Structural implications of lipoarabinomannan glycans from global clinical isolates in diagnosis of Mycobacterium tuberculosis infection

In Mycobacterium tuberculosis (Mtb), surface-exposed Lipoarabinomannan (LAM) is a key determinant of immunogenicity, yet its intrinsic heterogeneity confounds typical structure–function analysis. Recently, LAM gained a strong foothold as a validated marker for active tuberculosis (TB) infection and has shown great potential in new diagnostic efforts. However, no efforts have yet been made to model or evaluate the impact of mixed polyclonal Mtb infections (infection with multiple strains) on TB diagnostic procedures other than antibiotic susceptibility testing. Here, we selected three TB clinical isolates (HN878, EAI, and IO) and purified LAM from these strains to present an integrated analytical approach of one-dimensional and two-dimensional Nuclear Magnetic Resonance (NMR) spectroscopy, as well as enzymatic digestion and site-specific mass spectrometry (MS) to probe LAM structure and behavior at multiple levels. Overall, we found that the glycan was similar in all LAM preparations, albeit with subtle variations. Succinates, lactates, hydroxybutyrate, acetate, and the hallmark of Mtb LAM-methylthioxylose (MTX), adorned the nonreducing terminal arabinan of these LAM species. Newly identified acetoxy/hydroxybutyrate was present only in LAM from EAI and IO Mtb strains. Notably, detailed LC/MS-MS unambiguously showed that all acyl modifications and the lactyl ether in LAM are at the 3-OH position of the 2-linked arabinofuranose adjacent to the terminal β-arabinofuranose. Finally, after sequential enzymatic deglycosylation of LAM, the residual glycan that has ∼50% of α−arabinofuranose -(1→5) linked did not bind to monoclonal antibody CS35. These data clearly indicate the importance of the arabinan termini arrangements for the antigenicity of LAM.

The human adapted Mycobacterium tuberculosis complex (MTBC) exhibits a strong phylogeographical population structure, with some lineages occurring globally and others showing a strong geographical restriction (23,24). Among these lineages L2 and L4 are the most widespread globally, with L2 dominating in East Asia. L1 and L3 occur mainly in regions around the Indian Ocean. L5 and L6 are highly restricted to West Africa, whereas L7 is almost exclusively found in Ethiopia. Geographic location can introduce variability for TB screening because of heterogeneity in TB strain or clade prevalence. Our hypothesis was that Mtb clinical isolates have a wide spectrum of virulence, which is lineageassociated, modulates host immune response, and determines bacterial load in patients with pulmonary tuberculosis. The laboratory passaged strain H37Rv exhibits intermediate virulence, causing 50% macrophage lysis. We took this as a reference strain in this study.
Based on this information, in our work of global LAM characterization, we selected Mtb strains EAI from L3, IO from L1 and HN878 from L2 compared with H37Rv. LAM was purified in sufficient quantities from each strain to perform thorough analyses focusing primarily on NMR at first so that all features can be assessed in the native molecule, followed by enzymatic digestion and mass spectrometry analyses on released oligoarabinofuranosides. Analyses were carried out without any downstream derivatization to maintain the integrity of all substitution/s. The objectives of this study were to map the LAM phenotype in bacterial strains that cause TB disease in TB endemic geographical areas and examine whether any epidemiologically relevant structural characteristics were associated with those strains. Our study provides a comprehensive systematic analysis of the evidence for diversity in LAM in particular in bacterial strains that are of clinical relevance.

Isolation of LAM from clinical isolates
The TB clinical isolates represent three geographically distinct lineages (as referenced in (25)) wherein, HN878 (East Asia lineage), T17-IO (The Philippines/Rim of the Indian Ocean lineage) East African-Indian 91-0079-EAI (India and East Africa lineage) represent the most globally predominant lineages outside of those typed to the Europe and Americas lineage. The laboratory-type strain Mtb H37Rv was used as a reference strain and represents the Europe and Americas lineage. We investigated the intact LAM by extensive 1D and 2D NMR spectroscopy to find out differences in glycosidic linkages as well as small-molecule modifications. The findings were supported by mass spectroscopy on the enzyme-digested LAM terminal-arabinan fragments since these arrangements were presumed to be antibody binding structures.

Acylation substitutions in LAM
Succinyl residues are linked through an ester bond and only one set of acylated ring proton was observed in each of the HSQC spectra for HN878, EAI, and IO-LAM ( Fig. 2A). The cross peak at δ 4.85 ppm ( 1 H) and δ 79.2 ppm ( 13 C) was attributed to the H3/C3 of an arabinose ring. In fact, all the acyl functionalities should be linked through this proton unless they occupy the primary -OH groups. However, the absence of any downfield proton that could be attributed to H5 of Araf or H6 of Manp supports this finding. In addition, the absence of any anomeric proton in 1,3-bond correlation (in TOCSY) with the acylated ring proton ruled out acylation at the 2position of Araf (Fig. 2B). RvLAM on the other hand was found to have two sets of acylated ring protons ( Fig. 2A) with different intensities. The major cross peak was the H3/C3 of the Araf ring as found in all four LAM, and the minor cross peak was found at δ 4.83 ppm ( 1 H) and δ 81.1 ppm ( 13 C). Unfortunately, the weak peak could not affirm 1,3-bond correlation (TOCSY) with any anomeric proton, in the event the succinate was at the 2-position. The presence of a succinyl residue at the 2-position of 3,5-linked Araf was reported earlier in arabinogalactan isolated from Mtb H37Rv (27). The TOCSY of HN878 LAM (Fig. 2B) showed two sets of protons corroborating to a -C(O)-CH 2 -CH 2 -C(O)-structure unequivocally establishing a succinyl monoester residue. For confirmation, additional support was sought from enzymatic digestion of LAM. As a representative structure, the extracted ion chromatogram (LC/MS of enzyme digested LAM arabinan) showed same retention time for the ion m/z 1101.34 [M-H] corresponding to Suc 1 Man 2 Ara 5 for all the 3× LAM species (HN878, EAI and IO). This indicated that the monosuccinylated Man 2 Ara 5 has a similar structure in all LAM. MS-MS was performed on this ion to reveal the sequence (Fig. S4). This showed that Ara 5 structures are always linear in LAM (21). The analysis also supports the location of succinate at the 3-position of 2-linked Araf, as shown by NMR and described before. One interesting aspect was noted about the succinyl residues in EAI and IO-LAM. The 1 H-chemical shifts of the succinyl protons merged at δ 2.6 ppm giving an overlapping signal for 4× protons instead of usual δ 2.4 and δ 2.     (Fig. 5I). Collectively, these specific ions suggest that the α-hydroxy butyrate group is attached to the 2-linked Araf, which has only 3-position, available as a secondary -OH group, for substitution.

Confirmation of α-acetoxy butyrate using LC/MS-MS
We were able to investigate m/z 673.

Confirmation of MTX and MSX using NMR and LC/MS-MS
The presence of methylthioxylose (MTX) residue is evident in the NMR spectra of all LAM preperations-HN878, EAI, IO, and H37Rv. The correlation (TOCSY) of H5a and H5b, the diastereotopic protons, (δ 2.6 and 2.7 ppm) with δ 4.12 (H2), δ 4.18 (H3), δ 4.31 (H4) ppm respectively confirm (26,29) the presence of MTX. In addition, the sulfoxy variation of MTX, i.e., methylsulfoxylose (MSX) was also found to be present. The presence of MTX/MSX in relatively high NMR peak intensity signifies that there could be multiple sites that bear MTX and/or MSX. So far, it has been shown that MTX in LAM is present one per molecule of LAM (26). Our current observation indicates otherwise.
We had reported earlier that following digestion with endoarabinanase and LC/MS of released ararbinan fragments of LAM, low-intensity ions corresponding to

Overall abundance of LAM arabinan termini
The LC/MS profile of endoarabinanase released LAM arabinan termini revealed that all four LAM samples have abundant Ara 6 (EAI > IO > HN878 > H37Rv) and its mannose capped variations that are not succinylated (Fig. 8A). Two most prominent motifs are Ara 6 and Man 2 Ara 6 (H37Rv > HN878 > IO > EAI). Among the monosuccinylated versions, it seemed that the linear Ara 4 and Ara 5 with and without mannose capping dominated, although one needs to be cautious that this relative quantification is based on ion intensity only and is not absolute. It appears that four termini (Ara 4 , Ara 5 , Man 2 Ara 5 , and Ara 6 ) bear the succinates (Fig. 8B). In addition, relative succinylation (Fig. 8C) estimated from LC/ MS ion abundance analysis reveals that HN878, EAI-and IO-LAM have 45% succinylated termini (green/orange segments). Based on the LC/MS ion abundance, Figure 8D showed that EAI-LAM has approximately 10-12% less Man capping (green/yellow segment) compared with the LAMs from three other strains in this study.

Sequential enzymatic digestion
In order to unequivocally establish if the acyl substituents were on the terminal end Manp or not, LAM from each strain was first treated with α-mannosidase (Jack beans), and the residual glycan was passed through a Biogel P4 column. The 1 H-NMR of digested LAM revealed that the acyl groups and MTX/MSX residues were still present in all 4× LAM molecules (Fig. S7, A and B). TOCSY experiment on the digested RvLAM sample, chosen as the pilot because of its available amount, confirmed the presence of succinates along with MTX/MSX. However, MTX/MSX cross peaks were of relatively low intensity. The possibility of slower nonspecific digestion of MTX-Manp linkages cannot be ruled out. Major changes were noticeable in the anomeric region of HSQC NMR spectrum (Fig. S7C). The two sets of overlapping protons for β-D-Araf-(1→2)-Araf arising from Man capped and uncapped arabinan termini merged into one peak, thereby confirming almost complete removal of mannose caps, as expected. A weak residual signal for t-α-D-Manp-(1→2)-Manp supports this observation. Unlike the parent RvLAM molecule, the peak for Manp-α-(1→6) appears more intense than 2,6linked α-Manp (Fig. 9A). Although there is a collateral loss in structure, it does not deter any alteration in antibody recognition activity.
α-Mannosidase-treated LAM was subsequently digested with endoarabinanase. This allowed us to characterize the released arabinan termini (Ara 4 , Ara 5 , and Ara 6 ) bearing acyl groups and MTX, without heterogeneity arising from mannose capping. We found that monosuccinylated Ara 4 and Ara 5 were major linear species with lesser amounts of branched Ara 6 (Fig. 9B).
Abundant nonacylated Ara 5 and Ara 6 were found in the RvLAM digest compared with the other LAM species in this study.

Antigenic determinant in LAM
LAM extracted and purified from the Mtb strains were run on a 10-20% Tricine gel, followed by periodic acid-silver staining and transferred to nitrocellulose membrane for a Western Blot using mouse mAb CS35. The smear for RvLAM appeared to be running slower than the clinical isolates (Fig.  10A) indicating its larger size. As expected, CS35 mAb recognized LAM from each isolate equally well (Fig. 10A).
LAM, α-mannosidase-treated LAM, and endoarabinanasetreated enzyme resistant cores were analyzed by SDS/PAGE (Fig.  10B) followed by Western Blot using mAb CS35 (Fig. 10B1). As evident from the SDS/PAGE analyses, (Fig. 10B) there was only a shift of 2 kDa after mannosidase treatment and selective removal of Man capping did not abrogate any antibody binding activity. On the other hand, following the endoarabinanase treatment, which afforded only partial removal of the arabinan termini, the resistant core shifted down to 10-15 kDa (Fig. 10C) and lost all binding capacity to the antibody (Fig. 10C1). Compositional analyses of this material showed Ara:Man ratio of 0.9:1 indicating 57% of Ara still remaining and linkage analyses of this core by GC/MS showed some 2 Manp, 2 Araf are present along with large amounts of 5linked Araf (Fig. S8, A-C). . A, all have Ara 6 (EAI > IO > HN878 > H37Rv) and its mannose capped variations that are not succinylated. B, it appears that four termini (Ara 4 , Ara 5 , Man 2 Ara 5 , and Ara 6 ) bear majority of the succinates. C, relative overall succinylation (45%) was found to be similar in LAM (arabinan termini; Ara 4 , Ara 5 , Ara 6 ) from clinical isolates and considerably less in LAM from H37Rv (30%). D, relative overall mannose-capping of Ara 4 , Ara 5 , Ara 6 was observed in the order: H37Rv > HN878 > IO > EAI. For simplicity and relevance to the nonreducing end, we restricted our analysis for Ara 4 , Ara 5 , Ara 6 as released by the enzyme; higher arabinan (Ara 7 -Ara 9 ) was also visible in LC/MS with lower abundances.   -α(1→5)), or mannan backbone is not responsible for binding to CS35. Only Ara 4 , Ara 5 , Ara 6 specifically terminating with a Araf-β(1→2) is contributing to mAb binding.

Docking studies
The X-ray crystallographic investigation of synthetic methyl oligoarabinofuranosides (Ara 4 /Ara 6 )-CS35Fab recognition has been reported (32). The binding domain of CS35Fab and the conformational relevance of the nonreducing end of LAM (as represented by these synthetic substrates) are clearly understood. We attempted docking studies (PyRx-Vina (33) using hypothetical (MM2 energy minimized)(Ara 4 /Ara 6 ) with or without a succinyl residue at the 3-position of the 2-linked-Araf and CS35Fab. The lowest docking score (and root-meansquare deviation of atomic position; rmsd = 0) clearly showed that the binding of the succinylated Ara 4 /Ara 6 to CS35-Fab is unfavorable in comparison to nonsuccinylated Ara 4 /Ara 6 ( Fig.  11 I and Ia). It is apparent (Discovery Studio visualizer) that the succinyl residue tends to pose outward the (Fig. 11 II and IIa) binding groove of CS35Fab indicating conformational mismatch (Figs. 11 and S9).

Discussion
Current knowledge on the structure of LAM has resulted primarily from detailed studies on a few selected laboratory strains of M.tb (5,34), M. bovis bacillus Calmette-Guérin (35), M. smegmatis (3), and Mycobacterium kansasii (22). Considerable effort has been invested in correlating particular structural features with aspects of the immunopathogenesis of TB (34). An outcome of these efforts is the consensus that the mannose-caps of ManLAM constitute the single most important structural entity engaged in phagocytosis by macrophages and subsequent events such as inhibition of phagosome/lysosome fusion and immunomodulation. Although most of the studies were based on LAM isolated from a culture, a promising start has been made with providing the description of LAM from urine of a TB positive patient and Mtb infected guinea pig/mouse lung only lately (21).
In our present study, we show that there is no marked difference among the LAM preparations from TB clinical isolates HN878, EAI, IO, and the laboratory reference strain H37Rv in their glycan structure. Notably the shared features are arabinan termini motifs, glycosidic linkages, and consistent substitution with succinates. In support, the 1 H-13 C correlation spectra revealed that all 4xLAM samples have similar glycosidic linkages as reported for Mtb LAM (36,37) pointing to the presence of distinct Ara 4 , Ara 5 , and Ara 6 with and without Man capping.
The MTX/MSX substitution is common in all LAMs. MTX/ MSX is a particular substituent ever since its discovery (26,29) has been implicated in various functions such as regulatory (38), affinity to various antibodies, and as a contributor to a potent antigenicity (20). The distribution of this substitution is one/two per molecule of LAM. Therefore, in the context of antigen:(antibody) n type recognition of LAM, MTX/MSXtermini may not have significant contribution toward affinity. For EAI-LAM, the terminal β-Araf (t-Ara) appeared to be more abundant, whereas 2-linked Manp in lower abundance was observed by linkage analysis suggesting relatively lower mannose capping in EAI-LAM (Fig. S8C). Although succinylation is the most pronounced covalent modification of the LAM-arabinan, the TOCSY NMR of EAI-LAM (Fig. 3) Figure 11. Docking studies of Ara6 (I and Ia) and SucAra6 (II and IIa) with CS-35Fab (3HNS.pdb) on hydrophobic surface: Purple ribbon, Heavy Chain and Yellow ribbon: Light Chain. Best pose (lowest score/binding energy (kCal/mol) and rmsd = 0) has been presented. Succinylation at the 3-position of ring-C-Araf is unfavorable to binding compared to Ara6. Amino acid residues around Ara6 (Ia) were found to have similar orientation as reported (32). Upon succinylation, the binding conformation of the ligand (SucAra6) appeared (II and IIa) to be flipped with succinyl-residue posing outward. In the context of LAM-macromolecule, this may not be viable for binding. revealed that there are other hitherto unreported acyl functions present in lesser abundance than the succinates, and mere presence of these features could dictate the antigenicity of LAM from these strains.
The MS analyses also supported the possibility of the presence of a lactate and acetate covalently linked in the nonreducing ends of LAM. In our present work, we exploited 2D NMR on LAM, MS and MS-MS analyses of enzyme released fragments and unambiguously established that these short chain carboxylic acids occupy the 3-OH position of the 2-linked Araf (i.e., penultimate Araf). Only recently lactate and other short chain acids have been implicated to be present only as a contaminant in a LAM preparation (39). We conclusively show that these are covalently attached to the LAM-arabinan. Incidentally, the presence of lactate groups, associated with the arabinomannan domain of LAM, was also reported in M.tb, M. smegmatis and M. leprae (40).
α-Mannosidase-digested LAM was subsequently treated with endoarabinanase allowing us to characterize the released arabinan termini bearing acyl groups and MTX, without additional heterogeneity arising due to mannose capping. We found that succinyl residues mainly decorate the linear Ara 4 and Ara 5 .
To date, our understanding of the biological significance of noncarbohydrate modifications of the arabinomannan remains speculative. To address this, we performed simple experiments and tested each variant against mAb CS35.
We set out to investigate the impact of structural variations in LAM onto the mAb binding. We hypothesized three scenarios of LAM in TB patients. The case of variability in mannose capping was addressed with α-mannosidase-digestion, the impact of the lack or presence of acylation was addressed with deacylation experiment, and the importance of structural arrangements in the arabinan termini was addressed by releasing the termini by digestion with endoarabinanase.
We monitored the effect of Man cap removal by indirect ELISA whereby LAM/LAM derivatives were immobilized and subsequently detected by CS35, a well-known anti LAM murine mAb having strong binding affinity to LAM termini motifs. Earlier we reported (10) that CS35 is a broad-spectrum mAb with preferential affinity for Ara 6 . It also binds to the linear Ara 4 /Ara 5 and to a lesser extent to the (Man 2 ) 2 Ara 6 . Therefore, it is conceivable that the abundance of Ara 6 should preferentially dictate the response to CS35 as LAM-binding epitopes. If, however, Ara 6 is less abundant than (Man 2 ) 2-Ara 6 , total mAb response may equate the sum of the responses from binding to all available Ara 6 and (Man 2 ) 2 Ara 6 . The LC/ MS profile of all the four LAMs after digestion with endoarabinanase (Fig. 8A) revealed that the relative abundance of Ara 6 in HN878, EAI, IO, and H37Rv is 6.7%, 16.1%, 11.3%, and 5.3%, respectively and that of (Man 2 ) 2 Ara 6 is 12.6%, 8.4%, 10.9%, and 16.3%, respectively. It is understandable that for EAI and IO LAM, the response from CS35 will have approximately 2-fold more contribution from Ara 6 than for HN878 and H37Rv. This is what is supported by the ELISA (gray line; Fig. 12A, I-IV) as removal of Man caps does not influence the abundance of original Ara 6 . EAI and IO LAM did not show any drastic change in OD values after digestion with mannosidase (Fig. 12A, II and III). This justifies the reduced OD values for the mannosidase-digested HN878 and H37Rv LAM from their natives (blue line) by approximately 15-30%. These results indicate that the use of α-mannosidase as pretreatment reported for the detection of urinary LAM may be inconsistent (41). Ara 6 , Ara 4/5 , the major antigenic and abundant epitopes, have one common feature at the nonreducing end, i.e., β-D-Araf-(1→2)-α-D-Araf-(1→5)-α-D-Araf-(1→5)-. According to our findings, all the acyl groups as well as lactyl ether are at the 3-position of the 2-linked penultimate Araf. Steric hindrance, offered by these acyls, may not allow the acylated epitopes to bind to CS35. The docking studies indicate that the binding of succinylated Ara 4 /Ara 6 is less favored than their nonsuccinylated versions. Conformational mismatching may be responsible, thereby supporting our hypothesis. It is expected that deacylation, achieved by mild alkaline hydrolysis, should enable better availability of antigenic epitopes and consequently, enhancement of mAb-response and better sensitivity of immunoassay may be achieved ( 1 H NMR of deacylated H37Rv and EAI-LAM are presented in Fig. S12). Our initial observation by indirect ELISA using 1 μg LAM (blue line)/ deacyl LAM (orange line) and 2-fold concentration gradient of CS35 showed dramatic drop in OD values due to deacylation (Fig. 12A, I-IV). This was not unexpected because immobilization of LAM on polystyrene-ELISA plates relies on the PI anchor of LAM, and deacylation should remove these. Therefore, inefficient immobilization should lead to loss of antigen during washing and consequently, the mAb response should drop. To circumvent the antigen immobilization issue, the TB capture-ELISA (18) was employed. CS35 was used as the capture, and biotinylated A194-01 was used as the detection antibody while the LAM/deacyl-LAM concentration was varied with 2-fold dilution. Deacylated LAM from all the four strains reached signal saturation at 8-10-fold lower concentration than their native forms (Fig. 12B, I and II). These results demonstrate that the acyl groups impart some inhibitory effect towards intact LAM detection by capture ELISA, and deacylation is perhaps a rapid, cheap, and efficient way to achieve better sensitivity with the existing mAbs. However the caveat is, there is no evidence whether acyl functionalities such as mono-succinyl ester are present in LAM in biofluids.
Not unexpected, a drastic (70-80%) loss in CS35 response was observed with arabinanase resistant LAM-core compared with intact LAM (yellow line; Fig. 12A, I-IV) due to the loss of termini motifs.

Conclusion
It is evident that the structure of LAM, from different demographic origin of Mtb species, varies. Our study centered on the contribution of nonreducing arabinan termini from a diagnostic perspective. The variation in relative abundance of non-mannose-capped, mannose-capped terminus, may not have a profound impact on immunoassay sensitivity if the mAb is preferentially specific to non-mannose-capped terminus like CS35. Different acyl groups found at the 3-position of the 2-linked Araf of the antigenic epitopes partially inhibit Ab binding. Deacylation may prove to be advantageous as far as improving capture ELISA is concerned. Besides succinyl mono-ester, the presence of α-hydroxy/acetoxy butyryl ester in EAI and IO LAM is intriguing and needs further work to unveil biological implications. Importantly, when β-D-Araf-(1→2)-α-D-Araf-(1→5)-α-D-Araf-(1→5)-arrangement is lost from the nonreducing terminus, antigenic response to CS35 was compromised pointing that overall, abundance of specific arrangement of arabinan termini is crucial for the LAM-mAb recognition.
The biological significance of the structural features and their variations among strains remains to be determined. That said, the finding of covalent substituents such as succinates, methylthioxylose (MTX), and other acyl functions modifying the glycan in all M.tb strains suggests that M.tb has evolved strategies as other prokaryotes to promote their survival in its host environments. From the diagnostic perspective, the importance of specific arabinan nonreducing-end motifs, irrespective of its modifications cannot be understated. These could be the ultimate feature requirement for TB specific Ab binding.

Preparation of M.tb cell biomass
Stocks of M. tuberculosis (M.tb) were originally received from the Trudeau mycobacterial culture collection (H37Rv), the laboratory of Dr James Musser (HN878), or the laboratory of Dr Sebastian Gagneux (T17 -"IO" and 91-0079 -"EAI"). All growth conditions were performed by incubation at 37 C. Seed stocks were made by colony isolation, subculture to 7H9-OADC (100 ml), and freezing 0.8 ml samples in 20% glycerol.
Genotyping was confirmed for the seed stocks by spoligotyping. Working stocks (3 × 1 ml) were used to inoculate 7H10-OADC plates with passage of bacterial lawns to 1 L Fernbach flasks containing GAS media (42); these cultures were twice passaged to achieve 16 L cultures of each strain. Cells were harvested, washed with water, weighed, and cell biomass ranging from 70 to 80 g was inactivated by gamma irradiation (Ce137 irradiation, 2.4 MRad, adsorbed dose). Cells were lyophilized and stored at -80 C until use.

Purification of Lipoarabinomannan
LAM was extracted from lyophilized cells as reported (32). Briefly, chloroform:methanol:water (10:10:3) extraction to remove polar surface lipids, cell lysis, and detergent partition to achieve LAM enriched fraction (36). Ethanol precipitation (-20 C) produced LAM-enriched material. The precipitate was collected and resuspended at 50 mg/ml in endotoxin-free water and digested with proteinase K (0.1 mg/ml) to remove any residual proteins. After dialysis in running DI water (3500 MWCO cassette), the presence of LAM was confirmed by SDS-PAGE gel with periodic silver stain. Subsequently, tandem size-exclusion chromatography (Sephacryl-200 HR and Sephacryl-100 HR, 320 ml each) in LPS running buffer (0.2 M NaCl, 0.01 M Deoxycholic acid, 0.001 M EDTA in 0.01 M Tris, pH 8.0) was performed. Fractions (2.5 ml) were analyzed by SDS-PAGE. Pure LAM fractions were pooled and sequentially dialyzed (12,000 MWCO) against LPS dialysis buffer (minus deoxycholate), 1 M NaCl, running DI water, and endotoxinfree water. LAM was qualified by 1 H-NMR. ImageJ (43,44) was used to determine LAM concentration resolved on SDS-PAGE gel along with monosaccharide analysis.

Monosaccharide composition
LAM samples were hydrolyzed with 2 M trifluoroacetic acid. The resulting monosaccharides were converted to alditol acetates and analyzed on a TSQ 8000 Evo triple-quad GC-MS (Thermo Scientific). The mass spectrum was scanned from m/z 50 to 500, data analysis was performed using Chromeleon Chromatography data system (Thermo Scientific). Quantitation was based on 3-Omethyl -glucose, used as an internal standard (45).

NMR experiments
All NMR spectra were recorded with 5 mg LAM in D 2 O (0.56 ml) at 25 C. All PRESAT 1 H NMR was recorded on a Bruker Neo 400 MHz NMR instrument. 1 H-13 C correlation (HSQC) NMR spectra were recorded with 128 scans and relaxation delay 1.2 s on Bruker Neo 400 MHz NMR instrument. The default Bruker parameters were used for recording spectra. TOCSY ( 1 H-1 H correlation) NMR spectra were recorded with 256 scans on Varian Innova 500 MHz instrument. All chemical shifts are based on the reference to the HOD peak at 4.64 ppm. The default Varian parameters were used for recording spectra. Spectra were processed using MestReNova x64.
α -mannosidase treatment conditions LAM (1 mg/0.2 ml water) was treated with 0.085 ml (5 units) of α-Mannosidase (Jack Beans; Sigma; 3.5 mg/ml; pH 5.5) and incubated at 37 C (water bath) for 20 h. The reaction mixture was then frozen at -80 C for 2 h. After inactivation of the enzyme at 80 C for 15 min, the reaction mixture was purified over Biogel-P4 column (30 cm × 1 cm, water).
Digestion with endoarabinanase to release oligoarabinosides LAM (100-200 μg) from each strain was digested with endoarabinanase as previously described (21) and the released arabinan fragments were separated from the resistant core by nanosep 3K Omega (Pall Corporation) and centrifuging at 14,000g for 20 min at room temp. An aliquot of samples was analyzed by SDS-PAGE to ensure digestion of LAM. The flow-through containing the released oligoarabinosides was analyzed by LC/MS.

Immunoassays by Indirect and Capture ELISA
Indirect ELISA Indirect ELISA was carried out as previously described (46). LAM samples (1 μg/ml) were taken in coating buffer (0.05 M sodium carbonate and sodium bicarbonate, pH 9.6) and applied to the 96-well plate (Corning, Costar) at 4 C overnight. Wells were blocked by 1% BSA (Sigma Aldrich) in 1× PBS. Purified CS35 (5 μg/ml) was serially diluted 2-fold to derive a concentration curve and added to the wells and incubated at room temperature (90 min). The plates were then washed with 1× PBS containing 0.05% Tween-80 and incubated at room temperature (90 min) with the anti-mouse IgG alkaline phosphatase conjugated secondary antibody (1:2500 dilution). Subsequently, the alkaline phosphatase activity was measured (405 nm) after addition of p-nitrophenyl phosphate (pNPP) (Kirkegard and Perry Laboratories).

Capture ELISA
A 96-well polystyrene high binding microplate (Corning, Costar) was coated with CS35 (10 μg/ml in 1× PBS; 4 C overnight). The plates were blocked (1% BSA in 1× PBS; 60 min). LAM and derivatives (in 1× PBS at the final concentration of 12.5 ng/ml) were serially diluted 2-fold and applied in duplicate to the plate/s. After incubation (90 min, RT) and wash, detection antibody (Biotinylated A194 hu mAb; 250 ng/ml) was applied. Streptavidin-Horseradish Peroxidase (R&D Systems) (25 min) followed by Ultra TMB ELISA chromogenic substrate (Thermo Scientific) were added to develop color. The reaction was stopped by adding 2 M sulfuric acid, and the optical density was measured at 450 nm.

Conditions for LC/MS
Enzyme-released oligosaccharides were subjected to ultraperformance liquid chromatography (UPLC) separation on a Waters Acquity UPLC H-Class system in line with a Bruker MaXis Plus quadrupole time-of-flight (QTOF) mass spectrometer (MS). Separation was performed in gradient mode with a Waters Atlantis T3 3.0 μm column (2.1 × 150 mm) at 40 C. Mobile-phase components were 10 mM ammonium acetate in water (A) and 10 mM ammonium acetate in acetonitrile (B). The flow rate was 0.3 ml/min. The data were acquired in the negative electrospray ion (ESI) mode with a mass-tocharge ratio (m/z) range of 110−4000 at 1 Hz scan rate. For MS-MS experiments, the MS was in multiple reaction monitoring (MRM) scan mode with collision-induced dissociation (CID) energies of 40 eV and 60 eV on target masses with m/z width of 10 Da. Internal instrument mass-scale calibration was performed by infusing the Agilent ESI-L low-concentration tuning mix. Instrument controls were performed via the Bruker HyStar v4.1 software package. Data were processed using Bruker Compass 2.0 Data Analysis 4.4 software.

Docking studies
The methyl oligoarabinofuranosides with proper stereochemistry were drawn using Chemdraw 20.0 and 3D mm2 energy minimization was done. The protein structures of CS35Fab (3HNS and 3HNT) were obtained from Protein Data Bank. Docking studies were performed (3HNS with Ara 6 / SucAra 6 and 3HNT with Ara 4 /SucAra 4 ) on PyRx-Vina platform. The.pdbqt files were further analyzed using Discoverystudio-visualizer.

Data availability
All data are contained within the manuscript, either in the main body or in the supporting information submitted with the manuscript Supporting information-This article contains supporting information. spectra in the CSU core facilities respectively. We also acknowledge Dr Abraham Pinter from Rutgers University for providing mAb 194-01.