involved in the regulation and biosynthesis of
CP andm/z347.19, includingNaMYC2a,NaMYB8,
andNaAT1, which encodes a hydroxycinnamoyl-
coenzyme A:putrescine acyltransferase re-
sponsible for CP biosynthesis ( 16 ), as well as
two polyphenol oxidases (PPOs),NaPPO1and
NaPPO2, which are located on chromosomes 7
and 8, respectively (Fig. 3C). Coexpression
analyses of a microarray dataset of irMYB8
lines ( 23 ) usingNaAT1andNaDH29as baits
revealed a cluster of genes in theNaAT1group-
ings known to be involved in the biosynthesis
of CP, includingNaPAL1,NaPAL2,Na4CL1,
andNaC3H. A berberine bridge enzyme-like
(BBL) gene,NaBBL2, was highly coexpressed
withNaAT1and decreased in its induced ex-
pressions in irMYB8lines (Fig. 3C and fig.
S14A). We revisited the mQTL dataset and
found thatNaBBL2, located on chromosome 3,
was associated withm/z347.19 (Fig. 3C), al-
beit at reduced statistical significance (P=
0.0013). Time-resolved microarray data ( 24 )
of herbivory-elicited expression ofNaAT1,
NaPPO2, andNaBBL2revealed thatNaAT1
andNaPPO2showed similar induction pat-
terns inN. attenuatawild type (WT), whereas
NaBBL2washighlyinducedat1hourand
retained its induction at later time points,
albeit at reduced levels (fig. S14B). Moreover, the
herbivory-elicited inductions ofNaAT1,NaPPO1,
NaPPO2, andNaBBL2were reduced in an RNA
sequencing (RNA-seq) transcriptome dataset of
irMYC2lines (fig. S14C).
Although the number and order of the bio-
synthetic steps form/z347.19 accumulations
remained elusive, we hypothesized that pos-
sible oxidation and acylation reactions were
likely required. We therefore focused on oxi-
dases and acyltransferases collectively im-
puted from the multi-omics analysis. Candidate
genes included three acyltransferases,NaAT1,
NaAT2, andNaAT3; three polyphenol oxidases,
NaPPO1,NaPPO2, andNaPPO3; and a BBL
gene,NaBBL2. We evaluated the in vivo func-
tions of the candidate genes as well as NaAT1
as a positive control by silencing their expres-
sion inN. attenuatausing virus-induced gene
silencing (VIGS). Consistent with a previous
analysis, VIGS ofNaAT1abolishedm/z347.19
accumulation ( 14 ). Similarly, silencingNaPPO1,
NaPPO2, andNaBBL2expression also trun-
cated the elicitation ofm/z347.19 (Fig. 3D and
fig. S15). Untargeted metabolomics analysis
of the VIGS-silenced plants revealed that
silencingNaPPO1andNaPPO2truncated
m/z347.19 accumulations without changes
in other phenolamides (Fig. 3D and fig. S16).
These results suggest thatNaPPO1,NaPPO2,
andNaBBL2are required for the in vivo pro-
duction ofm/z347.19.
Previous analysis has suggested that the
additional C 6 H 8 O residue ofm/z347.19 is
produced from the fatty acid oxylipin cascade,
which converts C18 polyunsaturated fatty acids
released from biological membranes during
stresses, wounding, and herbivory ( 25 )topro-
duce green leaf volatiles (GLVs) enriched in
reactive C6 derivatives ( 14 ). We measured
herbivory-elicited GLVs in the same glasshouse-
grown MAGIC RIL population used for the im-
putation (Fig. 3) and conducted correlational
analysis with herbivory-elicitedm/z347.19 ac-
cumulations. (Z)-3-hexenal–derived volatiles,
such as (Z)-3-hexenyl-propanoate and (Z)-3-
hexenol, were the most significantly positively
correlated metabolites withm/z347.19, where-
as 1-hexanol, linalool, and other elicited vola-
tiles were not (Fig. 3F and fig. S17).
The C6 aldehydes, with their molecular for-
mula of C 6 H 8 O, are the most reactive alde-
hydes produced from the GLV pathway and
have been hypothesized ( 14 )tobethemissing
substrates for the biosynthesis ofm/z347.19.
Consistent with previous analyses ( 26 ), stably
silencing LIPOXYGENASE2 (irLOX2) in
N. attenuata, which controls the first committed
step in the GLV pathway, abolishes C6 alde-
hydes production and total GLV emissions
( 14 ),andstablysilencedcrossesofirLOX2
and irLOX3(irLOX2×irLOX3) completely elim-
inatedm/z347.19 production (Fig. 3F). Addi-
tionally, silencingNaHPL(with an antisense
construct, asHPL)—which catalyzes the for-
mation of the initial C6 GLV product, (Z)-3-
hexenal, and its isomer, (E)-2-hexenal—results
in considerable time-dependent reductions
of GLVs inN. attenuata( 27 ) and reducedm/z
347.19 accumulations to ~⅓of the WT levels
(Fig. 3F). From these results, we surmised that
in response toEmpoascaprobing,m/z347.19
is produced by a three-pronged metabolic
pathway composed of the LOX2–HPL-GLV
pathway, the LOX3-JA–regulated phenylpro-
panoid pathway, and the polyamine pathway,
which are condensed by NaPPO1, NaPPO2,
and NaBBL2 using CP and (Z)-3-hexenal or
(E)-2-hexenal to producem/z347.19 (Fig. 3E).
To test this hypothesis, we isolated purified
NaPPO1, NaPPO2, and NaBBL2 proteins with
N-terminal hexahistidine tags after expres-
sion inEscherichia coli(fig. S18). Incubation
of either NaPPO1 or NaPPO2 (NaPPO1/2) with
CP and (Z)-3-hexenal yielded am/z347.19 peak
with a MS/MS spectrum and retention time
identical to that of them/z347.19 induced in
N. attenuataleaves (fig. S19). Additionally,
by-products of doubly charged CP dimers at
m/z250.13 ([M+2H]2+,C 26 H 36 N 4 O 6 2+), which
are not detected in OS-induced leaves of
N. attenuataWT (data S2), were also produced
in vitro (Fig. 4A and fig. S20). These doubly
charged CP dimers were only produced when
the quantities of the (Z)-3-hexenal substrate
were lower than those of the CP substrate
in vitro. NaPPO1/2 showed little-to-no activ-
ity when incubated with CP and (E)-2-hexenal
(Fig.4A).NaBBL2alonecouldnotuseCP
and (Z)-3-hexenal as the substrates to pro-ducem/z347.19, and under in vitro conditions,
the addition of NaBBL2 in the presence of
NaPPO1/2, CP, and (Z)-3-hexenal did not signif-
icantly increase the production ofm/z347.19
(figs. S21 and S22 and table. S2). As PPOs have
broad substrate specificities that can accept both
hydroxy benzenes and/or ortho-dihydroxylated
benzenes as substrates ( 28 ), we further ex-
plored the substrate specificities of NaPPO1
and NaPPO2 by incubating NaPPO1/2 and
(Z)-3-hexenal with CoP or with chlorogenic
acid (CGA), which contains the same aromatic
dihydroxylation pattern as CP. However, no
new products were found, which indicates that
CoP and CGA are not accepted as substrates by
NaPPO1/2 (fig. S23). Together, these data reveal
that NaPPO1/2 accepts CP and (Z)-3-hexenal as
substrates to producem/z347.19 in vitro.Biosynthetic logic of the reactive
m/z347.19 chemistry
To elucidate the chemical structure ofm/z
347.19, we attempted to isolate and purify
them/z347.19 using inducedN. attenuataleaf
material and enzyme assay–derived products.
However, several attempts failed because of the
instability ofm/z347.19. Although relatively
stable in ammonium-acetate buffer (pH 4.8),
when concentrated either by rotatory evap-
oration or freeze drying,m/z347.19 rapidly
decomposed (table S1). These observations
indicate thatm/z347.19 is reactive and un-
stable at high pH. We modified the purification
procedures form/z347.19 to produce large
quantities from enzymatic assays under weak
acidic conditions and purifiedm/z347.19 using
solid-phase extraction under argon atmo-
spheres. The purifiedm/z347.19 was then sub-
jected to nuclear magnetic resonance (NMR)
analysis, which elucidated its structure as
a CP-5–(Z)-3-hexenal compound (hereafter
referred to as CPH) (fig. S24 and data S3).
CPH’s half-life is only ~22 hours in acidified
methanol-d 3 (0.1% formic acid) at room tem-
perature in darkness (fig. S25 and data S3).
CPH contains both the reactive moiety of an
a,b-unsaturated aldehyde derived from (Z)-3-
hexenal, which is electrophilic, and an amine
feature of CP, which is nucleophilic.
CPH results from the biochemical union of
so-called direct (CP) and indirect [(Z)-3-hexenal]
defense metabolism, and we hypothesized
that CPH was the metabolic trait underlying
Empoascanonhost resistance. We suggest
two possible mechanisms of action: The rapid
polymerization of the electrophilic and nucle-
ophilic groups could occlude the mouthparts
of probingEmpoascaleafhoppers, or the
a,b-unsaturated aldehyde may function as
a protein cross-linker that disablesEmpoasca
proteins ( 29 ). We propose a three-step biosyn-
thetic mechanism for the production of CPH—
NaPPO1/2 oxidizes CP to the corresponding
caffeoyl quinone derivative and activates (Z)-3-Baiet al.,Science 375 , eabm2948 (2022) 4 February 2022 6of9
RESEARCH | RESEARCH ARTICLE