Biochemical and structural characterization of the BioZ enzyme engaged in bacterial biotin synthesis pathway

Biotin is an essential micro-nutrient across the three domains of life. The paradigm earlier step of biotin synthesis denotes “BioC-BioH” pathway in Escherichia coli. Here we report that BioZ bypasses the canonical route to begin biotin synthesis. In addition to its origin of Rhizobiales, protein phylogeny infers that BioZ is domesticated to gain an atypical role of β-ketoacyl-ACP synthase III. Genetic and biochemical characterization demonstrates that BioZ catalyzes the condensation of glutaryl-CoA (or ACP) with malonyl-ACP to give 5’-keto-pimeloyl ACP. This intermediate proceeds via type II fatty acid synthesis (FAS II) pathway, to initiate the formation of pimeloyl-ACP, a precursor of biotin synthesis. To further explore molecular basis of BioZ activity, we determine the crystal structure of Agrobacterium tumefaciens BioZ at 1.99 Å, of which the catalytic triad and the substrate-loading tunnel are functionally defined. In particular, we localize that three residues (S84, R147, and S287) at the distant bottom of the tunnel might neutralize the charge of free C-carboxyl group of the primer glutaryl-CoA. Taken together, this study provides molecular insights into the BioZ biotin synthesis pathway.

This is an interesting and important study that puts yet another new and exciting spin on biotin biosynthesis. Namely, a "third way" of synthesis of the key C7 dicarboxylic acid pimelate precursor. Known pathways of C7 biosynthesis are the BioC/BioH pathway in E.coli and BioI/bioW in Bacillus subtilis. This new BioZ pathway looks to be present in a-proteobacteria such as Agrobacterium tumefacians (At), Rhizobium (Rh) and Brucella melitensis (Bm). They present strong evidence to support the assignment of BioZ as a novel FabH-like ketosynthase (KS) enzyme that catalyses the decarboxylative condensation between C3 malonyl-thioester and C5 glutaryl-thioester (they suggest as acyl carrier proteins (ACPs) thioesters) to give the pimeloyl-thioester. They show that this occurs on an ACP -they used the E.coli ACP for these studies. They show that of the 3 BioZs they tried (Bm, Rhiz and At BioZ) both the Bm and Rhiz complement E. coli BioC and Bioc/BioH mutant. They also determine the crystal structure of At BioZ to 1.99A res PDB code: 6KUE -to be released). It is a very interesting structure and compares well with the KS superfamily which has a conserved catalytic triad of Cys115, His255 and Asn285. They then go on to make mutants of these residues t0 confirm their roles in function and activity (C115A, H255A and N285A) in complementation, assays and growth curves. Figure 3 is very nice but I would render the catalytic triad residues in each of the four structures. They also investigate the binding of ACP to BioZ by making models using other KS and ACP complexes. They suggest that BioZ uses C3 malonyl-ACPand C5 glutaryl-ACP as substrates. Does this suggest that binds two ACPs at once? They use the recent FabB-ACP complex (PDB: 5KOF) from Burkart and Tsai (Naturer Chem Biol, 2019) to model how the BioZ:ACP complex forms. They should make sure they state that this interesting and very useful complex is a chemically-crosslinked complex between E.coli Fab and ACP. Fig. 5 needs to be re-drawn so that the two complexes have the same relative orientation -I would suggest they overlay on the two catalytic triads -please fix/centre on the conserved cysteines (FabH has C163, H333, H298). They use this model to identify possible Arg residues that are involved in ACP binding -mutational analysis of these 4 Arg residues supports this. Throughout they use Ec ACP but there is no discussion about what % conservation there between the Ec ACP and Agrobacterium tumefacians ACP? This should be discussed.
Overall -an interesting paper and important addition -the paper needs to be proof read and English corrected by a native English speaker.
Other points.
Discussion: The biotin part should be in the introduction. The discussion should describe the mechanism -does glutaryl-ACP bind first and transfer glutarate to the Cys residue? Can they capture this? Then it should bind malonyl-ACP, decarboxylate then catalyse the condensation between the C5 and the C2 to give pimeloyl-ACP. Please expand this. 3 BioZ homologs -show sequence alignment of these 3.
Is there an expt to show that BioZ does not catalyse the reaction between glutaryl-CoA and malonyl-CoA to give pimeloyl-CoA? It is referred to but no data.
Could not find Table S1 -please make sure it is available.   In this manuscript, the authors claim to have elucidated a third route to access the pimeloyl-ACP, an intermediate en route the biosynthesis of biotin. Biotin biosynthesis is fragmented across bacteria, this is abundantly clear, and this manuscript attempts to answer a question that will be of broad interest to the community. This reviewer is not competent to judge the phylogenetic work and will restrict his comments to the enzymology and structural aspects of this work, in addition to the writing, all three of which are severely lacking. The following are the major deficiencies in this work: 1. Major language editing is required. Several sentences, particularly in the results section make no sense, at all. The figure legends need to be descriptive. The reviewer is completely unable to understand what is being shown in Figure S8, much of the biochemical validation rests on that figure and it is impossible to understand what is implied there. If the authors haven't done so already, they are encouraged to recruit a professional manuscript preparation service. Do not insert figure callings repeatedly in the middle of sentences. Abbreviations are not to be defined in figure legends, pymol does not need to be credited in figure legends, too many other instances to parse out.
2. Biochemical validation of BioZ activity: the assertion that BioZ accepts two ACP bound substrates needs to be validated using careful in vitro biochemical work. The biochemical validation, at the moment, is extremely shallow. At a minimum, the authors need to be demonstrate whether or not BioZ possesses two distinct ACP binding sites or not, what is the order of substrate binding, and if the substrate binding is sequential. MS1 based assays, without MS2 fragmentation, are not conclusive. In such a rich spectra containing innumerable ions, it is easy to "find what you are looking for" without accounting for alternate hypotheses. Given the data presented in this study, this reviewer is not convinced that the biochemical role of BioZ is as claimed. Contemporary analytical procedures do not justify the use of data shown in Fig. 2a, without securing firm standards of all species that they claim to detect, as conclusive.
3. BioZ structure description in the second half of Page 7 is entirely superfluous and can be omitted. The description of the catalytic triad can be simplified, a lot, Docking the ACP into the BioZ structure cannot be achieved using manual means only, crystal structures of both interaction partners are in hand, better computational rigor needs to be applied. Mutations of the basic residues can hit a lot of off-target things, again the authors are blind to alternate hypotheses. At the very least, the use of biophysical techniques such as ITC and/or SPR is essential to map out the ACP:BioZ binding stoichiometries. Manual modeling followed by in vivo testing does not justify the hypothesis here. Again, the reviewer does not find justification that two ACPs can bind to BioZ, or, that ACP binding can be sequential given the structural data presented. This is the crux of the biochemical activity, one ACP binds, transfers the payload to the active site, and then the second ACP binds. The structural data needs to support this hypothesis. At the moment, it does not.

Reviewer #1
Overall comment: Biotin is an important micronutrient whose early biosynthetic steps are very diverse. Thus far, three different pathways have been identified for synthesis of the pimelic acid precursor, namely the BioC/BioH pathway, the BioI/BioW pathway and the BioZ pathway. BioZ has been shown to complement E. coli BioH, analyzed to be a beta-ketoacyl-ACP synthase III (KAS III), and predicted to synthesize pimeloyl-ACP (Ref. 30), but its catalytic function has not been characterized. In the current work, the authors first successfully repeated the previous BioZ complementation of BioH (and also BioC) in E. coli and did a thorough analysis of the evolutionary origin of the protein. Subsequently, AtBioH was successfully expressed and biochemical assays were carried to demonstrate that the enzyme is indeed able to condense malonyl-ACP and glutaryl-ACP to form pimeloyl-ACP, thus convincingly establishing the catalytic function of the enzyme. Next, the enzyme was crystallized and its structure was solved at 1.99 Angstrom to reveal that it is indeed a KAS III resembling the many previously solved structures of other members in the same category. Finally, site-directed mutagenesis and functional assays were used to identify the catalytic triad and the structural motif (composed of four positive arginine residues) responsible for recognizing the ACP substrate. These results provide unambiguous evidence for the proposed biological function of BioZ and establish the structural basis for this function, although neither the catalytic function nor the molecular mechanism is new.

Reply:
We appreciate referee 1 for the overall summary and positive evaluation on the significant importance of this work in the context of biotin metabolism.

Q1:
A caveat in the structural part of the work is the lack of a binding pocket responsible for recognition of the glutaryl moiety of the glutaryl-ACP substrate.
This binding pocket is where BioZ differs from all other members of KAS and is thus suggested to be identified to allow a better understanding of the catalytic mechanism. This pocket could be readily recognized or speculated in the 2 / 30 structural comparisons that had been carried in the work. The roles of the suspected amino acid residues could then be readily established by site-directed mutagenesis and the assays that were already used in characterization of the wild-type enzyme or the mutants.

Reply:
We do agree with this comment raised by Reviewer 1. We have combined molecular docking and site-directed mutagenesis to define this primer loading-tunnel (new_fig.8a-b). The binding of BioZ to glutaryl-CoA was demonstrated by our ITC experiment (new_fig.8c). Similarly, BioZ also binds to glutaryl-ACP efficiently in our ITC assays (new_fig.S11). More interesting, we localized that three crucial residues (S84, R147, and S287) at the bottom of this tunnel can neutralize the charge of free C-carboxyl group of glutaryl-CoA

Q2:
The presentation of title and abstract gives the impression that the work discovers a new, alternative pathway for the biosynthesis of pimelic acid or biotin. For example, the abstract states that 'Here we report a new mechanism that BioZ bypass the canonical route to begin biotin synthesis.' This, however, is not true. As mentioned above, the BioZ pathway was essentially established by the previous study presented in Ref. 30. What the authors have done is to define the substrates and establish the catalytic function of the enzyme and to solve its crystal structure. It is also noted that the catalytic mechanism is also not new. Thus, both the title and abstract should be modified to more accurately report the findings. Writing of the other parts of the work is clear and easy to understand.

Reviewer #2
Overall comment: This is an interesting and important study that puts yet another new and exciting spin on biotin biosynthesis. Namely, a "third way" of synthesis of the key C7 dicarboxylic acid pimelate precursor. Known pathways of C7 biosynthesis are the BioC/BioH pathway in E. coli and BioI/bioW in Bacillus subtilis. This new BioZ pathway looks to be present in a-proteobacteria such as Agrobacterium tumefacians (At), Rhizobium (Rh) and Brucella melitensis (Bm). They present strong evidence to support the assignment of BioZ as a novel FabH-like ketosynthase (KS) enzyme that catalyses the decarboxylative condensation between C3 malonyl-thioester and C5 glutaryl-thioester (they suggest as acyl carrier proteins (ACPs) thioesters) to give the pimeloyl-thioester. They show that this occurs on an ACP -they used the E. coli ACP for these studies. They show that of the 3 BioZs they tried (Bm, Rhiz and AtBioZ) both the Bm and Rhiz complement E.
coli BioC and BioC/BioH mutant. They also determine the crystal structure of AtBioZ to 1.99A res PDB code: 6KUE -to be released). It is a very interesting structure and compares well with the KS super-family which has a conserved catalytic triad of Cys115, His255 and Asn285. They then go on to make mutants of these residues to confirm their roles in function and activity (C115A, H255A and N285A) in complementation, assays and growth curves. They should make sure they state that this interesting and very useful complex is a chemically-cross-linked complex between E. coli Fab and ACP.
Reply: It is a good comment. As reviewer 2 mentioned, we do use the FabB-ACP cross-linked complex (PDB: 5KOF) reported by Milligan et al. 2 to model the possible binding of BioZ to ACP (Fig. 6). Also, we have followed referee 2 suggestion to rephrase the statement into "chemically cross-linked complex" accordingly (on P10 of original version with changes tracked). The recent structures of four KAS enzymes cross-linked with ACP (FabF-ACP, FabB-ACP, FabZ-ACP & FabA-ACP) from Burkart's research group 2,3 visualized that dimeric KAS enzyme can be bound by two ACP (one on each monomer). BioZ is evolutionarily-relevant member within the KAS family enzyme, whose structure is determined to feature with a FabH-like dimeric configuration. Thus, we predict that BioZ might follow a similar rule in binding two ACP, resembling other paradigm KAS member, such as FabB. As referee 2 recommended, we have applied ITC to assay the stoichiometry of BioZ binding to three ACP species (C3-ACP, C5-ACP, and ACP alone, in Fig. 6e, S11 and S19). These results consistently verified that BioZ binds to ACP is at the ratio of 1:1. One of ITC results is given as follows: new_Fig.6e ITC analysis of BioZ binding to C3-ACP Q3: Fig. 5 needs to be re-drawn so that the two complexes have the same relative orientation -I would suggest they overlay on the two catalytic triadsplease fix/centre on the conserved cysteine (FabB has C163, H333, H298). They use this model to identify possible Arg residues that are involved in ACP binding -mutational analysis of these 4 Arg residues supports this.
Reply: In agreement with Reviewer 2, we also believe that it might be important to have the same relative orientation when comparing the two structures (FabB/ACP & BioZ/ACP) with certain similarity. Thus, we tried to generate such images for the above two complexes. While, we encountered some problems. In current situation, the catalytic triads are buried in the inner of the binding pocket of BioZ. In catalytic triads-centering orientation, the structure of ACP covered the most part of other three subunits (another ACP and BioZ dimer). A similar scenario is seen with the cross-linked complex of    Reply: It is a constructive comment that benefits readers to easily follow it.
We have rephrased all the MS data, adding the calculated (predicted) and experimental value (new_Fig.S10-S14). In brief, as for the phosphopantetheine (Ppan)-linked pimeloyl group on ACP, the calculated mass is 484.53 in our LC-MS/MS measurement, highly close to that of its theoretical value, 484.1644 (Fig. S10c).
new_Fig.S10c LC-MS/MS determination of pimeloyl modification of ACP Q6: Fig.S10a is the crystal structure of the AtBioZ dimer -it has electron density for glutarate and glutarate is labeled in green -was glutarate bound?
There is no discussion about this -please expand! Reply: We apologized for our improper presentation causing confusion/misunderstanding to the referee 2. We have re-organized this figure to eliminate the possible misleading information (new_Fig.S16a-b). First, structural architecture of the dimeric BioZ is generated in new_Fig.S16a.
Because that we are not successful in harvesting crystal structure of BioZ and its substrate glutaryl-CoA (ACP), we performed an alternative method of molecular docking to address this question. The acyl group of glutaryl-ACP, glutarate (colored green in new_Fig.S16b) was subjected to molecular docking with BioZ enzyme. The rough model of BioZ-glutarate complex was created to show the approximate location of glutaryl moiety.
new_Fig.S16a-b Structural illustration of dimeric AtBioZ docked with glutarate a. Ribbon structure of AtBioZ in dimer b. Structure of glutarate-bound AtBioZ revealed by molecular docking

Reviewer #3
Overall comment: In this manuscript, the authors claim to have elucidated a third route to access the pimeloyl-ACP, an intermediate for the biosynthesis of biotin. Biotin biosynthesis is fragmented across bacteria, this is abundantly clear, and this manuscript attempts to answer a question that will be of broad interest to the community. This reviewer is not competent to judge the phylogenetic work and will restrict his comments to the enzymology and structural aspects of this work, in addition to the writing, all three of which are severely lacking. The following are the major deficiencies in this work: Reply: In addition to overall summary, the referee 3 presented constructive and critical comments on this work. We appreciated these helpful criticisms.
Also, we tried our best to address these questions/suggestions (point-to-point). Reply: First, we would like to thank referee 3 for the suggestion of language improvement. We have revised the whole manuscript in response to all three referees' suggestion. As for figure legends, some of them are rephrased in this revision (seen in the original version with changes tracked). As you suggested, we have reorganized new MS/MS data in which the calculated value vs the theoretical value have been introduced (new_Fig.S10 and S12-S14). In fact, the referee 2 also give a similar suggestion on this figure. Respectfully, we presented an improved Fig. S12 as an example. As for the BioZ reaction product keto-pimeloyl ACP, the he calculated mass of Ppan-linked keto-pimeloyl moiety is 498.13, which is almost identical to its theoretical value of 498.143 (new_Fig. S12).
new_Fig. S12 LC MS/MS determination of keto-pimeloyl ACP, the BioZ reaction product In addition, the web link of PyMol is removed. In fact, the reason we define abbreviations in figure legends is aimed to benefit readers to clearly understand them. In this revision, we have deleted unnecessary and/or redundant the explanation of abbreviations.
Q2: Biochemical validation of BioZ activity: the assertion that BioZ accepts two ACP-bound substrates needs to be validated using careful in vitro biochemical work. The biochemical validation, at the moment, is extremely shallow. At a minimum, the authors need to demonstrate whether or not BioZ possesses two distinct ACP binding sites or not, what is the order of substrate binding, and if the substrate binding is sequential. MS1 based assays, without MS2 fragmentation, are not conclusive. In such a rich spectra-containing inumerable ions, it is easy to "find what you are looking for" without accounting for alternate hypotheses. Given the data presented in this study, this reviewer is not convinced that the biochemical role of BioZ is as claimed. Contemporary analytical procedures do not justify the use of data shown in Fig. 2a, without securing firm standards of all species that they claim to detect, as conclusive.
Reply: As for these comments, they are indeed constructive. Whereas, it seems likely that Reviewer 3 misunderstand the key points of BioZ action.
Probably, this is due to the confused presentation in last version. Here, we would like to clearly explain it as follows: i) The results of gel filtration and EGS-based chemical cross-linking have confirmed that BioZ is a dimeric protein (new_Fig.S5b&d). This is verified by its x-ray crystal structure. Also, it is generally consistent with all the other paradigm member of KAS enzymes, such as FabH and FabB. iii) Bioinformatics analyses indicated that BioZ is an evolutionarily-relevant member within the KAS family enzyme, whose structure is determined to feature with a FabH-like dimeric configuration. iv) Though we are not accessible to the complex structure of BioZ chemically cross-linked with ACP, we alternatively exploited ITC technique to determine the interplay between BioZ and acyl-ACP (acyl-CoA or ACP alone) (new_Fig.6e, 8c and S11). The stoichiometry of BioZ to acyl-ACP (CoA) is consistently found to be around 1:1. v) We have provided a modified version of MS/MS for all the four C7-ACP intermediates/products, namely 5-keto-pimeloyl-ACP, 5-hydroxyl-pimeloyl-ACP, enoyl-pimeloyl-ACP, and pimeloyl-ACP (new_Fig.3b, S10, and S12-S14). As Reply: As for this point, our understanding might be different from that of Reviewer 3. This is because that BioZ structure is previously-unknown, and deserves to be described appropriately, rather than "omitted". In the revision, we have compromised to shorten this part accordingly.

Q4:
The description of the catalytic triad can be simplified, a lot.

Reply:
Although the catalytic triad is very important, we have simplified it to satisfy the requirement of Referee 3.

Q5:
Docking the ACP into the BioZ structure cannot be achieved using manual means only, crystal structures of both interaction partners are in hand, better computational rigor needs to be applied.

Reply:
As recommended by Reviewer 3, we have tried to model BioZ-ACP complex structure using the protein-protein docking tool ClusPro developed by Kozakov et al 8 . We analyzed top 10 candidate models from docking results (snapshot of molecular docking-1). The location of ACP protein was closed to the binding pocket of BioZ in 8 of 10 models. However, the Ser36 residue (colored yellow) of ACP didn't orientate to the binding pocket. We speculated that these results were largely relevant to the N-terminal flexible tail of ACP. As in these models, the tail inserted into the binding pocket. In the other 2 models, ACP protein located in the dimeric interface of BioZ and thus were not considered. Next, we removed this tail of ACP to optimize docking (snapshot of molecular docking-2). Only one (No. 8) of 10 candidate models we obtained exhibits a structure that supported proposed substrate-binding model (new_Fig.S19a). Notably, it returned an implication for structure-to-function relationship. These predicted arginine residues are further consolidated by site-directed mutagenesis. Similar scenarios were seen with BioJ 9,10 and BioH 5 .

Reviewer #1
Overall comment: Most concerns have been satisfactorily addressed. The conclusions are significantly strengthened with the additional experimental results.

Reply:
We appreciate referee 1 for the positive evaluation on this revision. Second, we should apologize for our inappropriate (unclear) presentation of ITC data, which might cause the referee 1's misunderstanding to some extent.
In fact, we carried out three independent ITC experiments (accessible to source data), and a representative graph is given in the manuscript. Thus, resultant stoichiometry values (N and Kd) from three independent experiments are given in an average ± SD. Because of the referee 1' concern with ITC data precision, we re-analyzed and processed the raw data with an updated version Microcal PEAQ-iTC system rather than the routine iTC200 system (last generation). The resultant data is given (new_fig.8c). Of note, the Kd value has been updated from 3.250±0.636 μΜ to 2.64±0.23μM.
In addition, we also rephrased the figure legend of Reply: It is a good comment. As the reviewer 1 suggested, we have modified it accordingly (new_fig. S13). As for the two peaks "y16+-H2O", we have double-checked the raw MS/MS data. We appreciated the referee 1's careful reading, and eliminated this error introduced by mistake in the graph editing/organization. Here, we presented a correct version (new_fig. S13).
Similarly, we also checked two additional figures (new_fig. S10 and new_fig. S14) and highlighted the peaks for mass calculation with pink arrows.