Autophagy controls resource allocation and protein storage accumulation in Arabidopsis seeds

Autophagy takes place in seeds, as evidenced by detection of autophagosomes and high expression of ATG genes in Arabidopsis, and the seeds of the atg5 mutant exhibit early browning as well as low and early accumulation of 12S globulin.


Introduction
Macroautophagy (hereafter termed autophagy) is a universal degradation mechanism that facilitates the clearing of unwanted constituents from eukaryotic cells. Autophagy comprises the formation of a cytosolic double-membrane vesicle, termed an autophagosome, that engulfs and sequesters cytoplasmic material such as damaged organelles and protein aggregates prior to their degradation (Liu and Bassham, 2012). Autophagy is essential for the recycling of cellular material and controls nitrogen remobilization at the whole-plant level. In Arabidopsis and maize, autophagy is essential for grain filling (Guiboileau et al., 2012;Li et al., 2015;Masclaux-Daubresse et al., 2017). The autophagy machinery needs the products of many AUTOPHAGY (ATG) genes. ATG genes were discovered by the pioneering work of Professor Yoshinori Ohsumi (Nobel Prize in Physiology or Medicine, 2016), which consisted of screening autophagy-defective yeast mutants under starvation conditions (Tsukada and Ohsumi, 1993). Yeast homologous genes have been found in plants and animals for almost all the ATG genes. Among the 50 ATG genes discovered in yeast, 18 are part of the central autophagy machinery Yang and Bassham, 2015) and are absolutely essential for the formation of autophagosomes. In Arabidopsis, about 40 ATG genes, among which most are in the core machinery list, have been identified. They are either single genes or members of family genes (Zientara-Rytter and Sirko, 2016), but not all of them have been functionally characterized . ATG5 gene is a unique copy gene in all animal and plant species studied so far. It is involved in the ATG5-ATG12 conjugation system that is essential for the formation of the ATG8-PE (phosphatidylethanolamine) conjugate that characterizes the autophagosome membrane (Yoshimoto et al., 2004). Arabidopsis atg5 mutants have been characterized by many research groups (Thompson et al., 2005;Yoshimoto et al., 2009;Guiboileau et al., 2012;Lee et al., 2013;Minina et al., 2013;Izumi et al., 2014;Sakuraba et al., 2014) and they present the typical phenotype of many other atg mutants (Doelling et al., 2002;Phillips et al., 2008;Yoshimoto et al., 2009;Wang et al., 2011;Michaeli et al., 2016), namely smaller rosette size, hypersensitivity to N and C starvation, reduced yield, and defects in nitrogen remobilization to the seeds. Using 15 N tracer experiments, Guiboileau et al. (2012) showed that atg5, as well as atg9 and atg18a, were strongly affected in N remobilization to the seeds. Despite its lower N remobilization from source leaves, seeds of atg5 (and of other atg mutants) presented higher percentage N content than those of the wild-type (WT), suggesting that in atg5 the seed sink strength for N relied on N sources originating from post-flowering uptake (Guiboileau et al., 2012). Differences in seed quality between atg and WT could be due to the strong effect of the atg mutations on the leaf metabolism of C and N, including photosynthesis and N remobilization (Guiboileau et al., 2013;Masclaux-Daubresse et al., 2014). However, it could also be due to seed metabolism and development. Indeed, transcriptome data that can be consulted on public databases such as the BAR (http://bar.utoronto.ca) or Genevestigator (https://genevestigator.com/gv/) sites shows that most of the Arabidopsis ATG genes are up-regulated in rosettes with ageing and in seeds during development. This led us to hypothesize that autophagy could play a physiological role in the seed, independently of its role in the mother plant.
In this study, seed development and seed protein contents were monitored in order to identify differences in N resource management in atg5 by comparison with WT seeds. In order to disconnect the effects of the senescence and autophagy processes, the sid2-atg5 double-mutant was used in this study. This mutant is affected in salicylic acid synthesis and exhibits a delayed senescence and a higher seed biomass compared to the single atg5 mutant, but an earlier senescence and a lower seed biomass compared to the Col wild-type (Guiboileau et al., 2012). Because seed filling and development can be influenced by the metabolism of the mother plant, and since autophagy is highly sensitive to nutrient availability, the seed phenotypes were monitored on seeds obtained from plants grown with ample (high) or limiting (low) nitrate supplies. Comparison of the phenotypes obtained in these conditions facilitates the identification of features related to mother-plant resource supply.
Dry seeds were stratified for 48 h at 4 °C in the dark and then sown on a sand substrate according to Masclaux-Daubresse and Chardon (2011 Pools of 100 seeds were dried at 70 °C for 24 h and weighed on a microbalance (XS3DU, Mettler Toledo, Viroflay, France).

Cloning and plant transformation
The Prom-ATG8f::GFP (green fluorescent protein) construct was created using the Gateway® technology (ThermoFisher Scientific). The ATG8f (At4g16520) promoter (1827 kb) was amplified by highfidelity Taq (Phusion High-Fidelity DNA Polymerase, ThermoFisher Scientific) by PCR, using the primers listed in Supplementary  Table S1 at JXB online, and flanked with attL recombination sites. pDONR207 (Invitrogen) was used as the entry vector and pGWB4 (Nakagawa et al., 2007) was used as the expression vector for the agro-transformation.
The Prom-Actin::GFP::ATG8f construct was created using a pEZS-CL vector (S. Cutler and D. Ehrhardt, Carnegie Institution for Science, Stanford, CA) whose 35S promoter region was replaced by the ACTIN2 (At3g18780) promoter. A digestion with the restriction enzymes SacI and NcoI was performed to remove the 35S promoter from the plasmid. The sequence of the ACTIN promoter was amplified with primers that provided compatible ends with these two restriction enzymes and then inserted into the plasmid. The Prom-Actin::GFP::ATG8f construct was extracted from the pEZS-CL vector with the NotI restriction enzyme, which generates blunt ends. This construct was finally inserted in the pCAMBIA vector (Cambia, Canberra, Australia) previously digested with SmaI.
Silique and seed developmental stages Determination of silique and seed developmental stages was performed according to Baud et al. (2008). Just before the opening of the floral bud, a cotton thread was tied around the bud pedicle. This moment coincides with the pollination step and was used as a reference for the age determination of siliques and seeds. The developmental unit used in this study is days after fertilization (DAF).

RNA isolation
When harvested, Arabidopsis siliques were immediately frozen in liquid nitrogen, then stored at -80 °C. For RNA isolation, three frozen siliques were ground using a mortar and pestle previously cooled with liquid nitrogen. In order to avoid the formation of a viscous paste, 250 µl of extraction buffer (0.1 M LiCl, 0.1 M Tris pH 8, 10 mM EDTA, 1% SDS, 1.5% β-mercaptoethanol) was added to the mortar during grinding. The powder obtained was treated twice with 450 µl of phenol/chloroform 5:1 (P1944, Sigma), vortexed, and centrifuged (13 000 g, 15 min, 4 °C). The supernatant phase was transferred to a 2-ml Eppendorf tube and 400 µl of 8 M LiCl solution was added and slowly mixed by turning the tube several times. RNAs were precipitated for 1 h at -80 °C and the tube was then centrifuged (13 000 g, 30 min, 4 °C). Then, 1 ml of 70% ethanol was added to the tube and it was centrifuged again (13 000 g, 30 min, 4 °C). The supernatant phase was removed and the pellet containing the RNA was dried by pipetting. Dry pellets were rehydrated with 40 µl of RNase-free water (average yield: 1 µg µl -1 ). After extraction, RNAs were purified using a Micro Bio-Spin column (7326206, Bio-Rad). Insoluble polyvinylpolypyrrolidone (PVPP, 5 mg) was placed on the column and then 40 µl of previously extracted RNA was added. The column was placed in an open 2-mL Eppendorf tube and centrifuged (1000 g, 2 min, room temperature). The collected RNAs were purified four times through the column (average yield: 0.5 µg µl -1 ), and then 2 µg of RNA was treated with the DNase I, RNase-free kit (EN0521, ThermoFisher Scientific). An inhibitor of RNase, RiboLock RNase Inhibitor (E00382, ThermoFisher Scientific) was added to the reaction mixture according to the manufacturer's protocol.

Reverse transcription and quantitative RT-qPCR
Reverse transcription of extracted RNA into cDNA was realized using the M-MuLV Reverse Transcriptase kit (EP0352, ThermoFisher Scientific) according to the manufacturer's protocol.
The RT-qPCRs were run on a CFX 96 thermocycler (Biorad) using a first step at 95 °C for 5 min and then 40 cycles of 5 s at 95 °C, 20 s at 60 °C, and 20 s at 72 °C. A final step consisted in an increase of 0.1 °C s -1 to 95 °C.
The primers used for RT-qPCR had an average length of 20 bases and were designed in order to amplify fragments between 100 and 240 bp (see Supplementary Table S1). All primers presented an efficiency of 100 ± 5%. For all RT-qPCR analyses, EF-1α (At5g60390) and APC2 (At2g04660) were used as reference genes (Dekkers et al., 2012;Srivastava et al., 2017).

Seed discolouration with sodium hydroxide
Fixed seeds were discoloured for at least 2 h at 37 °C in a sodium hydroxide solution (NaOH 200 mM, SDS 1%) then rinsed twice in water for 5 min. At this step, young seeds became pink due to the oxidation of tannins. Seeds were then discoloured for 30 min in a 5% bleach solution, then rinsed twice in water for 5 min. This discolouration technique was used on seeds after GUS staining.

GUS staining of seeds and siliques
Siliques were harvested, fixed and then incubated in a freshly prepared infiltration buffer (50 mM pH 7 phosphate buffer, 10 mM EDTA, 1% Triton X-100, 2 mM X-GlucA). After infiltration under vacuum, samples were placed in the dark at 37 °C for 2-24 h depending on the tissue. After incubation, samples were embedded in a resin, or directly observed, or discoloured by sodium hydroxide. White-light imaging was performed using a DMRB microscope (Leica).

Sample sections of siliques
For embedding of silique in resin or wax, samples have to be dehydrated in ethanol first. Fixed siliques were therefore immersed successively in solutions with an increasing gradient of ethanol (30%, 50%, 70%, and 100%, completed with PBS buffer) for at least 1 h for each solution.

Resin embedding of siliques
After GUS staining, dehydrated siliques were immersed successively in solutions with an increasing gradient of the Technovit 7100 kit resin (14655, VWR International, Radnor, PA, USA), namely 30%, 50%, 70%, and 100% resin completed with ethanol, for 12-24 hours for each solution. At the last step, samples had to stay in pure resin for at least 5 d. Then, siliques were placed in silicone moulds. A Hardener II solution provided with the kit was added to the resin and poured into the moulds. When they became hard, the blocks of resin were taken out of the moulds and silique sections with a thickness of 10 µm were taken using a rotating microtome (Leica).

Wax embedding of siliques
Wax embedding was used before immunolocalization on slides. After fixation and dehydration, siliques were immersed in a solution containing 50% wax (polyethylene glycol/1-hexadecanol 9:1 w/v) and 50% ethanol for 12 h at 40 °C, and then three times in a solution of 100% wax for 12 h each at 40 °C. Siliques were then placed in silicone moulds and covered with pure wax. When the wax became hard at room temperature, the blocks were taken out of the moulds and silique sections with a thickness of 10 µm were taken using a rotating microtome (Leica).

Immunolocalization of autophagosomes
Immunolocalization was carried out on siliques from plants expressing the Prom-Actin::GFP::ATG8f construct. Siliques were embedded in wax and cut using a microtome, then the sections were placed on Polysine slides (P4981, ThermoFisher Scientific). The slides were then successively immersed twice in ethanol for 10 min, in a solution of 50% ethanol and 50% PBS for 5 min, in a solution of 30% ethanol and 70% PBS for 5 min, and twice in a PBS solution for 5 min. When the wax was removed, the slides were immersed twice in a citrate buffer (4261, Promega) for 1 min at 90 °C and rinsed in PBS buffer for 5 min. The slides were then immersed in a bovine serum albumin solution (PBS, BSA 1%) for 1 h at room temperature. Next, 500 µl of 1/500 diluted primary anti-GFP monoclonal antibody (Clontech) produced in mice was deposited on the slides, which were then incubated overnight at 4 °C in a humid compartment. The slides were rinsed three times in PBS for 10 min before the addition of 500 µl of 1/250 diluted secondary antibody, Alexa Fluor 488 (ThermoFisher Scientific) produced in goat. A coverslip was placed on each slide with a solution composed of 50% glycerol and 50% PBS. Confocal observations were carried out using a TCS SP5 microscope (Leica).

Seed discolouration with chloral hydrate
In order to observe embryo development, siliques from 2-12 DAF were opened and the seeds were discoloured for 2-24 h in a chloral/ glycerol/water solution (8:1:2, w/v/v) (Xu et al., 2016). Discoloured seeds were then placed between a slide and coverslip and were observed using a Nomarski objective (DIC) on an AxioObserver microscope (Zeiss).

Quantification of total proteins
During harvesting, siliques were opened and seeds removed from the valves. Developing seeds were then frozen in liquid nitrogen and lyophilized. Then, 2 mg of dry seeds was ground using a ball mill with 50 µl of an extraction buffer [50 mM pH 7.5 Hepes, 3% SDS, 10% glycerol, 150 mM NaCl, 1× protease inhibitor (cOmplete Protease Inhibitor Cocktail; 11697498001, Roche), 0.1 M proteasome inhibitor MG132 (M7449, Sigma), 20 mM DTT). Ground seeds were then centrifuged (13 000 g, 15 min, room temperature) and 10 µl of the supernatant was used for total protein quantification using the 2D-Quant kit (GE80-6483-56, GE Healthcare, Little Chalfont, UK) according to the manufacturer's protocol.

SDS-PAGE electrophoresis
Protein extraction was performed from 2 mg of dry seeds for each sample according to the protocol used by Deruyffelaere et al. (2015). Protein separation, Coomassie colouration, and western blots were also carried out according to Deruyffelaere et al. (2015).
Polyclonal 12S globulin and 2S albumin sera were kindly provided by Sabine d'Andréa and used as described in d' Andréa et al. (2007). Quantification of protein bands was performed using the ImageJ software (https://imagej.nih.gov/ij/).

Expression of ATG genes increases during silique development
The expression level of the 34 ATG-related genes listed in Supplementary Table S1 was monitored during seed development on siliques (pericarp plus seeds) harvested at 4, 12, 20, and 28 d after flowering (DAF).
The relative expression of almost all the ATG genes increased during seed development and especially during maturation (after 12 DAF; Fig. 1). Only three genes showed decreased expression during silique development, namely ATG1c and ATG13a that belong to the AT1/ATG13 complex, which plays a role in the induction of the autophagy process, and ATG4b that is involved in the ATG8 conjugation system. The majority of the genes participating in the ATG8 conjugation system were expressed early compared with the other functional ATG groups, indicating that this system could be involved in another process. In the ATG9 complex involving ATG2 and ATG18, ATG9 showed optimal expression levels at 12 DAF and ATG18f and ATG18h, as well as ATG7 and ATG8e in the ATG8 complex, optimal expression levels at 20 DAF.
The data suggested that autophagy was clearly induced from the early stage of silique development and that it increased sharply after 20 DAF.

Tissue-specific expression of ATG8f genes in siliques and seed
The expression pattern of several ATG8 genes was then investigated using Prom-ATG8::UIDA fusions in pericarps and in seeds. GUS staining in the seeds at the globular (4 DAF), heart (6 DAF), bent (10 DAF), and mature (12 DAF) stages was more or less intense depending on the promoters and constructs used. Thus, we observed that all the Prom-ATG8::UIDA fusions were expressed mostly in the chalaza from the globular stage onwards, and in the  Table  S1) and normalized to a synthetic reference gene that combines EF-1α and APC2, taking the reference Ct value as √(EF-1α × APC2). Gene expressions were then normalized to the highest value of each gene and are displayed according to the scale bar. Data are the means of the expression values measured on siliques of three plants from three different plant cultures. Genes clustering was performed using XLSTAT. DAF, days after fertilization. embryo at different stages ( Fig. 2a and Supplementary  Fig. S1). The most spectacular GUS staining was observed with the ATG8f promoter, which developed an intense blue signal in embryos of mature green seeds (Fig. 2a, b and Supplementary Fig. S1). In the pericarp, the Prom-ATG8f::UIDA GUS staining was mainly detected in the vasculature of the pericarps (Fig. 2c) and funiculus (Fig. 2d, e). Transgenic plants carrying Prom-ATG8f::GFP constructs confirmed promoter activity in the veins of the pericarp and showed that it was restricted to the phloem companion cells and phloem parenchyma cells (Fig. 2f,  g). Thus, a typical single-file alignment of autofluorescent chloroplasts was detected in companion cells, whereas the chloroplasts were located at the cell periphery of the phloem parenchyma tissue (Cayla et al., 2015).

Visualization of autophagosomes in embryos
Many attempts to visualize autophagosomes in seeds were tried using confocal microscopy on fresh or dry seeds using the Prom-35S::GFP::ATG8, Prom-Actin::GFP::ATG8, and Prom-Ubi::GFP::ATG8 transformants that have been previously used successfully for observations on roots or leaves in our laboratory. Unfortunately, however, GFP fluorescence observations on seed material proved to be difficult and not convincing. Therefore, seed material was embedded and autophagosome structures were observed through GFP immunolocalization performed on the seeds of both WT and atg5 transgenics harbouring the Prom-Actin::GFP::ATG8f construct. Immunolocalization with an anti-GFP antibody and using the Alexa Fluor 488 secondary antibody revealed many fluorescent dots in the WT background (Fig. 3a, c, e) that were absent in atg5 background (Fig. 3b, d, f). The size of the fluorescent structures was less than 1 µm (Fig. 3e), which is in good agreement with the size of autophagosomes or pre-autophagosomal structures. The negative controls with the secondary antibody alone are shown in Supplementary  Fig. S2.
atg5 mutant seeds exhibit abortion and early browning Seed abortion was monitored by counting the viable fullydeveloped seeds in siliques. Because nutrition of the mother plant impacts significantly on metabolism and seed production in autophagy mutants, analyses were performed on plants grown under low-(2 mM) and high-(10 mM) nitrate conditions (see phenotypes of the Col and atg5 rosettes in Supplementary Fig. S3). A strong difference was observed between mutants and the WT for seed viability and seed colour, but no significant differences were detected between the two nitrate regimes (Fig. 4 and Supplementary Fig. S4, showing 10 mM and 2 mM, respectively). Significantly lower seed viability was observed as early as 12 DAF in atg5 relative to Col (21% and 33% of seed abortion in atg5 under high-and low-nitrate conditions, respectively). Low-nitrate conditions also increased seed abortion in Col (from 0 to 6%). In atg5, seed abortion was not systematically observed in all siliques. We also observed that atg5 seeds turned brownish from 16 DAF onwards, while Col seeds remained green longer and turned brown only at the end of their development, i.e. from 20 DAF (Fig. 4b and Supplementary Fig. S4b, showing 10 mM and 2 mM, respectively).
Observation of embryos after chloral hydrate treatment and using Nomarski optics revealed that arrest of development occurred between 8 and 12 DAF, mainly at the 'torpedo' stage (Fig. 5). The atg5 embryo presented stocky cotyledons compared to the wild-type.

atg5 mutant seeds have altered storage protein accumulation
Low seed weight of the atg5 mutant after 20 DAF Only viable seeds were considered and selected for further analyses. There was no significant difference in dry weight (DW) between atg5 and Col seeds during development until after 20 DAF under high-nitrate conditions, and until after 16 DAF under low-nitrate conditions (Fig. 6a, b). At 20 DAF, when grown under low nitrate, the dry weight of atg5 seed was slightly lower than that of Col seed. At maturity, the dry seed (DS) of atg5 was significantly lighter than that of Col seed (Fig. 6a, b) in both nutritional conditions. The single atg7 and the double atg4a-4b mutants also produced dry seed with lower weight than the wild-type (Supplementary Fig. S5).

High and early accumulation of total protein in the atg5 mutant
There was no significant difference in DW between atg5 and Col seeds until 20 DAF, meaning that protein content per seed was a good indicator of protein concentrations at these stages. Higher protein contents were measured in atg5 seeds relative to Col seeds (Fig. 6c, d), and differences were significant under both low and high nitrate, except at 8 and 20 DAF under high nitrate (Fig. 6c). This indicated that protein accumulation was globally higher and occurred earlier in the atg5 mutant than in the WT.

Increased ratio of high/low molecular-weight of total proteins in the atg5 mutant
In order to distinguish the nature of the proteins accumulated in atg5 seeds, protein profiles of dry seeds were examined on Coomassie Blue-stained SDS-PAGE gels ( Fig. 7 and Supplementary Fig. S6). The differences in profiles between atg5 and Col were similar under both high- (Fig. 7) and low-( Supplementary Fig. S6) nitrate conditions, as attested by quantification of the band intensities performed after gel scanning. The results showed that high molecularweight proteins (from 200 to 37 kDa) were significantly more abundant in atg5 than in Col seeds, while in contrast, low molecular-weight proteins, mainly representing globulin and albumin, were significantly less abundant in atg5 than in Col seeds. The single atg7 and the double atg4a-4b mutants presented the same profile in Coomassie gels, with around 30% of high-weight (HW) proteins compared to 20% of HW in the wild-type (Supplementary Fig. S7). The ratio of HW/LW for the sid2-atg5 double-mutant was also higher compared to those of sid2 ( Supplementary Fig. S7).

Low and early 12S globulin content in the atg5 mutant
Antibodies raised against the 12S globulins and 2S albumins were used to quantify the accumulation of the storage proteins during seed development. The kinetics of 12S and 2S accumulation in seeds of atg5 and Col plants, evaluated on western blots, showed that proteins, mostly 12S, accumulated earlier in atg5 seeds than in Col seeds (Fig. 8a). Signals for the 12S, 2S, and 12S-precursors were clearly observed at 16 DAF in atg5 seed extracts, but were only very slightly visible in Col seed extracts (Fig. 8a, b). At 20 DAF, 12S storage proteins appeared less abundant in atg5 seeds than in Col seeds, while 12S-precursors appeared as several distinct bands that were absent from Col seed extracts (Fig. 8a). The ratio of 12S to 12S-precursors increased between 16 and 20 DAF in the Col seeds but remained unchanged in the atg5 seeds (Fig. 8c).
The lower 12S globulin content in atg5 dry seeds relative to Col dry seeds was also observed in the sid2-atg5 doublemutant compared to the sid2 mutant (Fig. 9a, b). Nitrate nutrition had no effect on the accumulation of storage protein, as the difference between Col and atg5 was observed under high-and low-nitrate conditions ( Supplementary Fig.  S8a,b). In addition, the lower 12S globulin content compared to Col was also observed in the atg7 and atg4a-4b mutants ( Supplementary Fig. S8c,d).

Discussion
In this study, we show that autophagy gene expression increases during seed development, especially in the embryo, and during the maturation phase that corresponds to the formation of the oil and protein bodies in the embryo (Baud et al., 2008). At earlier developmental stages, all the ATG8 genes were moderately expressed in the embryo and more strongly in the chalaza, which corresponds to the location  where amino acids are released from the phloem, at the end of the funiculus vasculature, through the UmamiT efflux systems (Müller et al., 2015;Tegeder and Masclaux-Daubresse, 2018). Interestingly, the Prom-ATG8f::UIDA fusion showed a strong expression in the phloem companion cells of the pericarp and in the vasculature of the funiculus. As ATG gene expression is a preliminary to autophagy activity (Bernard et al., 2015), finding ATG8f promoter activity in the pericarp, funiculus, and chalaza maternal tissues suggests that autophagy plays a role in the nutrient import into the seeds, for development as well as for nutrient storage (Zhang et al., 2007). The strong expression in the zygote embryo at maturity also suggests a role of autophagy in the accumulation of protein and/or lipid reserves.
Although it was not possible to investigate the role of autophagy separately in maternal and zygote tissues, the study of seed features during development in the atg5 mutant, and the use of different nitrate regimes to grow the plants, provided some clues to this question. Following seed formation on plants grown under low-and high-nitrate conditions, we observed that in both situations early (before 12 DAF) abortion in atg5 was quite high. This phenotype was not nitratedependent but it may be explained by a defect in the capacity of mother tissues to drive nutrients to the seeds. We then presume that autophagy defects in phloem companion cells impaired seed nutrient capture, thus resulting in the seed abortion.
Although viable seeds could be obtained in atg5, they were different from the WT. Differences were visible early during development and, as a first symptom, we detected that chlorophyll decay and seed browning appeared earlier, as if seed development was accelerated. Development acceleration was confirmed by the earlier protein storage deposition in atg5 viable seeds.
As the atg5 mutant, as well as all the atg mutants, exhibited an early senescence that could interfere with the autophagy process, the sid2-atg5 double-mutant, was used to obtain plants with longer life. The phenotype of this mutant, i.e. low 12S storage protein production, was comparable to that of atg5, showing that the observed results were due to the lack of autophagy and not to an early senescence. Fig. 7. Protein profiles of wild-type Col and atg5 dry seeds produced under high-nitrate conditions. Total proteins extracted from dry seeds were separated on SDS-PAGE gels and stained with Coomassie Blue (a). The same quantity of protein was loaded in each lane. The graphs in (b) represent the average of the band intensities (AU, arbitrary units) measured on three biological repeats. Significant differences between Col and atg5 are indicated by the red dots on the left (t-test, P<0.05). Differences between Col and atg5 protein contents are presented in orange when the content was higher in atg5 than in Col, and in blue when the content was higher in Col than in atg5. The orange and blue colours correspond to the high-weight (HW, >37 kDa) and low-weight (LW, <37 kDa) proteins, respectively. The sum of the band intensities in the Col and atg5 protein extracts is indicated in (c) as the loading control. Similar results were obtained on seeds from plants grown with low nitrate and are presented in Supplementary Fig. S6.
The atg5 viable seeds did not seem to suffer from nitrogen starvation as might have been expected. As reported by Guiboileau et al. (2012), we indeed found that the percentage N content in seeds was higher in atg5 than in the WT (data not shown), and although seed DW was similar in atg5 and the WT until 16 DAF, seed protein concentration was higher in atg5 relative to the WT. This feature was nitrate independent. All these results then suggested that the abortion of seeds certainly reduced the overall sink N demand and made the mother resources sufficient to feed the atg5 viable seeds properly.
The higher protein content in atg5 seeds was not paralleled by an increase in 12S and 2S storage proteins. The reduced production of storage proteins observed in all the atg mutants tested was mostly manifested for 12S proteins and to a lesser extent for the 2S. We showed that the higher protein content of the atg5 mutant seed was probably correlated with higher storage protein precursors. This observation then questions the role of autophagy in the processing of seed storage proteins. Considering all the pathways of storage protein deposition, some key steps could be controlled by the autophagy process. Precursors of 12S globulins are synthesized in the endoplasmic reticulum and Golgi from proforms and transported to the protein storage vacuoles (PSVs) via multi-vesicular bodies (MVBs). Then, vacuolar processing enzymes (VPEs, mainly VPEβ) convert the precursors into seed storage protein (Shimada et al., 2003;Otegui et al., 2006). It has been reported that atg mutants suffer from oxidative stress (Masclaux-Daubresse et al., 2014) and hence the role of reactive oxygen species throughout the course of storage protein processing should be considered. The accumulation of oxidized proteins and organelle components could potentially affect (i) the maturation of the proforms into 12S precursors, (ii) the transport of the precursors in the MVBs, and/or to the PSVs, and (iii) the processing of precursors into mature storage protein.
Whether autophagy plays a role in storage protein processing is unclear, but this has been suggested in maize by Reyes Fig. 8. Kinetics of 12S and 2S protein accumulation in wild-type Col and the atg5 mutant during seed development. Seeds were obtained from plants grown under high-nitrate conditions. Western blots using (a) 12S and (b) 2S storage protein antibodies are shown. Total proteins extracted from the same number of seeds were loaded in each lane. Ratios of the 12S proteins to 12S precursors (c) were determined using ImageJ software. Means (±SD) are shown (n=3). DAF, days after fertilization; P12S, precursors of 12S proteins (indicated by arrows). et al. (2011). These authors found that zeins (a class of prolamine proteins) were delivered to aleurone PSVs through pre-vacuolar compartments consisting of multi-layered membranes that suggested autophagosome structures. However, the absence of ATG8 decoration in the membranes of these structures led them to the conclusion that they were not typical autophagosomes. As alterations in these structures has not been verified in the recently obtained maize autophagy mutants (Li et al., 2015), it cannot be said whether they are actually macroautophagy-dependent.
A recent publication from Gao et al. (2015) identified the FREE1-ESCRT component as required for MVB biogenesis and protein transport to the PSV. The authors showed a strong defect in storage protein processing in the free1 mutant, leading to the accumulation of precursors of storage proteins just as we observed in atg5 seeds. They also showed that the loss of function of FREE1 disturbed the autophagy pathway, promoting autophagosome-MVB fusion events and decreasing autophagosome fusion to the vacuole.
Our results showed that, like the free1 mutation, the atg5 mutation affects storage protein deposition, and thus that both the FREE1-dependent ESCRT and the autophagy mechanisms are needed for seed storage protein maturation. How these two pathways contribute together to PSV formation needs further investigation. Whether autophagy contributes to storage protein processing that facilitates VPE trafficking to the MVBs or PSVs is a possibility that needs to be considered. Nevertheless, our study describes a new role for autophagy that is distinct from the usual degradation function assigned to this intracellular process up to now.

Supplementary data
Supplementary data are available at JXB online. Fig. S1. Localization of the expression of ATG8c, e, f, and g during embryo development in seeds. Fig. S2. Autophagosomes observed in embryos. Fig. S3. Development of Col and atg5 rosettes. Fig. S4. Browning phenotype of atg5 seeds from plants grown under low-nitrate conditions. Fig. S5. Dry weight of seeds from Col and mutant plants. Fig. S6. Protein profiles of dry seeds of Col and atg5 produced under low-nitrate conditions. Fig. S7. Protein profiles of dry seeds of Col and atg mutants produced under low-nitrate conditions. Fig. S8. 12S globulin and 2S albumin protein contents in dry seeds of Col and atg mutants. Table S1. Primers used for RT-qPCR analyses and cloning.