SILIP: a novel stable isotope labeling method for in planta quantitative proteomic analysis


  • Jennifer E. Schaff,

    1. Department of Plant Pathology, Center for the Biology of Nematode Parasitism, North Carolina State University, Raleigh, NC 27695-7253, USA
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Flaubert Mbeunkui,

    1. Department of Molecular and Structural Biochemistry, North Carolina State University, Raleigh, NC 27695-7622, USA
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Kevin Blackburn,

    1. Department of Molecular and Structural Biochemistry, North Carolina State University, Raleigh, NC 27695-7622, USA
    Search for more papers by this author
  • David McK. Bird,

    1. Department of Plant Pathology, Center for the Biology of Nematode Parasitism, North Carolina State University, Raleigh, NC 27695-7253, USA
    Search for more papers by this author
  • Michael B. Goshe

    Corresponding author
    1. Department of Molecular and Structural Biochemistry, North Carolina State University, Raleigh, NC 27695-7622, USA
    Search for more papers by this author

(fax +1 919 515 2047; e-mail


Due to ease of manipulation, metabolic isotope coding of samples for proteomic analysis is typically performed in cell culture, thus preventing an accuratein vivo quantitative analysis, which is only achievable in intact organisms. To address this issue in plant biology, we developed SILIP (stable isotope labeling in planta) using tomato plants (Solanum lycopersicum cv. Rutgers) as a method that allows soil-grown plants to be efficiently labeled using a 14N/15N isotope coding strategy. After 2 months of growth on 14N- and 15N-enriched nitrogen sources, proteins were extracted from four distinct tomato tissues (roots, stems, leaves and flowers), digested, and analyzed by LC/MS/MS (data-dependent acquisition, DDA) and alternating low- and elevated-energy MS scans (data-independent acquisition, MSE). Using a derived relationship to generate a theoretical standard curve, the measured ratio of the M (monoisotopic) and M-1 isotopologues of 70 identified 15N-labeled peptides from 16 different proteins indicated that 15N incorporation was almost 99%, which is in excellent agreement with the 99.3%15N-enriched nitrate used in the soil-based medium. Values for the various tissues ranged from 98.2 ± 0.3%15N incorporation in leaves to 98.8 ± 0.2% in stems, demonstrating uniform labeling throughout the plant. In addition, SILIP is compatible with root-knot nematode (Meloidogyne spp.) development, and thus provides a new quantitative proteomics tool to study both plant and plant–microorganism systems.


Over recent years, methods for measuring the protein abundance among biological samples using LC/MS or LC/MS/MS have been developed utilizing a variety of analytical approaches (Domon and Aebersold, 2006; Ong and Mann, 2005; Ong et al., 2003a; Ono et al., 2006; Qian et al., 2004; Shen and Smith, 2005; Zimmer et al., 2006). These approaches fall into two major categories according to whether either stable isotope labels or tags are incorporated into the protein or peptide components prior to mass spectrometry (MS) analysis. In the label-free approach, proteins from each treatment group are digested with proteases and analyzed separately, and the measured electrospray intensities for each peptide component are compared with the corresponding peptide intensities from samples in other treatment groups to determine treatment-induced protein abundance changes (Levin et al., 2007; Ono et al., 2006; Tabata et al., 2007). The second approach involves the incorporation of unique stable isotope labels into the peptides from each treatment group, followed by pooling equivalent amounts of sample from each treatment prior to LC/MS or LC/MS/MS analysis. Based on the isotope coding used, each labeled peptide contains a distinctive mass signature that indicates the sample of origin and whose intensity can be used to determine the relative changes in protein abundance between samples during the same analysis (Goshe and Smith, 2003; Gruhler et al., 2005; Hsu et al., 2003; Tao and Aebersold, 2003).

Isotope coding of proteins using stable isotope labels has been accomplished in a variety of ways, including the use of isotope-coded chemically reactive tags such as ICAT (isotope-coded affinity tags) (Gygi et al., 1999) and iTRAQ (isotope tagging for relative and absolute protein quantitation) (Ross et al., 2004) or enzymatic labeling during proteolysis using 18O-enriched water (Liu et al., 2007; Yao et al., 2001). Although each of these approaches has been applied for proteomics analysis in a variety of systems including plants (Dunkley et al., 2004; Jones et al., 2006; Rudella et al., 2006), limitations intrinsic to the labeling chemistry can result in unwanted by-products and incomplete labeling due to the presence of sample contaminants. As labeling is performed after protein isolation, variations during sample handling prior to labeling cannot be evaluated. Furthermore, because ICAT labels and enriches for cysteinyl peptides, only a limited number of peptides may be used for quantification. In the case of iTRAQ, essentially all the peptides generated by proteolysis with trypsin will become labeled, thus allowing multiple peptides to be used to determine a specific protein abundance. However, each peptide requires multiple labeling (one modification at the N-terminus and the C-terminal Lys in addition to any Lys mis-cleavage sites based on the trypsin digestion efficiency), as well as an MS/MS event to reveal the label-specific reporter ions to be used in quantification, both of which contribute to increased sample complexity that may impede MS measurements. The use of 18O-enriched water has the benefit of labeling every peptide at the proteolytic step, but unwanted back exchange results in loss of the label (Staes et al., 2004; Yao et al., 2003), thus requiring additional work-up to ensure sufficient incorporation of two 18O atoms at the scissile carboxylate (Sevinsky et al., 2007).

An alternative isotope coding approach for protein quantification utilizes in vivo metabolic labeling, whereby various nutritional sources enriched in heavy isotopes, such as 15N and 13C, are supplied to cells or organisms during growth (Goshe and Smith, 2003; Gruhler and Matthiesen, 2007; Snijders et al., 2005). Typically, metabolic isotope labeling is performed using stable isotope labeling with amino acids in cell culture (SILAC) (Ong et al., 2002), in which a specific isotope-coded amino acid, such as Leu, Ile, Lys or Arg labeled with 15N and 13C, is added to the growth medium containing an auxotrophic cell line and thus is incorporated into all proteins at the corresponding residue (Everley et al., 2004; Gehrmann et al., 2004; Ong et al., 2003b). Alternatively, the growth medium can be supplemented with isotopically enriched salts, such as 15N-based nitrogen sources (Lafaye et al., 2005; Wu et al., 2004). SILAC has been used to label cultured Arabidopsis thaliana cells (Gruhler et al., 2005; Hsu et al., 2003; Tao and Aebersold, 2003), and methods employing 15N labeling to code Arabidopsis plantlets grown in liquid suspension cultures were recently reported (Engelsberger et al., 2006; Huttlin et al., 2007). The major analytical advantage of these labeling techniques over the chemical-based isotope tagging strategies is that equivalent amounts of the control and experimental sample (based on the number of cells or the mass of plant tissue) can be mixed prior to homogenization, protein extraction and digestion, thus providing an internal control for all the steps of sample preparation and analysis. Although these methods provide the ability to study various aspects of plant biology using isotope-coded mass spectrometry, they are limited to liquid media. In our research on soil-borne plant-associated microorganisms, including rhizobacteria and root-knot nematodes (Lohar and Bird, 2003; Lohar et al., 2004; Weerasinghe et al., 2005), it is absolutely necessary to use soil-grown plants to quantitatively study these symbiotic relationships.

To address the need for a soil-based 15N labeling method for plants, we developed SILIP (stable isotope labeling in planta), which allows whole plants to homogeneously incorporate 15N within their proteome. Tomato plants (Solanum lycopersicum cv. Rutgers) were grown in an optimized soil-based medium, which promoted the normal plant growth phenotype while facilitating uniform labeling across plant tissues according to the amount of 15N present in the supplied nitrogen source. The efficiency of label incorporation achieved by the SILIP method enabled LC/MS/MS identification of both 14N- and 15N-labeled peptides using data-dependent and data-independent acquisition methods. In addition, SILIP of tomato plants was found to be compatible with root-knot nematode (Meloidogyne spp.) development, and thus provides a new quantitative proteomics tool to study plant–microorganism interactions using mass spectrometry analysis.


Growth media and nitrogen source

As we are interested in using SILIP to study plant–nematode interactions, we wished to establish a soil-based medium that would support normal plant growth, provide a suitable matrix texture for efficient nematode inoculation, and provide controlled nutrient delivery. We also wished to maximize incorporation of each isotope into the plant by using compounds with the highest isotopic incorporation available at the lowest possible cost. A variety of soil-like growth media based on permutations of gypsum, river sand and vermiculite containing 15N-enriched nitrates were tested to determine their suitability for normal tomato and nematode growth. There was no statistical difference in the mass and height of plants grown in either of the sand-based media (data not shown), and because river sand was more adaptable for post-growth manipulation, it was used in subsequent experiments involving 15N incorporation.

Potassium nitrate (designated K15NO3) with a level of 15N incorporation of 99.3% as reported by the manufacturer is commercially available and was used as the 15N-enriched (heavy) source; normal (light) KNO3 is typically 99.6%14N. To utilize these salts in fertilizer, it was necessary to alter the composition of the nitrogen sources (in our case, Hoagland’s fertilizer) from the standard formulation. In particular, calcium nitrate and potassium phosphate were replaced with calcium phosphate and potassium nitrate. The amount of potassium nitrate in our modified Hoagland’s fertilizer (Table 1) was adjusted to deliver a level of nitrate identical to the original Hoagland’s fertilizer.

Table 1.   Formulation of nitrate and phosphate salts used in the soil-based medium for stable isotope labeling in planta (SILIP) of tomato plants
Salt compositionaIon concentrations (mm)b
  1. aOnly the components that were modified in the SILIP fertilizer are listed.

  2. bThe concentration of the nitrogen source is equivalent in both fertilizers.

Original Hoagland’s fertilizer
 Potassium nitrate KNO3 (5 mm)5500
 Calcium nitrate Ca(NO3)2 (5 mm)01050
 Potassium phosphate KH2PO4 (1 mm)1001
 Total ion concentration61551
Modified Hoagland’s fertilizer for SILIP
 Potassium nitrate [14N] or  [15N]KNO3 (15 mm)151500
 Calcium phosphate Ca3(PO4)2 (1 mm)0032
 Total ion concentration151532
Ion concentration ratios
 Modified Hoagland’s for  SILIP/Hoagland’s2.510.62

Tomato plant growth in 14N- and 15N-enriched soils

Tomato plants fertilized with modified Hoagland’s fertilizer containing light or heavy nitrogen were grown side by side under optimal light and temperature conditions. Similarly, tomato plants fertilized with modified Hoagland’s fertilizer were compared to plants grown in traditional Hoagland’s fertilizer. The differences in plant color, shape, size and stability were visually assessed. No morphological differences were observed between plants grown on the traditional or modified media, nor were any differences observed after 3 weeks when either of the nitrogen isotopes were used (Figure 1). Growth was extended to 8 weeks, and plants from 14N- and 15N-enriched soils were harvested to evaluate any differences in plant height and mass. As indicated in Figure 2, no statistically significant differences in height (Figure 2a) and mass (Figure 2b) were observed for plants grown in the 14N- and 15N-enriched soil.

Figure 1.

 Tomato plants generated by stable isotope labeling in planta (SILIP).
Tomato plants grown in river sand and supplied with modified Hoagland’s fertilizer containing 14N (light) and 15N (heavy) potassium nitrate as the sole nitrogen source for 3 weeks. Based on visual inspection, no significant difference was observed between plants grown on the two media.

Figure 2.

 Phenotype comparison of SILIP tomato plants.
The 14N- and 15N-labeled plants were compared with respect to (a) height and (b) mass after 8 weeks of growth. Based on measurements of five plants, no statistically significant difference was found between the isotope-coded plants.

Identification of SILIP-coded proteins in tomato tissues

To determine SILIP labeling efficiency, plants were grown for 8 weeks, and tissues from the roots, stems, leaves and flowers were harvested. The proteins from each tissue were isolated, proteolytically digested with trypsin, and the resulting peptides were analyzed by LC/MS/MS. Peptide identification based on retention time, peptide mass and fragment masses generated during collision-induced dissociation (CID) of the charged peptide was performed using data generated from data-dependent or data-independent acquisition.

The first mode of analysis was to assess the ability of data-dependent and data-independent SILIP peptide identification for a 14N/15N-labeled sample containing equivalent amounts of protein from each plant. During LC/MS/MS using data-dependent acquisition (DDA), ions are selected according to intensity and subsequently fragmented using CID. The masses of the peptide and its corresponding fragment ions were searched against a tomato protein database using the Mascot algorithm. An example of the mass spectra acquired using DDA is presented in Figure 3 for the doubly charged peptide ([M + 2H]2+) LTYDEIQSK. Figure 3(a) shows a single MS survey scan at one point during the peptide elution profile. The monoisotopic [M + 2H]2+ions at m/z 548.7779 correspond to the 14N-labeled peptide, while the monoisotopic [M + 2H]2+ions at m/z 554.2640 correspond to the 15N-labeled peptide. The m/z difference of 5.4861 deconvolutes to 10.9722 Da, indicating the incorporation of 11 15N atoms into the peptide during SILIP, and this is in agreement with the number of nitrogen atoms present in the peptide. For the subsequent DDA MS/MS analysis, the monoisotopic [M + 2H]2+ions were selected for CID based on their intensity, and product ion spectra for the 14N- and 15N-labeled peptide are shown in Figure 3(b,c). The fragmentation patterns for both precursor ions are identical, except that the product ions for the 15N-labeled peptide have a greater mass than the corresponding product ions in the 14N-labeled peptide according to the number of nitrogen atoms. For each tissue analyzed by LC/MS/MS using DDA, an average of 100 unique peptide matches to 40 proteins was obtained with an average of 2.5 peptides per protein.

Figure 3.

 Mass spectra of SILIP-labeled peptides acquired using a data-dependent LC/MS/MS analysis.
(a) MS spectrum of the [M + 2H]2+peptide LTYDEIQSK from an LC/MS/MS analysis of a mixture of 14N/15N-labeled peptides obtained during a survey scan.
(b, c) Based on the intensities of the monoisotopic [M + 2H]2+ions, the 14N-labeled peptide (m/z 548.7779) and the 15N-labeled peptide (m/z 554.2640) were sequentially selected for CID and produce the corresponding product ion spectra (b) and (c), respectively. On the basis of the masses of the y-ion series, the peptide can be identified as being 14N- or 15N-labeled.

The SILIP samples were also analyzed by LC/MSE, a relatively new data-independent approach (Plumb et al., 2006; Silva et al., 2005, 2006a,b). LC/MSE is a means of peptide fragmentation analysis whereby the mass spectrometer alternates between MS scans that use either no collision energy to generate intact peptide precursor ions or ‘elevated’ collision energy, which generates peptide fragmentation spectra. Thus, during LC/MSE, all precursors are fragmented data-independently during elution in parallel, as opposed to serially as in the case of DDA, and the data-processing software is used to re-associate product ions with their corresponding precursor ions for database searching. The fundamental principle that makes this mode of acquisition possible is that product ions exactly co-elute with their precursors. For LC/MSE analysis of SILIP peptides, both 14N- and 15N-labeled versions of the peptide co-elute, and therefore their corresponding product ions also co-elute, as shown in Figure 4 for the SILIP-labeled peptide LTYDEIQSK. In Figure 4(a), the ion peak chromatograms for the 14N- and 15N-labeled peptides based on the signal generated by the monoisotopic [M + 2H]2+ions produced in the MS scanning channel are observed to co-elute, and appear as a 14N/15N-labeled pair in the MS scan in Figure 4(b). During the CID event, which occurs between each MS scan acquisition, all eluting ions (including the 14N/15N-labeled peptide pair) are simultaneously fragmented at an elevated energy to produce a complex product ion spectrum, one of which is shown in Figure 4(c). As the 14N/15N-labeled peptide pair is subjected to CID, the composite elevated-energy product ion spectrum contains the fragment ion series corresponding to both the 14N- and 15N-labeled versions of the peptide. For each tissue analyzed, LC/MSE yielded an average of 400 unique peptide matches to 180 proteins, and the results for an analysis of the four tomato tissues (stems, flowers, leaves, roots) are presented in Tables S1–S4 with a total of 600 proteins identified by 2492 peptides to produce an average of 4.2 peptides per protein. Overall, the LC/MSE analysis provides a fourfold increase in identifications compared to LC/MS/MS using DDA with the same LC separation, and this is consistent with other reports in which MSE data provided higher proteome coverage than traditional LC/MS/MS analysis (Plumb et al., 2006; Silva et al., 2006a). This increase is due to the parallel mode of acquisition, as the instrument duty cycle is equally distributed between the MS scan for precursor intensity measurements and the MSE scan for simultaneous fragmentation analysis of all precursors, which generates more peptide data for both quantification (MS) and identification (MSE) than can be obtained by DDA measurements.

Figure 4.

 Mass spectra of SILIP-labeled peptides acquired using a data-independent LC/MSE analysis.
(a) Monoisotopic [M + 2H]2+extracted-ion chromatograms of the 14N-labeled (top) and 15N-labeled (bottom) peptides.
(b) MS spectrum of the [M + 2H]2+peptide LTYDEIQSK from LC/MSE analysis of a mixture of 14N/15N-labeled peptides obtained during the MS scan.
(c) Elevated-energy product ion spectrum (MSE) of all the fragment ions generated by the eluting peptides with the indicated fragment ion series corresponding to both the 14N-labeled (•) and 15N-labeled (bsl00072) versions of the peptide. The y-ion series for both peptides generated by MSE is nearly identical to that produced for each individual peptide during the data-dependent acquisition shown in Figure 3.

Although the SILIP-labeled sample contained a 1:1 stoichiometry of the 14N- and 15N-labeled proteins, the intensity of the 15N-labeled peptide is greater than that of the 14N-labeled peptide. The deviation from the ideal 1:1 ratio is partially attributed to combining the 14N- and 15N-labeled samples at the protein stage. During development, we separately analyzed each 14N- and 15N-labeled sample to verify that peptides could be confidently identified using both DDA and MSE approaches. This was important to make sure that we were not obtaining any artifacts due to mixing. At this stage, the protein samples were mixed in a 1:1 ratio and analyzed as a combined SILIP-labeled sample. The MS spectrum presented in Figure 3(a) originates from the combined sample and represents a single scan that was used to initiate the subsequent DDA CID analysis during peptide elution, while the spectrum in Figure 4(b) (based on the integrated peak areas in Figure 4a) was generated by LC/MSE analysis. However, in both cases, these spectra only reflect the signal of the doubly charged SILIP-labeled peptides. To accurately determine the abundance of each peptide, the intensities for each isotopologue for all the charge states for a given peptide are used to calculate its abundance in order to eliminate any minor differences in retention time and protonation state due to isotope effects. Based on this type of analysis using ProteinLynx Global Server (PLGS) (Waters Corporation,, we have determined that our SILIP method is consistent with other stable isotope coding approaches, with a precision of 10–25% (Goshe and Smith, 2003), and is well within the regime of measuring a biologically significant twofold change in protein abundance.

Determination of 15N incorporation using SILIP

In any metabolic labeling experiment, the extent of heavy atom incorporation is limited by the enrichment composition of the heavy atom source present in the medium used for growth (Huttlin et al., 2007). Although the SILIP approach was effective in isotope coding plants and their proteomes as assessed by LC/MS/MS analysis of peptides according to their monoisotopic precursor and the subsequent fragment ion masses, the extent of isotope incorporation must be determined from the peptide isotopic distribution. This can be accomplished by analyzing the MS data for the heavy isotope-labeled peptides, where isotopologues observed below the m/z of the monoisotopic mass (M) correspond to a peptide population that contains a certain amount of the light isotope. This is illustrated in the mass spectrum for the 15N-labeled peptide LTYDEIQSK in Figure 3(a) (also similar to Figure 4b), where a monoisotopic [M + 2H]+ ion at m/z 554.2640, corresponding to the peptide containing 11 15N-atoms, is observed together with the isotopologues [M-1 + 2H]2+and [M-2 + 2H]2+at m/z 553.7705 and m/z 553.2771, respectively, which correspond to peptides populations with one or two fewer 15N atoms relative to the monoisotopic ion.

The ratio of the intensities of the M-1 and M isotopologues can be used to measure the extent of heavy atom incorporation achieved by SILIP in a manner similar to assessing deuterium incorporation into amino acids and peptides (Goshe and Anderson, 1995, 1999). With a 100%15N-enriched nitrogen source, the maximum level of 15N incorparation for the peptides/proteins would be 100%, but as the level of incorporation of the heavy atom drops below 100%, the abundance of the M isotopologue decreases and that of M-1 increases, as illustrated in Figure 5(a). Although all atoms contribute to the isotopologue intensities within a peptide isotopic cluster, only the nitrogen atoms need to be considered, as the isotope coding used in SILIP is supplied by the nitrogen source (K15NO3). To determine a relationship between the percentage of 15N incorporation and the ratio of the M-1/M intensities, several SILIP-labeled peptides identified during LC/MS/MS analysis were used to generate theoretical intensities for various percentages of 15N incorporation for the peptide isotopologues as shown in Figure 5(a). Based on the ion abundance for all peptide isotopologues, the intensity of the M-1 ion is the most sensitive with regard to the percentage of 14N contained in the 15N-enriched nitrogen source, and the M isotopologue was the most insensitive. Thus a relationship was derived to determine the percentage of 15N incorporation within a peptide using the M-1 and M ion intensities. As shown in Figure 5(b), the percentage of 15N incorporation was found to be proportional to the ratio of the [M-1 + 2H]2+/[M + 2H]2+isotopologue intensities. However, in order to determine the percentage of 15N incorporation using the M-1 and M intensities for any given peptide, the M-1/M isotope ratio needs to be normalized. It was determined that the 15N incorporation was proportional to the M-1/M ratio multiplied by the inverse number of nitrogen atoms contained within the peptide. As shown in Figure 5(c), the normalized M-1/M ratio serves as the input for a second-order polynomial to determine the level of 15N incorporation over a range between 96.5% and 100%. This provides a very convenient method for determining the percentage of atom enrichment for any metabolic isotope labeling strategy using a nitrogen source, and it should be noted that a similar relationship for 13C labeling would be determined using the same approach.

Figure 5.

 Determination of 15N incorporation using SILIP.
Four representative 15N-labeled peptides containing different numbers of nitrogen atoms as identified by LC/MS/MS are used to illustrate the method for determining the percentage of 15N incorporation.
(a) Prediction of the isotopic distribution of the doubly charged peptide LTYDEIQSK using the ICR-2LS software for various 15N incorporation levels. When considering all the isotopologues over a limited range of 15N incorporation percentages, the relative intensity of the M-1 peak increases the most, and the monoisotopic M peak decreases the least.
(b) Percentage of 15N incorporation (from 95 to 100%) versus the ratio of the theoretical abundances of isotopologues M-1 and M for the peptides VPFLFTIK (C50H77N9O10), FGEAVWFK (C50H66N10O11), LTYDEIQSK (C48H77N11O18) and YSLAPLVPR (C48H78N12O12).
(c) Normalized percentage of 15N incorporation versus the normalized ratio of the M-1/M intensities. Normalization of the M-1/M ratio was achieved by multiplying by the inverse number of nitrogen atoms (n) contained in each peptide based on the amino acid composition. Using non-linear regression analysis, these data were best fitted to a quadratic equation, = 239.2 x2 − 99.7 + 100.0, which was used to determine the percentage of 15N incorporation for all the peptides listed in Table 2.

Based on the relationship derived in Figure 5(c), the extent of 15N incorporation was determined for a selected number of proteins identified using multiple unique peptides for each tissue type. A comparison between the experimental and predicted MS spectrum obtained for the doubly charged 15N-labeled peptide VPFLFTIK is shown in Figure 6, in which the experimentally measured isotope distribution and the predicted isotope distribution of the same peptide based on the M-1/M ratio are in agreement (98.4%15N incorporation). Using this enrichment calculation approach, the percentage of 15N incorporation was determined for peptides corresponding to a number of proteins across four tomato tissues (roots, stems, leaves and flowers), and the results are presented in Table 2. The extent of 15N enrichment was found to be consistent across tissue types. For example, Figure 7 presents the mass spectra for the 15N-labeled peptide FGEAVWFK across flower, leaf and stem tissues. For this peptide, the 15N enrichment was found to vary between 98.1% for leaves and 98.9% for stems. In most cases, proteins were detected in only a subset of the four tissues due to tissue-specific expression differences for these proteins in tomato. The overall 15N incorporation in tomato across all tissues using our SILIP approach was determined to be 98.6 ± 0.4%, a value that is in excellent agreement with the maximum achievable level of 99.3% based on the 15N-enriched nitrate source.

Figure 6.

 Comparison of the experimental and theoretical mass spectra of a SILIP-labeled peptide.
The mass spectrum (black) is for the doubly charged 15N-labeled peptide VPFLFTIK identified by LC/MS/MS analysis. Using the quadratic equation established in Figure 5(c), the peptide was determined to have a 15N incorporation level of 98.4%. Using this value, a theoretical mass spectrum was generated using ICR-2LS (red). The two spectra essentially overlap, demonstrating the utility of measuring the M-1/M isotopologue ratio as a means of rapidly determining the percentage of 15N incorporation.

Table 2.   Representative list of SILIP-labeled tomato proteins identified by LC/MSE
Accession numberProtein namePeptide sequenceaNo. nitrogen atomsb15N incorporationc,d (%)
  1. aFor each identified protein, unique peptides were detected as doubly charged precursor ions with less than 10 ppm mass measurement accuracy and displayed a discernable isotopic distribution pattern. Amino acid residues within parentheses are not part of the identified peptide, but are those occurring within the protein sequence and illustrate trypsin substrate cleavage specificity. All proteins included here were among those with a high PLGS score and a false-positive rate of ≤5%. A complete list of SILIP-labeled peptides appears in the Supporting information.

  2. bNumber of nitrogen atoms contained in each peptide based on the amino acid composition.

  3. cThe percentage of 15N incorporation for each peptide was calculated according to the equation established in Figure 5(c). The percentage of 15N incorporation for each protein and tissue was expressed as the average percentage of 15N incorporation ± SD according to the number of identified peptides.

  4. dND, not detected.

  5. eThis peptide contains an extra nitrogen atom due to carboxyamidomethylation of the cysteinyl residue. As this extra nitrogen is not derived from the nitrogen source used in SILIP, it is simply considered an additional mass modification and thus is not used for determination of the percentage of 15N incorporation. It should be noted that inclusion of the nitrogen produces a value within the error associated with the standard curve established in Figure 5(c).

SGN-U312436Chlorophyll a/b binding protein precursor   97.8 ± 0.4 98.6 ± 0.1 98.9 ± 0.1 
SGN-U31253133 kDa oxygen-evolving protein of photosystem II   98.2 ± 0.2 98.5 ± 0.3 98.8 ± 0.1 
SGN-U31257223 kDa subunit of the oxygen-evolving system of photosystem II   98.3 ± 0.1   
SGN-U312871Oxygen-evolving complex precursor   98.3 ± 0.1   
SGN-U312623Cathepsin D inhibitor    98.6 ± 0.1  
SGN-U31338421 kDa seed protein precursor    98.4 ± 0.4  
SGN-U312826Trypsin inhibitor V   98.7 ± 0.0  
SGN-U313362ATP synthase beta chain, mitochondrial precursor     98.8 ± 0.1 
SGN-U314312Glyceraldehyde-3-phosphate dehydrogenase     98.8 ± 0.2 
SGN-U313212Light-harvesting chlorophyll a/b binding protein     99.0 ± 0.1 
SGN-U314673Histone H4   98.4 ± 1.6  
SGN-U315168Xyloglucan endotransglycosylase     98.6 ± 0.2 
SGN-U312368Pathogenesis-related protein 10      98.6 ± 0.2
SGN-U313362ATP synthase beta chain, mitochondrial precursor      98.7 ± 0.1
SGN-U315615Glutathione-S-transferase 1      98.5 ± 0.4
SGN-U312978Nucleoside-diphosphate kinase      98.7 ± 0.21
   Leaves98.2 ± 0.3 (= 15)   
   Flowers 98.5 ± 0.5 (= 17)  
   Stem  98.8 ± 0.2 (= 22) 
   Roots   98.6 ± 0.2 (= 16)
   Tomato (all tissues)98.6 ± 0.4 (n = 70)
Figure 7.

 Mass spectra of a SILIP-labeled peptide identified from several tomato tissue samples.
The doubly charged 15N-labeled peptide FGEAVWFK was detected in proteins isolated from tomato tissues comprising (a) flowers, (b) leaves and (c) stems, with measured percentages of 15N incorporation of 98.6, 98.1 and 98.9%, respectively. The peptide is unique to the light-harvesting chlorophyll a/b binding protein (SGN-U313212) and was not detected in the root tissue sample. These data illustrate the ability of SILIP to uniformly label the entire plant proteome with a soil-based medium that can be used for differential quantitative proteomic analysis between entire plants or their tissue-specific components.


A large body of work has emerged over the past several years in which stable isotopes are metabolically incorporated into a variety of systems using isotopically enriched growth media for quantitative differential proteomics studies (Harada et al., 2006; Krijgsveld et al., 2003; Lafaye et al., 2005; McClatchy et al., 2007; Ross et al., 2004). Most of these approaches have centered on the study of bacterial systems or cultured cells. While these isotope coding methods are useful in some experimental contexts, other more complex experiments seek to measure protein abundance changes under conditions that cannot be replicated in cell culture. A primary research interest in our laboratories centers on understanding the interactions between plants and parasitic root-knot nematodes (Meloidogyne spp.), which can only be studied in planta. Although plants may be grown in liquid media for experimental studies (Engelsberger et al., 2006; Huttlin et al., 2007; Lanquar et al., 2007), these growth conditions are quite different from the environmental conditions typically encountered in a soil-based growth medium. Furthermore, in order to study these plant–microorganism interactions, the host plants need to be grown in a medium with properties similar to soil. In the present study, we describe a method of using a soil-like growth medium that allows efficient and essentially complete labeling of the nitrogen atoms for all proteins throughout the tomato plant using the heavy isotope (15N). This stable isotope labeling in planta (SILIP) is accomplished using a medium consisting of clean river sand and modified Hoagland’s fertilizer that promotes normal tomato plant growth using metabolic incorporation of the 14N/15N isotope-coded label to enable differential quantitative proteomics studies to be carried out between treatment or study groups with whole tomato plants. For our future proteomic studies of nematode–host interactions, a soil-based growth medium is essential to study the changes occurring in both the plant and the microorganism. While the present study involves the use of tomato plants only, SILIP should most certainly be extended to a wide variety of other plant systems for quantitative proteomic studies.

In any proteomic study involving isotope coding to differentiate between two distinct samples, peptide identification and quantification need to be performed and correlated to the level of relative protein abundance. This is readily achieved using our SILIP approach due the high efficiency of isotope coding of proteins across all tissues. Based on our measurements, the overall level of isotope incorporation was determined to be >98.6%, which is essentially the maximum afforded by the 99.3%15N-enriched nitrogen source. Although SILAC has been shown to be compatible with mammalian cell cultures with essentially full incorporation of the isotope-coded amino acid into the cellular proteome, only 70–80% incorporation of the label was achieved in Arabidopsis plant cell cultures due to their autotrophic nature (Gruhler et al., 2005). These lower levels of SILAC labeling efficiency can impair both identification and quantification of peptides due to the lower abundance of the monoisotopic ion for the heavy labeled peptide, which thus escapes data-dependent selection and fragmentation. In our combined 14N/15N-labeled sample (1:1) analyzed by LC/MS/MS using DDA, there were essentially equal numbers of identified 14N- and 15N-labeled peptides (data not shown), indicative of a relative equivalent abundance between the peptide monoisotopic ions that generate equivalent product ion spectra (Figure 3). Although use of 14N/15N labeling has been achieved in Arabidopsis plants in suspension cultures (Huttlin et al., 2007; Nelson et al., 2007), such cultures cannot be used to study plants that require long-term growth in soil or to study plant–microorganism interactions that are vital for plant growth.

The SILIP approach is also amenable to data-independent approaches for quantitative proteomics, as demonstrated by the use of a novel mode of peptide fragmentation and identification, MSE, which was used to characterize the protein composition of various tomato tissues analyzed in this study (Table 2). Unlike classical LC/MS/MS analysis, this new mode of peptide fragmentation acquires qualitative information for all peptides at all points in chromatographic time in parallel, allowing more complete proteome coverage than LC/MS/MS. The efficiency of SILIP enhances LC/MSE analysis, as co-elution of the 14N/15N-labeled peptide pair produces equivalent product ion spectra proportional to precursor ion abundance (Figure 4). In fact, MSE was found to provide an approximately fourfold improvement in proteome coverage than LC/MS/MS with DDA, and will be used in our work to study plant–nematode parasitism.

In this study, SILIP was developed using tomato plants, but the method is amenable to use in other plants by generating an appropriate soil in which the only difference is in the isotope implemented for labeling (Table 1). In developing such a method, it is necessary to assess the efficiency of isotope labeling. This was accomplished by developing a simplified methodology for determining the extent of 15N labeling of peptides. In this approach, a high-resolution (>10 000) mass spectrometer was used to measure the isotopic envelope intensities for 15N-labeled peptides. A relationship was derived between the intensity of the monoisotopic (M) and M-1 isotopologue (Figure 5) which was used to calculate experimentally the level of heavy isotope incorporation. The developed approach is general and can be used to rapidly screen samples for use in quantitative proteomics studies to ensure an appropriate level of label incorporation into the target proteins. For example, based on the nature of the isotope label and the level of enrichment in the isotope source (e.g. 15N-, 13C- or 13C/15N-labeled amino acids), a standard curve can be generated based on the various percentages of isotope enrichment and the ratio of the M-1/M isotopologue intensities. As only the M-1/M ratio is considered in the analysis, the isotope sources need to be enriched to at least 95%; otherwise more error is introduced in determination of the percentage of incorporation and requires the inclusion of the M-2 isotopologue, and perhaps others, into the calculation.

The use of isotope coding in quantitative proteomic methods allows two distinct samples to be compared within the same LC/MS analysis; however, the error associated with these measurements is dependent upon the point at which the isotope labeling is performed. As bottom-up proteomics involves mass spectrometry analysis of peptides obtained from proteolytic digestion of isolated proteins, it would be analytically advantageous to perform the isotope labeling at the earliest point in the sample preparation. In this manner, sample losses encountered during sample work-up will be equally shared between the two samples, thus preserving the initial protein abundance stoichiometry. In regard to minimizing errors, SILIP is ideally suited, as equivalent plant mass obtained from the control (14N-labeled) and experimental (15N-labeled) groups can be combined prior to any sample preparation steps. This is important when trying to isolate plant organelles (Dunkley et al., 2004) or membrane components (Mitra et al., 2007) that require a multitude of steps to be performed in parallel prior to isotope labeling of the proteins or peptides while trying to circumvent the unwanted consequences of downstream chemical isotope labeling that can result from incomplete labeling and the generation of unwanted by-products (Kota and Goshe, 2007).

In addition to an early isotope labeling step, SILIP provides complete proteome labeling throughout the plant (Table 2 and Tables S1–S4). This provides a mechanism for quantification of all detected proteins and any post-translation modifications such as phosphorylation (Goshe, 2006). As every protein is isotope-coded, a variety of fractionation techniques can be employed to increase identification, such as 1D/2D gel electrophoresis, isoelectric focusing, size-exclusion chromatography, or reversed-phase sorbents (Mbeunkui et al., 2006, 2007) to separate the proteins prior to proteolytic digestion. Peptides produced after proteolytic digestion can be further fractioned by strong cation or anion-exchange chromatography prior to LC/MS/MS. Affinity enrichment techniques such as immobilized metal affinity chromatography (IMAC) and the use of titanium oxide (TiO2) supports can be implemented to enrich for phosphopeptides, thus allowing SILIP to be used to quantify changes in phosphoproteomic abundance and elucidate signaling pathways under a variety of conditions, such as plant hormone treatments. Overall, SILIP provides plant biologists with a new tool for performing quantitative proteomic measurements for virtually any soil-based plant and their various tissues and modes of development while providing a method for those studying plant–microorganisms interactions and other agriculturally important crop plants.

Experimental procedures


The nitrogen source containing natural 14N (designated as K14NO3) was purchased from Sigma-Aldrich (, and the 15N-enriched source (99.3%, 15N incorporation designated as K15NO3) was from Spectra Stable Isotopes ( Ammonium bicarbonate was purchased from Fluka ( Sequencing grade-modified trypsin was obtained from Promega ( Acetonitrile (HPLC grade) and formic acid (ACS reagent grade) were from Aldrich ( Water was distilled and purified using a High-Q 103S water purification system ( All other chemicals were obtained from Sigma-Aldrich and used without further purification unless otherwise noted.

Plant materials and growth conditions

Tomato (Solanum lycopersicum cv. Rutgers) seeds were sterilized in bleach for 10 min and then rinsed four times for 5 min each with water (distilled and deionized) before planting. Plants were grown in a greenhouse using standard culture practices. Seedlings were first germinated in river sand and subsequently transplanted into four media: sand, river sand plus vermiculite, gypsum (Oil-Dri Corporation of America, and growth pouches (Mega International, Tomato seeds were watered daily until germination and then fertilized with 0.2x Hoagland’s fertilizer containing the modifications outlined in Table 1. The 15N-labeled medium was prepared using K15NO3 in place of the natural abundance salt (K14NO3). After 2–3 weeks of germination, the amount of fertilizer was increased to 0.4× for the remaining growth period. The height and mass of plants were measured after 2 months of growth in labeled and unlabeled media, and plant tissues were collected and frozen in liquid nitrogen for subsequent protein extraction. Statistical analysis of plant mass and height variation was performed using JMP statistical analysis software (SAS Institute,

Tomato protein extraction

Fresh tissues (2 g) from labeled and unlabeled tomato plants were suspended in a polyethylene scintillation vial containing 1.0 mm zirconia/silica beads (BioSpec Products, in 6 ml methanol:water (60:40). Tomato tissues were homogenized for 2 min using a modified reciprocating saw to agitate the tissues. Equal volumes (2 ml) of homogenate and cold tetrahydrofuran were mixed, vortexed and incubated at −20°C for 10 min to precipitate proteins and remove fat and pigments from the sample. Next, the homogenate was centrifuged at 10 000 g for 3 min and the pellet was suspended in 1 ml resuspension buffer (6 m urea in 50 mm ammonium bicarbonate pH 8.0). A volume of 4 ml tetrahydrofuran was then added and the sample was vortexed to re-precipitate the proteins. After centrifugation, the supernatant was removed and the pellet was washed with 4 ml of cold 70% ethanol and briefly air-dried. Proteins were extracted from the pellet using 2 ml morpholine buffer pH 8.3 (6 m urea, 2% SDS, 5 mm EDTA, 0.1% w/v Brij-56, 0.1% v/v morpholine) followed by acetone precipitation. The protein pellet was finally suspended in the resuspension buffer, and the protein concentration was determined using the BCA protein assay (Pierce,

Protein digestion and preparation of samples

Proteins (100 μg) from 14N-labeled, 15N-labeled and a 1/1 mixture of 14N-labeled/15N-labeled samples from each tomato tissue were treated with 0.1% RapiGest (Waters Corporation) in 50 mm ammonium bicarbonate, pH 7.8, to denature the proteins. Protein reduction was performed using 10 mm Tris(2-carboxyethyl) phosphine (TCEP) with incubation at 55°C for 30 min. Alkylation of cysteinyl residues was performed using 10 mm iodoacetamide in the dark at room temperature for 30 min. Proteolysis was conducted using a 1:50 trypsin-to-protein ratio, and digestion was allowed to proceed overnight at 37°C. Following digestion, the reaction was quenched by lowering the pH by adding HCl to a final concentration of 250 mm. After incubating the sample at 37°C for 2 h, it was centrifuged at 14 000 g for 10 min at 4°C, and the resulting peptides were purified by solid-phase extraction using C18 Extract-Clean columns (Alltech Associates Inc., connected to a PrepSep 12-port vacuum manifold unit (Thermo Fisher Scientific, as previously described (Mitra et al., 2007). The eluates were concentrated to dryness using vacuum centrifugation, and stored at −80°C until LC/MS/MS analysis could be performed.

Liquid chromatography–tandem mass spectrometry analysis

After digestion and desalting, samples containing approximately 3 μg of peptide digest (based on the BCA assay of pre-digested protein) were solubilized in 0.1% formic acid in 5% acetonitrile and analyzed by nanoscale capillary LC/MS using a nanoAcquity ultra-performance liquid chromatograph (UPLC) coupled to a Q-Tof Premier mass spectrometer (Waters Corporation). The nano-LC separation was performed using a C18 reverse-phase column (BEH stationary phase, 1.7 μm particle size) with an internal diameter of 75 μm and length of 250 mm (Waters Corporation), with a binary solvent system comprising 99.9% water and 0.1% formic acid (mobile phase A) and 99.9% acetonitrile and 0.1% formic acid (mobile phase B). Samples were initially pre-concentrated and desalted online at a flow rate of 5 μl/min using a Symmetry C18 trapping column (internal diameter 180 μm, length 20 mm) (Waters Corporation, After each injection, peptides were eluted into the NanoLockSpray ion source (Waters Corporation) at a flow rate of 300 nl/min with the following gradient: 2–7% mobile phase B over 1 min, 7–40% B over 120 min, 40–95% B over 1 min, isocratic at 95% B for 6 min, and a return to 2% B over 1 min. The column was re-equilibrated at initial conditions (2% mobile phase B) for 22 min prior to the next injection. The lockmass calibrant peptide standard, 100 fmol/ml glu-fibrinopeptide B, was infused into the NanoLockSpray ion source at a flow rate of 600 nl/min, and was sampled during the acquisition at 30 sec intervals. The mass spectrometer was operated in V-mode at a resolution of at least 10 000. For data-independent acquisition, full scan (m/z 50–1990) LC/MS data were collected using the ‘expression’ mode of acquisition with a 1 sec scan interval for both normal and elevated-energy data channels (Silva et al., 2005). Data were collected at a constant collision energy setting of 4 V during low-energy MS mode, whereas a step from 15 to 30 V of collision energy was used during the high-energy MSE mode. After the LC/MSE analysis, an equivalent amount of each sample was analyzed by LC/MS/MS using data-dependent acquisition (DDA) with dynamic exclusion to minimize multiple MS/MS events for the same precursor. A maximum of eight precursor ions were interrogated per MS survey scan with an optimal collision energy determined by the software on-the-fly based on precursor m/z and charge state.

Data processing and protein identification

LC/MSE data were database-searched using PLGS2.3/IDENTITYE software (Waters Corporation) with default search parameters specifying a 5% false-positive rate. For LC/MS/MS data, pkl files generated by PLGS2.3 were searched using Mascot 2.2 (Matrix Sciences Ltd, with 50 ppm and 0.05 Da tolerances for precursor and product ions, respectively. All searches were conducting using a tomato protein database consisting of 34 000 sequences (

Calculation of protein 15N incorporation from peptide isotopic distributions

Determination of the level of 15N incorporation for each detected peptide for a given protein involved several processing steps based upon the predicted isotopic distributions as calculated using ICR-2LS software (available at For a given isotopic distribution of a peptide, the measured intensities corresponding to the monoisotopic M and the M-1 ions were used to determine the percentage of 15N incorporation based on a standard curve using simulated 15N incorporation levels. Based on these calculated values, a relationship applicable over the 15N incorporation range of 96.5–100% was established between the background-corrected peak intensities of the isotopologues M-1 and M (of any charge state) and the number of nitrogen atoms (n) contained within the peptide based on the amino acid composition, corresponding to the quadratic equation ax2 + bx c, where y is the percentage of 15N incorporation and x is the value of (M-1)/n × M. Data plotting and regression analysis were performed using Origin 4.1 (


This work was supported by a grant from the United States Department of Agriculture (NRI 2006-35604-16739) to D.B. and M.G., and a Major Research Instrumentation grant from the National Science Foundation (NSF DBI-0619250) to M.G. and D.B. The authors thank the research agencies of North Carolina State University and the North Carolina Agricultural Research Service for continued support of our biological mass spectrometry research.