Making the Most of 3D Electron Diffraction: Best Practices to Handle a New Tool

Round 1

Reviewer 1 Report

       The work is interesting and well for the journal. But I think before it is accepted, some comments are required to be responded.

1. Are the three key aspects also essential for inorganic compounds? The authors are required to give more presentation or discussion in their manuscript.

2. How to solve the structures of the examples in the work? Are they all from Autochem or mainly refined from original models after correct indexing? The authors are required to give more presentation in the experiment part for the small molecules and the proteins.

3. When do we have to consider the contribution of dynamical diffractions? Are there some rules or methods to support the concern. The authors are welcome to give some suggestion in the revised work.

4. I don't think it is enough to have the correct absolute structure only by the R values. The authors are required to show other experiment data like optical properties in the manuscript to support their refinement with the consideration of dynamical diffraction.

5. Some writing need to be checked. For example, I suggest the authors only use "CAP" in the title of Fig.2 to shorten the words but using CrysAlis^Pro all over the manuscript. And please add the title for the table in 3.3 section, page 8.

Author Response

The work is interesting and well for the journal. But I think before it is accepted, some comments are required to be responded.

1. Are the three key aspects also essential for inorganic compounds? The authors are required to give more presentation or discussion in their manuscript.

In section 4.3 we have added the sentence: “At the other end of the spectrum, inorganic compounds will often yield useful diffraction up to very high resolutions; in this case, further shortening the DD below typical values will lead to improved resolution and completeness which would be useful e.g. for charge-density studies or dynamical refinement.”

Regarding dynamical refinement, all advantages are as valid for inorganic compounds, as demonstrated e.g. in the recent paper by Klar et al. – we explicitly mention this in section 4.2 now.

We did not yet encounter a scenario where cryo-transfer would have been relevant for an inorganic compound.

2. How to solve the structures of the examples in the work? Are they all from or mainly refined from original models after correct indexing? The authors are required to give more presentation in the experiment part for the small molecules and the proteins.

As stated in section 2.3, the final small-molecule structures were solved ab-initio using direct methods (SHELXT) after processing in CAP. We have added “using direct methods” to make this more explicit. AutoChem was used for on-line processing, the results of which were used as starting point for offline processing in some instances, which however has no effect on the outcome, as AutoChem is set up to use the same CAP-SHELXT-SHELXL pipeline. For the protein, the structure was solved using molecular replacement and further run through a standard CCP4 pipeline (section 2.3).

3. When do we have to consider the contribution of dynamical diffractions? Are there some rules or methods to support the concern. The authors are welcome to give some suggestion in the revised work.

A full discussion of limitations of the kinematical approximation and for which samples dynamical refinement is most advantageous is beyond the scope of our paper. We explicitly refer the reader to more specialized publications, and have added another sentence on the physical foundations of why kinematical computations can work at all:

“Making use of the kinematical approximation still yields satisfactory results, as data collection off zone axes [1], mosaicity on the nano-scale [44], and inelastic scattering [45] mitigate dynamical effects [9]. However, the usefulness of using full dynamical computations for refinement has now been demonstrated for a range of samples and techniques [34,46–49].”

4. I don't think it is enough to have the correct absolute structure only by the R values. The authors show other experiment data like optical properties in the manuscript to support their refinement with the consideration of dynamical diffraction.

Assignment of absolute structure from dynamical refinement should not be based on R-values, but rather the intensity bias z-score, which directly quantifies the confidence level in the correct assignment, based on the difference between the Icalc values for wrong and correct enantiomorph. We clarify this by the modified sentence: “To obtain a more quantitative measure for the confidence level of absolute structure assignment (as originally suggested by Le Page et al. [51]), we compute the background noise-adjusted z-score for the bias of observed intensities towards those computed from the correct absolute configuration following the definition of Klar et al. [49]. Z-scores higher than 3σ were found for even the simplest computation, with values reaching up to 10.0σ for data processing in CAP and full refinement in Jana2020, as shown in Figure 2(d).:

As in our manuscript we exclusively study samples with known absolute structure (which is correctly determined, as for many more e.g. in Klar et al. 2023), we see no necessity for additional experiments.

5. Some writing need to be checked. For example, I suggest the authors only use "CAP" in the title of Fig.2 to shorten the words but using CrysAlis^Proall over the manuscript. And please add the title for the table in 3.3 section, page 8.

The Caption of Fig. 2 has been changed. The table in section 3.3 appeared erroneously in the Word file sent for review and is now deleted.

Reviewer 2 Report

This study reports systematic 3D ED methodology for structure determination of small molecules/proteins. The authors introduce three points of experimental design unique to 3D ED experiments; 1) whether the sample is hydrated under vacuum in ED measurement, 2) dynamical refinement for 3D ED, 3) optimal crystal-to-detector distance (DD). This paper is well organized overall and offers a new direction for 3D ED fields. Therefore, this paper can be accepted to the Journal, Symmetry. The followings are my short comments.

- In particular, dynamical refinement strategies are quite attractive. However, obviously there will be cases where exact structure determination is difficult using only the proposed workflow. In such cases, the existing full refinement procedure should be applied. Therefore, the reviewer recommends that the authors discuss additional points to consider regarding whether this process is adaptable or not.

- Why is there a structural difference in symmetry between the dehydrated form under vacuum (Figure 1b) and the anhydrous form at room temperature (Figure 1f)? Is there any possible way to keep hydration except for the cryo-transfer system?

- DD determines the size and intensity of the diffraction spots. So adjustment of DD plays a crucial role in diffraction experiments. In this regard, is there any cons that detection intensity can be problematic by long DD?

Author Response

This study reports systematic 3D ED methodology for structure determination of small molecules/proteins. The authors introduce three points of experimental design unique to 3D ED experiments; 1) whether the sample is hydrated under vacuum in ED measurement, 2) dynamical refinement for 3D ED, 3) optimal crystal-to-detector distance (DD). This paper is well organized overall and offers a new direction for 3D ED fields. Therefore, this paper can be accepted to the Journal, Symmetry. The followings are my short comments.

- In particular, dynamical refinement strategies are quite attractive. However, obviously there will be cases where exact structure determination is difficult using only the proposed workflow. In such cases, the existing full refinement procedure should be applied. Therefore, the reviewer recommends that the authors discuss points to consider regarding whether this process is adaptable or not.

The question in which scenarios a more sophisticated dynamical refinement (including the actual atom coordinates) is required or not is discussed in section 4.2. We have slightly revised the second paragraph of 4.2 to make the discussion more transparent. For more details, readers are referred to more specialized literature (in particular, Klar et al.).

- Why is there a structural difference in symmetry between the dehydrated form under vacuum (Figure 1b) and the anhydrous form at room temperature (Figure 1f)? Is there any possible way to keep hydration except for the cryo-transfer system?

As mentioned at the end of section 3.1, it is known from literature that the two pathways to obtain trehalose anhydrous by either dehydration or high-temperature crystallizations result in the two different polymorphs. An alternative to cryo-preservation is given by environmental sample holders, which is mentioned including pertinent references at the beginning of section 4.1. We hence believe that no further changes are required.

- DD  the size and intensity of the diffraction spots. adjustment of DD plays a crucial role in diffraction experiments. In this regard, is there any cons that detection intensity can be problematic by long DD?

In our case of a counting hybrid-pixel detector and integration using a sufficiently sophisticated processing package, no difference in integrated SNR is expected. However, this might not be the case in other setups. To clarify, we have added a sentence to section 4.3: “Also, while our measurements have been conducted on a counting detector with no background signal, an instrument fitted with a conventional CMOS or CCD detector might yield worse signal-to-noise ratios at larger DD as more background noise is collected from the larger integration regions required to fit the magnified peaks.”

Reviewer 3 Report

Dear Authority,

The manuscript entitled ‘Making the most of 3D ED: Best practices to handle a new tool’ presents the three-dimensional electron diffraction (3D ED/MicroED) carried out in Rigaku’s laboratories. Optimal 3D ED/MicroED measurement studies on submicron small-molecule and protein structures help to understand cryo-transfer for hydration, radiation dose, effective distance between crystal and detector distance, and finding structure through dynamical refinement.

I think, the paper includes important information and data which will be useful for literature and will be reference paper for future studies. It could be considered for publication after minor correction according to following comments/recommendations;

1- Abstract is not good enough, improve it with the numerical results and write it again succinctly. Abstract section should be concisely reflected the content and summarize the problem, the method, the results, and the conclusions. I would recommend to put important data and outcomes.

2- Some sentences are so long so that it is hard to follow the topic and there are some parts grammatically wrong. Please go over the manuscript again for proof reading

3- In section 3.3, there is a table without caption. Please correct this or move the table to supplementary section.

4- Whole manuscript needs to reorganized in ‘word’ program such as numbering, page design

The manuscript can be published in Symmetry after these minor corrections.

Best wishes,

Comments for author File: Comments.pdf

Author Response

The manuscript entitled ‘Making the most of 3D ED: Best practices to handle a new tool’ presents the three-dimensional electron diffraction (3D ED/MicroED) carried out in Rigaku’s laboratories. Optimal 3D ED/MicroED measurement studies on submicron small-molecule and protein structures help to understand cryo-transfer for hydration, radiation dose, effective distance between crystal and detector distance, and finding structure through dynamical refinement.

I think, the paper includes important information and data which will be useful for literature and will be reference paper for future studies. It could be considered for publication after minor correction according to following comments/recommendations;

1- Abstract is not good enough, improve it with the numerical results and write it again succinctly.  section should be concisely reflected the content and summarize the problem, the method, the results, and the conclusions. I would recommend important data and outcomes.

We agree and have changed the abstract accordingly, giving more space to the outcome of our experiments.

2- Some sentences are so long so that it is hard to follow the topic and there are . Please go over the manuscript again for proof reading

The writing of the paper has been revised by several native-speaker authors to make sure that there are no language issues, specifically regarding too long sentences and grammatical mistakes. Changes are annotated in the revised version.

3- In section 3.3, there is a table without caption. Please correct this or move the table to supplementary section.

This table was placed erroneously; it belongs to the supplementary tables.

4- Whole manuscript needs to  in ‘word’ program such as numbering, page design

 We have added page and line numbers, while keeping a plain design. Further editing will be performed after the final revisions have been made.

The manuscript can be published in Symmetry after these minor corrections.

Reviewer 4 Report

Check for grammatical errors.

Some sentences are too long, which makes it challenging to follow the author's message. Breaking down complex sentences into smaller and more concise statements to enhance clarity.

It would be beneficial also to discuss any known limitations.

Author Response

Check for grammatical errors.

Some sentences are too long, which makes it challenging to follow the author's message. Breaking down complex sentences into smaller and more concise statements to enhance clarity.

 The writing of the paper has been revised by several native-speaker authors to make sure that there are no language issues, specifically regarding too long sentences and grammatical mistakes. Changes are annotated in the revised version.

It would be beneficial also to discuss any known limitations.

Specifically for the section on detector distances, we extended the discussion of the validity of our conclusion in different experimental conditions. For the other discussed aspects, we discuss the scope and limitations of the chosen approaches thoroughly. A discussion of the scope and limitations of 3D ED as a whole is beyond the scope of this paper, we refer the reader explicitly to a range of review papers which discuss this in-depth in the first paragraph of section 1.

Reviewer 5 Report

The paper by Truong et al. shows relevant results obtained on organic samples with 3D Ed electron diffraction collected on a novel electron diffractometer. The authors demonstrate that with this new instrument all the most advanced structure solution and refinement methodologies can be successfully applied, obtaining accurate structures and determining also the correct absolute configuration.

The advantages of using cryoprotection sample preparation methods for analyzing vacuum sensitive samples that can de-hydrate inside the TEM is successfully discussed and demonstrated. The same instrument is also applied to the challenging problem of protein structure determination.

This referee thinks that this paper should be published in its actual form.

Minor remarks

In the figure 4 (g) (h); (i) (j) what is discuss in the text does not appear clear. The images seems to be very similar and it is not possible to appreciate the better quality of (g)  over (h) and (i) over (j). Can you please improve that?

English quality is good

Author Response

The paper by Truong et al. shows relevant results obtained on organic samples with 3D Ed electron diffraction collected on a novel electron diffractometer. The authors demonstrate that with this new instrument all the most advanced structure solution and refinement methodologies can be successfully applied, obtaining accurate structures and determining also the correct absolute configuration.

The advantages of using cryoprotection sample preparation methods for analyzing vacuum sensitive samples that can de-hydrate inside the TEM is successfully discussed and demonstrated. The same instrument is also applied to the challenging problem of protein structure determination.

This referee thinks that this paper should be published in its actual form.

Minor remarks

In ig)  overi) over (j). Can you please improve that?

Indeed, the improvement at long DD was hard to see. We hence have added panels showing a close-up. We believe that those highlight the better definition of the peaks (especially for crystal 1) very well.

Reviewer 6 Report

This manuscript falls within the context of 3DED, and more precisely within the framework of the use of the Rigaku XtaLAB Synergy-ED electron diffractometer for the structure resolution of proteins or small molecules. It aims to provide concrete information to future users of diffractometers, particularly non-microscopists, to enable them to choose the conditions for acquiring and processing data from the point of view of some parameters: influence of cryogenics on the samples, effect of irradiation dose, choice of camera length. Moreover, it aims to provide a rapid method for determining the absolute structure of small molecules. 

The introduction clearly situates the framework of the work. The context is strongly documented by a relevant and rich bibliography. Sample preparation and acquisition conditions are precisely described. The study of the effect of cryogenics on the compounds presented in this article is sufficiently detailed and factual, likewise the data acquisition and processing. 

The discussion resumes each result. it shows the relevance of cryogenic conditions. It offers a rapid method for the first identification of the absolute structure. It also offers a simple calculation to define the optimal camera length, based on parameters easily accessible to users. The tables given at the end of the manuscript contain the information and parameters needed to assess the quality of the crystallographic structures.

I recommend the publication of this article, with a few minor corrections:

page 3 line 8: Please give information on the HyPix-ED detector. At least the number of pixels and the pixel size. Ideally also the maximal number of frames / second. The pixel size (100 µm) is given in discussion, but I think that the information should be given earlier.

page 3 line 22: “Morphological properties (size, thickness) optimal for electron diffraction”. If the paper aims to address non-microscopist crystallographers, these subjective criteria, well known to microscopists, should be further described.

page 3 line 31: “1µm or 2µm diameter” should be “1µm or 2µm apparent diameter”

page 3 line 36: “See Supplementary Tables for details” should be “See Supplementary Cristallographic Tables for details”

page 4 figure 1: The figure is relevant but images and diffraction patterns are too small. Coloured circles are hard to see. The figure should be enlarged to make it easier to read.

page 6 line 16: Jana 2006

page 6 line 24: “L-(S-)-tyrosine”. “L-tyrosine” seems to have been the chosen denomination throughout the rest of the paper. Please stay consistent.

page 6 line 25: “see Methods”: There is no “methods” part in the paper. All the process details are reported in the tables in the part “Supplementary Cristallographic Tables”.

page 6 line 32: Same as line 24.

page 6 line 33: Don't skip a line

page 8 line 22: Don't skip a line

page 8 table: This table is the same as table 3 (Supplementary Cristallographic Tables). It should be removed.

page 9 line 1: “Methods”: There is no “methods” part in the paper.

page 9 line 7: (figure caption): "at XXX sigma". Please clarify.

page 9 last line: Don't skip a line.

page 10 last line: Don't skip a line.

page 11 line 6 (Figure caption) : “selected area aperture of apparent diameter 2 µm” instead of “diameter 2 µm”

page 19 table 4. It would be more consistent to present the results in the same way as in Tables 1-3.5.

Comments for author File: Comments.pdf

Author Response

This manuscript falls within the context of 3DED, and more precisely within the framework of the use of the Rigaku XtaLAB Synergy-ED electron diffractometer for the structure resolution of proteins or small molecules. It aims to provide concrete information to future users of diffractometers, particularly non-microscopists, to enable them to choose the conditions for acquiring and processing data from the point of view of some parameters: influence of cryogenics on the samples, effect of irradiation dose, choice of camera length. Moreover, it aims to provide a rapid method for determining the absolute structure of small molecules. 

The introduction clearly situates the framework of the work. The context is strongly documented by a relevant and rich bibliography. Sample preparation and acquisition conditions are precisely described. The study of the effect of cryogenics on the compounds presented in this article is sufficiently detailed and factual, likewise the data acquisition and processing. 

The discussion resumes each result. it shows the relevance of cryogenic conditions. It offers a rapid method for the first identification of the absolute structure. It also offers a simple calculation to define the optimal camera length, based on parameters easily accessible to users. The tables given at the end of the manuscript contain the information and parameters needed to assess the quality of the crystallographic structures.

I recommend the publication of this article, with a few minor corrections:

page 3 line 8: Please give information on the HyPix-ED detector. At least the number of pixels and the pixel size. Ideally also the maximal number of frames / second. The pixel size (100 µm) is given in discussion, but I think that the information should be given earlier.

Those specifications are now given in section 2.2.

page 3 line 22: “Morphological properties (size, thickness) optimal for electron diffraction”. If the paper aims to address non-microscopist crystallographers, these subjective criteria, well known to microscopists, should be further described.

We added a sentence making the morphological parameters more explicit: “For typical organic compounds or proteins as studied in this work, a crystal thickness of the order of 500 nm along the directions covered by the rotation scan is desirable for an optimal trade-off between signal strength and reasonable levels of background due to inelastic scattering.“

page 3 line 31: “1µm or 2µm diameter” should be “1µm or 2µm apparent diameter”

Changed.

page 3 line 36: “See Supplementary Tables for details” should be “See Supplementary Cristallographic Tables for details”

Changed.

page 4 figure 1: The figure is relevant but images and diffraction patterns are too small. Coloured circles are hard to see. The figure should be enlarged to make it easier to read.

We revised the figure, significantly enlarging the real-space images and diffraction patterns.

page 6 line 16: Jana 2006

Changed.

page 6 line 24: “L-(S-)-tyrosine”. “L-tyrosine” seems to have been the chosen denomination throughout the rest of the paper. Please stay consistent.

We have changed the labeling; however, as the S-/R- nomenclature is more general and technically more correct, we have added the explicit remark “L- and inverted D-enantiomer (equivalently referred to as S- and R-enantiomer)”

page 6 line 25: “see Methods”: There is no “methods” part in the paper. All the process details are reported in the tables in the part “Supplementary Cristallographic Tables”.

Changed (deleted “Methods”).

page 6 line 32: Same as line 24.

page 6 line 33: Don't skip a line

page 8 line 22: Don't skip a line

page 8 table: This table is the same as table 3 (Supplementary Cristallographic Tables). It should be removed.

We suspect that those issues were artefacts in the version of the manuscript sent for review.

page 9 line 1: “Methods”: There is no “methods” part in the paper.

Changed to section 2.3.

page 9 line 7: (figure caption): "at XXX sigma". Please clarify.

Changed

page 9 last line: Don't skip a line.

page 10 last line: Don't skip a line.

See above.

page 11 line 6 (Figure caption) : “selected area aperture of apparent diameter 2 µm” instead of “diameter 2 µm”

Changed.

page 19 table 4. It would be more consistent to present the results in the same way as in Tables 1-3.5.

We use table entries and definitions common within the small-molecule (1-3, 5) and protein (4) crystallography communities, respectively, which are divergent due to the quite different processing, refinement and validation pipelines (and historical reasons). Trying to get those on a common denominator would lead to an awkward outcome outside of the most basic parameters.

Reviewer 7 Report

Please see attached word document.

Comments for author File: Comments.pdf

Author Response

Introduction:

Page 2, paragraph 2, the author mentioned that 'some modifications to optimize performance for 3D ED' has been implemented. Can they elaborate on the point?

The implemented changes pertain to control electronics allowing for better synchronization and rotation speed control. To keep trade secrets, we prefer not to disclose any further details on that matter.

Consider adding https://doi.org/10.3390/sym13112131.

We thank the reviewer for this very pertinent suggestion, especially given that our result confirms the outcome of that paper only in a very limited way – we do see a slight improvement of quality (albeit less than expected by many), but no actual decrease. We have added the reference in two places 4.1 in the discussion:  “It was recently reported by Yang et al. [56], that cryogenic data collection even decreased data quality for test systems including an anhydrous sugar, highlighting that the impact of temperature for vacuum-insensitive samples is highly sample-dependent and warrants further study.”

Material and Methods:

Page 3, paragraph I, how long did it take to make IO lamellae?

Approximately 20 minutes per crystal. This is now explicitly specified.

Page 3, paragraph 1, 'The grid was then directly cryo-transferred while remaining on the sample holder\’, was the grid mounted on the cryo-transfer holder during FIB milling, or was the grid loaded onto the holder after milling?

As stated, the grid is loaded on the cryo-transfer holder right after plunging and stays there at all times. Differing from other FIB systems, the JIB-4000Plus features a TEM-type load-lock and stage/gonio, such that this workflow becomes possible.

Page 3, paragraph 2, the authors mentioned that data were collected on identical Rigaku diffractometers located at various sites. Could data of the same sample (or same batch of sample) collected at different sites be scaled/merged together? I suppose even data collected at different camera length/detector distance can be scaled/merged together.

This should indeed be possible without major obstacles, but was not yet attempted (or required for the data discussed in the paper).

Page 3, paragraph 3 and elsewhere, consider use rotation speed or tilt speed instead of scan speed.

While “Tilt speed” is often used in electron microscopy, “Scan speed” is the common term in Single-Crystal (X-ray) diffraction. As this is the target audience for the paper, we would opt for keeping this nomenclature.

Results:

Page 5, paragraph 2, 'Due to the rather high symmetry point group, merging datasets did not lead to a significant increase of data completeness'. Was it because of preferred orientation of the crystals? I don't see how symmetry could affect the improvement of data completeness by merging.

Due to the high symmetry (resulting in high multiplicity), the single grain provides sufficient completeness (84.6%) for solution and refinement, which it would not in e.g. a triclinic space group. Redundancy, I/sigma and Rpim would likely be improved by merging of grains; however, due to the proof-of-principle nature of the study, we considered it a more pertinent experiment design to keep things focused and forgo merging (which would add yet another discussion point).

To clarify, we rephrased the sentence as follows: “Due to the rather high-symmetry point group, a dataset of sufficient completeness could be collected from a single crystal.”

Page 6, paragraph 1, most 3D ED/MicroED work done so far relied on kinematical approximation. Avoiding collecting data along major zone axes help reducing dynamical scattering. Mosaicity of the crystal and incoherence also reduce the dynamical effect.

We have added the sentence: “Making use of the kinematical approximation still yields satisfactory results, as data collection off zone axes [1], mosaicity on the nano-scale [44], and inelastic scattering [45] mitigate dynamical effects [9].”, specifically to refer the reader to pertinent literature (Kolb et al. on ADT, Gallagher-Jones et al. on mosaicity, Latychevskaia and Abrahams on inelastic scattering, Saha et al. for a recent general review of the issue).

Page 8, paragraph 2, having only 3 pixels between adjacent Bragg spots could cause problem when processing data with other software, although the authors were able to process the data using CrysAlisPT0. Since the crystal diffracted to 2.15Å, why didn't the authors use a longer camera length (detector distance)? At 1800mm, the detector can cover up to 1.2Å, which is not needed in this case.

Indeed, 1.2Å is optimistic to say the least. However, note that the detector is rectangular, and along the short direction, we will get a cut-off at 2.4Å already. Knowing that a 3 pixel spacing is possible, 1800mm was a reasonable choice.

Was 1.2Å at the edge or the corner of the detector?

It is at the edge; while technically we might specify the corner, we consider this the more relevant/honest quantity. We clarified this in the manuscript.

Page 8, table 3 is inserted by mistake here?

Yes, seems this somehow spuriously slipped into the version as sent out for review.

Page 9, caption of Figure 3, 2Fo-Fc map at 'XXX' sigma. Please fill in the number

Changed

Page 9, paragraph I, when setting up unattended data collection, were the crystals picked manually or automatically? The corresponding author is an expert in this aspect.

The crystals were selected manually. CrysAlisPro does not yet implement automatic selection of candidate crystals; also, there is good reason to believe that for this particular sample with strongly intergrown and aggregated crystals of uneven morphologies, a sophisticated selection algorithm going far beyond conventional image processing approaches (e.g. by Smeets et al.) would be required.

To clarify, we changed the sentence to “Using automated screening collection (unattended queue mode, 4° collection range) from ≈ 100 manually selected candidate crystals on two grids, we identified two well-diffracting, only weakly intergrown crystals at starkly different orientations with respect to the sample grid (Figure 4(a-f)).”

Page 9, Did the accuracy of unit cell parameters improve when data was collected at longer camera length (or shorter camera length)?

We did not see a systematic relation. It might be an interesting point for further studies, as there would be arguments in favor of both directions, but we would not expect the effect to be too strong.

Page 9-10, based on visual inspection, as well as data processing statistics, the signal- to-noise was better at shorter camera length, especially for weak reflections. The 1/0(1) and data redundancy were also higher for data collected at shorter camera length. I am puzzled by why the for data collected at short camera length.

In order to better visualize the better definition of peaks we have added close-up panels of the reciprocal-space projections to Figure 4. We believe that especially for crystal 1, it is now evident that peak overlap is decreased, both within the crystal and with the spurious intergrown one. Regarding the improved refinement metric, we would like to highlight, that better I/sigma as well as lower Rpim (arising from higher redundancy or otherwise) indicates a better precision of reflection intensity, but not an improved accuracy. However, it is the latter, which eventually determines the quality of the refinement as witnessed by, e.g., R1. Accuracy might indeed be increased by more favorable integration conditions, such as better peak separation and UB definition, despite a penalty in precision.

To discuss this in a more explicit way, especially for crystal 1, we have rephrased parts of the pertinent paragraph as follows: “Despite the unproblematic geometry, we clearly observed the best refinement metric R1 from the longest DD of 1289 mm (Figure 4(o), Table 5) at similar precision indicators Rint and Rpim (Figure 4(p,q)). This suggests that the higher data redundancy achieved at low DD is outweighed by higher accuracy of integrated intensities due to better peak fitting and decreased overlap specifically with a weak intergrown second crystal.”

(page 20 Table 5), although the numbers of total (unique) reflections used for refinement were similar for Cystine crystal I, 653mm, 1029mm and 1289mm, thetable) differed quite a lot. The numbers of total measured reflections were 7866, 6143 and 4983, and the number of strong (unique) reflections were 818, 669 and 571 for data collected at 653mm, 1029mm and 1289mm, respectively. The R1 for all reflections were 0.170, 0.160 and 0.154. Could it be that some of the symmetry related reflections (not measured using longercamera length) were strongly affected by dynamical scattering when measured at653mm?

In our understanding, if this was the case, the most immediate measure would be an elevated value of Rint. For neither of the two crystals we find the shortest DD to have a significantly higher value. (also see below). Note that the strong reflections are indeed listed in the table (“observed”).

Were the repeated measurements carried out over identical crystallographic orientations?

Yes. To clarify, we now explicitly state this in the paper: “For each of the crystals, we conducted three data collections at identical crystal orientation and scan range, but different settings of DD and a fine frame slicing/scan width of 0.15° for optimal resolution of peaks in three dimensions, […]”

Were there any changes in crystal orientation between measurements (i.e. induced by the beam or the crystal rotated by itself between measurements)?

We double-checked the rotation matrices from indexing refinement and can confirm that orientation differences are less than 1 degree (up to ambiguity arising from the Laue group) for both crystals.

Usually refining against less reflections would yield lower R1 value, 0.164 for 818 strong reflections, 0.140 for 669 strong reflections, and 0.122 for 571 strong reflections. The different is less for R1(all): 0.170 for 880 reflections, 0.160 for 882 reflections and 0.154 for 877 reflections.

Thank you for this observation. Indeed, both R1(obs) and R1(all) should be included. We have added R1(all) into the tables. This does not impact the qualitative conclusions.

Overall, I think the choice of camera length is sample dependent. It is a balance between having enough separation between adjacent Bragg spots and maximum achievable resolution.

We think that this discussion is included rather thoroughly at the end of section 4.3. We also added a sentence on influence of detector characteristics (on request by another reviewer), and the new sentence part “despite a possible penalty in completeness or precision (though not accuracy) of reflection intensities”. As we are highlighting, we believe that for small molecules, especially the cases of sub-optimal crystal morphologies (as in the presented case) and unknown heterogeneous mixtures may benefit from longer DD.

Page 11, caption of Figure 4, please add detector distance for (a) and (d).

We added the DD to the caption.

Discussion

Page 13, paragraph 2 (section 4.3), again, the detector coverage up to 1.7Å is at the corner or the edge of the detector?

Again, it refers to the edge, which is now explicitly mentioned.

References:

Please check the format of preprint citations (ie. publications on ChemRxiv).

We have changed preprint citations to include repository and doi.

Crystallographic Tables:

Please include , number of strong reflections (F2 > 20(F2), and number of restraints in all tables for small molecule refinement. Where possible, the authors refined the structure without using the command 'EXTI'. In some cases, they used it to remove a number of NPDs in ADPs. I think it is a reasonable approach. Although X-ray extinction is not presented in electron diffraction the use of 'EXTI' improves the weighting of low angle reflections, in turn improving the numerical values of ADPs. Having said that, if 'EXTI' is used, the values of ADPs become less interpretable.

The tables were revised. Indeed, using the EXTI can be used as a “fudge factor” to compensate for dynamical effects to a very low-order approximation, mitigating negative ADPs (as discussed e.g. in Yang et al. J Appl Crystallogr 55, 1583–1591 (2022)) - in any case, we believe that ADPs in both kinematical and dynamical refinement should be considered with a lot of caution in ED, a discussion which however would be far beyond the scope of the paper.

Table l: Trehalose anhydrous (100K) had more number of parameters. This was due to the fact that several H atoms were refined freely in that model.

This is correct.

Table I: In some cases, the gap between (all) and RI(F2 > 20(F2)) is very large, suggesting Iow agreement between the model and weak reflections. Can the authors investigate the reason behind it? Was the measurement of weak reflections less reliable due to incident loss or the threshold for counting set too high?

As the HyPix-ED is a fast-counting hybrid-pixel detector, coincidence losses do not play a role at any conceivable peak intensity (different from monolithic detectors), and there is no reason to believe that the accuracy of weak reflections would be lower. We hence suspect this to be rather due to unmodeled effects such as dynamical diffraction, redistributing intensity into weak reflections.

Table 4, please add standard deviation for the unit cell parameters. P should be italic in "P41212”

P has been changed to italic; unit cell SD are now included.

Table 5, comments for Cystine 1 were mentioned above. The refinement of Cystine 2 are less reliable since the number of parameters are far less than 10x of the number  to see how well the model fits the data. was much less than that for data collected at 777mm (251).  Cystine I case. Refining against less reflections would usually yield better R1I hope the authors could reanalyze the data and carefully adjust their claims made in section 3.3 (Cystine part).

We thank the reviewer for their careful analysis and valuable suggestions on section 3.3. While the qualitative conclusions stay unchanged, the reviewer comments helped us a lot to clarify and contextualize the results.