Human Mutation

Funding information Health and Labor Sciences Research Grants, Grant/Award Number: H29‐nanchitou(nan)‐ ippan‐051; Grant‐in‐Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan, Grant/Award Number: 16K09962; AMED, Grant/Award Number: JP17ek0109276 Abstract Mitochondrial acetoacetyl‐CoA thiolase (T2, encoded by the ACAT1 gene) deficiency is an inherited disorder of ketone body and isoleucine metabolism. It typically manifests with episodic ketoacidosis. The presence of isoleucine‐derived metabolites is the key marker for biochemical diagnosis. To date, 105 ACAT1 variants have been reported in 149 T2‐deficient patients. The 56 disease‐associated missense ACAT1 variants have been mapped onto the crystal structure of T2. Almost all these missense variants concern residues that are completely or partially buried in the T2 structure. Such variants are expected to cause T2 deficiency by having lower in vivo T2 activity because of lower folding efficiency and/or stability. Expression and activity data of 30 disease‐associated missense ACAT1 variants have been measured by expressing them in human SV40‐transformed fibroblasts. Only two variants (p.Cys126Ser and p.Tyr219His) appear to have equal stability as wild‐type. For these variants, which are inactive, the side chains point into the active site. In patients with T2 deficiency, the genotype does not correlate with the clinical phenotype but exerts a considerable effect on the biochemical phenotype. This could be related to variable remaining residual T2 activity in vivo and has important clinical implications concerning disease management and newborn screening.

In the biosynthetic direction, thiolases catalyze the formation of a carbon-carbon bond through a Claisen condensation mechanism (from two acetyl-CoA molecules) and in the reverse, degradative direction a C-C bond is broken through thiolysis (in the presence of CoA), resulting in chain shortening of the acyl chain by two carbon atoms (in case the substrate is an unbranched acyl chain) or by three atoms (in case the substrate is a 2-methyl-branched acyl chain), such as for example catalyzed by the T2 (Figure 1; Haapalainen, Meriläinen, & Wierenga, 2006;Song et al., 1994). No cofactors are required for the catalytic activity of thiolases, and each thiolase catalyzes the reaction in both directions. The crystal structures of several thiolases have been reported (Haapalainen et al., 2006;Kiema et al., 2019). From this structural information as well as from extensive sequence alignment, a classification of thiolases has been proposed (Anbazhagan et al., 2014).
The crystal structure of the wild-type human T2 thiolase tetramer has been reported in 2007 . Ketone bodies (acetoacetate and 3-hydroxybutyrate) are important energy sources for most tissues, particularly the brain. Ketone body synthesis begins in the liver by β-oxidation of free fatty acids to output acetyl-CoA and acetoacetyl-CoA. T2 in the liver catalyzes the Claisen condensation of two acetyl-CoA molecules into acetoacetyl-CoA. In extrahepatic tissues, T2 is responsible for thiolytic cleavage of acetoacetyl-CoA into two molecules of acetyl-CoA. T2 deficiency causes episodic ketoacidosis. This indicates that T2 deficiency impedes ketolysis to a greater extent than ketogenesis. The abundant amount of T1 in the liver likely compensates for T2 deficiency in ketogenesis . Potassium ions specifically enhance the activity of T2 but do not change that of T1 and other thiolases, therefore the potassium ion-activated acetoacetyl-CoA thiolase assay remains the gold-standard test for the T2 enzyme assay (Middleton, 1973).
Herein, we review 105 ACAT1 variants that have been reported in 149 patients with T2 deficiency; we use the term "disease-associated F I G U R E 1 The reactions catalyzed by the T2 thiolase. (a) The biosynthetic reaction: The substrates are two molecules of acetyl-CoA. (b) The degradative reaction: The substrates are 2-methylacetoacetyl-CoA (or acetoacetyl-CoA) and CoA. In both directions, the reaction mechanism proceeds via a covalent intermediate, in which the nucleophilic cysteine, Cys126 in human T2, becomes acetylated in the biosynthetic as well as in the degradative reactions ACAT1 variants" to refer to variants associated with T2 deficiency. A discussion on non-disease-associated ACAT1 variants is beyond the scope of this review. We discuss important structural features of human T2 and the location of the disease-associated missense ACAT1 variants in the context of the crystal structure of human T2. To increase the understanding of this rare disease, we also discuss its clinical and laboratory implications.

| THE TGENE AND DISEASE-ASSOCIATED VARIANTS
The human ACAT1 gene (NCBI reference sequence: NG_009888.1) is located on chromosome 11q22.3-q23.1, spanning approximately 27 kb. This gene contains 12 exons interspersed by 11 introns. The 5ʹ-flanking region lacks a classic TATA box, but it contains two CAAT boxes and is GC rich. These features are characteristic of housekeeping genes. Human T2 complementary DNA (cDNA; NCBI reference sequence: NM_000019.3) spans about 1.5 kb. It encodes a precursor protein (NCBI reference sequence: NP_000010.1) composed of 427 amino acids, including a leader polypeptide of 33-amino acid (Kano et al., 1991).
The sequence of human T2 is shown in Figure 3. The available data on ACAT1 variants associated with T2 deficiency are shown in three tables. Table 1,2 have the information on the diseaseassociated missense variants. These two tables also describe information on the location of the variant site with respect to the structure, in particular, whether the side chain of a residue is buried or whether it is exposed to bulk solvent. Table 1 lists the missense variants that have also been characterized with respect to (a) expression efficiency and (b) catalytic activity properties. The experimental details related to these characterizations are provided in the Supporting Information. For some variants, this information is available for expression at three temperatures; 30, 37, and 40°C. As can be seen in Table 1, there is generally a good correlation between the results obtained at different temperatures (e.g., whenever the data of expression at three temperatures are available, then the expression levels are the highest at 30°C and the lowest at 40°C). In addition, the activity recovery is generally never higher than the expression recovery. Table 3 lists other disease-associated variants (ATG initiation codon, insertions, deletions, duplications, nonsense and aberrant splicing). Figure 4 depicts the location of the diseaseassociated variants with respect to the exons of the ACAT1 gene.

| STRUCTURAL FEATURES OF THE T2 THIOLASE
Human T2 is initially synthesized in the cytosol as a 45-kDa precursor that matures, following mitochondrial entry, to a homotetramer of 41-kDa subunits (Fukao et al., 1990;Middleton, 1973).
The leader peptide is cleaved off on entry into the mitochondria. The overall structure of the tetramer is shown in Figure 5 and Figure S1.
The active site is located at the interface of the tight dimer, as shown in Figure 6 and Figure S2. The construct used for the protein crystallographic studies starts at residue Val34 (which was changed into an alanine to provide better yields when expressed as a recombinant protein in Escherichia coli) and the C-terminus is residue Leu427. In the crystal structure (PDB code 2IBW), residues Pro37 to Leu427 are well ordered and are included in the final model  for each of the four chains of the tetramer.
The N-and C-terminal residues are far away from the catalytic site, being on the opposite site of the subunit. The built model of each subunit has the distinct conserved thiolase superfamily fold that can be subdivided into the N-terminal domain, loop domain, and C-terminal domain Kiema et al., 2014Kiema et al., , 2019. The N-and C-terminal domains have the same βαβαβαββtopology and these two domains jointly form a five-layered α-β-α-β-α structure. The central α-layer consists of the two active site helices: Nα3 of the N-terminal domain and Cα3 of the C-terminal domain. The structure of the N-terminal domain of T2 is made by residues Pro37-Ser155 and Asn287-Leu309, whereas the C-terminal domain F I G U R E 2 Schematic drawing showing the T2 thiolase reaction in the synthetic direction. Two molecules of acetyl-CoA are converted into CoA and acetoacetyl-CoA. The role of the four catalytic residues (Cys126, Asn353, His385, Cys413 of human T2) is highlighted. These residues protrude into the catalytic site from the four catalytic loops (the CxS, NEAF, GHP, and CxG loops, respectively, shown in bold). Cys126 is the nucleophilic cysteine and Cys413 is the acid/base cysteine. The side chains of Asn353 (fixing Wat98) and His385, as well as the main chain N-atoms of the CxS and CxG loops, contribute to the two oxyanion holes (OAH1 and OAH2, shown as shaded semicircles). These oxyanion holes stabilize the negative charge that develops during the reaction on the thioester oxygen atom of the reaction intermediates, being therefore also critically important for catalysis. The short-curved arrows visualize the breaking/forming of bonds ABDELKREEM ET AL.  .
The functional site of each of the subunits of the T2 tetramer is very extensive, including not only the catalytic residues but also the residues that shape the CoA binding site. These residues are part of F I G U R E 3 The sequence of the human mitochondrial acetoacetyl-CoA thiolase (T2, UniProt code: P24752) with nomenclature of secondary structure, sequence fingerprints, and loops. The N-terminal region is the mitochondrial leader sequence, which is cleaved off on entry into the mitochondria. The secondary structure is obtained from the structure of the human T2 (PDB code: 2IBW) using the ESPript 3.0 server (Robert & Gouet, 2014) and shown above the sequence. An asterisk (*) above the sequence marks every tenth residue. The mature sequence starts at Val34, indicated by a black circle (•) above the sequence. Important active site loops that are near the catalytic site are identified below the sequence with their sequence fingerprint. The nomenclature of the functional regions of the loop domain (residues 156-286) is also given below the sequence. The structural properties of the latter loop regions are visualized in Figure 7 and Figure S3 T A B L E 1 Missense ACAT1 variants associated with T2 deficiency, with available expression and activity data (n = 30)    Figure 4). Exon 11 contains the highest number (n = 15), followed by exon 6 (n = 11), exon 5 (n = 9), and exons 7 and 9 (each = 8  (Nguyen et al., 2017). This highly conserved residue is changed into glutamine (c.623G>A, p.Arg208Gln) in two other families (Sakurai et al., 2007). The importance of this residue for the enzymatic function is discussed in subsequent sections.

F I G U R E 5
The structure of the T2 tetramer (PDB entry 2IBW), complexed with CoA. The bound CoA molecules are shown as stick models. The two tight dimers (below and above; side view) are assembled into tetramers via the four tetramerization loops (in the middle). "cationic" labels one of the cationic loops, which points to the 3′-phosphate of the CoA bound in the active site of the opposing dimer. Stereo view is provided in Figure S1 F I G U R E 6 The structure of the T2 tight dimer (PDB entry 2IBW). (a) Top view (view approximately down the local two fold axis of the tight dimer). (b) Side view (rotated by 90°around the horizontal with respect to the top view, same view as in Figure 5). The bound CoA molecules are shown as stick models. In the left subunit, the Ndomain, loop domain, and C-domain are colored as purple, blue, and green ribbons, respectively. In the right subunit, the N-domain, loop domain, and C-domain are colored as yellow, orange, and cyan ribbons, respectively. "cationic" and "tetra" identify the cationic and tetramerization loops, respectively. Stereo views are provided in Figure S2 The second most common disease-associated ACAT1 variant is c.1006-1G>C that has been identified in 13 families, most of which are Vietnamese. It affects a highly conserved point at the splice acceptor site of intron 10, altering the Shapiro and Senapathy score from 67 to 49.5 (Shapiro & Senapathy, 1987). cDNA analysis of T2-deficient patientʼs fibroblasts revealed that c.1006-1G>C is associated with exon 11 skipping . Exon 11 skipping causes a frameshift of the coding sequence, which is predicted to exert drastic effects on the variant T2 protein, truncating it prematurely with loss of 53 C-terminal residues. Indeed, the T2 activity and protein were virtually absent in fibroblasts of a patient homozygous for c.1006-1G>C variant . The third most common disease-associated ACAT1 variant and the most common missense variant is c.578T>G (p.Met193Arg) that has been detected in eight families, most of which are from India. This is followed by c.455G>C (p.Gly152Ala) that was found in six families. Transient expression analysis of both p.Met193Arg and p.Gly152Ala variant T2 cDNAs revealed no residual enzyme activity (Abdelkreem, Akella, et al., 2017;Zhang et al., 2004).

| Disease-associated missense ACAT1 variants
The disease-associated missense ACAT1 variants (n = 56) are not uniformly distributed across the ACAT1 gene (Figure 4). Exons 6 and 11 contain the highest number of such variants (nine for each), 4.2 | Structure-function relationship of missense variants whose expression levels are equal or greater than 25% that of wild-type The 30 disease-associated missense ACAT1 variants listed in Table 1 concern those variants for which the catalytic and expression properties have been determined at 37°C. All variants listed in Table   1 have low activities; c.431A>C (p.His144Pro) variant has the highest activity (25% that of wild-type T2). For only two variants, c.377G>C (p.Cys126Ser) and c.655T>C (p.Tyr219His), the expression level is similar to that of wild-type; however, the catalytic activity for these two variants is 0%. From the structure analysis, it can be seen that both residues are essential for enzyme function. Tyr219 interacts both with the potassium ion and with the adenine moiety of CoA and Cys126 is the nucleophilic cysteine (Figure 1) Figure 6a). (b) Side view (same as Figure 6b). The loop domain protrudes out of Nβ4 and ends at Nβ5 of the N-terminal domain (Figure 3). The covering loop "cov" is in orange, the cationic loop is in green, the adenine loop is in red and the pantetheine loop is in purple. "tetra" identifies the tetramerization loop. The Lα2 and Lα3 helices are also labeled. The bound CoA molecule is shown as a stick model. Stereo views are provided in Figure S3 ABDELKREEM ET AL.

| 1655
loop that interacts with the loop domain. Each of these four residues is in a surface loop. From the structure, it is predicted that these four variants would allow the formation of the tetramers assembly but nevertheless, the catalytic activity is very low for p.Asp186Tyr, p.Arg208Gln, and p.Asn282His. Only p.Ile323Thr has a catalytic activity of 20% that of wild-type T2. Further information on the structure-function relationship of these four variants is given below.
The p.Asp186Tyr variant (T2 protein level, 33%; measured activity, 0%) is in the covering loop (Figure 8). This loop stabilizes the conformation of the Cβ1-Cα1 loop and the Cβ4-Cβ5 loop. The latter loop provides the acid/base cysteine, Cys413, which is an essential catalytic residue and this loop also contributes to OAH2.
These two functionalities are essential for the catalytic properties and therefore it is predicted that this variant inactivates T2 thiolase.
The p.Arg208Gln variant (T2 protein level, 50%; measured activity, 0%) is in Lα3. The Arg208 side chain makes a hydrogen bond interaction with OD2(Asp212) and the backbone oxygen of Leu267, which is in a loop region immediately after the adenine binding loop. The latter interaction stabilizes this loop at the correct position for substrate binding (Figure 8).
The p.Asn282His variant (T2 protein level, 50%; measured activity, 0%) is in the pantetheine loop, at the end of the loop F I G U R E 8 Missense variants of residues in loops on the surface of the T2 tetramer (PDB entry 2IBW). The visualized loop residues are either in the loop domain (panels a, b, c) or interact with the loop domain (panel d). Expression of variant T2 cDNAs containing these variants produces T2 protein levels of 25% or higher compared to wild-type T2, as discussed in the text. These panels are zoomed-in views, of the loop domain (same view as in Figure 7b)

| Structural analysis of all missense variants
For each of the missense variants listed in Table 1, information is provided in the last column concerning the location of the variant residue with respect to the structure of T2. Considering the structure of the tetramer, three residue categories have been defined, being either (a) completely buried ("buried"), or (b) near the surface, being partially buried ("surface") or (c) having a side chain that points towards the solvent ("exposed side chain"). For some of the latter residues, the side chain is interacting closely with the rest of the protein. If the solvent exposed side chain is only loosely interacting with the rest of the protein, then for variants which do not change much the side chain properties (e.g., no change in polarity), it is predicted that the in vivo folding efficiency or stability will be similar to that of wild-type. This simple prediction scheme is not valid whenever the variant concerns a proline and/or glycine. In Table 1, it concerns seven out of the 30 listed variants, which are either buried or surface residues. Except for p.His144Pro (T2 protein level, 50%; measured activity, 25%), these variants are poorly expressed (protein level is equal or less than 10%).
Variants that are predicted not to change the folding or stability properties, but nevertheless are observed to be disease-associated, identify residues that are important for the catalytic properties of the native tetramer assembly. The current set of disease-associated variants listed in Table 1 includes five such residues, Cys126, Met193, Arg208, Tyr219, and Ile323. The p.Cys126Ser and p.Tyr219His variants are both expressed at the same level as that of wild-type, as discussed in the previous section. The p.Arg208Gln and the p.Ile323Thr variants have also been discussed in the previous section. The p.Met193Arg variant concerns a residue whose side chain points toward bulk solvent and is therefore predicted not to interfere with folding or stability. However, the side chain of Met193 points into the narrow pantetheine binding tunnel. The experimental data show that the bulkier and more polar arginine side chain does not allow proper folding of this T2 thiolase variant.
Most of the residues listed in Table 1 (16 out of 30) are completely buried in the structure. These variants are predicted to adversely affect the folding and/or stability and therefore are predicted to have lower catalytic efficiency. Indeed, the expression levels of all these variants are less than 25% that of wild-type, except for p.Asn93Ser, p.Asp186Tyr, and p.Asn282His (Figure 8). The latter two variants are located in loops near the surface of the tetramer, which apparently allows for partial folding, but the mature protein has low catalytic activity, as discussed above. Asn93 is buried at the dimer interface and the variant residue is predicted to prevent assembly of the functional tetramer, as also discussed above. Table 2 also lists missense disease-associated variants, but for these variants, there are no folding/stability or activity data. These variants have also been mapped onto the structure, and also for this set, most of the variants (23 out of 26) concern residues that are completely or partially buried. These 23 variants are predicted to produce expression levels less than that of wild-type. Some of these variants concern residues that are located at the monomer-monomer interface of the tight dimer. It is possible that these variants may be expressed in a soluble, monomeric form, but these forms are likely not active.
It will be particularly interesting to find out the expression and activity properties of the variants of Table 2, which are classified as having "exposed side chain". It concerns the three variants c.578T>C (p.Met193Thr), c.760G>A (p.Glu254Lys), and c.1160T>C (p.Ile387Thr).
The p.Met193Thr variant concerns a residue which is located at the beginning of helix Lα2, and whose side chain is usually a hydrophobic residue, pointing into the narrow pantetheine binding tunnel. Like for the p.Met193Arg variant (Table 1) It points into the catalytic site and is therefore predicted to interfere with proper catalytic function.

| Disease-associated splice variants
Most aberrant splicing-associated variants affect splice donor (n = 9) and acceptor (n = 6) sites in the ACAT1 gene ( Figure 4; Table 3). These The resulting frameshift replaces the last 54 amino acid residues, including the catalytically essential residues His385 and Cys413, with 69 different C-terminal residues (Fukao, Boneh, Aoki, & Kondo, 2008 (Fukao, Horikawa, et al., 2010. Without this information, c.951C>T might wrongly be regarded only as a benign synonymous variant. Of note, RNA sequencing is a useful technology to reveal abnormally spliced transcripts.

| Disease-associated deletion/insertion/ duplication variants
These variants are also listed in Table 3. It is difficult to predict how such variants would affect the folding and/or stability of the T2 protein. In such cases, expression analysis of variant cDNAs has not been routinely performed.
Furthermore, five large deletions/insertions/duplications have been reported in the ACAT1 gene. g.20623_29833delinsGTAA includes deletion of exons 6-11 (Nguyen et al., 2017). c.731-46_752del (a 68bp deletion) involves the splice acceptor site of intron 7, causing exon 8 skipping (Fukao, Song, et al., 1995). The other three variants could be attributed to Alu elements-mediated unequal homologous recombination Zhang et al., 2006). We established multiplex ligation-dependent probe amplification (MLPA) analysis for ACAT1, which is useful to identify these large gene rearrangements . Of note, a recent ACAT1 minigene experiment demonstrated that insertion of AluY-partial AluSz6-AluSx in an antisense direction within intron 9 has a negative effect on exon 10 inclusion. This effect is (a) distance dependent-the shorter the distance between the antisense Alu element and exon 10, the greater the skipping of exon 10; (b) additive with that of an ESE variant (c.951C>T) in exon 10; and (c) canceled by the c.941C>G substitution at the first nucleotide of exon 10, which optimizes the splice acceptor site of intron 9. Accordingly, intronic antisense Alu elements have a negative splicing effect on close downstream exons, particularly when splice acceptor sites are suboptimal (Nakama et al., 2018).

| Other disease-associated variants
Five nonsense ACAT1 variants have been reported (Table 3; Figure   4). mRNAs with premature termination are mostly subjected to nonsense-mediated decay. In the case of c.814C>T (p.Gln272*), this also causes skipping of exon 8 in 25% of transcripts. This phenomenon was previously termed as nonsense-associated alternative splicing  and recently designed as exon skipping caused by a variant at an ESE sequence (discussed above).

| Biochemical and laboratory significance
The genotype exerts a considerable effect on the biochemical phenotype of patients with T2 deficiency. Based on the T2 enzymatic activity detected on expression of variant cDNAs, patients with T2 deficiency can be divided into two categories: Those with "mild" variants, in whom at last one of the two variant alleles retains some residual T2 activity, and those with "severe" variants, in whom none of the two variant alleles has any residual T2 activity . Patients with mild variants can develop episodic ketoacidosis as severe and frequent as those with severe variants; however, the above mentioned isoleucine-catabolic intermediates, essentially TIG, tend to be more subtle in the former patients not only in stable states but also during acute ketoacidosis , Fukao et al., 2012 Fukao et al., 2014Fukao et al., , 2018. Of note, secondary carnitine deficiency is rare in T2 deficiency but if present, it may suppress β-oxidation and modify the clinical manifestation of T2 deficiency from ketoacidotic to hypoketotic hypoglycemic events (Alijanpour et al., 2019).
Patients with T2 deficiency had been thought to be asymptomatic between episodes unless a previous severe episode of ketoacidosis causes irreversible neurological damage. However, an increasing body of evidence indicates that chronic neurological impairment, mainly extrapyramidal manifestations, can exist independent of frank ketoacidosis even in patients with T2 deficiency confirmed at the molecular level (Buhaş et al., 2013;Fukao et al., 2018;Paquay et al., 2017). In vitro studies indicate that 2MAA and 2M3HB exert neurotoxic effects (Leipnitz et al., 2010;Rosa et al., 2005).
Given the large number of private (occurring only in one family) disease-associated ACAT1 variants, T2 deficiency lacks an obvious correlation between the genotype and the clinical phenotype, including the age at onset, severity and frequency of ketoacidotic episodes, and eventual outcome. This is evident from the provided comprehensive list of patients with T2 deficiency reported in the literature (Supporting Information Table). Environmental/acquired factors, such as ketogenic triggers, considerably contribute to the clinical presentation (Thümmler, Dupont, Acquaviva, Fukao & De Ricaud, 2010). However, several reports show that T2 deficiency has variable clinical phenotypes even among patients who share not only identical genotype but also similar environmental factors (Abdelkreem, Alobaidy, et al., 2017;Fukao et al., 2012;Köse et al., 2016;Nguyen et al., 2017;Thümmler et al., 2010). Proper acute and preventive treatment seems crucial for a favorable outcome (Hori et al., 2015;Nguyen et al., 2017).

| CONCLUDING REMARKS
Functional studies of 30 disease-associated missense T2 variants have been performed in vitro, using the potassium ion-activated acetoacetyl-CoA degradation assay and for all these variants low activity (equal or less than 25% that of wild-type T2) is observed (Table 1). From the available information, patients with T2 deficiency can be divided into those with "mild" variants, in whom at least one of the two variant alleles retains some residual T2 activity, and those with "severe" variants, in whom none of the two variant alleles has any residual T2 activity. However, patients with mild variants can develop episodic ketoacidosis as severe and frequent as those with severe variants . This raises questions whether the T2 activity measured in vitro using acetoacetyl-CoA as a substrate fully reflects the in vivo T2 deficiency, and whether it is better to use 2-methylacetoacetyl-CoA (or both acetoacetyl-CoA and 2-methylacetoacetyl-CoA) as specific substrates. Indeed, 2-methylacetoacetyl-CoA thiolase assay is more sensitive for detecting isoleucine catabolism deficiencies (Middleton & Bartlett, 1983). This substrate is not currently commercially available but can be prepared by published protocols. For one disease-associated variant (p.Ile323Thr; 25% expression, 20% remaining activity) the variant could affect the structure of the loop that shapes the binding pocket of the 2-methyl group. The experimental data suggest that the mature enzyme variant is fully active, but from the structural analysis it is predicted that the activity for the 2-methylacetoacetyl-CoA substrate could be much more affected. In the latter case, the isoleucine catabolism is much more affected than the ketone body metabolism. In any case, the analysis of the structural context of the missense variants shows that they concern also residues that are not near the active site, and it suggests that in almost all cases the identified disease-associated ACAT1 variants concern residues that are buried in the mature protein (Table 1) and therefore are predicted to deteriorate the stability and/or folding properties of the respective T2 variants, thereby decreasing the capacity to efficiently degrade 2-methylacetoacetyl-CoA as well as acetoacetyl-CoA.
Several patients with T2 deficiency developed chronic neurological impairment, mainly extrapyramidal, independent of frank ketoacidosis (Buhaş et al., 2013;Fukao et al., 2018;Paquay et al., 2017). In vitro studies indicate that 2MAA and 2M3HB exert neurotoxic effects (Leipnitz et al., 2010;Rosa et al., 2005). Therefore, accumulated isoleucine-catabolic metabolites may contribute to neurological impairment in patients with T2 deficiency Paquay et al., 2017). Accordingly, T2 deficiency should be considered not only as a ketolytic defect but also as a defect in isoleucine catabolism with the potential for insidious cerebral toxicity. However, it is likely that other genetic or environmental factors contribute to the neurotoxic effect of isoleucine metabolites, explaining why only a minority of T2-deficient patients have such neurological manifestations independent of the occurrence of severe metabolic crises. This is an important topic for future research.
Another unresolved issue is the therapeutic implications; the effectiveness of carnitine supplementation and protein, particularly isoleucine, restriction in preventing chronic neurological impairment remains to be determined .
Finally, many reported ACAT1 variants in patients diagnosed with T2 deficiency lack the experimental proof of decreased T2 activity (Tables 2 and 3). Structural analysis and in silico tools are useful to predict the pathogenic effect, but such predictions are not always true (see Table 1 and Section 4.4). In these cases, further laboratory studies to demonstrate the decreased T2 activity of such variants are required to confirm the diagnosis.