Novel Tonoplast Transporters Identiﬁed Using a Proteomic Approach with Vacuoles Isolated from Cauliﬂower Buds

Young meristematic plant cells contain a large number of small vacuoles, while the largest part of the vacuome in mature cells is composed by a large central vacuole, occupying 80% to 90% of the cell volume. Thus far, only a limited number of vacuolar membrane proteins have been identiﬁed and characterized. The proteomic approach is a powerful tool to identify new vacuolar membrane proteins. To analyze vacuoles from growing tissues we isolated vacuoles from cauliﬂower ( Brassica oleracea ) buds, which are constituted by a large amount of small cells but also contain cells in expansion as well as fully expanded cells. Here we show that using puriﬁed cauliﬂower vacuoles and different extraction procedures such as saline, NaOH, acetone, and chloroform/methanol and analyzing the data against the Arabidopsis ( Arabidopsis thaliana ) database 102 cauliﬂower integral proteins and 214 peripheral proteins could be identiﬁed. The vacuolar pyrophosphatase was the most prominent protein. From the 102 identiﬁed proteins 45 proteins were already described. Nine of these, corresponding to 46% of peptides detected, are known vacuolar proteins. We identiﬁed 57 proteins (55.9%) containing at least one membrane spanning domain with unknown subcellular localization. A comparison of the newly identiﬁed proteins with expression proﬁles from in silico data revealed that most of them are highly expressed in young, developing tissues. To verify whether the newly identiﬁed proteins were indeed localized in the vacuole we constructed and expressed green ﬂuorescence protein fusion proteins for ﬁve putative vacuolar membrane proteins exhibiting three to 11 transmembrane domains. Four of them, a putative organic cation transporter, a nodulin N21 family protein, a membrane protein of unknown function, and a senescence related membrane protein were localized in the vacuolar membrane, while a white-brown ATP-binding cassette transporter homolog was shown to reside in the plasma membrane. These results demonstrate that proteomic analysis of highly puriﬁed vacuoles from speciﬁc tissues allows the identiﬁcation of new vacuolar proteins and provides an additional view of tonoplastic proteins. glycerol immersion objective for onion cells coupled to Leica TCS SP2 spectral confocal and multiphoton microscope (Leica Microsystems). Fluorescing cells were imaged at an excitation wave-length of 488 nm and the emission signal was recovered between 495 and 530 nm for GFP fusion, and excitation of 543 nm and emission signal recovered between 555 and 620 nm for DsRed fusion. Images were processed with the Leica confocal software.

Young meristematic plant cells contain a large number of small vacuoles known to fuse during cell expansion and finally form a single large central vacuole, which occupies 80% to 90% of the cell volume (Marty, 1999). Cell expansion is, therefore, closely linked with increase of vacuolar volume. To reach the full cell size without expending too much energy, large amounts of inorganic ions are transported into the vacuole during vacuolar growth, concomitantly followed by a substantial water influx (Maeshima, 2001;Martinoia et al., 2007). In adult cells the vacuole plays a central role in cytosolic homeostasis. The concentration of metabolites in the cytosol and other metabolic active compartments has to be tightly regulated to facilitate the optimal functioning of a large number of metabolic pathways. Uptake, storage, and release of solutes into and out of the vacuole are thus of pivotal importance for a plant cell. Our knowledge of vacuolar transport processes is mainly derived from studies on vacuoles and tonoplast vesicles isolated from fully expanded cells (Maeshima, 2001;Martinoia, et al., 2007).
Two vacuolar proton pumps, the vacuolar H 1 -ATPase (V-ATPase) and the vacuolar pyrophosphatase (V-PPase) generate an electrochemical gradient, which can be used to drive solute uptake into the vacuole (Rea and Sanders, 1987;Maeshima, 2001). The membrane potential across the tonoplast enables the accumulation of anions. In the case of organic acids they may be additionally trapped within the vacuole by changing their protonation state within the acidic lumen. Cations, such as sodium or calcium, and uncharged solutes such as Suc can be accumulated by use of antiporters, which exchange protons with the corresponding solute. In addition, glutathionated, glucuronated, or glucosylated compounds may be transported into the vacuole by ATP-binding cassette-type transporters (Rea et al., 1998;Theodoulou, 2000;Martinoia et al., 2002).
Several vacuolar membrane proteins have been identified during the last years. In terms of abundance in the membrane the aquaporins are by far the most prevalent tonoplastic proteins (Maeshima, 2001). The sodium/proton antiporter has probably attracted most interest since overexpression of this transporter leads to salt-tolerant plants (Apse et al., 1999;Zhang and Blumwald, 2001). The members of the CAX family exchange calcium against protons and at least some of these transporters are also able to exchange cadmium against protons and thus are of potential interest for phytoremediation (Hirschi et al., 2000;Hirschi, 2001;Cheng et al., 2003).
NRAMP is a transporter implicated in iron export from the vacuole, since deletion mutants in this transporter exhibit a slow growth phenotype under low iron conditions, and double deletion mutants AtNRAMP3xAtNRAMP4 cannot mobilize iron from the seed vacuoles (Thomine et al., 2000(Thomine et al., , 2003Lanquar et al., 2005). Recently two vacuolar key transporters have been identified at the molecular level: a K 1/ Ca 21 channel (Peiter et al., 2005) corresponding to the welldescribed slow vacuolar channel and a nitrate antiporter responsible for the accumulation of nitrate, belonging to the family of the ClC channels (De Angeli et al., 2006). Furthermore, a malate transporter has been identified (Emmerlich et al., 2003). This transporter does not correspond to the described vacuolar dicarboxylate channel (Hurth et al., 2005) but plays an important role in vacuolar storage of malate. Glutathionated, glucuronated, and possibly malonylated compounds are substrates of ATP-binding cassette transporters. In view of the multitude of compounds identified in the vacuole it must be postulated that a large number of vacuolar transporters still await identification and characterization. One attempt to gain more insights into the vacuolar proteome is to isolate highly enriched vacuolar preparations and subject them to a proteomic approach. Such an approach has been undertaken for Arabidopsis (Arabidopsis thaliana) and barley (Hordeum vulgare; Carter et al., 2004;Sazuka et al., 2004;Shimaoka et al., 2004;Szponarski et al., 2004;Endler et al., 2006;Jaquinod et al., 2007).
Each group of researchers used different purification and separation methods. Interestingly, the overlap between identified proteins is relatively low. Between 34 and 650 different proteins, including soluble and membrane proteins were detected. From these 23 to 195 were reported to be membrane proteins. This discrepancy can be attributed to, for example, different amounts of proteins, and individual techniques used in the respective approaches. All six groups of researchers identified at least three subunits of the V-ATPase. Besides tonoplastic proteins, nonvacuolar proteins from other organelles like mitochondria, plasma membrane, and chloroplasts were additionally detected. Therefore, for all newly identified putative vacuolar proteins final proof for vacuolar localization has to be shown. However, this aspect has only been addressed by Endler et al. (2006) and Jaquinod et al. (2007). An alternative strategy to identify membrane proteins from organelles has been presented by Dunkley et al. (2006). Using density gradients in combination with isotope tagging they could differentiate membrane proteins from different organelles.
As mentioned above, vacuoles undergo different developmental stages. To analyze vacuoles from growing tissues we isolated corresponding organelles from cauliflower (Brassica oleracea) buds, which are mainly constituted of small and only partially expanded cells. Here we show that using different treatments for the purification of the tonoplast and analyzing these against the Arabidopsis database, 102 membrane proteins and 214 soluble cauliflower proteins could be identified. From the 102 identified membrane proteins 32.3% of the peptides could be assigned as vacuolar membrane proteins, comprising nine vacuolar membrane proteins. We identified 57 proteins containing at least one membrane spanning domain with an unknown subcellular localization. To verify whether some of the newly identified proteins are indeed tonoplastic we constructed and expressed five GFP fusions, four of which are localized in the tonoplast.

Purification of Tonoplast Proteins
Using the procedure published by Marty-Mazars et al. (1995), an average of 1 mg tonoplast proteins was purified from 750 g of cauliflower heads. The purity of the Nycodenz-purified fraction was assessed by using marker enzymes and the ATPase assay in the presence of specific inhibitors (Table I) as well as by western blots using specific antibodies (Fig. 1A). Based on the enzymatic analyses we determined that the enriched tonoplast fractions were contaminated up to 1.3% with plasma membranes, chloroplasts, mitochondria, Golgi, and endoplasmic reticulum (ER; see Table I). Estimating the contamination of the vacuolar fraction using the western-blot technique shows that this fraction is only slightly contaminated (see Fig. 1A). The vacuolar marker g-TONOPLAST INTRINSIC PROTEIN (TIP) is strongly enriched in the vacuolar preparation compared to the microsomal fraction. In contrast, a strong signal for the PLASMA MEMBRANE AQUAPORIN1 (PAQ1) could be detected in the microsomal fraction. The corresponding signal was only very week in the vacuolar fraction and based on the ratio of the bands between the microsomal fraction and the purified vacuoles a contamination of less than 5% can be estimated. The signals of the ER and plastidic markers BIP and ATPase (subunit a) were weaker than that of the PAQ1 and could not be detected in the vacuolar preparation. Therefore, it can be concluded that the vacuolar preparation is only marginally contaminated with other membranes. This highly purified vacuolar membrane preparation was used to perform the vacuolar proteomic analysis. et al. (2003) already demonstrated the importance of using complementary methods to identify as many membrane proteins as possible. However, such an approach has not been undertaken using plant vacuoles. Therefore, we used five different extraction procedures: namely alkaline (NaOH) and saline (KI) washing, acetone and chloroform/methanol (C/M) extraction, and treatment with sodium phosphate (at pH 6) followed by C/M extraction (C/M pH 6; Fig. 1B).

Ferro
To identify the hydrophobic proteins present in the different extracts, proteins were further fractionated by SDS-PAGE (Fig. 1B). Prior to mass spectrometry (MS) we separated the tryptic peptides by reversed phase chromatography (C18) and analyzed them by nanospray ionization MS/MS with an ion trap. Proteins were identified by SEQUEST searches using the National Center for Biotechnology Information Arabidopsis database (http://www.ncbi.nlm.nih.gov/). Protein identifications were based on the Peptide and Protein-Prophet data analysis software (Keller et al., 2002;Supplemental Table S1; cutoff of 0.9). The Arabidopsis database was chosen because cauliflower and Arabidopsis are closely related Brassicacea with highly conserved protein sequences. The Arabidopsis database is well annotated and has a size that is compatible with statistical tools for data analysis. Using higher search spaces (e.g. the Viridiplantae section of the National Center for Biotechnology Information database) increases the risk of false-positive protein identifications and severely hampers the use of statistical tools such as Peptide and Protein-Prophet.
Using the membrane fractions described above we identified 102 cauliflower membrane proteins with our proteomic approach (Table II). Further information on additional identified proteins (214 soluble proteins) is available in the Supplemental Table S1. Table II lists all identified membrane proteins sorted according to the number of their putative transmembrane spanning domains (TMDs) that were determined using the Aramemnon database (http://aramemnon.botanik.uni-koeln. de/) and the MIPS database (http://mips.gsf.de/). We included information about the molecular masses, the pI, the subfractions in which each protein was identified, i.e. acetone, KI, NaOH, C/M, or C/M pH 6, the number of analyzed gel slices, and the overlap to other vacuolar proteomic studies (Carter et al., 2004;Sazuka et al., 2004;Shimaoka et al., 2004;Szponarski et al., 2004;Endler et al., 2006;Jaquinod et al., 2007). The number of unique peptides detected in the different treated fractions is indicated. A total of 35 of 102 membrane proteins (34.3%) are identified by a unique peptide hit.
We detected 28 different membrane proteins of 258 peptides in the subfraction of KI (eight slices), 36 of 290 peptides in the subfraction of acetone (eight slices), 13 of 33 peptides in the subfraction of NaOH (four slices), 72 of 1,107 peptides in the subfraction of NaOH (10 slices), 27 of 109 peptides in the subfraction of C/M (four slices), and 18 of 75 peptides in the subfraction of C/M pH 6 (four slices; Table II; Supplemental Table  S1). A total of 55% (56 membrane proteins out of 102) of the membrane proteins were found solely in one of the washing or extraction procedures applied (Table II). Comparing the replicates of membranes washed with NaOH (Table II) we identified 83% more membrane proteins by cutting the gel into 10 slices (75 membrane proteins) than when cutting the gel into only four (13 membrane proteins). Interestingly, in all subfractions the total number of peptides from known tonoplast proteins ranged between 30% and 50%. However, in Table I. Enzymatic characterization of the tonoplast fraction Specific activities of marker enzymes in membrane fractions isolated from cauliflower inflorescence. Marker enzyme activities for mitochondria (CCO), plastids and mitochondria (F-ATPase), Golgi apparatus (IDPase), ER (CCR), and plasma membrane ATPase (P-ATPase) are presented together with the marker enzyme activity for tonoplast (V-ATPase) in microsomal and tonoplast-enriched fractions. Relative specific activity (degree of enrichment).
the C/M subfraction the total number of peptides from known tonoplast proteins was only 12%. These results demonstrate the importance of high resolution protein fractionation and that the application of different extraction methods increases the number of peptides that can be identified. We identified nine intrinsic membrane proteins, which are already known to be localized in the tonoplast. Calculating the total number of peptides we detected 33.5% (626 of 1,870) of all peptides belonging to tonoplast proteins (Fig. 2). If the nonmembrane spanning subunits of the V-ATPase are included, 57.1% (1,655 of 2,899) of the peptides could be attributed to the tonoplast. Based on the number of total peptides detected, the three major tonoplast proteins are the V-PPase (AtAVP-3; 345 peptides), the integral V 0 -ATPase subunit a3 (89 peptides), and the integral V 0 -ATPase subunit a2 (86 peptides). In addition, two other V-PPases (the homologs to AtAVP2 and AtAVPL1), the tonoplast intrinsic protein (the homolog to AtTIP1.2), and the integral V 0 -ATPase subunit c and two Ca 21 -ATPases (the homolog to AtACA4 and AtACA11; Geisler et al., 2000;Baxter et al., 2003) were identified. Moreover, 57 membrane proteins (26.4% peptides) are still not localized (Fig. 2) and 36 membrane proteins (40.1% peptides) could be attributed to already known nonvacuolar proteins. Consequently, for more than 56% (57 proteins of 102) of the membrane proteins no function has yet been attributed (Fig. 2).
To obtain an overview of the coverage of our proteomic data we had a closer look at the vacuolar ATPase. The V-ATPase is a large membrane bound multisubunit enzyme complex composed of two functional domains: a water soluble V 1 domain and a membrane embedded V 0 domain (Fig. 3). The Arabidopsis V-ATPase contains 13 distinct functional subunits (Sze et al., 2002). We detected all subunits of the V-ATPase complex except subunits VHA-c" and VHA-e ( Fig. 3; Supplemental Table S1). The identification of 11 out of 13 V-ATPase subunits indicates the high coverage of the vacuole proteome reported here.

Specification of New Tonoplast Proteins
In our search for new transporters and channels in the tonoplast we were interested in novel vacuolar membrane proteins with more than two transmembrane domains. To provide further independent evidence for a vacuolar localization of the newly identified putative tonoplast proteins we transiently transformed both Arabidopsis protoplasts and onion (Allium cepa) epidermal cells with five GFP fusion proteins and analyzed their respective localization (Fig. 4). As a vacuolar control we used the C-terminal DsRed-tagged KCO1 channel from Arabidopsis (At5g55630), known to reside in the vacuolar membrane (Czempinski et al., 2002).
Synthesis of the recombinant KCO1TDsRed fusion protein in both Arabidopsis protoplasts or epidermal onion cells resulted in red fluorescence at the tonoplast ( Fig. 4, C and K). The selected putative vacuolar membrane proteins (At5g45370, At1g16390, At4g28770, and At1g06470) fused to GFP are targeted to the vacuolar membrane ( Fig. 4, B, F, I, and M). In transformed Arabidopsis protoplasts fluorescence is clearly surrounding the central vacuole (Fig. 4, B and F). The tonoplast localization of two proteins (At5g45370 and At4g28770) was additionally demonstrated by colocalizing of the GFP fusion proteins and the KCO1T DsRed fusion protein (Fig. 4,D and L). In contrast to the four membrane proteins described above, one GFP fusion protein identified in our analysis, namely At1g17840, was localized at the plasma membrane ( Fig. 4P), indicating the importance of localization studies in proteomic analysis.
The function of the four newly localized tonoplast proteins is still unknown. The gene At1g16390 encodes a putative organic cation transporter with 11 TMDs. Its product was detected in the KI subfraction as a single hit in two different gel slices. Proteins At1g06470 and At5g45370 were also detected as single hits in the acetone subfraction. The gene At1g06470 encodes an Figure 1. Evaluation of vacuolar membrane purity by western blot analysis (A) and fractionation strategy for cauliflower tonoplast proteins (B). A, Purified vacuoles were treated with five different extraction methods: KI and alkaline (NaOH) washing, acetone and C/M extraction, and treatment with sodium phosphate pH 6 followed by C/M pH 6 extraction, subjected to one-dimensional SDS-PAGE followed by in gel digestion and LC/MS/MS. B, For analysis of purification 10 mg of total membrane proteins (totM) and tonoplast proteins (vacM) were loaded on each lane. Compartment specific antibodies were used for westernblot analysis: chloroplastic ATPase a-subunit (ATP-a; 55 kD), luminal binding protein (BIP; 73 kD), plasma membrane aquaporin (PAQ1; 30 kD), and g-TIP (25 kD).  Table II. Putative vacuolar membrane proteins sorted according to the number of their putative TMDs AGI Acc. No., Arabidopsis Genome Initiative accession number; AGIs in italics correspond to other possible proteins matching with the same peptides or part of them, AGIs in bold correspond to novel localized membrane proteins. TMD, Putative transmembrane domains were assumed according to the Aramemnon database (http://aramemnon.botanik.uni-koeln.de/) and the MIPS database (http://mips.gsf.de/). Loc., Subcellular localization predictions according to the Aramemnon database [CL, Chloroplast; ED, endosomes; MT, mitochondria; PM, plasma membrane; (PM), novel plasma membrane protein; PR, peroxisomes; T, tonoplast; (T), novel tonoplast protein; U, unknown]. PP, Protein-Prophet scores above 0.9 were accepted; at this cutoff, the rate of false-positive protein identifications is less than 10%; l, lower PP score but additional localization experiments (in bold); kD, predicted M r in kD; tonoplast fraction treated with KI, potassium iodide; NaOH, sodium hydroxide; C/M pH 6, treatment with sodium phosphate buffer at pH 6 followed by C/M; numbers in parentheses mean the number of gel slices. The fraction in which each protein was identified is indicated by the unique number of peptides. Ref, Indicates Arabidopsis homolog(s) found by: a, Carter et al. (2004); b, Endler et al. (2006);c, Sazuka et al. (2004); d, Shimaoka et al. (2004); e, Szponarski et al. (2004); and f, Jaquinod et al. (2007)   Our data demonstrates that it is possible to identify new tonoplast proteins from cauliflower by comparison with peptides from the Arabidopsis databank.

DISCUSSION
The main goals of this work were, first to identify new putative vacuolar proteins of a vacuolar preparation containing a large number of small vacuoles from developing cells, second, to use plant material of which large amounts of vacuolar membranes can be quickly and easily isolated, and third, to prove that new candidate proteins are indeed targeted to the vacuolar membrane. Despite the important role of the vacuole for temporary storage, detoxification, and cytosolic homeostasis in plant cells only a limited number of vacuolar proteins have been identified so far.
Since the vacuolar membrane contains less than 1% of the total cellular proteins and as this membrane is less protein dense compared to most other cellular membranes, small contaminations may result in the detection of a large number of nonvacuolar proteins. The heterogeneity observed in previous studies (Carter et al., 2004;Sazuka et al., 2004;Shimaoka et al., 2004;Szponarski et al., 2004;Jaquinod et al., 2007) may therefore be due to the different plant material like Arabidopsis mesophyll cells (Carter et al., 2004;Sazuka et al., 2004) or cell suspension culture (Shimaoka et al., 2004;Szponarski et al., 2004), or to different preparation techniques. The fact that a small contamination of the vacuolar preparation can result in the detection of a considerable number of nonvacuolar proteins is also shown in this work (35.3% proteins corresponding to 40.1% peptides; see Fig. 2). In this context it should be mentioned that the inner mitochondrial membrane, as well as the plasma membrane and other internal membranes, contain far more proteins compared to the vacuolar membrane. In density gradients, the latter membranes together with non-rRNA containing ER are the lightest membranes. The fact that more peptides that could be attributed to PIPs rather than TIPs were identified is astonishing. It may be explained that small, nonexpanding vacuoles contain low amounts of aquaporins, since the water permeability in lytic vacuoles is mainly regulated at the transcription level. In   contrast, plasma membrane aquaporins are regulated by phosphorylation and appear to be less regulated at the transcription level (Hu et al., 2003;Luu and Maurel, 2005). This hypothesis is sustained by the western blots showing that the PAQ1 signal is much stronger than the TIP signal in the microsomal fraction. This is also in line with experiments analyzing PIPs and TIPs in developing fruits (Hu et al., 2003).
Furthermore, for most other membranes many more membrane proteins have been identified and characterized. In addition, the vacuole is a single continuous compartment transected by tubular-like transvacuolar strands and coupled with the intimate associations with other organelles during stress adaptation (Reisen et al., 2005). Another possible source of contaminants is the basal level of autophagy and the internalization of cytosolic proteins, as well as organelles into the vacuole. Therefore, a comparison of peptides obtained from a vacuolar proteomic will overestimate the number of nonvacuolar proteins.
A further limiting aspect in a vacuolar proteomic approach is the minute amount of membranes that can be isolated. An in-depth proteomic approach where low abundance components of the membrane will also be identified requires washing steps and fractionation. Starting with Arabidopsis mesophyll cells, which probably allow a better purification than suspension cultured cells due to the larger differences in the density of vacuoles and protoplasts, only limited material can be isolated. However, techniques are improving fast and will allow to detect more proteins even with limited amounts of membranes. The scope of our study was also to identify membrane proteins expressed in the early phase of vacuole development. As a model plant for this study we have chosen cauliflower. Using cauliflower buds, vacuoles from different developmental stages were isolated, resulting in a preparation of small and large vacuoles (see Fig. 1; Dozolme et al., 1995). A further advantage of cauliflower is that the plant material is not a limiting factor and that the mechanical slicing technique allows isolation of large amounts of highly purified vacuoles (Marty and Branton, 1980;Marty-Mazars et al., 1995). However, a database for comparison is required to allow a complete or almost complete coverage of the peptides. Cauliflower is a Brassicacea as is Arabidopsis and thus shares very high sequence identity (95% identity for aquaporins [Barrieu et al., 1998], 87% to 97% identity for other membrane proteins). Therefore, it is not surprising that most of the peptides we identified could be assigned to an Arabidopsis homolog, proving that cauliflower is a suitable model plant. Different washings and extractions of the vacuolar preparation resulted in the identification of different proteins, underlining the necessity to isolate large amounts of membranes to obtain a high coverage.
In total, we identified 102 membrane and 214 soluble proteins, including seven nonmembrane spanning domains of the V-ATPase (see Table II; Supplemental  Table S1). This implies that 32.3% of our tonoplast fractions belong to membrane proteins. Of these, three membrane proteins (2.9% of all membrane proteins) were identified in the various washing and extraction procedures (Table II), including the V-PPase (homologs to AtAVP-3), the integral V 0 -ATPase subunit a3, and the putative plasma membrane intrinsic protein 3a (homologs to AtPIP2.7). Moreover, only 27 membrane proteins (26.5%) were detected in three of the five treatments. Less than the half of all membrane proteins (46 of 102 proteins) were identified in at least two procedures (Table II). Figure 5 shows the cross correlation of identified cauliflower membrane proteins in our study with the three largest Arabidopsis proteomic approaches of the vacuolar membrane (Carter et al., 2004;Shimaoka et al., 2004;Jaquinod et al., 2007). Well-known proteins of the chloroplast, mitochondria, plasma membrane, and the ER proteins AtACA1, calnexin 1, and AtRER1A were excluded from the analysis. Of 76 vacuolar and putative vacuolar membrane proteins 40 proteins (52%) were exclusively identified in our study, suggesting a higher relative abundance of these proteins in cauliflower buds vacuoles compared to Arabidopsis leaf mesophyll and cell culture vacuoles. Interestingly, when the corresponding Arabidopsis genes for the newly discovered vacuolar and putative vacuolar membrane proteins were analyzed for their expression pattern by GENEVESTIGATOR (www.genevestigator.ethz.ch), a very large part of them was mainly expressed in young, developing parts (elongation zone, epidermal atrichoblast, root hair zone, shoot apex) of the plant (Supplemental Fig. S1). This result confirms the interest to perform vacuolar proteomics in different tissues and at different developmental stages.
We detected 11 out of 13 subunits (Sze et al., 2002). Subunit VHA-a is abundant in all five different fractions, whereas only subunits VHA-c" and VHA-e were not identified. VHA-c" occurs with low abundance and VHA-e is a very small protein (approximately 8 kD) and, therefore, probably migrates too fast for detection in the 10% SDS-PAGE as used under our conditions. Until now, no tonoplast proteome is available containing these two subunits (Carter et al., 2004;Sazuka et al., 2004;Shimaoka et al., 2004;Szponarski et al., 2004;Figure 5. Overlap between different proteomic analyses of the tonoplast. The Venn diagram compares membrane proteins identified in cauliflower with membrane proteins identified in the three largest Arabidopsis tonoplast proteomic analyses (Carter et al., 2004;Shimaoka et al., 2004;Jaquinod et al., 2007). Numbers in parentheses indicate the total number of identified proteins with at least one transmembrane domain in the particular study, excluding contaminating proteins of the proteomic analyses. The prediction of TMDs by TMHMM was considered for membrane proteins identified by Carter et al. (2004).  Endler et al., 2006;Jaquinod et al., 2007). Calculating the total number of peptides from all subunits of the V-ATPase we detected 1,029 peptides. Compared to the total number of peptides from the nonvacuolar ATPases, only 146 peptides from four subunits of the F-ATPase (FHA-b, FHA-e, FHA-g, and FHA-D) and 31 peptides from the P-ATPase were identified (see Supplemental Table S1). To conclude, 85.3% of peptides (1,029 of 1,206) of all ATPases are of vacuolar origin. This comparison probably reflects much more the enrichment of the vacuolar membrane in our fraction than the total proportion of proteins. It is also interesting to note that it has been reported that in young tissues the V-PPase is the predominant pump (Maeshima, 2001). Despite the fact that our proteomic study is not quantitative, the large number of peptides corresponding to this proton pump indicates that this may also be true for cauliflower buds.
To get a hint on the reliability of identification and for the vacuolar localization we made GFP fusion constructs for five unknown proteins (see Fig. 4). Four of them were targeted to the tonoplast, the fifth to the plasma membrane. The first of the novel tonoplast proteins, At1g16390, has similarities to organic cation transporters like At1g79360 with 11 TMDs. Organic cation transporters influence the plasma concentration of many endogenous molecules and of an even wider array of xenobiotic compounds. They are expressed in many tissues and play a critical role in fundamental cellular processes (Bednarczyk et al., 2003). However, their function in plants has not been investigated.
The second novel tonoplast protein with nine TMDs, At5g45370, belongs to the nodulin N21 family. The closest homologs are At4g19185 and At3g45870 with 74% and 66% similarity, respectively. At5g45370 belongs to the drug/metabolite transporter superfamily (Jack et al., 2001), however, until now, no member of the family has been characterized.
Our third tonoplast candidate, At1g06470, is an integral membrane protein with eight predicted TMDs. The closest homologs, At2g25520 and At5g25400, both with 28% similarity, belong to a group of phosphate translocators that function as antiport systems using inorganic phosphate and phosphorylated C 3 , C 5 , or C 6 compounds as counter substrates (Flü gge, 1999). Preliminary tests investigating whether this gene could code for the vacuolar phosphate transporter indicated that the corresponding deletion mutants were not affected in vacuolar phosphate and sugar contents (data not shown).
The fourth tonoplast membrane protein, At4g28770, is a putative senescence-associated protein with three TMDs. There is just one close homolog, At2g20230, exhibiting a similarity of 75%. During senescence the vacuole has an important role in degradation of proteins and possibly also in temporary storage of sugars and amino acids. Furthermore, many compounds stored in the vacuole are retranslocated to young, growing tissues or seeds during senescence. The increase of the corresponding transcript level during senescence may indicate that this membrane protein may be related to such a process.
We are aware that from the localization of five proteins we cannot directly extrapolate the number of tonoplast proteins in our preparation. However, if we assume that around 80% of the novel identified proteins containing at least two membrane spanning domains are vacuolar 27 so far unknown vacuolar membrane proteins have to be postulated to exist.
In conclusion, despite the fact that the vacuole contains less than 1% of the cellular protein, a large number of proteins can be identified when large amounts of highly purified vacuoles are used in combination with the proteomic approach. This will allow for more detailed investigation into the role of vacuolar transporters in storage and homeostasis. Vacuoles are likely to fulfill functionally differing roles in various plants and tissues. Using the vacuolar proteomic approach will therefore not only help to identify new vacuolar proteins but allow us to elucidate the functional roles of vacuoles in a given plant and tissue.

General
Restriction enzymes were purchased from New England Biolabs. Oligonucleotides were obtained from Mycrosynth. DNA sequencing was conducted by the University of Zü rich. Unless otherwise indicated, all chemicals were purchased from Sigma. Escherichia coli strain XL-1 Blue (Stratagene) was generally employed for DNA cloning procedures. All cloning procedures were conducted using standard methods previously described by Sambrook et al. (1989).

Plant Material and Growth Conditions
Cauliflower (Brassica oleracea) buds were obtained at commercial fresh produce outlets. Arabidopsis (Arabidopsis thaliana) cells were grown in Murashige and Skoog medium as previously described (May and Leaver, 1993). The cell cultures were maintained at 25°C in an orbital shaker (150 rpm) under normal illumination. Cells were collected 7 d after subculture (Millar et al., 2001).

Tonoplast Preparation
Intact vacuoles and tonoplast fragments from cauliflower buds were prepared as previously reported (Leigh et al., 1979;Marty and Branton, 1980). Briefly, 750 g of fresh tissue was sliced three times in 500 mL of ice-cold collection medium (1 M sorbitol, 5 mM EDTA, 4 mM dithiothreitol [DTT], 50 mM Tris-HCl, pH 7.6) using a motor-driven tissue slicer. All subsequent manipulation was done at 4°C. The filtrate was pooled and centrifuged at 8,000g for 15 min. The sediments (crude vacuole homogenate) containing intact vacuoles, nuclei, mitochondria, plastids, and tissue fragments were resuspended in 2.5 mL of 15% (w/v) Nycodenz (Axis-Shield) in isolation medium (1.5 M sorbitol, 1 mM EDTA, 4 mM DTT, 10 mM Tris-HCI, pH 7.6) and pooled before filtration through one layer of Miracloth (Chicopee Mills, Inc.) prewetted with 15% (w/v) Nycodenz in isolation medium. The filtrate (16 mL) was divided into four tubes and overlaid first with 5 mL of isolation medium containing 8% (w/v) Nycodenz and 4 mL of isolation medium lacking Nycodenz. The discontinuous gradients were centrifuged for 45 min at 100,000g in a slowly accelerated and very slowly decelerated swing-out rotor. After centrifugation the vacuoles were found at the interface of the 0% and 8% Nycodenz layers. Vacuoles were lysed by hypotonic treatment in 10 mL of 10 mM Tris-HCI, pH 7.6. The vacuole membranes were sedimented at 140,000 g for 30 min. The sediments of vacuole membrane were extensively drained and the purity of the final fraction was assessed using marker enzymes . The purified fraction was directly solubilized for electrophoresis or resuspended in the appropriate elution solution (see subfractionation of tonoplast proteins) for extraction.

Subfractionation of Tonoplast Proteins
To remove most of the soluble proteins, the purified vacuole membrane fractions were individually treated. The tonoplast fraction (0.2 mg) was washed either by resuspending them in (1) 80% acetone; (2) in 50 mM MOPS/ NaOH pH 7.8, 1 mM DTT containing 0.1 M NaOH (final pH approximately 10); (3) in 20 mM HEPES-KOH pH 7.2 containing 0.3 M KI. After 30 min on ice, the mix was centrifuged (17,600g, 15 min, 4°C) to separate two fractions: the supernatant containing membrane associated proteins released by the respective treatment and the pellet containing the integral membrane proteins. The pelleted proteins were suspended in 30 mL Laemmli buffer (Laemmli, 1970) and analyzed by SDS-PAGE. In an additional approach, hydrophobic tonoplast proteins were extracted from purified vacuolar fraction using a C/M (5:4, v/v) mixture Marmagne et al., 2004). The tonoplast proteins (1 mg) were resuspended in 200 mL of 50 mM MOPS/NaOH pH 7.8, 1 mM DTT (C/M), or 20 mM sodium phosphate buffer pH 6 (C/M pH 6) and 1.8 mL cold C/M (5:4, v/v) was slowly added. The resulting mixtures were stored for 15 min on ice before centrifugation (4°C) for 20 min at 17,600g. The organic phase, which contains the hydrophobic proteins soluble in C/M was used for further protein analyses. C/M-soluble proteins were dried under nitrogen and the pellet proteins were then suspended in 30 mL Laemmli buffer (Laemmli, 1970) and analyzed by SDS-PAGE.

Protein Determination
Protein concentrations were determined by the dye-binding method of Bradford (1976) using bovine serum albumin as a standard.

SDS-PAGE
The buffer system of Laemmli (1970) was used for SDS-PAGE. Slab gels were 1 mm thick and consisted of a 10% (w/v) acrylamide resolving gel and a 4% (w/v) acrylamide stacking gel. Polypeptides in the resolving gel were stained with Coomassie Brilliant Blue (Fairbanks et al., 1971). For each protein fraction the gels were cut into four to 10 slices (see Table II), and each gel segment was subjected to trypsin in gel digest as described (Shevchenko et al., 1996).

Mass Spectrometric Protein Identification
Tryptic peptides of each fraction were resuspended in 5 mL of 5% acetonitrile and 0.2% formic acid (v/v) in water and loaded on laboratorymade silica capillary columns (i.d. of 75 mm, length of 9 cm; BGB Analytik AG) packed with C18 reversed-phase material (Magic C18 resins; 5 mm, 200-Å pore; Michrom BioResources). The peptide mixture was separated and eluted by a gradient from 5% to 65% acetonitrile over 2 h followed by an increase up to 80% during an additional 15 min. The flow rate at the tip of the column was adjusted to approximately 200 nL/min. LC was coupled on-line to an LCQDeca XP ion trap mass spectrometer (Thermo Finnigan) equipped with a nanospray ionization source. Mass analysis was performed with a spray voltage of 2.0 to 2.5 kV and one MS full scan followed by three data-dependent MS/MS scans of the three most intense parent ions. The dynamic exclusion function was enabled to allow two measurements of the same parent ion during 1 min followed by exclusion for 1 min.

Analysis and Interpretation of MS Data
The SEQUEST software (Thermo Finnigan) was used to search the Arabidopsis protein database (www.ncbi.nlm.nih.gov/, 3701, 01.12.2004). Arabidopsis and cauliflower are closely related species facilitating the detection of homologous proteins (e.g. 95% identity for aquaporins; Barrieu et al., 1998). Furthermore, this database was chosen because it is well annotated and has a size that is compatible with the statistical tools used for the evaluation of our proteomics data (see below). Data files were created by the SEQUEST software for every MS/MS scan with a total ion count of at least 5 3 10 4 , minimal peak count of 35, and a precursor ion mass in the range of 300 to 2,000 mass-to-charge ratio. Data were searched against the database restricted to tryptic peptides without modifications (except for carboxyamidomethylated Cyss 57.0513 and oxidized Mets 15.9994), allowing a parent mass error tolerance of 2 D and daughter ion error tolerance of 0.8 D.
To assess false-positive identification rates statistically, we performed Peptide and Protein-Prophet data analyses (Keller et al., 2002). These statistical models allow for the assessment of false-positive identification rates and we chose a cutoff of 0.9 for accepting protein identity, suggesting a falsepositive identification rate of 10% (see Table II; Supplemental Table S1). Supplemental Table S1 lists all identified membrane proteins including the Atg number, the total and unique number of peptides assigned to the protein, the protein probability calculated by Protein-Prophet, the peptide sequences, the sequence coverage in percent, and the mass differences of the theoretical and experimental masses observed for the precursor ions. In most instances, we identified the same protein from different biochemical fractions further substantiating the reliability of the identification (see Table II). In some cases where we accepted proteins with a lower Protein-Prophet score (but with an acceptable SEQUEST score [see above]) we performed additional experiments to demonstrate their vacuolar localization.
To further validate the identification procedures, we cloned the respective genes and analyzed the localization of these proteins by transient GFP localization assays (see Fig. 4; Table II). For the SEQUEST-based data interpretation we proceeded as follows: We accepted cross-correlation scores (Xcorr) of at least 2.5 for doubly, and 3.5 for triply charged ions. For peptide identifications with a dCN (normalized difference in correlation score, giving the differences between the front ranking and the following possible hit) lower than 0.1, the spectra of lower ranking hits were also examined. Identifications with a dCN of 0.0 resulting from different members of protein families, isoforms, or redundant database entries that could not be distinguished by the identified peptides are given in Table II.

GENEVESTIGATOR
The expression pattern of novel vacuolar membrane proteins and putative vacuolar membrane proteins identified by MS was analyzed using the Meta-Analyzer of GENEVESTIGATOR (Zimmermann et al., 2004). Signal intensity levels of all genes were checked using the digital northern tool. Genes showing almost always an absent call were excluded from the analysis.

Isolation of RNA
Plant material that was flash frozen in liquid nitrogen and stored at 280°C was ground in liquid nitrogen and transferred into a sterile 2 mL Eppendorf tube. RNA from different tissues (leaves, roots, flowers, stems, or protoplasts) of Arabidopsis was extracted using an RNA isolation kit, according to the manufacturer's instructions (RNeasy for plant material, Qiagen). Following extraction of total RNA, a DNase treatment (DNA-free, Ambion) was used to eliminate contamination with genomic DNA conditions. Quantification and purity of RNA were determined spectrophotometrically according to the method  Sambrook et al. (1989). The integrity of RNA samples was determined by formaldehyde agarose gel electrophoresis using a 1.2% gel prestained with ethidium bromide (Lehrach et al., 1977).

Plasmid Construction
To construct GFP fusion proteins we amplified, via PCR, the entire cDNA of At1g06470, At1g16390, At1g17840, At4g28770, and At5g45370, using the Expand High Fidelity PLUS PCR system (Roche) and either cDNA generated from Arabidopsis tissues (first-strand cDNA synthesis kit, Amersham Pharmacia Biotech) or RIKEN Arabidopsis full-length cDNA clones (Seki et al., 1998(Seki et al., , 2002, respectively, as templates. The primers we used are listed in Table  III. The obtained DNA fragments were cleaved with XbaI and XhoI (see Table  III) and inserted in frame in the vector pGFP2 (Kost et al., 1998) leading to C-terminal fusions to GFP.

Transformation of Arabidopsis Cell Culture or Onion Epidermal Cells and Confocal Laser Scanning Microscopy
Cells from an Arabidopsis cell suspension culture (Millar et al., 2001) were transiently transformed with purified plasmid DNA using Qiagen columns according to the manufacturer's protocol (Qiagen). The GFP fusion constructs were introduced into Arabidopsis protoplasts (Frangne et al., 2002) by a polyethylene glycol-mediated transformation procedure (Jin et al., 2001). Alternatively, transient expression of the fusion constructs was carried out after biolistic transformation into onion (Allium cepa) epidermal cells using the Helium Biolistic particle delivery system (Bio-Rad; Takeuchi et al., 1992).
Expression of the GFP fusion proteins was monitored 2 and 3 d after transformation by confocal laser scanning mircroscopy using a Leica DM IRE2 microscope with a 633 Plan-Apochromat oil immersion objective for protoplasts or with a 203 Plan-Apochromat glycerol immersion objective for onion cells coupled to Leica TCS SP2 spectral confocal and multiphoton microscope (Leica Microsystems). Fluorescing cells were imaged at an excitation wavelength of 488 nm and the emission signal was recovered between 495 and 530 nm for GFP fusion, and excitation of 543 nm and emission signal recovered between 555 and 620 nm for DsRed fusion. Images were processed with the Leica confocal software.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Expression pattern of newly identified vacuolar and putative vacuolar membrane proteins was investigated using Meta-Analyzer of GENEVESTIGATOR.
Supplemental Table S1. Table lists soluble and membrane proteins identified in different subfractions.