GyrI-like proteins catalyze cyclopropanoid hydrolysis to confer cellular protection

GyrI-like proteins are widely distributed in prokaryotes and eukaryotes, and recognized as small-molecule binding proteins. Here, we identify a subfamily of these proteins as cyclopropanoid cyclopropyl hydrolases (CCHs) that can catalyze the hydrolysis of the potent DNA-alkylating agents yatakemycin (YTM) and CC-1065. Co-crystallography and molecular dynamics simulation analyses reveal that these CCHs share a conserved aromatic cage for the hydrolytic activity. Subsequent cytotoxic assays confirm that CCHs are able to protect cells against YTM. Therefore, our findings suggest that the evolutionarily conserved GyrI-like proteins confer cellular protection against diverse xenobiotics via not only binding, but also catalysis.

Along these lines, does the bacterial strain containing the YtkR7 knockout display a phenotype (resulting from the build up of YTM, from which YtkR7 is presumably providing protection against)?
The authors state in the main text and show in Supplementary Figure 8 that both DMSO and low pH increase YtkR7-mediated hydrolysis of YTM. Did the authors determine if the background (uncatalyzed) reaction is also promoted under similar conditions? If the reaction does not occur in the absence of YktR7 or the other GyrI-like proteins, under any conditions, the authors should state this explicitly.
To further probe the catalytic mechanism of the GyrI-like CCHs, the authors solve the X-ray structure of a catalytically deficient variant of lin2189 with 1. The authors do not clearly show the interactions between lin2189 and 1. Also, the figures, including the supplementary information, does not include any electron density maps for the reviewer and readers to judge the quality of their model or clearly visualize the protein-ligand interactions. At the very least, the authors should show the ligand electron density as well as residue side chains and water molecules involved in binding.
Because the observed ligand-binding mode likely does not represent a catalytically relevant one, a docking approach is used to model the latter. However, there is no clear representation of the modeled ligand-binding mode or an indication of how it differs from that observed by crystallography. (The mutations introduced into the crystallized variant might be the cause of the observed, alternative binding mode.) Whereas the docked mode is described in the text by the authors, interpretation should be left to the reader via a clear visual presentation of the binding pocket, including residues making contact with the ligand. Also, the authors offer no evaluation of the predicted docking mode. They do state that "YTM can be docked well into the lin2189 pocket." (Supplementary Figure 17) This statement means nothing. Finally, do the authors observe significant structural differences between the apo-and ligand-bound lin2189 binding modes? metaMD simulations, which further confirms our identification of this substrate entrance." This statement is unclear. Moreover, the MD simulations cannot confirm such an entrance.
The authors state, "In apo-CCHs, E157 is protonated and E185 is deprotonated, which has been confirmed by pKa calculations on the MD trajectory ( Supplementary Fig. 22)." MD simulations cannot confirm pKa values. However, the apo-lin2189 structure may support such an assignment. Both E185 and E157 are somewhat buried and appears to engage with each other via an H-bonding contact.
The authors may want to consider moving Figure 4 and the discussion of the presented results to an earlier section in the paper.
Other Issues: Figure 1 legend does not explain the IC50 values shown for compounds 1, 2, 3 and 4.

Reviewer #2 (Remarks to the Author):
In this work, the Authors present a characterization of a particular subfamily of the GyrI-like proteins, and examine possible reasons leading to resistance, related to YTM cyclopropyl hydrolysis. The topic seems relevant, however, the way it is presented is too condensed and cryptic to permit the reader to get familiar with the problem and to have a clear picture of the gap filled by this work.
To probe the catalytic mechanism of CCHs, the Authors carried out both experimental and computational work on a member of CCH subfamily. In order to get a crystal, they made two mutations impairing the catalytic activity. In this way, they are no longer aiming at probing the catalytic mechanism, but just the binding one. However, I would say these mutations may also affect the biding properties of the protein, since they, for instance, affect the local electrostatic potential in a quite dramatic way. The Authors should have expected that these mutations very likely affect binding and should comment on this.
Then, in order to increase the sampling of the binding event the Authors used an enhanced sampling technique, namely metadynamics. Then, by using Principal (not principle) Component analysis, they observe a high fluctuation of the loop B and C distance. However, this quantity is one of the two collective variables accelerated by the metadynamics method, so how reliable can be this observation? Also the subsequent observations are flawed by the same issue. The correct usage of metadynamics would be to wait for convergence, which on a 2D space could be not straightforward, and to derive considerations on the reconstructed free energy surface, rather than commenting on individual events observed in a biased trajectory and ascribing them to a possible induced fit effect.
In the main text, they say that "In apo CCHs, E157 is protonated and E185 is deprotonated, which has been confirmed by pKa calculations on the MD trajectory", but in the Methods they say that "All amino acid residues of the protein were modelled according to their protonation state at neutral pH", so how do they estimate pKa of titratable residues in the MD runs? How therefore are derived the data in Supp. figure 22?
The protein was put in a water box 16A away from the protein, but this distance should be monitored during the dynamics, unless they put some restraints on translation or rotation of the system. "(NVT ensemble, 1bar)" is evidently a mistake.
"the simulation temperature was 310 K, and the sampling temperature ΔT was 298 K" please clarify.
"The convergence of our simulations was judged by using the free energy difference between states X and A at 10 ns intervals. Once the resulting data remained stable over time, the simulation was considered as converged. Each metadynamics simulation lasted 120 ns, and the results were analyzed upon convergence." Many things in this sentence should be rephrased and clarified, also because convergence is probably the most delicate aspect of metadynamics, as in many other enhanced sampling approaches. How precisely are defined X and A? how many recrossings have been observed? which is the threshold used for convergence? etc..
Overall, I should recommend a much clearer presentation of the logical flow of the work and a more careful and appropriate usage of the computational tools.

Reviewer #3 (Remarks to the Author):
GyrI-like proteins have been reported to confer antibiotic resistance through binding of small molecules. In this report, Yuan and co-workers report the catalytic function of GyrI-like proteins that hydrolyzes the cyclopropyl ring of cytotoxic cyclopropanoids. The authors initially identified this catalytic function through gene deletion study, which revealed the disappearance of two hydrolysis product peaks in the ytkR7 gene knock out mutants. Subsequent characterizations of YtkR7 homologues showed that many of the GyrI-like proteins have the same capability to hydrolyze cyclopropanoids. In addition, the authors conducted structural characterization of lin2189, which is a YtkR7 homolog that can hydrolyze multiple cyclopropanoid substrates. While the apo structure of this protein has been previously reported, the authors succeeded in acquiring the co-crystal of 1 with catalytically inactive mutant lin2189. Protein-ligand docking was used to predict the binding of 1 and 2 to lin2189 in the catalytically relevant form. In addition, metaMD simulations were performed to observe the potential movement in loops B and C for substrate entrance and exit. Finally, the authors confirmed that the cytotoxicity of the hydrolyzed product is much lower than the cyclopropanoids and the growth of E. coli in presence of 1 is improved upon expression of the cyclopropanoid cyclopropyl hydrolases. Finally, SPR analysis was performed to show that three GyrI-like proteins, SbmC, Rob and YtkR7, are capable of binding antibiotics as one of the detoxification methods. Overall, the work presented is clear and well rounded. Although the structure of the protein has been reported previously, additional computational modeling revealed additional information regarding the catalytic properties of the GyrI-like proteins. Based on the importance of the GyrI-like protein in antibiotic resistance, the reviewer judges that this work is of interest to broad audience and is suitable for publication in Nature Communications with minor revision.
1. Because this is the first example for the catalysis of GyrI proteins, it is informative if the authors provide the kinetic values (Km and kcat) of one of the identified cyclopropyl ring hydrolases. The enzymatic traits such as pH and temperature dependency should be also presented.
2. Can the authors explain how they find the key acidic residues E157 and E185 before crystallization briefly?
3. The authors report that 1694 GyrI proteins that has similar catalytic residues as lin2189. Are the genes encoding them located close to yatakemycin gene cluster? Can the authors predict the substrate for these enzymes from the point of view of the bioinformatic aspect? 4. How do the cyclopropyl ring hydrolases bind with the antibiotics except yatakemycin? Do they also bind with the substrate-binding cage? The authors should explain the difference of the binding mode between the cyclopropyl ring hydrolases and the reported GyrI protein structures. 5. Last sentence on page 5 seems to imply that substrates other than 1 and 2 were used for the in vitro and in vivo studies, but it is unclear which molecules were actually tested.
6. Did the authors observe the formation of the spontaneously oxidized product of 7? 7. Isolated compounds do not appear to be entirely pure based on the NMR spectra provided.
8. The HPLC peak intensities of the assays comparing the activities of the mutants are different for the double mutant compared to the other single mutants. Is there any reason for this difference?
The following are our point-by-point responses.
Reviewer #1 (Remarks to the Author): The authors are the first to report convincing evidence of cyclopropanoid cyclopropyl hydrolase activity by a subfamily of GyrI-like proteins. The study focuses partially on YtkR7, the gene product of a GyrI-like protein encoded within the same biosynthetic gene cluster of Yatakemycin, a cyclopropanoid antitumor antibiotic isolated initially from Streptomyces sp. TP-A0356.
Importantly, such an activity is potentially relevant to mechanisms of resistance to DNA-alkylating agents such as Yatakemycin. Moreover, this work suggests resistance related functions for single-domain GyrI-like proteins whose biological functions remain largely uncharacterized. Findings like those reported here suggest a biological importance of GyrI-like protein functions, which are wide-spread in all prokaryotes, eukaryotes, and archaea. Currently, their functions remain unknown. As such, the discovery of the described functions by GyrI-like proteins is indeed novel and noteworthy.
Authors provide convincing genetic, biochemical and additional bioinformatic-related evidence that YtkR7 is responsible for the metabolism and inactivation of 1 (YTM) and that other related GyrI-like proteins catalyze similar reactions. Importantly, the authors' expertise in the activity of Yatakemycin and related compounds, as well as other protein biosynthetic gene clusters, is well-established. The review is confident in the biochemical aspects of the work.
However, the discussions and analyses of structural aspects of the observed cyclopropanoid cyclopropyl hydrolase are unclear and, in their current state, contribute weakly to the manuscript.
Whereas the reviewer feels that major changes (and corrections) are needed (see below), a revised manuscript should be required for acceptance.

Major Concerns:
If indeed YtkR7 functions as a cyclopropanoid cyclopropyl hydrolase, the authors should that the protein shows significant rate acceleration over the background reaction, especially given the reactivity of strained cyclopropane rings.
Response: Sorry for the misunderstanding. In fact, YTM is very stable under all our reaction conditions. We also did the control assays. For example, in Figure 1d(i), the control assay contained only the reaction buffer and the substrate YTM. We have modified the corresponding legends. In Figure 1d(ii), the reason for the trace amount of 5 production is that YtkR7 was not completely inactivated by boiling.
Also, to better demonstrate YtkR7's enzymatic properties and power as a catalyst, the authors may want to show a Michaelis-Menton analysis and determine the kinetic parameters. Importantly, quantitative insights regarding catalysis might be important for determining or establishing the contributions to cellular resistance to compounds like YTM.

Response: From the HPLC analyses of the fermentation products, we can estimate that the yield of yatakemycin in the ytkR7 mutant is about one third that of the wild type.
Along these lines, does the bacterial strain containing the YtkR7 knockout display a phenotype (resulting from the build up of YTM, from which YtkR7 is presumably providing protection against)?
Response: Factually, the ytkR7 knockout strain displays a bald phenotype, that is, the spore production is very scarce. The phenotype is restored normally when we perform the complementation using the native ytkR7. However, we also observed the bald phenotype in other gene knockout mutants that could not produce YTM. So we cannot correlate the phenotype with YTM production. Response: Sorry for the misunderstanding. We usually make a stock solution of YTM in DMSO. As mentioned above, YTM is very stable under all our reaction conditions. We also did the control assays that contained only the reaction buffer and the substrate. As stated in the text "As expected, incubation of YtkR7 with 1 afforded 5, which was not observed for the control assay (Fig. 1d).", the underlined words have explained that the enzyme YtkR7 is absolutely necessary. We have also modified the corresponding legends.

The authors state in the main text and show in Supplementary
To further probe the catalytic mechanism of the GyrI-like CCHs, the authors solve the X-ray structure of a catalytically deficient variant of lin2189 with 1. The authors do not clearly show the interactions between lin2189 and 1.

Our MD simulation reveals that loop A, B and C underwent noticeable motion which
is in excellent agreement with our crystal structures as discussed above.
The authors state, "Notably, the complex crystal structure is identical to snapshots from our metaMD simulations, which further confirms our identification of this substrate entrance." This statement is unclear. Moreover, the MD simulations cannot confirm such an entrance.

Response:
We thank the reviewer's comments, and we rewrote this sentence in a more precisely way.
The authors state, "In apo-CCHs, E157 is protonated and E185 is deprotonated, which has been confirmed by pKa calculations on the MD trajectory ( Supplementary Fig. 22)." MD simulations cannot confirm pKa values. However, the apo-lin2189 structure may support such an assignment.
Both E185 and E157 are somewhat buried and appears to engage with each other via an H-bonding contact.

Response: We thank the reviewer's suggestions. MD simulation itself cannot confirm the pKa value. However, we have calculated the pKa values on 500 structures extracted from MD simulations so that it is more reliable statistically. We add this point in the Method part.
The authors may want to consider moving Figure 4 and the discussion of the presented results to an earlier section in the paper.

Response: CCHs are the first GyrI-like proteins disclosed as directly involved in antibiotic
resistance, so we think it is significant to keep the figure in main text. We have added a discussion part to discuss our results.
Other Issues: Figure 1 legend does not explain the IC50 values shown for compounds 1, 2, 3 and 4.
Response: Sorry for our carelessness. We added the information in Figure 1a.  Response: We thank the reviewer pointing this out. The distance seems to be a little bit far.
Possibly the proton comes from water instead of E157. Considering E157 and E185 are much closer to each other, proton transfer in Figure 3e is more likely from E157. We correct this detail in Figure 3.
The color differences corresponding to sequence similarity in Supplementary Figure 9 are not clear.

Response: We have included the identity and E-value for each protein in the figure legends,
so the color differences might be accessory.
Supplementary Figure 15. One interesting aspect of the lin2189 structure is how differs from known GyrI-like proteins. This topology figure presents a good opportunity to show this difference. CCHs using YTM as substrate) has no label or units.

Response: Sorry for our carelessness. The Y axis indicates the relative capacity of 5
production. We have added the information.
The legend of Supplementary Figure 22 should read, "The calculated pKa values of E157 and E185 for apo-lin2189, lin2189-YTM and lin2189-5". The standard error is indicated by an error bar.

Response: We changed this accordingly.
Reviewer #2 (Remarks to the Author): In this work, the Authors present a characterization of a particular subfamily of the GyrI-like proteins, and examine possible reasons leading to resistance, related to YTM cyclopropyl hydrolysis. The topic seems relevant, however, the way it is presented is too condensed and cryptic to permit the reader to get familiar with the problem and to have a clear picture of the gap filled by this work.
To probe the catalytic mechanism of CCHs, the Authors carried out both experimental and computational work on a member of CCH subfamily. In order to get a crystal, they made two mutations impairing the catalytic activity. In this way, they are no longer aiming at probing the catalytic mechanism, but just the binding one. However, I would say these mutations may also affect the biding properties of the protein, since they, for instance, affect the local electrostatic potential in a quite dramatic way. The Authors should have expected that these mutations very likely affect binding and should comment on this.

Response:
We thank the reviewer's comments. Indeed, we would like to have minimum influence on the protein and close to the native state. The native enzyme will hydrolyze the substrate. To obtain enzyme-substrate complex crystal structure, mutation around the binding pocket is unavoidable, which leads to some changes of the protein.
The principles that we chose the key residues are as follows: In 2004 We thus mutated lin2189 E185, and found that it lost the enzymatic activity.
Subsequently, we screened lin2189 E185L-YTM complex, and found that the electron density map of YTM is not clear enough. But from the information gained from the complex structure, we designed the E157 and E185 double mutant. Then, in order to increase the sampling of the binding event the Authors used an enhanced sampling technique, namely metadynamics. Then, by using Principal (not principle) Component analysis, they observe a high fluctuation of the loop B and C distance. However, this quantity is one of the two collective variables accelerated by the metadynamics method, so how reliable can be this observation?

Response:
We first would like to thank the review pointing this typo and we corrected it in the main text.

The PCA analysis is a widely used and well recognized method in biological study
which is featured by the following frequently cited papers. Also the subsequent observations are flawed by the same issue. The correct usage of metadynamics would be to wait for convergence, which on a 2D space could be not straightforward, and to derive considerations on the reconstructed free energy surface, rather than commenting on individual events observed in a biased trajectory and ascribing them to a possible induced fit effect.
Response: As explained above, our CVs have been well selected. We are always careful of our simulations. Our metaMD converged very well as indicated by the following plot.

However, it is probably not suitable to show this basic information in this experimental dominated story.
We agree with the reviewer that the FES is one of the most interesting parameters for metaMD. We have investigated the FES for two different free energy states in our previous

work: Yuan, S. et al. Activation of G-protein-coupled receptors correlates with the formation of a continuous internal water pathway (2014) Nat Commun, 5:4733 (citation=57)
However, in this work, we are only interested in how the structure changes upon ligand binding and product leaving, instead of the free energy surface of different states.
Thus, we only focus on the process of ligand binding/unbinding as shown in movies. We only show this plot to the review as following: In the main text, they say that "In apoCCHs, E157 is protonated and E185 is deprotonated, which has been confirmed by pKa calculations on the MD trajectory", but in the Methods they say that "All amino acid residues of the protein were modelled according to their protonation state at neutral pH", so how do they estimate pKa of titratable residues in the MD runs? How therefore are derived the data in Supp. figure  The protein was put in a water box 16A away from the protein, but this distance should be monitored during the dynamics, unless they put some restraints on translation or rotation of the system.

Response:
We don't add any restrain on the system. 16 A is an initial distance. During MD simulations, the box size will fluctuate to some extent since every atoms are moving all the time during simulations. However, the box shape is kept by the periodic boundary conditions (PBC) within the MD simulation tools. This is a standard procedure for modern MD simulation.

Response:
We corrected this accordingly in the Method part.
"the simulation temperature was 310 K, and the sampling temperature ΔT was 298 K" please clarify.

Response: In the equation 1 of the Method section, T is the temperature of the system. The
CVs sample an ensemble at a temperature T+ΔT which is higher than the system temperature T (well-tempered metaMD). The parameter ΔT can be chosen to regulate the extent of free-energy exploration: ΔT=0 corresponds to standard unbiased molecular dynamics, ΔT→∞ corresponds to standard metadynamics. We add this information to the Method section as well.
"The convergence of our simulations was judged by using the free energy difference between states X and A at 10 ns intervals. Once the resulting data remained stable over time, the simulation was considered as converged. Each metadynamics simulation lasted 120 ns, and the results were analyzed upon convergence." Many things in this sentence should be rephrased and clarified, also because convergence is probably the most delicate aspect of metadynamics, as in many other enhanced sampling approaches. How precisely are defined X and A? how many recrossings have been observed? which is the threshold used for convergence? etc..

Response:
We rephrased this sentence accordingly in the SI file. The converge of well-tempered metaMD indicated the Gaussian height. As indicated by above, our metaMD converged very well.

State X is a reference state, while A is the state after additional biases added to state X.
If the Gaussian height is close to zero between state A and X, the simulation is converged as shown by the plot mentioned above. We also added this information to the Method section.
Overall, I should recommend a much clearer presentation of the logical flow of the work and a more careful and appropriate usage of the computational tools.

Response:
We thank the reviewer's comments again, and we clarified all issues accordingly.
Reviewer #3 (Remarks to the Author): GyrI-like proteins have been reported to confer antibiotic resistance through binding of small molecules. In this report, Yuan and co-workers report the catalytic function of GyrI-like proteins that hydrolyzes the cyclopropyl ring of cytotoxic cyclopropanoids. The authors initially identified this catalytic function through gene deletion study, which revealed the disappearance of two hydrolysis product peaks in the ytkR7 gene knock out mutants. Subsequent characterizations of YtkR7 homologues showed that many of the GyrI-like proteins have the same capability to hydrolyze cyclopropanoids. In addition, the authors conducted structural characterization of lin2189, which is a YtkR7 homolog that can hydrolyze multiple cyclopropanoid substrates. While the apo structure of this protein has been previously reported, the authors succeeded in acquiring the co-crystal of 1 with catalytically inactive mutant lin2189. Protein-ligand docking was used to predict the binding of 1 and 2 to lin2189 in the catalytically relevant form. In addition, metaMD simulations were performed to observe the potential movement in loops B and C for substrate entrance and exit. Finally, the authors confirmed that the cytotoxicity of the hydrolyzed product is much lower than the cyclopropanoids and the growth of E. coli in presence of 1 is improved upon expression of the cyclopropanoidcyclopropyl hydrolases. Finally, SPR analysis was performed to show that three GyrI-like proteins, SbmC, Rob and YtkR7, are capable of binding antibiotics as one of the detoxification methods. Overall, the work presented is clear and well rounded.
Although the structure of the protein has been reported previously, additional computational modeling revealed additional information regarding the catalytic properties of the GyrI-like proteins. Based on the importance of the GyrI-like protein in antibiotic resistance, the reviewer judges that this work is of interest to broad audience and is suitable for publication in Nature Communications with minor revision.
1. Because this is the first example for the catalysis of GyrI proteins, it is informative if the authors provide the kinetic values (Km and kcat) of one of the identified cyclopropyl ring hydrolases. The enzymatic traits such as pH and temperature dependency should be also presented. 2. Can the authors explain how they find the key acidic residues E157 and E185 before crystallization briefly? We thus mutated lin2189 E185, and found that it lost the enzymatic activity.

Response
Subsequently, we screened lin2189 E185L-YTM complex, and found that the electron density map of YTM is not clear enough. But from the information gained from the complex structure, we designed the E157 and E185 double mutant.
To make it clear, we added "Previous structural and biochemical studies of BmrR revealed that the glutamate residue Glu-253 has a noticeable effect on its ligand binding and transcription activation 17 . This acidic residue is highly conserved among the GyrI-like proteins 11 , we thus constructed a lin2189 variant (E185L). Interestingly, this variant lost the enzymatic activity and allowed us to obtain a complex structure of lin2189 E185L and 1 in which the electron density for the substrate is partially visible. Detailed investigation of the electron density map implicated that several residues in the solvent accessible channel (e.g., E157) potentially interfered with substrate binding. Therefore, we designed a series of variants based on E185L. As expected, we successfully solved the complex structure of a catalytically-deficient form lin2189 E157A E185L and 1 with complete density at a resolution of 1.2 Å." to the main text.
3. The authors report that 1694 GyrI proteins that has similar catalytic residues as lin2189. Are the genes encoding them located close to yatakemycin gene cluster? Can the authors predict the substrate for these enzymes from the point of view of the bioinformatic aspect?
Response: From our enzymatic assays with the selected CCHs, we only found YtkR7, C10R6, and SHJG_8481 are, respectively, close to yatakemycin (YTM), CC-1065, and a cryptic biosynthetic gene cluster, which is homologous to that of YTM. For other tested CCHs, we did not observe any such biosynthetic gene clusters. These 1696 proteins are much more like evolutionarily conserved proteins rather than those acquired by horizontal gene transfer, and thus they should play beneficial roles to their hosts. Here, we show that they have the ability recognize YTM and CC-1065. It might be an accidental event during protein evolution. We regard these GyrI-like proteins as an intracellular scavenger for xenobiotics.
4. How do the cyclopropyl ring hydrolases bind with the antibiotics except yatakemycin? Do they also bind with the substrate-binding cage? The authors should explain the difference of the binding mode between the cyclopropyl ring hydrolases and the reported GyrI protein structures.

Response:
The GyrI-like proteins possess a duplicate βαββ configuration and appear to have been adapted for small-molecule binding. They each possess a different solvent accessible cage. Structural studies of BmrR for drug binding and multispecificity reveals that aromatic side chains lining the drug pocket form a rigid, aromatic platform for diverse drugs to dock.
Strange but comprehensible is that our tested GyrI-like have the ability to bind structurally unrelated compounds. The different set of aromatic and hydrophobic residues might determine their substrate spectrum. We have added some discussion in the text.
5. Last sentence on page 5 seems to imply that substrates other than 1 and 2 were used for the in vitro and in vivo studies, but it is unclear which molecules were actually tested.

Response: We only tested the two substrates (yatakemycin and CC-1065). From our assays,
we found that all the tested CCHs can hydrolyze yatakemycin, which is also consistent with its higher DNA alkylating efficiency. However, only five of the tested CCHs hydrolyze CC-1065. Recently, we have engineered a strain in a modified CC-1065 producing strain and isolated gilvusmycin (related to CC-1065) and its cyclopropyl hydrolyzed product, which indicates that the protein C10R6 recognize gilvusmycin. Moreover, from figure 1a, we can see that duocarmycins are similar to the left part of CC-1065, so it is predictable that duocarmycins are proper substrates of CCHs. So we will see at least this family of compounds can be as the substrates.
6. Did the authors observe the formation of the spontaneously oxidized product of 7?