Somatic diversity of the immunoglobulin repertoire is controlled in an isotype-specific manner

2.1 IgE immune response in answer to immunization with ph-Ox

We first analyzed the IgG1- and IgE-mediated immune response in WT BALB/c mice. Mice were immunized with the T cell-dependent hapten ph-Ox coupled to ovalbumin 13 over a 49-day period. The immunization protocol is schematically depicted in Fig. 1. IgG1 antibodies were readily found a few days after immunization (Fig. 2a), rose further after a first booster immunization and remained on a high level. Strong support exists in the literature that during this phase-affinity maturation takes place 4–7, 14 and long-lived plasma cells are generated 8. Specific IgE antibodies were only detectable after the first booster immunization, peaked at day 21, and then declined rapidly, until the next encounter with antigen (Fig. 2b). These results clearly show that IgE responses are short-lived. This may be partly due to the very short half-life time of IgE 15, 16 but it also is reflective of a poor recruitment of long-lived plasma cells in the IgE immune response (Fig. 2b, and unpublished observations). The responses therefore reflect the efficiency of the recruitment of short-lived IgE AFC. These AFC have not necessarily undergone extensive somatic mutations 8. IgE responses do show "memory" in the sense of a swift response to secondary and tertiary immunization 8. In mice with a deletion of the 25 C-terminal amino acids of the IgE cytoplasmic tail (IgE^KVKΔtail) 10 specific IgG1 titers were similar to those in WT mice, as expected; however, IgE titers were 70% lower than in WT mice after immunization with ph-Ox. Titers only increased slowly, indicative of a poor recruitment of AFC in the immune response.

2.2 Determination of the relative affinity of ph-Ox-specific IgE antibodies

Next, we assessed the relative affinity of anti-ph-Ox specific IgE antibodies in two ways. Because IgE antibody levels are measured in about a 1,000-fold excess of IgG1 antibodies we first depleted the serum antibodies with protein G-Sepharose for competing excess IgG antibodies (a reversed "classical" affinity measurement) 16. Depletion of the IgG did not markedly increase the titers in WT mice, indicating that IgE antibodies of sufficient affinity that were not influenced by competing IgG1 antibodies, were produced (Fig. 2c). In IgE^KVKΔtail mice specific IgE responses were sensitive to IgG depletion by protein G (Fig. 2c). The latter result implies that the affinity of the IgE antibodies was sufficiently low to be influenced by competing IgG1 antibodies. These results were confirmed by a second approach where the relative affinity 17 of anti-ph-Ox-specific IgE and IgG1 antibodies was measured using "low-haptenated" and "high-haptenated" ELISA plates 17. In WT mice we could not detect differential binding to the differently haptenated ELISA plates. In IgE^KVKΔtail mice, a twofold difference in titer between the highest-(ph-Ox₃₅-BSA) and the lowest haptenated ELISA-plate (ph-Ox₁₁-BSA) was seen at each time point at which specific IgE was detected (Fig. 3b), but was most impressive at day 21. These results and dilution experiments (data not shown) proved that these results were independent of the titer of IgE antibodies. Using the paired t-test, statistical significance at the level of 0.1>p>0.001 was calculated. The absence of the cytoplasmic tail of IgE causes a low-titered response of a poorer quality. The cytoplasmic tail of Ig receptors could thus influence both the efficiency of the somatic mutation process in germinal centers and the efficiency of the recruitment of (high-affinity) B cell clones into the AFC pool.

Therefore, we studied the pattern and extent of somatic diversification in germinal center B cells early and late in the immune response. We compared the occurrence of somatic mutations and the use of DJ elements in the prototype anti-ph-Ox heavy chain variable region, V_HOx1 14, 18, in IgM, IgG1 and IgE antibodies, both in WT and IgE^KVKΔtail mice. A schematic representation of the experimental protocol is given in Fig. 1. Spleen cells of three BALB/c and three IgE^KVKΔtail mice that were immunized at days 0, 14 and 43 13 were isolated on days 7, 14, 21, 43 and 49, and pooled. Germinal center B cells were isolated based on the expression of the surface marker CD45R and the binding of peanut agglutinin (PNA). This procedure excludes plasma cells that do not bind PNA 19. Germinal center B cells were found with an average frequency of 5%. RNA was prepared from 10⁶ cells, followed by cDNA synthesis. For PCR, primers complementary to Cμ, C γ1 or Cϵ constant regions were combined with a specific primer (Fig. 1) for the leader region of V_HOx1 4, 20. A proofreading polymerase was used throughout these experiments to minimize experiment-based mutations.

2.3 Evaluation of somatic diversity of IgM antibodies

To get an impression about the pool of somatic variants used, we first sequenced 36 μ-heavy chain clones from day 7 to day 49 of immunization (Fig. 4) and aligned the sequences with the ph-Ox-specific hybridoma H26 20, expressing the V_HOx1 germ-line sequence. As expected, V_HOx1 was the preferentially amplified V_H prototype. However, in 20–40% of the primary sequences of all isotypes a repeated pattern of substitutions was found. These V_H regions resembled the MC101 sequence 21, 22 (marked with one asterisk) and most likely represent a V_H subset that is closely related to V_HOx1. As observed before, the CDR3 region displayed a high degree of diversity, generated by VDJ rearrangement. The length differed from 7 to 17 amino acids. All four J elements were used, with a preference for J3 7 and J4. Interestingly, during the whole period of immunization not a single identical sequence was found twice (Table 1, Fig. 4). Table 1–4 also summarize the frequency and type of mutations. In these tables, the MC101-like sequences are not represented, because the germ-line sequence of this V_H member has not been established. The number of mutations per sequence did not vary significantly during the course of immunization and replacement mutations were more frequent than silent mutations. In summary, a highly diverse repertoire of V_HOx1-based antibodies accumulated in the IgM pool, with no indication of one preferred type during the course of immunization.

2.4 Molecular analysis of the somatic diversity following an IgG1 response

A rather different situation was found in γ1-heavy chains. We analyzed 81 γ1-specific heavy chain sequences from day 7 up to day 49 of immunization (Fig. 5). In the primary and late (day 43) response, MC101-like sequences as well as the prototype V_HOx1 sequence were observed. The CDR1 region at position 31–35 showed a higher rate of amino acid exchanges than in μ-heavy chains, particularly late in the immune response (e.g. Ser-31, Tyr-32 and Gly-33). In the CDR2 region, fidelity for amino acid substitutions is comparable to μ-heavy chains, with a few new amino acid substitutions (e.g. Ser at position 56). Highly different sequence motifs were detected in the CDR3 region. During the primary immunization period (days 7 and 14) the sequence motifs are very similar to those in the μ-heavy chains. The length of the CDR3 region varies from 7 to 15 amino acids; however, the total number of different DJ elements was lower (Table 2, Fig. 5). The length of the CDR3 region decreased during the secondary response. Clones with a short CDR3 most likely represent high-affinity anti-ph-Ox antibodies 21. Nine different sequence motifs, which varied only by single amino acid substitutions, dominated CDR3 at day 21. Still, no sequence was identical over the entire V_H region, and an immediate interdependence of the sequenced clones is not visible. Late in the secondary response (day 43) a new pool of anti-ph-Ox heavy chains accumulated. The CDR3 region of these resembled the primary set found at days 7 and 14. Also, the MC101-like sequence is found again (marked with one asterisk). At 7 days after the secondary booster immunization, the immune response is as restricted as at day 21, with a short CDR3 region. However, additional sequence motifs occurred, neither identical nor related. The number of mutations and the ratio between replacement and silent mutations was not significantly different from those found in μ-heavy chain sequences (Table 1). Summarizing, the pool of anti-ph-Ox γ1-heavy chains is more restricted than that of μ-heavy chains, in particular as to the length and composition of CDR3 early in the secondary and tertiary response.

2.5 Evaluation of somatic diversity following an IgE response

We next asked, whether the IgE response follows a maturation pattern similar to the IgG1 response. We analyzed 65 specific ϵ-V_H gene segments (Fig. 6). During the whole course of immunization the IgE response was rather restricted. Again, MC101-like sequences were found early in the primary and late secondary response. Late in the primary (day 14), early in the secondary (day 21) and in the tertiary response (day 49) a very restricted repertoire was found, with a clear relationship between the sequences. With "a clear relationship" we mean their similarity, in the sense of for instance CDR3 composition, but without a direct "hereditary" relationship. This may reflect the low titer found in the IgE response, with a low number of IgE-specific germinal centers. The reduction in length of the CDR3 region was less pronounced than for γ1 and a few dominant motifs characterized CDR3. Diversity returned in the sequences obtained from day 43. At day 49 new CDR3 motifs showed up. None of these variants were found in the sequence motifs at earlier time points. The ratio of replacement to silent mutations was not significantly changed (Table 3).

Summarizing, the IgE response is more restricted than the IgG1 response. The length of the CDR3 region is longer than that found in IgG1 antibodies, suggesting a lower affinity. It differs strongly from the sequence motifs found in the CDR3 regions of IgG1. However, there is no doubt that the IgE response matures. A set of antigen receptors with a distinct accumulation of mutations is selected and preferentially expanded.

2.6 Comparison of somatic diversity of IgE antibodies in IgE^KVKΔtail mice with those of IgM antibodies in WT mice

We finally asked whether the poor affinity maturation seen in IgE^KVKΔtail mice was due to an insufficient somatic mutation process or a poor recruitment of high-affinity B cells. Analysis of IgM-and IgG1-derived V_HOx1 sequences in IgE^KVKΔtail mice revealed a pattern as found in WT mice (data not shown). Fifty-seven IgE-derived sequences were analyzed (Fig. 7). Again, in the primary and late secondary response MC101-like sequences were found. The primary response was rather different from that in WT mice, particular in the CDR3 region. Only late in the primary response (day 14) did one clone type resemble a dominant clone type seen in WT mice. The secondary response is characterized by a CDR3 pattern that is strongly represented in γ1 antibodies. As for the other isotypes, the response was more diverse at day 43, with longer CDR3 regions. A most significant difference to WT BALB/c mice was detected at day 49. Sequences were more diverse than in WT mice, with long CDR3 regions, resembling the pattern seen in μ-heavy chains. The ratio of replacement to silent mutations was not significantly changed (Table 4). These results indicate that the mechanism of somatic mutation in IgE antibodies is not interfered with in IgE^KVKΔtail mice, but that the selection of B cells with high-affinity receptors is impaired.

2.7 Expression of anti-ph-Ox-specific IgE Fab fragments on the surface of phages

Our premise that a short CDR3 region is indicative of a higher affinity anti-ph-Ox antibody is based on the work of Griffiths et al. 21. To support this interpretation, we expressed anti-ph-Ox-specific IgE Fab fragments on the surface of M13 phages 23 and measured their affinity for ph-Ox with surface plasmon resonance analysis. We first cloned two different ph-Ox-specific light chains, using primer combinations as described by Berek et al. 4. Light chain variant I comprised V_KOx1 and the J_K5 element. The histidine at position 34 identifies this light chain as a low-affinity anti-ph-Ox variant 4. This light chain was cloned into phagemid pComb3 23, resulting in plasmid pEL3. Light chain variant II showed an exchange of histidine to asparagine, resulting in a high-affinity anti-ph-Ox light chain variant. Cloning of this variant into pComb3 resulted in pEL6. Two V_H regions were chosen: WT IgE d7–1 and WT IgE d22–170 as representatives of a long and a short CDR3 region, respectively. They were then re-amplified and subcloned in pComb3, generating plasmids pEL15 and pEL16, respectively, in pEL3, generating plasmids pEL8 and pEL12, respectively, and in pEL6, generating plasmids pEL10 and pEL14. The four plasmids that functionally expressed anti-ph-Ox-specific Fab fragments (pEL8 and pEL12 as well as pEL 10 and pEL14) and the four control plasmids (pEL3, pEL6, pEL15 and pEL16) were analyzed for their binding affinity using surface plasmon resonance technology. ph-Ox OVA and OVA as reference were immobilized to an appropriate sensorchip. Fab fragments, expressed on the surface of M13 phage, were directly injected onto the sensorchip. The relative response units, expressed as differences (Fc1–Fc2) of ph-Ox OVA (Fc1) and OVA (Fc2) are proportional to binding affinities and are summarized in Table 5. Injection of pEL3 and pEL6, which express only the (low- and high-affinity) light chain does not show binding to ph-Ox. Similarly, injection of the control plasmids pEL15, which expresses the heavy chain segment WT IgE d7–1, and pEL16 which expresses the heavy chain segment WT IgE d22–170 do not show a measurable response. Injection of pEL8, which expresses both the light chain and the WT IgE d7–1 heavy chain segment representing a long CDR3, shows a marginal response. In contrast, injection of pEL12, which expresses the light chain segment together with the WT IgE d22–170 (pEL12) heavy chain segment representing a short CDR3, results in a remarkable increase in binding affinity. Measurements of pEL10 and pEL14 (heavy chains as described before in combination with a high-affinity light chain variant) showed an additional increase in binding affinities if compared to pEL8 and pEL12. This result fully supports our evaluation of the sequence motifs presented in Fig. 4–7, where we interpret short CDR3 regions as representative of high-affinity and long CDR3 regions as representative of low-affinity antibody variants.

4.1 Immunizations

Mice were immunized subcutaneously (s.c.) with 20 μg ph-Ox₁₁-OVA, precipitated in alum with Bordetella pertussis as adjuvant. At days 14 and 43, mice received booster immunizations with 20 μg ph-Ox₁₁-OVA s.c. and 20 μg ph-Ox₁₁-OVA intraperitoneally (i.p.). ph-Ox was conjugated to OVA and BSA according to Mäkelä et al. 13.

4.2 ELISA assays

Serum was taken at different times (days 0, 7, 14, 21, 28, 43 and 49) after immunization. Sera were depleted for IgG as described by Yu et al. 10, 16. For measuring relative affinities of ph-Ox-specific IgE and IgG1, ELISA plates were coated with BSA coupled to varying densities of the hapten ph-Ox (ph-Ox₁₁-, ph-Ox₂₇- or ph-Ox₃₅)-BSA. Specific IgG1 and IgE antibodies to ph-Ox were detected with alkaline phosphatase-conjugated isotype-specific goat anti-mouse antibodies (Southern Biotechnology Associates) or rat anti-mouse IgE monoclonal antibody (EM-95–3). Titers are expressed as the reciprocal serial dilution, at which a half-maximal absorbance value was measured.

Because of the low titer of specific IgE in the serum, sera of three mice per data point were pooled. Measurements of individual mice were made to validate the use of pooled sera. The measurements in individual sera were in full agreement with those obtained in pooled sera.

4.3 Flow cytometry

Mouse spleens were homogenized and erythrocytes were lysed isotonically. Subsequently, the obtained single-cell suspension was prepared for fluorescence activated cell sorting using a FACS-Vantage flow cytometer (Becton Dickinson). Phycoerythrin-labeled antibody to CD45R (clone 6B₂), a B cell marker and the lectin PNA (peanut agglutinin) coupled with FITC were used to stain the germinal center B cells 4. Double-positive cells were sorted. For further RNA extraction, 10⁶ double-positive B cells were used.

4.4 RNA extraction

B cells (10⁶) were homogenized with QIAshredder according to manufacturer's instructions. The cell lysate was used for further total RNA extraction using RNeasy (Qiagen). About 20 μ total RNA per 10⁶ B cells was isolated; 3 μg total RNA was used for further cDNA construction.

4.5 Reverse transcription-PCR

cDNA was synthesized using oligo(dT) priming. After cDNA synthesis, antigen specific μ-, γ1- and ϵ-heavy chain variable regions were amplified by nested PCR 30. The primers bind specifically to the leader sequence of the V_HOx1 heavy chain variable region and to the constant region (CH1 or CH2 or CH4) of μ-, γ1- and ϵ-heavy chains, respectively. PCR products were subcloned into the pCR-Script cloning vector (Stratagene) and sequenced. Cloning of an anti-ph-Ox-specific light chain was achieved with a semi-nested approach: nested 5′ primers (leader V_LOXA and leader V_LOXB), complementary to the leader region of the ph-Ox-specific V element V_KOx1 were used; for cloning purposes, V_LOXB contains an internal Sac 1 restriction site. As 3prime; primer Lκ-reverse2, complementary to the 3prime; end of the constant region of the κ-locus with an internal Xba 1 restriction site was used. PCR products were subcloned into the pCR-Script cloning vector (Stratagene) and sequenced. For all amplifications a proof-reading DNA polymerase was used (Pfu polymerase, Stratagene).

Nested PCR conditions were: 2 min at 95°C (primary denaturation); 45 s at 95°C; 45 s at 59°C; 1 min at 72°C; (35 cycles); 3 min 72°C (final extension).

4.6 Primers

The following primers were used: oligo(dT): 5′-GAGAGAGAGAGAGAGAGAGAACTAGTCTCGAGTTTTTTTTTTTTTTTTT-3prime;; leader V_H OXA: 5′-TCCTCTCTAGAGCCCCCATCAGAGCATGGC-3prime;; leader V_H OXB: 5′-GCCTGGTTGAATTCCAAGCTGTGTC-3prime;; CH4ϵ-reverse: 5′-GATATTGTTTTCTCCAGTTTC-3prime;; CH2ϵ-reverse: 5′-CAGTTTGTGCAAGTGTATCAG-3prime;; CM-reverse: 5′-CAGGGGGCTCCTGCAGGAGACGAGG-3prime;; CG1-reverse: 5′-CAGCAGCTGCAGGGGCCAGTGGATAG-3prime;; CG2-reverse: 5′-GGGAAATAGCTCTAGACCAGGC-3prime;; leader V_LOXA: 5′-CCTGCTAATCAGTGCCTCAGTCATAATA-3prime;; leader V_LOXB: 5′-TCCAGAGGAGAGCTCCAAATTGTTCTC-3prime;; Lκ-reverse2: 5′-ACCTTTGTCTTCTAGACTAACACTCATTC-3prime;.

4.7 pComb3 cloning

The anti-ph-Ox-specific light chain was subcloned in pComb3 23 by digesting the PCR product with Sac 1 and Xba 1 followed by direct insertion into the pComb3 plasmid. Anti-ph-Ox specific heavy chains were re-amplified with a 5′ primer (leaderV_H OXC), with an internal Xho 1 restriction site and a 3prime; primer (CH1Comb) with an internal Spe 1 restriction site. Internal restriction sites were designed to guarantee in frame fusion of the light chain with the pelB-leader, and the heavy chain with the pelB-leader and the gIII phage surface protein, respectively;leader V_H OXC: 5′-GTGTCCTGTCCCTCGAGCAGGTGCAGC-3prime;; CH1Comb: 5′-GACAGGTCGAACACTAGTTAGGATAGTCC-3prime;. PCR-con-ditions were as described before.

4.8 Generation of phages

pComb3 plasmids, containing in frame fusions with anti-ph-Ox-specific light and heavy chains were transformed to E. coli strain XL1-blue. E. coli were grown to an optical density of OD₆₀₀≥0.5. Helper phage VCS-M13 (10¹¹/ml) was added. After incubation (1 h at 37°C) kanamycin was added (final concentration 35 μg/ml) and incubated overnight at 30°C. After centrifugation (20 min, 4,000×g), phage were precipitated with 4% PEG 8000 and 3% NaCl for 2 h on ice. After centrifugation (20 min, 11,000×g) the phage pellet was dissolved in 1 ml TBS (10 mM Tris-HCl pH 7.5, 150 mM NaCl). Titer was determined by diluting the phage from 10^–9 to 10^–11 and re-infection of 100 μl E. coli XL1-blue (OD₆₀₀>0.5) (10 min, room temperature). Infected XL1-blue were plated on LB-amp plates. The generated phage were diluted to a titer of 10¹¹ plaque-forming units. The phages were directly used for surface plasmon resonance analysis.

4.9 Surface plasmon resonance analysis

Using Biacore-J, the sensorchip CM5 (Biacore) was activated and coupled with ph-Ox-OVA and OVA as reference, respectively, according to the manufacturers specifications. 9,000 resonance units (RU) were coupled for each. Analysis was performed on the Biacore J. Phage (10¹¹) were injected for 3 min. Phages were eluted for 2 min. Binding curves were calculated using the BIA-viewer software program. Relative response units after the elution step are indicative for binding affinities.