Structure–function analyses of the G729R 2-oxoadipate dehydrogenase genetic variant associated with a disorder of l-lysine metabolism

2-Oxoadipate dehydrogenase (E1a, also known as DHTKD1, dehydrogenase E1, and transketolase domain-containing protein 1) is a thiamin diphosphate-dependent enzyme and part of the 2-oxoadipate dehydrogenase complex (OADHc) in l-lysine catabolism. Genetic findings have linked mutations in the DHTKD1 gene to several metabolic disorders. These include α-aminoadipic and α-ketoadipic aciduria (AMOXAD), a rare disorder of l-lysine, l-hydroxylysine, and l-tryptophan catabolism, associated with clinical presentations such as developmental delay, mild-to-severe intellectual disability, ataxia, epilepsy, and behavioral disorders that cannot currently be managed by available treatments. A heterozygous missense mutation, c.2185G→A (p.G729R), in DHTKD1 has been identified in most AMOXAD cases. Here, we report that the G729R E1a variant when assembled into OADHc in vitro displays a 50-fold decrease in catalytic efficiency for NADH production and a significantly reduced rate of glutaryl-CoA production by dihydrolipoamide succinyl-transferase (E2o). However, the G729R E1a substitution did not affect any of the three side-reactions associated solely with G729R E1a, prompting us to determine the structure–function effects of this mutation. A multipronged systematic analysis of the reaction rates in the OADHc pathway, supplemented with results from chemical cross-linking and hydrogen–deuterium exchange MS, revealed that the c.2185G→A DHTKD1 mutation affects E1a–E2o assembly, leading to impaired channeling of OADHc intermediates. Cross-linking between the C-terminal region of both E1a and G729R E1a with the E2o lipoyl and core domains suggested that correct positioning of the C-terminal E1a region is essential for the intermediate channeling. These findings may inform the development of interventions to counter the effects of pathogenic DHTKD1 mutations.

iments by using fibroblast cell lines derived from individuals with the (c.2185G3 A, p.G729R) DHTKD1 mutation (1,3). Also, the recent findings by Xu et al. (6) provided evidence for a novel relationship between AMOXAD and inherited peripheral neuropathy associated with CMT2Q, in addition to a mitochondrial dysfunction reported by these authors earlier (5). However, the molecular mechanisms leading to the pathogenesis of metabolic disorders associated with DHTKD1 mutations are poorly understood, as is the function of the 2-oxoadipate dehydrogenase complex (OADHc) in the L-lysine degradation pathway.
It has been assumed that similar to the mitochondrial 2-oxoglutarate (OG) dehydrogenase complex (OGDHc, also known as the ␣-ketoglutarate dehydrogenase complex, M r of 5 ϫ 10 6 ) in the tricarboxylic acid (TCA) cycle, the thiamin diphosphate (ThDP)-dependent E1a (an ␣ 2 homodimer; EC 1.2.4.2; 103.077 Da/per monomer) is the first component of the OADHc, which along with its specific dihydrolipoamide glutaryltransferase (E2a, so far unidentified gene localization) and dihydrolipoamide dehydrogenase (E3) carry out the principal reactions for glutaryl-CoA formation according to the overall reaction in Equation 1 and detailed chemistry in Equations 2-5 as follows: Our recent in vitro studies led to the remarkable recognition that E1a could recruit the dihydrolipoamide succinyltransferase (E2o) and E3 components of the OGDHc for its function; however, the assembled OGDHc and OADHc are not functionally redundant (10,11). Each complex displayed preference for its own substrate in accord with its function in the TCA cycle and in the L-lysine degradation pathway. Our further studies in vivo point to the existence of a hybrid 2-oxo acid dehydrogenase complex, in which both E1a and E1o participate along with E2o and E3 (9). A cross-talk between the OGDHc in the TCA cycle and OADHc in L-lysine catabolism was suggested to be of potential relevance to EoE (7). Whole-exome sequencing and rare variant burden analysis revealed an overabundance of putative, potentially damaging DHTKD1 mutations in multiple individuals with EoE, implicating DHTKD1 in the genetic eti-ology of EoE (7). The authors also identified seven potentiallydamaging genetic variants in the DHTKD1 homologue OGDHL, which encoded a putative 2-oxoglutarate dehydrogenase-like protein (7). It was suggested that mutations in both DHTKD1 and OGDHL, and their role in impaired mitochondrial function, are molecular pathways linked to progression of allergic inflammation (7).
Using multiple approaches, we herein unravel the effect of the common (c.2185G3 A, p.G729R) DHTKD1 mutation on protein-protein interactions in the human OADHc. Our major findings are as follows. The G729R E1a substitution does not affect the rate of OA decarboxylation in Equation 2, or the rates of the E1a-ThDP-enamine's three side-reactions (including reactive oxygen species (ROS) production). Rather, the bulk of the rate reduction in NADH production with the G729R E1a substitution (k cat, NADH ϳ24-fold lower compared with E1a) is the result of the slower rate of glutaryl transfer from the S8-glutaryldihydrolipoyl-E2o to CoA (k glutaryl-CoA ϳ37-fold slower compared with E1a) in Equation 4. Hydrogen/deuterium exchange MS (HDX-MS) and chemical cross-linking MS (CL-MS) revealed local conformational changes induced by the G729R E1a substitution in its C-terminal region that could affect E1a-E2o interactions. The HDX-MS experiments indicate that the interaction of E1a with E2o did not induce largescale conformational changes in E1a; however, local backbone amide proton perturbations were indeed observed in the N-terminal region, the ThDP and Mg 2ϩ -binding fold, and in the C-terminal region. In contrast, the HDX-MS data provide no clear evidence for major changes in deuterium uptake upon binding of G729R E1a to E2o at any hydrogen exchange time. The great number of cross-links formed between the C-terminal region of both E1a and G729R E1a with the E2o lipoyl domain (LDo) and E2o core domain (CDo) suggested that a correct positioning of the C-terminal region of E1a with E2o is essential for intermediate channeling in OADHc.

G729R E1a is expressed in patient-derived fibroblasts with ␣-aminoadipic and ␣-ketoadipic aciduria
The compound heterozygous common mutation (c.2185A3G (p. G729R)) in DHTKD1 was identified in a few cases with AMOXAD and is predicted to be disease-causing; however, the mechanisms of disease are poorly understood (1,3). Earlier, we reported four cases that were heterozygous for this mutation (3). The DHTKD1 mRNA expression levels were reported as normal; however, the level of the DHTKD1-encoded E1a protein was low or undetectable suggesting that the protein level could be affected. We have reinvestigated E1a protein expression using primary antibodies against E1a (GeneTex Inc., Irvine, CA) (Fig. 1). It was found that fibroblasts from an individual that carried a heterozygous (c.2185A3 G (p.G729R) and WT) DHTKD1 mutation had a normal level of E1a expression compared with that in control fibroblasts (Fig. 1A) (3). A lower level of E1a expression was revealed in the case compound heterozygous for the (c.2185A3 G (p.G729R) and c.86dup (p.Y63IfsX3)) DHTKD1 mutations (Fig. 1A); however, the protein level was detectable. Here, we also studied E1a protein Consequences of (c.2185G3 A (protein G729R)) DHTKD1 mutation expression in fibroblasts of a case compound heterozygous for (c.2143C3 T (p.R715C) and c.2185A3 G (p.G729R)) DHTKD1 (Fig. 1B). To evaluate whether the (c.2143C3 T (p.R715C) and c.2185A3 G (p.G729R)) DHTKD1 mutations rendered the E1a R715C/G729R protein level temperature-sensitive, we cultured the fibroblasts as well as control fibroblasts at 30, 37, and 40°C (Fig. 1B). The E1a R715C/G729R protein expression was lower in the DHTKD1-deficient cell line, but it was not temperature-sensitive (Fig. 1B). The data clearly indicated that on (c.2185A3 G (p.G729R)) DHTKD1 mutation, the E1a protein is expressed at a detectable level.
Thermal shift analysis of the purified recombinant E1a and G729R E1a proteins was used to determine whether the G729R E1a substitution affects protein stability (Fig. 1C). The calculated values of T m ϭ 48.85 Ϯ 0.12°C (E1a) and of T m ϭ 48.87 Ϯ 0.0°C (G729R E1a) were similar. This demonstration not only ruled out a stability issue being the source of the pathogenicity for the G729R E1a protein, but it also ruled out the concept of a misfolding disease resulting from the G729R E1a substitution. We next characterized the functional consequences of the c.2185G3 A DHTKD1 mutation on the encoded G729R E1a protein.

G729R E1a substitution affects the overall OADHc activity but not the E1-specific activity or any of the three side-reactions emanating from the E1a-ThDP-enamine intermediate
There are several intermediates on the pathway in Scheme 1 converting OA to glutaryl-CoA, those covalently bound to ThDP on the E1a and S8-glutaryldihydrolipoyl-E2o. Below, the fate of each intermediate and product will be compared for E1a and G729R E1a to help pinpoint the origin of the rate reduction produced by the pathogenic substitution. Our recent findings revealed that the E1a has recruited the E2o component of the TCA cycle OGDHc for its function in the L-lysine degradation pathway generating glutaryl-CoA (10 -12). Although the E3 component is also essential for this function, the same E3 enzyme serves all such 2-oxo acid dehydrogenase complexes in a given cell; hence, its participation is less surprising.
First, it was determined that the G729R E1a variant assembled with E2o and E3 into OADHc catalyzes the conversion of OA and NAD ϩ to glutaryl-CoA and NADH (ϩH ϩ ) according to the overall reaction in Equation 1 and Scheme 1. Indeed, the catalytic efficiency of the G729RE1a of 7.7 ϫ 10 3 M Ϫ1 s Ϫ1 was ϳ50-fold smaller than that for E1a, whereas similar values of K m, OA were determined in the overall NADH assay (Table 1). Next, an E1-specific kinetic assay was employed using an external chromophoric two-electron acceptor 2,6-dichlorophenolindophenol to replace E2o and E3, and this led to similar E1-specific activities of E1a (0.82 Ϯ 0.19 mol⅐min Ϫ1 ⅐mg E1a Ϫ1 ) and G729R E1a (0.67 Ϯ 0.13 mol⅐ min Ϫ1 ⅐mg G729R E1a Ϫ1 ). It was therefore concluded that the G729R E1a substitution did not affect the rate of decarboxylation of the OA, i.e. the formation of the E1a-ThDP-enamine intermediate (Equation 2, Scheme 1, and Table 1).
In the absence of E2o and E3, the E1a-ThDP-enamine intermediate has three alternate pathways (see Scheme 1, lower right-hand quadrant for E1a-ThDP-enamine side-reactions), 3. The E1a expression in mouse liver was used as a positive control. For null allele, the E1a protein was not detected because of the premature stop codon in DHTKD1. GAPDH denotes glyceraldehyde 3-phosphate dehydrogenase. B, immunoblot analysis and quantification of DHTKD1-encoded E1a levels in control and in patient-derived fibroblasts with R715C/G729R E1a substitution. Top, immunoblot analysis and quantification of DHTKD1-encoded E1a levels at 30 and 37°C. CS denotes citrate synthase. Expression was normalized using CS. The average E1a/CS for control fibroblasts at 30°C was set to 1. Bottom, immunoblot analysis and quantification of DHTKD1-encoded E1a levels at 37 and 40°C. The average E1a/CS for control at 37°C was set to 1. Error bars indicate S.D. The molecular mass standards are indicated on immunoblots. C, thermal stabilities of E1a and G729R E1a (see under "Experimental procedures: for details).
Consequences of (c.2185G3 A (protein G729R)) DHTKD1 mutation two nonoxidative and the third oxidative, and all three scenarios have precedents in ThDP enzymology.
The simplest side-reaction (the result of C2␣-protonation of the enamine) is conversion of E1a to a 2-oxoadipate decarboxylase, yielding glutaric semialdehyde from a nonoxidative decarboxylation. The rate of formation of glutaric semialdehyde by G729R E1a was also unaltered compared with E1a and could not be considered an efficient E1a side-reaction ( Table 1).
Third, the products and kinetics of their formation were determined for both half-reactions of the oxidative side-reaction in Scheme 1, lower right-hand quadrant. In the reductive half-reaction, reduction of molecular O 2 to superoxide/H 2 O 2 by G729R E1a and by OADHc assembled with E2o and E3 was evaluated. We hypothesized that the disease-causing G729RE1a substitution that led to a reduced NADH production could affect the superoxide/H 2 O 2 production (ROS) by E1a and thereby could contribute to oxidative stress in mitochondria, in accord with findings reported by others (5,7). The expected primary product is superoxide anion radical generated via one-electron reduction of O 2 with concomitant production in the oxidative half-reaction of the C2(␣-hydroxy)-␦carboxybutylidene-ThDP radical-E1a in Scheme 1 (henceforth called the ThDP-enamine radical) (12,14,15). The X-band EPR spectra of the G729R E1a incubated aerobically with OA (10 mM) revealed formation of the ThDP-enamine radical species (Fig. 2B). The concentration of ThDP-enamine radical species in the G729R E1a-active centers was estimated as 0.3 M (0.1% occupancy of active centers). Earlier, similar concentrations of ThDP-enamine radical species were reported for E1a from OA (0.2-0.5 M (0.1-0.12% occupancy)) ( Fig. 2A) (12,14,15), indicating that this side-reaction was also unaffected by the G729R substitution. The H 2 O 2 detected in the reaction assay is a product of dismutation of superoxide radical anion, either spontaneously or by the superoxide dismutase present in the H 2 O 2 assay (see Scheme 1, lower right-hand quadrant). Again, similar catalytic efficiency of H 2 O 2 production resulted by E1a and by G729R E1a by themselves or when assembled into OADHc with E2o and E3: 2.67 ϫ 10 3 M Ϫ1 s Ϫ1 (E1a) and 1.47 ϫ 10 3 M Ϫ1 s Ϫ1 (G729R E1a) ( Fig. 2C and Table 1).
It could be concluded that neither the rate of decarboxylation in E1a nor any of the three side-reaction rates of the E1a-ThDP-enamine were significantly affected by the G729R E1a substitution. This suggested that we explore the possibility that assembly of the E1a and E2o components is compromised by the G729R E1a substitution, resulting in reduced overall activity.

G729R E1a substitution does not change the strength of the E1a-E2o interaction
Before turning to kinetic experiments involving both E1a and E2o components, we had to establish whether or not the strength of the E1a⅐E2o subcomplex was impacted by the G729R E1a substitution. To help identify the loci of, and to determine the strength of interaction between E1a or G729R E1a and E2o, truncated versions of the E2o were created: the E2o 1-173 di-domain comprising the LDo; the linker region and part of the CDo; and the C-terminal E2o 144 -386 core domain, consisting of the C-terminal part of the linker and the entire CDo (see Fig. 3 for E2o domain assignment). We once again used an approach created at Rutgers to introduce an external N-(1-pyrene) maleimide fluorescent label onto the unique Cys 37 in the E2o 1-173 di-domain and onto the unique Cys 178 in the E2o 144 -386 core domain (16). Titration of the pyrene-labeled E2o 1-173 di-domain and the pyrene-labeled E2o 144 -386 core domain by E1a and by G729R E1a enhanced the pyrene fluorescence intensity, suggesting that a more hydrophobic environment was created around the pyrene fluorophore on their binding (16). The following dissociation constants were calculated: K d ϭ 1.12 M (E1a) and K d ϭ 0.98 M (G729R E1a) from titration of the pyrene-labeled E2o 1-173 di-domain; and K d ϭ 3.98 M (E1a) and K d ϭ 4.79 M (G729R E1a) from titration of the pyrene-labeled E2o 144 -386 core domain. These data suggested that E1a binding to the E2o 1-173 di-domain and to the E2o 144 -386 core domain was not significantly changed by the G729R substitution. In comparison, for E1o, the values of K d ϭ 0.041 M for the pyrene-labeled E2o 1-173 di-domain and of K d ϭ 0.061 M for the pyrene-labeled E2o 144 -386 core domain were reported by us earlier (16), suggesting a different binding mode of E1a and E1o with the same E2o component.

Kinetic evidence for impaired reductive glutarylation of E2o by G729R E1a
The reductive glutarylation of E2o by G729R E1a and OA was studied using a model reaction where E2o was replaced by its LDo (E2o 1-87 ) in the reductive glutaryl transfer (see Equation 3 and Scheme 1 (17)(18)(19)). On decarboxylation of OA by G729R E1a in the reaction mixture containing LDo, both glutarylated and unglutarylated forms of LDo could be detected by ESI FT-MS at different times of incubation. For G729R E1a, the A, X-band EPR spectrum of the ThDP-enamine radical species on E1a from OA as reported earlier (11). B, X-band EPR spectrum of the ThDP-enamine radical species observed on G729R E1a from OA. G729R E1a (29.8 mg/ml, concentration of active centers of 289 M) in 100 mM HEPES (pH 7.5) containing 0.15 M NaCl, 0.5 mM ThDP, 2.0 mM MgCl 2 , and 5% glycerol was mixed with 10 mM OA aerobically. The mixture was immediately transferred into an EPR tube and was flash-frozen in liquid nitrogen after ϳ30 -40 s of incubation with 2-oxoadipate. C, H 2 O 2 activities of the E1a and G729R E1a. The rate of H 2 O 2 production by E1a and by G729R E1a assembled to OADHc was measured as reported earlier (12,14,15). Each value represents the mean Ϯ S.D. (n ϭ 4).

Consequences of (c.2185G3 A (protein G729R)) DHTKD1 mutation
fraction of reductively glutarylated LDo versus total LDo (glutarylated plus unglutarylated), when plotted against incubation time, enabled calculation of the k glutarylation of 1.8 s Ϫ1 (at 25°C) and k glutarylation of 3.3 s Ϫ1 (at 37°C) with only 50% of the LDo being glutarylated even after 17 min of the reaction (Fig. 4). When compared with a k glutarylation of 7.6 s Ϫ1 (at 37°C) for E1a in the same model reaction, it became evident that there is indeed a reduction in the rate of reductive glutarylation of E2o, but the reduction in the rate could not account for the loss of NADH production by G729R E1a. It should be noted that the calculated values of K m, LDo were rather similar for E1a (8.9 M) and G729R E1a (12 M). The value was not different from that determined for E1o (K m, LDo ϭ 5.4 M), the established partner of E2o (Fig. S1).

Transthioglutarylation from S8-dihydrolipoyl-E2o to CoA in the E2o core domain forming glutaryl-CoA is significantly impaired by the G729R E1a substitution
The next step in the mechanism after reductive glutarylation of E2o to form S8-glutaryl-dihydrolipoyl-E2o (as outlined in Scheme 1) is transfer of the glutaryl group to CoA (Equation 4). A method was developed to measure the rate of enzymatic synthesis of the glutaryl-CoA by MALDI-TOF MS. When measured for E1a and for G729R E1a, both assembled with E2o and E3 into OADHc, a much slower rate was in evidence for G729R E1a (Table 1) (20,21). At 10 min reaction time, the CoA present in the NADH assay was completely converted to glutaryl-CoA by E1a assembled with E2o and E3. In clear contrast, under the same experimental conditions, including reaction time, with the G729R E1a, both CoA (m/z ϭ 766.414) and glutaryl-CoA (m/z ϭ 880.284) could still be detected in the reaction assay with only ϳ40% of the CoA converted to glutaryl-CoA ( Fig. 5A and Fig. S2). The second-order rate constants of k ϭ 0.245 M Ϫ1 s Ϫ1 (E1a) and k ϭ 0.0066 M Ϫ1 s Ϫ1 (G729R E1a) were calculated from the time-dependent kinetic studies (Fig. 5B). Apparently, the bulk of the rate reduction in NADH production with the G729R E1a substitution (k cat, NADH ϳ24-fold lower compared with E1a) is the result of the slower rate of glutaryl transfer from the S8-glutaryldihydrolipoyl-E2o to CoA (k glutaryl-CoA ϳ37-fold lower compared with E1a). This provides clear evidence that the G729R E1a substitution encoded by the pathogenic DHTKD1 mutation affects intermediate channeling, i.e. inter-component assembly in the OADHc, because it is the rate of a reaction taking place exclusively on the E2o that is being affected.

HDX-MS reveals different deuterium uptake patterns for E1a and G729R E1a and implies multiple conformations in solution
The results so far summarized suggested that inter-component interactions are impaired by the pathogenic G729R E1a substitution. Before undertaking studies of E1a-E2o interactions, we turned to HDX-MS to study potential conformational changes resulting from the G729R E1a substitution, as may be reflected by changes in deuterium incorporation. Online digestion of the E1a by pepsin, followed by FT-MS analysis of the resulting peptic peptides, gave evidence for 63 peptides  The relative intensity of the glutarylated versus total LDo (glutarylated ϩ unglutarylated) was plotted versus time. The data points were fit to single exponential equation (y ϭ yo ϩ a ϫ (1 Ϫ e Ϫkt )), and the trace is a nonlinear regression fit to initial rate conditions. The rate constants were calculated from the initial slope.
Consequences of (c.2185G3 A (protein G729R)) DHTKD1 mutation ( Table S1) with many of them partially overlapping and providing 90% sequence coverage for E1a. Similar peptides resulted on pepsin digestion of the G729R E1a according to FT-MS analysis (data not shown). A brief interpretation of the results is merited in view of the large amount of information content. In Fig. 6, there are two types of plots. In Fig. 6A, they are presented as "mirror"' images of pepsinolytic peptides of E1a and G729R E1a with the sequence assignment on the x axis and time course of intensity change on the y axis as ⌬D values (%). Although the results appear to create continuous curves, in fact, each x value represents a distinct pepsinolytic peptide. The true power of the method and of the results is most pronounced in the difference plot (Fig. 6B), where corresponding values for E1a are subtracted from those for G729R E1a, producing a ⌬⌬D intensity difference on the y axis (in Da units). The comparative HDX-MS analysis of the E1a and G729R E1a presented in  Fig. S3 for percent deuterium incorporation of E1a and G729R E1a mapped onto the modeled structure of E1a). The difference HDX-MS plot presented in Fig. 6B indicates that whereas the G729R E1a substitution did not induce large-scale conformational changes, local backbone amide proton perturbations were indeed observed. A positive ⌬⌬D in the difference plot implies that some peptides in the G729R E1a are more open to deuterium uptake compared with E1a, whereas a negative ⌬⌬D suggests that some peptides in the E1a are more open to deuterium uptake than the G729RE1a. For example, peptides comprising residues 24 YQTERGVYGYRPRKPESREPQGAL 47 in the N-terminal region and residues 401 SPEEVVRATRLAFEYQRQFRK 421 near the ThDP-and Mg 2ϩ -binding region of G729R E1a displayed an obvious decrease in deuterium incorporation, indicating that those regions are likely less open to the deuterium uptake. In Fig. 6C, peptides that become less open for deuterium uptake on the G729RE1a substitution are color-coded in blue on the modeled E1a structure, whereas peptides more open for deuterium uptake are color-coded in red. The data reveal obvious changes in deuterium uptake on G729R E1a substitution. Also, Consequences of (c.2185G3 A (protein G729R)) DHTKD1 mutation deuterium uptake by the peptide comprising residues 401-421 in G729R E1a was reduced by 1.4 Da at 10 min of exchange, and even after 90 min of exchange, a 2.3 Da mass reduction still persisted (see Fig. 7 for deuterium uptake difference plot). The residue Lys 37 in peptide 24 -47 was identified as a "hot spot"linking site by CL-MS (see under "Chemical cross-linking of E1a or G729R E1a with E2o identifies different inter-component interaction loci"), suggesting regional conformational change in this region that perhaps impacts the interaction of E1a and E2o. In contrast, an increase in deuterium uptake was observed in the peptide 256 SATGDVLSHLTSSVDL 271 in G729R E1a suggesting that it is more open to deuterium exchange compared with E1a (Fig. 7). Importantly, peptides from the C-terminal region of G729R E1a comprising residues 680 ISGGEAKWLLQSG 692 , 719 MCDSAEEGVDGDTVN 733 (including the substitution site at G729), 797 VDPKKVKTLVF-CSGKHFYSLVKQ 819 , and 888 RLVGRPPLPVPAVG 901 , experienced a significant increase in deuterium uptake level compared with E1a, suggesting more accessibility to the deuterium in response to the G729R E1a substitution (Fig. 7, bottom  panels). These findings clearly signal conformational changes induced by the G729R E1a substitution.

Consequences of (c.2185G3 A (protein G729R)) DHTKD1 mutation Observation of EX1 HDX kinetic behavior further supports two solution conformations of E1a
A closer examination of the high-resolution MS signals revealed several peptides with bimodal mass envelope distributions that could be curve-resolved into two closely-spaced peaks, an indication of "EX1 kinetics" (22). Unlike those peptides behaving according to "EX2 kinetics," characterized by mass peaks gradually shifting to higher mass in MS-based analysis with peaks moving to the right, the spectrum of "EX1 peptides" displayed two populations: a higher m/z envelope, which has undergone an unfolding event (to the right), and a lower m/z envelope, which has not (to the left in the deconvoluted spectrum) (23,24). Among the "EX1 peptides" identified in this study are peptides encompassing residues 256 -271, 680 -692, and 797-819 (see time course of HDX for the three peptides with side-by-side presentation of deconvoluted spectra for E1a and G729R E1a in Fig. 8A). Occurrence of EX1 kinetics provides a direct measure of a conformational transition. These findings indicate that there likely are at least two conformations of E1a under physiological conditions, in accord with the interpretation of the difference mass spectrum in Fig. 6B. In each of these peptides, the G729R E1a substitution results in a shift to a more open conformation in these regions (more deuterium absorbed in the "open" conformation), which could compromise the formation of a catalytically-competent E1a conformation required for productive interaction between the E1a and E2o components in OADHc.
Where the quality of the data allowed, the first-order rate constant could be estimated for the conformational interconversion by fitting the change in the relative abundance to a simple rate expression ln([A]/[A] 0 ) ϭ Ϫkt, which in turn led to the half-time of unfolding (t1 ⁄ 2 ϭ ln2/k). As an example, for peptide 256 -271 at 0.5 min, 20% in E1a and 47% of G729R E1a have undergone an unfolding event, respectively. This corresponds to a 3-fold decrease in the t1 ⁄ 2 of unfolding of G729R E1a (3.4 min) compared with E1a (10 min).
These findings could also be placed in a broader context. Excellent X-ray crystallographic studies reported for the related Mycobacterium smegmatis 2-ketoglutarate decarboxylase (MsKGD (25,26)) have already alluded to multiple conformations of some of the peptides reported here, as indicated by the sequence alignment for selected peptides (see Fig. 8, B and C). For example, the peptide comprising residues 256 -271 is part of the E1a amino acid sequence, which on alignment with MsKGD corresponds to the MsKGD active center loop comprising residues 560 -580 (26). It was suggested that this flexible loop adopts at least three distinct conformations: an opened state that shows conformational disorder in the available crystal structures to allow the substrate to enter the active site; a closed "lid" conformation that sequesters the active site to isolate the Michaelis complex or the post-decarboxylation intermediate; and an outer conformation that creates a tunnel for incoming substrate acceptor (26).

HDX-MS of E1a or G729R E1a with E2o present
To address the question as to which site on E1a interacts with which site on E2o, and whether these loci could be changed by the G729R E1a substitution, we carried out HDX-MS analysis by comparing the deuterium uptake of the peptic peptides that originated from E1a and G729R E1a in the absence and in the presence of E2o (Fig. 9). For a complete set of deuterium uptake curves for peptic peptides originating from E1a, G729R E1a, and the E1a⅐E2o and G729R E1a⅐E2o subcomplexes, see Fig. S4. The major findings are as follows.
(i) The HDX-MS experiments indicate that the interaction of E1a with E2o did not induce large-scale conformational changes in E1a; however, local backbone amide proton perturbations were indeed observed.
(ii) In contrast to the evidence for the E1a-E2o interaction, the HDX-MS data provide no clear evidence for major changes in deuterium uptake upon binding of G729R E1a to E2o at any H/D exchange time points analyzed, suggesting only a transient

Consequences of (c.2185G3 A (protein G729R)) DHTKD1 mutation
interaction between the two components that could be too weak to be captured by the method used ( Fig. 9B and Fig. S4).
(iii) Importantly, as shown in Fig. 9, A and C, the region from the N-terminal end of E1a comprising residues 24 -47 experi-enced significant protection from deuterium uptake in the subcomplex with E2o, similar to E1o (corresponding to the region comprising residues 27-40) (16). Another region that experienced significant protection from deuterium uptake on compl-

Consequences of (c.2185G3 A (protein G729R)) DHTKD1 mutation
exation with E2o comprises residues 485-518, which according to a recently reported X-ray structure of human E1a is part of the novel extended linker region (residues 497-527) connecting two halves of the E1a monomer (27) (PDB number is not available yet). In the C-terminal region, the peptides comprising residues 589 -609, 646 -664, and 847-874 also experienced protection from deuterium uptake on interaction with E2o.

Chemical cross-linking of E1a or G729R E1a with E2o identifies different inter-component interaction loci
We next carried out chemical CL-MS in binary E1a⅐E2o and G729R E1a⅐E2o subcomplexes (16). The motivation to establish the loci of interaction in the binary subcomplexes was that neither the reported X-ray structures of the core domain of the E2s (28 -30) nor the 7-Å resolution X-ray structure of the fulllength mammalian E2o provided information regarding the lipoyl domain and subunit-binding region (31), presumed to interact with the E1 and E3 components. Also, there is not an X-ray structure for any of the 2-oxo acid dehydrogenase complexes due to their size (M r of 4 -10 million) and the flexible and distant attachment of the E1 and E3 components to the E2 core (32)(33)(34)(35)).
An MS-cleavable zero-length cross-linker (1,1Ј-carbonyldiimidazole (CDI) with a spacer length of 2.6 Å and C␣-C␣ distance of ϳ16 Å to be bridged) (36) was employed to react with the ⑀-amino side chain of the lysine residues of the full-length E1a or G729R E1a in subcomplex with E2o. Two types of cross-linked peptides were identified: (i) intra-component cross-links, i.e. two lysines within the same subunit or between two identical subunits being cross-linked (recall that E1a is a dimer, whereas E2o is probably a trimer), and (ii) the inter-component cross-links, where cross-linking between the E1a and E2o components takes place at a max-
(i) The comparison of the CL-MS data revealed numerous intra-component cross-links for both E1a (85 cross-links) and for G729R E1a (87 cross-links) (Tables S2 and S3). The most frequently identified intra-component cross-links were between E1a residues located in the N-terminal region (Lys 37 , Lys 72 , Lys 148 , and Lys 155 ), residues in the ThDP-and Mg 2ϩbinding region (Lys 300 , Lys 450 , and Lys 456 ), and those located in the C-terminal region (Lys 767 , Lys 818 , Lys 826 , Lys 827 , and Lys 852/ 854) that could reflect a spatial proximity of these regions in the E1a dimer assembly. However, there were also some unique intra-component cross-links identified that are specific for E1a or for G729R E1a only (Table 2). Thus, Lys 244 in E1a forms unique cross-links with Lys 450 and Lys 456 from the ThDP-and Mg 2ϩ -binding region, whereas in G729R E1a, the Lys 244 forms a unique cross-link with Lys 759 from the C-terminal region (Table 2). Also, in G729R E1a, the Lys 300 from the ThDP-and Mg 2ϩ -binding region forms other cross-links with residues Lys 537 , Lys 628 , and Lys 800/801 , which are not detected in E1a. The data suggest that the G729R E1a substitution could induce regional conformational changes of E1a, in agreement with the HDX-MS data presented above, thus likely affecting its interaction with the E2o component.
(ii) On interaction of E1a or G729R E1a with the E2o component, multiple inter-component cross-links were identified for the first-time: 63 cross-linked residue pairs were identified for the E1a⅐E2o subcomplex (Table S4) and 66 for the G729R E1a⅐E2o subcomplex (Table S5). The MS dataset related to Tables S1-S5 has been deposited to the Proteome Xchange Cosortium via PRIDE (PXD017792). In summary, the following cross-linked Lys residues were identified on both E1a and G729R E1a: Lys 148 and Lys 188 from the N-terminal region; Lys 300 and Lys 450 from the ThDP-and Mg 2ϩ -binding region; and Lys 818 , Lys 826/827 , and Lys 852/854 from the C-terminal region. On E2o, the cross-linked Lys 24 (Tables S4 and S5). Intriguingly, a great number of cross-links formed between the Lys residues from the C-terminal region of E1a and of G729R E1a and those from the lipoyl domain of E2o have been detected, suggesting that the C-terminal region of E1a/G729R E1a could be important for interaction with E2o and for substrate channeling. Almost half of the cross-linked Lys residues in E2o are located within its catalytic/core domain, where the glutaryl group is transferred from S-glutaryldihydrolipoyl-E2o to CoA with the formation of glutaryl-CoA in Equation 4, whereas Lys 98 , Lys 150 , and Lys 159 from the linker region could be considered as "hot spot" lysines (i.e. participate in multiple cross-links) for the E1a-E2o interaction.
Additionally, some unique inter-component cross-links were identified for the E1a⅐E2o and G729R E1a⅐E2o subcomplexes (Fig. 10A, where the meridian represents the E2o component with its three domains: lipoyl domain, linker region, and core domain), suggesting that local conformational changes induced by the G729R substitution could lead to different cross-links. In the G729R E1a⅐E2o subcomplex, the Lys 818 from the C-terminal E1a region forms a few unique cross-links with Lys 78 from the lipoyl domain of E2o, and with Lys 240 and Lys 342 , both from the E2o core. In the E1a⅐E2o subcomplex only two unique cross-links were identified: Lys 69 from the N-terminal region of E1a forms a cross-link with Lys 172 from the E2o linker region, and Lys 886 from the C-terminal region of E1a forms a cross-link with Lys 342 from the E2o core domain (see Fig. 10B where lysine residues involved in unique inter-component cross-links in E1a and in G729R E1a are mapped onto the structure of E1a). HDX-MS and chemical cross-linking data both provide evidence for local conformational changes in the C-terminal region of E1a that could affect E1a-E2o interaction and intermediate channeling. Although none of the three side-reactions of the E1a-ThDP-enamine intermediate are affected, the glutaryl group's channeling through the OADHc and formation of the glutaryl-CoA product are significantly impaired by the G729R E1a substitution (see Figs. S7-S10 where all MS/MS spectra for cross-links identified in E1a, G729R E1a, E1a⅐E2o subcomplex, and G729RE1a⅐E2o subcomplex are presented).

Conclusions
Data presented in this paper provide the first evidence that a heterozygous missense mutation (c.2185G3 A (p.Gly729Arg)) identified in the DHTKD1 gene encoding E1a and linked to pathogenesis of AMOXAD is associated with impairment of OADHc function. Sequence alignment of all known DHTKD1encoded proteins (E1a's) and human E1o revealed that the C-terminal region is highly conserved among the DHTKD1 species (and in E1o); however, the Gly 729 residue is unique for E1a species but is not present in human E1o or in human E1olike protein (Fig. S6).
Biochemical studies compared the behavior of G729R E1a to E1a itself, both as isolated components and when assembled to OADHc. Tracing the reactions along the path in Scheme 1, they conveniently fall into two categories, those pertinent to E1a by itself and those requiring the E2o (second) and even E3 (third) component of OADHc. The rates of neither the decarboxylation of OA nor of the three side-reactions emanating from the enamine intermediate were significantly affected by the G729R Table 2 Unique

intra-component (both intra-subunit and inter-subunit) cross links identified in E1a and in G729R E1a with CDI cross-linker
a m indicates oxidized Met; B indicates S-acetamido-Cys; also consider the common deamination of Asn to Asp and Gln to Glu (from N to D and from Q to E); cross-links were evaluated using software tool Merox 1.6.6 (46).

Consequences of (c.2185G3 A (protein G729R)) DHTKD1 mutation
substitution, perhaps not surprising in view of the distant location of the site of substitution from the ThDP-Mg 2ϩ active center. The three subsequent reactions, (i) reductive glutarylation of E2o by E1a and OA, (ii) glutaryl transfer from S8-glutaryldihydrolipoyl-E2o to CoA producing glutaryl-CoA (a reaction whose mechanism of transthiolesterification we had examined recently (38)), and (iii) NADH production as a measure of the rate of the overall reaction of the complex, were indeed measurably reduced, well beyond experimental error. The specific measurement especially of glutaryl-CoA formation rates clearly implied that interaction between the E1a and E2o was compromised by the G729R E1a substitution, leading to the reduced overall NADH activity and suggesting the mass spectrometric experiments to identify any structural impact of the substitution. HDX-MS, specifically the difference spectrum in Fig. 6B, provides strong evidence that the G729R E1a has several regions displaying different H/D exchange uptake and suggesting that there is measurable conformational change accompanying the G729R substitution.
Identification of peptic peptides reporting an EX1 kinetic behavior for HDX gave further evidence for the presence of at least two solution conformations of the E1a and showed how the single G729R E1a substitution could shift the conformational equilibrium to a more open one.
The HDX-MS experiments indicate that the interaction of E1a with E2o did not induce large-scale conformational changes in E1a; however, local backbone amide proton perturbations were indeed observed in the N-terminal region, the ThDP-and Mg 2ϩ -binding fold, and in the C-terminal region in Fig. 9. In contrast to the evidence for E1a-E2o interaction, the HDX-MS data provide no clear evidence for major changes in deuterium uptake upon binding of G729R E1a to E2o at any H/D exchange time points analyzed, suggesting only a transient interaction between the two components that could be too weak to be captured by the method used.
The HDX-MS experiments were followed by CL-MS experiments using a zero-length cross-linker to identify potential differences in the loci of E1a-E2o interactions resulting from the G729R E1a substitution. Among the unique differences in the cross-linking map in Fig. 10, it is notable that Lys 886 of E1a, but Lys 818 of G729R E1a, interacts with the core Lys 342 of E2o, once more confirming conformational changes with the substitution. Given that Lys 342 is proximal to the active center of the E2o core domain (38, 39) responsible for glutaryl-CoA formation, this change could diminish the rate of glutaryl-CoA for- Consequences of (c.2185G3 A (protein G729R)) DHTKD1 mutation mation detected by kinetics. Additional interaction of note is Lys 818 of G729R E1a cross-linked to Lys 78 of E2o of the lipoyl domain. This cross-link could also contribute to the lower rate observed.
The functional and structural studies presented herein provide insight into the molecular basis of the c.2185G3 A (p.G729R)) pathogenic mutation in DHTKD1 that impaired the function of the OADHc in the L-lysine metabolic pathway. On the basis of our MS findings, a simple explanation of the reduced activity of the G729R E1a variant is steric interference originating from the additional E1a-E2o interactions, partly blocking the lipoylation site and the active center of E2o. We also note, however, that the reduction in the rate of NADH produced by a factor of 40 -50 by the G729R E1a substitution, rather small in energetic terms, is probably the result of a number of contributions.
Having here identified the step in the complex mechanism that is responsible for the slower rate in NADH production, as well as the likely loci where the "misalignment" of the interaction of substituted E1a and E2o originates, the challenge here created is to the medicinal chemist to design agents to counteract the consequence of the pathogenic mutation.

Protein expression and purification
Expression and purification of human E1o, E1a, E2o, and E3 was as reported by us earlier (10). For construction of plasmid, expression, and purification of E2o-derived proteins, the pET-15b-E2o plasmid encoding the N-terminal His 6 -tag E2o was used as template, and the amplification primers and their complements were used for site-directed mutagenesis to introduce a TAA stop codon in place of Lys 88 and Glu 174 (the numbers include Met as the first amino acid). The following E2o-derived proteins were created: E2o 1-87 (LDo, comprising lipoyl domain and partially a linker region) and E2o 1-173 di-domain (comprising LDo and linker region (10)). The gene encoding C-terminally truncated His 6 -tagged E2o 144 -386 core domain was synthesized by ATUM (Newark, CA).

Construction of plasmid, expression, and purification of G729R E1a
The gene encoding the C-terminal His 6 -tagged E1a was synthesized by ATUM. The DHTKD1 gene was codon-optimized for expression in Escherichia coli cells and was inserted into pET-22b(ϩ) through the NdeI and XhoI restriction sites, and the resulting plasmid was expressed in BL21(DE3) cells. Cells were grown in LB medium supplemented with 50 g/ml ampicillin containing 1 mM MgCl 2 and 0.50 mM thiamin-HCl. Protein expression was induced by 0.5 mM isopropyl 1-thio-␤-Dgalactopyranoside for 15 h at 18°C (10). The G729R E1a variant was created by using QuikChange II XL site-directed mutagenesis kit. The following amplification primer containing the desired G279R mutation and its complement were used for the mutagenesis reaction: 5Ј-GCTGAGGAGGGTGTTGACCGT-GATACGGTTAACATGTTTG-3Ј and 5Ј-CAAACATGTTA-ACCGTATCACGGTCAACACCCTCCTCAGC-3Ј.
The DNA substitution was confirmed by sequencing at GENEWIZ, LLC (South Plainfield, NJ). The G729R E1a dis-played good expression and was purified to homogeneity by using Ni-Sepharose high-performance affinity media.

Activities of the E1a and of OADHc in different assay systems
(i) Overall and E1a-specific activities were measured as reported by us earlier (10,11,14).
(ii) Formation of glutaric semialdehyde by E1a and by G729R E1a was measured in an alcohol dehydrogenase-coupled reaction assay containing: 0.1 M Tris⅐HCl (pH 7.5), 150 mM NaCl, 0.5 mM ThDP, 2.0 mM MgCl 2 , 0.25 mM NADH, 1 mM DTT, 1% glycerol, 0.08 mg/ml horse liver alcohol dehydrogenase (ADH), and 1.0 mM OA. Reaction was initiated by E1a or by G729R E1a (0.10 mg/ml) after 1 min of equilibration at 37°C. The rate of NADH oxidation was measured at 340 nm from the initial slope of the recorded progress curve.
(iii) The rate of H 2 O 2 production by E1a and G729R E1a and by assembled OADHc was measured using a fluorescent Amplex UltraRed reagent as was reported by us recently for E1o and for OGDHc (11).
(iv) CD spectroscopy was used for the detection of the (S)-2-hydroxy-3-oxoheptanedioic acid, a carboligation product formed in the reaction of the E1a-ThDP-enamine intermediate with glyoxylate as an acceptor as reported by us earlier (11,14).
(v) Reductive glutarylation of the E2o lipoyl domain by E1a and by G729R E1a was by FT-MS as reported by us earlier (11,17,38).

Enzymatic synthesis of glutaryl-CoA
For the enzymatic synthesis of glutaryl-CoA, the E1a (0.15 mg) or G729R E1a (0.15 mg) was assembled with E2o (0.30 mg) and E3 (0.75 mg) into OADHc in 0.15 ml of 0.1 M Tris⅐HCl (pH 7.5) containing 0.3 M NH 4 Cl, 0.5 mM ThDP, and 2.0 mM MgCl 2 at 25°C. The mass ratio of the E1a/E2o/E3 components of 1:2:5 (mg/mg/mg) was selected from an optimization of the overall NADH assay. After 40 min of incubation, an aliquot containing 0.005 mg of E1a or 0.020 mg of G729R E1a was withdrawn and was placed into 0.4 ml of the reaction assay containing all components necessary for the overall NADH assay. The reaction was initiated by addition of OA (2.5 mM) and CoA (300 M) after 1 min of equilibration at 37°C in a Thermomixer (Eppendorf). The reaction was terminated at different times by acidification on addition of 2.5% TFA to a final concentration of 0.1%, and samples were kept at 4°C. The glutaryl-CoA formed in the reaction was purified according to the protocol reported (21). The samples were analyzed with a 355-nm laser in a negative mode on the ultrafleXtreme MALDI spectrometer (Bruker Daltonics, Bremen, Germany). For each spectrum, five subspectra with 400 laser shots with 400 Hz frequency were acquired. The spectra were analyzed using flexAnalysis 3.0 software (Bruker Daltonics, Bremen, Germany) and were smoothed using the Savitzky-Golay algorithm after baseline subtraction.

Electron paramagnetic resonance spectroscopy
The sample preparation and EPR experiment for detection of ThDP-E1a-and G729R E1a-bound radical species was as reported by us earlier (11, 14, 15).

Consequences of (c.2185G3 A (protein G729R)) DHTKD1 mutation Fluorescence spectroscopy
All fluorescence spectra were recorded at 25°C by using a Varian Cary Eclipse fluorescence spectrophotometer. Titration experiments were conducted in 100 mM HEPES (pH 7.5), containing 0.15 M NaCl and 1% glycerol. The E2o 1-173 di-domain and E2o 144 -386 core domain (50 M), each containing a single cysteine residue at positions 38 and 179, respectively, were labeled with N-(1-pyrene) maleimide according to a protocol reported by us earlier (16). The titration experiment was conducted as reported previously (16).

Immunoblotting
Control (GM23964, Coriell Institute for Medical Research) and DHTKD1 R715C/G729R patient-derived fibroblasts (a gift of Dr. Jose E. Abdenur, University of California, Irvine) were cultured in Dulbecco's modified Eagle's cell culture medium supplemented with 10% fetal bovine serum. The samples for the immunoblot analysis of AMOXAD patient's fibroblasts in Fig.  1A were exactly as reported in (3). Cells were lysed in RIPA lysis and extraction buffer supplemented with protease inhibitor mixture and centrifuged 10 min at 12,000 ϫ g at 4°C, and total protein was determined by the BCA method. Proteins were separated on a Bolt TM 4 -12% Bis-TrisPlus Gel, blotted onto a nitrocellulose membrane, and were detected by using the following primary antibodies against DHTKD1-encoded protein (GTX32561, Lot no. 821800318, GeneTex), citrate synthase (GT1761) (GTX628143, Lot no. 41113, GeneTex), and glyceraldehyde-3-phosphate dehydrogenase (H00002597-D01P, Lot no. EB071, RRID:AB_1503253, Abnova). Proteins were detected and visualized using IRDye 800CW or IRDye 680RD secondary antibodies (LI-COR Biosciences, Lincoln, NE) and Image Studio Lite software (version 5.2, LI-COR Biosciences). Equal loading was checked by Ponceau S Staining Solution and citrate synthase and glyceraldehyde-3-phosphate dehydrogenase signals.

Thermal shift assay
Thermal shift assay for E1a and G729R E1a purified by Houten and co-workers (54) was performed using QuantStudio 3 real-time PCR system (Applied Biosystems). Purified proteins were diluted to a final concentration of 0.5 mg/ml in 100 mM KH 2 PO 4 (pH 7.2) containing 1 mM MgCl 2 , 0.21 mM ThDP, and 20% glycerol. A 17.5-l aliquot of protein was mixed with 2.5 l of 8ϫ SYPRO Orange in a final volume of 20 l. The 8ϫ SYPRO Orange was prepared by diluting the dye by 250-fold in H 2 O. The reactions were kept at 25°C for 2 min, and then the temperature was increased from 25 to 95°C at a rate of 0.05°C/s. The change in the fluorescence intensities of SYPRO Orange was monitored as a function of temperature and was analyzed by Protein Thermal Shift Software 1.3. Each reaction was performed in five replicates.

Mass spectrometry
(i) HDX-MS analysis was conducted as described by us earlier using a 7T Bruker Daltonics FT-MS instrument (16,17,(40)(41)(42). Peptides were identified from nondeuterated samples by a customized program DXgest (43), which matches experi-mental peptide mass with theoretically-generated peptic peptide mass by using statistical data for the pepsin cleavage pattern under H/D exchange. Mass tolerance was set at 2.0 ppm. H/D exchange data for each individual peptide at various time points were processed using QUDeX-MS (44) and HX-Express (45). No back-exchange correction was needed for the purpose of comparative analysis. All time points were run in duplicates. The deuterium uptake level and deuterium incorporation were plotted with the mean value of two independent experiments.
(ii) For CL-MS analysis of the E1a⅐E2o and G729R E1a⅐E2o binary subcomplexes, the CDI was used. First, the E1a (1 nmol, 33 M subunits) in 20 mM K 2 HPO 4 (pH 7.5) containing 0.5 mM ThDP, 1 mM MgCl 2 , 0.15 M NaCl, and 10% glycerol was mixed in 15 l with E2o (1 nmol, 33 M subunits) in the same buffer system. After 30 min of incubation at 20°C, 1 l of CDI (1 M) dissolved in DMSO was added, and the cross-linking reaction was conducted at 15°C for 45 min. To quench the reaction, the reaction mixture was diluted to 50 l with reaction buffer above, and 1 M Tris⅐HCl (pH 8.0) was added to a final concentration of 20 mM. Cross-linked E1a-E2o was identified by SDS-PAGE (7.5%).
Next, cross-linked samples were subjected to tryptic and Glu-C in-solution double digestion. An aliquot containing 1 nmol of the total protein was withdrawn from each reaction mixture and was placed into 70 l of 8 M urea in 100 mM NH 4 HCO 3 , and the samples were incubated at 60°C. After 20 min of incubation, 2 l of 200 mM DTT was added, and the samples were incubated for an additional 40 min. Next, 3.5 l of 200 mM iodoacetamide was added, and the samples were incubated at room temperature in the dark for 30 min. Then, 1 l of 200 mM DTT was added, with a subsequent dilution of the reaction mixture with 0.85 ml of 100 mM NH 4 HCO 3 . The Glu-C digestion was carried out at 50:1 protein to Glu-C (w/w) ratio at 37°C. After 4 h of incubation, the reaction mixtures were subjected to tryptic digestion at 100:1 protein to trypsin (w/w) ratio at 37°C. After overnight digestion, the reaction was terminated by addition of 2 l of 95% formic acid. Digested samples were desalted on a SepPak SPE C-18 column, dried using a SpeedVac centrifuge (Savant), and then dissolved in 100 l of 20% acetonitrile (0.05% formic acid).
Peptide enrichment was carried out by using an SCX trap cartridge. The enriched peptides were desalted by a SepPak SPE C-18 column and were dried using a SpeedVac centrifuge. Cross-linked peptides were analyzed by nano-LC-MS/MS (Dionex Ultimate 3000 RLSC nanosystem interfaced with Q Exactive HF (Thermo Fisher Scientific)) as reported by us earlier (16) and were evaluated by using software tool Merox 1.6.6 (46). Samples were loaded onto a self-packed 100-m by 2-cm trap (Magic C18AQ; 5 m and 200 Å; Michrom Bioresources, Inc.) and washed with buffer A (0.1% TFA) for 5 min at a flow rate of 10 l/min. The trap was brought in-line with the analytical column (self-packed Magic C18AQ; 3 m and 200 Å; 75 m by 50 cm), and peptides were eluted at 300 nl/min using a segmented linear gradient of 4 -15% solution A in 30 min (solution A, 0.2% formic acid; solution B, 0.16% formic acid, 80% acetonitrile), 15-25% solution B in 40 min, 25-50% solution B in 44 min, and 50 -90% solution B in 11 min. Mass spectrometric data were acquired using a data-dependent acquisition pro-Consequences of (c.2185G3 A (protein G729R)) DHTKD1 mutation cedure with a cyclic series of a full scan with a resolution of 120,000, followed by MS/MS (higher-energy C-trap dissociation; relative collision energy, 27%) of the 20 most intense ions and a dynamic exclusion duration of 20 s. LC-MS/MS peak lists were generated using the ProteoWizard 3.0.18156 (48) software package and searched against the SwissProt database (proteome ID: UP000000558, search was done by using searchGUI-3.3.117 incorporated with Peptideshaker 1.16.42 (49,50)). Search parameters were as follows: fragment mass error, 20 ppm; parent mass error, 5 ppm; fixed modification, carbamidomethylation on cysteine; potential modifications during initial search, methionine oxidation and acetylation on protein N termini and up to one missed tryptic cleavage. This search showed that no contaminant protein above 5% was present in the sample. Cross-links were evaluated using the software tool Merox 1.6.6 (46) with the following parameters: precursor 5 ppm; fragment 15 ppm mass accuracy; CDI was set as crosslinker. Only Lys-Lys cross-linked peptides were analyzed. The cross-linked peptides were manually validated, and the results with false discovery rate (FDR) Ͻ0.01 limit were analyzed (FDR was calculated by Merox software). The results were visualized online using the XVIS website (51). The MS proteomics data have been deposited to the ProteomeXchange Consortium (52) via the PRIDE (53) partner repository with the dataset identifier PXD017792. In Figs. S7-S10 are presented all MS/MS spectra for cross-links identified in E1a, G729R E1a, E1a⅐E2o subcomplex, and G729R E1a⅐E2o subcomplex.

Data availability
The MS proteomics data have been deposited to the Pro-teomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD017792. All remaining data are contained within the article and supporting information.