Journal of the American Chemical Society d by simulations for such that th 975%fce,2.5% ed Inova spectro eld the both 11.7 CEST t .The All of th ng a mi ng time that th of 235 kH betwe 103.91nd133.5 able of e s-pea 50(0987 and paard to s tha f) PG cross-peaksc 0 here a sp upling du 23 f ang e s of (33 ms 64 m btaining ts each c -128 CEST A39G FF d a radi hundreds of mill onds leads to a ver o tion effe ub he 2nd0296 (S in lt.and unlik FID)and d13 33b dat (al typical"HSQC CEST 14 from 104.36 to 13396 ppm alts of the study establish that CES that fou 0 t1171 d-LL EST pre Case simply by insp tudy by experir spectro opy Analysis. All of the dat MATERIALS AND METHODS the te CEST the has D SH? eled Ark CEST int 10%DO buffer (PH .).Th added peptide and by simulations for kex values ranging from 20 to 200 s−1 showing that accurate exchange parameters and chemical shifts can be obtained even from fits on a per-residue basis. On a practical note, we suggest that when a system is investigated for the first time, a relatively high ν1 value (25−50 Hz) should be employed along with a long mixing time, TEX, in order to facilitate the detection of excited states. Once such states are observed, experiments can be performed at multiple B1 field strengths. Single-residue fits require data recorded at no less than a pair of B1 field strengths, and low fields have the advantage that narrower lines are obtained, albeit often at the expense of smaller effects. As a final note, it is worth comparing the advantages of CEST to the magnetization exchange transfer approach for quan￾tifying slowly exchanging protein systems.32,67−70 Here longitudinal magnetization is transferred between states during a mixing time that is placed between pairs of chemical shift evolution periods (t1 and t2 in a 2D data set), leading in favorable cases to the observation of cross-peaks connecting correlations derived from interconverting sites. The time dependencies of the evolution of cross- and diagonal peaks are subsequently fit to extract exchange parameters. For a two-site exchanging system it is straightforward to show that the maximum intensity of the cross-peaks is less than pG × pE, placing a rather severe restriction on what the minimum value of pE must be so that quantifiable cross-peaks can be obtained. Only one or two exchange cross-peaks could be reliably identified in magnetization exchange spectra recorded on either of the SH3 (pE = 2.4%) or FF (pE = 1.7%) domains considered here, so this approach for quantifying exchange and obtaining chemical shifts of the excited state was not an option. In contrast, in the CEST experiment, application of a radio frequency field at the position of the minor-state correlation for a mixing period of several hundreds of milliseconds leads to a very significant amplification effect that substantially improves the sensitivity of the experiment relative to magnetization exchange.37 As a result, and unlike magnetization exchange experiments, it is not a requirement that excited-state correlations be observed in the spectra. Furthermore, because “typical” HSQC data sets are analyzed (all correlations derived from the major conformer), resolution is not nearly as limiting as in mag￾netization exchange experiments, where diagonal peaks from both states as well as cross-peaks must be well resolved. The results of the present study establish that CEST spectros￾copy will play an important role in characterizing protein excited states that exchange in the slow regime, providing a complement to CPMG relaxation dispersion. Robust excited￾state chemical shifts are available from fits of CEST data, in many cases simply by inspection, facilitating structural studies of rare conformers and significantly increasing the exchange time-scale window that is amenable to detailed study by experimental NMR spectroscopy. ■ MATERIALS AND METHODS NMR Samples. A [U-15N,2 H]-labeled protein L sample dissolved in 50 mM sodium phosphate, 0.05% NaN3, 10% D2O buffer (pH 6.0) was prepared as described previously.71 The sample was very concentrated (∼4 mM) to test for experimental artifacts (see the text). An Abp1p SH3 domain exchanging complex was produced in a manner discussed previously,28 comprising ∼1.5 mM [U-15N,2 H]- labeled Abplp SH3 domain and unlabeled Ark1p peptide (17 residues) in a 50 mM sodium phosphate, 100 mM NaCl, 1 mM EDTA, 1 mM NaN3, 10% D2O buffer (pH 7.0). The concentration of added peptide was such that the sample was 97.5% free, 2.5% peptide-bound. A second Abp1p−Ark1p sample, fully bound, was prepared for direct measure￾ments of chemical shifts in the bound state. The A39G FF domain was expressed and purified as described previously for other FF variants.11,43 The sample consisted of ∼2 mM [U-15N]-labeled protein dissolved in a 50 mM sodium acetate, 100 mM NaCl, 10% D2O buffer (pH 5.7). NMR Spectroscopy. Experiments were performed on Varian Inova spectrometers (11.7 and 18.8 T) equipped with room￾temperature triple resonance probes. The weak 15N B1 field (applied during TEX) was calibrated in a 1D fashion using the approach of Guenneugues et al.72 with the pulse sequence shown in Figure 2 by focusing on an isolated protein peak that did not show exchange broadening. The measured inhomogeneity was ∼10% at both 11.7 and 18.8 T and relatively constant over the range of B1 values used in the experiments. All of the experiments involving protein L were performed at 25 °C, 11.7 T, and ν1 = 28 Hz with a 1 H decoupling field strength of 2.35 kHz during TEX = 0.4 s (see Figure 2). A series of 31 2D data sets were acquired, corresponding to 15N offsets between 103.91 and 133.52 ppm, obtained at 50 Hz (0.987 ppm) intervals. Experiments on the Abp1p−Ark1p system were recorded at 1 °C and 11.7 T. Sixty 2D data sets were obtained, with the 15N carrier centered at frequencies corresponding to 104.16 to 133.81 ppm at a spacing of 25 Hz (0.494 ppm) during TEX = 0.4 s. A pair of 15N B1 fields corresponding to field strengths of 12.8 and 26.6 Hz were used. Data at a third field strength (52.6 Hz) were recorded; here a spacing of 30 Hz (0.593 ppm) was employed. In all of the experiments, 1 H decoupling during TEX was achieved using a 2.3 kHz field with composite pulse decoupling consisting of repeated applications of a 90x240y90x element. Each 2D data set was recorded with acquisition times of (33 ms, 64 ms), eight transients, and a delay of 1.5 s between scans, corresponding to 19 h (ν1 = 12.8, 26.6 Hz) and 16 h (ν1 = 52.6 Hz) for each complete series. CEST experiments were performed on the A39G FF domain at 1 °C and 11.7 or 18.8 T. Data recorded at 11.7 T comprised a series of 2D spectra with 15N offsets ranging between 104.16 and 133.81 ppm obtained in increments of 0.494 ppm (25 Hz) for ν1 fields of 26.2 and 13 Hz and 0.296 ppm (15 Hz) when the B1 field was 6.2 Hz. Acquisition times of (35.6 ms, 64 ms) in (t1, t2) were used, along with eight transients per free induction decay (FID) and a relaxation delay between scans of 1.5 s for total data collection times of 20 h (ν1 = 26.2 and 13 Hz) and 33 h (ν1 = 6.2 Hz). A TEX value of 500 ms was used for all of the experiments, with a 2.46 kHz 1 H decoupling field (90x240y90x). CEST experiments were performed at 18.8 T using a 3.67 kHz 1 H decoupling field, ν1 = 14.8 or 28.9 Hz, and TEX = 500 ms. 15N carrier frequencies were positioned from 104.36 to 133.96 ppm in steps of 0.3085 ppm (25 Hz). Similar acquisition parameters were used as described for data acquired at 11.7 T, except that four tran￾sients per FID were recorded, for a total acquisition time of 19.5 h for each data set. CPMG relaxation dispersion profiles13 were recorded at 11.7, 14.0, and 18.8 T as described previously,73 using a constant-time CPMG interval26 (60 ms) with 1 H continuous-wave decoupling along with 25 νCPMG values ranging from 17 to 1000 Hz, with three duplicate points for error analysis. Dispersion data from all static magnetic field strengths were analyzed simulataneously. Data Processing and Analysis. All of the data sets were processed using the NMRpipe program,74 with peak intensities quantified using FuDA;75 subsequently, the spectra were visualized using SPARKY.76 CPMG relaxation data were analyzed in the standard manner using the program CATIA,75 which numerically propagates the Bloch−McConnell equations.77 Errors in the CPMG data were estimated on the basis of 2−3 repeat measurements. Errors in the CEST intensity profiles were estimated on the basis of the scatter in regions of 1D profiles that did not contain any intensity dips. The CEST intensity profiles were analyzed using an in-house-written python program, ChemEx (available upon request), that minimizes the standard χ2 equation Journal of the American Chemical Society Article 8158 dx.doi.org/10.1021/ja3001419 | J. Am. Chem. Soc. 2012, 134, 8148−8161