Structural Basis for the Asymmetry of a 4-Oxalocrotonate Tautomerase Trimer

Abstract

Tautomerase superfamily (TSF) members are constructed from a single β−α−β unit or two consecutively joined β−α−β units. This pattern prevails throughout the superfamily consisting of more than 11,000 members where homo- or heterohexamers are localized in the 4-oxalocrotonate tautomerase (4-OT) subgroup and trimers are found in the other four subgroups. One exception is a subset of sequences that are double the length of the short 4-OTs in the 4-OT subgroup, where the coded proteins form trimers. Characterization of two members revealed an interesting dichotomy. One is a symmetric trimer, whereas the other one is an asymmetric trimer. One monomer is flipped 180° relative to the other two monomers so that three unique protein-protein interfaces are created that are composed of different residues. A bioinformatics analysis of the fused 4-OT subset shows a further division into two clusters with a total of 133 sequences. The analysis showed that members of one cluster (86 sequences) have more salt bridges if the asymmetric trimer forms, whereas the members of the other cluster (47 sequences) have more salt bridges if the symmetric trimer forms. This hypothesis was examined by the kinetic and structural characterization of two proteins within each cluster. As predicted, all four proteins function as 4-OTs, where two assemble into asymmetric trimers (designated R7 and F6) and two form symmetric trimers (designated W0 and Q0). These findings can be extended to the other sequences in the two clusters in the fused 4-OT subset, thereby annotating their oligomer properties and activities.

Graphical Abstract

Introduction

The tautomerase superfamily (TSF) consists of more than 11,000 members that sort into five major subgroups.¹ The members share a β−α−β building block and frequently have a catalytic amino-terminal proline.^1-5 The 4-oxalocrotonate tautomerase (4-OT) subgroup is the largest of the five subgroups and the individual members are constructed from a single β−α−β unit (58-84 amino acids) to form homo- or heterohexamers,^6-9 whereas those of the other four subgroups are composed of two consecutively joined β−α−β units (110-150 amino acids) to form trimers.^10-12 The predominance of this structural theme suggests that a gene fusion event took place early in the evolutionary history of the TSF leading to the diversification of function that is seen today.

Within the 4-OT subgroup, there is a subset of sequences double the length of that of the short 4-OTs.^1,13 These 4-OTs form a separate subgroup that still connects to the short 4-OTs in a sequence similarity network (SSN).¹³ More importantly, they might share features with possible progenitors for the other four subgroups in the TSF, which are of the double length represented by the fused 4-OTs. In view of the potential significance of the members in this subset, they have been subjected to kinetic, mechanistic, and structural analysis.¹³

A majority of the sequences in this subset (133 sequences) falls into 2 clusters identified as the “Linker 2” and “Fused 4-OT” clusters (47 and 86 sequences, respectively) in the SSN.^1,13 Structural analysis of one protein (designated “linker 2”) in the Linker 2 cluster showed that it is a symmetrical trimer (Figure 1A) that consists of three αβ interfaces (where α and β refer to the N- and C-terminal region of the adjoining β−α−β units, respectively). Each interface has a single catalytic Pro-1. The enzyme carries out the canonical 4-OT tautomerization reaction where Pro-1 (functioning as a general base) transfers a proton from the 2-hydroxy group of 2-hydroxymuconate (2-HM, 1, Scheme 1) to C-5 of the product (2).

Figure 1.

Symmetric and asymmetric trimers characterized in the fused 4-OT subset of sequences in the 4-OT subgroup.¹³ A. Linker 2 showing the typical TSF trimer consisting of three αβ interfaces, each containing a catalytic Pro-1. B. Fused 4-OT showing the αα, ββ, and αβ interfaces with two Pro-1, no Pro-1, and one Pro-1 (respectively and clockwise).

Scheme 1.

TSF-catalyzed Reactions

One striking feature about the first protein characterized in the Fused 4-OT cluster was the absence of P3 rotational symmetry. The protein, designated “fused 4-OT”, functions as an enzyme and forms an asymmetric trimer where one of the monomers is flipped 180° relative to the other two monomers in the trimer (Figure 1B).¹³ As a result, there are three unique interfaces (designated αα, ββ, and αβ) composed of different residues.¹³ The αα interface has two N-terminal prolines, the ββ interface does not have an N-terminal proline, and the αβ interface has a single Pro-1. It is not known how the different active site configurations contribute to the overall activity, but the enzyme catalyzes the same reaction as linker 2 (1 to 2) with nearly comparable catalytic efficiency (k_cat/K_m ~ 10⁶-10⁷ M⁻¹s⁻¹, Scheme 1).^1,13

In view of the unusual nature of the asymmetric oligomerization mode, a bioinformatics analysis was carried out to identify possible explanations. One difference between the sequences in the Linker 2 and Fused 4-OT clusters is the length of the linker loop between the two consecutively fused β−α−β units and the presence of a cis-prolyl peptide.¹³ In the Linker 2 cluster, the linker loop is much shorter with only 4 amino acids, whereas in the Fused 4-OT cluster the linker loop is about 10 amino acids long.¹³ These differences were ruled out as a possible explanation because an engineered variant that uncoupled the consecutively fused β−α−β units by removing the covalent bond (to make a heterohexamer) still showed the asymmetric arrangement.^1,13

A second possibility focused on the collection of hydrophilic and hydrophobic interactions at these interfaces. An analysis of the interactions showed that the fused 4-OT had a total of ten salt bridges, predominantly at the αα interface where there are six salt bridges. An artificially constructed symmetric model only had six salt bridges. The same observation holds for linker 2, but only when the symmetric trimer is formed. In this case, there are nine salt bridges, but the artificially constructed asymmetric model generated five salt bridges. Hence, it appears that a higher number of salt bridges is responsible for the more stable trimer, whether it be asymmetric or symmetric.

This hypothesis was examined by a bioinformatics analysis of the subset of sequences, followed by a kinetic and structural analysis of four targeted proteins – two from the Fused 4-OT cluster (designated R7 and F6) and two from the Linker 2 cluster (designated W0 and Q0). The former two are predicted to form asymmetric trimers, whereas the latter two are predicted to form symmetric trimers. These predictions were validated by structural analysis and reinforced by a detailed examination of the interface interactions. Not surprisingly, all four proteins display 4-OT activity. The combined observations suggest that the remaining sequences in the Fused 4-OT cluster will produce mostly asymmetric trimers, and those in the Linker 2 cluster will generate mostly symmetric trimers.

Experimental procedures

Materials.

All chemicals, biochemical, and enzyme substrates were obtained from sources described elsewhere.^7,8,1,13,14 Chromatographic resins (DEAE Sepharose fast flow and Phenyl-Sepharose 6 fast flow) and columns (PD-10 Sephadex G-25M, Econo-Column, and Spectra/Chrom LC) were purchased from commercial vendors, as described.⁷ The Amicon stirred cell concentrators and the ultrafiltration membranes (10,000 Da, MW cutoff) were purchased from EMD Millipore Inc. (Billerica, MA). Coomassie Blue R-250 dye for staining sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gels was adapted from a literature source.¹⁵

Bacterial Strains, Plasmids, Growth, and Expression Conditions.

Escherichia coli strain BL21-Gold(DE3) was obtained from Agilent Technologies (Santa Clara, CA). The genes encoding the targeted fused 4-OTs were codon-optimized for expression in E. coli, synthesized, and cloned into the expression vectors pJ411 (pJ411-R7, pJ411-F6, pJ411-W0, pJ411-Q0) by DNA2.0, Inc., now ATUM (Newark, CA).^13,14 The gene encoding R7 was from Burkholderia sp. RPE67 (UniProt accession: A0A060NYR7) and the gene encoding F6 was from Caballeronia arationis (UniProt accession: A0A157ZJF6). The gene encoding W0 was from Advenella mimigardefordensis DPN7 (UniProt accession: W0PD56) and the gene encoding Q0 was from Bordetella trematum (UniProt accession: A0A157L8Q0). Cells were grown overnight (~20 h) at 37 °C (220 rpm) in Luria–Bertani (LB) media (1 L in a 2 L flask), supplemented with ZYM-5052 auto induction media¹⁶ and kanamycin (50 μg/mL). The cultures were inoculated with 2 mL of pre-culture (grown overnight). Typically, the yield is 5 g of cells/L. Cells were re-suspended in 20 mM Na₂HPO₄ buffer, pH 7.2 (10 mL, buffer A) made 1 mM in phenylmethylsulfonyl fluoride.

General Methods.

Steady-state kinetic assays were performed on an Agilent 8453 diode-array spectrophotometer at 22 °C. Nonlinear regression data analysis was performed using the program Grafit (Erithacus Software Ltd., Staines, U.K.). Protein concentrations were determined by the Waddell method.¹⁷ SDS-PAGE was carried out on denaturing gels containing 12% polyacrylamide.¹⁸ Electrospray ionization mass spectrometry was performed on an Orbitrap Fusion mass spectrometer (Thermo, San Jose, CA), housed in the Proteomics Facility in the Center for Biomedical Support at the University of Texas, Austin. Samples were prepared according to previously published procedures.¹⁹

Construction of the SSN for the Level 2 4-OT Subgroup Containing the Canonical 4-OT.

The network was generated using an in-house version of the Pythoscape software,²⁰ tailored for use with available hardware. All-by-all pairwise comparisons of 133 Level 2 4-OT subgroup sequences, as curated in the Structure-Function Linkage Database (SFLD), were obtained using the BLAST algorithm. The nodes (each representing a sequence) are arranged using Prefuse force directed OpenCL layout provided by the Cytoscape software,²¹ weighted by BLAST bit score of the associated edges. Edges are drawn between two nodes only if the mean similarity between the two sequences is at least as significant as the bit score of 96 chosen to illustrate similarity relationships for this network.

Purification of R7, F6, W0, and Q0.

The proteins were purified based on the protocols described elsewhere for the cis-CaaD homologue encoded by PputUW4_01749 in Pseudomonas sp. UW4 (UniProt K9NIA5, designated Ps01740)¹⁴ and the fused 4-OT from Burkholderia lata (ATCC 17760, UniProt accession: Q392K7).¹³ Cells were disrupted by sonication (2 cycles of 3 min at 1 s pulses, 40% duty cycle/50% output) using an Ultrasonic Processor Sonicator model W-385 (Heat Systems-Ultrasonics, Inc, Newtown, CT). After centrifugation (25,400g for 45 min), the clear supernatant was treated as described below. Protein samples were concentrated using a 50-mL Amicon Stirred Cell, and the target proteins were typically identified in fractions by activity using 2-hydroxymuconate (Scheme 1) and visualization by SDS-PAGE analysis.

For R7, the clear supernatant (~10 mL) was applied to a hand-packed DEAE-Sepharose column (~20 mL bed volume), which had previously been equilibrated in buffer A. The column was washed with buffer A (3 × 20 mL), and the flow-through (~60 mL) was collected and concentrated to ~10 mL. The final protein concentration was ~9 mg/mL. SDS-PAGE analysis showed near homogenous protein. Aliquots were flash-frozen in liquid nitrogen and stored at −80 °C until further use.

For W0, the clear supernatant (~10 mL) was made up to 3M NaCl, where the ground NaCl was added in 3 aliquots to the chilled, stirring solution over a 2 h period. The resulting solution was centrifuged at 18,000 rpm (25,400g for 30 min). The supernatant was applied to a hand-packed Phenyl-Sepharose column (~20 mL bed volume), which had previously been equilibrated in buffer B (20 mM Na₂HPO₄ buffer made 3M in NaCl, pH 7.3). The column was washed with buffer B (3 × 20 mL), and the retained proteins were eluted with a linear gradient (buffer B to buffer A, total volume 100 mL). Fractions were collected at 1 min intervals (~2.5 mL). Typically, W0 eluted around 2.5 M NaCl. The appropriate fractions were pooled, concentrated to ~ 1 mL, and desalted using a PD-10 Sephadex column, which had previously been equilibrated in buffer A. Fractions were collected as described above and those containing W0 were pooled and concentrated to ~3 mL. The sample was applied to a hand-packed DEAE-Sepharose column (~20 mL bed volume), which had previously been equilibrated in buffer A. After washing the column with buffer A (2 × 20 mL), the retained proteins were eluted with a linear gradient (buffer A to buffer C, total volume 100 mL). Buffer C is 20 mM Na₂HPO₄ buffer made 500 mM in NaCl, pH 7.3 Typically, W0 eluted around 100 mM NaCl. The appropriate fractions were pooled and concentrated to ~1 mL. The resulting concentrate was desalted using a PD-10 column equilibrated with buffer A, and concentrated to ~10 mL. The final protein concentration was 3.5 mg/mL. Aliquots were flash-frozen in liquid nitrogen and stored at −80 °C until further use.

For F6 and Q0, the clear supernatant (~10 mL) was applied to a hand-packed DEAE-Sepharose column (~20 mL bed volume), which had previously been equilibrated in buffer A (20 mM Na₂HPO₄ buffer, pH 7.3). The column was washed with buffer A (3 × 20 mL), and the flow-through (~60 mL) was collected and concentrated to 30 mL. The solution was chilled on ice, and ground (NH₄)₂SO₄ was added in aliquots over a 2 h period to the stirring solution to make it 0.8 M in (NH₄)₂SO₄. There was no obvious precipitation. A portion of the solution (~15 mL) was applied to a hand-packed Phenyl-Sepharose column (~20 mL bed volume), which had previously been equilibrated in buffer D [20 mM Na₂HPO₄ buffer made 0.8 M in (NH₄)₂SO₄, pH 7.3]. The column was washed with buffer D (2 × 20 mL) and the retained proteins were eluted with a linear gradient (buffer D to buffer A, total volume 100 mL), followed by 2 washes of deionized H₂O (2 × 20 mL). Fractions were collected dropwise (~1 mL portions). Both proteins eluted at ~ 0.1 M (NH₄)₂SO₄. Fractions showing near homogenous protein were pooled and concentrated to 15 mL and 12 mL for F6 and Q0, respectively. The final protein concentration was 2.5 mg/mL for both F6 and Q0. Aliquots were flash-frozen in liquid nitrogen and stored at −80 °C until further use.

Steady State Kinetics.

The assays were carried out at 22 °C in 10 mM Na₂HPO₄ buffer (pH 7.35) using 2-hydroxymuconate (2-HM) and phenylenolpyruvate (PP) as substrates (Scheme 1).^7,13,14 Various stock solutions of 2-HM (28-44 mM) and PP (68-93 mM) were prepared by dissolving the appropriate amount of the crystalline free acids in absolute ethanol. Substrate concentrations ranged from 15-600 μM for 2-HM and 100-800 μM for PP. The initial enzyme tested (R7) was found to be unstable at dilute concentrations. Accordingly, each enzyme was diluted 10-fold into 100 mM Na₂HPO₄ buffer made 300 mM in NaCl (final pH 7.2). Subsequent dilutions were made up in 100 mM Na₂HPO₄ buffer (pH 7.2) and used within 15 min. For kinetics using 2-HM, stock solutions of 0.0280 mg/mL, 0.190 mg/mL, 0.0103 mg/mL, and 0.0438 mg/mL were made up for R7, W0, Q0, and F6, respectively. These stock solutions correspond to final enzyme concentrations of 0.00843 μM, 0.0359 μM, 0.00213 μM, and 0.0177 μM (respectively for R7, W0, Q0, and F6). For kinetics using PP, stock solutions of 0.866 mg/mL, 0.358 mg/mL, 0.263 mg/mL, and 0.263 mg/mL were made up for R7, W0, Q0, and F6, respectively. These stock solutions correspond to final enzyme concentrations of 0.173 μM, 0.0677 μM, 0.0555 μM, and 0.213 μM (respectively for R7, W0, Q0, and F6).

The enol-keto tautomerization of 2-HM was monitored by following the decay of the enol form at 315 nm (ε = 10000 M⁻¹ cm⁻¹). The enol-keto tautomerization of PP was monitored by following the decay of the enol-form at 305 nm (ε = 6200 M⁻¹ cm⁻¹).^13,14 Initial rates were determined, plotted against the substrate concentration, and fitted to the Michaelis-Menten equation using Grafit. For all proteins, the kinetic parameters are based on the monomer molecular mass.

Crystallization.

Initial crystallization conditions for R7, F6, W0, and Q0 were identified using sparse-matrix screening with a Phoenix crystallization robotic system (Art Robbins Instruments). Conditions for each protein were optimized using the sitting drop vapor diffusion technique at room temperature. R7 crystallized in 0.2 M (NH₄)₂SO₄, 15-30% PEG 5000 MME, pH 6-7.5. F6 crystallized in 8% ethylene glycol, 20-23% PEG 10000, pH 7.5. W0 crystallized in 0.2 M (NH₄)₂SO₄, 27-37% PEG 5000 MME, pH 6-6.5. Q0 crystallized in 0.2 M (NH₄)₂SO₄, 27-33% PEG 4000, pH 7.5. All crystals were cryoprotected with mother liquor supplemented with 30% glycerol and flash-frozen in liquid nitrogen to be shipped to a national synchrotron source.

Data collection, Processing, Structure Determination and Refinement.

Diffraction data were collected from both the Advanced Light Source beamline 5.0.3 (ALS, Berkeley, CA) and the Advanced Photon Source (APS) beamline 23-ID-D (Argonne National Laboratories). The data sets were indexed, integrated, and scaled using HKL-2000.²² The structures were determined by molecular replacement (MR) using Phaser-MR²³ and Autobuild²⁴ from the PHENIX suite of programs. The asymmetric fused 4-OT crystal structure with all side chains removed (74% sequence identity, PDB entry 6BLM) was used as a search model for the initial phases for the new structures (i.e., R7, F6, W0, and Q0). The monomeric alanine truncated version of fused 4-OT was used as a search model for the initial estimate of the R7, F6, W0, and Q0 structure factors. Subsequently, using high-resolution electron density maps, chain identity was assigned. Structure refinement was performed using Phenix Refine.²⁵ TLS parameters were included in the refinement of all structures.²⁶ W0 has considerable twinning as detected by the PHENIX software, Xtriage.²⁷ Hence, W0 was refined using the twin law -h, -k, l. The final structures were evaluated during and after refinement using Molprobity.²⁸ The refinement statistics for the structures are summarized in Table 1. The structures show good statistics in data collection and geometry in refinement with the exception of W0, in which the twinned crystals resulted in a high R_work/R_free in model refinement. All figures were prepared with PyMol (The PyMOL Molecular Graphics System, Version 1.8 Schrödinger, LLC.).²⁹

Table 1.

Crystallographic Data Collection and Refinement Statistics

Data Collection	R7	F6	W0	Q0
Space group	C21	P212121	P3	P43
Cell dimensions
a, b, c (Å)	149.7, 49.9, 48.0	48.1, 69.4, 93.7	83.8, 83.8, 149.6	72.9, 72.9, 82.2
α, β, γ (deg)	90.0, 100.5, 90.0	90.0, 90.0, 90.0	90.0, 90.0, 120.0	90.0, 90.0, 90.0
Resolution (Å)	47.22-1.83(1.90-1.83)*	42.87-1.39(1.44-1.39)*	41.93-2.10(2.176-2.101)*	35.84-1.80(1.86-1.80)*
Wavelength (Å)	1.0332	1.0332	1.0332	1.0332
Rsym	0.1314(0.4454)*	0.0838(0.4634)*	0.1433(0.5292)*	0.1112(0.6750)*
Rpim	0.0783(0.2877)*	0.0335(0.1901)*	0.0673(0.3031)*	0.0507(0.3148)*
CC1/2ϒ	0.988(0.790)*	0.998(0.86)*	0.992(0.75)*	0.996(0.738)*
I/σ	8.2(1.7)*	18.1(3.0)*	10.9(1.8)*	11.5(1.9)*
Completeness (%)	97.1(76.0)*	99.9(99.4)*	99.5(97.1)*	98.8(96.6)*
Redundancy	3.5(2.9)*	7.0(6.7)*	5.4(3.8)*	5.6(5.3)*
Refinement
Resolution (Å)	47.22-1.83(1.90-1.83)*	42.87-1.39(1.44-1.39)*	41.93-2.10(2.17-2.10)*	35.84-1.80(1.86-1.80)*
Unique reflections	29709(2277)*	62808(6174)*	68347 (6690)*	39332(3844)*
Rwork	0.1869(0.2557)*	0.1680(0.1992)*	0.2750 (0.3102)*	0.1719(0.2127)*
Rfree±	0.2285(0.2918)*	0.1865(0.2105)*	0.3335 (0.3678)*	0.2054(0.2136)*
No. of atoms
Protein	2697	2673	6978	2607
Ligand/ion	24	12	N/A	N/A
Water	384	462	125	400
B-factor (Å2)
Protein	28.6	15.8	28.9	26.3
Ligand/ion	33.6	20.9	N/A	N/A
Water	36.7	27.0	23.4	35.0
R.M.S Deviations
bond lengths (Å)	0.007	0.011	0.010	0.005
bond angles (deg)	0.87	1.27	1.19	0.82
Ramachandran plot (%)
Favored	99.72	99.45	93.85	98.90
Allowed	0.28	0.55	5.46	1.10
Outliers£	0.00	0.00	0.68	0.00
Molprobity score^	1.35^/98th percentile¥	1.12^/98th percentile¥	2.61^/35th percentile¥	0.97^/100th percentile¥

^*Values for the corresponding parameters in the outermost shell in parenthesis.

^ϒCC_1/2 is the Pearson correlation coefficient for a random half of the data, the two numbers represent the lowest and highest resolution shell respectively.

^±R_free is the R_model calculated for 5% of the reflections randomly selected and omitted from refinement.

^£There are two Ramachandran outliers in W0, which correspond to alanine residues (Ala 47 chain A, Ala 47 chain D) both with strong electron density.

^{^}MolProbity score is calculated by combining clashscore with rotamer and Ramachandran percentage and scaled based on X-ray resolution.

^¥The percentage is calculated with 100th percentile as the best and 0th percentile as the worst among structures of comparable resolution.

Results and Discussion

Interface Analysis of Linker 2 and Fused 4-OT.

Structural analysis of a representative in each of the two clusters revealed two different oligomerization arrangements: the enzyme designated linker 2 from Pusillimonas strain T7-7 formed a symmetric trimer with three identical αβ interfaces (Figure 2A), whereas the enzyme designated fused 4-OT from Burkholderia lata formed an asymmetric trimer with three different interfaces, αα, ββ, and αβ (Figure 2B).¹³ The initial hypothesis that differences in the linker region between the two β−α−β units (length and configuration) might be responsible for the different states was ruled out because severing the linker still produced an asymmetric arrangement (albeit a heterohexamer).¹³ Hence, the three different interfaces in the asymmetric trimer (αα, ββ, and αβ) were analyzed for interactions that might govern asymmetric trimer formation (Figure 2C-F).

Figure 2.

A) Schematic representation viewed from above of the symmetric linker 2. Each interface is labeled as αβ, Pro-1 is shown as a sphere at each interface, and the interface is depicted with a dotted gray line. B) Schematic representation viewed from above of asymmetric fused 4-OT. Each interface is labeled as αα, αβ, and ββ. Pro-1 is shown as a sphere and the interface is depicted with a dotted gray line. C) Side view of asymmetric fused 4-OT αα interface. The dotted line shows the explored interface area where Pro-1 is shown as a sphere. Arrows indicate turn of each monomer for panels D, E, and F. D) Surface analysis of αβ interface (Chains C and B) of asymmetric fused 4-OT. E) Surface analysis of αα interface (Chains A and B) of asymmetric fused 4-OT. F) Surface analysis of ββ interface (Chains C and A) of asymmetric fused 4-OT. The main interacting surface is shown within dotted black line.

Surface analysis of the three interfaces of fused 4-OT showed a combination of potential hydrophobic and hydrophilic (i.e., hydrogen bonds and salt bridges) interactions (Figure 2D-F). The hydrophobic interactions were quantified by PISA analysis, which calculates solvent-free energy gain (ΔG) as the difference in total solvation energies of isolated and interfacing structures, where a negative ΔG corresponds to hydrophobic interactions.³⁰ The values for the αα and ββ interfaces were compared with that of the αβ interface. In fused 4-OT, the energy is quite different for the three interfaces: for the αα interface, the ΔG is −5 kcal/mol; for the ββ interface, the ΔG is −12.9 kcal/mol; and for the αβ interface, the ΔG is −11 kcal/mol. Formation of an asymmetric trimer results in increased hydrophobic interactions in the ββ interface as compared with the αβ interface, but decreased hydrophobic interactions in the αα interface.

We next examined the polar interactions at each interface. Unlike the symmetric linker 2 trimer with three identical αβ interfaces, the three interfaces in the asymmetric fused 4-OT trimer have substantially different hydrophilic interactions with a striking predominance of salt bridges. Specifically, the αα interface is composed of two adjacent monomers where each monomer begins with the first β−α−β unit, resulting in the reciprocal formation of three pairs of salt bridges for a total of six salt bridges (Figure 3A, Table 2). Both the αβ and ββ interfaces contribute an additional two salt bridges each (Figure 3B,C, Table 2). Thus, there is a significant gain in hydrophilic interactions at the αα interface. Had fused 4-OT formed a symmetric trimer, only six salt bridges (two for each αβ interface) would form (Table 2).

Figure 3.

Ribbon representation showing the salt bridges of the three interfaces in the asymmetric fused 4-OT. Monomer A is shown in green; monomer B is shown in teal; monomer C is shown in taupe. Pro-1 is shown as a sphere. Salt bridge interactions are highlighted in stick form and dotted magenta line. A) αα interface showing three reciprocal salt bridge pairs for a total of six salt bridge interactions. B) αβ interface showing two unique salt bridge interactions (R123-D78 and D114-K81). C) ββ interface showing a conserved salt bridge (K108-E04) and a unique salt bridge interaction (R82-D59).

Table 2.

List of Salt Bridges in Fused 4-OT and Linker 2^a

fused 4-OT PDB: 6BLM
Asymmetric			No. of SB
Interface	Residue	Residue
αα x2	E04A,B	K108B,A	10
αα x2	D13A,B	R123B,A
αα x2	K16A,B	D114B,A
αβ	D78B	R123C
αβ	K81B	D114C
ββ	R82A	D59C
ββ	K108A	E04C
Symmetric model			No. of SB
Interface	Residue	Residue
αβ x3	D78	R123	6
αβ x3	K81	D114
6
linker 2 PDB: 5UNQ
Asymmetric model			No. of SB
Interface	Residue	Residue
αβ	R39	E97	5
αβ	D66	R101
αβ	E73	K119
ββ x2	E97	R99
Symmetric			No. of SB
Interface	Residue	Residue
αβ x3	R39A,B,C	E97B,C,A	9
αβ x3	D66A,B,C	R101B,C,A
αβ x3	E73A,B,C	K119B,C,A

^aResidues involved in salt bridge interactions in the naturally occurring oligomerization states (shaded background in bold) and corresponding artificially generated molecular models (white background). The rightmost column shows the total number of salt bridges in the naturally occurring states and in the artificially generated models.

These observations raise the question of whether a greater number of salt bridges can be formed for all 4-OTs in the subset of 133 sequences if they adopt the asymmetric trimer. This is particularly interesting for linker 2, which exists as a symmetric trimer. To address this question, an artificial, asymmetric model of linker 2 was generated by flipping one monomer and inspecting the three different interfaces in the resulting construct, and comparing them to those in the symmetric structure. PISA analysis for the linker 2 symmetric structure shows a ΔG between −9.9 to −11.2 kcal/mol for the three αβ interfaces, whereas the asymmetric model shows less favorable hydrophobic interactions in both αα and ββ interfaces (ΔG = −8.8 and −9.7 kcal/mol, respectively). Furthermore, the asymmetric model has substantially less hydrophilic interactions. Instead of the three salt bridges found in each αβ interface (for a total of nine salt bridges) of the symmetric linker 2 trimer, there are only five salt bridges (three in the αβ, two in the ββ interface, none in the αα interface) in the asymmetric model (listed in Table 2). Thus, unlike fused 4-OT, there would be a substantial energy cost for linker 2 if it had assumed an asymmetric state.

Bioinformatics Analysis.

With these observations in hand (Figure 3, Table 2), a bioinformatics analysis along with manual curation was carried out on the 2 clusters identified as the “Linker 2” and “Fused 4-OT” clusters (47 and 86 sequences, respectively) (Figure 4, top panel).¹³ A preliminary analysis showed that four salt bridges are conserved in the Fused 4-OT cluster (Glu-4/Lys-108, Asp-13/Arg-123, Lys-16/Asp-114, Lys-81/Asp-114). The first three are found at the αα interface and the last one is found at the αβ interface. The conservation of these three salt bridges at the αα interface suggested that they might stabilize an asymmetric trimer. Hence, R7 (with all four salt bridges) and F6 (missing one of the salt bridges) were chosen for in-depth kinetic and structural analysis. In the Linker 2 cluster, the Lys-81/Asp-114 salt bridge was conserved in several members, but not in others. Hence, Q0 (with the Lys-81/Asp-114 salt bridge) and W0 (without the Lys-81/Asp-114 salt bridge) were chosen for in-depth kinetic and structural analysis. A subsequent sequence alignment of fused 4-OT, linker 2, and the four candidates for study show additional salt bridges (Figure 4, bottom panel).

Figure 4.

In the top panel, the two different clusters are shown in a sequence similarity network.¹³ In a red gradient, we observe conservation of salt bridges in fused 4-OT, where white reflects the absence of salt bridges and darkest red reflects four salt bridges, as indicated by the legend. The salt bridges are listed in the text and the candidate proteins are indicated. Fused 4-OT and linker 2 are shown as squared shape nodes. The bottom panel shows a sequence alignment of asymmetric fused 4-OT, symmetric linker 2, and four sequences in the two clusters identified as the Fused 4-OT and Linker 2 clusters, respectively. (For clarity, only the sections of the sequences that interact are shown.) The sequences of R7 and F6 (in the Fused 4-OT cluster) align best with the sequence of the asymmetric fused 4-OT (top three sequences above the gray spacer), while the sequences of W0 and Q0 (in the Linker 2 cluster) align best with the sequence of linker 2 (bottom three sequences below the gray spacer). The top numbering corresponds to the amino acid sequence of R7, and the bottom numbering corresponds to the amino acid sequence of Q0. Salt bridge pairs from the asymmetric αα interface are color-coded to their respective partners based on the fused 4-OT mapped interactions.

Purification and Characterization of the R7, F6, W0, and Q0.

The four proteins were overproduced in E. coli BL21(DE3) and purified to near homogeneity. The yields are ~30 mg/L of culture for F6 and Q0, ~42 mg/L of culture for W0, and ~ 90 mg/L of culture for R7. ESI-MS analysis of the purified proteins showed a single major signal corresponding to the monoisotopic molecular mass of 12,781.00 Da (calc. 12,912.04 Da) for R7 (123 amino acids), 13,535.13 Da (calc. 13,666.18 Da) for W0 (127 amino acids), 12,439.79 Da (calc. 12,570.83 Da) for Q0 (123 amino acids), and 12,679.03 Da (calc. 12,810.07 Da) for F6 (123 amino acids). The mass differences of 131 Da between the theoretical and the observed masses indicate that the translation-initiating methionine was post-translationally removed, resulting in the mature, active proteins with an N-terminal proline.³¹ Smaller signals were observed for Q0 at 12,455 Da and 12,471 Da. The mass difference corresponds to 16 Da, which could correspond to the gain of oxygen. Fused 4-OT and linker 2 have 127 and 123 amino acids, respectively.

Kinetic Properties of R7, F6, W0, and Q0.

The proteins were screened for activity using 2-hydroxymuconate (2-HM in Scheme 1) and phenylenolpyruvate (PP in Scheme 1), a di- and monocarboxylate substrate, respectively. The kinetic parameters are collected in Tables 3 and 4. Using 2-HM, Q0 is the most efficient enzyme of the four, with the highest k_cat/K_m value (followed closely R7 and F6) (Tables 3). W0 is the least efficient with a ~41-fold lower k_cat/K_m value than that of Q0 (Tables 3). Using PP, Q0 and W0 have comparable k_cat/K_m values that are 3- to 7-fold higher than those of R7 and F6 (Table 4). The k_cat/K_m values measured for fused 4-OT using 2-HM and PP are comparable to these values except for W0 (using 2-HM) and R7 (using PP). The k_cat/K_m value for fused 4-OT using 2-HM is 14.7 times higher than that of W0 and the k_cat/K_m value for fused 4-OT using PP is 6.3 times higher than that of R7. The k_cat/K_m values measured for linker 2 using 2-HM and PP are higher than these values especially W0 (using 2-HM) and R7 (using PP). In these cases, the k_cat/K_m value for linker 2 using 2-HM is 184 times higher than that of W0 and the k_cat/K_m value for linker 2 using PP is 43.3 times higher than that of R7.

Table 3.

Kinetic parameters for R7, F6, Q0, and W0 using 2-HM^a

Enzyme	Km (μM)	kcat (s−1)	kcat/Km (M−1s−1)
R7	206 ± 33	257 ± 16	(1.3 ± 0.2) × 106
F6	86 ± 7	89 ± 2	(1.04 ± 0.9) × 106
Q0	96 ± 16	760 ± 39	(7.9 ± 1.4) × 106
W0	n.d.b	n.d.b	(1.9 ± 0.4) × 105

^aThe kinetic parameters were measured by the assay described in the text.

^bnot determined

Table 4.

Kinetic parameters for R7, F6, Q0, and W0 using PP^a

Enzyme	Km (μM)	kcat (s−1)	kcat/Km (M−1s−1)
R7	n.d.b	n.d.b	(3.0 ± 0.8) × 104
F6	722 ± 53	51 ± 2	(7.0 ± 0.6) × 104
Q0	n.d.b	n.d.b	(1.5 ± 0.1) × 105
W0	n.d.b	n.d.b	(2.1 ± 0.4) × 105

^aThe kinetic parameters were measured by the assay described in the text.

^bnot determined

Based on the k_cat/K_m values, the enzymes prefer the dicarboxylate substrate, which is consistent with the active site configuration (i.e., two arginine residues) and the trend reported elsewhere.¹ As will be discussed below, we determined the structures of four proteins (R7, F6, W0, Q0), in addition to fused 4-OT and linker 2. A structural comparison of the α,β interfaces in all six proteins shows that Pro-1 is positioned between two arginines (as it is for canonical 4-OT)¹, and that the active sites are nearly superimposable (Figure S1). As noted elsewhere, this configuration favors the dicarboxylate substrate (2-HM), but will process a monocarboxylate substrate (PP).¹ The decreased activities observed for some of the enzymes might be due to subtle changes in the active sites. Structural analysis of the αα and ββ interfaces in the three asymmetric proteins again shows that the active sites are again nearly superimposable (Figure S2). In the αα interfaces, a catalytic Pro-1 is positioned between two arginines and it appears that there could be two active sites within each αα interface. (For this reason, there could be three active sites per trimer and so the kinetic parameters are based on the monomer mass). In the ββ interfaces, proline is present, but not catalytic (the prolyl nitrogen is in the peptide bond), and positioned between two arginines.

W0 could not be saturated with 2-HM. Hence, the reported k_cat/K_m was determined from the linear region of the Michaelis-Menten fit. Likewise, R7, Q0, and W0 could not be saturated with PP so that the reported values of k_cat/K_m were calculated similarly.

Structure Determination of F6, R7, Q0, W0.

Crystallographic analysis was used to determine the trimeric arrangement for two members in the Fused 4-OT cluster (R7 and F6) and two members in the Linker 2 cluster (W0 and Q0). The four proteins were crystallized, and diffraction data were collected. One monomer from the fused 4-OT structure (PDB code 6BLM), rather than the trimer of fused 4-OT, was used as a search model to identify the solution for molecular replacement so that the oligomerization is independently assigned without bias from the search model.¹³ All four proteins show the typical tautomerase superfamily (TSF) fold where each monomer of the trimer consists of two consecutively fused β−α−β subunits (Figure 5, leftmost column).³

Figure 5.

Each panel shows a 4-OT trimer. To the far left, the overall ribbon representation of each trimer with Pro-1 highlighted as spheres. To the right, each interface is shown labeled and with transparent cartoon. Salt bridge formations are highlighted in stick mode. A) R7 asymmetric fused 4-OT where the flipped monomer A is shown in the darkest green (bottom monomer), and monomers B and C are shown in lighter shades of green. B) F6 asymmetric fused 4-OT where the flipped monomer A is shown in dark orange (bottom monomer), and monomers B and C are shown in lighter shades of yellow. C) W0 symmetric 4-OT trimer where the monomers are positioned equally, and each monomer is shown in a different shade of purple. D) Q0 symmetric 4-OT trimer where the monomers are positioned equally, and each monomer is shown in a different shade of blue.

The structure for R7 (1.8 Å resolution) contains continuous density from Pro-1 to Arg-123 (total 123 amino acids) for each monomer in the trimer. The trimer adopts an asymmetric arrangement like that of fused 4-OT (Figure 5A). This is not surprising in view of the high sequence identity (74%) with fused 4-OT (Figure 4, bottom panel). In particular, the residues forming the salt bridges in the αα interface are all conserved, resulting in a structure that looks almost identical to that of fused 4-OT (Figure 4, bottom panel). For the αα interface, these pairs involve Glu-4/Lys-104 (corresponding to Glu-4 and Lys-108 in fused 4-OT), Asp-13/Arg-119 (corresponding to Asp-13 and Arg-123 in fused 4-OT), and Lys-16 participates in a triad with Asp-44 and Asp-110 (from an adjacent monomer) (corresponding to Lys-16 and Asp-114 in fused 4-OT) (Figures 3A, 5A, and 6). There is one less salt bridge in the ββ interface because a shortening of the loop containing Asp-59 precludes the formation of the Asp-59/Arg-82 salt bridge that is observed in fused 4-OT (Figure 6). However, the Glu-4/Lys-104 salt bridge is present. In the αβ interface, there are two salt bridges (Glu-74/Arg-119 and Lys-77/Asp-110), corresponding to those in fused 4-OT (Asp-78/Arg-123 and Lys-81/Asp-114). Had R7 adopted a symmetric conformation, these two αβ interface salt bridges would be the only ones conserved in the structure resulting in a total of six salt bridges (Figure 6). Because of the asymmetric arrangement, there are a total of nine salt bridges with the αα interface contributing the most (six salt bridges) (Figure 6).

Figure 6.

The top panel shows an alignment of six proteins (three from the Fused 4-OT cluster and three from the Linker 2 cluster), highlighting the salt bridge pairs found in each interface of the asymmetric trimer. In the αα interface, each interaction is reciprocal, so that each salt bridge is found twice in the same interface. In the αβ and ββ interfaces, each salt bridge is found only once. At the rightmost panel, the total number of salt bridges for the asymmetric arrangements is calculated. The bottom panel shows an alignment of the same six proteins highlighting the salt bridge pairs found in the αβ interface of the symmetric trimer. All salt bridges found in αβ interfaces in proteins were considered. At the rightmost panel, the total number of salt bridges for the symmetric arrangements is calculated. Salt bridge pairs are highlighted in bold.

The structure for F6 (1.4 Å resolution) contains continuous density from Pro-1 to Arg-123 (total 123 amino acids) for each monomer in the trimer, where the high resolution confirms the asymmetric trimer arrangement (Figure 5B). The salt bridges are mostly conserved between fused 4-OT and F6 with two exceptions (one in the αα interface and one in the ββ interface). In the αα interface, Lys-16 (in both fused 4-OT and F6) is still sandwiched between the Asp-44 in the same subunit and Asp-110 from an adjacent monomer (Figure 5B). A conserved salt bridge between Glu-4 and Lys-104 is also present. However, Arg-119 (equivalent to Arg-123 in fused 4-OT) does not participate in a salt bridge because its counterpart in the salt bridge in fused 4-OT and R7, Asp-13, is replaced with Val-13. In the αβ interface, there are no changes in the number of salt bridges with the same two salt bridges present (Asp-74/Arg-119 and Asp-110/Lys-77). In the ββ interface, there is a single salt bridge (Glu-4/Lys-104) (Figure 6). Similar to R7, the Asp-59/Arg-82 salt bridge is not present in the ββ interface due to the shortening of the loop containing Asp-59. Because of the asymmetric arrangement, there are a total of seven salt bridges with the αα interface contributing the most (four salt bridges) (Figure 6). Had F6 adopted a symmetric conformation, these two αβ interface salt bridges would be the only ones conserved in the structure resulting in a total of six salt bridges (Figure 6).

The structures of two members (W0 and Q0) in the Linker 2 cluster were also determined. W0 crystallized with four single monomers in each asymmetric unit (space group P3), and the trimer is generated by applying the symmetric vector. The structure for W0 (2.1 Å resolution) contains residues Pro-1 to Arg-122 (total 127 amino acids) with a loop containing residues 54 and 55 disordered due to a lack of electron density. However, there is no doubt that the structure is that of a symmetric trimer (Figure 5C). The most significant observation about the W0 trimer is that the residues forming salt bridges at the αα, ββ, or αβ interfaces in fused 4-OT are mostly not present in W0 (Figures 5C and 6). They are replaced with non-interacting residues (Figure 6). The only possible salt bridge is between Lys-16 and Asp-49 in the symmetric trimer, reinforced by the intramolecular salt bridge with Asp-104. Hence, formation of the asymmetric W0 trimer would result in no salt bridges, whereas the symmetric structure of W0 has three salt bridges, one salt bridge per αβ interface (Lys-16/Asp-49) (Figure 6).

The structure of Q0 adopts a symmetric trimer, which allows for a maximal number of salt bridges. The structure for Q0 (1.6 Å resolution) contains residues Pro-1 to Ala-123 (total 123 amino acids) in each monomer of the symmetric trimer (Figure 5D). Similar to W0, Q0 lacks most of the residues that would form salt bridges in the αα, ββ, or αβ interfaces in a hypothetical asymmetric trimer. The only possible electrostatic interactions are between Lys-16 and Asp-109 in the αα interface, and between Lys-76 and Asp-109 in the αβ interface in this hypothetical asymmetric trimer. In contrast, the symmetric trimer observed in the crystal structure of Q0 allows for the formation of up to 12 salt bridges. In the symmetric trimer, the interaction between Lys-16 and Asp-49 at the αβ interface is present as it is in other members of the Linker 2 cluster. In addition, there are new salt bridges that are favorable to the formation of a symmetric trimer and these interactions appear to be unique to Q0. A salt bridge is formed between Lys-76 and Asp-109, stabilized by an intramonomer salt bridge with Asp-104 in each αβ interface. Furthermore, three salt bridges are form between Arg-103 and Asp-66 from each monomer at the αβ interface, where inter- and intramonomer interactions form a hexameric salt bridge network (Figures 5D and 6). Thus, in Q0, the formation of twelve salt bridges favors a symmetric trimer (Figure 6).

Based on these structural observations, the number of salt bridges could be a factor in trimer arrangement, whereas the hydrophobic and other polar interactions do not show significant differences. As such, this observation could be useful in predicting the quaternary structures of the timers. Salt bridges are strong non-covalent interactions that can be easily attenuated by salt or exposure to solvent. However, when shielded in a hydrophobic region like those found in protein-protein interfaces, the interaction can be stronger. When one of the residues is lost, it would leave a lone charge in the middle of the hydrophobic patch. In asymmetric trimers, a set of residues is conserved, allowing the formation of salt bridges at the αα interface. In symmetric trimers, the residues forming the salt bridges are not conserved as charged residues, and the advantage an asymmetric trimer is thus also lost.

The structural observations coupled with the bioinformatics analysis (Figure 7) suggest that most members of the Fused 4-OT cluster will form asymmetric trimers (orange nodes) and most members of the Linker 2 cluster will form symmetric trimers (blue nodes). There are two notable exceptions. One is the blue node (lower left) suggesting salt bridges characteristic of a symmetric trimer. The node connects to multiple orange nodes within the Fused 4-OT cluster. However, the sequence is not that of a fused 4-OT, but of a truncated one (72 amino acids). The other one is the orange node (upper right) suggesting salt bridges characteristic of an asymmetric trimer. The node connects to multiple blue nodes within the Linker 2 cluster. This sequence is under investigation.

Figure 7.

The Linker 2 and Fused 4-OT clusters in a sequence similarity network, color coded by conserved salt bridges. The square-shaped blue node is linker 2 and the square-shaped orange node is fused 4-OT. The orange nodes represent sequences with a greater number of the conserved salt bridges shown in the sequences of fused 4-OT, R7, F6 (first three lines, top panel, Figure 6), as compared to the salt bridges shown in the bottom panel of the figure. Conservation of these salt bridges suggests that the proteins/enzymes represented by these sequences will form asymmetric trimers. The blue nodes represent sequences with a greater number of the conserved salt bridges shown in the sequences of W0, Q0, and linker 2 (last three lines, bottom panel, Figure 6), as compared to the salt bridges shown in the top panel of the figure. Conservation of these salt bridges suggests that the proteins/enzymes represented by these sequences will form symmetric trimers. The gray nodes have an equal number of “asymmetric” and “symmetric” salt bridges so the configuration is not clear.

Our interest in the fused 4-OTs stems from the idea that they might share features with possible progenitors for the other four subgroups in the TSF. One obvious feature is the trimeric arrangement. If a fused 4-OT-like protein served as a progenitor for the other subgroups, then the asymmetric and symmetric structures could provide two different scaffolds and potentially different possibilities for diversification. Because the asymmetry doesn’t seem to affect the enzymes’ fitness as 4-OTs they can continue to function as 4-OTs. This proposal has two implications. First, it suggests that there might be asymmetric trimers in the other subgroups. Second, directed evolution studies using both the asymmetric and symmetric structures might have different outcomes. Both possibilities are under study.

Conclusions

A bioinformatics analysis of a fused 4-OT subset (133 sequences) in the tautomerase superfamily indicates that members of one cluster (86 sequences) have more salt bridges if the asymmetric trimer forms, whereas members of the other cluster (47 sequences) have more salt bridges if the symmetric trimer forms. Kinetic and structural characterization of two proteins within each cluster supported this hypothesis. All four proteins function as 4-OTs, where two assemble into asymmetric trimers (designated R7 and F6) and two form symmetric trimers (designated W0 and Q0). These findings can be extended to the other sequences in the two clusters, thereby annotating their oligomer properties and activities.

ABBREVIATIONS

ALS

Advanced Light Source

APS

Advanced Photon Source

DEAE

diethylaminoethyl

ESI-MS

electrospray ionization mass spectrometry

hh4-OT

heterohexamer 4-oxalocrotonate tautomerase

2-HM

2-hydroxymuconate

Kn

kanamycin

LB

Luria-Bertani

MR

molecular replacement

MME

monomethyl ether

4-OT

4-oxalocrotonate tautomerase

PP

phenylenolpyruvate

PEG

polyethylene glycol

PISA

Protein Interfaces Surfaces and Assemblies

RMSD

root-mean-square

SSN

sequence similarity network

SDS-PAGE

sodium dodecyl sulfate-polyacrylamide gel electrophoresis

SFLD

Structure-Function Linkage Database

TSF

tautomerase superfamily

TLS

translation-libration-screw-rotation