Expansion of a core regulon by transposable elements promotes Arabidopsis chemical diversity and pathogen defense

Abstract

Plants synthesize numerous ecologically specialized, lineage-specific metabolites through biosynthetic gene duplication and functional specialization. However, it remains unclear how duplicated genes are wired into existing regulatory networks. We show that the duplicated gene CYP82C2 has been recruited into the WRKY33 regulon and indole-3-carbonylnitrile (ICN) biosynthetic pathway through exaptation of a retroduplicated LINE retrotransposon (EPCOT3) into an enhancer. The stepwise development of a chromatin-accessible WRKY33-binding site on EPCOT3 has potentiated the regulatory neofunctionalization of CYP82C2 and the evolution of inducible defense metabolite 4-hydroxy-ICN in Arabidopsis thaliana. Although transposable elements (TEs) have long been recognized to have the potential to rewire regulatory networks, these results establish a more complete understanding of how duplicated genes and TEs contribute in concert to chemical diversity and pathogen defense.

Fig. 1

4OH-ICN is synthesized under ETI-like responses. a Schematic of tryptophan (L-Trp)-derived specialized metabolism in A. thaliana. White arrows denote the presence of additional enzymes. ANI, aci-nitro indole; ICY, indole cyanohydrin. b. LC-DAD-FLD-MS analysis of camalexin (top), ICN (middle), and 4OH-ICN (bottom) in seedlings elicited with indicated MAMPs and bacterial strains for 27 h. Data represent mean ± SE of four replicates of 15 ± 2 seedlings each. Different letters denote statistically significant differences (P < 0.05, one-factor ANOVA coupled to Tukey’s test). ICN totals consist of the sum of ICN and methanolic degradation product ICA-ME. 4OH-ICN totals consist of the sum of aqueous and methanolic degradation products 4OH-ICA-ME and 4OH-ICA, respectively

Fig. 2

Intraspecific variation in WRKY33 affects 4OH-ICN and immunity. a Schematic of WRKY33 proteins in Col-0, Col-0 wrky33, Ler-1, and Di-G. Black boxes denote WRKY domains (W), nuclear localization signal (NLS), or C-terminal domain (CTD). b LC-DAD-MS analysis of camalexin, ICN, and 4OH-ICN in seedlings inoculated with Psta for 24 h. Data represent mean ± SE of four replicates of 15 ± 2 seedlings each. c Bacterial growth analysis of Pst in surface-inoculated leaves. Middle and right panels were pre-treated with 20 μM dex for 6–8 h. Data represent mean ± SE of 4 (left); 11, 6, 7, 7, 7 (middle); 6, 6, 8 (right) replicates of 15 ± 2 seedlings each. CFU, colony-forming units. Different letters in b, c denote statistically significant differences (P < 0.05, one-factor ANOVA coupled to Tukey’s test). Source data of Figs. 2b and 2c are provided as a Source Data file

Fig. 3

WRKY33 directly activates 4OH-ICN biosynthetic genes. a qPCR analysis of 4OH-ICN regulatory and biosynthetic genes in seedlings co-treated with 20 μM dex and Psta for 9 and 12 h. Different letters denote statistically significant differences (P < 0.05, one-factor ANOVA coupled to Tukey’s test). Lowercase and uppercase letters denote comparisons across 9 and 12 h time points, respectively. Data represent mean ± SE of 4, 5, 4, 5 (9 h) and 6, 6, 6, 5 (12 h) replicates of 15 ± 2 seedlings each. b Schematic of FOX1 and CYP82C2 loci, indicating nt positions of W-box-containing regions (W). c ChIP-PCR analysis of W-box-containing regions upstream of FOX1 and CYP82C2 in wrky33/DEX:WRKY33-flag plants co-treated with 20 μM dex (D) or mock solution (M) and Psta for 9 h. Dashed line represents the fivefold cutoff between weak and strong TF-DNA interactions. Data represent median ± SE of four replicates of 15 ± 2 seedlings each. Source data of Figs. 3a and 3c are provided as a Source Data file

Fig. 4

Regulatory neofunctionalization of CYP82C2. a (Left) Phylogenetic species tree. (Right) HPLC-DAD analysis of 4OH-ICN in seedlings inoculated with Psta for 30 h. Data in box plots represent median (center line), 25th percentile (lower box limit), 75th percentile (upper box limit), and full range of variation (whiskers) for n = 13, 9, 3, 6, 6, 6, 6, 3, 3, 6 replicates of 15 ± 2 seedlings each. Data were pooled from several independent experiments with A. thaliana as the positive control. 4OH-ICA and 4OH-ICA-ME are aqueous and methanolic degradation products of 4OH-ICN, respectively. DW, dry weight; n.d., not detected. b (Left) Phylogenetic species tree. (Right) Synteny map of CYP82C genes. Gray arrows or rectangles represent non-CYP82C genes. Gray dotted lines represent large ( > 500 nt) sequence gaps. c, d qPCR analysis of 4OH-ICN and sideretin biosynthetic genes in seedlings inoculated with Psta (c) or grown in iron-deficient medium (d). Data in c represent mean ± SE of 3 (12 h Arabidopsis lyrata) or 4 (all other) replicates of 15 ± 2 seedlings each. Data in d represent mean ± SE of 3 replicates of 15 ± 2 seedlings each. Asterisks denote statistically significant differences of stress-treated relative to untreated samples (P < 0.05, two-tailed t-test). Source data of Figs. 4a, 4c, and 4d are provided as a Source Data file

Fig. 5

TE EPCOT3 is a CYP82C2 enhancer. a mVISTA plot of CYP82C2 upstream sequence, indicating nt positions of unique (EPCOT1–3; gray boxes) and conserved regions (≥ 70% identity, pink) among homologous sequences. Also indicated are positions of W-boxes (green) and WRKY33-specific motifs (blue) present (solid lines) or absent (dashed lines) in each homologous sequence, known WRKY33 TFBSs (diamonds) and ChIP-tested regions (W1–5). Al, A. lyrata; Ah, Arabidopsis halleri; Bs, Boechera stricta; Cg, Capsella grandiflora; Cr, Capsella rubella; TSS, transcriptional start site. b Epigenetic map of CYP82C2 upstream sequence, indicating positions of significant H3K4me2 (blue–gray bars) and H3K27me3 (purple bars). c (Left) Schematic of EPCOT3 and related LINE retrotransposons in A. thaliana, indicating positions of CYP82C2 and reverse-transcriptase (RT) domains. See also Supplementary Note 1. Dashed box outlines W-boxes (green lines) and/or WRKY33-binding motifs (blue lines) within EPCOT3/EPLs. (Right) Phylogenetic maximum likelihood tree. d (Upper left) Schematic of CYP82C2 and AlCYP82C2 transgenic loci used for WRKY33 transactivation experiments. (Lower left) RT-PCR images of CYP82C2, AlCYP82C2, and NbACTIN1 in N. benthamiana leaves co-transfected with DEX:WRKY33-flag and CYP82C2 or AlCYP82C2 locus, and incubated with 1 μM flg22 and mock solution (0.5% DMSO) or 20 μM dex for 30 h (CYP82C2/AlCYP82C2) or 24 hr (NbACTIN1). Data represent five replicates (three leaf discs each). (Lower right) RT-PCR images of CYP82C2, AlCYP82C2, and EIF4A1 in A. thaliana cyp82C2 protoplasts transfected with CYP82C2 or AlCYP82C2 locus and elicited 6 h with 1 μM flg22. As original CYP82C2 primers detect endogenous transcription downstream of the cyp82C2 T-DNA insertion (see CYP82C2 + cyp82C2-2, second row), a second set of primers (CYP82C2*, Supplementary Data 2) flanking the insertion was used to test WRKY33 transactivation (see CYP82C2, first row). Data represent four replicates of 2.5 × 10⁵ protoplasts each. e ChIP-PCR analysis of W-box-containing regions (W) within EPLs in wrky33/DEX:WRKY33-flag plants co-treated 9 h with 20 μM dex or mock solution and Psta. Data represent median ± SE of four replicates (~210 seedlings each). Dashed line represents fivefold cutoff between weak and strong TF-DNA interactions. Source data of Figs. 5d and 5e are provided as a Source Data file

Fig. 6

Model of regulatory neofunctionalization of CYP82C2. An ancestral gene with roles in iron-stress responses (CYP82C4) underwent gene duplication in a progenitor species to A. thaliana and A. lyrata, leading to ancestral CYP82C2. Subsequent speciation led to ancestral A. thaliana and A. lyrata. In the former species, a significant degree of retroduplication, mutagenesis, and transposition events occurred, culminating with the formation of W-box and WRKY33-specific sequences in the ancestral EPCOT3 and its integration upstream of CYP82C2. Subsequent epigenetic modifications in A. thaliana were necessary to permit WRKY33 binding and CYP82C2 activation. Features in black have a hypothesized function, whereas features in gray have no known function. Double-dashed line indicates features omitted from view (e.g., CYP82C3)