Structures of Exopolysaccharides Involved in Receptor-mediated Perception of Mesorhizobium loti by Lotus japonicus*

In the symbiosis formed between Mesorhizobium loti strain R7A and Lotus japonicus Gifu, rhizobial exopolysaccharide (EPS) plays an important role in infection thread formation. Mutants of strain R7A affected in early exopolysaccharide biosynthetic steps form nitrogen-fixing nodules on L. japonicus Gifu after a delay, whereas mutants affected in mid or late biosynthetic steps induce uninfected nodule primordia. Recently, it was shown that a plant receptor-like kinase, EPR3, binds low molecular mass exopolysaccharide from strain R7A to regulate bacterial passage through the plant's epidermal cell layer (Kawaharada, Y., Kelly, S., Nielsen, M. W., Hjuler, C. T., Gysel, K., Muszyński, A., Carlson, R. W., Thygesen, M. B., Sandal, N., Asmussen, M. H., Vinther, M., Andersen, S. U., Krusell, L., Thirup, S., Jensen, K. J., et al. (2015) Nature 523, 308–312). In this work, we define the structure of both high and low molecular mass exopolysaccharide from R7A. The low molecular mass exopolysaccharide produced by R7A is a monomer unit of the acetylated octasaccharide with the structure (2,3/3-OAc)β-d-RibfA-(1→4)-α-d-GlcpA-(1→4)-β-d-Glcp-(1→6)-(3OAc)β-d-Glcp-(1→6)-*[(2OAc)β-d-Glcp-(1→4)-(2/3OAc)β-d-Glcp-(1→4)-β-d-Glcp-(1→3)-β-d-Galp]. We propose it is a biosynthetic constituent of high molecular mass EPS polymer. Every new repeating unit is attached via its reducing-end β-d-Galp to C-4 of the fourth glucose (asterisked above) of the octasaccharide, forming a branch. The O-acetylation occurs on the four glycosyl residues in a non-stoichiometric ratio, and each octasaccharide subunit is on average substituted with three O-acetyl groups. The availability of these structures will facilitate studies of EPR3 receptor binding of symbiotically compatible and incompatible EPS and the positive or negative consequences on infection by the M. loti exo mutants synthesizing such EPS variants.

Rhizobia are soil bacteria that are able to form nitrogen-fixing endosymbiotic relationships with their host legume plants.
In this mutualistic interaction, the host plant secretes flavonoids that stimulate compatible rhizobia to synthesize specific lipochitooligosaccharide signaling molecules called Nod factors (1,2). The host plant perceives these Nod factors using membrane-bound receptor-like kinases that contain extracellular LysM domains (3)(4)(5). Perception of the Nod factors results in deformation and curling of root hairs and also the induction of cortical cell division to form nodule primordia (6). The bacteria entrapped in the curled root hairs form microcolonies and then enter the host through infection threads that are formed by localized digestion of the plant cell wall within the curl, followed by inverted growth of the cell wall/plasma membrane down the length of the root hair (7,8). The rhizobia grow down the infection thread by cell division, and the infection thread ramifies as it reaches the growing nodule primordia. Eventually the rhizobia are released from the infection thread into the plant cell cytoplasm surrounded by a membrane of plant origin and develop into the nitrogenfixing bacteroid form. The nodules formed may be either determinate (e.g. Lotus spp., Phaseolus spp., and Glycine max) or indeterminate (e.g. Medicago spp., Vicia spp., and Trifolium spp.) depending on the host (9). Determinate nodules have a short-lived meristem, and the nodules grow by plant cell expansion and division, resulting in nodules progressing through well defined developmental stages. In contrast, indeterminate nodules have a persistent meristem, and infection is continuous, resulting in all developmental stages being present within a single nodule.
Rhizobia produce a variety of cell surface-associated glycolipids and polysaccharides, as well as extracellular polysaccharides that also contribute to a successful symbiotic plant-microbe interaction (10 -15). Exopolysaccharides (EPS) 4 in particular are required for the formation and development of infection threads in indeterminate symbioses, of which the best characterized are those that occur between Sinorhizobium meliloti and Medicago species and between Rhizobium leguminosarum and Vicia species. Rhizobial EPS are usually branched or linear high molecular mass (HMM) heteropolymers or low molecular mass (LMM) oligomers composed of repeating units of 2-9 neutral or acidic sugar residues such as D-glucose, D-galactose, D-mannose, L-rhamnose, and D-glucuronic and D-galacturonic acids. Glycosyl residues are either ␤or ␣-linked and may be substituted with O-succinyl, pyruvyl, O-acetyl, or O-methyl groups. The occurrence of these substitutions in the polymer is often regulated by environmental factors and may be either stoichiometric or non-stoichiometric. The uronic acids and some of the non-carbohydrate substituents provide acidic character and net negative charge to the EPS. The most extensively studied exopolysaccharide from S. meliloti strain 1021, succinoglycan, or EPS I, is a polymer consisting of an octasaccharide repeat unit composed of seven D-glucose and one D-galactose residues substituted with 6-O-succinyl, 6-O-acetyl, and 4,6-pyruvyl groups (Fig. 1). It is produced in both HMM and LMM (monomer, dimer, and trimer repeats of the octasaccharide) forms (16 -20). Many S. meliloti strains but not strain 1021 also produce a second EPS, EPS II or galactoglucan, that consists of partially 6-O-acetylated D-glucose and 4,6-pyruvylated D-galactose disaccharide [␤-D-Galp-(133)-␣-D-Glcp- (13] n (21). Galactoglucan is able to replace the symbiotic requirement for succinoglycan in those strains that make it. In R. leguminosarum, the EPS polymer consists of a pyruvylated and O-acetylated octasaccharide repeat unit with a terminal non-reducing galactose residue ( Fig. 1) (22)(23)(24). The role of EPS in symbiosis has also been extensively studied in the broad host range strain Sinorhizobium fredii NGR234, which produces a nonasaccharide EPS (Fig. 1) (25). The LMM forms of NGR234 EPS consist of the nonasaccharide monomer and an octasaccharide processed from it (26).
EPS may contribute to the establishment of the rhizobiumlegume symbiosis in multiple ways, including physical roles such as mediating attachment, preventing agglutination, contributing to control of calcium ion fluxes, or providing protection against oxidative stress (27)(28)(29)(30)(31). However, studies of S. meliloti 1021 and S. fredii NGR234 have implicated EPS as playing a signaling role (20,26,28,(32)(33)(34). These studies suggested that succinoglycan trimers isolated from S. meliloti and the monomer oligosaccharides from Rhizobium sp. NGR234 were the symbiotically active species, as these species corrected the symbiotic defects of the EPS mutants when added exogenously (20,26,34). However, no plant receptors for the molecules were identified, and recent work has suggested that the S. meliloti LMM species are not required for symbiosis (35).
The genetics and biochemistry of succinoglycan biosynthesis in S. meliloti 1021 have been extensively characterized, and most biosynthetic genes except one required to add the terminal glucose to the branch have been identified (13,36). Mutations that abolish synthesis and some that modify succinoglycan structure cause symbiotic defects. For example, mutants in exoY, which initiates the first step in succinoglycan biosynthesis, still develop infection foci in curled root hairs but rarely initiate formation of infection threads. HMM succinoglycan is cleaved by two glycanases ExsH and ExoK (37), to produce a pool of monomers, dimers, and trimers of the octasaccharide repeat unit.
Recently, rhizobial EPS has also been shown to play a role in infection thread formation in determinate symbioses. Mesorhizobium loti strain R7A mutants affected in different stages of the EPS biosynthesis pathway exhibit different symbiotic phe- 4 The abbreviations used are: EPS, exopolysaccharide; COSY, homonuclear correlation spectroscopy; HMBC, heteronuclear multiple-bond correlation spectroscopy; HMM, high molecular mass (high molecular weight); HSQC, heteronuclear single-quantum correlation spectroscopy; LMM, low molecular mass (low molecular weight); PMAA, partially methylated alditol acetate; ROESY, rotating frame nuclear Overhauser effect spectroscopy; TOCSY, total correlation spectroscopy; XR ESI/MALDI-FT-ICR-MS, eXtreme resolution electrospray ionization/matrix assisted laser desorption time of flight-Fourier transform-ion cyclotron resonance mass spectrometry; 1D, one-dimensional; 2D, two-dimensional.  (22)(23)(24); and S. fredii NGR234 (25). "2/3", 3-or 2-linked; "OHB", hydroxybutanoate; "Pyr", pyruvate; "OAc", acetate; "O-Suc", succinate. SEPTEMBER 30, 2016 • VOLUME 291 • NUMBER 40 notypes on the model legume Lotus japonicus Gifu depending on the step at which EPS biosynthesis is halted. Mutants in genes affecting synthesis of the EPS backbone (exoB and exoA) form nitrogen-fixing nodules but with a delay compared with wild-type R7A (38). In contrast, mutants involved in synthesis of the EPS branch (e.g. exoU and exoO) form infection foci within curled root hairs, but only a few short infection threads are observed, and these fail to extend (38,40). Several lines of evidence suggest that M. loti EPS acts as a positive signaling molecule. A Tn5 transposon mutant with an insertion within the 3Ј end of exoU had a similar dry-colony morphology to a null exoU mutant, but compositional analysis revealed it produced a small quantity of wild-type EPS. This mutant nodulated L. japonicus normally. In vivo complementation experiments performed using mixtures of the exoU mutant and a nodA mutant strain, which produced wild-type EPS but was unable to form nodules, showed that EPS allowed infection thread extension to the base of root hairs but not through to the cortical cell layer (38).

Structures of High and Low Molecular Mass EPS of M. loti R7A
A search for plant mutants able to form nodules with the M. loti exoU mutant led to the identification of a specific L. japonicus receptor EPR3 that directly perceives M. loti EPS. EPR3 is a membrane-bound receptor-like kinase containing three LysM domains within the amino-terminal region. It was proposed that it differentiates between compatible and incompatible forms of EPS to regulate the passage of the beneficial bacteria through the plant's epidermal cell layer (40). An octasaccharide component of M. loti strain R7A EPS bound directly to the purified ecto-domain of EPR3 in in vitro binding assays, thus confirming the signaling role of LMM EPS in symbiosis.
These studies and future studies of receptor-ligand binding require the determination of the structure of both HMM and LMM EPS from M. loti strain R7A. In this study, we report the determination of the structure of both the HMM and LMM EPS from M. loti R7A, including the location of the four non-stoichiometric O-acetyl substituents.

Results
We have previously reported that GC/MS composition analysis of R7A EPS showed the presence of Glc, Gal, GlcpA, and RibfA (38). The stereochemical configurations of Glc, Gal, GlcA, and RibA (carboxyl reduced to Rib) were all assigned as "D" based on comparison of the retention times of their trimethylsilyl-(S)-(ϩ)-2-butyl glycosides with the authentic standards of D-Glc, D-Gal, D-GlcA, and D-Rib. Below, we determine the structures of both HMM and LMM EPS from R7A.
Glycosyl Linkage Analyses of HMM EPS-Glycosyl linkage determination was performed by methylation analysis of PMAA derivatives before and after reduction of carboxyl groups with sodium borodeuteride. Before reduction, GC/MS analysis of the PMAAs showed the presence of 3-linked Galp, 4,6-linked Glcp, 6-linked Glcp, and 4-linked Glcp in a 1.0:0.93: 1.1:3.3 ratio. To identify the GlcpA and RibfA linkages, methylation analysis was also performed after carboxyl group reduction (supplemental Fig. 1). In the case of GlcA, carboxyl reduction resulted in its conversion to a 4-linked Glc-6-D 2 residue as indicated by the presence of primary fragment ions of m/z 118 and m/z 235 in its PMAA derivative. The PMAA of 4-linked Glcp normally gives ions of m/z 118 and 233, which were also observed; however, after carboxyl reduction, the appearance of the m/z 235 ion provided proof of a 4-linked Glc residue that was derived from 4-linked GlcpA due to the two deuterium atoms present at C6 (supplemental Fig. 1A). A small amount of terminal GlcpA (characterized by primary fragment ions of m/z 163 and 207) was also observed (supplemental Fig. 1B). The RibfA was converted to Ribf-5D 2 with two deuterium atoms at C5. The GC/MS analysis showed that its PMAA derivative was consistent with it being present as a terminal residue (supplemental Fig. 1C). Thus, this EPS consists of 3-linked Galp, 4,6-linked Glcp, 6-linked Glcp, and 4-linked Glcp and 4-linked GlcpA in an approximate 1:1:1:3:1 ratio and with terminal RibfA suggesting a polymer built of eight glycosyl residues (residues A-H).
NMR Analysis of Glycosyl Sequence in HMM EPS-Initial proton NMR analysis of HMM EPS showed poor spectral resolution due to extreme viscosity but indicated extensive O-acetylation of the polymer (supplemental Fig. 2). The location of O-acetyl groups on the M. loti R7A EPS is discussed further below. However, to simplify the EPS glycosyl residue structural analysis, the sample was de-O-acetylated prior to NMR analysis, which significantly improved the quality of the spectra. The anomeric signals of the RibfA (residue A) and GlcpA (residue B) resonated at 4.96 and 5.27 ppm, respectively (Table 1 and Fig. 2). Following the connectivities and measuring chemical shift values in the COSY, TOCSY, and HSQC spectra yielded the complete chemical shift assignments of these two residues. The carbon chemical shifts of the RibfA residue were consistent with reported values for an unsubstituted ␤-anomer (41). The ␤-configuration was also supported by the J 1-2 coupling constant of less than 2 Hz; the reported values for ␣and ␤-anomers of RibfA are 4.1 and 1.7 Hz, respectively (41). The chemical shifts for H1/C1 of GlcAp (B) H1/C1 of 5.27/100.4 ppm point to an ␣-anomer, and the downfield carbon shift for C4 at 82.4 ppm (B4) is indicative of glycosylation at this position, a result that is consistent with the methylation data demonstrating a 4-linked GlcA residue (supplemental Fig. 1A). The HMBC spectrum demonstrated that the RibfA residue was linked to the 4-position of the GlcpA residue by showing a cross-peak between the RibfA H1 at 4.96 ppm and the C4 of GlcpA at 82.4 ppm (A H1 /B C4 , Fig. 3A) and also coupling between the C1 of RibfA and H4 of GlcpA (109.7/3.49 ppm) (A C1 /B H4 , Fig. 3B). The NOE cross-peak at 5.27/3.55 ppm (B H1 /Glc H4 ) (Fig. 3C) showed a close contact between H1 of GlcAp and H4 of a Glcp residue, providing proof of a 134 linkage between GlcpA and Glcp. This linkage was also supported by the HMBC experiment, which showed GlcpA C1 to Glcp H4 coupling at 100.4/ 3.55 ppm (B C1 /Glc H4 ) (Fig. 3B). Thus, these data support the following terminal trisaccharide glycosyl sequence as being a part of the oligosaccharide branch of the EPS repeat unit structure: ␤-D-RibfA-(134)-␣-D-GlcpA-(134)-D-Glcp-(13 (trisaccharide unit A-B-C). Additionally, the methylation data, which showed that a small portion of GlcpA was present as a terminal linkage (supplemental Fig. 1A), suggested that RibfA is occasionally absent, resulting in terminal ␣-D-GlcpA-(134)-␤-D-Glcp-(13 disaccharide sequence. Indeed, we were able to observe weak NMR signals of a spin system due to the terminal GlcpA residue with the H1/C1 at 5.31/100.4 ppm (residue Bt) ( Table 1 and Fig. 2).
The remaining anomeric signals in the spectra of HMM EPS were extensively overlapped so that a complete assignment for these residues was not possible. Most of these signals resonated near 4.5 ppm and formed a cluster corresponding to five protons. A signal at 4.49/4.13 ppm in the TOCSY suggested that one of the anomeric protons in this cluster belonged to a ␤-Gal residue (residue H). Its assignment, including chemical shifts of H1/C1 through H5/C5, was possible by using the chemical shift prediction calculation CASPER tool (42) (supplemental Table  1). The remainder of the group of signals around 4.5 ppm were therefore from ␤-Glc residues. The anomeric ␤-configuration was assigned by comparison of carbon and proton chemical shifts with common literature values. Some overlapping peaks near 4.6 ppm were also characterized as belonging to ␤-Glc by the upfield chemical shift of their H2 protons (3.30 -3.40 ppm). The COSY and TOCSY spectra ( Fig. 2B and supplemental Fig.  3), together with the HSQC spectrum ( Fig. 2A), showed that two Glcp ring systems had H6/C6 resonances at 4.16 -3.90/ 68.9 -69.2 ppm (see Table 1); the downfield chemical shifts for C6 supported the assignment of these resonances to the two Glcp residues that are substituted at C6, i.e. the 6-and 4,6linked Glcp residues (residues D and E, respectively). The H6/C1 coupling at 3.83/103.3 ppm in the HMBC spectrum (  Table 1.
Because of the extensive overlap of the anomeric signals of the various Glcp residues, no further definite structural information was available from the spectra of HMM EPS. Therefore, we concentrated our efforts on the structural characterization of LMM EPS.
Determination of Glycosyl Sequence of De-O-acetylated LMM EPS-To determine the monosaccharide sequence of EPS, we attempted analysis of LMM EPS from M. loti R7A. Co-extraction and co-precipitation of associated cyclic ␤-glucan along with EPS (38) made the analysis difficult. We took advantage of the availability of an M. loti R7A ndvB mutant that produces both HMM and LMM EPS but is unable to synthesize cyclic ␤-glucans (38). The size exclusion chromatography elution profile of crude LMM R7AndvB EPS indicated the presence of only one major fraction that eluted slightly ahead of the standard of maltoheptaose (supplemental Fig. 4), suggesting that LMM EPS was represented by only one type of oligosaccharide of homogeneous size. This was confirmed by MALDI-TOF MS displaying a molecular ion corresponding to RibAGlcAHex 6 OAc 3 (Fig. 6B). Initial comparative NMR analysis of HMM and LMM exopolysaccharides from both R7A and R7AndvB indicated their structural similarity (see also MALDI MS analysis of O-acetylated LMM EPS below). The PMAA linkage analysis with methylation done after carboxyl reduction indicated the presence of 6-linked Glcp, 4-linked Glcp, terminal RibfA, and 4-linked GlcpA in an approximate ratio of 2:3:1:1, with some terminal GlcpA and traces of a 3-linked Galp residue. The latter was detected in a smaller quantity and underwent partial degradation and modification in that it was converted to its furanose form, and in addition, the C1 of Gal was partially converted to a methyl glycoside during carboxyl esterification by methanolysis. As a result, we observed an early eluting peak in GC/MS due to Gal with a diagnostic electron ionization fragment at m/z 206 (supplemental Fig. 1D). These findings suggest that the galactose residue was at the reducing end of the LMM EPS. Unlike the HMM R7A polymer, no PMAA corresponding to the branching 4,6-linked Glcp was found in the LMM EPS, suggesting it could be an attachment point for a new subunit of the polymer. This result indicates that the LMM EPS is likely a single repeat unit of EPS.
These initial PMAA composition data on ndvB LMM EPS were supported by 1D and 2D NMR analyses on de-O-acetylated EPS. The 1D proton spectrum showed the same general signal pattern as that of de-O-acetylated HMM EPS but was of much better quality and higher resolution as a result of the small molecular weight of LMM EPS. As in the case of de-Oacetylated HMM EPS, ␤-RibfA and ␣-GlcpA were easily identified, and their chemical shifts were assigned from the COSY, TOCSY, and HSQC spectra. Likewise, the inter-residue connections between O-1 of RibfA and O-4 of GlcpA, as well as that from O-1 of GlcpA to O-4 of a ␤-Glc residue (trisaccharide unit

Structures of High and Low Molecular Mass EPS of M. loti R7A
SEPTEMBER 30, 2016 • VOLUME 291 • NUMBER 40

JOURNAL OF BIOLOGICAL CHEMISTRY 20949
A-B-C), were confirmed by correlations in both ROESY and HMBC spectra ( Fig. 4 and Table 2). However, in contrast to the spectra of HMM EPS, TOCSY and ROESY of LMM EPS were able to resolve the four ␤-Glcp spin systems whose anomeric proton resonances clustered around 4.5 ppm, such that a complete assignment of these residues became possible. Two of these were found to be 4-linked ␤-Glcp, and the other two were 6-linked ␤-Glcp residues. In addition to these residues, there was a set of two almost identical spin systems with chemical shifts consistent with 4-linked ␤-Glcp. Their anomeric proton signals resonated at 4.7 ppm with a combined integral of one proton, suggesting that these spin systems corresponded to two variants of the same 4-linked ␤-Glcp residue in the oligosaccha-ride. Finally, there were two variants of the 3-linked Galp residue, one ␣and one ␤-anomeric configuration (residues H␣ and H␤, respectively   Table 2. All spectra were recorded at 70°C. Dotted circle in B highlights an H4 residue of 3-␤-Galp. This low intensity signal was locally increased and does not represent actual abundance in comparison with intensities of chemical shifts of the remaining residues.   Table 2. All spectra were recorded at 70°C. Hex, gluco or galactosyl residue. ppm, respectively, see Table 2) identifying them as the two anomeric forms of the reducing end residue of the LMM oligosaccharide. ROESY and HMBC spectra were used to define the remainder of the monosaccharide sequence ( Fig. 4 and Table  2). Unambiguous ROESY and HMBC correlations were found between H3 of 3-␤-Galp (residue H␤) and one of the two 4-␤-Glcp anomeric signals near 4.7 ppm (residue G(␤)), as well as between H3 of 3-␣-Galp (residue H␣) and H1 of the other 4-␤-Glcp (residue G(␣)). At the other end of the sequence, the 4-linked ␤-Glcp (residue C) that is proximal to the 4-␤-GlcpA residue was identified by ROESY and HMBC correlations with H1 of GlcA pinpointing the H4/C4 chemical shifts of that residue to 3.59/78.4 ppm. TOCSY and HSQC showed that this signal is part of the spin system whose H1 resonated at 4.512 ppm. H1 of residue C in turn showed ROESY correlations with signals at 4.18 and 3.86 ppm, which arose from H6a and H6b of residue D, demonstrating the linkage of residue C to O-6 of residue D. The ROESY cross-peaks of the anomeric signal of residue D matched with the H6 resonances associated in the TOCSY spectrum with residue E. The anomeric signal of the latter exhibited a ROESY crosspeak at 3.66 ppm, which coincided with H4 of both residues F and G. However, the linkage from residue E to residue F was clearly evidenced by an HMBC correlation of H1 of residue E to a signal at 79.   EPS was almost identical to the octasaccharide, except that residue E was 4,6-␤-linked in the polymer and 6-␤-linked in the octasaccharide (refer to previous sections). This indicated that residue E contained both the possible branching and polymerization sites in HMM EPS. Therefore, we hypothesized that the LMM octasaccharide would be a subunit of the HMM EPS. If this was the case, the NMR chemical shifts of the different spin systems of LMM and HMM EPS should be the same, except for those involved in the branch point, i.e. residues E (Glcp) and H (Galp). To test this hypothesis, we overlaid and compared the 1 H-13 C HSQC spectra of the de-O-acetylated HMM and LMM and identified the signals that were unique in each spectrum (Fig. 5, red, LMM octasaccharide; black, HMM polymer). The two spectra were almost identical with few exceptions, which were accounted for by the different polymerization states of the two saccharides. Most of the signals unique to HMM EPS belonged to a 4,6-linked Glcp (residue E). These signals were shifted downfield in the proton dimension relative to the corresponding 6-linked Glcp residue in LMM EPS. These differences in proton, as well as in the carbon, chemical shifts of residue E in the two saccharides are in excellent agreement with what is predicted by the CASPER chemical shift calculation tool (42) (see supplemental Table 1). The replacement of a 6-linked Glc by a 4,6-linked Glc is accounted for by the polymerization of the octasaccharide, which forms a bond between O-4 of residue E and O-1 of residue H. Involvement of the latter residue in the polymerization was confirmed by the presence of two HSQC signals appearing only in the spectrum of HMM, namely those of H2 and H5 of ␤-Gal. Conversely, upfield anomeric carbon chemical shifts of ␣-Gal and ␤-Gal, as well as H3 of ␣-Gal appeared only in the spectrum of the octasaccharide where the Gal residue, being located at the reducing end, is able to undergo mutarotation, giving rise to both ␣and ␤-anomers. The final difference observed in the two spectra was the presence of terminal GlcA only in the spectrum of HMM EPS. As already shown by the linkage data, some repeating units in the polymer lack the riburonic acid residue, whereas the GlcA in LMM EPS is stoichiometrically substituted with RibfA.  The only signals that are different in both spectra are those belonging to the residues involved in connecting the subunits to each other (residues E and H) and those belonging to terminal GlcA (residue Bt), which is substoichiometric in HMM EPS but stoichiometric in LMM EPS. The chemical shift changes observed in residues E and H upon polymerization are consistent with predicted values as described previously (42). Taken together, these data provide strong evidence that the octasaccharide derives from the polymer and is the repeating unit of HMM EPS. In addition, this analysis complements and provides missing information for residues E and H in the polymer ( Table 1).  39 and 1395.38 consistent with the same octasaccharide with three and two O-acetyl groups, respectively. These data, together with the structural analysis described above, support the conclusion that the LMM EPS from both R7A and its ndvB mutant is an octasaccharide, representing one repeat unit of the EPS polymer (RibAGlcAGlc 5 Gal) n substituted with one to three O-acetyl groups.

Localization of O-Acetyl Groups in the LMM Monomeric
A series of NMR experiments on O-acetylated R7AndvB LMM EPS supported the presence of eight glycosyl residues (A-H), consistent with NMR analysis of the de-O-acetylated LMM EPS described above. To streamline data interpretation, we marked the position of the O-acetyl substituent(s) on the glycosyl residue with a subscripted number, e.g. A 3 and A 2,3 for 3-O-acetylated and 2,3-di-O-acetylated residue A, respectively. Different NMR variants of the same glycosyl residue, likely due to different substitution patterns by O-acetyl groups on neighboring residues, were labeled with prime marks, e.g. EЈ, EЉ, Eٞ, respectively. The 2D TOCSY spectrum showed nearly 30 individual monosaccharide residues (Fig. 7A). Tracing connectivities in the COSY spectrum and reading the corresponding carbon chemical shifts from the HSQC spectrum allowed us to assign chemical shifts to most of the protons and carbons of 23 of these residues (Table 3). Downfield carbon chemical shift displacements relative to the corresponding monomeric sugars were used to identify glycosylated positions and allowed us to classify the linkage type of each residue. The anomeric configuration of each residue was determined by J 1,2 proton-proton coupling constants and comparison of carbon and proton chemical shifts with common literature values. Thus, we found three terminal ␤-RibfA residues (A, A 2,3 , and A 3 ), one 4-␤-Gl-cpA residue (B), seven 4-␤-Glcp residues (C, CЈ, G, GЈ, F 2 , F 3 , and F 3 Ј), 10 6-␤-Glcp residues (D, D 3 , D 3 Ј, D 3 Љ, E, EЈ, EЉ, Eٞ, E 2 , and E 2 Ј), and the ␣and ␤-forms of the reducing-end Galp residue (H␣ and H␤, respectively) ( Table 3). Downfield proton chemical shift displacements were indicative of O-acetylation. The three ␤-RibfA residues were distinguished by their acetylation patterns in that A 2,3 was 2,3-di-O-acetylated, A 3 was 3-O-acetylated, and A was not acetylated. One of the 4-␤-Glcp residues was 2-O-acetylated (F 2 ), two were 3-O-acetylated (F 3 and F 3 Ј), and four were not acetylated (C, CЈ, G, and GЈ). Three of the 6-␤-Glcp residues were 3-O-acetylated (D 3 , D 3 Ј, and D 3 Љ), two were 2-O-acetylated (E 2 and E 2 Ј), and five were not acetylated (D, E, EЈ, EЉ, and Eٞ). Residues G and GЈ were identified as adjacent to the reducing end of Gal (H) due to crosspeaks to a resonance at 82.9 ppm (C-3 of ␤-Galp) in the HMBC spectrum (Fig. 7B). Both ROESY and HMBC confirmed that the RibfA residues were attached to O-4 of GlcpA. Correlations were also found between GlcpA H1 and H4 (ROESY) and C4 (HMBC) of residues C and CЈ (data not shown). The anomeric protons of residues E, EЈ, EЉ, and E 2 Ј showed HMBC crosspeaks to C4 of residue F 3 , which resonated near 75 ppm (Fig. 7). This chemical shift, which is unusually upfield for a 4-linked ␤-Glc residue, is due to the ␤-shielding effect of the O-acetyl group in the 3-position (43).
Because of overlap of the anomeric signals in the spectra, it was not possible to identify specific further connectivities between residues. However, the HMBC spectrum provided a way to define each residue more specifically (Fig. 7). Whereas the HSQC data identified the glycosylated position in each residue, the HMBC spectrum showed the linkage position of its reducing-side neighbor. In other words, each residue could be defined according to the form [3x-␤-Glc-13y], where x is the residue's own linkage (4 or 6), and y is the substituted position (3, 4, or 6) of its reducing-side neighbor. Each of the five Glc residues in the repeating unit is uniquely specified by this definition. Residue C is 34-␤-Glc-136; residue D is 36-␤-Glc-136; residue E is 36-␤-Glc-134; residue F is 34-␤-Glc-134;  (Table 3). Table 4 summarizes annotations for all identified glycosyl residues in the O-acetylated octasaccharide. The anomeric protons of residues E, EЈ, EЉ, Eٞ, E 2 , F 2 , F 3 , and F 3 Ј displayed HMBC cross-peaks between 78 and 79 ppm, demonstrating their linkage to the 4-position of their neighbor, whereas residues C, CЈ, D, D 3 , D 3 Ј, and D 3 Љ had HMBC correlations near 69 ppm, from their linkage to the 6-position of their neighbor. Residues G and GЈ had H1/C3 cross-peaks to a carbon at 83 ppm, which is the Gal C3.
The results of NMR assignments and O-acetylation pattern in the octasaccharide were further supported by positive-mode Fourier transform ion cyclotron (FT-ICR) electrospray ionization (ESI) mass spectrometric analysis (supplemental Fig. 5 and supplemental Table 2). A distribution of up to four acetyl groups was observed in the LMM EPS octasaccharide [RibAGlcAHex 6 OAc (1)(2)(3)(4) ϩ Na] ϩ , with the majority of species possessing three O-acetyl substitutions consistent with results from negative-mode MALDI-MS analysis (Fig. 6). Interest-ingly, loss of the penturonic acid (residue A) from the major structure was accompanied with loss of one O-acetyl group, suggesting a conserved O-acetylation of the terminal RibA (supplemental Table 1, ions I and II). This was particularly observed in collision-induced dissociation MS/MS fragmentation of ion m/z 1461.4017, [RibAGlcAHex 6 OAc 3 ϩ Na] ϩ , by detection of a daughter ion at m/z 1273.3705 (IX) [GlcAHex 6 OAc 2 ϩ Na] ϩ , due to loss of RibA(OAc) (residue A) (supplemental Table 2 and supplemental Fig. 5).
Further assignments of R7A LMM EPS were possible by examination of the FT-ICR-ESI fragmentation pattern of a biotinylated derivative of the EPS octasaccharide ( Fig. 8 . Determination of the linkage positions x and y for each 3x-␤-Glc-13y residue in the native octasaccharide. The residues are labeled according to Table 3, and the numbers associated with the residue labels refer to the proton numbers, e.g. The green lines were shortened to avoid excessive crowding. For example, residue E 2 shows two sets of cross-peaks in the TOCSY spectrum, one correlating to H2 (designated E 2 2), at 4.79 ppm in F2, and one correlating to H1 (designated E 2 1), at 4.711 ppm in F2. Both of these sets include a cross-peak at 4.20 ppm, corresponding to H6 of residue E 2 . The downfield chemical shift of this signal indicates that residue E 2 is 6-linked, i.e. x ϭ 6. The HMBC spectrum shows a cross-peak correlating H1 of residue E 2 (E 2 1) to a signal at 79.5 ppm, which is consistent with the chemical shift of C-4 of a ␤-glucose residue, therefore y ϭ 4, and thus we conclude that residue E 2 is of the type 63␤-Glc-134.
HexOAc ϩ H] ϩ (fragment ion XII    on either of residues E or F. Furthermore, identification of [HexLinkerBiotin ϩ H] ϩ (fragment ion X) and [Hex 2 LinkerBiotin ϩ H] ϩ showed that residues G and H were non-acetylated (Fig. 8, A and B). Fragment ions are summarized in Table 5. Considering all NMR, MALDI-MS, FT-ICR ESI MS and MS/MS, and glycosyl linkage results together, we conclude that the LMM EPS consists of eight glycosyl residues in the sequence shown in Table 4

Discussion
M. loti strain R7A is a microsymbiont of the model legume L. japonicus. Recent studies using this model system have indicated wild-type rhizobial EPS plays a role in enhancing the establishment of the determinate symbiosis, whereas truncated EPS produced by certain R7A exo mutants (e.g. exoU) plays a negative role and prevents the establishment of infection threads and their passage through the epidermal cell layer. It is proposed that LMM forms of both types of EPS are perceived by the LysM receptor-like kinase, EPR3, to regulate infection thread passage. Genetic evidence using rhizobial mutants (38) indicates that the positive function of wild-type EPS is dominant over the negative effect of truncated EPS, but whether this reflects different binding affinities of the two types of EPS to EPR3 remains to be determined. The detailed structures of both HMM and LMM EPS reported in this study are crucial for the biochemical characterization and understanding of the molecular mechanisms by which EPS is perceived in symbiosis. Structural studies of EPS are also important for characterization of the various phenotypes observed with M. loti exo mutants and for understanding the differences between symbiotically compatible and incompatible EPS, which may explain some of the variation in symbiotic efficiencies seen between rhizobial strains producing very similar Nod factors.
In this work, we show that the repeating oligosaccharide unit of HMM EPS and LMM EPS oligosaccharide share a similar structure in that they consist of highly O-acetylated (  of fourth glucose (asterisked above) of that octasaccharide in the formation of the polysaccharide. The terminal residue of the branching oligosaccharide was found to be D-riburonic acid, which to our knowledge has only been reported in EPS from one other microorganism, reported as Rhizobium (Sinorhizobium) meliloti IFO 13336 (44). Indeed, the structure of the carbohydrate portion of R7A EPS is consistent with that reported for IFO 13336, with the exception that the branched-chain terminal RibfA in R7A was ␤-D-RibfA and not ␣-D-RibfA, and the presence and location of the O-acetyl groups were not determined in IFO 13336 EPS. Amemura et al. (44) determined their structure using a method involving successive fragmentation with two ␤-D-glycanases, and hence their study provides strong complementary support for the structure reported in this paper. Also, Amemura et al. (44) concluded that RibfA is in the ␣-anomeric configuration because the NMR chemical shift of its anomeric proton is greater than 5 ppm. However, the anomeric proton chemical shifts of aldofuranoses tend to be greater than 5 ppm in both ␣and ␤-anomeric configurations (45,46). Given the similarities of R7A and IFO 13336 EPS, it seems likely that IFO 13336 was misidentified as S. meliloti, and indeed the strain is listed as belonging to Mesorhizobium sp. in the NITE Biological Resource Center (NBRC) Culture Collection.
We identified heterogeneity in EPS O-acetylation, showing that on average an octasaccharide unit is substituted with three O-acetyl groups out of six possible O-acetylation sites within four glycosyl residues (Table 4 and Fig. 9). O-Acetyl groups are well known to migrate between neighboring hydroxyl groups of carbohydrates (47). Therefore, it is possible that the variability in location is due to chemical rearrangement rather than specific enzymatic action. Although migration from O-3 to O-4 is

TABLE 5 High resolution MS analysis of biotin conjugate of M. loti R7A EPS octasaccharide and proposed compositions
For mass spectra and structure assignments refer to Fig. 8. SEPTEMBER 30, 2016 • VOLUME 291 • NUMBER 40 precluded in the 4-linked residues, it would be possible in 6-linked residues and in fact has been observed in Glc monomers (48). However, the complete absence of 4-O-acetylation in any of the residues observed in the octasaccharide indicates that migration is likely not a factor. The genetic basis for the addition of the three O-acetyl groups remains to be identified. The structures of M. loti EPS, succinoglycan from S. meliloti 1021, and EPS from Rhizobium tropici CIAT 899 (16 -20, 49) share many similarities. In this type of EPS polymer, every new octasaccharide subunit attaches via its reducing-end Gal to C4 of the fourth glycosyl residue of the polymer. In most of the octasaccharide polymers identified thus far, it is a branching 4,6-glucose. Hence, each EPS subunit consists of an O-acetylated tetrasaccharide repeating backbone with the remaining glycosyl residues forming the branching tetrasaccharide. The backbone tetrasaccharide and the two first side-chain residues seem to be conserved in these rhizobial strains. The proteins required for the assembly of this hexasaccharide (ExoY, -A, -L, -M, -O, and -U) have been identified for succinoglycan (13,50). The genetic organization of the corresponding genes is also conserved among the three species, which all make EPS with an octasaccharide repeating unit, and also S. fredii strains NGR234 and HH103, which make an EPS with a nonasaccharide repeating unit (compare structures in Figs. 1 and 9 with supplemental Fig. 6) (25, 51) and a six-residue backbone. Interestingly, this backbone is identical to the hexasaccharide described above, and the three-residue branch is attached to the fifth residue of the backbone. Hence, several diverse rhizobia synthesize similar EPS with the differences restricted to the last two or three residues of the branch and the types and positioning of the non-carbohydrate substituents. It seems likely that these differences play a critical role in host specificity during establishment of the rhizobium-legume symbiosis. Currently, nothing is known about the biological properties bestowed on the EPS by riburonic acid or the mechanisms by which it is synthesized and added to the EPS. In addition, the gene(s) required for the addition of glucuronic acid to the M. loti EPS are unknown. We are currently addressing these questions. The branch chains of the EPS discussed above contribute to the acidic character of the EPS (uronic acids, pyruvylated or succinylated neutral sugars) as well as their sequence heterogeneity. Although the functions of non-carbohydrate modifications of rhizobial EPS are not fully understood, there is evidence of a pivotal role in the symbiosis. In an S. meliloti exoZ mutant defective in expression of acetyltransferase, lack of O-acetylation of succinoglycan caused increased susceptibility of the polymer to cleavage by endogenous ExoK and ExsH endo-1,3-1,4-␤-glycanases. This resulted in increased production of LMM EPS, thus suggesting that O-acetylation may play a regulatory role in EPS length modifications. In contrast, an exoH mutant defective in succinylation produced only HMM EPS (37,52,53). The exoZ mutant was able to establish symbiosis with alfalfa but with reduced efficiency of infection thread formation and reduced nodule invasion (36,54), whereas the exoH mutant was unable to initiate infection thread formation. Lack of pyruvylation of succinoglycan caused by an exoV mutation resulted in the production of only monomer units, indicating pyruvylation is required for succinoglycan polymerization (50).

Structures of High and Low Molecular Mass EPS of M. loti R7A
In other bacteria, acetyl groups may also contribute to rheological, conformational, and physical properties of EPS. These changes provide protection, allowing early stages of biofilm formation, cell aggregation, and steric prevention of enzymatic degradation by soil microorganisms (55,56). Therefore, they may perform similar functions in rhizobia.
The LMM EPS characterized in this work is a monomeric octasaccharide with an average of three acetyl group substitutions. This octasaccharide is recognized by the plant EPR3 kinase receptor resulting in a positive response (40). It seems likely that the polysaccharide produced by the M. loti exoU mutant that initiates a negative response from the EPR3 receptor is a pentasaccharide with reduced acetylation and acidity compared with the octasaccharide. It will be of interest to determine the role of the O-acetyl groups as well as the effect of polysaccharide chain length on molecular recognition by the receptor and to determine how these opposing responses control passage through the root hair and the infection process.
A signaling role of LMM EPS in the symbiotic interaction with plants was also proposed for S. meliloti 1021 and S. fredii NGR234. In the case of S. meliloti, a trimer repeat of the octasaccharide was the most symbiotically active molecule (20,34), whereas only monomeric octa-and nonasaccharides were identified for NGR234, indicating these were the symbiotically active species (26). However, recent studies on succinoglycan in S. meliloti 1021 suggest no requirement for LMM EPS in host invasion in indeterminate symbiosis and a more critical role of succinylation of the polysaccharide rather than molecular size (35).Ourdiscoverythatonlyanacetylatedoctasaccharidemonomer is found in the crude LMM fraction rather than dimers or trimers of the repeating octasaccharide is in agreement with the data reported for S. fredii NGR234 (26), suggesting that a LMM EPS size requirement for symbiotic interaction could be both microbe-and host plant-dependent. Further studies on symbiotically compatible and incompatible exopolysaccharides and their contrasting effects on EPR3-mediated control of the rhizobial infection processes are underway.

Experimental Procedures
Isolation and Purification of M. loti R7A EPS-The wild type and an ndvB mutant of M. loti strain R7A, strain R7AndvB (38), which does not make cyclic ␤-glucans, were used for the isolation of EPS. The strains were grown in glucose/rhizobium-defined medium (Rhizobium defined medium containing 10 mM glucose) (38), and HMM EPS was precipitated from stationary phase culture supernatants with 3 volumes of ethanol (EtOH) and purified by chromatography as described previously (38). Traces of cyclic ␤-glucan were removed from chromatographed HMM EPS by extensive dialysis against H 2 O using a 50,000-Da MWCO dialysis tubing. The supernatant recovered from the 3-volume EtOH HMM EPS precipitation was concentrated by rotary evaporation, and 9 volumes of EtOH (v/v) were added, allowed to stand overnight at 4°C, and centrifuged at 5000 ϫ g for 20 min to form crude LMM EPS precipitate. The crude LMM was desalted and purified from monomeric glucose on a Bio-Gel P2 column (120 ϫ 1.5-cm inner diameter) or by dialysis against H 2 O using 1000-Da MWCO dialysis tubing and was later resolved on a Superdex Peptide 10/300 FPLC column with 50 mM ammonium acetate used as eluent. Eluting fractions were monitored by refractive index and by a multiple wavelength detector (Agilent Technologies 1200) at 210, 230, and 280 nm.
Chemical Composition Analysis-To identify the glycosyl residues in the EPS, trimethylsilyl methyl glycosides were prepared as described previously (38,57). The linkage analysis of neutral sugars and uronic acids was done by preparing and analyzing PMAAs (57) or PMAAs prepared after mild methyl carboxyl esterification and reduction of uronic acids prior to permethylation. The details of this chemical modification were previously reported (38). Stereochemical configuration of the glycosyl residues was performed following the method of Gerwig et al. (39,58) with optically active (S)-(ϩ)-2-butanol (Sigma).
De-O-acetylation of EPS-HMM and LMM EPS were de-Oacetylated in 12.5% NH 4 OH for 12 h at 35°C. Alternatively samples were de-O-acetylated with 10 mM KOH for 5 h at 20°C, and the reaction was quenched with HCl. Salts were removed by dialysis (1000-Da MWCO) against several changes of deionized water.
Mass Spectrometric Analysis of Native LMM EPS-The LMM EPS fraction was recovered from the 9-volume EtOH precipitate and Superdex Peptide 10/300 column, respectively, dissolved in water, mixed 1:1 (v/v) with 0.5 M 2,4,6-trihydroxyacetophenone monohydrate matrix in methanol, and spotted onto a MALDI plate. MALDI-TOF MS spectra were acquired on an Applied Biosystems AB SCIEX TOF/TOF 5800 system in the negative reflectron detection-mode and processed with Data Explorer (Applied Biosystems). High resolution mass spectrometry was performed on a Bruker Solarix XR ESI/ MALDI-FT-ICR-MS instrument equipped with a 7-tesla magnet. Spectra were acquired in the positive ESI-mode, externally calibrated with NaOTFA cluster ions, and processed using Bruker Data Analysis version 4.1.
NMR Analysis of EPS-Native and de-O-acetylated HMM and LMM EPS were exchanged twice in 99.9% D 2 O and finally dissolved in 100% D 2 O (Cambridge Isotope Laboratories, Andover, MA) at a final concentration of 6 mg/ml. 1D proton and 2D 1 H-1 H (COSY, TOCSY, NOESY, ROESY, and 1 H-13 C (HSQC and HMBC) spectra were acquired at 70°C on a Varian 600 MHz spectrometer equipped with 3-mm cold probe and processed using MestReC/Nova software (Mestrelab Research, Santiago de Compostela, Spain). For 2D acquisitions, the solvent signal was suppressed with a standard solvent suppression experiment in one-dimensional proton NMR. COSY experiments were recorded using sets of 1024 time increments