From network to phenotype: the dynamic wiring of an Arabidopsis transcriptional network induced by osmotic stress

Abstract

Plants have established different mechanisms to cope with environmental fluctuations and accordingly fine-tune their growth and development through the regulation of complex molecular networks. It is largely unknown how the network architectures change and what the key regulators in stress responses and plant growth are. Here, we investigated a complex, highly interconnected network of 20 Arabidopsis transcription factors (TFs) at the basis of leaf growth inhibition upon mild osmotic stress. We tracked the dynamic behavior of the stress-responsive TFs over time, showing the rapid induction following stress treatment, specifically in growing leaves. The connections between the TFs were uncovered using inducible overexpression lines and were validated with transient expression assays. This study resulted in the identification of a core network, composed of ERF6, ERF8, ERF9, ERF59, and ERF98, which is responsible for most transcriptional connections. The analyses highlight the biological function of this core network in environmental adaptation and its redundancy. Finally, a phenotypic analysis of loss-of-function and gain-of-function lines of the transcription factors established multiple connections between the stress-responsive network and leaf growth.

Keywords: growth regulation; mild osmotic stress; short‐term stress response; transcription factors; transcriptional network.

Figure 1. Mannitol‐induced transcriptional changes of the selected TFs in proliferating, expanding, and mature leaf tissue

The expression of the 20 genes encoding TFs was measured 24 h after mannitol treatment during the proliferating (n = 192 plants), expanding (n = 16 plants), and mature (n = 16) leaf developmental stage. Expression levels in wild‐type plants transferred to mannitol‐induced stress were compared to those transferred to control conditions at the same developmental stage. Data information: Data are presented as mean ± SEM, n = 4 independent experiments. FC = fold change. *FDR < 0.05, unpaired two‐sided Student's t‐test.

Figure 2. Four groups of transcriptional induction upon exposure to mannitol

1. A–D

   Based on a threshold of log₂(FC) > 1, the 20 TFs were categorized into four groups. The first group contains TFs that reached the log₂(FC) threshold 40 min after mannitol treatment (A), the second group reached the threshold after 1 h (B), the third group after 2 h (C), and the fourth group after 4 h (D). The arrow indicates the initial upregulation of every group.

Data information: Data are presented as mean ± SEM. n = 4 independent experiments. FC = fold change. FDR values are available in Table EV1.

Figure 3. The regulatory connections of the osmotic stress‐responsive GRN

1. The significant regulatory interactions identified by nCounter Nanostring at 1 h, 2 h, 4 h, 8 h, and 24 h after induction of overexpression of a TF.

2. The confirmed regulatory interactions between the 20 TFs part of the GRN, according to transient expression assays (n = 4 biological repeats). Green arrows represent activation and red arrows repression.

3. Heatmap of the significant regulations upon induction of overexpression of the five members of the core network, the activators (green) ERF6, ERF59, ERF98, and the repressors (red) ERF8 and ERF9. Color code represents FDR‐corrected P‐values with thresholds at FDR = 0.01, 0.05 and 0.1.

Data information: In (A), data are extrapolated from estimated averages, n = 3 independent experiments, FDR‐corrected P < 0.1 (mixed model analysis, user‐defined Wald tests). The thickness of the arrows represents the FDR value. In (B), data are presented as averages, n = 3 independent experiments. The intensity of the color of the arrows represents the strength of the regulation according to the TEA values and the thickness the FDR value of the nCounter Nanostring experiment. In (C), data are represented as FDR‐corrected P‐values, n = 3 independent experiments (mixed model analysis, user‐defined Wald tests). Source data are available online for this figure.

Figure EV1. Expression profiles of WRKY15 target genes

Expanding leaf tissue (third leaf – 15 DAS) of WRKY15‐GR was harvested 1 h, 2 h, 4 h, 8 h, and 24 h after transfer to dexamethasone. Expression values were normalized against the control line.

1. WRKY15 target genes (significantly differentially expressed during one time point) that showed a gradually increasing expression pattern.

2. WRKY15 target genes that showed an oscillating expression pattern.

Data information: data are presented as mean ± SEM, n = 3 independent experiments, *FDR < 0.1 (mixed model analysis, user‐defined Wald tests).

Figure 4. Phenotypic analysis of loss‐of‐function (LOF) and gain‐of‐function (GOF) lines of the TFs under control conditions

1. A

   At 22 days after stratification (DAS), leaf series of the LOF lines were made and the rosette area was calculated as the sum of the area of all individual leaves (n = 10 plants). The rosette area is presented relative to the corresponding wild type.

2. B

   Average area of the individual leaves of rap2.6L and myb51, two knockout lines with a smaller average rosette area.

3. C

   Rosette area of two or three independent GOF lines per TF (calculated as the projected area or the sum of the leaves, n = 10 plants) germinated and grown on DEX‐containing medium. Measurements were performed at 22 DAS relative to the control line. “Independent line 1” is the line with the highest overexpression level.

4. D

   A representative picture of the rosette of GOF lines with significant growth phenotypes in both independent lines at 22 DAS. Scale bar is 1 cm.

5. E, F

   Pavement cell number and area of the third leaf at 22 DAS of LOF (E) and GOF lines (F) that showed a significant rosette area phenotype.

Data information: In (A, C), data are presented as mean ± SEM, n = 3 independent experiments, *P < 0.05 (Tukey's test). In (B), data are presented as mean ± SEM, n = 3 independent experiments, *P < 0.05 (mixed model, partial F‐tests). In (E, F), data are presented as mean ± SEM, n = 3 independent experiments. ^. P < 0.1, *P < 0.05, **P < 0.01, ***P < 0.001 (Tukey's test).

Figure 5. Expression of four GA biosynthesis and degradation enzymes in wild‐type plants upon mannitol treatment and in GOF lines showing an altered growth phenotype in at least two independent lines

1. The expression level of two GA degradation and two GA biosynthesis genes in expanding leaf tissue (third leaf at 15 DAS) of wild‐type plants 20 min, 40 min, 1 h, 2 h, 4 h, 8 h, 12 h, 16 h, and 24 h after mannitol treatment. The fold changes (FC) were calculated relative to control conditions.

2. The expression of GA2‐OX6 and GA20‐OX1 in expanding leaf tissue (third leaf – 15 DAS), 8 h after transfer to DEX‐containing medium to induce overexpression. The FC was calculated relative to control conditions.

3. The effect of ERF6, ERF9, ERF98, and WRKY15 on the GA2‐OX6 promoter determined with transient expression assays. The relative luminescence was calculated relative to the control, 35S::GUS (n = 4 biological repeats).

Data information: In (A), data are presented as mean ± SEM, n = 4 independent experiments, *FDR < 0.05 (unpaired two‐sided Student's t‐test). In (B), data are presented as mean ± SEM, n = 3 independent experiments, *FDR < 0.1 (mixed model analysis, user‐defined Wald tests). In (C), data are presented as mean ± SEM, n = 3 independent experiments.

Figure 6. Phenotypic analysis under mild osmotic stress of loss‐of‐function (LOF) and gain‐of‐function (GOF) lines of every TF

1. At 22 DAS, leaf series were made of the LOF lines (n = 10 plants) grown under control conditions or mild osmotic stress conditions, and the reduction under mild osmotic stress was calculated. The rosette area reduction is presented relative to the corresponding wild type.

2. Leaf series of stz and erf11 mutants grown on control and mild osmotic stress conditions.

3. The rosette area of two or three independent GOF lines (n = 10 plants) germinated on DEX‐containing control or mild osmotic stress medium was measured at 22 DAS, and the reduction by mild osmotic stress was calculated relative to the control line. “Independent line 1” is the line with the highest overexpression level.

Data information: In (A, C), data are presented as mean ± SEM, n = 3 independent experiments, *P < 0.05 (Tukey's test). In (B), data are presented as mean ± SEM, n = 3 independent experiments, *P < 0.05 (mixed model, partial F‐tests).

Figure 7. Five different effects of two TFs on a common target gene

1. A–E

   The effect of the individual and the combination of two TFs on the expression of target genes ERF11 (A), ERF5 (B), MYB51 (C), RAP2.6L (D), and ERF6 (E). The relative luminescence was calculated relative to the control, 35S::GUS (n = 4 biological repeats). Green represents activation, red repression, and gray absence of regulation.

Data information: Data are presented as mean ± SEM, n = 3 independent experiments.

Figure EV2. Phenotypic analysis of crosses between two GOF lines of the core network

1. A–C

   All members of the core network were crossed with either each other, referred to as double cross, or with the control line (GFP‐GR), referred to as single cross and were germinated and grown on dexamethasone (n = 8 plants). At 22 DAS, the projected rosette area of the double and single cross and the GFP‐GR was measured. Double crosses resulted in a synergistic (A), an additive (B), or a negative (C) phenotype and are depicted in orange, green, and blue connection, respectively.

Data information: Data are presented as mean ± SEM, n = 3 independent experiments, *P < 0.05.

Figure 8. Phenotypes of the double crosses between core network members

The gain‐of‐function lines of all members of the core network were crossed either with each other, referred to as double cross, or with the control line (GFP‐GR), referred to as single cross. The projected rosette area of the F1 double and single crosses and the GFP‐GR line, germinated and grown on DEX, was measured at 22 DAS. A representative picture of the single and double crosses is depicted. The values represent the relative increase or decrease in rosette size of the double crosses compared to their parental single crosses. The outcome of every double cross is classified into three groups: additive (green connections), negative (blue connection), and synergistic (orange connections) phenotype. The precise measurements can be found in Fig EV2.

Figure 9. Overview of the transcriptional events following osmotic stress

1. We speculate that under normal conditions, ERF8 represses the other network genes. Upon mild osmotic stress (indicated by a red arrow), some genes of the network can be phosphorylated (PTM), a hypothesis based on the literature and the abundance under normal conditions. Subsequently, the expression of the network genes increases during four groups of transcriptional induction. The direct transcriptional regulations of the core network members (ERF6, ERF8, ERF9, ERF59, and ERF98) are depicted with green and red arrows, representing activation and repression, respectively. The regulatory connections of the core network members that were identified when evaluating the effect of two TFs together on a shared target gene are depicted with dashed green and red arrows, representing activation and repression, respectively. The latter regulatory interactions occur in the presence of a necessary transcriptional partner. The color of the nodes represents the strength of the induction. FC = fold change.

2. Schematic representation of the feed‐forward loop the network is composed of. Upon input, such as mild osmotic stress, activators of the core network are induced (A) and activates downstream TFs (C). These TFs or another unknown component could induce the expression of the repressors of the core network (B), which in turn leads to the repression of the downstream TFs (C), restoring the original state of the network.