Structural Analysis of Pathogenic Mutations Targeting Glu427 of ALDH7A1, the Hot Spot Residue of Pyridoxine- Dependent Epilepsy

Received: 10 September 2019 Revised: 17 October 2019 Accepted: 23 October 2019 DOI: 10.1002/jimd.12184

ORIGINAL ARTICLE

Structural analysis of pathogenic mutations targeting Glu427 of ALDH7A1, the hot spot residue of pyridoxine- dependent epilepsy

Adrian R. Laciak1 | David A. Korasick2 | Kent S. Gates1,2 | John J. Tanner1,2

1Department of Chemistry, University of Missouri, Columbia, Missouri Abstract 2Department of Biochemistry, University Certain loss-of-function mutations in the gene encoding the lysine catabolic of Missouri, Columbia, Missouri enzyme aldehyde dehydrogenase 7A1 (ALDH7A1) cause pyridoxine- dependent epilepsy (PDE). Missense mutations of Glu427, especially Correspondence John J. Tanner, Department of Glu427Gln, account for ~30% of the mutated alleles in PDE patients, and thus Biochemistry, University of Missouri, Glu427 has been referred to as a mutation hot spot of PDE. Glu427 is invariant Columbia, MO 65211. in the ALDH superfamily and forms ionic hydrogen bonds with the nicotin- Email: [email protected] amide ribose of the NAD+ cofactor. Here we report the first crystal structures Communicating Editor: Brian Fowler of ALDH7A1 containing pathogenic mutations targeting Glu427. The mutant

Funding information enzymes E427Q, Glu427Asp, and Glu427Gly were expressed in Escherichia coli National Institute of General Medical and purified. The recombinant enzymes displayed negligible catalytic activity Sciences, Grant/Award Number: compared to the wild-type enzyme. The crystal structures of the mutant R01GM093123; High-End + Instrumentation Grant, Grant/Award enzymes complexed with NAD were determined to understand how the Number: S10OD018483; National mutations impact NAD+ binding. In the E427Q and E427G structures, the nic- Institutes of Health, Grant/Award otinamide mononucleotide is highly flexible and lacks a defined binding pose. Number: P30 GM124169 In E427D, the bound NAD+ adopts a “retracted” conformation in which the nicotinamide ring is too far from the catalytic Cys residue for hydride transfer. Thus, the structures revealed a shared mechanism for loss of function: none of the variants are able to stabilise the nicotinamide of NAD+ in the pose required for catalysis. We also show that these mutations reduce the amount of active tetrameric ALDH7A1 at the concentration of NAD+ tested. Alto- gether, our results provide the three-dimensional molecular structural basis of the most common pathogenic variants of PDE and implicate strong (ionic) hydrogen bonds in the aetiology of a human disease.

KEYWORDS ALDH7A1, enzyme kinetics, missense mutation, PDE, pyridoxine-dependent epilepsy, X-ray crystallography

Abbreviations: AA, α-aminoadipate; AASAL, α-aminoadipate semialdehyde; ALDH, aldehyde dehydrogenase; ALDH7A1, aldehyde dehydrogenase 7A1; NMN, nicotinamide mononucleotide; P6C, Δ1-piperideine-6-carboxylic acid; PA, L-pipecolic acid; PDE, pyridoxine-dependent epilepsy; PLP, pyridoxal 50-phosphate.

1 | INTRODUCTION variants E427Q, E427D, and E427G. None of the variants exhibit measurable catalytic activity, consistent with a Pyridoxine-dependent epilepsy (PDE) is an epileptic role in PDE pathogenesis. The crystal structures show encephalopathy characterised by seizures, usually occur- that the catalytic defect results from NAD+ adopting ring in the first hours of life, which are unresponsive to non-native conformations, which are incompatible with standard antiepileptic drugs, but can be controlled by catalysis. Also, the mutations reduce the amount of active pharmacologic doses of pyridoxine.1 The most common tetrameric ALDH7A1 under the conditions tested. These cause of PDE is autosomal recessive inheritance of cer- results provide a molecular-structural explanation of the tain mutations in the gene encoding the enzyme most common pathogenic variants of PDE. α-aminoadipate semialdehyde dehydrogenase, also known as ALDH7A1 or antiquitin. ALDH7A1 is part of lysine catabolism and catalyses the NAD+-dependent oxi- 2 | RESULTS dation of α-aminoadipate semialdehyde (AASAL) to α-aminoadipate (AA). The mutational spectrum of PDE 2.1 | The mutations E427Q, E427D, and spans over 113 different mutations within the 18 exons of E427G compromise catalytic activity the ALDH7A1 gene, including approximately 60 missense mutations.1-4 The estimated carrier frequency of The purified recombinant mutant variants E427Q, ALDH7A1 mutations is 1:127, and the estimated inci- E427D, and E427G were tested for catalytic activity using dence of ALDH7A1 deficiency is 1:64 352 births.1 The steady-state assays that monitor the production of biochemical basis of PDE is a decrease of the ubiquitous NADH. As a control, the activity of the wild-type enzyme enzyme cofactor pyridoxal 50-phosphate (PLP). The cellu- was assayed as a function of AASAL concentration with lar pool of PLP is depleted via the irreversible reaction of NAD+ fixed at a saturating concentration of 2.5 mM. PLP with Δ1-piperideine-6-carboxylic acid (P6C), the The enzyme displays Michaelis-Menten behaviour cyclised form of AASAL.5 Treatment with pyridoxine (Figure 1B-D), characterised by kinetic constants of −1 addresses the PLP deficiency and typically provides ade- kcat = 0.38 ± 0.02 seconds and Km for AASAL of 517 quate seizure control, yet 75% of individuals with PDE ±86μM. The resulting catalytic efficiency (kcat/Km)of − − have intellectual disability and developmental delay.2,3 735 seconds 1 M 1 is within a factor of 2 to 6 of the The missense mutation of Glu427 to glutamine values reported previously.10-12 In contrast, none of the (Glu427Gln) is the most commonly reported pathogenic mutant variants displayed detectable activity across the variant in the PDE literature,1 and this mutation is esti- entire range of AASAL concentration (Figure 1B-D). The mated to be present in 30% of PDE patients with results for E427Q and E427G agree with previous studies European ancestry.5-7 In addition to the Gln mutation at reporting an absence of catalytic activity in crude extracts residue 427, mutations to Asp (Glu427Asp) and Gly of cells expressing these variants.5,13 To our knowledge, (Glu427Gly) were reported in a cohort of 18 North Amer- ours is the first report of E427D being inactive. ican patients with PDE.8 Because of its high frequency of mutation in PDE patients, Glu427 has been referred to as a mutational hotspot of the disease.8 2.2 | The nicotinamide mononucleotide Glu427 plays a key role in binding the cofactor of NAD+ bound to E427Q and E427G is NAD+. Crystal structures of ALDHs, including highly flexible and lacks a defined pose ALDH7A1, show that Glu427 hydrogen bonds to the hydroxyl groups of the NAD+ nicotinamide ribose The crystal structure of E427Q complexed with NAD+ (Figure 1A). Presumably, these interactions help stabilise was determined in space group C2 at 2.06 Å resolution the extended conformation of NAD+, and ensure that the from a crystal grown in the presence of 15 mM NAD+. nicotinamide ring is close to the catalytic cysteine The asymmetric unit contains eight protein chains (Cys330) and positioned optimally for the hydride trans- arranged in two tetramers. The tetramer observed in the fer step of the catalytic mechanism. The importance of crystal has been shown to be the active form of these hydrogen bonds is highlighted by the fact that ALDH7A1.14 We note this is the same crystal form that Glu427 is one of only a few invariant residues conserved has been used to solve structures of wild-type throughout the entire ALDH superfamily.9 ALDH7A115,16 and several PDE mutant variants.11 To better understand the molecular basis of the most The electron density was sufficient to model the loop common pathogenic variants of PDE, we have deter- containing Gln427 (“427-loop”) in all eight chains of the mined the catalytic properties, crystal structures, and asymmetric unit (Figure 2A). The conformation of the oligomeric states of the ALDH7A1 pathogenic mutant 427-loop of E427Q is very similar to that of the wild-type LACIAK ET AL. 3 enzyme (Figure 2B). Therefore, the defect in catalysis is highly flexible (Figure 2A). This was true for all eight not due to a major structural perturbation of the protein chains in the asymmetric unit. We note that the 427-loop. NMN is also disordered in wild-type ALDH7A1 struc- Strong electron density was observed for only the tures when NAD+ is included in the crystallisation at ADP portion of NAD+, suggesting the nicotinamide 5 mM (eg, see PDB code 4ZUK).15 However, the cofactor mononucleotide (NMN) group of the bound cofactor is is fully resolved by strong electron density when included in the crystallisation in the higher concentration range of 11 to 15 mM (PDB codes 2J6L, 6O4B-C, 6O4E, 6O4H).11,16 Because 15 mM NAD+ was used for crystallisation of E427Q, the lack of electron density for the NMN is meaningful. The apparent partial binding of the cofactor under these conditions is consistent with the

possibility of the mutation significantly increasing the Km for NAD+ compared to wild-type. The resolved ADP fragment of NAD+ bound to E427Q exhibits an atypical conformation. Although the adenosine of NAD+ binds the canonical location, the pyrophosphate group has rotated into an unusual “out” conformation in six of the eight chains (chains A-F) in the asymmetric unit (Figure 2A,B). In chains A and E, density for only the “out” conformation is observed, while strong density for both the “out” and a more typi- cal (“in”) conformation is present in chains B-D,F (eg, Figure 2A). In the “out” conformation, the NMN appears to be sampling the solvent region outside of the active site (Figure 2B). We note the “out” conformation has not been observed in any other ALDH7A1 structure. The structure of E427G-NAD+ (2.15 Å resolution) tells a similar story. As in E427Q-NAD+, the NMN of NAD+ bound to E427G is disordered, and there is electron density evidence suggesting an “out” conformation in some of the chains (Figure 2C). We note that electron density for the 427-loop is weaker than in the other mutant variants, particularly at residue 427 (Figure 2C). This is likely due to the increased flexibility introduced by the mutation of Glu to Gly. Nevertheless, the density was sufficient for modelling the backbone of the loop in all eight chains (Figure 2C). The model shows that residue 427 has shifted by 2 to 3 Å away from binding site

FIGURE 1 Structural context of NAD+ binding and steady- state kinetic data for Glu427 variants. A, Conformation and interactions of NAD+ bound to ALDH7A1 in the active conformation (PDB 6O4C). For reference, NAD+ and protein are coloured sand and grey, respectively. Steady-state kinetic data for (B) E427Q, (C) E427D, and (D) E427G. In each panel, the black symbols are data for the wild-type enzyme, and the black curve is the fit to the Michaelis-Menten equation. The circles, squares, and triangles represent different replicate experiments performed at the same substrate concentration. Three trials were performed for each enzyme. The kinetic constants for the wild-type enzyme are −1 kcat = 0.38 ± 0.02 seconds and Km = 517 ± 86 μM 4 LACIAK ET AL.

FIGURE 2 The impact of E427Q/D/G mutations on the structure of ALDH7A1. A, Electron density for NAD+ and the 427-loop of E427Q. The cage represents polder omit maps (3σ). B, Superposition of E427Q-NAD+ (sand) with the wild-type enzyme (white, PDB code 2J6L). The dashed lines represent hydrogen bonds in the wild-type enzyme between Glu427 and NAD+. C, Electron density for NAD+ and the 427-loop of E427G. The cage represents polder omit maps (3σ). D, Superposition of E427G-NAD+ (sand) with the wild-type enzyme (white, PDB code 2J6L). The arrow denotes the 3-Å shift of residue 427 caused by the mutation. E, Electron density for NAD+ and the 427-loop of E427D. The cage represents polder omit maps (3σ). The asterisk denotes the C4 atom of the nicotinamide, which is the hydride acceptor of the ALDH reaction. F, Superposition of E427D-NAD+ (sand) with the wild-type enzyme (white, PDB code 2J6L). The curved arrow denotes the 7-Å shift in the nicotinamide caused by the mutation for the nicotinamide ribose (Figure 2D). Thus, unlike several interactions not found in the wild-type enzyme- E427Q, the mutation to Gly apparently perturbs the local cofactor complex: the carboxamide of the nicotinamide protein conformation. hydrogen bonds with Asp427 and Gly299, the nicotinamide ribose hydrogen bonds with Asp427, and the nicotinamide ring stacks against Phe429. 2.3 | NAD+ binds to E427D in a retracted, The retracted conformation is incompatible with inactive pose catalysis. The hydride transfer step of the ALDH mechanism requires proximity of the nicotinamide to cata- The crystal structure of E427D complexed with NAD+ lytic Cys330. In the wild-type enzyme, the hydride was determined in space group P21 at 2.06 Å resolution. acceptor atom of the nicotinamide (C4) is 3.0 Å from We note this crystal form is new for ALDH7A1. The elec- the S atom of Cys330. The corresponding distance in tron density is strong for the entire NAD+ cofactor in all E427D is 4.4 Å, which is unsuitable for hydride transfer four chains in the asymmetric unit. Likewise, the map (Figure 2F). guided the modelling of the 427-loop in all four chains. A structure of E427D complexed with the product AA The density in chain A provides the clearest picture of was also determined (1.75 Å resolution) to understand the 427-loop conformation, so we focus our analysis on how this mutation affects recognition of the aldehyde this chain (Figure 2E). substrate. The pose and interactions of AA bound to The NAD+ adopts an inactive conformation. While E427D are identical to those of the wild-type enzyme the adenosine is anchored in the canonical binding site, (Figure S1). This result supports the hypothesis that the NMN has withdrawn from active site by 4 to 7 Å retraction of NAD+ from the catalytic Cys is the main (Figure 2F). The retracted conformation is stabilised by cause of the catalytic defect of E427D. LACIAK TABLE 1 Small-angle X-ray scattering analysis

ALDH7A1 E427Q E427D E427G AL ET

1.2 mg/ml 2.3 mg/ml 4.7 mg/ml 1.1 mg/ml 2.1 mg/ml 4.3 mg/ml 1.4 mg/ml 2.9 mg/ml 5.7 mg/ml 1.6 mg/ml 3.2 mg/ml 6.5 mg/ml . Guinier analysis

Rg (Å) 38.3 ± 0.2 38.3 ± 0.2 38.3 ± 0.2 37.4 ± 0.2 37.8 ± 0.2 37.6 ± 0.2 36.9 ± 0.2 38.2 ± 0.2 38.5 ± 0.2 37.8 ± 0.2 37.4 ± 0.2 37.6 ± 0.2 −1 qmin (Å ) 0.01087 0.01087 0.01087 0.01087 0.01087 0.01087 0.01087 0.01087 0.01087 0.01087 0.01087 0.01087

qRg range 0.42-1.29 0.42-1.29 0.42-1.29 0.41-1.30 0.41-1.29 0.41-1.29 0.40-1.30 0.41-1.29 0.42-1.30 0.41-1.29 0.41-1.28 0.41-1.29 R- 0.937 0.991 0.988 0.969 0.981 0.994 0.957 0.996 0.998 0.98 0.988 0.998 Squared I(0) 16.06 ± 0.07 31.1 ± 0.1 65.2 ± 0.2 15.29 ± 0.06 31.5 ± 0.1 61.0 ± 0.2 22.26 ± 0.08 47.7 ± 0.2 95.5 ± 0.3 22.83 ± 0.09 50.2 ± 0.2 103.8 ± 0.3 P(r) analysis

Rg (Å) 37.5 ± 0.2 37.4 ± 0.1 37.5 ± 0.8 36.6 ± 0.1 36.6 ± 0.1 36.6 ± 0.1 37.8 ± 0.1 37.9 ± 0.8 37.7 ± 0.8 37.2 ± 0.2 37.2 ± 0.2 37.6 ± 0.4

Dmax (Å) 115 107 107 102 108 106 113 114 106 113 105 109 q-Range 0.0125-0.2082 0.0120-0.2088 0.0137-0.2082 0.0131-0.2160 0.0142-0.2121 0.0142-0.2121 0.0125-0.2099 0.0114-0.2093 0.0125-0.2076 0.0125-0.2138 0.0131-0.2132 0.0120-0.2127 − (Å 1 ) Total 0.92 0.95 0.93 0.94 0.87 0.92 0.96 0.94 0.82 0.91 0.79 0.76 quality estimate Porod 350 000 326 000 315 000 237 000 237 000 255 000 256 000 259 000 272 000 291 000 251 000 271 000 volume (Å3)

SAXS Mr (kDa) MoWa 211 000 221 000 230 000 175 000 180 000 192 000 158 000 185 000 206 000 196 000 186 000 205 000 b Vc 191 300 190 000 193 000 154 000 159 000 168 000 155 000 170 000 180 000 170 000 167 000 178 000 Model fitting Dimer χ2 17 50 110 16 26 62 21 52 120 26 49 140 Tetramer χ2 0.52 0.39 0.65 1.0 1.1 1.9 1.4 1.1 2.3 1.0 1.2 2 Dimer: N/Dc N/Dc N/Dc 40:60 (0.83) 29:71 (0.72) 26:74 (1.1) 18:82 (1.4) 14:86 (1.0) 9:91 (2.3) 33:67 (0.74) 22:78 (0.82) 13:87 (1.6) tetramer (χ2) SASBDB code SASDGH4 SASDGJ4 SASDGK4 SASDGL4 SASDGM4 SASDGN4 SASDGP4 SASDGQ4 SASDGR4 SASDGS4 SASDGT4 SASDGU4 aCalculated using the SAXS MoW method18 as implemented in PRIMUS.19 Calculatedb using the volume of correlation method20 as implemented in PRIMUS.19 Using a dimer-tetramerc ensemble did not improve the fit compared to the tetramer-only model. 5 6 LACIAK ET AL.

2.4 | Mutation of Glu427 reduces the 3 | DISCUSSION amount of tetrameric ALDH7A1 The three ALDH7A1 Glu427 variants studied display We previously showed that the active form of ALDH7A1 negligible catalytic activity, consistent with clinical stud- is a tetramer, and the binding of NAD+ promotes tetra- ies implicating them in the pathology of PDE.1,5-8 We mer assembly.14 Because the mutant variants fail to bind note that PDE patients who are biallelic for E427Q gener- NAD+ in the active conformation, we tested the ability of ally experience relatively early disease onset, which is these enzymes to form the tetramer. Small-angle X-ray consistent with the lack of activity of this mutant.7 scattering (SAXS) was used to investigate the oligomeric The impact of mutating Glu427 has also been studied structure.17 in ALDH2. Mutation of Glu427 of ALDH2 to Gln or Asp As previously observed, the addition of NAD+ decreases catalytic efficiency by factors of 40 and (10 mM) to wild-type ALDH7A1 results in complete tet- 6, respectively.21,22 We found that these mutations have a ramer formation as indicated by the excellent agreement much more profound effect on the catalytic activity of (χ2 < 1) between the experimental SAXS curve and the ALDH7A1. A possible explanation for this difference is theoretical SAXS curve calculated from the crystallo- that the NAD+ bound to ALDH2 is stabilised by addi- graphic tetramer model (Table 1, Figure S2). This result tional interactions not found in ALDH7A1. Trp168 of indicates that the binding of NAD+ shifts the dimer- ALDH2 (Phe194 in ALDH7A1) forms two hydrogen tetramer equilibrium of the wild-type enzyme over- bonds with the pyrophosphate of NAD+ (Figure S3).23 whelmingly to the tetramer, in agreement with our These extra interactions may partially compensate for the previous study.14 loss of Glu427 in the ALDH2 variants. All three ALDH7A1 Glu427 mutant variants were Because Glu427 is invariant in the ALDH superfamily tested under the same experimental conditions (10 mM and participates directly in binding NAD+, it is perhaps not NAD+) and similar protein concentration ranges and surprising that mutations of Glu427 of ALDH7A1 are patho- compared to this wild-type result. Interestingly, all three genic. However, from a physicochemical perspective, one mutant variants are compromised in tetramer forma- might consider the mutation E427Q to be conservative. After tion. In each case, an ensemble model consisting of both all, glutamine is isostructural with glutamate, and glutamine the dimer and the tetramer structures provided better has the potential to hydrogen bond to the nicotinamide fits to the experimental data than the tetramer alone. ribose hydroxyl groups. The fact that glutamine is never Ensemble analysis with MultiFoXS shows that the observed at residue 427 suggests that the negative charge of E427Q, E427D, and E427G samples contain 26% to 40%, glutamate is essential for optimal catalytic function. 9% to 18%, and 13% to 33% dimer, respectively (Table 1, The essentiality of Glu427 is likely due to its ability to Figure S2). form strong hydrogen bonds. In his seminal book on Two other metrics were also utilised to reveal discrep- hydrogen bonding, George Jeffrey described the three cate- ancies in the oligomeric state of wild-type vs mutant gories of hydrogen bonds: strong, moderate, and weak.24 ALDH7A1: experimentally determined molecular mass Those between Glu427 and the hydroxyls of NMN are

(Mr) from SAXS and Porod volume (VP). Wild-type sam- “strong” by virtue of the negative charge of glutamate. ples showed a SAXS Mr of 211 to 230 kDa (Table 1), This type of hydrogen bond is also known as “ionic hydro- which is within 5% of the theoretical Mr of 222 kDa. In gen bonds”. In contrast, glutamine can only form moder- contrast, the E427Q samples showed a SAXS Mr of 175 to ate hydrogen bonds with a hydroxyl group. Our results 192 kDa (Table 1). Note that at the highest concentration show that, apparently, moderate hydrogen bonds are not tested, the Mr of E427Q is 13% lower than that of the tet- sufficient in energy to keep the NMN group in the proper + ramer. Similarly, the SAXS Mr values of E427D and position for catalysis, and thus the recognition of NAD E427G are consistently lower (7%-29%) than expected for by ALDH7A1 depends critically on the ability of Glu427 to a predominantly tetrameric solution (Table 1). Further, form strong hydrogen bonds with the cofactor. We may the range of VP reveals an overall decrease in average VP conclude that the aetiology of this human disease has of 26%, 21%, and 18% for E427Q, E427D, and E427G com- roots in the fundamental principles of physical chemistry. pared to wild-type (Table 1). Because of the relationship Strong hydrogen bonds between residue 427 and the 17 + of VP to approximate Mr, these data are also consistent NMN may help promote the unfurling of NAD during + with decreased Mr in the mutants compared to wild-type. binding. The predominant form of NAD in solution is a Overall, these data are consistent with a decreased aver- compact folded conformation in which the nicotinamide age in solution Mr for the Glu427 variants as compared to and adenine rings stack together with an inter-ring dis- wild-type, suggesting the mutations perturbed the dimer- tance of 5.2 Å.25 In contrast, NAD+ bound to ALDH7A1 tetramer equilibrium. (and other Rossmann fold enzymes26) is highly extended, LACIAK ET AL. 7 with an inter-ring distance of ~13 Å. Therefore, the Table S1. The mutations were verified by Sanger sequenc- enzyme must supply sufficient interactions to stabilise ing. Wild-type ALDH7A1 and the mutant variants were the extended conformation over the folded form. We expressed in E. coli and purified as described observed that NAD+ adopts well-defined, extended con- previously.11,27 formations when bound to either the wild-type enzyme or E427D, and both enzymes form strong hydrogen bonds with the NMN. In contrast, NAD+ exhibits disor- 5.2 | Kinetic assays der in E427Q and E427G, which are incapable of forming strong hydrogen bonds. The “out” conformation of ALDH7A1 enzymatic activity was measured by monitor- NAD+ observed in these variants perhaps suggests that ing NADH production at a wavelength of 340 nm using the bound cofactor is sampling folded conformations. an Epoch 2 Microplate Reader. The reaction assay buffer Thus, strong hydrogen bonds appear to be critical for pro- contained 50 mM pyrophosphate buffer (pH 8.0). moting the unfolding of NAD+ during the process of Enzyme stock solutions were prepared by diluting to the binding to ALDH7A1. desired concentration with 50 mM pyrophosphate buffer at pH 8.0 supplemented with 2.5 mM NAD+. AASAL was used as the variable substrate (16-2000 μM) with NAD+ 4 | CONCLUSION fixed at 2.5 mM. AASAL was synthesised and quantitated as previously reported.10 The assays contained a final Our results provide the three-dimensional molecular struc- enzyme concentration of 0.07 μM wild-type and 5 μM tural basis of the most common pathogenic variants of PDE. E427Q/D/G. Triplicate data sets were collected, and The lack of catalytic activity of E427Q, E427D, and E427G kinetic constants were obtained by fitting the initial rate suggests that ALDH7A1 is very sensitive to mutations at data to the Michaelis-Menten equation globally using Glu427. This is consistent with the universal conservation of Origin 2019. this residue in the ALDH superfamily. In E427Q and E427G, the bound cofactor is highly flexible and lacks a defined conformation for the NMN group. In E427D, the 5.3 | Protein crystallisation cofactor adopts a novel inactive pose in which the nicotinamide ring is retracted from the active site. Thus, the struc- Crystallisation trials were performed at 20C using sitting tures revealed a shared mechanism for loss of function: drop vapour diffusion. Wild-type ALDH7A1 crystals were none of the variants are able to stabilise the nicotinamide prepared as described previously11 and crushed to make group of NAD+ in the pose required for catalysis. a microseed stock to aid crystallisation of the mutant Here we uncovered a third distinct molecular mechanism variants. by which genetic mutations in the ALDH7A1 gene impair Crystals of E427Q and E427D complexed with NAD+ ALDH7A1 enzymatic function. These results build upon pre- were grown by co-crystallisation using 5 mg/ml enzyme + vious studies of mutations targeting the oligomer interfaces of and 15 mM NAD . The reservoir contained 0.2 M MgCl2, ALDH7A127 and a set of mutations affecting residues in the 21% (w/v) PEG 3350, and 0.1 M Bis-Tris (at pH 6.2 for AASAL binding site.11 The interface mutations abrogated cat- E427Q and pH 5.7 for E427D). Crystallisation trials (with alytic function by impairing formation of the active tetramer. microseeding) were set up using an Oryx8 robot (Douglas In contrast, mutations in the AASAL site resulted in subtle Instruments). The crystals were prepared for low temper- perturbations of the structure and dynamics of the aldehyde ature data collection by soaking in a cryobuffer consisting substrate site, without affecting cofactor binding. The Glu427 of the reservoir supplemented with 18% (v/v) ethylene variants revealed a third mechanism involving a defect in glycol and 5 mM NAD+, followed by flash-cooling in liq- NAD+ binding, accompanied by a reduction in the amount uid nitrogen. of active tetramer formed. Crystals of E427D complexed with AA were grown in

0.2 M MgCl2, 0.1 M Bis-Tris pH 5.4, and 25% (w/v) PEG 3350. Prior to crystallisation, the enzyme (6 mg/ml) was 5 | MATERIALS AND METHODS incubated overnight with 80 mM AA. Microseeding with wild-type crystals was employed. The cryoprotectant con- 5.1 | Protein expression and purification sisted of the reservoir supplemented with 18% (v/v) ethylene glycol. Synthetic genes encoding E427Q, E427D, and E427G Crystals of E427G complexed with NAD+ were grown were generated using the Agilent QuikChange Lightning by co-crystallisation using 5 mg/ml enzyme and 5 mM site-directed mutagenesis kit using the primers listed in NAD+. Wild-type crystals were used for microseed 8 LACIAK ET AL. matrix-screening28 using Hampton Index screen. Drops individual conformations have refined occupancies of were set using an Oryx8 robot (Douglas Instruments). 0.47 to 0.53. Similarly, the NAD+ in E427G-NAD+ was Data collection quality crystals grew in Hampton Index modelled as a single conformation in five chains (occu- condition 71 (0.2 M NaCl, 0.1 M Bis-Tris pH 6.5, and 25% pancy = 1.0) and as dual conformations in three chains (w/v) PEG 3350). The cryoprotectant consisted of the res- (occupancy = 0.41-0.59). The NAD+ in E427D-NAD+ ervoir supplemented with 18% (v/v) ethylene glycol. was modelled as a single conformation in all chains X-ray diffraction data sets were collected at ALS (occupancy = 0.86-0.93). The occupancy of AA in E427D- beamline 4.2.2 using a Taurus-1 CMOS detector in shut- AA was fixed at 1.0. terless mode. Each data set consisted of 900 images cover- ing a rotation range of 180 with a total exposure time of 360 seconds. In some cases, two 180 scans, optimised to 5.4 | Small-angle X-ray scattering collect high and low resolution, were collected from the same crystal. The data sets were integrated and scaled Shutterless SAXS data collection was performed at with XDS.29 Merging of low and high resolution scans beamline 12.3.1 of the Advanced Light Source through was done with XSCALE.30 Intensities were converted to the SIBYLS Mail-in High Throughput SAXS program.38 amplitudes with AIMLESS.31 Data processing statistics Prior to SAXS analysis, purified protein samples were are listed in Table S2. We note that the deposited struc- passed over a Superdex 200 10-30 size-exclusion chroma- tures use the historic numbering of ALDH7A1 to be con- tography column in the presence of a buffer containing sistent with other ALDH7A1 structures in the PDB. The 50 mM HEPES (pH 8.0), 100 mM NaCl, 2% (v/v) glycerol, two numbering schemes differ by 28 residues, and as and 1 mM dithiothreitol. Samples were then sup- such, Glu427 is Glu399 in the deposited structures. plemented with 10 mM NAD+ and dialysed overnight at Except for E427D-NAD+, the space group is C2, and 4C against a buffer containing 50 mM HEPES (pH 8.0), the asymmetric unit contains eight protein chains 100 mM NaCl, 2% (v/v) glycerol, 1 mM dithiothreitol, arranged in two tetramers. This is the same crystal form and 10 mM NAD+. that was used previously to determine structures of wild- SAXS data were collected on a Pilatus detector type ALDH7A115,16 and several mutant variants.11 operating in shutterless mode, writing frames every Crystallisation trials of E427D-NAD+ generated a new 0.3 seconds. Buffer subtracted SAXS curves were aver-

ALDH7A1 crystal form with space group P21 and the fol- aged using SAXS FrameSlice by averaging the first lowing unit cell dimensions: a =82Å,b = 128 Å, 5 frames for the Guinier region (total 1.5 seconds), the c = 88 Å, and β = 101. The asymmetric unit contains first 10 second frames for the Porod region (total 3 sec- four protein chains arranged as a tetramer. onds), and the first 20 frames for the high q region PHENIX32 was used for refinement, and COOT33 was (total 6 seconds). PRIMUS19 was used to inspect the used for model building. The starting model for refine- merged data and to derive SAXS parameters. The max- ment of the C2 structures was prepared from the coordi- imum particle dimension was estimated from calcula- nates of wild-type ALDH7A1 complexed with NAD+ tions of the pair distribution function using GNOM39 (PDB 4ZUK) or AA (PDB 4ZUL) by removing ligands via PRIMUS. Theoretical SAXS curves were calculated and solvent, truncating the mutated residue and the cata- using FoXS40 and MultiFoXS.40 lytic Cys to Ala, and deleting the mobile C-terminus (last 12 residues). Initial crystallographic phases for the ACKNOWLEDGMENTS E427D-NAD+ structure were obtained by molecular We thank Jesse Wyatt for synthesising AASAL, Jay Nix replacement as implemented in MOLREP34 via CCP4i35 for assistance with data collection at Advanced Light using a monomer search model prepared from the coor- Source beamline 4.2.2, and Katherine Burnett for collect- dinates of wild-type ALDH7A1 complexed with NAD+ ing SAXS data through the SIBYLS mail-in program. (PDB 4ZUK). Polder omit maps aided in modelling Research reported in this publication was supported by ligands and the 427-loop.36 All the structures were vali- the National Institute of General Medical Sciences of the dated using the PDB validation server and MolProbity.37 National Institutes of Health under award number Refinement statistics are listed in Table S2. R01GM093123 (to J.J.T.). This research used resources of Partial occupancy of ligands was addressed as follows. the Advanced Light Source, which is a DOE Office of Sci- NAD+ in E427Q-NAD+ was modelled as a single confor- ence User Facility under contract No. DE- mation with occupancy of 1.0 in four of the eight chains AC02-05CH11231. Additional support for the SIBYLS in the asymmetric unit. In the other four chains, dual beamline comes from the National Institutes of Health conformations were modelled such that the occupancies project ALS-ENABLE (P30 GM124169) and a High-End of the A and B conformations summed to 1.0; the Instrumentation Grant S10OD018483. LACIAK ET AL. 9

CONFLICT OF INTEREST 8. Bennett CL, Chen Y, Hahn S, Glass IA, Gospe SM Jr. Preva- The authors declare no competing financial interest. lence of ALDH7A1 mutations in 18 north American pyridoxine-dependent seizure (PDS) patients. Epilepsia. 2009; 50:1167-1175. AUTHOR CONTRIBUTIONS 9. Perozich J, Nicholas H, Wang BC, Lindahl R, Hempel J. Rela- A.R.L., D.A.K., and J.J.T designed experiments. tionships within the aldehyde dehydrogenase extended family. A.R.L. and D.A.K. performed experiments. A.R.L., Protein Sci. 1999;8:137-146. D.A.K., and J.J.T. analysed data and wrote the article. All 10. Korasick DA, Wyatt JW, Luo M, et al. Importance of the C- authors contributed to data interpretation. All authors terminus of aldehyde dehydrogenase 7A1 for oligomerization reviewed the results and approved the final version of the and catalytic activity. Biochemistry. 2017;56:5910-5919. manuscript. 11. Laciak AR, Korasick DA, Wyatt JW, Gates KS, Tanner JJ. Structural and biochemical consequences of pyridoxine- ETHICS STATEMENT dependent epilepsy mutations that target the aldehyde binding site of aldehyde dehydrogenase ALDH7A1. FEBS J. This article does not contain any studies with human or 2019. https://febs.onlinelibrary.wiley.com/doi/full/10.1111/ animal subjects performed by the any of the authors. febs.14997 12. Koncitikova R, Vigouroux A, Kopecna M, et al. Role and struc- DATABASES tural characterization of plant aldehyde dehydrogenases from Coordinates and structural factor amplitudes have been family 2 and family 7. Biochem J. 2015;468:109-123. deposited in the Protein Data Bank under the following 13. Coulter-Mackie MB, Li A, Lian Q, Struys E, Stockler S, accession codes: 6O4K, 6U2X, 6O4I, and 6O4L. SAXS Waters PJ. Overexpression of human antiquitin in E. coli: enzy- data sets have been deposited in the Small-Angle Scatter- matic characterization of twelve ALDH7A1 missense mutations associated with pyridoxine-dependent epilepsy. Mol Genet ing Biological Data Bank under the following accession Metab. 2012;106:478-481. codes: SASDGH4, SASDGJ4, SASDGK4, SASDGL4, 14. Korasick DA, White TA, Chakravarthy S, Tanner JJ. NAD(+) SASDGM4, SASDGN4, SASDGP4, SASDGQ4, SASDGR4, promotes assembly of the active tetramer of aldehyde dehydro- SASDGS4, SASDGT4, SASDGU4. genase 7A1. FEBS Lett. 2018;592:3229-3238. 15. Luo M, Tanner JJ. Structural basis of substrate recognition by ORCID aldehyde dehydrogenase 7A1. Biochemistry. 2015;54:5513-5522. David A. Korasick https://orcid.org/0000-0002-6337- 16. Brocker C, Lassen N, Estey T, et al. Aldehyde dehydrogenase 2085 7A1 (ALDH7A1) is a novel enzyme involved in cellular defense John J. Tanner https://orcid.org/0000-0001-8314-113X against hyperosmotic stress. J Biol Chem. 2010;285:18452- 18463. 17. Korasick DA, Tanner JJ. Determination of protein oligomeric REFERENCES structure from small-angle X-ray scattering. Protein Sci. 2018; 1. Coughlin CR 2nd, Swanson MA, Spector E, et al. The genotypic 27:814-824. spectrum of ALDH7A1 mutations resulting in pyridoxine 18. Fischer H, de Oliveira Neto M, Napolitano HB, Polikarpov I, dependent epilepsy: a common epileptic encephalopathy. Craievich AF. Determination of the molecular weight of pro- J Inherit Metab Dis. 2019;42:353-361. teins in solution from a single small-angle X-ray scattering 2. van Karnebeek CD, Tiebout SA, Niermeijer J, et al. Pyridoxine- measurement on a relative scale. J Appl Crystallogr. 2010;43: dependent epilepsy: An expanding clinical spectrum. Pediatr 101-109. Neurol. 2016;59:6-12. 19. Konarev PV, Volkov VV, Sokolova AV, Koch MHJ, Svergun 3. Stockler S, Plecko B, Gospe SM Jr, et al. Pyridoxine dependent DI, IUCr. PRIMUS: a windows PC-based system for small- epilepsy and antiquitin deficiency: clinical and molecular char- angle scattering data analysis. J Appl Crystallogr. 2003;36:1277- acteristics and recommendations for diagnosis, treatment and 1282. follow-up. Mol Genet Metab. 2011;104:48-60. 20. Rambo RP, Tainer JA. Accurate assessment of mass, models 4. Scharer G, Brocker C, Vasiliou V, et al. The genotypic and phe- and resolution by small-angle scattering. Nature. 2013;496: notypic spectrum of pyridoxine-dependent epilepsy due to 477-481. mutations in ALDH7A1. J Inherit Metab Dis. 2010;33:571-581. 21. Sheikh S, Ni L, Hurley TD, Weiner H. The potential roles of 5. Mills PB, Struys E, Jakobs C, et al. Mutations in antiquitin in the conserved amino acids in human liver mitochondrial individuals with pyridoxine-dependent seizures. Nat Med. aldehyde dehydrogenase. JBiolChem. 1997;272:18817- 2006;12:307-309. 18822. 6. Plecko B, Paul K, Paschke E, et al. Biochemical and molecular 22. Ni L, Sheikh S, Weiner H. Involvement of glutamate characterization of 18 patients with pyridoxine-dependent epi- 399 and lysine 192 in the mechanism of human liver mitochon- lepsy and mutations of the antiquitin (ALDH7A1) gene. Hum drial aldehyde dehydrogenase. J Biol Chem. 1997;272:18823- Mutat. 2007;28:19-26. 18826. 7. Mills PB, Footitt EJ, Mills KA, et al. Genotypic and phenotypic 23. Perez-Miller SJ, Hurley TD. Coenzyme isomerization is integral spectrum of pyridoxine-dependent epilepsy (ALDH7A1 defi- to catalysis in aldehyde dehydrogenase. Biochemistry. 2003;42: ciency). Brain. 2010;133:2148-2159. 7100-7109. 10 LACIAK ET AL.

24. Jeffrey GA. An Introduction to Hydrogen Bonding. New York: 36. Liebschner D, Afonine PV, Moriarty NW, et al. Polder maps: Oxford University Press; 1997. improving OMIT maps by excluding bulk solvent. Acta 25. Smith PE, Tanner JJ. Molecular dynamics simulations of Crystallogr D Struct Biol. 2017;73:148-157. NAD+ in solution. J Am Chem Soc. 1999;121:8637-8644. 37. Davis IW, Leaver-Fay A, Chen VB, et al. MolProbity: all-atom 26. Bottoms CA, Smith PE, Tanner JJ. A structurally conserved contacts and structure validation for proteins and nucleic acids. water molecule in Rossmann dinucleotide-binding domains. Nucleic Acids Res. 2007;35:W375-W383. Protein Sci. 2002;11:2125-2137. 38. Dyer KN, Hammel M, Rambo RP, et al. High-Throughput SAXS 27. Korasick DA, Tanner JJ, Henzl MT. Impact of disease-linked for the Characterization of Biomolecules in Solution: A Practical mutations targeting the oligomerization interfaces of aldehyde Approach. Totowa, NJ: Humana Press; 2014:245-258. dehydrogenase 7A1. Chem Biol Interact. 2017;276:31-39. 39. Svergun DI, IUCr. Determination of the regularization parame- 28. Till M, Robson A, Byrne MJ, et al. Improving the success rate ter in indirect-transform methods using perceptual criteria. of protein crystallization by random microseed matrix screen- J Appl Crystallogr. 1992;25:495-503. ing. J Vis Exp. 2013;78. https://doi.org/10.3791/50548. 40. Schneidman-Duhovny D, Hammel M, Tainer JA, Sali A. FoXS, 29. Kabsch W. XDS. Acta Crystallogr D Biol Crystallogr. 2010;66: FoXSDock and MultiFoXS: single-state and multi-state struc- 125-132. tural modeling of proteins and their complexes based on SAXS 30. Kabsch W. Integration, scaling, space-group assignment and profiles. Nucleic Acids Res. 2016;44:W424-W429. post-refinement. Acta Crystallogr Sect D. 2010;66:133-144. 31. Evans PR, Murshudov GN. How good are my data and what is SUPPORTING INFORMATION the resolution? Acta Crystallogr D Biol Crystallogr. 2013;69: Additional supporting information may be found online 1204-1214. in the Supporting Information section at the end of this 32. Adams PD, Afonine PV, Bunkoczi G, et al. PHENIX: a compre- article. hensive python-based system for macromolecular structure solution. Acta Crystallogr Sect D. 2010;66:213-221. 33. Emsley P, Lohkamp B, Scott WG, Cowtan K. Features and How to cite this article: Laciak AR, development of coot. Acta Crystallogr D Biol Crystallogr. 2010; Korasick DA, Gates KS, Tanner JJ. Structural 66:486-501. analysis of pathogenic mutations targeting Glu427 34. Vagin A, Teplyakov A. MOLREP: an automated program for molecular replacement. J Appl Crystallogr. 1997;30:1022-1025. of ALDH7A1, the hot spot residue of pyridoxine- – 35. Potterton E, Briggs P, Turkenburg M, Dodson E. A graphical dependent epilepsy. J Inherit Metab Dis. 2019;1 10. user interface to the CCP4 program suite. Acta Crystallogr D https://doi.org/10.1002/jimd.12184 Biol Crystallogr. 2003;59:1131-1137.