Main

Clonal expansion is an essential feature of the immune response. B lymphocytes bearing antigen-specific immunoglobulins undergo this process in the germinal centre, a specialized microanatomical compartment where B cells also diversify their immunoglobulin genes through somatic hypermutation1,2,3,4. GC B cells expressing mutated surface immunoglobulins with the highest affinity are then positively selected by iterative cycles of cell division, somatic hypermutation and selection5,6,7,8,9,10, endowing the host with high-affinity humoral immunity4.

GC B cells divide and mutate in the dark zone (DZ), and then migrate to the light zone (LZ) where they capture antigen through surface immunoglobulin and present it as peptide bound to major histocompatibility complex class II (pMHCII) to cognate follicular helper T cells (TFH) 4,10,11,12. Migration between the two zones is mediated by the chemokine receptors CXCR4 and CXCR5, with 50% of DZ cells migrating to the LZ, and 10% returning to the DZ from the LZ within 6 h5,10,13. Moreover, B cells in the two GC zones alternate between distinct genetic programs reflecting cell division in the DZ and selection in the LZ, but do so independently of local cues received in the two zones10,14. However, the precise mechanism by which the highest affinity cells are selected, and whether cell divisions and immunoglobulin mutations in the DZ are regulated, remains unknown14.

To determine whether the amount of antigen internalized by GC B cells governs the extent of clonal expansion, we titrated the amount of antigen delivered to GC B cells using antibodies that target DEC205, an endocytic receptor that carries antigen to intracellular MHCII-containing compartments10,15,16,17,18. GC responses were initiated by priming mice with ovalbumin (OVA), followed by boosting with OVA coupled to the hapten 4-hydroxy-3-nitrophenylacetyl (NP–OVA)9. Antigen-specific B-cell responses were tracked by adoptive transfer of B1-8hi immunoglobulin heavy chain knock-in B cells, which are specific for NP when they express immunoglobulin lambda (Igλ) light chains19. To measure the relative expansion of B cells receiving graded amounts of antigen, GCs were induced in mice that received a mixture of B1-8hi DEC205+/+ and B1-8hi DEC205−/− B cells at a 5:95 ratio. Graded doses of antigen were delivered to DEC205+/+ GC B cells using the chimaeric DEC205 antibody fused to cognate antigen, OVA (αDEC–OVA, Fig. 1a)20. Whereas control injections with PBS had no effect, injection with 10 μg of αDEC–OVA resulted in selective expansion of the B1-8hi DEC205+/+ GC B cells (Fig. 1b, c and Extended Data Fig. 1). Decreasing the dose of antigen delivered, by mixing αDEC–OVA with a chimaeric DEC205 antibody carrying the control irrelevant antigen Plasmodium falciparum circumsporozoite protein (αDEC–CS), resulted in decreased expansion of B1-8hi DEC205+/+ GC B cells that was proportional to the dose of αDEC–OVA (Fig. 1b, c). Consistent with the idea that pMHCII-mediated selection occurs in the LZ and cell division in the DZ4,10, selective dose-dependent expansion of B1-8hi DEC205+/+ GC B cells was already evident at 48 h in the DZ but only later in the LZ (Fig. 1d, e). In contrast, the B1-8hi DEC205−/− GC B-cell population contracted in proportion to the amount of antigen delivered to the B1-8hi DEC205+/+ GC B-cell population (Fig. 1b and Extended Data Fig. 1c). Thus, increasing the amount of cognate antigen presented by a subset of GC B cells to TFH cells leads to their proportional and selective expansion at the expense of GC B cells that present less antigen.

Figure 1: The amount of antigen captured and presented by GC B cells regulates their expansion.
figure 1

a, Protocol for be. i.p., intraperitoneally; s.c., subcutaneously. (1.5–5) × 106 B1-8hi DEC205+/+ and B1-8hi DEC205−/− B cells (≈ (1.5–5) × 105 Igλ+, NP-specific B cells) at a 5:95 ratio were transferred into OVA-primed wild-type mice, which were boosted with NP–OVA. After 6 days, mice were injected with PBS or αDEC–OVA mixed with αDEC–CS at ratios of 1:0, 1:3, 1:9, or 1:39. Lymph nodes were analysed 2, 3 and 4 days after injection. b, Proportion of B1-8hi DEC205+/+ and B1-8hi DEC205−/− GC B cells 48 h after treatment. c–e, Mean fraction of DEC205+/+ B cells among B1-8hi GC (c), DZ (CD86CXCR4+, d), and LZ (CD86+CXCR4, e) cells. Error bars, s.e.m. Data represent 2–3 independent experiments at each time point with a total of 4–6 mice per condition for all time points.

PowerPoint slide

To examine the mechanism by which increased T-cell help leads to selective GC B-cell expansion, we sought to measure cell division in the GC. Traditional dye-based methods to monitor cell division are unsuitable in this context because B cells divide extensively and lose most of the dye before entering the GC. To circumvent this problem, we combined transgenes encoding the tetracycline transactivator (tTA) protein expressed under the Vav promoter and a histone H2B–mCherry fusion protein driven by a doxycycline (DOX)-regulated promoter (tTA–H2B–mCh, Extended Data Fig. 2a)21,22. Under steady state conditions, tTA is expressed in haematopoietic cells and induces high levels of H2B–mCh expression (Extended Data Fig. 2b). Administration of DOX represses further H2B–mCh synthesis, and as a result the H2B–mCh dilutes in proportion to cell division (Extended Data Fig. 2c)23.

To determine whether tTA–H2B–mCh can be used to track antigen-specific GC B-cell division in vivo, we repeated the prime-boost protocol described above using B1-8hi tTA–H2B–mCh B cells. As expected, non-proliferative follicular B cells did not dilute H2B–mCh after DOX treatment (Fig. 2a). In contrast, after 36 h, a spectrum of discrete peaks of H2B–mCh expression corresponding to cell divisions became evident among GC B cells, and by 84 h, GC B cells had completely diluted H2B–mCh (Fig. 2a). Thus, tTA–H2B–mCh can be used to monitor cell division in the GC.

Figure 2: T-cell help regulates the number of GC B-cell divisions.
figure 2

a, H2B–mCh fluorescence among B1-8hi tTA–H2B–mCh B cells within the follicular (upper) or GC (lower) compartments of untreated mice (black) or after 36 (red) or 84 (blue) hours on DOX. Solid grey represents non-fluorescent cells. b, H2B–mCh fluorescence among B1-8hi tTA–H2B–mCh GC B cells in mice that received a 5:95 mixture of B1-8hi tTA–H2B–mCh and B1-8hi DEC205−/− B cells and were treated with PBS, αDEC–CS or αDEC–OVA for 72 h and administered DOX for 60 h before analysis. c, Mean fraction of B1-8hi tTA–H2B–mCh GC B cells treated as in b. Error bars, s.e.m. d, Per cent DZ cells among B1-8hi tTA–H2B–mCh GC B cells after treatment with PBS or αDEC–CS (control) or αDEC–OVA for 48 h and DOX for 36 h. Each symbol represents one mouse and lines represent mean values. Data represent 3–4 independent experiments for all time points with 2–3 mice for each condition per experiment. *P < 0.005; **P < 0.001; ***P < 0.0001, two-tailed Mann–Whitney test.

PowerPoint slide

To determine whether the amount of antigen captured by GC B cells influences their degree of proliferation, we delivered additional antigen to GC B cells using αDEC–OVA. After 60 h on DOX, B1-8hi DEC205+/+ tTA–H2B–mCh GC B cells from mice treated with control αDEC–CS or PBS were nearly evenly distributed among mChMed, mChLo and mCh groups, representing cells that underwent progressively more cell division. In contrast, after 60 h on DOX, 50% of GC B cells targeted with αDEC–OVA became mCh and 40% were mChLo (Fig. 2b, c). Therefore, increased antigen capture and presentation leads to increased rates of cell division by GC B cells.

Concomitant with the increased rate of cell division, there was a change in the zonal distribution of GC B cells. Whereas control cells equilibrated between DZ and LZ at a 2:1 ratio on average, αDEC–OVA-targeted GC B cells were found almost exclusively in the DZ (90%, Fig. 2d).

To examine cell cycle distribution during selection, we labelled GC B cells with an intravenous pulse of the nucleoside analogs 5-ethynyl-2′-deoxyuridine (EdU) followed 1 h later by 5-bromo-2′-deoxyuridine (BrdU), allowing separation of GC B cells into the earliest (EdUBrdU+), mid/late (EdU+BrdU+) and post- (EdU+BrdU) S phase periods of the cell cycle (Fig. 3a and Extended Data Fig. 3). Control GC B cells in early and mid/late-S phases were found in both zones; post-S phase cells were primarily in the DZ (Fig. 3b). This supports the idea that GC B cells can initiate S phase in the LZ and then migrate to the DZ to divide10. Moreover, the fact that early S phase cells are distributed at a 2:1 DZ:LZ ratio suggests that, once in the DZ, GC B cells can initiate two cell divisions on average. In contrast, inducing selection by injecting αDEC–OVA shifted early S phase cells among selected GC B cells to the DZ (86%), whereas slightly fewer early S phase cells were in the DZ (59%) among non-selected cells (Fig. 3b). We conclude that increasing the amount of antigen captured and presented to TFH cells increases the proportion of cells initiating S phase in the DZ. These findings indicate that GC B cells presenting the highest levels of antigen initiate additional cell divisions in the DZ before returning to the LZ.

Figure 3: Selective expansion involves increased S phase initiation in the DZ and longer DZ residence time.
figure 3

a, EdU and BrdU incorporation among GC B cells. OVA-primed mice received 5 × 106 B1-8hi DEC205+/+ and B1-8hi DEC205−/− B cells at a 15:85 or 50:50 ratio. On day 6 after boosting with NP–OVA, mice receiving the 15:85 or the 50:50 B-cell transfer were injected with αDEC–OVA or PBS, respectively. Two days later, mice received EdU and 1 h later BrdU and were analysed after 30 min (detailed in Extended Data Fig. 3a). b, Per cent DZ cells among early (E), mid/late (M/L), and post- (P) S phase cells. Black squares denote αDEC–OVA treatment; white squares denote PBS control. Data represent two independent experiments with 6 or 7 mice per condition in total. Squares indicate mean values; error bars, s.e.m. c, d, OVA-primed mice received 5 × 106 B1-8hi PAGFP+ and B1-8hi DEC205−/− B cells at a 15:85 ratio or 5 × 106 B1-8hi PAGFP+ B cells alone and were boosted with NP–OVA. For mice receiving mixed B-cell transfers, αDEC–OVA was injected on day 6. On day 8, DZ cells were photoactivated (c) and GC B cells were analysed by flow cytometry (detailed in Extended Data Fig. 4a) (d). e, Per cent LZ cells among photoactivated B1-8hi PAGFP+ GC B cells at 0 or 6 h after DZ photoactivation. Each symbol represents one mouse. Data represent multiple independent experiments. **P = 0.0022, two-tailed Mann–Whitney test.

PowerPoint slide

We used photoactivation to test whether selected B cells presenting higher levels of antigen reside longer in the DZ10. Primed mice received a mixture of B1-8hi DEC205+/+ B cells expressing a photoactivatable green fluorescent protein (B1-8hi PAGFP+) and B1-8hi DEC205−/− B cells. On day 6 of the GC response, we induced selection by injecting αDEC–OVA. DZ cells were photoactivated and lymph nodes were processed for flow cytometry either 0 or 6 h after photoactivation (Fig. 3c and Extended Data Fig. 4). Under control conditions, 57% of photoactivated GC B cells migrated to the LZ after 6 h10. In contrast, B1-8hi PAGFP+ cells undergoing selective expansion migrated from the DZ to the LZ at half this rate (Fig. 3d, e). Thus, GC B cells undergoing selective expansion as a result of increased antigen presentation reside longer in the DZ, allowing them to undergo additional cell divisions before returning to the LZ. We conclude that TFH cells regulate the number of cell cycles a GC B cell initiates during each passage through the DZ, and that they do so in direct proportion to the amount of antigen captured and presented.

To determine whether variable numbers of cell divisions in the DZ are associated with selection during a polyclonal immune response, we immunized non-immunoglobulin transgenic, tTA–H2B–mCh mice with NP–OVA. Mice were given DOX at day 12.5 after immunization, and 36 h later mChHi and mChLo GC B cells, representing lower and higher rates of proliferation, respectively, were isolated by cell sorting (Fig. 4a, b). VH186.2 family genes were then analysed for the characteristic high affinity anti-NP W33L mutation24. We found that 35.9%, of the highly divided GC B cells were W33L+; in contrast, only 6% of the cells that underwent the fewest divisions were W33L+ (P < 0.0001, Fig. 4b). Thus, selection of high-affinity GC B cells in a polyclonal response is associated with increased rates of proliferation. Moreover, because each division can introduce mutations that increase or decrease affinity, even high-affinity clones will rapidly partition into subclones that undergo different rates of proliferation, death or differentiation. This is probably reflected in the observation that mChLo cells are not all W33L+, and that W33L+ cells are not exclusively in the mChLo fraction.

Figure 4: Increased cell division in polyclonal GCs is associated with higher immunoglobulin affinity and somatic hypermutation.
figure 4

ad, tTA–H2B–mCh mice were immunized with NP–OVA (ac) or YU2-gp120 (d) and administered DOX for 36 h before purification. a, Representative histogram displaying mChHi and mChLo gates. b, Frequency of W33L mutation among VH186.2 clones in mChHi and mChLo GC B cells on day 14 after immunization. The number of W33L+ clones among VH186.2 sequences is shown in the centre of the pie chart and was compared using Fisher’s exact test. Data are pooled from 2 independent experiments with 2 mice each. c, d, Somatic hypermutation in the JH4 intron among mChHi and mChLo GC B cells on day 7 after NP–OVA (c) and day 9 after immunization with HIV-1YU2 gp120 (d). Data in c and d are pooled from 2 independent experiments with 2–3 mice each. Over 160 clones in c and 140 clones in d were analysed for mChHi and mChLo GC B cells and 51 clones in c and 94 clones in d were analysed for follicular B cells. Fo, follicular B cells. **P = 0.0033; ***P < 0.0001, χ2 test with Yates correction.

PowerPoint slide

Each round of cell division in the GC is predicted to produce 1 somatic mutation per 103 base pairs in immunoglobulin genes8. The finding that TFH cells control the number of B-cell divisions per GC cycle indicates that they may also regulate hypermutation. To determine whether there is a correlation between selection, proliferation and somatic hypermutation in the GC, we immunized tTA–H2B–mCh mice with NP–OVA, purified mChHi and mChLo GC B cells by cell sorting and analysed the intron downstream of JH4 for somatic mutations, because this region is targeted for somatic hypermutation but is not subject to selection25. GC B cells that had diluted H2B–mCh by undergoing a greater number of divisions were significantly more mutated than those that had undergone fewer divisions (P < 0.0001, Fig. 4c).

High levels of mutation are required to produce broadly neutralizing antibodies to human immunodeficiency virus (HIV)-1 and are believed to be an impediment to vaccine development26. To determine whether mutations accumulate differentially among GC B cells responding to HIV-1 gp120, we immunized tTA–H2B–mCh mice with HIV-1YU2 gp120 and compared GC B cells that had undergone different levels of division. Similar to NP–OVA, the more divided mChLo B cells in the anti-gp120 GCs were significantly more mutated than either mChHi GC or follicular B cells (Fig. 4d). We conclude that B cells undergo a variable number of divisions in the DZ before returning to the LZ, and that increased cell division is associated with higher immunoglobulin affinity and increased somatic hypermutation.

The GC is a site of intense immunoglobulin diversification and selection, from which high-affinity B cells emerge that seed the memory and plasma cell compartments3,4,27. Affinity-based immunoglobulin selection is accomplished iteratively. B-cell clones expanding and mutating in the DZ travel to the LZ to compete for help from TFH cells, with only a fraction of LZ cells selected to return to the DZ to continue cycling10. Our experiments indicate that the number of GC B-cell divisions per DZ cycle is variable, ranging from 1 to 6, and is regulated by the amount of antigen captured in the LZ. By capturing more antigen and dividing a greater number of times in the DZ during each interzonal cycle, high-affinity GC B cells outcompete lower affinity cells that capture less antigen and divide fewer times. The magnitude of T-cell help provided in the LZ therefore regulates the behaviour of a DZ cell—inducing selected cells to re-enter the cell cycle a variable number of times before migrating back to the LZ. This finding is consistent with and may explain the observation that the switch from DZ to LZ phenotype is independent of cues received in the DZ14. Moreover, the least proliferative GC B cells have the fewest affinity-enhancing mutations and are disproportionately found in the LZ (Figs 2d and 4b). The precise contributions of death and differentiation to the eventual disappearance of these cells from the GC remain to be determined.

An important feature of differential cell division in the dark zone is that each round of division is associated with increased accumulation of somatic hypermutation. Consistent with this idea, highly proliferative germinal centre B cells have both higher immunoglobulin affinity and a greater number of somatic mutations. Our experiments reveal a feed-forward loop in the germinal centre, whereby somatic hypermutation is greatest among germinal centre B cells whose antibodies provide the highest affinity template for further diversification. This finding may be particularly relevant for designing immunization strategies that elicit broadly neutralizing antibodies to pathogens, like HIV-1, that require exceptionally high levels of mutation.

Methods Summary

Vav-tTA transgenic mice were bred with Col1A1-tetO-H2B-mCherry mice to produce tTA–H2B–mCh mice21,22. B1-8hi, DEC205−/− and PAGFP transgenic mice were described previously10,19,28. Resting B cells from indicated mice were purified by magnetic cells sorting (MACS) and transferred intravenously into recipient mice that were primed 2–6 weeks previously by intraperitoneal immunization with 50 μg of OVA in alum. Adoptive transfer recipients were boosted subcutaneously with 25 μg of NP14–OVA. αDEC–OVA and αDEC–CS chimaeric antibodies were produced by transient transfection of 293T cells and were injected subcutaneously as indicated. Polyclonal GCs were generated by immunizing tTA–H2B–mCh mice intraperitoneally and subcutaneously with 50 μg and 12.5 μg of NP14–OVA or 12.5 μg and 6.25 μg of HIV-1YU2 gp120 protein in alum, respectively. DOX was administered by injecting 1.6 mg intraperitoneally and 0.2 mg subcutaneously and by supplementing the drinking water with DOX (2 mg ml−1) and sucrose (10 mg ml−1). Intravital imaging and photoactivation were performed as described previously10,18,29. Immunoglobulin sequence analysis was performed using MacVector 12.7 and IMGT/V-QUEST.

Online Methods

Mice

C57/BL6, B6.SJL, and Col1A1-tetO-H2B–mCherry mice were purchased from Jackson Laboratories. Vav-tTA transgenic mice that had been backcrossed to C57/BL6 were obtained from S. W. Lowe and crossed to Col1A1-tetO-H2B–mCherry mice to produce tTA–H2B–mCh mice21,22. B1-8hi, DEC205−/−, and PAGFP transgenic mice were described previously10,19,28. All experiments were performed with authorization from the Institutional Review Board and the IACUC at The Rockefeller University.

B-cell transfer and culture

Resting B cells were purified by forcing spleen tissue through a 40-μm mesh into complete RPMI media (Gibco) with 6% serum. Single-cell suspensions were purified by magnetic cells sorting (MACS) using CD43 beads, according to the manufacturer’s instructions (Miltenyi Biotec). Indicated cell numbers were transferred intravenously into recipient mice. To assess proliferation in vitro, resting B cells purified from tTA–H2B–mCh mice were labelled at 37 °C in 5 μM carboxyfluorescein succinimidyl ester. Labelled B cells were then stimulated with lipopolysaccharide and IL-4 for 72 h as previously described30. DOX was added to the cultures at 500 ng ml−1.

Immunizations and treatments

C57BL/6 or B6.SJL male recipient mice (6–8 weeks of age) were primed by intraperitoneal immunization with 100 μl containing 50 μg of OVA (Grade V, Sigma) precipitated in alum at a 2:1 ratio in PBS. Two to six weeks after priming, mice received adoptive cell transfers of indicated B cells and were boosted the following day with 25 μg of NP14–OVA (Biosearch Technologies) in hind footpads. Popliteal lymph nodes were collected for flow cytometric analysis. To generate polyclonal GCs, mice were immunized with 50 μg intraperitoneally and 12.5 μg in the hind footpads of NP14–OVA in alum. HIV-1YU2 gp120-specific GCs were induced by immunizing mice intraperitoneally and in hind footpads with 12.5 μg and 6.25 μg of gp120 protein, respectively, in alum. αDEC–OVA and αDEC–CS chimaeric antibodies were produced by transient transfection in 293T cells, as described20. 10 μg of chimaeric antibody in PBS was injected into footpads of mice at indicated time points, except in Fig. 3b in which 5 μg of chimaeric antibody was injected into footpads. Mice were administered DOX by intraperitoneal injection of 1.6 mg DOX (Sigma) in PBS and hind footpad injection of 0.2 mg DOX in PBS. Mice were maintained on DOX by adding DOX (2 mg ml−1) and sucrose (10 mg ml−1) to the drinking water for the indicated periods of time.

Flow cytometry

Lymph nodes were collected by forcing tissue through 40-μm mesh into complete RPMI media (Gibco) with 6% serum. Single cell suspensions were treated at 4 °C for 10 min with 1 μg ml−1 anti-CD16/32 (2.4G2, Bio-X-Cell) and then stained for 25 min at 4 °C. Anti-B220, CD38, CD86, CD45.1 and CD45.2 antibodies were from eBioscience. FAS, CXCR4, CD45.1, CD45.2, Igλ1-3, GL7, streptavidin–phycoerythrin and streptavidin–allophycocyanin were from BD Biosciences. Streptavidin–Alexa Fluor 488 was from Invitrogen. For cell cycle and S phase analysis, mice were injected intravenously with 2 mg BrdU (Sigma-Aldrich) and 1 mg EdU (Life Technologies) in PBS. Cells were then stained for surface antigens as described above and processed using an anti-BrdU-FITC kit (BD Biosciences) and Click-iT EdU-Pacific Blue kit (Life Technologies) according to manufacturers’ protocols. All samples were analysed on a BD Fortessa. GC B cells were gated as live/single, B220+, CD38 and FAS+. DZ and LZ GC B cells were further gated as CXCR4+CD86 and CXCR4CD86+, respectively. CD45.1 and CD45.2 allotypic markers were used to trace adoptively transferred B cells of genotypes B1-8hi DEC205+/+ (CD45.1+), B1-8hi DEC205−/− (CD45.1+CD45.2+), and B1-8hi DEC205+/+ tTA–H2B–mCh (CD45.2+) within either C57/BL6 (CD45.2+) or B6.SJL (CD45.1+) recipient mice.

Photoactivation

Intravital imaging and photoactivation were performed as described previously10,18,29. Anaesthesia was induced with 100 mg ketamine, 15 mg xylazine and 2.5 mg acepromazine per kg of body weight and maintained with 1.25% isoflurane in 100% oxygen. A double-edged razor blade was used to shave hind legs. Mice were restrained on a 37 °C stage warmer (BioTherm Micro S37; Biogenics) and an incision was made behind the knee to allow exposure of the popliteal lymph node. Once exposed, the lymph node was restrained with a metal strap and visualized with a microscope objective and a 40 °C objective heater. To label LZ-resident follicular dendritic cells, 1 μg of the red fluorescent protein tdTomato conjugated to NP was injected in hind footpads one day before imaging10,18. To photoactivate DZ B cells, B1-8hi PAGFP+ GC B cells were photoactivated external to NP–tdTomato-labelled FDCs in the LZ by scanning with a femtosecond-pulsed multiphoton laser tuned to 820 nm wavelength and imaged at 940 nm wavelength, as previously described10,18. Cell motility was monitored immediately following photoactivation to ensure cell viability. Incisions were sutured and mice were allowed to recover for 6 h before flow cytometric analysis. To determine the accuracy of this strategy, B1-8hi PAGFP+ GC B cells were photoactivated in the DZ of explanted lymph nodes and analysed immediately for surface phenotype. Imaging experiments were performed with an Olympus BX61 upright microscope (Olympus ×25 1.05 NA Plan objective), fitted with a Coherent Chameleon Vision II laser (Rockefeller University Bio-Imaging Resource Center).

Immunoglobulin sequence analysis

B cells were purified from spleen and lymph node tissues of immunized mice using CD43 magnetic beads (MACS). Single cell suspensions were stained as described above and sorted using a FACS Aria II (Becton Dickinson). GC B cells were gated as live/single, B220+CD38FAS+ (Fig. 4c) or B220+CD38FAS+GL7+ (Fig. 4d) and mChHi or mChLo. Genomic DNA was extracted from sorted cell populations and PCR was performed from DNA corresponding to 1,000–5,000 cells using Phusion HF (New England Biolabs). PCR products were gel extracted and cloned into Zero Blunt TOPO vectors (Invitrogen). For VH186.2 sequence analysis, GC B cells were further gated as Igλ+ and clones were amplified as described previously10,15. Immunoglobulin sequences were analysed using the IMGT/V-QUEST system to identify W33L mutations. JH4 intronic sequences were amplified using 5′-TCCTAGGAACCAACTTAAGAGT-3′ and 5′-TGGAGTTTTCTGAGCATTGCAG-3′ primers and 35 cycles with an annealing temperature of 57 °C and an extension time at 72 °C for 1 min. High-quality traces were analysed using MacVector 12.7 for base-pair mismatches and deletions as compared to the germline sequence. Both mismatches and deletions were counted as mutations. To calculate mutation frequency, the total number of mutations from all clones was summed and divided by the total number of base pairs analysed for each group. Mutation frequencies among different groups were compared and analysed for statistical significance by the χ2 test with Yates correction using Prism software v. 5.0 (Graphpad).