On 2018 Feb 01, Andrew Willetts commented:

The poor quality of the modelling studies of 3,6-diketocamphane monooxygenase (3,6-DKCMO) presented by Isupov et alresults from a combination of a number of well characterised deficiencies (1) in relevant structural and biochemical features that were used to develop their proposals for this Type II Baeyer-Villiger monooxygenase (2) from camphor-grown Pseudomonas putida NCIMB 10007.

An important contributory factor that stymies their studies is Isupov et al’s mistaken understanding of the nature and significance of the facial diastereoselectivity of the hydride transfer that occurs between the donated FNR cofactor and the enzyme in the active site of this flavin-dependent two-component monooxygenase (fd-TCMO [3]). Both Isupov et al’s structural and modelling studies place a significant dependence on comparative data for the luciferase from Vibrio harveyi, the first fd-TCMO to be recognised and well-characterised at both the structural and functional levels (4). However, while it has been conclusively shown that 3,6-DKCMO and the two enantiocomplementary 2,5-diketocamphane monooxygenases (2,5-DKCMOs) from camphor-grown P.putida NCIMB 10007 deploy requisite FNR exclusively in (si)-facial hydride transfers (5-7), the luciferase from V.harveyi (8,9) and all other bacterial luciferases thus characterised to date (10), deploy FNR exclusively in (re)-facial hydride transfers. This fundamental dichotomy between 3,6-DKCMO and the luciferase is not taken into account by Isupov et al who incorrectly state that ‘all enzymes of the bacterial luciferase superfamily catalyse their reaction on the si side of the ring’.

Consequential errors that result from this significant misunderstanding occur principally, but not exclusively, in Sections 3.7 (Comparison with other bacterial luciferase-family proteins) and 3.8 (The reaction mechanism) of Isupov et al’s paper. Typically, if this important biochemical difference had been appreciated, then Figure 5 of their paper might have been interpreted differently. Also, because 3,6- and the isoenzymic 2,5-DKCMOs have been shown to exhibit the same extremely high (si)-facial diastereoselectivity (5-7) with respect to the hydride transfers that characterise key biochemical events in their active sites, the outcome of which is the successive formation and stabilisation of their respective Criegee intermediates (11), Isupov et al’s prediction that these enantiocomplementary enzymes will exhibit ‘a different mode of cofactor binding’ seems highly unlikely.

The established differences in facial diastereoselectivity between these particular fd-TCMOs may help to explain why the commercially available (Sigma Aldrich) flavin reductase component of the luciferase from Vibrio harveyi failed to promote any significant electrophilic biooxidation of a small number of tested organosulfides by purified preparations of 3,6-DKCMO (12). Similar low levels of product(s) were detected in both experimental and equivalent control reactions (eg, <0.01% sulfoxide and <0.002% sulfone formed after 30h incubation with methyl phenyl sulphide +/- 3,6-DKCMO). It was concluded that the negligible levels of oxidation observed were principally, if not exclusively, the result of abiotic autooxidation, and consequently this particular research initiative was abandoned in mid-1996. Also, because they were outside the remit of the PhD programme of Jean Beecher (supervisor Dr Andrew Willetts, degree awarded 1997, University of Exeter), no equivalent studies of potential nucleophilic biooxidation of ketone substrates were considered or undertaken (13,14). Consequently, Dr Beecher’s thesis is notable for the total absence of any relevant content relating to either electrophilic or nucleophilic biooxidations with a hybrid P.putida-V.harveyi multienzyme complex. The claims in Isupov et al that ‘the commercially available Vibrio harveyi flavin reductase (Sigma) was able to demonstrate activity with purified 36DKCMO oxygenating enzyme in biotransformation reactions (McGhie, 1998)’, and that ‘commercially available NADH-FMN oxidoreductase from Vibrio harveyi has successfully been used for reduction of cofactor in activity measurements (McGhie, 1998)’ are incompatible with the above pre-1998 outcomes. McGhie is an accredited co-author of Isupov et al’s paper, and McGhie (1998) is in reference to her PhD awarded by the University of Exeter (supervisor Dr Littlechild). Most significantly, there is a total absence of either supporting data, or corresponding experimental protocols, or discussion relevant to both electrophilic and nucleophilic biooxidations by a hybrid P.putida-V.harveyimultienzyme complex in McGhie’s 1998 thesis, the sole relevant entry being a single sentence on p74 which claims that the hybrid complex can support ‘lactonising activity’ (= nucleophilic biooxidation), citing the source as ‘Beecher, personal communication’, which is clearly inconsistent with Jean Beecher’s pre-1998 studies (vide infra). The incorrect claim included in McGhie’s PhD thesis and the equivalent two incorrect claims included in Isupov et al are clearly interrelated. References. 1. Willetts, A. & Kelly, D.P. (2016). Microorganisms, 4, 38: 2. Willetts, A. (1997). Trends Biotechnol., 15, 55-62: 3. Ellis, H.R. (2010). Arch. Biochem. Biophys. , 497, 1-12: 4. Campbell, Z.T., Weichsel, A., Montfort, W.R. & Baldwin, T.O. (2008). Biochem. 48, 6085-6094: 5. Grogan, G. (1995). PhD Thesis, University of Exeter: 6. Beecher, J.E., Grogan, G., Roberts, S. & Willetts, A. (1996). Biotechnol. Lett., 18, 571-576: 7. Kelly, D.R., Knowles, C.J., Mahdi, J.G., Wright, M.A., Taylor, I.N., Roberts, S., Wan, P., Grogan, G., Pedragosa-Moreau, S. & Willetts, A.(1996). Chem. Commun., 20, 2333-2334: 8. Yamazaki, S., Tsai, L. & Stadyman, T.C. (1980). J. Biol. Chem., 255, 9025-9027: 9. Yamazaki, S., Tsai, L. & Stadyman, T.C., Teshima, T., Nakaji, A. & Shiba, T. (1985). Proc. Nat. Acad. Sci. USA, 82, 1364-1366: 10. Villa, R. & Willetts, A. (1997). J. Mol. Catal. B: Enzym., 2, 193-197:11. Yachnin, B.J., Sprules, T., McEvoy, M.B., Lau, P.C.K. & Berghuis, A.M. (2012). J. Amer. Chem. Soc., 134, 7788-7795: 12. Willetts, A. & Beecher, J.E. (1995; 1996). Laboratory records, unpublished data: 13. Willetts, A. (1996). Laboratory records, unpublished data: 14. Beecher, J.E. (1997). PhD Thesis, University of Exeter.

On 2017 Jul 12, Andrew Willetts commented:

I thank Drs Littlechild and Isupov for their recent Comment, which raises some interesting new issues.

As previously (1-3),the terms ‘model/structural model’ refer exclusively to material presented in Sections 3.5 – 3.7 in Isupovet al(4), and not the experimentally determined crystalline structures of 3,6-diketocamphane monooxygenase (3,6-DKCMO) presented in Sections 3.2 and 3.3. As in (5), misconceptions will arise if that important distinction is not appreciated and respected.

Isupov et al’s study (4) of 3,6-DKCMO, a flavin-diffusible two-component monooxygenase (TCMO) from camphor-grown Pseudomonas putida NCIMB 10007, determines to high resolution (1.9 A and low R-values) the three-dimensional structures of both the native enzyme and the corresponding oxidised flavin (FMN)-bound complex, which have been deposited with the PDB (code: 4UWM). In strict contrast, however, their modelling proposals detailed in sections 3.5 - 3.7 offer no sound biochemical perspective on which amino acid residues promote the formation and stabilization of the C4a-hydroperoxyflavin intermediate essential for catalytic activity, and consequently offer no mechanistic insight on 3,6-DKCMO. The binding of FMN proposed by Isupov et al, which they report as retained by the reduced form of the flavin FNR, is based on the electron density shown in Figure 3. Significantly, it is not possible to accurately assign the orientation of the isoalloxazine ring of FMN from these crystallographic data, and consequently the locations of N5 and the heterocyclic moiety of this tricyclic ring, which both serve as key determinants of functional activity, are ill-defined. Figures 4a and 5 therefore have little experimental basis.

It is likely that a combination of structural and biochemical issues may be contributing to this striking dichotomy between high resolution and low functional definition.

Structural issues - how valid are Isupov et al’s resolved crystal structures?

As a flavin-diffusible-TCMO, FMN has no direct involvement with the sequence of biochemical events catalysed by the oxygenating moiety of 3,6-DKCMO. However, at the time the structural studies were undertaken (6), 3,6-DKCMO was mistakenly characterised as a flavoprotein using bound FMN as a coenzyme (7), which would explain why Isupov et alchose this biochemically inert form of the cofactor in developing their structural studies. The structure of the FMN complex was solved by MR with the refined structure of native apo 3,6-DKCMO, which in turn was solved by dependency on MR of a synthetic α2 dimer of the luciferase from Vibrio harveyi (6) derived from Fisheret al’s (8) 2.4A resolution of the crystal structure of the native apo α/β heterodimeric form of this bacterial enzyme (PDB code: 1luc). This dependency on outcomes from Fisher et al’s 1995 study is important in this context because it predates the recognition of the relevance of structural allostery in influencing the conformation, and hence functional activity, of ligand-regulated heteromultimeric proteins (review: 9). Thus whereas 3,6-DKCMO is homodimeric, V.harveyi luciferase is an α/β heterodimer in which the β subunit, although not catalytically active, is known to have a profound allosteric effect on the 3-D shape and hence catalytic activity (>5-fold) of the α subunit resulting from a structural allosteric pathway triggered by the binding of the fully reduced flavin cofactor (FNR) as the effector ligand. A more recent study of this luciferase by Campbell et al(10) recognised and partially characterised this cofactor-triggered allosteric effect. Specifically it highlighted the importance of a mobile loop adjacent to the active site in promoting the relevant conformational changes, a structural feature that was crystallographically disordered in Fisheret al’s study (8). Thus their 1995 crystallographic studies (8) will have been compromised because these multiple major structural effects were not taken into account, and as a consequence, any structural study reliant on the synthetic α2 dimer (6) derived therefrom will be similarly compromised.

Biochemical issues - factors relevant to the active site that could result in low functional definition

1. The respective Kd values for FMN & FNR differ by >500% (1) which calls into question Isupov et al’s active site modelling proposals (1 – 3). It is likely that any FMN binding is either random or at best very weak, as evidenced by other flavin-diffusible TCMOs (11).

2. 3,6-DKCMO and V.harveyi luciferase have experimentally confirmed opposite facial diastereoselectivity with respect to the alignment of FNR and divestment of its reducing power within the active sites of the two enzymes (12). This factor was not taken into account by Isupov et al’s modelling studies which were developed in the mistaken belief that 3,6-DKCMO like ‘all other enzymes of the bacterial luciferase superfamily catalyse their reaction on the (si)-side of the ring’.

3. The tricyclic ring structure of F420 is an unwise precedent for modelling FNR into the active site because it lacks the catalytically crucial N5 atom.

Littlechild and Isupov’s Comment leaves unresolved two other interrelated issues (3), providing only evasion and obfuscation. The claim (4) that ‘Partial sequencing of the large CAM plasmid has now identified a flavin reductase adjacent to the 3,6-DKCMO gene on the CAM plasmid’ was made on the basis of partial sequencing of a corrupted sample of plasmid plus chromosomal DNA (13). The claim is incompatible with directly relevant seminal research by Iwaki et al (14) which reported no evidence for an FR-coding open-reading frame on the CAM plasmid. The three principal authors of (14) are also cited as co-authors of (4), which was published 30 months later, yet incomprehensibly the latter includes no details of or reference to their own directly relevant prior seminal research.

A number of these considerations will also be highly relevant to any equivalent structural studies being conducted (5) on either of the two the isoenzymic forms of 2,5-diketocamphane monooxygenase from Pseudomonas putida NCIMB 10007 (14).

References: 1. Willetts, A.; Kelly, D.R. (2016). Microorganisms,4, 38, doi: 10.3390/microorganisms4040038: 2. Willetts, A. (2016). PubMed Commons, Oct 27: 3. Willetts, A. (2017). PubMed Commons, Jun 05: 4. Isupov, M.N.; Schröder, E.; Gibson, R.P.; Beecher, J.; Donadio, G.; Saneei, V.; Dcunha, S.A.; McGhie, E.J.; Sayer, C.; Davenport, C.F.; Lau, P.C.; Hasegawa, Y.; Iwaki, H.; Kadow, M.; Balke, K.; Bornscheuer, U.T.; Bourenkov, G.; Littlechild, J.A. (2015). Acta Crystallogr. D, 71, 2344; 5. Littlechild, J.A.; Isupov, M.N. (2017). PubMed Commons, Jun 08: 6. Isupov, M.N.; Lebedev, A.A. (2008). Acta Crystallogr. D, 64, 90: 7. Taylor, D.G.; Trudgill, P.W. (1986). J. Bacteriol., 165, 489: 8. Fisher, A.J.; Raushel, F.M.; Baldwin, T.O.; Rayment, I. (1995). Biochem., 34, 6581: 9. Raman, S.; Taylor, N.; Genuth, N.; Fields, S.; Church, G.M. (2014). Trends Genet., 30, 521: 10. Campbell, Z.T.; Weichsel, S.; Montfort, W.R.; Baldwin, T.D. (2009). Biochem., 48, 6085: 11. Ellis, H.R. (2010). Arch. Biochem. Biophys., 497, 1; 12. Villa, R.; Willetts, A. (1997). J. Mol. Catal.: B Enzymatic, 2, 193: 13: Littlechild, J.A.; Isupov, M.N (2017). Submission to the Editors, Acta Crystallogr. D, Mar 18: 14. Iwaki, H.; Grosse, S.; Bergeron, H.; Leisch, H.; Morley, K.; Hasegawa, Y.; Lau, P.C.K. (2013). Appl. Environ. Microbiol., 79, 3282.

On 2017 Jun 13, Andrew Willetts commented:

None

On 2017 Jun 13, Andrew Willetts commented:

None

On 2017 Jun 08, Michail N Isupov commented:

We dispute the comments made on PubMed by Willetts regarding the paper ‘The oxygenating constituent of 3,6-diketocamphane monooxygenase from the CAM plasmid of Pseudomonas putida: the first crystal structure of a type II Baeyer-Villiger monooxygenase’ Michail N Isupov, Ewald Schröder, Robert P Gibson, Jean Beecher, Giuliana Donadio, Vahid Saneei, Stephlina A Dcunha, Emma J McGhie, Christopher Sayer, Colin F Davenport, Peter C Lau, Yoshie Hasegawa, Hiroaki Iwaki, Maria Kadow, Kathleen Balke, Uwe T Bornscheuer, Gleb Bourenkov, Jennifer A Littlechild, Acta Cryst. D Biol. Cryst, 71, 2015. In both pubmed comments Willetts fails to understand that it is the fit of the model to experimental data (crystallographic R-factor and FreeR) that serves as a measure of the quality of crystallographic structures and not the sequence similarity of the molecular replacement model. The use of an artificial dimer of bacterial luciferase, which has only 16% sequence similarity to 3,6-diketocamphane monooxygenase , to obtain the lost phase information and solve the well refined structure of the 3,6-DKCMO as reported in the paper is a very impressive use of molecular replacement which has already been reported in an earlier Acta Cryst paper (Isupov and Lebedev, 2008 Acta Cryst.D 64, 90-98) and was selected for a presentation at the CCP4 annual workshop in 2007 on Molecular Replacement. The X-ray data that was collected for both structures reported in our paper extends to beyond 2 Å, and the R-factors of the refined models, which were deposited in the Protein Data Bank, were compliant with their deposition criteria (pdb ref. Rose, P. W., Prlic´, A., Bi, C., Bluhm, W. F., Christie, C. H., Dutta, S., Green, R. K., Goodsell, D. S., Westbrook, J. D., Woo, J., Young, J., Zardecki, C., Berman, H. M., Bourne, P. E. & Burley, S. K. (2015) Nucleic Acids Res. 43, D345–D356). The 3,6-DKCMO structural paper was refereed by experienced crystallographers and found acceptable for publication in this well respected structural biology journal. The X-ray data shows the FMN molecule to make a number of hydrogen bonds in the monooxygenase enzyme active site. Willetts believes that the FNR (the reduced form of FMN) binding to the enzyme will result in a completely different hydrogen bonding pattern to that observed in the reported structure of the FMN monooxygenase complex. However he does not suggest any experimental evidence for this except for the increase of the enzyme affinity for the reduced form of the cofactor. We suggest that binding of the reduced flavin molecule FNR with the isoalloxazine ring in the ‘butterfly’ conformation as shown results in retention of the hydrogen bonding observed in the structure of the FMN monooxygenase complex and in an increase of hydrophobic interactions. The latter would result in the observed large increase in affinity of enzyme for the reduced cofactor. The H-bonding between N5-H and a Ser/Thr residue has not been observed in any structure determined to date of the bacterial luciferase enzyme family. In our 3,6-DKCMO structure there appears to be no Ser/Thr amino acid residue that could take on this role within the enzyme active site. The other comments of Willetts do not directly concern the main topic of the paper which describes the structural studies on the oxygenating subunits of 3,6-DKCMO, since they address the reductase enzyme. The question which we and others had investigated regarding whether this enzyme was found on the CAM plasmid or whether it was an enzyme on the Pseudomonas genome was unanswered for a number of years. We unfortunately omitted to include a reference from Iwaki et al, 2013 on this topic. We have agreed to include this reference in our next structural paper on the related 2,5-DKCMO oxygenating subunits which will shortly be submitted.

Jennifer Littlechild and Michail Isupov

On 2017 Jun 05, Andrew Willetts commented:

The modelling studies of 3,6-diketocamphane monooxygenase(3,6-DKCMO) from camphor-grown Pseudomonas putida NCIMB 10007 presented by Isupov et al in this paper are characterised by serious weaknesses (1) which result from a combination of reasons (2) including deploying molecular replacement resolution based on another protein with only 16% sequence homology, a value significantly below the lower limit accepted by structural biologists as highly likely to generate randomness (3). However, in addition to these major flaws, this paper is most notable for the inclusion of an unsupported claim to have identified a flavin reductase (FR) gene adjacent to the 3,6-DKCMO gene on the CAM plasmid, a claim which is contra to well-founded pre-existing knowledge, and which only serves to undermine the validity of the science data base.

For over 50 years, 3,6-DKCMO was mistakenly believed to be a loosely linked multienzyme complex consisting of a flavoprotein monooxygenase supported by another subunit that generated the reduced flavin coenzyme (reduced FMN = FNR) necessary for catalytic activity, and which has been variously described as ‘electron transport oxidase’ (4), NADH;(acceptor) oxidoreductase (5), or NADH dehydrogenase (6). However, recent seminal studies reported in May 2013 by Iwaki et al (7) confirmed that 3,6-DKCMO is not a flavoprotein, but rather an FMN-dependent member of the two-component monooxygenase (TCMO) group (8). Like other similar TCMOs, 3,6-DKCMO requires a separate FR to supply readily diffusible FNR as the reduced flavin cofactor necessary to undertake successful oxygenating activities. Extensive studies of the PCR amplified isolated pure CAM plasmid DNA by Iwaki et al confirmed unequivocably that there was no orf corresponding to a FR anywhere on the plasmid. Rather, Iwaki et al’s comprehensive study identified Fred, a chromosome-coded 36 kDa homodimeric FR as the principal source of FNR for 3,6-DKCMO during late log phase growth of P.putida on camphor-based minimal medium, a conclusion confirmed by a second more recent study (1).

The claim included by Isupov et al in their November 2015 publication that ‘Partial sequencing of the large CAM plasmid has now identified a flavin reductase adjacent to the 3,6-DKCMO gene on the CAM plasmid (Littlechild and Isupov, unpublished data)’ is incompatible with Iwaki et al’s extensively researched conclusions published 30 months previously (7). Their unsupported claim defies any obvious logical explanation. This is all the more so as firstly, Isupov et al were aware as long ago as 2008-9 that their relevant CAM plasmid DNA samples were contaminated with chromosomal DNA from P.putida (9), and secondly as Profs Iwaki, Hasegawa and Lau (the senior authors of [7]) are listed as co-authors of Isupov et al’s November 2015 publication. In this latter respect it is highly significant that Isupov et al’s paper fails to include a single reference to or citation of Iwaki et al’s seminal research and its highly relevant outcomes reported 30 months previously.

Perhaps one or more of the senior authors of either or both relevant publications can provide some insight, enlightenment, and/or explanation of these bizarre contradictions and inconsistencies?

References: 1. Willetts, A. & Kelly, D.R. (2016). Microorganisms,4, 38, doi: 10.3390/microorganisms4040038: 2. Willetts, A. PMC (2016). Oct 27: 3. Rost, B. (1999). Protein Eng 12 85-94: 4. Conrad, H.E., Lieb, K. and Gunsalus, I.C. (1965). J. Biol. Chem., 240, 4029-4037: 5. Trudgill, P.W., DuBus, R. and Gunsalus, I.C. (1966). J. Biol. Chem.,241, 1194-1205; 6. Taylor, D.G and Trudgill, P.W. (1986). J. Bact., 165, 489-497: 7. Iwaki, H., Grosse, S., Bergeron, H., Leisch, H., Morley, K., Hasegawa, Y. and Lau, P.C.K. (2013). Appl. Envron. Microbiol., 79, 3282-3293; 8. Ellis, H.R. (2010). Arch. Biochem. Biophys., 497, 1-12: 9. Littlechild, J.A & Isupov, M.N (2017). Submission to the Editors, Acta Cryst. D.

On 2016 Oct 27, Andrew Willetts commented:

I submitted a Letter for consideration for publication in Acta Cryst. Section D addressing a number of significant deficiencies (both errors and omissions) that characterise Isupov et al’s open access publication, but this was rejected by the Section Editor.

To attempt to redress the balance, here I provide an outline of just some of the points raised in that Letter to allow a wider audience the opportunity to reflect on the merits or otherwise of Isupov et al’s model. Science progresses by debate and discussion, not by censorship.

Comments on modelling the native apo structure of 36DKCMO

Type II Baeyer-Villiger monooxygenases (1) such as the 3,6-diketocamphane monooxygenase (36DKCMO) induced in camphor-grown Pseudomonas putida NCIMB 10007 are not true flavoproteins, as the flavin serves as a cosubstrate rather than a coenzyme. Along with the various bacterial luciferases, they are members of the FMN-dependent two-component monooxygenases (2).

Because of the relatively low-grade nature of the collected data, it was necessary to use Molecular Replacement (MR) to achieve resolution of the native apo structure. However, the choice of a synthetic α2 dimer of the luciferase from Vibrio harveyi for structure determination is misconceived, given that the two enzymes differ significantly in key structural and functional characteristics. Thus whereas 36DKCMO is homodimeric (3), the luciferase from V. harveyi is an α/β heterodimer (4). Of particular relevance, research (5) that combined extensive crystallographic studies undertaken on the α/β luciferase complex allied to mutational changes to the key amino acid residue βTyr151, has confirmed earlier predictions (6) that the β subunit plays an important allosteric role in establishing the catalytically active form of the α subunit, a feature that the synthetic α2 dimer does not take into account. The importance of this dynamic structural relationship is emphasised by a comparison which demonstrated that the quantum efficiency of the bioluminescent reaction undertaken by the α subunit was 5+ orders of magnitude lower than that undertaken by the α/β heterodimer (7).

Comments on modelling the flavin-bound 36DKCMO by Isupov et al

Isupov et al‘s use of FMN as the ligand of choice for the development of modelling studies of the active site of 36DKCMO is doubly problematical: i) Firstly, 36DKCMO is not a flavoprotein and does not generate requisite reduced FMN (FNR) in situfrom pre-bound FMN coenzyme. Rather, studies have confirmed that it accepts FNR generated from FMN by one or more competent flavin reductases (FR) present in camphor-grownP. putida NCIMB 10007 (8,9), rapid free diffusion of FNR between the two participating enzymes being a characteristic of the FMN-dependent two-component monooxygenases (2). There is a >500-fold difference in the dissociation constant (Kd) of FNR (0.24+/- 0.02 μM), and FMN (125.3 +/- 5.1 μM) recorded for 36DKCMO which suggests that any interaction of the oxidised flavin cosubstrate with the monooxygenase will be comparatively random. ii) Secondly, FMN and FNR differ in certain key structural characteristics which are likely to significantly influence the orientation of the flavin within the 3-D structure of 36DKCMO. The tricyclic isoalloxazine ring that is the basic functional feature of all flavins is planar in the fully oxidised state, whereas it is bowed through the N5-N10 axis into the so-called ‘butterfly’ conformation in FNR. The observed extent of bowing of the reduced flavin can be as high as 35^o (10), and is an idiosynchratic characteristic of any given FNR-dependent enzyme, Although the significance of the difference in conformation between the oxidised and reduced forms of the isoalloxazine ring of flavin cofactors is acknowledged by Isupov et al, they chose an arbitrary deviation from polarity (20^o) for the reduced isoalloxazine ring of FNR within the proposed active site of the enzyme, based on the past precedent of Adf (11). However, Adf is an F420-dependent enzyme, and the 5-deazoisoalloxazine ring of F420 differs from the tricyclic ring of FNR in a number of important structural and functional features, most significantly the absence of the key heteroatom N5 (vide supra). Because no illustration of their proposed FNR model is presented by Isupov et al, nor is it stated about which axis of the tricyclic ring the 20^o conformational modification was modelled in, it is not possible to critically examine their claim that FNR ‘appears to retain the hydrogen bonding observed for the oxidised FMN’. However, very large difference in the Kd values for the two forms of the flavin cosubstrate recorded with 36DKCMO (vide infra) is not compatible with this claim.

That Isupov et al’s model is significantly flawed is confirmed by a comparative study of related enzymes. Many crystal structures of other Class C (FMN-dependent) and Class D (FAD-dependent) two-component flavin-dependent oxygenases (12) have been solved recently, and although the similarity between them is generally low, the overall pattern of structural folding is well preserved (13). Above all, the feature of H-bonding interactions between the N5-H of the relevant flavin cofactor and a hydroxyl group of Ser or Thr in the active site of the enzyme, which is crucial for subsequent C4a-hydroperoxyflavin formation and stabilisation, is conserved in every other TCMO characterised at this level (14, 15). It is significant that none of the schematic drawings of the active site of flavin-bound 36DKCMO complex presented by Isupov et al are consistent with this key functional feature of TCMOs.

In summary, Isupov et al’s model is likely to be a poor basis for gaining insight into the molecular mode of action of 36DKCMO and other Type II Baeyer-Villiger monooxygenases.

References: 1. Willetts, A. (1997). Trends Biotechnol., 15, 55-62: Ellis, H.R. (2010). Arch. Biochem. Biophys., 497, 1-12: 3. Jones, K.H., Smith, R.T, & Trudgill, P.W. (1993). J. Gen. Microbiol., 139, 797-805: 4. Hastings, J.W., Potrikus, C.J., Gupta,S.C., Kurfurst, M. & Makemson, J.C. (1985). Adv. Microb. Physiol., 26, 235-291: 5. Campbell, Z.T., Weichsel, A., Montfort, W.R. & Baldwin, T.O. (2009). Biochem., 48, 6085-6094: 6. Meighen, E.A. & Bartlet, I. (1980). J. Biol. Chem., 255, 11181-11187: 7. Waddle, J. & Baldwin, T.O. (1991). Biochem. Biophys. Res. Commun., 178, 1188-1193: 8. Willetts, A. & Kelly, D.R. (2014). Microbiology , 160 , 1784-1794: 9. Willetts, A. & Kelly, D.R. (2016). Microorganisms,4, 38, doi: 10.3390/microorganisms4040038: 10. Lennon, B.W., Williams, C.H. & Ludwig, M.L. (1999). Prot. Sci., 8, 2366-2379: 11. Aufhammer, S.W., Warkentin, E., Berk, H., Shima,S., Thauer, R.K. & Ermler, U. (2004). Structure, 12, 361-370: 12. van Berkel, W.J., Kamerbeek, N.M. & Fraaije, M. (2006). J. Biotechnol., 124, 670-689: 13. Chaiyen, P., Fraaije, M.W. & Mattevi, A. (2012). Trends Biochem. Sci., 37, 373-380: 14. Thotsaporn, K., Chenprakhon, P., Sucharitakul, J., Mattevi, A., Chaiyen, P. (2011). J. Biol. Chem., 286, 28170-28180: 15. Visitsatthawong, S., Chenorakhon, P., Chaiyen, P., Surawatanawong, P. (2015). J. Amer. Chem. Soc., 137, 9363-9374.

On 2016 Oct 27, Andrew Willetts commented:

I submitted a Letter for consideration for publication in Acta Cryst. Section D addressing a number of significant deficiencies (both errors and omissions) that characterise Isupov et al’s open access publication, but this was rejected by the Section Editor.

To attempt to redress the balance, here I provide an outline of just some of the points raised in that Letter to allow a wider audience the opportunity to reflect on the merits or otherwise of Isupov et al’s model. Science progresses by debate and discussion, not by censorship.

Comments on modelling the native apo structure of 36DKCMO

Type II Baeyer-Villiger monooxygenases (1) such as the 3,6-diketocamphane monooxygenase (36DKCMO) induced in camphor-grown Pseudomonas putida NCIMB 10007 are not true flavoproteins, as the flavin serves as a cosubstrate rather than a coenzyme. Along with the various bacterial luciferases, they are members of the FMN-dependent two-component monooxygenases (2).

Because of the relatively low-grade nature of the collected data, it was necessary to use Molecular Replacement (MR) to achieve resolution of the native apo structure. However, the choice of a synthetic α2 dimer of the luciferase from Vibrio harveyi for structure determination is misconceived, given that the two enzymes differ significantly in key structural and functional characteristics. Thus whereas 36DKCMO is homodimeric (3), the luciferase from V. harveyi is an α/β heterodimer (4). Of particular relevance, research (5) that combined extensive crystallographic studies undertaken on the α/β luciferase complex allied to mutational changes to the key amino acid residue βTyr151, has confirmed earlier predictions (6) that the β subunit plays an important allosteric role in establishing the catalytically active form of the α subunit, a feature that the synthetic α2 dimer does not take into account. The importance of this dynamic structural relationship is emphasised by a comparison which demonstrated that the quantum efficiency of the bioluminescent reaction undertaken by the α subunit was 5+ orders of magnitude lower than that undertaken by the α/β heterodimer (7).

Comments on modelling the flavin-bound 36DKCMO by Isupov et al

Isupov et al‘s use of FMN as the ligand of choice for the development of modelling studies of the active site of 36DKCMO is doubly problematical: i) Firstly, 36DKCMO is not a flavoprotein and does not generate requisite reduced FMN (FNR) in situfrom pre-bound FMN coenzyme. Rather, studies have confirmed that it accepts FNR generated from FMN by one or more competent flavin reductases (FR) present in camphor-grownP. putida NCIMB 10007 (8,9), rapid free diffusion of FNR between the two participating enzymes being a characteristic of the FMN-dependent two-component monooxygenases (2). There is a >500-fold difference in the dissociation constant (Kd) of FNR (0.24+/- 0.02 μM), and FMN (125.3 +/- 5.1 μM) recorded for 36DKCMO which suggests that any interaction of the oxidised flavin cosubstrate with the monooxygenase will be comparatively random. ii) Secondly, FMN and FNR differ in certain key structural characteristics which are likely to significantly influence the orientation of the flavin within the 3-D structure of 36DKCMO. The tricyclic isoalloxazine ring that is the basic functional feature of all flavins is planar in the fully oxidised state, whereas it is bowed through the N5-N10 axis into the so-called ‘butterfly’ conformation in FNR. The observed extent of bowing of the reduced flavin can be as high as 35^o (10), and is an idiosynchratic characteristic of any given FNR-dependent enzyme, Although the significance of the difference in conformation between the oxidised and reduced forms of the isoalloxazine ring of flavin cofactors is acknowledged by Isupov et al, they chose an arbitrary deviation from polarity (20^o) for the reduced isoalloxazine ring of FNR within the proposed active site of the enzyme, based on the past precedent of Adf (11). However, Adf is an F420-dependent enzyme, and the 5-deazoisoalloxazine ring of F420 differs from the tricyclic ring of FNR in a number of important structural and functional features, most significantly the absence of the key heteroatom N5 (vide supra). Because no illustration of their proposed FNR model is presented by Isupov et al, nor is it stated about which axis of the tricyclic ring the 20^o conformational modification was modelled in, it is not possible to critically examine their claim that FNR ‘appears to retain the hydrogen bonding observed for the oxidised FMN’. However, very large difference in the Kd values for the two forms of the flavin cosubstrate recorded with 36DKCMO (vide infra) is not compatible with this claim.

That Isupov et al’s model is significantly flawed is confirmed by a comparative study of related enzymes. Many crystal structures of other Class C (FMN-dependent) and Class D (FAD-dependent) two-component flavin-dependent oxygenases (12) have been solved recently, and although the similarity between them is generally low, the overall pattern of structural folding is well preserved (13). Above all, the feature of H-bonding interactions between the N5-H of the relevant flavin cofactor and a hydroxyl group of Ser or Thr in the active site of the enzyme, which is crucial for subsequent C4a-hydroperoxyflavin formation and stabilisation, is conserved in every other TCMO characterised at this level (14, 15). It is significant that none of the schematic drawings of the active site of flavin-bound 36DKCMO complex presented by Isupov et al are consistent with this key functional feature of TCMOs.

In summary, Isupov et al’s model is likely to be a poor basis for gaining insight into the molecular mode of action of 36DKCMO and other Type II Baeyer-Villiger monooxygenases.

References: 1. Willetts, A. (1997). Trends Biotechnol., 15, 55-62: Ellis, H.R. (2010). Arch. Biochem. Biophys., 497, 1-12: 3. Jones, K.H., Smith, R.T, & Trudgill, P.W. (1993). J. Gen. Microbiol., 139, 797-805: 4. Hastings, J.W., Potrikus, C.J., Gupta,S.C., Kurfurst, M. & Makemson, J.C. (1985). Adv. Microb. Physiol., 26, 235-291: 5. Campbell, Z.T., Weichsel, A., Montfort, W.R. & Baldwin, T.O. (2009). Biochem., 48, 6085-6094: 6. Meighen, E.A. & Bartlet, I. (1980). J. Biol. Chem., 255, 11181-11187: 7. Waddle, J. & Baldwin, T.O. (1991). Biochem. Biophys. Res. Commun., 178, 1188-1193: 8. Willetts, A. & Kelly, D.R. (2014). Microbiology , 160 , 1784-1794: 9. Willetts, A. & Kelly, D.R. (2016). Microorganisms,4, 38, doi: 10.3390/microorganisms4040038: 10. Lennon, B.W., Williams, C.H. & Ludwig, M.L. (1999). Prot. Sci., 8, 2366-2379: 11. Aufhammer, S.W., Warkentin, E., Berk, H., Shima,S., Thauer, R.K. & Ermler, U. (2004). Structure, 12, 361-370: 12. van Berkel, W.J., Kamerbeek, N.M. & Fraaije, M. (2006). J. Biotechnol., 124, 670-689: 13. Chaiyen, P., Fraaije, M.W. & Mattevi, A. (2012). Trends Biochem. Sci., 37, 373-380: 14. Thotsaporn, K., Chenprakhon, P., Sucharitakul, J., Mattevi, A., Chaiyen, P. (2011). J. Biol. Chem., 286, 28170-28180: 15. Visitsatthawong, S., Chenorakhon, P., Chaiyen, P., Surawatanawong, P. (2015). J. Amer. Chem. Soc., 137, 9363-9374.

On 2017 Jun 05, Andrew Willetts commented:

The modelling studies of 3,6-diketocamphane monooxygenase(3,6-DKCMO) from camphor-grown Pseudomonas putida NCIMB 10007 presented by Isupov et al in this paper are characterised by serious weaknesses (1) which result from a combination of reasons (2) including deploying molecular replacement resolution based on another protein with only 16% sequence homology, a value significantly below the lower limit accepted by structural biologists as highly likely to generate randomness (3). However, in addition to these major flaws, this paper is most notable for the inclusion of an unsupported claim to have identified a flavin reductase (FR) gene adjacent to the 3,6-DKCMO gene on the CAM plasmid, a claim which is contra to well-founded pre-existing knowledge, and which only serves to undermine the validity of the science data base.

For over 50 years, 3,6-DKCMO was mistakenly believed to be a loosely linked multienzyme complex consisting of a flavoprotein monooxygenase supported by another subunit that generated the reduced flavin coenzyme (reduced FMN = FNR) necessary for catalytic activity, and which has been variously described as ‘electron transport oxidase’ (4), NADH;(acceptor) oxidoreductase (5), or NADH dehydrogenase (6). However, recent seminal studies reported in May 2013 by Iwaki et al (7) confirmed that 3,6-DKCMO is not a flavoprotein, but rather an FMN-dependent member of the two-component monooxygenase (TCMO) group (8). Like other similar TCMOs, 3,6-DKCMO requires a separate FR to supply readily diffusible FNR as the reduced flavin cofactor necessary to undertake successful oxygenating activities. Extensive studies of the PCR amplified isolated pure CAM plasmid DNA by Iwaki et al confirmed unequivocably that there was no orf corresponding to a FR anywhere on the plasmid. Rather, Iwaki et al’s comprehensive study identified Fred, a chromosome-coded 36 kDa homodimeric FR as the principal source of FNR for 3,6-DKCMO during late log phase growth of P.putida on camphor-based minimal medium, a conclusion confirmed by a second more recent study (1).

The claim included by Isupov et al in their November 2015 publication that ‘Partial sequencing of the large CAM plasmid has now identified a flavin reductase adjacent to the 3,6-DKCMO gene on the CAM plasmid (Littlechild and Isupov, unpublished data)’ is incompatible with Iwaki et al’s extensively researched conclusions published 30 months previously (7). Their unsupported claim defies any obvious logical explanation. This is all the more so as firstly, Isupov et al were aware as long ago as 2008-9 that their relevant CAM plasmid DNA samples were contaminated with chromosomal DNA from P.putida (9), and secondly as Profs Iwaki, Hasegawa and Lau (the senior authors of [7]) are listed as co-authors of Isupov et al’s November 2015 publication. In this latter respect it is highly significant that Isupov et al’s paper fails to include a single reference to or citation of Iwaki et al’s seminal research and its highly relevant outcomes reported 30 months previously.

Perhaps one or more of the senior authors of either or both relevant publications can provide some insight, enlightenment, and/or explanation of these bizarre contradictions and inconsistencies?

References: 1. Willetts, A. & Kelly, D.R. (2016). Microorganisms,4, 38, doi: 10.3390/microorganisms4040038: 2. Willetts, A. PMC (2016). Oct 27: 3. Rost, B. (1999). Protein Eng 12 85-94: 4. Conrad, H.E., Lieb, K. and Gunsalus, I.C. (1965). J. Biol. Chem., 240, 4029-4037: 5. Trudgill, P.W., DuBus, R. and Gunsalus, I.C. (1966). J. Biol. Chem.,241, 1194-1205; 6. Taylor, D.G and Trudgill, P.W. (1986). J. Bact., 165, 489-497: 7. Iwaki, H., Grosse, S., Bergeron, H., Leisch, H., Morley, K., Hasegawa, Y. and Lau, P.C.K. (2013). Appl. Envron. Microbiol., 79, 3282-3293; 8. Ellis, H.R. (2010). Arch. Biochem. Biophys., 497, 1-12: 9. Littlechild, J.A & Isupov, M.N (2017). Submission to the Editors, Acta Cryst. D.

On 2017 Jun 08, Michail N Isupov commented:

We dispute the comments made on PubMed by Willetts regarding the paper ‘The oxygenating constituent of 3,6-diketocamphane monooxygenase from the CAM plasmid of Pseudomonas putida: the first crystal structure of a type II Baeyer-Villiger monooxygenase’ Michail N Isupov, Ewald Schröder, Robert P Gibson, Jean Beecher, Giuliana Donadio, Vahid Saneei, Stephlina A Dcunha, Emma J McGhie, Christopher Sayer, Colin F Davenport, Peter C Lau, Yoshie Hasegawa, Hiroaki Iwaki, Maria Kadow, Kathleen Balke, Uwe T Bornscheuer, Gleb Bourenkov, Jennifer A Littlechild, Acta Cryst. D Biol. Cryst, 71, 2015. In both pubmed comments Willetts fails to understand that it is the fit of the model to experimental data (crystallographic R-factor and FreeR) that serves as a measure of the quality of crystallographic structures and not the sequence similarity of the molecular replacement model. The use of an artificial dimer of bacterial luciferase, which has only 16% sequence similarity to 3,6-diketocamphane monooxygenase , to obtain the lost phase information and solve the well refined structure of the 3,6-DKCMO as reported in the paper is a very impressive use of molecular replacement which has already been reported in an earlier Acta Cryst paper (Isupov and Lebedev, 2008 Acta Cryst.D 64, 90-98) and was selected for a presentation at the CCP4 annual workshop in 2007 on Molecular Replacement. The X-ray data that was collected for both structures reported in our paper extends to beyond 2 Å, and the R-factors of the refined models, which were deposited in the Protein Data Bank, were compliant with their deposition criteria (pdb ref. Rose, P. W., Prlic´, A., Bi, C., Bluhm, W. F., Christie, C. H., Dutta, S., Green, R. K., Goodsell, D. S., Westbrook, J. D., Woo, J., Young, J., Zardecki, C., Berman, H. M., Bourne, P. E. & Burley, S. K. (2015) Nucleic Acids Res. 43, D345–D356). The 3,6-DKCMO structural paper was refereed by experienced crystallographers and found acceptable for publication in this well respected structural biology journal. The X-ray data shows the FMN molecule to make a number of hydrogen bonds in the monooxygenase enzyme active site. Willetts believes that the FNR (the reduced form of FMN) binding to the enzyme will result in a completely different hydrogen bonding pattern to that observed in the reported structure of the FMN monooxygenase complex. However he does not suggest any experimental evidence for this except for the increase of the enzyme affinity for the reduced form of the cofactor. We suggest that binding of the reduced flavin molecule FNR with the isoalloxazine ring in the ‘butterfly’ conformation as shown results in retention of the hydrogen bonding observed in the structure of the FMN monooxygenase complex and in an increase of hydrophobic interactions. The latter would result in the observed large increase in affinity of enzyme for the reduced cofactor. The H-bonding between N5-H and a Ser/Thr residue has not been observed in any structure determined to date of the bacterial luciferase enzyme family. In our 3,6-DKCMO structure there appears to be no Ser/Thr amino acid residue that could take on this role within the enzyme active site. The other comments of Willetts do not directly concern the main topic of the paper which describes the structural studies on the oxygenating subunits of 3,6-DKCMO, since they address the reductase enzyme. The question which we and others had investigated regarding whether this enzyme was found on the CAM plasmid or whether it was an enzyme on the Pseudomonas genome was unanswered for a number of years. We unfortunately omitted to include a reference from Iwaki et al, 2013 on this topic. We have agreed to include this reference in our next structural paper on the related 2,5-DKCMO oxygenating subunits which will shortly be submitted.

Jennifer Littlechild and Michail Isupov

On 2017 Jun 13, Andrew Willetts commented:

None

On 2017 Jun 13, Andrew Willetts commented:

None

On 2017 Jul 12, Andrew Willetts commented:

I thank Drs Littlechild and Isupov for their recent Comment, which raises some interesting new issues.

As previously (1-3),the terms ‘model/structural model’ refer exclusively to material presented in Sections 3.5 – 3.7 in Isupovet al(4), and not the experimentally determined crystalline structures of 3,6-diketocamphane monooxygenase (3,6-DKCMO) presented in Sections 3.2 and 3.3. As in (5), misconceptions will arise if that important distinction is not appreciated and respected.

Isupov et al’s study (4) of 3,6-DKCMO, a flavin-diffusible two-component monooxygenase (TCMO) from camphor-grown Pseudomonas putida NCIMB 10007, determines to high resolution (1.9 A and low R-values) the three-dimensional structures of both the native enzyme and the corresponding oxidised flavin (FMN)-bound complex, which have been deposited with the PDB (code: 4UWM). In strict contrast, however, their modelling proposals detailed in sections 3.5 - 3.7 offer no sound biochemical perspective on which amino acid residues promote the formation and stabilization of the C4a-hydroperoxyflavin intermediate essential for catalytic activity, and consequently offer no mechanistic insight on 3,6-DKCMO. The binding of FMN proposed by Isupov et al, which they report as retained by the reduced form of the flavin FNR, is based on the electron density shown in Figure 3. Significantly, it is not possible to accurately assign the orientation of the isoalloxazine ring of FMN from these crystallographic data, and consequently the locations of N5 and the heterocyclic moiety of this tricyclic ring, which both serve as key determinants of functional activity, are ill-defined. Figures 4a and 5 therefore have little experimental basis.

It is likely that a combination of structural and biochemical issues may be contributing to this striking dichotomy between high resolution and low functional definition.

Structural issues - how valid are Isupov et al’s resolved crystal structures?

As a flavin-diffusible-TCMO, FMN has no direct involvement with the sequence of biochemical events catalysed by the oxygenating moiety of 3,6-DKCMO. However, at the time the structural studies were undertaken (6), 3,6-DKCMO was mistakenly characterised as a flavoprotein using bound FMN as a coenzyme (7), which would explain why Isupov et alchose this biochemically inert form of the cofactor in developing their structural studies. The structure of the FMN complex was solved by MR with the refined structure of native apo 3,6-DKCMO, which in turn was solved by dependency on MR of a synthetic α2 dimer of the luciferase from Vibrio harveyi (6) derived from Fisheret al’s (8) 2.4A resolution of the crystal structure of the native apo α/β heterodimeric form of this bacterial enzyme (PDB code: 1luc). This dependency on outcomes from Fisher et al’s 1995 study is important in this context because it predates the recognition of the relevance of structural allostery in influencing the conformation, and hence functional activity, of ligand-regulated heteromultimeric proteins (review: 9). Thus whereas 3,6-DKCMO is homodimeric, V.harveyi luciferase is an α/β heterodimer in which the β subunit, although not catalytically active, is known to have a profound allosteric effect on the 3-D shape and hence catalytic activity (>5-fold) of the α subunit resulting from a structural allosteric pathway triggered by the binding of the fully reduced flavin cofactor (FNR) as the effector ligand. A more recent study of this luciferase by Campbell et al(10) recognised and partially characterised this cofactor-triggered allosteric effect. Specifically it highlighted the importance of a mobile loop adjacent to the active site in promoting the relevant conformational changes, a structural feature that was crystallographically disordered in Fisheret al’s study (8). Thus their 1995 crystallographic studies (8) will have been compromised because these multiple major structural effects were not taken into account, and as a consequence, any structural study reliant on the synthetic α2 dimer (6) derived therefrom will be similarly compromised.

Biochemical issues - factors relevant to the active site that could result in low functional definition

1. The respective Kd values for FMN & FNR differ by >500% (1) which calls into question Isupov et al’s active site modelling proposals (1 – 3). It is likely that any FMN binding is either random or at best very weak, as evidenced by other flavin-diffusible TCMOs (11).

2. 3,6-DKCMO and V.harveyi luciferase have experimentally confirmed opposite facial diastereoselectivity with respect to the alignment of FNR and divestment of its reducing power within the active sites of the two enzymes (12). This factor was not taken into account by Isupov et al’s modelling studies which were developed in the mistaken belief that 3,6-DKCMO like ‘all other enzymes of the bacterial luciferase superfamily catalyse their reaction on the (si)-side of the ring’.

3. The tricyclic ring structure of F420 is an unwise precedent for modelling FNR into the active site because it lacks the catalytically crucial N5 atom.

Littlechild and Isupov’s Comment leaves unresolved two other interrelated issues (3), providing only evasion and obfuscation. The claim (4) that ‘Partial sequencing of the large CAM plasmid has now identified a flavin reductase adjacent to the 3,6-DKCMO gene on the CAM plasmid’ was made on the basis of partial sequencing of a corrupted sample of plasmid plus chromosomal DNA (13). The claim is incompatible with directly relevant seminal research by Iwaki et al (14) which reported no evidence for an FR-coding open-reading frame on the CAM plasmid. The three principal authors of (14) are also cited as co-authors of (4), which was published 30 months later, yet incomprehensibly the latter includes no details of or reference to their own directly relevant prior seminal research.

A number of these considerations will also be highly relevant to any equivalent structural studies being conducted (5) on either of the two the isoenzymic forms of 2,5-diketocamphane monooxygenase from Pseudomonas putida NCIMB 10007 (14).

References: 1. Willetts, A.; Kelly, D.R. (2016). Microorganisms,4, 38, doi: 10.3390/microorganisms4040038: 2. Willetts, A. (2016). PubMed Commons, Oct 27: 3. Willetts, A. (2017). PubMed Commons, Jun 05: 4. Isupov, M.N.; Schröder, E.; Gibson, R.P.; Beecher, J.; Donadio, G.; Saneei, V.; Dcunha, S.A.; McGhie, E.J.; Sayer, C.; Davenport, C.F.; Lau, P.C.; Hasegawa, Y.; Iwaki, H.; Kadow, M.; Balke, K.; Bornscheuer, U.T.; Bourenkov, G.; Littlechild, J.A. (2015). Acta Crystallogr. D, 71, 2344; 5. Littlechild, J.A.; Isupov, M.N. (2017). PubMed Commons, Jun 08: 6. Isupov, M.N.; Lebedev, A.A. (2008). Acta Crystallogr. D, 64, 90: 7. Taylor, D.G.; Trudgill, P.W. (1986). J. Bacteriol., 165, 489: 8. Fisher, A.J.; Raushel, F.M.; Baldwin, T.O.; Rayment, I. (1995). Biochem., 34, 6581: 9. Raman, S.; Taylor, N.; Genuth, N.; Fields, S.; Church, G.M. (2014). Trends Genet., 30, 521: 10. Campbell, Z.T.; Weichsel, S.; Montfort, W.R.; Baldwin, T.D. (2009). Biochem., 48, 6085: 11. Ellis, H.R. (2010). Arch. Biochem. Biophys., 497, 1; 12. Villa, R.; Willetts, A. (1997). J. Mol. Catal.: B Enzymatic, 2, 193: 13: Littlechild, J.A.; Isupov, M.N (2017). Submission to the Editors, Acta Crystallogr. D, Mar 18: 14. Iwaki, H.; Grosse, S.; Bergeron, H.; Leisch, H.; Morley, K.; Hasegawa, Y.; Lau, P.C.K. (2013). Appl. Environ. Microbiol., 79, 3282.

On 2018 Feb 01, Andrew Willetts commented:

The poor quality of the modelling studies of 3,6-diketocamphane monooxygenase (3,6-DKCMO) presented by Isupov et alresults from a combination of a number of well characterised deficiencies (1) in relevant structural and biochemical features that were used to develop their proposals for this Type II Baeyer-Villiger monooxygenase (2) from camphor-grown Pseudomonas putida NCIMB 10007.

An important contributory factor that stymies their studies is Isupov et al’s mistaken understanding of the nature and significance of the facial diastereoselectivity of the hydride transfer that occurs between the donated FNR cofactor and the enzyme in the active site of this flavin-dependent two-component monooxygenase (fd-TCMO [3]). Both Isupov et al’s structural and modelling studies place a significant dependence on comparative data for the luciferase from Vibrio harveyi, the first fd-TCMO to be recognised and well-characterised at both the structural and functional levels (4). However, while it has been conclusively shown that 3,6-DKCMO and the two enantiocomplementary 2,5-diketocamphane monooxygenases (2,5-DKCMOs) from camphor-grown P.putida NCIMB 10007 deploy requisite FNR exclusively in (si)-facial hydride transfers (5-7), the luciferase from V.harveyi (8,9) and all other bacterial luciferases thus characterised to date (10), deploy FNR exclusively in (re)-facial hydride transfers. This fundamental dichotomy between 3,6-DKCMO and the luciferase is not taken into account by Isupov et al who incorrectly state that ‘all enzymes of the bacterial luciferase superfamily catalyse their reaction on the si side of the ring’.

Consequential errors that result from this significant misunderstanding occur principally, but not exclusively, in Sections 3.7 (Comparison with other bacterial luciferase-family proteins) and 3.8 (The reaction mechanism) of Isupov et al’s paper. Typically, if this important biochemical difference had been appreciated, then Figure 5 of their paper might have been interpreted differently. Also, because 3,6- and the isoenzymic 2,5-DKCMOs have been shown to exhibit the same extremely high (si)-facial diastereoselectivity (5-7) with respect to the hydride transfers that characterise key biochemical events in their active sites, the outcome of which is the successive formation and stabilisation of their respective Criegee intermediates (11), Isupov et al’s prediction that these enantiocomplementary enzymes will exhibit ‘a different mode of cofactor binding’ seems highly unlikely.

The established differences in facial diastereoselectivity between these particular fd-TCMOs may help to explain why the commercially available (Sigma Aldrich) flavin reductase component of the luciferase from Vibrio harveyi failed to promote any significant electrophilic biooxidation of a small number of tested organosulfides by purified preparations of 3,6-DKCMO (12). Similar low levels of product(s) were detected in both experimental and equivalent control reactions (eg, <0.01% sulfoxide and <0.002% sulfone formed after 30h incubation with methyl phenyl sulphide +/- 3,6-DKCMO). It was concluded that the negligible levels of oxidation observed were principally, if not exclusively, the result of abiotic autooxidation, and consequently this particular research initiative was abandoned in mid-1996. Also, because they were outside the remit of the PhD programme of Jean Beecher (supervisor Dr Andrew Willetts, degree awarded 1997, University of Exeter), no equivalent studies of potential nucleophilic biooxidation of ketone substrates were considered or undertaken (13,14). Consequently, Dr Beecher’s thesis is notable for the total absence of any relevant content relating to either electrophilic or nucleophilic biooxidations with a hybrid P.putida-V.harveyi multienzyme complex. The claims in Isupov et al that ‘the commercially available Vibrio harveyi flavin reductase (Sigma) was able to demonstrate activity with purified 36DKCMO oxygenating enzyme in biotransformation reactions (McGhie, 1998)’, and that ‘commercially available NADH-FMN oxidoreductase from Vibrio harveyi has successfully been used for reduction of cofactor in activity measurements (McGhie, 1998)’ are incompatible with the above pre-1998 outcomes. McGhie is an accredited co-author of Isupov et al’s paper, and McGhie (1998) is in reference to her PhD awarded by the University of Exeter (supervisor Dr Littlechild). Most significantly, there is a total absence of either supporting data, or corresponding experimental protocols, or discussion relevant to both electrophilic and nucleophilic biooxidations by a hybrid P.putida-V.harveyimultienzyme complex in McGhie’s 1998 thesis, the sole relevant entry being a single sentence on p74 which claims that the hybrid complex can support ‘lactonising activity’ (= nucleophilic biooxidation), citing the source as ‘Beecher, personal communication’, which is clearly inconsistent with Jean Beecher’s pre-1998 studies (vide infra). The incorrect claim included in McGhie’s PhD thesis and the equivalent two incorrect claims included in Isupov et al are clearly interrelated. References. 1. Willetts, A. & Kelly, D.P. (2016). Microorganisms, 4, 38: 2. Willetts, A. (1997). Trends Biotechnol., 15, 55-62: 3. Ellis, H.R. (2010). Arch. Biochem. Biophys. , 497, 1-12: 4. Campbell, Z.T., Weichsel, A., Montfort, W.R. & Baldwin, T.O. (2008). Biochem. 48, 6085-6094: 5. Grogan, G. (1995). PhD Thesis, University of Exeter: 6. Beecher, J.E., Grogan, G., Roberts, S. & Willetts, A. (1996). Biotechnol. Lett., 18, 571-576: 7. Kelly, D.R., Knowles, C.J., Mahdi, J.G., Wright, M.A., Taylor, I.N., Roberts, S., Wan, P., Grogan, G., Pedragosa-Moreau, S. & Willetts, A.(1996). Chem. Commun., 20, 2333-2334: 8. Yamazaki, S., Tsai, L. & Stadyman, T.C. (1980). J. Biol. Chem., 255, 9025-9027: 9. Yamazaki, S., Tsai, L. & Stadyman, T.C., Teshima, T., Nakaji, A. & Shiba, T. (1985). Proc. Nat. Acad. Sci. USA, 82, 1364-1366: 10. Villa, R. & Willetts, A. (1997). J. Mol. Catal. B: Enzym., 2, 193-197:11. Yachnin, B.J., Sprules, T., McEvoy, M.B., Lau, P.C.K. & Berghuis, A.M. (2012). J. Amer. Chem. Soc., 134, 7788-7795: 12. Willetts, A. & Beecher, J.E. (1995; 1996). Laboratory records, unpublished data: 13. Willetts, A. (1996). Laboratory records, unpublished data: 14. Beecher, J.E. (1997). PhD Thesis, University of Exeter.