Epitope Recognition of Antibodies against a Yersinia pestis

Jan 30, 2014 - Today, the process of selecting carbohydrate antigens as a basis for active vaccination and the generation of antibodies for therapeuti...
0 downloads 0 Views 2MB Size
Download PDF
Letters pubs.acs.org/acschemicalbiology

Epitope Recognition of Antibodies against a Yersinia pestis Lipopolysaccharide Trisaccharide Component Felix Broecker,†,‡ Jonas Aretz,†,‡,§ You Yang,†,§ Jonas Hanske,†,‡,§ Xiaoqiang Guo,†,§ Anika Reinhardt,†,‡ Annette Wahlbrink,† Christoph Rademacher,† Chakkumkal Anish,*,† and Peter H. Seeberger*,†,‡ †

Max Planck Institute of Colloids and Interfaces, 14424 Potsdam, Germany Institute of Chemistry and Biochemistry, Freie Universität Berlin, 14195 Berlin, Germany

‡

S Supporting Information *

ABSTRACT: Today, the process of selecting carbohydrate antigens as a basis for active vaccination and the generation of antibodies for therapeutic and diagnostic purposes is based on intuition combined with trial and error experiments. In efforts to establish a rational process for glycan epitope selection, we employed glycan array screening, surface plasmon resonance, and saturation transfer difference (STD)-NMR to elucidate the interactions between antibodies and glycans representing the Yersinia pestis lipopolysaccharide (LPS). A trisaccharide epitope of the LPS inner core glycan and different LPS-derived oligosaccharides from various Gram-negative bacteria were analyzed using this combination of techniques. The antibodyglycan interaction with a heptose substructure was determined at atomic-level detail. Antibodies specifically recognize the Y. pestis trisaccharide and some substructures with high affinity and specificity. No significant binding to LPS glycans from other bacteria was observed, which suggests that the epitopes for just one particular bacterial species can be identified. On the basis of these results we are beginning to understand the rules for structurebased design and selection of carbohydrate antigens.

A

specifically recognize LPS of Y. pestis and are suitable for the specific detection of Y. pestis bacteria, since they bind only weakly to LPS of other Gram-negative bacteria, such as Escherichia coli, Neisseria meningitidis, and Salmonella typhi.4 As a first step toward the rational design of carbohydrate antigens for developing high-affinity antibodies, we define the carbohydrate−antibody interactions by employing a combination of synthetic glycan array screening, SPR analysis, and STDNMR spectroscopy. We define epitope recognition patterns, binding affinities, and carbohydrate−antibody interactions on the molecular level. Monoclonal antibodies (mAbs) directed against the triheptose motif of the Y. pestis LPS inner core structure [L-α-D-Hepp-(1→ 7)-L-α-D-Hepp-(1→3)-L-α-D-Hepp] (where Hep is L-glycero-Dmanno-heptose) were generated as described previously.4 Three mAbs expressed by clones 1B7, 1E12, and 3C11 were purified and subjected to denaturing SDS-PAGE analysis, showing characteristic bands of the heavy and light chains and no significant protein impurities (Supplementary Figure S1). Glycan array analysis helped us to uncover the structural and chemical elements that determine the specificity and selectivity of LPS recognition by these antibodies. A custom glycan array

ntibodies against cell-surface carbohydrate antigens hold great promise for therapy and diagnosis of a variety of human diseases. Cancer diagnosis1 and the detection of biowarfare agents, such as bacteria that cause anthrax2,3 or plague,4 are just some examples. Carbohydrate-based vaccines confer protection against a number of infectious diseases.5 Vaccines relying on isolated bacterial polysaccharides have been successfully marketed, including those against Haemophilus inf luenzae type B, Neisseria meningitidis, and Streptococcus pneumoniae.5 Still, the exact features of a glycan that gives rise to a good antibody response regarding epitope recognition and cross-reactivity are poorly understood. Consequently, the process of selecting carbohydrate antigens to generate highly selective antibodies against a certain pathogen is not a straightforward process.6,7 Defining the best antigens is a crucial step in developing highly specific anti-carbohydrate antibodies for diagnostic and therapeutic applications. Recent advances in the chemical synthesis of complex carbohydrates allow for access to structurally defined oligosaccharide antigens relevant for human diseases.8,9 Rational design based on synthetic oligosaccharide antigens remains to be demonstrated for the generation of pathogen-specific high-affinity antibodies. Synthetic oligosaccharides serve as tools to dissect carbohydrate−antibody interactions at the molecular level.7 Recently, we generated antibodies against the triheptose motif of the LPS inner core of Yersinia pestis.4 These antibodies © 2014 American Chemical Society

Received: December 18, 2013 Accepted: January 30, 2014 Published: January 30, 2014 867

dx.doi.org/10.1021/cb400925k | ACS Chem. Biol. 2014, 9, 867−873

ACS Chemical Biology

Letters

Figure 1. Epitope mapping of monoclonal antibody 1B7 using glycan arrays. (a) Overview of synthetic glycan epitopes employed in this study. Heptagons represent heptose, blue squares GlcNAc, yellow circles galactose, and green circles mannose residues. (b) Exemplary microarray scan of 1B7 incubated on the glycan array at 10 μg/mL. The spotting pattern is shown to the left of the scan. Large circles indicate glycans spotted at 1 mM, and small circles glycans spotted at 0.1 mM. Proteins (CRM197 and Spacer dummy) were spotted at 1 μM. For a complete scan of this microarray, refer to Supplementary Figure S2. The spacer dummy is a GlcNAc-BSA conjugate that contains the immunogenic spacer moiety composed of pentyl and adipoyl groups. (c) Quantification of binding signals to glycan epitopes 1−3 inferred by glycan array. Bars show mean + SD of four spots (1 mM).

glycan was equipped with an aminopentyl linker at the reducing end, allowing for orientation-specific immobilization on the microarray surface. As controls, the array included the diphtheria toxin variant CRM197, a carrier protein commonly employed to increase immunogenicity of glycan antigens in conjugate vaccines, which we used to generate antibodies to 3 and a GlcNAc-BSA dummy conjugate representing the generic spacer moiety generated during glycoconjugate synthesis4 as well as coupling buffer only (Figure 1b). Successful spotting of all compounds was verified as described in the Methods section. Purified mAbs at concentrations ranging from 0.01 to 50 μg/mL were subjected to glycan array analysis. All three mAbs specifically recognized triheptose 3, its substructure diheptose 2, and to a lesser extent, heptose 1 (Figure 1c and Supplementary Figure S2), while no significant binding was observed to any other LPS structures present on the array or any of the controls. Although we have qualitatively confirmed spotting for all compounds, we cannot rule out the possibility that charged oligosaccharides may have been spotted with lower efficiency, resulting in lower binding signals. Binding quantification expressed as mean fluorescence intensity, MFI, revealed that 2 and 3 are bound approximately at the same level (for 1B7 at 10 μg/mL, 3708 ± 843 MFI, mean ± SD of n = 4 microarray spots, and 4018 ± 404 MFI, respectively), while binding to 1 was weaker (1661 ± 157 MFI) (Figure 1c).

was prepared using synthetic oligosaccharides ranging from mono- to tetrasaccharides that reflect the LPS inner core structures of different Gram-negative bacteria (Figure 1a).4,10,11 In addition to the Y. pestis inner core tetrasaccharide12−14 L-α-DHepp-(1→7)-L-α-D-Hepp-(1→3)-L-α-D-Hepp-(1→5)-Kdo 10 (where Kdo is 3-deoxy-α-D-manno-oct-2-ulosonic acid), and its substructures, the triheptose hapten used to generate mAbs 1B7, 1E12, and 3C11, L-α-D-Hepp-(1→7)-L-α-D-Hepp-(1→3)L-α-D-Hepp 3, diheptose L-α-D-Hepp-(1→3)-L-α-D-Hepp 2 (unpublished data) and heptose L-α-D-Hepp 1, the inner core tetrasaccharide specific for N. meningitidis15 α-D-GlcNAc-(1→ 2)-L-α-D-Hepp-(1→3)-L-α-D-Hepp-(1→5)-α-Kdo 4 and its substructures, trisaccharide L-α-D-Hepp-(1→3)-L-α-D-Hepp(1→5)-α-Kdo 7, disaccharide L-α-D-Hepp-(1→5)-Kdo 9, and trisaccharide α-D-GlcNAc-(1→2)-L-α-D-Hepp-(1→3)-L-α-DHepp 11 (unpublished data) were spotted on the array. Trisaccharide 7 is common to most Gram-negative bacteria.16 The array also included the conserved core tetrasaccharide of H. inf luenzae17−19 L-α-D-Hepp-(1→2)-L-α-D-Hepp-(1→3)-L-αD-Hepp-(1→5)-α-Kdo 6, the tetrasaccharide L-α-D-Hepp-(1→ 3)-L-α-D-Hepp-(1→5)-[β-L-Ara4N-(1→8)]-α-Kdo 8 common to most Proteus strains,20,21 the tri-Kdo α-Kdo-(2→8)-α-Kdo(2→4)-β-Kdo 5 and Kdo 12. The capping tetrasaccharide of the Leishmania lipophosphoglycan22−24 13 was included on the array as a nonrelated control oligosaccharide hapten. Each 868

dx.doi.org/10.1021/cb400925k | ACS Chem. Biol. 2014, 9, 867−873

ACS Chemical Biology

Letters

Figure 2. Kinetic parameters of mAb 1B7 binding to different glycan epitopes by SPR. Representative, reference-subtracted sensorgrams representing binding kinetics to epitopes 1−3, 7, and 10 are shown. When applicable, a 1:1 binding model was used to determine affinities and rate constants; respective fitted curves (black) are overlaid. Where kinetic determinants kon and/or koff could not be measured, respective affinity plots are shown to the right of the sensorgrams. Numbers embedded in the sensorgrams indicate the respective concentrations of the glycans (analytes) in μM. KD and theoretical Rmax values are shown within the plots.

Apparently, the larger epitopes are preferentially bound by mAbs generated using 3. Interestingly, the Y. pestis inner core tetrasaccharide 10 that contains 3 but carries an extra Kdo moiety at the reducing end was not recognized on the glycan array by any of the mAbs (Figure 1b and Supplementary Figure S2). Likewise, the Kdo-containing haptens 7 and 8 that contain the diheptose motif found in 2 are not bound significantly by the mAbs. The presence of a terminal Kdo moiety, perhaps by introducing a charge by virtue of its carboxyl group, limits binding of the antibodies specific for 3. It has been noted previously that charged moieties within carbohydrates can be crucial determinants for antibody recognition.25 In addition, no binding was observed to 11 that has just one additional GlcNAc moiety at the nonreducing terminus, indicating that GlcNAc may mask the diheptose motif in this structure on the glycan array. The results indicate that the antibodies selectively bind glycan epitopes with at least one heptose residue at the nonreducing end and no terminal Kdo residue. To expand the results obtained by glycan array analysis, we determined binding specificities and kinetics by surface

plasmon resonance (SPR) experiments. Briefly, a CM5 sensor chip was immobilized with about 10,000 response units (RUs) of α-mouse IgG antibody, which was used to capture the mAbs. Then, oligosaccharide haptens were flowed over this chip and response unit (RU) changes were monitored. This setup allowed for the determination of binding specificities as well as the binding affinities expressed as dissociation constants (KD) that can be divided into association and dissociation rates (kon and koff, respectively). Binding to haptens 1−3 was verified. Representative sensorgrams of mAb 1B7 are shown in Figure 2 and those of 1E12 and 3C11 in Supplementary Figures S3 and S4, respectively. All affinities and rate constants are summarized in Table 1. Binding to heptose 1 by 1B7 was characterized by fast association and dissociation rates, not allowing for determination of the rate constants (Figure 2a). When using a steady-state affinity model, the KD value was determined to be 71 μM. KD values of 1E12 and 3C11 were in the same range (121 and 143 μM, respectively), and binding was also characterized by fast association and dissociation rates (Supplementary Figures S3a, S4a and Table 1). In contrast, 869

dx.doi.org/10.1021/cb400925k | ACS Chem. Biol. 2014, 9, 867−873

ACS Chemical Biology

Letters

Table 1. Affinity Parameters of Monoclonal Antibodies to Different Glycan Epitopes Inferred by SPRa mAb 1B7

1E12

3C11

epitope/analyte 1 2 3 7 10 1 2 3 7 10 1 2 3 7 10

kon [M‑1 s‑1] b

fast 7.8 (±0.2) × 104 12.0 (±2.1) × 104 slowc slowc fastb 11.5 (±5.1) × 104 14.5 (±3.8) × 104 slowc slowc fastb 5.8 (±4.3) × 104 13.5 (±1.0) × 104 slowc slowc

koff [s‑1]

KD

fast 266 (±15) × 10−4 5.6 (±1.4) × 10−4 nd nd fastb 283 (±124) × 10−4 6.7 (±0.1) × 10−4 nd nd fastb 107 (±100) × 10−4 6.3 (±3.2) × 10−4 nd nd

71 (±8) μM 342 (±29) nM 4.7 (±0.4) nM ≫256 μM ≫128 μM 121 (±26) μM 246 (±2) nM 4.6 (±0.6) nM ≫256 μM ≫128 μM 143 (±17) μM 164 (±50) nM 4.6 (±2.0) nM ≫256 μM 47.7 (±12.9) μM

b

a

Values given in brackets are standard deviations of two independent measurements. mAb, monoclonal antibody; nd, not determined (insufficient fitting of the 1:1 binding model). bAssociation or dissociation rates were too fast to be measured by the instrument. cAssociation rates were too slow to be measured by the instrument.

Figure 3. Epitope mapping of the heptose 1−3C11 interaction by STD-NMR spectroscopy. (a) Binding epitope of heptose 1 as indicated by the normalized STD effect displayed in the color code. (b) STD build-up curves of the carbon-bound protons according to numbering in left panel. STD amplification factors were obtained from STD-NMR spectra of 200 μM 1 in the presence of 2 μM antibody obtained at saturation transfer delays of 1, 2, 4, and 4 s on a 600 MHz spectrometer. STD effect was derived from initial slopes of fitted STD-AF function and normalized to the proton with maximum STD effect.

kinetic models to fit the data, the 1:1 binding Langmuir model and a two-state reaction model, which assumes a conformational change of the antibody−analyte complex. Both models yielded similar good fittings, and we chose the 1:1 binding model to determine the kinetic parameters, as this model makes

the binding to diheptose hapten 2 by all three antibodies was characterized by slower association and dissociation rates and consequently were in the measurable range of the instrument (Figure 2b and Supplementary Figures S3b, S4b). To determine KD, kon, and koff values, we evaluated two different 870

dx.doi.org/10.1021/cb400925k | ACS Chem. Biol. 2014, 9, 867−873

ACS Chemical Biology

Letters

fewer assumptions. For hapten 2, kon values ranging from 5.8 (3C11) to 11.5 × 104 M−1 s−1 (1E12) and koff values ranging from 107 (3C11) to 283 × 10−4 s−1 (1E12) were determined. The calculated KD values of all three mAbs to 2 were lower than those to 1, ranging from 164 nM (3C11) to 342 nM (1B7), indicating higher affinities to this larger substructure of 3 and confirming the microarray data that showed higher binding signals to 2 when compared with 1. Binding of mAbs to hapten 3 was characterized by kon rates in the same order of magnitude as to hapten 2, ranging from 12.0 (1B7) to 14.9 × 104 M−1 s−1 (1E12). In contrast, koff values were lower, ranging from 5.6 (1B7) to 6.7 × 10−4 s−1 (1E12), which resulted in KD values in the low nanomolar range (4.6−5.2 nM) (Figure 2c, Supplementary Figures S3c, S4c). Kdo-containing haptens were not bound at detectable levels as judged by glycan array analysis, although most of these glycans contain structural motifs also found in haptens 1−3 (Figure 1, Supplementary Figure S2). To verify these findings, we subjected glycan 7 that contains the structural motif of 2 at the nonreducing end as well as 10 containing the structural motif of 3 at the nonreducing end to SPR analysis. Dosedependent binding of 7 was observed for all three mAbs (Figure 2d, Supplementary Figures S3d, S4d). However, compared to 2, the association rates for all three antibodies were lower, below the detection limit of the instrument, indicating markedly lower affinity to this hapten. Consequently, when applying a steady-state affinity model, KD values could not be determined when using concentrations even up to 256 μM of 7, indicating KD values probably in the high micromolar or even millimolar range. Low affinity binding was likely too weak to be detectable by glycan array experiments where low-affinity antibodies might be eluted in the washing steps, a concern that has been raised previously.26 Likewise, dose-dependent binding to glycan 10 was detected by SPR, but the association rates were significantly lower compared with 3 and below the measurable range of the instrument (Figure 2e, Supplementary Figures S3e, S4e). Steady-state affinity models yielded KD values likely in the high micromolar range (KD > 128 μM) for mAbs 1B7 and 1E12 and 47.7 μM for 3C11. mAb 3C11 bound better to the natural LPS antigen present on Y. pestis bacteria than 1B7 and 1E12 (data not shown). This can be explained by the comparably high affinity of 3C11 to 10, representing the LPS inner core structure of this bacterium. While antibody binding to hapten 1 was observed using the glycan array and the affinities were even lower for all three mAbs, no significant binding was observed for 3C11 binding to 10 on the glycan array (Supplementary Figure S2). The negative charge imposed by the carboxyl group of the Kdo moiety is counteracted by the free amino group in 10 that is primarily present as a zwitterionic species in the near-neutral pH 7.4 of the SPR running buffer. However, when this glycan is immobilized on the array surface via this amino group, 10 is likely present mainly as a negatively charged species. This observation gives further credence to the notion that the charge imposed by Kdo moieties reduced binding of the herein investigated mAbs. Additional studies, for instance, by using Nacetylated derivatives of 7 or 10 for SPR measurements, are needed to confirm this hypothesis. To further investigate binding characteristics, we examined the interactions between mAb 3C11 and glycans 1−3 by saturation transfer difference (STD)-NMR. The dissociation rates of the 3C11-2 and -3 complexes were too low for any detectable STD effects (Supplementary Figure S5). The

complex of 3C11 with 1, which was characterized by a fast dissociation rate was a suitable candidate for further analysis by STD-NMR. STD effects of this particular interaction were quantified by determining the initial slopes of build-up curves employing a series of saturation times (1, 2, 4, and 6 s), and the highest STD effect was set to 100%, as described27 (Figure 3). Assessment of antibody binding at a 100:1 ratio of carbohydrate ligand to protein revealed that the contact surface area of 1 with 3C11 was mainly located around H6 (100% STD effect) and H2 (87%) and to a lesser extent around H3 and H4 (84% and 85%, respectively), while contributions of H5 and H71,2 (51% and 72%, respectively) were comparably weak. The STD effect around the anomeric proton, H1, could not be determined as the corresponding signals overlapped with the solvent signal (Supplementary Figure S5). Interactions between the unnatural aminopentyl linker at the reducing end (protons a−e) and 3C11 were comparably low except for He,1,2, the most distant to the reducing end of the heptose sugar with an STD effect of 79%. This might be due to nonspecific interactions with this distant part of the aminopentyl linker to the binding surface of 3C11. Overall, STD-NMR data indicated that the binding surface of 3C11 recognized 1 mainly around C2−C6 of the heptose sugar. Strong binding around C2 might be attributed to the axial conformation of the hydroxyl group at that position that is found in mannose as well. Moreover, C6, exhibiting a strong STD effect, is part of the heptose side chain that is not found in any mammalian glycans. Binding of 3C11 is mediated through structural motifs found in bacterial but not mammalian carbohydrates. Neither 3C11, nor the two other mAbs we investigated bind to the mannose-containing glycan 13 in glycan array experiments (Figure 1 and Supplementary Figure S2). Mannose and heptose differ only in the C7-OH. The binding surface of 3C11 to the di- and triheptoses 2 and 3 was further examined by competition STD-NMR experiments with equimolar amounts of 1 and 2 or 1 and 3, respectively. Addition of 2 or 3 to 1 completely diminished any STD effect imposed by 1 (Supplementary Figure S5), thus showing that the binding site of 3C11 for 1 is identical to that of 2 and 3, and at the same time confirming the tight binding of 2 and 3 to 3C11 observed using glycan array (Figure 1 and Supplementary Figure S2) as well as SPR measurements (Figure 2, Supplementary Figures S3, S4). Stronger binding to oligoheptoses 2 and 3 compared with 1 might be attributed to a higher avidity of antibodies binding to heptose presented in a di- (2) and trivalent (3) fashion; however, more complex structural motifs perhaps accounting for higher affinities to the oligoheptoses presented by 2 and 3 cannot be ruled out. In summary, by combining glycan arrays, SPR, and STDNMR spectroscopy as three complementary techniques, the interactions between three monoclonal antibodies and the corresponding glycan antigens provided atomic-level detail of antibody−carbohydrate interactions. Glycan arrays allowed for semiquantitative mapping of epitope recognition in a highthroughput manner. Binding candidates were further verified by SPR that yielded also kinetic parameters such as affinities and association and dissociation rates. STD-NMR spectroscopy provided detailed information about the interaction surfaces between antibodies and carbohydrate antigens only for antibody−carbohydrate interactions of relatively low affinity. mAbs generated against triheptose hapten 3 specifically recognized 3 and its substructures diheptose 2 and heptose 1, with decreasing affinity from larger to smaller epitopes. Strikingly, any related inner core LPS structures do not or 871

dx.doi.org/10.1021/cb400925k | ACS Chem. Biol. 2014, 9, 867−873

ACS Chemical Biology

Letters

(2) Tamborrini, M., Werz, D. B., Frey, J., Pluschke, G., and Seeberger, P. H. (2006) Anti-carbohydrate antibodies for the detection of anthrax spores. Angew. Chem., Int. Ed. 45, 6581−6582. (3) Tamborrini, M., Holzer, M., Seeberger, P. H., Schürch, N., and Pluschke, G. (2010) Anthrax spore detection by a luminex assay based on monoclonal antibodies that recognize anthrose-containing oligosaccharides. Clin. Vaccine Immunol. 17, 1446−1451. (4) Anish, C., Guo, X., Wahlbrink, A., and Seeberger, P. H. (2013) Plague detection by anti-carbohydrate antibodies. Angew. Chem., Int. Ed. 52, 9524−9528. (5) Astronomo, R. D., and Burton, D. R. (2010) Carbohydrate vaccines: developing sweet solutions to sticky situations? Nat. Rev. Drug Discovery 9, 308−324. (6) Serruto, D., and Rappuoli, R. (2006) Post-genomic vaccine development. FEBS Lett. 580, 2985−2992. (7) Oberli, M. A., Tamborrini, M., Tsai, Y. H., Werz, D. B., Horlacher, T., Adibekian, A., Gauss, D., Möller, H. M., Pluschke, G., and Seeberger, P. H. (2010) Molecular analysis of carbohydrateantibody interactions: case study using a Bacillus anthracis tetrasaccharide. J. Am. Chem. Soc. 132, 10239−10241. (8) Hecht, M. L., Stallforth, P., Silva, D. V., Adibekian, A., and Seeberger, P. H. (2009) Recent advances in carbohydrate-based vaccines. Curr. Opin. Chem. Biol. 13, 354−359. (9) Adamo, R., Nilo, A., Castagner, B., Boutureira, O., Berti, F., and Bernardes, G. J. L. (2013) Synthetically defined glycoprotein vaccines: current status and future directions. Chem. Sci. 4, 2995−3008. (10) Yang, Y., Martin, C. E., and Seeberger, P. H. (2012) Total synthesis of the core tetrasaccharide of Neisseria meningitidis lipopolysaccharide, a potential vaccine candidate for meningococcal diseases. Chem. Sci. 3, 896−899. (11) Yang, Y., Oishi, S., Martin, C. E., and Seeberger, P. H. (2013) Diversity-oriented synthesis of inner core oligosaccharides of the lipopolysaccharide of pathogenic Gram-negative bacteria. J. Am. Chem. Soc. 135, 6262−6271. (12) Holst, O. (2007) The structures of core regions from enterobacterial lipopolysaccharides - an update. FEMS Microbiol. Lett. 271, 3−11. (13) Pohanka, M., and Skládal, P. (2009) Bacillus anthracis, Francisella tularensis and Yersinia pestis. The most important bacterial warfare agents - review. Folia Microbiol. (Praha) 54, 263−272. (14) Vinogradov, E. V., Lindner, B., Kocharova, N. A., Senchenkova, S. N., Shashkov, A. S., Knirel, Y. A., Holst, O., Gremyakova, T. A., Shaikhutdinova, R. Z., and Anisimov, A. P. (2002) The core structure of the lipopolysaccharide from the causative agent of plague, Yersinia pestis. Carbohydr. Res. 337, 775−777. (15) Kahler, C. M., and Stephens, D. S. (1998) Genetic basis for biosynthesis, structure, and function of meningococcal lipooligosaccharide (endotoxin). Crit. Rev. Microbiol. 24, 281−334. (16) Holst, O. Chemical structure of the core region of lipopolysaccharides. In Endotoxin in health and disease (1999) Brade, H., Opal, S. M., Vogel, S. N., and Morrison, D. C., Eds., pp 115−154, Marcel Dekker, New York. (17) Masoud, H., Moxon, E. R., Martin, A., Krajcarski, D., and Richards, J. C. (1997) Structure of the variable and conserved lipopolysaccharide oligosaccharide epitopes expressed by Haemophilus influenzae serotype b strain Eagan. Biochemistry 36, 2091−2103. (18) Turk, D. C. (1984) The pathogenicity of Haemophilus influenzae. J. Med. Microbiol. 18, 1−16. (19) Murphy, T. F., and Apicella, M. A. (1987) Nontypable Haemophilus influenzae: a review of clinical aspects, surface antigens, and the human immune response to infection. Rev. Infect. Dis. 9, 1−15. (20) Vinogradov, E., and Perry, M. B. (2000) Structural analysis of the core region of lipopolysaccharides from Proteus mirabilis serotypes O6, O48 and O57. Eur. J. Biochem. 267, 2439−2446. (21) Vinion-Dubiel, A. D., and Goldberg, J. B. (2003) Lipopolysaccharide of Burkholderia cepacia complex. J. Endotoxin Res. 9, 201−213.

only weakly interact with the antibodies, including the Y. pestis core tetrasaccharide 10. Thus, the presence of a Kdo moiety limits antibody binding, probably due to the introduction of a charge imposed by the carboxylic group of Kdo. All of the antibodies we studied recognized the natural antigen present on Y. pestis bacteria and LPS derivative of this bacterium.4 Notably, 3C11 that binds the Y. pestis core tetrasaccharide 10 with the highest affinity also binds Y. pestis bacteria strongest (data not shown). Even though affinity to tetrasaccharide 10 was shown to be relatively low, antibodies bind to the natural antigen present on bacteria. The dense, multivalent presentation of LPS on the bacterial surface of ∼60,000 LPS molecules per μm2 may be responsible for this observation.28 Significantly lower or negligible binding of these antibodies to related Gram-negative bacteria is in line with the failure of the mAbs to bind any LPS structures from other bacteria on the glycan array (Figure 1 and Supplementary Figure S2). This study elucidates the binding characteristics of antibodies that are currently being developed for a highly sensitive detection system for Y. pestis. In a broader sense, understanding the structural features of antibody− carbohydrate interactions will enable the design of synthetic carbohydrate antigens for antibody development as well as the selection of antibodies with desired specificities.

■

METHODS

■

ASSOCIATED CONTENT

Details of the purification of monoclonal antibodies, the preparation of microarrays, microarray binding assays, surface plasmon resonance, and saturation transfer difference-NMR are provided in the Supporting Informations. S Supporting Information *

Details of the purification of monoclonal antibodies, the preparation of microarrays, microarray binding assays, surface plasmon resonance and saturation transfer difference-NMR, as well as supplemental figures. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected]. *E-mail: [email protected]. Author Contributions §

These authors contributed equally to this work.

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We thank the Max Planck Society, the Körber Foundation and the German Federal Ministry of Education and Research (grant no. 0315447) for generous financial support. C.R. is supported by the German Research Foundation through an Emmy Noether fellowship (RA1944/2-1) and the Max Planck Society. J.H. thanks the Fonds der Chemischen Industrie for a stipend. A.R. thanks the Studienstiftung des Deutschen Volkes for a fellowship.

■

REFERENCES

(1) Wang, C. C., Huang, Y. L., Ren, C. T., Lin, C. W., Hung, J. T., Yu, J. C., Yu, A. L., Wu, C. Y., and Wong, C. H. (2008) Glycan microarray of Globo H and related structures for quantitative analysis of breast cancer. Proc. Natl. Acad. Sci. U.S.A. 105, 11661−11666. 872

dx.doi.org/10.1021/cb400925k | ACS Chem. Biol. 2014, 9, 867−873

ACS Chemical Biology

Letters

(22) Hewitt, M. C., and Seeberger, P. H. (2001) Automated solidphase synthesis of a branched Leishmania cap tetrasaccharide. Org. Lett. 3, 3699−3702. (23) Hewitt, M. C., and Seeberger, P. H. (2001) Solution and solidsupport synthesis of a potential leishmaniasis carbohydrate vaccine. J. Org. Chem. 66, 4233−4243. (24) Anish, C., Martin, C. E., Wahlbrink, A., Bogdan, C., Ntais, P., Antoniou, M., and Seeberger, P. H. (2013) Immunogenicity and diagnostic potential of synthetic antigenic cell surface glycans of leishmania. ACS Chem. Biol. 8, 2412−2422. (25) Schoonbroodt, S., Steukers, M., Viswanathan, M., Frans, N., Timmermans, M., Wehnert, A., Nguyen, M., Ladner, R. C., and Hoet, R. M. (2008) Engineering antibody heavy chain CDR3 to create a phage display Fab library rich in antibodies that bind charged carbohydrates. J. Immunol. 181, 6213−6221. (26) Rillahan, C. D., and Paulson, J. C. (2011) Glycan microarrays for decoding the glycome. Annu. Rev. Biochem. 80, 797−823. (27) Mayer, M., and James, T. L. (2004) NMR-based characterization of phenothiazines as a RNA binding scaffold. J. Am. Chem. Soc. 126, 4453−4460. (28) Troelstra, A., de Graaf-Miltenburg, L. A., van Bommel, T., Verhoef, J., Van Kessel, K. P., and Van Strijp, J. A. (1999) Lipopolysaccharide-coated erythrocytes activate human neutrophils via CD14 while subsequent binding is through CD11b/CD18. J. Immunol. 162, 4220−4225.

873

dx.doi.org/10.1021/cb400925k | ACS Chem. Biol. 2014, 9, 867−873