Mieap forms membrane-less organelles involved in cardiolipin metabolism

Summary Biomolecular condensates (BCs) are formed by proteins with intrinsically disordered regions (IDRs) via liquid-liquid phase separation. Mieap/Spata18, a p53-inducible protein, participates in suppression of colorectal tumors by promoting mitochondrial quality control. However, the regulatory mechanism involved remains unclear. Here, we report that Mieap is an IDR-containing protein that drives formation of BCs involved in cardiolipin metabolism. Mieap BCs specifically phase separate the mitochondrial phospholipid, cardiolipin. Mieap directly binds to cardiolipin in vitro. Lipidomic analysis of cardiolipin suggests that Mieap promotes enzymatic reactions in cardiolipin biosynthesis and remodeling. Accordingly, four cardiolipin biosynthetic enzymes, TAMM41, PGS1, PTPMT1, and CRLS1 and two remodeling enzymes, PLA2G6 and TAZ, are phase-separated by Mieap BCs. Mieap-deficient cells exhibit altered crista structure, leading to decreased respiration activity and ATP production in mitochondria. These results suggest that Mieap may form membrane-less organelles to compartmentalize and facilitate cardiolipin metabolism, thus potentially contributing to mitochondrial quality control.


INTRODUCTION
Biomolecular condensates (BCs) in cells are also known as liquid droplets because of their liquid-like nature.BCs are composed of proteins, nucleic acids, and other macromolecular components, [1][2][3] and they are formed by proteins with intrinsically disordered regions (IDRs) via liquid-liquid phase separation (LLPS).Importantly, BCs function as membrane-less organelles (MLOs) that compartmentalize and facilitate cellular biological reactions.BCs are not surrounded by lipid bilayers, which enable facile exchange of reactants and products with their surroundings, expediting biological reactions in cells.Furthermore, theoretically, BCs are able to appear and disappear within a cell in response to cellular stress and/or subcellular circumstances, exhibiting spatiotemporally dynamic properties.On the basis of these features, the concept of MLOs is fundamentally different from the long-standing concept of membrane-bound organelles, such as nuclei, lysosomes, mitochondria, and endoplasmic reticulum, which are stable structures that segregate biomolecules and biological reactions using lipid bilayer membranes.Multiple MLOs may exist within a cell, possibly regulating cellular biological reactions and activities, including transcriptional regulation, 4 signal transduction, 5,6 immunity, 7 centrosome activity, 8 and mitosis. 9This list is rapidly growing.
Metabolic reactions are believed to be highly organized through spatiotemporal clustering and compartmentalization of sequential enzymes and substrates/intermediates at subcellular sites, which maximizes efficiency of linked reactions. 10Without this mechanism, small, toxic intermediates, derived from metabolic reactions, rapidly diffuse throughout the cytoplasm.Thus, the concept of subcellular compartmentalization of metabolic reactions has been anticipated for a long time.It was initially conceived as a ''metabolon,'' a structural-functional complex of sequential metabolic enzymes and substrates/intermediates. 11The ''metabolon'' concept predicted that sequential enzymes and cellular structural elements form a supramolecular complex for metabolic reactions.3][14][15] Among them, ''purinosomes'' involved in purine metabolism, were the first demonstration of enzyme clustering, revealed by live-cell imaging of six sequential enzymes for de novo purine synthesis. 16Although MLOs that compartmentalize and expedite metabolic reactions in cells could address regulatory mechanisms for sequential metabolic reactions, there has been no clear evidence for metabolic MLOs until now.
Cardiolipin (CL) is an important phospholipid for the following reasons.[1] It is a phospholipid dimer, with two phosphate residues and four fatty acyl chains. 17 [2] Because of this unique structure, it forms a cone shape, contributing to curvature of the lipid membrane and to maintenance of mitochondrial cristae. 18 [3] CL is the only phospholipid that is specific to mitochondria and is mainly located at the inner mitochondrial membrane and contact sites. 18 [4] CL interacts with many mitochondrial membrane proteins, including electron transport chain complexes involved in oxidative phosphorylation, and the ADP/ATP carrier.Interaction with CL promotes activity of these proteins. 18 [5]  CL stabilizes the structural assembly and activity of respiratory super-complexes at the inner mitochondrial membrane. 19 [6] It interacts with cytochrome c at the outer surface of the inner mitochondrial membrane, which supports cellular viability by maintaining stable respiratory charged residues in IDR1 and IDR2, respectively.In contrast, IDR3 and IDR4 formed clusters of positively charged residues, characterized by repeats of R and S (Figures 2B-2E). 36lthough amino acid sequences of the IDRs are evolutionarily divergent compared to the structured regions, the distribution of IDRs and clusters of positively charged residues in IDR3 and IDR4 are evolutionarily conserved (Figures 2C, 2E, and S2). 37In addition, there is an evolutionarily conserved hydropathic character in the Mieap sequence as a whole, the N-terminal half being hydrophilic and the C-terminal half being hydrophobic (Figures 2D, 2E, and S2). 37This implies that Mieap protein may act as a cellular biosurfactant.These molecular features in Mieap IDRs are consistent with the concept of the ''evolutionary signature'' that Zarin et al. previously proposed. 38They found that although the amino acid sequences of IDRs are poorly conserved in alignment, most disordered regions contain multiple molecular features that are preserved as an ''evolutionary signature'', which can be used to predict IDRs from their amino acid sequences in yeast.

Material state and dynamics of Mi-BCs are determined by specific regions of Mieap
To map sites responsible for the physical state and dynamics of Mi-BCs, using a confocal microscope, we examined cells expressing EGFP-Mieap WT (WT) and three deletion-mutant forms, EGFP-Mieap DCC, D275, and D496 (DCC, D275, and D496) (Figure 3A).
To investigate protein dynamics in Mi-BCs, we performed fluorescence recovery after photobleaching (FRAP) studies for WT, DCC, D275, and D496 condensates.During observations up to 60 s after spot-bleaching, with the bleaching depth being 82.2 G 8.1%, fluorescence intensity of WT, DCC, D275, and D496 condensates recovered to 9.9 G 4.0%, 4.0 G 1.0%, 94.3 G 4.0%, and 17.8 G 3.1% of their initial values, respectively (Figures 3B and S3A).Fluorescence recovery of the D275 condensates almost achieved their initial value within the 60-s observation period, indicating the most fluid state.
When we performed less intense laser exposure, which reduced the bleaching depth to 21.2 G 4.4%, fluorescence recovery increased 60 s after spot-bleaching (WT, 50.4 G 8.8%; DCC, 32.6 G 8.8%; D275, 94.9 G 13.7%; D496, 61.5 G 10.1%) (Figure S3B).The fluorescence recovery rate increased when the number of bleached molecules was small, suggesting that availability may be a rate-limiting factor.
We further examined slow fluorescence recovery up to 15 min.WT and D496 condensates showed continuous fluorescence recovery, which reached 80.4 G 10.2% and 94.0 G 9.5% of the initial value within 15 min, respectively (Figure 3C).In contrast, DCC condensates reached equilibrium at 37.4 G 6.3% of their initial value within 15 min (Figure 3C).These data suggested that WT condensates, as well as D496 condensates, consist mainly of mobile materials, but protein availability from their surroundings is limited.Therefore, FRAP analysis data also suggested that Mi-BCs possibly exist in mitochondria.

Mi-BCs phase-separate the mitochondrial phospholipid, CL
To identify molecules targeted for phase separation by Mi-BCs, we screened available fluorescence-tagged mitochondrial proteins and mitochondrial fluorescence probes using confocal live-cell imaging.EGFP-BNIP3, EGFP-NIX, AcGFP1-Mito, DsRed2-Mito, and SYBR Green I (a probe for mitochondrial DNA) were localized at mitochondria, but none of them were incorporated into to Mi-BCs (Figures 3D-3F and Video S3).
However, 10-N-nonyl acridine orange (NAO) was specifically incorporated into Mi-BCs (Figure 3G and Video S4).NAO targets CL. 39 CL binds to >60 mitochondrial proteins via its hydrophobic and electrostatic interactions. 40As described previously, positively charged residues are enriched in Mieap IDRs and Mieap has its C-terminal half hydrophobic region.CL carries two negative charge phosphate residues and four hydrophobic fatty acyl chains. 41Sequence data reveal the amphiphilicity of Mieap, and Mi-BCs consist of an electron-dense phase that is positive for OsO 4 , 42 according to TEM analysis, suggesting that Mi-BCs may contain unsaturated lipids.Therefore, we suspected that CL is a bona-fide target for phase separation by Mi-BCs.
To determine whether Mieap binds directly to CL, we performed a fat blot assay, 43 in which binding of GST-Mieap to CL, phosphatidylcholine (PC), and phosphatidylethanolamine (PE) was evaluated on lipid-dotted membranes.GST-Mieap bound to CL but not to PC or PE (Figure 3H).

Mi-BCs are possible membrane-less organelles involved in CL metabolism
We performed mass spectrometric analyses of CL in A549 cells with and without enforced expression of exogenous Mieap protein by Ad-Mieap infection.The total amount of CL per cell was higher in Ad-Mieap infected cells than in cells without the infection (Figure 4A).Broad CL species showed higher absolute values in cells infected with Ad-Mieap than in non-infected cells (Figure 4B), suggesting the role of Mieap in CL biosynthesis.In contrast, relative amounts of most CL species did not change.However, Mieap significantly increased the proportions of CL72:5, CL72:6, CL70:6, CL68:5, and CL68:6, and decreased the proportions of CL66:3, CL66:4, CL68:2, CL68:3, CL68:4, and CL70:4 (Figure 4B), suggesting the role of Mieap in CL remodeling.Therefore, these results suggest that Mieap is involved in CL metabolism.Therefore, we speculated that Mi-BCs may function as MLOs to compartmentalize and facilitate CL metabolic reactions.To validate this hypothesis, we examined whether Mi-BCs phase-separate enzymes are sequentially involved in CL metabolism (Figure 4C). 20,44,45Thus, we examined involvement of the following EGFP-tagged enzymes required for CL metabolism in Mi-BCs by performing confocal live-cell imaging: EGFP-TAMM41, EGFP-PGS1, EGFP-PTPMT1, and EGFP-CRLS1 (involved in CL biosynthesis); EGFP-PLA2G6 (related to CL hydrolysis by phospholipase A 2 activity); and EGFP-TAZ (involved in CL remodeling). 20,45All of these enzymes localized at mitochondria and were subsequently incorporated into Mi-BCs (Figures 4D-4I and S4A-S4F, and Video S5).Interestingly, all of these CL metabolic enzymes tended to be concentrated in the Mieap-depleted phase in Mi-BCs (Figures 4D-4I and Video S5).
It is important to classify droplet-driver and droplet-client proteins in Mi-BCs.Recently, a bioinformatics-based analysis program (the FuzDrop method) was developed by Fuxreiter's group, 48 which can predict and identify droplet-driver or droplet-client proteins of MLOs.Utilizing this program, we examined if Mieap and all the cardiolipin metabolism-related proteins found in the Mi-BCs (PRELI, TAMM41, PGS1, PTPMT1, CRLS1, PLA2G6, and TAZ) act as drivers or clients.As a result, the program predicted that Mieap acts as a driver of the MLOs and that PRELI, PGS1, CRLS1, PLA2G6, and TAZ act as clients of the MLOs (Figure S5).

Mieap protein is highly concentrated in mitochondrial BCs via its C-terminal hydrophobic region
Since CL biosynthetic enzymes and NAO are located at the inner mitochondrial membrane, we assumed that 3D imaging analysis of the relationship between Mi-BCs and enzymes/NAO would provide the most definitive information on the location of Mi-BCs at mitochondria.
Using fluorescence-tagged CL biosynthetic enzymes, we tried to confirm whether the enlarged Mi-BCs are truly localized within mitochondria by performing live-cell imaging analysis with tomographic 3D reconstruction.As shown in Figure 5A, Videos S6, and S8, enlarged Mi-BCs are fully enveloped by EGFP-TAMM41, signals of which are continuously localized from tubular mitochondria to all around the surfaces of spherical Mi-BCs.Supporting the results of CL biosynthetic enzymes, we further confirmed that NAO is also continuously localized from tubular mitochondria, all round and inside of spherical Mi-BCs (Figure 5B, Videos S7, and S8).
Since CL biosynthetic enzymes tend to localize at the surface of Mi-BCs (Figures 4D-4J), we compared signals of mApple-TOMM20 (the outer mitochondrial membrane) and EGFP-CRLS1 (the inner mitochondrial membrane) around Mi-BCs.As shown in Figures S6A-S6D, the signals of EGFP-CRLS1 were always detected inside the signals of mApple-TOMM20, compared to those of PLD6 (the outer mitochondrial membrane) that are completely consistent with the signals of mApple-TOMM20, suggesting that Mi-BCs are present within mitochondria, possibly facing the inner mitochondrial membrane at their surfaces.Similar results were obtained from the same analysis using another probe for EGFP-TAMM41 (Figure S6E).
Similarly, using EGFP-TAMM41 and NAO as mitochondrial probes, we determined whether DCC, D275, and D496 BCs are located inside mitochondria.As shown in Figures 5C, 5E, Videos S6, and S8, both DCC and D496 BCs are also fully enveloped by EGFP-TAMM41, whose signals are continuously derived from tubular mitochondria.Moreover, DCC and D496 BCs are also stained by NAO as a clear picture of tubular mitochondria and each condensate (Figures 5F, 5H, Videos S7, and S8).On the other hand, D275 condensates are never related to the signals of EGFP-TAMM41 or NAO (Figures 5D, 5G, and Videos S6-S8).These results suggest that DCC and D496 BCs are located inside mitochondria, whereas D275 BCs are present outside mitochondria.
We further examined partitioning behaviors of WT, DCC, D275, and D496 proteins by performing analysis of Intensity Ratios (IRs) of each protein in BCs and cytoplasm. 49As shown in Figure 5I, IR values (condensates/cytoplasm) of WT, DCC, and D496 are 158.62G 40.74, 178.81 G 34.07, 8.29 G 1.92, and 153.14 G 25.66 (mean G SD), respectively.This implies that IR of WT, DCC, and D496 are more than 18 times higher than that of D275.All WT, DCC, and D496 BCs are localized within mitochondria, whereas D275 condensates are present in cytoplasm.Therefore, these IR results suggest that Mieap protein tends to be highly partitioned and concentrated in mitochondrial BCs, and that this propensity is possibly determined by the C-terminal hydrophobic region of Mieap, which could mediate interaction of Mieap with CL/CL-related phospholipids.

Both N-terminal hydrophilic and C-terminal hydrophobic regions are required to generate multi-phase structure of Mi-BCs
To explore the mechanism responsible for multi-phase structure in Mi-BCs, using EGFP-TAMM41, we examined whether CL metabolic enzymes are phase-separated by DCC, D275, and D496 condensates (Figures 6A-6D).As expected, D275 condensates did not phase-separate EGFP-TAMM41 (Figure 6C).Importantly, although both DCC and D496 are located in mitochondria, EGFP-TAMM41 was phase-separated and incorporated in the Mieap-depleted phase of D496 condensates (Figure 6D), whereas EGFP-TAMM41 was mainly localized across the surfaces of DCC condensates, which did not generate multi-phase structures (Figure 6B).These results suggest that both the N-terminal hydrophilic and C-terminal hydrophobic regions are required to form multi-phase droplets.The region of IDR1-3 and two CCs may be critical to the interaction with CL metabolic enzymes to generate the internal enzyme-containing phase (the Mieap-deficient phase) in Mi-BCs.We further explored the internal structure of the Mieap-depleted phase with EFGP-PGS1 or EGFP-TAMM41.Importantly, we found that CL biosynthetic enzymes, EGFP-PGS1 or EGFP-TAMM41, formed condensates in the Mieap-depleted phase, which wetted the interface between the Mieap-containing phase and the Mieap-depleted phase in either WT or D496 BCs (Figures 6E-6L).These results suggest that enzymatic reactions between CL metabolic enzymes and their substrates may occur at the interface between the Mieap-containing phase (Mieap and substrates) and the Mieap-depleted phase (CL metabolic enzymes).

Mieap functions in mitochondrial quality control via regulation of CL metabolism
CL alterations cause mitochondrial dysfunction. 20,21Mieap is thought to be involved in mitochondrial quality control. 25,26,28,31Therefore, we hypothesized that Mieap contributes to mitochondrial quality control by regulating CL metabolism.To test this hypothesis, we evaluated mitochondrial status relative to CL integrity in the presence or absence of Mieap protein in cells and mice.
First, we examined respiration rate, mitochondrial ATP production rate, crista morphology, and ROS levels, in control and Mieap-knockdown (KD) LS174T cells under physiological conditions, all of which reflect CL integrity.Flux analysis indicated that respiration and ATP production rates of Mieap-KD cells were significantly lower than those of control cells (Figures 7A, 7B, and S7).TEM analysis revealed that cristae of Mieap-KD cells decreased, and their morphology became indistinct and irregular, compared to that of control cells (Figures 7C and 7D).ROS levels increased in Mieap-KD cells (Figure S8).Consistently, the total amount of CL in control cells was higher than in Mieap-KD cells (Figure 7E), and control cells showed higher absolute values for almost all CL species than Mieap-KD cells (Figure 7F), suggesting the role of Mieap in CL biosynthesis.However, physiological Mieap significantly increased the relative values of CL72:4, CL72:5, CL70:4, CL68:3, and CL68:4, and decreased relative values of CL64:4, CL66:4, CL66:5, CL68:6, and CL70:6 in LS174T cells (Figure 7F), suggesting the role of Mieap in CL remodeling.
Second, utilizing a Mieap-deficient colorectal cancer cell line HCT116, in which the promoter of the Mieap gene is completely methylated, 25 we examined whether re-expression of Mieap affects respiration rate and mitochondrial ATP production rate in these cells.As shown in Figures 7G-7J and S7, re-expression of Mieap protein by Ad-Mieap infection significantly increased respiration rate and mitochondrial ATP production in HCT116 cells.
Third, we analyzed crista morphology in kidney and liver tissues of Mieap-knockout (KO) mice by performing TEM analysis.In Mieap À/À kidney mitochondria, irregularly dilated lamellar structures without distinct OsO 4 staining were observed (Figures 7K and 7M).A decrease in normal crista structure was a common characteristic of Mieap À/À mitochondria in the kidney and liver (Figures 7K-7N).
To evaluate gender bias in these mice, we further examined whether there is any difference in obesity between male and female in Mieapdeficient mice.As shown in Figures S10A, S10B, S10E, S10F, Mieap-deficient mice of both sexes developed obesity that persisted throughout their lives.This obesity was prominent during middle and old age, from 44 to 104 weeks (Figures S10E and S10G) (mean value GSE; Mieap +/+ 36.060G 0.509 g [n = 74], Mieap +/À 36.835G 0.371 g [n = 139], Mieap À/À 37.679 G 0.461 g [n = 90] in male, and Mieap +/+ 30.110G 0.514 g To confirm that the body weight gain in Mieap-deficient mice reflects increased fat deposition, we evaluated the amount of subcutaneous fat and intra-abdominal fat (perigonadal fat and perirenal fat) in four representative litter pairs (Mieap +/+ and Mieap À/À ).As shown in Figure S11, the amount of fat in Mieap À/À mice is obviously higher than in Mieap +/+ mice in all four pairs.Therefore, we suggest that the body weight gain in Mieap-deficient mice is due to obesity.
Reduced mitochondrial respiratory activity in adipose tissues has been suggested as a factor contributing to obesity. 50Brown fat tissue (BAT) is essential to heat production via both its respiratory activity, which generates a proton gradient, and uncoupling protein 1 (UCP1)mediated proton leakage across the mitochondrial inner membrane. 51,52Therefore, BAT mitigates obesity.Importantly, as a CL-binding protein, activity of UCP1 is stabilized by CL. 53 Thus, we examined the status of mitochondrial cristae in BAT of Mieap À/À mice.We confirmed that normal crista structure was significantly decreased in Mieap-deficient BAT (Figures 8D and 8E).
These results of Figures 8A-8E, S10, and S11, taken together, suggest that obesity may be a long-term consequence of mitochondrial dysfunction due to CL alteration in tissues of Mieap-deficient mice, including BAT (Figure 8F).
Finally, we examined endogenous Mieap expression in vivo.To achieve this, utilizing WT, Mieap-KO, and p53-KO mice and anti-mouse Mieap antibody, we performed immunohistochemical (IHC) analysis on 20 tissues/organs, including brain (cerebellar cortex), brain (cerebroventricle), heart, liver, kidney, lung, stomach, small intestine, pituitary gland, eye, harderian gland, salivary gland, thyroid, bladder, testis, epididymis, seminal vesicle, prostate, uterus, and fallopian tube.Specific signals of Mieap were detected in all examined tissues/organs by comparing data from WT and Mieap-KO (Figures S12-S16).Interestingly, basal expression of endogenous Mieap protein is not regulated by p53.Importantly, patterns of Mieap expression are remarkably similar to those of cytochrome c in IHC and IF (Figures S12-S16), suggesting that endogenous Mieap protein is localized at mitochondria.
Considering the results of partitioning behavior of Mieap and its mutants in Figure 5I, since endogenous Mieap protein is expressed at mitochondria in most tissues/organs, endogenous Mieap protein could be highly concentrated in mitochondrial droplets under physiological conditions.This implies that physiological Mieap protein likely has a significant impact on mitochondrial quality control in all tissues/cells containing mitochondria.On the basis of all these results, we suggest that Mieap maintains mitochondrial quality by regulating CL metabolism.

DISCUSSION
Although the importance of multi-enzyme complexes in metabolic enzyme reactions has been recognized, it remains unclear how this complex of enzymes efficiently and safely enables sequential enzymatic reactions by preventing diffusion of intermediates.A recent report suggested that concentration of multiple enzymes and substrates/intermediates in a restricted space could mediate efficient sequential enzymatic reactions by preventing diffusion of intermediates. 54In that model, multiple copies of upstream and downstream enzymes involved in sequential enzymatic reactions are assembled into a single cluster, called an ''agglomerate.''According to this model, ''once an upstream enzyme produces an intermediate, although the probability of the intermediate being processed by any individual downstream enzymes is low, the probability that the intermediate will be processed by one of the many downstream enzymes in the ''agglomerate'' can be high.'' 54Therefore, based on this model, molecular crowding of enzymes, substrates, and intermediates in a restricted space could enable efficient sequential enzymatic reactions mediated by multiple enzymes.
The ''agglomerate'' concept is promising, but an important question remains.How can so many diverse molecules, including multiple enzymes, substrates, and intermediates, be gathered, concentrated, and compartmentalized in a single restricted space?What drives formation of the ''agglomerate''?We speculate that BCs could organize agglomerates as metabolic BCs. 557][58][59][60] Since BCs are not surrounded by lipid bilayers, theoretically, they exhibit spatiotemporal dynamic properties within a cell in response to cellular stress and/or subcellular circumstances.More importantly, while BCs contain hundreds of molecules, a few scaffold proteins can drive formation of these MLOs. 1 If there are proteins that can organize metabolic BCs as scaffolds, 57 these agglomerates could enable efficient metabolic reactions.
In the present study, we obtained the following evidence supporting our hypothesis that Mi-BCs may regulate CL metabolism: [1] Mieap drives formation of droplets in mitochondria.[2] Mi-BCs phase-separate NAO, a specific probe for CL, and Mieap directly binds to CL. [3] Mi-BCs compartmentalize and concentrate all four sequential enzymes for CL biosynthesis (TAMM41, PGS1, PTPMT1, and CRLS1).[4] Mi-BCs compartmentalize and concentrate two enzymes for CL remodeling (PLA2G6 and TAZ).[5] The presence or absence of Mieap protein is closely related to an increase or decrease in various species of CL, respectively.[6] Mieap protein specifically increases the proportion of several species of CLs.[7] Mieap deficiency is related to changes in crista structure and CL metabolism in cells, and crista structure in vivo.This evidence suggests that Mieap is a scaffold protein that drives formation of metabolic BCs to compartmentalize and concentrate enzymes, substrates, and intermediates that are involved in CL biosynthesis and remodeling, leading to molecular crowding within Mi-BCs that promotes efficient catalysis of CL metabolic reactions (Figure 9). 61Supporting this model, a recently developed bioinformatics-based program 48 (the FuzDrop method) predicted that Mieap acts as a driver protein of MLOs and that PGS1, CRLS1, PLA2G6, and TAZ act as client proteins of MLOs (Figure S5).Mi-BCs exhibit properties of multi-phase droplets, in which there are two phases, a Mieap-containing phase and a Mieap-depleted phase.Interestingly, CL and Mieap occur in the Mieap-containing phase, whereas all CL biosynthesis and remodeling enzymes, including TAMM41, PGS1, PTPMT1, CRLS1, PLA2G6, and TAZ, are predominantly segregated into the Mieap-depleted phase.This result suggests that substrates, intermediates, and products for CL metabolism do not reside in the same phase as their catalytic enzymes.Such a relationship between substrates and enzymes in multi-phase droplets is not limited to Mi-BCs but is seen in other droplets.In terms of RNA processing droplets formed by FMRP and CAPRIN1, RNA and phosphorylated FMRP form multi-phase droplets, in which the deadenylation enzyme, CNOT7, and the substrate, polyA-RNA, are segregated into different phases, but this leads to faster deadenylation rates. 62hy don't substrates and enzymes occur in the same phase of metabolic BCs?A recent study demonstrated that sequestration of enzymes to a membrane-less compartment that is away from, but adjacent to substrates, can accelerate reactions much faster than when the enzymes are mixed with the substrates in the same compartment. 63Concentration of enzymes and substrates in a single phase might result in substrate inhibition. 63Therefore, separation of enzymes from their substrates via LLPS could facilitate enzymatic reactions by mitigating substrate inhibition.In this case, the interface between the enzyme and substrate phases would be the site of the reaction. 63Consistent with this hypothesis, we observed accumulation of CL biosynthetic enzymes such as PGS1 and TAMM41 at the interface of the Mieap-depleted phase (Figures 6E-6L).Therefore, enzymatic reactions of CL enzymes and CL substrates may occur at the interface between the Mieap-containing phase (Mieap and substrates) and the Mieap-depleted phase (CL metabolic enzymes) (Figure 9). 61eric et al. reported the mechanism for generation of multi-phase structures in droplets. 64By performing in vivo and in vitro experiments, the authors demonstrated that layered droplet organization is caused by differences in droplet surface tension, facilitating sequential RNA processing reactions in a variety of RNP bodies.In their experiments, F1B and NPM1 formed multi-phase droplets in which F1B droplets tended to be encapsulated within NPM1 droplets.They found that F1B droplets tended to wet hydrophobic surfaces, whereas NPM1 droplets tended to wet hydrophilic surfaces.Wetting refers to the contact between liquids and surfaces, which depends on surface tension.Therefore, in the aqueous phase, F1B droplets with high surface tension tended to be enveloped by NPM1 droplets with lower surface tension.
Since Mieap may be positioned with its N-terminal domain facing outward at the surfaces of Mi-BCs (Figure S1B), Mi-BCs could exist in a hydrophilic environment.If so, according to the theory of Feric et al., the Mieap-containing phase may be enveloped by the Mieap-depleted phase, because the former is more hydrophobic than the latter.However, the authors also pointed out the presence of a surfactant that modulates surface tension alters or inverts the organization of multi-phase droplets.Since Mieap could serve as a biosurfactant, Mieap may modulate the surface tension of the Mieap-containing phase and the Mieap-depleted phase in a hydrophilic environment.Therefore, the relation between the two phases could be inverted in Mi-BCs.
We hypothesize the following model for sequential CL-metabolic reactions promoted by Mieap (Figure 9). 61Mieap may stably interact with the substrate (PA) via its C-terminal, hydrophobic, structured region, which exhibits a specific, strong interaction.On the other hand, Mieap weakly and transiently interacts with CL-metabolizing enzymes via its N-terminal hydrophilic region, which exhibits multiple interactions.In this model, once Mieap attracts the substrate with its C-terminal region, Mieap enables sequential CL metabolic reactions by transiently interacting with the enzyme corresponding to the substrate at the interface between the Mieap-containing phase and the Mieapdepleted phase, and then changing enzymes until mature CL is produced.Therefore, interactions of the N-terminal hydrophilic region of Mieap with CL enzymes and the C-terminal hydrophobic region of Mieap with CL/CL-related phospholipids may be critical to drive formation of multi-phase organization of Mi-BCs.In summary, (1) concentration of enzymes and substrates, (2) segregation of enzymes and substrates into distinct sub-compartments of metabolic droplets, (3) interfacial catalysis, and (4) biosurfactant activity of Mieap could foster highly efficient sequential enzymatic reactions for CL metabolism in Mi-BCs.We suggest that Mi-BCs may be the first known metabolic MLOs to regulate CL synthesis and remodeling (Figure 9). 61, I-J).After BCs were formed, the relationship between EGFP-TAMM41 and each mutant BCs was analyzed (A-D), or the Mieap-depleted phase was analyzed on wetting of EGFP-PGS1 (E, F, I, J) or EGFP-TAMM41 (G, H, K, and L).
BCs are often thought to accelerate enzymatic reaction rate by merely increasing local concentrations of enzymes and substrates (mass action).However, Peeples and Rosen reported that concentrating enzymes and substrates alone results in decreasing enzymatic reaction rates by substrate inhibition due to high concentrations of substrates. 65They demonstrated that in addition to mass action, a scaffold-induced decrease in K M is critical to accelerate enzymatic reactions in BCs. 65In their synthetic system where the SUMOylation enzyme cascade is recruited into engineered condensates, they showed that having both enzyme and substrate bound simultaneously to proximal sites in a scaffold oligomer, which leads to decreased K M, is required to enhance the enzymatic reaction.Our Mi-BC model is compatible with their findings; however, further investigation of this hypothesis is required.
In the present study, we demonstrated that Mieap-deficient LS174T cells exhibited altered CL metabolism, decreased respiration activity, increased ROS levels, and manifested abnormal crista structure, all of which are consistent with phenotypes induced by CL alteration.We also found that Mieap-deficient mice exhibit decreased numbers and morphological abnormalities of mitochondrial cristae in kidney and liver tissues.Furthermore, our body weight analysis of Mieap +/+ , Mieap +/À , and Mieap À/À mice clarified increased obesity in Mieap-deficient mice, which is likely attributable to mitochondrial dysfunction in various tissues, including BAT.Therefore, we assume that all these phenotypes in Mieap-deficient cells and mice reflect mitochondrial dysfunctions related to abnormal CL metabolism.So far, autophagy and proteostasis are two major mechanisms in mitochondrial quality control. 66In addition to these, we suggest that the Mieap-regulated pathway is the third mechanism for mitochondrial quality control, in which Mieap maintains integrity of mitochondria by regulating CL metabolism.It accomplishes this through homeostasis of the inner mitochondrial membrane by regulating CL metabolism, stabilizing oxidative phosphorylation, and suppressing mitochondrial ROS generation (Figure 8F).
Previously, we reported that although Mieap-deficient mice did not suffer from intestinal dysfunction, Mieap-deficient Apc Min/+ mice exhibited remarkable intestinal tumor generation and malignant transformation, compared to Mieap-WT Apc Min/+ mice. 28Furthermore, mitochondria in Mieap-deficient tumors revealed abnormal morphology, including fewer cristae and enlarged, spherical mitochondria.These results support the role of Mieap in mitochondrial quality control through control of CL metabolism in response to oncogenic stress.In addition to Apc Min/+ mice, we also observed tumor suppressive role of Mieap in a thyroid cancer mouse model. 67Therefore, Mieap-regulated mitochondrial quality control could be critical in tumor suppression by promoting CL metabolism, which leads to upregulation of respiratory activity and downregulation of mitochondrial ROS generation.Considering the role of Mieap in p53 function, we suggest that Mieap could act as a spatiotemporal and dynamic regulator/modulator of CL metabolism to suppress tumor initiation and progression. 61Recently, we found that Mieap-deficient sperm in Mieap-KO mice cause in vitro infertility due to mitochondrial ROS elevation and impaired sperm motility (unpublished data).Therefore, in addition to cancer and obesity, it is possible that alterations of Mieap-regulated mitochondrial quality control also promote infertility (Figure 8F).Further investigation is required to clarify the precise role of Mieap in the regulation of CL metabolism.
A recent study reported that PGS1, PTPMT1, and CRLS1 form a large mitochondrial CL synthesis complex in human cells, with a molecular mass of 700-800 kDa. 68This large CL synthesis complex includes multiple CL-binding proteins.These observations are very similar to our results that Mi-BCs contain all CL synthesis enzymes (Figures 4D-4I) and several CL-binding proteins (unpublished data).However, IF experiments on PGS1 or CRLS1 in that study indicated that neither PGS1 nor CRLS1 showed visible Mi-BC-like structures in mitochondria. 68Our results from live-cell imaging of EGFP-fused CL synthesis enzymes, including PGS1 and CRLS1, also showed that none of the EGFP-tagged (G) The kinetic profile of the OCR using the Seahorse XF Real-Time ATP rate assay in HCT116 cells infected with Ad-Mieap or Ad-empty.(H-J) Quantitative assessment of OCR (H), mitochondrial ATP production rates (I), and total ATP production rates (J) of the HCT116 cells as in (G).Data are shown as means G SD (n = 9).(K) Morphology of kidney mitochondria of Mieap +/+ and Mieap À/À mice with TEM.Scale bars, 1 mm (upper panels) and 200 nm (lower panels).(L) Ratios of crista area per mitochondrial section of Mieap +/+ and Mieap À/À mouse kidneys.Quantitative data were obtained from Mieap +/+ kidney mitochondria (n = 190) and Mieap À/À kidney mitochondria (n = 234) in TEM images and displayed in a violin plot.(M) Morphology of liver mitochondria of Mieap +/+ and Mieap À/À mice with TEM.Scale bars, 1 mm (upper panels) and 200 nm (lower panels).(N) Ratios of crista area per mitochondrial section of Mieap +/+ and Mieap À/À mouse livers.Quantitative data were obtained from Mieap +/+ liver mitochondria (n = 146) and Mieap À/À liver mitochondria (n = 134) in TEM images and displayed in a violin plot.(A, B, D, H-J, L, N For construction of plasmids containing EGFP-MieapDCC (pEGFP-MieapDCC), the nucleotide sequence of pEGFP-Mieap between two Pst I restriction sites was deleted by digestion with Pst I.The remainder was self-ligated, additionally deleting c.810C using the QuikChange Site-Directed Mutagenesis Kit (Agilent) with primers Mut-F3 and Mut-R3 to make the deletion mutation in-frame.
For construction of plasmids containing EGFP-MieapD275 (pEGFP-MieapD275), the nucleotide sequence of pEGFP-Mieap between the Bgl II and Sma I restriction sites was deleted by digestion using Bgl II and Sma I.After blunting with T4 DNA polymerase (Thermo Fisher Scientific), the remainder was self-ligated.
For construction of plasmids containing EGFP-MieapD496 (pEGFP-MieapD496), the nucleotide sequence of pEGFP-Mieap was deleted between the EcoR I and Kpn I restriction sites by digestion using EcoR I and Kpn I.After blunting with T4 DNA polymerase (Thermo Fisher Scientific), the remainder was self-ligated.
For construction of plasmids containing TagRFP-T-Mieap (pTagRFP-T-Mieap), the nucleotide sequence of pEGFP-Mieap between the Nhe I and Xho I restriction sites containing EGFP was replaced with nucleotide sequence of pTagRFP-T-EEA1 (Addgene #42635) between the Nhe I and Xho I restriction sites containing TagRFP-T, by digestion using Nhe I and Xho I.
For construction of plasmids containing Mieap DCC (pMieap DCC), pEGFP-Mieap DCC was digested at the Kpn I restriction sites to obtain the nucleotide sequence of Mieap DCC, and ligated into pcDNA3.1 (+) (Thermo Fisher Scientific) cut with the same enzyme, Kpn I.
For construction of plasmids containing Mieap D274 (pMieap D274), the nucleotide sequence of Mieap D274 was PCR-amplified from pEGFP-Mieap D275 using the primers, D274-F and D274-R.PCR products were digested with Kpn I, and ligated into pcDNA3.1 (+) (Thermo Fisher Scientific) cut with the same enzyme, Kpn I.
For construction of plasmids containing Mieap D496 (pMieap D496), pEGFP-Mieap D496 was subjected to inverse PCR using the primers, DEGFP-F and DEGFP-R to delete the nucleotide sequence of EGFP from pEGFP-Mieap D496, and the product was self-ligated using KOD-Plus-Mutagenesis Kit (TOYOBO).
Prior to construction of plasmids containing TagRFP-T-Mieap deletion mutants, the nucleotide sequence of TagRFP-T was PCR-amplified using the primers, TagRFP-T-F and TagRFP-T-R.PCR products were digested with Hind III and EcoR V, and ligated into pcDNA3.1 (+) (Thermo Fisher Scientific) cut with the same enzymes (pcDNA-N-TagRFP).The nucleotide sequence of Mieap was PCR-amplified using the primers, G35-F and G35-R.PCR products were digested with EcoRV and PspOMI, and ligated into pcDNA-N-TagRFP cut with the same enzymes (pG35).
For construction of plasmids containing TagRFP-T-Mieap DCC (pTagRFP-T-Mieap DCC), pG35 was subjected to inverse PCR using the primers, DCC-F and DCC-R, and the product was self-ligated using KOD-Plus-Mutagenesis Kit (TOYOBO).
All primers are listed in Table S1.

Other constructs
For construction of plasmids containing EGFP-BNIP3 (pEGFP-BNIP3), plasmids containing FLAG-BNIP3 (pCMV-Tag2B-BNIP3) were constructed in advance.For construction of the pCMV-Tag2B-BNIP3, the nucleotide sequence of BNIP3 was PCR-amplified using the primers, BNIP3-F and BNIP3-R.PCR products were ligated into the pCR-Blunt II-TOPO vector (Thermo Fisher Scientific) and sequenced.Inserted products were digested with EcoR I and Xho I, and ligated into the pre-digested pCMV-Tag2B (Agilent) cut with the same enzyme.The nucleotide sequence of pCMV-Tag2B-BNIP3 was digested at the EcoR I and Xho I restriction sites, and subsequently blunted with T4 DNA polymerase (Thermo Fisher Scientific).pEGFP-C1 (Clontech) was digested with Bgl II, blunted with T4 DNA polymerase, self-ligated, digested with EcoR I and Sma I, and ligated with the fragment of pCMV-Tag2B-BNIP3.For construction of plasmids containing EGFP-NIX (pEGFP-NIX), plasmids containing FLAG-NIX (pCMV-Tag2B-NIX) were constructed in advance.For construction of the pCMV-Tag2B-NIX, the nucleotide sequence of NIX was PCR-amplified using the primers, NIX-F and NIX-R.PCR products were ligated into the pCR-Blunt II-TOPO vector (Thermo Fisher Scientific) and sequenced.Inserted products were digested with EcoR I and Xho I, and ligated into pre-digested pCMV-Tag2B (Agilent) cut with the same enzyme.The nucleotide sequence of pCMV-Tag2B-NIX was digested at the EcoR I and Xho I restriction sites, and subsequently blunted with T4 DNA polymerase (Thermo Fisher Scientific).pEGFP-C1 (Clontech) was digested with Bgl II, blunted with T4 DNA polymerase, self-ligated, digested with EcoR I and Sma I, and ligated with the fragment of pCMV-Tag2B-NIX.
For construction of the plasmid backbone containing EGFP (pN-EGFP), the nucleotide sequence of EGFP, excluding the stop codon, was PCR-amplified using the primers, N-EGFP-F and N-EGFP-R.PCR products were digested with Hind III and BamH I, and ligated into pcDNA3.1 (+) (Thermo Fisher Scientific) cut with the same enzymes.
For construction of the plasmid backbone containing EGFP (pC-EGFP), the nucleotide sequence of EGFP was PCR-amplified using the primers, C-EGFP-F and C-EGFP-R.PCR products were digested with BamH I and Not I, and ligated into pcDNA3.1 (+) (Thermo Fisher Scientific) cut with the same enzymes.
For construction of plasmids containing EGFP-TAMM41 (pEGFP-TAMM41), the nucleotide sequence of TAMM41, excluding the stop codon, was PCR-amplified using the primers, TAMM41-F and TAMM41-R.PCR products were digested with Nhe I and BamH I, and ligated into pC-EGFP cut with the same enzymes.
For construction of plasmids containing EGFP-PGS1 (pEGFP-PGS1), the nucleotide sequence of PGS1, excluding stop codon, was PCRamplified using the primers, PGS1-F and PGS1-R.PCR products were digested with Hind III and BamH I and ligated into pC-EGFP cut with the same enzymes.
For construction of plasmids containing EGFP-PTPMT1 (pEGFP-PTPMT1), the nucleotide sequence of PTPMT1, excluding the stop codon, was PCR-amplified using the primers, PTPMT1-F and PTPMT1-R.PCR products were digested with Nhe I and Hind III and ligated into pC-EGFP cut with the same enzymes.
For construction of plasmids containing EGFP-CRLS1 (pEGFP-CRLS1), the nucleotide sequence of CRLS1 was PCR-amplified using the primers, CRLS1-F and CRLS1-R.PCR products were digested with BamH I and Not I and ligated into pN-EGFP cut with the same enzymes.
For construction of plasmids containing EGFP-PLA2G6 (pEGFP-PLA2G6), the nucleotide sequence of PLA2G6, excluding the stop codon, was PCR-amplified using the primers, PLA2G6-F and PLA2G6-R.PCR products were digested with Nhe I and Kpn I and ligated into pC-EGFP cut with the same enzymes.
For construction of plasmids containing EGFP-TAZ (pEGFP-TAZ), the nucleotide sequence of TAZ, excluding the stop codon, was PCRamplified using the primers, TAZ-F and TAZ-R.PCR products were digested with Nhe I and Kpn I and ligated into pC-EGFP cut with the same enzymes.
For construction of plasmids containing EGFP-PRELI (pEGFP-PRELI), the nucleotide sequence of PRELI, excluding the stop codon, was PCR-amplified using the primers, PRELI-F and PRELI-R.PCR products were digested with Nhe I and BamH I and ligated into pC-EGFP cut with the same enzymes.
For construction of plasmids containing EGFP-LONP1 (pEGFP-LONP1), the nucleotide sequence of LONP1, excluding the stop codon, was PCR-amplified using the primers, LONP1-F and LONP1-R.PCR products were digested with Hind III and BamH I and ligated into pC-EGFP cut with the same enzymes.
For construction of plasmids containing EGFP-PLD6 (pEGFP-PLD6), the nucleotide sequence of PLD6, excluding the stop codon, was PCRamplified using the primers, PLD6-F and PLD6-R.PCR products were digested with Nhe I and Kpn I and ligated into pC-EGFP cut with the same enzymes.
All primers are listed in Table S1.

Adenoviral infection
Infection of cell lines was carried out by adding viral solutions to cell monolayers, incubating them at 37 C for 120 min with brief agitation every 20 min.This was followed by addition of culture medium and return of the infected cells to the 37 C incubator.

Calculation of intensity ratio
To evaluate partitioning of EGFP-Mieap and deletion mutant proteins, EGFP-Mieap WT, DCC, D275, and D496 were expressed in A549-cont cells to generate condensates by infection with Ad-EGFP-Mieap WT, DCC, D275, and D496, respectively.EGFP intensity of condensates and cytoplasm was measured.Because EGFP intensity of these condensates was higher than the intensity of 0.4 mg/mL His-EGFP solution for standard curve, we chose intensity ratio rather than partition coefficient for the parameter of this partitioning experiments. 49Intensity ratio was calculated as (Intensity of condensates-Intensity of background)/(Intensity of cytoplasm-Intensity of background), where Intensity of condensates, Intensity of cytoplasm, and Intensity of background are the mean intensities of condensates, cytoplasm, and PBS acquired by the identical conditions (laser wavelength, 488 nm; laser transmissivity, 0.01%; detection wavelength, 500-600 nm; voltage, 350 V) on an FLUOVIEW FV3000 confocal laser scanning microscope (Olympus).Intensity ratio data were obtained from 40 cells for each construct.

Expression and purification of GST and GST-Mieap
Escherichia coli (BL21) cells transformed with expression vectors were grown in 200 mL of Luria-Bertani medium at 37 C until the OD600 was between 0.55 and 0.6.Protein expression was induced with 100 mM IPTG, and bacteria were subsequently incubated for 3 h at 25 C.After harvesting bacteria by centrifugation at 3000 3 g for 10 min at 4 C, pellets were lysed with lysis buffer (1% Triton X-100 buffered in PBS supplemented with 1 mM Phenylmethylsulfonyl fluoride), and sonicated (20 3 30 s bursts with 10 s rest between bursts).Insoluble material was removed by centrifugation at 10,000 rpm for 30 min at 4 C. Supernatant was incubated with glutathione Sepharose 4B (Cytiva) pre-equilibrated with lysis buffer at 4 C overnight.After the beads were washed twice with lysis buffer, proteins were eluted with elution buffer (50 mM glutathione diluted in 50 mM Tris-HCl, pH 8.0), and dialyzed at 4 C overnight against PBS.

Lipid-binding analysis
For lipid-binding analysis, protein-lipid interactions on lipid-spotted membranes were evaluated with fat blot assays. 43Natural CL, PC, and PE derived from bovine heart (Olbracht Serdary Research Laboratories) were diluted with chloroform/methanol/1N HCl (80:80:1).1 mL of each diluted lipid was spotted onto PVDF membranes (Cytiva) for antigen-antibody reactions using anti-Mieap antibody or nitrocellulose membranes (Cytiva) for antigen-antibody reactions using an anti-GST antibody to align spots with increasing amounts of lipids ranging from 0 to 667 pmol.Here, approximate molarities of CL, PC, and PE calculated from molecular weights of tetralinoleoyl CL, distearoyl PC, and distearoyl PC were used, respectively.After membranes were blocked with blocking buffer (3% fatty acid-free BSA diluted in 50 mM Tris-HCl, 150 mM NaCl, pH 7.5) for 1 h, membranes were incubated with 2.5 mg/mL of GST-Mieap or GST protein diluted in blocking buffer containing 0.1% Tween 20 overnight.Membranes were incubated with primary antibody (rabbit anti-Mieap antibody or rabbit anti-GST antibody) diluted in blocking buffer containing 0.06% Tween 20 (1:1000) for 3.5 h, and subsequently a secondary antibody (goat anti-rabbit antibody conjugated to horseradish peroxidase) diluted in blocking buffer containing 0.06% Tween 20 (1:10000) for 1 h.ECL Western Blotting Detection Reagents (Cytiva) was used to detect HRP and chemiluminescence was visualized with an ImageQuant LAS 4000 system (Cytiva).

Lipid preparation
Lipid preparation was performed as described previously. 76,77Briefly, total lipids were extracted from samples using the Bligh-Dyer method. 78n aliquot of the organic phase was added to an equal volume of methanol before being loaded onto a DEAE-cellulose column (Wako Chemical) pre-equilibrated with chloroform.After successive washes with chloroform/methanol (1:1, v/v), acidic phospholipids were eluted with chloroform/methanol/HCl/water (12:12:1:1, v/v), followed by evaporation to dryness to yield a residue was soluble in methanol.

Real-time ATP rate assay
LS174T-cont and Mieap-KD cells were seeded at a density of 2.5310 4 cells/well (n = 9) on a Seahorse XF24 Cell Culture Microplate.Cells were incubated at 37 C in a humidified chamber with 5% CO 2 .18 h after seeding, culture medium was replaced with XF DMEM medium pH 7.4 supplemented with 25 mM glucose and 2 mM L-glutamine through three washes.HCT116 cells were seeded at a density of 0.8310 6 cells/60-mm dish (n = 9).Cells were incubated at 37 C in a humidified chamber with 5% CO 2 .24 h after seeding, cells were treated with Ad-Mieap or Ad-empty.24 h after infection, cells were reseeded at a density of 4310 4 cells/ well (n = 9) on a SeahorseXF24 Cell Culture Microplate.20 h after reseeding, culture medium was replaced with XF DMEM medium pH 7.4 supplemented with 25 mM glucose and 2 mM L-glutamine through three washes.
After cells were incubated at 37 C in a non-CO 2 incubator for 60 min, cell culture plates were loaded into a Seahorse XFe24 Analyzer.Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were recorded before and after serial injections of oligomycin and rotenone/antimycin A to yield final concentrations of 0.5 mM.

Flow cytometric analysis
LS174T-cont and Mieap-KD cells cultured under normal conditions were harvested by trypsin-EDTA treatment.After adding complete growth media to inactivate trypsin, cells were centrifuged, washed with PBS, and incubated with 5 mM 2 0 ,7 0 -dichlorofluorescin-diacetate (Sigma) for 20 min at 37 C.After being washed with PBS, cells were immediately analyzed with an EC800 flow cytometry analyzer (Sony) using the 488-nm line.

Primers
The information of all PCR primers is indicated in Table S1.

QUANTIFICATION AND STATISTICAL ANALYSIS FRAP data quantification
Fluorescence recovery rates were calculated using cellSens Imaging Software (Olympus), in which the intensity initially acquired after bleaching was set to 0 and the pre-bleaching intensity was set to 1.The normalized average fluorescence recovery was plotted in JMP 14.2.0 (SAS).

Crista data quantification
For quantification of crista data, crista area and outlines of mitochondrial sections in TEM images were marked manually using Adobe Photoshop CC, where normal crista morphology was identified by the presence of lamellar structures with distinct OsO 4 staining.Aberrant cristalike structures that were not observed in mitochondria of WT were excluded.Subsequently, the ratio of crista area per mitochondrial section was calculated from the indicated number of mitochondria in legends of Figures 7D, 7L, 7M, and 8E, using ImageJ. 72

Data visualization
Visualization of the experimental data subjected to statistical analyses were performed using Graph Builder engine in JMP 14.2.0 (SAS).When the data were visualized using violin plots, boxplots were overlaid.The center line in the box indicates the median.The bottom and top of the box indicate the 25 th and 75 th percentiles.The whiskers extend 1.5 times the interquartile range (IQR) from the top and bottom of the box unless the minimum and maximum values are within the IQR.The values which fall above or below the whiskers are plotted individually as outliers.

Figure 1 .
Figure 1.Mieap forms mitochondrial biomolecular condensates (A) Live-cell imaging and the 3D reconstruction showing the spatial relationship between the Mi-BCs (EGFP-Mieap) and the mitochondrial outer membranes visualized with mApple-TOMM20.The A549 cells were co-infected with Ad-EGFP-Mieap and Ad-mApple-TOMM20.Left: a cell image.Right: higher magnification of the area indicated by the dashed square in the left panel and a line-scan of fluorescence intensities along the dashed arrow.Scale bars, 10 mm (left panel) and 2 mm (right panel).See also Video S2. (B) Super-resolution images showing the spatial relationship between Mi-BCs (EGFP-Mieap) and mitochondrial outer membranes visualized with mApple-TOMM20.Scale bar, 1 mm.(C) z stack images of the cell in (A).Scale bar, 10 mm.(D) 3D reconstruction of the cell shown in (A).See also Video S2.

Figure 2 .
Figure 2. Mieap is an IDR-containing protein that has a potential to drive LLPS (A) Phylogenetic spread of Mieap orthologs annotated with OrthoDB v10.Red sectors indicate present species.Light blue sectors indicate missing species.(B) Proportion of amino acid residues in each domain of the Mieap protein.(C) Multiple sequence alignment for Mieap orthologs in representative eukaryotes.Black and gray boxes indicate 100% and 80% identical residues among eukaryotes, respectively.Blue letters indicate IDRs annotated by VL3-BA.Orange letters indicate coiled-coil regions annotated by COILS.(D) Schematic of the domain structure of Mieap.The dashed vertical line indicates the boundary of gross hydrophilic and hydrophobic halves, separated by IDR3 and the adjacent structured region.Asterisks indicate clusters of positively charged residues.(E) Sequence analyses of Mieap protein.VL3-BA prediction of IDRs on the amino acid sequence of Mieap, in which bold lines indicate IDRs; DisMeta, meta-prediction of IDRs on the amino acid sequence of Mieap; COILS; coiled-coil regions annotated on the amino acid sequence of Mieap using a 21-residue sliding window; Hydro, hydrophobicity of Mieap using a 9-residue sliding window; NCPR, the linear net charge per residue of Mieap using a 5-residue sliding window.

Figure 3 .
Figure 3. Material state and dynamics of Mi-BCs and phase-separation of the mitochondrial phospholipid cardiolipin by Mi-BCs (A) EGFP-Mieap and the three deletion-mutant forms.The schematic indicates wild-type (WT) and three deletion mutants (DCC, D275, and D496) of EGFP-Mieap protein.Numbers indicate amino acid residues.(B) Normalized average fluorescence recovery in the FRAP experiment.EGFP-Mieap, EGFP-Mieap DCC, EGFP-Mieap D275, and EGFP-Mieap D496 were expressed in A549 cells to generate condensates by infection with Ad-EGFP-Mieap, Ad-EGFP-Mieap DCC, Ad-EGFP-Mieap D275, and Ad-EGFP-Mieap D496, respectively.Each condensate was subjected to spot-bleaching using a 488-nm laser at 10% laser power with 11.6 ms/mm exposure time and followed up for 60 s n = 15 condensates for each construct.Data shown are means G SD. (C) Normalized average fluorescence recovery in the FRAP experiment with weaker laser exposure as in (B).Laser power was weakened to 1.4% and exposure time was shortened to 1.4 ms/mm.Observation duration was expanded to 15 min after photobleaching entire condensates.n = 10 condensates for each construct.Data shown are means G SD. (D-F) Screening of the mitochondrial molecules involved in phase-separation by Mi-BCs.Ad-EGFP-BNIP3 (D), Ad-EGFP-NIX (E), and Ad-AcGFP1-Mito (F) were co-infected with Ad-Mieap and Ad-TagRFP-T-Mieap in A549 cells.Whether each mitochondrial fluorescence probe is phase-separated by Mi-BCs was examined

Figure 3 .
Figure 3. Continued with live-cell imaging analysis in A549 cells.EGFP-BNIP3 (D), EGFP-NIX (E), and AcGFP-mito (F) were not incorporated into Mi-BCs.Lower right: line-scan of fluorescence intensities along the dashed arrow.Scale bars, 10 mm.See also Video S3. (G) Live-cell imaging showing phase-separation of CL by Mi-BCs.CL was visualized by 10-nonylacridine orange bromide (NAO) in A549 cells.A549 cells were infected with Ad-Mieap and Ad-TagRFP-T-Mieap, and subsequently treated with NAO (200 nM).NAO was incorporated into Mi-BCs.Lower right: line-scan of fluorescence intensities along the dashed arrow.Scale bar, 10 mm.See also Video S4.(H) Lipid-binding analysis of GST-tagged Mieap protein.GST-Mieap or GST was incubated with membranes on which increasing amounts of CL, phosphatidylcholine (PC), and phosphatidylethanolamine (PE), ranging from 0 to 667 pmol, were spotted.Protein-lipid interactions were visualized using an anti-Mieap antibody and/or anti-GST antibody, as indicated.

Figure 5 .
Figure 5. Continued A549 cells were infected with Ad-TagRFP-T-Mieap WT (B), Ad-TagRFP-T-Mieap DCC (F), Ad-TagRFP-T-Mieap D275 (G), or Ad-TagRFP-T-Mieap D496 (H) to form BCs, and after BCs were formed, mitochondrial inner membrane was visualized by NAO.(I) Partitioning behavior of EGFP-Mieap WT, DCC, D275, or D496 protein in condensates and cytoplasm, displayed in violin plot.A549 cells were infected with Ad-EGFP-Mieap WT, DCC, D275, or D496 to generate condensates, and after BCs were formed, the intensity ratio (IR) of condensates and cytoplasm was measured.n = 40 cells for each construct in A549 cells.**p < 0.01, ***p < 0.001, ****p < 0.0001, two tailed Student's t test.When the data were visualized using violin plots, boxplots were overlaid.The center line in the box indicates the median.The bottom and top of the box indicate the 25 th and 75 th percentiles.The whiskers extend 1.5 times the interquartile range (IQR) from the top and bottom of the box unless the minimum and maximum values are within the IQR.The values which fall above or below the whiskers are plotted individually as outliers.

Figure 6 .
Figure 6.Both N-and C-terminal regions of Mieap are required to generate the multi-phase structure of Mi-BCs (A-D) Comparison of phase-separating behaviors on the CL metabolic enzyme, TAMM41 (EGFP-TAMM41), between BCs formed by TagRFP-T-Mieap WT (A), DCC (B), D275 (C), and D496 (D).Right: line-scan of fluorescence intensities along the dashed arrow.Scale bars, 10 mm.(E-H) CL metabolic enzymes wet the interface in the Mieap-depleted phase of Mi-BCs.Distributions of the CL metabolic enzymes, EGFP-PGS1 (E, F) and EGFP-TAMM41 (G, H), in the Mieap-depleted phase of Mi-BCs are shown.Lower right: line-scan of fluorescence intensities along the dashed arrow.Scale bars, 2 mm.(I-L) CL metabolic enzymes wet the interface in the Mieap D496-depleted phase of D496-BCs.Distributions of the CL metabolic enzymes, EGFP-PGS1 (I, J) and EGFP-TAMM41 (K, L), in the Mieap D496-depleted phase of BCs formed by the D496 mutant are shown.Lower right: line-scan as in (E-H).Scale bars, 2 mm.HeLa cells were transfected with pEGFP-TAMM41 (A-D, G, H, K, and L) or pEGFP-PGS1 (E, F, I, and J), and subsequently infected with Ad-TagRFP-T-Mieap WT (A, E-H), Ad-TagRFP-T-Mieap DCC (B), Ad-TagRFP-T-Mieap D275 (C), or Ad-TagRFP-T-Mieap D496 (D, I-J).After BCs were formed, the relationship between EGFP-TAMM41 and each mutant BCs was analyzed (A-D), or the Mieap-depleted phase was analyzed on wetting of EGFP-PGS1 (E, F, I, J) or EGFP-TAMM41 (G, H, K, and L).

Figure 7 .
Figure 7. Mieap contributes to mitochondrial quality control by promoting CL metabolism (A) Oxygen consumption rates (OCR) of LS174T-cont and Mieap-KD cells under normal conditions calculated with a flux analyzer.Data are shown as means G SD (n = 9).(B) Mitochondrial ATP production rates of LS174T-cont and Mieap-KD cells under normal conditions calculated with a flux analyzer, using a Seahorse XF real-time ATP rate assay.Data are shown as means G SD (n = 9).(C) Morphology of mitochondria of LS174T-cont and Mieap-KD cells with transmission electron microscopy (TEM).Scale bars, 2 mm.(D) Ratio of crista area per mitochondrial section of LS174T-cont and Mieap-KD cells.Quantitative data were obtained from cont mitochondria (n = 197) and Mieap-KD mitochondria (n = 329) in TEM images and displayed in a violin plot.(E) Quantitative assessment of total CL by mass spectrometric analysis.LS174T cells with (Cont) and without (Mieap-KD) endogenous Mieap expressions were subjected to mass spectrometric analysis.Data shown are means G SE (n = 6).(F) Quantitative and rate assessments of CL species by mass spectrometric analysis.LS174T cells were analyzed as described in (E).Absolute values of selected CL species are shown as the amount of substance per cell (left panel).Relative values of selected CL species are shown as % of total CL (right panel).Data shown are means G SE (n = 6).(G)The kinetic profile of the OCR using the Seahorse XF Real-Time ATP rate assay in HCT116 cells infected with Ad-Mieap or Ad-empty.(H-J) Quantitative assessment of OCR (H), mitochondrial ATP production rates (I), and total ATP production rates (J) of the HCT116 cells as in (G).Data are shown as means G SD (n = 9).(K) Morphology of kidney mitochondria of Mieap +/+ and Mieap À/À mice with TEM.Scale bars, 1 mm (upper panels) and 200 nm (lower panels).(L) Ratios of crista area per mitochondrial section of Mieap +/+ and Mieap À/À mouse kidneys.Quantitative data were obtained from Mieap +/+ kidney mitochondria (n = 190) and Mieap À/À kidney mitochondria (n = 234) in TEM images and displayed in a violin plot.(M) Morphology of liver mitochondria of Mieap +/+ and Mieap À/À mice with TEM.Scale bars, 1 mm (upper panels) and 200 nm (lower panels).(N) Ratios of crista area per mitochondrial section of Mieap +/+ and Mieap À/À mouse livers.Quantitative data were obtained from Mieap +/+ liver mitochondria (n = 146) and Mieap À/À liver mitochondria (n = 134) in TEM images and displayed in a violin plot.(A, B, D, H-J, L, N) **p < 0.01, ***p < 0.001, ****p < 0.0001, two tailed Student's t test.(E, F) *p < 0.05, **p < 0.01, ***p < 0.001, two tailed paired t test.When the data were visualized using violin plots, boxplots were overlaid.The center line in the box indicates the median.The bottom and top of the box indicate the 25 th and 75 th percentiles.The whiskers extend 1.5 times the interquartile range (IQR) from the top and bottom of the box unless the minimum and maximum values are within the IQR.The values which fall above or below the whiskers are plotted individually as outliers.

FFigure 8 .
Figure 7. Mieap contributes to mitochondrial quality control by promoting CL metabolism (A) Oxygen consumption rates (OCR) of LS174T-cont and Mieap-KD cells under normal conditions calculated with a flux analyzer.Data are shown as means G SD (n = 9).(B) Mitochondrial ATP production rates of LS174T-cont and Mieap-KD cells under normal conditions calculated with a flux analyzer, using a Seahorse XF real-time ATP rate assay.Data are shown as means G SD (n = 9).(C) Morphology of mitochondria of LS174T-cont and Mieap-KD cells with transmission electron microscopy (TEM).Scale bars, 2 mm.(D) Ratio of crista area per mitochondrial section of LS174T-cont and Mieap-KD cells.Quantitative data were obtained from cont mitochondria (n = 197) and Mieap-KD mitochondria (n = 329) in TEM images and displayed in a violin plot.(E) Quantitative assessment of total CL by mass spectrometric analysis.LS174T cells with (Cont) and without (Mieap-KD) endogenous Mieap expressions were subjected to mass spectrometric analysis.Data shown are means G SE (n = 6).(F) Quantitative and rate assessments of CL species by mass spectrometric analysis.LS174T cells were analyzed as described in (E).Absolute values of selected CL species are shown as the amount of substance per cell (left panel).Relative values of selected CL species are shown as % of total CL (right panel).Data shown are means G SE (n = 6).(G)The kinetic profile of the OCR using the Seahorse XF Real-Time ATP rate assay in HCT116 cells infected with Ad-Mieap or Ad-empty.(H-J) Quantitative assessment of OCR (H), mitochondrial ATP production rates (I), and total ATP production rates (J) of the HCT116 cells as in (G).Data are shown as means G SD (n = 9).(K) Morphology of kidney mitochondria of Mieap +/+ and Mieap À/À mice with TEM.Scale bars, 1 mm (upper panels) and 200 nm (lower panels).(L) Ratios of crista area per mitochondrial section of Mieap +/+ and Mieap À/À mouse kidneys.Quantitative data were obtained from Mieap +/+ kidney mitochondria (n = 190) and Mieap À/À kidney mitochondria (n = 234) in TEM images and displayed in a violin plot.(M) Morphology of liver mitochondria of Mieap +/+ and Mieap À/À mice with TEM.Scale bars, 1 mm (upper panels) and 200 nm (lower panels).(N) Ratios of crista area per mitochondrial section of Mieap +/+ and Mieap À/À mouse livers.Quantitative data were obtained from Mieap +/+ liver mitochondria (n = 146) and Mieap À/À liver mitochondria (n = 134) in TEM images and displayed in a violin plot.(A, B, D, H-J, L, N) **p < 0.01, ***p < 0.001, ****p < 0.0001, two tailed Student's t test.(E, F) *p < 0.05, **p < 0.01, ***p < 0.001, two tailed paired t test.When the data were visualized using violin plots, boxplots were overlaid.The center line in the box indicates the median.The bottom and top of the box indicate the 25 th and 75 th percentiles.The whiskers extend 1.5 times the interquartile range (IQR) from the top and bottom of the box unless the minimum and maximum values are within the IQR.The values which fall above or below the whiskers are plotted individually as outliers.
-Mieap, EGFP-MieapDCC, EGFP-MieapD275, and EGFP-MieapD496 were expressed in A549-cont cells to generate condensates by infection with Ad-EGFP-Mieap, Ad-EGFP-MieapDCC, Ad-EGFP-MieapD275, and Ad-EGFP-MieapD496, respectively.FRAP experiments were performed on an FLUOVIEW FV3000 confocal laser scanning microscope (Olympus), using a 60x/1.4NA oil immersion objective (Olympus).Condensates were subjected to spot-bleaching or full-bleaching (bleaching entire condensates).For spot-bleaching, the bleaching area was unified to a diameter of 1.38 mm.Condensates were imaged for 6 s, acquiring 30 images prior to spot-bleaching or 50 s, acquiring 5 images prior to full-bleaching.Photobleaching employed a 488-nm laser at 10% laser power with 11.6 ms/mm exposure time or 1.4% laser power with 1.4 ms/mm exposure time.Time-lapse images were acquired at 0.2-ms intervals for 60 s or 10 s intervals for 15 min.Spot-bleaching data for each construct were acquired from 15 different condensates.Full-bleaching data of each construct were acquired from 10 different condensates.