Gel-based proteomic map of Arabidopsis thaliana root plastids and mitochondria

Background Non-photosynthetic plastids of plants are known to be involved in a range of metabolic and biosynthetic reactions, even if they have been difficult to study due to their small size and lack of color. The morphology of root plastids is heterogeneous and also the plastid size, density and subcellular distribution varies depending on the cell type and developmental stage, and therefore the functional features have remained obscure. Although the root plastid proteome is likely to reveal specific functional features, Arabidopsis thaliana root plastid proteome has not been studied to date. Results In the present study, we separated Arabidopsis root protein fraction enriched with plastids and mitochondria by 2D-PAGE and identified 84 plastid-targeted and 77 mitochondrion-targeted proteins using LC-MS/MS. The most prevalent root plastid protein categories represented amino acid biosynthesis, carbohydrate metabolism and lipid biosynthesis pathways, while the enzymes involved in starch and sucrose metabolism were not detected. Mitochondrion-targeted proteins were classified mainly into the energetics category. Conclusions This is the first study presenting gel-based map of Arabidopsis thaliana root plastid and mitochondrial proteome. Our findings suggest that Arabidopsis root plastids have broad biosynthetic capacity, and that they do not play a major role in a long-term storage of carbohydrates. The proteomic map provides a tool for further studies to compare changes in the proteome, e.g. in response to environmental cues, and emphasizes the role of root plastids in nitrogen and sulfur metabolism as well as in amino acid and fatty acid biosynthesis. The results enable taking a first step towards an integrated view of root plastid/mitochondrial proteome and metabolic functions in Arabidopsis thaliana roots.


Background
Plastids are plant cell organelles possessing diverse roles in energy metabolism, biosynthetic reactions and other metabolic activities. All plastid types are enclosed by a double envelope membrane and they contain several copies of a semiautonomous circular genome encoding circa 100 proteins, which are involved in photosynthesis, transcription and translation. However, a vast majority of plastid-localized proteins are encoded by a nuclear genome. These proteins are translated on cytosolic ribosomes and transported into the plastid using the information buried in the N-terminal transit peptide [1,2]. The transit peptide selectively interacts with different translocon components hosted by a specific plastid, thereby discriminating between the plastids of different types and/or ages [1,3]. The cell type and surrounding tissue together with environmental factors determine the direction of plastid development [4].
Plastids are divided in distinct groups depending on structure, (pigment) composition and functional properties. The best characterized plastid type is the chloroplast, which are abundant in green tissues. Chloroplasts are the sites of photosynthesis and contain an elaborate internal membrane system called the thylakoids, in which the large pigment-protein complexes involved in photosynthetic electron transfer reactions are embedded. Flowers and fruits contain chromoplasts, which are rich in yellow and orange carotenoids, while roots (and other non-photosynthetic tissues) contain leucoplasts, which can be further classified as elaioplasts storing lipids, proteinoplasts containing crystalline protein bodies, and amyloplasts storing starch [5]. Chloroplasts are able to capture light energy and convert it into chemical form as ATP and reducing equivalents of NADPH, while the non-photosynthetic plastids are dependent on the import of external sugar phosphates and ATP to the plastid.
The non-photosynthetic plastids are involved in a range of metabolic and biosynthetic reactions, such as biosynthesis of starch [6], carotenoids [7] and lipids [8] as well as assimilation of nitrogen [9]. Although the proteome of a certain plastid type is likely to reflect and reveal specific functional features, there are limited data available on the proteomes of non-photosynthetic plastids. Thus far, the proteomes of potato tuber amyloplasts [10], wheat endosperm amyloplasts [11,12], chromoplasts from bell pepper and sweet orange fruits [13,14], non-differentiated plastids from tobacco cell culture [15], rice etioplasts [16], and Medicago truncatula root plastids [17,18] have been studied, while there is no information available on the root plastid proteome of Arabidopsis thaliana.
In the present study, we have taken the same strategy used in most studies focusing on non-photosynthetic plastid proteomes: enrichment of root plastids, separation of the proteins using 2D-gel electrophoresis followed by in-gel digestion and identification of the protein spots using LC-MS/MS [e.g. 10,11,17]. The identified proteins reveal a complex regulatory and metabolic network present in Arabidopsis thaliana root plastids and mitochodria, and enable a first step towards an integrated view of root plastid/mitochondrial proteome and metabolic functions of roots.

Results
Due to the small size (diameter of root plastids ranging from 1.8 to 3.0 μm as compared to 5-10 μm in chloroplasts) and lack of color, root plastids have been difficult to study, and therefore they are still poorly characterized [19]. Nevertheless, it has been shown that the morphology of root plastids is very heterogeneous and the plastid shape (often described as amoeboid), size, density and subcellular distribution varies depending on the cell type and the stage of development [19]. Additionally, root plastids are abundant in stromules, the stroma filled tubules which can interconnect different plastids [20]. As the size of mitochondria (0.1-0.5 μm × 1-2 μm) falls in the same range as the root plastids, it is not surprising that the sample representing enriched root plastids also contained mitochondria, with a minor contamination by cytosolic proteins (Fig. 1; Additional file 1: Figure S1). These contaminants are frequently detected also in other studies using the same approach [10,11,17]. The preparation was further used to study the leucoplastidic and mitochondrial proteome of Arabidopsis roots.
Separation of organellar root proteins by isoelectric focusing using the pH range from 3 to 11 followed by SDS-PAGE and Sypro Ruby staining revealed 139 distinct protein spots that could be identified in at least three independent biological replicates (Additional file 2: Figure S2). We selected the spots for LC-MS/MS analysis based on the following criteria: (i) the spots were clearly visible in three biological replicates, (ii) the spots were separated from neighboring spots, and (iii) the spots were not too big and intensive, because such spots usually contain a high number of different proteins. Low molecular weight protein spots were excluded, as they usually do not provide enough peptides for protein identification. Moreover, some spots showing the same molecular weight but different pI were selected (e.g. 1-4, 6 and 7-9 in Additional file 2: Figure S2 A), because we wanted to see if these spots represented different isoforms of the same protein. The proteins identified with high confidence and represented by at least 3 peptides, expected molecular weight and isoelectric point (sequence based predictions) are presented in Additional file 3: Table S1. The plastid-targeted proteins were classified into nine, and the mitochondriontargeted proteins into seven functional categories (Fig. 2) based on localizations and functions annotated in Fig. 1 Analysis of the root protein sample. 20 μg of crude root protein extract (C), the pellet representing root plastids (P) and supernatant representing cytosol (S) were separated by SDS-PAGE, transferred to a PVDF membrane and immunolabelled with roottype FNR (RFNR; root plastid marker), alternative oxidase (AOX1/2; mitochondrial marker) and nitrate reductase (NR; cytosolic marker) antibodies UniProt and/or according to recent literature (e.g. [12] and other references presented in text). In plastids, the most prevalent categories represented amino acid biosynthesis (31%) and carbohydrate metabolism (20%), while in mitochondria the proteins involved in energetics (41%) formed the largest category (Fig. 2). It should be noted, however, that the proteins not included in these categories may still have an (unknown) role in these pathways, or have other important roles in root metabolism.
The L-arginine biosynthesis pathway was well represented, as five enzymes of the pathway were identified from our root plastid sample (Table 1, Fig. 3). Acetylglutamate kinase (NAGK, At3g57560), catalyzing phosphorylation of N-acetylglutamate was identified in spot 92. N-acetylglutamate phosphate is then converted to Nacetylglutamate semialdehyde by N-acetylglutamyl- phosphate reductase (AGPR, At2g19940, spots 70 and 71). In the next step, acetylornithine aminotransferase (WIN1, At1g80600), detected in spots 62 and 63, transfers an amino group from a second glutamate to form N 2 -acetylornithine. Subsequently, the produced ornithine is further used by ornithine transcarbamomylase (OTC, At1g75330, spot 85) to form citrulline. Argininosuccinate synthase (AS, At4g24830) detected in spot 51 and 52 converts L-citrulline to argininosuccinate that is the direct substrate for synthesis of L-arginine [25].
The branched-chain amino acids (BCAAs), namely Lisoleucine, L-valine and L-leucine, belong to the nine of 20 amino acids that cannot be synthetized de novo by humans nor animals. The BCAAs synthesis is known to take a place inside of plastids [26], and indeed all four enzymes catalyzing subsequent steps in conversion of Lthreonine or pyruvate to different BCAAs were identified in our sample enriched with root plastids (Table 1, Fig. 3). Acetolactate synthase (also known as acetohydroxyacid synthase, AHAS, spot 58) is the first enzyme in the BCAA synthesis pathway that catalyzes the condensation of two molecules of pyruvate to 2-acetolactate or condensation of pyruvate and α-ketobutyrate to 2aceto-2-hydroxybutyrate. In plants, AHAS works as a heterodimer formed by a catalytic and a regulatory subunit [27,28]. In Arabidopsis, two genes for the regulatory subunits have been identified, and one of them (At5g16290) was detected in spot 58. Ketolacid reductoisomerase (KARI, At3g58610, present in spots 36 and 37) isomerizes and then reduces two acetohydroxyacids to Fig. 3 Simplified scheme of the amino acid biosynthesis pathways and glycolysis in plants. Black arrows represents the reactions catalysed by the enzymes identified in plastid enriched fraction, grey arrows indicate reactions catalysed by the enzymes which were not detected in the current work. AHAS, acetolactate synthase small subunit; 1AGPR, probable N-acetyl-gamma-glutamyl-phosphate reductase; AS, argininosuccinate synthase; ATP-PRT, ATP phosphoribosyltransferase; 1BACAT, branched-chain-amino-acid aminotransferase; 3CHS, chorismate synthase; CS, cysteine synthase; DapB, 4-hydroxy-tetrahydrodipicolinate reductase; 1DapF, diaminopimelate epimerase; DHAD, dihydroxy-acid dehydratase; DHQ-SDH, bifunctional 3-dehydroquinate dehydratase/shikimate dehydrogenase; EPSP-S, 3-phosphoshikimate 1-carboxyvinyltransferase; FBA, fructose-bisphosphatase aldolase; GAPCp1, plastid localized glyceraldehyde-3-phosphate dehydrogenase; GPI, glucose-6-phosphate isomerase 1; HK, homoserine kinase; IGPS, imidazole glycerol phosphate synthase; KARI, ketol-acid reductoisomerase; LysA, diaminopimelate decarboxylase; NAGK, acetylglutamate kinase; OAT, ornithine aminotransferase; OTC, ornithine carbamoyltransferase; PAT-AT, bifunctional aspartate aminotransferase and glutamate/ aspartate-prephenate aminotransferase; PGDH1, D-3-phosphoglycerate dehydrogenase; PSAT, phosphoserine aminotransferase; TCA, tricarboxylic acid cycle; TPI, triose phosphate isomerase; TS, threonine synthase; TSβ, tryptophan synthase beta chain; WIN1, acetylornithine aminotransferase  produce dihydroxyacids. This reaction is the second step in the parallel pathway towards L-valine and Lisoleucine biosynthesis [22,29]. Dihydroxyacid dehydratase (DHAD, At3g23940), performing the third step in synthesis of L-isoleucine from 2-oxobutanoate or synthesis of L-valine from two molecules of pyruvate [30], was identified in spots 30, 43 and 45. Finally, one of the three plastid localized isoforms of branched-chainamino-acid aminotransferase (BCAT At3g49680) [31] was detected in spots 64 and 66. BCATs catalyze the last step of all the three BCAA biosynthesis pathways, converting the respective α-keto acids to amino acids [32]. The Shikimate pathway is a multistep metabolic route that converts the primary metabolites phosphoenolpyruvate (PEP) and erythrose-4-P to chorismate. Chorismate is further used as a precursor for synthesis of the three aromatic amino acids (L-phenylalanine, L-tyrosine, and L-tryptophan) as well as other aromatic compounds (like pigments and hormones) required for plant growth, development and defense [33]. In our studies, we have identified three enzymes involved in the shikimate pathway, the bifunctional 3dehydroquinate dehydratase/shikimate dehydrogenase (DHQ/SDH; At3g06350, in spot 30), 5enolpyruvylshikimate-3-phosphate (EPSP) synthase (also referred as 3-phosphoshikimate 1-carboxyvinyltransferase; At2g45300, spots 51, 52) and chorismate synthase (CHS, At1g48850, spots 71, 73) ( Table 1, Fig. 3). The first enzyme catalyzes two subsequent reactions in the pathway leading to conformation of 3-dehydroquinate into shikimate [34]. EPSP synthase transfers an enolpyruvoyl moiety from PEP to shikimate 3-phosphate via a chemically unusual reaction involving C-O bond cleavage of PEP rather than P-O bond cleavage as in most PEP-using enzymes [35,36]. The product of this reaction, EPSP, is further transformed to chorismate via 1,4-trans elimination of the phosphate and hydrogen by chorismate synthase. In plants, several different enzymes compete for chorismate in post-chorismate metabolic pathways. In one of these metabolic branches, chorismate is converted into L-tryptophan. The last two steps of this pathway are catalyzed by tryptophan synthase α subunit and β subunits, respectively. Tryptophan synthase subunit β (TSβ, At5g54810) was detected in spot 65, corroborating the earlier results suggesting that in addition to green parts of the plant, L-tryptophan is also synthesized in Arabidopis roots [37,38]. Also, the post-chorismate metabolic branch that leads to synthesis of L-phenylalanine and Ltyrosine was represented in root plastid samples, as the bifunctional aspartate aminotransferase and glutamate/ aspartate-prephenate aminotransferase (PAT-AT; At2g22250) [39] was identified in spots 64 and 66 ( Table  1, Fig. 3).
In non-photosynthetic plastids (as well as in chloroplasts during the dark period), the oxidative pentose phosphate pathway generates reducing power in the form of NADPH to be used in numerous metabolic pathways, including nitrogen and sulfur assimilation. NADPH is oxidized by the root-type ferredoxin-NADP + oxidoreductase enzymes (RFNR1, At4g05390 in spot 77 and RFNR2, At1g30510 in spot 78), which subsequently reduce root-type ferredoxin proteins [45]. Ferredoxinnitrite reductase (NiR; At2g15620), identified in spots 45, 51 and 52, oxidizes ferredoxin for the reduction of nitrite to ammonia, which then enters the cell metabolism via the amide and amine nitrogen of amino acids (e.g. via glutamine synthetase At5g35630 in spot 55). ATP sulfurylase (ATP-S; At3g22890, spots 56 and 57), in turn, is the first enzyme of the sulfate assimilation pathway (Table 1). It catalyzes the activation of inorganic sulfate by transferring sulfate to the adenine monophosphate moiety of ATP to form the high-energy compound adenosine 5′-phosphosulfate (APS) [46]. APS is then reduced to sulfite (by APS reductase) and sulfide (by sulfite reductase), which also requires reducing equivalents of ferredoxin. Thereafter, sulfide is incorporated into cysteine, which is a precursor for methionine as well as a major component of glutathione and phytochelatins, which are needed for the induction of abiotic stress responses [47].

Lipid metabolism
The biosynthesis of glycerolipids and fatty acids occurs in plastids (and endoplasmic reticulum). Generation of acetyl-CoA by the pyruvate dehydrogenase enzyme complex (see above) is followed by the formation of malonyl-CoA, catalyzed by acetyl-CoA carboxylase [48,49]. In Arabidopsis, plastid acetyl-CoA carboxylase is a multienzyme complex, composed of biotin carboxylase, biotin carboxyl carrier protein, and αand β-subunits of carboxyltransferase [50]. Indeed, three of these proteins (biotin carboxylase, At5g35360, spot 47; biotin carboxyl carrier protein, At5g16390, spot 86; carboxyl transferase subunit β, AtCg00500, spots 47 and 59) were identified in our root plastid sample (Table 1). Then, another multienzyme complex named fatty acid synthase performs cyclic condensation of two carbon units: b-ketoacyl-acyl carrier protein (ACP) synthase III (KASIII) starts fatty acid chain elongation by performing the condensation reaction of malonyl-acyl ACP and acetyl-ACP [51,52], while KASI and KASII are the condensing enzymes for the elongation of the carbon chain from C4 to C18 [53]. KASI (At5g46290) was detected in spots 59 and 85. After the condensing reaction, the 3-ketoacyl-ACP is reduced by 3ketoacyl-ACP reductase (also known as 3-oxoacyl-ACP reductase, At1g24360, detected in spot 97), dehydrated, and finally the enoyl-ACP reductase (At2g05990, detected in spots 82 and 83) completes the formation of saturated fatty acids [54]. The presence of these enzymes emphasizes the importance of root plastids for the fatty acid biosynthesis in Arabidopsis.

Root mitochondria: components of citric acid cycle and respiratory electron transfer
In line with the main role of mitochondria in producing ATP and reducing equivalents for biosynthetic reactions, 41% of the identified mitochondrial proteins were classified into the energetics category (Fig. 2, Table 2; Additional file 4: Table S2). The citric acid cycle oxidizes acetyl-CoA and produces NADH and FADH 2 , which are required for the mitochondrial electron transfer and synthesis of ATP. In our sample, multiple components of the citric acid cycle were detected. First, we identified various components of the pyruvate dehydrogenase complex producing acetyl-CoA, including dihydrolipoyl-lysine-residue acetyltransferase components 2 (At3g13930, spots 22-25, 28, 30, 79, 86, 115) and 3 (At1g54220, spots 30, 45, 86), as well as EI (pyruvate dehydrogenase E1 component subunit beta-1, At5g50850, spots [79][80][81]86). In the first committed step of citric acid cycle, citrate synthase (CS) functions in converting acetyl-CoA to citrate [55]. Indeed, the CS4 (At2g44350) protein was identified in spots 56-58 and 65. Citrate is further converted to isocitrate by the aconitase (ACO) enzyme(s) [56]. In Arabidopsis, four genes encode the distinct isoforms ACO1-4. It is intriguing that the ACO2 (At4g26970) and ACO3 (At2g05710) isoforms, detected in the present study, were distributed in multiple spots indicating differential processing and/or other types of post-translational modifications of the ACO isoforms (Fig. 4, Table S1, 2). The same ACO isoforms were identified in a previous proteomic study on Arabidopsis mitochondrial proteome, but there only ACO3 was present in multiple spots [57]. Isocitrate is subsequently decarboxylated to 2-oxoglutarate by isocitrate dehydrogenase, represented by isocitrate dehydrogenase family protein (At5g14590 in spot 65) and regulatory subunits 1 (At4g35260, spot 72) and 2 (At2g17130, spots 62 and 63). 2-oxoglutarate, in turn, is decarboxylated to succinyl-CoA (by alpha-ketoglutarate dehydrogenase), which is then converted to succinate by the action of succinyl-CoA synthetase (or ligase) [58]. Succinyl-CoA ligase subunits alpha-1 (At5g08300) and beta (At 2 g20420) were identified in spots 75-78 and 85, and in spots 62 and 63, respectively. Next, succinate is oxidized to fumarate by succinate dehydrogenase complex, which participates in both the citric acid cycle and the respiratory electron transfer [59]. At5g66760, detected in spots 17, 18 and 20 encodes the flavoprotein subunit of succinate The respiratory electron transfer chain is composed of complex I (NADH coenzyme Q reductase), complex II (succinate dehydrogenase), complex III (cytochrome bc1 complex) and complex IV (cytochrome c oxidase). Complex I is a multisubunit complex of at least 49 subunits [60], which accepts electrons in the form of NADH from the citric acid cycle. The 75 kDa protein At5g37510 (spots 1, 11, 13, 18, 21, 45, and 46) and gamma carbonic anhydrase (At1g47260 in spots 92 and 93) represented complex I (Table 2, S1). As indicated above, succinate dehydrogenase complex performs a dual function in citric acid cycle and electron transfer [61]. Complex III, in turn, was represented by prohibitin-3 (At5g40770, spots 93, 95-97, 99) and cytochrome c1 1 (At3g27240, spot 88), while we could not detect any complex IV subunits in our sample. Moreover, ATP-synthase subunits were present in multiple spots: ATP synthase subunit alpha (ATMG01190) in spots 30, 37, 43-50, 55, 56, 58, 63, 64, 79, 83, 87, 89 Mitochondria, like all other organelles, are in constant interaction with other cellular compartments. Interaction occurs via transport of various substrates, intermediates and end-products between the compartments as well as via various signaling pathways. Porins, or voltage-dependent anion-selective channels (At5g15090, At5g67500, At3g01280), detected in multiple spots (95-97, 99, 100), enable the exchange of ions and small molecules, e.g. NADH and ATP across the mitochondrial outer membrane, thus linking mitochondrial reactions strictly to cellular metabolism.

Amino acid biosynthesis in root plastids: building blocks and regulatory elements
The presence of enzymes involved in the biosynthesis of amino acids appears to be common feature among the non-green plastids [12,13,17]. In all higher plants Llysine is synthesized from L-aspartate via a DAPdependent pathway that is functional also in algae, bacteria and fungi containing cellulose in their cell wall [62]. Similarly, L-aspartate is used as a substrate also for the synthesis L-threonine. It has been shown already decades ago that intact, illuminated chloroplasts are able to synthesize all the aspartate derived amino acids from exogenous [ 14 C] aspartate, indicating that they contain all the enzymes necessary for L-lysine and L-threonine synthesis [63,64]. It is likely that the biosynthetic pathways are present also in non-green plastids, as in addition to Arabidopsis root plastids some of the involved enzymes have been found also in Medicago truncatula root plastids [17] and amyloplasts from developing wheat endosperm [12].
L-histidine is synthesized through a series of enzymatic reactions taking place in chloroplasts [65,66]. As the pathway is extremely energy-consuming (41 ATP/1 His) [67], compartmentalization to chloroplasts suggests that photosynthesis might provide energy to the process [68]. However, the enzymes involved in histidine biosynthesis have also been detected in non-green plastids [12,17]. Indeed, L-histidine has been implicated in regulation of biosynthesis of other amino acids, in chelation and transport of metal ions, and in plant reproduction and growth [69][70][71]. Moreover, L-histidine biosynthesis is related to a number of other metabolic pathways including synthesis of purines, pyrimidines and folates [68]. It is worth noting that also adenylosuccinate synthase (At3g57610) was detected in our analysis (Table 1). It catalyzes the first step in the de novo synthesis of adenosine monophosphate (AMP) [72], which as a purine nucleotide is a structural component of nucleic acids as well as a precursor for various vitamins, coenzymes and hormones [73][74][75][76][77]. Indeed, it has been shown that the plastid-localized de novo synthesis pathway is the major purine synthesis pathway in plants [73].
L-serine is an important amino acid that is an intermediate in several metabolic pathways, serves as a precursor in formation of other amino acids (like L-glycine, L-cysteine) and is involved in synthesis of phospholipids and purines [78,79]. In plants, L-serine biosynthesis proceeds via three pathways. D-3-phosphoglycerate dehydrogenase (PGDH1), detected in multiple spots, has recently been suggested to be particularly important in non-photosynthetic tissues, as compared to the photorespiration dependent glycolate pathway functioning in the leaves [80][81][82].
As enzymes involved in biosynthesis of BCAAs are targets for commercially available and potential new herbicides, the biosynthetic pathways have been an object of extensive studies [26,29]. Similarly, the loss of the shikimate pathway in animal lineages has made it an interesting research object as a source of valuable nutritional molecules and potential target of new antibiotics and herbicides. Therefore the fact that a number of enzymes involved in BCAA biosynthesis and shikimate pathway were identified in the root plastid enriched sample is particularly interesting. Recent studies have also shown that dihydroxyacid dehydratase (DHAD) is widely expressed in both vegetative and generative tissues and its activity is important for the accurate accumulation of all three BCAAs in roots [30]. Plants with reduced level of the DHAD enzyme were found to show short root phenotype as well as hypersensitivity to salt stress [30]. It should be noted that the branched-chain-amino-acid aminotransferase BCAT (and other enzymes involved in leucine biosynthesis) are also used for synthesis of glucosinolates through the methionine chain elongation pathway [83,84], underlining the importance of root plastids for secondary metabolism and defense reactions of plants. Also polyamines and nitric oxide play important roles in the regulation of developmental processes and stress responses [85,86]. Importantly, L-arginine is not only a precursor for both of these compounds [87], but it is also an essential amino acid for protein synthesis possessing a high nitrogen: carbon ratio. Therefore Larginine often serves as a nitrogen storage, especially in underground storage organs and roots [88][89][90]. These properties position enzymes involved in L-arginine biosynthesis in the focus of current research aiming at resolving root originated stress responses.

Impact of plastidial glycolytic pathway enzymes on primary metabolism, stress responses and developmental processes
Glycolysis is the major primary metabolic pathway in living organisms where hexoses are oxidized to provide energy, reducing power and precursors for biosynthetic reactions [91]. In plants, there are two glycolytic pathways operating in parallel in the cytosol and in the plastids, and they interact through selective transporters present in plastid envelope [92]. In addition to their role(s) in primary metabolism, several plastid enzyme isoforms appear to be involved in regulation of developmental processes and induction of stress responses. For instance, the plastidial isoform of glucose-6-phosphate isomerase has been postulated to be essential for starch synthesis in floral initiation [93] as well as for survival in low oxygen conditions [94], while the expression of the triose phosphate isomerase gene is regulated via developmental stimuli [95]. FBA, in turn, is known to be crucial both for sugar metabolism and for signaling [96]. In addition for the well-studied function of FBA 3 (At2g01140) in glycolysis, the enzyme has been suggested to be important in the induction of responses to biotic and abiotic stresses as well as in the growth and developmental processes [96]. Indeed, the mRNA level of FBA was shown to increase in response to salicylic, abscisic and gibberellic acid [96][97][98], drought, salt, as well as cold stress [97,99,100]. Interestingly, FBA 3 was localized in plastoglobules of chloroplasts and non-green plastids of Arabidopsis plants [96]. Recently, FBA has been reported as a target for lysine methylation [101] and cysteine glutathionylation [102], which may affect the conformation, activity or stability of the enzyme.

Amyloplasts versus Arabidopsis root plastids
Several of the identified root plastid proteins, such as phosphoserine aminotransferase, arginosuccinate synthase, threonine synthase, malate dehydrogenase, fructose-bisphosphate aldolase and PPiase were also identified in the amyloplast fraction of potato tuber [10] and/or wheat ears or seeds [11,12]. Nevertheless, the abundant proteins involved in starch and sucrose metabolism detected in wheat and potato tuber amyloplasts [10,11] were not well represented among Arabidopsis root plastid proteins, which imply that the root plastids do not play a major role as a long-term carbohydrate storage in Arabidopsis. Alternatively, as the amyloplasts are fragile and difficult to isolate, the possibility that amyloplasts have damaged and escaped from our root plastid sample during the isolation cannot be excluded. It is also worth noting that in highly purified amyloplast fraction a number of mitochondrial proteins were identified (for instance mitochondrial porin) [10][11][12], which may indicate either a true dual location of some proteins, or co-purification of amyloplasts and mitochondria.
In addition to a number of plastid proteins (Table 1), also a range of mitochondrial proteins were detected, mostly representing citric acid cycle and electron transfer chain components (Table 2). In agreement with previous studies [57], many of these proteins were present as multiple spots, suggesting that mitochondrial proteins are targets of multiple post-translational modifications. Indeed, it has been recently shown that organellar proteins are frequently modified by e.g. phosphorylation [103,104], Lys acetylation [105] and methylation [106,107]. The functional consequences of these modifications, however, are still unknown. We also reliably identified some proteins, which may exist in multiple compartments or which have not been annotated as plastid or mitochondrial proteins. For instance jacalinrelated lectins are carbohydrate binding proteins involved in defense signaling pathways [108]. In addition to the membrane-bound isoforms, plants contain soluble lectins in the cytosol, nuclei, chloroplasts as well as in the vacuole and in apoplast [109][110][111][112]. Whether the jacalin-related lectin detected in several spots (see Additional file 3; Table S1) represents a protein truly accumulating in the mitochondria or root plastids remains to be elucidated.

Conclusions
In this study, we have identified 84 root plastid-targeted and 77 mitochondrium-targeted proteins and show that Arabidopsis root plastids have a broad biosynthetic capacity representing synthesis of amino acids and fatty acids as well as carbon, nitrogen and sulfur assimilation. This work underlines the importance of root plastids for secondary metabolism and root originated stress responses, and shows that Arabidopsis root plastids do not have a major role in long-term storage of carbohydrates. Currently we do not know whether the plastids present in our root plastid fraction represent one uniform plastid type, or multiple distinct types of leucoplasts, which may have specialized functional features. Moreover, the impact of cell or tissue types and developmental stage on the structural and functional properties of root plastids remains an interesting topic for future research.

Plant material
Arabidopsis thaliana ecotype Columbia-0 seeds, purchased from Nottingham Arabidopsis Stock Centre (NASC; UK), were surface sterilized and plated on the uppermost surface of square plates (½ x Murashige and Skoog with vitamins (Duchefa) in 50 mM MES buffer pH 5.7, 0,8% plant agar (Duchefa). The plants were grown on vertical plates under 16/8 h light/dark photoperiod at 120 μmol photons m − 2 s − 1 at 23°C for four weeks.

Root plastid isolation SDS-PAGE and Western blotting
Root plastids were isolated as described in [113] with the following modifications. 500 mg roots of four week old plants were ground in pre-cooled mortar in 1.5 ml homogenization buffer (50 mM Tris-HCl, pH 7.5, 330 mM sorbitol, 1 mM EDTA, 1 mM MgCl2, 0,1% BSA, PierceTM Protease inhibitor, 1 tablet/10 ml). The homogenate was filtered through one layer of pre-soaked Miracloth and centrifuged (4000 x g, 3 min, 4°C). Supernatant was collected as a cytosolic fraction, and the plastid enriched pellet was gently re-suspended in 50 μl homogenization buffer by using small brush and pipetted with a wide tip on 1.8 ml 10% Percoll followed by centrifugation (4000 x g 5 min, 4°C). The plastid fraction was collected and washed twice with 500 μl homogenization buffer. The final pellet was resuspended in 50 μl shock buffer (5 mM sucrose, 10 mM Hepes-NaOH, 5 mM MgCl 2 ) and left on ice for 5 min.

2D gel electrophoresis and staining of gels
For 2D electrophoresis, root plastid samples of approximately 150 μg protein with rehydration buffer (8 M urea, 2 M thiourea, 4% CHAPS, 100 mM dithiothreitol, and 0.5% 3-11 NL IPG buffer) was freshly prepared, covered with aluminum foil, and vortexed with maximum speed for 2 h. A protein isoelectric focusing unit was used with NL (non-linear) 18 cm IPG strips (Bio-Rad) for pH ranges 3-11. IPG strips were rehydrated in 340 μl rehydration buffer overnight. Protein samples were cup loaded directly onto the strips, with the cup placed close to the anodic end of the strip. The first-dimension separation was performed using the Ettan IPGphor 3 isoelectric focusing system. The proteins were focused at 150 V for 3 h and at 300 V for 3 h, then the voltage was raised to 1000 V during 6 h and to 10,000 V during 2 h, and then kept at 10,000 V for 3 h [114]. After electrofocusing, the proteins in the strips were reduced with 130 mM dithiothreitol (DTT) and alkylated with 135 mM iodoacetamide (IAA) in equilibration buffer (375 mM Tris-HCl pH 8.8, 6 M urea, 20-30% glycerol, 2% SDS). Second-dimension electrophoresis was performed on 14% linear SDS gels.
After electrophoresis, gels were stained with SYPRO Ruby protein stain® according to the manufacturer's instructions as follows. Gel fixation was performed in 40% methanol, 10% trichloroacetic acid for 3 h, followed with three washes in water for 10 min each. Subsequently, gels were covered with 330 ml of SYPRO Ruby protein stain and left overnight in darkness with continuous gentle agitation. After staining, gels were rinsed in 10% ethanol, 7% acetic acid for 45 min followed by washing in water. Stained proteins were visualized using 300 nm UV transilluminator and imaged with the Geliance 1000 imager using Cy3 filter (PerkinElmer). Thereafter, the same gels were silver stained to enable cutting of the protein spots. MS compatible silver staining was performed according to [115]. After rinsing in 20% ethanol and water for 15 min each, gels were sensitized by incubation in 200 ml 1.2 mM Na 2 S 2 O 3 ·5 H 2 O for 90 s. Subsequently, gels were washed two times for 20 s with water and silver stained. Staining was performed in 12 mM AgNO 3 for 30 min on the platform shaker followed by rinsing with water for 20 s. Afterwards, gels were developed in 217 mM K 2 CO 3 , 0,6 mM Na 2 S 2 O 3 , 0,07% formalin by shaking gently on the platform shaker for 2-5 min until the spots were clearly visible. The development was stopped by shaking the gels for 2 min in stop solution (2.5% acetic acid, 400 mM Tris-HCl).

MS analysis, database searches and functional classification
Selected protein spots from 2D gels were excised manually and subjected to in-gel digestion with 0.2-0.3 μg trypsin (Promega V5111) at 37°C for 18 h. After digestion, tryptic peptides were extracted from the gel pieces with acetonitrile (ACN) and subsequently with 50% ACN / 5% HCOOH solution. Total extracts were pooled, dried in vacuum centrifuge and stored at − 20°C. Directly prior to MS analyses, the peptides we dissolve in 10 μl 0.1% HCOOH by vortexing, incubating at 37°C for 15 min. Five μl peptides were injected to LC-MS/MS analysis performed on a nanoflow HPLC system (Easy-nLC1000, Thermo Fisher Scientific) coupled to the Q Exactive mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) equipped with a nano-electrospray ionization source. Peptides were first loaded on a trapping column (100 μm ID × 2 cm) and subsequently separated inline on an analytical column (75 μm ID × 15 cm). The packing material for the both columns was ReproSil-Pur 5 μm 200 Å C18-AQ (Dr. Maisch HPLC GmbH, Ammerbuch-Entringen, Germany). The mobile phase consisted of water with 0.1% formic acid (solvent A) or acetonitrile/water (80:20 (v/v)) with 0.1% formic acid (solvent B). A 10 min gradient from 8 to 43% B was used to elute peptides. MS data was acquired automatically by using Thermo Xcalibur 4.0 software (Thermo Fisher Scientific). An information dependent acquisition method consisted of an Orbitrap MS survey scan of mass range 300-1750 m/z followed by HCD fragmentation for 10 most intense peptide ions.
Data files were searched for protein identification using Proteome Discoverer 2.2 software (Thermo Fisher Scientific) connected to an in-house server running the Mascot 2.5.1 software (Matrix Science). Data was searched against SwissProt database (version 2017_03). The database search settings included a taxonomy filter 'Arabidopsis thaliana", trypsin as an enzyme, carbamido methyl as fixed modification and methionine oxidation as variable modification. Two missed cleavages were allowed. A significance threshold of p < 0.05 was used.
Classification of proteins was performed using MapMan software (http://mapman.gabipd.org/mapman-download) and the results were manually verified using the annotations of UniProt database.