Molecular and structural properties of three autoimmune IgG monoclonal antibodies to histone H2B.

In systemic autoimmune diseases such as lupus the immune system produces autoantibodies to nuclear antigens including DNA and histone molecules. In the present study, we describe three monoclonal IgG antibodies that have been obtained from lupus-prone MRL/lpr mice. These three antibodies react with the amino terminus of histone H2B, a region of the molecule that is accessible in chromatin. Using a series of overlapping H2B synthetic peptides and structural analogues, we have mapped the different epitopes recognized by these antibodies. We have also sequenced the combining sites (variable regions) of the antibodies and modeled their interactions with the corresponding epitopes. Overall, the data suggest that the mechanisms of interaction with antigen are different for each of the three antibodies, even though they all react with the amino-terminal domain of the histone H2B molecule. The results also suggest that the binding between these antibodies and histone H2B is different from that between most antibodies and conventional protein antigens since the heavy chain complementarity-determining region 3 appears to play only a limited role in the three antibodies tested. The study of the interaction between self-antigens and spontaneously occurring autoantibodies may help us elucidate the mechanisms driving the expansion of self-reactive lymphocytes.

In systemic autoimmune diseases such as lupus the immune system produces autoantibodies to nuclear antigens including DNA and histone molecules. In the present study, we describe three monoclonal IgG antibodies that have been obtained from lupus-prone MRL/ lpr mice. These three antibodies react with the amino terminus of histone H2B, a region of the molecule that is accessible in chromatin. Using a series of overlapping H2B synthetic peptides and structural analogues, we have mapped the different epitopes recognized by these antibodies. We have also sequenced the combining sites (variable regions) of the antibodies and modeled their interactions with the corresponding epitopes. Overall, the data suggest that the mechanisms of interaction with antigen are different for each of the three antibodies, even though they all react with the amino-terminal domain of the histone H2B molecule. The results also suggest that the binding between these antibodies and histone H2B is different from that between most antibodies and conventional protein antigens since the heavy chain complementarity-determining region 3 appears to play only a limited role in the three antibodies tested. The study of the interaction between self-antigens and spontaneously occurring autoantibodies may help us elucidate the mechanisms driving the expansion of self-reactive lymphocytes.
In systemic autoimmune diseases such as systemic lupus erythematosus (SLE), 1 self-reactive B cells produce autoantibodies against nucleosomes, the building blocks of chromatin. The nucleosome core particle is composed of a central octamer of two molecules of each of the histones H2A, H2B, H3, and H4, surrounded by approximately two turns of DNA (1). The core histones interact among themselves via their central globular domains, whereas the positively charged amino termini of these histones are accessible on the external surface of the nucleosome core particle (1). Histone H1 binds outside the core particle to the linker DNA (20 -60 base pairs), joining adjacent core particles to form the nucleosomal array.
SLE autoantibodies can recognize a diverse array of epitopes located on the surface of the nucleosome. Some SLE antinuclear antibodies can recognize multimolecular determinants, whereas other antibodies are directed against individual components such as DNA or histones, reviewed in (2). Among antibodies to individual core histones, those reacting with histone H2B are a particularly frequent specificity (3)(4)(5)(6). Antibodies to H2B are also encountered in other autoimmune diseases, such as systemic sclerosis (7) as well as in human immunodeficiency virus-infected individuals (8). Anti-H2B antibodies in SLE patients almost always recognize determinants that are located in the amino terminus (residues 1-25) of the histone molecule (9). Several mouse strains, such as the autoimmune MRL/lpr mice, spontaneously develop a syndrome with autoantibody production and glomerulonephritis that resembles SLE (10). In the present study, we report the molecular and structural characteristics of three IgG monoclonal antibodies to histone H2B that have been generated from nonimmunized MRL/lpr mice.

EXPERIMENTAL PROCEDURES
Monoclonal Antibody (mAb) Generation-The generation of the hybridomas was performed using standard techniques. Splenocytes from spontaneously autoimmune (nonimmunized) MRL/lpr mice were fused with the nonsecreting plasmocytoma SP2/0, and culture supernatants were screened by enzyme-linked immunosorbent assay (ELISA) for reactivity with total histones (11). Positive hybridomas were cloned twice by limiting dilution, and isotypes were determined using antiisotype reagents from Roche Molecular Biochemicals. mAbs were purified from culture supernatants by affinity chromatography on a recombinant protein G-Sepharose column (Amersham Pharmacia Biotech). Three anti-H2B IgG mAbs were independently generated from three nonimmunized MRL/lpr mice. Some properties of two of these mAbs, LG2-2 (IgG2a) and LG11-2 (IgG2a), have been previously reported (6,(12)(13)(14), whereas the third mAb, PR1-1 (IgG2b), is entirely new to this study.
Nucleosomes and Synthetic Peptides Used as Antigens-Nucleosomes were prepared from calf thymus as described previously (15) and purified on a 5-29% (w/v) sucrose gradient. The nucleosome preparations were characterized by 1.5% agarose gel electrophoresis, and the histone content was checked by SDS-polyacrylamide gel electrophoresis.
Thirteen synthetic peptides covering different segments within the region 1-25 of H2B were used in this work (Table I). They were synthesized using Fmoc chemistry as described previously (16). Calf thymus sequences were used for all syntheses. An additional cysteine residue was added to some peptides to allow their specific conjugation to carrier protein. In one series of analogues, the NH 2 and COOH termini of the peptides were acetylated and carboxamidated, respectively. These blocked peptides were assembled using Fmoc chemistry on a Fmoc-2,4-dimethoxy-4Ј-(carboxymethyloxy)benzhydrylamine resin. Peptides were purified by reversed-phase high performance liquid chromatography (HPLC) using a Perkin-Elmer preparative HPLC system on an aquapore ODS 20-m column (100 ϫ 10 mm) (17). The elution was achieved by a linear gradient of aqueous 0.1% trifluoroacetic acid (A) and 0.08% trifluoroacetic acid in 80% acetonitrile, 20% water (B) at a flow rate of 6 ml/min with UV detection at 220 nm. The homogeneity of each peptide was assessed by analytical HPLC on a Beckman instrument with a nucleosil C 18 , 5-m column (4.6 ϫ 150 mm) using a linear gradient of 0.1% trifluoroacetic acid in water and acetonitrile containing 0.08% trifluoroacetic acid at a flow rate of 1.2 ml/min. Peptide identity was established by mass spectrometry using a protein time-offlight apparatus (Bruker Spectrospin, Wissembourg, France). Synthesis of control peptide 44 -67 of the SmD1 protein was described previously (18). Peptide 1-13 C was conjugated to ovalbumin using m-maleimidobenzoyl N-hydroxysuccinimide ester.
ELISA-The indirect ELISA procedure used to measure the binding of mAbs to nucleosome was as described previously (15) using microtiter plates (Falcon, Oxnard, CA; catalog number 3912) coated overnight with 50 -400 ng/ml mononucleosomes in phosphate-buffered saline (PBS) pH7.4. To study mAb binding to H2B synthetic peptides, ELISA plates were coated overnight at 37°C with each peptide diluted in 0.05 M carbonate buffer, pH 9.6. In each assay, mAbs were also tested in a noncoated well incubated with coating buffer as a control. Saturation of plates was obtained by adding PBS containing 0.05% Tween (PBS-T) and 0.4% bovine serum albumin. The subsequent steps of the test were performed as described previously (12) using mAbs diluted in PBS-T, 0.4% bovine serum albumin and rabbit anti-mouse IgG conjugated to horseradish peroxidase diluted 1:5000 in PBS-T. The final reaction was visualized by addition of 3,3Ј,5,5Ј-tetramethyl benzidine in the presence of H 2 O 2 . Inhibition experiments were performed as described previously (12) using peptide 1-25 H2B directly coated on the plastic at the appropriate concentration (0.1, 0.125, or 0.4 M), and the mAbs were incubated for 1 h at 37°C and 1 h at 4°C with the various competitor peptides (maximal peptide concentration tested in the fluid phase, 10 M). The reaction was revealed as described above. IC 50 values are defined as the amount (expressed in M) of competitor peptide necessary to inhibit the maximal antibody binding by 50%.
Variable Region Messenger RNA Sequencing-Poly(A) ϩ RNA was isolated from hybridoma cells using oligo(dT)-cellulose according to a previously described method (19). The variable region nucleotide sequences were determined by dideoxy sequencing as previously reported (20 -22). The nucleic acid sequences were compared with immunoglobulin sequences in the GenBank data library (23). The variable region sequences were assigned to known gene families or germline genes by comparison with previously published sequences.
Modeling-All molecular modeling was performed on either a Silicon Graphics Personal IRIS 4D/25 or a Silicon Graphics Indigo2 workstation. Initial models were constructed using XABgen, a suite of programs for homology-based antibody model generation (24,25). A model of the antigen-binding site for each antibody, formed by the six complementarity-determining region (CDR) loops, was constructed using the heavy and light chain variable region sequences for each antibody. The output of the XABgen program, a model of the antigen-binding site in Protein Data Bank format, was used as the initial starting point for all further modeling.
Refinement of all molecular models was performed using the DRE-IDING II force field and biograf software (BIOSYM/Molecular Simulations, Can Diego CA). The initial step in the refinement process involved minimizing the energy of the entire structure to convergence. After this, each structure was subjected to molecular dynamics calculations (200 ps), during which time only the atoms within the 6 CDR loops were permitted to move, while the motion of all the other atoms in the model was maintained constant. The energy of the entire structure was again minimized to convergence following the dynamics calculations.
The next step in the process was to construct models of the antigenic peptides of interest, namely peptides 1-13 and 1-25 of H2B. Peptide models were generated in a manner similar to that used for generating a model of human immunodeficiency virus coat protein gp120 (26). In general, a peptide was constructed using the peptide builder within the main biograf program, the energy of the resulting structure was minimized to convergence, and molecular dynamics calculations were performed on the entire peptide structure, again followed by energy minimization.
The final step in our modeling procedure involved docking each peptide into the proposed binding site of the relevant antibody. Refinement of each antigen-antibody complex included the following. 1) The energy of the entire complex was minimized to convergence; 2) molecular dynamics calculations (100 ps) were performed, during which time only the atoms of the 6 CDR loops, and those of the peptide antigen were permitted to move, while the motion of all other atoms within the structure were maintained constant; 3) the energy of the entire complex was then again minimized.
During the molecular dynamics calculations, the atomic positions of the peptide antigen and of the six CDR loops on each antibody were permitted to move. A comparison of the structure of the peptide antigen bound to its target before and after extensive molecular dynamics calculations resulted in an average root mean square difference of 5 Å. Thus, the peptide conformation within the antigen-antibody complex was significantly different from the original peptide structure indicating that the molecular dynamics procedure resulted in a sampling of the available peptide conformational space.

RESULTS AND DISCUSSION
In preliminary experiments (data not shown), we determined that the three mAbs described in this study (PR1-1, LG2-2, and LG11-2) specifically react with histone H2B but not with the other core histone molecules (H2A, H3, and H4) nor with DNA. Direct ELISA tests with overlapping peptides encompassing the whole H2B molecule indicated that all three mAbs recognized epitopes in the amino-terminal peptide 1-25 (data not shown). We therefore further characterized the reactivity of these mAbs against the histone H2B amino terminus. mAbs PR1-1 and LG2-2 strongly reacted with the amino-terminal peptide 1-13 but not with peptide 6 -18. In contrast, mAb Microtiter plates were coated with 1 M peptide and allowed to react with mAbs PR1-1 (1 g/ml), LG2-2 (0.5 g/ml), and LG11-2 (0.5 g/ml). Anti-mouse IgG-peroxidase conjugate was diluted 1:5000. Peptide 1-13(C) was used as an ovalbumin conjugate (carrier to peptide molar ratio, 1:8

P E P A K S A P A P K K G (C)
1-20 LG11-2 did not bind peptide 1-13 but strongly reacted with peptide 6 -18 (Table II). We then tested the ability of various H2B amino-terminal peptides in solution to inhibit the interaction between the three mAbs and peptide 1-25 immobilized on the ELISA plate. As shown in Table III, the binding of PR1-1 mAb to peptide 1-25 was inhibited by peptides 1-25, 1-20, and 1-13 but not at all by peptides 1-6, 6 -18, and 13-18, indicating thus that residues 1-13 are required for antibody recognition and that residues surrounding the serine residue at position 6 (Ser 6 ) are important. Like PR1-1, LG2-2 binding was strongly inhibited by peptides 1-25, 1-20, and 1-13 but moderately by peptide 1-6 and not at all by peptides 6 -18 and 13-18, indicating that residues 1-6 (with a few residues in the segment 7-13) constitute most of the epitope recognized by this mAb. Similar experiments were performed with LG11-2 (Table  III). The binding of this antibody to peptide 1-25 was inhibited by peptides 1-25, 1-20, and 6 -18 but not by peptides 1-13 and 13-18, therefore confirming that the central segment of the 1-25 peptide (residues 6 -18) contains the major part of the LG11-2 epitope. No reaction was observed with any of these three antibodies against the unrelated control peptide 44 -67 of SmD1 (Table III). The epitopes recognized by the anti-H2B mAbs in our study are present on "native" structures, since all three mAbs react with purified calf thymus mononucleosomes when tested by ELISA (Table IV). These determinants are also accessible in chromatin, since the mAbs reacted with HEp-2 nuclei by immunofluorescence staining (not shown).
The H2B amino terminus is characterized by the presence of several positively charged lysine residues that are probably important for interaction with DNA in the nucleosome. We have therefore evaluated the role of two of these residues, Lys 5 and Lys 11 , by replacing them with alanine side chains in mutant 1-20 peptides. We also tested the importance of the first proline residue (Pro 1 ) by similarly replacing it with an alanine. The results of the antibody inhibition binding assays to these modified peptides are listed in Table III. With respect to the two mAbs specific for peptide 1-13, the binding of PR1-1 is totally abolished when Pro 1 is substituted with Ala 1 and moderately affected when Lys 11 is replaced with an alanine residue, whereas the Lys to Ala 5 substitution has no effect on PR1-1 binding. As with PR1-1, LG2-2 reactivity is abolished by the Pro to Ala 1 substitution and only moderately affected by the Lys to Ala 11 replacement. In contrast with PR1-1, LG2-2 binding is only partially decreased by the Lys to Ala 5 replacement. Previous results indicated that LG11-2 is specific for a determinant located between residues 6 and 18, and as expected, alanine substitution at position 1 (Pro 1 to Ala 1 ) does not affect binding. The Lys to Ala 5 substitution reduces LG11-2 binding only slightly, whereas reactivity is totally abolished by the Lys to Ala 11 substitution.
We further characterized the reactivity of LG11-2 mAb (which reacts with residues located within the segment 6 -18 of histone H2B) using various analogues of the parent peptide 6 -18 (Table I). These peptide analogues were acetylated at their NH 2 termini and carboxamidated at their COOH termini, and three of them (named R1, R2, and R3) also included se-quence alterations from the original 6 -18 sequence. Blocked peptides can assume a three-dimensional conformation different from conventional peptides that may result in alteration of antibody recognition (27,28). The results of these binding studies are shown in Table V. As expected from the results described earlier in this paper, neither PR1-1 nor LG11-2 reacted with any of the 6 -18 analogs, since both antibodies require residues 1-6 for binding. In contrast, LG11-2, which reacts with peptide 6 -18, also recognized the blocked 6 -18 peptide (Table V). However, LG11-2 did not bind to any of the analogue peptides, R1, R2, or R3 (Table V). Of these three analogues, R1 is the closest to the original sequence since the only differences between R1 and 6 -18 occur at the NH 2 and COOH extremities, whereas the core sequence APKKGSKKA (residues 9 -17) is identical in both peptides (Table I). The lack of binding of LG11-2 to this analogue suggests that the terminal residues of the 6 -18 peptide may be involved in direct contact with the LG11-2 binding site or that some of these residues (e.g. Pro 8 , which was replaced in R1 by an acetylated alanine) play a critical role in the overall conformation of the epitope.

TABLE VI V gene and gene segment use in anti-H2B monoclonal antibodies
All three mAbs were obtained from independent fusions from three different MRL/lpr mice. The D segment of mAb LG11-2 is only 6 nucleotides long and too short for its germline gene to be identified. The PR1-1 D segment is coded by the Q52 gene that has rearranged in inverted orientation.  Inhibition experiments were conducted with the three anti-H2B mAbs in the fluid phase: PR1-1 (1 g/ml), LG2-2 (0.5 g/ml), LG11-2 (0.5 g/ml), and peptide 1-25 absorbed on the ELISA plate (0.4 M in the case of PR1-1 and 0.1 M in the cases of LG2-2 and LG11-2). Various H2B and analogue peptides were added as inhibitors in the fluid phase. The results are expressed in terms of IC50 defined as the M amount of competitor peptide necessary to inhibit maximal antibody binding by 50%. We then examined the heavy and light chain variable (V) region sequences of the three anti-H2B mAbs. The sequences of these V regions are available from GenBank/EMBL under accession numbers X67621, X67624, AF143907, AF143908, AF143909, and AF143910. To our knowledge, these are the only anti-H2B antibody variable region sequences to have been reported. The V, D, and J gene segment usage of our mAbs is summarized in Table VI, and the amino acid sequences of the CDRs are listed in Table VII. The variable regions of these mAbs are encoded by different D, J H , V, and J gene segments, and although the J558 V H gene family is used by all three mAbs, they utilize three different germline V H genes. Therefore, even though all three mAbs recognize determinants located in the amino terminus of histone H2B, they use different genetic elements to create their combining sites. To help us understand the interactions between these three mAbs and the H2B amino terminus, we modeled the structures of the three antibody-antigen complexes. Despite the uncertainties associated with these computational procedures, the predicted structures are a valuable aid, as they can provide a possible explanation for the binding interactions that have been experimentally determined. The structural elements used to generate the anti-H2B variable region models are listed in Table VIII, and the results of these modeling studies are depicted in Fig. 1.
In the following descriptions of these interactions, the variable region residues are numbered on the line with the residue abbreviations according to the Kabat nomenclature (29), and the H2B peptide residues are numbered in superscript characters (see Table I for the H2B sequence). Both PR1-1 and LG2-2 react with the peptide 1-13 of H2B. For PR1-1 (Fig. 1, A and  B), the heavy chain interactions involve Tyr-27 (located just outside of CDR1) with Pro 3 , CDR2 Asp-52 with Lys 11 and Lys 12 , and CDR2 Glu-56 with Lys 11. Only one residue in the heavy chain CDR3 is involved, namely Asp-101 with Pro 1 . The light chain CDR1 is not implicated, whereas several CDR2 residues are involved, Tyr-49 with Lys 5 , Trp-50 with Pro 8 and Pro 10 , Phe-55 with Pro 1 , and Thr-56 with Glu 2. From the light chain CDR3, only Tyr-91 interacts with Ser 6.
For LG11-2, which reacts and was modeled with the peptide 1-25 of H2B, the main interactions of the heavy chain CDRs were with lysine residues at the center of the peptide, with Thr-28 and Thr-30 from CDR1 interacting with Lys 16 , Asp-31 from CDR1 interacting with Lys 11 , Lys 16 , and Lys 20 , whereas Asp-52 from the CDR2 interacts with Lys 11 (Fig. 1, E and F). The light chain CDRs interact with determinants located at both peptide termini. Asp-55 and Tyr-49 from CDR2 interact with Pro 1 , whereas Tyr-32 from CDR1 interacts with Lys 5. At the carboxyl terminus, interactions involve charged residues. From the light chain CDR1, Asp-27 interacts with Lys 24 , Asp-28 with Lys 24 and Asp 25 , Lys-30 with Asp 25 , whereas Lys-53 from the CDR2 also interacts with Asp 25. In our model, neither the heavy nor the light chain CDR3 plays any significant role in peptide binding.
Despite the limitations and uncertainties inherent to the modeling process, these studies are in agreement with the results obtained by ELISA, indicating that Pro 1 is required for the binding of both LG2-2 and PR1-1 and that Lys 11 is important for the optimal binding of all three antibodies. Interestingly, the replacement of Lys 5 with an alanine residue had moderate or no effect on antibody reactivity. This is also in good agreement with the model that indicates that this residue is only involved in distant hydrogen bonds, and this also suggests that the Lys to Ala 5 replacement does not dramatically affect the overall structure of the 1-20 H2B peptide recognized by the set of antibodies studied.
A remarkable and unexpected feature of these modeling studies is that they suggest that the heavy chain CDR3s play only a limited role in direct binding to the histone H2B peptide (although these CDR3s can still play an important role by contributing to the overall V region structure). The heavy chain CDR3 is extremely diverse, since it results from the rearrangement of three different gene segments (V H , D, and J H ) and from the extensive joining flexibility that takes place during this rearrangement process. The heavy chain CDR3 is centrally located within the antibody-combining site and is critical for antigen binding. For instance, in anti-chromatin autoantibodies directed against DNA or DNA-histone complexes, the heavy chain CDR3 often contains multiple cationic residues (usually arginine) that are critical for contact with the negatively charged DNA (reviewed in Refs. 2 and 30). For two of our mAbs (LG11-2 and PR1-1), the limited role of heavy chain CDR3 can be in part attributed to the very short lengths of their heavy chain CDR3s (4 residues). In contrast to these two mAbs, a survey of mouse heavy chain CDR3 sequences indicated that   mAb  HCDR1  HCDR2  HCDR3  LCDR1  LCDR2  LCDR3   PR1-1  SYWMH  KIDPSDSETHYNQKFKD  PLDY  KASQNVGTDVS  WASNRFT  EQYSSYPLT  LG2-2  SYVMY  YINPYNDGTKYNEKFKG  PGDGYPFDY  RSSQSIVHSNGNTYLE  KVSNRFS  FQGSHVPYT  LG11-2  DYDMH  AIDPETGGTAYNQKLKG  EVDY  KSSQSLLDSDGKTYLN  LVSKLDS  WQGTHFPWT their average length is 8.7 (31). Our results are reminiscent of studies conducted with several anti-lysozyme antibodies that contained heavy chain CDR3s of various lengths (32)(33)(34)(35). These observations indicated that the involvement of the heavy chain CDR3 in lysozyme binding is proportional to its length. In contrast, as the heavy chain CDR3 shortens, the heavy chain CDR1 and CDR2 become gradually more involved and interact with a greater number of residues on the lysozyme molecule.
Our analysis therefore supports the view that the length of the heavy chain CDR3 can dramatically influence the importance of the interactions of the other regions of the combining site.
The results obtained in this study are important for our understanding of autoimmune phenomena. They support previous studies indicating that the autoantibodies to histone H2B are mostly directed against a unique segment of the molecule, namely the amino terminus domain. Our data also confirm FIG. 1. Three-dimensional structural models of the interaction between the three anti-H2B mAbs, PR1-1 (A and B), LG2-2 (C and  D), LG11-2 (E and F), and their corresponding histone H2B peptides, 1-13 (A, B, C, and D) and 1-25 (E and F). Each antibody-peptide interaction is represented both by a side view (A, C, and E), where the peptide is on the left and the combining site on the right, and by a top view (B, D, and F), where the peptide lies on top of the combining site. Color codes are as follows. Purple, peptide. Antibody light chain: yellow, framework; white, CDR1; red, CDR2; yellow, CDR3. Antibody heavy chain: cyan, framework; green, CDR1; cyan, CDR2; blue, CDR3. that, despite its limited size, the H2B amino terminus includes several overlapping self-epitopes. There is a general consensus that self-reactive B cells in SLE are directly stimulated by nuclear antigens (2, 30) and there is also increasing evidence that these self-antigens can be presented to the immune system as a result of the apoptotic process (36). Therefore, these results suggest that the histone H2B amino-terminal domain is particularly exposed and immunogenic in chromatin that has been fragmented and adulterated following apoptotic cell death. Our study also helps us to understand the interactions that take place among the various CDR residues in the antibody variable regions and the sequential H2B epitope(s). The immune response to a linear determinant of histone H2B is particularly unique since other autoantibody responses in SLE often involve complex, multimolecular epitopes (37). Future site-directed mutagenesis studies of the anti-H2B autoantibodies CDRs will allow us to elucidate the molecular basis of these interactions.