Autophagy Controls Sulphur Metabolism in the Rosette Leaves of Arabidopsis and Facilitates S Remobilization to the Seeds.

Sulphur deficiency in crops became an agricultural concern several decades ago, due to the decrease of S deposition and the atmospheric sulphur dioxide emissions released by industrial plants. Autophagy, which is a conserved mechanism for nutrient recycling in eukaryotes, is involved in nitrogen, iron, zinc and manganese remobilizations from the rosette to the seeds in Arabidopsis thaliana. Here, we have compared the role of autophagy in sulphur and nitrogen management at the whole plant level, performing concurrent labelling with 34S and 15N isotopes on atg5 mutants and control lines. We show that both 34S and 15N remobilizations from the rosette to the seeds are impaired in the atg5 mutants irrespective of salicylic acid accumulation and of sulphur nutrition. The comparison in each genotype of the partitions of 15N and 34S in the seeds (as % of the whole plant) indicates that the remobilization of 34S to the seeds was twice more efficient than that of 15N in both autophagy mutants and control lines under high S conditions, and also in control lines under low S conditions. This was different in the autophagy mutants grown under low S conditions. Under low S, the partition of 34S to their seeds was indeed not twice as high but similar to that of 15N. Such discrepancy shows that when sulphate availability is scarce, autophagy mutants display stronger defects for 34S remobilization relative to 15N remobilization than under high S conditions. It suggests, moreover, that autophagy mainly affects the transport of N-poor S-containing molecules and possibly sulphate.

1. Introduction

Macroautophagy consists in the formation of cytosolic double membrane vesicles named autophagosomes that engulf and sequester unwanted cytoplasmic constituents to drive them to the central vacuole. The vacuole hydrolases and proteases then degrade unwanted material and release amino acids and sugars to be recycled [1]. In plants, macroautophagy is enhanced in senescing leaves. By contributing to the degradation of useless cytoplasmic components and altered organelles, it facilitates the release of nutrients that are then available to the source for sink mobilization [2,3]. In n class="Species">Arabidopsis and maize, it was shown using mutants that functional autophagy is essential for the remobilization of nitrogen from vegetative organs to the seeds [4,5]. More recently, Chen et al. [6] showed that enhancing autophagy in Arabidopsis by overexpressing ATG8 genes leads to a better N remobilization to the seeds.

Autophagosome formation incorporates the products of many AUTOPHAGY (ATG) genes, among which eighteen are essential for the formation of autophagosomes [7]. The ATG5 gene is one of these core-machinery genes. Like other autophagy mutants, the n class="Gene">atg5 mutants display a hypersenescent phenotype, having accumulated a large amount of proteins in their rosettes during ageing [5,8]. The 15N labelling experiments showed that N remobilization to the seeds was decreased by 40% in the atg5 mutants relative to the wild type, and the combination of atg5 mutations with salicylic acid (SA) deficient mutations partially abolished the senescence phenotypes but not the N remobilization deficiency. Recently, it was shown that overexpressing ATG5 in Arabidopsis stimulated autophagosome formation, autophagic flux, delayed senescence and increased resistance to necrotrophic pathogen, plant growth, seed set and seed oil content [9].

Because the cargoes taken in charge by macroautophagy for degradation are not only proteins but also organelle fragments, autophagy is certainly not only involved in the recycling of proteins or nitrogen sources. Recent advances show that it can also contribute to the recycling of n class="Chemical">lipids and of any kind of micronutrients contained in the macromolecules [3,10,11]. Accordingly, it was recently shown, using 57Fe labelling and measuring metal budgets, that the efficiencies of iron, zinc and manganese translocation from vegetative organs to the seeds were dramatically decreased in the atg5 mutant [10].

Like nitrogen, n class="Chemical">sulphur (S) is an essential element for growth and metabolic functioning in plants [12]. It is an important constituent of proteins due to S-containing amino acids like methionine and cysteine. S fertilization in crops has been an agricultural concern since the 1980s. Due to the decrease of atmospheric sulphur dioxide (SO2) from industrial emissions, S deposition into the soil was strongly reduced, leading to the increasing occurrence of S deficiency in crops, at least in Western Europe [13,14]. Brassicaceae plants are highly S-demanding, as they have high concentrations of S-containing secondary metabolites, and S limitation can severely impact their seed yield and quality [15]. Therefore, the number of studies on S use efficiency started increasing since the 1980s.

In this study we used the same genotypes as Guiboileau et al. [5] to investigate the role of autophagy on n class="Chemical">sulphur remobilization to the seeds. Simultaneous labelling with 34SO42− and 15NO3− was performed in order to compare the N remobilization efficiency (NRE) and the S remobilization efficiency (SRE). Our results indicate that SRE was surprisingly higher under high sulphur conditions than under sulphur limitation in both autophagy mutants and control lines. However, the atg5 mutants were less efficient in S remobilization than the control lines under both conditions. In addition, by comparison with N remobilization, we show that low S conditions exacerbate the defects of autophagy mutants for S remobilization, suggesting the occurrence of a selectivity with regard to autophagy in the different S sources available in plants for remobilization.

2. Materials and Methods

2.1. Plant Material and Growth Conditions

Arabidopsis mutants n class="Gene">atg5-1 [SALK_020601], atg5-2 [SAIL_129B07], sid2.atg5-1 and NahG.atg5-2. were the same as used before by Guiboileau et al. [5]. The plants were grown on sand on either high sulphate supply (High S; 0.266 mM SO42−, 10 mM NO3−) or low sulphate (Low S; 0.016 mM SO42−, 10 mM NO3−) conditions. For the S and N remobilization experiment, the plants were sown on sand and cultivated under high S conditions for four weeks in short days (8 h light–16 h dark, 150 µmoL/m2/s). Then, the plants were labelled for five days using a labelled high S nutrient solution containing 15N and 34S isotopes (15N 10% enrichment and 34S 2.5%). After 5 d labelling, some of the plants were harvested for isotopic ratio measurement, to estimate the level of 34S and 15N uptake over five days by the different genotypes. The pots containing sand and roots of the remaining plants, dedicated to measure 34S and 15N remobilizations, were rinsed carefully several times with deionised water. Then, half of these plants were transferred to unlabelled low S conditions and half to unlabelled high S conditions. The plants were then grown in short days (8 h light) for eight weeks before the transfer to long days (16 h light). Similar day/night temperatures (21 °C day, 17 °C night) were maintained until seed maturity.

2.2. 15N Labelling and Tracing

The plants were harvested at the end of their reproduction cycle, when both vegetative tissues and seeds were dry. At harvest, the plants were divided into the following parts: rosette, stem (including cauline leaves), silique envelopes (pericarp) and seeds. The dry weight (DW) of all these compartments was determined. Then, the n class="Chemical">34S and 15N abundances in all collected samples were determined as described in Diaz et al. [16]. The 34S abundance was calculated as atom percent and defined as A(34S)% = [100 × (34S)/(34S + 35S)] and the 15N abundance as A(15N)% = [100 × (15N)/(15N + 14N)]. The S and N concentrations (S% and N%) as mg of element per 100 mg of dry weight were also determined. The enrichment (E%) was calculated as (Asample% – Acontrol%). Acontrol% was determined for unlabelled samples. The absolute quantity of 34S contained in a sample was defined as [DW × E(34S)% × S%] and the absolute quantity of 15N contained in a sample as [DW × E(15N)% × N%]. The partitioning of 34S was calculated for each organ (rosette, stem, pericarp and seeds) as the proportion of the quantity of 34S in each organ relative to the total quantity in the whole plant, and expressed as % of the whole plant. As such, the sum of the partitionings (rosette, stem, pericarp and seeds) was 100%. The same was done for the 15N partitioning. Unlabelled samples were harvested in order to determine the natural abundance of 34S and 15N.

2.3. Metabolite Profiling and Sulphate Measurements

Metabolite profiling and metabolomics data processing were performed according to Masclaux-Daubresse et al. [17]. The plant material used for the metabolomics analyses consisted in the rosette of 60-days-old plants grown under high S or low S conditions and short days. n class="Chemical">Sulphate concentrations were measured on the dry matter obtained from the 15N labelled plants at seed maturity. Sulphate was extracted from ca. 25 mg of freeze dried matter (DM), which was initially mixed up with 1 mL of 50% ethanol [18]. After incubation at 45 °C for 1 h, the extract was centrifuged at 10,000g for 10 min, and the supernatant was collected. The ethanol extraction step was reiterated on the pellet, and the supernatant was pooled with the previous one. The pellet was then re-extracted in 1 mL of distilled water at 95 °C for 1 h. The extracts were centrifuged at 10,000g for 10 min. All the supernatants were pooled and evaporated under vacuum (Concentrator Evaporator RC 10.22, Jouan, Saint-Herblain, France). The dry tissue was re-suspended in 0.5 mL of ultrapure water and filtered to remove any plant residues. The sulphate concentration was determined by high performance liquid chromatography (HPLC, DX100, Dionex Corp., Sunnyvale, CA, USA). The eluent solution consisted of 1.8 mM Na2CO3 and 1.7 mM Na2HCO3, and was pumped isocratically over an AS17 guard column.

2.4. Statistical Analysis

The 34S and n class="Chemical">15N labelling experiments were carried out in two independent biological repeats (R1 and R2), performed as two consecutive culture cycles in the same growth chamber. An ANOVA Newman–Keuls (SNK) comparison was performed using XLSTAT (13.2). As R1 and R2 provided very similar data, the mean and SD were calculated based on the R1 and R2 data (n = 12; 6 plants per genotype for each repeat, R1 and R2). For metabolomics, an ANOVA was first used to select the significant metabolites. The subset of significant metabolites was then used to perform hierarchical clustering using MEV_4_8_1 free software (https://sourceforge.net/projects/mev-tm4/).

3. Results

3.1. Hypersensitivity of Autophagy Mutants to Sulphur Limitation

It is well known that autophagy mutants are hypersensitive to n class="Chemical">carbon and nitrogen starvations [2]. When submitted to chronic sulphur limitation, atg5 mutants displayed earlier leaf yellowing phenotypes (Figure 1A), which suggested defects in the management of sulphur resources.

Figure 1

The biomass of atg5 autophagy mutants is affected under both low sulphur (Low S) and high sulphur (High S) conditions. (A) Phenotype of atg5 mutants under high S and low S conditions 60 days after sowing. (B) Schematic representation of the labelling experiments performed on atg5 mutants (atg5-1, atg5-2, atg5-1.sid2, atg5-2.NahG) and control lines (Col-0, sid2 and NahG). The plants were sown on sand and cultivated under high S conditions for four weeks in short days (8 h light–16 h dark, 150 µmoL/m2/s1). Then, the plants were labelled for five days using a high S nutrient solution containing 15N and 34S isotopes. After 5 d labelling, half of the plants were transferred to unlabelled low S conditions and half to unlabelled high S conditions. After eight weeks, the plants were transferred to long days (16 h light), maintaining similar day/night temperatures (21 °C day, 17 °C night) until seed maturity. (C) The biomasses of the rosette, stems, pericarps and seeds were determined at harvest on the 34S and 15N labelled plants (dry weight, DW, g). The numbers above the atg5 mutant bars indicate the % of biomass decrease in mutants relative to their respective control lines. Data are the adjusted means and SD from two biological repeats with six plants each (n = 12). The different letters indicate values of total biomass significantly different at p < 0.05 (n = 12) between atg5 mutants and control lines, as determined using an ANOVA Newman–Keuls (SNK) comparison.

It is also well known that autophagy mutants accumulate salicylic acid. The combination of atg mutations with n class="Gene">sid2 mutation or with NahG overexpression has already been used by Guiboileau et al. [5] to investigate the SA-independent autophagy effects on N remobilization. We have thus used the same genotypes as Guiboileau et al. to determine how biomass partitioning and sulphur allocation are modified in plants defective in autophagy, independently of the salicylic acid effect. Nitrogen and sulphur labelling were performed on four-week-old plants providing 34SO42− and 15NO3− every day during five days (Figure 1B). Since N remobilization has already been monitored by Guiboileau et al. [5], we performed double labelling with 34S and 15N in order to use 15N as a control of the source to sink nutrient fluxes. After labelling, plants were watered with isotope-free solutions and grown under low or high sulphur conditions until seed maturity. Under both low and high sulphur conditions, the atg5-1, atg5-2, sid2.atg5-1 and NahG.atg5-2 plants were (64% to 78%) smaller at harvest than their respective control lines (i.e., Col-0 for atg5-1 and atg5-2, sid2 for sid2.atg5-1, and NahG for NahG.atg5-2; Figure 1C). However, the decrease of plant biomass in the atg5 mutants relative to the control lines was globally similar under low S and high S conditions.

Histograms representing the partition of dry weight (DW) in the rosettes, stem, pericarp and seeds (Figure 2) show that the patterns were similar under low S and high S for all the genotypes. Under both low S and high S, the partition of the biomass in the n class="Chemical">rosettes of atg5 mutants was more important relative to the control lines, and the biomass allocated to siliques (both pericarps and seeds) was lower in the atg5 mutants than in the control lines. This was consistent with the lower harvest index already reported by Guiboileau et al. for autophagy mutants [5]. It can be noticed that the partition of DW in the stems was poorly modified in the atg5 mutants by comparison with the control lines (less than 10%), irrespective of S conditions (Fig2A–B). Thus, it showed that autophagy mainly impacted the rosette and seed parts.

Figure 2

The atg5 autophagy mutants display a lower harvest index than the control lines under both low S and high S. From the measurement of the dry weights of the rosette, stem, pericarp and seeds presented in Figure 1A, the partitioning of the dry matter in each organ was computed. The numbers above the atg5 mutant bars indicate the % of significant increase or decrease in each mutant relative to its respective control line. Data are the adjusted means and SD from two biological repeats with six plants each. The different letters indicate values significantly different at p < 0.05 (n = 12), as determined using an ANOVA Newman–Keuls (SNK) comparison.

3.2. Autophagy Mutants Hyper-Accumulate Sulphur

As expected, total sulphur concentrations (S%; n class="Chemical">mg.100 mg-1 DW) were lower in the plants grown under low S than in those grown under high S (Figure 3). In the control lines, S% was higher in the seeds than in the other organs, irrespective of S conditions. This indicated that a large part of the S resources in plants was allocated to the seeds at harvest in the control lines. This feature was in clear contrast with that of atg5 mutants, whose S% in the rosette, stem and pericarp (i.e., dry remains) remained almost as high as the S% measured in their seeds. It is noticeable that the S% in all the organs of atg5 mutants were higher than those measured in the control lines, except in seeds under high S.

Figure 3

Sulphur concentrations are higher in the atg5 autophagy mutants than in the control lines under both low S and high S. S concentrations (S% as mg.100 mg-1 DW) in the rosette, stem, pericarp and seeds of atg5 mutants (atg5-1, atg5-2, atg5-1.sid2, atg5-2.NahG) and control lines (Col-0, sid2 and NahG) were determined on plants grown under low S and high S. Data are the adjusted means and SD from two biological repeats with six plants each. The different letters indicate values significantly different at p < 0.05 (n = 12), as determined using an ANOVA Newman–Keuls (SNK) comparison.

The reason why sulphur concentrations were higher in the n class="Gene">atg5 plants than in controls is certainly related to their lower growth rate, that resulted in lower S dilution. Nevertheless, the strong accumulation of S, in their dry remains especially, suggests that autophagy mutants displayed defects in S reallocation during plant development.

Although the genotype effects on the S% in the rosette, stem and pericarp were similar under low S and high S, we can observe that the genotype effect on the S% in seeds is different depending on high S and low S conditions. The S concentration in seeds was lower in the control lines than in n class="Gene">atg5 mutants under low S, but similar under high S. This reveals that S limitation was more critical in the control lines than in atg5 mutants under low S. Again, this could be explained by the fact that atg5 plants were smaller than control plants.

The values of S% measured in seeds raise two questions: (i) How can the S% (S concentration) in the seeds of the atg5 mutants be maintained as high as in the control lines, while S remobilization to the seeds is lower in the n class="Gene">atg5 mutants relative to the control lines? (ii) Why is the S% in the seeds of atg5 similar to that of the control lines, while the S% in the other organs are much lower in atg5 mutants than in the control lines? The explanation of this intriguing observation is that the lower S flux in autophagy mutants led to (i) a lower amount of S allocated to the reproductive organs and (ii) less seeds produced [19]. Indeed, the biomass of the rosette (source leaves) was 2 times lower in the atg5 mutants than in the control lines at harvest, irrespective of S conditions. The biomass of the seeds of the atg5 mutants was 4.5 and 6.2 times lower than that of the control lines under low S and high S, respectively. The S content (in mg) in the atg5 mutants was 3 and 6 times lower than that of the controls under low S and high S, respectively. The 34S content (in mg) in atg5 mutants was 3.5 and 3.2 times lower than that of the control lines under low S and high S, respectively. Therefore, we see that the absence of autophagy in Arabidopsis had a stronger effect on seed production than on the allocation of S to the seeds. As a result, the autophagy mutants produced less seeds properly filled with S and with N [5].

3.3. Sulphur Remobilization from the Rosette to the Seeds and Pericarp Is Less Efficient in atg5 Mutants

The 34S labelling was performed before transferring the plants to S limitation. The total n class="Chemical">34S quantity absorbed by the different plants was globally similar, irrespective of genotypes and sulphur conditions (Figure S1A). However, the 34S allocation at harvest was different in the control lines and atg5 mutants (Figure S1B,C).

The 34S partitioning (also called n class="Chemical">34S harvest index, 34SHI) gives a clear picture of the 34S remobilization from the rosette to the stems, pericarp and seeds, independently of the differences that might exist in plant biomass (Figure 4). Surprisingly, 34S remobilization was more efficient under high S than under low S (Figure 4). Almost 80% and 65% of the 34S was remobilized from the rosettes to the rest of the plant in control plants under high S and low S, respectively. In atg5 mutants, this amount was significantly lower, representing only 50% and 25% under high S and low S, respectively. In the control plants, the largest proportion of 34S was found in the seeds, irrespective of the nutritive conditions, and only 13%–12% has remained in the stems and pericarps. The proportion of 34S remaining in the stems of atg5 (13% in average) was globally similar to that found in the control lines. Only the 34S remobilization to the seeds was significantly different between the control lines and atg5 mutants. The allocation of 34S to the seeds was 5 times and 2 times lower in atg5 than in the control lines under low S and high S, respectively (Figure 4).

Figure 4

The partitioning of 34S in the rosette and seeds of atg5 autophagy mutants was significantly different from that of the control lines under both low S and high S. The partitioning of 34S in each organ was calculated as the % of 34S in each organ relative to the total quantity of 34S in the whole plant (see material and methods and supplemental Figure S1). Data are the adjusted means and SD from two biological repeats with six plants each. The different letters indicate values significantly different at p < 0.05 (n = 12), as determined using an ANOVA Newman–Keuls (SNK) comparison.

3.4. Relative Rates of 34S and 15N Remobilizations Are Different in Control Lines and atg5 Mutants Under Low S Conditions

Under both low S and high S, the 34S and n class="Chemical">15N remobilizations to the seeds were significantly higher in the control lines than in atg5 mutants (Figure 5A–D). The N and S remobilization defects in atg5 mutants were independent of SA. As in the case of 34S, we observed that the 15N remobilization to the seeds was more efficient under high S than under low S (Figure 5B,D). This revealed that the S status of the plants was important for both N and S nutrient fluxes at the whole plant level.

Figure 5

The remobilization of 34S and 15N to the seeds is differentially modified by sulphate availability in atg5 autophagy mutants and control lines. The partitioning of 34S (A,C) and 15N (B,D) in the rosette, stem, pericarp and seeds of atg5 mutants and control lines under low S (A,B) and high S (C,D) conditions. The ratios of 34S and 15N partitions in the seeds show different patterns depending on genotypes under high S and low S conditions (E). Data are the adjusted means and SD from two biological repeats with six plants each. The different letters indicate values significantly different at p < 0.05 (n = 12), as determined using an ANOVA Newman–Keuls (SNK) comparison.

To compare the 34S and n class="Chemical">15N fluxes in both S conditions, the partitionings of 34S (34SHI) and 15N (15NHI, 15N harvest index) in seeds were compared considering the 34SHI/15NHI ratio. Under both low S and high S conditions, the remobilization of 34S in the control lines was more efficient than the remobilization of 15N (34SHI/15NHI ratio above 150–200%) (Figure 5E). This indicated that the mobilization of S from the rosette to the seeds was probably not exclusively supported by S amino acids. Under high S, the 34SHI/15NHI ratio in atg5 mutants was mostly similar to that of the control lines. However, under low S, this ratio was interestingly much lower in atg5 mutants (globally equal to 100%) than in the control lines (remaining equal to 150–200%, as under high S). Such discrepancy indicated that under low S (i), 34S remobilization to the seeds was strongly reduced in the atg5 mutants relative to the control lines and that (ii) under low S conditions, the S forms remobilized from the source organs to the seeds in atg5 mutants were certainly mainly S amino acid forms. It is likely that the decrease in 34S remobilization in atg5 mutants relative to control lines under low S was due to defects in mobilizing N-poor S-containing molecules.

3.5. Metabolite Accumulation in the Rosette Leaves of atg5 Mutants Is Exacerbated under Low S Irrespective of Salicylic Acid Production

In order to determine whether amino acids, and especially S-containing amino acids accumulated in the leaf tissues of autophagy mutants, metabolite profiling was performed using GC–MS (gaz chromatography coupled to mass spectrometry). We first compared the metabolite relative contents in atg5-1 and Col-0 under high S and low S (Figure 6). Results showed that, relative to Col-0, most of the metabolites accumulated in n class="Gene">atg5-1 under both low S and high S. Only urate and threitol contents were lower in atg5-1 than in Col-0. Surprisingly, glucose, fructose, galactose, mannose, fumarate and xylose accumulated in atg5-1 under low S, relative to Col-0, but were lower under high S. Like other amino acids, cysteine and methionine accumulated significantly in atg5-1. Such accumulation was consistent with the defects observed for N and S remobilization in atg5 mutants. To determine whether the accumulation of metabolite was salicylic acid-dependent, we then compared the metabolome of atg5-1, atg5-1.sid2, Col-0 and sid2 under low S conditions (Figure 7). Again, we observed that, globally, metabolites were more abundant in atg5 mutants vs. the control lines, irrespective of SA (Figure 7). The few metabolites depleted in atg5-1 vs. Col-0 (urate, threitol, linoleic acid, spermidine, dehydroascorbate and porphine) were also depleted in atg5-1.sid2 compared to sid2. This then proved that the hyperaccumulation of both C and N metabolites in atg5-1 under low S was SA-independent. This feature was consistent with the SA-independent effect of atg5 mutation on N and S remobilization (Figure 5A–D).

Figure 6

Hierarchic clustering of metabolite relative contents in atg5-1 and Col-0. Metabolite relative contents in the rosettes of atg5 and Col-0 wild type genotypes, grown under high sulphate and low sulphate conditions for 60 days, were measured using GC–MS. An ANOVA was performed to identify metabolites significantly modified by genotype or sulphate availability. The hierarchic clustering of significant metabolites was performed using MEV4. The clusters show metabolites less abundant in atg5 relative to Col-0 under both high and low sulphate conditions (green), less abundant in atg5 relative to Col-0 under high S but more abundant under low S (orange), more abundant in atg5 relative to Col-0 under both high and low sulphate conditions (red), and not different in atg5 and Col-0 but modified by sulphate availability (blue). The red arrows indicate the position of cysteine and methionine. Three biological repeats are shown for each genotype and condition.

Figure 7

Hierarchic clustering of metabolite relative contents in atg5-1, atg5-1.sid2, sid2 and Col-0 under low S conditions. The metabolite relative contents in the rosettes of atg5, atg5-1.sid2, sid2 and Col-0 plants grown under low sulphate conditions for 60 days were measured using GC-MS. An ANOVA was performed to identify metabolites significantly modified by genotype. The hierarchic clustering of significant metabolites was performed MEV4. The clusters show metabolites less abundant in atg5-1 and atg5-1.sid2 relative to Col-0 and sid2, respectively (blue), more abundant in atg5-1 and atg5-1.sid2 relative to Col-0 and sid2, respectively (red), and more abundant in atg5-1 only by comparison with the three other genotypes (green). The red arrow indicates the position of cysteine. Three biological repeats are shown for each genotype.

In order to determine whether autophagy could impair the translocation of inorganic S from the rosette leaves to the seeds, the concentrations of the n class="Chemical">sulphate remaining in the rosettes of plants after seed production were measured using the same plant material used for the 34S remobilization assay. Interestingly, although sulphate availability in the soil was limited under low S conditions, the concentrations of sulphate in the dry rosettes of the atg5-1 and atg5-2 mutants at harvest were twice as high as those found in the dry rosettes of the Col-0 wild type (Figure 8). Sulphate concentrations were also significantly higher in atg5-1.sid2 and atg5-2.NahG than in sid2 and NahG, respectively. This indicates that sulphate was blocked in the rosettes of the atg5 mutants, contributing to the defect of S remobilization to the seeds.

Figure 8

Sulphate concentrations in the rosette of atg5 autophagy mutants are significantly higher than in the rosettes of control lines under low S. Sulphate was measured on the dry remains of the rosettes at seed maturity. Data are the adjusted means and SD from six plant repeats The different letters indicate values significantly different at p < 0.05 (n = 6), as determined using an ANOVA Newman–Keuls (SNK) comparison.

4. Discussion

There are now several lines of evidence that autophagy is involved in the remobilization of macro- and micro-elements in plants [4,5,6,10]. Using 34S isotopes, we show here that autophagy is controlling n class="Chemical">sulphur remobilization in a SA-independent manner and is essentially blocking the S flux from the rosette leaves to the seeds. The absence of difference between the atg5 mutants and the control lines for 34S partitioning in the stem and pericarps reveals that the steps mainly controlled by autophagy are (i) the export of S forms from the source rosette leaves and (ii) the import of the S-mobilized forms into the seeds. This feature is in good accordance with the expression levels of autophagy genes, that were shown to be increased in senescing leaves and in seeds during maturation. The S forms transported from source to sink in plants are not well known. Sulphur remobilization may involve organic forms, as glutathione, glucosinolates, methiine (S-methylcysteine sulphoxide) [20] and the methionine and cysteine S amino acids, and also inorganic forms such as sulphate, at least when SO42− availability for plants is sufficient.

The metabolite profiling performed on autophagy mutants and control lines showed that most of the amino acids, including cysteine and n class="Chemical">methionine, accumulated in the rosettes of atg5 mutants irrespective of S nutrition. This feature is consistent with the metabolic profiling previously reported by Masclaux-Daubresse et al. [17]. In their reports, the authors also showed that glutathione highly accumulated in the rosette leaves of autophagy mutants, possibly due to excessive oxidative stress. Glutathione storage in rosette leaves of autophagy mutants may impair S remobilization.

The proteomic analyses performed by Havé et al. [21] showed that the MTO2 (n class="Gene">methionine overaccumulator 2; threonine synthase) protein was less abundant in atg5 mutants than in the control lines. MTO2 was shown to lead to methionine overaccumulation. Therefore, the lower MTO2 content in atg5 mutants might explain the accumulation of methionine in the atg5 rosettes. Higher abundances of enzymes involved in cysteine, methionine and glucosinolate biosynthesis in atg5 mutants irrespective of SA and of growth conditions have also been reported by Havé et al. [21]. These differences, relative to the control lines, may also explain the increase of S amino acids measured in autophagy mutants (Figure 9). Nevertheless, the defects in methionine, cysteine and glutathione translocation from the leaves to the seeds in autophagy mutants remain to be explored.

Figure 9

Schematic representation of the effect of atg5 mutation on sulphur assimilation in plants grown under low S conditions. Protein accumulation or depletion in atg5 and atg5.sid2 mutants has been reported by Havé et al. [21]. Proteins overaccumulated in atg5 or atg5.sid2 vs. Col-0 or sid2 are presented in red boxes. Proteins depleted are presented in blue boxes. ATPS, ATP sulphurylase; APR, adenosine 5-phosphosulphate reductase; SiR, sulphite reductase; OAS, O-acetylserine; OASTL, O-acetylserine (thiol) lyase; SOT16, cytosolic sulphotransferase 16.

As said before, autophagy is not only involved in the recycling of proteins; it can also contribute to the degradation of n class="Chemical">lipids, and especially membrane lipids [21,22]. Its impact on lipid metabolism and turnover could contribute to impair the S mobilization from cell to cell and from source organs to sinks. How autophagy impairs the mobilization of S from lipids and proteins and whether there are specific cargoes degraded by autophagy under sulphate limitation remains to be investigated.

The lower S remobilization in autophagy mutants could also be due to a lower translocation of the sulphate itself. The n class="Chemical">34S labelling did not reveal any defect in the sulphate uptake in atg5 mutants. Accordingly, the transcriptomic and proteomic data reported by Havé et al. [17,21] did not reveal any changes in the transcript or protein levels of the sulphate transporters in the atg5 mutants. However, the higher concentrations of sulphate measured at the end of seed maturity, in the rosettes of the atg5 mutants grown under low S conditions, strongly suggest that sulphate was not as well assimilated in autophagy mutants as in the control lines, and that the remobilization from the rosettes to the seeds was impaired in autophagy mutants.

Still, the comparison of SRE (34SHI) and n class="Chemical">NRE (15NHI) shows that S remobilization was higher than N remobilization under high sulphate conditions (2-fold). This means that the conventional S-mobile forms for source to sink mobilization in plants (both control and atg5 mutants) are not only N-containing molecules (like amino acids, methiine or glutathione, for example) but also N-free molecules. Interestingly, the fact that under low S, the 34SHI/15NHI ratio was lower in the atg5 mutants (from 200% to 100%) than in the control lines, indicates that SRE was more affected than NRE by low S in the atg5 mutants. This also indicates that some S-mobile forms were missing in atg5 relative to the control lines. These missing S-mobile forms were certainly N-poor S-containing molecules. Although their nature remains to be determined, our results suggest that the composition of the S-mobile pools in atg5 mutants and in control lines are different, at least under low S conditions.

In conclusion, our study shows that (i) S remobilization from the rosette to the seeds is impaired in the n class="Gene">atg5 mutants when compared to the control lines, irrespective of SA accumulation and S conditions; (ii) many metabolites, among which S-free amino acids, as well as methionine and cysteine, over-accumulate in the rosette leaves of the atg5 mutants; and (iii) the nature of the S-mobile molecules available for remobilization is different in the atg5 and the control lines, at least under low S conditions. Under low S, it is likely that the absence of autophagy affects the transport of N-poor S-containing molecules and that the SA effect is partially involved in this process.