An effective strategy for exploring unknown metabolic pathways by genome mining

Article abstract of DOI:10.1021/ja401535g

Plants allocate an estimated 15-25% of their proteome to specialized metabolic pathways that remain largely uncharacterized. Here, we describe a genome mining strategy for exploring such unknown pathways and demonstrate this approach for triterpenoids by functionally characterizing three cytochrome P450s from Arabidopsis thaliana. Building on proven methods for characterizing oxidosqualene cyclases, we heterologously expressed in yeast known cyclases with candidate P450s chosen from gene clustering and microarray coexpression patterns. The yeast cultures produced mg/L amounts of plant metabolites in vivo without the complex phytochemical background of plant extracts. Despite this simplification, the product multiplicity and novelty overwhelmed analytical efforts by MS methods. HSQC analysis overcame this problem. Side-by-side HSQC comparisons of crude P450 extracts against a control resolved even minor P450 products among ～100 other yeast metabolites spanning a dynamic range of >10 000:1. HSQC and GC-MS then jointly guided purification and structure determination by classical NMR methods. Including our present results for P450 oxidation of thalianol, arabidiol, and marneral, the metabolic fate for most of the major triterpene synthase products in Arabidopsis is now at least partially known.

Full text of DOI:10.1021/ja401535g

Article
pubs.acs.org/JACS
An Effective Strategy for Exploring Unknown Metabolic Pathways by
Genome Mining
Dorianne A. Castillo,† Mariya D. Kolesnikova,† and Seiichi P. T. Matsuda*,†,‡
†Department of Chemistry and ‡Department of Biochemistry and Cell Biology, Rice University, Houston, Texas 77005, United States
S
* Supporting Information
ABSTRACT: Plants allocate an estimated 15−25% of their
proteome to specialized metabolic pathways that remain
largely uncharacterized. Here, we describe a genome mining
strategy for exploring such unknown pathways and demon-
strate this approach for triterpenoids by functionally character-
izing three cytochrome P450s from Arabidopsis thaliana.
Building on proven methods for characterizing oxidosqualene
cyclases, we heterologously expressed in yeast known cyclases
with candidate P450s chosen from gene clustering and
microarray coexpression patterns. The yeast cultures produced mg/L amounts of plant metabolites in vivo without the
complex phytochemical background of plant extracts. Despite this simplification, the product multiplicity and novelty
overwhelmed analytical efforts by MS methods. HSQC analysis overcame this problem. Side-by-side HSQC comparisons of
crude P450 extracts against a control resolved even minor P450 products among ∼100 other yeast metabolites spanning a
dynamic range of >10 000:1. HSQC and GC−MS then jointly guided purification and structure determination by classical NMR
methods. Including our present results for P450 oxidation of thalianol, arabidiol, and marneral, the metabolic fate for most of the
major triterpene synthase products in Arabidopsis is now at least partially known.
are converted to >20 000 triterpenoids.11,12 Genome mining13
INTRODUCTION
■
has been enormously successful in functionally characterizing
Natural selection constructs in each plant species a versatile
oxidosqualene cyclases (OSCs) by heterologous expression.14
genetic arsenal for biosynthesizing specialized (secondary)
In Arabidopsis, this approach15,16 has identified far more OSC
metabolites that are accessible for defense and other needs of
sessile organisms.1−3 These natural products are collectively
products (triterpenes) than have been found by direct analysis
of plant material.7b,17,18
much greater in number than primary metabolites and
Extending genome mining to the subsequent metabolism of
triterpenes to triterpenoids was envisioned a decade ago.19,20
Since then, heterologous expression has been used to study
glycyrrhizin biosynthesis21 and other oxidative pathways22 of
pentacyclic triterpenes to known products that were identified
mainly by GC−MS or LC−MS comparisons with authentic
standards. Others23,24 tackled the difficult problem of
elucidating unknown pathways for P450 oxidation of two
Arabidopsis triterpenes, thalianol and marneral. Their earlier
discovery of coregulated gene clusters in plants25 facilitated
judicious choices of candidate P450s clustered with the
thalianol and marnerol synthases. Gene overexpression
demonstrated the P450 oxidations in planta. However, GC−
MS gave only partial chemical structures of the novel products,
a result underscoring the reliance of MS-based methods26 on
databases and authentic standards. Even with standards for the
few known oxygenated derivatives of these unusual triterpene
skeletons,27−29 GC−MS would be challenged to identify the
regio- and stereochemistry of oxidation unambiguously.
structurally more diverse. Biosynthetic pathways of primary
metabolism are well established, whereas the numerous
pathways leading to specialized metabolites in plants are mostly
unknown.1
Cytochrome P450s play a major role in specialized
metabolism. The Arabidopsis genome contains ∼245 cyto-
chrome P450s, of which only ∼20% have been functionally
characterized.4,5 Genomics can reliably identify most P450s of
primary metabolism by homology because these genes are well
conserved across families. However, elucidating specialized
metabolic pathways is more difficult, and few of the estimated
200 Arabidopsis P450s of specialized metabolism have been
characterized.6 These few studies generally start from a known
natural product as a basis for identifying the P450s. This
approach depends on knowledge of the specialized metabolite,
most of which are yet unidentified.7 Filling this gap in
metabolomics and functional genomics is the topic of this
Article.
A large branch of specialized metabolism is terpenoid
biosynthesis.3 A few linear terpenes (C10, C15, C20, C30, and
C40) undergo cyclization, P450 oxidation, and other enzymatic
reactions to make >40 000 terpenoids.8,9 Within the C30 subset,
oxidosqualene is cyclized to >100 triterpene skeletons,10 which
Received: February 11, 2013
Published: April 9, 2013
© 2013 American Chemical Society
5885
dx.doi.org/10.1021/ja401535g | J. Am. Chem. Soc. 2013, 135, 5885−5894

Journal of the American Chemical Society
Article
Building on this previous work and our experience in
functionally characterizing OSCs,13,15,16,30−33 we used heterol-
ogous expression and 2D NMR to develop an effective strategy
for exploring unknown pathways. Relative to experiments in
planta, heterologous expression34 eliminated the background
interference of extraneous phytochemicals and supplied ample
amounts of P450 products for NMR analysis. HSQC then
became a powerful tool for analyzing crude extracts, which
often contain a variety of triterpenoids due to the product
multiplicity16,31,32 that is common in specialized metabolism.
Heterologous expression and HSQC thus overcame major
hurdles in exploring unknown metabolic pathways.
Here, we apply this genome mining strategy to study the
enzymatic oxidation of three Arabidopsis OSC products:
thalianol, arabidiol, and marneral. Each OSC was hetero-
logously expressed with a coregulated Arabidopsis P450, which
was functionally characterized by identifying the major
oxidation products using GC−MS and 2D NMR. Our
experimental design is directly applicable to the elucidation of
other unknown P450 pathways of triterpenoid metabolism,
where the novel isomeric products might confound the high-
throughput methods of metabolomics17b and dereplication.35
This general approach and its variations36,37 may be useful in
exploring how plants and other organisms creatively exploit the
powerful genetic machinery available for specialized metabo-
lism.38
oxidation. These structural ambiguities would not likely be
resolved by MS methods, including high-resolution LC−MS/
MS.
We then turned to 2D NMR for mixture analysis.42 Unlike
1D NMR, HSQC of the crude yeast extracts resolved many
triterpenoid methyl signals. Often these and other correlated
1H and 13C chemical shifts revealed structures of novel
triterpenoid products prior to purification. Moreover, compar-
ing HSQC spectra of crude P450 yeast extracts against the
corresponding control distinguished P450 metabolite signals
from the numerous unidentified peaks. These facile side-by-side
or overlaid comparisons in Topspin software quickly identified
even many minor (1%) metabolites in the crude extracts. GC−
MS and NMR then guided chromatographic purification for
classical structure determination.
7β-Hydroxythalianol Is the Primary CYP708A2Meta-
bolite of Thalianol. CYP708A2 is known to metabolize
thalianol.23 To determine the location and configuration of this
oxidation, we studied thalianol oxidation in JBY575 yeast
transformed with thalianol synthase (THAS1, At5g48010),
CYP708A2 (At5g48000), and ATR2 (Scheme 1). GC−MS
Scheme 1. Oxidation of Thalianol by CYP708A2 To Form 3
(Solid Arrow), Accompanied by Oxidation to 4 Either
Directly by CYP708A2 or via 5 by a Yeast P450 (Dashed
Arrows)
RESULTS
■
Experimental Design. We studied the P450 oxidation of
three triterpenes (thalianol, arabidiol, and marneral) by
heterologous expression in vivo in yeast strains coexpressing
the OSC, the candidate P450, and its redox partner ATR2 (an
Arabidopsis NADPH-cytochrome P450 reductase). The yeast
strains efficiently produced the foreign triterpene substrates in
vivo, thus avoiding the complications of performing in vitro
experiments. The resulting P450 metabolites were analyzed by
GC−MS and 2D NMR.
More specifically, we selected candidate P450s based on gene
cluster analysis39 and coexpression40 relative to each triterpene
OSC (Figures S2 and S3), as done previously.23,24 Recombi-
nant yeast strains were constructed by transforming the JBY575
strain41 with an OSC, a P450, and ATR2. Three analogous
control strains were made without the Arabidopsis P450. The
OSCs were available from previous work.13,30,32 Two P450s
(CYP705A1 and CYP71A16) were obtained as plasmids from
resource centers. Another P450 (CYP708A2) and ATR2 were
synthesized commercially with codon optimization. Determi-
nation of the correct protein sequence for ATR2 is described in
text accompanying the sequence alignment shown in Figure S1.
The three yeast strains and their controls were grown in liquid
cultures. The putative P450 metabolites were obtained from
cells as a crude ethanol extract comprising triterpenoids,
triterpenes, yeast sterols, and other lipids.
An efficient assay for P450 metabolites in the crude extracts
was needed for the many trial cultures for each of the several
yeast strains. However, recognizing the metabolites as minor
components in a complex mixture was challenging because
authentic standards and spectral data were generally unavail-
able. Following others23 in using a dereplication approach, we
initially resorted to spectral comparisons against the substrate.
For anticipated hydroxylations, GC−MS of trimethylsilyl
(TMS) ethers showed peaks with the expected molecular ion
but gave limited clues about the location and stereochemistry of
analysis of the crude product as TMS ethers (Figure 1A)
showed peaks for thalianol (2; m/z 498) and two putative
hydroxythalianols (3 and 4; m/z 586) that were absent in a
control culture lacking CYP708A2. Comparison of the MS
fragmentation of 3 and 4 versus 2 suggested that hydroxylation
occurred in ring B or C, but the exact position was ambiguous.
A previous in planta study of the same CYP708A2 oxidation23
reached a similar conclusion based on GC−MS spectra
resembling those of 3 and 4. Reversed-phase HPLC (Figure
1B) provided purified samples of 3 and 4 that were analyzed by
NMR (Figure 1C). 2D NMR correlations (Figure 2)
established the location and configuration of hydroxylation
for 3 and 4 as 7β and 7α, respectively. The minor product 7-
ketothalianol (5) was partially purified by HPLC, with structure
determination by 2D NMR. Additionally, one or more of some
very minor triterpenoids observed by GC−MS might have
originated from enzymatic oxidation of thalianol (Figures
S5,S6).
Did CYP708A2 generate a mixture of 3−5 or a single 7-
oxygenated product that was further metabolized by one of the
5886
dx.doi.org/10.1021/ja401535g | J. Am. Chem. Soc. 2013, 135, 5885−5894

Journal of the American Chemical Society
Article
A−H, which were analyzed by GC−MS as TMS ethers. GC−
MS of the main polar fraction (E) (Figures 3A and S9)
Figure 1. Oxidation of thalianol by CYP708A2. (A) GC−MS of the
crude extract as the total ion chromatogram (TIC) and MS of 3. (B)
HPLC purification separated 3 and 4. (C) Homogeneity of purified 3
1
Figure 3. Oxidation of arabidiol by CYP705A1. (A) GC−MS and (B)
and 4 (except for fatty acids, marked by *) as judged by H NMR.
1
800 MHz H NMR of chromatographic fraction E (saponified P450
metabolites; 7 and 8 proved to be families of isomers). (C) HSQC
(600 MHz) comparison of control and P450 spectra of the
unsaponified crude extracts; the asterisk denotes a minor P450
metabolite.
suggested side-chain oxidation of arabidiol to a secondary
alcohol (7; m/z 586) and a nonisomeric tertiary alcohol (8;
m/z 498) derivative. The third peak in the triterpenoid region
(tR 12−27 min) represented coeluting mono-TMS ethers of
(20R)- and (20S)-dammarenediols (9a and 9b), which are
PEN1 cyclization byproducts rather than P450 metabolites.
Had we relied on GC−MS or LC−MS analysis, we would have
annotated CYP705A1 as a hydroxylase that makes side-chain
oxygenated derivatives of arabidiol.
Anticipating a mixture containing comparable amounts of 7,
8, and 9a/9b, we proceeded with NMR analysis. Unexpectedly,
1D NMR (Figure 3B) showed one major component with only
four methyl singlets and none of the usual side-chain
resonances of tricyclic triterpenoids. 2D NMR revealed that
the side chain had been cleaved at the C14−C15 bond, leaving
a C19 fragment (Scheme 2).44 We informally denote this 13R
C19 methyl ketone (10a) as 14-apo-arabidiol (borrowing from
carotenoid nomenclature). SciFinder substructure searching
uncovered no remotely similar compounds except for 14-
hydroxy analogues of nonbiological origin with different
stereochemistry in the ABC ring system.45
Figure 2. Classical 2D NMR confirmed the structure of 3 and 4. (A)
HMBC and COSYDEC correlations established hydroxylation at C7.
(B) NOESY correlations established the C7 configurations. Distances
from H7 to remote (red) and nearby (blue) hydrogen atoms are
shown for the major side-chain conformation (B3LYP/6-31G*
geometry).
seven S. cerevisiae P450s? To clarify this ambiguity, which is
often encountered when P450s are heterologously expressed,43
we took aliquots of a culture at ca. 11 h intervals and measured
the relative amounts of 3 and 4 by GC−MS (Table S1). The
ratio of 3:4 was ∼10:1 at early time points and decreased to 2:1
at saturation. This trend is consistent with initial P450
oxidation to 3, which is then metabolized further by oxidation
to 5 and reduction to 4 by CYP708A2 and/or yeast enzymes.
CYP705A1 Cleaves the Arabidiol Side Chain to a C19
Product. Metabolism of arabidiol (6) was studied analogously
in cultures of JBY575 expressing arabidiol synthase (PEN1,
At4g15340), CYP705A1 (At4g15330), and ATR2. Flash
chromatography of the saponified ethanol extract gave fractions
Among the minor products in the saponified mixture was the
13S epimer 10b, which represents partial epimerization during
saponification.46 Subsequent experiments confirmed that only
10a is formed when workup excluded saponification. Side-by-
side HSQC comparison of crude extracts from the CYP705A1
construct against a control lacking the P450 (Figure 3C)
showed 10a, 7, and 8 only in the CYP705A1 extracts, whereas
5887
dx.doi.org/10.1021/ja401535g | J. Am. Chem. Soc. 2013, 135, 5885−5894

Journal of the American Chemical Society
Article
mixture of three species, the underivatized aldehyde and the E
and Z TMS enol ethers.
Scheme 2. Cleavage of the Arabidiol Side Chain To Form
10a
GC−MS and HSQC were used to compare the triterpenoid
profile in CYP71A16 cultures and control cultures lacking the
P450. GC−MS (Figure 4A) showed TMS derivatives of a
both cultures contained arabidiol and at lower levels
dammarenediols, other PEN1 byproducts, and yeast sterols.
Why did we initially overlook 14-apo-arabidiol in the GC−
MS results? Not anticipating a C19 product, we did not
exhaustively investigate the early fatty-acid region of the
chromatogram. Even after the NMR identification of 10a and
10b, locating their GC−MS peaks was not trivial. Both epimers
were present as mixtures of mono- and bis-TMS ethers, and
their MS fragmentation showed only weak ions for losses of
CH3 and TMSOH. Their abundant low-mass ions (Figure S8)
resembled mass spectra of early eluting lipids commonly
observed in yeast extracts more than ions of tricyclic
triterpenoids (Figures S6,S7). After this humbling experience,
we routinely surveyed P450 product profiles by analyzing crude
extracts using both HSQC and GC−MS.
CYP71A16 Hydroxylates Marneral Selectively at C23.
Cultures of JBY575 transformed with marneral synthase
(MRN1, At5g42600), CYP71A16 (MRNO, At5g42590), and
ATR2 (At4g30210) were extracted, chromatographed, and
analyzed spectrally as in the CYP708A2 and CYP705A1 studies.
In this case, the mixture of P450 metabolites, unoxidized
triterpenes, yeast sterols, and other lipids was further
complicated by the presence of the OSC product as a
combination of aldehyde marneral (11a) and alcohol marnerol
(11b) (Scheme 3). This duality, which has been noted in both
Figure 4. Hydroxylation of marneral at C23 by CYP71A16. (A) GC−
MS and (B) HSQC comparison of crude extracts from cultures with
(“P450”) and without (“control”) CYP71A16. The second GC−MS
peak for 11a corresponds to a TMS enol ether of the aldehyde. (C)
GC−MS of the CYP71A16 products 12a and 12b after HPLC
purification.
Scheme 3. Marneral Hydroxylation by CYP71A16
hydroxymarnerol product (12b; m/z 588) and a small
hydroxymarneral peak (12a) that were unique to the P450
culture. Both cultures contained marnerol (11b; m/z 500) and
marneral (11a; m/z 498 underivatized). HSQC, which avoided
the complications of TMS derivatization, cleanly showed the
same pattern (Figure 4B) and indicated hydroxylation at C23
or C24. 1D NMR was used to detect and identify trace
amounts of 3- and 23-aldehyde species, which gave distinctive
chemical shifts in the 9.5−10.5 ppm region.
Flash chromatography on silica gel and reversed-phase
HPLC furnished samples of 12a and 12b (Figure 4C). 2D
NMR showed the hydroxylation to be at the Z allylic methyl
28
1
(C23). H NMR data reported for 12b were consistent with
our NMR results but insufficient for unambiguous structure
identification due to the modest chemical-shift precision
( 0.01 ppm) and lack of 13C NMR data. Very minor
unidentified side-chain oxygenated marnerol derivatives,
possibly CYP71A16 metabolites, were observed in GC−MS
and HSQC spectra (Figures S10,S11). We detected only traces
plant24,47 and yeast24,30 extracts, extended to the P450
metabolites (12a and 12b). The aldehyde reduction might
arise from yeast metabolism and/or workup with or without
saponification. Another analytical complication was TMS
derivatization, which converted each aldehyde moiety into a
5888
dx.doi.org/10.1021/ja401535g | J. Am. Chem. Soc. 2013, 135, 5885−5894

Journal of the American Chemical Society
Article
(<0.1%) of 23-aldehyde species, such as the marnerol
derivative, for which NMR data have been reported.28 Thus,
CYP71A16 hydroxylates marnerol/marneral in high accuracy
specifically at C23.
HSQC Gives a Comprehensive Profile of Substances in
Crude Extracts. Although 2D NMR is used mainly for
structure determination of purified compounds, we found
HSQC to be valuable for analyzing crude extracts from yeast
cultures (Figures 3C and 4B). These figures showed only the
upper intensity levels for ∼0.1% of the total HSQC spectral
area. Digging deeper and more broadly into a similar HSQC
spectrum (Figure 5) revealed a multitude of minor P450
metabolites, OSC triterpene byproducts, yeast sterols, and
other lipids. Some spectral regions were degraded by artifacts
from major peaks (sinc ripples from FID truncation, peak
shoulders from imperfect shimming, and sidebands from
adiabatic decoupling), but >95% of the spectral area showed
peaks spanning a dynamic range of >17 000:1 (Figure 5E). This
typical HSQC of a crude yeast extract contained thousands of
peaks representing ∼100 compounds ranging from <1 μg to >1
mg.
The expanded HSQC snapshots in Figure 5A and B illustrate
the resolution of many upfield methyl signals of minor sterols
and triterpenoids. Comparing their correlated 1H and 13C
chemical shifts with literature data can reveal the location of
double bonds and functional groups in the tricyclic ring system
and nearby side chain. The H-5α signals (Figure 5D) gave
complementary structural insights. HMBC and COSYDEC
spectra can augment the HSQC results but are more
appropriate for simpler mixtures due to their greater signal
overlap and lower sensitivity.
Figure 5 illustrates the power and versatility of HSQC
analyses. HSQC can be used classically with other 2D NMR
methods for determining the structure of unknown substances.
The samples need not be pure. For example, the structure
determination of 10a and 10b was performed on a 4:1 mixture
containing many minor triterpene metabolites (Figures 3B and
S20−S22). HSQC can also be used to identify known
1
compounds by matching pairs of H−13C correlated chemical
shifts with literature data. This identification is highly reliable
1
when a good match ( 0.002 ppm for H, 0.02 ppm for 13C)
is obtained for distinctive signals representing C−H pairs that
span the triterpenoid structure. The sterols and triterpenes in
Figure 5E were identified by this method. The utility of HSQC
for structure determination, quantification, and compound
identification in complex mixtures is well-known.42
DISCUSSION
■
This work demonstrates the power of combining genome
mining, heterologous expression, and HSQC analysis to
elucidate unknown biosynthetic pathways and identify their
novel products. We illustrate this strategy by functionally
characterizing triterpene oxidation by three Arabidopsis P450s
and rigorously determining the structures of the triterpenoid
metabolites. We show that CYP708A2 oxidizes thalianol to 7β-
hydroxythalianol (3), that CYP705A1 cleaves the side chain of
arabidiol to give a C19 methyl ketone (10a), and that
CYP71A16 hydroxylates an allylic methyl of marneral/marnerol
to 23-hydroxymarneral/23-hydroxymarnerol (12a/12b). The
three P450 oxidations illustrate diverse outcomes: hydrox-
ylation of an allylic CH2 in the ring system, side-chain cleavage
resembling mammalian pregnenolone synthesis, and hydrox-
ylation of an allylic methyl. Side-chain hydroxylation and
desaturation are other potential outcomes that were not
observed as significant products.
The hydroxylation of marneral/marnerol at C23 by
CYP71A16 resembles marneral/marnerol metabolism in the
iris family.29 Whereas the iris pathway leads to fragrant irones,
the pathway in Arabidopsis obviously does not. The seeming
coincidence of 23-hydroxylation in these distant taxa merely
reflects the broad distribution of P450 families in angiosperms48
and the limited number of triterpene skeletons and common
oxidation sites. Arabidopsis and iris evidently recruited and
modified a CYP71 gene for the same commonplace
Figure 5. HSQC spectrum of crude extracts from yeast expressing
PEN1, ATR2, CYP705A1, and CYP705A2. The latter P450 generates
little or no additional arabidiol oxidation products, as judged by
comparison with the HSQC spectrum of Figure 4B. (A and B)
Expansions showing methyl singlets for major metabolites (6, 10a),
unidentified minor CYP705A1 metabolites (E1−E6, F1−F2, and G1),
PEN1 cyclization byproducts (B1−B5), and yeast sterols (B6−B7).
Peak labels correspond to flash chromatography fractions B−G; P450
metabolites are underlined. (C) Overview of the upfield methyl region.
(D) Expansion showing the H-5α signals (doublets in F2, actual
chemical-shift scale in F1). (E) Positive HSQC peak intensities span a
dynamic range of >17 000:1. Negative peak intensities cover a similar
dynamic range. Identity of peaks B1−B7: B1, 14-epithalianol; B2,
(13S)-malabarica-14(27),17,21-trienol; B3, (13S)-malabarica-
14E,17,21-trienol; B4, 13R epimer of B2; B5, 13R epimer of B3; B6,
lanosterol; B7, 4,4-dimethylcholesta-8,24-dienol.
5889
dx.doi.org/10.1021/ja401535g | J. Am. Chem. Soc. 2013, 135, 5885−5894

Journal of the American Chemical Society
Article
hydroxylation of the same triterpene skeleton. Such convergent
evolution is unsurprising in specialized metabolism.1,49
considering the promiscuity of specialized metabolism.53 Even
in well-established pathways of primary mammalian metabo-
lism, such ambiguities can be difficult to resolve.54 Identification
of final metabolic products and their biological roles would
provide a needed framework for evaluating the many divergent
experimental results and determining how closely the pathways
correlate to gene clusters in Arabidopsis.
The recent completion of functional characterization for the
13 Arabidopsis OSCs16 indicates 13 triterpenes as major OSC
products (Figure 6): cycloartenol (13) derived from CAS1
cyclization, lupanediol (14) and lupeol (17) from LUP1, β-
amyrin (16) from LUP2 and LUP4, camelliol C (18) from
LUP3, tirucall-7,24-dienol (19) and isotirucallol (20) from
LUP5, arabidiol (6) from PEN1, baruol (21) from PEN2, 19
again from PEN3, thalianol (2) from PEN4, marneral (11a)
The side-chain cleavage of arabidiol is uncommon among
P450 oxidations.4a,50 The mechanism could resemble that of
pregnenolone synthesis from cholesterol,51 which involves
sequential P450 hydroxylations at C22 and C20, followed by
cleavage with loss of isocaproaldehyde. Another class of
cleavage mechanisms (Figure S4) is exemplified by the P450
oxidation of nerolidol to 4,8-dimethyl-1,3,7-nonatriene
(DMNT) and presumably but-1-en-3-one.52 DMNT and its
C16 homosesquiterpene analogue TMTT are observed in
Arabidopsis leaf6a and root.52 TMTT is synthesized in leaf from
(E,E)-geranyllinalool by CYP82G1 oxidative cleavage, which
can also efficiently convert nerolidol to DMNT.6a Interestingly,
the arabidiol side chain is an exact substructure of nerolidol.
Moreover, oxidative cleavage of nerolidol by CYP82G1 and of
arabidiol by CYP705A1 occurs at identical positions, removing
the same C11 substructure and (presumably) converting the
tertiary alcohols to methyl ketones. CYP82G1 and CYP705A1
are in the same CYP71 clan. If DMNT is cleaved from
arabidiol, this would represent an interesting variant of
convergent evolution in specialized plant metabolism.1,49 Is
the target arabidiol metabolite for Arabidopsis 14-apo-arabidiol
or the unidentified side-chain fragment (perhaps DMNT)?
Several recurrent themes pervade this work: the greater
difficulty of exploring unknown versus partially known
metabolic pathways; the power of heterologous expression in
vivo to generate large amounts of metabolite free of a complex
phytochemical background (tempered by the potential for
generating artifacts of yeast metabolism); the formation of
minor enzymatic products, as commonly observed in
specialized metabolism; the challenge of detecting the often
novel metabolites during product isolation; the importance of
flexibility in selecting isolation methods ranging from chemical
transformation to the many variants of solvent extraction, with
or without saponification; the need for analytical methods that
are not dependent on authentic standards; and the value of
using MS and HSQC analysis concurrently at all stages of
purification.
HSQC analysis of crude mixtures proved to be a powerful
tool for exploring unknown pathways. With the ability to
disperse signals in two dimensions over an intensity range of
>10 000:1, HSQC resolved many minor triterpenoids that were
obscured by peak overlap in GC−MS. Differential 2D NMR
analyses42f were especially useful in comparing crude extracts of
cultures with and without a candidate P450 gene. Side-by-side
HSQC comparisons in Topspin software smoothly identified
metabolite peaks absent in the control spectrum. Structural
insights were gleaned from comparisons against the known
chemical shifts of the triterpene substrate, which are
1
reproducible to 0.001 ppm for H and 0.01 ppm for 13C
NMR.16,18,31−33
Our results provide the first experimental support for gene
clustering proposed for arabidiol metabolism4a,23,39b and add to
existing support23,24 for thalianol and marneral clusters. Can
these clusters be exploited in silico to predict the step order and
length of an entire metabolic pathway, such as helvolic acid
biosynthesis in fungi?33 Despite strong evidence from gene
coexpression, gene clustering, parallel observations in planta,23
and reaction efficiency with high regio- and stereoselectivity in
recombinant yeast, might different P450s with compatible
spatial and temporal expression perform the triterpene
oxidations in planta? Simple answers may be elusive
Figure 6. Currently known metabolic fate of the primary products of
triterpene synthases (OSCs) in Arabidopsis. *CYP705A5 appears to
metabolize the product of CYP708A223 (i.e., 3) and is separately
reported to be involved in flavonoid biosynthesis.55
5890
dx.doi.org/10.1021/ja401535g | J. Am. Chem. Soc. 2013, 135, 5885−5894

Journal of the American Chemical Society
Article
and extracted with ethanol aided by sonication or bead beating. After
further centrifugation to remove cell debris, the ethanol fraction was
collected by decanting and evaporated to a residue that was partitioned
between water and ether. The ether layers were evaporated to a “crude
extract”, which was analyzed by GC−MS and HSQC for triterpenoid
content. The extract was then purified chromatographically, with
characterization of individual fractions by GC−MS and NMR. In an
alternative isolation method used occasionally, the cell pellet was
extracted by a modified Folch method (2:1 CH2Cl2−methanol),
followed by partitioning of the evaporated residue between MTBE and
water, with or without saponification (10% KOH in 80:20 ethanol−
water at 70 °C for 1 h). Control cultures lacking the P450 plasmid
were grown under conditions and workup procedures identical to
those used for the corresponding P450 strain.
All cultures were grown at 30 °C with shaking at 250 rpm in
Erlenmeyer flasks loosely capped with aluminum foil. The experiments
were repeated many times at different scales; 100 mL cultures in 250
mL Erlenmeyer flasks gave consistently good production of oxy-
genated triterpenoids and yeast (1.5−2 g cells/100 mL). Larger
cultures (e.g., 250 mL in 1 L flasks) gave lower amounts of oxygenated
products relative to OSC products, perhaps because of insufficient
aeration.
Chromatographic Purification. The “crude” or Folch extracts
were subjected to flash chromatography using Strata solid-phase
extraction (SPE) cartridges (1, 2, or 5 g of silica gel; 55 μm particle
size, 70 Å pore size) from Phenomenex (Torrance, CA) in a standard
SPE vacuum manifold. This purification was occasionally supple-
mented by preparative TLC on silica gel 60 glass-backed plates (250
μm layer). Often further purification was done by preparative HPLC
on an Agilent 1100 system using UV detection at 210 nm, a 250 mm ×
21.2 mm C18 column, and a linear gradient of 85−100% methanol in
water at ca. 23 °C.
GC−MS Analysis. GC−MS was done on an Agilent 6890 GC
interfaced to a 5973N MS detector and containing a Restek 30 m ×
0.25 mm i.d. Rxi-35Sil column (35% phenyl−65% methylpolysiloxane,
0.25 μm film thickness). Triterpenoid samples were derivatized as
TMS ethers by heating 1−5% of the sample in 60 μL of 2:1 pyridine−
bis(trimethylsilyl)trifluoroacetamide for 1 h. After pulsed splitless
injection of this mixture (2 μL) at 280 °C, the oven temperature was
held at 110 °C for 1 min, increased at 30 °C/min to 270 °C, and then
increased at 0.5 °C/min to 280 °C (held 8 min). The helium carrier
gas was maintained at a constant flow of 2.7 mL/min. Electron-impact
MS data were collected in full-scan mode (m/z 50−650) and analyzed
with Agilent Chemstation software.
NMR Analysis. NMR spectra were acquired at 25 °C in CDCl3
solution on 600−800 MHz instruments equipped with a cryogenic
probe. Referencing was to internal tetramethylsilane. Spectra were
processed with Topspin 3 software. 2D NMR spectra were obtained
with standard pulse sequences, usually with echo-antiecho phase
encoding, shaped pulses, and adiabatic decoupling (HSQC). HSQC,
HMBC, COSYDEC (F1-decoupled COSY), and NOESY spectra were
acquired using the following respective parameters: acquisition time,
0.2−0.3, 0.5, 0.5, 0.5 s; d1 delay, 1.2, 1.1, 1.0, 1.0 s; pulse sequence,
hsqcedetgpsp.3, hmbcetgpl2nd, cosyjdqf, noesyetgp (or near equiv-
alents in vnmrj); F1 spectral width, 30, 70, 1.4, 1.4 ppm; o2p (center of
F1 window), 56, 45, 2.8, 2.8 ppm; number of FIDs, 512−1024, 400−
1024, 332, or 444 (600 or 800 MHz with d6 = 0.2 s), 500−800.
COSYDEC spectra were acquired with a d6 constant-time period of
0.2 s (sometimes also 0.3 s). The NOESY mixing time was 0.5 s. 2D
spectra were processed with 4-fold linear prediction in F1 and π/2.5-
shifted qsine apodization in F2 and F1 (HSQC, HMBC, NOESY)
using 4k and 2k points (si). The t1away macro (70−90%) was used as
needed to attenuate t1 noise. Reduced F1 windows (resulting in limited
spectral aliasing) and 256−512 t1 increments allowed measurement of
chemical shifts to ca. 0.01 ppm (13C) or 0.001 ppm (1H). The F1
ppm labels in figures were adjusted to minimize confusion from
aliasing.
from PEN5, seco-β-amyrin (22) from PEN6, and lanosterol
(23) from LSS1. Cycloartenol is further metabolized to plant
membrane sterols and lupanediol to trinorlupeol (15).18 β-
Amyrin and lupeol are found unmodified in stem wax, silique,
and bud.18 With our results for P450 oxidation of thalianol,
arabidiol, and marneral/marnerol, the metabolic fate of most of
the primary products of Arabidopsis OSCs is now at least
partially known. This emerging picture is consistent with
observations7b,17a,18 suggesting that Arabidopsis does not
oxidize the nonsteroidal triterpenes that it produces con-
stitutively.
CONCLUSIONS
■
A substantial portion of plant genomes is dedicated to
specialized metabolic pathways that are taxonomically narrow,
fast evolving,56 tissue specific,1 catalytically promiscuous,53 and
often silent.38 These pathways contain a wealth of information
about selective pressures, defense mechanisms, P450 function,
and plant metabolomes. However, specialized pathways are
barely explored. Genomic analysis usually57 fails because the
narrowly distributed pathways lack true orthologs. High-
throughput MS-based metabolomics often57 struggles to
identify novel specialized metabolites, whose structural diversity
can defy the largest MS databases and collections of authentic
standards. Traditional natural products isolation42a can solve
structures but overlooks silent metabolism and rarely connects
chemical structures with systems biology.58,59 We demonstrate
an alternative strategy for exploring specialized pathways that
incorporates elements of genomics (to optimize heterologous
expression), metabolomics (for complementary60 HSQC and
MS dereplication of crude extracts), and natural products
methods (for isolation and final structure elucidation). This
strategy was successful in all three systems that were studied.
Continued efforts should ultimately provide a comprehensive
accounting of Arabidopsis triterpenoids. Applied more generally
to exploring unknown metabolic pathways, this strategy may
begin to fill a large knowledge gap in metabolomics and
functional genomics.
EXPERIMENTAL SECTION
■
Subcloning and Yeast Transformation. CYP705A1 and
CYP71A16 were obtained from the Riken Bioresource Center and
the Arabidopsis Biological Resource Center (ABRC), respectively.
CYP708A2 and ATR2 were synthesized and codon optimized for yeast
by GenScript. These genes were subcloned into yeast expression
vectors pRS424Gal, pRS423Gal, or pRS426Gal. Construction of
similar plasmids has been described for thalianol,13 arabidiol,32 and
marneral30 synthases (OSCs). Appropriate sets of OSC, P450, and
ATR2 plasmids were then used sequentially to transform Saccha-
romyces cerevisiae strain JBY575 (MATa ura3−52 trpl-A63 leu2−3,112
his3-A200 ade2 Gal+)41 by the lithium acetate method. The resulting
transformants (JBY575[pGCF4.0, pDCR5.1, pDCR4.1], JBY575-
[pMDK3.6, pDCR9.3, pDCR4.1], and JBY575[pXQ11.2, pDCR8.3,
pDCR4.1]) were used to study P450 metabolism of thalianol,
arabidiol, and marneral heterologously in vivo. Analogous control
strains lacking the foreign P450 were prepared similarly. Specific
details for constructing each transformant are given in the Supporting
Information.
Yeast Culture and Product Isolation. Each JBY575 transformant
was selected on synthetic complete medium lacking uracil, tryptophan,
and histidine, supplemented with 2% glucose, and solidified with 1.5%
agar. Cells from the resulting colonies were grown in 10 mL liquid
cultures in the above medium, followed by inoculation into 100 mL
batches of inducing medium (the same medium with galactose in place
of glucose). At saturation the cells were harvested by centrifugation
Dynamic range in HSQC spectra was estimated from ratios of peak
heights. Manual peak picking was used to choose the largest positive
and negative signals (2 516 671 792 and −2 549 601 544, terminal
5891
dx.doi.org/10.1021/ja401535g | J. Am. Chem. Soc. 2013, 135, 5885−5894

Journal of the American Chemical Society
Article
CH3 and CH2 peaks of long-chain impurities) and the smallest signals
that were distinguishable from noise. A small positive peak (144 480)
was chosen in the weakest clearly defined olefinic doublet (10 Hz
coupling) among a group of similar doublets.
compatibility with quantum mechanical predictions of NMR chemical
shifts (Tables S5−S7).
ASSOCIATED CONTENT
* Supporting Information
■
Isolation of Thalianol Oxidation Products. The crude extract of
a preliminary 100 mL culture of JBY575[pGCF4.0, pDCR5.1,
pDCR4.1] showed likely thalianol metabolites23 by GC−MS.
Purification by preparative TLC (developed with 1:2 ether/hexanes)
gave the major metabolite 3, which was tentatively identified as 7β-
hydroxythalianol by NMR. The minor oxygenated thalianols were
investigated with a larger culture (5 × 100 mL). The crude ethanol
extract was subjected to flash chromatography (1 g silica SPE
cartridge; elution with gradients of ether/hexanes). The 1:1 ether−
hexanes fractions contained 3 and two minor compounds. As indicated
in Figure 4C, preparative reversed-phase HPLC separated the three
oxygenated thalianols. 2D NMR suggested the two minor compounds
to be the 7α-hydroxy and 7-keto thalianols 4 and 5. A third experiment
(17 × 100 mL) using the same extraction and purification procedures
gave ample amounts of 3, 4, and 5 for rigorous structure determination
by HSQC, HMBC, COSYDEC, and NOESY (see the Supporting
Information). Cultures of the control strain (lacking pDCR5.1)
showed no evidence of 3, 4, or 5 by GC−MS.
Isolation of Arabidiol Oxidation Products. The crude ethanol
extract from combined 8 × 100 mL cultures of JBY575[pMDK3.6,
pDCR9.3, pDCR4.1] was saponified, followed by GC−MS analysis.
Flash chromatography on a 2 g silica SPE cartridge with ether−
hexanes gradients gave fractions A−H (listed as (% ether in hexanes),
triterpenoid content): A (5% ether), squalene; B (15−40% ether),
PEN1 products, including 14-epithalianol; C (50% ether), yeast
metabolites, arabidiol (6); D (50% ether), 6, 14-apo-arabidiol epimers
(10a, 10b), an unidentified side-chain oxygenated malabaricatrienol,
ergosterol; E (100% ether), 7, 8, 9a, 9b, 10a, 10b, unknowns (see
Figure 5); F, G, and H (100% ether): continuing elution of side-chain
oxygenated tricyclic triterpenes (see Figure S9).
S
Details of subcloning and yeast transformations, alignment and
evaluation of coding sequences for ATR2, microarray data
showing coexpression of OSCs and P450s, possible mecha-
nisms for P450 cleavage of arabidiol, tables of NMR chemical
shifts and their quantum mechanical predictions, and figures of
NMR and mass spectra. This material is available free of charge
via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
Corresponding Author
matsuda@rice.edu
■
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank Kevin R. MacKenzie (University of Houston) for
helpful discussions about NMR and Bill Wilson (Rice
University) for generous technical and editorial contributions.
This work was funded in part by The Robert A. Welch
Foundation (C-1323). NMR instrumentation was supported by
the UT-Houston Structural Biology Center, the W. M. Keck
Foundation (University of Houston), and the John S. Dunn Sr.
Gulf Coast Consortium for Magnetic Resonance (Rice
University).
REFERENCES
■
The main polar fraction (E) contained 10a as the major arabidiol
metabolite with some contamination of 10b ascribed to partial
epimerization of 10a during saponification. To obtain a full profile of
arabidiol oxidation products, we repeated the experiment (10 × 100
mL) but did not saponify the crude ethanol extracts. GC−MS and
HSQC analysis showed no formation of 10b. GC−MS of five
independent control cultures for each of the saponified and
nonsaponified experiments showed no formation of 10a or other
P450 products. Similarly, HSQC spectra of the two combined sets of
control extracts each showed no 10a (Figure 2C). Multiple control
experiments were motivated by the 4-fold variability in GC−MS
triterpenoid levels observed among different P450 cultures.
(1) Pichersky, E.; Gang, D. R. Trends Plant Sci. 2000, 5, 439−445.
(2) Weng, J. K.; Philippe, R. N.; Noel, J. P. Science 2012, 336, 1667−
1670.
(3) Croteau, R.; Kutchan, T. M.; Lewis, N. G. In Biochemistry and
Molecular Biology of Plants; Buchanan, B., Gruissem, W., Jones, R.,
Eds.; American Society of Plant Physiologists: Rockville, MD, 2000; pp
1250−1318.
(4) (a) Bak, S.; Beisson, F.; Bishop, G.; Hamberger, B.; Hofer, R.;
̈
Paquette, S.; Werck-Reichhart, D. Cytochromes P450; The Arabidopsis
Book, American Society of Plant Biologists: Rockville, MD, 2011; doi:
10.1199/tab.0144. (b) Schuler, M. A.; Duan, H.; Bilgin, M.; Ali, S.
Phytochem. Rev. 2006, 5, 205−237. (c) Nelson, D. R. Hum. Genomics
2009, 4, 59−65.
Isolation of Marneral Oxidation Products. GC−MS analysis of
crude extracts from a trial 100 mL culture of JBY575[pXQ11.2,
pDCR8.3, pDCR4.1] showed mainly marnerol, marneral, and a
hydroxymarnerol. The Folch extract from a larger experiment (18 ×
100 mL) was purified on a 5 g silica SPE cartridge using ether/hexanes
gradients. Marnerol and other MRN1 products (eluting with 10−15%
ether) were well separated from traces (<1%) of putative side-chain
oxygenated marnerols (eluting with 30% ether) and from the
hydroxymarnerol (eluting with 100% ether). Preparative reversed-
phase HPLC provided the purified hydroxymarnerol, which was
identified by 2D NMR as the 23-hydroxy derivative 12b. In another
experiment (10 × 100 mL), the crude ethanol extract was purified on
an SPE column (CH2Cl2 elution, followed by gradients of ether/
hexanes). Preparative reversed-phase HPLC then gave purified
samples of 11a and 11b, 12a, and 12b. The minor and major 23-
hydroxy metabolites 12a and 12b were characterized by GC−MS
(Figure 4C) and 2D NMR (see the Supporting Information).
Metabolites 12a and 12b were undetectable in GC−MS and HSQC
(Figure 3A and B) of several control cultures.
(5) The number of Arabidopsis P450s excludes pseudogenes.
Estimates of characterized P450s are derived from literature reports
and the annotated list of Arabidopsis P450s at www-ibmp.u-strasbg.fr/
∼CYPedia/index.html.
(6) Leading references for some Arabidopsis P450s linked to specific
natural products. Homoterpenes: (a) Lee, S.; Badieyan, S.; Bevan, D.
R.; Herde, M.; Gatz, C.; Tholl, D. Proc. Natl. Acad. Sci. U.S.A. 2010,
107, 21205−21210. Camalexins: (b) Beets, C. A.; Huang, J. C.;
Madala, N. E.; Dubery, I. Planta 2012, 236, 261−272. Arabidopyr-
ones: (c) Weng, J. K.; Li, Y.; Mo, H.; Chapple, C. Science 2012, 337,
960−964. Glucosinolates: (d) Chen, S.; Glawischnig, E.; Jorgensen,
K.; Naur, P.; Jorgensen, B.; Olsen, C. E.; Hansen, C. H.; Rasmussen,
H.; Pickett, J. A.; Halkier, B. A. Plant J. 2003, 33, 923−937. Many
early steps of specialized metabolism (e.g., oxidosqualene formation)
are shared with biosynthetic pathways that are broadly distributed
among plants and constitutively expressed. Examples of known P450
oxidations in early, shared polypropanoid biosynthesis: (e) Pan, Y.;
Michael, T. P.; Hudson, M. E.; Kay, S. A.; Chory, J.; Schuler, M. A.
Plant Physiol. 2009, 150, 858−878.
1
NMR Signal Assignments. 13C and H NMR assignments for 3,
4, 5, 10a, 10b, 12a, and 12b are presented in Tables S2,S3. As
described in section 12 of the Supporting Information, triterpenoid
structures were elucidated from 2D NMR correlations, comparisons
with reported NMR chemical shifts for triterpenes,13,16,30−33 and
(7) (a) von Roepenack-Lahaye, E.; Degenkolb, T.; Zerjeski, M.;
Franz, M.; Roth, U.; Wessjohann, L.; Schmidt, J.; Scheel, D.; Clemens,
S. Plant Physiol. 2004, 134, 548−559. (b) D’Auria, J. C.; Gershenzon, J.
5892
dx.doi.org/10.1021/ja401535g | J. Am. Chem. Soc. 2013, 135, 5885−5894

Journal of the American Chemical Society
Article
Curr. Opin. Plant Biol. 2005, 8, 308−316. The number of secondary
metabolites has markedly increased since the 2005 census of 170, due
mainly to LC−MS/MS identification of phenylpropanoids, glucosino-
lates, and other polar substances. However, few new P450 metabolites
of terpenes and triterpenes have been discovered in Arabidopsis.
(8) Hamberger, B.; Bak, S. Philos. Trans. R. Soc., B 2013, 368,
20120426.
(9) Misawa, N. Curr. Opin. Biotechnol. 2011, 22, 627−633.
(10) Xu, R.; Fazio, G. C.; Matsuda, S. P. T. Phytochemistry 2004, 65,
261−291.
(34) Duan, H.; Schuler, M. A. Phytochem. Rev. 2006, 5, 507−523.
(35) Lang, G.; Mayhudin, N. A.; Mitova, M. I.; Sun, L.; van der Sar,
S.; Blunt, J. W.; Cole, A. L.; Ellis, G.; Laatsch, H.; Munro, M. H. J. Nat.
Prod. 2008, 71, 1595−1599.
(36) Heterologous expression in E. coli to study sesquiterpene
biosynthesis: Harada, H.; Shindo, K.; Iki, K.; Teraoka, A.; Okamoto, S.;
Yu, F.; Hattan, J.; Utsumi, R.; Misawa, N. Appl. Microbiol. Biotechnol.
2011, 90, 467−476.
(37) Sainsbury, F.; Saxena, P.; Geisler, K.; Osbourn, A.; Lomonossoff,
G. P. Methods Enzymol. 2012, 517, 185−202.
(11) Sheng, H.; Sun, H. Nat. Prod. Rep. 2011, 28, 543−593.
(12) Hill, R. A.; Connolly, J. D. Nat. Prod. Rep. 2012, 29, 780−818.
(13) Fazio, G. C.; Xu, R.; Matsuda, S. P. T. J. Am. Chem. Soc. 2004,
126, 5678−5679.
(38) Lewinsohn, E.; Gijzen, M. Plant Sci. 2009, 176, 161−169.
(39) (a) Osbourn, A.; Papadopoulou, K. K.; Qi, X.; Field, B.; Wegel,
E. Methods Enzymol. 2012, 517, 113−138. (b) Wada, M.; Takahashi,
H.; Altaf-Ul-Amin, M.; Nakamura, K.; Hirai, M. Y.; Ohta, D.; Kanaya,
S. Gene 2012, 503, 56−64.
́
(14) Xue, Z.; Duan, L.; Liu, D.; Guo, J.; Ge, S.; Dicks, J.; O’Maille, P.;
Osbourn, A.; Qi, X. New Phytol. 2012, 193, 1022−1038.
(40) Ehlting, J.; Provart, N. J.; Werck-Reichhart, D. Biochem. Soc.
Trans. 2006, 34, 1192−1198.
(15) Phillips, D. R.; Rasbery, J. M.; Bartel, B.; Matsuda, S. P. T. Curr.
Opin. Plant Biol. 2006, 9, 305−314.
(41) Corey, E. J.; Matsuda, S. P. T.; Bartel, B. Proc. Natl. Acad. Sci.
U.S.A. 1994, 91, 2211−2215.
(16) Morlacchi, P.; Wilson, W. K.; Xiong, Q.; Bhaduri, A.; Sttivend,
D.; Kolesnikova, M. D.; Matsuda, S. P. T. Org. Lett. 2009, 11, 2627−
2630.
(42) Review: (a) Robinette, S. L.; Bruschweiler, R.; Schroeder, F. C.;
̈
Edison, A. S. Acc. Chem. Res. 2012, 45, 288−297. HSQC of an
Arabidopsis extract: (b) Lewis, I. A.; Schommer, S. C.; Hodis, B.; Robb,
K. A.; Tonelli, M.; Westler, W. M.; Sussman, M. R.; Markley, J. L. Anal.
Chem. 2007, 79, 9385−9390. Quantifying triterpenoids in plant
extracts by HSQC: (c) Kontogianni, V. G.; Exarchou, V.; Troganis, A.;
Gerothanassis, I. P. Anal. Chim. Acta 2009, 635, 188−195.
Quantitative HSQC of mixtures: (d) Hu, K.; Westler, W. M.;
Markley, J. L. J. Am. Chem. Soc. 2011, 133, 1662−1665. Covariance
2D NMR in mixture analysis: (e) Zhang, F.; Bruschweiler-Li, L.;
Bruschweiler, R. J. Am. Chem. Soc. 2010, 132, 16922−16927.
Differential 2D NMR mixture analysis to establish biological function:
(f) Pungaliya, C.; Srinivasan, J.; Fox, B. W.; Malik, R. U.; Ludewig, A.
H.; Sternberg, P. W.; Schroeder, F. C. Proc. Natl. Acad. Sci. U.S.A.
2009, 106, 7708−7713.
(43) Cankar, K.; van Houwelingen, A.; Bosch, D.; Sonke, T.;
Bouwmeester, H.; Beekwilder, J. FEBS Lett. 2011, 585, 178−182.
(44) 2D NMR of this mixture of 7, 8, 9a, 9b, 10a, and 10b resolved
all of the 1H and 13C NMR signals of 10a and 10b and gave
correlations sufficient for their signal assignments and definitive
structure determination (Tables S2−S5).
(45) Corey, E. J.; Cheng, H. Tetrahedron Lett. 1996, 37, 2709−2712.
(46) Epimer 10a was slightly favored in the Boltzmann distribution
based on B3PW91/6-311G(2d,p)//B3LYP/6-31G* relative energies:
10a, C13−C14 rotamers, 0.09 and 0.96 kcal/mol; 10b, C13−C14
rotamers, 0.00 and 4.08 kcal/mol; predicted ratio of epimers/
rotamers: 41.7, 9.6; 48.6, 0.0.
(47) Go, Y. S.; Lee, S. B.; Kim, H. J.; Kim, J.; Park, H. Y.; Kim, J. K.;
Shibata, K.; Yokota, T.; Ohyama, K.; Muranaka, T.; Arseniyadis, S.;
Suh, M. C. Plant J. 2012, 72, 791−804.
(17) (a) Lytovchenko, A.; Beleggia, R.; Schauer, N.; Isaacson, T.;
Leuendorf, J. E.; Hellmann, H.; Rose, J. K.; Fernie, A. R. Plant Methods
2009, 5, 4. (b) Quanbeck, S. M.; Brachova, L.; Campbell, A. A.; Guan,
X.; Perera, A.; He, K.; Rhee, S. Y.; Bais, P.; Dickerson, J. A.; Dixon, P.;
Wohlgemuth, G.; Fiehn, O.; Barkan, L.; Lange, I.; Lange, B. M.; Lee, I.;
Cortes, D.; Salazar, C.; Shuman, J.; Shulaev, V.; Huhman, D. V.;
Sumner, L. W.; Roth, M. R.; Welti, R.; Ilarslan, H.; Wurtele, E. S.;
Nikolau, B. J. Front. Plant Sci. 2012, 3, 15.
(18) Shan, H.; Wilson, W. K.; Phillips, D. R.; Bartel, B.; Matsuda, S.
P. T. Org. Lett. 2008, 10, 1897−1900.
(19) Suzuki, H.; Achnine, L.; Xu, R.; Matsuda, S. P. T.; Dixon, R. A.
Plant J. 2002, 32, 1033−1048.
(20) Haralampidis, K.; Trojanowska, M.; Osbourn, A. E. Adv.
Biochem. Eng. Biotechnol. 2002, 75, 31−49.
(21) Seki, H.; Sawai, S.; Ohyama, K.; Mizutani, M.; Ohnishi, T.;
Sudo, H.; Fukushima, E. O.; Akashi, T.; Aoki, T.; Saito, K.; Muranaka,
T. Plant Cell 2011, 23, 4112−4123.
(22) (a) Shibuya, M.; Hoshino, M.; Katsube, Y.; Hayashi, H.;
Kushiro, T.; Ebizuka, Y. FEBS J. 2006, 273, 948−959. (b) Fukushima,
E. O.; Seki, H.; Ohyama, K.; Ono, E.; Umemoto, N.; Mizutani, M.;
Saito, K.; Muranaka, T. Plant Cell Physiol. 2011, 52, 2050−2061.
(c) Han, J. Y.; Hwang, H. S.; Choi, S. W.; Kim, H. J.; Choi, Y. E. Plant
Cell Physiol. 2012, 53, 1535−1545. (d) Huang, L.; Li, J.; Ye, H.; Li, C.;
Wang, H.; Liu, B.; Zhang, Y. Planta 2012, 236, 1571−1581.
(23) Field, B.; Osbourn, A. E. Science 2008, 320, 543−547.
(24) Field, B.; Fiston-Lavier, A. S.; Kemen, A.; Geisler, K.;
Quesneville, H.; Osbourn, A. E. Proc. Natl. Acad. Sci. U.S.A. 2011,
108, 16116−16121.
(25) Qi, X.; Bakht, S.; Leggett, M.; Maxwell, C.; Melton, R.;
Osbourn, A. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 8233−8238.
(26) Recent advances in LC−MS have overcome some of the
dependence on authentic standards: Wu, H.; Guo, J.; Chen, S.; Liu, X.;
Zhou, Y.; Zhang, X.; Xu, X. J. Pharm. Biomed. Anal. 2013, 72, 267−
291.
(48) Mizutani, M. Biol. Pharm. Bull. 2012, 35, 824−832.
(49) Pichersky, E.; Lewinsohn, E. Annu. Rev. Plant Biol. 2011, 62,
549−566.
(50) Guengerich, F. P. Chem. Res. Toxicol. 2001, 14, 611−650.
(51) Strushkevich, N.; MacKenzie, F.; Cherkesova, T.; Grabovec, I.;
Usanov, S.; Park, H. W. Proc. Natl. Acad. Sci. U.S.A. 2011, 108, 10139−
10143.
(27) Domingo, V.; Arteaga, J. F.; Quilez del Moral, J. F.; Barrero, A.
F. Nat. Prod. Rep. 2009, 26, 115−134.
(52) Tholl, D.; Sohrabi, R.; Huh, J. H.; Lee, S. Phytochemistry 2011,
(28) Ritzdorf, I.; Bartels, M.; Kerp, B.; Kasel, T.; Klonowski, S.;
Marner, F.-J. Phytochemistry 1999, 50, 995−1003.
72, 1635−1646.
(53) Weng, J. K.; Noel, J. P. Cold Spring Harbor Symp. Quant. Biol.
2012, DOI: 10.1101/sqb.2012.77.014787.
(29) Marner, F. J. Curr. Org. Chem. 1997, 1, 153−186.
(30) Xiong, Q.; Wilson, W. K.; Matsuda, S. P. T. Angew. Chem., Int.
Ed. 2006, 45, 1285−1288.
(54) Wassif, C. A.; Brownson, K. E.; Sterner, A. L.; Forlino, A.;
Zerfas, P. M.; Wilson, W. K.; Starost, M. F.; Porter, F. D. Hum. Mol.
Genet. 2007, 16, 1176−1187.
(55) Withers, J. C.; Shipp, M. J.; Rupasinghe, S. G.; Sukumar, P.;
Schuler, M. A.; Muday, G. K.; Wyatt, S. E. Am. J. Bot. 2013, 100, 183−
193.
(31) Lodeiro, S.; Xiong, Q.; Wilson, W. K.; Kolesnikova, M. D.;
Onak, C. S.; Matsuda, S. P. T. J. Am. Chem. Soc. 2007, 129, 11213−
11222.
(32) Kolesnikova, M. D.; Obermeyer, A. C.; Wilson, W. K.; Lynch, D.
A.; Xiong, Q.; Matsuda, S. P. T. Org. Lett. 2007, 9, 2183−2186.
(33) Lodeiro, S.; Xiong, Q.; Wilson, W. K.; Ivanova, Y.; Smith, M. L.;
May, G. S.; Matsuda, S. P. T. Org. Lett. 2009, 11, 1241−1244.
(56) Zust, T.; Heichinger, C.; Grossniklaus, U.; Harrington, R.;
Kliebenstein, D. J.; Turnbull, L. A. Science 2012, 338, 116−119.
5893
dx.doi.org/10.1021/ja401535g | J. Am. Chem. Soc. 2013, 135, 5885−5894

Journal of the American Chemical Society
Article
(57) Counterexamples include combining chemical genomics and
metabolomics to elucidate unknown pathways. For leading references,
see: Bocobza, S.; Willmitzer, L.; Raikhel, N. V.; Aharoni, A. Plant
Physiol. 2012, 160, 1160−1163.
(58) Kliebenstein, D. J. Annu. Rev. Phytopathol. 2012, 50, 155−173.
(59) Mintz-Oron, S.; Meir, S.; Malitsky, S.; Ruppin, E.; Aharoni, A.;
Shlomi, T. Proc. Natl. Acad. Sci. U.S.A. 2012, 109, 339−344.
(60) HSQC establishes stereochemistry and carbon connectivity,
while MS supplies information about heteroatoms. Although their
relative strengths differ between hydrogen-poor polypropanoids and
stereochemically rich terpenoids, neither method is effective alone, as
illustrated in a phenylpropanoid study: Xu, Y.-J.; Foubert, K.;
Dhooghe, L.; Lemiere, F.; Maregesi, S.; Coleman, C. M.; Zou, Y.;
Ferreira, D.; Apers, S.; Pieters, L. Phytochemistry 2012, 79, 121−128.
5894
dx.doi.org/10.1021/ja401535g | J. Am. Chem. Soc. 2013, 135, 5885−5894