Class switch recombination (CSR) occurs between highly repetitive sequences called switch (S) regions and is initiated by activation‐induced cytidine deaminase (AID). CSR is preceded by a bidirectional transcription of S regions but the relative importance of sense and antisense transcription for CSR in vivo is unknown. We generated three mouse lines in which we attempted a premature termination of transcriptional elongation by inserting bidirectional transcription terminators upstream of Sμ, upstream of Sγ3 or downstream of Sγ3 sequences. The data show, at least for Sγ3, that sense transcriptional elongation across S region is absolutely required for CSR whereas its antisense counterpart is largely dispensable, strongly suggesting that sense transcription is sufficient for AID targeting to both DNA strands.
The generation of primary repertoire of immunoglobulin (Ig) and T‐cell receptor requires a developmentally regulated and lineage‐specific recombination between variable (V), diversity (D) and joining (J) segments through V(D)J recombination. The process is regulated by accessibility control elements including transcriptional promoters and enhancers (Hesslein and Schatz, 2001; Krangel, 2003; Cobb et al, 2006). The Eμ enhancer, located between JH segments and Cμ exons, has a critical role in germ‐line (GL) transcription of DH and JH segments and in V(D)J recombination. This enhancer regulates the production of a set of sense and antisense transcripts: Iμ transcripts that initiate at the Iμ promoter associated with Eμ enhancer, μ0 transcripts that initiate at a promoter upstream of DQ52, the most distal DH segment, and potentially DH and JH antisense transcripts (Perlot et al, 2005, 2008; Afshar et al, 2006; Bolland et al, 2007).
Upon antigen challenge, mature B cells undergo two other diversification processes: somatic hypermutation (SHM) that introduces point mutations at the V‐region exons and class switch recombination (CSR), which further diversifies antibody constant (C) domains. Both processes require transcription of the target sequences and the single‐stranded DNA‐specific cytidine deaminase, activation‐induced cytidine deaminase (AID). However, how does AID specifically target some sequences (e.g., V regions) while sparing adjacent sequences (e.g., C regions) remains largely unknown (Di Noia and Neuberger, 2007; Peled et al, 2008; Maul and Gearhart, 2010).
CSR occurs between highly repetitive S sequences located upstream of all the C genes except Cδ and involves double‐strand breaks at the recombining S sequences (Stavnezer et al, 2008). Targeted deletion of the whole S sequences severely impairs CSR, demonstrating their critical importance for efficient CSR (Shinkura et al, 2003; Khamlichi et al, 2004). Transcription of S sequences is derived from GL promoters called I promoters upstream of S sequences. The transcripts run through a non‐coding I exon and intronic S sequences and undergo polyadenylation downstream of the C exons. Splicing enables fusion of I exon to the C region yielding sterile transcripts (Chaudhuri and Alt, 2004). GL transcription is regulated in a major part by the 3′ regulatory region (3′ RR) located downstream of the IgH locus (Khamlichi et al, 2000; Manis et al, 2002). The transcriptional orientation of S sequences is important for CSR in vivo as inversion of Sγ1 decreases CSR to IgG1 (Shinkura et al, 2003). There is evidence that AID associates with the chromatin of the target S sequences in a GL transcription‐dependent manner through a direct interaction with the elongating transcription machinery (Nambu et al, 2003). In addition, a high density of RNA polymerase II (Pol II) molecules was detected at S regions indicating a stalling of Pol II molecules (Rajagopal et al, 2009; Wang et al, 2009).
During transcriptional elongation, GL transcripts form RNA–DNA hybrids with the template strand exposing the single‐stranded, non‐template strand. The main function of GL transcripts is thought to promote accessibility of S sequences through co‐transcriptional generation of R loops and possibly other structures that provide the substrates for AID (reviewed in Chaudhuri and Alt (2004)). In this context, it was found that AID‐initiated mutations were detectable at some 150 bp downstream of I promoters, rise sharply at S sequences before falling off at their downstream boundaries, and are undetectable at C regions (reviewed in Di Noia and Neuberger (2007)). However, there is clear evidence that both DNA strands are targeted by AID in vivo (Rada et al, 2004; Xue et al, 2006). Therefore, AID must somehow gain access to the presumably protected cytosines of the C‐rich template strand.
How does AID achieve this function is presently unclear. Besides the hypothesis that transcription bubbles may be sufficient for AID attack, there are several non‐mutually exclusive possibilities: (1) intracellular RNase H may degrade the RNA of the DNA/RNA duplex resulting in a collapsed structure in which misaligned repeats are exposed as single‐stranded regions on both strands (Yu et al, 2003), (2) transient supercoiling of DNA upstream of elongating Pol II molecules may have a role for AID targeting (Shen and Storb, 2004), (3) other factors such as the single‐stranded DNA binding protein RPA may target AID to non‐R loop‐forming sequences (Chaudhuri et al, 2004) and (4) splicing of the GL transcripts may be involved in the collapse or destabilization of the DNA/RNA duplex (reviewed in Yu and Lieber (2003)). In this context, previous work showed that mutation of the donor splice site of Iγ1 exon inhibited CSR to IgG1 despite normal GL transcription (Hein et al, 1998), (5) the antisense transcripts at S regions (Julius et al, 1988; Apel et al, 1992; Morrison et al, 1998; Perlot et al, 2008) may somehow contribute to AID targeting.
In the latter context, the expression pattern of antisense transcripts appears to mirror that of their sense counterparts: they are synthesized before CSR, they are produced constitutively at Sμ region, and they are induced at the same time and in the same stimulation conditions as the sense transcripts at downstream S regions (Perlot et al, 2008). Several hypotheses have been put forward to account for the function of antisense transcripts during CSR including promotion of Pol II stalling, stabilization of secondary structures upon potential collision of transcriptional complexes moving in opposite directions, and contribution to the modulation of chromatin topology beyond simple transcriptional opening (Roa et al, 2008; Teng and Papavasiliou, 2009; Maul and Gearhart, 2010).
In this study, we provide evidence that antisense switch transcripts are largely dispensable for CSR in vivo. We also report the striking finding that elongation past premature polyadenylation (PolyA) and pause sites is more efficient in μ locus than in γ3 locus.
Insertion of premature transcriptional termination sites in μ and γ3 regions
The transcriptional terminator cassette (hereafter called pAp cassette) was designed so that it provides polyA sites in both sense and antisense orientations. In the sense orientation, the pAp cassette was made up of a synthetic polyA site and pause site (Levitt et al, 1989) followed by SV40 late polyA site and its own pause site (Carswell and Alwine, 1989). In the antisense orientation, the cassette consists of SV40 early polyA and pause sites followed by a synthetic polyA site. The cassette thus provides two polyA sites in the sense orientation and three polyA sites in the antisense orientation (Supplementary Figure S1A). The pAp cassette was inserted at 168 bp downstream of Iμ exon (the knock‐in will be referred to as pApμ), 251 bp downstream of Iγ3 exon (hereafter pApγ) or ∼500 bp downstream of Sγ3 core repeats (hereafter Sγ3‐pAp) (Supplementary Figure S1B–G). All analyses were carried out on homozygous mice. In order to rule out the introduction of any mutation in the pAp cassette upon the integration events and the deletion of the selectable marker at the three insertion sites, the cassette was PCR‐amplified from genomic DNA of the corresponding splenic B cells and sequenced, no mutation was found (Supplementary Figure S1A and data not shown).
Analysis of early B‐cell development in pApμ mice
Preliminary observations showed somewhat surprisingly no sign of immunodeficiency in pApμ mice, which would be expected if V(D)J recombination and transcription were completely shut down. To check at which developmental stage the mutation exerts its effect, we first analysed B‐cell populations in the bone marrow by flow cytometry. A ∼2.5‐fold decrease was seen for the B220+ population in pApμ mice (∼22%) compared with WT controls (∼59%) (Supplementary Figure S2A). Double positive B220+IgM+ population was decreased by half in pApμ bone marrows (∼6%) compared with WT controls (∼13%) (Supplementary Figure S2B). A slight accumulation was found for mutant pro‐B cells (B220+IgM−CD43high) (∼30% for pApμ and ∼20% for WT) and a decrease in pre‐B cell population (B220+IgM−CD43low) (∼70% for pApμ and ∼80% for WT) (Supplementary Figure S2C). The results show that insertion of the pAp cassette in μ intron slightly impairs, but does not block early B‐cell development.
Normal V(D)J recombination in pApμ mice
In order to further characterize the step at which the impairment occurs, we analysed V(D)J recombination at the IgH locus. Genomic DNA was prepared from sorted pro‐B cells (B220+IgM−CD43high) and pre‐B cells (B220+IgM−CD43low) and was subjected to PCR using a set of specific primers. No obvious decrease was detected in DH to JH rearrangement in pApμ pro‐B cells compared with WT controls. Likewise, VH to DHJH rearrangement was normal be it for the proximal VH7183 or the distal VHJ558 gene families (Supplementary Figure S3 and data not shown). VκJκ was also normal (Supplementary Figure S3). Thus, the mutation has no apparent effect on V(D)J recombination in pApμ mice.
Decreased GL transcription in the bone marrow of pApμ mice
The lack of effect of the pApμ insertion on V(D)J recombination led us to investigate whether the decrease in B‐cell population correlates with a decrease in V(D)J transcription. To this end, total RNA from magnetically sorted CD19+ bone marrow B cells was subjected to RT–PCR using a degenerate primer specific for the largest VHJ558 gene family and a primer specific for the Cμ1 exon. VDJCμ transcript levels were clearly decreased in pApμ mice compared with WT controls (Figure 1A), indicating a partial inhibition of transcriptional elongation, though we cannot exclude the involvement of checkpoints in the bone marrow.
In order to analyse the extent of the transcriptional read‐through independently of selection mechanisms that operate against the B cells that fail to express a functional (pre‐) B‐cell receptor, the pApμ mutation was brought into an Rag2−/−‐deficient background. Total RNA was prepared from bone marrow purified CD19+ B cells and subjected to RT–PCR using primers that are specific for μ0 or Iμ exons (primers 3′DQD1‐Fw/DDJH1‐Rev and IμFw/IμRev, respectively) or for the spliced forms of μ0 and Iμ transcripts (μ0‐Fw/Cμr and Iμ‐Fw/Cμr) (Figure 1B). While initiation of μ0 and Iμ exon transcripts in pApμ/Rag2−/− mice was comparable to that of Rag2−/− controls (Figure 1C and D), a mild decrease (∼1.5‐fold) was detected for both μ0 and Iμ spliced transcripts (reflecting a bypass of the pAp cassette) in pApμ/Rag2−/− mice versus Rag2−/− controls (Figure 1C and D). The data indicate that the pApμ insertion has no effect on the initiation step of μ0 and Iμ GL transcription and that it impairs but clearly does not totally block transcriptional elongation through Sμ region.
Analysis of splenic B cells in non‐immunized pApμ mice
The decrease seen in B220+IgM+ population in pApμ bone marrow may lead to an alteration of B‐cell homoeostasis in the periphery. Therefore, we sought to know if the mutation would affect the accumulation of the mutant B cells in the spleen. To this end, we resorted to flow cytometry by using an anti‐B220, anti‐IgM and anti‐IgD antibodies. We found that the mutant B cells replenished normally the spleen at levels comparable to WT controls. In fact, we consistently found a slight but reproducible (n=3) increase in B220+IgM+ and B220+IgD+ populations in pApμ spleens compared with WT controls (Supplementary Figure S4A and B). Altogether, the data show that replenishment of the mutant spleens occurs normally with a slight accumulation of IgM‐ and/or IgD‐expressing populations, thus excluding any major alteration of late B‐cell homoeostasis in pApμ mice.
Decreased CSR in activated pApμ B cells
We then asked if and how the pApμ mutation would affect CSR to downstream isotypes. To this end, we resorted to ex vivo activation of purified splenic B cells from WT, pApμ and AID−/− mice. B cells were cultured in the presence of anti‐CD40+IL4 which induce switching to IgG1, of LPS+anti‐IgD‐dextran which induce switching to IgG3 and IgG2b or of LPS+TGF‐β+B‐LYS which induce switching to IgA. Surface expression was monitored by flow cytometry at day 4.5 using an anti‐B220 antibody and anti‐IgG1, anti‐IgG3 or anti‐IgA antibodies. Surface expression was decreased ∼1.3–2.3‐fold for all isotypes tested but was clearly not extinguished in pApμ activated B cells (Figure 2A and B). Thus, insertion of the pAp cassette upstream of Sμ results in a generalized but mild defect in CSR to downstream isotypes.
The decreased CSR at the genomic level correlates with a decreased μ GL transcription in pApμ mice
In order to check that the decrease in CSR seen by FACS occurs at the genomic level, we resorted to DC‐PCR (Chu et al, 1992). At day 4.5 post‐stimulation, genomic DNA was prepared from WT, pApμ and AID−/− splenic B cells that have been induced to switch to IgG1 or IgG3 and subjected to DC‐PCR. CSR to Cγ1 and Cγ3 was clearly decreased in stimulated splenic mutant B cells (Figure 3A), demonstrating that the impairment of CSR in pApμ mice occurs indeed at the genomic level.
To analyse GL transcription, total RNA was prepared at day 2 from activated WT and pApμ B cells, was reverse transcribed and amplified in semi‐quantitative conditions using isotype‐specific GL transcript primers (Figure 3B). We will hereafter use the term GL transcripts meaning the ‘classical’ spliced GL transcripts derived from I promoters. Iγ3–Cγ3 and Iγ2b–Cγ2b GL transcript levels were comparable in WT and pApμ B cells activated with LPS+anti‐IgD‐dextran. The same pattern was found for Iγ1–Cγ1 GL transcripts in anti‐CD40+IL4‐activated B cells (Figure 3B and D). With regard to μ transcripts, Iμ exon transcripts (reflecting initiation) were equally abundant in WT and pApμ B cells. Additionally, we used IμF and pA1 primers that would detect mutant transcripts that end up at the pAp site and probably some μ pre‐mRNAs that reach Cμ polyA site. Transcripts in the form of Iμ–pA1 (485 bp) were clearly present in pApμ splenic B cells in both stimulation conditions (Figure 3B and data not shown). In contrast, spliced Iμ–Cμ GL transcripts were decreased (∼3‐fold) in activated pApμ B cells compared with WT controls (Figure 3B and D) regardless of the stimulation condition (data not shown).
We also analysed post‐switch GL transcripts in the form of Iμ–Cx (Cx being any switched isotype), which are a good marker of the efficiency of CSR. Given the location of the pApμ insertion site, one might expect that the vast majority of CSR events will preserve the pAp cassette in the newly generated hybrid Sμ–Sx regions. Thus, the defect in CSR can be directly correlated with the effect of the mutation. Total RNA at day 4.5 post‐stimulation was prepared from WT, pApμ and AID‐deficient mice and amplified with Iμ–Cx primers (x being μ, γ1, γ3, γ2b) or Iμ–pA1 primers. Post‐switch transcripts corresponding to all isotypes tested were readily detectable, but were clearly decreased (3–4‐fold decrease depending on the isotype) in pApμ mice (Figure 3C and D). Again, unspliced Iμ‐pA1 transcripts were detected in pApμ mice but not in control mice as expected (Figure 3C).
The identification of antisense transcripts in primary B cells (Perlot et al, 2008) led us to investigate whether the decreased CSR in pApμ mice correlates with a decrease in sense or antisense transcription. To this end, we used a strand‐specific RT–PCR (Bolland et al, 2007; Perlot et al, 2008) to analyse sense and antisense transcription upstream and downstream of the pAp insertion site. As shown in Figure 3E, unprocessed sense and antisense transcripts were detected both upstream and downstream of the insertion site in WT and pApμ mice.
Altogether, the results demonstrate that the pApμ mutation has no effect on downstream pre‐switch GL transcripts and that a substantial fraction of μ and post‐switch GL transcripts can elongate past the pAp site and undergo splicing. Thus, the premature pAp site impairs but does not completely stop elongation of μ GL transcripts or post‐switch GL transcripts. The data also show that there is a correlation between the defect of CSR at the genomic level and the decrease of μ GL transcript levels. In contrast, we could not conclude about the relative importance of sense versus antisense transcription regarding the impairment of CSR.
Complete and specific abrogation of CSR in pApγ mice
In order to analyse the effect of pApγ mutation on CSR, we resorted to ex vivo activation of pApγ splenic B cells. Surface expression of IgG1 was comparable in WT and pApγ B cells when activated with anti‐CD40+IL4 (Figure 4A). In contrast, while surface expression of IgG3 was readily detectable in WT B cells activated with LPS+anti‐IgD‐dextran, that of their pApγ counterparts was completely absent (Figure 4B). The same conclusion could also be drawn from serum and culture supernatant analyses by ELISA. IgG2b titres of pApγ sera or supernatants from LPS+anti‐IgD‐dextran stimulation were comparable to those of WT controls. In contrast, only background levels of IgG3 were detected in the sera or culture supernatants from pApγ mice (Figure 4C and D).
The results show that insertion of the pAp cassette upstream of Sγ3 sequence leads to a complete and specific extinction of IgG3 production in vivo and ex vivo and that the block is B‐cell autonomous and cannot be ascribed to some inhibitory signals in vivo.
The absence of CSR to IgG3 in pAp‐γ mice correlates with a shutdown of sense transcription
In order to demonstrate that the shutdown of CSR to Cγ3 occurs at the genomic level, genomic DNA was prepared from WT, pApγ and AID−/− splenic B cells induced to switch to IgG1 or IgG3 and subjected to DC‐PCR. While CSR to IgG1 was comparable between WT and pApγ splenic B cells induced with anti‐CD40+IL4, the data clearly showed a total absence of CSR to IgG3 in pApγ B cells activated with LPS+anti‐IgD‐dextran (Figure 5A). Thus, abrogation of CSR to IgG3 occurs at the genomic level.
In an attempt to establish closer links between the defect of CSR at the genomic level and the potential impairment of sense or antisense transcription, we analysed GL transcripts from WT or pApγ splenic B cells activated with LPS+anti‐IgD‐dextran. The levels of Iγ2b–Cγ2b GL transcripts were comparable in WT and mutant pApγ B cells (Figure 5B and D). In contrast, a drastic decrease was found for spliced Iγ3–Cγ3 transcripts in pApγ B cells despite normal transcription initiation (Iγ3F–Iγ3R primers) (Figure 5B and D). The transcripts in the form of Iγ3–pA1 (570 bp) were readily detected (Figure 5B) further demonstrating that Iγ3 promoter is in an open configuration, and that most detectable transcripts that originate from Iγ3 fail to bypass the pAp site. Hence, the low level of GL transcription across Sγ3 is clearly not sufficient for detectable CSR to IgG3 in pApγ3 mice.
We also analysed post‐switch transcripts at day 4.5 post‐stimulation. Iμ–Cγ2b transcripts were equally abundant in WT and mutant B cells (Figure 5C and D). As expected, IμF–pA1 transcripts were not detected in pApγ B cells (Figure 5C). More importantly, Iμ–Cγ3 post‐switch transcripts were barely detectable in pApγ splenic B cells (Figure 5C and D) further confirming the lack of CSR to IgG3 in pApγ B cells.
These data demonstrate that transcription initiation from Iγ3 promoter is intact in pApγ mice and that the vast majority of γ3 GL transcripts fail to efficiently elongate past the pAp site. Thus, the data strongly suggest that failure to efficiently transcribe Sγ3 region in the sense orientation is likely to account for the specific lack of CSR to IgG3 in pApγ mice.
In order to prove this, we used strand‐specific RT–PCR to analyse sense and antisense transcription. Sense transcripts were readily detected upstream of the insertion site in both WT and pApγ mice (Figure 5E, see also Figure 8). In contrast, sense transcripts downstream of the insertion site were barely detectable in pApγ mice compared with WT mice, further confirming the strong termination of sense transcription by the pAp cassette. Interestingly, antisense transcripts were readily detected downstream of the insertion site in both WT and pApγ mice, indicating that antisense transcription of Sγ3 occurred and was not hindered by the mutation. In contrast, only a faint signal was detected for antisense transcripts upstream of the insertion site in pApγ mice, suggesting that few Pol II molecules have managed to elongate past the pAp cassette. We conclude that the abrogation of CSR to IgG3 in pApγ mice correlates with the lack of sense transcription of Sγ3 and that antisense transcription of Sγ3 in the absence of its sense transcription is not sufficient for CSR to IgG3.
Normal CSR to IgG3 in Sγ3‐pAp mice
In order to provide further functional support to the notion that it is the sense transcription of Sγ3 that counts for efficient CSR to IgG3, we generated a third mouse line in which the pAp cassette was inserted downstream of Sγ3. We reasoned that because of the bidirectional function of the pAp cassette, Sγ3 would be transcribed in the sense orientation while its antisense transcription should be blocked. Therefore, should CSR to IgG3 occur, the correlation with sense transcription would be more firmly grounded.
In order to analyse the effect of Sγ3‐pAp mutation on CSR, we first resorted to ex vivo activation of splenic B cells. No obvious difference could be detected between WT and Sγ3‐pAp activated B cells regarding surface expression of IgG1 and, somewhat surprisingly, of IgG3 (Figure 6), indicating normal CSR to IgG3. Thus, in stark contrast to the situation in pApγ mice, insertion of the pAp cassette downstream of Sγ3 does not affect CSR to IgG3.
Normal CSR to IgG3 in the absence of Sγ3 antisense transcription in Sγ3‐pAp mice
We confirmed this result at the genomic level by using DC‐PCR. No obvious difference could be seen between Sγ3‐pAp and WT control regarding Sμ>Sγ3 switching (Figure 7A). To check that no AID‐induced mutation occurred at the pAp cassette, we again amplified the cassette from genomic DNA of activated B cells and definitively excluded this possibility (data not shown).
The Iγ3–Cγ3 and Iγ2b–Cγ2b pre‐switch transcripts were equally abundant in WT and Sγ3‐pAp activated B cells (Figure 7B). Conversely, Iμ–Cγ3 and Iμ–Cγ2b post‐switch transcript levels were comparable in WT and mutant B cells further confirming normal CSR to IgG3 and IgG2b in Sγ3‐pAp mice (Figure 7C).
In order to analyse the contribution of sense and antisense transcription to the normal CSR to IgG3 in Sγ3‐pAp mice, we used the strand‐specific RT–PCR assay. Sense transcripts were readily detected upstream and downstream of the insertion site in both WT and Sγ3‐pAp mice. Likewise, antisense transcripts were readily detected downstream of the insertion site in both WT and Sγ3‐pAp mice. More importantly, antisense transcripts were undetectable upstream of the insertion site in Sγ3‐pAp mice compared with WT control (Figure 7D, see also Figure 8). We conclude that in Sγ3‐pAp mice, normal CSR to IgG3 correlates with sense transcription of Sγ3. These results, together with the data from pApγ mice, show that antisense transcription of Sγ3 does not significantly contribute to normal CSR to IgG3.
Insertion of a transcriptional terminator upstream of Sγ3 efficiently blocks γ3 GL transcription elongation and results in a total and specific abrogation of CSR to IgG3, whereas μ GL transcriptional elongation is only mildly affected in similar conditions, leading to a moderate decrease of CSR to downstream isotypes. We conclude that transcription elongation past the pAp cassette at μ and γ3 loci is differentially regulated in vivo, although we cannot presently exclude the remote possibility that we may be dealing with highly selected mature B‐cell clones in pApμ mice but not in pApγ mice.
Analysis of early B‐cell development in pApμ mice revealed a decrease in B220+ population that was not due to a failure to rearrange the variable gene segments but to a partial inhibition of μ transcription. While this indicates that the pAp cassette does not completely block elongation at μ locus, we cannot exclude the implication of selection mechanisms (von Boehmer and Melchers, 2010) so that only clones that have successfully expressed a functional pre‐BCR and BCR would survive. However, the fact that the blockade at μ0 and Iμ elongation step is not complete in Rag2−/− background, while initiation is normal, clearly shows that the ability to bypass the pAp cassette is inherent to μ transcription elongation.
Despite the structural similarities between μ and γ3 transcription units, there are some important differences. In particular, Eμ enhancer acts both as an enhancer and as a GL promoter at μ locus, whereas Iγ3 GL promoter has no known enhancer function. Additionally, while Sμ is transcribed throughout the B‐cell development, Sγ3 is selectively transcribed after B‐cell activation. Another difference stems from the fact that Eμ/Iμ‐derived transcription is constitutive (Li et al, 1994) whereas Iγ3‐derived transcription is inducible.
Taking into account the fact that μ and γ3 introns are of relatively the same size, the distance of the pAp cassette from the splice sites of μ and γ3 sense transcripts, is unlikely to account for the difference in the phenotypes of pApμ and pApγ mice. Interestingly, while sense μ transcripts are readily detected downstream of the pAp cassette in pApμ mice, their γ3 counterparts are barely detectable in pApγ mice, suggesting that γ3 transcripts downstream of the cleavage site(s) (3′ of the polyA sites of the pAp cassette) undergo extensive degradation before reaching Sγ3 (where the generated DNA/RNA duplex may possibly contribute to stabilize them) and the Cγ3‐1 acceptor splice site. Strikingly, when the pAp cassette is moved downstream of the Sγ3 (in Sγ3‐pAp mice), the transcripts downstream of the pAp cassette appear to be just as abundant as their WT counterparts.
Thus, within the framework of current models of co‐transcriptional processing (Moore and Proudfoot, 2009; Richard and Manley, 2009), the failure of Pol II molecules to efficiently transcribe Sγ3 in pApγ mice stems, at least in part, from a failure to protect the 3′ RNA transcripts from degradation before splicing. In the case of Sγ3‐pAp mice, the insertion site lies at some 2 kb from Cγ3‐1 acceptor splice site and downstream of the core Sγ3 where the majority of R loops form (Huang et al, 2006), and where Pol II molecules stall (Wang et al, 2009). Hence, one possibility is that the high density of stalled Pol II at Sγ3 (and perhaps the stable DNA/RNA duplex structures) protects the transcripts against cleavage, enabling Pol II molecules to rapidly reach the Cγ3‐1 acceptor splice site. Thus, in this scenario, the density of Pol II arises as a critical determinant at the basis of the phenotypes of the three mutant mouse lines, and more specifically in the differential ability to ignore premature termination sites. Whether Pol II molecules recruit or lack specific factors at the three pAp sites is presently unknown. Whatever the case, these factors need not be recruited in a developmentally regulated manner as described for instance for membrane versus secreted forms of μ transcripts (Takagaki et al, 1996), as the differential ability to bypass the pAp sites is manifest in the very activated splenic B cells of the corresponding mouse lines.
Previous studies involved enhancers in the stimulation of elongation past premature or endogenous termination sites (Yankulov et al, 1994; Krumm et al, 1995; Blau et al, 1996; Brown et al, 1998, Sawado et al, 2003). Thus it is plausible that Eμ enhancer is involved in the elongation of μ transcripts, which may confer to Sμ some features that would be missing in downstream S regions, for instance, by acting as a powerful entry site for Pol II as a high density of Pol II molecules was detected at Eμ/Iμ region (Rajagopal et al, 2009; Wang et al, 2009; Pavri et al, 2010). In addition, different sets of transcripts run through Sμ region throughout B‐cell development before, and after, the onset of V(D)J rearrangement. Such continuous transcription may imprint on Sμ chromatin some structural and/or epigenetic marks that distinguish it from downstream S regions. Moreover, after CSR, post‐switch transcripts, though decreased, are readily detectable in pApμ mice irrespective of the acceptor switch site or the stimulation condition, suggesting that Eμ enhancer, which remains intact in any CSR event, continues to activate transcription elongation of the hybrid Sμ/Sx sequences.
In stark contrast to pApμ, the only set of Pol II molecules that reach the pAp cassette in pApγ mice, before CSR, are those that initiate transcription from Iγ3 promoter. Therefore, the strong inhibition of transcriptional read‐through in pApγ mice could, at least in part, be due to a weaker GL promoter. Taking into account the fact that the 3′RR controls Iγ3‐derived GL transcription (Khamlichi et al, 2000; Manis et al, 2002), one might speculate that the 3′RR acts at the elongation step downstream of the pApγ site. An effect of a locus control region on elongation has been reported in the β‐globin locus (Sawado et al, 2003). We propose that Eμ enhancer and the 3′RR may facilitate the transition of Pol II molecules from stalling to efficient elongation at Sμ and downstream S regions, respectively.
By using an inducible switch substrate in a B lymphoma line, it was suggested that there exists a linear correlation between the transcription level of Sα and the efficiency of switching to IgA (Lee et al, 2001). Our results support this notion and extend it to Sμ and Sγ3 in vivo. More specifically, we provide evidence that, at least for Sγ3, there is a correlation between the level of sense transcription and the efficiency of CSR to IgG3.
While we could not conclude on the relative importance of sense and antisense transcription of Sμ for the targeting and efficiency of CSR, the data from pApγ and Sγ3‐pAp mice confer the crucial role to the sense transcription of Sγ3 in CSR to IgG3. These data exclude a major role for the putative collision of opposite transcription machineries in the efficiency of CSR. They cannot be explained by a double‐stranded RNA made up of sense and antisense transcripts at the polyA sites that might mask these sites (reviewed in Faghihi and Wahlestedt (2009)). Rather, the recruitment of the critical factors for CSR to IgG3 seems to be primarily conditioned by the efficient transcription of Sγ3 sequence in the sense orientation.
It thus appears that sense transcription is sufficient for AID activity on both DNA strands of transcribed Sγ3, indicating that a double‐stranded RNA structure is not mandatory for AID targeting and action. How does AID attack the presumably protected template strand is presently unknown and several mechanisms have been proposed (see Introduction). It is also unclear whether AID targets both strands at the same time in vivo or whether it attacks the non‐template strand first. From the pApγ model (in which CSR to IgG3 is inhibited despite antisense transcription of Sγ3), it would appear that mutation of the non‐template strand is the limiting step suggesting a stepwise process, and indicating that sense transcription‐mediated generation of structures such as R loops is a crucial factor for the efficiency of CSR in vivo. This would confer a major role to the G content of the sense transcript as suggested previously (Reaban et al, 1994; Daniels and Lieber, 1995; Mizuta et al, 2003; Shinkura et al, 2003; Yu et al, 2003).
Yet, it is still puzzling how AID targets the exposed arm of the loop while it is recruited to the transcribed strand. Recent studies indicate that PKA is specifically recruited to S regions to promote localized phosphorylation of AID and binding of RPA (Vuong et al, 2009), and that the stalling factor Spt5 associated with RNAp II facilitates AID recruitment to its targets (Pavri et al, 2010). Whether these associations occur during antisense transcription is presently unclear. Nonetheless, it could be that during sense transcription, stalled RNAp II, Spt5 and PKA recruit increased number of AID molecules and allow sufficient time for its access to the exposed strand. Alternatively, the move of AID from the transcribed strand to the exposed strand may be an active and regulated process that requires specific cofactors and/or post‐translational modifications of AID.
It has been noted (Roa et al, 2008) that if antisense transcription had a major role, one would expect a mutation pattern downstream of the antisense start site similar to that of the sense start site, which is clearly not the case in vivo (Xue et al, 2006) although more complex scenarios may be envisaged. Thus, our results provide a basis for the prediction that transcription‐associated high rate of mutation correlates with sense transcription during SHM.
Materials and methods
GL transcription in bone marrow
B cells from bone marrow were sorted by using CD19‐magnetic microbeads and LS columns (Miltenyi Biotec) and total RNA was reverse transcribed (Invitrogen) and subjected to PCR. Quantification was as described (Haddad et al, 2010).
Spleen cell cultures
After removal of red blood cells, a single‐cell suspension from spleens of 6–8‐week‐old mice was obtained. Ex vivo stimulation was as described (Boboila et al, 2010; Zhang et al, 2010). Negatively sorted B cells were activated with 25 μg/ml of LPS (Sigma) and anti‐IgD‐dextran (Fina BioSolutions) at 3 ng/ml or with 0.5 μg/ml of anti‐CD40 (eBiosciences) and 20 ng/ml of IL4 (eBiosciences) or with 25 μg/ml of LPS, 3 ng/ml of anti‐IgD‐dextran, 10 ng/ml of IL4, 2 ng/ml of TGF‐β (R&D systems), 5 ng/ml of IL5 (R&D systems) and 5 ng/ml of B‐LYS (R&D systems). At days 2 and 4.5, aliquots of cells were removed for RNA preparation.
Flow cytometry analysis of CSR
At day 4.5, activated B cells were stained with anti‐B220‐APC and anti‐IgG1‐FITC, with anti‐B220‐APC and anti‐IgG3‐FITC or with anti‐B220‐APC and anti‐IgA‐FITC (Biolegend). Data were obtained on 3 × 104 viable cells by using a Coulter XL apparatus (Beckman Coulter, Fullerton, CA).
ELISA on serum and culture supernatants
GL transcription in stimulated splenic B cells
For GL transcription, the primers, the RT–PCR conditions and the expected sizes of the PCR products have been described (Oruc et al, 2007; Haddad et al, 2010). pA1, GATGGAGAGCGTATGTTAGTAC. Additional primers used in this assay are listed in the Supplementary data. For unprocessed sense and antisense transcripts, primers downstream of pApμ and RT–PCR conditions were as described (Perlot et al, 2008). Briefly, reverse transcription is performed with strand‐specific primers. The latter generate sense‐ and antisense‐specific cDNAs, respectively, which are then used as templates for two different PCRs that control each other: sense‐specific and antisense‐specific PCR leading to overlapping products (Bolland et al, 2007; Perlot et al, 2008). Additional controls include reverse transcription with random hexamers (which do not allow distinction between sense and antisense transcripts), and controls for genomic DNA contamination.
Negatively sorted B cells were activated as described above. At day 4.5, live cells were sorted with dead cell removal kit (Miltenyi) as per manufacturer's instructions. Genomic DNA was purified and subjected to DC‐PCR essentially as described (Chu et al, 1992). The primers used are listed in the Supplementary data.
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
Conflict of Interest
The authors declare that they have no conflict of interest.
We thank C‐A Reynaud and J‐C Weill for critical reading of the manuscript, S Aoufouchi for helpful comments and M Cogné for warm support. We also thank T Honjo for providing AID‐deficient mice. ZO was supported by fellowships from the MREN, the FRM and EMBO‐STF. This work was supported by ARC (Grant SFI20101201441), ANR (ANR‐07‐BLAN‐0080‐03), INCa (No 07/3d1616/PL‐96‐052/NG‐NC, EpiEMPREINTE), LCC—Comité de Haute‐Garonne and Cancéropôle GSO: L'instabilité génétique comme signature péjorative de la maladie (ACI 2007–2009).
- Copyright © 2011 European Molecular Biology Organization