Induction of Ig Somatic Hypermutation and Class Switching in a Human Monoclonal IgM⁺ IgD⁺ B Cell Line In Vitro: Definition of the Requirements and Modalities of Hypermutation

Abstract

Partly because of the lack of a suitable in vitro model, the trigger(s) and the mechanism(s) of somatic hypermutation in Ig genes are largely unknown. We have analyzed the hypermutation potential of human CL-01 lymphocytes, our monoclonal model of germinal center B cell differentiation. These cells are surface IgM⁺ IgD⁺ and, in the absence of T cells, switch to IgG, IgA, and IgE in response to CD40:CD40 ligand engagement and exposure to appropriate cytokines. We show here that, CL-01 cells can be induced to effectively mutate the expressed V_HDJ_H-Cμ, V_HDJ_H-Cδ, V_HDJ_H-Cγ, V_HDJ_H-Cα, V_HDJ_H-Cε and VλJλ-Cλ transcripts before and after Ig class switching in a stepwise fashion. In these cells, induction of somatic mutations required cross-linking of the surface receptor for Ag and T cell contact through CD40:CD40 ligand and CD80: CD28 coengagement. The induced mutations showed intrinsic features of Ig V(D)J hypermutation in that they comprised 110 base substitutions (97 in the heavy chain and 13 in the λ-chain) and only 2 deletions and targeted V(D)J, virtually sparing C_H and Cλ. These mutations were more abundant in secondary V_HDJ_H-Cγ than primary V_HDJ_H-Cμ transcripts and in V(D)J-C than VλJλ-Cλ transcripts. These mutations were also associated with coding DNA strand polarity and showed an overall rate of 2.42 × 10⁻⁴ base changes/cell division in V_HDJ_H-C_H transcripts. Transitions were favored over transversions, and G nucleotides were preferentially targeted, mainly in the context of AG dinucleotides. Thus, in CL-01 cells, Ig somatic hypermutation is readily inducible by stimuli different from those required for class switching and displays discrete base substitution modalities.

The process of V(D)J gene somatic hypermutation diversifies Abs, thereby providing the structural basis for selection by Ag of higher-affinity mutants and the maturation of the immune response. This process occurs within the germinal center (GC),³ where it requires T cell help and engagement of the surface B cell receptor for Ag (BCR) and remains one of the most intriguing features of the T cell-dependent immune response (1, 2). Somatic Ig V(D)J gene hypermutation is thought to be operative at the centroblastic stage (3). At the centrocytic stage, B clones with a BCR with high affinity for the inducing Ag would undergo Ag-driven positive selection, while autoreactive B cells or low-affinity clones undergo negative selection through apoptosis (4). In vivo and in vitro studies have suggested that Ig hypermutation displays: 1) a prevalence of point-mutations together with occasional insertions and deletions (5, 6); 2) an intrinsic preference for certain “hotspots” (7); 3) a dependence on initiation of transcription, A > T bias, and DNA strand polarity (2, 5, 8, 9); 4) a dependence on cis-acting elements, including the intronic and 3′ enhancers in the κ locus (10, 11); and, finally, 5) a preference for secondary Ig isotypes (12). However, the lack of a well-defined in vitro model of GC differentiation has limited our understanding of the requirements for the induction, the modalities, and the mechanisms of hypermutation.

CD40:CD40 ligand (CD40L) engagement in association or not with BCR cross-linking in the presence of various cytokines has led to the induction of proliferation and isotype switching, but not somatic hypermutation (13–16). Consistent with the primary role of T cells in GC formation in vivo (1, 17, 18), Ig somatic mutations have been induced in vitro in mouse and human B cells in the presence of T cell help and upon BCR engagement (19–21). This, together with the finding that certain monoclonal B cell lines, such as the murine 18.81 cells (22) and a human follicular lymphoma line (23), mutate spontaneously in vitro in the absence of specific triggers, provided impetus for the identification of the Burkitt’s lymphoma cell line BL2, which was found to accumulate somatic mutations in the expressed IgM upon BCR cross-linking and co-culture with activated T cells (24). However, these cells appear to be frozen at the surface (s)IgM⁺ sIgD⁻ phenotype and are incapable of switching to downstream Ig isotypes and undergoing concomitant phenotypic differentiation. A cell line that enables analysis of the requirements and the modalities of somatic hypermutation as it relates to Ig class switching and other GC differentiative processes would constitute a more appropriate and useful model of physiological Ig hypermutation.

We have analyzed the Ig somatic hypermutation potential of our recently identified monoclonal model of GC B lymphocyte differentiation, human CL-01 cells. These B cells express a founder centroblast-like phenotype, including sIgM, sIgD, CD38, and CD77 (15, 16, 25, 26). Following engagement of CD40 by CD40L and exposure to the appropriate cytokines, they undergo a coordinated maturation program that includes Ig class switching to all seven downstream isotypes, progression through phenotypic GC stages, and differentiation to memory-like B cells and plasma cells. We show here that CL-01 cells can be induced to not only switch to IgG, IgA, and IgE, but also to effectively mutate the V_HDJ_H and V_λJ_λ gene segments, while sparing the C_H and Cλ regions. CD40 and CD80 coengagement by T cell CD40L and CD28 is necessary, in addition to BCR engagement, for the induction of Ig hypermutation. Mutations accumulated in a stepwise fashion before and after class switching and were distributed throughout the entire V(D)J gene segment, indicating a lack of selection by Ag. These mutations showed preference for transitions over transversions, biased targeting of G within the AG dinucleotide, and evidence of strand polarity.

Materials and Methods

CL-01 cells

The human B cell line CL-01 has been described (15). CL-01 cells are sIgM⁺ sIgD⁺ and monoclonal, as shown by blotting and Southern hybridization with labeled J_H probes and by the expression of unique V_HDJ_H-Cμ and V_HDJ_H-Cδ transcripts (27). These cells display on both chromosomes the switch (S)μ, σδ, and Sγ3, Sγ1, Sα1, Sγ2, Sγ4, Sε, and Sα2 regions in germline configuration (15). These cells also express the phenotype of GC founder centroblasts, including CD38 and CD77. Upon engagement of CD40 by CD40L and exposure to appropriate cytokines, CL-01 cells undergo a coordinated program of GC differentiation involving characteristic phenotypic changes and switching to all downstream isotypes, i.e., IgG3, IgG1, IgA1, IgG2, IgG4, IgA2, and IgE, eventually giving rise to plasmacytoid elements and memory-like B cells (15, 16, 25). sIgM⁺ sIgD⁺ CL-01 cells were cultured in RPMI 1640 medium (Life Technologies, Gaithersburg, MD) supplemented with 10% heat-inactivated FCS, 2 mM L-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin (FCS-RPMI) at <10⁵/ml.

T cells

CD4⁺ T cells were positively selected from normal PBMCs, prepared by fractionation through Histopaque 1077 (Sigma, St. Louis, MO) using anti-CD4 mAb-conjugated magnetic beads (Miltenyi Biotech, Auburn, CA). They were cultured in FCS-RPMI and expanded by weekly stimulation with a feeder cell mixture containing irradiated (1200 rad) PBMCs, 100 μg/ml of PHA (Life Technologies), and 100 U/ml of human rIL-2 (Genzyme, Cambridge, MA). For T:B cell coculture experiments, CD4⁺ T cells were used at least 2 wk after their last activation with feeder cells and were incubated for 6 h with 20 ng/ml of PMA (Sigma) and 500 ng/ml of ionomycin (Calbiochem-Novabiochem, San Diego, CA) before culture with B cells.

B:T cell cultures

CL-01 cells were cultured at 0.5 × 10⁶ cells/well in the presence or absence of 2.5 × 10⁶ irradiated (4000 rad) CD4⁺ T cells or 1 × 10⁶ irradiated (4000 rad) human CD40L-transfected 293 cells (CD40L-293 cells) in a flat-bottom 6-well plate (5.0 ml FCS-RPMI volume). For B:T cell cocultures, plates were previously coated with 1:800 OKT3 mAb (Ortho Diagnostics Systems, Raritan, NJ) for 2 h at room temperature. To cross-link the BCR, CL-01 cells were reacted for 2 h at 4°C with Sepharose-conjugated rabbit Ab to human Ig light chain (2 μg/ml; Irvine Scientific, Santa Ana, CA) and then washed with cold PBS. After 7 days of culture, CL-01 cells were collected, freed of dead cells and debris by fractionation through Histopaque 1077 (Sigma), reacted again with anti-BCR Ab, washed, and reseeded over a new layer of irradiated T cells or CD40L-293 cells (in plates coated or not coated with anti-OKT3 mAb) in the presence or absence of cytokines. At day 14 of culture, CL-01 cells were harvested for total RNA extraction. In selected B:T cell cocultures, T cell conditioned medium (TCM) obtained from the culture fluids of T cells activated for 1 wk with 1:800 OKT3 mAb and IL-2 (100 U/ml) was used at a concentration of 1:5; IL-4 (Genzyme) and IL-10 (Schering-Plough, Kenilworth, NJ) were used at 100 U/ml and 100 ng/ml, respectively. The CD40:CD40L, CD80:CD80L, and CD30:CD30L interactions were blocked by preincubating T or B cells with saturating amounts (30 μg/ml) of mouse 24–31 mAb to human CD40L (Ancell, Bayport, MN), mouse CD28.2 mAb to human CD28 (PharMingen, San Diego, CA), mouse Ber-H2 mAb to human CD30 (Dako, Carpinteria, CA), or mouse BB1/B7-1 mAb to human CD80 (PharMingen).

PCR amplification of V(D)J transcripts

RNA was extracted from 2 × 10⁶ cells using the RNeasy total RNA kit (Qiagen, Cathsworth, CA). mRNA was reverse transcribed using the SuperScript preamplification system for first strand cDNA synthesis (Life Technologies). V_H3DJ_H-C_H transcript cDNAs were amplified with a V_H3 leader-specific sense primer LV_H3(I) (5′-ATGGAG(CT)TTGGGCTGA(CG)CTGG(CG)TTT(CT)T-3′) (27) and with the antisense-specific primers C_H1-μ (5′-GTTGCCGTTGGGGTGCTGGAC-3′) (spanning Cμ nucleotides 268–288), Cδ (5′-TCCAGCAGTGGCGCCAAGGCGAG-3′) (Cδ, 220–242), universal primer C_H1-γ (5′-CAAGCTGCTGGAGGGCACGGT-3′) (Cγ, 206–226), C_H1-α (5′-CTAGGCACTGTGTGCCGGCAGGGT-3′) (Cα, 209–232), or C_H1-ε (5′-CGAGACGGTCAGCAAGCTGATGG-3′) (Cε, 201–223), using Cloned Pfu DNA polymerase (Stratagene, La Jolla, CA) and the reaction buffer provided by the manufacturers with 30 cycles, each consisting of a 1-min denaturation at 94°C, a 1-min annealing at 58°C, and a 1-min extension at 72°C. The amplification was completed by an additional 10-min extension at 72°C. To amplify Vλ1-Jλ-Cλ transcripts, the Vλ1 leader-specific sense primer LVλ1(I) (5′-ATG(GA)CC(TG)GCT(CT)CCCTCTCCTCCT-3′) (λ-chain leader, 1–23) and the Cλ-specific antisense primer Cλ(II) (5′-CGTCAGGCTCA GATAGCTGCTG-3′) (Cλ, 202–223) were used with the PCR conditions described above. The PCR cDNA products were purified with the PCR purification kit (Qiagen) and ligated into pCR-Script SK vector (pCR-Script Cloning Kit, Stratagene, La Jolla, CA), which was used to transfect in XL1-Blue MRF supercompetent cells (Stratagene). The (positive) white bacterial colonies were screened by PCR (28) using the V_H3 leader sense primer LV_H3 (II) (5′-GTTGCTATTTTAAAAGGTGTCCAGTGT-3′) (CL-01 heavy chain leader, 21–57) and the consensus J_H antisense primer (5′-CGGTCACCGTCTCCTCA-3′) for V_HDJ_H-C_H clones and the internal Vλ1 leader sense primer LVλ1 (II) (5′-ACCCCTCCTCACTCACTGTG CAG-3′) (CL-01 λ-chain leader, 23–45) together with the internal Cλ antisense primer Cλ (I) (5′-TTGGCTTGAAGCTCCTCAGAGGA-3′) (Cλ, 42–64) for VλJλ-Cλ clones. The individual colonies that had been directly used as PCR templates were seeded onto fresh Luria-Bertani medium plates and expanded overnight. The clones containing the V_HDJ_H-C_H or VλJλ-Cλ transcripts were selected for single-stranded conformational polymorphism (SSCP) analysis.

Detection of mutated V_HDJ_H-C_H and VλJλ-Cλ transcripts by SSCP

Mutated V_HDJ_H-C_H and VλJλ-Cλ transcripts were identified by SSCP analysis (29). cDNAs for SSCP analysis were amplified by PCR (30 cycles of a 1-min denaturation at 94°C, a 1-min annealing at 58°C, and a 1-min extension at 72°C) using the cloned cDNA inserted into pCR-Script SK vector as template in a 10-μl reaction volume with Taq DNA polymerase (Life Technologies) in the presence of 1 μCi [α-³²P]dCTP (NEN Life Sciences, Boston, MA) (3000 Ci/mmol). The internal V_H3 leader sense primer LV_H3(II) and C_H1 antisense primer Cμ (5′-AGACGAGGGGGAAAAGGGTT-3′) (Cμ, 18–37), Cδ (5′-TGGGGAACACATCCG GAGCCTTG-3′) (Cδ, 8–30), universal Cγ (5′-GAAGACCGATGGGC CCTTGGTGGA-3′) (Cγ, 4–27), Cα (5′-GACCTTGGGGCTGGTCGG GGAT3′) (Cα, 3–24), or Cε (5′-CGGAGGTGGCATTGGAGG-3′) (Cε, 54–71) were separately used for V_HDJ_H-Cμ, V_HDJ_H-Cδ, V_HDJ_H-Cγ, V_HDJ_H-Cα, or V_HDJ_H-Cε analysis. To analyze the λ-chain, the Vλ1 leader sense primer LVλ1(II) and the Cλ antisense primer Cλ (I) were used. Optimal sensitivity is achieved when SSCP is performed using 100- to 350-bp DNA fragments. Therefore, our amplified >400-bp fragments were digested with KpnI and then diluted 1:15 in 10 mM EDTA, 0.1% SDS. The labeled, cleaved, and diluted DNAs were mixed with an equal volume of sequencing stop solution containing 95% formamide, 20 mM NaOH, 20 mM EDTA, 0.05% bromphenol blue, and 0.05% Xylene cyanol. Samples were denatured for 10 min at 98°C, chilled on ice, and immediately loaded in 3-μl aliquots onto a 6% acrylamide gel (20:1 acrylamide:bis) with 1× Tris/boric acid/EDTA buffer (100 mM Tris, 90 mM boric acid, and 1 mM EDTA) containing 10% glycerol. Electrophoresis was at room temperature for 18 h at 6 W. Gels were autoradiographed on Kodak X-Omat AR film (Eastman Kodak, Rochester, NY).

Sequencing Ig V(D)J-C transcripts

The clones displaying an altered electrophoretic mobility in SSCP gel were analyzed by sequenced to confirm and characterize the nature of the mutations. Plasmids were extracted using QIAprep Spin Miniprep kit (Qiagen) and sequenced on both strands using Taq dideoxy terminator cycle sequencing kit and a 373 automatic sequencer (Applied Biosystems, Foster City, CA). Sequences were compared with the unmutated CL-01 V_HDJ_HC_H1 and VλJλ-Cλ sequence from CL-01 cells cultured in medium only using the MacVector v.5.0 software (International Biotechnologies, New Haven, CT). Sequencing of a total of 20 heavy chain and 10 λ-chain transcripts defined as negative by SSCP analysis revealed a lack of somatic mutations in all of them (not shown). Sequencing of a total of 15 heavy chain and 15 λ-chain transcripts defined as positive by SSCP analysis yielded 100% concordance with the SSCP analysis results and demonstrated that a single nucleotide change in the 375 bp of the V_HDJ_H or the 336 bp of the VλJλ cDNA sequence was sufficient to alter DNA mobility in the SSCP gel (not shown).

Mutational analysis

The census of the somatic point-mutations was performed by counting identical mutations in more than one transcript (of any isotype) only once. It was assumed that identical base changes in different transcripts were shared mutations, although some of these mutations might have arisen as hotspots and therefore could actually be independent mutations. The obtained values were used for the compilation of Tables I (last column), II, III, IV, and V. For the compilation of the fourth column of Table I, all point-mutations found in all sequenced transcripts were counted. For the fifth column of Table I, the somatic point-mutations found in different transcripts of the same isotype were counted only once.

Table I.

In vitro somatic point-mutations in Ig V_HDJ_H and VλJλ gene sequences expressed by induced CL-01 cells

VHDJH	Mutated/Total Transcripts	Mutated Transcripts Sequenced	Mutations in All Sequenced Transcripts	Unique Mutations in Transcripts of Different Isotypesa	Unique Mutations in Transcripts of All Isotypesb
μ	16/112	7	15	14	14
δ	20/110	14	30	29	28
γ	17/120	12	59	49	43
α	9/86	4	10	9	6
ε	3/24	1	7	7	5
Total VHDJH	65/452	38	121	108	96
VλJλ	18/179	9	16	13	13

^aIdentical mutations in different transcripts of the same isotype were assumed not to be independent and were counted only once.

^bIdentical mutations in different transcripts of the same and different isotypes were assumed not to be independent and were counted only once.

Comparison of the observed with the expected frequency of replacement (R) and silent (S) somatic point-mutations was performed using the inherent mutation rate of the CL-01 V_HDJ_H and VλJλ sequences, calculated by the Inh. Sus. Calc. program version 1.0 for the Macintosh (30) and a binomial distribution model, as reported by Chang and Casali (30). The comparisons of the observed with the expected number of mutations for each individual nucleotide residue to each of the three other nucleotides were performed using a contingency table (χ² test). The expected frequency of mutations was calculated by taking into account the base composition of the unmutated CL-01 V(D)J sequence; that is, it was corrected by the frequency of occurrence of the individual nucleotides, or di-, tri-, or tetranucleotides considered within the CL-01 V(D)J sequence assuming randomness.

Results

CL-01 cells switch to IgG, IgA, and IgE, and hypermutate the V(D)J gene in both primary and secondary isotypes upon BCR engagement and exposure to T cells

Human CL-01 cells are monoclonal and express sIgM and sIgD with a λ-chain. The V_HDJ_H segment consists of a V_H26–3.7-like gene (90% identity) (31), rearranged to DN4 and D5 genes, and an unmutated J_H6c gene. The VλJλ segment consists of a 1c.10.2/DPL2 Vλ1-like gene (94% identity) (32), rearranged to a Jλ2 gene with a single change, probably a point-mutation. Multiple sequences at different times of V_HDJ_H-Cμ, V_HDJ_H-Cδ, and VλJλ-Cλ cDNAs from cultured CL-01 cells revealed no somatic mutations (not shown). To determine whether CL-01 cells could be induced to hypermutate the expressed Ig V(D)J genes, these cells were reacted with an anti-light chain Ab to cross-link the BCR and then cultured with activated human T cells. After 14 days, mRNA was extracted from the cultured cells and reverse transcribed. Ig V(D)J-C cDNAs were amplified using the CL-01 V_H or Vλ gene leader sense primer in combination with an antisense Cμ, Cδ, Cγ, Cα, Cε, or Cλ primer and were cloned into appropriate vectors for further “nested” PCR amplification (in the presence of [³²P]dCTP) of V(D)J-C cDNAs. Amplified [³²P]-labeled V(D)J-C cDNAs were digested with KpnI for SSCP analysis. Those V(D)J-C cDNAs defined as mutated by SSCP were then sequenced (Fig. 1A).

FIGURE 1.

Induction of somatic hypermutation and Ig class switching in human IgM⁺ IgD⁺ CL-01 cells. A, Strategy used to analyze somatic mutations in induced CL-01 cell Ig V(D)J genes. B, BCR engagement and CD40:CD40L and CD80:CD28 coengagements are crucial for the induction of hypermutation. Bars indicate percent of V(D)J gene transcripts mutated as determined by SSCP in CL-01 cells cultured for 14 days with medium only; CD40L-293 cells, TCM, IL-4 (100 U/ml), and IL-10 (100 ng/ml) after BCR engagement; activated T cells alone; activated T cells after BCR engagement; activated T cells and anti-CD28 mAb after BCR engagement; activated T cells; anti-CD40L mAb after BCR engagement; and activated T cells and anti-CD30 mAb after BCR engagement. The number of transcripts analyzed from CL-01 cells cultured activated T cells upon BCR engagement are shown in the Table I. The numbers of transcripts analyzed from all other culture conditions are given in Results.

As expected from the detection of abundant IgG, IgA, and IgE in the culture fluids (not shown), V_HDJ_H-Cγ, V_HDJ_H-Cγ-Cα, and V_HDJ_H-Cγ-Cε cDNAs were readily amplified, in addition to V_HDJ_H-Cμ and V_HDJ_H-Cδ cDNAs, from CL-01 cultured with T cells after BCR engagement. By SSCP analysis, 16 of 112 (14.3%) V_HDJ_H-Cμ transcripts were mutated, i.e., they displayed a gel mobility different from that of corresponding transcripts from CL-01 cultured in medium alone, based on the analysis of 50 identical V_HDJ_H-Cμ and 50 identical VλJλ-Cλ transcripts (Table I and Fig. 1B). Also mutated were 20 of 110 (18.2%) V_HDJ_H-Cδ, 17 of 120 (14.2%) V_HDJ_H-Cγ, 9 of 86 (10.5%) V_HDJ_H-Cα, 3 of 24 (12.5%) V_HDJ_H-Cε transcripts, and 18 of 179 (10.1%) V_λJ_λ-C_λ transcripts (Table I and Fig. 1B). In contrast, no mutated transcripts were found in CL-01 cells cultured with anti-BCR Ab in the absence of activated T cells (50 V_HDJ_H-Cμ, 50 V_HDJ_H-Cδ, and 50 V_HDJ_H-Cγ transcripts analyzed) or cultured with activated T cells without BCR engagement (50 V_HDJ_H-Cμ, 50 V_HDJ_H-Cδ, and 50 V_HDJ_H-Cγ transcripts analyzed). Thus, after BCR engagement and exposure to activated T cells, CL-01 cells effectively Ig class switch and mutate both primary (V_HDJ_H-Cμ and V_HDJ_H-Cδ) and secondary (V_HDJ_H-Cγ, V_HDJ_H-Cα, and V_HDJ_H-Cε) heavy chain transcripts, as well as VλJλ-Cλ transcripts.

CD40:CD40L and CD80:CD28 coengagement is required for the induction of Ig somatic hypermutation

As T cell-derived soluble factors have been implicated in sustaining Ig V gene somatic hypermutation (21), we assessed the ability of these factors to substitute T cells in the induction of Ig mutation in CL-01 cells. No mutated primary or secondary Ig transcripts were detected in CL-01 cells cultured after BCR engagement with TCM further enriched with IL-4 and IL-10 (not shown), even upon CD40 engagement by CD40L-293 cells (50 V_HDJ_H-Cμ, 50 V_HDJ_H-Cδ, and 50 V_HDJ_H-Cγ transcripts analyzed) (Fig. 1B). To further explore the role of CD40 in the induction of hypermutation, CD40L on the activated T cells was blocked using a mouse mAb to human CD40L. The failure of the CL-01 cells cultured with this anti-CD40L mAb, but not the putatively irrelevant anti-CD30 mAb, to hypermutate the expressed Ig even upon BCR engagement (60 V_HDJ_H-Cμ and 40 VλJλ-Cλ transcripts analyzed from anti-CD40L Ab cultures and 50 V_HDJ_H-Cγ and 52 VλJλ-Cλ transcripts from anti-CD30 Ab cultures) indicates that CD40 engagement by CD40L, not mere T:B cell contact, is necessary to induce the mutational machinery in CL-01 cells (Fig. 1B).

Because the CD28 activation pathway has been reported to be essential for the Ab response to T cell-dependent Ags in the mouse (33), we addressed the role of this costimulatory molecule in the induction of somatic hypermutation. Blocking of CD28 on the surface of activated T cells using an anti-CD28 Ab made these T cells ineffective inducers of hypermutation (50 V_HDJ_H-Cμ, 50 V_HDJ_HCδ, and 50 V_HDJ_H-Cγ transcripts analyzed) when added to CL-01 cells after BCR engagement (Fig. 1B). That CD28 was required to engage CD80 (the CD28 complement on the B cell surface) to induce Ig hypermutation was further indicated by the lack of mutations (50 V_HDJ_H-Cμ, 50 V_HDJ_H-Cδ, and 50 V_HDJ_H-Cγ transcripts analyzed) in CL-01 cells cultured with activated T cells in the presence of a blocking mouse anti-CD80 mAb (Fig. 1B). Thus, in addition to BCR engagement, engagement of the CD40:CD40L and CD80:CD28 costimulatory pairs is necessary for the induction of somatic Ig hypermutation in human B cells.

The mutations induced in CL-01 cells show intrinsic features of Ig V(D)J somatic hypermutation, but not evidence of Ag-driven selection

The 38 heavy chain and 9 λ-chain transcripts that appeared mutated by SSCP analysis were sequenced, and all contained somatic point-mutations. These V_HDJ_H-Cμ, -Cδ, -Cγ, -Cα, -Cε, and VλJλ-Cλ sequences were compared with those of the V_HDJ_H-C_H1 and VλJλ-Cλ transcripts expressed by the CL-01 cultured alone (unmutated templates) (Fig. 2). A total of 123 mutations were found in the 38 V_HDJ_HCμ, V_HDJ_H-Cδ, V_HDJ_H-Cγ, V_HDJ_H-Cα, and V_HDJ_H-Cε transcripts (575 bp each). A total of 121 were point-mutations confined to the V_HDJ_H sequence (375 bp), one was a deletion of three nucleotides in the framework region (FR) 3 of a V_HDJ_H-Cγ transcript (γ013), and one was a C→ G S mutation at residue 87 of the C_H1 sequence of a V_HDJ_H-Cμ1 transcript (μ015) (Fig. 2A). The 121 V_HDJ_H point-mutations corresponded to an overall frequency of 1.22 × 10⁻³ changes/base (Table I), >30-fold the PCR amplification error rate with high-fidelity Pfu DNA polymerase (~10⁻⁶ bases/cycle, i.e., 4.0 × 10⁻⁵ changes/base in 30 cycles) (p < 0.001). The single nucleotide change in the C_H1 region of V_HDJ_H-Cμ1 transcript μ015 represented a frequency of 1.35 × 10⁻⁴ changes/base, which was threefold the PCR error rate. The 121 V_HDJ_H nucleotide changes comprised 96 independent mutations, 17 of which were observed in multiple transcripts. The 96 independent point-mutations consisted of 68 R, 27 S, and one stop codon mutations and yielded an overall mutation rate of 2.42 × 10⁻⁴ base changes/cell division.

FIGURE 2.

The somatic mutations induced in vitro in CL-01 cells are scattered through the entire Ig V(D)J gene sequence and preferentially target selected nucleotides. A, Schematic depiction of the V_HDJ_H-C_H and the VλJλ-Cλ sequences of mutated transcripts (as defined by SSCP analysis) from CL-01 cells cultured with activated T cells after BCR engagement. Bars depict S mutations, lollipops depict R mutations, solid boxes (■) depict deletions, and the cross (×) depicts a stop codon. B, Sequences containing the 121 point-mutations identified in the 38 V_HDJ_H-C_H and the 16 point-mutations in the 9 VλJλ-Cλ transcripts from CL-01 cells cultured with activated T cells after BCR engagement as compared with the sequences of the unmutated V_HDJ_H and VλJλ templates (unstimulated CL-01 cells). Different nucleotide changes found in different transcripts are in many cases listed in the same row, resulting in a total of only 6 rows for V_HDJ_H and 2 rows for VλJλ, instead of a total of 38 and 9, respectively, actual mutated transcripts sequenced.

Seventeen mutations were found in the 9 VλJλ-Cλ transcripts sequenced (539 bp each). Sixteen were point-mutations confined to the VλJλ sequence (336 bp), and one was a 3-nucleotide deletion in the FR2 of transcript λ013. No mutations were found in the Cλ1 sequence (203 bp). The 16 VλJλ point-mutations corresponded to an overall frequency of 5.4 × 10⁻⁴ changes/base (Table I), a rate >10-fold the PCR error rate for 30 cycles (p < 0.001). These mutations comprised 13 independent mutations, 3 of which were observed in multiple transcripts. The 13 independent mutations consisted of 7 R and 6 S mutations and yielded an overall mutation rate of 1.08 × 10⁻⁴ base changes/cell division, a value lower than that in V_HDJ_H (p <0.05).

The highest load of mutations was found in the single V_HDJ_H-Cε transcript sequenced (Table I). Overall, the V_HDJ_H-Cγ transcripts bore a mutational load two- to threefold greater than that of the V_HDJ_H-Cμ, -Cδ, -Cα, or VλJλ-Cλ transcripts (Table I). In the V_HDJ_H-C_H transcripts, the 96 independent point-mutations targeted 83 nucleotide residues scattered throughout the V_HDJ_H sequence, with no preferential segregation to complementarity-determining regions (CDRs) or FRs (Fig. 2 and Table II). Likewise, in the VλJλ-Cλ transcripts, the 13 independent mutations targeted 13 nucleotide residues scattered throughout the VλJλ sequence, with no preferential segregation to CDRs or FRs (Fig. 2). In both the V_HDJ_H and VλJλ segments, the number of R mutations in CDRs was lower than that theoretically expected by chance alone, and, therefore, inconsistent with a positive selection of R mutations by Ag (Fig. 2 and Table II).

Table II.

Somatic point-mutations induced in CL-01 cells do not accumulate preferentially in V(D)J CDRs^a

	CDRs		FRs
	R	S	R	S
VHDJHb	21 [22]c	8 [6]	47 [50]	19 [18]
VλJλb	1 [2]	1 [1]	6 [7]	5 [3]
Total	22 [24]	9 [7]	53 [57]	24 [21]

^aIn addition to the R and S mutations, an Amber stop codon was found in the heavy chain FR3 region.

^bIdentical mutations in different transcripts of the same and different isotypes were assumed not to be independent and were counted only once.

^cThe [expected] number of mutations was calculated by multiplying the number of observed total (R and S) mutations by the R mutation frequency inherent to the V(D)J gene CDR or FR sequence and the relative size of those regions in the unmutated V(D)J sequence (unstimulated CL-01 cells). The Ig V(D)J gene inherent frequency of mutation was calculated with Inh. Sus. Calc. program v 1.0 (30).

The mutations induced in CL-01 cells show a bias for transitions over transversions and preferentially target G nucleotides and AG and CT dinucleotides

Randomly occurring point-mutations are expected to be one-third transitions and two-thirds transversions, but the 96 unique V_HDJ_H point-mutations were equally divided between transitions and transversions (p < 0.01) (Table III), and the 14 unique VλJλ point-mutations consisted of almost twice as many transitions (n = 9) as transversions (n = 5) (p < 0.01). In the V_HDJ_H transcripts, G nucleotides were mutated at a frequency (46.9% of total mutations) about 50% higher than that expected by chance alone (33.3%) after correcting for base composition, i.e., normalizing for the relative occurrence of G in the unmutated V_HDJ_H sequence template (p < 0.05), with G → A transitions accounting for 42.2% of the total G mutations and 43.2% of the total transitions (Table III). Preferential mutation of G was associated with scarcity of mutations in A (p < 0.001) and T (p < 0.001) and stochastic frequency of mutations in C (Table III). A preference for transitions over transversions and G nucleotide targeting could be discerned in the VλJλ transcripts, but the small sample size did not allow a meaningful statistical conclusion.

Table III.

Nature of the base substitutions in the Ig V_HDJ_H gene segment of induced CL-01 cells

Transitions	G → A	A → G	C → T	T → C
44 [32]	19a [9.7]b	5 [7.6]	13 [7.8]	7 [6.8]
Transversions
	G → C	A → C	C → A	T → A
23 [32]	9 [9.7]	1 [7.6]	9 [7.8]	4 [6.8]
	G → T	A → T	C → G	T → G
29 [32]	17 [9.7]	1 [7.6]	9 [7.8]	2 [6.8]
Total mutations	G → N	A → N	C → N	T → N
96 [96]	45* [29.2]	7** [22.8]	31 [23.5]	13** [20.4]

^aIdentical mutations in different transcripts of the same and different isotypes were assumed not to be independent and were counted only once.

^bThe [expected] number of mutations (from a given nucleotide residue to another given nucleotide residue) was normalized for the base composition of the unmutated V_HDJ_H sequence. It was calculated by multiplying the frequency of occurrence of the nucleotide target of mutation in the unmutated sequence (unstimulated CL-01 cells) by the total number of observed mutations, and dividing this product by three. For instance, the expected number of G → A mutations was calculated by multiplying 0.30 (G frequency of occurrence in the unmutated V_HDJ_H sequence) by 96 = 29.2, divided by 3 (as G → A, G → C, and G → T mutations have all the same theoretical probability to occur) = 9.7.

^*p < 0.05;

^**p < 0.001.

Certain sequence motifs have been suggested to be preferentially targeted by the hypermutation machinery (2, 6, 8, 34–36). We considered the 16 possible dinucleotides identifiable in V_HDJ_H the sequence and made a census of all the point-mutations targeting these dinucleotides (the VλJλ sequence was not considered because of the small number of point-mutations) (Table IV). All 16 dinucleotides were targeted by mutations except for TT. The AG dinucleotide and its inverse repeat CT were found to be mutated at a frequency significantly higher than that expected by chance alone (p < 0.001) (Table IV). This was not due to a mere G and C preference (Table IV), as G and C were mutated at a significantly lower frequency when occurring outside than within AG and CT (38 and 45% vs 62 and 55% of total G and C mutations, respectively). In addition, GA and TC, which contain the same nucleotides in inverse order, contained 2.3- and 3.4-fold fewer G mutations than AG and CT (Table IV).

Table IV.

Somatic point-mutations induced in the V_HDJ_H gene segment of CL-01 cells preferentially target AG and CT dinucleotides

Dinucleotides	Dinucleotides in Unmutated Sequence		Nucleotide Changes in Dinucleotides of Mutated Sequences
	Number of occurrencesa	Percent of the total dinucleotides	Observed number of mutationsb,c	Expected number of mutationsc,d	Observed to expected mutation ratio
AG	28	7.5	32	14.4	2.22*
GA	26	7.0	14	13.4	1.04
TA	16	4.3	2	7.8	0.26*
GC	18	4.8	13	9.2	1.41
CG	14	3.7	7	7.1	0.99
AC	27	7.2	10	13.8	0.72
AA	17	4.6	4	8.6	0.47*
AT	16	4.3	2	8.3	0.24*
CT	26	7.0	25	13.4	1.87*
GT	26	7.0	17	13.4	1.27
CA	30	8.0	5	15.4	0.32*
TG	27	7.2	12	13.8	0.87
CC	22	5.9	15	11.3	1.33
GG	44	11.8	22	22.7	0.97
TT	12	3.2	0	6.2	0.00*
TC	25	6.7	12	12.9	0.93
Total	374	100	192	192.0	1.00

^aThe V_HDJ_H segment of CL-01 cells comprises 375 nucleotides. Therefore, the number of occurrences is 375–1, as dinucleotides are identified in two different frames (i.e., beginning with residue 1 and residue 2).

^bIdentical mutations in different transcripts of the same and different isotypes were assumed not to be independent and were counted only once.

^cThe number of mutations in any dinucleotide cohort is twice the number of actual somatic point-mutations because each mutation is counted twice as being shared by two partially overlapping dinucleotides.

^dThe expected number of mutations for any given dinucleotide takes into account the base composition of the unmutated V_HDJ_H sequence (i.e., it was corrected for the sequence base composition). It was calculated by multiplying the frequency of occurrence of each of the dinucleotides considered in the unmutated sequence by the total number of observed point-mutations in the V_HDJ_H region.

^*p < 0.001.

GAG and AGG trinucleotides are preferential targets of point-mutations as a result of the preferential targeting of G within the AG dinucleotide

We examined all occurrences of AGN, NAG, CTN, and NCT trinucleotides to verify whether the observed concentration of somatic point-mutations in the AG and the CT dinucleotides could in fact reflect a preference for the trinucleotides that harbor these dinucleotides (Table V). We also examined AGC and TAC and their inverse repeats GCT and GTA, as these trinucleotides have been suggested to be preferential targets of hypermutation (36). Finally, we considered all other trinucleotides that had been targeted by 10 or more point-mutations in the CL-01 V_HDJ_H sequence. The trinucleotides most frequently targeted by mutations were AGA, GAG, AGG, GGC, GAC, ACC, GCT, and GAC (p < 0.01). AGC and TAC, and the inverse repeat of the latter, GTA, contained a number of mutations not higher than that expected by chance, and so did CTC, GGG, CTG, CCT, and CAG.

Table V.

Somatic point-mutations induced in the V_HDJ_H gene segment of CL-01 cells preferentially target GAG trinucleotides^a

^aThe trinucleotides shown in this table are those targeted by 10 or more point-mutations and those that have been suggested to be hotspots (36).

^bIdentical mutations in different transcripts of the same and different isotypes were assumed not to be independent and were counted only once.

^cThe [expected] number of mutations was normalized for the base composition of the unmutated V_HDJ_H sequence. The [expected] number of mutations targeting any given trinucleotide was calculated by multiplying the frequency of the occurrence of each of the trinucleotides considered in the unmutated sequence by the total number of point-mutations.

^dThe [expected] number of mutations targeting the underlined dinucleotide in or out of the context of the trinucleotide, respectively, was calculated by multiplying the frequency of occurrence of the dinucleotide in or out of the context of the trinucleotide, respectively, by the total number of point-mutations that occurred in that given dinucleotide.

^*p < 0.01.

In AGA, GAG, and AGG, AG was the preferential target of mutations, containing 10 of the 10, 16 of the 21, and 9 of the 10 mutations, respectively, found in these trinucleotides, with 8 of the 10, 15 of the 16, and 9 of the 9 mutations targeting the dinucleotide G residue (p < 0.001) (Table V). In the unmutated V_HDJ_H sequence, G occurred 69 times outside AG, AGA, GAG, or AGG and 36 times as part of these di- or trinucleotides, but was the target of mutation 16 and 29 times, respectively, in different transcripts (p < 0.001). AG was targeted by mutations at the expected frequency whether occurring in or outside AGA, GAG, and AGG trinucleotides. Thus, the high frequency of mutations in AGA, GAG, and AGG trinucleotides reflected the preferential targeting of G within the AG dinucleotide.

The RGYW tetranucleotide motif is virtually spared by somatic point-mutations

The consensus RGYW motif, where R is a purine (A or G), Y is a pyrimidine (C or T), and W is A or T, has been identified as a mutational hotspot (35). Census of the CL-01 Ig V_HDJ_H sequence for the presence of the RGYW tetranucleotide revealed that this motif occurred 11 times in 6 different iterations (AGCA twice, AGCT once, AGTA twice, AGTT none, GGCA once, GGCT three times, GGTA twice, and GGTT none) and was mutated more frequently than expected only in the GGCT iteration (p < 0.01), in which 6 of 9 mutations targeted the terminal CT dinucleotide. The lower frequency of mutation of the AGCT iteration, also containing a CT dinucleotide, suggested that the high mutability of GGCT might be attributed to a target bias of its two intrinsic and partially overlapping GGC and GCT trinucleotides.

Somatic point-mutations accumulate stepwise concomitant with Ig class switching

The ability of sIgM⁺ sIgD⁺ CL-01 cells to switch to IgG, IgA, and IgE allowed us to address the relationship between somatic hyper-mutation and Ig class switching. A genealogical tree was constructed using the V_HDJ_H sequences derived from CL-01 cells cocultured with T cells in a single 14-day culture after BCR engagement. Identical mutations in different transcripts were considered to be shared mutations, although some of these might have arisen as hotspots and therefore could actually be independent mutations. There are three branches to the genealogical tree we generated, spanning four, three, and two generations (Fig. 3). The first branch stemmed from a first-generation putative intermediate bearing a 56 G → A transition, which could be tracked down to two fourth-generation IgG-switched elements bearing 5 and 10 point-mutations (γ013 and γ002), through a putative second-generation intermediate bearing the 56 G → A and a 267 G → A transition and a third-generation putative intermediate bearing four shared mutations, flanked by an IgD intermediate bearing the two second-generation shared mutations plus a third mutation (336 G → A). The second branch stemmed form a first-generation putative intermediate bearing a 20 C → T transition. This gave rise to two third-generation elements, one still expressing IgD (δ075), the other switched to IgE (ε001), both with seven point-mutations, of which two shared the 115 C → A transversion and the original 20 C → T transition. An unswitched IgD element bearing five point-mutations and a putative IgD/IgM second-generation element with two mutations that were conserved in the third-generation elements were also part of this branch. Finally, the third branch consisted of a first-generation element characterized by a 355 C → A transversion, which gave rise to IgM (μ015) and IgA (α006) second-generation elements bearing three and five point-mutations, of which only the original 355 C → A transversion was shared. Thus, the mutational machinery was active throughout class switching in CL-01 cells, and point-mutations accumulated in a stepwise fashion in IgG, IgA, and IgE.

FIGURE 3.

Stepwise accumulation of somatic point-mutations correlates with class switching from IgM to IgG, IgA, or IgE in CL-01 cells. Depicted is the genealogical tree constructed using related V_HDJ_H sequences of V_HDJ_H-Cμ, V_HDJ_H-Cδ, V_HDJ_H-Cγ, V_HDJ_H-Cα, and V_HDJ_H-Cε transcripts from a single culture of CL-01 cells incubated, after BCR engagement, with activated T cells for 14 days. Thin frames depict putative intermediate sequences, and point-mutations are indicated by their codon number and the nature of the base change. Vertical bars depict S mutations, and lollipops depict R mutations.

Discussion

The present findings show that human CL-01 cells can be induced to hypermutate the expressed Ig V(D)J genes in vitro with modalities that are congruent with those inferred from studies in vivo. Together with our other findings (15, 16, 25, 26), they also show that in this monoclonal cell line somatic hypermutation is induced in the context of a coordinated program that recapitulates GC B cell differentiation, whereby CL-01 cells that are sIgM⁺ sIgD⁺ switch to IgG, IgA, and IgE and eventually give rise to plasmacytes and memory-like B cells. In CL-01 cells, somatic hypermutation is induced in all primary and secondary V_HDJ_H-C_H, as well as VλJλ-Cλ transcripts, when the appropriate stimuli are applied, i.e., BCR engagement and T cell contact, allowing for engagement of at least the two costimulatory molecule pairs CD40:CD40L and CD80:CD28.

The ability of CL-01 cells to undergo Ig somatic hypermutation and class switching as part of an integrated differentiation program has allowed us to determine that the requirements for the induction of these two central GC maturational processes are different. As we have shown (15, 16, 25), CD40 engagement by soluble trimeric CD40L or CD40L-expressing 293 cells in the absence of T cells effectively induces CL-01 cells, as well as freshly isolated normal human IgM⁺ IgD⁺ B cells, to switch to IgG, and IgA, and, in the presence of IL-4, to IgE. As we show here, CD40:CD40L engagement is necessary but not sufficient for the induction of Ig somatic hypermutation. CD40L-expressing 293 cells cannot substitute for T cells in inducing somatic hypermutation, even in the presence of TCM, further enriched with IL-4 and IL-10, and after BCR engagement. The failure of these switched B cells to accumulate mutations further points to somatic hypermutation and class switching as two independent and discrete processes and indicates that engagement of additional B:T cell costimulatory molecule pair(s) is indispensable to trigger the mutational machinery.

Our present findings strengthen those showing that both BCR engagement and T cell help are required to trigger Ig hypermutation (21, 24) and extend them by providing evidence that B:T cell contact allowing for CD40:CD40L and CD80:CD28 coengagement is necessary, in addition to BCR engagement, to induce this process. Furthermore, by showing that anti-CD30 mAb fails to interfere with the induction of the mutational machinery, they provide an explanation for the putatively normal somatic hypermutation process observed in CD30-deficient mouse mutants (37) and indicate that the anti-CD40L, anti-CD80, and anti-CD28 mAbs ablated Ig hypermutation not by merely reducing or abrogating B:T cell contact, but rather by specifically interfering with CD40: CD40L or CD80:CD28 coengagement and related signaling. Whether, after BCR engagement, CD40:CD40L and CD80:CD28 coengagement is sufficient to induce Ig hypermutation or the co-engagement of these costimulatory pairs mediates the induction of crucial T cell surface molecules and/or factors remains to be determined.

Ig somatic hypermutation and class switching are intimately linked in the GC, where point-mutations account for a vast majority of the changes, which include only scarce insertions and deletions (38). Ig hypermutation is thought to begin in centroblasts before class switching (15, 39). It is not known whether high-affinity GC centrocytes that have undergone isotype switching undergo further hypermutation, although it has been suggested that isotype switching does not terminate hypermutation (40). Our present findings at the clonal level suggest that the onset of somatic hypermutation is not related to Ig class switching to IgG, IgA, or IgE and further emphasize that these are two independent processes. They also show that the modalities of in vivo GC hypermutation are reflected in CL-01 cells, in which, vis-à-vis of 137 somatic point-mutations, only two codon deletions were found, both involving triplets and both leaving the transcripts in frame. The conservation of shared mutations among primary and secondary isotypes and the observation of nonmutated Ig secondary isotype transcripts suggest that B cells accumulate somatic mutations along cell divisions before and after isotype switching occurs.

In induced CL-01 cells, V_HDJ_H-Cγ and -Cε transcripts bore a load of somatic mutations approximately twofold greater than that of V_HDJ_H-Cα transcripts or their primary V_HDJ_H-Cμ and V_HDJ_H-Cδ counterparts (Table I). The higher frequency of mutations in CL-01 cell V_HDJ_H-Cγ transcripts in vitro extends findings in vivo showing, in centroblasts, a heavier load of mutations in V_HDJ_H-Cγ than in V_HDJ_H-Cμ transcripts (3, 41) and indicates that this differential mutational load may reflect an inherent feature of the integrated hypermutation and class-switching processes. This contention would be further supported by the 1000-fold higher mutation rate in a V_HDJ_H-Cγ construct than in its V_HDJ_H-Cμ counterpart after stable integration into an established B cell line, (42, 43). In CL-01 cells, the frequency of mutations in V_HDJ_H-Cγ, V_HDJ_H-Cμ, and V_HDJ_H-Cδ, but not V_HDJ_H-Cα, transcripts was about two- to fourfold higher than in VλJλ-Cλ transcripts. A similar higher load of mutations in the heavy chain than the light chain of Ag-selected Abs has been interpreted to reflect the dominant role of the V_HDJ_H over the V_LJ_L segment in providing the structural correlate for Ag binding (44, 45). Our in vitro findings in a system putatively devoid of nominal antigenic pressure suggest that more efficient targeting of mutations to the heavy chain is an intrinsic property of the hypermutation machinery.

In vivo studies have shown that Ig V(D)J somatic point-mutations are not generated randomly in terms of base substitution, preference, and distribution (hotspots) (6, 36, 46). In induced CL-01 cells, G bases were preferentially targeted with G → A transitions, which are especially copious in all transcripts of different heavy chain isotypes. The preferential G targeting in CL-01 cells differs from the lack of such a bias in productive and non-productive human V_HDJ_H rearrangements ex vivo (27, 47–53) and contrasts with the preferential targeting of A in the mouse (8, 36). A preferential targeting of G nucleotides has been observed in vitro in human BL2 cells (24), Chinese hamster cells (54), and mouse B cells (55) and in vivo in exotherms that do not have GCs (56), suggesting the mechanism(s) driving such a mutational preference is phylogenetically conserved and is operational in the absence of the GC microenvironment. A preferential G targeting has also been found in MSH2 mismatch repair protein-deficient mice (57, 58) and may be related to the activity of this protein, which corrects mismatches mainly at G and C (58).

The identification of sequence-specific preferences by the somatic hypermutation machinery is important as it may provide clues in determining the molecular mechanism of this process. In addition to base targeting bias, mutational hotspots have been identified throughout Ig V(D)J gene sequences (6, 36, 46). While some of these hotspots may reflect the application of selection force(s) on the gene product, others may be inherent to the nucleotide sequence targeted by mutations, as suggested by the analysis of “passenger” or nonselectable genes (59, 60). The putative lack of the application of a nominal positive or negative selective pressure to CL-01 cells under the culture conditions used here should allow for an insight into the base preference(s) and hotspot(s) that are inherent to the mutational machinery. In our experiments, the striking bias for G mutations (more than half of all V_HDJ_H point-mutations) would reflect a marked mutational preference for the AG dinucleotide, whether isolated or in the context of AGA, GAG, and AGG trinucleotides.

For convention, all the present mutations in CL-01 cells were recorded from the coding strand, although it is not known on which strand mutations occur. The high G:C mutation ratio (1.6) in CL-01 cells resembles the marked bias toward mutation of purines in the coding strand, especially G → A transitions as in sheep Vλ regions (61). As emphasized by Storb, one cannot detect a strand bias if there is not also a G/C or A/T bias (8). In CL-01 cells, the higher than expected frequency of G mutations recorded from the coding strand is consistent with either preferential G targeting by the mutational machinery in the nontranscribed (“top”) strand or preferential C targeting by the mutational machinery in the transcribed (“bottom”) strand. In CL-01 cells, this strand polarity is further supported by the finding that T accumulated almost twice as many mutations as A and is in agreement with the proposed strand polarity of somatic mutations in the experimental mouse (2, 5, 8), although, in the mouse, strand polarity has been associated with an A > T bias.

The Ser encoded by AGC or AGT, particularly at position 31 of Ig V(D)J sequences, has been reported to constitute a hotspot both in vivo (2, 36) and in vitro (24). AGC and AGT codons occur five times in the CL-01 Ig V_HDJ_H sequence, and only two of them were mutated, neither at codon 31 (AGC). A comparable sparing of AGC and AGT codons by mutations has been found in several human Ig V(D)J sequences in vivo (38) and in vitro (20). In general, many AGC/T sequences are not mutational hotspots, and the local targeting of mutations is not simply dependent on the two or three bases surrounding the hotspot (2). Other structural features, such as palindromes, may well be important (62). Likewise, a scarcity of mutations was found in the different iterations of the RGYW motif in the of CL-01 cell V_HDJ_H gene sequence, despite the originally proposed hotspot nature of this sequence (35). This suggests that the hypermutation process has been subjected to evolutionary pressure to yield substitutions over the whole V(D)J region and that aspects of local secondary structure are also likely to contribute to the formation of mutational hotspots. These and other mechanistic issues, including those related to the activation of the mutational machinery, can be optimally addressed by further use of CL-01 cells.