Six nonnodulating plant mutants defective for Nod factor-induced transcriptional changes associated with the legume-rhizobia symbiosis

Abstract

As the legume-rhizobia symbiosis is established, the plant recognizes bacterial-signaling molecules, Nod factors (NFs), and initiates transcriptional and developmental changes within the root to allow bacterial invasion and the construction of a novel organ, the nodule. Plant mutants defective in nodule initiation (Nod^-) are thought to have defects in NF-signal transduction. However, it is unknown whether WT plants respond to NF-independent bacterial-derived signals or whether Nod^- plant mutants show defects in global symbiosis-associated gene expression. To characterize plant gene expression in the establishment of the symbiosis, we used an Affymetrix oligonucleotide microarray representing 9,935 Medicago truncatula expressed sequences. We identified 46 sequences that are differentially expressed in plants exposed for 24 h to WT Sinorhizobium meliloti or to the invasion defective S. meliloti mutant, exoA. Eight of these genes encode nucleolar proteins, which are implicated in ribosome biogenesis. We also identified differentially expressed transcription factors, signaling components, defense response proteins, stress response proteins, and several previously uncharacterized genes. NF appears both necessary and sufficient to induce most changes. Six of seven Nod^- M. truncatula mutants (nfp, dmi1, dmi2, dmi3, nsp1, and nsp2) showed no transcriptional response to S. meliloti, suggesting that the encoded proteins are required for initiating new transcription. The Nod^- mutant hcl, however, exhibits a reduced transcriptional response to S. meliloti, indicating that the machinery responsible for initiating new transcription is at least partially functional in this mutant.

In the Medicago truncatula/Sinorhizobium meliloti symbiosis, the two partners undergo a complex developmental process resulting in the production of a newly formed organ, termed the root nodule, in which bacteria reduce molecular dinitrogen for use by the plant. In initiation of the symbiosis, the bacteria produce a lipochitooligosaccharide-signaling molecule, Nod factor (NF) (1), in response to plant-derived flavonoid molecules in the soil. The plant responds to NF within seconds, exhibiting ionic fluxes and a calcium-spiking response in root hair cells (2–5). Within hours of exposure to S. meliloti, plant root hairs become deformed as a result of altered growth (6), and root cells initiate the transcription of nodulation-specific genes (7–10). By 24 h the pericycle and inner cortical cells of the root become activated in preparation for cell divisions that will give rise to the nodule (11, 12). Within 48 h of exposure to S. meliloti, bacteria become trapped within tightly curled root hairs, and plant-derived infection structures (infection threads) form within root hair cells to allow bacteria entry into the plant (12). Purified NF triggers many symbiotic responses in the plant, but it cannot provoke tight root hair curling or infection thread formation. Over the course of 2–3 weeks, bacteria differentiate into nitrogen-fixing bacteroids, and the plant forms a fully developed nodule.

Several bacterial mutants show defects early in the establishment of the symbiosis. Bacterial mutants that cannot synthesize NF, such as SL44 (13), fail to induce calcium spiking (14), early gene expression (7), or nodule formation on the plant (S.R.L., unpublished data). Bacterial mutants such as exoA, with defects in exopolysaccharide biosynthesis, elicit plant cell divisions but fail to successfully invade root hairs or induce gene expression changes associated with infection or nodule formation in M. truncatula (15–17).

Nonnodulating (Nod^-) M. truncatula mutants have been identified and characterized based on their responses to bacterial signals. The Nod^- plant mutant nfp shows no measurable response to S. meliloti inoculation (18). The Nod^- mutants dmi1 and dmi2 exhibit a rapid calcium flux and root hair swelling in response to NF but do not show calcium spiking, early nodulation-associated gene expression, or cortical cell divisions (3, 19, 20). The dmi3 mutant differs from dmi1 and dmi2 by exhibiting a calcium spiking response, the nsp1 and nsp2 mutants additionally exhibit root hair branching responses, and the hcl mutant exhibits root hair curling and cortical cell division foci (19–22). All Nod^- plant mutants, with the exception of the hcl mutant, fail to transcriptionally up-regulate the three nodulation-induced genes thus far examined (20–22). A comparison of transcriptional profiles for plant mutants with differing root hair responses may identify genes important for root hair deformation or curling.

Several plant genes are differentially regulated in the establishment of the symbiosis. A peroxidase gene (RIP1) and an annexin gene (MtANN1) are induced in the plant from 6 to 48 h after exposure to the bacterial symbiont (7, 23). In addition, several genes, termed early nodulins (ENODs) or nodulins, are induced in rhizobia-inoculated plants (8, 9, 24, 25). M. truncatula EST sequencing projects have identified 189,919 ESTs from 47 different cDNA libraries. The sequences of these ESTs are compiled at The Institute of Genomic Research and have been assembled into 17,610 tentative consensus sequences (TCs) based on sequence overlap (26). By using data from these M. truncatula EST projects, bioinformatics approaches have identified candidate genes that are predicted to have nodule-specific expression based on representation in nodule-specific cDNA libraries (16, 27, 28).

The initiation of the symbiosis triggers a wide variety of physiological changes in preparation for bacterial infection and nodule construction. We hypothesized that these changes might be accompanied by different plant transcriptional events related to bacterial identification, plant signal transduction, or root hair deformation. To test this hypothesis, we used a composite Affymetrix oligonucleotide chip (29) representing a large, targeted portion of the M. truncatula transcriptome. We initially identified a set of 46 plant genes that are differentially regulated in M. truncatula after 24 h of exposure to S. meliloti. Using the expression of these genes, we have tested bacterial mutants defective for NF production and infection, and the application of purified NF for the ability to alter gene expression at 24 h, and propose that NF is the principal signal recognized by plants at this stage of the symbiosis. Further, we have examined the transcriptional responses of seven Nod^- M. truncatula mutants (nfp, dmi1, dmi2, dmi3, nsp1, nsp2, and hcl) to S. meliloti and found that only one mutant, hcl, shows any measurable response for these genes, suggesting that the signal transduction pathways must converge to produce transcriptional changes at this stage of the symbiosis.

Materials and Methods

Plant Growth and Bacterial and NF Treatments. M. truncatula Gaertner cv. Jemalong A17, nfp (18), dmi1–1 (20), dmi2–1 (20), dmi3 (20), nsp1–1 (20), nsp2–2 (21), and hcl-1 (22) seeds were prepared and planted as described in ref. 16. S. meliloti strain Rm1021 is a streptomycin-resistant derivative of WT field isolate SU47 (30). The S. meliloti mutant SL44 has a deletion in the nodD1-nodABC region, and the S. meliloti mutant TJ170 has a Tn5 insertion in nodC, rendering both mutants unable to make the GlcNAc backbone of NF (13, 31). The exoA bacterial mutant (Rm7031) is described in ref. 17. Bacteria were grown in liquid tryptone-yeast extract (32) supplemented with appropriate antibiotics at the following concentrations: 500 μg of streptomycin per ml and 50 μg of neomycin per ml. For inoculations, bacteria were pelleted, washed in 0.5× buffered nodulation medium (BNM) (5), and resuspended in 0.5× BNM at an OD₆₀₀ of 0.05. One microliter of this bacterial suspension or NF was spotted onto the root tips of 6-day-old plants. NodRm-IV (Ac, S) was purified as described in ref. 2. Root segments from just above the root tip to 2.5 cm above the inoculation site were harvested 24 h after inoculation. One plant from each condition was examined 4 weeks later to confirm the appropriate nodulation phenotype, depending on genotype and treatment.

Affymetrix GeneChip Construction. We commissioned the construction of a custom S. meliloti/M. truncatula oligonucleotide chip with an 18-μm feature size and a 12.8-mm chip size (Affymetrix, Santa Clara, CA). The bacterial portion of the chip will be described elsewhere (M. Barnett, personal communication). The 9,935 TCs (The Institute of Genomic Research mtgi v.4.0, www.tigr.org/tdb/mtgi; ref. 26) from M. truncatula are represented by using 11 probe pairs (perfect match and mismatch) per gene with 3′ biased probe sets. In the current version of mtgi v.7.0, the TCs have been recompiled and renamed. Because the probe pairs for the oligonucleotide chip were chosen based on mtgi v.4.0 and the perceived 3′ end of each “gene” in this compilation, we refer to the TCs herein by their mtgi v.4.0 numbers. “TCs” are not always synonymous with “genes.” Multiple TCs may represent nonoverlapping ends of a gene; alternatively, a single TC may represent multiple genes in a closely related family. For the sake of simplicity, we assume that each TC represents a single gene and note exceptions to this assumption in the text. To facilitate identification of genes important for the root-based symbiosis, we chose TCs with at least one EST represented in uninoculated root, in S. meliloti-inoculated root (24, 48, or 72 h postinoculation), or in nodule libraries. In mtgi v.4.0, ESTs from 13 such libraries had been generated. All TCs from these libraries, with the exception of 500 TCs specific to a library from phosphate-starved roots (MHRP-), were printed on the chip. Other sequences used on the chip included genes that were likely to be of interest based on inferred roles in pathogenesis, calcium metabolism, or nodule biogenesis: TCs from Phytophthora-exposed roots or oligogalacturonide-treated roots; sequences representing the ethylene biosynthetic enzymes 1-aminocyclopropane-1-carboxylate synthase and 1-aminocyclopropane-1-carboxylate oxidase, 10 calcium pumps, and three phospholipase Cs (E. Engstrom, personal communication); TCs that were annotated at The Institute of Genomic Research to have roles in calcium homeostasis; M. truncatula homologs of pathogen responsive genes; known nodulins; and tissue-specific controls (courtesy of Kathryn Vanden-Bosch, University of Minnesota). As indicators for labeling bias, the 5′, middle, and 3′ sections of three genes (Actin/TC32575, S-adenosyl-l-methionine synthetase/TC31862, and Poly-Ubiquitin/TC39313) and all standard Affymetrix chip controls were included.

RNA Preparation and Affymetrix Chip Hybridizations. RNA was purified with TRIzol (Invitrogen) as described in ref. 16. For each preparation, ≈100 μg of total RNA was isolated from 60 root segments.

Affymetrix experiments were performed as described in the Affymetrix technical manual. Thirty micrograms of total RNA was used for SuperScript double-stranded cDNA synthesis (Invitrogen). Biotin-labeled cRNAs were synthesized by using the BioArray High Yield RNA Transcription Labeling Kit (Enzo Diagnostics). The quality of cDNA and cRNA synthesis was monitored by using agarose gel electrophoresis. Hybridization, washing, staining, and scanning were performed as described in the Affymetrix technical manual, at the Stanford Protein and Nucleic Acid Facility (Stanford, CA). Three independent biological replicates were performed for each treatment or genotype examined.

Expression Analysis. Pixel values were extracted from scan files by using Affymetrix mas 5.0 software, and .CEL files were analyzed by using dchip 1.3 (www.dchip.org; ref. 33). Probe sets specifying bacterial genes were masked before analysis. Extracted values were normalized across 109 independent chips by using invariant set normalization, which identifies a set of genes with small within-subset rank differences across all arrays, and uses these as the basis for fitting a normalization curve (34). Model-based analysis was performed by using perfect match (PM) only analysis, compiling data from three independent biological replicates for each condition (34). Data from analyzed chips were visually inspected to monitor hybridization, background changes, and array outliers. Pairwise comparisons were performed, and a lower 90% confidence bound (LCB) and fold change (FC) was determined for each comparison. Genes with an absolute value of the LCB of 2-fold or higher and with intensities differing by >100 between the two conditions were considered for further analysis. As an independent analysis method, the Significance Analysis of Microarrays (SAM) method was used (35). Genes for which transcriptional changes were >2-fold, with the expectation that no false positives would be chosen, were identified by using sam 1.21 (http://www-stat.stanford.edu/~tibs/SAM/index.html). For negative FC values, the log₂ FC is based on a transformed FC that is continuous about 1. For example, for a FC of -2, the transformed FC = 0.5 and the log₂ FC = -1. To group genes with similar expression patterns, hierarchical clustering using cluster 3.0 (http://bonsai.ims.u-tokyo.ac.jp/~mdehoon/software/cluster/software.htm) was performed on log₂ FCs based on Pearson correlation coefficient and centroid linkage clustering (36).

Real-Time RT-PCR. Real-time RT-PCR was performed by using a DNA Engine Opticon 2 Continuous Fluorescence Detection System (MJ Research, Cambridge, MA) from a first-strand template cDNA (Invitrogen) made by an Oligo(dT) primer and the DyNAmo SYBR Green qPCR kit (Finnzymes, Helsinki). In each reaction, 0.3 μM primer and 10 ng of cDNA was used. PCR reactions for each of three biological replicates were performed in triplicate. The initial denaturing time was 5 min, followed by 40 PCR cycles of 95°C for 10 s, 50°C for 10 s, and 72°C for 15 s. A melting curve was run after the PCR cycles. Primer sequences used are as follows: TC31841, ATTCTCATAGAGCTCACACAACCAG, TGTAAAGTTCCATCACTCCGATG; TC33471, ATGGA AT T TGCTGTAGA AGT TGA AG, A A A ACATTACTTGCCTTGATGTCTC; TC29161, TTTCTTCAACCAAATTTCATTTCTC, AAGTTTAGATCCACCACATCTTGAA; TC35954, GGGAGATTATTTCCTTCAATATGCT, GTGGTCTTCAAGAGTTATTTCTCCA; and TC40340, AAGGAGACTTTGAGAAAGCTGTGTA, ACATACCTAAATGCAGTCACTAGCC. Actin was used as the internal control gene (ATAAAGGAGAAGCTTGCATATGTTG, GGAACATAGTTGAACCACCACTAAG). Data were quantitated by using opticon monitor 2 (MJ Research). Normalized transcript levels were calculated as described in ref. 37.

Results

Identification of M. truncatula Genes Differentially Expressed During the Establishment of the Symbiosis. To focus on transcriptional changes elicited in the establishment of the symbiosis, we identified a set of plant genes that are differentially expressed in response to S. meliloti. We designed an Affymetrix oligonucleotide chip (29) representing 9,935 M. truncatula tentative consensus sequences (TCs), the sequences of which are derived mainly from M. truncatula root EST libraries (see Materials and Methods). We used this chip to compare transcriptional profiles of buffer-treated plants with those inoculated with S. meliloti for 24 h. Within this time frame, plant root hair cells exhibit alterations in ionic levels and in growth, and plant pericycle and cortical cells become activated for cell division.

Two criteria were used for defining a set of genes that showed differential expression in triplicate experiments. First, we calculated an average FC and a LCB for fold changes for each TC by using dchip analysis software. Because LCB takes the spread of the data into account, it is likely a better predictor of transcriptional changes than is FC (34). To conservatively choose genes that show consistent expression changes, we selected TCs for which the absolute value of the LCB (|LCB|) is ≥2-fold. Second, we used the SAM method (35); TCs identified by using LCB calculations were also identified by using the SAM method. By these strict criteria, we identified a set of 46 of the 9,935 M. truncatula TCs that show differential expression 24 h after inoculation with S. meliloti (Fig. 1).

Fig. 1.

Expression patterns of 46 TCs that are differentially expressed in the plant after 24 h of exposure to S. meliloti. Data from oligonucleotide chip experiments were pooled across three biological replicates, and a pooled log₂ FC was calculated for each pairwise comparison described below. A clustering map of genes with similar expression patterns is shown to the left of the table; a positive log₂ FC is represented by a red square and a negative log₂ FC is represented by a blue square. For this map, the brightest blue represents a log₂ FC of -2.3 and the brightest red represents a log₂ FC of 4.5. The first column of the table indicates the TC number of the gene examined. The log₂ FC from 12 pairwise comparisons is shown (columns 2–13). Five comparisons are between WT plants exposed to buffer and to WT bacteria (Rm1021, column 2), 100 nM NF (column 3), the bacterial mutant SL44 (column 4), the bacterial mutant TJ170 (column 5), or the bacterial mutant exoA (column 6). Seven comparisons are between mutant plants exposed to buffer or Rm1021 (columns 7–13). Log₂ FC numbers are colored in red or blue to indicate values ≥1 or ≤-1, respectively. Table cells are colored in yellow to indicate genes predicted by the SAM method to change transcript levels ≥2-fold. Sequence homologies are shown (far right column) to indicate potential functions of identified TCs.

This method was successful in identifying three known nodulins (Fig. 1). MtANN1 (TC40340, LCB = 2.2, FC = 2.8), ENOD11 (LCB = 3.8, FC = 6.8), and soybean nodulin 26 ortholog (TC28561, LCB = 3.3, FC = 4.9) exhibited gene expression changes above the thresholds selected (Fig. 1). Two additional M. truncatula early nodulins, RIP1 (7) and ENOD40 (38), showed an increased fold change but had a LCB of 1.5 (FC = 1.6) and 1.6 (FC = 2.0), respectively, and were not identified by the SAM method. The RIP1 transcript is expressed in uninoculated root tissues, and transcript levels increase in the presence of S. meliloti (7). These data suggest that our strict criteria will exclude some genes showing a shift in expression with a <2-fold change.

We independently verified the expression patterns of 5 of the 46 TCs by using real-time RT-PCR (Table 1). In all cases, the observed change in transcript abundance was similar when using both methods. In most cases, log₂ FC calculations derived from real-time RT-PCR experiments were more extreme than those derived from oligonucleotide chip hybridization experiments.

Table 1. Verification of expression patterns of five TCs by using real-time RT-PCR.

	Log2 FC
TC	RT-PCR	Chip
TC31841	1.9	3.0
TC33471	4.4	3.0
TC29161	2.8	1.6
TC35954	–2.2	–1.5
TC40340	1.9	1.4

In several cases, we identified coordinately regulated sets of genes with similar or complementary functions. Of the 42 TCs with a LCB ≥ 2.0, indicating induction by bacteria, 9 have homology to nuclear or nucleolar components necessary for ribosome construction, ribosomal RNA processing, and cell proliferation (TC28762, TC28786, TC29010, TC33689, TC37151, TC39281, TC39286, TC39961, and TC41559). In several cases, multiple TCs were found to encode closely related genes or the same gene. Within 24 h of bacterial inoculation, the root pericycle and inner cortical cells of M. truncatula become activated for cell divisions and contain a round, centrally located nucleus with a large nucleolus (12). In addition, two TCs induced by S. meliloti have homology to signal transduction components that may play a role in embryogenesis (TC29161 and TC33471).

In preparation for infection, the plant must distinguish between symbionts and pathogens. Seven TCs that encode defense-related or stress-protection proteins were induced in response to S. meliloti (TC28391, TC31841, TC35893, TC36043, TC36848, TC38809, and TC39533). Two sequences have motifs suggesting participation in pathways involving the plant hormone ethylene: TC36043 encodes an oxidoreductase with similarity to both flavonol synthases and 1-aminocyclopropane-1-carboxylic acid oxidases, which produce ethylene, and TC28391 encodes a homolog of an ethylene-responsive DNA binding protein (39).

Four TCs had a LCB ≤ -2.0 indicating down-regulation in the symbiosis (TC28746, TC35880, TC35954, and TC36479). TC35880 encodes a previously identified M. truncatula phosphate transporter that is suppressed during colonization by symbiotic mycorrhizal fungi (40).

The Plant Transcriptional Response Is Mediated by the Bacterial Signaling Molecule NF. To determine what bacterial factors are required for eliciting plant gene expression at this stage of the symbiosis, we evaluated transcriptional changes for the 46 previously identified TCs after exposure to bacterial mutants that could not produce NF (SL44 and TJ170), a bacterial mutant that can produce NF but is defective in infection (exoA), or a single application of purified NF. Most of the 46 plant TCs that are differentially expressed in response to S. meliloti did not show differential expression (|FC| ≥ 2.0 and identification by SAM) in response to the SL44 or TJ170 S. meliloti mutants (Fig. 1), suggesting that S. meliloti must make NF to induce these changes. Although four TCs (TC31841, TC37151, TC39281, and TC39286) that are typically induced in response to S. meliloti showed a FC ≤ -2.0 in response to SL44 or TJ170 (Fig. 1), we expect that these changes are not relevant to the symbiosis because they were not identified by SAM, were inconsistent between the two mutants, and show reversed expression patterns when compared with the response to WT S. meliloti.

The exoA S. meliloti mutant is defective in plant invasion and late nodulin induction (16, 17). Because the NF biosynthetic genes are intact in this mutant, we infer that the mutant elicits early plant responses that depend on NF, including calcium spiking and root hair deformation, and that it acts as a continuous source of NF. All 46 TCs that were identified from the response to WT S. meliloti were also differentially expressed in response to the exoA mutant (Fig. 1). These data suggest that continuous application of NF, in the absence of infection, is sufficient for modulating the expression of all 46 TCs.

NF elicits ionic changes and growth behaviors in plant roots but fails to induce tight root hair curling or infection thread formation (41). Treating plant roots with a single dose of purified 100 nM NF elicited a transcriptional response in the 46 TCs similar to that observed in response to S. meliloti inoculation (Fig. 1). However, 35 of the 46 genes showed reduced amounts of transcriptional change in response to NF when compared with S. meliloti. All four TCs that are down-regulated by S. meliloti are also down-regulated by NF (Fig. 1, first four TCs listed). Thirty-one TCs that were up-regulated in response to S. meliloti were also up-regulated in response to NF but were often not induced to the same level by using NF (Fig. 1). Eleven TCs that were induced in response to S. meliloti showed a FC ≤2.0 in response to NF and were not identified by the SAM method; however, with the exception of TC28561, these TCs typically showed low levels (FC ≥ 1.4) of induction by NF (Fig. 1). We suggest that NF also mediates the induction of these genes, because they are not induced by bacterial mutants that cannot make NF and are induced by purified NF to low levels, but that maximal induction requires the presence of both NF and S. meliloti or a continuous application of NF to the plant.

Differential Transcriptional Responses of Plant Nod^- Mutants to S. meliloti. Seven Nod^- plant mutants have been identified that fail to form infection threads upon inoculation and have altered root hair responses to NF or S. meliloti: nfp shows no response to NF; dmi1, dmi2, and dmi3 show root hair swelling; nsp1 and nsp2 show root hair branching (18, 20, 21); and hcl shows a root hair curling response to S. meliloti (22). The first six fail to initiate cortical cell divisions in response to bacteria, whereas hcl exhibits a reduced number of cell divisions compared with WT plants (22). To investigate whether the identified gene expression changes described above correlate with root hair or cell division responses, we monitored gene expression in Nod^- plant mutants inoculated with WT S. meliloti.

The six Nod^- plant mutants that are defective in root hair curling or the initiation of cell divisions (nfp, dmi1, dmi2, dmi3, nsp1, and nsp2) failed to exhibit any obvious (|LCB| ≥ 2.0) transcriptional responses to S. meliloti (data not shown). Whereas gene expression changes of the 46 identified genes showed a |LCB| ≥ 2.0 in the WT plants, none of these genes showed a |LCB| ≥ 1.5 or a |FC| ≥ 2.0 with identification by the SAM method in the six mutant lines (Fig. 1 and data not shown). However, in hcl, 25 of the 46 TCs showed a |FC| ≥ 2.0, and were identified by the SAM method. Many of the remaining TCs, including six of the nine identified TCs encoding nuclear or nucleolar-localized proteins (TC28786, TC29010, TC28762, TC33689, TC39961, and TC41559), showed less extreme changes than those observed in WT plants (Fig. 1). We interpret these data to indicate that at least some of the machinery responsible for triggering gene expression changes is functional in hcl.

Discussion

In the development of a new organ, transcriptional changes are expected to accompany morphological changes. In the symbiosis between legumes and rhizobia, we additionally expect to observe transcriptional changes important for bacterial recognition and invasion. We have identified a set of 46 M. truncatula TCs that are differentially expressed during the first 24 h of exposure to the symbiotic partner S. meliloti. Using the expression of these TCs as an assay for plant perception of bacterial signals, we have shown that in this early phase of the symbiosis, the bacterial signal NF is the principal signaling molecule eliciting plant gene expression changes. Further, we have demonstrated that the initiation of the plant transcriptional response requires at least six plant gene products, previously demonstrated to control a wide variety of physiological responses to NF.

Eight plant TCs that encode nucleolarly localized proteins were up-regulated within 24 h of S. meliloti inoculation and play potential roles in ribosome biogenesis. TC41559 encodes the 60S subunit of the ribosome. Five TCs encode proteins important for rRNA processing (TC29010, TC28762, TC33689, TC39961, and TC37151). The processing of rRNA involves cleavage of the primary transcript and modification of nucleotides by methylation or pseudouridylation. The Drosophila melanogaster minifly gene (TC29010 and TC37151) is important for proper cleavage of the primary transcript and for pseudouridylation (42). The Xenopus laevis brix gene (TC28762) is also important for cleavage (43) and fibrillarin (TC39961) is important for methylation (44). Two TCs (TC39281 and TC 39286) encode nucleolin, which has several roles in ribosome biogenesis including preribosome synthesis and the transport of preribosomal particles from the cytoplasm to the nucleus (45). These nucleolar proteins are typically induced in rapidly dividing cells, such as tumor cells; several mammalian tumor suppressor genes affect ribosome formation (46).

We identified several genes that are involved in plant–microbe or plant–pathogen interactions, suggesting that the plant employs common proteins in these responses. In addition to the S. meliloti symbiosis, M. truncatula also forms a symbiosis with mycorrhizal fungi to increase phosphate uptake from the soil; some of the signaling components important for nodulation are also important for mycorrhization (47). A gene encoding a phosphate transporter is suppressed in the plant upon S. meliloti inoculation; this gene is also suppressed upon mycorrhizal colonization (40), indicating shared regulatory steps between the two symbioses. We also identified a homolog of ferritin that is induced in response to S. meliloti (TC39533); ectopic expression of ferritin results in a tolerance to oxidative damage and pathogens (48). Finally, the sugar beet (Beta vulgaris L.) nematode resistance gene is induced in response to nematode feeding (49, 50) and is also induced in response to S. meliloti (TC38809).

We propose that NF is the main signal to which plants respond within the first 24 h of symbiotic development. Differential expression of all 46 TCs requires that bacteria produce NF (because bacteria that cannot make NF do not elicit these changes), and purified NF alone induces most early transcriptional responses in the plant. NF and S. meliloti both elicit calcium spiking, transcriptional changes, and root hair deformations of the plant (2, 6, 7). In M. sativa, NF triggers the formation of bacteria-free pseudonodules, indicating an ability to provoke cortical cell divisions (11). The local application of NF can cause directional changes in root hair growth resulting in root hair curling (51). Our results suggest that many of the responses to S. meliloti may be associated with bacterial binding or continuous localized delivery of NF; however, the initiation of the transcriptional program leading to infection may be triggered by NF alone.

Because NF triggers multiple responses in the plant, and Nod^- plant mutants are blocked at different stages of the establishment of the symbiosis, we hypothesized that the Nod^- plant mutants would exhibit a hierarchy of transcriptional changes in response to S. meliloti. Surprisingly, six of seven M. truncatula Nod^- mutants showed no measurable transcriptional changes in response to bacterial inoculation. These data suggest that all six of the genes identified through genetic analysis are required for transducing the NF signal into a transcriptional response at 24 h. Indeed, the sequence of the cloned DMI2 and DMI3 genes, which both have kinase domains, suggests a role in signal transduction (52–54). Identification of the NSP1 and NSP2 genes may be important for determining the subsequent signal transduction steps leading to new transcription. Based on our data, we expect that the initial calcium spiking, root hair swelling, and root hair deformation responses do not require gene expression changes. However, transcriptional studies of the hcl mutant suggest that later stages of symbiotic development, involving root hair curling and cell divisions, require new gene expression.

Our experimental goal was to discover a representative set of genes altered in expression during the establishment of the symbiosis. These genes were used to successfully test the effects of different stimuli on the plant and dissect the response of plant mutants to bacterial signals. The construction of the array and the technical limitations of the analysis suggest that other genes may be found to respond transcriptionally to bacteria in the early phase of the symbiosis. First, the oligonucleotide chip was constructed from EST data and represents only a fraction of the entire M. truncatula genome. Second, we developed stringent criteria to identify differentially expressed M. truncatula TCs with high confidence, thus eliminating genes with weak or inconsistent expression changes. Finally, we assayed gene expression at a single time point, and under one set of environmental conditions, potentially missing genes that are transiently expressed before or after 24 h.

Our examination of gene expression in M. truncatula has identified plant genes that are differentially expressed during the establishment of the M. truncatula/S. meliloti symbiosis. We identified genes that are differentially expressed in response to both S. meliloti and to the purified signaling molecule NF. We also genetically defined, using plant and bacterial mutants, at what point in the development of the symbiosis transcriptional changes are triggered. Further, our studies have identified induced developmental and biochemical pathways for future study regarding the role of these pathways in the establishment of the M. truncatula/S. meliloti symbiosis.