Pyruvate dehydrogenase complex deciency: updating the clinical, metabolic and mutational landscapes

Background : Pyruvate dehydrogenase complex (PDC) catalyzes the irreversible decarboxylation of pyruvate into acetyl-CoA which ultimately generates ATP. PDC deciency can be caused by alterations in any of the genes encoding its several subunits, and the resulting phenotype, though very heterogeneous, mainly affects the neuro-encephalic system. The aim of this study is to describe and discuss the clinic, metabolic and genotypic proles of thirteen PDC decient patients, thus seeking to establish possible genotype-phenotype correlations. Results : The mutational spectrum revealed that seven patients (54 %) carry mutations in the PDHA1 gene , encoding the E1α subunit, ve patients (38 %) carry mutations in the PDHX gene, encoding the E3 binding protein, and the remaining patient (8 %) harbors mutations in the DLD gene, encoding the E3 subunit. These data corroborate PDHA1 mutations as the predominant cause of PDC deciency, though revealing a notable prevalence of PDHX mutations among Portuguese patients, most of them carrying a seemingly private mutation (p.R284X). The biochemical analyses revealed high lactate and pyruvate plasma levels whereas de ratio L/P was under 16; enzymatic activities, when compared to control values, revealed to be independent from the genotype and ranged from 8.5% to 30% which may be considered a cut-off value for primary PDC deciency. Concerning the clinical features, all patients displayed developmental delay/psychomotor retardation, the severity of which seems to correlate with the type and localization of the mutation carried by the patient. The therapeutic options essentially go through the administration of a ketogenic diet and supplementation with thiamine, although arginine aspartate intake revealed to be benecial in some patients. Moreover, the in silico analysis of the missense mutations present in this PDC decient population allowed to understand the molecular mechanism underlying these pathogenic variants. Conclusion : The identication of the disease-causing mutations, together with the functional and structural characterization of the mutant protein variants, allows to get insight on the severity of the clinical phenotype and the selection


Background
Pyruvate dehydrogenase complex (PDC) de ciency, rst described in 1970 by Blass and colleagues (1), is an inborn error of mitochondrial energy metabolism. The pyruvate oxidation route, that bridges the cytosolic glycolytic pathway and the mitochondrial tricarboxylic acid cycle (2), involves not only PDC but also pyruvate transport and the ancillary metabolic routes associated with various cofactors. PDC, which irreversibly decarboxylates pyruvate to acetyl coenzyme A, comprises three functional (E1 or pyruvate dehydrogenase, E2 or dihydrolipoamide transacetylase, and E3 or dihydrolipoyl dehydrogenase) and one structural (E3BP or E3 binding protein, formerly designated by protein X) components, and the regulatory PDC kinases and PDC phosphatases, which control the complex activity via phosphorylation/dephosphorylation (3). The E1 subunit catalyzes the rst, rate-limiting and irreversible step of the reaction which takes place in two active sites formed at the interface between their α and β subunits, each requiring a thiamine pyrophosphate cofactor and magnesium ions for activity (4,5).
The great majority (77.3%) of patients with PDC de ciency harbor mutations in the X-linked PDHA1 gene which encodes the E1α subunit. Hemizygous males are usually symptomatic whereas heterozygous females present variable expression due to random patterns of X-inactivation in different tissues. Therefore, the clinical manifestations depend on the proportion of cells expressing the mutant E1α subunit. The remaining cases are caused by mutations in the PDHB gene encoding the E1β subunit (4.3%), in the DLAT gene encoding the E2 subunit (1.5%), in the DLD gene encoding the E3 subunit (6.2%) and nally in the PDHX gene encoding the E3BP subunit (10.7%) (6).
The clinical presentation is extremely heterogeneous, ranging from a fatal lactic acidosis and progressive neurological and neuromuscular degradation in the neonatal period to a chronic neurological dysfunction and neurodegenerative condition. Inadequate removal of lactate but mainly of pyruvate (7) results in lactic acidemia which, together with the blood lactate/pyruvate ratio ≤20, is recognized as an important biomarker (8,9). The primary phenotypic manifestation corresponds to an impairment of neurological and/or developmental functions, leading to a wide range of symptoms and clinical features: hypotonia, seizures, structural and functional brain abnormalities, ataxia, respiratory distress (like apnea and hypoventilation), facial dysmorphism, and peripheral neuropathy. According to their severity, the phenotypes can be divided into three main categories: neonatal, infantile and benign (10)(11)(12).
The existing therapeutic strategies target: i) the metabolic pathway, bypassing energy production via a ketogenic diet (11), ii) the regulatory system of the dysfunctional PDC, using xenobiotic inhibitors (13,14), or iii) cofactor(s) stimulation of the residual PDC activity with supplementation of these (15,16).
Nevertheless, none of these treatments is su ciently effective and the responsiveness is markedly individual. Thus, the precise diagnosis, including the identi cation of the genetic defect and the phenotypic consequences, is becoming increasingly decisive for selecting the appropriate therapeutic strategy. Knowledge on the functional and structural impact of the identi ed mutations on the resulting PDC component variant provides relevant information for informed decisions on the therapeutic approaches.
In this study, we describe and discuss the pro les of thirteen Portuguese PDC de cient patients with reference to their clinical data, biochemical ndings and results of DNA analysis, thus seeking to establish possible genotype-phenotype correlations. This is the rst report on phenotypic and genotypic variability in a subset of PDC de cient patients diagnosed in Portugal.

General information
The broad phenotypic spectrum of PDC de ciency, common to many other genetic mitochondrial disorders, often hampers the achievement of a rapid diagnosis. Although among the patients suspected to be PDC de cient based on clinical signs /symptoms and biochemical date, namely elevated plasma lactate (L) and pyruvate (P) levels with low L/P ratio and / or impaired PDC activity, we only considered those whose genetic diagnosis was accomplished. Accordingly, a total of 13 individuals, all of European ancestry, were enrolled in this study (Table 1).
The 13 patients, seven males and six females, were born between 1983 and 2018, their current ages varying from 20 months to 36 years, with a median age of 15.5 years. Consanguinity was reported only in three families harboring PDHX mutations. Among this group of 13 patients, only two were siblings, born from a triplet pregnancy after in vitro fertilization (two affected males and one unaffected female) who carry a PDHA1 mutation. No affected relatives are known in any of these families.
All patients showed the rst symptoms either in the neonatal period ( ve individuals) or during infancy (eight individuals). However, in some cases, the diagnosis was only achieved later, mainly among the older patients who presented a less severe clinical picture; the most striking example is the de ciency in E3 (also designated by dihydrolipoamide dehydrogenase, DLD), which was only diagnosed when the patient was 17 years old.
All patients but one are alive; the deceased patient is a boy carrying the c.1132C>T mutation in PDHA1 (generating the p.R378C E1α variant); the death occurring at 3.3 years and being caused by cardiorespiratory arrest.

Genetic ndings
This cohort of 13 patients revealed a mutational spectrum encompassing ten different mutations distributed by three different genes: ve in PDHA1, three in PDHX and two in DLD. The results revealed that seven patients carried mutations in PDHA1 (~54 %), ve carried mutations in PDHX (~38%), and one carried mutations in DLD (~8 %) ( Table 1).
It is interesting to note that most patients with mutations in PDHA1, presenting an X-linked inheritance pattern, are males (6/7) in whose families no consanguinity was reported, mutations likely arrived in all by a de novo mechanism, and mosaicism was absent. On the contrary, most patients with mutations in PDHX, displaying an autosomal recessive mode of inheritance, are females (4/5) and consanguinity was detected in three cases. Finally, the single patient with mutations in DLD, also displaying an autosomal recessive mode of inheritance, is a compound heterozygous female with no consanguinity reported in her family.
Concerning the type of mutations found in these patients, their severity is also predictably different according to the affected gene. Indeed, a single mutant allele in PDHA1 is su cient to cause PDC de ciency, including in females who carry two alleles.

Biochemical ndings
All patients, regardless of the carried mutation, were suspected of PDC de ciency by their elevated plasma lactate and pyruvate levels, ranging from 3.0 to 17.0 mmol·L -1 and from 0.27 to 0.81 mmol·L -1 , respectively. Lactate/pyruvate ratios, as expected, were always ≤16, i.e. in the normal range with a single exception. PDC de ciency was con rmed by determination of PDC activity. Enzymatic assays were performed either with circulating lymphocytes (12 patients) or cultured broblasts (two patients), both tissues having been analyzed in patient 11. All patients had their enzymatic activities con rmed in a second independent sample and the results always matched. Interestingly, a single false negative result was rstly determined and precisely in lymphocytes of a female patient later identi ed with a PDHA1 mutation. In our experience, the cut-off value for considering a primary PDC de ciency should be ≤ 30% of control activity and, indeed, all patients but one presented enzymatic activities ranging from 8.5% to 30%. The single exception was one of the siblings carrying the p.R378C variant in PDHA1, who presented 40 % of normal control activity (Table 1), surprisingly the one whose symptoms manifested earliest in the neonatal period.

Clinical features
The clinical features observed in these patients' cohort are displayed in Table 1 and, according to the most recently published data (6), can be roughly divided into two categories: one caused by PDHA1 and PDHX mutations, and the other caused by DLD mutations. Indeed, the clinical phenotypes manifested by patients carrying PDHA1 and PDHX mutations are quite variable and almost undistinguishable.
In the latter group, concerning neurological features, all individuals presented development delay/psychomotor retardation, which ranged from severe (eight patients) to moderate (three patients) or very mild (one patient who attended normal school). Hypotonia was also observed in all individuals, with a single exception (Patient 11) who nevertheless could display it at an earlier age. Seizures were reported for half the individuals, especially those carrying PDHA1 mutations (5/7 patients). On the contrary, ocular manifestations were more frequent in individuals harboring PDHX mutations (4/5 patients).
Microcephaly, dystonia and ataxia were also reported in half the patients of both mutational groups. As expected, facial dysmorphism was only detected in a single female patient carrying a PDHA1 mutation. The patient with DLD de ciency revealed moderate developmental delay/psychomotor retardation, seizures, hypotonia, dystonia and ataxia. Basal ganglia abnormalities were exclusively observed in patients carrying mutations in the PDHA1 gene. Cerebral atrophy, but not cerebellar atrophy, was detected mainly in patients carrying PDHX mutations. However, among patients carrying PDHA1 mutations, cerebral atrophy, when present, was always associated with cerebellar atrophy (Figures 1 and 2). The individual harboring mutations in the DLD gene (Patient 13) presented partial agenesis of corpus callosum.

Treatment
All the patients are under therapeutic measures which encompass the most frequent modalities: ketogenic diet, thiamine supplementation and antiepileptic drugs. Three patients with PDHA1 mutations and three with PDHX mutations are under ketogenic diet with perceived clinical bene ts. The long-term thiamine supplementation, as PDC cofactor, is prescribed in almost all patients (11/13), and antiepileptics drugs only in those presenting seizures (Table 1). Finally, it is interesting to mention that three of the patients carrying mutations in the PDHA1 gene were under arginine aspartate supplementation, with bene cial effects.
In silico analysis of missense mutations Bioinformatic analysis using the PolyPhen-2 server (17) suggested that all mutations but one affecting E1α subunit are most probably damaging. The E1α p.R302H, p.R378C and p.R378H variants displayed a score of 1.000 (sensitivity 0.00 and speci city 1.00), p.F205L displayed a score of 0.919 (sensitivity 0.81 and speci city 0.94), and p.R253G displayed a score of 0.007 (sensitivity 0.96 and speci city 0.75) thus being considered benign. As for E3 subunit, the p.P87S variant is predicted to be most probably damaging, with a score of 0.996 (sensitivity 0.55 and speci city 0.98).
To complement the information of the PolyPhen-2 server, which solely bases its predictions on the polypeptide sequence and overlooks other structural and functional details, such as e.g. cofactor binding and interaction with other proteins, we obtained and thoroughly inspected structural models of each E1α variant. The models obtained for the E1α p.R253G, p.R378C, p.R378H and p.F205L variants have been recently described by our group, attempting to understand the molecular mechanisms underlying the pathogenicity of the corresponding mutations (18). All mutations result in putative loss of H-bonds, electrostatic or hydrophobic interactions between the side chains of the substituted amino acids and neighboring residues, with predicted effects on Ploop destabilization, inter-subunit interactions and proper oligomeric assembly, as well as interaction with other PDC components. Herein, we additionally generated a structural model of the E1α p.R302H variant ( Figure 3). R302 is located at one end of the P-loop A, its side chain being within electrostatic and H-bonding distance to the side chain or main chain carbonyl moieties of Y287, R288, Y289, H290 (active site residue) and G298, all residues belonging to the same P-loop. Upon substitution by a histidine residue, most of the possible interactions between its side chain and other residues in the P-loop are lost. The single remaining H bond is that between the side chain imidazole and the main chain carbonyl of G298. Therefore, in the p.R302H E1α variant, the net loss of four possible side chain interactions with other P-loop A residues ( Figure 3) is likely to contribute to a more disordered loop and consequently lower enzymatic activity. Notably, the degree of disorder in P-loop A has been negatively correlated with E1 enzymatic activity, since an ordered loop favors TPP binding, which itself promotes P-loop A order (19).
To further understand the pathogenicity of the DLD mutation generating the E3 p.P87S variant, a structural model was obtained (Figure 4) based on the reported 3D crystallographic structure of another disease-causing E3 variant (PDB entry 6I4T) (20). As observed in Figure 4, P87 is located in a helix that lines with the avin adenine dinucleotide isoalloxazine ring and contains the active site disul de composed of C45 and C50. Substitution of P87 by a serine is likely to alter the exibility and thus the overall stability of the respective helix, possibly affecting the enzymatic activity.

Discussion
PDC de ciency diagnosis is extremely challenging due to a phenotypic presentation that can be observed in many other disorders, especially those causing altered mitochondrial metabolism (6,21,22). Given this context, we aimed to present the rst report on Portuguese PDC de cient patients combining information on the associated clinical, biochemical, enzymatic and genotypic spectra. The mutational spectrum of PDC de ciency in this group of patients revealed ten different mutations affecting three genes, PDHA1 (54 %), PDHX (38 %) and DLD (8 %). Concerning each gene de ciency prevalence, these data generally agree with literature surveys (6,9) and also with several studies focused on different populations (8,23,24). However, the frequency of mutations in the PDHA1 gene is lower than the usually reported average of 75-80 %.
The most striking evidence is the relatively high incidence of E3BP de ciency, mostly caused by a single PDHX variant, p.R284X, half the cases originating from the Azores Islands. This E3BP variant was rst described by our group (25) and, until now, only another Portuguese patient has been reported to carry this mutation (26). Furthermore, the mutational spectrum of E3BP de ciency in Portugal includes very severe mutations, leading to null alleles. Nevertheless, the older patients surprisingly reached adulthood, in line with the high proportion of long-term survival among reported E3BP de cient patients (8,27,28). In general, an overwhelming majority of the mutations hitherto identi ed in the PDHX gene are, as in this work, deletions, nonsense mutations, point mutations at intron-exon boundaries, or even large intra-genic rearrangements, expected to result in a complete absence of E3BP protein (8,(28)(29)(30). Despite this fact, the patients retain considerably signi cant PDC activity (20-30 %), taking into account the expected impairment on PDC assembly. On the one hand, as a structural subunit devoid of enzymatic activity, E3BP does not directly contribute to the complex catalytic activity. On the other hand, the signi cantly truncated E3BP, if present at all in the cell, would likely compromise the structure of the E2/E3BP PDC core and binding of the E3 component. Both E2 and E3BP components have a similar structure and domain organization, despite only E2 being catalytically active (31). However, the possibility of the E2 core directly binding to the E3 enzyme may underlie the observed residual PDC activity (28)(29)(30)32). In their recent work, Prajapati and collaborators report a non-uniform stoichiometry of the E2/E3BP PDC core. The imbalanced distribution of E2 and E3BP constituents of the trimeric units results into structurally dynamic E1 and E3 clusters (33). Moreover, for one of the proposed models of E. coli, PDC core is a fully functional E2 homotrimer operating in a "division-of-labor" mechanism, including binding of E3 component (33,34).
Concerning the mutational spectrum of E1α de ciency, ve different PDHA1 mutations were identi ed but a single one (p.R378C) in non-consanguineous patients. Almost all mutations affect an arginine codon (35) and those situated in exons 10 and 11 cause a severe phenotype, because the resulting protein variants present very low enzyme activity. On the contrary, the two mutations located in exon 7 originate moderate (p.F205L) or very mild (p.R253G) phenotypes. Interestingly, mutations affecting codon 378 are considered particularly lethal (9). Indeed, from our male patients carrying the p.R378C mutation, one deceased at three years of age and the twins, presently aged 8 years, display a severe clinical picture. However, a female patient bearing the p.R378H substitution reached the adulthood, probably due to a lyonization effect.
Regarding a possible genotype-dependent phenotypic presentation, our data is roughly suggestive of such a correlation. Effectively, the patients harboring the most deleterious mutations (p.R302H and p.R378C in PDHA1 gene, and all the mutations in PDHX gene) present the most severe phenotypes, involving serious psychomotor retardation, hypotonia and seizures, whereas those carrying less severe mutations accordingly display a better clinical outcome. The most puzzling observation concerns the female carrying mutations in DLD gene. Despite being a compound heterozygote bearing two severe mutations, her clinical course was satisfying until 2018 when she suffered an acute metabolic decompensation originating spastic tetraparesis with gait and language loss. Although she partially recovered language, currently she presents only a moderate to severe psychomotor handicap. Despite the E3 subunit being common to other enzyme complexes, such as α-ketoglutarate dehydrogenase and branched-chain amino acid dehydrogenase, this patient did not display the associated biochemical or clinical phenotypes.
Irrespectively of our patients cohort size, the majority of our PDC de cient patients remarkably reached adulthood, as opposed to several other reports (9,23,24,36). Concerning the therapeutic measures to which these individuals are subjected, it is clear they are only palliative, since all but one patient continue presenting clinical features ranging from moderate to severe forms. The single exception is Patient 2 who seems to represent an exceptional case, because his treatment only encompasses thiamine (E1 subunit cofactor) and arginine aspartate (potential protein stabilizer) supplementation (37). Indeed, this patient carried the p.R253G mutation that originates an E1α subunit whose in silico and in vitro analyses with the recombinant E1 variant exhibited proneness to aggregation and low enzymatic activity.
In conclusion, the identi cation of the disease-causing mutations, together with the functional and structural characterization of the mutant protein variants, allows to get insight on the severity of the clinical phenotype and the selection of the most appropriate therapy, namely the option for a ketogenic diet.

Cohort of patients
This study included all PDC de cient patients whose diagnosis was con rmed at the molecular level: thirteen individuals comprising a pair of monozygous twin siblings, 46% being females and 54% males. Since patients originated from all regions of Portugal, this cohort is representative of the whole population. Diagnosis of patients, suspected by high lactate and pyruvate plasma levels and respective lactate/pyruvate ratio < 20, was con rmed by reduced PDC activity (ca <30 % of laboratory control mean: 1734 ± 455 pmol·min -1 ·mg protein -1 ; range: 1279-2189; n = 70) in peripheral lymphocytes and/or cultured broblasts originating from skin biopsy, and also by identi cation of the causative mutation(s).
This study was approved by the local Ethics Committees and informed consents were obtained from the patients or their parents, who were also enrolled in the study, whenever necessary and possible. Declaration of Helsinki was also strictly observed.

Phenotypic evaluation
The physicians following these patients completed a questionnaire involving a wide range of parameters, namely: general characteristics of the patients, clinical features, brain malformations, biochemical ndings, genetic ndings, and current therapy.

Sample Collection and Preparation
Blood samples were collected after overnight fasting by venipuncture into EDTA-containing tubes for plasma separation and quanti cation of lactate and pyruvate levels, and into heparin-containing tubes for peripheral blood mononuclear cells (PBMC) isolation.
PBMC were separated at room temperature on a Ficoll-Paque gradient. Fibroblasts were cultured in Dulbecco's Modi ed Eagle Medium supplemented with 10% newborn calf serum and 1% antibiotic/antimycotic solution. Pelleted cells were resuspended in homogenization buffer (80 mM KH 2 PO 4 , pH 7.4, 2 mM EDTA). PBMC suspensions were immediately disrupted by sonication, whereas broblast suspensions were treated with 5 mM dichloroacetate for 15 minutes at 37 °C. The reaction was blocked by addition of a stopping solution (25 mM NaF, 25 mM EDTA, 4 mM DTT), and cells were disrupted by three freeze/thaw cycles.

PDC Activity Assay
Enzymatic activity was measured using a radiochemical method based on the release of 14  Preparation of Genomic DNA, RNA and cDNA Genomic DNA and, eventually, total RNA were isolated from peripheral blood leukocytes using the Puregene Cell and Tissue kit (Gentra Systems) and the Trizol method, respectively; 5 g of total RNA were used for the reverse transcription reaction (Amersham First Strand cDNA Synthesis kit, GE Healthcare Bio-Science Corp.).

PCR ampli cation of PDC coding genes
The complete sequence of each gene was obtained by PCR ampli cation of individual exons, including intronic boundaries, or overlapping fragments of the respective cDNA. The PDC subunits under analysis together with their coding genes, approved symbols and reference sequences are listed in Table  S1, whereas primer sequences designed for each gene ampli cation are listed in Table S2.
Sequence Analysis PCR and RT-PCR products were puri ed from solution or directly from agarose gels, using Isolate II PCR and Gel Kit (Bioline). PCR forward or reverse primers were added to the puri ed products from each individual sample and submitted to bi-directional Sanger sequencing. All chromatograms corresponding to PCR and RT-PCR fragments were analyzed using the BLAST program (NCBI).

In silico analysis of PDC-E1 mutations
To better establish genotype-phenotype correlations regarding the missense mutations in PDHA1, we undertook an in silico analysis of protein variants resulting from the described mutations. Besides evaluating the potential pathogenicity of the mutations using the PolyPhen-2 server (17), we sought to obtain structural models of the protein variants through two complementary strategies, previously described for the p.R253G, p.R378C, p.R378H, and p.F205L variants (18), and herein extended to p.R302H: i) submitting the sequence of the amino acid substituted variant to the SwissModel server and retrieving the corresponding model; and ii) using the Mutagenesis tool in Pymol (version 1.7) (39) employing the structure of WT PDC-E1 (PDB entry 3EXE) as template (19).

Declarations
Data can be made available upon reasonable request to the corresponding authors.

COMPETING InterestS
The Authors declare they have no con ict of interest.    Figure 1 Magnetic resonance images of brain lesions of an 8 month old patient carrying a PDHA1 mutation. Hyperintense bilateral and symmetrical lesions in the thalami, globus pallidus and putamina, on T2 weighted images, hypointense on T1 and with diffusion restriction suggestive of acute lesions; there is no diffusion restriction suggestive of chronic lesions. a) axial T2 b) axial T1 IR c) axial DWI d) axial ADC.

Figure 2
Magnetic resonance spectroscopy of a patient carrying a PDHA1 mutation. Spectroscopy with TE = 135 ms in left thalamus; slight reduction of NAA, increased choline and lactate.

Figure 4
In silico analysis of pyruvate dehydrogenase complex E3 p.P87S variant. Cartoon representation of homodimeric PDC E3 crystallographic structure (PDB entry 6I4T; one monomer represented in blue, the other in grey). Flavin adenine dinucleotide (FAD) cofactor in yellow sticks. Right panel, zoom-in on the location of the P87S substitution. P87 is located in an α-helix which contains the active site cysteine residues C80 and C85. Substitution of P by S will likely affect the helix structure and disturb the proximity between the active site disulphide and the FAD cofactor. Structural model of PDC-E1 p.R302H variant was obtained by loading the structure of WT PDC-E3 (PDB entry 6IT4) into Pymol and applying the mutagenesis tool to generate all possible rotamers of the substituting amino acid side chain.

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