1. Introduction

Most cells in the human body die quietly by apoptosis, a crucial step in the natural order of tissue development, homeostasis, and immune surveillance. The ingestion and clearance of dying apoptotic cells, known as efferocytosis (from the Latin effere, to carry to the grave), is more than just a critical housekeeping process, but a central immune regulatory axis. Efferocytosis is a distinct form of phagocytosis, a discrimination which hinges on the dominant, apoptosis-programmed “eat-me” signal phosphatidylserine (PtdSer), a membrane phospholipid confined to the inner leaflet in healthy viable cells and externalized upon apoptosis. PtdSer can be recognized by efferocytes, most commonly macrophages, which engulf the apoptotic cells and digest the biomass of the corpse. When clearance fails and apoptotic cells progress to secondary necrosis, persistent corpses signal local efferocytic insufficiency, recruiting circulating myeloid reinforcements. Yet, if rapid and silent clearance cannot be restored, cell death becomes a disease driver, where inflammatory modalities combine with damage-associated molecular patterns (DAMPs)^[1] to shift protective alarm signals into chronic pathological signaling.

Efferocytosis is generally associated with the classical binary framework of “eat-me” and “don’t-eat-me” signals. While necessary, this logic presents insufficient architecture. Macrophages express a strikingly diverse repertoire of efferocytosis receptors (e.g., MER receptor tyrosine kinase (MERTK), AXL receptor tyrosine kinase (AXL), T-cell immunoglobulin and mucin-domain containing protein (TIM)-4, integrins, scavenger receptors) and soluble bridging molecules (e.g., GAS6, PROS1, MFG-E8, C1q). Whether this abundance reflects redundancy, hierarchy, or cooperative amplification across engulfment rounds remains unresolved. It may be better to understand efferocytic macrophages as signal hubs that simultaneously coordinate cytoskeletal remodeling, metabolic reprogramming, and transcriptional responses than as simple engulfment triggers.

Additionally, the temporal dynamics of efferocytosis remain incompletely defined, particularly when a single macrophage engulfs multiple apoptotic targets in succession. Across tissues and species, efferocytosis appears disproportionately executed by subsets of macrophages, characterized by unequal corpse burdens and divergent levels of engagement^[2-5]. Live tracking of apoptosis and phagosomal acidification revealed that some macrophages remain non-efferocytic, while others can repeatedly accumulate large corpse burdens, with no evident feedback mechanism imposing an upper limit on continued uptake^[2]. Each round of uptake may further alter the macrophage’s identity, and the logic governing the uptake of the first corpse may not govern the fifth.

Macrophage efferocytosis is perhaps better understood as a state of being rather than a discrete event. Classifying macrophages by polarization labels in vitro (e.g., M1, M2, or cytokine-defined surrogates), without accounting for what they ingest, how much, and how often, predictably yields contradictory conclusions in inflammation, fibrosis, and cancer in vivo^[6,7]. Under sustained clearance demand, the distinction between homeostatic and pathogenic efferocytosis becomes increasingly blurred, arguing for a framework that defines efferocytic macrophages as a coherent functional state shaped by continual corpse processing. We propose that this macrophage state can build upon the cardinal points described by Blériot et al.^[6], with four efferocytosis-defined coordinates (Figure 1). These variables are regulated by permissive (gas) and restraining (brake) signals on efferocytic activity, akin to shifting gears in a manual transmission rather than a simple on/off switch. Together, these coordinates and their regulators determine whether a macrophage ingests an occasional corpse (if any) or sustains the continuous clearance cycles required for homeostatic maintenance and inflammatory resolution. This review covers efferocytosis across its pan-tissue landscape, the receptor-ligand systems that govern it, and its role in health and disease, while proposing a multidimensional framework for defining the efferocytic macrophage state and the computational outlook needed to resolve it in complex datasets.

Display Full Size

Download

Figure 1. An expanded cardinal framework of macrophage identity incorporating efferocytic states. Macrophage identity encompasses a cardinal framework^[6], to define a coherent functional state (A); Efferocytosis-defined coordinates consist of the following: WHAT is being eaten, referring to distinct dying cell types; WHO is eating, relative to tissue origin, ontogeny, and baseline activation state; HOW OFTEN do macrophages eat, illustrating numerical and serial efferocytosis; and WHERE does clearance occur, depicting a macrophage navigating inflammatory and resolving signals within the same niche (B); The central macrophage integrates input from these overlaid axes into its metabolic state, with functional outcomes shaped by their cumulative and interactive demands rather than by any single variable in isolation (C).

2. A Pan-Tissue Tour of Homeostatic Efferocytosis

Daily cellular turnover in humans is estimated to reach 1% of our total cells, corresponding to ~300 billion cells^[8-10]. This roughly amounts to a volleyball’s worth of dead cell biomass each month. Despite this continuous burden, healthy tissue histology reveals little to no evidence of dead cells, reflecting the extraordinary efficiency of efferocytosis that is largely handled by highly specialized populations of resident tissue macrophages (rTMs), elaborated below. Since uncleared dead cell biomass disrupts tissue architecture, releases damage signals, and serves as a persistent source of inflammation, corpse removal is essential to maintain tissue and systemic homeostasis.

The body’s highest-turnover tissues include those driven by perpetual regenerative demands (spleen, liver, bone marrow, thymus) and those exposed to the external environment (gut, lung, skin). Short-lived erythrocytes, neutrophils, intestinal and gastric epithelial cells jointly account for nearly 96% of daily cellular turnover^[8]. In the spleen, red pulp macrophages handle the majority of red blood cell clearance^[11], while follicular tingible body macrophages in the white pulp remove waves of dying B cells generated during germinal center responses, thereby clearing potential self-antigens to limit autoimmunity^[12]. The liver serves as a complementary erythrocyte clearance site^[13], where Kupffer cells (KC) are uniquely adapted to consume a large and diverse spread of dying cells. An estimated 700 million hepatocytes are generated in the healthy adult liver every day^[14], implying equivalent ongoing clearance of their aged and apoptotic predecessors. To take up apoptotic hepatocytes, KCs residing within the hepatic sinusoidal vasculature project through the endothelium^[15,16]. The liver’s blood-facing position further renders KCs indispensable regulators of circulating immune cell homeostasis. Together with splenic and bone marrow macrophages, KCs clear tens of billions of short-lived neutrophils daily^[17,18]. KCs in collaboration with liver sinusoidal cells further remove activated or senescent T cells from circulation to promote T cell compartment contraction following immune responses, earning the liver its designation as a “T cell graveyard”^[19,20].

Beyond aged cells, macrophages must clear the apoptotic burden generated by immune selection and hematopoiesis. T cell development in the thymus is a major source of apoptosis, as 95% of progenitors undergo cell death during selection^[21,22]. A small population of resident thymic macrophages is remarkably efficient in preventing this accumulation by removing several dozen corpses per day^[23,24]. To preserve normal maintenance of the bone marrow niche, clearance of apoptotic progenitors is carried out predominantly by rTMs and to a lesser degree by mesenchymal stromal cells (MSCs), thus preserving MSC capacity for osteoblastic differentiation and homeostatic bone remodeling^[25]. Recently, efferocytic macrophages were shown to provide a major source of self-antigen, notably to CD8 T cells, by sampling proteins from healthy living cells at low levels under steady-state conditions, revealing a distinct endocytic trafficking mechanism for presentation of these antigens^[26].

Epithelial cells from mucosal tissues are continuously cleared by intestinal, alveolar, or dermal rTMs in the gastrointestinal tract, lung, and skin, respectively, preventing the buildup of dead cell material and release of inflammatory debris^[27-30]. Constant exposure to these apoptotic cells is considered to imprint an anti-inflammatory program on rTMs^[31,32]. As in the bone marrow, non-professional phagocytes can also contribute to local cell clearance at barrier sites. For example, during hair follicle regeneration, when most follicular stem cells die, a small surviving fraction engulfs and clears multiple corpses before being instructed by local cues to return to their normal maintenance function^[4].

By contrast, tissues with limited regenerative capacity and low cellular turnover, such as the central nervous system and heart, require particularly stringent regulation of efferocytic activity. Carefully coordinated efferocytosis appears most critical in the brain^[33,34], where nearly half of all neurons die and are efficiently cleared by microglia during development^[35-37]. Meningeal macrophage-mediated efferocytosis of apoptotic lymphatic endothelial cells appears required for postnatal refinement of the dural lymphatic vasculature, with disruption linked to social behavior deficits^[38]. In the adult brain, microglia and astrocytes cooperate in corpse clearance, with astrocytes engulfing small apoptotic fragments and microglia removing neuronal soma and larger dendritic structures, together eliminating up to 10 million neuronal corpses daily^[39,40]. Similarly, primitive macrophages in the developing heart remove corpses to allow for proper neuronal innervation and development of the three-dimensional cardiac vasculature^[41-43]. In adults, cardiac rTMs manage low levels of cardiomyocyte turnover but also repurpose efferocytic machinery to fulfill their unique role of removing aged mitochondria-containing exophers released by cardiomyocytes^[44,45]. Thus, macrophages throughout the body precisely tune efferocytosis to local demand, discriminating between viable cells, apoptotic cells, and cellular byproducts.

3. From Sense to Synapse in Efferocytic Recognition

Cycles of corpse detection, tethering, processing, and reset require an arsenal of regulated receptors, bridging molecules, and signaling networks (Figure 2). Accordingly, Table 1 compiles efferocytic receptor-ligand pairs and their documented functional outcomes by cargo in human cells or preclinical models.

Display Full Size

Download

Figure 2. The intertwined choreography of efferocytosis spanning corpse detection, engulfment, and digestion. Efferocytosis proceeds through sequential functional steps to clear apoptotic cells (corpses). Corpse detection is initiated by “find-me” signals to orient macrophages to dying cells. Tether captures the apoptotic target through receptor-ligand interactions with “eat-me” signals, driving cytoskeletal reorganization and phagocytic cup formation, while “don’t-eat-me” signals on healthy bystanders prevent inappropriate engulfment. Digest encompasses phagolysosomal processing of the internalized cargo with metabolic rewiring to reset the macrophage to eat again, closing the loop and enabling serial corpse clearance. TAMs: TYRO3, AXL, MERTK.

Macrophages are hard-wired to interpret cues associated with cell death as signals to migrate, prepare for cytoskeletal rearrangement, and activate metabolically^[46]. Apoptotic cells release extracellular “find-me” factors, including adenosine triphosphate (ATP), sphingosine-1-phosphate (S1P), and C-X3-C motif chemokine ligand 1 (CX3CL1) that are respectively sensed by macrophage surface receptors purinergic receptor P2Y2 (P2Y2), sphingosine-1-phosphate receptor 1 and receptor 2 (S1P1/2), and C-X3-C motif chemokine receptor 1 (CX3CR1), which guide macrophage chemotaxis towards sites of death^[38,47-51].

Once macrophages “follow their nose” and migrate towards the corpse, they tether to and “taste” their presumptive meal, either via direct recognition of PtdSer or via bridging molecules. For example, macrophage-secreted milk fat globule-EGF-factor 8 (MFG-E8) binds to PtdSer and is recognized by integrins αvβ3 and αvβ5 to facilitate homeostatic clearance^[52,53]. In addition, macrophages generally express TYRO3-AXL-MERTK receptor family (TAM) receptors, an acronym derived from the receptors TYRO3 protein tyrosine kinase (TYRO3), AXL, and MERTK which mediate corpse engulfment via the PtdSer-binding bridging molecules growth arrest-specific protein 6 (GAS6) and protein S1 (PROS1)^[54-58]. Upon calcium-dependent recognition of GAS6/PROS1-coated apoptotic cells, TAM receptor tyrosine kinase signaling promotes anti-inflammatory and pro-resolution responses^[57,59,60].

Cell corpses can also be detected via direct PtdSer receptors like TIM-4^[61], TIM-3^[62], TIM-1^[63], BAI1^[64,65], TREM2^[66], Stabilin-2^[67], RAGE^[68], CD300b^[69], CD14^[70,71], and CD36^[72,73], although their expression varies widely among myeloid populations^[7]. Notably, TIM-4 is an rTM-specific PtdSer receptor across the heart, liver, lung, kidney, and brain that is transcriptionally conserved between mice and humans^[74]. While TIM-4 lacks a canonical signaling domain, it appears to be remarkably “sticky” and exhibits specific, high-affinity binding to PtdSer^[61]. In contrast to TIM-4, its family member TIM-3, which is expressed more broadly by leukocytes and some non-immune cells, acts as a context-sensing receptor whose PtdSer-dependent engulfment is actively inhibited by high mobility group box 1 (HMGB1) released from necrotic cells^[64,75].

Following corpse recognition, efferocytosis proceeds through a coordinated cascade of cytoskeletal signaling events that drive physical engulfment of the bound cell. Engagement of PtdSer, either through direct receptor binding or ligand-bridged activation of TAM receptors, coordinates with integrin rearrangement to activate the ELMO-DOCK180 complex, which in turn activates Rac1 and Rho family GTPases^[76,77]. These GTPases promote actin polymerization and membrane protrusion around the apoptotic target, forming the efferocytic cup that progressively closes around the cargo and seals into a phagosome^[78]. While this cytoskeletal machinery must be actively reset after cargo internalization to enable subsequent engulfment rounds, evidence suggests that macrophages engage multiple targets at once^[79,80].

Once internalized via PtdSer receptors, cargo is routed through a liver kinase B1 (LKB1)/AMP-activated protein kinase (AMPK)-dependent autophagy-like pathway that diverts self-peptides from antigen presentation, potentially via Rab17-mediated endosomal recycling, for immunologically silent debris clearance^[81-84]. Additionally, corpses opsonized with complement proteins like C3 or C1q can be recognized by CD209 (DC-SIGN), scavenger receptor class F member 1 (SCARF-1), CD91 (LRP1), or a C1QR-CD93 complex to facilitate immunologically silent clearance^[85-90]. These homeostatic clearance mechanisms stand in contrast to phagocytosis of corpses opsonized with antibodies, where corresponding Fc receptors (CD64 and CD16) drive ITAM-dependent nuclear factor kappa B (NF-κB) signaling and pro-inflammatory activation^[91].

Whether efferocytic receptors function redundantly or as a coordinated network integrating graded and spatiotemporally resolved cues is unclear. While TIM-4 and MERTK appear to act synergistically in tethering and engulfment, respectively^[92,93], how the broader repertoire of receptors spatially organizes, integrates signals at the synapse, and responds to multiple engulfment rounds requires further investigation.

4. Sustaining Continual Efferocytosis–From Engulfment to Reprogramming

Metabolism of corpses should not merely be considered as the final step in the efferocytic cycle, but an ongoing program required for efferocytosis to be sustained over time. Since apoptotic cell production exceeds the number of rTMs available for clearance, macrophages must engage multiple corpses in rapid succession to maintain tissue homeostasis. A single engulfment can double macrophage biomass, a burden amplified during inflammation when multiple corpses may be ingested in less than an hour. Macrophages meet this demand by coupling extracellular PtdSer-directed recognition of dying cells to intracellular sensing of incoming biomass, deploying metabolic programs that sustain serial clearance^[79].

The linking of detection, engulfment, and metabolism is demonstrated by the concurrent nature of these processes. Apoptotic cell-secreted factors further metabolically prime macrophages for efferocytosis, notably by upregulating SGK1, which drives glucose transporter SLC2A1 translocation to the macrophage membrane to pre-emptively fuel the energy demands of engulfment^[94]. Within minutes of corpse detection by TAM receptors, macrophages release paused RNA polymerase II at loci encoding early growth response (EGR) family transcription factors (EGR1, EGR2, EGR3), which drive gene programs associated with cytoskeletal motility and lysosomal function to enable continual efferocytosis^[79]. Internalized apoptotic cells are routed through the LC3-associated phagocytosis (LAP) pathway to promote phagosome maturation and lysosomal fusion, which acidifies the phagosome to enable the enzymatic degradation of cargo and release of signaling metabolites to reprogram macrophage metabolism^[95]. This transition is coordinated by WDFY3, which links actin disassembly to autophagy machinery via interactions with GABARAP family proteins, such as ATG5, and LC3-II^[96]. Accordingly, blocking lysosomal acidification with the vacuolar ATPase inhibitor bafilomycin prevented subsequent corpse uptake by macrophages that had already eaten, but not previously unfed macrophages^[79,80], demonstrating that lysosomal capacity can be rate-limiting for continual efferocytosis.

That macrophage metabolic reprogramming is cargo-driven rather than a generic response to membrane engulfment is supported by the finding that internalization of apoptotic cells expressing PtdSer, but not empty PtdSer-coated beads, upregulated metallothioneins in HoxB8-derived macrophages, consistent with increased intracellular metal burden imposed by ingested cargo^[79]. Furthermore, engulfment of apoptotic cells activates liver X-receptor alpha (LXRα), retinoid X receptor alpha (RXRα), and peroxisome proliferator-activated receptor gamma (PPARγ)-dependent metabolic programs to manage the substantial input of lipid and cholesterol from the ingested corpses^[97-99].

Beyond corpse removal and digestion, TAM-receptor mediated efferocytic rewiring drives macrophage production of remodeling factors such as IL-10, resolvins, oxysterols, TGFβ, and VEGF^[97,100-102], as well as the antioxidant mitochondrial-derived peptide humanin^[103]. In the neonatal heart, an age-dependent efferocytosis-driven metabolic shift stimulates macrophage production of thromboxane A2, which acts as a cardiomyocyte mitogen to promote wound healing after myocardial injury^[34]. Similarly, macrophages exposed to hepatocyte debris, but not synthetic phagocytic substrates (latex beads or liposomes), produce Wnt ligands to promote local tissue remodeling^[104].

The identity of the corpse itself may also influence the functional outcomes of macrophage efferocytosis. Uptake of apoptotic neutrophils enhanced IL-4-induced a “pro-resolving” polarization in mouse bone marrow-derived macrophages (BMDMs), marked by Chil3, Ear2, and PPARγ/LXR targets^[105], while BMDM pre-feeding with apoptotic hepatocytes, but not T cells, induced immune tolerance gene programs upon IL-4 stimulation^[106]. Whether these divergent outcomes reflect differential receptor engagement at the synapse, metabolic sensing of ingested corpses, or both remains unresolved.

In the lung, alveolar macrophage efferocytosis of apoptotic neutrophils promoted continual efferocytosis and pro-resolving metabolic programs via neutrophil-derived myeloperoxidase (MPO) and UCP2-dependent mitochondrial reprogramming, but at the expense of anti-bacterial infection protection, whereas ingestion of MPO-deficient neutrophils or epithelial cells preserved it^[107]. In a complementary study, apoptotic epithelial cells during influenza infection were preferentially taken up by lung interstitial macrophages rather than alveolar macrophages^[108]. These findings underscore that efferocytosis is not intrinsically pro-resolving but is profoundly shaped by macrophage identity and corpse composition, as molecules from the meal can remain bioactive within the efferocytic macrophage. Detailed discussion of post-efferocytosis metabolic and anti-inflammatory programs can be found in recent comprehensive reviews^[46,109,110].

4.1 “Gas-and-brakes” tuning of efferocytic throughput

Macrophages actively adapt their clearance capacity to match steady-state turnover or resolve elevated apoptotic burden during inflammation. Yet efferocytosis cannot proceed unchecked. Here we present a “gas-and-brakes” framework in which permissive signals accelerate engulfment while inhibitory pathways apply restraint.

4.2 Efferocytic “Gas”

To sustain clearance capacity within inflamed tissues, efferocytic signaling is coupled to proliferative/mitogenic, autophagic, and cytoskeletal reset mechanisms. In BMDMs, DNase2a-mediated degradation of apoptotic cell DNA generates free nucleotides that activate DNA-PKcs-mTORC2/Rictor signaling to promote Myc-dependent macrophage proliferation^[111]. This digestion-induced proliferative response cooperates with MERTK-driven survival and mitogenic signaling downstream of corpse recognition to expand the efferocyte pool and promote immune resolution^[109]. Autophagic bottlenecks during efferocytosis can be alleviated by enhancing LC3-mediated degradation^[112], while conversion of apoptotic cell-derived arginine into putrescine maintains Rac1 cytoskeletal activation in BMDMs, thereby resetting macrophages for rapid, successive rounds of corpse engulfment^[113].

To meet such energetic demands, efferocytosis is often supported by oxidative phosphorylation (OXPHOS), as observed in rTMs inhabiting oxygen-rich niches, including liver KCs, alveolar macrophages, and large peritoneal macrophages^[97,114]. Consistent with this context dependence, hypoxia within atherosclerotic plaques suppressed efferocytosis, likely through MERTK downregulation^[115]. Nevertheless, metabolic flexibility enables certain rTMs to sustain efferocytosis under diverse and often energetically constrained conditions. Low-oxygen environments resembling those of the spleen, bone marrow, and thymus were unexpectedly shown to enhance continual efferocytosis in HoxB8-derived, splenic, and thymic macrophages in vitro, as well as apoptotic thymocyte clearance in vivo^[116]. Even under hypoxia, macrophages can adapt by increasing glycolytic flux and remodeling mitochondrial function to preserve ATP production despite reduced reliance on OXPHOS.

4.3 Efferocytic “Brakes”

The metabolic support of efferocytosis, as well as receptor-mediated corpse recognition, is not unidirectional. Rather, there is a parallel set of “brakes” to constrain macrophage efferocytic capacity under certain circumstances at both the corpse and macrophage levels, which are highly tunable through local tissue context and inflammatory cues.

Efferocytic brakes generally seek to reduce corpse recognition and engulfment or impair intracellular degradation to restrict subsequent rounds of eating^[113]. The canonical “don’t-eat-me” signal on healthy cells is CD47, a ligand for signal regulatory protein alpha (SIRP⍺), an inhibitory receptor expressed by macrophages^[117]. Cell-surface expression of CD47 is rapidly lost during apoptosis, but frequently retained in pathological contexts. Inflammatory cell death similarly inhibits efferocytosis, as necroptotic cells are cleared far less efficiently than apoptotic ones^[118]. This efferocytic resistance arises from reinforcing mechanisms of CD47 upregulation, as shown in atherosclerosis and metabolic dysfunction-associated steatohepatitis (MASH, formerly referred to as non-alcoholic steatohepatitis/NASH)^[119,120]. Similarly, CD47 and CD24, another “don’t-eat-me” signal that binds to Siglec-10 on macrophages, are often overexpressed in cancer, suppressing tumor cell phagocytosis by macrophages and enabling immune escape^[121-123].

Efferocytic throughput can also be gated by metabolic constraints imposed by local tissue signals. For example, necroptotic cells release prostanoids such as thromboxane that suppress macrophage OXPHOS and impair the energetic needs of continuous uptake^[124]. Notably, resolvin D1 (RvD1) reverses thromboxane-mediated efferocytic suppression in microglia by restoring OXPHOS through increased glutamine uptake^[124,125], directly linking pro-resolving lipid mediators to metabolic rescue of efferocytic capacity. Inflammatory and pattern-recognition receptor (PRR) signaling can provide an additional layer of metabolic control, tuning efferocytic capacity in a context-dependent manner rather than exerting uniformly pro- or anti-efferocytic effects. While certain pathways such as TLR3 and TLR9 may enhance efferocytosis, TLR2 and TLR4 signaling as well as other canonically inflammatory cues such as LPS or IFNγ polarization may antagonize uptake through metabolic reprogramming that suppresses OXPHOS^{[116,126-128]}.

Efferocytosis is further constrained by inflammatory modulation of surface receptor availability. In models of bacterial infection, cardiac ischemia-reperfusion injury, and atherosclerosis, pro-inflammatory signaling through the p38 MAPK pathway suppresses expression of phagocytic receptors and induces proteases like a disintegrin and metalloproteinase 17 (ADAM17) that cleave membrane-bound MERTK^[129-132]. Consequently, both apoptotic cell uptake and downstream anti-inflammatory signaling were impaired. A related mechanism operates in rheumatoid arthritis, where IL-17 drives JAK-STAT3-dependent upregulation of ADAM17, resulting in MERTK shedding from synovial macrophages and impaired efferocytosis^[133]. Nevertheless, regulated proteolytic control of MERTK may serve physiological roles, such as limiting excessive efferocytosis in vulnerable tissues like the retina, underscoring the context-dependent consequences of receptor shedding^[134]. Macrophage MERTK can also be physically obscured by local neutrophil extracellular traps (NETs) or cleaved by NET-associated enzymes like ADAM17^[135]. Macrophages in turn secrete DNases to control local NET abundance, suggesting that the NET-DNase balance actively tunes efferocytic receptor availability across the course of an immune response from inflammation to resolution^[136].

While corpse recognition depends on the balance of “eat-me” and “don't-eat-me” signals, additional factors shield macrophage access to these cues. Mucins and glycocalyx components such as hyaluronan attach to the extracellular domains of transmembrane receptors such as CD43 and CD44 and physically obscure both “eat-me” PtdSer and “don’t-eat-me” CD47/SIRPα signals through steric and electrostatic inhibition^[137-139]. During apoptosis, cytoskeletal remodeling and membrane retraction release glycocalyx from its membrane anchors, permitting blebbing and PtdSer exposure on apoptotic protrusions that are then recognizable by macrophages^[138]. Glycocalyx, a previously underappreciated regulator of immune cell function, thus emerges as a contextual gatekeeper of efferocytosis. Whether its pathological remodeling in autoimmune type 1 diabetes, cancer, and other disease contexts similarly modulates dead cell clearance rates remains under investigation^[140-142].

Finally, recent work in the context of tissue transplantation has demonstrated that PtdSer-dependent recognition is sufficient to drive macrophage clearance of even viable cells, as elevated PtdSer on live xenogeneic donor cells triggers AXL-mediated engulfment by recipient macrophages in a process known as “phagoptosis”^[143,144]. This process could be inhibited by reversing PtdSer upregulation on donor cells or enforcing CD47-mediated inhibitory signaling. These findings highlight that efferocytic recognition is fundamentally gated by the balance of “eat-me” and “don’t-eat-me” signals, rather than cell death.

5. Dysregulated Efferocytosis as a Disease Driver

5.1 When the dead pile up

Steady state tissue maintenance and effective resolution of acute immune responses rely on timely and efficient efferocytosis, while pathological inflammation can be defined in part by its failure, resulting in corpse accumulation. In these inflammatory settings where cell death increases, efferocytosis becomes a throughput-limited process governed by the balance between dying cell burden and the availability of competent efferocytes. Reflecting this imbalance, standard in vitro efferocytosis assays deliberately model elevated corpse load with apoptotic cell-to-macrophage co-culture ratios of 3:1 to 5:1^[145]. In vivo, chronically inflamed tissues often exhibit substantial loss of rTM numbers, further skewing this balance towards excess corpses^[146]. Although such inflammatory conditions promote robust infiltration of theoretically efferocytosis-competent myeloid cells (e.g., monocyte-derived-macrophages and neutrophils) into tissues, these recruited populations often fail to fully compensate, and corpse burden persists. Accordingly, ongoing efforts are focused on defining the causes of dead cell accumulation and persistence in these contexts, including alterations in macrophage populations or states, receptor expression and signaling, and metabolic constraints that limit sustained efferocytic capacity.

In chronic inflammatory liver diseases such as MASH, extensive hepatocyte apoptosis and necrosis fuel ongoing tissue injury compounded by progressive rTM KC loss^[10,16]. Corpse clearance relies on the KC surface receptor TIM-4, whose partial knockdown by small interfering RNA (siRNA) (~75%) exacerbated corpse accumulation, inflammation, and fibrosis, even in the presence of other efferocytic receptors on KCs and infiltrating myeloid populations^[147]. A complementary constraint on efferocytosis acts at the level of efferocyte processing capacity and survival. In murine models, hyperactivation of the hypoxia-inducible factor 2 alpha (HIF2α)-mechanistic target of rapamycin (mTOR) axis in KCs inhibits the lysosomal regulator TFEB, inducing lysosomal stress that impaired efferocytosis and promoted KC death^[148]. Multifactorial efferocytic defects are also observed in type 2 diabetes (T2D), where monocyte-derived-macrophages isolated from T2D patients exhibit cell-intrinsic defects in efferocytosis efficiency in vitro compared to healthy donors^[149]. This deficiency in efferocytosis in diabetic wound healing may be exacerbated by macrophage-extrinsic factors in vivo, as excessive NET deposition may physically obscure dead cells from recognition by efferocytic receptors and contribute to diabetic wound healing and inflammation^[150]. In the context of atherosclerosis, failure to digest NETs via DNase activity similarly promotes defective corpse clearance within plaques^[136,151]. Together, disruption of efferocyte identity and post-engulfment processing, often compounded by metabolic stress, reduces efferocytic throughput in disease.

Environmental exposures may impose yet another constraint on efferocytosis that contributes to corpse persistence. Polystyrene microplastics (PS-MPs) accumulate broadly across tissues and are particularly enriched in macrophages^[152-154]. Ingestion of PS-MPs suppressed continual efferocytosis in vitro and in vivo by dysregulating macrophage metabolism and elevating the reactive metabolite methylglyoxal (MGO), which impaired phagolysosomal function and corpse degradation, effects that were reversed by Glo1, an MGO-detoxifying enzyme^[155]. These findings raise concern that increasing environmental microplastic exposure may impose a previously unappreciated metabolic brake on efferocytosis with consequences for tissue homeostasis^[156].

5.2 When the meal is pathogenic and efferocytosis maladaptive

Although efferocytosis is considered immunologically silent, its outcomes depend on the inflammatory context. In a murine model of tumor necrosis factor (TNF)-driven sepsis or systemic inflammatory response syndrome (SIRS), efferocytosis of apoptotic neutrophils triggered caspase-8-dependent, inflammasome-independent pyroptosis in macrophages, while TAM receptor engagement by PtdSer-coated beads in the presence of TNF recapitulated this death in vitro^[157]. Notably, TNF pre-treatment protected macrophages against efferocytosis-associated death, while TAM receptor inhibition in mice partially protected against SIRS pathology, demonstrating that inflammatory signals can override homeostatic efferocytic programming to exacerbate disease.

In DAMP-prone inflammatory environments, macrophages repeatedly engulf lipid-rich phagocytes in metabolic dysfunction-associated steatotic liver disease (MASLD, formerly referred to as non-alcoholic fatty liver disease/NAFLD) or “foamy cells” in atherosclerosis. These inherit the engulfed cholesterol burden and become lipid-associated macrophages (LAMs) that can progress to a pro-inflammatory state, promoting NLRP3 inflammasome activation and cell death^[101]. As newly recruited monocytes in turn ingest these cholesterol-laden corpses via TREM2, each successive meal perpetuates the foamy macrophage phenotype, suggesting that strategies to boost efferocytosis in plaques may require concurrent metabolic support to be therapeutic^[158]. However, the TREM2-dependent LAM phenotype was shown to be protective in mouse models of MASLD, as TREM2 deficiency in KCs or recruited monocyte-derived-macrophages impaired clearance of dying cells and exacerbated MASH and/or fibrosis^[159,160]. The yin/yang biology of TREM2^[161,162] thus reflects a balance between acute protective efferocytosis and reinforcement of pathogenic macrophage states, as in cardiovascular and neurodegenerative disease^[163-167] driven by chronic exposure to pathogenic cargo and functions beyond efferocytosis^[163].

The functional outcome of engaging the canonical efferocytosis receptor MERTK is also highly context-dependent^[168]. In experimental models of cardiomyopathy and inflammatory arthritis, reduced MERTK signaling, either by MARCH2-NR1H2 axis deficiency^[169] or increased MERTK cleavage by ADAM17^{[102,130,133]}, inhibited effective cell clearance and promoted inflammation and pathology. However, contrary to non-signaling TIM-4^[147], MERTK was found to promote fibrosis in MASH through ERK and TGFβ signaling independently of corpse clearance, such that MERTK deficiency reduced liver damage, myofibroblast activation, and collagen deposition^[170].

Likewise, efferocytosis has divergent effects in cancer. Macrophage populations within metastatic livers are highly heterogeneous, with subsets exhibiting divergent metabolic and immunoregulatory programs, which may influence the impact of various efferocytic pathways^[171]. In some cases, such as intrahepatic cholangiocarcinoma, blockade of MERTK-efferocytosis can be pathogenic by enabling secondary inflammation, thereby rendering the local environment less permissive to tumor growth^[172]. Conversely, MERTK-mediated efferocytosis has emerged as a critical immunosuppressive node in the tumor microenvironment (TME), as efferocytosis of dying cells suppresses pro-inflammatory and type I interferon responses, reinforcing immune tolerance^[173]. Importantly, these immunoregulatory effects are coupled to broader tissue repair programs, including angiogenesis, which supports tumor growth and limits effective anti-tumor immunity. Consistent with this, hypoxia-induced cell death in the melanoma TME was shown to drive pro-angiogenic programs in efferocytic macrophages, promoting vascularization and tumor progression^[174]. Similarly, in glioblastoma, hypoxia-driven necrosis drove immunosuppressive programs in macrophages fueled by transglutaminase 2 (TGM2)^[175]. TGM2, a cross-linking enzyme that can be induced by both LXR signaling^[176] and engulfment of apoptotic cargo^[175], plays important roles in efferocytosis by stabilizing the phagocytic portal during efferocytosis initiation and acting as a cell-surface coreceptor for MFG-E8^[177,178]. These findings establish a feed-forward axis in which hypoxia drives cell death, efferocytosis, macrophage reprogramming, angiogenesis, and immune suppression^[174,175]. Metabolic programming further stratifies tumor-associated macrophage states, as demonstrated in triple-negative breast cancer, where efferocytosis-driven pro-resolution macrophages and inflammatory macrophages define mutually exclusive but convergent resistance pathways^[179]. In pancreatic ductal adenocarcinoma liver metastasis, lysosomal metabolism is pathologically harnessed to sustain efferocytosis through macrophage-derived progranulin regulating cystic fibrosis transmembrane conductance regulator (CFTR)-dependent lysosomal acidification and cargo degradation^[171]. This enables liver X receptor alpha/retinoid X receptor alpha (LXRα/RXRα)-mediated macrophage polarization and arginase-1 upregulation, while genetic or pharmacological inhibition of progranulin-dependent efferocytosis restores tumor immune surveillance via CD8 T cells and limits liver metastasis^[171].

Defective efferocytosis is one of the markers of autoimmune diseases such as systemic lupus erythematosus^[10,180-184], characterized by humoral responses and the production of auto-antibodies. Aberrant glycosylation may trigger this immunogenic efferocytosis in autoimmunity^[185,186]. Recently, defective N-terminal glycosylation of apoptotic cell-derived RNAs was shown to drive intracellular nucleic acid sensor activation and inflammatory signaling during efferocytosis^[187]. Altered glycosylation further increases the visibility of PtdSer, potentially leading to overactive efferocytosis and macrophage presentation of PtdSer antigens. These findings may provide a link between immunogenic efferocytosis and auto-antibodies against RNA and PtdSer.

6. Computational Gaps in Deconvolving Efferocytosis Events in Big Data

While single-cell sequencing (scRNA-seq) has revolutionized our understanding of macrophages and improved our ability to study efferocytosis, current analysis workflows remain technically limited in resolving engulfment events from technical doublets (Figure 3). Standard single-cell quality-control pipelines identify and discard 10-40% of mixed transcriptomes, routinely excluding doublets and multiplets known to distort clustering, inflate cluster-defining genes, and obscure real biology^[188-190]. Macrophage efferocytosis should represent an important exception to this rule. Mixed transcriptomes may reflect bona fide engulfment events rather than technical artifacts, yet current methods largely fail to distinguish between the two because they rely on transcriptomic nearest-neighbor distances rather than biological provenance^[191].

Display Full Size

Download

Figure 3. Deconvolving doublets to resolve the efferocytic macrophage state. Schematic illustrating biological and technical sources of mixed transcriptomes that can be labeled as “doublets” in computational workflows. Left: a macrophage that has internalized a T cell through efferocytosis. Middle: a macrophage engaged with a T cell via antigen presentation at an immune synapse. Right: a macrophage and a T cell co-encapsulated in a single reverse emulsion droplet. The mixed transcriptomes produced by co-encapsulation and immune synapses can be resolved using current computational tools, while true biological engulfment states remain unresolved.

By comparison, myeloid scRNA-seq datasets are routinely flagged for doublet removal at higher rates than lymphocyte datasets, though this empirically noted difference has not been formally benchmarked in the published literature. While T cells are small (~8 μm), non-adherent, and readily dissociated into clean single-cell suspensions with low aggregation and minimal ambient RNA contamination, macrophages are larger and more heterogeneous in size, measuring ~20 μm at rest and expanding to ≥ 30 μm upon activation or efferocytosis, with harsher dissociation protocols to lift adherent macrophages in vitro, increasing the risk of artifactual co-encapsulation and background RNA release. Whether this differential is real and consistent across tissues and platforms could be settled directly using an atlas that contains both myeloid and lymphoid compartments from the same donor^[192]. In practice, standard scRNA-seq library preparation workflows usually passage cells through 70 µm, 40 µm, or even 30 μm cell strainers prior to microfluidic loading, physically excluding the most enlarged and heavily cargo-loaded efferocytic macrophages before a single read is generated. Depending on the tissue type, single-nucleus RNA-seq (snRNA-seq) circumvents this size constraint because isolated nuclei are uniformly small, regardless of cellular activation state. However, the tradeoff is the loss of cytoplasmic transcripts, where much of the cargo transcriptome, as well as the metabolic and functional signatures of macrophage efferocytosis, resides. Moreover, intact nuclei derived from engulfed cells remain subject to the same doublet detection and removal logic. Determining which nucleus was taken up by which macrophage, and whether we can track this “inheritance”, remains unknown. As a result, no current single-cell platform captures the efferocytic state without systematic loss or misclassification. Given the dynamic nature of the macrophage state, indiscriminate doublet removal may overrepresent efferocytic potential while underrepresenting actual efferocytic events.

An underlying question, rarely addressed in atlasing work, is whether the cargo's transcripts persist long enough to be detected at all. Internalized apoptotic material is routed into an acidifying phagolysosome whose degradative machinery turns over nucleic acids^[75], and tracking studies that exploit species- or cell-specific transcripts have put a timescale on this decay. Using human-specific transcripts (e.g., APOL1) to follow ingested apoptotic-cell RNA, Lantz and colleagues reported that internalized cargo transcripts become largely undetectable by approximately six hours after engulfment^[3]. Any cargo signal must additionally survive the one-to-four hours of dissociation, staining, and loading that precede capture. By contrast, a macrophage's own transcriptional response to a corpse is rapid and durable: engulfment triggers RNA polymerase II pause release and metabolic reprogramming on the order of an hour^[79]. The eater therefore begins rewriting its transcriptome on a timescale faster than the decay of the meal’s transcriptome. This has a direct and underappreciated consequence for experimental design; single-cell datasets sampled at or beyond the ~6 h window will largely have lost the cargo transcripts that would mark a recent engulfment event, retaining instead the macrophage's reaction to the meal. The cargo-transcript signal can be recovered experimentally where the eater and cargo differ in origin, for example through xenogeneic (cross-species) feeding or the use of cell type-specific markers for cargo; but in routine human-tissue atlases, where eater and cargo share a transcriptome and are sampled long after engulfment, such tracers are unavailable^[193]. This limitation reframes the computational objective: rather than relying on the detection of two intact genomes within a single barcode, a more tractable classifier would distinguish macrophages bearing the transcriptional and metabolic aftermath of engulfment from doublets that lack any such coordinated response.

A macrophage that has recently engulfed apoptotic cargo may transiently contain inherent transcripts from the phagocyte and “passenger” transcripts from the engulfed apoptotic body^[3]. Efforts to computationally deconvolve efferocytosis-associated biology have therefore relied on signature-based approaches that score transcriptional states rather than resolve discrete engulfment events, as exemplified in efferocytosis-related gene sets used to stratify prognosis and microenvironmental features at the cohort level in patients with gastric cancer^[194]. What remains missing is a supervised or semi-supervised machine-learning framework trained explicitly to distinguish technical doublets from biologically meaningful eater-cargo composites. Classifying previously discarded macrophage populations with calibrated probabilities should span at least three categories: technical artifact, ambient-contaminated singlet, and true efferocytic composite. Such a model would need to learn the joint structure of efferocytic programs (receptor engagement, LAP-autophagy, lysosomal flux, metabolic adaptation) and cargo-lineage transcript signatures, while simultaneously controlling for ambient RNA and dissociation stress in a discrimination task that current doublet callers were never designed to perform^[195].

Transcriptomics alone is unlikely to train such a model robustly across tissues in health and disease contexts. Protein-level measurements of receptor abundance and cleavage state (e.g., heterogeneity in TIM-4 surface expression^[74] or whether MERTK is intact or shed^[134]) provide direct evidence of engulfment competence that RNA cannot. Spatial transcriptomics and high-resolution imaging can confirm whether a macrophage physically occupied an apoptotic niche, or merely shares a gene program with one, while metabolomic and lipidomic readouts capture cargo burden as an orthogonal axis. A rescue model that integrates these layers would make event-level efferocytosis measurable for the first time in human tissue. Until that framework exists, a two-lane analytical workflow should be adopted: one applying conventional strict quality control for atlas construction and reproducible differential expression, and the second retaining flagged macrophage candidates under explicit uncertainty for event-aware reanalysis in practice, retaining barcodes in an ambiguous doublet-probability window (e.g., 0.25-0.75) only when they co-express canonical efferocytic genes (such as MERTK, SLC2A1, AXL, TIM-4, PPARG) alongside lineage-discordant transcripts, and carrying them forward with confidence intervals rather than hard labels. The goal is not to weaken quality control, but to avoid treating all mixed profiles as expendable, because in the context of efferocytosis, some of these profiles capture the biology of interest.

Finally, modulators of efferocytosis are often evaluated by their effect on the fraction of cells that engulf a corpse; however, some cells rapidly ingest multiple corpses while others remain bystanders within the same population^[2-4]. This heterogeneity in efferocytic activity underscores the need for single-cell methodologies to define what distinguishes “big eaters” from “small eaters” and whether high-throughput clearance arises from feed-forward programs induced by initial engulfment or reflects pre-established efferocytic predispositions. Moreover, rTMs survive weeks to months in tissues and likely clear far more corpses over their lifespan than short-term, single-event assays can capture within a few hours. Continual efferocytosis in vivo thus emerges from the intersection of what is being eaten, who is eating, where clearance occurs, and how often macrophages are challenged. These demands can only be fully captured by studying efferocytosis as a long-term, context-dependent state through advanced computational approaches.

7. Reframing Efferocytosis as a Programmable Disease-Shaping Axis

Foundational studies established apoptotic cell clearance as a central homeostatic and immunoregulatory process rather than a passive garbage disposal function^[10,196-198]. More recent work has substantially expanded this view by demonstrating that efferocytic capacity is dynamically regulated through transcriptional licensing^[96,113], receptor-state biology, metabolic fitness, and cytoskeletal resetting^[96,113]. It is also becoming apparent that the identity or state of the apoptotic cargo shapes macrophage fate or functional responses to eating^[106,107]. As a result, efferocytosis should be considered a flexible, self-renewing macrophage state rather than a discrete engulfment event. Cell type-specific efferocytosis has been shown to drive divergent macrophage states in vivo, including functionally distinct alveolar macrophage subsets determined by the nature of engulfed corpses^[107]. While efferocytosis is generally an adaptive and protective response^[199], pathogenic efferocytic macrophage states can be reinforced by toxic or metabolically burdensome cargo^[164-167] or exploited in cancer to support immunosuppression and tumor progression^{[171,172,175]}.

These complexities raise fundamental questions about therapeutic strategy^[200] (Table 2). Thus far, clinical trials targeting efferocytosis pathways have generally involved either antagonism of TAM or TIM family receptors in an effort to reduce tumor immunosuppression, or targeting the CD47-SIRPα axis to reduce “brakes” on efferocytic clearance, which showed limited benefit in patients^[201,202]. Novel preclinical approaches to “boost” efferocytosis have shown promise in a more diverse array of disease contexts, but how these insights can be translated to the clinic remains a significant challenge. A recent study using a porcine model of atherosclerosis demonstrated that a macrophage-targeted nanoparticle which blocked the CD47-SIRPα signaling axis reduced cell death and inflammation in plaques, notably without inducing the anemia typical of anti-CD47 therapies^[203]. Enhancing lysosomal degradation can be protective in inflammatory disease, as shown by columbamine-mediated rescue of efferocytosis and resolution in colitis^[112]. Another approach demonstrated that efferocytosis could be boosted by chimeric receptors that coupled PtdSer receptors TIM-4 or BAI1 to the cytoplasmic adapter ELMO1, linking apoptotic cell recognition with cytoskeletal motility. Cells expressing these chimeric receptors were shown to attenuate pathology in models of colitis, hepatotoxicity, and nephrotoxicity^[204]. By contrast, MERTK engagement has shown mixed outcomes across disease contexts^[205], while phytochemical modulation of efferocytosis and autophagy exhibits context-dependent efficacy^[206]. Cleavage-resistant TREM2 variants that evade ADAM17-mediated shedding improve debris clearance in some settings^[207], yet sustained TREM2-dependent programs may also reinforce disease-perpetuating macrophage states under chronic exposure to pathogenic cargo. The challenge moving forward is not simply increasing clearance, but selectively reprogramming efferocytes to match tissue, disease stage, and cargo composition.

Addressing this challenge requires a shift in both experimental and analytical frameworks. For clinical relevance, macrophage state will need to be defined across the transcriptome, proteome, morpholome, and secretome, with conservation across species from flies^[2,208] to mice and humans, in addition to ensuring in vitro experiments capture in vivo relevance. A major computational hurdle remains the inability to resolve efferocytic events within high-dimensional datasets, leading to frequent misclassification or exclusion of macrophages actively engaged in clearance. Experimental designs should explicitly control apoptotic cargo identity, uptake timing, and inflammatory milieu, alongside artificial-intelligence and machine-learning approaches capable of simulating therapeutic intervention points without presupposing causality. Rather than a binary function, efferocytosis operates as a long-term, context-dependent program that actively shapes disease trajectories. Across tissues, what a macrophage eats, how frequently it eats, and how efficiently that cargo is processed together emerge as major determinants of downstream transcriptional, metabolic, and immunological outcomes with therapeutic implications.

Authors contribution

Sawden M, Wadsworth II MH, Fabre T, Lebeaupin C: Conceptualization, writing-original draft, writing-review & editing.

Conflicts of interest

All authors are employees of Pfizer, Inc., Ltd. The authors declare no other competing interests.

Ethical approval

Not applicable.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Availability of data and materials

Not applicable.

Funding

None.

Copyright

© The Author(s) 2026.