The idiotypic network in the regulation of autoimmunity: Theoretical and experimental studies.

The regulation of autoimmunity is a key issue in fundamental immunology. Despite outstanding achievements on this front, we currently have more questions than answers. The idea of an immune network as a regulatory mechanism is quite attractive, since it enables us to explain the selectivity (specificity), and moreover the clonality, of the regulation. Nevertheless it remains unclear how this mysterious network of immune cells is organized, how it operates, and how it exerts control over autoimmunity. This article presents an attempt to understand how the immune network functions and how it controls autoreactivity. We present a mathematical model of the immune network that is based on principles of immune network organization and function that we arrived at from a survey of the available literature. To test the principles on which the mathematical model is based, we studied the model and compared the different responses to antigen that it generated with the results obtained from experimental studies of immune response. The modeled kinetics of idiotype and anti-idiotype in response to the administration of antigen are in good agreement with the experimental kinetics of idiotypic and anti-idiotypic antibodies. To obtain evidence of the existence of idiotypic mechanisms for regulating autoimmunity, we studied a mathematical model containing autoclones and compared the model results with data from experimental studies in a model of autoimmune hemolytic anemia in mice. Because the results from the theoretical and the experimental studies coincide, there is justification to conclude that autoreactive lymphocytes are normal components of the immune network within which they are regulated. We discuss a possible molecular/cellular mechanism for negative control of autoreactive cells as affected by anti-idiotypic antibodies.

Phosphocholine-binding antibody activities are hierarchically encoded in the sequence of the heavy-chain variable region: dominance of self-association activity in the T15 idiotype.

A methodology based on the representation of each amino acid of a protein sequence by the electron-ion interaction potential and subsequent analysis by signal processing was used to determine the characteristic or common frequency (in Hz) that reflects the biological activity shared among phosphocholine (PC)-binding antibodies. The common frequency for the variable portion of the heavy chain (VH) of the PC-specific antibodies is found to be at f = 0.37 Hz. The VH sequences of the PC-binding antibodies exhibit three subsites for the PC moiety where hypervariable region 2 (CDR2) plays a role in the interaction with the phosphate group. Mutations in this VH region have an impact on the ability of mutant variants to bind PC and its carrier molecule, as well as on the characteristic frequency shift toward f = 0.12 Hz for mutants failing to bind both hapten and carrier. The VH sequence of mutants that retain the ability to bind PC still shows f = 0.37 Hz, suggesting that this frequency determines PC binding. However, this statement was not confirmed as mutation in another PC subsite impairs PC binding but retains both the phosphate-group recognition and the frequency at f = 0.37 Hz. Herein, this finding is discussed to promote the idea that the VH sequence of the PC-binding antibodies encodes the subsite for phosphate-group binding as a dominant functional activity and that only CDR2 of the T15-idiotype antibodies together with FR3 region form an autonomous self-association function represented by the T15VH50-73 peptide with f = 0.37±0.05 Hz. Thus, these data confirmed that T15VH50-73 peptide might be used in superantibody technology.

Plant-produced idiotype vaccines for the treatment of non-Hodgkin's lymphoma: safety and immunogenicity in a phase I clinical study.

Plant-made vaccines have been the subject of intense interest because they can be produced economically in large scale without the use of animal-derived components. Plant-made therapeutic vaccines against challenging chronic diseases, such as cancer, have received little research attention, and no previous human clinical trials have been conducted in this vaccine category. We document the feasibility of using a plant viral expression system to produce personalized (patient-specific) recombinant idiotype vaccines against follicular B cell lymphoma and the results of administering these vaccines to lymphoma patients in a phase I safety and immunogenicity clinical trial. The system allowed rapid production and recovery of idiotypic single-chain antibodies (scFv) derived from each patient's tumor and immunization of patients with their own individual therapeutic antigen. Both low and high doses of vaccines, administered alone or co-administered with the adjuvant GM-CSF, were well tolerated with no serious adverse events. A majority (>70%) of the patients developed cellular or humoral immune responses, and 47% of the patients developed antigen-specific responses. Because 15 of 16 vaccines were glycosylated in plants, this study also shows that variation in patterns of antigen glycosylation do not impair the immunogenicity or affect the safety of the vaccines. Collectively, these findings support the conclusion that plant-produced idiotype vaccines are feasible to produce, safe to administer, and a viable option for idiotype-specific immune therapy in follicular lymphoma patients.

A common idiotype in IgE and its relation to recognition of the grass pollen allergen Phl p 2.

The variable regions of allergen-specific IgE, the isotype mediating allergic responses, are poorly defined to date. In this study we define the character of human antibody binding sites recognizing Phl p 2, a major allergen from timothy grass pollen. Independently raised specificities developed by phage display technology tended to have common sequence motifs (idiotypes), such as IGHV4-31 germline gene origin and heavy chain complementarity-determining region (CDR) 3 length and sequence. They also combined with highly related light chain sequences. Such heavy chain variable domain-encoding transcripts have also been found in the IgE-encoding transcriptome of yet other grass pollen allergic subjects. Altogether these data argue that a common idiotype is used to establish specific antibodies with a potential to mediate allergic responses to Phl p 2. Such a restriction may contribute to the limited molecular diversity observed in some IgE populations.

Identification of an idiotypic peptide recognized by autoantibodies in human immunodeficiency virus-1-infected individuals.

Antibodies against HIV-1 proteins in HIV-1-infected individuals share a cross-reactive idiotype defined by the monoclonal antiidiotypic antibody 1F7 (5). Using a computer algorithm based on the molecular recognition theory, regions of inverse hydropathy between the variable sequence of 1F7 and human monoclonal anti-HIV-1 antibodies were identified, which are assumed to be involved in idiotype-antiidiotype contacts. A peptide was designed from the proposed contact in the variable heavy chain framework 3-complementarity determining region 3 (FR3-CDR3) of human antibodies and was synthesized. This peptide is recognized by the antiidiotype 1F7 and inhibits the binding of 1F7 to human anti-HIV-1 antibodies which express the 1F7 idiotype. A survey of normal and HIV-1-infected sera revealed the presence of antibodies in infected sera which bind to the FR3-CDR3 peptide. The biological relevance of autoantibodies against a self idiotope associated with HIV-1 infection is discussed in the context of the regulation of the antibody response to HIV-1.

Idiotype protein vaccination in combination with adjuvant cytokines in patients with multiple myeloma--evaluation of T-cell responses by different read-out systems.

Anti-idiotypic T cells were analyzed in myeloma patients (n=18) vaccinated with idiotypic protein together with the adjuvant cytokines granulocyte-macrophage colony-stimulating factor (GM-CSF) and/or interleukin-12 (IL-12). In the group given IL-12/GM-CSF, 78% developed idiotype specific T cells as compared to 22% in the group given only IL-12 (proliferation/ELISPOT assays) (p<0.05). The percentage of immune-responding patients increased when quantitative real time polymerase chain reaction assays for cytokines were included. A predominance of a Th1 (IFN-gamma/TNF-alpha) immune response was noted in the IL-12 group while a Th2 (IL-5) response prevailed in the IL-12/GM-CSF group (p=0.053). Application of multiple read-out systems improved the characterization of the immune response.

Crystal structure of an Fv-Fv idiotope-anti-idiotope complex at 1.9 A resolution.

Anti-idiotopic antibodies react with unique antigenic features, usually associated with the combining sites, of other antibodies. They may thus mimic specific antigens that react with the same antibodies. The structural basis of this mimicry is analyzed here in detail for an anti-idiotopic antibody that mimics the antigen, hen egg-white lysozyme. The crystal structure of an anti-hen-egg-white lysozyme antibody (D1.3) complexed with an anti-idiotopic antibody (E5.2) has been determined at a nominal resolution of 1.9 A. E5.2 contacts substantially the same residues of D1.3 as lysozyme, thus mimicking its binding to D1.3. The mimicry embodies conservation of hydrogen bonding: six of the 14 protein-protein hydrogen bonds bridging D1.3-E5.2 are structurally equivalent to hydrogen bonds bridging D1.3-lysozyme. The mimicry includes a similar number of van der Waals interactions. The mimicry of E5.2 for lysozyme, however, does not extend to the topology of the non-polar surfaces of E5.2 and lysozyme, which are in contact with D1.3 as revealed by a quantitative analysis of the contacting surface similarities between E5.2 and lysozyme. The structure discussed herein shows that an anti-idiotopic antibody can provide an approximate topological and binding-group mimicry of an external antigen, especially in the case of the hydrophilic surfaces, even though there is no sequence homology between the anti-idiotope and the antigen.

Defining the structural correlates responsible for loss of arsonate affinity in an IDCR antibody isolated from an autoimmune mouse.

Immunization of the autoimmune mouse strain (M x A) Id/lpr with Ars-KLH, has been shown to elicit a prolonged anti-Ars IdCR response similar to that found in A/J mice. Cell fusion of splenocytes from a diseased mouse previously immunized with Ars-KLH resulted in a monoclonal antibody, 1-52.30, that was found to express the strain A major cross-reactive idiotype, but failed to bind Ars. Nucleotide sequence analysis demonstrated that 1-52.30: (a) used the "canonical" combination of gene segments associated with this idiotype, and (b) exhibited a pattern of somatic mutation consistent with selection for high affinity Ars binding. Two amino acids, VL 91 and 93, were mutated in 36-65, the germline equivalent of the IdCR antibodies, to 1-52.30-like residues (91G-->D, 93T-->M). The results of the mutagenesis showed that changing a single light chain residue, VL 91, from glycine to aspartic acid, resulted in a dramatic loss of Ars binding activity.

Monitoring the formation of soluble immune complexes composed of idiotype and anti-idiotype antibodies by electron microscopy.

We have previously used immunoelectron microscopy (IEM) to generate a three-dimensional map of idiotypic (Id) and isotypic epitopes on the Fab arms of HGAC 39 (Roux et al., 1987, Proc. natn. Acad. Sci. U.S.A. 84, 4984-4988), a mouse IgG3 monoclonal antibody (Mab). In this report, we analyse the geometry of the various types of immune complexes formed by the interaction of HGAC 39 with Mab directed against four mapped epitopes. Moreover, by sampling of reaction mixtures over time, we show that the kinetics of each of the subpopulations of immune complexes, as defined by geometric configuration, can be determined. The data show that for each antibody (Ab)-HGAC 39 combination the rate of immune complex formation was greatest during the first 1.5-3.5 min but that additional complexes formed through the remainder of the half hour assay period. As anticipated, complexes composed of even number units predominated (primarily dimers and tetramers) and most of these were in the form of closed rings. The data also suggest that the location and orientation of the epitopes on HGAC 39 to which the monoclonal antibodies were bound has an influence on the types of immune complexes generated. Specifically we observed that those anti-idiotype Abs that bind to the distal tip of Fab arms (i.e. in the CDR) are less likely to produce bivalently associated ringed dimers than antibodies that bind to epitopes that are proximal to the CDR and that project laterally from the surface of the Fab arms. These data are interpreted in terms of restrictions on hinge mediated flexibility and steric inhibition between adjacent Fab arms on HGAC 39.

Molecular basis for the lack of mimicry of Brucella polysaccharide antigens by Ab2gamma antibodies.

The crystal structure of the complex of an anti-Id Fab with an Fab specific for a Brucella polysaccharide antigen has previously been reported (Evans et al., 1994, J. Mol. Biol. 241, 691-705). To complement this study, the binding characteristics and immunological properties of this Ab2 and two others raised with a second anti-Brucella antibody were investigated, including quantitative kinetic measurements by surface plasmon resonance. The affinities of the Fabs from the Ab2s for the Ab1s were three orders of magnitude greater than those estimated for the antigen, but the Ab2s failed to induce antigen-binding Ab3s, that is, they were of the Ab2gamma type. The avidities of the Ab1s for antigen were however within one order of magnitude of their avidities for Ab2. Tests of 16 other anti-Brucella polysaccharide antibodies showed that the two idiotopes were not present in them, and in confirmation of the lack of a dominant idiotope, N-terminal sequencing of their H and L chains showed a wide variety of V genes were employed in the immune response to the Brucella polysaccharides. The failure of the Ab2 to induce antigen-reactive Ab3 thus appears to be due to neither intrinsic affinity nor idiotope frequency, but arises instead from structural reasons, for example, the incomplete penetration of the Ab2 into the binding-site cleft of the Ab1. The surface topography of polysaccharide antigens and their binding-sites thus appears to be especially difficult for Ab2s to mimic and will restrict their routine use as surrogates for T-cell independent polysaccharide antigens.

Structural characterization of viral neutralizing monoclonal antibodies to hepatitis B surface antigen.

In this study we describe the viral neutralizing activity of murine monoclonal antibodies (MAb) specific for hepatitis B surface antigen (HBsAg). This viral neutralizing activity was assessed in vitro by employing Hepatitis Delta Virus (HDV) and human hepatocytes as target cells. To further characterize these viral neutralizing antibodies we generated a panel of anti-idiotypic (anti-Id) reagents and serologically characterized these antibodies for epitope specificity, Id specificity, and Id heterogeneity. Direct binding and competitive inhibition solid phase enzyme immunoassay have demonstrated that two murine MAb specific for HBsAg (anti-HBs), designated A1.2 and A3.1, recognize similar or overlapping epitopes on HBsAg, while monoclonal anti-HBs, designated A2.1 recognizes a unique HBsAg epitope. Further, Id analysis using monoclonal and polyclonal anti-Id reagents have identified both a private and a cross-reactive Id, respectively, on the anti-HBs, A1.2 preparation. The source of the idiotypic cross-reactivity between A1.2 and A3.1 has been identified, using Western blot analysis, to conformational determinants expressed by the heavy (H) and light (L) chains of these monoclonal anti-HBs. Lastly, the intrastrain antibody repertoire induced following HBsAg immunization was found to be relatively restricted in heterogeneity by clonotype analysis using isoelectric focusing and affinity immunoblot analysis. Interspecies variability in the anti-HBs response was observed based on epitope recognition using purified anti-HBs from a variety of species.

Molecular mapping of the 8.12 SLE-associated idiotype specificity at the single amino acid level.

The 8.12 idiotype is expressed in elevated titer in the serum of patients with systemic lupus erythematosus and is a marker for a subpopulation of anti-DNA antibodies that possess a V(lambda)II encoded light chain. This study utilized a eukaryotic expression system to identify the structural basis for expression of this idiotype. Reversion of the 8.12+ DSC light chain to the hslv215.23/DPL11 germline gene reveals that the 8.12 idiotype is encoded in the germline. The 8.12+ DSC and the 8.12 AS17 light chains, both belonging to the V(lambda)II family, were subjected to site directed mutagenesis, to localize amino acids important for expression of the 8.12 idiotype. Point mutations were performed in CDR1, CDR2, FR3 and CDR3, in positions where the 8.12+ DSC differs from the 8.12-AS17. Amino acids in CDR1 and the CDR2 proximal region of FR3, but not the J proximal region of CDR3, play a crucial role in 8.12 reactivity. The 3-D structure of Mcg, a human IgG1, with which DSC shares a sequence homology of 92.3% has been examined to visualize the effect of each of the mutations and to identify the surface on DSC that comprises the idiotype.

Molecular characterization and structural modeling of immunoglobulin variable regions from murine monoclonal antibodies specific for hepatitis B virus surface antigen.

We have characterized structurally the V regions of a set of murine monoclonal antibodies designated A1.2, A3.1, and A2.1, which recognize a group-specific epitope associated with hepatitis B virus surface antigen (HBsAg). The selection of these antibodies for this characterization was based on data which indicated that A1.2 and A3.1 recognize an overlapping epitope, while A2.1 recognizes a different group-specific epitope, on the HBsAg molecule. In addition, a conformation-dependent cross reactive Id is expressed on both A1.2 and A3.1, but not on A2.1. We have determined the primary sequence structures of these three monoclonal antibodies to HBsAg (anti-HBs), and have aligned them to evaluate V region sequence homology and identify potential regions of structural homology which provide a basis for the HBsAg epitope recognition and the cross reactive Id. Both A1.2 and A3.1 express VH regions which are highly homologous to the VH NP gene family (V186-2), both use members of the DSP2 D region gene family and utilize the JH 2 and JH 1 J gene segments, respectively. Alternatively, A2.1 is related to the VH J558 gene family and expresses a fusion of the DFL16.1 and DQ52 D gene regions in conjunction with the MH 1 gene segment. Each of these three monoclonal anti-HBs utilize light chains from the V kappa 21 and the J kappa 4 gene families. Primary amino acid sequence data were employed to construct computer generated models of the A1.2, A3.1, and A2.1 V regions to determine potential antigen combining site structures and the basis for the expression of the cross reactive Id. These results are discussed in terms of potential interaction sites with HBsAg and V region sites involved in Id expression.

An alternative ELISA for T4 determination based on idiotype anti-idiotype interaction and a latex method for anti-idiotype monoclonal antibody selection.

This paper is the first report on the use of an idiotype-anti-idiotype monoclonal antibody reaction to develop an enzyme immunoassay for thyroxine (T4). We have developed a monoclonal antibody against T4, named 1F10 of IgG1 subclass and KA 5.21 x 10(8) M-1 which was used to obtain anti-idiotypic monoclonal antibodies. Anti-idiotypic antibodies were selected by a novel method, a passive agglutination assay with the idiotype monoclonal 1F10 absorbed on latex particles and subsequently characterized by RIA. One of these anti-idiotype antibodies, named 5B3--type beta antibody--of IgG1 subclass, was used to develop an enzyme-linked T4 idiotype-anti-idiotype immunosorbent assay. The T4 calibration curve, using the 1F10 idiotypic antibody adsorbed to solid phase and the 5B3 anti-idiotypic antibody conjugated to alkaline phosphatase (ALP), shows adequate performance in the range between 0.7-25 micrograms% of the analyte. The reliability of the proposed method is demonstrated by the correlation coefficient r = 0.74, found between T4 measured by RIA and our assay, with a panel of sera from euthyroid, hypothyroid and hyperthyroid individuals. The correlation coefficient was r = 0.93 within assays and r = 0.88 between assays. These results provide the basis for a new non isotopic assay for the study and diagnosis of T4-related human disease and provides a model to develop immunoassays for other haptens and small molecules of clinical interest.

Characterization of phage-displayed recombinant anti-idiotypic antibody fragments against coronavirus-neutralizing monoclonal antibodies.

Murine coronaviruses provide useful animal models for human neurological disorders such as multiple sclerosis. In an effort to better understand the mechanisms involved in protection from coronavirus infection, we are studying the role of the idiotypic network in the modulation of viral infectivity. We have explored the feasibility of using single-chain antibodies displayed on phage surfaces for the isolation of recombinant anti-idiotypic antibodies (anti-Ids) with antigen-mimicking properties, which has proven to be difficult with conventional hybridoma approaches. A phage-display library containing more than 10(8) different antibody specificities was screened for the presence of anti-Ids by successive rounds of panning with three different in vitro neutralizing and in vivo protective antiviral monoclonal antibodies. After five rounds of panning, between 32% and 84% of all individual clones tested showed antibody-binding in an enzyme-linked immunosorbent assay (ELISA). Although several clones showed identical antibody sequences, a number of different clones were identified and further characterized. None of the selected clones induced the production of antiviral or neutralizing antibodies or conferred reproducible protection from viral challenge in BALB/c and C57BL/6 mice. These results demonstrate that anti-Ids can be isolated from a phage-display library, although high-affinity antigen-mimicking phages with antiviral protective capacities were apparently not represented in this library. This argues for the development of more diverse phage-display libraries.

The use of defined peptides in characterizing idiotypes.

The structural correlates of idiotypes have been sought in several antibody systems. The cumulative results suggest that the hypervariable regions (or complementarity-determining regions) of the heavy and light chains are the structural basis of idiotypes. However, in most cases, it is exceedingly difficult to associate a particular idiotypic determinant with a specific amino acid sequence. Recently, synthetic peptides were used to induce antibodies specific for predefined determinants in intact proteins. These findings led us and others to use synthetic peptides corresponding to the hypervariable regions/complementarity-determining regions to induce anti-idiotypes. These novel anti-idiotypic antibodies are easy to prepare, and are ideal reagents for structural and genetic studies of antibody responses.

Perspectives on antigenicity and idiotypy.

The recent crystal determination of a lysozyme-antilysozyme complex provides a three-dimensional prototype of the manner in which contacts in idiotype-anti-idiotype interactions may be realized. Such interactions can be approximated by two complementary "flat" surfaces. Each IDR (autoantigenic locus) location might provide a particular recognition feature between two interacting partners. The combinatorial manner in which IDR domains are recognized by anti-idiotypic antibodies describe the repertoire of private and public (crossreactive) idiotopes of an antibody. Several interesting features emerge from consideration of the Ab contact residues in the crystal structure. First, framework residues are implicated in contacting the antigen: Thr 30 (FR1) of the heavy chain and Tyr 49 (FR2) of the kappa light chain. Both of these residues lie within predicted IDRs. Framework regions have recently been suggested to be involved in several anti-idiotypic systems, although such regions have, in the past, been disregarded based solely upon sequence analysis. The surface variability analysis, which identifies the repertoire of complementary interacting surfaces, depicts the immunoglobulin as having more variability than generally thought. This variability may also extend to T cell receptors since T cell chains express an extensive surface variable repertoire similar to that of the immunoglobulin light chains (Kieber-Emmons and Köhler, unpublished). Second, the D region plays a critical role in the generation of the antilysozyme combining sites. Similarly, the D segment makes up the largest component of an IDR. Third, while the CDR3 of the heavy chain contributes most to the antibody-lysozyme complex it is not the most surface-exposed (see Novotny, this issue). Nevertheless, surface variability analysis indicates that this region is generally immunodominant which is also observed experimentally. Together, these results indicate that perhaps certain IDR regions are intrinsically more antigenic. Idiotypic structures must be accessible for antibody recognition and binding. From a structural viewpoint, a single antibody molecule has a continuum or several different combining sites. Subsequently, a single residue can be contained in several overlapping idiotypic determinants. Surface variability analysis suggests that the hypervariable regions of Igs provide a diverse idiotope repertoire that can be utilized for binding. Monoclonal antibodies have been shown to have multiple specificities and this capacity for multiple binding is also intrinsic to the definitions that have emerged for anti-idiotypic antibodies.(ABSTRACT TRUNCATED AT 400 WORDS)

Topographic analysis with monoclonal anti-idiotopes: probing the functional anatomy of immunoglobulin variable domains.

Attempts to correlate immunoglobulin variable domain functional properties with variable domain primary structure have been valuable, but these efforts have suggested that to more fully account for variable domain function in terms of structure will require knowledge of molecular relationships in three dimensions. In this review we describe generally applicable methods, using monoclonal anti-idiotopes, for the determination of spatial relationships of idiotopes relative to one another and relative to two orienting structural markers of variable domains: 1) the hapten-binding site and 2) the junction of the variable and constant domains. Using these methods it has been possible to construct an oriented idiotope map which spans the variable domain along an axis connecting the paratope and the variable domain-constant domain junction. In addition, it has been possible to correlate idiotope position with other properties of idiotope expression. This approach may contribute to the development of predictive principles of idiotope expression.

Rational design of peptide vaccines for autoimmune disease: harnessing molecular recognition to fix a broken network.

Autoreactive T-cells and antibodies are found at low levels in normal individuals and are thought to be kept at bay by regulatory T-cells and a network of idiotypic and anti-idiotype-bearing antigen receptors on lymphocytes as well as idiotypic anti-idiotypic antibodies. Disruption of this network by genetic, environmental and unknown factors is thought to result in autoimmune diseases. An obvious, ideal and specific therapy for such disorders would be to harness this regulatory network to re-establish immunologic homeostasis. In practice, however, this is not an easy task as most autoimmune diseases involve polyclonal responses to self antigen. Thus, we are faced with the conundrum of not knowing which autoreactive idiotype-bearing antibody or antigen receptor(s) to target in order to restore or induce network regulatory function. The thesis of this review is that understanding a fundamental property governing peptide/protein shape can be used in part to circumvent the problems of self reactivity and polyclonality in autoimmune disorders. More specifically, an algorithm has been developed to design peptide vaccines with shapes that are thought to be complementary in contour to self epitopes which seem to be the focus of autoimmunity. In theory, such complementary shapes should be engendered in certain autoreactive antigen receptors--these complementary constructs consequently represent receptor mimetics. By targeting an immune response against such mimetics, one generates a polyclonal anti-idiotype response that matches the complexity of the autoimmune response itself. This article will describe the algorithm for vaccine design, summarize the in vitro and in vivo evidence for its efficacy and discuss possible therapeutic utility in human autoimmune diseases.

Architecture of idiotypic networks: percolation and scaling behavior.

We investigate a model where idiotypes (characterizing B lymphocytes and antibodies of an immune system) and anti-idiotypes are represented by complementary bit strings of a given length d allowing for a number of mismatches (matching rules). In this model, the vertices of the hypercube in dimension d represent the potential repertoire of idiotypes. A random set of (with probability p) occupied vertices corresponds to the expressed repertoire of idiotypes at a given moment. Vertices of this set linked by the above matching rules build random clusters. We give a structural and statistical characterization of these clusters, or in other words of the architecture of the idiotypic network. Increasing the probability p one finds at a critical p a percolation transition where for the first time a large connected graph occurs with probability 1. Increasing p further, there is a second transition above which the repertoire is complete in the sense that any newly introduced idiotype finds a complementary anti-idiotype. We introduce structural characteristics such as the mass distribution and the fragmentation rate for random clusters, and determine the scaling behavior of the cluster size distribution near the percolation transition, including finite size corrections. We find that slightly above the percolation transition the large connected cluster (the central part of the idiotypic network) consists typically of one highly connected part and a number of weakly connected constituents and coexists with a number of small, isolated clusters. This is in accordance with the picture of a central and a peripheral part of the idiotypic network and gives some support to idealized architectures of the central part used in recent dynamical mean field models.