The polar oxy-metabolome reveals the 4-hydroxymandelate CoQ10 synthesis pathway

Abstract

Oxygen is critical for a multitude of metabolic processes that are essential for human life. Biological processes can be identified by treating cells with 18O2 or other isotopically labelled gases and systematically identifying biomolecules incorporating labeled atoms. Here we labelled cell lines of distinct tissue origins with 18O2 to identify the polar oxy-metabolome, defined as polar metabolites labelled with 18O under different physiological O2 tensions. The most highly 18O-labelled feature was 4-hydroxymandelate (4-HMA). We demonstrate that 4-HMA is produced by hydroxyphenylpyruvate dioxygenase-like (HPDL), a protein of previously unknown function in human cells. We identify 4-HMA as an intermediate involved in the biosynthesis of the coenzyme Q10 (CoQ10) headgroup in human cells. The connection of HPDL to CoQ10 biosynthesis provides crucial insights into the mechanisms underlying recently described neurological diseases related to HPDL deficiencies1-4 and cancers with HPDL overexpression5.

Conflict of interest statement

Competing Interests Statement

A.C.K. has financial interests in Vescor Therapeutics, LLC. A.C.K. is an inventor on patents pertaining to KRAS regulated metabolic pathways, redox control pathways in pancreatic cancer, targeting GOT1 as a therapeutic approach, and the autophagic control of iron metabolism. A.C.K is on the SAB of Rafael/Cornerstone Pharmaceuticals. A.C.K has been a consultant for Diciphera Pharmaceuticals. M.E.P. has options in Raze Therapeutics, is the recipient of travel funds from Thermo Fisher Scientific and consulted for aMoon Ventures. R.S.B., Q.S., and M.E.P. are co-inventors on a patent filing on aspects of CoQ10 metabolism. The other authors declare no competing interests.

© 2021. The Author(s), under exclusive licence to Springer Nature Limited.

Figures

Extended Data Figure 1.

A robust method for 18O2 labelling of human cells. a, Schematic of 18O2 labelling. A closed system chamber is flushed multiple times with N2 to remove 16O2. A gas mixture containing 18O2 and CO2 is pulsed into the closed chamber to reach the desired oxygen concentration. At the assay endpoint, the chamber is opened, cells are extracted, and metabolites separated and identified by liquid chromatography-tandem mass spectrometry (LC-MS/MS). b, Oxygen measurements after N2 flush, followed with or without pulses of O2:CO2 gas mixture in the closed chamber containing tissue culture plates and media (n=3 technical replicates each). c-e, Oxygen percentage of O2-labelling experiments performed at 3% (c), 1% (d), and 0.2% (e) 16O2 or 18O2. (n=2 technical replicates each). f, Cells were treated with several concentrations of MG132, DFO (an iron chelator), IOX1 (a dioxygenase inhibitor), or in combination at 5% O2 for 24 hours. Immunoblots of HIF1α with ERK2 as a loading control. g, Immunoprecipitation of HIF1α to determine its hydroxylation (P564-OH) levels by the indicated inhibitors. Immunoblots of HIF1α and HIF1α P564-OH are shown, with ERK2 serving as a loading control. Experiments were performed once for optimization of drug concentrations (f, g). Graphs represent mean ± s.d. (c-e).

Extended Data Figure 2.

18O2 labelling of human cells reveals the oxy-metabolome. a, Schematic of the approach used to identify 18O-labelled features and metabolites that were labelled by 18O. “n” represents the number of features or metabolites identified in MIAPACA2 cells grown at 3% 18O2. See b for details. b, Summary of total and percentage of identified 18O-labelled features for each cell line and oxygen tension as described in a. c, Venn diagram demonstrating the overlap of unique 18O-labelled metabolite and features identified for each oxygen condition per cell line. d, Total number of unique dioxygenase-dependent, 18O-labelled metabolites and features identified in each cell line and condition. Features were categorized into predicted or not predicted/unknown 18O-labelled metabolites, based on known oxygen-dependent metabolic pathways, and sensitivity or insensitivity to IOX1 (dioxygenase inhibitor) treatment. e, Number of 18O-labelled metabolites detected in cells grown in 3%, 1%, and 0.2% 18O2 for 24 hours in two sets of experiments. The overlap of the total number of detected 18O-labelled metabolites and features in both experimental sets are shown. f, Venn diagram representing the distribution of common and unique 18O-labelled metabolites identified in each cell line. g, List of the 46 unique 18O-labelled metabolites that were identified in f and categorized into known oxygen-dependent metabolic pathways. ** represents metabolites that have matching MS2 spectra, but need to be validated due to multiple metabolite isomers.

Extended Data Figure 3.

Fractional 18O labelling of metabolites and features identified in human cells by 18O2 labelling. a, Heatmap representing the median fractional 18O labelling of the 49 metabolites and features in the indicated cell lines and oxygen tensions. “**” represents metabolites that have matching MS2 spectra, but need to be validated due to multiple metabolite isomers. The red arrow indicates a highly labelled unknown metabolite. b, Correlation matrixes demonstrating the Spearman rs value based on the fractional 18O labelling of the 46 metabolites and features across the indicated cell lines and oxygen tensions. c, Fractional 18O labelling of unknown feature (167.0339 in negative ion mode, elution time of 8.2 minutes) in MIAPACA2, A498, and SKNDZ cells grown in 3%, 1%, and 0.2% 18O2, and treated with vehicle or IOX1 (dioxygenase inhibitor) for 24 hours (n=3). d-i, Fractional 18O labelling of metabolites and features by 18O2 across multiple cell lines in response to different oxygen tensions, treated with or without IOX1 (dioxygenase inhibitor) for 24 hours. 18O labelling of predicted (d-e), not predicted (f-h), and unknown (i) metabolites or features are shown for the indicated cell line. ** represents metabolites that have matching MS2 spectra, but need to be validated due to multiple metabolite isomers. n=3 biologically independent samples for each group and condition in all experiments. Graphs represent mean ± s.e.m. and were compared using one- (b-d) or two-way ANOVA (a, e-f), followed by Tukey post-hoc test (*p<0.05, ^p<0.01, %p<0.005, #p<0.0001).

Extended Data Figure 4.

Total levels of unlabelled and 18O-labelled metabolites identified in human cells. a, Schematic of the carnitine biosynthesis pathway. Dioxygenases, TMLH (Trimethyllysine hydroxylase) and BBOX (butyrobetaine, 2-oxoglutarate dioxygenase), are shown in orange boxes, and 18O labelling is indicated in blue with arrows. b-d, Total intracellular levels of unlabelled and 18O-labelled γ-butyrobetaine from cells grown in 3%, 1%, and 0.2% 18O2 with the indicated reagents for 24 hours (n=3). e, Schematic of methionine salvage pathway. ADI1 (Acireductone dioxygenase 1), a dioxygenase, is shown in orange, and 18O labelling is indicated in blue with arrows. f-h, Total intracellular levels of unlabelled and 18O-labelled methionine from cells grown in 3%, 1%, and 0.2% 18O2 with the indicated reagents for 24 hours (n=3). i, Schematic of methionine oxidation by 18O-labelled reactive oxygen species with arrows. j-m, Total intracellular levels of unlabelled and 18O-labelled methionine sulfoxide from cells grown in 3%, 1%, and 0.2% 18O2 with the indicated reagents for 24 hours (n=3). “n” represents the number of biologically independent experiments for each group and condition. Graphs (mean ± s.e.m.) were compared using two-tailed Student t-test (c-d, g-h, k-m) or one-way (b, f, j) ANOVA, followed by Tukey post-hoc test (*p<0.05, ^p<0.01, %p<0.005, #p<0.0001).

Extended Data Figure 5.

Identification of 18O-labelled 4-HMA in human cells. a, Tandem mass spectra (MS2) of homogentisate (HGA) standard, 4-hydroxymandelate (4-HMA) standard, and unlabelled (167.0344m/z) feature precursors, and the respective product fragments. Mass differences between the precursor and product ions reflect loss of one CO2. The red line indicates fragmentation of the precursor ion into the two product ions. The structure of the precursor and product ions are depicted on the left. b, MS2 of unlabelled (167.0344 m/z), +one 18O (169.0387 m/z), and +two 18O (171.0428 m/z) labelled 4-HMA precursors, and the respective product fragments. Mass differences between precursor and product ions, reflects loss of unlabelled and +one 18O-labelled CO2. The red line indicates precursor ion fragmentation into two product ions. The structure and position of 18O-labelled (blue and arrow) 4-HMA are depicted on the left. c-d, Total levels of unlabelled and 18O-labelled 4-HMA levels in A498 (c) and SKNDZ (d) cells grown in 3%, 1%, and 0.2% 18O2, and treated with or without IOX1 (dioxygenase inhibitor) for 24 hours. n=3 in biologically independent replicates for each group and condition. Graphs represent mean ± s.e.m. and were compared by two-way ANOVA (c-d), followed by Bonferroni post-hoc test (*p<0.05, ^p<0.01, %p<0.005, #p<0.0001).

Extended Data Figure 6.

4-HMA is a tyrosine-derived metabolite synthesized from tyrosine in human cells. a, Schematic of known and proposed pathways involved in 4-HMA biosynthesis found in the literature. A. orientalis biosynthesizes 4-HMA from 4-hydroxyphenylpyruvate (4-HPPA), via hydroxymandelate synthase (HmaS), an Fe-dioxygenase. 4-HMA also has been proposed to be made from tyramine in rabbits by radioactive tracing studies. However, the proposed pathway was never formally demonstrated, as indicated by the dotted lines and box. b, Fractional labelling of Phe, Tyr, 4-HPPA, 4-HPLA, and 4-HMA from cells grown at 3% O2 with or without 13C9-Tyr or 13C6-Phe for 24 hours (n=5 for each group). c, Total intracellular levels of unlabelled and 13C-labelled Tyr, 4-HPPA, 4-HPLA, and 4-HMA from cells grown at 3%, 1%, and 0.2% O2 with 13C-Tyr for 24 hours (n=5 for each group). d, Total intracellular levels of unlabelled and 13C-labelled Tyr, 4-HPPA, 4-HPLA, and 4-HMA from cells grown in 13C-Tyr at 3% O2 with the indicated reagents for 24 hours (n=5 for each group). “n” represents the number of biologically independent replicates for each group and condition. Graphs represent mean ± s.e.m. and were compared by two-way ANOVA (b-d), followed by Tukey post-hoc test (*p<0.05, ^p<0.01, %p<0.005, #p<0.0001).

Extended Data Figure 7.

Human HPDL is an ortholog of A. orientalis HmaS. a-b, Phylogenetic tree of HPD, HPDL, and HmaS cDNA (a) and protein (b) sequences across several model organisms. c, Protein sequence alignment of HPD, HPDL and HmaS. Catalytic histidines involved in coordinating the iron ion needed for activity are highlighted in red. Specific residues in Steptomyces avermitilis and Pseudomonas fluorescens HPD have been mutated in other studies, and the human equivalent mutations are as indicated; hydrophobic (blue), polar (green) amino acids and proline (yellow). The HPD P239T mutant decreases HGA production and generates oxopinone. The N241I/L mutation abolishes HGA production by HPD. The HPD S226A mutations blocks HGA production. However, the mutation in the equivalent site in HMS (S201A) does not affect the generation of 4-HMA. The F337V/L mutation in HPD decreases HGA synthesis and allows slight production of 4-HMA. d, Growth curve of MIAPACA2 cells with sgRNAs at 21% O2. (n=3 technical replicates for each cell line, performed at least twice). e, Unlabelled and 13C8-labelled 4-HMA from PATU-8902 cells grown in 13C9-Tyr at 21% 16O2 for 24 hours. (n=3). Immunoblots of HPDL levels from PATU-8902 cells expressing control and HPDL sgRNAs. ERK2 serves as a loading control. f, Unlabelled and 13C-labelled Tyr, 4-HPPA, 4-HPLA from MIAPACA2 cells were grown in 13C9-Tyr at 21% 16O2 for 24 hours (from Figure 2a). (n=5). g, Unlabelled and 13C-labelled Tyr, 4-HPPA, 4-HPLA from MIAPACA2 sgHPDL #3 cells were grown in 13C9-Tyr at 21% 16O2 for 24 hours (from Figure 2b). (n=5). “n” represents the number of biologically independent replicates for each group and condition, unless indicated (d). Graphs are represented as mean ± s.d. (d) or s.e.m. (e-g) and were compared by two-tailed Student t-test (e), or two-way ANOVA (d, f-g), followed by Tukey post-hoc test (*p<0.05, ^p<0.01, %p<0.005, #p<0.0001).

Extended Data Figure 8.

Expression of HPDL affects 13C9-Tyr labelling of CoQ10. a, Schematic of known and unknown components of the CoQ10 biosynthesis pathway in humans. R reflects the polyprenyl tail that is attached to 4-HB. b, Fractional labelling and total levels of CoQ10 from MIAPACA2 grown in unlabelled, 13C9-Tyr-, and 13C6-Phe-labelled media for 24 hours at 3% 16O2. (n=5). c, Fractional labelling and total levels of CoQ10 from MIAPACA2 cells grown in 13C9-Tyr at 21%, 3% and 0.1% 16O2 for 24 hours (n=5). d, Fractional labelling and total levels of CoQ10 from MIAPACA2 cells grown in 13C9-Tyr containing media, with the indicated compounds at 3% 16O2 for 24 hours (n=5). e, Fractional labelling and total intracellular levels of unlabelled and 13C6-labelled CoQ10 from the indicated MIAPACA2 cells. Cells were grown in 13C9-Tyr at 21% 16O2 for 24 hours (n=5). f, Fractional labelling of CoQ10 from indicated MIAPACA2 cells. Cells were grown in 13C9-Tyr at 21% 16O2 for 24 hours. (n=5). g, Unlabelled and 13C6-labelled CoQ10 levels from PATU-8902 cells grown in 13C9-Tyr media (n=3). h, Extracellular concentrations of 13C8-labelled 4-HMA released from MIAPACA2 cells expressing control and HPDL sgRNAs. Cells were grown in 13C9-Tyr at 21% 16O2 for 24 hours at low (LD) and high (HD) cell densities (n=3). Representative images of LD and HD cells are shown. i, The effect of cell density on total intracellular levels of unlabelled and 13C6-labelled CoQ10 from the indicated MIAPACA2 cells. Cells were grown in 13C9-Tyr at 21% 16O2 for 24 hours at LD and HD (n=3 for each group). “n” represents the number of biologically independent experiments for each group and condition. Graphs represent mean ± s.e.m. were compared using two-tailed Student t-test (g), one- (d-f, h-i) or two- (b-c) way ANOVA, followed by Tukey post-hoc test (*p<0.05, ^p<0.01, %p<0.005, #p<0.0001).

Extended Data Figure 9.

CoQ10 synthesis is important for growth in 3D, but not 2D, conditions. a, Schematic of pulse-chase study using tyrosine-derived intermediates shown in c. Cells were labelled with 13C9-Tyr for two weeks before being grown in 12C9-Tyr or 13C9-Tyr with or without unlabelled 4-HPPA, 4-HPLA, 4-HMA, and 4-HB for 24 hours at 21% O2. b, Growth curve of MIAPACA2 cells with the indicated intermediates and times at 21% O2. (n=3 technical replicates for each cell line, performed at least twice). c, Total levels and fractional labelling of unlabelled and 13C-labelled metabolites in the CoQ10 headgroup biosynthesis pathway in humans, as described in a (n=4). Endogenous 4-HB is below the limit of detection. d, Schematic of known and potential enzymes and intermediates in the CoQ10 headgroup biosynthesis pathway in humans and yeast. Dotted lines reflect potential pathways and enzymes. e, Immunoblot of the indicated MIAPACA2 cells. ERK2 is the loading control. Experiment was performed twice to check for knockout efficiency. f, Growth in 2D culture of MIAPACA2 cells (n=4). g-h, Total levels (g) and fractional labelling (h) of CoQ10 in MIAPACA2 cells. Cells were grown in 13C9-Tyr at 21% 16O2 for 24 hours (n=5). i, Growth in 3D culture of MIAPACA2 cells (n=4) after three days. j, Representative confocal fluorescent images of the indicated MIAPACA2 cells. Images are representative of three independent experiments. k, Fractional labelling of intracellular and mitochondrial CoQ10 from 13C9-Tyr in the indicated MIAPACA2 cells (n=4). l, Immunoblot of total, cytosolic, and mitochondrial fractions from k. Subcellular fractionation was performed twice to determine localization of HPDL. “n” represents the number of biologically independent experiments for each group and condition, unless indicated (b). Graphs (mean ± s.e.m.) were compared by two-way ANOVA (b-c,f-I,k), followed by Tukey (b-c,f-h,k) or Dunnett’s (i) post-hoc test (*p<0.05, ^p<0.01, %p<0.005, #p<0.0001).

Extended Data Figure 10.

HPDL is important for a subset of PDAC tumours. a, Orthotopic pancreatic tumour weight from the indicated MIAPACA2 cells. b-c, Tumour images (b) and weights (c) from second experiment set of orthotopic pancreatic tumour xenografts from MIAPACA2 cells expressing control or HPDL sgRNA with coHPDL WT or catalytically inactive mutant after 6 weeks post-injection. First experiment set can be found on a. d-g, Tumour images (d, f) and weight (e, g) of orthotopic (d-e) or subcutaneous (f-g) pancreatic tumour xenograft of PATU-8902 cells expressing control or HPDL sgRNA after 5 weeks post-injection. h-i, Representative (h) and quantification (i) of H&E and immunohistochemistry for cleaved caspase 3 (CC3), phospho-histone H3 (p-HH3), and the death to proliferation ratio (CC3:p-HH3) from MIAPACA2 tumours from a. j, Overall and progression-free survival of HPDL high (n=44) and low (n=96) expressing PDAC tumours from the TCGA dataset. “n,” and each point represents the number of biologically independent experiments for each group and condition. Survival curve (j) was compared using the two-sided Log-rank (Mantel-Cox) test. Graphs (median ± max/min (a, c, e, g, i)) were compared by two-tailed Mann Whitney test (e, g), one-way ANOVA (a, c, i), followed by Holm-Sidak post-hoc test (*p<0.05, ^p<0.01, %p<0.005, #p<0.0001).

Extended Data Figure 11.

HPDL-dependent CoQ10 biosynthesis pathway. The canonical tyrosine catabolism, HPDL-dependent (red), and HPDL-independent (purple) CoQ10 biosynthetic pathways are shown as indicated. The HPDL-independent pathway was proposed from earlier studies in rats. Dotted lines represent unknown pathway or transport steps. Potential intermediates and enzymes are proposed within the 4-HMA, HPDL-dependent and-independent pathways.

Extended Data Figure 12.

Applications of gaseous labelling. Our system for gaseous labelling can label cells with a wide range of isotopically labelled gases to study their incorporation into metabolites, lipids, nucleotides, proteins, and any other components of cells to understand the mechanisms of the biological effects of these gases.

Figure 1.

Analysis of the oxy-metabolome identifies a highly labelled metabolite, 4-hydroxymandelate, in human cells. a, Heatmap representing the median fractional 18O labelling of the 46 metabolites and features across the indicated oxygen tensions in MIAPACA2 cells. “**” represents metabolites that have matching MS2 spectra, but need to be validated due to multiple metabolite isomers. The red arrow indicates a highly labelled unknown metabolite. b, Mass spectra (MS1) of unlabelled (−167.0344m/z) and 18O2-labelled features (−169.0387m/z and −171.0426m/z) from MIAPACA2 cells grown in 3% 16O2 or 18O2 for 24 hours. The mass shifts of one (Δ2.0043) and two (Δ4.0082) 18O atoms are shown. c, Liquid chromatography-mass spectrometry traces showing retention times of metabolite precursors of homogentisate (HGA) and 4-hydroxymandelate (4-HMA) standards, and a highly 18O2-labelled feature (−167.0344m/z) from MIAPACA2 cells grown in 3% 18O2 for 24 hours. d, Unlabelled and 18O-labelled 4-HMA levels in MIAPACA2 cells grown in 3%, 1%, and 0.2% 18O2, and treated with MG132 (proteasome inhibitor), IOX1 (dioxygenase inhibitor), DFO (iron chelator), and/or physiological levels of ascorbate for 24 hours. (n=3). e, Schematic of 18O2-labelling of 4-HMA by an unknown Fe-dioxygenase and substrate in human cells. “n” represents the number of biologically independent replicates for each group and condition. Graphs represent mean ± s.e.m. and were compared by two-way ANOVA (d), followed by Tukey post-hoc test (*p<0.05, ^p<0.01, %p<0.005, #p<0.0001).

Figure 2.

4-HMA is derived from tyrosine and synthesized by HPDL (4-hydroxyphenylpyruvate dioxygenase-like) in human cells. a, Total levels of unlabelled and 13C8-labelled 4-HMA from MIAPACA2 cells expressing control (sgTomato; sgTOM and sgLuciferase; sgLUC), and HPDL sgRNA guides. Cells were grown in 13C9-Tyr at 21% 16O2 for 24 hours (n=5). b, Total levels of unlabelled and 13C8-labelled 4-HMA from MIAPACA2 sgHPDL #3 cells with or without expression of sgRNA-resistant HPDL-FLAG wild-type (WT) and catalytically inactive HPDL mutants (H258A, and H163/258A). Cells were grown in 13C9-Tyr at 21% 16O2 for 24 hours (n=5). S6K serves as a loading control (a-b). c, Total levels of 4-HMA generated from enzymatic assays using HPDL-FLAG immunoprecipitated from MIAPACA2 sgHPDL #3 cells with or without expression of sgRNA-resistant HPDL-FLAG WT. d, Total levels of 4-HMA generated from enzymatic assays using HPDL-FLAG immunoprecipitated from MIAPACA2 sgHPDL #3 cells with or without expression of sgRNA-resistant HPDL-FLAG WT, or catalytically inactive mutants (H258A, and H163/258A). e, Schematic of the canonical tyrosine catabolism pathway and proposed non-canonical tyrosine pathway. “n” represents the number of biologically independent replicates for each group and condition. Graphs are represented as mean ± s.e.m. and were compared by one- (d) or two-way ANOVA (a-c), followed by Tukey post-hoc test (#p<0.0001).

Figure 3.

HPDL and 4-HMA participate in the human CoQ10 headgroup biosynthesis pathway. a, Unlabelled and 13C6-labelled CoQ10 levels from MIAPACA2 cells grown in 13C9-Tyr media at low (LD) and high density (HD) (n=3). b, 3D growth assays in the indicated MIAPACA2 cells (n=4). c, Fractional labelling of CoQ10 from 13C9-Tyr in the indicated MIAPACA2 cells grown in 3D conditions (n=4). d, Oxygen consumption rates (OCR) from the indicated MIAPACA2 cells grown in 3D conditions (n=5). e-f, 3D growth assays (e) or OCR (f) in the indicated MIAPACA2 cells treated with or without 1mM 4-HMA or 4-HB (n=4). g, Summary of HPDL and 4-HMA in the human CoQ10 headgroup biosynthesis pathway. More details can be found in Extended Data Figure 11. “n” represents the number of biologically independent experiments for each group and condition. Graphs (mean ± s.e.m) were compared using one- (b-d, f) or two (a, e, g) way ANOVA, followed by Tukey post-hoc test (*p<0.05, ^p<0.01, %p<0.005, #p<0.0001).