Natural variation in temperature-modulated immunity uncovers transcription factor bHLH059 as a thermoresponsive regulator in Arabidopsis thaliana

Temperature impacts plant immunity and growth but how temperature intersects with endogenous pathways to shape natural variation remains unclear. Here we uncover variation between Arabidopsis thaliana natural accessions in response to two non-stress temperatures (22°C and 16°C) affecting accumulation of the thermoresponsive stress hormone salicylic acid (SA) and plant growth. Analysis of differentially responding A. thaliana accessions shows that pre-existing SA provides a benefit in limiting infection by Pseudomonas syringae pathovar tomato DC3000 bacteria at both temperatures. Several A. thaliana genotypes display a capacity to mitigate negative effects of high SA on growth, indicating within-species plasticity in SA—growth tradeoffs. An association study of temperature x SA variation, followed by physiological and immunity phenotyping of mutant and over-expression lines, identifies the transcription factor bHLH059 as a temperature-responsive SA immunity regulator. Here we reveal previously untapped diversity in plant responses to temperature and a way forward in understanding the genetic architecture of plant adaptation to changing environments.

Temperature impacts plant immunity and growth but how temperature intersects with endogenous pathways to shape natural variation remains unclear. Here we uncover variation between Arabidopsis thaliana natural accessions in response to two non-stress temperatures (22˚C and 16˚C) affecting accumulation of the thermoresponsive stress hormone salicylic acid (SA) and plant growth. Analysis of differentially responding A. thaliana accessions shows that pre-existing SA provides a benefit in limiting infection by Pseudomonas syringae pathovar tomato DC3000 bacteria at both temperatures. Several A. thaliana genotypes display a capacity to mitigate negative effects of high SA on growth, indicating withinspecies plasticity in SA-growth tradeoffs. An association study of temperature x SA variation, followed by physiological and immunity phenotyping of mutant and over-expression lines, identifies the transcription factor bHLH059 as a temperature-responsive SA immunity regulator. Here we reveal previously untapped diversity in plant responses to temperature and a way forward in understanding the genetic architecture of plant adaptation to changing environments.

Author summary
Temperature has a profound effect on plant innate immune responses but little is known about the mechanisms underlying natural variation in transmission of temperature signals to defence pathways. Much of our understanding of temperature effects on plant immunity and tradeoffs between activated defences and growth has come from analysis of the common Arabidopsis thaliana genetic accession, Col-0. Here we examine A. thaliana genetic variation in response to temperature (within the non-stress range-

Introduction
Analysis of phenotypic variation is a means to identify genes and networks underlying complex traits [1]. Environment shapes plant phenotypes and is a driver of adaptation to new habitats [2][3][4]. Temperature, as one key environmental variable, impacts plant physiology, growth and responses to abiotic and biotic stresses [5,6]. As temperature fluctuations across the globe increase, it is important to determine how plants integrate temperature signals with plant developmental and stress programs, and the genetic networks enabling resilience to climate change.
There has been recent progress in elucidating processes that coordinate temperature with plant endogenous pathways. Phytochromes act as thermosensors, coupled with their central integrative role in light quality perception and signalling [7][8][9]. In Arabidopsis thaliana, phytochrome B (phyB) regulates the bHLH transcription factor phytochrome interaction factor 4 (PIF4) to prioritize growth over immune responses at elevated temperatures via the De-Etiolated 1 (DET1) and Constitutive Photomorphogenic 1 (COP1) photomorphogenic regulatory module [10,11]. Also, sumoylation events mediatated by SUMO E3 ligase SIZ1 affect COP1/ PIF4-dependent growth-defense tradoffs at high temperatures [12]. In the cold, membranebound NAC transcription factor NTL6 is released to induce disease resistance [13]. The Early Flowering 3 (ELF3) scaffold protein for a temperature-responsive transcription repressor evening complex has a directly thermosensitive prion-like domain [14]. Therefore, temperature signals influence transcriptional regulation of immunity and growth.
Coordination between temperature and plant resistance to pathogen infection is determined by phytohormone pathways with contrasting roles in growth and defence, and by temperature effects on in planta microbial metabolism and infectivity [6,[15][16][17][18]. In A. thaliana, the two major protective layers against microbial pathogens: cell surface-based pattern-triggered immunity (PTI) and intracellular effector-triggered immunity (ETI) respond differently to ambient temperature, with PTI being preferentially activated at elevated and ETI at lower temperatures [17,19,20]. Heat stress events can, however, drastically reduce PTI responsiveness [21]. Gradual depletion of the alternative histone H2A.Z in nucleosomes with increasing temperature [22] is associated with increased PTI-dependent gene expression at the expense of ETI [19]. Hence, temperature effects registered at the chromatin are also important for plant immunity outputs.
The plant stress hormone, salicylic acid (SA), mediates basal and systemic immunity to biotrophic and hemi-biotrophic pathogens by reprogramming cells for defence via the transcriptional co-regulator, nonexpressor of PR1 (NPR1) [23]. In A. thaliana, pathogen-induced SA is generated mainly by the isochorismate synthase1 (ICS1) pathway [24,25]. Induction of ICS1 expression and pathogen resistance in A. thaliana basal and ETI responses are compromised at temperatures above 23-24 o C [5,16,26]. In A. thaliana accession Col-0, increased SA was also responsible for plant stunting after a shift from 23˚C to near chilling conditions (5˚C) [27]. Exposure of other A. thaliana accessions to 10˚C revealed genotype-specific expression patterns for~75% cold-regulated transcripts [28], highlighting the extent of natural variation in temperature-modulated gene expression. Additionally, SA is an inducer of thermogenesis in certain plant species [29], broadening its role in temperature responses.
Here we investigate A. thaliana natural genetic variation in immunity and growth responses to two temperature regimes (22/20˚C and 16/14˚C). Our aim was to assess the phenotypic space in immunity x growth interactions over a non-stress temperature range for this species. Using SA accumulation in leaves of 105 genetically diverse accessions as a first proxy for defence homeostasis, we uncover variation in temperature modulation of SA and in the relationship between leaf SA and biomass. At both temperatures, there is a measurable benefit of high initial SA levels on plant resistance to a leaf-infecting bacterial pathogen, Pseudomonas syringae pathovar tomato DC3000 (Pst DC3000), after it has passed a stomatal barrier (poststomatal resistance). A genome-wide association study of temperature x SA variation identifies the bHLH059 transcription factor as a new thermoresponsive immunity component.

SA chemotyping of A. thaliana plants at two temperature regimes
To measure temperature-modulated SA accumulation we selected 105 A. thaliana accessions from the HapMap population based on genetic diversity and geographical distance [45]. Most accessions (80%) originate from Eurasia populations and we included naturalized lines from America, Africa, New Zealand and Japan (S1 Table). Individual plants were grown in separate pots to avoid competition/shading and, as a randomized design in controlled cabinets, kept at 16˚C/14˚C or 22˚C/20˚C and 12 h light/dark cycle within the non-stress range for A. thaliana [46]. We then determined biomass and SA contents of 5-week-old plants under each temperature regime. Because there was a strong correlation between fresh and dry plant weights (S2 Table), we used above-ground fresh weight (FW) as a measure of biomass.
To quantify SA in a large number of samples, we used a high-throughput SA biosensorbased luminescence method [47,48] (Methods). This provided total SA measurements in medium (Ws-0 and Col-0) and high SA (C24 and Est-1) accumulating accessions [41,49] with an accuracy comparable to GC-MS (S1A Fig). The biosensor method was less reliable for quantifying low levels of free SA, the biologically active form (S1B Fig) [50]. There was a high correlation between free and total SA amounts in GC-MS assays of 15 tested 5-week-old accessions with contrasting SA contents at 22˚C (S2 Fig). We therefore used biosensor-based total SA as a measure of SA accumulation at the two temperatures. As plant age influences SA accumulation and outputs [40], we assessed whether differential SA accumulation between accessions is captured reliably at 5 weeks. For this, total SA was quantified in five accessions which in pilot studies had shown low (Sha, Col-0), intermediate (Est-1) or high (An-1) total SA contents, together with a Col-0 isochorismate synthase SA biosynthesis mutant sid2-1 [24], over a 7-week time course. Total SA accumulation trends seen in 5-week-old plants persisted over the course of development from 4-7 weeks regardless of flowering time (S3 Fig).

Genetic variation in A. thaliana SA-growth tradeoffs
At each temperature there was considerable genetic variation in plant biomass and total SA levels between accessions (Fig 1A and 1B and S1 Table). Surprisingly, total SA did not show a general tendency to increase in plants grown at 16˚C compared to 22˚C, although biomass at 16˚C was lower (Fig 1A-1C). Therefore, increased SA at cooler temperatures reported previously for accession Col-0 [13,27], and also found here for Col-0 (S1 Table), appears not to be generalizable for A. thaliana. Moreover, comparing total SA contents with biomass in each accession revealed that at 16˚C and 22˚C there was an extremely weak negative correlation between total SA levels and above ground FWs (Fig 2A and 2B). One third of accessions with total SA contents >1 μg/g FW had a biomass above the median at each temperature, suggesting that there is within-species genetic and/or phenotypic plasticity in SA-growth tradeoffs.
Negative effects of defence on growth in high SA backgrounds might be mitigated by imposing a higher induction threshold for SA immunity. We therefore measured expression of the SA-responsive pathogenesis-related 1 gene (PR1) at 22˚C in selected 5-week-old accessions with high total SA amounts (>1 μg/g FW) and high biomass (>1.5 g) (Ven-1, PHW-13, Kas-2), accessions with high total SA and low biomass (>0.5 g) and some leaf necrosis (Gy-0, Spr1-2, Mz-0), or with low total SA (<0.3 μg/g FW) and varied biomass (Mrk-0, Col-0, Oy-0) (Fig 2C-2F). For these nine tested accessions, high total SA was accompanied by elevated PR1 expression but not always stunting (Fig 2C-2F). The data suggest that mechanisms other than responsiveness to SA reduce antagonism of growth in some A. thaliana genetic backgrounds. Whether natural variation in SA x growth tradeoff at 22˚C also involves cold-regulated mechanisms of growth reduction such as phytochrome signalling [7], auxin transport [51] and/or gibberellic acid and inhibitory DELLA interplay [52,53], remains unclear. Analysis of biomass and total SA levels in 105 A. thaliana accessions grown for 5 weeks at two temperature regimes (22˚C and 16˚C) reveals natural variation in SA homeostasis in response to temperature. Data are represented as boxplots. A) Aboveground fresh weight (FW) of plants grown at 16˚C (n = 420, four biological replicates) and 22˚C (n = 315, three biological replicates). Statistical difference according to kruskal-wallis non parametric test with p-value<0.05 is indicated above the plot with a star B) Leaf total SA contents of plants grown at 16˚C (n = 420) and 22˚C (n = 315). Statistical difference according to kruskal-wallis non parametric test is indicated above the plot. NS = not significant. C) Distribution of mean leaf total SA differences in accessions between temperature regimes (n = 105). Accessions with higher SA contents at 16˚C than 22˚are at the positive side and accessions with higher SA contents at 22˚C than at 16˚C on the negative side of the plot. Accessions with little or no change in SA contents in response to temperature score around 0. https://doi.org/10.1371/journal.pgen.1009290.g001

SA accumulation in response to temperature is genotype-specific
Because SA amounts in leaves of different accessions did not relate strongly with reduced growth, we examined the effect of temperature on SA homeostasis regardless of biomass. From the initial 105 A. thaliana accessions, we selected lines that accumulated higher total SA at 22˚C than 16˚C (Ven-1, Mz-0, Nok-3), lines that showed no variation in total SA levels between the two temperatures (Se-0, NFA-8), and lines displaying higher total SA at 16˚C than . Letters indicate significant differences after FDR multiple testing correction in one-way ANOVA. D) Leaf total SA contents in 5-week-old plants of nine A. thaliana accessions grown at 22˚C (n = 3 biological replicates). Letters indicate significant differences after FDR multiple testing correction in oneway ANOVA. Data was log(10) transformed for statistical analysis. E) Expression of PR1 relative to SAND reference gene in 5-week-old plants of 9 A. thaliana accessions grown at 22˚C (n = 3 biological replicates). Letters indicate significant differences after FDR multiple testing correction in one-way ANOVA. Data was log(10) transformed for statistical analysis. F) Visual growth phenotypes of 5-week old A. thaliana accessions examined in c) to e). at 22˚C (Fei-0, Ei-0, Bay-0, Est-1) (S1 Table and Fig 3A). Reference accessions Col-0 and Ler-0 had relatively small but opposing total SA responses to temperature (S1 Table and Fig 3A). We concluded that even a moderate change in temperature within the normal range of A. thaliana (here 6˚C) exposes variation in SA pathway homeostasis. To further validate our findings, we repeated total SA quantitation by GC-MS analysis in Ven-1, Mz-0, Fei-0, Ei-2 and Col-0 grown at 16˚C and 22˚C and observed similar temperature-dependent patterns as with the biosensor-based method (Figs 3A and S4A). Temperature-dependent expression of the SAresponsive gene PR1 in Ven-1, Mz-0, Fei-0, Ei-2 and Col-0 correlated with total SA contents (S4B Fig), suggesting that different SA contents impact SA-dependent defense gene expression. Because differential SA accumulation in response to temperature might be driven by changes in SA biosynthesis, we measured expression of three SA biosynthesis genes in the same leaf tissues of these accessions. Isochorismate synthase 1 (ICS1) and avrPphB SUSCEPTIBLE 3 (PBS3) are key enzymes for pathogen-induced SA synthesis [25] and phenylalanine ammonia lyase 4 (PAL4) regulates SA basal accumulation [54]. None of these genes showed differential regulation in response to temperature in Ven-1, Mz-0, Fei-0, Ei-2 and Col-0 (S10C-S10E Fig). These data suggest that temperature-dependent SA accumulation within this 16˚C-22˚C range is not determined by strongly altered expression of SA biosynthesis genes.

Temperature-modulated SA impacts bacterial pathogen growth in leaves
With differences of up to 5 μg total SA/g FW in accessions grown under the two temperature regimes (Fig 1C), we anticipated temperature-dependent variation in immune responses between accessions, as suggested by the PR1 expression profiles of selected genotypes at 22˚C (Figs 2E and S4B). We spray-inoculated leaves of 11 5-week-old accessions showing diverse total SA levels at 16˚C and 22˚C ( Fig 3A) with virulent Pst DC3000 at these two temperatures. Pst DC3000 produces the JA-Ile mimic coronatine (COR) which promotes reopening of leaf stomata to counter bacterial PAMP-induced stomatal closure and increase bacterial entry to the leaf apoplast [18]. Host-produced SA promotes stomatal closure and post-stomatal resistance to Pst DC3000 [18,55,56]. At 3 h post inoculation (hpi), Pst DC3000 levels inside leaves were unchanged between temperatures in each accession but showed up to 10-fold differences between accessions (S5A Fig), suggesting that differences in early stomatal entry of bacteria into leaves is not a major variable between 16˚C and 22˚C in these accessions. After measuring Pst DC3000 growth in leaves of plants at 4 dpi, we found a robust inverse correlation between temperature-modulated total SA accumulation and bacterial growth across accessions ( Fig  3B). Thus, in accessions showing a rise in total SA between 16˚C and 22˚C there was increased resistance to Pst DC3000 and the opposite trend was observed in plants which had reduced total SA between 16˚C and 22˚C ( Fig 3B). Accessions which responded negligibly to temperature at the level of SA accumulation showed no difference in temperature effects on Pst DC3000 infection ( Fig 3B). These data reveal a positive relationship between temperaturemodulated total SA accumulation and limitation of Pst DC3000 growth in leaves of the 11 tested A. thaliana accessions.
We observed variation between accessions in the degree to which total SA differences impacts bacterial resistance. For example in Ler-0, a rise of only 0.38 μg/g FW total SA between 16˚C and 22˚C resulted in a substantial (1.5 log 10 ) reduction in bacterial numbers ( Fig 3B). Est-1, with a much higher total SA differential (1.68 μg/g FW) between temperatures, showed only a small (0.5 log 10 ) difference in bacterial growth, whereas in Fei-0 a 2.56 μg/g FW total SA change translated to a 2.0 log 10 bacterial growth difference ( Fig 3B). All accessions in

PLOS GENETICS
Arabidopsis thaliana natural variation and genetic architecture of temperature-modulated immunity immunity. When cultured on liquid M9 minimal salt medium containing sorbitol as carbon source over a 56 h time course, Pst DC3000 grew more slowly at 16˚C than at 22˚C during the exponential phase (S6 Fig). This result emphasizes the influence of A. thaliana host genotype in determining temperature effects on bacterial growth in leaves. In order to delineate genotype-dependent stomatal vs apoplastic defenses at 16˚C and 22˚C, we syringe-infiltrated Pst DC3000 bacteria into leaves of five of the 11 differentially responding genotypes (Ven-1, Mz-0, Se-0, Fei-0 and Col-0) to bypass stomata, and performed a bacterial growth time course over 4 days at the two temperatures (S7A- S7E Fig). We observed similar temperature-dependent bacterial growth outcomes as in the spray inoculated plants (compare S7 Fig with Fig 3B). Therefore, we propose that apoplastic immunity contributes substantially to the observed genetic variation in temperature effects on Pst DC3000 infection.

High SA accumulation prior to infection increases bacterial immunity
We have shown that A. thaliana SA amounts before infection correlate positively with resistance to virulent Pst DC3000. Next we tested whether temperature effects on Pst DC3000-induced SA might also contribute to resistance. For this, leaves of accessions Ven-1, Mz-0, Se-0, Fei-0, Ei-2 and Col-0 grown at 16˚C or 22˚C were sprayed with Pst DC3000 or buffer (mock) and total SA measured at 24 hpi. Similar temperature effects on total SA accumulation were observed in these accessions after mock treatment as in untreated plants (compare Fig 4 (mock) and Fig 3A). After Pst DC3000 inoculation, there were no significant temperature differences in total SA accumulation between accessions (Fig 4). Hence, accessions with lower starting (basal) SA at 16˚C or 22˚C induced SA to comparable levels as high initial SA accumulators at 24 hpi (Fig 4). We concluded that temperature modulated SA accumulation before infection is an important determinant of A. thaliana immunity to Pst DC3000 bacteria in the 16˚C to 22˚C range.

SA underlies differential temperature effects on resistance to bacteria
We tested whether the observed temperature-dependent differences in bacterial resistance between accessions are determined by SA levels. For this, we introduced into different A. thaliana accessions a bacterial NahG (salicylate hydroxylase) gene which breaks down SA to catechol [57]. A single NahG transformant line from each accession was selected after checking total SA depletion at the temperature the parental accession produced highest SA amounts (S8 Fig). The SA-depleted (NahG) accessions, together with existing Ler-0 and Col-0 NahG lines [58,59] grown at 16˚C or 22˚C, were spray-inoculated with Pst DC3000 and bacterial titers measured in leaves at 3h and 4 dpi. There was again no detectable temperature effect on Pst DC3000 early stomatal entry to leaves at 3 hpi (S5B Fig). In contrast to the parental responses (Fig 3B), corresponding NahG lines had lost temperature-dependent differential resistance to Pst DC3000 growth at 4 dpi ( Fig 3C). This loss was also observed in Col-0 sid2-1 (Fig 3C). These data suggest that temperature-regulated SA accumulation directly or indirectly underlies the observed temperature effects on resistance to virulent Pst bacteria.
Notably, variation in Pst DC3000 growth between wild-type accessions persisted in the corresponding NahG transgenic lines that was independent of temperature ( Fig 3B and 3C). For example, up to a 1000-fold difference in Pst DC3000 titers was observed between the most susceptible (Nok-3 and Ler-0) and resistant (Mz-0 and Ei-2) genotypes ( Fig 3C). These data highlight a substantial contribution of SA-independent processes in limiting Pst DC3000 growth which, unlike the SA-dependent resistance, are unaffected by changes in temperature within the 16˚C to 22˚C range. We concluded that there is within-species genetic variation in both temperature-dependent SA and temperature-independent (non-SA) defences shaping A. thaliana post-stomatal immune responses to bacteria.

Genetic architecture of SA regulation by temperature in A. thaliana
After assessing SA homeostasis in response to temperature in 105 A. thaliana accessions, we examined whether specific phenotypes fit a global distribution pattern, using the coefficients of a GLM (Generalized Linear Model; Materials and Methods) to colour-coded phenotypes at occurrence sites. No obvious geographic or climatic distribution patterns were found for temperature-dependent total SA regulation (Methods) (Fig 5A; representing only extended European accessions for clarity).
Broad sense heritability of SA accumulation was calculated to be 0.79 at 16˚C and 0.76 at 22˚C, indicating a sizable genetic underpinning to this trait. To explore the trait genetic architecture we performed temperature x total SA association mapping on 99 accessions using the GWAPP tool [60] and coefficients of the GLM as a phenotype (Methods). One major peak on the upper arm of chromosome 4 contained six significant single nucleotide polymorphisms (SNPs) after Bonferroni multiple testing correction (Fig 5B and S3 Table). Two additional peaks were found on chromosomes 1 and 4, each with one significantly associated SNP after Bonferroni correction (Fig 5B and S3 Table). Immediate and neighbouring genes within 10 kb each side of the significant SNPs were considered as candidates (S3 Table) [61]. Two SNPs on the upper arm of chromosome 4 fall in the bHLH059 transcription factor gene from AtbHLH group XI [62], in which a T-DNA insertion in Col-0 led to slightly increased resistance to a virulent strain of the oomycete pathogen Hyaloperonospora arabidopsidis [63]. Since bHLH059 knock-out mutants are not sterile (S10F Fig) [63] we use bHLH059 here rather than its initial name UNFERTILIZED EMBRYO SAC 12. None of the six remaining significant SNPs considered by the GWAPP tool was in a gene related to SA biosynthesis/signalling, temperature responses, defence or cell death regulation (S3 Table) except for At4g02600, a homologue of barley mildew resistance locus 1 (ATMLO1) [64]. However, ATMLO1 expression was found to be specific to early development [64]. In the GWAS analysis, A. thaliana genes involved in thermosensory regulation, such as PIF4, PhyB, NTL6 or genes controlling alternative histone H2A. Z recruitment [7,10,13,21], were not found to be associated with temperature-dependent SA regulation. Because Bonferroni is conservative, we extended the list of candidate genes in the

Comparison of bHLH059 functions with thermosensory immune regulator PIF4
As bHLH059 is supported by two significantly associated SNPs on chromosome 4 (Fig 5B and  5C and S3 Table), we investigated its role in temperature-dependent SA accumulation and immunity. Using TAIR.10 sequence data to identify bHLH059 genomic polymorphisms with the Col-0 reference genome (http://signal.salk.edu/atg1001/3.0/gebrowser.php) we examined sequences from accessions with extreme or intermediate temperature x SA phenotypes that were also used for the Pst DC3000 infection assays (Fig 3). Variation was uncovered in bHLH059 coding and regulatory sequences (Fig 5C). While accessions Bay-0 and Se-0 have several deletions and nucleotide exchanges in the bHLH059 coding sequence relative to Col-0, all other considered accessions only display one SNP in the coding sequence ( Fig 5C). Since this SNP leads to a synonymous mutation, we reasoned that variation in expression of bHLH059 rather than protein sequence might underlie temperature-modulated SA and/or bacterial resistance. Indeed, bHLH059 expression measured in five key accessions revealed differential temperature regulation (S9 Fig). Whereas bHLH059 expression was stable between 22˚C and 16˚C in Ven-1 and Mz-0 with higher SA accumulation at 22˚C, it increased at 22˚C in Fei-0, Ei-2 and Col-0 from the other phenogroup with reduced SA at 22˚C (S9 Fig).
Although bHLH059 lacks a diagnostic SNP profile for its phenogroups in promoter and genic sequences (Fig 5C), the gene has a phenogroup expression pattern. Therefore, trans-regulatory elements might play a role in temperature-dependent bHLH059 expression.
A T-DNA line with an insertion in the last intron of bHLH059 leading to a truncated transcript (SALK_13303; bhlh059-13 verified by qRT-PCR (Fig 6A and S4 Table)), and a β-estradiol inducible bHLH059 transgenic line (βE::bHLH059) from the TRANSPLANTA Col-0 collection [65], were selected for phenotyping. Because the bHLH TF PIF4 and its closest homologue PIF5 are temperature-sensing immunity regulators under control of the PhyB thermosensory pathway in Col-0[10], we explored redundancy or cooperativity between PIF4/PIF5 and bHLH059 by including the pif4-2 pif5-3 double mutant and a PIF4::PIF4HA pif4-101 over-compensating transgenic line [66,67] in our assays. By performing a 2 way-ANOVA on expression of BHLH059, PIF4 and PR1, SA contents and Pst DC3000 bacterial counts at 0 and 4 dpi in response to genotype and temperature in this set of mutant lines, we found that temperature significantly affects all our variables except bHLH059 expression where it is marginally significant (p-value = 0.08) and bacterial counts at 0 dpi, in line with results shown in S5 Fig. We therefore tested for differences between mutant lines within each temperature regime to determine effects of bHLH059 and PIF4 misregulation on SA and bacterial resistance (Fig 6). In 5-week-old Col-0 plants at 16˚C, bHLH059 expression was low and increased 0.5-fold at 22˚C (Fig 6A). As expected, bhlh059-13 had no detectable full-length transcript whereas the estradiol-untreated βE::bHLH059 line expressed 2-fold higher bHLH059 than Col-0 at both temperatures (Fig 6A). PIF4 expression was undetectable in pif4-2 pif5-3 and was elevated in the PIF4::PIF4HA pif4-101 line at 16˚C compared to Col-0, and further boosted in PIF4::PIF4HA pif4-101 leaves at 22˚C (Fig 6B). Loss or gain of PIF4 expression, respectively in pif4-2 pif5-3 and PIF4::PIF4HA pif4-101, did not alter bHLH059 expression at either temperature ( Fig 6A). Reciprocally, PIF4 expression which was higher than bHLH059, did not change within the 6˚C temperature range in Col-0 or bhlh059-13 and βE::bHLH059 lines (Fig 6B). Therefore, bHLH059 and PIF4/PIF5 do not influence each other's expression under the tested conditions.

bHLH059 has features of a temperature-responsive immunity regulator
We quantified total SA in the above lines at 16˚C and 22˚C. Col-0 SA levels decreased with increased temperature (Fig 6C) as observed before (Fig 3A). At 16˚C, the bhlh059-13 mutant had similar total SA amounts as Col-0 but, unlike Col-0, maintained the same SA level at 22˚C (Fig 6C). Strikingly, 2-fold over-expression of bHLH059 in the βE::bHLH059 line led to low total SA accumulation at both temperatures ( Fig 6C). Therefore, mis-regulation of bHLH059 alters SA accumulation in response to temperature, with higher bHLH059 expression dampening SA levels, similar to the behaviour of accessions Fei-0 and Ei-2 (Figs 3A and S9). We found that pif4-2 pif5-3 did not alter temperature modulation of total SA but that PIF4 over expression (in PIF4::PIF4HA pif4-101) reduced total SA in plants grown at 16˚C and further at 22˚C (Fig 6C). These data suggest that bHLH059 and PIF4 operate differently in transmitting temperature information to SA accumulation.
Next we tested whether the temperature x SA profiles in the above lines tally with changes in SA-based immunity by quantifying PR1 expression and Pst DC3000 growth in 5-week-old plants at 16˚C and 22˚C. PR1 expression and resistance to Pst DC3000 correlated with SA regulation by temperature in these lines with two exceptions (Fig 6C-6E). At 22˚C, PR1 expression in bHLH059-13 was lower than expected based on its SA accumulation and resistance to Pst DC3000 (Fig 6C-6E). By contrast, the PIF4 overexpression line (PIF4::PIF4HA pif4-101) exhibited equivalent low SA and PR1 expression but higher Pst DC3000 susceptibility than Col-0 at 22˚C (Fig 6C-6E). At 3 hpi, Pst DC3000 bacterial entry was similar between temperature regimes and lines and neither genotype nor temperature had significant influence ( Fig  6F), suggesting that the observed bHLH059 and PIF4 effects on bacterial resistance are mainly after initial stomatal entry. Taken together, these data suggest that bHLH059 participates in temperature regulation of SA immunity and do not support a conjunction of bHLH059 and PIF4 pathways in this temperature response.

bHLH059 mis-regulation does not alter A. thaliana development
A. thaliana PIF4 has an important role in thermomorphogenesis in which it negatively regulates pathogen immunity to favour plant growth at higher temperature [6,10]. Mis-regulation of PIF4 and its homologues alters hypocotyl and petiole elongation, growth and onset of flowering. We therefore compared developmental phenotypes of the bhlh059-13, βE::bHLH059, pif4-2 pif5-3, PIF4::PIF4HA pif4-101 lines and Col-0 after 5 weeks at 16˚C and 22˚C and a 12 h light/dark cycle, as in the previous assays. The bhlh059-13 mutant and βE::bHLH059 over expression line resembled Col-0 in stature (Fig 7A). By contrast, pif4-2 pif5-3 plants were stunted at 22˚C and PIF4::PIF4HA pif4-101 had longer petioles (Fig 7A), as reported [10,68]. Hypocotyl lengths were similar between bHLH059 lines and Col-0 at both temperatures, but were shorter in pif4-2 pif5-3 and longer in PIF4::PIF4HA pif4-101 at 16˚C and 22˚C (Fig 7B) [6]. Deviations from Col-0 were also observed in above-ground FWs of the tested PIF4 lines at 16˚C and 22˚C, but not in the bHLH059 lines (Fig 7C). Flowering time was similar in all lines at 16˚C but delayed in the PIF4 lines at 22˚C (Fig 7D). Delayed flowering was reported for PIF-deficient lines grown at~22˚C in short day conditions, whereas PIF over-expressors accelerated flowering [68,69]. Why this latter trend is not observed under our growth conditions remains unclear. Phenotypic analysis of an independent bHLH059 T-DNA insertion line (SALK_010825C; bhlh059-01) [63] compared to wild-type Col-0 showed that it behaves similarly to bhlh059-13 at 16˚C and 22˚C at the level of bHLH059 and PR1 expression, total SA accumulation and plant biomass (S10A-S10E Fig). These data lead us to conclude that bHLH059 expression in A. thaliana Col-0 impacts temperature modulation of SA immunity without markedly altering developmental traits and that bHLH059-related thermosensory processes affecting SA accumulation and immunity to Pst DC3000 are distinct from those con-

Discussion
Here we explored A. thaliana natural variation in response to temperature impacting SA accumulation, growth and resistance to bacterial (Pst DC3000) infection. One aim was to

PLOS GENETICS
determine differential temperature effects on immunity within a non-stress range, taking SA as an initial proxy for plant defence status. A second aim was to identify potential benefits and costs of accumulating high or low SA at a particular temperature. By testing 105 genetically diverse A. thaliana accessions, we uncover variation in total leaf SA accumulation between the 16˚C and 22˚C temperature regimes. We establish that increased SA amounts do not always correlate with reduced biomass, indicating a capacity of certain A. thaliana genotypes to mitigate negative effects of high SA levels on growth. Using a set of 11 selected accessions covering the range of observed temperature-modulated SA and growth responses, we detect a robust positive relationship between total SA in leaves prior to infection and restriction of Pst DC3000 growth, representing a possible benefit of accumulating SA. From an association study of temperature x SA in 99 of the 105 accessions, we identify bHLH059 as a strong candidate for thermoresponsive control of SA immunity to Pst DC3000. This analysis uncovers diversity in plant responses to temperature and a way forward to understand the genetic architecture of plant adaptation to changing environments.
We were able to group A. thaliana accessions into three broad classes based on increased, decreased or stable total SA contents associated with the 16˚C-22˚C temperature difference (Figs 1C and 3A). A. thaliana Col-0, the most studied accession for temperature effects on immunity and growth [16,17,26,27,30,70], showed a comparatively weak negative SA accumulation trend with increased temperature (Figs 3A and 6C). In a previous study, higher SA in Col-0 plants grown at 5˚C compared to 23˚C contributed to growth retardation at the chilling temperature [27], consistent with tradeoffs between induced plant defences and growth [35,38].
A. thaliana autoimmunity phenotypes leading to growth inhibition and necrosis have been linked to increased SA [5,20,33]. It is therefore striking that~33% of the 105 accessions with >1μg/g FW total SA retained a biomass above the median of the tested genotypes (Fig 2A and  2B). That certain accessions with high SA exhibited increased resistance to Pst DC3000 or PR1 expression without a measurable biomass penalty (eg. Ven-1, Kas-2, PHW-13 in Figs 2C-2E and 3A and 3B), points to genotypic variation in the threshold at which SA leads to autoimmunity [28]. Presence of genetic modifiers of SA-related autoimmunity and stress sensitivity are evident from studies of different A. thaliana accessions [41,71]. Alterations in the hormone network controlling SA crosstalk with growth-promoting pathways might buffer against SA negative effects in some genotypes [72]. Notably, SA signalling contributes positively to petiole elongation in the A. thaliana Col-0 shade avoidance growth response [66]. Also, defences and growth were effectively uncoupled in A. thaliana Col-0 in a Jasmonate-Zim Domain (JAZ) repressor x phyB sextuple mutant, indicating that perturbation of the hormone transcriptional network can reduce defence-growth tradeoffs [43]. A. thaliana accession C24 displays an unusual broad-ranging tolerance to stress encounters with little negative impact on growth [73]. For accessions with different SA-growth relationships identified in our analysis (Fig 2C  and 2D), it will be interesting in future studies to pinpoint underlying stress network properties and whether the apparent benefit of high SA on bacterial resistance creates vulnerabilities to other environmental stresses or conditions. Phenotypic characterization of 11 differential accessions revealed a positive relationship between total SA accumulation in response to temperature and post-stomatal restriction of Pst DC3000 growth in leaves, with plants being more resistant at the temperature the respective unchallenged accession accumulated higher SA (Fig 3). SA amounts in plants within the studied 6˚C temperature range might thus be a predictor of basal resistance strength. Some accessions (eg. Ven-1, Nok-3, Ler-0, Fei-0) displayed more than 50-fold differences in Pst DC3000 growth at 4 dpi between the two temperatures ( Fig 3B) which is similar to the differential growth observed between virulent and avirulent Pst strains in leaves of A. thaliana accessions Col-0 or Ws-2 [70]. Therefore, even a moderate temperature change can have a similar impact on bacterial infection as pathogen effector-triggered immunity. Quantifying Pst DC3000 titres in the corresponding NahG-transgenic lines showed that SA depletion abolished the temperature effect on Pst DC3000 growth in all of the tested accessions (Fig 3). Hence, the temperature effect on resistance to bacterial growth appears to be an SA-dependent trait in these accessions. Huot et al (2017) established that SA signalling represents a major temperature-sensitive resistance node in A. thaliana accession Col-0, assessed over a warmer temperature range of 23-30˚C, and that high temperature suppression of immunity to Pst DC3000 was independent of PhyB and PIF4 [16]. These and our data emphasize the importance of temperature differences within the normal range experienced by A. thaliana on effectiveness of SA-based pathogen immunity. In Pst DC3000-inoculated or A. thaliana autoimmune backgrounds, increased SA accumulation at 20-22˚C compared to 28-30˚C was associated with higher expression of the ICS1 and PBS3 SA biosynthesis genes [15,35,70]. By contrast, we found that changes in basal SA levels between 16˚C and 22˚C were not accompanied by altered ICS1 or PBS3 expression in five A. thaliana accessions (S4 Fig), suggesting that basal SA accumulation is controlled by another mechanism across this moderate temperature range.
The differences in Pst DC3000 titres remaining between NahG-expressing accessions ( Fig  3C), expose a contribution of SA-independent processes to variation in resistance to virulent bacteria, as also indicated by a screen of 1041 A. thaliana accessions in response to spray-inoculated Pst DC3000 [74]. Mz-0, our most resistant accession, displays an autoimmune phenotype at 22˚C with chlorosis, stunting and high SA accumulation due to a hyperactive allele at the ACD6 locus that is also present in Est-1 [49]. Nevertheless, the Mz-0 NahG line retained strong SA-independent resistance to Pst DC3000 (Fig 3C). In Col-0, post-stomatal basal and effector-triggered immunity to Pst DC3000 strains can be divided into parallel SA-dependent and SA-independent resistance branches [56,75,76]. SA-independent resistance provides some protection against pathogens that can disable SA pathways. Our and other analyses suggest that part of that resilience might lie at the level of maintaining SA-independent immunity over a range of temperatures [17,19].
The differences in SA contents regulated by temperature between A. thaliana accessions were not linked to a geographical distribution pattern ( Fig 5A) and therefore it is not known whether this represents an adaptive trait to local climatic conditions [77]. Loci strongly associated with climate variables were enriched in amino acid-changing SNPs, indicating the presence of adaptive alleles [2]. There is increasing evidence for microhabitat effects such as edaphic conditions, intraspecific competition, herbivore distribution and altitude playing roles in local adaptation [3,78,79]. A study on the genetic basis to local adaptation of A. thaliana in Europe revealed several associated loci related to immunity [77]. Also, defence and cold response processes were associated with adaptive climate variables [2]. Taken together, these data suggest that temperature modulation of plant defences impact local adaptation.
Association mapping allowed us to link variation in temperature-dependent total SA to three loci on two chromosomes, with a strongly supported QTL on chromosome 4 ( Fig 5B). In A. thaliana Col-0, bHLH059 has features of a thermoresponsive immunity component because loss or mild over-expression of this gene disturbed temperature effects on SA accumulation and basal resistance to Pst DC3000 bacteria (Fig 6E). Interestingly, bHLH059 was identified as a weak negative component of Col-0 immunity and, in yeast 2-hybrid assays, as a potential defence hub connected to multiple NLRs [63].
Our comparative physiology and immunity phenotyping of PIF4/5 and bHLH059 misexpressed lines (Figs 6 and 7) suggest that these factors act independently in transmitting or processing temperature stimuli to immune and growth responses. Also, bHLH059 was not identified as a PIF4 transcriptional target [80]. Therefore, in line with Huot et al (2017) findings, we think it unlikely that temperature modulation of A. thaliana SA defences involves PIF4/5 signalling. Loss-of-function bHLH059 mutations in accession Col-0 enhanced resistance to Pst DC3000 without an obvious physiological or developmental cost at 22˚C (Figs 6E  and 7). It will be interesting to test if manipulating expression of bHLH059 and/or its paralogue bHLH007 (At1g03040) (61) in A. thaliana increases survival against pathogen infection over a range of temperatures.

Plant growth conditions
A.thaliana plants were grown under controlled conditions at 16±1˚C (day) and 14±1˚C (night) or 22±1˚C (day) and 20±1˚C (night), 60±10% relative humidity, 200μE m 2 s -1 light intensity and 12h day/night cycle. Seeds were first stratified in soil at 4˚C for 3 d. Plants were grown in individual 0.8l pots with commercial potting soil pretreated with entomopathogenic nematodes. Pots were distributed into trays in a fully randomized design for temperature x SA screening. One plant per accession was grown in each replicate. Data from three independent experiments (biological replicates) at 22˚C and four at 16˚C were used for analysis. Plants were grown in parallel in two growth chambers after ensuring replicability of results between chambers.

Bacterial infection and A. thaliana physiology assays
For Pst DC3000 infection assays and bHLH059/pif4 mutant characterization, plants were distributed in trays in a randomized design by genotype row. SA and gene expression assays were performed on three plants for each genotype in three independent experiments. Bacterial early entry into leaves through stomata was determined by measuring in planta bacterial titers at 3 hpi in a total of nine plants (with three individual plants per independent experiment). At 4 dpi, bacterial growth was determined in a total of 18 plants (derived from three independent experiments). Random groups of six genotypes were tested in parallel at each temperature with at least one wild type in each group. Trays were distributed randomly in a phytotron growth chamber and moved once a week to a new position. For spray-inoculation of 5-weekold plants with Pst DC3000, bacterial suspensions at 0.15 OD 600 in 10 mM MgCl 2 were used, as described [84]. Pst DC3000 infiltration into leaves was performed using a concentration of 0.0005 OD 600 injected with a flat-topped 1ml syringe.

Salicylic acid measurements
Luminescence produced by induction of an SA degradation operon coupled to a LUX cassette was used to measure SA in leaves using the biosensor-based method, as described [47,48]. Total and free SA was quantified by GC-MS as described [85]. For both methods, 100-200 mg leaf samples were frozen in liquid nitrogen and disrupted by a tissue lyzer (Retsch). Material was suspended in 250μl NaOAc 0.1M pH5.5 and mixed. Samples were centrifuged in a microfuge for 15 min, 200μl supernatants transferred to a 96 well PCR plate and treated with 4U of almond beta-glucosidase (Sigma) at 37˚C for 1.5 h. For bio-sensor measurements, 30μl sample was transferred to a 96 well black optiplate (Perkin Elmer) containing 60μl LB medium. A standard curve for SA (Sigma) was used in a volume of 10μl complemented with 20μl β-glucosidase-treated leaf extract of Col-0 NahG leaves to mimic leaf samples. The standard curve was designed and tested to measure SA concentrations from 0 to 20 μg total SA/g leaf fresh weight. The transgenic Acinetobacter luminescent strain was grown as described [85] and 50 μl bacterial suspensions (OD 600 0.4) were added to each optiplate sample before incubating at 37˚C for 1h. The optiplate was then read by a luminometer (Berthold technologies) measuring luminescence emitted by each sample in 1/3s. Mean luminescence taken from three plate readings was used to calculate total SA concentrations. Because luminescence increase in the standard curve was non-linear we interpolated the data points using the approxfun() function in R (Cran 3.2.2.).
Gas chromatography-mass spectrometry (GC-MS)-based analyses of plant metabolites (SA and SAG) (S4A Fig) was performed as detailed in Hartmann et al 2018 [86], with modifications. 50 mg of pulverized, frozen leaf samples were extracted twice with 1 ml of MeOH/50 mM NaPO4 pH 6 in H 2 O (80:20, v/v). For internal standardization, 1 μg of D9-Pip, D9-NHP, D5-SA, indol-3-propionic acid (IPA), salicin, ribitol were added. 600 μl of the extract were evaporated to dryness. Free hydroxyl groups of the analytes were converted into their trimethylsilyl derivatives by adding 20 μl of pyridine, 20 μl of N-methyl-N-trimethylsilyltrifluoroacetamide (MSTFA) containing 1% trimethylchlorosilane (v/v) and 60 μl of hexane. The solution was heated to 70˚C for 30 min and after cooling, samples were diluted with hexane, and 2 μl of the solution was separated on a gas chromatograph (GC 7890A; Agilent Technologies) equipped with a Phenomenex ZB-35 (30m x 0.25mm x 0.25μm) capillary column. The following GC temperature settings were used: 70˚C for 2 min, with 10˚C / min to 320˚C, 320˚C for 5 min. Mass spectra were recorded in the electron ionization mode between m/z 50 and m/z 750 with a 5975C mass spectrometric detector (Agilent Technologies). Metabolites were analyzed using the Agilent MSD ChemStation software. For quantification, substance peaks of selected ion chromatograms.

Genome wide association study (GWAS)
Broad sense heritability of total SA contents in both temperature environments was estimated as H 2 = σ 2 accession/ σ 2 accession+ σ 2 residuals based on a linear model using log 10 -transformed total SA data for normalization. Variation in temperature-dependent total SA was expressed using the coefficients of the glm model: Total SA~accession � temperature+replicate fitted with a gamma distribution to account for variation due to temperature as well as between biological replicates. Association mapping was performed on 99 accessions using the GWAPP online application https://gwas.gmi.oeaw.ac.at/ [59]. We chose to further normalize the glm coefficients with a box cox power transformation and used the accelerated mixed model (AMM) to account for population structure [59]. SNPs were considered as significantly-associated with phenotype when they withstood Bonferroni multiple-testing correction.

Quantitative RT-qPCR
RNA was extracted from liquid Nitrogen-frozen plant material using a my-budget Plant RNA kit (Bio Budget technologies Gmbh) according to manufacturer's instructions. cDNA was synthesized from 1μg plant RNA using M-MLV reverse transcriptase (Promega) following the manufacturer's protocol. RT-qPCR was performed with IQ SYBR Green supermix (Bio Rad) on a CFX Connect Real time system (Bio Rad). RT-qPCR primer sequences are listed in S3 Table. Relative expression of test genes was measured against SAND (At2g28390) as a stable reference gene [87]. Relative expression of genes was assessed in three plants per genotype, each from an independent experiment.

Statistical analysis
We performed statistical analyses on at least three samples from three biological replicates using R package version 3.3.1. Comparison between two groups was done using a two-tailed Student t-test or non-parametric Kruskal Wallis test. Multiple comparisons were performed by one-way ANOVA and Tukey's multiple testing correction applied, except for data in Fig  2C-2E where false discovery rate was used. Raw data for Figs 1, 2A-2D, 3A and 5A are provided in S1 Table. For Fig 6 data, an initial 2 way ANOVA was performed. Correlations were assessed using Pearson's product moment correlation coefficient.