The Structures and Biological Activities of the Lipo-oligosaccharide Nodulation Signals Produced by Type I and I1 Strains of Bradyrhixobium japonicum*

lipo-oligosac- charide signal molecules that induce deformation of root hairs and meristematic activity on soybeans. 8 39.7 with respect to 2,2-dimethyl-2-silapentane-5-sul- fonate. The GARP (35) sequence was used for 13C decoupling during acquisition. One-dimensional ROESY experiments used the following pulse se- quence: selected 190"-selected 180°-t-acquired, where the selective pulses were calibrated DANTE (36) pulse trains. The exorcycle (37) phase cycle was applied to the selective 180" pulse. The selective 90" pulse was 8.9 ms, and the refocusing delay t was 5.6 ms. The continuous wave spin-lock pulse was 500 ms. Data were processed typically with a lorentzian-to-gaussian weight-ing function applied to t2 and a shifted squared sine bell function and zero filling applied to tl. Processing was performed with Felix software (Hare Research, Inc.).

J. S.), and NATO Grant GlOl88 (to H. P. S., G. S., and R. W. C.). The and Sciences grant (to H. P. S.), a Fulbright postdoctoral fellowship (to costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18  These lipo-oligosaccharides are N-fatty acylated chitin oligomers. The nod genes that determine host specificity dictate variations in the type of N-acyl substituent present on the terminal glucosamine and in substituents that are present on the reducing N-acetylglucosamine. The terminal N-acylglucosamine can also be acetylated at C-6. A single species of Rhizobium can produce several lipo-oligosaccharides. In the case of Rhizobium meliloti, the major lipo-oligosaccharide is a tetramer with hexadecadienoic acid (c16:2) as the N-acyl substituent and a sulfate group at C-6 of the reducing N-acetylglucosamine. This molecule has been designated as NodRm-Iv(c16:z,s) (8) after the nomenclature of Spaink et al. (6). The terminal N-acylglucosamine is frequently acetylated at C-6 and is designated as NodRm-IV(Ac,C16,2,S) (11). Minor amounts of lipotri-and tetrasaccharides containing hexadecanoic (C16:o), hexadecenoic (C16:l), or hexadecatrienoic (C16:.) acid have also been reported for R. meliloti (9). The unsaturated fatty acyl residue and the sulfate group are required for the specific interaction with Medicago (8). Rhizobium leguminosarum bv. viciae produces a lipo-pentasaccharide in which the terminal Nacylglucosamine contains octadecatetraenoic acid (ClSz4) as the acyl substituent and is acetylated at C-6 (6). There are no substitutions on the reducing N-acetylglucosamine. Both the Cp34 and the O-acetyl substituents are required for the specific interaction with the legume host (6). Enzymes that are involved in the synthesis or addition of these substituents are encoded by the host specificity genes nodEF (required for the synthesis of ClSz4) and nodL (required for O-acetylation) (6). Lipotetrasaccharides, rather than pentasaccharides, that contain vaccenic acid (C18:lA~d as the N-acyl substituent are also synthesized by R. leguminosarum bv. viciae (6).
The above lipo-oligosaccharides are all from Rhizobium species that have a symbiotic relationship with hosts that form indeterminate nodules. Hosts such as soybean and bean form determinate nodules. The differences between these two types of nodules have been described in a recent review (12). B. japonicum strains are symbionts of soybean, but can also have other hosts, such as siratro. Our previous report (7) showed that B. japonicum USDAllO produces one major lipo-oligosaccharide; however, as detected by thin-layer chromatography (14), B. japonicum strain USDA135 produces several lipo-oligosaccharides. The USDAllO lipo-oligosaccharide has a pentasaccharide backbone that contains octadecenoic acid (Cla,l) as the N-acyl substituent and a 2-O-methylfucosyl residue at C-6 of the reducing N-acetylglucosamine and is designated as NodBj-V(C18:1,MeFuc) (7). The structures of the USDA135 Nod metabolites are described in this report. Another report describes the structures of a family of lipo-oligosaccharides from the broad host range (which includes soybean) Rhizobium sp. NGR234 (13). This strain produces lipo-oligosaccharides having pentasaccharide backbones with C 1 s : l A l l as the N-acyl substituent and a 2-0-methylfucose at C-6 of the reducing N-acetylglucosamine. The 2-0-methylfucosyl residue can also be sulfated or acetylated. In addition, the N-acylglucosamine is also N-methylated and contains either none, one, or two carbamoyl groups at C-3, C-4, and/or C-6. A similar N-methylated monocarbamoylated Nod metabolite has also been isolated from Azorhizobium caulinodans; however, this metabolite contains D-arabinose, rather than 2-O-methylfucose, linked to C-6 of the reducing N-acetylglucosamine (45).
The B. juponicum species is divided into two major groups with both B. juponicum USDAllO and USDA135 members of the Type I strains. Type I1 strains are quite different from Type I strains with regard to their DNA homology and the type of extracellular polysaccharide produced and also in that they belong t o the cowpea miscellany and therefore may have a broader host range than Type I strains (15). To understand the differences in the ability of l j p e I and I1 strains to nodulate different hosts, it is necessary to determine the structural differences between the lipo-oligosaccharides produced by these strains. In this report, we describe the structures and biological activities of several lipo-oligosaccharides from E . japonicum USDA135 (a Type I strain) and the structures of the lipo-oligosaccharides from strain USDA61 (a Type I1 strain).
EXPERIMENTAL PROCEDURES Bacterial Strains and Growth Conditions-B. japonicum strains were maintained on Rhizobium defined yeast extract (RDY) agar as described (16). The cells were grown in liquid RDY medium a t 30 "C until the cultures reached a n As,, of 0.5-0.6. The cells were then washed and diluted to anAsOo of 0.1 in minimal medium (17) containing glycerol as the carbon source and sodium glutamate as the nitrogen source. Seed extract (Glycine mar cv. Essex or Williams) or genistein (2 PM final concentration) (5) was added, and the bacteria were grown at 30 "C for a n additional 40 h. Strains used were the wild-type B. japonicum USDA135, USDA110, and USDA61 (18).
Detection of Lipo-oligosaccharides by Thin-layer Chromatography fI4)"Cells were grown in liquid RDY medium a t 30 "C until the cultures reached an AeO0 of 0.5-0.6. Bacteria were pelleted in a microcentrifuge, washed once with liquid minimal medium, and diluted in this medium to an A,,, of 0.1. Cells were then induced by the addition of 2 PM genistein or soybean seed extract. At the time of genistein or root exudate addition, 50 pCi of [I4C]acetate (56 mCi/mmol, 1 Ci = 37 GBq; ICN) was added, and the cultures were incubated overnight. The induction of the nodulation genes was indirectly monitored by the induction of @-galactosidase in a strain containing a nodY-lac2 fusion (i.e. ZB977) (19). Supernatants of labeled cultures were extracted with 1butanol and applied to octadecyl silica TLC plates (Sigma) as described (14). Plates were dried and exposed to x-ray film (Kodak X-Omat A R ) for 2-6 days at room temperature.
Purification of Lipo-oligosaccharides-The Nod metabolites were purified as described previously (7). The cells were pelleted, and the supernatants were extracted with 0.33 volume of distilled 1-butanol. The butanol layer was collected, and the butanol was removed by rotary evaporation. The residue was resuspended in acetonitri1e:water (1:l) and chromatographed using 60% acetonitri1e:water on a Silica Gel 60 column (1.6 x 100 cm; Pharmacia LKB Biotechnology Inc.). Fractions containing Nod metabolites were further analyzed and purified by HPLCl using a Pharmacia SuperPac Pep-S column (5 pm, 5 x 250 mm). The eluent from the HPLC column was monitored a t 206 nm.
Assay for Biological Activity of Nod Metabolites-Seeds of Glycine soja PI468397 were surface-sterilized and germinated as previously described (16). Cortical cell division activity was tested following the spot inoculation method originally described by Turgeon and Bauer (20). Two-day-old seedlings were placed in plastic pouches containing 5 The abbreviations used are: HPLC, high pressure liquid chromatography; GC-MS, gas chromatography-mass spectrometry; FAB-MS, fast atom bombardment mass spectroscopy; TG, thioglycerol; TOCSY, total correlation spectroscopy; ROESY, rotating frame nuclear Overhauser effect spectroscopy; HSQC, heteronuclear single quantum coherence; Cb, carbamoyl; Gro, glycerol; FID, free induction decay. ml of plant nutrient solution (16) and allowed to grow overnight in the dark. At the time of inoculation, the position of the smallest emergent root hairs, visible in a dissecting microscope at magnification x 50, and the root tip were marked on the top face of the plastic pouch. The top face of the plastic pouch was slit with a razor blade and rolled back to expose the root. Prior to inoculation, a single Amberlite bead was transferred with forceps to a position above the root tip -80% of the distance between the root tip and the smallest emergent root hairs. Droplets containing different amounts of purified Nod metabolites in a volume of 30-50 nl were delivered by micropipette to the same position as the Amberlite bead. The droplets were allowed to dry on the root surface for 10-15 min, and the pouches were taped closed. To avoid undesirable binding of Nod metabolites to the plastic, a sterile straw was placed next to the root to hold the plastic at a distance for the first 2 h after inoculation. Plants were then transferred to a plant growth room with a 16-h lighU8-h dark photo period. Roots were analyzed for cortical cell division and nodule formation by following the clearing method described by Truchet et al. (21).
Hair deformation activity was determined as previously described using Vicia sativa subsp. nigra (6) or G. soja (7) as test plants.
Chemical Analysis of Lipo-oligosaccharides-Glycosyl composition analysis was performed by GC-MS analysis of alditol acetates prepared as described in York et al. (22). Glycosyl linkage analysis was performed by GC-MS analysis of partially methylated alditol acetates prepared by the procedure of Hakomori modified as described by York et al. (22). Analysis was performed using a 30-m SP2330 capillary column (Supelco, Inc.). Fatty acids were identified by GC-MS analysis of their methyl esters prepared by acid-catalyzed methanolysis (23) in methanolic 4 M HCl at 80 "C for 14 h. Fatty acids were also isolated by alkaline hydrolysis (1.7 M NaOH) in dimethyl sulfoxide at 80 "C for 14 h, followed by acidification and extraction into chloroform. Methyl esters of the fatty acids released by alkaline hydrolysis were prepared by methanolysis in methanolic 1 M HCl at 80 "C for 1 h. Analysis was performed using a 30-m capillary DBl column (J & W Scientific). The fatty acyl residue that was attached to glucosamine was determined by mild methanolysis in dry methanolic 1 M HC1 a t 80 "C for 1 h, followed by trimethylsilylation and analysis by GC-MS using a 15-m DB1 column (8,11,25). The location of the double bond in the C,,:, fatty acyl residue was determined by the preparation and analysis of dimethyl disulfide ethers of the C,,:, methyl ester (26). The resulting products were identified by GC-MS using a DB1 column. FAB-MSAnalysis-FAB-MS was carried out on a VG-ZAB SE instrument at an accelerating voltage of 8 kV in the positive mode with thioglycerol (TG) or glycero1:m-nitrobenzyl alcohol (1:l) as the matrix. The samples were dissolved in dimethyl sulfoxide, and -2-10 pg in 1 pl was applied to the probe. Tandem MS-MS analysis was performed using a JEOL HXllO/HXllO mass spectrometer operated a t 10-kV accelerating potential. Spectra acquired by MS-1 are averaged profile data as recorded by a JEOL complement data system. These spectra were acquired from m / z 0 to 3000 at a rate that would scan from m / z 1 to 6000 in 1 min. A filtering rate of 300 Hz and an approximate resolution of 1500 were used in acquiring these spectra. Ions were produced by liquid secondary ion mass spectrometry. Collisionally induced dissociation was performed in the third field free region using helium as the collision gas. The helium pressure was sufficient to attenuate the primary ion beam by 75%, and the collision cell was floated at 3 kV. The samples were dissolved in dimethyl sulfoxide as described above. Glycero1:mnitrobenzyl alcohol (1:l) was used as the matrix for the tandem MS-MS analyses.
NMR Analysis-Prior to analysis, the samples w' ere suspended in 2H20 and lyophilized. This process was repeated three times. All spectra were recorded on a Bruker AMX 600 MHz spectrometer using deuterated dimethyl sulfoxide as the solvent. Two-dimensional double quantum filtered COSY (27), TOCSY (28, 29), ROESY (30), and HSQC (31) data sets were collected in phase-sensitive mode using the time proportioned phase incrementation (32) method. In all experiments, low-power presaturation was applied to the residual HDO signal.
For the homonuclear experiments, typically 512 FIDs of 2048 complex data points were collected, with 64 scans/FID for the TOCSY data and 128 scans for the double quantum filtered COSY and ROESY data. The spectral width was set to 5000 Hz, and the carrier placed at the residual HDO peak. The TOCSY pulse program contained a 130-ms MLEV17 (33) spin-lock pulse, and the ROESY experiment used a 200-ms continuous wave spin-lock pulse flanked by two 90" pulses for offset compensation (34).
For the HSQC spectrum, 256 FIDs of 4096 complex points were acquired, with 256 scans/FID. The spectral width in the carbon dimension was set to 120 ppm, with the carrier a t 6 70 referenced to dimethyl sulfoxide a t 8 39.7 with respect to 2,2-dimethyl-2-silapentane-5-sulfonate. The GARP (35) sequence was used for 13C decoupling during acquisition.
One-dimensional ROESY experiments used the following pulse sequence: selected 190"-selected 180°-t-acquired, where the selective pulses were calibrated DANTE (36) pulse trains. The exorcycle (37) phase cycle was applied to the selective 180" pulse. The selective 90" pulse was 8.9 ms, and the refocusing delay t was 5.6 ms. The continuous wave spin-lock pulse was 500 ms.
Data were processed typically with a lorentzian-to-gaussian weighting function applied to t2 and a shifted squared sine bell function and zero filling applied to t l . Processing was performed with Felix software (Hare Research, Inc.).

RESULTS
Purification of B. japonicum Nod Metabolites- Fig. 1 shows a thin-layer chromatogram of the Nod metabolites from strains USDA110, USDA135, and USDA61. Strain USDAllO produced one major Nod metabolite, with trace amounts of several others, while strains USDA135 and USDA61 both produced several Nod metabolites. The Nod metabolites from the various strains were purified by HPLC. Fig. 2 shows the HPLC profile of the Nod metabolites from strain USDA135. Four fractions (Fl-F4) were isolated, with fractions F3 and F4 present in the largest amounts. The identification of the various HPLC fractions was determined by TLC analysis. Because strain US-DA135 produced larger amounts of fractions F3 and F4, they were characterized in the greatest detail. These results are described below.
Composition and Glycosyl Linkage Analysis-The glycosyl compositions of fractions F3 and F4 were determined by the preparation and GC-MS analysis of alditol acetates and trimethylsilyl methylglycosides. Both fractions F3 and F4 had a 1:5 ratio of 2-0-methylfucose to N-acetylglucosamine.
Methylation analysis of both fractions F3 and F4 gave a 1:3:1 ratio of terminal to 4-linked to 4,g-linked N-acetylglucosamines. Lower amounts of terminal 2-0-methylfucose were also detected. The lower value for the partially methylated alditol acetate of terminally linked 2-0-methylfucose was probably a result of some loss due to the volatility of its partially methylated alditol acetate.
The fatty acid components of both fractions F3 and F4 showed the presence of C16:o, octadecanoic (C18:.), and Cls:l fatty acids. Since small amounts of Cleo, ClR0, and CIB:. could be due to slight contamination by membrane phospholipids, it was necessary to identify those fatty acid residues that are part of the Nod factor preparations. Therefore, the fatty acyl components of fractions F3 and F4 were determined by mild methanolysis, preparation of trimethylsilyl ethers, and GC-MS analysis. This method was used since mild methanolysis readily liberates the methylglycoside of N-acylglucosamine and thus permits the identification of the fatty acyl moiety that is still attached to the glucosamine (8,11,25). Using this procedure, it . . . . e .
-+ -+ -+ was found that fraction F3 contained both N-hexadecanoylglucosamine and N-octadecenoylglucosamine, while fraction F4 contained only N-octadecenoylglucosamine. The electron impact and chemical ionization spectra for the trimethylsilyl methylglycosides of these components are shown in Fig. 3. The (M + H)' ions were at rnlz 674 and 648 for the trimethylsilyl methylglycosides of N-octadecenoylglucosamine and N-hexadecanoylglucosamine, respectively (Fig. 3A). A fragment ion at m l z 204 (the fragment containing C-3 and C-4) was found for both N-acylglucosaminosyl residues, and the characteristic C(2)-C(3) fragment ions at mlz 395 and 369 were observed for N-octadecenoylglucosamine and N-hexadecanoylglucosamine, respectively (Fig. 3B). Other fragment ions were consistent with those reported for the trimethylsilyl derivatives of N-acylglucosamine methylglycosides (38). The presence of both Nhexadecanoylglucosamine and N-octadecenoylglucosamine in fraction F3 indicated that this fraction contained a mixture of at least two molecules, one with an N-hexadecanoyl substituent and the other with an N-octadecenoyl substituent.
Our previous paper (7) reported the location of the double bond in the Clel fatty acyl substituent of NodBj-V(CIB,l,MeFuc) from strain USDAllO to be between carbons 9 and 10, i.e. oleic acid, while all other studies on C1,,,-containing Nod metabolites reported the presence of CIR:lAll. Therefore, the location of the double bond in the CIR1 present in fraction F4 was examined using methods that greatly increased the Cis:. recovered from the Nod metabolite. Saponification with 1.7 M NaOH in dimethyl sulfoxide at 80 "C for 14 h increased by 5-10 fold, over that previously reported (7), the amount of C I R :~ liberated from the Nod metabolite. However, when the temperature was increased to 100 "C, the recovery of CIR1 was greatly 4. reduced. The release of C1s:l from the Nod metabolites was also increased by performing methanolysis in methanolic 4 M (rather than 1 M) HCl at 80 "C for 14 h. The location of the double bond in the fatty acid released by these methods was determined by preparing the dimethyl disulfide derivatives of the fatty acid methyl esters (26). Analysis of this derivative by GC-MS (data not shown) gave mass fragments at m l z 145, 245, and 213. The ions at r n l z 145 and 245 result from fragmentation between the carbons that carry the dimethyl disulfide groups and show that the double bond was between carbons 11 and 12. The fragment at r n l z 213 is due to the loss of methanol (-32 atomic mass units) from the ion at r n l z 245. Thus, the octadecenoyl component in this Nod metabolite is vaccenic acid. The fatty acyl component of the Nod metabolite from USDA110, previously identified as oleic acid (7), will be re-examined using the above methods to ensure greater release of the fatty acyl component from that Nod metabolite.
FAB-MS of Fractions F3 and F P T h e FAB-MS spectra for fractions F3 and F4 are shown in Fig. 4 ( A and B , respectively).
The (M + H)' ions observed for fraction F4 were at r n l z 1416 and 1458, with the ion at rn I z 1458 being of greatest intensity. The ion at r n l z 1416 is due to the presence of a small amount of non-0-acetylated (i.e. -42 atomic mass units) metabolite. It is likely that the presence of the non-0-acetylated metabolite in this fraction is due to the loss of this labile substituent during sample preparation. It should be noted that fraction F3, which contains the largest amount of this same non-0-acetylated molecule, is well separated from fraction F4 during HPLC purification (see Fig. 2). A TG adduct was observed for molecules carrying an unsaturated fatty acyl residue. Hence, the (M + H+ TG)' ion, r n l z 1566 (+108), is due to the TG adduct of fraction F4. Fragment ions at r n l z 468 (present but not shown in Fig. 4), 671, 874, and 1077 and their TG adducts were also observed. The structure shown in Fig. 4 is consistent with this fragmentation pattern and with the chemical data described above. The ion at r n l z 468 shows that the 0-acetyl and N-octadecenoyl groups are present on the terminal glucosamine. The difference of 203 atomic mass units between fragment ions is consistent with a sequence of 3 additional N-acetylglucosaminosyl residues. The mass difference, 381 atomic mass units, between the The only branching glycosyl residue found during methylation analysis was a 4,6-linked N-acetylglucosamine (see above). Thus, it is likely that the terminal 2-0-methylfu-cosy1 residue is linked to C-6 of the reducing N-acetylglucosamine. Confirmation of this linkage was obtained by twodimensional NMR analysis and is discussed below. These data are consistent with fraction F4 being NodBj-V(Ac,Cls,l,MeFuc).
The FAB-MS spectrum of fraction F3 (Fig. 4.4) shows that it fraction F3 is probably due to the loss of the labile 0-acetyl group during sample preparation since fraction F2, which contains only Nod-Bj-V(C16,0,MeFuc) (discussed below), was separated with base-line resolution from fraction F3 (see Fig. 2).
The structures shown in Fig. 4 were confirmed by FAB-MS analysis of peracetylated or prereduced (with NaB2H4) and peracetylated fraction F3. The results are shown in Fig. 5 (A  and B ) . Peracetylation without prereduction was done in dimethyl sulfoxide using N-methylimidazole as the catalyst (39). The FAB-MS spectrum of the peracetylated products of fraction F3 is shown in Fig. 5A. The fragment ions are consistent with the presence of a mixture of two peracetylated Nod metabolites, one containing an N-octadecenoyl substituent and the other an N-hexadecanoyl substituent. Also notice that both peracetylation products were present as N-methylimidazolium glycosides. Reduction (with NaB2H4) prior to peracetylation also resulted in a mixture of two reduced peracetylated Nod metabolites containing N-octadecenoyl and N-hexadecanoyl substituents (Fig. 5B) ions. m-Nitrobenzyl alcohol was used as the matrix for the spectrum shown in Fig. 5B, and the TG adducts are noticeably absent.
FAB-MS analysis was also performed on fractions F1 and F2. Not enough of these fractions was obtained for a complete chemical or NMR analysis. Fraction F2 gave a spectrum (data not shown) that is consistent with this component being NodBj- 1518. The presence of TG adducts indicates that this molecule contains an unsaturated fatty acyl substituent. Since the TG adducts often give more intense ions, the TG adduct of the fragment ion at rnlz 1007, rnlz 1115, was also observed. As with the other Nod metabolites, the difference between r n l z 1496 and 1115 is 381 atomic mass units and is consistent with the reducing end of this molecule containing a 2-0-methylfucosyl-N-acetylglucosamine disaccharide component. An unsaturated fatty acyl substituent (which would give rise to TG adducts) that is consistent with the molecular size of this molecule is a hexadecenoyl substituent. Thus, it is proposed that this Nod metabolite is NodBj-V(C16:.,MeFuc). Fatty acid analysis of fraction F1 (data not shown) also showed the presence of hexadecenoic acid; however, not enough material was available to determine the location of the double bond.
NMR Analysis-The proton NMR spectrum of NodBj-V(Ac,C18,1,MeFuc) is shown in Fig. 6. The resonance at 6 4.93 (Jl,2 = 2.4 Hz) is consistent with that reported for the anomeric proton of the reducing a-N-acetylglucosamine of other Nod metabolites (6,8,11). The resonances between 6 4.34 and 4.50 (J1,2 = 9 Hz) are due to the anomeric protons of the 0-linked N-acyland N-acetylglucosaminosyl residues as reported for other Nod metabolites (6,8,11). The resonance at 6 5.00 (Jl,2 = 3.7 Hz) is consistent with an a-linked 2-0-methylfucosyl residue and is identical to that reported for the Nod metabolite from B. juponi- ~~I cum USDAllO (7). Because of the reducing N-acetylglucosamine, this Nod metabolite exists as a mixture of d panomers; therefore, a second minor doublet at 6 4.98 is due to the 2-0-methylfucosyl residue attached to the reducing p-Nacetylglucosaminosyl residue of the anomeric mixture. The singlet at 6 3.46 is due to the methoxy protons of the 2-0-methylfucosyl residue, and the singlet of lower intensity at 6 3.44 is a second methoxy proton resonance due, again, to the anomeric effect of the reducing N-acetylglucosaminosyl residue. The resonance at 6 1.13 is due to the H-6 methyl protons of the 2-0methylfucosyl residue.
The linkage of the 2-0-methylfucosyl residue to C-6 of an N-acetylglucosaminosyl residue (known to be the reducing residue from the FAB-MS data described above) was confirmed by NMR analysis. An HSQC spectrum showed a set of cross-peaks (labeled G6 in Fig. 7 0 ) at 6 68.716 3.65 and 3.82 corresponding to C-6/H-6 cross-peaks of an N-acetylglucosaminosyl residue and shifted downfield from the other C-6 atoms at 6 60-61. This 7-8-ppm downfield shift for C-6 is consistent with a glycosyl linkage at that position. Both one-and two-dimensional TOC-SYs (data not shown) show that the two H-6 atoms (6 3.65 and 3.82) that are coupled to this downfield C-6 belong to the reducing N-acetylglucosaminosyl residue. A selective one-dimensional ROESY experiment (Fig. 7C), in which the 2-0-methylfucosyl H -l ( 6 5.00) was irradiated, enhanced the upfield signal of one of these H-6 atoms (6 3.65). Fig. 7 (A and B ) shows that this H-6 is present in both NodBj-V(Ac,Cls:l,MeFuc) and Nod-Bj-V(C1s:l,MeFuc), respectively. These data, together with the FAB-MS and methylation data described above, show that the 2-0-methylfucosyl residue is linked to C-6 of the reducing-end N-acetylglucosaminosyl residue.
The location of the 0-acetyl group was also deduced from NMR analysis. The proton spectrum (Fig. 6) of NodBj-V(Ac,Cls,l,MeFuc) shows a sharp singlet at 6 2.1 that is due to the 0-acetylmethyl protons. The proton signals at 6 4.11 (J5,6 = 7.7 Hz, J6,6 = 12.6 Hz) and 6 4.35 (J5,6 < 1 Hz) are due to the terminal N-acylglucosamine H-6 atoms since they were shown by two-dimensional TOCSY (spectrum not shown) to be connected to the unique H-4 resonance (6 3.18) of the terminal unsubstituted C-4 of this residue. The downfield position of these H-6 atoms is characteristic of an 0-acetyl substitution. Therefore, these data, together with the FAB-MS data, show that the 0-acetyl group of NodBj-V(Ac,Cls:l,MeFuc) is at C-6 of the terminal N-acylglucosaminosyl residue.
The resonance at 6 5.40 (see Fig. 6) is due to the vinyl protons of the Cls,l fatty acyl component. The small H(9)-H(10) cou- pling constant indicates a cis-configuration. Resonances typical for the methylene and methyl protons of the fatty acyl group are present as indicated in Fig. 6.
Analysis of Nod Metabolites from Strain USDA6l"Small amounts of these metabolites were purified as described above. atomic mass units greater than that at rnlz 1213. This 43atomic mass unit increase is consistent with an added carbamoyl (Cb) group, as has been reported for the NGR234 and A. caulinodans Nod metabolites (13,45). The fragment ion at m l z 469 would dictate that the carbamoyl group is located on the terminal N-octadecenoylglucosamine, indicating that this molecule is NodBj-IV(Cb,Cls,l,MeFuc). it has an added methyl group. Based on the previous reports for the NGR234 and A. caulinodans Nod metabolites (13,45), it was likely that this methyl group was present as an N-methyl group on the terminal N-acylglucosamine. This was confirmed by methanolysis of fraction F3 in methanolic 1 M HCl a t 80 "C, followed by hydrolysis in 4 M HCl a t 100 "C for 18 h. The glycosyl residues were reduced and acetylated as described (22). Analysis by GC-MS showed the presence of alditol acetates of both N-methylglucosamine and glucosamine. The mass spectrum (data not shown) of the alditol acetate of this N-methylglucosamine (with a 2H atom a t C-1 due to reduction with NaB2H4) shows the characteristic primary fragments a t rnlz 374 and 159. In addition, small amounts of glycerol, fucose, and 2-0-methylfucose were detected, even though the strong hy- The (M + H)' molecule at m l z 1458 has a fragmentation pattern identical to that of NodBj-V(Ac,Cla,.,MeFuc) found in fraction F2 (described above) and in strain USDA135 (also described above). It is likely that this is residual fraction F2 material that was not completely separated from fraction F3.
The (M + H)' molecule a t m l z 1473 has a fragmentation pattern (shown in Fig. 8B) that is consistent with that described above for fraction F1 ( m l z 14161, but with carbamoyl (+43 atomic mass units) and methyl (+14 atomic mass units) groups (one each) added to the terminal N-octadecenoylglu- cosamine, resulting in a fragment ion at rnlz 483. As described above, the only methylated glucosamine found in fraction F3 was N-methylglucosamine, indicating that this molecule has an N-methyl group. The location of the carbamoyl group could be at C-3, C-4, or C-6. These data indicate that this molecule is NodBj-V(Cb,C18:.,NMe,MeFuc).

The (M + H)' molecule a t m l z 1501 is found in fraction F4.
Its molecular size is 43 atomic mass units units larger than NodBj-V(Ac,Cla,.,MeFuc) ((M + HI' 1458), indicating that it has an added carbamoyl group. The fragment ion at rnlz 511 dictates that this added carbamoyl group is located on the terminal N-octadecenoyl-0-acetylglucosamine. If the location of the 0-acetyl group is at C-6 as described above for the US-DA135 metabolite, then the carbamoyl group would be located a t C-3 or C-4. These data indicate that this molecule is NodBj-V(Ac,Cb,Cls,.,MeFuc).
Not enough of fractions F2 and F3 were obtained to perform methylation or NMR analysis; therefore, the linkages and anomeric configurations could not be determined. However, based on the structures for the USDA135 Nod metabolites described above and those previously reported (3,(6)(7)(8)(9)13), it is likely that the N-acetylglucosaminosyl residues are P-linked and that the fucosyl and 2-0-methylfucosyl residues are a-linked to C-6 of the reducing-end N-acetylglucosamine. The location of the carbamoyl group on the terminal N-acylglucosamines of these Nod metabolites is not known. Larger amounts of these various metabolites are being purified to confirm these structures and to determine their biological activities.
Biological Activity of B. japonicum Nod Metabolites -Previous investigations have shown that V: sativa is a useful test plant with respect to its reaction to Nod metabolites in that its root hairs are readily deformed (Had activity) by a broad variety of Nod metabolites (6,8). Fig. 9 shows that fractions Fl-F4 from USDA135 have Had activity on V: sativa subsp. nigra at nanomolar concentrations. Thus, Had activity on V: satiua occurs with these Nod metabolites in which the N-acyl substituent can be an N-hexadecenoyl, N-hexadecanoyl, or Noctadecenoyl substituent. In addition, Nod metabolites with or without the 0-acetyl group were active, indicating that there is not an absolute requirement for the 0-acetyl group for this activity.
A previous paper has shown that NodBj-V(Cle,.,MeFuc) from strain USDAllO has Had activity on G. soja and siratro at 100 PM and no activity on alfalfa even when present at a 10,000-fold higher concentration (7). Under the experimental conditions used, where the entire root was exposed to the Nod metabolite, nodule formation or cortical cell division was not detected on either G. soja or siratro. However, using the spot inoculation procedure (see "Experimental Procedures"), outgrowths on the roots of G. soja were observed (Fig. 1OA). These structures appeared at the point of inoculation with either NodBj-V(Cle,l,MeFuc) or NodBj-V(Ac,Cla:.,MeFuc). When 1.5 ng of either compound was applied to the roots, 3 of 12 plants showed one or more of these structures. This ratio increased to 9 of 12 plants when 15 ng was applied. Control plants inoculated with solvent alone did not show any of these structures. The swellings do not show the typical nodule anatomy as they do not contain an internal vascular tissue (data not shown). However, they do not appear to be lateral roots since the methylene blue-stained meristem is not cone-shaped, and it does not originate in the inner cortex as in real lateral roots (Fig. 1OB). Closer examination of these Nod metabolite-induced swellings showed them to be very disorganized, with mitotically active cells dispersed near the epidermis. This is similar to soybean nodule development reported by Calvert  cell death. In some respects, these structures resembled the popcorn pseudonodules elicited by certain B. japonicum mutants that are defective in their lipopolysaccharides (41). It is possible that induction of normal nodule structures requires the presence of additional signals, besides the Nod metabolites, from the bacterium.

DISCUSSION
In this report, we have shown that both Type I and I1 B.
japonicum strains produce a variety of Nod metabolites. The structures of these Nod metabolites are summarized in Fig. 11. The types of Nod metabolites produced appear to be straindependent. Strain USDAllO produces one major metabolite, NodBj-V(Cls:.,MeFuc) (7). This strain can also produce lesser amounts of NodBj-V(Ac,Cle,.,MeFuc (data not shown). In addition to the USDAllO Nod metabolites, strain USDA135 pro- stimulate cell division on G. soja. Due to insufficient amounts, the other Nod metabolites from USDA135 have not been tested for their ability to induce cortical cell division. The Type I1 strain USDA61 produces eight Nod metabolites in addition to the NodBj-V(Clg,.,MeFuc) and NodBj-V(Ac,Cle,l,MeFuc) molecules, which are also produced by the Type I strains. Two of these additional Nod metabolites have chitin pentasaccharide backbones, while the other six have tetrasaccharide backbones. All have an octadecenoyl group as the N-acyl substituent. A number of these metabolites have carbamoyl andor N-methyl substituents located on the terminal N-octadecenoylglucosamine. In this respect, they are similar to those molecules reported for NGR234 and A. cauZinodans (13,45). Four of these metabolites are unique in that the reducingend N-acetylglucosamine contains a branching fucose and is glycosidically linked to glycerol. These four metabolites have tetrasaccharide backbones. This is the first report of Nod metabolites in which the reducing-end N-acetylglucosamine does not exist as a free reducing sugar. It is possible that these molecules represent end products with unique biological properties or intermediates in the biosynthesis of these B. japonicum Nod metabolites.
, , " " , Recently, the structures of several Nod metabolites from the broad host range Rhizobium sp. NGR234 have been reported (13). All of these metabolites contain 2-0-methylfucose at C-6 of the reducing N-acetylglucosamine, while varying in the fatty acyl substituent, i.e. either N-hexadecanoyl or N-octadecenoyl. Other NGR234 Nod metabolites contain carbamoyl groups at C-3, C-4, and/or C-6 of the N-methyl-N-acylglucosamine as well as sulfate or acetate a t C-3 or C-4, respectively, of the 2-0methylfucosyl residue. Since one of the hosts of NGR234 is soybean, these results would indicate that the 2-0-methylfucosyl residue is required for nodulation of soybean. However, another possibility is that the 2-0-methylfucosyl residue is important in extending the host range. Both B. japonicum and NGR234 have broad host ranges in comparison to R. leguminosarum or R. meliloti.
That some substituents, such as 2-O-methylfucose, carbamoyl, and N-methyl groups, may be involved in extending the host range has some support in the literature. Recently, it was reported that nods has homology to methyltransferases that use S-adenosylmethionine as the methyl donor (45). It was suggested that this gene is responsible for the N-methylation of the N-acylglucosamine of the Nod metabolites (45). When this gene was transferred into Rhizobium fredii USDA257, its host range was extended t o include Leucuenu, not normally a host of USDA257 (46). Other reports have shown that mutations in B. juponicum genes nod2 and nodVW result in restriction of the host range, i.e. the symbionts no longer nodulate siratro, but still nodulate soybean (47,48). Examination of the Nod metabolites from a nod.2-mutant has shown that they do not contain the 2-0-methylfucosyl residue.' Thus, it is possible that certain structural modifications of the Nod metabolites are required for infecting a broad range of hosts.
Further work on the structures and biological activities of Nod metabolites from various B. japonicum Type I (16,42,43) and I1 (18,44) mutants is in progress to determine structurefunction relationships of the various host specificity genes.