X-ray crystal structure of the COG0325 protein YlmE from Streptomyces coelicolor A3(2)

Cover Page

The following handle holds various files of this Leiden University dissertation:

http://hdl.handle.net/1887/63154

Author: Tassoni, R.

Title: Structural characterization of bacterial proteins involved in antibiotic resistance and peptidoglycan biosynthesis

Issue Date: 2018-06-27

6. X-ray crystal structure of the COG0325 protein YlmE from Streptomyces coelicolor A3(2).

Abstract

COG0325 members are Fold Type III, pyridoxal phosphate (PLP)-binding proteins that are very conserved throughout the phylogenic tree, from bacteria to yeast and mammals, but whose biochemical function is still unknown. Sequence annotation revealed that COG0325 proteins are most similar to the N-terminal domain of alanine racemases, and thus, they are annotated as putative alanine racemases.

Here, the gene encoding for the COG0325 protein YlmE from Streptomyces coelicolor A3(2) was cloned, and heterologously expressed in Escherichia coli. The recombinant protein was purified, crystallized, and the X-ray crystal structures of apo- and holo-YlmE were determined. YlmE folds as a classical (α/β)8 barrel with the PLP reversibly bound to Lys40 in the core of the structure. Since no binding could be detected by ITC and crystal soaking between YlmE and D- or L-Ala, other amino acids, and small metabolites, the role of YlmE as a racemase was questioned. Despite the overall structural similarity of YlmE to Alr, the proteins show major differences in the residues surrounding the PLP cofactor, and notably in the oligomerization state in solution. In conclusion, the cellular function of COG0325 and YlmE still remains elusive.

Introduction

YlmE from Streptomyces coelicolor A3(2) is a member of the COG0325 family, that comprises Fold Type III, pyridoxal phosphate (PLP)-binding proteins widely distributed among the majority of bacterial species and most eukaryotes, including yeast, Caenorhabditis elegans, Drosophila, Arabidopsis, Zea mays, zebrafish, chicken, and mammals including humans [215]. The first published X-ray crystal structure of a COG0325 family member was that of YBL036c (protein data bank (PDB) entries 1CT5 and 1B54) from Saccharomyces cerevisiae, that was selected for structural studies in the context of the Structural Genomics Project with the stated proposition that the three-dimensional structure rather than the sequence could better aid in identifying its biological function [67]. After YBL036c, the structures of homologous proteins YggS from Escherichia coli (PDB 1W8G), PipY from Synechococcus elongatus (PDB 5NLC and 5NM8), and hypothetical proteins from Agrobacterium tumefaciens and Bifidobacterium adolescentis (PDB 3R79 and 3CPG, respectively), and an engineered, target protein OR70 from the Northeast Structural Genomics Consortium (PDB 3SY1) were deposited in the PDB. There is high sequence and structural conservation between COG0325 proteins from different species (Fig. 1), and, notably, between COG0325 proteins and the N- terminal domain of alanine racemases (Alr’s). YlmE from S. coelicolor shares 23.3% sequence identity (44.7% similarity) with the N-terminal domain of the Alr from the same microorganism, and was tentatively annotated as a putative Alr.

However, the putative racemase activity of YlmE was not supported by experimental evidence.

In this work, the gene SCO2080 encoding for YlmE in S. coelicolor A3(2) was cloned, heterologously expressed in E. coli, crystallized, and the X-ray crystal structures of apo- and holo-YlmE were solved. To study if PLP is essential for protein folding, a YlmE mutant was produced which carried a Met in position 40 instead of the PLP-binding Lys residue. Binding of YlmE to D- and L-Ala, and other amino acids and small metabolites was investigated by Isothermal Titration Calorimetry (ITC) and by crystal soaking, but no evidence of binding to any of the tested compounds was obtained. Thus, the crystal structures of YlmE and Alr from S. coelicolor (chapter 5) were compared with a particular focus on the PLP-binding site, to try to understand the differences in the activity of the two proteins.

Figure 1. Multiple amino acid sequence alignment of the COG0325 proteins from different species. Sco

= Streptomyces coelicolor A3(2); Sce = Saccharomyces cerevisiae; Eco = Escherichia coli; Bsu = Bacillus subtilis; Hss = Homo sapiens. The sequence alignment showing structural elements of YlmE was generated with ESPript3.0. α-helices are shown as large coils labeled α, 310-helices are shown as small coils labeled η, β-strands are shown as arrows labeled β and β-turns are labeled TT. Identical residues are shown on a red background, conserved residues are shown in bold on a yellow background.

Materials and Methods

Gene cloning, protein expression and purification

The construct encoding for YlmE fused to a N-terminal histidine tag (NHis-YlmE) was made by Lizah van der Aart by amplifying wild type ylmE (SCO2080, appendix 1) from the genomic DNA of Streptomyces coelicolor A3(2) using primers 2080FW and 2080RV (Table 1), and cloning it into pET-15b with NdeI and BamHI as restriction sites. To improve expression and make mutagenesis easier, the ylmE sequence was codon optimized for expression in E. coli (appendix 1) and the synthetic gene was purchased from GeneArt Gene Synthesis (Invitrogen). Primers EC_FWD and EC_REV (Table 1) were used for amplification of the gene and subsequent cloning into pET-28a(+) using restriction enzymes NcoI and XhoI, which resulted in a C-terminal His-tagged YlmE (CHis-YlmE). Point mutation of Lys40 into Met was done using mutagenic primers K40M_FWD and K40M_REV (Table 1). The mutated amplicon was sub-cloned into pET-28a(+) as for the wild type gene with a C-terminal His-tag (CHis-K40M-YlmE).

Table 1. Primers.

Primer name Sequence (5’-3’)

2080FW ctaggaattcatatgacggaccgtaagcacgaactc

2080RV gactggatccttacccgagcctgggtcggactc

EC_FWD ctgtgcaaccatggggaccgatcgtaaac

EC_REV ctgtgcaactcgagacccagacgag

K40M_FWD cctgattgttgttaccatgacctatccggc

K40M_REV gccggataggtcatggtaacaacaatcagg

All constructs for expression of YlmE were transformed into Escherichia coli BL21 (DE3)pLysS electro-competent cells. E. coli cultures were incubated in LB medium at 37°C until an OD600 of 0.6 was reached, and gene expression was induced by addition of 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) followed by incubation overnight at 16°C. The cultures were pelleted at 5500´g and re- suspended in 20 mM Tris-HCl, pH 8.0, 10 mM NaCl. Bacteria were disrupted by French-press and the solution was ultra-centrifuged at 25000´g for 45 min at 4°C.

The supernatant was applied to a 5-ml pre-packed HisTrap HP column (GE Healthcare) pre-equilibrated in 20 mM sodium phosphate buffer, pH 8.0 and 500 mM NaCl. The column was washed with ten column volumes of the same buffer solution containing 50 mM imidazole. His-tagged recombinant proteins were eluted with a concentration of 250 mM imidazole. The fractions containing YlmE were loaded on a PD-10 desalting column (GE Healthcare) for fast removal of the imidazole and the protein was further purified by size exclusion chromatography using a Superose-12 column (GE Healthcare). The collected fractions were analyzed by SDS-PAGE and those containing pure YlmE were pooled together, concentrated using a Centriprep centrifugal filter unit (10 kDa cut-off, Millipore),

and flash-frozen in liquid nitrogen. The final buffer for NHis-YlmE was 25 mM MES pH 6.5, 100 mM NaCl, whereas CHis-YlmE and CHis-K40M-YlmE were stored in 20 mM Tris-HCl, pH 8.0, 100 mM NaCl. The purification yielded ~30 mg, ~20 mg, and ~5 mg of protein per L of bacterial culture for NHis-YlmE, CHis-YlmE, and CHis-K40M-YlmE, respectively.

Crystallization conditions

Purified NHis-YlmE, CHis-YlmE, and CHis-K40M-YlmE were concentrated up to 10, 5, and 8 mg ml^-1 respectively. Crystallization conditions were screened by sitting-drop vapor-diffusion using the commercial screens JCSG+ and PACT premier (Qiagen) at 20°C with a drop size of 500 nL. The reservoir solution (75 μL) was pipetted by a Genesis RS200 robot (Tecan) and the drops were made by an Oryx6 robot (Douglas Instruments). After ~2 months, NHis-YlmE crystals grew in condition number G1 of JCSG+, which consisted of 0.1 M HEPES, pH 7.0, 30% (w/v) Jeffamine ED-2001. Crystallization was optimized by a pH and buffer system screening, and crystals grew in 0.1 M MES, pH 6.5, 30% (w/v) Jeffamine ED-2001. However, crystallization conditions were not satisfactorily reproducible, and would not allow to efficiently screen for ligand binding by soaking. More reproducible crystallization conditions were achieved after additive screening using the Additive Screen HR2-428 Reagents (Hampton) in 0.1 M MES, pH 6.5, 30% (w/v) Jeffamine ED-2001, 10 mM praseodymium(III) acetate hydrate. Pr(III) acetate hydrate led to rapid nucleation and formation of small needle crystals, which were diluted into fresh crystallization buffer, crushed by vortexing with MicroSeed Beads (Molecular Dimensions), and seeded into fresh drops without any additives. Seeding proved to be a reproducible crystallization protocol, and yielded good quality crystals of NHis-YlmE. Quality diffracting crystals were also obtained by mixing purified NHis-YlmE at a final concentration of 4 mg mL^-1 with 0.1 M MES, pH 6.5, 30% (w/v) Jeffamine ED-2001, and 10 mM praseodymium(III) acetate hydrate in a ratio of 1:1:0.5. CHis-YlmE crystallized in condition E12 of the PACT premier screening, which consisted of 0.2 M disodium malonate and 20%

PEG3350. The structure of CHis-YlmE was obtained by diffracting the crystals grown in the initial screening. Crystals of CHis-K40M-YlmE grew in condition H10 of the JCSG+ screening, which consisted of 0.1 M BIS-Tris, pH 5.5, 25%

PEG3350, 0.2 M ammonium acetate. Crystallization conditions were optimized, and crystals for diffraction were prepared in 0.1 M BIS-Tris, pH 5.5, 20%

PEG3350, 0.2 M ammonium acetate. All crystals were verified to be protein using a fluorescence microscope with a U-MWU2 filter (Olympus). The crystals were mounted on cryo-loops and soaked in a solution that consisted of mother liquor and 10-15% glycerol before flash-cooling in liquid nitrogen. Prior to cryo-cooling, a number of wild type or K40M-YlmE crystals were soaked in possible ligands (Results, Table 4 for complete list) in order to test for binding.

X-ray data collection

X-ray data collection was performed at the ESRF (Grenoble, France). An overview of the data collection and refinement statistics for the four data sets is presented in Table 2. All data sets were auto-processed by the EDNA Auto processing package that used XDS [127] to integrate the intensities and AIMLESS [68] to scale and merge the intensities. The structures were solved by molecular replacement with MOLREP [261] using 1W8G as a search model from the CCP4 suite [277], and manually refined using REFMAC [189] and Coot [64]. Where suggested by the data, multiple conformations were used for amino acid side chains, and the B factors were refined for each conformation.

Analytical gel filtration

Superdex 200 10/300 GL analytical gel filtration column (GE Healthcare) was equilibrated in 20 mM Tris-HCl, pH 8.0, 50 mM NaCl at a flow rate of 0.2 ml min^-1 and calibrated using Gel Filtration Standard #151-1901 (Bio-Rad). CHis-YlmE (100 μL), either free or mixed with pulled-down, purified nucleic acids, was loaded on the column. The eluate was simultaneously monitored at 280, 260, and 230 nm by attaching the column to an ÄKTA purifier or an ÄKTA pure system (GE Healthcare).

Native-Polyacrylamide Gel Electrophoresis (PAGE)

Native polyacrylamide gels with polyacrylamide concentrations of 7.5, 10, and 12.5% were loaded in native conditions with 10 μL YlmE sample at a final protein concentration of 2 or 4 mg mL^-1. Gels were electrophoresed at 180 V for 30 min, and proteins were stained by Coomassie Brilliant Blue G-250 (Bio-Rad).

Isothermal Titration Calorimetry (ITC)

Isothermal Titration Calorimetry (ITC) experiments were performed on a MicroCal VP-ITC (Malvern) instrument. Data analysis was done in Origin using the software provided by Malvern. Before each experiment, proteins were dialyzed overnight at 4°C, and concentration was accurately measured by Nanodrop 2000 Spectrophotometer (Thermo Fisher Scientific) using extinction coefficients of 9970 M cm^-1 for YlmE and 61,679 M cm^-1 for Alr. The ligands tested for binding were weighed and dissolved in the dialysis buffer. Samples were thoroughly degassed prior to loading in the sample cell or syringe. All experiments were conducted at 25°C, with an injection program consisting of a first 2-μL injection, followed by injections of 7 μL, with a spacing of either 240 or 300 s. All ITC experimental conditions are summarized in Table 3.

Table 2. Data-collection and refinement statistics.

Apo-NHis- YlmE

Holo-NHis- YlmE

CHis-K40M- YlmE

Holo-CHis- YlmE

Beamline ID23-1 ID29 ID29 ID30B

Wavelength (Å) 0.968622 0.979235 1.07227 0.979312

Detector Pilatus 6M_F (DECTRIS)

Pilatus3_6M (DECTRIS)

Pilatus 6M_F (DECTRIS)

Pilatus3_6M (DECTRIS)

Frames 1007 2440 2400 2400

Oscillation (°) 0.15 0.05 0.15 0.05

Exposure time per

frame, total (s) 0.049, 49.343 0.02, 48.8 0.037, 88.8 0.02, 48 Resolution range

(Å)

44.04-1.43 (44.04-5.54/

1.48-1.43)

42.97-1.79 (42.97-6.93/

1.85-1.79)

44.01-1.41 (44.01-5.46/

1.46-1.41)

45.93-2.13 (45.93-8.25/

2.20-2.13) Space group P 1 21 1 P 21 2 21 P 1 21 1 P 65 2 2

Unit-cell parameters a, b, c, α, β, γ

44.93, 37.99, 62.39, 90.00, 101.44, 90.00

37.58, 45.05, 128.92, 90.00,

90.00, 90.00

44.99, 38.03, 62.67, 90.00, 101.97, 90.00

106.06, 106.06, 88.73, 90.00, 90.00, 120.00 Number of

observations

105290 (1890/7153)

90443 (1617/8394)

119069 (2283/5611)

212624 (3364/18709) Number of unique

reflections

36863 (679/2983)

21270 (438/1997)

38010 (737/2589)

16844 (340/1596) Completeness (%) 96.0

(95.0/79.8)

99.4 (97.6/98.5)

94.5 (99.3/66.4)

99.4 (96.3/99.3)

Rpim 0.039

(0.029/0.512)

0.062 (0.039/0.421)

0.027 (0.011/0.670)

0.073 (0.036/0.670)

CC(1/2) 0.997

(0.996/0.647)

0.992 (0.988/0.568)

0.999 (0.999/0.531)

0.991 (0.991/0.561)

áI/s(I)ñ 9.7

(24.4/1.4)

8.0 (18.7/1.7)

12.9 (51.8)

6.1 (17.5/0.9) Multiplicity 2.9

(2.8/2.4)

4.3 (3.7/4.2)

3.1 (3.1/2.2)

12.6 (9.9/11.7)

R factor (%) 21.5 25.8 20.9 26.5

Rfree (%) 22.1 26.9 21.3 26.9

RMS Z scores

Bond lengths (Å) 1.08 0.95 1.01 0.72

Bond angles (°) 1.12 1.08 1.13 0.91

Number of atoms 1952 1816 1886 1798

Protein 1789 1678 1706 1725

PLP - 15 - 15

Jeffamine (JEF) 5 - - -

PEG - - 27 -

Water 158 123 153 58

Ramachandran plot (%)

Preferred regions 97 98 100 98

Allowed regions 3 2 0 2

Outliers 0 0 0 0

Numbers in parentheses refer to inner/outer shell.

Table 3. ITC experimental conditions.

Sample cell Syringe

NHis-YlmE concentration

(μM)

Molecule Conc

(mM) Buffer pH

70 D-alanine 10 20 mM sodium phosphate 7.0

70 L-alanine 10 20 mM sodium phosphate 7.0

116 β-alanine 10 50 mM Tris-HCl, 100 mM NaCl 8.0

88 D-glutamate 10 20 mM sodium phosphate 7.0

98 L-glutamate 10 20 mM sodium phosphate 7.0

85 L-glutamine 10 20 mM sodium phosphate 8.0

40 L-arginine 5 25 mM Tris-HCl, 10 mM NaCl 8.0

84 L-valine 10 50 mM Tris-HCl, 100 mM NaCl 8.0

40 L-proline 5 25 mM Tris-HCl, 10 mM NaCl 8.0

116 L-carnosine 10 50 mM Tris-HCl, 100 mM NaCl 8.0 124 D-lactate 100 50 mM Tris-HCl, 100 mM NaCl 8.0 124 α-ketoglutarate 100 50 mM Tris-HCl, 100 mM NaCl 8.0

98 Pyruvate 10 20 mM sodium phosphate 7.0

116 Pantothenate 10 50 mM Tris-HCl, 100 mM NaCl 8.0

85 GABA 10 20 mM sodium phosphate 8.0

44 ATP 1 50 mM Tris-HCl, 10 mM NaCl 7.5

59 GlcNAc 5 50 mM Tris-HCl, 10 mM NaCl 7.5

Results Holo-YlmE

The X-ray crystal structure of YlmE in complex with the PLP cofactor was solved for the two constructs with the His-tag either at the N- or C-terminus. For both datasets, phases were found by molecular replacement (MR) using PDB 1W8G as the search model, and the structures were refined to a resolution of 1.79 Å in space group P 2₁ 2 2₁ for NHis-YlmE, and to a resolution of 2.13 Å in space group P 6₅ 2 2 for CHis-YlmE (Table 2). Both constructs crystallized as monomers, with only a single YlmE chain in the asymmetric unit (AU) (Fig. 2a,b). In the NHis-YlmE structure, good electron density was found for most protein residues, and amino acids 3-233 were included in the model. However, poor electron density was found

Figure 2. Crystal structure of holo-YlmE, obtained from the NHis-YlmE (a) and CHis-YlmE (b) constructs. Both models are in ribbon representation colored according to secondary structure element.

(a) Β-strands are colored in lawn green, α-helixes in orange, and loop regions in grey. Labels indicate secondary structure elements. (b) Β-strands are colored in red, α-helixes in green, and loop regions in

b a

c d

grey. The N- and C-termini of the protein are indicated as N and C, respectively, in both figures. (c) The pyridoxal phosphate (PLP) cofactor is covalently bound to the side chain of Lys40, and is stabilized in the active site by hydrogen bonds (dashed lines) with Met210, Ser211, Gly228, and Thr229. PLP and the residues at hydrogen bond distance or at a distance <4 Å are shown as sticks colored according to atom type (C in green for PLP and grey for amino acids, N in blue, O in red, S in yellow, and P in magenta). The 2mFO-DFC electron density map (blue chicken wires, 1 σ) is clipped around the PLP cofactor. The image was produced using the X-ray crystal structure of holo-NHis-YlmE. (d) Superposition of the PLP-binding sites in holo-NHis-YlmE (teal) and holo-Chis-YlmE (grey).

around residues Glu135, Glu136, Gly137, Gly138, Arg139, and Gly140, and were not modelled in the structure as they are most likely part of a flexible loop. The electron density of CHis-YlmE allowed to model residues 2-235, including residues 135-139, that were visible, although they had high B-factors and poor electron density around the side chains. YlmE folds as a (α/β)₈ barrel, with the core of the barrel formed by eight parallel β-strands (β1-8), surrounded by eight α-helices (α1- 8) (Fig. 2a). In both holo-structures, positive electron density in the mF_O-DF_C map clearly indicated the presence of the PLP cofactor covalently bound to Lys40 via an internal aldimine bond between the aldehyde group of PLP and Nε of the Lys side chain (Fig. 2c). The PLP is stabilized by key hydrogen bonds between the guanidinium group of Arg226 and the pyridinic nitrogen of PLP, and between Ser211, Gly228, and Thr229, and the phosphate moiety of the cofactor. The position of the His-tag at the N- or C-terminus of YlmE did not influence the folding of the protein, and the final models of NHis- and CHis-YlmE superpose well with an average root-mean-square deviation (rmsd) between corresponding Cα atoms of 0.60 Å. The average rmsd was found to be 0.16 Å when comparing only the Cα atoms of the residues surrounding the PLP, as shown by the good superposition of the PLP-binding sites of the two proteins (Fig. 2d).

Apo-YlmE

Crystallization of NHis-YlmE without PLP allowed the structural solution of apo- YlmE. The initial model was obtained by MR using PDB 1W8G as the search model, and the structure was refined to a resolution of 1.43 Å in space group P 1 21 1 (Table 2). Apo-YlmE crystallized as a monomer, with only one protein chain in the AU. Good electron density was found for most of the structure, and residues 5-238 could be modelled. However, poor electron density was found around residues Glu135, Glu136, Gly137, Gly138, Arg139, and Gly140, as in the structure of holo-NHis-YlmE. No electron density could be found for the PLP cofactor. Since the protein plated for crystallization was mostly bound to PLP, it is possible that either apo-YlmE was present in small amount in the purified protein and preferentially crystallized in the conditions used, or the PLP was lost during incubation for crystallization due to time or reactivity with buffer components, i.e.

Jeffamine or HEPES.

The structure of apo-NHis-YlmE shows good superposition to the structure of holo-NHis-YlmE, with an average rmsd difference of 0.52 Å between corresponding Cα atoms (Fig. 3a). Major differences were found in the α-helix α8 and in the C-terminus of the protein, which is more distant from the PLP-binding site in the structure of apo-YlmE than in holo-YlmE (Fig. 3b). It is possible that in the presence of the cofactor, the C-terminus of the protein is hindered by hydrogen bonds between Thr229 and the phosphate moiety of the cofactor. In the apo-YlmE structure, Thr229 is not anchored to the PLP, and its Cα undergoes a shift of 1.19 Å. The same shift of the C-terminus of the protein was observed in the structures of apo- and holo-PipY from Synechococcus elongatus [258].

Interestingly, the α-helix α8 gets closer to the PLP-binding site when PLP is not present (Fig. 3b).

Figure 3. Structural comparison of apo- and holo-NHis-YlmE. (a) Superposition of apo-YlmE (orange) and holo-YlmE (dark cyan), both in ribbon representation. The PLP cofactor and its surrounding residues are represented as sticks colored according to atom type (C in dark cyan, O in red, N in blue, P in magenta, and S in yellow). (b) Zoom on the PLP-binding site of YlmE. C indicates the C-terminus of holo-YlmE.

CHis-K40M-YlmE

To further investigate the structural differences between apo- and holo-YlmE, Lys40 was mutated into a Met residue, so that the protein could not link to PLP.

The resulting mutant construct CHis-K40M-YlmE crystallized, and crystals diffracted to a resolution of 1.41 Å. The structure was refined in space group P 1 21

1 with a final R-factor of 20.9 % and Rfree of 21.3 % (Table 2, Fig. 4a). Amino acidic residues from position 3 to 236 could be modelled in the electron density, but, like for the structures of apo- and holo-NHis-YlmE, poor electron density was found for residues 135-139, which were not included in the final model of the protein. The Lys40Met mutation did not affect the overall fold of the protein. The structures of

a b

K40M-YlmE and of apo- NHis-YlmE can be superposed with an average rmsd value between Cα atoms of 0.24 Å. Superposition of K40M-YlmE with the structures of holo-NHis- and CHis-YlmE yields an average rmsd of 0.51 Å and 0.69 Å, respectively. Like for apo-NHis-YlmE, the C-terminus of the protein and the α-helix α8 were shifted outward of the PLP-binding site (Fig. 4b). The structure of K40M-YlmE confirmed that the absence of PLP causes subtle structural rearrangements of the PLP-binding site. It is not known is these structural shifts have a biological significance.

Figure 4. Crystal structure of mutant CHis-K40M-YlmE. (a) The protein is represented as ribbons colored according to secondary structure elements (β-strands in dark purple, α-helixes in yellow, and loop regions in grey). The N- and C-termini are indicated as N and C, respectively. Superposed to apo- NHis-YlmE. (b) Superposition of CHis-K40M-YlmE (olive, green sticks) and apo-NHis-YlmE (brown, orange sticks).

Oligomerization state of YlmE in solution

With the exception of the ornithine decarboxylase paralogue [129], Fold Type III, PLP-binding proteins form homodimers in a similar way to the example given for Alr (chapter 5). The dimerization happens via head-to-tail interaction between two molecules and is mediated by an additional domain to the α/β-barrel, such as the C-terminal domain of Alr (chapter 5). However, YlmE lacks such a domain, so that the oligomerization state of the protein cannot be easily predicted. Here, the oligomerization state of YlmE in solution was investigated by analytical size exclusion chromatography (SEC) and by native-PAGE electrophoresis. Based on the single elution peak of YlmE (~2 mg mL^-1) by SEC (Fig. 5a), a molecular weight of approximately 21.5 kDa could be calculated, which is comparable to the theoretical molecular weight of monomeric YlmE, that is 26.3 kDa. However, the results from SEC are not supported by native-PAGE electrophoresis, which show

a b

the formation of four migration bands (Fig. 5b), and suggest the existence of different oligomerization states for YlmE. To gain further information on YlmE oligomerization state in solution, experimental results were compared to theoretical

Figure 5. Oligomerization state of YlmE. (a) Chromatogram from the analytical size exclusion chromatography of YlmE. YlmE eluted as a single peak with an elution volume of 15.7 mL, corresponding to a molecular weight of 21.5 kDa, suggesting a monomeric protein. (b) By native-PAGE, YlmE was shown to exhibit at least four bands. The same electrophoretic pattern was observed for YlmE concentrations of 10 and 5 mg mL^-1 loaded either on a 12.5%, 10% (not shown), or 7.5%

polyacrylamide gel. The purity of the samples was assessed by SDS-PAGE on a 12.5% polyacrylamide gel. (c) Ribbon representation of a putative dimer calculated by the Eppic webserver using the structure of holo-CHis-YlmE as input. The two subunits are colored in light blue and coral. N and C indicate the N- and C-termini, respectively. The PLP cofactors are represented as sticks and are colored according to atom type (C in green, N in blue, O in red, and P in magenta). Residues Leu12, Arg30, Gln31, and Gly54 that are predicted to mediate the protein-protein interaction are shown in sticks and colored according to atom type. (d) Closer view of the dimerization interface of the dimer shown in (c), with one subunit shown as a coral ribbon and the other in surface representation colored according to the electrostatic potential.

a b

c d

-25.00 25.00 75.00 125.00 175.00 225.00 275.00 325.00 375.00 425.00 475.00 525.00 575.00

14.50 15.00 15.50 16.00 16.50 17.00 17.50

Abs (mAU)

Elution volume (mL)

Free YlmE

10 mg mL^-1 5 mg mL^-1 10 mg mL^-1 5 mg mL^-1

12.5% Native-PAGE 7.5% Native-PAGE 12.5 % SDS-PAGE 31

6.5 14.4 45 66.7 116.3 200

Peak fractions ^kDa

calculations performed on the web servers PDBePISA⁴ and Eppic⁵ [58]. Both PDBePISA and Eppic aim at estimating the biological significance of protein assemblies based on the structure of a protein. Calculations were done by inputting the X-ray crystal structures of both NHis- and CHis-YlmE in the holo- state. For both structures, the PDBePISA server could not find any significant biological oligomerization states of YlmE, while Eppic found one possibly significant dimerization mode when the crystal structure of holo-CHis-YlmE was used in the calculation. The dimerization interface calculated by the Eppic server involves residues Leu12, Arg30, Gln31, Gly54, close to the α1-helix in both the interacting monomers (Fig. 5c,d). Nevertheless, the dimerization interface calculated by the Eppic web server would limit YlmE oligomerization to a dimer, but results from native-PAGE electrophoresis suggests at least four different oligomers. Thus, if higher oligomeric states of YlmE exist, it remains to be elucidated how they form. One possibility is that polymerization is mediated by the positive and negative potential electrostatic patches present of the surface of YlmE (not shown). However, experimental evidence is required before drawing any conclusions. The fact that YlmE elutes as a monomer by SEC suggests that protein-protein interactions are rather weak. Possibly, the presence of a substrate of YlmE might strengthen dimers and/or higher aggregation states. If that were the case, it could help to elucidate which one is functionally relevant. However, this experiment still cannot be performed because no biological function is assigned to YlmE.

Amino acid and small metabolite binding

YlmE shares 23.3% sequence identity and 44.7% sequence similarity to the N- terminal domain of Alr, and has been annotated as a putative Alr. The yeast COG0325 homolog YBL036c was reported to catalyze the L- to D-Ala racemization, and a role as broad spectrum racemases rather than solely Alr was suggested for COG0325 proteins [67]. Based on this hypothesis, recombinant NHis-YlmE was tested for binding to several amino acids, small molecules and metals (Table 4 for complete list) by Isothermal Titration Calorimetry (ITC) and/or by crystal soaking. However, none of the tested compounds showed binding to YlmE. Although these experiments provide substantial evidence for the absence of binding, binding of YlmE to one or more of the tested compounds cannot be ruled out only on the basis of ITC and crystallographic soaking. ITC can produce false negatives, especially in the case of low-affinity binding events, which can have K_D constants in the mM range. As for crystallographic soaking, ligand permeability and binding might be affected by several factors, including crystal packing or solubility of the ligand itself. Thus, different experimental techniques should be employed

4 http://www.ebi.ac.uk/msd-srv/prot_int/cgi-bin/piserver

5 http://eppic-web.org/ewui

before drawing any definitive conclusions. Since previous literature studies reported Ala racemization by homologous COG0325 proteins [67,137], binding of YlmE to L- and D-Ala should be tested further, for example by NMR.

Table 4. Ligand binding screenings.

Compound ITC Crystal

soaking Compound ITC Crystal

soaking Compound ITC Crystal soaking

D-alanine ✓ 2-pyrrolidone-5-

carboxylate ✓ 5-aminopantothenate ✓

L-alanine ✓ L-carnosine ✓ GABA ✓

β-alanine ✓ ✓ D-lactate ✓ ✓ ATP ✓

D-glutamate ✓ D-glucose ✓ GlcNAc ✓ ✓

L-glutamate ✓ ✓ Sucrose ✓ MurNAc ✓

L-glutamine ✓ α-ketoglutarate ✓ ✓ Bactopeptone ✓

Pyroglutamate ✓ Pyruvate ✓ CuCl2 ✓

L-arginine ✓ ✓ D-cycloserine ✓ Cl2Co ✓

L-valine ✓ Pantothenate ✓ ✓ CuSO4 ✓

L-proline ✓ ✓ α-lactose ✓

Structural comparison or YlmE and Alr

YlmE sequence similarity to the N-terminal domain of Alr suggests that the two proteins also share structural features. Thus, the X-ray crystal structures of YlmE and Alr were compared in detail in order to identify key differences or similarity that might also have an effect on the biological function of the proteins. As predicted based on their sequence similarity, the overall fold of YlmE resembles the N- terminal domain of Alr (Fig. 6a,b). A few exceptions include the first α-helix of the barrel, that is longer in YlmE than in Alr, and the flexible loop of YlmE (residues 135-139), that is missing in Alr. Although the fold topology of YlmE and Alr is conserved, the two α/β-barrels superpose poorly, with a rmsd of 18.1 Å between corresponding Cα atoms. Remarkably, in both structures, the PLP is localized in the same position in the middle of the upper aperture of the barrel, but little conservation is shared by amino acidic residues in proximity of the cofactor. Only three amino acids are conserved, which are the Lys that directly binds PLP (Lys40 in YlmE, Lys46 in Alr), and the Arg and Ser residues that are involved in the stabilization of the cofactor in the active site by hydrogen bonding to the nitrogen of the pyridine ring in the case of Arg (Arg226 in YlmE, Arg237 in Alr) and one oxygen of the PLP phosphate in the case of the Ser (Ser211 in YlmE and Ser222 in Alr). The identity of the residues around the PLP plays a crucial role in the selection of the ligand by steric hindrance, and the relevance of the low conservation observed between YlmE and Alr cannot be underestimated.

Moreover, the entryway to the active site in Alr was shown to be a narrow cavity formed after dimerization of the protein (Fig. 6d). Instead, YlmE appears to be mostly a monomer in solution, and the PLP-binding site, probably the active site of the protein, is exposed on the surface in direct contact with the solvent (Fig. 6c).

Figure 6. Comparison of the structures of YlmE and Alr. (a), (b) Superposition of Alr (PDB 5FAC, orange) and YlmE (light blue) in ribbon representation. (c), (d) Surface representation of YlmE (c, light blue) and Alr (d, subunit A in coral is bound to PLP, subunit B in dark cyan). The PLP cofactor is shown in stick representation colored according to atom type (C in green, O in red, N in blue, and P in magenta). In YlmE, the PLP is in contact with the solvent, while in Alr it is buried at the end of the entryway to the active site.

a b

c d

Discussion

YlmE from S. coelicolor A3(2) is a member of the COG0325 family currently with an unknown biological function. Here, the X-ray crystal structure of YlmE was solved both in the apo- and holo-state. YlmE is a (α/β)8 barrel protein, belonging to the Fold Type III, PLP-binding family. The PLP cofactor is covalently bound to Lys40, in the middle of the upper aperture of the barrel core. As previously observed for other COG0325 crystal structures [67], the structure of YlmE deviates from the typical (α/β)₈ barrel in that the first element of the structure is a α-helix instead of a β-strand. Having an α-helix as first secondary structure element of the barrel is a common feature of Fold Type III PLP-binding proteins, e.g. ornithine decarboxylases [133], and Alr’s [231]. The (α/β)₈ barrel, or triosephosphate isomerase (TIM) barrel, from the first enzyme found to have this fold, is one of the most common folds in the PDB [275]. TIM barrel enzymes catalyze very diverse reactions and often present a phosphate-binding site, like the PLP-binding protein Alr [191]. Unlike most Fold Type III, PLP-binding proteins, such as Alr or ornithine decarboxylase (Odc), YlmE lacks the β-strand domain that in the other proteins mediates dimerization. Hence, YlmE is likely to be a monomeric protein. The two YlmE constructs presented in this work differed for the position of the His-tag, which was either at the N- or at the C-terminus of the heterologous protein. Both constructs crystallized as monomers, and the overall fold is fully conserved, showing that the position of the His-tag does not affect the structure of the protein, nor changes its oligomerization state in the crystals. The monomeric state of the protein in solution was confirmed by analytical gel filtration, which gave an apparent molecular weight of 21.5 kDa for the CHis-YlmE (the calculated molecular weight is 26.3 kDa). Our results based on His-tagged YlmE find confirmation in previously published structures (PDB 1W8G, 3SY1, 3R79, 3CPG 5NLC, 5NM8), and in particular the structure of YBL036c (PDB 1CT5 and 1B54) [67], which did not contain any purification tags and still crystallized and behaved as a monomer. Nevertheless, YlmE formed four bands when electrophoresed in native conditions. Similarly, Knight et al. [137] showed heterogeneous patterns by native-PAGE for COG0325 members corresponding to PDB entries 1CT5, 3R79, 3CPG, and 3SY1, but did not investigate the polymerization modality of these proteins. TIM barrels are known to form very diverse oligomerization states that were suggested to favor their thermodynamic stability [137,224]. Even though biologically relevant dimerization interfaces were not found by PBDePISA, the Eppic webserver found a possible dimerization site that involves residues of the a1-helix. However, the polymerization modes of YlmE were not further investigated in this study.

In vitro experiments, ITC and crystal soaking, were performed for detecting the binding of YlmE to amino acids and other small molecules. Under the experimental conditions used, YlmE did not show any binding to the tested

compounds, in agreement with previous reports by Ito and colleagues [119], who could not detect racemase or other enzymatic activity by the homologous, recombinant protein YggS from E. coli towards any of the tested amino acids.

Nevertheless, other studies [67,137] reported the racemization of D- to L-Ala for COG0325 protein members. The authors argued that the observed racemase activity is characteristic to only a particular polymerization state of the COG0325 proteins, and not to the monomeric proteins, but the nature of the catalytically active aggregation state was not explicated [137]. The YlmE fold is not dependent on the covalent bond with PLP, but small structural differences were observed between the crystal structures of apo- and holo-YlmE in regions close to the PLP- binding site. Similar structural rearrangements were also observed in the apo- and holo-structures of PipY [258], and they were brought in support of the hypothesis that COG0325 proteins may have a role in delivering PLP to PLP-dependent proteins [47,137,143,258]. Indeed, YlmE reversibly binds PLP, similar to other PLP-dependent proteins. Thus, further evidence must be gathered before a function in PLP-delivery can be confirmed or confuted for YlmE.

In conclusion, the solution of the structure of YlmE from S. coelicolor A3(2) and other COG0325 members revealed a family of conserved, PLP-binding proteins, but their biochemical function still remains to be elucidated.