The Arabidopsis thaliana Col-0 ecotype and erf109 knock-out mutant lines (SALK_150614) were obtained from the Arabidopsis Biological Resource Center. Seeding holder was made by punching 96 holes on a foam board and filled by the hydroponic solution which solidified with 1% Agargel (A3301, Sigma, MO, United States) containing essential nutrients as described (Rodríguez-Celma et al., 2013), which included 5 mM KNO₃, 2 mM MgSO₄, 2 mM Ca(NO₃)₂, 2.5 mM KH₂PO₄, 70 μM H₃BO₃, 14 μM MnCl₂, 1 μM ZnSO₄, 0.5 μM CuSO₄, 10 μM NaCl, 0.2 μM Na₂MoO₄, 40 μM FeEDTA, 4.7 mM MES, with pH 5.7. Then, the seeding holders were floated on the hydroponic solution inside boxes. Seeds were stratified on seeding holders at 4°C in the dark for 1 day and subsequently transferred to acclimatize under continuous light (90 μmol m^–2 s^–1) and were grown at 21°C in a growth chamber. The hydroponic solution was renewed every 3 days. Plants were transferred to Fe-sufficient or Fe-deficient (−Fe, 0 μM Fe²⁺ supplemented with 100 μM 3-(2-pyridyl)-5,6diphenyl-1,2,4-triazine sulfonate to chelate trace iron) hydroponic solution for treatment. Shoot tissue was harvested for all experiments.

For RNA-seq and quantitative RT-PCR (RT-qPCR) analysis, Col-0 and erf109 plants were grown in hydroponic solution for 10 days and transferred to Fe-sufficient or Fe-deficient hydroponic solution for an additional 3 days. Three replications for each treatment and genotype were applied for data collection. For Fe concentration analysis, plants were grown in hydroponic solution for 18 days under continuous light and transferred to Fe-sufficient or Fe-deficient solution under continuous light for 3 days. Two replications for each treatment were applied for data collection. For chlorophyll fluorescence and chlorophyll content measurements, plants were grown in Fe-sufficient hydroponic solution for 14 days and transferred to Fe-sufficient or Fe-deficient solution under continuous light for 3 days. Three replications for each treatment were applied for data collection.

For RNA-seq, RNA extraction, sequencing, and annotation were conducted as described (Rodríguez-Celma et al., 2013). Total RNA was extracted from shoots of Col-0 and erf109 plants using the RNeasy Plant Mini Kit (Qiagen), following the manufacturer’s instructions. The cDNA libraries were constructed with equal amounts of total RNA following the manufacturer’s protocol (Illumina, CA, United States), enriched by PCR amplification and subjected to paired-end sequencing on an Illumina Genome Analyzer II. Data collection was conducted as previously described (Mortazavi et al., 2008). Reads from RNA-seq were mapped to the Arabidopsis genome version TAIR10 by Bowtie2 (Langmead and Salzberg, 2012), other unmappable reads were mapped to Arabidopsis genome version TAIR10 using BLAT (Kent, 2002). Read counts were calculated and normalized by means of RackJ package¹ and the TMM-quantile method (Robinson and Oshlack, 2010), respectively. Normalized read counts were further converted into RPKM (Reads Per Kilobase Million) values.

To obtain a comprehensive catalog of differentially expressed genes (DEGs), three subsets of DEGs were selected based on mean RPKM compared to control with a Z-test P-value < 0.05: “Col-Fe” indicates genes that were differentially expressed between Fe-deficient and Fe-sufficient Col-0 plants, “erf109-Fe” indicates genes that were differentially expressed between Fe-deficient and Fe-sufficient erf109 mutant plants, and “erf109-regulon” indicates genes that were differentially expressed between erf109 and Col-0 plants under Fe-sufficient conditions.

For DEG clustering, the DEGs of ‘‘Col-Fe,’’ ‘‘erf109-Fe,’’ and ‘‘erf109-regulon’’ were used as input for the MACCU software² (Lin et al., 2011) to build co-expression networks. The network was based on co-expression relationships with a Pearson’s coefficient greater than or equal to 0.8. The network was visualized using Cytoscape software ver. 3.7.2.³ The color was set to show up-regulated (red) and down-regulated (blue) DEGs from log2 fold-change ranging from 2 to −2.

Differentially expressed genes with the same regulation pattern in “Col-Fe” and in the “erf109-regulon” were selected for functional analysis using the singular enrichment analysis tool in agriGO (Tian et al., 2017), which was applied with the default parameters for A. thaliana. The biological process result was visualized using the REVIGO (Supek et al., 2011) tool with the default parameters for A. thaliana GO terms. All the DEGs of “Col-Fe,” “erf109-Fe,” and “erf109-regulon” were put into the MapMan software (Thimm et al., 2004) with log2 fold-change values for visualizing functional categories. The “Ath_AGI_TAIR9_Jan2010” database in MapMan was applied for mapping. The heatmaps were plotted in R (version 4.1.0).

Total RNA was extracted with TRIzol™ Reagent (Invitrogen, MA, United States) following the manufacturer’s instructions. The HiScript II 1st Strand cDNA Synthesis Kit (Bionovas, Toronto, ON, Canada) was used for cDNA preparation. RT-qPCR was performed on the CFX Connect Real-Time PCR Detection System (Bio-Rad, CA, United States) using iQ™ SYBR ^® Green Supermix (Bio-Rad, CA, United States). For each reaction, 0.3 μM of the forward and the reverse primers and 100 ng of cDNA template were added. All primers used in this study are listed in Supplementary Table 3. The relative gene expression level was determined according to the 2^–ΔΔCt method (Livak and Schmittgen, 2001) and fold-changes were calculated relative to Col-0 values. For each sample, the mRNA abundances of the target genes were normalized to those of the UBIQTIN10 (At4g05320) gene. Three biological replicates were run for each gene. The final figures were plotted in R (version 4.1.0).

Shoots were rinsed with 1% HCl and at least three times with deionized water, dried at 100°C for 1 h, and kept at 70°C for 3 days. NIST SRM 1573a (tomato leaves) served as a standard. Dried shoots were weighed into a Teflon vessel and digested with 5 ml of 65% HNO₃ (Merck, Darmstadt, Germany) and 2 ml H₂O₂ (Choneye Pure Chemicals, Taipei, Taiwan) in a MarsXpress microwave digestion system (MARS 5 Xpress; CEM). ICP-optical emission spectrometry (ICP-OES; PerkinElmer, MA, United States) was used for metal element analyses. Determination of metal concentration was conducted as described (Shanmugam et al., 2011). The final figures were plotted in R (version 4.1.0).

The plant material was acclimated in the dark for 20 min before the measurements. Chlorophyll fluorescence was determined at room temperature using a PAM-2100 (Heinz Walz GmbH, Effeltrich, Germany) fluorometer. Minimal (Fo) and maximal (Sies and Menck, 1992) fluorescence yields of dark-adapted samples were recorded after applying a saturating pulse of light. Maximal quantum yield of photosystem II (Fv/Fm) was calculated from the given Fo and Fm-values using the equation: Fv:Fm = (Fm−Fo):Fm. The effective quantum yield of photosystem II (yield) was calculated according to the equation: Y = (Fm′−Ft): Fm′ = ΔF: Fm′. A Soil Plant Analysis Development (SPAD) chlorophyll meter (Spectrum Technologies, Inc., IL, United States) was applied to measure the chlorophyll content in each line. The SPAD value represents the index of relative chlorophyll content (Richardson et al., 2002). The final figures were plotted in R (version 4.1.0).

Data were analyzed by applying a Student’s t-test with a P-value < 0.05.

To explore the effects of ERF109 on the Fe content and light responsiveness in leaves, we first collected phenotypic data of erf109 mutant plants. After 72 h of Fe deficiency, no visible differences between the mutant and its wild-type were observed (Figure 1A). However, the Fe concentration of both Col-0 and erf109 plants was decreased in response to Fe deficiency (Figure 1B). Notably, erf109 mutant plants exhibited significantly higher Fe levels than the wild-type under both growth conditions (Figure 1B). The maximum quantum yield of photosystem II (Fv/Fm) was reduced upon Fe-deficiency in both Col-0 and erf109 plants, with erf109 plants exhibiting significantly lower values than Col-0 under Fe-deficiency (Figure 1C). A similar pattern was observed for the quantum yield of PSI and yield, the difference between the erf109 and Fe-deficiency Col-0 was, however, not statistically significant (Figure 1D). SPAD values, providing a read-out for the chlorophyll content, were reduced in both genotypes under Fe-deficient conditions, with significantly lower values in erf109 plants under Fe-sufficient and Fe-deficient conditions (Figure 1E). Together, these results indicate that in erf109 plants the chlorophyll content and photosynthetic efficiency was decreased relative to the wild-type despite higher Fe concentrations in the leaves even under Fe-sufficient conditions, suggesting that the additional Fe may be not physiologically available.

To uncover a possible role of ERF109 in the Fe deficiency response of Arabidopsis shoots, we subjected wild-type plants and erf109 mutant lines to transcriptional profiling using RNA-seq technology. Genes that were differentially expressed between the mutant and its wild-type are here referred to as the erf109-regulon, while Fe-responsive genes of the wild-type and mutant plants were designated Col-Fe and erf109-Fe, respectively. To explore the global regulation pattern of the three DEG subsets, heatmaps of log fold-change values were generated (Figure 2A). Conspicuously, highly similar expression patterns were observed for Col-Fe and the erf109-regulon. A subset of 681 of the total 955 DEGs in the erf109-regulon overlapped with the Col-Fe DEGs (Figure 2B), with 149 up-regulated and 515 down-regulated genes in both Col-Fe and the erf109-regulon (Figure 2D). The significant overlap and the similar direction of gene regulation in Fe-deficient wild-type plants and erf109 mutants indicate a partial constitutive Fe deficiency response in shoots of erf109 mutant plants. Under Fe-deficient conditions, a subset of 416 out of 1,654 DEGs that responded to Fe starvation in the wild-type were also responsive to the Fe regime in erf109 mutants; the majority of Col-Fe genes (1,238 DEGs), however, were not affected in the mutant (Figure 2C). The subset of overlapping DEGs revealed a similar direction of expression in both Col-Fe and erf109-Fe, with 124 genes that were induced and 273 genes that were repressed under Fe starvation (Figure 2E). From the 1,238 Col-0-specific Fe-responsive DEGs that were not affected in erf109 mutant plants under Fe deficiency, 103 induced and 433 repressed genes were regulated similarly in the erf109-regulon but lost their differential expression in the erf109-Fe dataset (Figure 2F). The function of the 681 DEGs that were comprised in both the Col-Fe subset and the erf-109 regulon were further analyzed by their gene ontology. Among these genes, 149 up-regulated DEGs are involved in the response to abiotic stimulus and cellular nitrogen compound metabolism (Figure 2G), and 515 down-regulated DEGs are associated with the response to biotic stimulus, wounding, chitin, and the immune system (Figure 2H). These results indicate that knock-out of ERF109 induces gene expression patterns that were to a large part similar to those observed in Fe-deficient wild-type plants and predicted to be involved in establishing immunity.

To further explore the changes in the transcriptome induced by either Fe starvation, dysfunctional ERF109, or both conditions, a co-expression network was constructed for the gene subsets Col-Fe, erf109-Fe, and erf109-regulon. Total DEGs were divided into two main clusters by their co-expression relationships: group 1 comprises mainly down-regulated genes, while group 2 comprises mostly up-regulated genes, which can be further subdivided into chloroplast-encoded and nucleus-encoded genes (Figure 3A). In the Col-Fe network, group 1 genes are connected to group 2 DEGs, which in turn is connected to a group of chloroplast-encoded DEGs which were significantly up-regulated (Figure 3A). The latter subset is not significantly regulated within the erf109-regulon (Figure 3B). Under Fe-deficient conditions, genes in erf109-Fe lost most of the response within both groups; some group 2 DEGs were even down-regulated (Figure 3C). Taken together, the network revealed that the overlap of the erf109-regulon and Col-Fe partly comprises the down-regulated genes in group 1, and the erf109-regulon completely abolishes the up-regulation of chloroplast-encoded genes. Moreover, defective ERF109 dramatically affected the gene expression pattern observed under Fe-deficient conditions in wild-type plants.

To further investigate the function of the DEGs in the erf109-regulon, a gene ontology analysis using the MapMan software was employed. The analysis revealed that the most significant down-regulated DEGs are involved in biotic and abiotic stress responses, including R genes that code for proteins recognizing specific pathogen effectors, induction of the respiratory burst, signaling, and defense genes such as PR-proteins. Genes putatively involved in processes related to biotic stress such as hormone signaling, cell wall, proteolysis, and transcription factors were significantly reduced in both Col-Fe and erf109-regulon (Figures 4A,B). These defense-related DEGs comprise transcription factors such as DWARF AND DELAYED FLOWERING (DDF), SG2-type R2R3-MYB (MYB), WRKY, a putative zinc-finger (RHL), and other ERF family proteins (Figure 4C) as well as signaling proteins such as calmodulin-like proteins (CML), MAP kinases (MPK), and other defense-related proteins including a cysteine-rich receptor-like protein kinase (CRK), a PLANT U-BOX (PUB) protein, PROTEIN KINASE 2A (APK2A), and SUPPRESSOR OF BIR1 (SOBIR1) (Figure 4D). The observation that these stress response-associated genes were commonly affected in Col-Fe and in the erf109-regulon suggests that ERF109 governs the regulation of Fe-responsive immunity-associated genes that are part of the Fe deficiency response.

In addition, induction of genes encoding photosystem subunits was observed in both Col-Fe and in the erf109-regulon, including a large fraction of genes encoding proteins involved in photosynthetic electron transport which showed increased transcription in Col-Fe and, to a lesser extent and partly different from those of Col-Fe, in erf109 plants (Figures 5A,B). In Col-Fe, the most highly expressed genes were encoded by plastidic DNA, including genes encoding proteins associated with the reaction centers of PSI and PSII, electron transfer proteins, NAD(P)H dehydrogenases, and ATP synthases (Figure 5C). By contrast, in the erf109-regulon, other genes associated with PSI and PSII, ferredoxin, and light-harvesting chlorophyll a/b-binding proteins were higher expressed than in the wild-type. When erf109 plants were subjected to Fe-deficient conditions, genes involved in these processes were rather repressed, showing a pattern that was clearly distinct from the other two transcriptomes (Figure 5C).

Chlorophyll plays an essential role in light harvesting and electron transfer in photosynthesis (Schreiber et al., 1995). The expression of genes encoding enzymes in tetrapyrrole biosynthesis such as HEME2, GUN4, CHLI1, and PORB was enhanced in both Col-Fe and in the erf109-regulon, but this induction was abolished in Fe-deficient erf109 plants (Figure 5D). A comparison between the genotypes suggests that ERF109 is mainly required for mounting defense responses under conditions of Fe deficiency. In addition, defects in ERF109 under Fe-deficiency is accompanied by partial repression of photosynthesis subunits and genes involved in chlorophyll biosynthesis.

To investigate how ERF109 influences cellular Fe homeostasis in leaves, RT-qPCR of a subset of Fe-responsive genes was selected, including four clade Ib bHLH proteins (bHLH38/39/100/101) that are critically required for the induction of all Fe deficiency responses (Wang et al., 2013), a gene encoding the Fe chelator nicotianamine (NICOTIANAMINE SYNTHASE4, NAS4) that is involved in metal translocation within the plant (Nozoye, 2018), and Fe storage protein FERRITIN1 (FER1) (Harrison and Arosio, 1996). The expression of bHLH38/39/101 was significantly increased in erf109 mutant plants and even more so after the Fe deficiency treatment (Figure 6). The expression of NAS4 and FER1 was higher in erf109 than in Col-0 plants in both Fe-sufficient and Fe-deficient conditions (Figure 6). Mutations in ERF109 increased bHLH38/39/101 and NAS4 expression, mimicking Fe deficiency. While FER1 was down-regulated in Fe-deficient Col-0 plants, FER1 transcript levels were increased in erf109 mutants under Fe-sufficient conditions, reflecting the higher Fe content in leaves of erf109 mutants (Figure 6).

To validate the differential expression pattern observed in the RNA-seq analysis, RT-qPCR of DEGs selected from the groups of defense-related, photosynthesis-related, and chloroplast-encoded genes was carried out. The criterion for the selection of these DEGs was based on the significance of the fold-changes and the molecular function of the genes. In wild-type plants, ERF109 was down-regulated in response to Fe-deficiency treatment; no ERF109 transcripts could be detected in the knock-out lines (Figure 7). Decreased expression of the calmodulin-like protein CML37, transcription factors involved in the defense response, in negotiating between abscisic acid (ABA) and ethylene (WRKY40, ERF13), and the exocyst subunit EXO70 family protein B2 (EXO70B2) was observed in Fe-deficient Col-0 plants and in erf109 mutant plants under both Fe-sufficiency and Fe-deficiency (Figure 7). By contrast, the expression of the ferredoxin-NADP reductase FNR2, photosystem II light-harvesting protein LHCB3, and the chlorophyll biosynthesis gene GUN4 was significantly increased upon Fe-deficiency in Col-0 but not in erf109 mutant plants (Figure 7). CONSERVED IN THE GREEN LINEAGE AND DIATOMS 27 (CGLD27) was up-regulated in both Fe-deficient Col-0 and erf109 plants (Figure 7). However, chloroplast-encoded genes such as the subunit of the chloroplast NAD(P)H dehydrogenase complex NDHI, photosystem II reaction center protein PSBH, and a protein required for photosynthesis I assembly and stability (YCF4) were only responsive to Fe deficiency in Col-0 plants (Figure 7). All expression patterns were in accordance with the RNA-seq data.