Sulphur systems biology-making sense of omics data.

Systems biology approaches have been applied over the last two decades to study plant sulphur metabolism. These 'sulphur-omics' approaches have been developed in parallel with the advancing field of systems biology, which is characterized by permanent improvements of high-throughput methods to obtain system-wide data. The aim is to obtain a holistic view of sulphur metabolism and to generate models that allow predictions of metabolic and physiological responses. Besides known sulphur-responsive genes derived from previous studies, numerous genes have been identified in transcriptomics studies. This has not only increased our knowledge of sulphur metabolism but has also revealed links between metabolic processes, thus indicating a previously unexpected complex interconnectivity. The identification of response and control networks has been supported through metabolomics and proteomics studies. Due to the complex interlacing nature of biological processes, experimental validation using targeted or systems approaches is ongoing. There is still room for improvement in integrating the findings from studies of metabolomes, proteomes, and metabolic fluxes into a single unifying concept and to generate consistent models. We therefore suggest a joint effort of the sulphur research community to standardize data acquisition. Furthermore, focusing on a few different model plant systems would help overcome the problem of fragmented data, and would allow us to provide a standard data set against which future experiments can be designed and compared.

Introduction

Regulation of cellular processes in plants occurs as the result of developmental programmes or the necessity to respond to external signals. Regulation does occur at all tiers of information conversion from DNA to physiology. Epigenetics, histone and chromatin modifications, regulation of transcription, of translation, of protein stability, and at the level of enzyme activity, all modify the response flow. At the next layer, communications between cells and tissues coordinate the responses through hormones and long-distance signals to attain the physiological responses necessary for survival and propagation.

Advances in technical capabilities for measuring large numbers of biochemical compounds (Fiehn ) have allowed high-throughput analyses to be performed on biological materials, producing large increases in the quantities of data obtained. The aim of data cataloguing and the definition of systems biology as we consider it here (Klipp ) is to measure all available components and functions of a cell or tissue, or at least as many as possible, and to use mathematical modelling to understand the underlying network and cooperativity. The intention of this approach is to yield a holistic understanding of a system by capturing all its parts at a given state or condition. This additionally includes system-wide responses that, based on our current understanding, may seemingly be unrelated to n class="Chemical">sulphur nutrition, and which would be missed in targeted approaches. The ultimate goals are to identify candidate genes, elucidate gene functions, and understand physiological processes (Fiehn ; Stitt and Fernie, 2003). Systems biology should also aim to look at the dynamics of the response of a system over time or in relation to developmental states.

One of the first example of ‘omics’ approaches was DNA sequencing, which eventually yielded the sequence information of whole organisms, for example Arabidopsis thaliana (The Arabidopsis Genome Initiative, 2000), thus laying the grounds for the area of genomics. There were obviously precursors of high-throughput analyses when sequence information was not yet available, such as amplified fragment length polymorphism (AFLP) mapping (Howarth ). Over the last two decades, this has been followed by a wide variety of new, upcoming omics-technologies aimed at determining the various functional entities in a cell, namely DNA, RNA, metabolites, proteins, enzyme activities, and flux dynamics. Improvements in technical approaches continue to provide data even more efficiently and faster, for example high-throughput sequencing technologies such as RNA sequencing (RNA-seq) (Weber, 2015). Coupled to this is an increasing need for improved biostatistical approaches, generally termed bioinformatics, to catalogue and analyse the data being generated (Rhee ).

Of these ‘omics’ approaches, metabolomics analyses provide the most complex data sets because the analytes are constituted of chemical compounds with a huge range of molecular masses and diverse physicochemical properties (such as hydrophobicity and ionic strengths) and have to be extracted and analysed using multiple methods (Fiehn ; Stitt and Fernie, 2003; Watanabe ). In contrast, DNA and RNA in genomics and transcriptomics are composed of only four different, closely related chemicals, nucleotides, with only a few modifications that add only slightly to their complexity. The initial concept was to use unbiased approaches to capture the metabolic state of a system, that is without challenging the plant with, for example, a stress such as sulphate-deficient conditions (Roessner ; Kusano ). Soon, however, environmental challenges were applied or different developmental states compared in order to obtain more informative data. Compared to classical differential screening approaches, systems biology produces a multitude of data in order to obtain a holistic response pattern rather than concentrating on individual parts of metabolism. An iterative analytical phase has now been reached where the results of omics-based screening approaches that have identified candidate genes are being subjected to (for example) reverse genetics for a further round of validation again by omics approaches (Aarabi ). This is still the classical reductionist approach, but with the intent of analysing the system in a holistic manner. The eventual goal of systems approaches with respect to sulphur metabolism is to unravel gene function and to generate a network scheme and a model for plant sulphur metabolism that allows predictive biology.

In this review, we concentrate on the systems biology of plant sulphur metabolism (‘sulphur-omics’) in Arabidopsis, especially on transcriptome, metabolome, and proteome studies. These approaches began in the early years of this century with the first papers on transcriptomics using DNA arrays (Hirai ; Maruyama-Nakashita ; Nikiforova ). We try to answer the question as to whether ‘omics’ approaches have provided us with a better understanding, and we discuss what needs to be done in the future to further enhance our knowledge. Finally, we ask whether we are already at a level of understanding to bring about agronomical improvements in terms of sulphate use efficiency, crop quality, and production of medicinal compounds.

What is the status of sulphur-omics?

Conceptually, systems approaches are designed to provide as much and as unbiased data as possible. These data provide initial information on the subject of investigation, here plant sulphur metabolism. For data validation, an iterative cycle of analytical procedures is necessary (Fig. 1A), aimed at understanding the molecular responses and physiological processes regulating plant sulphur metabolism and homeostasis. The initial results are then analysed in an iterative process using systems or more targeted approaches to yield a better understanding of the system, for example by employing altered conditions or mutants of selected candidate genes identified through forward and reverse genetics.

Fig. 1.

Schematic representation of sulphur systems biology. (A) Experimental flow chart. Raw data is acquired by employing ‘omics’ technologies to results from plants exposed to altered S availability and by conducting targeted experiments. These data are then analysed by various data-mining steps such as annotation of genes, metabolites, or proteins. External knowledge is included for data mining. Bioinformatics analyses, again with the use of external knowledge, help to sort the data and to identify significant changes or correlations. Data interpretation leads to the identification of candidate genes, proteins, or processes, which results in the generation of hypotheses or models to explain the observed responses. These models need to be validated in an iterative analysis, for example using transgenic approaches, mutants, or altered conditions, i.e. the identified candidates are subjected to a second analytical cycle. Eventually, data interpretation should result in the formulation of a theory that explains the aspects of S metabolism that have been investigated, and thus our knowledge of plant S metabolism is increased. This knowledge may then be exploited for plant breeding and the generation of new crop varieties. Crop quality validation and field testing might employ the same analytical circle. (B) Identification of the function of SDI genes. An example of an experimental flow chart focusing on SDI genes using an iterative research cycle. AFLP, amplified fragment length polymorphism; GLS, glucosinolate; indOX, inducible overexpression; KO, knockout; OAS, O-acetylserine; OX, overexpression; qRT-PCR, quantitative reverse transcription-polymerase chain reaction; QTL, quantitative trait locus; SDI, sulfur deficiency-induced; SERAT, serine acetyltransferase; slim1, sulfur limitation1; SNP, single nucleotide polymorphism; TF, transcription factor.

Transcriptomics of Arabidopsis with respect to sulphur metabolism

Arabidopsis is well established as a valuable model plant system (Scholl ). The release of its genome sequence (The Arabidopsis Genome Initiative, 2000) increased its value as a research tool and boosted approaches aimed at deciphering holistic rather than individual plant responses to particular conditions. Such systems approaches have been applied to study plant sulphate metabolism, with the aim of systematically unravelling the molecular responses. While early analyses focussed on gene and metabolite responses of sulphate uptake, reduction, and assimilation processes, systems approaches have sought to examine the connections and interplay within the system as a whole. This has been based on the inherent assumption that a response to (e.g.) sulphate starvation will not only affect sulphur metabolism per se but also other interconnected and downstream processes. It is obvious that in this context different parts of plants, such as roots, leaves, or seeds, will show both general and also specific responses. Furthermore, developmental aspects have to be taken into account when analysing plant responses to sulphate withdrawal or resupply over time. Transcriptomics studies as part of sulphur systems biology were pioneered by Hirai , Nikiforova , and Maruyama-Nakashita using Arabidopsis. A transcriptome analysis of sulphur (S-)deprived Arabidopsis seedlings was performed for different durations of S starvation in order to address the development of the response over time (Nikiforova ). In addition, Arabidopsis seedlings were treated with O-acetylserine (OAS), the immediate precursor of cysteine biosynthesis (Hirai ). OAS accumulates upon S deprivation and early research considered it as an S-starvation signal (Saito, 2000; Hopkins ), which was indeed subsequently demonstrated (Hubberten ). Due to the technical limitations at the time, these pioneering studies were performed on macro-arrays, each comprising about 10 000 random cDNAs (Hirai ; Nikiforova ), or using Affymetrix 8K chips with probes for ~8000 genes (Maruyama-Nakashita ). The differentially expressed genes that were identified included some that were already known to be responsive to S status, such as sulphate transporters (Smith ; Hawkesford, 2000), which confirmed the validity of the approach. More interestingly, information on novel genes was obtained. Thus, alongside known genes, these early studies provided a catalogue of genes that as yet had no assignment of their function in response to S-deficient growing conditions.

The number of transcriptome studies that have been conducted on Arabidopsis is still quite low with only 14 in total (Table 1). It is justifiable to include arrays of plants exposed to selenium (Van Hoewyk ) as it acts as a competitor with sulphur, thus mimicking S deprivation. Among related Brassicaceae species a sulphate starvation study was performed on rapeseed (Buhtz et al., 2008, 2010). Despite the fact that rapeseed has a high requirement for sulphate (Girondé ), there is a lack of transcriptomics studies on this subject. With respect to Arabidopsis, the tissues and conditions investigated in the early studies were already quite diverse (Table 1). This provided a wealth of information, but it made comparisons between studies difficult as each experiment was based on very specific conditions with respect to the sulphate levels applied and/or the tissues examined. For example, the tissues studied in response to S-deprivation included whole seedlings grown on agar plates (Nikiforova ), seedlings separated into leaves and roots (Hirai ; Maruyama-Nakashita , 2005, 2006), and developing seeds (Higashi ). Subsequent studies have examined whole seedlings exposed to S deprivation in submerged seedling cultures followed by resupply in order to score for recovery processes (Bielecka ), hydroponically grown root tissues exposed to S deprivation and separated into fractions of various cell types (Iyer-Pascuzzi ), and studies where S deprivation has been one factor among other combined stresses (Barciszewska-Pacak ; Forieri ). Sulphate starvation has been used as a condition to investigate phloem-specific micro-RNAs in rapeseed (Buhtz ). Although only a subset of the phloem RNA fraction was analysed, results regarding the regulatory function of miRNA-395 were substantiated in further studies employing Arabidopsis (Kawashima et al., 2009, 2011). Sulphate metabolism in response to acid rain conditions has been investigated, with high inputs of S under low pH conditions (Liu ). Acid rain is an ecological and a health problem in many countries due to combustion of fossil fuels releasing SO2. In North America and Europe, SO2 emissions have been successfully reduced over recent decades due to legislative measures that have regulated industrial and domestic use of fossil fuels. However, this has consequently reduced sulphur inputs into agro-ecological systems, which in turn has triggered research into its agricultural impact (Haneklaus ; Menz and Seip, 2004). From the molecular perspective, several transcriptome datasets on wheat in relation to responses to S nutrition may be the primary resource for studying the effects of sulphur inputs (Table 1).

Table 1.

Transcriptome analyses related to sulphur metabolism

Experiment	Species	Tissue	Type	ID	References
–S	Arabidopsis	Seedling	Macroarray		Nikiforova et al. (2003)
		Leaf, Root	Macroarray		Hirai et al. (2003)
		Leaf, Root	Affymetrix 8K Chip		Maruyama-Nakashita et al. (2003)
		Leaf, Root	Agilent oligo microarray	E-MEXP-211	Hirai et al. (2005)
		Root	GeneChip ATH1	GSE5688	Maruyama-Nakashita et al. (2005)
		Seed	GeneChip A-AFFY-2	E-ATMX-1	Higashi et al. (2006)
		Root	GeneChip ATH1	GSE4455	Maruyama-Nakashita et al. (2006)
		Root cell types	GeneChip ATH1	GSE30100 GSE30099 GSE30098	Iyer-Pascuzzi et al. (2011)
		Root cell types	Agilent-custom promoter array	GSE30166	Iyer-Pascuzzi et al. (2011)
		Seedling	GeneChip ATH1	GSE64972	Bielecka et al. (2015)
		Seedling, Leaf	Illumina HiSeq 2000	GSE66599	Barciszewska-Pacak et al. (2015)
		Leaf, Root	GeneChip ATH1	GSE81347	Aarabi et al. (2016)
		Leaf	GeneChip 1.0 ST	GSE93048	Dong et al. (2017)
		Root	GeneChip 1.1 ST	GSE77602	Forieri et al. (2017)
	Oilseed rape	Leaf, Root, Phloem	LC Sciences dual colour	GSE20263	Buhtz et al. (2010)
+Se	Arabidopsis	Leaf, Root	GeneChip ATH1	GSE9311	Van Hoewyk et al. (2008)
–O2	Arabidopsis	Seedling	GeneChip ATH1		Branco-Price et al. (2008)
+acid rain S	Arabidopsis	Leaf	GeneChip ATH1	GSE52487	Liu et al. (2014)
–S	Triticum aestivum	Leaf	GeneChip Array	E-MEXP-1415	Howarth et al. (2008)
		Root	GeneChip Array	E-MEXP-1694	Bo et al. (2014)
		Root	GeneChip Array	GSE61679	Gupta et al. (2017)
		Grain	Illumina HiSeqTM PE125/PE1		Yu et al. (2018)
		Grain	NimbleGen microarray	E-MTAB-1782	Dai et al. (2015)
		Grain	NimbleGen microarray	E-MTAB-1920	Vincent et al. (2015)

Transcriptome analyses related to n class="Chemical">sulphur metabolism

A common feature of all systems biology approaches is that they yield vast amounts of data (Kopriva ). Hence, statistical methods have had to be developed or adapted to deal with this (Klipp ; Xia, 2018). In the context of n class="Chemical">sulphur systems biology, such methods were already being applied to the early transcriptomics data sets. Especially when attempting to correlate transcriptomics and metabolomics data (Nikiforova ), it was inevitably necessary to apply bioinformatics approaches in order to allow data interpretation and the development of models (Hirai ; Hirai and Saito, 2004; Nikiforova et al., 2004, 2005a). Results are often displayed as correlation networks (Nikiforova ). This kind of approach is aimed at filtering the data to remove the ‘noise’ of variability associated with gene expression and metabolite contents, and in doing so to highlight differences that are statistically significant (Massonnet ).

One constraint of systems approaches such as transcriptomics, proteomics, or metabolomics is the fact that even if concentration differences per se are determined, they may not represent changes in activities of relevant proteins or enzymes, or of metabolite fluxes. An example of such a situation where transcriptomics would not reveal an important gene is the transcription factor sulfur limitation1 (SLIM1, AT1G73730), which has been identified through genetic screening of Arabidopsis mutants (Maruyama-Nakashita ) and has been shown to control a major part of the S-starvation response (Kawashima ; Wawrzyńska and Sirko, 2014). As far as current data suggest, SLIM1 itself is not, or is only marginally, transcriptionally regulated upon S deprivation. EIN3 (AT3G20770), a major factor involved in ethylene signalling, has been shown to modulate SLIM1 binding activity to its target gene promoters (Wawrzyńska and Sirko, 2016). As the authors suggest, this probably interferes with the S deficiency-dependent induction of target genes by SLIM1. However, they do not exclude the possibility that further regulators might be involved in shaping the response to S deprivation. To unravel the complexity of the regulation of plant S metabolism it is therefore obvious that despite the wealth of data provided by systems approaches, targeted analyses need to be combined in order to reveal the cellular and physiological responses to S deprivation (Fig. 1A).

Deposition of systems biology results in databases allows data to be revisited when new knowledge is available, such as improved gene annotation, and this can not only confirm initial assumptions but also provide novel information (Fig. 1A; Nikiforova ; Hoefgen and Watanabe, 2017). Recently, Henríquez-Valencia have conducted a comparative meta study using existing data sets together with novel bioinformatics approaches. This led to the identification of transcription factor networks that provide new candidate genes for sulphate research that would not otherwise have been identifiable in individual experimental set-ups. This also highlights the need for further transcriptomics studies to be provided to the scientific community to advance our knowledge. With increasing depositions of data related to S metabolism, including data on species other than Arabidopsis (Table 1), such approaches will have a greater impact on the generation of hypotheses. New candidate genes and biochemical processes interconnected to plant S metabolism will be identified as a result of these systems-based and targeted approaches. It is a matter of ongoing debate, probably driven by individual research interests, as to whether only ‘robust’ processes that occur under a variety of conditions and in various plants are relevant or whether ‘specific’ responses that occur under only certain conditions are the most meaningful for improving our understanding of plant sulphur physiology.

While initial high-throughput analyses can lead to the generation of hypotheses (Nikiforova et al., 2003, 2005a, 2005b), these need to be tested experimentally for further validation (Fig. 1A). Cataloguing alone is insufficient to develop knowledge of processes and, eventually, to exploit them for plant breeding and crop production. Validation efforts necessarily need to employ all levels of molecular biology and bioinformatics-based approaches in an iterative manner (Hoefgen and Watanabe, 2017). An example is the investigation of predicted hub genes (Nikiforova ) through a mutational approach (Falkenberg ). Three transcription factors, IAA13, IAA28, and ARF-2 (ARF1-Binding Protein), in a network responsive to S deprivation have been identified as being connected to multiple downstream and upstream interactors, and thus constitute hubs, making it likely that they represent important genes (Mähler ). Falkenberg subsequently showed that these transcription factors indeed play a role in controlling certain aspects of plant sulphate metabolism, and thus validating the assumption that identification of correlative network hubs is indeed a tool that can be used to identify relevant target genes—in this case linking S deprivation to auxin signalling. In fact, IAA28 may constitute the link between auxin signalling, S starvation, and alterations in root development (Rogg ; Falkenberg ; De Rybel ), although this remains to be demonstrated functionally. A link to auxin had been postulated previously (Nikiforova ). A further example is the identification of the functional roles of sulfur deficiency induced 1 (SDI1) and SDI2 (Fig. 1B). An AFLP study on wheat identified SDIs as being strongly responsive S-deprivation genes (Howarth ) and they were also identified in early macroarray studies on S-deprived and OAS-treated Arabidopsis (Hirai ; Nikiforova ). However, the function of the SDI genes was not clear from these initial studies. A combination of a bioinformatics approach to OAS-related responses (Hubberten ) and a mutational approach coupled with transcriptomics and metabolomics analyses (Aarabi ) revealed that SDI1 and SDI2 interact through protein–protein binding with a previously described transcription factor, MYB28. Upon S deprivation in Arabidopsis, this binding down-regulates MYB28 transcription and consequently reduces the biosynthesis of glucosinolates (Gigolashvili ; Sønderby ). In functional terms, this may divert S resources from secondary to primary metabolism. Interestingly, Hubberten additionally revealed a group of OAS-responsive genes that are co-regulated under various conditions, termed OAS-cluster genes. Co-regulated expression hints at the existence of common upstream regulatory control mechanisms, which would be worth investigating.

Transcription factors of Arabidopsis related to sulphur metabolism

The responses of plant S metabolism to changes in the availability of sulphate are well described (Davidian and Kopriva, 2010). In terms of regulation, several candidate genes have been identified (Table 2). ‘Omics’ approaches do not usually identify post-transcriptional or post-translational modifications unless they are specifically designed to indicate modifications such as persulphidation (Aroca ) or DNA methylation (Huang ). Proteomics approaches are suited to identify protein modifications but the number of such studies on responses to S deprivation is low, even for Arabidopsis. Even less information is available regarding the signal molecules that induce the S-deprivation response. Results that indicate the involvement of phosphorylation originate from targeted and not from systems analyses, except for the potential involvement of sucrose non-fermenting-1-related protein kinases (SNRKs; Iyer-Pascuzzi ) that is suggested based on transcriptomics of different root cell types in Arabidopsis starved of sulphate, nitrate, and phosphate. OAS has been considered as a potential signal and evidence has accumulated to substantiate its signalling function (Saito, 2000; Hirai and Saito, 2004; Hubberten et al., 2012a, 2012b; Aarabi ). But exactly how OAS is sensed is still unknown as neither the receptor nor the signal transduction chain has yet been identified, although SDI1 and SDI2 seem to be induced by OAS. In addition to OAS, several other intermediates of the sulphate assimilation pathway have been suggested as signals, including sulphate (Rouached ), sulphite (Brychkova ; Naumann ), sulphide with its role in persulphidation (Ma ; Aroca ), glutathione (GSH), and cysteine. The problem is that the sulphate assimilation pathway reacts to changes in any of its metabolite concentrations with correlated changes of other metabolites of the pathway, making it difficult to discern individual effects. For OAS, this could be experimentally resolved by expressing a serine acetyltransferase (SERAT) gene under control of an inducible promoter, and by the finding that OAS is possibly related to stress-induced reactive oxygen species (ROS) that are induced under conditions where no further changes of the S-containing metabolites are detected (Hubberten ). Receptors have not been identified in transcriptomics studies. A mutational approach might provide this information, but although such an approach did identify SLIM1 (Maruyama-Nakashita ) as a transcription factor (TF) that controls certain parts of the sulphate starvation response, it did not identify a receptor of S-containing metabolites. Various hormones have been generally implicated in regulating aspects of sulphate metabolism (Falkenberg ; Amtmann and Blatt, 2009; Gojon ; Rubio ; Wawrzyńska and Sirko, 2016) but their exact involvement remains still elusive. A cytokinin receptor, CRE1/AHK4 (cytokinin response 1/Arabidopsis histidine kinase 4), has been identified and suggested to play a role in the regulation of sulphate uptake (Maruyama-Nakashita ). This receptor has been previously determined to modulate phosphate starvation responses by inhibiting phosphate transporter expression and to down-regulate sulphate uptake in roots under conditions with sufficient P and S supply (Gojon ). The TF phosphate starvation response1 (PHR1) is also known to be associated with control of sulphate metabolism as shoot-to-root sulphate transport is affected in phr1 mutants and the accumulation of sulphoquinovosyl diacylglycerol (SQDG) decreases about 2-fold relative to the wild-type under P-deprived conditions (Rouached ; Pant ). These findings indicate the existence of crosstalk between P and S metabolism (Gojon ).

Table 2.

Transcription factors and regulators suggested to be associated with S metabolism

Transcription factor	AGI Code	Regulation	References
SLIM1	AT1G73730	S response, S metabolism	Maruyama-Nakashita et al. (2006)
HY5	AT5G11260	S assimilation (APR)	Lee et al. (2011); Koprivova and Kopriva (2014)
MYB28	AT5G61420	Aliphatic glucosinolate	Celenza et al. (2005); Bielecka et al. (2015); Gigolashvili et al. (2007a, 2008); Hirai et al. (2007); Sønderby et al. (2007); Malitsky et al. (2008); Davidian and Kopriva (2010)
MYB29	AT5G07690	Aliphatic glucosinolate
MYB76	AT5G07700	Aliphatic glucosinolate
MYB34	AT5G60890	Indolic glucosinolate
MYB51	AT1G18570	Indolic glucosinolate
MYB122	AT1G74080	Indolic glucosinolate
SDI1	AT5G48850	Glucosinolate (MYB)	Aarabi et al. (2016)
SDI2	AT1G04770	Glucosinolate
NF-YA2	AT3G05690	Development, S, N, P responses	Henríquez-Valencia et al. (2018)
RVE2	AT5G37260	Germination, Circadian rhythm
MSA1 (SHM7)	AT1G36370	S-adenosylmethionine	Huang et al. (2016)
PHR1	AT4G28610	S, P responses, Sulphate shoot-to-root flux	Gojon et al. (2009); Rouached et al. (2011); Pant et al. (2015)
IAA28	AT5G25890	Auxin signalling, lateral root	Falkenberg et al. (2008)
IAA13	AT2G33310	Auxin signalling, embryonic root
ARF1-BP (ARF2)	AT5G62010	Auxin signalling, plant ageing
OBP2 (DOF)	AT1G07640	Glucosinolate	Skirycz et al. (2006)
Calmodulin binding IQD protein (IQD1)	AT3G09710	Glucosinolate	Levy et al. (2005)
miRNA395		Sulphate transporter (SULTR2;1), Sulphur assimilation (ATPS)	Kawashima et al. (2009, 2011); Buhtz et al. (2010); Liang et al. (2012)
EIN3	AT3G20770	Ethylene signalling, SLIM1	Wawrzyńska and Sirko (2016)
ARF12	AT1G34310	Auxin response, root development	Bielecka et al. (2015)
ARR16	AT2G40670	Cytokinin signalling, root
ATAF1 (NAC)	AT1G01720	Abscisic acid biosynthesis
CO-like Yabby	AT1G73870	Auxin homeostasis
DREB A-4	AT2G44940	–
HAT14 (HB)	AT5G06710	–
MADS	AT4G33960	–
MYB9	AT5G16770	Suberin in seed coat
MYB31	AT1G74650	–
MYB45	AT3G48920	–
MYB52	AT1G17950	Secondary cell wall
MYB53	AT5G65230	Lateral root
MYB54	AT1G73410	Secondary cell wall
MYB71 (MYB305)	AT3G24310	–
MYB75 (PAP1)	AT1G56650	Anthocyanin
MYB93	AT1G34670	Lateral root
Trihelix	AT3G10040	Hypoxia response
WRKY56	AT1G64000	–
ZAT12 (C2H2)	AT5G59820	Abiotic/oxidative stress
ZAT6 (C2H2)	AT5G04340	S&P response, root development

A comparative study dedicated to identifying TFs that respond to sulphate deprivation and resupply but not to N or P starvation (Bielecka ) yielded several candidates (Table 2). In the same study a set of known TFs, in particular those related to aliphatic and indolic glucosinolate biosynthesis (MYB28, 29, 76, 34, 51, 122), were correlated with the expression of glucosinolate pathway genes. Furthermore, TFs shown to be responsive to S availability included those regulating anthocyanin biosynthesis, mainly MYB75 (PAP1, production of anthocyanin pigment1) together with a set of TFs probably controlled by PAP1, namely MYB90 (PAP2), MYB113, and MYB114 for anthocyanin; TT8 (TRANSPAREN TTESTA 8), bHLH, TTG1 (TRANSPARENT TESTA, GLABRA1), WD40, and TTG2 (TRANSPARENT TESTA GLABRA2), and WRKY for flavonoids. Such a link to anthocyanin and flavonoid biosynthesis is obvious as S-starved plants (in common with N- and P-starved plants) display accumulation of reddish pigments in leaves (Nikiforova ; Wulff-Zottele ). Although induced by S deprivation, these TFs might be also be part of a more general stress rescue system (Whitcomb ). Most of the TFs suggested by Bielecka have not yet been validated through further analyses, but they provide a valuable data resource for future research. Unravelling the signals, the receptors, the signalling cascade, the TFs, and other regulators that control plant S metabolism is an ongoing challenge.

Metabolomics of Arabidopsis with respect to sulphur metabolism

The sum of all the metabolites in a cell or tissue is referred to as the metabolome. The number of metabolomics studies dedicated to the response to S deprivation or S resupply in Arabidopsis is low (Table 3). At the level of primary metabolite composition, we postulate that most plants will share related responses as primary metabolites constitute those metabolic pathways that are present in all plants to serve the basic functions of life (Pichersky and Gang, 2000). Thus, it is justifiable to compare S-related metabolome studies even between different species (Fig. 1A; Table 3). These studies display common changes in the form of reductions of tissue levels of sulphate, thiols, and other S-containing metabolites, while other metabolites such as OAS accumulate (Nikiforova et al., 2003, 2004, 2005a, 2005b; Hirai ; Maruyama-Nakashita ). They also show unpredicted effects on metabolite composition, for example reduction in chlorophyll, protein, and RNA contents, and accumulation of N-rich compounds, such as allantoin, asparagine, glutamine, and putrescine. Flavonoids accumulate as an effect of MYB75 expression, as indicated by the reddish colour of leaf tissues. With respect to secondary metabolites, more caution has to be exercised in drawing conclusions; for example, there is a reduction in glucosinolate content upon sulphate deprivation in Brassicaceae species such as Arabidopsis and rapeseed, but not in other plants where these compounds are not present (Gigolashvili et al., 2007a, 2007b, 2008; Hirai ).

Table 3.

Metabolome analyses related to sulphur metabolism

Experiment	Species	Tissue	Type*	References
–S	Arabidopsis	Seedling	GCMS, LCMS, HPLC	Nikiforova et al. (2003)
		Leaf, Root	FTMS, CE, HPLC	Hirai et al. (2004)
		Leaf, Root	FT-MS, CE, HPLC	Hirai et al. (2005)
		Seedling	GCMS, HPLC, IC, ICP-OES	Bielecka et al. (2015)
		Root	LC-MS/MS, HPLC, IC, ICP-OES	Forieri et al. (2017)
		Seedling	GC-TOF-MS, LCMS	Zhang et al. (2011)
	Brassica rapa	Leaf	GC-TOF-MS, HPLC	Sung et al. (2018)
	Barley	Leaf, Root	LC-QTOF-MS	Ghosson et al. (2018)
S	Wheat	Leaf, Grain	GCMS	Zorb et al. (2012)
32S, 34S	Arabidopsis	Seedling	DI-ICR-FT-MS	Gläser et al. (2014)
		Leaf, Root	UPLC-FT-MS	Giavalisco et al. (2011)
	Onion	Bulb	FT-ICR-MS	Nakabayashi et al. (2013); Nakabayashi and Saito (2017)
	Garlic	Bulb	FT-ICR-MS	Nakabayashi et al. (2016)

* CE, capillary electrophoresis; DI-ICR, direct-infusion ion-cyclotron-resonance; FT, Fourier transform; FTICR, Fourier transform ion cyclotron resonance; GC, gas chromatography; HPLC, high-performance liquid chromatography; IC, ion chromatography; ICR, ion-cyclotron resonance; ICP-OES, inductivity coupled plasma optical emission spectrometer; LC, liquid chromatography; MS, mass spectrometry; QTOF, quadrupole time of flight; TOF, time of flight; UPLC, ultra-performance liquid chromatography.

Metabolome analyses related to n class="Chemical">sulphur metabolism

It can be argued that metabolomics only hints at affected pathways and does not provide detailed information with respect to proteins or genes that are actually involved in modulating the metabolic composition. The metabolite composition can be assumed to reflect the integration of all the transcription, translation, enzyme activity, and flux dynamics that are relevant to all previous regulatory responses and biochemical steps. As such, metabolomics data contain information on how a plant adapts to a given stress such as S deprivation. Shifts in metabolic patterns might provide information on the pathways affected, but very little on the individual genes or enzymes involved in generating the overall output. The accumulation of intermediates that are otherwise not present in the metabolome is helpful in identifying bottlenecks in biosynthetic pathways and the genes, enzymes, or rescue mechanisms that are involved. OAS accumulates upon S deprivation (Hirai ; Nikiforova ) and it is an intermediate of cysteine biosynthesis that is normally present in low concentrations in plant tissues, thus hinting at altered activities of its associated enzymes SERAT and O-acetylserine(thiol)lyase (OASTL). These studies eventually led to the functional determination of the cysteine synthase complex (CSC; Saito, 2000; Wirtz and Hell, 2006; Koprivova and Kopriva, 2014; Watanabe ). A combined transcriptomics and metabolomics analysis of conditions where OAS accumulates without S being depleted (Lehmann ; Espinoza ; Caldana ) has suggested that OAS is a putative signal molecule that can induce a set of genes, termed OAS-cluster genes (Hirai ; Hubberten ; Aarabi ). OAS has been identified as a precursor of cysteine biosynthesis by the CSC, and as a regulator of this complex for its own synthesis at enzyme-activity level (Wirtz and Hell, 2006). Specifically, in Arabidopsis OAS may serve as a signal to down-regulate glucosinolate biosynthesis in favour of shifting remaining S resources towards primary metabolism through activating SDI1 and SDI2 (Saito, 2000; Aarabi ). As SDIs have been found to be present in all plant species examined (Howarth ; Aarabi ), they may have assumed additional biological functions, such as to mediate responses to stresses involving ROS (Lehmann ) and in developmental processes (Espinoza ; Caldana ). However, this remains to be validated.

The ultimate goal of a plant is to retain homeostasis in order to support growth and eventual propagation, and hence the relative stability of the metabolome even under S stress conditions. We postulate that under insufficient S availability, all regulatory efforts are aimed at keeping a balanced metabolite composition or at the activation of compensatory measures to substitute for depleted metabolites, such as depletion of ROS-protective GSH under S-deprived growth conditions (Whitcomb ). Consequently, at the level of the metabolome, changes in metabolite contents only occur when resources fall below certain thresholds for sustaining biosynthesis. Flux analyses can provide information on how metabolite levels are kept constant through altered fluxes (Hawkesford and De Kok, 2006; Rennenberg and Herschbach, 2014). We suggest integrating metabolomics with other targeted or non-targeted data such as transcriptomics or flux analysis and then examining the combined data for correlations. The use of heavy isotopes (34S) has recently provided novel information (Table 3). While previous metabolomics studies have concentrated on primary S-containing metabolites such as cysteine, methionine, glutathione, and S-adenosylmehtionine and secondary metabolites as glucosinolates and sulpholipids, the use of 34S-sulphate has allowed the identification of more than 140 previously unknown S-containing compounds that differentially react to sulphate deprivation or to different gene mutations (Gläser ). The authors concluded by calling for a re-evaluation of sulphur metabolism. However, several of these novel compounds are not yet annotated. Whether low-concentration S compounds can indeed exert effects on plant metabolism needs to be evaluated. A particular case is the identification of health-promoting compounds in garlic and onion (Nakabayashi et al., 2013, 2016; Nakabayashi and Saito, 2017) (Table 3) where heavy-isotope labelling has helped to identify the relevant compounds and pathways. Secondary metabolite analysis mostly provides information within a plant family; however, the methods presented in these studies are applicable to other species and are thus helpful for sulphur systems biology.

Proteomics

For Arabidopsis, only a few studies are available that have been dedicated to the determination of proteome changes in leaves or roots upon S deprivation (Table 4). A combined transcriptomics and proteomics study on Arabidopsis seeds that investigated the response to S deprivation (Higashi ) corroborated the shift of storage proteins in favour of those containing less S amino acids by blocking C-terminal degradation of low-S 12S globulins and reducing the amount of S-rich 2S albumins. Similar findings were also found in seed proteomics studies of wheat (Grove ) and rapeseed (D’Hooghe ) under S deprivation, allowing the assumption that this constitutes a general response. However, the molecular mechanisms and regulatory control of this response still remain to be elucidated. In addition, these studies revealed that S-responsive genes such as sulphate transporters and APS reductase were induced in seeds under S limitation together with several genes related to ROS protection, indicating the presence of ROS stress in S-deprived seeds. The genes identified overlapped with those found in earlier studies on Arabidopsis exposed to S stress (Hirai ; Nikiforova ). It was hypothesized that ROS accumulation might be the result of reduced thiol availability in the S-deprived seeds, and that seeds possess mechanisms to counteract these effects to maintain viability. Extended S starvation in rapeseed has been shown to impair viability and germination ability (D’Hooghe et al., 2013, 2014). These proteomic studies of rapeseed exposed to S limitation combined with metabolomics studies on leaves (D’Hooghe ) and seeds (D’Hooghe et al., 2013, 2014) corroborate the findings described for Arabidopsis, with additional changes in the seed lipid composition in favour of long fatty-acid chains and impairment of photosynthesis in the leaves. In addition, these studies have also indicated a link with ethylene and jasmonate metabolism, as has been described for Arabidopsis (Nikiforova ; Wawrzynska ; Wawrzyńska and Sirko, 2016). As such, proteomics studies in other Brassicaceae species might compensate for the lack of studies on Arabidopsis. A recent study focussed on the gasotransmitter H2S and L-cysteine desulphhydrase 1 (DES1) used a specialized proteomics approach to identify persulphidated proteins in Arabidopsis (Aroca et al., 2017, 2018). Persulphidation is believed to affect a variety of biological functions, including stress responses and carbon metabolism, and displays a new type of regulation that is assumed to counteract the nitrosylation-mediated response of the gasotransmitter NO that presumably acts on the same proteins. We anticipate a detailed systems biology-based study on the effects of persulfidation being carried out in the future. A proteomics study on the Tibetan alpine plant Lamiophlomis rotata revealed that H2S as an important player in adaptation to high-altitude stresses, and linked S metabolism to this adaptation process and to oxidative stress (Ma ). Proteomics studies in response to heavy metal stresses have not provided much novel information with respect to S proteomics per se. Specific heavy metal stresses that have been examined include exposition to cadmium (reviewed in (Villiers ; Bagheri ) of various plants and tissues such as Arabidopsis leaves (Semane ), poplar leaves and roots (Kieffer ), and roots of Brassica juncea (Alvarez ). In addition, chromium (Yildiz and Terzi, 2016) and arsenite expositions (Dixit ) have been investigated. However, these studies are still relevant as they relate to S metabolism, because GSH and phytochelatin synthesis are essential to detoxify metal ions and/or alleviate the consequent effects of ROS.

Table 4.

Proteome analyses related to sulphur metabolism

Experiment	Species	Tissue	Type	References
–S	Arabidopsis	Seed	2DE, MALDI-TOF, LC-MS/MS	Higashi et al. (2006)
	Oilseed rape	Leaf	2DE, ESI LC-MS/MS	D’Hooghe et al. (2013)
	Brassica napus	Seed	2DE, ESI LC-MS/MS	D’Hooghe et al. (2014)
		Leaf	2DE, MALDI-TOF/TOF	Yildiz and Terzi (2016)
-S, +Cd	Spinach	Leaf	SDS-PAGE	Bagheri et al. (2017)
+Cd	Arabidopsis	Leaf	2DE, ESI LC-MS/MS	Semane et al. (2010)
	Poplar	Leaf, Root	2DE, MALDI-TOF/TOF	Kieffer et al. (2009)
	Brassica juncea	Root	2DE, nano-LC-MS/MS	Alvarez et al. (2009)
S	Wheat	Grain	2DE, MALDI-TOF/TOF	Grove et al. (2009)
S, +As (III)	Rice	Leaf	2DE, MALDI-TOF/TOF	Dixit et al. (2015)
H2S	Arabidopsis	Leaf	LC-MS/MS	Aroca et al. (2017)
	Lamiophlomis rotata	Leaf	2DE, MALDI-TOF/TOF	Ma et al. (2015)

2DE, two-dimensional gel electrophoresis; ESI LC-MS/MS, electrospray ionization liquid chromatography with tandem mass spectrometry; MALDI-TOF, matrix-assisted laser desorption ionization-time of flight; nano-LC-MS/MS, nanoscale liquid chromatography coupled to tandem mass spectrometry; SDS-PAGE, polyacrylamide gel electrophoresis.

Proteome analyses related to n class="Chemical">sulphur metabolism

The analysis of persulphidation in plants of the des1 mutant provides an example of a repeated analytical cycle employing ‘omics’ analysis as illustrated in Fig. 1A. Another example of the approach illustrated in the figure is that in the initial phase, ‘omics’ studies were conducted on wild-type plants exposed to S deprivation (Table 1) and then in an iterating analytical cycle, mutants were analysed as a result of targeted research efforts (Aroca ) or from candidate genes (e.g. n class="Chemical">SDI1 and n class="Chemical">SDI2; Fig. 1B) identified from previous ‘omics’ studies (Howarth ; Aarabi ).

Bioinformatics of Arabidopsis with respect to sulphur metabolism

Sulphur systems biology has developed over the last two decades to become a widely used analytical approach in research on S physiology. Right from the beginning, the sheer amount of data produced by ‘omics’ studies made it difficult to identify relevant information. Hence, early in the development sulphur systems biology, bioinformatics approaches were employed to determine correlations among responses and to deduce correlation networks (Nikiforova et al., 2003, 2005a; Hirai and Saito, 2004; Hirai ; Ohta ; Usadel ; Obayashi and Kinoshita, 2010). Approaches such as gene-to-gene correlations, gene-to-metabolite correlations, clustering, principal component analyses, and batch-learning self-organizing maps (BL-SOM) helped to organize data into relevant related units and aided the generation of hypothesis, and hence the identification of candidate genes or processes (Fig. 1A). Examples of the successful use of BL-SOM are the identification of the involvement of S metabolism in the biosynthesis of medicinally active compounds in plants (Rai ), the identification of a correlation between S metabolism and the MYB transcription factor PAP1 that controls anthocyanin biosynthesis (Hirai ), and the functional elucidation of SDI1 and SDI2 (Aarabi ) (Fig. 1B). Obviously, an important aspect is the clear definition of the experimental question, and the experimental and applied biostatistical approaches that are taken: without these, data interpretation will remain incomplete and unsatisfactory. A statistical challenge is the gap between the number of samples, such as the conditions or plant genotypes, and the much greater number of determined values, such as genes, metabolites, proteins, or other data. To resolve this challenge, dimensionality reduction methods have been developed and are widely applied to ‘omics’ data (Steinfath ; Weckwerth, 2008).

Another factor to consider for successful S systems biology is standardization of experiments and data (Salek ). This mainly concerns to the experimental side: how plants are grown, how material is processed, and how values are measured. As demonstrated by recent studies (Kopriva ; Henríquez-Valencia ), the available S transcriptome data have been obtained from plants grown under different conditions and various tissues and developmental stages have been used. There is still an overlap between these studies that can indicate the most robust responders, usually those with the biggest increments in change of expression or metabolite content between samples (Kopriva ). Among them, an NADPH oxidoreductase (AT1G75280) has been speculated but not proven to act as an isoflavonoid reductase (IFR) that is active in anthocyanin/flavonoid biosynthesis to provide ROS-protection capacity under S-deprived, and hence GSH-deprived, conditions (Nikiforova ). This may well link to the correlation of PAP1 with sulphate metabolism under S-deprived conditions discussed earlier (Hirai ). Another example is the identification of the OAS-cluster genes (Hubberten ), which correlate various conditions where OAS is accumulated in tissues with a set of co-expressed genes. Here, the link between these diverse conditions and S metabolism is not yet established. Thus, bioinformatics in support of sulphur systems biology has proven successful as early findings have laid the basis for later detailed research (e.g. OAS-cluster genes) or have provided candidate genes (e.g. IFR) whose function in plant S metabolism can be further examined.

An important problem is that subtle differences in (e.g.) gene expression might be difficult to identify. For example, the transcriptional changes of TFs are usually low, with thresholds often involving changes of only 1.5- or 2-fold, which may scarcely be above background variation (Nikiforova ; Maruyama-Nakashita ; Bielecka ; Forieri ). Differences in data generated by different laboratories also exist, even when methods are standardized. Massonnet organized a number of independent laboratories to conduct the same experiment using similar genotypes and standardized growth conditions with leaf phenotypes, transcriptomics, and metabolomics as the output, or with material produced in one lab being distributed to the others for analysis. But despite this standardization differences in the data produced were still observed. The variations were suggested to be attributable to variability in the plants and sample handling (i.e. human factors) as well as to slight differences in growth conditions (light quality, temperature, and water). Moreover, comparability of S deprivation is further complicated as this is a dynamic process and is dependent on the specific level of deprivation that is applied (Whitcomb ; Henríquez-Valencia ). It would be helpful for the advancement of our knowledge of S metabolism if an agreement could be reached on standard conditions, plant lines, and procedures, and on a more systematic and complete catalogue of the systems response to distinct S deprivation and resupply conditions. This would be helpful as a blueprint on which to base future experiments using other cultivars, conditions, or mutants, the results of which could then be compared back to this master data set. Such a blueprint has been provided in the case of senescence (Watanabe ). The data that are currently available do still allow meta-analysis of the S transcriptome and can yield suggestions for novel TFs that potentially play roles in S metabolism. For example, Henriquez-Valencia deduced a network of known and putatively correlated regulators and TFs, which in particular suggested that NF-YA2 (AT3G05690, nuclear transcription factor Y subunit A-2) and RVE2 (AT5G37260, reveille2) act as upstream regulators of the S deprivation response (Table 2). NF-YA2 and RVE2 display connections to five S-related response modules. NF-YA2 is induced by S depletion and has been shown to be associated with the regulation of several developmental processes, such as flowering and leaf and root system architecture, and to respond to depletions of N and P. RVE2 is reduced by S depletion and has been shown to be associated with germination and control of the circadian rhythm. Network analysis has suggested that IAA28 (AT5G25890) acts as a regulator in the S-starvation response network, although its expression is not markedly changed upon S depletion of seedlings (Nikiforova et al., 2003, 2005a, 2005b; Falkenberg ; Hoefgen and Watanabe, 2017).

Conclusions and outlook

Sulphur systems biology has provided novel information, especially with respect to fundamental research findings. It is an inherent feature of transcriptomics approaches that they cannot identify when regulation occurs post-transcriptionally or even post-translationally, for example through changes of enzyme activities or regulatory properties. An example is the first TF identified to control S metabolism, SLIM1 (Maruyama-Nakashita ), which is not (or hardly) transcriptionally altered upon S deprivation. SLIM1 instead seems to be modulated by a protein–protein interaction with EIN3, a positive regulator in the ethylene response pathway (Wawrzyńska and Sirko, 2016). Thus, systems biology studies provide a certain subset of information, usually based on differential accumulation of molecules. Hence, integrating targeted analyses or agronomic data is necessary to gain a holistic understanding of the system (Fig. 1A). Sulphur systems approaches have already served to build models that have allowed novel candidate genes to be identified and confirmed and, in the subsequent iterative process of applying targeted and non-targeted analyses of omics-derived candidates, have allowed further details to be uncovered (Fig. 1A, B).

As S moieties are a key determinant of a vast number of biomolecules and biochemical processes, it is not surprising that systems approaches have highlighted processes that seemingly appear to be unrelated to bona fide S assimilation and the biosynthesis of primary S-containing metabolites. Sulphur-omics is a rich source of candidate genes, many of them still awaiting detailed examination. The huge flood of data that has been produced through sulphur systems biology approaches has led to results that have helped to elucidate plant S physiology. Systems biology will continue to support the identification of novel genes and the validation of candidate genes as a standard tool of molecular biology.

What might be required to increase our knowledge beyond what has currently been achieved? A more systematic analysis of plants, especially of the model Arabidopsis, exposed to defined conditions of S availability would be helpful to provide a blueprint and to correlate future research findings to a master data set. The current systems biology data available for Arabidopsis are derived from only a limited number of conditions, genotypes, and developmental time-points (Tables 1, 3, 4). Even when considering other plant species, the database is only marginally greater. In the case of proteomics, the data set available is insufficient. There is not only a need for cataloguing changes in protein contents, but also for changes in protein activities and protein modifications, such as phosphorylation, persulfidation, and glutathionylation. A recent study on protein persulphidation provides an example of what is urgently needed (Aroca ). Likewise, flux analyses at a systems level are generally missing and would help in the construction of consistent functional networks. This does not necessarily simply mean that more data is better data, but that the S research community should agree on standards, on approaches, and on data storage and exchange. We would even propose a joint research effort between different laboratories despite the redundancy of data it might include. A broader and reliable database would be a good resource for future bioinformatics attempts to deduce relevant conclusions. A debate needs to be had as to whether fragmentation of data should be avoided by concentrating research on model systems such as Arabidopsis or whether analysis of multiple plant species with diverse biology would eventually provide a better understanding. With ever-improving sequence technologies there are hardly any restrictions with regards to the availability of genome sequence data, and resources that provide data on genetic variability are available for many crop species, such as rapeseed, rice, and wheat, besides the Arabidopsis model system (Scholl ).

In the case of bioinformatics, there is a need to improve data interpretation and model building in order to close the gap between bioinformatics and the biological interpretation of data. The identification of relevant candidates, pathways, and processes is still knowledge-driven rather than being provided in an objective manner by bioinformatics prediction tools. Studies such as a recent analysis of network topologies and their relation to stability against mutational variation (Mähler ) might be helpful for n class="Chemical">sulphur systems biology. Further, bioinformatics studies and systems biology studies should aim at understanding dynamic processes over time rather than just the current snapshot view of S metabolism.

Sulphur is not an isolated entity within the biochemistry of a plant. It is instead interactive, cross-influencing and being influenced by numerous other processes, the foremost of which are the links to ROS tolerance and detoxification, and the interplay with other mineral nutrient ions (Kopriva and Rennenberg, 2004; Kruse ; Forieri ; Zuchi ). Aspects such as photosynthesis (Nikiforova ; Wulff-Zottele ; Naumann ) and seed protein quality (Galili and Höfgen, 2002; Galili ) are also worth considering in greater detail within this context.

With regards to the application of sulphur systems biology for agronomy, there are no reports yet in relation to new varieties released to the market. However, systems biology has become a standard element in the analytical toolbox for crop research (Langridge and Fleury, 2011; Reynolds and Langridge, 2016; Heyneke ; Casartelli ). Application of knowledge originating from plant S research needs to take into account agricultural procedures such as fertilization regimes, cropping systems, soil parameters, and water availability. The specific needs of certain crops for S supply also have to be considered, for example the high demand of rapeseed (Bloem ). Sulphate supply is also considered necessary for legume nodule functioning (Krusell ) and root–mycorrhiza interactions (Sieh ). Not least, S metabolism is dependent on nutrient interactions under natural conditions (Zuchi ; Forieri ).

We have reviewed sulphur systems biology and looked separately at transcriptomics, metabolomics, proteomics, and bioinformatics. However, systems biology inherently integrates all existing information from genotype to phenotype. We are convinced that integration of the results of all ‘omics’ technologies and of classical biochemical, physiological, and agronomical experiments will eventually lead to breakthrough results (Fig. 1A). Systems biology is not about ‘omics’ technologies but about a holistic view of the highly complex biological system of the plant and trying to capture its function through understanding all (or at least as many) parts as possible using integrative approaches, i.e. high-throughput ‘omics’ determinations supported by targeted approaches. The ultimate goal of Sulphur-omics is to understand the underlying network scheme, to model it mathematically, and to use this for predictions—at the cellular, organ, and whole-plant level. Sulphur systems biology has the chance to provide a showcase for nutrient systems biology.