RESEARCH ARTICLE
◥PLANT SCIENCE
Natural historyÐguided omics reveals plant defensive
chemistry against leafhopper pests
Yuechen Bai^1 , Caiqiong Yang^1 , Rayko Halitschke^1 , Christian Paetz^2 , Danny Kessler^1 , Konrad Burkard^1 ,
Emmanuel Gaquerel^3 , Ian T. Baldwin^1 , Dapeng Li4,5*
Although much is known about plant traits that function in nonhost resistance against pathogens, little
is known about nonhost resistance against herbivores, despite its agricultural importance.Empoasca
leafhoppers, serious agricultural pests, identify host plants by eavesdropping on unknown outputs of
jasmonate (JA)–mediated signaling. Forward- and reverse-genetics lines of a native tobacco plant
were screened in native habitats with native herbivores using high-throughput genomic, transcriptomic,
and metabolomic tools to reveal anEmpoasca-elicited JA-JAZi module. This module induces an
uncharacterized caffeoylputrescine–green leaf volatile compound, catalyzed by a polyphenol oxidase in
a Michael addition reaction, which we reconstitute in vitro; engineer in crop plants, where it requires
a berberine bridge enzyme-like 2 (BBL2) for its synthesis; and show that it confers resistance to
leafhoppers. Natural history–guided forward genetics reveals a conserved nonhost resistance
mechanism useful for crop protection.
B
eing at the bottom of most terrestrial
food chains, plants are continuously
attacked by herbivores and pathogens
( 1 , 2 ). Research into plant traits that
provide resistance against these biotic
agents has primarily focused on nonhost re-
sistance to pathogens ( 3 – 5 ) and host resistance
to herbivores ( 6 , 7 ). This difference in empha-
sis likely reflects the greater physiological au-
tonomy of herbivores, which are selective in
choosing plants to attack, coupled with the
challenge of discovering resistance traits of
hosts that herbivores refuse to attack. Plants
rendered defenseless by the abrogation of de-
fense pathways can be attacked by nonhost
herbivores in nonchoice assays in the labora-
tory ( 8 , 9 ). However, these assays do not cap-
ture the selective procedures by which insects
choose their host plants in nature, which lim-
its the inferences that can be drawn from these
laboratory studies about nonhost resistance.
Because of the paucity of field studies, the
mechanisms and metabolic traits underlying
nonhost resistance against herbivores remain
largely unknown.
We found thatNicotiana attenuataplants—
transformed to silence the signaling that me-
diates inducible expression of host resistance
traits—when released into the wild, are con-
tinuously assessed and attacked by nonhost
insect herbivores when rendered defenseless
( 10 ). Among these opportunistic herbivores
is theEmpoascaleafhopper—a major pest
common to many crops. These leafhoppers
probe nonhost plants to eavesdrop on a plant’s
jasmonate (JA) signaling, which is elicited
upon probing ( 11 ). However,N. attenuata’s
portfolio of JA-elicited specialized metabo-
lites, such as alkaloids, protease inhibitors,
diterpene glycosides (17-HGL-DTGs), and elic-
ited volatiles, which are effective against host
herbivores, were excluded as nonhost resist-
ance traits ( 11 ). Because many agricultural
pests may be opportunistic herbivores, under-
standing mechanisms of nonhost resistance
against insects could accelerate the breed-
ing of durable resistance in crops. To date,
themostcommonlyusedstrategiestocon-
trol agricultural pests are insecticidal sprays
and ectopic expression of insecticidal pro-
teins, both of which have ecological draw-
backs ( 12 , 13 ).
To uncover the JA-elicited nonhost resistance
traits ofN. attenuata, we adopted a forward-
genetics strategy. We planted a replicated pop-
ulation of 650 recombinant inbred lines (RILs)
from a 26-parent multiparent advanced gen-
eration intercross (MAGIC) population into
a native habitat in Arizona, USA (Fig. 1A and
fig. S1). In this setting,Empoascaleafhop-
pers are abundant and damage their native
host cucumbers (Cucurbita foetidissima). JA-
deficientN. attenuatalines were attacked by
the leafhoppers ( 11 ) at rates that varied within
the MAGIC populations (Fig. 1A). We quan-
tifiedEmpoascaattack levels in 1907 indi-vidual plants of 674 RILs and parental lines
of the field-grown MAGIC population and
constructed a multi-omics dataset based on
high-throughput analyses of phytohormones,
transcriptomes, and metabolomes. This multi-
omics dataset was produced from leaves elic-
ited by a simulated herbivory treatment. We
mimicked herbivore attack by treating stand-
ardized puncture wounds (W) with oral secre-
tions (OS) ofManduca sextalarvae (W + OS)
to remove confounding factors caused by the
stochastic nature of insect attack in nature
and to capture transiently expressed genes
and metabolites in these samples from field-
grown plants (Fig. 1B and fig. S2).
To analyze associations among the genetic
and metabolic responses of this multi-omics
dataset, we first focused on JA signaling–
related genes and used previously acquired
knowledge ofN. attenuataleaf chemistry
( 14 , 15 ) to construct a coassociation network
of the JA-dependent module. This network
considered not only the correlations among
metabolites, phytohormones, and gene expres-
sions but also the shared single-nucleotide
polymorphisms (SNPs) inferred from meta-
bolic quantitative trait locus (mQTL) or expres-
sion QTL (eQTL) analyses for each of these
components(Fig.1C,fig.S3,anddataS1).
This coassociation network revealed that
JAs and JA-related genes nucleated by the JA-
regulated phenolamide master transcription
factor (TF) regulator NaMYB8 ( 16 )formedage-
netic hub that clustered with induced phenola-
mides, such asN-coumaroylputrescine (CoP),
N-caffeoylputrescine (CP),N-feruloylputrescine
(FP), and malonylated 17-HGL-DTGs (Fig. 1C).
More peripheral to this hub were glycosylated
17-HGL-DTG precursors, such as lyciumoside
I, lyciumoside IV, attenoside, and nicotiano-
side III, and other specialized metabolites,
such as nicotine, acylsugars, and flavonoids
(fig. S3). ANaJAZigene clustered centrally
to JAs but was distant fromNaJAR4and
NaCOI1, which suggests the engagement of
JA signaling.
To further resolve the components in the co-
association network responsible forEmpoasca
susceptibility, we conducted a pairwise corre-
lational analysis among the omics datasets and
Empoascaabundance and leaf area damaged
in the RILs of the MAGIC populations (Fig.
1D). The putrescine-derived phenolamides
and malonylated 17-HGL-DTGs were negatively
correlated with theEmpoascanumbers and
damage, whereas glycosylated 17-HGL-DTG
precursors were positively correlated. JA-
related genesNaMYB8,NaLOX3,NaAOC,
NaOPR3,NaJAR6, andNaWIPKexhibited
the highest negative correlation scores with
theEmpoascanumbers and damage. There
was considerable heterogeneity in the expres-
sion of the JA-related family of JAZ genes, with
the expression ofNaJAZa,NaJAZd,NaJAZf,RESEARCH
Baiet al.,Science 375 , eabm2948 (2022) 4 February 2022 1of9
(^1) Department of Molecular Ecology, Max Planck Institute for
Chemical Ecology, D-07745 Jena, Germany.^2 Department of
Biosynthesis/NMR, Max Planck Institute for Chemical
Ecology, D-07745 Jena, Germany.^3 Institut de Biologie
Moléculaire des Plantes du CNRS, Université de Strasbourg,
Strasbourg, France.^4 National Key Laboratory of Plant
Molecular Genetics, CAS Center for Excellence in Molecular
Plant Sciences, Institute of Plant Physiology and Ecology,
Chinese Academy of Sciences, Shanghai, China.^5 CAS-JIC
Center of Excellence for Plant and Microbial Sciences
(CEPAMS), Institute of Plant Physiology and Ecology,
Chinese Academy of Sciences, Shanghai, China.
*Corresponding author. Email: [email protected] (Y.B.);
[email protected] (I.T.B.); [email protected] (D.L.)