Anal. Chem. 2010, 82, 3266–3269
Demixing of Severely Overlapping NMR Spectra through Multiple-Quantum NMR Manjunatha Reddy G. N. and Stefano Caldarelli* Universite´ Paul Ce´zanne (Aix-Marseille III) Equipe Chimiome´trie et Spectroscopie, Institut des Sciences Mole´culaires de Marseille ISM2-UMR-CNRS-6263. Faculte´ des Sciences et Techniques, Service 512 13397 Marseille Cedex 20 France We introduce an NMR method to help in the analysis of complex mixtures. The spectra of molecular fragments are obtained as the traces of a correlation spectrum of the regular 1H NMR spectrum on one dimension with the one of the highest possible 1H multiple-quantum (MaxQ) order. As this latter is a function of the number of distinguishable protons in a given molecular fragment, the analysis of a series of multiple-quantum spectra is required to achieve a complete assignment. This MaxQ NMR approach is likely to perform best in the case of signals concentrated in a very narrow frequency range, which is a challenging situation commonly encountered in many relevant analytical problems such as the characterization of extraction fractions (oil, plants, tissues), biological fluids, or environmentally relevant samples. As a demonstration, we apply the MaxQ NMR analysis to a mixture of 11 poly- and monocyclic aromatic hydrocarbons. The characteristic high resolution of molecular NMR associated with the use of sophisticated multidimensional experiments has allowed the investigation of molecular systems of remarkable complexity. Nonetheless, the limits of current methods are put to the test by the necessity of analyzing increasingly larger and complicated systems, characterized by important spectral overlap. Examples of particularly crowded proton NMR spectra are observed in biomacromolecules and perhaps even more dramatically in multicomponent mixtures such as extract fractions, biological fluids, or environmentally relevant samples. Overlapping resonances are often generated by distinct molecular fragments, localized either on the same molecule, particularly for large molecular objects, or in different ones, in the case of mixtures. The isolation of the subspectra of each of these spin systems is an effective strategy for the analysis of such crowded NMR spectra. This can be achieved, typically, by NMR-based diffusion measurements, if the fragments possess or are induced with different translational diffusivities,1-3 or by correlation multidimensional experiments,4 to help detect isolated spin systems, or * To whom correspondence should be addressed. E-mail: s.caldarelli@ univ-cezanne.fr. (1) Johnson, C. S., Jr. Prog. Nucl. Magn. Reson. Spectrosc. 1999, 34, 203–256. (2) Viel, S.; Ziarelli, F.; Caldarelli, S. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 9696–9698. (3) Caldarelli, S. Magn. Reson. Chem. 2007, 45, S48–S55. (4) Lin, M.; Shapiro, M. J. Anal. Chem. 1997, 69, 4731–4733.
3266
Analytical Chemistry, Vol. 82, No. 8, April 15, 2010
by a combination of the two.5-7 If the analysis concerns a series of samples, methods based on covariance analysis8 have been proposed as an alternative9 to improve the resolution of the spectra, although its performance in the case of overlapping peaks is not assured. We describe here a purely NMR approach that improves the capability of individuating the single molecular constituents in a mixture. The method capitalizes on the simplified spectra associated with high-quantum order coherences, up to the limiting case of the maximum-quantum (MaxQ) observable coherence, which is always a unique singlet. Although forbidden according to quantummechanical selection rules, the simplifying role of recording NMR multiple-quantum (MQ) transitions has long been recognized.10-14 While homonuclear MQ spectroscopy has found limited applications in 1H NMR, most notably for spin-counting,15 it has flourished in 13 C analysis with INADEQUATE and its variants.16,17 Common utilizations of MQ coherences are for filtering out specific NMR signals18,19 or using combinations of MQ-orders to produce simpler correlation spectra, as in e-COSY.20 Elaborate spectral filters exploiting high-order coherences have been developed to select specific spin topologies.21 Recently, 1H MQ NMR has been used to simplify spectra including heteronuclear couplings by relying on spin state selection for molecules.22-24 (5) Barjat, H.; Morris, G. A.; Swanson, A. G. J. Magn. Reson. 1998, 131, 131– 138. (6) Vitorge, B.; Jeanneat, D. Anal. Chem. 2006, 78, 5601–5606. (7) Viel, S.; Caldarelli, S. Chem. Commun. 2008, 2013–2015. (8) Eads, C. D.; Noda, I. J. Am. Chem. Soc. 2002, 124, 1111–1118. (9) Cloarec, O.; Dumas, M.-E.; Craig, A.; Barton, R. H.; Trygg, J.; Hudson, J.; Blancher, C.; Gauguier, D.; Lindon, J. C.; Holmes, E.; Nicholson, J. Anal. Chem. 2005, 77, 1282–1289. (10) Munowitz, M.; Pines, A. Science 1986, 233, 525–531. (11) Munowitz, M.; Pines, A. Adv. Chem. Phys. 1987, 66, 1–152. (12) Norwood, T. J. Prog. Nucl. Magn. Reson. Spectrosc. 1992, 24, 295–375. (13) Norwood, T. J. Encyclopedia of Magnetic Resonance; John Wiley & Sons: Chichester; 2007, pp 3181-3188. (14) Sinton, S.; Pines, A. Chem. Phys. Lett. 1980, 76, 263–267. (15) Hughes, C. E. Prog. Nucl. Magn. Reson. Spectrosc. 2004, 45, 301–313. (16) Bax, A.; Freeman, R.; Frenkiel, T. A. J. Am. Chem. Soc. 1981, 103, 2102– 2104. (17) Lesage, A.; Auger, C.; Caldarelli, S.; Emsley, L. J. Am. Chem. Soc. 1997, 119, 7867–7868. (18) Piantini, U.; Sorensen, O. W.; Ernst, R. R. J. Am. Chem. Soc. 1982, 104, 6800–6801. (19) Shaka, A. J.; Ray, F. J. Magn. Reson. 1983, 51, 169–173. (20) Griesinger, C.; Soerensen, O. W.; Ernst, R. R. J. Am. Chem. Soc. 1985, 107, 6394–6396. (21) Levitt, M. H.; Ernst, R. R. Chem. Phys. Lett. 1983, 100, 119–123. (22) Baishya, B.; Reddy, G. N. M.; Prabhu, U. R.; Row, T. N. G.; Suryaprakash, N. J. Phys. Chem. A 2008, 112, 10526–10532. 10.1021/ac100009y 2010 American Chemical Society Published on Web 03/23/2010
An underexploited feature of the Maximum-Quantum coherence order of a spin system is that it involves and summarizes the contribution of all the participating spins. Thus, a MaxQ-1Q correlation spectrum produces an effective demixing of the spectral components. This feature should be most practical in the analysis of mixtures and of macromolecules and is the key of the experiment proposed here. METHODS Sample Preparation. The test mixture was prepared by dissolving the selected molecules (naphthalene, antharacene, phenanthrene, fluoranthane, triphenylene, o-terphenyl, phenol, acetanilide, benzophenone, biphenyl, and dibenzyl) in CDCl3, which was taken from a new bottle without further purification. The concentration of each solute was of about 5 mM. NMR Spectroscopy. All the experiments were performed at room temperature using a Bruker Avance-500 NMR spectrometer equipped with a cryogenic probe capable of generating gradients fields up to 55 G/cm. 1H multiple-quantum coherences were excited and detected with the basic pulse sequence (π/2)-τ(π)-τ-(π/2)φ-t1-(π/2)-t2, where the value of τ is optimized empirically on a one-dimensional version of the sequence (i.e., with a fixed t1 value at 3 µs) to obtain the best intensity of the MQ coherence of choice, p. The phase φ is chosen to select odd or even MQ orders. Unwanted coherences were effectively dephased using a couple of sine-shaped pulse field gradients, G1 and G2, placed before and after the last π/2 pulse, respectively. The ratio of the gradient pulse was selected to fulfill G2 ) p × G1,25 with G2 being kept fixed at 20.79 G/cm and G1 varied accordingly. The indirect dimension was described by acquiring 256 t1 increments, for 62 min total acquistion time. The 2D spectra were processed to produce a 512 × 1024 matrix and are displayed in magnitude mode. The indirect (MQ) dimension was plotted with a reduced scale δ1R ) δ1/p. With this choice, the singlet in the MaxQ dimension is found at the average position of the 1H signals participating in its creation, p
R ) δmaxQ
∑ i)1
δi1Q p
RESULTS AND DISCUSSION We analyzed, as a proof of principle, a set of 11 poly- and monocyclic aromatic hydrocarbons: naphthalene, antharacene, phenanthrene, fluoranthane, triphenylene, o-terphenyl, phenol, acetanilide, benzophenone, biphenyl, and dibenzyl, which corresponds to 104 aromatic protons for 28 distinguishable chemical shifts. The region of the aromatic signals of the 1H 1D NMR spectrum is shown in Figure 1. The protons resonate in a span of about 2 ppm, with about 80% of the signals concentrated in the 7.1-8.1 ppm region. The size of the aromatic proton spin systems in this test set varies from two to five protons, which fixes the highest number of excitable quanta in this case to five. The series of all possible (23) Baishya, B.; Suryaprakash, N. J. Chem. Phys. 2007, 127, 214510. (24) Reddy, G. N. M.; Row, T. N. G.; Suryaprakash, N. J. Magn. Reson. 2009, 196, 119–126. (25) Keeler, J.; Clowes, R. T.; Davis, A. L.; Laue, E. D. In Nuclear Magnetic Resonance, Pt C; Academic Press Inc.: San Diego, CA, 1994; Vol. 239, pp 145-207.
Figure 1. 1H NMR spectrum of the 11 molecule test mixture described in the text. The total number of protons is 104, with 28 different chemical shifts.
MQ-1Q correlation spectra for this system is shown in Figure 2. A progressive simplification of the correlation peaks pattern is observed along the spectral series, due to the progressive collapse of the resonances in the MQ dimension and to the filtering out of spin systems incapable of producing a given MQ coherence. All spectra were recorded with the same experimental time and thus have similar sensitivities. The survival of a good sensitivity with increasing quantum number along this series of spectra is somewhat unexpected as the higher quantum coherence order are more difficult to excite. However, this can be understood as the composition of two competing effects: the decrease of the efficiency of the MQ excitation with the MQ order and the progressive reduction of the spread of the signals in the 2D diagram, which produces a signal enhancement. The 1Q-1Q correlation spectrum provides information and complexity equivalent to a correlation spectroscopy (COSY) spectrum and was not analyzed further. At the other end of the series lies the 5Q-1Q correlation diagram, corresponding to MaxQ for the six fragments with five protons, which is the simplest spectrum. Figure 3a illustrates as MaxQ correlations can be used to identify the pure spectra of the mixture components, exploiting the property that all protons in the fragment contribute to a unique signal in the MaxQ dimension. The information in the MaxQ correlation is thus tantamount to total correlation spectroscopy (TOCSY) but with a simplified layout that enhances the resolution by a large factor. The single resonance position of each fragment in the MaxQ dimension is located, in the reduced scale used here, at the average frequency of the participating protons. Thus, the 1 H spectra of two spin systems can be separated by this method if the difference of their average chemical shift exceeds the resolution of the MaxQ dimension. In the examples shown here, the signal lifetime in the indirect dimension was of the order or longer of the acquisition time in the MQ dimension. The length of this latter was thus the factor limiting the resolution. The multiplet structure is maintained in the MQ correlation spectra, which can be used to extract J Analytical Chemistry, Vol. 82, No. 8, April 15, 2010
3267
Figure 2. Series of MQ-1Q correlation spectra of the test mixture, with the coherence order p ranging from 1 to 5 (A-E). The MQ dimension uses a reduced frequency scale δ1R equals to δ1/p, with p as the coherence order. In this reduced scale, the MaxQ signal appears at the average of the resonance values of the protons that contribute to its creation. The spectral complexity reduces with the number of quanta, due to a progressive simplification of the spectrum in the MQ dimension and to the filtering out of some molecular fragments (see text). The 5Q-1Q correlation presents only MaxQ signals.
Figure 3. (A) Expansion of the relevant region of the 5Q-1Q correlation spectrum and (B) details of the MaxQ signal of benzophenone. The 1D 1H spectrum is also shown for comparison of the multiplet structure.
Figure 4. (A) Expansion of the relevant region of the 4Q-1Q spectrum. MaxQ traces are indicated by arrows, while (MaxQ-1). (B) Details of the MaxQ signal of phenanthrene. The 1D 1H spectrum is also shown for comparison of the multiplet structure.
couplings (Figure 3b). With this information it was thus possible to assign the MaxQ signal to the 5-proton molecular fragments of phenol (δ1R ) 6.96) o-terphenyl (δ1R ) 7.13), dibenzyl (δ1R ) 7.18), acetanilide (δ1R ) 7.29), biphenyl (δ1R ) 7.43), and benzophenone (δ1R ) 7.59). Roughly half of the molecules of the mix did not appear in the 5Q spectrum, as they were filtered out due to an insufficient size of their spin systems, and thus the lower MQ orders spectra must be analyzed for completeness. The peaks in the 4Q correlation
spectrum (Figure 4) arise from two different kinds of molecular fragments, with 4 and 5 protons. For the former ones, 4Q is the MaxQ, and thus a single correlation trace links the 1Q and the 4Q dimension. Conversely, 5-spin fragments possess several (up to five distinguishable) 4Q coherences, and thus their signal is more complex but easily recognized on the basis of the knowledge extracted from the 5Q spectrum (see boxes in Figure 4). The 4Q spectrum allows the unambiguous identification of six more 4-proton molecular fragments: o-terphenyl (δ1R ) 7.15), fluoran-
3268
Analytical Chemistry, Vol. 82, No. 8, April 15, 2010
thene (δ1R ) 7.70), naphthalene (δ1R ) 7.72), anthracene (δ1R ) 7.78), phenanthrene (δR1 ) 8.01), and triphenylene (δR1 ) 8.21). Portions of all molecules in the test set are identified already at this level. The fragments with two or three protons can similarly be identified in the 2Q and 3Q spectrum, completing the identification of all aromatic spin systems. Note that the composition and number of molecules of the mixture considered here resemble typical chromatographic test sets for the analysis of aromatic hydrocarbons. Therefore, the resolving power of the proposed approach has the potential to be comparable, within its range of applicability, to chromatographic methods. It is noteworthy that the experimental efficiency observed allows already reasonable recording times, for S/N ratios in excess of 10. In the case discussed here, these latter were dominated by the requested resolution in the MQ dimensions and thus can be further compressed, in principle, using faster acquisition/process(26) Frydman, L.; Scherf, T.; Lupulescu, A. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 15858–15862. (27) Kupce, E.; Nishida, T.; Freeman, R. Prog. Nucl. Magn. Reson. Spectrosc. 2003, 42, 95–122. (28) Bruschweiler, R.; Zhang, F. L. J. Chem. Phys. 2004, 120, 5253–5260. (29) Richard, R.; Ernst, G. B.; Wokaun, A. Principles of Nuclear Magnetic Resonance in One and Two Dimensions; Clarendon Press: Oxford, U.K., 1990. (30) Glaser, S. J. J. Magn. Reson., Ser. A 1993, 104, 283–301. (31) Khaneja, N.; Kramer, F.; Glaser, S. J. J. Magn. Reson. 2005, 173, 116–124.
ing schemes.26-28 Finally, it should be noted that the efficiency of the excitation of a given multiple quantum coherence relies on the specificity of the J-coupling networks,29 as for many other familiar NMR experiments involving single-quantum coherences (COSY, TOCSY, etc.).30 Optimal and as uniform as possible excitation schemes should be thus devised for multiple-quantum excitation-reconversion as they were for the single-quantum polarization transfer experiments.31 For example, in the experiment shown here, the 5Q efficiency varied between 5 and 25% of the 1Q spectrum. Thus, and as it has been the case for other experiments, the usefulness of MaxQ in the case of specific mixtures of interest (for example, biofluids) should be assessed by building a database of spectra of the relevant molecules. ACKNOWLEDGMENT We are grateful to the Agence National de la Recherche (ANR) for support under Grant ANR-08-BLAN-273-01 and to Spectropoˆle (Fe´de´ration des Sciences Chimique FR 1739) for privileged spectrometer time.
Received for review January 3, 2010. Accepted March 5, 2010. AC100009Y
Analytical Chemistry, Vol. 82, No. 8, April 15, 2010
3269
