V_H and Vκ Segment Structure of Anti-Insulin IgG Autoantibodies in Patients with Insulin-Dependent Diabetes Mellitus

Abstract

In some patients with insulin-dependent (type I) diabetes mellitus (IDDM), autoantibodies to insulin are present at diagnosis. After initiation of the treatment with not only animal but also human insulin, anti-insulin, mainly IgG, autoantibodies become a major component of the autoimmune response in virtually all IDDM patients. Their structure, however, is still relatively unknown. We analyzed the structure of the V_H and Vκ segments of three human IgG mAb derived from three IDDM patients. The sequences of V_H genes of two IgG, mAb13 and mAb48, were 98.3 and 96.6% identical with those of the H11 and 1.9III genes (V_HIII family), respectively. The sequence of the V_H gene of the third IgG, mAb49, was 98.6% identical with that of the 51p1 gene (V_HI family). All three IgG mAb used VκIII segments. The VκIII gene sequences of mAb13 and mAb49 were 97.9 and 98.9% identical, respectively, to that of the kv3g gene; the mAb48 Vκ gene sequence was 96.5% identical to that of the kv328 gene. The V_H and/or Vκ segments of these anti-insulin IgG mAb are similar to Ig V genes expressed in the fetal, and adult normal and autoimmune B cell repertoires. The nucleotide differences displayed by the three anti-insulin IgG mAb V_H gene sequences, when compared with those of the closest reported germ-line genes, were concentrated in the CDR (6.2 × 10^–2 and 0.8 × 10^–2 difference/base in CDR and FR, respectively; p < 0.01, χ² test), and yielded a significantly higher putative replacement (R) to silent (S) mutation ratio in the CDR (12.0) than in the framework (0.2). The concentration of nucleotide differences in the CDR and their high R:S putative mutation ratios were consistent with the hypothesis that these expressed V_H genes underwent a process of somatic mutation and Ag-driven clonal selection. That such differences constituted somatic point-mutations was formally proved in IgG mAb13, by differentially targeted PCR amplification and Southern blot hybridization of the mAb13-producing cell line DNA. The putative germ-line gene that gave rise to the expressed V_H segment was cloned using genomic DNA from PMN of the same patient whose B cells were used for the generation of this mAb. Overall, in the anti-insulin IgG mAb V_H and VκIII genes, the (putative and verified) somatic point-mutations yielded 27 amino acid replacements, of which 14 nonconserved. Four of these resulted in positively charged residues, three Arg and one His. Additional two single and three tandem Arg residues were present in the H chain CDR3 of the mAb13 and mAb48, respectively. Thus, the rearrangement of Ig V genes that are commonly expressed in the human B cell repertoire, in conjunction with a process of somatic mutation and Ag-driven clonal selection can underlie the emergence of high affinity anti-insulin IgG autoantibodies in IDDM patients.

The genetic composition and the somatic changes of antibodies undergoing affinity maturation to different foreign Ag have been analyzed thoroughly in mice (1–4). In humans, the few reported studies of specific antibodies induced by foreign, mainly complex, Ag, including inactivated rabies virus vaccine (5, 6), HIV-1 (7), HSV-1 (8), and Haemophilus influenzae type b capsular polysaccharide (9), have suggested that the Ig V genes used by these antibodies extensively overlap with the pool of Ig V genes used by natural antibodies or autoantibodies and are expressed in the fetal and adult B cell repertoires. They also showed that some of these antibodies underwent a process of somatic hypermutation and Ag-driven clonal selection. Similar features seem to be displayed by human autoimmune disease-related autoantibodies, such as anti-DNA antibodies in SLE patients⁵ (10–12) and rheumatoid factors (RF)⁶ in rheumatoid arthritis patients (13–16). However, although these autoantibodies appear to be specific and display a high affinity for the relevant self Ag, they might be induced by unrelated cross-reacting, perhaps foreign, Ag. For instance, it has been suggested that at least some anti-DNA antibodies can be elicited by cross-reacting structures on bacteria (17). As emphasized by Thomas (18), the administration of recombinant human insulin for therapeutic purposes provides a unique opportunity for the structural analysis of specific autoantibodies actually induced by a self Ag, to which naturally occurring antibodies exist in the normal B cell repertoire (19, 20).

In patients with insulin-dependent (type I) diabetes mellitus (IDDM), antiislet cell surface and anti-insulin receptor IgG autoantibodies are present in the circulating blood before the development of overt disease and possibly play a major immunopathologic role by destroying the insulin-producing β-cells in the pancreas (reviewed in Ref. 21). Some circulating anti-insulin and antiproinsulin autoantibodies also exist in these patients before the development of overt disease (22–24). Their titer increases dramatically after administration with not only animal but also human insulin and they may complicate the therapeutic treatment (25). The structure of the V_H segments of six anti-insulin IgM antibodies appearing after insulin treatment, has been recently reported (18, 26, 27). This, however, may not reflect the structure of the bulk of the specific high avidity anti-insulin autoantibodies, which are mainly IgG (28).

We report the complete structure of the V_H and Vκ segments of three anti-insulin IgG mAb generated from IDDM patients treated with recombinant insulin. The three IgG mAb were specific and displayed relatively high affinity for human insulin. They used three distinct V_H genes in combination with VκIII genes. The V genes used by these self Ag-induced IgG mAb are similar to those commonly expressed in the human B cell repertoire at different stages of ontogeny. When compared with those of the closest reported germ-line V genes, their sequences displayed a number of differences that were consistent in nature and distribution with those resulting from an Ag-dependent clonal selection. In one IgG mAb, these differences were formally proved to represent somatic point-mutations. Thus, in IDDM patients, anti-insulin IgG autoantibodies are heterogeneous in structure and can undergo affinity maturation through a somatic mutation and clonal selection process driven by homologous insulin.

Materials and Methods

Generation and analysis of human anti-insulin IgG mAb

The monoreactive anti-human insulin IgG1 mAb13-, IgG3 mAb48-, IgG1 mAb49-producing cell lines were established from three IDDM patients treated with human recombinant insulin, by EBV transformation and somatic cell hybridization techniques, and have been reported (28). There, it was stated that the subject source of mAb48 was a newly diagnosed IDDM patient. In fact, careful reinspection of the clinical records showed that this patient received few injections of human recombinant insulin before the withdrawal of the peripheral blood used as a source of B cells for the generation of mAb48. The mAb bindings to human recombinant insulin (Eli-Lilly, Indianapolis, IN), ssDNA, human thyroglobulin, tetanus toxoid, and BSA were measured using specific ELISA (28–30). The mAb affinity for human insulin was expressed as K_d calculated as described (28–30).

Cloning and sequencing of expressed Ig V_H and Vκ genes

mRNA was extracted from the mAb-producing cells and first strand cDNA was synthesized using MMTV reverse transcriptase (6, 14, 16, 31). The sense HA-1 and HI-3 and antisense HI-1 oligonucleotide primers were synthesized and used to amplify the V_H gene cDNA. The HI-3 primer sequence was identical to a leader sequence (5′ TTGGGCTGTGCTGGGTTTTCCT 3′) that is relatively conserved among members of the V_HIII family. The degenerate HA-1 primer sequence (5′ GGGAATTCATGGACTGGACCTGGAGG(AG)TC(TC)TCT(GT)C 3′) was highly similar to a leader sequence that is relatively conserved among members of the V_HI family (6, 16, 31). The anti-sense HI-1 Cγ primer sequence (5′ TAGTCCTTGACCAGGCAGCC 3′) was the reverse-complement of a 5′ portion of the IgG C region gene conserved in all four human IgG subclasses (6, 31). The sense VκI-II, VκIII, VκIV, and anti-sense Cκ oligonucleotide primers were synthesized and used to amplify the expressed Vκ gene cDNA (16). The degenerate VκI-II primer sequence (5′ AGCTCCTGGGGCT(GC)CT(AG)(AC)TGCTCT 3′) was similar to a leader portion of both the VκI and VκII genes; the VκIII primer sequence (5′ TCTCTTCCTCCTGCTACTCTGGCT 3′) was identical to a leader portion of the VκIII gene; the VκIV primer sequence (5′ ATGGTGTTG-CACACCCAGGTCTTC 3′) was identical to a leader portion of the VκIV gene. The anti-sense Cκ oligonucleotide primer sequence (5′ CTGCTCATCAGATGGCGGGAAGA 3′) was the reverse complement of a 5′ sequence of the human Cκ gene (16). PCR was performed in a 50 μl volume containing 10 mM Tris hydrochloride, 1.5 mM MgCl₂, 200 μM of each dNTP, 10 pmol of each oligonucleotide primer, and 2.5 U of Taq DNA polymerase (Perkin Elmer Cetus, Norwalk, CT) for 25 cycles (1 min at 94°C, 2 min at 55°C, and 2 min at 72°C). Amplified DNA was cloned into pCR1000 vector (Invitrogen, La Jolla, CA) and dideoxy sequencing was performed using plasmid dsDNA prepared from selected bacterial clones as reported (6). Each V_H and V_K gene sequence was determined by analyzing at least four independent clones isolated from an identical source. Intraclonal nucleotide variations were less than 0.001/base, and were excluded from analysis. Sequences were analyzed using the software of the Genetic Computer Group of the University of Wisconsin, Release 6, the GenBank and EMBL databases, and a Model 6000-410 VAX computer (Digital Equipment Corp., Marlboro, MA) (6).

Analysis of genomic V_H segment from PMN DNA and hybridoma DNA

PCR amplifications of V_H segment sequences from genomic DNA of the mAb13-producing cells and genomic DNA from autologous PMN using different combination of the 13CDR1, HI-2, and HI-3 oligonucleotide primers were performed. The sequence of the sense 13CDR1 primer (5′ TTCACCTTCAGTGGGTACTGGAT 3′) encompassed a FR1-CDR1 portion (residues 79–101) of the mAb13 V_H gene that was different in two nucleotides from the corresponding area of the germ-line H11 gene. The sequence of the anti-sense HI-2 primer (5′ TAAACAGCCGTGTCCTCGGCTC 3′) was the reverse complement of a FR3 sequence (residues 281–260) conserved among members of the V_HIII gene family, and differed in one base from the corresponding sequence of the mAb13 V_H gene. Genomic DNA (100 ng) was subjected to PCR (50 μl volume) with 10 pmol of the HI-2 oligonucleotide and 10 pmol of either the 13CDR1 or the HI-3 oligonucleotide. Each cycle consisted of denaturing, annealing, and extension steps of 1 min, and at temperatures of 94, 62, and 72°C, respectively. After 30 cycles, the products amplified using genomic DNA extracted from autologous PMN or from mAb13-producing cells were fractionated on 1.7% agarose gel containing 1 μg/ml of ethidium bromide. DNA was transferred to a filter membrane and hybridized with the γ-³²P-labeled oligonucleotide 13CDR1 at 61°C. The membrane was washed twice with 2X SSC/0.5% SDS at room temperature for 10 min, and twice with 1X SSC/0.5% SDS at 61°C for 20 min, before exposure to Kodak XAR film. The PCR product amplified from autologous PMN genomic DNA using the HI-3 and HI-2 primer pair was also fractionated on 2% low melting agarose gel. The DNA fragment were isolated, cloned, and sequenced.

Results

Generation of anti-human insulin IgG mAb-producing cell lines and mAb analysis

The insulin-specific IgG1 mAb13, IgG3 mAb48, and IgG1 mAb49 were produced by cell lines established using circulating B cells from three IDDM patients who had been administrated recombinant human insulin (28). These patients had never been treated with heterologous insulin. Their H chain isotypes, L chain types, and K_d values for human insulin were reported previously (28) and are listed in Table I; their dose-saturable bindings to insulin, ssDNA, human thyroglobulin, tetanus toxoid, and BSA are depicted in Figure 1.

Table I.

V_H, D, and J_H, and Vκ and Jκ segments of human IgG mAb to homologous insulin

			VH Segment								VκIII Segment
			Closest germ-line genea (family)		Number of nucleotide differences						Closest germ-line gened (family)		Number of nucleotide differences
Clone	H L Chains	Kd (M)		Nucleotide % identity	CDR		FR					Nucleotide % identity	CDR		FR
					R	S	R	S	Db	JHc			R	S	R	S	Jκ5e
mAb13	IgCI, κ	5.0 × 10−7	H11 (VHIII)	98.3	4 (0.8)f	0	0 (2.9)	1	DXP'1	JH2	kv3g (VκIII)	97.9	1 (1.5)	0	5 (4.5)	0	Jκ1
mAb48	lgG3, κ	8.0 × 10−7	1.9111 (VHIII)	96.6	5 (1.6)	1	0 (5.9)	4	D21-9	JH6	kv328 (VκIII)	96.5	7 (1.7)	1	1 (5.1)	1	Jκl
mAb49	IgGI, κ	2.5 × 10−7	51 p1 (VHI)	98.6	3g (0.6)	0	1 (2.4)	0	DN1	JH4	kv3g (VκIII)	98.9	1 (0.6)	0	2 (1.7)	0	Jκl

^aThe sequence of the germ-line H11 gene has been reported by Rechavi et al. (32); the sequence of the 1.9111 gene has been reported by Berman et al. (33); the sequence of the 51 p1 gene has been reported by Schroeder et al. (34).

^bD genes have been reported by Ichihara et al. (36).

^cJ_H genes have been reported by Ravetch et al. (41).

^dThe sequence of kv3g and kv328 has been reported by Chen et al. (43).

^eJκ genes have been reported by Heiter et al. (44).

^fExpected numbers of R mutations calculated as reported in Results.

^gIncludes the G to T change at position 146, resulting in a G to V amino acid mutation immediately preceding the CDR2 sequence.

FIGURE 1.

Binding of the IgG mAb13, IgG mAb48, and IgG mAb49 to human recombinant insulin (●), tetanus toxoid (■), ssDNA (△), BSA (□), and thyroglobulin (△). The Ag-binding activity of each mAb is expressed as optical absorbance at 492 nm.

V_H segments of anti-insulin IgG mAb

The nucleotide and deduced amino acid sequences of the anti-insulin IgG mAb V_H segments are depicted in Figure 2, A and B, respectively. The nucleotide differences and predicted amino acid changes when compared with the reported closest germ-line V_H gene sequences are summarized in Table I. The mAb13 V_H gene sequence displayed the highest degree of identity with that of the germ-line H11 gene (V_HIII family) (32), differing in four nucleotides within the coding region. All three nucleotide differences in the CDR resulted in putative amino acid changes, whereas that in the FR3 was silent. The mAb48 V_H gene sequence displayed the highest degree of identity with that of the germ-line 1.9III gene (V_HIII family) (33). Five of the six nucleotide differences in the CDR resulted in putative amino acid changes. All five differences in the FR were silent. The mAb49 V_H gene sequence displayed the highest degree of identity to that of the 51p1, a (germ-line) V_H gene expressed in the fetal liver (34). Both pairs of nucleotide differences in the CDR and FR2 resulted in amino acid changes. One of the FR2 amino acid changes, a Val instead of Gly at position 49, was adjacent to the CDR2, and was considered in our nucleotide distribution calculations as belonging to the CDR2 (Table I). Thus, the comparison of the anti-insulin IgG V_H gene sequences with those of the closest reported germ-line genes showed a total of 19 nucleotides differences. These were significantly more frequent in the CDR, 6.2 × 10^–2 difference/base, than in the FR, 0.8 × 10^–2 difference/base (p < 0.01, χ² test), and yielded putative nucleotide R:S mutation ratios of 12.0 (12:1) in the CDR and 0.2 (1:5) in the FR.

FIGURE 2.

Nucleotide (A) and deduced amino acid (B) sequences of the V_H genes of the IgG mAb to human insulin. The top sequence in each cluster is given for comparison and represents the reported germ-line V_H gene sequence displaying the highest degree of identity to those of the expressed V_H genes. The H11 and 1.9III V_H genes belongs to V_HIII family (32, 33). 51p1 is a fetal expressed gene member of the V_HI family (34). Dashes indicate identities. Solid lines on the top of each cluster depicts CDR. Small letters denote leader introns. 13G12 is the germ-line sequence we obtained from PMN DNA of the subject whose B cells were used for the generation of mAb13 (see Materials and Methods and Results). The sequences or reverse complements of the sequences encompassed by the oligonucleotide primers HI-3, HI-2, and 13CR1 are underlined. The present sequences are available from EMBL/GenBank/DDBJ under accession numbers D-16832, D-16833, D-16835, and D-16837.

D and J_H genes of anti-insulin IgG mAb H chain, and CDR3 and FR4 structure

The D segment sequences were 28 to 49 bases in length, and displayed the highest degree of identity to those of the germ-line DXP’1, D21-9, and DN1 genes, respectively (35–37) (Fig. 3A). When compared with the germ-line DN1 gene sequence, that of the mAb49 D segment displayed two adjacent nucleotide differences, TG instead of CA, which are also shared by the expressed fetal M26 D gene (36), suggesting a polymorphism of the DN1 gene or the existence of a variant gene tentatively designated as DN2 (39, 40). The mAb13, mAb48, and mAb49 used truncated forms of the germ-line J_H2, J_H6, and J_H4 genes (41), respectively (Fig. 3B). The mAb48 J_H6 segment nucleotide sequence was identical to that of the reported J_H6b gene, except for one base (40); that of the mAb49 J_H4 segment was identical to the reported J_H4b gene sequence (40).

FIGURE 3.

Nucleotide (A) and deduced amino acid (B) sequences of the D and J_H segments of the mAb to human insulin. Germ-line D gene sequences are given for comparison. Dashes indicate identities. The deduced amino acid sequences are separated in CDR3 and FR4. The present sequences are available from EMBL/GenBank/DDBJ under accession numbers D-16833, D-16835, and D-16837.

The deduced amino acid sequences of the expressed D-J_H genes were segregated into CDR3 and FR4 according to Kabat et al. (42), and are depicted in Figure 3C. The length, 10 to 20 amino acids, and the sequences of the CDR3, were highly divergent. The deduced amino acid sequences of the FR4 were relatively conserved.

Vκ and Jκ segments of anti-insulin IgG mAb

The nucleotide and deduced amino acid sequences of the mAb13, mAb48, and mAb49 VκIII segments are depicted in Figure 4, A and B, respectively. Their nucleotide differences and predicted amino acid changes, when compared with the reported closest germ-line VκIII gene sequences, are summarized in Table I. The mAb13 Vκ gene sequence displayed highest degree of identity with that of the germ-line kv3g gene (VκIII subgroup) (43), differing in one base in the CDR, and in five bases in the FR. The mAb49 Vκ gene sequence was identical to that of the kv3g gene except for one base in the CDR3 and two in the FR. The mAb48 Vκ gene displayed the highest degree of identity to that of the germ-line kv328 gene (VκIII subgroup) (43), differing in eight bases in the CDR, and in two bases in the FR. Thus, overall comparison of the anti-insulin IgG mAb VκIII gene sequences with those of the closest germ-line genes showed a total of 18 nucleotide differences. These yielded seven and eight putative amino acid replacements in the CDR and FR, respectively. All three antiinsulin IgG mAb used a similar form of the Jκ1 gene (44) (Fig. 4, C and D). However, mAb13 and mAb48 Jκ1 predicted amino acid sequences each displayed one residue difference compared with the Jκ1 sequence of mAb49.

FIGURE 4.

Nucleotide (A and C) and deduced amino acid (B and D) sequences of the VκIII and Jκ genes of the IgG mAb to human insulin. The top sequence in each cluster is given for comparison and represents the reported germline (kv3g, kv328, and Jκ1) gene sequence (43) displaying the highest degree of identity to the expressed VκIII genes of the cluster. Dashes indicate identities. Solid lines on the top of each cluster depicts CDR. The present sequences are available from EMBL/GenBank/DDBJ under accession numbers D-16834, D-16836, and D-16838.

Somatic point-mutations in IgG mAb13 V_H segment

The high degree of conservation of a number of V_H genes, including H11, 1.9III, and 51p1, in the human population suggested that anti-insulin IgG mAb13, mAb48, and mAb49 V_H segments constituted somatically mutated forms of the germ-line H11, 1.9III, and 51P1 V_H genes, respectively, or closely related V_H genes. Although this hypothesis was further supported by the distribution and the high R:S mutation ratios of the putative nucleotide changes in the anti-insulin IgG mAb V_H segments, it needed to be formally validated at least in one mAb. To formally demonstrate that the nucleotide differences of the mAb13 V_H gene sequence compared with that of the germ-line H11 gene represented somatic point-mutations, we performed PCR amplifications using ad hoc designed oligonucleotide primers and genomic DNA from autologous PMN or genomic DNA from the mAb13-producing cell hybridoma. The sense 13CDR1 primer, encompassing most of the mAb13 CDR1 and a 5′ flanking FR1 sequence, and differing in two nucleotides from the corresponding H11 gene sequence, was used in conjunction with the antisense HI-2 primer, encompassing a FR3 sequence identical in the expressed mAb13 and H11 V_H genes, except for a T instead of G in position 267 (Fig. 2A). The two combined primers yielded an amplification product when used in PCR involving genomic DNA from the hybridoma but not from autologous PMN (Fig. 5A). The size of the amplified product (~200 bp) was consistent with that of the sequence spanning residues 89 to 281 of the mAb13 V_H gene. Utilization of the same antisense (HI-2) primer in conjunction with the sense HI-3 primer encompassing a leader sequence stretch shared by the mAb13 and H11 genes (Fig. 2A), resulted in the amplification of a ~430 bp product from both hybridoma and PMN genomic DNA (Fig. 5B). Southern blot analysis showed that the ~430 bp DNA amplified from hybridoma hybridized with the γ-³²P-labeled l3CDR1 oligonucleotide probe, although that from PMN did not (Fig. 5C). These experiments suggested that the expressed mAb13 V_H gene constituted a somatically mutated form of the germ-line H11 or a H11-like gene. To identify the autologous germ-line gene that putatively gave rise to the expressed mAb13 V_H gene, the DNA amplified from PMN DNA was cloned and sequenced. Ten V_HIII family gene sequences were derived (data not shown). Three clones contained sequences more than 99% identical with that of H11, and the consensus sequence (13G12) is shown in the alignment with the H11 sequence (Fig. 2A). The nucleotide sequence of 13G12 differed from that of H11 in two bases that were both in the CDR2. One of the two nucleotide differences (G instead of C at position 176) was shared with the mAb13 V_H sequence, further suggesting that the 13G12 was the germ-line V_H gene that gave rise to the mutated mAb13 gene. The comparison of the germ-line 13G12 gene sequence with that of the expressed mAb13 V_H segment suggested a total of four somatic point-mutations. Three of them were located in the CDR and one in the FR3. All differences in the CDR yielded amino acid replacements and that in the FR3 was silent.

FIGURE 5.

Evidence somatic hypemutation in the mAb13 V_H gene: A, Etidium bromide staining of amplified DNA fractionated in agarose gel electrophoresis (10 μl of reaction mixture were applied to each lane). Using the 13CDR1 and HI-2 oligonucleotide primers (see Materials and Methods and Results), an amplification product of appropriate size (last 3′ portion of the V_H segment, ~200 bp) was obtained by priming genomic DNA from mAb13-producing cells (hybridoma DNA), but not genomic DNA from autologous PMN (PMN DNA); B, Etidium bromide staining of amplified DNA fractionated in agarose gel electrophoresis (10 μl of reaction mixture were applied to each lane). Using the V_H gene leader (HI-3) and FR3 (HI-2) specific oligonucleotide primers (see Materials and Methods and Results), amplification products of identical and appropriate size (~430 bp) were obtained by priming genomic DNA from the mAb13-producing cells (hybridoma DNA) and autolobous PMN; C, Southern blot hybridization of the PCR products shown in B with the γ-³²P-labeled oligonucleotide probe (13CDR1) encompassing the CDR1 sequence of the expressed mAb13 V_H gene (see Materials and Methods and Results). A positive signal was detected only with DNA amplified from the mAb13-producing B cells.

Overall configuration of V_H and Vκ segments of human anti-insulin IgG mAb

In the absence of negative or positive selective pressure on a gene product, R and S mutations are scattered randomly throughout the protein sequence. If a DNA segment displays a number of R mutations lower than that expected by chance only, it is likely that a selective pressure to maintain the protein structure was exerted. Conversely, if a DNA segment displays a number of R mutations higher than that expected by chance, it is likely that a positive selective pressure to mutate was exerted. The numbers of expected R mutations in the anti-insulin IgG mAb V_H and VκIII segment CDR and FR were calculated using the formula n × Rf × CDRf (or FRf), where n is the total number of observed mutations, Rf is the expected proportion of R mutations (0.75, see Ref. 45), and CDRf is the relative size of the CDR (or FR) (0.22 and 0.78 for the CDR and FR, respectively, of the H11, 1.9III, and 51p1 V_H genes; 0.25 and 0.75 for the CDR and FR, respectively, of the VκIII segment). Consistent with a clonal selection by Ag, the mAb13, mAb48, and mAb49 V_H segments displayed higher and lower numbers of R mutations in the CDR and FR, respectively, than those theoretically expected (Table I). The mAb48 VκIII segment, but not those of mAb13 and mAb49, displayed a similar clear pattern of higher and lower number of R mutations in the CDR and FR, respectively. The major role in Ag binding is putatively played by the antibody V_H segment (46, 47). Thus, we calculated the probabilities that the excess R mutations arose by chance in the antiinsulin mAb V_H segment CDR and FR using the binomial distribution model p = [n!/k!(n – k)!] q^k (1 – q)^{n – k}, where q = Rf × CDRf or Rf × FRf is the probability a R mutation will locate to the V_H segment CDR (q = 0.22 × 0.75) or FR (q = 0.78 × 0.75), and k = number of observed R mutations in the CDR or FR (48). The likelihood that the excess R mutations arose by chance in the mAb13, mAb48, and mAb49 V_H segment CDR were p = 0.003, p = 0.012, and p = 0.008, respectively. The probability that the scarcity of R mutations in the mAb49 V_H segment FR resulted from chance was p = 0.17. The mAb13 and mAb48 V_H segment FR did not display any R mutation. Thus, the mAb13-, mAb48-, and mAb49-producing cells were subjected to positive and negative pressures to mutate the expressed V_H segment CDR and FR structures, respectively.

The features of the putative and proved amino acid changes in the mAb13, mAb48, and mAb49 V_H and VκIII segments are summarized in Table II. Of the total 27 changes observed, 14 were nonconserved. Four of them (one in the mAb13 V_H segment, two in the mAb48 VκIII segment, and one in the mAb49 VκIII segment) resulted in positively charged residues, three of which were Arg and one His. The three anti-insulin mAb H chain CDR3 differed in their Arg content. Three tandem Arg residues were present in the mAb48 H chain CDR3, possibly stemming from N segment additions. Two single Arg residues were in the of mAb13 H chain CDR3 and no Arg was present in the CDR3 of mAb49 H chain. No Pro residue was found in any of the three mAb H chain CDR3. However, in all three anti-insulin IgG mAb, a Pro was encoded by the last codon of the VκIII gene. In addition, in the mAb48 VκIII-Jκ1 junction, one extra Pro residue, encoded by a CCG codon, possibly generated by joining of the CC following the coding sequence of the kv328 gene with the G preceding that of the Jκ1 gene, was present (43, 44).

Table II.

Deduced amino acid changes in V_H and V_L segments of human IgG mAb to homologous insulin

	VH Segment				VκIII Segment
Clone	Region	Residue	Germ-line	Expressed	Region	Residue	Germline	Expressed
mAb13	CDR1	31	S polar	G polar	FR1	10	T polar	I
						14	S polar	C polar
	CDR2	57	S polar	T polar	CDR1	28	S polar	N polar
		59	T polar	Ra positive	FR2	49	Y polar	S polar
					FR3	76	S polar	N polar
						77	S polar	N polar
mAb48	CDR1	31	S polar	A	CDR1	28	S polar	T polar
		33	A	G polar		29	V	I
	CDR2	52	S polar	W	FR3	79	Q polar	E negative
		56	S polar	N polar	CDR3	90	Q polar	H positive
		61	A	T polar		93	N polar	R positive
						94	W	L
mAb49	CDR1	33	A	T polar	FR2	40	P	A
	FR2	38	R positive	O polar	FR3	77	S polar	R positive
		49	G polar	V	CDR3	93	N polar	T polar
	CDR2	66	G polar	D negative

^aNonconserved amino acid changes are underlined.

Discussion

We analyzed the structure of the V_H and Vκ segments of three human specific IgG mAb induced by recombinant homologous insulin. We found that: 1) these IgG mAb utilized V_H and VκIII genes that are commonly expressed in human fetal and adult B cell repertoires; 2) when compared with those of the closest germline V genes, the antiinsulin IgG mAb V segment sequences displayed a number of differences, which were formally proved in the mAb13 V_H segment to represent somatic point-mutations; 3) the putative and verified somatic point-mutations in the antiinsulin IgG mAb displayed a distribution and nature consistent with those resulting from a process of clonal selection driven by Ag.

This is, to the best of our knowledge, the first structural analysis of human insulin-induced autoantibodies of the IgG class. Previous studies have been limited to IgM (15, 18, 26). Of the six reported insulin-induced IgM mAb derived from patients with IDDM, one (26, 27) used a V_HV gene, V_H 251, closely related to the 83p2 gene, one (18) used a possibly mutated form of the 20p3 V_HI gene, and the remaining four (18) used 22-2B and/or 22-2B-like V_HIII elements. The 83p2 and 20p3 V_H genes are expressed in the developmentally restricted fetal B cell repertoire (34). The 22-2B gene is commonly expressed in the human B cell repertoire to encode different specificities, including, as we showed, that of a high affinity anti-rabies virus IgA mAb105 induced by active immunization (5, 6). The insulin-induced specific IgG mAb used three distinct V_H genes. mAb49 utilized a V_HI gene, 51p1, different from 20p3, the gene utilized by the anti-insulin IgM mAb 22 (18), but like 20p3 expressed in the fetal repertoire (34). The remaining two anti-insulin IgG, mAb13 and mAb48, utilized V_H genes, H11 and 1.9III, respectively, which are not closely related to 22-2B, but belong to the same family, V_HIII (32, 33). In addition, the V_H genes of the insulin-induced specific IgG mAb are used by natural antibodies or autoantibodies in healthy subjects, specific autoantibodies in autoimmune patients, and foreign Ag-induced specific antibodies (49). For example, the sequence of the mAb13 V_H segment (H11) is closely related to those of the natural polyreactive IgM mAb55 (5, 6) and the high affinity IgM RF mAb60 (14). The sequence of the mAb49 V_H segment (51P1) is closely related to that of the V_H segment of the anticardiolipin and ssDNA R149 mAb (11), to that of the V_H segment of the anticardiolipin Kim13.1 mAb (50), and to that of the V_H segment of the monoreactive anti-IgG IgM RF-TS1 mAb (51). The sequence of the mAb48 V_H segment (1.9III) is closely related to those of the anti-dsDNA Kim4.6 mAb (52), and the high affinity anti-IgG IgM RF-TS3 and RF-SJ3 autoantibodies (51). Finally, the V_H 1.9III and/or 1.9III-like genes are utilized by specific antibodies induced by vaccination with Haemophilus influenzae type b capsular polysaccharide (9).

Thus, the V_H genes segments of the insulin-induced specific IgG antibodies belong to the same clusters of genes used by insulin-induced IgM antibodies and overlap with those used by natural antibodies and specific autoantibodies, as well as with those used by specific antibodies induced by foreign Ag. This is further emphasized by the antiinsulin IgG mAb D genes, DXP’1, D21-9, and DN1, which are among the most commonly expressed D genes in the adult B cell repertoire (37, 40, 49). The apparent overutilization of V_HIII family genes by the six of the total nine (the three IgG reported plus the six IgM reported by Thomas et al. (18, 26, 27)) specific antibodies induced by insulin, a self Ag, is remarkably similar to the well-documented overutilization of the same family genes by the specific antibodies induced by complex foreign Ag, even very different in nature, including rabies virus (5, 6) and H. influenzae type b capsular polysaccharide (9); and it possibly reflects the overepresentation of V_HIII gene-expressing clonotypes in the human B cell repertoire, as suggested by the analysis of cDNA libraries constructed using circulating B cells from adults (53–55). Thus, the expressed, actually available antibody repertoire in the human may be smaller than predicted and shaped by selection forces, at least in part, different than those paradigmatic of maximal Ag-binding diversity.

The three anti-insulin IgG mAb Vκ segments displayed highly similar sequences. The mAb49 and mAb13 VκIII segments (96% identical in nucleotide sequence) were encoded by the same gene (kv3g) or closely related genes. The mAb48 VκIII segment was encoded by a gene highly similar to but distinct from kv3g. The high degree of similarity of the three insulin-induced specific IgG mAb VκIII sequences could be consistent with the hypothesis that conserved sequences in these VκIII chains provide the structural correlate for insulin binding in these autoantibodies. However, a VκIII sequence (ka3d1) displaying only three amino acid differences from that of the antiinsulin IgG mAb49 VκIII segment and six differences from that of the mAb13 VκIII segment is used by the specific D1 RF IgG mAb (13). Further, a possibly mutated form of the kv328 VκIII segment, used by the anti-insulin IgG mAb48, is used to encode the Vκ chain of RF (43) and of the anti-insulin RBC cold agglutinin mAb 20 (56). Thus, it is unlikely that the VκIII segments provide a specific and critical contribution to insulin-binding in the present IgG autoantibodies. This is consistent with the demonstration by Kabat and Wu (57) that identical V_H or V_L segments can encode antibodies with different specificities.

The insulin-induced IgG mAb V_H segments displayed a number of putative somatic point-mutations, which were formally verified in mAb13 by differentially targeted PCR amplification and cloning and sequencing of the germ-line gene that gave rise to the expressed V_H segment, using genomic DNA from the mAb13-producing cell line or genomic DNA from PMN of the same patient whose B cells were used for the generation of this mAb. The higher and lower numbers of amino acid R mutations in the IgG mAb V_H segment CDR and FR, respectively, than those expected by change only, were consistent with exertion of a positive antigenic pressure to mutate the structure of the CDR and a negative antigenic pressure to mutate that of the FR (58–61). This contention was further strengthened by the high (12.0) and low (0.2) R:S mutation ratios in the IgG V_H segment CDR and FR, respectively, and by the calculated statistically nonsignificant probability that the numbers of R mutations observed in these CDR and FR occurred by chance. Clear traces of Ag selection could be detected also in the mAb48 VκIII segment, but not in those of mAb13 and mAb49. This would be consistent with the primary role played by the Ig V_H segment in Ag-binding (47, 48) and is reminiscent of the minimal traces of Ag-selection found in the Vκ chains of the anti-insulin IgG mAb elicited in the BALB/c mice by injection with heterologous insulin (62). Thus, our findings suggest that, in humans, an affinity maturation process may contribute to the generation of specific IgG autoantibodies to homologous insulin. The cellular elements recruited by insulin into such an affinity maturation process may include the anti-insulin IgG-producing cell precursors, that, as we showed, are present in newly diagnosed IDDM patients before the initiation of the therapeutic treatment (28) and/or anti-insulin natural autoantibody-producing cell precursors, including putatively point-mutated anti-insulin IgM-producing cells (18).

The anti-insulin IgG mAb V_H or VκIII segments did not display obvious structural motifs, perhaps responsible for insulin-binding, as found in other autoantibodies to the same Ag. For instance, in BALB/c mice, heterologous insulin-induced IgG mAb Vκ segments containing clusters of up to six Ser in their CDR1 (see mAb 123 and mAb 125 in Ref. 62). A cluster of five Ser is also present in the CDR2 of the V_H segment of the antiinsulin human IgM mAb 19 (18). No similar Ser clusters were present in the insulin-induced human IgG mAb. In these mAb, the assortment of amino acid changes brought about by the somatic mutation process appears to be heterogeneous (Table II). Most amino acid replacements differed in location and no obvious common pattern in amino acid distribution and/or nature could be discerned. Fourteen of the putative and verified 27 amino acid replacements were nonconserved. Of these, four were basic amino acids, an Arg in the mAb13 V_H segment, an Arg and a His in the mAb48 VκIII segment, and an Arg in the mAb49 VκIII segment (Table II). Basic amino acids, particularly Arg, are known to bind insulin, and this interaction is the rationale for the clinical use of protamin to delay insulin absorption (18).

The H chain CDR3 is a crucial element in the V_H chain-mediated Ag-binding and constitutes an important source of diversity in the expressed antibody repertoire (34–40, 49, 57, 63–65). Its structure is the result of a complex interplay of somatic rearrangement events, often involving, at least in the human, more than one D gene, and germ-line “unencoded” nucleotide additions (34–40, 49). The amino acid sequences and lengths of the insulin-induced specific IgG mAb H chain CDR3 were highly divergent. mAb13 H chain CDR3 displayed two single Arg and mAb48 H chain CDR3 an Arg triplet. Arg residues have been observed in the H chain CDR3 of two human anti-insulin IgM mAb and in those of two mouse anti-insulin IgG mAb (18, 61). Arg-rich H chain CDR3 are also characteristic of anti-DNA autoantibodies and are thought to play a major role in DNA-binding (46, 66). Tandem Pro are present in the H chain CDR3 of two of the reported human IgM mAb (18), and are consistently generated at the Vκ-Jκ recombination junction of antiinsulin murine IgG mAb (62). None of the anti-insulin IgG H chain CDR3 displayed tandem Pro. However, a tandem Pro was generated at the IgG mAb48 Vκ-Jκ recombination site, although a single Pro was present at the site of the mAb13 and mAb49 Vκ segments.

Thus, the anti-insulin IgG mAb H CDR3 and Vκ-Jκ junctional sequences contain elements that are thought to play a role in insulin-binding in both human IgM and mouse IgG anti-insulin autoantibodies. In human anti-insulin IgG mAb, the possible primary role played by the H chain CDR3 in insulin-binding might be complemented by the amino acid substitutions accumulated in the V_H and Vκ segments as a result of the insulin-driven affinity maturation, traces of which have been documented. By increasing the overall avidity of the circulating antiinsulin autoantibodies, such an affinity maturation process may play an important role in the complications associated with insulin therapy. Similar complications, possibly due to a progressively higher avidity of the circulating specific autoantibodies, have been observed during the administration of natural and human rIFN (67, 68). The precise contribution to Ag-binding of individual amino acid changes accumulated in the Ig V_H and Vκ segments and of that of the H chain CDR3, are currently analyzed in our laboratory by site-directed mutagenesis, sequence deletion and shuffling experiments using the insulin-induced IgG mAb and an efficient in vitro human Ig expression system.