A chemical complementation approach reveals genes and interactions of flavonoids with other pathways


  • Lucille Pourcel,

    1. Center for Applied Plant Sciences and Department of Molecular Genetics, Ohio State University, Columbus, OH, USA
    Search for more papers by this author
    • Present address: Université de Genève, Département de Botanique et Biologie Végétale, Sciences III, 30 Quai Ernest-Ansermet, 1211 Genève, Switzerland.
  • Niloufer G. Irani,

    1. Center for Applied Plant Sciences and Department of Molecular Genetics, Ohio State University, Columbus, OH, USA
    Search for more papers by this author
    • Present address: Department of Plant Systems Biology, Flanders Institute for Biotechnology, and Department of Plant Biotechnology and Genetics, Ghent University, 9052 Ghent, Belgium.
  • Abraham J. K. Koo,

    1. Department of Energy Plant Research Laboratory, Michigan State University, East Lansing, MI, USA
    Search for more papers by this author
  • Andres Bohorquez-Restrepo,

    1. Center for Applied Plant Sciences and Department of Molecular Genetics, Ohio State University, Columbus, OH, USA
    Search for more papers by this author
  • Gregg A. Howe,

    1. Department of Energy Plant Research Laboratory, Michigan State University, East Lansing, MI, USA
    Search for more papers by this author
  • Erich Grotewold

    Corresponding author
    • Center for Applied Plant Sciences and Department of Molecular Genetics, Ohio State University, Columbus, OH, USA
    Search for more papers by this author

For correspondence (e-mail grotewold1@osu.edu).


In addition to the classical functions of flavonoids in the response to biotic/abiotic stress conditions, these phenolic compounds have been implicated in the modulation of various developmental processes. These findings suggest that flavonoids are more integral components of the plant signaling machinery than traditionally recognized. To understand how flux through the flavonoid pathway affects plant cellular processes, we used wild-type and chalcone isomerase mutant (transparent testa 5, tt5) seedlings grown under anthocyanin inductive conditions, in the presence or absence of the flavonoid intermediate naringenin, the product of the chalcone isomerase enzyme. Because flavonoid biosynthetic genes are expressed under anthocyanin inductive conditions regardless of whether anthocyanins are formed or not, this system provides an excellent opportunity to specifically investigate the molecular changes associated with increased flux through the flavonoid pathway. By assessing genome-wide mRNA accumulation changes in naringenin-treated and untreated tt5 and wild-type seedlings, we identified a flavonoid-responsive gene set associated with cellular trafficking, stress responses and cellular signaling. Jasmonate biosynthetic genes were highly represented among the signaling pathways induced by increased flux through the flavonoid pathway. In contrast to studies showing a role for flavonoids in the control of auxin transport, no effect on auxin-responsive genes was observed. Taken together, our data suggest that Arabidopsis can sense flavonoids as a signal for multiple fundamental cellular processes.


As part of their adaptation to the environment and often in response to biotic or abiotic cues, plants accumulate a large number of small molecules derived from specialized metabolism, which may be broadly classified into various large groups, including phenolics, terpenoids, glucosinolates and alkaloids. In contrast to phytohormones (Santner and Estelle, 2009), such specialized metabolites have not been typically associated with the integration of environmental signals or the plant genetic programs that result in growth and development. Instead, products of specialized metabolism, often referred to as secondary metabolites, are generally considered outputs of growth and developmental pathways, for example providing chemical protection against pathogens (e.g. phytoalexins; Dixon, 2001) or in response to unfavorable environmental conditions (e.g. osmoprotectants; Rontein et al., 2002), and functioning as signals for the interaction of plants with other organisms (e.g. volatiles; Pichersky et al., 2006).

Plant hormone perception often involves a receptor, as shown for auxin, gibberellins, abscisic acid and jasmonoyl-isoleucine [JA-Ile, the bioactive form of jasmonate (JA)] (Santner and Estelle, 2009; Lumba et al., 2010). The hormone–receptor interaction triggers activation of a signal transduction cascade that ultimately results in changes in gene expression, which are often unique for each hormonal pathway (Nemhauser et al., 2006). Nevertheless, considerable integration and coordination between hormone signaling pathways exists, for example by one hormone controlling the synthesis or degradation of another, or by a phytohormone playing an integrator role, as recently proposed for auxins (Jaillais and Chory, 2010). As part of such a model, a specialized metabolite that affects the biosynthesis, transport or activity of the integrating phytohormone may substantially interfere with hormone signal integration, and hence plant development and growth. Such auxin transport-modulating activities have been demonstrated for flavonoids (Jacobs and Rubery, 1988; Murphy et al., 2000; Brown et al., 2001; Muday and Murphy, 2002; Peer et al., 2004; Peer and Murphy, 2007).

Flavonoids are one group of specialized plant compounds for which the biosynthetic pathway has been extensively studied in many plant taxa (Winkel-Shirley, 2001). Flavonoids are phenolic compounds with a C6–C3–C6 carbon structure, derived from the general phenylpropanoid pathway. Flavonoid pigments, such as the anthocyanins or the condensed tannins (proanthocyanidins), participate in important eco-physiological functions (Grotewold, 2006; Lepiniec et al., 2006). In Arabidopsis, genetic and molecular characterization of transparent testa (tt) mutants that display reduced seed coat proanthocyanidin pigmentation has significantly contributed to establishing the architecture of the pathway (Lepiniec et al., 2006). For example, TT4 encodes chalcone synthase (CHS) (Feinbaum and Ausubel, 1988), the first committed step in the flavonoid pathway (Figure 1a). The Arabidopsis chalcone isomerase (CHI) enzyme, which catalyzes the second step in the pathway and is encoded by the TT5 locus (Shirley et al., 1992), utilizes tetrahydroxychalcone as a substrate and converts it to naringenin (Nar). Consequently, the tt5 mutant fails to accumulate naringenin or any of the downstream pathway products, including anthocyanins and flavonols. However, addition of naringenin to tt5 rapidly and robustly complements the accumulation of flavonols and anthocyanins of seedlings grown under conditions of high light and sugar, and low nitrogen, referred to as anthocyanin inductive conditions (AIC) (Poustka et al., 2007; Pourcel et al., 2010). Thus, complementation of the induced pathway by naringenin in tt5 provides a powerful system to investigate how increased flux through the flavonoid pathway affects Arabidopsis gene expression changes.

Figure 1.

Anthocyanin induction system.

(a) Schematics of the anthocyanin pathway. The naringenin induction is indicated, as well as the tt5, tt6, tt3 and fls1 mutations.

(b) Expression of flavonoid genes in tt5-1 and WT seedlings treated with naringenin. Expression of PAP1, PAP2, CHI, CHS, DFR, F3H and FLS1. Transcript accumulation was measured by quantitative real-time PCR (= 3). Asterisks indicate a significant difference between no AIC and AIC (*0.1 >  > 0.05; ** 0.05). AIC, anthocyanin inductive conditions.

The regulation of flavonoid biosynthesis is well understood. As in many other plants (Quattrocchio et al., 2006), a regulatory complex consisting of R2R3-MYB, basic helix-loop-helix and WD-repeat proteins controls Arabidopsis anthocyanin biosynthesis (Ramsay and Glover, 2005). At least four partially redundant factors (PAP1/AtMYB75, PAP2/AtMYB95, AtMYB113 and AtMYB114) provide the R2R3-MYB function (Borevitz et al., 2000; Gonzalez et al., 2008). High sucrose induces the expression of PAP1 (Kranz et al., 1998; Teng et al., 2005) and low nitrogen induces the expression of PAP2 (Lea et al., 2007), suggesting that these two proteins gear most anthocyanin induction under AIC.

Here, we investigated the role of flavonoids as regulators of gene expression, specifically the potential role of flavonoids as signaling molecules. By increasing flux through the flavonoid pathway through feeding of a pathway intermediate (naringenin) that bypasses the absence of CHI function, we identified a flavonoid-responsive gene set associated with cellular trafficking, stress responses and cellular signaling. Our results permit us to hypothesize that flavonoids function as ‘buffering’ molecules to modulate responses induced by biotic or abiotic stress conditions.


Complementation of the tt5 mutant with naringenin reveals novel regulatory mechanisms

To investigate the effects on gene expression of increased flux through the flavonoid pathway by feeding of a pathway intermediate (naringenin) that bypasses the absence of the CHI enzyme, we used the AIC previously described (Poustka et al., 2007; Pourcel et al., 2010). Briefly, after stratification at 4°C, WT (wild-type, ecotype Landsberg erecta) or tt5-1 seeds were germinated for 3 days under high light conditions in plain liquid sucrose medium without a nitrogen source (AIC, see Methods S1), or under the same conditions but in the dark (no AIC). In WT (Figure 1b), PAP1 mRNA steady-state levels were only slightly increased by the AIC. However, expression of PAP2, CHS, CHI and the flavanone 3-hydroxylase (F3H) and flavonol synthase (FLS1) genes was significantly (< 0.05) higher in AIC compared with no AIC (Figure 1b, light versus dark gray). These results suggest that PAP2 is probably the major flavonoid biosynthesis gene regulator under these conditions.

When the experiment was performed on tt5-1 seedlings, a very different situation was observed (Figure 1b). Under AIC, although PAP1 and PAP2 mRNA accumulation was substantially lower than in WT, mRNA accumulation for all the flavonoid genes tested remained at levels comparable to WT (CHS, FLS1 and the dihydroflavonol reductase gene DFR) or higher (F3H). The tt5-1 mutant allele was generated by ionizing radiation and harbors an inversion within the gene (Shirley et al., 1992). Although tt5-1 does not show any CHI enzymatic activity (Shirley et al., 1992), the CHI transcript is present at the same level as in WT in microarray data (data not shown).

It is unclear whether the low levels of PAP1 and PAP2 observed in tt5 are sufficient for sustained flavonoid biosynthetic gene expression, or whether a regulator other than PAP1 and PAP2 (e.g. MYB113 or MYB114; Gonzalez et al., 2008) compensates for the low PAP1/PAP2 mRNA accumulation. However, from the perspective of this study, the relevant observation is that all anthocyanin biosynthetic genes are expressed at high levels under AIC in tt5, despite tt5 not accumulating any anthocyanins. Thus, in sharp contrast with previous studies that investigated gene expression changes induced by the expression of pathway regulators (Bruce et al., 2000; Tohge et al., 2005), AIC-grown tt5-1 seedlings incubated with naringenin provide a powerful system to investigate the consequences on gene expression of increasing flux through the flavonoid pathway.

Naringenin induces genome-wide expression changes

To uncover genome-wide effects of increased flux through the flavonoid pathway, gene expression analyses were performed using Arabidopsis ATH1 GeneChip® Affymetrix microarrays representing 22 810 genes and mRNA obtained from 3-day-old tt5 or WT Arabidopsis seedlings (2000–4000 per plate) grown under AIC and either mock-treated with 1% ethanol (−Nar) or with 100 μm naringenin (+Nar) for 24 h. All microarray experiments were performed using biological duplicates at least. Four possible comparisons were considered for analysis of the results (Figure 2): (i) tt5 −Nar versus tt5 +Nar, (ii) WT −Nar versus WT +Nar, (iii) tt5 −Nar versus WT −Nar, and (iv) tt5 +Nar versus WT +Nar. The first two comparisons are expected to uncover genes with mRNA steady-state levels that are affected by naringenin, pathway intermediates or the (increased) accumulation of flavonoids. The last two comparisons are anticipated to identify effects induced by the absence of the CHI enzyme. Because our main objective for these studies was to uncover gene expression changes induced by flux through the flavonoid pathway, special attention was given to genes represented by the first two. In anticipation that gene expression changes induced by pathway flux or pathway intermediates are not as robust as those induced by over-expression of transcription factors (which combines direct activation of the genes and increased pathway flux), we used less stringent statistical criteria than in previous studies involving regulatory proteins (Morohashi and Grotewold, 2009; Xie et al., 2010).

Figure 2.

Microarray experiments.

Schematic representation of the strategy employed with WT and tt5 and complementation with naringenin: representation of the states of the three components (induction by naringenin, CHI expression and anthocyanin accumulation) in seedlings treated with and without naringenin. Four comparisons were considered as indicated by Roman numerals: (I) and (II) correspond to genes differentially expressed after naringenin induction of tt5 and WT, respectively; (III) and (IV) correspond to genes differentially expressed between tt5 and WT, without or with naringenin induction, respectively. The numbers of genes up- and down-regulated by naringenin are indicated in gray boxes.

The addition of naringenin to tt5 resulted in induction of 32 genes and repression of 17 genes (< 0.06). Similarly, addition of naringenin to WT resulted in induction of 42 genes and repression of 28 genes (< 0.06) (Figure 2, Table 1 and Table S1). Among the genes affected by Nar addition, six were induced (Table 1) and six were repressed (Table S1) in both tt5 and WT.

Table 1. Naringenin-induced genes in WT and/or tt5
TranscriptGene descriptiontt5WT
Ratio (+Nar/−Nar)PRatio (+Nar/−Nar)P
  1. Fold changes in the gene expression between non-treated and naringenin-treated WT and tt5-1 seedlings. Relative expression values correspond to the mean of scaled and normalized triplicates. A 1.5-fold up-regulation with a P value < 0.06 was considered to be significant.

  2. a

    At5g52050 is not significantly induced but up-regulation was confirmed by quantitative RT-PCR.

At3g49780Phytosulfokine 4 (PSK4)2.800.014.570.00
At5g05340Peroxidase, putative (PER41)
At1g15520ABC transporter (PDR12)3.470.032.020.03
At2g38870Protease inhibitor, putative1.780.011.900.03
At5g50200High-affinity nitrate transporter, wound-responsive element 3 (NRT3.1, WR3)1.660.011.840.01
At1g30700FAD-binding domain-containing protein1.740.051.750.02
At4g12490Lipid transfer protein2.350.103.690.02
At2g29470Glutathione S-transferase (GST21)1.120.412.560.01
At3g46230Heat shock protein (HSP17.4-CI)1.330.352.320.01
At5g05410DRE-binding protein0.950.412.020.05
At1g16030Heat shock protein 70 (HSP70), putative1.200.322.020.00
At1g78340Glutathione Stransferase, putative (GSTU22)1.410.211.980.01
At1g60600UbiA prenyltransferase1.020.481.830.05
At2g35980Harpin-induced (YLS9)1.310.221.820.04
At3g28740Cytochrome P4500.750.291.800.03
At5g12030Heat shock protein 17.6A (HSP17.7-CII)1.420.141.790.01
At1g47980Expressed protein1.360.251.780.05
At5g48430Expressed protein0.930.361.770.04
At1g74010Strictosidine synthase1.810.091.760.04
At4g36430Peroxidase, putative1.320.141.730.05
At3g18250Expressed protein1.060.401.730.02
At1g66830Kinase, putative1.630.221.720.03
At3g27880Expressed protein1.150.361.680.00
At5g26655MYB family transcription factor1.020.471.660.01
At2g02010Glutamate decarboxylase, putative1.510.191.640.02
At1g67920Expressed protein0.780.211.610.01
At3g01260Aldose 1-epimerase1.090.391.600.02
At3g54070Ankyrin repeat0.990.491.580.03
At4g34135UDP-glucosyl transferase1.160.071.570.03
At5g4854033 kDa secretory protein-related1.340.251.570.01
At3g61570Intracellular transport protein USO1-related0.920.241.550.01
At5g19880Peroxidase, putative1.160.271.540.01
At2g38900Serine protease inhibitor0.970.391.530.04
At3g23085hAT-like transposase1.020.381.520.06
At1g67100Lateral organ boundaries domain protein 401.080.371.520.00
At3g47730ABC transporter0.910.261.520.01
At5g13880Expressed protein1.00.501.520.02
At3g22910Ca2+-ATPase, putative1.280.171.510.01
At1g74310Heat shock protein 101 (HSP101)1.160.341.510.04
At1g35710Transmembrane protein kinase, putative1.000.501.510.05
At3g53150UDP-glucosyl transferase1.070.211.500.06
At1g59860Heat shock protein (HSP17.8-CI)1.170.361.500.04
At1g64380AP2 domain-containing transcription factor, putative (NAR1)2.620.001.090.28
At5g14700Cinnamoyl CoA reductase-related2.510.050.710.38
At5g59310Lipid transfer protein 4 (LTP4)2.470.051.460.31
At2g47950Expressed protein2.250.051.510.16
At3g20210Vacuolar processing enzyme, putative (DELTAVPE)
At1g51660Mitogen-activated protein kinase kinase, putative2.040.020.870.34
At1g08080Carbonic anhydrase2.
At1g74430MYB family transcription factor (MYB95)1.970.011.440.30
At1g02400GA2-oxidase, putative1.870.021.150.40
At4g19810Glycosyl hydrolase1.830.011.110.30
At5g49520WRKY family transcription factor1.810.021.600.17
At1g24140Matrixin family protein1.810.020.970.48
At5g1620050S ribosomal protein-related1.760.031.240.39
At2g35290Expressed protein1.740.050.960.39
At1g21910AP2 domain-containing transcription factor1.740.050.640.33
At3g54420Class IV chitinase (CHIV)1.670.011.480.04
At3g61250MYB family transcription factor (MYB17)1.630.000.990.48
At4g21390S-locus lectin protein kinase1.630.050.900.38
At2g32990Glycosyl hydrolase1.630.020.670.02
At5g10100Trehalose-6-phosphate phosphatase, putative1.570.011.020.47
At2g16900Expressed protein1.520.041.390.04
At2g05940Protein kinase, putative1.520.061.240.39
At4g29030Glycine-rich protein1.510.051.100.40
At4g01090Extra-large G-protein-related1.510.051.200.33
At3g54820Aquaporin, putative1.500.030.940.38
At5g52050aMATE efflux protein-related1.550.150.860.41

To validate the microarray results, we chose genes that were significantly differently expressed between the various conditions (e.g., WT versus tt5, + and -Nar, Figure 2) according to our statistical criteria, and investigated the corresponding mRNA levels by quantitative RT-PCR. We selected 18 naringenin-modulated genes (Figures S1 and S2). The quantitative RT-PCR experiments confirmed statistically significant (< 0.1) differences in mRNA accumulation for three of seven genes induced in WT, and six of eight genes induced in tt5 (Figure S1). For genes differentially expressed between WT and tt5, four of eight were confirmed by quantitative RT-PCR in −Nar conditions and six of seven were confirmed under +Nar conditions (Figure S2). These results suggest that, even the not very stringent cut-off conditions used for the microarray analyses are sufficient to uncover significant mRNA accumulation differences (70% of the identified gene expression differences validated by quantitative RT-PCR). Gene Ontology (GO) analysis of the molecular functions associated with each of the four comparisons (Figure 2) showed that genes significantly up- or down-regulated in tt5 or WT correspond primarily to the categories ‘interaction with environment’, ‘defense functions’ and ‘cellular transport’, with a minor representation of categories that include ‘cell development and differentiation’ (Figure S3).

Genes induced by PAP1 have been previously analyzed using plants over-expressing this regulator (pap1-D) (Tohge et al., 2005). pap1-D plants also accumulate more anthocyanins compared to WT, but the intermediate cyanidin 3-glucoside (C3G), which is abundantly present in tt5-1 seedlings complemented with Nar (Poustka et al., 2007), was not detected (Tohge et al., 2005). Comparing PAP1 up-regulated genes with those affected by naringenin addition in both tt5 and WT, it became evident that these two approaches to induce flavonoid pathway flux have very different effects on mRNA accumulation (Table S2a,b).

Flavonoids modulate the expression level of core pathway enzymes

Comparison of the WT and tt5 relative expression values showed that the steady-state mRNA levels of CHS, F3H, DFR, FLS1 and the anthocyanidin synthase gene ANS are significantly higher in tt5 compared to WT in the absence of pathway flux (−Nar, Table S3a). We further confirmed these results by quantitative RT-PCR in the presence of naringenin (WT +Nar/tt5 +Nar) (Figure S2). These findings are in agreement with a previous study in which the protein levels of CHS, F3H, FLS1 and ANS were monitored in seedlings and found to be higher in tt5 compared to WT (Pelletier et al., 1999). Our microarray results also identified other flavonoid enzymes induced in tt5, including an anthocyanin 5-O-glucosyltransferase (At4g14090) and a gene encoding a protein with a fold similar to CHI (At5g05270; Table S3a), genes encoding two enzymes from the phenylpropanoid pathway [4-coumarate:CoA ligase 3 (4CL3, At1g65060) and ferulate-5-hydroxylase 1 (CYP84A1, At4g36220)], and a UDP-glucosyl transferase (UGT84A2, At3g21560) (Table S3c). Examination of the WT/tt5 comparison for the pathway transcription factors showed no significant changes, except for an induction in expression of TTG2 and TT8 in tt5 versus WT in the presence or absence of naringenin (Table S3b). TT8 coordinately regulates the flavonoid biosynthetic pathway, together with the MYB transcription factors PAP1 and TT2 (Lepiniec et al., 2006). Thus, the induction of expression of flavonoid enzymes in tt5 may be a direct consequence of the increase in TT8 transcript levels.

In contrast to the differences in gene expression observed between tt5 and WT, our results indicate that naringenin does not have a considerable effect on expression of the structural enzymes of the flavonoid pathway (Table S3a), the transcription factors that are known to regulate the pathway (Table S3b), or the majority of phenylpropanoid pathway genes (Table S3c) in either tt5 or WT seedlings.

Addition of naringenin to flavonoid mutants reveals multiple induction mechanisms

To better understand the mechanisms by which naringenin modulates gene expression, we chose to further investigate in detail the regulation and function of seven of the naringenin-modulated genes (Figure 3). These were selected based on their potential involvement in flavonoid subcellular trafficking, flavonoid detoxification or flavonoid signaling functions. These genes encode the ABC transporter PDR12 (At1g15520), a putative transcription factor NAR1 (At1g64380) that belongs to the AP2/ERF super-family, a gluthatione S-transferase GST21 (At2g29470), the phytosulfokine peptide hormone PSK4 (At3g49780), the peroxidase PER41 (At5g05340), and the high-affinity nitrate transporter NRT3.1 (also known as the wound-responsive element WR3, At5g50200). We also selected for further analysis a multi-drug efflux transporter (MATE, At5g52050), which, despite showing a P value > 0.6 in the microarray analyses, was validated by quantitative RT-PCR as induced by naringenin in tt5 (Figure 3 and Figure S1). The induction of PSK4 by naringenin in both WT and tt5 was confirmed by quantitative RT-PCR (Figure S1), while the induction of PDR12, NRT3.1 and PER41 by naringenin was confirmed in a statistically robust fashion in either tt5 or WT, but not in both genotypes (Figure 3 and Figure S1). Consistent with the microarray results, NAR1 and the MATE gene were only significantly induced by naringenin in tt5, but not in WT (Figure 3). In contrast, GST21, which was identified as induced by naringenin in WT, also showed a significant induction in tt5 (Figure 3 and Figure S1).

Figure 3.

Validation of candidate genes up-regulated after naringenin treatment.

Expression of the candidate genes PDR12, NAR1, GST21, PSK4, NRT3.1, PER41 and the MATE gene in 4-day-old WT, tt51, tt6-1, tt3-1 and fls1-1 seedlings treated or not with Nar. Transcript accumulation was measured by quantitative RT-PCR (= 3). Asterisks indicate a significant difference in expression between −Nar and +Nar (*0.1 >  > 0.05; ** 0.05).

To establish whether naringenin or a flavonoid pathway intermediate downstream of naringenin is responsible for induction of these genes, we examined their expression profile in other flavonoid mutants grown under AIC with and without naringenin supplementation. We chose the DFR, F3H and FLS1 mutants tt3-1, tt6-1 and fls1-1, respectively, because they represent key branch points in the pathway (Figure 1a and Figure S4), allowing us to group the possible effects of flavonoids into four types, i.e. those of naringenin itself (I), flavonols (II), an intermediate in the pathway downstream of naringenin but not a flavonol (III), and a compound downstream of DFR, most likely an anthocyanin (IV).

The quantitative RT-PCR analyses of the seven genes in the tt3-1, tt6-1 and fls1-1 mutants provided interesting but unexpected patterns (Figure 3). The induction of four of the seven naringenin-induced genes (PDR12, NAR1, PER41 and the MATE gene) was significantly reduced in tt3 (Figure 3). This suggests that a product downstream of DFR is responsible for induction of these genes, or that a pathway intermediate that accumulates in tt3, but not in WT, inhibits their induction by naringenin. If a product downstream of DFR were responsible for the induction, then the induction would also be significantly blocked in the tt6 mutant. This was not the case for any of the four genes (Figure 3), suggesting that perhaps a product upstream of DFR has an inhibitory effect. Among others, such inhibitory compounds may be dihydroflavonols or flavonols (Figure 1a). If the inhibitory compounds were dihydroflavonols, a similar reduction would be observed in the fls1 mutant, which was not the case for any of the genes (Figure 3). If the inhibitory compounds were flavonols, naringenin would induce expression in fls1, which is what we observed for PER41 (< 0.05) and the MATE gene (< 0.01), with a similar pattern but less statistical support for NAR1 and PDR12 (Figure 3). These results suggest that one or more flavonol compounds play a repressive role in steady-state mRNA accumulation for PDR12, NAR1, PER41 and the MATE gene. Moreover, the ability of naringenin to induce PDR12, NAR1 and NRT3.1 expression in tt5 suggests that a compound downstream of TT3/DFR that is specifically present in tt5 (perhaps a chalcone) has an activating role on expression of these genes.

Naringenin appears to be the inducer for PSK4 and, to a lesser extent, for GST21 (Figure 3). In fact, both genes are significantly induced in tt6 after naringenin treatment (< 0.01). PSK4 is robustly and very significantly (< 0.01) induced by naringenin in all the mutants tested (Figure 3). To further investigate the regulation of the PSK4 gene, we measured its promoter activity after induction with naringenin. We used transgenic plants containing the PSK4 promoter fused to the uidA reporter gene encoding β-glucuronidase (GUS) (Matsubayashi et al., 2006). Five-day-old ProPSK4:uidA seedlings were grown on AIC with or without Nar. GUS activity (see Methods S1) was increased in seedlings induced by naringenin, supporting the finding that the PSK4 promoter is induced in the presence of naringenin (Figure 4). Taken together, the chemical and genetics approaches used here resulted in identification of unknown flavonoid pathway intermediates that affect mRNA accumulation for a number of genes involved in several distinct cellular processes.

Figure 4.

Pattern of PSK4 promoter activity in wild-type seedlings.

Expression of the ProPSK4:uidA cassette in 5-day-old seedlings grown under AIC, without (a) and with (b) induction with 100 μm naringenin.

Jasmonate (JA) biosynthesis is altered in tt5

To uncover additional patterns in the genes modulated by naringenin, we utilized the ATTED-II toolbox (http://atted.jp/) (Obayashi et al., 2007) to identify clusters of co-expressed genes. We considered all the genes (total 255 genes) from the microarray that were induced (≥ 1.5-fold) in tt5 by the presence of naringenin, without any P value restriction. One of the identified clusters was particularly striking, as it contained at least 14 genes (Figure S5a) identified as induced by naringenin in tt5 (but with weak statistical support for individual genes). Among these 14 genes, ten had been previously implicated in JA biosynthesis (Table 2 and Figure S6), including two 13-lipoxygenases (At1g17420/LOX3 and At1g72520), an allene oxide cyclase (At3g25780/AOC), a 12-oxophytodienoate reductase (At2g06050/OPR3) and a 4-coumarate:CoA ligase (At1g20510/OPCL1). In addition, this cluster also contained three JA-responsive genes (At1g17380/JAZ5, At2g34600/JAZ7 and At5g13220/JAZ10) encoding jasmonate-zim-domain (JAZ) repressor proteins that function in the core JA signaling pathway (Chung et al., 2008; Kazan and Manners, 2012) (Table 2).

Table 2. Jasmonic acid-related genes up-regulated in tt5 by naringenin
TranscriptGene descriptiontt5+/− NarWT+/−Nar
  1. No P value restriction was imposed. +/− Nar indicates the ratio of expression in tt5 seedlings with and without naringenin.

At1g17380Jasmonate ZIM-domain protein 5 (JAZ5/TIFY11A)1.71.0
At1g17420Lipoxygenase (LOX3)2.71.0
At1g17750Leucine-rich repeat receptor kinase, putative (PEPR2)1.60.7
At1g205104-coumarate:CoA ligase (OPCL1)2.00.8
At1g72520Lipoxygenase, putative1.80.7
At2g0605012-oxophytodienoate reductase (OPR3)1.81.1
At2g34600Jasmonate ZIM-domain protein (JAZ7/TIFY5B)2.10.7
At3g25780Allene oxide cyclase (AOC3)1.91.1
At2g46510Abscisic acid-inducible bHLH-type (AIB)1.61.0
At5g13220Jasmonate ZIM-domain protein (JAZ10/TIFY9)1.80.6

To further investigate whether the microarray data indicated genes in the JA biosynthetic pathway (as a group) as showing increased accumulation upon naringenin treatment in tt5, we applied the gene set enrichment analysis (GSEA) algorithm (Subramanian et al., 2005). We ranked 105 genes of the JA pathway (listed in Table S4) from high to low based on the correlation between their expression levels under non-treated and naringenin-treated conditions (Figure S5b,c). Genes up-regulated under +Nar conditions compared with −Nar conditions were located at the top of the list, whereas genes that are down-regulated were located towards the bottom. We found that JA biosynthetic and JA-responsive genes were located towards the top of the list, with normalized enrichment scores of 1.53 and 1.25, respectively, and false discovery rates (q value) of 0.20 and 0.18, respectively. A false discovery rate of 0.25 was used in these studies as a cut-off. A glucosinolate biosynthetic gene set of similar size as the JA biosynthetic gene set was used as a negative control for enrichment in the comparison of tt5 with and without Nar (normalized enrichment score = −78; false discovery rate = 0.77; Figure S5d). Consistent with the weak statistical support obtained in the microarray experiments, we did not detect a statistically significant difference in expression of these nine genes when their expression was assayed individually by quantitative RT-PCR, mostly because of the very large difference between biological replicates (Figure S7).

To further examine possible connections between flavonoid and JA biosynthesis, we investigated the consequences of using wounding to increase jasmonate levels, which are usually low under normal growth conditions (Chung et al., 2008). We also explored the expression of JA biosynthetic genes, JA-responsive genes and JA levels in tt5 and WT. To achieve this, we manually wounded 11-day-old tt5-1 plants grown on MS agar containing 2.5% sucrose with or without addition of 100 μm naringenin. Subsequently, we monitored expression of the LOX3, OPR3 and JAZ5 genes by quantitative RT-PCR. We determined that these three genes were induced after wounding in both tt5 −Nar and tt5 +Nar plants (Figure 5a–c). LOX3 mRNA accumulation is significantly (< 0.05) higher under −Nar conditions, compared to +Nar conditions, at 30 min after wounding (Figure 5a). Similarly, the steady-state levels of OPR3 and JAZ5 were significantly higher in the absence of naringenin, 60 min after wounding (Figure 5b,c). Therefore, it appears that JA genes responded faster and more strongly to wounding in the absence of flavonoids. We performed similar experiments comparing WT and tt5 at 11 and 15 days after germination, collecting samples at various time points after wounding (Figure S8). As determined for tt5 −Nar/+Nar plants, significantly higher expression of LOX3 and OPR3 was observed in tt5 compared to WT at 30–45 min after wounding, with the difference in gene expression between tt5 and WT decreasing, approximately 60 min after wounding (Figure S8b,d). These results suggest that the absence of flavonoids in tt5 has a temporal quantitative effect on the response of young Arabidopsis plants to wounding.

Figure 5.

Expression of JA and other stress response genes in wounded tt5 −Nar and tt5 +Nar plants.

Ten-day-old tt5-1 plants grown on 2.5% sucrose solid medium with or without naringenin were subjected to wounding. Samples were collected at 0, 5, 30 and 60 min. Transcript accumulation was measured by quantitative RT-PCR (= 3). Asterisks indicate a significant difference in expression between −Nar and +Nar (*0.1 <  < 0.15; **< 0.1).

(a–c) JA-related genes LOX3, OPR3 and JAZ5.

(d–f) Early stress response genes MPK3, WRK40 and At1g32920.

To further investigate these effects, we measured the levels of JA and its active derivative jasmonoyl-l-isoleucine (JA-Ile) by LC-MS in green tissues extracts obtained from manually wounded 10-day-old tt5 plants grown under the same conditions as described above (MS agar containing 2.5% sucrose with or without naringenin). Consistent with the observed gene expression results (Figure 5), JA and JA-Ile levels were significantly higher in wounded tt5 seedlings without naringenin, compared to wounded naringenin-treated tt5 seedlings (Figure 6a,b). Our results demonstrate that JA (Figure 6a) and JA-Ile (Figure 6b) are produced in significantly larger quantities in tt5 when not treated with naringenin, particularly at shorter times after wounding. We performed similar JA and JA-Ile analyses in 10-day-old tt5 and WT seedlings grown under AIC. Similarly, JA (Figure 6c) and JA-Ile (Figure 6d) increase faster in tt5, compared to WT, after wounding. Moreover, within the time frame of the experiment, the levels of JA in WT fail to reach the levels present in tt5. A similar but less pronounced effect was also observed for JA-Ile, with WT accumulating 10-15% less of this compound than tt5 (Figure 6d).

Figure 6.

Jasmonate quantification after wounding of Arabidopsis seedlings.

Ten-day-old tt5-1 seedlings without or with naringenin (a, b) and tt5-1 and WT (c, d) grown on 2.5% sucrose solid medium were subjected to wounding. Samples were collected at 15, 30 and 60 min after wounding the plants, and JA (a, c) and JA-Ile (b, d) levels were measured by LC-MS (= 3). Asterisks indicate a significant difference in expression between pairs (*< 0.1). Error bars denote standard deviation.

Flavonoids modulate the induction of other stress genes

Our results showing that induction of JA is modulated by flavonoids beg the question of whether flavonoids function as overall stress response modulators. We investigated whether flavonoids could modulate the expression of rapid wound-responsive (RWR) genes, which are known to respond quickly to mechanical wounding (Walley et al., 2007; Walley and Dehesh, 2010). We used quantitative RT-PCR to investigate the expression of MPK3, WRKY40 and At1g32920, encoding a MAPK signal transduction component, a WRKY transcription factor and a protein of unknown function, respectively. Expression of these three genes is activated 5 min after wounding in tt5 −Nar plants, but their expression peaks significantly later, approximately 30 min after wounding, in tt5 +Nar plants. Moreover, the expression of MPK3 and At1g32920 is significantly higher in tt5 −Nar plants compared with tt5 +Nar plants 5 min after wounding (Figure 5d,f), and WRKY40 expression is higher in tt5 −Nar plants 60 min after wounding (Figure 5e).


Here we describe the consequences of altered flux through the flavonoid biosynthetic pathway on cellular mRNA accumulation in Arabidopsis young seedlings. In contrast to other studies, the flavonoid flux changes were not a consequence of altering the expression of flavonoid pathway regulators. Instead, we increased flux by adding an early pathway precursor, naringenin, to tt5 or WT plants under conditions in which all pathway genes are already expressed at a high level (AIC). As anticipated, the effects on mRNA accumulation were modest, but this strategy resulted in identification of a flavonoid-responsive gene set distinct from that controlled by pathway regulators.

Feedback regulation in the flavonoid pathway

The possibility of positive feedback regulation of biosynthetic genes by pathway intermediates was recognized early on, when the effect of flavonoid biosynthesis mutants on accumulation of flavonoid mRNAs (Shirley et al., 1995) and proteins (Pelletier et al., 1999) was investigated. Our microarray results support and expand these studies, showing that CHS, F3H, DFR, F3'H, FLS1 and ANS mRNA levels are significantly higher in tt5 compared to WT, especially when naringenin is added to the growth medium (Figure S2 and Table S3a).

Flavonoids affect the accumulation of a discrete set of mRNAs unrelated to flavonoid biosynthesis

The ability to induce flux through the flavonoid pathway by adding naringenin to seedlings grown under AIC provides an opportunity to investigate how flavonoids themselves affect genome-wide mRNA levels (Figure 2 and Table 1). Combining the addition of naringenin with use of several mutants (tt5-1, tt3-1, tt6-1 and fls1-1) that block the flavonoid pathway at various points enabled us to determine what group of compounds are the ones most likely contributing to the changes in mRNA accumulation.

Our studies specifically focused on a set of seven genes induced by naringenin in tt5, WT or both (Figure 3). PDR12 is one of 15 members of the pleiotropic drug-resistant group of ABC super-family of transporters, and is induced by a number of compounds including the diterpenoid sclareol, cycloheximide (van den Brule and Smart, 2002), salicylic acid, ethylene, methyl jasmonate (Campbell et al., 2003), phytoprostanes (non-enzymatically formed oxylipins; Mueller et al., 2008) and lead (Lee et al., 2005). Expression of PRD12 promoter–GFP fusions indicated that PDR12 expression is localized to the plasma membrane (Lee et al., 2005). PDR12 was recently shown to mediate uptake of the phytohormone abscisic acid into plant cells. Abscisic acid is necessary for the timely closure of stomata in response to drought stress, as well as for normal seed germination and lateral root development (Kang et al., 2010). This study further suggested that PDR12, through the uptake of abscisic acid, indirectly controls lead tolerance and terpenoid export. In our experiments, PDR12 may function as a general xenobiotic transporter that is induced when plants accumulate intermediates in the flavonoid pathway. Indeed, the ability of naringenin to induce PDR12 in tt5, but not in tt6, tt3 or fls1 (Figure 3), suggests that induction is perhaps conferred by a compound downstream of DFR that is specifically present in tt5 seedlings treated with naringenin.

GST21 was significantly up-regulated by naringenin in both WT and tt5. Arabidopsis encodes a super-family of 47 gluthatione S-transferase (GST) genes (Wagner et al., 2002). Among them, TT19 (GST26) is necessary for anthocyanin and proanthocyanidin accumulation (Kitamura et al., 2004, 2010). Similar to Petunia hybrida AN9, GST26 has been proposed to participate in the escort of anthocyanins to the tonoplast (Mueller et al., 2000). GST21 and GST26 belong to different GST sub-classes, based on amino acid similarity and gene organization (Wagner et al., 2002). In silico analyses (Genevestigator; https://www.genevestigator.com/) suggest that GST21 expression is induced after pathogen attacks. Therefore, it is possible that GST21 acts in the same way as PDR12, participating in xenobiotic transport and re-localization when xenobiotics over-accumulate in the cell after abiotic stress. Naringenin induces GST21 expression only in WT, tt5 and tt6, suggesting that naringenin itself is most probably the inducing molecule.

NAR1 encodes a transcription factor of the AP2/ERF super-family (Nakano et al., 2006) that is involved in the regulation of a variety of biological processes related to growth and development (Boutilier et al., 2002), and participates in multiple responses to environmental stress (Gutterson and Reuber, 2004). Our results indicate that NAR1 is only induced by naringenin in tt5 (Figure 3), which suggests that, similar to PDR12, it may be induced by a compound downstream of DFR. Expression profiles after various stress conditions (Genevestigator) show that NAR1 is substantially induced by senescence and by 2,4-dichlorophenoxyacetic acid, a synthetic auxin molecule that is used as an herbicide (Hanson and Trewavas, 1982). NAR1 expression was also found to be up-regulated in dormant seeds during imbibition, and a function in dormancy maintenance was suggested (Hanson and Trewavas, 1982). It is tempting to speculate that some anthocyanins (e.g. C3G) may form in Arabidopsis in response to particular biotic or abiotic stress conditions, and that such chemical entities modulate the stress response pathway.

Phytosulfokines (PSKs) are sulfated pentapeptide [Tyr(SO3H)-Ile-Tyr(SO3H)-Thr-Gln] growth factors (e.g. PSK-α) that regulate cell proliferation, and were first isolated from conditioned media of Asparagus officinali mesophyll cell cultures (Matsubayashi et al., 2001). The production of PSK-α correlates with signal transduction pathways for auxins and cytokinins (Matsubayashi et al., 1999). There are five PSK genes in Arabidopsis, all of which were expressed in the seedling system investigated here. Only PSK4 was affected by naringenin, and was induced in WT as well as the tt5, tt3, tt6 and fls1 mutants, suggesting that expression of this gene is modulated by naringenin itself, rather than by a pathway product. Moreover, the induction of PSK4 expression by naringenin appears to be most dramatic in the root (Figure 4). Flavonoids, particularly flavonols, are known to modulate root growth through modulation of auxin transport (Buer and Muday, 2004; Zazimalova et al., 2007). Naringenin may be another key flavonoid modulating root architecture through modulation of a signal peptide, such as PSK4.

Another signaling process highlighted by these analyses is the relationship between anthocyanin induction and nitrate uptake. Naringenin treatment in tt5 induced expression of At5g50200, a gene encoding a wound-responsive protein (WR3), also identified as a high-affinity nitrate transporter (NRT3.1; Okamoto et al., 2006). A relationship between nitrate limitation and flavonoids has been reported previously (Peng et al., 2007). In fact, genome-wide mRNA accumulation analyses show an induction of anthocyanin biosynthetic genes (CHS, CHI, F3H, DFR and ANS) in Arabidopsis plants grown under limiting nitrate conditions through induction of the PAP1 transcription factor. This induction of anthocyanins was proposed to occur as a means to protect plants from the photo-damage resulting from nitrogen limitation (Peng et al., 2007). Based on our results, expression of NRT3.1 appears to be induced by a downstream product of DFR, and other nitrate high-affinity transporters were not induced by naringenin. High or fast accumulation of anthocyanin intermediates may be sensed as a potential nitrogen deficiency stress for the plant. It would be interesting to determine whether the nitrate content is higher in tt5 after naringenin induction.

Peroxidases are oxidoreductases that may be involved in flavonoid oxidation (Pourcel et al., 2007). After an environmental stress, peroxidases scavenge hydrogen peroxide (H2O2) through flavonoid oxidation, contributing to detoxification (Pourcel et al., 2007). Expression of the PER41 peroxidase was previously shown to be stress-induced (Zhang et al., 2007; Rius et al., 2008). PER41 may induce oxidation of anthocyanins or intermediates participating in an unknown detoxification process, for example. Further analyses of the mutant and in vitro activity of the enzyme are necessary to uncover the mechanism of action.

At5g52050 encodes a proton-coupled multi-drug and toxin extrusion transporter (MATE) that was up-regulated by naringenin in tt5 (Figure 3). Little is know about this particular MATE. However, MATE transporters are known to be associated with flavonoid transport. For example, the Arabidopsis TT12 MATE is involved in uptake of proanthocyanidins into the vacuole of endothelial cells of seeds (Debeaujon et al., 2001). Therefore, the MATE identified in our study may be induced when anthocyanins are newly formed, in order to transport them to a particular subcellular compartment. Moreover, according to our results, At5g52050 may be induced by a downstream product of naringenin that may be C3G itself. However, we cannot rule out the possibility that a flavonol compounds may also play a repressive role in At5g52050 steady-state mRNA accumulation.

Flavonoid homeostasis modulates stress-induced JA biosynthesis

The induction of JA biosynthetic genes as a group by naringenin in tt5 suggested the possibility that flavonoids modulate the levels of this phytohormone. To establish whether this was the case, we induced JA accumulation by wounding, and measured JA and JA-Ile levels in tt5 with and without naringenin (Figure 6). The levels of JA and JA-Ile were always significantly lower when flavonoids were formed, compared to seedlings without flavonoids. A role for flavonoids in modulating accumulation of JA and JA-Ile was also supported when WT and tt5 seedlings were compared (Figure 6c,d). JA has long been known as an inducer of anthocyanin accumulation in Arabidopsis (Loreti et al., 2008; Shan et al., 2009) and other plants (Curtin et al., 2003; Shimizu et al., 2010). Recently, JAZ proteins were shown to directly bind to TT8 and PAP1 (a basic helix-loop-helix transcription factor and a MYB transcription factor, respectively) that are involved in regulation of the flavonoid pathway (Qi et al., 2011). Our results suggest the existence of a negative feedback regulation in which JA induces flavonoids, which then modulate the levels of JA (and JA-Ile) accumulation.

Flavonoids as stress response buffers?

Why are jasmonate levels controlled by flavonoids? We hypothesize that (some) flavonoids function as ‘buffering’ molecules to modulate responses induced by biotic or abiotic stress conditions. If this is the case, we expect that expression of genes induced by stress will be modulated as well. Here we showed that the induction of rapid wound-responsive (RWR) genes, which are known to respond quickly to mechanical wounding (Walley et al., 2007; Walley and Dehesh, 2010), was significantly delayed and reduced when naringenin was added to wounded tt5 plants (Figure 5d–f). Therefore, these results are consistent with a possible function of flavonoids as general modulators of stress responses.

It is interesting to view these findings in the context of the xenohormesis phenomenon, a theory that has gained considerable interest in an attempt to explain how many plant molecules modulate key regulatory steps in animal physiology, ultimately resulting in a variety of health benefits (Lamming et al., 2004; Howitz and Sinclair, 2008). Most significantly, fatty acid oxidation products (represented by jasmonates in plants and prostanglandins in animals) participate in similar wound responses in both kingdoms (Schultz, 2002). Plant-derived compounds, such as salicylic acid, which is known to inhibit JA biosynthesis in plants (Glazebrook et al., 2003; Koornneef et al., 2008; Leon-Reyes et al., 2010), also inhibit the formation of prostaglandins (Langberg et al., 2003). This inhibition results from direct binding of salicylic acid to cyclooxygenase (COX) involved in prostaglandin biosynthesis. Flavonoid have previously been shown to indirectly inhibit COX-2 activation (Nicholas et al., 2007; Seong et al., 2011; Tahanian et al., 2011), suggesting that flavonoids can modulate hormones in animals.

Experimental procedures

Plant materials

The Arabidopsis tt5-1 (CS86, At3g55120), tt3-1 (CS84, At5g42800), tt6–1 (CS87, At3g51240), fls1-1 (FLAG_533E06, At5g08640) and per41-1 (SALK_081257, At5g05340) mutants were obtained from the Arabidopsis Biological Resource Center (Columbus, OH). The tt5-1, tt3-1 and tt6-1 mutants are in the Ler background, fls1-1 is in the WS background, and per41-1 is in the Col-0 background.

Gene expression analyses

For RT-PCR experiments, green tissues from 30 to 40 seedlings were used for each RNA extraction. Total RNA was extracted using an RNeasy plant mini kit according to the manufacturer's instructions (Qiagen, http://www.qiagen.com). cDNA was synthesized from 1 μg total RNA using Superscript reverse transcription enzyme II (Invitrogen, http://www.invitrogen.com), and used as a template for quantitative PCR amplification in an Applied Biosystems 7500 real-time PCR system (http://www.appliedbiosystems.com) using SYBR Green (Applied Biosystems) as the fluorescent reporter. Primers were designed to generate unique 100-200 bp fragments (Table S5). For normalization, EF1aA4 was used as a reporter gene. Biological triplicates were used for each experiment. Student's t test was used as a statistical tool, using one-tailed distribution and two-sample equal variance.

Microarray analysis

Total RNA was prepared following the TRIzol reagent protocol (http://www.invitrogen.com). The RNA was further processed at the Columbus Children's Research Institute (http://genomics.nchresearch.org/). Either biotin-labeled cRNA or biotin-labeled cDNA was synthesized from 25 ng total RNA, hybridized to ATH1 whole-genome chips (Affymetrix Corporation, http://www.affymetrix.com) and visualized by biotin staining using streptavidin R/phycoerythrin. The scanned images were further analyzed using microarray data analysis packages. Background adjustment and cross-chip normalization were performed by robust multi-chip analysis in AffylmGUI (version 1.5.4), using a linear model to summarize probe level expression values (Irizarry et al., 2003). Statistical and additional packages, R statistics (http://cran.r-project.org/), Tcl/Tk (http://www.activestate.com/Products/ActiveTcl/) and Bioconductor packages (http://www.bioconductor.org/) were downloaded for use with AffylmGUI. The linear values of normalized and scaled relative expression levels were used for further analysis. Mean values of the triplicate sets were determined, and the ratios calculated.

Quantification of JA and JA–Ile levels

Extraction and measurement of endogenous JA and JA-Ile were performed as described previously (Chung et al., 2008; Koo et al., 2009). Dihydro-JA and [13C6]JA-Ile were added to the frozen plant sample at the start of extraction as internal standards for JA and JA-Ile quantification, respectively. Compounds were separated on an Ascentis C18 column (2.7 μm particle size, 2.150 mm column width) using an Acquity ultra performance liquid chromatography system (Waters, http://www.waters.com). A Quattro Premier XE tandem quadrupole mass spectrometer (Waters) was used in electrospray negative mode to detect JA (m/z 209→59), dihydro-JA (211→59), JA-Ile (322→130) and [13C6]JA-Ile (328→136). Data analysis was performed using MassLynx 4.1 software (Waters).


We thank Y. Matsubayashi (Yoshikatsu Matsubayashi, National Institute for Basic Biology, Myodaiji, Okazaki, Aichi, Japan) for sharing with us the PSK4:uidA line, K. Saito (Metabolic Function Research Team Plant Science Center, Tsurumi, Yokohama, Kanagawa, Japan) for the microarray data on pap1-D plants, and B. Read (California State University at San Marcos, San Marcos, CA, USA) for assistance with early microarray experiments. We acknowledge support from the Chemical Sciences, Geosciences and Biosciences Division, Office of Basic Energy Sciences, Office of Science, US Department of Energy (grant number DE-FG02-91ER20021 to G.H.) for the JA/JA-Ile measurements. Support for this project was also provided by US Department of Energy grant number DE-FG02-07ER15881, Agricultural and Food Research Initiative competitive grant number 2010-65115-20408 from the US Department of Agriculture National Institute of Food and Agriculture, and US National Science Foundation grant number MCB-1048847 to E.G.