Involvement of nodS in N -Methylation and nodU in 6- O Carbamoylation of Rhizobium sp. NGR234 Nod Factors*

Although Rhizobium sp. NGR234 and Rhizobium fredii USDA257 share many traits, dysfunctional nodSU genes in the latter prohibit nodulation of Leucaena species. Accordingly, we used R. fredii transconjugants harboring the nodS and nodU genes of NGR234 to study their role in the structural modification of the lipo-oligosac-charide Nod factors. Differences between the Nod factors mainly concern the length of the oligomer (three to five glucosamine residues in USDA257 and five resi- dues only in NGR234) and the presence of additional substituents in NGR234 ( N -linked methyl, one or two carbamoyl groups on the non-reducing moiety, acetyl or sulfate groups on the fucose). R. fredii ( nodS ) transcon- jugants produce chitopentamer Nod factors with a N linked methyl group on the glucosaminyl terminus. In- troduction of nodU into USDA257 results in the formation of 6- O -carbamoylated factors. Co-transfer

Symbiotic soil bacteria of the genera Azorhizobium, Bradyrhizobium, and Rhizobium (collectively termed rhizobia) interact with the roots of legumes to form nodules in which atmospheric nitrogen is reduced to ammonia. Signal exchange between the symbionts regulates the expression of both bacterial and plant genes involved in nodule development (1,2). Flavonoids excreted by the legume roots activate both common and host-specific nod genes which direct the synthesis of lipooligosaccharide Nod factors. Secretion of these hormone-like substances into the plant rhizosphere induces root hair deformation and curling (3,4), the formation of preinfection threads (5), and nodule-like structures (6). On a number of legumes, Nod factors permit various natural or engineered Nod Ϫ mutants to enter the legume roots and to form nitrogen-fixing nodules (4,7). All Nod factors so far identified are ␤-1,4-linked tri-or tetra-or pentamers of N-acetyl-D-glucosamine, N-acylated at the non-reducing end, and N-acetylated on the other residues (1,8). Essential differences among the Nod factors of the various species concern the substituents linked to both ends of the chitinic backbone. Among the substitutions found on the terminal non-reducing N-acetyl-D-glucosamine are an N-methyl group, carbamoyl groups, acetyl groups, and various fatty acids. Similarly, the reducing N-acetyl-D-glucosamine residue may be substituted with a sulfate group or with D-arabinose, L-fucose, or 2-O-methylfucose. Furthermore, this additional saccharide may be acetylated or sulfated (1,9).
Mutations in the nodABC genes, which are common to all rhizobia, abolish Nod factor production. nodC shares homology with chitin synthases (10), an observation which has been supported by in vitro studies (11). It thus seems likely that the first step in Nod factor synthesis involves the assembly of N-acetyl-D-glucosamine subunits by the N-acetylglucosaminyltransferase coded by nodC (12). When the growing chain reaches three to five residues, NodB probably removes the N-acetyl moiety of the non-reducing end (11). NodA is an Nacyltransferase which links the acyl chain to the free NH 2 group on the oligomers synthesized by the NodC and NodB proteins (13,14).
Host-specific nod genes are involved in the addition of extra groups to the core lipo-oligosaccharides. Modified factors permit nodulation of specific plants (16). For example, the nodH and nodPQ genes of R. meliloti are involved in the sulfation of the oligosaccharide signals (16). NodL shares homology with O-acetyltransferases; in vitro studies showed that NodL 6-Oacetylates the non-reducing glucosaminyl residue (17). nodE and nodF are involved in synthesis of the lipid chain (18,19), while nodZ and nolO play a role in fucosylation of Bradyrhizobium japonicum nodulation signals (20,21). The nodSU genes of Rhizobium. sp. NGR234 form an operon, the activity of which is required for nodulation of Leucaena leucocephala (22). nodSU are also present in R. fredii USDA257 (23), B. japonicum (24), and Azorhizobium caulinodans (25), but no correlation exists between the ability to nodulate Leucaena and the presence of nodSU in these organisms (23). Moreover the nodSU genes from B. japonicum are unable to complement a nodS Ϫ mutant of NGR234 (24). On the other hand the transfer of nodSU from NGR234 to R. fredii USDA257 confers on the latter the ability to nodulate Leucaena (23). NodS shares similarities with S-adenosylmethyl transferases (25,26). Support for this latter function has been obtained by in vitro labeling studies. The role of NodU remains to be elucidated.
Functions of nodS and nodU of NGR234 were determined by introducing nodSU, nodS, and nodU into R. fredii USDA257 and by analyzing the Nod factors produced by the transconjugants. Analysis of the Nod factors produced by nodS Ϫ and nodU Ϫ mutants of NGR234 confirmed that the product of the nodS gene is involved in N-methyltransferase activity and that the 6-O-carbamoyltransferase activity is dependent upon expression of the nodU gene.

EXPERIMENTAL PROCEDURES
Bacteriological and Molecular Methods-Rhizobia, along with their relevant characteristics, are listed in Table I. Microbiological techniques were performed as described in Lewin et al. (22). Nodulation tests were conducted in Magenta TM -type Leonard jars (22). DNA sequences were determined using the dideoxy chain termination procedure using double-stranded templates and the Sequenase II TM kit (U. S. Biochemical Corp.).
Placement of nodU Under Control of the nodSU nod Box-The cohesive ends of the 250-base pair PvuI-EcoRI fragment containing the nod box of the nodSU operon were filled in and ligated into the SmaI site of pBS(ϩ). Orientation of the insert in the resulting plasmid pNBSU was checked by sequencing. Concomitantly, the nodS gene (as well as the accompanying upstream sequences) was deleted from pA18 by double digestion with BamHI and EcoRI and replaced by the nod box of pNBSU (Fig. 1). The resulting plasmid pRAF25 was introduced into R. fredii USDA257 using the helper plasmid pRK2013.
Purification of Nod Factors-Nod factors were isolated from apigenin-induced (10 Ϫ6 M) culture media. Solid-phase extraction onto large scale C 18 reverse-phase columns, followed by washing with ethyl acetate and preparative HPLC 1 chromatography was performed essentially as described previously (27). Fractions reacting positively with anthrone were repurified on a C 18 reverse-phase analytical column. A 20 -90% methanol gradient was used for elution. Further purification was achieved on a propylamine-linked analytical column using a linear gradient of 100% acetonitrile to water/acetonitrile (80:20 (v/v)).
Tomato Cell Suspension Assay-Fifty ml of the appropriate cultures in RMM3 (27) were centrifuged (4,000 ϫ g, 4°C), applied to 3-ml Sep-Pak C18 cartridges (Waters Associates, Milford, MA), and the Nod factors eluted with methanol. Dried samples were dissolved in Me 2 SO/ ethanol, and appropriate dilutions added to suspension cultures of tomato cells (28).
Analytical Methods-Gas chromatography was performed on an OV1-bound capillary column (0.32 mm ϫ 30 m, Spiral, France) using a temperature gradient of 100 -280°C (2°C/min). Mass spectra were recorded on an Autospec instrument (Fisons, VG-Analytical, Manchester, United Kingdom). The acceleration voltage was 8 kV. A glycerol matrix (1 l), either acidified with 0.13 M trifluoroacetic acid (1 l) or spiked with 10% sodium iodide (1 l), was used in the positive-ion mode. Glycerol-thioglycerol (1:1) was the matrix used for ionization in the negative-ion mode. B/E-linked scans were performed to study metastable ion decay in the first field-free region without the use of a collision gas. NMR spectra were recorded on a Bruker 400 MHz spectrometer (9.4 Tesla) using deuterated dimethyl sulfoxide as the solvent and as the secondary reference ( 13 C, ␦ ϭ 39.5 ppm).

Nod Factors Produced by nodU Ϫ and nodSU Ϫ Mutants-
Reverse-phase HPLC analysis of the supernatants from NGR⍀26 (ϭ nodU Ϫ ) (see Fig. 2) cultures revealed the same two Nod factor peaks as in the overproducing strain. The first peak corresponds to a mixture of sulfated compounds, the second to non-sulfated Nod factors. FAB-MS spectra of NodNGR factors of the mutant strains were compared with those previously published (27). In the positive-ion mode, the spectra of compounds from peak A gave [MϪHϩNa] ϩ ions by sodium attachment of m/z 1597, 1569, 1554, and 1526 which are characteristic of sulfated Nod factors. Fragmentation of the non-reducing sugar yielded similar molecular ions which correspond to the glucosaminyl residue at the non-reducing end. They differ from analogous fragments of Nod factors originating from the overproducing strain NGR(pA28) by the absence of bis-carbamoylated ion products (lacking m/z 526 and 498 which correspond to species with two carbamates and C 18:1 or C 16:1 , respectively).
Peak B contains Nod factors with an acetylated or nonsubstituted 2-O-methylfucose. As with the sulfated molecules, they differ from those of the NGR(pA28) by the presence of only one carbamoyl group (Fig. 2). Again, this is clearly seen on the B1 oxonium ion fragment (data not shown). Thus, a mutation in nodU seems to suppress the production of bis-carbamoylated products (Fig. 2). Analyses of the supernatants of the nodSU Ϫ mutant (ϭ NGR⍀25) (22) revealed a dramatic reduction in Nod factor production in comparison with the wild-type or overproducing strains (data not shown). This was shown both by thin layer chromatography and by direct measurement of the amounts of chitin oligomers in the supernatants using the tomato cell suspension assay (28,29). In contrast, the nodU Ϫ mutant produced similar amounts of Nod factors to the wild-type bacterium. Together, the low amounts and complex mixtures of the biologically active products produced by the nodSU Ϫ strain rendered their analysis difficult. Nevertheless, the FAB-MS spectra in the negative-ion mode gave [MϪH] Ϫ ions predominately at m/z 1494, 1537, 1466, and 1509 which correspond to non-N-methylated but fucosylated, sulfated, C 18:1 -acylated, mono-or non-carbamoylated pentamers, and non-methylated, fucosylated, sulfated, C 16:1 -acylated, mono-or non-carbamoylated pentamers, respectively. Exhaustive butanol extraction of the membrane fraction derived from NGR⍀25 cells after lysis with a Frensh press did not reveal significant amounts of Nod factors. Together, these data show that expression of nodS and nodU modulates the synthesis, methylation, and bi-carbamoylation of NodNGR factors.
Nod Factors Produced by R. fredii Transconjugants Harboring nodSU, nodS, and nodU-To circumvent problems linked to mutagenesis (e.g. operon structure, duplication of genes, etc.), we studied Nod factors produced by R. fredii transconjugants harboring nodSU (pA18) or nodS (pA26) and a truncated nodU (pA16) (Fig. 1). All three transconjugants are able to nodulate Leucaena (22). As the published structures of the USDA257 factors (30) and those ascertained in this laboratory differed only by an additional molecule containing a C 16:1 acyl chain, this approach is valid (Fig. 2).
Using the tomato cell suspension assay, we found that R. fredii (pA16) transconjugants produce 20 times more Nod factors than wild-type USDA257. HPLC analysis of the supernatants revealed three peaks (A-C) in the ratios 0.35:1:0.6, respectively. FAB spectra measured from the peaks showed a fragmentation sequence separated by 203 mass units which is characteristic of N-acetyl-D-glucosamine oligomers. Peak A gave [MϩNa] ϩ at m/z 1424 and an associated peak (5%) yielded [MϩNa] ϩ at m/z 1410. The former corresponds to Nod factors containing five N-acetyl-D-glucosamine residues, a C 16:1 acyl chain, 2-O-methylfucose, and an additional methyl group. The smaller peak (5%) lacks the 2-O-methyl substitution on the fucose seen in wild-type factors.
[MϩNa] ϩ ions of peak B arose at m/z 1452. Again, this mass is 14 Da more than that of chitopentameric Nod factors of wild-type USDA257 possessing a C 18:1 acyl chain. Sodium attached molecular ions from peak C [MϩNa] ϩ arise at m/z 1454, 2 Da higher than those of peak B which correspond to an acyl chain of C 18:0 rather than C 18:1 . To locate the additional methyl group (14 Da), all products were studied in the presence of an acidified matrix to enhance the formation of MH ϩ ions and their fragmentation. It is clear that the methyl group is borne on the ion fragment corresponding to the non-reducing sugar (oxonium B1 ion). This result was confirmed by the metastable decay of MH ϩ ions (B/E scans) (Fig.  3B). The ion fragment at m/z 440 which arises from decomposition of the m/z 1430 (peak B) corresponds to the non-reducing D-glucosamine end substituted by a C 18:1 acyl chain and an additional methyl group (14 Da).
Localization of the methyl group was performed by (a) 1 H NMR analysis, (b) 13 C NMR analysis, and (c) using gas chromatography of the hydrolysis products from peak B. The 13 C NMR spectra gave a signal at ␦ ϭ 27.3 ppm which corresponds to N-CH 3 . In 1 H NMR (Fig. 4B), the singlet at ␦ ϭ 2.7 ppm corresponds to the chemical shift of a methyl group bound to an amide nitrogen as in NodNGR and other N-methylated Nod factors (26,27). This 1 H NMR singlet is absent from the Nod factors of R. fredii USDA257 (30). Moreover, comparison of the gas chromatograph retention times of peracetyl derivatives of authentic N-methyl-D-glucosamine and of the acid hydrolysates of peak B, confirmed the presence of an N-methyl group on the glucosaminyl end.
Another observation of relevance to the role of nodS concerns the fact that wild-type USDA257 produces a majority of pentamers with relatively small amounts of tetramers and trimers (30). Introduction of nodS into R. fredii results not only in N-methylation of the Nod factors, but also in the complete disappearance of products containing three or four glucosamine residues (Fig. 2). Even though pA16 lacks only the carboxyl terminus of nodU, and the Omega (⍀) insertion in pA26 is very close to the N terminus, USDA257 transconjugants harboring either pA16 or pA26 have the same phenotype (22). Given these data, it is not surprising that the Nod factors produced by R. fredii (pA16) and R. fredii (pA26) are identical.
To discriminate between the roles played by nodS and nodU, the nod box of the nodSU operon was fused to the nodU gene. This way, the nodU gene could be expressed independently of nodS but under the control of the same promoter (Fig. 1). The resulting plasmid pRAF25 confers the ability on USDA257 to nodulate Leucaena leucocephala (data not shown). In contrast to USDA257(nodS) transconjugants, R. fredii containing nodU (pRAF25) produces comparable amounts of Nod factors to those made by the wild-type USDA257 (tomato cell suspension culture assay). The HPLC profile of USDA257(pRAF25) transconjugants shows three peaks, D-F, which are present in the proportions 1:0.4:0.2, respectively. Using similar techniques to those described above, components with three, four, and five N-acetyl-D-glucosamine residues were found with a predominance of the pentamer as in wild-type USDA257. Mass spectrometry showed that the methylfucose was still present. Similarly, fraction D gave B1 ions at m/z 441 and 398. The latter corresponds to glucosamine bearing a C 16:1 acyl chain as in Nod factors of wild-type USDA257. In the former, the shift up of 43 mass units is indicative of the presence an additional carbam-oyl group. Similarly, fractions E and F contain mixtures of molecules that are either mono-or non-carbamoylated and possess either C 18:1 or C 18:0 acyl chains and three, four, or five N-acetyl-D-glucosamine residues (Fig. 2). B/E spectra performed on the different [MϩH] ϩ ions (Fig. 3A) confirmed the presence of an additional 43 mass units on the non-reducing sugar. Thus, introduction of nodU into R. fredii induces a partial mono-carbamoylation at the non-reducing end.
Conjugation of both nodS and nodU (pA18) into R. fredii produced Nod factors than can be separated into three HPLC peaks which gave [MϩH] ϩ ions at m/z 1445, 1473, and 1475, indicating a chitopentameric backbone (data not shown). B1 ions of these peaks at m/z 455, 483, and 485 correspond to N-acetyl-D-glucosamine possessing acyl chains of C 16:1 or C 18:1 or C 18:0 , respectively, that are shifted up by 57 mass units in comparison with ions from USDA257 Nod factors. This 57 Da difference corresponds to the sum of the mass of two groups: 43 Da for carbamate and 14 Da for the N-methyl group. B/E spectra performed on the [MϩH] ϩ ion at m/z 1473 confirmed that these groups are attached to the non-reducing end (Fig.  3C). A 1 H NMR signal at ␦ ϭ 2.7 ppm corroborated the presence of an N-methyl group which also gave a signal at ␦ ϭ 27.2 ppm in 13 C NMR (Fig. 4A). The same spectrum confirmed the presence of a 6-O-carbamoyl group at ␦ ϭ 156.5 ppm (Fig. 4A). The minor products (5% of the total) detected by FAB-MS also correspond to products containing five N-acetyl-D-glucosamine residues which are partially carbamoylated but not 2-O-methylated (data not shown).
Location of the Carbamoyl Groups-13 C NMR analysis in FIG. 2. Structure of the major Nod factors produced by Rhizobium. sp. NGR234, R. fredii USDA 257, and various derivatives thereof. n is the number of N-acetyl-D-glucosamine residues. R1 represents acyl chains, the carbon length and double bonds of which are indicated in the table (Carb ϭ carbamate group). The fatty acid compositions of these products were confirmed by GC analysis using authentic standards and by 1 H NMR (data not shown). N-Acylated C 18:O molecules were also observed in Nod factors produced by the wild-type strain NGR234, but not the over-producing strain NGR(pA28) (S. Jabbouri, unpublished results). deuterated-Me 2 SO of Nod factors produced by the over-producing strain NGR234(pA28) gave three signals at 156.5, 156.1, and 155.7 ppm, indicative of carbamate groups possibly at C-3, -4, or -6 (data not shown). In addition, FAB-MS spectra of the non-sulfated products showed B1 ion fragments at m/z 526, 483, and 440, which correspond to bis-, mono-, and non-carbamoylated species, respectively, that are N-methylated and possess a C 18:1 acyl chain. Since it was not possible to separate these compounds by HPLC, all attempts to identify the exact positions of the carbamoyl groups failed. Accordingly, another approach based on ␤-elimination of the C-3 (or C-4) substituents on the B1 oxonium ion, and using factors produced by different strains, was chosen. Indeed, the H-C(2) or H-C(5) proton of B1 ions generated by fragmentation of [MϩH] ϩ ions is acidic and leads to facile ␤-elimination of the C-3 (or C-4) substituent. Thus, the metastable ion spectra (B/E scans) of the ion fragment at m/z 483 corresponding to the mono-carbamoylated species (of NodNGR(pA28) factors) gave fragments at m/z 465 and 422 (Fig. 5B). This loss of 61 Da (carbamic acid) or 18 Da (H 2 0) suggests that a carbamoyl group is partially present at C-3 or at C-4. Data obtained by ␤ elimination do not exclude, however, that carbamoyl groups in mono-carbamoylated NodNGR factors could be at C-6. Nod factors of Azorhizobium caulinodans (which are carbamoylated at C-6) were used to verify this technique (see Fig. 5A).
Constant B/E scan spectra of the ion fragment at m/z 483 of Nod factors (mono-carbamoylated species) produced by NGR⍀26 (ϭ NodU Ϫ ) shows an ion fragment at m/z 422 which originated from ␤ elimination of carbamic acid (Fig. 5C). This indicates that the carbamoyl group produced by NodU Ϫ strains is at least partially located on C-3 or C-4. Periodate oxidation followed by reduction with NaBD 4 (26) did not give cleavage products containing a carbamoyl group, ruling out carbamoylation at position C-6. In contrast, periodate cleaves the noncarbamoylated Nod factors which are also present, as revealed by a 4 mass units shift up of the B1 ion following reduction by NaBD 4 (data not shown). In other words, mutational analysis suggests that nodU codes for an enzyme with 6-O-carbamoyltransferase-like activity. It is evident, however, that 3-O-(and/or 4-O-) carbamoyl transferases must also exist in NGR234. This is confirmed by the two singlets at ␦ ϭ 156.1 and ␦ ϭ 155.7 ppm in the 13 C spectra of nodU Ϫ mutants of NGR234 (data not shown).
Additional proof that nodU directs C6-O-carbamoylation was obtained by examination of the metastable ion spectra performed on m/z 469 and 483 of mono-carbamoylated B1 ions formed in the FAB spectrum of Nod factors produced by USDA257 containing nodU (USDA257(pRAF25)) as well as USDA257 and nodSU (USDA257(pA18)) transconjugants. The latter ion species bear an additional N-linked methyl group. ␤ elimination yielded ion fragments at m/z 451 and 465, respectively, which originate from the loss of 18 Da (ϭ H 2 O), showing that the carbamate cannot be at C-3 or C-4 (Fig. 5, D and E). Moreover, these sugar rings can be oxidized by periodate which is indicative of free OH groups at C-3 and C-4. 13 C NMR spectra of the major product of R. fredii(pA18) transconjugants gave a singlet at ␦ ϭ 156.5 (Fig. 4A) which suggests a single carbamoyl group. The resonance of carbon C-6, identified by DEPT (distortionless enhancement by polymerization transfer) analysis, showed a down-field shift of 3 ppm. This confirms that the carbamoyl group is positioned on C-6. This shift, relative to a free -CH 2 OH group was not observed in supernatants obtained from NGR⍀26 (which bear a carbamoyl group at C-3 or C-4). Moreover, carbamoylation at position C-3 affects the proton resonance of the neighboring N-CH 3 proton group (␦ ϭ 0.03 ppm). DISCUSSION We previously noted the striking nucleotide and amino acid sequence homologies between nodS of NGR234 and USDA257 (22,23). Similarly, the nodU gene of NGR234 (EMBL, Gen-Bank, and DDBJ accession number X89965) shows significant similarities to the amino acid sequeces of NodU of R. fredii USDA257 (23), B. japonicum USDA110 (24), A. caulinaudans (25), and ORF10 of Nocardia lactamdurans (31) (data not shown). ORF10 of Nocardia is implicated in the biosynthesis of the cephamycin family of carbamoylated ␤-lactam antibiotics. Nevertheless, the fact that such well conserved and apparently functional genes behave differently in Rhizobium sp. NGR234 and B. japonicum USDA110 suggest that the NodU protein also may act on other substrates such as lipopolysaccharides as has recently been reported (32). We are currently investigating this possibility.
Our physical/chemical data suggest that the nodS gene product is involved in N-linked methyltransferase activity, while that of the nodU gene is probably a 6-O-carbamoyltransferase. An impediment to elucidating the role of nodSU in NGR234 is the probable existence of other genes encoding similar functions. Mutations in nodS drastically reduce Nod factor production, but suppression of nodU has no effect on Nod factor levels (28). Thus, definitive analysis of nodS could only be obtained by introducing it into USDA257. Although both nodSU exist in USDA257, a deletion in the promoter region drastically reduces transcription of the operon (23). In spite of the extraordinary similarities between the two genomes, USDA257 produces Nod factors that differ from those of NGR234 by the absence of acetyl and sulfated groups on the 2-O-methylfucose moiety and which lack both the N-linked methyl group as well as carbamoyl groups on the non-reducing sugar. In R. fredii harboring nodS of NGR234, tri-and tetrameric Nod factors are no longer produced, while N-methylation of the acyl chain of the pentameric species occurs. NodU causes partial 6-O-mono-carbamoylation of the non-reducing sugar but does not control the length of the oligomer. Together, nodSU yield pentamers that are N-methylated and mono-carbamoylated on C-6. Surprisingly, the activity of either gene is sufficient to confer nodulation of Leucaena species. Indeed, there is no apparent combination of N-linked methyl and carbamoyl groups with the number of saccharide repeating units which permits nodulation of Leucaena. This contrasts starkly with the requirement of sulfated Nod factors of R. meliloti for nodulation of Medicago sativa (16). Nevertheless, mutations in nodS prevent NGR234 from nodulating Leucaena sp. (22,23). It should be noted, however, that nodSU of both B. japonicum and R. fredii are unable to complement nodSU Ϫ mutants of other (brady)rhizobia, including NGR234 (24).