Overlap of Proteome Changes in Medicago truncatula in Response to Auxin and Sinorhizobium meliloti 1[W][OA]

We used proteome analysis to identify proteins induced during nodule initiation and in response to auxin in Medicago truncatula . From previous experiments, which found a positive correlation between auxin levels and nodule numbers in the M. truncatula supernodulation mutant sunn (supernumerary nodules), we hypothesized (1) that auxin mediates protein changes during nodu- lation and (2) that auxin responses might differ between the wild type and the supernodulating sunn mutant during nodule initiation. Increased expression of the auxin response gene GH3 : b -glucuronidase was found during nodule initiation in M. truncatula , similar to treatment of roots with auxin. We then used difference gel electrophoresis and tandem mass spectrometry to compare proteomes of wild-type and sunn mutant roots after 24 h of treatment with Sinorhizobium meliloti , auxin, or a control. We identiﬁed 131 of 270 proteins responding to treatment with S. meliloti and/or auxin, and 39 of 89 proteins differentially displayed between the wild type and sunn . The majority of proteins changed similarly in response to auxin and S. meliloti after 24 h in both genotypes, supporting hypothesis 1. Proteins differentially accumulated between untreated wild-type and sunn roots also showed changes in auxin response, consistent with altered auxin levels in sunn . However, differences between the genotypes after S. meliloti inoculation were largely not due to differential auxin responses. The role of the identiﬁed candidate proteins in nodule initiation and the requirement for their induction by auxin could be tested in future functional studies. Can Proteome Analysis Determine Whether the Difference in Auxin Content between the Wild Type and sunn Mutant Explains

Many legume species form symbioses with nitrogenfixing bacteria called rhizobia. The symbiosis starts with the exudation of flavonoids by the legume host, which stimulate the synthesis of so-called Nod factors by the bacterial partners. Nod factors are lipochitin oligosaccharides that are perceived by plant receptors and trigger a series of events leading to bacterial invasion and development of a nodule. Inside the nodule, rhizobia convert atmospheric nitrogen into ammonia, which is exported to the plant in exchange for carbohydrates.
The development of a nodule starts with the reinitiation of cortical and pericycle cell divisions in the root in the zone of root hair emergence. In most deter-minate legumes, like Lotus japonicus or soybean (Glycine max), cortical cell divisions start in the outer cortex and lead to the formation of a nodule with a determinate meristem. In indeterminate legumes, like Medicago truncatula or pea (Pisum sativum), cortical cell divisions start in the inner cortex and often occur simultaneously to cell divisions in the pericycle located opposite xylem poles (Timmers et al., 1999). A small nodule primordium is formed that later differentiates into a nodule and is invaded by rhizobia (Kinkema et al., 2006).
Nod factors have many functions during nodule initiation, one of which is thought to be an indirect effect on plant hormone signaling and transport (Ferguson and Mathesius, 2003;Roudier et al., 2003;Kondorosi et al., 2005). Auxin is likely to be an important regulator of nodulation. Auxin is crucial in plant development for the regulation of cell division and elongation, embryogenesis, and meristem formation in plants. The location of auxin maxima in the plant, regulated in part through polar auxin transport, is essential for the correct placement of meristems and organs (Blilou et al., 2005;Heisler et al., 2005). Auxin is mainly synthesized in the young leaves, and transported to the root tip in an acropetal direction, which reverses at the root tip. Auxin transport is facilitated by auxin importers, including AUX1 and PGPs (P-glycoproteins), and auxin exporters, including PIN (pin-formed) proteins and other PGPs (for a recent review, see Kramer and Bennett, 2006). Polar auxin transport can be inhibited by synthetic auxin transport inhibitors, for example, N-(1-naphthyl) thalamic acid, or internal regulators, for example, flavonoids (Brown et al., 2001). Auxin transport and localization are also important for nodule development, although the exact role of auxin during nodulation is not fully understood.
Nodule initiation in indeterminate legumes is preceded by an inhibition of acropetal auxin transport at the site of nodule initiation (Mathesius et al., 1998;Wasson et al., 2006), which can also be triggered by application of purified Nod factors (Boot et al., 1999). Just before and during nodule initiation, auxin-responsive reporter gene expression is localized in the dividing inner or outer cortical cells in indeterminate (white clover [Trifolium repens] and M. truncatula; Mathesius et al., 1998;Huo et al., 2006) and determinate (L. japonicus) legumes (Pacios-Bras et al., 2003), respectively. Moreover, an auxin-inducible A2-type cyclin, is induced by Nod factors and is located in dividing cells of a nodule primordium (Roudier et al., 2003), indicating that one of the main roles of auxin during nodulation is in the regulation of mitosis (Kondorosi et al., 2005). Application of N-(1-naphthyl)thalamic acid or the auxin action inhibitor p-chlorophenoxyisobutyric acid, but also high concentrations of auxin, led to reduced nodule numbers in M. truncatula, whereas treatment with low external auxin concentrations increased nodule numbers (van Noorden et al., 2006), suggesting that optimal auxin levels and auxin action are required for nodule development. In addition, it was shown that silencing of PIN genes in M. truncatula inhibits nodulation (Huo et al., 2006). Both MtPIN (Huo et al., 2006) and MtLAX (a homolog of AUX1) expression was observed in developing nodules, possibly to channel auxin into a developing nodule primordium (de Billy et al., 2001).
Nodule numbers are strictly controlled by the plant host via a long distance ''autoregulation '' mechanism (Caetano-Anollès and Gresshoff, 1991). Autoregulation of nodulation (AON) is a mechanism by which a plant inhibits further nodule formation on a root system after the first few nodules have been formed. AON is regulated by a Leu-rich repeat receptor-like kinase acting in the shoot, which has been identified in several legumes (Krusell et al., 2002;Nishimura et al., 2002;Searle et al., 2003;Schnabel et al., 2005). It is thought that inoculation of the root with rhizobia triggers an (unknown) signal that moves to the shoot, activates autoregulation, and then causes transport of another unidentified signal to the root, where it inhibits further nodules from forming (Caetano-Anollès and Gresshoff, 1991;Oka-Kira and Kawaguchi, 2006).
Studies in supernodulating mutants have suggested that auxin is most likely to be a positive regulator required for nodule development, at least in indeterminate legumes. Two supernodulating mutants of M. truncatula, skl (an ethylene-insensitive mutant with root-controlled increases of nodule numbers; Penmetsa and Cook, 1997;Prayitno et al., 2006b) and sunn (supernumerary nodules, an AON mutant ;Schnabel et al., 2005), show increased expression of the auxin response gene GH3 in inoculated roots compared to the wild type (Penmetsa et al., 2003). Direct measurements of auxin (indole-3-acetic acid [IAA]) content in sunn have demonstrated approximately 3-fold increased levels of auxin in the root segment susceptible to nodulation, compared to the wild type, both before and after inoculation of roots with rhizobia. sunn also shows approximately 3 times increased levels of auxin transport from the shoot to the root (van Noorden et al., 2006). These increased levels of auxin transport and content correlate with increased numbers of nodules in sunn roots. The skl mutant of M. truncatula has increased nodule numbers at the site of the first inoculation, and shows increased local auxin transport and increased PIN gene expression at the inoculation site after 24 h (Prayitno et al., 2006b).
In addition, it was shown that the transport of auxin from the shoot to the root is involved in the autoregulation mechanism. Inoculation of wild-type roots with rhizobia caused a decrease of auxin loading from the shoot to the root within 24 h, whereas this long-distance inhibition of auxin transport did not occur in the sunn mutant (van Noorden et al., 2006). This suggests that certain levels of auxin are necessary for nodule development and that a lack of auxin after autoregulation could be limiting nodule numbers.
In this study we used proteome analysis to identify proteins involved in nodule initiation that may also be regulated by auxin, to extend our knowledge on the role of auxin during nodule initiation. We hypothesized (1) that auxin has a positive role in nodule initiation and (2) that the difference in auxin content between the wild type and sunn mutant could (partially) explain their nodulation phenotypes. First, we characterized local expression of GH3:GUS preceding nodule initiation to pinpoint the best time point for analysis. We then used proteome analysis as a tool to reveal broad differences or similarities in protein accumulation that could test the previous hypotheses, and to identify proteins involved in auxin responses during nodulation. To test hypothesis 1, we compared the proteomes of wild-type root segments corresponding to the inoculation zone 24 h after treatment with either Sinorhizobium meliloti, the symbiont of M. truncatula, IAA, or a sham control. To test hypothesis 2, we compared proteomes of wild-type and sunn root segments following the same treatments as above. Our results show increased auxin localization and a large overlap in proteins induced by rhizobia and auxin at 24 h after inoculation (ai), and support a positive role for auxin during nodule initiation in M. truncatula.
examined GH3:GUS expression in hairy roots of wildtype composite plantlets after inoculation with S. meliloti and auxin. At least 20 roots were examined for each treatment, with similar results. GH3:GUS expression in uninoculated wild-type roots was located mainly in the vascular bundle (Fig. 1, A and B). Root tips (including the meristem and root cap) were also stained in about 30% of roots. Treatment with 1 mM IAA for 24 h caused GH3:GUS expression to spread to the cortical cells as well as the vascular bundle along the whole root (Fig. 1,C and D). After spot-inoculation of the roots with S. meliloti, increased GH3:GUS expression could be seen within 24 h ai in vascular and cortical cells around the inoculation site (Fig. 1, E and F). This staining was located in a patch of several millimeters in length in approximately 60% of the roots, but in 40% almost the whole roots showed blue staining. After 48 h, cortical and pericycle cells could be seen that had just divided. These cells showed strong GH3:GUS expression in contrast to nondividing cells (Fig. 1, G and H). In general, auxin was located in vascular and dividing cells in the root. At 24 h after treatment, there was a similar strong induction of GH3:GUS expression in most cell types in response to auxin and inoculation, indicating that increased auxin levels inside the root at this time point could be involved in preparation for nodule initiation. Thus, the 24-h time point was chosen for subsequent proteome analysis.

Proteome Analysis
Proteome analysis was used to compare roots treated with rhizobia (24 h ai), with 1 mM auxin for 24 h, or a control, all at the zone of emerging root hairs. Four biological repeats of 50 root segments each were collected of each treatment. Analysis by difference gel electrophoresis (DIGE) was performed that included a dye swap between Cy3 and Cy5 (Supplemental Table S1). Two-dimensional (2D) gels of proteins labeled with CyDyes showed approximately 3,800 spots across all 12 gels (Fig. 2) in a pI range of 4 to 7. All spots were statistically compared for treatment and genotype effects by two-way ANOVA. Overall, 297 of the 3,800 spots (approximately 7.5%) showed statistically significant changes in response to either treatment or genotype (P , 0.05, two-way ANOVA; Table I; Supplemental  Tables S2 and S3). The average fold change of all significant changes was 2.74 (SD 5 1.69, n 5 649), with a range of 1.17 to 13.1.
To identify differentially accumulated proteins, all samples were combined for a 2D gel stained with Coomassie Brilliant Blue, resolving approximately 1,000 spots (Fig. 2). We could confidently match 201 of the 297 differentially accumulated proteins between DIGE and Coomassie gels. The remaining 96 proteins were too low in quantity to detect on the Coomassie gels due to the dye's lower staining sensitivity. Using MALDI-TOF/TOF tandem mass spectrometry (MS/ MS), we could identify 140 of the 201 differentially accumulated protein spots excised from Coomassie gels ( Fig. 2; Supplemental Table S4). Of the identified protein spots, 21 were successfully matched to two or more tentative consensus (TC) sequences, suggesting that these spots contained two or more proteins, which could not be resolved on the 2D gel, as observed in other studies (Watson et al., 2003). However, 15 of the 21 proteins matched two TCs with the same annotated function, and only six proteins matched to two or more TCs with different annotated functions. The identified proteins matched to 108 different TCs; 28 were found twice or more, indicating the existence of different protein isoforms.

Changes in Protein Profiles during Inoculation with Rhizobia and Auxin Treatment
Our first hypothesis was that auxin mediates some of the protein changes during the early stages of nodule initiation. To test this hypothesis and to identify proteins potentially induced by auxin during nodulation, we compared the proteomes of root segments treated with S. meliloti, 1 mM IAA, and a control after 24 h.
Overall, the accumulation level of 270 protein spots was significantly changed (P , 0.05, two-way ANOVA) between the three treatments (Supplemental Table S2); approximately 50% of the spots could be identified (Table II).
In the wild-type roots, the accumulation level of 174 protein spots was significantly changed (P , 0.05, Student's t test) in response to S. meliloti inoculation and 208 in response to auxin (P , 0.05, Student's t test), compared to the controls. Interestingly, the majority (154) of proteins that changed after inoculation also changed in abundance after treatment with auxin, whereby both treatments resulted in similar accumulation levels of the same proteins (Table I; Supplemental Table S2). Figure 3 shows the overlaps and differences in protein changes between the different treatments in wild-type roots according to their functional classification in the Kyoto Encyclopedia of Genes and Genomes (KEGG; Kanehisa et al., 2004).
In the sunn mutant, rhizobia inoculation caused the accumulation level of 103 protein spots to significantly change (P , 0.05, Student's t test) compared to the control. Auxin altered the accumulation of 169 protein spots (P , 0.05) compared to the control. Similar to the wild type, the majority of proteins showed similar changes in response to auxin and rhizobia inoculation, with the accumulation levels of 92 protein spots changing in both the inoculated roots and the roots treated with auxin ( Fig. 3; Table II; Supplemental Table S2). Protein changes in sunn were in general very similar to changes in the wild type in response to auxin and S. meliloti (Fig. 3). In both genotypes, more proteins were induced by auxin than by S. meliloti. Overall, the results showed that the majority of proteins that were up-or down-regulated in response to S. meliloti at 24 h were also up-or down-regulated by a 24-h auxin treatment, supporting our first hypothesis.

Comparison of Protein Changes between the Wild Type and sunn
Our second hypothesis was based on two main differences between wild-type and sunn roots: sunn roots were shown to have higher levels of auxin in the inoculation zone and this was accompanied by twice the number of nodules in that zone (van Noorden et al., 2006), Therefore, we expected that there would be an interaction between the effect of auxin and the effect of S. meliloti inoculation between the wild type and sunn. We broke this hypothesis down into three subquestions.
(1) Do the genotypes differ in their response to external auxin? We expected that both genotypes have in general a similar response to auxin. ANOVA for treatment effects (Table I; Supplemental Table S2) showed that there were 223 auxin-inducible proteins, of which 142 (63%) were induced/reduced similarly and significantly in both the wild type and sunn. An additional 58 proteins were only altered in the wild type by auxin, and 23 were only changed in sunn by auxin. However, in all of those cases, the proteins showed a similar, although not statistically significant, trend for changed levels in the other genotype. Overall, these data suggest that both the wild type and sunn respond van Noorden et al. Figure 2. Proteome reference map for M. truncatula roots. Alignment of a DIGE (A) with the Coomassie gel (B) and the resulting reference map (C). A is an example of a (false-color) DIGE-labeled gel showing differential expression of two samples and an internal control, labeled with Cy3, Cy5, and C2, respectively. White protein spots indicate proteins not differentially displayed; colored spots are induced in one of the treatments (e.g. spots 308,405,141,226,178). The labeled proteins were marked as landmarks on both gels to aid in the alignment in DeCyder software. C, DIGE gel (Cy2 channel only) showing the position of the differentially accumulated and identified proteins listed in Supplemental Table S1. Proteins were separated across an isoelectric focusing range of pI 4 to 7. similarly to auxin in general, but that more proteins are inducible by auxin in the wild type than in sunn, possibly because of already higher endogenous auxin levels in sunn.
(2) Are the proteins that differ between wild-type and sunn roots also auxin inducible? To analyze the genotype effects between the wild type and sunn, we calculated ANOVAs for genotype effects only (Table  III; Supplemental Table S3). We found 27 significant changes between genotypes in control roots, 32 in rhizobia-inoculated roots and 29 in auxin-treated roots. Of 27 proteins that differed significantly between wild-type control and sunn control roots, 19 were also significantly affected by auxin treatment (Table I; Supplemental Table S3). Of those 19, 15 were inducible by auxin in both genotypes and four only in the wild type. In addition, more than 50% of proteins differing between genotypes in rhizobia-inoculated and auxintreated roots could also be induced by auxin in both genotypes. These data support the idea that the differing auxin levels in sunn and the wild type in the examined root zone could lead to changes in protein accumulation level in control (untreated) roots.
(3) Are the proteins that differ between S. melilotiinoculated wild-type and sunn roots also different in their auxin response? Of 196 rhizobia-induced proteins, 81 showed similar accumulation patterns between wildtype and sunn roots after inoculation. However, there were 93 proteins that were only significantly altered in the wild type and 22 that were only changed in sunn after inoculation with S. meliloti (Table II; Supplemental Table S2). To find out if these differences could be due to the altered auxin levels in sunn and the wild type, we examined whether these proteins also showed a different response to auxin in the two genotypes. Of the 93 proteins only induced by rhizobia in the wild type, 36 showed similar responses to auxin in both genotypes, 40 showed significant induction by auxin in the wild type only, and three were significantly induced by auxin in sunn only. Of the 22 proteins only induced by rhizobia in sunn, 14 showed similar responses to auxin in both genotypes, five were induced by auxin only in sunn, and two were induced by auxin only in the wild type. Overall, the majority of proteins that were specifically induced by rhizobia in only one genotype did not differentially change in response to auxin between genotypes.

DISCUSSION
Our study was based on previous observations of a positive correlation between auxin content and auxin transport with the numbers of nodules formed on the roots of M. truncatula (Prayitno et al., 2006b;van Noorden et al., 2006). The hypotheses tested in this study were that auxin is a positive regulator of nodule initiation and that the difference in auxin content between the wild type and sunn mutant could partially explain their nodulation phenotypes. Our primary tool in detecting responses to rhizobia and auxin was proteome analysis.

Proteome Analysis
Three main large-scale tools for the analysis of gene expression changes are commonly used in biological systems, proteomics, transcriptomics, and metabolomics. Using the DIGE technology, proteome analysis can be carried out similarly to a microarray experiment in that two samples are compared on one gel by analyzing the ratio of two fluorescent labels between two samples for each protein (Ü nlü et al., 1997). Together with the use of a pooled internal standard that combines all samples, this has been shown to reduce gel-to-gel variation and increase statistical confidence in the results (Alban et al., 2003). Compared to other proteomics studies that have analyzed the effect of S. meliloti inoculation on M. truncatula roots using either silver or Coomassie staining, our analysis, using DIGE, showed approximately 10-fold more significant protein accumulation changes (Bestel-Corre et al., 2002;Prayitno et al., 2006a). Compared to a microarray analysis of M. truncatula that included a 24-h time point and found approximately 500 gene expression changes (Lohar et al., 2006), a change of approximately 200 proteins in our analysis is comparable, especially considering that only approximately 3,800 proteins were analyzed compared to 6,000 cDNAs in that study. The range and average of ratios in abundance were also Differentially displayed protein spots in both rhizobia and auxin treatment.
van Noorden et al.   Lohar et al. (2006). While there were some similarities in the proteins induced in the microarray and proteomics studies, most of the changes were found to affect different genes, and this could be due to the different half-lives of the transcripts and proteins in the cell as well as the slightly different growth conditions in the different experiments. Of 89 proteins identified in our study as significantly changed by rhizobia at 24 h (in wild-type roots), 28 matched to genes with the same annotated function that were also found to be affected in a transcriptome study (Lohar et al., 2006; Table II), whereas 61 proteins (matching to 44 TCs) were not found in the transcriptome study. In contrast to transcriptome analyses, proteomics was able to reveal the accumulation of many protein isoforms from one transcript, thus giving a more accurate picture of the actual gene products. For example, 27 individual PR10 and related Pprg2 protein spots were identified in the roots, which matched to nine different TCs. These were overall the most abundant proteins. However, transcriptome analysis only found one PR10 transcript in M. truncatula roots at 24 h, and this was not affected by rhizobia (Lohar et al., 2006). In addition, our study found multiple isoforms of metabolic enzymes, for example, malate dehydrogenase, Fru-bisphosphate aldolase, and glyceraldehyde-3-P dehydrogenase (Table II), where only one gene appeared to be affected at the transcript level (Lohar et al., 2006). In general, the up-or downregulation of the proteins and genes was similar in both studies. Compared to transcriptome analysis, our study showed a lack of transporters and signaling components. e P values (Student's t test) and ratios for proteins differentially accumulated between wild-type control and wild-type rhizobia-inoculated roots.
f P values (Student's t test) and ratios for proteins differentially accumulated between wild-type control and wild-type auxin-treated roots. g P values (Student's t test) and ratios for proteins differentially accumulated between sunn control and sunn rhizobiainoculated roots. h P values (Student's t test) and ratios for proteins differentially accumulated between sunn control and sunn auxin-treated roots.
i Protein spot matched successfully to multiple TCs with the same predicted identity, as shown in Supplemental Table S4.
We used GH3:GUS expression in hairy roots to compare timing and location of auxin response to auxin and rhizobia. Some variation in GUS staining was observed between individual roots. Similarly, this variation was observed in white clover fully transformed with GH3:GUS (Larkin et al., 1996), and could be due to differences in transgene numbers or expression in individual hairy roots and possible effects of the rol genes on auxin sensitivity (Nilsson and Olsson, 1997). We confirmed that GH3:GUS expression is located in cortical cells preceding cell division, and in the first dividing pericycle and cortical cells of a nodule primordium (24-48 h ai). This was similar to observations in white clover (Mathesius et al., 1998) and L. japonicus (Pacios-Bras et al., 2003) and in M. truncatula hairy roots transformed with the DR5 auxin reporter showing expression in nodule primordia 72 h ai (Huo et al., 2006). Thus, it appears that auxin responses occur at the site of nodule initiation. This conclusion was supported by the proteomics results that showed that 89% of the proteins affected by inoculation with S. meliloti at 24 h also changed in response to auxin. In addition, the trend toward up-or down-regulation of these proteins was very similar after both treatments. At the moment it is impossible to determine the exact internal auxin concentration in the cortical cells undergoing division, and how that correlates to the concentrations of auxin induced by the external application of 1 mM auxin and the time of treatment with auxin. It is also difficult to find out if the responses we measured in root segments reflect responses in specific cell types that could differ between auxin and S. meliloti treatment. For example, S. meliloti causes cell divisions in pericycle and cortical cells (compare with Fig. 1H; Timmers et al., 1999), which require auxin signaling (Roudier et al., 2003;Kondorosi et al., 2005); external auxin treatment stimulates pericycle division and lateral root initiation Casimiro et al., 2001). However, in general, the results suggest that increased auxin responses in the root could be responsible for a number of the changes in gene expression following inoculation with rhizobia 24 h ai. In the future this hypothesis could be further tested by comparing auxin responses in inoculated roots of auxin response mutants; however, none has been described yet in M. truncatula.
Most of the proteins that were affected by both auxin and S. meliloti at 24 h included metabolic enzymes, proteins involved in energy metabolism and protein processing (Fig. 3). These changes could be due to a general requirement for metabolites used for the preparation for cell division induced by S. meliloti and auxin. It is interesting to note that only one protein that changed in response to auxin treatment (and only in the sunn mutant) was previously identified as an auxin-inducible protein, a 2,4-D-inducible glutathione S-transferase, although this protein was actually down-regulated by auxin. This protein was also found to be downregulated in response to early mycorrhizal infection in M. truncatula in a study that also found other evidence for auxin-induced protein changes during mycorrhization (Amiour et al., 2006). The microarray study by Lohar et al. (2006) showed an increase in expression of two auxin-inducible genes and reduction of two auxin-repressed proteins at 24 h ai. Neither these nor other known auxin response proteins were identified in our study, and this is possibly due to their low accumulation level. Other proteins of interest that could play a role during nodulation in response to auxin are 12-oxophytodienoic acid 10,11-reductase (OPR), isoflavone reductase, a late-embryogenesis-like protein, a number of redox-related proteins, and a large group of PR10 and Pprg2 proteins.
OPR converts 12-oxophytodienoic acid into 3-oxo-2-(2#-Z-pentenyl)-cyclopentane-1-octanoic acid, a precursor of jasmonic acid (JA; Schaller and Weiler, 1997). JA inhibits Nod factor signal transduction by modulating the calcium signature of responding root hair cells and prevents early nodulation gene accumulation, resulting in the inhibition of nodule formation (Sun et al., 2006). In addition, the application of a volatile derivative of JA, methyl jasmonate, to the shoot of L. japonicus inhibits nodule formation, suggesting that JA plays a Table III. Changes in protein accumulation levels between the wild type and sunn The table shows the identified proteins that were differentially accumulated in response to genotype effects. The table shows separate effects between the wild type and sunn for proteins differing in their abundance between genotypes in the controls, the rhizobia-inoculated roots, and the auxin-treated roots. Positive value indicates up-regulation in sunn; negative value indicates down-regulation in sunn compared to wild type; ns, not significant. e P values (Student's t test) and ratios for proteins differentially accumulated between wild-type and sunn control roots. f P values (Student's t test) and ratios for proteins differentially accumulated between wild-type and sunn rhizobia-inoculated roots.
g P values (Student's t test) and ratios for proteins differentially accumulated between wild-type and sunn auxin-treated roots. h Protein spot matched successfully to multiple TCs with the same predicted identity as shown in Supplemental Table S4. part in long-distance regulation of nodule numbers (Nakagawa and Kawaguchi, 2006). OPR was strongly induced in response to S. meliloti and auxin, suggesting that JA synthesis is activated during the early stages of nodulation, possibly to limit the numbers of nodules. OPR was more strongly induced in sunn than in the wild type, and this could suggest that JA is locally induced as part of the nodule initiation process.
Isoflavone reductase, an enzyme of the isoflavonoid biosynthetic pathway, was induced by auxin and S. meliloti. Various members of the flavonoid pathway are induced during nodulation (El Yahyaoui et al., 2004;Lohar et al., 2006), and this could be due to a requirement for flavonoids as Nod gene inducers or regulators of auxin transport (Subramanian et al., 2006;Wasson et al., 2006). In Arabidopsis (Arabidopsis thaliana), it has been suggested that a feedback mechanism exists among auxin accumulation, flavonoid induction, and subsequent regulation of auxin transport by flavonoids (Peer et al., 2004). The induction of flavonoid biosynthetic enzymes by auxin treatment in our study supports those findings.
Our data also showed an effect of auxin and rhizobia on proteins involved in redox control, which has emerged as an important regulator of cell division (Jiang and Feldman, 2005). The most important redox couples include ascorbic acid/dehydroascorbate and reduced/oxidized glutathione. In addition, thioredoxin, glutaredoxin, and peroxidases also play a role in redox control. It is likely that auxin might partially influence cell division through its effects on redox control, as changes in the auxin distribution altered both the patterns of redox state and cell division in maize (Jiang et al., 2003). Our study showed evidence for the induction of redox-related proteins by auxin and a connection of redox control with nodulation: Four enzymes contributing to a more reduced redox state were upregulated in auxin-and in S. meliloti-treated roots, including glutathione reductase, monodehydroascorbate reductase (MDAR), a peroxidase, and a thioredoxindependent peroxidase. MDAR and a peroxidase were also found to be induced at 24 h in a microarray study (Lohar et al., 2006). Ascorbate peroxidase was induced in meristems compared to nonmeristematic tissue in M. truncatula (Holmes et al., 2006). Altogether the results suggest that auxin might play a role in inducing enzymes that form a more reduced redox state, which could be important for nodule initiation.
Various biotic and abiotic conditions have been shown to induce members of the JA biosynthesis pathway, including OPR (Sasaki-Sekimoto et al., 2005); the flavonoid pathway (Stafford, 1997); and changes in redox control (Mittler et al., 2004). Additionally, there is evidence for cross talk between these pathways; for example, JA responses mediate induction of antioxidant enzymes (Sasaki-Sekimoto et al., 2005). Different members of these stress response pathways are induced in response to different pathogens and environmental stresses. Their induction during nodulation might reflect the possibility that symbiotic rhizobia are recognized as pathogens initially, and that selective down-regulation of defense responses allows symbiosis under appropriate conditions (Mithö fer, 2002). A comparison of the individual protein isoforms induced by different stimuli will be necessary in an individual plant species to identify specific members that might be responsible for certain responses.
The PR10 protein class was the second highest class (18%) of identified proteins that changed accumulation after both rhizobia inoculation and auxin treatment. A microarray study in M. truncatula showed several PR10 proteins to be differentially regulated at later (3-10 d ai) stages of nodulation (El Yahyaoui et al., 2004), and some of these proteins were also differentially accumulated in meristems of M. truncatula (Holmes et al., 2006). The function of the PR10 class of proteins is so far unknown. They were originally described as pathogen-related proteins (van Loon et al., 2006). However, the PR10 class of proteins has key roles in many plant developmental programs as PR10 is also induced by auxin (Poupard et al., 2001), salicylic acid, JA (McGhee et al., 2001), abscisic acid (Colditz et al., 2004), ethylene (Prayitno et al., 2006a), and biotic and abiotic stress (Hashimoto et al., 2004). PR10 proteins are also known to bind fatty acids, flavonoids, steroids, cytokinins, and brassinosteroids (Mogensen et al., 2002;Markovic-Housley et al., 2003), and could thus link different hormone responses. For example, the correct ratio of auxin and cytokinin signaling is likely to be important in nodule initiation (Gonzalez-Rizzo et al., 2006;Murray et al., 2007;Tirichine et al., 2007). In Brassica napus, overexpression of a pea PR10 protein caused elevated cytokinin and decreased ABA levels, suggesting that changes in expression of PR10 could have an effect on hormone-regulated developmental processes (Srivastava et al., 2006). From our analysis it appears that the specific isoforms of PR10 might have distinct roles as they were very specifically altered in response to the different treatments and genotypes. A PR10 protein (MtN13) was previously found to be expressed specifically during nodulation; thus, this class of proteins might have evolved specific functions for different developmental processes (Gamas et al., 1998). Furthermore, six isoforms of PR10 were found to be induced by the oomycete pathogen Aphanomyces euteiches in M. truncatula, in a proteome study (Colditz et al., 2004); only three showed coinduction by either drought or ABA. The three PR10 isoforms specifically induced by the pathogen were also induced in our study by S. meliloti and auxin (represented in our gels by spots 353, 387, and 356). Thus, a particular combination of isoforms of PR10 proteins might be characteristic for responses to certain elicitors.
It was surprising to find only a few proteins induced by rhizobia but not by auxin, for example, infectionrelated proteins. A possible explanation could be that only a few cells are involved in infection of the whole root segment harvested, and, thus, changes in protein accumulation in those cells would be diluted. Interestingly, we found three ATPases that were induced in S. meliloti but not auxin in the wild type. Visualization of extracellular ATP in M. truncatula was shown to be high in growing root hairs (Kim et al., 2006). Redirection of root hair growth is part of the infection process of rhizobia, and the induction of the three ATPases could thus indicate a role in root hair curling and infection of rhizobia.
Can Proteome Analysis Determine Whether the Difference in Auxin Content between the Wild Type and sunn Mutant Explains Their Nodulation Phenotypes?
Roots of the sunn mutant were shown to have approximately 2-to 3-fold increased auxin levels at the zone of nodule initiation (van Noorden et al., 2006), which are accompanied by higher long-distance auxin transport in sunn (van Noorden et al., 2006) and increased GH3 expression levels during nodulation (Penmetsa et al., 2003). Roots of the sunn mutant are also shorter than those of the wild type (Penmetsa et al., 2003;van Noorden et al., 2006), and this could also be a result of higher auxin levels, as high levels of auxin inhibit root growth, including in M. truncatula (van Noorden et al., 2006).
Overall, there was a high similarity between wildtype and sunn proteomes, and genotype effects were less numerous than treatment effects. Compared to 270 proteins with significant changes in response to one of the treatments (Supplemental Table S2), there were only 89 proteins that showed significant genotype effects (Supplemental Table S3). Contrary to our expectation, sunn did not show a more extensive response to either auxin or rhizobia than the wild type. This is in contrast to a microarray study that showed more extensive gene expression changes in a sunn mutant than in the wild type during nodulation, although this study examined later time points (3-10 d ai) compared to ours (El Yahyaoui et al., 2004). There was also no evidence that the proteins induced by S. meliloti or auxin in the wild type but not in sunn were present at higher background levels in sunn control roots due to their higher internal auxin levels. It is possible that proteome analysis is not sensitive enough to determine differences in accumulation levels of proteins that might exist between the wild type and sunn, especially if those changes only occur in a few cells out of the total root segments analyzed here. Alternatively, there might be feedback mechanisms that are activated in sunn to compensate for high auxin levels.
The response to auxin was very similar between the wild type and sunn in the proteome analysis; it was mirrored by very similar root growth responses to auxin in the wild type and sunn (van Noorden et al., 2006). Some notable differences were observed, including the expression of two protein disulfate isomerases, OPR, isoflavone reductase, a Gly-rich RNA-binding protein, albumin 2, two PR10 proteins, and a thioredoxin-dependent peroxidase. The Gly-rich RNA-binding protein (TC93436) was a protein specifically induced by auxin but also significantly higher in sunn than in the wild type.
There were also some interesting differences in protein accumulation between the wild type and sunn in response to inoculation with S. meliloti; these included OPR, albumin 2, a trypsin inhibitor, six PR10 proteins, and the thioredoxin-dependent peroxidase.
As discussed above, the differential accumulation level of PR10 proteins and the thioredoxin-dependent peroxidase could indicate altered stress response and/ or redox control during nodulation. Whereas the PR10 proteins were more highly accumulated in the wild type, the thioredoxin-dependent peroxidase was present at higher levels in sunn under all three treatment conditions. It would be interesting in the future to assess redox state in the sunn mutant to find out if this might contribute to its supernodulation phenotype.

CONCLUSION
Using DIGE combined with MS/MS, we identified 131 of 270 proteins differentially displayed after treatment of M. truncatula roots with S. meliloti or auxin, and 39 of 89 proteins differentially displayed between the wild type and the sunn mutant, the highest number of proteins identified from M. truncatula roots during nodule initiation. We found evidence for coregulation of a large number of proteins by rhizobia and auxin, suggesting that these proteins could be regulated by auxin changes during the early stages of nodulation. Protein accumulation patterns also indicated that some of the differences between the wild type and sunn could be due to their different internal auxin levels, although in general both genotypes showed very similar responses to rhizobia and to auxin. The differentially accumulated candidate proteins could be used in future studies, for example, using RNA interference or overexpression, to determine their function during nodulation.

Plant Growth and Treatments
Medicago truncatula 'Jemalong' genotype A17, which was used as the wildtype control, and its derivative sunn mutant (Penmetsa et al., 2003;Schnabel et al., 2005) were used in this study. Seeds were scarified with sandpaper and surface sterilized for 10 min with 6.25% (w/v) hypochlorite, followed with six washes with sterile water. Additional sterilization included immersion in a solution of 200 mg L 21 augmentin for 6 h at 29°C and washing six times with sterile water. Sterilized seeds were spread on petri dishes containing 0.8% (w/v) agar (Grade J3; Gelita Pty Ltd.) and moistened with a drop of sterile water to prevent the seeds from drying. Seeds were then germinated by incubating the plate upside-down in the dark overnight at 29°C. The germinated seeds were transferred to large petri dishes (diameter 15 cm, 150 mL) containing nitrogen-free Fåhraeus agar medium (Fåhraeus, 1957), with the seed placed at a distance of 4 cm from the top of the plate. Plates were kept vertical and the bottom half of each plate was sealed with Nescofilm (Bando Chemical Ind. Ltd.), with the sides covered for two-thirds with black paper to reduce light intensity around the root. An aluminum foil spacer was placed between the lid and the petri dish to allow for air exchange. Plates were incubated in a growth chamber at 21°C for a 16-h-day/8-h-night cycle with a photon flux density of 100 mmol m 22 s 21 .
The rhizobia strain used for inoculating the roots was Sinorhizobium meliloti strain 1021. S. meliloti were grown on agar plates containing Bergensen's modified medium (BMM; Rolfe et al., 1980) for 3 d at 28°C. For plant inoculations, 5 mL of liquid BMM was inoculated from a single bacterial colony and incubated at 28°C overnight in a shaker. The bacterial culture was diluted with liquid BMM medium to an optical density of 0.1 at the wavelength of 600 nm. Three-day-old seedlings were flood inoculated at the root tip with 10 mL of the rhizobial culture. Control roots were sham inoculated with the same amount of sterile liquid BMM medium. For proteome analysis, the inoculation zone was marked and the root segment comprising the inoculation zone (1 cm length, 5 mm either side of the inoculation) was harvested after 24 h, and the apical 2 mm of the root tip was removed.
Auxin treatments of roots was done by placing 3-d-old seedlings of similar length onto plates containing IAA, which was dissolved as a stock solution in ethanol at 1 mM and then diluted to the correct final concentration. All control plates contained equivalent amounts of ethanol. For proteome analysis, the zone of the root corresponding to the inoculation zone of the S. melilotiinoculated roots was marked and harvested 24 h after transfer to the IAAcontaining plates (1 mM final concentration). This concentration was chosen because it showed physiological effects in M. truncatula, including stimulation of pericycle cell divisions (data not shown; van Noorden et al., 2006). Again, the root tip was removed from the analyzed root segments and the root segment was of 1 cm length, comparable to that of the inoculated roots.

Plant Transformation with the GH3:GUS Reporter Gene
For promoter:GUS analysis, a 722-bp region upstream of the ATG start codon of the GH3 gene from soybean (Glycine max; Hagen et al., 1991) was PCR amplified from genomic DNA using the following primers: GH3-F ggg gac aag ttt gta caa aaa agc agg ct and GH3-R ggg gac cac ttt gta caa gaa agc tgg gtg. The resulting PCR product was verified by sequencing and subcloned into the pKGWFS7 vector (Karimi et al., 2002) using GATEWAY technology (Invitrogen). The construct was transformed into the Agrobacterium rhizogenes ARqua1 strain (Boisson-Dernier et al., 2001) using electroporation. Transformation of M. truncatula wild-type plants was performed as described by Boisson-Dernier et al. (2001), with identical conditions as described (Wasson et al., 2006). For nodulation experiments, NH 4 NO 3 was omitted from the media and agar was replaced by Phytagel (Sigma). Hairy root plantlets were grown in a growth chamber at 25°C with a 16-h-light period at 150 mmol m 22 s 21 light intensity.
For sectioning, roots were embedded in 3% DNA grade agarose (Progen Biosciences) and sectioned on a Vibratome (1000 Plus; Vibratome Company) at 100 mm thickness. Sections were mounted on glass slides in water and examined with a Leica DMBL microscope (Leica Microsystems) and photographed with a mounted SPOT RT slider CCD camera (Diagnostic Instruments).

Protein Extraction
The root proteins were extracted using a TCA-acetone precipitation method as described before (Mathesius et al., 2001). For each sample, root segments from about 50 to 60 seedlings were harvested and immediately frozen in liquid nitrogen before extraction.

Experimental Design for the DIGE Experiment
The experiment included six different treatments (wild type or sunn, each treated with S. meliloti, auxin, or a control), and for each treatment four biological repeat samples were generated. Each ''repeat'' included approximately 50 to 60 root segments, which were grown on separate occasions. To allow for differences in CyDye intensity, two of each of the biological repeat samples were labeled with Cy3 and the other two with Cy5 labels. Each gel contained one Cy3 and one Cy5 sample, as well as an internal control consisting of a mixture of all 24 samples, which was labeled with Cy2. All possible combinations of pairwise comparisons between the samples were included, as recommended in the Amersham Biosciences Ettan DIGE User Manual. For the exact composition of samples of each of the 12 gels, see Supplemental Table S1.

Protein Quantification and Minimal DIGE Labeling of the Protein Sample
The pH of the protein samples was adjusted to pH 8 to 8.5 and the total protein concentrations of the samples were quantified using the Bradford protein assay (Bio-Rad). Bovine serum albumin over a concentration range from 0 to 1 mg mL 21 was used as a standard.
The protein samples were labeled using the CyDye DIGE Fluors (minimal dyes) for Ettan DIGE kit (Amersham Biosciences), according to the manufacturer's instructions with minor modifications. Protein samples, 100 mg each, were labeled with 400 pmol amine reactive cyanine freshly dissolved in anhydrous dimethyl formamide. The labeling reaction was incubated on ice in the dark for 30 min. The reaction was terminated by addition of 10 nmol Lys. Equal volumes of 23 sample buffer (7 M urea, 2 M thiourea, 20 mg/mL DTT, and 2% Bio-Lyte 3-10) were added to each of the labeled protein samples, and the two samples plus the internal standard were mixed.

2D Gel Electrophoresis
The combined protein samples were separated by isoelectric focusing using linear precast immobilized pH gradient gel strips of 24 cm length with a pH gradient of 4 to 7 (Amersham Biosciences). Immobilized pH gradient strips were rehydrated and focused as described previously (Mathesius et al., 2001).
The SDS-PAGE gels were 1-mm-thick, self-cast gels containing 12.5% acrylamide without gradient; the gels were cast using the EttanDaltsix system (Amersham Biosciences). The gel-cast unit was assembled using low fluorescent glass plates (Amersham Biosciences), which are compatible with the DIGE system. First-dimension strips were equilibrated as described (Mathesius et al., 2001) and placed on the second-dimension gels. SDS-PAGE was carried out at 20°C in running buffer (25 mM Tris, pH 8.3, 192 mM Gly, and 0.1% SDS) until the bromphenol blue front had reached the bottom of the gel (600 V, 10 mA, and 2.5 W per gel for the first hour; 600 V, 40 mA, and 13 W per gel for the following 4-5 h).

Gel Imaging
Immediately after running the second dimension, the DIGE-labeled proteins were visualized using a Typhoon Trio laser scanner (Amersham Biosciences). The Cy3 images were scanned using a 532-nm laser and a 580-nm band pass (BP) 30 emission filter. Cy5 images were scanned using a 633-nm laser and a 670-nm BP30 emission filter. Cy2 images were scanned using a 488-nm laser and an emission filter of 520-nm BP40. All gels were scanned at 100-mm resolution. The photo multiplier tube was set to ensure maximum pixel intensity between 40,000 and 60,000 pixels. Images were cropped to remove areas extraneous to the gel image using ImageQuant Version 5.2 (Amersham Biosciences) prior to analysis. The Coomassie-stained gel was scanned identically to the Cy5 images for matching to the DIGE gels. For presentation in the figure, it was additionally scanned at 600 dpi as a TIFF file on a flatbed scanner (UMAX Powerlook III; UMAX Technologies).

Gel Analysis
Spot detections and analyses were done using DeCyder Version 5 (Amersham Biosciences), a 2D gel analysis software package designed specifically for DIGE. The difference in gel analysis (DIA) module of the DeCyder was used for protein spot detection and quantification. The maximum number of spots was set at 10,000. Protein spots with a slope .1.5 and a volume ,3,000 were excluded to eliminate nonproteinaceous spots (e.g. dust particles with a steep slope) from further analysis. All gels were normalized before analysis using the internal Cy2-labeled standard. For the normalization, the DIA module calculates the log volume ratios of all spots on a gel against the internal standard, and creates a model histogram of spot frequency against log volume ratios with the center at a log volume ratio of zero, assuming that most spots are not differentially accumulated between samples. For the histogram selection, the scatter parameter was set to ''maximum volume'' and the threshold mode was set to 2 times the SD. The biological variation analysis (BVA) module was used for gel-to-gel matching of spots and quantitative comparisons of protein accumulation across multiple gels. Gel maps were then matched to a master map using the BVA software module; however, most abundant spots were manually checked to see if the matching was correct. Two-way ANOVA followed by Student's t tests was performed using the BVA module. The results were extracted using the XML toolbox module of the DeCyder software. To match the Coomassie-stained image to the DIGE gels, the Coomassie image was imported as a spot map into DeCyder software and matched in the BVA module after applying landmarks across the gel (see Supplemental  Fig. S1).

Colloidal Coomassie Staining
For protein identification, 2D gel electrophoresis was performed, with 500 mg of (unlabeled) proteins as starting material from A17 or sunn roots of a pooled sample of all treatments used in the DIGE analysis. Coomassie staining was done as described (Mathesius et al., 2001).

Mass Spectrometry
Protein spots of interest were excised from the preparative gels stained with Coomassie blue using sterile scalpel blades. The excised gel pieces were destained, digested, and analyzed using MALDI-TOF/TOF MS with an Applied Biosystems 4800 Proteomics Analyser as described by Prayitno et al. (2006a).

Database Searches
Mass spectra and ion data generated by MALDI-TOF MS/MS were used to search for protein identification against the M. truncatula EST database (MtGI; http://compbio.dfci.harvard.edu/tgi/cgi-bin/tgi/gimain.pl?gudb5medicago) using the Mascot Daemon Version 2.1.0 software program (Matrix Science), which at the time of analysis (February 2006) contained 226,923 entries. For peptide matching, a maximum of one miscleavage per peptide, and peptide modifications by oxidation of Met and carbamidomethylation of Cys were allowed. The peptide mass tolerance and ion mass (MS/MS) accuracy used for peptide matching were 100 ppm and 0.4 D, respectively. The confidence of peptide matches was based on the significant value of the MOWSE score (.35), the mass accuracy, number of peptide matches, and the percentage of sequence coverage. Because calculations of the expected M r and pI are unreliable from EST data as EST sequences are often incomplete, we did not compare expected and actual molecular weight and pI of the proteins. The search parameters for the single peptide matches included no miscleavage, no allowance of any modifications, a peptide charge of 1 1 , and a peptide tolerance of 50 ppm unless de novo sequences of at least eight amino acids obtained. The raw MS data of proteins that did not match significantly to the current M. truncatula EST database are available on request.

Protein Classification
Identified proteins have been assigned to a hierarchical functional classification modeled on KEGG database (Kanehisa et al., 2004) by homology relationship Weiller, 2006, 2007).
To detect if a certain functional category is statistically overrepresented in a group of proteins (as shown in Fig. 3) compared to the rest of the identified proteins, the P value for all functional categories throughout the classification was calculated using the hypergeometric distribution. This P value represents the probability that the intersection of the set of proteins belonging to the given functional classification occurs by chance.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Table S1. Labeling scheme for DIGE gels.
Supplemental Table S2. Protein accumulation levels following rhizobia and auxin treatment.
Supplemental Table S3. Protein accumulation level changes between the wild type and sunn.
Supplemental Table S4. Identity of protein spots as determined by MS.