Identification of Degradants of a Complex Alkaloid Using NMR
Cryoprobe Technology and ACD/Structure Elucidator
Gary E. Martin,* Chad E. Hadden, David J. Russell, Brian D. Kaluzny,
Jane E. Guido, Wayne K. Duholke, Bruce A. Stiemsma, and Thomas J. Thamann
Rapid Structure Characterization Group, Global Pharmaceutical Sciences, Pharmacia Corp.
Kalamazoo, Michigan 49001-0199
Ronald C. Crouch
NMR Applications Laboratory, Varian Inc., Palo Alto, California 94303
Kirill Blinov, Mikhail Elyashberg, Eduard R. Martirosian
Advanced Chemistry Development, Moscow 117513, Russian Federation
Sergey G. Molodtsov
Novosibirsk Institute of Organic Chemistry,
Siberian Branch of Russian Academy of Science, Lavrentiev Avenue 9, Novosibirsk 630090, Russia
Antony J. Williams
Advanced Chemistry Development, Toronto, Ontario, M5H 3V9, Canada
Paul L. Schiff, Jr.
Department of Pharmaceutical Sciences, School of Pharmacy, University of Pittsburgh, Pittsburgh, PA 15261
Received June 6, 2002
Identification of degradants of pharmaceuticals is a necessary challenge of the drug development process
following the subjection of candidate molecules to a variety of physico-chemical stresses. It would be desir-
able to be able to conduct such studies on a minimal amount of material. As a prototypical study, the isola-
tion and identification of degradants of a sample of the complex indoloquinoline alkaloid, cryptospirolepine,
was undertaken after prolonged storage in DMSO solution using a combination of cryogenic NMR probe
technology and CASE (Computer-Assisted Structure Elucidation) programs. None of the starting alkaloid
remained after storage; a chromatogram of the DMSO solution demonstrated the presence of >25 compo-
nents in the mixture. The two most abundant degradation products were identified as the known alkaloid
cryptolepinone (~35%) and an unprecedented rearrangement product, DP-2, (~16%).
J. Heterocyclic Chem., 39, 1241 (2002).
Introduction.
Structural characterization of degradants of pharmaceu-
ticals is a necessary challenge of the drug development
process. Candidate molecules are routinely subject to
physicochemical stress challenges of various types to
understand their stability and degradation behavior.
Ideally, to facilitate this process, it would be desirable to be
able to conduct such studies on minimal amounts of mate-
rial, which imposes stringent sensitivity demands on the
NMR spectrometer used in the work, mandating the use of
small volume, high sensitivity NMR probes or cryogenic
NMR probe technology. Mass spectrometric sensitivity is
seldom an issue and should be an integral part of the struc-
ture characterization process. In addition, to expedite the
process, it would also be highly desirable to be able to
employ computer-assisted structure elucidation (CASE
systems) to assist the investigator.
As a model study, a 2.5 mg sample of the complex
spiro nonacyclic alkaloid cryptospirolepine (1) that had
been stored in a sealed 5 mm NMR tube in d6-DMSO for
a prolonged period (~10 years) was examined to evaluate
the combined utilization of cryogenic NMR probe tech-
nology and computer-assisted structure elucidation in the
characterization of unknown degradants of a complex
molecule. [1]
Nov-Dec 2002 1241
1

G. E. Martin, C. E. Hadden, D. J. Russell, B. D. Kaluzny, J. E. Guido, W. K. Duholke, B. A. Stiemsma, T. J. Thamann,
R. C. Crouch, K. Blinov, M. Elyashberg, E. R. Martirosian, S. G. Molodtsov, A. J. Williams and P. L. Schiff, Jr.
Computer-Assisted Structure Elucidation (CASE).
Computer-Assisted Structure Elucidation (CASE) pro-
grams are a viable means to assist even highly competent
NMR spectroscopists to elucidate complex chemical struc-
tures where there are 2D NMR data in a more efficient
manner. There have been a number of reports in the litera-
ture that have described expert systems [2-11] intended for
this purpose. In these systems, 2D NMR data are pre-
sented in the form of atom-to-atom connectivities between
atoms of the molecule, which serve as restrictions in the
structure generation process in accord with the given mol-
ecular formula. Typical data sets may be comprised of:
homonuclear COSY or TOCSY data; NOESY or ROESY
correlation data; direct heteronuclear shift correlation
spectra such as 1H-13C HMQC or HSQC, or more recently
1H-15N direct correlation spectra; and lastly long-range
heteronuclear shift correlation data. The latter, until a few
years ago, were exclusively restricted to statically-opti-
mized 1H-13C HMBC data. More recently, long-range 1H-
15N experiments at natural abundance have received con-
siderable attention, forming the topic of recent, compre-
hensive reviews [12]. In addition, the development of
numerous, accordion-optimized long-range heteronuclear
shift correlation experiments has been reported and is the
topic of a recent review [13]. Regardless of the ensemble
of 2D NMR data acquired and utilized, the preferred
approach has been based on the generation of structures
based on the prediction of 13C NMR shifts of candidate
molecules consistent with the atom-to-atom connectivity
information extracted from the various 2D NMR data sets.
The experience gained with the development and utiliza-
tion of expert systems has shown that the best results can
be achieved when there is a close, and highly synergistic
interaction between the spectroscopist and the computer
program being used. Ideally, a competent spectroscopist
will be allowed the broad possibility of using his/her expe-
rience and knowledge about properties of the molecule
being studied to impose additional restrictions for the
types of structures generated. Quite simply, if the spectro-
scopist knows that the molecule being structurally interro-
gated with his/her NMR experiments is a steroid, there is
little reason to allow a CASE program the freedom of gen-
erating inappropriate and unrelated molecular structures.
The successful implementation of such possibilities cre-
ates a "symbiotic" relationship between the human mind
and the computer, ultimately synergizing the structure elu-
cidation process. Unfortunately, most of the expert sys-
tems devised thus far [2-11] have rather restricted ability to
apply a priori information, thus precluding the use of user-
defined structural fragments during the elucidation
process. Little attention was paid in the past to detecting
contradictions in 2D data and/or in providing methods for
their resolution. It is known that one frequent source of
contradictions is, for example, the observation of correla-
tions in COSY or long-range heteronuclear shift- correla-
tion spectra that correspond to four or more bonds [14]. In
general, CASE programs are "tuned" for shorter correla-
tion pathways. Unusual long-range heteronuclear correla-
tion pathways have recently been reviewed by Araya-
Maturana and co-workers [15]. Different ways of elimi-
nating contradictions have been proposed. One possible
approach is to iteratively search for contradictions by mul-
tiple repetition of the structure generation process [14].
With each cycle, one correlation to a weak 2D peak is
added to the structure generation process. If any correla-
tion is across more bonds than the number set by default,
the structure generator portion of the program will not pro-
duce any structures after such a correlation has been
added, indicating the presence of contradictions. Another
approach to overcoming this difficulty is by using a sto-
chastic algorithm of structure generation that requires
application of computer complexes on parallel processors
[10]. The common drawback of the methods extant in the
literature [2-11] is the inability of these programs of inter-
acting with a diverse sub-structure base accompanied by
related 13C NMR sub-spectra. Our experience has shown
that quite frequently important structural information
about an unknown structure can be derived from such
base.
The aforementioned drawbacks are largely overcome in
the StrucEluc program [14]. This system is capable of
using, in addition to 2D NMR spectral data, a library with
500,000 fragments that have 13C NMR chemical shift
assignments associated with them [15]. In those cases
when the number of available 2D NMR correlations is
insufficient to impose restrictions on the structure genera-
tion process (in this case the number of possible structures
can be enormous and generation times unacceptable), the
system searches for appropriate fragments in the library in
accordance with their sub-spectra. Of the fragments
found, the fragments meeting the restrictions arising from
2D spectra, are kept. The admissible combinations of
good fragments are "projected" on the set of all atoms of a
molecule. Consequently, the program builds and visualizes
molecular connectivity diagrams (MCD); fragments,
atoms, and connectivities of different length are graphi-
cally represented. At this point in the elucidation process,
a capable spectroscopist has the opportunity to analyze
MCD's and to make his/her revisions (specify carbon
atoms hybridization set by the program, defining the possi-
ble chemical "neighborhood" with heteroatoms, etc.). It is
envisioned that the chemist may also introduce fragments
that in his/her opinion should be present in a molecule.
Subsequently, both user fragments and those from the
library are used for MCD creation.
Chemists commonly employ a strategy of using the
assigned spectra of chemically related structures, whenever
they're available, in deducing the structure of a new
1242 Vol. 39

Identification of Degradants of a Complex Alkaloid Using NMR Cryoprobe Technology
compound. Generally this approach is rather successful. In
order to implement this method within the StrucEluc pro-
gram, algorithms enabling the automatic generation of the
user's fragments library, have been incorporated. To create
the latter, the chemist first enters related structures having
assigned 13C NMR spectra into the user database; then frag-
ments are created according to certain rules from these
structures. In this case the probability that the resulting data-
base will contain fragments that are actually present in the
unknown molecule under study is relatively high. If neces-
sary, the program is also able to sequentially "construct" the
molecule of the unknown compound from the fragments by
overlapping common atoms or forming fragments sets that
will be then used for structure generation under the given
molecular formula. The generated structures can be veri-
fied using NMR and IR filters [15].
To provide the means for the preliminary analysis of 2D
NMR data for non-contradiction, the StrucEluc program
uses a heuristic algorithm that is capable in most cases (90-
95%) of detecting the presence of contradictions as well as
revealing the reasons for the contradiction. Then the pro-
gram tries to automatically eliminate any contradictions
found by lengthening those connectivies that according to
the analysis results may have length contradictory to the
one set by default. In the present example, we have
employed the Structure Elucidator program for the identi-
fication of isolated degradants of the complex alkaloid
cryptospirolepine to explore the flexibility of the program
in terms of its ability to elucidate the correct structure and
for the ability of a competent spectroscopist to interact
with it in a convenient manner.
Chromatography:
The sample of cryptospirolepine was first interrogated
by LC/MS methods to see how much of the starting alka-
loid (MH+ = 505) remained in the sample – none was
found. A preparative chromatographic system was then
developed for the sample, which was found to contain 26
components ranging from a major component DP-1
(~35%) to numerous components in the range of 2-3%
(Figure 1). Initial effort was directed at isolating the two
major components, DP-1 and DP-2 (~35 and ~16% of the
total sample, respectively, based on peak area). The two
major degradation products (DP-1 and DP-2) of cryp-
tospirolepine (1) were isolated by reversed-phase, semi-
preparative HPLC. A DMSO solution of the degraded
alkaloid was diluted with mobile phase and injected onto a
semi-preparative HPLC system consisting of a 21.2 x 250
mm, Kromasil C18 column with an acetonitrile-aqueous
trifluoroacetic acid mobile phase. Detection was accom-
plished at 270 nm. Collected fractions of DP-1 and DP-2
were concentrated and desalted via trapping on a 10 x 250
mm, Kromasil C18 column. Eluent from the trapping
column containing the degradation products was freeze-
dried to yield about 1.1 mg of DP-1 at 96% purity by
HPLC and about 200 µg of DP-2 at 95% purity by HPLC.
NMR samples of about ~0.5 mg and ~100 µg, respectively,
were used for the structural characterization effort.
DP-1 – Cryptolepinone.
The isolated degradant, DP-1, which constituted the
major component of the chromatogram shown in Figure 1
on the basis of peak area percent, was readily identified by
mass spectrometry, giving a parent ion, MH+ = 249, sug-
gesting that the cryptospirolepine (1) had presumably been
split in half during degradation. Using only GCOSY and
GHSQC spectra, the structure was quickly identified (<20
min) as cryptolepinone (2) [17-20]. A quick (~ 1 h)
GHMBC spectrum was acquired, solely to provide a few
long-range connectivities, i.e. those from the N-methyls
for the data to be fed into the CASE program.
The ACD/ StrucEluc [21] software package is composed
of a number of software modules including both 1D and
2D NMR processing and NMR prediction of both 1H and
13C spectra from input chemical structures. Both NMR
prediction modules offer the ability to construct user data-
bases of structures and nuclear assignments that can be
used to train the prediction algorithms. The 1D and 2D
NMR and the mass spectral data were fed to the StrucEluc
2
Nov-Dec 2002 1243
Figure 1. Chromatogram showing the distribution of peaks from a
degraded sample of cryptospirolepine (1) after prolonged storage in d6-
DMSO. The two most abundant components, DP-1 and DP-2, were iso-
lated in the present study. On the basis of LC/MS data, no cryp-
tospirolepine (1) remained in the sample at the beginning of this study.

G. E. Martin, C. E. Hadden, D. J. Russell, B. D. Kaluzny, J. E. Guido, W. K. Duholke, B. A. Stiemsma, T. J. Thamann,
R. C. Crouch, K. Blinov, M. Elyashberg, E. R. Martirosian, S. G. Molodtsov, A. J. Williams and P. L. Schiff, Jr.
program. Because of the relatively unconstrained data set
(the lack of complete long-range data), the program gener-
ated a total of 208 structures as output. When the gener-
ated family of structures was sorted based on the match
factor (d) distinguishing the deviation between the experi-
mental and the predicted 13C spectra, cryptolepinone had
the lowest standard deviation (0.0). Such a deviation iden-
tifies the fact that the chemical structure already exists in
the assigned literature databases. The ability to construct
User Databases is useful in that it can preclude the "re-elu-
cidation" of structures as a database grows and becomes
more complete, thereby resulting in considerable time sav-
ings in the elucidation process.
The combined input of a set of GHSQC data and accu-
rate mass and fragmentation data is a potentially useful
way of establishing the identity of known compounds.
The StrucEluc program produced a total of 14 structures
from these input data. The output file generated, sorted on
the basis of the 13C average chemical shift deviation with a
5.0 ppm maximum deviation, are shown in Figure 2.
That cryptolepinone (2) was the primary component of
the degraded sample of cryptospirolepine (1) is interesting
and suggests, not surprisingly, that the spiro center of the
parent molecule was a labile point in the structure and
prone to oxidative degradation in DMSO solution,
although a mechanism to explain the formation of 2 is not
readily apparent.
DP-2.
Mass spectrometry on the second most abundant isolate
from the degraded sample gave a molecular ion, MH+ = 479,
which suggested a molecular formula of the DP-2 isolate as
C32H22N4O, based on a loss of 26 Da (C2H2) relative to well
established molecular formula of cryptospirolepine (1) of
C34H24N4O. Significant fragment ions were observed in an
MS/MS experiment at 464, 447, 435, 432, 247, 232, and 217
Da. It is interesting to note that the 232 daughter ion corre-
sponds to cryptolepine minus a proton, suggesting a substi-
tuted cryptolepinyl moiety as a subcomponent in the struc-
ture of the degradant. This type of intuitive insight by the
experienced spectroscopist should, ideally, be able to be
utilized by a CASE program, which StrucEluc allows.
Physically, the sample of DP-2 dissolved in d6-DMSO
was initially reddish-orange in color. The initial proton
spectrum at 500 MHz was rather broadened, suggesting
that the sample was protonated from the chromatographic
isolation. A small quantity of ammonia gas (headspace
gas from a bottle of conc. ammonium hydroxide) was
bubbled through the sample causing the color to shift to
deep purple, consistent with the extended conjugation of
cryptolepine (3) or a cryptolepinyl substructure as implied
by the mass spectral data [22-25]. A similar ammonia gas-
induced color shift was noted during the characterization
of the alkaloid cryptolepicarboline, which also contains an
11-cryptolepinyl moiety in its structure [26].
1244 Vol. 39
Figure 2. Structures generated by ACD/Structure Elucidator sorted on the basis of 13C average deviation (<5 ppm maximum). It is interesting that five
of the fourteen structures are indoloquinoline analogs which are plausible if this had been an unknown structure.

Identification of Degradants of a Complex Alkaloid Using NMR Cryoprobe Technology
A GCOSY spectrum (Figure 3) and a GHSQC spectrum
(Figure 4) were obtained using a Varian INOVA 500 MHz
three channel NMR spectrometer equipped with a Nalorac 3
mm MIDTG-500-3 gradient inverse triple resonance probe.
The experiments gave a proton spectrum containing two
methyl signals that could be assigned as N-methyls (4.56 and
5.09 ppm – typical of cryptospirolepine) and a total of six-
teen protonated aromatic CH's that were subgrouped into
four four-spin systems by the GCOSY data. The isolated 11-
methine singlet of cryptospirolepine from the indoloben-
zazepine-derived portion of the molecule was absent, as was
the indole NH resonance. This observation, coupled with
MS/MS fragment ions and the observed color change on
treatment with ammonia gas, further strengthened the
hypothesis of the structure of DP-2 containing an 11-cryp-
tolepinyl moiety. The GHSQC experiment required an
overnight data acquisition (17 h) using a conventional 3 mm
gradient inverse triple resonance probe, suggesting an acqui-
sition time of at least 3-4 days to acquire a usable HMBC
spectrum. A 500 msec ROESY experiment gave four
N-methyl to aromatic methine correlations (see 4 and 6). A
phase-cycled 8 Hz optimized HMBC spectrum was recorded
next at 500 MHz using a Varian 500 MHz gradient inverse 5
mm triple resonance Chili-probe [27]. The initial,
overnight HMBC spectrum contained 32 readily assigned
responses; there was considerable overlap in the region from
7.3-7.47 ppm complicating signal assignments somewhat.
Then, the available data were interpreted by the authors in
addition to being loaded into ACD/ StrucEluc.
Interpretation of the HMBC data quickly led to the deduc-
tion that, as suspected, the purple color of the solution in the
NMR tube was due to the presence of an 11-cryptolepinyl
species (4) contained in the structure. The initial run of
ACD/ StrucEluc, performed in parallel with human data
interpretation, gave an initial output of ca. 2000 structures,
which when filtered afforded 107 11-crytolepinyl-containing
structures. From the preliminary HMBC data, it was antici-
pated that the correct structure would necessarily locate a
carbonyl resonance (167.4 ppm – typical of the 4-quinolone
species contained in some Cryptolepis alkaloids) within 3 to
4 bonds of one of the terminal aromatic proton resonances
(8.05 ppm). None of the 107 filtered structures met this
requirement of the data set in a reasonable fashion – i.e.,
none were indoloquinoline-based that could reasonably be
derived from the degradation of cryptospirolepine (1) itself.
A second phase-sensitive, phase-cycled HMBC spectrum
optimized for 6 Hz was acquired overnight and contained a
total of 46 assignable responses, shown in Figure 5. A com-
parison of the heavily congested region of the conventional
and phase-sensitive HMBC spectra is shown in Figure 6.
As is readily noted from the comparative data presented in
Figure 6, there is substantially better resolution in the highly
congested region of the spectrum when phase-sensitive data
are acquired. Interpretation of these data was again under-
taken concurrently with the processing of the data using
ACD/ StrucEluc [28]. This resulted in the output of almost
3300 structures after a generation period of almost 8 hours.
After the removal of duplicate structures 355 structures
remained. Examination of the elucidation results showed
that structure 4 in the output table was consistent and indolo-
quinoline-based as expected (vide infra). This structure is
4
Nov-Dec 2002 1245
Figure 3. GCOSY spectrum of DP-2 acquired using a Varian INOVA 500
MHz three channel NMR spectrometer equipped with a Nalorac
MIDTG-500-3 gradient triple resonance NMR probe.
3

G. E. Martin, C. E. Hadden, D. J. Russell, B. D. Kaluzny, J. E. Guido, W. K. Duholke, B. A. Stiemsma, T. J. Thamann,
R. C. Crouch, K. Blinov, M. Elyashberg, E. R. Martirosian, S. G. Molodtsov, A. J. Williams and P. L. Schiff, Jr.
5,5'-dimethyl-5'H-10,11'-biindolo[3,2-b]quinolin-11(5H)-
one. Structures were sorted on the basis of the agreement
between the calculated 13C chemical shift (dA) and the
"observed" 13C shifts taken from the HMBC spectrum.
1246 Vol. 39
Figure 5. Phase-sensitive HMBC spectrum of the aromatic region of an ~100 mg of DP-2 in 150 mL d6-DMSO in a 3 mm NMR tube acquired overnight
using a Varian INOVA 500 MHz three channel NMR spectrometer equipped with a 5 mm gradient inverse triple resonance cryogenic NMR probe.
Figure 4. GHSQC spectrum of DP-2 acquired overnight using a Varian INOVA 500 MHz three channel NMR spectrometer equipped with a Nalorac
MIDTG-500-3 gradient triple resonance NMR probe.

Identification of Degradants of a Complex Alkaloid Using NMR Cryoprobe Technology
When the 11-cryptolepinyl fragment was included as one
of the user fragments the elucidator produced 1268 struc-
tures in less than 10 minutes. This filtered to 111 structures
after the removal of duplicates. The structure with the best
combination of 13C and 1H deviations was displayed at
position 2 (Figure 7) and was again 5,5'-dimethyl-5'H-
10,11'-biindolo[3,2-b]quinolin-11(5H)-one (7) consistent
with the expected structure. It is also worth noting that this
structure has the best agreement when considered on the
basis of calculated vs. observed 1H chemical shift data.
A long-range 1H-15N CIGAR-HMBC experiment was
also performed with an optimization from 3-6 Hz [13, 29-31]
using the Chili-probe over a long weekend (~72 h). Long-
range correlations were observed for the two N-methyl
groups - 4.56/109.7 and 5.09/158.4 ppm. The former was
consistent with the N-methyl chemical shift of the N-methyl
contained in the indolobenzazepine-derived portion of cryp-
tospirolepine, and cryptolepinone. The latter, in contrast,
didn't agree well at all, but rather was in almost exact agree-
ment with the corresponding N-methyl resonance of cryp-
tolepine. One other long-range correlation was observed in
the CIGAR-HMBC spectrum, that being a 5JNH correlation
from the H1 resonance of the 11-cryptolepinyl moiety to the
N10 nitrogen resonance (exo N=C) at 230 ppm. In compari-
son, the corresponding 15N resonance of unsubstituted cryp-
tolepine is observed at 207.8 ppm. The 15N chemical shift
data and observed long-range 1H-15N correlations were also
fed as constraints to StrucEluc [32].
In the absence of the 11-cryptolepinyl fragment the elu-
cidator produced about 4700 structures over about 15 hours.
Removal of duplicate structures gave 334 final structures
with the 5,5'-dimethyl-5'H-10,11'-biindolo[3,2-b]quinolin-
11(5H)-one (7) again positioned second on the basis of 13C
shift and first based on 1H shift data. When the 11-cryp-
tolepinyl fragment and the 15N correlation data were used
111 structures were generated in 20 minutes with 5,5'-
dimethyl-5'H-10,11'-biindolo[3,2-b]quinolin-11(5H)-one
(7) again located at position 2 in the table of structures
sorted on the basis of chemical shift. A comparison of the
Nov-Dec 2002 1247
Figure 7. When the 11-cryptolepinyl fragment was included as one of the user fragments the Structure Elucidator produced 1268 structures in less than
10 minutes, which were filtered to 111 after the removal of duplicate structures. The structure with the best combination of 13C and 1H deviations was
displayed at position 2 and again was the 5,5'-dimethyl-5'H-10,11'-biindolo[3,2-b]quinolin-11(5H)-one, consistent with the expected structure.
Figure 6. Comparison of the congested region of the long-range 1H-13C
heteronuclear shift correlation spectra of DP-2. The conventional, phase-
cycled HMBC data are shown on the left; phase-cycled, phase-sensitive
HMBC data are shown on the right. Processing was identical but the data
on the left were acquired using 200 increments of the evolution time, ni,
while 224 hypercomplex increments were used to acquire the data shown
in the right panel. All of the long-range correlations are well resolved in
the phase-sensitive data, facilitating interpretation. In addition, some
weak responses are observed in the phase-sensitive data that were not
observed in the conventional, phase-cycled experiment.

G. E. Martin, C. E. Hadden, D. J. Russell, B. D. Kaluzny, J. E. Guido, W. K. Duholke, B. A. Stiemsma, T. J. Thamann,
R. C. Crouch, K. Blinov, M. Elyashberg, E. R. Martirosian, S. G. Molodtsov, A. J. Williams and P. L. Schiff, Jr.
computation times and results obtained using Structure
Elucidator are collected in Table 1.
The long-range data were interpreted manually to give the
structural fragment shown by 5. Key correlations observed in
the HMBC spectrum are shown on the structure. ROESY cor-
relations are denoted by double-headed arrows. Valence
requirements and the empirical formula were satisfied by one
additional sp2 carbon, linking the carbonyl, the nitrogen, and
the N-methyl containing bridge to give a second indoloquino-
line moiety in place of the starting indolobenzazepine-contain-
ing structural unit of the starting cryptospirolepine to give 6.
Linking the N-cryptolepinonyl substructure represented
by 6 to the 11-cryptolepinyl substructure shown by 4
established the final structure of DP-2 as 5,5'-dimethyl-
5'H-10,11'-biindolo[3,2-b]quinolin-11(5H)-one, which is
shown by 7.
Complete 1H and 13C, and partial 15N chemical shift
assignments and observed long-range heteronuclear cou-
plings are summarized in Table 2.
It is interesting to note that the spiro-center of cryp-
tospirolepine (1) was at the focal point of the degradation
chemistry thus far examined. It will be interesting to see if
this trend continues for the remainder of the numerous
compounds represented in the chromatogram in Figure 1.
The formation of cryptolepinone (2) can be assumed to be
7
5 6
1248 Vol. 39
Table 1
Parameters used and results output from StrucEluc calculation.
Cryptolepine N–H Number of Number of structures Generation DP-2 position
fragment HMBC connectivities (generated/after time (by CNMR /
usage usage duplicates removal) by HNMR)
Yes Yes 4 2158/111 20 min 2/1
Yes No 2 1268/111 10 min 2/1
No Yes 576 4683/334 15 h 2/1
No No 288 3282/355 8 h 2/1
Figure 8. Structures generated by the second Structure Elucidator run using phase-sensitive, phase-cycled HMBC data. The 8 structures shown, from a
total of 355 generated, were those with a 13C average chemical shift deviation of < 10.0 ppm. The fourth structure agrees with the proposed structure of
the DP-2 degradant of cryptospirolepine, as shown by 7.

Identification of Degradants of a Complex Alkaloid Using NMR Cryoprobe Technology
a relatively straight-forward oxidative degradation. In
contrast, however, there is no readily obvious chemical
pathway to explain the much more complex conversion of
cryptospirolepine (1) to DP-2 (7).
Mechanistic considerations that must necessarily be
implicit for such a conversion to have occurred are still
being considered. The structures of additional degradants
represented by the chromatogram shown in Figure 1 is
ongoing and will form the basis for future reports. Of
necessity, the rigorous structural determination of samples
still considerably smaller than DP-2 will mandate the uti-
lization of NMR cryoprobe technology.
Conclusions.
The significant improvement in sensitivity offered by
cryogenic NMR probes can be applied to the elucidation of
novel, unknown structures. The added sensitivity can be
used to reduce data acquisition times for experimental data
that are obligatory and can make feasible the acquisition of
data such as long-range 1H-15N 2D heteronuclear shift corre-
lations that would be inaccessible in practical periods of time
when using conventional NMR probe technology. These
data can also be used to advantage with CASE (Computer-
Assisted Structure Elucidation) programs such as ACD/
StrucEluc to aid structural chemists in the characterization of
unknowns. As shown in this example, a group of spectro-
scopists with intimate knowledge of the structure elucidation
of this family of indoloquinoline alkaloids still required
about three days of interpretation time once all of the data
were in hand. In contrast, using StrucEluc, under the worst-
case scenario, plausible structural hypotheses would be
available to the chemist for his/her consideration overnight
once all of the data were in hand. It should also be noted that
a CASE program such as StrucEluc could be used while data
acquisition is on-going, allowing the investigator to refine
his/her thinking "on the fly". This approach should offer
advantages in the identification of natural products or the
identification of degradation products of unknown types aris-
ing from compounds of known structure.
EXPERIMENTAL
The original sample of 2.5 mg of cryptospirolepine (1) was
prepared in ~500 µL of deuterodimethylsulfoxide, 99.96% D and
then sealed in a standard 5 mm NMR tube. On initial prepara-
tion, the sample was a red-orange color. After the elucidation of
the structure of cryptospirolepine [1] the sealed NMR tube was
stored at ambient temperature for ~10 years with no special pre-
caution taken. On prolonged storage, the sample darkened to a
deep brown color.
After the NMR tube was cut open, preliminary interrogation of
the sample by LC/MS showed that none of the starting alkaloid
remained in the sample. Analytical HPLC showed the sample to
contain 26 components (Figure 1) ranging from the major com-
ponents DP-1 (~35%) and DP-2 (~16%) (based on peak area
uncorrected for relative response factors) to numerous small
components in the 2-3% range. The deuterodimethylsulfoxide
solution was diluted with an aqueous acetonitrile-water-trifluo-
roacetic acid mobile phase and injected onto a 21.5 x 250 mm
semi-preparative Kromasil C18 column. Detection was accom-
plished by monitoring at 270 nm.
Mass spectrometric data for the isolated samples were acquired
either on PE Sciex Q-Star time of flight mass spectrometer (low res-
olution and MS/MS measurements or using a Finnigan MAT-900ST
mass spectrometer operating in the micro-ESI (micro electrospray
ionization) mode. High resolution mass spectral data for the DP-1
and DP-2 isolates were obtained on a few micrograms of each of the
isolates dissolved in mobile phase consisting of 50:50
methanol:water with 2% formic acid added. Accurate mass mea-
surement was carried out by linear E-scan peak matching at a reso-
lution near 8,500 (m/∆m, 10 % valley definition) using selected ref-
erence ions from PEG 400 to bracket the various sample pseudo-
molecular ions. The accurate masses of isolate DP-1 and DP-2
were measured as their protonated ions. These data established the
molecular formulae of cryptolepinone (2) and the unknown, DP-1
Nov-Dec 2002 1249
Table 2
1H, 13C, and 15N Chemical Shifts, Long-Range Heteronuclear and
ROESY Correlations observed for 5,5'-dimethyl-5'H-10,11'-
biindolo[3,2b]quin-olin-11(5H)-one (7).
Position d1H d13C d15N Long-range 1H-13C or Important
1H-15N Correlations ROESY
(w = weak; Correlations
vw = very weak;
nd = not determined)
1 7.40 120.1 --- C4a, C3
2 7.47 131.5 --- C4, C11a
3 7.06 117.4 --- C1, C4a
4 8.58 125.8 --- C3, C4b(w), C11a
4a --- 115.5 --- ---
4b --- 142.2 --- ---
5 --- --- 158.6 ---
5-Me 5.09 39.8 --- C4b,C5a, N5 H4, H6
5a --- 133.6 --- ---
6 8.70 117.9 --- C8, C9a
7 7.92 129.3 --- C5a, C9
8 7.53 125.2 --- C6, C7, C9a
9 7.61 124.4 --- C5a, C7, C10, N11 ( 5JNH)
9a --- 123.8 --- ---
10 --- 133.3 --- ---
10a --- 136.9 --- ---
11 --- --- 230.0 ---
11a --- 161.4 --- ---
1' 6.77 112.6 --- C2', C3', C4a'
2' 7.36 121.4 --- C4', C11a'
3' 7.38 128.6 --- C1', C4a'
4' 8.66 124.4 --- C2', C3', C4a', C4b', C11a'
4a' --- 117.7 --- ---
4b' --- 132.9 --- ---
5' --- --- 109.4 ---
5'-Me 4.56 37.0 --- C4a', C5a', N5' H4', H6'
5a' --- 141.4 --- ---
6' 8.08 116.6 --- C8', C9a'
7' 7.80 132.2 --- C5a', C9'
8' 7.29 121.8 --- C6', C7', C9a'
9' 8.05 126.1 --- C5a', C7', C10', C10a'(vw)
9a' --- 124.9 --- ---
10' --- 167.1 --- ---
10a' 130.0 ---
11' --- --- nd ---
11a' --- 141.4 --- ---

G. E. Martin, C. E. Hadden, D. J. Russell, B. D. Kaluzny, J. E. Guido, W. K. Duholke, B. A. Stiemsma, T. J. Thamann,
R. C. Crouch, K. Blinov, M. Elyashberg, E. R. Martirosian, S. G. Molodtsov, A. J. Williams and P. L. Schiff, Jr.
(7). The latter, with an empirical formula of C32H22N4O based on
high resolution measurements, represents a loss of 26 Da relative to
the structure of cryptospirolepine (1). Fragmentation observed in
the MS/MS experiments is discussed in the text.
NMR data for an ~500 µg sample of cryptolepinone (2) were
acquired by dissolving the sample in ~150 µL of deuterodi-
methylsulfoxide (99.996% D, Cambridge Isotope Laboratories)
and then transferring the sample to a 3 mm NMR tube (Wilmad).
The data were acquired on a two channel Varian Inova 400 MHz
NMR spectrometer equipped with a Nalorac Z•SPEC MIDG-
400-3 gradient inverse 3 mm NMR probe. The sample was iden-
tified as cryptolepinone (2) from COSY and GHSQC data.
NMR data for DP-2 (7) were obtained by halving the sample and
dissolving ~100 µg of the isolate in ~150 µL of deuterodimethylsul-
foxide (99.996% D, Cambridge Isotope Laboratories) and then
transferring the sample to a 3 mm NMR tube (Wilmad). All data
were acquired on a Varian Inova 500 MHz three channel NMR
instrument equipped with either a Nalorac Z•SPEC MIDTG-500-
3 gradient inverse triple resonance probe or a Varian 5 mm gradient
inverse triple resonance cryogenic Chili-probe operating at 25 K.
Homonuclear NMR data were acquired using the conventional 3
mm NMR probe as were the GHSQC data, which were acquired
overnight. On initially recording a proton reference spectrum, the
aromatic proton resonances were broad and suggestive of traces of
acid contamination from the isolation procedure. The sample was
initially orange in color. After being treated with several bubbles of
ammonia gas from the headspace of a bottle of conc. ammonium
hydroxide, the color shifted to a deep purple color consistent with
the extended conjugation of cryptolepine with commensurate sharp-
ening of the proton resonances. Long-range 6 Hz optimized 1H-13C
HMBC data, both magnitude calculated and phase-sensitive, were
acquired overnight using the Chili-probe running the 3 mm tube
coaxially in the 5 mm probe. Processing details are described in the
figure captions. Long-range 1H-15N CIGAR-HMBC data [13,29-
31] optimized for 3-6 Hz were acquired in approximately 72 h over
a weekend, and gave correlations to 3 of the 4 nitrogen resonances
contained in the structure of DP-2. Chemical shifts for the 3 nitro-
gens observed were consistent with 15N chemical shifts for similar
molecules in this series of alkaloids [29].
REFERENCES AND NOTES
* To whom inquiries should be addressed: e-mail: gary.e.martin
@pharmacia.com; +269-833-6283; fax +269-833-2030.
[1] The sample used for this study was the original sample of
cryptospirolepine prepared when the original structure elucidation was
undertaken in late 1991. See: A. N. Tackie, G. L. Boye, M. H. M. Sharaf,
P. L. Schiff, P., Jr., R. C. Crouch, T. D. Spitzer, R. L. Johnson, J. Dunn, D.
Minick, and G. E. Martin, J. Nat. Prod., 56, 653 (1991).
[2] M. E. Munk, J. Chem. Inf. Comput. Sci., 38, 997 (1998).
[3] K. Funatsu, S. Sasaki, J. Chem. Inf. Comput. Sci., 36, 190
(1996).
[4] C. Peng, S. Yuan, C. Zheng, and Y. Hui., J. Chem. Inf.
Comput. Sci., 34, 805 (1994).
[5] C. Peng, S. Yuan, C. Zheng, Y. Hui, H. Wu, K. Ma, and X.
Han, J. Chem. Inf. Comput. Sci., 34, 814 (1994).
[6] C. Peng, S. Yuan, C. Zheng, Z. Shi, and H. Wu, J. Chem. Inf.
Comput. Sci., 35, 539 (1995).
[7] J.-M. Nuzillard and G. Massiot, Tetrahedron, 47, 3655
(1991).
[8] C. Steinbeck, Angew. Chem. Int. Ed. Engl., 35, 1984 (1996).
[9] C. Steinbeck, J. Chem. Inf. Comp. Sci., 41, 1500 (2001).
[10] J. Meiler, M. C. Will, J. Chem. Inf. Comp. Sci., 41, 1535
(2001).
[11] T. Lindel, J. Junker, M. Koeck, Eur. J. Org. Chem., 3, 573
(1999).
[12] G. E. Martin and C. E. Hadden, J. Nat. Prod., 63, 543 (2000);
R. Marek and A. Lycˇka, Curr. Org. Chem., 6, 35 (2002).
[13] G. E. Martin, "Qualitative and Quantitative Exploitation of
Heteronuclear Coupling Constants," Ann. Rep. NMR Spectrosc., Vol 46,
G. A. Webb, Ed., Academic Press, New York, 2002, pp. 37-100.
[14] K. A. Blinov, M. E. Elyashberg, S. G. Molodtsov, A. J.
Williams, and E. R. Martirosian, Fresenius J. Anal. Chem., 369, 709
(2001).
[15] R. Araya-Maturana, T. Delgado-Castro, W. Cardona, and B.E.
Weiss-Lopez, Curr. Org. Chem., 5, 253 (2001).
[16] M. E. Elyashberg, K. A. Blinov, and E. R. Martirosian, Lab.
Automat. Inf. Management, 34, 15 (1999).
[17] M. H. M. Sharaf, Ph.D. Dissertation, School of Pharmacy,
University of Pittsburgh, 1993, pp. 177-192.
[18] A. Paulo, E. T. Gomes, P. J. Houghton, J. Nat. Prod., 58,
1485 (1995).
[19] M. M. Cooper, J. A. Lovell, and J. A. Joule, J. A. Tetrahedron
Lett., 37, 4283 (1996).
[20] M. H. M. Sharaf, P. L. Schiff, Jr., A. N. Tackie, and G. E.
Martin, J. Heterocyclic Chem. 35, 1365 (1998).
[21] ACD/StrucEluc, V5.08, Advanced Chemistry Development, 90
Adelaide Street W., Suite 702, Toronto, Ontario, M5H 3V9, Canada. All
calculations were performed on a Pentium 1 GHz with 512 Mbytes of RAM
[22] E. Gellert, Raymond-Hamet, and E. Schlittler Helv. Chim.
Acta, 34, 642 (1991).
[23] D. Dwuma-Badu, J. S. K. Ayim, N. I. Y. Fiagbe, J. E. Knapp,
P. L. Schiff, Jr., and D J. Slatkin, J. Pharm. Sci., 67, 433 (1978).
[24] S. Y. Ablordeppy, C. D. Hufford, R. F. Bourne, and D.
Dwuma-Badu, Planta Med., 56, 416 (1990).
[25] A. N. Tackie, M. H. M. Sharaf, P. L. Schiff, Jr., G. L. Boye, R.
C. Crouch, and G. E. Martin, J. Heterocyclic Chem., 28, 1429 (1991).
[26] M. H. M. Sharaf, P. L. Schiff, Jr., A. N. Tackie, C. H. Phoebe,
Jr., L. Howard, C. Meyers, C. E. Hadden, S. K. Wrenn, A. O. Davis, C. W.
Andrews, D. Minick, R. L. Johnson, J. P. Shockcor, R. C. Crouch, and G.
E. Martin, Magn. Reson. Chem., 33, 767 (1995).
[27] Work in the author's laboratory and results recently reported
(see: W. Reynolds and R. G. Enriquez, Magn. Reson. Chem. 39, 531
(2001)) have shown that with very small samples it is disadvantageous to
use gradient-selected long-range experiments. Better results are obtained
with conventional non-gradient phase-cycled experiments.
[28] Only strong (2J or 3J) peaks were entered for the first HMBC
experiment using a 6Hz optimization. A total of 31 peaks were entered.
Only 12 strong 3J peaks were entered for the COSY experiment. Peaks in
the second HMBC experiment, optimized at 8 Hz were manually separated
into two groups (2J/3J and 3J/4J). A total of 22 2,3J peaks and 23 3,4J peaks
were entered. The molecular formula of C32H22ON4 was included as input.
[29] G. E. Martin and C. E. Hadden, J. Nat. Prod., 63, 543 (2000).
[30] G. E. Martin and C. E. Hadden, Magn. Reson. Chem., 38, 251
(2000).
[31] C. E. Hadden, G. E. Martin, and V. V. Krishnamurthy, J.
Magn. Reson., 2000 38, 143 (2000).
[32] A total of three long-range 1H-15N correlations were added to
the input dataset. One was entered as a long-range (4-5J) coupling.
1250 Vol. 39