Read an Excerpt
Modern NMR Approaches to the Structure Elucidation of Natural Products
Volume 1: Instrumentation and Software
By Antony Williams, Gary E. Martin, David Rovnyak The Royal Society of Chemistry
Copyright © 2016 The Royal Society of Chemistry
All rights reserved.
ISBN: 978-1-84973-383-0
CHAPTER 1
New Directions in Natural Products NMR: What Can We Learn by Examining How the Discipline Has Evolved?
GARY E. MARTIN, ANTONY J. WILLIAMS AND DAVID ROVNYAK
In 1992, in a laboratory that is set in the Amazonian Rainforest, Sean Connery (in the guise of researcher Robert Campbell and the movie Medicine Man) made an injection into a mass spectrometer and identified a new natural product. Almost 25 years later, even with the miniaturization of analytical instrumentation and the incredible achievements in both sensitivity and mass resolution, it is not possible to elucidate a natural product structure in an automated manner using mass spectrometry. The only way that a complex natural product structure can be identified using this technique is by ensuring significant fragmentation and validating masses against a database. The dream of being able to use spectroscopy to elucidate automatically a molecular structure that is embodied in the movie Medicine Man is, however, much closer to reality in a laboratory setting using NMR spectroscopy.
Ultimately, the capability of modern NMR as a technique is the sum total of the assembly of a group of technologies paired with scientific acumen. NMR spectroscopists drive hardware and software in synergy to perform both data generation and analysis. With regard to natural product structure elucidation, the ultimate goal of NMR spectroscopists is to extract knowledge via the manipulation of magnetization and to determine, to the greatest extent possible, molecular-level detail regarding the spatial distribution, atom-to-atom connectivities, and orientations of atoms and bonds. What is achievable today using analytical spectroscopy tools applied to structure elucidation challenges is, relative to just a few years ago, truly breathtaking.
From this point forward, it is extremely difficult to predict how the future of natural products NMR will evolve in the decades to come. Twenty years ago, could we have imagined that acquiring long-range 1H–15N heteronuclear multiple-bond correlation (HMBC) spectra would become a routine and integral part of alkaloid structure elucidation studies? Even with one of the authors having a direct hand in inaugurating those experiments, he would not have dared to predict the now routine usage of them. Indeed, 5 years ago, would we have predicted the now burgeoning development of pure shift NMR methods? No, we probably would not have had the foresight to predict where we are now with these methods and the impact that they are already having on our ability to probe the structures of increasingly complex natural products, often available in only very limited quantities. Could we have anticipated the significantly increased "reach" afforded an investigator for 1H–13C or 1H–15N heteronuclear shift correlation studies when using the newly reported LR-HSQMBC experiment developed as a complement to the venerable HMBC experiment? Again, probably not. Most recently, would we have even dared to imagine that it would be possible to perform 13C–15N correlations at natural abundance using 1H detection in the just reported HCNMBC experiment? Despite one of the authors' own considerable experience acquiring 1,1- and 1,n-ADEQUATE experiments on sub-milligram samples, he would not have had the nerve to make such an audacious forecast. Nevertheless, there is an example of just such a spectrum per- formed on a 4 mg sample of strychnine using a 1.7 mm MicroCryoProbe (Bruker) in the chapter on alkaloids (see Volume 2, Chapter 10)! None of the three techniques just noted existed when we started to assemble these volumes just four short years ago.
Similar events can be pointed to in terms of hardware developments. In the 1980s, NMR studies were routinely conducted in 5 mm NMR tubes. That changed in 1992 with the introduction of 3 mm NMR probes in one of the authors' laboratories, and changed several additional times with the introduction of 1.7, 1.0, and microcoil NMR probes. Smaller diameter tube formats foreshadowed the even more profound change embodied in the development and now widespread availability of helium-cooled cryogenic NMR probes, and more recently the liquid nitrogen-cooled Prodigy probes offered by Bruker BioSpin. Then we saw the diameter of cryogenic probes "shrink" to 3 mm first and then to 1.7 mm. With the shrinking coil diameter of cryogenic NMR probes, sample requirements have correspondingly plummeted. Using a 1.7 mm MicroCryoProbe, one of the authors has demonstrated the acquisition of pure shift HSQC spectra of a sample of ~3 µg of a "heavy" drug metabolite (MW 661) generated using incubation with a recombinant enzyme in 14 h. Natural products, of course, can also be interrogated at the same level. Readers interested in very low-level natural product structure investigations are directed to several relatively recent reviews by Molinski and co-workers. Individually, all of these changes have been significant. In concert, the impact of cryogenic NMR probe technology has had a profound effect on what is possible in terms of natural product structure elucidation.
By way of providing a real-world example of what is feasible today regarding the structure elucidation of a complex natural product, we next consider what modern NMR experiments can provide in terms of connectivity and correlation data. When applied in tandem and in conjunction with accurate mass measurements, these data underpin the elucidation process, whether it be manual analysis or, as discussed later, performed by a computer. Beyond fundamental 1D proton and carbon reference spectra, there is a plethora of 2D NMR experiments available to an investigator sitting at the console of a modern high-field NMR spectrometer. Where to begin?
Strategies will vary from one investigator or laboratory to the next. One of the authors (G.M.) prefers to run a proton spectrum immediately followed by a multiplicity-edited HSQC spectrum. Within the past year, the HSQC experiment choice has become more powerful with the availability of pure shift variants of the experiment, which collapse all but anisochronous geminal methylene resonances to singlets, thereby improving both resolution and sensitivity.
If we employ strychnine as a model compound, the information content of strychnine can be described as illustrated in Figure 1.1.
To illustrate the homonuclear decoupling in the pure shift HSQC spectrum, a segment of the aliphatic region encompassing the H12, H23a/b, H16 and H8 resonances is shown in Figure 1.3. An expansion of the contour plot is shown in Figure 1.3a. The H12, H16 and H8 correlations are collapsed to singlets while the 23-methylene resonances are collapsed from doubled doublets to a pair of doublets. The vicinal coupling of both H23a and H23b to the H22 vinyl proton (1H–12C) is collapsed since the likelihood of a 13C resonance being adjacent to the detected 1H–13C resonant pair is 1 in 10 000. In contrast, for the methylene protons, both are on the same 13C and hence are unaffected by the BIRD-based decoupling applied during acquisition, leaving them as a pair of doublets. Figure 1.3b shows the high-resolution proton spectrum (A) and the phased traces extracted at the 13C shifts of C12 (B) and C23 (C).
Following the acquisition of some form of an HSQC spectrum, typical structure elucidation strategies will probably next acquire COSY data, with which most readers will likely be familiar. Homonuclear correlation data can be used to subgroup the proton resonances into discrete spin systems. For strychnine, the various spin systems in the structure of the molecule are subgrouped by color as shown in Figure 1.4. We refer to these types of figures as "correlation diagrams" or, to interject some humor, "spaghetti diagrams," the origin of this euphemistic label becoming obvious as the diagrams become more complex based on the type of experiment being applied and the nature of the data extracted.
Following the acquisition of a proton spectrum, HSQC, and COSY data, investigators will typically begin to try to deduce how the various parts of the molecule are interconnected. From an exact mass measurement that will provide the empirical formula of the molecule being investigated and the HSQC spectrum, the number of protonated carbons can be readily deter- mined. Most structure elucidation strategies next embark on the acquisition of a long-range 1H–13C heteronuclear shift correlation spectrum. The HMBC experiment described in 1986 by Bax and Summers is probably the most widely cited NMR experiment ever described, and has been the subject of numerous reviews. Aside from the incorporation of adiabatic pulses, the experiment has changed relatively little since its inception and there is currently not a real-time pure shift version of the experiment available, although the pseudo-3D tilt-HMBC was reported in 2013 by Sakhaii et al.
Prior to embarking on the acquisition of HMBC data, a decision must be made on the optimization of the long-range delay. Typically, 10 Hz has probably been the most commonly used optimization, although the data in the example that follows were acquired with an 8 Hz optimization, with the low-pass J-filter that that is utilized to reject unwanted 1JCH correlations optimized for 145 Hz. Despite the inclusion of a low-pass J-filter in the pulse sequence, signal-to-noise ratios with modern cryogenic NMR probes are so high that as one goes down towards the noise floor of the spectrum, the 13C satellites can frequently be observed symmetrically displaced about the proton and carbon chemical shift coordinates, i.e. the location of the direct correlation response in the HSQC spectrum. The data shown were acquired with 256 increments of the evolution time used to digitize the second frequency domain. Unlike the HSQC experiment, which affords only one-bond correlations governed by the 1JCH coupling constant, there is no such filtering of long-range correlations in an HMBC spectrum and nJCH correlations where n = 2–4 are routinely observed. In contrast, a recently reported modification of the 1,n-ADEQUATE experiment, inverted 1JCC 1,n-ADEQUATE does allow the differentiation of 1JCC correlations form nJCC correlations where n ≥ 2. Most commonly, 2JCH and 3JCH correlations will be observed in HMBC spectra with longer range correlations, e.g.4JCH and 5JCH correlations, being observed less commonly. As we shall see with strychnine, however, the rigid skeletal framework of the molecule greatly facilitates the observation of longer range couplings. All of the correlations extracted from an 8 Hz optimized HMBC spectrum of strychnine are shown superimposed on the structure in Figure 1.5, and it should now be obvious where the euphemistic label "spaghetti diagram" mentioned earlier came from.
When the data contained in the 8 Hz HMBC spectrum of strychnine are interpreted, correlation diagrams can be constructed that are limited to the number of bonds involved in the correlation. First, consider Figure 1.6, which highlights only the 2JCH correlations.
Going from the manageable number of 2JCH correlations shown on the structure in Figure 1.6 to Figure 1.7, which shows all of the 3JCH correlations, the level of complexity in the interpretation of HMBC data becomes more apparent. There are significantly more 3JCH than 2JCH correlations in the typical HMBC spectrum. Usefully, however, 3JCH correlations span hetero-atoms incorporated into the skeletal framework and also make it possible to begin linking structural moieties together that would not generally be possible from HSQC and COSY data alone.
The rigid skeletal framework of strychnine facilitates a significant number of 4JCH correlations. Indeed, the number of 4JCH correlations observed is essentially identical with the number of 2JCH correlations (Figure 1.8). As molecules become more complex, they also in many cases become more proton deficient. When the ratio of hydrogens to heavy atoms (C, N, O, S) falls below 2, a postulate known as the "Crews Rule" was suggested, which states that such a hydrogen to heavy atom ratio may render the structure of a molecule difficult and in some cases impossible to deduce. However the Crews rule was based on what were "standard" data sets acquired for structure elucidation at the time that it was suggested. Now, there are a number of experiments available with longer "reach" that probably mandate adjusting the Crews Rule ratio downwards.
It is interesting that for strychnine, there are also a significant number of 5JCH correlations, as shown in Figure 1.9. In part, the large number of 4JCH and 5JCH correlations can be attributed to the acquisition of these data using a cryogenic NMR probe, for this example a 600 MHz Bruker TXI 1.7 mm gradient triple resonance MicroCryoProbe.
As noted in the caption of Figure 1.8, in 2014 Williamson and co-workers developed the LR-HSQMBC experiment. That experiment is a refocused single quantum-based long-range experiment. The refocusing facilitates one-band decoupling during acquisition as in the D-HMBC experiment and refocusing also prevents small heteronuclear couplings from being antiphase at the end of the pulse sequence as in the HMBC experiment. The antiphase character of weak long-range correlations can lead to their cancellation when HMBC data are magnitude calculated for presentation and interpretation. In contrast, LR-HSQMBC spectra are phase sensitive. Generally, LR-HSQMBC should be considered a "second-tier" long-range heteronuclear shift correlation experiment. HMBC data should be utilized to "prune" the number of correlations in the LR-HSQMBC spectrum to just those that were not observable in an HMBC experiment, which will typically be the longer range, smaller correlations. In the initial investigation of the LR-HSMQBC experiment, cervinomycin A2, which is quite proton deficient, was used as a model compound. From DFT calculations carried out in conjunction with the NMR investigation, heteronuclear coupling constants <0.5 Hz were readily visualized by the LR-HSQMBC experiment when it was optimized at 2 Hz. The availability of such correlation data for very small coupling constants was later shown to have a significant impact on calculation times for the Structure Elucidator CASE program when the 2 Hz LR-HSQMBC data were included in the data input file.
When a 2 Hz optimized LR-HSQMBC spectrum of strychnine was com- pared with the 8 Hz HMBC data that we have been using as an example, constructing a connectivity diagram comprised of only those correlations seen in one experiment or the other gives the following result. Clearly, there are some correlations (red arrows in Figure 1.10) that are observed in the HMBC data that are not observed, for whatever reason, in the 2 Hz LR-HSQMBC spectrum. In contrast, however, there are significantly more correlations due to very small coupling constants in the LR-HSQMBC spectrum than were observed in the HMBC data.
(Continues...)
Excerpted from Modern NMR Approaches to the Structure Elucidation of Natural Products by Antony Williams, Gary E. Martin, David Rovnyak. Copyright © 2016 The Royal Society of Chemistry. Excerpted by permission of The Royal Society of Chemistry.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.