INTRODUCTION

TOP ABSTRACT INTRODUCTION RESULTS AND DISCUSSION CONCLUSIONS MATERIALS AND METHODS ACKNOWLEGMENTS LITERATURE CITED

Legumes are valuable agricultural and commercial crops that serve as important nutrient sources for both humans and animals. For example, alfalfa (Medicago sativa) is an important forage crop with over 24 million acres planted annually with an annual U.S. value approaching 6 billion dollars (U.S. Department of Agriculture-National Agricultural Statistics Service, 2002). Legumes are characterized by symbiotic relationships with both nitrogen-fixing bacteria and arbuscular mycorrhizal fungi (Barker et al., 1990). These host-symbiont interactions result in the ability to fix atmospheric nitrogen and effect mutualistic and defense-related biosynthetic pathways such as the isoflavones, which have been reported to possess antimicrobial, anticarcinogenic, and other health-promoting properties (Dixon, 1999). Other secondary metabolites in legumes such as the triterpenes have been associated with defense and are of particular interest as novel pharmaceuticals (Small, 1996; Haridas et al., 2001).

The study of legume biology using many of the agriculturally important legumes such as soybean (Glycine max) and alfalfa is complicated by the large genome size and complex ploidy of these species. Fortunately, barrel medic (Medicago truncatula) has a smaller diploid genome that yields more manageable genetics. These traits, along with its autogamous nature, short generation time, and prolific seed production have made barrel medic a useful model legume (Barker et al., 1990; Cook et al., 1997; Cook, 1999; Bell et al., 2000; Trieu et al., 2000).

The impressive achievements in genome and expressed sequence tag (EST) sequencing have yielded a wealth of information for many model organisms, including the plants Arabidopsis and barrel medic. Unfortunately, sequence information alone is insufficient to answer questions concerning gene function, developmental/regulatory biology, and the biochemical kinetics of life. To address these questions, more comprehensive approaches that include quantitative and qualitative analyses of gene expression products are necessary at the transcriptome, proteome, and metabolome levels. Transcriptome approaches using microarray and serial analysis of gene expression technologies are powerful tools; however, mRNA abundances may only represent putative function because there is still a questionable correlation between mRNA and protein levels (Futcher et al., 1999; Gygi et al., 1999). In contrast, proteomics provides a more direct assessment of biochemical processes by monitoring the actual proteins performing the enzymatic, regulatory, and structural functions encoded by the genome and transcriptome. Recent improvements in high-resolution two-dimensional PAGE (2-DE; Klose and Kobalz, 1995; Görg et al., 1999), increased content of protein and nucleotide databases, and increased capabilities for protein identification utilizing modern mass spectrometry methods such as matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOFMS; Pappin et al., 1993; Yates, 1998a, 1998b; Corthals et al., 2000) have made the large-scale profiling and identification of proteins a dynamic new area of research in plant biology.

Although there is a substantial amount of work in the literature on bacterial (Guerreiro et al., 1999; Morris and Djordevic, 2001), yeast (Futcher et al., 1999), and human proteomes (Anderson et al., 2001; Stensballe and Jensen, 2001), there is relatively less information on plant proteomes (van Wijk, 2001). Costa and coworkers have identified proteins from xylem and needles of maritime pine (Pinus pinaster; Costa et al., 1998, 1999), and Tsugita and coworkers have worked on the rice (Oryza sativa) proteome with some success (Tsugita et al., 1994). Both of these groups have relied heavily on Edman sequencing, which suffers due to the inability to sequence proteins blocked at the N terminus. More recently, researchers have reported on subcellular proteomes such as the chloroplast membrane (Peltier et al., 2000, 2002) whereas others have focused on single tissues including Arabidopsis seeds (Gallardo et al., 2001), Arabidopsis mitochondria (Kruft et al., 2001; Millar et al., 2001), maize (Zea mays) root tips (Chang et al., 2000), and barrel medic roots (Mathesius et al., 2001, 2002). To date, there has been no large-scale project to identify proteins from multiple tissues of the same plant species.

The objective of the present work was to survey the organ-/tissue-specific proteomes of the model legume barrel medic, to provide an overview of the barrel medic proteome, and to serve as a basis for future proteome comparisons of genetic mutants, biotically, abiotically, and/or environmentally challenged plants. The survey was accomplished using 2-DE to produce reference maps of protein extracts from leaves, stems, roots, flowers, seed pods, and cell suspension cultures. MALDI-TOFMS peptide mass fingerprinting was used to identify 304 proteins. HPLC coupled with quadrupole time-of-flight tandem mass spectrometry (LC/MS/MS) was used to validate marginal MALDI-TOFMS protein identifications. The identified proteins are discussed and classified based on putative functions determined through similarity (Bevan et al., 1998). Database search results are quantified and strategies discussed. The expression levels quantified by 2-DE are compared with mRNA levels quantified by EST counting.

	RESULTS AND DISCUSSION

TOP ABSTRACT INTRODUCTION RESULTS AND DISCUSSION CONCLUSIONS MATERIALS AND METHODS ACKNOWLEGMENTS LITERATURE CITED

2-DE Reference Maps and Protein Identifications of Barrel Medic Tissues

2-DE reference maps were obtained for barrel medic leaves, stems, roots, flowers, seed pods, and cell suspension cultures and are provided in Figure 1. To qualitatively survey the proteins visualized by 2-DE, a total of 551 proteins (i.e. approximately 96 arbitrary protein spots per gel including positive molecular mass marker controls and negative gel blank controls) were excised from each of the organ-/tissue-specific Coomassie-stained 2-DE gels and analyzed by mass spectrometry. Typically, high-quality MALDI-TOFMS peptide mass maps were obtained, and representative spectra are provided in Figure 2. Of the 551 protein spots processed, 304 proteins were successfully identified and are listed in Table I.

View larger version (107K): [in this window] [in a new window]   Figure 1.   2-DE proteome reference maps were obtained for A, leaf; B, stem; C, root; D, flowers; E, seed pods; and F, cell suspension cultures. Proteins that were identified in this study are marked with arrows and numbers. The numbers correlate with protein identifications listed in Table I. 2-DE was performed using 0.75 to1.0 mg of protein, linear 11-cm IPG strips (pH 3-10), and a 12% (w/v) total acrylamide SDS second dimension. Gels were stained overnight with Coomassie Brilliant Blue R-250, destained the next day, and images recorded.

View larger version (36K): [in this window] [in a new window]   Figure 2.   Representative peptide mass maps obtained using MALDI-TOFMS illustrating good data quality but differences in protein identification success dependent upon the database queried. Mass spectral peaks are labeled with monoisotopic mass-to-charge ratio (m/z) values used for database searching. A, Stromal 70-kD heat shock-related protein (HSP70, accession no. Q02028) was successfully identified in seed pods (pds#7) using the NCBI databases. B, Isoflavone reductase (accession no. BE325778) from seed pods (pds#39) was identifiable only through use of the EST databases.

Supplemental Table I (see www.plantphysiol.org) contains extensive data that document the analytical rigor of the protein identifications. These data include an assigned protein spot number (see Fig. 1), an arbitrary peptide mass fingerprint data quality (PMFQ) score of 1 to 5 (with 5 being best, see "Materials and Methods") to allow assessment of data quality, the number of peptides matched, m/z accuracy and SD of peptides matched, percent protein coverage, theoretical molecular mass and pI, experimental molecular mass and pI, the database accession number of the best match and the databases that yielded concurrent identifications, LC/MS/MS data for select proteins, and the organism to which the matching protein was identified through similarity. For protein identifications determined using the SwissProt and National Center for Biotechnology Information (NCBI) databases, the organism reported in supplemental Table I is that from which the protein or gene was directly sequenced. In the case of most ESTs, protein identifications were first made to barrel medic ESTs that were not annotated. These ESTs were annotated by comparison with The Institute for Genomic Research (TIGR) gene indices or through similarity to other organisms via BLAST. The organism yielding the highest similarity score is the organism reported for EST database identifications in Supplemental Table I. Protein function is also classified and recorded in Supplemental Table I. A minimum of four peptides is statistically necessary to qualify as a confident match (Pappin et al., 1993). Use of additional criteria such as those listed above are advised and increase the confidence in the protein identification. Most proteins identified in Table I had high confidence identifications; however, a small number (23) of the original proteins were identified using only four peptides that had poor m/z accuracies (i.e. above 30 ppm). These protein identifications were considered marginal and were further interrogated using LC/MS/MS. LC/MS/MS data were queried against the same three databases (NCBI, SwissProt, and dbESTothers) used to query MALDI-TOFMS data. The majority of identifications were found to be valid, but four MALDI-TOFMS proteins were revealed as misidentified. The correct LC/MS/MS identifications for these four are reported in Table I. Tandem data was also used to confirm a specific MALDI-TOFMS identified protein questioned by a reviewer in leaves (spot no. 51) that had a minimal four matching peptides and low sequence coverage. This identification was confirmed using LC/MS/MS. These results are provided in Figure 3 and include a search score from dbESTothers (12 peptides matched and Mascot score of 513), representative TOFMS data, and tandem TOF/MS/MS data. Nine proteins from the original list of 23 marginal identifications could not be validated by LC/MS/MS due to limited sample, therefore, were omitted from Table I.

View larger version (31K): [in this window] [in a new window]   Figure 3.   Representative LC/MS/MS data obtained on an ABI Qstar Pulsar for leaves (spot no. 51) confirming the identification of this protein as a TPR repeat protein (accession no. AW694998) as suggested by MALDI-TOFMS peptide mass fingerprinting. The data include: A, database search score and peptides successfully identified; B, example TOF/MS; and C, tandem TOF/MS/MS mass spectra for the peptide observed at m/z 677.62.

Database Query Strategies and Success Rates

In an attempt to maximize our protein identification success rate for barrel medic proteins, we have used protein (SwissProt), nucleotide (NCBI), and EST databases (dbESTothers, and barrel medic-only ESTs from NCBI) for queries of experimental peptide mass maps (Mann and Wilm, 1994; Pappin et al., 1993; Yates, 1998aa, 1998b; Choudhary et al., 2001). The specific databases used to successfully identify each individual protein are reported in Table I, and a summary of the protein identification success rates is provided in Table II. In most cases, the resulting peptide mass maps were of high quality; however, this did not always translate to successful protein identification.

The average protein identification success rate for all tissues using only the protein databases (SwissProt and NCBInr) was 25%, whereas the average protein identification success rate for all tissues using the EST database was 46% (see Table II). Interestingly, the average overlap in the number of proteins identified in both databases was only 15%; thus, searching both databases was complementary and not necessarily redundant. For example, the peptide maps provided in Figure 2 are of similar high quality; however, spectra 2b could not be identified successfully in the SwissProt or NCBI databases and could only be identified successfully through EST database queries. This complementary searching strategy yielded a final protein identification success rate of 55% for our representative protein set.

Strategies using multiple database queries have enhanced our ability to identify proteins even in the absence of a genomic sequence. Our overall success rate of 55% is good when compared with other reports focused on organisms without sequenced genomes. For example, a recent publication concerning pea (Pisum sativum) chloroplast proteins reported a success rate of 15% using mass spectrometry and Edman sequencing (Peltier et al., 2000), whereas a barrel medic root proteome article reported a success rate of 37% (Mathesius et al., 2001). Our protein identification success rates are approaching those for organisms with sequenced genomes. For example, identification success rates of 54% using MS only (Kruft et al., 2001) and 69% (Millar et al., 2001) using MS, immunoblotting, and Edman sequencing were reported for Arabidopsis mitochondrial proteomes. Further, protein identification success rates in human proteome projects are approximately 60% (Stensballe and Jensen, 2001). We expect protein identification success rates to continually increase as the population of unique ESTs continues to increase, as full-length EST sequences are generated, and as genomic sequence of barrel medic becomes available (Comment, 2002).

The average length of barrel medic ESTs used to successfully identify proteins in all organ/tissues was 597 ± 177 nucleotides (or 199 ± 59 amino acids). For proteins in the 30-kD range or less, this represents complete or almost complete sequence coverage by the EST; thus, our confidence in these identifications is very high. For larger proteins this only represents partial protein sequence; however, our data demonstrate that the current EST information is sufficient to allow confident identifications. Additional experimental data such as number of peptides matched, m/z accuracy, molecular mass, and pI provide additional confirmation of identification. It is logical that a strategy including both protein and nucleotide databases would yield greater protein identification rates as some mRNAs, such as mitochondrial and chloroplast-encoded mRNAs (i.e. Rubisco large subunit), do not contain poly(A⁺) tails (Sugiura and Takeda, 2000). These poly(A⁺) tails are used in the initial stages of affinity purification of mRNAs in the cDNA/EST library generation process (Sambrook et al., 1989). Messenger RNAs without poly(A⁺) tails pass through the affinity purification process and are unlikely to be sequenced. These proteins are poorly represented in the EST libraries but are present in many of the protein databases. Therefore, querying both provides greater identification success rates.

Protein Identifications and Functional Classifications

Putative protein functional classifications were assigned based on similarity to better understand the biological processes encompassed by the proteins identified using a 2-DE proteomics approach. Summaries of protein functions observed in the barrel medic proteome are provided in Figure 4. Protein functions were assigned using the protein function database Pfam (http://www.sanger.ac.uk/Software/Pfam/; Bateman et al., 2002) or Inter-Pro (http://www.ebi.ac.uk/interpro/; Apweiler et al., 2001). Protein function was categorized into 13 classes as previously described for Arabidopsis (Bevan et al., 1998). The "unclear" protein class included proteins that were successfully matched to putative proteins from such sources as the Arabidopsis genomic sequence but do not yet have a known function. Most proteins could be unambiguously classified; however, a small number of proteins were associated with multiple functions. Classifications for these proteins were based on their predominate function. Discussions concerning a portion of the proteins observed and their functional role are presented below in relation to the tissue in which they were observed.

View larger version (50K): [in this window] [in a new window]   Figure 4.   Summary of the distribution of tissue specific identified protein classes as determined using the protein function database Pfam (http://www.sanger.ac.uk/Software/Pfam/) and classification schema previously reported for Arabidopsis (Bevan et al., 1998).

Leaves

Photosynthetic enzymes dominated the 2-DE profiles of leaf tissue. Approximately 40% of the leaf protein mass visualized with Coomassie staining can be attributed to a small number of enzymes including the large subunit of Rubisco (26.1%), Rubisco small subunit (2.8%), Rubisco activase (3.2%), and oxygen-evolving protein (6.4%). Most of these proteins appear as multiple spots, and the reported percentages are estimates including all identified spots. The relatively high concentrations of the abundant photosynthetic enzymes demonstrate the importance of these enzymes; however, the prominence of these proteins, specifically Rubisco, in specific regions of the gel, generally contributes to lower quality 2-DE gels and prevents the observation of moderate or lower abundance proteins due to their relatively lower concentrations and the limited dynamic range of common 2-DE staining techniques including Coomassie. Other proteins involved in photosynthesis and carbon fixation were observed in leaf, including: PS1 iron-sulfur protein, ATP synthase, glyceraldehyde 3-phosphate dehydrogenase, malate dehydrogenase, triose phosphate isomerase, tartrate dehydrogenase, and Fru biphosphate aldolase. Many of these photosynthetic enzymes were also observed at lower levels in other green tissues such as stems and immature seed pods.

Several signal transduction proteins were observed in leaves, including the multiple domain protein remorin. Remorin binds simple and complex galacturonide and its C-terminal region has functional similarities to viral intercellular communication proteins (Reymond et al., 1996). Other proteins involved in protein destination or transport included chaperonin 21 precursor, an ankryin repeat protein, and an ATP-binding cassette transporter. Ankyrin repeat proteins have been associated with protein-protein interaction (Gorina and Pavletich, 1996), transcriptional regulation (Batchelor et al., 1998), and transcription inhibition (Jacobs and Harrison, 1998). ATP-binding cassette transporters are membrane-localized proteins that transport small hydrophilic molecules across membranes and include an ATP-binding domain (Higgins, 1992; Jasiñski et al., 2001). Interestingly, other membrane localized proteins were identified and included a chloroplast membrane-associated 30-kD protein (Li et al., 1994) and ATP synthase. The identifications of membrane proteins are important because these proteins are generally underrepresented in 2-DE proteomic studies due to low solubility (Molloy et al., 1998). The observation of plant proteins in 2-DE relative to their general average hydropathicity score has been discussed recently (Millar et al., 2001). Additional proteins identified in leaf tissues included: two cell division proteins, filamentous temperature sensitive protein K homolog cell division protein, miotic cyclin B1-1, DNA mismatch repair protein, RNA-binding protein, transcription factor, and a Gly-rich cell wall structural protein.

Stems

The 2-DE reference map of barrel medic stem proteins was of better quality than that of leaves, primarily due to a lower abundance of Rubisco. Many of the same photosynthetic and carbon metabolism enzymes reported above for leaf were also identified in stems. In addition, several members of the ATP complex associated with energy metabolism were observed. Proteins involved in protein destination and storage were also identified and included the 26S proteasome AAA-ATPase subunit and a 20S proteasome subunit alpha type 7 protein. The 26S proteasome is responsible for protein degradation of endogenous proteins.

Proteins involved in secondary metabolism are of specific interest to our functional genomics project focused on natural products (National Science Foundation Plant Genome Research Project no. 0109732). Several secondary metabolic enzymes were identified in stems and included cinnamoyl-CoA reductase, which plays a role in lignin biosynthesis, and isoflavone reductase-like oxidoreductase, an enzyme involved in phytoalexin production. Stems also revealed several kinases including adenosine kinase, fructokinase, Rib-phosphate pyrophosphokinase, uridylate monophosphate kinase, and nucleoside diphosphate kinase1. A number of RNA binding proteins thought to be important in transcription were also observed. Multiple ribosomal proteins including 40S and 60S ribosomal proteins were identified and function in protein synthesis.

Roots

The roots of legumes are of special interest because of their role in the characteristic symbiotic relationships formed with microorganisms. Although recent articles have been published on the proteomes of barrel medic nodulated root (Bestel-Corre et al., 2002) and uninoculated root (Mathesius et al., 2001), we have included roots as part of our survey for completeness and comparison. Approximately 24% of root proteins identified in this report were associated with plant disease/defense and included peroxidases, superoxide dismutases, ripening related protein, abscisic acid (ABA)-responsive protein, and chitinase. Peroxidases are generally involved in hydrogen peroxide detoxification and are induced by bacterial infection (Cook et al., 1995; Peng et al., 1996). Peroxidases also play a major role in lignin biosynthesis (Lewis and Yamamoto, 1990; Davin and Lewis, 1992). Several glucanases were also identified. These normally constitutively expressed proteins are induced in response to fungal and viral elicitation (Meins et al., 1992). Proteins involved in secondary metabolism of the flavonoid/isoflavonoid pathway made up another 8% of the identified root proteins. Similar to leaves, several membrane-localized proteins such as ATPase and cytochrome C oxidase were also observed in roots.

Relative to other tissues, a larger percentage (i.e. 15%) of the barrel medic root proteins were identified as putative proteins or unannotated proteins. These proteins could be confidently linked to specific ESTs or predicted open reading frames whose functions are still unknown. The observation of unannotated proteins provides experimental evidence of putative/predicted proteins that offer exceptional opportunities in gene annotation (Mann and Pandey, 2001). Because roots appear to have the largest percentage of proteins of unknown function, it is possible that many of these proteins may be specific to legumes and may be involved in microbial interactions characteristic of legumes.

The root 2-DE reference map and protein identifications reported here are consistent with the previous studies by Mathesius et al. (2001) in young, uninoculated barrel medic roots, and by Bestel-Corre et al. (2002) using roots inoculated with Glomus mosseae or Sinorhizobium meliloti. Similar to our results listed above, Mathesius and coworkers reported 5% of their identified root proteins to be associated with flavonoid metabolism and 18% with defense and stress response, yielding a total of 23% defense-related proteins. Further, the total overlap in identified root proteins between the current study and the detailed report by Mathesius and coworkers was over 50%. These included heat shock 70 protein, protein disulfide isomerase, glyceraldehyde-3-phosphate dehydrogenase, isoflavone reductase and chalcone isomerase, a glucosidase and a Cys proteinase, ascorbate peroxidase, alpha-fucosidase, and a ripening-related protein. Many of these proteins had very similar molecular mass and pI values in both studies. For example, cytochrome c oxidase was reported to have a gel molecular mass/pI of 37 kD/4.2 by Mathesius and coworkers, whereas it was observed at a molecular mass/pI of 36 kD/4.9 in the present study. Similarly, ripening related protein had an experimental molecular mass/pI of 16 kD/5.8 in this study and 18 kD/5.5 or 17 kD/6.2 (isoforms) in the Mathesius et al. work. Interestingly, some proteins demonstrated varied slightly between studies. For example, VcCyp was observed at a gel molecular mass/pI of 22 kD/6.3 in the current study as opposed to molecular mass/pI 20 kD/8.8 in Mathesius and coworkers. These slight inconsistencies may represent real differences in posttranslational modifications of the proteins or may be the result of experimental variability.

Proteins identified in all three investigations include a peroxidase precursor, cytochrome c oxidase subunit 6, VcCyp (cyclophilin), a superoxide dismutase, and ABA-responsive protein. Only the ABA-responsive protein and VcCyp were reported to be constitutively expressed by Bestel-Corre et al. (2002), whereas the others proteins common to all three investigations were identified by them as symbiosis-related proteins. Bestel-Corre also identified and reported profucosidase as a symbiosis-related protein. This protein was identified in the current report using uninoculated roots.

Interestingly, two proteins identified in this investigation were not found in either of the other two studies. Acidic glucanase was observed as a relatively abundant protein in the present report (rts#39), but due to its pI of 8.4 and the fact that Mathesius and coworkers' first dimension immobilized pH gradient (IPG) pH range was 4 to 7, it was not present on their gels. We also identified three isoforms of chitinase, all with a pI above 7, that are missing in the Mathesius et al. work. Bestel-Corre et al. (2002) used a pH of 3 to 10 first dimension IPG; thus, these proteins should be visible in their gels. Unfortunately, the total number of identified proteins in the Bestel-Corre report was limited, and these proteins were not identified by them.

Overall, these three reports (this report; Mathesius et al., 2001; Bestel-Corre et al., 2002) provide a wealth of information on the barrel medic root proteome. There are significant similarities between the reference maps that serve as landmarks and can be used for navigation through the root proteome. For example, ABA-responsive protein is one of the most abundant root proteins in each of these investigations. Its relative position can be used to locate PR10 (a highly abundant low-molecular mass protein reported by Mathesius and coworkers next to ABA-responsive protein, rts#80, that was not identified in the present study) in the present and other studies based on similarity. Unfortunately, absolute comparisons of the proteome reference maps are not always straightforward as demonstrated by the differences in molecular mass and pI values shown above for VcCyp.

Flowers

The proteome of flowers contained proteins from almost every functional category. The major portion (38%) of the identified proteins was associated with energy production including glycolysis, pyruvate metabolism, and the tricarbonylic acid (TCA) cycle. Another 21% of the identified proteins were involved with protein synthesis or protein destination. For example, peptidyl prolyl isomerase accelerates protein folding by catalyzing cis-trans isomerization in oligopeptides. Several proteins identified were related to disease/defense or involved in secondary metabolism, such as chalcone isomerase. These enzymes are commonly associated with flower pigmentation or UV protection and serve as important defense proteins in developing seeds. One of the proteins identified specifically in the flower proteome was profilin. Profilin normally binds to monomeric actin to prevent polymerization, although under certain conditions it can promote the polymerization of actin. It occurs in all organs, but is most abundant in mature pollen, making it more likely to be identified in flowers. Many proteins associated with oxidative responses were also identified in flowers. Low levels of a few photosynthetic enzymes were observed due to collection of green sepals with the flowers.

Seed Pods

The intact seed pod proteome was generated from tissue containing both seed and pod tissue. The proteins visualized and identified in the barrel medic seed pod proteome consisted primarily of globulins or seed storage proteins that serve as a nitrogen/nutritional source for developing plants. Several members of the superfamily of "cupins" were identified in barrel medic seed and included 7S and 11S globulins (Dunwell, 1998). The 11S globulins are non-glycosylated proteins and include glycinin and legumin (Hayashi et al., 1988; Duranti et al., 1995). The 7S proteins are a series of similar but progressively larger variations of the same subunit and include vicilin, convicilin, and legumin. It is also interesting to note that 85% of the proteins in this group have been matched to other legumes, suggesting a high level of sequence similarity in legume storage proteins. All of the barrel medic seed storage proteins were observed at multiple molecular masses and pIs. These may represent various stages of protein synthesis and degradation, posttranslational processing not observable at the genome or transcriptome level, or may be the products of multigene families. Similar variations in observed isoforms have been reported for Arabidopsis 12S seed storage proteins in mature and developing seeds (Gallardo et al., 2001).

A significant number of disease-/defense-related proteins were observed in seed pods including peroxidases, osmotin, and ABA-responsive protein. These proteins help defend the plant in early stages of development. Other proteins associated with carbon metabolism, nutrient acquisition, and protein syntheses were also observed. These proteins supply necessary nutrients to the developing plant. Several photosynthetic proteins were observed and are attributed to the collection of immature green seed pods.

Cell Suspension Cultures

Cell suspension cultures were initiated from barrel medic root calli (Dixon, 1980) and their proteome surveyed. Cell culture proteins were extracted with a Tris buffer and, thus, consisted primarily of cytosolic proteins. Most of the identified proteins from cell cultures could be classified in four categories: energy (24%), protein destination and storage (24%), metabolism (22%), and disease/defense (18%). The defense proteins were primarily composed of pathogenesis-related proteins. The most abundant proteins identified were an ABA-responsive protein and a class 10 PR protein. Other disease/defense proteins identified included selenium-binding protein, catalase, and peroxiredoxin. Several of the metabolic enzymes identified in cells were not identified in any other tissue. One of these, 12-oxophytodienoate reductase, is associated with the conversion of 12-oxophytodienoic acid to jasmonic acid.

In some instances, more than one protein was identified with high confidence in each protein spot. For example, spot cls#82 contained peptides that could be associated with both ABA-responsive protein and leghemoglobin. Interestingly, leghemoglobin was identified as a root nodule-specific isoform (Gallusci et al., 1991). This protein is root specific and is induced during nodulation; however, it is generally not observed at appreciable levels in uninoculated roots. Thus, the observation of leghemoglobin is unique here, and this protein may be induced by the cell culturing process. Further, it may also suggest a "memory" effect or root-specific expression pattern observed in the cell cultures that were originally generated from root material (Dixon, 1980). Although many flavonoid-related proteins were observed in other tissues such as root and stem, none were identified in the limited set of proteins surveyed in unchallenged cell cultures.

The proteome of suspension cell cultures is of special interest because the tissue is relatively homogeneous and, therefore, provides a good model tissue system for experiments directed toward integrated functional genomic studies of natural products (https://www.fastlane.nsf.gov/servlet/showaward?award=0109732). Future work will focus on generation of an extensive 2-DE proteome reference map of suspension cell cultures and the changes in the proteome after biotic and abiotic elicitation.

Tissue-/Organ-Specific Expression of Proteins

Many of the proteins identified were redundant as an average of 61% were identified in one or more tissues of barrel medic. The remaining 39% were identified in only one tissue and have the potential of being uniquely expressed in specific tissues/organs based on our limited dataset. The quantities of redundant and potentially unique proteins identified in each specific tissue are summarized in Figure 5. Many of the putative unique proteins are related to the primary function of the specific tissue. For example, photosynthetic enzymes such as PSI iron-sulfur protein and plastid specific ribosomal proteins were only identified in leaves. Other proteins identified only in a specific tissue include the seed storage proteins glycinin, convicilin, and legumin in seed pods. Profilin, a known pollen allergen, was also identified in flowers. These are limited examples illustrating the unique nature of the proteome, but we are hopeful that continued evaluation of the tissue- and organelle-specific proteomes of barrel medic will yield further insight into the specialized functionality of these tissues.

View larger version (11K): [in this window] [in a new window]   Figure 5.   Bar graph summarizing the number of redundant proteins identified in more than one tissue (A) and the number of putative tissue-specific proteins identified in a single tissue only (B). The graph is segregated by tissue. A total of 61% of the proteins were found to be redundant and 39% were found to be putatively tissue specific. Guarantee of specificity at this stage is difficult due to the limited size of the reported protein dataset relative to the total proteome.

Comparison of Barrel Medic Proteome and Transcriptome

A better understanding of the relationship between mRNA and protein abundances is needed to elucidate the processes and regulation of transcription and translation. Several recent publications present conflicting views concerning the correlation of mRNA and protein levels. Gygi et al. (1999) suggested that there is a poor correlation between most yeast mRNAs and protein levels with the exception of only the most abundant proteins. In contrast, Futcher et al. (1999) reported a good correlation between yeast mRNA abundances, measured by both SAGE and microarray chips, and protein abundances.

Given the large abundance of EST information for barrel medic (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi/), a simple comparison of identified protein levels with their corresponding mRNA levels was performed. Currently, over 145,000 EST sequences from approximately 20 different non-subtractive, non-normalized (J. White, TIGR, personal communication) cDNA libraries are available (Covitz et al., 1998; Cook, 1999; Bell et al., 2000; Gyorgyey et al., 2000). It is possible that a select few sequences from these libraries are being held back by the contributors, but these are few and specialized, and should have a minimal affect on the following comparisons. The cDNA libraries were used to estimate or "count" the relative expression level of a particular barrel medic transcript based on the repetitive occurrence of sequences from the same mRNA (Audic and Claverie, 1997; Ewing et al., 1999). The relative abundances of the top 200 ESTs for barrel medic leaves, stems, uninoculated roots, flowers, seed pods, and elicited cell cultures were quantified in this manner and are provided in Supplemental Table II (see www.plantphysiol.org). The relative abundances for the ESTs were generated using cDNA libraries originating from similar tissues; however, these tissues were from multiple and separate origins. Comparisons were based on functional annotation and not necessarily on specific protein or GenBank numbers, i.e. oxygen-evolving protein as opposed to P14226. Although this comparison is not of high analytical rigor, it does provide insight into correlation of protein and mRNA levels.

Although the proteins were arbitrarily chosen across pI and molecular mass ranges, most represent relatively abundant proteins typical of 2-DE and CBB 250 staining. Based on the 2-DE protein quantification results presented here, 67% of the identified proteins were in the top 100 most abundant proteins visualized with Coomassie, whereas 97% of the proteins identified were in the top 200 most abundant proteins. Thus, identified proteins were compared with the top 200 most abundant tissue-specific ESTs in related cDNA libraries. The percentages of the identified proteins observed by 2-DE that were also observed in the top tissue specific ESTs are summarized in Table III. This summary reveals that an average of 50% of the identified proteins were observed in the top 200 tissue-specific ESTs. An evaluation of the top 100 tissue-specific ESTs shows that 40% of proteins identified in 2-DE experiments were also observed in the 100 most abundant tissue-specific ESTs. These results suggest a moderate level of correlation between mRNA and protein. For example, leaf proteins such as the photosynthetic enzymes Rubisco small subunit and oxygen-evolving protein appear to be highly correlated with their respective mRNA levels.

Interestingly, some highly expressed proteins such as Rubisco large subunit were not observed in the EST libraries. As mentioned earlier, we believe that this is due to the chloroplast-encoded nature of certain mRNAs, such as Rubisco large subunit, which do not contain poly(A⁺) tails necessary for purification and cDNA library preparation (Sambrook et al., 1989).

Highly abundant leaf ESTs not represented in the protein data to date included aquaporins, chlorophyll-binding proteins, and cytochrome B6. This apparent lack of correlation can be explained by the integral thylakoid membrane nature of these proteins. It is commonly accepted that integral membrane proteins are underrepresented in 2-DE due to poor solubilization. Lipoxygenase also appeared in the top 100 clones of five tissue-specific EST libraries; however, it was never identified in the protein dataset. Plants express both cytosolic and chloroplast isoforms of lipoxygenase, most of which have a molecular mass of approximately 100 kD. A possible explanation for the absence of this protein from the protein data could be the inherent discrimination against high-molecular mass proteins encountered during isoelectric focusing using IPG strips of fixed gel composition (Candiano et al., 2002).

The lack of correlation between mRNA and protein could not always be explained. For example, identified stem proteins included acid phosphatase, actin, and osmotin; however, these proteins were absent or of very low abundance in the stem-specific EST library. Other proteins identified but not represented in the EST libraries included: RNA-binding protein and ankyrin repeat protein in flowers and hydroxyacyl glutathione hydrolase in roots. Interestingly, elongation factor 1-alpha was observed as a highly expressed EST (top 50) in all tissues but was not observed in the protein set. The lack of correlation may be due to the relative turnover rates of both transcripts and proteins, or translational controls such as codon bias (Gygi et al., 1999), mRNA secondary structure (Wang and Wessler, 2001), or upstream open reading frame repression (Wang and Wessler, 1998).

Based on the limited comparison above, we estimate a moderate 50% correlation between protein and mRNA levels. This value suggests a correlation that is higher than that reported by Gygi et al. (1999) but lower than that reported by Futcher et al. (1999). If the limitations imposed by the chloroplast-encoded proteins, poor representation of membrane proteins in 2-DE, and our limited protein dataset are taken into account, a higher correlation than that reported may be possible. Although a significant level of correlation is perceived, there are still many specific examples that show poor correlation.

	CONCLUSIONS

TOP ABSTRACT INTRODUCTION RESULTS AND DISCUSSION CONCLUSIONS MATERIALS AND METHODS ACKNOWLEGMENTS LITERATURE CITED

To date, we have identified over 300 proteins in specific tissues of barrel medic. Protein identifications using only protein databases were 25% successful even with good peptide mass fingerprints. Significant increases in protein identification success rates were achieved by using EST sequence databases. Using complementary protein, nucleotide, and EST sequence libraries, we were able to achieve a protein identification success rate of 55% for our representative protein dataset. We consider this a relatively high success rate in the absence of a genomic sequence and in comparison with other plant proteomic projects. Tentative consensus searches currently are being performed and confirm many of the proposed identifications in this study (Asirvatham et al., 2002b); however, this topic will be discussed in a separate publication.

The 2-DE profiles of various barrel medic tissues provide reference maps for future proteomic comparisons of genetic mutants, biotically and abiotically challenged plants, and/or environmentally challenged plants. The identified proteins provide a survey of those proteins observable using current technology and also serve to define the limitations of the reported proteomics approach. For example, it will be difficult to study other physiological processes besides photosynthesis and carbon metabolism in leaves using current proteomic technologies due to the very high level of these proteins in leaves. Further, the proteins identified serve as physiological markers of tissue-specific protein expression. Based on the limited dataset, 39% of all the identified proteins were only identified in a single tissue. These putative unique proteins provide valuable insight into the specialized physiological function of each of the tissues. For example, a comparison of roots and root-derived cell cultures can yield insights into the physiological phenomena associated with the dedifferentiation of root tissue during establishment of a suspension cell culture.

A comparison between the levels of the identified proteins and mRNA levels quantified through EST counting was performed. It is estimated that on average 50% of the proteins appear to be correlated with their corresponding mRNA levels; conversely, 50% are not. Information on both transcript and protein levels can be utilized for targeting potential regulatory genes that are characterized by high transcript but low protein levels.

The proteins identified in this study as unclear or putative represent unique opportunities to probe molecular function. Systematic perturbations and monitoring of these proteins would be expected to yield insight into function. These abundant but unclassified proteins have been linked to specific ESTs and, thus, establish the feasibility to experimentally monitor both the protein and mRNA. The relatively high abundance of these proteins further stresses the biological but unknown importance of these proteins in barrel medic.

This report provides a comprehensive overview of the barrel medic proteome and provides a good foundation for future comparative proteomic efforts associated with this important model plant. The importance of barrel medic is further emphasized by the recent recommendation from the National Academy of Sciences that the goals of the National Plant Genome Initiative for 2003 through 2008 should focus on a small number of key species including barrel medic (http://books.nap.edu/books/0309085292/html/index.html). This work serves as a major step in this direction for a key plant species. As we seek to better understand gene function and to study the holistic biology of systems, it is inevitable that we study the proteome.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION RESULTS AND DISCUSSION CONCLUSIONS MATERIALS AND METHODS ACKNOWLEGMENTS LITERATURE CITED

Plant Material and Protein Extraction

Differentiated plant tissues were collected from barrel medic (Medicago truncatula cv Jemalong A17) grown in an environmentally controlled growth chamber and maintained under standard conditions (Asirvatham et al., 2002a). Eight-week-old plants were used for leaf and stem tissue. The top two apical unfolded trifoliates were sampled for leaf tissue, and stem tissue was restricted to the first two apical internodes. Flowers included all stages from buds until petal browning and all parts except the peduncles. Green seed pods were collected from a variety of developmental stages (including very young pods to those with maturing seeds) of 3-month-old plants. Roots were collected from seedlings grown in perlite 2 weeks after planting. Total protein from these tissues was extracted according to a reported method (Tsugita et al., 1994). In brief, tissues (0.4-1.0 g) were ground in liquid N₂ and proteins precipitated at 20°C with 10% (w/v) TCA in acetone containing 0.07% (w/v) 2-mercaptoethanol for at least 45 min. The mixture was centrifuged at 35,000g at 4°C for 15 min, and the precipitates were washed with acetone containing 0.07% (w/v) 2-mercaptoethanol, 1 mM phenylmethylsulfonyl fluoride, and 2 mM EDTA. Pellets were dried by vacuum centrifugation and solubilized in 8 M urea, 4% (w/v) CHAPS, 20 mM DTT, 0.1% (v/v) Biolytes (pH 3-10; Bio-Rad Laboratories, Hercules, CA; Molloy et al., 1998).

Cell cultures derived from barrel medic cv Jemalong A17 roots were grown in the dark in shaker flasks and suspended in Schenk and Hildebrandt (SH) medium with transfer to fresh medium every 2 weeks. Cells were harvested 4 d after transfer, washed once with fresh SH medium and once with SH:water (1:1 [v/v]), ground in liquid N₂, and extracted with 40 mM Tris (pH 9.5), 50 mM MgCl₂, 2% (w/v) polyvinylpolypyrrolidone, 1 mM phenylmethylsulfonyl fluoride, and 120 units mL¹ endonuclease (catalogue no. E8263, Sigma, St. Louis) by sonication (Molloy et al., 1998). After centrifuging at 12,000g, 4°C, for 10 min, proteins in the supernatant were precipitated on ice with 12% (w/v) TCA, centrifuged, and washed with cold acetone. The pellet was air dried and resuspended in solubilization buffer.

Protein Quantification and Electrophoresis

Protein concentrations of all tissue extracts were quantified using the Bradford method (Bradford, 1976) and a commercial dye reagent (Bio-Rad) with bovine serum albumin as a standard. Eleven-centimeter immobilized pH gradient (IPG) strips (linear, pH 3-10) from Bio-Rad were rehydrated at 20°C with 0.75 to 1.0 mg of protein in 300 µL for 15 to 16 h. Focusing was carried out in a Bio-Rad Protean IEF Cell for a total of 35,000 volt hours. After focusing, strips were equilibrated with reduction and then with alkylation buffers, loaded onto a 12% (w/v) acrylamide gel, and run at 25 mA gel¹ (Asirvatham et al., 2002a). Gels were stained overnight with Coomassie Brilliant Blue R-250 and destained the next day. Gel images were digitized with a Bio-Rad FluorS equipped with a 12-bit camera. Experimental molecular mass and pI were calculated from digitized 2-DE images using standard molecular mass marker proteins and the linear calibration option of Genomic Solutions HT Analyzer software (Genomic Solutions, Ann Arbor, MI).

Digestions and MALDI-TOFMS

Protein spots were excised from the gel, washed twice with water for 15 min, and destained with a 1:1 (v/v) solution of acetonitrile and 50 mM ammonium bicarbonate while changing solutions every 30 min until the blue color of Coomassie was removed. 2-DE gel spots were then dehydrated by washing twice with 100% acetonitrile and dried by vacuum centrifugation. Gel plugs were rehydrated with a solution of 10 ng µL¹ bovine trypsin (Roche) in 25 mM ammonium bicarbonate and digested for 4 to 6 h at 37°C. The enzymatic digestions were stopped with the addition of 10% (v/v) formic acid, and the supernatant was saved. Gel plugs were extracted once with 25 µL acetonitrile:water (1:1 [v/v]) and once with 25 µL of 100% (w/v) acetonitrile. Supernatants were combined and taken to dryness. Peptides were resuspended in 2% (w/v) formic acid:acetonitrile (1:1 [w/v]), mixed 1:1 with matrix (10 mg mL¹ -cyano-4-hydroxycinnamic acid in same solvent), and spotted for MALDI-TOFMS. Mass spectra were obtained with a PerSeptive Biosystems DE-STR at an instrument resolution exceeding 10,000 and internally mass calibrated by matching to at least one and often more autolytic trypsin peaks (906.5049, 1153.5741, 2163.0570, and 2273.1602). Database search results were reprocessed with a reiterative search algorithm (Intellical, XXXX, XX) at 20 ppm that recalibrates m/z based on the best hit. Intellical software is part of the ABI Proteomics Solutions 1 software. If the best match is a real match, the identification confidence score will increase after reiterative calibration. If the best match is a false positive, the score will generally decline. The process was especially useful when trypsin autolytic peaks were of low abundance or absent. Resultant peptide mass fingerprints were assigned an arbitrary quality score (PMFQ) to quantify the quality of the peptide fingerprint and are reported in Supplemental Table I. The PMFQ scores were assigned based on the relative number of analyte peptides observed and their relative intensities as compared with the most abundant trypsin autolytic peptide peaks (2,163 and 2,273). If no peptides were observed or if analyte peptides were less than 10% of the trypsin autolytic peaks, a PMFQ value of 0 was assigned. If fewer than five peptides with relative intensities less than the trypsin peaks were observed, then a PMFQ of 1 was assigned. If five or more analyte peptides with intensities approximately equal to the trypsin autolytic peaks were observed, then a PMFQ value of 3 was assigned. If significantly more peptides were observed with a relative intensity greater than the trypsin autolytic peaks (but trypsin peaks still >10% for internal m/z calibration) were observed, then a PMFQ value of 4 (approximately 10 peptides) or 5 (>10 peptides) was assigned. Both MALDI-TOFMS peptide fingerprints illustrated in Figure 2 have a PMFQ of 5.

Database Queries and Protein Identifications

The peptide mass fingerprints were compared with sequences in: (a) NCBInr database (release January 1, 2002), (b) SwissProt database (release January 1, 2002), and/or (c) dbESTothers (NCBI; release January 1, 2002), (d) and/or a subset of dbESTothers (NCBI) consisting of approximately 145,000 barrel medic EST sequences, dated November 15, 2001, and queried using MS-Fit (http://prospector.ucsf.edu) in an automated mode using Proteomic Solutions 1 software from Applied Biosystems (Foster City, CA). Mass spectra were de-isotoped, baseline corrected, and threshold adjusted before database searching. Database searches were performed using a 100-ppm mass accuracy with a minimum requirement of four peptide matches from a submission list of typically 30 peptides. The maximum number of missed cleavages was set at one. The only user-defined modification specified was carbamidomethylation of Cys; however, the software default considered possible modifications of N-terminal Gln to pyro-Glu, oxidation of Met, and protein N terminus acetylation. When peptide mass fingerprints were matched to sequences in the EST databases, functional information was obtained by BLASTX (NCBI; http://www.ncbi.nlm.nih.gov/BLAST/) of the sequence or reference of the clone identifier to the barrel medic gene index (MtGI; http://www.tigr.org/tdb/mtgi/). The theoretical molecular mass and pI of the identified protein were then calculated using GPMAW (Lighthouse data) and compared with the experimental molecular mass calculated from the digitized 2-DE images. Protein identifications were evaluated on the basis of multiple variables including the number of peptides matched, mass error (m/z accuracy), percent coverage of the matched protein with 10% of the full-length protein set as the minimum value, quality of the peptide maps, intensity of the matched peaks (18%-20% minimum), similarity of experimental and theoretical protein molecular masses and pIs, and species from which the sequence was matched. For EST matches, the percent coverage was calculated by dividing the number of matched amino acids by the total number of amino acids in the protein sequence returned from the BLASTX or MtGI searches.

LC/MS/MS

Select digest mixtures were analyzed by nanoscale HPLC coupled with LC/MS/MS. Data were obtained using an ABI QSTAR Pulsar (Applied Biosystems) hybrid quadrupole time-of-flight mass spectrometer. The instrument m/z was calibrated with standards supplied by the manufacturer. Separated peptides were introduced into the mass spectrometer from an HPLC system equipped with an autosampler (LC Packings, San Francisco). Separations were achieved using an LC Packings nanoscale pepmap column (15 cm × 75 µm i.d., 3 µm, 100 Å, C18) and a linear binary gradient (solvent A was 1% [v/v] formic acid in 95%:5% [v/v] water:acetonitrile, whereas solvent B was a 0.8% [v/v] formic acid in 5%:95% [v/v] water:acetonitrile). The linear gradient was 95% (w/v) A:5% (w/v) B (0 min) to 60% (w/v) A:40% (w/v) B over 33 min, then ramped to 5% (w/v) A:95% (w/v) B at 37 min and held at 5% (w/v) A:95% (w/v) B until 42 min, where it was returned to 95% (w/v) A:5% (w/v) B 48 min and allowed to reequilibrate to 95% (w/v) A:5% (w/v) B 60 min. Nanoscale-ESI was performed using a Protona interface and nanoelectrospray needles (silver-coated glass capillary, New Objective, Woburn, MA). Mass spectra datasets were searched against NCBInr, SwissProt, dbESTothers, and mtEST databases using Mascot (http://www.matrixscience.com). The search results were validated as described for the peptide mass fingerprint results.

EST Counting and Protein Relative Abundance Estimates

Barrel medic ESTs were extracted from dbEST (http://www.ncbi.nlm.nih.gov/, accessed November 4, 2001). ESTs were assembled into tentative consensus sequences by TIGR to generate the barrel medic gene index (MtGI, http://www.tigr.org/tdb/tgi.shtml). The MtGI release of September 7, 2001 was used to count the occurrence of barrel medic genes in six different EST datasets including leaf (one cDNA library of developing leaf, 7,831 ESTs), stem (one library of developing stem, 10,314 ESTs), root (three libraries of uninoculated root, 6,593 ESTs), flower (one library of developing flower, 3,404 ESTs), seed pod (one library of developing seed and one library of developing pod, 4,587 ESTs), and cell suspensions (one library of elicited cell suspensions, 8,926 ESTs). The barrel medic genes were then sorted in the descending order on their EST counts for each dataset and used in the comparison with proteomic data.

Protein abundances were calculated using the normalized spot volume of each protein determined with HT Analyzer software (Genomic Solutions) as previously reported (Asirvatham et al., 2002a).

	ACKNOWLEGMENTS

TOP ABSTRACT INTRODUCTION RESULTS AND DISCUSSION CONCLUSIONS MATERIALS AND METHODS ACKNOWLEGMENTS LITERATURE CITED

We thank Dr. Richard Dixon for scientific discussion and editorial comments. We thank Drs. Zhentian Lei and Aaron Elmer for their assistance in performing LC/MS/MS analyses.