Thank you for the thought-provoking comments! The supplement of this author response explains why we think that the through-space scalar ¹⁹F-¹⁹F couplings observed in our work come about by direct contacts rather than some indirect mechanism.

Thank you for the careful review!

1. I’m a bit confused about the gamma gauche effect when talking about two monofluorinated methyl groups in Leu. Both 19F nuclei can simultaneously be trans relative to Cgamma, so it’s not clear that this helps with stereo assignment, even though the correlation between JHF and 19F shift clearly shows it to be correct.  Is it conceivable that Calpha_Cdelta gamma gauche effects contribute to the 19F chemical shift (as they do for 13Cdelta)?

Response: Marking the chemical shifts of the two ¹⁹F spins of a diFLeu residue as “high-field” and “low-field”, respectively, this relative order is also observed for the singly fluorinated residues FLeu1 and FLeu2 located at the same site. At least, this has been so for the two examples we have, GB1-d and PpiB-d. To explain this, there is not only the γ-gauche effect, but also the local chemical environment. (If both ¹⁹F nuclei were simultaneously trans relative to Cγ, the γ-gauche effect would be the same but their chemical environments would probably still be different.) It is not self-evident, however, that CH₂F groups populate the same rotamers in monofluoro- and difluoro-leucine residues, especially as the conformations with two parallel C-F bonds are energetically disadvantaged (Marstokk and Møllendal, 1997). We agree that the preservation of high-field and low-field shifts is a purely empirical observation for which exceptions are bound to be discovered in the future.

The Cα-Cδ γ-gauche effect implies that the 3-bond Cα-Cδ coupling is proportional to the ¹³C-chemical shift of Cδ (https://doi.org/10.1007/BF00202043). (Thank you for pointing out the effect – it made us note that Figure S5 had mislabelled the cross-peaks of Leu12 in wild-type GB1. It means that the Cα-Cδ γ-gauche effect is maintained for the FLeu CH₃ groups in GB1-1 and GB1-2.) Further transmission of the Cα-Cδ γ-gauche effect onto the ¹⁹F spin is conceivable but difficult to document because it would be a second-order effect much less prominent than the direct γ-gauche effect with fluorine or the influence of the immediate chemical environment. Fluorination already seems to upend the Cα-Cδ γ-gauche effect for the CH₂F groups in GB1-1 and GB1-2: Figure S6a and b shows that the ¹³C chemical shift of the C^δ1H₂F group of residue 12 is greater than that of the C^δ2H₂F group, although (according to the solution structure 3GB1) the C^δ2H₂F group is trans to the α-carbon in the wild-type protein.

1. For the HOESY measurements, I suspect the NOE effect to be very sensitive to internal motion due to the closeness of wH and wF.  Heteronuclear 1H-19F NOE can be negative or positive, depending on applicable spectral densities and some comments may be helpful.

Response: For ¹H-¹H NOEs and a rigid 2-spin model, the sign changes for a rotational correlation time τc = sqrt(5)/2ω_H, which is 0.44 ns on a 400 MHz NMR spectrometer (ω_H denoting the Larmor frequency of ¹H). Although ω_H and ω_F are not the same, the ¹H-¹⁹F NOE changes sign at a very similar correlation time (τc = 0.46 ns). Internal motions are far less prominent for CH₂F than CH₃ groups (which are famous for rotating even at zero Kelvin by quantum-mechanical tunnelling). For the CH₂F group in 1-fluoropropane the intrinsic energy barrier between staggered rotamers has been calculated to be 4-5 kcal/mol (Feeney et al., 1996), which is comparable to the energy barrier for rotation about the central C-C bond in n-butane. Therefore, the H-F substitution is an efficient way to decrease the rate of exchange between the three staggered rotamers of a methyl group (arguably more so than, e.g., for the CH₃ group of a methionine residue). Experimentally, we did not observe negative ¹H-¹⁹F cross-peaks with the CH₂F groups, apart from NOEs with water. We’ll happily add a sentence or two about this.

1. 19F line widths are reported to be 7-15 Hz, which corresponds to R2 values of 20-40/s, but this seems fast considering the 60-ms TOCSY mixing times used.  Could incomplete decoupling or isotope effects contribute to these line widths?  Perhaps a R1rho number would be helpful.

Response: We now measured R_1ρ values. They are indeed considerably smaller than the R₂* relaxation rates determined from the NMR line widths, see the table provided in the Supplement. R_1ρ values are even slower for deuterated diFLeu residues. For concise notation, diFLeu residues, where all five protons of the isopropyl group are replaced by deuterium, are referred to in the following as diFLeu-D and the GB1-d sample made with diFLeu-D as GB1-d-D.

The slower R_1ρ (¹⁹F) relaxation rates of diFLeu-D residues versus diFLeu residues demonstrates the impact of the nearest protons on ¹⁹F relaxation. The effect correlates with solvent exposure: it is greatest for the most buried residue 5, still significant for the next most-buried residue 7 and the buried fluorine of residue 9, and less for the more solvent-exposed ¹⁹F spins. We conclude that dipolar relaxation matters particularly for buried and immobilised CH₂F groups. Applying the CPMG sequence with a lower frequency increases the relaxation rates (especially for residue 7), indicating a chemical exchange contribution. Unfortunately, we have no spectrometer to repeat the CPMG experiment at a different magnetic field strength, i.e., we cannot determine quantitative chemical exchange rates from relaxation dispersion experiments. Notably, the rate of 180 degree pulses in the CPMG sequence has little effect on residue 5, which shows the broadest lines and the largest and smallest ³J_HF coupling constants and ¹⁹F chemical shifts. We conclude that the CH₂F groups of this residue are largely immobilised and their line widths are predominantly governed by relaxation rather than exchange broadening.

The R₂* values derived from the line widths observed in the 1D ¹⁹F-NMR spectrum are greater than expected based on the R_1ρ data, regardless of whether we use GARP or WALTZ decoupling at 2800 Hz.

We tried obtaining narrower ¹⁹F NMR signals by ²H decoupling of GB1-d-D, but the line widths of most of the diFLeu-D residues were broader than for diFLeu residues in GB1-d measured with ¹H decoupling. Maybe this is due to unresolved 4-bond couplings with β-protons (we are not equipped to decouple ¹H and ²H simultaneously), but we never saw direct cross-peaks between ¹⁹F and β-protons. Heterogeneities due to incomplete deuteration (90% at the branch point of the isopropyl group) might contribute too, but the importance of this is unclear to us. Certainly, isotope effects have a large impact on ¹⁹F chemical shifts. As the solvent contained 90% H₂O/10% D₂O, deuterated amides could also contribute to isotope effects. If so, shouldn’t we expect the resulting line shapes to be unsymmetric? We do not observe this but will follow up with a measurement in the presence of less D₂O.

1. Line 146-147: “faster rotation of the CH2F group about the Cgamma-Cdelta bond results in slower transverse relaxation”.  This may well be true, but the magnitude of this effect seems larger than expected considering the modest chemical shift difference of ~10ppm.  Could crank-shaft sidechain motions, previously suggested to be responsible for different Cdelta 13C relaxation rates, play a role?

Response: We assume (but do not know for sure) that the correlation time governing dipolar ¹⁹F relaxation as well as CSA relaxation is mostly determined by rotation about the Cγ-Cδ bond, because this conformational change requires minimal structural adjustments of the surrounding protein environment. X-ray structures of PpiB with fluorinated leucine residues show that CH₂F groups populate staggered rotamers but populating single rotamers is the exception, not the rule. We imagine that jumps between rotamers can be fast when solvent-exposed, just like for lysine side chains, but cannot exclude smaller amplitude motions within the energy well of a single rotamer. We find it hard to picture a crankshaft motion about the Cγ-Cδ bond that would be equally benign in terms of structural conservation as a simple rotation of the CH₂F group, at least for motions changing the conformation by a similar angle. Our sentence suggests, rather than claims, that the line width is governed by the speed of rotation of the CH₂F groups.

1. Line 199: “different” from what?  Perhaps use “multiple”?

Response: Agreed, multiple is the better term.

1. Line 201: “greater conformational freedom than suggested by 3GB1”. This is a bit of a philosophical issue, but the width of an “NMR bundle” does not reflect motional freedom but the certainty at which the structure that agrees best with NMR restraints can be determined.  If not, measuring fewer restraints would make the protein more dynamic.

Response: Fair point, as the NMR structure was determined with the aim of presenting the best approximation to the average structure rather than a conformational ensemble. We’ll point this out in the revised version.

1. Line 322: “the third example of the gamma-gauche effect”. Probably correct, but perhaps useful to remind the reader of how commonly this is used in 13C analysis, including proteins (e.g. https://link.springer.com/article/10.1007/BF00202043 )

Response: Thank you for pointing out the reference. Our wording needs to be more accurate and only refer to the γ-gauche effect with regard to CH₂F groups.

1. There is a considerable amount of older literature on TS-JFF couplings, with an empirically determined very steep distance dependence. See e.g. Bakhmutov and references therein.  Perhaps including some reference to this historic work would be helpful, e.g. https://doi.org/10.1002/mrc.1260231117

Response: Through-space couplings go back to the 1960s and are described in NMR textbooks (e.g. H. Günther, NMR-Spektroskopie, Georg Thieme Verlag Stuttgart 1983). We prefer referring to Ernst, L.; Ibrom, K. Angew. Chem., Int. Ed. Engl. 1995, 34, 1881 and Mallory, F. B. et al. J. Am. Chem. Soc. 2000, 122, 4108 for establishing plausible exponential distance dependences. We’ll happily cite those articles.

1. Lines 358-362, c=chi; d=delta

Response: thank you for pointing out the typos.

1. Please include the RF field strength and mixing scheme (DIPSI?) used for the TOCSY spectrum.

Response: yes, DIPSI-2 mixing, using 4200 Hz RF field strength.

1. Can the authors provide approximate TS-JHH values based on the cross/diagonal peak ratios?

Response: we used the GB1-d-D sample to measure the ratio of cross-peaks versus diagonal peaks in TOCSY experiments recorded with increasing mixing times. The figure in the Supplement shows the data for three different cross-peaks, which are prominent in the FF-TOCSY spectrum. Using the approximation of a 2-spin system, the fits indicate through-space J_FF couplings of the order of up to 2-3 Hz. We’ll include the data in the revised manuscript.

1. Trivialities: Spell out CFPS upon first use; Juszewski is really Kuszewski (3 times).

Response: We’ll spell out CFPS as cell-free protein synthesis. Thank you for pointing out the embarrassing misspelling of John Kuszewski’s name.

Thank you for the careful assessment. We propose the following amendments.

1. The authors observe ‘through-space’ J-couplings between ¹⁹F^δ atoms of the leucines, which are supposedly in spatial contact. Due to the Fermi contact mechanism, the J-couplings show that the respective s-electrons are correlated in their motion. How do the authors imagine this electronic polarization transfer? Given that the fluorines have strong negative partial charges, would they interact directly in an attractive or repulsive way? Could it also be that the hydrogens in the CFH₂ group acquire partial positive charges and are then attracted to the fluorine of the neighboring leucine, such that the transfer is achieved as F…H-C-F? Maybe this is unlikely, since Teflon is very hydrophobic. An educated discussion would be very helpful.

Response: The quantum mechanical underpinning of through-space ¹⁹F-¹⁹F J-couplings was clarified decades ago. It is complicated and not the topic of the present work. Scalar ¹⁹F-¹⁹F couplings via through-space F…H interactions appear unlikely as we did not observe any through-space couplings between ¹⁹F and ¹H. Furthermore, C-F groups are known not to be good H-bond acceptors, the protons of CH₂F groups are barely acidic and we would expect that a “three-bond” interaction is weaker than a direct interaction. Teflon presents a unique situation because the net dipole moments in Teflon, like in CF₃ groups, are reduced by the multitude of C-F bonds pointing in different directions. Regarding through-space couplings, we propose to direct the reader to the comprehensive 2013 review by Hierso in Chem. Rev.

1. Can the authors estimate the size of the J_FF-couplings from the intensities of the TOCSY-spectra?

Response: we have done this. See our response to Ad Bax’s review.

1. The spectra in Figure 3 were recorded at different pH and buffer for the different GB1 analogues (pH 6.5 MES vs pH7.5 HEPES). Is there a reason? How much would the pH affect the chemical shifts? This should be discussed in the text. Also, the Figures of other spectra do not mention the pH and buffer conditions. They should be indicated throughout.

Response: Accidentally, different buffers were used for different preparations. The ¹⁹F-NMR spectra did not vary significantly with the buffer (see supplement to this response which shows the ¹⁹F-NMR spectra of two different GB1-d preparations, one in MES buffer pH 6.5, the other in HEPES buffer pH 7.5). In the spectra of the manuscript, the buffer used for GB1-1 and GB1-2 was HEPES, pH 7.5, whereas the buffer of GB1-d was MES, pH 6.5. This will be clarified in the revised version.

1. The authors mention ¹⁹F T₁ relaxation times of 0.3 s (line 121). Can they estimate the T₂s? They indicate line widths of 7–15 Hz. Would this be 20–50 ms? Do they have information from spin-echos?

Response: we measured the transverse relaxation times. See our response to Ad Bax’s review.

1. Figure 9: a plot of the ³J_HF constants vs. the ¹⁹F shifts would be very helpful to visualize and quantify the γ-gauche effect.

Response: We published such a plot in our Biochemistry article of PpiB made with fluoroleucines. In GB1, the fluoroleucine residues are in contact with each other and, as shown in the present manuscript, the ¹⁹F chemical shifts change greatly depending on whether the neighbouring leucine residue is fluorinated or not. Therefore, the chemical environment obscures the correlation between ³J_HF couplings and ¹⁹F chemical shifts. Table S3 presents a clear overview of the correlation and the existence of outliers.

1. Apparently the GB1 construct contains an additional MASMGT sequence at the N-terminus. Although no effect is expected, it would be good to mention this in the main text.

Response: will do in the revised version

1. Figure S3: CD melting. Is there a reason for the larger (absolute) ellipticity of wild-type GB1? Please mention/discuss.

Response: Temperature denaturation curves are often reported with the vertical axis labelled from 0 (folded) to 1 (unfolded). We have chosen to present the original data, which are affected by somewhat different sample concentrations. We’ll correct the figure legend to state that the concentrations were approximately 0.3 mg/mL. The ¹H NMR data leave no doubt about the structural integrity of the fluorinated samples.

1. Figure S5: is this a natural-abundance ¹³C HSQC? This should be mentioned in the main text.

Response: yes, natural abundance. This will be clarified in the revised version in the figure legends of Figures S5 and S6.

Presentation:

The accessibility of the manuscript could be improved by considering the following:

1. The nomenclature is hard to follow. (2S,4S)-5-fluoroleucine, (2S,4R)-5-fluoroleucine and 5,5’-difluoro-L-leucine are FLeu1, FLeu2 and diFLeu. Apparently FLeu1, FLeu2 and diFLeu are leucines labeled in the δ1, δ2 or δ1+δ2 respectively (if I am not mistaken). This should be clearly indicated in the drawing of Figure 1 (not in the legend). A standard IUPAC NMR nomenclature of all leucine atoms would be also very helpful in this Figure. Could FLeu1, FLeu2 and diFLeu be replaced by something like F^δ1Leu, F^δ2Leu, F^δ1δ2Leu throughout the text?

Response: We’ll change Figure 1 to show the δ₁ and δ2 labels. The distinction between the different isomers is important and, in general, we prefer not to relegate this information to a small superscript.

1. Throughout the figures: the authors are very terse with labeling in the figure graphics. While this emphasizes the data, it makes them hard to grasp. In particular:

2.1. Figure 2: indicating the δ1, δ2 color code directly in the graphic would be very helpful.

2.1. All 1D spectra: the stereospecific assignments would be better indicated by δ1, δ2 instead of a red dot only for δ1.

2.3. All 2D spectra: individual peaks should be labeled with 2D, stereospecific assignment information whenever possible.

2.5. Figure S5: please provide color code in graphic.

Response: We have thought carefully about the labelling and striking the right balance between labelling and cluttering.

We’ll provide a colour code in Figure 2.

We used red dots instead of δ₁ labels to make it graphically obvious that the ¹⁹F chemical shift of the δ₁ group can be both high-field and low-field of the δ₂ signal. In addition, some of the red dots come close to each other, leaving little space for labelling with characters.

Labelling every cross-peak in the 2D spectra would lead to, we think, clutter that is more distracting than helpful.

Except for Figure 9, which is already rich in labels, we plotted the 1D ¹⁹F spectrum along the horizontal axis of each 2D spectrum together with the assignments. In Figure 6, where it matters most, we also included the stereospecific assignment information.

We’ll label the spectra of Figure S5 to clarify their belonging to GB1-1 and GB1-2, respectively.

1. line 286: please indicate the size of the ‘different 3JHF for residue 5’.

Response: We’ll quote the couplings and refer to Table S3.

1. lines 358–360: the Greek characters were lost.

Response: Thank you for alerting us to the typos.