To characterize light-dependent proline accumulation in Arabidopsis, an in vitro experimental system was designed: Fourteen-day-old plantlets were treated with high intensity light (550 µE m⁻² s⁻¹) or deprived of light for up to 5 days, and subsequently salt or ABA-triggered proline accumulation was monitored periodically ( Figure 1A ). When plants were irradiated with strong light for several days, proline levels accumulated up to three times higher as compared to plants kept under standard light conditions ( Figure 1B ). Proline accumulation was compromised in the p5cs1-1 mutant (Székely et al., 2008), suggesting that the P5CS1 gene controls the rate-limiting step in proline accumulation in these conditions. In standard light conditions 1-day 10 µM ABA and 150 mM NaCl treatments lead to two or five times higher proline contents, respectively. When plants were deprived of light, proline accumulation was considerably smaller: 1 day of dark adaptation reduced proline levels from 40% to 60% of light cultured plants, whereas 5 days in darkness prevented the enhancement of proline content ( Figure 1C ). The negative effect of dark on proline accumulation may derive from the lack of adequate light or photoreceptor-derived signals reducing proline biosynthesis and/or inducing catabolism. Alternatively, if energy shortage prevents proline accumulation in dark, then externally added sugar should compensate for the absence of light. To test energy dependence, proline concentrations were measured in dark-adapted plants in the presence of various concentrations of sucrose. Proline contents in dark-adapted plants were similar in the presence of 0% and 2% (W/V) sucrose in the culture medium, whereas 4% (W/V) sucrose significantly enhanced proline accumulation. Proline levels in these conditions were, however, still far inferior to those in illuminated plants, which accumulated five to ten times more proline ( Figure 1D ). When proline content was compared in plants treated with sucrose, glucose, or mannitol, only minor differences were observed ( Figure S2 ). These results suggest that sugar-dependent glycolysis cannot compensate for the lack of light signals or other photosynthesis-derived metabolites such as NADPH.

Light generates specific signals, which can trigger biosynthetic and/or repress catabolic pathways. Light signals are perceived by specific photoreceptors, each of them possessing well-defined sensitivity to a particular spectrum of light (Franklin and Quail, 2010; Chaves et al., 2011). To test the effect of light quality on proline metabolism, dark-conditioned plants were transferred to white or monochromatic red, far-red or blue lights with or without simultaneous salt stress ( Figure 2A ). The light intensities for each light qualities (white light: fluorescent cool white, 4200 K, 60 µE m⁻² s⁻¹, monochromatic blue: 470 nm, 15 µE m⁻² s⁻¹, red: 660 nm, 15 µE m⁻² s⁻¹, or far-red: 730 nm, 5 µE m⁻² s⁻¹ light), were sufficient to saturate photoreceptors and light signaling cascades but not the photosynthetic electron transport (Wolf et al., 2011; Adam et al., 2013). In the absence of salt stress, proline levels increased under white, red and blue light, but remained unchanged in darkness or under far-red light ( Figure 2B ). Proline concentrations increased more than 10-fold in salt-treated plants under white or red light, whereas under blue light 5-fold enhancement was measured. Salt stress could only slightly augment free proline content in plants kept in dark or illuminated by monochromatic far-red light ( Figure 2B ). The effect of white and monochromatic light on proline accumulation was similar in Columbia 0 (Col-0) and Wassilewskija (WS) ecotypes ( Figure S3 ). These results suggest that to promote proline accumulation, red is the most efficient component of the light spectrum followed by blue, whereas far-red light is insufficient.

To investigate the molecular background of light-dependent proline accumulation, expression patterns of the key metabolic genes P5CS1 and PDH1 were monitored under different light regimes with or without salt treatment. Transcript levels of P5CS1 were significantly higher in plants illuminated with white, red, and blue light than in plants kept in dark or illuminated by far-red light. Salt treatment enhanced P5CS1 expression in all light conditions, and transcript levels were highest under white and red lights followed by blue, but were moderately enhanced in far-red light or in darkness ( Figure 2C , Figure S4 ). PDH1 expression was downregulated by white, red, and blue light and not affected significantly under far-red light. Salt treatment repressed PDH1 expression even more in most light conditions, including in darkness ( Figure 2C , Figure S4 ). These results suggest that besides white light, red, and blue monochromatic lights are efficient in promoting P5CS1 and suppressing PDH1 expression, which ultimately leads to high levels of proline accumulation when plants are exposed to salt stress.

In order to investigate whether photosynthetic electron transport is required for proline accumulation, leaves were treated with 3-(3′,4′-dichlorophenyl)-1,1-dimethylurea (DCMU) in combination with salt ( Figure S5 ). DCMU binds irreversibly to the acceptor side of photosystem II thereby inhibiting linear electron transport, which results in altered fast chlorophyll a fluorescence (OJIP) kinetics. Upon full inhibition, the J (F_2ms) step equals the maximum fluorescence (F_M or P) intensity (Tóth et al., 2007), as seen also in Figure S5C . The DCMU treatment alone had no effect on proline levels, whereas NaCl-induced proline accumulation was significantly reduced ( Figure S5A ). In the presence of DCMU P5CS1 induction was slightly reduced in salt-treated plants, whereas PDH1 was upregulated in both salt-treated and control plants ( Figure S5B ).

The bZIP-type transcription factor HY5 is a key positive regulator of light-dependent gene expression that controls transcription of thousands of light-induced genes (Cluis et al., 2004; Toledo-Ortiz et al., 2014). A recent ChIP-seq analysis of HY5 binding sites revealed that this transcription factor recognizes conserved cis-acting elements in more than three thousand genes, both under red and blue light (Hajdu et al., 2018). Genome-wide mapping of HY5 binding sites revealed that this TF can recognize the promoter regions of the P5CS1 and PDH1 genes ( Figures S6 , S7 ). While peak of the reads were mapped close to the transcription initiation site of P5CS1, maximum reads were localized to 0.5 kb upstream of the PDH1 transcription initiation site. Another recent study revealed binding of HY5 to the 5′ UTR and a distal upstream region of P5CS1 (Feng et al., 2016). Promoter and 5′ UTR regions of P5CS1 contain a number of predicted regulatory sequence motifs, including a G-box (CACGTG) at +172 bp in the 5′ UTR and a C-box (GACGTC) in the promoter, at −59 bp distance from the transcription start site, which can serve as binding sites of HY5 ( Figure 3A ) (Fichman et al., 2015). The PDH1 promoter contains one conserved C-box motif in the promoter, at −553 bp distance from the transcription start site ( Figure 3A ).

To verify HY5 binding in the identified regulatory regions, ChIP followed by quantitative PCR (ChIP-qPCR) assays were performed on P5CS1 and PDH1 promoter fragments containing the predicted G-box and C-box sequence motifs. When compared to intergenic regions, specific enrichment in qPCR-amplified C-box- and G-box-containing P5CS1 and PDH1 promoter fragments was detected in the HY5-YFP-immunoprecipitated DNA samples ( Figure 3B ). Enrichment of HY5 binding to P5CS1 G-box and C-box regions was around 14 to 16 times higher while it was 5 times higher on C-box region of the PDH1 promoter than on a control intergenic region ( Figure 3B ). ChIP-qPCR experiments therefore confirmed that HY5 interacts in vivo with the selected 5′ UTR and promoter regions of both the P5CS1 and the PDH1 genes. To compare our results with previously reported HY5 assays, ChIP assay was performed with primers used to amplify Region 2 and Region 5 of P5CS1, as defined by Feng et al. (2016). Region 2 is a 140 bp fragment, in the 5′ UTR (from −5 to +135 bp), flanked by C-box and G-box sequences. Region 5 is a 147 bp fragment in the upstream region (from −2129 to −2276 bp), which corresponds the previously described “Essential for Memory Fragment” (EMF) (Feng et al., 2016). In our experimental conditions enrichment of Region 2 in ChIP assay was similar to fragments containing G-box and C-box elements, whereas enrichment was a magnitude lower when HY5 binding was tested for Region 5, corresponding to EMF ( Figure 3B ).

To verify that the conserved C-box and G-box motifs are indeed the targets of HY5, electrophoretic mobility shift assays (EMSA) were performed using 56 bp dsDNA fragments containing the native regulatory sequences or their mutated forms in which the conserved CACGTG or GACGTC sequence motifs were altered, eliminating the core ACGT sequence ( Figure 4A ). Complex formation of HY5 with ds oligonucleotides corresponding to wild-type P5CS1 and PDH1 promoter fragments was observed in the EMSA assays. Complexes between HY5 and oligoes carrying the mutated G-box or C-box sequences were, however, not formed or were detected at much lower level ( Figure 4B ). These experiments confirmed that the C-box and G-box sequences are indeed responsible for HY5 binding to the P5CS1 promoter (−59 bp) or the 5′ UTR (+172 bp) regions as well as binding of the PDH1 promoter (−553 bp) region.

To study the function of HY5 in proline metabolism, free proline contents and transcript levels of P5CS1 and PDH1 genes were compared in wild-type (WS) and hy5hyh mutant plants carrying knockout mutations for both HY5 and the closely related HYH genes (Hajdu et al., 2018). Wild-type and hy5hyh double mutant plants were conditioned to dark as described above, and were subsequently treated by salt under white and monochromatic red or blue light ( Figure 5A ). Compared to plants kept in darkness, proline levels were enhanced by illumination with both white and red or blue monochromatic lights. When compared to wild type, proline levels were not affected or were slightly lower in illuminated hy5hyh mutants without salt treatment ( Figures 5B , S8 ). Salt stress enhanced proline contents three to six times in illuminated plants, while proline accumulation was around 50% lower in the hy5hyh mutant when compared to wild-type plants in the same conditions ( Figures 5B , S8 ). When plants were kept in darkness, proline levels were only slightly increased by salt treatment, and enhancement was similar in both genotypes.

P5CS1 transcript levels were low in both wild-type and hy5hyh mutants without salt treatment with minor induction by illumination. P5CS1 expression was clearly induced by 6 h of salt treatment in illuminated plants, reaching approximately 50% lower transcripts in the hy5hyh mutant than in wild-type plants ( Figures 5C , S9B ). In all light conditions transcript levels were higher after 6 h of stress than after 24 h ( Figure 5C ). PDH1 expression was reduced by illumination and by salt stress in all light conditions. Genotype-dependent differences in PDH1 transcript levels were however less pronounced and downregulation was more variable ( Figures 5D , S9C ). The P5CS1 transcription pattern positively correlated with changes in proline levels, indicating that biosynthesis is essential in defining proline accumulation in these conditions, while PDH1-controlled catabolism can fine-tune proline levels. These data indicate that HY5 (and possibly HYH) is a positive regulator of proline accumulation by contributing to the expression of the P5CS1 gene with a minor role in the control of PDH1 expression.

Arabidopsis thaliana plants were either Col-0 or WS ecotype. The hy5hyh double mutant (Holm et al., 2002) has WS background. Basic conditions of plant growth were described before (Aleksza et al., 2017). Briefly: seeds were surface sterilized and germinated on 1/2MS culture medium containing 0.5% (W/V) sucrose. Plants were grown in vitro in growth chambers under 120 µE m⁻² s⁻¹ illumination (white light) using a 8 h light/16 h dark cycle, and 22°C/18°C temperature cycle for 14 days.

For high-intensity light treatment 14-day-old plants were illuminated with white light with 550 µE m⁻² s⁻¹ light intensity in growth chambers. For dark conditioning, 14-day-old plantlets were transferred to dark, and incubated in the absence of light for up to 5 days in the same conditions (medium, temperature). For subsequent light induction, plants were transferred to either white light (fluorescent cool white, 4200 K, 60 µE m⁻² s⁻¹), or monochromatic blue (470 nm, 15 µE m⁻² s⁻¹), red (660 nm, 15 µE m⁻² s⁻¹), or far-red (730 nm, 5 µE m⁻² s⁻¹) light, or kept in dark for up to three further days. The primary criteria for setting the fluence rate of light was to reach saturation of signaling cascades triggered by phytochrome and cryptochrome photoreceptors. The most studied light responses are saturated at the light intensities described above (Wolf et al., 2011; Adam et al., 2013). Dark-conditioned plants (including those kept in constant darkness) were transferred and handled under green light.

To induce proline accumulation, plants were cultured on the surface of thin-layer liquid culture medium (10 ml medium/13 cm diameter Petri dish), using nylon mesh to prevent submergence. For stress, liquid media were supplemented with 150 mM NaCl or treated with 10 µM ABA. Proline levels were determined in plants for up to 3 days as described.

Proline content was determined by the ninhydrin-based coloritmetric method as described (Abraham et al., 2010). Alternatively, a microtiter-scale colorimetric reaction was used, which was based on a recent paper (Lee et al., 2018) with some modifications. Plant material (approximately 50 mg fresh weight/sample) was ground and 20 µl of 1% (W/V) sulfosalicylic acid was added per mg FW tissue. After centrifugation at top speed (15.000 rpm) for 5 min at 4°C in a microcentrifuge, the supernatant was removed and mixed with acidic ninhydrin [1,25% (W/V) ninhydrin in 80% (V/V) acetic acid] in 1:2 ratio, and incubated at 95°C for 30 min. The reaction was terminated on ice, and absorbance was measured at 510 nm in a plate reader (MULTISKAN GO, Thermo Scientific) using a 1:2 mixture of sulfosalicylic acid and acidic ninhydrin as reference. The system was calibrated with standard curves with known concentrations of proline. Anthocyanine accumulation was not visible in the plants after these treatments. Experiments were repeated three times and four to six replicates were used to determine proline levels in a treatment.

Fluorescence measurements were carried out at room temperature with a Handy-PEA instrument (Hansatech Instruments Ltd, UK). Plants were dark-adapted for 30 min and detached leaves were then placed in a modified Handy-PEA leaf clip. The leaf sample was illuminated with continuous red light (3500 µE m⁻² s⁻¹, 650 nm peak wavelength; the spectral half-width was 22 nm; the light emitted by the LEDs is cut off at 700 nm by a NIR short-pass filter). The light was provided by an array of three light-emitting diodes focused on the sample surface. The first reliably measured point of the fluorescence transient is at 20 µs, which can be taken as F₀ (O). The length of the measurements was 1 s.

Fourteen-day-old in vitro-grown plants (Col-0 ecotype) were transferred to 150 mM NaCl and/or sprayed with 50 µM DCMU solution. OJIP fluorescence was measured 3 and 24 h after DCMU and salt treatments. Proline accumulation was measured 24, 48, and 72 h after DCMU treatment, whereas gene expression was measured after 24 h.

To test transcript levels of selected genes, quantitative RT-PCR (qRT-PCR) was performed on cDNA templates obtained from total RNA samples. RNA isolation was performed with Nucleo Spin RNA isolation kit (Macherei-Nagel). Total RNA was DNase treated with TURBO DNA-free™ Kit (Invitrogen by Thermo Fisher Scientific). First-strand cDNA synthesis of 1.5 µg of total RNA was carried out with RevertAid M-MuLV Reverse Transcriptase (Fermentas), using random hexamers. Real-time PCR was carried out with the ABI 7900 Fast Real Time System (Applied Biosystems). The protocol in 45 cycles was 15 s at 95°C, followed by 1 min at 60°C. The specificity of the amplifications was verified using the ABI SDS software. Expression of the P5CS1 (AT2G39800) and PDH1 (AT3G30775) genes was monitored by qRT-PCR as described (Aleksza et al., 2017). Normalized transcript levels were calculated by the modified 2^−ΔΔCt method using averages of actin2 (AT2G37620) and UBQ1 (AT3G52590) Ct values as reference (Livak and Schmittgen, 2001; Vandesompele et al., 2002). In relative expression data of the figures, reference was obtained on non-treated plants at the start of the experiment (e.g. dark-adapted plants, just before light and stress treatments). Statistical analysis was made on 2^−ΔΔCt values of three replicates corresponding to cDNA templates and RNA samples isolated from three different Petri plates. Experiments were repeated at least twice. Primers used in qRT-PCR experiments are listed in Table S1 .

The average amplification efficiencies of each primer pair used in the qRT-PCR experiments were derived from the slope of the amplification curve in the exponential phase of three different reactions from three different samples. The corresponding PCR efficiency was calculated according to the formula: E = 10 (1/slope) (Svec et al., 2015). Each primer showed high amplification efficiency from 1.99 to 2.03. Sequences of the PCR primers are available in Table S1 .

The ChIP protocol by Werner Aufsatz (http://www.epigenome-noe.net/researchtools/protocol.php?protid=13) was applied with the following modifications. Fourteen-day-old hy5 mutant plants expressing HY5-YFP fusion proteins from the HY5 promoter (Hajdu et al., 2018) were fixed in 1% (V/V) formaldehyde solution. Chromatin samples were sonicated on ice six times for 10 s using a Vibra Cell sonicator (SONICS & MATERIALS Inc., Danbury, CT, USA) at 10% power. Sonicated and diluted chromatin samples were pre-cleared by 20 µl (bed volume) of binding control agarose beads (Chromotek GmbH, Germany) for 1 h at 4°C. An aliquot of the pre-cleared chromatin solution was saved for the input sample and the rest of the material was precipitated using 12.5 µl (bed volume) of GFP-Trap agarose beads (Chromotek GmbH, Germany) for 16 h at 4°C. Precipitated chromatin was eluted from the beads, and along with the input sample, it was de-crosslinked and DNA was extracted using the Silica Bead DNA Gel Extraction Kit (Thermo Scientific). The final volume of purified DNA samples was about 45 µl. 1.5 µl of the eluate was analyzed in qPCR reactions. Primers were designed to amplify genomic regions around the putative HY5 binding sites. Standard series were prepared from 10-fold dilutions of the input DNA samples. ChIP-related qPCR primers are listed in . ChIP data were analyzed and presented according to the “percent of input” method (Haring et al., 2007). Experiments were repeated three times.

The pET28a vector carrying the full-length HY5 cDNA fragment (Hajdu et al., 2018) was introduced into Escherichia coli BL21 DE3 Rosetta cells (New England 513 Biolabs). Proteins were purified on His-Select Nickel affinity gel (SIGMA). 2 µg of purified protein was incubated for 30 min with 2 pmol biotin-labeled DNA (respective P5CS1 and PDH1 oligonucleotide sequences are available in ). DNA fragments and complexes were separated in 4% (W/V) native polyacrylamide gel, then blotted to HyBond-N⁺ nucleic acid transfer membrane (Amersham). DNA fragments were crosslinked to the membrane with UV light (UV Stratalinker, Stratagene). DNA fragments were detected by an immune reaction with Streptavidin-conjugated horseradish-peroxidase (Thermo Scientific) using the LightShift Chemiluminescent EMSA Kit (Thermo Scientific). Signals were developed with a chemiluminescent substrate (Supersignal West-Thermo Scientific) and detected in Fusion FX western blot and gel documentation imaging device (Vilber). Experiments were repeated twice.

Promoter sequence analysis was performed with AthaMap tool (http://www.athamap.de). Oligonucleotides were designed and analyzed by IDT OligoAnalyzer (https://eu.idtdna.com/calc/analyzer). Oligonucleotides used in this study are listed in Table S1.

Statistical analyses (one-way and two-way ANOVA, means comparisons by Tukey tests) were performed using the OriginPro 2018 software version 9.5 (OriginLab Corporation, Northampton, MA, USA). In case of one-way ANOVA the differences between means were determined Tukey test or by Duncan’s multiple range test and labeled in all diagrams by different letters. When two-way ANOVA was used, the means comparison were made with Tukey test. Data were processed and in some experiments Diagrams were prepared with MS Excel 14.7.7, and figures were assembled with MS Powerpoint 14.7.7 and Adobe Photoshop CS5.1.