Arabidopsis HEAT SHOCK TRANSCRIPTION FACTORA1b overexpression enhances water productivity, resistance to drought, and infection.

Heat-stressed crops suffer dehydration, depressed growth, and a consequent decline in water productivity, which is the yield of harvestable product as a function of lifetime water consumption and is a trait associated with plant growth and development. Heat shock transcription factor (HSF) genes have been implicated not only in thermotolerance but also in plant growth and development, and therefore could influence water productivity. Here it is demonstrated that Arabidopsis thaliana plants with increased HSFA1b expression showed increased water productivity and harvest index under water-replete and water-limiting conditions. In non-stressed HSFA1b-overexpressing (HSFA1bOx) plants, 509 genes showed altered expression, and these genes were not over-represented for development-associated genes but were for response to biotic stress. This confirmed an additional role for HSFA1b in maintaining basal disease resistance, which was stress hormone independent but involved H₂O₂ signalling. Fifty-five of the 509 genes harbour a variant of the heat shock element (HSE) in their promoters, here named HSE1b. Chromatin immunoprecipitation-PCR confirmed binding of HSFA1b to HSE1b in vivo, including in seven transcription factor genes. One of these is MULTIPROTEIN BRIDGING FACTOR1c (MBF1c). Plants overexpressing MBF1c showed enhanced basal resistance but not water productivity, thus partially phenocopying HSFA1bOx plants. A comparison of genes responsive to HSFA1b and MBF1c overexpression revealed a common group, none of which harbours a HSE1b motif. From this example, it is suggested that HSFA1b directly regulates 55 HSE1b-containing genes, which control the remaining 454 genes, collectively accounting for the stress defence and developmental phenotypes of HSFA1bOx.

Introduction

Water limitation is experienced by all terrestrial plants and is a major evolutionary force within plant populations (Heschel ; Morison ). Moderate limitation of water availability diverts resources away from growth into protective responses restricting stomatal conductance and CO2 uptake, thus limiting photosynthesis and plant growth (Schulze, 1986, ; Boyer, 1970; Condon ; Morison ). Molecular genetic studies using Arabidopsis thaliana often define the survival of dehydration stress as drought tolerance (Liu ; Passioura, 2007), but generally do not address its effects on plant productivity. The term water productivity describes the relationship between yield of the harvestable product and water loss, and is important when looking at plant productivity in water-limiting environments (Passioura, 1977; Monteith, 1984, 1993; Condon ; Steduto ; Morison ; Bechtold ).

Drought is often accompanied by elevated air and leaf temperatures; therefore, leaves experience additional evaporative demand due to an increase in leaf to air vapour pressure difference (VPD; Turner, 2004). Consequently, there may be cross-talk between heat and dehydration stress signalling networks. For example DREB2A, a dehydration-responsive transcription factor (TF), has been shown to have a dual function in Arabidopsis, regulating the responses to dehydration and heat stress (Sakuma ). The signalling by DREB2A is routed through activation of a heat shock TF gene, HSFA3, leading to the expression of heat shock protein genes (Schramm ; Yoshida ). Similarly, MULTIPROTEIN BRIDGING FACTOR1c (MBF1c) has been proposed to regulate the response to temperature stress in Arabidopsis (Suzuki ). MBF1c overexpression leads to improved tolerance to heat, osmotic, and biotic stress (Suzuki ), and its regulon includes DREB2A and two class B HSF genes (Suzuki ).

In all eukaryotes, HSFs are the central component of the cellular heat stress response, with their basic structure and the cis regulatory heat shock elements (HSEs) of their target genes being highly conserved (Wu, 1995; Morimoto, 1998; Nover ; Baniwal ). Plants, compared with all other eukaryotes, have large HSF families (Nover ; Czarnecka-Verner ). For example, the Arabidopsis HSF family consists of 21 genes, with the proteins they encode being divided into three structural classes (A, B, and C; Nover ; Kotak , 2007). There are 15 class A HSFs, which act as transcription activators. The B class HSFs may be transcriptional repressors and/or co-activators by interacting with class A HSF genes (Czarnecka-Verner ; Hahn ; Ikeda ). No function has been ascribed to the single HSFC1 in Arabidopsis (Nover ).

The four clade A1 members in Arabidopsis (HSFA1a HSFA1b, HSFA1d, and HSFA1e) are key regulators in the early response to heat (Lee ; Prändl ; Panchuk ; Busch ; Yamada ; Liu ). When HSFA1b is overexpressed in Arabidopsis and tomato (Solanum lycopersicum) it improves thermotolerance (Prändl ) and heat-shock induced chilling tolerance, respectively (Li ). Similarly in tomato, SlHSFA1 is a regulator of thermotolerance, and reduced expression of SlHSFA1 leads to heat stress sensitivity (Mishra ). However, the size of plant HSF gene families has prompted the suggestion that some HSFs have evolved to regulate responses to other stresses (Kotak ). Many HSF genes are induced in response to environmental stresses other than heat (Miller and Mittler, 2006; Swindell ). For example, HSFA2 shows strong transcriptional activation during high light, anoxia, salinity, and bacterial infection (Panchuk ; Rizhsky ; Miller and Mittler, 2006; Nishizawa ; Schramm ; Banti ). HSFA9 has a key role in seed maturation (Kotak ), and rice OsHSFA4a is implicated in tolerance to cadmium toxicity (Shim ). It has been demonstrated that HSFB1 and HSFB2b are negative regulators of resistance to Alternaria brassicicola (Kumar ). This agrees with observations that induced thermotolerance can have negative effects on basal and R gene-mediated resistance (Noël ; Wang ). Thus, a role for HSFs in plants’ responses to a range of environmental stresses appears likely.

Due to the connection between heat and drought stress, the hypothesis was tested that among the Arabidopsis class A HSF family, some could play a direct role in a drought response independent of heat shock, and display dual functions similar to DREB2A. Furthermore, given the influence some HSF A class genes have on growth and development (Kotak ; Liu ), their influence might extend to effects on lifetime traits such as water productivity (Morison ). To test this hypothesis, the focus of this study was on two HSF genes, HSFA1b and HSFA2, whose constitutive overexpression promotes thermotolerance and enhances expression of genes responsive to leaf water status such as DREB2A and ASCORBATE PEROXIDASE2 (APX2; Panchuk ; Nishizawa ; Ogawa ; Galvez-Valdivieso ).

Material and methods

Plant material and growth conditions

Plants were grown in both controlled-environment and glasshouse conditions exactly as described by Bechtold . All transgenic 35S:HSFA1bOx lines were produced by Agrobacterium-mediated transformation using the floral dip method (Clough and Bent, 1998) in three different laboratories: Tübingen (HSFA1bOx1/Ws-2), Essex (HSFA1bOx2 and -3/Col-0), and Warwick (HSFA1b::mRFP_B/Col-0). The transgenic and mutant genotypes hsfA1a/hsfA1b, HSFA2Ox, hsfA2-1, MBF1cOx, and mbf1c-1 have been described previously (Busch ; Suzuki , 2008; Nishizawa ).

Oil seed rape transformation

A doubled haploid Brassica napus L. cv. Q6 was transformed using an Agrobacterium-mediated tissue culture approach as previously described (Sparrow ) using a pGreen Ti vector harbouring a Cauliflower mosaic virus (CaMV) 35S: nptII coding sequence (Supplementary Fig. S3A available at JXB online; Hellens ) and a 35S:AtHSFA1b cDNA fusion from pJIT30 (Guerineau ). Segregation of T-DNA loci in progeny from the primary transformant was determined by PCR of genomic DNA using primers for the nptII gene (forward, 5´-TGAATGAACTGCAGGACGAG-3´; reverse, 5´-AGCCAACGCTATGTCCTGAT-3´). The two transgenic lines generated in this study were confirmed to be independent by DNA gel blotting (Fig. S3B) using standard procedures described previously (Hellens ). The blot was probed with a 32P-labelled nptII DNA fragment amplified by PCR from pGreen0029 (Hellens ) and washed at 0.1× SSC (65 °C).

Application of stresses

Five-week-old plants were subjected to a 15min heat stress at 36 °C, 78% relative humidity maintaining VPD at 1 kPa. Drought stress and determination of water productivity, rosette biomass, and harvest index (HI) were carried out exactly as described previously (Bechtold ). Virulent Pseudomonas syringae pv tomato DC3000 (Pst) infection were carried out on 4-week-old plants by vacuum infiltration, or dipping in cultures of the bacteria at a density of 105 colony-forming units (cfu) ml–1 in 10mM MgCl2 buffer containing 1% (v/v) Silwett L-77. Bacterial growth in leaves was monitored at 0, 2, and 4 d post-infection as described by Innes . For oomycete infection, Hyaloperonospora arabidopsidis strain WAC09 (Hpa) was used. Spores were extracted in 10ml of water from plant leaves prior to infection, and 2-week-old seedlings were inoculated and counted as described by Muskett .

Gas exchange measurements

Transpiration and photosynthesis parameters were measured as previously described (Lawson and Weyers, 1999). Briefly, the response of A to changes in the intercellular CO2 concentration (c ), and the response of A to changes in photosynthetic photon flux density (PPFD) from saturating to subsaturating levels was measured using a combination of red and white LEDs (PP Systems, Amesbury, MA, USA) at ambient CO2 concentration (390 μmol mol–1), leaf temperature of 20 (±1) °C, and a VPD of 1 (±0.2) kPa. Snapshot measurements were carried out in the glasshouse, and readings were taken at steady-state rates of A and stomatal conductance (g s) at current atmospheric [CO2].

Hormone, hydrogen peroxide (H2O2), and glutathione (GSH) determinations

H2O2 and GSH measurements were carried out as described previously (Bechtold ). Hormone measurements were analysed using an adapted method described in Forcat . A 20mg aliquot of freeze-dried leaf samples was extracted three times for 1h in a rotary extractor using 10% (v/v) methanol, 1% (v/v) acetic acid in a total volume of 1100 μl. After the third extraction step 200 μl of the samples were used to analyse salicylic acid (SA) and jasmonic acid (JA) levels as described in Bechtold .

Transcriptomics, gene expression, and microarray data analysis

RNA was extracted from fully expanded leaves of stressed and non-stressed plants exactly as described by Bechtold . The comparison of Ws-2 versus HSFA1bOx1 was carried out using Agilent Arabidopsis 3 Oligo Microarrays, and the data were processed as previously described (Bechtold ). Three independent experiments using individual rosettes per experiment and one dye swap were carried out. Up-regulated genes were determined as >2-fold in HSFA1bOx1 at a 5% false discovery rate (FDR). Raw data from these experiments and for those from HSFA2Ox1 plants can be found on the ArrayExpress database at http://www.ebi.ac.uk/microarray-as/ae. Quantitative real-time reverse transcription–PCR (qRT–PCR) of cDNA using the SYBR Green (Sigma Ltd, UK) chemistry was carried out as described previously (Bechtold ) using the primers listed in Supplementary Table S10 at JXB online. The data were normalized against cyclophilin (Rossel ).

CATMA arrays

RNA extraction from rosettes of four pooled plants, and the comparison of Col-0 versus HSFA2Ox was carried out using the CATMA (version 3) microarray (Allemeersch ). Normalization and analysis of the array was carried out using LimmaGUI, a graphical front end for the limma (Linear Models for MicroArray; Wettenhall and Smyth, 2004) package for R available from Bioconductor (http://www.bioconductor.org).

Bioinformatics

Public data sets

The publicly available microarray data sets accessed for this study (see legend of Fig. 4) can be found at ArrayExpress (http://www.ebi.ac.uk/arrayexpress). Raw fluorescence data from 35S:MBF1c (ID: E-GEOD-5539) arrays were downloaded from the Array Express database and analysed as previously described for Affymetrix microarrays (Bechtold ).

Fig. 4.

The HSE1b motif is recognized in vivo in the promoters of TF genes regulated by HSFA1b overexpression. (A) The coloured letters show the consensus sequence, generated by MEME (see the Materials and methods), for the HSE1b motif present in the promoter regions of 55 HSFA1bOx1-up-regulated genes. (B) Expression, determined by qRT–PCR of HSE1b-containing TF genes in hsfA1a/hsfA1b plants subjected to 15min at 22 °C (white bars) or 37 °C (grey bars). VPD was maintained at 1 kPa. All differences in the heat-stressed samples are significant (P < 0.05; Student’s t-test) except for ZAT6. (C) Expression of the HSE1b-TF genes and HSFA1b-responsive genes without the HSE1b element (below the dotted line) in the presence (white bars) and absence (grey bars) of the protein synthesis inhibitor cycloheximide in plate-grown seedlings. The data are the means (±SEM) of two separate experiments, totalling six plates per treatment and three technical replicates per assay. (D) PCR amplification of ChIP promoter fragments of three genes containing a single HSE1b element (see Supplementary data) but no canonical HSE element (+HSE1b no HSE). The same procedure was carried out on three genes containing canonical HSE motif(s) but no HSE1b motif (no HSE1b +HSE; see Supplementary data). Gels showing PCR amplicons from positive control, input DNA (lane 1); negative control, no antibody control precipitation (lane 2); and ChIP DNA (lane 3). The result presented here is one of four representative experiments. The chromatin was immune-precipitated from fully expanded leaves of non-stressed 5-week-old HSFA1b-mRFP_B plants with anti-RFP antibody. (E) ChIP-PCR of the promoter regions of the seven HSE1b-TF genes from the same immune-precipitated samples as in D.

Analysis of Gene Ontology (GO)

This was done using the software packages in the Database for Annotation, Visualization and Integrated Discovery v6.7 (DAVID; http://david.abcc.ncifcrf.gov/; Huang ).

Promoter analysis using MEME

Sequences 500bp upstream of the predicted transcription start site were retrieved from TAIR (version 8; www.arabidopsis.org). Promoters of the top 50 expressed genes were searched using MEME (3.5.7). A position-specific scoring matrix (PSSM) was generated and used to scan the promoters of all 352 up-regulated genes to identify those enriched for occurrences of the motif (Supplementary Table S7 at JXB online). To identify putative direct targets of HSFA1b, each promoter sequence from genes up-regulated in HSFA1bOx1 was scored for over-representation of the HSE1b motif detected by MEME. For each promoter, the matrix similarity score (Kel ) was computed at each position in the sequence. A P-value for each score was computed from a score distribution by applying the HSE1b PSSM to a random sequence 100 million bases in length, which was generated by a third-order Markov model learned from the whole Arabidopsis genome. A score for potential multiplicity was calculated by taking the top k non-overlapping hits and computing the binomial probability for the presence of k sites within the sequence of length n. Genes that had a binomial P-value ≤0.05 were classed as over-represented for the motif and therefore putative direct targets of HSFA1b. Regulatory sequence analysis was performed using the APPLES software framework to scan for hits of the MEME motif in promoters using binomial testing.

Chromatin immunoprecipitation followed by PCR (ChIP-PCR)

ChIP was carried out using fully expanded leaves of 5-week-old HSFA1b:mRFP_B plants according to Saleh . A red fluorescent protein (RFP)-specific antibody (anti-RFP, AB62341; Abcam, Oxford, UK) was used to precipitate the HSFA1b:mRFP–DNA complexes from chromatin. The primers used were promoter specific, spanning the HSE1b or canonical HSE elements in the respective genes (Supplementary Table S10 at JXB online). PCR was carried out on the immunoprecipitated DNA, input DNA (before precipitation), and on no antibody control-precipitated DNA. Products were separated on 1.5% (w/v) TAE agarose gels and visualized under UV light after staining with ethidium bromide.

Cycloheximide treatments

Plants were grown aseptically on half-strength Murashige and Skoog (MS) medium containing 3% (w/v) sucrose and incubated under the short-day growth conditions described above. Plates of 10-day-old AtHSFA1bOx-1/Ws-2 and Ws-2 seedlings were sprayed with 10mM cycloheximide or water and after 4h their RNA was extracted and used to determine gene expression. Expression of selected genes was determined by qRT–PCR as described above.

Results

Initially, three transgenic Arabidopsis lines were used for this study, HSFA1bOx-1/Ws-2, HSFA1bOx-2/Col-0, and HSFA2Ox/Col-0. These lines showed 50- to 160-fold, overexpression compared with their wild-type controls (Supplementary Fig. S1A–C at JXB online). A HSFA1bOx-3/Col-0 line with 34-fold induction of HSFA1b was also included in part of the study (Supplementary Fig. S1A), as was a line overexpressing an HSFA1b-RFP fusion (Supplementary Fig. S2A). HSFA2 expression was not affected in HSFA1bOx and vice versa (Supplementary Fig. S1B, C). An hsfA1a/hsfA1b double mutant was also used because single mutants do not have diminished responses to heat stress (Busch ). HSFA1b is expressed in all organs under a range of environmental conditions (Supplementary Fig. S1D; Miller and Mittler, 2006). Consequently, expression of the CaMV 35S:HSFA1b transgene was enhanced but not ectopic.

HSFA1b is involved in dehydration and drought stress responses

Detached rosettes of HSFA1bOx plants dehydrated more slowly, while hsfA1a/hsfA1b plants showed the opposite phenotype (Fig. 1A). At 20% relative soil water content (rSWC), intact HSFA1bOx plants did not wilt, unlike wild-type controls (Fig. 1B). Increased dehydration tolerance may divert resources away from growth and limit photosynthesis (Passioura, 2007; Morison ). Therefore, seed yield was measured in these plants when well watered and after exposure to progressive drought stress to 20% rSWC, followed by re-watering. HSFA1bOx plants showed reduced soil drying rates compared with controls (Fig. 1C). In contrast, hsfA1a/hsfA1b plants had significantly faster drying rates (Fig. 1C). Seed yield was elevated in both watered and droughted HSFA1bOx lines compared with the wild type (Fig. 1D), although no reciprocal difference was observed in hsfA1a/hsfA1b plants.

Fig. 1.

HSFA1b regulates drought tolerance, seed yield, and water productivity. (A) Rate of weight loss of detached 5-week-old rosettes of HSFA1b-overexpressing lines (HSFA1bOx1 and HSFA1bOx2) and the hsfA1a/hsfA1b double null mutant, and their wild-type controls Ws-2, Col-0, and Ws-0, respectively (n=6). (B) Typical phenotype of HSFA1b-overexpressing plants compared with their controls after 14 d without water. (C) Rate of water loss in HSFA1bOx and hsfA1a/hsfA1b plants compared with their wild-type controls averaged over 13 d with no watering (n=8). (D) Seed yield in HSFA1b-overexpressing plants and their wild-type controls after water withdrawal to 25% relative soil water content (rSWC) followed by re-watering to seed set (grey bars; n=8) or watered controls throughout this period (black bars; n=8). (E) Water productivity in HSFA1bOx and hsfA1a/hsfA1b plants in well-watered (80% rSWC, n=10; black bars) and water-limited (40% rSWC, n=10; grey bars) conditions. (F) Harvest index from the plants in E. All data are presented as means (±SEM). The asterisks (*) denote significant differences (P ≤ 0.05; Student’s t-test) between the overexpressing or mutant lines and their controls.

HSFA1b overexpression influences seed yield under limited watering conditions

The robustness of the seed yield phenotype was tested in either well-watered (80% rSWC) or water-limited (40% rSWC) growth regimes. The ratio of seed yield to total above-ground biomass (HI) is a component of water productivity (Passioura, 2007; Morison ; Bechtold ). At 40% rSWC, HSFA1bOx plants showed significant increases in water productivity and HI compared with their wild-type controls (Fig. 1E, F; see also Fig. 6F, G). Conversely, hsfA1a/hsfA1b plants showed lowered water productivity and HI at 40% rSWC (Fig. 1E, F). Significant differences were also detected in the 80% rSWC treatment of HSFA1bOx2 plants (Fig. 1E, F; see also Fig. 6F, G). No consistently significant effects of HSFA1b overexpression were observed for seed weight or viability (Supplementary Table S1 at JXB online). There was no effect of altered HSFA1b expression on the capacity for photosynthetic carbon assimilation, stomatal conductance, or instantaneous transpiration efficiency (Supplementary Table S2).

Overexpression of HSFA1b in oil seed rape increases harvest index and seed yield

To establish whether the effects of HSFA1b overexpression are conserved in other Brassicaceae, the 35S:HSFA1b chimaeric gene was transformed into B. napus (oil seed rape). Two independent single locus transgenic oil seed rape lines (BnHSFA1bOx#1 and BnHSFA1bOx#3) overexpressing Arabidopsis HSFA1b 109-fold (SD±33; n=4) and 59-fold (SD±32; n=4), respectively, showed the same improved productivity traits of seed yield and HI (Supplementary Fig. S3C, D at JXB online). BnHSFA1bOx plants had a bushier flowering phenotype than their azygous siblings (Supplementary Fig. S3E).

HSFA1b overexpression influences basal resistance to two pathogens

While HSFA1b overexpression positively influences plant productivity, drought tolerance, and thermotolerance, such plants could be more susceptible to pathogens (see Introduction). However, HSFA1bOx plants showed increased resistance to the bacterial pathogen Pst after inoculation either by vacuum infiltration (Fig. 2A) or by dipping (Fig. 2B), and Hpa (Fig. 2D). Conversely, hsfA1a/hsfA1b plants showed decreased resistance to these pathogens (Fig. 2C, D).

Fig. 2.

HSFA1b regulates basal resistance to a bacterial and an oomycete pathogen. Colonization of virulent Pst on HSFA1bOx plants (A, B) or hsfaA1a, hsfA1b, hsfA1a/hsfA1b knockout mutants (C) compared with wild-type controls at 2 d (white bars) and 4 d (grey bars) post-inoculation. Bacteria were inoculated by vacuum infiltration (A, C) or by dipping (B). Data are representative of at least two independent experiments for each method (n=6). The inocula recovered from leaves at day 0 were 2.37 log cfu ml–1 (± 0.43). (D) Spore yields from 12-day-old HSFA1bOx and hsfA1a/hsfA1b plants (n ≥7) inoculated 5 d previously with 5×104 spores of Hyaloperonospora arabidopsidis pv. WACO9. The asterisks (*) denote significant differences (P ≤ 0.05; Student’s t-test) between the overexpressing or mutant lines and their controls.

HSFA1b overexpression effects the expression of >500 genes

A pair-wise comparison of the transcriptome of HSFA1bOx1 with Ws-2 under non-stress conditions revealed 352 and 157 differentially expressed genes (DEGs) with >2-fold and <0.5-fold altered expression, respectively (P <0.05; FDR q <0.05; Supplementary Table S3 at JXB online). Many of these DEGs were classed as heat stress responsive (Supplementary Tables S3, S4) and highly significantly overlapped with microarray data from heat-stressed plants (Table 1). The overlap with microarray data sets from plants infected with virulent Pst and with microarray data collated from infection of Arabidopsis with three different isolates of Hpa (Eulgem ; Supplementary Table S5) was also highly significant (Table 1; Supplementary Table S3). However, the significance of the overlap between HSFA1b-regulated genes and drought-responsive genes was much lower in comparison with those affected by heat or pathogen infection (Table 1; Supplementary Tables S3, S4). While for heat, Pst, and Hpa, there were many HSFA1bOx1-responsive genes common to two or all stresses (Fig. 3A), 63, 53 and 30%, respectively, were responsive only to a single stress (Fig. 3A).

Table 1.

Hypergeometric distribution test for commonality of DEGs from publicly available microarrays of stress-exposed Arabidopsis plants and the HSFA1bOx1/Ws-2 comparison (Supplementary Table S3 at JXB online)

	DEGs from the publically available stress microarray data (n)	Genes from the stress microarrays present in the HSFA1bOx1/Ws-2 data set (n)	Overlapping DEGs (n)	P-value
Heat	815	397	161	1.4×10–133
Drought	4407	397	94	0.003
Pst	1314	397	124	6.5×10–61
Hpa	224	147	33	7.7×10–38

Fig. 3.

HSFA1b-responsive genes are also responsive to heat stress, H2O2, and infection by Hpa and Pst. (A) The Venn diagram shows the overlap of HSFA1bOx1-responsive genes with those responsive to heat stress (database ID, E-GEOD-5628), infection with virulent Pst (E-GEOD-5520), or Hpa (up to three isolates; Supplementary Table S5 at JXB online; Eulgem ). The significances of the overlaps can be found in Table 1. (B) Expression of six HSFA1b-responsive genes (mean ±SEM; n ≥4) classified as controlled by H2O2-mediated signalling (Supplementary Table S4 at JXB online; Gadjev ). The data are from HSFA1bOx2 (white bars), HSFA1bOx3 (black bars), and HSFA1b:mRFP_B (grey bars) 5-week-old non-stressed plants and Col-0 controls using real-time qRT–PCR. Values are significant between transgenic lines and Col-0. (C and D) Foliar levels of H2O2 (C) and GSH (D) for HSFA1bOx, hsfA1a/hsfA1b, and wild-type 5-week-old non-stressed plants (n=6). The differences marked with an asterisk (*) are significant at P ≤ 0.05 (Student’s t-test).

H2O2 signalling but not stress hormone signalling is stimulated in HSFA1bOx plants

GO analysis revealed no significant enrichment of abscisic acid (ABA)- and SA-responsive DEGs in the HSFA1bOx1/Ws-2 microarray data set (Supplementary Table S6 at JXB online). There was significant enrichment of JA-responsive genes (Supplementary Table S6), but their increased expression in HSFA1bOx2 or -3 plants could not be verified. The levels of SA, JA, and ABA were not consistently affected by HSFA1b overexpression in the different HSFA1bOx lines (Supplementary Table S7).

From the same GO analysis, there was enrichment of H2O2-responsive genes (Supplementary Table S4 at JXB online). A more specific group of H2O2-responsive genes, based on expression patterns in mutants with altered H2O2 levels (Gadjev ) also revealed a significant overlap (P ≤ 0.0001; hypergeometric distribution test; Supplementary Table S4). Six of these genes, chosen as the most differentially expressed in HSFA1bOx1 plants (Supplementary Tables S3, S4), were confirmed as such in the Col-0 HSFA1bOx lines (Fig. 3B). HSFA1bOx and hsfA1a/hsfA1b plants had higher and lower foliar levels of H2O2, respectively (Fig. 3C). Foliar H2O2 levels are often associated with increased levels of the thiol antioxidant GSH (see Discussion), and these were significantly elevated in the HSFA1bOx lines (Fig. 3D).

Fifty-five promoters of HSFA1bOx DEGs contain a novel HSE variant

HSFs interact with HSEs [(TTCnn)GAAnnTTC] in the promoters of target genes (Nishizawa ; Larkindale and Vierling, 2008; Kumar ). Using MEME, a motif searching algorithm (see the Materials and methods), a novel version of HSE (here called HSE1b; Fig. 4A) was identified in the promoter regions of 55 HSFA1bOx1 DEGs (Supplementary Tables S3, S7 at JXB online). It was hypothesized that these genes could constitute an HSFA1b regulon in HSFA1bOx plants. To test this hypothesis, the focus was on seven HSE1b-containing TF genes (HSFA7A, HSFB2b, HSFB2a, MBF1c, MYB, TFIIS, and ZAT6; Supplementary Tables S3, S10 at JXB online). With the possible exception of ZAT6, the induction in expression of these HSE1b-TF genes was inhibited in heat-stressed hsfA1a/hsfA1b plants (Fig. 4B), suggesting that they are regulated by clade A1HSFs in wild-type plants.

To discriminate between genes harbouring or lacking HSE1b motifs, HSFA1bOx1 and Ws-2 plants were treated with cycloheximide, an inhibitor of protein synthesis. In cycloheximide-treated HSFA1bOx1 plants, the transcription of genes directly regulated by HSFA1b would be unaffected since the level of HSFA1b protein would be high enough to persist and exert control in the absence of its synthesis (Yamada ). In contrast, the expression of genes indirectly affected by HSFA1b overexpression would be inhibited, since the levels of their transcripts would be dependent on synthesis of the TFs regulating their expression. The transcript levels of the seven HSE1b-TF genes were unaffected by cycloheximide (Fig. 4C), whereas the transcript levels of seven DEGs not harbouring a HSE1b promoter motif were lowered by the treatment (Fig. 4C).

To establish in HSFA1bOx plants whether HSFA1b interacts with promoters harbouring the HSE1b element, ChIP followed by PCR was carried out using a 35S::HSFA1b:RFP fusion line (HSFA1bOx-mRFP_B). C-terminal fusions of proteins do not affect HSFA1b function (Prändl ). The HSFA1bOx-mRFP_B line showed 165-fold induction of HSFA1b expression (Supplementary Fig. S2A at JXB online), 1.5- to 3-fold overexpression of the seven HSE1b-TF genes (Supplementary Fig. S2B), and enhanced resistance to Hpa and Pst (Supplementary Fig. S2C, D). To demonstrate the specificity in vivo of HSFA1b, three genes were selected (Fig. 4D) which harbour a single HSE1b element in their promoters (Supplementary Table S7). These genes do not have any other HSE-like motif present in their promoter regions (Supplementary data). A further three genes were selected (Fig. 4D) that harbour only a core HSE (GAAnnTTC; Larkindale and Vierling, 2008), and no match to the consensus HSE1b sequence (see Supplementary data). The promoter segments for the three HSE1b-containing genes showed amplification of DNA recovered after precipitation with the anti-RFP antibody (Fig. 4D). In contrast, the three genes which only harbour a core HSE consistently failed to give a PCR amplicon from the same ChIP preparations (Fig. 4D). Therefore, in non-stressed HSFA1bOx plants, promoters harbouring the HSE1b element can be specifically recognized in vivo by HSFA1b at least when overexpressed. As with most of the 55 HSE1b-containing genes (Supplementary Table S3), the seven HSE1b-containing TF genes contain both core HSE and HSE1b motifs in their promoter regions (Supplementary data). ChIP-PCR experiments revealed that HSFA1b binds in vivo to the promoters of the TF genes (Fig. 4E). Based on the analysis of the ‘positive’ and ‘negative’ control promoters in these experiments (Fig. 4D), it was concluded that HSFA1b most probably recognizes the HSE1b element in each TF gene.

MBF1c is part of the HSFA1b regulon and controls resistance to Pst and Hpa

Of the seven HSE1b-containing TF genes regulated by HSFA1b (Figs 5B, D; Supplementary Table S3 at JXB online), MULTIPROTEIN BRIDGING FACTOR1c (MBF1c) has already been studied extensively in the context of tolerance to heat and osmotic stress and resistance to pathogen infection (Suzuki , 2008, 2011). The overexpression of HSE1b-containing genes, and especially the seven TF genes, could be responsible for the phenotypes observed in HSFA1bOx plants (Figs 1, 2). A corollary of this is that overexpression of some HSE1b-containing genes would reproduce all or part of the phenotypes observed in HSFA1bOx plants. Microarray data from 35S:MBF1c plants (Suzuki ; here called MBF1cOx) were compared with the microarray data set from HSFA1bOx1 plants (Supplementary Table S8). There was a significant (P < 0.0001; hypergeometric distribution test) overlap of 24 genes between the 463 and 352 up-regulated genes of MBF1cOx and HSFA1bOx1, respectively (Supplementary Table S8). None of these 24 genes harbours a HSE1b element (Supplementary Table S3). MBF1c has been reported to regulate the expression of HSF genes (see Introduction), but the expression of HSFA2a, HSFB2b, and HSFA7a was unaffected in MBF1cOx plants (Fig. 5A; Supplementary Table S8). MBF1cOx and mbf1c-1 plants (Suzuki ) were analysed for water productivity and resistance to pathogen infection. Significant resistance to Pst and Hpa was observed in MBF1cOx plants (Fig. 5B, D) and there was increased susceptibility of mbf1c-1 to Pst infection (Fig. 5C). However, there were no significant increases in H2O2, GSH, and SA levels, HI, and water productivity of MBF1cOx plants in comparison with Col-0 (Supplementary Fig. S4A–C; Figs 5E, F).

Fig. 5.

Basal resistance but not water productivity is enhanced in MBF1cOx plants. (A) Analysis of gene expression of HSFA1b-responsive HSF genes and MBF1c in leaves of 5-week-old HSFA1bOx and MBF1cOx plants. White bars, HSFA1bOx1; light grey bars, HSFA1bOx2; dark grey bars, MBF1cOx. (B) Colonization of Pst in MBF1cOx and HSFA1bOx2 compared with Col-0 at 2 d (white bars) and 4 d (grey bars) post-inoculation (n=6). The inocula recovered from leaves at day 0 was 2.3 log cfu ml–1 (±0.01). (C) Colonization of Pst in mbf1c-1 plants compared with Col-0 at 2 d (white bars) and 4 d (grey bars) post-inoculation (n=6). (D) Spore yields from 12-day-old MBF1cOx and Col-0 plants (n ≥7) inoculated 5 d previously with 5×104 spores of Hpa. Data for B–D are combined from two separate experiments. (E) Water productivity in MBF1cOx, HSFA1bOx2, and Col-0 plants (n=11) in well-watered (80% of maximum rSWC) and water-limited (40% rSWC) conditions; the data represent the mean (±SEM). (F) Harvest index from the plants and conditions in E. The differences marked with an asterisk (*) are significant at P ≤ 0.05 (Student’s t-test).

Overexpression of HSFA2 does not result in phenotypes similar to HSFA1bOx plants

No improvements in water productivity, HI (Supplementary Fig. S3A, B at JXB online), and immunity to Hpa and Pst (Supplementary Fig. S4D–G), or increases in H2O2, GSH, and SA levels were observed in HSFA2Ox plants (Supplementary Fig. S4A–C). A microarray comparison between HSFA2Ox and Col-0 revealed only 43 DEGs (Supplementary Table S9) in contrast to the 509 for HSFA1bOx1 (Supplementary Table S3). The overlap between data sets was 14 genes, of which 10 are heat stress responsive (Supplementary Table S9).

Discussion

HSFA1b controls a developmental component to drought tolerance and water productivity

The data presented show that HSFA1b is a determinant of drought/dehydration tolerance when overexpressed in Arabidopsis (Fig. 1A–C). In addition, HSFA1b fulfils the same role in wild-type plants since reciprocal effects on these parameters were observed in the hsfA1a/hsfA1b mutant (Fig. 1A, C). This effect of HSFA1b overexpression on drought/dehydration tolerance did not involve changes in the expression of DREB2A or many other ABA- or dehydration-responsive genes (Supplementary Tables S3, S4 at JXB online). Furthermore, HSFA1b-regulated genes were not as over-represented in microarray data sets from plants subjected to drought stress compared with those suffering infection or heat (Table 1). Instead, the enhanced drought tolerance and water productivity of HSFA1bOx plants (Figs 1A–E, 5E) are traits connected to the increase in HI (Figs 1F, 5F), revealing a developmental component to the HSFA1bOx water productivity phenotype. Overexpression of Arabidopsis HSFA1b in transgenic oil seed rape plants (Supplementary Fig. S3B) supports this interpretation since clear changes in seed yield and HI were observed in this species (Supplementary Fig. S3C–E). The drought response and water productivity phenotypes of hsfA1a/hsfA1b plants, in many cases, were the opposite of those of the HSFA1bOx plants (Fig. 1A, C, E, F). Thus the fecundity of both wild-type and HSFA1bOx plants under differing water regimes is influenced by the constitutive expression of HSFA1b, consistent with the properties of a robust water productivity trait (Morison ). To the authors’ knowledge, there has been no single gene, when overexpressed in transgenic plants, specifically identified as influencing the HI component of water productivity (Passioura, 1977; Morison ). Biomass water ratio (BWR) is a component of water productivity (Morison ). In laboratory conditions, BWR is considered equivalent to water use efficiency (WUE; Morison ) and therefore single gene manipulations which influence WUE could also promote water productivity. These would include ERECTA (Masle ) and the TF genes STRESS-RESPONSIVE NAC1 (Hu ), HARDY (Karaba ), and NUCLEAR FACTOR-YB1 (Nelson ). A strong growth-defective phenotype has been observed in a quadruple knockout mutant of the clade A1 HSFs (Liu ), but the microarray analysis of HSFA1bOx1 plants (Supplementary Tables S3, S4) did not reveal any enrichment of genes associated with development.

HSFA1b overexpression enhances basal resistance without compromising thermotolerance or yield

HSFA1bOx plants show enhanced resistance to virulent Hpa and Pst (Fig. 2A–D; Supplementary Fig. S2C, D at JXB online) while hsfA1a/hsfA1b plants show enhanced susceptibility to these pathogens (Fig. 2A–D). In general, there is much evidence of cross-talk between abiotic and biotic stress signalling (Fujita ; Miller and Mittler, 2006; Swindell ). Heat stress can induce programmed cell death which is associated with a burst of reactive oxygen speicies, which links biotic and heat stress signalling cascades (Vacca ; Larkindale and Vierling, 2008) with SA and ABA signalling (Dat , ; Larkindale and Knight, 2002; Larkindale and Huang, 2004; Larkindale ). This may explain the negative interaction between resistance to biotrophic pathogens and sudden exposure to high temperatures (Wang ). Furthermore, HEAT SHOCK COGNATE70-1 (HSC70-1) overexpressing plants, which show enhanced thermotolerance, are negatively affected in basal and R gene-mediated resistance (Noël ). In contrast, HSFA1bOx plants reveal an important positive relationship in the signalling between heat and biotic stress responses.

The enhanced and diminished resistance to infection in HSFA1bOx and hsfA1a/hsfA1b plants, respectively (Fig. 2A–D), did not significantly involve SA-, JA-, and ABA-dependent signalling (Supplementary Table S4 at JXB online) or alterations in the levels of these hormones (Supplementary Table S6). HSFA1b-directed signalling could be mediated by H2O2 since genes responsive to it (Gadjev ) were significantly over-represented in the microarray data (Supplementary Table S4), selected genes from this group showed elevated expression in three HSFA1bOx lines (Fig. 3B), and enhanced levels of H2O2 were detected in HSFA1bOx plants (Fig. 3C). GSH and H2O2 levels are often correlated with one another in plants showing altered basal resistance (de Gara ; Mateo ; Bechtold ; Dubreil-Maurizi ). However, here, while GSH levels were enhanced in the HSFA1bOx lines (Fig. 3D), SA levels were not altered (Supplementary Table S6). Furthermore, the enhanced immunity of HSFA1bOx plants may also have been due to the overexpression of single genes such as HSP90.1 and LURP1 (LATE UP-REGULATED IN RESPONSE TO H. PARASITICA RECOGNITION1; Supplementary Table S3), which promote resistance to Pst and Hpa, respectively (Hubert ; Knoth and Eulgem, 2008).

Many Arabidopsis mutants that constitutively express ABA and/or SA signalling pathways, or are primed for resistance to infection, show diminished fecundity (Dietrich ; Heidel and Dong, 2006; Mateo ; van Hulten ; Bechtold ). Clearly, by using SA- and ABA-independent basal disease resistance (Supplementary Tables S4, S6 at JXB online), HSFA1bOx plants were not compromised in seed yield or fitness (Fig. 1D–F; Supplementary Table S1) or thermotolerance (Prändl ; Panchuk ; Busch ).

The HSE1b promoter motif suggests discrimination in HSFs binding to their cognate genes

Bioinformatics identified a modified HSE element associated with 55 genes up-regulated by HSFA1b overexpression (Fig 4A; Supplementary Table S7 at JXB online). Of the seven HSE1b-TF genes, six showed lowered expression in heat-stressed hsfA1a/hsfA1b plants (Fig. 4B) as well as all being overexpressed in non-stressed HSFA1bOx plants (Figs. 4C, 5B; Supplementary Table S3, Fig. S2B). From the cycloheximide experiments (Fig. 4C), it can be suggested that overexpressed HSFA1b directly regulates HSE1b-containing genes. The ChIP-PCR experiments (Fig. 4D) on genes containing either a single HSE1b or a single HSE showed that HSFA1b, at least when overexpressed under non-stressed conditions, specifically binds to the former. This suggests that HSFA1b recognizes HSE1b motif(s) in the promoters of the seven TF genes (Fig. 4E), supporting the conclusion from the cycloheximide experiments (Fig. 4C) that these genes are directly regulated by HSFA1b.

It must be emphasized that the functioning of the HSE1b element in wild-type plants remains to be established, but in support of the observations here a transcriptome analysis of hsfA1a/hsfA1b compared with wild-type plants showed that under heat stress, HSFA1a and HSFA1b could co-regulate the expression of >100 genes, most of which do not contain perfect HSEs (Busch ).

MBF1c expression is regulated by HSFA1b in HSFA1bOx plants and contributes to the basal resistance phenotype

Direct regulation of HSE1b-TF genes in HSFA1bOx plants suggests that they could regulate in turn some of the remaining 454 genes (Supplementary Table S3 at JXB online), thus extending the HSFA1bOx1 network to indirectly regulated genes. The example provided here comes from considering the interaction of HSFA1b with MBF1c. From the data presented (Figs 4B–E, 5A; Supplementary Table S3, Fig. S2B), it is concluded that under non-stressed conditions, overexpressed HSFA1b directly regulates MBF1c expression via its interaction with the HSE1b motif in the MBF1c promoter. MBF1cOx and HSFA1bOx1 plants share altered expression of 24 genes (Supplementary Table S8), none of which containd a HSE1b motif (Supplementary Table S3). These genes would be classed as being indirectly regulated by HSFA1b. It is suggested that it is the combination of this direct and indirect regulation of the 509 genes (Supplementary Table S3) that determines the range of observed phenotypes of HSFA1bOx plants (Figs 1, 2; Supplementary S3; Prändl ; Panchuk ). MBF1cOx plants have enhanced basal resistance (Fig 6B, D), but did not show any enhancement of water productivity or HI (Fig. 6E, F). Thus the improved basal resistance of HSFA1bOx plants may be due to its direct control of MBF1c expression, in turn altering expression of downstream genes that contribute to the resistance phenotype. These observations contrast with recent studies which suggest that MBF1c acts upstream of an SA-dependent thermotolerance pathway, routed through DREB2A, HSFB2b, and HSFB2a (Suzuki , 2011). While up-regulation of HSFB2a, HSFB2b (and HSFA7a) expression in HSFA1bOx plants was readily measured (Figs. 4B, 5A; Supplementary Table S3), no effect of MBF1c overexpression was noted on the expression of these genes under non-stressed conditions (Fig. 5A). From the microarray data (Supplementary Table S3), no altered DREB2A expression was noted in HSFA1bOx1 plants.

Overexpression of HSFA2 does not phenocopy HSFA1bOx plants

From the parallel studies on HSFA2Ox plants (Supplementary Figs S1, S4 at JXB online) it is evident that not all A-class HSFs control a broad spectrum of resistances to abiotic and biotic challenges or influence plant development. Although both HSFA1b and HSFA2 are implicated in thermotolerance (Prändl ; Nishizawa ), they control early and late responses to heat stress, respectively (Li ). In particular, it is suggested that early responding HSF genes such as HSFA1b appear to have developed as regulators of much larger gene networks in comparison with late responding HSF genes such as HSFA2, as evidenced from the present microarray analyses (Supplementary Tables S3, S9). It has been proposed that all clade A1 HSFs are regulators of the same environmental stress responses, which implies a high degree of redundancy (Liu ). However, HSFA1d and HSFA1e directly regulate HSFA2 expression during thermotolerance and high light responses (Nishizawa-Yokoi ), while HSFA1b does not impact on HSFA2 expression, and vice versa (Supplementary Fig. S1B, C; Busch ). Rather, these observations suggest distinct but overlapping regulons for each of the clade A1 HSFs.

Supplementary data

Supplementary data are available at JXB online.

Figure S1. HSFA1b and HSFA2 expressed in leaves of transgenic Arabidopsis plants and HSFA1b organ-specific expression in wild-type plants.

Figure S2. Phenotypes of HSFA1bOx-mRFP_B plants.

Figure S3. Overexpression of HSFA1b in oil seed rape improves seed yield and HI.

Figure S4. Foliar levels of SA, GSH, H2O2, HI, and water productivity in HSFA2Ox and MBF1cOx plants and response to Pst and Hpa of HSFA2Ox plants.

Table S1. Seed weight and viability of HSFA1bOx and hsfA1a/hsfA1b plants.

Table S2. Photosynthesis measurements.

Table S3. Microarray comparison of HSFA1bOx1/Ws-2, promoter analysis, and microarray data comparisons.

Table S4. GO analysis of HSFA1bOx1-responsive genes.

Table S5. Genes responsive to Hpa interaction with different RPP genes (Eulgem ).

Table S6. Stress hormone levels.

Table S7. Occurrence of the HSE1b motif in putative direct target gene promoters.

Table S8. Microarray comparison of MBF1cOx/HSFA1bOx1 and GO analysis of genes responsive to MBF1cOx.

Table S9. Microarray analysis of HSFA2Ox/Col-0 and comparison with HSFA1bOx1 genes.

Table S10. Primer sequences used in quantitative real-time PCR and ChIP-PCR.