A novel DNAJ protein, TCAIM, drives proteolysis of α-ketoglutarate dehydrogenase and regulates mitochondrial metabolism.

Mitochondria play essential roles in metabolism, and the proteostasis system is important for almost all biological processes that occur in this double-membrane-enclosed organelle, including metabolic functions. In this study, we identied a novel mtHSP70 co-chaperone DNAJC protein, TCAIM, that specically binds to the E1 subunit of α-ketoglutarate dehydrogenase (OGDH). Using the single-particle cryo-EM technique, we determined the binding structures of TCAIM and OGDH. We further demonstrated that by binding to the native form of OGDH, TCAIM specically mediates the degradation of OGDH and regulates its activity in vivo in a mtHSP70- and LONP1-dependent manner. Moreover, the lack of TCAIM changes metabolism in mice and primary cells, whereas overexpression of TCAIM decreases mitochondrial bioenergetics. Thus, this study revealed a novel mechanism by which mitochondrial metabolism could be regulated by selective degradation of an important metabolic enzyme, mediated by a DNAJC protein. centrifugation. Free while total was obtained ℃ for reaction were measured on a


Introduction
Mitochondria are the central organelles for bioenergetics and are crucial for many metabolic pathways, including biosynthesis and catabolism of glucose, fatty acids, and amino acids. The tricarboxylic acid (TCA) cycle not only generates nicotinamide adenine dinucleotide (NADH) and reduced avine adenine dinucleotide (FADH2) for ATP production, but also provides intermediates for various metabolic processes. Some TCA cycle enzymes are large protein complexes, such as the α-ketoglutarate dehydrogenase complex (OGDHC), located in the mitochondrial matrix. The function of these macromolecules needs to be maintained and regulated in a timely manner to coordinate with biochemical processes occurring inside and outside the mitochondria 1 . OGDHC consists of three different subunits: the E1 subunit, 2-oxoglutarate dehydrogenase (OGDH); the E2 subunit, dihydrolipoyllysine succinyltransferase (DLST); and the E3 subunit, D-2-hydroxyglutarate pyruvate transhydrogenase (DLD).
This enzyme catalyzes the reaction between α-ketoglutarate and CoA to reduce NAD + to NADH and generate succinyl-CoA and CO 2 2 . OGDHC is a rate-limiting enzyme of the tricarboxylic acid (TCA) cycle and plays a unique role in various metabolic pathways. Its substrate, α-ketoglutarate, is located at the crossroads of the TCA cycle and glutamine catabolism. Therefore, regulation of its function has been of interest for decades. OGDHC activity is regulated by the NAD + /NADH ratio, ADP/ATP ratio, and Pi concentration 3 . However, little is known about how these subunits and the stoichiometry of the protein complex are regulated post-translationally.
One of the most important mechanisms of post-translational regulation is protein degradation, an important part of protein homeostasis, or "proteostasis." In the mitochondria, proteostasis is essential for maintaining its metabolic functions, and its malfunction is a cause of many metabolic disorders related TCAIM was identi ed as a "T cell activation inhibitor, mitochondrial" but its biochemical function has not been fully clari ed 12,13 . Given its speci c binding to OGDH, a subunit of the OGDHC enzyme complex, we reasoned that it would affect mitochondrial function. Indeed, overexpression of TCAIM changed mitochondrial morphology, displayed a more fragmented network ( Fig. 2A), and decreased the maximum oxygen consumption rate (OCR) (Fig. 2B, Fig S2A), which is similar to the effect of OGDH knockdown ( Fig  S2 C-D). Surprisingly, the decrease in mitochondrial respiration was coupled with increased membrane potential and elevated mitochondrial ROS generation, as determined by JC-1 and MitoSOX staining, respectively ( Fig. 2C and 2D). These results indicated that TCAIM decreases oxidative phosphorylation is by suppressing electron transport chain (ETC) but profoundly affect energy metabolism. Similarly, overexpression of TCAIM also led to a decrease in the maximum extracellular acidi cation rate (ECAR) capacity, indicating that glycolysis was also affected by TCAIM overexpression (Fig. 2E, Fig S2E).

Loss of TCAIM-altered metabolism in mice
To further investigate the biological function of TCAIM, we established a Tcaim knockout (KO) mouse model using CRISPR/Cas9 and validated the deletion of Tcaim by sequencing and qPCR (Fig S3A and B). Compared to their wild-type (WT) littermates, these mice displayed reduced bodyweight, without a change in food intake ( Fig. 3A and B). Consistent with these phenotypes, Tcaim KO mice exhibited increased oxygen consumption compared with WT mice, indicating an increased metabolic rate (Fig. 3C). The human protein atlas database showed that the liver and kidney are among the organs with the highest Tcaim expression levels (https://www.proteinatlas. org/), and our qPCR results also con rmed this ( Fig  S3C). Moreover, 24 h of fasting followed by 16 h of feeding caused an elevation in Tcaim expression in mouse livers (Fig. 3D). The fact that the Tcaim mRNA level is the highest in the kidney and liver prompted us to speculate that loss of Tcaim likely causes metabolic changes in serum. Indeed, the serum metabolome of Tcaim KO mice was signi cantly different ( Fig. 3E and Fig S3D). Several glycerophospholipids, such as phosphatidylcholine, phosphatidylglycerol, and phosphatidic acid were at the top of the list of reduced metabolites in the serum of KO mice, and glycerophospholipid metabolism was the most enriched category by KEGG analysis (Fig. 3F, Fig S3E, and Table S2). We further con rmed the changes in lipid metabolism in mice by measuring cholesterol levels in mouse serum. In line with our metabolome results, compared with the serum from WT mice, the serum of Tcaim KO mice contained signi cantly less free and total cholesterol ( Fig. 3G and H). Interestingly, we also observed that compared to WT mice, the protein malonylation level was signi cantly decreased in the liver of Tcaim KO mice, but not the protein acetylation or succinylation levels (Fig. 3I, Fig S3F). This increase in protein malonylation levels could be mimicked by treating cells with malonate to increase malonyl-CoA levels ( Fig S3G). Since SIRT5 has been reported to regulate protein malonylation, we also tested whether SIRT5 expression was altered in Tcaim KO mice but found no difference in SIRT5 protein levels between WT and KO mice ( Fig  S3H).
TCAIM displays little effect on OGDH and OGDHC enzyme activities in vitro The type I and type II DNAJ protein (DNAJA and DNAJB) families are usually involved in protein folding and refolding by recruiting HSP70 to their clients, while the functions of the type III DNAJ protein (DNAJC) family are rather diverse, with some of them exhibiting HSP70-independent functions 14 . We asked whether the speci c binding between OGDH and TCAIM interferes with the enzymatic functions of OGDH and OGDHC. Adding TCAIM to puri ed OGDH in vitro slightly suppressed its maximum activity (Vmax) and decreased the Km for its substrate α-KG ( Fig. 4A and B, Fig S4A and B). Applying TCAIM to OGDHC in vitro had no obvious effect on Vmax or the α-KG Km of OGDHC under different pH conditions, even though the OGDHC activities decreased with an increase in buffer pH value (Fig. 4C). Therefore, these results indicated that TCAIM does not directly regulate the activity of OGDH or OGDHC. Since some DNAJ proteins are involved in client folding and refolding under stress conditions 15,16 , we further tested whether TCAIM helps OGDH maintain its structure or function under heat stress. An increase in temperature decreased the activity of puri ed OGDH, but the application of TCAIM protein had no effect on its function (Fig. 4D, Fig S4D). Moreover, the presence of TCAIM did not affect the thermal stability of the OGDH protein in vitro (Fig. 4E), while overexpression of TCAIM in HEK293T cells also showed no effect on the thermal stability of the OGDH protein in vivo (Fig. 4F). Interestingly, TCAIM seemed to bind poorly to heat-inactivated OGDH (Fig S4E), suggesting that TCAIM may not recognize misfolded OGDH protein.
TCAIM reduces OGDH protein level and the activity of OGDHC in vivo Next, we investigated whether TCAIM affects OGDHC activity in vivo. Transient or stable overexpression of TCAIM in HeLa or HEK293T cells signi cantly reduced the enzymatic activity of OGDHC ( Fig. 5A-D).
We also noticed a signi cant reduction in OGDH protein levels with transient overexpression of TCAIM in HeLa cells, whereas the OGDH protein level only slightly decreased in HeLa cells stably overexpressing TCAIM ( Fig. 5E and Fig S5A). A similar pattern of OGDH protein levels was also observed in TCAIM transiently or stably overexpressed in HEK293T cells (Fig S5B and C). Combined with our in vitro ndings ( Fig. 4), these results suggested that the decrease in OGDHC activity is most likely due to the reduction of OGDH protein, but not a direct inhibition of the enzyme complex (Fig S5D-G). In line with these ndings, OGDH protein levels, as well as the enzymatic activities of OGDHC, were signi cantly increased in liver lysates of Tcaim KO mice compared to WT mice ( Fig. 5F and G). We wondered whether TCAIM overexpression affects the proteome of mitochondria and decreases the levels of other mitochondrial proteins. To test this possibility, we examined the other two subunits of OGDHC, as well as irrelevant proteins, such as SIRT3, PHB1, and COX4. The levels of the tested mitochondrial proteins remained unchanged, with the exception of DLST, the E2 subunit of OGDHC, was also reduced in HeLa cells. Similarly, we did not observe an increase in the protein levels of the other tested candidates in liver lysates of Tcaim KO mice, compared to those of WT mice ( Fig. 5F and Fig. 5H).
Single particle cryo-EM resolves the binding between TCAIM and native OGDH Since TCAIM binds to functional OGDH in vivo and in vitro, and this binding has little effect on OGDH function directly, but rather mediates OGDH degradation, we decided to elucidate how TCAIM binds to native OGDH. To evaluate how OGDH interacts with TCAIM, we established the structure of the TCAIM-OGDH complex using single-particle cryo-EM at an overall resolution of 2.86 Å (Fig S6A-C, Table S3).
With apoOGDH (hsOGDH only, PDB: 7WGR, Fig. 6A) structure modelling and alpha-fold predicted TCAIM structure modelling, we built an atomic model of the hsOGDH-TCAIM complex (Fig. 6B). In the hsOGDH-TCAIM complex, OGDH presented a con guration similar to that of apoOGDH, while TCAIM was depicted as a triangle lying on the α/β1 domain of one molecule of the OGDH dimer (Fig. 6B). TCAIM adopts a three-repeat ααβββ-fold conformation, with one αβββ following a long α helix. The rst and second repeats closely interacted with OGDHa ( Fig. 6C/F). Among these binding sites, the C-terminal beta-strand (aa 467-490) displayed the closest binding distance to the surface of OGDHa and formed a hydrogen bond between Ser453 of TCAIM and Glu399 of OGDH (Fig. 6D). This binding site in TCAIM is conserved among different species, from Drosophila to humans ( Fig. 6G and Fig S6D). In addition, Glu399 in OGDH homologues is conserved between Drosophila and humans ( Fig. 6H).
In the hsOGDH-TCAIM complex, the density of the cofactor TPP and Mg2+ could be clearly observed in the predicted TPP binding pocket compared to apoOGDH (Fig. 6C). In the pocket, well-de ned Mg2+ ions were coordinated with the side chains of Asp430 and Asn463, which are conserved in OGDH homologues.
We asked whether the binding of TCAIM affects the structure of OGDH and compared the structures of the OGDH dimer and the TCAIM-bonded OGDH dimer. No obvious difference was observed between these two structures with a RMSD of 0.3 Å (Fig. 6E). This result suggests that the binding of TCAIM alone does not change the structure of OGDH, nor does it rely on conformational changes in OGDH.

TCAIM mediates OGDH degradation in a mtHSP70dependent manner
We next aimed to decipher the mechanism by which TCAIM downregulates the protein level of OGDH. We rst checked the mRNA level of OGDH in TCAIM transiently overexpressing HeLa cells compared to that of EGFP transiently overexpressing cells (Fig. 7A), and the mRNA level of Ogdh from the livers of Tcaim KO mice and their WT littermates (Fig. 7B). We found that the mRNA level of OGDH showed no signi cant change in either case, and thus, the change in transcription could not be the cause of the decrease in OGDH protein levels regulated by TCAIM.
Thus, we examined the mechanism by which TCAIM mediates the degradation of OGDH. Because TCAIM is a DNAJ protein, we rst tested whether mitochondrial HSP70 (HSPA9) is necessary for TCAIMmediated OGDH degradation. Indeed, HSPA9 knockdown attenuated the OGDH reduction (Fig. 7C). Moreover, the knockdown of LONP1 slightly attenuated the effect of TCAIM (Fig. 7D), but not knockdown of AAA Proteases CLPP or AFG3L2( Fig S7A&B). Interestingly, applying 20 S protease inhibitor, MG132 also failed to attenuate the degradation of OGDH, indicating this mechanism is likely through a 20S proteasome independent pathway (Fig S7C). These results indicate that the degradation of OGDH mediated by TCAIM may depend on the LON protease system, but not on the ubiquitination and CLP protease systems.
From our structural data, the last three beta-sheets of TCAIM displayed the closest distance to the surface of OGDH, indicating that this region may be responsible for TCAIM-OGDH binding. To test this hypothesis, we transiently overexpressed TCAIM mRNA that lacks this region (C-terminal beta sheet deletion CBD-TCAIM) to evaluate its effect on OGDH protein levels. Based on the prediction by Alphafold 17 , lack of this region does not cause dramatic structure change ( Fig S7D). However, the absence of this region signi cantly reduced the interaction between TCAIM and OGDH and abolished the effect of TCAIM on the protein level of OGDH ( Fig. 7E-F). Moreover, the absence of this region signi cantly attenuated the ability of TCAIM to reduce the maximum oxygen consumption rate in HeLa cells (Fig. 7G & Fig S7E).

Discussion
The mitochondrial proteostasis system plays an important role in maintaining mitochondrial function and ageing 18 . In this study, we found that TCAIM, a novel DNAJ/Hsp40 protein, regulated metabolism, especially mitochondrial functions and lipid metabolism, at the cellular and organismal levels. Interestingly, TCAIM speci cally targeted and mediated the degradation of native OGDH in a HSPA9-LONP1-dependent way. This nding suggests that besides maintaining the mitochondrial proteome and functions, the proteostasis system may also be directly involved in metabolic regulation, probably through selective degradation of native metabolic enzymes.
The OGDHC protein complex plays a unique and important role in the TCA cycle, as it is one of the ratelimiting enzymes of the citric acid cycle, and has also been reported to be the major site of ROS generation in the TCA cycle 19 . Moreover, its substrate, α-ketoglutarate, links the TCA cycle and glutamine/amino acid catabolism and serves as an essential substrate for epigenetic modi cation 20 . However, in addition to allosteric regulation by its cofactors, substrates, and products, little is known about its post-translational regulation. We previously reported that its activities can be regulated by lysine acetylation, indicating the post-translational regulation of this enzyme 10 . In this study, we revealed a novel post-translational regulatory mechanism that determines the protein level of the E1 subunit of the OGDHC complex, and consequently affects the in vivo activity of this enzyme.
TCAIM was originally identi ed as a T-cell activation inhibitor by Sawitzki and colleagues, as its expression level was signi cantly reduced during T-cell activation 12,13 . A recent publication also reported a profound impact of TCAIM knock-in (KI) on T cell metabolism 21 . However, changes in TCAIM expression are coupled with changes in the expression of many genes involved in glycolysis, TCA cycle, and other metabolic pathways. These sequential modi cations in metabolic gene expression may compensate for and even reverse the primary function of TCAIM in metabolism. Therefore, it is not surprising that they observed that TCAIM KI inhibited cholesterol biosynthesis gene expression, but the serum cholesterol level decreased in our KO mouse 21 . Nevertheless, consistent with the decrease in serum cholesterol levels, the protein malonylation level was also reduced in the liver lysates of TCAIM KO mice.
Malonyl-CoA is a unique precursor of fatty acid synthesis 22 , and increasing malonyl-CoA levels by applying malonate leads to an increase in protein malonylation levels. Therefore, even though it is indirect, the protein malonylation level may re ect the malonyl-CoA level in the liver cells and indicate a change in lipid biosynthesis.
TCAIM contains a J domain structure as well as an HPD motif after its mitochondrial leading sequence, but it is not followed by a glycine/phenylalanine-rich domain. Further studies revealed that it binds to the mitochondrial Hsp70 (mtHsp70) homologue, HSPA9. Moreover, its regulatory function is dependent on the presence of HSPA9. Taken together, our data suggested that TCAIM belongs to the type III DNAJ protein or DNAJC family. Unlike the members of type I or type II DNAJ protein families, DNAJC proteins display diverse functions and do not seem to bind non-native clients 14 . TCAIM also displays a similar aspect: in vitro binding assays as well as cryo-EM results show that it binds to native and functional OGDH, while its binding to heat-inactivated OGDH decreases signi cantly. The binding of DNAJC proteins to their mature, folded proteins play roles, such as "remodeling" of large multiprotein complexes and affecting the stability of protein-protein interactions 23,24 . But to our knowledge, DNAJ proteins have not been reported to drive protein degradation through direct target recognition as TCAIM does. However, we cannot eliminate the possibility that protein complexes of the OGDHC will be remodeled, followed by the degradation of OGDH, a replaceable E1 subunit of this protein complex 25,26 .
Recent studies have shown that DNAJ protein could assist the degradation of substrates of E3 ubiquitin ligase by stabilizing the ligase or by enhancing the binding between the ligase and substrate 27,28 . These regulations of protein degradation by the DNAJ protein are indirect; recognition and degradation of target proteins depend on the traditional E3 ubiquitin-protein ligase complex, not on the DNAJ protein per se. Thus, in contrast to TCAIM-OGDH regulation, the DNAJ proteins in these two studies are indirectly involved in protein degradation and are not essential for recognition or degradation of the target proteins.
The mitochondrial HSP70 protein plays several important roles in mitochondrial proteostasis. It is essential for the import and folding of mitochondrial precursor proteins 29 ; it also promotes the degradation of misfolded proteins with the help of LONP1 30 . Although not found in mitochondria, cytosolic HSP70 protein facilitates the degradation of protein aggregates 31 . Therefore, it may not be surprising that we found that mtHSP70 plays a role in TCAIM-mediated OGDH degradation, that is mtHSP70 can facilitate its degradation of the native protein OGDH using TCAIM to recognize the target. LONP1 is an AAA + (ATPase associated with a variety of cellular activities) protease in the mitochondrial matrix and plays multiple roles in maintaining the mitochondrial proteome 32,33 . Recent research has also shown that it cooperates with the mtHSP70 chaperone system to assist mitochondrial protein folding 34 . Currently, there is no speci c inhibitor of LONP1. The peptide aldehyde MG132 is one of the best inhibitors of Lon protease, but it is much more effective against the 20S proteasome and inhibits protein degradation via the ubiquitin-proteasome pathway 9,35,36 . Although several studies have found that MG132 can inhibit LONP1 function, the mechanism of action of MG132 on LONP1 is unclear, and the inhibitory effect appears to be substrate dependent 33,37 . We did not observe any inhibitory effect of MG132 on OGDH degradation up to 40 µM, which had a deleterious effect under our cell culture conditions. Nevertheless, LONP1 plays a role in TCAIM-mediated degradation of OGDH and most likely functions downstream of mtHSP70 as a protease.
Selective protein degradation is a key mechanism of post-translational regulation in cells, and plays an important role in various biological processes, particularly signal transduction 38,39 . In mitochondria, protein degradation is likely to affect metabolism because mitochondria are a crucial hub for metabolites. However, little is known about whether this protein degradation can serve as a speci c regulation of certain metabolic pathways, or whether it can only globally impact all metabolism occurring in the mitochondria. This study demonstrated that TCAIM can speci cally target and mediate degradation of OGDH, highlighting the importance of speci c regulation of metabolism through selective protein degradation, and raising the possibility that the DNAJC family proteins may play an important role in this type of post-translational regulation of mitochondrial metabolism. Moreover, overexpression of the DNAJC protein could speci cally target and mediate the degradation of a client protein, which could become a novel strategy for targeted protein degradation.
Several questions need to be addressed in the future. We found that refed after fasting upregulated TCAIM expression, suggesting that TCAIM may play a role in the regulation of nutrient metabolism. However, we still lack direct evidence regarding the circumstances or stress conditions in which this mechanism is regulated and becomes essential. We tested different tissue culture conditions, including low and high glucose levels, with or without glutamine, and applying a pro-oxidant or antioxidant. None of these modi ed the effect of TCAIM on OGDH in TCAIM-overexpressing cells (data not shown).
Moreover, even though CompPASS analysis is good at identifying speci c interacting proteins by reducing false positive data in IP-mass spectrometry experiments, it may remove real speci c interactions 10,40 . Therefore, despite our data demonstrating that OGDH is the most speci c interacting protein of TCAIM, we cannot eliminate the possibility that TCAIM has more binding proteins than OGDH and regulates their functions through a similar protein degradation mechanism or through completely different pathways.
In conclusion, the present study demonstrated that the mitochondrial proteostasis system could speci cally regulate metabolism mediated by the DNAJC protein TCAIM by speci cally binding to native OGDH and driving its degradation in a HSPA9-LONP1-dependent manner. This nding not only highlights a novel function of mitochondrial proteostasis in metabolism but also indicates a new eld of DNAJCmediated post-translational regulation of native proteins. Moreover, the potential of this new native protein degradation mechanism should be further explored in basic and translational research.

Methods
Cell Lines and Reagents HEK 293T and HeLa cell lines were acquired from ATCC and cultured in DMEM/high glucose (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (FBS; Gibco, Waltham, MA, USA), 100 U/mL penicillin, and 100 mg/mL streptomycin (basal media, Shanghai, China). All cell lines tested negative for Mycoplasma. Cells were grown in a humidi ed incubator at 37 ℃ with 5% CO 2 .  40 . Lentiviruses were generated by transfection with the target or negative control vectors using the packaging plasmid mix and a ratio of 1:2 polyethyleneimine (PEI) (Polyscience, USA) into 293T cells. Pseudovirus was harvested 48 h later, ltered, and used to transduce HeLa cells in the presence of 10 µg/mL polybrene. Infected HeLa cells were maintained in 1 µg/ml puromycin for three days and expanded to perform the experiments. Transfection e ciency was con rmed by western blotting.

Real-time PCR
Total RNA was isolated from prepared cell samples using TRIzol (TianGen, Beijing, China) and reverse-

Transfection of mRNA
Cells were seeded in a 6-well-plate to 70-90% con uency and transfected with 1-2 µg/well mRNA using Lipofectamine 2000 (11668019, Thermo Fischer Scienti c). Complexes at an mRNA (µg) to reagent (µl) ratio of 1:2 were prepared for most cell lines. In vitro OGDH activity OGDH activity was measured by 2,6-dichlorophenolindophenol (DCPIP) reduction rate at 610 nm in a reaction with 60 µg OGDH, 50 mM KH 2 PO 4 (pH 7.5), 0.5 mM thiamine pyrophosphate (ThDp), 1.0 mM MgCl 2 and 0.08 mM DCPIP at 37 ℃. After equilibrating in a microplate reader (SYNERGY H1, BioTek) for 3-5 min, 2 mM α-ketoglutarate was added to three replicate mixtures to initiate the reaction, and the nal volume was 200 µL. OGDH activity is shown as the DCPIP reduction rate based on the enzyme kinetics curve at 610 nm for the rst 5 min. For the heated OGDH assay, 2 µM OGDH was preheated at 42 or 45 ℃ for 10 min and then incubated with 10 µM TCAIM or BSA protein for 10 min. BSA and TCAIM proteins alone were negative controls. For the TCAIM heat-inactivation assay, 10 µM TCAIM was preheated at 60 ℃ for 10 min and added to the OGDH activity assay system to evaluate its effects on OGDH activity.

Cellular thermal shift assay
For Km and Vmax of the OGDH assay, 2 µM OGDH was incubated with or without 10 µM TCAIM for 20 min in advance. The OGDH activity was measured at different concentrations of α-ketoglutarate. The initial velocity (V0) at different concentrations of α-ketoglutarate was analyzed using Prism for Km and Vmax values.
In vitro OGDHC activity OGDHC activity assay using cell lysate 293T cells and HeLa cells transfected with 2 µg TCAIM mRNA were washed twice with PBS and collected with lysis buffer (50 mM Tris pH8.0, 150 mM NaCl, 1% Chaps, 1 mM PMSF, and complete EDTA-free protease inhibitor cocktail). The cells were lysed on ice for 20 min and the supernatant was separated by centrifugation (4 ℃, 12,000 × g) for 20 min. Protein concentrations were determined using Bradford assay. Whole cell lysate (20 µg) was added to the OGDHC activity system as the in vitro activity described previously. OGDHC activity was measured by the NADH production rate at 344 nm excitation / 460 nm emission from 4 mM α-ketoglutarate. Normalized OGDHC activity was indirectly determined by the ratio of total enzyme activity to the OGDH protein expression level.

Protein expression and puri cation
Recombinant OGDH and TCAIM strains were cultured at 37 ℃ in 2 L LB medium supplemented with 50 µg/mL kanamycin and 35 µg/mL chloramphenicol. Protein expression was induced at OD 600 ≈ 0. Amicon-Ultra-4 10K centrifugal device from Millipore, ash-frozen in liquid nitrogen, and stored at −80 ℃ until use.
For OGDH-TCAIM complex puri cation, the OGDH and TCAIM cell pellets were combined and resuspended in the lysis buffer above, followed by the same puri cation procedure. After Ulp1 protease digestion, the sample was loaded onto a HisTrapTM HP column twice and washed with 65 mM, 125 mM,

Sample preparation and cryo-EM data acquisition
Four microliters of freshly puri ed OGDH or OGDH-TCAIM144 at a concentration of 1.5 mg/mL was applied to glow-discharged 300-mesh Quantifoil R1.2/1.3 holey carbon grids (Quantifoil N1-C14nCu30-01). Grids were blotted for 3.0 s at 4 ℃ and 100% humidity on an TFS Mark IV Vitrobot before being plunge-frozen in liquid ethane cooled by liquid nitrogen. Frozen OGDH grids were transferred to a TFS Titan Krios electron microscope (Thermo Fisher Scienti c) operating at 300 kV, equipped with a Gatan BioQuantum energy lter (slit width 20 eV). Images were recorded using a K2 Summit direct electron detector in super-resolution mode. Data acquisition was performed using the SerialEM 3.8 41  Cryo-EM data processing The drift correction of all image was performed using MotionCor2 42 with 2 × 2 binning, and a doseweighted sum of all frames from each movie was used for all image-processing steps. After whole-image CTF estimation using CTFFIND-4.1 43 , the remaining steps were performed using cryoSPARC 44 . Particles were auto-picked, and after several rounds of 2D classi cation, good particles were selected for further 3D analysis. These particles were used to generate initial models for 3D classi cation and 3D re nement.
All re nements followed the gold-standard resolutions estimated based on the gold-standard Fourier shell correlation (FSC) = 0.143 criterion. The number of particles in each dataset and other details related to the data processing are summarized in Supplementary Fig S6 and Fig S7. Model building, re nement, and validation Atomic models of OGDH and OGDH-TCAIM144 were manually built and adjusted using COOT 45 . The models were then subjected to global re nement and minimization in real-space re nement using PHENIX 46 with a secondary structure.

Metabolic cage and bodyweight measurements
TCAIM heterozygous knockout mice were bred for two generations to obtain TCAIM homozygous knockout mice and wild-type littermates. All animals were raised in an SPF environment, and their bodyweight was recorded twice per week after weaning. They were transferred to the Comprehensive Lab

Mitochondrial metabolism measurement using Seahorse instrument
The oxygen consumption rate was determined using an Agilent Seahorse XF Cell Mito Stress Test Kit (Agilent Technologies, USA, 103015-100), and glycolytic function was determined using an Agilent Seahorse XF Glycolysis Stress Test Kit (Agilent Technologies, USA, 103020-100). HeLa cells were plated at 1 × 10 4 cells/well on a Seahorse XF Cell Culture Microplate (Agilent Technologies, USA, 102416-100). One hour before the assay, the medium was changed to 180 µL/well of OCR XF base medium (Seahorse XF DMEM (Agilent Technologies, USA, 103575-100), 1 mM pyruvate, 2 mM glutamine, 25mM glucose; pH 7.4) or ECR XF base medium (Seahorse XF DMEM, 1 mM pyruvate, 2 mM glutamine; pH 7.4). The cell culture microplate was placed in a 37 ℃ non-CO 2 incubator for 1 h. Different compounds were loaded into the appropriate ports of a hydrated sensor cartridge (OCR: 2 µM oligomycin, 0.75 µM FCCP, 0.5µM Rotenone and antimycin A; ECR: 10mM glucose, 2µM oligomycin, 50mM 2DG). Oxygen respiration status and pH change were then collected and analyzed using Agilent Seahorse XFe/XF analyzers, according to the manufacturer's instructions.

ROS measurement
Cells were trypsinized, collected by centrifugation for 5 min at 500 x g, and then incubated with MitoSOX (5 µM) (Thermo Scienti c, USA, m36008) diluted in DMEM for 10 min at 37 ℃. After staining, the cells were washed with PBS and analyzed using a BD FACSVerse ™ Flow Cytometer with 510 nm excitation and 580 nm bandpass emission lters. The data were analyzed using FlowJo V10.
Mitochondrial membrane potential Cells were trypsinized, collected by centrifugation for 5 min at 500 x g, and then incubated with JC-1 (2 µM) (Thermo Scienti c, USA, M34152) diluted in DMEM for 30 min at 37 ℃. After staining, the cells were washed with PBS and analyzed using a BD FACSVerse™ Flow Cytometer at 488 nm excitation with 527 nm and 586 nm bandpass emission lters. The data were analyzed using FlowJo V10.

Mitochondrial morphology
Cells were seeded on vitreous cell culture to 30-50% con uence and then stained with 50nM MitoTracker Red (Thermo Scienti c, USA) in DMEM for 30 min at 37 ℃. After washing with PBS, the cells were recovered in complete media for 1 h, and images were acquired with a Leica DMi8 uorescent microscope using a 100X oil lens.

Mouse metabolomics analysis
Whole blood samples were collected from heart punctures, then incubated at room temperature for 1 h, followed by 10 min of centrifuging at 1000 x g. Serum was collected by carefully transferring the supernatant into tubes, and samples were stored at -80 ℃ until measurement. 300µL of Methanol (containing 5 µg/mL 2-Chloro-L-phenylalanine as an internal standard) was added to 100µL of each serum and then mixed using a vortexer for 1 min. Then the mixture was centrifuged at 13,000 rpm, 4 ℃ for 10 min. The supernatant was then transferred to a sample vial for detection. An inhouse quality control (QC) was prepared by mixing equal amounts of each sample and passing through a Mass spectrometry was performed in both positive and negative ion modes. The optimized parameters were as follows: Capillary voltage: 3.5 kV; drying gas ow: 10 l/min; gas temperature: 325 ℃; nebulizer pressure, 20 psig; fragmentor voltage: 120 V; skimmer voltage: 45 V; mass range: m/z 50-3000.
Dietary treatment, food intake 7 days prior to dietary treatment, mice were individually housed. Fasting for 20 h was accomplished by placing mice in a new cage without food and with ad libitum access to water. Refeeding entailed allowing fasted mice ad libitum access to normal chow for 8 h.
Primary hepatocyte isolation and culture Primary hepatocytes were obtained from 2-to 3-month-old mice. The mice were killed by carbon dioxide, and the thoracic and abdominal cavities were opened. A ow of perfusion buffer (HBSS without Ca/Mg, 0.75% NaHCO 3 , 0.06mM EDTA) began through the ventriculus sinister, and the portal vein was immediately cut. When the blood was removed, the liver was perfused with collagenase buffer (HBSS with Ca/Mg, collagenase type 2 (Worthing, USA, LS004176) (to light medium brown), 5mM CaCl 2 , 0.75% NaHCO 3 ). Primary hepatocytes were resuspended in DMEM/high glucose, passed through a 70µm cell strainer, and centrifuged for 2 min at 520rpm. After another wash with DMEM/high glucose by centrifugation for 2 min at 520 rpm the resulting hepatocytes were plated in a 10 cm dish coated with collagen type I.

Serum cholesterol measurement
Cholesterols were measured using a Cholesterol Assay Kit (Abcam ab65390) according to the manufacturer's instructions. Brie y, sera were obtained as described in the metabolomic analysis method, and low-density or very-low-density lipoprotein (LDL/VLDL) was separated from high-density lipoprotein (HDL) by precipitation and centrifugation. Free cholesterol was incubated without cholesterol esterase, while total cholesterol was obtained with cholesterol esterase in the assay buffer provided by the kit. After   TCAIM has no signi cant effect upon OGDH and OGDHC activities in vitro A. OGDH activity was measured by DCPIP reduction rate at 610 nm. Puri ed TCAIM proteins (red), 95 ℃ heat-treated TCAIM (gray) were added to the OGDH activity assay system to evaluate their effects on OGDH activity. Buffer and TCAIM proteins alone were used to replace OGDH for the same activity assay as a negative control (**p < 0.01, n = 3).
B. The Km and Vmax of OGDH with (red) or without (blue) 10 μm TCAIM was calculated by the V0 at different concentrations of α-ketoglutarate using an in vitro assay.
C. The Km and Vmax of OGDHC with (red) or without (blue) 10 μm TCAIM is calculated by the V0 at different concentrations of α-ketoglutarate under different pH conditions using an in vitro assay. D. OGDH (Blue), 42 ℃ -treated OGDH incubated with protein suspension buffer, TCAIM or BSA protein (red), as well as Buffer, BSA and TCAIM proteins alone (grey), as negative controls, measured for OGDH TCAIM decreases OGDH protein level and OGDHC activities in vivo A. OGDHC activities were calculated using OGDHC assay system from 20 μg of puri ed cell lysates of TCAIM mRNA-transfected HeLa cells or that of Luciferase mRNA-transfected HeLa cells (p < 0.05, n = 6). B. OGDHC activities were calculated using OGDHC assay system from 20 μg of puri ed cell lysates of TCAIM mRNA-transfected HEK293T cells or that of Luciferase mRNA-transfected HEK293T cells (p < 0.0001, n = 12).
C. OGDHC activities were calculated using OGDHC assay system from 20 μg of puri ed cell lysates of TCAIM stable overexpression HeLa cells or that of control HeLa cells (p < 0.0001, n = 17).
D. OGDHC activities were calculated using OGDHC assay system from 20 μg of puri ed cell lysates of TCAIM stable overexpression HEK293T cells or that of control HEK293T cells (p < 0.0001, n = 17).
E. The protein levels of OGDH, TCAIM-HA and GAPDH in HeLa cells that transiently overexpressed EGFP or TCAIM-HA mRNA were tested by immunoblotting (left). The band intensities of OGDH were quanti ed using ImageJ and normalized to the band intensities of GAPDH (right) (p < 0.01, n = 3).
F. The protein levels of OGDH, DLST, DLD, SIRT3 and GAPDH in the liver lysates of TCAIM KO mice and that of WT littermates were tested by immunoblotting.
H. Protein levels of OGDH, TCAIM-HA and SIRT3, COX4, PHB1 in HeLa cells that transiently overexpressed EGFP or TCAIM-HA mRNA were tested by immunoblotting.
B. Cryo-EM structure of hsOGDH-TCAIM complex, density map (left panel), hsOGDH-TCAIM complex structure (middle and right panels), the structure shown as cartoon, chain-a colored in blue, chain-b in violet and TCAIM in yellow.
C. Cartoon representation of OGDH-TCAIM complex, chain-a colored in green, chain-b in blue and TCAIM by rainbow. TPP shown as sticks.
D. Close-up view of interaction site of OGDH and TCAIM, residues involved in interactions shown as sticks.
E. Superimposition of OGDH apo and OGDH from the OGDH-TCAIM complex.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.