The integrity of mitochondria may be compromised owing to oxidative stress, starvation, ischemia–hypoxia, and aging (Spees et al., 2006; Gustafsson and Dorn, 2019; Liu et al., 2021), leading to energy exhaustion, ROS overproduction, and Ca²⁺-induced cell apoptosis (Bock and Tait, 2020). Mitophagy is an acute response to stress under changing developmental, bioenergetic, and environmental conditions that enable cells to meet the demands of metabolic reprogramming, mitochondrial quality control, and cell differentiation (Gustafsson and Dorn, 2019). Mitophagy is a process of cargo-specific autophagy that eliminates damaged mitochondria to regulate mitochondrial quality and quantity (Kim et al., 2007). During mitophagy, serine/threonine-protein kinase PINK1 is stabilized on the membranes of unwanted mitochondria for the subsequent recruitment of E3 ubiquitin-protein ligase parkin from the cytoplasm. Unwanted mitochondria then are marked by parkin-mediated ubiquitination in the outer mitochondrial membrane and recognized by autophagosomes (Pankiv et al., 2007; Matsuda et al., 2010; Okatsu et al., 2010). Other mitophagy pathways independent of ubiquitination are mediated by direct interaction between mitophagy receptors, including Bcl-2/adenovirus E1B 19-kDa protein-interacting protein 3-like (BINP3)/NIX, and several autophagosome proteins (Youle and Narendra, 2011; Zhang T. et al., 2016; Koentjoro et al., 2017).

As well as mitophagy, mitochondria constantly undergo changes in their position and morphology to deal with stress and to meet the cell’s demands. Changes in intracellular position are driven by the attachment and movement of mitochondria along dynamic cytoskeletal tracks (Liesa et al., 2009; Rafelski, 2013). Changes in track arrangements, interactions between mitochondria and organelles including the endoplasmic reticulum and plasma membrane, and mitochondrial fission and fusion are the main factors that influence mitochondrial intracellular movements (Chen et al., 2005; Lackner et al., 2013; Rafelski, 2013). Changes in morphology most commonly involve fission and fusion dynamics. Fusion reverses the effects of stress on the cell by allowing functional mitochondria to complement dysfunctional ones, whereas fission can lead to cleansing of daughter mitochondria by mitophagy (Youle and van der Bliek, 2012).

Mitochondria movements are not constrained within cells but also take place between cells. Mitochondrial transfers occur both under physiological conditions, e.g., in tissue homeostasis and stemness maintenance and in pathological conditions such as hypoxia, inflammation, and cancer (Liu et al., 2021). The transferred cargos may contain either healthy or damaged mitochondria. Healthy mitochondria are transferred from donor cells to protect recipient cells from oxidative stress and apoptosis and to enhance their mitochondrial respiration. In the case of stroke, astrocytes release healthy mitochondria that enter neurons to promote ATP production and viability (Islam et al., 2012; Liu et al., 2014; Hsu et al., 2016). Transfers of healthy mitochondria also occur between cancer cells to promote their survival during chemotherapy (Lou et al., 2012; Desir et al., 2016; Zampieri et al., 2021). On the other hand, stressed cells can transfer damaged mitochondria to recipient cells to ease their burden of impaired mitochondria. This occurs between injured retinal ganglion cells and adjacent astrocytes and between acute leukemia T cells and bone-marrow-derived stem cells (Davis et al., 2014; Wang et al., 2018).

Macrophages have two main activation states: M1 and M2 polarization (Figure 2; Locati et al., 2020). M1-polarized macrophages are activated by interferon-γ (IFN-γ), lipopolysaccharide (LPS), granulocyte-macrophage colony-stimulating factor, or tumor necrosis factor (TNF; Biswas and Mantovani, 2010). The toll-like receptor (TLR) for LPS and receptors for cytokines such as IFN-γ are thus activated and induce subsequent expression of transcription factors nuclear factor kappa-B (NF-κB), interferon regulatory factor 3 (IRF-3), and signal transducer and activator of transcription 1 (STAT1; Sengupta et al., 1996; Sica and Bronte, 2007). These transcription factors are transported into the nucleus, where they upregulate genes related to M1-polarized macrophages (Taniguchi et al., 2001; Gao et al., 2018). M1-polarized macrophages exhibit enhanced phagocytosis mediated by increased secretion of pro-inflammatory cytokines and chemotactic factors; thus, they facilitate the removal of non-self components (Sica and Mantovani, 2012) and play important parts in Th1-mediated immune responses (Biswas and Mantovani, 2010). M2-polarized macrophages are stimulated by interleukin-4 (IL-4) or interleukin-10 (IL-10) signaling, which induces signal transducer and activator of transcription 6 (STAT6), interferon regulatory factor 4 (IRF-4), and peroxisome proliferator-activated receptor γ (PPARγ; Odegaard et al., 2007; Czimmerer et al., 2018). M2-polarized macrophages can be further divided into M2a, M2b, and M2c subgroups. The M2a and M2b phenotypes are activated by IL-4 and promote an immune response mediated by Th2 (Stein et al., 1992). By contrast, M2c inhibits the immune response and favors tissue remodeling after activation by IL-10 or glucocorticoids (Curtale et al., 2013, 2017).

Macrophages undergo mitochondria-related metabolic reprogramming during activation (Haschemi et al., 2012; Huang et al., 2014; Jin et al., 2014). M1-polarized macrophages depend mainly on glycolysis as the first line of defense, whereas M2-polarized macrophages largely rely on oxygen consumption by mitochondrial respiration for their long-term functions (Haschemi et al., 2012; Galvan-Pena and O’Neill, 2014). Increased glucose utilization in IL-4-stimulated macrophages requires activation of the mechanistic/mammalian target of rapamycin complex 2 pathway, which operates in parallel with the IL-4Rα-STAT6 pathway to facilitate M2 activation via induction of IRF-4 (Huang et al., 2016). PPARγ-coactivator-1b (PGC-1b) induces mitochondrial biogenesis and is also indispensable for M2 polarization (Vats et al., 2006), and cell-autonomous lysosomal-based lipolysis and fatty-acid oxidation fuel the mitochondrial metabolism to maintain the M2 phenotype (Huang et al., 2014).

In addition to metabolism alteration, mitochondrial damage-associated molecular patterns (DAMPs) such as mtDNA and byproducts of mitochondrial respiration such as ROS have important roles in the initiation and transduction of signals in the immune response, especially in M1 activation (Nakahira et al., 2011; Zhou et al., 2011). The synthesis of mtDNA, which is induced after the engagement of TLRs, is crucial for NACHT and leucine-rich repeat protein 3 (NLRP3) signaling in M1-polarized macrophages; dysregulated NLRP3 inflammasome activity results in uncontrolled inflammation (Zhong et al., 2018). ROS are essential bactericidal components generated primarily via the phagosomal NADPH oxidase machinery by phagocytes including macrophages (Lambeth, 2004; West et al., 2011). ROS promote production of pro-inflammatory cytokines in response to LPS via decreasing the dephosphorylation of mitogen-activated protein kinases (MAPKs) including c-Jun N-terminal kinase, extracellular signal-regulated kinase, and p38 MAPK phosphorylation (Bulua et al., 2011). ROS also contribute to NLRP3 inflammasome activation (Sorbara and Girardin, 2011). Moreover, mitochondrial ROS are critical to the differentiation of M2-polarized macrophages (Angajala et al., 1605). In a study of TAMs, which are similar to M2-polarized macrophages in terms of their pro-angiogenic and immune-suppressive functions, inhibition of superoxide production was shown to specifically block the differentiation of M2 macrophages (Zhang et al., 2013). However, the inhibitory effect of ROS elimination on macrophage differentiation was overcome when macrophages were polarized to the M1 phenotype (Zhang et al., 2013).

Cell–cell communications occur frequently between macrophages and adjacent tissue cells. Macrophages integrate into host tissues; this entails their specialization in response to the local environment (Varol et al., 2015). Local cells imprint the specific functions of macrophages (Varol et al., 2015), as exemplified by AMs, osteoclasts, and microglia. Alveolar epithelial cells are a major source of colony-stimulating factor 2 (CSF-2), which is necessary for the differentiation of AMs (Guilliams et al., 2013). The differentiation and function of osteoclasts are regulated by the balance of receptor activator of NF-κB ligand (RANKL) and osteoprotegerin produced by osteoblasts (Takayanagi et al., 2002; Boyle et al., 2003; Ikebuchi et al., 2018). Molecules of neuronal origin control microglial motility and functions via chemotaxis, neurotransmitters, and purinergic and adenosine signaling pathways (Crain et al., 2009; Mead et al., 2012; Limatola and Ransohoff, 2014).

Cell–cell communications between macrophages and tissue cells also facilitate tissue-specific functions of macrophages and contribute to the development and specific functions of resident tissues (Figure 3). The most direct type of cell–cell communication is based on the prototypical macrophage function, phagocytosis (Nagata, 2018). For example, macrophages in spleen red pulp phagocytose red blood cells (RBCs) to facilitate iron circulation (Rodrigues et al., 2017). Slight modifications of the cell membranes of RBCs, such as those associated with RBC senescence or damage, are sensed by macrophages, which phagocytose such RBCs and return iron to erythroid progenitors (Korolnek and Hamza, 2015). Microglia (macrophages in the central neural system) contribute to neural synapse maturation and brain development by synaptic pruning (Paolicelli et al., 2011).

In addition to phagocytosis, other forms of cell–cell communication exist between macrophages and their surrounding cells. Microglia can also release signaling molecules including brain-derived neurotrophic factor and microvesicles (MVs) containing cytosolic proteins, lipids, and microRNAs to regulate synaptic activity (Parkhurst et al., 2013; Maas et al., 2017). Macrophages in adipose tissues regulate lipid metabolism and insulin sensitivity in adipocytes via paracrine effects of noradrenalin (Pirzgalska et al., 2017; Flaherty et al., 2019). Perivascular macrophages in capillaries attenuate vascular endothelial-cadherin phosphorylation in endothelial cells to limit blood vessel permeability and maintain vascular integrity (He et al., 2016; Lapenna et al., 2018). BMMs develop into osteoclasts, where they coordinate with osteoblasts for bone modeling and remodeling through both cell contacts and ligand–receptor interactions (Boyle et al., 2003). The best-studied such mutual interactions are RANKL signaling in osteoclasts and its reverse signaling in osteoblasts, which maintains the balance between osteoclast maturation and function (Takayanagi et al., 2002; Ikebuchi et al., 2018). After osteoclast maturation, the proton pump in osteoclasts acidifies the resorption organelle and releases lytic enzymes to realize bone resorption (Boyle et al., 2003). In general, macrophages and tissue cells interact both directly and indirectly to maintain physiological functions of tissues.

Damaged mitochondria are generally degraded by mitophagy, a cell-autonomous activity (Davis and Marsh-Armstrong, 2014). In the case of ganglion cells and astrocytes, transcellular mitophagy (hereafter called transmitophagy) has been proposed as a means of extracellular degradation that lightens the burden of stressed cells (Davis et al., 2014). Recently, different groups of macrophages have been reported as potential handlers of damaged mitochondria during transmitophagy. Accumulating evidence also indicates that the mitophagy level inside macrophages influences macrophage polarization and function (Table 3; Kim et al., 2016; Larson-Casey et al., 2016; Zhao et al., 2017; Bhatia et al., 2019; Patoli et al., 2020; Zhang et al., 2020). Thus, transmitophagy may exert influences on both donor cells and recipient macrophages to regulate tissue homeostasis.

As mitochondria are centers of energy metabolism and coordinate aerobic respiration fueled by either glucose or lipids (Schon et al., 2012), mitochondrial transfers involving macrophages are not only used for transmitophagy but also meet metabolic ends, as exemplified by adipocytes in white adipose tissue (WAT) and adipose tissue macrophages (ATMs). In a recent study focusing on mitochondrial transfers between adipocytes in WAT and ATMs, nearly half of the ATMs in mice internalized mitochondria from neighboring adipocytes under physiological conditions (Brestoff et al., 2020). The significance of these mitochondrial transfers was confirmed in a mouse model of obesity fed a high-fat diet, where adipocyte-to-macrophage mitochondrial transfers were drastically decreased (Brestoff et al., 2020).

Adipose tissue macrophages regulate the glucose utilization and energy expenditure of adipocytes (Pirzgalska et al., 2017; Flaherty et al., 2019), and obesity induces a phenotype switch in ATMs (Lumeng et al., 2007). In turn, changes in ATM phenotypes contribute to WAT inflammation and obesity-induced insulin resistance (Han et al., 2013; Kratz et al., 2014). Indeed, decreased adipocyte-to-macrophage mitochondrial transfers are related to an obesity-induced inflammatory state in WAT. First, the pro-inflammatory environment induced by IFN-γ, LPS activation, and M1 polarization contributes to a macrophage-intrinsic impairment in mitochondrial uptake (Brestoff et al., 2020). Second, the uptake ability of adipose-resident macrophages depends on the heparan sulfate biosynthesis pathway, which has been reported to show anti-inflammatory effects (Brestoff et al., 2020). Inhibition of heparan sulfate biosynthesis leads to aberrant mitochondrial uptake, accompanied by decreased energy expenditure and fat accumulation in adipose tissue (Brestoff et al., 2020).

Intriguingly, gene enrichment analysis of mitochondria-recipient macrophages has defined a transcriptionally distinct subgroup of macrophages that are capable of mitochondrial internalization, with traits resembling those of anti-inflammatory macrophages (Brestoff et al., 2020). More investigations are needed to trace the derivation and characteristics of the transcriptionally distinct macrophage subgroups, as well as their response to the metabolic state in obesity, in order to understand the complexity of ATMs. Moreover, as evidence suggests that ATMs are highly plastic according to their surrounding environment, and obesity is strongly associated with the number, derivation, and functional changes of ATMs (Lumeng et al., 2008; Xu et al., 2013), regulation of ATM subgroups toward mitochondrial uptake subgroups is a promising therapeutic approach for obesity and related metabolic syndromes.

In addition to the favorable effects on surrounding cells of mitochondrial transfers from macrophages, macrophages themselves are influenced by transferred mitochondria. BMSCs modulate the function of AMs by transferring mitochondria via both contact-dependent and paracrine routes; this modulation can be exploited as a mechanism for BMSC therapy for acute lung injury (Jackson et al., 2016; Morrison et al., 2017). In Escherichia coli-treated co-culture system, MSCs were shown to transfer mitochondria to human macrophages via tunneling nanotubes (TNTs); this enhanced their phagocytic capacity and facilitated the antimicrobial effects of the BMSCs (Jackson et al., 2016). In a later study using a transwell system for macrophage and BMSC co-culture, BMSCs significantly increased the proportion of M2-polarized and phagocytic macrophages via transferring BMSC-derived extracellular vesicles (EVs) containing healthy mitochondria (Morrison et al., 2017).

In murine models, adoptive transfer of murine AMs treated with MSC-derived EVs protects mice from LPS-induced lung injury by alleviating the inflammatory cell recruitment, suggesting that anti-inflammatory M2 polarization occurs when AMs receive healthy mitochondria (Morrison et al., 2017). Importantly, BMSC-conditioned medium taken from rhodamine-6G-pretreated MSCs with dysfunctional mitochondria could not cause such changes, indicating the presence of functional mitochondria rather than mitochondrial components that induce such activation changes (Morrison et al., 2017).

Instead of mitochondria-derived immune signals, an increased energy supply from functional mitochondria is responsible for enhanced phagocytosis in AMs; the ATPase inhibitor oligomycin completely reversed the effect of MSC-conditioned medium on BMM phagocytosis (Morrison et al., 2017). However, AMs are a heterogeneous group of cells, especially in acute lung injury, which is characterized by acute immune response and subsequent tissue repair (Duan et al., 2012; Short et al., 2014). Whether self-replicative tissue-resident AMs or monocyte-derived AMs recruited during the acute immune response enable mitochondrial recipients to undergo M2 polarization and change of function remains unclear. In addition, mitochondrial transfers enhance the function of recipient macrophages by improving their mitochondrial bioenergetics (Phinney et al., 2015). Unhealthy mitochondria extruded by BMSCs still exhibit residual membrane potential, which provides evidence for their mitochondrial membrane integrity and fusion ability (Phinney et al., 2015). Depolarized but not totally fragmented mitochondria undergo mitochondrial fusion in macrophages for reutilization (Phinney et al., 2015). In addition to mitochondria, these vesicles contain miR451, miR1202, miR630, and miR638, which represses TLR expression, thereby tolerizing macrophages to mitochondrial-transfer-induced inflammation caused by excessive mtDNA (Phinney et al., 2015).