The cin Quorum Sensing Locus of Rhizobium etli CNPAF512 Affects Growth and Symbiotic Nitrogen Fixation*

Rhizobium etli CNPAF512 produces an autoinducer that inhibits growth of Rhizobium leguminosarum bv. viciae 248 and activates the Agrobacterium tumefaciens tra reporter system. Production of this compound in R. etli is dependent on two genes, named cinR and cinI , postulated to code for a transcriptional regulator and an autoinducer synthase, respectively. NMR analysis of the purified molecule indicates that the R. etli autoinducer produced by CinI is a saturated long chain 3-hydroxy-acyl-homoserine lactone, abbreviated as 3OH-(slc)-HSL. Using cin - gusA fusions, expression of cinI and cinR was shown to be growth phase-dependent. Deletion analysis of the cinI promoter region indicates that a regulatory element negatively controls cinI expression. Mutational analysis revealed that expression of the cinI gene is positively regulated by the CinR/3OH-(slc)-HSL

Although bacteria are unicellular organisms, they often show group behavior. For this, bacteria have to monitor their own population size. This can be achieved by means of autoinduction. Cell-cell communication using N-acyl-homoserine lactone (AHL) 1 signals is one of the few known mechanisms through which bacteria can communicate with each other and is a widespread phenomenon in Gram-negative bacteria (1,2), including plant-associated bacteria (3,4). AHLs mainly vary with respect to the length (4 -14 carbons) and the substituent (H, O, or OH) at the third carbon of the acyl side chain. The AHL signal is released into the environment, either by passive diffusion, as observed for 3O-C 6 -HSL in Vibrio fischeri and Escherichia coli cells (5) or by a combination of diffusion and active efflux in Pseudomonas aeruginosa (6) and accumulates with growth of the bacterial population. At least in V. fischeri, the signal freely diffuses back into the cells such that its intracellular concentration also rises as a function of the increase in bacterial population. Transduction of this information to response regulators of gene expression leads to the elaboration of an appropriate phenotype at high cell densities.
Using the Agrobacterium tumefaciens tra reporter system to detect autoinducer molecules, members of the genus Rhizobium showed the greatest diversity, with some producing as few as one and others producing as many as seven detectable signals (7). In Rhizobium leguminosarum bv. viciae, the cin locus encodes a master regulatory system. Mutation of cinIR abolishes the production of N-(3R)-hydroxy-7-cis-tetradecenoyl-L-homoserine lactone (3OH-C 14:1 -HSL), also termed "small", and reduces the synthesis of AHLs produced by the enzymes encoded by raiI, traI-like, or rhiI (8). The reduced levels of C 6 -HSL and C 8 -HSL and decreased rhiR expression cause a repression of the rhizosphere-expressed genes in cinI or cinR mutants (8 -10). Furthermore, 3OH-C 14:1 -HSL induces the stationary phase (9) whereas mutation of cinI has little effect on growth or nodulation of the host plant (8,11).
Rhizobium etli CNPAF512, a nitrogen-fixing symbiont of Phaseolus vulgaris, produces at least seven different autoinducer molecules (12). Rosemeyer et al. (12) identified the raiIR quorum-sensing system in R. etli. Examination of different rai mutants for nodulation of beans showed that raiI is involved in the restriction of nodule number, whereas nitrogen-fixing activity per nodule is not affected. The culture supernatant of a raiI mutant revealed only three different autoinducer molecules. One of them induces a growth-inhibitory effect on R. leguminosarum bv. viciae 248, similar to the low molecular * Parts of this study were presented at the American Society for Microbiology Conference on Cell-Cell Communication in Bacteria, July 6 -9, 2001, Snowbird, UT. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  weight bacteriocin "small", which is common in fast-growing rhizobia. The properties, growth inhibition, and autoinducer activity, are features reported for 3OH-C 14:1 -HSL, produced by R. leguminosarum bv. viciae (13).
Here we report on the cin locus, the second quorum-sensing system in R. etli CNPAF512 that is expressed under both free-living and symbiotic conditions and is involved in the production of a 3OH-(slc)-HSL (slc, saturated long chain). Despite high sequence conservation, the cin locus of R. etli and R. leguminosarum bv. viciae appear to control different functions. Mutational analysis of R. etli revealed that the cin system regulates growth and fulfills a key role in bacteroid differentiation and nitrogen fixation.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Plasmids, and Culture Conditions-E. coli was grown in Luria-Bertani (LB) medium at 37°C (14). Rhizobium was grown at 30°C in TY or AMS medium (15). A. tumefaciens NT1 was grown in AB medium at 28°C (16). Antibiotics were added as appropriate. To study bacterial growth over a long period of time, overnight cultures of the strains were diluted in 10 mM MgSO 4 to an absorbance at 595 nm (A 595 ) of 0.3 (dilution ϳ10-fold). Subsequently, these cultures were diluted 10fold after which 295 l of growth medium was inoculated with 5 l of bacterial suspension (total dilution ϳ6000-fold). Bacteria were grown, and the absorbance was measured automatically each 30 min during at least 6 days in a BioscreenC (Labsystems Oy). For each time point, the average optical density was calculated from five independent measurements.
Plant Experiments, Bacteroid Isolation, and Preparation of Bacteroid-free Supernatant-Phaseolus vulgaris cv. Limburgse vroege seedlings were planted in Snoeck medium, which is optimized for in vitro growth of common bean. 2 Plants were inoculated and grown essentially as described by Michiels et al. (17). Acetylene reduction activity was determined 3 weeks after inoculation. For expression analysis during symbiosis, bacteroids were purified from plant material by differential centrifugation (17). This protocol was slightly adapted for the extraction of bacteroid autoinducers. Nodules from 1-2 plants (Ϯ 1.5 g) were collected in a Falcon tube (15 ml) containing 0.2 g of polyvinyl polypyrolidone, and magnesium phosphate buffer was added to a final volume of 6 ml and subsequently crushed. Next, the plant material was removed by differential centrifugation. To lyse the bacteria, the suspension was supplemented with SDS (final concentration, 1%) and proteinase K (final concentration, 100 g/ml) and incubated for 1 h at 37°C. After incubation, the cell debris was removed by centrifugation at 6000 rpm after which the cell-free supernatant was immediately extracted to isolate the autoinducers.
Qualitative and quantitative analysis of ␤-glucuronidase (GusA) activity was performed as described elsewhere (17,18). For transmission electron microscopic analysis (TEM), thin sections of 3-week-old nodules were prepared as described by Xi et al. (19), and analyzed in a Zeiss EM 900 electron microscope.
Extraction and Detection of Autoinducers-Rhizobium strains and A. tumefaciens transformants were tested for activation of the A. tumefaciens tra reporter system and for bacteriostatic activity toward the sensitive strain R. leguminosarum bv. viciae 248 as described by Schripsema et al. (13). To extract autoinducers, strains were first grown to the stationary phase. Cell-free supernatant from either a free-living liquid culture or from symbiotic bacteroids (see above) was extracted, and the autoinducers were detected on TLC by the tra reporter system as described by Rosemeyer et al. (12).
Isolation of the R. etli 3OH-(slc)-HSL-Ethyl acetate extracts from 10 liters of stationary phase cultures of FAJ4010 were reconstituted in 50% acetonitrile in water and subjected to solid-phase extraction using Waters OASIS HLB cartridges. Fractions were eluted with an increasing concentration of methanol in water (50 -100%, v/v). Positive fractions on the A. tumefaciens tra reporter and R. leguminosarum bv. viciae 248 growth inhibition assays were collected, dried, and redissolved in 50% acetonitrile in water and applied to a C18 Phenomenex Bondclone HPLC column. Fractions were eluted with a linear gradient of acetonitrile in water (40 -100%, v/v) over a 30-min period at a flow rate of 1 ml/min and monitored at 200 nm. Positive fractions were re-chromatographed using an isocratic mobile phase (50% acetonitrile in water; 1 ml/min), and the active subfraction was analyzed by NMR spectroscopy.
DNA Techniques and Nucleotide Sequencing-Standard techniques were used for DNA manipulations (14). Restriction enzymes were used according to the manufacturer's instructions. DNA probes for Southern hybridization were labeled with digoxigenin. An ordered series of sequencing clones was obtained via restriction enzyme mapping of pFAJ4003 and ExoIII deletion procedures (Erase-A-Base ® Promega). Nucleotide sequencing of cinR, cinI, and the flanking regions was accomplished by using the A.L.F. sequencer (Amersham Biosciences).
Cloning of the cinR-cinI Gene Region and Construction of Mutants-The 5.3-kb EcoRI fragment from pFAJ4000 containing cinR, cinI, orf123, and orf140 was first cloned into pBluescriptIIKs ϩ yielding pFAJ4003. A PstI-EcoRI fragment of pFAJ4003, lacking the 5Ј-end of orf140, was EcoRI-PstI subcloned in the broad host range vector pLAFR3 (pFAJ4012; Fig. 1). Furthermore, pFAJ4013 was made by insertion of the 5.3-kb EcoRI fragment from pFAJ4003 in pPZP200, in which part of the multiple cloning site between XhoI and PacI was deleted.
A fragment containing cinR and cinI was amplified via PCR using primers Rhi15 (5Ј-ATGGGAATTCATCCAGTGCCGAGGAGATAC-3Ј) and Rhi16 (5Ј-TAGAGGATCCTCGGCATCATCATCACCTCG-3Ј) and pFAJ4003 as template DNA. The resulting 3-kb fragment was digested with EcoRI and BamHI and cloned in pUCNotI-⌬S (pUCNotI derivative lacking the SphI recognition site via restriction digest and blunt ligation) yielding plasmid pFAJ4004 (Fig. 1). The 2.2-kb BamHI fragment from pHP45⍀-Km containing a Km r cassette was ligated into the unique SphI site of cinI in pFAJ4004 after blunting of the fragments. The cinRcinI::Km locus was further cloned into the sacB suicide vector pJQ200uc1 as a 5.2-kb NotI fragment to obtain pFAJ4006. The nonpolar cinR point mutation and an additional frameshift were introduced via the QuickChange™ site-directed mutagenesis (Stratagene) using primers Rhi32 (5Ј-CTATGCCATGCTGGTACCATGCCCAGAAGCACG-3Ј) and Rhi33 (5Ј-CGTGCTTCTGGGCATGGTACCAGCATGGCATAG-3Ј) on pFAJ4004. As a result of the mutation, a new and unique KpnI site (GGTACC; mutations are shown in bold: insertion of A; substitution G with C) was created in cinR. Subsequently, the 3-kb NotI fragment containing the mutated cinR gene was inserted into pJQ200uc1 creating pFAJ4007. The cinR::Sp mutation was made by introducing the 2-kb spectinomycin resistance cartridge from pHP45⍀ into the unique KpnI site within cinR of pFAJ4007 after blunting of the fragments, resulting in pFAJ4009. The plasmids pFAJ4006, pFAJ4007, and pFAJ4009 were introduced into R. etli CNPAF512, and double recombinants were selected as described previously (12) creating a cinI mutant (FAJ4006) and two cinR mutants (FAJ4007, FAJ4009; Fig. 1). A cartridge containing the promoterless gusA gene and a spectinomycin resistance gene from pWM5 was removed with SmaI and ligated into the XhoI site of pFAJ1327 (containing raiI, raiR, and orf1) after blunting of the vector, yielding plasmid pFAJ4010. For the construction of a R. etli raiI mutant strain, pFAJ4010 was introduced into the R. etli CNPAF512 wild-type strain, yielding FAJ4010. Tri-parental conjugation of pFAJ4006 (cinI::Km) into FAJ4010 resulted in the raiI::Sp cinI::Km double mutant FAJ4013.
Construction of Fusions-Plasmid-borne P cinR -gusA and P cinI -gusA fusions were constructed in the promoter probe vector pFAJ1703 (20). The cinR promoter region was amplified by PCR with FAJ1337 as a template using primers Rhi15 and Tn5-B (5Ј-GGTTCCGTTCAG-GACGCTAC-3Ј). The resulting 1.2-kb fragment was cloned into the HpaI site of pFAJ1703 after blunting of the fragments, yielding plasmid pFAJ4011. To construct pFAJ4014 (cinI-gusA fusion carrying a mutant cinR gene region), the 1.8-kb EcoRI-SphI fragment of pFAJ4007 was blunt-end ligated into the HpaI site of pFAJ1703. pFAJ4015 is a KpnI deletion derivative of the cinI-gusA fusion pFAJ4014 (Fig. 1).

Identification of the cin Locus of R. etli CNPAF512-A
Tn5induced mutant library of R. etli CNPAF512 was screened using the growth inhibition assay (13). The mutant FAJ1337 no longer inhibited growth of R. leguminosarum bv. viciae 248. By means of inverse PCR on EcoRI-digested genomic DNA of FAJ1337 with transposon-specific primers, ϳ10 kb of the transposon flanking DNA of FAJ1337 was amplified and cloned. Subsequently, PstI fragments of the amplified DNA were subcloned into pUC18 and tested for autoinducer production in E. coli using the tra system of A. tumefaciens as a reporter (7). One positive clone was obtained, and the corresponding 7.5-kb PstI fragment, carrying part of the cloning vector, was subsequently used as a probe in a Southern hybrid-ization of a R. etli CNPAF512 genomic library, constructed in pLAFR1 (21). A 5.3-kb EcoRI fragment from clone pFAJ4000 was found to give a positive hybridization signal. FAJ1337 was complemented for growth inhibition of the sensitive strain R. leguminosarum bv. viciae 248 by both pFAJ4000 and pFAJ4012 ( Fig. 1).
DNA Sequence Analysis of the R. etli CNPAF512 cin Locus-DNA sequence analysis of the cloned 5.3-kb EcoRI fragment revealed five complete open reading frames (ORFs) as illustrated in Fig. 1. The ORFs were identified and the start codon was assigned on the basis of the GC content (22), the preferential codon usage (searchcutg, GCG-package Wisconsin), and similarity with known genes.
One ORF (726 bp) codes for a putative protein of 241 amino acids with a calculated molecular mass of 27.3 kDa. The putative protein is similar to several LuxR-type transcriptional activators such as CinR (96% amino acid identity) of R. leguminosarum bv. viciae (AAF89989), CerR (30% identity) of Rhodobacter sphaeroides (AAC46021) and RaiR (31% identity) of R. etli (AAC38173). Because of the high amino acid identity of the R. etli putative protein with CinR, it was given the same name. Alignment of R. etli CinR with E. coli NarL (23), designates a helix-turn-helix motif between residues 196 and 220. In silico analysis of the cinR non-coding region revealed a putative terminator (nucleotides 2615-2650, ⌬G ϭ Ϫ26.3). PCR analysis with a Tn5-specific primer combined with primers within the coding sequence of cinR or cinI (see below) indicated that the transposon in FAJ1337 is inserted between nucleotides 423 and 424 of cinR.
Immediately downstream of cinI, a short ORF123 (368 bp), located on the opposite strand, encodes a putative response regulator of the CheY family with 94% identity to the CheY like protein of R. leguminosarum bv. viciae (AAF89991), 40% identity to a probable response regulator of Mesorhizobium loti (BAB49462) and 35% identity to the FixL receiver domain of R. etli (AAG00949). The R. etli response regulator encoded by ORF123 contains the conserved residues D14, D58, T86 en K106 (numbered with respect to ORF123), which are part of the essential active site of CheY (26) in which D58 can be phosphorylated.
ORF140 (420 bp) is located upstream of cinR and codes for a protein similar to a hypothetical protein (AAG2039) of Halobacterium sp. NRC-1 (51% identity), and an unknown protein of Bacillus subtilis (CAB11811) (39% identity). Upstream of ORF123, a Met-tRNA gene (74 bp) (tRNAscan-S.E. v. 1.11) is found with the anticodon (CAT) located between nucleotides 4364 and 4366 of the 5.3-kb EcoRI fragment. This gene shows perfect (100%) DNA sequence identity to the Met-tRNA gene of R. leguminosarum bv. viciae (AF210630) and a M. loti sequence (AP002999) and is similar to a Rhizobium sp. NGR234 sequence (AE000079) (91% identity). Analysis of the intergenic region between ORF123 and the Met-tRNA gene indicates the presence of a putative terminator sequence downstream of the tRNA gene (nucleotides 5002-5068, ⌬G ϭ Ϫ22.9).
NMR Analysis of the Isolated Compound Produced by R. etli CinI-An NMR analysis of the compound produced by R. etli CinI was conducted. As a control, 3OH-C 14:1 -HSL was synthesized (data not shown) and analyzed. The NMR data of the synthetic compound are in agreement with previously recorded data (13).
The 1 H NMR spectrum of the R. etli compound contains all characteristic signals of a 3-hydroxyacyl-homoserine lactone. Evidence for the homoserine lactone moiety is constituted by the signal at 6.33-6.24 ppm (amide NH) and the characteristic butyrolactone signals at 4.52, 4.47, 4.27, 2.77, and 2.10 ppm. The line shapes and splitting patterns are in good agreement with those of synthetic acyl-homoserine lactones. Moreover, the line at 3.98 ppm is similar to the CH(OH) resonance in 3OH-C 14:1 -HSL. However, the characteristic double bond signals between 5 and 6 ppm observed for 3OH-C 14:1 -HSL, as well as the signals around 2.00 ppm of the protons on adjacent carbon atoms are absent. On the basis of its chromatographic properties (TLC, HPLC), the R. etli HSL is likely to possess a long chain fatty acid group. This allows to tentatively assign the spectrum of the R. etli autoinducer produced by CinI to a saturated long chain 3-hydroxy-acyl-homoserine lactone, which is clearly different from the structure of R. leguminosarum 3OH-C 14:1 -HSL. In the subsequent part we will refer to the R. etli autoinducer as 3OH-(slc)-HSL.
Gene Regulation of cinI and cinR-To study the cell densitydependent expression of the cin locus, cinR-gusA (pFAJ4011) and cinI-gusA (pFAJ4014) fusions were constructed. The cinR gene in pFAJ4014 was inactivated by site-directed mutagenesis. To determine whether a promoter is present in the cinR-cinI intergenic region, a second cinI-gusA (pFAJ4015) fusion, containing a 632-bp upstream region of cinI (Fig. 1), was also constructed. As shown in Fig. 2A, cinI expression from pFAJ4014 under free-living conditions in a wild-type background increased with the cell density and reached a plateau (1500 -2400 units) as the culture entered into the stationary phase. cinI expression from pFAJ4015 displayed a similar cell density-dependent pattern of expression (Fig. 2B). However, two differences between the two fusions can be noticed. Firstly, induction of cinI expression from pFAJ4015 starts at a lower absorbance compared with pFAJ4014. Secondly, the maximum expression level of cinI from pFAJ4015 is approximately 4-fold higher, compared with that of pFAJ4014. None of the cinI-gusA fusions are expressed in cinR or cinI mutant backgrounds (Fig.  2, A and B), demonstrating that transcription of cinI requires both CinI and CinR.
A threshold cell density (approximately A 595 ϭ 0.6) seems to be required to observe a very low cinR expression in wild-type and cinR or cinI mutant backgrounds (Fig. 2C). Although expression of cinR remains low, it reaches a maximum (ϳ20 units) as soon as cells enter the stationary phase. The observation that both cinR and cinI expression is maximal during the same stage of growth is in agreement with a role of CinR in the regulation of cinI expression.
The observation that cinI is expressed in either the presence or absence of the cinR promoter region, demonstrates that cinR and cinI are likely organized into different transcriptional units. Furthermore, the overall high expression level of cinI (minimal 100-fold higher than cinR), suggests that transcription of cinI in both pFAJ4014 and pFAJ4015 is controlled by a promoter in the cinR-cinI intergenic region. This is in agreement with the presence of a putative terminator downstream of cinR. The difference in cinI expression levels between pFAJ4014 and pFAJ4015, is likely caused by a negative regulation at the level of the putative cinI promoter.
Expression under Symbiotic Conditions-Expression of the cinI-gusA and cinR-gusA fusions was monitored in isolated bacteroids, obtained from 21-day-old bean nodules, induced by wild-type CNPAF512, the cinR (FAJ4007), and cinI mutant (FAJ4006). The presence of the plasmid-borne gusA-fusions pFAJ4011, pFAJ4014, and pFAJ4015 in the different strains did not affect their symbiotic performance (data not shown). As illustrated in Fig. 3, expression of the cinR fusion (pFAJ4011) is low (ϳ40 units) and does not significantly differ in the three genetic backgrounds. Expression of the short cinI fusion (pFAJ4015) is significantly higher (ϳ8-fold) compared with the long cinI fusion (pFAJ4014) in wild-type background, similar to what was observed under free-living conditions. However, in contrast to free-living conditions, expression of the short cinI fusion is less dependent on the presence of CinI because cinI expression from pFAJ4015 is still observed in cinI mutant (FAJ4006) bacteroids. Similarly to free-living conditions, cinI gene expression requires the presence of CinR. Finally, it should be noted that the expression level of cinI in bacteroids is at least 10-fold lower compared with free-living conditions. Shortly after inoculation (24 h), cinI-gusA expression (pFAJ4015) was observed on the surface of root hairs during colonization of wild-type and the cinI mutant (FAJ4006) (data not shown). Light microscopic analysis, 48 h after inoculation of the roots, localized pFAJ4015 expression in the infection tread, formed by wild-type and the FAJ4006 mutant (Fig. 4, A and B). However, no cinI-gusA expression from pFAJ4015 was observed in the cinR mutant (FAJ4007) during colonization (data not shown) or in the infection tread (Fig. 4C). Based on the intensity of the Gus-staining, the fusion seems less expressed in a cinI mutant background (FAJ4006) compared with the wild-type background, similarly as observed for bacteroid expression.
Analysis of Autoinducers-The cell-free culture supernatant of the R. etli wild type, the cinR and the cinI mutant, was extracted with ethyl acetate, and analyzed by TLC combined with the A. tumefaciens tra reporter system (3). As was previously published by Rosemeyer et al. (12), wild-type R. etli CNPAF512 produces at least seven autoinducers (Fig. 5, lane  A). The raiI mutant, FAJ4010 (Fig. 5, lane C) produces 3OH- FIG. 2. Expression of cinR-and cinI-gusA fusions. cinI expression from pFAJ4014 (A) or pFAJ4015 (B) and cinR expression from pFAJ4011 (C) was examined quantitatively in a cell density-dependent manner. Curves representing the expression in the wild-type background are indicated by a diamond, in a cinI mutant background (FAJ4006) by a gray triangle, and in a cinR mutant background (FAJ4009) by a square. The absorbance at 595 nm of wild-type R. etli CNPAF512 containing pFAJ4011, pFAJ4014, or pFAJ4015 (cross) is shown. The growth of the other strains is similar to that of the wild type. The graphs are representative of experiments carried out independently several times. (slc)-HSL (F-AI1) and two other active molecules (F-AI2, F-AI3), which co-migrate with compounds extracted from sterile, non-inoculated TY medium (Fig. 5, lane E). Ethyl acetate extracts from AMS medium or water were negative in this test (data not shown). Both cinR and cinI mutants (FAJ4006, FAJ4007, FAJ4009) secrete all of the autoinducers except 3OH-(slc)-HSL (Fig. 5, data shown for FAJ4006 in lane B) as was also observed in the growth inhibition assay. As illustrated in Fig. 5 lane D, the extract prepared from a raiIcinI double mutant FAJ4013, contains two active spots (F-AI2, F-AI3), co-migrating with active molecules present in sterile, non-inoculated TY. Taken together, these data suggest that two genetic systems, rai and cin, are required for the synthesis of all of the autoinducers by R. etli CNPAF512 as detected by the Agrobacterium tra reporter system.
To confirm that the cin locus is solely responsible for the production of the 3OH-(slc)-HSL molecule, a plasmid containing the 5.3-kb EcoRI fragment, pFAJ4013, was introduced in the A. tumefaciens NT1 strain (lacking the Ti plasmid, containing plasmid pJM749), a derivative lacking synthesis of autoinducers (Fig. 5, lane G). TLC analysis revealed the production of a compound co-migrating with the R. etli 3OH-(slc)-HSL (Fig.  5, lane F). Moreover, using this extract, growth inhibition of R. leguminosarum bv. viciae 248 was observed in a growth inhibition assay (data not shown).
Finally, the autoinducer production was analyzed during symbiosis. For this, the lysate, prepared from R. etli bacteroids, was extracted with ethyl acetate and the extracted autoinducers from one nodulated plant were analyzed on TLC for compounds activating the A. tumefaciens tra system. Less different autoinducers seem to be produced under symbiotic conditions compared with free-living growth (Fig. 5, lane H). One of the wild-type bacteroid autoinducer compounds, S-AI1, co-migrates on TLC with 3OH-(slc)-HSL (F-AI1). The bacteroid autoinducer S-AI3 co-migrates with a compound, extracted from root material and able to activate the tra reporter system (Fig.  5, lane I). Taken together, these results indicate that at least three different autoinducer molecules are produced by R. etli during symbiosis (S-AI1, S-AI2, S-AI4).
Mutation of the rai and cin Autoinducer System Affects Growth of Rhizobium-To examine the phenotypic relevance of the rai and cin system in R. etli, a rai-cin double mutant was constructed. To monitor growth, absorbance was measured in a BioscreenC over a 6-day period (total dilution 6000-fold). Growth of the raiR mutant (FAJ1329) and the cinI mutant (FAJ4006) was clearly delayed compared with wild-type growth in AMS mannitol medium as can be seen from Fig. 6. The lag phase was prolonged by 14.5 and 24.5 h, respectively. Noteworthy, the growth pattern of the cinR mutants, FAJ4007 and FAJ4009, as well as the growth of a raiI mutant (FAJ4010) was not different from that of the wild type (Fig. 6). Entry into the exponential phase was even more delayed (79 h more than the wild type) when growth of the raiIcinI double mutant, FAJ4013, was examined (Fig. 6). Furthermore, growth of the cinI (g ϭ 6.8 h) and raiR (g ϭ 8.3 h) mutants as well as that of the double raiIcinI mutant (g ϭ 8.8 h) was marked by a 1.5-to 2-fold increase in generation time compared with the wild type (g ϭ 4.5 h). The observed difference in lag phases was less pronounced when the culture was only 1000-fold diluted whereas no difference was obtained with a 100-fold-diluted preculture (data not shown). Although a similar increase in generation time was also observed when bacteria were grown in AMS succinate medium (data not shown), the difference in the lag phases was less pronounced, indicating that the raicindependent regulation of growth is complex and depends on the growth medium (carbon source) used.
Symbiotic Phenotype of cinR and cinI Mutants-The R. etli cinR and cinI mutants were tested for their ability to nodulate common bean (Phaseolus vulgaris cv. Limburgse vroege) and to fix nitrogen. No significant differences in kinetics of appearance of the first nodules were observed (Table I). Moreover, the plant and nodule dry weight as well as the nodule number, determined 21 days after inoculation, were not significantly different between plants nodulated by the wild-type or the mutant strains (Table I). However, a clear impact of the cin system on nitrogen fixation could be observed because inoculation of bean plants with the R. etli cinR mutants (FAJ4007 and FAJ4009) or the cinI mutant (FAJ4006) resulted in a statistically lower acetylene reduction activity (ARA) per plant (30 -40% of wild-type ARA; Table I). Furthermore, inoculation with the FAJ4013 raiIcinI double mutant decreased nitrogen fixation per plant even further (27% of wild-type ARA; Table I). Autoinducer production, growth, and the tested symbiotic features are similar for FAJ4009 and FAJ4007, the polar and non-polar cinR mutant. This observation supports further the hypothesis that cinR and cinI are likely organized into different transcriptional units.
To further examine the effect of cin mutations at the bacteroid level, sections of nodules, formed by wild-type and mutant strains were analyzed by transmission electron microscopy (TEM). This analysis indicated that cinI mutant bacteroids were always individually packed in the symbiosome membrane (SM) (Fig. 7A), whereas wild-type symbiosomes usually contained multiple bacteroids (Fig. 7B). Furthermore, cinI mutant bacteroids were surrounded by a minimal symbiosome space (SS) compared with wild-type bacteroids. These results indicate that the cin system fulfills a key role during symbiosome development. DISCUSSION We have characterized the cin locus in R. etli CNPAF512, involved in the synthesis of 3OH-(slc)-HSL, containing a saturated long acyl chain. In our current model, cinI codes for the autoinducer synthase and cinR for the transcriptional regulator that binds the 3OH-(slc)-HSL. The latter complex activates cinI expression. The chromatographic properties of 3OH-(slc)-HSL are very similar to 3OH-C 14:1 -HSL, produced by R. leguminosarum bv. viciae. Moreover, both compounds induce growth inhibition of R. leguminosarum bv. viciae 248. In contrast, major differences between both cin loci with respect to growth under free-living conditions and symbiotic performance of the corresponding mutants were observed, indicating that both molecules may perform different functions in R. etli and R. leguminosarum bv. viciae.
Mutational and expression analysis revealed that cinR and cinI have distinct promoters. Expression of cinR is low both under free-living and symbiotic conditions and is cell densitydependent. Expression of cinI is also regulated in a cell densitydependent way and reaches a maximal expression level in the stationary phase. Furthermore, the cinI gene is expressed in bacteroids. Expression of cinI requires CinR both under freeliving conditions and during symbiosis. Expression levels of cinI differ depending on the extent of the upstream region in the plasmid-borne cinI-gusA fusion construct. Possibly, the DNA sequence upstream of the putative cinI promoter in the plasmid construct competes for a trans-acting factor that is required for cinI transcription, such as CinR. Alternatively, several reports from the literature suggest complex regulation of genes involved in quorum-sensing, in particular the occurrence of negative regulators. Examples of such negative regulators are: EsaR in Pantoea stewartii (27), TraS in A. tumefaciens (28) and RsaL in P. aeruginosa (29). Moreover, Lithgow et al. (8) showed that expression of cinI is significantly reduced when the symbiotic plasmid pRL1JI is present in R. leguminosarum bv. viciae, resulting in a reduction in the level of 3OH-C 14:1 -HSL. The mechanism of pRL1JI-mediated repression of cinI expression or the identity of the factors enhancing or relieving this repression have not yet been identified.
cinI is likely positively autoregulated in the bacteroids even though a clear expression level could be observed in a cinI mutant background. CinR may stimulate expression of cinI even in the absence of 3OH-(slc)-HSL. Whether CinR is activated through an autoinducer molecule produced by RaiI (see further) or whether a plant compound is able to activate CinR is yet unknown. Teplitski et al. (30) noticed that various species of higher plants can secrete substances, chemically different from bacterial AHLs but mimicking their activity. Also, extraction of non-inoculated Phaseolus vulgaris bean roots revealed the production of a compound activating the Agrobacterium tra reporter system, as illustrated in this work, making a crosstalk between both partners in the symbiosis quite reasonable. However, in contrast to the previously identified plant compounds (30), the P. vulgaris active molecules were found in the ethyl acetate fraction. Our suggestion of a possible cross-talk between a prokaryote and the eukaryotic host, was demonstrated in the case of P. aeruginosa, the opportunistic pathogen of immunocompromised individuals. The P. aeruginosa quorum-sensing signal molecule 3O-C 12 -HSL stimulates interleukin-8 production in pulmonary epithelial cells (31) and may modulate the host immune response by suppressing cytokine production in macrophages (32). Up to now, several autoin-  ducer systems have been described in Rhizobium sp. but the cin system of R. etli is the first proven to be expressed in the infection tread and in differentiated bacteroids. In contrast, the rhi system, which is shown to be important for interaction with legumes and which is specific to R. leguminosarum bv. viciae, is only expressed in the rhizosphere (10). We have illustrated here for the first time that nitrogen-fixing bacteroids produce autoinducers in the nodules under conditions that are quite different from free-living growth. In vivo production and excretion of AHLs was also observed in lung tissues of mice infected with P. aeruginosa (33). Furthermore, autoinducers other than AHLs, such as quinolones (34) and diketopiperazines (DKP) (35), have been described in Gram-negative bacteria. The observation that DKP can be generated via non-enzymatic cyclization of linear dipeptides at extremes of pH and temperature (36) can offer an explanation for the presence of active compounds in sterile, non-inoculated TY medium.
Analysis of quorum-sensing in R. etli is further complicated by the fact that besides the cin system, involved in the synthesis of 3OH-(slc)-HSL, a second system, rai, regulates production of several other molecules with autoinducer activity (12). This is particularly striking when the effect of mutations in the cin or/and rai genes on growth of R. etli is analyzed: RaiR fulfills a key regulatory role, and it is shown that CinR negatively affects growth when 3OH-(slc)-HSL is not produced. Growth was even more delayed in a raiIcinI mutant. Both systems seem to interact at the level of unknown genes directly or indirectly controlling growth of the bacteria. The R. etli autoinduction systems clearly affect the symbiotic properties of the bacterium. The rai system is involved in restriction of the nodule number, whereas nitrogen fixation activity is not affected (12). On the other hand, the decreased acetylene reduction activity of plants nodulated by cin mutant strains and the bacteroid morphology, illustrate the important role the cin quorum-sensing system plays in R. etli during symbiosis. A disrupted communication, as observed in the cin mutants, results in an arrest of bacteroid differentiation. Also, in contrast to wild-type bacteroids, cinI mutants are individually enclosed in a symbiosome and are devoid of a large symbiosome space. The cin quorum-sensing system is a prerequisite to complete the differentiation cycle of the bacterium. In contrast, it was shown that 3OH-C 14:1 -HSL, produced by CinI in R. leguminosarum bv. viciae, is not required for the formation of effective nodules (11). This observation was also confirmed by Lithgow et al. (8) in a plant experiment where mutation of cinI had little effect on growth or nodulation. In summary, R. etli 3OH-(slc)-HSL and R. leguminosarum bv. viciae 3OH-C 14:1 -HSL share some common characteristics but elicit different phenotypic changes in their respective producing cells.