in efferocytosis gene expression¹⁹, it may be that a biomarker panel that combines genomic risk variants and expression profiling⁷⁶ could have utility in diagnosing individuals at risk for clinical events due to necrotic core expansion and plaque vulnerability.

Additionally, therapies that appear to specifically reactivate efferocytosis are already being tested in humans. In the oncology field, Weissman and colleagues have recently found that a major mechanism by which cancer cells evade the tumoricidal macrophage is by upregulating ‘don’t eat me’ molecules on their surface^{77, 78}. Accordingly, humanized antibodies and decoy molecules have been developed that interrupt these pathways and restore the phagocytosis of malignant cells⁷⁹. Preliminary studies suggest that these therapies may have an acceptable toxicity profile in non-human primates⁸⁰ and phase I cancer studies are already underway (ClinicalTrials.gov: NCT02678338). Assuming that these treatments are found to be safe and effective, there will be a very unusual opportunity to leverage the experience of the immuno-oncology field and rapidly translate these treatments into the cardiovascular realm.

Further, diseases associated with impaired efferocytosis seem particularly susceptible to combination therapies. In cancer models, anti-tumor antibodies such as rituximab synergize with pro-efferocytic therapies to dramatically accelerate tumor clearance^{79, 81}. There is a clear mechanistic justification to consider a similar approach in cardiovascular disease, particularly given the link between vascular inflammation and the imbalance of efferocytosis ligand expression. As described above, levels of functional Mertk⁷², LRP1⁴⁵ and Mfge8⁴³ all appear to be reduced under pro-inflammatory conditions. The hypothesis that anti-inflammatory therapies may restore pro-efferocytic signaling has specifically been tested in the case of CD47, which is now known to be directly downstream of TNF-α. Humans treated with the anti-TNF-α antibodies Infliximab or Etanercept have reduced in vivo expression of CD47, and mice treated with combination anti-TNF-α/anti-CD47 therapy display a modest incremental reduction in atherosclerosis over anti-CD47 alone⁴². Given that patients treated with anti-TNF-α therapy for rheumatological conditions appear to be protected from myocardial infarction and other adverse cardiovascular outcomes⁸², there is a strong rationale for combining directed anti-inflammatory and pro-efferocytic therapies in the treatment of atherosclerosis.

A final consideration is that pro-efferocytic therapies may provide an actionable opportunity for the oft-mentioned concept of precision cardiovascular medicine⁸³. Genetic studies pursuing the heritable component of cardiovascular disease have shown that the majority of genome-wide significant loci confer risk for myocardial infarction independently of all classical risk pathways, suggesting a novel mechanism of action which is not being addressed by lipid lowering or antihypertensive therapies^{38, 84, 85}. Given that the top GWAS locus is now known to be associated with a reduction in intravascular ‘eat me’ ligand expression¹⁹, it is highly likely that these individuals will particularly benefit from genotype-driven, pro-efferocytic therapy, similar to oncology patients who receive tailored chemotherapy directed against their personal cancer mutation. While it is expected that all subjects with atherosclerosis will benefit from pro-efferocytic therapies, pharmacogenomic-based approaches could lead to outsized effects and even induce plaque regression.

Areas for future study

Although much has been learned about the role of efferocytosis in atherosclerosis over the last decade, several aspects of this process are yet to be explored. A better understanding of the nuances surrounding these pathways in health and disease should enhance our fundamental understanding of vascular biology and facilitate the translation of these findings from bench to bedside.

The first is the potential that pro-efferocytic therapies may have unanticipated drawbacks. For example, CD47 blocking antibodies are generally well tolerated, but do induce erythrophagocytosis and anemia under some conditions^{42, 77}. While this toxicity can be ameliorated with dose escalation or reduced-dose combination therapy approaches^{42, 80}, other pro-efferocytic therapies may not be as specific for diseased tissue. Thus, the spatial and temporal changes in ‘eat me’ and ‘don’t eat me’ ligands that must occur in order to render a cell susceptible to phagocytic clearance need to be defined. Likely, the ratio and physical co-localization of these molecules provides an integrated ‘signature’ that determines a cell’s edibility¹⁰. Such information will explain how the body protects against off-target clearance of healthy tissue and should inform the development of pro-efferocytic therapies that do not induce removal of viable cells. Along these lines, the local delivery of pro-efferocytic therapies (e.g. with drug eluting stent⁸⁶ or targeted nanoparticle technology⁸⁷) may prove highly effective, and could have a superior safety profile.

Second, the link between cancer and cardiovascular disease requires further investigation^{19, 42}. The recent finding that a common pathway may underlie growth of both tumors and atheromas raises questions about whether the clonal hypothesis of atherosclerosis needs to be revisited⁸⁸. Cancer stem cells are now known to express high levels of CD47 and they may use this factor to evade immune surveillance^{89, 90}. Whether a similar process occurs on a pre-atherosclerotic clone could be investigated with advanced lineage tracing modalities. Recent studies have suggested there is far more plasticity of vessel wall cells during atherogenesis than previously appreciated^{91, 92}, and it may be that efferocytosis pathways are involved in such cell-fate decision making and expansion of a diseased subset of cells.

Third, the role of the vascular SMC in efferocytosis must be defined. While many of the studies described above focus on macrophage phagocytosis, it is clear that other vascular cells also participate in efferocytosis during plaque development. Elegant in vivo lineage tracing studies have now unequivocally confirmed that SMCs ‘de-differentiate’ and assume a ‘macrophage-like’ phenotype during atherogenesis, based on their ability to upregulate markers previously thought to be ‘macrophage-specific’, such as MAC-2^{91, 92}. While the physiologic properties of these ‘phenotypically-modulated’ cells remains an area of intense investigation, in vitro studies suggest that these cells have reduced phagocytic capacity compared to bona fide bone marrow-derived macrophages⁹³. Because careful quantification of advanced lesions from lineage-traced mice has shown that a significant percentage of lesional ‘macrophages’ are actually of smooth muscle and not myeloid origin⁹², we must determine if their efferocytic program is governed by pathways different from those reported in classical macrophages. If so, different therapies may be required to target specific components of the evolving lesion.

Finally, the intersection with related physiologic pathways and vascular disorders should be pursued. For example, several classical cardiovascular risk factors, including dyslipidemia, smoking and shear stress have been linked to efferocytosis^94–96. Statins and PPAR-γ agonists have been shown to enhance efferocytosis, and experimental data supports the concept that enhancing blood flow or targeting of the angiotensin pathway could ameliorate defects in efferocytic signaling^{31, 95–97}. Exciting new work exploring novel processes related to atherosclerosis including necroptosis (programmed cell necrosis)⁹⁸ and autophagy⁹⁹ has shown that these mechanisms may also regulate efferocytosis, and could become viable theranostic targets, as well. While efferocytosis has been studied in post-infarct cardiomyopathy⁷¹, we do not yet know its role in other vascular diseases such as restenosis, pathologic angiogenesis or aneurysm disease. Given the simultaneous link to atherosclerosis, acute inflammation and clearance of apoptotic cardiomyocytes after myocardial infarction, pro-efferocytic therapies should be considered as a potential therapy for the early window after an acute coronary syndrome.

Conclusions

While the preceding 50 years have been marked by major improvements in survival from cardiovascular disease, many of these trends have begun to plateau or even worsen in recent years¹⁰⁰. Some of these changes may be due to increases in obesity-related conditions such as diabetes, but others are likely due to the diminishing returns being achieved with the addition of new drugs that target risk factors for which effective therapies already exist. The recent discovery that efferocytosis is impaired in vascular disease represents a major advance in our understanding of the root cause of atherosclerosis and why necrotic debris accumulates over time. The advent of pro-efferocytic therapies that allow for the removal of diseased and apoptotic cells could allow for an entirely new platform of treatment that specifically targets the necrotic core, should these agents prove to be specific and safe in ongoing oncology trials.