Substrate binding to wild type ThAOS.
The original paper that characterized ThAOS showed that it could accept Gly, L-Ala and L-Ser amino acids and acetyl-CoA, pimeloyl-CoA and palmitoyl-CoA acyl-thioester substrates34. It was also shown that ThAOS catalysed both KBL (Gly/acetyl-CoA) and AONS (L-Ala/pimeloyl-CoA) reactions. We used this information in our previous work to measure an expanded substrate scope of ThAOS using a relatively high-throughput kinetic assay which uses the thiol-reactive DTNB reagent (Ellman’s) to monitor CoASH release (Fig. 1A)36. The ThAOS enzyme was active with amino acids l-Ala, Gly, l-Ser and also l-Aba, and could catalyse condensation with acetyl-, propionyl-, butyl-, hexanoyl- and octanoyl-CoA thioester substrates. Therefore, we concluded that the ThAOS biocatalyst has unusually broad substrate scope which suggests it be further expanded.
In this new study we wanted to further explore substrate binding so we took advantage of the inherent UV-vis properties of the PLP-dependent ThAOS. The PLP UV-vis spectrum gives insights into substrate binding, as well as the formation of key intermediates during the catalytic mechanism40. The most intensively studied members of the AOS family have been SPT, ALAS and AONS with a view to understand their substrate specificity within the biosynthetic pathways in which they operate13,14,41. These combined studies have led to a consensus catalytic mechanism of the AOS family with various evidence to support the structure and role(s) of key intermediates and active site residues (Fig. 2). The AOS enzymes exhibit characteristic PLP spectroscopic properties with a clear shift in the UV-vis spectrum upon binding of the amino acid substrate to form the PLP:amino acid external aldimine. The binding of the second, acyl-CoA substrate is thought to cause a conformational change (proposed by Dunathan42) that allows deprotonation of the proton at C-a to generate the key quninonoid/carbanion intermediate. This species subsequently reacts in the C-C forming, Claisen-like condensation reaction to generate a b-keto acid intermediate which is readily decarboxylated to the amino ketone product. The observation of key intermediates, such as the quinonoid, is dependent on the specific AOS/substrate combination.
We studied the binding of ThAOS with its amino acid substrates using UV-vis spectroscopy, according to the methods of Webster et al. and Raman et al that were used to study both E. coli AONS and S. paucimobilis SPT9,43. Recombinant, purified ThAOS displays a smaller, broad absorbance between 330–360 nm and a larger, broad absorbance between 380–480 nm corresponding to the tautomeric forms of the PLP-bound internal aldimines (Fig. S1A and Fig. 2). Titration of the core substrate Gly (0–16 mM) led to a small decrease in the 330 nm peak and a small increase in the 430 nm peak (Fig. S1B). A similar titration of L-Ser (0–16 mM) led to a small increase in the peak at 330 nm and a decrease in the 430 nm peak. With L-Ala (0–16 mM) there was little change in the 330 nm absorbance, a slight decrease and shift (by ~ 2-4nm to 426–428 nm) in the 430 nm peak and the observation of a small, broad peak between 490–505 nm. Taken together this spectroscopic data confirms that these substrates bind to the ThAOS to form PLP:amino acid external aldimines and that L-Ala binding leads to a small amount of PLP:L-Ala quinonoid formation.
Having established amino acid binding, we then explored what happens upon addition of the second substrate, acetyl-CoA to the external aldimine forms of ThAOS. Little change was observed with Gly and L-Ser, but we were interested to observe the clear formation of a peak at 501 nm (with a shoulder at 472 nm) with L-Ala as the substrate, suggesting this substrate combination generates an intense quinonoid species (Fig. S2). The observation of such an intermediate suggests subtle differences in the way each substrate binds to and is stabilized by the enzyme. These results suggested to us that ThAOS could potentially bind a greater range of substrates than previously thought – either non-productively to form a ThAOS:PLP-amino acid external aldimine, or productively, to go on to product aminoketone formation in the presence of an acyl-CoA substrate (Fig. 2). Moreover, we hoped that the ThAOS could be engineered to further expand its substrate scope.
Rational engineering of ThAOS for an expanded substrate scope.
We previously determined the crystal structure of the ThAOS homo-dimer in the internal aldimine form with the PLP bound cofactor (PDBs: 7POA, 7POB and 7POC)36. In the absence of a substrate bound structure we modelled the PLP:Gly external aldimine from the crystal structure of RcALAS (PDB: 2BWP), a homologue from the AOS family. This model highlighted a close interaction (3.5 Å) between the proposed position of the substrate amino acid sidechain and that of Val79 from the other ThAOS monomer within the dimer structure (Fig. 3A). Thus, Val79 appeared to be an important site for amino acid substrate selectivity, and it was targeted for site-directed mutagenesis. This residue is conserved as a Val side chain in all KBL enzymes but replaced by a Ser, Leu and Thr in other AOS enzymes such as SPT, ALAS, AONS and the recently discovered Alb29 (Fig. 3B). This lack of conservation also suggests that changes at this residue might alter the substrate scope, but would not adversely impact the catalytic activity. Our initial hypothesis was that ThAOS V79 variants would display an expanded amino acid substrate scope.
Activity screening of ThAOS V79 variants identifies improved biocatalysts.
Initially, we aimed to reduce size of the Val79 side-chain, ideally permitting the entry of bulkier amino acids besides those already turned over by WT ThAOS. As proof of concept, we prepared the enzymes ThAOS V79A and V79G as initial variants to investigate their substrate scope. The mutant biocatalysts were expressed and purified as the wild type ThAOS (Fig. S5) and were screened against a panel of amino acids (16 mM) and acetyl-CoA (1 mM), by monitoring the reaction via the colorimetric DTNB thiol detection screen at 50°C36,43. In our screen we included all 20 proteinogenic amino acids, plus the unnatural amino acids (UAAs) l-Aba, dl-Alg, l-Cpa, l-Cpg, l-Hsr, l-Nle, l-Nva, l-Orn, l-Phg, l-Pra, l-Trl, L-Oas and a-Ile, (Fig. S3).
Of the 33 amino acid substrates tested with acetyl-CoA as the acyl-thioester substrate, ThAOS V79A and ThAOS V79G displayed activity with 16 amino acids (Table 1). These split into 9 UAAs (l-Aba, dl-Alg, l-Cpa, l-Cpg, l-Hsr, l-Nva, l-Pra and l-Oas) and 7 proteinogenic amino acids (l-Ala, l-Asp, Gly, l-Ile, l-Ser, l-Thr and l-Val). Both variants were significantly faster than the wild type biocatalyst for the original set of four amino acid substrates (l-Aba, l-Ala, Gly and l-Ser), and in general ThAOS V79G was faster than ThAOS V79A. These variants were kinetically characterised in more detail using the DTNB assay, confirming ThAOS V79G to be faster than ThAOS V79A in most cases (Fig. S4, Table S2). With the ThAOS V79A variant, the catalytic efficiency (kcat/KM) for all four original amino acid substrates was improved substantially (by 2.4x – 64.8x fold). Similarly, the ThAOS V79G variant showed improved catalytic efficiency with these four reactions by (2.7x – 57.5x fold), with both variants displaying the highest activity with l-Ser. Having identified Val79 as a critical residue involved in substrate binding, we attempted to obtain other useful and/or interesting ThAOS variants by performing saturation mutagenesis at this position (Table S3). To allow high throughput analysis this expanded set of V79 variants were prepared and expressed in a mini culture format, taking advantage of the thermostability of the biocatalysts.
Under the standard culture conditions, no expression was detected for seven of the possible 17 new variants (ThAOS V79D, F, H, N, P, Q and W). Two variants (ThAOS V79K and V79Y) expressed but were insoluble, but eight (ThAOS V79C, E, I, L, M, R, S and T) were obtained in soluble form (Table S3). These soluble variants were purified from the cell free extract via a heating step at 80°C for 30 minutes, prior to centrifugation (Fig. S5). The variants which were not stable to this treatment were purified by the full Ni2+ affinity chromatography (IMAC) method. The purified ThAOS variants were screened against the broad range of amino acid substrates with acetyl-CoA and the release of CoASH observed using the DTNB assay. It was clear that the ThAOS V79S variant displayed increased activity relative to the wild type ThAOS with the core set of four amino acids (l-Aba, l-Ala, Gly and l-Ser), but also accepted the UAAs, L-Cpg and L-Pra (Fig. 3). Of the other variants, ThAOS V79M and ThAOS V79R turned over only l-Ser to a lesser extent and ThAOS V79T was only active with l-Ala. Amongst the other soluble mutants, ThAOS V79C and ThAOS V79E displayed no detectable activity with the amino acid substrates tested. The ThAOS V79I variant showed the same substrate profile as wild type but with a decreased rate, whilst the ThAOS V79L turned over l-Cpg as well as the original four substrates, and at a slightly improved rate than the wild-type biocatalysts. The retention of activity observed with the ThAOS V79I/L variants is understandable since these are all similar, hydrophobic side-chains.
We sought to understand the differences in the activities of the ThAOS V79 variants by simple consideration of the chemistry of each side chain (Fig. S6). It appears that the presence or absence of a substituent (e.g. methyl) at the β-position of the amino acid side chain influences not only the substrate scope but also the catalytic activity. Furthermore, this explains the relative lack of distinction between the ThAOS V79A and ThAOS V79G variants. It is clear that the residue at this position in the ThAOS structure is one of the important determinants of amino acid substrate selectivity. It is also notable that the Val79 residue is not conserved across other members of the AOS family that display differences in their amino acid selectivity e.g. in the well characterized S. paucimobilis SPT which uses L-serine as a substrate, this residue is Ser102 (Fig. 2B). The residues in this 13 amino stretch are part of a key flexible loop where the AOS enzymes display conformational flexibility in key residues involved in substrate binding and catalysis. The AOS enzymes have another flexible loop containing the conserved “PATP” motif near the C-terminus that also undergoes conformational change. Our work on S. paucimobilis SPT captured the key PLP:L-Ser external aldimine complex (PDB: 2W8J) showed that key arginine residues (R378 and R390) are involved in recognition of the carboxylate of L-Ser; the side-chain of Arg378 swings into place upon substrate binding and then “passes on” this carboxylate to Arg390 once the C16-CoA substrate binding has caused the conformational change that induces deprotonation at C-a (Fig. 2)43. The resulting PLP:L-Ser quinonoid intermediate is the nucleophile that forms the C-C bond in the b-keto acid intermediate. Mutagenesis of these Arg residues showed that the highly-conserved Arg378A was x40 fold less active than the wild type and this variant could not stabilize the quinonoid. It is interesting to note that KBL is the only AOS member that does not decarboxylate this unstable intermediate (2-amino-3-ketobutyrate, AKB) to give aminoacetone. It protects the AKB by forming a complex between KBL and L-threonine dehydrogenase (TDH). Together KBL and TDH are involved in a glycine/acetyl-CoA metabolic cycle with L-Thr via the AKB intermediate16. The fact that ThAOS in isolation can catalyse aminoketone formation with various substrates underscores its unusual, but useful, properties.
Expanded acyl-CoA substrate scope of ThAOS V79 variants.
Our next goal was to determine the impact, if any, of variations at Val79 on the acyl-CoA thioester substrate scope. In our previous study, we noted that wild-type ThAOS turned over all four of the amino acid substrates (L-Aba, L-Ala, Gly, L-Ser) with acetyl-, propionyl-, butyryl-, hexanoyl- and octanoyl-CoA35. Therefore, we used the DTNB assay to screen the three most promising variants (ThAOS V79G, V79A and V79S) derived from the amino acid substrate screen with acyl-CoA thioesters of increasing acyl chain length (C2, C3, C4, C6, C8, C10 and C12), as well as benzoyl-CoA (Table S4-S6).
We were pleased to observe that the acyl-CoA substrate range of ThAOS V79G, V79A and V79S was retained across the C2-C8 chain length when compared to wild type ThAOS. We noted that the ThAOS V79G variant was active with all four core amino acids and C10 and C12 acyl-CoA substrates and could also condense Gly and L-Ala with benzoyl-CoA albeit at very slow rates (Table S4). It was also pleasing to observe that two of the new UAA substrates (L-Cpg, L-Hsr) could also be condensed with acyl-CoA thioesters up to C12 (Table S4). The ThAOS V79A and V79S variants also featured a slightly improved acyl-CoA substrate scope relative to the WT ThAOS (Tables S5 and S6). The ThAOS V79S is distinguishable in that it is the one variant that was able to catalyse condensation of the UAA substrate L-Pra with C2-C8 acyl-CoAs at modest rates. Furthermore, this variant displayed detectable activity with L-Ala, Gly, L-Ser and L-Cpg and benzoyl-CoA (Table S5).
Acyl-CoA independent quinonoid formation in the ThAOS V79G variants.
Since the ThAOS V79G variant displayed the broadest activity (Table 1) we have begun to probe how it behaves in the presence of a sub-set of substrates. Here we monitored formation of the key quinonoid intermediate between 490–510 nm to explore the ThAOS variants in terms of amino acid substrate binding alone, and in the presence of the acyl-CoA thioester. The ThAOS V79G variant was incubated with Gly, L-Ser and L-Ala. When compared with wild type ThAOS we noted similar changes to the UV-vis spectrum when the amino acid was added, although in each case there was a small, but clear absorbance between 480–520 nm indicating the formation of a ThAOS V79G PLP:amino acid quinonoid intermediate (Fig. S7 A-C). We also noted acetyl-CoA binding to the ThAOS V79G PLP:Gly complex led to enhancement of this peak (Fig. S7A). In contrast, addition of the acetyl-CoA to the ThAOS V79G PLP:L-Ser complex led to formation of a broad, intense quinonoid signal between 490–520 nm (Fig. S7B), with a shoulder at 468 nm. Moreover, addition of the acyl-CoA thioester substrate to the ThAOS V79G PLP:L-Ala external aldimine generated the most intense quinonoid signal between 490–520 nm (Fig. 7C), also with a shoulder 468 nm). Clearly, the removal of the Val79 side-chain has impacted the formation and stabilization of the ThAOS PLP:amino acid complex and binding of the acyl-CoA thioester substrate leads to deprotonation at C-a and quinonoid formation, supporting the overall proposed AOS catalytic mechanism (Fig. 2). Since the ThAOS V79G variant also accepted L-Asp, L-Thr and L-Val (Table 1) we looked at quinonoid formation with these three substrates and were pleased to observe the characteristic formation of the peak around 500 nm of varying intensities for each amino acid (Fig. S7 D-F). It is clear that this relatively easy assay is a useful tool to study not only substrate binding, but also formation of this key catalytic quinonoid intermediate.
Quinonoid formation has been studied in only a few members of the AOS family. We have studied the E. coli AONS and provided supporting evidence for quinonoid formation using L-L-Ala methyl ester (L-Ala-OMe) as a substrate mimic18. This ligand bound to the enzyme to form the PLP:L-Ala-OMe external aldimine and upon addition of pimeloyl-CoA two new species were observed by UV-Vis spectroscopy; an intense absorption at 486 nm indicative of formation of the AONS: PLP:L-Ala-OMe quinonoid species and a novel peak at 454 nm which was due to accumulation of the β-ketoacid methyl ester aldimine complex that cannot undergo enzymatic decarboxylation. Taken together it suggests the clever use of the Me-ester substrate mimic prevents decarboxylation and is a useful probe of the AOS mechanism (Fig. 2). This Me-ester trick was also recently used by Chun and Narayan and colleagues to stall the SxtA AONS and, in the presence of D2O, allow selective a-deuteration of a panel of 24 natural and UAA a-amino ester substrates38.
Another elegant example of quinonoid trapping is the study of S. paucimobilis SPT by Ikushiro and colleagues. They initially used an acyl-CoA thioether substrate mimic (S-(2-oxoheptadecyl)-CoA, a nonreactive thioester analogue of palmitoyl-CoA, to “trick” the SPT PLP:L-Ser complex27. This mimic bound to the enzyme and caused a conformational change which led to deprotonation at C-a and PLP:L-Ser quinonoid formation observed at 493 nm. Since the quinonoid could not react with the acyl-CoA thioether the authors used NMR to measure a rate acceleration of C-a deprotonation of more than 100 fold upon binding of the second substrate. This study also supported the over-arching AOS mechanism that we proposed based on our earlier work on the AOS enzyme E. coli AONS9 .More in-depth kinetic analyses (both steady-state and stopped flow) of the ALAS enzyme that uses L-Ala and succinyl-CoA have also been reported by Ferreira and colleagues and from these combined studies the key conserved residues involved in AOS catalysis have been proposed13. The results presented here are the first to use knowledge of the catalytic mechanism to explore rational mutagenesis of an AOS enzyme and generate a biocatalyst with a much broader substrate specificity. In further studies, we aim to investigate formation, stability and breakdown of these quinonoid species which should provide more detailed insights into the substrate specificity across the whole AOS family.
Thermostability and spectroscopic properties of selected ThAOS variants with expanded substrate scope.
Having obtained a sub-set panel of four competent biocatalyst variants via saturation mutagenesis at ThAOS V79 (V79A, V79G, V79L, V79S), we set out to further investigate if the mutations had affected the thermostability of the biocatalyst. This was an important consideration in terms of selecting the best ThAOS variant that could be used for application in chemical synthesis. Since the ThAOS is from a thermophilic bacterium we decided to study their respective thermostabilities at elevated temperatures (Fig. S8). The wild type ThAOS was previously found to be stable to incubation for two hours at 70°C21. The ThAOS V70L variant displayed comparable residual activity (92.1%) to the WT ThAOS (88.6%), while the V79A variant (77.6%) and V79S (79.4%) mutants were slightly less stable. We then tested how stable the variants were when incubated at 70°C (Fig. S8A). Unfortunately, the ThAOS V79G variant which exhibited the biggest improvement in substrate scope during the screening process, displayed a significant drop in activity (retaining 56.0%) compared to the wild type biocatalyst. We then analysed how stable the variants were upon incubation at 90°C for 30–120 mins. The WT was to reduced ~ 15% activity after 120 mins and the V79S, V79G and V79L displayed varying retention of activity (45%, 5% and < 5% respectively). However, ThAOS V79A retained ~ 60% activity after incubation at this elevated temperature. By evaluating both the improved substrate scope and thermostability we then decided to take forward the ThAOS V79A variant for scaled-up synthesis of a ThAOS-derived pyrrole product in combination with the KPR.
The ThAOS V79 variants are viable biocatalysts with alternative, inexpensive acyl N-acetylcysteamine (SNAc) substrates.
Across the AOS family it has been shown that the acyl-CoA thioester substrates display low µM kinetic constants and high turnover rates that is ideal for synthesis. However, the high cost of acyl-CoAs prohibits their use in biocatalysis unless this issue can be overcome by inclusion of an acyl-CoA recycling system. This was achieved in our previous work with wild type ThAOS where we used acyl-CoA synthetase (ACS) to convert the CoASH product back into acyl-CoA (Ashley). Alternatively, a number of simple CoA substrate mimics have been reported that can be used in place of the acyl-CoA.44 Of these, the most widely-used are thioesters of the low-molecular- weight CoA fragment, N-acetylcysteamine (SNAc, Fig. 5). Enzymes capable of using these simple acyl-SNAc thioester substrates in place of an acyl-CoA thioester or pantetheine thioester include thioesterases, ketosynthases, ACP synthases and many others, (reviewed in 44).
Therefore replacing the expensive acyl-CoA substrates of AOS enzymes with SNAc-thioesters would greatly improve the synthetic potential of these biocatalysts in preparing valuable aminoketone products or being combined into chemo- and bio-catalytic cascades. The only issue that could prevent using these substrate mimics is a kinetic penalty, since the acyl-SNAc substrates display very high mM binding constants. An AOS homologue, M. wollei SxtA, has recently exhibited improved activity with acyl-SNAcs in the presence of a pantetheine-like auxiliary molecule, which mimics the presence of the acyl-CoA pantetheine arm39. This work prompted us to explore the use of acyl-SNAc in combination with the newly-discovered ThAOS V79 variants.
Our hypothesis is based on the proposed AOS catalytic mechanism (Fig. 2), that only upon acyl-CoA binding will the enzyme catalyse formation of the nucleophilic quinonoid intermediate. However, since the ThAOS V79A and V79G variants generate a quinonoid in the presence of the amino acid substrate alone, we hoped that these biocatalysts would react with acyl-SNAc substrates, albeit at high concentrations. Acetyl-SNAc (N,S-diacetylcysteamine) was therefore prepared from cysteamine hydrochloride and acetic anhydride using a published method (see SI). The wild type ThAOS and three active variants (V79A, V79G and V79S) were screened for Claisen-condensation activity with acetyl-SNAc against a panel of three amino acids (L-Aba, L-Ala and Gly) with the DTNB assay at 50°C (Fig. S9). It is important to note that wild type ThAOS exhibited no detectable activity with any of the substrate combinations. In contrast, we were delighted to observe activity with all of the ThAOS variants. The ThAOS V79A displayed activity with acetyl-SNAc and L-Aba and Gly, the ThAOS V79G variant was active with L-Aba, L-Ala and Gly and the ThAOS V79S variant was also active with all three substrates with Gly displaying a turnover of 0.14 min-1.
The results of this screen allowed us to test the synthetic usefulness of the new reaction using combinations of ThAOS variants and amino acids with acetyl-SNAc. To do this we coupled the AOS biocatalytic reaction with the KPR (at elevated temperature) as we had done previously. The reaction was performed in the presence of methyl acetoacetate (MAA), with the aim of capturing the aminoketone product as a pyrrole, 1 (Fig. 5). The three biocatalysts were screened using high catalyst loadings (5 mgmL-1) with Gly, acetyl-SNAc and MAA (32 mM each) at a variety of elevated temperatures for 16 hours. Reactions were analysed by HPLC as described in our original paper36. The results showed that ThAOS V79A worked best with Gly and acetyl-SNAc at 60 oC with 35% analytical yield of pyrrole 1 (Fig. 5). We explored if the pyrrole yield could be increased with the higher loading of acetyl-SNAc (32–800 mM). However, no significant increase in yield was observed (Fig. S10).
Since the ThAOS V79A variant was the most thermostable under long incubation times at elevated temperatures (Fig. S8) it was used in a preparative scale reaction (10 mL, 32 mM glycine, 32 mM acetyl-SNAc, 32 mM MAA, 15 mg/mL ThAOS V79A, 100 mM HEPES, pH 7.5, 16 hrs). The pyrrole product 1 was extracted and purified and its structure confirmed using NMR (Fig. S11 and S12). Taken together, this result shows that the rationally engineered ThAOS V79A variant could be used to prepare various products using inexpensive acyl-thioester starting materials.
X-ray structure of the improved, engineered ThAOS V79A in complex with the inhibitor L-Pen.
Inspired by the success of the engineering campaign to deliver improved variants of ThAOS at residue V79, we endeavored to crystallise the variants and solve their crystal structures with bound ligands. This would allow a comparison with the wild type ThAOS and might provide molecular insight into the origin of the broadened substrate scope of the engineered variant. Screening revealed that ThAOS V79A gave crystals that were not suitable for structural studies and we also screened unsuccessfully with various natural amino acid and UAA substrates that came from the screens (Fig. S3).
As a final probe to study ligand binding we used the well known PLP enzyme inhibitor L-penicillamine (L-Pen) and monitored the reaction using UV-vis spectroscopy. We used L-Pen since we had studied the related AOS enzyme S. paucimobilis SPT described by Lowther et al.45. The D-Pen enantiomer is an FDA-approved drug used for the treatment of Wilson’s disease and also binds specifically to PLP enzymes46. Incubation of L-Pen (1 mM) to the wild type and both ThAOS V79A and V79G variants rapidly decolourised both reactions and led to the formation of a broad peak between 310–380 nm, with an absorbance maximum at 330 nm (Fig. S13 A, B and C). This data is consistent with the formation of a covalent ThAOS PLP:L-Pen ring-closed thiazolidine adduct and this was refined to an excellent fit and geometry. We subsequently used the L-Pen ligand in structural studies and when ThAOS V79A crystals were soaked with L-Pen, prior to flash cooling and data collection, clear ligand density was observed in each active site (Fig. 6A). The crystal structure was determined to 1.5 Å resolution in P1 space group using the wild type ThAOS structure (PDB ID: 7POA) as a molecular replacement model. The asymmetric unit contained two chains forming the stable dimer, and the final model refined to an Rwork of 0.16 and Rfree of 0.19 (Fig. S14). The data collection and refinement statistics are shown in Table S7.
L-Pen was captured in the active site of ThAOS V79A in a conformation typical of amino acids co-crystallised in the active site of AOS enzymes. The L-Pen substrate amine is covalently linked to the PLP cofactor, and the substrate carboxylate is chelated by Arg366. The five-membered ring of the inhibitory thiazolidine intermediate is clearly resolved, and sits in the position of an amino acid substrate sidechain. Most interesting is the distinction between WT ThAOS and the engineered, more active variant, ThAOS V79A. For this reason, the structure of ThAOS internal aldimine was overlaid onto the structure of L-Pen-bound ThAOS V79A. Despite a long history of use as an inhibitor of PLP-dependent enzymes, this is the first published crystal structure of a PLP enzyme in complex with l-Pen.
The two chains of the ThAOS V79A variant align with around 0.4 Å RMSD Cα to wild type ThAOS structure over the 397 residues. There is no significant structural change in the active site in the V79A variant with PLP:L-Pen bound when compared to the wild type PLP-bound ThAOS, with no shift in the loop around the altered residue. The distance between A79 and the thiazolidine methyl group and is 4Å and modelling the distance to the native V79 gives a distance of 3 Å to the thiazolidone (Fig. 6B). It appears that the removal of the valine side chain increases the volume of the ligand binding site without any large-scale remodelling of this region.
Comparison of penicillamine inhibitor binding in structural homologues.
Searching the PDB for the L- and D- forms of the penicillamine ligand identifies two unpublished structures of cysteine desulfurase (CSD) enzymes with structural homology to ThAOS with bound penicillamine ligand: NifS from Helicobacter pylori (PDB ID: 7XES); and SufS from Bacillus subtilis (PDD ID: 7XEN). There is also a structure of the SufS without penicillamine (PDB ID: 7XEN). The monomers of these two proteins align to ThAOS-V79A with overall RMSD Cα of 1.3 Å (7XES) and 1.2 Å (7XEN) (Fig. S15A). The primary structural difference between ThAOS and the cysteine desulfurase enzymes is in the position of the N-terminal region of these proteins and an extended beta-elbow that is present in the cysteine desulfurases (Fig. S15A). In ThAOS-V79A, the first 40 amino acids adopt an alpha helix and extended loop arrangement, forming a large interface with the partner chain (Fig. S15B); whereas, in the cysteine desulfurase enzymes the N-terminal region is shorter; and, in the case of the B. subtilis enzyme, it forms a twisted helix that mainly participates in interactions with its own chain (Fig. 15). These regions are shifted by a rotation of 110° between the cysteine desulfurase enzymes and ThAOS. The beta-elbow region in the cysteine desulfurase enzymes participates in the dimerisation interface and due to this additional region, the quaternary arrangement of these enzymes differs from ThAOS (Fig. 16A-B). It appears that it is the entrance to this cavity that has been engineered in the more active ThAOS V79 mutants.
The active sites of NifS and SufS both have a bound penicillamine ligand captured in the external aldimine form between the penicillamine and PLP cofactor (Fig. S16A). There position of the PLP and penicillamine is well conserved between the three proteins other than the ring closure to form the thiazolidine in the ThAOS. There is a clear cavity at the dimer interface in each enzyme where the substrate can bind (Fig. 16B). There is only 20% sequence identity between ThAOS and the bacterial CSDs, and residue conservation in the active site is limited to histidine (H136), and arginine (R366); the active arginine (R243 in ThAOS) is conserved, but in different sequence positions in the NifS and SufS proteins (Fig. S17). The wider active site and ligand binding regions of the three proteins differ considerably, with distinct ligand binding tunnels in the three enzymes (Fig. S17).