About mitochondrial fusion and fission-related proteins were first discovered in yeast (107), drosophila (108), etc. Later, homologous proteins were found in humans and/or other mammals. MFN1 and MFN2 involved in OMM fusion in mammals are homologies of fuzzy onions (Fzo) (64). Similarly, Ugo1 is a OMM protein that coordinates the fusion events of OMM and IMM, but no obvious homologues of Ugo1 have been found in higher eukaryotes (109). The discovery of SLC25A46, a member of the mitochondrial solute carrier family 14 (SLC25), which matched Ugo1 in yeast (Schizosaccharomyces japonicus), confirmed that the OMM and IMM can fuse independently in mammalian cells (65). SLC25A46 may be recruited to the OMM as a pre-fission factor and interact with the IMM remodeling protein mitofilin (Fcj1), the IMMT homolog of mammal (65). Knocking down SLC25A46 causes delay in mitochondrial fission or elongation (65), reduced expression of SLC25A46 increases the stability of MFN1 and MFN2 (110), which both leading to mitochondrial hyperfusion. Interestingly, SLC25A46 mutation also results in hyperfusion of mitochondria, but with down-regulated OXPHOS, because excessive fusion seriously damages the subcellular structure of mitochondria, showing that the cristae structure is destroyed or even disappeared (111).

Mitochondrial carrier homolog 2 (MTCH2, SLC25A50), located at OMM, is a direct regulator of mitochondrial fusion/elongation in mouse embryonic fibroblasts (MEFs) and embryonic stem cells (ESCs) (66). As an effective binding partner of BID, MTCH2/MIMP (Met-induced mitochondrial protein) interacts with truncated BID (tBID) at OMM, promoting tBID translocation to mitochondria and ultimately inducing OMM permeabilization (MOMP) to accelerate apoptosis (67, 68).

Similar to Ftsz of prokaryotic (69), MSTO1 (Misato homolog 1), mainly located to the cytoplasm, is a protein involved in the fusion and network formation of OMM (69). Currently, MSTO1 is only known to be related to the fusion of OMM. It is worth exploring how MSTO1 participates in the fusion process, whether there is structural interaction with OMM, or whether MSTO1 promotes the mitochondrial fusion by enhancing the chemotaxis of other factors.

Intriguingly, remolding membrane phospholipid content may affect mitochondrial fusion. For example, mitochondrial phospholipase D (MitoPLD) belongs to Phospholipase D (PLD) signaling enzyme superfamily. A series of metabolic cascades catalyzed by MitoPLD results in mitochondrial tubular morphology, that is, promote mitochondrial fusion. Concretely, due to the special N-terminal domain, MitoPLD is anchored on OMM, which can hydrolyze cardiolipin (CL) on the surface of mitochondria to generate phospholipid acid (PA) (70). At certain points in time when MFNs tethered mitochondria together, MitoPLD modifies the lipid surface of the OM of the opposite mitochondria in a PA-dependent manner to perform the subsequent steps of mitochondrial fusion (70). Later studies found that PA absorption phosphatase (Lipin 1β) produced by MitoPLD hydrolysis converts PA into diacylglycerol (DAG) and terminates the lipid-signaling pathway of browning activated by PA (71). The catalytic domain of Lipin 1β locates at the tip of small mitochondria, or in a spot-like manner on tubular mitochondria, converting long mitochondrial tubules into medium-size fragments (71). Lipin 1β facilitates mitochondrial fission (71), and likewise, skeletal muscle of Lipin-1-deficient patients and extensor digitorum longus muscle of Lipin-1-deficient fatty liver dystrophy mice have been observed with mitochondrial aggregation or large shape (112–114). In conclusion, MitoPLD promotes mitochondrial fusion by lowering Lipin 1β through PA generation. On the other hand, PA generated by the MitoPLD is converted to DAG, which eventually leads to the increase of DAG. An animal study revealed that DAG probably reduces body fat mass by stimulating thermogenesis in brown adipose tissue (BAT) while inducing lipolysis in white adipose tissue (WAT) (115). From a more macroscopic perspective, MitoPLD shows that there is a correlation between cell metabolism and mitochondrial dynamics.

Moreover, mammalian FAM73A (Miga1) and FAM73B (Miga2) were discovered as orthologs of drosophila mitoguardin (miga) gene (72). As membrane integrins on OMM, MIGA1 and MIGA2 promote the formation of MitoPLD dimers and inhibit its degradation (72). Therefore, miga acts downstream of MitoPLD, and both miga and MitoPLD act downstream of MFNs to promote mitochondrial fusion. In addition to the effect on mitochondrial morphology, miga is also involved in the regulation of ΔΨ(m), mitochondrial ATP and ROS production (72). Therefore, miga may be a pleiotropic factor that affects mitochondrial morphological dynamics and cell metabolism.

Mgm1, the first member of the Dynamin family identified in yeast, is required for mitochondrial fusion of IMM and is also associated with the structure maintenance of crista (116), and the human homolog of Mgm1 is OPA1 (73). OPA1 is a dynamin-related GTPase that plays an important role in promoting fusion of IMM, maintaining cristae structure, promoting IMM structure and integrity (74) and controlling apoptosis (117, 118). OPA1 exists in two forms, namely, membrane-bound Long-OPA1 (L-OPA1) and Short-OPA1 (S-OPA1), the hydrolysate of L-OPA1 limited to the membrane space. Mitochondrial fusion is believed to depend on the presence of both L- and S-OPA1 (119), and various stress conditions destroy these complexes and trigger the complete transformation of L-OPA1 into S-OPA1, thus inhibiting mitochondrial fusion (119, 120). In recent years, it has been found that the L-OPA1 is sufficient to mediate mitochondrial fusion and has anti-apoptotic properties (121). Therefore, the morphology of OPA1 has a great influence on mitochondrial morphology.

The proteolytic sites S1 and S2 of OPA1 are encoded by exons 5 and 5B, respectively (122). The IM peptidase OMA1 and i-AAA protease YME1L cut OPA1 separately at S1 and S2, resulting in the aggregation or transformation of L-/S-OPA1 and interfering with mitochondrial fusion. The OMA1 cleaves L-OPA1 in S1 to generate S-OPA1 forms c and e (75). In Oma1 knockout cells, fusion ability is preserved in the presence of L-OPA1, while the c and e forms of S-OPA1 are blocked, thus maintaining the tubular mitochondrial network and protecting cells from apoptosis (121). L-OPA1 protein in Yme1l knockout cells is equivalent to that in WT cells, but S-OPA1 accumulates due to OMA1 hydrolysis (121). This unbalanced processing of OPA1 results in the mitochondrial fragmentation, which also indicates that S-OPA1 may be involved in mitochondrial fission. According to the current conclusions, L-OPA1 is beneficial for promoting mitochondrial fusion, while S-OPA1 appears to be detrimental to mitochondrial fusion. However, S-OPA1, despite its lack of fusion function, has recently been shown to maintain cristae and mitochondrial energy through its GTPase activity (123), to some extent facilitating cell survival under stress.

Intriguingly, IM fusion has a different pattern from OM fusion. This is mainly reflected in the fact that OM fusion requires MFNs to appear on two opposite membranes, while for IM fusion, OPA1 regulates the fusion in an asymmetric way, which occurs as long as the unilateral IM contains OPA1 (124). Speaking of asymmetric regulation of OPA1, we have to mention a negatively charged acidic phospholipid CL. CL is enriched in IMM, and as long as there is sufficient concentration (15%-20%) on the other side of the mitochondria, L-OPA1 and CL pair to induce membrane-tethering and further induce mitochondrial fusion (125). S-OPA1 promotes the bridging between L-OPA1 and CL (125). However, fusion does not occur when the CL concentrations are low on the opposite or both sides of OPA1 (125). At this time, the interrelationship of homotypic trans-OPA1 mediates tethering of IM, which is conducive to the formation and maintenance of cristae (125). CL is re-assigned from IMM to OMM during the induction of damage signals such as mitochondrial damage and depolarization (126). In aging mitochondria, CL is oxidized by cyt c peroxidase and the content of CL decreases (127). Therefore, this regulatory mode of OPA1- CL^high bridging contributes to the specificity of mitochondrial fusion and may be involved in preventing the fusion of healthy mitochondria with damaged/senescent mitochondria and improving the efficiency of cells. However, the mechanism by which OPA1 combined with CL to regulate mitochondrial fusion remains unclear and needs to be explored experimentally.

In addition, CL is required for optimal activity of a variety of mitochondrial carrier proteins. For instance, CL endows ADP/ATP carrier (AAC) 2 with optimal activity by promoting the stable association between AAC2 and respiratory supercomplexs, and reduces the travel distance of cyt c between ETC complex III and IV, contributing to the efficiency of OXPHOS (128). AACs are responsible for the exchange of ADP/ATP, specifically the transfer of ATP formed by OXPHOS from IMM to the intermembrane space (IMS) (129). In addition to regulating ATP production (coupled energy conversion) in mitochondria, AAC is shown to mediate the uncoupled energy conversion (130). Therefore, in the process of regulating mitochondrial fusion with OPA1, CL is likely to associate mitochondrial dynamics with cell metabolism by influencing AACs.

ARL2 (ADP-ribosylation factor (ARF) like 2) exists in the IMS and is uniformly distributed along the mitochondria in a spot-like pattern with MFN1/2 (76). The dominant negative mutation of ARL2[T30N] results in a decrease in mitochondrial fusion rate (24h) and a loss of specific mitochondrial motility (30h), and the dominant activating mutant ARL2[Q70L] leads to a significant increase in mitochondrial tubular structure (48h), independent of its effect on microtubules (76). Although there are some differences in the dynamics of cellular effects, as indicated activation time in parentheses, ARL2 generally promotes mitochondrial fusion during constitutively activation, and its reversal of mitochondrial fragmentation in the absence of fusion factors requires the presence of OPA1 and at least one MFN (76).

Richard A. et al. conducted the above systematic studies on ARL2 also found ELMOD2 (ELMO domain containing 2), which is essential for ARL2 to exert its active functions (131). ELMOD2 is an ARL2 GTPase-activating protein (GAP) that acts downstream of ARL2, and its influence on mitochondrial morphology is independent of GAP activity (77). In MEFs with MFN1 or MFN2 deficiency, the effect of ELMOD2 on the reverse of mitochondrial fragmentation is to a lower degree than that of ARL2, and ELMOD2 and ARL2 have no significant effect in MFN1 and MFN2 double-knockout cells (78). This also suggests that ARL2 and ELMOD2 both act upstream of MFNs, and ARL2 activation without ELMOD2 cannot induce mitochondrial elongation, so ELMOD2 is an effector necessary for ARL2 to promote mitochondrial elongation and tubular structure maintenance (77). Interestingly, ELMOD2 and ARL2, as well as MFN1, MFN2, and MitoPLD, are all located in a regular spot-like discrete distribution along the mitochondria forming bands (77), suggesting that they may form some special complexes in the mitochondria to function together, or there are unknown binding domains on the mitochondria, which recruits these factors to alter mitochondrial dynamics.

In the process of mitochondrial fusion, various elements of OM, IM, and IMS work together to change the dynamics of mitochondria, forming a complex and precise regulatory network ( Table 1 .). Among them, CL and DAG are involved in the metabolic regulation of OXPHOS and lipid metabolism, respectively, suggesting that mitochondrial dynamics is related to cell metabolism.

Compared with wild type (WT) cells, FAM73b KO BMDMs show a severe decline in the basal OXPHOS level and the mitochondria are in a state of fragmentation (26). However, there is almost no difference in ECAR between WT and FAM73b KO BMDMs after LPS (0.5 µg/mL) stimulation and the difference in OCR are converged (26). These results suggest that FAM73b KO has little effect on the glycolysis ability of macrophages, and affects OXPHOS mainly in the resting state. Therefore, we speculate that mitochondrial dynamics affect macrophage metabolism, but are biased and depend on specific cell state.

In bone marrow mesenchymal stem cells (BMSCs), inhibition of miR (microRNA)-150 mitigates LPS (0.5 µg/mL)-induced apoptosis in RAW264.7 (84), suggesting that the metabolic status of RAW264.7 is changed. The co-culture of BMSC also reduces ROS production of RAW264.7 after LPS stimulation (84). Mitochondrial respiratory chain is the main source of intracellular ROS, and the antioxidant system of mitochondria is responsible for removing ROS (132), so mitochondria play an important role in this process. As expected, the inhibition of miR-150 leads to the expression of MTCH2 (84), which is related to mitochondrial fusion (66). In addition, it is indicated that increased MTCH2 promotes mitochondrial transfer from BMSCs to RAW264.7 (84), so miR-150 may regulate the metabolism of RAW264.7 by affecting the morphology of transferred mitochondria. Unfortunately, this research did not cover studies related to mitochondrial dynamics, so this conclusion needs to be validated experimentally. On the one hand, miR-150 has been shown to negatively regulate brown-like transformation of white adipose tissue in inguinal peritoneum of mice, so regulates metabolism (133). In AML (acute myelocytic leukemia) cells, MTCH2 deletion has been shown to reduce glucose entry into the mitochondrial TCA cycle, thereby inducing glutamine to maintain the TCA cycle in the form of oxaloacetate, as well as reduced mitochondrial pyruvate levels and increased nuclear pyruvate and pyruvate dehydrogenase (PDH) levels (97). Further, in the cell nucleus, PDH converts pyruvate into acetyl-CoA and increases histone acetylation. This series of metabolic changes promotes the differentiation of AML (97). Therefore, we hypothesized that MTCH2, which affected mitochondrial fusion, is transferred from BMSCs to RAW264.7, which may promote TCA circulation by affecting macrophage metabolism, thus enhancing OXPHOS and alleviating LPS-induced apoptosis.

Cardiomyocytes-specific deletion of Yme1l (cYKO) in mice inhibits the formation of S-OPA1 form d, and activated OMA1 accelerates the proteolysis of OPA1, leading to the accumulation of S-OPA1 form c and e, triggering mitochondrial fragmentation and leading to dilated cardiomyopathy and heart failure (98). The cYKO shifts the cardiomyocytes from lipid metabolism to the glucose metabolism, and this metabolic shift leads to heart failure in cYKO mice (98). However, double knockout mice (additional OMA1-deficient in Yme1l ^-/- mice) show normal cardiac function by preventing OPA1 cleavage (98). The additional muscle specific-knockout of Yme1l maintains cardiac function without restoring the mitochondrial morphological defects of cYKO cardiomyocytes (98). Interestingly, the loss of Yme1l in skeletal muscle systematically impairs glucose homeostasis, inhibits insulin secretion, and thus reduces cardiac glucose uptake, alters cardiac metabolism, and slows down cellular involuntary metabolism, demonstrating the physiological importance of tissue crossover disturbance (98). Metabolic intervention in cYKO mice fed a high-fat diet replicates the protective effect of hmYKO (heart and muscle-specific deletion of Yme1l) mice (98). In conclusion, coordinated mitochondrial dynamics or balanced mitochondrial fusion and fission plays a key role in cardiac metabolism and cardiac function. YME1L and OMA1 are still lacking in research in macrophages, but they affect mitochondrial dynamics by shearing L-OPA1 to produce different forms of S-OPA1, which may affect macrophage metabolism in this process, which also provides new ideas for studying the macrophage metabolism regulation.

When mitochondria are exposed to selective stress in cells, mitochondrial hyperfusion occurs prior to mitochondrial fission (79). In this process, mitochondrial IM protein SLP-2 cooperates with L-OPA1 and MFN1 to induce hyperfusion, which in turn promotes the production of OXPHOS and ATP (79).

Mitochondrial Sirtuins (mtSIRT), including SIRT3, SIRT4 and SIRT5, are important parts of the sirtuin family (SIRT1-7) and possess enzyme activity such as NAD⁺-dependent deacylases and ADP-ribotransferase, which endowing mtSIRT with the function of regulating energy metabolism (134).

On the one hand, SIRT3 promotes mitochondrial function by regulating the activity of metabolic enzymes, such as by binding to NDUFA9 subunits of ETC complexes I and deacetylation to maintain OXPHOS (99). SIRT3, on the other hand, directly binds to OPA1 at its lysine site (K926 and/or K931) for deacetylation, and enhances mitochondrial respiratory complex activity (80). Since mitochondrial morphology depends on the integrity of the coiled-coil domain at C-terminal of OPA1 (81), where these two acetylated lysine residues of OPA1 located, therefore, SIRT3 deacetylation promotes mitochondrial fusion (80). According to the TEM results, in SIRT3-KO heart, mitochondria are mostly clustered together, forming multiple island-like shapes, and the OMM is continuous, but the IMM fusion is defective (80), which also is consist with the function of OPA1 responsible for IMM fusion and reminds us that OMM fusion can occur independently of IMM fusion.

Down-regulation of SIRT4 in aging chondrocytes was accompanied by excessive fragmentation of mitochondrial network, elevated ROS levels, and impaired ΔΨ(m) (135). Overexpression of SIRT4 saved mitochondrial fragmentation and restored mitochondrial homeostasis in aging chondrocytes (135). Therefore, SIRT4 has the potential to promote mitochondrial fusion. In addition to reported dehydrogenase activity (136, 137), in fibroblast models, SIRT4 expression inhibits mitochondrial respiration under basal conditions, resulting in reduced electron transport system capacity (ETS) and ΔΨ(m) (82). The up-regulated SIRT4 is accompanied by a higher level of mitochondrial fusion regulatory factor L-OPA1 expression, resulting in a higher degree of mitochondrial elongation/fusion and inhibiting mitochondrial fission and autophagy (82), and this phenomenon can also be repeated by upregulation of SIRT4 by inhibiting miR-15b (82). Meanwhile, the overexpression of SIRT4 in Muller glial cells increased the L-OPA/S-OPA1 ratio and MFN2 (138); SIRT4 positively regulates OPA1 and MFN1 in skeletal muscle (139); In mammary epithelial cells, SIRT4 deletion reduces MFN1/2 and increases DRP1 and Fis1 (140). Taken together, these evidences support the role of SIRT4 in promoting mitochondrial fusion.

SIRT5 has deacetylase, deglutarylase and desuccinylase activities and participate in a variety of intracellular metabolic activities, such as the TCA, ETC, glycolysis, fatty acid β-oxidation and other processes (141, 142). For tumor cells, SIRT5 supports cell transformation and promotes breast carcinoma proliferation and tumorigenesis (143). Therefore, functional evaluation of SIRT5 should refer to its specific cell environment. Glutamine is converted into glutamate and ammonia in mitochondria by glutaminase (GLS), and glutamate is converted into α-ketoglutarate (α-KG) through glutamate dehydrogenase 1 (GLUD1) to participate in TCA (144). SIRT5 inhibits glutamine metabolism and regulates ammonia production by desuccinylation of GLS at Lys245 and/or Lys32, thus affecting autophagy (83). However, SIRT3 but not SIRT5 is co-immunoprecipitation with GLUD1 (83). After SIRT5 silencing or inhibition (by MC3482), autophagy is promoted, the expressions of MFN2 and OPA1 are decreased, and mitochondria are short-round shapes that cluster around the nucleus (83). In contrast, SIRT5 overexpression leads to the increase of MFN2 or OPA1, which is conducive to mitochondrial fusion, with increased number of mitochondria, elongated morphology, high ΔΨ(m), and reduced lactate production (83). STIR5, which regulates glutamine homeostasis, is also involved in mitochondrial dynamics regulation, suggesting there is a correlation between cellular metabolism and mitochondrial dynamics.

The relationship between SIRT3/4/5 and macrophage metabolism has been emphasized in recent years. Generally speaking, SIRT3 promotes OXPHOS and FAO (fat acid oxidation), SIRT4 promotes glycolytic anabolism, and SIRT5 changes the metabolic state of macrophages by regulating PTM of key metabolic enzymes (100, 101, 145, 146). For example, in BMDM, SIRT3 deacetylates at Lys83 to activate PDHA1 (pyruvate dehydrogenase E1 α), then PDHA1 catalyzes the decarboxylation of pyruvate into acetyl-CoA and promotes OXPHOS, inhibits glycolysis, which helpful for reducing inflammation of macrophages (100). Interestingly, the ability of SIRT3 to promote OXPHOS and support FAO may lead to immune tolerance in monocytes (147). During the immune tolerance process, the increased SIRT4 controls the expression of PDK1 and SIRT1, thereby transforming the FAO pathway into glucose oxidation, promoting pyruvate to enter glycolysis, which is conducive to restoring immune homeostasis (145). The expression of SIRT4 is reduced in hepatocellular carcinoma (HCC) tumor tissues (101). By increasing the expression of lipid catabolisc genes, such as MCAD (medium chain acyl-CoA dehydrogenase), CPT1 (carnitine palmitoyl transferase 1), etc., SIRT4 knockdown activates the FAO-PPARδ (peroxisome proliferator-activated receptor-δ)-STAT3 signal pathway and promotes the M2 markers (CD206, Arg-1) expression of TAM (101). This has also proved that SIRT4 is not conducive to FAO metabolic pathway from the opposite side. Moreover, SIRT4 silencing increases the expression of mitochondrial genes, including PGC1α, EER-α, CytC, CoxV (mitochondrial cytochrome oxidase V) and IDH3α (isocitrate dehydrogenase 3 alpha) in TAMs (101). PKM2 (M2-type pyruvate kinase), the physiological substrate of SIRT5, is a key determinant of the aerobic glycolytic transition of LPS-activated macrophages (102). In LPS (0.1 µg/mL)-stimulated BMDM, SIRT5-knockdown -regulated hypersuccinylation inhibits the pyruvate kinase activity of PKM2, promotes PKM2 dimerization entering into the nucleus, and ultimately mediates the up-regulation of IL-1β (102). Similarly, SIRT5 promotes inflammation of primary macrophages by promoting p65 acetylation (146). SIRT5 mediates the deglutarylation of GLUD1 in IL-4-polarized BMDM, thereby enhancing GLUD1 enzyme activity and promoting the production of αKG, which has a positive effect on the polarization of M2-like macrophages (148). It has also been shown that after the inhibition of SIRT5 by glutamine, PDH activity was enhanced due to the inhibition of PDHA1 desuccinylation, which promoted the polarization of IL-4 treated BMDM, and the glutamine metabolite α-KG played a role in the inhibition of SIRT5 (149).This seems to be the opposite of the previous sentence in which SIRT5 promotes α-KG production, possible causes include different disease models (eosinophilic chronic rhinosinusitis or burn sepsis) or different amounts of IL-4 stimulation (10 ng/mL or 50 ng/mL), but the specific reasons remain to be further explored (148, 149).

Therefore, mtSIRT is an important link between cellular metabolism and mitochondrial dynamics. SIRT3/4/5 promotes mitochondrial fusion by affecting OPA1 or/and MFN2, which produces different regulation on cell metabolism. SIRT3 promotes mitochondrial respiration, while SIRT4 and SIRT5 inhibit mitochondrial respiration in fibroblasts, MDA-ME-231 or C2C21. As for macrophages, SIRT3/4/5 affect macrophage metabolism in OXPHOS, glycolysis and FAO, illustrating that mtSIRT reconciles the metabolic state of macrophages to meet their demands. Unfortunately, these researches do not cover the change of mitochondrial dynamics under the remolding of macrophage metabolic state. Thence, there are complex regulatory mechanisms for mtSIRT to reconcile cellular metabolism and mitochondrial dynamics, and more researches are urgently needed to unearth their connection.

In conclusion, mitochondrial fusion is driven by multifactor and has undergone a biological process from OM fusion to IM fusion. In addition to mitochondrial fusion-related proteins regulating its morphology, ΔΨ(m), cristae formation, etc., some proteins have been confirmed to be involved in regulating macrophage metabolism ( Table 2 ). However, there are few and scattered articles relating mitochondrial dynamics to macrophage metabolism, but we still conclude with a gratifying possibility. Whether there is a direct connection between mitochondrial fusion and macrophage metabolism and whether there is a logical relationship such as primary and secondary or causality is worthy of in-depth exploration.