Dual impact of elevated temperature on plant defence and bacterial virulence in Arabidopsis

Environmental conditions profoundly affect plant disease development; however, the underlying molecular bases are not well understood. Here we show that elevated temperature significantly increases the susceptibility of Arabidopsis to Pseudomonas syringae pv. tomato (Pst) DC3000 independently of the phyB/PIF thermosensing pathway. Instead, elevated temperature promotes translocation of bacterial effector proteins into plant cells and causes a loss of ICS1-mediated salicylic acid (SA) biosynthesis. Global transcriptome analysis reveals a major temperature-sensitive node of SA signalling, impacting ~60% of benzothiadiazole (BTH)-regulated genes, including ICS1 and the canonical SA marker gene, PR1. Remarkably, BTH can effectively protect Arabidopsis against Pst DC3000 infection at elevated temperature despite the lack of ICS1 and PR1 expression. Our results highlight the broad impact of a major climate condition on the enigmatic molecular interplay between temperature, SA defence and function of a central bacterial virulence system in the context of a widely studied susceptible plant–pathogen interaction.

P lant diseases represent one of the most important causes of crop loss worldwide 1 ; therefore, understanding the mechanisms underlying disease development is critical for developing effective disease control measures as part of global efforts to enable crop yields commensurate with increasing demand 1,2 . Weather plays a large role in determining the outcome of plant-pathogen interactions, and it has been noted that disease epidemics are more likely to occur when environmental conditions are suboptimal for the plant 3,4 . Responding to combined stresses (e.g., abiotic plus biotic) is challenging for plants because the response needed to mitigate one stress often can exacerbate another 4,5 . Breeding efforts to enhance yield typically reduce genetic diversity, which increases vulnerability to disease, and is also likely to negatively impact the resilience of plant immunity under adverse environmental conditions 1,6 . Increasing our understanding of how specific environmental factors affect the host and the pathogen as well as their interactions can inform strategies for developing robust crop resistance under increasingly unpredictable climate conditions.
The ability of the bacterial pathogen Pseudomonas syringae pv. tomato DC3000 (Pst DC3000) to cause disease in Arabidopsis thaliana (hereafter Arabidopsis) has made it a popular model for studying plant-pathogen interactions 7 . Two important virulence mechanisms employed by Pst DC3000 to cause disease are the phytotoxin coronatine (COR) and the type III secretion system (T3SS), which translocates bacterial effectors into host cells 7 . Conflicting results exist with respect to the effect of temperature on these virulence mechanisms. Elevated temperature has a negative effect on the expression of both COR-related and T3SSrelated genes in vitro 8 . However, elevated temperature did not affect Pst DC3000 production of COR in planta 9 . Whether elevated temperature affects the production and translocation of type III bacterial effectors into plants is unknown.
Salicylic acid (SA) is a major plant defence hormone important for both local and systemic resistance against biotrophic and hemi-biotrophic pathogens, such as Pst DC3000 10 . In Arabidopsis, pathogen-induction of SA biosynthesis occurs predominantly through the isochorismate pathway involving ICS1 11 . Following SA induction, the master regulator, NPR1, accumulates in the nucleus where it interacts with TGA and WRKY transcription factors (TFs) to promote transcriptional reprogramming 12 . Among the many genes induced by SA, PR1 is one of the most widely used markers for SA signalling in Arabidopsis 13 . SA-mediated defence has been well established as a crop protecting mechanism; for example, an SA synthetic analogue, benzothiadiazole (BTH), can provide resistance resulting in increased yield in multiple crops, including wheat 14 and maize 15 . Additionally, over-expression of NPR1 has been shown to improve disease resistance in rice 16 . Basal defence against Pst DC3000 and induction of SA during effector-triggered immunity have been shown to be compromised at elevated temperature 17,18 . However, it is unclear whether either of these outcomes results from a direct impact of temperature on the SA pathway, as SA-deficient mutants were reported to retain temperature sensitivity during basal defence 18 , and loss of effector-triggered immunity-induced SA may be an indirect effect resulting from temperature-mediated loss of upstream resistance (R) protein function. SA also plays a role in pattern-triggered immunity 19 , which has been reported to not be suppressed at elevated temperature 20 .
It was not clear until recently how plants sense elevated temperature. The phyB red-light photoreceptor is a negative regulator of the PIF4 growth-promoting TF 21 , and was shown to function as a thermosensor in plants 22,23 . At elevated temperatures, heat inactivation of phyB results in de-repression of PIF4regulated genes, enabling growth 22 . Another recent study suggests that PIF4 mediates defence suppression at elevated temperature; however, all the pif mutants tested retained temperature-sensitive pathogen growth 24 .
Although previous research has studied the effects of elevated temperatures on effector-triggered immunity or pattern-triggered immunity, our knowledge of the impact of elevated temperature on disease development-from pathogen virulence systems and host defence to temperature sensing-during a susceptible plant-pathogen interaction remains fragmented, preventing formulation of a cohesive model that could guide future research. We sought to address this knowledge gap by studying the model susceptible Arabidopsis-Pst DC3000 pathosystem. Contrary to prevailing results obtained in vitro, we discovered that elevated temperature has a positive effect on the function of the T3SS in planta, and a negative, but phyB/PIF-independent effect on SA-mediated defence in the host, resulting in overall enhanced disease. Remarkably, although BTH-induction of the temperature-sensitive PR1/ICS1 branch is compromised at 30°C, BTH-mediated protection against Pst DC3000 persists. We propose an integrated model illustrating the interplays between temperature, SA-mediated defence and the function of the T3SS in the context of a susceptible plant-pathogen interaction.

Results
Elevated temperature enhances disease susceptibility. Temperatures between 27-30°C are considered 'moderately elevated' for Arabidopsis, which is more susceptible to Pst DC3000 infection in this temperature range 18,[24][25][26][27] . However, some of these studies used plants that were grown at acclimated temperatures for long periods of time before disease assays 18,27 , resulting in plants with dramatically different morphology, including exaggerated hypocotyl and petiole elongation, due to PIF4-induction of auxin 28 . Although periods of elevated temperature vary in nature, to minimize the impact of physiological differences confounding our study, we first assessed the effect of a shorter, 2-day temperature acclimation period on infection of Arabidopsis plants by Pst DC3000. Four-week-old plants were acclimated to test chambers at 23°C (control) or 30°C (test) for only 48 h prior to syringe-infiltration with Pst DC3000. Morphological differences were greatly reduced between test and control plants acclimated for 48 h relative to those acclimated for 7 days (Supplementary Fig. 1a-c). We observed a 30-fold increase in bacterial growth as well as a dramatic increase in disease-associated chlorosis in plants at 30°C relative to those at 23°C (Fig. 1a, b). In contrast, Pst DC3000 grew similarly at 23°C and 30°C in vitro ( Supplementary Fig. 2), suggesting that the effect of elevated temperature on Pst DC3000 multiplication is plant-dependent.
We next conducted Pst DC3000 infections in plants that were (i) acclimated and kept at 23°C following infection (23°C → 23°C), (ii) acclimated at 30°C and shifted to 23°C (30°C → 23°C), (iii) acclimated at 23°C and shifted to 30°C following infection (23°C → 30°C) or (iv) acclimated and kept at 30°C following infection (30°C → 30°C). We found that the postinfection temperature is the determining factor in disease susceptibility ( Supplementary Fig. 3). Although plants were consistently more susceptible to Pst DC3000 at 30°C relative to plants at 23°C, the degree of enhanced disease susceptibility was more consistent following a 2-day acclimation at 30°C ( Supplementary Fig. 3). Therefore, we used a 2-day acclimation to 30°C prior to inoculation for the remainder of our experiments.
PIF4 was recently proposed as a regulator of growth promotion and immunity suppression at elevated temperature 24 . To determine whether PIFs are responsible for enhanced susceptibility at elevated temperature in our system, disease assays were conducted with wild type (WT, Col-0) and pif1 pif3 pif4 pif5 29 (hereafter, pifq) mutant plants 1 dpi, 2 dpi and 3 dpi at both 23°C and 30°C. No difference in bacterial growth was observed between the WT and pifq mutant plants at 30°C for each time point assessed ( Supplementary Fig. 4a), indicating that PIFs do not play a major role in mediating elevated temperaturedependent enhancement of disease susceptibility to Pst DC3000 inside the infected leaves (i.e., when bacteria are infiltrated into the leaf apoplast). Next, we examined the possibility that enhanced susceptibility at elevated temperature is due to heat inactivation of the phyB photoreceptor, which is a positive regulator of SA-induced PR1 gene expression and has been shown to be a thermosensor for photomorphogenesis 22,23,30 . To test this, we assessed disease susceptibility in temperature stable phyB transgenic lines (35S::PHYB Y276H , referred to as YHB hereafter) 31,32 . We observed no difference in disease phenotype between YHB and Landsberg erecta (Ler) WT plants at either temperature ( Supplementary Fig. 4b). Taken together, we conclude that enhanced susceptibility at elevated temperature is independent of the phyB/PIF pathway.
Elevated temperature enhances bacterial type III secretion. We next examined whether enhanced multiplication of Pst DC3000 in planta at 30°C requires bacterial virulence factors. Specifically, growth of two bacterial mutant strains, Pst DC3000 hrcC − (hereafter hrcC − , defective in the T3SS 33 ) and Pst DC3118 (defective in COR production 34 ), was compared with growth of Pst DC3000 in Arabidopsis plants kept at 23°C or 30°C. The hrcC − mutant strain had no detectable increase in growth at 30°C (Fig. 1c). In contrast, while disease-associated leaf chlorosis was greatly reduced in Pst DC3118-infected plants (Fig. 1d), growth of this strain was 400-fold higher at 30°C than at 23°C (Fig. 1c), indicating that enhanced growth of Pst DC3000 in planta at 30°C requires a functional T3SS but not COR.
We found a requirement of the T3SS for enhanced disease at 30°C surprising because previous in vitro studies had shown a negative effect of elevated temperature on the T3SS 8 . To resolve this unexpected dilemma, we directly examined T3SS-mediated bacterial effector translocation into plant cells at 30°C using Pst DC3000 strains with either the AvrPto, AvrPtoB, HopU1 or HopG1 effectors fused to the CyaA reporter 35 . To ensure differences in effector translocation were not influenced by bacterial populations, plant samples were first selected based on having similar bacterial populations at the 4-6 h post-infiltration (hpi) time point used for the translocation assay ( Supplementary  Fig. 5a, b). A hrcC − mutant strain carrying the P nptII ::avrPto-CyaA plasmid was used as a negative control for the translocation assay. We observed a significant increase (2-4-fold) in translocation of all tested effectors at 30°C compared to that at 23°C (Fig. 1e), suggesting that, contrary to the prevailing notion from in vitro studies, increased Pst DC3000 virulence at 30°C is linked to increased translocation of bacterial effector proteins.
The SA pathway is compromised at elevated temperature. As SA-mediated defence plays a major role in protecting Arabidopsis against virulent Pst DC3000, we investigated the possibility that the SA pathway is compromised at 30°C. We measured ICS1 and PR1 marker gene expression as well as SA metabolite levels 24 hpi with mock or Pst DC3000. We observed significant Pst DC3000induction of both ICS1 (7-fold) and PR1 (60-fold) in plants kept at 23°C, whereas neither gene was induced by the pathogen at 30°C (Fig. 2a). Similarly, total SA was induced by Pst DC3000 to levels 7-fold higher than in mock-infiltrated plants at 23°C, but no significant increase was detected at 30°C (Fig. 2b). To determine whether loss of SA can account for enhanced susceptibility at elevated temperature, we tested bacterial growth at 1 dpi, 2 dpi and 3 dpi in both WT (Col-0) and sid2-2 (hereafter, ics1) mutant plants. As previously reported 18 , ics1 plants were more susceptible to Pst DC3000 than WT plants at 23°C, showing 70-fold more bacterial growth at 2 dpi, which is prior to the bacterial population reaching saturation (Fig. 2c). Under our experimental conditions, the ics1 mutant showed only a slight (3-fold) increase in bacterial growth at 1 dpi, but showed similar bacterial loads at 2 dpi and 3 dpi at 23°C vs. 30°C (Fig. 2c). As the disease phenotype of the ics1 mutant at 23°C resembles that observed in WT plants at 30°C, we speculated that loss of pathogen-induced SA in WT plants is likely a major mechanism for enhanced disease at elevated temperature.
Tissue was collected at 4-6 hpi for quantification of cAMP, which was normalized by total protein. Higher levels of cAMP indicate more translocation of bacterial effectors. All data are representative of three independent experiments; graphical data are presented as the mean ± standard error of the mean (s.e.m.), with n = biological replicates. Letters indicate statistical significance based on a two-factor ANOVA with Tukey's HSD post hoc analysis (p < 0.05); samples sharing letters are not significantly different. Asterisks indicate statistical significance based on a Student's t test (***p < 0.001) of pairwise comparisons for each individual effector strain at 23°C vs. 30°C; 'ns' indicates no significance assessed at the same time point as used for translocation assays to confirm differences in translocation are not due to differences in bacterial population ( Supplementary Fig. 5c). We observed a significant increase (2-fold) in bacterial effector translocation into ics1 vs. WT plants at 23°C (Fig. 2d), indicating that loss of SA enables more effector translocation. However, there was also significantly more translocation of effector proteins into both WT and ics1 mutant plants at 30°C (4-5-fold, Fig. 2d), indicating that, at elevated temperature, both loss of SA biosynthesis in the host and increased bacterial translocation of effector proteins likely contribute to enhanced disease development.
Suppression of SA responses at 30°C is pathogen independent. We next used BTH to directly interrogate the effect of elevated temperature on SA biosynthesis and signalling in a pathogen-free system. BTH is widely used as a surrogate for the SA signal, and is a potent inducer of SA response genes in Arabidopsis 10,36 . BTH significantly induced both ICS1 (10-fold) and PR1 (>2500-fold) gene expression as well as total SA levels (15-fold) in plants kept at 23°C with no significant induction of ICS1, PR1 or SA levels observed at 30°C (Fig. 3a, b). Similarly, BTH-induction of callose deposition, which is thought to reinforce plant cell walls against pathogen penetration 37 , was observed at 23°C (8-fold) but not at 30°C (Fig. 3c, d). In contrast, although slightly reduced at elevated temperature, flg22 elicited a strong callose response at both 23°C (90-fold) and 30°C (50-fold; Supplementary Fig. 6a, b), indicating that the observed effect of temperature on BTHinduced callose deposition is likely due to compromised SA signalling rather than an effect on the callose synthase enzyme.
NPR1 nuclear localization is retained at 30°C. Elevated temperature negatively affects R protein-mediated disease resistance 19,20 . Zhu et al. (2010) showed that loss of R protein nuclear localization contributes to compromised R-mediated defence at elevated temperature (28°C) 38 . NPR1 is a key regulator of SA signalling and accumulates in the nucleus upon SA signal perception 10,12 . As nuclear localization of NPR1 is required for PR1 gene induction 39  YFP) were also generated as controls. A nuclear YFP signal was observed in mock-treated and BTH-treated NPR1-Y1 plants at both 23°C and 30°C, although the signal was extremely weak in mock-treated plants at both temperatures ( Fig. 4a). There was no observable effect of treatment (chemical or temperature) on YFP signal detected in the p35S::YFP control, and no YFP signal was detected in the parent npr1 plants ( Supplementary Fig. 8a, b). To independently confirm NPR1 nuclear localization, subcellular fractionation experiments were also conducted. NPR1-YFP was observed in whole-cell lysate and both fractions of BTH-treated samples at both temperatures (Fig. 4b). Western blotting of UGPase and H3 proteins, which were used as cytosolic-fraction and nuclear-fraction specific markers, respectively, showed significant enrichment within their respective fractions (Fig. 4b). Thus, the confocal microscopic and nuclear fractionation data both show that elevated temperature does not prevent BTH-induced nuclear accumulation of the NPR1 protein.
To confirm the loss of BTH-induction of PR1 protein accumulation, western blot analysis using a PR1 antibody was also conducted with fractionated protein samples confirmed to have NPR1 localized to the nucleus. Similar to PR1 gene expression (Fig. 3a), PR1 protein was only detectable in BTHtreated plants at 23°C (Supplementary Fig. 9b). Therefore, in addition to loss of SA biosynthesis, there appears to be a negative effect of elevated temperature on SA-mediated signalling  Because changes in basal gene expression due to temperature will impact fold change, we analysed the BTH vs. mock DEGs in two ways: (i) fold change (BTH/mock) and (ii) expression levels (23°C BTH vs. 30°C BTH). k-means cluster analysis (k = 9) was conducted with the 2820 BTH-responsive DEGs, and a heat map was generated to visualize gene expression patterns. Clusters 1-7 contain genes induced by BTH (1833 genes), with cluster 1, 2 and 3 genes exhibiting compromised induction, cluster 4 genes having similar induction, and cluster 5, 6 and 7 genes showing enhanced induction at 30°C; clusters 8 and 9 contain genes suppressed by BTH at 23°C but compromised in suppression at 30°C (987 genes, Supplementary Fig. 10). For the analysis based on expression levels, a 2-fold cut-off was used to categorize genes as follows: group A, expression lower at 30°C vs. 23°C (956 genes); group B, expression similar at 30°C and 23°C (725 genes); and group C, expression higher at 30°C vs. 23°C (152 genes; Fig. 5a). As 810 out of 987 BTH-suppressed genes showed higher expression at 30°C, these genes were kept in a single group D (Fig. 5a). A comparison of gene distribution by cluster and group reveals that 42% of genes that did not show differential induction based on cluster analysis (i.e. cluster 4) showed compromised expression at 30°C based on group analysis (i.e. group A), indicating that cluster-based and group-based analyses have different sensitivities (Supplementary Table 2). Based on meta-analysis of publicly available microarray datasets (clusters only) and functional annotations using GO analyses, BTH-induced genes (clusters 1-7 / groups A-C) are involved in   If we consider genes induced by BTH at 23°C (groups A and B, 1681 genes) as the typical SA response genes, a clear temperature-sensitive bifurcation is revealed at the 24 h time point assessed in our study, with 57% of BTH-induced genes 'temperature-sensitive' (group A, 956 of 1681 genes) and 43% of BTH-induced genes 'temperature-insensitive' (group B, 725 of 1681 genes). Based on this analysis, known positive regulators of SA biosynthesis, including EDS1, PAD4, CBP60g and SARD1 10 , are predominantly in the PR1/ICS1 branch (group A), while negative regulators, such as WRKY54, MYC2 and several MYC2-regulated ANACs 10 , are in groups B and C (Fig. 5b, c,  Supplementary Table 4). NPR1 and two of its known regulators, NPR3 10 and SUMO3 40 , are in group B (Fig. 5b), which is consistent with our finding that NPR1 induction (and localization) is not impaired at 30°C. Downstream of NPR1, many SAR-related genes, including PR1, PR2, PR5, ALD1 and FMO1 10 , appear in group A; however, several positive regulators, including WRKY54 10 and ZAT6 41 are in group B (Fig. 5b, c).
We next conducted analysis of motif enrichment (AME) 42 , using 1000 bp upstream of each gene, to identify TFs that may be involved in regulation of genes within each cluster and    Table 5a); however, the group-based analyses were more informative. While genes in groups A and B had an over-representation of the W-box motif bound by WRKY TFs, known to regulate genes induced by both biotic and abiotic stresses 43 , only genes in group A had an over-representation of the as-1 element bound by TGA TFs 44 , which are known to regulate SA-responsive PR genes 10 (Supplementary Table 5b). Genes containing cis elements bound by TF regulating circadian responses, such as CCA1 45 , were overrepresented in group B, which is the temperature-insensitive group. Additionally, genes in group C had an over-representation of cis elements bound by TFs involved in jasmonic acid (JA) and abscisic acid (ABA) signalling as well as the CAMTA2 and CAMTA3 TFs, both of which are negative regulators of SA 46,47 . Finally, group D genes had an over-representation of cis elements bound by PIFs, MYCs and multiple TFs involved in growth-related processes.
Enhanced susceptibility is not due to major SA antagonists.
Our GO and promoter analyses showed that plants treated with BTH at 30°C have an upregulation of genes involved in JA, ethylene (ET) and ABA signalling pathways (Supplementary Tables 3  and 5), which are known to antagonize SA biosynthesis and signalling [48][49][50] . We conducted further experiments to determine whether these pathways are also upregulated during Pst DC3000 infection at elevated temperature. We found no measurable effect of temperature on pathogen-induction of ET-responsive genes, EIN2 and ERF6, following infection; however, both JA pathway marker genes, LOX2 and MYC2, as well as the NCED3 gene involved in ABA biosynthesis were induced to higher levels by Pst DC3000 at 30°C compared to those at 23°C (Fig. 6a, c, e; Supplementary  Fig. 14a, c). ABA metabolite levels were also increased following pathogen infection at 30°C (Supplementary Fig. 14e). Therefore, while SA biosynthesis and signalling are compromised, Pst DC3000-induction of JA-mediated signalling and ABA biosynthesis are enhanced at elevated temperature.
To test whether ABA, JA or ET has a causal role in enhanced susceptibility at elevated temperature, we conducted disease assays at 23°C and 30°C using the aba2-1 51,52 (hereafter aba2) ABA-deficient mutant, the dde2-2 53 JA-deficient mutant (defective in the ALLENE OXIDE SYNTHASE gene; hereafter aos), the myc2 myc3 myc4 54 (hereafter myc2/3/4) JA-signalling mutant and the ein2-1 55 (hereafter ein2) and ein3 eil1 48 ET-signalling mutants. As expected, all five mutants showed a significant (5-10-fold) reduction of Pst DC3000 growth at 23°C compared to WT plants (Fig. 6b, d, f; Supplementary Fig. 14b, d). However, based on the retention of temperature sensitivity in the mutant plants and similar levels of susceptibility between these mutants and the corresponding WT plants at 30°C, we conclude none of these hormone signalling pathways is primarily responsible for enhanced susceptibility to Pst DC3000 at elevated temperature.
We also found enrichment of CAMTA-targeted cis elements in genes more highly induced and/or expressed following BTH treatment at 30°C (Supplementary Table 5). CAMTA1, CAMTA2 and CAMTA3 are TFs that function redundantly in repressing SA biosynthesis at temperatures between 19°C and 22°C 47,56 . This repression is relieved upon pathogen perception or in response to cold temperatures (~4°C), enabling SA accumulation in these conditions 56,57 . As expected, the camta2 camta3 double mutant (hereafter camta2/3) showed enhanced resistance against Pst DC3000 relative to WT plants at 23°C (90-fold less growth; Fig. 6g). Unlike the other mutants of SA negative regulators tested, the camta2/3 mutant retained partially heightened resistance at 30°C (13-fold less growth; Fig. 6g), indicating that CAMTAs may contribute to enhanced susceptibility at elevated temperature. In support of this, total SA in the camta2/3 mutants was significantly elevated at 30°C, with levels similar to pathogen-treated WT plants at 23°C (Fig. 6h). However, although ICS1 and PR1 gene expression was constitutively elevated in the camta2/3 mutant relative to WT plants at 23°C, expression of these genes remained compromised at 30°C (Fig. 6i), indicating that loss of ICS1 and PR1 gene expression at   BTH confers resistance at 30°C without inducing ICS1 or PR1. BTH is an SA analogue available commercially as a means of crop disease control 58 . The dramatic effect on SA biosynthesis and signalling at 30°C raises the possibility that BTH may fail to protect plants against pathogens at elevated temperature. To directly test this possibility, we conducted BTH protection assays in WT plants at 23°C and 30°C. To our surprise, BTH-treated WT plants had 150-200-fold less bacterial growth and showed no disease symptoms relative to the mock-treated controls at both temperatures (Fig. 7a, b). Even more surprising is that BTH protection occurred with no induction of the ICS1 and PR1 marker genes or accumulation of SA at 30°C (Fig. 7c, d). Thus, BTH can fully protect Arabidopsis against Pst DC3000 in the absence of the canonical SA marker genes, ICS1 and PR1. We next tested whether BTH-mediated protection at 30°C requires the core SA signalling pathway. BTH protection assays were conducted in WT, npr1 mutant and tga2 tga5 tga6 59 (hereafter tga2/5/6) triple mutant plants. Mock-treated npr1 mutant plants had 30-fold higher bacterial growth relative to mock-treated WT plants at 23°C; BTH no longer provided protection in the npr1 mutant plants at either temperature (Fig. 7e). Similarly, mocktreated tga2/5/6 mutant plants had 10-fold more bacterial growth than mock-treated WT plants at 23°C (Fig. 7f). BTH protection is completely absent in tga2/5/6 mutants at 30°C (Fig. 7f), indicating that NPR1, TGA2, TGA5 and TGA6 are required for BTHmediated protection at elevated temperature.
Next, we tested the effect of BTH on translocation of bacterial effectors to determine whether BTH-mediated restriction of virulence proteins into plant cells contributes to protection at elevated temperature. A higher bacterial inoculum was used in BTH-treated plants than in mock-treated plants to enable equal bacterial populations in plants at 4 hpi ( Supplementary Fig. 15a). We observed a significant reduction (2-3-fold) in translocation of AvrPto-CyaA effector protein in BTH-treated plants relative to the mock-treated controls at both temperatures (Fig. 7g). However, although the trends were consistent across all experiments, the effect size of BTH on bacterial effector translocation was more variable at 30°C ( Supplementary  Fig. 15b). From these data, we conclude that BTH restriction of bacterial effector translocation contributes to protection at both temperatures, but there are likely other mechanisms involved in BTH-induced disease resistance.

Discussion
According to the long-standing 'disease triangle' dogma in plant pathology, plant-pathogen interactions can only be fully understood within the context of environment 60 . Elevated temperature has been shown to break down effector-triggered immunity and promote disease in many plant-pathosystems 61 , from fungal infection of wheat 62 to viral infection of tobacco 63 . While some insight regarding the molecular mechanisms involved in loss of effector-triggered immunity has been gained 64 , the underlying cause for a potentially general enhancement of disease susceptibility in compatible plant-pathogen interactions (i.e. in the absence of effector-triggered immunity) was poorly understood prior to this study. We provide evidence that enhanced multiplication of virulent Pst DC3000 inside Arabidopsis leaves at elevated temperature is independent of the phyB/PIF thermosensing pathway (Supplementary Fig. 4a, b). Instead, enhanced bacterial multiplication at 30°C requires a functional T3SS (Fig. 1c), and, contrary to the widely held notion based on studies of the T3SS in vitro 8 , is associated with increased translocation of bacterial effectors into plant cells (Fig. 1e). We also show that enhanced susceptibility of Arabidopsis under our conditions (i.e. when bacteria are pressure-infiltrated directly into the plant leaves) is linked to loss of SA, and the SA-deficient ics1 mutant   Fig. 7 BTH protection against Pst DC3000 at 30°C requires NPR1 and TGAs but occurs without ICS1 and PR1 expression. a Bacterial growth in mock-or BTH-pre-treated plants (n = 4) 3 days after vacuum-infiltration with Pst DC3000. b Disease symptoms three dpi in plants in a. c SA marker gene expression (n = 3) and d SA metabolite quantification (n = 4) 24 hpi of plants in a. Gene expression and LCMS data were processed and analysed as described in Fig. 2. Bacterial growth in mock-or BTH-pretreated WT, e npr1 and f tga2/5/6 mutant plants (n = 4) 3 days after vacuum-infiltration with Pst DC3000. g Translocation of bacterial effector proteins in plants (n = 4) pre-treated with mock or BTH 24 h before syringe-infiltration with Pst DC3000(P nptII ::avrPto-CyaA). Additional mock-treated plants (n = 4 for each temperature) were infiltrated with hrcC − (P nptII ::avrPto-CyaA strains) as a negative control. Tissue was collected at 4 hpi for quantification of cAMP, which was normalized by total protein.
All data are representative of three independent experiments; graphical data are presented as the mean ± s.e.m. with n = biological replicates. Letters indicate statistical significance based on a two-factor ANOVA with Tukey's HSD post hoc analysis (p < 0.05); samples sharing letters are not significantly different. Data in e and f were analysed in two groups based on temperature as indicated by the prime symbol (′). Symbols in g denote statistical significance based on a one-factor ANOVA with Dunnett's post hoc analysis (***, ### p < 0.001) using the mock-treated sample at each temperature as the means for comparison exhibits a largely temperature-insensitive disease phenotype, with a similar level of susceptibility as WT plants at 30°C (Fig. 2c). RNA-seq analysis of combinations of temperature and BTH treatments, reveal temperature-sensitive and temperatureinsensitive branches in the SA gene expression network (Fig. 5).
We also provide evidence that BTH confers robust protection against disease at 30°C even in the absence of canonical ICS1 and PR1 gene expression (Fig. 7a-c). Collectively, these results highlight the multi-faceted impact of elevated temperature on the molecular interplay between temperature, SA-mediated defence and the function of a central bacterial virulence system in one of the best studied susceptible plant-pathogen interactions. We considered three potential models that could explain how elevated temperature affects the compatible host-pathogen interaction to enable enhanced disease. Based on previous in vitro analyses showing a negative effect of elevated temperature on the expression of virulence-associated genes 8 and our data showing loss of pathogen-induced SA biosynthesis (Fig. 2b), Model 1 predicts that elevated temperature negatively affects both pathogen virulence and SA-mediated defence. In this model, enhanced disease would result from increased host susceptibility in spite of compromised bacterial virulence. However, the inability of the non-pathogenic hrcC − mutant strain to grow more in plants at elevated temperature indicates that a functional T3SS is required for enhanced bacterial growth at 30°C (Fig. 1c). Moreover, our results show that effector translocation in planta is more efficient at 30°C (Fig. 1e), effectively refuting Model 1.
Alternatively, Model 2 proposes that elevated temperature enhances bacterial virulence, which then causes loss of SA biosynthesis to promote disease. Pst DC3000 generated COR, which is known to specifically target and antagonize SA biosynthesis by activating JA-mediated signalling 50 , was previously shown to be unaffected by elevated temperature in planta 9 . Although symptom development was greatly reduced in plants infected with a CORdeficient mutant, growth of this strain was enhanced similarly to that of Pst DC3000 at 30°C (Fig. 1c, d), thereby eliminating COR-mediated suppression of SA biosynthesis as a potential mechanism. As bacterial effector translocation is increased at 30°C, it is possible that an effector-mediated process may contribute to loss of SA biosynthesis at elevated temperature. However, the loss of BTH-induced SA production at 30°C (Fig. 3b), which is a pathogen-free treatment, prompts rejection of Model 2.
Therefore, we propose a third model whereby elevated temperature both antagonizes SA accumulation in the host and promotes translocation of virulence proteins by the pathogen resulting in enhanced disease (Fig. 8). Increased bacterial effector translocation into the ics1 mutant relative to WT plants at 23°C (Fig. 2d) suggests that loss of SA may result in increased 'permeability' of the cell to T3SS-mediated pathogenesis. This enhanced permeability would result in the observed temperature sensitivity of the ics1 mutant to effector delivery. We typically observe a positive correlation between bacterial effector translocation and bacterial multiplication. However, at 30°C, there is significantly more bacterial effector translocation into ics1 mutant relative to WT plants with no further increase in bacterial populations (Fig. 2c, d). This suggests there is a minimum threshold of bacterial effector translocation required for maximal bacterial multiplication, after which no further positive effect may be observed.
One outstanding question to be addressed is the mechanism responsible for loss of SA at elevated temperature. Based on in vitro activity assays, which show the ICS1 enzyme maintains >90% maximal activity from 4 to 37°C 65 , and our data showing complete loss of ICS1 gene induction at 30°C, it seems likely that loss of SA biosynthesis and accumulation occurs upstream of ICS1 gene expression. However, we cannot rule out the possibility of additional changes in ICS1 and other SA biosynthesis protein modifications at elevated temperature that may affect their activities. In this study, we discovered upregulation of a number of known negative regulators of ICS1-mediated SA biosynthesis, including components of JAmediated 50 and ET-mediated 48 signalling and ABA 46 biosynthesis (Fig. 6a, c, e; Supplementary Fig. 14a, c, e). However, all hormone mutants tested showed similar levels of susceptibility as the WT plants at 30°C (Fig. 6b, d, f; Supplementary Fig. 14b, d), indicating that none of these pathways is solely responsible for SA suppression at elevated temperature.
Retention of resistance in the camta2/3 mutants and in BTHtreated WT plants at 30°C in the absence of PR1 and ICS1 gene induction (Figs. 6g, i and 7a, c) is striking. In particular, the observed retention of BTH-induced NPR1 nuclear accumulation (Fig. 4), and the requirement for NPR1 and the TGA2, TGA5, TGA6 TFs (Fig. 7e, f) indicates that the core SA signalling pathway is utilized to facilitate this resistance at 30°C. Global transcriptome analysis of BTH-regulated genes revealed a temperature-sensitive bifurcation in the SA signalling pathway at the 24 h time point assessed in our study, with~60% of genes in the canonical PR1/ICS1 temperature-sensitive branch, and~40% of genes in the temperature-insensitive branch (Figs. 5 and 8). Based on the lack of induction of PR1 and ICS1 by either Pst DC3000 or BTH alone (Figs. 2a and 3a) or in combination at 30°C (Fig. 7c), it is possible that a major part of the SA-regulated transcriptome is dispensable for BTH-mediated protection against Pst DC3000 (Fig. 8). This surprising result indicates that either the temperature-insensitive branch of the SA transcriptome and/or a gene expression-independent process underlies BTH-mediated Fig. 8 Model for the Arabidopsis-Pst DC3000 interaction at elevated temperature. At elevated temperature (depicted by the sun), pathogeninduction of SA biosynthesis via the ICS1 enzyme is blocked and translocation of bacterial T3E proteins is enhanced to promote disease. SA signalling induced by the SA synthetic analogue, BTH, is also affected, with the PR1/ICS1 branch (group A in Fig. 5a) no longer induced. However, BTH-mediated resistance against Pst DC3000 is still conferred at elevated temperature in a NPR1-and TGA2/5/6-dependent manner. It is possible that genes in the temperature-insensitive branch, group B and/or genes in the elevated temperature + BTH-induced branch, group C, are involved in BTH-mediated resistance under this condition. Solid arrows represent direct positive interactions, dotted arrow represents induction of ICS1 in response to pathogen detection and bar-headed lines represent inhibitory interactions resistance against Pst DC3000 at 30°C. We chose the 24 h time point for gene expression analysis because this is when PR1, PR2 and PR5 genes are most highly expressed following BTH treatment 36 . However, it is important to note that this provides only a snap shot of the dynamic transcriptional landscape, and future research should examine whether temperature-sensitive genes may be induced or suppressed at earlier or later time points.
In summary, we have studied the impact of an important climate condition, elevated temperature, on a widely studied compatible plant-pathogen interaction. Our results highlight pathogen-induced SA production as a key temperature-vulnerable step in the SA defence network. In addition, our study revealed a surprisingly positive effect of elevated temperature on bacterial translocation in planta, which challenges the long-standing notion based on previous in vitro studies that an efficient T3SS requires a low temperature (e.g., 18-20°C) 8 . Surprisingly, in spite of the increase in pathogen virulence, we still observed BTH-mediated protection against Pst DC3000 infection. This finding supports the continued use of BTH as a crop protectant even at elevated temperatures. Because the SA pathway is an integral component of the plant immune system, we hope that the fundamental insights gained from this study will stimulate future research to uncover additional temperature-sensitive and temperature-insensitive nodes of the plant immune system. This information should prove useful for genetic manipulation of climate-relevant components of the plant immune system to enhance plant resilience to combined adverse abiotic and biotic conditions.
Phytohormone extraction and quantification. Phytohormones were extracted and quantified as previously described 66 with some modifications. Leaf tissue between 10 and 50 mg (fresh weight, FW) was flash-frozen in liquid nitrogen, ground and extracted at 4°C overnight (~16 h) using 0.3-0.5 mL of ice-cold extraction buffer (methanol:water (80:20 v/v), 0.1% formic acid, 0.1 g L −1 butylated hydroxytoluene, 100 nM ABA-d 6 ). Filtered extracts were quantified using an Acquity Ultra Performance Liquid Chromatography (UPLC) system (Waters Corporation, Milford, MA) as previously described 66 , except the capillary voltage, cone voltage and extractor voltage were set to 3.5 kV, 25 V and 5 V, respectively, and the desolvation gas and cone gas were set to flow rates of 600 L h −1 and 50 L h −1 , respectively. Selected ion monitoring (SIM) was conducted in the negative ES channel for SA (m/z 137 > 93), SA glucoside (SAG; m/z 299.1 > 137), ABA (m/z 263.1 > 153.1) and the internal ABA-d 6 standard (m/z 269.1 > 159.1). Parent > daughter SIM pairs, as well as the optimal source cone and collision energy voltages for each compound monitored were determined using Quan-Optimize software. Analyte responses based on peak area integrations relative to the internal standard was determined using QuanLynx v4.1 software (Waters, Milford, MA). Both the SA and SAG analytes were quantified based on the SA standard curve and ABA was quantified based on the ABA standard curve to calculate the sample concentrations (nM), which were converted to ng using the molecular weight of the compound and the extraction volume, and were then normalized by sample FW in g. SA and SAG concentrations were combined and reported as total SA. RNA extraction and qPCR. RNA was extracted from flash-frozen, ground leaf tissue with the ToTally RNA kit following the manufacturer's protocol (Ambion). Samples were digested with DNaseI (Roche) to remove any genomic DNA contamination, and then purified using the RNeasy Mini kit (Qiagen). M-MLV reverse transcriptase (RT, Life Technologies) was used to synthesize cDNA. For all genes of interest,~1.5 ng of cDNA template was used for quantitative PCR (qPCR), with expression normalized to the PP2AA3 internal control gene using the equation 2 −ΔCT , where ΔC T is C T target gene -C T PP2AA3 (see Supplementary Table 6 for primer sequences). All qPCR reactions were performed using the SYBR® Green master mix (Life Technologies) and 7500 Fast Real-Time PCR system (Applied Biosystems, Foster City, California), with three technical replicates and a minimum of three biological replicates per experimental treatment.
RNA sequencing and data analysis. Three biological replicates of each experimental treatment were selected based on quality (RIN score 7) using a BioAnalyzer Agilent 2100. Samples were pooled on two lanes of an Illumina HiSeq 2500 Rapid Run flow cell (v1) and sequenced in a 1 × 50 bp single-end format using Rapid SBS reagents. Details regarding sequencing, read counts and data analysis are provided in Supplementary Note 1.
Callose accumulation. Following temperature acclimation and chemical treatment, leaves were harvested and cleared in 100% ethanol overnight. Cleared leaves were fixed with a 75% ethanol, 25% acetic acid solution for 2 h, after which leaves were washed consecutively with 75% ethanol, 50% ethanol and 150 mM K 2 HPO 4 pH 9.5 for 15 min. Finally, leaves were stained in an aniline blue solution (0.1%, 150 mM K 2 HPO 4 pH 9.5) overnight at 4°C. Callose deposits were visualized using an Olympus IX71 inverted microscope with a 120-watt metal halide lamp (X-Cite series 120) using a DAPI filter (Semrock, excitation 377/50 and emission 447/60). Images shown are at ×10 magnification. Callose counts were processed using ImageJ (Rasband W.S., National Institutes of Health, USA). Images were first converted to 32-bit grey-scale, after which the threshold of the image was adjusted so that only callose deposits were visible over the background. Callose deposits were then counted using the analyse particles tool. Four callose measurements were collected per leaf; each individual leaf was collected from a different plant. Six to eight plants were evaluated per treatment.
Preparation and selection of transgenic lines. The pNPR1::NPR1-YFP and p35S:: YFP constructs were generated by Gateway cloning and then used for Agrobacterium-mediated transformation of npr1-6 (SAIL_708F09) and Col-0 plants. Details for cloning and screening of transgenic lines are described in Supplementary Note 2.
Disease and BTH protection assays. Pst DC3000 inoculum was prepared at a concentration of~1-3 × 10 6 colony-forming units (CFU) mL −1 (an optical density at an absorbance wavelength of 600 nm (OD 600 ) of~0.001) as described in Supplementary Note 3. For time course disease assays with ics1 and pifq mutants, a lower inoculum of~4-5 × 10 5 was used to assess disease phenotypes in the mutant plants prior to saturation of bacterial growth. Bacterial infiltration and quantification are described in Supplementary Note 3.
In vitro growth assay. An overnight culture of Pst DC3000 (grown in LM + Rif) was diluted to an OD 600 of 0.05 (LM only), which was then divided into six 125 mL flasks. Three flasks each were incubated with shaking (225 RPM) at 23°C or 30°C. Bacterial populations were assessed at each time point by measuring the OD 600 with a spectrophotometer (DU 800, Beckman Coulter).
Effector translocation assay. Following temperature treatments, plants were infiltrated using a needless syringe with a high inoculum (2-4 × 10 7 CFU mL −1 , 0.25 mM MgCl 2 ) of Pst DC3000 or the hrcC − mutant carrying the P nptII ::avrPto-CyaA plasmid 35 , or Pst DC3000 containing either the P tac ::avrPtoB-CyaA 67 , P tac :: hopU1-CyaA 67 or P tac ::hopG1-CyaA 68 plasmid. For translocation experiments using plants pre-treated with mock or BTH, BTH-treated plants were infiltrated with inoculum~1.5 times higher than that used to infiltrate mock-treated plants. This was done to enable samples with equivalent bacterial populations 4 hpi. Leaf discs were harvested using a biopsy punch 4-6 hpi for both bacterial population quantification and cAMP quantification, which was normalized by total plant protein. cAMP was extracted and quantified using the Direct cAMP ELISA kit (ENZO) according to the manufacturer's protocol. Total protein was quantified using a Quickstart Bradford assay (BioRad) according to the manufacturer's protocol.
Nuclear fractionation and western blotting. Following temperature and chemical treatments, a minimum of 0.5 g (FW) leaf tissue was harvested and the mass recorded prior to flash freezing in liquid nitrogen. After grinding, cell lysate was isolated using the CelLytic PN Isolation/Extraction Kit (Sigma), using the semipure fractionation method according to the manufacturer's protocol. Following isolation and fractionation, the whole-cell lysate and cytosolic fractions were diluted with an equal volume of 4× SDS Laemmli sample buffer (125 mM Tris-HCl, pH 6.8; 4% (w/v) SDS; 20% glycerol; 0.02% bromophenol blue; 5% (v/v) βmercaptoethanol) while the nuclei pellet was resuspended in 100 µl of 2× SDS Laemmli sample buffer, resulting in a nuclear-fraction sample with an 8-fold higher concentration than the whole-cell or cytosolic-fraction samples. All samples were boiled for 10 min at 95°C, equal volumes of each were loaded in 4-12% SDS-PAGE gradient gels (NuPAGE, Novex) and run for 40-50 min at 200 V. Proteins were transferred to PVDF membranes at room temperature for 1 h at 25 V. Primary antibodies used were α-GFP (1:7500, Abcam, cat# ab290), for detection of NPR1-YFP; α-PR1 69 (1:5000, gift from Xinnian Dong, Duke University); α-UGPase (1:3000, Agrisera, cat# AS05 086) and α-H3 (1:10,000, Agrisera, AS10 710). The secondary antibody used for all blots was a goat α-rabbit (1:20,000, ThermoFisher Scientific, cat# 31460). A third protein gel was run and stained with Coomassie to enable visualization of protein loading. Uncropped images for all western blots are provided in Supplementary Fig. 16.
Confocal microscopy. Images were taken on a Zeiss 510 Meta Confocal Laser Scanning system configured on a Zeiss AxioObserver.Z1 inverted microscope (Carl Zeiss Microscopy, Thornwood, NY) using a Zeiss ×63 C-Apochromat water immersion objective (NA 1.2). Sequential imaging with a Kalman averaging of 4 and pinhole set to 120 µm was used to capture images from a single confocal plane. Bright field (BF) images were recorded using an Argon 514 nm laser. YFP was excited with an Argon 514 nm laser set at 20% and fluorescence emission was recorded using a 535-565 band pass filter. Aim Image Browser Software (Zeiss LSM) was used to add scale bars and for adjustments to brightness and contrast, which were as follows: NPR1-Y1 images, contrast 60%; npr1 images, contrast 65%; 35S::YFP YFP images, brightness 52%, contrast 55%; 35S::YFP BF images, brightness 48%, contrast 55%. Images were also adjusted for sharpness and contrast using Microsoft PowerPoint as follows: all images, sharpness + 50%; npr1 BF images, contrast (+20%); 35S::YFP YFP images, −40% contrast; 35S::YFP BF images, +20% contrast.
Statistical analysis. A minimum of three independent experiments were done for all assays unless otherwise indicated. Sample size used for experiments is indicated in the figure legends and was selected based on previously published work of similar experiments shown to be sufficient for statistical analyses. Statistical significance was determined using a Student's t test (Excel) for pairwise comparisons, a one-way analysis of variance (ANOVA) followed by a Dunnett's test (Prism 6, GraphPad Software, Inc.) for comparisons of multiple test samples to the same control, or by conducting a 2 × 2 factorial ANOVA followed by Tukey's honest significant difference (HSD) test (RStudio (https://www.rstudio.com/) for multivariate analyses. In the case of unequal variances as determined by the Brown-Forsythe test (α ≤ 0.05, Prism 6, GraphPad Software, Inc.), data were logtransformed prior to conducting statistical analysis.
Data availability. All the RNA-seq reads for this work have been submitted to NCBI under BioProject PRJNA325245. The authors declare that all other data supporting the findings of this study are within the manuscript and its supplementary files or are available from the corresponding author upon request.