The Lipopolysaccharide of Sinorhizobium meliloti Suppresses Defense-Associated Gene Expression in Cell Cultures of the Host Plant Medicago truncatula

In the establishment of symbiosis between Medicago truncatula and the nitrogen-fixing bacterium Sinorhizobium meliloti, the lipopolysaccharide (LPS) of the microsymbiont plays an important role as a signal molecule. It has been shown in cell cultures that the LPS is able to suppress an elicitor-induced oxidative burst. To investigate the effect of S. meliloti LPS on defense-associated gene expression, a microarray experiment was performed. For evaluation of the M. truncatula microarray datasets, the software tool MapMan, which was initially developed for the visualization of Arabidopsis (Arabidopsis thaliana) datasets, was adapted by assigning Medicago genes to the ontology originally created for Arabidopsis. This allowed functional visuali- zation of gene expression of M. truncatula suspension-cultured cells treated with invertase as an elicitor. A gene expression pattern characteristic of a defense response was observed. Concomitant treatment of M. truncatula suspension-cultured cells with invertase and S. meliloti LPS leads to a lower level of induction of defense-associated genes compared to induction rates in cells treated with invertase alone. This suppression of defense-associated transcriptional rearrangement affects genes induced as well as repressed by elicitation and acts on transcripts connected to virtually all kinds of cellular processes. This indicates that LPS of the symbiont not only suppresses fast defense responses as the oxidative burst, but also exerts long-term influences, including transcriptional adjustment to pathogen attack. These data indicate a role for LPS during infection of the plant by its symbiotic partner.

Plants and animals have the capacity to recognize and attack pathogenic microorganisms such as fungi, viruses, or bacteria. But, in contrast to animals, plants lack a system of circulating fluids, like blood and lymph, as well as specialized immune cells. Therefore, each plant cell has to possess the ability to recognize different pathogenic molecules and to respond with a broad range of defense reactions. In recent years, it has become more and more evident that there are two general systems of plant immunity (Gomez-Gomez and Boiler, 2002; Niirnberger et al., 2004). Cultivarspecific resistance is realized on a gene-for-gene basis. In this context, certain plant cultivars carrying a specific resistance gene are able to mount defense reactions against pathogenic microorganisms that carry a corresponding avirulence gene. These defense reactions usually include induction of a hypersensitive response (HR) to inhibit the spread of a pathogen as fast as possible (Dixon and Harrison, 1994). Cultivarspecific resistance is discussed to resemble the animal's adaptive immune system, although it does not involve specialized immune cells or recombinatory events on the genetic level. The second type of plant immunity is the ability to distinguish between self and nonself. This strategy is known from animals as innate immunity and this term is also applied to the plant system (Gomez-Gomez and Boiler, 2002). Recognition of foreign organisms in the plant's surrounding often does not lead to such vigorous responses like the activation of a HR, but rather induces general preparations to a possible invasion by microorganisms. These defense mechanisms include induction of oxidative burst, expression of defense-related genes, production of antimicrobial compounds, like phytoalexins, and reinforcement of the plant cell wall (Benhamou, 1996;Ebel and Mithofer, 1998;Heath, 2000). Although these responses are also activated during cultivar-specific resistance, hypersensitive cell death is absent during basal defense. The plant's innate immunity is based on recognition of microbial signal molecules, termed general elicitors or pathogen-associated molecular patterns (PAMPs). Unlike the products of avirulence genes in cultivar-specific resistance, PAMPs are expressed by a wide range of microorganisms and are usually indispensable for their biological fitness. Among the PAMPs are long-known and well-characterized elicitors like (l,3-l,6)-hepta-/3-glucoside isolated from cell walls of Phytophthora sojae (Sharp et al., 1984a(Sharp et al., , 1984b, chitin from cell walls of higher fungi (Bartnicki-Garcia, 1968; Ren and West, 1992) LPS are also important signal molecules in symbiotic signaling. In the Rhizobium-legume symbiosis, evidence for the involvement of LPS in the establishment of symbiosis has been gained from mutants defective in LPS synthesis. In different kinds of Rhizobialegume associations, alterations in the LPS structure lead to delayed nodulation, abortion of infection threads, formation of nonfixing nodules, and induction of plant defense reactions (Carlson et al., 1987;Lagares et al., 1992;Perotto et al., 1994;Niehaus et al., 1998;Campbell et al., 2002). Albus et al. (2001) showed that LPS of the microsymbiont Sinorhizobium meliloti is able to suppress elicitor-induced oxidative burst in suspension cultures of the alfalfa (Medicago saliva) host plant. Because this LPS does induce oxidative burst in cell cultures of the tobacco nonhost plant, the suppression effect is not a general feature of the rhizobial LPS, but is speculated to be a very specific feature in the establishment of a symbiotic relationship. Recent results identified the LipidA substructure of the S. meliloti LPS to be the suppressor-active portion (Scheidle et al., 2005). However, the extent to which rhizobial LPS is able to suppress the host's defense mechanisms has remained poorly characterized. To gain more information on the capacity of S. meliloti LPS to interfere with the plant's defense machinery, we performed microarray experiments to monitor the global expression of defense-related genes. Defense reactions were induced by yeast invertase (Basse et al., 1992). By treatment of suspension-cultured Medicago truncatula cells with either S. meliloti LPS and yeast invertase or yeast invertase alone, the effect of S. meliloti LPS on the level of gene expression could be addressed. Comparison of defense-related gene expression in the two kinds of treatment showed that induction of many genes strongly correlated to plant defense (e.g. Phe ammonia lyase [PAL] and isoflavone reductase [IFR]) was significantly suppressed after LPS pretreatment. Using MapMan software (Thimm et al., 2004;Usadel et al., 2005), it was furthermore possible to make a functional correlation between the suppression effect calculated for all genes displayed on the microarray and their specific function inside the cell.

Invertase Induces Defense-Associated Gene Expression in M. truncatula Cell Suspension Cultures
In this study, M. truncatula suspension-cultured cells derived from roots were used as a test system for the perception of microbial signal molecules. To verify the responsiveness of the cell culture, invertase was used as a well-established elicitor of plant defense reactions. In all four biological replicates used, the addition of invertase to a final concentration of 50 /ig mL"1 led to the production of hydrogen peroxide (H2O2), with an average maximal concentration (±se) of 3.54 jam H2O2 (±0.15) 15 min after elicitation (Fig. 1). The oxidative burst assay was used as a test system for the uniform perception of the elicitor in different experiments. Only responsive cell cultures were used for subsequent expression profiling.
For transcriptome-profiling studies, M. truncatula cell suspensions were treated for 4 h with either invertase, S. meliloti LPS, invertase + S. meliloti LPS, or water as a control. Cells were harvested, frozen, and processed for transcriptome profiling. Based on analysis of all four biological replicates in comparison to water-treated suspension cells, LPS treatment alone hardly showed any effect on gene expression in M. truncatula cell cultures. A complete list of genes meeting the prerequisites , or water as a negative control (black diamonds). The production of H2O2 was determined utilizing the luminol-based chemiluminescence assay. Oxidative burst measured for one of the four biological replicates is shown. P < 0.05, A > 5.5, and n > 6 (P, A, and w describe the statistical relevance, signal intensity, and the number of agreeing replicates, respectively) is included in Supplemental Table SI. No gene in this list was identified as induced 2-fold or more and only two genes (tentative consensus [TC]88927 and TC78069) were repressed more than 2-fold. These results indicate that M. truncatula cell cultures do not respond strongly to the LPS of the symbiotic partner. In contrast, in suspension cells treated with invertase, 336 genes were identified to be at least 2-fold induced with a statistical significance of P < 0.05, whereas 43 genes were identified to be repressed at least 2-fold. Table I lists the 20 genes that are most strongly induced or repressed, respectively, and the complete list of genes meeting the prerequisites P < 0.05, A > 5.5, and n > 6 is included in Supplemental Table SI Another gene strongly induced after elicitation with invertase was a chitinase involved in the defense against fungal pathogens. Furthermore, transcripts encoding two germin-like proteins were highly abundant in the elicited cell cultures. Germin-like proteins are ubiquitous plant proteins encoded by diverse multigene families. Although the function of most germinlike proteins could not be elucidated so far, nectarin I, a germin-like protein present in the nectar of tobacco plants, was found to exhibit superoxide dismutase activity (Carter and Thornburg, 2000). Additionally, a gene encoding a cyanogenic /3-glucosidase was strongly induced after elicitation. Cyanogenesis has been described in the context of protection against herbivores, but, due to the high toxicity of HCN, other roles in plant defense were postulated (Poulton, 1990).
Six of the 20 most repressed genes encode proteins involved in cell wall degradation. Because reinforcement of the plant cell wall is one of the major responses upon pathogen attack, it was not unexpected to find the expression of degradative enzymes repressed.
Manual classification of transcription-profiling data according to published literature revealed that, among the 336 genes induced 2-fold or more after elicitation with invertase, more than 170 have been described to be involved in plant defense or induced under biotic stress conditions. Grouping of all genes induced at least 2-fold into functional categories enables the illustration of different plant responses upon elicitation (Fig. 2). These include expression of several signaling components, activation of different hormone pathways, expression of proteins involved in protection from oxidative stress, as well as induction of transcriptional regulators. Besides the two largest categories that are composed of genes with unknown function and those that did not fit into any of the given functional categories (others), two categories (secondary metabolism and biotic stress) comprise 14% and 16% of all genes induced at least 2-fold, respectively. These findings are in general agreement with previous studies (Suzuki et al., 2005) and confirm the responsiveness of the utilized M. truncatula suspension cell line on the transcriptome level.

The Visualization Tool MapMan Allows Fast and Easy Evaluation of Transcriptome-Profiling Datasets on the Functional Level
Transcriptome-profiling studies can give valuable information on global cellular responses; however, evaluation and interpretation of the datasets are time consuming and require detailed knowledge of the existing literature. Bioinformatic tools can help to organize datasets according to preexisting biological knowledge and therefore enable comprehensive interpretation on the functional level. MapMan is a visualization tool developed to display plant genomic datasets onto pictorial diagrams (Thimm et al, 2004). The software organizes genes into blocks of cognate functions to allow even tentatively assigned genes and individual members of gene families to be displayed. This structure allows detection of general trends in the system's response and, in cases where the ontology is further developed, more precise resolution at the single pathway level. Recently, MapMan has been extended to enable statistical evaluation of differences in the responses of sets of genes assigned to different biological functions, allowing more validated interpretation of global cellular responses (Usadel et al., 2005)

. Because MapMan was initially designed to process datasets obtained from Arabidopsis (Arabidopsis thaliana) ATH1 microarrays, adaptation of this tool for employment with other microarray formats and plant species was an important step. At the beginning of the year, Urbanczyk-Wochniak et al. (2006) published the extension of MapMan for displaying Solanaceous datasets. The basis for this adaptation is the classification of the genes represented on the microarray according to the ontology used in MapMan.
To use MapMan as a tool to visualize transcriptomic datasets obtained from M. truncatula Mtl6kOLIl and Mtl6kOLIlPlus microarrays, we assigned the 16,470 TCs represented on the Mtl6kOLIlPlus array to MapMan BINS and subBINS based on a combination of automated BLAST searches against The Arabidopsis Information Resource 6.0 database and protein domain information. TCs lacking a good hit in Arabidopsis were assigned manually according to The Institute for Genomic Research (TIGR) annotation. It has to be Table I  MapMan is designed to allow for the creation of userdefined diagrams and visualization maps, this tool is well suited for employment in a symbiotic context. To demonstrate the value of the visualization of transcriptome data using MapMan, the dataset obtained from invertase-treated suspension cell cultures described above was used for a proof-of-concept study. In  Table S1 . The number of genes allocated to each functional category is indicated, the functional categories are defined.

. Overview of the effect of invertase treatment on M. truncatula suspension cells Listed are the 20 most strongly induced and repressed genes obtained from transcriptomic profiling of invertase-treated cells in comparison to watertreated cultures. Suspension-cultured M. truncatula cells were treated for 4 h with either 50 /ig mL"1 invertase or water as a reference. Oligo ID
cultures. Visualization of the cell's general metabolism confirms this cellular response (Fig. 3A). Furthermore, organization into hierarchical modules of biological relevance facilitates a more detailed resolution of global responses. For example, the three modules representing the secondary metabolism pathways flavonoids, phenylpropanoids, and phenolics, as well as N-misc, are clearly induced in this MapMan visualization. The strong induction of the pathways involved in lignification and phytoalexin synthesis was detected by standard evaluation methods as described above, whereas induction of the alkaloid metabolism (N-misc) had remained undetected. According to the M. truncatula TIGR gene index 7.0, most Medicago genes contained in this BIN cannot be assigned unambiguously to a specific cellular function, rendering detection of an induction in alkaloid biosynthesis via standard evaluation methods impossible. Nevertheless, automatic assignment of Medicago genes to the corresponding BIN as described in "Materials and Methods" leads to the identification of otherwise undetected response patterns, such as induction of the alkaloid pathway highlighting the applicability of the gene classification used to employ MapMan for the evaluation of Medicago datasets. Stimulation of alkaloid formation upon elicitation with fungal signal molecules has been shown before for different plant species (Blechert et al., 1995). Although production of alkaloids in Arabidopsis is poorly understood, induction of the berberine biosynthetic pathway upon wounding was described (Cheong et al., 2002). Therefore, induction of alkaloid metabolism upon elicitation can also be expected in M. truncatula. Another conspicuous response of invertase-treated cell cultures is the transcriptomic rearrangement of cell wall metabolism. In this context, the mixture of induced and repressed genes can be easily interpreted by closer examination of this module. Genes encoding for pectin methylesterases were induced upon elicitation with invertase. The transcriptional activation of pectin methylesterases upon herbivore attack has been shown before (von Dahl et al., 2006). In this context, demethylation of the cell wall pectin has been proposed to enable cross-linking via calcium ions, resulting in reinforcement of the cell wall. Other enzymes transcriptionally up-regulated upon invertase treatment are cellulose synthases and polygalacturonase inhibitor proteins, whereas proteins involved in cell wall degradation (i.e. cellulases, polygalacturonases, etc.) are repressed. Cell wall reinforcement is known to be a crucial response upon pathogen attack, but transcriptional initiation of this process besides the activation of lignifying enzymes or other secondary metabolic pathways has been rarely reported (Schenk et al., 2000). Another response that has been proposed to be induced following pathogen attack is the degradation of fatty acids by /3-oxidation. Although the cluster of lipid metabolism did not show any marked changes, closer examination of the transcriptional changes of this cluster reveals a minor induction of the /3-oxidation pathway, as well as of the glyoxylate cycle. This is in accordance with a previous study that monitored transcriptional adaptations upon attempted infection of Arabidopsis with Pseudomonas syringae pv tomato (Scheideler et al., 2002). Although it is tempting to speculate that mobilization of fatty acids might be utilized for generation of energy, it should be emphasized that, on the other hand, our data indicate a repression of starch degradation. This could hint to a role of j8-oxidation in processes different from energy generation, like the provision of precursor molecules for biosynthetic pathways. Furthermore, our data show induction of the oxidative pentose phosphate pathway that has been shown to be essential in supplying NADPH for the activity of NADPH oxidase during oxidative burst (Pugin et al., 1997). Another response toward elicitation is induction of the cluster representing the synthesis of Asp family amino acids. This pattern is mainly caused by the induction of several S-adenosyl-methionine synthetase enzymes involved in the production of the phytohormone ethylene that has been identified as one of the key players in orchestrating plant disease resistance (Feys and Parker, 2000). Finally, a trend in the induction of transcripts involved in glutathione and ascorbate metabolism can be detected.  Figure   3B visualizes transcriptomic datasets with respect to different cellular responses. In this representation, the most striking response observed was for a class of genes assigned as involved in biotic stress. In fact, only five of these genes appear to be suppressed upon invertase-mediated elicitation of cell cultures. Three of these are annotated as protease inhibitor proteins and could possibly be involved in mechanisms different from inhibition of proteases of pathogenic origin. Other patterns that can be observed are the induction of several genes involved in the response to different kinds of stress apart from drought and salt stress. Cross talk between the responses toward different kinds of stresses is a frequent observation, although stronger induction of wound-induced responses could have been expected (Desikan et al., 2001; Cheong et al., 2002). The class of miscellaneous genes involved in the response to abiotic stress contains four highly induced genes. All four are annotated as encoding germin-like proteins, which are well known to be elicitor-inducible proteins. The above-discussed induction of genes involved in protection from oxidative stress is visualized in more detail in Figure 3B, demonstrating that, apart from ascorbate and glutathione metabolism, little transcriptional reprogramming can be observed upon elicitation.

Induction of antioxidative proteins like dismutases and catalases is missing despite the potential need for scavenging ROI produced during oxidative burst. This could hint to a role of ascorbate and glutathione cycle intermediates in addition to detoxification of ROI. It has been proposed that glutathione itself is involved in signaling events leading to the activation of plant defense responses (Dron et al., 1988; Gomez et al., 2004). On the other hand, a function as a superoxide dismutase has been described for a germin-like protein from the tobacco nectar, nectarin I.
Although such a function remains purely speculative, the above-described induction of four germin-like proteins could be due to their function in the scavenging of ROI. Finally, some transcriptional rearrangements can be observed in the cluster of development-associated genes predominantly showing repression of this class. The transcriptomic adaptation of genes involved in cell maintenance and development has been described before (Schenk et al. 2000). However, no agreement on the single gene level could be made by comparing our results to the list of regulated genes described earlier.
In summary, the clear patterns of plant response upon elicitation visualized using MapMan are in accordance with existing literature and validate the usefulness of this software tool for the evaluation of transcriptome datasets. Nevertheless, it has to be emphasized that interpretation of the results leads to identification of genes, which did not follow the general trend observed in the class to which they were assigned. This can be explained either by opposing regulation of transcripts marking isoenzymes with different cellular functions or by assigning such genes to a wrong BIN.

The LPS of the Microsymbiont S. meliloti Suppresses Elicitor-Induced Transcriptional Reprogramming
To investigate the effect of S. meliloti LPS on defenseassociated gene expression in M. truncatula, we treated suspension cell cultures simultaneously with invertase and S. meliloti LPS with final concentrations of 50 /xg mL"1 each. This concentration suppresses the invertaseinduced oxidative burst in M. truncatula suspensioncultured cells to nearly 90% (Fig. 1). In contrast, concomitant treatment of cell cultures with invertase and LPS of the phytopathogen X. campestris pv campestris did result in oxidative burst with 91% maximal H2O2 production as compared to elicitation with invertase alone (Supplemental Fig. SI). These results indicate that the suppression effect is a specific feature of the M. truncatula-S. meliloti (legume-Rhizobia) interaction, which is further substantiated by elicitation of oxidative burst by S. Table I shows that, for all 20 genes most strongly induced upon invertase treatment, there is also strong induction following combined treatment with invertase and S. meliloti LPS. Nevertheless, in each case, the induction rate is markedly reduced, giving a first hint to moderating activity of the rhizobial LPS. The same effect can be observed for 19 of the 20 genes most strongly repressed upon elicitation with invertase. To investigate general patterns of cellular adaptation induced by simultaneous subjection to invertase and S. meliloti LPS, the dataset was visualized using MapMan. Figure 4 shows a representation of the cell's basic metabolic pathways in response to the combined treatment. The overall response shows strong induction of secondary metabolism, some rearrangements in cell wall metabolism, and induction of several transcripts involved in ascorbate and glutathione metabolism. In general, the responses observed upon treatment with invertase and rhizobial LPS resembled transcriptional rearrangements followed by elicitation with invertase alone. Nevertheless, the described response patterns are induced to a lesser extent in the presence of S. meliloti LPS. To investigate a possible suppression effect of the rhizobial LPS on the global level, the differences between induction rates for each gene of both datasets were calculated. Genes for which rhizobial LPS had a moderating effect on transcriptional adjustment were characterized by negative values, whereas for genes that were even more strongly altered after simultaneous treatment with invertase and S. meliloti LPS compared to elicitation with invertase alone, the calculated values were positive. To analyze the global effects of S. meliloti LPS on defense-associated gene expression, the resulting values for differences in transcriptional response ratios were visualized by MapMan, using a representation of the cell's general metabolism (Fig. 5). It has to be emphasized that, unlike the diagrams discussed before, this representation does not visualize an unprocessed microarray dataset, but the calculated differences between two datasets that were compared. The comparison of the two datasets clearly showed that S. meliloti LPS has a moderating effect for nearly all genes that are displayed in Figure 5. This affects genes that are induced by invertase elicitation as well as genes that are down-regulated in response to the pathogenic stimulus. The visualization shows very few genes for which the rhizobial LPS seems to have a stimulating effect on defense-associated transcriptional rearrangement. A closer examination of these genes showed that, with one exception (TC80744, a gene annotated as unknown), the relative response ratios for these transcripts were comparably low in both datasets that were compared. Thus, the calculated positive values for the differences in response ratios could be due to the inaccuracy of the M values obtained in the microarray experiments. Of the 323 genes induced 2-fold or more after invertase treatment, which were compared to the dataset of cells treated with invertase and S. meliloti LPS, only 19 transcripts showed a positive value for the differences in relative response ratios. For all other genes, the calculated values indicated a moderating effect of rhizobial LPS on elicitor-induced transcription activity (i.e. values of differences in relative response ratios were negative). Nevertheless, the datasets obtained by the microarray experiment do not indicate a fixed degree for the suppression of transcriptional reorganization. Closer examination of nodulespecific genes did confirm the suppression effect of S. meliloti LPS on those nodulins that were induced upon invertase treatment (e.g. TC77131 encoding a nodulin 12A precursor). Nevertheless, significant transcriptional rearrangements in response to rhizobial LPS alone could not be observed.

meliloti LPS in tobacco cells (Albus et al., 2001). For transcriptional profiling analyses, cells were incubated with invertase and S. meliloti LPS for 4 h, separated from the medium, frozen, and RNA extracted and processed for microarray hybridizations. Transcriptomic reprogramming induced by this treatment combining pathogenic and symbiotic stimulus was compared to the transcriptome profiles obtained by elicitation with invertase alone. Comparison of the two datasets was made possible by using the same RNA of water-treated cells as a reference. A list of all genes in the dataset obtained from the combined treatment meeting the prerequisites P < 0.05, A > 5.5f and n > 6 is included in Supplemental Table SI. Fourteen of the 20 most strongly induced genes are also among the 20 genes with highest induction rates after elicitation with invertase listed in Table I. Comparison of gene induction rates given in
To confirm the moderating effect of rhizobial LPS on elicitor-induced transcriptional regulation, we monitored gene induction of eight genes using real-time reverse transcription (RT)-PCR. TCs selected for this approach included six genes induced upon invertase treatment and two genes that were repressed after elicitation and for which the microarray approach had shown a strong suppression effect mediated by S. meliloti LPS. It should be mentioned that, for the peroxidase gene (TC85174), no unique primers could  be designed and melting curves indicated amplification of at least two different transcripts. Therefore, real-time RT-PCR monitored the abundance of a mixture of at least two different peroxidase genes. For all six TCs that were identified via microarray hybridization to be induced upon elicitation, induction could be confirmed (Fig. 6). Furthermore, induction of the genes was significantly suppressed in cells treated simultaneously with invertase and S. meliloti LPS (P < 0.05; t test). Regarding the two TCs that were identified to be repressed upon invertase treatment only for TC81805 encoding a MYB-related transcription factor, repression of transcriptional activity could be confirmed. For both of these genes, rhizobial LPS did not show any significant moderating effect on transcriptional regulation (data not shown). Therefore, the suppression effect could only be confirmed for genes that are induced during elicitation, although the calculated differences in relative response ratios visualized in Figure 5 clearly indicate that S. meliloti LPS has a moderating effect on both induction and repression of genes.
These results show two important characteristics of S. meliloti LPS-mediated suppression of defenseassociated gene expression: (1) Global visualization of the differences in relative transcript levels of cells elicited with invertase and cells concomitantly treated with invertase and S. meliloti LPS shows that LPS has a moderating effect on the transcriptional adaptation of genes induced as well as on those repressed upon elicitation; (2) the suppression effect applies to all kinds of cellular processes, indicating early interference of LPS with signal transduction pathways.
However, by which means rhizobial LPS targets transmission of recognition of a pathogenic signal molecule into plant defense responses remains elusive. A different kind of suppressor molecule of rhizobial origin, the cyclic l,3-l,6-/3-glucans from the soybean (Glycine max) symbiont Bradyrhizobium japonicum have been shown to succeed in competition with the /3-glucan elicitor of P. sojae for its binding sites (for review, see Mithofer, 2002). This competitive binding not only inhibits the production of phytoalexins (Mithofer et al., 1996), but also earlier defense-associated signaling events (i.e. an increase in cytosolic Ca2+ concentration [Mithofer et al., 1999]). However, unlike S. meliloti LPS, the cyclic /3-glucans from B. japonicum do not suppress, but rather induce, production of ROI in the host plant. Another mechanism that could be responsible for S. meliloti LPS-mediated suppression of plant defense is the interference with basal elements of the signal transduction pathways responsible for the activation of defense responses. In the case of the phytopathogenic fungus Mycosphaerella pinodes, it has been shown that a glycopeptide termed supprescin is able to suppress plant defense at different levels, including induction of PAL transcription (Wada et al., 1995). The mechanism of suppression is likely to involve interference of the supprescin glycopeptide with a plasma membrane ATPase via direct interaction (Kato et al., 1993). Furthermore, the supprescin of M. pinodes does interfere with proteinase inhibitor metabolism in the plant plasma membrane that is essential for activation of defense responses (Shiraishi et al., 1994). It has been shown in studies on pea (Pisum sativum) plasma membranes and epicotyls that elicitation leads to a rapid and biphasic increase in levels of phosphatidylinositol-4,5-bisP and inositol 1,4,5-triP, whereas addition of the suppressor molecule inhibited this effect (Toyoda et al., 1992(Toyoda et al., , 1993. These findings show that inhibition of compounds of very early signaling cascades is used by pathogens as a comprehensive mechanism of suppression. Another possibility as to how suppressors could operate in interfering with defense responses is active recognition of molecules via specific receptors. This idea has been proposed for the glycopeptide supprescin from M. pinodes by Shiraishi et al. (1994), although to date there is no proof for such a mode of action. However, specific recognition of suppressors via receptor proteins could mediate fine-tuned regulation of defense responses that are switched on or off upon concomitant recognition of an elicitor and a suppressor. In contrast, inhibition of early signaling processes necessarily blocks all downstream processes without any possibility of regulation. In the case of suppressors of pathogenic origin, the presence of a receptor protein without any other function appears unlikely because it would cause an immense evolutionary disadvantage, rendering the plant susceptible to pathogens producing the appropriate suppressor molecule. Nevertheless, in the case of symbiotic interactions, the presence of receptors specific for suppressors of symbiotic origin would enable different kinds of beneficial responses, including fine-tuned suppression of plant defense.
Surface polysaccharides act as bacterial signals in pathogenic and symbiotic interactions (Djordjevic et al., 1987). Although, to date, the role of the rhizobial LPS in the establishment and maintenance of symbiosis could not be elucidated in detail, there is evidence for its importance in the interaction. LPS mutants of Rhizobium leguminosarum as well as S. meliloti show severe defects in the establishment of symbiosis with their host plants (Vedam et al., 2004). In some cases, induced nodules are small, poorly infected, and show signs of plant defense (e.g. accumulation of cell wall material and phenolic compounds and autofluorescence [Perotto et al., 1994;Niehaus et al., 1998]). In both studies, bacteria were able to induce and colonize infection threads, but failed to differentiate after infecting nodule tissue. The death of rhizobial mutants during the infection process was thought to be due either to impairment of bacterial protection against unfavorable conditions or to inability to interfere with the host's defense response. Another mutant study showed that not only the presence, but also the correct modification, of LPS is needed for optimal nodulation activity (Keating et al., 2002). Furthermore, purified LPS of the microsymbiont has been shown to promote infection in the case of the R. leguminosarum bv trifoliiwhite clover (Trifolium repens) symbiosisand to cause the production of novel proteins (Dazzo et al., 1991). The principle of the ability of rhizobial LPS to suppress plant defense responses in the host was achieved by demonstrating the suppression effect on elicitor-induced oxidative burst (Albus et al., 2001). Nevertheless, it is known that, during symbiotic interaction, prolonged production of H2O2 can be detected in the infection threads and the infection zone of the nodule, a process most likely involved in regulation of nodulation (Santos et al., 2001). The fact that oxidative burst during the infection process does not lead to abortion of the symbiotic interaction is due to the high activity of antioxidative proteins (Dalton et al., 1998). Therefore, it was a crucial step to demonstrate the suppression of plant defense responses different from oxidative burst. This study demonstrates that S. meliloti LPS has the capacity to interfere with the transcriptional response toward pathogenic signaling, further emphasizing the importance of LPS during successful symbiotic interaction.

Plant Material
Medicago truncatula Jemalong cell suspension cultures were obtained and maintained as described by Scheidle et al. (2005).

Cultivation of Bacteria and Isolation of LPS
Sinorhizobium meliloti wild-type strain 2011 (Casse et al., 1979) was grown on tryptone yeast (Saccharomyces cerevisiae) agar plates supplemented with 0.4% Glc (w/v) at 28°C for 3 d. Bacteria were washed from the plates with 0.9% NaCl (w/v) and centrifuged at 5,000g and 4°C for 20 min. Cells were washed three times with cold water, resuspended in water, and LPS was extracted via the hot phenol-water method (Westphal and Jann, 1965). The water phase was dialyzed for 3 d against water. Proteins and nucleic acids were removed by treatment with 100 /ig mL"1 DNase I (Serva), 15 /xg mL"1 RNase (Serva), and 150 /xg mL"1 proteinase K (Roche). Fractions were dialyzed against water for 2 d and lyophilized. The lyophilisate was resuspended in water and purified by ultracentrifugation at 100,000g and 4°C for 8 h. Centrifugation was repeated and the pellet resuspended in water and lyophilized. For the experiments, the LPS was dissolved in water and ultrasonicated on ice for 20 min.

Treatment of Plant Cell Cultures
For each biological replicate of microarray hybridizations, an appropriate number of M. truncatula suspension cell cultures was pooled. A sample of the pooled cells was sieved and 4 X 2 g of cells were removed for oxidative burst measurements. The remaining cell suspensions were realiquoted to sterile 100-mL flasks and incubated under standard cultivation conditions for 4 to 5 h. After determination of oxidative burst using the cells removed from the pooled cultures had confirmed responsiveness of the suspension cells, realiquoted cultures were treated with water (500 fiL), 50 \l% mL"1 yeast invertase (Sigma), 50 /ig mL"1 S. meliloti LPS, or a combination of 50 fig mL"1 invertase and 50 /ig mL"1 S. meliloti LPS. After treatment for 4 h, each cell culture was sieved to remove the cells from the medium and cells were rapidly frozen in liquid nitrogen. Four independent biological replicates were used for subsequent isolation of RNA.

Oxidative Burst Measurements
To confirm responsiveness of the cell suspensions used for microaiTay hybridization, 2 g of cells were resuspended in 8 mL of preincubation medium (3% Sue [w/v] in 4% Murashige and Skoog [v/v]; Murashige and Skoog, 1962). Cells were incubated under culture conditions for 3 h. Oxidative burst was determined using the H2O2-dependent chemiluminescence reaction according to Warm and Laties (1982). After cell cultures were subjected to one of the treatments used in these experiments, 200 /xL of suspensioncultured cells were added to 700 fiL 50 mM phosphate buffer (pH 7.9) and 100 /LtL 1.2 mM luminol (Sigma). The reaction was started by adding 100 fiL of 14 mM potassium-hexacyanate (Fluka). Luminescence was monitored using a Sirius luminometer (Berthold).

RNA Extraction and Microarray Hybridization
Total RNA from suspension-cultured cells was extracted using TRIzol reagent (Sigma) and purified using Microcon-30 columns (Millipore). Twenty micrograms of RNA were used to synthesize Cy3-and Cy5- The classification based on the similarity to Arabidopsis proteins was then automatically checked against domains that have been assigned to a MapMan category. Furthermore, Medicago TCs that did not meet the prerequisites for an automatic draft assignment were classified manually according to their respective TIGR annotation (TIGR release 7.0). Finally, all draft assignments were corrected manually for potential mistakes.

RT-PCR
To validate results obtained by microarray hybridizations, RT-PCR was performed for eight genes using gene-specific primers (Supplemental Table  S2). Peroxidase-specific primers (TC 85182) were not unique and led to the amplification of at least two different transcripts. Primers had a calculated melting temperature of 53°C ± 0.4°C and amplifications were not longer than 300 bp. RT-PCR was performed using the QuantiTect SYBR Green RT-PCR kit (Qiagen). Fifty nanograms of DNasel-treated total RNA (Serva) were used in a total volume of 25 fiL. Amplification was monitored using the Opticon real-

ACKNOWLEDGMENTS
Manuela Meyer is gratefully acknowledged for her assistance in the synthesis of microarray targets and in microarray hybridization. Axel Nagel is acknowledged for his support in maintenance and updating of MapMan software and databases. Thomas Bekel and Alex Goesmann are acknowledged for providing information on protein sequences and domains via the SAMS system. Moreover, we thank Thomas Patschkowski for a critical reading of the manuscript.