Reassignment of the human aldehyde dehydrogenase ALDH8A1 (ALDH12) to the kynurenine pathway in tryptophan catabolism

The kynurenine pathway is the primary route for l-tryptophan degradation in mammals. Intermediates and side products of this pathway are involved in immune response and neurodegenerative diseases. This makes the study of enzymes, especially those from mammalian sources, of the kynurenine pathway worthwhile. Recent studies on a bacterial version of an enzyme of this pathway, 2-aminomuconate semialdehyde (2-AMS) dehydrogenase (AMSDH), have provided a detailed understanding of the catalytic mechanism and identified residues conserved for muconate semialdehyde recognition and activation. Findings from the bacterial enzyme have prompted the reconsideration of the function of a previously identified human aldehyde dehydrogenase, ALDH8A1 (or ALDH12), which was annotated as a retinal dehydrogenase based on its ability to preferentially oxidize 9-cis-retinal over trans-retinal. Here, we provide compelling bioinformatics and experimental evidence that human ALDH8A1 should be reassigned to the missing 2-AMS dehydrogenase of the kynurenine metabolic pathway. For the first time, the product of the semialdehyde oxidation by AMSDH is also revealed by NMR and high-resolution MS. We found that ALDH8A1 catalyzes the NAD+-dependent oxidation of 2-AMS with a catalytic efficiency equivalent to that of AMSDH from the bacterium Pseudomonas fluorescens. Substitution of active-site residues required for substrate recognition, binding, and isomerization in the bacterial enzyme resulted in human ALDH8A1 variants with 160-fold increased Km or no detectable activity. In conclusion, this molecular study establishes an additional enzymatic step in an important human pathway for tryptophan catabolism.

To date, the KP pathway genes and their corresponding enzymes have not been identified beyond ACMSD (10,11), although the metabolic pathway was published 53 years ago and has made its way into numerous biochemistry textbooks. One limiting factor for studying KP enzymes is that the identification of their mammalian genes has proved difficult. Initial characterization of the KP enzymes was performed from animal liver extracts (1). Although these studies verified the activities and transformations of the KP, they were unable to provide much insight into the individual enzyme structures and mechanisms. Study of the KP enzymes stagnated until the discovery of an analogous KP in some bacteria (12)(13)(14)(15)(16)(17) and that 2-nitrobenzoate biodegradation shares many of the downstream proteins with the eukaryotic kynurenine pathway (18 -20). An additional difficulty for studying the KP enzymes, especially downstream proteins, is that several of the metabolic intermediates of the pathway are unstable and commercially unavailable. As shown in Scheme 1, the substrates for ACMSD and its downstream neighbor, 2-aminomuconate semialdehyde dehydrogenase (AMSDH), are both unstable and spontaneously cyclize via a pericyclic reaction to their respective pyridine products, QA and picolinic acid (PA) (2).

Extension of the mammalian kynurenine pathway
(HMSDH) family possesses two conserved arginine residues that are involved in substrate recognition and an isomerization activity (22).
In this work, we have identified a human enzyme annotated as a retinal dehydrogenase (ALDH8A1) that carries the hallmarks of an aminomuconate semialdehyde dehydrogenase. An overexpression system was constructed, and recombinant ALDH8A1 was tested for activity with two muconic semialdehyde substrates. The activity of selected active-site variants was also investigated, and the reaction products were verified with NMR and high-resolution MS. All evidence suggests that ALDH8A1 should be reconsidered as the aldehyde dehydrogenase of the kynurenine pathway of tryptophan catabolism.

Identification of ALDH8A1 as a potential member of the kynurenine pathway
To continue studying the KP pathway at the molecular level, the next pressing step is to identify a mammalian AMSDH, especially the human version. We performed a BLAST search with pfAMSDH as the search sequence. The results revealed a human protein, ALDH8A1 (initially designated ALDH12), with 44% amino acid sequence identity to pfAMSDH. However, ALDH8A1 is currently annotated in the NCBI gene database as a retinal dehydrogenase. It was assigned as cis-retinal dehydrogenase based on its ability to oxidize 9-cis-retinal faster than all-trans-retinal, even though it was most active with benzaldehyde rather than the retinal substrates (23). It was also noted in the original characterization that ALDH8A1 shares the closest nucleotide and protein sequence similarity with AMSDH, but it was not tested for such activity, presumably due to difficulty obtaining 2-aminomuconate semialdehyde (2-AMS). Here, we present evidence that ALDH8A1 may be more appropriately considered human AMSDH of the KP. AMSDH belongs to the HMSDH family. Our alignment of ALDH8A1 against AMSDH and several members of the HMSDH family showed not only high overall conservation but also that, in addition to residues required for aldehyde dehydrogenase activity (Asn-155, Glu-253, and Cys-287), residues responsible for substrate recognition only in HMSDH enzymes are conserved in ALDH8A1, namely Arg-109 and Arg-451 ( Fig. 1, generated with ESPript (24)).
A homology structure model of ALDH8A1 was built using the iTASSER server (25). An overlay of the homology model and pfAMSDH is shown in Fig. 2A. The model shows full coverage of the human enzyme sequence, which overlays well with the bacterial enzyme with a root mean square deviation of 1.15 Å for 472 C-␣ carbons. In addition, all of the catalytically essential active-site residues from pfAMSDH (Arg-109, Asn-155, Glu-253, Cys-287, and Arg-451 by ALDH8A1 numbering) are in the same location in the homology model as in the pfAMSDH structure (Fig. 2B).

ALDH8A1 can perform the NAD ؉ -dependent oxidation of 2-aminomuconic semialdehyde
The next question to arise was whether ALDH8A1 is able to catalyze the NAD ϩ -dependent oxidation of 2-AMS to 2-ami-nomuconate (2-AM). Pursuant to this end, an overexpression system was generated. The synthesized gene for human ALDH8A1 was ligated into pET-28a(ϩ) vector with a cleavable N-terminal His 6 tag and transformed into Escherichia coli BL21(DE3) competent cells. The expressed protein was purified by nickel affinity chromatography (Fig. 3A), and its ability to oxidize 2-AMS in a coupled-enzyme assay with ACMSD was tested. As shown in Fig. 3B, ACMS has a broad absorbance band at 360 nm. Upon ACMSD-catalyzed decarboxylation to 2-AMS, the absorbance maximum red-shifted to 380 nm and then decayed as 2-AMS was nonenzymatically converted to PA. The inclusion of purified ALDH8A1 and 1 mM NAD ϩ in an otherwise identical assay prevented the red shift, and instead, a broad absorbance band around 350 nm was observed that has been previously assigned as the oxidized product, 2-AM, and NADH ( Fig. 3C) according to the reaction scheme shown in Fig. 3D. These results resemble those observed from pfAMSDH and show that ALDH8A1 is able to rapidly oxidize 2-AMS in solution. The presence of the expected product in the coupled-enzyme assay was also verified by NMR spectroscopy. As shown in Fig. 4, the reaction mixture contained resonances consistent with 2-AM and expected cross-peaks in the 1 H-1 H COSY spectrum to show connectivity. Proton resonances were assigned based on similarity with 2-hydroxymuconic acid, which has been rigorously characterized (26).

Characterization of the reaction product of the ALDH8A1-catalyzed reaction
In initial studies of AMSDH, the identity of the product was inferred based on knowledge of the substrate structure and the catalytic cycle of the dehydrogenation reaction. The in crystallo characterization of the AMSDH reaction revealed that, in addition to oxidation of the aldehyde to its corresponding carboxylic acid, pfAMSDH also isomerizes the 2,3-double bond inside the active site prior to substrate oxidation (22). In the process of trying to determine the conformation of the product of the reaction catalyzed by ALDH8A1, we noticed that the product, 2-AM ( max 330 nm), is unstable and is nonenzymatically bleached in a single kinetic phase with a half-life of 67 min (Fig. 5A). High-resolution mass spectra were acquired of the 2-AM decay product after purification by HPLC. Mass spectra were collected in negative mode, and as shown in Fig. 5B, the parent ion matches tautomerized hydroxymuconate, 2-oxo-3-hexenedioate, with 2.55-ppm mass accuracy. Furthermore, fragment ions from cleaving at either side of the keto group can be observed with nominal masses of 113 and 85 Da.
The downstream enzyme of AMSDH performs deamination on 2-AM to produce 2-hydroxymuconate, which is expected to tautomerize to its ␣,␤-unsaturated ketone form, 2-oxo-3-hexenedioate, as shown in Scheme 2. The deamination reaction is not known to be coupled to any other reaction, so it is expected to be thermodynamically preferred. As such, it may proceed nonenzymatically at a slower rate. To lend credence to the proposed nonenzymatic deamination followed by tautomerization, ACMSD-AMSDH coupled-enzyme assays were performed in H 2 O and D 2 O in separate experiments, and the Extension of the mammalian kynurenine pathway reaction products were monitored by NMR spectroscopy (Fig. 6). When performing the reaction in H 2 O (Fig. 6A), two doublets around 5.8 ppm and a doublet of doublets at 7.15 ppm can be observed to decay while a new doublet at 6.17 ppm and a doublet of triplets at 6.9 ppm arise. These new resonances are consistent with 2-oxohexenedioate observed in the study of 2-hydroxymuconate tautomerization (26). Alternatively, upon enzymatic decarboxylation and oxidation performed in D 2 O by ACMSD and AMSDH, respectively, the 1 H NMR spectrum shows two doublets at 7.1 and 5.8 ppm, corresponding to the protons on carbons 4 and 5 (H b and H c ) of 2-AM, respectively (Fig. 6B). The proton at the 3-position (H a ) is replaced with deuterium by running the ACMSD reaction in D 2 O (Scheme 3). The two doublets coalesce into a single resonance at 6.9 ppm over time. The decay of the resonance at 5.8 ppm indicates that the proton at the 5-position can eventually exchange with solvent, and the shift of the doublet at 7.1 to a singlet at 6.9 ppm implies that a chemical change takes place in addition to simple exchanging of protons for deuterons. The most likely candidate for such a chemical change is the replacement of the nitrogen at

Extension of the mammalian kynurenine pathway
the 2-position with oxygen derived from water, i.e. deamination of 2-AM. Thus, the NMR spectra of the initial and final products of the ALDH8A1 reaction are consistent with 2-AM being the initially formed product, which then spontaneously deaminates and tautomerizes.

Determination of the kinetic parameters of ALDH8A1 and selected site-directed mutants
Although the coupled-enzyme assays presented above show that 2-AMS can serve as a substrate for ALDH8A1, such experiments are not amenable to extraction of enzymatic kinetic parameters. Instead, a substrate analog, 2-hydroxymuconate semialdehyde (2-HMS), in which the amino group of 2-AMS has been replaced with a hydroxyl group to prevent the nonenzymatic cyclization reaction, is used to determine kinetic parameters. The 1 H NMR spectrum of 2-HMS can be found in Fig. 7 along with corresponding 1D NOESY spectra, which show not only through-space interactions between protons but also in-phase resonances for the enol tautomer, which was previously implicated as an intermediate in the 2,3-bond isomerization reaction. The observation of resonances consistent with the enol form of 2-HMS in solution lends credence to the previously proposed tautomerization mechanism in pfAMSDH by showing that the enol form is energetically accessible. ALDH8A1 exhibits typical steady-state kinetics when acting on 2-HMS (Fig. 8A). The data were fitted with the Michaelis-Menten equation to provide a k cat , K m , and k cat /K m of 0.42 s Ϫ1 , 590 nM, and 7.1 ϫ 10 5 M Ϫ1 s Ϫ1 , respectively. A submicromolar K m is at the lower end for the KP enzymes; however, such high commitment may be necessary to efficiently compete with the rapid decay of its substrate to PA.
To further investigate the specificity of ALDH8A1 for ␣-substituted muconate semialdehydes, several active-site mutants were constructed. The two strictly conserved residues among the HMSDH family previously shown to be responsible for substrate recognition and binding, Arg-109, and Arg-451, were mutated to alanine, and their kinetic parameters for 2-HMS were determined. As summarized in Table 1, deletion of Arg-109 by mutation to alanine generated a variant with a similar turnover number but approximately 160-fold increased K m as compared with WT (Fig. 8B). No detectable activity could be measured for the R451A variant. Additionally, the active-site asparagine (Asn-155) responsible for stabilizing tetrahedral, oxyanion intermediates in general and involved in substrate isomerization in AMSDH was mutated to alanine, aspartic acid, and glutamine. The activity of the mutants was too low to determine kinetic parameters; however, specific activities for N155A, N155D, and N155Q were 7.4 Ϯ 0.1, 20 Ϯ 1, and 0.31 Ϯ

Extension of the mammalian kynurenine pathway
0.03 nmol/mg/min, respectively. Of the same mutants in pfAMSDH, N169D also showed the highest activity (22). In other aldehyde dehydrogenases, mutation of the corresponding asparagine to alanine or aspartic acid reduced the activity by approximately 1,000-fold or below detectable limits (27,28).

Conclusion
The human enzyme ALDH8A1 (ALDH12) was shown to catalyze the NAD ϩ -dependent oxidation of 2-AMS with catalytic efficiency comparable with pfAMSDH. Mutation of the activesite residues, which were shown to be heavily involved with substrate recognition, binding, and isomerization in the bacterial enzyme, resulted in variants with 100-fold increased K m or no detectable activity. As such, the ALDH8A1 enzyme, which was previously assigned as a cis-retinal dehydrogenase, should be reassigned as human AMSDH. It was also shown that the reaction product, 2-AM, can spontaneously deaminate in solution, ultimately forming 2-oxo-3-hexenedioate. This work thus establishes that the aldehyde dehydrogenase of the kynurenine pathway, first discovered 53 years ago from liver extracts (1), is ALDH8A1 (ALDH12). The kynurenine pathway of the tryptophan catabolic pathway in humans is therefore extended to AMSDH.

Cloning and site-directed mutagenesis
A DNA sequence that codes for human ALDH8A1 (accession number AF303134) was purchased from DNASU (Arizona State University) and ligated into pET28a(ϩ) vector with NheI and HindIII restriction sites, creating an N-terminal His 6tagged construct. The resultant plasmid was transformed into E. coli cell line BL21 (DE3), which was stored at Ϫ80°C as a 20% (v/v) glycerol stock. Overexpression systems for R109A, R451A, N155A, N155Q, and N155D were constructed by PCR overlap extension using the WT as the starting template. The forward primers were 5Ј-CCATGGACATTCCCGCGTCTGTGC-AGAA and 5Ј-CTGCTGGCTCATCGCGGAGCTGAACCTT for R109A and R451A, respectively, and 5Ј-GCTGGTCTGAT-CAGCCCCTGGGCTTTGCCACTCTACTTGCTGACC, 5Ј-GCTGGTCTGATCAGCCCCTGGCAGTTGCCACTCTAC-TTGCTGACC, and 5Ј-GCTGGTCTGATCAGCCCCTGGG-ACTTGCCACTCTACTTGCTGACC for N155A, N155Q, and N155D, respectively. Nucleotides corresponding to the mutated residues are underlined.

Protein preparation
For all cultures, antibiotic selection under kanamycin was used. Cultures were started by streaking the appropriate glycerol stock onto an LB-agar plate, which was incubated overnight at 37°C. A single colony was selected for further incubation in 15 ml of LB-Miller broth at 37°C with 220 rpm shaking until an A 600 of approximately 0.6 was achieved. Then 50 ml of LB-Miller broth was inoculated to an A 600 of 0.0002 and incubated at 37°C with 220 rpm shaking. Finally, once the 50-ml flask reached an A 600 of approximately 0.6, it was used to inoculate 6 liters of LB-Miller broth in 12 2-liter baffled flasks to an A 600 of 0.0002. The flasks were incubated at 37°C with 220 rpm shaking. Upon reaching an A 600 of 0.5, isopropyl ␤-D-1-thiogalactopyranoside was added to a final concentration of 800 M to induce protein expression, the temperature was lowered to

Extension of the mammalian kynurenine pathway
16°C, and the culture was incubated for an additional 12 h. Cells were harvested by centrifugation at 8,000 ϫ g and resuspended in 50 mM KP i , 150 mM NaCl buffered to pH 8.0 with 0.1% (v/v) ␤-mercaptoethanol. Protein was released by cell disruption (LS-20, Microfluidics), and the cell debris was removed by centrifugation at 27,000 ϫ g.
The protein of interest was purified by nickel affinity chromatography. Clarified, cell-free extract was applied to a nickelnitrilotriacetic acid column and eluted with an imidazole gradient. The running and elution buffers were 50 mM KP i , 150 mM NaCl buffered to pH 8.0 with 5 mM 1,4-DTT with the elution buffer also containing 500 mM imidazole. The purified protein was then desalted to 25 mM HEPES buffer, pH 7.5, 5 mM 1,4-DTT, 1 mM NAD ϩ , 5% glycerol (w/v); concentrated to approximately 1 mM by 30-kDa centrifugal filters; flash frozen in liquid nitrogen; and stored at Ϫ80°C until use. ACMSD used for coupled-enzyme assays was prepared as reported previously (11, 29 -33).
The coupled-enzyme assays were initiated by addition of excess ACMSD (1 M) to rapidly convert all ACMS to 2-AMS, which is in turn was converted to 2-aminomuconic acid by ALDH8A1. In the absence of ALDH8A1 and 1 mM NAD ϩ , the 2-AMS nonenzymatically decays to picolinic acid. Catalytic parameters were obtained using 2-HMS as the substrate. The decrease in absorbance as 2-HMS ( max at 375 nm, ⑀ 375 of 43,000 M Ϫ1 cm Ϫ1 ) and NAD ϩ are converted to 2-hydroxymuconic acid and NADH (⑀ 375 of 1,900 M Ϫ1 cm Ϫ1 ) was measured with an Agilent 8453 diode-array spectrometer. The reaction rate was calculated as the change in absorbance divided by the sum of the extinction coefficients of 2-HMS and NADH. Initial rates versus substrate concentration were fitted with the Michaelis-Menten equation.
Nonlinear least squares regression was performed with Orig-inPro 8.5.

NMR spectroscopy
All NMR spectra were recorded on a Bruker (Billerica, MA) Avance III HD 500-MHz spectrometer equipped with a Cryo-Prodigy Probe at 300 K running TopSpin 3.5pl6. Spectra were recorded in D 2 O or 90% H 2 O, 10% D 2 O and referenced to residual solvent ( 1 H, 4.70 ppm). One-dimensional 1 H spectra (pulse sequence, zg30) were recorded with 1-s relaxation delays, 65,536 data points, and multiplied with an exponential function for a line broadening of 0.3 Hz before Fourier transformation. Double quantum-filtered COSY (pulse sequence, cosygpmfppqf) spectra were acquired with spectral widths of 13.0 ppm with 2,048 ϫ 128 data points and a relaxation delay of 2.0 s. 1D gradient-selected NOESY (pulse sequence, selnogp) spectra Scheme 2. Proposed spontaneous decay mechanism for 2-aminomuconate.

Extension of the mammalian kynurenine pathway
were recorded with a mixing time of 0.3 s and a 2-s relaxation delay and multiplied with an exponential function for a line broadening of 3 Hz before Fourier transformation. All NMR data were processed using MestReNova NMR version 11.0.3 software.

Mass spectrometry
The AMSDH reaction product, 2-aminomuconate, was isolated for mass spectrometric characterization by reverse-phase HPLC with an InertSustain C 18 column (5-m particle size, 4.6-mm inner diameter ϫ 100 mm; GL Sciences Inc.) on a Dionex Ultimate 3000 HPLC equipped with a diode-array detector (Sunnyvale, CA). The crude reaction mixture was ultrafiltered (10-kDa cutoff) to remove protein, and separation was achieved using isocratic elution with 95% H 2 O, 5% acetonitrile, and 0.5% formic acid. Mass spectra were collected on a maXis plus quadrupole-TOF mass spectrometer equipped with an electrospray ionization source (Bruker Daltonics). The instrument was operated in the negative ionization mode in the range 50 Յ m/z Յ 1,500 and calibrated using ESI-L Low Concentration Tuning Mix (Agilent Technologies). Samples were introduced via syringe pump at a constant flow rate of 3 l/min. Relevant source parameters are summarized as follows: capillary voltage, 3500 V with a set end plate offset of Ϫ500 V; nebulizer gas pressure, 0.4 bar; dry gas flow rate, 4.0 liter/min; source temperature, 200°C. Mass spectra were averages of 1 min of scans collected at a rate of 1 scan per second. Collision-

Extension of the mammalian kynurenine pathway
induced dissociation was achieved using a set collision energy of Ϫ20 eV. OtofControl software version 6.3 was used for data acquisition, and Compass Data Analysis software version 4.3 (Bruker Daltonics) was used to process all mass spectra. mMass software version 5.5.0 was used for all exact mass calculations (34).