Use of Raman and Raman optical activity to extract atomistic details of saccharides in aqueous solution

Sugars are crucial components in biosystems and industrial applications. In aqueous environments, the natural state of short saccharides or charged glycosaminoglycans is floating and wiggling in solution. Therefore, tools to characterize their structure in a native aqueous environment are crucial but not always available. Here, we show that a combination of Raman/ROA and, on occasions, NMR experiments with Molecular Dynamics (MD) and Quantum Mechanics (QM) is a viable method to gain insights into structural features of sugars in solutions. Combining these methods provides information about accessible ring puckering conformers and their proportions. It also provides information about the conformation of the linkage between the sugar monomers, i.e., glycosidic bonds, allowing for identifying significantly accessible conformers and their relative abundance. For mixtures of sugar moieties, this method enables the deconvolution of the Raman/ROA spectra to find the actual amounts of its molecular constituents, serving as an effective analytical technique. For example, it allows calculating anomeric ratios for reducing sugars and analyzing more complex sugar mixtures to elucidate their real content. Altogether, we show that combining Raman/ROA spectroscopies with simulations is a versatile method applicable to saccharides. It allows for accessing many features with precision comparable to other methods routinely used for this task, making it a viable alternative. Furthermore, we prove that the proposed technique can scale up by studying the complicated raffinose trisaccharide, and therefore, we expect its wide adoption to characterize sugar structural features in solution.

cal details. As a results we proposed a simulation protocol that offers a quality Raman/ROA spectra that is also extremely cost-efficient to simulate. The aim of current work is to show how our previously developed protocol can be used to access to a plethora of structural features of saccharides in solution. We now show that we when combining the experimental spectra with our protocol we can obtain various features of saccharides in solution (puckering, glycosidic bond, mixtures, larger saccharides) which was not the point and most were not even tackled on our former work. In fact, our present work shows that our previous protocol with just small fine tuning can tackle significantly more complex tasks that those we used for its development, including state of the art cases for ROA/Raman techniques such as trisaccharides. To sum up, our former work focuses on technicalities of simulating Raman/ROA spectra, while the current work focuses of using the developed protocol to interpret experi-2 mental data and access various structural features of saccharides in solution. These are two completely different aspects, although we obviously agree with the reviewer that without our previous methodological work, this work will not be possible and we certainly build on our previously developed technique to show how powerful it is.
We never hide this fact, and our previous work is heavily cited in the present one to make clear its origin. Still, to avoid any misinterpretation, we have added some clarifying text to the manuscript that we believe emphasizes the differences with our previous work.

Question
To understand properly all methods used here, I read the authors' previous work (Palivec et al. PCCP.,2020, 22, 1983, together with the Reply to Reviewers that was added as the SI. After careful reading, I still have concerns about scaling of calculated intensities. In the previous paper to which the reader is referred, you mentioned "The average magnitudes of ROA and Raman intensities are experimentally related (I(ROA) I(Raman)x10 −4 ), however, we optimize both spectral intensities separately. As a result, the ratio is not necessarily preserved during the optimization", Is it the case also in this paper? If yes it should be mentioned somewhere, and discussed, especially because all simulated and experimental ROA/Raman spectra are presented here in some puzzling arbitrary units, where ROA and Raman intensities are comparable, although one can expect rather 10 −4 ratio of ROA/Raman intensities.
In my opinion it will be puzzling for future readers. I have a problem with it because you are losing information about CID ratios. As we all know, CID ratios are inherently associated with ROA, and their calculation and comparison with experiment can improve the reliability of structural conclusions (Polavarapu, CHIRALITY 26:539-552 (2014)). At least I would suggest that the original ROA/Raman ratios of both experimental and simulated spectra should be preserved during all the scaling, and minimization of the cost function.
I'm curious how would it change the final results?

Answer
The reviewer is concerned about scaling of calculated Raman/ROA spectral intensities to match experimental ones using two different intensity scaling factors and asks how would this change our results. Yes, we are using the simulation method developed in our previous work. Therefore we are using two different scaling factors, which does not preserve the CID ratios as was already explained and thoroughly discussed in the methodological paper. Firstly, we would like to stress that our methodology is completely agnostic to absolute intensities of Raman/ROA spectra. Therefore, our final structure prediction results would remain completely unchanged. This is because their evaluation is based upon optimizing overlap integral values, which are independent on individual intensities. Therefore, the only thing that would change with using same scaling factors is the visual comparison of Raman/ROA spectra.
Moreover, we would like to stress that the CID shape remains unchanged during the whole process, up to a constant. This is because we scale Raman/ROA with two factors(ρ Raman , ρ ROA ) and so Therefore, still, our spectra can be used to produce CID which can be compared to experimentally observed CID. Lastly, we checked the difference between separate scaling factors and single scaling factor on three moieties and the average difference is negligible, i.e., 2.7%.
To summarize, our two factor scaling have no effect whatsoever on our structure predictions and the difference as compared to a single scaling factor is negligible, therefore, visually there would be essentially no difference upon switching to a single scaling factor(up to 2.7% change in intensities). Altogether, change to a single scaling factor would effectively have no effect on current manuscript at all.
Nevertheless, we agree that using two scaling factors and not preserving CID ratios is an important part of our approach and therefore we are now discussing it in the manuscript.
Moreover, we agree that using ideas that are discussed in (Polavarapu, CHIRALITY 26:539-552 (2014)) are valuable, e.g., the concept of robustness, and could be used to further improve the reliability of structural predictions of our methodology.

R1Q3
Question About bibliography. I would suggest to enrich the bibliography with the latest papers on ROA of trisaccharides, or polysaccharides e.g. 10.1039/C9CP00472F, or 10.1016/j.saa.2018.08.017, where in the latter, raffinose ROA spectrum is discussed.

Answer
The discussion now includes these works as they certainly justify the interest of this work to cover larger saccharides.

R2Q1
Question This manuscript reports a detailed structural study of six model monosaccharides in water using Raman, ROA and sometimes NMR data, coupled with a combined molecular dynamic (MD) and quantum mechanics (QM) approach. One well-known challenge in interpreting the VOA spectra observed is how to properly account for water solvation. In should be cited.

Answer
We thank the review for the suggestions and we agree. The citations have been added and are now discussed in the introduction.

Question
There are some words used whose meaning is ambiguous. For example, the authors used "mobile" often in the text. "Unfortunately, many available structure characterization techniques are inadequate when applied to mobile molecules such as saccharides". All molecules are mobile in solution. I think the authors want to emphasize "flexibility" of saccharides and should change the wording accordingly.

Answer
We agree and we changed the text accordingly.

R2Q3
Question "ROA is a Raman-based spectroscopic method where the difference in right/left circularly polarised light spectra is recorded." A more precise definition of ROA should be provided.

Answer
We change the text to provide better and more precise definition of ROA: "ROA is a Raman-based spectroscopic method where the difference in Raman spectra when using right(I R )/left(I L ) circularly polarised light spectra is recorded, i.e., I ROA =I R -I L ."

R2Q4
Question "The advantage of Raman/ROA techniques when studying saccharides is a transparent vibrational spectral region between 75 and 3100 cm −1 ," Does this refer to water, the solvent, which has weak Raman in the region?

Answer
Yes, but also to the transparency of the optics used. Since the Raman effect is a small shift away from the incident photons, optics that are transparent in the visual range can be used even though we are probing vibrational frequencies. See below for limitations in IR.

Question
Can one apply the same approach to VCD spectra of chiral molecules in water? How well will it work? Can the authors make some comments on this aspect?

Answer
The computational approach can easily be applied to VCD(as documented in ref. 62). The problem with VCD on carbohydrates is experimental. While application of Raman/ROA on carbohydrates result in strong signals, IR/VCD on this group of molecules is very weak, making them less appealing targets for VCD analysis(as can be seen, e.g., in Chirality 2008, 20, 446;JACS 2004, 126, 9496;Org. Biomol. Chem., 2007, 5, 1104Carbohydr. Res. 2004, 339, 2713. Combine that with water being a strong absorber of IR radiation and limitations to range due to limitations to optical transparency, IR/VCD on carbohydrates would be limited to the range 1000-1500 cm −1 . This is severely limiting the amount of information that can be compared to the predicted spectra, and therefore Raman/ROA is the usual target when characterizing saccharides in solution.

Question
In conclusion, this is an interesting and well-performed study, which fully deserves publication in PLOS Computational Biology pending the modifications suggested above.

Answer
We thank the reviewer for his kind words, and we hope that the changes and clarifications provided above are satisfactory. 8