Active site conformational changes upon reaction intermediate biotinyl-5'-AMP binding in biotin protein ligase from Mycobacterium tuberculosis

Abstract

Protein biotinylation, a rare form of post-translational modification, is found in enzymes required for lipid biosynthesis. In mycobacteria, this process is essential for the formation of their complex and distinct cell wall and has become a focal point of drug discovery approaches. The enzyme responsible for this process, biotin protein ligase, substantially varies in different species in terms of overall structural organization, regulation of function and substrate specificity. To advance the understanding of the molecular mechanism of biotinylation in Mycobacterium tuberculosis we have biochemically and structurally characterized the corresponding enzyme. We report the high-resolution crystal structures of the apo-form and reaction intermediate biotinyl-5'-AMP-bound form of M. tuberculosis biotin protein ligase. Binding of the reaction intermediate leads to clear disorder-to-order transitions. We show that a conserved lysine, Lys138, in the active site is essential for biotinylation.

Introduction

Biotin-dependent carboxylases are enzymes that have multiple important metabolic functions.^1,2 These enzymes share a requirement for biotin, which is covalently bound to a biotin carboxyl-carrier protein (BCCP) domain common to all these carboxylases. Biotin protein ligase (BPL, EC 6.3.4.15) catalyses the ATP-dependent ligation of biotin to a BCCP domain, in two reaction steps:

In the first reaction biotin and ATP are condensed into the intermediate biotinyl-5'-AMP liberating pyrophosphate. The BPL then recruits BCCP and transfers the biotin from biotinyl-5'-AMP to the BCCP protein substrate. BPL and down-stream enzymes that use biotinylation as covalent modification have recently been investigated as potential drug targets demonstrating the importance of the enzyme for potential medicinal applications and to understand in mechanistic terms the biochemical processes associated with it.³

Several structures of bacterial and archaeal BPLs have been determined.^4–10 The Escherichia coli BPL, called the biotin-induced repressor A (BirA), has been extensively studied as a BPL prototype.^4–7 E. coli BirA is a type II BPL and consists of three domains: an N-terminal DNA-binding domain with a helix-turn-helix motif, a ligase domain with a central β-sheet and a C-terminal SH3-like domain.⁶ E. coli BPL is a bifunctional protein. In addition to its biotin ligase function, it also functions as a biotin synthesis repressor.¹¹ Binding of the BCCP domain and binding of biotin or any analog of the intermediate biotinyl-5'-AMP to E. coli BPL are mutually exclusive.¹² BPL assembly with BCCP induces BPL dimerization, which in turn leads to binding and subsequent repression of the biotin biosynthesis operon.^11,12

In contrast, BPLs in thermophilic archaea and mycobacteria are type I BPLs, which lack the N-terminal DNA-binding domain and hence are incapable of regulation at the expression level. The type I BPL catalytic ligase domain and the C-terminal SH3-like domain have a similar structure to those of type II BPLs. Dimerization of some type I BPLs has been reported,^8,12 but it is independent of ligand binding and the dimerization interfaces are distinct from those seen in prototypical type II E. coli BPL.^5,8,12 In M. tuberculosis BPL (Rv3279c) dimerization under specific crystallization conditions has been observed.¹³ However, gel filtration and dynamic light scattering data showed that the protein is monomeric in solution.^13,14

The aim of this study was to add new structural and functional findings on M. tuberculosis BPL where data have been missing and to integrate them into a complete and useful picture of this enzyme. We report the first structure of the biotin-5′-AMP-bound form of M. tuberculosis BPL at 1.7-Å resolution and compare it to the apo-conformation of the enzyme, which we determined at 1.8-Å resolution. Analysis of these structures revealed substantial conformational changes that took place upon biotin-5′-AMP binding. We validated the active site of M. tuberculosis BPL by mutating Lys138 into a serine, which completely abolishes BPL activity.

Results

M. tuberculosis BPL is monomeric irrespective of ligand binding

We used analytical gel filtration to determine the BPL association state in solution (Fig. 1). The molecular weight of the full-length BPL based on its amino acid sequence is ∼28 kDa. The retention volume of BPL (12.45 mL) from the calibrated gel filtration column gives a molecular weight of 26.7 kDa [Fig. 1(A)], unambiguously indicating that the enzyme is monomeric in solution. The elution profile of BPL did not change when the protein is incubated with biotin and ATP implying that it remains monomeric in the presence of these ligands [Fig. 1(B)]. This is in agreement with published data that demonstrate that M. tuberculosis BPL is monomeric in solution regardless of which substrate is present.¹⁴

Figure 1.

BPL is a monomer in solution. A, Elution profile of BPL from an analytical gel filtration column calibrated with mass standards (inset). BPL elutes in a volume of 12.45 mL with an apparent molecular weight of a monomer. B, Elution profile of BPL in the presence of 40 µM biotin and 3 mM ATP. BPL elutes in a volume of 12.04 mL. There is only a minor change in the peak position implying that BPL remains monomeric in the presence of ligand.

M. tuberculosis BPL structure in the absence and presence of the step 1 reaction product biotinyl-5'-AMP

We have determined the crystal structures of the apo- and biotinyl-5′-AMP-complexed forms of M. tuberculosis BPL at 1.8 Å and 1.7 Å resolution, respectively [Fig. 2(A,B)]. The overall structure of the apo- and ligand-bound BPLs is similar with a root mean square deviation in Cα atomic positions of 0.47 Å [Fig. 2(C)]. The overall fold of M. tuberculosis BPL is in agreement with a 2.8-Å resolution structure of the apo-form of the enzyme.¹³ The protein contains two domains in a dumbbell shaped arrangement. At the N-terminal part there is a catalytic domain (residues 1–217), with a BPL-LplA-LipB fold (Pfam family: PF03099). Its central part is a β-sheet consisting of seven strands (β1↑ β2↑ β3↓ β7↑ β6↓ β5↑ β4↓). Five helices flank the β-sheet: α2 is on one side; α1, α3, α4 and α5 are on the other side. The C-terminal SH3-like domain (residues 218–265) is composed of five anti-parallel β-strands (β12↑ β8↓ β9↑ β10↓ β11↑) that form a small β-barrel.

Figure 2.

Structures of apo- and biotinyl-5′-AMP-bound BPL. A, Apo-BPL (2CGH). The catalytic domain is colored green and the C-terminal SH3-like domain is colored blue. Secondary structure elements are labeled. B, The structure of biotinyl-5′-AMP-bound BPL (4OP0). The domains are colored as in A. Sequence segments undergoing disorder-to-order transitions, when comparing the apo and biotinyl-5′-AMP-bound structures of M. tuberculosis BPL, are colored red and their residue range is given. Biotinyl-5′-AMP is shown as a stick model. C, The apo- and biotinyl-5′-AMP-bound BPL structures superposed. D, The electron density map of biotinyl-5′-AMP is shown in black (2mF_o − DF_c map contoured at the 1σ level). E, Schematic representation of the BPL active site generated by LIGPLOT (http://www.ebi.ac.uk/thornton-srv/software/LIGPLOT). Residues are labeled according to the sequence of 4OP0. Carbon atoms are colored black, oxygen atoms red, nitrogen atoms blue, sulfur is yellow and the phosphorus atom is colored orange. F, BPL active site with the side chains of key residues shown by stick presentation.

In the crystal of the ligand-bound BPL there is well interpretable electron density for biotinyl-5′-AMP, implying that ATP and biotin react to form this product of the first BPL reaction step [Fig. 2(D)], as previously reported for BPL from Pyrococcus horikoshii.⁹ Three regions that lack electron density in the crystal of apo-BPL, including residues 1–7, 63–77, and 162–171, are evident in biotinyl-5′-AMP-bound BPL [Fig. 2(B)]. The latter two of these segments are near the BPL active site, and their folding is likely to be induced by biotinyl-5′-AMP binding, as observed in the equivalent enzymes from E. coli and P. horikoshii.^5,9 This data is also in agreement with the report of an inhibitor-bound M. tuberculosis BPL structure in which the same disorder-to-order transition was observed.³ A short helix and loop is formed by residues 162–172, which together with the two additional strands formed by residue segment 63–77, as well as strands β6 and β7 of the central β-sheet, compose the ligand binding pocket. The interactions between biotinyl-5′-AMP and M. tuberculosis BPL are similar to those reported for other ligand-bound BPL structures [Fig. 2(E,F)],^3,5,8 demonstrating that biotinyl-5′-AMP binding and hence the overall active site architecture is conserved across type I and type II BPLs. The invariant active site Lys138 tightly interacts with the oxygen atoms of both biotinyl and AMP moieties of biotinyl-5′-AMP [Fig. 2(E)].

In addition, residues 1–7 form a helix at the N-terminus in the reaction intermediate-bound BPL. As this area is remote from the active site, its folding may be induced by crystal contacts specific to the monoclinic biotinyl-5′-AMP-bound BPL crystal form (cf., Table I).

Table I.

X-ray Structure Determination and Refinement Statistics

Crystals	apo-BPL	Biotinyl-5'-AMP-BPL
Wavelength (A)	0.9364	0.9762
Space group	P212121	P21
Cell dimensions (A) and angles (°)	a = 62.1, b = 81.0, c = 101.9	a = 41.7, b = 75.4, c = 77.6, β = 97.9
Resolution rangea	39.4–1.80 (1.90–1.80)	31.4–1.70 (1.80–1.70)
Total unique reflections	48344	51119
Completeness (%)a	99.8 (99.9)	97.4 (95.9)
Mean I/σ(I)a	36.3 (3.9)	15.6 (3.2)
Multiplicitya	16.0 (8.5)	12.0 (7.1)
Rint (%)a	4.4 (55.9)	9.8 (52.6)
Rsigmab (%)a	1.7 (26.2)	4.1 (31.1)
Refinement statistics
Reflections	45951	48546
Rwork/Rfree (%)a	18.5 (22.4)	19.5 (24.4)
Number of atoms
Protein	3868	4197
Ligand	—	76
Solvent	342	340
Relative mean square deviation from standard values
Bond length (Å)	0.018	0.016
Bond angle (°)	1.78	1.82
Ramachandran plot most favored regions	92.2	91.2
Additional allowed regions	7.8	8.4
Generously allowed regions		0.5

^aValues in parenthesis are for the highest resolution shell.

^bR_sigma = Σ[sigma (|Fo|²)]/Σ[|Fo|²]. |Fo|² are intensities of the merged reflections, referred to in the SHELX-97 manual.

Lys138 is essential for M. tuberculosis BPL activity

We have used a coupled enzyme assay to measure the catalytic activity towards ATP of BPL [Fig. 3(A)]. Wild type BPL is active and its kinetic parameters for ATP consumption are K_M = 0.20 ± 0.04 mM and the k_cat = 0.017 s⁻¹ (k_cat/K_M = 0.085 s⁻¹ mM⁻¹) [Fig. 3(A)]. The kinetic parameters determined here differ slightly from those reported by Purushothaman et al., ¹⁴ likely due to the higher concentrations of substrate they assayed in the linear velocity range. A mobility shift assay was employed to measure BCCP biotinylation [Fig. 3(C)]. We subsequently used these assays to test the role of the Lys138 of M. tuberculosis BPL. The lysine to serine substitution completely abolishes the ligase activity [Fig. 3(B)] and renders BPL unable to biotinylate its protein substrate BCCP [Fig. 3(C)] confirming the essential role of this conserved active site lysine in BPL catalysis.

Figure 3.

BPL K¹³⁸S is inactive. A, Steady state kinetics of BPL for its substrate ATP. BPL reaction velocities are plotted against the increasing concentration of ATP. B, Measurement of BPL activity. The decrease in OD₃₄₀ is a reflection of BPL activity, measured by the consumption of NADH. No decrease is evident in the mutant K¹³⁸S. C, Electrophoretic mobility shift in a non-denaturing polyacrylamide gel assessing the ability of BPL to biotinylate BCCP. All reactions contained BCCP, ATP and biotin as substrates. Biotinylation of BCCP results in increased mobility toward the anode.¹⁵ Lane 1, in the absence of BPL there is no biotinylation of BCCP. Lane 2 and 3 show a shift due to biotinylation of BCCP catalyzed by wild type BPL at two concentrations, 2 and 8 μM BPL, respectively. Lanes 4 and 5 contained samples of the same reactions as in lane 2 and 3, however, the wild type BPL was substituted with the mutant K¹³⁸S at the two concentrations, 2 and 8 μM.

Discussion

In this contribution, we have been able to determine and characterize the BPL structure from M. tuberculosis in the presence of the reaction intermediate biotinyl-5′-AMP. This allowed comparison with related structures, in particular with a recent structure of BPL in which the active site is bound by the inhibitor Bio-AMS.³ The structures of both these ligand-bound BPLs are nearly identical, their root mean square deviation of Cα atomic positions being 0.38 Å (Fig. 4). The structure of the binding pocket in biotinyl-5'-AMP-bound BPL is also similar to that of M. tuberculosis BPL bound by Bio-AMS [cf., Fig. 4(A) and Fig. 4(B)].³ Binding of the inhibitor leads to the same type of disorder-to-order transitions that are evident in the structure of BPL bound by the reaction intermediate.³ The two ligands adopt a similar conformation in the active site [Fig. 4(D)] and the same constellation of hydrogen bonds is formed between their respective ligands and the protein backbone [Fig. 2(E)].³ In particular, the active site residue Lys138 is hydrogen bonded to the two ligands in the same manner.

Figure 4.

Comparison of M. tuberculosis BPL structures. A, Structure of the biotinyl-5′AMP-bound BPL (4OP0) shown as a reference structure for those depicted in B and C. These are shown separately for clarity. Colors and labels are those used in Figure 2. B, Structure of BPL bound by the inhibitor Bio-AMS (3RUX³). C, Structure of dehydrated BPL (3L1A¹³). Regions that undergo disorder-to-order transitions in ligand-bound BPL are colored red. D, All three BPL structures superposed. Only those regions adopting structure in the presence of ligand are depicted, as are the superposed biotinyl-5′-AMP and Bio-AMS. The root mean square deviation of Cα atomic positions between 4OP0 and 3RUX is 0.38 Å and that between 4OP0 and 3L1A is 0.64 Å.

Interestingly, the structure of a M. tuberculosis apo-BPL from a dehydrated crystal (3L1A)¹³ reveals that the residue segments 63–77 and 162–171 are ordered, as observed for M. tuberculosis BPL in the presence of the reaction intermediate [Fig. 4(C)]. These however adopted a different conformation to both the reaction intermediate and inhibitor-bound forms [Fig. 4(D)]. These differences, although likely biased by enhanced crystal contacts due to crystal dehydration, may indicate that the biotinyl-5′-AMP disorder-to-order transition is more transient than suggested by a sole binary comparison of the apo- and the biotinyl-5′-AMP-bound forms.

Most residues interacting with the biotinyl moiety have the same conformation in the apo- and the ligand-bound form, whereas the AMP binding residues largely reside in those regions that are disordered in the apo-structure [Fig. S1]. A reasonable scenario for substrate binding based on our structural data is therefore the following: biotin binds and induces the disorder-to-order transition of residues 63–77, which subsequently supplies important binding sites such as W74, R69, and A75 for binding ATP [Fig. 2(E,F)]; ATP binds and induces the folding of the residues 162–171 [Fig. 2(B)]. Should ATP bind first, its interaction with residues like W74 and R69 may occlude the biotin-binding pocket of the active site. In the ATP-bound structure of P. horikoshii BPL (1X01), the ligand pocket is indeed closed by ATP and prevents further binding of biotin.⁹

The residue Lys138 has been recognized as a key functional residue in BPLs and other proteins with BPL-LplA-LipB domains.^6,16–18 Lys138 is postulated to orientate the substrates biotin and ATP and stabilize the transition state of the intermediate.⁸ The P. horikoshii BPL mutant K¹¹¹A, analogous to the K¹³⁸S mutant of M. tuberculosis BPL, binds biotin and ATP, but cannot form biotinyl-5′-AMP.⁹ In that structure the phosphate of ATP is not in the optimal position to react with the biotin carboxyl group. In our ligand-bound BPL structure, Lys138 forms critical hydrogen bonds to the phosphate group of the AMP moiety and the carboxyl group of the biotin moiety [Fig. 2(E)]. The important role of Lys138 in BPL catalysis is strongly supported by all available structural data and is corroborated by our biochemical evidence demonstrating that substitution of this lysine in M. tuberculosis BPL abolishes its activity in vitro.

Materials and Methods

Cloning, protein expression, and purification

The open reading frame of the M. tuberculosis bpl/birA gene (Rv3279c) was amplified by PCR from genomic DNA using KOD polymerase with the sense primer 5'-AAAACCATGGCCGACCGCGATCGGCTCAG-3' and antisense primer 5'- ATTAGAATTCGCGCGAGTTAACGCAAATGCACCAC-3'. The PCR product was ligated to the NcoI and EcoRI sites of the expression vector pETM11, which contained an N-terminal poly-histidine tag, which was cleaved from the recombinant fusion protein with tobacco etch virus (TEV) protease. The construct was transformed into E. coli Rossetta (DE3)plysS cells for protein expression. Transformed bacteria from a single clone were cultured in Luria-Broth medium containing 50 μg mL⁻¹ kanamycin at 37°C until the culture reached an optical density of 0.6 at 600 nm. Protein expression was then induced with isopropylthiogalactopyranoside at a final concentration of 0.1 mM and the culture incubated overnight at 20°C. The bacteria were harvested by centrifugation at 3000g and 4°C. The culture pellets were resuspended in 20 mM Tris-HCl (pH 8.0), 300 mM NaCl, 10 mM imidazole, 0.02% (v/v) β-mercaptoethanol (BME), with mini-EDTA-free protease inhibitor (Roche) at concentrations recommended by the manufacturer. Cells were lysed by sonication using a Bandelin Sonoplus HD3200 sonicator set to pulse with an on-off cycle of 0.3–0.7 s and an amplitude of 45% for a total of 3 min. The sample was cooled on ice throughout. The lysate was centrifuged at 38,700g for 1 h at 4°C. The supernatant was loaded onto a nickel-nitrilotriacetic acid (Ni-NTA) resin that had a bead volume of 2 mL (Qiagen). The resin was washed with 20 column volumes of 20 mM Tris-HCl (pH 8.0), 300 mM NaCl, 10 mM imidazole, 0.02% (v/v) BME. Bound proteins were eluted with 20 mM Tris-HCl (pH 8.0), 200 mM NaCl, 300 mM imidazole, 0.02% (v/v) BME. The eluted protein was cleaved with 0.5 µM TEV protease overnight at 4°C, while dialyzing against 20 mM Tris-HCl (pH 8.0), 50 mM NaCl, 2 mM dithiothreitol (DTT), 1 mM EDTA. Contaminants in the dialyzed protein solution were removed by reloading the sample onto the Ni-NTA column, which bound the cleaved tag and the protease. The recombinant BPL was collected in the flow-through. The protein was further purified by gel filtration using a Superdex 75 column (GE Healthcare) equilibrated with 10 mM HEPES (pH 7.5), 50 mM NaCl, 1 mM DTT. The column was calibrated using the low molecular weight calibration kit from Amersham. To assess if the presence of substrate changed the oligomeric state of BPL we added 40 µM biotin and 3 mM ATP to the purified apo-BPL and incubated these for 30 min before re-loading the sample onto a Superdex 75 column (GE Healthcare). The BPL mutant, K¹³⁸S, was produced in the same manner as apo-BPL. The point mutation was introduced into the BPL coding sequence using the Stratagene site-directed mutagenesis kit using standard protocols recommended by the manufacturer.

X-ray structure determination

All crystals used in this study were grown by the hanging drop vapor diffusion method at 20°C. Apo-BPL crystals grew in drops composed of 0.1M Tris-HCl (pH 8.5), 0.2 mM trimethylamine N-oxide, 20% (w/v) PEG-2000 monomethyl ether. The crystallization drop was prepared by mixing 1 µL of this solution with 1 µL of 15 mg mL⁻¹ apo-BPL in 10 mM HEPES (pH 7.5), 50 mM NaCl, 1 mM DTT. Crystals of the biotinyl-5'-AMP-BPL complex were grown from equal volumes of 10 mM biotin, 10 mM ATP, and 21 mg mL⁻¹ BPL incubated at 37°C for 30 min prior to setting up the crystallization drop. The drop was prepared by mixing 6 µL of this sample with 1 µL well solution composed of 0.1M Tris-HCl (pH 7.8), 0.2M Li₂SO₄, 25% (w/v) PEG-4000. The crystals were mounted in loops directly from the drop, without cryo-protectant, and flash frozen in liquid nitrogen. X-ray diffraction datasets were collected at 100 K, using the synchrotron radiation beam line BW7A of the DORIS storage ring at EMBL/DESY, Hamburg, Germany. The data were processed with Denzo and Scalepack.¹⁹

The apo-BPL structure was solved by molecular replacement using CaspR²⁰ and the E. coli BirA structure 1HXD as the search model. Atomic positions and individual isotropic B-factors were refined using the default geometric restraints of the program Refmac5.²¹ TLS tensors were used to model the anisotropic effect. Five percent of the unique reflections were excluded form refinement and served as data for cross-validation to monitor model fitting. Model building between refinement cycles was performed with XtalView.²² The diffraction and refinement statistics are summarized in Table I. Residues 8–64, 77–162, 171–265 of chain A, and residues 8–65, 77–163, 169–265 of chain B could be modeled on electron density.

The structure of the biotinyl-5'-AMP-BPL complex was solved by molecular replacement using the refined apo-structure as search model, using the same refinement and modeling procedure as for the apo-structure. Geometric restraints of biotinyl-5'-AMP used during refinement and model building were generated using PRODRG.²³ All graphic representations of models were created using PyMol.²⁴

Electrophoretic mobility shift assay

After the first affinity chromatography step BPL was further purified by gel filtration using as buffer 10 mM Tris-HCl (pH 7.5), 0.1 mM DTT, 5% (v/v) glycerol, 200 mM KCl. A transformed E. coli strain containing a BCCP domain expression plasmid was grown in Luria-Broth media and purified as described.²⁵ The recombinant BCCP domain contains the 87 C-terminal residues of E. coli BCCP¹⁵ and has a molecular weight of 9.3 kDa. The recombinant BCCP domain was concentrated by vacuum drying to 60 µM. The biotinylation reaction was performed as described.²⁵ The biotinylation reaction mixture contained 40 mM Tris-HCl (pH 8.0), 3 mM ATP, 5.5 mM MgCl₂, 1 mM DTT, 100 mM KCl, 60 µM BCCP, 40 µM biotin and 2 or 8 μM of wild type BPL or K¹³⁸S. The proteins were preincubated for 20 min at 20°C and their mobility was analyzed in an 8% Tris-borate-EDTA native polyacrylamide gel.

Enzyme activity assay

We used a coupled enzyme assay that measured the quantity of AMP released during the second step of BPL's reaction.²⁶ Spectrophotometric data were recorded using a PowerWaveX Select spectrophotometer (Bio-Tek Instruments) and Greiner Bio-one UV-transparent, flat-bottom 96-well plates. Data was recorded using KC4 Kineticalc version 3.01 (Bio-Tek Instruments) and analyzed using GraphPad Prism 5 version 5.03 (Graphpad Software). The final concentration of reagents in one 300 µL reaction volume was 2.5 mM phosphoenolpyruvate, 0.2 mM NADH, 11.5 U myokinase, 9.9 U pyruvate kinase, 12.3 U lactate dehydrogenase, 100 µM biotin, 100 µM BCCP, 80 mM Tris-HCl (pH 8.0), 400 mM KCl, 11 mM MgCl₂, 0.2 mM DTT, and 0.2 mg mL bovine serum albumin. The concentration of ATP was varied between 0.1 and 4 mM. The concentration of protein used in each reaction was 10 nM BPL or K¹³⁸S. The reaction was initiated upon addition of biotin and the consumption of NADH monitored at 340 nm at 30°C for 1 h.

Glossary

BCCP

biotin carboxyl-carrier protein

BirA

biotin induced repressor A

BME

β-mercaptoethanol

BPL

biotin protein ligase

TEV

tobacco etch virus.

Supporting Information

Additional Supporting Information may be found in the online version of this article.

Supplementary Information Figure S1.