Advertisement
Research ArticleArticle

Aldehyde Oxidase Contributes to All-Trans-Retinoic Acid Biosynthesis in Human Liver

Guo Zhong, Chris J. Seaman, Erickson M. Paragas, Huaqing Xi, Karla-Luise Herpoldt, Neil P. King, Jeffrey P. Jones and Nina Isoherranen
Drug Metabolism and Disposition March 2021, 49 (3) 202-211; DOI: https://doi.org/10.1124/dmd.120.000296

Abstract

All-trans-retinoic acid (atRA) is a critical endogenous signaling molecule. atRA is predominantly synthesized from retinaldehyde by aldehyde dehydrogenase 1A1 (ALDH1A1), but aldehyde oxidase (AOX) may also contribute to atRA biosynthesis. The goal of this study was to test the hypothesis that AOX contributes significantly to atRA formation in human liver. Human recombinant AOX formed atRA from retinaldehyde (Km ∼1.5 ± 0.4 µM; kcat ∼3.6 ± 2.0 minute−1). In human liver S9 fractions (HLS9), atRA formation was observed in the absence of NAD+, suggesting AOX contribution to atRA formation. In the presence of NAD+, Eadie-Hofstee plots of atRA formation in HLS9 indicated that two enzymes contributed to atRA formation. The two enzymes were identified as AOX and ALDH1A1 based on inhibition of atRA formation by AOX inhibitor hydralazine (20%–50% inhibition) and ALDH1A1 inhibitor WIN18,446 (50%–80%inhibition). The expression of AOX in HLS9 was 9.4–24 pmol mg−1 S9 protein, whereas ALDH1A1 expression was 156–285 pmol mg−1 S9 protein measured by liquid chromatography–tandem mass spectrometry (LC-MS/MS) quantification of signature peptides. The formation velocity of atRA in the presence of NAD+ correlated significantly with the expression of ALDH1A1 and AOX protein. Taken together, the data show that both AOX and ALDH1A1 contribute to atRA biosynthesis in the human liver, with ALDH1A1 being the high-affinity, low-capacity enzyme and AOX being the low-affinity, high-capacity enzyme. The results suggest that in the case of ALDH1A dysfunction or excess vitamin A, AOX may play an important role in regulating hepatic vitamin A homeostasis and that inhibition of AOX may alter atRA biosynthesis and signaling.

SIGNIFICANCE STATEMENT This study provides direct evidence to show that human AOX converts retinaldehyde to atRA and contributes to hepatic atRA biosynthesis. The finding that AOX may be responsible for 20%–50% of overall hepatic atRA formation suggests that alterations in AOX activity via drug-drug interactions, genetic polymorphisms, or disease states may impact hepatic atRA concentrations and signaling and alter vitamin A homeostasis.

Introduction

Vitamin A (retinol) is an essential endogenous compound in all chordates. Retinol is first oxidized by retinol dehydrogenases to yield all-trans-retinaldehyde, which is further metabolized to all-trans-retinoic acid (atRA) (Napoli, 2012; Kedishvili, 2013). The active retinoid atRA binds to retinoic acid receptors and regulates various physiological processes, such as cell differentiation, T-cell activation, spermatogenesis, and lipid and glucose metabolism (Kedishvili, 2013). Liver, the major vitamin A storage organ, is an important site of retinoid metabolism. Aldehyde dehydrogenase 1A1 (ALDH1A1), one of the three members in the ALDH1A family in humans, is believed to be the major enzyme responsible for atRA biosynthesis in the liver, and the enzyme activity is NAD+-dependent (Arnold et al., 2015a, 2016). However, neither chemical inhibition of Aldh1a1 by WIN18,446 (Arnold et al., 2015a) nor global knockout of Aldh1a1−/− (Fan et al., 2003) in mice completely abolishes liver atRA biosynthesis. WIN18,446 inhibited 44% of atRA formation in mouse liver (Arnold et al., 2015a), and Aldh1a1−/− mice are healthy and fertile at adulthood, with similar hepatic atRA concentrations as wild-type mice (Fan et al., 2003). Taken together, these findings suggest that other enzyme(s) contribute to hepatic atRA formation and maintenance of vitamin A homeostasis.

It has been proposed that aldehyde oxidase (AOX) contributes to atRA biosynthesis in select organs, including the liver. AOX is a cytosolic enzyme expressed in most animals and encoded by a single gene in humans. AOX generally metabolizes aldehydes and N-containing heterocyclic compounds (Pryde et al., 2010; Rashidi and Soltani, 2017), but unlike ALDH1As, catalysis by AOX does not require the cofactor NAD+. AOX enzymes in rabbit and mouse liver have been reported to catalyze retinaldehyde oxidation to yield atRA (Tomita et al., 1993; Huang et al., 1999). The data showing that retinaldehyde concentrations increased in the harderian gland of Aox4−/− mice and atRA concentrations decreased in the harderian gland and skin of Aox4−/− mice in comparison with wild-type mice (Terao et al., 2009) suggest that AOX may play an important role in regulating atRA biosynthesis. The AOX family of enzymes has also been reported to be involved in testicular atRA biosynthesis in mice (Beedle et al., 2019), and hydralazine, an AOX inhibitor, decreased atRA formation in mouse liver S10 fractions by 45% (Arnold et al., 2015a). Furthermore, both NAD+-dependent and -independent retinaldehyde metabolism was observed in soluble extracts of human liver and kidney (Ambroziak et al., 1999), strongly suggesting that AOX is involved in atRA biosynthesis. However, AOX is not typically included when atRA formation is considered. This is largely due to the lack of Aox knockout mouse models. The only available Aox4−/− mouse model did not show any major morphologic abnormalities or lethal phenotype that would be expected if atRA biosynthesis was significantly impaired (Terao et al., 2009, 2020). In addition, at present no quantitative characterization of AOX contribution to atRA biosynthesis in specific tissues has been published. In this study, we hypothesized that aside from ALDH1A1, AOX makes a significant contribution to atRA formation in human liver. We aimed to quantitatively characterize atRA formation mediated by AOX in human liver and delineate the relative contribution of ALDH1A1 and AOX in hepatic atRA formation. This characterization is important for addressing altered atRA synthesis in human liver in various disease states and due to drug-drug interactions.

Materials and Methods

Chemicals, Reagents, Tissues, and Recombinant Enzymes.

Retinaldehyde, atRA, atRA-d5, NAD+, NADPH, raloxifene, and hydralazine were purchased from Millipore Sigma (Burlington, MA). WIN18,446 was from Acros Organics (Geel, Belgium). All LC-MS/MS-grade solvents were purchased from Thermo Fisher Scientific (Waltham, MA). Pooled human liver S9 fractions (lot number 3212595, 27 donors with 12 male and 15 female) were purchased from Corning (Corning, NY). Human liver samples from three individual donors—HL136 (39 years old, Caucasian male), HL146 (10 years old Caucasian male), and HL149 (63 years old, Caucasian female)—were obtained from the tissue bank of University of Washington School of Pharmacy.

Expression and Purification of Recombinant Human ALDH1A1, AOX, and Cellular Retinol Binding Protein 1 and Preparation of Human Liver S9 Fractions.

Recombinant human AOX was expressed, purified, and quantified as published previously (Barr et al., 2013; Paragas et al., 2017). Human recombinant ALDH1A1 was expressed in Escherichia coli as previously described (Arnold et al., 2015b) with several modifications. Briefly, frozen cell pellets were thawed on ice, resuspended with 25 ml lysis buffer (1 mg/ml of lysozyme, 500 mM NaCl, 20 mM Tris, 5 mM imidazole, pH 7.4), and lysed at room temperature for 15 minutes with gentle shaking. The resuspended mixture was centrifuged at 18,000 g for 20 minutes. The supernatant was loaded into a HisTrap HP column (GE Healthcare Bio-Sciences, Marlborough, MA) using Biologic DuoFlow chromatography system (Bio-Rad Laboratories, Hercules, CA). The column was washed with 5× column volumes of wash buffer (20 mM Tris-HCl, 500 mM NaCl, 20 mM imidazole, pH 7.4), and ALDH1A1 was eluted with five column volumes of elution buffer (20 mM Tris-HCl, 500 mM NaCl, 300 mM imidazole, pH7.4). Fractions containing ALDH1A1 were concentrated and exchanged into HEDK buffer (10 mM Hepes-NaOH, 0.1 mM EDTA, 0.5 mM DTT, 100 mM KCl, pH 8.0) using a HiLoad 16/60 Superdex200 column (GE Healthcare Bio-Sciences) following the manufacturer’s instructions. After the protein concentration was determined by a bicinchoninic acid protein assay (Thermo Fisher Scientific), purified ALDH1A1 was stored at −20°C in HEDK buffer with 2 mM tris(2-carboxyethyl)phosphine and 50% glycerol.

The gene encoding CRBP1 was ordered from Integrated DNA Technologies (Coralville, IA) as a gBlock and cloned into pET28b (Novagen, Madison, WI) using Gibson assembly between the NdeI and XhoI endonuclease restriction sites. Cloning at this site incorporates an N terminal, thrombin cleavable, His-tag. The DNA and amino acid sequences are listed in Supplemental Table 1. The plasmid encoding CRBP1 was then transformed into the T7 Expression strain of E. coli (New England Biolabs, Ipswich, ME) for protein expression. Cells were grown in Luria-Bertani (LB) medium supplemented with 50 mg/l of kanamycin at 37°C until an opptical density 600 of 0.6 was reached. Protein expression was induced by addition of isopropyl-thio-β-d-galactopyranoside to a final concentration of 1 mM. After induction, cells were allowed to grow for 16 hours at 18°C before harvesting by centrifugation. Cells pellets were lysed by sonication (2.5 minutes in 2-second pulses) in 50 mM Tris (pH 8), 150 mM NaCl, 20 mM imidazole, 0.05 mg/ml DNase, 0.05 mg/ml RNase, and 1 mM phenylmethylsulfonyl fluoride. Lysate was centrifuged at 25,000 g for 25 minutes. Lysate supernatants were applied to HisTrap FF columns (GE Healthcare, Uppsala, Sweden) for purification by immobilized metal affinity chromatography on an AKTA pure fast protein liquid chromatography system (GE Healthcare). The protein of interest was eluted over a linear gradient of 20–500 mM imidazole in a background of 50 mM Tris (pH 8) and 150 mM NaCl after washing with ∼10 column volumes wash buffer (elution buffer with 20 mM imidazole). Peak fractions were concentrated in 3 kDa molecular weight cutoff centrifugal filters, sterile filtered (0.22 μm), and dialyzed overnight to remove imidazole into a buffer containing 25 mM Tris (pH 8) and 150 mM NaCl. After dialysis, thrombin was added at a dilution of 1:2000 (v/v) and incubated overnight at 4°C. To remove cleaved His-tag fragments and any un-cleaved protein, the solution was applied to an equilibrated HisTrap FF column, and the flow-through was collected. The collected protein was again concentrated using a 3 kDa molecular weight cutoff centrifugal filter and applied to a Superdex 200 Increase 10/300 GL size exclusion chromatography column (GE Healthcare) using 25 mM Tris (pH 8) and 150 mM NaCl. Purified CRBP1 was stored at −20°C with 20% glycerol.

To prepare HLS9 fractions, liver samples (140–170 mg) from three donors were homogenized on ice with homogenizing buffer [50 mM potassium phosphate (KPi) containing 250 mM sucrose and 50 mM KCl, pH 7.4] in a 5:1 ratio of buffer to liver weight. After centrifugation at 9000 g at 4°C for 30 minutes, the supernatant (HLS9) was collected, aliquoted, and stored at −80°C until use. The protein concentration in each human liver S9 sample was measured using a bicinchoninic acid protein assay (Thermo Fisher Scientific).

General Protocol of Enzyme Incubations and atRA Quantification Using LC-MS/MS.

Incubations were performed according to the following protocol unless otherwise stated: HLS9 or purified recombinant enzymes were first preincubated at 37°C for 3 minutes in 100 mM KPi buffer (pH7.4) in a final volume of 100 µl. The reactions were initiated by adding retinaldehyde. The incubation times were as described for each experiment in the following sections. When possible, a 30-second incubation time was used to avoid substrate depletion and time-dependent loss of enzyme activity. In incubations that had low product formation rates as a result of chemical inhibition or low enzyme expression, longer incubation periods (up to 3 minutes) were employed to allow sufficient sensitivity and detection of atRA formation while maintaining incubation time within the linear range. All the incubation times were sufficiently short to avoid time-dependent loss of or decrease in enzyme activity for ALDH1A1 or AOX. Protein and time linearity was confirmed for all systems employed. After the predefined incubation time, the reactions were terminated by adding an equal volume of ice-cold acetonitrile (ACN) with 20 nM atRA-d5 [internal standard (IS)]. The samples were centrifuged at 15,000 g at 4°C for 10 minutes, and the supernatants were used for LC-MS/MS analysis. This method is referred to as “ACN precipitation method” in the following sections. For incubations with low product formation, incubation volume was increased to 500 µl. After adding an equal volume of ice-cold ACN with 4 nM atRA-d5 to stop the reaction, samples were processed using a hexane extraction method (Arnold et al., 2015b). In this method, each sample was extracted with 8 ml of hexanes and centrifuged at 1000 g for 5 minutes, and the organic layer was transferred into a glass tube and dried under N2 flow. The residue was resuspended with 100 µl of 80% ACN for LC-MS/MS analysis. Incubation time and concentrations of protein, substrate, inhibitor, and cofactor used for incubations are specified for each experiment in the following sections. All incubations were performed under yellow light to prevent retinoid degradation and isomerization. All cofactor solutions were freshly prepared before incubations. As minor nonenzymatic formation of atRA was observed in incubations with retinaldehyde, incubations with no enzyme were included in each experiment and used for background subtraction. Unless described otherwise, all incubations were performed as duplicates or triplicates and repeated three times on separate days.

The formation of atRA in incubations was measured by LC-MS/MS with AB Sciex 5500 quadrupole linear ion trap mass spectrometer (AB Sciex LLC, Framingham, MA) using positive ion athmospheric pressure chemical ionization probe and an Agilent 1290 Infinity ultra high performace liquid chromatography (Agilent Technologies, Santa Clara, CA) with a Kinetex C18 column (1.7 µm, 100 Å, 2.1 × 100 mm; Phenomenex, Torrance, CA) for chromatographic separation. The mobile phase consisted of H2O with 0.1% formic acid (solvent A) and ACN with 0.1% formic acid (solvent B). The flow rate was 0.45 ml/min, and the column temperature was set at 40°C. The gradient started at 50% B, increased to 95% B within 4.0 minutes, and then was kept at 95% B for 1 minute before returning to initial conditions. The injection volume was 20 µl per sample. atRA was monitored with multiple-reaction monitoring transitions of mass-to-charge ratio (m/z) 301 >205 and 301 >123, and atRA-d5 was monitored with m/z 306 >208. All MS/MS parameters were set as previously described (Arnold et al., 2015b). Peaks of atRA and IS were integrated using Analyst 1.6.3. All atRA quantification methods followed best practices as previously reported for quantification of atRA (Czuba et al., 2020). The peak area ratio of atRA to IS was used to quantify atRA formation. For quantification, standard curves of atRA (five concentrations ranging from 2 to 37.5 nM) were constructed by spiking atRA into KPi buffer, and standard curve samples were processed in parallel with incubation samples. The ratio of atRA to atRA-d5 (IS) peak area was plotted against atRA concentration (r2 values > 0.98 for all curves) and was used to calculate the concentration of atRA formed in incubations (Arnold et al., 2015b). At least one representative experiment in replicates was analyzed by two different people in the laboratory to validate the analyses.

atRA Formation by Recombinant Human AOX, ALDH1A1, and HLS9.

To measure enzyme kinetic parameters of atRA formation from retinaldehyde by AOX, 10 nM recombinant AOX was incubated for 30 seconds with 11 concentrations of retinaldehyde ranging from 100 nM to 7.5 µM. Samples were processed with the ACN precipitation method. Enzyme kinetic parameters, including kcat and Michaelis-Menten constant (Km), were obtained using GraphPad Prism 5.0 (GraphPad Software, La Jolla, CA) by fitting the Michaelis-Menten equation to the data. Enzyme kinetic parameters of atRA formation from retinaldehyde with recombinant human ALDH1A1 were measured as described previously (Arnold et al., 2015b). To test cofactor dependence of atRA formation with HLS9, pooled HLS9 (0.02 µg protein µl−1) were incubated with 2 µM retinaldehyde in the absence or presence of various cofactors. Tested cofactors and their final concentrations in incubations were 2 mM NAD+, 2 mM NADP+, and 1 mM NADPH. Incubation time was 30 seconds. Samples were processed with the ACN precipitation method. To perform kinetic characterization of atRA formation with HLS9, pooled HLS9 (0.005 mg protein ml−1) were incubated in the presence or absence of 2 mM NAD+ for 40 seconds with various concentrations of retinaldehyde ranging from 10 nM to 10 µM. Samples were processed with the hexane extraction method. Enzyme kinetic parameters, including the maximum atRA formation velocity (Vmax) and Michaelis-Menten constant in biologic matrix (Km), were obtained using GraphPad Prism 5.0 by fitting the Michaelis-Menten equation to the data. As two enzymes were involved in atRA formation in HLS9 when NAD+ was present, a two-enzyme Michaelis-Menten equation (eq. 1) was fit to the data to obtain kinetic parameters:Embedded Image(1)In eq. 1, Km of enzyme 1 (Km1) was constrained to be the value obtained from incubations performed without NAD+, and Vmax1 constrained to be larger than 400 pmol minute−1 mg−1 S9 protein. To build Eadie-Hofstee plots, atRA formation velocity (v) was plotted against v/[S], in which S is the substrate concentration.

Effects of CRBP1 on atRA Formation by AOX and ALDH1A1.

The effect of CRBP1 on atRA formation by AOX and ALDH1A1 was tested with recombinant proteins and pooled HLS9. CRBP1-bound retinaldehyde (holo-CRBP1) was prepared by incubating 4 µM retinaldehyde for 5 minutes at room temperature with 4 µM CRBP1 in 100 mM KPi buffer (pH 7.4). The incubations were conducted with recombinant enzymes (5 nM AOX; 20 nM ALDH1A1) or pooled HLS9 (0.005 mg protein ml−1) in 100 mM KPi buffer containing 2 mM NAD+. The total incubation volume was 300 µl for recombinant enzymes and 1000 µl for pooled HLS9. Incubations were initiated by adding holo-CRBP1 (200 nM final concentration) or retinaldehyde (200 nM final concentration). The reaction was stopped by adding an equal volume of ice-cold ACN with 4 nM atRA-d5. The incubation time was 1, 3, and 2 minutes for recombinant AOX, ALDH1A1, and HLS9, respectively. Samples were processed using the hexane extraction method and analyzed using LC-MS/MS.

Inhibition of atRA Formation by AOX and ALDH1A1 Inhibitors Hydralazine, Raloxifene, and WIN18,446.

Both hydralazine and raloxifene are known AOX inhibitors (Pryde et al., 2010; Strelevitz et al., 2012), whereas WIN18,446 is a selective ALDH1A inhibitor (Arnold et al., 2015a). To build IC50 curves of hydralazine and raloxifene against AOX and to test whether WIN18,446 inhibits AOX activity, 10 nM recombinant human AOX was incubated with 0.5 µM retinaldehyde and five to eight concentrations of raloxifene (1 nM–5 µM), hydralazine (1 nM–100 µM), and WIN18,446 (1–500 µM) for 2 minutes. All samples were processed with the ACN precipitation method. The IC50 values were determined by nonlinear regression using GraphPad Prism 5.0 as described previously (Thatcher et al., 2011).

To test whether hydralazine and raloxifene inhibit ALDH1A1 activity, 20 nM ALDH1A1 was incubated with 0.2 µM retinaldehyde, 2 mM NAD+, and hydralazine (10 nM-1 mM) or raloxifene (1 nM-5 µM) in buffer containing 150 mM KCl and 50 mM Hepes-NaOH (pH 8.0). Incubations were initiated with the addition of NAD+. The incubation time was 5 minutes. Control groups were incubations with no inhibitor. All samples were processed with the ACN precipitation method. The percentage of control activity was calculated and plotted against inhibitor concentrations.

To determine the relative contribution to atRA formation by AOX and ALDH1A1 in human liver S9 fractions, pooled HLS9 and HLS9 from 3 individual liver donors (0.01 mg protein ml−1) were incubated with inhibitors and 0.2 µM retinaldehyde in the presence or absence of 2 mM NAD+ for 30 seconds. Tested inhibitors included WIN18,446 and hydralazine (final concentration 250 µM). Control groups contained no inhibitor. Samples were processed with the hexane extraction method, and the percentage of control activity was calculated. Incubations were also performed with 1 µM retinaldehyde as a substrate to test the impact of substrate concentration on enzyme contribution. Final concentrations of WIN18,446 and hydralazine were 500 µM in these experiments. Samples were processed with the ACN precipitation method.

Quantification of AOX and ALDH1A1 in HLS9 Fractions Using LC-MS/MS.

AOX and ALDH1A1 protein concentrations in HLS9 were measured based on previously published protocols with minor modifications (Arnold et al., 2016). Stable isotope–labeled peptides (SIL-peptides) were ordered from Thermo Fisher Scientific and used as IS for AOX and ALDH1A1 quantification (Table 1). SIL-peptides of AOX and ALDH1A1 were labeled with [13C615N2]arginine and [13C615N2]lysine, respectively.

View this table:
TABLE 1

Peptide sequences and MS parameters for AOX and ALDH1A signature peptides selected for protein quantification using LC-MS/MS

Labeled amino acids in the SIL-peptides are shown as italicized letters. SIL-peptides of AOX and ALDH1A1 were labeled with [13C615N2]arginine and [13C615N2]lysine, respectively. ALDH1A2 and ALDH1A3 were not detected in any HLS9 samples used in this study.

Mouse liver S9 fractions were used as blank matrix (Arnold et al., 2016) and diluted to 0.5 mg ml−1 with 100 mM ammonium bicarbonate buffer (pH 7.8). Various amounts of purified recombinant human AOX (0–4 pmol) and ALDH1A1 (0–10 pmol) were spiked into the diluted mouse liver S9 fractions (20 µl per sample) to make standard curve samples as described previously (Arnold et al., 2016). HLS9 fractions (10 µg, 0.5 mg/ml) and standard curve samples were first incubated with 4 µl of 100 mM dithiothreitol and 10 µl of 100 mM ammonium bicarbonate buffer (pH 7.8) for 20 minutes at room temperature. After 5 μl of 10% sodium deoxycholate was added, samples were incubated at 95°C for 5 minutes, and then 4 μl of 200 mM iodoacetamide was added, and the samples were incubated at room temperature for 20 minutes. The protein was digested with trypsin at 37°C for 15 hours at a 1:25 trypsin/protein (w/w) ratio. Trypsin digestion was stopped by the addition of 20 μl ice-cold ACN containing 8% trifluoroacetic acid and internal standards (50 nM SIL-peptide for AOX and 100 nM for ALDH1A1). Samples were centrifuged at 18,000 g at 4°C for 30 minutes, and supernatant was collected for LC-MS/MS analysis.

AOX and ALDH1A1 were quantified via measurement of the signature peptides (Table 1) using an AB Sciex 5500 QTrap quadrupole linear ion trap mass spectrometer coupled with an Agilent 1290 HPLC and an Aeris peptide XB-C18 column (50 × 2.1 mm; 1.7-μm particle size). The MS parameters used are listed in Table 1. MS transitions for signature peptides of ALDH1A2 and ALDH1A3 were also monitored. Other MS parameters and the chromatographic conditions were as described previously (Arnold et al., 2016). The ratio of signature peptide peak area to SIL IS peptide area was used to construct standard curves and calculate AOX and ALDH1A1 concentrations in HLS9 fractions. The correlation between the protein expression levels and the atRA formation velocity was analyzed by first log-transforming the data and then testing the correlation via Pearson correlation. All measurements were performed in duplicates on three separate days.

Predictions of atRA Formation Velocity and Intrinsic Clearance of Retinaldehyde in HLS9, and the Relative Contribution of AOX and ALDH1A1 to atRA Formation.

The intrinsic clearance (CLint) of atRA formation from retinaldehyde was calculated using eq. 2:Embedded Image(2)

To scale the activity from the recombinant enzymes to HLS9, intersystem extrapolation factor (ISEF) was calculated using eq. 3:Embedded Image(3)in which CLint, pooled liver S9 is the measured intrinsic clearance in the pooled HLS9 sample, and [enzyme] is the enzyme expression measured by LC-MS/MS in the pooled HLS9 sample. ISEFAOX is calculated with CLint measured in the absence of NAD+, whereas ISEFALDH1A1 is calculated with CLint measured in the presence of NAD+ minus the CLint measured in the absence of NAD+.

The atRA formation velocity in individual HLS9 samples was predicted based on the measured AOX and ALDH1A1 expression levels and the kinetics of atRA formation by recombinant enzymes using eq. 4:Embedded Image(4)in which venzyme is the predicted atRA formation velocity by a particular enzyme, [enzyme] is the enzyme expression measured in individual HLS9 samples prepared in this study, and [S] is the total substrate concentration. The average fold error (afe) of predictions of atRA formation velocity in three HLS9 samples from individual donors was calculated using eq. 5:Embedded Image(5)in which vpredicted and vobserved are the predicted and observed atRA formation velocities of the individual HLS9 samples with 1 µM retinaldehyde as substrate. The afe values ranging from 0.5 to 1.5 were considered acceptable.

The predicted fraction metabolized by a particular enzyme (fm) was calculated using eq. 6:Embedded Image(6)in which vtotal is the total atRA formation velocity (vtotal = vALDH1A1 + vAOX), and venzyme is calculated from eq. 4, in which [enzyme] is the average value of the enzyme expression measured in three HLS9 samples from individual donors.

The percent inhibition (%inhibition) of atRA formation in HLS9 caused by CRBP1 was predicted using eq. 7:Embedded Image(7)in which %inhibitionenzyme is the observed percent inhibition of the recombinant enzyme activity caused by CRBP1. Predicted fm values by AOX and ALDH1A1 in the presence of CRBP1 (cfm,enzyme) were calculated using eq. 8:Embedded Image(8)

In both eqs. 7 and 8, fm,enzyme indicates the observed relative contribution of a particular enzyme to atRA formation.

Results

Kinetics of atRA Formation by AOX and ALDH1A1 and Inhibition by Raloxifene, Hydralazine, and WIN18,446.

Recombinant human AOX oxidized retinaldehyde to yield atRA (Fig. 1A) with a Km of 1.5 ± 0.4 µM and kcat of 3.6 ± 2.0 minute−1 (mean ± S.D., n = 3), resulting in a CLint of 2.6 ± 1.5 µl minute−1 pmol−1. In comparison, the Km of retinaldehyde oxidation with ALDH1A1 was 0.25 µM, and kcat was 0.38 minute−1, resulting in a CLint of 1.5 µl minute−1 pmol−1 (Supplemental Fig. 1). AOX inhibitors raloxifene and hydralazine both inhibited atRA formation by recombinant AOX, with IC50 values <1 µM (Fig. 1, E and F). The IC50 value of hydralazine was about 3-fold higher than that of raloxifene (Table 2). WIN18,446 did not inhibit AOX activity at inhibitor concentrations up to 500 µM, whereas both AOX inhibitors, raloxifene and hydralazine, exhibited inhibitory effects on recombinant ALDH1A1, although at higher inhibitor concentrations than those causing AOX inhibition (Fig. 1, B–D). Based on these data, WIN18,446 is a selective inhibitor of ALDH1A1 at concentrations up to 100 µM. The difference in IC50 values between AOX and ALDH1A1 for raloxifene was only 10-fold, suggesting raloxifene is not sufficiently selective as an inhibitor of AOX. In contrast, hydralazine can be used as a selective and potent inhibitor of AOX based on the >4000-fold difference in the IC50 values between AOX and ALDH1A1 (Table 2).

Fig. 1.
Fig. 1.

Characterization of atRA formation by recombinant human AOX, and inhibition of AOX- and ALDH1A1-mediated atRA formation by raloxifene, hydralazine, and WIN18,446. (A) shows a representative Michaelis-Menten plot of atRA formation with recombinant AOX. (B and C) show the inhibition of recombinant ALDH1A1-mediated atRA formation by raloxifene (B) and hydralazine (C). (D–F) show the inhibition of recombinant AOX-mediated atRA formation by WIN18,446 (D), raloxifene (E), and hydralazine (F). For determination of IC50 values, the bottom and the top of IC50 curves were constrained to be >0 and <120, respectively. Each panel shows a representative plot obtained from one of the three experiments performed on separate days with open circles showing data for recombinant AOX and open squares showing data for recombinant ALDH1A1. Data points and error bars represent mean and S.D. of triplicates within the experiment.

View this table:
TABLE 2

IC50 values of raloxifene, hydralazine, and WIN18,446 measured against recombinant AOX and ALDH1A1 enzymes (mean ± S.D., n = 3)

The selectivity index was calculated by dividing the IC50 value of the nontarget enzyme by the IC50 value of the target enzyme.

atRA Formation in HLS9.

The potential enzymes forming atRA in human liver all use different cofactors, and therefore, cofactor dependence can be used to differentiate the enzyme contributions to atRA formation. In HLS9, atRA formation was observed in the absence of cofactors (NAD+/NADP+/NADPH; Fig. 2A), with a substrate concentration of 2 µM. A linear Eadie-Hofstee plot (Fig. 2B) indicated that a single enzyme in HLS9 contributed to atRA formation in the absence of cofactors. Vmax obtained by fitting the single-enzyme Michaelis-Menten equation to the data was 646 ± 132 pmol minute−1 mg−1 S9 protein, and Km of retinaldehyde was 1.4 ± 0.9 µM (mean ± S.D., n = 3), similar to the Km value measured with recombinant AOX. The CLint for atRA formation was 615 ± 397 µl minute−1 mg−1 S9 protein (mean ± S.D., n = 3). Based on the fact that this activity did not require a cofactor, the enzyme responsible for this activity is likely AOX.

Fig. 2.
Fig. 2.

Characterization of atRA formation in human liver. (A) shows the dependence of atRA formation on different cofactors in pooled HLS9. Each data point represents a mean of triplicate assays performed on separate days. Horizontal lines and error bars represent mean and S.D. of the measurements over the different days. The retinaldehyde concentration used in the incubations was 2 μM. Nonenzymatic atRA formation was subtracted from the total atRA formation to calculate the velocity of enzymatic atRA formation in each group. (B and C) show retinaldehyde concentration–dependent atRA formation by HLS9 in the absence (B) and presence (C) of NAD+. Representative Eadie-Hofstee plots of atRA formation by HLS9 are shown, with the Michaelis-Menten plots of atRA formation shown in insets. Each Eadie-Hofstee plot is a representative plot obtained from one of the three experiments performed on separate days. Individual data points shown as open circles and error bars represent mean and S.D. of triplicates within the day. The inset in (B) shows the fit of a single-enzyme Michaelis-Menten equation to the data, and the inset in (C) shows the fit of a two-enzyme Michaelis-Menten equation to the data as described in Materials and Methods. [S], substrate concetration; v, velocity.

The addition of NADPH and NADH to the incubations did not result in any increase in atRA formation velocity in comparison to incubations without a cofactor, indicating that cytochrome P450 and NADH-dependent enzymes play a negligible role in retinaldehyde oxidation in human liver (Fig. 2A). In the presence of NAD+, the velocity of atRA formation measured with 2 µM retinaldehyde was about 1.8-fold higher than that measured in the absence of NAD+. The Eadie-Hofstee plot (Fig. 2C) indicated that two enzymes in HLS9 fractions were responsible for atRA synthesis in the presence of NAD+. The low-affinity enzyme was assumed to be AOX with the kinetic parameters described above. As the two-enzyme Michaelis-Menten model was fitted to the data, the high-affinity enzyme Vmax and Km values were 518 ± 288 pmol minute−1 mg−1 S9 protein and 0.38 ± 0.13 µM (mean ± S.D. n = 3), respectively. The CLint of the high-affinity enzyme was 1299 ± 370 µl minute−1 mg−1 S9 protein (mean ± S.D., n = 3), and the overall CLint was 1914 µl minute−1 mg−1 S9 protein. Based on the requirement of NAD+ for the activity of the observed second enzyme, this activity is likely ALDH1A1. Taken together, these data strongly support that both ALDH1A1 and AOX contribute to atRA biosynthesis in human liver.

Predicted Contribution of AOX and ALDH1A1 to atRA Formation in HLS9.

The signature peptide of AOX for quantification was selected based on the MS signal intensity and peptide stability in matrix (Supplemental Fig. 2). The selected human AOX and ALDH1A1 quantification peptides were observed in HLS9 but not in mouse liver S9, demonstrating the specificity of these peptides (Fig. 3, A and B). The expression level of ALDH1A1 in HLS9 (156–285 pmol mg S9 protein−1) was on average 13-fold higher than the expression of AOX (9.4–24 pmol mg S9 protein−1) (Fig. 3C). No expression of ALDH1A2 or ALDH1A3 was detected in the HLS9 (data not shown).

Fig. 3.
Fig. 3.

Characterization of expression of AOX and ALDH1A1 and their predicted contribution to atRA formation in human liver. (A and B) show the LC-MS/MS measurement of AOX and ALDH1A1 protein expression. The quantified MS/MS transitions of AOX (m/z 605.3 > 963.5) and ALDH1A1 (m/z 795.4 > 879.5) signature peptides are shown in black and orange, respectively. (A) shows the LC-MS/MS chromatogram of a sample of mouse liver S9 (MLS9) fraction spiked with 1 pmol purified recombinant human AOX and 2.5 pmol purified recombinant human ALDH1A1 (standard sample). The inset in (A) shows MLS9 without adding any recombinant human AOX and ALDH1A1. (B) shows a representative chromatogram of pooled human liver S9 fraction and detection of AOX and ALDH1A1 in human liver. (C) shows the correlation between AOX and ALDH1A1 protein expression, and (D and E) show the correlation between the ALDH1A1 (D) and AOX (E) protein expression and the observed atRA formation velocity measured at 1 µM retinaldehyde. “Activity (NAD+)” indicates the total velocity measured in the presence of NAD+ that reflects contribution by both AOX and ALDH1A1; “Activity (No NAD+)” indicates the velocity measured in the absence of NAD+ reflecting only AOX activity; Activity (NAD+-No NAD+) = Activity (NAD+) - Activity (No NAD+). Each symbol represents an individual HLS9 sample. Symbols represent the mean values of three individual experiments conducted on separate days in duplicates. The dashed lines show the linear regression between the enzyme expression and atRA formation velocity; P values shown in the panels indicate whether the slope of the corresponding line is significantly different from zero. (F) shows the dependence of predicted ALDH1A1 and AOX contributions to atRA formation (fm) on retinaldehyde concentrations. The solid lines indicate predicted fm of atRA formation by ALDH1A1 (orange) and AOX (black) in HLS9 fractions.

Using CLint obtained in the pooled HLS9 and recombinant enzymes, ISEFAOX and ISEFALDH1A1 were calculated to be 25 and 6, respectively. The predicted velocity of atRA formation mediated by AOX in the three individual livers ranged from 586 to 846 pmol minute−1mg−1 S9, and the predicted total velocity including AOX and ALDH1A1 ranged from 943 to 1366 pmol minute−1mg−1 S9. When compared with the observed formation velocities in the three individual livers, AOX-mediated formation of atRA had an afe of 1.35, and the total velocity had an afe of 0.71, both within 2-fold of the observed velocities measured without (371–655 pmol minute−1mg−1 S9) and with (1377–1861 pmol minute−1mg−1 S9; Fig. 3) NAD+. With NAD+, atRA formation velocity measured with 1 µM retinaldehyde positively correlated with the enzyme expression of ALDH1A1 (r = 0.99; P = 0.011) and AOX (r = 0.98; P = 0.022) in HLS9 (Fig. 3, D and E). However, the AOX expression did not show a correlation with atRA formation velocity measured in the absence of NAD+ (r = -0.80; P = 0.20; Fig. 3E), possibly suggesting that not all MS-quantified AOX in the liver is catalytically active (Fu et al., 2013). The expression of ALDH1A1 showed a weak correlation with the expression of AOX (r = 0.94; P = 0.065; Fig 3C).

The fraction of atRA formation by AOX and ALDH1A1 (fm) is expected to vary with retinaldehyde concentration because of the differences in the Km values for the two enzymes. Based on the AOX and ALDH1A1 expression and atRA formation kinetics, the fm,AOX was predicted to increase from 0.36 to 0.73, and fm,ALDH1A1 was predicted to decrease from 0.64 to 0.27 as retinaldehyde concentration increased from low nanomolar concentrations to saturating concentrations (5 µM) (Fig. 3F). As such, determination of the relative importance of each enzyme to atRA formation in human liver depends on substrate concentrations and should be assessed at biologically relevant concentrations of retinaldehyde.

Observed Relative Contribution of AOX and ALDH1A1 to atRA Formation in HLS9.

To determine the quantitative contributions of ALDH1A1 and AOX to atRA formation in human livers, inhibition of atRA formation in human liver by hydralazine (selective AOX inhibitor) and WIN18,446 (selective ALDH1A1 inhibitor) was characterized. Inhibitor concentrations used in HLS9 incubations were first tested with purified recombinant enzymes to confirm selectivity and efficacy of inhibition. At their selected concentrations, hydralazine and WIN18,446 caused >95% inhibition of their target enzymes, AOX and ALDH1A1, without inhibiting the nontargeted enzymes (Fig. 4, A and B). In HLS9 incubations without NAD+ cofactor, hydralazine inhibited >95% of atRA formation, whereas WIN18,446 did not lead to any significant inhibition (Fig. 4, C and D). These data confirm that AOX is the enzyme responsible for the non–NAD+-dependent formation of atRA. In the presence of NAD+, hydralazine inhibited only 24% ± 6% (mean ± S.D., n = 4) of atRA formation in HLS9 with a substrate concentration of 0.2 µM, whereas WIN18,446 inhibited 81% ± 6% (mean ± S.D., n = 4) of atRA formation (Fig. 4E). The predicted fm of ALDH1A1 (0.53) underpredicted ALDH1A1 contribution to atRA formation in HLS9 at 0.2 µM retinaldehyde, whereas that of AOX (0.47) was overestimated (Fig. 3F). With 1 µM retinaldehyde as a substrate, hydralazine and WIN18,446 inhibited 47% ± 15% and 56% ± 11% (mean ± S.D., n = 4) of the total atRA formation activity (Fig. 4F), respectively, demonstrating the decreased contribution of ALDH1A1 to atRA formation at higher substrate concentrations. The observed percent inhibition at 1 µM retinaldehyde was higher than the predicted fm of ALDH1A1 (0.38) and lower than the predicted fm of AOX (0.62) (Fig. 3F). The combination of hydralazine and WIN18,446 abolished >95% of atRA formation in HLS9 samples when incubated in the presence of NAD+, suggesting that AOX and ALDH1A1 are the only enzymes contributing to atRA formation in HLS9.

Fig. 4.
Fig. 4.

Effect of selective AOX and ALDH1A1 inhibitors on atRA formation by recombinant enzymes and human liver S9 fractions. Retinaldehyde concentrations used in incubations were 0.2 (left column) and 1 μM (right column), and inhibitor concentrations (HDZ and WIN) were 250 μM in left panels and 500 μM in right panels. In (A and B), inhibition assays were performed using recombinant human ALDH1A (open triangles) and AOX (open circles). In (C–F), inhibition assays were done using HLS9 fractions, and (C and D) show incubations in the absence of NAD+, whereas (E and F) show incubations in the presence of NAD+. Bars represent mean values of three individual experiments performed on separate days, and each data point represent an individual experiment done in duplicates. Different bars and symbols represent pooled HLS9 (open circles, filled gray bars) and HLS9 fractions from three individual donors HL136 (open triangles and dashed bars), HL146 (filled triangles and white bars), and HL149 (open squares and dotted bars). HDZ, hydralazine; WIN, WIN18,446.

Effects of CRBP1 on AOX and ALDH1A1 Activity.

To determine whether binding to CRBP1 affects the formation of atRA by AOX or ALDH1A1 and their relative importance in hepatic atRA formation, atRA formation velocity was measured with recombinant enzymes and HLS9 using holo-CRBP1 as the substrate. As shown in Fig. 5A, in the presence of CRBP1, the formation of atRA by AOX and ALDH1A1 decreased about 85% and 55% in comparison with control incubations that contained no CRBP1. As AOX and ALDH1A1 contribute about 20% and 80% to atRA formation in HLS9 in the absence of CRBP1, the activity of atRA formation in HLS9 was predicted to decrease by 61% in the presence of CRBP1. The fm values by AOX and ALDH1A1 were estimated to be 0.08 and 0.92, respectively, in the presence of CRBP1 (Fig. 5B). The decrease in atRA formation caused by CRBP1 was tested in HLS9, and the addition of CRBP1 decreased atRA formation by 63% when compared with the no-CRBP control (Fig. 5A), a level of inhibition consistent with predictions.

Fig. 5.
Fig. 5.

Effects of CRBP1 on atRA formation by recombinant AOX, ALDH1A1, and human liver S9. (A) shows atRA formation velocity when free retinaldehyde (−CRBP) or holo-CRBP1 (+CRABP) is used as a substrate. The bars represent the mean percent inhibition from three individual experiments conducted on separate days, and error bars represent S.D. (B) shows the predicted fraction of retinaldehyde metabolized by AOX and ALDH1A1 in the presence and absence of CRBP1. The substrate concentration (retinaldehyde or CRBP1-bound retinaldehyde) was 0.2 µM in all experiments.

Discussion

To date, the majority of the knowledge regarding the role of specific atRA biosynthesis enzymes is derived from studies in developmental biology and altered retinoid signaling during embryonic development. However, the biologic importance of vitamin A metabolizing enzymes defined during embryonic development does not always translate to postnatal life. For example, the embryonic knockout of Aldh1a2 leads to embryonic death in mice (Niederreither et al., 2002), but neither tamoxifen-induced postnatal global knockout of Aldh1a2 or germ cell–specific deletion of Aldh1a2 caused any abnormal phenotype (Beedle et al., 2019). A similar discrepancy was observed with atRA hydroxylase Cyp26a1. Cyp26a1 knockout is embryonically lethal, but postnatal knockout has no major effects on mouse health (Zhong et al., 2019a). Although knockout mouse models have allowed considerable progress in defining the enzymology of vitamin A and retinoid homeostasis, knowledge of the role of specific enzymes in atRA biosynthesis in human tissues is still limited. This includes limited knowledge of the identity of the enzymes synthesizing atRA in the human liver, the main storage organ for vitamin A. The goal of this study was to determine the main enzymes synthesizing atRA in the human liver and test the hypothesis that AOX contributes to atRA biosynthesis in the human liver.

This study unequivocally shows that AOX forms atRA from retinaldehyde and that in human liver ALDH1A1 and AOX contribute to atRA biosynthesis. Recombinant AOX exhibited about 6-fold higher Km and 1.6-fold higher CLint than ALDH1A1 for atRA formation, showing that ALDH1A1 is a high-affinity, low-capacity atRA synthesizing enzyme, whereas AOX is a low-affinity, high-capacity enzyme. According to the human protein atlas data base, AOX protein is expressed in the kidney, bladder, pancreas, endocrine tissues, and reproductive tissues (https://www.proteinatlas.org/ENSG00000138356-AOX1/tissue). This suggests that AOX may contribute to atRA synthesis in the testis, among other tissues. However, studies in healthy mice have shown that AOX activity is not quantitatively significant in the mouse testis (Arnold et al., 2015a) but rather that AOX plays a role in testicular atRA synthesis in mice when ALDH1A activity is impaired (Beedle et al., 2019). This is in agreement with the findings in the human testis showing that ALDH1A1 and ALDH1A2 are responsible for >95% of atRA formation in this tissue (Arnold et al., 2015). In contrast to the testis, AOX appears to contribute about 50% of atRA biosynthesis in mouse liver (Arnold et al., 2015a). This suggests a tissue-specific role of AOX that likely depends on the expression levels of AOX and the complement of ALDH1A enzymes expressed in the different tissues. The results of this study support a similar role of AOX in atRA biosynthesis in human liver as in the mouse liver, with AOX contributing between 33% and 70% of atRA biosynthesis depending on retinaldehyde concentration. The reported total concentrations of retinaldehyde in the liver and adipose are 100–200 pmol/g (∼150 nM) (Ziouzenkova et al., 2007; Kane et al., 2008). These concentrations are much below the Km toward AOX (1.5 µM) but similar to the Km toward ALDH1A1 (0.25 µM). Because of the difference in Km values for AOX and ALDH1A1, the predicted and observed fm values of retinaldehyde by AOX show a clear dependence on retinaldehyde concentration within the evaluated range of 20–2,000 nM retinaldehyde. As the concentration of retinaldehyde increases, fm,AOX increases due to the saturation of the high-affinity ALDH1A1. This also suggests that in the case of ALDH1A dysfunction (inhibition, downregulation, or genetic defect) or excess retinoid/vitamin A (saturation of ALDH1A1), AOX plays an important role in regulating hepatic vitamin A homeostasis. The relative contribution of AOX (30%–70%) to atRA biosynthesis is supported by the scaling of atRA formation from recombinant enzymes, the quantification of AOX and ALDH1A1 protein expression in the liver, the cofactor dependence of atRA formation, and the inhibition of atRA formation by selective ALDH1A and AOX inhibitors WIN18,446 and hydralazine. These tools developed in the current study allow differentiation of ALDH1A and AOX contributions to atRA formation and can be further applied to other critical retinoid responsive tissues, such as the skin, hematopoietic cells, and adipose tissue, to determine the enzymes contributing to atRA biosynthesis. Such information will be useful in predicting how inhibition or genetic variability in AOX and ALDH1A enzymes will affect retinoid homeostasis and signaling.

In tissues that contribute to vitamin A homeostasis, CRBP1 is expressed, and retinoids are believed to be bound to retinoid binding proteins (Noy, 2000; Napoli, 2016). As such, CRBP1 can impact retinoid metabolism and the enzymology of atRA biosynthesis. The presence of CRBP1 decreased the atRA formation velocity by both ALDH1A1 and AOX. This finding is in agreement with previous demonstration that with CRBP1, the activity of ALDH1A1-mediated atRA formation decreased about 52% when compared with the activity measured without CRBP1 (Arnold et al., 2015). Similarly, with 0.2 µM CRBP1-bound retinaldehyde as a substrate, atRA formation velocity measured in the E2 fraction (associated with AOX) from soluble extracts of human liver was about 70% to 80% lower than that measured without CRBP1 (Ambroziak et al., 1999). However, as shown here, CRBP1 binding decreased atRA formation by both ALDH1A1 and AOX, and therefore the impact of CRBP1 on the fm values by AOX and ALDH1A1 is likely modest. In addition, CRBP1 binding is unlikely to have a significant impact on atRA formation in vivo, as CRABP1 in human liver is expected to be bound by hepatic retinol [1.2–162 µM; (Zhong et al., 2019b)], which has a Kd of ∼18.6 nM to CRBP1 compared with Kd of retinaldehyde ∼44 nM; (Ambroziak et al., 1999; Silvaroli et al., 2016).

The results of this study show that the human liver possesses a high atRA biosynthesis capacity. The observed atRA formation velocity in the HLS9 observed here (1043–1861 pmol minute−1mg−1 S9 protein) is 31- to 56-fold higher than the observed atRA formation velocity in human testis (33.3 pmol minute−1mg−1 S10 protein at 1 µM retinaldehyde) (Arnold et al., 2015). This difference in atRA formation velocities is comparable to the difference between human testicular (4.3 ± 0.5 pmol g−1 testis) and hepatic tissue atRA concentrations (14–580 pmol g−1, median 43 pmol g−1 liver) (Arnold et al., 2015; Zhong et al., 2019b). This finding supports the previous report in human testis that variability in tissue atRA formation kinetics could be used to predict the variability in tissue atRA concentrations (Arnold et al., 2015). The velocity of atRA formation in HLS9 was successfully predicted at a single concentration of retinaldehyde in this study when AOX and ALDH1A1 protein expression and specific ISEF values were applied to individual donors. As reported in the literature, with xenobiotic substrates of AOX, AOX activity is often poorly scaled from one system to another (e.g., recombinant enzyme to cytosol, in vitro to in vivo), and recombinant AOX typically underpredicts the activity in native environment (Pryde et al., 2010; Fu et al., 2013), suggesting that atRA biosynthesis may be more efficient by AOX in vivo than predicted from the HLS9 here. This finding was confirmed in the current study, with the data suggesting an ISEF value of 20 for AOX. Surprisingly, ALDH1A1 also had an ISEF value of 6 in this study when previous work in the human testis indicated an ISEF value of approximately 2 for ALDH1A. As ISEF values may vary for different tissues and for different enzymes, these findings emphasize the importance of independent experiments such as selective inhibition studies to determine the relative contributions of individual enzymes to atRA formation.

In conclusion, this study shows that human AOX is responsible for 30%–70% of atRA biosynthesis in human liver. The importance of AOX and ALDH1A1 in atRA biosynthesis and in regulation of atRA signaling is an intriguing intersection between xenobiotic and endogenous compound metabolism. As many AOX inhibitors are therapeutic drugs, the administration of these drugs may affect atRA gradients and signaling in the liver and other tissues. Determining whether AOX inhibition in the liver or in other organs alters atRA concentrations and biosynthesis requires further study. Additionally, reported single nucleotide polymorphisms of human AOX (Hartmann et al., 2012; Hutzler et al., 2014; Foti et al., 2017) may have an impact on in vivo retinaldehyde clearance and tissue atRA gradients. As such the impact of altered AOX activity on atRA concentrations and vitamin A homeostasis should be considered in preclinical and clinical studies.

Authorship Contributions

Participated in research design: Zhong, Isoherranen.

Conducted experiments: Zhong, Seaman, Xi, Paragas, Herpoldt.

Contributed new reagents or analytic tools: Paragas, Jones, Herpoldt, King.

Performed data analysis: Zhong, Seaman.

Wrote or contributed to the writing of the manuscript: Zhong, Seaman, Paragas, Herpoldt, King, Jones, Isoherranen.

Footnotes

Abbreviations

ACN
acetonitrile
afe
average fold error
ALDH1A
aldehyde dehydrogenase 1A
AOX
aldehyde oxidase
atRA
all-trans-retinoic acid
CLint
intrinsic clearance
CRBP
cellular retinol binding protein
fm
fraction metabolized
HLS9
human liver S9 fraction
IS
internal standard
ISEF
intersystem extrapolation factor
KPi
potassium phosphate
LC-MS/MS
liquid chromatography–tandem mass spectrometry
m/z
mass-to-charge ratio
SIL-peptide
stable isotope–labeled peptide

References

  1. ↵
    1. Ambroziak W,
    2. Izaguirre G, and
    3. Pietruszko R
    (1999) Metabolism of retinaldehyde and other aldehydes in soluble extracts of human liver and kidney. J Biol Chem 274:33366–33373.
    OpenUrlAbstract/FREE Full Text
  2. ↵
    1. Arnold SL,
    2. Kent T,
    3. Hogarth CA,
    4. Griswold MD,
    5. Amory JK, and
    6. Isoherranen N
    (2015a) Pharmacological inhibition of ALDH1A in mice decreases all-trans retinoic acid concentrations in a tissue specific manner. Biochem Pharmacol 95:177–192.
    OpenUrlCrossRefPubMed
  3. ↵
    1. Arnold SL,
    2. Kent T,
    3. Hogarth CA,
    4. Schlatt S,
    5. Prasad B,
    6. Haenisch M,
    7. Walsh T,
    8. Muller CH,
    9. Griswold MD,
    10. Amory JK, et al.
    (2015b) Importance of ALDH1A enzymes in determining human testicular retinoic acid concentrations. J Lipid Res 56:342–357.
    OpenUrlAbstract/FREE Full Text
  4. ↵
    1. Arnold SL,
    2. Stevison F, and
    3. Isoherranen N
    (2016) Impact of sample matrix on accuracy of peptide quantification: assessment of calibrator and internal standard selection and method validation. Anal Chem 88:746–753.
    OpenUrl
  5. ↵
    1. Barr JT,
    2. Jones JP,
    3. Joswig-Jones CA, and
    4. Rock DA
    (2013) Absolute quantification of aldehyde oxidase protein in human liver using liquid chromatography-tandem mass spectrometry. Mol Pharm 10:3842–3849.
    OpenUrlCrossRef
  6. ↵
    1. Beedle MT,
    2. Stevison F,
    3. Zhong G,
    4. Topping T,
    5. Hogarth C,
    6. Isoherranen N, and
    7. Griswold MD
    (2019) Sources of all-trans retinal oxidation independent of the aldehyde dehydrogenase 1A isozymes exist in the postnatal testis. Biol Reprod 100:547–560.
    OpenUrlPubMed
  7. ↵
    1. Czuba LC,
    2. Zhong G,
    3. Yabut KC, and
    4. Isoherranen N
    (2020) Analysis of vitamin A and retinoids in biological matrices. Methods Enzymol 637:309–340.
    OpenUrl
  8. ↵
    1. Fan X,
    2. Molotkov A,
    3. Manabe S,
    4. Donmoyer CM,
    5. Deltour L,
    6. Foglio MH,
    7. Cuenca AE,
    8. Blaner WS,
    9. Lipton SA, and
    10. Duester G
    (2003) Targeted disruption of Aldh1a1 (Raldh1) provides evidence for a complex mechanism of retinoic acid synthesis in the developing retina. Mol Cell Biol 23:4637–4648.
    OpenUrlAbstract/FREE Full Text
  9. ↵
    1. Foti A,
    2. Dorendorf F, and
    3. Leimkühler S
    (2017) A single nucleotide polymorphism causes enhanced radical oxygen species production by human aldehyde oxidase. PLoS One 12:e0182061.
    OpenUrl
  10. ↵
    1. Fu C,
    2. Di L,
    3. Han X,
    4. Soderstrom C,
    5. Snyder M,
    6. Troutman MD,
    7. Obach RS, and
    8. Zhang H
    (2013) Aldehyde oxidase 1 (AOX1) in human liver cytosols: quantitative characterization of AOX1 expression level and activity relationship. Drug Metab Dispos 41:1797–1804.
    OpenUrlAbstract/FREE Full Text
  11. ↵
    1. Hartmann T,
    2. Terao M,
    3. Garattini E,
    4. Teutloff C,
    5. Alfaro JF,
    6. Jones JP, and
    7. Leimkühler S
    (2012) The impact of single nucleotide polymorphisms on human aldehyde oxidase. Drug Metab Dispos 40:856–864.
    OpenUrlAbstract/FREE Full Text
  12. ↵
    1. Huang D-Y,
    2. Furukawa A, and
    3. Ichikawa Y
    (1999) Molecular cloning of retinal oxidase/aldehyde oxidase cDNAs from rabbit and mouse livers and functional expression of recombinant mouse retinal oxidase cDNA in Escherichia coli. Arch Biochem Biophys 364:264–272.
    OpenUrlCrossRefPubMed
  13. ↵
    1. Hutzler JM,
    2. Yang Y-S,
    3. Brown C,
    4. Heyward S, and
    5. Moeller T
    (2014) Aldehyde oxidase activity in donor-matched fresh and cryopreserved human hepatocytes and assessment of variability in 75 donors. Drug Metab Dispos 42:1090–1097.
    OpenUrlAbstract/FREE Full Text
  14. ↵
    1. Kane MA,
    2. Folias AE, and
    3. Napoli JL
    (2008) HPLC/UV quantitation of retinal, retinol, and retinyl esters in serum and tissues. Anal Biochem 378:71–79.
    OpenUrlCrossRefPubMed
  15. ↵
    1. Kedishvili NY
    (2013) Enzymology of retinoic acid biosynthesis and degradation. J Lipid Res 54:1744–1760.
    OpenUrlAbstract/FREE Full Text
  16. ↵
    1. Napoli JL
    (2012) Physiological insights into all-trans-retinoic acid biosynthesis. Biochim Biophys Acta 1821:152–167.
    OpenUrlCrossRefPubMed
  17. ↵
    1. Napoli JL
    (2016) Functions of intracellular retinoid binding-proteins. Subcell Biochem 81:21–76.
    OpenUrlCrossRef
  18. ↵
    1. Niederreither K,
    2. Vermot J,
    3. Schuhbaur B,
    4. Chambon P, and
    5. Dollé P
    (2002) Embryonic retinoic acid synthesis is required for forelimb growth and anteroposterior patterning in the mouse. Development 129:3563–3574.
    OpenUrlAbstract/FREE Full Text
  19. ↵
    1. Noy N
    (2000) Retinoid-binding proteins: mediators of retinoid action. Biochem J 348:481–495.
    OpenUrlAbstract/FREE Full Text
  20. ↵
    1. Paragas EM,
    2. Humphreys SC,
    3. Min J,
    4. Joswig-Jones CA,
    5. Leimkühler S, and
    6. Jones JP
    (2017) ecoAO: a simple system for the study of human aldehyde oxidases role in drug metabolism. ACS Omega 2:4820–4827.
    OpenUrl
  21. ↵
    1. Pryde DC,
    2. Dalvie D,
    3. Hu Q,
    4. Jones P,
    5. Obach RS, and
    6. Tran TD
    (2010) Aldehyde oxidase: an enzyme of emerging importance in drug discovery. J Med Chem 53:8441–8460.
    OpenUrlCrossRefPubMed
  22. ↵
    1. Rashidi M-RR and
    2. Soltani S
    (2017) An overview of aldehyde oxidase: an enzyme of emerging importance in novel drug discovery. Expert Opin Drug Discov 12:305–316.
    OpenUrl
  23. ↵
    1. Silvaroli JA,
    2. Arne JM,
    3. Chelstowska S,
    4. Kiser PD,
    5. Banerjee S, and
    6. Golczak M
    (2016) Ligand binding induces conformational changes in human Cellular Retinol-binding Protein 1 (CRBP1) revealed by atomic resolution crystal structures. J Biol Chem 291:8528–8540.
    OpenUrlAbstract/FREE Full Text
  24. ↵
    1. Strelevitz TJ,
    2. Orozco CC, and
    3. Obach RS
    (2012) Hydralazine as a selective probe inactivator of aldehyde oxidase in human hepatocytes: estimation of the contribution of aldehyde oxidase to metabolic clearance. Drug Metab Dispos 40:1441–1448.
    OpenUrlAbstract/FREE Full Text
  25. ↵
    1. Terao M,
    2. Garattini E,
    3. Romão MJ, and
    4. Leimkühler S
    (2020) Evolution, expression, and substrate specificities of aldehyde oxidase enzymes in eukaryotes. J Biol Chem 295:5377–5389.
    OpenUrlAbstract/FREE Full Text
  26. ↵
    1. Terao M,
    2. Kurosaki M,
    3. Barzago MM,
    4. Fratelli M,
    5. Bagnati R,
    6. Bastone A,
    7. Giudice C,
    8. Scanziani E,
    9. Mancuso A,
    10. Tiveron C, et al.
    (2009) Role of the molybdoflavoenzyme aldehyde oxidase homolog 2 in the biosynthesis of retinoic acid: generation and characterization of a knockout mouse. Mol Cell Biol 29:357–377.
    OpenUrlAbstract/FREE Full Text
  27. ↵
    1. Thatcher JE,
    2. Buttrick B,
    3. Shaffer SA,
    4. Shimshoni JA,
    5. Goodlett DR,
    6. Nelson WL, and
    7. Isoherranen N
    (2011) Substrate specificity and ligand interactions of CYP26A1, the human liver retinoic acid hydroxylase. Mol Pharmacol 80:228–239.
    OpenUrlAbstract/FREE Full Text
  28. ↵
    1. Tomita S,
    2. Tsujita M, and
    3. Ichikawa Y
    (1993) Retinal oxidase is identical to aldehyde oxidase. FEBS Lett 336:272–274.
    OpenUrlCrossRefPubMed
  29. ↵
    1. Zhong G,
    2. Hogarth C,
    3. Snyder JM,
    4. Palau L,
    5. Topping T,
    6. Huang W,
    7. Czuba LC,
    8. LaFrance J,
    9. Ghiaur G, and
    10. Isoherranen N
    (2019a) The retinoic acid hydroxylase Cyp26a1 has minor effects on postnatal vitamin A homeostasis, but is required for exogenous atRA clearance. J Biol Chem 294:11166–11179.
    OpenUrlAbstract/FREE Full Text
  30. ↵
    1. Zhong G,
    2. Kirkwood J,
    3. Won K-J,
    4. Tjota N,
    5. Jeong H, and
    6. Isoherranen N
    (2019b) Characterization of vitamin A metabolome in human livers with and without nonalcoholic fatty liver disease. J Pharmacol Exp Ther 370:92–103.
    OpenUrlAbstract/FREE Full Text
  31. ↵
    1. Ziouzenkova O,
    2. Orasanu G,
    3. Sharlach M,
    4. Akiyama TE,
    5. Berger JP,
    6. Viereck J,
    7. Hamilton JA,
    8. Tang G,
    9. Dolnikowski GG,
    10. Vogel S, et al.
    (2007) Retinaldehyde represses adipogenesis and diet-induced obesity. Nat Med 13:695–702.
    OpenUrlCrossRefPubMed
Back to top

In this issue

Download PDF
Article Alerts
Email Article
Citation Tools

Research ArticleArticle

Retinoic Acid Synthesis by Aldehyde Oxidase

Guo Zhong, Chris J. Seaman, Erickson M. Paragas, Huaqing Xi, Karla-Luise Herpoldt, Neil P. King, Jeffrey P. Jones and Nina Isoherranen
Drug Metabolism and Disposition March 1, 2021, 49 (3) 202-211; DOI: https://doi.org/10.1124/dmd.120.000296
Twitter logo Facebook logo Mendeley logo

Jump to section

Advertisement