Selecting the Most Appropriate NMR Experiment to ... - ACS Publications

Article pubs.acs.org/jnp

Selecting the Most Appropriate NMR Experiment to Access Weak and/or Very Long-Range Heteronuclear Correlations Josep Saurí,*,† Yizhou Liu,† Teodor Parella,‡ R. Thomas Williamson,† and Gary E. Martin† †

NMR Structure Elucidation, Process & Analytical Chemistry, Merck & Co. Inc., 126 E. Lincoln Avenue, Rahway, New Jersey 07065, United States ‡ Servei de Ressonància Magnètica Nuclear, Universitat Autònoma de Barcelona, E-08193 Bellaterra, Barcelona, Catalonia, Spain S Supporting Information *

ABSTRACT: Heteronuclear long-range NMR experiments are well established as essential NMR techniques for the structure elucidation of unknown natural products and small molecules. It is generally accepted that the absence of a given nJXH correlation in an HMBC or HSQMBC spectrum would automatically place the proton at least four bonds away from the carbon in question. This assumption can, however, be misleading in the case of a mismatch between the actual coupling constant and the delay used to optimize the experiment, which can lead to structural misassignments. Another scenario arises when an investigator, for whatever reason, needs to have access to very long-range correlations to confirm or refute a structure. In such cases, a conventional HMBC experiment will most likely fail to provide the requisite correlation, regardless of the delay optimization. Two recent methods for visualizing extremely weak or very long-range connectivities are the LR-HSQMBC and the HSQMBC-TOCSY experiments. Although they are intended to provide similar structural information, they utilize different transfer mechanisms, which differentiates the experiments, making each better suited for specific classes of compounds. In this report we have sought to examine the considerations implicit in choosing the best experiment to access weak or very long-range correlations for different types of molecules.

N

signal is strongly dependent on the congruence of the delay optimization, Δ, and the actual scalar coupling constant, nJCH. Those factors individually or, in some cases, combined can prevent the visualization of expected cross-peaks in the final spectrum, either because they are weak or because they are very long-range heteronuclear responses, in both cases due to a small nJCH. In addition, there is a cos(πJHHΔ) term in the detected component of magnetization that can cause accidental cancelation of cross-peaks. For instance, a proton with a JHH = 8 Hz having an active nJCH = 8 Hz can cancel out, giving no signal when the experiment is optimized to 8 Hz. HSQMBC was proposed as an alternative to the HMBC experiment to obtain long-range proton−carbon correlations.3 The experiment differs from HMBC in that only single-quantum 13C coherences evolve during t1, affording spectra that can be phased rather than requiring magnitude calculation. However, HSQMBC still leads to line shapes with mixed phases due to JHH evolution during the INEPT transfers. The HSQMBC experiment has been further improved over the past decade with several modifications, both in an effort to recover expected missing signals and also to facilitate the accurate measurement of longrange heteronuclear coupling constants.4 Nonetheless, the signal intensity of any given correlation is still an intrinsic

MR spectroscopy is a powerful analytical technique widely used by chemists for structure elucidation of various types of molecules, ranging from inorganic compounds, to proteins, to small synthetic molecules and natural products.1 Chemists engaged in structure elucidation have available a “tool kit” composed of 1D and 2D NMR experiments that are (almost) always utilized for the full characterization of such compounds, i.e., classic 1D 1H and 13C and 2D techniques including COSY, ROESY or NOESY, HSQC, and HMBC. Heteronuclear long-range correlation experiments are crucial experiments to connect structural fragments via nonprotonated carbons or across heteroatoms. HMBC is undoubtedly the most widely used NMR experiment for observing long-range heteronuclear correlations.2 The most basic version of the pulse sequence employs only a few RF pulses, making it very robust in terms of RF inhomogeneities as well as the most sensitive experiment providing long-range proton−carbon connectivities. However, there are several important issues associated with the HMBC experiment: (a) the detected proton magnetization is antiphase with respect to the active carbon (2IxSz); (b) both multiple quantum coherences and homonuclear couplings (JHH) evolve during the entire t1 evolution period, giving tilted multiplet structures along the F1 dimension; (c) cross-peaks exhibit highly distorted complex phases due to the evolution of JHH during the entire sequence; and (d) the signal intensity depends on the function sin(πnJCHΔ). Thus, the observed © XXXX American Chemical Society and American Society of Pharmacognosy

Received: February 15, 2016

A

DOI: 10.1021/acs.jnatprod.6b00139 J. Nat. Prod. XXXX, XXX, XXX−XXX

Journal of Natural Products

Article

reach out to as many protons as the HSQMBC-TOCSY. Hence, HSQMBC-TOCSY and LR-HSQMBC are proposed as complementary but alternative tools to the traditional HMBC and HSQMBC experiments. The intention of this work is to make a comparative assessment on which experiment, providing weak and/or very long-range connectivities (2 Hz optimized LR-HSQMBC vs 8 Hz optimized HSQMBC-TOCSY with a 60 ms mixing time for the TOCSY), is a better choice in terms of general sensitivity. We also sought to illustrate which experiment affords better performance in terms of number of signals obtained, which in turn will depend on the nature of the compound being studied. The comparison was done using two different model compounds reflecting two different molecular scenarios: proton-rich retrorsine, 1, was used as the first model, and cervinomycin A2, 2, was used as a proton-deficient compound example. Finally, it will also be shown that the same conclusions can be extended to the application of both experiments to 1 H−15N heteronuclear correlation.

factor limiting the ability of both the HMBC and HSQMBC experiments to visualize either weak or very long-range heteronuclear correlations. HSQC-TOCSY is an elegant alternative experiment used to obtain long-range heteronuclear correlations by using a somewhat different approach.5 Instead of a direct transfer mechanism based on the heteronuclear coupling between the proton and the long-range-coupled carbon, HSQC-TOCSY utilizes an out-and-back transfer mechanism from a proton to its directly bound carbon via 1JCH followed by a TOCSY transfer via the H−H coupling network. The downside is that no correlations to nonprotonated carbons can be obtained from this experiment. Two different techniques have recently been described to access weak and/or very long-range correlations. The first to be reported was the LR-HSQMBC experiment,6 which is based on an optimum choice of the delay between the real nJCH and the average value Δ, which is usually set to 2−4 Hz in order to maximize the signal intensity response for those correlations exhibiting very small long-range heteronuclear coupling constants, according to the sin2(πnJCHΔ) function. The pulse sequence uses a refocused INEPT to restore in-phase magnetization so that the application of heterodecoupling during acquisition is allowed. The primary drawback of the experiment is that, when optimized for correlations with small n JCH (2−4 Hz), the pulse sequence becomes quite long (lasting from 250 to 500 ms), so that losses due to T2 relaxation could become an issue in some cases. The experiment also requires relatively high digitization in F1 (512 to 768 increments are recommended), although evaluations of nonuniform sampling techniques to shorten experiment times or increase S/N are under way in our laboratories. However, the experiment is, in any case, much more sensitive than any of the variants of the ADEQUATE experiments that are based on a carbon−carbon transfer following the initial 1JCH transfer.7 Hence, LRHSQMBC was proposed as a complementary tool to the HMBC and basic HSQMBC experiments, and its use is especially indicated for proton-deficient molecules where obtaining 4J, 5J, and 6J can be critical.8 The second method is the hybrid, hyphenated HSQMBCTOCSY experiment,9 which is based on an initial nJCH magnetization transfer followed by a JHH transfer mechanism. Hence, the experiment borrows transfer mechanisms from both the HSQMBC and TOCSY experiments. HSQMBC-TOCSY follows the idea of the HSQC-TOCSY experiment but uses an initial long-range heteronuclear transfer mechanism so that nonprotonated carbons are also accessible. Very importantly, the signal intensity of a given cross-peak will depend on the function sin(πnJCHΔ) cos(2πJHHτm) for a spin system including only two protons, i.e., an initial strong nJCH transfer and a subsequent TOCSY transfer via H−H that is dependent on the homonuclear 3JHH coupling constants. Because a strong nJCH is required as an initial step, the delay is usually optimized to 8 Hz, which generally establishes the strongest 3J correlations. Due to the inclusion of the TOCSY mixing time (30−60 ms), the duration of the pulse sequence can be 150−180 ms, which is considerably shorter than the LR-HSQMBC pulse sequence with consequently fewer losses due to T2 relaxation and correspondingly better relative sensitivity. In a similar way, one can also use a COSY transfer instead of a TOCSY pulse train, as described in the original reference.9 Although the HSQMBCCOSY pulse sequence is shorter (∼130 ms) because it avoids the TOCSY mixing time period, the experiment is not able to

■

RESULTS AND DISCUSSION Proton-Rich Compounds. Retrorsine (1, C18H25NO6) is an alkaloid that has been reported in the literature as a model compound for the development of edited ADEQUATE experiments.10 For this particular study, it was employed as a proton-rich natural product example because it incorporates well-defined proton spin systems that would allow an efficient TOCSY transfer, therefore making the molecule a good candidate for the HSQMBC-TOCSY experiment. Figure 1 illustrates a comparison of HSQMBC-TOCSY and LRHSQMBC experiments (see the Supporting Information for the pulse sequence schematics and the Bruker TopSpin 3.1 pulse sequence code), both experiments were acquired and processed identically to compare their relative performance. From a simple visual inspection it is readily apparent that the former provided a greater number of cross-peaks. In total, the HSQMBC-TOCSY experiment afforded 34 additional correlations that were absent from the LR-HSQMBC (Table S1 in the Supporting Information). Among those, one is a six-bond correlation, six are five-bond correlations, 10 are four-bond correlations, 10 are three-bond correlations, and seven are twobond correlations. In contrast, there are only four correlations observed in the LR-HSQMBC spectrum that did not appear in the HSQMBC-TOCSY spectrum. As noted in the original report, the LR-HSQMBC experiment can give rise to weak signals even in the very late segment of the t1 increments, and it is therefore advisible to acquire the experiment with a reasonably high number of t1 increments. In the case of retrosine, raising the number of t1 increments up to 640 provided four additional correlations with respect to the data acquired with 256 t1 increments. In all cases, the additional correlations from the HSQMBCTOCSY experiment arise due to the homonuclear transfer B



Article

Figure 1. HSQMBC-TOCSY (A) and LR-HSQMBC (B) experiments of retrorsine, 1. Both experiments were acquired under the same acquisition parameters using 2048 points in F2 and 32 transients for each one of the 256 t1 increments. In A the nJCH delay optimization was set to 8 Hz and the TOCSY mixing time was set to 60 ms. For the spectrum shown in B the nJCH delay was optimized for 2 Hz. Total experiment time was 3 h 24 min in A and 4 h 7 min in B. Note that the difference in the total experimental time is due to the different delay optimization used in each experiment. Data were processed in exactly the same way using a 90° shifted squared-sine apodization function and applying zero-filling to 4K in F2 and 1K in F1. Thirty-four additional correlations were obtained in A, most of them highlighted in the insets. All additional correlations obtained in the HSQMBCTOCSY that do not appear in the 2 Hz LR-HSQMBC are summarized on the structure shown in C. Color code: black arrows correspond to twobond correlation, green arrows signify three-bond correlation, blue arrows denote four-bond correlation, red arrows indicate five-bond correlation, and the purple arrow identifies a six-bond correlation from H-6 to carbonyl C-16.

of 0.2 Hz as well (Figure S1 in the Supporting Information for details). However, the signal intensity is rather large for such a small coupling constant. Since 3JC4−H6b is a strong heteronuclear correlation (7.6 Hz experimental value, Figure S2), and there is also a strong TOCSY from H-6b and H-7, as shown in Figure 2B, a strong cross-peak was observed for the 4JC4−H7 correlation, regardless of the small four-bond heteronuclear coupling constant. Another important consideration is the coexistence of different magnetization transfer pathways all potentially contributing to the enhancement of the signal intensity of a given correlation. For instance, in the case of the 4JC4−H7 correlation peak, an additional potential pathway could arise via 3JC4−H1 + JH1−H7. Finally, it is worth noting that a weak, but very real five-bond correlation is obtained from H-9a/b to C-4 through a weak TOCSY between H-6b and H-9 via H-7. Proton-Deficient Compounds. Cervinomycin A2 (2, C29H21NO9) is a highly proton-deficient antibiotic that was first isolated and characterized by O̅ mura and co-workers in 1982.12 This highly proton-deficient molecule was employed as a model compound for the initial development of the LRHSQMBC experiment and has more recently been used to assess the impact of LR-HSQMBC data in computer-assisted structure elucidation (CASE) programs.8 Here it is used to illustrate how a severely proton-deficient compound hampers the performance of HSQMBC-TOCSY compared to the LR-

mechanism employed in the latter (TOCSY) half of the experiment, which requires a proton spin system to propagate the information obtained in the first part (HSQMBC) of the pulse sequence. If the initial long-range correlation(s) that are established in the HSQMBC step are strong enough, and threebond correlations generally are, magnetization will be propagated as far as the TOCSY is capable of “reaching out”. A closer examination reveals that the vast majority of the additional correlations took place in the bottom part of the structure where the compound is highly protonated. Figure 1C summarizes all those additional correlations; the length of the correlation pathway is designated by the color-coded arrows. It is also noteworthy that there is a dearth of additional correlations specifically where the molecule is less protonrich, e.g., −O−C-11−C-12 skeleton. As noted above, an important benefit of the HSQMBCTOCSY experiment is the signal intensity dependence of the relayed responses derived by the TOCSY block. Owing to the transfer mechanism of the experiment, strong correlations to a remote site(s) in the molecular framework can be obtained even in the case where the active heteronuclear coupling constant is close to zero. An example is shown in Figure 2C, which shows the four-bond correlation between carbon C-4 and proton H-7. DFT calculations predicted a 4JC4−H7 = 0.2 Hz, whereas the experimental measurement11 agreed, giving a value C



Article

Figure 2. (A) 1H reference spectrum of retrorsine. (B) 1D sel-TOCSY (mixing time 60 ms) after exciting the H-6b proton in retrorsine. (C) 1D slice taken at the C-4 carbon frequency extracted from the HSQMBC-TOCSY experiment shown in Figure 1A. Clearly, H-6b has a strong TOCSY correlation to H-7, which accounts for the high signal intensity of the C-4−H-7 cross-peak visualized in the HSQMBC-TOCSY experiment, despite its coupling constant (4JC4−H7) being close to zero. Cross-peaks arising through the TOCSY transfer are red labeled in C. Note that in C some peaks present phase distortions due to zero-quantum artifacts arising from the TOCSY transfer.

performance.8 The correlation in question was specifically the 4 JCH from H-1 to C-27. Since the atom connectivity from H-1 to C-27 crosses a heteroatom plus a carbonyl group, the TOCSY interval of the HSQMBC-TOCSY experiment has no way to progress through these positions, whereas a direct heteronuclear long-range transfer mechanism is able to provide such a correlation. Figure 4 shows 1D slices extracted for the carbon C-27 from both experiments. The slice from the 2 Hz LR-HSQMBC spectrum in Figure 4B is shown in an 8-fold vertical expansion with respect to the slice from the HSQMBCTOCSY experiment. Although the LR-HSQMBC is less sensitive than the HSQMBC-TOCSY, the superior performance of the former in proton-deficient compounds amply justifies the acquisition of these data. It is also worth noting that the acquisition of an LR-HSQMBC spectrum, especially in a sample-limited context, is highly preferred over any variant of the 1,n-ADEQUATE experiment.7c,d LR-HSQMBC and HSQMBC-TOCSY are both well suited for observing long-range 1H−15N heteronuclear correlations, which can be of considerable importance when dealing with the structure elucidation of any complex natural product. The same considerations outlined above apply for 1H−15N experiments. The only consideration that should be taken into account is that, in general, the long-range 1H−15N coupling constants tend to be a bit smaller than 1H−13C coupling constants. Therefore, it is advisible to optimize the HSQMBC step in the HSQMBC-TOCSY experiment to 5 Hz instead of 8 Hz

HSQMBC insofar as obtaining weak or very long-range correlations. Figure 3A and B show an expansion of the HSQMBC-TOCSY and LR-HSQMBC spectra containing the responses for the aromatic H-9 and H-10 resonances, respectively. When compared to the HSQMBC-TOCSY experiment, 14 additional correlations were observed in the LR-HSQMBC spectrum, and eight of them corresponded to these two aromatic protons, as shown by the insets in Figure 3B. The additional correlations observed in the LR-HSQMBC spectrum are primarily very long-range correlations: there was one 6JCH, four 5JCH, eight 4JCH, and one additional 3JCH correlation. In contrast, there were only two correlations observed in the HSQMBC-TOCSY spectrum that were not observed in the 2 Hz LR-HSQMBC spectrum (Table S2). One of the two additional correlations from the HSQMBC-TOCSY spectrum was a 5JCH between H-30 and C-7 due to a small 4JHH coupling between H-30 and H-5, and the other was a 5JCH correlation from H-7 to C-30 due to a small 4JHH between H-7 and H-5. On the basis of a cursory comparison of the data shown in Figure 3, it is clear that the proton-deficient nature of cervinomycin A2 renders the HSQMBC-TOCSY experiment disadvantageous when compared to LR-HSQMBC, with only scant additional information being gleaned from the HSQMBC-TOCSY spectrum. As recently reported, there was a key correlation obtained from the 2 Hz LR-HSQMBC data that had a very significant impact on CASE program D



Article

Figure 3. Expanded areas corresponding to the (A) HSQMBC-TOCSY and (B) LR-HSQMBC spectra, showing correlations for the aromatics protons H-10 and H-9 of cervinomycin A2 (2). For these two protons, eight additional correlations were observed in B as marked in the chemical drawing. Color code: green arrows signify three-bond correlation, blue arrows denote four-bond correlation, red arrows indicate five-bond correlation, and the purple arrow identifies a six-bond correlation. Both experiments were acquired under the same acquisition parameters as 2K data points with 32 scans accumulated for each of the 640 t1 increments acquired. Spectral windows in F2 and F1 were set 4504.5 and 25 153.8 Hz, respectively. HSQMBC-TOCSY was optimized for nJCH = 8 Hz with the TOCSY mixing time optimized for 60 ms, giving a total acquisition time of 8 h 10 min. The LR-HSQMBC experiment was optimized for 2 Hz, giving a total acquisition time of 9 h 57 min. Prior to Fourier transformation of the data, zero-filling to 1024 points in F1 and 4096 points in F2 and 90° shifted squared-sine apodization were applied in both dimensions.

the former vs ∼500 ms for the latter). Note that, because N-19 is located in a proton-rich segment of the structure, the number of correlations obtained with the HSQMBC-TOCSY experiment was superior to the results for the LR-HSQMBC experiment. However, when a 1D slice from N-9 is considered, the LR-HSQMBC provides a somewhat better S/N ratio (Figure 5D) than the corresponding HSQMBC-TOCSY (Figure 5C), despite the greater length of the pulse sequence. This observation can be easily explained by examining the DFT-calculated 1H−15N coupling constant values from N-9 (2JN9−H8 = −1.62 Hz, 3JN9−H4 = −0.60 Hz, 3JN9−H11a = −2.84 Hz, 3JN9−H11b = +0.07 Hz, 3JN9−H13 = −2.19 Hz), because none of them are larger than 2−3 Hz. Thus, the best solution would probably be to optimize the 1H−15N HSQBMC-TOCSY to 5 Hz, which would provide a more uniform response for both nitrogen atoms. In summary, as demonstrated in this report, LR-HSQMBC and HSQMBC-TOCSY are complementary experiments for obtaining weak or very long-range heteronuclear correlation

typically used for 1H−13C experiments, although this cannot be applied as a general rule. A good example is the case of strychnine (3), where long-range 1H−15N coupling constants for N-19 are substantially larger than those for N-9.13

As shown in Figure 5, when the 1H−15N HSQMBC-TOCSY experiment was optimized to 8 Hz + 60 ms, N-19 showed much better S/N than the corresponding 2 Hz LR-HSQMBC (Figure 5A vs B, respectively). The difference is likely due to the pulse sequence length of the two experiments (∼180 ms in E



Article

Figure 4. 1D slices corresponding to C-27 resonances extracted from the (A) HSQMBC-TOCSY and (B) LR-HSQMBC experiments. Although the S/N is lower in B, the performance of the LR-HSQMBC is superior, affording three additional correlations with respect to the HSQMBC-TOCSY spectrum. Correlations obtained from each experiment are highlighted in the corresponding structures shown, using the following color coding: green corresponds to three-bond correlation; blue denotes four-bond correlation; red indicates five-bond correlation; and purple identifies six-bond correlations. As noted in a CASE program performance evaluation, the 4JCH C-27−H-1a/b correlations were critical to program performance, manifested as a 600-fold reduction in the calculation time when these data were included in the program input file.9

Figure 5. 1D slices extracted from the 1H−15N HSQMBC-TOCSY and the 1H−15N LR-HSQMBC13 experiments for both N-19 and N-9 nitrogens of strychnine. (A) N-19 slice from the 1H−15N HSQMBC-TOCSY optimized for 8 Hz + 60 ms, (B) N-19 slice from the 1H−15N LR-HSQMBC optimized for 2 Hz, (C) N-9 slice from the 1H−15N HSQMBC-TOCSY optimized for 8 Hz + 60 ms, and (D) N-9 slice from the 1H−15N LRHSQMBC optimized for 2 Hz. 1H spectrum of strychnine is shown on the bottom of each panel for reference. Note that distortions in the line shapes arise either due to zero-quantum contributions in the HSQMBC-TOCSY or due to JHH modulation in the LR-HSQMBC experiment. For more details about the nature of these distortions the reader is referred to the original research articles.6,9

F



■

information. Their output will depend on the nature of the compound being investigated due to their different transfer mechanisms. It can be concluded that HSQMBC-TOCSY is the more desirable technique for proton-sufficient molecules (e.g., H:C ≈ 1) since it offers better sensitivity with excellent response characteristics. In contrast, LR-HSQMBC is far superior when it comes to severely proton-deficient scenarios (H:C ≪ 1). LR-HSQMBC can also be considered as a “highsensitivity” technique despite the 2 Hz optimization and F1 digitization requirements when compared to more timeconsuming experiments such as the 1,n-ADEQUATE experiments that rely on 13C−13C homonuclear transfers.7c,d

■

EXPERIMENTAL SECTION

■

ASSOCIATED CONTENT

Article

AUTHOR INFORMATION

Corresponding Author

*Tel: (+1) 732.594.5371. Fax: +1 732 594 9456. E-mail: josep. [email protected]. Notes

The authors declare no competing financial interest.

■ ■

ACKNOWLEDGMENTS T.P. gratefully acknowledges financial support from the MINECO (project CTQ2012-32436). REFERENCES

(1) (a) Reynolds, W. F.; Mazzola, E. P. In Progress in the Chemistry of Organic Natural Products; Kinghorn, A. D.; Falk, Y.; Kobayashi, J., Eds.; Methodology, Springer: Heidelberg, 2015; pp 223−310. (b) Breton, R. C.; Reynolds, W. F. Nat. Prod. Rep. 2013, 30, 501−524. (c) Reynolds, W. F.; Enríquez, R. G. J. Nat. Prod. 2002, 65, 221−244. (d) Williams, A. J.; Martin, G. E.; Rovnyak, D. Modern NMR Approaches to the Structure Elucidation of Natural Products, Vol. 1: Instrumentation and Software; Royal Society of Chemistry, October, 2015. (e) Williams, A. J.; Martin, G. E.; Rovnyak, D. Modern NMR Approaches to the Structure Elucidation of Natural Products, Vol. 2: Data Acquisition and Applications to Compound Classes; Royal Society of Chemistry, anticipated April 2016 release, in press. (2) (a) Bax, A.; Summers, M. F. J. Am. Chem. Soc. 1986, 108, 2093− 2094. (b) Furrer, J. Concepts Magn. Reson., Part A 2012, 40A, 101−127. (c) Furrer, J. Concepts Magn. Reson., Part A 2012, 40A, 146−169. (d) Furrer, J. Concepts Magn. Reson., Part A 2014, 43A, 177−206. (3) Williamson, R. T.; Marquez, B. L.; Gerwick, W. H.; Kover, K. E. Magn. Reson. Chem. 2000, 38, 265−273. (4) Parella, T.; Espinosa, J. F. Prog. Nucl. Magn. Reson. Spectrosc. 2013, 73, 17−55. (5) Parella, T.; Sanchez-Ferrando, F.; Virgili, A. J. Magn. Reson. 1997, 126, 274−277. (6) Williamson, R. T.; Buevich, A. V.; Martin, G. E.; Parella, T. J. Org. Chem. 2014, 79, 3387−3394. (7) (a) Köck, M.; Reif, B.; Fenical, W.; Griesinger, C. Tetrahedron Lett. 1996, 37, 363−366. (b) Reif, B.; Köck, M.; Kerssebaum, R.; Kang, H.; Fenical, W.; Griesinger, C. J. Magn. Reson., Ser. A 1996, 118A, 282−285. (c) Martin, G. E. Annual Reports in NMR Spectroscopy; Webb, G. A., Ed., Elsevier: London, 2011; Vol. 74, pp 215−291. (d) Martin, G. E.; Reibarkh, M.; Buevich, A. V.; Blinov, K. A.; Williamson, R. T. eMagRes 2014, 3, 15−234. (8) Blinov, K. A.; Buevich, A. V.; Williamson, R. T.; Martin, G. E. Org. Biomol. Chem. 2014, 12, 9505−9509. (9) Saurí, J.; Marcó, N.; Williamson, R. T.; Martin, G. E.; Parella, T. J. J. Magn. Reson. 2015, 258, 25−32. (10) Martin, G. E.; Williamson, R. T.; Dormer, P. G.; Bermel, W. Magn. Reson. Chem. 2012, 50, 563−568. (11) Saurí, J.; Espinosa, J. F.; Parella, T. Angew. Chem., Int. Ed. 2012, 51, 3919−3922. (12) O̅ mura, S.; Iwai, Y.; Hinotozawa, K.; Takahashi, Y.; Kato, J.; Nakagawa, A.; Hirano, A.; Shimizu, H.; Haneda, K. J. Antibiot. 1982, 35, 645−652. (13) Williamson, R. T.; Buevich, A. V.; Martin, G. E. Tetrahedron Lett. 2014, 55, 3365−3366.

General Experimental Procedures. Samples were prepared as follows: a 1.1 mg sample of retrorsine (1) was dissolved in 35 μL of CD3OD; 0.5 mg of cervinomycin A2 (2) was dissolved in 35 μL of CDCl3; and 4.5 mg of strychnine (3) was dissolved in 35 μL of CDCl3. All samples were then respectively transferred to a 1.7 mm NMR tube (Bruker). All of the NMR data were acquired using a 500 MHz Bruker three-channel AVANCE III NMR spectrometer equipped with a Bruker gradient triple resonance TXI 1.7 mm TCI 1H/13C−15N MicroCryoProbe. Data shown in Figure 1 were acquired as 2048 data points with 32 scans accumulated for each of the 256 t1 increments accumulated. Spectral windows in F2 and F1 were set at 3501.4 and 22 638.4 Hz, respectively. The HSQMBC-TOCSY experiment was optimized for n JCH = 8 Hz with the TOCSY mixing time optimized for 60 ms, giving a total acquisition time of 3 h 24 min. The LR-HSQMBC data were acquired with the same digitization in both F2 and F1 and the same spectral width; the experiment was optimized for 2 Hz, giving a total acquisition time of 4 h 7 min. Prior to Fourier transformation of the data, zero-filling to 1024 points in F1 and 4096 points in F2 was employed and 90° shifted squared-sine apodization was applied in both dimensions. Data shown in Figure 2B (1D selective TOCSY) were acquired in 25 s using 8 scans. A 45 ms Gaussian pulse was used to selectively invert H6b. Data shown in Figure 3 were acquired as 2048 data points with 32 scans accumulated for each of the 640 t1 increments. Spectral windows in F2 and F1 were set at 4504.5 and 25 153.8 Hz, respectively. HSQMBC-TOCSY was optimized for nJCH = 8 Hz with the TOCSY mixing time optimized for 60 ms, giving a total acquisition time of 8 h 10 min. The LR-HSQMBC experiment was optimized for 2 Hz, giving a total acquisition time of 9 h 57 min. Prior to Fourier transformation of the data, zero-filling to 1024 points in F1 and 4096 points in F2 and 90° shifted squared-sine apodization were applied in both dimensions. Data shown in Figure 5 were acquired as 2048 data points with 64 scans accumulated for each of the 80 t1 increments. Spectral windows in F2 and F1 were set at 5411.3 and 9123.7 Hz, respectively. HSQMBC-TOCSY was optimized for n JNH = 8 Hz with the TOCSY mixing time optimized for 60 ms, giving a total acquisition time of 2 h 41 min. The LR-HSQMBC experiment was optimized for 2 Hz, giving a total acquisition time of 3 h 9 min. Prior to Fourier transformation of the data, zero-filling to 1024 points in F1 and 4096 points in F2 and 90° shifted squared-sine apodization were applied in both dimensions.

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jnatprod.6b00139. Additional figures, pulse sequence schemes with full details, as well as pulse program codes for Bruker (PDF) G


Selecting the Most Appropriate NMR Experiment to ... - ACS Publications

Recommend Documents