4-alkyl-L-(Dehydro)proline biosynthesis in actinobacteria involves N-terminal nucleophile-hydrolase activity...

4-alkyl-L-(Dehydro)proline biosynthesis in actinobacteria involves N-terminal nucleophile-hydrolase activity...





Assignment of lmbA and its homologues in ALDP formation


Comparative analysis of the available biosynthetic gene clusters of ALDP-containing metabolites21,22,23,24,25,26,27, that is, lincomycin A (lmb), anthramycin (ant), porothramycin (por), sibiromycin (sib), tomaymycin (tom) and hormaomycins (hrm), revealed a set of highly conserved genes, for example, lmbB1, lmbB2, lmbY and lmbA in lmb cluster, from which the deduced proteins are thus readily correlated with the common reactions occurring during the biosynthesis of ALDP residues (Fig. 2). The oxidase pairs, for example, LmbB1 and LmbB2, are known to be involved in the transformation of L-tyrosine into pyrroline intermediate 2 or 3 (refs 17, 18, 19, 20), whereas the roles of LmbY- and LmbA-like proteins remain unclear.



Figure 2: Comparative analysis of available ALDP-related biosynthetic gene clusters.
Figure 2


The genes coding for ALDP formation are coloured. ant, for anthramycin; por, for porothramycin; sib, for sibiromycin; tom, for tomaymycin; hrm, for hormaomycins and lmb, for lincomycin A. For sequence identities of deduced proteins, LmbA is homologous to Ant6 (72.3%), Por11 (72.9%), SibY (58.0%), TomL (50.0%) and HrmG (64.6%); LmbY is homologous to Ant14 (49.8%), Por15 (48.2%), SibT (53.5%), TomJ (50.2%) and HrmD (46.3%); LmbB1 is homologous to Ant12 (40.8%), Por13 (44.6%), SibV (54.4%), TomH (52.8%) and HrmF (46.2%); and LmbB2 is homologous to Ant13 (31.2%), Por14 (30.9%), SibU (38.7%), TomI (37.3%) and HrmE (42.9%).






LmbY and its counterparts (that is, Ant14, Por15, SibT, TomJ and HrmD), which share sequence homology with putative coenzyme F420-dependent oxidoreductases, are believed to function in olefinic double bond reduction at a later stage of ALDP formation16,28. The biosynthesis of the F420 cofactor involves glutamyl transfer activity29. Despite the lack of experimental evidence, this activity has long been speculated to arise from LmbA and its counterparts (that is, Ant6, Por11, SibY, TomL and HrmG) because they are homologous to various γ-GTs. γ-GTs catalyse γ-glutamyl transfer/hydrolysis with Ntn-hydrolase activity through an acyl-nucleophile enzyme intermediate2. Mechanistically, the production of 4 in the biosynthetic pathways of ALDP residues could occur through the hydrolysis of 3 to release the diene group (Fig. 1b). This process would require nucleophilic attack by water onto the α-keto group of the side chain to produce the co-product, oxalic acid, either directly or indirectly after the formation of an acyl-nucleophile enzyme intermediate. Therefore, in the search for an alternative to the cofactor-forming proteins for supporting LmbY-like activity, the mechanistic consistency inspired us to ask whether LmbA and its counterparts share a similar acyl transfer/hydrolysis process with γ-GTs during the conversion of pyrroline intermediate 3 into diene product 4. However, no evidence exists suggesting that the known γ-GTs and other Ntn-hydrolase fold proteins possess C–C bond cleavage activity.


Correlation of related genes with ALDP formation


To validate the necessity of LmbA-like activity for ALDP formation, we inactivated lmbA in the lincomycin-producing S. lincolnensis strain (Supplementary Fig. 1). This mutation substantially lowered the yield of lincomycin A by60-fold relative to that of the wild-type strain; then, the in trans expression of lmbA in the ΔlmbA mutant strain partially restored the production capacity, producing lincomycin A in a yield of 20% of that produced by the wild-type strain (Fig. 3a). Clearly, LmbA plays a vital role in lincomycin biosynthesis. A similar complementation effect was found by substituting lmbA with ant6, the lmbA homologue in ant cluster (Fig. 3a), demonstrating the functional identity between LmbA and its pathway-specific counterparts for ALDP formation.



Figure 3: In vivo functional characterization of lmbA and its orthologue lmbA2991.
Figure 3



(a) Product profiles of S. lincolnensis strains, including the wild-type control (I) and mutants (II, for the ΔlmbA single mutation; III, for homologous complementation of the ΔlmbA mutant by lmbA; IV, for heterologous complementation of the ΔlmbA mutant by ant6; V, for ΔlmbA-lmbA2991 double mutation; VI, for the ΔlmbD single mutation). (b) The assembly of PPL and 6 that is mediated by a discrete NRPS system involving LmbD activity.






The ΔlmbA mutant strain still produced lincomycin A (Fig. 3a), indicating that an orthologue outside lmb cluster partially compensates for the loss of lmbA. Indeed, sequence analysis of the S. lincolnensis genome revealed lmbA2991 (48.4% identity to lmbA) (Supplementary Fig. 2). As anticipated, further inactivation of this gene in the ΔlmbA mutant strain led to the complete abolishment of lincomycin production (Fig. 3a). Importantly, either the single (ΔlmbA) or the double (ΔlmbA-lmbA2991) mutant strain accumulated intermediate 6, the thiol ergothioneine (EGT) S-conjugated lincosamide derivative, but did not produce the specific ALDP residue PPL; in contrast, the ΔlmbD mutant control strain produced both (Fig. 3a)30. We recently established that a discrete non-ribosomal peptide synthetase system involving LmbD activity catalyses the assembly of 6 and PPL to provide a key EGT S-conjugated intermediate (7) in the lincomycin biosynthetic pathway (Fig. 3b)30. Clearly, the significant decrease (for the ΔlmbA single mutation) or complete abolishment (for the ΔlmbA-lmbA2991 double mutation) of lincomycin production could be attributed to the lack of LmbA or both LmbA and LmbA2991 activities in the mutant strain, which caused PPL supply and, thus, its subsequent assembly with 6 for molecular tailoring and maturation, to fail.


Typical autoproteolytic activity for self-activation


Despite numerous attempts, the expression of LmbA and its counterparts (for example, Ant6 and SibY) remained highly refractory to various approaches for soluble protein preparation in Escherichia coli. Because the interaction of functionally associated proteins may result in a solubilization effect, we developed an E. coli system to co-express LmbA-like proteins with their upstream enzymes in the pathways, that is, LmbB1-like dioxygenases. Then, we examined the availability of soluble proteins using an in vivo activity assay in the presence of L-DOPA (1) prior to purification. The control systems included the recombinant E. coli strains, each of which expressed dioxygenase alone, and their abilities to transform 1 (colourless, [M+H]+m/z: calcd. 198.0761 for [C9H12NO4]+, found 198.0756) into pyrroline 2 or 3 (bright yellow, [M+H]+m/z: calcd. 212.0553 for [C9H10NO5]+, found 212.0556) were confirmed by the characteristic colour change and high-performance liquid chromatography with mass spectrometric detection (HPLC-MS). The permutation of LmbA and its homologues with LmbB1 and its homologues (for example, Ant12 and SibV) revealed the active pair of Ant6 and Ant12 (Supplementary Fig. 3). Their combination in E. coli produced a new compound (5, [M+H]+m/z: calcd. 140.0706 for [C7H10NO2]+, found 140.0707) that shares its molecular weight with the expected product diene 4 (5 was demonstrated in vitro to be a tautomer of 4). Because the transformation of 2 or 3 occurred, the recombinant Ant6 protein should be active in E. coli and, thus, soluble under the conditions in which Ant12 was co-expressed.


We then purified Ant6 from the co-expressing E. coli system. Three related products with different molecular weights were obtained (Fig. 4a), consistent with the notion that Ant6 is a γ-GT-like Ntn-hydrolase fold protein that is capable of auto-catalytically processing its full-length (67 kDa) precursor into the large (47 kDa) and small (20 kDa; this subunit appeared obscure, partially because of protein instability) subunits to construct a functional heterodimer (Supplementary Fig. 4). Precursor cleavage appeared to occur during the expression and purification process, and the full-length remains (60% of Ant6-related protein products) could be unfolded/misfolded products, which were resistant to further conversion. The structure of the Ant6 precursor was predicted online using the I-TASSER approach, which revealed high similarity to the crystal structures of Bl-γ-GT-T399A (PDB: 4Y23) from Bacillus licheniformis and Ec-γ-GT-T391A (PDB: 2E0W) from E. coli K-12. Thus, both of these were selected for homology modelling to identify the proteolytic site of Ant6 (Supplementary Fig. 5). Bl-γ-GT-T399A and Ec-γ-GT-T391A are γ-GT precursor mimics31,32, in which the catalytically important nucleophilic residue Thr is mutated to Ala, and they lack the self-processing activity for the hydrolysis that occurs between Thr and the immediately upstream residue Glu/Gln. Comparative analysis identified the corresponding residues D429 and T430 in the Ant6 precursor, which constitute a motif that is highly conserved in all LmbA-like proteins (Supplementary Fig. 6). This finding suggested that the hydrolytic cleavage of the Ant6 precursor proceeds between these two residues to produce the large (429-aa) and small (192-aa) subunits; these sizes are in good agreement with the experimental values described above. To validate the necessity/variability of D429 and T430 residues, we prepared Ant6 variants by single mutation of D429A, T430A, T430S and T430C and double mutation of D429A&T430A. Although Ant6-D429A, Ant6-T430A, Ant6-T430C and Ant6-D429A&T430A were observed solely in full-length forms and accordingly lost their self-processing abilities, the purification of Ant6-T430S revealed the47-kDa large and20-kDa small subunits that arise from the cleavage of the67-kDa precursor, albeit in lower yields (Fig. 4a). These findings demonstrate the specificity of Ant6’s autoproteolytic activity in which the most similar residue Ser can partially substitute for T430 in the autoproteolytic process.



Figure 4: Analysis of autoproteolytic activity on SDS–PAGE gels.
Figure 4



The asterisk indicates an unknown impure protein associated with the purification of γ-GT homologues/variants. (a) Ant6. Lane 1, protein standard; Lane 2, Ant6 wild-type; Lane 3, Ant6-T430S; Lane 4, Ant6-T430C; Lane 5, Ant6-D429A&T430A; Lane 6, Ant6-T430A; Lane 7, Ant6-D429A. (b) LmbA2991. Lane 1, protein standard; Lane 2, LmbA2991 wild-type; Lane 3, LmbA2991-D420A; Lane 4, LmbA2991-T421A; Lane 5, LmbA2991-D420A&T421A; Lane 6, LmbA2991-T421C; Lane 7, LmbA2991 T421S. (c) Reconstituted LmbA2991 heterodimer (HD) large (L) and small (S). Lane 1, protein standard; Lane 2, heterodimer wild-type; Lane 3, LmbA2991 HD L-D420A&S; Lane 4, LmbA2991 HD L&S-T1A; Lane 5, LmbA2991 HD L&S-T1C; Lane 6, LmbA2991 HD L&S-T1S.






LmbA2991, the orthologue of LmbA in S. lincolnensis, was readily prepared in E. coli. As expected, the purification produced its (perhaps unfolded/misfolded) full-length (67-kDa) precursor and the large (46-kDa) and small (21-kDa) subunits (Fig. 4b), and the quality achieved was much higher than that found using Ant6. Most LmbA2991 precursor products (85% of the related protein products) were self-cleaved, as demonstrated by gel electrophoresis. Additionally, the cleaved products, particularly the small unit, appeared to be more stable than those produced from the Ant6 precursors. N-terminal sequencing of the purified products of LmbA2991 revealed two sequences: ‘GSSHHHHHHS’, which is shared by the full-length precursor and the 420-aa large subunit (consistent with the designed construct in which the target protein was produced in an N-terminally 6 His-tagged form), and ‘TCHLDVVDRW’, which is specific for the 197-aa small subunit (Supplementary Fig. 7), whose molecular weight was further established by high-resolution (HR)-MS (m/z: calcd. 21312.16, found 21312.63) (Supplementary Fig. 8). Similar to Ant6 precursor processing, the autoproteolysis of the LmbA2991 precursor occurs specifically between the conserved residues D420 and T421, leading to the release of Thr, which resides at the N-terminus of the small subunit in the newly generated heterodimer. In contrast, both the single mutations of LmbA2991 (D420A, T421A, T421S and T421C) and the double mutation (D420A&T421A) completely abolished its autoproteolytic activity. Indeed, all enzyme variants were present solely in full-length precursor forms (Fig. 4b), indicating the stringency of LmbA2991 in self-processing, for which the key nucleophilic residue Thr is irreplaceable.


Unusual Ntn-hydrolase activity for C-C bond cleavage


Because pyrroline 2 or 3 is unstable and its purification and homogeneous chemical synthesis are difficult19, we prepared this substrate in situ using LmbB1-catalysed oxidation reaction to examine the activities of Ant6 and LmbA2991 in vitro. LmbB1 was expressed and isolated from E. coli. Its dioxygenase activity was then reconstituted by incubation with FeSO4.7H2O, ascorbic acid and dithiothreitol33, resulting in an active enzyme that was capable of rapidly converting L-DOPA into pyrroline 2 or 3 for the following tests (Fig. 5a).



Figure 5: Examination of Ntn-hydrolase activity by HPLC-ESI-MS analysis.
Figure 5



Pyrroline substrate (2 or 3) was prepared in situ through LmbB1-mediated transformation of L-DOPA during the conversion to diene product (5) in the absence or presence of the catalysts. (a) Validation of Ntn-hydrolase activity of Ant6 and LmbA2991. I, negative control (without the catalyst Ant6 or LmbA2991); II, Ant6; III, LmbA2991 and IV, 5 standard. (b) Production of imine diene 5 through the tautomerization of enamine diene 4, which results from either LmbB1 and Ant6 (or LmbA2991)-mediated cascade transformations (the derivatization of the co-product oxalic acid is indicated) or chemical deprotection of the synthesized mimic 4′. (c) Evaluation of self-activation dependence for Ntn-hydrolase activity of Ant6 and LmbA2991. I, Ant6-D429A; II, Ant6-T430A; III, Ant6-T430S; IV, Ant6-T430C; V, Ant6-D429A&T430A; VI, LmbA2991-D420A; VII, LmbA2991-T421A; VIII, LmbA2991-T421S; IX, LmbA2991-T421C and X, LmbA2991-D420A&T421A. (d) Independent examination of Ntn-hydrolase activity of LmbA2991. I, LmbA2991 HD L&S; II, LmbA2991 HD L-D420A&S; III, LmbA2991 HD L&S-T1A; IV, LmbA2991 HD L&S-T1S and V, LmbA2991 HD L&S-T1C.






The addition of the Ant6 catalyst, which was a mixture of the full-length and cleaved forms, into the above reaction mixture led to the nearly complete conversion of pyrroline 2 or 3 to 5 ([M+H]+m/z: calcd. 140.0706 for [C7H10NO2]+, found 140.0705) (Fig. 5a), the product also observed in the E. coli system in which both Ant6 and the dioxygenase Ant12 were co-expressed. This finding is consistent with the hypothesis that a diene product is yielded through the elimination of the terminal 2-C unit of the side chain. This conversion was scaled up, and subsequent spectral analysis suggested that 5 unlike 4, which has previously been proposed to be an enamine resulting directly from the cleavage of 2 or 3, is a tautomeric imine (Fig. 5b). To examine the tautomerization tendency of 4, we synthesized a stable enamine mimic of 4 (4′, Fig. 5b), by protecting the carboxylate with a tert-butyl group and the amino group with a tert-butyloxy carbonyl group. Indeed, after deprotection, product analysis revealed 5 but no 4, indicating that 4 is extremely unstable and tends to be converted to tautomer 5 by a diene shift. Meanwhile, the Ant6-containing reaction mixture was subjected to selective extraction and derivatization with N-methyl-N-(tert-butydimethylsilyl)-trifluoroacetamide (MTBSTFA) (Fig. 5b), leading to the observation of the derivative 8 by gas chromatography with MS detection (Supplementary Fig. 9). Therefore, the co-product is oxalic acid. This result explains the fate of the terminal 2-C unit of the side chain and confirmed that the cleavage of 2 or 3 proceeds through hydrolysis. The production of 5 was also observed in the reaction mixture in which Ant6 was replaced with an equal amount of LmbA2991 (Fig. 5a), confirming that LmbA2991 shares this unusual Ntn-hydrolase activity for C–C bond cleavage during the conversion of pyrroline 2 or 3. However, the conversion efficiency of LmbA2991 was substantially decreased (37% of that of Ant6) under the same reaction conditions, despite the catalyst mixture being of higher quality and stability and containing more cleaved (active) forms (as determined by the extent of self-activation).


Catalytic similarities and differences of Ant6 and LmbA2991


Taking advantage of the cascade assay system involving LmbB1 activity, we examined whether each mutant Ant6 or LmbA2991 catalyst functions in the conversion of pyrroline 2 or 3 to diene 5. All full-length variants lacking autoproteolytic activity, including the derivatives from both Ant6 (that is, D429A, T430A, T430C and D429A&T430A) and from LmbA2991 (that is, D420A, T421A, T421S, T421C and D420A&T421A), were incapable of catalysing the hydrolysis of 2 or 3; in contrast, the variant Ant6-T430S, which does possess autoproteolytic activity for self-processing, was capable of this hydrolysis (Fig. 5c), albeit with activity lower than that of Ant6 wild type as judged by the yields of 5 and apparent reaction rates/efficiencies (Supplementary Fig. 10). These results demonstrate that the unusual Ntn-hydrolase activities of both Ant6 and LmbA2991 for C–C bond cleavage completely depend on their autoproteolysis activities, because precursor cleavage is essentially a self-activation process.


To determine whether LmbA2991 shares the capacity for Ntn exchange with Ant6 (for example, between Thr and Ser/Cys) for the conversion of pyrroline 2 or 3 to diene 5, we first co-expressed the large and small subunits of LmbA2991 as two separated proteins and reconstituted the heterodimer through their interaction in E. coli (Fig. 4c and Supplementary Figs 4 and 11). This strategy by-passes the protein self-activation process because the LmbA2991-T421S/C precursor is resistant to autoproteolysis and cannot be converted to a functional heterodimer form; thus, we could evaluate the Ntn-hydrolase activities independently. The resulting LmbA2991 heterodimer formed an active complex in solution that exhibited hydrolytic activity towards 2 or 3 comparable to that of the catalyst prepared by LmbA2991 self-activation (Fig. 5d and Supplementary Fig. 10). We next evaluated the necessity/variability of D420 at the C-terminus of the large subunit, and in particular, T1 at the N-terminus of the small subunit; these two residues correspond to highly conserved D420 and T421 residues in the full-length LmbA2991 precursor. Subjecting the large subunit to D420A mutation had no effect on the conversion of 2 or 3 (Fig. 5d and Supplementary Fig. 10), thereby excluding D420 from playing a role in the C–C bond cleavage process; in contrast, subjecting the small unit to T1A, T1C or T1S mutation completely eliminated Ntn-hydrolase activity for the production of 5 (Fig. 5d). Clearly, unlike Ant6 catalysis, the role of the key residue Thr in LmbA2991 catalysis as an internal nucleophile for precursor splitting or as a newly released Ntn for carbon chain hydrolysis is absolutely irreplaceable.






Source link


Comments

Popular posts from this blog

Alcohol, Aging, and Curing Cancer