Unexpected Diversity in the Fine Specificity of Monoclonal Antibodies That Use the Same V Region Gene to Glucuronoxylomannan of Cryptococcus neoformans¹

Cryptococcosis is predominantly a disease of immunocompromised individuals, and those with impaired cellular immunity are at particularly increased risk (1). Cryptococcus neoformans is surrounded by a polysaccharide capsule, a type II T-independent Ag (2, 3), which is being investigated for vaccination purposes (4, 5, 6, 7), because it can induce protective Abs. Furthermore, peptide mimotopes of the capsular polysaccharide can elicit capsule-binding mAbs that are protective (8).

The main component of the capsule is a large molecular mass polysaccharide called glucuronoxylomannan (GXM),³ which composes ∼88% of the capsule (9). The reactivity with absorbed rabbit sera to antigenic determinants in GXM have defined five serotypes (A, B, C, D, and AD) (10). GXM consists of repetitive units comprised of mannan trimers substituted with single glucuronic acid and xylose sugars, with the main compositional difference between serotypes involving the extent of xylosylation and the type of linkage between xylose and the mannan backbone (11). GXMs are also modified by the extent and location of O-acetyl groups (12, 13). O-Acetylation is required for the binding of numerous mAbs to GXM (6, 14, 15). The Ab response to GXM has been shown to be highly restricted in V region usage. The majority of mAbs are constructed using a V region combination consisting of V_H7183, J_H2, V_k5.1, and J_k1 (16). mAbs to GXM that use this V region configuration are known as class II Abs (17).

Attempts to understand Ab recognition of GXM have been limited by an inability to precisely define the motifs that constitute their respective epitopes and an uncertainty on the number of different epitopes present on the polysaccharide. Unlike proteins, comparable techniques such as deletion analysis and peptide scanning are unavailable for polysaccharide Ags. Because polysaccharides and Abs are large heterodisperse molecules, neither crystallography nor nuclear magnetic resonance (NMR) is a feasible option for solving the structure of Ag-Ab complexes. Peptide mimetics have been successfully used to discriminate among closely related Abs to polysaccharides, but this technology cannot identify the specific carbohydrate structure recognized by the Ab and, therefore, cannot predict the compositional similarities of the actual carbohydrate epitopes. Currently, the only means to identify a carbohydrate epitope is to chemically synthesize an oligosaccharide or to digest the polysaccharide into small units and demonstrate that the carbohydrates bind at the Ag-combining site of the Ab. Unfortunately, these techniques either have not been successful so far or are presently unavailable for the study of GXM.

Given these limitations, an alternative approach to studying the fine specificity of GXM-binding mAbs was undertaken, such that epitopes and paratopes were examined concomitantly. In this study, we approached fine specificity mapping of polysaccharide-binding Abs by characterizing their reactivity with defined C. neoformans capsular mutants. The analysis of mAb binding with mutant cells revealed differences in mAb fine specificity that were not distinguishable by other methods. As a consequence of this study, it is now apparent that Abs to GXM that are derived from a restricted response that uses the V_H7183, J_H2, V_k5.1, and J_k1 gene families recognize unique epitopes, and for one Ab, the importance of epitope conformation was established.

The mAbs used in these studies originate from hybridomas generated from mice immunized with GXM-tetanus toxoid conjugate and have been described elsewhere. mAb 2H1 and 18B7 are IgG1 (6), whereas mAbs 12A1, 13F1, and 21D2 are IgMs (6, 16, 18). The isotypes of other Abs used in this study are listed within Results.

A 2-day-old culture of C. neoformans strain B-3501, grown in Sabouraud dextrose broth, was diluted to 8 × 10⁵ organisms per milliliter in Sabouraud dextrose broth and grown for 6 h at 30°C with a rotation speed of 150 rpm. B-3501 is a serotype D strain that was chosen for this study because all mAbs to be analyzed were known to bind this serotype, and strain B-3501 has been the predominant strain used in C. neoformans genetic studies. Yeast cells were collected by centrifugation at 320 × g, for 10 min at room temperature, washed twice in PBS (137 mM NaCl, 2.7 mM KCl, 1.5 mM KH₂PO₄, 8.5 mM Na₂HPO₄), and then resuspended in PBS to a final density of 1 × 10⁷ organisms per milliliter. A volume of 5 ml of this suspension was added to a 60-mm petri dish for mutagenesis. Mutagenesis was performed in a Stratalinker 1800 equipped with an 8-W bulb emitting UV light at a wavelength of 254 nm. To ensure consistent irradiation of the samples, the lamps were prewarmed by a 2-min run. Organisms were irradiated at various energies (15,000, 30,000, or 50,000 μJ), diluted in PBS, and plated onto Sabouraud dextrose agar. Plates were incubated at 30°C for 3 days, and the colonies were screened for loss of reactivity to mAb 2H1 by a colony immunoblot assay. mAb 2H1 was used as the screening mAb, because it is representative of the class II mAbs to GXM (17). mAb 2H1 is a protective Ab that binds O-acetylated, but not de-O-acetylated, GXM (6). As a control, duplicate samples were plated on synthetic dextrose agar (0.67% yeast nitrogen base without amino acids, 2% glucose, 0.2% drop-out mix lacking uracil, 50 μg/ml uracil, and 0.1% 5-fluoro-orotic acid (FOA) (19)) to test for ura5 mutants. No more than a few colonies grew on the control plates containing FOA consistent with a relative paucity of mutational events among the irradiated cells. The percentage of cells that survived treatment was also determined by comparing the number of CFU on Sabouraud dextrose agar from irradiated samples to the number of CFU from nonirradiated cells.

GXM was isolated from the supernatant of 3-wk-old cultures grown in Sabouraud dextrose broth at 30°C with shaking at 150 rpm and purified as described (20) with minor modifications. Briefly, culture supernatant was filtered through 0.45-μm filters to remove residual cryptococci. The polysaccharide was precipitated by the addition of sodium acetate (10% (w/v)) and 2.5 vol of ethanol. After 2–3 days, the precipitate was collected and dissolved in water, and the amount of carbohydrate was estimated by the phenol-sulfuric method (21). After adjusting the solution to 0.2 M NaCl, hexyldecyltrimethylammonium bromide (CTAB; Sigma-Aldrich, St. Louis, MO) powder (minimum of 3 g of CTAB per gram of carbohydrate) was slowly stirred into the solution to selectively precipitate GXM. The GXM precipitate was collected by centrifugation, dissolved in 2 M NaCl, and dialyzed extensively against 1 M NaCl to remove the CTAB and then against distilled water. The GXM solution was lyophilized to collect GXM, which was weighed and dissolved in water to generate a solution of defined concentration for experimental use.

Colonies were lifted from agar plates to nitrocellulose filters (0.45 μm; Osmonics; Fisher Scientific, Pittsburgh, PA) and air dried for 15 min. The filters were then blocked to prevent nonspecific protein binding by incubating in 5% nonfat milk in TBS (10 mM Tris (pH 7.2), 150 mM NaCl) with 10 mM sodium azide (TBS-A) for 30–60 min at room temperature with mild shaking. The milk solution was then replaced by a primary Ab solution containing the GXM-binding mAb of interest in TBS-A at a concentration of 13.3 nM. The filters were incubated with the primary Ab for 60 min. The primary Ab solution was decanted, and the filters were washed three times with TBS-A. The filters were subsequently incubated with a 1 μg/ml secondary Ab in TBS-A for 45–60 min. Alkaline phosphatase-coupled goat anti-mouse IgG1 or goat anti-mouse κ was used as the secondary Ab (Southern Biotechnology Associates, Birmingham, AL. The filters were then washed three times with TBS-A before alkaline phosphatase detection. To detect the colonies that expressed the GXM epitope bound by the individual Abs, filters were incubated with the alkaline phosphatase substrates 5-bromo-4-chloro-3-indolyl phosphate/nitroblue tetrazolium (Kirkegaard & Perry Laboratories, Gaithersburg, MD). At the completion of the reaction, the filters were rinsed briefly with distilled water and air dried in the dark at room temperature.

To identify colonies to which mAb 2H1 did not bind, dried filters were compared with the original agar plates. Potential mutants were transferred to new Sabouraud dextrose agar plates, grown for 3 days at 30°C, and retested by the colony immunoblot assay. The identified mutants are collectively referred to as epitope-deficient capsular (EDC) mutants.

Ten milligrams of lyophilized GXM were resuspended in deuterium oxide (D₂O; 99.9%; Cambridge Isotope Laboratories, Cambridge, MA) and dissolved overnight at room temperature. Approximately 600 μl of each sample was analyzed in a 5-mm NMR tube at 80°C in a Varian (Palo Alto, CA) 600-MHz spectrometer (20). ¹H chemical shifts were measured relative to an external sodium 4,4-dimethyl-4-silapentane-1-sulfonate standard. The data presented are an average of 128 scans.

C. neoformans cells were collected from culture medium by centrifugation (320 × g; room temperature) and washed five times with distilled water. One-half of these collected cells were resuspended in distilled water with no further treatment, and the other half of the cells were de-O-acetylated by resuspending the cells in distilled water that had been adjusted to pH 11 by the addition of ammonium hydroxide and incubated at room temperature for 1 h with continuous slow mixing (12). Cells were collected by centrifugation as above and washed five times with 30–40 ml of PBS. Cells were resuspended in 10 ml of PBS, counted using a hemacytometer, and used in the Hestrin assay (22, 23)

To use the positive/negative selection of the URA5-based complementation vector for C. neoformans, spontaneous variants of the EDC mutants with mutations in the URA5 gene were selected on synthetic dextrose agar containing FOA. A single FOA-resistant colony for each EDC mutant was selected and single-cell streaked five times on FOA-containing agar to ensure a stable phenotype. The capsular phenotype was confirmed for each mutant following this selection procedure.

Cryptococcal cells were grown in yeast peptone dextrose (YPD) broth for 2 days and then diluted to 2 × 10⁶ organisms/ml in fresh YPD broth and grown at 30°C at 225 rpm (24). When cells reached a density of ∼6–8 × 10⁶ cell/ml, cells were collected by centrifugation at 2150–4600 × g for 5–15 min at 4°C and washed twice with cold ultrapure water. Cells from 200 ml or less of YPD culture were resuspended in 50 ml of cold electroporation buffer (10 mM Tris (pH 7.4), 1 mM MgCl₂, 270 mM sucrose); 200 μl of 1 M DTT was added; and cells were incubated on ice for 5–15 min. Cells were collected as before, washed twice in cold electroporation buffer, and then resuspended in a minimal amount of cold electroporation buffer (∼2–7 × 10⁹ cells/ml).

For transformation, 40 μl of cryptococcal cells, in electroporation buffer, and 350 ng of the pC5T-20 genomic library DNA (derived from the C. neoformans Mat a strain WSA-20) were added to ice-cold, 2-mm-gap electroporation cuvettes. The library was linearized with Meganuclease I-Sce-I (Roche Applied Science, Indianapolis, IN) to expose the terminal telomeric sequences, phenol:chloroform purified, and resuspended in ultrapure water before its use. Cells were electroporated on the low-power setting of an ECM600 BTX electroporator with 475 kV of charging voltage, 25 μF of capacitance, and 480 Ω of resistance. These settings routinely yielded a pulse constant of 18–20 ms. Following electroporation, the cells were plated directly onto synthetic complete medium lacking uracil and incubated at 30°C. Ura5⁺ colonies were screened for a reversion of their mAb 2H1 phenotype by the colony immunoblot assay.

A direct ELISA was performed to assess the ability of the GXM-binding mAbs to recognize purified GXM from the EDC mutants. Ninety-six-well microtiter plates (Fisher Scientific) were coated with 10 μg/ml GXM (in a 50-μl volume, PBS) either at 37°C for 1 h or overnight at 4°C. Nonspecific binding sites in the wells were blocked with 200 μl of 2% BSA in PBS (blocking solution). Following removal of the blocking solution, 50 μl of the primary Ab solutions at a concentration of 13.3 nM was added to the appropriate wells and incubated at 37°C for 1 h. Wells were washed three times with 300 μl/well TBS containing 0.1% Tween 20 and 1 mM sodium azide (TBS-TA). A 50-μl volume of a 1 μg/ml solution of alkaline phosphatase-coupled goat anti-mouse κ chain (Southern Biotechnologies) in PBS was added to all wells, and plates were incubated at 37°C for 1 h. Wells were washed as above, and 50 μl of a 1 mg of p-nitrophenyl phosphate per milliliter of carbonate buffer (pH 9.6) (50 mM Na₂CO₃, 1 mM MgCl₂·6H₂O) was added to each well, and the absorbance was measured at a wavelength of 405 nm.

Three encapsulated C. neoformans mutants that failed to bind mAb 2H1 were isolated using a negative-selection strategy (Fig. 1,A). The mutants were identified at the conclusion of immunoblotting by the absence of colony staining with the alkaline phosphatase detection stain 5-bromo-4-chloro-3-indolyl phosphate/nitroblue tetrazolium. The lack of reactivity of mAb 2H1 for each of these mutants was confirmed by chemiluminescence detection of HRP-coupled secondary Ab, agglutination assays, and FACS analysis (data not shown). The inability of mAb 2H1 to bind the mutants was not due to an acapsular phenotype, because each mutant had a capsule visible in an India ink suspension (Fig. 1 B). No differences in colony morphology or capsule thickness were apparent when the mutant cells were compared with wild-type cells.

The frequency of colonies negative for binding to mAb 2H1 was ∼1 in 5–10,000 after mutagenesis, where 85% or more of the plated cells were killed. EDC1 was identified in a population of irradiated yeast that received 30,000 μJ of energy with 86.2% killing. EDC3 and EDC4 had received 30,000 μJ (95.6% killing) and 50,000 μJ (99.1% killing), respectively.

NMR analysis of GXM from EDC4 indicated a significant loss of O-acetylation relative to that from the wild-type strain B-3501 (Fig. 2). GXM from the mutant strain revealed an almost complete loss of the four acetylation peaks found in GXM of the wild-type strain between 2.12 and 2.26 ppm. Small signals were detectable at ∼2.03 and 2.24 ppm in GXM from the mutant strain but not at ∼2.20 or 2.22 ppm consistent with the loss of most acetylation, a modification that is required for recognition by the majority of mAbs that recognize GXM (6, 14, 15). The signal at ∼2.03 ppm was not apparent in the wild-type strain and is consistent with a small amount of acetylation occurring at a different location in GXM from EDC4.

The chemical shifts for the anomeric protons of the mannosyl residues are similar in the wild-type and EDC4 strains, suggesting that the composition of the repeat unit structures (i.e., chemotype) is the same (Fig. 2). The chemical shifts observed here are in agreement with those previously determined (20) and classify strain B-3501 as chemotype 2 and consisting of M1 (51%), M2 (19%), M5 (18%), and M6 (12%) structural reporter groups (Fig. 3). The small −0.02 shift in some of the signals in EDC4 may be a consequence of an altered acetylation pattern. However, the difference in the intensities of each signal between the wild-type and EDC4 strains suggest a different ratio of structural reporter groups within the chemotype.

To rapidly determine whether O-acetylation was affected in all of the independently generated mutants, the colorimetric Hestrin assay was performed on whole cells. The results of this assay indicated that the amount of acetylation on EDC1, -3, and -4 was below the limit of detection of the assay (data not shown). De-O-acetylating GXM by alkaline hydrolysis did not completely remove the reactive O-acetyl groups from B-3501. Interestingly, a separate mutant that retained some reactivity with mAb 2H1 also retained some O-acetyl groups after treatment (data not shown).

FOA-resistant isolates of each mutant (designated by the FOA suffix) were generated so that a genomic library containing a URA5 selection marker could be used for complementation purposes. EDC4-FOA was complemented by a 6-kb genomic fragment that contained the complete open reading frame of the CAS1 gene (GenBank accession no. AF355592) from C. neoformans, which is necessary for the O-acetylation of GXM (14). Thus, given the lack of O-acetylation in the EDC mutants and the presence of the intact CAS1 open reading frame in the complementing genomic fragment, the CAS1 gene was presumptively identified as sufficient to complement the mutation in EDC4-FOA. To formally test this hypothesis, the non-CAS1 sequence was removed from the plasmid and the plasmid was electroporated into EDC4-FOA. Approximately 20–50% of the Ura5⁺ CFU obtained were also detectable by mAb 2H1. A higher percentage of double positives was expected, because telomeric expression vectors improve transformation efficiencies in C. neoformans (25); therefore, it is probable that some degradation of the telomeric sequences occurred, resulting in a lower than expected number of double-positive cells. Nonetheless, the number of double-positive cells greatly exceeded the number of spontaneous Ura5⁺ revertants and could have only occurred through the presence of the complementing gene. This plasmid also complemented the nonreactive mAb 2H1 phenotype of EDC1-FOA and EDC3-FOA, strongly suggesting that the three independent mutants were either carrying mutations in the CAS1 gene or that extrachromosomal expression of CAS1 was masking the mutant phenotype.

The binding of 32 different mAbs (13 IgM, 14 IgG1, 1 IgG3, 4 IgA) to the capsular mutants was studied by the immunoblot assay. EDC1, -3, and -4 shared similar Ab reactivity profiles (Fig. 4). This suggested a similarity in the capsules of the mutants and was consistent with the genetic complementation data showing that the mutants harbored similar genetic and capsular defects. Because O-acetylation is required for GXM binding by most GXM-binding Abs (6, 14, 15), the results were compared with prior results where the reactivity of the various mAbs for de-O-acetylated GXM was tested by direct ELISA. This was of particular interest, because the genetic screen used here selected for mutants in the epitope of mAb 2H1 that was considered representative of the class II category, which includes a large group of Abs believed to bind to a common, or closely related, epitope(s). Surprisingly, some class II mAbs (12A1, 7E5 (IgM), 25G12 (IgA), and 18G9 (IgA)), previously identified to require O-acetylation for binding by ELISA (6), still retained their ability to bind the EDC mutants by the immunoblot analysis (Fig. 4).

Five mAbs (2H1, 21D2, 12A1, 13F1, and 18B7) were chosen for further analysis, because they represent Abs that are protective (2H1, 12A1, and 18B7) and nonprotective (13F1, 21D2) in a murine model of disease using 10⁷ serotype D C. neoformans organisms and Abs that were previously shown to require O-acetylation (2H1, 12A1, 13F1, and 18B7) or not to require O-acetylation (21D2). Furthermore, mAb 18B7 is currently in clinical evaluation for the treatment of cryptococcosis. The colony immunoblot assay demonstrated a difference between two protective Abs, because mAb12A1 bound to the EDC mutants and mAb 2H1 did not bind (Fig. 4). Interestingly, mAb 12A1 was previously shown to require O-acetylation for binding GXM; yet, the lack of O-acetylation on GXM in the context of the mutant colonies did not appear to significantly influence mAb recognition of the polysaccharide. mAb 18B7 appeared to bind the mutants; however, the extent of binding was slightly above the background staining of the negative controls. Both nonprotective Abs (13F1, 21D2) demonstrated relatively low levels of reactivity for the wild-type strain by colony immunoblot but differed in their reactivities for the capsular mutants: mAb 13F1 did not bind to the mutants, and mAb 21D2 had a greater reactivity to the mutants than to the wild type.

The colony immunoblot assay was limited by the qualitative results obtained and the uncertainty about whether whole cells or GXM shed from the surface had been retained on the membrane. Hypothetically, this latter point would be particularly relevant if the polysaccharide existed in slightly different conformations depending on the surface to which it was attached. ELISAs with purified GXM from each of the four EDC mutant strains was used to provide a secondary means to confirm the immunoblot results (Fig. 5). The direct ELISA demonstrated an almost complete loss of mAb 2H1 binding to the GXM from EDC1, EDC3, and EDC4 (Fig. 5,A). The direct ELISA for mAb 21D2 confirmed the result observed qualitatively by colony immunoblot; that is, an increase in binding to the acetyl-deficient GXM of the mutants (Fig. 5,B). There was a >3-fold increase in binding to EDC4 and >40-fold increase in binding to EDC1 and EDC3 relative to the binding of the mAb to the wild-type strain. The ELISA for mAb 12A1 revealed similar binding curves for mutant and wild-type GXMs; however, there may be fewer binding sites for the Ab on the GXM of EDC4 as judged by the lower absorbance reading (Fig. 5,C). The ELISA for mAb 18B7 (Fig. 5,E) established that the mAb was capable of weakly binding the mutants at the highest concentration. Thus, the ELISA results confirmed the data obtained for mAbs 2H1, 21D2, 12A1, and 18B7 by the colony immunoblot method (Fig. 4). The ELISA results with mAb 13F1, however, disagree with the data from the colony immunoblot, because this mAb bound to the purified GXM from the mutants by ELISA (Figs. 4 and 5,D). Further examination of this discrepancy demonstrated that mAb 13F1 reacted weakly with GXM from the parent and the mutants by immunodiffusion (data not shown), but failed to bind the purified GXM of the mutants when applied to the nitrocellulose filters used in the colony immunoblot assay (Fig. 6). If the purified GXM is assumed to have the same composition and primary structure as is on the surface of the cell, then this discrepancy suggests that mAb 13F1 was recognizing a conformational epitope that was not accessible in the immunoblot assay but accessible by ELISA

The immunoreactivity patterns of 32 mAbs initially screened by immunoblotting did not suggest a isotype-related effect on the reactivity of the mAbs for GXM from the mutants (Fig. 4). Therefore, the differences in reactivity must be a consequence of somatic mutations within the variable regions of the individual mAb. A previous study of the Ig H chain sequences of these mAbs combined with analysis of H and L chain productive and nonproductive rearrangements concluded that many of these mAbs are derived from the same progenitor B cell (16). A comparison of the immunoblot data for EDC1, -3, and -4 with the B cell genealogy based on the molecular data (16) is shown in Fig. 7. The majority of the Abs in this lineage did not bind the mutants. However, two points in the lineage become relevant when examining the immunoblot data. First, the somatic mutations that yielded mAb 12A1 resulted in a variable region capable of binding GXM in the absence of acetylation. Secondly, further somatic mutations of the mAb 12A1 variable domain reversed this binding capability, because mAb 18B7 partially bound the mutants.

Our results indicated that the antigenic complexity of GXM is much greater than suspected. A general model of our understanding of the epitopes for the protective and nonprotective Abs based on prior characterizations (Fig. 8, left) and after the data with the EDC mutants are included in the analysis (Fig. 8, right) is diagrammed.

Serologic, functional, and molecular characterizations of the Ab response to GXM have grouped Abs according to efficacy (protective or nonprotective), quellung reactions (rim or puffy), immunofluorescent staining (annular or punctuate), and requirement of O-acetylation for binding (6, 26, 27, 28). In general, such groupings are well defined and nonoverlapping such that protective Abs give rim reactions or annular staining patterns, nonprotective Abs give puffy reactions or punctate patterns, and Abs that did not require O-acetylation for binding are nonprotective. These observations suggested that the highly restricted response to GXM, where almost all mAbs studied to date use the same V region (V_H7183, J_H2, V_k5.1, and J_k1) and only differ by a few somatic mutations, resulted from a response elicited by a few immunodominant antigenic determinants in GXM.

By altering the capsular structure indirectly through genetic manipulation, we discerned fine specificity differences between mAbs that previously have been considered to bind to the same epitope. Three independent mutants of C. neoformans were isolated in a screen that selected for the absence of mAb 2H1 binding, and all were deficient in O-acetylation. Their ability to bind mAb 2H1 was restored when complemented by CAS1, which is necessary for the O-acetylation of GXM (14). Hence, our results independently support the involvement of CAS1 in the O-acetylation of GXM and further established the importance of O-acetylation for the reactivity of the mAbs 2H1, 13F1, and others using genetically defined mutants (14, 29, 30). For EDC4, a small amount of acetylation was detectable in the ¹H NMR spectrum of its GXM at two locations. The signal at 2.24 ppm had the same resonance as that found in the spectrum of the wild-type GXM, whereas the signal at 2.03 ppm suggested that a small amount of acetylation was located in an area of the molecule that contained more electron shielding. Determining whether this was because acetylation occurred at a new location or a secondary structural change in the polysaccharide would require in-depth structural analysis of GXM that is beyond the scope of our study. Nonenzymatic migration of O-acetyl groups occurs on meningococcal C polysaccharide, and this could potentially explain the spectrum change observed (31). Loss of acetylation in the other two mutants was determined by the Hestrin assay, which lacks the sensitivity of NMR but was adequate to confirm that each of our mutants was deficient in O-acetylation. The NMR spectrum for EDC4 GXM also suggested a possible change in the ratio of the structural repeat units. Consequently, subtle changes to the base structure of GXM cannot be ruled out for any of the mutant GXMs. Nonetheless, the alterations in mAb binding to the mutants are most likely a consequence of impaired GXM O-acetylation given that the chemotype remained unchanged and that O-acetylation has already been biochemically established as essential for mAb binding.

All prior studies had suggested that protective class II mAbs to GXM recognized an epitope dependent on O-acetylation. Our results demonstrated that mAb 12A1 still bound GXM in the absence of acetylation. This observation does not contradict prior findings (6, 29) that mAb 12A1 recognizes an epitope that requires acetylation, because loss of acetylation decreases binding in the assays used. However, the fact that mAb 12A1 reacted with GXM from each of the mutants suggested that the mAb can bind to more than one epitope and/or that acetylation is not required for all epitopes. This insight explains why mAb 12A1 retained significantly lower agglutination activity for a previously described de-O-acetylation variant than mAb 2H1 despite the fact that both mAb had the same agglutination end point with the wild-type strain (29). The implication of a protective Ab with more than one epitope is noteworthy. The fuzzy nature of Ab-Ag recognition (32, 33) in the case of mAb 12A1 suggests that there may be O-acetylated and de-O-acetylated epitopes that induce protection, or that this Ab can be protective or nonprotective depending on the GXM structure that is presented to it.

mAb 18B7 is being evaluated as an immunotherapeutic agent for the treatment of cryptococcosis (7). mAb 18B7 responded similarly to mAb 2H1 and other class II Abs in all assays previously conducted (7). Its sequence is very similar to mAb 2H1, with only 4 of 18 amino acid differences occurring in the binding pocket that recognizes a 12-mer peptide mimotope, and the apparent affinity constants for the two mAbs differ by only 3-fold (7, 34). Despite these strong similarities, a difference in specificity was apparent in their abilities to bind the EDC mutants. This result suggested that mAb 18B7 can either recognize an epitope that is de-O-acetylated or that its higher apparent affinity permits recognition of a quantitatively reduced O-acetylated epitope that mAb 2H1 cannot recognize. mAb 18B7 derives from the same progenitor cell as mAb 12A1, which our data showed bound de-O-acetylated GXM. Thus, if mAb 18B7 is binding a de-O-acetylated epitope, then the somatic mutations that establish mAb 18B7 as a unique Ab provide this mAb with binding characteristic of both mAbs 2H1 and 12A1. Our findings for mAbs 18B7 and 12A1 could also be explained by the possibility of conformational isomers of each Ab, as recently demonstrated for an IgE Ab (35), and cannot be discounted. mAb 18B7 may be a better candidate for clinical use than mAb 2H1, because it would be less likely to select de-O-acetylated variants during Ab therapy. Such variants can conceivably arise spontaneously during infection, may differ in virulence from the original strain, and would not be recognized by protective Abs that require acetylation to be present (29, 36, 37).

mAbs 13F1 and 21D2 are both nonprotective Abs, but the former requires O-acetylation, and the latter does not. Within the context of the mutant background, additional insights into how these two mAbs are different were gained. Previously, it appeared that the epitope recognized by mAb 21D2 was not affected by O-acetylation (6). However, our immunoblot and ELISA data suggested that acetyl groups were inhibiting the binding of the Ab, because there was increased binding to the de-O-acetylated GXM mutants. These data provided an insight into the molecular mechanism of Ab recognition by this nonprotective Ab and would predict a binding pocket that does not easily accommodate the acetyl group or that it recognizes a conformational epitope that is found more frequently in GXM with reduced O-acetylation. Furthermore, the contrast of increased mAb 21D2 binding and decreased mAb 2H1 binding to the de-O-acetylated mutants may suggest that these mAbs recognize the same oligosaccharide within GXM, but its acetylation state determines Ab specificity. Further analysis is required to validate this hypothesis, but it raises the intriguing notion that protective and nonprotective epitopes of GXM may occupy similar regions of the polysaccharide.

mAb 13F1 did not bind GXM from the mutants when adsorbed to nitrocellulose but retained its binding ability in ELISA and immunodiffusion assays. This suggests that the epitope recognized by mAb 13F1 is conformational, and that acetylation is a structural requirement. Conformational epitopes in polysaccharides are known in bacterial systems (38, 39) but have not, until now, been shown to be important in GXM. Interestingly, this may explain why this mAb binds to serotype A and D strains in annular and punctuate immunofluorescent patterns, respectively, and thus why it protects mice infected with a serotype A strain (28). Serotype A and D contain different quantities of xylose in their GXM, and individual strains differ in O-acetylation content (12, 20, 40, 41). Hence, O-acetylation in the context of the different GXM compositions could yield different capsular conformations and thus different immunofluorescent patterns. The role of O-acetylation in determining structural conformation is further reinforced by immunofluorescent studies that demonstrate altered capsule staining by mAbs 2H1 and 12A1 on a de-O-acetylated spontaneous variant of C. neoformans (29). Courtauld-Koltun modeling of acetylated LPS from Salmonella typhi suggests that the bulky O-acetyl groups reside on the surface of the molecule and create a rigid structure that could prevent intermolecular interactions of some Abs with the sugars of the backbone (42). Such modeling of the mannan backbone of GXM has depicted several possible secondary conformations, but, unlike the LPS of S. typhi, it is postulated that acetyl groups would decrease the rigidity of one model (12). Therefore, the lack of O-acetylation in the capsules of the mutants would be predicted to influence polysaccharide secondary structure and translate into different antigenic structures being presented in the different Ab-based assays. Although not a part of the detailed analysis, a separate mutant was isolated that had a partially acetylated capsule. The acetylation could not be removed under the Hestrin assay conditions. A similar result was observed for the wild-type strain, because some of the acetyl groups remained after chemical treatment. This may suggest that some of the acetyl groups on GXM are not always solvent accessible and could again suggest a relationship between GXM conformation and epitope accessibility.

The vast majority of the mAbs tested did not recognize the mutants, confirming prior findings that they require the O-acetylation modification for binding, and is in agreement with the findings of the C. neoformans CAS1 mutant (14, 15), which was also unable to bind the majority of Abs tested. Analysis of a B cell lineage of these Abs demonstrated that somatic mutations produced Abs with different fine specificity as can be observed from their binding abilities to the acetyl-deficient mutants. The significance of confirming the contribution of the variable domain recognition of GXM is important in light of new data showing that the isotype of human chimeric Abs to GXM, which contained identical variable domains, can affect the fine specificity of the Ab (43). Site-directed mutagenesis studies of mAbs 12A1 and 13F1 have elucidated two amino acids responsible for their immunofluorescent staining differences (44). Closer examination of the amino acid differences among the Abs to GXM evaluated with the EDC mutants may similarly reveal residues important in binding the polysaccharide. Our success at using mutants to dissect the antigenic complexity of a capsular polysaccharide in C. neoformans suggests that a similar approach may be fruitful for other capsulated pathogenic microbes where current methods of epitope identification are intangible. Furthermore, the antigenic complexity revealed by these findings suggests that mutants may be identified that lack nonprotective-eliciting GXM epitopes for use in conjugate vaccines.

Beyond the insights for the specificity of individual mAbs, these data collectively show that the Ig response to GXM, although restricted in its V_H usage, does not necessarily recognize a limited number of Ag-binding sites. Class II mAbs recognize all C. neoformans serotypes and presumably bind to the common antigenic determinant factor 1, defined by differentially absorbed rabbit sera (6, 10). Our observations that class II mAbs recognize distinct epitopes implies that factor 1 is antigenically complex and consists of epitopes dependent and independent on O-acetylation. Thus, the discovery of epitope-specific heterogeneity among class II mAbs indicates that a diverse and complex humoral response occurred in response to GXM-tetanus toxoid conjugate, which includes the recognition of acetylation-dependent conformations.

We thank John Perfect and Brian Wickes for advice on molecular genetics in C. neoformans, Guilhem Janbon for sharing information on the CAS1 mutant, and the Casadevall laboratory members for their suggestions. We also thank Brian Wickes for the genomic library, Sean Cahill for performing the ¹H NMR analysis, and Gwladys Fremand for assistance early in the course of this work. We are grateful to Oscar Zaragoza for critical reading of the manuscript.