1916

DOI 10.1002/pmic.200800443

Proteomics 2009, 9, 1916–1924

RESEARCH ARTICLE

Ratio-dependent significance thresholds in reciprocal 15N-labeling experiments as a robust tool in detection of candidate proteins responding to biological treatment Sylwia Kierszniowska, Dirk Walther and Waltraud X. Schulze Max Planck Institut für Molekulare Pflanzenphysiologie, Golm, Germany

Metabolic labeling of plant tissues with 15N has become widely used in plant proteomics. Here, we describe a robust experimental design and data analysis workflow implementing two parallel biological replicate experiments with reciprocal labeling and series of 1:1 control mixtures. Thereby, we are able to unambiguously distinguish (i) inherent biological variation between cultures and (ii) specific responses to a biological treatment. The data analysis workflow is based on first determining the variation between cultures based on 15N/14N ratios in independent 1:1 mixtures before biological treatment is applied. In a second step, ratio-dependent SD is used to define p-values for significant deviation of protein ratios in the biological experiment from the distribution of protein ratios in the 1:1 mixture. This approach allows defining those proteins showing significant biological response superimposed on the biological variation before treatment. The proposed workflow was applied to a series of experiments, in which changes in composition of detergent resistant membrane domains was analyzed in response to sucrose resupply after carbon starvation. Especially in experiments involving cell culture treatment (starvation) prior to the actual biological stimulus of interest (resupply), a clear distinction between culture to culture variations and biological response is of utmost importance.

Received: May 26, 2008 Revised: October 9, 2008 Accepted: October 30, 2008

Keywords: Detergent resistant membrane domain / 15N labeling / Plasma membrane proteins / Quantitative proteomics / Stable isotope labeling

1

Introduction

Full metabolic labeling with the heavy 15N isotope has become a widely used tool in quantitative proteomic experiments in plant science. As autotrophic organisms, plants can easily be metabolically labeled by replacing the inorganic nitrogen source containing natural abundance of nitrogen isotopes with a nitrogen source enriched in the 15N isotope. Most commonly used forms of nitrogen labeling involve Correspondence: Dr. Waltraud X. Schulze, Max Planck Institut für Molekulare Pflanzenphysiologie, Am Mühlenberg 1, 14476 Golm, Germany E-mail: [email protected] Fax: 149-3315678403

© 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

K15NO3 and 15NH415NO3 in the plant growth medium. The first biological experiments involving 15N-labeling in plants were carried out on suspension cell cultures [1–3]. More recently, successful labeling of whole plants and seedlings in liquid culture has been achieved and a thorough analysis of full versus partial 15N labeling in Arabidopsis plants has revealed that both methods yield comparable results [4]. A general workflow of the data processing involved in 15Nlabeling experiments was outlined [5] and most common pitfalls and problems in protein identification in labeling experiments due to an increased number of isobaric amino acids in the fully labeled proteome of Arabidopsis have been described [6]. However, even in repeated 15N-labeling experiment using the same labeling scheme, it is not possible to efficiently distinguish between effects of the applied biologi-

www.proteomics-journal.com

1917

Proteomics 2009, 9, 1916–1924

cal treatment and inherent differences between the sets of labeled and unlabeled suspension cell cultures or seedlings. Thus, paired experiments involving reciprocal labeling are an ideal means of filtering candidate proteins responsive to the biological treatment versus such proteins that display differences between control and treatment independently of the biological stimulus applied. In a recently published large scale analysis of phosphorylation in response to elicitor treatments of plant suspension cell cultures, reciprocal labeling has already been employed to confirm candidate proteins responding to the biological treatment in both replicate sets [7]. Reciprocal labeling has also been used in a comparison of the quantitative proteomic methods of 15N metabolic labeling and differential fluorescent gel electrophoresis [8]. However, the latter study did not use abundance ratios from metabolic labeling as the primary criterion for selection of regulated proteins. Adaptation to changes in the nutrient environment in plants has until now been extensively studied by monitoring changes in transcript levels [9–11]. So far, only sucrose induced responses to carbon starvation has been studied on the level of proteins by a global analysis of protein phosphorylation dynamics [12]. Nutrient uptake and adaptation to a changing resource environment is thought to also involve rapid dynamic changes in protein abundances directly at the plasma membrane. Thus, the aim of this study was to use quantitative proteomics to monitor dynamic changes in membrane microdomain composition under changing carbon (sucrose) availability. Detergent resistant domains are considered to have many functions in regulation of plasma membrane composition [13]. Typical functions associated with lipid rafts in yeast and mammalian cells, such as receptor internalization or endocytosis have also been observed in plant cells [14–17]. However, the proteins involved at the molecular and mechanistic level have not been identified. Here, we use a set of experiments with reciprocally labeled treatments in combination with 1:1 mixtures prior to treatment to develop a robust workflow and criteria of how to efficiently filter for candidate proteins responsive to the biological treatment of sucrose starvation and resupply. The workflow especially aims at those experiments in which the biological response is superimposed on a larger degree of inherent biological variation between cell cultures.

2

Materials and methods

2.1 Metabolic labeling of suspension cell cultures and sucrose starvation-resupply Full metabolic 15N-labeling of Arabidopsis thaliana Col-0 suspension cell cultures was carried out as described [1]. Briefly, in the labeled cell cultures the nitrogen source was replaced © 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

with K15NO3 as the only nitrogen source yielding a fully 15Nlabeled proteome within two weeks of growth in the labeling medium. Labeled and unlabeled Arabidopsis cell cultures were starved for sucrose over two days by replacing the sucrosecontaining growth medium with a medium without sucrose. In one of the cell culture sets, sucrose was added back for 5 or 20 min, while the other set of cultures remained untreated. Sucrose was resupplied to a final concentration of 40 mM. 2.2 Cell culture harvesting and sample mixing Cells were harvested by vacuum suction through a metal filter plate and were immediately frozen in liquid nitrogen and ground to fine powder. Material was mixed on basis of equal amounts of powderized fresh weight. At least for plant cell cultures in the biological context described here, the weight of ground cell material was found to be good estimate of protein content. 2.3 Preparation of plasma membrane detergent resistant domain For preparation of detergent resistant membranes, microsomal membranes were isolated from frozen and powderized cell suspension cultures using an extraction buffer (100 mM Hepes-KOH, 10% w/v glycerol, 5 mM EDTA, 0.6% w/v PVP K-25, 5 mM ascorbic acid) containing protease inhibitor cocktail (Sigma Aldrich). Total protein extract was filtered through Miracloth (Calbiochem) and centrifuged at 10 0006g for removal of the cell debris. Microsomes were obtained by ultra-centrifugation at 100 0006g and were resuspended in a buffer containing 5 mM potassium phosphate pH 7.8, 5 mM KCl and 0.1 mM EDTA. Plasma membrane was separated from the microsomes over a two-phase system with PEG/Dextran with 6.4% w/w Dextran, 6.4% w/w PEG and 5 mM KCl [18]. The upper phase was collected and purified once again over the two-phase system. Finally, the upper phase was diluted 3 times and centrifuged at max speed 120 0006g for 1 h. Plasma membranes were resuspended in 50 mM Tris-HCl pH 7.5, 3 mM EDTA and treated with Triton-X100 at a protein to detergent ration 1:13 and a final detergent concentration of 1% for 30 min on ice with continues shaking at 60 rpm. Treated plasma membrane was combined with 2.4 M to the final concentration of 1.8 M sucrose and overlaid by a sucrose step gradient from 1.6 to 0.15 M sucrose and centrifuged at 250 0006g for 18 h. A ring, below 0.15 M sucrose concentration was visible and collected for further analysis as the “detergent resistant membrane fraction”. The collected DRM fraction was diluted in 25 mM Tris-HCl 7.5, 150 mM NaCl, 5 mM EDTA buffer and centrifuged at 200 0006g for 1 h. DRM pellets were resuspended in 8 M urea, 2 M thiourea for in-solution tryptic digest. After reduction in 0.5 mM DTT, and after alkylation of cysteine groups in 2.5 mM iodoacetamide, proteins were digested with LysC for 3 h. www.proteomics-journal.com

1918

S. Kierszniowska et al.

Subsequently, the solution was diluted fourfold with 10 mM Tris-HCl pH 8 before digestion of protein with trypsin (Promega). Digested peptides were desalted over C18 STAGE-tips [19] before mass spectrometric analysis. 2.4 Mass spectrometric analysis and protein identification Tryptic peptide mixtures were analyzed by LC/MS/MS using nanoflow HPLC (Proxeon Biosystems, Denmark) and an Orbitrap hybrid mass spectrometer (LTQ-Orbitrap, Thermo Electron, USA) as mass analyzer. Peptides were eluted from a 75 mm analytical column (Reprosil C18, Dr. Maisch GmbH, Germany) on a linear gradient running from 4 to 64% acetonitrile in 90 min and sprayed directly into the LTQ-Orbitrap mass spectrometer. Proteins were identified by MS/MS by information-dependent acquisition of fragmentation spectra of multiple-charged peptides. Up to five data dependent MS/MS spectra were acquired in the linear IT for each FTMS full scan spectrum acquired at 30 000 FWHM resolution settings with an overall cycle time of approximately one second. Fragment MS/MS spectra from raw files were extracted as DTA-files and then merged to peak lists using default settings of DTASuperCharge version 1.18 (msquant.sourcforge.net) with a tolerance for precursor ion detection of 50 ppm. Fragmentation spectra were searched against a nonredundant Arabidopsis protein database (TAIR8, version 2008-04; 31921 entries; www.arabidopsis.org) using the MASCOT algorithm (version 2.2.0; Matrix Science, UK, www.matrixscience.com). The database contained the full Arabidopsis proteome and commonly observed contaminants (human keratin, trypsin, lysyl endopeptidase), thus no taxonomic restrictions were used during automated database search. The following search parameters were applied: trypsin as cleaving enzyme, peptide mass tolerance 10 ppm, MS/ MS tolerance 0.8 Da, one missed cleavage allowed. Carbamidomethylation of cysteine was set as a fixed modification, and methionine oxidation was chosen as variable modifications. “15N metabolic labeling” was chosen as a quantitative method for MASCOT database searching, allowing identification of labeled and unlabeled peptides within the same database search. Only peptides with a length of more than five amino acids were considered. In general, peptides were accepted without manual interpretation if they displayed a MASCOT score greater than 32 (as defined by MASCOT p,0.01 significance threshold), peptides with a score greater than 24 (as defined by MASCOT p,0.05 significance threshold) were manually inspected requiring a series of three consecutive y or b ions to be accepted. Original MS/MS spectra considered for quantitation in this analysis will be deposited in the Promex spectral library (http://promex.mpimp-golm.mpg.de/home.shtml). Using the above criteria for protein identification, the rate of false peptide sequence assignment as determined by the “decoy database” function implemented in MASCOT v. 2.2.0 © 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Proteomics 2009, 9, 1916–1924

was 3.4% on a 95% confidence level, indicating increased ambiguity in protein identification as has recently been reported [6]. In result tables, peptide assignment to proteins followed the MASCOT default settings, i.e., each redundant peptide was primarily assigned to the highest scoring protein. For proteins identified by ambiguous peptides only, all possible isoforms are listed in the result tables. Isoforms of protein only appear in the tables as separate protein entry if they were assigned at least one unique peptide. In result tables, only peptides with more than one peptide assignment are reported. 2.5 Quantitative protein analysis Ratios between labeled and unlabeled forms of each tryptic peptides were calculated in MSQuant version 1.4.3a29 (released 2007-10-11; msquant.sourceforge.net). Quantitative information was taken from extracted ion chromatograms of labeled and unlabeled form of each identified peptide. Thereby, co-elution of both peptide forms was made a requirement and it was manually inspected in MSQuant that the pairs of labeled and unlabeled forms actually fit with the expected isotope envelope distributions. Peptides that did not meet these criteria were omitted from the analysis. Intensity ratios of labeled 15N-form to unlabeled 14Nform of each identified peptide were averaged across all peptides belonging to the same protein within one experimental set. Peptides conserved in multiple members of a protein family were identified using the “show sub-sets” option in MASCOT, and the respective peptides present in multiple proteins were excluded from quantitative analysis if the redundant peptides displayed ratios significantly different (p,0.05; w2-test) from unique peptides of the same protein. Peptides meeting the criteria for sequence identification, but for which only 14N forms or only 15N forms were quantified, were manually assigned the ratios 0.1 (14N-form only) or 10 (15N-form only). This affected only peptides, for which the pairing labeled or unlabeled peak was at noise level. Since quantitative information was extracted from full scan spectra with very low level of noise as obtained in the Orbitrap mass analyzer, no minimum threshold was set for quantitation [20]. Protein abundance ratios were converted into log2 values and were normalized to the average log2 ratio of all identified proteins within each experimental dataset. Only those proteins were considered for further analysis, for which intensity ratios were obtained in both of the paired reciprocal experimental sets. Ratios of 15N to 14N forms and the respective SD as calculated in MSQuant for each identified peptide and the number of peptides used for quantitation for each protein are presented in Supporting Information Tables 1 and 2. The workflow applied to define significant differences between protein ratios upon biological treatment is described and developed in the Results section. The average relative error of quantitation for all 389 quantified proteins was 9.6%. www.proteomics-journal.com

Proteomics 2009, 9, 1916–1924

2.6 Sucrose uptake analysis and expression profiling of marker genes In order to check the starvation status of labeled and unlabeled cell cultures, the expression of sucrose-starvation induced marker genes was analyzed by quantitative PCR as described [10]. The following marker genes for sucrose starvation were used: At1g08630, At2g18050, At2g18700, At2g19800, and At3g47340. Sucrose uptake into cells was measured by an enzymatic assay involving conversion of sucrose to glucose and then monitoring the reaction of glucose to 6-phosphogluconate under production of NADPH. Changes in NADPH abundance are recorded by UV-absorbance at 340 nm [21].

3

Results

Previous biological experiments involving a reciprocal labeling setup used the log2 values of ratios between treated and untreated samples as a measure to identify responsive and nonresponsive proteins. Ratios of treated to untreated samples were averaged for each protein and any average protein ratio outside a 3 s interval was considered a responsive candidate, where s is the SD of all averaged ratios [7]. The workflow described here modifies this approach by taking into account the ratio-dependent SD, thereby allowing the discovery of responsive protein candidates in which the amplitude of the response is more subtle and which may also

1919 be overlaid by inherent biological variations. In addition, multiple testing corrections are applied assigning each candidate protein a specific p-value. 3.1 Design of sucrose starvation-resupply reciprocal labeling experiment For reciprocal labeling, two sets of labeled and unlabeled suspension cell cultures were created from the same labeled or unlabeled parent culture. In one experimental setup, the labeled cell culture was resupplied with sucrose after starvation, while in the reciprocal experimental setup the unlabeled cell culture was resupplied with sucrose. After harvesting, equal amount (gfw) of labeled and unlabeled cell material was combined for joint protein extract and mass spectrometric analysis (Fig. 1). A 1:1 mixture from labeled and unlabeled parent cell cultures without biological treatment was used as control to define the expected distribution of protein abundance ratios in mixtures not subjected to additional treatment (general control, Fig. 1). Another 1:1 mixture from the parent cell cultures was collected after starvation treatment, but before sucrose resupply (starvation control, Fig. 1). In total, four such reciprocal block experiments were carried out, each of which consisted of two paired biological replicates. Results from all experiments were combined. In reciprocal labeling experiments, three classes of results are anticipated. Most of the proteins not responsive to the biological treatment are expected to be identified in a ratio of 15N-form to 14N-form of 1:1. In these cases, the log2

Figure 1. Workflow of the reciprocal labeling experiments. Two biological replicate experiments are carried out in parallel: in one case, the 14 N cells are subjected to the biological treatment and 15N cells are used as control. In the second case, the 15N cells are used for biological treatment, while the 14N cells remain as untreated control. 1:1 mixtures of equal protein amount from untreated 14N and 15N cells taken before starvation treatment (general control) and after starvation treatment (starvation control) are used to define inherent differences between the cell cultures and the technical variance.

© 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

www.proteomics-journal.com

1920

S. Kierszniowska et al.

values of the ratios in both experimental sets are close to zero. For proteins displaying a specific response to the biological treatment, in one of the experimental sets the ratio of 15 N-form to 14N-form is expected to be high, while in the reciprocal experiment the same protein ratio is expected to be low (Fig S1 of Supporting Information). For those proteins, which are inherently different between the two differently labeled cell cultures, but do not respond to the biological treatment, ratios are expected to be consistently high or low in both of the experimental setups. These proteins become more abundant in those experimental setups, in which unlabeled and labeled cell cultures undergo separate treatments prior to the actual biological experiments (such as starvation treatment, as in this study). Thus, proteins displaying a more or less pronounced difference between cultures, but also being responsive to the biological treatment are more difficult to capture. Therefore, the aim of the data analysis workflow described here is to efficiently differentiate between the different classes of results in order to filter for those protein candidates that show clear responses to a biological treatment despite the confounding influence of “culture effects”. 3.2 Characterization of variation in 1:1 mixtures In a first step, repeated measurements of a 1:1 mixture of 15 N cells and 14N cells were used to assess the variation associated with the mixing procedure and the inherent difference between the labeled and unlabeled cell cultures after nutrient starvation. Proteins with ratios deviating from the 1:1 ratio in one measurement are expected to also have deviating ratios in the independent mixtures if these ratios refer to those proteins, which are inherently abundant at different levels between the two cultures. Thus, if ratios from one 1:1 mixture are plotted against the ratios from the second 1:1 mixture, the proteins displaying inherently different expression levels between the labeled and unlabeled cell culture distribute along a diagonal line, and deviation from this line is due to technical effects (Fig. 2). In total, 272 proteins were quantified in three 1:1 mixtures before starvation (general control; black dots, Fig. 2), and 346 proteins were quantified in at least one of the four 1:1 mixtures after starvation treatment (starvation control; white triangles, Fig. 2). Indeed, protein abundance ratios mainly scatter around the center of the graph in the general control samples, while the number of proteins distributing along the 457 diagonal largely was found to be dependent on additional treatments, such as carbon starvation, applied to both cultures before the start of the biological stimulus (starvation control). The process of starvation seemed to introduce greater culture to culture variation (white triangles, Fig. 2) compared to mixtures analyzed from undisturbed cell cultures grown side by side (black dots, Fig. 2). Log2 values of 15N to 14N ratios in 1:1 mixtures for general control and starvation control show normal distribution as © 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Proteomics 2009, 9, 1916–1924

Figure 2. Repeated measurements of independent 1:1 mixtures before and after starvation of cell cultures, but prior to differential resupply with sucrose. Log2 values for each protein ratio from one measurement are plotted against Log2 values of ratios of the same protein in the other measurement. Black dots refer to protein abundance ratios before starvation, white triangles refer to protein abundance ratios after starvation, red triangles indicate proteins outside the significance threshold (see text for explanation).

indicated by a Gaussian Fit over the histogram of binned values (Fig. 3). In the starvation control, slightly more data points are present with extreme ratios. 3.3 Definition of significance thresholds The SD of log2 values of ratios from four independent mixtures of cell cultures after starvation was used to define the expected SD of ratios in bins of log2 values. Protein ratios were classified into bins of equal width of 0.25 across the range of observed average ratios. Bin size of 0.25 was chosen up to log2 value 2 and was changed to 0.5 for larger log2 values to ensure that each bin contains at least three data points. Within each bin, the average and SD of all log2 values was calculated. We found a linear relationship (r2 = 0.844) between the absolute ratio bin value and measured SD (Fig. 4). Thus, the ratio-dependent SD for each given log2 value in the experimental conditions can be calculated as follows sd ¼ bl
x þ b0

(1)

where sd is the ratio-dependent SD, x is the measured log2 value of 15N to 14N ratio, b0 and b1 are the coefficients derived from the linear regression between experimental SD and absolute ratio bins. For data points in the smallest bin www.proteomics-journal.com

1921

Proteomics 2009, 9, 1916–1924

Figure 4. Average SD log2 values of 15N to 14N ratios from four independent 1:1 mixtures as calculated for equally sized bins of log2 values (width = 0.25) of protein abundance ratios from a reference measurement.

3.4 Biological experiment: Changes in DRM composition in response to sucrose resupply after starvation

Figure 3. Distribution of log2 values of 15N/14N ratios in a 1:1 mixture taken as general control before starvation and as starvation control after starvation treatment. A Gaussian fit indicates normal distribution.

(log2 values between 0 and 0.5) the SD of the bin was kept constant. The local SDs calculated from equation 1 were applied to define a local confidence interval, which allows defining data points significantly inside or outside this interval. The distance of each data point to the 457 diagonal line was used as a comparative measure and was calculated as follows d¼

jx  yj sqrt

(2)

where x and y are the log2 values of 15N to 14N ratios from each of the reciprocal experiments. Statistical significance of differential protein abundance was assessed by calculating pvalues associated with the observed distances to the diagonal compared to the local SD and assuming normal distribution (Fig. 3). Specifically, for each data point the ratio between the “distance” and the local SD was calculated and the p-value was calculated by a 2-tailed t-distribution. Subsequently, a multiple testing correction was applied to the whole data set using the false discovery rate (FDR) method introduced by Benjamini and Hochberg [22]. Reported proteins correspond to a cut-off FDR of 5%. Using the above criteria, the false positive rate of incorrect classification of proteins was of 0.14% (1 out of 272) in the general control and 1.4% (5 out of 348) in starvation control. © 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

The above described classification strategy was applied to experiments in which Arabidopsis cell suspension cultures were subjected to carbon starvation by replacing the growth medium with medium without sucrose. After two days of starvation, one of the cultures was resupplied with sucrose, while the other remained untreated. Starvation status of the cell suspension cultures was confirmed by increased expression of sucrose starvation induced genes (Fig. S2 of Supporting Information). Following the experimental outline described in Fig. 1, two such complete sets of reciprocal experiments were carried out using the same parent cell cultures: in one experimental set, sucrose was resupplied for 5 min, while in the other experimental set sucrose was resupplied for 20 min. Sucrose uptake into cells over time was monitored by enzymatic assay (Fig. S3 of Supporting Information). After 5 min, sucrose uptake was barely detectable, whereas after 20 min, sucrose had already accumulated in the cells to significantly higher levels as in the starved condition. Thus, we aim at monitoring protein rearrangement processes at the plasma membrane as cells undergo transition from a status inactive in sucrose uptake to a state of active sucrose influx. For the analysis of reciprocal experimental data sets, ratios from both experiments were plotted against each other (Fig. 5). Each data point represents a given protein that had been quantified in both of the reciprocal experiments. For each data point p-values are calculated based on the above procedure and all log2-values of protein abundance ratios are subjected to Benjamini–Hochberg false discover rate multiple testing correction [22]. Data points significantly (p,0.05) deviating from the diagonal are considered candidate proteins with changing abundance in www.proteomics-journal.com

1922

S. Kierszniowska et al.

Proteomics 2009, 9, 1916–1924

arrange abundance in DRM. A 14-3-3 protein was also recruited to DRM 20 min after sucrose resupply. The observed dynamics in plasma membrane ATPases and 14-3-3 proteins are supported by a reported sucrose-dependent increase in phosphorylation and 14-3-3 binding for plasma membrane ATPases AHA1 and AHA2 in Arabidopsis seedlings [12]. Strongest dynamic changes occur for proteins involved in vesicle transport, lipid metabolism, and transport (Supporting Information Table 1d; Fig. 6). For example, adaptin (At5g22780) and a sterol glucosyltransferase (At3g07020) was identified with a large increase in abundance in DRM after 20 min of sucrose supply, while a GDSLlipase/hydrolase (At1g71120) and a phosphoglyceride transfer family protein (At1g22530) was strongly enriched in DRM 5 min after sucrose supply and then rapidly was depleted from DRM within 20 min. Our results indicate that DRM are membrane microdomains in which rapid changes of protein composition lead to adaptation of cellular processes, such as sucrose uptake.

4

Figure 5. Results from all replicates of reciprocal biological experiments. Log2 values of 15N to 14N ratios from one experiment are plotted against log2 values of 15N to 14N ratios from the reciprocal experiment. Blue and Red symbols indicated those proteins which show significant reciprocal response, whereas Yellow symbols indicate those proteins which are classified as representing inherent differences between the labeled and the unlabeled cell cultures. Calculated significance thresholds are indicated as dashed lines.

DRM after sucrose resupply (“biologically responsive”), while data points close to the 457 diagonal line represent proteins that differed between 14N culture and 15N culture based on differences in the starvation procedure independently of the sucrose treatment (culture effects, “nonresponsive”). Using this data analysis procedure described above, for a total of 389 proteins quantitative information was available from both reciprocal experiment pairs. Thus about 50% of the 774 identified proteins were subjected to the data analysis. With application of the multiple testing corrections, 48 unique proteins (14%) were found significantly (p,0.05) responsive to sucrose resupply after starvation (Supporting Information Table1d). Most of these proteins (90%) have previously been experimentally localized to plasma membrane or they were predicted as plasma membrane or extracellular proteins according to subcellular localization database SUBA [23]. In detail, an aquaporin water channel and two plasma membrane ATPases, AHA3 and AHA4, were found to re© 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Discussion

In order to characterize the protein dynamics in Arabidopsis plasma membrane DRM fractions under changing sucrose availability we developed a robust data analysis workflow allowing to distinguish between proteins differing due to culture to culture variations and proteins significantly responding to the biological treatment. Herein, reciprocal labeling in combination with the novel approach of using ratio-dependent SD has proven to be a simple and robust method to specifically identify proteins responsive to the treatment of interest. One important observation from the experiments carried out in this study is the fact that culture to culture variation is strongly increased upon starvation treatment compared to completely untreated cell cultures (Fig. 2). However, it is not particularly surprising that two independent cell cultures (the labeled and unlabeled) will produce slightly different responses to a starvation treatment although it is technically applied in the same manner to both sets of cultures. Among the proteins showing large culture to culture variations were mainly found abundant proteins of primary metabolism (e.g., glyceraldehyde phosphate dehydrogenase), ribosomal proteins and several proteins with unknown functions. The culture to culture variations introduced by the starvation treatments were as great as changes introduced upon sucrose resupply with respect to the maximum response amplitude (i.e., “distance”) and with respect to the number of responding proteins (Fig. 5). However, the data analysis procedure described here allows to specifically distinguish both types of responses: applying the data analysis workflow to the starvation control data set (1:1 mixtures after starvation; Fig. 1) resulted in 1.4% of the proteins being considered as candidate proteins, while applying the criteria to the sucrose www.proteomics-journal.com

Proteomics 2009, 9, 1916–1924

1923

Figure 6. Degree of change of proteins identified in DRM fractions 5 and 20 min after sucrose resupply to starved cells. “Degree of Change” describes the distance of each data point to the 457 diagonal line described in Fig. 5. Positive values indicate enrichment in DRM fractions, negative values indicate depletion compared to the starvation control. Error bars indicate SD as calculated from RSD of ratios per protein (Supporting Information Table 1). Asterisks indicate proteins identified as significantly (p,0.05) responsive after application of multiple testing correction {Benjamini, 1995 #2172}.

resupply reciprocal experimental data set (Fig. 2), 14% of all quantified proteins are considered as significant candidates. The combination of the requirement for reciprocal data points with the statistical analysis workflow including multiple testing corrections presented here indeed successfully excludes typical co-purifying proteins such as ribosomal proteins, plastid proteins and high abundant cytosolic proteins from the candidate list within the chosen cut-off error rate of 5%. Other quantitative proteomic approaches such as ITRAQ differ from the reciprocal labeling strategy in that they rely on the same pool of starting material, which is being divided into treatments and controls. Thus, instead of “culture effects” differences between samples may be introduced in differential handling or labeling efficiencies. The central element in our proposed data analysis workflow is the thorough implementation of suitable control experiments and the combined analysis of paired biological replica sets. The controls in our workflow are two specific 1:1 mixtures (general control, starvation control) from which significance thresholds are defined. Thus, although the workflow described here is rather specific for experimental designs relying on two different sets of source material (e.g., as in metabolic labeling), the principle of running two biological replica experiments in parallel for joint data analysis, as well as the implementation of suitable 1:1 control mixtures could also well be applied to multiplexed ITRAQ experiments. © 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Our study is the first global analysis of dynamics in detergent resistant domains in response to an external stimulus in plants, which, in this case, was sucrose supply after sucrose starvation. The results provide an overview of 48 unique proteins that significantly change their abundance in DRM at different time points of sucrose resupply. The experiment was designed to compare dynamic changes of protein abundances in DRMs at an early time point, in which sucrose had not yet accumulated inside the cells (5 min; see Fig. S3 of Supporting Information) with a later time point in which transport machinery has been established and activated in the plasma membrane and uptake of sucrose is visible by an increased sucrose accumulation (20 min). During this transition of cells from one steady state in starvation to a new steady state in sucrose supplied medium, we observe strong changes in DRM composition with respect to vesicle transport and membrane lipid modifications. Consistent with our findings, clathrins and syntaxins have been identified as typical dynamic lipid raft proteins in mammalian and yeast systems [13]. It will be interesting in future experiments to combine the analysis of time-dependent protein modifications [12] with the analysis of DRM dynamics in order to understand the underlying molecular mechanisms during changes in membrane compartment compositions. Future in-depth analysis of the sucrose responsive candidate proteins will be carried out to address these processes. In summary, we believe that the design of reciprocal labeling experiments in proteomic analyses efficiently increases the confidence in candidate identification in rewww.proteomics-journal.com

1924

S. Kierszniowska et al.

sponse to biological treatments. Since in biological experiments a varying degree of inherent variation between sets of plants, seedlings or cell cultures can exist, the reciprocal experiments (i) include a complete biological replicate of the experimental system and (ii) in combination both data sets allow specific definition of responding proteins. By implementing independent 1:1 control mixtures labeled and unlabeled material just before biological treatment, the a priori variation can be assessed specifically for each experiment.

The authors would like to thank Dirk Hincha and Lothar Willmither for stimulating discussions and critical comments. Thanks goes to Witold Szymansky for assistance in cell culture handling and membrane preparations. WS was funded by an Emmy-Noether Fellowship of the German Research Foundation (DFG). The authors have declared no conflict of interest.

5

References

Proteomics 2009, 9, 1916–1924 [9] Morcuende, R., Bari, R., Gibon, Y., Zheng, W. et al., Genomewide reprogramming of metabolism and regulatory networks of Arabidopsis in response to phosphorus. Plant Cell Environ. 2007, 30, 85–112. [10] Osuna, D., Usadel, B., Morcuende, R., Gibon, Y. et al., Temporal responses of transcripts, enzyme activities and metabolites after adding sucrose to carbon-deprived Arabidopsis seedlings. Plant J. 2007, 49, 463–491. [11] Scheible, W. R., Morcuende, R., Czechowski, T., Fritz, C. et al., Genome-wide reprogramming of primary and secondary metabolism, protein synthesis, cellular growth processes, and the regulatory infrastructure of Arabidopsis in response to nitrogen. Plant Physiol. 2004, 136, 2483–2499. [12] Niittylä, T., Fuglsang, A. T., Palmgren, M. G., Frommer, W. B., Schulze, W. X., Temporal analysis of sucrose-induced phosphorylation changes in plasma membrane proteins of Arabidopsis. Mol. Cell. Proteomics 2007, 6, 1711–1726. [13] Simons, K., Toomre, D., Lipid rafts and signal transduction. Nat. Rev. Mol. Cell Biol. 2000, 1, 31–39. [14] Russinova, E., Borst, J. W., Kwaaitaal, M., A., C.-D. et al., Heterodimerization and endocytosis of Arabidopsis brassinosteroid receptors BRI1 and AtSERK3 (BAK1). Plant Cell 2004, 16, 3216–3229. [15] Horn, M. A., Heinstein, P. F., Low, P. S., Receptor-mediated endocytosis in plant cells. Plant Cell 1989, 1, 1003–1009.

[1] Engelsberger, W. R., Erban, A., Kopka, J., Schulze, W. X., Metabolic labeling of plant cell cultures with K15NO3 as a tool for quantitative analysis of proteins and metabolites. Plant Methods 2006, 2, 1–11.

[16] Etxeberria, E., Baroja-Fernandez, E., Munoz, F. J., PozuetaRomero, J., Sucrose-inducible endocytosis as a mechanism for nutrient uptake in heterotrophic plant cells. Plant Cell Physiol. 2005, 46, 474–481.

[2] Lanquar, V., Kuhn, L., Lelièvre, F., Khafif, M. et al., 15N-metabolic labeling for comparative plasma membrane proteomics in Arabidopsis cells. Proteomics 2007, 7, 750–754.

[17] Bhat, R. A., Miklis, M., Schmelzer, E., Schulze-Lefert, P., Panstruga, R., Recruitment and interaction dynamics of plant penetration resistance components in a plasma membrane microdomain. Proc. Natl. Acad. Sci. USA 2005, 102, 3135– 3140.

[3] Harada, K., Fukusaki, E., Bamba, T., Sato, F., Kobayashi, A., In vivo 15N-enrichment of metabolites in suspension cultured cells and its application to metabolomics. Biotechnol. Prog. 2006, 22, 1003–1011. [4] Huttlin, E. L., Hegeman, A. D., Harms, A. C., Sussman, M. R., Comparison of full versus partial metabolic labeling for quantitative proteomic analysis in Arabidopsis thaliana. Mol. Cell. Proteomics 2007, 6, 860–881. [5] Palmblad, M., Bindschedler, L. V., Cramer, R., Quantitative proteomics using uniform (15)N-labeling, MASCOT, and the trans-proteomic pipeline. Proteomics 2007, 7, 3462–3469. [6] Nelson, C. J., Huttlin, E. L., Hegeman, A. D., Harms, A. C., Sussman, M. R., Implications of 15N-metabolic labeling for automated peptide identification in Arabidopsis thaliana. Proteomics 2007, 8, 1279–1292. [7] Benschop, J. J., Mohammed, S., O’Flaherty, M., Heck, A. J. et al., Quantitative phospho-proteomics of early elicitor signalling in Arabidopsis. Mol. Cell. Proteomics 2007, 6, 1705–1713. [8] Hebeler, R., Oeljeklaus, S., Reidegeld, K. A., Eisenacher, M. et al., Study of early leaf senescence in Arabidopsis thaliana by quantitative proteomics using reciprocal 14N/15N labeling and difference gel electrophoresis. Mol. Cell. Proteomics 2008, 7, 108–120.

© 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

[18] Marmagne, A., Salvi, D., Rolland, N., Ephritikhine, G. et al., Purification and fractionation of membranes for proteomic analyses. Methods Mol. Biol. 2006, 323, 403–420. [19] Rappsilber, J., Ishihama, Y., Mann, M., Stop And Go Extraction tips for matrix-assisted laser desorption/ionization, nanoelectrospray, and LC/MS sample pretreatment in proteomics. Anal. Chem. 2003, 75, 663–670. [20] Venable, J. D., Wohlschlegel, J., McClatchy, D. B., Park, S. K., Yates, J. R. I., Relative quantification of stable isotope labeled peptides using a linear ion trap-Orbitrap hybrid mass spectrometer. Anal. Chem. 2007, 79, 3056–3064. [21] Geigenberger, P., Lerchl, J., Stitt, M., Sonnewald, U., Phloem-specific expression of pyrophosphatase inhibits long distance transport of carbohydrates and amino acids in tobacco plants. Plant Cell Environ. 1996, 19, 43–55. [22] Benjamini, Y., Hochberg, Y., Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. Royal Stat. Soc. 1995, 57, 289–300. [23] Heazlewood, J. L., Verboom, R. E., Tonti-Filippini, J., Small, I., Millar, A. H., SUBA: the Arabidopsis Subcellular Database. Nucleic Acids Res. 2007, 35, D213–218.

www.proteomics-journal.com