Dear Anonymous reviewer,

Thank you very much for your positive evaluation of my manuscript.

On line 180 of the current manuscript the author indicates that the sensitivity of the PRESERVE-TROSY would depend on a number of factors, including protein size. Noting that the length of the PRESERVE-TROSY element is not trivial might the authors predict over what sizes of molecule the gains should be apparent. Perhaps, starting with the experimental ubiquitin data of Figure 2D and including effects of the HN T2s the author might be able to provide some guidelines in this respect. Of course the T1s will increase with molecular size, compensating for the shorter T2s.

As pointed out in my general comment, the presented experiments may provide improved sensitivity for perdeuterated proteins (probably only up to a certain size/tumbling correlation time), and other molecules with long ¹H and ¹³C/¹⁵N T₁. At the present stage, it is difficult to give more detailed guidelines than these general rules.

It is not immediately clear to me how best to set-up an experiment in terms of choice of trec values. I note that the author shows examples where trec varies from 1 ms to several hundreds of ms. The author provides an eqn for SNR from which the effect of trec can be calculated, but perhaps there are some intuitive rules that could be provided.

The experiments are intended to be useful when short overall experimental times are desired, requiring fast repetition rates. In practice, the behaviour will be different for different molecular sites (see Fig 2D) characterized by different ¹H T₁.

Figure 4. How were the different beta values selected? Just from the equation for SNR?

A series of spectra was recorded with varying flip angle, and the spectrum with the highest average SNR was plotted.

How does the SNR compare between a SOFAST-TROSY spectrum with lamda optimized (such as in Figure 5A) and a second spectrum recorded with the same sequence but where lamda=0 (ie, effectively no PRESERVE), but of course in both cases using selective 1H pulses to ensure a bath of protons for restoration of signal. Gains of 20% or so are noted in Figure 5 relative to beta=90o, but by effectively deleting the PRESERVE element and thus shortening the pulse sequence could this gain be compensated, especially for larger systems (ie, recording SiR at 5oC)?

For the different experimental data shown in this manuscript, only for highly deuterated proteins (ubiquitin and SiR) a substantial sensitivity gain was observed when compared to a conventional BEST-TROSY implementation (as available in our NMRlib package). See also my general comments on the manuscript.

Dear Eriks,

Thanks a lot for your positive evaluation of my manuscript. I will implement all suggested improvements in a revised version of the manuscript (or the proofs).

Dear Dr. Parella,

Thank you very much for your positive evaluation of my manuscript. Please see also the general comments I have posted to respond to all reviewers.

In my opinion, the reader may be confused when similar terms are used to refer to Trec i trec. I would suggest using different terminology for one of the two.

I propose to avoid the additional T_rec term and replace it by t_rec+t_acq !

A more detailed experimental part is missing, where more information about the acquisition conditions could be found. For example, it is necessary to mention the total acquisition time required for each spectrum and specify the number of increments, scans or the potential use of NUS. The reader is interested in these methods as rapid acquisition techniques and it is relevant for each sample and concentration, how much spectrometer time is necessary.

As the purpose of the experimental data is to illustrate the performance of the variable flip angle adjustment in PRESERVE, and in some cases the sensitivity comparison between different pulse schemes recorded under identical acquisition conditions, I feel that these details do not add any useful additional information. In practice, the choice of acquisition parameters will depend on sample concentration, NMR spectrometers, desired spectral resolution, etc. Typically, the proposed experiments are only of interest if short experimental times are required and feasible. Of course, NUS may be an alternative or complement for reducing acquisition times.

It would be interesting to add some comments on the potential effects of rapid pulsing on practical aspects of sample heating or the generation of unwanted artefacts.

Figure 4a serves this purpose. Spectra with very clean line shapes are observed at highest possible repetition rates, demonstrating the performance of this particular sequence under fast pulsing conditions. Actually, TROSY is particularly amenable to fast pulsing as it does not require heteronuclear decoupling during ¹H detection. Note that artifact free spectra are only obtained with a 2-step phase cycle of the PRESERVE sequence (as explained in the manuscript).

It is interesting to conclude in which real cases these methods can become relevant. I have the impression that these methods should aimed at deuterated proteins in very high magnetic fields but are hardly better than those existing in the case of protonated proteins or 13C applications.

I agree ! See my general comments.

There is a very important aspect that is not discussed in most of the publications describing sofast NMR methods. It is essential to know the “signal saturation percentage” of the proposed preserve-TROSY and SOFAST-TROSY sequences. For example, from the data of FIG. 4B, what would be the absolute sensitivities of a reference SOFAST-TROSY experiment acquired with beta=90º using a long trec time of 2-3 s? The diagrams in this figure only display relative intensities but no reference is made about absolute intensities. As another relevant data, It should be interesting to compare absolute signal intensities with beta=60 and trec=300ms vs beta=43 and trec=1ms. The reader needs to know how critical it can be to use very short trec values.

Again, the aim of the experimental data was not to discuss any possible absolute sensitivity gain, by adjusting the recycle delay separately for different flip angles. These possible enhancements are generally quite moderate as calculated from theory and experimental considerations, such as the higher saturation of “passive” spins at short repetition rates (see for example Schanda, Prog NMR Spectrosc. 2009). Variable flip angle experiments are only of interest if (t_rec+t_acq) << T₁ (figure 1).

When speaking about sensitivity enhancements it must be done in terms of the total acquisition time, so the term “sensitivity enhancement per time unit” should be more appropriate to have a clear idea of ​​the real sensitivity and the acquisition time needed.

In my understanding, the notation “sensitivity” means SNR for a given experimental time. Therefore, making a difference between “sensitivity” and “real sensitivity” does not make much sense ?!

SOFAST-TROSY is probably interesting for 15N but the author also provides examples on 13C. Some comments on the real sensitivity of the spectra in Fig. 5B, 5C and 5D compared to conventional TROSY and HSQC data acquired under the same experimental time would be valuable to understand the suitability of these 13C experiments. In the case of Fig. 5D, some comments on the use of the SOFAST sequence and the effect of band-selective pulses on J-coupled systems such as those found in sugar would also be advisable. Are selective pulses really necessary? In all these 13C applications it would also be very informative to provide the degree of absolute signal saturation losses suffered when acquired with trec=0.

Once again, figure 5 is only intended to show that these pulse sequences can be applied to various ¹H-¹⁵N or ¹H-¹³C spin systems yielding clean spectra and allowing for variable flip-angle adjustment. Spin-coupling evolution can be easily inferred from the “replacement” schemes of the selective pulses applied. Whether they are necessary (LRE effect) depends on the application.