The MtMMPL1 early nodulin is a novel member of the matrix metalloendoproteinase family with a role in Medicago truncatula infection by Sinorhizobium meliloti.

We show here that MtMMPL1, a Medicago truncatula nodulin gene previously identified by transcriptomics, represents a novel and specific marker for root and nodule infection by Sinorhizobium meliloti. This was established by determining the spatial pattern of MtMMPL1 expression and evaluating gene activation in the context of various plant and bacterial symbiotic mutant interactions. The MtMMPL1 protein is the first nodulin shown to belong to the large matrix metalloendoproteinase (MMP) family. While plant MMPs are poorly documented, they are well characterized in animals as playing a key role in a number of normal and pathological processes involving the remodeling of the extracellular matrix. MtMMPL1 represents a novel MMP variant, with a substitution of a key amino acid residue within the predicted active site, found exclusively in expressed sequence tags corresponding to legume MMP homologs. An RNA interference approach revealed that decreasing MtMMPL1 expression leads to an accumulation of rhizobia within infection threads, whose diameter is often significantly enlarged. Conversely, MtMMPL1 ectopic overexpression under the control of a constitutive (35S) promoter led to numerous abortive infections and an overall decrease in the number of nodules. We discuss possible roles of MtMMPL1 during Rhizobium infection.

Rhizobium infection and nodulation are the two key developmental processes involved in establishing the nitrogen-fixing symbiosis between legume plants and the appropriate soil microsymbiont. Both of these processes require Nod factors, lipo-chito-oligosaccharidic molecules that are synthesized by rhizobia in response to root-secreted flavonoids and play an essential role in the activation of symbiotic genetic programs in specific host plants (for review, see Oldroyd and Downie, 2004;Riely et al., 2004;Stacey et al., 2006). The plant responses to Nod factors and rhizobia integrate various internal cues, notably nutritional (particularly regarding carbon and nitrogen status), developmental (only specific root regions are responsive), and autoregulatory signals. Indeed, both the location and the number of nodules and infections are tightly controlled by the plant by at least two mechanisms, either ethylene dependent or independent, which lead to a superinfection/supernodulation phenotype when altered by mutations (skl, affected in ethylene perception [Penmetsa and Cook, 1997;Penmetsa et al., 2003]; har1 and sunn, affected in a Leu-rich repeat receptor kinase gene [Krusell et al., 2002;Schnabel et al., 2005]).
In Medicago truncatula, as in many other temperate legumes, Rhizobium infection takes place via plant cell structures called infection threads (ITs), which form in curled, infected root hairs in the presence of Rhizobium, probably following Nod factor-dependent cytoskeleton reorganization (for review, see Brewin, 2004;Gage, 2004). The IT initially develops an invagination of the root hair wall, thereby producing an inwardly growing cylinder of wall material bounded by a membrane and containing the bacteria embedded in a matrix. In addition to bacteria and secreted bacterial products, such as exopolysaccharides (EPSs) and lipopolysaccharides (LPSs), the IT lumen contains many plant components in common with the extracellular matrix (ECM; Rae et al., 1992;Wisniewski et al., 2000;Rathbun et al., 2002), such as extensins and other plant-derived glycoproteins that certainly modulate its physicochemical properties.
In legumes with indeterminate nodules, like M. truncatula, ITs first grow inward as a branched network from the root hairs to the newly divided cortical cells of the nodule primordium, following the pathway created by preformed cytoplasmic columns (named pre-ITs) in aligned activated cells. Once the nodule meristem differentiates within this primordium (about 3-4 d postinoculation [dpi] in M. truncatula A17) and the nodule starts to grow continuously out from the root, new IT branches grow outward and behind this apical meristem (Monahan-Giovanelli et al., 2006). The newly divided cells generated by the nodule meristem (zone I) are then infected within the adjacent so-called infection zone II, where bacteria are liberated from unwalled outgrowths of ITs termed infection droplets (Brewin, 2004). The mechanism for bacterial release within nodules might partly overlap with that proposed for IT initiation and progression, and depend on localized plant cell wall degradation and accumulation of osmotically active compounds leading to swelling of the IT droplet (for review, see Timmers et al., 2005). After release, rhizobia, now called bacteroids, remain surrounded by a plant membrane, the peribacteroid (or symbiosome) membrane, with features of the host plasma membrane and additional specific components. Nitrogen fixation takes place in the nodule zone III and requires the coordinated differentiation of both bacteroids and their host plant cells. A senescent zone IV, in which nodule cells and bacteroids undergo degradation, forms in older nodules proximally to zone III, usually about 4 weeks postinoculation in M. truncatula A17.
A number of infection-defective mutants have been identified in the macro-and microsymbionts, showing that both partners are involved in this process. Plant mutants affected in early infection stages (IT initiation or progression) have been reported in pea (Pisum sativum; Tsyganov et al., 2002), Lotus japonicus (Bonfante et al., 2000;Tansengco et al., 2003;Yano et al., 2006), and M. truncatula (Catoira et al., 2000;Limpens et al., 2003;Kuppusamy et al., 2004). From these, only lyk3 has been cloned, encoding a putative Nod factor receptor kinase (possibly a component of the so-called entry receptor; Limpens et al., 2003). Plant mutants affected in bacterial release have also been described in pea (Tsyganov et al., 1998;Morzhina et al., 2000), L. japonicus (Imaizumi-Anraku et al., 1997), and M. truncatula . Because most of the corresponding genes have not yet been cloned, important information about the way the plant controls ITs is still lacking. In addition, RNAi constructs knocking down the expression of a SymRK receptorlike kinase were also shown to alter the bacterial uptake process in Sesbania rostrata (Capoen et al., 2005) and M. truncatula (DMI2; Limpens et al., 2005).
Other sources of information rely on the identification of plant genes whose expression is up-regulated during Rhizobium infection in roots or in the nodule infection zone, such as MtLEC4 (Mitra and Long, 2004), MtENOD11, and MtENOD12 (Pichon et al., 1992;Journet et al., 2001); MtN1 and MtN6 (Gamas et al., 1996or PsRNE1 (Rathbun et al., 2002). However, the lack of mutants (or knock-down constructs) in these genes has so far hampered direct investigations of their function. We show here that MtMMPL1, a M. truncatula gene identified by transcriptomics, also accompanies the Rhizobium infection process. Our attention was attracted to this gene because it encodes a putative protein belonging to the large matrix metalloendoproteinase (MMP) family, of particular interest in relation to developmental processes, notably invasion related.
MMPs are structurally related zinc (Zn)-containing endopeptidases thought to play a key role in the breakdown of the ECM. Their importance is well documented in animals in the context of many normal biological processes involving remodeling of connective tissues (e.g. embryonic development, organ morphogenesis, angiogenesis, wound healing, and apoptosis), as well as pathological processes (notably arthritis, cancer metastasis, and inflammation; for review, see Nagase and Woessner, 1999). In animals, MMP expression is tightly regulated both at the transcript level (by growth factors, hormones, chemical agents, or physical stress) and protein level (by the control of their activation from precursor proteins and also by endogenous inhibitors; Nagase and Woessner, 1999). Several MMPs have been described in plants, with different but still unclear functions and without determining their physiological substrates. The soybean (Glycine max) SMEP1 protein was the first plant MMP to be purified (Graham et al., 1991;McGeehan et al., 1992) and found to be transcribed only in mature leaves (Pak et al., 1997). Five MMP genes (At1/5-MMP) were then shown to be expressed with different patterns in Arabidopsis (Arabidopsis thaliana), with no hint of their possible function (Maidment et al., 1999). An insertion mutant in At2-MMP was later characterized and exhibited alterations in plant growth and development, notably with late flowering and early senescence (Golldack et al., 2002). Another MMP gene, Cs-1MMP, was identified in cucumber (Cucumis sativus) from a screen of genes up-regulated during programmed cell death in cotyledons (Delorme et al., 2000). Finally, a soybean gene, GmMMP2, was found to be activated during compatible and incompatible interactions with the oomycete pathogen Phytophtora sojae or the bacterial pathogen Pseudomonas syringae pv glycinae and suggested to be involved in a novel defense mechanism (Liu et al., 2001).
Considering the importance of ECM and tissue remodeling during nodulation and infection, we decided to further characterize MtMMPL1 as an interesting example of a plant MMP and to make use of reverse genetics tools recently made available for M. truncatula to explore its possible function during symbiosis.

MtMMPL1: A Novel Variant of the MMP Family
MtMMPL1 is a M. truncatula nodulin gene that was first identified by a subtractive screen of a young nodule cDNA library (designated at that time MtN9; Gamas et al., 1996). Subsequently, MtMMPL1 was reidentified following large-scale EST sequencing approaches, as a cluster of nodule-specific ESTs (designated MtC40019 in the Medicago EST Navigation System [MENS] database and TC95584 in The Institute for Genomic Research [TIGR] MtGI database), and by macroarray/microarray analyses of nodule versus root gene expression (El Yahyaoui et al., 2004). BLASTN interrogation against other M. truncatula EST clusters indicated that MtMMPL1 is a member of a multigene family (see MENS database, http://medicago.toulouse.inra.fr/ Mt/EST/), which was confirmed by Southern analysis (data not shown), while BLASTX/P searches against Swiss-Prot (SP)/TrEMBL protein databases revealed homologies with the MMP family. Because the initial BLAST analyses suggested that available MtMMPL1 cDNA and ESTs corresponded to truncated transcripts, we carried out a 5# RACE approach. We thus obtained a 1,290-nucleotide-long cDNA, likely to be full size considering the size of the MtMMPL1 transcript detected by northern analysis. The corresponding predicted open reading frame begins 15 bp from the 5# end and encodes a 316-amino acid-long putative protein. Its closest homolog in the SP/TrEMBL databases is Gm-MMP2 (62% conserved amino acids, E value of 10 243 ), a soybean MMP transcriptionally induced in response to pathogenic infections (Liu et al., 2001). More generally, as shown in Figure 1, MtMMPL1 exhibits several hallmarks of the MMP family (Massova et al., 1998). Indeed, it has the typical prepro-enzyme structure found in Zn endopeptidases with: (1) a signal peptide (probability of 0.991, as determined with SignalP3.0; Bendtsen et al., 2004) predicted to be cleaved between positions 25 and 26; (2) a propeptide containing the highly conserved octapeptide P-R-C-G-V-P-D-I called the Cys switch or autoinhibitory region, which chelates the active site Zn ion and is cleaved to generate an active metalloproteinase; and (3) a sequence similar to the MMP catalytic domain, comprising three Zn-binding His residues accompanied by a ''Met turn'' motif responsible for the structural integrity of the Zn-binding site.
In all known MMP catalytic domains, a Glu residue is found next to the first His residue. This Glu is essential for protease activity because it acts as a nucleophilic group promoting attack of a water molecule on the substrate. A Gln was found instead of a Glu in MtMMPL1, and this is why MtN9 was renamed MtMMPL1 (for MMP-like 1). A Glu-to-Gln substitution generated by site-directed mutagenesis in animal MMPs (human gelatinase A and MMP-9) resulted in a dramatic decrease in protease activity (Crabbe et al., 1994;Rowsell et al., 2002). We therefore verified that the Glu-to-Gln substitution in MtMMPL1 did not result from a cloning or a sequencing artifact. We examined MtMMPL1 ESTs from independent cDNA libraries as well as MtMMPL1 genomic DNA sequences, obtained both by PCR amplification and from the ongoing M. truncatula genome sequencing program. Although the sequencing and annotation of the bacterial artificial chromosome carrying the MtMMPL1 gene region (CR962135, chromosome 5) was not completed (phase 2 clone, see TIGR Web site, http://www.tigr.org/tdb/ e2k1/mta1/), it was possible to conclude that MtMMPL1 comprises a single exon. Moreover, we discovered that MtMMPL1 belongs to a cluster of four related genes that all contain the Glu-to-Gln substitution, present within a region of about 13 kb flanked by transposonlike sequences (see Supplemental Fig. S1 for an alignment of the corresponding predicted proteins).

The Active Site Glu-to-Gln Variant Is Found in Other Legume MMP-Like Putative Proteins
Because the Glu-to-Gln variant had been confirmed in several M. truncatula MMP-like proteins, we then looked for similar variants in other species, using public nucleotide and protein databases. We first conducted a ScanProsite search on the SP and TrEMBL databases (release 50.2 and release 33.2, respectively) to identify proteins containing the following motif: To explore a larger range of species, we then carried out a TBLASTN search on dbEST (36,649,443 sequences), using a 31-amino acid sequence centered on the MMP active site (WDLETVAMHQIGHLLGL-DHSSDVESIMYPTI). A total of 502 hits were found in ESTs from 52 species (seven animals and 45 plants, including trees, monocots, and dicots), among which 101 corresponded to the variant (Gln) site. These Gln variants were found exclusively in legume species (one from Trifolium pratense, 14 from M. truncatula, and 86 from soybean). These came from 17 cDNA libraries (one from T. pratense, 10 from M. truncatula, and six from soybean) and often, but not always, corresponded to stress responses (e.g. response to salicylic acid, with 19 soybean ESTs). Legume MMP ESTs with a nonvariant (Glu) site were also found but overall with a lower frequency (one from T. pratense; six from M. truncatula, but not in the libraries where the Gln variants were found; 31 from soybean). In addition, two other active site variants were found at a very low frequency among soybean ESTs, with one Glu-to-Pro and one Glu-to-Asp substitution.
A phylogenetic tree analysis was carried out on all predicted plant MMP proteins or protein fragments containing at least the C switch and the catalytic domain (in total, 38 sequences from 18 plant species). In addition, this analysis included eight representatives of animal MMPs belonging to the same subfamily (M10B in PROSITE nomenclature). To compare fragments of similar lengths and therefore generate a more reliable tree, only sequences spanning the region from the Cys switch to the Met turn were considered. Figure  2 presents the resulting tree, showing clearly that the eight animal MMPs form a subgroup separated from the plant MMPs and that MtMMPL1 stands in a group of closely related proteins, which contains the six (Gln) legume MMP variants as well as GmMMP2. The fact that the soybean SMEP1 protein does not belong to this group indicates that this group does not simply correspond to legume sequences and suggests a very distinct function for SMEP1. None of the proteins of this group possesses the binding site for a calcium ion and a second Zn ion, known as structural Zn, found in many MMPs (Massova et al., 1998). The precise function of this site in plant MMPs is unclear, however, because the metalloproteinase activity of GmMMP2 was experimentally confirmed (Liu et al., 2001). All plant MMPs identified so far carry one conserved Glu residue located just upstream of the predicted catalytic site ( Fig. 1), never found in animal MMPs (generally carrying a Phe or a Leu residue at this position). The fact that this residue seems to be conserved may indicate that it is part of the active site in plants.

MtMMPL1 Transcription Is Specifically Associated with Sinorhizobium meliloti Infection
Using quantitative reverse transcription (qRT)-PCR analysis, we compared MtMMPL1 transcript levels in symbiotic and nonsymbiotic conditions. MtENOD11, a repetitive Pro-rich protein gene induced by purified Nod factors and by S. meliloti infection (Journet et al., 2001;Boisson-Dernier et al., 2005), was used as a positive control for all these qRT-PCR studies.
Consistent with MtMMPL1 EST distribution, we could not detect significant MtMMPL1 transcription in M. truncatula shoots, stems, flowers, or seed pods (data not shown), while we confirmed a strong MtMMPL1 induction in young nodules (Fig. 3A). Using this sensitive method, MtMMPL1 expression could not be detected in Nod factor-treated root samples, in contrast to MtENOD11 (data not shown). MtMMPL1 transcripts were detected in wild-type M. truncatula roots at 3 dpi with S. meliloti, while practically undetectable The Cys switch motif, the structural Zn-binding site, and the catalytic Zn-binding site are indicated by green, blue, and red rectangles, respectively. The alignment was done with Multalin (Corpet, 1988). The consensus sequences corresponding to the predicted Cys switch, the structural Zn-/calcium-binding site, and the catalytic Zn-binding site are indicated by green, blue, and red boxes, respectively. *, Conserved His residues involved in Zn binding within the catalytic domain and the structural Zn-/calcium-binding site. The arrow points to the Glu residue found to be replaced by a Gln in MtMMPL1 and Gm-c1017. The arrowhead indicates a Glu residue that is conserved in plant MMPs and not found in animal MMPs. The region chosen for the phylogenetic tree analysis shown in Figure 2 is indicated by a dotted line (from the Cys switch to the catalytic Zn-binding site).
at 1 dpi and in noninoculated roots (Fig. 3B). MtMMPL1 transcripts were about 5 to 14 times less abundant than MtENOD11 mRNA in infected root and nodule samples. Figure 3B also shows that, at 3 dpi, MtMMPL1 expression was much more strongly induced in the hypernodulating M. truncatula mutants sunn (TR122 allele; Sagan et al., 1995) and skl (Penmetsa and Cook, 1997) than in wild-type M. truncatula. However, MtMMPL1 was about 3-fold less expressed in skl than in sunn, which contrasts with MtENOD11, which is transcribed at a similar level in both mutants. Another difference between these two genes is that MtMMPL1 was not expressed during symbiotic interactions between M. truncatula and the arbuscular Figure 2. Phylogenetic tree of plant and selected animal MMP fragments. Bootstrap values are indicated as percentages. All analyzed fragments are of similar size and span the conserved region from the Cys switch to the Met-turn motif (Fig. 1). Available short plant MMP sequences not containing both of these motifs were not considered. The animal MMP representatives were taken from the M10B family in the Prosite nomenclature, which also contains all plant MMPs. The variants with a Gln residue in the predicted active site are underlined. These belong to two closely related subgroups (circled with a continuous line). All analyzed animal MMPs belong to one group, shown by a dotted line.  (Journet et al., 2001; data not shown).
To determine more precisely the spatio-temporal pattern of MtMMPL1 expression, we then carried out in situ hybridizations (ISHs) with 35 S-radiolabeled MtMMPL1 antisense riboprobes (Fig. 4). As expected, hybridization with the MtMMPL1 sense probe control did not give a detectable signal (data not shown). To study early infection stages, we carried out ISH on sections of roots that had been spot inoculated with S. meliloti and harvested at 2 dpi. At this stage, under our experimental conditions, ITs just begin to form in infected roots hairs, concomitantly with cell divisions in the inner root cortex. Figure 4, A and B, shows an example of an epidermal cell bearing an IT at an early stage of development, to which MtMMPL1 transcripts were found to be specifically associated in several serial sections. No MtMMPL1 hybridization signal was observed in noninfected dividing cortical cells. Once the sequence of the MtMMPL1 genomic region became available, we could confirm this expression pattern in infected roots, using a promoterTGUS fusion generated with a 2.4-kb region upstream of the MtMMPL1 start codon and introduced in M. truncatula by Agrobacterium rhizogenes-mediated root transformation (Fig. 4C).
In young nodules, MtMMPL1 transcripts located to inner nodule tissues where ITs develop and ramify within newly divided cells (Fig. 4, D-F), as previously observed with several other early nodulin genes (de Carvalho Niebel et al., 1998;Gamas et al., 1998). In older, well-differentiated nodules, MtMMPL1 transcripts were only detected in the infection zone II where bacteroids are liberated from ITs and it could clearly be seen at high magnification that the hybridization signal was closely and exclusively associated with ITs (Fig. 4, G-I).
To confirm that MtMMPL1 expression is tightly correlated to S. meliloti infection, we then analyzed situations where infection was impaired due to various mutations in bacterial or plant symbiotic genes/ locus. Thus, we used an exoA mutant of S. meliloti, defective for EPS production, which shows IT abortion in root hairs and elicits small empty nodules (Yang et al., 1994). MtMMPL1 induction was drastically reduced in roots at 3 dpi, as judged by qRT-PCR (Table I) or in isolated 10-d-old nodules (microarray analyses; S. Moreau and P. Gamas, unpublished data). We suggest that the low level of residual MtMMPL1 expression was associated with abortive infections. In contrast, almost normal MtMMPL1 transcript levels were found in nodules induced by bacA and fixLJ mutants of S. meliloti, which are defective in nitrogen fixation but not infection defective (S. Moreau and P. Gamas, unpublished data).
We also carried out qRT-PCR analysis of MtMMPL1 expression in three early symbiotic M. truncatula mutants altered in the infection process, namely, nsp1 (B85 allele), hcl (B56 allele), and lin (D8 allele). The nsp1 mutant is unable to trigger cortical cell division and is defective in root hair curling and IT formation in response to S. meliloti (ccd 2 , hac 2 , and inf 2 phenotype, respectively), but exhibits normal root hair deformation and very limited MtENOD11 induction in response to Nod factors (Catoira et al., 2000). The hcl mutant is hac 2 ccd 1 (Catoira et al., 2001) and the lin mutant is ccd 1 hac 1 , but all infections are arrested within root hairs where ITs fail to elongate (Kuppusamy et al., 2004). We found that, following S. meliloti inoculation, MtENOD11 was induced in lin and hcl roots (Table I) but not in nsp1. Using the same samples, we were unable to detect MtMMPL1 expression in nsp1 and hcl, while in lin expression was around 10-fold lower than in wild-type M. truncatula and similar to the level obtained in wild-type M. truncatula in response to S. meliloti exoA (Table I).
Taken together, these results lead us to conclude that MtMMPL1 expression is specifically associated with S. meliloti infection and triggered at the onset of IT formation. . The level of expression was monitored by real-time PCR analysis using isolated wild-type M. truncatula nodules at 4, 10, and 14 dpi, compared to noninoculated nitrogen-starved root samples (time point 0; A). B, Gene expression in S. meliloti-inoculated roots (at 0, 1, and 3 dpi) is compared between wild-type M. truncatula and sunn and skl supernodulant mutants. All values were normalized using an EF-1a housekeeping gene as an internal control. The error bar depicts the variation between two biological repetitions and two technical repeats.

RNA Interference and Ectopic Overexpression of MtMMPL1
We further investigated the functional role of MtMMPL1 by an RNAi approach. An RNAi construct, covering most of the MtMMPL1 translated and 3# untranslated regions, was cloned into pRedRoot II. This vector, derived from pRedRoot (Limpens et al., 2004), allows transformed roots to be selected both by their resistance to kanamycin and their expression of the fluorescent DsRED1 protein. This RNAi construct led to an approximately 4-fold reduction in MtMMPL1 transcript levels by comparison with roots transformed with the empty vector roots (at 6 dpi with S. meliloti), as estimated by qRT-PCR analysis of pooled transformed roots (Fig. 5A). MtMMPL1 belongs to a multigene family, but, as judged by ESTs, no other family member is expressed at a comparable level in young nodules or nodulated roots. The phenotype conferred by the RNAi construct is thus likely to result from the alteration of MtMMPL1 expression.
The symbiotic competence of MtMMPL1 RNAi roots was compared with that of control roots (empty vector transformed). Similar kinetics of nodulation and nodule numbers were observed (Fig. 6A). MtMMPL1 RNAi nodules were clearly functional because plants C and F show the histochemical revelation of a pMtMMPL1TGUS fusion, while other sections correspond to ISH analysis with 35 S-labeled MtMMPL1 riboprobes (7-mm-thin sections). B, E, and H, Dark-field microscopy, with the hybridization signal appearing as white dots. A, D, G, and I, Bright-field microscopy, with the signal appearing as dark dots. A and B, Spot-inoculated roots, 2 dpi. C, S. meliloti-inoculated root (6 dpi). D to F, Four-day-old nodules. G to I, Twenty-day-old nodules, with I being a magnification of a zone II region. ITs are shown by arrows and root cortical cell divisions by asterisks. Note that in all cases hybridization signals are very close to ITs. Bars 5 50 mm. grew vigorously for several weeks in a medium without combined nitrogen. In fact, these nodules were slightly larger (about 1.3-fold) than control nodules in three out of the four experiments performed. However, while the overall structure of nodules appeared normal, vibratome (50 or 70 mm) and semi-thin (4 mm) nodule sections revealed that, within the infection zone, ITs were on average significantly larger in diameter in comparison to those found in control nodules (Fig. 7, A and B). These enlarged ITs were not accompanied by obvious signs of plant defense reactions such as autofluorescence and thickening of IT walls. Electron microscopic observations of ultra-thin sections (Fig. 7, C and D) confirmed the presence of very abundant (but otherwise apparently normal) bacteria filling these enlarged ITs. Bacteroid differentiation and bacterial release appeared normal in the nodule zones II and III, associated with type 1 to type 4 bacteroids (Vasse et al., 1990).
To explore whether IT enlargement was accompanied by a difference in bacterial viability, we counted the bacteria recovered from crushed nodules (following surface sterilization). Using nodules harvested at 21 or 44 dpi, we found a substantial increase in the number of bacterial colonies recovered from MtMMPL1 RNAi nodules in comparison to control nodules (Table II).
We also examined the phenotype of transgenic roots expressing MtMMPL1 constitutively under the control of the 35S promoter, which resulted in a more than a 1,000-fold increase in MtMMPL1 transcript accumulation in noninoculated roots (Fig. 5B). Under these conditions, the number of S. meliloti-induced nodules was decreased by about 60% (Fig. 6B). These nodules were fully functional and indistinguishable from control nodules in terms of structure or number of viable rhizobia. However, small regions of the root displayed a large number of infections aborting in the epidermis, as shown in Figure 7E, which was never seen in control roots (Fig. 7F). It thus seems that expressing MtMMPL1 before the initiation of S. meliloti infection leads to a decrease in the number of productive infections.

DISCUSSION
In this study, we investigated the possible role in the nitrogen-fixing symbiosis of a nodulin gene, MtMMPL1, which encodes a member of a family of proteins frequently involved in ECM and tissue remodeling.

MtMMPL1, a Novel and Specific Marker for S. meliloti Infection
The combined use of ISH, a promoterTGUS fusion, and infection-defective mutants allowed us to establish that MtMMPL1 transcription accompanies IT development from root hairs to the nodule primordium and then subsequently in the nodule infection zone. MtMMPL1 was not found to be induced by purified Nod factors, and during the preinfection stage, unlike MtENOD11, a well-characterized early nodulin gene also activated by rhizobial infection. MtENOD11 is activated before the formation of ITs (Boisson-Dernier et al., 2005), as shown with a nodF nodL S. meliloti mutant (Ardourel et al., 1994;Catoira et al., 2000), which elicits root hair curling but no IT initiation. In contrast, we have no indication that MtMMPL1 is induced before IT formation because MtMMPL1 was not induced by the nodF nodL mutant in a M. truncatula sunn background (TR122 allele), which amplifies many nodulin gene responses (data not shown). Finally, MtMMPL1 was substantially less induced in a skl  1.4 10 23 6 1.3 10 23 7.6 10 23 6 1.6 10 23 1.9 10 22 6 8.8 10 23 5.2 10 25 6 3.5 10 25 7.0 10 24 6 6.9 10 24 1.1 10 23 6 9.0 10 24 than in a sunn mutant background, which is not the case for MtENOD11. MtMMPL1 thus shows both common and distinct features in comparison to MtENOD11 and is therefore quite complementary, e.g. for characterizing the phenotype of early symbiotic plant mutants or for identifying cis-and trans-regulatory elements controlling infection-induced genes.

MtMMPL1 Has an Impact on IT Structure and Rhizobium Multiplication
Nodules with enlarged ITs and more viable rhizobia were observed when MtMMPL1 expression was decreased by RNAi. The bacteria recovered on plates from crushed nodules probably corresponded mostly to those present in ITs because differentiated bacteroids found in indeterminate nodules are unable to divide when cultured (Mergaert et al., 2006).
There are several reports of enlarged ITs, either due to Rhizobium or plant host mutations. Thus, alterations in the EPSs or in the LPSs of various rhizobia (R. leguminosarum bv viciae, S. meliloti, Azorhizobium caulinodans) lead to the formation of enlarged ITs with thickened cell walls, often associated with plant defense reactions, and to the production of ineffective nodules in their plant host (pea, M. truncatula, and S. rostrata, respectively; Niehaus et al., 1993Niehaus et al., , 1998Perotto et al., 1994;Mathis et al., 2005). Rhizobium surface polysaccharides are thought to play an important role in plant-bacteria communication, for the invasion process via IT (EPSs) and after the bacterial release stage  (LPSs), possibly to evade plant immune responses (for review, see D' Haeze and Holsters, 2004). To our knowledge, viability of nodule bacteria was determined only in the study of an LPS-and EPS-defective mutant of A. caulinodans, where about 100-fold fewer viable bacteria were recovered from S. rostrata 2-monthold nodules than with wild-type A. caulinodans (Mathis et al., 2005). Altered IT structure has also been reported for several plant mutants. Thick abortive ITs were observed in root hairs of infection-defective mutants in L. japonicus (Yano et al., 2006), pea (Sagan et al., 1994), and M. truncatula (TE7 mutant: Benaben et al., 1995; rpg mutant: J.-F. Arrighi and C. Gough, personal communication). Enlarged ITs were also found in nodules exhibiting bacterial release defects, as with Mtsym6 and nip mutants or DMI2 RNAi lines in M. truncatula (Tirichine et al., 2000;Veereshlingam et al., 2004;Limpens et al., 2005), or premature degradation of nodule tissues, as with Pssym33, Pssym40, or Risfix pea mutants (Novak et al., 1995;Tsyganov et al., 1998).
Enlarged ITs due to MtMMPL1 RNAi differed strikingly from these cases because they were found in fully functional nodules with an increased Rhizobium accumulation and without any sign of plant defense responses. It could be argued that the MtMMPL1 RNAi used here corresponded to a weak mutant allele, for which an altered IT structure was perhaps the most easily observed phenotype. Bacterial accumulation in ITs could thus result from nonoptimal bacterial release but without problems of viability because these bacteria were wild type. Bacterial surface components are  (B and D) or in 35STMtMMPL1 roots (E). A, C, and F, Control samples, transformed with an empty cloning vector. A and B, Bright-field microscopy of 4-mm sections. C and D, Electronic microscopy of a section made in the nodule infection zone. E and F, Bright-field microscopy of S. meliloti hemATlacZ-inoculated roots at 7 dpi after b-galactosidase activity detection (note the numerous abortive infections in E, in comparison to F). Transverse and longitudinal sections of ITs are shown by pink and black arrows, respectively (note the enlarged ITs in B and D); green arrowheads point to nucleoli. Bars 5 50 mm, except in C and D (5 mm).
indeed likely to be important for Rhizobium protection (D' Haeze and Holsters, 2004) against the harsh conditions encountered in ITs, notably with abundant reactive oxygen species (Pauly et al., 2006) and possibly also defensin-like and thionin-like proteins Mergaert et al., 2003;Silverstein et al., 2005; see below). Alternatively, it may be that rhizobial accumulation in MtMMPL1 RNAi nodules was due to a change in IT structure or internal composition, making it more hospitable for bacteria.

Possible Roles of the MtMMPL1 MMP-Like Protein during Rhizobial Infection
Considering the preproprotein structure of MtMMPL1 and the localization of the corresponding transcript, MtMMPL1 is probably exported into the IT cell wall or lumen. The lumen matrix shares many components with the ECM (Rae et al., 1992;Rathbun et al., 2002), and therefore it is interesting to have identified a MMP-like nodulin that can influence IT structure. Because MtMMPL1 is expressed as soon as ITs are formed, it is likely to play a role in IT growth or to be induced by this process. In the previous section, we proposed two alternative ways in which an MMP could influence rhizobial infection.
First, MtMMPL1 could play a positive role in infection, contributing to the formation of the IT cell wall or ECM. Decreasing MtMMPL1 expression could lead to a modified IT structure with an indirect effect on bacterial accumulation. Major proteins in the IT ECM are extensin-like Hyp-rich glycoproteins (HRGPs), hypothesized to play a role both in polar growth of the IT and in the control of bacterial divisions, which takes place only at the IT tip where the ECM is not rigidified by HRGP cross-linking (Wisniewski et al., 2000). The fact that MtMMPL1 production needs to be tightly coordinated with other IT components could explain why overexpressing MtMMPL1 under the control of the 35S promoter led to frequent infection defects.
Second, the primary role of MtMMPL1 could be to control the number of infecting bacteria, for two reasons: first, no obvious defects in IT growth or bacterial release were observed in MtMMPL1 RNAi roots, and, second, the phylogenic tree analysis showed a higher level of homology with an MMP associated with a stress/defense reaction (Gm-MMP2) than with other plant MMPs involved in developmental processes (e.g. At2-MMP). In this respect, we note that MtMMPL1like ESTs come from legume stress-response libraries (such as the response to salicylic acid in soybean). The legume host may control not only the number of productive S. meliloti infections (via ethylene-dependent and -independent pathways) but also the number of infecting bacteria within ITs. As mentioned above, various plant proteins induced during nodulation could have a defense-related function and be involved in this kind of control, notably extensin-like proteins and antimicrobial peptides (defensins) that could inhibit cytokinesis and induce bacteroid differentiation (Mergaert et al., 2006). In this scenario, MtMMPL1 would be one (among several) elements involved in the fine balance between promoting and restricting rhizobial infection. Expressing MtMMPL1 before infection has started, as in the 35STMtMMPL1 transgenic roots, could lead to frequent abortions and a decrease in nodule number (indirectly leading to abortive hyperinfections). Residual nodules, however, would be normal because the 35S promoter is certainly not as strong as the MtMMPL1 promoter in the nodule infection zone.

The Mechanism of MtMMPL1 Activity Raises Intriguing Questions
The key amino acid residue (Glu) known to be involved in the metalloproteinase activity of MMPs is replaced by a Gln residue in MtMMPL1. This Glu-to-Gln substitution prevents protease activity in human MMPs (Crabbe et al., 1994;Rowsell et al., 2002). We found this mutation not only in a cluster of MtMMPL1-like genes, but also within ESTs from two other legume species (soybean and T. pratense). This mutation appeared recently during evolution, because it seems to be legumespecific, but it is not possible to deduce with the current data available whether the MtMMPL1 paralogs were generated before or after M. truncatula speciation.
One hypothesis is that there is a compensatory mutation elsewhere in MtMMPL1-like genes enabling some protease activity to be restored. We could not find a collagenase activity in an MtMMPL1 extract synthesized in vitro (data not shown), but this does not, of course, rule out possible protease activity with natural plant substrates and under the particular conditions existing in ITs. Indeed, Dow et al. (1998) showed that a Xanthomonas campestris metalloprotease was enzymatically active on defense-related ECM HRGPs but not on model substrates such as casein. Another possibility is that a functional MtMMPL1 does not need an active protease site and that its substrate(s) may be bound without degradation, resulting in substrate protection from proteases of bacterial or plant origin. This would be somewhat reminiscent of Srchi24, a chitinase homolog induced during nodule development in S. rostrata and lacking a Glu residue essential for hydrolytic activity, hypothesized to trap Nod factors to protect them or to facilitate interactions with a receptor protein (Goormachtig et al., 2001). To push speculation even further, it can be imagined that Rhizobium, as for X. campestris, produces proteases that can degrade ECM HRGPs and that MtMMPL1 provides a way to control them. To clarify these points, it will be necessary to establish the precise subcellular localization of MtMMPL1, to identify protein(s) interacting with MtMMPL1, and to test them as possible substrates.

CONCLUSION
This study highlights MtMMPL1 as a particularly useful gene for studying various aspects of the infection The Role of MtMMPL1 in Rhizobial Infection process, including IT structure. Furthermore, MtMMPL1 represents a novel variant of the MMP family. Even though its precise role remains to be elucidated, this is the first member of this biologically important protein family with a clear function in plant-microbe symbiotic associations.
Medicago truncatula seeds were surface sterilized and germinated on inverted agar plates in the dark for 3 d at 8°C and 1 d at 20°C. Plants used for qRT-PCR were grown in aeroponic caissons. Plant growth chamber conditions were the following: temperature, 22°C; 75% hygrometry; light intensity, 200 mE m 22 s 21 ; light-dark photoperiod, 16 h/8 h. Control root (T 0 ) and nodule samples from wild-type M. truncatula Gaertn 'Jemalong' genotype A17 were prepared as described by El Yahyaoui et al. (2004). For analysis of inoculated roots (1 and 3 dpi) of M. truncatula A17, supernodulating mutant sunn (TR122 allele; Sagan et al., 1995;Penmetsa et al., 2003) and skl (Penmetsa and Cook, 1997), A17 mutants lin (D8 allele; C. Gough and J.F. Arrighi, unpublished data), hcl (B56 allele; Catoira et al., 2001), and nsp1 (B85 allele; Catoira et al., 2000), chamber and harvesting conditions were the same, but roots were grown directly on a nitrogen-free medium. In the case of Nod factor experiments, M. truncatula A17 plants were treated by 10 28 M Nod factor after 4 d of nitrogen starvation, and whole root systems were harvested before and at 1, 3, 6, 24, and 48 h after treatment and then frozen in liquid nitrogen.

Analysis of Gene Expression by qRT-PCR
RNA was extracted using the SV total RNA extraction kit (Promega) according to the manufacturer's recommendations. RNA quality was checked using a Bioanalyser (Agilent Technologies), and the absence of DNA contamination was verified by a PCR reaction with EF1-a primers. RT-PCR was performed using the SuperScript reverse transcriptase II (Invitrogen) on 3 mg of plant total RNA, to which 80 pg of in vitro transcribed human desmin RNA were added as an external standard to check the RT efficiency. In addition, an EF1-a housekeeping gene was used as an internal standard to check the plant RNA quality. qPCR was conducted on a Roche Lightcycler system (Roche Diagnostics) according to the manufacturer's recommendations. The primer sets used in the different experiments were: 5#-CTTTGCTTGGTGCTGTTTA-GATGG-3# and 5#-ATTCCAAAGGCGGCTGCATA-3# (EF1-a); 5#-CAGCCT-CAGTCCTCCAAATCACA-3# and 5#-TAGGCCTGAGGTCACAGAGGT-3# (desmin); 5#-TTCAAAGGTCTGGGACTTGG-3# and 5#-GCAAAACCAAGG-GACAAAGA-3# (MtMMPL1); and 5#-CAGCCTCCACCTAGCATCCA-3# and 5#-CCACATGCAAAGATGGGACG-3# (MtENOD11). The specificity of primer pairs was confirmed by sequencing the PCR amplicons and analyzing their dissociation curves.

RACE PCR
RACE PCR was carried out using a 5#/3# RACE kit, 2nd generation (Roche Diagnostics). The resulting MtMMPL1 full-length cDNA sequence is accessible in the EMBL database (accession no. Y18249).

Cloning Procedures
For the RNAi construct, we followed the procedure described by Limpens et al. (2004) to generate the MtMMPL1 RNAi construct in pRedRoot II. Inverted repeats were created in pRNAi by two successive cloning steps. An MtMMPL1 1.3-kb fragment was PCR amplified from a cDNA clone, using primers corresponding to flanking pBluescript vector sequences with additional AscI-SpeI and SwaI-BamHI restriction sites, respectively (5#-GCAGAT-GGCGCGCCTCTAGAACTAGTAGATCC-3# and 5#-GGGATTTAAATGGAT-CCATAGGGCGAATTGAGTACC-3#). The PCR product was subsequently cloned BamHI-SpeI and AscI-SwaI into the pRNAi vector, using gel-purified fragments. The resulting inverted repeat construct was inserted KpnI-PacI into the pRedRoot II binary vector, again using gel purified DNA fragments.

Plant Transformation
Roots were transformed using Agrobacterium rhizogenes, following essentially the procedure described by Boisson-Dernier et al. (2005). This method was modified by transferring the transformed plants (checked by red fluorescence at 3 weeks posttransformation) to growth pouches watered with 6 mL of the mineral plant growth medium used for aeroponic cultures (Journet et al., 2001). Plants were inoculated 1 week after transfer, with 1 mL of S. meliloti suspension at an OD 600 5 0.05. Pouches were watered every week alternatively with water and growth medium.

Histochemical Staining and Microscopy Studies
GUS staining (using only 5-bromo-4-chloro-3-indolyl-b-glucuronic acid UA) and double staining for both GUS and b-galactosidase activities after inoculation with a S. meliloti strain carrying a constitutive hemA-lacZ fusion (Ardourel et al., 1994) were performed as described by Boisson-Dernier et al. (2005). For simple b-galactosidase activity, we used 5-bromo-4-chloro-3-indolyl-b-D-galactopyranoside (MP Biomedicals) instead of Magenta-Gal. Roots and nodule sections (50 mm thick) were done in 4% agarose with a vibrating microtome (Leica VT 1000S), and stained samples were observed with a Zeiss Axiophot light microscope (Carl Zeiss). Examination of ITs was repeated in three independent experiments, including one with double blind scoring of RNAi and control (empty vector-transformed) roots.
Histology of the nodule was performed after fixation in 2.5% of glutaraldehyde buffered in 0.1 M sodium phosphate buffer, pH 7.2, dehydrated in an alcohol series and embedded in Technovit 7100 resin (Hereaus Kulzer). Sections (4 mm thick) were observed after counterstaining in a 0.2% aqueous toluidine blue solution. For electron microscopy studies, following the glutaraldehyde fixation step, nodules were postfixed in 1% osmium phosphate buffer solution; we then proceeded to epon embedding steps as previously described (Vasse et al., 1993). Grids were examined using a Hitachi H600 electron microscope, and images were recorded on Kodak film 4489 Estar thick base.
ISHs were carried out as described by de Billy et al. (2001).

Counting Bacteria Recovered from Nodules
Nodules were cut from roots at the indicated times. They were surface sterilized using the following procedure: immersion for 30 s in 70% ethanol, wash with distillate water (five times), immersion for 30 s in 25% (v/v) commercial bleach, and final wash with water (six times). Sterilized nodules were ground with a glass tube in 100 mL of water. Aliquots of serial dilutions (10-to 10 5 -fold) were spread on tryptone yeast plates supplemented with 6 mM calcium chloride.

Phylogenetic Tree Analysis
Protein segments of similar lengths were manually defined, as indicated. Amino acid sequences were then aligned using ClustalW (http://clustalw. genome.ad.jp/) and analyzed with the PHYLIP software package (http:// www.csc.fi/molbio/progs/phylip/doc/main.html). Bootstrap values were obtained from 1,000 replicates. The branch lengths indicate the frequency of the corresponding clade in the set of bootstrap trees.
Sequence data from this article can be found in the GenBank/EMBL data libraries under accession number Y18249.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Multiple alignment analysis of putative proteins encoded by MtMMPL1 and the cluster of MtMMPL1-like genes located on the bacterial artificial chromosome clone CR962135.
Supplemental Figure S2. Electronic northern of MtMMPL1 EST cluster (MtC40019) and of two closely related EST clusters.