The HSF-like Transcription Factor TBF1 Is a Major Molecular Switch for Plant Growth-to-Defense Transition

TBF1 Is a TF that Binds to the TL1 cis Element Enriched in Defense-Related Gene Promoters

The TL1 cis element (consensus sequence GAAGAAGAA) in the ER-resident genes is essential for their activation in response to SA induction [3]. We examined publically available microarray data (see Supplemental Information available online) and found that the TL1 cis element is enriched in promoters of genes regulated by elf26 (the first 26 amino acids of EF-Tu) (p < 0.001) and flg22 (p < 0.01), indicating that this novel element may play a role in MTI. Interestingly, we detected no significant enrichment of TL1 when all of the BTH (benzothiadiazole; an SA analog)-affected promoters were analyzed even though the element was first discovered in the SA-induced, NPR1-dependent ER-resident genes [3].

Next, we searched the TF that binds to TL1 using the TFSEARCH database and found HSFs of Saccharomyces cerevisiae and Drosophila melanogaster as potential candidates. The Arabidopsis genome contains 21 HSF-like genes. Several reports have indicated the involvement of the HSFs in immediate heat response, acquired thermotolerance, sensing of reactive oxygen species (ROS), and seed development [11, 12]. To identify a candidate gene for TBF1, we first examined publically available expression data for profiles of the Arabidopsis HSF family members (see Supplemental Information). Only one gene family member, HSF4 (also known as HsfB1; At4g36990), was strongly induced by BTH as well as virulent and avirulent Pseudomonas syringae pv. maculicola (Psm) ES4326 bacteria. Because Arabidopsis HSF4 and its tomato homolog do not functionally complement the yeast hsf1 mutant strain [13] (Daniel Neef and Dennis Thiele, personal communication) and its overexpression has little effect on heat-shock protein expression or thermotolerance [14, 15], we speculated that HSF4 does not encode a typical heat-shock factor. Its pathogen-inducible expression pattern suggests that it has a novel function related to plant immunity and is a candidate for TBF1.

In the current study, we demonstrated HSF4 to be the TL1-binding TF (TBF1) through a yeast one-hybrid (Y1H) assay in which the fragment of BiP2 promoter containing multiple functional TL1 cis-elements [3] was used as bait. We constructed two yeast bait strains containing the wild-type (WT) and the mutant (mTL1) BiP2 promoters, respectively (Figures S1A and S1B). Expression of TBF1-AD (activation domain) in Strain 1 activated both HIS3 and LacZ reporters driven by the WT BiP2 promoter (Figures 1A and 1B). The binding specificity of TBF1 to TL1 was confirmed in Strain 2 where the two single nucleotide substitutions in the mTL1 core binding sequence blocked the induction of LacZ, whereas the control HIS3 reporter with the WT TL1 was induced normally (Figure 1B).

TBF1 binding to the TL1 cis element was further demonstrated by electrophoretic mobility shift assay using protein extracts from WT and an insertional knock-out TBF1 mutant, tbf1 (Figures S1C and S1D). WT displayed an upshifted band whose intensity was further enhanced in the extract made from plants treated with SA (Figure 1C). This band was diminished in the tbf1 mutant extract, indicating that TBF1 is required for the DNA-protein complex formation. Competition assay using nonradioactive TL1 and mutant mTL1 probes indicated that the observed TBF1 binding was specific to the TL1 consensus sequence.

To test TBF1 DNA-binding activity in vivo, we generated transgenic tbf1 plants expressing a translational fusion between TBF1 and GFP driven by the endogenous TBF1 promoter (TBF1p:TBF1-GFP). Because the fusion protein was proven to be biologically active through genetic complementation of the tbf1 mutant phenotype (Figure S1E), we used it for chromatin immunoprecipitation (ChIP). As shown in Figure 1D, using six pairs of primers spanning different regions of the BiP2 promoter, we detected sequence enrichment corresponding to the TL1-containing region 2 in both uninduced and SA-treated samples, region 3 in the SA-treated sample, and region 4 in the uninduced sample. No enrichment was found in regions 1, 5, and 6 that do not contain the TL1 element.

TBF1 Is a Major Molecular Switch Involved in Transcriptional Reprogramming Induced by SA and elf18

In the test for TBF1 dependency, we found that the induction of BiP2 and CRT3, whose promoters contain multiple copies of TL1 elements, was compromised in the tbf1 mutant and in npr1-1 (Figure 2A). Consequently, the BiP2 protein accumulation was inhibited in the SA-treated tbf1 mutant plants (Figure S2A). In contrast, induction of two other ER-resident genes, BiP3 and CRT1, which do not have TL1 in their promoters, was not affected in tbf1 (Figure 2B), confirming the specificity of TBF1 to TL1.

We next performed a genome-wide transcriptional profiling on WT and tbf1 plants upon SA and elf18 treatments to determine the global effect of TBF1. We found 1,269 and 1,792 TBF1-dependent genes differentially regulated by SA and elf18, respectively (p < 0.05, fold change > 2). However, only ~8% of these genes were regulated by both signals (Figure 2C; Table S1), indicating that TBF1 controls distinct output genes in SAR and MTI. Moreover, the total numbers of significantly induced and repressed genes (the top heatmaps in Figures 2D and 2E), the degrees of TBF1 dependency (the middle heatmaps), and the numbers of TL1 cis elements present in the gene promoters (the bottom heatmaps) indicate that TBF1 plays a greater role in SA- and elf18-mediated transcription repression than in induction. This finding is in agreement with the previous work indicating that class B-Hsfs mainly act as repressors of target gene expression [16, 17].

Gene ontology (GO) analysis identified a significantly enriched cluster of SA-induced secretory pathway genes (p ≤ 0.001; Figure 2D “membrane proteins,” Table S1), recapitulating our earlier findings [3]. In addition, we identified several major SA- and elf18-mediated and TBF1-dependent functional categories including genes known to encode defense-related proteins, ribosomal proteins, chloroplast function-related proteins, and abiotic stress signal transducers (Figures 2D and 2E; Table S1; described in more detail in Supplemental Information) [18]. We postulate that these genes are likely directly controlled by TBF1 because their regulatory regions contain the TL1 element (Table S1).

Expression changes observed in the microarray experiment were further confirmed through qRT-PCR of independent biological samples on 26 selected genes representing various GO categories listed in Figures 2D and 2E (Figures S2B–S2E).

TBF1 Plays a Key Role in the Growth-to-Defense Transition

Upon SA treatment, TBF1 downregulates genes encoding chloroplast proteins (Figure 2D)—an effect that is known to be associated with SA [19]. Interestingly, chloroplast function-related genes were even more profoundly repressed by elf18 (Figure 2E), indicating that TBF1 might play a general role as a major molecular switch in the growth-to-defense transition. To test this, we first measured the growth of both WT and the tbf1 mutant plants. As shown in Figure 3A, in the absence of SA or elf18, the tbf1 mutant grew at a similar rate as WT. In the presence of elf18 or increasing SA concentrations, however, the growth of the WT plants was significantly inhibited. This effect was partially alleviated in the tbf1 mutant. In contrast, another MAMP signal, flg22, exerted a similar growth-suppressing effect on both WT and tbf1 seedlings (Figure S3A).

We performed an additional series of tests to determine the stress responses mediated by TBF1. Even though the tbf1 mutant has been shown to have normal heat-induced marker gene expression (Figure S3B) and plays no detectable role in the heat-shock response [20], its defect in the induction of multiple chaperone genes prompted us to examine the unfolded protein response (UPR). When the plants were treated with the UPR inducer tunicamycin, seedling survival rate for tbf1 was only ~20% compared to ~60% for WT (Figure 3B), indicating that TBF1 plays a role in UPR.

The capacity of TBF1 to efficiently secrete antimicrobial PR1 protein was evaluated next. Whereas the PR1 transcript induction and total PR1 protein levels were unchanged in tbf1 (Figure 3C; Figure S3C), secreted PR1 in the intercellular wash fluid was dramatically reduced in the tbf1 mutant, as in the control bip2 dad1 double mutant [3], compared to WT. This phenotype was complemented in the transgenic line carrying a genomic fragment encompassing upstream regulatory sequences (3,554 bp) and the coding region (1,047 bp) of TBF1 (TBF1 complementation). The defect in secretion of anti-microbial proteins in the tbf1 mutant correlated with more than 1 log higher growth of the bacterial pathogen Psm ES4326 compared to WT and the complementation line (Figure 3D). In response to SA induction, less than 1 log reduction in Psm ES4326 growth was observed in the tbf1 mutant as opposed to the ~2 log reduction detected in the WT plants (Figure 3E). SA-inducible defenses were restored in the TBF1 complementation line, whereas npr1-1 was completely deficient in establishing resistance.

We next examined whether the tbf1 mutant is capable of mounting effective MTI using elf18 and flg22. WT plants pre-treated with either signal showed a more than 1 log reduction in Psm ES4326 growth compared to mock-treated samples (Figure 3F; Figure S3D). The tbf1 mutant, on the other hand, completely failed to establish the resistance induced by elf18. Interestingly, this defect is specific to elf18 because flg22-induced resistance was intact in the tbf1 mutant (Figures S3D and S3E). Although previous studies showed that flg22 and elf18 induce largely overlapping sets of genes [21], our results, however, are consistent with the genetic data showing that the perception of elf18, but not flg22, specifically requires the ERQC mechanism [6, 7, 22, 23] and with the fact that TBF1 controls the induction of these ERQC genes (Figure 2A). The molecular mechanism underlying this TBF1 requirement has yet to be determined.

Another observation from the Psm ES4326 infection experiment was the near-normal response to elf18 observed in the SA-insensitive npr1-1 mutant (Figure 3F). Even though there was an overall increase in Psm ES4326 growth in npr1-1, elf18 pretreatment could still result in a more than 1 log reduction in pathogen growth similar to that observed in the WT. This result is in accordance with the expression profiling data showing that elf18 and SA induce distinct sets of genes (Figure 2C). Moreover, a recently published study demonstrated that MTI induced by flg22 and elf18 is largely intact in the sid2 mutant, which is deficient in SA synthesis [24]. MTI and SAR are two temporally and spatially separate immune responses. The former occurs locally and immediately upon pathogen challenge, whereas the latter is a systemic response induced after the local response.

Translation of TBF1 Is Controlled by uORFs Sensitive to Cellular Metabolic Changes

Our expression profiling and genetic data show that TBF1 is a major molecular switch that, upon pathogen challenge, turns on multiple defenses and inhibits primary growth and development (Figures 2 and 3). To understand how TBF1 itself is regulated, we analyzed its expression patterns upon SA treatment and detected the maximum accumulation of TBF1 transcript 4 hr after the treatment (Figure 4A). We further demonstrated that functional NPR1 is required for SA-mediated TBF1 transcription. Interestingly, TBF1 also plays a role in regulating NPR1 as the NPR1 transcript levels were diminished in tbf1 (Figure 4B), suggesting a feedback mechanism between these two key immune regulators.

In addition to transcriptional regulation, analysis of the TBF1 mRNA through the 5′ and 3′ rapid amplification of cDNA ends (RACE) detected two additional upstream open reading frames (uORFs) 5′ of the TBF1 start codon (Figure 4C). The second uORF (At4g36988) is well-conserved among TBF1 homologs in other plant species [25]. To test whether these two uORFs influence translation initiation of TBF1, we made a fusion between the 5′ UTR of TBF1 containing both uORFs, the first exon of TBF1 and the GUS reporter (abbreviated as uORF1-uORF2-GUS). We also created three additional constructs with the start codon mutated (ATG to CTG) for uORF1 (uorf1-uORF2-GUS), uORF2 (uORF1-uorf2-GUS), or both uORFs (uorf1-uorf2-GUS). These reporter constructs were driven by the 35S promoter to allow detection of only translational differences. We first transiently expressed these constructs in Nicotiana benthamiana leaves and quantified the GUS activities (Figure 4D). Using activity of the WT construct as a control, we detected 1.5- and 3.5-fold increases in GUS activities in uorf1-uORF2-GUS and uORF1-uorf2-GUS, respectively. Like the uorf2 mutant, mutating both uORFs in uorf1-uorf2-GUS resulted in a 3.5-fold elevation in GUS activity over WT. These observations suggest that both uORFs have inhibitory effects on TBF1 translation, with the effect of uORF2 epistatic to that of uORF1.

To further understand the regulatory mechanism of TBF1 translation during plant defense, we quantified GUS activities of these translational fusions in stable transgenic Arabidopsis lines in response to Psm ES4326 carrying the avirulent effector, avrRpt2. We found that perception of this avirulent bacterial strain, which can induce MTI, ETI, and SAR, caused a rapid increase in GUS activity in the uORF1-uORF2-GUS transgenic lines (Figure 4E). Interestingly, this increase was not observed in the uorf1-uorf2-GUS transgenic lines. These results suggest that the Psm ES4326/avrRpt2 challenge could rapidly alleviate the inhibitory effects of the uORFs on translation of the downstream gene. To determine whether the endogenous TBF1 was indeed translated in the plant cell upon pathogen challenge, we conducted a polysome profiling experiment (Figure 4F). We found a significant increase in TBF1 transcript in the polysomal fractions of the gradient within 30 min of Psm ES4326/avrRpt2 inoculation (Figure 4G), consistent with the results from the study using the GUS reporter (Figure 4E). Moreover, this polysomal association appeared to be transient as the TBF1 transcript decreased 1 hr post inoculation. In contrast, polysomal association of the housekeeping gene UBQ5 transcript was not significantly increased upon pathogen challenge (Figure S4).

Both uORFs of TBF1 are enriched in aromatic amino acids, particularly in phenylalanine (Phe). Intriguingly, this Phe enrichment is evolutionarily conserved for uORF2 among the TBF1 homologs in other plant species [25]. We hypothesize that pathogen challenge may result in a decrease in cellular Phe concentration and allow translation initiation of TBF1 downstream of uORFs.

Because repeated efforts in measuring Phe concentrations in plants infected with Psm ES4326/avrRpt2 have not produced consistent results, we employed a yeast-based reporter system (see Experimental Procedures). This system involves the use of the yeast strain aro7, which is auxotrophic for Phe and tyrosine, and the reporter containing the mouse DHFR (dihydrofolate reductase), which has been engineered to be less stable [26] and resistant to methotrexate (MTX) [27]. In the presence of MTX, yeast growth becomes dependent on the recombinant DHFR reporter expression. We inserted uORF1-uORF2-TBF1^{1st exon} upstream of the DHFR reporter to generate uORF1-uORF2-DHFR and transformed this construct into aro7. The resulting transformant was cultured in the presence of MTX and growth was recorded over the course of time.

As shown in Figure 5A, under the Phe-rich conditions (75 mg/L; standard Phe concentration), the yeast strain carrying the DHFR displayed a much higher growth rate than the strain carrying the uORF1-uORF2-DHFR, showing the inhibitory effects of the uORFs on DHFR translation. However, under the Phe-restricting conditions (15 mg/L), both strains grew at similar rates, suggesting that low Phe level released the inhibitory effects of uORFs on DHFR translation (Figure 5A). To ensure that the TBF1^{1st exon} –DHFR fusion protein was not toxic to yeast cells, we grew the strains in the absence of MTX and noted no significant difference in their growth (Figure S5). The uORFs of TBF1 appear to be specifically sensitive to Phe starvation because aspartic acid (Asp) starvation caused by addition of tobramycin (TOB), a known inhibitor of yeast tRNA^Asp aspartylation [28], did not eliminate the difference in growth rate between the strain carrying uORF1-uORF2-DHFR and the strain carrying DHFR.

We next sought to understand how Phe starvation alleviates uORFs-mediated repression of TBF1 translation. In yeast, amino acid starvation results in a series of events including uncharged tRNAs accumulation, activation, and autophosphorylation of GCN2, GCN2-mediated phosphorylation of eukaryotic initiation factor 2α (eIF2α), and translation initiation of GCN4 downstream of uORFs [29–32]. To determine whether a similar mechanism controls the translation of TBF1, we first performed northern blot analysis to measure charged and uncharged tRNA levels following inoculation with Psm ES4326/avrRpt2. We detected a dramatic increase in both charged and uncharged tRNA^Phe within 30 min of the bacterial inoculation, which persisted for 8 hr (Figure 5B). In contrast, only a moderate increase in charged tRNA^Asp level was observed under the same conditions. No uncharged tRNA^Asp was detected. This is consistent with our hypothesis that pathogen challenge in plants can lead to specific changes in Phe metabolism.

Because a functional and stress-inducible GCN2-eIF2α pathway has been found in Arabidopsis [33], we investigated whether the pathogen-induced accumulation of uncharged tRNA^Phe could lead to phosphorylation of eIF2α. In an immunoblot experiment, we detected a rapid accumulation of the phosphorylated eIF2α in Arabidopsis leaves infected with Psm ES4326/avrRpt2 (Figure 5C). The phosphorylated eIF2α may facilitate ribosome reattachment to the TBF1 translation start codon downstream of uORFs to initiate TBF1 translation.

Taken together, we concluded that TBF1 expression is tightly controlled not only at the transcriptional level by NPR1 but also at the translational level through uORFs. Pathogen challenge, which causes a temporary increase in uncharged tRNA^Phe accumulation, triggers eIF2α phosphorylation and consequently derepression of TBF1 translation.

Translational Analysis of uORF1-uORF2-GUS

The DNA fragment containing the 5′UTR and the first exon of TBF1 (designated uORF1-uORF2) was PCR amplified using primers TBF1 5′UTR-GW-F and TBF1 5′UTR-GW-R (Table S3) and cloned into the Gateway vector pDONR207 (Invitrogen). Two A-to-C point mutations were introduced, either separately or together, into the start codons (ATG) of uORF1 and uORF2. The WT and mutant uORF1-uORF2 sequences were inserted downstream of the constitutive 35S promoter and upstream of the coding region of the GUS reporter in pMDC140 through recombination [44]. The resulting translational reporters were transformed into Col-0 WT plants or transiently expressed in Nicotiana benthamiana using Agrobacterium tumefaciens [45]. For Arabidopsis stable transgenic lines, two independent T₃ lines homozygous for each construct were chosen for quantitative GUS assay [3] at 0, 0.5, 1, 2, 3, 4, and 8 hr after Psm ES4326/avrRpt2 infiltration (OD_600nm = 0.02). Further detailed experimental procedures are described in the Supplemental Information.