138
Research Article
Received: 21 August 2007 Revised: 12 October 2007 Accepted: 16 October 2007 Published online in Wiley Interscience: 20 December 2007
(www.interscience.com) DOI 10.1002/mrc.2141
Using indirect covariance spectra to identify
artifact responses in unsymmetrical indirect
covariance calculated spectra
Gary E. Martin,a∗ Bruce D. Hilton,a Kirill A. Blinovb and Antony J. Williamsc
Several groups of authors have reported studies in the areas of indirect and unsymmetrical indirect covariance NMR processing
methods. Efforts have recently focused on the use of unsymmetrical indirect covariance processing methods to combine various
discrete two-dimensional NMR spectra to afford the equivalent of the much less sensitive hyphenated 2D NMR experiments, for
example indirect covariance (icv)-heteronuclear single quantum coherence (HSQC)-COSY and icv-HSQC–nuclear Overhauser
effect spectroscopy (NOESY). Alternatively, unsymmetrical indirect covariance processing methods can be used to combine
multiple heteronuclear 2D spectra to afford icv-13C– 15N HSQC-HMBC correlation spectra. We now report the use of responses
contained in indirect covariance processed HSQC spectra as a means for the identification of artifacts in both indirect covariance
and unsymmetrical indirect covariance processed 2D NMR spectra. Copyright c© 2007 John Wiley & Sons, Ltd.
Keywords: NMR; indirect covariance; unsymmetrical indirect covariance; artifact forecasting; GHSQC-TOCSY; GHSQC-COSY
Introduction
The first reported application of indirect covariance processing
of heteronuclear 2D NMR data was in the late 2004 report
of Zhang and Bru¨schweiler who applied the method to a
gradient heteronuclear single quantum coherence GHSQC-TOCSY
spectrum to produce an indirect covariance (icv)-13C–13C-COSY
(icv) connectivity plot.[1] A footnote in that 2004 reportmentioned
the possibility of proton resonance overlap causing artifacts
in the icv-13C–13C-COSY connectivity plots but the authors
did not elaborate further on that observation. Using several
model compounds, in 2005 the present authors reported the
analysis of two types of artifacts in icv-13C–13C-COSY connectivity
plots derived from indirect covariance processing of inverted
direct response (IDR)-GHSQC-TOCSY spectra.[2] In an effort to
eliminate artifact responses, the indirect covariance processing
algorithm was modified to allow two 2D NMR data matrices
to be coprocessed. By decomposing an IDR-GHSQC-TOCSY
spectrum into the correspondingpositively andnegativelyphased
subspectra and then processing them back together using the
unsymmetrical indirect covariance processing algorithm one type
of artifact was eliminated and the other was rendered diagonally
asymmetric, thereby allowing the second type of artifact response
to be eliminated by symmetrization of the icv-13C–13C-COSY
connectivity plot.
While the elimination of artifacts from icv-13C–13C-COSY
connectivityplots isausefulattributeof theunsymmetrical indirect
covariance processing algorithm, a far more useful capability
resides in an investigator’s ability touse thealgorithmtocoprocess
discretely acquired 2D NMR spectra. Examples have included:
coprocessing 1H–13C GHSQC and GHMBC spectra to yield
the equivalent of an m,n-ADEQUATE spectrum[3]; coprocessing
GHSQC and GCOSY spectra to yield icv-HSQC-COSY spectra[4,5];
coprocessing GHSQC and nuclear Overhauser effect spectroscopy
(NOESY) data to produce the equivalent of icv-HSQC-NOESY
spectra[6]; and coprocessing 1H–13C GHSQC and various 1H–15N
S
1514
1011
13
12
8
N
6
5
16
17
4
3
1
2
1
Scheme 1. Structure of naphtho[2′,1′ : 5,6]naphtho [2′,1′ : 4,5]thieno[2,3-
c]quinoline.
long-range correlation spectra to derive icv-13C–15N HSQC-HMBC
(the notation HSQC-HMBC defines the heteronuclides involved,
13C in the case of the HSQC portion of the acronym and 15N
for the HMBC portion) heteronuclear shift correlation spectra.[7–9]
It should be noted that the unsymmetrical indirect covariance
processingmethod isnot restrictedtogradient-basedexperiments
∗ Correspondence to: Gary E. Martin, Schering Plough Research Institute, Rapid
Structure Characterization Laboratory, Mail Stop S7-D1-1, 556 Morris Ave,
Summit, NJ 07901, USA. E-mail: gary.martin@spcorp.com
a Schering-Plough Research Institute, Rapid Structure Characterization Labora-
tory, Pharmaceutical Sciences, Summit, NJ 07901, USA
b AdvancedChemistryDevelopment,MoscowDivision,Moscow117504, Russian
Federation, Russia
c ChemZoo, Inc., Wake Forest, NC 27581, USA
Magn. Reson. Chem. 2008; 46: 138–143 Copyright c© 2007 John Wiley & Sons, Ltd.

139
Identification of artifacts in indirect covariance spectra
with perhaps the sole exception of 1H–15N long-range correlation
experiments, which cannot be reasonably performed without
resorting to gradients.
Experimental
Alldatawereacquired for thepresent studyusingaVarian600 MHz
NMR spectrometer equipped with a 5 mm Cold Probe operating
at an rf coil temperature of 20 K. The sample used for the data
acquisition was prepared by dissolving ∼1 mg of naphtho[2′,1′:
5,6]naphtho [2′,1′: 4,5]thieno[2,3-c]quinoline (1)[10] (Scheme 1)in
∼200 µl deuterochloroform (CIL), after which the sample was
transferred to a 3 mm NMR tube (Wilmad) using a flexible Teflon
needle and a gas-tight syringe (Hamilton). Spectra acquired
included a 1H reference spectrum, a GCOSY spectrum (15 min),
a GHSQC spectrum (75 min) and an 18 ms IDR-GHSQC-TOCSY
spectrum (6 h). Experiments used the standard pulse sequences
contained in the vendor-supplied pulse sequence library andwere
used without modification. Indirect and unsymmetrical indirect
covariance processing was done using algorithms provided in
ACD/Labs SpecManager v10.02 software. Data matrices were
processed to afford identically digitized spectra 2K × 1 K points;
spectral widths in F2 and F1 were not identical. Unsymmetrical
indirect covariance processing times were typically a few seconds.
Results and Discussion
While examining the icv-13C–15N HSQC-HMBC spectrum of the
antitumor alkaloid vinblastine, an artifact was observed due to
the overlap of the H2 methine and 24-O-methyl singlets.[11] The
Figure 1. (A) GHSQC spectrum of naphtho[2′,1′ :5,6]naphtho[2′,1′ :4,5]thieno[2,3-c]quinoline (1) recorded at 25 ◦C in CDCl3 at 600 MHz. (B) Indirect
covariance spectrum, icv-HSQC, calculated from the GHSQC spectrum shown in Panel A. Off-diagonal responses are observed only for overlapped proton
resonances. (C) An expansion of the region from 132–137 ppm plotted at a lower threshold level shows several of the weaker pairs of off-diagonal
responses in the icv-HSQC spectrum that are barely visible in Panel B. As will be noted from Panel A, responses are generated in the indirect covariance
processed spectrum shown in Panel B with overlap ranging from partial to full.
Magn. Reson. Chem. 2008; 46: 138–143 Copyright c© 2007 John Wiley & Sons, Ltd. www.interscience.wiley.com/journal/mrc

140
G. E. Martin et al.
Figure 2. (A) The IDR-GHSQC-TOCSY spectrum of 1 recorded at 600 MHz with an 18 ms mixing time. Direct responses are inverted and plotted in red;
relayed responses have positive intensity and are plotted in black. Legitimate, vicinal correlations for the overlapped H6 and H15 protons are designated
by solid black lines. Type I artifact responses arise between pairs of direct or relayed correlation responses and have a negative phase in the indirect
covariance processed spectrum shown in Panel B and are designated by solid red lines. Type II artifact responses arise between direct and relayed
responses of different spin systems and have positive phase in the indirect covariance processed spectrum and are designated by dashed red lines
in Panel B. Type I responses can be assigned by visual inspection; Type II responses can be assigned only through analysis of the spectrum.[2] (B) The
icv-13C–13C correlation plot derived by the indirect covariance processing of the IDR-GHSQC-TOCSY spectrum shown in Panel A is presented in Panel B.
Legitimate 13C–13C correlation responses are denoted by solid black lines. Type I artifact responses are shown by solid red lines; Type II artifact responses
are denoted by dashed red lines.[2] Additional artifact responses are observed when plots are prepared with a deeper threshold.
www.interscience.wiley.com/journal/mrc Copyright c© 2007 John Wiley & Sons, Ltd. Magn. Reson. Chem. 2008; 46: 138–143

141
Identification of artifacts in indirect covariance spectra
Figure 3. (A) Indirect covariance spectrum, icv-HSQC, calculated from the GHSQC spectrum of 1. Off-diagonal responses denote proton resonance
overlaps that can give rise to artifact responses in the icv-13C-13C-COSY correlation plot prepared by indirect covariance processing of the GHSQC-TOCSY
or IDR-GHSQC-TOCSY spectrum. Off-diagonal responses in panel A are color coded identically to Fig. 1(B) and (C). (B) The icv-13C–13C-COSY plot prepared
by the indirect covariance processing of the IDR-GHSQC-TOCSY spectrum shown in Fig. 2(A). Off-diagonal responses designated with solid black lines
correspond to legitimate 13C–13C-COSY correlations. Off-diagonal responses denoted with solid red lines correspond to Type I artifact responses
predicted by the icv-HSQC spectrum shown in Panel A. Off-diagonal responses designated with dashed red lines correspond to Type II responses. The
blue boxes highlight the Type I artifact responses, contained in Panel B that are identified by the off-diagonal responses in the icv-HSQC spectrum shown
in Panel A. The dashed blue box identifies the Type II response origin, which as shown in Fig. 2, arises from the overlap of a direct response from one spin
system and the relayed response from an overlapped, second spin system.
former was expected to exhibit a two-bond correlation to the
N1 resonance; the correlation to the 24-O-methyl resonance, six
bonds distant from N1, was clearly an artifact. Titrating the sample
with d6-benzene demonstrated that aromatic solvent induced
shifts (ASIS)[12] could remove the degeneracy and eliminate the
artifact response. Perhaps more importantly, it was also shown
that the indirect covariance processed GHSQC spectrum (icv-
HSQC hereafter) showed off-diagonal responses that predicted
the location and identity of the O-methyl artifact response in
the 13C–15N icv-HSQC-HMBC spectrum. On that basis, we wanted
to explore the possibility of using off-diagonal responses in the
icv-HSQC spectrum to identify artifact responses in icv-13C–13C-
COSY correlation plots derived by indirect covariance processing
of GHSQC-TOCSY spectra as well as artifact responses in icv-
HSQC-COSY/-TOCSY spectra derived by unsymmetrical indirect
covariance coprocessing of gradient or nongradient HSQC and
either COSY or TOCSY spectra.
As a model compound for the present study, we again
elected to employ the complex polynuclear heteroaromatic,
naphtho[2′,1′:5,6]naphtho[2′,1′:4,5]thieno[2,3-c]quinoline (1)[10]
used in our initial study of artifacts in icv-13C–13C-COSY spectra.[2]
The molecule has a congested 1H NMR spectrum at 600 MHz
and multiple proton resonance overlaps that lead to artifact re-
sponses in the icv-13C–13C-COSY correlation plot derived from
the IDR-GHSQC-TOCSY spectrum (see Fig. 2 for an explanation of
the origins of the artifact responses).[2] In a similar fashion, artifact
responses are also anticipated in icv-HSQC-COSY spectra of 1 due
to proton resonance overlap.
The GHSQC spectrum and the icv-HSQC result obtained from
the indirect covariance calculation are shown in Fig. 1(A) and (B),
respectively. As will be noted from the icv-HSQC plot, there
are pairs of off-diagonal responses observed due to proton
resonance overlap of varying degrees that range from the
essentially complete overlap of the H6 and H15 resonances to
partial overlaps that give much weaker off-diagonal pairs in the
icv-HSQC spectrum, as shown in Fig. 1(C).
Subjecting the IDR-GHSQC-TOCSY spectrum shown in Fig. 2(A)
to indirect covariance processing affords the icv-13C–13C-COSY
connectivity plot shown in Fig. 2(B). Off-diagonal responses in
Fig. 2(B) correlated by solid black lines represent legitimate
13C–13C-COSY correlation responses. Off-diagonal responses
designated by solid red and dashed red lines correspond to Type
I and Type II artifact responses, respectively.[2] Fig. 3 compares
the indirect covariance spectrum calculated from the GHSQC
spectrum shown in Panel A with the icv-13C–13C-COSY correlation
plot calculated using indirect covariance processing shown in
Panel B. Legitimate vicinal 13C–13C-COSY correlations are denoted
by solid black lines; Type I artifacts are designated by solid red
lines; Type II artifacts are denoted by dashed black lines. The solid
blue boxes linking Panels A and B identify the origins of the Type I
artifact responses identified from the off-diagonal response in the
icv-HSQC spectrum shown in Panel A. The single dashed blue box
identifies the origins of one of the Type II artifact responses that
arise from the overlap of a direct response from one spin system
and a relayed response from a second, overlapped spin system.
While forecasting and identifying artifact responses in a
icv-13C–13C-COSY correlation plot is interesting, being able
Magn. Reson. Chem. 2008; 46: 138–143 Copyright c© 2007 John Wiley & Sons, Ltd. www.interscience.wiley.com/journal/mrc

142
G. E. Martin et al.
Figure 4. (A) IDR-GHSQC-TOCSY spectrum of 1 recorded at 600 MHz with an 18 ms mixing time. Direct responses are inverted and are plotted in red;
TOCSY responses have positive intensity and are plotted in black. A high resolution proton spectrum is plotted above Panel A. Connectivities are
shown in the less congested region of the spectrum. Direct proton/carbon correlation responses are labeled.[10] (B) icv-HSQC-COSY spectrum calculated
using the unsymmetrical indirect covariance processing algorithm from discretely acquired GHSQC and GCOSY spectra. A projection is shown above
Panel B. (C) icv-HSQC spectrum of 1. A projected 13C spectrum is shown above Panel C. Off-diagonal responses correspond to protonated carbons with
overlapped protons. The F1 13C shift corresponds to the F1 axis of the icv-HSQC-COSY spectrum shown in Panel B. The off-diagonal pair of responses in
Panel C denoted with the red bar corresponds to the overlapped proton resonances associated with the C6 and C15 resonances at 125.4 and 129.8 ppm,
respectively. Red arrows from the off-diagonal responses in Panel C to the corresponding F1 chemical shifts in Panel B denote the carbons associated
with overlapped protons. Correlations at the proton chemical shifts of H14 (7.84 ppm), H6/H15 (8.52 ppm), and H5 (9.09 ppm) and the C6 (125.4 ppm)
and C15 (129.8 ppm) correspond to artifact correlations. These pairs of responses in the icv-HSQC-COSY spectrum are designated by vertical red bars.
Off-diagonal responses at 134.4 and 132.4 ppm in Panel C designated by the green bar correspond to another pair of 13C shifts with proton resonance
overlaps. Green arrows from Panel C to Panel B identify these pairs of artifact responses in a manner analogous to that just described from the H6/H15
proton resonances. Artifact correlations in the icv-HSQC-COSY spectrum shown in Panel B are identified by red hatched boxes. Openblue boxes designate
legitimate correlations that are either not observed in the 18 ms IDR-GHSQC-TOCSY spectrum shown in Panel A or responses that are below the threshold
of the spectrum shown in Panel A.
to forecast the location of artifact responses in icv-HSQC-
COSY spectra calculated using unsymmetrical indirect covariance
processing is a much more useful application of the resonance
overlap information data derived by icv-HSQC spectrum. Figure 4
shows the comparison of the unsymmetrical indirect covariance
calculated icv-HSQC-COSY spectrum of 1 with the IDR-GHSQC-
TOCSY spectrum recorded with an 18 ms mixing time shown
previously inFig. 2(A). Inparticular, theability touseunsymmetrical
indirect covariance processing methods to calculate an icv-HSQC-
COSY spectrum from the discrete gradient or nongradient HSQC
and COSY spectra has the potential to provide investigators with
considerable timesavingsversushaving toacquire thehyphenated
2D data.
In addition, this type of post-acquisition processing also allows
investigators access to the information content of HSQC-COSY
spectra when they might have to wait before they again have
access toa spectrometeronwhich toacquire thedataorworse still,
when they may no longer have the sample due to decomposition,
consumption for biological or other testing, etc.
Visual comparison of the unsymmetrical indirect covariance cal-
culated icv-HSQC-COSY and acquired IDR-GHSQC-TOCSY spectra
shown in Fig. 4(B) and (A), respectively, quickly confirms that there
are responses present in the icv-HSQC-COSY spectrum that are not
present in the experimental spectrum. There are two possibilities
for the additional responses. First, as expected, proton resonance
overlaps confirmed by the icv-HSQC spectrum (Figs 1(B), (C), and
www.interscience.wiley.com/journal/mrc Copyright c© 2007 John Wiley & Sons, Ltd. Magn. Reson. Chem. 2008; 46: 138–143

143
Identification of artifacts in indirect covariance spectra
3(A)) are expected to give rise to artifact responses. Second,
there also exists the possibility that the higher relative sensitivity
of the GHSQC and GCOSY experiments, when compared to the
IDR-GHSQC-TOCSY experiment, can allow legitimate responses
to be observed that are either very weak or even absent in the
IDR-GHSQC-TOCSY spectrum.
Analyzing the icv-HSQC spectrum, 13C chemical shift pairs
can be determined for the off-diagonal response pairs. Plotting
the GHSQC indirect covariance spectrum beside the icv-HSQC-
COSY spectrum, the off-diagonal response pairs of the former
correspond to 13C shifts in the latter at which artifact responses
may be observed as shown in Fig. 4. Using the overlapped H6
and H15 proton resonances at 8.52 ppm as an example, off-
diagonal responses are observed in Fig. 4(C) for C6 (125.4 ppm)
and C15 (129.8 ppm denoted with red antidiagonal line). Tracking
horizontally from Fig. 4(C) to (B), pairs of responses at these 13C
chemical shifts are observed at theproton shifts of H14 (7.84 ppm),
H6/H15 (8.52 ppm), and H5 (9.09 ppm). The vertical red lines in
Fig. 4(B) denote the pairs of responses identified by the off-
diagonal responses in Fig. 4(C). The responses in red, hatched
boxes at 7.84/125.4 and 9.09/129.8 ppm are artifact responses.
The responses at 8.52/125.4 and 8.52/129.8 ppm are both direct
responses from isolated spin systems andhence are not correlated
to one another. Off-diagonal responses at 134.4 and 132.4 ppm
in Panel C correlated with a green anti-diagonal line represent
another pair of artifact correlations that are tracked back into
Fig. 4(B). In a similar fashion, the other off-diagonal correlations
shown in Fig. 4(C) could be used to track artifacts in the highly
congested region of the icv-HSQC-COSY spectrum in the region
from 7.75–8.1 ppm in F2 and from 134.0–136.5 ppm in F1.
In addition to the artifact responses already discussed in the
icv-HSQC-COSY spectrum shown in Fig. 4(B), there are also several
additional responses in blue, hatched boxes. The responses in this
group are legitimate correlation responses that are either below
the threshold of the IDR-GHSQC-TOCSY contour plot shown in
Fig. 4(A) or not observed at all because of the length of the mixing
time used. The ability to observe ‘distant’ responses in a four-spin
system is a function of the mixing time in the case of a GHSQC-
TOCSY experiment or the extent of digitization in the F1 frequency
domain in the case of the GCOSY spectrum used to calculate the
icv-HSQC-COSY spectrum as shown in Fig. 4(B).
Conclusions
Subjecting a GHSQC spectrum to indirect covariance processing
gives rise to an autocorrelated icv-HSQC spectrum in which
the only off-diagonal responses observed are due to carbon
resonances with overlapped, directly attached protons as shown
in Fig. 1. As shown in Fig. 3, the off-diagonal responses in the
icv-HSQC spectrum can be used to identify artifact responses
in GHSQC-TOCSY and IDR-GHSQC-TOCSY spectra subjected
to indirect covariance processing to derive icv-13C–13C-COSY
correlation plots. Perhaps the most valuable aspect of the present
report is the demonstrated ability to use off-diagonal response
information from the icv-GHSQC spectrum to identify artifacts
in icv-HSQC-COSY spectra calculated from discretely acquired
gradient or nongradient HSQC and COSY spectra using the
unsymmetrical indirect covariance processing algorithm.
Overlapping proton resonances, as initially observed in the
late 2004 report of Zhang and Bru¨schweiler,[1] can give rise to
artifact responses in icv-13C–13C-COSYcorrelationplots calculated
by indirect covariance processing of GHSQC-TOCSY and IDR-
GHSQC-TOCSY spectra. In a similar fashion, proton resonance
artifacts can also lead to artifacts in icv-HSQC-COSY spectra
calculated using unsymmetrical indirect covariance processing
as shown in the present study, or in icv-13C–15N HSQC-HMBC
as recently demonstrated.[11] However, as demonstrated in the
present study, off-diagonal responses contained in the icv-HSQC
spectrum resulting from the indirect covariance processing of
a GHSQC spectrum can be used to ‘forecast’ the location
of artifact responses in indirect covariance and unsymmetrical
indirect covariance calculated spectra of various types. Hopefully,
the ability to determine the location of artifact responses in
calculated icv-HSQC-COSY spectra will facilitate the use of these
and other calculated spectra, affording investigators considerable
spectrometer time savings. It will also be interesting to explore
the application of icv-HSQC spectra in the examination of spectra
derived by covariance processing of HMBC to afford icv-1H–1H-
COSY spectra as recently reported by Mu¨ller and coworkers.[13]
Some examples of time savings have been reported but
as a further example, we note that 13C–15N heteronuclear
correlation spectroscopy cannot be reasonably undertaken at
natural abundance while, in contrast, these correlation spectra
can be generated algorithmically. This capability alone further
underscores the value of the unsymmetrical indirect covariance
calculation of hyphenated and other 2D NMR spectra.
References
[1] Zhang F, Bru¨schweiler R. J. Am. Chem. Soc. 2004; 126: 13180.
[2] Blinov KA, Larin NI, Kvasha MP, Moser A, Williams AJ, Martin GE.
Magn. Reson. Chem. 2005; 43: 999.
[3] Blinov KA, Larin NI, Williams AJ, Zell M, Martin GE. Magn. Reson.
Chem. 2006; 44: 107.
[4] Blinov KA, Larin NI, Williams AJ, Mills KA, Martin GE. J. Heterocycl.
Chem. 2006; 43: 163.
[5] Martin GE, Hilton BD, Irish PA, Blinov KA, Williams AJ. J. Nat. Prod.
2007; 70: 1393.
[6] Blinov KA, Williams AJ, Hilton BD, Irish PA, Martin GE. Magn. Reson.
Chem. 2007; 45: 544.
[7] Martin GE, Hilton BD, Irish PA, Blinov KA, Williams AJ. Magn. Reson.
Chem. 2007; 45: 624.
[8] Martin GE, Hilton BD, Irish PA, Blinov KA, Williams AJ. J. Heterocycl.
Chem. 2007; 44: 1219.
[9] Martin GE, Hilton BD, Irish PA, Blinov KA, Williams AJ. Magn. Reson.
Chem. 2007; 45: 883.
[10] Hadden CE, Martin GE, Luo J-K, Castle RN. J. Heterocycl. Chem. 2000;
37: 821.
[11] Martin GE, Hilton BD, Blinov KA, Williams AJ. J. Nat. Prod. 2007; in
press.
[12] For reviews of ASIS (Aromatic Solvent Induced Shifts) see the
following: (a) Laszlo P. Prog. Nucl. Magn. Reson. Spectrosc. 1967; 3:
231; (b) Ronayne J, Williams DH. Annu. Rev. NMR Spectrosc. 1969; 2:
83.
[13] Schoefberger W, Smrecˇki V, Vikic´-Topic´ D, Mu¨ller N. Magn. Reson.
Chem. 2007; 45: 583.
Magn. Reson. Chem. 2008; 46: 138–143 Copyright c© 2007 John Wiley & Sons, Ltd. www.interscience.wiley.com/journal/mrc