The chicken Ig light chain 3′-enhancer is essential for gene expression and regulates gene conversion via the transcription factor E2A

Deletion of the MAR in the chicken IgL locus

A knockout construct for the MAR (Fig. 1A) was transfected into DT40 cells; neomycin-resistant clones were selected and analysed for homologous recombination. DT40 cells contain a rearranged and a non-rearranged Ig allele, which, after Pst I digestion, yield a 3.0- and a 2.4-kb fragment using an IgL constant region probe in Southern blot analysis (Fig. 1B). Insertion of the neomycin gene at the rearranged allele leads to the formation of a 1.7-kb band instead of the 3.0-kb band. Homologous recombinants were analysed for IgM surface expression by FACS (Fig. 2A). Insertion of the neomycin gene in place of the MAR generated an IgM^– phenotype. This is not surprising since the neomycin gene introduces a stop codon, which causes premature termination of transcription leading to a non-functional IgL message. To assess the effect of deleting the MAR without interference from the inserted neomycin gene, we stably transfected these cells with a tamoxifen-inducible form of Cre recombinase. Upon induction of the recombinase, approximately 15% of cells had become IgM⁺. These were likely to be the cells in which the Cre recombinase had catalysed the excision of the neomycin gene, leaving only the loxP sites in its place. To test this, IgM-positive and -negative cells were sorted and examined for their Ig genotype. Fig. 1C shows that indeed the IgM-positive cells had lost the neomycin gene, while the IgM-negative cells retained it. This phenotype remained stable after several weeks of culture. The level of surface expression in the Ig-positive population was very similar to that of the parental DT40 cells (Fig. 2A). However, we wished to examine RNA levels and utilised real-time PCR to quantitate IgL expression after deletion of the MAR. Fig. 2C shows that RNA levels are strongly reduced after the insertion of the neomycin gene (MAR neo). Similar expression levels were seen in cells that had been induced by tamoxifen, but failed to recombine out their neomycin gene (Tmx–). However, there was an approximately twofold increase in the levels of IgL RNA in MAR knockout clones in which the neomycin gene had been removed (MAR del) (Fig. 2C). Hence, the MAR in the chicken light chain locus acts as a negative regulator. This is in keeping with findings in the murine system where the MAR also acts as a negative regulator, lowering the probability of premature Igκ rearrangement 23.

Deletion of the 3′-enhancer in the chicken IgL locus

Next, we examined the effect of deleting the 3′-enhancer. The knockout construct (Fig. 3A) was stably transfected into DT40 cells and homologous recombinants were identified by Southern blotting. Using the restriction enzymes Bam HI and Sac I, the non-rearranged and the rearranged IgL alleles give rise to a 7.4- and a 6.2-kb band, respectively (Fig. 3B). Insertion of the neomycin gene into the non-rearranged allele would yield a 4.9-kb band, although no such band was seen in any of the clones analysed. Deletion of the 3′-enhancer by neomycin insertion into the rearranged allele leads to the formation of a 6.2-kb band, which is the same size as the non-rearranged allele. The creation of a homologous recombinant (3′E neo) is therefore characterised by the loss of the 7.4-kb band. To detect recombination between the loxP sites after the induction of Cre recombinase expression, we used PCR to amplify the enhancer regions with the primers shown (Fig. 3B). The WT locus gives rise to a single 748-bp PCR product, while the loxP recombinant (3′E del) gives rise to an additional band of 321 bp (Fig. 3D). No detectable PCR product was generated from alleles containing the neomycin gene, as the resultant fragment is too big to be amplified efficiently. Using FACS to assess IgM expression (Fig. 2B), we determined that replacement of the enhancer by neomycin (3′E neo) was sufficient to give rise to an IgM^– phenotype. After deletion of the neomycin gene by Cre recombinase, these cells remained IgM^–. In fact, somewhat reduced surface expression is observed when compared to the neomycin-containing knockout (Fig. 2B). Real-time PCR analysis of IgL expression levels (Fig. 2C) revealed that the differences seen in the FACS analysis are also apparent at the RNA level. Upon insertion of the neomycin gene (3′E neo), transcription was much reduced, falling to 1/50th of the WT IgL. Once the neomycin gene had been removed, expression was reduced another 100-fold in two independent Cre transfectants (3′E del.1 and 3′E del.2). Presumably, the presence of the promoter driving the neomycin gene is able to maintain low levels of expression. In its absence, the locus becomes fully inactive. In future analyses, it will be interesting to determine whether or not deletion of the 3′-enhancer leads to full chromatin condensation. In conclusion, our experiments have shown that the 3′-enhancer is essential not only for the establishment of an active locus, but also for maintenance of IgL gene expression in a mature B cell.

The 3′-enhancer contributes to the regulation of GC

In the mouse model, deletion of the intron or the 3′ enhancer does not stop surface expression of Ig. This means that it has been possible to assess the role of the intron and 3′-enhancer in regulating the levels of somatic hypermutation 24, 25. We have demonstrated above that deletion of the IgL 3′-enhancer leads to a more than 1000-fold reduction in transcription. There is now extensive evidence that transcription is required for SHM or GC to occur (reviewed in 10). Thus, in order to examine the role of the IgL 3′-enhancer in the regulation of GC, we sought to reduce rather than abolish enhancer activity. Within the chicken 3′-enhancer, five CANNTG consensus sites for basic-region helix-loop-helix (bHLH) proteins have been identified. We wondered whether overexpression of the dominant-negative Id proteins would suffice to diminish enhancer activity. First, we carried out transient transfection experiments in mouse A20 cells. Fig. 4 demonstrates that co-expression of either mouse Id1 or Id3 will reduce enhancer activity by 50%, indicating that bHLH proteins contribute significantly to enhancer function. Overexpression of the Id proteins is therefore a valid approach to reduce 3′-enhancer function. For a convenient GC assay, we utilised clone 18 DT40 cells. This cell line transcribes its IgL gene at high levels, but is IgM surface negative since the light chain gene carries a frameshift nonsense mutation 26. Upon GC, this lesion is repaired using upstream pseudogenes and the cells revert to being IgM surface positive. Therefore, in clone 18 DT40 cells, GC can easily be quantified by measuring the appearance of IgM-positive cells. We generated stable Id1 and Id3 transformants of clone 18 cells. RT-PCR analysis of selected subclones demonstrated that Id1 and Id3 are transcribed (Fig. 5). In the case of Id1 where commercial antiserum is available, we confirmed by Western blotting that Id1 protein is present in the cells (data not shown). Subsequently, fluctuation analysis was carried out for WT clone 18 cells and two independent Id1- and Id3-expressing clones. Individual subclones were cultured for 4 wk and then analysed for IgM surface expression. Fig. 6A shows that in the presence of Id1 and Id3, the appearance of IgM-positive clones is reduced fivefold. Id3 transfectants appear to be more closely clustered. However, statistical analysis revealed that this spread is not significantly different from that seen for Id1 transformants. Id proteins have been reported to exert their effect via changes in the cell cycle 22. To ensure that reduced GC is not simply due to reduced proliferation rates, we measured cell growth in DT40 clone 18 and the Id1 and Id3 transformants. As Fig. 6B demonstrates, no changes in cell doubling times were observed. Taken together, our data indicate that the 3′-enhancer regulates GC, which is reduced by Id protein overexpression.

E47 strongly up-regulates GC

The major target of Id proteins in B cells is E47, which is inducible in mature B cells 20, 27. We have previously demonstrated that Id1 and Id3 are able to inhibit chicken E47 17, which is the most likely mediator of the effect observed here. As Id acts as a negative regulator of GC, we wondered whether transfection of E47 would suffice to activate GC. A chicken E47 expression vector 17 was stably transfected into clone 18 cells and fluctuation analysis was once again used to assay GC rates. As shown in Fig. 7A, E47 had a dramatic effect on clone 18 cells. In individual E47-expressing subclones, up to 90% of cells became IgM⁺ within the 4 wk of our analysis. The average rate of GC increased from 5% in clone 18 cells to 35% in clone 18 + E47 cells. Two additional, independent transfectants expressing E47 yielded average conversion rates of 31% and 34%. Once again, the growth rates of E47-expressing cells were similar to those of untransfected cells. Thus, E47 is a strong positive regulator of GC.

We wondered whether indirect pathways might be responsible for the effects of Id and E47 on GC. Id3 has previously been reported to prevent the induction of AID gene expression in mouse primary B cells, while E47 activated expression 28. We therefore used real-time PCR to analyse whether chicken AID gene expression is affected by the overexpression of Id or E47 in DT40 cells. We find that, relative to WT clone 18 cells, AID expression levels in the Id1, Id3 and E47 transformants are unchanged (Fig. 7B). Although there appears to be a small up-regulation in all three transformants, statistical analysis of the data using an unpaired t-test indicates that the increase is not significant, yielding p values of 0.36, 0.19 and 0.11 for Id1, Id3 and E47, respectively. Furthermore, we see the same increase for both the Id and E47 transformants while of course GC is down-regulated in the Id- and up-regulated in the E47-expressing clones. We conclude that AID gene expression in the chicken is not significantly affected by the levels of Id1, Id3 or E47 expressed in our system. Several factors could explain the apparent discrepancy between our results and those obtained in murine cells 28. Firstly, the chicken AID gene may be regulated by distinct sequences. Secondly, mouse Id proteins may be less effective in repressing chicken rather than mouse E47. Alternatively, the protein level of Id in our experimental system may not suffice to see a reduction in AID expression levels, although this is unlikely since the expression levels suffice to strongly affect GC. Furthermore, recent reports indicate that transcription factors other than E47 act as critical regulators of AID gene expression. AID gene activity correlates with occupancy of the Pax5 but not the E2A binding sites in this gene 29. In addition NF-κB and STAT-6 have been shown to regulate AID gene expression in activated human B cells 30.

Previous results in cell lines have indicated that SHM in cell lines correlates directly with the rates of transcription of the Ig genes 1, 10. We therefore wished to examine whether the changes in GC observed here correlate with IgL transcription rates. Real-time PCR on the IgL gene demonstrates that both Id1 and Id3 transformants displayed a fivefold decrease in IgL transcription (Fig. 7C). This result is consistent with our transient transfection data showing that Id reduces enhancer activity. Interestingly, reduced expression did not lead to significant changes in surface expression. Presumably, other factors, which operate independently of IgL message levels, control the levels of Ig surface expression. For example, Ig heavy chain, the Igα or Igβ proteins, which are required for Ig surface expression, could be limiting. Our findings that Id overexpression reduces IgL RNA levels in chicken cells are entirely consistent with data obtained in other systems. For example, IgL gene expression is reduced in Id1-transgenic mice 31. The down-regulation observed in the stable transformants is higher (fivefold) than that seen in transient transfections (twofold). One reason for this may lie in the ability of E2A to modify chromatin. E2A is able to interact directly with the SAGA complex, which catalyses histone acetylation 32. This would have a more pronounced effect in stable transfections where the introduced DNA is assembled into higher-order chromatin. In contrast, transiently transfected DNA is unlikely to have acquired such a structure and the effects would not be seen. Secondly, we cannot rule out that there are additional elements within the IgL gene that can respond to bHLH proteins, although no such binding sites have been described.

We next wondered whether the increase in GC after E47 transfection is once again mirrored by IgL transcription rates. As Fig. 7C demonstrates, overexpression of E47 did not cause a significant increase in IgL RNA levels. This is surprising since E47 is thought to be a strong transcriptional activator in mature B cells and there are five potential E2A binding sites within the IgL 3′-enhancer. However, these results are entirely consistent with transient transfection experiments carried out in our laboratory. Using either A20 or HEK293 cells, we were unable to up-regulate IgL 3′-enhancer activity by cotransfection of murine or chicken E47 or E12 (data not shown). These data indicate that E47 protein levels in the cell are not limiting for Ig transcription in mature B cells. In conclusion, our experiments demonstrate that E47 is an important positive regulator of GC. E47 is the first transcription factor identified as an activator of GC whose function does not depend on increasing the rate of transcription.

Upon transfection of E47, GC rates increased sevenfold. We wondered whether such a large increase was due to an increase in the number of cells attempting GC or whether there is an increase in the number of GC events per V gene. To this end, we sequenced the V genes of three clones of unsorted clone 18 + E47 cells, which had been cultured for 4 wk. These clones had become 40.2%, 85% and 90.8% IgM surface positive. GC events are shown in Fig. 8. Our data clearly demonstrate that the increase in surface IgM-positive cells correlates with an increase in the number of GC events per V gene sequence. While the data set is relatively small, there is no doubt that the number of GC events per sequence is dramatically higher than those obtained with unsorted, untransfected clone 18 cells, where less than five conversion events would be expected per 100 sequences.

Our results show that Id reduces both GC and transcription, while E47 increases GC but does not alter Ig message levels. One reason for this differential regulation may be the ability of Id proteins to interact with other families of transcription factors such as Ets and Pax proteins 29, 33, 34. Extensive functional analysis has not been carried out on the IgL 3′-enhancer and it may well be possible that members of the Pax and Ets families contribute to IgL 3′-enhancer activity. Interference with these could cause reduced Ig transcription. Furthermore, the differential regulation of GC and gene transcription could be achieved if E47 is not limiting for Ig transcription but is limiting for GC. In this scenario, Id would reduce expression through a combination of inhibiting E47 and other transcription factors. Reduced transcription levels would then correlate with reduced GC. Upon overexpression of E47, transcription levels would not increase since E47 is not limiting, yet GC would increase in line with higher E47 levels. Similarly, Storb and colleagues observed that the inclusion of a CANNTG consensus bHLH binding site in a synthetic V transgene increased the level of somatic hypermutation in this substrate without a notable increase in transcription rates 21, supporting the notion that distinct mechanisms regulate transcription and gene modification. However, it is also possible that E47 does increase the rate with which RNA polymerase is recruited. This could lead to increased recruitment of AID, introducing lesions that render the newly transcribed message unstable. The steady state level of the Ig RNA would therefore not change. However, this is unlikely since treatment of clone 18 cells with trichostatin A increases GC with a concomitant increase in transcription 35, suggesting that increased rates of transcription would be detectable in DT40 clone 18 cells. We therefore favour a model whereby the recombination in GC can be activated independently of transcription. In line with this model, deletion of the recombinase-activating sequence in the mouse Igκ locus reduces κ rearrangement without affecting the levels of transcription 9.

As yet, we do not understand the mechanism by which E47 acts to increase GC. We have excluded the possibility that the observed effect is due to an increase in AID expression and have demonstrated that IgL message levels do not change. E47 may also act directly to recruit the AID complex to the locus. Such an interaction may require intermediary proteins. Alternatively, E47 may exert its effect through changes in the chromatin structure. Effects of E47 on chromatin structure are well documented 32, 36, 37, although we do not as yet understand how changes in chromatin structure affect AID activity. There is evidence that another epigenetic modification, namely DNA demethylation, is associated with increased mutability 38, suggesting that other changes in chromatin structure may also affect GC and SHM.

Plasmids

To generate the MAR knockout vector, the rearranged IgL V gene was amplified by PCR using primers P1 and P2 (see Table 1 for all primer sequences). The flanking restriction enzyme sites Xho I and Sal I were used to subclone the resultant 1.2-kb fragment into a pBSII(KS)+ vector, in which the Hind III site had been replaced by a Bgl II site. Into this vector, we inserted a 2.5-kb genomic Eco RI fragment spanning the CL region of the IgL gene. A Bam HI fragment carrying the neomycin gene with flanking loxP sites (obtained from J. M. Buerstedde) was subsequently inserted into the Bgl II site (see Fig. 1A). To generate the 3′-enhancer knockout construct, the primer pairs P3/P4 and P5/P6 were used to amplify the sequences 3′ and 5′ of the IgL 3′-enhancer. These were inserted into the pBSII(SK+) vector using the restriction sites engineered into the primers. The floxed neomycin gene was inserted into the Bam HI site of the resultant vector. A 2.5-kb Eco RI fragment spanning the CL gene was subsequently inserted into the 5′ Eco RI site of our construct.

In order to introduce the Cre recombinase into the knockout cell lines, we subcloned the Hind III fragment of the pANmerCREmer vector 39 into the Hind III site of the pcDNA3.1/hyg vector (Invitrogen). Chicken E47 and mouse Id1 and Id3 expression constructs have been described 17. The chicken IgL 3′-enhancer was amplified using primer pair P7/P8, subcloned into the pCR2.1TOPO vector (Invitrogen) and then transferred to the pGL3promoter luciferase reporter vector (Promega) using Sac I and Bgl I.

Transfection and cell culture

Subclones of the chicken bursal B cell line DT40 26, 40 were maintained in RPMI medium with 7% foetal calf serum and 3% chicken serum, penicillin and streptomycin, and 50 µM β-mercaptoethanol. The subclone used carries a heterozygous mutation of the XRCC2 gene. Heterozygocity in this locus does not have any known phenotype with regard to gene expression. For stable transfection, 1 × 10⁷ cells were electroporated (500 V, 25 µF) with 20 µg linearised plasmid and selected using 2 mg/mL Geneticin or 1.0–1.4 mg/mL hygromycin. Induction of Cre expression was carried out at 100 nM 4-hydroxy-tamoxifen for 3 days. Homologous recombinants were identified either by Southern blotting (Maniatis) or by PCR using primer pair P9/P10. A20 cells (mouse B lymphocytes) were maintained and transfected as described 17 using 10 µg reporter plasmid, 10 µg effector plasmid and 5 µg of the internal control vector CMV-β-gal. Luciferase and β-galactosidase assays were carried out as reported 17.

FACS analysis

Cells were stained with FITC-conjugated goat anti-chicken IgM antibody (Bethyl Laboratories), at a dilution of 1 : 100. The percentage of FITC-positive whole cells (as determined by forward scatter and side scatter) from a total of 10 000 events was determined using a FACSCalibur (BD) analyser. FITC-conjugated goat anti-IgM antibody-stained cells were also sorted on a Mo-Flo cell sorter (Dakocytomation). To determine the rate of cell growth, equal volumes of cells from the various subclones were analysed on a BD-FACSCalibur flow cytometer with an equal volume of BD-Calibrite beads. The number of live cells counted while acquiring 5000 beads was recorded at the time points shown in Fig. 6B, with the ratio of cells to one bead displayed.

PCR analysis

To assess gene expression levels, we isolated RNA on RNeasy columns (Qiagen) and carried out reverse transcription reactions using 2 µg RNA. The RNA was annealed to oligo(dT) and reverse-transcribed using Super RT enzyme (HT Biotechnology, UK). Real-time PCR was carried out using the QuantiTect SYBR Green PCR kit (Qiagen, UK) in a volume of 50 µL. Samples were denatured at 94°C for 15 min. Then, 40 cycles of 15 s denaturation at 95°C, 30 s annealing at 65°C and 30 s extension at 72°C were carried out. The primer pair P11/P12 was used to detect IgL expression and the primer pair P13/P14 to detect AID expression. Values were normalised to β-actin expression detected by primer pair P15/16. Samples were run and analysed on the 7000 Sequence Detection System (Applied Bioscience). Id expression in stably transfected cells was analysed as described 17. The primer pair P17/18 was used to detect Id1 expression, primer pair P19/20 Id3 expression, and primer pair P21/22 β-actin expression.