A Monoclonal Antibody to Cryptococcus neoformans Glucuronoxylomannan Manifests Hydrolytic Activity for Both Peptides and Polysaccharides*

Studies in the 1980s first showed that some natural antibodies were “catalytic” and able to hydrolyze peptide or phosphodiester bonds in antigens. Many naturally occurring catalytic antibodies have since been isolated from human sera and associated with positive and negative outcomes in autoimmune disease and infection. The function and prevalence of these antibodies, however, remain unclear. A previous study suggested that the 18B7 monoclonal antibody against glucuronoxylomannan (GXM), the major component of the Cryptococcus neoformans polysaccharide capsule, hydrolyzed a peptide antigen mimetic. Using mass spectrometry and Förster resonance energy transfer techniques, we confirm and characterize the hydrolytic activity of 18B7 against peptide mimetics and show that 18B7 is able to hydrolyze an oligosaccharide substrate, providing the first example of a naturally occurring catalytic antibody for polysaccharides. Additionally, we show that the catalytic 18B7 antibody increases release of capsular polysaccharide from fungal cells. A serine protease inhibitor blocked peptide and oligosaccharide hydrolysis by 18B7, and a putative serine protease-like active site was identified in the light chain variable region of the antibody. An algorithm was developed to detect similar sites present in unique antibody structures in the Protein Data Bank. The putative site was found in 14 of 63 (22.2%) catalytic antibody structures and 119 of 1602 (7.4%) antibodies with no annotation of catalytic activity. The ability of many antibodies to cleave antigen, albeit slowly, supports the notion that this activity is an important immunoglobulin function in host defense. The discovery of GXM hydrolytic activity suggests new therapeutic possibilities for polysaccharide-binding antibodies.

naturally occurring catalytic antibodies to carbohydrates have not been described (22). Previous studies in our laboratory have led to the development of a murine monoclonal antibody (mAb) library against glucuronoxylomannan (GXM), the major component of the polysaccharide capsule in the fungal pathogen Cryptococcus neoformans (23)(24)(25)(26)(27). Earlier studies suggested that two of these mAbs, 3E5 and 18B7, possess hydrolytic activity against a peptide antigen mimetic known as P1 (28,29). Using two-dimensional NMR, variable region identical isotype switch variants of 3E5 were shown to hydrolyze P1 at different rates, with IgG3 possessing no catalytic activity. These studies, along with a report by another group, illustrated that class switching can impose different structural constraints on the V domain, thereby altering binding affinity and causing the emergence of proteolytic activity (28 -30). Mass spectrometry revealed that 3E5 hydrolyzed the 12-amino acid P1 (SPNQHT-PPWMLK) at positions 1, 2, and 10, whereas NMR data suggested that the closely related 18B7 antibody hydrolyzed P1 at different positions. Here, we confirm and characterize the catalytic activity of 18B7 against the P1 peptide with mass spectrometry and kinetic analyses using Förster resonance energy transfer (FRET) techniques. We also present evidence that the same antibody catalyzes hydrolysis of a synthetic heptasaccharide and the C. neoformans polysaccharide capsule.
Previous studies have revealed the presence of a serine protease-like active site in a handful of proteolytic Igs, which is typically composed of a Ser-His-Asp catalytic triad (31)(32)(33)(34). Many variations in orientation and residue composition on the traditional serine protease triad have been described in proteolytic enzymes, and it is possible that either some of these known active sites or completely novel active sites are responsible for the catalytic activity of some Igs (35). A potential serine protease-like triad was previously identified in a model of 3E5 IgG1 and the crystal structure of 3E5 IgG3 (PDB code 4HDI) and is conserved in the 18B7 protein sequence (28,36). This triad matches the location of confirmed proteolytic residues that have been described in two unrelated catalytic antibodies; however, the crystal structures of these antibodies are not known (32,34). To understand whether this or other motifs are associated with catalytic antibodies, we developed a structural template algorithm to compare three-residue motifs in known protein structures and applied it to the full set of antibodies in the Protein Data Bank structure database, ultimately allowing us to predict and test the catalytic activity of other antibodies.

Results
mAb 18B7 Sequencing-Since the 18B7 mAb to GXM was initially isolated and characterized, several studies over the intervening years have found nucleotide and amino acid sequences for the V region containing discrepancies at a number of positions in both H and L chains (24,37,38). These earlier sequences were obtained by either direct mRNA sequencing of the hybridoma cell line or by sequencing cDNA generated by the hybridoma through RT-PCR. Discrepancies are most likely due to either sequencing errors or accumulation of mutations that have occurred in the hybridoma over time. To investigate the basis for catalytic activity and to generate a structural model of 18B7, an accurate amino acid sequence was determined through protease digestion and mass spectrometry of the same antibody stock used in all experiments in this study. Mass chromatogram analysis showed complete coverage of the V H and V L regions with no peptides matching more than one of the possible 18B7 sequences, fully resolving all five V L discrepancies and all five V H discrepancies. Based on these data, 18B7 differs from the closely related 3E5 mAb at only 3 amino acids in the V L and 15 positions in the V H (supplemental Fig. S1).
mAb 18B7 Peptide Hydrolysis-Our laboratory has previously shown that certain isotypes of the GXM-binding mAb 3E5 are capable of specifically cleaving the P1 peptide mimetic at positions 1, 2, and 10 (28). Matrix-assisted laser desorption/ ionization time-of-flight (MALDI-TOF) mass spectrometry (MS) analysis following incubation of P1 with mAb 18B7 resulted in specific cleavage at positions 5 and 6 ( Fig. 1A). This result confirms prior NMR data suggesting that mAb 18B7 hydrolyzes P1 in a different pattern from 3E5 and illustrates that small changes in the paratope can affect catalytic activity (28). It is interesting that such a slight variation in antibody sequence translates into a marked change in the peptide cleavage site. To investigate the selectivity of catalytic activity for specific residues in the P1 peptide, a library of alanine-substituted P1 variants was synthesized and incubated with mAb 18B7, with products analyzed by mass spectrometry (supplemental Fig. S2). As the alanine position in P1 varied, we observed the emergence of additional cleavage sites; however, no residue substitution eliminated P1 fragmentation (Fig. 1B).
To study the kinetics of peptide hydrolysis by antibody, we developed a FRET protocol making use of internally quenched fluorescent peptide substrates, which were conjugated to an N-terminal Mca fluorophore and a C-terminal Dnp quencher, leading to increased fluorescence with a cleaved peptide ( Fig.  2A). In this work, we refer to peptides with these two modifications by appending "q" to their names. mAb 18B7 hydrolyzes the P1 FRET peptide (P1q) but not the irrelevant 1Fq peptide (DYKDDDK), illustrating that epitope specificity is important in the catalytic mechanism (Fig. 2, B and C). Both P1q and 1Fq were hydrolyzed rapidly by proteinase K, a broadly specific serine protease (Fig. 2, B and C). We note that even after 64 h, 18B7 has managed to hydrolyze only about 60% of the available P1q. Hydrolysis of P1q was nearly completely inhibited by preincubation of 18B7 with both 1000 and 50 M PMSF, suggesting the antibody uses a serine protease-like catalytic mechanism (Fig.  2D). Purified C. neoformans GXM at 10 and 100 g/ml was also found to inhibit P1q hydrolysis by 18B7 in a concentration-dependent manner, suggesting that P1q competes with GXM for the same binding site (Fig. 2E). To better represent the degree of P1q inhibition under these conditions, we calculated the area under the curve (AUC) and compared them using an ordinary one-way analysis of variance (ANOVA) and post hoc analysis by the Tukey test for multiple comparisons (Fig. 2, F and G). Both PMSF and GXM inhibition showed significant differences in the ANOVA with post hoc analysis finding significant differences between all pairwise conditions. Using varying 18B7 and substrate concentrations over 24 -72 h, we measured reaction velocities and fit the data to the classic Michaelis-Menten model for enzyme kinetics (Fig. 3, A and B). The kinetic rate constants K m , V max , and K cat were calculated for each antibody concentration from the nonlinear models (Fig. 3C). Depending on antibody concentration, the K m values ranged from 46 to 55 M, and the K cat values were near 10 Ϫ2 s Ϫ1 . V max was largest with 1 mg/ml 18B7 at 0.2 M/s and decreased to 0.03 M/s with decreasing antibody.
18B7 Hydrolysis of Polysaccharide-Because mAb 18B7 was generated by immunizing mice with GXM, we sought to determine whether its catalytic activity extended to polysaccharide substrates. We first used a synthetic heptasaccharide substrate containing the M2 motif (Fig. 4A), which is a major structural component of GXM in serotype A C. neoformans (39). This heptasaccharide has been previously shown to bind antibodies to GXM, including 18B7, and it has been found to elicit nonprotective antibodies in mice (40,41). We incubated 300 M heptasaccharide alone, with 720 g/ml 18B7, or with 18B7 preincubated with 1 mM PMSF. Incubations were performed at 37°C for 22 h, and fragmentation of the heptasaccharide was detected by MALDI-TOF MS (Fig. 4B). The heptasaccharide alone showed no visible fragmentation and a peak at ϳ1278 Da corresponding to the intact substrate. Although the intact heptasaccharide was still present in the sample incubated with 18B7, many higher intensity peaks appeared below 600 Da. These peaks were assigned to fragments of the heptasaccharide corresponding with cleavage along the mannan backbone. When PMSF-inhibited 18B7 was used, however, the intensities of these new peaks were nearly eliminated.
To further examine mAb 18B7-mediated hydrolysis of cryptococcal polysaccharide, we used, as a substrate, heat-killed C. neoformans cells that had been induced for capsule growth and washed repeatedly to remove any extracellular or loosely attached polysaccharide. Washed cells were incubated at an equal concentration in PBS, varying concentrations of 18B7, or with 50 g/ml of an irrelevant murine IgG1. Aliquots of supernatant were collected at regular time intervals and analyzed for shed polysaccharide by capture ELISA (Fig. 5A). Compared with the cells incubated with no antibody, the addition of 1 g/ml 18B7 led to significantly higher amounts of GXM in the supernatant at 7 days and later (Fig. 5B). The addition of 10 g/ml 18B7 led to larger amounts of GXM released over time with significant increases at day 3 and later. The presence of 50 g/ml 18B7 caused significantly lower levels of GXM at days 3 and 5, but it ultimately resulted in higher levels of GXM at day . mAb 18B7 hydrolysis of the P1 peptide differs from that of the related 3E5 mAb. A, MALDI-TOF mass spectra of the P1 peptide alone and in the presence of mAb 18B7 following incubation at 37°C. Fragmentation of P1 is evident at positions 5 and 6 following incubation with 18B7. B, 18B7 was incubated with P1 and 12 P1 derivatives with an alanine substituted at each position. Each row represents a different peptide incubated with 18B7. Locations of alanine substitutions are indicated in red, and vertical bars indicate the locations that each peptide was hydrolyzed. *, sites of P1 hydrolysis by mAb 3E5 were previously determined and are shown for reference (28).
23 and later. The addition of nonspecific IgG1 resulted in slightly elevated GXM release at days 23 and 30, but similar levels to the antibody-free sample at all other time points, and much lower than measured with 10 g/ml mAb 18B7. Representative images of fungal cells from each condition are shown in Fig. 5C with capsules visible by india ink stain.
Changes in capsule antigen concentration due to mAb 18B7mediated hydrolysis were examined by immunostaining of cells using a fluorescently labeled 18B7 probe (18B7-AF568) (Fig.  6A). Compared with the cells at day 0, cells incubated for 39 days with no antibody showed increased 18B7-AF568 binding by flow cytometry. A similar reactivity was observed for cells . mAb 18B7 catalytic activity is inhibited by serine protease inhibitors and cryptococcal polysaccharide. A, schematic illustrates the FRET system used to assay catalytic activity against peptide substrates. Increased fluorescence at 405 nm indicates peptide hydrolysis. B, P1q was incubated either alone or with 18B7 for 64 h. Proteinase K (100 g/ml) was added at 64 h. C, 1Fq was incubated either alone or with 18B7 for 64 h, with 100 g/ml proteinase K added at 64 h. D, P1q was incubated with mAb 18B7 alone or after preincubation with either 50 or 1000 M PMSF. E, P1q was incubated with 18B7 alone or after preincubation with 10 or 100 g/ml purified cryptococcal polysaccharide as a competitive inhibitor. F, AUC was calculated for each PMSF inhibition curve in D. G, AUC was calculated for each GXM inhibition curve in E. All AUC measurements were compared by ordinary one-way ANOVA and the post hoc Tukey test for multiple comparisons (****, p Ͻ 0.0001).
incubated with 1 g/ml 18B7. Cells incubated with 10 and 50 g/ml 18B7 showed higher binding of 18B7-AF568. A small population of cells with low 18B7-AF568 reactivity was also observed in the presence of 50 g/ml 18B7. The cells incubated with irrelevant IgG1 showed the lowest 18B7 reactivity. Capsule imaging showed binding intensity profiles consistent with the flow cytometry data (Fig. 6B). In multiple instances, the capsule edge shows no reactivity to 18B7-A568. Cell wall staining with Uvitex2b was comparable among all samples suggesting little or no cell wall remodeling. These data are consistent with quantitative changes in capsular epitopes due to 18B7mediated catalysis that results in enhanced GXM binding of 18B7, possibly as a result of exposing new antigenic regions in the capsule by facilitating antibody accessibility.
The size distribution of GXM molecules released into the supernatant by the five samples of heat-killed cells was compared at 10 weeks by quasi-elastic or dynamic light scattering (DLS). Supernatant was collected from each set of cells, lyophilized, and concentrated 40-fold to obtain useful correlation functions by DLS (Fig. 7A). Over 10 independent measurements, the mean effective diameter (ED) of particles released from cells incubated with either 1 or 10 g/ml 18B7, were significantly smaller than that of the other conditions (Fig. 7B). The ED and polydispersity determined from all 10 measurements were used to estimate a log normal distribution of particle sizes for each sample (Fig. 7C). Additionally, a multimodal size distribution was estimated indicating two distinct populations of particles that shifted down in size in the presence of 1 or 10 g/ml 18B7 (Fig. 7D). Interestingly, the presence of 50 g/ml 18B7 reversed the trend observed with the lower concentrations and produced larger sized particles, which we attribute to mAb aggregation.
Catalytic Motif Analysis of PDB Structures-Our laboratory has previously published the crystal structure of the mAb 3E5 to GXM, whose V L region differs from that of mAb 18B7 by only 3 residues (36). A possible serine protease-like catalytic triad (Ser-26 -His-98 -Asp-1) was identified in the structure's V L region and is conserved in the 18B7 sequence. Given that well over 100,000 protein structures are present in the PDB, we developed a structural template algorithm to compare the putative catalytic site of anti-GXM mAbs with other antibody and protease structures. A schematic for this algorithm is shown in Fig. 8A. First, a 3-residue seed motif was chosen, and its template coordinates are iteratively averaged with all similar templates from a given set of structures (the mean dataset). The resulting mean template is then superimposed with all templates from another set of structures (the query dataset) to identify matches. Each 3-residue template consists of 24 Cartesian points in 3-dimensional space corresponding to five backbone and three side chain points per residue. The similarity between two templates was calculated by the root-mean-square deviation (r.m.s.d.) following optimal superposition as well as a distance-based r.m.s.d. (d.r.m.s.) of points in each template. Matching templates were classified as those with either an r.m.s.d. or d.r.m.s. similarity of Յ1 Å. A mean template was generated based on the 3E5 Ser-26 -His-98 -Asp-1 motif and a mean dataset of all 1665 unique antibody structures identified in the PDB. When this mean template was used to query the full set of antibody structures, 133 structures with a matching motif were identified (Fig. 8B). The full list of all 133 matching motifs is available in supplemental Table S1. Included in the matching structures were 14 of 63 (22.2%) known catalytic antibodies and 119 of the 1602 (7.4%) other antibodies, indicating a significant enrichment of the motif in antibodies with catalytic activity (p Ͻ 0.0001, Table 1). The putative 3E5 motif, however, was not present in any of the 10,302 hydrolase structures, which included known peptidase, glycosylase, esterase, and other hydrolytic enzyme structures (Fig. 8C). The number of structures analyzed in each hydrolase or antibody category is shown in Table 1. When the seed motif was changed to represent the classic catalytic triads of serine proteases trypsin, subtilisin, or the  For each antibody concentration, these data were fit to the Michaelis-Menten equation for a single-step bimolecular reaction using nonlinear regression. B, Michaelis-Menten models from A are shown in a Lineweaver-Burk plot. C, kinetic parameters K m , V max , and K cat were calculated for each antibody concentration.
cytomegalovirus (CMV) assemblin protease, our algorithm identified non-overlapping sets of enzyme structures but no antibody structures (Table 1). Trypsin belongs to superfamily A of proteases with mixed nucleophiles (PA clan), which contains both serine and cysteine proteases, whereas subtilisin and CMV protease belong to clans that exclusively utilize serine nucleophiles. We note that the trypsin seed motif was also found in a number of cysteine proteases, illustrating that this algorithm can identify similar motifs even with nucleophile substitutions. Residue Conservation, Activity Confirmation, and Phylogenetic Analysis-Sequence conservation was also compared between all 61 known unique catalytic antibody V L chains and a random sample of 61 other antibody V L chains (Fig. 9, A and B). Results showed increased conservation among catalytic antibodies at positions 26 and 99, which correspond to two residues of the putative catalytic triad in the aligned set of sequences. The proportion of catalytic antibodies with a serine at position 26 and a histidine at position 99 were also significantly higher (Fig. 9C). We then inferred the species of origin and family for each immunoglobulin structure by determining each molecule's highest scoring germ line V region genes. Based on the pairwise edit distance between the V region sequences, we cal-culated phylogenetic trees for all V L sequences ( Fig. 10A) and all V H sequences (Fig. 10B). Color-coded circular tracks in these figures illustrate the antibody species, family, whether it is a known catalytic antibody, and whether it contains the putative 3E5 catalytic motif. This analysis showed that most known catalytic antibodies in the PDB, as well as those with the 3E5 motif, share closely related V L sequences. There was no similar grouping when the same analysis was performed with V H sequences. To experimentally validate the results of our structural template algorithm, we obtained one of the putative catalytic antibodies possessing the presumed catalytic 3E5 motif and tested it for protease activity. We chose the M2 anti-FLAG antibody because of its commercial availability and its importance as a biological reagent. The FLAG peptide (DYKDDDK) was incubated at 200 M either alone or with 960 g/ml M2 antibody for 5 days at 37°C. Fragmentation of the peptide was analyzed by MALDI-TOF MS at both time 0 and again at day 5 ( Fig. 11). Results show that at time 0, the spectra of both conditions are very similar with peaks at 1013.4 Da corresponding to the intact peptide. By day 5, the sample incubated with M2 antibody has lost the intact peptide peak and several lower mass peaks appear in the spectrum, suggesting that the FLAG peptide has been hydrolyzed.

Discussion
Naturally occurring catalytic antibodies are associated with many positive and negative roles in autoimmune disease, homeostasis, and infection (4 -21). However, the field still has many unanswered questions. The biological importance of catalytic activity in antibodies is still debated, and the prevalence, diversity, and origin of catalytic activity in antibody molecules remains poorly understood. In this work, we report and characterize the peptidase activity of 18B7, a monoclonal antibody to GXM, which is the major polysaccharide component of the C. neoformans capsule. We also report the hydrolytic activity of the same antibody against a synthetic oligosaccharide substrate and the fungal capsule, providing the first example of a naturally occurring glycolytic antibody. Furthermore, we provide insights into the prevalence and origin of catalytic activity in antibodies by identifying a putative proteolytic active site and developing a structural template algorithm to find similar motifs in known antibody structures.
Previous studies indicate that a related mAb to GXM known as 3E5 has hydrolytic activity against the 12-amino acid P1 peptide (28,29,36). That work also suggests that 18B7, which differs from 3E5 in three V L positions and 15 V H positions, hydrolyzes P1 with a different pattern. Mass spectrometry data presented here confirmed this prior hypothesis by showing that P1 was hydrolyzed by 18B7 at positions 5 and 6 near the center of the peptide, whereas 3E5 hydrolyzes P1 at positions 1, 2, and 10. The fact that P1 is cleaved by 3E5 near both the N terminus and near the C terminus raises the possibility that the mAb can bind P1 in multiple orientations relative to the active site. It is also possible that the mAb has multiple active sites, although this has not been shown to occur in any catalytic Abs. Another possible explanation is that in the conformation adopted by bound P1, both the C and N termini are close to the active site. This, however, is not indicated by NMR, crystallography, or modeling studies (28,36,42), with the caveat that solution and crystal peptide conformations could differ. Based on the different P1 hydrolysis sites observed with 18B7 and 3E5 mAbs, these data indicate that either the peptide binds the two mAbs with distinct orientations or that the catalytic active sites in the two mAbs are different. Regardless of the explanation, the results indicate that a few amino acid differences in the paratope can lead to major differences in the specificity of catalysis.
The substitution of alanine for each residue in the P1 sequence did not prevent hydrolysis by 18B7 but instead altered the cleavage sites. The largest changes in hydrolysis pattern were evident with alanine substitutions in the TPPW motif, which is a common mimotope for peptide mimetics of GXM (43). These results suggest that differences in the paratope and CDRs between 18B7 and 3E5 affect the positioning of P1 relative to the hydrolytic active site. Furthermore, the position of the peptide relative to the active site is also affected by alanine substitution within the P1 mimotope.
The use of FRET peptides allowed a much more detailed characterization of the enzymatic properties of 18B7 peptidase activity. mAb 18B7 hydrolytic activity was specific for the P1 peptide and was competitively inhibited by the presence of increasing amounts of GXM, indicating that epitope recognition by the Ig V domain paratope is required for catalysis. Inhibition of P1 catalysis was also evident in the presence of the serine protease inhibitor PMSF, which functions by covalently attaching to the nucleophilic serine of a catalytic triad (44 -47). This result suggests that a serine protease-like mechanism is responsible for the 18B7 peptidase activity. The kinetic measurements of 18B7-mediated P1 hydrolysis illustrate that the reaction rate depends on both antibody and substrate concentration. The data also fit the classic Michaelis-Menten model for a single step enzymatic reaction with no indication of cooperativity between the mAb's two binding sites. Kinetic constants calculated from these data are similar to those that have been published for other catalytic antibodies and indicate a much less efficient catalyst than the typical enzyme (32,48). Depending on antibody concentration, the calculated K cat val-ues indicate one catalytic event approximately every 25-37 s. The catalytic efficiency of this reaction, calculated as K cat /K m , is 10 5 -fold lower than that of the most efficient enzymes. Despite this low efficiency and reaction rate, however, an antibody with these kinetic parameters could easily exert profound effects in vivo given the high serum concentrations (mg/ml) and long half-lives (2-3 weeks) of immunoglobulin molecules.
To answer the question of whether 18B7's catalytic activity could play any biologically relevant role, we tested the antibody's ability to hydrolyze a synthetic oligosaccharide and the C. neoformans polysaccharide capsule. Although glycosidase antibodies have been intentionally generated by immunization with transition state analogues, none have been found following immunization with a naturally occurring antigen (22). Incubation of 18B7 with a synthetic heptasaccharide containing one of the major structural units present in GXM suggested that the antibody was able to hydrolyze the oligosaccharide substrate, breaking at least two glycosidic bonds in the mannose backbone. Interestingly, this activity was inhibited by PMSF despite the significant mechanistic differences between glycosidases and serine proteases. Although it is possible that the same nucleophile is used for both substrates, the mechanism of PMSF inhibition may also differ between heptasaccharide and peptide hydrolysis by 18B7. For example, PMSF attachment to a proteolytic serine may sterically block another residue from participating in heptasaccharide hydrolysis.
We then used heat-killed fungal cells that had been washed repeatedly to remove any extracellular material or polysaccharide as a substrate for 18B7. The result of this experiment showed significantly increased levels of GXM released from the capsule in the presence of 10 g/ml 18B7. The increase was less pronounced with 1 g/ml 18B7. A higher 18B7 concentration of 50 g/ml paradoxically decreased levels of GXM release, likely due to the cross-linking and aggregation of fungal cells, which was visualized by india ink stain, effectively locking the capsule in place. This phenomenon has been described previously with live C. neoformans and concentrations of 18B7 50 g/ml and higher (49). At these higher antibody concentrations, the reduction in GXM released due to cross-linking overwhelms the amount of GXM hydrolyzed by 18B7. Notably, the presence of a control IgG1 did not result in cell aggregation or a consistent pattern of increased GXM release. Immunofluorescence staining of cells at day 39 and analysis by flow cytometry revealed that despite the presence of bound unlabeled antibody, cells that had been incubated with 10 and 50 g/ml 18B7 had increased capsular epitopes when compared with the control samples. Such epitope changes are consistent with 18B7 hydrolysis of the capsule, potentially exposing new antigenic regions. DLS measurements of GXM particles released in solution echoed these results with smaller sized particles present in the 1 and 10 g/ml 18B7 samples. Although no difference was seen between the control IgG1 and the no antibody samples, the 50 g/ml 18B7 sample contained GXM particles with a larger ED. However, the multimodal size distribution for this sample indicated that the ED increase was due to the lack of small particle intensity rather than the presence of larger sized particles. Thus, the result is consistent with the reduced release of hydrolyzed GXM due to extensive cross-linking in the presence of 50 g/ml 18B7.
It is unclear whether the same catalytic active site is responsible for both peptide and polysaccharide hydrolysis by 18B7, but it is fascinating that this antibody displays catalytic activity for two extremely different substrates. Enzymes, however, are usually evolved to carry out a very specific reaction with high efficiency. We note that at least one other example of an antibody displaying catalytic activity for two dissimilar substrates has been published (50), describing a murine anti-DNA Ig that is capable of hydrolyzing both double-stranded DNA as well as a peptide mimetic. Whether such broad substrate specificity is common in catalytic antibodies is unknown. It is possible that Igs with a propensity for catalysis have one or more relatively nonspecific active sites encoded in complementarity-determining region (CDR) and framework region residues in the germ line. Additionally, whether catalytic activity is selected during somatic diversification or, more likely, within the germ line remain open questions. B cell selection for increasing catalytic rates appears implausible due to the requirement for extended B cell receptor (BCR) occupancy by antigen. One possible mechanism for the selectability of slow nucleophiles in the somatic diversification of antibodies is based on the observation that immunization with artificial electrophiles can induce proteolytic antibodies through covalent attachment to BCR nucleophilic residues (33,51,52). Another proposed possibility involves BCR signal transduction through productive use of the free energy released from antigen hydrolysis (52).
Although the three-dimensional structure of 18B7 has not been solved, a crystal structure is available for the closely related 3E5 catalytic mAb to GXM (36). A putative serine protease-like catalytic triad was identified in the 3E5 V L region near the paratope consisting of Ser-26, His-98, and Asp-1 (28). These residues are conserved in the 18B7 V L. Although the distances between these residues are longer than for a classic catalytic triad, the flexibility of antibody CDR loops, especially upon binding antigen, and the N terminus may allow the residues to occasionally assume a configuration that facilitates catalysis. This is supported by the temperature dependence of  catalytic activity in the 3E5 mAb, which displayed very little activity at 25°C (28). Furthermore, the increased distances and lack of an oxyanion hole may partially account for the antibody's slow hydrolysis rates. We also note that the potent inhibition by PMSF supports the importance of a serine proteaselike nucleophile. To identify other antibodies that possessed the same motif, we developed a structural template algorithm, IgMotif, to create an average or mean template for a given seed motif, and then compared the mean template with all possible motifs from other known structures in the PDB. Similar approaches have been used successfully to identify structural motifs and catalytic triads in three-dimensional protein structures (53-56). The IgMotif algorithm is able to identify structures with motifs very similar to that of any user-specified seed containing one acidic, one basic, and one nucleophilic residue. Using IgMotif with initial seed motifs consisting of well characterized catalytic triads from structures of trypsin, subtilisin, and CMV protease allowed us to confirm the algorithm's specificity. Each of these enzymes is a member of a different protease clan with independent evolutionary origins. As classified by the MEROPS database, trypsin is a member of the PA clan, which contains both serine and cysteine proteases; subtilisin is in the serine protease SB clan, and CMV protease is a member of the SH clan, which employs a unique Ser-His-His triad (57). IgMotif identified non-overlapping sets of structures containing the motif for each enzyme when run against 10,302 hydrolase structures but identified no inappropriate matches in glycosylase, threonine, aspartic, and metalloprotease datasets. No motifs similar to these canonical enzyme triads were identified in 1665 unique antibody structures, which included 63 catalytic antibodies, indicating that any of the Ig catalytic motifs are structurally distinct from those found in these classic serine proteases. We note that the template-matching algorithm will fail to identify structures that are classified as proteases but only contain partial structures without the catalytic residues or whose catalytic triads have been modified through mutagenesis or the presence of certain inhibitors. It also may fail to match structures of lower resolution with less accurate atomic coordinates. When the hypothetical 3E5 catalytic triad was used as a seed, IgMotif identified an ϳ3-fold enrichment of the motif in catalytic antibody structures when compared with antibodies with no annotation of catalytic activity. No similar motifs were identified in any of the hydrolase structures. This result suggests that these residues are either involved in the catalytic mechanism of some antibodies or are part of a common structural component to a subset of catalytic antibodies. We also validated the association of the putative 3E5 catalytic motif with proteolytic activity by showing that M2 mAb, which is not known to have any peptidase activity, was able to hydrolyze the short FLAG peptide. Although this is a frequently used antibody in the laboratory, it is likely that peptidase activity is not observed under typical experimental time scales and conditions of low concentration and temperature. Sequence conservation analysis also supported these possibilities, as His-98 was significantly more common in catalytic antibody sequences than in a random sample of antibodies with unknown catalytic activity. By separating both V L and V H sequences of analyzed antibodies by TABLE 1 Putative 3E5 catalytic motif is enriched in catalytic antibody structures Table displays the total number of structures in both antibody (Ab) and hydrolase datasets that were analyzed with the IgMotif algorithm. A number of structures were eliminated from each group during several pre-processing steps to remove duplicate proteins and structures with incomplete atomic coordinates. Three-residue motifs were only identified for the remaining structures following all processing steps. Hits in each structure group are shown in the bottom half of the table for each of several seed motifs, with dashes indicating no hits. The proportion of antibodies containing the 3E5 motif was compared between the catalytic antibody group and the non-catalytic antibody group using a 2 two-sample test for equality of proportions. their edit distance, it was possible to visualize the grouping of closely related L and H chains. Many of the catalytic antibodies and those containing the 3E5 putative triad share closely related V L sequences, suggesting that certain V L germ line lineages are predisposed to producing catalytic antibodies. No such pattern was observed with the V H sequences. This result is consistent with several studies that have isolated Ig catalytic activity to the L chain (6 -8, 32, 34, 58 -60), although others have proposed catalytic motifs in the H chain (31,48). It is unclear why the V L would be more likely to contain catalytic activity than the V H , but it is possible that antibodies with catalytic H chains have simply been infrequently identified and are thus under-represented in the literature and the PDB. We also note that the C H is known to have significant effects on the catalytic activities of at least two different antibodies (28,29,50). Thus, if the hydrolytic active site is present in the V L , this would imply that major inter-domain and inter-chain conformational changes can be propagated in immunoglobulins, a topic that warrants further investigation (61,62).

Catalytic Abs
Monoclonal antibodies have become one of the most versatile options for pharmaceutical development and have been approved to treat a wide variety of illnesses. A better understanding of the characteristics and prevalence of naturally occurring catalytic antibodies could prove useful in designing antibody-based therapies with desired catalytic activities. Specifically, the identification of natural mAbs that hydrolyze polysaccharide antigen broadens the possible role for this function in immunity. In patients with cryptococcosis, soluble GXM in the serum and cerebrospinal fluid (CSF), which can be present at milligram/ml concentrations, has been implicated in numerous immunosuppressive effects (63)(64)(65)(66). Whether the presence of a natural catalytic antibody response to GXM affects the amount of soluble polysaccharide in vivo remains to be seen, but it is possible that this activity plays a mitigating or exacerbating role in pathogenesis. The administration of catalytic mAbs to patients with cryptococcosis could provide a new approach to therapy by effectively clearing antigen and mitigating the polysaccharide's immunosuppressive effects.

Experimental Procedures
mAb 18B7 Sequencing and Molecular Models-18B7 was purified as described previously (37). The 18B7 V region protein sequence was determined by the Johns Hopkins Mass Spectrometry and Proteomics Facility (Baltimore, MD). Protein digestion with a trypsin/LysC mixture of 40 g of 18B7 was performed following reduction with DTT and alkylation with   iodoacetamide. 500 ng of the desalted protein digest was analyzed with liquid chromatography tandem mass spectrometry on a Q Exactive Plus (Thermo Scientific). Mass chromatogram and peptide analysis was performed using Scaffold (version 4.6.2; Proteome Software, Portland, OR) and PEAKS (version 8.0, Bioinformatics Solutions Inc., Waterloo, Ontario, Canada).
Structural models of 18B7-IgG1 and 18B7-IgG3 were generated by performing in silico mutations and subsequent energy minimization of the 2H1-IgG1 and 3E5-IgG3 crystal structures in UCSF Chimera. Molecular graphics images were produced using the UCSF Chimera package from the Computer Graphics Laboratory, University of California, San Francisco (67). Peptide Synthesis-The P1 peptide and an alanine substitution P1 library were synthesized using Fmoc chemistry on a  microwave-assisted peptide synthesizer (Liberty; CEM Corp.) at the Rockefeller University Proteomics Resource Center (New York). FRET peptides were synthesized using Fmoc chemistry with an N-terminal Mca fluorophore and a C-terminal Dnp quencher at The Johns Hopkins University Synthesis and Sequencing Facility (Baltimore, MD).
Proteolysis and Kinetic Assays-To measure hydrolysis of P1 and its alanine substitution library, 18B7 at 1.98 mg/ml in PBS was incubated overnight at 37°C with 103 M peptide. Peptide hydrolysis was detected with MALDI-TOF MS at the Columbia University Protein Core Facility (New York). Kinetic assays with FRET peptides were performed by detecting fluorescence with the SpectraMax M5 microplate reader (Molecular Devices). All kinetic measurements were performed at 37°C with an excitation wavelength of 320 nm, an emission wavelength of 405 nm, and an emission cutoff filter of 325 nm. Wells were set up in triplicate, and 30 readings were taken per time point per well with a 3-s mix between time points. Photomultiplier tube sensitivity was set to low, and opaque shallow-well 96-well microplates were used with an optically clear plate adhesive to prevent evaporation. Inhibition studies with 50 or 1000 M PMSF were carried out following a 1-h pre-incubation of PMSF with antibody at 37°C. Inhibition experiments with purified GXM at either 10 or 100 g/ml also began with preincubation of the inhibitor with 18B7 for 1 h at 37°C. Inhibition data were analyzed by calculating the AUC for each condition and performing an ordinary one-way ANOVA and the post hoc Tukey test for multiple comparisons (****, p Ͻ 0.0001). AUC analysis was performed in GraphPad Prism (version 7, GraphPad Software, La Jolla, CA). Michaelis-Menten models were generated by first calculating the initial reaction velocity v 0 for each enzyme, substrate, and inhibitor condition, defined by the slope of the early linear phase of fluorescence production over time. For catalytic antibody samples, v 0 was defined as the slope of the average relative fluorescence units curve between 0.5 and 2.5 h with readings every 5 min. The maximal velocity (V max ) and Michaelis-Menten constant (K m ) were determined via nonlinear fit to Michaelis-Menten Equation 1, (Eq. 1) K cat was determined by V max ϭ K cat [E] 0 , with a murine IgG1 molecular mass of 150 kDa.
Heptasaccharide Hydrolysis-The synthesis of a heptasaccharide comprising the major M2 structural motif of C. neoformans GXM was described previously (41). Heptasaccharide was incubated at a concentration of 300 M in PBS, PBS with 720 g/ml mAb 18B7, or PBS with 18B7 and 1 mM PMSF. Incubations were performed at 37°C for 22 h. Samples were analyzed for heptasaccharide fragmentation by MALDI-TOF MS in a 2,5-dihydroxybenzoic acid matrix at the Albert Einstein College of Medicine Proteomics Facility (Bronx, NY). Fragmentation analysis for heptasaccharide was done as described previously for polysaccharides (68,69). All glycosidic cleavages (Z-, Y-, B-, and C-type) as well as cross-ring cleavages (X-and A-type) were examined. Fragments were assigned based upon the closest matches with charge state consideration.
C. neoformans Growth and Polysaccharide Purification-C. neoformans serotype A strain H99 (ATCC 208821) cells from a frozen stock were grown overnight at 30°C with agitation (150 rpm) in Sabouraud rich medium. To induce capsule growth, 20 l of this starter culture was used to inoculate larger cultures grown at 37°C in minimal medium, which is composed of glucose (15 mM), MgSO 4 (10 mM), KH 2 PO 4 (29.4 mM), glycine (13 mM), and thiamine-HCl (3 M), pH 5.5 (70). Alternatively, capsule induction was performed in 10% Sabouraud medium buffered at pH 7.3 with 50 mM MOPS (71). Capsular polysaccharide was purified as described previously (72). Briefly, fungal cells were collected by centrifugation and washed three times in distilled water. Washed cell pellets were suspended in 15 ml of dimethyl sulfoxide (DMSO) and incubated at room temperature for 2 h with periodic agitation. Samples were then centrifuged, and the DMSO supernatant was extensively dialyzed against distilled water for 48 h with periodic replacement of the dialysate. Dialyzed samples were then lyophilized, yielding purified capsular polysaccharide. Extracellular polysaccharide (exo-PS) was purified by ultrafiltration as described previously (73). Briefly, fungal cells were pelleted by centrifugation at 4000 ϫ g, and the supernatant was filtered through a 0.22-m filter to remove any cells. Supernatant was then concentrated using an Amicon (EMD Millipore, Danvers, MA) ultrafiltration cell (100-kDa cutoff; 400-ml capacity) with stirring and Biomax polyethersulfone ultrafiltration discs (63.5 mm diameter) with a nitrogen stream used as the pressure gas. The viscous layer formed on the ultrafiltration disc was collected and lyophilized to yield purified exo-PS above the 100-kDa molecular mass cutoff. This process was repeated with the flow-through using a 10-kDa ultrafiltration disc and again with a 1-kDa ultrafiltration disc to yield exo-PS fractions of different molecular mass ranges.
Capture ELISAs for Measuring Cryptococcal Polysaccharide-C. neoformans GXM was measured by capture ELISA as described previously (74). Briefly, microtiter polystyrene plates were coated with goat anti-mouse IgM at 1 g/ml (SouthernBiotech, Birmingham, AL) and then blocked with 1% bovine serum albumin in PBS. The murine anti-GXM IgM mAb 2D10 was then added as a capture antibody at 10 g/ml. Next, unknown samples and a GXM standard at 1 g/ml were added to each plate and serially diluted. Completion of the assay was performed by successively adding the anti-GXM IgG1 mAb 18B7 at 10 g/ml and then alkaline phosphatase-labeled goat anti-mouse IgG1 at 1 g/ml (SouthernBiotech). Incubation at each step was performed for 1 h at 37°C or overnight at 4°C, and plates were washed three times between steps with 0.1% Tween 20 in Tris-buffered saline. The plates were developed with 1 mg/ml p-nitrophenyl phosphate in substrate buffer, and absorbance was measured at 405 nm. Standard curves were fit with a 4-parameter logistic regression model, and unknown GXM concentrations were interpolated from dilutions within the linear portion of the standard curve. Statistical analysis was performed in GraphPad Prism.
Capsule Shedding Experiments-The amount of cryptococcal polysaccharide shed from fungal cells was investigated in the presence of varying concentrations of antibody. C. neoformans cultures were grown to stationary phase under capsule-induc-ing conditions in Sabouraud medium diluted 1:10 in 50 mM MOPS. Cells were collected and heat-killed at 60°C for 1 h. Complete killing was confirmed by plating on solid Sabouraud medium. Heat-killed cells were washed 5-10 times with PBS to remove any unattached polysaccharide and collected by ultracentrifugation (15,000 ϫ g, 30 min, 4°C) until aliquots of the wash were confirmed to have no detectable cryptococcal polysaccharide by capture ELISA. Washed cells were suspended in a small volume of PBS, which was divided equally between the five capsule-shedding conditions. Each sample was diluted with PBS, pH 7.3, and 0.1% sodium azide to a final volume of 20 ml and a concentration of ϳ10 6 heat-killed cells/ml. Equal volumes of PBS, mAb 18B7, or a control IgG1 (MOPC-31C, Sigma) were added to each sample for the necessary final antibody concentration. Samples were then incubated at 37°C with agitation. Aliquots of 1 ml were collected from each sample at time 0 and at periodic time points for up to 40 days. All cell and wash aliquots were first centrifuged at 7000 ϫ g for 10 min to pellet any cells, and then the supernatant was filtered through a 0.22-m cellulose acetate microcentrifuge tube filter (Corning, Corning, NY) at 1000 ϫ g for 4 min. Filtered aliquots were stored at Ϫ20°C and analyzed at the same time, each sample in triplicate, for cryptococcal polysaccharide content by capture ELISA. Fungal capsules were visualized at day 0 and at periodic time points under light microscopy by mixing 2 l of india ink with 6 l of cells with an Olympus AX 70 microscope using a ϫ40 objective. Statistical tests were performed with GraphPad Prism.
Capsule Immunofluorescence-Cell aliquots collected from the capsule shedding experiment following a 39-day incubation with 18B7 (unlabeled) were thawed at room temperature and incubated with 5 g/ml mAb 18B7 directly conjugated to Alexa-568 (Molecular Probes, Inc.) and 1 g/ml Uvitex2b (specific to chitin in the cell wall) (Polysciences, Warrington, PA) in blocking solution (1% bovine serum albumin in PBS). Labeling incubation was done in microcentrifuge tubes under continuous rocking at room temperature for 1 h followed by three washes with PBS. Flow cytometry analysis of stained cells was done using a Beckman Coulter MoFlo XDP flow cytometer violet 405 nm, blue 488 nm, and yellow 561 nm lasers. Histograms for 18B7-A568 binding are presented as a percentage of the maximum count (modal option), Ͼ15,000 counts per sample. Data were analyzed with FlowJo software. Imaging of stained cells was done with an Olympus AX 70 microscope using an oil immersion ϫ100 objective. Samples were mounted onto glass coverslips with india ink counterstain for capsule visualization under bright field. Fluorescent filters for DAPI or rhodamine were used to image the cell wall or 18B7-A568 capsular staining using the same exposure time for all samples (80 and 125 ms, respectively).
Dynamic Light Scattering Measurements-Polysaccharide samples for DLS measurements were prepared from each of the five capsule shedding conditions described above. Approximately 7 ml of supernatant from each condition was filtered through a 0.22-m syringe filter to remove any cells. Exactly 6 ml of each supernatant sample as well as an equal volume of PBS as a negative control was lyophilized and resuspended overnight in 150 l of pure water for a 40-fold increase in concen-tration. The absence of any antibody in the samples was confirmed via a direct ELISA for IgG1. Samples were centrifuged at 10,000 ϫ g for 10 min to pellet any precipitate, and 50 l of supernatant was analyzed by DLS. ED and polydispersity of each sample were determined as described previously through quasi-elastic light scattering in a 90Plus/BI-MAS MultiAngle Particle Sizing analyzer (Brookhaven Instruments, Holtsville, NY) (70). Fluctuations in the intensity of scattered light over time originating from the random motions of particles in solution were processed by the autocorrelation function C(t), where t is the time delay. Particle size was determined by relating diameter to the translational diffusion coefficient. Polydispersity is the relative standard deviation of particle sizes in the sample. The ED and polydispersity were used to generate a log normal distribution of particle sizes for each sample. The multimodal size distributions of particles were obtained by a nonnegatively constrained least squares algorithm based on the intensity of light scattered by each particle. Mean ED was calculated from 10 individual measurements for each sample. The correlation function, polydispersity, log normal distribution, and multimodal size distribution for each sample were calculated from the combined data of all 10 measurements. Statistical tests were performed with GraphPad Prism.
Structural Analyses and IgMotif Template Algorithm-All scripts and raw data relating to the template algorithm described in this work are available and annotated in the Zenodo database (http://doi.org/10.5281/zenodo.192184). Unless otherwise noted, all analyses were developed using the R statistical language (version 3.3.1, R Foundation for Statistical Computing, Vienna, Austria). All published structures used in this work were obtained from the complete PDB database of 119,857 non-obsolete and non-empty structures as of July 9, 2016. The complete list of antibody-containing structures could not be accurately determined by a simple database web search because antibody structural annotations in the PDB are frequently inaccurate or incomplete, and many non-antibody structures possess annotated antibody-like domains or immunoglobulin folds. Instead, the antibody structure set was determined by performing a Needleman-Wunsch sequence alignment between each unique protein chain from all PDB structures and the combined set of all 2191 V region sequences from 10 different species in the International Immunogenetics Information System (IMGT) reference directory (75). PDB query chains consisted of the N-terminal 139 or fewer amino acids of sequences Ͼ86 amino acids in length. The IMGT reference V region sequences ranged in length from 106 to 119 amino acids. The top scoring alignment for each PDB query chain was recorded, and a cutoff score of Ϫ350 was determined for both V H and V L sequences by inspecting the resulting distribution and ensuring that non-antibody Ig-like sequences (e.g. T cell receptors) fell well below the cutoff, resulting in 2511 antibody-containing structures. Structures with duplicate antibody sequences were then removed from the dataset leaving 1780 unique antibody structures. The final set of 1665 antibody structures was obtained after removing 115 additional entries because they were found to contain incomplete side chain or only backbone atomic coordinates and thus could not be used to generate structural templates for superposition. From this set, 63 structures were confirmed to be known catalytic antibodies by performing a PDB query for the terms "catalytic antibody," "catalytic antibodies," "hydrolytic antibody," "hydrolytic antibodies," "catalytic immunoglobulin," "hydrolytic immunoglobulin," and "abzyme" followed by manually inspecting the literature for each result. Enzyme classification (EC) numbers listed in the PDB were used to generate the hydrolase dataset and to classify hydrolases (EC 3) as either glycosylases (EC 3.2) or serine (EC 3.4.21), cysteine (EC 3.4.22), threonine (EC 3.2.23), metalloproteases (EC 3.4.24) or aspartic proteases (EC 3.4.25). Enzyme structures containing antibody chains, duplicate sequences, or incomplete side chain or backbone coordinates were excluded.
All possible three-residue structural templates were generated for each analyzed PDB file. Templates were required to contain one acidic residue (His, Asp, Glu), one basic residue (His, Lys, Arg), and one nucleophilic residue (Ser, Cys, Thr, Tyr). Additionally, the distances between active atoms in the base-acid and base-nucleophile pairs were required to be Յ11 Å. Templates for each three-residue motif meeting these requirements were constructed with the three-dimensional coordinates for five backbone atoms (CA, CB, O, C, and N) and three side chain points (geometric center, center of mass, and a designated residue-specific side chain atom) per residue for a total of 24 points per template, regardless of the particular amino acids in the motif. In the case of NMR data, only the first model was considered. When calculating mean templates, only structures with a resolution Յ2.5 Å were considered. Comparison of identified three-residue motifs was performed by calculating both the r.m.s.d. and the d.r.m.s. between paired templates. Following optimal superposition of two 24-point templates, (s 1 , …, s 24 ) and (v 1 , …, v 24  Within each template, distances were calculated for each of the eight points per residue between the base-acid (ba), basenucleophile (bn), and acid-nucleophile (an) pairs. The d.r.m.s. between two templates was defined as the r.m.s.d. between their two distance sets, d s ϭ (s ba1 , …, s ba8 , s bn1 , …, s bn8 , s an1 , …, and s an8 ) and d v ϭ (v ba1 , …, v ba8 , v bn1 , …, v bn8 , v an1 , …, and v an8 ). Motifs were classified as matches if either the r.m.s.d. or d.r.m.s. measurements between the two templates was Յ1 Å. This cutoff criterion was chosen by inspecting the clustering of resultant similarity measurements and by comparing the ability of r.m.s.d. and d.r.m.s. criteria to discriminate between three-residue motifs.
Sequence Alignment and Phylogenetic Analysis-V L sequences for 61 known catalytic antibody structures and a random sample of 61 antibodies without catalytic activity were aligned using the Clustal Omega multiple sequence alignment tool with default parameters (76). The Shannon entropy was calculated at each position in the aligned sequences, and scores were normalized to 1 for low entropy residues and 0 for high entropy residues using the Bio3D R package (77). Of the 1665 analyzed antibody structures, phylogenetic trees were gener-ated for 1498 V L sequences and 1594 V H sequences based on the pairwise Levenshtein distance between sequences. The germ line gene family and species of origin for each sequence were inferred based on the highest scoring IMGT reference gene as described above. Sequences were further classified by whether they were known to have catalytic activity and whether they contained the 3E5 putative catalytic motif, as determined by the IgMotif template algorithm. Plots of the phylogenetic trees and color-coded classifications were generated using the circlize R package (78). All scripts and raw data used in these analyses are available and annotated on line (79).
M2 Antibody and FLAG Hydrolysis-The anti-FLAG M2 antibody and the FLAG peptide were obtained commercially (Sigma). FLAG peptide at a concentration of 200 M was incubated either alone or with 960 g/ml M2 mAb at 37°C for 5 days. Aliquots of each sample were taken on day 0 and day 5 and stored at Ϫ20°C until analysis by MALDI-TOF MS. Samples were spotted after a 20-fold dilution in MALDI matrix, which was prepared with 5 mg/ml ␣-cyano-4-hydroxycinnamic acid in 50% acetonitrile, 0.1% trifluoroacetic acid. Spectra were obtained at the Johns Hopkins Mass Spectrometry and Proteomics Facility on a Voyager-DE STR Biospectrometry work station (Applied Biosystems, Foster City, CA) in reflector mode and analyzed with the Data Explorer software (version 4.8, Applied Biosystems). Spectra figures were generated using GraphPad Prism.