Seeds of A. thaliana Columbia ecotype (Col-0), and vte1 (GK_111D07) and vte4 (SALK_036736) mutants were used in this study. vte1 and vte4 mutants have T-DNA insertions in the VTE1 and VTE4 genes (Porfirova et al., 2002; Bergmüller et al., 2003), which encode tocopherol cyclase and γ-tocopherol methyltransferase, respectively, so that the vte1 mutant lacks both α- and γ-tocopherol, and the vte4 mutant lacks α-tocopherol but accumulates γ-tocopherol (Figure 1).

Seeds were cold-stratified and sown in 0.1 L-pots in soil (Einheitserde GS90; Gebrüder Patzer) in a climate-controlled chamber with a 8-h day length provided by fluorescent light at 100 μmol m^-2 s^-1, a day/night temperature of 20/16.8°C and a relative humidity of 60/75% (day/night). Prior to treatments, plants were watered every third day during 23 days with high Pi nutrient solution containing 5 mM Ca(NO₃)₂, 5 mM KNO₃, 2 mM MgSO₄, 1 mM KH2PO₄, and 5 g/L iron chelate (EDTA FeNa). Next, plants were divided into four sets and were subjected to varying phosphate treatments (Figure 2). One set of plants was supplied with the same high Pi nutrient solution (containing 1 mM KH₂PO₄) every other day throughout the experiment (high Pi-plants). Contrasting Pi levels were then supplied to the other three set of plants by using 0.5 mM instead of 1 mM KH₂PO₄ in the nutrient solution (low Pi). This nutrient solution was supplied with 0.5 mM KCl to compensate for K deficiency. Typically, low Pi (0.1 mM) and high Pi (ranging between 0.5 and 2.5 mM) are used in studies on Pi starvation in A. thaliana (Williamson et al., 2001). However, our studies indicated that 0.5 mM Pi is an adequate concentration to induce priming effects ultimately resulting in enhanced growth (Figure 3).

Samplings were performed at midday (in the middle of the photoperiod) at the end of treatments. Whole rosettes of six individuals were used to estimate leaf water contents, chlorophyll levels, the F_v/F_m ratio, and the levels of plastochromanol-8 and vitamin E, as well as the TFs and hormonal profiling. Samples for biochemical and transcriptional analyses were collected, immediately frozen in liquid nitrogen and stored at -80°C until analysis.

Samples were weighed to estimate the fresh matter (FW), immersed in distilled water at 4°C for 24 h to estimate the turgid matter (TW) and then oven-dried at 80°C to constant weight to estimate the dry matter (DW). Relative water content (RWC) was then calculated as 100 × (FW - DW)/(TW - DW). For analyses of macro- and micronutrients (P, S, Ca, Mg, K, Na, Mn, Zn, Fe, Mo, B, Cu, and Si), dried samples were weighed, digested with HNO₃, and analyzed by inductively coupled plasma atomic emission spectroscopy (ICP-AES). For pigment analysis, measurements were performed using a SPAD 502 Plus chlorophyll meter. The maximum efficiency of the photosystem II (F_v/F_m ratio) was determined measuring the chlorophyll fluorescence of leaves by using a pulse-modulated fluorometer (Mini PAM; Walz, Effeltrich, Germany) as described by Genty et al. (1989). The extent of lipid peroxidation was estimated by measuring the levels of malondialdehyde (MDA) in leaves. MDA levels were estimated spectrophotometrically following the thiobarbituric acid-reactive assay considering the effect of potential interfering compounds, as described (Hodges et al., 1999).

For analyses of vitamin E and plastochromanol-8 contents, leaf samples (50 mg) were ground in liquid nitrogen and extracted with cold methanol (v/v) using ultra-sonication. After centrifuging at 8000 rpm for 10 min and 4°C, the supernatant was collected and the pellet re-extracted with the same solvent until it was colorless; then, supernatants were pooled, filtered, and injected into the HPLC. Tocochromanols were separated isocratically on a normal-phase HPLC system using a fluorescent detector as described (Cela et al., 2011). Compounds were identified by co-elution with authentic standards and quantified by using a calibration curve.

For the large-scale TF (a total of 1,880 TFs) profiling using qRT-PCR, total RNA was extracted from whole rosette leaves. Primer sequences are provided in Supplementary Table 10. Total RNA extraction, synthesis of cDNA, and qRT-PCR were performed as described (Caldana et al., 2007). ACTIN2 was used as reference gene. PCR reactions were run on an ABI PRISM 7900HT sequence detection system (Applied Biosystems Applera), and SYBR Green (Life Technologies) was used for visualizing amplified products.

Gene ontology (GO) enrichment analysis for the priming specific TFs was performed using PLAZA 3.0 using default settings (Proost et al., 2015).

Differentially expressed genes were visualized as heatmaps using multiple expression viewer (Mev¹; Saeed et al., 2003). Cluster analysis for differentially expressed genes was performed with Short Time-series Expression Miner (STEM) software using default settings (Ernst and Bar-Joseph, 2006).

For analyses of cytokinins, auxin, gibberellins, ABA, salicylic acid, jasmonates, the ethylene precursor, 1-aminocyclopropane-1-carboxylic acid, and melatonin, leaf samples (50 mg) were ground in liquid nitrogen and extracted with cold methanol:isopropanol:acetic acid (50:49:1, v/v/v) using ultra-sonication. After centrifuging at 8000 rpm for 10 min and 4°C, the supernatant was collected and the pellet re-extracted with the same solvent until it was colorless; then, supernatants were pooled, filtered, and injected into the UHPLC-MS/MS. Phytohormones were separated using an elution gradient on a reverse-phase UHPLC system and quantified using tandem mass spectrometry in multiple reaction monitoring mode as described (Müller and Munné-Bosch, 2011). Recovery rates were calculated for each hormone on every sample by using deuterated compounds.

Data was analyzed by using three-way and one-way factorial analysis of variance (ANOVA), and by additionally using Duncan post hoc tests to analyze for the effects of genotypes at each condition. In all cases, differences were considered significant at a probability level of P < 0.05. All statistical tests were carried out using the SPSS 20.0 statistical package.

In order to understand plant adaptive mechanisms to contrasting Pi availability, plants were grown in a controlled environment initially to attain uniform growth among all the genotypes under study. Later, one set of plants were supplied with the same Pi concentration (hereafter called high Pi), whereas half the concentration (hereafter called low Pi, see “Materials and Methods”) was supplied to plants in three different sets to understand plant response to contrasting Pi availability and priming effects. The first set included plants supplied with low Pi for 2 weeks and then returned to high Pi; the second set included plants grown under high Pi condition and then exposed to low Pi for 2 weeks just prior to samplings. To capture the effects of priming, plants were exposed to low Pi, returned to high Pi condition and then later exposed to second spell of low Pi (Figure 2).

Rosette biomass was not affected by Pi availability, but it was significantly influenced by the genotype and priming. Plant biomass did not differ between genotypes under high Pi, either in unprimed or primed plants. However, rosette biomass was lower in the vte1 mutant compared to wild-type plants at low Pi in unprimed plants, and in both mutants relative to the wild type in primed plants (Figure 3). Wild-type plants were the ones better adapted to low Pi in terms of biomass accumulation, so that priming had positive effects on plant growth. α-Tocopherol deficiency prevented the mutants to benefit from low Pi availability, as indicated by the smaller rosette biomass in both vte mutants compared to wild-type plants under primed conditions at low Pi (Figure 3). Despite these effects on growth, endogenous Pi concentrations did not differ between genotypes at any tested conditions (Figure 4). Furthermore, none of the other nutrients measured revealed any significant genotype-related difference (P < 0.05, ANOVA, Supplementary Figures 1, 2).

Tocochromanols, including α- and γ-tocopherol, as well as plastochromanol-8, were not detected in the vte1 mutant, while both γ-tocopherol and plastochromanol-8 accumulated in the absence of α-tocopherol in the vte4 mutant (Figure 5). In contrast, wild-type plants accumulated α-tocopherol in leaves, particularly at low Pi in primed condition. The contents of this antioxidant doubled at low Pi availability in primed compared to unprimed plants (Figure 5). The contents of γ-tocopherol also increased to a similar extent in the vte4 mutant at low Pi in primed plants only. Plastochromanol-8 contents did not follow the same variations (Figure 5).

Changes in photo-oxidative stress markers, including chlorophyll contents, the maximum efficiency of PSII photochemistry (F_v/F_m ratio), and the extent of lipid peroxidation, estimated as MDA accumulation (Figure 6) paralleled those of rosette biomass (Figure 3), though effects were observed to a much more limited extent, particularly at low Pi in primed plants. Both chlorophyll contents and the F_v/F_m ratio were lower in the vte1 mutant compared to wild-type plants under primed condition at low Pi. However, reductions in the F_v/F_m ratio were very small, the values in all plant genotypes being always above 0.75 (Figure 6). No significant differences in the extent of lipid peroxidation were observed between genotypes, Pi availability or priming (Figure 6).

Transcription factors, the major regulators of gene expression changes play a pivotal role in plant stress responses. To identify such transcriptional regulators that may have possible roles in the regulation of plant response to low Pi conditions and capture the priming effect on plant’s response to contrasting Pi availability, 1,880 Arabidopsis TFs expression was profiled under experimental conditions described above using quantitative real-time PCR (qRT-PCR). Obtained data was analyzed to identify TFs that respond differentially to varying low Pi conditions: (A) TFs responding to low Pi pre-treatment at high Pi, (B) TFs responding to a single low Pi episode, (C) Priming responsive TFs, and (D) TFs responding to recurrent low Pi stimuli (Figure 2).

In wild-type plants, a total of 454 TFs were differentially expressed with a 3-log₂ fold change as cut-off (either up- or down-regulated) in any of the four comparisons performed, representing ∼ 24% of TFs tested in this study (Figure 7A and Supplementary Table 1). Diverse expression patterns of these TFs under different conditions tested suggest a massive transcriptional re-programming involved in plant response to Pi availability. Differentially expressed TFs (DETFs) represented several TF families such as MYB, AP2-EREBP, bZIP, bHLH, AGL, and BBX, among others. These TFs were manually classified based on their expression pattern as either A, B or C specific (Supplementary Table 2). The “A” group contains 69 TFs that were specifically up- (43) or down-regulated (26) in response to low Pi pre-treatment at high Pi, representing ∼15% of total DETFs. Forty-six TFs (∼10% of total DETFs) expressed specifically to a single low Pi episode (B), but only four of them were up-regulated. Interestingly, ∼12% of DETFs (29 up- and 24 down-regulated) were specifically expressed in response to priming treatment (C). These include several TFs functioning in response to plant growth or stress responses. For example, SWI2C, a core component of the SWI/SNF-type chromatin-remodeling complex C (CRCs), was up-regulated in a priming specific manner. SWI2C is a growth regulator and has been identified to interact with DELLA proteins (Sarnowska et al., 2013). Anthocyanin accumulation is a typical phenotypic response under Pi starvation (Morcuende et al., 2007); interestingly MYB111 and MYB113 involved in the regulation of anthocyanin production (Tohge et al., 2013) were repressed specifically in primed plants. Next, GO enrichment analysis was performed for the DETFs that showed priming specific expression using PLAZA 3.0 (Proost et al., 2015). Significantly over-represented GO terms describing the biological process include “gibberellin biosynthesis process,” “ABA-activated signaling pathway,” “regulation of triglyceride catabolic process,” and “histone H3- and H4-acetylation” (Supplementary Figure 3). These GO terms indicate the underlying mechanism of priming effect on plant response to low Pi (Supplementary Figures 4–6).

Additionally, in order to visualize cluster profiles of the DETFs based on their expression direction and magnitude, STEM (STEM) was employed. A comparison of observed groups with those expected in random permutation enables to determine enrichment of the obtained clusters (Ernst et al., 2005; Ernst and Bar-Joseph, 2006). Such comparison of expression patterns resulted in profiles classified into 50 categories in the response of wild-type plants to contrasting Pi availability (Figure 7B). Out of the 50 possible clusters, eight clusters were found to be significant (P-value ≤ 0.05). Colored blocks in (Figure 7B and Supplementary Table 3) display expression profiles of significant clusters (clusters 6, 11, 20, 24, 31, 34, 44, and 45) arranged based on their significance. MYB TF family represented the highest number among the various cluster profiles, followed by AGL and bHLH TF families. Cluster profiles 6, 44, and 31 displayed 73, 67, and 47 TFs, respectively, representing larger cluster groups, among others. To understand the biological significance of the obtained cluster profiles, GO terms enriched in the TF clusters were analyzed. Profile 6 is enriched for TFs associated with GO terms “regulation of response to stimulus,” “response to jasmonic acid, endogenous stimulus, hormone, stress, acid chemical, oxygen containing compound, gibberellin, salicylic acid, and ethylene,” “regulation of signal transduction, jasmonic acid-mediated signaling pathway,” and “chromatin modification and organization,” among others. Profile 44 represents GO terms “regulation of cellular macromolecule biosynthetic process,” “nucleobase-containing compound metabolic process,” “organic substance biosynthesis and metabolic process,” and “root system development,” among others. Profile 31 represents “response to salt and osmotic stress, alcohol, lipid.”

Given the extensive expression changes of TFs under varying Pi availability in wild-type plants (Figure 7A), it was interesting to study how an altered tocopherol composition affects the expression of those transcriptional regulators under contrasting Pi availability. Toward this, we performed whole TF profiling in vte1 and vte4 mutants under the above described conditions (Figure 2). A total of 568 and 583 TFs were differentially expressed (both up- and down-regulated with 3-log₂ fold change as cut-off threshold) in vte1 and vte4 mutants compared to the wild type (Figure 8 and Supplementary Tables 4, 5) representing ∼30 and ∼31% of the total TFs analyzed, respectively. DETFs in the vte1 mutant are represented by a mixture of TF families, while TFs belonging to MYB, AGL, bZIP, bHLH TF families were abundant in the vte4 mutant compared to the wild type. To identify TFs that exhibit condition specific expression patterns, manual classification was performed (Supplementary Tables 6, 7) as described above (see also Figure 2). In the vte1 mutant, 49 TFs were specific to group A of which, 34 and 15 were up- and down-regulated, respectively, representing ∼9% of the total DETFs. vte4 mutant displayed ∼8% TFs specific to group A with 14 and 29 up- and down-regulated, respectively. Contrasting to the wild type, the vte1 mutant displayed more TFs specific to group B with 121 up-regulated and only 1 down-regulated (∼22% of DETFs). In the vte4 mutant, around 8% (35 up- and 8 down-regulated) of total DETFs responded specifically to group B. The TFs involved in the regulation of anthocyanin were up-regulated in both the mutants compared to wild-type plants, suggesting that the mutants were experiencing stress under low Pi compared to the wild type. For example, in the vte1 mutant, MYB112 was induced in group A and MYB114 up-regulated in both A and B, both known to be involved in the regulation of anthocyanin production (Tohge et al., 2013; Lotkowska et al., 2015). Furthermore, ∼8% TFs displayed a priming specific (group C) expression pattern, of which 28 and 16 were up- and down-regulated in the vte1 mutant compared to the wild type. Interestingly, several of these TFs showed an opposite priming specific expression pattern in wild-type plants. Priming repressed TFs in the wild type, MYB111 and MYB113 expression was induced in vte1 primed plants, which suggests that these plants could not benefit from priming.

Varying number of TFs representing group specific expression pattern among vte1 and vte4 mutants was observed which might suggest diverse roles for the different tocopherol forms. As contrasting to vte1, vte4 mutants displayed ∼24% (56 up- and 81 down-regulated) TFs with priming specific expression pattern compared to the wild type. MYB111 expression was also increased in primed plants. Expression of ZAT6, a negative regulator of Pi homeostasis (Devaiah et al., 2007b) was up-regulated in the vte4 mutant in a priming specific manner. Previously, expression of several ethylene signaling pathway genes such as EIN2, EIN3, CTR1, and ERF1 has been reported to be up-regulated in the vte4 mutant upon salt stress compared to wild-type plants (Cela et al., 2011). In the present study, expression of several ERFs was induced, while EIN and EIL3 expression was repressed in primed vte4 plants. GO enrichment analysis of the priming specifically expressed TFs revealed interesting over-represented GO terms such as “ethylene mediated signaling pathway,” “cytokinin activated signaling pathway,” “TF import to nucleus,” and “histone H3 K27 methylation” in vte1 mutant; “salicylic acid-mediated signaling pathway,” “sucrose-induced translational repression,” and “regulation of pectin biosynthesis and metabolism” in vte4 mutant. Activation of cytokinin signaling pathway in vte1 mutant indicate attenuated low Pi response in these plants, as cytokinins negatively regulate Pi starvation responses (Yang and Finnegan, 2010). Defense hormone SA-mediated signaling and pectin biosynthesis in the vte4 mutant suggests activated defense response networks in vte mutants in response to priming. Further, positive effect of priming on the growth observed in the wild type was compromised in vte mutants possibly due to the tradeoff toward defense.

Next, to understand those expression patterns observed among different Pi availability conditions in both the mutants, first we clustered the DETFs into profiles using STEM. A total of 50 possible cluster profiles were obtained for both mutants, of which, 7 (5, 6, 24, 27, 44, 45, and 47) and 9 (5, 6, 12, 14, 23, 24, 27, 28, and 44) clusters were found to be significant in vte1 and vte4 mutants, respectively (Figures 9A,B and Supplementary Tables 8, 9). Further, these significant clusters were compared with the significant clusters obtained from wild-type plants using STEM, where the TFs in each cluster from the mutants were compared with their magnitude of expression and direction in the wild type. Interestingly, such correlation revealed opposite patterns for many of the cluster profiles between the mutants and the wild type (Figures 9A,B). To learn about the involvement of tocopherols in mediating the plant response to contrasting Pi availability, these oppositely expressed TFs could serve as a valuable resource. Additionally, learning about the GO terms associated with these TFs would help to expand our knowledge in understanding the adaptive mechanisms associated with vitamin E under varying Pi availability. Interestingly, the largest cluster profiles 6 and 44 of wild-type plants showed good correlation with the cluster profiles of the vte1 mutant. Wild type profile 6 and 44 in the vte1 mutant displayed TFs representing similar GO terms such as “response to: gibberellin, hormone, endogenous stimulus, organic substance, ethylene, lipid, jasmonic acid, salicylic acid, auxin, ABA, organic cyclic compound,” “regulation of: signaling, signal transduction, cell communication,” and “negative regulation of: cellular, macromolecule biosynthesis process, nucleobase-containing compound metabolic process” among others. Wild type profile 44 correlated with vte1 profile 6 representing the GO terms “heterocyclic compound-, organic cyclic compound-, nucleic acid-binding,” and “protein dimerization activity.” Profile correlations between wild type and the vte4 mutant as well showed significant GO enrichment terms. Wild type profile 6 and vte4 profile 44 correlated TFs represented GO terms “response to gibberellin, stress, abiotic stress, osmotic stress, lipid, salicylates, ethylene,” “chromatin modification, organization,” and “histone modification,” among others. Wild type profile 23 and vte4 profile 24 displayed GO terms “cellular response to: ethylene stimulus, hormone stimulus, organic substance, endogenous and chemical stimulus,” “ethylene activated signaling pathway,”, “phosphorelay signal transduction system,” and “hormone-mediated signaling pathway”.

Hormonal profiling revealed genotype-related differences, particularly for salicylic acid contents, which increased significantly in the vte1 mutant compared to wild-type plants at low Pi, but in unprimed plants only. This effect was not observed in primed plants, because in this case salicylic acid contents increased similarly in the three genotypes at low Pi availability (Figure 10). Furthermore, enhanced jasmonic acid-isoleucine (JA-Ile) contents were observed in the vte4 mutant compared to wild-type plants and the vte1 mutant in primed plants at low Pi (Figure 10). No genotype-related differences were observed in the contents of cytokinins (Supplementary Figure 3), auxin (Supplementary Figure 7), gibberellins (Supplementary Figure 8), ABA, the ethylene precursor, 1-amino-cyclopropane-1-carboxylic acid, or melatonin (Supplementary Figure 9). Priming had significant effects on jasmonates, salicylates, and auxin contents; defense-related compounds such as jasmonic acid in particular, increasing, and indole-3-acetic acid contents decreasing, in primed plants (Figure 10 and Supplementary Figure 7). Low Pi availability increased the contents of oxo-phytodienoic acid and JA-Ile irrespective of priming, the latter particularly increasing in the vte4 mutant in primed plants (Figure 10).