Frequency and Characterization of Phenotypic Ig Heavy Chain Allelically Included IgM-Expressing B Cells in Mice¹

In all studied vertebrates, B lymphocytes are known to express only one of the Ig H chain (IgH)³ alleles, a phenomenon called allelic exclusion (1). This is also the case for Ig κ (2) and λ L chain genes, which, in addition, show isotypic exclusion (3) (as a general rule, either κ or λ is expressed but not both). Furthermore, the λ L chain undergoes subtypic exclusion (4). Altogether, these mechanisms ensure lymphocyte receptor monospecificity, a requisite for cells to function as the selective units of the immune system.

Several models have been proposed to explain allelic exclusion ever since it was first observed. The mainstream model relies on two key observations: 1) mature B cells carrying two productive V_HDJ_H alleles (V_HDJ_H⁺/V_HDJ_H⁺) are rare (5, 6); 2) in mice carrying one wild-type IgH allele and another in which the membrane exon has been mutated by homologous recombination, allelic exclusion is incomplete (7). In other words, both alleles are thought to rearrange until the first to achieve a productive rearrangement expresses the corresponding protein in the membrane, signaling the shutdown of further somatic recombination in the locus.

In the last few years, it has been shown that the assembly of the pre-B cell receptor (pre-BCR), composed of a μ-chain and of two invariant chains, λ5 and V_{pre B} (8, 9), is required to achieve IgH allelic exclusion during early B cell development (10).

Despite these recent findings, two questions remain unanswered: 1) What are the intermediate steps from the pre-BCR assembly to the loss of somatic recombination at the IgH locus? and 2) What is the extent of phenotypic IgH allelic exclusion? In the present study we focus on the latter question.

H chain phenotypic allelic inclusion has been described in BCR transgenic (11, 12) and gene-targeted animals (13), but it is presently not consensual whether IgH dual expressors (DEs) in normal mice can be detected and, if so, at what frequency. In an attempt to clarify this issue, we evaluated how tight allelic exclusion is and tried, by looking at the eventual exceptions, to gain insights into this mechanism.

We show that IgH DEs are present in normal mice at a frequency in the order of 1 in 10⁴ splenic IgM⁺ B cells. Single cell (or clonal) analysis of the rearranged loci in DEs showed that we were, indeed, detecting truly allelically included cells at the genetic and phenotypic levels. The analysis of these rearrangements revealed that most cells carry a D-distal V_H (notably of the J558 family), whereas the other allele is often a D_H-proximal one. Interestingly, whereas the complementarity-determining region (CDR) 3 length in the population of DEs can be very heterogeneous, both IgH alleles for each of these cells tend to have a similar CDR3 length. These results show that phenotypic IgH allelic inclusion is extremely rare. We discuss two non-mutually exclusive hypotheses to explain the origin of IgH DE B cells in normal mice

C57BL/6 and BALB/c animals were purchased from Iffa-Credo (L’Arbresle, France) and crossed in SPF-barrier-free conditions at the Pasteur Institute animal facilities. BC8 and CB20 congenic strains were crossed with C57BL/6 and BALB/c animals, respectively. Animals used in experiments were between 4 and 12 wk old.

Abs used for FACS stainings and cell sorting were the following : FITC anti-IgM^a (clone DS-1), PE anti-B220, anti-IgM^b (clone AF6-78), biotin anti-B220, and streptavidin cychrome tricolor (all from PharMingen, San Diego, CA). For each animal, >10⁷ spleen cells were stained. Cells were isolated in a FACStar^Plus cell sorter (Becton Dickinson Immunocytochemistry Systems, Sunnyvale, CA). Dead cells in sorted samples were scored by propidium iodide (PI) and/or trypan blue exclusion. Fluorescence was measured with a FACScan flow cytometer (Becton Dickinson) using the Cell Quest 3.1 software (Becton Dickinson).

Sorted dual (B220⁺IgM^a+IgM^b+) or single (B220⁺ IgM⁺) expressors were cultured at 37°C in an atmosphere of 5% CO₂ for 4 or 5 days in OptiMEM (Life Technologies, Rockville, MD) medium supplemented with 10% FCS, 25 μg/ml LPS (Difco, Detroit, MI), 50 mM 2-ME, 50 U/ml penicillin-streptomycin in the presence of a feeder layer of 3000 rad irradiated S17 stromal cells. For limiting dilution conditions, cells were plated at 3, 1, 0.3, and 0.1 cells/well. Ninety-six replicates were plated for each cell concentration.

Cells were cultured for 4 or 5 days and then fixed in 95% ethanol on slides according to standard procedures and incubated with FITC anti-mouse IgM^a and biotin anti-mouse IgM^b revealed by streptavidin-conjugated Texas Red (Southern Biotechnology Associates, Birmingham, AL). On an Axiophot Zeiss (Oberkochen, Germany) fluorescence microscope, we analyzed 0.05% p-phenylenediamine-buffered glycerol-mounted slides. For the estimation of the DE frequency in LPS cultures 10³-2 × 10⁴ splenocytes (per animal) were sorted as B220⁺IgM^a+IgM^b+. Background was standardized by staining cells from homozygous mice with both Abs. After culture, a minimum of 350 cells was scored per animal, and the total number of cells analyzed was >15,000. The number of cells analyzed from cultures of splenocytes sorted as B220⁺ cells was >5,000.

The %DE_{in vivo} was calculated based on the frequency of ex vivo stained splenic cells in the dual DE gate (%DE_sort) and on the frequency of DE in cultures of LPS-stimulated splenocytes sorted as DE (DE_cult), as estimated by intracytoplasmic stainings. Formally, %DE_{in vivo} = DE_sort × DE_cult × (f_sing + 2 × f_doub) × (100 − % dead cells in DE_sort) × (1/DE_popf)/100, where f_sing (f_doub) is the fraction of singlets (doublets) in the DE-sorted cells and DE_popf is the fraction of the DE population FACS profile distribution included in the DE gate. f_sing and f_doub were estimated by analyzing cells on the microscope after sorting; the theoretical DE population distribution in the FACS profile was calculated from the distributions of IgM^a+ and IgM^b+ single expressors, assuming that both alleles contribute equally for the membrane IgM density (13), similar to that of single expressors; dead cells are estimated by PI staining; since, in all experiments, (f_sing + 2 × f_doub) × (100 − % dead cells in DE_sort)/100 × (1/DE_popf) was estimated to be < 2, in Table I %DE_{in vivo} ≅ %DE_sort × DE_cult.

Supernatants from limiting dilution cultures were tested for the presence of IgM at day 10 of culture. Briefly, 96-well flat-bottom microtiter plates were coated overnight at 4°C at a concentration of 5 μg/ml with goat anti-total mouse μ-chain (Sigma, St. Louis, MO), anti-IgM^a (RS3.1), or anti-IgM^b (MB86). After blocking for 2 h with 1% BSA in PBS, wells were incubated with culture supernatants for 3 h at 37°C. Anti-μ HRP-labeled Ab (Jackson ImmunoResearch Immunotech, France) was then added, and plates were incubated for another hour at 37°C. Bound Abs were revealed with 2,2′-azino-bis(ethylbenzthiazoline-6-sulfonic acid (Sigma) in 50 mM phosphate citrate buffer containing 0.03% sodium perborate (Sigma). The absorbency was determined at 405/620 nm with an ELISA plate reader (Dynatech MR5000, Chantilly, VA). Washes between incubations were done with 0.1% Tween 20 in PBS. A well was considered to be positive when the absorbency measured was greater than the mean of the background plus three times the SD of the mean. Background staining for both anti-allotypic Abs was determined by repeating the ELISA reaction with sera from IgM^a or IgM^b mice.

DE double-sorted cells were diluted in 5% glucose and picked individually into Terasaky plates under a magnifying lens with a heat-extended capillar connected to a mouth piece. Alternatively, cells double sorted as DEs were set in limiting dilution conditions as previously described, the resulting clones being picked individually from Terasaky plates. In both cases, (V_H)DJ_H rearrangements were amplified as previously described (6), with minor modifications. After a preliminary study, it was shown that the 5′ V_H primers that would account for most rearrangements were: V_HA, 5′-GCGAAGCTTARGCCTGGGRCTTCAGTGAAG-3′; V_HB, 5′-GCGAAGCTTCTCACAGAGCCTGTCCATCAC-3′; V_HE, 5′-GCGAAGCTTGTGGAGTCTGGGGGAGGCTTA-3′; and V_HF, 5′-GCGAAGCTTWCTGGAGGAGGCTTGGTGCAA-3′.

Only these V_H primers were used in this analysis. The DJ_H-specific primers were: Q52, 5′-ACGTCGACGCGGACGACCACAGTGCAACTG-3′; and DFS, 5′-ACGTCGACTTTTGTSAAGGGATCTACTACTGT-3′. The Dye Terminator Sequencing (Perkin-Elmer, CA) kit was used to determine the sequence of (V_H)DJ_H Gene Clean II Kit (Bio101, Vista, CA) purified PCR fragments, according to the instructions of the manufacturers. Sequences were obtained in a ABI 370A DNA sequencer (Applied Biosystems, Foster City, CA). The efficiency of amplification of a single allele was 0.28 when single cells were used and 0.50–0.75 (depending on the experiment) when clones were used. V_H families were determined by BLAST search using partial V_H sequences.

To estimate the frequency of DEs in mature B cells, we took advantage of the IgH allotypic difference between C57BL/6 and BALB/c mice. As a first attempt to determine this frequency, splenocytes from C57BL/6 × BALB/c animals (B6CF1) were sorted as IgM^a+IgM^b+B220⁺ (Fig. 1,A). When the purity of the sorted population was evaluated, it became clear that the proportion of single expressors was still high (Fig. 1,B). To understand the failure in obtaining a pure population of DEs, 1:1 mixtures of BALB/c and C57BL/6 spleen cell populations that lack by definition IgM^a+IgM^b+ cells were analyzed. Fig. 1 C shows the results of flow cytometry analysis in which the frequency in the DE gate is plotted against the FSC-H. As the upper limit of the forward scatter (FSC-H) increases, single IgM^a/single IgM^b doublets and other aggregates of more than two cells start being included in the analysis, which could explain the increase in the putative DE frequency as a function of FSC-H, observed with both B6CF1 and a 1:1 mixture of splenocytes. On the other hand, the increased frequency of putative DEs in B6CF1 cells as compared with that of the BALB/c and C57BL/6 cells mixture could be due to: 1) doublets that would be in higher frequency in the B6CF1 suspension because cells came from the same tissue; 2) nonspecific binding of serum IgM to the surface of B lymphocytes; and 3) presence of real DEs.

To overcome these problems, we reevaluated the single or double IgM expression in the sorted population after in vitro stimulation with LPS in bulk cultures. At day 4 of culture, double-color intracytoplasmic stainings were performed, and cells were analyzed in a fluorescence microscope. DEs could be readily distinguished from single expressing cells (Fig. 2), and the results for B6CF1 splenocytes pooled from three independent experiments are shown in Table I. The initial percentage of B220⁺IgM⁺ in the DE gate (%DE_sort) was then corrected by the frequency of DEs in culture (%DE_cult) based on the intracytoplasmic stainings (see Table I and Materials and Methods). We found that the frequency of phenotypic IgH allelically included cells in normal mice is in the order of 1 in 10⁴ IgM⁺ splenic B cells. Finally, in a single independent experiment, IgH heterozygous mice with the C57BL/6 (C57BL/6 × BC8) or BALB/c (BALB/c × CB20) backgrounds were directly compared. We found that the frequency of DEs is not statistically different in these two strains, showing no influence of the mouse strain genetic background in this trait.

It has been suggested that the frequency of an Ag responding cell with a particular reactivity is in the order of 1 in 10⁴. Thus, it could be argued that, instead of DE cells, we were detecting IgM^a or IgM^b single expressors reacting with a given epitope on the anti-IgM^b or anti-IgM^a Abs, respectively. It follows that the formal demonstration that the cells we define as DEs are really phenotypic allelically included relies on the analysis of V_HDJ_H rearrangements. After sorting, single expressors still account for more than 90% of cells in the population sorted as IgM^a+IgM^b+, as shown in Table I. Therefore, to characterize rearrangements at the single cell level, further enrichment was necessary. Sorted cells were expanded by LPS stimulation for 3.5 days, stained, and subjected to a second round of cell sorting. Fig. 3 shows the correspondent dot plots at each step. The frequency of cells in the DE gate after sorting and culture was 6.8 ± 0.4 for cells previously sorted as DEs and 1.3 ± 0.2 for cells sorted as B220⁺ (in three independent experiments). The value of 6.8% is above the one found by intracytoplasmic staining and is likely to be an overestimation (see below). Nevertheless, it suggested that, by two consecutive sortings, enrichment for DEs could be achieved. As seen in Fig. 3 B, cells initially sorted as DEs using a broad gate resolve into three major populations of single IgM^a, IgM^b, and DEs. The DE population has a diagonal shape (representing around 1.5% of the cells, which is consistent with the data from intracytoplasmic stainings), indicating that cells with different levels of expression for both alleles, if they are present after the first sorting, represent a minority.

To confirm the enrichment for DEs, cells that were submitted to a second round of sorting (this time using a gate stricter than the DE gate) were LPS stimulated under limiting dilution conditions, and 5 days later supernatants from individual wells were tested by ELISA for the presence of IgM^a and IgM^b (immediately after sorting microscopic inspection showed a frequency of doublets below 10%). Forty-five percent of tested IgM⁺ supernatants were positive for both IgM^a and IgM^b; likewise, when cells double sorted as DEs were set again in culture for 2 additional days and then intracytoplasmic stained, 39% of the cells expressed both alleles. This result confirmed that enrichment for DEs had been achieved and that the frequency of 6.8% in the DE gate is, at least, a 2-fold overestimation.

The estimation of the frequency of DEs was based on the postulate that dual and single expressors respond equally to LPS stimulation. We, therefore, directly assessed the frequency of LPS responders and the clonal size after double sorting of B220⁺IgM⁺ and B220⁺ IgM^a+IgM^b+ cells. The frequency of DE responders was found to be 1 in 1.5 cells, similar to the one of single expressors (1 in 1.5). This frequency was 30% lower than that of the respective starting populations after one round of sorting, found to be 1 in 1 (14). This is probably due to the fact that, when cells were sorted for the second time, they were less able to proliferate than ex vivo sorted cells, because they had already been in culture for 3.5 days. Clone size of LPS-stimulated cells from both double sorted populations was analyzed by seeding one cell per well in Terasaky plates and scoring numbers of cells from each clone after 4 days. No statistically significant differences were observed since the clone size was 31.5 ± 24.9 and 33.0 ± 27.5 for dual and single expressor populations, respectively, in more than 50 clones scored for each population.

The results described above prompted us to characterize the V_HDJ_H rearrangements on cells from the population double sorted as DE. IgH-rearranged alleles were PCR amplified directly from micromanipulated double sorted single cells or from double sorted derived clones that were cultured in limiting dilution conditions. The set of primers used recognizes all D_HJ_H, and the majority of V_HDJ_H rearrangements with V_H genes from the 7183, Q52, S107, X24, 3660, J606, MRL-DNA4, and J558 murine V_H families and were previously described (6). Amplified products were sequenced directly. Even if technical limitations did not allow the detection of both rearrangements in many cells, the analysis of the simple V_HDJ_H sequences already suggested that, although not pure, the population double sorted as IgH DEs had a high frequency of real DEs. Phenotypic allelically included cells have necessarily two V_HDJ_H⁺ rearrangements, whereas single producers can be V_HDJ_H⁺/V_HDJ_H⁻ or V_HDJ_H⁺/DJ_H. For any given V_H family, V_HDJ_H⁺ alleles should have identical chances to be PCR amplified over V_HDJ_H⁻ alleles. The finding that 8 out of 11 V_HDJ_H sequences with D-proximal V_H (Q52 and 7183) were V_HDJ_H⁺ and that all 23 V_HDJ_H with V_H other than D-proximal ones were V_HDJ_H⁺, can be explained only if a large proportion of cells in the population double sorted as IgM^a+IgM^b+ were V_HDJ_H⁺/V_HDJ_H⁺ (in a population of single expressors, considering all V_HDJ_H, 30% are expected to be V_HDJ_H⁻).

The formal demonstration that we were dealing with genetic and phenotypic allelically included cells comes from the observation, shown in Table II, that 13 out of the 22 cells (clones) double sorted as DEs, for which both alleles were successfully amplified by PCR, are allelically included at the genetic level, i.e., they carry two productive V_HDJ_H rearrangements (the remaining cells were 5 V_HDJ_H⁺/V_HDJ_H⁻ and 4 V_HDJ_H⁺/DJ_H). The fact that we found 9 single expressors out of 22 cells (41%) after a second sorting for DEs rules out the possibility that, once in culture after the first sorting, DEs would down-regulate one of the alleles and become phenotypic allelically excluded.

In the double-sorted population, there is a close agreement between the DE frequency estimation based on intracytoplasmic staining, on ELISA, and on the description of IgH rearrangements (39%, 45%, and 59%, respectively). Altogether, these results consolidate our estimation of the in vivo double expressor frequency.

The V_H usage in DEs is summarized in Table II. A J558 V_H is typically used in one of the alleles whereas in the other allele any V_H can be used, including D-proximal V_H in six of the cells.

With respect to N and P additions and D element usage, DEs do not appear to have any particular feature. However, 3 (possibly only 2) out of the 19 alleles (16%; possibly only 11%) for which a D element reading frame (RF) can be assigned use RF2, which suggests an increase in the usage of this RF, which is around 3% immature lymphocytes (17, 18), although a larger sample would be of interest here. The bias for J_H4 usage is likely to result from a technical artifact since PCR products from rearrangements using J_H4 are smaller and thus easier to amplify. Interestingly, whereas the CDR3 length (in base pairs) can be very heterogeneous (43.4 ± 6.2), IgH alleles from the same cell tend to have a similar CDR3 size, which is illustrated by the CDR3 length correlation that was found for the pairs of V_HDJ_H⁺/V_HDJ_H⁺ (Fig. 4).

To our knowledge, the experiments presented here provide the first consistent effort to determine the frequency and characterize phenotypic IgH allelically included cells in a genetically nonmanipulated animal model. Here, we show that IgH DEs are in the order of 1 in 10⁴ of IgM⁺ splenic B cells. This result was based on intracytoplasmic staining of cells, previously sorted as DEs, that were then stimulated with LPS. Given the nature of this technique and the criteria that we used to score a cell as DE, this result should be considered as a minimal estimate. On the other hand, the estimation obtained by FACS analysis of similar cultures was about 6-fold higher and should be considered as an overestimation. This is due to the fact that by FACS analysis we are reiterating all the sources of artifact that influenced the first sorting. Furthermore, even when DE-sorted cultures are resorted using a DE gate stricter than the one that gives 6.8%, the frequency of real DEs is still below 50%, as judged by three independent methods. It should be noticed that, by intracytoplasmic staining of LPS-stimulated splenocytes, Kitamura and Rajewsky (7) found a frequency of 0.3% for IgH DEs. Assuming that 1.7% of IgM⁺ splenocytes are initially sorted as DEs and that the DE gate covers >50% of the theoretic DE distribution (see Materials and Methods), Kitamura and Rajewsky’s estimation would imply a DE frequency of >8.8% in cultures of these cells. This was not observed, and the discrepancy could be explained by differences in the criteria used to score a cell as DE by intracytoplasmic staining. Our findings are consistent with the notion that IgH allelic exclusion is an extremely tightly regulated mechanism, even when compared with that of other Ag receptor genes known to be stringently phenotypically excluded, such as the TCR β (19, 20) and δ (21) chains and the Ig κ (2, 22, 23, 24) and λ (4) L chains.

The finding that phenotypic allelically included cells do exist but are kept at very low frequencies raises some questions concerning the selective pressures that shaped allelic exclusion. It is now clear that allelically included cells can go through B cell development and mature (our unpublished observations; Ref. 13). Nevertheless, one question remains: would the vertebrate immune system be less efficient if a high proportion of cells would be allelically included? Assuming that the probability of negative selection of a newly formed lymphocyte (p) is a function of the number of receptors the cell can display (for example, if p for IgH single expressors is x, for IgH DEs it will be 2×), negative selection would be more severe on a DE population. However, in terms of the efficiency of lymphocyte output and in the absence of an estimation of x, it is difficult to argue that this difference is, indeed, relevant. At the level of B cell responsiveness, although some observations indicate that DEs become single expressors during the course of an affinity maturation response (3, 25), it is not known whether a single expressor would initially be recruited preferentially over a phenotypic allelically included cell. Additionally, we lack data on the efficiency of triggering in cells with a 2-fold (or higher) decrease in membrane receptor density. In contrast with the idea that allelic exclusion was selected per se, one could still consider it as an indirect consequence of other selective pressures. For instance, somatic rearrangement and extensive diversification at the junctional level in the IgH locus lead to two nonproductive rearrangements in a substantial fraction of B cell precursors (7, 17). In the absence of simultaneous allelic rearrangement, the selective pressure to recruit cells for further differentiation as soon as one productive rearrangement occurs could account for the allelic exclusion we observe.

Although the frequency of phenotypic IgH allelically included cells is extremely low, given that an adult mouse has 10⁸ B cells, at any time the animal has to cope with 10⁴ DEs. These cells express two different IgH variable domains from which one could be self reactive, inducing, once activated, an autoimmune condition. This is unlikely because, as previously discussed, before achieving maturation, DEs should be checked for both receptors. Presumably, negative selection is ensured even if only half of the receptors in one cell are self reactive. If a parallel with T cell physiology is allowed, this rather simplistic view does not entertain the possibility of some autoimmune hazard resulting from the activation, once in the periphery, of passenger autoreactive receptors in DEs that may have escaped deletion due to low levels of surface expression (26).

Presently, several non-mutually exclusive hypotheses can be envisaged to explain the origin of IgH DEs. One very simple interpretation is to consider these cells as rare escapees that would leak through the tight mechanism of allelic exclusion. More elaborate scenarios include: 1) de novo rearrangement in the periphery; and 2) synchronous V_H to (V_H)DJ_H rearrangement on both alleles occurring in a small fraction of cells. The former scenario is unlikely to explain our results. In fact, although B cell somatic recombination activity has been reported in germinal centers (27), it remains to be formally shown that a silenced IgH allele can rearrange and be expressed in the periphery; furthermore, B cells in germinal centers lose IgM surface expression and would not be detected in our assay. On the other hand, the synchronous rearrangement hypothesis would conciliate the existence of DEs with the mainstream feedback model of allelic exclusion. By “synchronous,” we mean that both alleles rearrange within the time window defined by the first rearrangement and the decay in the potential of recombination which would compromise rearrangement in the second allele. The length of the N region has been interpreted as a measure of the activity of TdT at the time of rearrangement (28). Assuming that TdT levels change in ontogeny, similar N region lengths in both alleles could be interpreted as a signature of synchronous rearrangement. In the described DEs, this is not observed, either in D-J_H or in V-DJ_H junctions. What about the D RF usage? The Dμ protein (transcribed and translated from a RF2 DJH, for all D elements except Q52) is thought to be involved in the shutdown of somatic recombination (29, 30), consequently driving the cell to an endpoint. This explains the RF2 counter selection observed in mature B cells. However, it is conceivable that, if cells rearrange D to JH at the same time and immediately after V_H rearranges in both alleles, the cell would not have time to accumulate Dμ, and RF2 counter selection should be diminished. Although 16% of D segments were found to be in RF2 (as compared with 3% (17) previously found for single expressors), we do not valorize this finding due to the low number of sequences (2) and therefore we lack altogether support for the synchronous rearrangement hypothesis.

One last remaining possibility is based on a report in which genetic allelically included cells were described (18). It was proposed that these allelically included cells arise when the second allele is given a chance to recombine because, although the rearrangement in the first allele was productive, the correspondent H chain failed to associate with the surrogate L chain. Furthermore, it has been shown that the pre-BCR mediates a shift in the B cell V_H gene repertoire, usage representation (31). Normally, pre-BCR formation is essential for B cell progression. In the c-kit⁺, cytoplasmic μH chain⁺ pre-B cell compartment, D-proximal V_H genes are preferentially expressed, but, notably, most μH from the Q52 family and the 81X gene fail to form a pre-BCR. Consequently, these elements are underrepresented in the V_HDJ_H⁺ pool from large pre-BII cells, which necessarily express the pre-BCR.

Could these pre-B genetic (but not phenotypic) allelically included cells correspond to a premature developmental stage of the mature DEs? The fact that close to one third of the DEs we found in the periphery use one Q52 V_H in one of the alleles argues in that direction. In fact, this observation indicates that these cells failed allelic exclusion because the chain from the first allele to rearrange V_HDJ_H⁺ failed to form a pre-BCR. If the second allele rearranges its V_H productively and the correspondent μ-chain associates with the surrogate L chain, then these cells can progress in differentiation.

We, therefore, propose a scenario where, upon conventional L chain expression, the μ-chain that did not pair to form the pre-BCR will be “rescued” in a fraction of genotypic allelically included cells and expressed at the surface, just like the μ-chain that first paired with the surrogate L chain. This fraction is likely to be very small, following the notion that μ-chains that first failed to associate with the surrogate L chain are less prone to pair with conventional L chains (32, 33). This would largely explain the difference between the frequency we estimated for phenotypic allelically included cells and the ≈2.5% (18) calculated for genetic allelically included B lymphocytes. Although based on only 2 VDJ⁺/VDJ⁺ cells out of 39 carrying two VDJ rearrangements, this value suggests that the frequency of cells carrying two productively rearranged VDJ alleles is higher than the one for DEs. In light of this hypothesis, the CDR3 length correlation would be interpreted as an additional physical constraint imposed by the conventional L chain, to achieve IgH surface expression of both alleles.

We thank Anne Louise and Mathias Haury for help in the use of the cell sorter. We also thank Prof. Guy Bordenave for giving us the BC8 and CB20 congenic strains. We are grateful to António Bandeira, Jocelyne Demengeot, and Pablo Pereira for helpful discussions and critical reading of the manuscript and to Ken McElreavey for polishing the English.