Discovery of extended product structural space of the fungal dioxygenase AsqJ

The fungal dioxygenase AsqJ catalyses the conversion of benzo[1,4]diazepine-2,5-diones into quinolone antibiotics. A second, alternative reaction pathway leads to a different biomedically important product class, the quinazolinones. Within this work, we explore the catalytic promiscuity of AsqJ by screening its activity across a broad range of functionalized substrates made accessible by solid-/liquid-phase peptide synthetic routes. These systematic investigations map the substrate tolerance of AsqJ within its two established pathways, revealing significant promiscuity, especially in the quinolone pathway. Most importantly, two further reactivities leading to new AsqJ product classes are discovered, thus significantly expanding the structural space accessible by this biosynthetic enzyme. Switching AsqJ product selectivity is achieved by subtle structural changes on the substrate, revealing a remarkable substrate-controlled product selectivity in enzyme catalysis. Our work paves the way for the biocatalytic synthesis of diverse biomedically important heterocyclic structural frameworks.

The manuscript (Manuscript ID: NCOMMS-22-50046) by Tobias A. M. Gulder, et al. entitled "Discovery of extended product structural space of the fungal dioxygenase AsqJ" explores the expanded substrate promiscuity of the fungal dioxygenase AsqJ towards a broad variety of substrates. Efficient synthesis routes based on solid and liquid phase peptide synthesis (SPPS/LPPS) or combinations thereof are developed. These routes facilitate the rapid synthesis of a broad range of functionalized substrates, for example with different substituents at the aromatic core structure, C3-epimerization, N-alkylation pattern, exchange of heteroatoms (nitrogens replaced by oxygens), as well as tricyclic substrates. This in turn enables for the first time an in-depth screening of the established natural quinolone pathway, where a broad tolerance towards different substituents at the core aromatic portion is shown, enabling the use of AsqJ for the efficient biocatalytic synthesis of quinolones and their epoxide precursors in future applications. Tricyclic precursors are transformed into the corresponding tricyclic quinazolinones following the recently described unnatural reactivity of AsqJ. This finding is a fascinating example of a fungal enzyme that produces a compound closely related to a plant natural product, namely vasicinone, thus from a different kingdom of life. It is even more captivating that this outcome is triggered by the substrate structure. Most importantly, the presented investigations reveal that AsqJ additionally gives access to yet unknown heterocyclic product classes of high biomedical importance. These are quinazolindiones (formed by oxidative cleavage of the aliphatic side chain of the substrate following a new mechanism described here for the first time) as well as hydroxylated benzodiazepindiones, which show high similarity to the blockbuster pharmaceutical lorazepam.
Taken together, the paper contains a large set of well-planned and executed experiments that reveal valuable biocatalytic applications of AsqJ and lead to the discovery of two novel reaction pathways. All claims are well substantiated by the reported analytical data. The discoveries add substantial knowledge on the substrate promiscuity of AsqJ and its substrate-directed product outcome, which lays the foundation for the development of biocatalytic synthetic access to a broad range of biomedically important heterocyclic small molecules using AsqJ. Therefore, this work will be interesting for a broad range of the versatile readership of Nature Communications.
The following minor points could be addressed to make the paper even more accessible: 1. To facilitate a quick and easy digest of the large number of biocatalytic results presented in the manuscript, it would be advantageous to generate a table that graphically captures the product outcome for each substrate in a single picture. This could be added to the ESI.
2. Similarly, such a tabulated overview on overall yield of the synthesized substrates would be a nice addition to the ESI.
3. One of the new reactions described in the paper leads to C-hydroxylation of substrate 48n, thereby generating a stereogenic center at C3 in the product. It would be interesting to know if this is a stereoselective process. This reaction could be scaled up and the formed product thoroughly analyzed, for example by product separation using HPLC on chiral material. The results of these investigations would also help further substantiate the mechanistic proposal via radical/ionic intermediates. 4. For the transformation of the proline-derived substrate 53, it is hypothesized the more polar sideproduct (in addition to tricyclic quinazolinone 63) to likely be the carboxylic acid 67. Regarding the side chain excision that happens for other substrates leading to quinazolindiones, this reaction would lead to ring opening of the five-membered proline ring, setting free a bicyclic aldehyde. As the molecular masses of the proposed 67 compared to such an alternative, hypothetical aldehyde product would be identical, it would be interesting to exclude such a reaction pathway. This could for example be achieved by synthesis of the bicyclic aldehyde and comparison of the spectroscopic data of the isolated intermediate versus the synthesized compound. These investigations would either corroborate the mechanistic proposal for quinazolindione formation, or further substantiate the hypothesis currently stated in the manuscript. Sincerely,

Kenji Watanabe
Reviewer #3 (Remarks to the Author): The manuscript authored by Einsiedler and Gulder used a series of substrate analogs to explore the substrate scope and reactivity of AsqJ. Following substrate preparation and enzymatic assays, HPLC, product isolation and NMR were carried out to characterize reaction products. Through changing the substituent from phenyl, isobutyl, methyl to proton (along with N-substitution), AsqJ showed diverse reactivity. Total four outcomes were described in which two of them were described in the previous studies. Two new reactivities are associated with quinazolindione and hydroxylation. The authors claimed that some of these new reactions are important since several of them are affiliated with currently used drugs. This reviewer appreciates the authors' efforts in synthesizing substrate analogs and exploiting the substrate promiscuity toward AsqJ. On the other hand, this reviewer has some concerns and questions regarding to what are new knowledge and insight that we readers can learn from this manuscript. Reply: We fully agree with the reviewer that mechanistic insights into the different mechanisms are important. We therefore provided detailed mechanisms for the well-investigated natural AsqJ pathway to quinolones (see ESI, Figures S57 and S58) as well as for the new pathway to quinazolinones (see ESI, Figure  S59) in addition to mechanistic proposals for substrate hydroxylation (see Figure 7b, main manuscript) and side-chain cleavage to quinazolindiones (see ESI, Figure S60). Furthermore, we have experimentally evaluated the impact of C3 stereochemistry on quinolone versus quinazolin(di)one pathways (see main manuscript, Figure 4 and respective text, page 4, lines 16ff.). Within the revised work, we also provide additional investigations on the stereochemical outcome of the hydroxylation reaction from substrate 49n to 50 (see main manuscript page 5; ESI, chapter 3.3; cf. replies to comments 2.3 and 3.2). In addition, we provide experimental mechanistic insights into quinazolindione formation, thereby validating that this indeed is a separate pathway and not a result of further modification of initially formed quinazolinones (tested by assays with quinazolinone 6, see main manuscript, page 2, lines 61 ff.) and determined which molecular portions get excised exactly by conducting experiments with 13 C-labelled substrate 23 (see Figure  3b and text main manuscript, page 3, line 1), which overall resulted in our mechanistic proposals for quinazolindione formation (see ESI, Figure S60). We thus believe that sufficient insights into mechanistic questions are now provided for this manuscript.
In depth analyses of the impact of catalytic residues in the enzyme active site or its proximity are not part of the current manuscript that already contains a tremendous amount of experimental work and insights. This analysis by itself will be very extensive and will thus be evaluated in future work in our laboratory, which will build on the current manuscript and will aim at changing AsqJ selectivities by strategic active-site mutations.

1
To facilitate a quick and easy digest of the large number of biocatalytic results presented in the manuscript, it would be advantageous to generate a table that graphically captures the product outcome for each substrate in a single picture. This could be added to the ESI.

Reply:
We thank the reviewer for this excellent idea. We added the corresponding table to the ESI at the end of chapter 3.2 (Table S2).

2
Similarly, such a tabulated overview on overall yield of the synthesized substrates would be a nice addition to the ESI.
Reply: Again, thank you very much for this suggestion that will make our results even more accessible. We added the corresponding table to the ESI at the end of chapter 2.2 (Table S1).

3
One of the new reactions described in the paper leads to C-hydroxylation of substrate 49n, thereby generating a stereogenic center at C3 in the product. It would be interesting to know if this is a stereoselective process. This reaction could be scaled up and the formed product thoroughly analyzed, for example by product separation using HPLC on chiral material. The results of these investigations would also help further substantiate the mechanistic proposal via radical/ionic intermediates.

Reply:
We thank the reviewer for this excellent suggestion. We now added these exact experiments as suggested and investigated the corresponding AsqJ product 50 (from 49n) by HPLC on a chiral phase (see ESI, chapter 3.3, Figure S44; manuscript, page 5, lines 43 ff.). With this analysis, we observed a ~1:1-mixture of enantiomers of 50, which might hint at a non-enzymatic quenching mechanism. However, it also has to be taken into account that 3-hydroxylated diazepines, such as lorazepam (51), appear to have a very short epimerization half-life time, which was shown to be about 21 minutes for 51. 1 While we used freshly prepared 50 and conducted downstream chiral analysis very quickly, it cannot be excluded that stereochemical information of an initially stereoselective process is lost before being determinable.

4
For the transformation of the proline-derived substrate 54, it is hypothesized the more polar side-product (in addition to tricyclic quinazolinone 64) to likely be the carboxylic acid 74. Regarding the side chain excision that happens for other substrates leading to quinazolindiones, this reaction would lead to ring opening of the five-membered proline ring, setting free a bicyclic aldehyde. As the molecular masses of the proposed 74 compared to such an alternative, hypothetical aldehyde product would be identical, it would be interesting to exclude such a reaction pathway. This could for example be achieved by synthesis of the bicyclic aldehyde and comparison of the spectroscopic data of the isolated intermediate versus the synthesized compound. These investigations would either corroborate the mechanistic proposal for quinazolindione formation, or further substantiate the hypothesis currently stated in the manuscript.
Reply: This is an important observation by the reviewer and we were happy to address these questions, thereby testing our hypothesis. We now performed the synthesis of the suggested alternative aldehyde sideproduct 76 (see ESI, chapter 2.3.). Comparison of the NMR spectrum of 76 with that of the isolated compound of the assay with proline-derived substrate 54 (see ESI, Figure S245) shows no match between these two compounds. This corroborates our initial proposal that the isolated compound is indeed the corresponding acid 74. We added this discussion to the main manuscript on page 7, lines 8ff.

1
To emphasize the usefulness of this enzymatic reaction toward biomedical applications, the author should consider demonstrate how to apply this reaction in making any of the listed compounds (at least one or two examples), e.g. 26, 27, 50 and 64.

Reply:
We fully agree with the reviewer that demonstration of application potential is highly important. As we are particularly interested in the chemo-enzymatic synthesis of biomedically interesting natural products, we had already prepared the plant metabolites deoxyvasicinone (64) and mackinazoline (72) in the initial manuscript. We have now further extended this work by adding a chemo-enzymatic synthesis of antifungal isovasicinone (66). The developed synthetic route uses commercially available L-4-Hydroxyproline (67), which via protected building block 68 can be readily transformed into the desired AsqJ precursor 69 by the synthetic methodology established within the current paper. Transformation of 69 by AsqJ indeed smoothly delivered the desired antifungal 66, hence demonstrating that both, the substrate synthesis procedures and the discovered AsqJ reactivities, can directly be applied to the production of compounds of biomedical value. We added the respective discussion to the main manuscript, page 6, lines 31 ff. and Figure 8d; the experimental data to the ESI (2.1, 2.2: substrate synthesis of 69; 3.2, 3.3: conversion of 69 and isolation of 66).