Assessing Site-Specific Enhancements Imparted by Hyperpolarized Water in Folded and Unfolded Proteins by 2D HMQC NMR

Hyperpolarized water can be a valuable aid in protein NMR, leading to amide group 1H polarizations that are orders of magnitude larger than their thermal counterparts. Suitable procedures can exploit this to deliver 2D 1H–15N correlations with good resolution and enhanced sensitivity. These enhancements depend on the exchange rates between the amides and the water, thereby yielding diagnostic information about solvent accessibility. This study applied this “HyperW” method to four proteins exhibiting a gamut of exchange behaviors: PhoA(350–471), an unfolded 122-residue fragment; barstar, a fully folded ribonuclease inhibitor; R17, a 13.3 kDa system possessing folded and unfolded forms under slow interconversion; and drkN SH3, a protein domain whose folded and unfolded forms interchange rapidly and with temperature-dependent population ratios. For PhoA4(350–471) HyperW sensitivity enhancements were ≥300×, as expected for an unfolded protein sequence. Though fully folded, barstar also exhibited substantial enhancements; these, however, were not uniform and, according to CLEANEX experiments, reflected the solvent-exposed residues. R17 showed the expected superposition of ≥100-fold enhancements for its unfolded form, coexisting with more modest enhancements for their folded counterparts. Unexpected, however, was the behavior of drkN SH3, for which HyperW enhanced the unfolded but, surprisingly, enhanced even more certain folded protein sites. These preferential enhancements were repeatedly and reproducibly observed. A number of explanations—including three-site exchange magnetization transfers between water and the unfolded and folded states; cross-correlated relaxation processes from hyperpolarized “structural” waters and labile side-chain protons; and the possibility that faster solvent exchange rates characterize certain folded sites over their unfolded counterparts—are considered to account for them.

NMR Spectroscopy. Sample temperature is an important aspect for the claims made in this study. In principle temperatures should be quite stable and uniform throughout the injections: both the protein solution waiting inside the NMR and the hyperpolarized water injected into the NMR are comixed and maintained at the target temperature -the protein pre-and post-injection solutions as regulated in-situ by the NMR console's VT system, and the hyperpolarized water as controlled by a heating-tape-based system. As long as the water is injected above ≈35 ˚C (the temperature at which it arrives from the Hypersense/Arduino system), this precalibrated approach should give a solution whose temperature matches that of the protein and hence is constant through space and time. Figure S1. Examining the reliability of HyperW's 2D NMR thermal stability experiments at 50 ˚C. (A, B) Central peak positions observed along the 1 H (F2) dimension of a 50 ˚C HyperW 2D acquisition, for residue A39 and for the indole W36 peak of the drkN SH3 (unfolded-form) protein -both of which relatively isolated (Fig. 7) and hence can be followed by 1D 1 H projections. The temperature-dependence of these peaks was independently measured by 2D HMQC NMR, leading to the actual temperatures indicated on the left-hand axes. Shown for completion (dashed lines) are the chemical shifts measured in the 2D HyperW HMQC spectrum for these residues, leading to ≈49.25±0.25 ˚C as the representative temperature of this experiment. (C) 1D traces leading to the results illustrated in panel (B), illustrating both the maxima but also the line shape changes undergone by the W36 indole (U) peak for two postinjection times. The dashed lines indicate the chemical shifts measured for the same peak by conventional 2D HMQC NMR at 47 and 50 ˚C. Notice as well the significant and clearly larger enhancement shown by the folded (F) signal of this residue, as well as its ensuing line narrowing with time.
Still, driven by the importance that temperatures have in defining folded/unfolded protein ratios, particularly for drkN SH3's 50 ˚C dissolution experiments, ancillary analyses were made. Panels A and B in Figure S1 summarize these, by plotting the central peak positions observed along the 1 H (F2) dimension of a 50 ˚C HyperW 2D acquisition, for two residues of the drkN SH3 protein that are both relatively isolated (and hence can be followed in the 1D projection) and show a relatively strong temperature dependence in their 1 H chemical shifts. Independent measurements collected by conventional HMQC NMR in the 47-50 ˚C range, allowed us to translate these HyperW 1 H peak positions into sample temperatures as a function of post-dissolution time. These values are also shown in the left-hand axes of these figure's panels, and paint coinciding pictures regarding the system's stabilization following the injection. According to these, there is a certain drop in the sample temperature immediately upon injection of the pre-heated water (to ≈48 ˚C), but a nearly perfect thermal stabilization at the targeted 50 ˚C temperature ca. 15 s past the injection. These examinations can be extended by analyzing the central peak positions displayed by the residues along the 1 H (F2) dimension, following full 2D processing of the HyperW data. These Fourier-averaged positions should reveal the most representative temperature (and peak intensities) reflected in the 2D HMQC experiments, and they end up being between 49 and 49.5 ˚C (dashed lines in Figures S1A, S1B). Further insight into how samples in 2D HyperW HMQC experiments reach their final temperatures can be gathered from the line shape changes exhibited by the traces illustrated in Figure S1C, which focus on the unfolded SH3 W36 indole resonance -one of the thermally-sensitive peaks that were examined. When analyzed in the (t 1 ,F2)-domain for two different post-dissolution times t 1 and compared with traces collected for fully-thermallyequilibrated, H 2 O-dissolved samples, one can notice that peaks are broader at the beginning of the HyperW series (with a maximum at the equivalent of ca. 48˚C), most likely reflecting a thermal distribution within the NMR tube. Subsequently, peaks sharpen up and coincide with the unfolded chemical shift recorded on a thermally polarized, thermally stabilized sample at 50 ˚C. This sharpening leads to dominating peak positions and intensities corresponding to ca. 49.25±0.25 ˚C, when Fourier processing the full 2D HyperW HMQC interferogram. In addition, notice how the 1D 17s post-injection trace highlights the orders-of-magnitude enhancements that the folded resonance in this residue gains from the hyperpolarized water over its unfolded counterpart -which of course is one of the paper's main findings.
The two dimensional spectra were acquired using a 2D HyperW 1 H-15 N HMQC sequence, similar to what was used in previous work (Fig. S2). 1-2 It is a 2D HMQC-based sequence, using a solute-specific Figure S2. Pulse sequence for the 2D HyperW 1 H-15 N HMQC used in this study. Full bars represent 90° hard pulses, and shapes represent amide-selective 90° and 180° pulses. The recycle delay d 1 was typically set to 0.037-0.1 s; water polarization was achieved during t DNP ≈120-180 min, and the subsequent dissolution and injection of hyperpolarized water occurred during t injection ≈2-3 s. Selective excitation of the amide protons was achieved using a PC9 polychromatic pulse, 5 refocusing with a REBURP pulse 6 centered at 8.5 ppm with a 3.0 ppm bandwidth, and typically N 1 =128 increments were collected. The sequence employed the indicated phase-cycling of the 15 N excitation and storage pulses, to reduce the water background and to deliver by a hypercomplex acquisition purely absorptive lineshapes. [3][4] Decoupling on the 15 N channel was done using GARP modulation 7 during the acquisition. excitation approach. [3][4] The amide downfield region is excited using a selective 90° pulse in order to maximize the use of the hyperpolarized exchanged sites while minimizing water depolarization.
PhoA (350-471) : Per protio and SSP results. It is of interest to compare the HyperW HMQC spectrum (Fig. 1, red) not only to a thermal equilibrium spectrum measured on the same postdissolution sample (Fig. 1, blue), but also against a thermal spectrum measured in an 82.5% H 2 O buffer under otherwise same conditions. A comparison between these data (Fig. S3) and Fig. 1 reveals that indeed with 82.5% water, one could observe peaks that are broadened beyond detection in the thermal post-dissolution sample, albeit with very poor sensitivity. Peaks are observed with a better sensitivity in the HyperW spectrum, thanks to a nearly ~500× enhancement. Unlike what had been previously reported for -synuclein, the sensitivity enhancements evidenced by PhoA's HyperW HMQC, do not appear to correlate with the electrostatic charges in the protein sequence ( Fig. 2). To explore potential correlations between the enhancements and the SSP values, Supporting Figure S4 compares both individual enhancements vs SSP scores, as well as the running-average enhancements for every three consecutive residues against the SSP average score of the same three residues. A modest correlation appears to emerge in the latter, but it is hard to ascertain. Figure S4. Correlations between the HyperW HMQC sensitivity enhancements calculated for resolved residues in the 15 N-labeled PhoA4 protein fragment, against SSP scores (absolute values) given in the literature 20 based on NMR 13 C α and 13 C β chemical shifts. (A) Average enhancements calculated for every three consecutive residues are compared against the absolute value of the average SSP score for the same three consecutive residues. (B) Idem but without the running average.
Variable temperature ZZ-exchange NMR on drkN SH3. ZZ-exchange is a kinetic experiment based on a 2D NMR 1 H-X chemical shift correlation, [9][10] with the addition of a mixing delay T during which the magnetization is stored along the z-axis while dynamics take place. The SH3 domain exists in two exchanging states (U and F), such that a given nucleus resonates at a frequency ω U in state U and ω F in state F. A 2D ZZ-exchange spectrum for this system (Fig. S5A) will thus contain two diagonal-peaks at U(ω U 13C , ω U 1H ) and N(ω F 13C , ω F 1H ) in the (F1, F2) frequency dimensions, and two cross-peaks at C1(ω F 13C , ω U 1H ) and C2(ω U 13C , ω F 1H ), due to the U⇋F exchange occurring during the mixing time. In order to obtain kinetic information, a series of ZZ-exchange spectra are recorded with a range of mixing delays. The dependence of the peak intensities on the mixing delay are then analyzed and fit to a kinetic exchange model (Fig. S5B): 11 (S1) and . R 1 U and R 1 F are the longitudinal relaxation rate → 22 = 1 + → 21 = -→ constants of magnetization in sites U and F, respectively, and I U (0) and I N (0) are the peak intensities in the unfolded and folded states, respectively, at T = 0. The factors A F and A U represent efficiency of coherence transfer after the mixing period T, and were determined as described previously. 12 The simultaneous fits to the data yield the first-order rate constants k FU = 31.0 ± 4 s -1 and k UF = 1.9 ± 0.4 s -1 . The populations at 50 °C are therefore: p U = 94.3% and p F = 5.7%; these ZZ-exchange results (Fig. S5), which take into account compensation for differences in the relaxation rates (R 1 and R 2 ) of the peaks. The folded state population of only 5.7% at 50 °C, is to be contrasted to the 55% observed at 27 °C. Populations and exchange rates at 37 °C were also calculated using a 1 H-15 N version of the ZZ-exchange experiment, as k FU,37C = 7.9 ± 2.8 s -1 , k UF,37C = 8.7 ± 2.9 s -1 , p U,37C = 48% and p F,37C = 52%. Table S1 summarizes these kinetic and population values, as derived by these measurements on SH3 at the three temperatures that we explored.  Methyl-TROSY NMR experiments on drkN SH3. The relative populations in Table S1 were derived from ZZ-exchange experiments performed on mostly deuterated (90% D 2 O) solutions. D 2 O as a solvent, however, has been reported to stabilize certain protein structures when compared to H 2 O, and to affect protein folding-unfolding kinetics. 21,22 The dDNP enhancements reported for drkN SH3, however, are done by comparing HyperW results arising from deuterated solutions, against thermal results arising from mostly protonated ones. It follows that in order to properly quantify the signal enhancement upon hyperpolarization, one must also take into account the potential differences in the populations observed for the folded states in D 2 O (the solvent used in the HyperW measurements) vs. in H 2 O (the solvent used for the thermal equilibrium measurements). As solvent exchanges prevent us from measuring the populations of drkN SH3's folded and unfolded forms by relying on the amide group resonances using D 2 O as solvent, a methyl labeled ( 13 CH 3 -ILVM, 2 H) drkN SH3 protein was expressed, and the populations of these two forms were quantified by integrating the peak intensities of ten residues in the folded and unfolded states using methyl-TROSY 1 H-13 C. 23 These experiments were carried out at 50 ˚C in both 90% H 2 O and in 100% D 2 O. The ensuing results are summarized in Table S2. As can be appreciated from these results, the populations of the folded states at 50 °C in D 2 O were indeed higher than those in H 2 O: 5.7 ± 0.6%.vs 4.0 ± 0.7%. At the same time, the deuterated solvent methyl-TROSY results were in excellent agreement with the ZZ-exchange measurements. These methyl-TROSY-derived populations were used in the simulations described throughout the Supporting and the Main texts; these populations were also used to rescale the kinetic rates in Table S1, as appropriate. Supporting Figure S6 compares CLEANEX-PM measurements at 37 °C, with the HyperW enhancements observed for the folded and unfolded residues of the drkN-SH3 domain. Supporting Figure S7 re-examines drkN SH3 HyperW's enhancements measured at 50 °C, assuming that post-injection temperatures were not as believed but instead lower by 3 ˚C -a difference that is still compatible with the peak positions recorded in the HMQC NMR spectra. Indeed, although Fig. S1 attests to the good thermal reliability of our setup, the linewidths of the HyperW NMR data yield a certain uncertainty in the temperature, which is bound by a lower limit of 47 ˚C. This plot is a recalculation of the enhancement data presented in Figure  8E, but with enhancements renormalized according to thermally polarized reference spectra measured at 47 ˚C. As evidenced by this plot, this still leads to a picture where folded-residue peaks are more enhanced by the hyperpolarized solvent than their unfolded-state counterparts. Table S3 summarizes the enhancements observed for the various folded and unfolded drkN SH3 residues at 50 ˚C, taking into account multiple dissolutions and the population considerations in Table S2. Comments indicate why the corresponding residues were not utilized in the paper's discourse/conclusions.  Figure S7. Same as Figure 8E in the main text, but assuming that the HyperW injection temperature had been misscalibrated and actually took place at 47 ˚C. Site-specific amide-water exchange rates lead to heterogeneities in the HyperW enhancement. However, the exchange rates in the folded state were expected to be slower relative to the unfolded state, due to protection factors and hydrogen bonds. Scheme 1 suggests that an additional magnetization transfer from an enhanced unfolded state residue to the same residue in the folded state can explain its observed sensitivity enhancements. Biases in the hyperpolarization of folded and unfolded residues could also arise from the different crossrelaxation behavior of these systems. To estimate how the HyperW signal enhancements will be affected by these exchanges, we computed the water and amide magnetizations for each conformation <H 2 O> z , <H N U > z , <H N F > z expected to arise in a process characterized by a forward reaction rate (proton transfers from H 2 O to H N ) k WU , k WF ; and a backward reaction rate k UW , k FW . These exchange rates are in fact related to each other by the water and protein molar fraction ratios X: A model based on the McConnell-Solomon equations [14][15][16] was implemented within a home-written Matlab ® (The Mathworks Inc.) code that involved numerical solution of the system of differential equations for different proton reservoirs, including chemical exchange and cross-relaxation between amide and aliphatic proton pools as well as with protons in the hyperpolarized H 2 O pool. This leads to the system of 7 differential equations: The relaxation matrix used in this model was generated using the Bloch-Redfield-Wangsness theory on a reduced spin system, 17 in combination with SpinDynamica. 18 Five spins were included to account for the polypeptide backbone and sidechain ( , , two aliphatic sidechain protons and , one labile sidechain protons ) and a reduced relaxation matrix with only longitudinal 1 2 terms for each spin present was utilized. A model-free approach with order parameters for each interaction was adopted, 19 with spectral densities given by the general form: The final relaxation matrix includes two spin order longitudinal terms of folded and unfolded conformations for the amide protons, two corresponding terms for the aliphatic protons, one for the sidechain labile proton, and one for the external water proton. Cross-relaxation between the amide and aliphatic spin pools were assumed to differ for the folded and unfolded states, given in each case by: ( 2 [ 2 (0) + 3 2 ( ) + 6 2 (2 )] Additional intrinsic relaxation rates 1/T 1 U,F were also added to the relaxation terms of each amide proton, in a search for an additional ad hoc parameter that might potentially explain drkN SH3's anomalous HyperW behavior. Order parameters and internuclear distances were chosen from the literature for Ubiquitin at room temperature, which is a fair assumption based on the very similar molecular weights of Ubiquitin and drkN-SH3 domain. The and rates will mostly depend on the internuclear amide/aliphatic distance (kept constant at 2.3 Å for the folded and unfolded states for simplicity) and on the internuclear correlation time , which was taken to be 3.4 ns for the folded state and 0.8 ns for the unfolded state of the protein. As purely intramolecular crossrelaxation models failed to predict larger folded than unfolded enhancements unless exchange rates k FW ≥k UF were invoked , Eq. (S6) was modified to enable the presence of intermolecular water-amide proton-proton cross-relaxation. This interaction was incorporated into the simulations in a manner similar to that in Eq. (S8); for simplicity, the same correlation times were assumed to control the intra-and inter-molecular cross-relaxation processes (3.4 ns for the folded, 0.8 ns for the unfolded H-H vectors). , in Eq. (S6) are the water and protein amide ⟨ 2 ⟩ ( ) ⟨ ⟩ ( ) and ⟨ ⟩ ( ) and aliphatic magnetizations at thermal equilibrium. Complementing Eq. (S6)'s time-dependence, the evolution of was artificially set to zero at t = n . TR (where TR is the experimental ⟨ ⟩ ( ) repetition time) to account for the depletion of protein magnetization arising due to the selective excitation pulses applied. Equation (S6) plus this reset condition were used for analyzing both the HyperW (Hyp) and the thermal equilibrium (TE) experiments that were carried out, which were recorded on the same samples under identical conditions -apart from their initial water polarization. The initial water magnetization was = ε . , where ε is the ⟨ 2 ⟩ (0) ⟨ 2 ⟩ ( ) enhancement factor over the thermal equilibrium polarization (ε = 200 for the Hyp experiment; and ε = 1 for the TE experiment). The initial polarization for the amide protons in the protein was assumed to be ; and were set equal to unity. For both cases (Hyp ⟨ ⟩ (0) = 0 ⟨ ⟩ (0) ⟨ ⟩ ( ) and TE) the equilibrium polarization was scaled according to the concentrations: and same holds for ; ⟨ ⟩ ( ) = ≡ ; ⟨ ⟩ ( ) = 1 aliphatic spin pools. In order to translate the magnetizations that will be predicted by these equations into observable signals, we further considered that in the full 2D HyperW 1 H-15 N HMQC experiment these will have to be converted into a 1 H coherence that transfers to and from the amide nitrogens: (S10).