|Home | About | Journals | Submit | Contact Us | Français|
Studies in autoantibody transgenic mice have demonstrated receptor editing rearrangements at antibody heavy and light chain loci. However, the physiologic role of heavy chain editing (VH replacement and rearrangement on the second allele) has been called into question. It is unclear if additional rounds of heavy chain rearrangement are driven by BCR specificity. Here we analyze the manner in which B cells undergo additional heavy chain (H chain) rearrangements in an anti-DNA heavy chain knock-in mouse, B6.56R. We find that rearrangements in 56R+ B cells tend to involve the D gene locus on both alleles and the most JH-proximal VH gene segments on the endogenous allele. As a result, some B cells exhibit V(D)J rearrangements on both H chain alleles, yet allelic exclusion is tightly maintained in mature 56R B cells. As B cells mature, a higher proportion expresses the non-transgenic heavy chain allele. Rearrangements on both H chain alleles exhibit junctional diversity consistent with TdT-mediated N-addition, and TdT RNA is expressed exclusively at the pro-B cell stage in B6.56R. Collectively, these findings favor a single, early window of H chain rearrangement in B6.56R that precedes the expression of a functional BCR. B cells that happen to successfully rearrange another heavy chain may be favored in the periphery.
B cells exhibit ordered gene rearrangements at immunoglobulin loci, starting with heavy chain D to J rearrangement, followed by VH to DJ and subsequently, light chain rearrangement. Based on immunophenotyping studies, fewer than one in ten thousand B cells expresses more than one kind of H chain (1). This restriction is a property referred to as allelic exclusion. The mechanistic basis for H chain allelic exclusion has been argued for decades (reviewed in (2)). According to the stochastic model, the probability of successful (productive or in-frame) H chain rearrangement is low. The likelihood of successful H chain rearrangement is reduced by the presence of multiple reading frames (including the varied functionality of D reading frames even when the overall rearrangement is in frame) and VH pseudogenes (3-5). If the probability of rearranging successfully on one H chain allele is low, the probability of rearranging successfully on both alleles in the same cell is obviously lower still. Coupled with a short time window for rearrangement and/or a high “crash factor” if defective or autoreactive rearrangements result in clonal deletion, a peripheral repertoire of B cells with mostly one H chain allele fully rearranged can be obtained (4, 6). According to the ordered or regulated model, there is sequential accessibility of the H chain gene segments (starting with D segments, followed by D-proximal and later distal VH gene segments (7)), and loci, through chromatin modifications including methylation and differential nuclear localization (8-10). As further quality control, feedback inhibition prevents rearrangement on the second allele if the first allele rearranges successfully. Alternatively, or in addition, there could be toxicity associated with the production of H chains in excess of L chains (11); however, in at least some transgenic systems where H chain allelic inclusion is enforced, B cells seem to survive without complications (12).
Compelling evidence for feedback inhibition of H chain rearrangement derives from antibody transgenic mice, some of which display very low levels of endogenous H chain rearrangement (13). On the other hand, other H chain transgenic and knock-in mice display high levels of endogenous H chain rearrangement and/or inactivation of the site-directed transgene by further rearrangement (14-16). Using an allelic series of H chain knock-ins (wherein the promoters and sites of integration are identical, but the VH sequences differ), several groups have demonstrated different frequencies of transgene inactivation and/or ongoing H chain rearrangement (15-18). These findings indicate that the sequence of the VH is of critical importance and have engendered the idea that rearrangements occur in response to BCR specificity (19, 20). An alternative explanation for the varying frequencies of H chain rearrangements in different H chain transgenic models is that B cells with productive endogenous H chain rearrangements are expanded to a different extent in each model. BCR specificity-driven H chain rearrangement predicts that H chain rearrangements occur after L chain rearrangement and BCR surface expression. Conversely, if H chain rearrangements arise “by mistake,” H chain rearrangements should occur at the pro-B cell stage, before L chain rearrangement.
To investigate the manner and timing of ongoing H chain rearrangement, we analyzed the rearrangements on the transgenic and endogenous H chain alleles of B6.56R mice. The B6.56R model has three important properties that can be exploited to study H chain rearrangement. First, the H chain is well characterized and is known to produce anti-dsDNA antibodies with all but a handful of L chains (17, 21, 22). Second, there is a clear precursor-product relationship because all B cells start out with the rearranged 56R H chain DNA. Third, additional H chain rearrangements can be detected phenotypically and genetically.
By characterizing B cells with additional H chain rearrangements from B6.56R mice, we find evidence in support of early H chain rearrangement. In B6.56R mice that are heterozygous for the 56R allele, we observe D gene rearrangements on the 56R allele, as well as VH rearrangements to D invaded sequences. VH gene rearrangements on the endogenous allele in 56R+ B cells are out-of-frame and appear to involve predominantly JH-proximal VH genes; on the other hand, rearrangements on the endogenous allele in 56R- B cells involve a wider diversity of H chains, including distal VH genes. We find genetic evidence of H chain rearrangements on both alleles in peripheral B cells but infrequent phenotypic allelic inclusion (surface expression of both H chain alleles), suggesting that B cells with productive H chain rearrangements on both alleles are produced, but counter-selected as they mature. H chain rearrangements on both alleles in B6.56R are characterized by N-additions at the rearrangement junctions, suggesting that rearrangement takes place at the pro-B cell stage, when the enzyme TdT is normally expressed. We demonstrate that TdT mRNA is expressed at highest levels during the pro-B cell stage in B6.56R B cell development, suggesting that most H chain rearrangements occur at this early stage in B6.56R mice. Based on these data and other findings in the literature, we propose that initial H chain rearrangement in B6.56R B cells represents a stochastic rearrangement of D genes and proximal VH genes, resulting in inactivation of 56R, or rearrangement of the endogenous allele. Additionally, because H chain rearrangement appears to occur mainly if not exclusively at the pro-B cell stage in B6.56R B cells, the autoreactive specificity of the BCR is unlikely to promote further H chain rearrangement. Rather, B cells that happen to have inactivated 56R are likely to have a selective advantage.
The generation of 56R site-directed transgenic mice and their backcrossing to the C57Bl/6 (hereafter B6) background have been described previously (19, 22). B6 and 129/Sv mice were purchased from Jackson Laboratories (Bar Harbor, ME). The 56R transgene was detected by PCR amplification of tail DNA (23). Mice used for these experiments were analyzed in the heterozygous state (56R on one allele and the endogenous H chain locus on the other allele) and bred and maintained at the University of Pennsylvania School of Medicine under an IACUC approved protocol.
DNA was isolated from spleen using a DNeasy Kit (Qiagen Inc., Valencia, CA). H chain rearrangements were amplified using several degenerate forward primers, MH1- MH7, as described in ref. (24). In addition, the following VH specific forward primers were designed for this analysis: 7183 5′- ACC ATT AGT AGT GGT GGT AG- 3′ (Ta=58°C); 3609N: 5′-GTA ATT ACA GTC AAA TCT GAT AA-3′(Ta= 57°C), MOPC21: 5′-ATT AGT AGT GGC AGT AGT AC-3′(Ta= 58°C), VH1: 5′-GAG ATT TAT CCT GGA AGT GGT AATAC-3′ (Ta= 53°C), in conjunction with the following JH reverse primers: JH4KI: 5′- CCA GTA GTC CAT AAC ATA GG- 3′, JH4endog: 5′- CCA GTA GTC CAT AGC ATA GTA- 3′, JH2: 5′- CTG TGA GAG TGG TGC CTT G- 3′, JH1: 5′- CAG ACA TCG AAG TAC CAG- 3′. All PCRs were performed with 1μM forward and reverse primers, 0.5mM dNTPs, 1U AmpliTaq Gold, 1.5mM MgCl2 in PCR buffer (Roche Applied Sciences, Indianapolis, IN). Amplification conditions: 94°C, 10 min.; [(94°C, 30 sec), (Ta defined above, 30 sec), (72°C, 30 sec)], 40-cycles; 72°C, 10 min. PCR products were cloned into a pCR 2.0 vector (Invitrogen Corp., Carlsbad, CA) and sequenced with the M13F sequencing primer. H chain rearrangements from hybridomas were amplified with the above degenerate primers and JH4 primers and sequenced. Rearrangement of D genes to 56R were PCR amplified as described previously (19). Presence of the 3′ end of the 56R allele was detected by PCR using the following primers: DJ KI F (5′-GAG GAG TAA ATA TTC CTA TGT- 3′) and JC intron R (5′- ATC TTC TTC AAA TGA GCC TCC- 3′) with Ta of 59.5°C and similar mix and cycling conditions to the other VH PCRs.
Spectratyping (PCR with fluorescently labeled primers followed by capillary electrophoresis for size separation) was performed to evaluate the VH gene usage of rearrangements on the endogenous allele in different B6.56R B cell populations. B220+CD93- splenic B cells were sorted on the basis of IgM allotype expression into IgMa+ and IgMb+ pools. For comparison of proximal and distal VH usage on the endogenous H chain allele, genomic DNA from each pool was amplified using three different VH specific primers in combination with the JH2-FAM primer: 7183.2.3: 5′-GCC ATT AAT AGT GAT GGT GGT AGC-3′ (Ta= 60°C); 7183.9.15: 5′- TAG TGG TGG TGG TGG TAA C-3′ (Ta= 56°C), VH6: 5′- CAA ATA AGA TTG AAA TCT GAT AAT TAT GC- 3′(Ta= 57°C). The 7183.2.3 primer was HPLC purified to reduce notching of amplicons. PCR products were resolved by capillary electrophoresis on the ABI 3100 instrument (Applied Biosciences, Foster City, CA). Peak sizes were interpolated using a ROX ladder (ABI). Capillary electropherograms were generated and analyzed using ABI Genotyper 3.7 or ABI Peak Scanner Software v1.0. The DNA amount was titrated so that fewer than 10 peaks were obtained per spectratype to facilitate more accurate counting of peak numbers.
Reverse transcription of RNA from sorted cells was performed using ABI TaqMan RT Reagents (Cat# N808-0234) with random hexamer primers per the manufacturer's instructions. RT-PCR on cDNA was performed with TaqMan Gene Expression Assays for β-actin (Mm 02619580_g1), Rag1 (Mm 01270936_m1), and TdT (Mm 00493500_m1) under conditions recommended by ABI on a Roche Light Cycler 480 real time instrument. Data were analyzed by the comparative threshold method (25), with β-actin to control for cDNA content and thymus from a B6 mouse as the reference sample.
Three splenic hybridoma panels, 680S, 680LPS, and 56R.old, generated for a previous study (22), were further characterized in this study. In addition, two new hybridoma panels (IgMb sort, IgMb LPS) were generated for this study using the method described in ref. (26): “IgMb sort” was a “spontaneous” fusion of B220+IgMb+ sorted splenocytes from two 6 month old B6.56R mice, and “IgMb LPS” was an LPS stimulated fusion from unsorted splenocytes of a B6.56R mouse. For the spontaneous fusion, splenocytes were removed from the mouse and, without further manipulation or stimulation in vitro, were fused immediately to a myeloma cell line. The IgMb LPS panel was screened for secretion of IgMb antibodies by ELISA, and positive clones were further characterized and screened for IgMa antibody secretion by ELISA as described in (26). The coating antibody for allotype ELISAs was goat-anti-mouse IgM-UNLB (Southern Biotech, Birmingham, AL), primary antibodies: Biotin-anti-mouse IgMb (1/5000, BD Pharmingen, San Diego, CA) or Biotin-anti-mouse IgMa (1/5000, BD Pharmingen); followed by SA-AKP (1/2000, BioLegend, San Diego, CA) and pNPP (Sigma-Aldrich, St. Louis, MO).
Bone marrow fractions were resolved by staining with the following antibodies: FITC or PE anti-IgMa(DS-1, BD Pharmingen), PE anti-IgMb(AF6-78, BD Pharmingen), AlexaFluor750 or PE-Cy7 anti-B220 (RA3-6B2, eBioscience), PE-Cy7 or APC anti-CD93 (AA4.1, eBioscience, San Diego, CA); biotin anti-CD43 (S7, BD Pharmingen), APC-Cy5.5 anti-CD19 (6D5, BioLegend), followed by SA-APC (BD Pharmingen) after red cell lysis (ACK Lysing Buffer, BioWhittaker, Walkersville, MD). Spleen fractions were resolved by staining with the following antibodies: FITC or PE anti-IgMa, PE anti-IgMb, APC or PE-Cy7 anti-CD93, PE-Cy7 anti-B220, APC anti-CD21(7G6, BioLegend), biotin anti-CD23 (B3B4, eBioscience), APC-Cy5.5 anti-CD19, followed by SA-APC-Cy7 (BD Pharmingen). Intracellular staining was performed following fixation and permeabilization with Caltag Laboratories Kit (#GAS-004) per the manufacturer's recommendations. Cells were stained with Rabbit anti-TdT, 20μg/mL (SuperTechs, Bethesda, MD), followed by 1:25 dilution of FITC anti-Rabbit IgG (SuperTechs, Bethesda, MD). Dead cells were excluded by EMA staining for the TdT analysis and 7-AAD was used for live-dead discrimination in non-permeabilized cells. Cells were analyzed on a FACSCanto or LSR II instrument (Becton Dickenson, Franklin Lakes, NJ). Data were analyzed using FlowJo Version 8.2 (Tree Star Inc, Ashland, OR).
The heterozygous B6.56R mouse has an in-frame 56R H chain rearrangement inserted in the JH locus on one allele (the 56R allele) and a normal B6 H chain locus on the other (the endogenous allele, Fig. 1a) (19). The two H chain alleles of B6.56R can be differentiated phenotypically using anti-allotype antibodies: anti-IgMa binds to antibodies expressed by the 56R allele and anti-IgMb binds to antibodies expressed by the endogenous allele. On the genetic level, rearrangements on the 56R allele and the endogenous allele can be analyzed in several ways. Rearrangements on the 56R allele can be distinguished from rearrangements using JH4 on the endogenous allele due to two base pair changes in the 56R JH4 gene compared to the germline JH4. In addition, the endogenous H chain allele has the full complement of JH gene segments (JH1-JH4), whereas the 56R allele only has JH4. Thus primers situated in JH1-JH3 will only amplify the endogenous allele.
As shown in Fig. 1b, the 56R allele can undergo three types of H chain rearrangement: 1) invasion by an upstream VH gene into the 56R cryptic heptamer (hereafter referred to as VH replacement), 2) invasion by an upstream DH gene into the 56R cryptic heptamer (hereafter referred to as D invasion), and 3) D invasion into the 56R cryptic heptamer and rearrangement of an upstream VH gene (19). Although each of these rearrangements has been described previously, their relative frequency, how and when during B6.56R B cell development they take place have not been well characterized.
To gain insight into the frequency of H chain rearrangement in B6.56R mice, different B cell subsets were analyzed for the expression of IgMa and IgMb. Fig. 2a shows the staining pattern of IgMa and IgMb antibodies in different bone marrow and splenic B cell subsets in a representative B6 (top row) and a heterozygous B6.56R mouse (bottom row). The B6 mouse only expresses the IgMb allotype and over 95% of the B cells in all of the subsets except the CD93+ bone marrow cells (which include surface immunoglobulin negative pre-B cells) express IgMb. In contrast, the expression of IgM is more complex in B6.56R. As has been described previously for B6.56R, there are IgMa+ and IgMa dim cells (22). Based on studies in the related 3H9 heavy chain transgenic mouse, the IgMa dim cells in B6.56R likely correspond to an anergic B cell population (27) and are enriched in the transitional splenic B cell pool in B6.56R (Fig. 2b and (22)). There are also approximately 0.5-10% IgM- cells in B6.56R (the prevalence of IgM- cells differs in different subsets). In the mature subsets these IgM- cells may have undergone isotype switching or may have more severely down-regulated IgM. Down regulation of IgM could also be the result of peripheral light chain editing (28). There is a progressive loss of IgMa dim and IgM- cells as cells progress from the bone marrow, through the transitional compartment in the spleen and acquire CD23 expression (Fig. 2b). The similarity between the CD93- bone marrow fraction and the splenic transitional compartment is somewhat surprising because normally CD93- cells are thought of as more mature recirculating cells (CD93- immature bone marrow cells comprise a very minor fraction of this population because nearly all of the cells are IgM+). In B6.56R, these data suggest that the pattern of IgM expression in CD93- bone marrow B cells has more in common with transitional cells than with follicular (CD23+) B cells. Perhaps these patterns point to different selection pressures in circulating compared to tissue based B cells. B6.56R mice also have a population of IgMb+ cells that comprises approximately 5-25% of the B cell compartment, depending upon the subset (Fig. 2c). In contrast to the shifts in the IgMa and IgM- populations, the population with the highest proportion of IgMb+ cells in most of the mice was the CD93-CD23- splenocyte population. The progressive increase in IgMb expression in peripheral B cell fractions suggests that selection and/or continued H chain rearrangement are contributing to alterations in the frequency of peripheral IgMb+ cells.
Co-staining with anti-IgMa and anti-IgMb antibodies can be used to detect B cells expressing both the transgenic and endogenous H chain alleles. The frequency of IgMa+IgMb+ B cells is very low in B6.56R spleen (< 0.6% of IgM+ cells, data not shown). Immature B cells in the bone marrow appeared to contain a higher proportion of IgMa+IgMb+ cells (as a proportion of IgM+ cells), but this increase is likely to be spurious, due to non-specific staining of the anti-IgMa antibody. A similar relative increase in non-specific staining was seen in immature BM B cells of a B6 mouse (which only expresses IgMb, data not shown).
We considered three non-mutually exclusive scenarios to account for the low frequency of IgMa+IgMb+ cells compared to the higher frequency of IgMb+ cells. First, the inactivation of 56R via rearrangement of upstream Ig genes may be a prerequisite for rearrangement of the endogenous allele, preserving allelic exclusion. Second, IgMa+IgMb+ B cells may be generated in the bone marrow but rapidly counter-selected. Third, IgMb+ B cells may down-regulate surface expression of the 56R (IgMa+) H chain. These scenarios were evaluated in B6.56R hybridomas and B cells.
To evaluate B6.56R cells for allelic exclusion, we analyzed both H chain alleles in two new hybridoma panels from splenic B cells, one fused spontaneously from IgMb+ sorted splenocytes and a second from an in vitro LPS stimulated panel that was screened for IgMb antibody secretion. All 38 IgMb hybridomas from these panels lacked amplifiable 56R DNA, suggesting that inactivation of 56R had taken place (Fig. 3A). In order to determine if 56R had been inactivated by rearrangement of upstream D and/or VH genes, IgMb+ hybridomas were screened for these rearrangements by PCR. Rearrangements of D genes to 56R were identified with primers that are situated upstream of DSP2 and DFL16 D genes and were detected in 5 out of the 38 hybridomas (Fig. 3A). These D invasions are non-functional because they do not contain a VH gene. Rearrangements of VH genes to 56R were detected in 12/38 hybridomas with a degenerate primer (MH1) that is predicted to recognize > 90% of VH genes (23). Five of these VH rearrangements on the 56R allele were successfully cloned and sequenced; all included an upstream D gene and all were out-of-frame (Fig. 3B). The remaining 21 hybridomas did not amplify D or VH rearrangements on the 56R allele. This could result from chromosome loss or because they contain D or VH gene rearrangements that are not recognized by the VH and D primer sets. These data show that 56R is genetically, rather than phenotypically, inactivated in IgMb expressing hybridomas. (By phenotypic inactivation, we mean that the B cell expresses only one heavy chain on the cell surface but has in-frame H chain rearrangements on both H chain alleles.)
To assess if inactivation of 56R is a prerequisite for endogenous H chain rearrangement, we characterized rearrangements on the endogenous H chain locus in 56R+ hybridomas. One-hundred and one 56R+ hybridomas from two previously characterized hybridoma panels (680S, 680LPS, described in (22)) were screened for rearrangements on the endogenous locus with the MH1 primer and a JH primer recognizing all four endogenous JH gene segments. Endogenous H chain rearrangements were detected in 3 out of 101 56R+ hybridomas; all were out-of-frame and each utilized the most JH-proximal functional VH in B6, 7183.2.3 (Fig. 4A). (7183.2.3 in the B6 strain is very similar in sequence to the 81X VH gene segment in BALB/c.) In comparison, endogenous allele rearrangements sequenced from IgMb+ hybridomas utilized VH genes from a wider range of families, including the distal J558 family.
To determine if rearrangement of the endogenous H chain allele is restricted to the most JH- proximal functional VH in 56R+ B cells, we analyzed the prevalence of 7183.2.3 rearrangements and two more distal VH gene segments on the endogenous allele in IgMa+ splenocytes compared to IgMb+ splenocytes. Rearrangements were amplified using DNA from IgM allotype sorted cells with a primer specific for VH 7183.2.3, a primer specific for a more distal member of the 7183 family, 7183.9.15, and a primer specific for a distal VH, J606.1.79 (hereafter VH6). Each VH primer was used in a separate amplification with the same FAM-labeled JH2 primer, which only amplifies the endogenous H chain allele (Fig. 4B). PCR products were analyzed by capillary electrophoresis, and the number of rearrangements per 1,000 cells was determined. Details of the calculation and this analysis (also referred to as spectratyping) are given in methods and representative peaks and data are given in Fig. 1 and Table 1 of the electronic supplement. In DNA from IgMa+ cells, 7183.2.3-JH2 rearrangements have the highest abundance of the VH gene segments tested, whereas the number of rearrangements amplified for the two more distal VH genes, 7183.9.15 and VH6, is negligible (Fig. 4B). The same rearrangements were amplified in IgMb+ cells and were analyzed similarly. Unlike the IgMa+ cells, IgMb+ cells display the more characteristic peripheral B cell distribution of VH usage, with more frequent involvement of 5' VH gene segments (reviewed in (29)). The frequency of 7183.2.3-JH2 rearrangement in IgMb+ cells is low, consistent with the reported counter-selection of 81X (81X is homologous to 7183.2.3, (30)).
Previous studies in 56R mice have documented the presence of N-nucleotides at H chain junctions (19, 31). We confirmed this, and further analyzed the N-additions at H chain junctions. Sixty-three unique sequences were derived from several sources of DNA from B6.56R mice: whole spleen, splenocytes sorted by IgM allotype expression, and hybridomas (Supplementary Figs. 2A and 2B). Sequences of rearrangements on both alleles, using several different VH genes, were obtained. Nearly all sequences on the 56R allele included an upstream D gene; therefore, two junctions were available for analysis on the 56R allele (V-D and D-56R). N-additions, P-additions, and deletion of nucleotides were found at junctions on both alleles. The average length and G/C content of each junction are summarized in Table 1. Nucleotide sequences consistent with N-additions were present in 86% and 97% of rearrangements on the endogenous and 56R allele, respectively. By comparison, the prevalence of N-additions in peripheral B cells of non-transgenic mice is 84% (32). N-nucleotides added by TdT tend to be G/C rich (33, 34). Accordingly, the G/C content of all four junctions analyzed on both alleles was greater than 50%. Overall, the sequence analysis of H chain junctions is consistent with the addition of N-nucleotides by TdT.
The presence of N-additions suggests that H chain rearrangement occurs at the pro-B cell stage in B6.56R mice. However, this assumes that the expression pattern of TdT in B6.56R mice resembles that seen in wild-type mice. To rule out the possibility that TdT is inappropriately expressed in the presence of 56R, TdT protein expression was compared in B6.56R and B6 B cell subsets by flow cytometry. The highest amount of TdT staining is observed in Hardy Fraction B-C' in both B6 and B6.56R mice (Fig. 5A). Importantly, TdT staining is diminished in Hardy Fraction D and is not increased in Hardy Fraction E cells of either allotype in 56R. Similarly, TdT protein is not found at higher levels in Hardy Fraction F, IgMa or IgMb expressing marginal zone B cells or follicular B cells in B6.56R compared to B6. However, TdT staining is slightly higher in B6.56R splenic transitional cells than in B6 splenic transitional cells in some animals but not in others (individual experiments are shown in supplementary fig. 3). To determine if TdT is re-expressed in B6.56R transitional cells, TdT RNA expression was compared in populations of sorted cells from B6 and B6.56R mice by quantitative RT-PCR. The highest level of TdT RNA was observed in pro-B cells (Hardy Fractions B-C') (Fig 5B). TdT transcript abundance was comparable to background levels after the pro-B cell stage in B6.56R mice, including in transitional cells (supplementary fig. 4). Also consistent with an early (pro-B) window for H chain rearrangement, RAG-1 transcripts were detected in pro-B cells, but not above background levels in splenic B6.56R transitional or more mature B cell fractions from spleen or bone marrow (Fig. 5C).
H chain rearrangement should cease once a functional H chain is expressed, yet it occurs in some B cells that harbor a functional H chain knock-in transgene (17, 31). Here, we investigated the pathways and timing of H chain rearrangement in B6.56R mice, which have an anti-dsDNA H chain knock-in transgene on one allele and a wild type H chain locus on the other. Three H chain rearrangement pathways were identified in 56R B cells: inactivation of 56R by rearrangement of upstream Ig genes, accompanied by rearrangement of the endogenous H chain allele; productive rearrangement on the endogenous allele without inactivation of 56R; and productive rearrangement on the 56R allele.
The cryptic heptamer present in the 56R VH allows the rearrangement of upstream Ig genes to inactivate or replace 56R. Sequences of upstream VH genes rearranged to 56R reveal that nearly all include an upstream D gene (19, 31). Terminal rearrangement of D genes to 56R also occurs. The VH cryptic heptamer has been proposed to be more compatible with rearrangement to a VH-RSS than D-RSS; however, VH replacement of 56R occurs less often than D gene invasion accompanied by conventional V to D rearrangement (herein and ref. (19)). D genes may be preferred rearrangement substrates due to the presence of the intronic enhancer Eμ, located downstream of JH4, which increases germline transcripts and chromatin accessibility of the D and JH loci (19, 35). Eμ dependent D-JH rearrangement occurs in lymphoid precursors and in CD4+CD8+ T cells, so it is possible that D rearrangement to 56R is a stochastic event (36, 37). Though this type of rearrangement cannot be defined as VH replacement, in the context of 56R, “D replacement” serves as a model of VH replacement. To what extent the presence of a truncated D gene cluster upstream of the 56R transgene promotes the occurrence of ongoing H chain rearrangement is unresolved. However, a consequence of D invasion into 56R is that the B cell must undergo further H chain rearrangement on either allele if it is to survive.
Rearrangement to a cryptic heptamer is less efficient than rearrangement involving conventional RSS (38, 39). It is therefore not surprising that rearrangement of the endogenous allele occurs in 56R+ B cells (Fig. 4A, 4B and 4D). Three (out of 101) 56R+ hybridomas were found to have an out-of-frame rearrangement on the endogenous H chain locus. Yet at the phenotypic level, H chain allelic exclusion appears to be tightly enforced.
The pattern of endogenous allele VH usage in 56R+ cells suggests that rearrangements to proximal VH genes are more frequent. 56R+ hybridomas with endogenous allele rearrangements all utilized proximal VH genes that are part of the 7183 family, including the most proximal VH gene in B6, 7183.2.3. Furthermore, spectratyping analysis suggests that IgMa+ B cells with completed VH-JH2 rearrangements on the endogenous allele exhibit a skewing towards the use of JH-proximal VH gene segments. Rearrangements involving VH 7183.2.3 are over 50 times more frequent than VH6 and ten times more frequent than another 7183 family member. The frequency estimates in this analysis are based on performing spectratyping analysis at limiting dilution of input DNA. This analysis assumes that the PCR assays for different VH genes have similar sensitivities. It is possible that the sensitivities of the assays are not identical. However, even if this were the case, the results for the rearrangement frequencies are significantly different in IgMb+ vs. IgMa+ B cells, suggesting that proximal VH gene rearrangements are indeed over-represented on the endogenous H chain allele in IgMa+ B cells.
The over-representation of proximal VH gene rearrangements on the endogenous allele in 56R+ B cells is consistent with less tightly controlled rearrangement of these gene segments. The rearrangement of proximal VH genes appears to occur more often in the setting of allelic inclusion (9, 40, 41). Endogenous allele rearrangement of genes in the proximal 7183 and Q52 VH families has been observed in human IgM transgenic mice, while rearrangement of the distal J558 family is effectively excluded (40). In M54 H chain transgenic B cells, high levels of H chain allelic inclusion are observed and most of the endogenous H chain rearrangements involve the four most proximal VH genes (40-42). Allelic inclusion in vivo at the H chain loci is uncommon but when observed in normal B cells, one allele often utilizes a J558 family VH, and the other harbors a D-proximal VH gene (1).
The timing of H chain rearrangement in B6.56R B cells provides insight into the potential stimulus for these rearrangements. Early H chain rearrangement in B6.56R B cells, occurring before the specificity of the BCR has been determined, would suggest that rearrangements are not participating in active receptor editing. Conversely, H chain rearrangements occurring after BCR expression could indicate that BCR specificity promotes further H chain rearrangement. H chain rearrangement in autoreactive H chain knock-in B cells has led to speculation that the H chain locus may become re-accessible for rearrangement when an autoreactive BCR is expressed.
Consistent with the possibility of late H chain rearrangement, one study reported H chain DNA breaks in spleen (as well as bone marrow) DNA of the quasi-monoclonal mouse (43). The quasi-monoclonal (QM) mouse has a H chain knock-in on one allele, all JH segments deleted on the other H chain allele and is homozygous for kappa deficiency (44). Therefore, the major pathway by which antibodies are diversified in the QM mouse is VH replacement. It is not clear, however, if the H chain breaks seen in QM mouse spleen represent an unique peculiarity of the QM mouse, if the breaks were selectively recovered from rare contaminating immature cells, or if H chain rearrangements can also occur in mature B cells from normal spleen.
Three studies describe increased frequencies of H chain rearrangement in heavy chain knock-in B cells that are derived from immune stimulated B cell populations. The first showed that when 3H9 H chain knock-in mice were immunized with KLH, the majority of splenic B cells did not retain 3H9 (19). In the second, QM mice were immunized with VSV and antibodies that bound and/or neutralized the virus were found to contain D invasions and VH rearrangements (45). In the third, graft vs. host disease was induced in 56R mice and it was observed that splenocytes that had lost the 56R knock-in DNA were more frequent in GVH mice than in control 56R mice (46). One could interpret all of these studies to mean that VH replacement occurs in peripheral B cells, but a plausible alternative is that B cells that express H chains other than the transgene are more likely to be activated, by virtue of their greater diversity or their lesser autoreactivity (22). Selection affecting the H chain knock-in confounds interpretations of peripheral H chain editing based purely on rearrangement frequencies. If instead one looks to non-transgenic B cells for insights into the timing of VH replacement, peripheral H chain rearrangement seems unlikely. Two groups have failed to document H chain breaks in splenocytes recovered from wild type inbred mice (43, 47).
On the other hand, there is evidence in favor of VH replacement in peripheral human B cells. B cells isolated from the synovium of patients with rheumatoid arthritis exhibited heavy chain rearrangements with shared somatic mutations and some of these probable clones harbored diverse 5′ ends with shared CDR3 sequences (suggesting that VH replacement occurred after the start of somatic mutation). RAG-1 mRNA was also detected in synovial B cells. The shared mutation pattern and RAG expression are suggestive of VH replacement in peripheral B cells (48). In another study, Wilson observed “hybrid” VH gene rearrangements involving different VH4 family members recovered from IgM−, IgD+, CD38+ tonsillar B cells (in healthy subjects); some of these hybrid sequences had shared somatic mutations, again raising the possibility of peripheral VH replacement (49). Unlike in 56R and other H chain knock-in models, the position of the rearrangement junction varied (several different putative cryptic heptamers appeared to be used) and there was minimal junctional modification. These differences suggested that other enzymatic machinery such as AID could be involved in peripheral VH replacement. However, recent sequence analysis of VH rearrangements in patients with hyper IgM syndrome and AID deficiency reveals probable stigmata of VH replacement and is therefore inconsistent with the possibility that all VH replacements in man are occurring via AID (50).
Several molecular features of the H chain rearrangements in B6.56R suggest that they occur early during B cell development. Rearrangements on the 56R allele include D invasions. Indeed, 33 out of 34 VH rearrangements that were recovered from the 56R allele had at least one D invasion. In non-transgenic mice, D to J rearrangement initiates in early lymphoid progenitor cells (51). Furthermore, the near ubiquity of N-additions and the pattern of TdT RNA expression in 56R B cells also suggest that H chain rearrangements occur mainly in early stage B cells, prior to BCR expression. Our TdT data are consistent with the recently published findings of normal TdT expression in B cell development in B6.56R mice by Nakajima and colleagues (52). Furthermore, quantitative RT-PCR showed that TdT and RAG transcripts were were not expressed above background levels in surface immunoglobulin positive cells. Taken together, these findings favor a model wherein rearrangement at the H chain loci occurs at the pro-B cell stage in B6.56R.
Early H chain rearrangement in B6.56R mice suggests that specificity of the BCR is not the stimulus for rearrangement. It is possible that the pre-BCR plays a role in ongoing H chain rearrangement. However, breaks at the cryptic heptamer are detected even in μMT mice, which do not express the pre-BCR (47). Furthermore, VH replacement occurs in mice expressing two non-functional VDJ rearrangements, suggesting that the presence of a functional VH rearrangement is not required for VH replacement (38, 39). Recently Nakajima and colleagues showed H chain rearrangements occur on either allele in B cells from B6.56R mice that are λ5 deficient (52). Collectively, these findings indicate that pre-BCR signaling is not required for ongoing H chain rearrangement in 56R.
The prevalence of VH replacement in non-transgenic B cells has been called into question. VH replacement to a cryptic heptamer should leave a “footprint” between 5 and 8 nucleotides from the original VH, which may be difficult to distinguish from N-additions and other modifications (53). Watson et al. screened 518 heavy chain sequences from pre-B and immature B cells from TdT deficient mice and found only one sequence with a possible VH footprint (54). If all B cells that failed to rearrange the first H chain allele productively were given the opportunity to rearrange the second H chain allele, a large fraction (approximately 4/9) would have non-productive VDJ rearrangements on both alleles after the second rearrangement attempt. These VDJ−/VDJ− B cells would then either undergo VH replacement or be eliminated. The absence of evidence for frequent VH replacement suggests that VH replacement may be infrequent. On the other hand, VH replacement appears to be more common in human B cells (with frequencies of up to 10% reported by some groups (20)).
Another possible explanation for the low frequency of VH replacement in normal mice is that VH replacement may inefficiently rescue B cells with autoreactive H chain rearrangements. As VH replacements tend to conserve part or all of the original CDR3, it is possible that some of the dysfunctional or autoreactive features of the preceding rearrangement are carried over into the new rearrangement. Furthermore, because VH replacement tends to increase the CDR3 by the length of the footprint region, the CDR3 of the new rearrangement is longer, which could contribute to multireactivity (53, 55). Despite these constraints, significant frequencies of B cells with VH replacements have been documented in mice with H chain knock-ins such as 56R, or in mice in which both H chain alleles contain non-productive rearrangements (38, 39).
Several lines of evidence point to the selective survival of B cells that have rearranged and express a H chain other than 56R. This survival advantage can be explained negatively, from the standpoint of the disadvantages faced by 56R+ B cells, as well as positively, from the standpoint of the advantages enjoyed by 56R− B cells. Nearly all B cells that retain the 56R H chain use a highly restricted subset of L chains (17, 22). Yet, even with these so-called L chain editors, a large fraction of B6.56R+ B cells retain dsDNA-binding activity (22, 56). Thus, like 3H9 B cells, 56R+ B cells must be restrained by other peripheral tolerance mechanisms including surface IgM down-regulation and follicular exclusion (57). Indeed, a large population of IgMdim cells is observed in spleen and BM of B6.56R mice, and may represent an anergic population (57). Alternatively, IgMdim cells in B6.56R mice may harbor antibodies that bind to self-antigens that reside in an antibody-accessible compartment of the cytosol such as the Golgi apparatus, preventing their effective surface expression (58). Another possibility is that IgMdim B cells may be undergoing L chain receptor editing. Kiefer et al. have observed Rag1 expression and DNA breaks at the κ and λ loci in transitional B cells that have down-regulated surface IgM in 56R/Vκ8 double transgenic mice (28). We have not observed a significant increase in Rag1 mRNA levels in B6.56R transitional cells, although we have not enriched for surface IgM− or IgMdim cells within the transitional B cell population.
In addition to negative regulation of 56R+ cells, 56R− B cells likely have a selective advantage in B6.56R mice. IgMb+ B cells exhibit a wider repertoire of L chains than 56R+ cells (LY and ELP, unpublished data); furthermore, they are significantly less likely to secrete anti-DNA antibodies than 56R+ (IgMa+) B cells (22, 59). As a result of being more diverse, IgMb+ cells are more likely to participate in immune responses (46). Accordingly, IgMb+ cells are preferentially expanded in more mature B cell populations in B6.56R spleen. These data show that tolerance mechanisms are preserved in B6.56R mice, and that less autoreactive B cells are favored, regardless of how or when they are generated.
We thank Jan Erikson, Craig Bassing, Michael Atchison, David Gasser, Melvin Bosma, Martin Weigert and members of our laboratories for helpful discussions. We also thank Pamela Nakajima and Melvin Bosma for sharing their data with us prior to publication. We thank members of the University of Pennsylvania flow cytometry, molecular diagnosis and DNA sequencing core facilities for their skilled technical assistance with these experiments.