Rearrangement of the human heavy chain variable region gene V3–23 in transgenic mice generates antibodies reactive with a range of antigens on the basis of V_HCDR3 and residues intrinsic to the heavy chain variable region

Abstract

To formulate a ‘logic’ for how a single immunoglobulin variable region gene generates antibodies with different antigen specificity and polyreactivity, we analysed chimeric antibodies produced in transgenic mice carrying the germ-line human V3–23 gene, multiple diversity (D) and joining (J) gene segments. Hybridomas producing antibodies encoded by the V3–23 gene in combination with different mouse V_κ genes were obtained by fusion of splenocytes from transgenic mice. All antibodies had human μ-chains and mouse light chains, were multimeric in structure and expressed the human V3–23 gene. Nucleotide sequence analyses of genes encoding the heavy and light chains of 12 antibodies in relation to antigen specificity highlighted the importance of heavy chain variable region CDR3 in determining reactivity with different antigens. However, the results also suggest that non-CDR3 sequences intrinsic to the V3–23 gene itself may be involved in, or determine, the binding of the chimeric antibodies to some of the antigens tested in the current study.

Introduction

Diversity of the primary antibody repertoire is the consequence of variation in the combination of non-contiguous gene segments for the heavy chain-variable (V_H), diversity (D) and joining (J_H) genes and the light chains V_L (V_κ or V_λ) and J_L (J_κ or J_λ) genes and imprecision in generating junctional regions [1]. After exposure to antigen, further diversity is created and affinity of the existing primary antibody repertoire is enhanced by somatic mutation [2]. Early studies of immunoglobulin variable region (IgV) gene use in healthy adult individuals and in patients with a range of diseases were based on the presumption that functional IgV genes have equal expression potential in the primary antibody repertoire. However, subsequent studies provided evidence that a small set of IgV genes are preferentially expressed [3,4]. It was further observed that this set of genes encoded antibodies with a wide range of antigen specificity, and that some reacted with more than one antigen (polyreactive or ‘natural’ antibodies) [5,6]. Among these genes, the V3–23 gene has been shown in some studies to encode a significant proportion of the primary repertoire in healthy adult individuals and also pathogenic antibodies in patients, such as antibodies to double-stranded DNA (dsDNA) in patients with systemic lupus erythematosus (SLE) [3,4,7,8]. These and related observations raised a number of important issues: (i) how a single heavy chain gene, often in germ-line configuration, could encode antibodies with different antigen specificity [6]; (ii) whether diversity is primarily dependent on CDR3 sequences, or involves sequences intrinsic to the heavy chain [9], in this case the V3–23 gene; and (iii) if, and how, polyreactive antibodies found in healthy individuals relate to potentially pathogenic autoantibodies in SLE patients [6].

In this study we sought to understand further the molecular basis of self and exogenous antigen recognition by antibodies encoded by the V3–23 gene. Specifically, we hoped to understand the contribution of different parts of antibody molecules encoded by V3–923 to antigen reactivity. In order to focus on the V3–23 gene, we raised hybridomas from mice transgenic for a human heavy chain minilocus containing the V3–23 gene [10]. We show here that the V3–23 transgene can recombine with different human D and J_H genes and mouse V_L genes to generate functional antibodies. Further, sequence analyses show that variation in the V_HCDR3 influences the pattern of polyreactivity. This is consistent with a previous study in which the importance of the V_HCDR3 in dictating the polyreactive phenotype of recombinant antibodies was established [11]. Intriguingly, the data also suggest that reactivity with IgG-Fc, collagen type I and II may involve residues intrinsic to V3–23 germ-line sequence itself. This is in agreement with recent genetic and crystallographic studies of human antibodies to the I antigen on erythrocytes and to IgG (rheumatoid factor) in which binding to the autoantigen was shown to depend on residues in the framework regions outside the conventional antibody binding site [12,13]. Based on these observations, and on the frequency of V3–23 gene use in the healthy adults, we propose that the potential of V3–23-encoded antibodies to bind antigens through conventional antigen binding sites, as well as framework residues, provide B lymphocytes expressing the gene with a selective advantage in the circulating B lymphocyte repertoire.

Materials and methods

Transgenic mice

The generation of transgenic mice from which the hybridomas derived has previously been described [11]. The C57Bl/6 X CBA F₁ mice colony used were descendants of founder 15 and carried copies of two human genomic cosmids containing the V3–23 gene segment, D_H segments, the six J_H genes, the μ-chain gene, and the V_H6 gene. This transgenic line rearranges the V3–23 gene only and expresses chimeric antibodies composed of human μ heavy chains paired with murine light chains [14].

Production and expansion of hybridomas

Hybridomas from two fusions of splenocytes were selected on the basis of human μ heavy chain production. A total of 779 clones were obtained at a fusion efficiency of approx. 0·8 × 10⁻⁴. The frequency of human μ-chain-producing lines within the total number of lines was 7%.

Antigen reactivity

Reactivity with antigens was determined at the same concentration of antibodies (1 or 10 μg/ml) using ELISA, immunoprecipitation, agglutination or immunofluorescence assays. In ELISA, microtitre plates were sensitized with antigens (10 μg/ml) by incubation at 4°C overnight and purified chimeric antibody dilutions added as described [14]. In all assays, positive and negative controls were included to confirm specificity and bound antibodies revealed with peroxidase-conjugated goat anti-human μ-chain. Chimeric antibody binding was considered positive if the highest point on the titration curve was > 5% of the maximum signal obtained for the positive control. Immunofluorescence was performed on multiblock sections of rat kidney, liver and stomach fixed in acetone at room temperature for 10 min, cytospins of Crithidia lucillae or Hep-2 cells. Bound antibodies were revealed with FITC-conjugated goat anti-human immunoglobulin.

V_H gene family analysis was determined by ELISA and by Western blotting using antibodies specific for V_H gene families (Dr G. Silverman, UCSD, San Diego, CA) [15]. Expression of cross-reactive idiotype was determined by ELISA using the MoAb 3H7 (kindly provided by Dr S. Suleyman and Professor J. Natvig, Oslo, Norway) [16].

RNA extraction and polymerase chain reaction

Total RNA was extracted with 4 m guanidinium isothiocyanate [17]. Human μ-chain cDNA was synthesized using RNase H reverse transcriptase (Superscript II; Gibco-BRL, Paisley, UK) and human μ-chain constant region antisense primer (Cμ1 5′-TGGAATTCTCACAGGAGACG-3′). Mouse κ-chain cDNA was synthesized using a constant region antisense primer (MCκ1 5′-GTAGAAGTTGTTCAAGAAGCACAC-3′). Polymerase chain reaction (PCR) was performed as previously described [18]. The primers used for the V3–23 PCR were a nested antisense human μ-chain primer (Cμ2 5′-TTAAGCTTGGGGCGGATGCACTC-3′) and a V3–23 leader peptide complementary primer (V3–23L 5′-GCGAATTCTTGTGGCTATTTTAAAAGG-3′). Amplification was carried out using 1 μl cDNA in a 30-μl mixture containing 1·5 mm dNTP, 67 μm Tris–HCl pH 8·8, 6·7 μm MgCl₂, 10 mm β-mercaptoethanol (β-ME), 6·7 μm EDTA, 170 μg/ml bovine serum albumin (BSA), 10% DMSO and primers to a final concentration of 1 μm each. After denaturation at 94°C for 7 min, 0·3 U of Taq polymerase was added and 30 cycles of amplification, consisting of denaturation at 94°C for 1 min, annealing at 56°C for 1 min and extension at 72°C for 1 min, were performed followed by a final 10-min extension at 72°C.

Mouse κ-chain V genes were amplified using a set of primers based on mouse V_κ signal peptide sequences in the EMBL database. These primers were MKS1 (5′-AGSTTSCTGCTAATCAKTG-3′), MKS2 (5′-CCCAGGTCYTYATATTWCT-3′), MKS3 (5′-TRTGGSTRCTKCTGCTCTG-3′), MKS4 (TTGGMWTCTTGTTGCTCTKGT-3′), MKS5 (5′-YGTATWCRTGYTKCTSTGGTT-3′), MKS6 (5′-TCTCCTGTTGCTCTGTTTT-3′), MKS7 (5′-TGTTCTGGATTCCTGYTTC-3′), MKS8 (5′-TKCTGYTCTGGKTMTCWWGG-3′), MKS9 (5′-AACTTCTGCTCTTMCTGCT-3′), MKS10 (5′-TCAGWTTCTTGGAYTTWTGCT-3′), MKS11 (5′-GCCCAGTTCCTGTTTCTG-3′) and MKS14 (5′-GYTGCTKYTGYTSTGGMT-3′) (degenerate nucleotides: Y = C/T; R = A/G; W = A/T; S = G/C; K = T/G; M = C/A). PCR was carried out using a second mouse constant region antisense primer, MCκ2 internal to MCκ1 (MCκ2 5′-AGATGTTAACTGCTCACTGGAT-3′) with Pfu-polymerase (Stratagene Europe, Amsterdam, The Netherlands). Thirty cycles of amplification of 1 min denaturation at 94°C, 1 min annealing at 56°C or 58°C and 1 min extension at 72°C were performed, followed by a final 10-min extension at 72°C. The signal peptide primer required to amplify a particular sequence was determined by two sets of reactions. Amplification reactions were first performed using three groups of primers, A (MKS1, 4, 6, 8 at annealing temperature 58°C), B (MKS2, 9, 10, 14 at annealing temperature 56°C), and C (MKS3, 5, 7, 11 at annealing temperature 58°C). A second series of reactions was then performed using fresh cDNA and individual MKS primers based on the earlier results.

Nucleotide sequencing

PCR products were excised from 1·5% low melting point (LMP) agarose gels and DNA extracted [19]. The PCR products were cloned using a TA cloning kit (Invitrogen, Groningen, The Netherlands). The DNA sequence of plasmid dsDNA minipreps was determined using Sequenase T7 DNA Polymerase (USB, Cleveland, OH) with manual sequencing or using Taq-FS DNA polymerase with automated sequencing (ABI Perkin Elmer Ltd, Warrington, UK). Sequence information was processed and compared with existing sequences in GenBank and EMBL databases using Lasergene software (DNAstar, Madison, WI).

Results

Establishment of hybridomas and antibody reactivity

Thirty-four human μ⁺ clones from two fusions were generated and antibody reactivity determined as described in detail elsewhere [14]. Fusion of unmanipulated splenocytes yielded 10 human μ-secreting clones, while the fusion of cells prestimulated with 10 μg/ml lipopolysaccharide (LPS) yielded 24 clones. The frequency human μ⁺ clones in the two fusions were 2·1% and 9·1%, respectively. Compared with the 3–5% frequency of human μ⁺ cells in the spleen of the transgenic mice, the data suggest that the LPS stimulation increased the frequency of human μ-secreting clones. Table 1 summarizes antibody binding specificity and Table 2 lists the molecular characteristics of the human V3–23 heavy chain and mouse V_κ genes used.

Table 1.

Summary of antigen binding specificity of V3–23-encoded chimeric monoclonal antibodies described in the study

Chimeric antibody	Antigens bound
CN1	Tg*
CN2	Tg
CN4	Tg*
CN5	Tg
CN6	Coll I
CN7	Coll I*, Hep-2
CN10	IgG-Fc, Coll I, Coll II, BSA*, TT*, Hep-2, Kid., Liv.,  Stom.
CN15	IgG-Fc, Coll I, Coll II, Tg, PC, TT, PPS, Hep-2
CN21	CL, PC
CN28	IgG-Fc, Coll I, Coll II
CN30	IgG-Fc, Coll I, Coll II, BSA, TT
CN31	IgG-Fc, Coll.I, Coll.II, CL, PC, BSA, TT
18/2	IgG-Fc, Coll.I, Coll.II, BSA, TT, ssDNA, PPS, Stom.

BSA, Bovine serum albumin; Coll. I and II, bovine collagen type I and II; CL, cardiolipin; ds/ssDNA, double/single-stranded DNA; IgG-Fc, Fc fragment of human IgG1; PC, phosphatidylcholine; PPS, pneumococcal polysaccharides; Tg, thyroglobulin; TT, tetanus toxoid; binding to intracellular antigens was studied using Hep-2 cells; Kid, rat kidney; Liv, rat liver; Stom., rat stomach.

^*The antibodies reacted with the marked antigens weakly and did not give clear titration curves. For comparison, the reactivity of the prototype polyreactive human V3–23 gene-encoded clone 18/2 is also given.

Table 2.

Molecular and specificity characteristics of monoclonal human V3–23 gene and mouse κ light chain encoded chimeric antibodies included in this study

Chimeric antibody	No. of antigens recognized	D gene used	Nucleotide length of the D gene in VHCDR3 (95–102)	No. of possible N-nucleotide additions	JH gene used	VHCDR3 length in amino acids	Vκ gene	Jκ gene
CN1	1	DIR2	6	6	4	5	19	2
CN2	1	DIR2	6	6	4	5	21A	2
CN4	1	DQ52	20	13	2	9	9	5
CN5	1	DQ52	20	13	2	9	9	5
CN6	1	D21/10+D9	10	0	4	8	21E	2
CN7	2	D21/10+D9	10	0	4	8	21E	2
CN10	9	D21/10	3	0	4	4	21A	2
CN15	8	DQ52	5	1	4	4	19	2
CN21	2	D21/10	19	4	4	10	19	2
CN28	3	D21/9	17	5	4	9	19	2
CN30	5	ND	ND	ND	ND	ND	6	2
CN31	8	D21/9+D3	20*	4	3	12	9	1

Assignment of the D genes is based on the analysis given in Fig. 3. V_HCDR3 length is defined as the number of amino acids between the arginine residue encoded by the V_H gene and the fourth amino acid position of the J_H genes (Fig. 2). For some of the chimeric antibodies the precise origin of the D gene remains uncertain and hence the number of N-nucleotide additions could not be determined accurately.

^*The first two nucleotide positions from the presumed D region could possibly be due to palindromic nucleotides (P-additions) [21]. ND, Not determined.

V3–23 gene and mouse V_κ gene expression and sequences

The rearranged V_H and mouse V_κ gene sequences in 12 clones were all in-frame (Figs 1–3). Six V_H rearrangements were identical to germ-line V3–23. Three clones (CN4, 5 and 30) each had one replacement nucleotide interchange in V_HCDR1 and 2. There were eight distinct DJ_H rearrangements. The lengths of the V_HCDR3 (codons 95–102; Kabat et al. [20]) varied from four to 12 amino acids. The lengths of the V_HCDR3 in four hybridomas were shorter than any rearranged V3–23 genes from human B cell hybridomas. The recombination process involved the use of nucleotides from the 3′ non-coding region of V3–23 (GA in CN6, 7 and 31 and G in CN4, 5 and 21), deletion of two to five nucleotides from 5′ J_H genes and the use of as few as two to three nucleotides from the putative D gene (clone CN10). Putative ‘N’ nucleotide additions was seen at the V_HD junctions of one clone (CN21 and possibly 4 and 5) and at the DJ_H junctions in five clones (CN4, 5, 21, 28 and 31). Putative ‘P’ nucleotides [21] were observed at the V_HD junction in clone CN31. In agreement with other studies of J_H gene use in human B cells [4], the majority of the cases rearranged into the J_H4 gene (all from the ‘a’ locus). As has previously been described, the 5′ rearrangement sites of the J_H4 segment tended to occur at specific sites downstream of the 5′ ends, with deletion of few nucleotides from the 5′ end [4].

Fig. 1.

Sequences of V3–23 gene rearrangements in chimeric antibodies described in the study. Numbering is according to Kabat et al. [20]. Codons at selected positions in the FRs and the CDRs are shown for brevity. All other codons not shown are identical to the germ-line V3–23. Sequences of the V3–23 germ-line gene and the J_H genes are given in bold. Dashes represent identity. Nucleotide changes leading to amino acid replacements are in uppercase and silent changes are in lowercase letters. The predicted amino acids are given above each nucleotide codon.

Fig. 3.

D_H region gene segment rearrangements in the chimeric antibody-producing hybridomas. Sequences are shown aligned to the most closely related germ-line D_H sequences. Positions of sequence homology are indicated by a dash. Underlined bases are likely to be from the 3′ non-coding region of the V3–23 gene. Putative P nucleotide additions are indicated by lower case letters. Possible N-additions are given in bold italics.

Comparison of the deduced amino acid sequences showed that some of the antibodies with similar antigen reactivity had V_HCDR3 of similar length and amino acid composition. However, some antibodies (CN10, 15, 28 and 31) reacted with the same antigens despite having different V_HCDR3 and different light chains. This observation suggests that reactivity with these antigens may involve sequences intrinsic to the V3–23 gene itself.

Somatic hypermutation

The rearranged V3–23 genes in nine clones were identical, or near identical, to the V3–23 germ-line gene. The single changes in clones CN4 and 5 were identical and led to a conservative interchange in CDR1. However, the serine codon AGC is an intrinsic hot spot for random mutations [22]. The nucleotide change from G to T in clone CN30 resulted in replacement of the glycine to cysteine at position 54 in the CDR2. Again, the glycine codon GGT at position 54 is a hot spot for random mutation in the V3–23 gene, although the replacement of G to T was not seen as a favoured replacement [22].

The rearranged V_κ genes in 11 clones were 99·3–100% homologous with the closest related germ-line gene (Fig. 2). The V_κ gene in clone CN30 was 93·3% homologous with the MUSIGKV8 gene. The rearranged V_κ genes in clone CN10 had two nucleotide differences from the MMVK21A germ-line gene and both resulted in conservative amino acid replacements. It is possible that the interchange in CDR1 could be due to somatic mutation, since it was absent in CN2. The second nucleotide replacement, A to G at position 73, of these two hybridomas could however, represent a new allotypic variant.

Fig. 2.

Sequences of mouse V_κ genes in the chimeric antibodies described in the study. Codons at selected positions in the FRs and the CDRs are shown for brevity. All other codons not shown are identical to the closest indicated germ-line mouse. Sequences of mouse V_κ genes and J_κ genes are given in bold. The predicted amino acids are given above each nucleotide codon of the top germ-line gene MMVK21A. Predicted amino acid sequences in the remaining germ-line and rearranged genes are indicated only if different from MMVK21A.

Discussion

This study describes the molecular structure and antigen binding activity of chimeric antibodies encoded by the human V3–23 heavy chain variable region gene and mouse V_κ genes expressed in transgenic mice. The transgenic mice carry two human genomic cosmids containing various gene segments including Cμ, J_H and D_H, as well as V3–23 [10]. All chimeric antibodies had human μ-chains expressing the V3–23 gene and mouse V_κ genes. Sequence analysis of the V3–23 gene from 12 hybridomas revealed eight different V_HCDR3 rearrangements utilizing varied combinations of D-J_H segments probably arising from eight to nine different precursor cells. Further, although the human minilocus contains sequences necessary for hypermutation, only three of the rearranged V3–23 genes had nucleotide changes resulting in amino acid interchanges. Whilst the lack of hypermutation was not unexpected, it was important, nonetheless, to determine that mutation in the V3–23 gene in these adult mice, which had been kept in non-sterile conditions, did not influence antibody reactivity. Lack of extensive mutation therefore confirms that the reactivity pattern with antigens seen in this study reflects reactivity of antibodies encoded by the V3–23 gene in germ-line, or near germ-line configuration. The information was also relevant to determine if mouse B cells expressing the V3–23 were selected as a result of interaction with mouse self antigens. In this respect it was noteworthy that the only clone with putative mutations in the mouse V_κ gene was obtained after LPS stimulation, which suggests that clones with mutations, presumably reacting with antigens encountered in vivo, in this founder colony of normal mice may have been functionally inactivated. However, among the 12 hybridomas, four used mouse V_κ21 genes (two V_κ21A and two V_κ21E), which was previously shown to be over-expressed in autoantibodies to DNA in the lupus-prone MRL/lpr mouse [23]. This gene group represents 4–8% of mouse V_κ germ-line genes; it is therefore over-represented in this panel of chimeric hybridomas [24]. The cause of this over-expression in the transgenic mice is not clear, but may suggest that the V3–23 gene could only associate with a limited number of mouse light chain V genes to produce functional antibodies.

Previous studies of the molecular basis of reactivity of antibodies encoded by the V3–23 with antigens were carried out using human B cell hybridomas selected on the basis of their specificity or idiotype expression [6,7]. By and large, these studies predicted that reactivity with self antigens was primarily the consequence of the V3–23 gene itself and independent of light chains. In all previous studies of V3–23 encoded antibodies however, the antibodies tested had the native pair of immunoglobulin heavy and light chains. In contrast, each of the functionally rearranged chimeric antibodies characterized in this study were co-expressed with mouse light chains. Despite the lack of selection for specificity or idiotype expression, the chimeric antibodies had a range of binding reactivities broadly reflecting the pattern seen in human V3–23-encoded antibodies (e.g. antibody 18/2). These data therefore suggest that conformations and contact residues that determine reactivity with some of the antigens tested in our study and in previous studies may primarily be dependent on the heavy chain. Further, although the data suggested that V_HCDR3 was crucial for binding to some antigens (cardiolipin, BSA, rat tissue sections and polysaccharides), the data also suggested that CDR1, CDR2 and/or framework residues may determine, or influence reactivity with IgG-Fc, collagen types I and II. This is consistent with studies showing that contact residues in antibodies encoded by the human V4–34 gene with the i/I antigen on erythrocytes lie in the FR1 [12,25,26]. In addition, crystallographic studies of a human rheumatoid factor antibody bound to its antigen (human IgG), revealed that the contact regions lay outside the ‘traditional’ antigen-binding pocket [13]. The exception to this suggestion however, appears to be binding to dsDNA, which none of the chimeric antibodies has demonstrated. Thus, these data suggest that binding to dsDNA may be crucially dependent on V_HCDR3 and/or influenced by the light chain in potentially pathogenic antibodies found in SLE patients.