Ambient temperature and genotype differentially affect developmental and phenotypic plasticity in Arabidopsis thaliana

Background Global increase in ambient temperatures constitute a significant challenge to wild and cultivated plant species. Forward genetic analyses of individual temperature-responsive traits have resulted in the identification of several signaling and response components. However, a comprehensive knowledge about temperature sensitivity of different developmental stages and the contribution of natural variation is still scarce and fragmented at best. Results Here, we systematically analyze thermomorphogenesis throughout a complete life cycle in ten natural Arabidopsis thaliana accessions grown under long day conditions in four different temperatures ranging from 16 to 28 °C. We used Q10, GxE, phenotypic divergence and correlation analyses to assess temperature sensitivity and genotype effects of more than 30 morphometric and developmental traits representing five phenotype classes. We found that genotype and temperature differentially affected plant growth and development with variing strengths. Furthermore, overall correlations among phenotypic temperature responses was relatively low which seems to be caused by differential capacities for temperature adaptations of individual accessions. Conclusion Genotype-specific temperature responses may be attractive targets for future forward genetic approaches and accession-specific thermomorphogenesis maps may aid the assessment of functional relevance of known and novel regulatory components. Electronic supplementary material The online version of this article (doi:10.1186/s12870-017-1068-5) contains supplementary material, which is available to authorized users.

The investigation of signaling pathways that translate temperature stimuli into qualitative and quantitative developmental responses has so far largely been limited to either seedling development or flowering time. However, it seems likely that temperature responses in different phases of development either require variations of a canonical signaling pathway or involve at least partially specific signaling components. To enable the dissection of thermomorphogenic signaling at different developmental stages, it is vital to gather a comprehensive understanding of the diversity of temperature reactions throughout plant development.
According to basic principles of thermodynamics, temperature-induced changes in free energy will affect the rates of biochemical reactions. As these effects should occur generally, albeit to different magnitudes, non-selective phenotypic responses can be expected to occur robustly and rather independently of genetic variation.
Such traits may therefore be indicative of passive, thermodynamic effects on a multitude of processes. Alternatively, robust temperature responses may be due to thermodynamic effects on highly conserved signaling elements. These may be attractive targets for classic mutagenesis screens to identify the relevant regulatory components. In contrast, natural variation in thermomorphogenesis traits is likely the consequence of variability in one or several specific signaling or response components. It may be addressed by quantitative genetic approaches to identify regulators that contribute to variable temperature responses. Such genes may represent attractive candidates for targeted breeding approaches.
In this study we aim to (i) provide a map of developmental phenotypes that are sensitive to ambient temperature effects throughout a life cycle in the model organism A. thaliana, (ii) identify traits that are robustly affected by temperature with little variation among different accessions, and ask (iii) which traits are affected differentially by different genotypes and thus show natural variation in temperature responses.
To realize this, we performed a profiling of numerous developmental and morphological traits which can be sorted into five main categories: juvenile vegetative stage, adult vegetative stage, reproductive stage, morphometric parameters and yield-associated traits. Phenotypes were analyzed in a subset of ten A. thaliana accessions which were grown at 16, 20, 24, and 28 °C in climate-controlled environments. Knowing that even a small randomly selected set of A. thaliana accessions covers a wide spectrum of genetic diversity [26], we chose to analyze commonly used lab accessions such as Col-0, Ler-1 and Ws-2, accessions known to react hypersensitively to elevated temperature (e.g., Rrs-7, [24,27], and parental lines of available mapping populations such as Bay-0, Sha, and Cvi-0.
In addition to a meta-analysis of the phenotypic data, we provide accession-specific developmental reference maps of temperature responses that can serve as resources for future experimental approaches in the analysis of ambient temperature responses in A. thaliana.

Plant material and growth conditions
Phenotypic parameters ( Fig. 1) were assessed in A. thaliana accessions that were obtained from the Nottingham Arabidopsis Stock Centre [28]. Morphological markers and time points of analyses are described in Additional file 1. Detailed information on stock numbers and geographic origin of Arabidopsis accessions are listed in Additional file 2. For seedling stage analyses, surface-sterilized seeds were stratified for 3 days in deionized water at 4 °C and subsequently placed on A. thaliana solution (ATS) nutrient medium [29]. Seeds were germinated and cultivated in climatecontrolled growth cabinets (Percival, AR66-L2) at constant temperatures of 16,20,24 or 28 °C under long day photoperiods (16h light/8h dark) and a photosynthetically active fluence rate (PAR) of 90 μmol·m -2 ·sec -1 of cool white fluorescent lamps. We refrained from including a vernalization step because the primary focus of this study was to record morphology and development in response to different constant ambient temperature conditions. Germination rates were assessed daily and hypocotyl, root length, and petiole angles were measured in 7 days old seedlings (n > 15) with ImageJ [30] and Root Detection All other analyses were performed on soil-grown plants cultivated in growth cabinets (Percival) at a PAR of 140 μmol·m -2 ·sec -1 and long day photoperiods (16h light/8h dark). After imbibition for 3 days at 4 °C, seeds were grown in individual 5 x 5 cm pots, which were randomized twice a week to minimize position effects. Relative humidity of growth cabinets was maintained at 70 % and plants were watered by subirrigation. Plants (n > 15) were photographed daily for subsequent determination of phenotypic parameters (leaf number, rosette area and petiole length) using Image J (http://imagej.nih.gov/ij/). Determination of developmental progression largely followed the stages defined in Boyes et al. [32]. The vegetative growth period was divided in a juvenile phase (germination to initiation of the fifth rosette leave) and an adult vegetative stage (initiation of the sixth rosette leave to floral transition). At transition to the reproductive growth phase, the number of leaves was determined by manual counting in addition to recording the number of days after germination.
Spectrophotometric determination of chlorophyll content was performed as described in [33].

Data analysis
Visualization and statistical analyses of the data were performed using the software R [34]. Box plots were generated using the boxplot function contained in the graphics package. Heat maps were generated using the heatmap.2 function contained in the gplots package.
ANOVAs for a single factor (either accession or temperature) and Tukey's 'Honest Significant Difference' test as post hoc test were performed using the anova and TukeyHSD function, respectively, which are both contained in the R stats package. Variation in phenotype expression was analyzed by 2-way ANOVA according to Nicotra [35] and Whitman and Agrawal [36] to test each phenotype for a significant effect of genotype (G, accession) or environment (E, temperature), and a significant genotype by environment interaction (GxE). Reaction norms for each analysis are shown in Additional file 3.

Q 10 temperature coefficient
The Q 10 temperature coefficient was calculated according to Loveys [37] where P w and P c are the trait values at the warmer and cooler temperatures, respectively. T w and T c represent the corresponding temperatures in °C. We computed the geometric mean of the six Q 10 values of all pairwise temperature combinations for each phenotypic trait to avoid artifacts caused by differential reaction norms/response shapes.

Index of phenotypic divergence (P st )
Calculation of the index of phenotypic divergence (P st [38,39]) as a measure to quantify variation in each phenotypic trait was calculated as previously described by Storz [38] as where σ b 2 is the variance between populations, and σ w 2 is the variance within populations. The ANOVA framework was used to partition the variances to get and σ w 2 .
Using the two factorial design, two types of indices of phenotypic variation of a trait/phenotype were considered separately.

Results
To assess phenotypic plasticity in a range of ambient temperatures, A. thaliana plants  As reported previously, Col-0 plants showed a decrease in developmental time until flowering with increasing ambient temperatures [11]. The transition from the vegetative to the reproductive phase at 28 °C occurred about 25 days earlier than at 16 °C (Fig. 2a). Similarly, the number of rosette leaves developed at time of bolting differed by approximately 26 leaves between 28 °C and 16 °C (Additional file 4b).
The observation that only a very limited number of phenotypes were insensitive to cultivation in different temperatures clearly illustrates the fundamental impact of ambient temperature on plant growth and development.

Natural variation of temperature responses
To assess whether the observed temperature responses in Col-0 are robust among A. thaliana accessions or which of the responses may be affected by natural variation, phenotypic profiling was performed in nine additional A. thaliana accessions parallel to the analysis in Col-0 (Additional files 4-13). Naturally, a panel of ten accessions does not comprehensively represent the world-wide gene pool of A. thaliana. However, it can be expected that even 10 randomly chosen natural accessions represent ~70 % of the allelic diversity in the A. thaliana gene pool [26].
Hence, the general assessment of thermo-responsive development in A. thaliana as well as the identification and discrimination between traits that generally seem to exhibit natural variation and those that may be genetically fixed within the gene pool is a realistic aim even with a set of 10 selected accessions.
To approximate and to compare temperature sensitivity of traits among different accessions, we calculated Q 10 values for each individual trait and phenotype class for each analyzed genotype [37]. The Q 10 quotient represents the factor by which a trait significant factor because cold treatment is explicitly recommended to induce earlier flowering for several Got-7 lines available at NASC/ABRC [40]. Natural variation in regulators such as FLM may contribute to this phenotype. However, as all accessions were able to flower at temperatures of 16 and 20 °C vernalization does not seem to be an essential requirement.
Taken together, juvenile and adult vegetative development remained highly conserved, whereas the reproductive stage and yield-associated traits showed higher variation between accessions and within individual accession, as indicated by the ranges/dimensions of the box plots in Fig. 3a. Here, high variation within a phenotype class indicates that temperature effects on individual traits within that class are highly variable. The strongest variation within accessions was observed for morphometric phenotypes such as hypocotyl and petiole length. In contrast, a high variation between accessions is indicative for differential responses of different genotypes which was most prominent in reproductive stage traits.  (Fig. 3a, b).
To assess genotype and temperature contributions in a more quantitative manner, we next used a variance partitioning approach [38,39,41,42]. Specifically, we calculated the index of phenotypic divergence (P st , [38]) at each analyzed temperature as a measure of genotype effects P st gen on the trait of interest (Additional file 16a). To complement this analysis, we also estimated the variation occurring across temperatures P st temp for each of the analyzed accessions (Additional file 16b), which enabled us to assess the temperature effect for the trait of interest for specific genotypes.

Genotype effects
The were generally very low (Fig. 4a), corroborating the impression gained from the analysis of Q 10 values (Fig. 3). However, genotype effects on later stages of adult vegetative development seem to increase with higher temperatures (Additional file 16a), which may be the significant effect observed in the ANOVA-based GxE interaction assessment.
Similarly, strong genotype effects at higher temperatures were also observed for reproductive traits. Here, P st gen values at 16°C were already considerably higher than for vegetative growth stages and increased further with elevated temperatures (Additional file 16a). A contrasting pattern of decreasing genotype effects with an increase in temperatures was observed for total plant height indicating that here, natural variation in growth is higher at lower temperatures. Yield-associated phenotypes in general showed only low genotype effects on variation, indicating that under our experimental conditions variation in trait expression in this category is primarily affected by temperature (Fig. 4a).
Other phenotypes display rather differential or less gradual genotype effects among different temperatures. For example, the genotype impact on variation in hypocotyl and petiole length sharply increases from 24 to 28°C, indicating a certain buffering capacity or a threshold for natural variation.
In some cases, such as flowering time, a strong genotype effect seems to correlate also with a strong general temperature sensitivity as indicated by the high betweenaccessions variability in Q 10 values (Fig. 4a and Fig. 3b). However, this does not seem to be a general principle. In case of root length, for example, low genotype effects were observed (Fig. 4a, b), even though the phenotype in principle was highly sensitive to a change in ambient temperature (Fig. 3b).

Temperature effects
We also used the variance partitioning approach to analyze the extent of the significant impact of temperature on phenotypic variation that was detected in the GxE interaction analysis (Additional file 15). Therefore, we calculated the index for chlorophyll content in Ler-1 vs. Bay-0). This is particularly obvious for yield-related traits such as total number of seeds per plant and silique as well as silique length.
Here, temperature effects on phenotype variation were low for Col-0, C24 and Bay-0, whereas considerably higher P st temp values were determined for the other accessions (Additional file 16b). Accessions which exhibit strong temperature effects on phenotypic variation may be interesting candidates for forward genetic approaches to identify the contributing molecular regulatory components.

Comparison of temperature and genotype effects
As each phenotypic trait has been assigned a value for genotype and temperature effects, they can easily be compared to assess which of the two has a stronger influence on the phenotypic plasticity. To allow a direct comparison of effects, we compared mean values for P st gen across all temperatures and P st temp across all accessions (Fig. 4a, b).  (Fig. 4b). While vegetative and reproductive phenotypes form tight clusters, morphometric phenotypes displayed a heterogenous pattern. In these traits, temperature responses seem to be affected by natural variation and may thus serve as candidate phenotypes for classic or quantitative forward genetic analyses.
Several yield-associated phenotypes such as total number of seeds, seed size and seed weight showed varying degrees of temperature sensitivity, likely caused by the partially distinct temperature effects on individual accessions (Fig.2b, Additional file 17).
The fundamental impact on temperature on the phenotypic responses is also reflected in the results of the principle component analysis (PCA). The PCA was performed on mean-centered and scaled data in order to allow integration of data with different scaling. PC1 which covered 50% of the observed variation, allowed a clear separation of samples via temperature (Fig. 5a). Here, the differentiation between 16 and 20 °C seems to be higher than the temperature changes from 20 to 24 °C and 24 to 28°C. PC2 explained ~16% of the variation and separated samples rather by genotype. Here, Rrs-7 and Got-7 showed a clear divergence from other genotypes. Again, this separation is already clear between 16 and 20°C whereas a further increase in temperature contributed little more to the separation. were retained in the analysis (Fig. 5b).

Correlation of phenotypic temperature responses
High correlations were detected among traits within the vegetative stage of development (e.g. juvenile and adult vegetative stage), and among traits within the reproductive phase (e.g. flowering traits and the onset of silique production). In addition, temperature-induced reduction in foliar surface correlated strongly with the decrease in developmental time in vegetative and reproductive phases. Similarly, the reduction in developmental times and foliar surface were moderately correlated to the effect on several seed-associated traits (Fig. 5b).
Model temperature phenotypes such as petiole and hypocotyl length showed a positive correlation and were in turn correlated or inversely correlated with several other phenotypes or trait groups. However, temperature responses in primary root length under these experimental conditions showed an even more robust connection to many other traits. Mostly, these were inverse correlations with the exception of other seedling traits which were positively correlated with primary root lengths (Fig.   5b).
Due to the differential genotype effects on variation we also wondered whether On the other hand, phenotypes that show a high degree of genotype and temperature effects might rather be influenced by one or more specific genes that contribute to trait expression in a quantitative manner. As such, these phenotypes would represent "active" temperature effects [45]. However, the involvement of specific signaling elements does not necessarily exclude influences via thermodynamics. In fact, the recently described thermosensing via phyB acts via the promotion of phyB P FR to P R conversion in a temperature-promoted manner [18,19].
Natural variation in thermomorphogenic responses could be caused by polymorphisms in signaling or response genes ranging from alteration in gene sequence to expression level polymorphism [46]. As they may provide keys to altered temperature responses that could be utilized in specific breeding approaches, identification of such genes would be of high interest.
In fact, natural allelic variation in the circadian clock components ELF3 and in the regulation of GIGANTEA have recently been shown to directly affect PIF4-mediated hypocotyl elongation in response to elevated temperatures [12,13,47]. Therefore, PIF4 and PIF4-regulating components could be important targets of adaptation to  [12,13,20,48]. However, a lack of general correlation among seedling growth and flowering time responses may indicate that these processes are not universally regulated via the same components. Alternatively, the impact of these signaling components on diverse phenotypes may be more prominent for specific alleles which may be reflected by the diversity in correlation patterns among individual accessions (Fig. 5c, Additional file 18).
In general, the intraspecific diversity in phenotypic changes in response to elevated ambient temperatures argue against a general explanation of morphological and developmental changes due to passive thermodynamic effects.
Exploiting natural genetic variation to identify genes that are involved in the regulation of temperature effects on specific traits can provide new leads for plant breeding. The work presented here may inspire new approaches for temperature research in non-reference accessions as some temperature responses were much more pronounced in accessions other than Col-0 (Fig. 3b). Specific approaches will depend on the focus on either yield-or biomass-associated traits.
In conclusion, our work provides a map that allows the dissection of thermomorphogenesis in phenotypic traits that are either robustly affected by temperature or traits that are differentially affected by temperature among different accessions. While robust temperature-sensitive phenotypes might indeed be caused by thermodynamic acceleration of metabolism or highly conserved signaling events, natural genetic variation of temperature responses implicate the relevance of specific regulatory cascades that can be instrumental to future breeding approaches. The datasets analysed during the current study is available from the corresponding author on request.

 Competing interests
The authors declare that they have no competing interests         Table S1.
A heatmap of individual P st gen and P st temp values and a scatter plot including standard