There has been considerable interest over the past decade in the utilization of direct and long-range 1H-15N heteronuclear shift correlation methods at natural abundance to facilitate the elucidation of small molecule structures. Recently, there has also been a high level of interest in the exploration of indirect covariance NMR methods. Our initial explorations in this area led to the development of unsymmetrical indirect covariance methods, which allow the calculation of hyphenated 2D NMR spectra such as 2D GHSQC-COSY and GHSQC-NOESY from the discrete component 2D NMR experiments. We now wish to report the utilization of unsymmetrical indirect covariance NMR methods for the combination of 1H-13C GHSQC and 1H-15N long-range (GHMBC, IMPEACH-MBC, CIGAR-HMBC, etc.) heteronuclear chemical shift correlation spectra to determine 15N-13C correlation pathways.

1. Utilizing Unsymmetrical Indirect Covariance Processing
to Define 15N-13C Connectivity Networks
Gary E. Martin,* Patrick A. Irish, and Bruce D. Hilton
Schering-Plough Research Institute
Rapid Structure Characterization Laboratory
Pharmaceutical Sciences
Summit, NJ 07901
Kirill A. Blinov
Advanced Chemistry Development
Moscow Department
Moscow 117513
Russian Federation
Antony J. Williams
Advanced Chemistry Development
Toronto, Ontario
Canada
15
Running title: N-13C Unsymmetrical Indirect Covariance NMR
Keywords: unsymmetrical indirect covariance NMR, 1H-13C GHSQC, 1H-15N
GHMBC, IMPEACH-MBC, CIGAR-HMBC, 15N-13C connectivity

2. Abstract:
There has been considerable interest over the past decade in the utilization of direct and
long-range 1H-15N heteronuclear shift correlation methods at natural abundance to
facilitate the elucidation of small molecule structures. Recently, there has also been a
high level of interest in the exploration of indirect covariance NMR methods. Our initial
explorations in this area led to the development of unsymmetrical indirect covariance
methods, which allow the calculation of hyphenated 2D NMR spectra such as 2D
GHSQC-COSY and GHSQC-NOESY from the discrete component 2D NMR
experiments. We now wish to report the utilization of unsymmetrical indirect covariance
NMR methods for the combination of 1H-13C GHSQC and 1H-15N long-range (GHMBC,
IMPEACH-MBC, CIGAR-HMBC, etc.) heteronuclear chemical shift correlation spectra
to determine 15N-13C correlation pathways.
2

3. 1
Numerous studies utilizing H-15N heteronuclear chemical shift correlation
experiments have been reported over the past decade and are the subject of several recent
reviews.1-5 Underscoring the high level of interest and growing importance of being able
to access 1H-15N long-range heteronuclear shift correlation data, there have been two
recent reports6,7 describing pulse sequences that allow the simultaneous acquisition of
long-range 1H-13C and 1H-15N HMBC data. The ability to exploit 1H-15N connectivities
is an important adjunct to small molecule structure elucidation, particularly for
pharmaceuticals and alkaloids. Considerable recent attention has also been focused on
the development of covariance NMR methods, initially by Brüschweiler and co-
workers.8-13 Insofar as potential for small molecule application, the most interesting
report described the capability of extracting carbon-carbon connectivity information from
a GHSQC-TOCSY spectrum using indirect covariance methods. Brüschweiler and
Zhang commented that proton resonance overlap could lead to artifacts in the calculated
carbon-carbon spectra but did not explore this observation. 11 We subsequently reported
the analysis of two types of artifacts observed in inverted direct response (IDR) GHSQC-
TOCSY spectra with overlapped proton resonance, which, in turn, prompted us to
explore the elimination of these artifacts via a method that we have named unsymmetrical
indirect covariance.14 Subsequent work has shown that it is also possible to
mathematically combine various discretely acquired 2D NMR spectra. The first effort in
this direction demonstrated the combination of 1H-13C GHSQC and GHMBC spectra to
afford the equivalent of an m,n-ADEQUATE spectrum.15 Subsequent studies have
demonstrated the calculation of GHSQC-COSY16 and GHSQC-NOESY17 spectra from
discretely acquired COSY, NOESY, and 1H-13C GHSQC 2D NMR spectra. Recently,
3

4. Kupče and Freeman7 have demonstrated the use of projection reconstruction techniques
15
to establish N-13C correlations at natural abundance, using vitamin B-12 as a model
compound for the study, also noting in parallel that indirect covariance methods can be
used to obtain homonuclear correlation spectra indirectly. Thus, we would now like to
demonstrate specifically, that unsymmetrical indirect covariance NMR methods can
15
indeed be used to derive N-13C connectivity information from discretely acquired 1H-
13
C GHSQC and 1H-15N HMBC spectra.
1
For the present study, multiplicity-edited H-13C GHSQCAD and 1
H-15N
GHMBC spectra were acquired using an ~5 mg sample of strychnine (1) dissolved in
~180 μL CDCl3 in a 3 mm NMR tube. The spectra were recorded at 500 MHz using a
Varian two channel spectrometer equipped with a 3 mm gradient inverse triple resonance
probe. The 1H-13C GHSQCAD spectrum was recorded in 36 m while the 1H-15N
GHMBC spectrum was recorded in 3 h 6 m. The spectra were acquired with identical
proton spectral widths and the data were processed to yield data matrices that were
identically digitized with F1,F2 dimensions of 512 x 2048 points.
The unsymmetrical indirect covariance matrix can be calculated by
C = RN * RCT [1]
where RN and RC correspond to the real data matrices from the long-range 1H-15N
GHMBC and 1H-13C multiplicity-edited GHSQCAD spectra, respectively. In the present
report, the GHSQCAD data are plotted with CH and CH3 resonances with positive phase
and CH2 resonances with negative phase. The 1H-13C data matrix is transposed to RCT
4

5. during processing. The data were acquired and processed so that there were equal
numbers of columns in the data sets, i.e. RN is N * M1 and RC is N * M2 to allow the
multiplication of the data matrices. In the present example, F2 spectral widths were
identical although that is not an absolute requirement. By definition, the following
formula is used to calculate each element Cij (i and j are row indices in the initial
matrices, correspondingly, RN and RC) of data matrix C:
Cij = (RN)ij * (RC)ij = (RN)i1 * (RC)j1 + (RN)i2 * (RC)j2 + (RN)ik * (RC)jk + (RN)iN * (RC)jN [2]
Each element of matrix C is the sum of products of values (RN)ik and (RC)jk. A
necessary condition is to have non-zero elements in equivalent positions in the rows of
(RN)i and (RC)j. For two “ideal” 2D NMR spectra, assuming zero noise in the data
matrices, the sum of a matrix element will be non-zero when rows (RN)i and (RC)j have
crosspeaks in the same position.
The resulting GHSQCAD and GHMBC data sets were subjected to
15
unsymmetrical indirect covariance processing to yield the N-13C long-range correlation
spectrum shown in Figure 1. The 1H-15N HMBC spectrum is shown in the top left panel;
the 1H-13C GHSQCAD spectrum is shown in the bottom right panel and has been
13 15
transposed to reflect the orientation of the C spectrum as the F2 axis in the N-13C
15
correlation spectrum shown in the top right panel. In the N-13C correlation spectrum,
15 13
N Chemical shift information is displayed in the F1 frequency domain while C
15
chemical shift information is presented in the F2 frequency domain. N-13C correlation
responses in the spectrum shown in Figure 1 are “modulated” via the nJNH correlation
5

6. response in the 1H-15N HMBC spectrum, the 13
C chemical shift information deriving
from the chemical shift of the carbon directly bound to the proton in question in the 1H-
13
C GHSQCAD spectrum.
39.6
18 N 20
19
17 16
7 15
8 14 21
N 13 22
155.2 9
10 12 23
O 11 O
1
15
N-13C long-range heteronuclear correlations observed in Figure 1 are shown on
15
the structure above. Correlations plotted using red contours in the N-13C correlation
spectrum shown in Figure 1 arise from correlations between methylene carbons and the
nitrogen; correlations plotted in black arise from correlations from methine (or methyl,
although there are no methyl resonances in the structure of strychnine) carbons to
nitrogen. The phase of the 13C resonance from the multiplicity-editing of responses in the
15
GHSQCAD spectrum is carried forward into the N-13C unsymmetrical indirect
covariance processed spectrum shown in the top right panel in Figure 1. Presumably,
13
methine and methylene carbons with the same C chemical shift that correlate to the
same nitrogen could partially or completely cancel, hence it may useful to consider the
acquisition of both conventional and multiplicity-edited 1H-13C GHSQCAD spectra when
6

7. the data are acquired since these data can be accumulated in a very reasonable periods of
time. An interesting corollary arises in the case of overlapped protons, one of which is
15
long-range correlated to a nitrogen resonance. By calculating the N-13C correlation
spectrum, as shown in Figure 1, the specific proton (via the 1H/13C response in the
15 15
GHSQC spectrum) correlating to the N resonance can be determined by the N-13C
correlation response.
15
In conclusion, N-13C heteronuclear chemical shift correlation spectra can be
derived through the unsymmetrical indirect covariance processing of 1H-13C GHSQCAD
and 1H-15N long-range heteronuclear chemical shift correlation spectra ( 1H-15N HMBC in
the present example, but we have obtained the same results with 1H-15N IMPEACH-
MBC18 and 1
H-15N CIGAR-HMBC19 experiments). The utilization of 15
N-13C
heteronuclear chemical shift connectivity information may prove useful in the structural
characterization of pharmaceuticals (>80% contain nitrogen in their structures) and
alkaloids as well as in the characterization of other nitrogenous compounds. These data
may also allow investigators to differentiate between overlapped protons, one of which is
15
long-range coupled to N, provided that the carbons directly bound to the overlapped
protons are resolved. Generally, it is highly unlikely that both the proton and carbon
resonances for two sites in a molecule will have identical 1H and 13C chemical shifts. It
will be quite interesting to see what other applications for 15N-13C heteronuclear chemical
shift correlation data will arise. In a parallel vein, it will be equally interesting to see
what other types of data can be beneficially co-processed using unsymmetrical indirect
covariance NMR methods. These investigations are presently underway in our
laboratories and will form the basis of forthcoming reports.
7

8. References:
1. G. E. Martin and C. E. Hadden, J. Nat. Prod., 65, 543 (2000).
2. R. Marek and A. Lycka, Curr. Org. Chem., 6, 35 (2002).
3. G. E. Martin and A. J. Williams, “Long-Range 1H-15N 2D NMR Methods,” in
Ann. Rep. NMR Spectrosc., vol. 55, G. A. Webb, Ed., Elsevier, Amsterdam, 200,
pp. 1-119.
4. P. Forgo, J. Homann, G. Dombi, and L. Máthé, “Advanced Multidimensional
NMR Experiments as Tools for Structure Determination of Amaryllidaceae
Alkaloids,” in Poisonous Plants and Related Toxins, T. Acamovic, S. Steward and
T. W. Pennycott, Eds., Wallingford, UK, 2004, pp. 322-328.
5. G. E. Martin, M. Solntseva, and A. J. Williams, “Applications of 15N NMR in
Alkaloid Chemistry,” in Modern Alkaloids, E. Fattorusso and O. Taglialatela-
Scafati, Eds., Wiley-VCH, 2007, in press.
6. M. Pérez-Trujillo, P. Nolis, and T. Parella, Org. Lett., 9, 29 (2007).
7. E. Kupče and R. Freeman, Magn. Reson. Chem., 45, 103 (2007).
8. R. Brüschweiler and F. Zhang, J. Chem. Phys., 120, 5253 (2004)
9. R. Brüschweiler, J. Chem. Phys., 121, 409 (2004).
10. F. Zhang and R. Brüschweiler, Chem. Phys. Chem., 5, 794 (2004).
11. F. Zhang, and R. Brüschweiler, J. Am. Chem. Soc., 126, 13180 (2004).
12. N. Trbovic, S. Smirnov, F. Zhang, and R. Brüschweiler, J. Magn Reson., 171, 177
(2004).
13. Y. Chen, F. Zhang, W. Bermel, and R. Brüschweiler, J. Am. Chem. Soc., 126,
15564 (2006).
8

9. 14. K. A. Blinov, N. I. Larin, M. P. Kvasha, A. Moser, A. J. Williams, and G. E.
Martin, Magn. Reson. Chem., 43, 999 (2005).
15. K. A. Blinov, N. I. Larin, A. J. Williams, M. Zell, and G. E. Martin, Magn. Reson.
Chem., 44, 107 (2006).
16. K. A. Blinov, N. I. Larin, A. J. Williams, K. A. Mills, and G. E. Martin, J.
Heterocyclic Chem., 43, 163 (2006).
17. K. A. Blinov, A. J. Williams, B. D. Hilton, P. A. Irish, and G. E. Martin, Magn.
Reson. Chem., 45, in press (2007).
18. C. E. Hadden, G. E. Martin, and V. V. Krishnamurthy, J. Magn. Reson., 140, 274
(1999); G. E. Martin and C. E. Hadden, Magn. Reson. Chem., 38, 251 (2000).
19. C. E. Hadden, G. E. Martin, and V. V. Krishnamurthy, Magn. Reson. Chem., 38,
143 (2000).
9

10. 15
N-13C unsymm. indirect covariance
40 N19 40
C16 C18 C20 C17 C14 C15
60 60
F1 Chemical Shift (ppm)
F1 Chemical Shift (ppm)
80 80
15
100 100
N
120 120
140 C8 C13 C11 140
N9
4.0 3.5 3.0 2.5 2.0 1.5 1.0 65 60 55 50 45 40 35 30 25 20
F2 Chemical Shif t (ppm) F2 Chemical Shif t (ppm)
1 13
H-15N GHMBC C
1.5
F2 Chemical Shift (ppm)
2.0
multiplicity-edited 2.5
1
H-13C GHSQC 3.0
3.5
4.0
65 60 55 50 45 40 35 30 25 20
F1 Chemical Shif t (ppm)

11. 15
Figure 1. N-13C (F1,F2) Heteronuclear chemical shift correlation spectrum (upper right panel) derived by the unsymmetrical
indirect covariance processing of discretely acquired, multiplicity-edited 1H-13C GHSQCAD (bottom right panel,
transposed to reflect the orientation of the 13C chemical shift axis in the final 15N-13C correlation spectrum CH/CH3
resonances are plotted in black and have positive phase, CH2 resonances are inverted and plotted in red) and the 1H-
15
N GHMBC spectrum (top left panel). The data were recorded at 500 MHz using an ~5 mg sample of strychnine in
180 μL CDCl3. Total data acquisition time for the two 2D spectra was < 4 h. The individual 2D NMR spectra were
processed to yield a pair of spectra comprised of 512 x 2048 points that were the subjected to unsymmetrical indirect
covariance processing. Processing time was approximately 4 sec. The N9 and N19 resonances of strychnine (1) are
observed at 155.2 and 39.6 ppm, respectively. Correlations are observed, as expected, between the C8, C11, and C13
resonances and the N9 amide nitrogen resonance. Correlations are observed from the C14, C15, C16, C17, C18, and
C20 resonances to the N19 aliphatic nitrogen resonance. There is no way to differentiate between the overlapped C8
and C16 resonances in the 15N-13C correlation spectrum, although the H8 and H16 resonances are resolved in the 500
MHz 1H-15N GHMBC spectrum, allowing an unequivocal assignment to be made.
11