the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 11 Jan 2021
Protein dynamics insights from 15N-1H (TROSY) HSQC
Abstract. Protein dynamic information is customarily extracted from 15N NMR spin-relaxation experiments. These experiments can only be applied to (small) proteins that can be dissolved to high concentrations. However, most proteins of interest to the biochemical and biomedical community are large and relatively insoluble. These proteins often have functional conformational changes, and it is particularly regretful that these processes cannot be supplemented by dynamical information from NMR.
We ask here whether (some) dynamic information can be obtained form the 1H line widths in 15N-1H HSQC spectra. Such spectra are widely available, also for larger proteins. We developed a computer program to predict amide proton line widths from (crystal) structures. As a calibration, we test our approach on BPTI. We find that we can predict most of the distribution of experimental amide proton line widths if we take the dipole-dipole interaction with at least 40 surrounding protons into account. When focusing our attention the outliers of the distribution, we find for BPTI a cluster of conformationally broadened 1HN resonances of residues in strands 10–15 and 36–40 of the beta sheet. Conformational exchange broadening of the 15NH resonances for these residues was previously reported using 15N relaxation measurements (Szyperski et al., J. Biomol. NMR 3, 151–164, 1993). There is little or no evidence for motional narrowing of the 1HN resonances, also in agreement with earlier data using 15N relaxation methods (Beeser et.al, J. Mol. Biol. 269, 154–164, 1997). We also apply our program to 42 kDa domain of the human Hsc70 protein. In this case, there is no previous 15N relaxation data to compare with, but we find, again from the outliers of the distribution, both exchange broadening and motional narrowing that appears to corroborate previous conformational insights for this domain.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(6306 KB)
-
This preprint has been withdrawn.
- Preprint
(6306 KB) - Metadata XML
- Revised submission
- BibTeX
- EndNote
Interactive discussion
Status: closed
-
RC1: 'Comment on mr-2020-38', Anonymous Referee #1, 20 Jan 2021
The article tries, and fails, to correlate measured line widths with line widths calculated from crystal structures, assuming dipolar relaxation as the dominant relaxation mechanism.
The ultimate conclusion is that amides with narrower line widths than average are mobile, whereas amides with broader line widths reflect us – ms dynamics. This is not at all new and the conclusion in line 215 is trivial.
Using M = 3 in equation 1 is inappropriate for a sensitivity-enhanced HSQC, as magnetization is not transverse throughout all delays of the PEP scheme.
Equation 4 applies to indistinguishable spins. It does not apply to amide protons with distinguishable chemical shifts (see Bothner-By et al., JACS 106, 811 (1984)
or equations 79 and 89 in Abragam, Chapter VIII).
As admitted by the author, Figure 4 shows no correlation. Figure 6 doesn’t show a significant correlation either, again as admitted by the author. This indicates failure of the entire concept. If coordinate precision is a major problem (line 222), the concept cannot be salvaged.
Line 167: The line width of the amide proton of Asp3 could very well be due to enhanced H-exchange, as the N-terminus is positively charged and the exchange base-catalyzed.
Line 187: Much more is known about the dynamics in BPTI than suggested by the manuscript, see, e.g., Grey et al., JACS 125, 14324 (2003).
Line 225: water molecules on protein surfaces are not known to contribute to dipolar relaxation.
Line 235: detailed dynamical data have been obtained for large proteins from methyl relaxation.
Line 280: independent movement of a domain relative to the rest of the protein produces narrower lines. This is not novel: there are countless examples, such as calmodulin, trigger factor etc.
Figure 3 does not specify the acquisition times used in the spectrum - too short acquisition times would cause artificial line broadening to a variable degree for amide protons with different intrinsic line width.
Legend of Fig. 1: EM and cosine not explained
Line 57: equal to what?
Line 94: us is not a unit.
Line 155: what is meant by flattered comparison?
Line 269: what is meant by ATP conformation?
In the opinion of this reviewer, the manuscript falls far short of the criteria of Magnetic Resonance in terms of scientific impact (absence of a clear advance) or scientific quality (wrong equations used, data mostly not provided and, hence, the conclusions of the manuscript cannot be reproduced independently). It would be unexpected if it were to pass the review process of any NMR journal.
Citation: https://doi.org/10.5194/mr-2020-38-RC1 -
AC2: 'Reply on RC1', Erik Zuiderweg, 21 Jan 2021
The comment was uploaded in the form of a supplement: https://mr.copernicus.org/preprints/mr-2020-38/mr-2020-38-AC2-supplement.pdf
-
AC2: 'Reply on RC1', Erik Zuiderweg, 21 Jan 2021
- AC1: 'Comment on mr-2020-38', Erik Zuiderweg, 21 Jan 2021
-
RC2: 'Comment on mr-2020-38', Anonymous Referee #2, 22 Jan 2021
This is a largely exploratory manuscript investigating the potential to obtain (semi-)quantitative results on protein dynamics based on backbone amide 1H linewidths in 1H-15N HSQC or TROSY spectra. As Zuiderweg states himself, the underlying idea to glean dynamic information from the HSQC spectrum is commonly used by NMR spectroscopists in a qualitative sense by identifying particularly sharp or broadened cross-peaks. Here, Zuiderweg provides a more comprehensive analysis by attempting to account for all contributions to the 1HN linewidth from various relaxation mechanisms and other effects, such as unresolved J-couplings and B0 inhomogeneity. Naturally, the dominant part of the relaxation is due to dipole-dipole relaxation, which is estimated for each backbone amide based on the high-resolution crystal structure of the protein in question.
The end result is that the calculated linewidths show only fair agreement with the experimental ones, but outliers appear to reliably identify backbone amides that are either undergoing large-amplitude fast internal dynamics (i.e., residues with low order parameters) or residues undergoing conformational exchange. Thus, we are left with the conclusion that 1HN linewidths cannot provide more detailed information than what is customarily obtained from a qualitative, first-glance interpretation of HSQC-type spectra. To this extent, the work clearly does not advance the field since it does not provide any substantial conclusions beyond current knowledge. Still, I appreciate the comprehensive, semi-quantiative analysis offered by Zuiderweg, which clearly shows the limitations of the proposed analysis. In essence, the work demonstrates that 1HN linewidths cannot be interpreted in terms of dynamics to any detailed extent. For these reasons, I believe that the study could be worth publishing.
Minor points:
p. 4, Eq [1]: I do not follow this equation fully: the last two factors are not defined and it is not clear why they appear in the equation. The first exponential assumes that each step in the reverse polarization transfer (after t1) contributes equally to the linewidth, but this is not true for all pulse sequences (it depends on the details of the PEP scheme, etc).
l. 94: aromatic ring flipping does not cause exchange linebroadening of amide protons (since the end states are identical).
Table 1: Please clarify what is listed in this table. By comparing with Fig. 4, I assume that the Sum of 1HN linewidths is taken over all residue pairs in the protein(?). This should be stated in the Table header (or footnote).
Figure 6 legend: "open diamonds" should be 'open squares'.
The text should be checked for typos, incorrect order of words, missing words, etc.
Citation: https://doi.org/10.5194/mr-2020-38-RC2 -
AC3: 'Reply on RC2', Erik Zuiderweg, 26 Jan 2021
The comment was uploaded in the form of a supplement: https://mr.copernicus.org/preprints/mr-2020-38/mr-2020-38-AC3-supplement.pdf
-
AC3: 'Reply on RC2', Erik Zuiderweg, 26 Jan 2021
-
RC3: 'Comment on mr-2020-38', Anonymous Referee #3, 23 Jan 2021
The manuscript proposed by Erick Zuiderweg presents a valuable attempt to rationalize the intensities measured on protein 1H-15N correlation spectra in order to get a qualitative description of the backbone dynamics at different timescales. The different parameters affecting the 2D correlation peak's intensities are exhaustively listed and their values estimated from a structural model of the protein. The concept is tested on the BPTI protein and, as noted by the author, modelling the different known contributions to the signal intensities fails to reproduce the experimental measurements. The relevance of the approach is defended by the observation that large deviations from the modelled intensities do cluster in regions of BPTI where specific dynamical features where previously reported from 15N relaxation measurements. A graphic based clustering of these "model deviating" signals is therefore proposed as a fast approach to get qualitative insights on the protein dynamics. The approach is applied to a large protein (Hsc70) enabling some observations to be made on its dynamical properties.
As stated by the author in his introduction, such an approach would be very valuable in the field of protein NMR, as we do share the general feeling that the information content of 15N-HSQC or TROSY are under-exploited, and the author's attempt to address this task is interesting. However, my general opinion is that the current state of this development is far too preliminary and would desserve more work to be published. The general applicability of such a method should be assessed by probing the concept on different class of proteins that display distinct and well documented dynamical features (depending on their size, geometry, experimental conditions of the study T° pH ...). The "correlation" approach is indeed the only way to go when the model fails to reproduce the experiment. Anyhow, some statistical assessment is necessary for the applicability of the proposed approach on other systems: for instance, what criteria should be used to identify a residue with abnormal intensities ? A quantitative description of the deviating between the theoretical model and the experimental values is clearly missing here.Some of the hypothesis made by the author are questionable. In particular, the assumption that intrinsic exchange rates are identical for all amide protons is probably not true since local chemical environment at the protein surface do modulate these exchanges with water. Such information may be obtained by simple 2D experiments such as the Het-SOFAST proposed by Paul Shanda and Bernhard Brutscher.
Relaxation mechanisms different from the dipole-dipole interaction may also contribute for the amide transverse relaxation: can we fully discard scalar relaxation ?
Small points:
- Equation 1 contains some mistakes:
I guess the two factors in the denominator are line-width (delta_nu) and not frequencies (delta missing)
The term describing the nitrogen relaxation doesn't make sense to me: why is the average <t1> considered ?
The amplitude of the peak on the F1 (15N) dimension depends on the magnetisation's level at the end of the t1 increment.-140 Please describe how the amide value of CSA is derived ? Is it reasonable to assume the same CSA for all amides ? could this not be one major reason of the observed deviations, since the amide may be engaged within a hydrogen bond modulating the distribution of electrons.
- 145 The author should mention the model they used to derive the linewidths from Sparky (Lorentzian fitting ? gaussian ?)
- Figure 4: There are some discrepancies between the text and the figure:
- orange points are below 11 Hz for Reduced Experimental Line width
- What is meant by "at opposite side of the diagonal ?" I suggest using "Upper triangle" and lower triangle regions- Figure 6:
- Legend : plain circle and squares- Figure 9 : Labelling the different domains of Hsc70 would be helpful to follow the dynamical description.
Citation: https://doi.org/10.5194/mr-2020-38-RC3 -
AC4: 'Reply on RC3', Erik Zuiderweg, 26 Jan 2021
The comment was uploaded in the form of a supplement: https://mr.copernicus.org/preprints/mr-2020-38/mr-2020-38-AC4-supplement.pdf
-
AC4: 'Reply on RC3', Erik Zuiderweg, 26 Jan 2021
-
CEC1: 'Comment on mr-2020-38', Gottfried Otting, 27 Jan 2021
It seems to me that this discussion deserves more careful consideration.
“as far as I can see, Abragam actually never defines what limits there are what one may call “like” spins.”
Abragam’s book, page 290: if the spins I and S are like spins, only the sum <I> + <S> is observed, whereas <I> and <S> are observed separately if the spins are unlike.
Following this definition, the resonances of amide protons in liquids are observed separately (their resonances usually don’t overlap), which makes them unlike spins. Of course, the situation is different in solids.
“I disagree with esteemed Prof. Axel Bothner-By”
To the best of my knowledge, the theory underpinning ROESY has not been questioned previously.
Regarding the discussion of the PEP scheme, more detail would be welcome: for a two-spin system, the PEP scheme transfers every term present at the end of the evolution time t1 into a term that is observable during acquisition (i.e. t2). Not all of these terms relax as transverse proton magnetization during both the delays b – c and d – e. For example, transverse relaxation tends to be much slower for 15N than 1H.
Citation: https://doi.org/10.5194/mr-2020-38-CEC1
Interactive discussion
Status: closed
-
RC1: 'Comment on mr-2020-38', Anonymous Referee #1, 20 Jan 2021
The article tries, and fails, to correlate measured line widths with line widths calculated from crystal structures, assuming dipolar relaxation as the dominant relaxation mechanism.
The ultimate conclusion is that amides with narrower line widths than average are mobile, whereas amides with broader line widths reflect us – ms dynamics. This is not at all new and the conclusion in line 215 is trivial.
Using M = 3 in equation 1 is inappropriate for a sensitivity-enhanced HSQC, as magnetization is not transverse throughout all delays of the PEP scheme.
Equation 4 applies to indistinguishable spins. It does not apply to amide protons with distinguishable chemical shifts (see Bothner-By et al., JACS 106, 811 (1984)
or equations 79 and 89 in Abragam, Chapter VIII).
As admitted by the author, Figure 4 shows no correlation. Figure 6 doesn’t show a significant correlation either, again as admitted by the author. This indicates failure of the entire concept. If coordinate precision is a major problem (line 222), the concept cannot be salvaged.
Line 167: The line width of the amide proton of Asp3 could very well be due to enhanced H-exchange, as the N-terminus is positively charged and the exchange base-catalyzed.
Line 187: Much more is known about the dynamics in BPTI than suggested by the manuscript, see, e.g., Grey et al., JACS 125, 14324 (2003).
Line 225: water molecules on protein surfaces are not known to contribute to dipolar relaxation.
Line 235: detailed dynamical data have been obtained for large proteins from methyl relaxation.
Line 280: independent movement of a domain relative to the rest of the protein produces narrower lines. This is not novel: there are countless examples, such as calmodulin, trigger factor etc.
Figure 3 does not specify the acquisition times used in the spectrum - too short acquisition times would cause artificial line broadening to a variable degree for amide protons with different intrinsic line width.
Legend of Fig. 1: EM and cosine not explained
Line 57: equal to what?
Line 94: us is not a unit.
Line 155: what is meant by flattered comparison?
Line 269: what is meant by ATP conformation?
In the opinion of this reviewer, the manuscript falls far short of the criteria of Magnetic Resonance in terms of scientific impact (absence of a clear advance) or scientific quality (wrong equations used, data mostly not provided and, hence, the conclusions of the manuscript cannot be reproduced independently). It would be unexpected if it were to pass the review process of any NMR journal.
Citation: https://doi.org/10.5194/mr-2020-38-RC1 -
AC2: 'Reply on RC1', Erik Zuiderweg, 21 Jan 2021
The comment was uploaded in the form of a supplement: https://mr.copernicus.org/preprints/mr-2020-38/mr-2020-38-AC2-supplement.pdf
-
AC2: 'Reply on RC1', Erik Zuiderweg, 21 Jan 2021
- AC1: 'Comment on mr-2020-38', Erik Zuiderweg, 21 Jan 2021
-
RC2: 'Comment on mr-2020-38', Anonymous Referee #2, 22 Jan 2021
This is a largely exploratory manuscript investigating the potential to obtain (semi-)quantitative results on protein dynamics based on backbone amide 1H linewidths in 1H-15N HSQC or TROSY spectra. As Zuiderweg states himself, the underlying idea to glean dynamic information from the HSQC spectrum is commonly used by NMR spectroscopists in a qualitative sense by identifying particularly sharp or broadened cross-peaks. Here, Zuiderweg provides a more comprehensive analysis by attempting to account for all contributions to the 1HN linewidth from various relaxation mechanisms and other effects, such as unresolved J-couplings and B0 inhomogeneity. Naturally, the dominant part of the relaxation is due to dipole-dipole relaxation, which is estimated for each backbone amide based on the high-resolution crystal structure of the protein in question.
The end result is that the calculated linewidths show only fair agreement with the experimental ones, but outliers appear to reliably identify backbone amides that are either undergoing large-amplitude fast internal dynamics (i.e., residues with low order parameters) or residues undergoing conformational exchange. Thus, we are left with the conclusion that 1HN linewidths cannot provide more detailed information than what is customarily obtained from a qualitative, first-glance interpretation of HSQC-type spectra. To this extent, the work clearly does not advance the field since it does not provide any substantial conclusions beyond current knowledge. Still, I appreciate the comprehensive, semi-quantiative analysis offered by Zuiderweg, which clearly shows the limitations of the proposed analysis. In essence, the work demonstrates that 1HN linewidths cannot be interpreted in terms of dynamics to any detailed extent. For these reasons, I believe that the study could be worth publishing.
Minor points:
p. 4, Eq [1]: I do not follow this equation fully: the last two factors are not defined and it is not clear why they appear in the equation. The first exponential assumes that each step in the reverse polarization transfer (after t1) contributes equally to the linewidth, but this is not true for all pulse sequences (it depends on the details of the PEP scheme, etc).
l. 94: aromatic ring flipping does not cause exchange linebroadening of amide protons (since the end states are identical).
Table 1: Please clarify what is listed in this table. By comparing with Fig. 4, I assume that the Sum of 1HN linewidths is taken over all residue pairs in the protein(?). This should be stated in the Table header (or footnote).
Figure 6 legend: "open diamonds" should be 'open squares'.
The text should be checked for typos, incorrect order of words, missing words, etc.
Citation: https://doi.org/10.5194/mr-2020-38-RC2 -
AC3: 'Reply on RC2', Erik Zuiderweg, 26 Jan 2021
The comment was uploaded in the form of a supplement: https://mr.copernicus.org/preprints/mr-2020-38/mr-2020-38-AC3-supplement.pdf
-
AC3: 'Reply on RC2', Erik Zuiderweg, 26 Jan 2021
-
RC3: 'Comment on mr-2020-38', Anonymous Referee #3, 23 Jan 2021
The manuscript proposed by Erick Zuiderweg presents a valuable attempt to rationalize the intensities measured on protein 1H-15N correlation spectra in order to get a qualitative description of the backbone dynamics at different timescales. The different parameters affecting the 2D correlation peak's intensities are exhaustively listed and their values estimated from a structural model of the protein. The concept is tested on the BPTI protein and, as noted by the author, modelling the different known contributions to the signal intensities fails to reproduce the experimental measurements. The relevance of the approach is defended by the observation that large deviations from the modelled intensities do cluster in regions of BPTI where specific dynamical features where previously reported from 15N relaxation measurements. A graphic based clustering of these "model deviating" signals is therefore proposed as a fast approach to get qualitative insights on the protein dynamics. The approach is applied to a large protein (Hsc70) enabling some observations to be made on its dynamical properties.
As stated by the author in his introduction, such an approach would be very valuable in the field of protein NMR, as we do share the general feeling that the information content of 15N-HSQC or TROSY are under-exploited, and the author's attempt to address this task is interesting. However, my general opinion is that the current state of this development is far too preliminary and would desserve more work to be published. The general applicability of such a method should be assessed by probing the concept on different class of proteins that display distinct and well documented dynamical features (depending on their size, geometry, experimental conditions of the study T° pH ...). The "correlation" approach is indeed the only way to go when the model fails to reproduce the experiment. Anyhow, some statistical assessment is necessary for the applicability of the proposed approach on other systems: for instance, what criteria should be used to identify a residue with abnormal intensities ? A quantitative description of the deviating between the theoretical model and the experimental values is clearly missing here.Some of the hypothesis made by the author are questionable. In particular, the assumption that intrinsic exchange rates are identical for all amide protons is probably not true since local chemical environment at the protein surface do modulate these exchanges with water. Such information may be obtained by simple 2D experiments such as the Het-SOFAST proposed by Paul Shanda and Bernhard Brutscher.
Relaxation mechanisms different from the dipole-dipole interaction may also contribute for the amide transverse relaxation: can we fully discard scalar relaxation ?
Small points:
- Equation 1 contains some mistakes:
I guess the two factors in the denominator are line-width (delta_nu) and not frequencies (delta missing)
The term describing the nitrogen relaxation doesn't make sense to me: why is the average <t1> considered ?
The amplitude of the peak on the F1 (15N) dimension depends on the magnetisation's level at the end of the t1 increment.-140 Please describe how the amide value of CSA is derived ? Is it reasonable to assume the same CSA for all amides ? could this not be one major reason of the observed deviations, since the amide may be engaged within a hydrogen bond modulating the distribution of electrons.
- 145 The author should mention the model they used to derive the linewidths from Sparky (Lorentzian fitting ? gaussian ?)
- Figure 4: There are some discrepancies between the text and the figure:
- orange points are below 11 Hz for Reduced Experimental Line width
- What is meant by "at opposite side of the diagonal ?" I suggest using "Upper triangle" and lower triangle regions- Figure 6:
- Legend : plain circle and squares- Figure 9 : Labelling the different domains of Hsc70 would be helpful to follow the dynamical description.
Citation: https://doi.org/10.5194/mr-2020-38-RC3 -
AC4: 'Reply on RC3', Erik Zuiderweg, 26 Jan 2021
The comment was uploaded in the form of a supplement: https://mr.copernicus.org/preprints/mr-2020-38/mr-2020-38-AC4-supplement.pdf
-
AC4: 'Reply on RC3', Erik Zuiderweg, 26 Jan 2021
-
CEC1: 'Comment on mr-2020-38', Gottfried Otting, 27 Jan 2021
It seems to me that this discussion deserves more careful consideration.
“as far as I can see, Abragam actually never defines what limits there are what one may call “like” spins.”
Abragam’s book, page 290: if the spins I and S are like spins, only the sum <I> + <S> is observed, whereas <I> and <S> are observed separately if the spins are unlike.
Following this definition, the resonances of amide protons in liquids are observed separately (their resonances usually don’t overlap), which makes them unlike spins. Of course, the situation is different in solids.
“I disagree with esteemed Prof. Axel Bothner-By”
To the best of my knowledge, the theory underpinning ROESY has not been questioned previously.
Regarding the discussion of the PEP scheme, more detail would be welcome: for a two-spin system, the PEP scheme transfers every term present at the end of the evolution time t1 into a term that is observable during acquisition (i.e. t2). Not all of these terms relax as transverse proton magnetization during both the delays b – c and d – e. For example, transverse relaxation tends to be much slower for 15N than 1H.
Citation: https://doi.org/10.5194/mr-2020-38-CEC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 824 | 708 | 109 | 1,641 | 75 | 59 |
- HTML: 824
- PDF: 708
- XML: 109
- Total: 1,641
- BibTeX: 75
- EndNote: 59
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1