Abstract

The biologically important carnitine biosynthesis pathway in humans proceeds via four enzymatic steps. The first step in carnitine biosynthesis is catalyzed by trimethyllysine hydroxylase (TMLH), a non-heme Fe(II) and 2-oxoglutarate (2OG)-dependent oxygenase, which catalyzes the stereospecific hydroxylation of (2S)-Nε-trimethyllysine to (2S,3S)-3-hydroxy-Nε-trimethyllysine. Here, we report biocatalytic studies on human TMLH and its 19 variants introduced through site-directed mutagenesis. Amino acid substitutions at the sites involved in binding of the Fe(II) cofactor, 2OG cosubstrate and (2S)-Nε-trimethyllysine substrate provide a basic insight into the binding requirements that determine an efficient TMLH-catalyzed conversion of (2S)-Nε-trimethyllysine to (2S,3S)-3-hydroxy-Nε-trimethyllysine. This work demonstrates the importance of the recognition sites that contribute to the enzymatic activity of TMLH: the Fe(II)-binding H242–D244–H389 residues, R391–R398 involved in 2OG binding and several residues (D231, N334 and the aromatic cage comprised of W221, Y217 and Y234) associated with binding of (2S)-Nε-trimethyllysine.

Introduction

Carnitine (l-3-hydroxy-4-N,N,N-trimethylaminobutyrate) is an important metabolite that facilitates the transport of long fatty acids from the cytosol into mitochondria in humans, other eukaryotes, many plants and microorganisms [13]. Humans obtain carnitine via diet (e.g. meat) or endogenous biosynthesis [46]. The carnitine biosynthesis pathway proceeds via four enzymatic steps from (2S)-Nε-trimethyllysine to l-carnitine (Figure 1). The first and the last steps are catalyzed by trimethyllysine hydroxylase (TMLH) and γ-butyrobetaine hydroxylase (BBOX), members of non-heme Fe(II) and 2-oxoglutarate (2OG)-dependent oxygenases [3,79]. The second step of carnitine biosynthesis is catalyzed by 3-hydroxy-Nε-trimethyllysine (HTML) aldolase, an enzyme that has not yet been identified in humans, whereas the third step is mediated by NAD+ dependent 4-trimethylaminobutyraldehyde (TMABA) dehydrogenase (ALDH9 in humans) [3,10]. Targeting enzymes involved in the carnitine biosynthesis pathway has a potential in drug discovery as a strategy for therapeutic intervention in cardiovascular diseases [11]. A biomedical relevance of carnitine biosynthesis has been exemplified by the development of 3-(2,2,2-trimethylhydrazine)propionate (also known as Mildronate or Meldonium), a substrate-competitive highly selective inhibitor of BBOX, which is used for the treatment of myocardial infarction [12]. Recent work also proposed that a dysregulation of the carnitine biosynthesis pathway via TMLH deficiency or mutations is linked with an increased risk for autism [1315].

The carnitine biosynthesis pathway.

Human TMLH catalyzes the stereospecific C-3 hydroxylation of (2S)-Nε-trimethyllysine to (2S,3S)-3-hydroxy-Nε-trimethyllysine in cells and in vitro [1620]. Mechanistic investigations confirmed high dependence of TMLH activity on the presence of a ferrous ion, 2OG cosubstrate, oxygen and ascorbate, a signature known for several other members of Fe(II)/2OG-dependent oxygenases [9,16,17]. Studies using recombinantly expressed human TMLH have shown that the enzyme also efficiently catalyzes C-3 hydroxylation of simplest (2S)-Nε-trimethyllysine analogs that differ in the length of the chain and the bulkiness of the quaternary ammonium ion [16,21]. The examination of structure–activity relationships suggested key structural elements of (2S)-Nε-trimethyllysine required for its ability to act as an excellent TMLH substrate; the presence of the quaternary trimethylammonium cation (NMe3+), the l-stereochemistry at C-2, the negatively charged C-1 carboxylate (COO) and the positively charged α-ammonium group (NH3+) being all essential [16].

Human TMLH and BBOX possess high substrate specificity and exhibit 29% sequence identity, which indicates significant homology, with most of the postulated recognition site residues being conserved (Figure 2) [8,22]. Comparison of protein sequences suggests that conserved residues include (i) the catalytic ferrous ion binding His–Asp–His triad, (ii) two Arg residues involved in binding of C-5 carboxylate of 2OG cosubstrate and (iii) the aromatic cage, comprised of Trp–Tyr residues, involved in the association with the positively charged quaternary trimethylammonium ion (Figure 2). Although the BBOX crystal structure has been determined in the presence and absence of γ-butyrobetaine (γBB) substrate [8,23], the structural information regarding TMLH is lacking, hence leading to an insufficient understanding of basic requirements needed for TMLH enzymatic activity.

Sequence alignment of human TMLH and BBOX.

Figure 2.
Sequence alignment of human TMLH and BBOX.

Residues involved in (2S)-Nε-trimethyllysine substrate binding, 2OG cosubstrate binding and ferrous ion chelation are indicated as green stars, blue triangles and red circles, respectively.

Figure 2.
Sequence alignment of human TMLH and BBOX.

Residues involved in (2S)-Nε-trimethyllysine substrate binding, 2OG cosubstrate binding and ferrous ion chelation are indicated as green stars, blue triangles and red circles, respectively.

Here, we report enzymatic and modeling studies on human TMLH and its recognition site variants that unravel the importance of amino acids involved in binding of Fe(II) cofactor, 2OG cosubstrate and (2S)-Nε-trimethyllysine substrate. Based on the crystal structure of the BBOX–γBB complex, it was suggested, but not experimentally proven, that certain residues in TMLH might contribute to its enzymatic activity [8]. The objective of this work is to mutate the proposed TMLH residues involved in binding of Fe(II) cofactor, 2OG cosubstrate and (2S)-Nε-trimethyllysine substrate, and examine such TMLH variants for enzymatic hydroxylation of (2S)-Nε-trimethyllysine. Three sets of TMLH variants were introduced, namely those that target binding of Fe(II), 2OG and (2S)-Nε-trimethyllysine, respectively: (i) Fe(II) cofactor is presumably chelated by His242, Asp244 and His389; (ii) 2OG cosubstrate possibly interacts with Arg391, Arg398 and Thr269; (iii) (2S)-Nε-trimethyllysine substrate was proposed to associate with the aromatic cage comprised of Tyr217, Trp221, Tyr234, while neighboring Tyr404 fixates Asp244 carboxylate; and (iv) the α-amino group and the C-1 carboxylate of Nε-trimethyllysine substrate also presumably interact with Asp231 and Asn334, respectively [8]. Notably, a single amino acid substitution (negatively charged Asp231 in TMLH, neutral Asn191 in BBOX) likely contributes to the specificity of TMLH over BBOX, presumably via the formation of an energetically strong salt bridge with the positively charged α-ammonium group of (2S)-Nε-trimethyllysine substrate [8].

Materials and methods

Site-directed mutagenesis of TMLH

Mutagenic primers were designed using PrimerX program and obtained from Biolegio (Nijmegen, The Netherlands). Sequences of primers are listed in Supplementary Table S1. pETDt_MalE_TMLH vector including the MBP-encoding gene MalE (N-terminally located 6×His tag and C-terminally located rTEV protease cleavage site) and TMLHa gene (lacked the N-terminal mitochondrial targeting sequence and hydrophobic sequences, EC 1.14.11.8) was used as before [22]. Mutants of TMLH were obtained from pEDTt_MalE_TMLH using a QuickChange II XL site-directed mutagenesis kit (Agilent). The cycling parameters used were as follow: 95°C (1 min), followed by 18 cycles of 95°C (50 s), 60°C (50 s) and 68°C (8 min), followed by a final elongation step at 68°C (8 min). After amplification, the reaction mixtures were incubated with Dpn I restriction enzyme at 37°C for 1 h to digest parental supercoiled dsDNA. The reaction products were subsequently transformed into E. coli XL 10-Gold ultracompetent cells. Briefly, 2 µl of the Dpn I treated product was transferred to a 50 µl aliquot of the ultracompetent cells. The reaction was incubated on ice for 30 min and heat-shocked at 42°C water bath for 45 s, followed by incubation on ice for 2 min. To this mixture, 0.8 ml of NZY+ broth was added and the mixture incubated for 1 h at 37°C with shaking at 200 rpm. The cells were spread on LB-ampicillin agar plates and then incubated at 37°C overnight. Then, 5 ml of LB-ampicillin medium was inoculated with a single bacterial colony and cultivated overnight at 37°C. Mutant plasmid DNA was purified using a Strataprep plasmid miniprep kit (Agilent). The desired mutation in the vector was verified by sequence analysis (BaseClear, Leiden, The Netherlands).

Expression and purification of TMLH variants

pGro7 vector was transformed into BL21-AI competent cells. A positive colony was selected to make BL21-AI competent cells containing pGro7 vector. In brief, 100 ml of LB medium (34 µg/ml chloramphenicol) containing the colony was inoculated until the OD600 reached 0.4–0.55. The culture was cooled on ice for 15 min and the cells were centrifuged (2700 g) for 15 min at 4°C. The cells were resuspended in 33 ml of RF1 buffer (100 mM RbCl, 50 mM MnCl2, 30 mM KOAc, 10 mM CaCl2, 15% Glycerol, at final pH 5.8). Cells were then incubated on ice for 15 min, followed by centrifugation (600 g) for 15 min at 4°C and then resuspended in 4 ml of RF2 buffer (10 mM MOPS, 10 mM RbCl, 75 mM CaCl2, 15% Glycerol, at final pH 6.5). Cells were incubated for another 15 min, then aliquoted and stored at −80°C. Mutant TMLH vectors were transformed into BL21-AI competent cells containing pGro7 vector.

Expression and purification of the MBP-TMLHa fusion construct and its 19 mutants were performed according to the reported method [16,22]. Wild-type (WT) and variant TMLH enzymes were expressed in Escherichia coli BL21-AI cells in LB growth medium supplemented with ampicillin and chloramphenicol at 37°C. Cells were grown until an OD600 of 0.5–0.6. The cells were then induced with 0.1 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) and 1% l-Arabinose and cultured for 4 h at 37°C. Cells were then harvested and lysed by sonication. MBP-TMLHa was purified by Ni-NTA affinity chromatography and concentrated using a 30 kDa MWCO filter (Amicon). Further purification was carried out using size-exclusion chromatography (SEC) using a Superdex-200 preparative grade column on an AKTA system, using 20 mM K2HPO4 (pH 7.0), 20 mM KCl, 2 mM DTT as the mobile phase. The WT TMLH and its 19 variants were monitored by SDS–PAGE on a 4–15% gradient polyacrylamide gel (Bio-Rad) and the concentrations were determined using the Nanodrop DeNovix DS-11 spectrophotometer. Fractions containing the enzyme were pooled and concentrated, aliquoted, flash-frozen in liquid nitrogen and stored at −80°C.

LC–MS experiments

LC–MS enzymatic experiments for TMLH and its 19 variants were carried out in phosphate buffer (20 mM) and KCl (20 mM) and DTT (2 mM) at pH 7.5. To a premixed solution of TMLH (3 µM), FeSO4 (500 µM), 2OG (2.5 mM) and ascorbate (5 mM), was added trimethyllysine (500 µM). Samples were incubated in an 1.5 ml Eppendorf vial for 30 min at 37°C in a Thermomixer (Eppendorf) at 600 rpm. The reaction mixture (100 µl) was then quenched with methanol (100 µl). The sample was analyzed by LC–MS on a Thermo Finnigan LCQ-Fleet ESI-ion trap (Thermo Fisher Scientific, Breda, the Netherlands). Analysis was achieved on Phenomenex Gemini-NX C18 column, 50 × 2.0 mm, particle size 3 µM (Phenomenex, Utrecht, The Netherlands). The mobile phase contained 0.1% formic acid in water and 0.1% formic acid in acetonitrile. The gradient process over 50 min with a flow of 0.2 ml/min applied 5–100% of acetonitrile with 0.1% formic acid. The percentage of hydroxylation was derived from the intensities of both peaks in LC–MS.

NMR experiments

Nuclear magnetic resonance (NMR) enzymatic experiments were conducted at 310 K in 20 mM Tris-D11·DCl (pD 7.5). To a premixed solution of TMLH/TMLH variant (3 or 10 µM), FeSO4 (100 µM), 2OG (2 mM) and ascorbate (500 µM) was added (2S)-Nε-trimethyllysine (500 µM). After shaking for 30 min at 37°C in an Eppendorf vial, the reaction mixture (500 µl) was quenched with 1 M DCl (5 µl), transferred into the NMR tube and recorded by 1H NMR. NMR spectra were recorded on a Bruker Avance III spectrometer paired with a 500 MHz magnet equipped with a Prodigy cryoprobe. 1H 1D spectra were acquired using presaturation to suppress the water signal with 64 or 128 transients and a relaxation delay of 12 s. 2D 1H–13C multiplicity-edited HSQC spectra were acquired using 1k points per transient, 16 transients per increment, a relaxation delay of 2 s and 256 increments. All experiments were carried out in replicates.

NMR kinetic assays using trimethyllysine

NMR kinetic experiments were performed by incubation of TMLH/TMLH variant (3 µM) with FeSO4 (100 µM), 2OG (2 mM), ascorbate (500 µM) and varying concentrations of (2S)-Nε-trimethyllysine (250 µM–2 mM) at 310 K in 20 mM Tris-D11·DCl (pD 7.5). After 10 min incubation, the enzymatic mixture was quenched with 1 M DCl (5 µl) and the reaction mixture (500 µl) was transferred into the NMR tube and recorded by 1H NMR.

NMR kinetic assays using 2OG

NMR kinetic experiments were performed by incubation of TMLH/TMLH variant (3 µM) with FeSO4 (100 µM), (2S)-Nε-trimethyllysine (2 mM), ascorbate (500 µM) and varying concentrations of 2OG (100 µM–5 mM) at 310 K in 20 mM Tris-D11·DCl (pD 7.5). After 10 min incubation, the enzymatic mixture was quenched with 1 M DCl (5 µl) and the reaction mixture (500 µl) was transferred into the NMR tube and recorded by 1H NMR.

Modeling of TMLH

The sequences of human TMLH and human BBOX show 29% identity over a length of 354 amino acids. These numbers are in general considered to fall in the homology modeling ‘twilight zone’, where modeling is not impossible, but the correct identification of a template and the creation of a correct alignment require more sophisticated techniques. However, the high local conservation of the active site residues allowed us to build a model. The standard homology modeling script in the YASARA Structure (PMID: 11948792 and 29086303) & WHAT IF (PMID: 2268628) Twinset was used to create models in the presence and absence of the ligands. We generated a model of human TMLH in the presence and absence of (2S)-Nε-trimethyllysine using the PDB file 4C5W (PubMed: 24571165) as a template. A model without substrates was created by removing the ligands γBB and NOG (N-oxalylglycine) from the template. A second model containing all the relevant ligands was created by replacing the γBB molecule in the template with a (2S)-Nε-trimethyllysine molecule obtained from PDB file 2H23 (PubMed: 16682405). Further detailed analysis of both models was done with YASARA.

Results and discussion

Based on the comparison of TMLH and BBOX protein sequences, we identified TMLH amino acid residues likely involved in binding of Fe(II) cofactor, 2OG cosubstrate and (2S)-Nε-trimethyllysine substrate. We employed QuikChange site-directed mutagenesis to introduce single amino acid substitutions, and we purified TMLH variants by Ni-NTA exchange and SEC (Supplementary Figures S1–S3 and Table S1). The enzymatic activity of recombinantly expressed human TMLH variants was examined by mass spectrometry (MS) and NMR-based assays (Supplementary Figures S4–S20). Initially, the WT TMLH was examined for enzymatic hydroxylation as a reference. In NMR spectrum, almost quantitative conversion (94%) of (2S)-Nε-trimethyllysine (500 µM) to (2S,3S)-3-hydroxy-Nε-trimethyllysine was observed in the presence of TMLH (3 µM), FeSO4 (100 µM), 2OG (2 mM) and ascorbate (500 µM) after 30 min at 37°C (Supplementary Figure S4); this result is in agreement with our recent findings [16,20]. NMR data also clearly indicated that TMLH-catalyzed hydroxylation is tightly coupled to the conversion of 2OG cosubstrate to succinate. Full conversion of (2S)-Nε-trimethyllysine (500 µM) to (2S,3S)-3-hydroxy-Nε-trimethyllysine was observed in the presence of an increased amount (10 µM) of TMLH under standard conditions (Figure 3a and Supplementary Figure S4). NMR-based kinetic analyses revealed that KM values for (2S)-Nε-trimethyllysine and 2OG are 1.18 mM and 420 µM, respectively (Figure 3b, Table 1 and Supplementary Figure S21). These results are in line with KM values of purified rat TMLH [24].

Enzymatic activity of human TMLH and its variants.

Figure 3.
Enzymatic activity of human TMLH and its variants.

(a) (2S)-Nε-trimethyllysine hydroxylation by WT TMLH and its variants as determined by NMR spectroscopy. (b) Kinetics analysis of WT TMLH and W221F, T269A and D244E variants, varying the concentration of (2S)-Nε-trimethyllysine substrate (left) and 2OG cosubstrate (right).

Figure 3.
Enzymatic activity of human TMLH and its variants.

(a) (2S)-Nε-trimethyllysine hydroxylation by WT TMLH and its variants as determined by NMR spectroscopy. (b) Kinetics analysis of WT TMLH and W221F, T269A and D244E variants, varying the concentration of (2S)-Nε-trimethyllysine substrate (left) and 2OG cosubstrate (right).

Table 1
Kinetics parameters for conversion of (2S)-Nε-trimethyllysine to (2S,3S)-3-hydroxy-Nε-trimethyllysine by TMLH and its variants
(2S)-Nε-trimethyllysine Vmax (µM min−1KM (mM) 
WT TMLH 147 ± 13 1.18 ± 0.13 
W221F 77 ± 9.0 1.47 ± 0.31 
T269A 72 ± 4.0 0.70 ± 0.09 
D244E 12 ± 0.75 0.32 ± 0.07 
2OG     
 WT TMLH 114 ± 9 0.42 ± 0.12 
 W221F 35 ± 3 0.35 ± 0.11 
 T269A 122 ± 20 1.97 ± 0.70 
 D244E 11 ± 0.6 1.12 ± 0.16 
(2S)-Nε-trimethyllysine Vmax (µM min−1KM (mM) 
WT TMLH 147 ± 13 1.18 ± 0.13 
W221F 77 ± 9.0 1.47 ± 0.31 
T269A 72 ± 4.0 0.70 ± 0.09 
D244E 12 ± 0.75 0.32 ± 0.07 
2OG     
 WT TMLH 114 ± 9 0.42 ± 0.12 
 W221F 35 ± 3 0.35 ± 0.11 
 T269A 122 ± 20 1.97 ± 0.70 
 D244E 11 ± 0.6 1.12 ± 0.16 

The catalytic triad His–Asp/Glu–His involved in the chelation of the active site ferrous ion is highly conserved among the superfamily of Fe(II)/2OG oxygenases that catalyze a biologically diverse set of chemical reactions [7,9,25,26]. Mutations of these conserved residues typically lead to inactive enzymes due to insufficient ability to bind a ferrous ion required for activation of molecular oxygen. Comparison of TMLH sequences from various species (eukaryotes and prokaryotes) also verifies the highly conserved nature of the His–Asp–His catalytic triad in these species [27]. Our enzyme activity studies showed that the His242Ala, Asp244Ala and His389Ala variants of TMLH (at 3 µM and 10 µM) typically do not catalyze C-3 hydroxylation of (2S)-Nε-trimethyllysine within limits of detection (Figure 3a and Supplementary Figures S6–S9). NMR studies, moreover, indicated that neither of these three variants facilitates the uncoupled turnover of 2OG to succinate. The result with His389Ala is largely in line with the lack of enzymatic activity of His389Leu in cellular assays [27]. Although the His–Asp/Glu–His triad appears to be essential for catalytic activities of numerous Fe(II)/2OG oxygenases, all three residues are not always mandatory. For instance, mutation studies on Factor Inhibiting HIF (FIH) showed that FIH only requires two histidines, not needing the entire His–Glu–His triad found in the WT FIH, for efficient iron chelation and enzymatic activity [28]. In addition, we substituted the iron-chelating Asp244 by Glu244, which contains a longer side chain, in TMLH. The Asp244Glu variant displayed a significant enzymatic activity (80% at 10 µM, 22% at 3 µM), implying that the mutated His–Glu–His triad in TMLH possesses a good affinity for binding a ferrous ion (Figure 3a and Supplementary Figure S10). C-3 hydroxylation was also observed to be tightly coupled to the conversion of 2OG to succinate by the Asp244Glu variant (Supplementary Figure S8). Moreover, the KM values obtained for Asp244Glu for acting on (2S)-Nε-trimethyllysine and 2OG were found to be 325 µM and 1.12 mM, respectively (Figure 3b and Table 1). It is noteworthy that members of Fe(II)/2OG dependent enzymes also have the ability to catalyze halogenation of non-activated C-H bonds; in such halogenases, Asp/Glu residues of the catalytic triad are replaced by smaller nonchelating Ala or Gly residues, thus providing space for binding of the halogen ion to the ferrous ion [7,26,29] The Asp244Ala variant of TMLH was examined for potential C-3 chlorination of (2S)-Nε-trimethyllysine; we detected neither any chlorinated product nor conversion of 2OG to succinate. Taken together, our mutagenesis studies on the iron-chelating His–Asp–His residues revealed that human TMLH requires a complete catalytic triad for efficient catalytic activity; however, it exhibits some flexibility with respect to the presence of Asp, as reflected by the observable enzymatic activity of the Asp244Glu variant.

Next, we explored residues potentially involved in binding of the 2OG cosubstrate. Structural and mechanistic work on Fe(II)/2OG oxygenases revealed that the negatively charged C-5 carboxylate of 2OG forms a salt bridge with conserved positively charged Arg/Lys residues located deep inside the active site [7,9,30]. For example, 2OG binding is mediated via arginine (e.g. TMLH, BBOX, PHD2) or lysine residues (e.g. JMJD2A, JMJD6, FIH). Our enzyme activity studies showed that the Arg398Ala variant of TMLH is inactive, implying that the abrogation of the 2OG binding leads to the lack of enzymatic activity (Figure 3a). The Arg398Lys variant, which in principle possesses the ability to associate with the C-5 carboxylate of 2OG, displayed significantly lower enzymatic activity (35%) when compared with WT TMLH (Figure 3a and Supplementary Figure S11). This result implies that binding of 2OG to Arg398Lys is not optimal, thus leading to a less active enzyme; similar observations were found in mutagenesis studies of human PHD2, a key cellular oxygen sensor [31]. Based on the BBOX structure and homology, another arginine residue (Arg391) and adjacent threonine residue (Thr269) in human TMLH were postulated to interact with 2OG via a salt bridge and an H-bond, respectively. The enzymatic assays with the Arg391Ala variant showed a lack of activity within limits of detection, suggesting that the guanidinium cation of Arg391 importantly contributes to the association with the C-5 carboxylate of 2OG (Figure 3a). In contrast, the Thr269Ala variant displayed high enzymatic activity (96% at 10 µM, 76% at 3 µM) for the conversion of (2S)-Nε-trimethyllysine to (2S,3S)-3-hydroxy-Nε-trimethyllysine; this result suggests that Thr269 does not play an important role in the binding of 2OG (Figure 3a and Supplementary Figure S12). Obtained KM values for (2S)-Nε-trimethyllysine and 2OG are 700 µM and 1.97 mM (Figure 3b and Table 1). Overall, our examinations of the 2OG-binding residues provide strong support for direct involvement of Arg391 and Arg398 residues in the association with the C-5 carboxylate of 2OG, the interactions that eventually contribute to efficient TMLH activity.

Having confirmed the importance of amino acids involved in binding of the Fe(II) cofactor and 2OG cosubstrate, we then explored TMLH residues potentially involved in the association with the (2S)-Nε-trimethyllysine substrate. Based on the crystal structure of the related BBOX-γBB complex, we hypothesized that the specific recognition of (2S)-Nε-trimethyllysine is driven by the energetically favorable cation–π interactions between the quaternary trimethylammonium cation with the aromatic cage of TMLH, comprised of electron-rich side chains of Trp221, Tyr217 and Tyr234. Substitutions of individual amino acids by small, nonaromatic alanine, i.e. Trp221Ala, Tyr217Ala and Tyr234Ala, led to inactive TMLH variants, highlighting the required presence of all three aromatic residues that constitute the aromatic cage of TMLH (Figure 3a). Replacement of larger Trp221 by somewhat smaller but still aromatic Phe showed that the Trp221Phe variant displayed high enzymatic activity (94% at 10 µM, 65% at 3 µM) (Figure 3a); this engineered aromatic cage well resembles the aromatic cage of BBOX from Pseudomonas sp. that contributes to efficient binding of γBB substrate [32]. More detailed 2D NMR analyses ultimately confirmed the formation of (2S,3S)-3-hydroxy-Nε-trimethyllysine by the Trp221Phe variant of TMLH (Supplementary Figure S13). Additional kinetic analyses on Trp221Phe showed that the KM value for (2S)-Nε-trimethyllysine is 1.47 mM, whereas the KM value for 2OG is 348 µM (Figure 3b and Table 1). We also examined the potential involvement of the phenolic OH group of Tyr404 in TMLH catalysis. The Tyr404Phe variant displayed considerably reduced enzymatic activity (32%), implying that the OH group of Tyr404 contributes to a stronger interaction with the trimethylammonium group of (2S)-Nε-trimethyllysine, presumably via energetically favorable electrostatic interactions (Figure 3a and Supplementary Figure S14).

The readout process of histones that contain posttranslationally modified trimethyllysine is mediated by epigenetic reader proteins that possess aromatic cages [33]. Most of such aromatic cages are comprised of 2–4 aromatic amino acids, whereas there are examples of readers that contain the negatively charged Asp or Glu residues instead of aromatic residues, in particular for the predominant recognition of dimethyllysine and methyllysine [33]. To further explore the composition of the aromatic cage on TMLH activity, we substituted aromatic Trp221 and Tyr217 by negatively charged Asp/Glu residues. The Trp221Glu variant was observed to be inactive for the C-3 hydroxylation of (2S)-Nε-trimethyllysine, whereas the succinate product was also not observed (Figure 3a). In line with this result, the Tyr217Asp and Tyr217Glu variants also did not display any enzymatic activity within limits of detection (Figure 3a). Although these three engineered TMLH variants could in principle recognize lower methylation states of lysine (in analogy with epigenetic recognition), we observed that they do not catalyze C-3 hydroxylation of (2S)-Nε-dimethyllysine and (2S)-Nε-methyllysine nor mediate conversion of 2OG to succinate (Supplementary Figures S15–S17).

Recent structure–activity relationship studies revealed that both the α-ammonium group (NH3+) and C-1 carboxylate (COO) moieties of (2S)-Nε-trimethyllysine play essential roles in TMLH binding and productive catalysis. It was postulated that the positively charged α-ammonium group forms an energetically favorable salt bridge with the negatively charged Asp231 of TMLH [8]. To test this proposal, we produced the Asp231Ala and Asp231Asn variants. Importantly, the Asp231Ala variant completely lost the enzymatic activity (Figure 3a and Supplementary Figures S18–S19). The Asp231Asn variant of TMLH well resembles the BBOX WT (Asn191 is in place of Asp231); our assays showed a significant reduction in enzymatic activity (28% remaining activity) for Asp231Asn (Figure 3a and Supplementary Figure S20). Due to the role of Asn191 residue in γBB recognition by BBOX, we also examined the analogous Asp231Asn variant of TMLH for potential hydroxylation of γBB to yield l-carnitine. We did not observe the formation of l-carnitine, implying that differences in substrate specificities between human TMLH and BBOX go beyond a single residue (i.e. Asp231/Asn191) alteration. Structural and bioinformatics analyses also suggested that Asn334 (in TMLH) and Asn292 (in BBOX) interact with the negatively charged C-1 carboxylate via H-bond. Mutating asparagine to alanine led to very poorly active Asn334Ala variant of TMLH (10% at 10 µM, Figure 3a), thus verifying the involvement of the side chain of Asn334 in direct interaction with the C-1 carboxylate of (2S)-Nε-trimethyllysine substrate. Collectively, the examinations of amino acids that presumably constitute the recognition site for (2S)-Nε-trimethyllysine binding show that human TMLH requires the aromatic cage comprised of three electron-rich aromatic residues, the negatively charged Asp231 and the neutral Asn334 for efficient binding and productive enzymatic conversion of (2S)-Nε-trimethyllysine to (2S,3S)-3-hydroxy-Nε-trimethyllysine.

Following the mutagenesis studies, we attempted to rationalize the observed degrees of enzymatic activities of TMLH variants by generating a structural model of human TMLH in complex with Zn(II) (as an inactive substitute for Fe(II)), N-oxalylglycine (NOG, as an inactive substitute for 2OG) and (2S)-Nε-trimethyllysine (Figure 4). Since no experimentally solved structure for human TMLH exists, we aimed to build a homology model of this protein based on the known BBOX structure [23]. The overall fold of the modeled TMLH appears to be similar to the one of the crystallographically determined BBOX (Supplementary Figure S22). A detailed analysis of the active site of the structure of BBOX and the model of TMLH reveals that the majority of important residues mentioned above are conserved (Figure 5). This includes the TMLH's metal-binding His242–Asp244–His389 triad, the residues surrounding the 2OG (Arg391 and Arg398), and the aromatic residues Trp221, Tyr217, Tyr234 and Tyr404 (Figure 5). Small changes and displacements in side chain positions have still occurred, as is to be expected in such a modeling process. Notably, the negatively charged Asp231 of TMLH is located in close proximity to the positively charged α-ammonium cation of the substrate, implying the existence of an energetically favorable salt bridge. Moreover, our TMLH model shows that the side chain of Asn334 is located within the van der Waals distance to the C-1 carboxylate of the substrate, indicating direct interaction via H-bond. Taken together, the TMLH model provides a valuable support of our experimentally observed enzymatic activities of TMLH variants.

The TMLH model.

Figure 4.
The TMLH model.

View on the active site of the modeled human TMLH (gray) in the presence of zinc ion (pink sphere), N-oxalylglycine (NOG, green) and (2S)-Nε-trimethyllysine (Kme3, yellow).

Figure 4.
The TMLH model.

View on the active site of the modeled human TMLH (gray) in the presence of zinc ion (pink sphere), N-oxalylglycine (NOG, green) and (2S)-Nε-trimethyllysine (Kme3, yellow).

View on the active sites of the overlapped structures of human TMLH (light gray) and human BBOX (dark gray) with labeled protein residues.

Figure 5.
View on the active sites of the overlapped structures of human TMLH (light gray) and human BBOX (dark gray) with labeled protein residues.

TMLH (light gray), NOG (green) and (2S)-Nε-trimethyllysine (yellow). BBOX (dark gray), NOG (blue) and γ-butyrobetaine (orange). Zinc ion (pink).

Figure 5.
View on the active sites of the overlapped structures of human TMLH (light gray) and human BBOX (dark gray) with labeled protein residues.

TMLH (light gray), NOG (green) and (2S)-Nε-trimethyllysine (yellow). BBOX (dark gray), NOG (blue) and γ-butyrobetaine (orange). Zinc ion (pink).

In conclusion, our biocatalytic studies of TMLH variants reveal the essential roles of individual active site residues that contribute to productive TMLH catalysis for the conversion of (2S)-Nε-trimethyllysine to (2S,3S)-3-hydroxy-Nε-trimethyllysine. The work demonstrates the importance of (i) the His242–Asp244–His389 triad for chelation of the active site ferrous ion, (ii) Arg391 and Arg398 residues involved in binding with the 2OG cosubstrate and (iii) the aromatic cage comprising of Trp221, Tyr217 and Tyr234 for the association with the positively charged trimethylammonium group of (2S)-Nε-trimethyllysine, and Asp231 and Asn334 residues that form a salt bridge and a H-bond with the backbone of (2S)-Nε-trimethyllysine substrate. Along with results from the examinations of the simplest (2S)-Nε-trimethyllysine analogs as substrates for human TMLH [16], this study defines molecular requirements for associations between TMLH and its cofactor, cosubstrate and substrate. It is envisioned that the understanding of the basic mechanisms of enzymes involved in the biomedically important carnitine biosynthesis will guide the rational drug design aimed at targeting the human carnitine biosynthesis pathway. Towards this aim, our enzymatic and modeling work characterizes TMLH residues that associate with the Fe(II) cofactor, 2OG cosubstrate and (2S)-Nε-trimethyllysine substrate, and hence essentially contribute to efficient enzymatic catalysis.

Abbreviations

     
  • 2OG

    2-oxoglutarate

  •  
  • BBOX

    γ-butyrobetaine hydroxylase

  •  
  • FIH

    Factor inhibiting HIF

  •  
  • HTML

    3-hydroxy-Nε-trimethyllysine

  •  
  • IPTG

    isopropyl β-d-1-thiogalactopyranoside

  •  
  • LC–MS

    liquid chromatography–mass spectrometry

  •  
  • MS

    mass spectrometry

  •  
  • NMR

    nuclear magnetic resonance

  •  
  • NOG

    N-oxalylglycine

  •  
  • SEC

    size-exclusion chromatography

  •  
  • TMABA

    4-trimethylaminobutyraldehyde

  •  
  • TMLH

    trimethyllysine hydroxylase

  •  
  • WT

    wild-type

  •  
  • γBB

    γ-butyrobetaine

Author Contributions

J.M. conceived and supervised the study. Y.W., Y.V.R., A.H.K.AlT., F.H.T.N. and D.C.L. performed experiments and analyzed results. H.V. carried out modeling studies. J.M. wrote the manuscript. All authors contributed to editing the manuscript.

Funding

We thank the Netherlands Organization for Scientific Research (NCI-TA 731.015.202) for financial support. The research has been aided by China Scholarship Council [No. 201606175202 to Y.W.] and the National Nature Science Foundation of China [No. 81702896 to Y.W.].

Competing Interests

The Authors declare that there are no competing interests associated with the manuscript.

References

References
1
Kerner
,
J.
and
Hoppel
,
C.
(
2000
)
Fatty acid import into mitochondria
.
Biochim. Biophys. Acta Mol. Cell Biol. Lipids
1486
,
1
17
2
Strijbis
,
K.
,
Vaz
,
F.M.
and
Distel
,
B.
(
2010
)
Enzymology of the carnitine biosynthesis pathway
.
IUBMB Life
62
,
357
362
3
Vaz
,
F.M.
and
Wanders
,
R.J.A.
(
2002
)
Carnitine biosynthesis in mammals
.
Biochem. J.
361
,
417
429
4
Dunn
,
W.A.
,
Rettura
,
G.
,
Seifter
,
E.
and
Englard
,
S.
(
1984
)
Carnitine biosynthesis from gamma-butyrobetaine and from exogenous protein-bound 6-N-trimethyl-l-lysine by the perfused guinea pig liver. Effect of ascorbate deficiency on the in situ activity of gamma-butyrobetaine hydroxylase
.
J. Biol. Chem.
259
,
10764
10770
PMID:
[PubMed]
5
Sachan
,
D.S.
and
Hoppel
,
C.L.
(
1980
)
Carnitine biosynthesis. Hydroxylation of N6-trimethyl-lysine to 3-hydroxy-N6-trimethyl-lysine
.
Biochem. J.
188
,
529
534
6
Servillo
,
L.
,
Giovane
,
A.
,
Cautela
,
D.
,
Castaldo
,
D.
and
Balestrieri
,
M.L.
(
2014
)
Where does Nε-trimethyllysine for the carnitine biosynthesis in mammals come from?
PLoS ONE
9
,
e84589
7
Herr
,
C.Q.
and
Hausinger
,
R.P.
(
2018
)
Amazing diversity in biochemical roles of Fe(II)/2-oxoglutarate oxygenases
.
Trends Biochem. Sci.
43
,
517
532
8
Leung
,
I.K.H.
,
Krojer
,
T.J.
,
Kochan
,
G.T.
,
Henry
,
L.
,
von Delft
,
F.
,
Claridge
,
T.D.W.
et al.  (
2010
)
Structural and mechanistic studies on γ-butyrobetaine hydroxylase
.
Chem. Biol.
17
,
1316
1324
9
Islam
,
M.S.
,
Leissing
,
T.M.
,
Chowdhury
,
R.
,
Hopkinson
,
R.J.
and
Schofield
,
C.J.
(
2018
)
2-oxoglutarate-dependent oxygenases
.
Annu. Rev. Biochem.
87
,
585
620
10
Vaz
,
F.M.
,
Fouchier
,
S.W.
,
Ofman
,
R.
,
Sommer
,
M.
and
Wanders
,
R.J.A.
(
2000
)
Molecular and biochemical characterization of rat γ-trimethylaminobutyraldehyde dehydrogenase and evidence for the involvement of human aldehyde dehydrogenase 9 in carnitine biosynthesis
.
J. Biol. Chem.
275
,
7390
7394
11
Tars
,
K.
,
Leitans
,
J.
,
Kazaks
,
A.
,
Zelencova
,
D.
,
Liepinsh
,
E.
,
Kuka
,
J.
et al.  (
2014
)
Targeting carnitine biosynthesis: discovery of new inhibitors against γ-butyrobetaine hydroxylase
.
J. Med. Chem.
57
,
2213
2236
12
Dambrova
,
M.
,
Liepinsh
,
E.
and
Kalvinsh
,
I.
(
2002
)
Mildronate
.
Trends Cardiovasc. Med.
12
,
275
279
13
Celestino-Soper
,
P.B.S.
,
Shaw
,
C.A.
,
Sanders
,
S.J.
,
Li
,
J.
,
Murtha
,
M.T.
,
Ercan-Sencicek
,
A.G.
et al.  (
2011
)
Use of array CGH to detect exonic copy number variants throughout the genome in autism families detects a novel deletion in TMLHE
.
Hum. Mol. Genet.
20
,
4360
4370
14
Celestino-Soper
,
P.B.S.
,
Violante
,
S.
,
Crawford
,
E.L.
,
Luo
,
R.
,
Lionel
,
A.C.
,
Delaby
,
E.
et al.  (
2012
)
A common X-linked inborn error of carnitine biosynthesis may be a risk factor for nondysmorphic autism
.
Proc. Natl Acad. Sci. U.S.A.
109
,
7974
7981
15
Longo
,
N.
(
2016
)
Primary carnitine deficiency and newborn screening for disorders of the carnitine cycle
.
Ann. Nutr. Metab.
68
,
5
9
16
Al Temimi
,
A.H.K.
,
Pieters
,
B.J.G.E.
,
Reddy
,
Y.V.
,
White
,
P.B.
and
Mecinović
,
J.
(
2016
)
Substrate scope for trimethyllysine hydroxylase catalysis
.
Chem. Commun.
52
,
12849
12852
17
Hulse
,
J.D.
,
Ellis
,
S.R.
and
Henderson
,
L.M.
(
1978
)
Carnitine biosynthesis. β-hydroxylation of trimethyllysine by an α-ketoglutarate-dependent mitochondrial dioxygenase
.
J. Biol. Chem.
253
,
1654
1659
PMID:
[PubMed]
18
Novak
,
R.F.
,
Swift
,
T.J.
and
Hoppel
,
C.L.
(
1980
)
N6-trimethyl-lysine metabolism. structural identification of the metabolite 3-hydroxy-N6-trimethyl-lysine
.
Biochem. J.
188
,
521
527
19
Leśniak
,
R.K.
,
Markolovic
,
S.
,
Tars
,
K.
and
Schofield
,
C.J.
(
2017
)
Human carnitine biosynthesis proceeds via (2S,3S)-3-hydroxy-Nε-trimethyllysine
.
Chem. Commun.
53
,
440
442
20
Reddy
,
Y.V.
,
Al Temimi
,
A.H.K.
,
White
,
P.B.
and
Mecinović
,
J.
(
2017
)
Evidence that trimethyllysine hydroxylase catalyzes the formation of (2S,3S)-3-hydroxy-Nε-trimethyllysine
.
Org. Lett.
19
,
400
403
21
Reddy
,
Y.V.
,
Al Temimi
,
A.H.K.
and
Mecinović
,
J.
(
2017
)
Fluorinated trimethyllysine as a 19F NMR probe for trimethyllysine hydroxylase catalysis
.
Org. Biomol. Chem.
15
,
1350
1354
22
Kazaks
,
A.
,
Makrecka-Kuka
,
M.
,
Kuka
,
J.
,
Voronkova
,
T.
,
Akopjana
,
I.
,
Grinberga
,
S.
et al.  (
2014
)
Expression and purification of active, stabilized trimethyllysine hydroxylase
.
Protein Expr. Purif.
104
,
1
6
23
Tars
,
K.
,
Rumnieks
,
J.
,
Zeltins
,
A.
,
Kazaks
,
A.
,
Kotelovica
,
S.
,
Leonciks
,
A.
et al.  (
2010
)
Crystal structure of human gamma-butyrobetaine hydroxylase
.
Biochem. Biophys. Res. Commun.
398
,
634
639
24
Vaz
,
F.M.
,
Ofman
,
R.
,
Westinga
,
K.
,
Back
,
J.W.
and
Wanders
,
R.J.A.
(
2001
)
Molecular and biochemical characterization of rat ε-N-trimethyllysine hydroxylase, the first enzyme of carnitine biosynthesis
.
J. Biol. Chem.
276
,
33512
33517
25
Straganz
,
G.D.
and
Nidetzky
,
B.
(
2006
)
Variations of the 2-his-1-carboxylate theme in mononuclear non-heme FeII oxygenases
.
ChemBioChem
7
,
1536
1548
26
Martinez
,
S.
and
Hausinger
,
R.P.
(
2015
)
Catalytic mechanisms of Fe(II)- and 2-oxoglutarate-dependent oxygenases
.
J. Biol. Chem.
290
,
20702
20711
27
Monfregola
,
J.
,
Cevenini
,
A.
,
Terracciano
,
A.
,
van Vlies
,
N.
,
Arbucci
,
S.
,
Wanders
,
R.J.A.
et al.  (
2005
)
Functional analysis of TMLH variants and definition of domains required for catalytic activity and mitochondrial targeting
.
J. Cell. Physiol.
204
,
839
847
28
Hewitson
,
K.S.
,
Holmes
,
S.L.
,
Ehrismann
,
D.
,
Hardy
,
A.P.
,
Chowdhury
,
R.
,
Schofield
,
C.J.
et al.  (
2008
)
Evidence that two enzyme-derived histidine ligands are sufficient for iron binding and catalysis by factor inhibiting HIF (FIH)
.
J. Biol. Chem.
283
,
25971
25978
29
Blasiak
,
L.C.
,
Vaillancourt
,
F.H.
,
Walsh
,
C.T.
and
Drennan
,
C.L.
(
2006
)
Crystal structure of the non-haem iron halogenase SyrB2 in syringomycin biosynthesis
.
Nature
440
,
368
30
Aik
,
W.
,
McDonough
,
M.A.
,
Thalhammer
,
A.
,
Chowdhury
,
R.
and
Schofield
,
C.J.
(
2012
)
Role of the jelly-roll fold in substrate binding by 2-oxoglutarate oxygenases
.
Curr. Opin. Struct. Biol.
22
,
691
700
31
Tarhonskaya
,
H.
,
Chowdhury
,
R.
,
Leung
,
I.K.H.
,
Loik
,
N.D.
,
McCullagh
,
J.S.O.
,
Claridge
,
T.D.W.
et al.  (
2014
)
Investigating the contribution of the active site environment to the slow reaction of hypoxia-inducible factor prolyl hydroxylase domain 2 with oxygen
.
Biochem. J.
463
,
363
372
32
Rydzik
,
A.M.
,
Leung
,
I.K.H.
,
Kochan
,
G.T.
,
Loik
,
N.D.
,
Henry
,
L.
,
McDonough
,
M.A.
et al.  (
2014
)
Comparison of the substrate selectivity and biochemical properties of human and bacterial [gamma]-butyrobetaine hydroxylase
.
Org. Biomol. Chem.
12
,
6354
6358
33
Taverna
,
S.D.
,
Li
,
H.
,
Ruthenburg
,
A.J.
,
Allis
,
C.D.
and
Patel
,
D.J.
(
2007
)
How chromatin-binding modules interpret histone modifications: lessons from professional pocket pickers
.
Nat. Struct. Mol. Biol.
14
,
1025
1040