nolO and noeI (HsnIII) ofRhizobium sp. NGR234 Are Involved in 3-O-Carbamoylation and 2-O-Methylation of Nod Factors*

Loci unique to specific rhizobia direct the adjunction of special groups to the core lipo-oligosaccharide Nod factors. Host-specificity ofnodulation (Hsn) genes are thus essential for interaction with certain legumes. Rhizobium sp. NGR234, which can nodulate >110 genera of legumes, possesses three hsn loci and secretes a large family of Nod factors carrying specific substituents. Among them are 3-O (or 4-O)- and 6-O-carbamoyl groups, an N-methyl group, and a 2-O-methylfucose residue which may bear either 3-O-sulfate or 4-O (and 3-O)-acetyl substituents. The hsnIII locus comprises a nodbox promoter followed by the genes nodABCIJnolOnoeI. Complementation and mutation analyses show that the disruption of any one of nodIJ, nolO, or noeI has no effect on nodulation. Conjugation of nolO intoRhizobium fredii extends the host range of the recipient to the non-hosts Calopogonium caeruleum and Lablab purpureus, however. Chemical analyses of the Nod factors produced by the NodI, NolO, and NoeI mutants show that the nolO andnoeI gene products are required for 3 (or 4)-O-carbamoylation of the nonreducing terminus and for 2-O-methylation of the fucosyl group, respectively. Confirmation that NolO is a carbamoyltransferase was obtained from analysis of the Nod factors produced by R. frediicontaining nolO; all are carbamoylated at O-3 (or O-4) on the nonreducing terminus. Since mutation of both nolO andnodU fails to completely abolish production of monocarbamoylated NodNGR factors, it is clear that a third carbamoyltransferase must exist. Nevertheless, the specificities of the two known enzymes are clearly different. NodU is only able to transfer carbamate to O-6 while NolO is specific for O-3 (or O-4) of NodNGR factors.

known examples. Genes of the second class are necessary for the interaction with certain, but not all, legumes. Expression of these genes permits nodulation of additional hosts, and for this reason, they have been termed host s pecificity of nodulation (ϭ hsn) genes. By definition, they are unique to one or a few rhizobia.
In 1990, Lerouge et al. (1) showed that the products of the nod genes are N-acylated oligomers of N-acetyl-D-glucosamine. Numerous other investigators have confirmed these findings (for reviews see Refs. [2][3][4][5]. Since these substances are the products of the nod genes, they are called Nod factors. Principal differences among the Nod factors of the various rhizobia concern the length of the core molecule as well as the substitutions to both the reducing and nonreducing residues. Presumably, the hsn genes are responsible for these substitutions. With the discovery of Nod factors, it became possible to correlate nod gene expression with Nod factor structure. In their pioneering studies, Roche et al. (6) demonstrated that the hsn loci, nodH and nodPQ, are responsible for the 6-O-sulfation of the reducing N-acetyl-D-glucosamine of NodRm factors. Later work has shown that the first step in Nod factor assembly is performed by an N-acetylglucosaminyltransferase coded by nodC (7). Then, a deacetylase coded by nodB removes the N-acetyl moiety from the nonreducing end of the N-acetylglucosamine oligosaccharides (8). Finally, an acyltransferase coded by nodA, links the acyl chain to the NH 2 -free carbon C-2 of the nonreducing end of the oligosaccharide (9). NodI and NodJ are involved in the export of Nod factors (10,11).
In Rhizobium sp. NGR234, three Hsn loci were discovered by transferring cosmids covering the symbiotic plasmid to heterologous rhizobia (i.e. rhizobia unable to nodulate NGR234 hosts), and asking if the transconjugants could form nodules on Vigna unguiculata (12,13). In this way, we showed that HsnII, which is responsible for the host-specific nodulation of Leucaena species, contains nodSU (14), which are involved in Nmethylation and 6-O-carbamoylation of NodNGR factors, respectively (15). HsnI contains five genes encoding a set of enzymes responsible for the synthesis GDP-fucose and its transfer by NodZ to NodNGR factors (16,17).
Here we present a molecular analysis of the HsnIII locus (12,13). In this, as in much of the work discussed above, we used the closely related R. fredii strain USDA257 because (a) it nodulates an exact subset of the NGR234 hosts; (b) many of the nod genes, and most of the essential, chromosomal genes, are extremely well conserved between the two strains (18 -20); and (c) perhaps because of (b), transconjugants are more stable in the R. fredii background than in any other rhizobia. Combined, these properties facilitate extension of host-range studies in which NGR234 clones are introduced into USDA257 on broad host range, multiple copy plasmids. Since the nodulation re-quirements of the two bacteria are known, the initial screening for particular hosts can be performed on legumes that are nodulated by NGR234 but not by USDA257.
These analyses show that HsnIII is downstream of nodABC, and includes the nodIJnolOnoeI genes. NoeI is required for 2-O-methylation of the fucose and NolO for 3-O (or 4-O)-carbamoylation of the nonreducing terminus of NodNGR factors.

EXPERIMENTAL PROCEDURES
Bacterial Genetics/Manipulation of DNA-Bacterial strains and plasmids used in this study are listed in Table I. Rhizobia were grown in/on RMM3 (21). Microbiological techniques were performed as described in Lewin et al. (14). Recombinant DNA procedures were used as described previously (19,22,23). DNA sequencing was performed using chain-termination inhibitors (24).
Construction of the Mutants NGR⍀nolO and NGR⍀nodI-Plasmid p6.8HR was linearized with the restriction enzymes BglII and EcoRI to liberate nodI, and nolO respectively. Concomitantly, the Sp R Omega fragment was excised from pHP45 with BamHI or EcoRI and ligated onto the linear p6.8HR fragments. After transformation into Escherichia coli XL.1 Blue, the plasmids were conjugated into a Rif R derivative of NGR234 with the help of plasmid pRK2013. Rif R Sp R Tc S colonies were checked for homologous recombination by Southern transfer. A 2.6-kb EcoRI fragment of pNG77 carrying noeI was first cloned into the EcoRI site of pBluescript SK. The resulting construct, pSK2.6E contains a unique StuI site that is located 26-pb downstream of the start codon (ATG) of noeI. Digestion of pSK2.6E with StuI permitted introduction of the Sp R Omega interposon (which was extracted from pHP45⍀ by digestion with SmaI) into the noeI gene (25). Finally, the entire noeI-Omega cassette was cloned into the XbaI and SalI sites of the suicide vector pJQ200sk (26). Triparental matings (including the helper plasmid pRK2013) were used to mobilize noeI-Omega into NGR234. Selection of NGR⍀noeI was performed as described previously (27).
Construction of nodU/nolO Double Mutants-The 2.2-kb HindIII/ BamHI fragment downstream of nodABCIJ, which contains nolO, was cloned into the HindIII and BamHI sites of pBluescript KS (giving pKS2.2B/H). Digestion of pBluescript KS with BamHI and HindIII led to deletion of the EcoRI site in the multiple-cloning site. Consequently, an EcoRI fragment carrying the Km R Omega interposon (25) could be cloned into the EcoRI site of nolO. The cassette containing nolO::⍀ was then cloned into XbaI and XhoI sites of pJQ200sk. Finally, the recombinant plasmid was transferred to two distinct nodU mutants: NGR⍀26 and NGR⌬1 that are a Sp R insertion mutant (in nodU) and a deletion mutant (nodSU) of NGR234, respectively (14). The two double mutants NGR⍀26⍀nolO and NGR⌬1⍀nolO were verified as described in Hanin et al. (27) Nodulation Tests-Seeds of Calopogonium caeruleum (Benth.) Hemsl. were purchased from the Inland and Foreign Trading Co., Indus  (28). Two replicate jars, each containing four seedlings, were used per treatment. Inoculation with 10 7 colony-forming units was performed 3 d after planting. All plants were grown at a daytime temperature of 30°C, a night temperature of 20°C, and a light phase of 16 h (including a 1-h stepped "sunrise" and a 1-h stepped "sunset"; maximum intensity of illumination was 350 mol⅐m Ϫ2 ⅐s Ϫ1 photosynthetically active radiation).
Analytical Methods-Separation of lipochitinoligosaccharides (LCOs) 1 by HPLC, their analysis by liquid secondary ion mass spectrometry (LSI-MS), B/E linked scans (where B/E ϭ fixed ratio of the magnetic (B) over the electric field (E) in daughter ion mass spectrometry), and one-dimensional NMR spectra were performed as described previously (15). 15 N-1 H-heteronuclear multiple quantum correlation NMR spectra of 15 N-enriched NodNGR factors were recorded in Me 2 SO-d 6 at 14 Tesla (600-MHz proton frequency) on an AMX-Z-600 spectrometer (Bruker GmbH, Karlsruhe, Germany). After alditol ace-tate derivatization, analyses of monosaccharides originating from the LCOs were performed by GC/MS (using an electron impact ion source) after separation on a 15-cm Supelco TM SP fused silica column (Hewlett Packard, Palo Alto, CA). The synthesis and analysis of monoacetylated glucosamines by MS/MS was performed as described in Ferro et al. 2

RESULTS
Host Range Extension-At the outset, large fragments (Х40 kb) of the symbiotic plasmid of Rhizobium sp. NGR234 (pNGR234a) were cloned into the nontransmissible cosmid vector pJB8 (22,30). Individual cosmids were mobilized into heterologous rhizobia by introducing the cis-acting DNA recognition site for conjugative DNA transfer (Mob site) of RP4 into the clones. This was accomplished by conjugating the Tn5-Mob vector pSUP5011 (31) into E. coli strains containing the pJB8 cosmids. Matings with another RP4 derivative permitted mobilization of the cosmids into various Agrobacterium/Rhizobium strains including Rhizobium loti strain NZP4010. R. loti (pWA54) transconjugants nodulated V. unguiculata at low frequency, and this locus was named HsnIII (12). Preliminary mapping showed that pWA54 partially overlaps with pWA46, but that the latter does not confer host-range extension on the transconjugants (13).
To search for a cosmid which would give a higher frequency of Nod ϩ transconjugants, a new clone bank was established in the broad host-range, transmissible, cosmid vector pRK7813 (32). Triparental matings were used to mobilize the individual pRK7813 cosmids into R. fredii strain USDA257. One clone Sites into which the spectinomycin resistant Omega cassette were inserted are marked by triangles. (ϩ) and (Ϫ) correspond to the phenotypes Fix ϩ or Nod Ϫ observed in USDA257 transconjugants containing different fragments of the HsnIII locus. nodI shares 67% identity (and 78% similarity) while nodJ shows 65% identity (and 77% similarity) to the same genes of R. leguminosarum (37).
(pNG77) was identified that shared homology with pWA54 and permitted USDA257 to nodulate C. caeruleum. Among the legumes tested, pNG77 extended the host range to Calopogonium and L. purpureus but had no effect on nodulation of the other plants (Table II). Examination of the nodulation capacity of R. fredii transconjugants harboring either HsnII (ϭnodSU, contained on pA18) or HsnIII (carried on pNG77), clearly demonstrated that these are proper hsn loci. The nodSU genes permit nodulation of Leucaena leucocephala (but not Calopogonium and Lablab), while the reverse is true for HsnIII (Table II).
Delimitation of the HsnIII Locus-A series of subclones in pRK7813 were generated and mobilized into USDA257. The resulting transconjugants were used to inoculate Calopogonium, Lablab, and other legumes. Interestingly, the 7.6-kb EcoRI fragment of pNG77, which contains regions on either side of nodABC, was unable to extend the host range (Table II).
On the other hand, clones containing the right most of two 6.8-kb HindIII fragments (p6.8HR) (Fig. 1), conferred the ability to nodulate both C. caeruleum and L. purpureus on the transconjugants. These data suggest that the HsnIII locus lies at the end of p6.8HR distal to nodABC. Surprisingly, when the transconjugants were reisolated from Calopogonium nodules, they had lost resistance to tetracycline (i.e. they were Km R Tc S ), but retained their ability to nodulate Calopogonium. DNA isolated from nodules induced by USDA257(pNG77), and hybridized against an internal nodC probe of NGR234, showed that the restriction pattern of the USDA257 nodC locus in the re-isolates had changed to that of NGR234. 3 Similar Km R Tc S reisolates were obtained from nodules induced by USDA257(p6.8HR) transconjugants. Interestingly, the restriction pattern of the USDA257(p6.8HR) nodC locus was identical to that obtained from USDA257(pNG77) transconjugants, suggesting that the common nod genes play a role in homologous recombination between the two species. R. fredii (pNG77) or R. fredii (p6.8HR) reisolates from C. caeruleum nodules were stable as judged by their ability to nodulate both Calopogonium and Lablab when reinoculated onto them (Table II).
Mutational and Sequence Analysis of HsnIII-As R. fredii transconjugants harboring p7.6E and p2.2B/H (Fig. 1) were unable to nodulate C. caeruleum while those containing p6.8HR were able to do so (Table II), the HsnIII locus is most probably located in the region of overlap between p7.6E and p2.2B/H. To delimit possible genes in this region, we sequenced the plasmids represented by p7.6E and p2.8HR. As confirmed by complete sequencing of pNGR234a (33), seven open reading frames are contained in this locus (Fig. 1). A noncoding sequence of 173 bp separates nodABC from the 1,005-bp nodI. After a gap of only 3 bp, this is followed by another ORF of 789 bp that is highly similar to nodJ. nolO is downstream of nodJ and is followed by part of noeI. Insertion of an Omega cassette into either nodI or nolO or noeI had no effect on the capacity of the mutant to form nodules (Table II).
RNA competition/hybridization experiments were used to see whether the loci contained on pNG77 and p6.8HR were inducible by flavonoids. One h after exposure to apigenin, all of the fragments contained on p6.8HR (e.g. the 1.7-and the 3.6-kb BamHI fragments) as well as the 7.6-kb EcoRI fragment of pNG77 hybridized, showing that the entire locus is inducible 3 B. Relić, unpublished results. (data not shown). Although these data do not constitute proof, they suggest that nodABCIJnolOnoeI are transcribed as one operon.
Involvement of HsnIII Genes in 2-O-Methylation of the Fucose-Thin layer chromatographic analyses of the supernatants from [ 14 C]glucosamine-fed cultures of NGR⍀nodI, NGR⍀nolO, and NGR⍀noeI did not reveal differences from the wild-type bacterium (data not shown). We were thus able to exclude a role for these genes in fucosylation or sulfation of NodNGR factors. Other possible roles of the hsnIII genes were studied by subjecting partially purified (i.e. eluted from a C 18reverse-phase column) Nod factors to acidic methanolysis. Hydrophilic compounds were separated from the hydrophobic components by hexane extraction. GC/MS analyses of the methyl-esterified hexane-fraction did not reveal any changes in the N-linked acyl chain attached to the nonreducing terminus. This way, we were able to rule out any affect of the hsnIII genes on fatty acid synthesis.
In another approach (and after alditol-acetate derivatization), the monosaccharide compositions of the mutants were compared with those produced by the wild-type bacterium. Interestingly, the peak corresponding to 2-O-methyfucose was absent in Nod factors produced by all three mutants, while those corresponding to fucose, N-acetylglucosamine (GlcNAc), and N-methylglucosamine (MeGlcN) were present (see Fig. 3). This indicates that one or combinations of all three mutations in hsnIII prevent 2-O-methylation of the fucose. Since nodAB-CIJnolOnoeI probably form an operon (see above and Fig. 1), and each of the Omega insertions has a polar effect on expression of the down-stream genes, the simplest interpretation of the phenotype is that NoeI is the methyltransferase. To test this, supernatants of NGR⍀noeI cultures were purified by reverse phase (RP)-HPLC, and the appropriate fractions were analyzed by LSI-MS in the positive ionization mode. Fragment ions generated from the nonreducing terminus were unchanged in comparison to those given by wild-type extracts, indicating that the N-methyl and the set of carbamoyl groups are present (Fig. 2C). In contrast, the mass of all pseudomolecular ions [M ϩ H] ϩ was 14 Da less than those produced by the wild-type bacterium. To localize the origin of the 14-Da ion, metastable ion spectra (MS/MS) of the major Nod factors were taken. It is clear from these analyses that the fucose moiety lacks the methyl group since the first significant fragment ion corresponds to the loss of fucose (m/z 146) rather than methylfucose (m/z 160). Also, the 1 H NMR spectrum lacks the signal of the 2-O-methyl at ␦ ϭ 3.35 ppm. Thus mutation on noeI seems to suppress the production of NodNGR-factors containing 2-O-methylated fucose (Fig. 3).
Nod Factors Produced by NGR⍀nolO-The LSI-MS spectra (positive ionization mode), of RP-HPLC-purified compounds extracted from the supernatants of NGR⍀nolO cultures revealed [M ϩ H] ϩ molecular ions which correspond to sulfated or acetylated molecules. Fragment ions from the nonreducing terminus differ from analogous ions of Nod factors of the wild-type bacterium by the absence of bis-carbamoylated products (absence of m/z 528, 526, and 498), which correspond to LCO species possessing two carbamates, acylated with C 18:0 , C 18:1 , and C 16:1 , respectively, and N-methylated. Molecules in which the fucose is acetylated gave pseudomolecular ions that were shifted down by 57 Da in comparison to ions from NodNGR factors ( Fig. 2A). This 57-Da difference corresponds to the sum of the mass of carbamoyl (43 Da) and methyl (14 Da) groups. B/E spectra were performed on several molecular ions of LCOs extracted from the NGR⍀nolO mutant. Among them, the [M ϩ H] ϩ ion at m/z 1459 revealed the following fragment ions: m/z 1313, 1092, 889, 686, and 483. Thus the ion m/z 1313 is derived from m/z 1459 by the loss of fucose (146 Da) rather than methylfucose (160 Da). Similarly, the loss of 43 Da must be from the nonreducing terminus, since no fragment ion corresponding to bis-carbamoylated ions (m/z 526) was observed. 13 C NMR showed the absence of signals at 60, 156.1, and 155.7 ppm, confirming that no methyl group was attached to the fucose and that only one carbamoyl group was present on the nonreducing terminus.

FIG. 4. Constant B/E scans of B1 oxonium ions (A-C) and periodic acid oxidation of the N-methylated, N-C
In this way, we were able to provisionally assign the 2-Omethyation function to NoeI, and carbamoylation at the nonreducing terminus to NolO. Examination of Nod factors produced by NGR⍀nodI also showed the loss of one carbamoyl and the 2-O-methyl group, confirming the organization of nodI-JnolOnoeI genes in the operon. Nevertheless, the height of the HPLC peaks of LCOs produced by the NodI Ϫ mutant were only half those produced by loading extracts prepared from the same quantity of wild-type bacteria (data not shown).
Nod Factors Produced by Various USDA257 Transconjugants-As shown above, when p7.6E, which contains nodABC-nodIJ (Fig. 1), was mobilized into USDA257, the transconjugants are unable to nodulate C. caeruleum. On the other hand, p6.8HR, which also contains nolO, was able to confer on R. fredii the capacity to nodulate C. caeruleum (Table II). This suggests that nolO (but not nodIJ) is necessary for nodulation of this plant. To determine the biochemical role of nolO, we studied Nod factors produced by USDA257(p6.8HR). LSI-MS spectra from the RP-HPLC peaks showed fragmentation sequences separated by 203 Da, which is characteristic of GlcNAc oligomers. The major peak, which corresponds to Nod factors containing five GlcNAc, N-C 18:1 , and methylfucose gave [M ϩ H] ϩ ions at m/z 1459 instead of m/z 1416 from wild-type USDA257. This 43-Da augmentation in mass corresponds to a carbamoyl group. It is obvious that the carbamoyl group originates from the nonreducing terminus since all fragments were shifted up by 43 Da, particularly the B1 ion which arose at m/z 469 rather than m/z 426. Analyses of all HPLC fractions showed partial carbamoylation of tri-, tetra-, and pentameric Nod factors (Fig. 3).
Nod Factors Produced by nodU/nolO Double Mutants-To establish whether all carbamoylated NodNGR factors are the products of NodU and NolO, double mutants in the two genes were made. Construction of the double mutant was performed by inserting the Sp R Omega cassette into nolO of NGR⌬1 in which a 9-kb fragment containing the complete hsnII (nodSU) locus has been deleted (see "Experimental Procedures"; Table  I not yield fragment ions corresponding to N-methylated or biscarbamoylated Nod factors (Fig. 2B), but the monocarbamoylated forms persisted (m/z 455, 469, and 471). Unexpectedly, the B1 oxonium ions of mass spectra generated from different fractions of the supernatants of NGR⌬1⍀nolO were similar to those from NGR⌬1, showing that mutation of nodU and nolO were not sufficient to completely abolish carbamoylation (data not shown). The 2-O-methyl group was however absent in the supernatants of cultures from NGR⍀nolO ( Fig. 2A) and NGR⌬1⍀nolO. Moreover the pseudomolecular ions from NGR⌬1⍀nolO or NGR⌬1 revealed a short Nod factor backbone (3 or 4 GlcNAcs) confirming the role of NodS in ensuring pentameric NodNGR factors (15). In some cases, the molecular ions of the tri-and tetrameric LCOs were similar to fragment ions from the pentameric forms (see Fig. 2B, at 1299.7, 1256.7, 1096, and 1053 Da), but the relative abundances of the noncarbamoylated and monocarbamoylated forms were reversed. MS and MS/MS analyses of pure HPLC fractions of these short LCOs clearly proved these structures. The monocarbamoylated forms were also observed in the LSI-MS spectra of LCOs purified from NGR⍀nodU⍀nolO.
Specific Carbamoylation by NolO-Further evidence that NolO specifies carbamoylation on positions other than O-6 was obtained by tandem mass spectrometry and by 15 N or 13 C NMR spectroscopy. Using model compounds, we showed that the metastable ion spectra of the oxonium ions from all isomers of mono-O-acetylated, N-acetylglucosamine were different and that 3-O-and 6-O-carbamates behaved as the corresponding O-acetates (4-O-acetate was not used as a reference compound). 2 In an attempt to locate their O-carbamoyl substitutions, these experiments were extended to NodNGR factors. Metastable decomposition of the B1 ion from Nod factors isolated from NGR⍀nodU showed the prominent loss of carbamic acid (Fig. 4B). This behavior was very similar to that found with 3-O-acetyl-N-acetylglucosamine in which the 3-O-substituent, acetic acid, was eliminated. It was different from that found with the 6-O isomer however, which was characterized by the loss of water. In contrast, the B1 ions from Nod factors produced by the NGR⍀nolO mutant were characterized by the prominent loss of water (Fig. 4C), clearly locating the remaining carbamoyl group on C-6. This ion spectrum was identical to that found with Nod factors from Azorhizobium caulinodans (NodARc) in which the carbamate group has been unambiguously assigned to O-6. Indeed, the metastable ion spectra of the fragment ion at m/z 483, which corresponds to monocarbamoy-lated NodNGR factors, gave fragment ions at m/z 465 and 422 (Fig. 4A) showing the loss of carbamic acid or water. This suggests that the carbamoyl group could be at either of positions O-3 or O-6.
Nod factors from USDA257(6.8HR), in which the nolO gene has been introduced, were submitted to the same type of analyses. Results identical to those obtained using the NGR⍀nodU mutant confirm the specificity of NolO for the 3-O position of the nonreducing terminus. Carbamoylation of the C-6 position did not occur as cleavage of the C-3-C-4 bond by periodic acid was not detected in LCOs produced from either NGR⍀nodU (Fig. 4D) or in USDA257(p6.8HR). In contrast, periodate oxidation of LCOs from NGR⍀nolO, followed by sodium borodeuteride reduction, showed a 4-Da shift in the B1 ions, which results from periodic acid cleavage of the two adjacent hydroxyl groups (Fig. 4E).
We attempted to corroborate these results using heteronuclear 15 N-1 H two dimensional NMR spectroscopy. To this purpose, Nod factors were isolated from large scale apigenininduced NGR234 cultures in which 15 (NH 4 ) 2 SO 4 (11% 15 Nenriched) replaced glutamate as the nitrogen source. Two-dimensional NMR spectra of acetylated Nod factors showed two 15 N/ 1 H cross peaks at ␦15 N /␦1 H Ϫ308/6.5 and Ϫ265/7.9, corresponding to 1 J H coupling within the carbamate group and the amide of GlcNAc units, respectively. Unfortunately, the difference of chemical shifts between 15 N atoms linked to O-3 (or O-4) versus O-6 was too small to be detected by this method.
NodU is responsible for the 6-O-carbamoylation of NodNGR factors (15). 13 C NMR spectra in deuterated Me 2 SO of NodNGR factors gave three signals at ␦ ϭ 156.5, 156.1, and 155.7 ppm (Fig. 5A) (15), indicative of carbamate groups at O-3, O-4, and O-6 ( Fig. 3). Mild alkaline hydrolysis (in 0.1 M KOH), which removes substituents linked by carbamic ester bonds, resulted in the complete disappearance of all three signals (Fig. 5F). The 13 C NMR spectrum of supernatants taken from NGR⍀nodU cultures (in which the nolO gene is still active), lack the signal at 156.5 ppm but retain those at 156.1 and 155.7 ppm (Fig. 5B). This implies that NolO carbamoylates either O-3 or O-4 but not O-6 of the nonreducing terminus (Fig. 6). 13 C NMR spectra of the major Nod factors produced by NGR⍀nolO (in which NodU is still active) only gave a signal at ␦ ϭ 156.5 ppm (Fig. 5C), confirming that NodU carbamoylates only position O-6 ( Fig. 6). 13 C NMR spectra from the double mutants (NGR⍀nodU⍀nolO or NGR⌬1⍀nolO) show a persistent signal at ␦ ϭ 156.5 ppm (Fig. 5D). Since the Omega interposon contains the transcrip- Boxes represent identical residues in all six sequences. nolO of NGR234 is 72% identical (and 82% similar) to nolO of B. japonicum, even though it is 211 amino acids longer (38). In the central domain, nolO of NGR234 shares 60% similarity and 40% identity with nodU of A. caulinodans (35). tion termination signals of the bacteriophage T4D gene 32 (34), it is highly unlikely that any residual 3 (or 4)-O-carbamoyltransferase activity of NolO remains in NGR⍀nolO. It is thus likely that a third carbamoyltransferase specific for the O-6 of the nonreducing terminus exists. Since no other genes homologous to either nodU or nolO were found on pNGR234a (33), this suggests that the third carbamoyltransferase must be on another replicon. DISCUSSION Since mutation of noeI has no effect on nodulation while host-range extension (complementation) studies with nolO render USDA257 Nod ϩ on C. caeruleum, it seems likely that nolO is the principal host range determinant of the hsnIII locus. R. fredii transconjugants harboring p6.8HR (which contains the complete nolO gene but only part of noeI) or p6.8HRCRep (which contains a frameshift in noeI) allow USDA257 to nodulate C. caeruleum. On the other hand, transconjugants harboring p7.6E (in which nolO is truncated) or p6.8HRERep (having a frameshift in nolO) are unable to nodulate. As the fucosyl moiety of USDA257 Nod-factors is 2-O-methylated, an equivalent of the noeI gene must be active in this strain. Yet USDA257 does not nodulate C. caeruleum, confirming that adjunction of the 2-O-methyl group to the fucose is not essential for nodulation of this species. Mutation of nodI has a polar effect on the expression of the nodABCI-JnolOnoeI operon and inhibits both the bis-carbamoylation and 2-O-methylation of NodNGR factors. Thus, through elimination it seems likely that NoeI is the 2-O-methyltransferase.
The predicted NolO protein is highly homologous to NodU of various rhizobia as well as a carbamoyltransferase of Nocardia lactamdurans (Fig. 7). From the present work, NolO seems to be specific to position O-3 or at least to positions O-3 and O-4 of the nonreducing terminus (Fig. 6). In A. caulinodans and R. loti, a single carbamate group was found at position O-6 (35) or position O-4 (36), respectively. Although position O-6 had clearly been established by a set of convergent methods, some doubt remained about positions O-3 and O-4 because of a possible isomerisation occurring during purification. Despite this reservation, it seems clear that at least two specific carbamoytransferases exist in rhizobia. The present work, based on LSI-MS, the absence of periodic cleavage, and 13 C NMR spectra confirm that NolO carbamoylates one of the secondary hydroxyl groups of the terminal non-reducing GlcNAc. The continued production of 6-O-carbamoylated NodNGR factors in the double mutant NGR⍀nodU⍀nolO, points to a third carbamoyltransferase specific of the O-6 position. The physical location of this third carbamoyltransferase remains to be determined, since no other NodU/NolO homologues are present on pNGR234a (33). Also, since 6-O-carbamoylated NodNGR factors cannot be detected when nodU is mutated (15), the second 6-O-carbamytransferase must be less efficient than NodU (a few percent of the 6-O-isomer cannot be ruled out by MS and NMR data) or repressed when nolO is activated.
Finally, since Leucaena sp. and Calopogonium/Lablab react differently to NodNGR factors modified by NodU and NolO respectively (Table II), it seems likely that both groups of plants have specific requirements for the position of the carbamoyl group. On the other hand, the absence of clear phenotypes with the NGR⍀nolO and NGR⍀nodU⍀nolO mutants is probably due to the third, carbamoyltransferase.