Characterization of the PLP-dependent aminotransferase NikK from Streptomyces tendae and its putative role in nikkomycin biosynthesis

Authors


Peter Macheroux, Institute of Biochemistry, Graz University of Technology, Petersgasse 12, A-8010 Graz, Austria
Fax: +43 316 873 6952
Tel: +43 316 873 6450
E-mail: peter.macheroux@tugraz.at
Karl Gruber, Institute of Molecular Biosciences, University of Graz, Humboldtstr. 50/3, A-8010 Graz, Austria
Fax: +43 316 380 9897
Tel: +43 316 380 5483
E-mail: karl.gruber@uni-graz.at

Abstract

As inhibitors of chitin synthase, nikkomycins have attracted interest as potential antibiotics. The biosynthetic pathway to these peptide nucleosides in Streptomyces tendae is only partially known. In order to elucidate the last step of the biosynthesis of the aminohexuronic building block, we have heterologously expressed a predicted aminotransferase encoded by the gene nikK from S. tendae in Escherichia coli. The purified protein, which is essential for nikkomycin biosynthesis, has a pyridoxal-5′-phosphate cofactor bound as a Schiff base to lysine 221. The enzyme possesses aminotransferase activity and uses several standard amino acids as amino group donors with a preference for glutamate (Glu > Phe > Trp > Ala > His > Met > Leu). Therefore, we propose that NikK catalyses the introduction of the amino group into the ketohexuronic acid precursor of nikkomycins. At neutral pH, the UV-visible absorbance spectrum of NikK has two absorbance maxima at 357 and 425 nm indicative of the presence of the deprotonated and protonated aldimine with an estimated pKa of 8.3. The rate of donor substrate deamination is faster at higher pH, indicating that an alkaline environment favours the deamination reaction.

Structured digital abstract

Abbreviations
DNPH

2,4-dinitrophenyl hydrazine

3′-EPUMP

3′-enolpyruvyl uridine monophosphate

HisC

histidinol phosphate aminotransferase

HPHT

hydroxypyridylhomothreonine

PEP

phosphoenolpyruvate

PLP

pyridoxal-5′-phosphate

PMP

pyridoxamine-5′-phosphate

Introduction

A variety of structurally related peptide nucleosides termed nikkomycins have been isolated from the culture filtrates of Streptomyces tendae and Streptomyces ansochromogenes [1–4]. Nikkomycins show structural similarity to the chitin synthase substrate UDP-N-acetylglucosamine and hence inhibit chitin biosynthesis in fungi and insects [5]. Owing to this inhibition nikkomycins have potential as potent fungicidal, insecticidal and acaricidal compounds. As they are non-toxic to mammals and easily degradable in nature, they are recognized as ideal fungicides for human therapy and insecticides for agriculture [6,7].

In 1970, the first producing strain for nikkomycins, S. tendae Tü901, was isolated from a soil sample near Nikko, Japan, as part of a programme to discover novel fungicides and insecticides for agricultural use [7,8]. Nikkomycins are composed of a peptidyl and a nucleoside moiety, which are synthesized separately and then linked by a peptide bond [9]. The peptidyl moiety consists of hydroxypyridylhomothreonine (HPHT, nikkomycin D) which is derived from l-lysine and 2-oxo-butyrate [10,11]. 5-Aminohexuronic acid, which is bound N-glycosidically either to uracil or to 4-formyl-4-imidazoline-2-one, forms the nucleoside moiety of nikkomycins Z and X, respectively [2]. Nikkomycins X and Z are the main components of nikkomycins in S. tendae and S. ansochromogenes, and also the most active compounds. The corresponding nucleoside moieties are composed of aminohexuronic acid and designated nikkomycin CX and CZ [12]. HPHT, as well as aminohexuronic acid, were isolated from the culture filtrate of S. tendae at a level three- to four-fold lower than that of nikkomycins X and Z [13].

The nikkomycin biosynthetic gene cluster in S. tendae consists of 22 structural genes, organized in three polycistronic operons and a monocistronically transcribed regulatory gene. Although functions have been assigned to many of these genes by sequence comparisons and gene knock-out experiments [13], the biosynthesis of nikkomycins is so far only partially understood. The biosynthesis of the peptidyl moiety commences with the deamination of l-lysine to the corresponding α-keto acid, catalysed by the aminotransferase NikC (SanL) [14]. After spontaneous cyclization and dehydration of the α-keto acid, the resulting piperideine-2-carboxylate is oxidized by the FAD containing enzyme NikD (SanK) to form picolinic acid [15]. Picolinic acid is activated and loaded to coenzyme A by the ATP-dependent picolinate-CoA ligase NikE (SanJ), and the resulting CoA-thioester is reduced to picolinaldehyde by the dehydrogenase NikA (SanN) [4,11,16]. An aldol condensation between picolinaldehyde and 2-oxobutyrate is catalysed by the aldolase NikB (SanM), generating 4-(2′-pyridinyl)-2-oxo-4-hydroxyisovalerate (POHIV) which is aminated by NikT (SanT) to 4-(2′-pyridinyl)-homothreonine (PHT). Hydroxylation of PHT by NikG (SanI) and NikF (SanH) at the pyridinyl moiety then yields the final peptide residue for nikkomycin biosynthesis [14,17,18].

Nikkomycins of the X series comprise a 4-formyl-4-imidazolin-2-one base, which originates from l-histidine. After l-histidine is activated and transferred to the carrier protein NikP1 (SanO), it is hydroxylated by NikQ (SanQ). NikP2 (SanP) is necessary for the hydrolytic release of β-OH-His from NikP1 (SanO) [2]. Further transformations lead to the imidazolone base which is possibly transferred to 5′-phosphoribosyl-1-pyrophosphate by NikR (SanR), an enzyme with significant sequence similarity to uracil phosphoribosyl transferases [13].

The enolpyruvyl transferase NikO (SanX) has been demonstrated to catalyse the formation of 3′-enolpyruvyl uridine monophosphate (3′-EPUMP) from uridine monophosphate and phosphoenolpyruvate [19]. This is totally different from the previously proposed pathway for the biosynthesis of the nucleoside moiety, where the enolpyruvyl moiety is attached to the 5′-position of uridine [20]. It can be expected that NikO also catalyses the analogous reaction with 5′-phosphoribofuranosyl-4-formyl-4-imidazolin-2-one instead of UMP, because nikO inactivated mutants produced neither the nikkomycins I, J, X and Z nor the nucleoside moieties CX and CZ, but accumulated the novel nucleoside ribofuranosyl-4-formyl-4-imidazolone in the culture filtrate [20]. Considering the introduction of the enolpyruvyl group at the 3′-position of UMP, a rearrangement of the carbon skeleton, catalysed by the enzymes which are cotranscribed with NikO, was postulated [19]. Structural genes for the biosynthesis of the aminohexuronic acid moiety are encoded on the nikIJKLMNO operon, and nikL, nikK and nikM disruption mutants accumulated bicyclic nikkomycins SX and SZ in the fermentation medium [13]. The characterization of the encoded enzymes will shed light on the reactions that lead to aminohexuronic acid. Here, we report the heterologous expression and biochemical characterization of NikK. Our studies show that the enzyme is a pyridoxal-5′-phosphate (PLP) dependent aminotransferase with l-glutamate as the preferred amino group donor. We propose that NikK catalyses the introduction of the amino group into the nucleoside moiety of nikkomycin antibiotics as shown in Scheme 1.

  • image(Scheme 1.)

[  Known enzymatic steps in the biosynthetic reactions leading from UMP to aminohexuronic acid (R = uracil or 4-formyl-4-imidazoline-2-one). The reaction studied here is that of NikK, the transfer of an amino group to the ketohexuronic acid. ]

Results

Genomic sequencing of the nikK gene

The nikK gene region was sequenced six times to confirm the DNA sequence deposited in the UniProt database (GenBank CAC80909.1). In all cases we detected two deviations from the deposited sequence: the codon for the amino acid in position 348 specified asparagine instead of aspartate, and the codon for position 349 specified aspartate instead of histidine (see Fig. S1). Accordingly, these alterations were taken into account for the generation of the synthetic gene used for heterologous expression of nikK.

NikK expression and native molecular weight determination

Several attempts to heterologously express the PCR-amplified nikK gene in Escherichia coli and Pichia pastoris were unsuccessful. Hence, a synthetic gene in E. coli codon usage was employed to express the protein. In addition, the soluble protein yield could be doubled to 5 mg·L−1 medium by concomitant expression of the chaperone genes dnaK, dnaJ, grpE, groEL and groES. The C-terminally hexahistidine-tagged protein was purified to near homogeneity by one-step Ni–nitrilotriacetic acid (Ni–NTA) affinity chromatography, yielding about 0.6 mg of NikK per gram of wet biomass (see Fig. S4). Size exclusion chromatography yielded a single peak with an elution volume corresponding to a molecular mass of 79 kDa.

UV-visible absorbance properties of NikK

The UV-visible absorbance spectrum of purified NikK indicated the presence of a PLP cofactor (Fig. 1). The features of the absorbance spectrum strongly depend on the pH (Fig. 1). At high pH, the spectrum is dominated by an absorbance maximum at 357 nm while at low pH the maximum absorbance shifts to longer wavelength (425 nm). The pH-dependent spectral changes were used to calculate a pKa value of 8.3 for the internal aldimine species that typically forms between the PLP cofactor and the side chain of a lysine residue (Fig. 1, insert) [21]. At low pH, an additional peak at 330 nm is visible, which is characteristic for the presence of pyridoxamine-5′-phosphate (PMP). No fluorescence emission was observed at 500 nm with purified NikK. The enzyme solution merely showed a fluorescence maximum at 385 nm.

Figure 1.

 Absorbance spectra of NikK at various pH values (6.5, solid line; 7.8, long dashed line; 8.1, medium dashed line; 8.6, short dashed line; 9.1, dotted line; 9.8, dash-dotted line). The arrows indicate the direction of change with increasing pH. The insert shows the pH dependence of the absorbance at 357 nm (•) and 425 nm (○).

Demonstration of aminotransferase activity

The deamination of l-alanine in the presence of α-ketoglutarate was observed by measuring the absorbance of the coloured complex formed by pyruvate and 2,4-dinitrophenyl hydrazine (DNPH) at 540 nm. The colour change in the presence of NikK enzyme clearly showed that l-alanine had been converted to pyruvate. No colour change is seen in the absence of α-ketoglutarate, which serves as an amino acceptor. Thus, it is evident that NikK has an aminotransferase activity.

Amino donor specificity: screening of amino acids

Half-transamination reactions of various amino acids have been followed at pH 8.0 and 9.0 (Table 1). The most active amino group donor is l-glutamate. Based on the spectral changes occurring in the presence of l-glutamate, shown in Fig. 2, kmax and KD values of the NikK catalysed deamination of l-glutamate were determined (Fig. 3 and Table 2). Other amino acids (Phe, Trp, Ala, His, Met, Leu) also show significant amino group donor activity. The spectral changes were negligible in the presence of Gly, Ile, Thr and Pro. With Cys, a rapid reaction is observed.

Table 1.   Reactivity of NikK with various amino acids. Reaction rates (kobs) were measured at a concentration of 10 mm of the respective amino acid at pH 8.0 and pH 9.0 by monitoring the absorbance decrease at 425 nm or 357 nm and the absorbance increase at 330 nm and globally fitting the kinetic traces.
 k (s−1)
pH 8.0pH 9.0
Glu0.5600.586
Phe0.1030.116
Trp0.0700.069
Ala0.0560.058
His0.0520.055
Met0.0500.050
Leu0.0450.052
Asp0.0310.031
Gln0.0200.022
Arg0.0130.016
Lys0.0110.011
Val0.0030.003
Cys4.2/0.0033.4/0.003
Ser0.0020.003
Gly00
Ile00
Thr00
Pro00
Figure 2.

 Spectral changes in the presence of 5 mm l-glutamate at pH 8.0 (A) and pH 9.0 (B). The solid lines show spectra of an NikK solution in the absence of glutamate; the dashed lines are spectra taken after 5 min incubation with glutamate.

Figure 3.

 Observed rate of the half-transamination reaction with l-glutamate shown in a Lineweaver–Burk linearization to determine the kmax and KD values by following the absorbance increase at 330 nm at pH 8 (•) and pH 9 (▪).

Table 2. kmax and KD values of the transamination reactions of NikK with l-glutamate.
 kmax (s−1)KD (mm)
pH 8.02.0 ± 0.1832.4 ± 5.5
pH 9.032 ± 22128 ± 38

Following time-dependent spectral changes after mixing NikK with the respective amino acid in a stopped-flow device, it was observed that the reactions conformed to a single-exponential process, except in the case of cysteine as substrate. The kobs values for several substrate concentrations were determined by monitoring the absorbance increase at 330 nm and the absorbance decrease at 425 nm at pH 8 or at 357 nm at pH 9 (Table 1). At a fixed amino acid concentration, the observed rate was independent of the wavelength used to monitor the reaction.

pH dependence of amino group transfer

To assess the pH dependence of amino group transfer, l-alanine was used as amino group donor at different pH values. As shown in Fig. 4, the decrease at 430 nm is ∼ 10 times faster at pH 9 than at pH 7. The pH dependence of the reaction rate correlates with the pH-dependent spectral changes described above (Fig. 1 and insert). Accordingly, the inflection point of the pH dependence shown in Fig. 4 is close to the determined pKa of the aldimine species (8.3).

Figure 4.

 Rate of the absorbance decrease during the half-transamination reaction with 7.5 mm l-alanine as amino group donor, plotted against pH. The sigmoidal function shows an inflection point at pH 8.3, the pKa of the Lys221–PLP Schiff base. The reaction rate increases with pH, indicating that the deprotonated Schiff base is the active form.

Homology model

blast searches against a non-redundant set of protein sequences (National Center for Biotechnology Information) showed that close homologues to NikK, with sequence identities of more than 50%, only occur in other Streptomyces strains followed by a few sequences of histidinol phosphate aminotransferases (HisC) from various organisms, exhibiting around 30% sequence identity. A blast search against the protein databank (PDB) identified the structure of HisC from E. coli in complex with histidinol phosphate (PDB code 1fg3, 25% sequence identity) as the most suitable template (Fig. S2). NikK was modelled as a dimer, because dimer formation was found to be essential for many aminotransferases in order to build up functional active sites [22–24] and because size exclusion chromatography indicated the occurrence of NikK dimers in solution (Fig. S3).

The modelled NikK structure exhibits the same three domains into which structures of aminotransferases can usually be divided, namely N-terminal arm, small domain and PLP binding domain [22,24] (Fig. 5). The model indicates different loop structures in the small domain (residues 33–40), however, as well as in the PLP binding domain (residues 49–65) (Fig. 6). Compared with the sequence of HisC from E. coli, NikK shows an insertion of 12 residues in the aforementioned loop regions. Interestingly, the different loop structures of residues 33–40 introduce three completely new interaction possibilities (C35, N34 and K33) for substrates in the active site, by shifting the corresponding loop towards the PLP cofactor. Concomitant with the shape alteration in this loop region, residues 49–65, which are predicted to form a small helix in NikK rather than an extended loop as in HisC, are affected as they are close in space. Furthermore, a tyrosine residue from one protomer, that in analogous aminotransferase structures is pointing towards the active site of the vice versa protomer [22,24–28] (Y55 in HisC corresponding to Y61 in NikK), is part of the affected residue range (Fig. 6). Other essential active site residues (Asp190, Tyr193, Ser220, Lys221 and Arg330) [24,29], however, are very well conserved in sequence and structure (see Doc. S1).

Figure 5.

 Overall structure of the NikK homology model. The enzyme was modelled as a dimer. One protomer is shown in green (left) whereas the other protomer is colour coded for the three domains typical for aminotransferases: N-terminal arm (salmon), small domain (light blue) and PLP binding domain (pale yellow). The PLP histidinol phosphate adduct, which was kept during modelling to ensure a maximum in plasticity of the active site, is shown in a stick representation and coloured in yellow. The figure was created with pymol (http://www.pymol.org).

Figure 6.

 Comparison of the homology model of NikK with HisC from E. coli. Left: superposition of the NikK homology model and the experimentally determined structure of HisC (PDB accession code 1fg3). Residues corresponding to NikK are coloured in dark green whereas HisC residues are shown in cyan. The PLP cofactor of HisC is shown in yellow as a reference point. All residues are labelled according to NikK numbering. Right: a colour-coded (NikK, dark green; HisC, cyan) cartoon representation of the two loop regions that show different conformations in the NikK model. In contrast to HisC, the loop region of residues 49–65 is predicted to form a helix in NikK and therefore adopts a completely different conformation, which results in a differently shaped binding pocket. Moreover, the region of residues 33–40 shows a different conformation as well, which results in three new interaction possibilities for a bound substrate. These changes most probably are necessary to accommodate the big substrate, ketohexuronic acid, in the active site of NikK.

The role of K221 in Schiff base formation

We generated an NikK variant where K221 is replaced by methionine (K221M). The isolated K221M variant displays a deep yellow colour; however, the UV-visible spectrum shows remarkable differences from the spectrum obtained with wild-type protein. In contrast with wild-type protein, the absorbance maximum of the K221M variant is shifted to lower wavelength (400 nm) resembling the absorbance spectrum of free PLP rather than the enzyme-bound Schiff base. The addition of glutamic acid (10 mm final concentration) to the K221M variant results in a shift of the absorbance maximum to around 410 nm, and formation of the external aldimine does not lead to deamination of the substrate.

Substrate docking

We docked three possible amino group donors and the proposed keto acid substrate in their external aldimine form into the active site of our NikK homology model. Characteristic interactions of the docked ligands were hydrogen bonding of pyridine N1 to D190, stacking interactions of the pyridine ring with F112 and phosphate group stabilization by hydrogen bonding to the main chain NH and side chain OH of S88 for the PLP moiety as well as salt bridge formation between the α-carboxylate of the amino acid/keto acid moieties and R343/R330.

Compared with structures of other aminotransferases (e.g. HisC from E. coli and aspartate aminotransferases from various organisms), additional interactions (Fig. 7A–D) of the ligands with the enzyme are formed due to the differently shaped loop region above the active site of the NikK model (residues 33–35) (Figs 6 and 7). Within this loop especially K33 was found to form an additional salt bridge with the α-carboxylate of the docked ligands. Apart from that, further interactions of the amino acid substrate side chains were facilitated by a salt bridge between the δ-carboxyl group and R229 (glutamate, Fig. 7B) and hydrophobic interactions of the leucine sec-butyl group with F19 (Fig. 7D).

Figure 7.

 Docked and energy minimized complex structures of the NikK homology model and three representative amino group donors as well as the ketohexuronic acid moiety built as external aldimines. (A) Ketohexuronic acid linked via a Schiff base to PMP building an external aldimine. Active site residues are shown as sticks and interactions between protein residues and substrate are indicated by the red dashed lines. The active site lysine K221 is coloured in orange. (B) Docked complex of NikK and glutamic acid external aldimine. Residues from a loop region above the active site are forming additional interactions with the δ-carboxyl group of the substrate. In addition, stacking interactions of a phenylalanine residue with the pyridine ring of PLP can be seen. (C) Docked external aldimine form of cysteine. It can be seen that there is no steric reason for cysteine not being accepted as amino group donor. (D) Leucine docked as external aldimine to the active site of NikK. Even though there are multiple polar interaction possibilities, the hydrophobic side chain of leucine gets accommodated, making hydrophobic interactions with an adjacent phenylalanine residue. All residues are labelled according to NikK amino acid numbering.

Docking of our proposed substrate for NikK, ketohexuronic acid in its external aldimine form, to the active site of our homology model gave a reasonable looking binding mode as well. Here, additional predicted ligand side chain interactions with the enzyme are hydrogen bonding between D89 and the OH groups of the sugar moiety as well as the interaction of the 4-formyl-4-imidazoline-2-one with the conserved R229 (Fig. 7A).

To compare the plasticity of the active sites of our NikK homology model with its template, HisC from E. coli, we superimposed the obtained docking pose for ketohexuronic acid onto the HisC active site by fitting the two PLP moieties to each other. This resulted in clashes with side chain and main chain atoms of HisC residues 54–58. These residues, however, correspond to residues 60–64 in NikK, which are part of one altered loop region (Figs 6 and 8).

Figure 8.

 Superposition of the binding mode obtained for the ketohexuronic acid docked as an external aldimine to NikK, with the active site structure of HisC. The orientation of the loop region (residues 49–65) in HisC would introduce clashes with the sugar moiety of the substrate. This is further evidence for the different structure of this loop region being essential for binding of ketohexuronic acid.

Discussion

In order to gain further insight into the biosynthesis of nikkomycins, we have expressed NikK to demonstrate its role as an aminotransferase involved in the last step of the biosynthesis of aminohexuronic acid from ketohexuronic acid (Scheme 1). The S. tendae protein shows 97% identity to the aminotransferase domain of SanB of S. ansochromogenes. It was shown by Li et al. [30] in 2000 that disruption mutants of sanB lost the ability to synthesize nikkomycins. The nikK gene encodes a 39.5 kDa protein which shows sequence similarities to several PLPs containing HisC, e.g. 44% similarity and 25% identity to the HisC from E. coli. Using a synthetic nikK gene optimized for E. coli codon usage, a sizable amount of soluble protein was produced. The yellow colour of the protein solution and the UV-visible absorbance properties of the protein indicated the presence of PLP. NikK was shown to form a dimer in solution, like many other PLP-dependent enzymes, with each protomer having a mass of approximately 40 kDa as judged from SDS/PAGE. For many PLP-dependent enzymes, dimer formation seems to be essential for forming a functional active site [24].

The PLP cofactor features distinct peaks in the UV-visible absorbance spectrum. The peak at 425 nm, which is prominent at lower pH values, represents the protonated form of the nitrogen in the Schiff base formed between the aldehyde group of the cofactor and the side chain amino group of an active site lysine residue. On the other hand, the peak at 357 nm observed at high pH shows the deprotonated form of the Schiff base nitrogen. The absorbance spectra intersect in a single isosbestic point at 380 nm. From the pH dependence of the UV-visible absorbance spectra, a pKa of 8.3 was determined. This value is significantly higher than the values of 6.6 and 6.8 determined for HisC (EC 2.6.1.9) and aspartate aminotransferase (EC 2.6.1.1), respectively. Apparently, in those enzymes the pKa increases successively during catalysis. In aspartate:2-oxoglutarate aminotransferase, pKa is expected to increase to 8.8 in the Michaelis complex that is formed with the substrate amino acid and to above 10 in the external aldimine. The low pKa value of the unliganded enzyme is explained by the strain on the imine–pyridine torsion angle of the Schiff base, which is also crucial for the successive increase of pKa [31,32]. On binding of the substrate amino acid, pKa increases and a proton is transferred from the protonated amino group of the substrate to the Schiff base nitrogen. The free amino group then nucleophilically attacks the imine, which has an increased electrophilicity due to protonation. In this so-called transaldimination step, the PLP substrate Schiff base is formed [32]. However, the imine nitrogen of another member of subfamily I aminotransferases, glutamine:phenylpyruvate aminotransferase from Thermus thermophilus, shows a comparably high pKa value of 9.3 [33].

The protonated Schiff base can also form an enolimine with an absorbance at 340 nm. We have observed an additional peak at 330 nm in the UV-visible absorbance spectrum of the purified NikK enzyme. However, the enolimine exhibits a typical fluorescence emission centred at around 500 nm which could not be detected in the case of NikK [21]. Therefore, we conclude that the absorbance at 330 nm is not due to the enolimine but rather from the PMP form of the cofactor, which might be formed during cell disruption and purification, when the protein is in contact with molecules that function as amino group donors.

Our homology model predicted K221 as a potential candidate for Schiff base formation with the PLP cofactor. In order to confirm this assignment we have exchanged K221 with leucine by site-directed mutagenesis of the nikK gene. The resulting K221M variant shows significant changes in the UV-visible spectrum with an absorbance maximum at 400 nm (Fig. 7). The absorbance spectrum of the K221M variant resembles that of free PLP suggesting that the variant is still capable of PLP binding; however, it is unable to form the internal Schiff base. Upon addition of a potential amino group donor, no conversion to PMP is observed, clearly demonstrating that K221 is the catalytic lysine in the active site of NikK.

Amino transfer reactions catalysed by PLP-dependent enzymes proceed in two steps, the first of which is always a transaldimination reaction, which starts by conversion of the internal aldimine formed by the PLP cofactor and the ε-amino group of the lysine residue in the active site into an external aldimine intermediate by replacing the lysine residue with the amino acid substrate [26]. Most probably, the mechanism is comparable to that of serine- glyoxylate aminotransferase, which was described by Karsten et al. The amino acid substrate binds with its α-amine protonated, and an enzymatic base accepts a proton from the amino group of the substrate. The resulting free amino group attacks the PLP cofactor at C4′, leading to a gem-diamine intermediate which then collapses to form the external aldimine. In the external aldimine, the α-proton is abstracted from the amino acid substrate by the ε-amino group of the lysine residue, resulting in the quinonoid intermediate. The most important function of PLP is the stabilization of carbanions generated by the loss of the proton at Cα of the substrate. The negative charge is delocalized by resonance into the π-electron system of the quinonoid intermediate. These properties of a suitable electron sink are further enhanced by a close interaction of the pyridinium nitrogen with an aspartate residue [34]. According to our homology model, this role is fulfilled by D190 in the active site of NikK. Proton transfer from the ε-amino group of the lysine to C4′ of the quinonoid intermediate leads to the ketamine which is finally hydrolysed to the keto acid corresponding to the substrate and the PMP form of the enzyme [35,36].

We have characterized the deamination step spectroscopically with several amino acids by observing the absorbance change at 330, 357 and 425 nm. The reaction rate was negligible with Ser, Gly, Ile, Thr and Pro as amino group donors while other amino acids, like Phe, Trp, Ala or His, showed significant amino donor activity (Table 1). Glutamate turned out to be the most active amino donor (Table 1), as is the case for the majority of aminotransferases. The transamination reaction with glutamate is pH dependent, with rates increasing at higher pH (see Figs 2 and 3, and Table 2). The pH dependence of the rate exhibits the same pKa as the protonation/deprotonation equilibrium of the Schiff base nitrogen indicating that the deprotonated form of the Schiff base is the active form of the enzyme that deprotonates the incoming amino acid zwitterion.

Very rapid spectral changes in the NikK spectrum are observed in the reaction with cysteine. However, the reaction does not proceed in a monophasic manner as observed with the other amino acids but seems to consist of a rapid and a slow phase (for rates see Table 1). We suppose that these observations are due to a side reaction and not to amino group transfer. Cysteine reacts with PLP resulting in the formation of a stable thiazolidine ring, a species that absorbs at 330 nm [37,38].

Our results demonstrate that NikK is an active transaminase with a preference for glutamate as amino group donor. The enzyme may then transfer the amino group to ketohexuronic acid to complete the synthesis of aminohexuronic acid used as a building block for nikkomycins (see Scheme 1). Unfortunately, up to now, the keto acid substrate is not available for testing. How the ketohexuronic acid is synthesized by the gene products of nikI, nikJ, nikL and nikM from the first metabolite 3′-enolpyruvyl-UMP is currently the subject of further investigations in our laboratory.

To get a first idea of what the NikK structure could look like, we built a homology model using the structure of HisC from E. coli as a template. Despite the low sequence identity (25%), NikK displays conservation of many essential active site residues such as D190, Y193, S220, K221 and R330. However, there are subtle differences between our model and other aminotransferase structures. Two loop regions (residues 33–40 and 49–65) are modelled in different conformations in the NikK homology model. This is most probably due to missing alignments for the residues from these parts of the structure. Still, considering that this is just a homology model, these new conformations would not only introduce additional interaction partners for a bound substrate in the active site of NikK, but also have a major influence on the size and plasticity of NikK’s binding pocket (Figs 6 and 7). Moreover, the change in loop conformations accompanied by the gain in active site space is in agreement with the increased spatial requirements of the ketohexuronic acid substrate compared with the amino group acceptor of HisC.

In addition, we obtained interesting results for docking of the ketohexuronic acid as well. The slightly changed shape of the binding pocket in the NikK homology model enabled the accommodation of our big substrate. Thus we propose that this change in shape is necessary to enable the amination of the ketohexuronic acid substrate to the aminohexuronic acid product. Further evidence for these shape alterations being a prerequisite for NikK to act as an aminotransferase on such a big substrate is that, by superposing the docking pose of the ketohexuronic acid onto the PLP cofactor in the HisC structure, clashes with side and main chain atoms of residues 54–58, from the second protomer, are found (Fig. 8).

Furthermore, we docked three different amino acids (Glu, Leu and Cys) and the proposed substrate, ketohexuronic acid, in their external aldimine form into the active site of the homology model, to get a first rationalization for the aminotransferase activity of NikK. Thereby, we were able to identify potential active site residues, which could provide important interactions with the ligands. In our obtained docking pose for glutamate, the polar charged side chain gets accommodated very well by forming a salt bridge between R229 and the δ-carboxyl group of the amino acid. In the pose obtained for leucine, on the other hand, favourable interactions are formed between F19 and the sec-butyl side chain of the amino acid. In general, the environment of the homology model’s active site is aligned by more polar than apolar residues and the only hydrophobic interaction of leucine is that with F19, which might explain the lower reactivity of this amino acid with NikK. The external aldimine form of cysteine seems to be easily accommodated in the active site of the NikK homology model. No structural explanation can be drawn from the docking calculations that would explain the non-acceptance of cysteine as amino group donor.

Materials and methods

Reagents

All chemicals were of the highest grade commercially available and were purchased from Sigma-Aldrich (St Louis, MO, USA), Fluka (Buchs, Switzerland) or Merck (Darmstadt, Germany). Ni–NTA agarose was from Qiagen (Hilden, Germany) and Sephadex resin from Pharmacia Biotech AB (Uppsala, Sweden). The chaperone plasmid set was from Takara (Shiga, Japan).

Determination of the nikK gene sequence

Genomic DNA was isolated from S. tendae Tü901/8c using the illustra™ bacteria genomicPrep Mini Spin Kit from GE Healthcare (Amersham, UK). The nikK gene region was amplified by PCR and several clones of the PCR product were analysed by DNA sequencing (Eurofins DNA; Ebersberg, Germany).

Cloning and expression in E. coli

The synthetic gene in E. coli codon usage, encoding for the S. tendae NikK, was ordered from DNA 2.0. It contained an NdeI restriction site at the 5′-end and a XhoI restriction site at the 3′-end, and was cloned into the NdeI/XhoI restriction sites of the pET21a vector (Novagen), generating the expression plasmid pET21a(nikK). The gene was designed without a stop codon, which allows expression of the protein with a C-terminal hexahistidine tag. The nikK sequence on the expression plasmid was verified by DNA sequencing (Eurofins DNA). Chemically competent E. coli BL21(DE3) cells were transformed with the chaperone plasmid pG-KJE8 (Takara Bio Inc.); these transformants were made competent and transformed with pET21a(nikK).

Expression of NikK with the help of co-expressed chaperones (dnaK, dnaJ, grpE, groES, groEL) was achieved by growing a 10 mL preculture in Luria–Bertani (LB) medium containing 50 μg·mL−1 ampicillin and 20 μg·mL−1 chloramphenicol overnight at 37 °C. The preculture was used to inoculate 750 mL LB medium containing 50 μg·mL−1 ampicillin and 20 μg·mL−1 chloramphenicol for plasmid selection, 1 mm pyridoxine for cofactor integration and 0.5 mg·mL−1 l-arabinose and 1 ng·mL−1 tetracycline for chaperone induction. The culture was incubated at 37 °C until a A600 of 0.5 and isopropyl thio-β-d-galactoside was added to a final concentration of 0.1 mm. The culture was incubated overnight at 20 °C and cells were harvested by centrifugation. The pellet was washed with 0.9% NaCl solution and stored at −20 °C.

Generation of NikK K221M and E265A/E266A/Q267A variants by site-specific mutagenesis

Site-directed mutagenesis was carried out according to the QuikChange® XL Site-directed Mutagenesis Kit from Stratagene (Santa Clara, CA, USA). The pET21a(nikK) plasmid served as template. The following primers and their complementary counterparts were used: 5′-C CAC GTA GTT CGT GTG AAT ACC TTC TCT ATG AGC TAC GGC CTG TCT GG-3′ for the NikK K221M variant and 5′-GT CGC TCC CTG GCGGCGGCG GCT GTT TTC ACT GCG ATT TG-3′ for the NikK E265A/E266A/Q267A variant. The underlined nucleotides denote the mutated codons. After mutagenesis, the sequence of the transformation construct was verified by DNA sequence analysis. Transformation and expression were carried out as described for wild-type NikK.

Cell disruption and purification

The pellet was thawed and resuspended in lysis buffer (50 mm phosphate buffer at pH 8.0, containing 300 mm NaCl and 10 mm imidazole), using 2 mL buffer per gram of wet cells. After adding a few milligrams of PLP, cells were disrupted by 30 min incubation with lysozyme and 0.5 s sonication pulses for 10 min while cooling on ice. The cell debris was removed by centrifugation at 18 000 g for 30 min at 4 °C.

The hexahistidine-tagged NikK was purified by Ni–NTA affinity chromatography, loading the supernatant onto an Ni–NTA column (Qiagen), previously equilibrated with lysis buffer. After loading of the filtered lysate, the column was washed with 10 column volumes of wash buffer (50 mm sodium phosphate, 300 mm NaCl, 20 mm imidazole, pH 8.0) and bound protein was recovered with elution buffer (50 mm sodium phosphate, 300 mm NaCl, 150 mm imidazole, pH 8.0). The purity of the eluted 3 mL fractions was determined using SDS/PAGE. Fractions containing NikK were pooled and concentrated using Centripreps (Millipore). The buffer was exchanged to 50 mm potassium phosphate containing 150 mm NaCl (pH 7.5) using a PD-10 column (GE Healthcare) and the deeply yellow solution of NikK protein was stored at 4 °C.

Native molecular mass determination

The molecular mass of the purified protein was determined by gel filtration chromatography in order to assess the oligomerization state of native NikK. The analytical column, a Superdex 200 10/300 GL (Pharmacia Biotech AB), was mounted on a fast protein liquid chromatography system (ÄKTA Explorer, Pharmacia Biotech AB). The column was equilibrated with 100 mm Tris/HCl buffer containing 150 mm NaCl at pH 7.5. A 100 μL sample containing 200 μg of NikK was applied to the column, and the elution volume was determined by monitoring the absorbance at 280 and 430 nm. The column was calibrated using a high-molecular-mass gel-filtration calibration kit (Amersham Biosciences; α-chymotrypsinogen A, 25 kDa; ovalbumin, 43 kDa; bovine serum albumin, 67 kDa; aldolase, 158 kDa; catalase, 232 kDa; and ferritin, 440 kDa).

Spectrophotometric measurements and determination of protein concentration

UV-visible absorbance spectra were recorded with a Specord 205 spectrophotometer (Analytik Jena, Jena, Germany) and fluorescence emission spectra with a Hitachi U-3210 fluorescence instrument (Tokyo, Japan). The protein solution contained 50 mm potassium phosphate buffer at pH 7.5. To measure protein solutions at various pH values, either 200 mm Tris (pH 8–10) or Mops (pH 6–7.9) buffers were used. Enzyme concentrations were measured using the Pierce BCA™ Protein Assay Kit from Thermo Scientific. Spectra of the NikK wild-type, NikK K221M and free PLP were taken in 100 mm potassium phosphate buffer, at pH 8.0.

Aminotransferase activity assay

The transfer of amino groups from l-alanine to α-ketoglutarate was assayed according to the procedure established by Reitman and Frankel [39], also described by Vedavathi et al. [40]. In our activity test, DNPH was added after a reaction time of 30 min.

Amino donor specificity and rapid reaction kinetics

The decrease of the absorbance bands at 425 and 357 nm and the increase at 330 nm were followed spectrophotometrically in order to assess the transfer reaction of the amino group from the donor amino acid to the PLP cofactor [21]. All experiments were carried out at 25 °C. The enzyme solution containing 100 μm NikK and 200 mm Tris/HCl was mixed with an equal volume of amino acid solution of varying concentrations in the same buffer and the absorbance decrease was measured immediately after mixing. The experiment was carried out at various pH, using stopped-flow equipment (SF-61DX2; Hi-Tech Scientific) and measuring time-dependent changes of the absorbance spectrum (300–700 nm) with a diode array detector or at single wavelengths (330, 357, 425 nm) with a photomultiplier.

Molecular modelling

The homology model of dimeric NikK was built using yasara [41,42]. The structure of HisC from E. coli (PDB accession code 1fg3) [24] which shares 25% sequence identity with NikK was manually selected as the template for modelling. In order to get a good plasticity of the NikK active site, a gem-diamine complex between PLP, histidinol phosphate and the active site lysine residue, present in the template structure, was kept during modelling. This model was further refined using secondary structure predictions (HHpred, http://toolkit.tuebingen.mpg.de/hhpred) [43,44] for previously unaligned regions using the program modeller v9.8 [45]. Optimization of the model was facilitated using a protocol described by Krieger et al. [42]. Briefly, a short molecular dynamics simulation was carried out under constant pressure and temperature conditions. The homology model was solvated in a water box, sodium ions were used to neutralize charges on the protein and the modelled ligand intermediate was parameterized automatically. The solvation box was 7.5 Å bigger than the protein in each direction. The whole system was optimized by a steepest descent energy minimization using the amber 03 force field [46,47]. Long-range electrostatic interactions were treated using the particle-mesh–Ewald method [48]. A trajectory was run for a total of 500 ps with snapshots taken every 25 ps. Every snapshot structure was minimized by a simulated annealing run and the energies were tabulated. The lowest energy snapshot was taken as the best optimized structure and analysed using yasara (http://www.yasara.org) and pymol (http://www.pymol.org).

For docking calculations the structures of ligands (external aldimines of glutamic acid, leucine, cysteine and the ketohexuronic acid) were built in maestro (http://www.schrodinger.com) and geometry was optimized using the opls force field [49,50]. Partial charges for the ligands were computed with the antechamber program of amber 11 using the semi-empirical AM1-bcc method [51]. The refined homology model was prepared according to the requirements of autodock v4.2 (addition of polar hydrogen atoms, addition of partial charges, assignment of atom types). The modelled ligand intermediate was deleted from the model prior to docking and the simulations were performed with autodock v4.2 [1] using the implemented Lamarckian genetic algorithm with a population size of 350 individuals and an average number of generations of about 900. The cubic energy grid was centred on the Nε atom of K221 and had an extension of 22.5 Å in each direction. While all ligand structures were treated as fully flexible to ensure a maximum degree of freedom during the simulations, all protein residues were kept rigid. For each ligand structure, 50 independent docking runs were performed and the resulting binding modes were clustered based on an rms deviation cut-off of 2.0 Å. The obtained structures were further subjected to geometry optimization in yasara.

Every protein–ligand complex was analysed individually to see whether it met the criteria for a mechanistically reasonable binding mode (angle between Nε of K221, C4A and C4 of PLP at around 90°, distance of N1 from PLP to the carboxyl group of D190 within 4 Å, salt bridge between the α-carboxyl group of the ligand and R343, superposition with the PLP cofactor).

Acknowledgements

This work was supported in part by the Fonds zur Förderung der wissenschaftlichen Forschung (FWF) through grant P19858. As part of the PhD programme ‘Molecular Enzymology’ (FWF W901 to P.M. and K.G.) the project also received support for a research stay abroad (to G.O.) and conference participation (to A.B and G.O.).

Ancillary