Legume adaptation to sulfur deficiency revealed by comparing nutrient allocation and seed traits in Medicago truncatula



Reductions in sulfur dioxide emissions and the use of sulfur-free mineral fertilizers are decreasing soil sulfur levels and threaten the adequate fertilization of most crops. To provide knowledge regarding legume adaptation to sulfur restriction, we subjected Medicago truncatula, a model legume species, to sulfur deficiency at various developmental stages, and compared the yield, nutrient allocation and seed traits. This comparative analysis revealed that sulfur deficiency at the mid-vegetative stage decreased yield and altered the allocation of nitrogen and carbon to seeds, leading to reduced levels of major oligosaccharides in mature seeds, whose germination was dramatically affected. In contrast, during the reproductive period, sulfur deficiency had little influence on yield and nutrient allocation, but the seeds germinated slowly and were characterized by low levels of a biotinylated protein, a putative indicator of germination vigor that has not been previously related to sulfur nutrition. Significantly, plants deprived of sulfur at an intermediary stage (flowering) adapted well by remobilizing nutrients from source organs to seeds, ensuring adequate quantities of carbon and nitrogen in seeds. This efficient remobilization of photosynthates may be explained by vacuolar sulfate efflux to maintain leaf metabolism throughout reproductive growth, as suggested by transcript and metabolite profiling. The seeds from these plants, deprived of sulfur at the floral transition, contained normal levels of major oligosaccharides but their germination was delayed, consistent with low levels of sucrose and the glycolytic enzymes required to restart seed metabolism during imbibition. Overall, our findings provide an integrative view of the legume response to sulfur deficiency.


Sulfur (S) is a macronutrient that is acquired by roots in the form of inorganic sulfate. Sulfate transport and reduction are essential components of plant metabolism, leading to the synthesis of cysteine (Cys) and methionine (Met), and of numerous metabolites derived from S–containing amino acids, such as cofactors and vitamins (e.g. biotin and thiamine). Sulfate is a precursor in the synthesis of S–adenosylmethionine (AdoMet), the universal methyl group donor (Ravanel et al., 2008), and of glutathione, which is involved in maintenance of the redox balance and mitigation of oxidative stress in response to environmental changes (Foyer et al., 2001). As a component of S–containing amino acids, sulfur regulates protein structure and metabolism, and determines the protein value of plant foods for nutrition of humans and animals that are unable to synthesize Met and are dependent on dietary sources of this amino acid. Sulfur assimilation is inter- connected with carbon (C) and nitrogen (N) metabolism (Kopriva et al., 2002; Hawkesford and De Kok, 2006). In particular, there are interactions between the amino acid pathways, balancing the flux of S, C and N, as reported for the aspartate pathways leading to synthesis of Met and two other essential amino acids (threonine and lysine) (Jander and Joshi, 2010).

In recent decades, the increased production of agricultural crops together with the reduction of sulfur dioxide emission and changes in agronomic practices, such as use of S-free fertilizers and pesticides (Vestreng et al., 2007), has led to S deficiencies in agricultural land in North America, Asia and Western Europe (Janzen and Bettany, 1984; McGrath and Zhao, 1995; Zhao et al., 1999a; Khurana et al., 2008). Moreover, due to its high solubility in water, sulfate is readily lost by leaching from soils. Many studies have reported the agronomic consequences of lack of sulfate availability in soils. It significantly affects plant growth, decreases yields (Janzen and Bettany, 1984; Zhao et al., 1999a,b; Hawkesford, 2000) and reduces plant resistance to pathogens and abiotic stresses (Rausch and Wachter, 2005; Kruse et al., 2007) as fewer S-containing defense compounds may be synthesized. In legumes, previous studies showed that S deficiency decreases N assimilation and fixation (Scherer and Lange, 1996; Sexton et al., 1998; Zhao et al., 1999b; Scherer et al., 2006; Varin et al., 2010) and modifies seed protein composition (Blagrove et al., 1976; Chandler et al., 1983, 1984; Gayler and Sykes, 1985; Spencer et al., 1990) by decreasing the abundance of storage proteins with the highest content of S–containing amino acid i.e. albumins and legumin-type globulins) while increasing the level of S–poor globulins (i.e. vicilins).

Because S deficiency represents an agronomic issue for crops, notably for the quality of legume products (i.e. seeds and forage), it is necessary to provide S fertilizers to meet crop S needs together with improving the efficiency of S use. To achieve this goal, increased knowledge of how plants respond to S deficiency is required. In the last decade, the plant response to S limitation has been studied by transcript, metabolite or protein profiling in Arabidopsis (Hirai et al., 2003, 2005; Nikiforova et al., 2003, 2005; Higashi et al., 2006; Hoefgen and Nikiforova, 2008) and cereals (Howarth et al., 2008; Lunde et al., 2008), and through dynamic studies of S and N mobilization in rapeseed (Brassica napus L.) (Dubousset et al., 2009, 2010; Abdallah et al., 2010, 2011). While the plant response to reduced S availability has been widely documented for the Brassica family, whose seeds contain large amounts of glucosinolates (i.e. S–rich compounds), it remains poorly studied in legumes, whose seeds are deficient in glucosinolates but rich in proteins (approximately 23–40%). Because seeds of legumes, such as pea (Pisum sativum L.) and soybean (Glycine max L.), are used for animal and human nutrition, identifying the mechanisms that control seed production and composition under challenging environmental conditions is important to improve or maintain seed yield and quality. To provide knowledge of the legume response to S deficiency, we used Medicago truncatula, whose seed composition resembles that of grain legumes (Djemel et al., 2005; Gallardo et al., 2007). In this paper, after exploring the phenotypic response of M. truncatula to S deficiency applied at various growth stages, we compared allocation patterns of S, C and N between plant organs with seed composition and germination performance. We reveal an impressive plasticity of the M. truncatula response to S deficiency that varied according to the timing of S stress and influenced seed composition and germination. Molecular processes were identified in leaves that may help maintain high remobilization rates to seeds in response to S deficiency, and other processes were identified in seeds that may explain their low germination performance. This combined analysis of plant and seed tissues represents a crucial step towards understanding nutrient management by plants under S-stress conditions and determining sulfate requirements in legumes for high yield and seed quality.


Phenotypic response of M. truncatula to S deficiency

To examine the phenotypic response of M. truncatula to various S-deficiency periods, sulfate was depleted at the mid-vegetative stage characterized by appearance of tertiary branches (S1, 844°C days, thermal time), at flowering (S2, 1111°C days) and during reproductive growth (S3, 1597°C days, i.e. seed filling of the first pods) (Figure 1a). The relative chlorophyll content of the last fully expanded leaves was monitored during development in these plants and in control plants to which sulfate was supplied until harvest (Figure 1b). It decreased steadily in leaves of S1 and S2 plants compared with control plants: by 50 and 40%, respectively, after 2000°C days. In contrast, S deficiency at the S3 stage did not affect leaf chlorophyll content, probably because of abundant S reserves in senescent leaves that may be remobilized to newly formed leaves in the later stages of growth. As a good correlation between chlorophyll content and photosynthetic characteristics had previously been observed (Griffin et al., 2004; Kato et al., 2004; Kumagai et al., 2009), our results suggest S deficiency at flowering or earlier reduced photosynthesis in young leaves, whereas S starvation during reproductive growth had no marked influence on this parameter.

Figure 1.

Effect of S deficiency on leaf chlorophyll content. (a) Developmental stages S1, S2 and S3 at which S was depleted. (b) Mean leaf chlorophyll content throughout plant development (polynomial curves) for S1, S2, S3 and control plants (Co). The arrows indicate the start of S deficiency. Values are means ± SE (= 8). °C day, thermal time.

Dry weight measurements of plant parts (Table 1) revealed that S deficiency starting at the mid-vegetative stage (S1) dramatically affects forage and seed yields. Importantly, we observed that S deficiency applied at flowering (S2) reduced forage yield but not seed yield, resulting in a high seed harvest index (i.e. seed yield to total biomass ratio). Significantly, a sharp increase in one-seed and one-pod weights but an unchanged number of seeds per pod were observed for S2 plants. This indicates the occurrence of adaptive mechanisms to increase seed weight and prevent seed yield losses when sulfate is deficient during flowering. Contrasting results were obtained for S3 plants, with a reduced number of seeds per pod and no change in one-seed weight compared to the control, indicating that plants in the reproductive phase lost the ability to maintain a high number of seeds per pod under S-stress conditions. There were also some interesting trends in seed yield (although not statistically significant due to use of the stringent Tukey test), being higher for S2 plants and lower for S3 plants.

Table 1. Effect of S deficiency on phenotypic characteristics at harvest (2246°C days)
Phenotypic characteristics at harvestControlS1S2S3
MeanSDMeanSD P S1/Co ratioMeanSD P S2/Co ratioMeanSD P S3/Co ratio
  1. Phenotypic characteristics measured after drying the tissues at 80°C, except dry mature seeds. An analysis of variance followed by Tukey's post hoc HSD test was performed to determine significant variations between plants subjected to S deficiency at various growth stages (S1, S2 and S3) and control plants (Co) (see Figure 1a). The ratio of values for S-starved plants versus control plants is indicated. The harvest index corresponds to the ratio of seed yield to total biomass. The arrows represent significant decreases and increases in yield or quality parameters, respectively, compared to control. SD, standard deviation.

Plant characteristics
Onset of flowering (thermal time in °C days)11338810921410.9020.961042690.4820.921174980.9041.04
Root weight per plant (g)
Forage yield per plant (g)658.12250.0000.34 image_n/tpj12350-gra-0001.png445.00.0020.68 image_n/tpj12350-gra-0002.png66140.9931.02
Pod weight per plant (g) image_n/tpj12350-gra-0001.png13.65.00.0051.86 image_n/tpj12350-gra-0003.png4.702.00.3650.64
Pod number per plant16223177.50.0000.10 image_n/tpj12350-gra-0001.png151400.9460.93111510.0820.69
Seed yield per plant (g) image_n/tpj12350-gra-0001.png2.71.10.0761.590.700.30.0770.41
Total plant biomass (g)819316.40.0000.38 image_n/tpj12350-gra-0001.png65100.0820.8079160.9720.98
Harvest index × 1001. image_n/tpj12350-gra-0003.png1.00.30.2970.53
Pod and seed characteristics
Seed number per pod8. image_n/tpj12350-gra-0001.png
One-pod weight with seeds (mg)557.1112140.0002.04 image_n/tpj12350-gra-0003.png104150.0001.89 image_n/tpj12350-gra-0004.png506.30.8520.91
One-pod weight without seeds (mg)334.893110.0002.82 image_n/tpj12350-gra-0003.png63120.0001.91 image_n/tpj12350-gra-0003.png235.70.2040.70
One-seed weight (mg) image_n/tpj12350-gra-0003.png2.40.40.7970.95

Differences in nutrient allocation patterns

To relate the phenotypic characteristics to nutrient allocation patterns, S, C, and N contents (i.e. mg nutrient per mg dry weight) in roots, straw, empty pods and mature seeds were measured (Figure S1a–d), and then the quantity of S, C and N (in mg) in the total dry biomass was calculated for each organ by multiplying the S, C and N contents by the total biomass at harvest (mg dry weight) (Figure 2). We also studied nutrient partitioning between tissues as a percentage relative to the total quantity of each nutrient in the whole plant. These data are included in Figure 2 using a color scale from white/yellow (low percentage of nutrient in the total dry biomass) to brown/black (high percentage of nutrient in the total dry biomass). The results revealed a reduced amount of S in all compartments of S1 plants and marked differences in the partitioning of assimilates between straw and roots compared to the control. While S was predominantly allocated to straw under non-limiting conditions (Figure 2d), it was mostly allocated to roots of S1 plants (Figure 2e). This was accompanied by a significant increase in the proportion of C and N allocated to roots. In contrast, there was a reduced quantity of S in roots of S3 plants and a trend towards a decreased quantity of S, C and N in mature seeds and pods (Figure 2b,c).

Figure 2.

Effect of S deficiency on S, C and N partitioning between plant parts. S, C and N quantities (mg) in the entire plant (a) and in tissues collected at 2200°C days, i.e. mature seeds (b), empty mature pods (c), straw (d) and roots (e). Values are mean absolute quantities and standard deviation (= 6). For each condition, the S, C, and N proportion (%) within the plant is shown using a color scale. Bars with different letters on the top represent values that are statistically significantly different at < 0.05. Asterisks indicate statistically significantly differences compared with the control (Co) plants (**< 0.01, *< 0.05).

In agreement with the hypothesis that adaptive mechanisms occur to avoid seed yield losses when S deficiency starts at flowering, we found a significant increase in C and N allocation to pods of S2 plants, resulting in doubling of C and N quantities in this compartment compared to control plants (Figure 2c). High quantities of C and N were also provided to seeds of S2 plants, at least twofold higher than S1 and S3 plants and similar to the control (Figure 2b). This reflects an activation of reserve mobilization to developing S2 pods that may help to sustain seed production (number of seeds per pod and yield, Table 1) and maintain C and N quantities in seeds. Moreover, there was a significant increase (by 85% compared to the control) in the proportion of S allocated to the pods and seeds of S2 plants, whereas the proportion of S decreased in straw and roots (Figure 2). This demonstrates the importance of S remobilization from source to sink organs for seed production under S deficiency.

We further investigated whether differences in nutrient allocation patterns influenced S, C and N contents in each tissue (Figure S1a–d). S content decreased in all compartments of S1 and S2 plants, especially roots and straw. In contrast, N content increased in the roots of S1 and S2 plants and the straw of S1 plants, indicating a strong imbalance of N/S in roots and vegetative parts. With regard to C content, the only significant difference was a decrease in the pods of S1 plants. The S3 plants did not vary in terms of C and N content. The only significant change was for S content, which decreased in S3 pods. S, C and N quantities per seed were also estimated (Figure S1e), revealing a reduced quantity of S in each S1 and S2 seed and a significant reduction of C and N in each S1 seed.

Sulfate remobilization and reduction are activated in leaves of S2 plants

The data presented above, demonstrating that S is remobilized from source organs to seeds in S2 plants, prompted us to measure the level of transcripts and metabolites related to sulfate reduction, assimilation and remobilization in leaves and developing seeds of S2 and control plants. To avoid any heterogeneity in leaf development for these analyses, flowers of primary branches were tagged on the day of opening, and then developing seeds and leaves at the base of the tagged pods were collected 13 days after flowering (i.e. 14 days after pollination, DAP; the start of seed filling, Gallardo et al., 2007). One gene selected for expression profiling by quantitative RT–PCR encodes a sulfate transporter (SULTR) sharing 89% sequence similarities with AtSULTR4.1, which is involved in remobilization of sulfate from the vacuole to the cytosol in Arabidopsis (Kataoka et al., 2004). This gene, with accession number Medtr7g022870 (M. truncatula genome version 3.5, Young et al., 2011), was previously named MtSULTR4.1 (Casieri et al., 2012). In contrast to Arabidopsis, whose genome contains two SULTR4 genes, there was only one such gene in the available genomic sequences of M. truncatula (Young et al., 2011), which suggests the existence of a single-copy gene for the sulfate transporter in this species. Eleven other genes were selected for expression profiling that encode six enzymes catalyzing sulfate reduction or assimilation (Figure 3 and Figure S2). Their sequences were retrived from the M. truncatula genome version 3.5 and/or from the M. truncatula Gene Index (http://compbio.dfci.harvard.edu/tgi/, see Table S3).

Figure 3.

Effect of S deficiency at flowering on the relative levels of transcripts, sulfate and metabolites related to S remobilization and metabolism in leaves. The sequences were retrieved from the M. truncatula genome version 3.5 and/or from the M. truncatula Gene Index (http://compbio.dfci.harvard.edu/tgi/, see Table S3 for accession numbers, different letters indicate different isoforms). The relative levels of transcripts (black bars) and metabolites (gray bars) are shown relative to the levels in control plants, which were set to 1. Values are means ± standard error of at least three biological replicates. Asterisks indicate statistically significantly differences between S2 and control plants (**< 0.01, *< 0.05). APS, adenosine 5′–phosphosulfate; PAPS, 3′–phosphoadenosine-5′–phosphosulfate; OAS, O–acetylserine; AdoMet, S–adenosylmethionine; GSH, reduced glutathione; GSSG, oxidized glutathione.

S2 leaves showed an up-regulation of gene expression for 5′–phosphosulfate reductase, sulfite reductase and a glutathione synthetase compared to the control (Figure 3), suggesting a need for sulfate reduction and assimilation in reproductive leaves of S2 plants. This was associated with a higher level of the Cys precursor O–acetylserine, which has been previously shown to increase in Arabidopsis seedlings subjected to S deficiency (Nikiforova et al., 2005), and a lower amount of reduced glutathione (GSH) in S2 leaves (Figure 3). However, S flow through the methionine metabolic pathways is likely to be maintained as there were no significant variations in the levels of Met or AdoMet between S2 and control leaves. The level of oxidized glutathione (GSSG) was also maintained in S2 leaves, which contrasts with the low level of GSH and suggests predominant GSH oxidation in response to S deficiency that may contribute to control of oxidative stress in S2 leaves. Moreover, there may be an increased loading of GSH into the phloem of S2 plants for transport to seeds, whose development fully depends on S from source organs. The unchanged level of GSH between S2 and control seeds reinforces this hypothesis (Figure S2). Higher expression of MtSULTR4;1 (threefold) was observed in S2 leaves (Figure 3), suggesting this transporter may provide sulfate to support the high demand for sulfate reduction and/or transport to seeds. Consistently, there was a significant decrease in the sulfate pool of S2 leaves compared to the control (Figure 3).

To obtain further information regarding allocation of sulfate to S2 seeds and its assimilation within the seed, the sulfate content of 14 DAP and mature seeds from S2 and control plants was determined, and the proportion of S accumulated in seeds in the form of sulfate relative to total S was calculated as described by Zuber et al. (2010a) (Figures S3a–c). Sulfate content was reduced by 70% (Figures S2 and S3b) in 14 DAP S2 seeds. These data, together with the transcript and metabolite profiles in Figure S2, are indicative of reduced allocation of sulfate to S2 seeds during the first stages of seed development (up to 14 DAP). Indeed, although no significant variation in expression of genes related to S assimilation was observed in S2 seeds at 14 DAP, there were lower levels of free Met and GSSG compared to the control (Figure S2). This was associated with increased levels of serine/O–acetylserine and glycine, which cannot be used respectively for the synthesis of Cys and glutathione when S flow through the sulfate reduction pathway is reduced (Figure S2). Significantly, the sulfate content and the proportion of S in the form of sulfate decreased by 69 and 64%, respectively, in S2 seeds between 14 DAP and mature stages, but remained stable in control seeds (Figure S3b,c). This shows that most of the sulfate entering the S2 seeds is assimilated during subsequent stages of seed filling and maturation. The levels of other anions (nitrate, phosphate and chloride) in mature S2 and control seeds showed no significant change, except for chloride, whose level increased in S2 seeds (Figure S3d), probably for osmotic pressure adjustment (Marschner, 1995).

Changes in seed proteins and sugars in response to S deficiency

To study the impact of the various nutrient allocation strategies (Figure 2 and Figure S1) on the seed protein complement, an important determinant of seed quality, a proteome analysis of mature seeds from the various plant samples was performed. It allowed quantification of 389 well-resolved spots (Figure S4), and revealed 192 spots whose abundance varied in seeds in response to one or several S-deficiency periods (Table S1). S restriction applied at flowering or earlier substantially modified seed protein composition: 164 and 132 spots varied in S1 and S2 seeds, respectively, compared to control seeds. In contrast, S deficiency during reproductive growth had little influence on the seed proteome, with only eight spots varying between S3 and control seeds. Interestingly, while some protein changes were common to the S1 and S2 seeds, others were specific to one S-deficiency period, as indicated in Figure S5, which shows the number of proteins up- and down-regulated in response to the S-deficiency periods. Eighty-two of these variable spots were annotated by using proteome reference maps of late maturing seeds (24 spots identified by Gallardo et al., 2003, 2007) or following liquid chromatography coupled to tandem mass spectrometry (Table S2). Of these 82 spots, 58 corresponded to globulin-type storage proteins. As expected, many globulins that increase in response to S deficiency (vicilins and convicilins) contain few S–containing amino acids (approximately 0.6% of their sequence), whereas the globulins whose levels decreased (legumins) contained a higher proportion of S–containing amino acids (approximately 1.5%) (Table S1). The S1 seeds were even more affected in terms of globulin composition than the S2 seeds. The proteome data (clickable map, spot characteristics and annotation) have been deposited at http://www.thelegumeportal.net/medicago-seed-proteome/.

Our study also revealed 24 non-storage proteins that are regulated by at least one S-deficiency period (four up-regulated by S stress and 20 down-regulated by S stress), some of which had not previously been reported to be regulated by S deficiency in seeds (Table S1). According to the KEGG (Kyoto Encyclopedia of Genes and Genomes) classification, these proteins are related to: (i) glycolysis (six spots down-regulated), (ii) folding/stability (five spots down-regulated), (iii) stress responses/detoxification with one and two spots up- and down-regulated, respectively, (iv) metabolism (three spots down-regulated), (v) seed maturation (three down- and three up-regulated spots), and (vi) hydrolase (one spot down-regulated). While the level of a glutathione-S–transferase (spot 123) decreased specifically in S1 seeds, the level of a lipoxygenase (spot 119) was enhanced. Also, the levels of a BiP protein (luminal binding protein, spot 221) and a seed biotin-containing protein (SBP65, spot 260) decreased in S2 and S3 seeds, respectively.

Interestingly, levels of all proteins related to glycolysis or glucose signaling decreased in S2 seeds (Figure 4a). This prompted us to measure the concentrations of seed sugars (Figure 4b). As previously observed by Djemel et al. (2005), only trace amounts of glucose and fructose were detected in mature seeds, with no reliable variations between samples. Mature S1 seeds appeared to be the most affected in terms of the level of major sugars, with a decrease of up to 55% in the level of sucrose and raffinose family oligosaccharides (RFOs), i.e. stachyose and verbascose. The levels of sucrose and verbascose also decreased in S2 seeds. In contrast, the seed sugar content of S3 seeds was similar to that of the control. These data indicate an alteration of seed sugar transport and/or metabolism when S restriction is applied at flowering or before, which may influence seed vigor, as sucrose and RFOs are sources of energy during germination.

Figure 4.

Effect of S deficiency on the level of glycolysis-related proteins and sugars in mature seeds. (a) Proteins related to glycolysis whose level varied in response to S deficiency (S1 and/or S2 compared to control, Figure 1). (b) Sucrose and RFO contents (mg g−1 dry weight ± SD). Asterisks indicate statistically significantly differences compared with the control (Co) (***< 0.001; **< 0.01, *< 0.05). In most cases, the arrows represent multiple steps. Enzyme names: (1) UDP-glucose pyrophosphorylase, spot 141; (2) fructose BiP aldolase, spot 299; (3) glyceraldehyde 3–phosphate dehydrogenase, spot 166; (4) enolase, spots 181 (top) and 339 (down); (5) aldose reductase, spot 345; (6) short-chain dehydrogenase/reductase SDR1, spot 187; (7) 3–phosphoglycerate dehydrogenase, spots 319; (8) ketol-acid reductoisomerase, spot 289. ABA, abscisic acid.

Germination speed is affected by S deficiency during plant growth

The germination characteristics of seeds from the various plant samples revealed that the onset of germination was delayed for S1 and S2 seeds (Figure 5). Indeed, the times required to achieve 1 and 10% germination were at least twice as long as for the control. This delay was coupled to a decreased germination speed for S1 seeds as their final germination percentage was reached later compared to the control (6.5 versus 5 days). Although there was no delay in the start of germination for S3 seeds, the final percentage of germinated seeds was lower than for the control (Figure 5a) and the germination speed was significantly slower: 9.2 days to reach the maximum percentage of germination compared to 5 days for the control (Figure 5b). The ungerminated S3 seeds were subjected to a tetrazolium test. All embryos stained red, confirming their viability.

Figure 5.

Effect of S deficiency on seed germination. (a) Kinetics of germination for control (Co), S1, S2 and S3 seeds. (b) Number of days to achieve 1%, 10%, 50% and the maximum percentage germination (Gmax).


The increased extent of soil S deficiency means that sulfate fertilization is essential for satisfactory crop production (Khurana et al., 2008). As sulfate is readily lost in soils by leaching, there is increasing interest in studying the plant response to S deficiency to identify plant markers of S needs and to improve S uptake and use efficiency. Although this response has been well documented for crucifer species, little is known of the legume response to S deficiency. Here, we used M. truncatula to provide insights into nutrient management by a legume plant in S-deficient environments and its consequence on seed characteristics. Figure 6 presents an integrated view of the main findings and the hypotheses raised.

Figure 6.

Integrated view of the main findings and the hypotheses raised (asterisks). Favorable and unfavorable plant/seed characteristics are shown in red and green, respectively.

Adaptation of M. truncatula to S deficiency depends on the ability of source organs to remobilize S, C and N to seeds

An interesting observation was the various strategies used by M. truncatula to survive and produce seeds depending on the growth stage at which S deficiency occurs. As has been well documented for other species (Hawkesford, 2000), early S deficiency affected leaf chlorophyll content and yield parameters. It also increased the N/S ratio (up to 45-fold in leaves), which has been proposed to be a good indicator of S deficiency in several species, including wheat (Triticum aestivum) (Zhao et al., 1995). In contrast, when applied at flowering, S deficiency led to larger pods and seeds without affecting the number of seeds per pod, thus preventing seed yield losses (Table 1). This differential phenotypic response revealed adaptive processes modulating seed weight, yield and quality under S deficiency. Such processes undoubtedly include reorientation of the metabolism towards accelerating seed production, as demonstrated in Arabidopsis (Hoefgen and Nikiforova, 2008). Our thorough analysis of S, C and N partitioning between the compartments of S-deprived and control plants allowed us to obtain further insight into these processes.

Impressive differences in nutrient partitioning were revealed between the various plants shown in Figure 1(a). In particular, the results demonstrated substantial remobilization of nutrients from source to sink organs in S2 plants (Figure 2). Indeed, although these plants were deprived of S at flowering, they were able to allocate S, C and N to their pods and seeds in quantities similar to the control (Figure 2 and Figure S1). Conversely, S1 plants were unable to allocate S, C and N to seeds, and favored root growth at the expense of the shoot (Table 1). Previous studies have demonstrated the role of C allocation to seeds in maintaining the embryo cell division capacity and storage-associated cell differentiation (Wobus and Weber, 1999). Moreover, it has been shown that the flow of sucrose to developing pea seeds increased in response to N deficiency and correlated with a high seed growth rate (i.e. cell division in the cotyledons, Munier-Jolain and Salon, 2003). Our data suggest that S2 plants used the same strategy of increasing the flux of nutrients from source organs towards developing pods to produce well-filled seeds (Figure 2). This adaptive response occurred specifically when S was depleted at the floral transition, which is characterized by an increased level of sucrose in the phloem sap (Corbesier et al., 1998). Hence, the adaptive response possibly involves sugar-signaling mechanisms, as previously observed in response to other abiotic stresses (Rolland et al., 2006).

The decreased chlorophyll content of S2 leaves during reproductive growth (Figure 1b) is characteristic of foliar senescence and suggests reduced photosynthetic rates. As observed in response to N deficiency, S2 plants were characterized by a reduced number of secondary branches, leading to low foliage yields (Table 1). This may explain, at least in part, how C was liberated for vigorous seed production and growth rate on the reproductive branches despite reduced photosynthesis. Moreover, the CO2-fixing potential of pod walls may compensate for the lower photosynthetic activities of leaves (Atkins et al., 1977). Possible adaptation of the leaf photosynthetic apparatus may also occur, as observed in response to N limitation in maize (Zea mays) leaves, which were more efficient in terms of CO2 assimilated per photosynthetic unit despite reduced chlorophyll content per unit leaf mass (Khamis et al., 1990).

Understanding the molecular mechanisms underlying the adaptive response

Understanding the processes enabling the plant to maintain its metabolism and to remobilize stored nutrients is of great interest for improving plant tolerance to sulfate deficiency. We observed that expression of MtSULTR4;1, the only putative vacuolar transporter of sulfate in M. truncatula, increased significantly, by threefold, in reproductive S2 leaves (Figure 3). Putative vacuolar sulfate transporters have also been shown to be up-regulated in the leaves of rapeseed and wheat in response to S deficiency (Dubousset et al., 2009; Shinmachi et al., 2010), and thus are good candidates for indicators of S needs in various species. These transporters have been shown to play a role in sulfate efflux from the vacuoles in roots to control the flux of sulfate channeled toward the xylem vessels (Kataoka et al., 2004). However, their roles in leaves remain unknown. MtSULTR4;1 transcripts were at least as abundant in leaves as in roots under S-starved conditions, suggesting an important function in leaves (Sieh et al., 2013).

In the present study, we observed a concomitant up-regulation in S2 leaves of genes encoding enzymes of sulfate reduction (Figure 3), suggesting that MtSULTR4;1 is a key player in remobilizing the pool of sulfate stored in leaf vacuoles for reduction and synthesis of S compounds and/or for sulfate transport to seeds. Its Arabidopsis homolog, AtSULTR4;1, is up-regulated in roots, along with genes of glutathione metabolism, in response to cadmium stress, which is known to induce an oxidative stress response (Herbette et al., 2006). This suggests a role for SULTR4;1 in redox homeostasis, as also proposed by Zuber et al. (2010b). MtSULTR4;1 may exert a similar function in leaves, which are prone to oxidative stress during reproductive growth, especially in challenging environments, and require accumulation of protective S compounds, such as glutathione, to control the level of free radicals (Zimmermann and Zentgraf, 2005; D'Hooghe et al., 2013). In support of this hypothesis, a glutathione synthetase gene was strongly up-regulated (12-fold) and the level of GSSG was maintained in reproductive S2 leaves (Figure 3). Glutathione functions as an antioxidant by scavenging ROS, resulting in oxidation of GSH to GSSG, which may contribute to maintenance of leaf metabolism and high remobilization efficiency for seed production. Met and AdoMet are other S metabolites whose levels were maintained in S2 leaves (Figure 3) and that may be used for protein synthesis and repair processes through AdoMet-dependent methyltransferases (Ogé et al., 2008) during the reproductive period. However, the S content and quantity per seed remained lower than in the control (Figure S1). This is in part explained by the reduced level of sulfate in S2 seeds (Figures S2 and S3). Hence, although remobilization processes appeared to be efficient in terms of preventing seed yield losses (Table 1), they did not provide S2 seeds with a level of S and sulfate equivalent to control seeds (Figures S1 and S3). We do not yet know in which forms S is transported to seeds under our conditions, but we suggest it may occur through transport of sulfate and reduced S forms, such as GSH or S–methylmethionine (Figures S2 and S3) (Tabe and Droux, 2001; Tan et al., 2010).

The low abundance of sucrose and glycolysis-related proteins in S1 and S2 seeds may affect the resumption of metabolic activities during early germination

The germination characteristics of the seed samples revealed that S deficiency at flowering or earlier delayed the onset of subsequent germination (Figure 5). Because the early imbibition stages are associated with resumption of metabolic activities from soluble carbohydrates and proteins stored in mature seeds (Rajjou et al., 2012), identifying the soluble sugars and proteins that did not accumulate in mature S1 and S2 seeds may provide information on the processes that are lacking for rapid resumption of metabolic activities. One marked change in the proteome of S1 and S2 seeds was the reduced level of enzymes of the glycolysis and related pathways (Figure 4a). This was accompanied by a significant decrease, up to 50%, in the level of sucrose (Figure 4b), which is presumably required for glycolysis to occur during seed imbibition given that mature M. truncatula seeds contain only trace amounts of glucose and fructose (Djemel et al., 2005). Glycolysis and the subsequent Krebs cycle play a key role in resumption of metabolic activities during early imbibition. Indeed, the heterotrophic nature of the seed makes it dependent on Krebs cycle activities to supply reducing equivalents for mitochondrial respiration, which in turn produce ATP, which is essential for completion of the earliest stages of germination (Bewley, 1997). The low levels of sucrose and glycolytic enzymes in mature seeds may therefore impair glycolysis and the subsequent Krebs cycle during imbibition and have a detrimental effect on the resumption of respiration. In addition, the proteome of S1 seeds was characterized by reduced accumulation of 3–phosphoglycerate dehydrogenase (Figure 4a), an enzyme of the so-called phosphorylated pathway catalyzing the first step of serine synthesis from the glycolytic intermediate 3–phosphoglycerate. The mean abundance of this enzyme also decreased in mature S2 seeds, but the P value (0.056) did not reach statistical significance. Because this pathway is thought to be the primary route for serine production in non-photosynthetic plant tissues, such as seeds (Cheung et al., 1968; Ho et al., 1999), a low level of 3–phosphoglycerate dehydrogenase in mature seeds may alter the reactivation of this pathway during early imbibition. This hypothesis is reinforced by the fact that serine is the precursor of Met, whose de novo synthesis during seed imbibition is essential for rapid germination (Gallardo et al., 2002). The results are thus indicative of a role for S in the accumulation during seed development of compounds that are important for the rapid switch from quiescence to active metabolism during imbibition.

Specific features of S1 seeds are linked to a reduced allocation of C to seeds

Unlike control and S2 seeds, for which germination was accomplished in 5 days, it took 6.5 days for the S1 seeds to reach the maximum percentage of germination (Figure 5b). In addition to low sucrose content, S1 seeds contained low amounts of stachyose, which is the most abundant RFO in mature M. truncatula seeds (Figure 4b). During seed imbibition, RFOs are rapidly degraded into the sucrose that is required for expansion and growth of the embryonic axes (Kuo et al., 1990; Rosnoblet et al., 2007). A study in M. truncatula demonstrated a correlation between the sucrose/RFO ratio and germination and radicle growth (Vandecasteele et al., 2011), suggesting that sucrose conversion into RFOs during seed development is linked to seed vigor during germination and seedling establishment. Therefore, we assume that the low RFO and sucrose levels in mature S1 seeds (Figure 4b) may be a contributory cause of slow germination. By inspecting the nutrient allocation pattern of S1 plants, we observed a drastic shift of C allocation to the root system, together with a twofold reduced quantity of C per seed (Figure 2 and Figure S1). This contrasts with the behavior of S2 plants, which employed a different strategy for survival by maintaining C allocation to their seeds. Thus, it may be argued that S2 plants were more efficient than S1 plants in converting sucrose into RFOs during seed development, probably because of a higher remobilization rate of C to seeds (Figure 2).

Specific features of S3 seeds

Although S deficiency applied during reproductive growth had little influence on yield parameters and S, C and N allocation to seeds, the S3 seeds displayed specific features. They started to germinate early, but their germination speed was slow. To further investigate the factors that may be responsible for slow germination, we searched for specific changes in the proteome of mature S3 seeds. Only four spots specifically varied in these seeds (Figure S5 and Table S1), and one of them, whose abundance decreased by approximately 30%, was identified. It corresponded to SBP65, a seed-specific biotinylated protein of 65 kDa that is degraded during germination (Duval et al., 1994; Dehaye et al., 1997). Its exact role is unknown, but possible functions include a sink for free biotin to sustain germination and a role in desiccation tolerance (Job et al., 2001; Boudet et al., 2006). Although further studies are required to specify its role and relationship with S deficiency, our results provide an indication of the importance of S availability during reproductive growth for SBP65 accumulation.

In conclusion, our results reveal an impressive plasticity of M. truncatula in terms of its response to S deficiency, and provide insights into processes that allow the legume plant to adapt to soils lacking S (Figure 6). Plants deprived of S at flowering efficiently remobilized and translocated nutrients from source organs to seeds. Our results suggest that MtSULTR4;1, a possible indicator of plant S needs, may play a pivotal role in remobilizing sulfate in leaves in order to maintain an active metabolism and sustain the allocation of resources, including sucrose, to seeds. Future studies are planned to specify the role of MtSULTR4;1 by using TILLING mutants in M. truncatula and pea (Dalmais et al., 2008; Le Signor et al., 2009). Total RFO content was maintained in S2 seeds, reflecting efficient conversion of sucrose into RFOs during seed development. However, S2 seeds were delayed in the onset of germination, consistent with the reduced level of sucrose and of proteins related to glycolysis that are important for resumption of metabolic activities. Our data also demonstrated the adverse impact of S deficiency at the mid-vegetative stage on C allocation to seeds, which may affect sugar metabolism during seed development and lead to a major reduction in the level of RFOs, which are important for rapid expansion of the embryonic axes. Finally, although plants deprived of S during reproductive growth were not significantly affected in their development, they yielded seeds with low germination performance and containing low amounts of SBP65, a putative indicator of germination vigor (Capron et al., 2000) that has not previously been related to S nutrition.

Experimental Procedures

Plant growth conditions

Medicago truncatula Gaertn. (A17 genotype) was grown in a greenhouse under a temperature of not more than 30°C during the day and not less than 19°C during the night. Artificial lighting was used to achieve 16 h light per day. Plants were individually grown in 3 liter pots under hydroponic conditions with vigorous aeration. For each condition (S1, S2, S3 and control plants, Figure 1), eight plants were randomized in two distinct blocks with independent nutrient reservoirs. The plants were not inoculated with Sinorhizobium sp. and the nitrogen supply was unlimited: 4 mm KNO3, 4 mm Ca(NO3)2·4 H2O, 0.3 mm MgSO4·7 H2O, 0.9 mm MgCl2, 0.2 mm NaCl, 0.72 μm Na2MoO4·2 H2O, 0.10 mm EDTA-Fe-Na.3H2O, 8.2 μm MnCl2·4 H2O, 1 μm CuCl2·2 H2O, 1 μm ZNCl2, 30 μm H3BO3, 1 mm K2HPO4 (pH adjusted to 6.3 using H3PO4 before addition of K2HPO4). Sulfate was depleted at the S1, S2 and S3 stages using the same solution lacking MgSO4·7 H2O but containing 1.16 mm MgCl2. Leaf chlorophyll content was measured on the ten topmost fully expanded leaves of each plant using a SPAD–502 chlorophyll meter (Minolta Camera Co. Ltd, http://www.konicaminolta.eu/). The plants were harvested at 2200°C days (expressed in thermal time with a base temperature of 5°C; see Bonhomme, 2000; Moreau et al., 2006), the tissues were separated and their weight was determined after oven-drying at 80°C for 48 h, except dry mature seeds, which were weighed after oven-drying at 30°C for 24 h. Analysis of variance (anova) followed by a stringent post hoc test (Tukey's HSD) was applied to reveal significant variations (< 0.05) in yield parameters between S-stressed and control plants (Statistica 7.0 software, StatSoft, http://www.statsoft.fr/index.php). Four biological replicates of S2 and control plants were grown under the same conditions, and flowers of primary branches were tagged on the day of opening (i.e. at 1 DAP). Reproductive leaves (at the base of the tagged pods) and seeds were collected at 14 DAP, frozen in liquid nitrogen, and stored at −80°C until mRNA extraction. The same experiment was repeated to collect additional tissues for measurements of anions and metabolites related to S metabolism.

Reverse transcription and quantitative PCR

Detailed information on the procedure used is provided in Method S1. Briefly, total RNAs were extracted as described by Chang et al. (1993) and treated with RNase-free RQ1 DNase (Promega, http://www.promega.com/). RNA samples were reverse-transcribed using the iScript cDNA synthesis kit (Bio–Rad, http://www.bio-rad.com/), and PCR reactions were performed using a LightCycler® 480 System (Roche Applied Science, https://www.roche-applied-science.com/). The primers used were designed in the 3′ regions to amplify fragments of 80–150 bp, and only primers with efficiencies of 80–100% were used (Table S3). The relative expression ratios for each gene/target were obtained using the ΔΔCt method (Pfaffl, 2001), taking efficiencies into account, with the elongation factor–1α gene as reference. The relative expression ratios (fold change in expression normalized to the reference gene and the control condition, see formula in Method S1) were used for statistical analyses. anova followed by Tukey post hoc test was used to reveal significant differences (< 0.05, = 4) in gene expression between the samples (Statistica 7.0 software).

Relative quantification of S-related metabolites

Metabolite extractions were performed from 5 mg of lyophilized tissue powder in 2 ml Eppendorf tubes with 150 μl cold methanol (LC-MS grade) containing 0.1 m HCl and 200 μm lactitol (internal standard). The tubes were incubated for 15 min at room temperature under agitation, then for 1 h at 4°C, and 100 μl chloroform (LC-MS grade) was added. The tubes were vortexed and placed for 5 min at 37°C. After addition of 167 μl of 0.1 N HCl, the tubes were vortexed again and centrifuged for 5 min at 15 000 g (4°C). Finally, 50 μL of the upper phase was dried in a centrifugal vacuum concentrator (UNIVAPO 100H; UniEquip, http://www.uniequip.com/). Metabolites were derivatized using 6–aminoquinolyl-N–hydroxysuccinimyl carbamate using the AccQ-Tag Ultra derivitization kit (Waters Micromass, http://www.waters.com/), and analyzed by ultra-performance liquid chromatography coupled to tandem mass spectrometry (see Method S2).

Measurements of S, C, N, sugars and anions

S, C and N content was determined on dried, ground tissue samples from six to seven biological samples using the Dumas method (Allen et al., 1974) on a Carlo Erba elemental analyzer NC2500 (Thermo Fisher Scientific, http://www.thermofisher.com/) adapted with a multi-separation column (polytétrafluoroéthylène, 2 mm length, internal and external diameters of 5 and 6 mm, respectively; CE Elantech, http://www.ceelantech.com/). Prior to analysis, 2 mg vanadium pentoxide was added to 5 mg tissues. Sugar measurements were performed using 30 mg lyophilized powder of mature seeds from four biological samples by HPLC (Carbopac PA–1 column, Dionex Corp., http://www.dionex.com/) as described by Vandecasteele et al. (2011). The concentration of anions in 50 mg lyophilized tissue powder from at least four biological samples was determined using a DX100 ion chromatograph with conductivity detector (Dionex Corp.) as described by Abdallah et al. (2010). All data were subjected to anova followed by Tukey post hoc test (Statistica 7.0 software).

Separation and identification of seed proteins

Proteins were extracted from five biological replicates of mature seeds in 67 μl/mg of thiourea/urea lysis buffer as described by Gallardo et al. (2007). After determining protein concentration (Bradford, 1976), 165 μg proteins (i.e. approximately 40 μl protein extracts) were separated in a non-linear pH 3–10 gradient (Immobiline DryStrip, 24 cm; GE Healthcare Life Sciences, http://www.gelifesciences.com/). Gel strips were rehydrated in the IPGphor system (GE Healthcare Life Sciences) for 7 h at 20°C with thiourea/urea lysis buffer containing 2% v/v Triton X–100 and 20 mm dithiothreitol. Isoelectric focusing was performed at 20°C in the IPGphor system for 7 h at 50 V, 1 h at 300 V, 2 h at 3.5 kV and 7 h at 8 kV. Prior to the second dimension, each gel strip was equilibrated as described by Gallardo et al. (2007). Proteins were then separated in 10% polyacrylamide gels as described by Gallardo et al. (2001). Gels were stained with Coomassie Brilliant Blue G–250 (Bio–Rad) as described by Mathesius et al. (2001), and image acquisition was performed using the Odyssey infrared imaging system (LI–COR Biosciences, http://www.licor.com/) at 700 nm with a resolution of 169 μm. Image alignment and spot quantification were performed using Progenesis SameSpots version 2.0 (Nonlinear Dynamics, http://www.nonlinear.com/). For each gel, normalized spot volumes were calculated as the ratio of each spot volume to the total spot volume in the gel. All data were submitted to statistical analyses by Progenesis SameSpots software. Differences with < 0.05 (Student's t test) were considered significant. In the present study, 58 spots were subjected to in-gel trypsin digestion as described by Shevchenko et al. (1996). Extracted peptides were analyzed by nanoLC-MS/MS using an LTQ ion trap mass spectrometer (ThermoFisher Scientific) or a Q–TOF Ultima mass spectrometer (Waters Micromass) as described by Belleannee et al. (2011). Detailed information on the databases and parameters used for protein identification are available in Table S2.

Germination and tetrazolium tests

The seeds were stored for 16 weeks at 20°C in the dark to release dormancy, and then germination was studied using four biological replicates of 30 seeds per condition. Seeds were scarified using fine sandpaper sheets and placed in 14 cm Petri dishes on Whatman paper No. 1 (GE Healthcare Bio-Sciences Corp., Systat Software Inc. USA, http://www.gelifesciences.com/) and imbibed with 3.5 ml distilled water. The seeds were incubated at 20°C in the dark, and the number of germinated seeds (with visible radical protrusion) was recorded during imbibition. The number of days to achieve 1%, 10%, 50% and the maximum percentage germination (Gmax) were estimated using non-linear regression analysis (3rd-order sigmoid, SigmaPlot 8.02 software, Systat Software Inc., San Jose, USA; http://www.sigmaplot.com/). The viability of un-germinated seeds was estimated using the tetrazolium test as described by Zuber et al. (2010b).


We thank Leonardo Casieri (UMR 1347 Agroécologie, Institut National de la Recherche Agronomique, Dijon, France) and the greenhouse team of INRA Dijon for help in setting up the experiment, Christophe Noroy (Master's student in the UMR 1347 Agroécologie) for designing and validating the primers, and Richard Thompson, Judith Burstin, Annabelle Larmure and Vanessa Vernoud, from the UMR 1347 Agroécologie, for helpful discussions and/or critical reading of the manuscript. This work was initiated through funding from the Regional Council of Burgundy (2009-9201AAO040S00680) and continued under the framework of the SERAPIS project (Fonds Unique Interministériel number F1209006 E, 2012–2015). Protein spot identification was performed under the framework of the Génoplante 2010 QualityLegSeed project (GPLA06036G).