Effects of Thiosulfate as a Sulfur Source on Plant Growth, Metabolites Accumulation and Gene Expression in Arabidopsis and Rice

Abstract

Plants are considered to absorb sulfur from their roots in the form of sulfate. In bacteria like Escherichia coli, thiosulfate is a preferred sulfur source. It is converted into cysteine (Cys). This transformation consumes less NADPH and ATP than sulfate assimilation into Cys. In Saccharomyces cerevisiae, thiosulfate promoted growth more than sulfate. In the present study, the availability of thiosulfate, the metabolite transformations and gene expressions it induces were investigated in Arabidopsis and rice as model dicots and monocots, respectively. In Arabidopsis, the thiosulfate-amended plants had lower biomass than those receiving sulfate when sulfur concentrations in the hydroponic medium were above 300 μM. In contrast, rice biomass was similar for plants raised on thiosulfate and sulfate at 300 μM sulfur. Therefore, both plants can use thiosulfate but it is a better sulfur source for rice. In both plants, thiosulfate levels significantly increased in roots following thiosulfate application, indicating that the plants absorbed thiosulfate into their root cells. Thiosulfate is metabolized in plants by a different pathway from that used for sulfate metabolism. Thiosulfate increases plant sulfide and cysteine persulfide levels which means that plants are in a more reduced state with thiosulfate than with sulfate. The microarray analysis of Arabidopsis roots revealed that 13 genes encoding Cys-rich proteins were upregulated more with thiosulfate than with sulfate. These results together with those of the widely targeted metabolomics analysis were used to proposes a thiosulfate assimilation pathway in plants.

Accession numbers: The microarray data reported in this article has been deposited to GEO with accession number GSE118293.

Introduction

Sulfur (S) is an essential element for plant growth. It is an integral component of proteins, amino acids and various secondary metabolites. Sulfur in proteins and amino acids participates in oxidation–reduction reactions and structural maintenance in the form of cysteine (Cys) residues. Plants absorb sulfur from their roots as sulfate which is then reduced and assimilated into Cys (Leustek et al. 2000, Saito 2004, Long et al. 2015).

Plasma membrane-bound ion transporters carry sulfate against the internal negative membrane potential gradient to drive the ion transport. Plants use mainly a proton/sulfate symport system for sulfate influx (Lass and Ullrich-Eberius 1984, Smith et al. 1995). Sulfate transporters are structurally related to a family of membrane-bound transporters with 12 membrane-spanning domains (Smith et al. 1995, Takahashi 2010). Once it is absorbed, sulfate is activated to adenylyl sulfate when it is combined with ATP by ATP sulfurylase. Adenosine 5′ phosphosulfate (APS) is reduced to sulfite by APS reductase. The reducing power is provided by glutathione (GSH). The sulfite is then reduced to sulfide by sulfite reductase. Sulfide then reacts with O-acetyl-L-serine (OAS) synthesized from serine and acetyl CoA by serine acetyltransferase (SAT). This reaction product is assimilated into Cys by cysteine synthase. Cys is an important starting material for the syntheses of Met, GSH, vitamins, coenzymes and iron-sulfur clusters (Droux 2004, Wirtz and Droux 2005, Van Hoewyk et al. 2008). Cystathionine γ-synthase is the key enzyme of Met synthesis (Chiba et al. 1999). The enzyme γ-glutamylcysteine synthetase is the key enzyme of GSH synthesis. In the pathways synthesizing sulfur-containing secondary metabolites like glucosinolates, APS is phosphorylated to 3′-phosphoadenosine-5′-phosphosulfate (PAPS).

In plants, reduced sulfur participates in redox regulation. Compounds with thiol groups such as Cys and GSH are involved in oxidation–reduction reactions (Buchanan and Balmer 2005). They also help fold and stabilize proteins by cleaving disulfide bridges (Haag et al. 2012). GSH is an intracellular antioxidant tripeptide. It participates in intracellular redox homeostasis (Foyer and Noctor 2011, Noctor et al. 2011) mainly via the thiol group in its Cys residue. GSH also detoxifies because of the high reactivity of its Cys residue. GSH polymerizes to phytochelatins which detoxify heavy metals (Mendoza-Cozatl et al. 2011, Rea 2012). GSH conjugates decompose peroxides (Dixon et al. 2002).

In recent years, a new sulfur assimilation pathway using thiosulfate as the substrate was discovered. When plants were exposed to hydrogen sulfide (H₂S), the excess sulfide was converted to persulfide which reacted with sulfite to form thiosulfate (Krüßel et al. 2014). A thiosulfate-specific sulfur transferase in Arabidopsis thaliana has been reported (Bauer and Papenbrock 2002) as was S-sulfocysteine synthase CS26 which synthesizes sulfocysteine from thiosulfate (Bermúdez et al. 2010). Therefore, thiosulfate may be assimilated in plants as a sulfur source. Bartels et al. (2007) reported that thiosulfate application to plant culture media increased thiosulfate levels in plants which suggests that thiosulfate was absorbed and utilized by plants. However, sulfate was also present in the culture media, so thiosulfate was not the sole sulfur source. Therefore, the direct availability of thiosulfate in plants remains unknown.

A thiosulfate metabolic pathway has been identified in Escherichia coli (Hryniewicz et al. 1990, Sekowska et al. 2000). In E. coli, thiosulfate may be preferentially assimilated into Cys when both thiosulfate and sulfate are present. The pathway assimilating thiosulfate into Cys consumes less NADPH and ATP energy than that which transforms sulfate to Cys (Nakatani et al. 2012). In Saccharomyces cerevisiae, thiosulfate also conserves metabolic energy and promotes growth better than sulfate (Funahashi et al. 2015).

Fertilizers containing thiosulfate are commercially available as ammonium thiosulfate, potassium thiosulfate, calcium thiosulfate and magnesium thiosulfate. Thiosulfate application acidifies soils after the S is oxidized. Thiosulfate molecules have unique effects on soil chemistry and biology. Ammonium thiosulfate improves micronutrient solubility (El-Tarabily et al. 2006). Mixing thiosulfate with urea ammonium nitrate (UAN) slows urea hydrolysis and reduces gaseous ammonia (NH₃) losses (Frazier et al. 1990, Goos 1985). Nevertheless, little is understood about thiosulfate availability and assimilation in plants. If plants can utilize thiosulfate as a sulfur source and assimilate it into Cys, then the thiosulfate assimilation pathway may conserve energy relative to sulfate transformation. In the present study, we examined the availability of thiosulfate as a sulfur source to plants. We also evaluated its effects on growth, metabolism, and gene expression. We used A. thaliana and Oryza sativa (rice) as model dicots and monocots, respectively.

Results

Thiosulfate stability in hydroponic culture medium

Thiosulfate in the hydroponic medium may have been oxidized during plant culture. To confirm thiosulfate stability, the sulfate ions (1,500 μM) in the MGRL medium were replaced with thiosulfate ions (750 μM) which were equimolar in terms of sulfur atoms. The solution was placed in a 50-ml conical tube at room temperature. Sulfate, thiosulfate and phosphate concentrations were determined for freshly prepared media by ion chromatography at 0, 19, 24, 43, 48, 66, 72, 91, 97 and 120 h (Supplementary Fig. S1). The thiosulfate concentration gradually decreased whereas the sulfate level increased. The phosphate concentration was nearly constant over the 120 h. The number of sulfur atoms from thiosulfate decreased by ∼2.5× more than the increase in the number of sulfur atoms over 120 h. Therefore, thiosulfate decreased both by oxidation to sulfate and by conversion to other sulfur-containing compounds. The thiosulfate in the medium disappeared after 5 d so the medium was replenished twice weekly in the subsequent experiments.

Growth of Arabidopsis with thiosulfate as a sulfur source

Wild-type A. thaliana ‘Columbia-0’ (Col-0) were hydroponically grown under short-day conditions for 6 weeks using sulfate or thiosulfate as the sole sulfur source. Sulfur in the medium was adjusted by adding Na₂SO₄ or Na₂S₂O₃. Short-day conditions were used to inhibit flowering to simplify the experimental design by omitting the effects of flowering. The sulfur concentrations used were 1,000 μM (control), 300, 100 and 30 μM. As shown in Supplementary Fig. S2, shoots grown with thiosulfate looked smaller than those with sulfate at 1,000 μM and 300 μM sulfur concentrations (Supplementary Fig. S2A). Fresh weights of shoots grown with thiosulfate were smaller than those treated with sulfate at 1,000 and 300 μM, but they were similar at 100 and 30 μM sulfur (Supplementary Fig. S2B). Fresh weights of roots were not different between thiosulfate and sulfate treatments at all sulfur concentrations (Supplementary Fig. S2C). Chlorophyll a content in shoots treated with thiosulfate was higher than that in sulfate treatments at 1,000 and 300 μM, but they were similar in treatments with 100 and 30 μM sulfur (Supplementary Fig. S2D), which may be due to the condensation by smaller biomass. Chlorophyll b content was not different between thiosulfate and sulfate treatments at any sulfur concentration (Supplementary Fig. S2E). Plants are able to use thiosulfate as a sulfur source, but high concentrations of thiosulfate are toxic to them. In treatments with 100 and 30 μM sulfur, plant fresh weights were not decreased by thiosulfate compared with sulfate. However, it is possible that 100 and 30 μM sulfur caused sulfur deficiency responses in Arabidopsis. Namely, the expression of the sulfur deficiency-responsive gene encoding the β subunit of β-conglycinin from a soybean seed storage protein that was introduced in Arabidopsis was induced at 100 μM sulfur, but not at 300 μM, when compared with the control concentration of 1500 μM (Hirai et al. 1995). Therefore, for the following sulfur index and microarray analyses, sulfur was applied to the plants in the form of sulfate or thiosulfate at 300 μM to avoid both severe toxicity from thiosulfate overdose and sulfur deficiency responses.

Sulfur index analysis of Arabidopsis plants

To obtain enough sizes of plant samples for the following analysis, wild-type Col-0 were hydroponically grown under short-day conditions for 8 weeks using sulfate or thiosulfate at 300 μM sulfur, and shoots and roots were subjected to sulfur index, ion chromatography, CN coda, ICP-MS, microarray and widely targeted metabolomics analyses. Fig. 1A, B shows root and shoot fresh weights, respectively, at sampling time. Fresh weights of plants grown with thiosulfate were significantly smaller than those raised on sulfate, nevertheless, both thiosulfate- and sulfate-grown plants had green leaves and did not show any stress symptoms. Sulfur-containing compound profiles in shoots and roots were compared between Arabidopsis plants grown with sulfate and those raised on thiosulfate. To this end, a sulfur index analysis was used. It is a combination of a sensitive LC-MS/MS analysis and a thiol-specific derivatization method using monobromobimane (mBBr) (Kawano et al. 2015, Tanaka et al. 2019, Yamada et al. 2019). As shown in Fig. 2A, thiosulfate levels were significantly higher in both the roots and shoots of the thiosulfate-treated plants than in those treated with sulfate. Thiosulfate in the roots of the thiosulfate-treated plants was >100× higher than it was in those of the plants receiving sulfate. Sulfite levels were also higher in both the roots and shoots of the thiosulfate-treated plants than they were in those for the plants treated with sulfate. In the roots of the thiosulfate-treated plants, sulfite levels were >50× higher than they were in those of the plants treated with sulfate (Fig. 2B). Sulfide and cysteine persulfide (Cys-S) levels were also significantly higher in the roots of thiosulfate-treated plants than they were in those of the plants treated with sulfate (Fig. 2C, E). In the shoots, the levels of Cys, glutathione persulfide (GS-SH) and Met decreased with increasing thiosulfate level (Fig. 2D, G and I).

Fig. 1

Fresh weights of Arabidopsis plants grown with sulfate or thiosulfate. Wild-type A. thaliana Col-0 plants were hydroponically grown with sulfate or thiosulfate at 300 μM of sulfur under short-day conditions for 8 weeks. Means with bars of 95% confidence interval (CI) of 12–14 biological replicates for fresh weights of roots (A) and shoots (B) are shown. Asterisks indicate significant differences between sulfate and thiosulfate treatments (Student’s t-test, *** P < 0.001, * P < 0.05).

Fig. 2

Relative contents of sulfur-containing compounds normalized to the internal standard in Arabidopsis according to sulfur index analysis. Roots and shoots of wild-type A. thaliana Col-0 plants grown with sulfate or thiosulfate at 300 μM of sulfur under short-day conditions for 8 weeks were analyzed. Means with bars of 95% CI of five biological replicates for relative levels of thiosulfate (A), sulfite (B), sulfide (C), Cys (D), Cys-S (E), GSH (F), GS-SH (G), GS-SSH (H) and Met (I) are shown. White bars are sulfate and black bars are thiosulfate. Asterisks indicate significant differences between sulfate and thiosulfate treatments (Wilcoxon signed-rank test, ** P < 0.01, * P < 0.05). N.D., not detected.

A scatterplot matrix of sulfur-containing compounds was constructed and analyzed to identify correlations among metabolites. As shown in Fig. 3, matrix patterns significantly differed between roots receiving thiosulfate and those given sulfate. In thiosulfate-treated roots, there was a positive correlation between thiosulfate dosage and most metabolites. The correlation between thiosulfate level and sulfite was 100% (Fig. 3A). In roots administered sulfate, thiosulfate levels were negatively correlated with other metabolites (Fig. 3B). There was a positive correlation between sulfite level and other metabolites in the thiosulfate-treated roots (Fig. 3A) but a negative correlation between sulfite level and metabolites in sulfate-treated roots (Fig. 3B). In shoots, the matrix patterns differed between plants given thiosulfate (Fig. 4A) and those receiving sulfate (Fig. 4B). As for the roots, the correlations among thiosulfate, sulfite and other metabolites were more strongly positive in shoots supplied with thiosulfate than in those shoots receiving sulfate. GSSSH was positively correlated with other metabolites in plants receiving thiosulfate. The correlation between GSSSH and GSH was 100% (Fig. 4A). In contrast, for the shoots grown with sulfate, the correlations between GSSSH and other metabolites were negative.

Fig. 3

Scatter plot matrix of sulfur-containing compounds in Arabidopsis roots determined by sulfur index analysis. The blue (A) and red (B) dots in the left triangular matrix show relative contents of metabolites normalized to the internal standard in plants grown with thiosulfate (A) or sulfate (B) as a sulfur source, respectively. The right triangular matrix shows Spearman’s rank correlation and the numbers in ellipse are the coefficients. The darker the blue color of the ellipse, the higher the positive correlation. Asterisks indicate significant correlations (n = 5, pairwise analysis, *** P < 0.001, ** P < 0.01, * P < 0.05).

Fig. 4

Scatter plot matrix of sulfur-containing compounds in Arabidopsis shoots determined by sulfur index analysis. The blue (A) and red (B) dots in the left triangular matrix show relative contents of metabolites normalized to the internal standard in plants grown with thiosulfate (A) and sulfate (B) as a sulfur source, respectively. The right triangular matrix shows Spearman’s rank correlation and the numbers in ellipse are the coefficients. The darker the blue color of the ellipse, the higher the positive correlation. Asterisks indicate significant correlations (n = 5, pairwise analysis, *** P < 0.001, ** P < 0.01, * P < 0.05).

Effects of thiosulfate on sulfate content in Arabidopsis plants

Sulfate was not included in the profile of the sulfur index analysis (Kawano et al. 2015). Therefore, sulfate levels in the shoots and roots of Arabidopsis were determined along with nitrate and phosphate by ion chromatography. As shown in Table 1, the sulfate and nitrate contents were ∼50% lower in the roots of the thiosulfate-treated plants than they were in the roots of the sulfate-supplied plants. In the shoots, however, there was no significant difference between treatments. No significant differences were observed between treatments for root or shoot phosphate levels.

Table 1

Sulfate, nitrate and phosphate concentrations in the shoots and roots of Arabidopsis

	Roots		Shoots
	Sulfate	Thiosulfate	Sulfate	Thiosulfate
Sulfate (μ mol g FW−1)	9.8 ± 0.5	5.1 ±1.9*	11.4 ± 2.5	10.6 ± 1.0
Nitrate (μ mol g FW−1)	30.0 ± 4.4	18.0 ± 3.7*	167.1 ± 36.4	175.4 ± 34.1
Phosphate (μ mol g FW−1)	3.6 ± 1.9	5.6 ± 4.3	7.7 ± 1.5	9.2 ± 1.3

Shoots and roots of wild-type A. thaliana Col-0 were grown either with sulfate or thiosulfate under short-day conditions for 8 weeks. Means ± standard deviations of four biological replicates are shown. Asterisks indicate significant differences between sulfate and thiosulfate treatments (Mann–Whitney U test, P < 0.05).

Effects of thiosulfate on total sulfur, nitrogen and carbon in Arabidopsis plants

To test whether thiosulfate affects the total sulfur contents in plants, total sulfur, together with nitrogen and carbon, was determined in shoots and roots (Table 2). There were no significant differences in the levels of any of these elements between thiosulfate and sulfate applications.

Table 2

Total carbon, nitrogen and sulfur in the shoots and roots of Arabidopsis thaliana ‘Columbia-0’ grown with sulfate or thiosulfate under short-day conditions for 8 weeks

	Roots		Shoots
	Sulfate	Thiosulfate	Sulfate	Thiosulfate
Carbon (mg g DW−1)	385.0 ± 18.4	386.6 ± 10.9	362.5 ± 5.6	358.6 ± 6.0
Nitrogen (mg g DW−1)	48.5 ± 2.5	51.3 ± 1.0	70.8 ± 1.7	71.8 ± 1.4
Sulfur (mg g DW−1)	14.1 ± 1.0	11.1 ± 0.5	7.4 ± 0.2	6.4 ± 0.6

Means ± standard deviations of five (for shoots) or three (for roots) biological replicates are shown.

DW, dry weight.

Microarray analysis of Arabidopsis plants

The sulfur index analysis shown in Fig. 2 indicates that thiosulfate application changed the sulfur-containing metabolite levels especially in the roots. The effects of thiosulfate application on gene expression profiles in roots were determined by extracting RNA from the roots of the wild-type Arabidopsis hydroponically grown for 8 weeks and simultaneously used for sulfur index analysis. The plants were subjected to microarray analysis with an Affymetrix GeneChip^TM Arabidopsis Gene 1.0 ST Array.

We used principal component analysis (PCA) to evaluate the overall gene expression profile. The thiosulfate and sulfate treatments generated separate clusters (Supplementary Fig. S3). Therefore, the thiosulfate application induced changes in the gene expression profile relative to that for the sulfate treatment in the roots of Arabidopsis plants.

The scatterplot of the normalized signals is shown in Supplementary Fig. S4. The numbers of probes with differentially expressed genes (DEGs) with fold change (log₂-FC) > 2 and < −2 by thiosulfate compared to sulfate were 153 and 91, respectively. The total was 244.

All 244 DEGs, fold changes with thiosulfate compared to sulfate, gene symbols, gene descriptions and their GOs are listed in Supplementary Table S1. The 153 DEGs upregulated by thiosulfate are shown in Table 3 and 91 DEGs dowregulated by thiosulfate are in Table 4. As shown in Table 3, the most strongly induced gene was At1g55390 which is described as a cysteine/histidine-rich C1 domain-containing protein. Twelve other genes were also described as Cys-rich proteins like At5g26130, At3g28650, At3g11390, At1g61840, At5g57625, At3g11370, CRK31, At2g02680, At5g02350, At1g55700, At5g54030 and At1g44030. As for genes involved in sulfur assimilation, the two 5′-adenylylsulfate reductase genes, APR1 and APR2, were induced by thiosulfate application. The MAM1 gene was induced by thiosulfate. It is involved in the synthesis of Met-derived glucosinolates. While thiosulfate repressed the BGLU30 gene which may participate in glucosinolate catabolism (Maruyama-Nakashita 2017) (Table 4). Among the 91 DEGs repressed by thiosulfate, 6 and 14 genes encoded pre-tRNA-Met and pre-tRNA-Phe, respectively.

Table 3

DEGs upregulated in Arabidopsis roots with thiosulfate compared with sulfate

Fold change	AGI code	Gene symbol	Description
5.60	AT1G55390	AT1G55390	Cysteine/histidine-rich C1 domain-containing protein
5.48	AT5G06900	CYP93D1	Cytochrome P450, family 93, subfamily D, polypeptide 1
4.39	AT1G80320	AT1G80320	2-Oxoglutarate (2OG) and Fe(II)-dependent oxygenase superfamily protein
4.26	AT2G24980	AT2G24980	Proline-rich extensin-like family protein
4.16	AT2G01520	MLP328	MLP-like protein 328
3.88	AT1G51830	AT1G51830	Putative leucine-rich repeat protein kinase
3.61	AT1G15540	AT1G15540	2-Oxoglutarate (2OG) and Fe(II)-dependent oxygenase-like protein
3.56	AT5G03995	AT5G03995	Hypothetical protein
3.49	AT4G29905	AT4G29905	Hypothetical protein
3.48	AT5G57150; AT5G57140	AT5G57150; PAP28	Transcription factor bHLH35; purple acid phosphatase 28
3.42	AT2G05440	GRP9	Glycine-rich protein 9
3.42	AT3G45130	LAS1	Lanosterol synthase 1
3.41	AT1G52820	AT1G52820	Putative 2-oxoglutarate-dependent dioxygenase
3.32	AT3G54070	AT3G54070	Ankyrin repeat-containing protein
3.29	AT4G08400	AT4G08400	Proline-rich extensin-like family protein
3.10	AT5G23220	NIC3	Nicotinamidase 3
3.05	AT1G05240; AT1G05250	AT1G05240; AT1G05250	Peroxidase 1; peroxidase 2
3.03	AT3G45970	EXLA1	Expansin-like A1
2.98	AT4G29450	AT4G29450	Leucine-rich repeat protein kinase-like protein
2.90	AT5G26130	AT5G26130	Cysteine-rich secretory proteins, Antigen 5 and Pathogenesis-related 1 protein superfamily protein
2.89	AT4G11320	AT4G11320	Putative cysteine proteinase
2.88	AT1G33900	AT1G33900	P-loop containing nucleoside triphosphate hydrolases superfamily protein
2.87	AT1G54950	AT1G54950	Hypothetical protein
2.85	AT5G06800	AT5G06800	myb-like HTH transcriptional regulator family protein
2.82	AT4G02270	RHS13	Protein root hair specific 13
2.77	AT3G17280	AT3G17280	F-box and associated interaction domains-containing protein
2.77	AT1G19250	FMO1	Flavin-dependent monooxygenase 1
2.77	AT2g29130	LAC2	Laccase 2
2.76	AT1G49030	AT1G49030	PLAC8 family protein
2.73	AT3G27960	AT3G27960	Tetratricopeptide repeat domain-containing protein
2.73	AT1G29670	AT1G29670	GDSL esterase/lipase
2.72	AT4G15393	CYP702A5	Cytochrome P450, family 702, subfamily A, polypeptide 5
2.71	AT5G06905	CYP712A2	Cytochrome P450, family 712, subfamily A, polypeptide 2
2.71	AT3G28650	AT3G28650	Cysteine/histidine-rich C1 domain-containing protein
2.70	AT1G73330	DR4	Protein drought-repressed 4
2.67	AT4G12550	AIR1	Auxin-induced in root cultures 1
2.67	AT5G06640	AT5G06640	Proline-rich extensin-like family protein
2.66	AT3G01420	DOX1	Alpha-dioxygenase
2.66	AT2G33850	AT2G33850	Hypothetical protein
2.63	AT3G07340	AT3G07340	Transcription factor bHLH62
2.63	AT2G42250	CYP712A1	Cytochrome P450, family 712, subfamily A, polypeptide 1
2.62	AT5G12420	AT5G12420	O-acyltransferase (WSD1-like) family protein
2.59	AT2G41830	AT2G41830	Hypothetical protein
2.56	AT5G67400	RHS19	Peroxidase 73
2.56	AT1G12950	RSH2	Root hair specific 2
2.55	AT4G01430	AT4G01430	Nodulin MtN21-like transporter family protein
2.54	AT1G67150	AT1G67150	Hypothetical protein
2.52	AT2G01530	MLP329	MLP-like protein 329
2.52	AT5G62920	ARR6	Two-component response regulator ARR6
2.52	AT2G46650	CB5-C	Cytochrome B5 isoform C
2.51	AT1G49450	AT1G49450	WD40 domain-containing protein
2.49	AT3G11390	AT3G11390	Cysteine/histidine-rich C1 domain-containing protein
2.48	AT3G44550	FAR5	Fatty acyl-CoA reductase 5
2.47	AT1G62180	APR2	5-Adenylylsulfate reductase 2
2.47	AT1G61840	AT1G61840	Cysteine/histidine-rich C1 domain-containing protein
2.45	AT2G27402	AT2G27402	Hypothetical protein
2.44	AT4G08380	AT4G08380	Proline-rich extensin-like family protein
2.44	AT5G57625	AT5G57625	CAP (Cysteine-rich secretory proteins, Antigen 5 and Pathogenesis-related 1 protein) superfamily protein
2.42	AT3G11370	AT3G11370	Cysteine/histidine-rich C1 domain-containing protein
2.41	AT1G52790	AT1G52790	Oxidoreductase, 2OG-Fe(II) oxygenase family protein
2.40	AT5G44585	AT5G44585	Hypothetical protein
2.39	AT3G53750	ACT3	Actin 3
2.39	AT1G55330	AGP21	Arabinogalactan protein 21
2.39	AT1G12040	LRX1	Leucine-rich repeat extensin-like protein 1
2.38	AT3G52970	CYP76G1	Cytochrome P450, family 76, subfamily G, polypeptide 1
2.37	AT1G51940	AT1G51940	LysM type receptor kinase-like protein
2.37	AT4G11470	CRK31	Putative cysteine-rich receptor-like protein kinase 31
2.37	AT1G70850	MLP34	MLP-like protein 34
2.37	AT4G08930	APRL6	Protein APR-like 6
2.36	AT5G06630	AT5G06630	Proline-rich extensin-like family protein
2.35	AT2G19090	AT2G19090	Hypothetical protein
2.35	AT4G20240	CYP71A28	Putative cytochrome P450 71A28
2.34	AT2G02680	AT2G02680	Cysteine/histidine-rich C1 domain-containing protein
2.34	AT2G38530	LTP2	Non-specific lipid-transfer protein 2
2.34	AT1G16060	ADAP	AP2-like ethylene-responsive transcription factor
2.34	AT1G16060	ADAP	AP2-like ethylene-responsive transcription factor
2.34	AT2G04440	AT2G04440	MutT/nudix-like protein
2.32	AT3G02610	AT3G02610	Acyl-[acyl-carrier-protein] desaturase
2.32	AT3G18400	NAC058	NAC domain-containing protein 58
2.32	AT1G72180	AT1G72180	Leucine-rich receptor-like protein kinase
2.31	AT5G66690	UGT72E2	UDP-glycosyltransferase 72E2
2.31	AT1G33840	AT1G33840	Hypothetical protein
2.31	AT2G24710	GLR2.3	Glutamate receptor 2.3
2.31	AT1G05240; AT1G05250	AT1G05240; AT1G05250	Peroxidase 1; peroxidase 2
2.28	AT3G62040	AT3G62040	Haloacid dehalogenase-like hydrolase superfamily protein
2.28	AT1G10470	ARR4	Two-component response regulator ARR4
2.27	AT4G30670	AT4G30670	Putative membrane lipoprotein
2.27	AT3G54580	AT3G54580	Proline-rich extensin-like family protein
2.26	AT5G02360	AT5G02360	DC1 domain-containing protein
2.26	AT1G61667	AT1G61667	Hypothetical protein
2.26	AT2G24710	GLR2.2	Glutamate receptor 2.2
2.26	AT5G44020	AT5G44020	HAD superfamily, subfamily IIIB acid phosphatase
2.26	AT4G35160	AT4G35160	O-methyltransferase family protein
2.25	AT2G30930	AT2G30930	Hypothetical protein
2.21	AT4G04610	APR1	5-Adenylylsulfate reductase 1
2.20	AT1G77060	AT1G77060	Carboxyvinyl-carboxyphosphonate phosphorylmutase
2.19	AT1G20930	CDKB2;2	Cyclin-dependent kinase B2-2
2.19	AT1G64200	VHA-E3	V-type proton ATPase subunit E3
2.18	AT4G34160	CYCD3;1	Cyclin-D3-1
2.18	AT1G55450	AT1G55450	S-adenosyl-L-Met-dependent methyltransferases superfamily protein
2.18	AT3G10720	AT3G10720	Pectinesterase 25
2.18	AT1G62990	KNAT7	Homeobox protein knotted-1-like 7
2.17	AT4G28250	EXPB3	Expansin B3
2.17	AT4G18610	LSH9	Hypothetical protein
2.16	AT2G44110	MLO15	MLO-like protein 15
2.16	AT5G55550	AT5G55550	RNA recognition motif-containing protein
2.14	AT1G61430	AT1G61430	G-type lectin S-receptor-like serine/threonine-protein kinase
2.14	AT3G23810	SAHH2	Adenosylhomocysteinase 2
2.13	AT3G60130	BGLU16	Beta-glucosidase 16
2.13	AT1G78120	AT1G78120	Tetratricopeptide repeat-containing protein
2.13	AT2G28250	NCRK	Receptor-like serine/threonine-protein kinase NCRK
2.13	AT4G35320	AT4G35320	Hypothetical protein
2.12	AT5G24030	SLAH3	SLAC1 homolog 3
2.12	AT2G40960	AT2G40960	Single-stranded nucleic acid binding R3H protein
2.12	AT1G70860	AT1G70860	SRPBCC ligand-binding domain-containing protein
2.12	AT2G31270	CDT1A	CDT1-like protein a
2.11	AT1G24100	UGT74B1	UDP-glucosyl transferase 74B1
2.11	AT5G02350	AT5G02350	Cysteine/histidine-rich C1 domain-containing protein
2.10	AT1G77530	AT1G77530	O-methyltransferase family protein
2.10	AT3G49330	AT3G49330	Invertase/pectin methylesterase inhibitor family protein
2.10	AT2G20080	AT2G20080	Hypothetical protein
2.09	AT1G55700	AT1G55700	Cysteine/histidine-rich C1 domain-containing protein
2.08	AT2G29480	GSTU2	Glutathione S-transferase tau 2
2.08	AT2G28990	AT2G28990	Putative LRR receptor-like serine/threonine-protein kinase
2.08	AT4G01450	AT4G01450	Nodulin MtN21-like transporter family protein
2.08	AT1G68710	AT1G68710	Putative phospholipid-transporting ATPase 9
2.08	AT4G37410	CYP81F4	Cytochrome P450, family 81, subfamily F, polypeptide 4
2.07	AT1G32450	NRT1.5	Nitrate transporter 1.5
2.07	AT1G23720	AT1G23720	Proline-rich extensin-like family protein
2.07	AT1G52800	AT1G52800	Oxidoreductase, 2OG-Fe(II) oxygenase family protein
2.06	AT5G07760	AT5G07760	Putative formin-like protein 21a
2.05	AT3G52370	FLA15	Fasciclin-like arabinogalactan protein 15
2.05	AT2G32990	GH9B8	Glycosyl hydrolase 9B8
2.05	AT1G10020	AT1G10020	Hypothetical protein
2.05	AT5G54030	AT5G54030	Cysteine/histidine-rich C1 domain-containing protein
2.04	AT1G44030	AT1G44030	Cysteine/Histidine-rich C1 domain family protein
2.04	AT4G35200	AT4G35200	Hypothetical protein
2.04	AT1G11670	AT1G11670	MATE efflux family protein
2.04	AT1G51850	AT1G51850	Leucine-rich repeat protein kinase family protein
2.04	AT4G24415	MIR824A	MicroRNA824A,
2.04	AT1G04250	AXR3	Auxin-responsive protein IAA17
2.04	AT5G16390	CAC1	Chloroplastic acetylcoenzyme A carboxylase 1
2.03	AT5G23010	MAM1	Methylthioalkylmalate synthase 1
2.03	AT2G47160	BOR1	Boron transporter 1
2.02	AT1G75060	AT1G75060	Hypothetical protein
2.02	AT5G57800	CER3	Protein ECERIFERUM 3
2.02	AT5G56970	CKX3	Cytokinin dehydrogenase 3
2.01	AT4G37540	LBD39	LOB domain-containing protein 39
2.01	AT1G14630	AT1G14630	hypothetical protein
2.01	AT3G55890	AT3G55890	Yippee family putative zinc-binding protein
2.01	AT5G65360	AT5G65360	Histone H3.1
2.01	AT1G64570	DUO3	Homeodomain-like superfamily protein
2.00	AT5G66560	AT5G66560	Phototropic-responsive NPH3 family protein
2.00	AT1G10480	ZFP5	Zinc finger protein 5

Table 4

DEGs downregulated in Arabidopsis roots with thiosulfate compared with sulfate

Fold change	AGI code	Gene symbol	Description
−2.02	AT1G05060	AT1G05060	Hypothetical protein
−2.02	AT3g24290	AMT1;5	Ammonium transporter 1;5
−2.02	AT4G00651	AT4G00651	Hypothetical protein
−2.03	AT5G47170	AT5G47170	Hypothetical protein
−2.04	AT3G46080	AT3G46080	Zinc finger protein ZAT8
−2.05	AT5G45990	AT5G45990	Putative crooked neck protein/cell cycle protein
−2.06	AT3G48450	AT3G48450	RPM1-interacting protein 4 (RIN4)
−2.06	AT4G19720	AT4G19720	Glycosyl hydrolase family protein with chitinase insertion domain
−2.06	AT2G29870	AT2G29870	Putative aquaporin NIP4
−2.06	AT5G35290	AT5G35290	Hypothetical protein
−2.06	AT5G67370	AT5G67370	Hypothetical protein
−2.07	AT1G74710	EDS16	Isochorismate synthase 1
−2.09	AT3G53305	CYP71B32	Cytochrome P450, family 71, subfamily B, polypeptide 32
−2.10	AT1G12940	NRT2.5	Nitrate transporter2.5
−2.10	AT2G29910	AT2G29910	F-box/LRR-repeat protein
−2.12	AT5G36980	AT5G36980	Hypothetical protein
−2.15	AT3G09390	MT2A	Metallothionein 2A
−2.16	AT2G21890	CAD3	Cinnamyl alcohol dehydrogenase-like protein 3
−2.16	AT5G57510	AT5G57510	Hypothetical protein
−2.17	AT1G74880	NDH-O	NAD(P)H:plastoquinone dehydrogenase complex subunit O
−2.19	AT3G26190	CYP71B21	Cytochrome P450 71B21
−2.19	AT2G15780	AT2G15780	Cupredoxin superfamily protein
−2.21	AT5G28810	AT5G28810	Hypothetical protein
−2.22	AT2G02340	PP2-B8	Phloem protein 2-B8
−2.22	AT3G45093	AT3G45093	Defensin-like protein 255
−2.23	AT1G57980	AT1G57980	Nucleotide-sugar transporter family protein
−2.23	AT1G71695	AT1G71695	Peroxidase 12
−2.24	AT5G37440	AT5G37440	Chaperone DnaJ-domain superfamily protein
−2.26	AT1G51035	AT1G51035	Hypothetical protein
−2.27	AT4G16820	PLA-I{beta]2	Phospholipase A1-Ibeta2
−2.27	AT1G51030	AT1G51030	Hypothetical protein
−2.28	AT5G13740	ZIF1	Zinc induced facilitator 1 protein
−2.28	AT4G12430	CPuORF26; AT4G12430	Uncharacterized protein; probable trehalose-phosphate phosphatase F
−2.29	AT1G69000	AT1G69000	Pre-tRNA tRNA-Met
−2.29	AT2G23020	AT2G23020	Pre-tRNA tRNA-Met
−2.29	AT2G23650	AT2G23650	Pre-tRNA tRNA-Met
−2.29	AT5G57885	AT5G57885	Pre-tRNA tRNA-Met
−2.29	AT5G67455	AT5G67455	Pre-tRNA tRNA-Met
−2.29	AT5G54865	AT5G54865	Pre-tRNA tRNA-Met
−2.31	AT5G37140	AT5G37140	P-loop containing nucleoside triphosphate hydrolases superfamily protein
−2.32	AT3G28193	AT3G28193	Hypothetical protein
−2.34	AT5G26950	AGL93	Agamous-like MADS-box protein AGL93
−2.35	AT5G12323	AT5G12323	Hypothetical protein
−2.38	AT1G31290	AGO3	Argonaute 3
−2.42	AT3G57157	AT3G57157	Short open reading frame 5
−2.44	AT2G30766	AT2G30766	Hypothetical protein
−2.46	AT5G43610	SUC6	Putative sucrose transport protein SUC6
−2.46	AT1G02480	AT1G02480	Pre-tRNA tRNA-Phe
−2.46	AT1G03640	AT1G03640	Pre-tRNA tRNA-Phe
−2.46	AT1G68970	AT1G68970	Pre-tRNA tRNA-Phe
−2.46	AT3G15585	AT3G15585	Pre-tRNA tRNA-Phe
−2.46	AT3G62285	AT3G62285	Pre-tRNA tRNA-Phe
−2.46	AT4G27395	AT4G27395	Pre-tRNA tRNA-Phe
−2.46	AT4G01870	AT4G01870	tolB protein-like protein
−2.46	AT4G03435	AT4G03435	Pre-tRNA tRNA-Phe
−2.46	AT5G54365	AT5G54365	Pre-tRNA tRNA-Phe
−2.46	AT5G54375	AT5G54375	Pre-tRNA tRNA-Phe
−2.46	AT5G03452	AT5G03452	Pre-tRNA tRNA-Phe
−2.46	AT5G11475	AT5G11475	Pre-tRNA tRNA-Phe
−2.46	AT5G25625	AT5G25625	Pre-tRNA tRNA-Phe
−2.46	AT5G55045	AT5G55045	Pre-tRNA tRNA-Phe
−2.46	AT5G55055	AT5G55055	Pre-tRNA tRNA-Phe
−2.49	AT1G73260	KTI1	Kunitz trypsin inhibitor 1
−2.51	AT4G16370	OPT3	Oligopeptide transporter
−2.51	AT2G18700	TPS11	Putative alpha, alpha-trehalose-phosphate synthase [UDP-forming] 11
−2.52	AT4G26460	AT4G26460	S-adenosyl-L-Met-dependent methyltransferases superfamily protein
−2.53	AT5G42530	AT5G42530	Hypothetical protein
−2.55	AT1G70390	AT1G70390	Putative F-box protein
−2.55	AT1G02230	NAC004	NAC domain-containing protein 4
−2.57	AT4G37760	SQE3	Squalene epoxidase 3
−2.59	AT5G45105	ZIP8	Zinc transporter 8 precursor
−2.62	AT5G49620	MYB78	Myb domain protein 78
−2.65	AT5G26170	WRKY50	Putative WRKY transcription factor 50
−2.69	AT3G01970	WRKY45	WRKY DNA-binding protein 45
−2.71	AT5G39690	NAC093	NAC domain-containing protein 93
−2.76	AT4G05370	AT4G05370	BCS1 AAA-type ATPase
−2.83	AT2G26400	ARD3	Acireductone dioxygenase 3
−2.83	AT5G27580	AGL89	Protein AGAMOUS-LIKE 89
−2.84	AT5G52720	AT5G52720	Copper transport family protein
−2.90	AT1G01680	PUB54	U-box domain-containing protein 54
−3.29	AT3G04525	AT3G04525	Pre-tRNA tRNA-Arg
−3.34	AT1G51020	AT1G51020	Hypothetical protein
−3.40	AT1G43910	AT1G43910	P-loop containing nucleoside triphosphate hydrolases superfamily protein
−3.46	AT2G46670	AT2G46670	CCT motif family protein
−3.52	AT1G21520	AT1G21520	Hypothetical protein
−3.59	AT1G69880	TH8	Thioredoxin H8
−3.77	AT1G51010	AT1G51010	Hypothetical protein
−4.04	AT1G18830	AT1G18830	Transport protein SEC31-like protein SEC31B
−5.00	AT2G30670	AT2G30670	NAD(P)-binding Rossmann-fold superfamily protein
−5.28	AT1G09080	BIP3	Protein BIP3
−5.77	AT3G60140	BGLU30, DIN2	Beta-glucosidase 30

Widely targeted metabolomics in Arabidopsis plants

To elucidate the effects of thiosulfate application on other metabolites that cannot be measured by sulfur index analysis, the widely targeted metabolomics, which quantifies hundreds of targeted metabolites in a high-throughput manner (Sawada et al. 2009, Sawada et al. 2017), was performed. The roots and shoots of Arabidopsis plants grown hydroponically for 8 weeks (as for the sulfur index and microarray analyses) were analyzed by widely targeted metabolomics for their metabolite content. Fold changes (thiosulfate/sulfate) and intensities (peak area normalized to the internal standards) of all metabolite contents detected in the analysis are listed in Supplementary Table S2. Root and shoot metabolites levels significantly differing between the thiosulfate and sulfate treatments are listed in Table 5. Among the 117 metabolites detected (Supplementary Table S2), 36 and 7 significantly changed in the roots and shoots, respectively, between the thiosulfate and sulfate applications (Table 5). Sixteen glucosinolates (GSLs) were detected (Supplementary Table S2). Among them six Met-derived GSLs were significantly decreased in the roots by thiosulfate application including 3-methylsulfinylpropyl-GSL, 4-methylsulfinylbutyl-GSL, 5-methylsulfinylpentyl-GSL, 7-methylsulfinylheptyl-GSL, 4-methylthiobutyl-GSL and 5-methylthiopentyl-GSL. In contrast, 8-methylthiooctyl-GSL and 1-methoxyindole-GS were significantly increased in the roots (Table 5). S-adenosyl-L-Met and its derivatives like S-methyl-L-Met, 5′-methyladenosine, 1-aminocyclopropane-1-carboxylic acid, choline phosphate and choline were also significantly decreased in the roots by thiosulfate application (Table 5).

Table 5

Metabolites significantly changed between thiosulfate and sulfate application in roots or shoots of Arabidopsis in widely targeted metabolomics

Annotation	Fold change (thiosulfate/sulfate)		t-test P-value (thiosulfate vs. sulfate)
	Roots	Shoots	Roots	Shoots
Metformin	0.5	1.2	0.042	0.335
1-Aminocyclopropane-1-carboxylic acid	0.7	0.8	0.039	0.173
Adenine	1.7	2.9	0.017	0.151
Phenylalanine	0.7	1.0	0.012	0.811
Deoxycytidine	0.6	0.6	0.421	0.044
Sinapate	0.6	0.8	0.026	0.207
Glutathione disulfide	3.8	0.5	0.006	0.304
Hypoxanthine	0.6	1.1	0.015	0.857
O-Phospho-L-serine	0.6	0.6	0.011	0.269
Quisqualic acid	1.2	0.8	0.677	0.029
Nicotinamide D-ribonucleotide	0.7	0.9	0.048	0.498
Pyridoxal phosphate	0.8	1.1	0.049	0.611
D-Alanyl-D-alanine	0.5	0.5	0.002	0.292
5′-Methylthioadenosine	0.7	0.7	0.004	0.120
S-Adenosyl-L-Met	0.7	1.0	0.002	0.811
Alpha-Methyl-histidine	0.8	0.9	0.049	0.799
3-Hydroxy-3-methylglutarate	0.4	0.3	0.005	0.057
Ethanolamine phosphate	0.6	0.6	0.010	0.202
Choline	0.8	1.0	0.001	0.802
Choline phosphate	0.6	0.6	0.004	0.033
N-Methylnicotinate	1.1	0.6	0.905	0.004
2-Phenylethyl-GSL	1.0	1.4	0.414	0.023
Sulforaphane	0.6	0.4	0.009	0.106
Esculetin	0.3	1.4	0.000	0.471
3-Methylsulfinylpropyl-GSL	0.3	1.1	0.014	0.331
4-Methylsulfinylbutyl-GSL	0.4	1.2	0.004	0.085
5-Methylsulfinylpentyl-GSL	0.6	1.6	0.029	0.001
7-Methylsulfinylheptyl-GSL	0.6	1.0	0.003	0.902
4-Methylthiobutyl-GSL	0.4	1.2	0.008	0.554
5-Methylthiopentyl-GSL	0.7	1.2	0.020	0.603
8-Methylthiooctyl-GSL	1.5	1.4	0.049	0.399
1-Methoxyindole-GSL	1.2	3.1	0.264	0.037
4-Methoxyindole-GSL	1.7	1.0	0.031	0.860
S-MethylMet	0.5	3.3	0.001	0.485
CDP-choline	0.6	0.9	0.029	0.798
Pantothenate	0.8	1.0	0.034	0.762
Nicotinamide	0.6	0.8	0.016	0.147
N,N-Dimethylglycine	0.8	0.9	0.016	0.405
Indole-3-carboxaldehyde	0.6	0.8	0.032	0.179
Malate	0.6	0.9	0.012	0.490
2-Oxoglutarate	0.6	2.1	0.049	0.178

Roots and shoots of A. thaliana Col-0 wild-type plants grown with sulfate or thiosulfate under short-day condition for 8 weeks were subjected to widely targeted metabolomics analysis. Analysis was performed with five biological replicates and the fold changes of thiosulfate to sulfate are shown.

GSL, glucosinolate.

Growth of rice with thiosulfate as a sulfur source

Thiosulfate treatment at more than 300 μM of sulfur repressed Arabidopsis growth (Fig. 2, Supplementary Fig. S2). The reducing conditions created by thiosulfate application may not have been conducive to Arabidopsis growth. It is known that lowland rice plants prefer reducing conditions, so we compared their growth response to thiosulfate application with that of Arabidopsis. O. sativa L. ‘Koshihikari’ plants were hydroponically grown for 4 weeks under a long-day condition. The sulfur concentration in the medium was set to 300 μM using sulfate or thiosulfate. The leaf stages did not significantly differ between sulfur sources: 4.2 ± 0.2 with sulfate and 4.3 ± 0.2 with thiosulfate [means ± 95% confidence interval (CI); n = 18, Student’s t-test]. Fig. 5A shows the fresh weights of the roots, the first + second leaves, the third leaves and the fourth leaves. There were no significant differences between treatments in terms of the fresh weights of both mature (first and second) and young (third and fourth) leaves. As shown in Fig. 5B, the SPAD values of the third or fourth leaves did not significantly differ between sulfate and thiosulfate.

Fig. 5

Fresh weights and SPAD values of rice plants grown with sulfate or thiosulfate. Oryza sativa L. “Koshihikari” plants were hydroponically grown with sulfate or thiosulfate at 300 μM of sulfur for 4 weeks. (A) Fresh weights of roots, first and second leaves, third leaves and fourth leaves are indicated by columns with diagonal, gray, black and white lines, respectively. (B) SPAD values of third and fourth leaves are indicated by black and white columns, respectively. Means with bars of 95% CI are shown. There were no significant differences between sulfate and thiosulfate when the same organs were compared (n = 15, Dunnett’s test).

Sulfur index analysis of rice plants

A sulfur index analysis (Kawano et al. 2015) was used to compare the sulfur-containing compound profiles in the shoots and roots between rice plants grown with sulfate and those raised on thiosulfate. As shown in Fig. 6, the levels of thiosulfate, sulfite, sulfide, Cys-S and GS-SH were much higher in the roots of thiosulfate-treated rice than in those of sulfate-treated rice. The root and shoot levels of other metabolites like Cys, GSH and Met did not significantly differ between thiosulfate- and sulfate-supplied rice plants.

Fig. 6

Relative contents of sulfur-containing compounds in rice normalized to the internal standard according to the sulfur index analysis. Roots and shoots of Oriza sativa L. “Koshihikari” plants grown with sulfate or thiosulfate at 300 μM sulfur for 4 weeks were analyzed. Means with bars of 95% CI of five biological replicates for relative contents of thiosulfate (A), sulfite (B), sulfide (C), Cys (D), Cys-S (E), GSH (F), GS-SH (G) and Met (H) are shown. White bars are sulfate and black bars are thiosulfate. Asterisks indicate significant differences between sulfate and thiosulfate treatments (n = 5, Wilcoxon signed-rank test, ** P < 0.01, * P < 0.05).

A scatterplot matrix of sulfur-containing compounds was analyzed for rice roots (Fig. 7) and shoots (Fig. 8). In the thiosulfate-treated roots, thiosulfate and sulfite were more strongly positively correlated with other metabolites (Fig. 7A) than those of the roots receiving sulfate (Fig. 7B). The correlation between thiosulfate and sulfite, between thiosulfate and sulfide and between sulfite and sulfide were 100% in rice roots supplied with thiosulfate (Fig. 7A). In shoots supplied with thiosulfate, the thiosulfate content was negatively correlated with sulfite and sulfide (Fig. 8A). In shoots receiving sulfate, the correlations between thiosulfate and sulfite and between thiosulfate and sulfide were positive (Fig. 8B). Correlations among GS-SH, Cys and Cys-S were positive in thiosulfate-treated shoots (Fig. 8A) and negative in shoots supplied with sulfate (Fig. 8B).

Fig. 7

Scatter plot matrix of sulfur-containing compounds in rice roots determined by sulfur index analysis. The blue (A) and red (B) dots in the left triangular matrix show relative contents of metabolites normalized to the internal standard in plants grown with thiosulfate (A) and sulfate (B) as a sulfur source, respectively. The right triangular shows Spearman’s rank correlation and the numbers in ellipse are the coefficients. The darker the blue color of the ellipse, the higher the positive correlation. Asterisks indicate significant correlations (n = 5, pairwise analysis, *** P < 0.001, ** P < 0.01, * P < 0.05).

Fig. 8

Scatter plot matrix of sulfur-containing compounds in rice shoots determined by sulfur index analysis. The blue (A) and red (B) dots in the left triangular matrix show relative contents of metabolites normalized to the internal standard in plants grown with thiosulfate (A) and sulfate (B) as a sulfur source, respectively. The right triangular shows Spearman’s rank correlation and the numbers in ellipse are the coefficients. The darker the blue color of the ellipse, the higher the positive correlation. Asterisks indicate significant correlations (n = 5, pairwise analysis, *** P < 0.001, ** P < 0.01, * P < 0.05).

Discussion

The present study suggests that plants utilize thiosulfate as a sulfur source and assimilate it into other sulfur-containing metabolites. The sulfur index analyses (Figs. 2, 6) show that hydroponic thiosulfate application highly increased thiosulfate levels in the roots and shoots of Arabidopsis and in the roots of rice. Thiosulfate application also altered the profiles of other sulfur-containing compounds especially in the roots. The levels of sulfite, sulfide, Cys-S and GS-SH were higher in thiosulfate-treated roots than they were in sulfate-supplied roots. The oxidation state of sulfur is +2 in thiosulfate and + 6 in sulfate, which may be the reason for higher levels of reduced state compounds like sulfide and Cys-S in thiosulfate-treated plants than in sulfate-treated plants. In bacteria like E. coli, thiosulfate assimilation consumes less energy than sulfate assimilation. The conversion of 1 mol sulfate into Cys require 2 mol ATP and 4 mol NADPH whereas the assimilation of 1 mol thiosulfate requires only 1 mol NADPH. In E. coli, thiosulfate may be preferentially assimilated into Cys when both thiosulfate and sulfate are available (Nakatani et al. 2012). In plants, sulfate must also first be activated by ATP into adenosine-5′-phosphosulfate then reduced to sulfite; therefore, similar to bacteria, thiosulfate may be a preferred sulfur source with reducing power and energy (ATP).

Although thiosulfate conversion may consume less energy than sulfate conversion, thiosulfate treatment with more than 300 μM sulfur concentration in terms of sulfur atoms inhibited Arabidopsis growth relative to those receiving sulfate. Thiosulfate application may have created redox conditions that were too reducing for Arabidopsis. In contrast, growth was similar for rice raised on thiosulfate and sulfate at 300 μM in terms of sulfur atoms. In this study, we used O. sativa L. ‘Koshihikari’, a lowland rice cultivar which prefers relatively reduced conditions compared to that of Arabidopsis. Upland plants like Arabidopsis are sensitive to ammonium toxicity and prefer more oxidized nitrogen forms like nitrate. On the other hand, ammonium is a major inorganic nitrogen source in paddy rice (Kronzucker et al. 2001, Britto and Kronzucker 2002, Tabuchi et al. 2007). Although we cannot exclude the possibility that high concentration of thiosulfate may inhibit rice growth, it is likely that lowland rice plants are more resistant and may prefer reduced sulfur forms like thiosulfate compared to Arabidopsis.

Arabidopsis roots supplied with thiosulfate contained about half as much sulfate as those receiving sulfate (Table 1). Plants may have absorbed sulfate oxidized from the thiosulfate in the medium (Supplementary Fig. S1). Alternatively, they may have a metabolic pathway to oxidize thiosulfate to sulfate. Nevertheless, thiosulfate treatment significantly increased root thiosulfate levels and altered the sulfur index (Figs. 2, 6) relative to the sulfate-treated plants. It is certain, then, that the roots absorbed thiosulfate. The scatterplot matrices of the sulfur-containing compounds significantly differed between the thiosulfate- and sulfate-treated plants (Figs. 3, 4, 7 and 8). Therefore, thiosulfate metabolism at least partially differs from that of sulfate. In Arabidopsis roots and shoots and in rice roots supplied with thiosulfate, there were strong positive correlations between thiosulfate and other metabolites, especially sulfite and sulfide (Figs. 3, 4, 7 and 8). In the shoots of rice supplied with thiosulfate, the correlations between thiosulfate and most other metabolites were negative. Thiosulfate content increased in response to thiosulfate application in Arabidopsis roots and shoots and in rice roots but not in rice shoots. Rice roots may more actively metabolize thiosulfate to other compounds than Arabidopsis roots, thereby leaving comparatively little remaining thiosulfate to translocate to the shoots.

Thiosulfate supply induced the At1g55390 encoding cysteine/histidine-rich C1 domain-containing protein expression the most (Table 3). Twelve other genes encoding Cys-rich proteins were also induced by thiosulfate (Table 3). Thiosulfate did not increase Cys according to the sulfur index analysis (Fig. 2), which may be because Cys concentrations must be maintained to be low in cells because Cys can reduce ferric ions and cause a Fenton reaction (Park and Imlay 2003). It was possible that assimilated Cys, then, is incorporated into Cys-rich proteins and this reaction may be more active in plants grown with thiosulfate than in those treated with sulfate.

The microarray analysis showed that thiosulfate induced the MAM1 gene involved in Met-derived glucosinolates synthesis and repressed the BGLU30 gene possibly involved in glucosinolate catabolism, leading to a possible hypothesis that thiosulfate increased glucosinolate levels. In our widely targeted metabolomics, six Met-derived glucosinolates (GCLs) were significantly decreased while 8-methylthiooctyl-GSL was increased in roots in response to thiosulfate application (Table 5), thus it is as yet unknown whether MAM1 and BGLU30 regulation affected glucosinolate content. It was reported that most of the genes for glucosinolate synthesis, including MAM1, are downregulated and the genes for glucosinolate catabolism, including BGLU30, are upregulated by sulfate deficiency (Maruyama-Nakashita 2017). In the present study, the thiosulfate-treated plants were sulfate-deficient but rich in reduced sulfur. In the previous studies, plants grown under sulfur deficiency with low sulfate levels were lacking in both sulfate and reduced sulfur. For this reason, MAM1 was upregulated and BGLU30 was downregulated in plants treated with thiosulfate despite their sulfate deficiency. The aforementioned genes may be regulated by reduced sulfur as well as sulfate.

Our widely targeted metabolomics analysis indicated that S-adenosyl-L-Met and its derivatives like S-methyl-L-Met, 5′-methyladenosine, 1-aminocyclopropane-1-carboxylic acid, choline phosphate and choline were also significantly decreased in roots treated with thiosulfate. S-adenosyl-L-Met is a Met derivative. In our sulfur index analysis, Met levels were lower in thiosulfate-treated Arabidopsis shoots than they were in the sulfate-treated Arabidopsis shoots (Fig. 2). Met levels were also lower in both the roots and shoots of sulfate-treated rice than they were in those of thiosulfate-treated rice (Fig. 6). It is possible that the sulfur from thiosulfate may be used primarily for thiol or persulfide synthesis rather than the production of Met and its derivatives.

In microarray analysis, the expression of six genes encoding pre-tRNA-Met was repressed by thiosulfate (Table 4), which may be ascribed to less available Met for protein synthesis as Met levels in thiosulfate-treated plants were decreased (Figs. 2, 6). Similarly, 14 genes encoding pre-tRNA-Phe were repressed by thiosulfate (Table 4), which may be due to the decrease in Phe levels by thiosulfate as observed in widely targeted metabolomics (Table 5).

The two 5′-adenylylsulfate reductase genes APR1 and APR2 were induced by thiosulfate application (Table 3). APR genes may be induced by sulfate deficiency in plants supplied with thiosulfate. Low sulfate concentrations in the medium upregulate APR expression (Takahashi et al. 1997, Koprivova et al. 2000). OAS increases under environmental sulfur deficiency. It triggers the upregulation genes encoding sulfur transport and assimilation, including APR (Kim et al. 1999, Ohkama-Ohtsu et al. 2004). OAS was included in the sulfur index analysis and the widely targeted metabolomics. In the present study, however, it was below the detection limit in both analyses and for both Arabidopsis and rice treated with sulfate or thiosulfate. OAS actively participates in Cys synthesis by combining with sulfide in plants supplied with thiosulfate since sulfide increases in response to thiosulfate application. Therefore, a signal other than OAS upregulated APR genes under thiosulfate treatment.

Based on the combined results of our sulfur index and microarray analyses and those of our previous studies on thiosulfate metabolism, we propose a possible plant thiosulfate assimilation pathway as shown in Fig. 9. Thiosulfate absorbed in cells is transported to the plastids where it reacts with OAS to produce S-sulfocysteine. This reaction is partially catalyzed by CS26 (Bermúdez et al. 2010, Birke et al. 2015). In E. coli, S-sulfocysteine is then reduced to Cys by NADPH releasing sulfite (Nakatani et al. 2012). The enzyme responsible for S-sulfocysteine reduction in plants has not yet been identified. Nevertheless, there was a remarkable increase in sulfite in response to thiosulfate application according to our sulfur index analysis (Figs. 2, 6). There was a strong correlation between thiosulfate and sulfite in Arabidopsis and rice supplied with thiosulfate (Figs. 3, 4 and 7). This correlation supports the existence of this pathway in plants. Sulfite is reduced to sulfide by sulfite reductases. The next step is Cys-S synthesis (Maruyama-Nakashita and Ohkama-Ohtsu 2017). In our sulfur index analysis, the sulfide level was increased and it may have reacted with Cys to produce Cys-S. An increase in Cys-rich proteins suggest that they are pools for reduced sulfur. Increases in sulfite, sulfide and Cys-S in response to thiosulfate treatment were observed in sulfur index analyses for Arabidopsis (Fig. 2) and rice (Fig. 6). Therefore, our putative thiosulfate assimilation model (Fig. 9) is also applicable to rice. The thiosulfate assimilation pathway in plants should consume less energy than that for sulfate assimilation as is the case with bacteria (Nakatani et al. 2012). In S. cerevisiae, thiosulfate promoted growth more than sulfate (Funahashi et al. 2015). No such difference in growth promotion was observed for either Arabidopsis or rice in our study. Although rice prefers reducing conditions, its growth rates were similar under both the thiosulfate and sulfate treatments in our condition. Further investigation is needed to determine whether using thiosulfate as a sulfur source enhances plant growth.

Fig. 9

Plant thiosulfate assimilation models. Red arrows indicate proteins whose genes were induced by thiosulfate. Blue arrows indicate compounds increased or decreased by thiosulfate.

Materials and Methods

Arabidopsis growth condition

Unless otherwise indicated, wild-type A. thaliana Columbia-0 were grown hydroponically in 50-ml conical tubes (Ohkama-Ohtsu et al. 2007) at 22°C under short-day with 8 h light (200 μmol⁻¹ m⁻² s⁻¹)/16 h dark. The MGRL medium (Fujiwara et al. 1992, Hirai et al. 1995) was used as a nutrient source and slightly modified as follows. Magnesium was supplied as MgCl₂. Sulfur in the medium was adjusted by adding Na₂SO₄ or Na₂S₂O₃. Hydroponic solutions were replaced twice weekly. For the sulfur index, widely targeted metabolomics, sulfate and microarray analyses, the shoots and roots of plants grown for 8 weeks under short-day conditions were harvested separately, immediately frozen with liquid N₂, and stored at −80 °C until extraction.

Measurement of chlorophylls a and chlorophyll b contents

Chlorophylls were extracted from Arabidopsis shoots by soaking them in dimethylformamide and chlorophylls a and chlorophyll b contents in the extract were determined by spectrometric analysis according to Porra et al. (1989).

Rice growth condition

Oryza sativa L. ‘Koshihikari’ seeds were surface-sterilized with sodium hydrochloride were germinated at 28 °C on nets floating in water purified by reverse osmosis. After 6 d, the seedlings were transferred to hydroponic medium (Makino et al. 1988) in 50-ml conical tubes at a density of one plant per tube. They were grown at 28 °C under 16 h light (250 μmol⁻¹ m⁻² s⁻¹)/8 h dark. Magnesium was supplied as equimolar MgCl₂. Sulfur in the medium was adjusted to 300 μM with Na₂SO₄ or Na₂S₂O₃. Hydroponic solutions were replaced twice weekly. After 4 weeks, the roots, third leaves, fourth leaves and the other parts of the shoots were harvested separately, weighed, immediately frozen in liquid N₂ and stored at −80 °C until the sulfur index analysis.

Evaluation of chlorophyll content in rice leaves by SPAD values

At 4 weeks after germination, rice plant chlorophyll content was evaluated using SPAD-502 Plus (KONICA MINOLTA, Tokyo, Japan). The SPAD values in the center of the third and fourth leaves were determined.

Sulfur index analysis

Plant samples were extracted and derivatized with mBBr as described in Minocha et al. (2008) and in Nishida et al. (2016). D-camphor-10-sulfonic acid sodium salt 10 μM was used as an internal standard. In our analysis, the reductant tris (2-hydroxyethyl) phosphine hydrochloride was eliminated in derivatization. Sulfur index analysis with LC-MS/MS was performed as described in Kawano et al. (2015). The relative contents of metabolites were calculated by normalizing their peak areas to the area of internal standard d-camphor-10-sulfonic acid sodium salt as unity in each sample.

Microarray analysis

Microarray analysis with Arabidopsis was performed in three biological replicates with sulfate or thiosulfate. Total RNA was extracted using the RNeasy mini kit (Qiagen, Hilden, Germany) combined with the RNase-Free DNase Set (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. The cRNA synthesis and the subsequent single-strand cDNA synthesis were performed with the Affymetrix GeneChip WT Plus Reagent Kit (Thermo Fisher Scientific, Waltham, MA, USA). The single-strand cDNAs was labeled with the Affymetrix GeneChip WT Terminal labeling kit (Thermo Fisher Scientific, Waltham, MA, USA). The cDNAs were hybridized with the Affymetrix Gene Chip Hybridization, Wash and Stain Kit (Thermo Fisher Scientific, Waltham, MA, USA) to the Affymetrix GeneChip^TM Arabidopsis Gene 1.0 ST Array (Thermo Fisher Scientific, Waltham, MA, USA). The scanned fluorescent images were analyzed with Affimetrix GeneChip Command Console Software (Thermo Fisher Scientific, Waltham, MA, USA) to generate a CEL file and with the Affimetrix Expression Console (Thermo Fisher Scientific, Waltham, MA, USA) to generate a CHP file. The CHP file was analyzed with Transcriptome Analysis Console Software (v. 4.0, 64-bit) (Thermo Fisher Scientific, Waltham, MA, USA). The Robust Multiarray Average method (Irizarry et al. 2003) was used to normalize the microarray signal values. The probes which were significantly differentially expressed with fold change (log₂-FC) < −2, or >2 between the thiosulfate and sulfate applications (Empirical Bayes test; P < 0.05) were selected for subsequent gene ontology (GO) analysis. In the GO analysis, categories matched with the first threshold (corrected P-value <0.1) were selected. If the first threshold was not found, the second threshold (corrected P-value <0.001) was used.

Analysis of sulfate, thiosulfate, nitrate and phosphate by ion chromatography

Sulfate, thiosulfate and phosphate in the medium were determined by ion chromatography (LC-20AD, CDD-10Asp; Shimadzu Corp., Kyoto, Japan) fitted with the Shim-pack IC-SA2 column (Shimadzu Corp., Kyoto, Japan). The mobile phase solution contained 12 mM NaHCO₃ and 0.6 mM Na₂CO₃ and flowed isocratically at 1.0 ml min⁻¹.

Sulfate, nitrate and phosphate concentrations in plant tissues were determined according to the methods of Maruyama-Nakashita et al. (2015) and Yamaguchi et al. (2016).

Analysis of total carbon and nitrogen

Total carbon (C), hydrogen (H) and nitrogen (N) contents were analyzed using 1–2 mg of plant powder in an elemental analyzer (Yanaco CHN Corder MT-5; Yanako, Kyoto, Japan).

Analysis of total sulfur

One to two milligrams of plant powder was digested with 200 µl concentrated nitric acid. Digestion was conducted in a heat block for 30 min at 95°C, followed by digestion and evaporation for 45 min at 120°C. The clear, digested samples were diluted to 1.5 ml with extra pure water. The digested samples were further diluted with extra pure water to 80 times and analyzed by ion chromatography (IC-2001, TOSOH, Japan). Using serial 30 µl injections, the diluted extracts were separated at 40°C using a TSK SuperIC-AZ column (TOSOH) at a flow rate of 0.8 ml min⁻¹ and with an eluent containing 7.5 mM NaHCO₃ + 1.1 mM Na₂CO₃ (Wako Pure Chemicals, Osaka, Japan). Anion mixture standard solution 1 (Wako Pure Chemicals) was used as a standard.

Widely targeted metabolomics

Plant shoots and root were lyophilized and 4 mg of powdered tissues were extracted with 1 ml extraction solvent consisting of 80% MeOH (v/v) and 0.1% formic acid (v/v). Two internal standards were used: 8.4 nM lidocaine (positive ion mode) and 210 nM camphor sulfonic acid (negative ion mode). About 25 µl of extracted solvent was lyophilized and 250 µl of LC-MS grade water was added to it. About 40 ng µl⁻¹ sample solution was measured by LC-QqQ-MS/MS (UPLC-TQS, Waters, Milford, MA, USA). The detection conditions of the metabolites were described in previous reports (Sawada et al. 2009, Sawada et al. 2017). The data matrices (sample × metabolites) were generated by mass lynx (Waters, Milford, MA, USA). The data were analyzed in R using the function and packages as follows: Welch’s t-test, “t.test” function; heatmap analysis, “pheatmap” package; PCA, “FactoMineR” package; volcanoplot analysis, “ggplot2” and “ggrepel” packages.

Statistical analysis

Unless otherwise stated, statistical analyses were performed in R or SPSS 23 (IBM, Armonk, NY., USA). The scatterplot matrix was generated by the pairs function in R.

Funding

JSPS KAKENHI [16K07639 and 15KT0028 to N.O.-O.].

Disclosures

The authors have no conflicts of interest to declare.

Acknowledgments

Microarray analysis with total RNAs was performed by Riken Genesis Co. Ltd. (Tokyo, Japan). Analysis of total carbon and nitrogen was conducted at The Service Center of the Elementary Analysis of Organic Compounds, Graduate School of Sciences, Kyushu University.