Articles

https://doi.org/10.1038/s41477-018-0220-z

Contribution of isopentenyl phosphate to plant

terpenoid metabolism
Laura K. Henry1,8, Suzanne T. Thomas2,8, Joshua R. Widhalm   3,4, Joseph H. Lynch1, Thomas C. Davis   5,
Sharon A. Kessler4,5, Jörg Bohlmann6, Joseph P. Noel   2,7* and Natalia Dudareva   1,3,4*

Plant genomes encode isopentenyl phosphate kinases (IPKs) that reactivate isopentenyl phosphate (IP) via ATP-dependent
phosphorylation, forming the primary metabolite isopentenyl diphosphate (IPP) used generally for isoprenoid/terpenoid bio-
synthesis. Therefore, the existence of IPKs in plants raises unanswered questions concerning the origin and regulatory roles
of IP in plant terpenoid metabolism. Here, we provide genetic and biochemical evidence showing that IP forms during specific
dephosphorylation of IPP catalysed by a subset of Nudix superfamily hydrolases. Increasing metabolically available IP by over-
expression of a bacterial phosphomevalonate decarboxylase (MPD) in Nicotiana tabacum resulted in significant enhancement
in both monoterpene and sesquiterpene production. These results indicate that perturbing IP metabolism results in measur-
able changes in terpene products derived from both the methylerythritol phosphate (MEP) and mevalonate (MVA) pathways.
Moreover, the unpredicted peroxisomal localization of bacterial MPD led us to discover that the step catalysed by phospho-
mevalonate kinase (PMK) imposes a hidden constraint on flux through the classical MVA pathway. These complementary
findings fundamentally alter conventional views of metabolic regulation of terpenoid metabolism in plants and provide new
metabolic engineering targets for the production of high-value terpenes in plants.

T
erpenoids, also referred to as isoprenoids, constitute one of
the largest and most chemically diverse classes of primary and
secondary metabolites in nature. These compounds serve a
broad range of physiological functions, including key roles in respi-
ration, photosynthesis, growth, development, reproduction, defence
and environmental sensing1,2. Terpenoids are also highly valued by
humans as fragrances, flavours, biofuels, nutritional supplements,
insecticides and pharmaceuticals3–5.
Despite their structural diversity, all terpenoids begin with two
universal five-carbon isoprene-like building blocks, isopentenyl
diphosphate (IPP) and dimethylallyl diphosphate (DMAPP). In
plants, both IPP and DMAPP are derived from two compartmen-
tally separated metabolically crosstalking routes, the mevalonic
acid (MVA) and methylerythritol phosphate (MEP) pathways
(Fig. 1a)6. While the MEP pathway is exclusively localized in plas-
tids, the MVA pathway distributes between cytoplasm, endoplasmic
reticulum and peroxisomes (Fig. 1a)6.
The MVA pathway generates IPP and DMAPP, which are
elongated by the ubiquitous enzymes farnesyl diphosphate syn-
thases (FPPSs). Generally, FPPSs catalyse the condensation of one
DMAPP molecule with two IPP molecules to produce farnesyl
diphosphate (FPP) and two molecules of pyrophosphate. FPP is an
essential metabolite used for sesquiterpene, homoterpene, triter-
pene, sterol, brassinosteroid and polyprenol biosynthesis2. It is gen-
erally considered that 3-hydroxy-3-methyl-glutaryl-CoA reductase
(HMGR) catalyses the rate-limiting step of the MVA pathway1,2,6,7;
however, HMGR-overexpression-based metabolic engineering
strategies were not sufficient to overcome the limitations of using
the cytosolic MVA pathway for high-yield terpenoid production in
plants8–12. Therefore, these results suggest that the MVA pathway is
regulated by additional as of yet unknown mechanisms governing
flow through the pathway and subsequently metabolite yield.
We recently discovered that, in addition to the classical MVA
and MEP pathway enzymes, plant genomes encode another IPP-
generating protein, isopentenyl phosphate kinase (IPK)13,14. In
plants, IPK localizes to the cytoplasm, where it transforms iso-
pentenyl phosphate (IP) and possibly dimethylallyl phosphate
(DMAP) to IPP and DMAPP via ATP-dependent phosphoryla-
tion. Paradoxically, IPK appears to augment terpenoid produc-
tion through both the MVA and MEP pathways14. While in plants
IPK appears to be involved in a metabolite reactivation process14,
in some bacteria and archaea, IPK catalyses an essential and final
step in the alternative MVA pathway (Fig. 1b)13,15. The alternative
MVA pathway bifurcates from the classical metabolic route follow-
ing mevalonate kinase (MK)-mediated phosphorylation of meval-
onate yielding phosphomevalonate (MVAP). In the classical MVA
pathway, MVAP undergoes phosphorylation catalysed by phos-
phomevalonate kinase (PMK) to produce mevalonate diphosphate
(MVAPP), which is subsequently subjected to decarboxylation cata-
lysed by mevalonate 5-diphosphate decarboxylase (MDD) (Fig. 1a).
In contrast, in the alternative MVA pathway, the order of reactions
is reversed wherein MVAP undergoes initial decarboxylation to IP
catalysed by a phosphomevalonate decarboxylase (MPD) followed
by ATP-dependent phosphorylation of IP catalysed by IPK (Fig. 1b).
The presence of genes encoding IPK in all sequenced plant
genomes indicates that modulating the ratios of IP to IPP and
DMAP to DMAPP may serve an unknown role in regulating ter-
penoid biosynthesis14. Moreover, yield enhancement of MVA and
MEP pathway-derived terpenoids in tobacco leaves following over-
expression of Arabidopsis thaliana IPK (AtIPK) further supports

1
Department of Biochemistry, Purdue University, West Lafayette, IN, USA. 2Jack H Skirball Center for Chemical Biology and Proteomics, Salk Institute for
Biological Studies, La Jolla, CA, USA. 3Department of Horticulture and Landscape Architecture, Purdue University, West Lafayette, IN, USA. 4Purdue Center
for Plant Biology, Purdue University, West Lafayette, IN, USA. 5Department of Botany and Plant Pathology, Purdue University, West Lafayette, IN, USA.
6
Michael Smith Laboratories, University of British Columbia, Vancouver, Canada. 7Howard Hughes Medical Institute, Salk Institute for Biological Studies,
La Jolla, CA, USA. 8These authors contributed equally to this work: Laura K. Henry, Suzanne T. Thomas. *e-mail: noel@salk.edu; dudareva@purdue.edu

Nature Plants | www.nature.com/natureplants

Articles NATurE PlAnTs

a Acetyl-CoA ER However, determining the physiological function of Nudix enzymes

AACT is complicated by their measurable in vitro substrate promiscu-
MVA pathway HMGS ity17,18. To identify putative IPP/DMAPP phosphatase candidates
HMGR
Plastid that may produce IP/DMAP in planta, we first expressed all Nudix
MK Pyruvate and GA-3P enzymes encoded by the Arabidopsis genome19 in Escherichia coli.
Cytoplasm
MVAP DXS Of the 27 heterologously expressed proteins, 16 produced soluble
DXR
enzymes that were assayed for small-molecule phosphatase activ-
MVAP
MEP pathway ity. The remaining genes, AtNudx 2, 4, 8, 10, 13, 16-19, 21 and 22
Peroxisome PMK
MCT
produced insoluble inclusion bodies and were not assayed. Most
MVAPP CMK of these Nudix enzymes were previously assayed against a panel
MDD
MDS
of phosphate-bearing substrates; however, isoprenoid diphos-
DMAP DMAPP IPP IP ? IP phates were notably absent19,20. We screened the soluble enzymes
? IDI
?
HDS
for catalytic activity with IPP as a substrate employing a modified
IPK IPK
HDR HDR malachite green assay for free phosphate detection. Only AtNudx1
DMAPP
IDI
IPP ? IPP
IDI
DMAPP
(At1g68760) and AtNudx3 efficiently catalysed dephosphoryla-
tion of IPP to IP, while the remaining enzymes exhibited low to no
FPPS GPPS activity (Supplementary Fig. 1). After optimizing for pH and mag-
FPP GPP nesium cation (Mg2+) dependence (Supplementary Fig. 2), steady-
state kinetic constants were determined for AtNudx1 and AtNudx3
Sesquiterpenes Sterols using isoprenoid mono- and diphosphate-containing compounds.
Monoterpenes
Both AtNudx1 and AtNudx3 catalysed dephosphorylation of
MVAPP, IPP, DMAPP, geranyl diphosphate (GPP) and FPP to the
b monophosphate products, MVAP, IP, DMAP, geranyl phosphate
Alternative MVA pathway
MPD IPK (GP) and farnesyl phosphate (FP), respectively, and did not catalyse
Acetyl-CoA MVAP IP IPP further dephosphorylation to their respective isoprenoid alcohols.
AtNudx1 utilized IPP, DMAPP, GPP and FPP with equivalent cata-
lytic efficiencies (kcat/KM) where KM is the Michaelis constant and
Fig. 1 | Terpenoid biosynthetic pathways in plants, and in some archaea kcat is the turnover number, while MVAPP was 100-fold less effi-
and bacteria. a, The MVA (left) and MEP (right) pathways responsible cient as a substrate (Table 1). In contrast, AtNudx3 preferred IPP
for terpenoid biosynthesis in plants. b, An alternative MVA pathway and DMAPP with catalytic efficiencies similar to AtNudx1 and with
discovered in some archaea and the Chloroflexi phylum of bacteria. This threefold higher catalytic efficiencies compared to AtNudx1 using
pathway follows the same steps as the plant classical MVA pathway GPP and FPP as substrates (Table 1).
until MVAP formation. MVAP is then converted by MPD to IP, which
is subsequently phosphorylated by IPK to IPP. AACT, acetoacetyl-CoA AtNudx1 activity against 8-oxo-dGTP is likely to be irrelevant
thiolase; CMK, 4-(cytidine 5ʹ​-diphospho)-2-C-methyl-d-erythritol kinase; in vivo. It was previously reported that AtNudx1 functions as an
DMAP, dimethylallyl phosphate; DMAPP, dimethylallyl diphosphate; 8-oxo-dGTPase on the basis of homology with bacterial and mam-
DXR, 1-deoxy-d-xylulose 5-phosphate reductoisomerase; DXS, 1-deoxy- malian Nudix superfamily members21. While AtNudx3 was unable
d-xylulose 5-phosphate synthase; FPP, farnesyl diphosphate; FPPS, to dephosphorylate 8-oxo-dGTP, AtNudx1 catalysed dephosphor-
farnesyl diphosphate synthase; GA-3P, d-glyceraldehyde 3-phosphate; ylation with catalytic efficiencies similar to those obtained with
GPP, geranyl diphosphate; GPPS, geranyl diphosphate synthase; HDR, IPP and DMAPP as substrates (Table 1). However, AtNudix1 only
(E)-4-hydroxy-3-methylbut-2-enyl diphosphate reductase; HDS, (E)-4- weakly prefers 8-oxo-dGTP compared to dGTP with (kcat/KM)8-oxo-
hydroxy-3-methylbut-2-enyl diphosphate synthase; HMGR, 3-hydroxy-
dGTP
/(kcat/KM)dGTP =​  2.6 (ref.  20) while intracellular concentrations of
3-methylglutaryl-CoA reductase; HMGS, 3-hydroxy-3-methylglutaryl- dGTP are expected to be significantly higher than 8-oxo-dGTP
CoA synthase; IDI, isopentenyl diphosphate isomerase; IP, isopentenyl (Supplementary Table 2).
phosphate; IPK, isopentyl phosphate kinase; IPP, isopentenyl diphosphate; To precisely define the structural basis for substrate selectivi-
MCT, 2-C-methyl-d-erythritol 4-phosphate cytidylyltransferase; ties of AtNudx1, we next obtained diffraction-quality crystals for
MDD, mevalonate diphosphate decarboxylase; MDS, 2-C-methyl- atomic-resolution protein X-ray crystallographic analyses. AtNudx1
d-erythritol 2,4-cyclodiphosphate synthase; MK, mevalonate kinase; MPD, crystallized without ligands (2.0 Å, Rwork =​  0.1823 and Rfree =​  0.2165)
phosphomevalonate decarboxylase; MVAP, mevalonate 5-phosphate; and with IPP bound (1.90 Å, Rwork =​  0.1827 and Rfree =​  0.2277)
MVAPP, mevalonate diphosphate; PMK, phosphomevalonate kinase. (Supplementary Table 1). IPP (pdb ligand ID IPR) and three mag-
nesium cations (Mg2+) were easily modelled in active-site electron-
density maps (Fig. 2a). The presence of IPP instead of the reaction
the contribution of IP formation to regulating the plant terpenoid product IP is clear in the active site of AtNudx1 and probably due
network14. However, the origin of IP and possibly DMAP in plants to the low pH values (pH 5.0) and low temperatures used during
is unresolved, leaving open questions about how the metabolism crystallization and data collection, 4 °C and −​273  °C, respectively.
of the universal five-carbon terpenoid building blocks, IPP and At pH 6.5, the specific activity of AtNudx1 using IPP as a substrate
DMAPP, is regulated in the plant kingdom. is sevenfold lower than activities measured at the optimal pH of 8.5
(at 37 °C) (Supplementary Fig. 2a).
Results and discussion Like most Nudix superfamily members, AtNudx1 employs
Origin of IP in plants. The absence of genes encoding MPDs in divalent cations for catalytic activity. Mg2+ is present in the crystal-
plant genomes indicates that, in plants, IP and possibly DMAP lization conditions used in this study. From the crystal structures,
may arise via a different route from the alternative MVA pathway Mg2+ octahedrally coordinates several water molecules, the side
found in certain bacteria and archaea (Fig. 1b). Recently, a unique chains of Glu 56 and Glu 60, the carbonyl oxygen of Gly 40 and the
two-domain (hydrolase/peptidase) member of the Nudix hydrolase diphosphate oxygens of IPP (Fig. 2b). The multivalent diphosphate
superfamily, AtNudx3 (At1g79690), was shown to dephosphory- group is also hydrogen bonded by His 42 and Arg 27. The C5 car-
late IPP and hydrolyse chromogenic dipeptide substrates in vitro16. bon chain of IPP is sequestered by the hydrophobic side chains of

Nature Plants | www.nature.com/natureplants

NATurE PlAnTs Articles
Table 1 | Kinetic parameters for recombinant A. thaliana Nudx1 and Nudx3
AtNudx1 AtNudx3
Substrate KM (μM) kcat (s )
−1
kcat/KM (M s )
−1 −1
KM (μM) kcat (s−1) kcat/KM (M−1 s−1)
(R)-MVAPP 240 ±​ 60 0.51 ±​ 0.06 2.1 ×​ 103 ±​ 0.6 125 ±​ 36 1.5 ±​ 0.2 1.2 ×​ 104 ±​ 0.4
DMAPP 10.7 ±​ 1.8 1.7 ±​ 0.07 1.6 ×​ 10  ±​ 0.3
5
44 ±​ 8 7.1 ±​ 0.3 1.6 ×​ 105 ±​ 0.3
IPP 8.0 ±​ 0.6 1.11 ±​ 0.02 1.4 ×​ 10  ±​ 0.1
5
36 ±​ 8 5.5 ±​ 0.3 1.5 ×​ 105 ±​ 0.3
GPP 8.3 ±​ 2.5 2.0 ±​ 0.3 2.5 ×​ 105 ±​ 0.8 53 ±​ 13 2.7 ±​ 0.3 5.1 ×​ 104 ±​ 0.5
FPP 8.7 ±​ 1.6 1.35 ±​ 0.06 1.6 ×​ 105 ±​ 0.3 58 ±​ 10 3.3 ±​ 0.3 5.7 ×​ 104 ±​ 0.5
8-oxo-dGTP 33 ±​ 2 4.5 ±​ 0.1 1.4 ×​ 10  ±​ 0.1
5
NA NA NA
Data are means ±​ s.d. (n =​ 3 independent experiments). NA, no detectable activity. No phosphatase activity was detected for AtNudx1 and AtNudx3 with IP, DMAP, MVAP, GP and FP.

a b c

R27
R27
K110 K110 GPP

E60
IPP
Y87 E60 Y87
G40 G40

E56 E56A
H42 H42
E59 E59

R55 R55

Fig. 2 | Substrate recognition by AtNudx1. a, IPP- and Mg2+-bound AtNudx1 showing unbiased Fo −​Fc electron density omit map at contour level 2σ​ for
IPP, active site Mg2+ (green), and coordinating waters (red), where Fo is the experimentally measured amplitude and Fc is the model-based amplitude.
b, Mg2+ and phosphate coordination scheme in wild-type AtNudx1–IPP. Side chains of key residues are shown. c, The GPP-bound E56A mutant of AtNudx1
(pdb 5GP0)22. Side chains of residues shown in b are shown here for comparison. Carbon, yellow; nitrogen, blue; oxygen, red; phosphorus, orange.

Ala 11, Val 12, Val 13, Ile 31, Ala 37, Leu 38, Phe 78, Phe 127, Pro 129,
Leu 130 and Leu 133 (Supplementary Fig. 3a).
While preparing this manuscript, apo and GPP-bound E56A
mutant structures of AtNudx1 were reported22. The apo structures
(pdb 5WWD) and IPP- and GPP-bound structures (pdb 5GP0)
superimposed with root mean squared deviations of 0.2969 Å and
0.3246 Å, respectively. Surprisingly, superposition of the previ-
ously reported catalytically impaired AtNudx1 mutant, E56A,
with GPP bound (pdb 5GP0) and our structure with IPP bound
show that the shared chemical features of these two ligands do
not superimpose (Fig. 2b,c). The E56A mutation probably pre-
vents Mg2+ and GPP from binding in productive conformations.
In our AtNudx1–IPP experimental structure, Glu 56 bicoordinates
with two of the active site Mg2+ ions. Mg2+ ions are absent in apo-
AtNudx1, due to the absence of the diphosphate group that prob-
ably initiates cation recognition and early stages of divalent cation,
active-site coordination.
We structurally aligned the AtNudx1–IPP complex with 8-oxo-
dGMP bound E. coli 8-oxo-dGTPase MutT (pdb 3A6U) and
human 8-oxo-GTPase MTH1 (pdb 3ZR0) to compare the chemi-
cal features governing substrate binding in these Nudix enzymes23,24
(Supplementary Fig. 3). Noticeably, residues important for bind-
ing the nucleotide substrate, Glu 34, His 28 and Asn 119 in MutT
(Supplementary Fig. 3b) and Asn 33, Asp 119, Asp 120 and Trp 117
in MTH1 (Supplementary Fig. 3c), are absent in AtNudx1. Despite
sharing the same fold, AtNudx1 and 8-oxo-dGTPases possess dis-
tinct active-site pockets for substrate recognition and catalysis.
These results combined with in vitro substrate specificity stud-
ies suggest that AtNudx1 activity with 8-oxo-dGTP is likely to be
inconsequential in vivo.
AtNudx1 and AtNudx3 contribute to IP, and possibly DMAP,
formation in planta. To investigate the in planta contribution of
AtNudx1 and AtNudx3 to isoprenoid production, we first analysed
their expression across different tissues using quantitative RT-PCR
(qRT-PCR) with gene-specific primers. Both Nudix genes were
found to be expressed in all tissues with AtNudx3 messenger RNA at
significantly higher levels than those of AtNudx1 (Fig. 3a). In addi-
tion, expression of a β​-glucuronidase (GUS) reporter under the con-
trol of AtNudix1 and AtNudix3 promoters further indicated that their
expression overlaps across different tissues (Supplementary Fig. 4).
We next used a reverse genetics approach and profiled terpe-
noids in Arabidopsis transfer DNA (T-DNA) insertion lines (nudx1-
1, nudx1-2, nudx3-1 and nudx3-2) (Supplementary Fig. 5) to
further examine the role of AtNudx1 and AtNudx3. No AtNudx1 or
AtNudx3 transcripts were detected in mutants with the exception of
the nudx1-2 mutant, which exhibited a 90% reduction in AtNudx1
expression (Fig. 3b). Emission of the most abundant sesquiterpene,
β​-carophyllene, from Arabidopsis flowers increased by 28–60%
(Fig. 3c). The concentration of the sterol sitosterol nearly doubled
in all nudx mutants, while the levels of campesterol and stigmas-
terol remained unchanged (Fig. 3d). Emission of the monoterpene
linalool from flowers was increased 148–503% in all nudx T-DNA
mutants (Fig. 3e), suggesting that AtNudx1 and AtNudx3 modulate
the ratios of IPP to IP and possibly DMAPP to DMAP in vivo.
The similar terpenoid metabolic profiles in the nudx1 and nudx3
mutants suggest that despite their differential expression levels
(Fig. 3a), both AtNudx1 and AtNudx3 regulate the availability of
metabolites contributing to both GPP- and FPP-derived terpenoids.
As AtNudx1 and AtNudx3 are localized in the cytoplasm20,21, the
observed effects on sterol levels and sesquiterpene emission may

Nature Plants | www.nature.com/natureplants

Articles NATurE PlAnTs

a 0.15
b
120 Nudx1 mRNA 120 Nudx3 mRNA

(picograms per 200 ng total RNA)

AtNudx1
0.10

levels (% of control)
AtNudx3
0.03 80 80

mRNA
mRNA
0.02
40 40
0.01

nd nd nd
0.00 0 0
Root Cauline Stem Rosette Silique Flower

-0

1
-0

2
2

3-
1-

3-
1-

ol
ol

dx
dx

dx
dx

C
C

nu
nu

nu
nu

P = 0.0046
P = 0.0031
c d

P = 0.0054
P = 0.0471
50 2.5 Col-0 **
P = 0.0013 nudx1-1 ** **
P = 0.0245 nudx1-2 *
β-Caryophyllene emission

40 ** P = 0.0029 2.0
P = 0.0181 * nudx3-1

Sterols (mg gFW–1)

h )
–1

* ** nudx3-2
30 1.5
–1
(pmol gFW

20 1.0

10 0.5

0 0.0
Col-0 nudx1-1 nudx1-2 nudx3-1 nudx3-2 Campesterol Stigmasterol Sitosterol

e 0.6
f 40
P = 0.0056 EV AtNudx1 AtNudx3
**
Emission (pmol gFW–1 h–1)

P = 0.0373

P = 0.0446
P < 0.0001 30
(pmol gFW–1 h–1)
Linalool emission

P = 0.0050
0.4 P = 0.0224 ****

P = 0.0082
*
P = 0.0064

P = 0.0024

P = 0.0054
20 *
P = 0.0310
*
0.2 * ** **
10
** **
**
0.0 0
Col-0 nudx1-1 nudx1-2 nudx3-1 nudx3-2 β-Ocimene Linalool β-Caryo- 5-Epi-
phyllene aristolochene

Fig. 3 | Role of Nudx1 and Nudx3 in vivo. a, Expression of AtNudx1 and AtNudx3 in different Arabidopsis tissues. b, AtNudx1 and AtNudx3 transcript levels in
Col-0 and T-DNA insertion mutants determined by qRT-PCR (means ±​ s.e.m., n =​ 3 biologically independent samples). c–e, Effect of AtNudx1 and AtNudx3
knockouts on sterol and terpenoid formation. β​-Caryophyllene (c) and linalool (e) emission from flowers of 5-week-old Arabidopsis inflorescences. Floral
volatiles were collected from approximately 50 inflorescences and analysed by gas chromatography mass spectrometry. d, Sterol levels in 8-day-old
Arabidopsis seedlings of Col-0 and nudx1 and nudx3 mutants. f, Transient overexpression of AtNudx1 and AtNudx3 under control of a CaMV-35S promoter
in tobacco leaves. All data are means ±​ s.e.m., n =​ 3 biologically independent samples, except n =​ 6 biologically independent samples for Col-0 in c and e.
*P <​ 0.05; **P <​ 0.01; ****P <​ 0.0001 (two-tailed Student’s t-test); FW, fresh weight; nd, not detected.

result from dephosphorylation of IPP and/or FPP (Table 1 and
Fig. 1a). In contrast, because monoterpene formation partially
relies on IPP imported into plastids from the cytoplasm (Fig. 1a),
the observed effects on monoterpene levels can result only from IPP
dephosphorylation.
Next, each nudx gene was transiently overexpressed in N. taba-
cum leaves (Supplementary Fig. 6), which emit both monoterpene
and sesquiterpene compounds. Twenty-four hours after infiltrating
Agrobacteria carrying either AtNudx1 or AtNudx3 expression con-
structs into tobacco, the emission of sesquiterpenes was decreased
by 57–88% compared to leaves infiltrated with Agrobacteria har-
bouring an empty vector (Fig. 3f). Emission of the monoterpenes
linalool and β​-ocimene decreased on average by 50% in tobacco
leaves overexpressing AtNudx1 and was lower, albeit not signifi-
cantly for β​-ocimene, in tobacco leaves overexpressing AtNudx3
relative to the control (Fig. 3f). Thus, overexpression of AtNudx1
and AtNudx3 resulted in an opposite metabolic phenotype to
that observed when AtIPK was overexpressed in tobacco leaves14.
Together with the Arabidopsis nudx1 and nudx3 mutants pro-
filed here (Fig. 3), and the ipk mutants analysed previously14, the
observed complementary phenotypes provide in vivo evidence that
Nudix and IPK catalyse opposing reactions to regulate IPP/IP and
possibly DMAPP/DMAP ratios (Fig. 1). Thus, in plant cells, IP and
DMAP formation is not the consequence of dephosphorylation by
nonspecific phosphatases but is instead the result of the catalytic
activity of specific Nudix enzymes. Moreover, because AtNudx1
and AtNudx3 dephosphorylate FPP (Table 1), we cannot exclude
the possibility that these enzymes also function to modulate the
FPP to FP ratio. The activities with prenyl diphosphate substrates
presented here are consistent with the recent report of a Rosa ×
hybrida Nudix enzyme, RhNUDX1, a homologue of AtNudx1 and
AtNudx3, that in rose petals converts GPP to GP for terpene alco-
hol formation in the cytoplasm25. Although AtNudx1 and AtNudx3
can also dephosphorylate GPP (Table 1), it is unlikely that they act

Nature Plants | www.nature.com/natureplants

NATurE PlAnTs Articles
on this substrate in planta as GPP is considered to be restricted to a 0.8
plastids in most plants.

RcMPD mRNA levels

0.6
Effects of increasing IP formation on terpenoid biosynthesis.
Our data indicate that IP/DMAP is formed from dedicated dephos-
0.4
phorylation of IPP/DMAPP catalysed by Nudix superfamily hydro-
lases (Fig. 3). Considering the limitations of the MVA pathway for
high-yield terpenoid production, we tested whether introducing de 0.2
novo IP production in plants increases flux toward downstream ter-
penoids. As plants lack genes encoding MPDs, we used a bacterial nd nd
0.0
MPD gene from Roseiflexus castenholzii encoding an enzyme previ- WT EV MPD-1 MPD-4 MPD-11
ously shown to possess strict specificity for the efficient decarboxyl-
ation of MVAP to IP13. The RcMPD gene was stably overexpressed in b WT EV MPD-1 MPD-4 MPD-11

N. tabacum under the control of the CaMV 35S promoter to create

P = 0.0003

P = 0.0006
a bifurcation in the canonical MVA pathway in order to produce IP

P = 0.0029
P = 0.0002

P = 0.0014
without dephosphorylating IPP. We hypothesized that if the endog- 5
enous IPK is capable of converting MPD-generated IP to IPP, then ***

P = 0.0212
P = 0.0009
***

P = 0.0002
RcMPD overexpression may result in the introduction of a novel

P = 0.0012
4 **
second branch of the MVA pathway resembling the alternative MVA *** **

Sterols (mg gFW–1)

pathway known to exist in some bacteria and archaea (Fig. 1b)13,15.

P = 0.0002
3 *
Metabolic analyses of three independent transgenic lines with *** **
***

P = 0.0002
different expression levels of the RcMPD gene, namely MPD-1,

P = 0.0023
MPD-4 and MPD-11 (Fig. 4a), revealed that all lines exhibited a 2
***
substantial increase in overall terpenoid production relative to wild-
type and empty-vector control plants without affecting the expres- 1 ***
**
sion of endogenous IPK, PMK and MDD. Sterol levels, including
cholesterol, stigmasterol, sitosterol and campesterol, were respec-
0
tively 3.2-, 4.2-, 3.2- and 3.7-fold higher in RcMPD transgenic Cholesterol Stigmasterol Sitosterol Campesterol
plants relative to controls (Fig. 4b). More elaborated downstream
terpenoid–quinone conjugate products were not affected, probably c
P = 0.002

P = 0.0028
P = 0.0075
due to limited availability of the aromatic building blocks necessary

P = 0.0028
P < 0.0001

for their biosynthesis (Supplementary Fig. 7). Finally, the leaves of 150 P = 0.0007
**
MPD transgenic plants emitted up to 4.1- and 7.4-fold more mono- **
and sesquiterpenes, respectively, than control plants (Fig. 4c). **
P = 0.0001
**

P = 0.0008
The increased production of terpenoids in RcMPD- ***

P < 0.0001
P = 0.0393
Emissions (pmol gFW–1 h–1)

P < 0.0001
overexpression lines indicates that endogenous IPK can access and ***
100

process de novo-produced IP to IPP and that the upstream portion
of the MVA pathway provides sufficient MVAP substrate to mea-
surably increase flux through the introduced alternative MVA path-
way. In comparison to previously generated AtIPK-overexpression
tobacco lines14, the RcMPD transgenics produced 1.5-fold more
sterols and up to 2-fold higher levels of mono- and sesquiterpenes,
thus suggesting that overall yields of terpenoid products from both
the MVA and MEP metabolic networks and crosstalk between the
two may be limited by endogenous IP levels.
To assess whether endogenous IPK in MPD transgenic plants
depleted all de novo generated IP, AtIPK was transiently overex-
pressed in the MPD-4 transgenic line (Fig. 5a). AtIPK overexpres-
sion in this background resulted in additional increases of 1.9- and
2.8-fold in emitted sesquiterpenes, β​-caryophyllene and 5-epi-aris-
tolochene, respectively, relative to the levels in the MPD-4 transgenic
line overexpressing an empty vector (Fig. 5b). These results suggest
that endogenous IPK activity was conditionally limiting in MPD
transgenics. Moreover, while sesquiterpene levels increased, levels
of the emitted monoterpenes, β​-ocimene and linalool, remained
unchanged (Fig. 5b). This effect on MEP-pathway-derived terpe-
noids may indicate that the IPP transporter involved in importing
cytoplasmic IPP into plastids, and/or plastidial enzymes acting
downstream of IPP, work near maximum capacity in these plant
backgrounds. In addition or alternatively, these results may suggest
that increased flux toward sesquiterpene formation occurred due to
the turnover of IP by IPK to relax FPP synthase inhibition, as IP and
DMAP competitively inhibit FPP synthase14.
Overexpression of HMGR and MPD amplifies flux toward down-
stream products. It is generally assumed that HMGR catalyses
P = 0.0065
***
***
* *** ***
**
50
0
β-Ocimene β-Caryo-
Linalool
phyllene
5-EPI-
aristolochene
Fig. 4 | Effect of Roseiflexus castenholzii MPD overexpression on
terpenoid formation in tobacco. a, RcMPD transcript levels in wild-type
and transgenic tobacco lines (empty vector control EV, MPD-1, MPD-4
and MPD-11) determined by qRT-PCR. Absolute transcript levels of
RcMPD are shown as picograms per 200 ng total RNA (means ±​ s.e.m.,
n =​ 3 biologically independent samples). b, Sterol levels in tobacco leaves
of wild-type, EV control- and RcMPD-overexpressing lines. Data are
means ±​ s.e.m., n =​ 5 biologically independent samples for WT and
MPD-11, n =​ 3 biologically independent samples for EV, and n =​ 6
biologically independent samples for MPD-1 and MPD-4. c, Emission
of monoterpenes, β​-ocimene and linalool, and sesquiterpenes,
β​-caryophyllene and 5-epi-aristolochene, from tobacco leaves of
wild-type, EV control- and RcMPD-overexpressing lines. Data are
means ±​ s.e.m., n =​ 6 biologically independent samples for WT, except
n =​ 3 biologically independent samples for EV and MPD-11, n =​ 5
biologically independent samples for MPD-1, and n =​ 8 biologically
independent samples for MPD-4. *P <​ 0.05; **P <​ 0.01; ***P <​ 0.001
(two-tailed Student’s t-test); nd, not detected.

Nature Plants | www.nature.com/natureplants

Articles
a 1
0.1
mRNA levels
0.01
nd nd
0.001
AtlPK
b MPD-EV MPD-IPK
2.7
P = 0.027
2.1
P = 0.0155
Emission (nmol gFW–1 h–1)
1.5
* RcMPD possesses a cryptic peroxisomal targeting signal blocked
0.9 * *
0.3
0.2
0.1
0.0
β-Ocimene Linalool
Fig. 5 | Effect of overexpression of AtIPK and AtHMGR1 on terpenoid
formation in RcMPD-4 tobacco transgenic line. a, Levels of AtIPK and
AtHMGR1 mRNAs in tobacco leaves of the RcMPD-4 transgenic line
infiltrated with Agrobacterium carrying the empty vector control (MPD-EV)
(black bars), the AtIPK construct (white bars) and the AtHMGR1 construct
(grey bars). Absolute transcript levels of AtIPK and AtHMGR1 are shown
as picograms per 200 ng total RNA (means ±​ s.e.m., n =​ 3 biologically
independent samples). b, Sesquiterpene and monoterpene emission from
RcMPD-4 transgenic tobacco leaves transiently overexpressing EV, AtIPK
and AtHMGR1. All data are means ±​ s.e.m. (n =​ 3 biologically independent
samples). *P <​ 0.05; **P <​ 0.01 (two-tailed Student’s t-test); nd,
not detected.
the rate-limiting step of the MVA pathway1,2,6,7. To test whether
increasing expression of HMGR in the RcMPD background fur-
ther enhances terpenoid production, we transiently overexpressed
AtHMGR in RcMPD transgenic tobacco leaves. Compared to
the MPD-4 line expressing an empty vector as a control, levels of
the monoterpenes β​-ocimene and linalool increased by 2.7- and
4.6-fold, respectively, in MPD-4 lines overexpressing AtHMGR
(Fig. 5b). Even higher levels were achieved for the sesquiterpenes
β​-caryophyllene and 5-epi-aristolochene, reaching 4.6- and 16.5-
fold increases, respectively (Fig. 5b). Taken together, the coexpres-
sion of AtHMGR with RcMPD in tobacco leaves has the potential to
enhance monoterpene formation by up to 20-fold and sesquiterpene
production by over 130-fold relative to their production in wild-
type plants (Figs. 4c and 5b). To investigate whether the AtHMGR-
RcMPD tobacco platform is suitable for heterologous production of
valuable terpenoids, Santalum album santalene synthase (SaSSy)26
was also coexpressed (Supplementary Fig. 8a). Overexpression of
SaSSy with AtHMGR and RcMPD resulted in almost tenfold higher
emitted levels of the non-native sesquiterpenes α​-exo-bergamotene
NATurE PlAnTs
and α​-santalene relative to overexpression of the SaSSy gene alone
in wild-type background (Supplementary Fig. 8b).
MPD-EV
MPD-IPK PMK contributes to controlling carbon flux through the MVA
MPD-HMGR pathway. The introduced bacterial RcMPD and endogenous NtPMK
genes both encode enzymes that use MVAP as a substrate (Fig. 1).
While PMK in plants resides in peroxisomes27, we assumed that the
bacterial RcMPD would a priori localize to the cytoplasm. In this
scenario, the increase in terpenoid production observed in RcMPD
transgenic lines may result from the introduction of an alternative
biosynthetic route that bypasses the peroxisomal MVAP import and
nd nd IPP export steps associated with the naturally occurring plant MVA
pathway (Fig. 1). To test this hypothesis, we first examined the in
P = 0.0032
AtHMGR1 planta subcellular localization of the introduced RcMPD protein.
GFP was fused to either the amino terminus or carboxy terminus
MPD-HMGR of RcMPD and transiently expressed in N. benthamiana leaves. The
** GFP signal for the C-terminal GFP fusion protein RcMDP–GFP
was detected in the cytoplasm (Supplementary Fig. 9). In contrast
and unexpectedly, the green fluorescence of GFP associated with
P = 0.0267
the N-terminal GFP fusion protein GFP–RcMPD was observed in
P = 0.0035
peroxisomes (Supplementary Fig. 9), suggesting that the bacterial
**
following fusion of GFP to the C terminus of RcMPD. Further
examination of the bacterial RcMPD sequence revealed a peroxi-
somal targeting signal type 2 (PTS2; (R/K) (L/V/I) X5 (H/Q) (L/A))
(Supplementary Fig. 10)28, which, despite not being present in the
N terminus, was perhaps still recognized by the plant peroxisomal
import machinery.
Given the peroxisomal localization of RcMPD, enhanced terpe-
noid production in RcMPD- and MPD4-AtIPK tobacco transgenics
indicates that generated IP can be transported out of this organ-
β-Caryo- 5-EPI-
phyllene aristolochene
elle probably in a similar way as IPP. These results also suggest that
peroxisomal MVAP levels are sufficient to achieve the observed
increases in overall terpenoid production (Fig. 5b). Therefore,
flux through the classical MVA pathway is, in part, limited by the
conversion of MVAP to MVAPP by PMK. To test this hypothesis,
Arabidopsis PMK was transiently overexpressed in wild-type tobacco
leaves under control of the CaMV 35S promoter (Fig. 6a). Relative
to the empty-vector control, AtPMK overexpression led to 4- and
44-fold increases in β​-caryophyllene and 5-epi-aristolochene levels,
respectively, with negligible effects on monoterpene levels (Fig. 6b).
Next, because HMGR is presumed to catalyse the rate-limiting step
of the MVA pathway1,2,6,7, we transiently coexpressed AtHMGR with
AtPMK in wild-type tobacco leaves (Fig. 6a). Similarly, compared
to empty-vector controls, transient overexpression of AtHMGR
alone led to increased emission of sesquiterpenes (2.8- and 9.4-fold
for β​-caryophyllene and 5-epi-aristolochene, respectively) with no
discernible effect on monoterpene levels (Fig. 6b). Compared to
overexpressing AtPMK and AtHMGR individually, overexpress-
ing the two genes together further increased β​-caryophyllene and
5-epi-aristolochene emission by 17- and 63-fold, respectively, rela-
tive to the empty-vector control (Fig. 6b). Coexpressing AtPMK and
AtHMGR also increased emission of monoterpenes, on average, by
4.6-fold (Fig. 6b), indicating that sufficient IPP levels are produced
in the cytoplasm to drive plastidial terpenoid biosynthesis. These
results demonstrate that PMK does indeed share control of flux with
HMGR through the plant MVA pathway.
PMK may exert control on flux through the MVA pathway by
pulling on the peroxisomal MVAP pool imported from the cyto-
plasm. This shift in the transport equilibrium toward peroxisomes
depends on PMK catalytic efficiency and PMK’s peroxisomal
concentration. In contrast to all other enzymes of the classical
MVA pathway except MKs, PMKs are encoded by single-copy
genes6. Previous biochemical characterization of Arabidopsis
PMK (AtPMK, At1g31910) as well as N. tabacum PMK (NtPMK,
XP_016504246) performed here, shows that PMKs possess high
Nature Plants | www.nature.com/natureplants
NATurE PlAnTs
a 1
WT-EV
WT-PMK
0.1
mRNA levels
0.01
nd nd
0.001
AtPMK
b which has long been considered to catalyse the rate-limiting step in
WT-EV
1,200
WT-PMK
900
Emission (pmol gFW–1 h–1)
600
400
P < 0.0001
300
P = 0.0148
200
***
100
* ***
0
β-Ocimene Linalool
Fig. 6 | Effect of AtPMK and AtHMGR1 overexpression on terpenoid
formation in tobacco leaves. a, Levels of AtPMK and AtHMGR1 mRNAs
in wild-type tobacco leaves infiltrated with Agrobacterium carrying the
empty vector control (WT-EV) (black bars), the AtPMK construct (white
bars), the AtHMGR1 construct (grey bars) and the AtPMK construct
with the AtHMGR1 construct (dotted bars). Absolute transcript levels
of AtPMK and AtHMGR1 are shown as picograms per 200 ng total RNA
(means ±​ s.e.m., n =​ 3 biologically independent samples). b, Sesquiterpene
and monoterpene emission from tobacco wild-type leaves transiently
overexpressing EV, AtPMK, AtHMGR1 and AtPMK and AtHMGR1 together.
All data are means ±​ s.e.m., n =​ 3 biologically independent samples, except
n =​ 6 biologically independent samples for WT-EV. *P <​ 0.05; **P <​ 0.01;
***P <​ 0.001 (two-tailed Student’s t-test).
specificity for MVAP with KM values of 12 µ​M (ref.  13) and 31 µ​M,
respectively (Supplementary Table 2). The catalytic efficiencies
(kcat/KM) of PMKs from both species were comparable (1.7 ×​  106
and 7.1 ×​  105 M−1 s−1 for AtPMK13 and NtPMK, respectively). These
efficiencies are also similar to AtIPK (1.0 ×​  105 M−1s−1) characterized
previously14 and NtIPK (XP_009616074) (2.3 ×​  105 M−1 s−1) charac-
terized here (Supplementary Table 2). As PMK is catalytically effi-
cient (Supplementary Table 2), and its encoding gene is one of the
most highly expressed MVA pathway genes6, unexplored transcrip-
tional and/or post-transcriptional regulation may control PMK
activity and metabolic flux through the MVA pathway.
Conclusion
Our study shows that Nudix hydrolases function as a part of key
regulatory machinery capable of modulating the metabolic out-
come of terpenoid metabolic networks in plants. Furthermore, the
work described here demonstrates that IP and possibly DMAP in
plants are not produced de novo, nor are they the result of cumu-
lative effects of nonspecific phosphatase activity. Instead, IP and
possibly DMAP originate from the active dephosphorylation of IPP
(DMAPP) by dedicated Nudix hydrolases that, together with IPK,
Nature Plants | www.nature.com/natureplants
Articles
WT-HMGR
coordinately regulate the concentration of IPP destined for higher-
order terpenoid biosynthesis. Introducing de novo IP formation by
WT-PMK–HMGR
overexpressing a bacterial RcMPD in tobacco revealed that a heter-
ologous, alternative route for MVA pathway metabolic flux increases
the overall production of terpenoids produced via both the MVA
and MEP pathways. The unexpected peroxisomal localization of the
overexpressed RcMPD, which competes for the same MVAP sub-
strate as PMK, demonstrates that the MVAPP formed by PMK is
probably a limiting factor for the biosynthesis of downstream MVA-
nd nd
derived terpenoids. Indeed, transient overexpression of AtPMK in
wild-type tobacco led to an increase in emitted sesquiterpenes, as
AtHMGR1
did overexpression of AtHMGR (Fig. 6b), the encoded enzyme of
P < 0.0001
P = 0.0008
the MVA pathway. Coexpression of AtPMK and AtHMGR substan-
tially enhanced the formation of both sesquiterpenes and mono-
WT-HMGR *** terpenes relative to the empty-vector controls (Fig. 6b), suggesting
WT-PMK–HMGR ***
P < 0.0001 that these two genes are positively epistatic and encode enzymes
that have major roles in controlling flux through the plant MVA
pathway. When AtHMGR was coexpressed with RcMPD, the latter
of which provides a bypass to the PMK-catalysed reaction, again,
P = 0.0002
***
synergistic effects were observed for both emitted sesquiterpenes
P < 0.0001
and monoterpenes (Fig. 5b). These results further support PMK
P = 0.0084
P = 0.0005
being an unsuspected regulatory hub in the plant MVA pathway and
***
*** hint at a role for IP in regulating the formation of both MVA and
** MEP pathway-derived terpenoids. Beyond advancing fundamental
knowledge, this study offers new strategies for high-level produc-
tion of economically valuable isoprenoids.
β-Caryo- 5-Epi-
phyllene aristolochene
Methods
Plant material. N. tabacum cv. Xanthi was used for generation of transgenic lines
and transient expression. Arabidopsis T-DNA insertion mutant lines, nudx1-1
(SALK_025320C), nudx1-2 (SAIL_236_D10), nudx3-1 (SAIL_554_G07) and
nudx3-2 (SALK_009963), were obtained from the Arabidopsis Biological Resource
Center. Homozygosity of obtained mutant lines were verified by PCR on isolated
genomic DNA using respective gene- and T-DNA border-specific primers
(Supplementary Data Set 1) designed using T-DNA Express Primer Design (Salk
Institute). All plant material was grown in a greenhouse or growth room under a
16 h light/8 h dark photoperiod.
Generation of transgenic tobacco plants overexpressing RcMPD. The RcMPD
open reading frame (ORF) was codon-optimized for plant systems and synthesized
by Clontech. It was subcloned into the binary vector pB2GW7 under the control
of the cauliflower mosaic virus 35S promoter using the Gateway LR Clonase II
(Invitrogen). Transgenic tobacco plants were obtained via Agrobacterium
tumefaciens (strain EHA105 carrying 35S-RcMPD) leaf disc transformation using
a standard transformation protocol29. Plants rooted on BASTA selection (1 mg l−​1)
were screened for the RcMPD presence using forward and reverse, RcMPD_qRT_
for and RcMPD_qRT_rev, primers (Supplementary Data Set 1). Untransformed
tobacco plants as well as plants transformed with empty vector were used as
controls in all experiments.
Transient overexpression in tobacco. For transient overexpression constructs,
ORFs of AtHMGR1 (G12571), AtPMK (U65928), AtIPK (GC104863), AtNudx1
(G50379) and AtNudx3 (U16680) were obtained from the Arabidopsis Biological
Resource Center and transferred from a Gateway-compatible entry vector into
the binary vector pB2GW7 under the control of the cauliflower mosaic virus 35S
promoter using the Gateway LR Clonase II (Invitrogen). For the santalene synthase
overexpression construct, full-length SaSSy was also subcloned into the binary
vector pB2GW7. Transient overexpression was achieved by A. tumefaciens (strain
EHA105 carrying the corresponding construct) infiltration of 2–3 leaves of wild-
type or RcMPD-4 transgenic tobacco plants as described previously30. Twenty-four
hours after infiltration, scent emission was analysed. For plants co-infiltrated with
multiple constructs, equal amounts of Agrobacterium cultures with OD600 nm 1.0
were mixed and infiltrated.
RNA isolation and qRT-PCR. RNA isolation from Arabidopsis and tobacco
tissues, cDNA synthesis and qRT-PCR analysis were performed as described
previously14. Gene-specific primers were designed using the PrimerExpress
software (Applied Biosystems) and are shown in Supplementary Data Set 1. For the
absolute quantification of gene transcript levels, respective cDNA fragments were
purified, diluted to several concentrations between 160 pg ml−1 and 1.28 pg ml−1,
and used to generate standard curves in qRT-PCR with gene-specific primers.
Absolute quantities of individual transcripts were calculated on the basis of
Articles
standard curves, and expressed as a picograms (pg) of mRNA per 200 ng of total
RNA or as a percentage of the expression in wild-type. Each data point represents
three biological and three technical replicates.
Metabolic profiling. Sterol extraction and analysis were performed as described
previously14. Floral and leaf volatiles were collected using a closed-loop stripping
system as described earlier14.
Subcellular localization. The RcMPD ORF was cloned into the binary vectors
pK7FWG2 and pK7WGF2 with and without stop codons to generate N- and
C-terminal GFP-fusion constructs, respectively. mCherry markers for peroxisome
(px-rk CD3-983) and mitochondria (mt-rk CD3-991) were co-infiltrated with GFP
constructs31. Constructs, markers and empty-vector controls were transformed into
Agrobacterium (EHA105) and infiltrated into 3-week-old Nicotiana benthamiana
leaves as previously described30. Plant tissues were analysed 1–2 days after
infiltration using the Nikon A1Rsi laser scanning confocal microscope as
described previously32.
Heterologous expression and purification of recombinant proteins. N. tabacum
PMK and IPK ORFs were codon-optimized for E. coli expression and synthesized
by SGI-DNA and Integrated DNA Technologies, respectively. The ORFs of
Arabidopsis Nudix genes were PCR-amplified from cDNA with gene-specific
forward and reverse primers (Supplementary Data Set 1), with the exception of
AtNudx3, which was obtained from the Arabidopsis Biological Resource Center
(clone U16680). Transit peptide sequences were excluded when present. ORFs
were cloned into a modified pET28b vector with an N-terminal His8-tag by the
In-Fusion cloning system (Takara Bio USA). ORFs for Nudix 4, 8–11, 16–18, 21–24
and 26 were also cloned into cold-shock expression vector pCold I DNA (Takara
Bio USA). The resulting constructs were transformed into BL21(DE3) E. coli and
grown in terrific broth medium supplemented with 50 μ​g  ml−1 kanamycin or
100 μ​g  ml−1 ampicillin. Protein expression was induced with 0.5 mM isopropyl
1-thio-β​-d-galactopyranoside at 18 °C for pET28b or 15 °C for pCold I DNA. After
20–24-h incubation, cells were collected by centrifugation and lysed by sonication
in 50 mM Tris-HCl pH 8.0 buffer, 0.5 M NaCl, 20 mM imidazole and 10% glycerol.
After removal of the cell debris by centrifugation, expressed proteins were purified
from the supernatants by immobilized metal-affinity chromatography with HisPur
Ni-NTA resin (ThermoFisher Scientific). Pure proteins were eluted with 50 mM
Tris-HCl pH 8.0, 0.5 M NaCl buffer supplemented with 250 mM imidazole followed
by buffer exchange to 50 mM Tris-HCl pH 8.0 and 0.2 M NaCl for storage. For
crystal screening and steady-state kinetic analysis, the N-terminal His8-tags of
AtNudx1 and AtNudx3 were removed with thrombin and remaining histidine-
tagged protein was removed by passing over HisPur Ni-NTA resin. AtNudx1 and
AtNudx3 were further purified by size-exclusion chromatography on a Superdex
200 16/60 column (GE Lifesciences) equilibrated and eluted with 50 mM Tris-HCl
pH 8.0 supplemented with 0.2 M NaCl and 2 mM dithiothreitol.
Enzyme assays. Phosphatase activity was monitored at 37 oC using BIOMOL Green
reagent (Enzo Life Sciences) to detect and quantify released phosphate at 623 nm
using a phosphate standard curve. The optimum pH for activity was determined
using a three-component buffer system of 50 mM acetic acid, 50 mM MES and
100 mM Tris-HCl with 0.1 mM IPP as substrate. Magnesium ion dependence
was investigated with 1 mM to 20 mM MgCl2. Steady-state parameters were
determined in 100 mM TAPS pH 8.5 with 10 mM MgCl2 and varied concentrations
of substrates IPP (Isoprenoids), DMAPP (Isoprenoids), GPP (Isoprenoids), FPP
(Isoprenoids) and 8-oxo-dGTP (TriLink Biotechnologies). Assays with 8-oxo-
dGTP included 0.05 U of inorganic pyrophosphatase (ThermoFisher Scientific) per
0.1 ml assay. Reactions were initiated by the addition of enzyme and quenched by
the addition of BIOMOL Green.
Activities of NtPMK and NtIPK were analysed by the lactate dehydrogenase/
pyruvate kinase coupled assay as detailed previously14. Briefly assays were
conducted in 100 mM Na+-Hepes pH 7.5 supplemented with 8 mM MgCl2 and
100 mM KCl with coupling enzymes pyruvate kinase/lactate dehydrogenase
(Sigma), 4 mM ATP, 2 mM phosphoenolpyruvate, 0.16 mM NADH and varied
concentrations of (R)-MVAP (Sigma) and IP (Isoprenoids) as substrate for NtPMK
and NtIPK, respectively. Reactions were monitored at 340 nm and 30 °C.
Controls included assays without enzyme and without substrate. All enzyme
assays were performed at an appropriate enzyme concentration so that the reaction
velocity was linear and proportional to the enzyme concentration during the
incubation time period. Kinetic data were fitted using Prism (GraphPad Software)
to the Michaelis–Menten equation to compute KM and kcat. At least triplicate assays
were performed for all data points.
Crystallization and structure solution. Crystallization trials were conducted
by the hanging-drop method using Hampton Research crystal screens. Typically,
1 μ​l of protein at 12 mg ml−1 in 50 mM Tris-HCl pH 8.0, 0.2 M NaCl and 2 mM
dithiothreitol was mixed with 1 μ​l of each reservoir solution and incubated at 4 ˚C
over a 500 μ​l reservoir solution. Ligand-bound structures were obtained by co-
crystallization with 5–10 mM ligand. Diffraction-quality AtNudx1 crystals were
obtained from a mixture of the protein with 0.1 M succinic acid pH 5.0, 20% PEG
NATurE PlAnTs
4000 and 0.3 M Mg(NO3)2. Crystals were flash-frozen in cryoprotectant of 17%
ethylene glycol and reservoir buffer plus substrate. X-ray diffraction data were
collected at beamlines 8.2.1 and 8.2.2 of the Advanced Light Source at Lawrence
Berkeley National Laboratory. Diffraction images were indexed and integrated
with iMosflm33, and the measured reflection intensities were scaled and merged
using CCP4 Aimless34. The initial structural elucidation of AtNudx1 was obtained
by molecular replacement with CCP4 MolRep35 using a search model derived
from Rickettsia Felis MutT (pdb entry 4KYX) with non-conserved amino-acid
residues pruned using CPP4 Chainsaw36. The structural model was refined with
Phenix Refine and inspected against electron-density maps and adjusted manually
in Coot37. Autobuilding was performed with Phenix Autobuild38. Subsequent
structure determinations of other forms of AtNudx1, which were crystallized
isomorphously, were initiated with the refined AtNudx1 model, after omission of
ligands and water molecules.
GUS staining and imaging. Approximately 2-kb regions upstream of each
AtNudx1 and AtNudx3 genes were cloned from Arabidopsis Col-0 genomic
DNA with primers containing attB Gateway linkers (pNDX1_F and pNDX1_R2;
pNDX3_F and pNDX3_R2; Supplementary Data Set 1). The PCR products were
inserted into the pDONR207 entry vector via a BP clonase I Gateway reaction
(Invitrogen) and the fragment was sequenced to verify the identity of the insert.
The insert was then moved from the pDONR207 entry vector into a pMDC163
expression vector (GUS expression vector) by LR clonase I Gateway reaction
(Invitrogen)39. This expression vector was then transformed into Agrobacterium
tumefaciens GV3101n and used for infiltration of A. thaliana Col-0 plants via the
floral dip method40. Hygromycin-resistant transformants were selected for GUS
colorimetric assays.
Tissues were fixed in 90% acetone for 40 min at −​20 °C, washed twice with
phosphate buffer (pH 7.0), added to GUS staining solution (0.1% Triton X-100,
10 mM EDTA, 2 mM ferrocyanide, 2 mM ferricyanide, 100 mM sodium dibasic
phosphate, 100 mM sodium monobasic phosphate and 4 mM 5-bromo-4-chloro-
3-indolyl-β​-glucuronide), vacuum-infiltrated for 5 min and incubated at 37 °C
until staining was visible or a maximum for 4 days. Tissues were then cleared in
70% ethanol for 2 days and then imaged in a 8:2:1 chloral hydrate/distilled water/
glycerol solution. Images were taken on a Nikon Eclipse Ti-2 inverted microscope
with a Nikon Digital Sight DS-Fi2 camera. Expression patterns were analysed
for each construct in at least three independent transgenic lines to account for
positional effects of the insert.
Reporting Summary. Further information on experimental design is available in
the Nature Research Reporting Summary linked to this article.
Data availability. Structural information for AtNudx1 and the AtNudx1–IPP
complex have been deposited in the Protein Data Bank, PDB ID codes 6DBY
and 6DBZ, respectively. Figures 2–6 and Supplementary Figs. 1–4 and 6–9 have
associated raw data available upon request to the corresponding authors.
Received: 29 March 2018; Accepted: 13 July 2018;
Published: xx xx xxxx
References
1. Hemmerlin, A., Harwood, J. L. & Bach, T. J. A raison d’être for two distinct
pathways in the early steps of plant isoprenoid biosynthesis? Prog. Lipid Res.
51, 95–148 (2012).
2. Tholl, D. Biosynthesis and biological functions of terpenoids in plants.
Adv. Biochem. Eng. Biotechnol. 148, 63–106 (2015).
3. Gershenzon, J. & Dudareva, N. The function of terpene natural products in
the natural world. Nat. Chem. Biol. 3, 408–414 (2007).
4. Tippmann, S., Chen, Y., Siewers, V. & Nielsen, J. From flavors and
pharmaceuticals to advanced biofuels: Production of isoprenoids in
Saccharomyces cerevisiae. Biotechnol. J. 8, 1435–1444 (2013).
5. Zerbe, P. & Bohlmann, J. Plant diterpene synthases: Exploring modularity
and metabolic diversity for bioengineering. Trends Biotechnol. 33,
419–428 (2015).
6. Vranová, E., Coman, D. & Gruissem, W. Network analysis of the MVA
and MEP pathways for isoprenoid synthesis. Annu. Rev. Plant Biol. 64,
665–700 (2013).
7. Hemmerlin, A. Post-translational events and modifications regulating plant
enzymes involved in isoprenoid precursor biosynthesis. Plant Sci. 203–204,
41–54 (2013).
8. Vickers, C. E., Bongers, M., Liu, Q., Delatte, T. & Bouwmeester, H. Metabolic
engineering of volatile isoprenoids in plants and microbes. Plant Cell Environ.
37, 1753–1775 (2014).
9. Wu, S. et al. Redirection of cytosolic or plastidic isoprenoid
precursors elevates terpene production in plants. Nat. Biotechnol. 24,
1441–1447 (2006).
10. Farhi, M. et al. Generation of the potent anti-malarial drug artemisinin in
tobacco. Nat. Biotechnol. 29, 1072–1074 (2011).
Nature Plants | www.nature.com/natureplants
NATurE PlAnTs
11. Muñoz-Bertomeu, J., Sales, E., Ros, R., Arrillaga, I. & Segura, J. Up-regulation
of an N-terminal truncated 3-hydroxy-3-methylglutaryl CoA reductase
enhances production of essential oils and sterols in transgenic Lavandula
latifolia. Plant Biotechnol. J. 5, 746–758 (2007).
12. Liao, P., Hemmerlin, A., Bach, T. J. & Chye, M.-L. The potential of the
mevalonate pathway for enhanced isoprenoid production. Biotechnol. Adv. 34,
697–713 (2016).
13. Dellas, N., Thomas, S. T., Manning, G. & Noel, J. P. Discovery of a metabolic
alternative to the classical mevalonate pathway. eLife 2, e00672 (2013).
14. Henry, L. K., Gutensohn, M., Thomas, S. T., Noel, J. P. & Dudareva, N.
Orthologs of the archaeal isopentenyl phosphate kinase regulate terpenoid
production in plants. Proc. Natl Acad. Sci. USA 112, 10050–10055 (2015).
15. VanNice, J. C. et al. Identification in Haloferax volcanii of phosphomevalonate
decarboxylase and isopentenyl phosphate kinase as catalysts of the terminal
enzyme reactions in an archaeal alternate mevalonate pathway. J. Bacteriol.
196, 1055–1063 (2014).
16. Karačić, Z. et al. A novel plant enzyme with dual activity: an atypical Nudix
hydrolase and a dipeptidyl peptidase III. Biol. Chem. 398, 101–112 (2017).
17. Kraszewska, E. The plant Nudix hydrolase family. Acta Biochim. Pol. 55,
663–671 (2008).
18. McLennan, A. G. Substrate ambiguity among the nudix hydrolases:
biologically significant, evolutionary remnant, or both? Cell. Mol. Life Sci. 70,
373–385 (2013).
19. Ogawa, T. et al. Molecular characterization of organelle-type Nudix
hydrolases in Arabidopsis. Plant Physiol. 148, 1412–1424 (2008).
20. Ogawa, T., Ueda, Y., Yoshimura, K. & Shigeoka, S. Comprehensive analysis of
cytosolic nudix hydrolases in Arabidopsis thaliana. J. Biol. Chem. 280,
25277–25283 (2005).
21. Yoshimura, K., Ogawa, T., Ueda, Y. & Shigeoka, S. AtNUDX1, an 8-oxo-7,
8-dihydro-2′​-deoxyguanosine 5′​-triphosphate pyrophosphohydrolase, is
responsible for eliminating oxidized nucleotides in Arabidopsis. Plant Cell
Physiol. 48, 1438–1449 (2007).
22. Liu, J. et al. Structural insights into the substrate recognition mechanism of
Arabidopsis GPP-bound NUDX1 for noncanonical monoterpene biosynthesis.
Molecular Plant 11, 218–221 (2018).
23. Svensson, L. M. et al. Crystal structure of human MTH1 and the 8-oxo-
dGMP product complex. FEBS Lett. 585, 2617–2621 (2011).
24. Nakamura, T. et al. Structural and dynamic features of the MutT protein
in the recognition of nucleotides with the mutagenic 8-oxoguanine base.
J. Biol. Chem. 285, 444–452 (2010).
25. Magnard, J.-L. et al. Biosynthesis of monoterpene scent compounds in roses.
Science 349, 81–83 (2015).
26. Jones, C. G. et al. Sandalwood fragrance biosynthesis involves sesquiterpene
synthases of both the terpene synthase (TPS)-a and TPS-b subfamilies,
including santalene synthases. J. Biol. Chem. 286, 17445–17454 (2011).
27. Simkin, A. J. et al. Peroxisomal localisation of the final steps of the mevalonic
acid pathway in planta. Planta 234, 903–914 (2011).
28. Reumann, S., Chowdhary, G. & Lingner, T. Characterization, prediction
and evolution of plant peroxisomal targeting signals type 1 (PTS1s).
Biochim. Biophys. Acta Mol. Cell Res. 1863, 790–803 (2016).
29. Curtis, I. S., Davey, M. R. & Power, J. B. Leaf disk transformation.
Methods Mol. Biol. 44, 59–70 (1995).
30. Sparkes, I. A., Runions, J., Kearns, A. & Hawes, C. Rapid, transient expression
of fluorescent fusion proteins in tobacco plants and generation of stably
transformed plants. Nat. Protoc. 1, 2019–2025 (2006).
Nature Plants | www.nature.com/natureplants
Articles
31. Nelson, B. K., Cai, X. & Nebenführ, A. A multicolored set of in vivo organelle
markers for co-localization studies in Arabidopsis and other plants. Plant J.
51, 1126–1136 (2007).
32. Yoo, H. et al. An alternative pathway contributes to phenylalanine
biosynthesis in plants via a cytosolic tyrosine:phenylpyruvate
aminotransferase. Nat. Commun. 4, 2833 (2013).
33. Powell, H. R., Battye, T. G. G., Kontogiannis, L., Johnson, O. & Leslie, A. G. W.
Integrating macromolecular X-ray diffraction data with the graphical user
interface iMosflm. Nat. Protoc. 12, 1310–1325 (2017).
34. Evans, P. R. & Murshudov, G. N. How good are my data and what is the
resolution? Acta Crystallogr. D 69, 1204–1214 (2013).
35. Vagin, A. & Teplyakov, A. MOLREP: an automated program for molecular
replacement. J. Appl. Crystallogr. 30, 1022–1025 (1997).
36. Stein, N. et al. CHAINSAW: a program for mutating pdb files used
as templates in molecular replacement. J. Appl. Crystallogr. 41,
641–643 (2008).
37. Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics.
Acta Crystallogr. D 60, 2126–2132 (2004).
38. Terwilliger, T. C. et al. Iterative model building, structure refinement and
density modification with the PHENIX AutoBuildwizard. Acta Crystallogr. D
64, 61–69 (2008).
39. Curtis, M. D. & Grossniklaus, U. A Gateway cloning vector set for
high-throughput functional analysis of genes in plants. Breakthr. Technol. 133,
462–469 (2003).
40. Clough, S. J. & Bent, A. F. Floral dip: a simplified method for
Agrobacterium-mediated transformation of Arabidopsis thaliana. Plant J. 16,
735–743 (1998).
Acknowledgements
This work was supported by the USDA National Institute of Food and Agriculture
Predoctoral Grant 2017-67011-26076 to L.K.H. and start-up funds from Purdue
University to J.R.W. and S.A.K. This work was also supported by the USDA National
Institute of Food and Agriculture Hatch Project number 177845.
Author contributions
J.P.N. and N.D. conceived the project; L.K.H., S.T.T., J.R.W., S.A.K., J.B., J.P.N. and
N.D. designed the experiments; L.K.H., S.T.T., J.R.W., J.H.L. and T.C.D. performed the
experiments; L.K.H., S.T.T., J.R.W., J.H.L., T.C.D., S.A.K., J.B., J.P.N. and N.D. analysed
the data; L.K.H., S.T.T., J.R.W., J.P.N. and N.D. wrote the paper. All authors read and
edited the manuscript.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary information is available for this paper at https://doi.org/10.1038/
s41477-018-0220-z.
Reprints and permissions information is available at www.nature.com/reprints.
Correspondence and requests for materials should be addressed to J.P.N. or N.D.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
 nature research | reporting summary
Corresponding author(s): Natalia Dudareva

Reporting Summary
Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Research policies, see Authors & Referees and the Editorial Policy Checklist.

Statistical parameters
When statistical analyses are reported, confirm that the following items are present in the relevant location (e.g. figure legend, table legend, main
text, or Methods section).
n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
An indication of whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.

A description of all covariates tested

A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistics including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND
variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)

For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.

For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated

Clearly defined error bars

State explicitly what error bars represent (e.g. SD, SE, CI)

Our web collection on statistics for biologists may be useful.

Software and code

Policy information about availability of computer code
Data collection GC-MS data were collected using MSD Chemstation v.E.01.00.237. LC-MS data were collected using ChemStation LC and CE Drivers
version A.01.13 [014]. qRT-PCR data were collected using StepOne v.2.2.2.

Data analysis GC-MS data were analyzed using GraphPad Prism v6.05. Diffraction images were indexed and integrated with iMosflm v.7.1.0, and the
measured reflection intensities were scaled and merged using CCP4 Aimless v.0.3.11. The initial structural elucidation of AtNudx1 was
obtained by molecular replacement with CCP4 MolRep v.11.2.08 using a search model derived from Rickettsia Felis MutT with
nonconserved amino-acid residues pruned using CPP4 Chainsaw v.6.4. The structural model was refined with Phenix Refine v.1.12 and
inspected against electron-density maps and adjusted manually in Coot v.0.8.9. Autobuilding was performed with Phenix Autobuild
v.1.12.
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors/reviewers
April 2018

upon request. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

1
Data

nature research | reporting summary

Policy information about availability of data
All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
- A list of figures that have associated raw data
- A description of any restrictions on data availability
Structural information for AtNudx1 and AtNudx1-IPP complex have been deposited at PBD under PDB ID codes 6DBY and 6DBZ, respectively. Figures 2, 3, 4,
5, 6, S1, S2, S3, S4, S6, S7, S8, and S9 have associated raw data available upon request.

Field-specific reporting
Please select the best fit for your research. If you are not sure, read the appropriate sections before making your selection.
Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/authors/policies/ReportingSummary-flat.pdf

Life sciences study design

All studies must disclose on these points even when the disclosure is negative.
Sample size A minimum sample size of 3 independent biological experiments was chosen to obtain valid statistical analyses. A sample size of 3 is accepted
as sufficient to perform statistical tests and to account for positional effects of the transgene in research using trangenic plants. In preliminary
screening (Supplementary Fig. 1) sample size was 2 as selected proteins were used in further investigation with statistical analysis.

Data exclusions No data was excluded

Replication At least three biological replicates were performed for each experiment. All attempts at replication were successful and all data included in
the manuscript.

Randomization Plants were randomized in greenhouse, and sample order was randomized for metabolite profiling.

Blinding Blinding is not necessary because results are quantitative for all experiments, and not subjective. Blinding was not possible in these
experiments, as samples were analyzed immediately following treatments, and thus the treatment was known to the investigator.

Reporting for specific materials, systems and methods

Materials & experimental systems Methods

n/a Involved in the study n/a Involved in the study
Unique biological materials ChIP-seq
Antibodies Flow cytometry
Eukaryotic cell lines MRI-based neuroimaging
Palaeontology
Animals and other organisms
Human research participants

Unique biological materials

Policy information about availability of materials
April 2018

Obtaining unique materials Seeds of transgenic tobacco plants generated for this study are available upon request.

2
Animals and other organisms

nature research | reporting summary

Policy information about studies involving animals; ARRIVE guidelines recommended for reporting animal research
Laboratory animals No laboratory animals were used in this study.

Wild animals No wild animals were involved in this study.

Field-collected samples This study did not involved field-collected samples.

April 2018