Induced production of specialized steroids by transcriptional reprogramming in Petunia hybrida

Abstract Plants produce specialized metabolites with defensive properties that are often synthesized through the coordinated regulation of metabolic genes by transcription factors in various biological contexts. In this study, we investigated the regulatory function of the transcription factor PhERF1 from petunia (Petunia hybrida), which belongs to a small group of ETHYLENE RESPONSE FACTOR (ERF) family members that regulate the biosynthesis of bioactive alkaloids and terpenoids in various plant lineages. We examined the effects of transiently overexpressing PhERF1 in petunia leaves on the transcriptome and metabolome, demonstrating the production of a class of specialized steroids, petuniolides, and petuniasterones in these leaves. We also observed the activation of many metabolic genes, including those involved in sterol biosynthesis, as well as clustered genes that encode new metabolic enzymes, such as cytochrome P450 oxidoreductases, 2-oxoglutarate-dependent dioxygenases, and BAHD acyltransferases. Furthermore, we determined that PhERF1 transcriptionally induces downstream metabolic genes by recognizing specific cis-regulatory elements in their promoters. This study highlights the potential of evolutionarily conserved transcriptional regulators to induce the production of specialized products through transcriptional reprogramming.


Introduction
Plants have a tremendous ability to produce and accumulate a wide variety of structurally complex and diverse specialized metabolites, also known as secondary metabolites, including bioactive alkaloids and terpenoids (1).These compounds are generally specific to species or groups of taxonomically related species and are present in limited quantities within plant tissues.They often confer adaptive advantages to the producing plants, especially in fluctuating environmental conditions.Humans have long exploited plant-derived natural products as medicines, dyes, perfumes, and industrial materials.It is important to develop plant and other biological resources to ensure a stable supply of these valuable chemicals.
Recent developments in molecular and genomics approaches have greatly facilitated the discovery and functional characterization of genes encoding metabolic enzymes involved in the biosynthesis of specialized metabolites in plants (2).In certain instances, such as with morphine, tropane alkaloids, and monoterpenoid indole alkaloids, an entire series of metabolic genes have been defined and exploited to reconstitute the biosynthesis pathway in heterogeneous microorganisms (3)(4)(5).However, to employ this synthetic approach successfully, it is critical to identify many metabolic genes that form an entire biosynthesis pathway.Conventional gene discovery, which largely relies on structural homology and expression profiles, has limitations, and new approaches are required.
To ensure metabolic flow through multiple enzymatic steps and rapid metabolic adjustment in response to developmental and environmental stimuli, the expression of metabolic genes in certain pathways is often coordinated by transcription factors (6,7).These factors positively and/or negatively regulate transcription of downstream target genes by specifically recognizing cis-regulatory elements in their promoters.Natural variation in and genetic manipulation of genes encoding such transcriptional regulators often leads to substantial metabolic alterations (8,9), as exemplified by anthocyanin-based color diversification in flowers and fruits (10)(11)(12) and the elimination of toxic substances in major crops (13)(14)(15).In contrast to microbial production mentioned above, this metabolic engineering approach, which relies on transcriptional regulation, does not necessarily require much knowledge of the downstream metabolic steps.
The plant-specific APETALA2/ETHYLENE RESPONSE FACTOR (AP2/ERF) family is one of the largest transcription factor families, with over 100 members in each species (16).A small group of ERFs from clade II of group IXa (16,17) are known transcriptional regulators in diverse specialized pathways from different plant lineages (18).These ERFs include CrORCAs (Octadecanoidderivative-responsive Catharanthus AP2-domain), which regulate monoterpenoid indole alkaloids in Madagascar periwinkle (Catharanthus roseus) (19,20); the homologous pair NtERF189 and NtERF199, which regulate nicotine biosynthesis in tobacco (Nicotiana tabacum) (21,22); SlJRE4 (jasmonate-responsive ERF 4), which regulates steroidal glycoalkaloids in tomato (Solanum lycopersicum) (23,24); and AaORA (Octadecanoid-responsive AP2), which regulates artemisinin in sweet wormwood (Artemisia annua) (25).The expression of the genes encoding the reported ERFs in this group and their downstream pathways are commonly induced by jasmonates (18), a class of fatty acid-derived phytohormones that play central roles in plant defense against herbivores and pathogens (26).Most ERF genes are found in tandem clusters in plant genomes (18).Clusters of 10, 5, and 5 ERF genes have been reported in the tobacco (27), tomato (24), and C. roseus (20) genomes, respectively.As genome sequences have become available in many plants, related clustered ERF genes have been discovered in many eudicots (18).It remains unclear whether these newly identified ERF genes also play regulatory roles in defense metabolism or other pathways in their respective species.
Petunia (Petunia hybrida) is an ornamental flowering species in the family Solanaceae, which produces and accumulates specialized compounds, such as steroids (28) and acylsugars (29), with protective properties.In this study, we identified a group of clustered genes in the petunia genome, named PhERF genes, which are related to NtERF189 and SlJRE4.We investigated the in planta function of PhERF1 as a transcriptional regulator in the biosynthesis of specialized steroids, such as petuniolides and petuniasterones, by analyzing the effects of transient PhERF1 overexpression in petunia leaves on the transcriptome and metabolome.Importantly, we had no a priori knowledge of the pathway under the control of this ERF, but we elucidated the entire biosynthetic pathway relying on this ERF using a multiomics approach.Our findings provide another example of ERFs regulating defense metabolism in plants and demonstrate a promising strategy for enhancing the production of bioactive specialized metabolites by reprogramming transcription networks using evolutionarily conserved transcription factors.
We examined PhERF expression levels in petunia leaves by transcriptome deep sequencing (RNA-seq) and plotted the data as heatmaps (Fig. 1B; Table S1).PhERF1, PhERF2, and their homoeologs PhERF1L1 and PhERF2L1 showed high expression levels in leaves, whereas the other members had limited or no expression (Fig. 1B; Table S1).To investigate the function of PhERF1, we transiently overexpressed it in petunia leaves using Agrobacterium-mediated transformation with a binary vector containing the full-length coding region of PhERF1 under the control of the constitutive cauliflower mosaic virus (CaMV) 35S promoter and conducted an RNA-seq analysis (Data Sets S1 and S2).We detected much higher transcript levels for PhERF1 and its close homologs, PhERF1L1 and PhERF1L2, but not for other PhERF genes (Fig. 1B; Table S1), indicating that PhERF1 does not regulate PhERF genes in general.Please note the possibility of short-read misalignment for PhERF1 with its closely related counterparts.Furthermore, we evaluated the response of petunia plants to methyl jasmonate (MeJA) exposure by RNA-seq analysis (Data Sets S3 and S4).PhERF10 was significantly induced by MeJA, and its expression was predominant in MeJA-treated leaves, whereas MeJA did not significantly affect the expression of the other PhERF genes (Fig. 1B; Table S1).
The PhERF genes appeared to be present in the petunia genome as a pair of homologous clusters, consisting of 15 PhERF genes represented by 2 scaffolds in the P. axillaris genome and 3 scaffolds in the P. inflata genome (Fig. 1C).PhERF1, PhERF2, PhERF4, PhERF9, and PhERF9L1 were not in these clusters.We observed partial synteny and similar gene arrangements among the ERF gene clusters in the petunia, tobacco, and tomato genomes, likely reflecting their common ancestry and diversification (Fig. 1C).Since PhERF1 belongs to the same clade as NtERF189 and SlJRE4 (Fig. 1A) and PhERF1 is highly expressed in petunia leaves (Fig. 1B), we chose this gene for characterization.

PhERF1 up-regulates metabolic genes involved in steroid biosynthesis
To explore the downstream genes regulated by PhERF1, we identified differentially expressed genes (DEGs) (Tables S2 and S3) in our RNA-seq data sets (Data Sets S1 and S2) that were up-regulated (log 2 -transformed fold-change [FC] > 1) or down-regulated (log 2 --transformed FC < −1) by PhERF1 overexpression.Gene Ontology (GO) analysis revealed an enrichment for terms related to sterol or steroid metabolism in the up-regulated DEGs (Table S4).
To better annotate the relevant biosynthetic pathways in petunia, we performed a BLASTP search using a set of tomato proteins (Table S5) as queries to identify petunia proteins involved in sterol metabolism.We annotated 117 genes from petunia involved in sterol metabolism and related pathways from acetyl-CoA to cholesterol, phytosterols, and specialized steroids (Table S6).The common upstream pathway splits into multiple downstream branches after the common intermediate cycloartenol (Fig. 2A).Fig. 1.Identification of petunia ERF transcription factors related to tomato JREs and tobacco ERF189.A) Phylogenetic tree of ERFs from various plant species.PhERFs from petunia (Ph; P. hybrida) predicted from the P. axillaris (v.1.6.2) and P. inflata (v.1.0.1)genomes; SlJRE1 (Solyc01g090300), SlJRE2 (Solyc01g090310), SlJRE3 (Solyc01g090320), SlJRE4 (Solyc01g090340), SlJRE5 (Solyc01g090370), and SlJRE6 (Solyc05g050790) from tomato (Sl; S. lycopersicum, ITAG release 2.40); NtERF189 (GenBank; AB827951) from tobacco (Nt; N. tabacum); CrORCA2 (GenBank; AJ238740), CrORCA3 (GenBank; AJ251250), CrORCA4 (GenBank; MT414982), CrORCA5 (GenBank; MT414981), and CrORCA6 (GenBank; MT414983) from Madagascar periwinkle (Cr; C. roseus); AaORA (GenBank; JQ797708) from sweet wormwood (Aa; A. annua); and AtERF13 (At2g44840) from Arabidopsis (At; A. thaliana) are included.ERFs, which belong to one clade with a branch point marked with a circle, are less conserved among tobacco, tomato, and petunia (see text).The percentage support from 1,050 bootstrap replicates is indicated at branch nodes.The scale bar indicates the number of amino acid substitutions per site.PhERFs predicted from the P. inflata genome are marked with asterisks.B) Heatmap illustration of PhERF expression levels in petunia leaves.Normalized expression values obtained by RNA-seq analysis in petunia leaves transiently overexpressing PhERF1 (left) or treated with MeJA (right).Genes with significant changes in expression upon PhERF1 overexpression or MeJA treatment (log 2 -transformed FC ≥ 1 or ≤ −1 and adjusted P-value < 0.05) are marked with asterisks.PhERFs predicted from the P. inflata genome are marked with asterisks.Details are available in Table S1.C) Clusters of ERF genes in tobacco, tomato, and petunia genomes.Petunia has a homologous pair of presumptive clusters, which are represented with five scaffolds from the P. axillaris and P. inflata genomes.Positions and orientations of ERF genes are shown as arrows.White arrows represent non-ERF genes, which are present near the clusters and conserved among species.Numbers of non-ERF genes present between the arrowheads are shown in parentheses.Homologous genes are linked with dashed lines.
Based on phylogenetic analysis with their tomato counterparts (31), we defined petunia metabolic proteins specific to each branch and those shared between branches (Fig. 2A; Fig. S1).We noticed that enzymes mediating most of the steps in the upstream part of the pathway before cycloartenol are encoded by multiple pairs of genes (Fig. 2A; Table S6), possibly hinting at complex regulation to meet metabolic demands for sustaining metabolite flow to multiple downstream branches.
We determined that PhERF1 induces one pair of homoeolog genes encoding either hydroxymethylglutaryl-CoA reductase (HMGR), squalene monooxygenase (SQO), or cycloartenol synthase (CAS) while down-regulating another pair of HMGR genes (Fig. 2A and B; Tables S2 and S3).In the downstream portion of the pathway, both pairs of homoeolog genes encoding sterol Δ24-isomerase (24ISO) were up-regulated by PhERF1 (Fig. 2A and  C, Table S2).In several species of the family Solanaceae, 24ISO was identified as a key enzyme that catalyzes the conversion of 24-methylenecholesterol to 24-methyldesmosterol, which is a precursor of certain specialized steroids, such as withanolides in Withania species (32) and physalins in Physalis species (33).Thus, the PhERF1-mediated up-regulation of 24ISOs implies the involvement of this ERF factor in transcriptional activation of a biosynthetic pathway leading to 24-methyldesmosterol-derived steroidal compounds in petunia like in other Solanaceae species.In line with this idea, with the exception of STEROL METHYL OXIDASE 4 (SMO4) genes thought to be specific for cholesterol biosynthesis (Fig. 2A) (31), PhERF1 up-regulated the expression of metabolic genes encoding the enzymes sterol methyltransferase 1 (SMT1), sterol methyl oxidase 1 (SMO1), ERG28-like protein (ERG28), cyclopropylsterol isomerase (CPI), sterol C14-demethylase (CYP51G), sterol C-5 desaturase/Dwarf 7-1 (DWF7-1), and sterol reductase (also known as DWF5).Notably, all these enzymes, whether specific to one branch or shared between branches, belong to the biosynthetic route contributing to specialized steroids (Fig. 2A and B; Table S2) (31).
We detected 100 up-regulated DEGs, including all 4 24ISO genes, and 5 down-regulated DEGs, which together formed 7 different types of 11 presumptive clusters, including 8 clusters containing 4 homoeologous pairs (Fig. 3).Most of the clustered DEGs and 25 homologous DEGs not in clusters encoded metabolic enzymes not assigned to specific pathways, belonging to the cytochrome P450 (CYP) (34), 2-oxoglutarate-dependent dioxygenase (DOX) (35), or BAHD acyltransferase (AT) (36) families (Table S2).We reconstructed phylogenetic trees using these CYPs, DOXs, and ATs from petunia and their homologs from petunia and tomato, which allowed us to classify each member into a few subgroups of each family (Figs.S2-S4).Related genes were usually located in close proximity in the petunia genome, suggesting cluster formation through tandem gene duplications (Fig. 3).

Induction of metabolic genes by PhERF1 and related proteins
We investigated the functional similarities and differences between PhERF1 and related proteins as transcriptional regulators.To this end, we delivered constructs encoding each full-length ERF protein fused to a 4× Myc tag at the C-terminal end via Agrobacterium-mediated infiltration of petunia leaves.To assess the transactivation of downstream genes by each introduced PhERF, we measured the expression levels of three PhERF1regulated metabolic genes, PhCAS1, Ph24ISO2, and PhCYP88C13.1, by reverse transcription-quantitative PCR (RT-qPCR) analysis (Fig. 4A).As transcriptional output depends on the abundance of the transcription factor, we estimated the relative accumulation of each Myc-tagged PhERF in the leaf extracts by immunoblot analysis (Fig. 4B) and normalized transcript levels by the measured protein abundance.We observed that the expression of PhCAS1, Ph24ISO2, and PhCYP88C13.1 was significantly induced, to similar degrees, upon overexpression of PhERF1 or PhERF2, indicating that these two PhERFs have comparable transcriptional activation potential (Fig. 4A).In contrast, overexpression of PhERF5 or PhERF7 did not significantly affect the transcript levels of the examined metabolic genes compared with the empty vector (EV) controls (Fig. 4A).We also determined that SlJRE4 and NtERF189 significantly induce the expression of PhCAS1 (in the case of SlJRE4) and Ph24ISO2 (in the case of SlJRE4 and NtERF189), by 6.2-to 11-fold (Fig. 4A).These results indicate that the ERFs from tomato and tobacco have transactivation activities in petunia, but their activities are weaker than those of PhERF1 and PhERF2.

PhERF1-mediated transcriptional regulation of metabolic genes through promoter regions
We asked whether the increased transcript levels of downstream metabolic genes measured by RT-qPCR were caused by direct transcription control by PhERF1 binding to promoter regions.Using the known binding site of NtERF189 (a GC-rich 10-nt element ( 38)) as a starting point, we computationally predicted putative cis-regulatory elements (E1 to E6) recognized by PhERF1 in the promoter sequences of PhCAS1, PhSMO1.2,Ph24ISO2, and PhCYP88C13.1 (Fig. 5A).E1 and E5 were indeed present as multiple overlapping sequences in these promoters.To test the direct transcriptional activation of the above promoters by PhERF1, we conducted a transient expression assay in Nicotiana benthamiana leaves via Agrobacterium-mediated infiltration using the β-glucuronidase (GUS) reporter gene placed under the control of the promoter sequences of PhCAS1 (−1,771 to −1; counted from the first ATG), PhSMO1.2 (−1,515 to −1), Ph24ISO2 (−2,538 to −1), and PhCYP88C13.1 (−1,502 to −1).We coinfiltrated each GUS reporter construct with EV (CK, as control) or with a construct overexpressing PhERF1; 2 days after infiltration, we harvested samples for RT-qPCR analysis of GUS and PhERF1 relative transcript levels.Compared with the EV controls, relative GUS transcript levels were elevated 7.4 to 11 times upon overexpression of PhERF1 (Fig. 5B), suggesting that PhERF1 mediates the transcriptional activation of the promoter fragments.
To determine whether the predicted cis-regulatory elements were required for PhERF1-dependent reporter activation, we mutated highly conserved GC dinucleotides in the elements (17) by PCR-based mutagenesis.We then placed GUS under the control of these mutated promoter fragments and repeated the transient transactivation assays.PhERF1 failed to induce the accumulation of GUS transcripts when the reporter gene was driven by the mutated PhSMO1.2 or PhCYP88C13.1 promoter, which each have one predicted cis element (Fig. 5A and B), suggesting that they are required for activation.As the PhCAS1 promoter harbors two elements, E1 and E2 (Fig. 5A), we mutated them individually and together.The mutated versions of the PhCAS1 promoter with mutated E1 (mE1) or mE2 showed a significant decrease in relative GUS transcript levels, although PhERF1 still induced GUS transcription from these mutated promoters.The mutation of both elements (mE1/2) abolished PhERF1-mediated transcriptional activation of the GUS reporter from the PhCAS1 promoter (Fig. 5B), suggesting that both elements are necessary for promoter activation.The Ph24ISO promoter contains two elements, E4 and E5 (Fig. 5A).We detected a complete loss of PhERF1-mediated  S6) were predicted based on similarities to their tomato counterparts (Table S5).The genes are represented as boxes placed beside enzyme abbreviations (see Table S5); red and blue boxes represent genes up-regulated and down-regulated by PhERF1, respectively.Downstream of cycloartenol, cholesterol-specific genes are shown in blue; phytosterol-specific genes are shown in red, and genes involved in both pathways are in purple.Petunia-specific steroids, such as petuniolides and petuniasterones, are presumed to be produced through a downstream branch that starts with a reaction mediated by sterol 24ISO.B) and C) Heatmap visualization of the expression levels of PhERF1-regulated genes shown in A and genes present in clusters Ai, Bi, Ci, and Da (see Fig. 3).24ISO genes are included in C. Please refer to Table S5 or Fig. 3 for gene names and labels.Normalized expression values obtained in leaves transiently overexpressing PhERF1 are shown.PhERF1-down-regulated genes are marked with asterisks.activation of GUS transcription following mutation of E4, but not E5 (Fig. 5B), indicating that PhERF1 mediates Ph24ISO promoter activation by recognizing the promoter element E4.A consensus of the cis-regulatory elements (E1, E2, E3, E5, and E6), which have been validated experimentally, is shown as a sequence logo (Fig. 5C).

Effects of PhERF1 overexpression on steroid metabolism
Given that PhERF1 up-regulates metabolic genes contributing to specialized steroid biosynthesis (Fig. 2A; Table S2), we examined the influence of PhERF1 overexpression on the accumulation of sterol and related steroidal compounds in petunia leaves.To allow the accumulation of metabolites, we harvested petunia leaves 8 days after Agrobacterium-mediated infiltration of the 35S:PhERF1 construct.We extracted metabolites and subjected them to gas chromatography-mass spectrometry (GC-MS) and liquid chromatography-mass spectrometry (LC-MS) analyses to measure sterols and specialized steroids, respectively.Specifically, we measured the relative levels of cholesterol, phytosterols (stigmasterol, β-sitosterol, and campesterol), 24methylenecholesterol, and 24-methyldesmosterol in leaf samples using targeted GC-MS analysis.24-Methyldesmosterol levels were significantly lower, by 46%, whereas campesterol accumulation increased by 40% with PhERF1 overexpression compared with EV controls (Fig. 6A and B, Table S7).These changes in sterol levels were in line with those seen for specialized steroids (see below), suggesting an increased metabolic flow to a branch leading to campesterol and specialized steroids and consumption of 24-methyldesmosterol, a product of 24ISO catalysis.This consumption was consistent with the increased expression of downstream genes specific to specialized steroid branches (see Fig. 2A).
We performed LC-MS analysis to detect the specialized steroids petuniolides and petuniasterones.Total ion chromatograms (TICs) revealed many higher peaks and new peaks in extracts obtained from petunia leaves overexpressing PhERF1 relative to the controls (Fig. 6C, Data Set S5), suggesting a PhERF1-mediated metabolic elicitation.We selected peaks based on similarity (with a tolerance of 0.01) to reported parental ion m/z values (values reported in (28) and values for metabolites, which were extracted with the keyword Petunia in Biological Source, in the Dictionary of Natural Products [http://dnp.chemnetbase.com])as well as average intensities of >0.2 in leaves overexpressing PhERF1 and FC of >10-fold among the samples.We then manually identified these peaks using MS/MS spectra (28) (Fig. S5), since authentic steroid standards were not available.Accordingly, we identified 5 peaks as 3 petuniolide derivatives and 11 peaks as 4 petuniasterone derivatives (Data Set S6).Two of the selected peaks were not assigned to any known metabolites (Data Set S6).Because individual peaks for petuniolide A, 12α-acetoxypetuniasterone D 7-acetate, and petuniolide E were difficult to ascertain in Fig. 6C, we show their extracted ion chromatograms (EICs) (Fig. 6D).We used the area of peaks (marked with asterisks in Fig. 6C) with the highest average values among those for each metabolite to calculate relative accumulation levels (Fig. 6E, Data Set S6).We observed significant increases (6.6-to 67-fold) upon PhERF1 overexpression for all seven steroidal metabolites (Fig. 6E), indicating that PhERF1 overexpression induces the accumulation of a group of specialized products.

PhERF1 overexpression broadly affects metabolism
To determine whether the PhERF1-mediated transcriptional changes and associated induction of specialized steroids affected primary and secondary metabolism in petunia leaves, we performed comprehensive metabolite profiling by GC-MS.We detected 203 peaks, including 96 identified or annotated metabolites in control leaves and leaves overexpressing PhERF1.We detected significant changes of over 50% between the samples only for campesterol, stigmasterol, methionine, homoserine, quinic acid, cis-3-O-caffeoyl quinic acid, and raffinose among the 96 known metabolites (Fig. 7, Data Set S7).Because of differences between extraction procedures, only free stigmasterol was quantified in Fig. 7, and both free and conjugated forms of the sterol were collectively measured in Fig. 6.We observed a significant decrease for free stigmasterol (Fig. 7) but not for the total forms of the sterol (Fig. 6), when PhERF1 was overexpressed, suggesting an increase of stigmasterol conjugates due to this overexpression.

Response of PhERF and metabolic genes to jasmonates
Jasmonate-induced gene expression is a characteristic feature of most homologs of PhERF genes in other species (18, 21, 24, 40) but not of most PhERF genes (as shown in Fig. 1B).To investigate the response of PhERF1-regulated metabolic genes to jasmonates in petunia leaves, we treated petunia plants with MeJA and examined the expression levels of these genes 24 h later by RNA-seq analysis.We selected DEGs that were up-regulated (over 2-fold) or down-regulated (below 0.5-fold) by the phytohormone treatment (Data Sets S3 and S4).We established that although a significant fraction of the PhERF1-regulated DEGs are induced by MeJA (Fisher's exact test: P < 0.001), many PhERF1-regulated DEGs were unaffected (Table S2, Fig. S6).
To better understand the underlying response, we performed a time-course analysis of the response to MeJA for selected PhERF and downstream metabolic genes.Accordingly, we measured relative transcript levels using RT-qPCR analysis following MeJA treatment for 1 to 24 h.We observed that MeJA did not induce PhERF1 expression but significantly increased the expression of PhERF7 and PhERF10, with maximum FC of 8.9-and 4.3-fold, respectively (Fig. S7).We also observed variable and noncoordinated increases in the expression of PhSMO1.2,PhCYP88C13.1, and Ph24ISO2 in response to MeJA, whereas the expression of PhCAS1 was not affected by the phytohormone (Fig. S7).These results suggest that many, but not all, PhERF1-regulated metabolic genes are positively regulated but not necessarily in a coordinated manner in terms of timing or magnitude.Relative expression levels in the controls were set to 1.For each reporter vector, significant differences relative to the controls were determined by Student's t-test; **P < 0.01, *P < 0.05.ns, not significant.C) Sequence logo of the consensus sequence for the cis-regulatory elements, E1, E2, E3, E4, and E6, which have been validated experimentally in B. One hundred hypothetical sequences representing the elements were used to create a logo by WebLogo (39).Sequences with the highest scores were used as representatives for E1 and E5.

Discussion
The overexpression of PhERF1 in petunia leaves led to the transcriptional reprogramming of many sterol biosynthesis and uncharacterized metabolic genes (Fig. 2, Table S2), resulting in the production of a group of specialized steroids, the petuniolides and petuniasterones (Fig. 6).Petuniasterones are characterized by the presence of a ketone group in the A-ring, whereas petuniolides have a spirolactone A-ring structure (41).These ergostanetype steroids have protective properties and may play a crucial role in plant defense (42).Petuniolides have been proposed to act on γ-aminobutyric acid (GABA) receptors in the central nervous system (43) and are generally more toxic to a wide range of insect larvae than petuniasterones (42).The overexpression of PhERF1 significantly increased the accumulation of three petuniolide derivatives and four petuniasterone derivatives (Fig. 6C-E).Although they did not all reach the thresholds for our selection criteria, we observed an increase in many peaks by LC-MS analysis upon PhERF1 overexpression (Fig. 6C, Data Set S5).These peaks may correspond to derivatives of known or unknown steroidal compounds.Further analysis will be needed to determine the exact identities and structures of these unknown peaks (Fig. 6C, Data Set S6) to fully understand their potential role.Fig. 6.Effects of PhERF1 overexpression on steroid accumulation in petunia leaves.PhERF1 was transiently overexpressed via Agrobacterium-mediated infiltration in petunia leaves under the control of the 35S promoter (OX).The pBI121 vector was used as the EV control.Leaves were harvested for metabolite analysis 8 days after infiltration.Relative contents of each metabolite were calculated based on area values of the corresponding peaks relative to those of internal standards.Values are means ± SD of biological replicates (n = 8 for control, n = 18 for PhERF1 OX).Significant differences relative to the controls were determined by Student's t-test.**P < 0.01; ns, not significant.A) Relative levels of cholesterol, campesterol, stigmasterol, β-sitosterol, 24-methylenecholesterol, and 24-methyldesmosterol, as determined by GC-MS analysis.B) TICs of GC-MS analysis.C) TICs of LC-MS analysis.Peaks that were used to calculate the levels shown in E are marked by asterisks.D) EICs of three steroidal compounds.E) Relative levels of petuniolide and petuniasterone derivatives.The area of peaks with highest average intensities for each metabolite was used to calculate their levels.

Shoji et al. | 9
PhERF1 overexpression induced the expression of a series of sterol biosynthesis genes, including 24ISO genes, which contribute to the production of these specialized steroids (Fig. 2, Table S2).
The pattern of activated genes in petunia is clearly different from that in tomato, where SlJRE4 activates metabolic genes involved in the cholesterol pathway contributing to steroidal Fig. 7. Effects of PhERF1 overexpression on a wide range of metabolites in petunia leaves.PhERF1 was transiently overexpressed via Agrobacterium-mediated infiltration in petunia leaves under the control of the 35S promoter (OX).The pBI121 vector was used as EV control.Leaves were harvested for comprehensive metabolite analysis 8 days after infiltration.Relative contents for each metabolite were calculated based on area values of the corresponding peaks relative to those of internal standards.A) Metabolic map.All metabolites detected in this analysis are marked by asterisks.Metabolites whose abundance significantly increased or decreased in PhERF1 OX samples (>1.5-fold) relative to the controls are shown in red and blue, respectively.Significant differences were determined by Student's t-test at P = 0.05.B) Contents of the increased or decreased metabolites.FC are indicated above the black bars.Values are means ± SD values of biological replicates (n = 8 for control, n = 18 for PhERF1 OX).Significant differences relative to the controls were determined by Student's t-test.**P < 0.01, *P < 0.05.glycoalkaloid production (23,24,44), although the sterol and cholesterol pathways are related (Fig. 2A) (31).In this regard, the activity of SMO4 from petunia needs to be experimentally verified, as its tomato homolog suggests that its activity is specific to the cholesterol pathway (31), which may not be consistent with our scenario (Fig. 2A).Furthermore, PhERF1 up-regulated many previously underannotated metabolic genes from the CYP, DOX, and AT families, most forming clusters composed of multiple nonhomologous metabolic genes in the petunia genome (Fig. 3, Table S2).One homologous pair of these clusters (cluster A in Fig. 3) that included 24ISO, CYP87, and CYP88 genes was reported in a previous study (45).Currently, limited information is available regarding the metabolic pathway downstream of the 24ISO-catalyzed step (45).These downstream steps are responsible for elaborating the steroidal backbone through extensive oxidation and other modifications.It is possible that PhERF1-regulated metabolic genes from the CYP, DOX, and AT families participate in these modification steps, although further research is necessary to elucidate the specific underlying metabolic details.A comparative approach assessing related steroids from a range of plants in the Solanaceae family, such as withanolides from Withania species (32) and physalins from Physalis species (33), would be beneficial for understanding how different metabolic enzymes contribute to the production of a diverse group of related steroidal compounds.It would be intriguing to investigate whether metabolic gene clusters analogous to those found in petunia exist in the species that produce related steroids.If so, clustering would allow us to pinpoint the relevant genes involved in the pathways.
We hypothesize that the metabolic effects of PhERF1 overexpression are primarily related to the production of specialized steroids.This assumption is supported by the enrichment of DEGs induced by PhERF1 in pathways related to the biosynthesis of steroidal compounds, rather than in other metabolic pathways (Tables S2 and S4).Comprehensive metabolite profiling further supports this idea, as few metabolites not related to the sterol biosynthesis pathway were significantly affected in their abundance by PhERF1 overexpression (Fig. 7).A substantial portion of the alterations observed in the levels of nonsterol metabolites could potentially result from indirect effects stemming from the redirection of metabolic flux caused by the ongoing metabolic changes, rather than being directly attributed to the PhERF1 transcription factor.The significant 7.8-fold increase of methionine (Fig. 7B) cannot be attributed to changes in the expression of genes directly involved in amino acid accumulation (Table S2).Similarly, overexpression and mutational loss of NtERF189 in tobacco have significant effects on the accumulation of metabolites not involved in the nicotine biosynthetic pathway (13).The metabolomic changes detected in tobacco do not necessarily overlap with those induced by PhERF1 overexpression in petunia.For example, methionine was not significantly affected by transient overexpression of NtERF189 in N. benthamiana leaves (13).The distinct metabolic changes induced by related ERFs in petunia and tobacco (13) suggest different mechanisms of metabolic regulation.A large, sustained change in the level of even a single amino acid may adversely affect plant growth and development and therefore may only be tolerable in a transient expression system.This notion underscores the potential of transient systems over the generation of conventional stable transgenic lines for economically viable plant-based bioproduction of specialized products.
Small groups of evolutionarily conserved ERF genes often form clusters in plant genomes (18).Unlike clusters of nonhomologous metabolic genes (46), ERF clusters consist of multiple homologous genes arranged in tandem, potentially resulting from repeated gene duplication.In ERF clusters from Solanaceae species, ERF genes at the outer edge of the clusters are relatively well conserved, whereas those in the central region are more variable among species (Fig. 1A and C).This pattern suggests that these clusters expanded in distinct lineages primarily through the addition of genes at the centers of clusters.PhERF1, PhERF2, and their homoeologs, specific to the petunia lineage, are considered the petunia counterparts of NtERF189 and SlJRE4, based on their genomic place within the central variable region of the clusters and their high expression (Fig. 1B and C).This notion is supported by the fact that PhERF1 and PhERF2 have higher transactivation activity than related regulators.The comparable expression levels (Fig. 1B) and transactivation activities (Fig. 4) of PhERF1 and PhERF2 suggest their functional similarity as regulators of specialized steroids, with potential redundancy as with the multiple CrORCAs regulating alkaloid biosynthesis in C. roseus (20).
In contrast to the reported homologs from other species, the expression of most PhERF genes, including PhERF1 and PhERF2, is not induced by jasmonates (Fig. 1B, Fig. S7).Although MeJA regulated a fraction of PhERF1-regulated metabolic genes (Fig. S6), we did not detect coordinated induction of gene expression (Fig. S7), supporting the idea that PhERF1 may not play a significant role in phytohormonal regulation.Further investigation will be required to determine whether and how the defense-related phytohormone regulates the biosynthesis of bioactive steroids in Petunia species, possibly through a PhERF1-independent mechanism.The differences in response to jasmonates among ERF regulators from different species suggest a regulatory change during speciation.
Our study shows that overexpressing the transcriptional regulator gene PhERF1 in petunia activated sterol biosynthesis and the expression of several mostly clustered metabolic genes, leading to the production of defense-related steroids.This approach of using evolutionarily conserved factors as genetic tools to stimulate the production of bioactive natural products, which has previously been proposed and demonstrated (18,47,48), has significant potential for bioproduction purposes, even in the absence of prior knowledge of the metabolic pathways involved.We plan to further examine the regulatory functions of the other ERF genes in the clade to fully test the notion that evolutionary proximity indeed results in the control of similar metabolic pathways.

Plant growth and treatment
Surface-sterilized seeds of P. hybrida cv.Mitchell diploid (W115) and N. benthamiana were germinated on half-strength Gamborg B5 medium solidified with 0.7% (w/v) agar and supplemented with 3% (w/v) sucrose.Two-week-old seedlings were transferred to soil (JA Nipi Horticulture Soil No. 1; Nippon Hiryo, Gunma, Japan) in pots and grown in a culture room at 25°C in a 16-h light/8-h dark photoperiod.Four-week-old petunia and 6-weekold N. benthamiana plants were used for the experiments.The topmost fully expanded leaves at comparable positions on the plants were employed in each experiment.
The plants were exposed to MeJA gas in 12-L air-tight plastic buckets.A piece of paper towel soaked with 1.2 mL of a volatile liquid form of MeJA (Fujifilm Wako, Osaka, Japan) was placed in the bucket with two individual plants.
The full-length amino acid sequences of the identified proteins were aligned using MUSCLE (52), and an unrooted phylogenetic tree was reconstructed using the neighbor-jointing algorithm with MEGAX (53) using default settings.
To search for significantly enriched GO terms among upregulated or down-regulated DEGs upon PhERF1 overexpression relative to all genes predicted in either P. axillaris or P. inflata genomes, the Plant Transcriptional Regulatory Map (http:// plantregmap.gao-lab.org/go.php)(54) was used with a P-value threshold set to 0.001.

Agrobacterium-mediated transient overexpression
Several vectors were used, including pGWB2 and pGWB417, for the individual overexpression of PhERF1, PhERF2, PhERF5, PhERF7, SlJRE4 (Solyc01g090340), and NtERF189 (GenBank: AB827951).pGWB5-based promoter reporter vectors were also generated for the PhCAS1, PhSMO1.2,Ph24ISO2, and PhCYP88C13.1 promoter fragments; the control vectors were pBI121 and pBI101.Additionally, the p19 vector was included for all Agrobacterium (Agrobacterium tumefaciens)-mediated infiltrations for silencing suppression (56).The pGWB417-based NtERF189 overexpression vector was generated previously (22).The vectors were introduced into Agrobacterium strain GV3101 by heat shock treatment.The various coding regions were individually transiently heterologously overexpressed in the youngest fully expanded leaves of plants as described (57).Bacterial suspensions of the ERF overexpression vector or the control vector pBI121 and p19 vector were combined at a ratio of 7:3 (v/v), and those for ERF overexpression or the control vector pBI101, promoter reporter vectors, and the p19 vector were mixed at a ratio of 7:7:6 (v/v/v).The leaves were harvested 2 days after infiltration for RNA and protein extraction and 8 days after infiltration for metabolite extraction.

RT-qPCR analysis
RNA isolation, RT, and qPCR were performed as described (21).EF1α (Peaxi162Scf00351g00312) was used as a reference to normalize expression levels.Primer sequences are in Table S8.Since primer specificities were not validated experimentally, the possibility of amplification of closely related sequences cannot be excluded.

Immunoblot analysis
Protein extraction, separation by electrophoresis, transfer to membranes, membrane incubation with antibodies and washes, and chemiluminescence detection were carried out as described (22).An anti-Myc tag polyclonal antibody (PA1-981; Thermo Fisher Scientific) and goat horseradish peroxidase-conjugated anti-rabbit IgG (H + L) (Thermo Fisher Scientific) were diluted at 1:2,000 and 1:10,000 and used as primary and secondary antibodies, respectively.The intensities of the bands corresponding to the Myc-tagged fusion proteins were quantified using ImageQuant TL software (GE Healthcare, Chicago, IL).These intensities were then utilized to normalize the relative expression levels of the PhERF1-regulated metabolic genes, which were determined through RT-qPCR analysis.

Metabolite analysis
Leaf tissues were ground into a fine powder with a mortar and pestle, and metabolites were extracted for LC-MS analysis as described (59).The extracts were analyzed using LC-QTOF-MS (LC, Waters Acquity UPLC system; MS, Waters Xevo G2 Q-Tof; Waters, Milford, MA).LC separation, MS detection, and MS/MS data acquisition were conducted as described (59) with the polarity of electrospray ionization applied in positive ionization mode.

Fig. 2 .
Fig. 2. PhERF1 regulates newly predicted genes involved in sterol biosynthesis.A) Diagram of sterol biosynthesis and related pathways.Petunia biosynthesis genes involved in the pathways (TableS6) were predicted based on similarities to their tomato counterparts (TableS5).The genes are represented as boxes placed beside enzyme abbreviations (see TableS5); red and blue boxes represent genes up-regulated and down-regulated by PhERF1, respectively.Downstream of cycloartenol, cholesterol-specific genes are shown in blue; phytosterol-specific genes are shown in red, and genes involved in both pathways are in purple.Petunia-specific steroids, such as petuniolides and petuniasterones, are presumed to be produced through a downstream branch that starts with a reaction mediated by sterol 24ISO.B) and C) Heatmap visualization of the expression levels of PhERF1-regulated genes shown in A and genes present in clusters Ai, Bi, Ci, and Da (see Fig.3).24ISO genes are included in C. Please refer to TableS5or Fig.3for gene names and labels.Normalized expression values obtained in leaves transiently overexpressing PhERF1 are shown.PhERF1-down-regulated genes are marked with asterisks.

Fig. 3 .
Fig.3.Multiple clusters of PhERF1-regulated genes exist in Petunia genomes.PhERF1-regulated genes (≥4) were considered to form clusters in the genomes when present in tandem with ≤4 nonrelevant genes between any two genes.Positions and orientations of PhERF1-regulated genes (arrows; 100 up-regulated and 5 down-regulated DEGs, whose details are in TablesS2 and S3, respectively) are represented on scaffolds (Scf) from the P. axillaris (v.1.6.2) and P. inflata (v.1.0.1) genomes with labels indicating clusters and positions (numbered from left to right on each scaffold).Five genes down-regulated by PhERF1 are shown with red labels.Numbers of unrelated genes present between the arrows are shown in the brackets.Arrow colors reflect protein classification: AT, BAHD-type acyltransferase; CYP, cytochrome P450; DOX, 2-oxoglutarate-dependent dioxygenase, 24ISO; sterol Δ24-isomerase.Four groups of clusters, A, B, C, and D, each of which is represented by a homologous pair of clusters from P. axillaris and P. inflata (respectively denoted a and i) were presumed in petunia, whereas clusters E, F, and G are each present in only one of the two genomes.Genes closely related between P. axillaris and P. inflata sequences (see Figs. S2-S4) are linked with dashed lines.

Fig. 4 .
Fig. 4. Transient overexpression of PhERF1 and related ERF genes selectively induces metabolic genes in petunia leaves.PhERF1 (Peaxi162Scf00672g00031), PhERF2 (Peaxi162Scf01094g00002), PhERF5 (Peaxi162Scf00475g00077), PhERF7 (Peaxi162Scf00475g00053), SlJRE4 (Solyc01g090340), and NtERF189 (GenBank; AB827951) cloned in-frame and upstream of the sequence encoding a 4× Myc tag in pGWB417 (37) were individually transiently overexpressed in petunia leaves under the control of the CaMV 35S promoter.The pBI121 vector was used as the EV control.A) Relative expression levels of PhCAS1 (Peaxi162Sch00263g00924), Ph24ISO2 (Peaxi162Sch00695g00218), and PhCYP88C13 (Peaxi162Sch00312g00220).Transcript levels were determined by RT-qPCR analysis and normalized to the abundance of the fusion proteins determined in B. The values in the controls (white bars) were set to 1. Values are means ± SD for biological replicates.Different lowercase letters indicate significant differences at P < 0.05, as determined by one-way ANOVA followed by a Tukey-Kramer test.B) Abundance of Myc-tagged ERF fusion proteins, as determined by immunoblot analysis with an anti-Myc antibody.Major detected bands are indicated with asterisks.Intensities of the bands were measured.Two biological replicates (#1 and #2) were analyzed.

Fig. 5 .
Fig. 5. PhERF1 mediates transcriptional regulation of PhCAS1, PhSMO1.2,Ph24ISO2, and PhCYP88C13.1 through their promoter regions.A) Diagram of the PhCAS1 (Peaxi162Sch00263g00924), PhSMO1.2(Peinf101Sch00284g004025), Ph24ISO2 (Peaxi162Sch00695g00218), and PhCYP88C13.1 (Peaxi162Sch00312g00220) promoter regions.Nucleotides were counted from the first ATG.Predicted cis-regulatory elements E1 to E6 are indicated as boxes.Note that E1 and E5 are present in multiple partially overlapping copies, each of which is 10 bp in length.Nucleotide sequences of the wild-type (WT) and mutated elements are shown with their positions and scores to the right; mutated nucleotides are in red.B) Transactivation analysis of WT promoter reporters, PhCAS1pro:GUS, PhSMO1.2pro:GUS,Ph24ISO2pro:GUS, and PhCYP88C13pro:GUS, and their mutant versions (element [E] names denoted with m).The reporter plasmids were introduced into N. benthamiana leaves through Agrobacterium-mediated infiltration with either control (CK) pBI101 or 35S:PhERF1 overexpression vector.Transcript levels of the reporter gene GUS and PhERF1 were determined by RT-qPCR.Values are means ± SD of the biological replications.Relative expression levels in the controls were set to 1.For each reporter vector, significant differences relative to the controls were determined by Student's t-test; **P < 0.01, *P < 0.05.ns, not significant.C) Sequence logo of the consensus sequence for the cis-regulatory elements, E1, E2, E3, E4, and E6, which have been validated experimentally in B. One hundred hypothetical sequences representing the elements were used to create a logo by WebLogo(39).Sequences with the highest scores were used as representatives for E1 and E5.