Standard Article

You have free access to this content

Immunoglobulin Gene Construction: Human

  1. Alexander H Lucas

Published Online: 28 MAY 2003

DOI: 10.1038/npg.els.0001172

eLS

eLS

How to Cite

Lucas, A. H. 2003. Immunoglobulin Gene Construction: Human. eLS. .

Author Information

  1. Children's Hospital Oakland Research Institute, Oakland, California, USA

Publication History

  1. Published Online: 28 MAY 2003

Introduction

  1. Top of page
  2. Introduction
  3. Two Genes-One Polypeptide Chain Hypothesis
  4. Experimental Verification of V and C Genes and their Somatic Rearrangement
  5. Genomic Structure of Immunoglobulin Genes
  6. Somatic Recombination
  7. Heavy Chain Class Switching
  8. Diversifiers
  9. Summary
  10. References
  11. Further Reading

This article provides an overview of the structure and dynamics of the genes encoding human immunoglobulin (Ig) molecules. Immunoglobulin genes are unconventional. They are fragmented in the genome; they rearrange during antigen-independent and antigen-dependent phases of B-lymphocyte development. And, they hypermutate. This exceptional genomic organization and somatic behaviour represent unique evolutionary solutions to the problem of generating a diverse repertoire of molecules capable of recognizing and eliminating a plethora of antigenically polymorphic and ever-changing infectious agents.

Our present understanding of the structure and organization of immunoglobulin genes derives from studies using the tools of modern molecular biology that permit a direct examination of genomic structure; however, prior to the era of gene cloning and sequencing, many experimental findings hinted that the genetic basis of antibody production was going to require unique explanations. By the middle of the twentieth century, it had been established that the humoral immune system had an enormous capacity to recognize foreign antigenic determinants. Antibodies could be prepared to a seemingly limitless number of antigens, including those associated with infectious agents as well as synthetic compounds. Thus the question arose as to what genetic mechanisms could account for this exceptional diversity. Were there separate genes encoding each antibody specificity or were there only a limited number of germline genes that were capable of mutating in such a way as to generate the antibody repertoire?

In the 1960s, the general structure of the IgG molecule was elucidated. An IgG molecule consists of four polypeptide chains: two identical heavy (H) chains and two identical light (L) chains linked by disulfide bonds (Figure 1). The H and L chains are each divisible into two regions: the variable (V) region, which is located at the N-terminal end of the polypeptide chain, and the constant (C) region, located at the C-terminal end. The V regions are so-named because the amino acid sequence in this portion of the molecule varies considerably between different antibodies. In contrast, the C region shows much less sequence variation. In the fully folded IgG molecule the H and L V regions associate to form the V domain, which comprises the antigen-combining site. An IgG molecule has two identical antigen-combining sites. The effector capabilities, such as complement binding and attachment to phagocytic cells, are located in the C-terminal portions of the molecule and are mediated by the C domain of the H chains. For a more detailed discussion of Ig structure, the reader may consult the relevant articles in the Encyclopedia.

thumbnail image

Figure 1. An immunoglobulin G (IgG) molecule.

Two Genes-One Polypeptide Chain Hypothesis

  1. Top of page
  2. Introduction
  3. Two Genes-One Polypeptide Chain Hypothesis
  4. Experimental Verification of V and C Genes and their Somatic Rearrangement
  5. Genomic Structure of Immunoglobulin Genes
  6. Somatic Recombination
  7. Heavy Chain Class Switching
  8. Diversifiers
  9. Summary
  10. References
  11. Further Reading

Hints as to the nature of the genes encoding immunoglobulin molecules came from a variety of experimental observations. The early structural studies of immunoglobulin revealed a fundamental duality: antigen recognition resided at one end of the molecule and effector function at the other end. Genetic and serological investigations had shown that antibodies were divisible into classes, e.g. IgM and IgG, and that class determined effector capability. Antibodies of the same specificity could occur in different classes. Furthermore, particular allotypes and idiotypes, serological markers associated with the N-terminal end of the immunoglobulin molecule, were shown to associate with different classes. Amino acid composition studies of myeloma proteins (homogeneous immunoglobulin molecules derived from plasma cell tumours) revealed a striking disparity in the distribution of amino acid residues in the Ig molecule. The N-terminal half of the H and L polypeptide chains had a much more diverse amino acid composition than did the C-terminal region of the molecule. Together, these findings contributed to the idea that an immunoglobulin polypeptide chain may be encoded by more than one gene (Koshland et al., 1969). Direct support for this idea was obtained in 1965 with the publication of the entire amino acid sequence of two human myeloma protein L chains (Hilschmann and Craig, 1965). A clear pattern of amino acid sequence variation was revealed. The primary sequence of the N-terminal portion of the L chain was highly variable between the two chains, whereas the sequence of the C-terminal half of the molecule was nearly invariant. Variability in primary structure could account for antigen-binding specificity differences among antibodies, and the shared constant regions were responsible for the class of the molecule. These observations prompted the hypothesis that two genes encoded a single immunoglobulin polypeptide chain; one gene encoded the V region and one gene encoded the C region (Dreyer and Bennett, 1965). At the time, this proposition was heretical for it stood in contrast to the accepted doctrine of one gene-one polypeptide. Many immunoglobulin sequencing studies followed this initial report, and they confirmed that both H and L polypeptide chains contained V and C regions. Furthermore, three regions of hypervariability were found to occur within V regions. These areas, called hypervariable or complementarity-determining regions, were later shown to be areas that contact antigen and play a pivotal role in determining the chemical character of the combining site.

Experimental Verification of V and C Genes and their Somatic Rearrangement

  1. Top of page
  2. Introduction
  3. Two Genes-One Polypeptide Chain Hypothesis
  4. Experimental Verification of V and C Genes and their Somatic Rearrangement
  5. Genomic Structure of Immunoglobulin Genes
  6. Somatic Recombination
  7. Heavy Chain Class Switching
  8. Diversifiers
  9. Summary
  10. References
  11. Further Reading

Up to this time, the immunoglobulin gene problem had been addressed by examining immunoglobulin molecules, i.e. phenotype. However, in the 1970s new methods were developed for purifying messenger ribonucleic acid (mRNA) and deoxyribonucleic acid (DNA), for cutting, cloning and sequencing DNA, and for generating antigen-specific monoclonal antibodies. These methods permitted a direct examination of immunoglobulin genotype. Experimental evidence for the two gene-one polypeptide hypothesis was reported in 1975 (Tonegawa et al., 1975). Molecular probes that could distinguish the V region from the C region, were used to demonstrate that immunoglobulin V and C genes in the mouse existed as distinct entities in germline DNA. In a plasma cell (actually a myeloma cell) however, the V genes and C genes had changed their relative genomic position such that they had become more closely associated. This study not only confirmed the hypothesis that the V and C regions of L chains were encoded by distinct genes, but demonstrated that immunoglobulin genes rearrange during B-lymphocyte development. Further heresy was to follow with the demonstration that immunoglobulin genes were capable of hypermutating (Weigert et al., 1970). Within the next few years the tools of modern molecular biology were brought to bear on the immunoglobulin gene problem. Immunoglobulin genes were among the first mammalian genes sequenced. At the present time, most of the human immunoglobulin gene H and L chain loci have been sequenced.

Genomic Structure of Immunoglobulin Genes

  1. Top of page
  2. Introduction
  3. Two Genes-One Polypeptide Chain Hypothesis
  4. Experimental Verification of V and C Genes and their Somatic Rearrangement
  5. Genomic Structure of Immunoglobulin Genes
  6. Somatic Recombination
  7. Heavy Chain Class Switching
  8. Diversifiers
  9. Summary
  10. References
  11. Further Reading

Human immunoglobulin genes comprise three linkage groups. The genes encoding the kappa (κ) and lambda (λ) L chain genes are located on chromosome 2 and 22, respectively, and the H chain genes are found on chromosome 14. Although these gene families share general features, their respective organization is unique, and therefore they will be discussed separately.

κ L chain genes

The genomic organization of the human κ locus is depicted in Figure 2. A single Cκ gene is located towards the centromere. A cluster of five J (joining) regions occurs immediately 5′ to the Cκ gene. A large cluster of 76 V gene segments is located further upstream of the J regions. They are organized into two groups: one group which is the most proximal to the J cluster contains 40 V gene segments, and another group which is distal to the J cluster contains 36 V gene segments. The proximal and distal groups appear to be derived from one another, as the sequences of many of the genes in one cluster are closely related to members in the other cluster. This organization indicates that the clusters are likely to have arisen from a common ancestral group of genes that underwent a duplication event. The majority of the V genes in the proximal group are in the same transcriptional polarity as the J regions and the Cκ region, whereas the V gene segments in the distal segment are in the opposite transcriptional polarity. Thus, the primordial duplication event probably occurred by inversion. Approximately one-half of the V segments in each group are nonfunctional genes, as they contain errors in their sequence that prevent transcription or translation. However, some of these nonfunctional Vκ genes have minor defects, such as single base changes resulting in a stop codon, and, therefore, reversion to functionality could occur relatively frequently. Indeed, the precise number of functional V gene segments is likely to vary between different individuals. The Vκ gene segments can be grouped into seven families or subgroups based upon sequence similarities.

thumbnail image

Figure 2. Genomic organization of human immunoglobulin genes. (a) κ Light chain locus (chromosome 2); (b) λ light chain locus (chromosome 22); (c) heavy chain locus (chromosome 14). Coding sequences are indicated by open boxes. Nonfunctional or pseudogenes are indicated by shading. Arrows indicate transcriptional polarity. Not drawn to scale.

λ L chain genes

The λ locus resembles the κ locus in that the V gene segments, the J regions and the C regions exist as separate entities. However, in contrast to the κ locus, the λ locus contains seven C genes and each has its own J region gene (Figure 2). Three of these seven C-J clusters are considered pseudogenes because they contain either in-frame stop codons or frame-shifting deletions that prevent their expression. The Vλ gene segment cluster is located upstream of the J-Cλ cluster and consists of approximately 70 members, of which 30–35 are classified as pseudogenes. Vλ genes are classified into 10 subgroups of related genes. Unlike Vκ genes, the Vλ genes are not duplicated into proximal and distal clusters. A single Vλ gene segment may recombine with either of the four functional J-C genes to form a complete λ L chain gene.

H chain genes

The H chain locus is more complex than either L chain locus (Figure 2). There are nine functional CH region genes: μ, δ, γ1, γ2, γ3, γ3, α1, α2 and ε. These encode the different classes or isotypes of immunoglobulin polypeptide H chains: IgM, IgD, the four subclasses of IgG, the two subclasses of IgA and IgE. A nonfunctional pseudogene, φε, lies between the γ1 and α1 C genes. The duplicated pattern, γ-γ-ε-α, suggests that the CH locus evolved by duplication of this group of genes.

The VH gene segments are located upstream from the CH genes. There are approximately 130 VH gene segments but only about 45 are functional. The VH gene segments have been assigned to seven subgroups based upon their sequence relatedness. Individual members of a family are not grouped together but are interspersed throughout the V gene cluster. The H locus contains six J regions. In contrast to either L chain locus, there is an additional cluster of elements called D regions that are located between the V and J clusters. Approximately 25 functional D region genes have been described.

Somatic Recombination

  1. Top of page
  2. Introduction
  3. Two Genes-One Polypeptide Chain Hypothesis
  4. Experimental Verification of V and C Genes and their Somatic Rearrangement
  5. Genomic Structure of Immunoglobulin Genes
  6. Somatic Recombination
  7. Heavy Chain Class Switching
  8. Diversifiers
  9. Summary
  10. References
  11. Further Reading

One of the most remarkable features of immunoglobulin genes is their ability to rearrange. These gene rearrangements occur at two distinct phases of B-lymphocyte development. The first phase is antigen-independent and occurs as part of the developmental programme that generates mature B cells from hematopoietic stem cells. In this phase, complete VH and VL genes and functional transcriptional units are created by gene recombination events. During B lymphocyte ontogeny, gene rearrangements first occur at the H chain locus and then at the L chain loci. The second phase of gene rearrangements occurs only in the H locus and is stimulated in B-cell clones following antigen activation, a process known as H chain class switching (see below).

L chain production

Although the κ and λ L chain loci differ somewhat in genomic organization, their mechanism of gene rearrangement is similar. Figure 3 depicts gene rearrangements in the κ locus. One particular Vκ gene segment recombines with a particular Jκ segment, resulting in a complete V region gene, a V-J fusion product. During this process, the region between the V and J elements is looped out. These gene rearrangements occur by site-specific recombination and are mediated by enzymes that recognize DNA sequences (recombination signal sequences) located at the 3′ end of each V gene segment and the 5′ ends of the J regions. Once V-J joining has occurred, a functional transcriptional unit is created. A primary transcript is generated from the assembled gene. This transcript consists of the leader region (a short sequence of nucleotides encoding the signal peptide that directs the polypeptide chain into the lumen of the endoplasmic reticulum during translation), V-J and the Cκ region. Interspersed between these coding regions, also known as exons, are intervening sequences or introns. The introns are spliced out of the primary transcript during RNA processing. The final mRNA contains contiguous nucleotides encoding the complete L chain polypeptide. The signal peptide is removed from the N-terminal end during processing in the endosome. Thus, a single mature L polypeptide chain is created from not two but three genes, V-J-C. The VL region is approximately 110 amino acids in length. The majority (95) of these residues are encoded by the V gene segment. The J region encodes about 13 residues, but during V-J joining extra nucleotides can be added and result in lengthening (and diversifying) of the V region. The first and second hypervariable regions are located in the V gene segment, whereas the third hypervariable region is located around the V-J junction.

thumbnail image

Figure 3. L chain production. The process of antigen-independent gene rearrangement that generates a functional L chain gene. A complete VL coding region is created by recombination between the V1 gene segment and J3. Splicing of the primary transcript from the V1-J3-Cκ assemblage generates an mRNA in which the V and C regions are contiguous.

It is important to note that individual B lymphocytes express a single functionally rearranged L (and H) chain gene product. The formation of a successful V-J fusion (or V-D-J for the H locus) prevents the other V and J elements from being rearranged on the same chromosome. In addition, V genes on the other parental chromosome (and for L chains on the two chromosomes carrying the other L chain type) are also silenced. These exclusionary processes ensure that a single B lymphocyte will express a single VH-VL combination.

The choice as to which V gene segment recombines with a particular J gene is thought to occur by a stochastic process. Two hundred distinct Vκ regions can be created by the combinatorial association of 40 functional Vκ genes and five Jκ genes. A similar calculation for the λ locus generates 280 Vλ regions (70 Vλ × 4 Jλ).

H chain production

Gene rearrangements also occur in the H chain locus, except the liaison consists of a fusion between three elements, V-D-J (Figure 4) Site-specific recombination occurs first between the D and J loci (D-J joining) to create a D-J fusion. Then, a VH gene segment recombines with D-J to form the mature rearranged V gene consisting of V-D-J. The enzymes that mediate L chain rearrangements also mediate rearrangements at the H locus. Recombination signal sequences are located at the 3′ end of each VH gene segment, at the 5′ and 3′ ends of the D regions and at the 5′ ends of the J regions. A consequence of V-(D)-J fusion is that other VH, D and JH genes on the same chromosome are silenced and prevented from being expressed within that particular B cell. As with the L chain, the V-D-J recombination is stochastic and, therefore, from 45 functional VH gene segments, 25 D and six J region genes, a total of 6750 distinct V assemblages are possible.

thumbnail image

Figure 4. H chain production. The process of antigen-independent gene rearrangement that generates a functional H chain gene. A complete VH coding region is created by recombination between the V2 gene segment, D3 and J4. Splicing of the primary transcript from the V2-D3-J4-Cμ assemblage generates an mRNA in which the V and C regions are contiguous.

A functional transcriptional unit is created once successful V-D-J assembly has occurred. A primary RNA transcript is generated from the rearranged DNA, and this transcript is then processed into a mature H chain mRNA encoding the leader sequence, the assembled V region (V-D-J) and the C region (Figure 4). During translation and protein processing the leader peptide is cleaved from the chain. The V region portion of the H chain polypeptide is therefore created from three separate genetic elements (V-D-J). These elements do not equivalently contribute to the overall length of a VH region, which is approximately 120 amino acids. The V gene segment encodes about the first 95 amino acids; the D segments are of variable length and can encode up to 10 or 12 amino acids; the J regions also vary slightly in length and contribute approximately 17 amino acids. The third hypervariable (or complementarity-determining region) of the VH region is created by the fusion of the D and J regions, whereas the first and second hypervariable regions are within the VH gene segment.

The CH genes depicted in Figure 2 and Figure 3 are shown simply as single blocks. The fine structure is actually more complex. As shown in Figure 5, the Cμ gene is divisible into coding (exons) and noncoding (introns) regions. The other CH genes are similarly organized. Most CH exons are approximately 300 bases in length, corresponding to a single immunoglobulin domain of approximately 100 amino acids, although there are smaller exons such as the leader sequence exon, the exon encoding the hinge region in IgG and the 3′ termini of the CH genes. Immunoglobulins occur both as cell surface molecules, where they function as receptors for antigen, and as secreted molecules. The sequences occurring at the 3′ end of the CH gene dictate whether the polypeptide molecule is either membrane-bound or secreted. The generation of these two forms is determined by processing of RNA. As shown in Figure 5, the primary H chain transcript contains the exons encoding the leader sequence, the assembled V region (V-D-J), the CH region and the secreted or membrane termini. Differential processing of the RNA generates either an mRNA with the secreted terminus or an mRNA with the membrane terminus. During processing, the introns are excised and the exons become fused into a contiguous chain. These RNA processing events occur at all productively rearranged CH loci, but the generation of IgM and IgD transcripts is unique. Resting virgin B lymphocytes express both IgM and IgD on their cell surface. mRNAs encoding these two isotypes are generated from a single primary transcript that spans the leader sequence, V-D-J, and the Cμ and Cδ genes. The primary transcript is processed such that two mRNAs are produced: one with the V-D-J module associated with the Cμ gene and the other with the V-D-J associated with the Cδ gene.

thumbnail image

Figure 5. Generation of membrane and secreted forms of immunoglobulin M (IgM). The Cμ gene is divided into exons (boxes) and introns (lines). The primary transcript generated from a rearranged H locus contains intron sequences and exon sequences encoding the signal or leader peptide (L), the assembled V region (VDJ), the four Cμ domains, the secreted (s) and membrane (m) termini. The transcript is differentially processed to generate a mature mRNA having either the secreted terminus or the membrane terminus. The introns are removed during processing such that the exons are contiguous.

Heavy Chain Class Switching

  1. Top of page
  2. Introduction
  3. Two Genes-One Polypeptide Chain Hypothesis
  4. Experimental Verification of V and C Genes and their Somatic Rearrangement
  5. Genomic Structure of Immunoglobulin Genes
  6. Somatic Recombination
  7. Heavy Chain Class Switching
  8. Diversifiers
  9. Summary
  10. References
  11. Further Reading

During the course of an immune response a shift generally occurs in the predominant class of antibody formed. Expression of the IgM class precedes the expression of the other downstream classes. This class switch allows the same combining site (antigen-binding specificity) to be expressed in the context of different H chain classes. CH switching is induced in individual B cells following appropriate antigenic stimulation and occurs by gene rearrangement. As described above, the initial transcriptional unit from the CH locus comprises the V-D-J module in association with both Cμ and Cδ. When switching recombination occurs the V-D-J module is relocated to the vicinity of a different CH gene that is downstream of the Cμ-Cδ region. This recombination involves looping out and deletion of the intervening genes. Recombination is mediated by recombinase enzymes and switch regions that are located immediately 5′ to the CH genes. Figure 6 depicts two consecutive rearrangement events. The first rearrangement involves recombination of V-D-J to the γ1 gene, resulting in the deletion of the Cμ, Cδ, and Cγ3 genes. The second rearrangement involves switching from γ1 to α2. This process is sequential and proceeds downstream with deletion of the upstream genes. Thus, B cells cannot revert to expression of classes encoded by the upstream CH genes once they have begun the switching process. Class switching is a mechanism that provides the important biological function of expressing the same VH in the context of different isotypes or effector classes.

thumbnail image

Figure 6. Class switching. Rearrangements in the H locus occur in the B cell following antigen stimulation. A VDJ module switches from Cμ to Cγ1 by site-specific recombination at switch (S) regions. Looping out of the intervening DNA occurs during this process, resulting in the deletion of CH genes. A subsequent switch occurs, resulting in the relocation of the VDJ module to the Cα2 gene.

Diversifiers

  1. Top of page
  2. Introduction
  3. Two Genes-One Polypeptide Chain Hypothesis
  4. Experimental Verification of V and C Genes and their Somatic Rearrangement
  5. Genomic Structure of Immunoglobulin Genes
  6. Somatic Recombination
  7. Heavy Chain Class Switching
  8. Diversifiers
  9. Summary
  10. References
  11. Further Reading

Considerable combining site diversity can be generated from the combinatorial processes described above. If we assume that H and L polypeptide chains can randomly associate, then the potential repertoire can be calculated as comprising ∼3 × 106 VH-VL combinations (6750 VH × 280 Vλ = 1.9 × 106 plus 6750 VH × 200 Vκ = 1.4 × 106). Somatic diversifiers exist that substantially expand V gene diversity and thereby extend the antigen recognition potential of these germline combinations. One such process is flexible joining. V-(D)-J recombination is not always precise, and nucleotides at the 5′ and 3′ ends of the pieces of DNA being joined can be differentially utilized to form different codons at the V-J junction. Furthermore, additional nontemplated nucleotides can be added to the ends of the DNA being joined. Flexibility in joining codons and addition of nontemplated nucleotides provide a rich source of sequence diversity in the third hypervariable region of both H and L chains.

A further diversifier is hypermutation. Initial evidence for this remarkable process was obtained in a study of mouse L chains, where single amino acid substitutions were observed among a set of VL regions which were derived from the same V gene (Weigert et al., 1970). Since this early observation, a large number of studies have documented the occurrence of somatic mutations in antibodies to a plethora of antigens. The signature of hypermutation is the increasing occurrence of point mutations in antibody V regions during the course of an immune response. Immunoglobulin gene mutation is called hypermutation because it occurs at a frequency that is many orders of magnitudes higher than that of normal genomic mutation. The mutations are targeted to the V region and may be preferentially directed to hotspots within the V gene. The enzyme(s) mediating the mutation is not known but an error-prone DNA polymerase is a good candidate. Although single base substitutions are the typical products of hypermutation, insertions and deletions may also occur. The process of somatic hypermutation provides an explanation for affinity maturation, a phenomenon where antibodies with higher binding affinity for antigen appear during the course of an immune response.

Summary

  1. Top of page
  2. Introduction
  3. Two Genes-One Polypeptide Chain Hypothesis
  4. Experimental Verification of V and C Genes and their Somatic Rearrangement
  5. Genomic Structure of Immunoglobulin Genes
  6. Somatic Recombination
  7. Heavy Chain Class Switching
  8. Diversifiers
  9. Summary
  10. References
  11. Further Reading

Human antibody diversity is generated by several mechanisms: the existence of multiple germline genes; the combinatorial assortment of these genes to form assembled V regions; the nucleotide diversity resulting from recombination and hypermutation; and the pairing of H and L chains. Although adaptive immune systems encoded by immunoglobulin genes are present in all vertebrates, notable differences exist between species (Litman et al., 1999). Immunoglobulin gene systems and T-cell receptors appear to have evolved from a primordial gene encoding a single immunoglobulin-like domain. Features common to the evolution of most immunoglobulin gene systems include their derivation from a primordial single domain gene, duplication, retention of segmentation and acquisition of a somatic recombination mechanism. Major histocompatibility complex molecules and other cell surface receptor systems appear to have evolved from this primordial immunoglobulin gene, and they are classified into a group of related structures termed the immunoglobulin gene superfamily.

References

  1. Top of page
  2. Introduction
  3. Two Genes-One Polypeptide Chain Hypothesis
  4. Experimental Verification of V and C Genes and their Somatic Rearrangement
  5. Genomic Structure of Immunoglobulin Genes
  6. Somatic Recombination
  7. Heavy Chain Class Switching
  8. Diversifiers
  9. Summary
  10. References
  11. Further Reading

Further Reading

  1. Top of page
  2. Introduction
  3. Two Genes-One Polypeptide Chain Hypothesis
  4. Experimental Verification of V and C Genes and their Somatic Rearrangement
  5. Genomic Structure of Immunoglobulin Genes
  6. Somatic Recombination
  7. Heavy Chain Class Switching
  8. Diversifiers
  9. Summary
  10. References
  11. Further Reading
  • Honjo T and Alt FW (1995) Immunoglobulin Genes, 2nd edn. New York: Academic Press.
  • Max EE (1999) Immunoglobulins: molecular genetics. Paul WE Fundamental Immunology, 4th ed., pp. 111182. Philadelphia: Lippincott-Raven
  • Podolsky SH and Tauber AI (1997) The Generation of Diversity: Clonal Selection Theory and the Rise of Molecular Immunology. Cambridge, MA: Harvard University Press.
  • Williams AF and Barclay AN (1988) The immunoglobulin superfamily - domains for cell surface recognition. Annual Review of Immunology 6: 381405.