The nematode resistance allele at the rhg1 locus alters the proteome and primary metabolism of soybean roots.

Heterodera glycines, the soybean cyst nematode (SCN), causes the most damaging chronic disease of soybean (Glycine max). Host resistance requires the resistance allele at rhg1. Resistance destroys the giant cells created in the plant's roots by the nematodes about 24 to 48 h after commencement of feeding. In addition, 4 to 8 d later, a systemic acquired resistance develops that discourages later infestations. The molecular mechanisms that control the rhg1-mediated resistance response appear to be multigenic and complex, as judged by transcript abundance changes, even in near isogenic lines (NILs). This study aimed to focus on key posttranscriptional changes by identifying proteins and metabolites that were increased in abundance in both resistant and susceptible NILs. Comparisons were made among NILs 10 d after SCN infestation and without SCN infestation. Two-dimensional gel electrophoresis resolved more than 1,000 protein spots on each gel. Only 30 protein spots with a significant (P < 0.05) difference in abundance of 1.5-fold or more were found among the four treatments. The proteins in these spots were picked, trypsin digested, and analyzed using quadrupole time-of-flight tandem mass spectrometry. Protein identifications could be made for 24 of the 30 spots. Four spots contained two proteins, so that 28 distinct proteins were identified. The proteins were grouped into six functional categories. Metabolite analysis by gas chromatography-mass spectrometry identified 131 metabolites, among which 58 were altered by one or more treatment; 28 were involved in primary metabolism. Taken together, the data showed that 17 pathways were altered by the rhg1 alleles. Pathways altered were associated with systemic acquired resistance-like responses, including xenobiotic, phytoalexin, ascorbate, and inositol metabolism, as well as primary metabolisms like amino acid synthesis and glycolysis. The pathways impacted by the rhg1 allelic state and SCN infestation agreed with transcript abundance analyses but identified a smaller set of key proteins. Six of the proteins lay within the same small region of the interactome identifying a key set of 159 interacting proteins involved in transcriptional control, nuclear localization, and protein degradation. Finally, two proteins (glucose-6-phosphate isomerase [EC 5.3.1.9] and isoflavone reductase [EC 1.3.1.45]) and two metabolites (maltose and an unknown) differed in resistant and susceptible NILs without SCN infestation and may form the basis of a new assay for the selection of resistance to SCN in soybean.

Soybean (Glycine max) is the world's most important legume crop; it is grown for both oil and protein coproducts (Messina and Liu, 1997). Many natural by-products are also derived from soybean, and several possess significant antimicrobial and pharmacological properties (Dixon and Sumner, 2003). The abundance and variety of soybean's bioactive factors suggest that the species has evolved and been domesticated in the presence of many persistent pathogen populations.
After water deficit, the leading causes of seed yield losses in soybean are biotic stresses from pathogen infections (Wrather et al., 1996). Among soybean pathogens, the soybean cyst nematode (SCN) causes the most damage. Yield losses in the world due to SCN infection have approached $2 billion a year (Niblack et al., 2004) and amounted to $1.5 billion in the United States alone for the 1996 to 1997 period (Chen et al., 2001). SCN preventative measures include crop rota-tion, the use of resistant cultivars, and nematicides. The use of resistant cultivars provides the most efficient and practical control of SCN.
Soybean breeding has heavily used three plant introductions (PIs) resistant to the cyst nematode (Brucker et al., 2005). The PIs 'PI88788', 'Peking', and 'PI437654' have been shown to carry resistance loci effective against multiple nematode races (Concibido et al., 2004). However, the resistance to SCN was quantitative and controlled by multiple loci. Previously identified were two quantitative trait loci that provided substantial resistance against the SCN Hg type 0 (a biotype of race 3) in near isogenic lines (NILs) derived from Peking via 'Forrest' (Meksem et al., 2001;Ruben et al., 2006). The resistance locus Rhg4 on linkage group A2 provided resistance against Hg type 0, whereas rhg1, the resistance locus on linkage group G, provided some resistance against nematodes of all known races. The rhg1 locus allele from Peking causes rapid necrosis of the giant cells and the formation of thickened cell walls around the defunct feeding site (syncytium) by 24 to 48 h after the SCN female starts to feed, some 4 to 5 d after infestation of the roots (Alkharouf et al., 2004(Alkharouf et al., , 2006Ithal et al., 2007aIthal et al., , 2007b. The activity of the resistance allele is associated with global changes in transcript abundance (TA) within the syncytium and systemic changes in the root system, with hundreds of transcripts altered. However, in plants, the correlations between transcript, protein, and metabolite abundances are low. Hence, key proteins may have been overlooked by transcript analysis. Furthermore, some proteins can alter the metabolome without alterations in their abundances (Edwards, 1996;Ito et al., 2003;Schad et al., 2005;Mungur, 2008). Therefore, key metabolites may remain undetected. Finally, protein-protein and protein-metabolite interactions are important in many plant responses to the environment (Geisler- Lee et al., 2008). Therefore, key interactions may remain undetected by the transcript analyses reported to date.
The rhg1 locus appears to be a multifunctional and multigenic locus (Ruben et al., 2006;Afzal, 2007;Iqbal et al., 2009). The locus contained several genes capable of altering the transcriptome, proteome, metabolome, and interactome, including a receptor-like kinase (RLK; Afzal and Lightfoot, 2007) inferred to alter root development (Afzal et al., 2008a(Afzal et al., , 2008b; an unusual laccase (Iqbal et al., 2009) involved in phenolic metabolism; and a predicted Na/H antiporter. Flanking the locus were two proteins of unknown function, one of which is transcribed in roots (Triwitayakorn et al., 2005;Ruben et al., 2006). Each of the genes was encoded by distinct alleles in resistant soybean compared with susceptible soybean, with significant amino acid and nucleotide changes (Ruben et al., 2006;Afzal et al., 2008a). Analysis in NILs suggested that the resistance alleles of the core set of three linked genes cannot be separated by recombination without the loss of resistance (Afzal, 2007;Afzal et al., 2008a).
Recent progress in the plant disease resistance field has shed light on some resistance (R) proteins and their effector molecules  for review, see Afzal et al., 2008b). However, the full set of interacting partners implicated in resistance has not been completely decoded. Equally, a number of systematic studies have made progress identifying the transcription abundances that change as a consequence of SCN infection (Alkharouf et al., 2004(Alkharouf et al., , 2006Ithal et al., 2007a;Klink et al., 2007aKlink et al., , 2007b. However, by 2009, there was no report that had addressed the difference in protein abundance in SCN-infested compared with noninfested roots. Proteomics databases and reference maps for Medicago truncatula (Lei et al., 2005) and soybean (Xu et al., 2006) have been constructed using two-dimensional (2D) gel electrophoresis. The 2D gels have also been employed in the identification of proteins expressed during seed filling (Mooney and Thelen, 2004;Hajduch et al., 2005), symbiotic nitrogen fixation (El Yahyaoui et al., 2004), for the detection of glycoproteins using iodine-labeled lectins (Basha and Roberts, 1981), and to explore the consequences of biotic stresses (Lei et al., 2005). However, reports of proteomic analysis of the pathogen-plant relationship have been sparse. These include infection of root hairs with Bradyrhizobium japonicum (Wan et al., 2005) and the analysis of pathogenesis-related proteins from tobacco necrosis virus-infected soybean plants (Roggero and Pennazio, 2008).
Recent proteomics projects were encouraged by marked improvements in 2D gel reproducibility. This was achieved mainly by the development of immobilized pH gradient strips (Gorg et al., 2000) and enhanced techniques to obtain soluble proteins (Molloy et al., 1998). Furthermore, image-analysis software with automated statistical analysis eased the laborious process of spot quantification. In addition, in-gel protease digestion, refinement of electrospray ionization tandem mass spectrometry (MS/MS), matrix-assisted laser desorption ionization time-of-flight mass spectrometry (TOF-MS), and protein identification using various databases (National Center for Biotechnology Information [NCBI], Protplot, or organism-specific EST databases) have improved. Accurate identification of femtomolar quantities of proteins is now routine.
The hypotheses explored in this paper were as follows: that expression of the set of candidate genes at rhg1 might alter the abundance of other plant proteins encoded by unlinked genes; that the altered proteins might form enriched clusters within the interactome; and that protein abundance changes have effects at the metabolite layer. In order to test these hypotheses, the key proteins underlying SCN resistance were identified by 2D gel electrophoresis of total proteins extracted from the roots of NILs with different alleles at rhg1 infested for 10 d with the SCN Hg type 0. Differences in protein abundance in response to both genotype and pathogen infestation were noted. TAs were determined for the genes corresponding to the differentially abundant protein spots. An interactome map generated on the basis of orthology between soybean and Arabidopsis (Arabidopsis thaliana) proteins helped identify a significant cluster of interacting proteins.

RESULTS
NILs polymorphic across the region encompassing the rhg1 locus were used for this study. The region between markers BARC-Satt214 (and Satt163) and Satt570 spans 12.7 centimorgans (3.16 Mb), with the rhg1 candidate gene cluster at about 4 centimorgans, close to the center (Meksem et al., 2001;Ruben et al., 2006;Afzal et al., 2008a). The NIL 34-3 was susceptible to the SCN Hg type used; in contrast, NIL 34-23 was resistant to SCN. NIL genotypes were confirmed by the Southern Illinois University, Carbondale TMD1 marker (GI:218937751) located in the intron of the RLK within the rhg1 locus (data not shown; Afzal et al., 2008a). Infection of the roots with Hg type 0 nematode at 10 d after infestation (dai) was confirmed by PCR amplification of the nematode 18S ribosomal gene from root-extracted DNA (Fig. 1).
Differentially abundant proteins and metabolites were identified from four two-way comparisons. The first was made between Hg type 0 infested and noninfested SCN resistant NIL 34-23. The second compared the infested resistant NIL 34-23 and the infested susceptible NIL 34-3. The third was made between the infested and noninfested susceptible NIL 34-3. The fourth compared the noninfested resistant NIL 34-23 and noninfested susceptible NIL 34-3. These comparisons were made in an effort to find overexpressed proteins in the susceptible plants infested with SCN Hg type 0 and proteins more abundant in the resistant NIL 34-23. Answers to three fundamental questions were sought. (1) Which proteins and metabolites are differentially abundant in the resistant NIL plant roots as a consequence of SCN infestation? (2) Which proteins and metabolites were at higher abundance in the infested resistant NIL plant roots and not in the infested susceptible NIL plant roots? And (3) were any proteins or metabolites at different abundance in resistant and susceptible NIL plant roots without SCN infestation?

Proteome Abundance Analysis
More than 1,000 protein spots were reproducibly resolved on each gel (Fig. 2). The pI for the majority of the spots ranged from pH 4 to 8, whereas the molecular masses for most proteins were between 20 and 100 kD (Fig. 2). Spot comparisons for the two-way analyses were subjected to the Student's t test with a significance (P) value set at 0.05. A total of 29 spots with a 1.5fold or higher abundance were chosen for further analysis. Nine spots were overrepresented in the resistant NIL 34-23-infested roots compared with noninfested roots (Table I). Intensities for 12 spots varied between the infested resistant and susceptible NIL roots (34-23-I versus 34-3-I comparison gels; Table II). Four protein spot abundances differed in the susceptible infested roots compared with noninfested roots (34-3-I versus 34-3-NI gels; Table III). Only two spots had higher abundance in the resistant noninfested compared with the susceptible noninfested roots (Table IV), showing that both NIL proteomes were closely isogenic in the absence of pathogen infestation.
For 25 of the 28 spots analyzed, the theoretical and experimental pI and molecular mass deviations were within 20% of the expected values (Tables I-IV). For two of the deviating spots (Glc-6-P isomerase [EC 5.3.1.9] and glutathione S-transferase [GST; EC 2.5.1.18]), the extent was marginally greater than 20%, whereas for one spot (ascorbate oxidase [EC 1.10.3.3]), the deviation was significant (Table II) because the predicted protein was a preprotein. The observed differences may be attributed to incorrect N-or C-terminal inferences, novel posttranslational modifications, or length variations between the homologous proteins.
For correctly annotated proteins, there is usually no more than 20% difference between experimental and theoretical molecular mass and pI values (Wan et al., 2005). However, setting 20% as an arbitrary cutoff may lead to the exclusion of some correctly assigned proteins that may migrate to unexpected positions on a 2D gel due to posttranslational modifications (Wan et al., 2005). Furthermore, some spots yielded more than one protein, not an unusual occurrence for 2D analysis of total protein samples involving proteins with similar pI and molecular mass values.

Functional Classification of Proteins
The soybean genome was not yet completely sequenced and annotated in 2009; about 1,200 sequence gaps remained. Nevertheless, most of the protein spots of different abundance were successfully identified using the Department of Energy (DOE) sequence, the soybean EST, NCBI, and Swissprot databases. A few sequences had no match to an EST, nucleotide, or protein sequence from soybean. These were identified on the basis of homology to proteins from the Arabidopsis or the M. truncatula databases. After identification, the proteins were classified into functional categories (Bevan et al., 1998). The largest functional group was represented by proteins implicated in disease or defense responses (45%), followed by proteins involved in energy-related metabolic functions (29%). About 11% of the proteins belonged to the protein destination and storage category (Fig. 3).

DISCUSSION
Among the significant sample of the high to moderate abundance proteins in soybean roots that could be visualized on 2D gels, finding just 30 significant differences (3%) was surprising. In comparison with metabolites, where 58 of 131 were altered (44%), the number was very low. TA analyses in soybean genotypes that differ by both rhg1 and Rhg4 alleles also showed a larger proportion of differences (Alkharouf et al., 2004(Alkharouf et al., , 2006Ithal et al., 2007a;Klink et al., 2007b). Therefore, protein analysis may be focusing on key differences by eliminating TA changes that do not result in protein abundance changes. In addition, the metabolite analyses may show important posttranslational events adding to metabolite changes.

Proteins and Metabolites Altered in Abundance in Roots after SCN Infestation in the Resistant NIL
Among the nine proteins found to be increased by more than 1.5-fold in the roots of infested compared with the noninfested SCN-resistant NIL 34-23 (Table I) were two GSTs. The GST 8 and GST 11 spots had  34-23 (bottom). Most protein spots were present under both conditions. The pI for a majority of the spots was in the pH 4 to 8 range, whereas the molecular mass for most proteins was between 20 and 100 kD. significant differences in molecular mass and pI (Table  I). Metabolites in the glutathione pathway included the precursors Gly and Ala that were decreased in abundance (Table V); however, glutathione was not detected. EST analysis of the soybean root library at 8 dai by SCN also suggested a role for GST in resistance to SCN infection (Alkharouf et al., 2004). Similarly, comparative microarray analysis of SCN-infected soybean roots showed increased GST transcripts by 8 dai (Ithal et al., 2007a;Klink et al., 2007). The GST pathway leads to the detoxification of certain compounds during periods of plant stress through conjugation with glutathione. Soybean GSTs mainly conjugate the predominant thiol, homoglutathione (Marrs, 1996). Thus, soybean GSTs may be involved in detoxification reactions during nematode-induced pathogenesis (Alkharouf et al., 2004). Puthoff and coworkers (2003) used an Arabidopsis Affymetrix Genechip to show that following sugar beet cyst nematode (BCN; Heterodera schactii) infestation, three of the genes with  altered TA encoded GSTs. BCN and SCN are closely related species, so the change in GSTs may be an example of an orthologous plant response to nematodes. However, the TA of GSTs in Arabidopsis responds to various stimuli (Marrs, 1996;Coleman et al., 1997;Chen and Singh, 1999) and so may be part of a general defense response. The protein encoded by an EST (GI:31468505) was increased in abundance by 2.5-fold. The protein did not show similarity to any protein of known function, so it was not in any pathway at MapMan (Thimm et al., 2004). The source EST library was generated after salicylic acid induction (salicylic acid sprayed daily at 2.0 mM for 4 d) in whole 14-d-old seedlings from the SCN-susceptible 'Kefeng 1' plant (Tian et al., 2004). Kefeng 1 was released for innate resistance to Soybean mosaic virus. Salicylic acid is a common second messenger protein used by plants to induce resistance responses that are both local and systemic to many pathogens. Here, the existence of a protein product encoded by that mRNA that was 2.5-fold larger than predicted from the genome and EST sequence and an association with SCN resistance were both novel.
Two spots identified as increased in abundance in SCN-infested roots were identical to the soybean NADH-quinone oxidoreductase (QOR; EC 1.6.99.5) isozymes (quinine oxidoreductases; Table I). The enzymes were probably both involved in ubiquinione biosynthesis in roots. No metabolites in the mevalonic acid and quinine pathways were altered in abundance (Table V). However, a cDNA library isolated from laser microdissected syncytia collected 8 d after SCN infes-tation showed enhanced abundance of the transcript encoding a QOR in a susceptible soybean cultivar (Klink et al., 2005). Similarly, a transcript encoding a QOR was increased in abundance by 4-fold following infestation of Arabidopsis roots with BCN (Puthoff et al., 2003). Increased abundance of QOR transcripts in tomato (Solanum lycopersicum) overexpressing the PTO protein has also been observed (Mysore et al., 2003). QOR was shown to act on quinines, a common class of metabolites implicated in energy production, defense against pathogens, and electron transport (Matvienko et al., 2001). QOR reactions produce semiquinones that bind and subsequently inactivate key cellular components such as lipids, proteins, nucleic acids, and carbohydrates (Testa, 1995). Superoxide anions and hydroxyl radicals are also produced through the reaction of semiquinones with molecular oxygen. The superoxide anions and hydroxyl radicals cause enzyme inactivation, trigger membrane lipid peroxidation, and nick DNA. Thus, QOR plays a critical role in cell toxicity associated with the hypersensitive response, particularly in plants combating microbial pathogens (Hammond-Kosack and Jones, 1996). Plant pathogenic cyst nematode genomes like SCN encode several enzymes used in pathogenesis that are bacterial in origin (Ithal et al., 2007a). Perhaps QOR inhibits these enzymes.
One protein with a more than 3.5-fold increase in intensity showed high percentage similarity to a dirigent protein that mediates stereospecific lignin precursor couplings (Table I) and may be involved in the process of cell wall modification/lignification   (Burlat et al., 2001). MapMan (Thimm et al., 2004) placed the enzyme in the lignin and lignan biosynthesis pathway with laccase (EC 1.11.1.7), the enzyme that works in tandem with dirigent proteins. Phe, the amino acid used as a lignin precursor, was decreased in abundance in coordination with the hydroxyl coumaryl alcohol derivative agmatine and the cellulose breakdown product N-acetylgalactosamine (Table V). Together, the data suggest that cell walls and protective lignin derivatives are being synthesized to a greater degree in SCN-infested resistant roots. The dirigent proteins have previously been implicated in insect resistance by conifers (Ralph et al., 2006) and cyst nematode resistance in soybean (Klink et al., 2005). Interestingly, the transcript encoding the dirigent-like protein was shown to be decreased in abundance in 3-dai compatible interactions but not in 8-dai compatible interactions (Klink et al., 2007a). A gene encoding a variant laccase is found within the rhg1 locus (Ruben et al., 2006;Iqbal et al., 2009). Therefore, lignin, lignan, and/or isoflavone biosynthesis may underlie some part of resistance to SCN, perhaps the broad systemic acquired resistance response to secondary infestations from late-hatching eggs (Iqbal et al., 2009). An additional protein was identified within the dirigent-like protein spot. The protein corresponded to that encoded by a root EST that was predicted to encode the F1-ATPase subunit of mitochondrial ATP synthase (EC 3.6.3.14). Combined, the two proteins appeared to be increased by 6.5-fold. Since ATP synthases are very highly conserved, the spot might also correspond to a protein from the nematode rather than the plant protein, although there were many more plant cells than nematode cells in the samples. ATP synthase mediates ATP production critical to resistance reactions. In addition, a peptide derived from the proteolytic cleavage of the chloroplast ATP synthase can result in the production of the inceptin elicitor that activates plant defense responses in cowpea (Vigna unguiculata) during insect herbivory (Schmelz et al., 2006(Schmelz et al., , 2007. Nematode resistance genes can be pleiotropic to genes for resistance to other insect herbivores (Nombela et al., 2003), and there is some evidence for the phenomenon at Rhg4 in soybean (C.R. Yesudas and D.A. Lightfoot, unpublished data), so this finding may yet be proven significant.
Another protein identified from the comparison of the resistant NIL infested and noninfested was identical to the Kunitz-type trypsin inhibitor, which has strong homology to the a-fucosidase protein (EC 3.2.1.51; Table I; Augur et al., 1995). The O-glycosyl hydrolase (EC 3.2.1.-) family encompasses 85 subfamilies. They hydrolyze the glycosidic bond between two or more carbohydrates or between a carbohydrate and a noncarbohydrate moiety. The protein detected here was from family 29, a lysosomal enzyme responsible for hydrolyzing the a-1,6-linked Fuc joined to the reducing end GlcNAc of the carbohydrate moieties of glycoproteins. Among the related metabolites, Glc, sorbose, Man, Gal, Fru, and N-acetylgalactosamine decreased in abundance 1.5-to 7.0-fold, while maltose increased 13-fold in resistant NILs with SCN infestations (Table V). Increased TA of the mRNA encoding Kunitz trypsin inhibitor transcript has been observed during plant pathogenesis within the syncytia at 8 dai (Klink et al., 2007a) and in roots at 12 dai (Alkharouf et al., 2006). The protein was associated with the hypersensitive response in leaves of soybeans (Park et al., 2001) and transgeneic tobacco (McManus et al., 1999). The protein was involved in the regulation of nitric oxide synthase and responses to snake venom proteins (see discussion in Park et al., 2001). Therefore, the lysosomal activity of the O-glycosyl hydrolase that is a Kunitz trypsin inhibitor seems to be important in root resistance to SCN. Determining the enzymespecific substrates and molecular bases of the dramatic increase in maltose will be a priority for future work.
Another protein spot increased in abundance corresponded to a b-1,3-endoglucanase (EC 3.2.1.39; Table  I). The metabolite substrate Glc was decreased ( Table  V). The gene family forms a well-characterized group of pathogenesis-related proteins that include cell wallhydrolyzing enzymes. Endoglucanases have been reported to be activated by both biotic and abiotic stresses. The pathogenesis protein PR2 is a b-1,3endoglucanase (Graham et al., 2003(Graham et al., , 2007. The enzyme acts directly in plant disease resistance through digesting the cell wall of the invading pathogen (Bishop et al., 2005). The released cell wall fragments may also act as elicitors or in resistance, resulting in an accentuated response (Graham et al., 2007).
The most significant changes in protein spot intensity on the 2D gels corresponded to a vegetative  Shown is the fold increase or decrease associated with the NIL genotype (S, susceptible; R, resistant) and the treatment (I, infested with SCN; NI, not infested). Metabolites are grouped by class. Metabolites that could not be named unequivocally were prefixed with y_ and given retention time or class (CHO is the abbreviation for carbohydrate). Metabolites altered more than 5-fold between genotype or by inoculation with SCN are underlined; those increased more than 1.5-fold or decreased more than 1.5-fold are shown in boldface.  Fig. 4; Table I; Xu et al., 2006). The enzyme is at a key point in glycolysis and gluconeogenesis. Several metabolites, substrates, and end products in those pathways were decreased in abundance (Table V); only maltose and ribonic acid were increased. The enzyme is a target of glutathionylation by GST in Arabidopsis (Ito et al., 2003) and was increased in abundance by 2-fold in peach (Prunus persica) fruit after salicylic acid treatment (Chan et al., 2007). Therefore, this protein may be involved in the energetics of resistance. Considering the VSP, in both soybean and Arabidopsis the VSP was induced as a consequence of methyl jasmonate treatment, wounding, and insect feeding (Liu et al., 2005). The recombinant protein added to the diets of acid gut insects can cause a significant delay in development and an increase in mortality. The transcript accumulated in Arabidopsis plants infested with the SBN (Puthoff et al., 2003). Therefore the VSP may act as a deterrent to feeding.

Proteins and Metabolites Altered by SCN Infestation of Resistant Compared with Susceptible NIL Roots
Twelve spots had differential abundance (Table II). Three of the 12 proteins were increased in abundance in the inoculated resistant NIL roots (previous section). There was a 2.7-fold increase in triose-P isomerase and VSP spot intensity (Fig. 4) and a 3.3-fold increase in a dirigent-like protein (Table II). Therefore, these proteins were increased by SCN infestation in resistant but not susceptible NILs. Proteins increasing abundance in response to both genotype and infestation are likely to be important to resistance to SCN.
Among the proteins altered only in the resistantto-susceptible infested NIL comparison (increased by the resistance allele at rhg1 independent of SCN infestation), the enolase (EC 4.2.1.11) protein was increased 3-fold (Table II). The enzyme converts 2-phosphoglycerate into the shikimate pathway precursor, phosphoenolpyruvate, during both glycolysis and gluconeogenesis. The direct metabolites were not detected by gas chromatography-mass spectrometry (GC-MS), but several sugars and Krebs cycle acids, citric acid and fumaric acid, were reduced in the resistant NIL (Table V). Therefore, increased glycolysis may occur in the SCN resistance response as it does during both tomato defense against the powdery mildew fungus (Li et al., 2006) and symbiosis between soybean root hairs infected with Bradyrhizobium japonicum (Wan et al., 2005).
Another protein increased in abundance was the aldehyde dehydrogenase-1A1-like enzyme (EC 1.2.1…) involved in the conjugation of isoprenoids in plants to various products, including carotenoids and terpenoids. The precursor metabolites inositol and myoinositol were significantly decreased in the resistant roots (Table V). Furthermore, plants with mutations in this enzyme accumulated hydrogen peroxide, suggesting a role for this protein in plant defense against oxidative stress (Kirche et al., 2004). Conversely, elevated hydrogen peroxide can mediate the nonenzymatic condensation of lignins and lignans in association with dirigent proteins. Suppression of susceptibility to toxin inhibition of mitochondrial metabolism in T-cytoplasm maize (Zea mays) by interference with the production of the Turf-13 protein can be mediated by aldehyde dehydrogenases. Therefore, there are several mechanisms by which this enzyme might contribute to resistance to SCN through constitutively higher abundance in resistant genotypes.
A 3-fold increase in the S-adenosyl homocysteine hydrolase (SAH; EC 3.3.1.1) protein spot was detected in SCN-infested resistant NILs (Table II). This enzyme is involved in converting adenosylselenohomocysteine to adenosine and selenohomocysteine, substrates in many methyltransferase reactions. None of the metabolites in this pathway was detected by GC-MS. Methylation reactions are involved in the synthesis and activation of many antimicrobial compounds (Mitsui et al., 1997). SAH can inhibit the synthesis of ethylene, slowing senescence and the hypersensitive response. Transcripts encoding SAHs were increased during the hypersensitive response in cassava (Manihot esculenta; Kemp et al., 2005) and alfalfa (Medicago sativa; Edwards, 1996). Therefore, the protein seems likely to contribute to resistance by being constitutively increased in resistant NIL roots.
Three of seven enzymes in the ascorbate metabolism pathway were constitutively more abundant in resistant NIL roots. Ascorbate oxidase, dehydroascorbate reductase (DHAR; EC 1.8.5.1; Fig. 5A), and monodehydroascorbate reductase (mDHAR; EC 1.6.5.4) were identified (Table II). The precursor threonate was slightly decreased by about 1.4-fold in the resistant NILs (inferred by Fourier transform ion cyclotron resonance MS; Afzal, 2007). The oxidation of ascorbic acid to the toxic dehydro-L-ascorbic acid has roles in impeding insect growth and development (Felton and Summers, 1993). DHAR and mDHAR are involved in the detoxification of hydrogen peroxide and were increased in barley (Hordeum vulgare) genotypes resistant to powdery mildew (El Zahaby et al., 1995) and maize overexpressing a superoxide dismutase gene (EC 1.15.1.1; Kingston-Smith and Foyer, 2000). Tobacco expressing a foreign mDHAR showed increased abiotic stress resistance (El Tayeb et al., 2007). Therefore, constitutively higher expression of the ascorbate pathway in resistant NILs may be involved in innate resistance to SCN.
A multicatalytic endopeptidase protein was increased (Table II). The protein was highly similar (89% identity, 96% similarity) to an EST from embryos of SCN-resistant soybean 'Jack' that encoded the 20S proteasome b-subunit A1 (EC 3.4.25.1), a central hub in the protein-to-protein interactome (Supplemental Table S1). The 20S proteasome was involved in the breakdown of oxidatively modified or damaged proteins (Giulivi et al., 1994;Lee et al., 2006) and in the activation of SAGA-dependent transcription (Shukla et al., 2009). In rice (Oryza sativa) challenged with Rhizoctonia solani, this protein was increased in both resistant and susceptible genotypes, although the in-crease in abundance was higher in the resistant line. A constitutive increase in the multicatalytic endopeptidase protein abundance in resistant soybean NILs may be involved in maintaining the resistance response by degrading nematode proteins, by reducing or repairing oxidative damage by activating transcription of a subset of genes, or by changing the rates of deactivation and degradation of proteins according to their contribution to defense.
Two protein spots with increased intensities were members of the chaperonin (EC 3.6.4.9) family of proteins (Table II), a 10-kD-like chaperonin and a 23-kD-like cochaperonin that assist in protein folding and are heat shock proteins (HSPs; Sangster and Queitsch, 2005). Several HSPs were identified as activated upon pathogen infection (Sangster and Queitsch, 2005;Wan et al., 2005;Klink et al., 2007aKlink et al., , 2007b. Overexpression of HSP70 in the resistant plants was reported to reduce root mass in Arabidopsis (Sung and Guy, 2003). In this study and previously, shorter roots and lower root masses were seen in plants that underwent incompatible reactions (Brucker et al., 2005;Afzal et al., 2008a). Furthermore, a different chaperonin appeared to interact with the RLK at rhg1. Therefore, the chaperonins present at higher abundance in resistant roots may underlie three phenomena; part of the SCN resistance; the reduced seedling vigor associated with resistance (Brucker et al., 2005;Afzal et al., 2008a); and the lower seed yield that results (Yuan et al., 2002).

Proteins Induced by SCN in the Infested Susceptible NIL Roots
Proteins in this class might be part of a defeated defense response or proteins induced by signals from SCN to root cells. Four proteins were identified ( Table   Figure 5. The regions of three 2D electropherograms displaying differences in protein expression between the infested resistant NIL and the infested susceptible NIL. The protein circled in A was glutathione dehydrogenase (Table II). B shows differentially abundant proteins in the susceptible infested compared with the susceptible noninfested NILs. The protein circled was a peroxidase (Table III). III). One, mDHAR, had been detected as constitutively increased in the resistant genotype. The mDHAR spot, therefore, was increased in both the resistant infested and susceptible infested NILs, but to a lesser degree. Therefore, the protein increase was a general response to SCN infestation enhanced by resistance to SCN directed by rhg1.
A spot with a 6-fold increased abundance was identified as a Gly-rich RNA-binding protein (RBP ;  Table III). RBP TAs were increased by a number of external stimuli, including wounding, abscisic acid, dehydration, stress, mercuric chloride treatment, and infection with TMV (Brady et al., 1993;Naqvi et al., 1998). The role of nucleus-localized RBP in adjustments to changing internal and external stimuli may be based upon interactions with several RNAs, proteins, and kinases. Therefore, the change in protein abundance during a compatible reaction with SCN may be part of a general response to infection or may be directed by the nematode.
Two spots corresponding to peroxidase (EC 1.11.1.7) proteins were increased in abundance (Table III; Fig.  5B). The substrates were inferred to be inorganic chemicals not detected by GC-MS. Peroxidase TAs have dual roles in plant defense, sometimes decreased in abundance to channel resources to more urgent regulatory needs (Moy et al., 2004) or increased, as in 8-d post-SCN infestation syncytial cells compared with root cells (Klink et al., 2007b). Peroxidases may have an active role in wheat (Triticum aestivum) stem rust disease (Seevers et al., 1971;Tyagi et al., 2000). The increase in peroxidase isozymes during pathogen infection may restrict pathogen growth. Alternatively, peroxidases could be involved in the synthesis of plant phenolics or the detoxification of peroxide substrates (Seevers et al., 1971).

Protein Comparisons between Noninfested Resistant and Noninfested Susceptible NILs
Proteins in this class were expected to be constitutively increased by the presence of the resistance allele at rhg1. Only Glc-6-P isomerase and isoflavone reductase showed increased abundance in the comparison (Table IV). Glc-6-P isomerase was a glycolytic enzyme that converts Glc-6-P into Fru-6-P. The metabolite Fru and product sorbose were increased about 2-fold, while the substrate precursor maltose decreased 13-fold in resistant NILs (Table V). Previously, Glc-6-P isomerase was among a number of glycolytic enzymes with increased abundance in cereals in response to the sheath blight fungus (Danson et al., 2000). Therefore, increased glycolysis in roots might contribute to greater readiness to mount a resistance response to SCN. Figure 6. Part of the predicted interactome for soybean showing six proteins from the 2D gel analysis (chaperonin, F1-ATPase, multicatalytic endopeptidase, TL protein, cytosolic HSP, and triose-P isomerase). The interactome was extended by one node to include 159 nodes (proteins) and 194 edges (interactions; Supplemental Table S1). Two proteins with previously characterized roles in disease resistance were identified within this part of the interactome: the TL protein induced in response to fungal pathogens (blue) and a cytosolic HSP in disease resistance signaling (green). All six proteins interacted through two to seven intermediate partners (for pink spot identification, see Supplemental Table S1). Autoregulation was predicted for the triose-P isomerase and the chaperonin. [See online article for color version of this figure.] Isoflavone reductase (EC 1.3.1.45) was also implicated in readiness against SCN infestation by a 3-fold increase (Table IV). The precursor Phe was slightly decreased in abundance in resistant genotypes (Table  V), possibly as a result of the more active isoflavone synthesis. Isoflavone reductase accumulation in plants increases with increased resistance toward pathogen infection (Kuc, 1995;Ithal et al., 2007a) by increased production of phytoalexins and deposition of lignin (Keen, 1992). Interactions with the laccases and dirigent proteins are expected. Therefore, the isoflavonoid phytoalexin synthesis pathways are strongly inferred to be under the control of the rhg1 locus.

Interactome Analysis
Ortholog analysis was conducted on 31,921 proteins from Arabidopsis (from The Arabidopsis Information Resource) and 62,199 proteins from soybean (from the DOE at www.phytozome.net). Shown at SoyGD (http://soybeangenome.siu.edu) was the score that was the likelihood for direct orthology ("in-paralogs"), while the other members of the gene family were "outparalogs" (multiple orthologs among species without a clear in-paralog). If a single protein could not be assigned as the best ortholog, InParanoid (Remm et al., 2001) assigned it to clusters as one-to-many (one protein in one species having many orthologs in the other) and many-to-many (many proteins in one species having many orthologs in the other). For clarity here, only one-to-one orthologs (in-paralogs) were considered. This more stringent predicted soybean interactome was compared with the proteins altered in abundance by SCN infestation (Fig. 6). Among them, proteins were spread into many clusters, as expected from pathway analysis (19 enzymes in 17 pathways). However, initially, four proteins were clustered in one small part of the interactome, comprising 16 proteins within the interactome map. Since this clustering represented a massive enrichment (40,000-fold over random), their interactions were expanded by another node. After two iterations, two more proteins were added. Thus, from the huge soybean interactome, a small portion of 159 proteins with significant resistance to SCN was identified, with an enrichment of 4,000-fold, suggesting significance for the protein cluster.
The predicted interactome showed that chaperonin, F1-ATPase, multicatalytic endopeptidase, thaumatinlike (TL) protein, cytosolic HSP, and triose-P isomerase were interacting indirectly through two to seven intermediates (Fig. 6). The cluster of all potential interacting neighbors showed 159 proteins in total, with 194 interactions. The protein functions included transcription factors, chaperonins, signal transduction factors, and metabolism involved in energy generation (Supplemental Table S1). The four proteins in the cluster with the most interacting partners were a GTP-binding family protein (28 partners), importin a-2 subunit (64 partners), importin b-2 subunit (22 partners), and 26S proteasome non-ATPase regulatory subunit 3a (21 partners). Therefore, the regulation of protein degradation and transport of proteins across the nuclear membranes may be important aspects of resistance to SCN.
Among the 159 proteins predicted to interact were many proteins previously associated with disease resistance. One was the TL protein induced by SCN as well as in response to pathogens (displayed in blue in Fig. 6). TL proteins were previously shown to be Figure 7. Relative TA ratios measured by quantitative reverse transcription-PCR. The fold change in relative TA in the resistant SCNinfested NIL compared with the resistant noninfested NIL is shown for 10 plant replicates pooled after tissue grinding for RNA extraction. Data shown are means 6 SD from three technical replicates. defense proteins that reduced the permeability of the cell walls to pathogenic fungi (Trudel et al., 1998). The TL protein in soybean roots was predicted to interact with the nuclear importin a-2 subunit-like node ( Fig.  6; Supplemental Table S1), a nodal protein. The chaperonin detected to be changed in protein abundance by SCN was also predicted to interact at this node. Therefore, the regulation of transport and modification of both plant and nematode-secreted proteins across the nuclear membranes during resistance to SCN will be a focus of future studies. The second protein previously linked to disease resistance was a cytosolic HSP (AT4G11260.1; displayed in green in Fig. 6). This protein interacted with the 26S proteasome non-ATPase regulatory subunit 3a (Table II). In Arabidopsis, HSP90 was required for complete RPS2-mediated signaling in response to pathogens (Takahashi et al., 2003) and was shown to interact with the disease resistance signaling component SGT1b (Azevedo et al., 2002). Therefore, during resistance to SCN, protein-protein interactions were predicted to play an important role in cellular functions. Eventually, the interactome map for soybean may be used to predict which proteins in a network respond to various external stimuli and to design new mechanisms of resistance.

Comparison of TA between Infested and Noninfested Resistant NILS
Seven of the 19 genes that encoded enzymes in pathways were chosen to compare protein with TAs, because mRNA-specific primers that produced single amplicons could be designed. GST8 (gene 1a; 1.57 6 0.38), quinine oxidoreductase (1.38 6 0.27), F1-ATPase (1.87 6 0.53), triose-P isomerase (1.25 6 0.34), and GST11 (gene 8; 1.31 6 0.24) were increased in TA when the comparison was between the infested and noninfested NILs. There was agreement between the 2D gel and real-time quantitative PCR results for these genes. However, the gene transcripts encoding trypsin inhibitor (22.43 6 0.34), the EST induced by salicylic acid (22.41 6 0.39), and quinine oxidoreductase (21.2 6 0.20) were decreased in TA but increased in protein abundance ( Fig. 7; Table VI). The protein abundance increase in response to SCN was inferred to be a posttranscriptional event for these three proteins that may reflect differences between protein and RNA halflives . Correlations between transcript and protein abundances have mostly been weak (Greenbaum et al., 2003;Yu et al., 2007). Therefore, root resistance to SCN was shown to be derived from changes in both TA and protein abundance.

CONCLUSION
Numerous attempts to determine the key transcripts involved in SCN resistance in soybean have been made (Alkharouf et al., 2004(Alkharouf et al., , 2006Klink et al., 2005Klink et al., , 2007aKlink et al., , 2007b. Variations in experimental design included analysis of compatible and incompatible interactions, plant responses to resistant and susceptible SCN biotypes, and analysis of transcript variations at different days post inoculation. None of the studies thus far sought to determine changes in actual protein abundance as a function of SCN infection. Previous studies relied heavily on genomic tools, mainly microarray analysis using soybean Affymetrix chips to measure TA. One weakness of TA analysis is that the abundances of transcripts and of the respective proteins may not correlate well. In this study, there was limited correlation between some transcripts and their encoded proteins but more between proteins and their substrates or metabolites. Therefore, results obtained from transcriptomic, proteomic, and metabolomic analyses must be integrated and compared with the interactomes to elucidate the nature of resistance to SCN and other phenomena. This study showed that differentially abundant proteins exist in the context of genotype and reaction following SCN infection and laid the groundwork for a systems biology approach to SCN resistance. In addition, proteins and metabolites in roots were discovered that may be used to predict the resistance of seedlings without SCN infestation at high throughput. Future research will explore the usefulness of these biomarkers further.

Materials
Seeds of soybean (Glycine max) were obtained from Southern Illinois University at Carbondale. Seeds of NIL 34-23 (resistant haplotype between markers Satt214 and Satt570) and NIL 34-3 (susceptible haplotype between Satt214 and Satt570) were used at the F5:13 generation. Genotypes were rhg1rhg1Rhg4Rhg4 for NIL 34-23 and Rhg1Rhg1Rhg4Rhg4 for NIL 34-3 (if rhg1derived resistance was recessive; for the counter evidence, see Afzal et al., 2008a).

Growth Conditions
Soybean plants were grown in tubes (5 cm 3 30 cm) placed in buckets of sand with 20 tubes in a completely randomized design. Each tube contained a 1:1 ratio of sand to soil mix. The buckets were placed in a water bath to maintain root zone temperature at 24°C and grown in a 14-h-light cycle, with aerial daytime temperature of 30°C and nighttime temperature of 22°C. The humidity was maintained at approximately 40% to 50% (v/v).
Soybean plants (infested and noninfested) were removed from the cones at 10 dai. The cones were transferred to a cold-water (4°C) container and briefly soaked until the sand-soil mixture became loose. The plant roots were subsequently cleaned with distilled water, blotted dry, cut from the shoots, and stored in a 280°C freezer.

DNA Extraction
DNA was isolated following a modified protocol by Afzal et al. (2008a). Briefly, 100 mg of plant tissue was homogenized and frozen, and 600 mL of preheated (65°C) extraction buffer (Tris-HCl [pH 8.0; 100 mM], EDTA [pH 8.0; 20 mM], NaCl [1.4 M], cetyl-trimethyl-ammonium bromide [2%, w/v], and 2-mercaptoethanol [0.4%, v/v]) was added. The samples were incubated at 65°C for 1 h, cooled, and centrifuged at 10,000g for 15 min. The supernatant was decanted, and 5 mL of RNase (5 mg mL 21 ) was added and incubated at 37°C for 1 h. The DNA in the aqueous phase was purified by phenol:chloroform: isoamyl alcohol (25:24:1) extraction followed by chloroform:isoamyl alcohol (24:1). DNA was precipitated by the addition of isopropanol and centrifugation at 12,000g for 5 min. The DNA pellets were washed twice with 70% (v/v) ethanol, dried, and finally dissolved in 30 mL of Tris buffer. DNA concentrations were calculated by measuring A 260 .

Total Protein Extraction and Quantification
Total protein extraction followed Hurkman and Tanaka (1986) and Hajduch et al. (2005). Briefly, frozen root samples from 10 plants were pooled together, ground to a fine powder, and 2 g of tissue was resuspended in 5 mL of Tris-buffered phenol (pH 8.8) and 5 mL of extraction buffer (0.1 M Tris-HCl, pH 8.8, 0.4% [v/v] b-mercaptoethanol, 10 mM EDTA, and 0.9 M Suc). The solution was vortexed vigorously and centrifuged at 5,000g for 15 min. After removal of the top phase (phenol), the bottom phase was back extracted with Tris-buffered phenol (5 mL) and an equal volume of the extraction buffer. Proteins were pelleted by centrifugation at 20,000g for 20 min and washed according to Hajduch et al. (2005). The pellet was dried and resuspended in the DeStreak rehydration agent (GE Healthcare). Protein concentration was determined using a noninterfering protein assay (Sheffield et al., 2006).

2D Gel Electrophoresis and Image Analysis
Protein extracts (275 mg) from infested NILs or noninfested NILs (34-23 and 34-3) were used for the 2D electrophoretic analysis. Briefly, samples were initially hydrated overnight on 17-cm Bio-Rad immobilized pH gradient gel strips (pH 3-10), and isoelectric focusing was performed using the Protean IEF Cell (Bio-Rad). Equilibration of the strips was according to the manufacturer's instructions. Linear SDS-PAGE gradient gels (8%-16%, w/v) were used to resolve proteins in the second dimension. A Bio-Rad Protean II apparatus was used for gel electrophoresis at 15 mA cm 21 for 30 min, followed by 25 mA cm 21 for approximately 5 h at ambient temperature (20°C 6 2°C). Gels were washed with distilled water and stained with SYPRO Ruby. SYPRO Ruby fluorescent dye used for gel staining provided both higher sensitivity and a broad linear range for accurate protein quantification and detection (Sheffield et al., 2006).

Image Analysis and Spot Picking
Acquisition of gel images used a high-resolution laser scanner (Typhoon 9410 from GE Healthcare) and a CCD camera linked to a GelPix protein spot excision system (Genetix). Images were analyzed with the Imagemaster 2D software (GE Healthcare). The analysis was fully automated unless there was a need to manually edit unresolved spots. Identical spots from the compared gels were chosen by assigning landmarks to each gel. The volume for each protein spot was normalized against the total spot intensity. The Student's t test (P , 0.05) was used to determine whether spot intensities between treatments were significantly different. Differentially abundant spots were excised manually or with the GelPix system.

Protein Digestion and Cleanup
Proteins were digested in-gel as described previously (Gabelica et al., 2002), with the exception that digestion was carried out at 37°C overnight with 6 mg mL 21 trypsin (Promega) in 50 mM NH 4 HCO 3 . The samples were initially extracted with 30 mL of 1% (v/v) formic acid and 2% (v/v) methyl cyanide followed by incubation at 30°C for 30 min on a shaking platform. For the second extraction, 60% (v/v) acetonitrile was used. The pooled extractions were separated into aliquots, lyophilized, and stored. Aliquots were resuspended in 2% (v/v) acetonitrile and 1% (v/v) formic acid solution for use. Proteins were further purified and cleaned using a 10-mL C18 ZipTip (Millipore) according to the manufacturer.

Database Searching
Proteins were identified via peptide sequencing using electrospray ionization MS/MS as described previously (Chen, 2006). Analyst QS software (Applied Biosystems) was used for spectral processing. The peptides were searched against the soybean and Medicago truncatula EST databases (downloaded from NCBI in January 2006), the nonredundant NCBI database, and the Swissprot database using MASCOT version 1.9. Input parameters for variable and fixed modifications were specified as oxidation of Met and carbamidomethylation of Cys, respectively. Positive identification was based on the following: (1) number of peptide sequences identified in a protein; (2) calculated and theoretical pI/kD; and (3) total MASCOT and MOWSE scores (http://www.matrixscience.com/help/scoring_help.html) at P , 0.05. Protein sequences derived from EST databases or without significant matches were searched against the NCBI BLASTX and TBLASTX databases for evidence of orthology and against the DOE soybean peptide library (www. phytozome.net) for evidence of identity.

Quantitative PCR Analysis
Total RNA was isolated using Trizol (Invitrogen) from both mock-infected and SCN-infected root samples. Total RNA was treated with DNaseI, and the absence of DNA was confirmed by PCR amplification using intron-flanking primers before reverse transcription. Reverse transcriptase SuperScript II (Invitrogen) was used to generate first-strand cDNA following the manufacturer's protocol. Primer pairs for the genes corresponding to differentially expressed proteins were designed using the IDT primer design tool (www. idtdna.com). Primers are listed in Supplemental Table S2. SYBR Green (iQ SYBR Green Supermix; Bio-Rad) was used to measure the number of amplicons after the quantitative PCR. The experiments were performed in triplicate. The soybean tubulin gene was used as an internal control (Iqbal et al., 2009). At the end of each experiment, the melting curve for each amplicon was determined at 62°C to 95°C, with readings every 1°C. The 2 2DDCT method was used to analyze the relative changes in TA as described previously (Livak and Schmittgen, 2001).

Metabolite Extraction, Identification, and Pathway Analysis
Metabolites were extracted and identified from nine replicated plant samples using adapted standard operating practices for GC-TOF-MS analysis of plant extracts as reviewed by Mungur (2008). Briefly, 15 mg of freeze-dried root tissue was ground in 2 mL of extraction mix (methanol:water, 1:1, v/v) on ice, shaken for 4 to 6 min at 4°C, and centrifuged for 2 min at 14,000g. About 1 mL of supernatant was collected as two aliquots, which were dried in a vacuum centrifuge. All samples and replicates were stored at -20°C under argon or nitrogen until further processing. The dried extracts were dissolved in 5 mL of methoxamine hydrochloride (20 mg mL 21 pyridine) and incubated at 30°C for 90 min with continuous shaking. Then, 45 mL of N-methyl-Ntrimethylsilyltrifluoroacetamide was added to derivatize polar functional groups at 37°C for 30 min.
GC-TOF-MS (Leco Pegasus II GC-TOF mass spectrometer; Leco) analyses were performed on an HP 5890 gas chromatograph with tapered, deactivated split/splitless liners containing glasswool (Agilent) and 1.5 mL of splitless injection at 230°C injector temperature. Before each injection, the liner was rinsed with a pure N-methyl-N-trimethylsilyltrifluoroacetamide injection (1 mL). The gas chromatograph was operated at a constant flow of 2 mL min 21 helium with a 30-m 3 0.32-mm (i.d.) 3 0.25-mm MDN35 column (Macherey-Nagel). The temperature gradient started at 80°C, was held isocratic for 2 min, and subsequently was ramped at 15°C min 21 to a final temperature of 330°C, which was held for 6 min. Twenty spectra per second were recorded between mass-to-charge ratios 85 and 500. Peak identification and quantification were performed using the Pegasus software package ChromaTOF 1.61 (Leco). Retention time shifts were corrected by linear interpolation using known metabolites as reference markers. All files were subsequently processed against a reference that was generated using a signal-to-noise threshold of 10 with automated peak identification based on mass spectral comparison with a standard NIST 98 library and available in-house customized mass spectral libraries.
GC-MS identified 131 metabolites, among which 58 were significantly altered by one or more treatment. Of the 58, 27 could not be unequivocally identified but 31 were named. A full description of the metabolite analysis will be published elsewhere. Here, metabolite data were examined for concordance with the pathways in which the protein enzymes were altered in abundance. The analysis used MapMan tuned for the soybean proteins that were known enzymes (Goffard and Weiller, 2006) or their Arabidopsis (Arabidopsis thaliana) orthologs (Thimm et al., 2004). MapMan showed the enzymes in pathways with their metabolites.

Protein-Protein Interactome Analysis
A Linux-based software, InParanoid (Remm et al., 2001), was used to detect orthologs between Arabidopsis and soybean. InParanoid works on the NCBI-based BLAST server, where 100% orthology is assigned for an E-value of 0.01 or lower. The first round of the program used BLOSSUM45 as the scoring matrix. The second and third rounds used BLOSSUM62 and BLOSSUM80, respectively. Proteins with greatest orthology (percentage similarity) between the two species were used for the analysis. Protein annotations for the orthologs of soybean were imported from http://www.biomart.org, which contained curated information for Arabidopsis proteins from The Arabidopsis Information Resource and Ensemble. Proteins of soybean that showed clear orthology to Arabidopsis were superimposed onto the Arabidopsis interactome kindly provided by Dr. Matt Geisler. Cytoscape 2.5.1 was used to build the interactome.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Table S1. The predicted interactome viewed in Figure 6.
Supplemental Table S2. Primers used in reverse transcription-PCR.

ACKNOWLEDGMENTS
We thank Dr. P. Gibson, O. Myers Jr., and M. Schmidt for assistance with germplasm development and maintenance from 1991 to 2000 and Gary and Mary Kinsel for assistance with protein analysis and interpretation. The support of the Southern Illinois University, Carbondale, College of Agriculture and Office of the Vice Chancellor for Research to D.A.L., A.J.A., and M.J.I. is appreciated. We thank the DOE Community Joint Sequencing Program for release of the whole genome shotgun reads and scaffolds.