Failure to repair damaged NAD(P)H blocks de novo serine synthesis in human cells

Abstract

Background: Metabolism is error prone. For instance, the reduced forms of the central metabolic cofactors nicotinamide adenine dinucleotide (NADH) and nicotinamide adenine dinucleotide phosphate (NADPH), can be converted into redox-inactive products, NADHX and NADPHX, through enzymatically catalyzed or spontaneous hydration. The metabolite repair enzymes NAXD and NAXE convert these damaged compounds back to the functional NAD(P)H cofactors. Pathogenic loss-of-function variants in NAXE and NAXD lead to development of the neurometabolic disorders progressive, early-onset encephalopathy with brain edema and/or leukoencephalopathy (PEBEL)1 and PEBEL2, respectively.

Methods: To gain insights into the molecular disease mechanisms, we investigated the metabolic impact of NAXD deficiency in human cell models. Control and NAXD-deficient cells were cultivated under different conditions, followed by cell viability and mitochondrial function assays as well as metabolomic analyses without or with stable isotope labeling. Enzymatic assays with purified recombinant proteins were performed to confirm molecular mechanisms suggested by the cell culture experiments.

Results: HAP1 NAXD knockout (NAXDko) cells showed growth impairment specifically in a basal medium containing galactose instead of glucose. Surprisingly, the galactose-grown NAXDko cells displayed only subtle signs of mitochondrial impairment, whereas metabolomic analyses revealed a strong inhibition of the cytosolic, de novo serine synthesis pathway in those cells as well as in NAXD patient-derived fibroblasts. We identified inhibition of 3-phosphoglycerate dehydrogenase as the root cause for this metabolic perturbation. The NAD precursor nicotinamide riboside (NR) and inosine exerted beneficial effects on HAP1 cell viability under galactose stress, with more pronounced effects in NAXDko cells. Metabolomic profiling in supplemented cells indicated that NR and inosine act via different mechanisms that at least partially involve the serine synthesis pathway.

Conclusions: Taken together, our study identifies a metabolic vulnerability in NAXD-deficient cells that can be targeted by small molecules such as NR or inosine, opening perspectives in the search for mechanism-based therapeutic interventions in PEBEL disorders.

Keywords: 3-Phosphoglycerate dehydrogenase; Inborn errors of metabolism; Inosine; Metabolite damage and repair; NAD(P)H hydration; NAXD; Nicotinamide riboside; Serine biosynthesis.

Fig. 1

Metabolic profiling of a HAP1 cell model of NAXD deficiency in standard (Iscove’s modified Dulbecco’s medium, IMDM) medium. A Schematic presentation of the NAD(P)HX repair system. Hydration damage (enzymatic or nonenzymatic) of NAD(P)H generates S- or R-NAD(P)HX forms that can interconvert and further react to form cyclic forms designated collectively as cNAD(P)HX. The repair enzymes (NAXD and NAXE) are shown in blue. The NAXE epimerase accelerates interconversion between the S- and R-NAD(P)HX forms, whereas the NAXD dehydratase stereospecifically acts on S-NAD(P)HX in an ATP-dependent manner. B HAP1 NAXDko cells accumulate damaged forms of NADH. Control and NAXDko cells were grown in IMDM for 72 h or 96 h (Supplementary Fig. S1A), and NAD(H)(X) metabolites were measured by LC–MS. All values were normalized to an internal standard and cell counts from replicate plates, and are means ± SDs of three replicate wells. Statistical significance was calculated using an equal variance, unpaired Student’s t-test. **p < 0.01; ***p < 0.001; ns, not significant. C, D Metabolite extracts from HAP1 control and NAXDko cells grown for 72 h and 96 h in IMDM were analyzed using untargeted GC–MS (C) and LC–MS (D) methods (72 h data is shown here, whereas 96 h data is shown in Supplementary Fig. S1B, C). For both the GC–MS and LC–MS data, metabolites with significant changes between the control and NAXDko cells at both time points are presented. For the LC–MS data plotted in Fig.1D and Supplementary Fig. S1C, the presence of metabolites in significantly enriched metabolite sets was used as an additional selection filter. GC–MS data were first normalized using the internal standard [U-¹³C]-ribitol. LC–MS data were normalized using the summed MS1 intensities of all metabolites within each sample. For both GC–MS and LC–MS data, all values presented are relative to controls and are means ± SDs of six biological replicates. Pathway analysis for the differential metabolites detected by LC–MS is presented in Supplementary Fig. S2. The detailed list of metabolites is compiled in Additional file 1. NAXD, NAD(P)HX dehydratase; NAXE, NAD(P)HX epimerase; GAPDH, glyceraldehyde 3-phosphate dehydrogenase; Acetyl-CoA, acetyl coenzyme A; ADP, adenosine diphosphate

Fig. 2

NAXD-deficient HAP1 cells show more pronounced phenotypes under galactose culture conditions. A The HAP1 NAXDko cells show impaired growth specifically in galactose medium. The experimental design is shown in the inset. HAP1 control and NAXDko cells were seeded in IMDM and allowed to grow for 24 h, and then the medium was changed to a basal medium with either glucose (25 mM) or galactose (25 mM). B Control and NAXDko cells were grown in glucose or galactose as indicated in panel A, and S-NADHX, R-NADHX, and NADH levels were measured after 72 h by HPLC–UV (NADPH and NAD⁺ measurements are shown in Supplementary Fig. S3). The NADHX burden (calculated as the sum of S- and R-NADHX levels divided by the NADH level) was increased in NAXDko cells grown in galactose compared with the same cells grown in glucose. C NAXDko cells grown in galactose show increased mitochondrial superoxide levels. HAP1 control and NAXDko cells were grown in glucose or galactose as indicated in panel A, cells were stained with MitoSOX Red after 72 h, and images were captured at an excitation wavelength of 488 nm. Fluorescence intensity was measured as relative fluorescence units (RFUs) and normalized to control cells grown in glucose. All values are means ± SDs of 3–6 biological replicates, and statistical significance was calculated using an equal variance, unpaired Student’s t-test. *p < 0.05; **p < 0.01, ****p < 0.0001; ns, not significant. For all panels, Co, Control; Ko, NAXDko; Glc, glucose medium; Gal, galactose medium

Fig. 3

NAXD-deficient HAP1 cells grown in galactose medium show perturbations in the serine biosynthesis pathway. The experimental scheme is shown in the blue inset. Control and NAXDko cells were seeded in IMDM and shifted to galactose-containing medium for 72 h before metabolite extraction and analysis by ICMS and HILIC-MS methods. The metabolite levels (peak areas) were normalized to summed MS1 peak intensities and mapped manually on the indicated central metabolic pathways. The HILIC analysis was performed to determine serine and glycine levels (shown in the other inset). The nutrient sources are marked with blue font. All values are means ± SDs of three biological replicates. Statistical significance was calculated using an equal variance, unpaired Student’s t-test; nonsignificant changes (p-value > 0.05) are not annotated with a symbol. **p < 0.01, ***p < 0.001; ****p < 0.0001. 6-PG, 6-phosphogluconate; S7P, sedoheptulose 7-phosphate; PRPP, phosphoribosyl pyrophosphate; FGAR, phosphoribosyl-N-formylglycinamide; AICAR, 5-aminoimidazole-4-carboxamide ribonucleotide; AMP, adenosine monophosphate; GMP, guanosine monophosphate; IMP, inosine monophosphate; CMP, cytidine monophosphate; UMP, uridine monophosphate; UDP, uridine diphosphate; Acetyl-CoA, acetyl coenzyme A; Gal, galactose; ICMS, ion-exchange mass spectrometry; HILIC-MS, hydrophilic interaction liquid chromatography- mass spectrometry

Fig. 4

NAXD-deficient HAP1 cells grown in galactose medium show inhibition at the PHGDH step of serine biosynthesis. A Tracer analysis with ¹³C₆-glucose demonstrates inhibition of the serine pathway in living NAXDko cells. The experimental scheme is shown in the inset. Control and NAXDko cells grown in galactose were pulsed with ¹³C₆-glucose for 10 min or 30 min (Supplementary Fig. S9A) followed by metabolite extraction and LCMS (ICMS and HILIC-MS) analysis. Fractional distribution of the label is shown for 3-phosphoglycerate, 3-phosphoserine, and serine. The labeling data for glycine, 6-phosphogluconate, and lactate are presented in Supplementary Fig. S9B. All values are means ± SDs of three biological replicates. Statistical significance was calculated using an equal variance, unpaired Student’s t-test. *p < 0.05; **p < 0.01; ***p < 0.001. B The dehydrogenase activity of recombinant human PHGDH is inhibited by NADHX. The difference in reaction mechanism catalyzed by the yeast (Ser3 and Ser33) and human (hPHGDH) PHGDHs is shown on the left. The assay was performed with purified, recombinant PHGDH (purification gel is shown in Supplementary Fig. S10A) at a substrate (3-PG) concentration of 40 µM and in the absence and presence of the indicated concentrations of purified S-, R-, or c-NADHX forms. All values are means ± SDs of at least three independent measurements. Co, Control; Ko, NAXDKo; Gal, galactose; MS, mass spectrometry; 3-PG, 3-phosphoglycerate; PHP, 3-phoshohydroxypyruvate; 2-KG, 2-ketoglutarate; D-2HG, D-2-hydroxyglutarate; PHGDH, 3-phosphoglycerate dehydrogenase

Fig. 5

Cytosolic, but not mitochondrial, expression of NAXD in NAXD-deficient HAP1 cells relieves inhibition of serine biosynthesis. A Schematic presentation of the human NAXD gene showing the transcription start sites of the different NAXD isoforms. The CytoNAXD and MitoNAXD rescue lines were generated by lentiviral transduction of the corresponding constructs into NAXDko cells. Validation of the rescue lines by qPCR analysis is presented in Supplementary Fig. S11A. B All cell lines were grown in a standard IMDM medium and S-NADHX, R-NADHX, and NADH levels were measured by HPLC–UV after 72 h (NADPH and NAD⁺ levels are shown in Supplementary Fig. S11B). C Intracellular levels of serine pathway metabolites were measured (by LCMS) after 72 h in galactose medium as shown in the inset. Data were normalized using the summed MS1 intensities of all metabolites within each sample. The levels of other lower glycolytic intermediates are shown in Supplementary Fig. S11C. All values are means ± SDs of six (panel B) and three (panel C) biological replicates. Statistical significance was calculated using ordinary one-way ANOVA (*p < 0.05; **p < 0.01, ****p < 0.0001). Co, Control_eGFP; Cyto, CytoNAXD rescue line; Mito, MitoNAXD rescue line; Ko, NAXDko_eGFP; Gal, galactose; MS, mass spectrometry

Fig. 6

Inosine partially rescues growth of NAXD-deficient HAP1 cells and drastically reduces de novo serine synthesis. A Inosine supplementation shows dose-dependent, partial rescue in growth of NAXDko cells. HAP1 control and NAXDko cells were grown in galactose media without or with supplementation of inosine at the indicated concentrations. The viable cell count was measured after 72 h (the count after 96 h and replots of the time course data relative to the control cell viability are shown in Supplementary Fig. S12B). B Control and NAXDko cells were grown in galactose without or with 5 mM inosine supplementation and indicated metabolites were quantified after 72 h by HPLC–UV. (NADPH and NAD⁺ levels are shown in Supplementary Fig. S12C.) C Control and NAXDko cells were grown without or with 5 mM inosine supplementation and indicated metabolites were measured after 72 h using ICMS and HILIC-MS (the complete mapped data are presented in Supplementary Fig. S13). Data were normalized using the summed MS1 intensities of all metabolites within each sample. All values are means ± SDs of at least three biological replicates. Statistical significance was calculated using an equal variance, unpaired Student’s t-test (*p < 0.05; **p < 0.01, ***p < 0.001; ****p < 0.0001; ns, not significant). Co, Control; Ko, NAXDKo; Gal, galactose; Ino, inosine; MS, mass spectrometry

Fig. 7

Nicotinamide riboside supplementation partially rescues growth and metabolic defects in NAXD-deficient HAP1 cells. A Control and NAXDko cells were grown in galactose without or with supplementation of NAD precursors (2 mM each), and viable cell count was measured using Acridine Orange and 4′,6-diamidin-2-phenylindole dyes after 72 h (this panel) or 96 h (Supplementary Fig. S14A; replots of the time course data relative to the control cell viability). B Control and NAXDko cells were grown without or with the indicated supplements (5 mM each), and viable cell count was measured after 72 h (replot of viable cell count in supplemented versus control conditions is shown in Supplementary Fig. S14D). C Control and NAXDko cells were grown in galactose without or with indicated supplements (5 mM each), and indicated metabolites were measured after 72 h by ICMS and HILIC-MS. Data were normalized using the summed MS1 intensities of all metabolites within each sample. All values are means ± SDs of 3–6 biological replicates. Statistical significance was calculated using an equal variance, unpaired Student’s t-test (A, B) or ordinary one-way ANOVA (C) (*p < 0.05; **p < 0.01, ***p < 0.001; ****p < 0.0001; ns, not significant). Co, Control; Ko, NAXDKo; Gal, galactose; NR, nicotinamide riboside; NA, nicotinic acid; NAM, nicotinamide mononucleotide; Ino, inosine; MS, mass spectrometry

Fig. 8

NAXD patient-derived fibroblasts and NAXDko whole brain organoids grown in galactose show impaired de novo serine synthesis. A Pathogenic mutations observed in patients are mapped on the NAXD gene. Case numbers are retained from the ones reported in Ref. [11], and further details on these cases are summarized in Table 1. Exons are shown as blue numbered boxes. B, C Control and patient fibroblasts were grown in galactose for 72 h, and NADHX (B) as well as serine synthesis pathway metabolites (C) were analyzed using ICMS or HILIC-MS methods. All values are means ± SDs of three independent replicates. Statistical significance was calculated using ordinary one-way ANOVA (*p < 0.05; **p < 0.01, ***p < 0.001; ****p < 0.0001). D Schematic representing WBO derivation from control and NAXDko human iPSCs. Control and NAXDko WBOs were grown in a rich medium for 75 days and shifted to a galactose containing medium for 72 h before analysis of NADHX (E) and serine synthesis pathway metabolites (F) by ICMS and HILIC-MS methods. Data were normalized using the summed MS1 intensities of all metabolites within each sample. For fibroblasts, values are means ± SDs of three biological replicates and statistical significance was calculated using ordinary one-way ANOVA. For WBOs, values are means ± SDs obtained from 8 samples (extracts), where each sample was obtained by pooling 2 organoids (the 16 organoids used in total per genotype were generated in 4 independent derivations). Statistical significance was calculated using an equal variance, unpaired Student’s t-test (*p < 0.05; **p < 0.01, ***p < 0.001; ****p < 0.0001). spNAXD, signal peptide NAXD; Co, Control; Ko, NAXDKo; WBO, whole brain organoid

Fig. 9

A working model summarizing key conclusions from this study. Upper panel: NAXD-deficient yeast and human cells display inhibition of de novo serine synthesis. In human cells, this effect is revealed only under certain conditions, such as cultivation in the presence of galactose instead of glucose. It is possible that this blockage of the serine synthesis pathway contributes to downstream perturbations in lipid metabolism, which were suggested on the basis of plasma metabolomic analyses in a NAXD patient [18]. Lower panel: NR and inosine partially rescue NAXD deficiency phenotypes via distinct mechanisms. Our results suggest that inosine exerts a rescue effect in HAP1 NAXDko cells by reducing the cells' reliance on the serine synthesis pathway for downstream metabolic processes, whereas NR supplementation presumably leads to positive effects via relieving PHGDH inhibition through reduced NADHX burden and/or increased NAD⁺/NADH ratio