Cryo-EM structure of Helicobacter pylori urease with an inhibitor in the active site at 2.0 Å resolution

Infection of the human stomach by Helicobacter pylori remains a worldwide problem and greatly contributes to peptic ulcer disease and gastric cancer. Without active intervention approximately 50% of the world population will continue to be infected with this gastric pathogen. Current eradication, called triple therapy, entails a proton-pump inhibitor and two broadband antibiotics, however resistance to either clarithromycin or metronidazole is greater than 25% and rising. Therefore, there is an urgent need for a targeted, high-specificity eradication drug. Gastric infection by H. pylori depends on the expression of a nickel-dependent urease in the cytoplasm of the bacteria. Here, we report the 2.0 Å resolution structure of the 1.1 MDa urease in complex with an inhibitor by cryo-electron microscopy and compare it to a β-mercaptoethanol-inhibited structure at 2.5 Å resolution. The structural information is of sufficient detail to aid in the development of inhibitors with high specificity and affinity.

The paper by Hartmut Luecke and co-workers describes the obtainment of a high resolution (2.01 Å) structure of urease from Helicobacter pylori using cryo-EM. The main conclusions of the work are that cryo-EM can be used to obtain structural data for ureases from bacterial human pathogens, valuable for the development of inhibitors with high specificity and affinity. Moreover, the paper describes a comparison of two cryo-EM structures of H. pylori urease complexed with a novel inhibitor (2-{[1-(3,5-dimethylphenyl)-1H-imidazol-2-yl]sulfanyl}-N-hydroxyacetamide) or with beta-mercaptoethanol, a molecule known to be a urease inhibitors since many years. The paper should be published with minor revisions as I indicate below, to improve the manuscript impact: In general, the paper does not follow a clear outline and that should be corrected. Topics are discussed without following a clear line of thought. Please revise this aspect.
Line 32-33 While it is true that in H. pylori urease is produced in the bacterium cytoplasm (where else should it be produced?) it should be specified that it is the extra-cytoplasmic form that is critical for the establishment of a near neutral pH around the bacterial cells.
Line 75-77 It should be specified that the first enzyme to be crystallized was urease from Canavalia ensiformis (jack bean), while the first urease crystal structure belonged to the bacterium Klebsiella aerogenes.
Line 79-80 It has been widely established that the ligand bridging the two Ni(II) ions in the active site of urease is a hydroxide anion. Thus, this is not "thought". Please add the following references that support this information:  (7), 520-530.

Lines 102-104
In the main text the authors say: "We report cryo-EM maps of H. pylori urease at 2.45 Å and 2.01 Ar esolution, the highest resolution to date for H. pylori urease, of sufficient detail to aid in drug development ( Fig. 1 and Supplementary Fig. 1 and 2)". However, Fig.1 and 1SI show the Cryo-EM density map at 2.01 Å resolution of dodecameric 1.1 MDa Helicobacter pylori urease with bound SHA, while Fig. 2SI describes the data collection and processing procedure. No pictures of the cryo-EM maps of H. pylori urease at 2.45 Å resolution are shown. The authors should provide them.
Line 111 Supplementary Fig.1 shows only U-SHA. Either change the sentence or the figure.
Lines 130-131 Here the authors talk about bacterial ureases. In fact, plant ureases are dimers of trimers.

Lines 144-149
Here the authors include a part describing kinetic characterization of the inhibition in a mainly structural paragraph. This should be corrected.
Lines 158-159 Papers more recent than Pearson (2000)  Lines 160-162 The sentence "Interestingly, the flap region shows the highest RMSD backbone variations of both UreA and UreB subunits (above 1.5 A), whereas the active site residues nearby exhibit very low RMSD between U-BME and U-SHA (Fig. 4B)." hints to a surprise by the authors, while it has been already established in a large number of papers (essentially in all papers reporting urease structures) that the urease active site is rigid, while the flap covering the active site is mobile. See comment above.
Line 163-164 In SPU, actually two molecules of BME were found in the active site cavity, one directly binding the Ni ions and the other one covalently bound to the Cys322 residue (SPU numbering). The authors should specify this point and indicate whether they see a similar situation in their structure. Line 192 The authors should use the term "apo" only to indicate, if necessary, the urease enzyme devoid of the essential Ni(II) ions, and indicate as "native" the form of the enzyme devoid of bound inhibitors.
Lines 219-220 The authors should indicate that hydroxamic acids cannot be used for medical purposes. This is a good paper about determining the structure of an enzyme with a drug bound that has high clinical relevance. The structure is important and should be published at once after the problems are fixed. Fortunately, none of the problems are fundamental to the paper or its findings. They are really just a case of overenthusiastic use of the data processing programs without a deep understanding of how this might cause artefacts. They can easily be remedied with a bit of reprocessing of the data with more caution.
1. The authors have performed 'density modification' of the cryoEM map which is dangerous at best and makes it impossible to determine if they have over-interpreted their data. The FSCs in Sfig 2 show a sharp drop which indicates a severe processing defect of some sort. There is also aliasing in Sfig 2a and the density differences in Sfig 1 are not believable without seeing the mask and the surrounding density. To sort this out: a. The authors should calculate the FSC's between unfiltered half maps, and with high resolution phase randomization (both are standard outputs of Relion which they likely already have).
b. They should calculate the FSC between the model and the non-density modified map that is the output of Relion and report this in the supplement as well.
c. They should plot the particle orientation distribution and evaluate it for coverage of Fourier space. d. They should check their CTF fits for aliasing artefacts and increase the sampling/padding if required.
2. The claims about structure based drug design need to be removed since they have not actually done this in this study. What they have done is solved the structure of an enzyme with a drug bound that was found in a high throughput screen. Many other cryoEM studies have shown drugs bound to complexes and there are no new methods in this paper. This does not take away from the quality or importance of the work, the authors should just tone down the hyperbole, in particular: a. L220-221 phrase about "that opens the door...EM" is unsupported and should be deleted.
b. Delete the paragraph from L240-248, just waffle that says nothing interesting.
3. For resolution reports in the text, use 2 significant digits 4. Ramachandran plots: On line 119 the authors seem to claim that their models being free of Ramachandran outliers means their maps are better. This is not correct-they have used Ramachandran restraints in Phenix which forces the model to have only allowed angles regardless if the data supports it. The models should be refined without Ramachandran restraints and all outlying rotamers carefully compared and deposited (the ones on line 124) in the pdb if they wish to make this comparison. Otherwise just delete it completely and be sure to specify in the methods the type of restraints that were imposed during model refinement.
Other minor edits Paragraph on L138-141 should be moved to the discussion Paragraph on L212-218 is redundant with intro -delete Reviewer #1: The paper by Hartmut Luecke and co-workers describes the obtainment of a high resolution (2.01 Å) structure of urease from Helicobacter pylori using cryo-EM. The main conclusions of the work are that cryo-EM can be used to obtain structural data for ureases from bacterial human pathogens, valuable for the development of inhibitors with high specificity and affinity. Moreover, the paper describes a comparison of two cryo-EM structures of H. pylori urease complexed with a novel inhibitor (2-{[1-(3,5-dimethylphenyl)-1H-imidazol-2-yl]sulfanyl}-Nhydroxyacetamide) or with beta-mercaptoethanol, a molecule known to be a urease inhibitors since many years. The paper should be published with minor revisions as I indicate below, to improve the manuscript impact: In general, the paper does not follow a clear outline and that should be corrected. Topics are discussed without following a clear line of thought. Please revise this aspect.

Line 32-33
While it is true that in H. pylori urease is produced in the bacterium cytoplasm (where else should it be produced?) it should be specified that it is the extra-cytoplasmic form that is critical for the establishment of a near neutral pH around the bacterial cells. We thank the reviewer for the comment, however the mention of "urease in the cytoplasm" in the abstract is not meant to state where urease is produced (indeed by ribosomes in the cytoplasm), but where the bulk of the acid-sensitive urease is located and where substrate (urea) arrives through the plasma membrane urea channel, UreI. We are aware of an unresolved controversy about the role of extracellular urease, but this is not the focus of this paper.
Line 75-77 It should be specified that the first enzyme to be crystallized was urease from Canavalia ensiformis (jack bean), while the first urease crystal structure belonged to the bacterium Klebsiella aerogenes.
We have now clarified this in the main text. We have altered "is thought" to "has been reported" and added the citation where the hydroxide anion mechanism is described: Benini, S.; Rypniewski, W. R.; Wilson, K. S.;Miletti, S.;Ciurli, S.;Mangani, S. Structure 1999, 7 (2), 205-216."A new proposal for urease mechanism based on the crystal structures of the native and inhibited enzyme from Bacillus pasteurii: why urea hydrolysis costs two nickels".

Lines 102-104
In the main text the authors say: "We report cryo-EM maps of H. pylori urease at 2.45 Å and 2.01 Å resolution, the highest resolution to date for H. pylori urease, of sufficient detail to aid in drug development ( Fig. 1 and Supplementary Fig. 1 and 2)". However, Fig.1 and 1SI show the Cryo-EM density map at 2.01 Å resolution of dodecameric 1.1 MDa Helicobacter pylori urease with bound SHA, while Fig. 2SI describes the data collection and processing procedure. No pictures of the cryo-EM maps of H. pylori urease at 2.45 Å resolution are shown. The authors should provide them. We have added cryo-EM density for U-BME to the Supplemental Information (Fig. 1). Supplementary Fig.1 shows only U-SHA. Either change the sentence or the figure.

Line 111
We have changed the figure and the sentence.

Lines 130-131
Here the authors talk about bacterial ureases. In fact, plant ureases are dimers of trimers. We have added "dimers of trimers" to the sentence.

Lines 144-149 Here the authors include a part describing kinetic characterization of the inhibition in a mainly structural paragraph. This should be corrected.
Our intent in measuring and providing IC50 values for the two inhibitors that we also analyze structurally is to highlight the IC50 difference of nearly 3 orders of magnitude. We could have chosen to provide this information after the presentation of structural data, however, we feel it is preferable for the reader to know this while reading the structural section of our results.

Lines 160-162
The sentence "Interestingly, the flap region shows the highest RMSD backbone variations of both UreA and UreB subunits (above 1.5 Å), whereas the active site residues nearby exhibit very low RMSD between U-BME and U-SHA (Fig. 4B)." hints to a surprise by the authors, while it has been already established in a large number of papers (essentially in all papers reporting urease structures) that the urease active site is rigid, while the flap covering the active site is mobile. See comment above. We have removed the word "interestingly".

Line 163-164
In SPU, actually two molecules of BME were found in the active site cavity, one directly binding the Ni ions and the other one covalently bound to the Cys322 residue (SPU numbering). The authors should specify this point and indicate whether they see a similar situation in their structure.
We do not observe density for a second BME near Cys321 (H. pylori numbering) at the tip of the flap in U-BME. We have added a sentence clarifying this point.

Line 192 The authors should use the term "apo" only to indicate, if necessary, the urease enzyme devoid of the essential Ni(II) ions, and indicate as "native" the form of the enzyme devoid of bound inhibitors.
This is an excellent point and we have changed all references to "apo" (which would imply the absence of cofactors such as nickel) to "native".

Lines 219-220
The authors should indicate that hydroxamic acids cannot be used for medical purposes. While there are known issues with hydroxamic acids as therapeutics (ie Graham F. Smith, in Progress in Medicinal Chemistry, 2011), at least one hydroxamic acid has been approved by the U.S. This discussion is specifically about substrate access to the active site in the spherical dodecameric arrangement of the H. pylori enzyme. We have made this more clear by adding the term dodecamer early in the first sentence.

The authors have performed 'density modification' of the cryoEM map which is dangerous at best and makes it impossible to determine if they have over-interpreted their data. The FSCs in Sfig 2
show a sharp drop which indicates a severe processing defect of some sort. There is also aliasing in Sfig 2a and the density differences in Sfig 1 are not believable without seeing the mask and the surrounding density. To sort this out: a. The authors should calculate the FSC's between unfiltered half maps, and with high resolution phase randomization (both are standard outputs of Relion which they likely already have). We thank the reviewer for the suggestions and comments. The recently published procedure of EM map density modification (Terwilliger et al., 2020) was not essential for our analysis as we obtained only modest improvements of nominal resolution and map quality. We have indeed analyzed the map-to-map FSCs between unfiltered half maps (generated by Relion with high resolution phase randomization before density modification) and this results in resolution estimates of 2.09 Å for U-SHA and of 2.55 Å for U-BME, respectively, as stated in the manuscript. The corresponding FCSs do not show a sharp drop which means that masking effects are not present. We have now added these FSC plots to Supplementary Fig. 2. It is only after the density modification (as described in the methods paper by Terwilliger et al.) that we get modest resolution estimation improvements of 0.08 Å and a 0.1 Å for U-SHA and U-BME, respectively. The sharp drop is only observed in the density modified FSC curves and corresponds to a resolution of twice the pixel size, which naturally is beyond the resolution limit we are considering.

b. They should calculate the FSC between the model and the non-density modified map that is the output of Relion and report this in the supplement as well.
We had determined the model-to-map FSC between the Relion maps that agree well with the FSC calculated from the half maps. We have added the Relion map-to-model FSCs to Supplementary Fig. 2. c. They should plot the particle orientation distribution and evaluate it for coverage of Fourier space. This plot has been added as Supplementary Fig. 3, however we would like to note that we do have tetrahedral symmetry (T), which significantly reduces unique orientational space.

d. They should check their CTF fits for aliasing artefacts and increase the sampling/padding if required.
The reviewer is correct in pointing out the CTF aliasing issue in the micrograph originally pictured in Supplementary Fig. 2. We had picked this micrograph of the lower-resolution U-BME data set because it showed the particles nicely due to a high defocus value. In general, only high-defocus micrographs which contribute little to the density of U-BME, displayed aliasing. We have re-run CTF estimation with CTFfind 4.1.10 using 2048-pixel box size for the amplitude spectrum instead of the default of 512 pixels, removing the aliasing as shown in updated Supplementary Fig 2. Incidentally, the larger box size does not change the parameters determined by CTFfind 4.1.10 by more than 0.004%. Furthermore, we do not observe any CTF aliasing for the U-SHA highresolution data set which was recorded with a smaller pixel size. Importantly, for both data sets, as described in the methods section, we used two rounds of the subsequent per-particle CTF refinement option of Relion-3 during the later stages of refinement. We have clarified this in the methods section and we have added a panel to Sup. Fig. 2 showing the gradient of per-particle defocus values within one representative micrograph. Here is the citation and an excerpt from the paper describing per-particle CTF refinement: Zivanov J, Nakane T, Forsberg BO, Kimanius D, Hagen WJ, Lindahl E, Scheres SH. New tools for automated high-resolution cryo-EM structure determination in RELION-3. Elife. 2018 Nov 9;7:e42166. doi: 10.7554/eLife.42166. PMID: 30412051; PMCID: PMC6250425.

CTF refinement
In RELION, per-micrograph CTF parameters are determined through wrappers to CTFFIND (Rohou and Grigorieff, 2015) or Gctf (Zhang, 2016). These programs fit CTFs to the Thon rings visible in the power spectra of (patches of) micrographs. In RELION-3, we have implemented a program to refine the CTF parameters, that is to re-estimate defocus and astigmatism, using a 3D reference structure. This allows CTF estimation to exploit both the phases and the amplitudes of the experimental images, instead of having to rely exclusively on their power spectra. This approach thus produces significantly more reliable estimates, due to higher signal-to-noise ratios and also because it does not require separation of the Thon rings from the background intensity of the power spectrum. This is illustrated in Figure 1. The increased stability of the CTF fitting can be leveraged to estimate independent defoci for individual particles. Similar functionality also exists in Frealign (Grigorieff, 2007) and cisTEM (Grant et al., 2018).
2. The claims about structure based drug design need to be removed since they have not actually done this in this study. What they have done is solved the structure of an enzyme with a drug bound that was found in a high throughput screen. Many other cryoEM studies have shown drugs bound to complexes and there are no new methods in this paper. This does not take away from the quality or importance of the work, the authors should just tone down the hyperbole, in particular: a. L220-221 phrase about "that opens the door...EM" is unsupported and should be deleted. We did not intend to state that we have carried out structure based drug design but that our work paves a way for doing so in the future. We have corrected the sentence so that our intention becomes clearer.
b. Delete the paragraph from L240-248, just waffle that says nothing interesting. In this paragraph, which is at the very end of the article, our aim is to provide an outlook of the challenges and developments of cryo-EM. We would prefer to keep this outlook and would like the editor to make the call if it needs to be deleted.

For resolution reports in the text, use 2 significant digits
This has been corrected.

Ramachandran plots:
On line 119 the authors seem to claim that their models being free of Ramachandran outliers means their maps are better. This is not correct-they have used Ramachandran restraints in Phenix which forces the model to have only allowed angles regardless if the data supports it. The models should be refined without Ramachandran restraints and all outlying rotamers carefully compared and deposited (the ones on line 124) in the pdb if they wish to make this comparison. Otherwise just delete it completely and be sure to specify in the methods the type of restraints that were imposed during model refinement. Ramachandran restraints will not necessarily remove all Ramachandran outliers from a model as can be seen in our re-refinement of the two 3.0 Å H. pylori urease crystal structures published in 2001 using the same Ramachandran restraints as we used for the refinement of our models with our cryo EM maps. For example, the 2001 apo crystal structure shows 7.1% Ramachandran outliers, after re-refinement with Ramachandran restraints, that percentage is reduced by less than 50% to 3.6%, We are stating now that we used Phenix Ramachandran restraints for refinements using both cryo EM as well as electron densities.

Fig 2 the data points are hard to see and interpret
We suspect the reviewer is referring to Fig. 3 of the manuscript (IC50 plots). We have replotted the data for more clarity.