Homologous Elements hs3a and hs3b in the 3′ Regulatory Region of the Murine Immunoglobulin Heavy Chain (Igh) Locus Are Both Dispensable for Class-switch Recombination*

Immunoglobulin heavy chain (IgH) genes are formed, tested, and modified to yield diverse, specific, and high affinity antibody responses to antigen. The processes involved must be regulated, however, to avoid unintended damage to chromosomes. The 3′ regulatory region of the Igh locus plays a major role in regulating class-switch recombination (CSR), the process by which antibody effector functions are modified during an immune response. Loss of all known enhancer-like elements in this region dramatically impairs CSR, but individual element deletions have no effect on this process. In the present study, we explored the hypothesis that an underlying functional redundancy in the homologous elements hs3a and hs3b was masking the importance of either element to CSR. Several transgenic mouse lines were generated, each carrying a bacterial artificial chromosome transgene that mimicked Igh locus structure but in which hs3a was missing and hs3b was flanked by loxP sites. Matings to Cyclization Recombination Enzyme-expressing mice established “pairs” of lines that differed only in the presence or absence of hs3b. Remarkably, CSR remained robust in the absence of both hs3a and hs3b, suggesting that the remaining two elements of the 3′ regulatory region, hs1.2 and hs4, although individually dispensable for CSR, are, together, sufficient to support CSR.

The immunoglobulin heavy and light chain (IgH and IgL) loci undergo multiple rounds of DNA recombination and somatic hypermutation as B lymphocytes form and then perfect their receptors for antigen. The process of heavy chain class-switch recombination (CSR) 3 takes place upon antigen stimulation of the B cell and allows for the development of daughter clones that can produce antigen-reactive antibody with the appropriate properties to combat the pathogen in question (e.g. ability to traffic to appropriate sites in the body; ability to directly destroy the organism through complement-mediated lysis; etc.).
Although activation-induced cytidine deaminase has been identified as an essential protein, central to both CSR and the somatic hypermutation of variable-region encoding genes (1), the means by which activation-induced cytidine deaminase activity is recruited to the appropriate DNA targets within the Ig loci remains largely unknown. Several studies have shown that the 3Ј regulatory region (3ЈRR) plays an important role in class-switch recombination, implicating it in this recruitment process (2)(3)(4)(5). The 3ЈRR is a series of DNase hypersensitivity sites, initially identified as enhancer-like elements, that map downstream of the Ig heavy chain constant region genes (6 -11) (see Fig. 1A). Analogous elements have been described downstream of each of the two C␣ genes in the human Igh locus (12,13), arguing that their function is evolutionarily conserved.
Class-switch recombination is preceded by transcription from "intronic" promoters that lie upstream of the constant region genes. These transcripts are termed "germline" (GLT) as they are generated prior to the DNA rearrangement between S (switch) regions that juxtaposes an assembled V H with a new C H (e.g. between S and S␥2a to place the V H upstream of C␥2a). Because of G/C-rich sequences in S regions, RNA transcripts remain associated with the template strand, displacing the G-rich, non-template strand (R loop formation), thereby providing a single-stranded DNA substrate for activation-induced cytidine deaminase (14,15). T lymphocytes "instruct" B cells to switch to particular heavy chain isotypes through the secretion of cytokines that, in turn, signal the activation of particular intronic promoters and the production of germline transcripts (e.g. IL-4 prompts ␥1 and ⑀ GLTs and a switch to IgG1 or IgE) (reviewed in Ref. 16).
The first indication that the 3ЈRR played a role in classswitch recombination arose from an analysis of mice in which the hs1.2 region (see Fig. 1A, 1,2) was replaced with a selectable marker gene (neo R ). In B cells carrying this modified Igh locus, germline transcripts from the ␥2a, ␥2b, ␥3, and ⑀ constant regions genes were reduced or undetectable, and class-switch recombination to these isotypes was significantly impaired (17). Although this defect turned out to be due to marker gene insertion and not to hs1.2 deletion (18), subsequent work supported the underlying suggestion that the 3ЈRR was important to the CSR process and that perturbation of the 3ЈRR interrupted proper functioning of that process. Most notably, deletion of the pair of elements hs3b and hs4 (Fig. 1A, 3b and 4), had a profound effect on CSR to all classes except IgG1, and deletion of the 28 kb spanning hs3a, hs1.2, hs3b, and hs4 dramatically reduced CSR to IgG1 as well (2)(3)(4)(5). Chromosome conformation capture techniques have revealed a cytokine-induced, tripartite interaction involving the Ig transcription unit, the relevant intronic promoter for germline transcripts, and the 3ЈRR, providing physical evidence for the role of 3ЈRR in the classswitch recombination process (19).
In the present study, we have sought to extend analysis of the 3ЈRR to more precisely determine which elements are expendable for CSR and, reciprocally, which are sufficient to maintain it. Others have recently shown that although deletion of both hs3b and hs4 profoundly reduces CSR to most IgH classes, either element deleted individually has no such effect (20,21). Because hs3b is an inverse replicate of hs3a, however, we thought it possible that the importance of these two elements to CSR was masked in single-element deletions because of an underlying functional redundancy. Notably, there is only one hs3-like element in the human 3ЈRR.
To address this hypothesis, we generated several independent transgenic mouse lines carrying a large segment of the Igh locus as transgene. In all the mouse lines, the Igh locus transgene lacked hs3a, which was shown previously to be individually dispensable for CSR (18). The Igh locus transgene also carried a loxP-flanked version of hs3b so that each mouse line could also generate a paired "sister" line in which both hs3a and hs3b were missing. Remarkably, we found that class-switch recombination took place efficiently within the bacterial artificial chromosome (BAC), whether or not the second hs3 element (hs3b) was present. The remaining DNA sequences, which included the 3ЈRR elements hs1.2 and hs4, were sufficient for robust class-switch recombination, although hs1.2 and hs4 are each individually dispensable for CSR (18,21). These findings demonstrate a remarkable flexibility in the structure/composition of the 3ЈRR capable of supporting classswitch recombination. Moreover, as discussed in more detail below, these findings point toward a new approach for further unraveling the mode of action of the 3ЈRR in this important process.

IgH⌬3alx3b BAC Construction
A BAC that carried the immunoglobulin heavy chain constant region genes of the murine Igh locus (129 mouse strain) was modified to create a functional IgH gene, delete the 3ЈRR sequence hs3a, and flank the 3ЈRR sequence hs3b with loxP sequences. The BAC was designed so that after loxP-mediated deletion of hs3b, it would lack the regions of homology shared by hs3a and hs3b (nt 1192-2154 and nt 24019 -24989, respectively; GenBank TM accession number AF450245). Development of the BAC was achieved through a sequence of homologous recombination events in bacterial cells (described below). Primer sequences used to clone specific fragments of DNA are provided in the supplemental data.
CHC BAC-CHC BAC is a BAC containing ϳ230 kb of DNA from the murine Igh locus. It begins 11 kb 5Ј of the J H cluster and extends 35 kb 3Ј of hs4.
V H Insertion-A 2.2-kb DNA segment containing the assembled B1-8 V H gene was cloned from the plasmid pIV H B1-8L2neo r (22). The encoded V H corresponds to that found in the (4-hydroxy-3-nitrophenyl) acetyl-binding antibody B1-8. The B1-8 V H gene used in the current studies carries a silent point mutation (codon 92) that inactivates the internal recombination signal sequence, eliminating the possibility of V H replacement (22).
The B1-8 V H gene fragment replaced the J H gene region (inserted between nt 1000 and 2350, GenBank accession number J00440). Upstream and downstream homology sequences were joined to the 2.2-kb fragment containing B1-8 V H using a PCR strategy (23) and inserted into the pSV1-RecA shuttle vector for homologous recombination with CHC BAC (methods of Yang et al. (24)). The resulting BAC, which lacks the J H region, was designated VDJC.
BAC constructs were analyzed at each stage of the modification process, both for the presence of the expected changes and for the absence of any other, unwanted modifications. One method used to screen for unwanted modifications was digestion with BamHI, EcoRI, and HindIII, respectively, and comparisons of the ethidium bromide-stained fragment profiles (data not shown). Expected changes induced by homologous recombination were confirmed by Southern blot (see "Results" and supplemental Fig. 1).

Generation of Transgenic Mouse Lines
IgH⌬3alx3b BAC was digested with NotI, and the insert was purified before transfer to the Mouse Genetics Core Facility, Memorial Sloan Kettering Cancer Center (New York, NY) for microinjection. Transgenic founders were identified in PCR 3RR Elements hs3a/hs3b Dispensable for CSR screens, using primer pairs specific for B1-8 V H and for loxPflanked hs3b (lx3b), respectively (primer sequences are provided in the supplemental data). Animals were bred and maintained in animal facilities at Hunter College (Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC)-accredited), City University of New York, and all mouse experiments were approved by the Hunter College Institutional Animal Care and Use Committee.
For genomic Southern screens for hs3b deletion, DNA was digested with BamHI and hybridized to hs3b (1.2-kb XbaI fragment from HS3.4 plasmid) (10). A 0.9-kb BamHI fragment was lost upon hs3b deletion. Genomic DNA was digested with HindIII and hybridized to hs3bD, a sequence that lies downstream of hs3b. A HindIII fragment detected by hs3bD increased from 3.8 to 4.2 kb upon hs3b deletion.

Checking BAC Integrity in Transgenic Mouse Lines
Methods for checking the integrity of the V H region and hs3b region were described above ("Generation of Transgenic Mouse Lines"). For other regions (S␥2a, S␥2b, hs4, hs1.2, C⑀), the strategy of Dunnick et al. (28) was followed (supplemental Fig. 3 and data not shown). The presence of functional constant region genes was further confirmed by expression (e.g. Ig a in mice with Igh b genetic background) and/or presence of allelespecific germline transcripts (see below). DNA probes used in supplemental Fig. 3 are as follows: S␥2b, 5.2-kb XbaI fragment from pBR1.4 plasmid (25, 29); Hs4, 2.4-kb PstI/HindIII fragment from HS3.4 plasmid (10).

Southern Blot Analyses
Southern blot analyses were performed as described previously, using ϳ10 -30 g of mouse tail-tip DNA or liver DNA (30).

Enzyme-linked Immunosorbent Assays (ELISA)
Mouse sera were assayed for immunoglobulin isotype levels, using a sandwich ELISA and standard methods. Most antisera were supplied by BD Biosciences (see supplemental data for a list of all antisera and monoclonal antibodies used).
All sera were diluted in duplicate, and each dilution was analyzed in duplicate (four tests/serum sample). Individual readings generally deviated less than 10% from the mean. Standard curves for each assay were generated with purified mouse immunoglobulin of the relevant class. Plates were read at wavelength 405 nm, using a PowerWave HT microplate spectrophotometer (BioTek, Winooski, VT), and data were analyzed by the software Gen5 (BioTek). Statistical tests for serum concentration differences among the mouse lines were performed with the GraphPad Prism software (version 5.0c).

In Vitro Stimulation of Splenic B Cells
Resting splenic B cells were enriched by negative selection (antibody to CD43), using a B cell isolation kit (catalog number 130-090-862; Miltenyi Biotec, Bergisch Gladbach, Germany). Experiments were performed in parallel on cells isolated from WT, IgH⌬3alx3b, and IgH⌬3a⌬3b transgenic mice. ϳ3 ϫ 10 6 B cells were cultured in each of the following conditions to induce class-switching to specific isotypes.
IgG1/IgE-Conditions were as above but with the addition of IL-4 (10 ng/ml) (PeproTech, catalog number 214-14). Cells were harvested and analyzed on day 4.

Flow Cytometry
Methods were essentially as described previously (32). For expression of IgM, spleen cells were stained with antibodies to B220, IgM a , and/or IgM b . B cells stimulated to undergo classswitch recombination were stained with antibodies for B220 and the relevant IgH class. A list of the reagents used for all flow cytometry experiments is provided in the supplemental data.
In each staining experiment of stimulated B cells, an isotypematched control antibody (with no specificity for murine cells or Ig) was included. For example, the FITC-conjugated rat IgG2a, monoclonal antibody specific for murine IgG3 (BD553403) was matched with a control rat IgG2a, monoclonal antibody with irrelevant specificity (BD554688).

Reverse Transcription-Polymerase Chain Reactions (RT-PCRs) to Detect Germline Transcripts
RNAs from stimulated splenic B cells were extracted with an RNeasy TM mini prep kit (Qiagen, Valencia, CA, catalog number 74106), and RT-PCR was performed with a OneStep RT-PCR kit (Qiagen, catalog number 210212). Primers specific for the germline transcripts from individual C H regions were as described (28) and are shown in supplemental Tables 1 and 2. Supplemental Table 1 provides restriction enzymes used to distinguish RT-PCR products from "a" allotype and "b" allotype transcripts, respectively. For IgA, germline transcripts from 3RR Elements hs3a/hs3b Dispensable for CSR AUGUST 5, 2011 • VOLUME 286 • NUMBER 31 both alleles were first amplified with consensus primers (I␣f and ALPHMURCH2). The products from this reaction were used as template with primers specific for the Igh a allotype (I␣f and ALPHMURCH2A), generating a secondary (or nested) PCR product only when ␣-germline transcripts from the Igh a chromosome were present.

Chromatin Conformation Capture Assays
Methods were as described previously (33). Briefly, isolated B cells were harvested either before culture or after stimulation for 48 h in LPS or in LPSϩIL-4. Harvested cells were incubated in formaldehyde to cross-link interacting proteins, nuclei were prepared, and non-cross-linked protein was removed by SDS. Triton X-100 was added to sequester the SDS, after which samples were digested with HindIII. Digested nuclei were subjected to intramolecular ligation and then treated with proteinase K and RNase A, after which DNA was purified. One hundred nanograms of ligated DNA were analyzed by 32 cycles of PCR (94°C/30 s, 56°C/30 s, and 72°C/30 s) in a 25-l reaction (PCR amplifications were confirmed to be in the linear range). A positive control template consisted of a mixture of the IgH⌬3alx3b BAC with two calreticulin (CalR) fragments in equal molar ratios. The mixture was subjected to HindIII digestion and then DNA ligation and was used to normalize PCR efficiency among the different primer pairs. The negative control template was treated in the same way but consisted of a BAC containing unmodified, endogenous Igh locus sequences mixed with CalR fragments. loxP sequences and a ␤-globin tag sequence unique to the IgH⌬3alx3b BAC were used to design a chromatin conformation capture primer that specifically tracked the hs4 segment of the BAC transgenic allele (see supplemental Table 3). As expected, this transgene-specific 3Ј primer generated PCR products when used with the various 5Ј primers (for E, I␥1, I␥3, and hs1.2) on the positive control but not on the negative control template (confirming its specificity).
PCR products were size-fractionated by electrophoresis (2% agarose gel), and gel pictures were analyzed by the SYNGENE GeneGenius bioimaging system. Cross-linking efficiency between two HindIII fragments was calculated as described previously (the ratio of the signal for PCR product from the IgH locus/signal for PCR product from CalR in a given cell type, divided by this ratio in the positive control template).

RESULTS
Construction of IgH⌬3alx3b BAC-In a previous study, it was shown that CSR was unimpaired in B cells from mice lacking the 3ЈRR element hs3a (18). More recently, deletion of hs3b alone was also shown to have no effect on CSR (20). To test the possibility that these nearly identical elements are functionally redundant, we established several, independent mouse lines with a fully functional, but hs3a-deficient, Igh locus transgene (BAC) and then tested whether subsequent loss of hs3b affected locus expression and/or CSR.
A bacterial artificial chromosome containing much of the murine Igh locus (CHC BAC) (28) was first modified to replace the J H region with an assembled V H gene (Fig. 1A). In subsequent steps, hs3a of the 3Ј regulatory region was deleted, hs3b was flanked by loxP sites, and a short human DNA sequence (397-bp ␤-globin cDNA segment) was inserted to uniquely tag this region of the BAC (Fig. 1A, TAG). hs3a deletion corresponded to the nucleotides shared by hs3a and hs3b, and, reciprocally, the hs3b region between loxP sites included all nucleotide sequences shared by these two 3ЈRR elements. The resulting BAC (IgH⌬3alx3b) extended from a site ϳ11 kb upstream of the J H cluster to a site ϳ35 kb downstream of hs4. Included in the BAC were D Q52 , the D H gene segment most proximal to the J H region, and all of the 3ЈRR elements thus far identified (hs1-7) (11) (Fig. 1A). Deletion of the J H region precluded D Q52 -J H recombination events on the BAC.
ods of Yang et al. (24). Southern blots were performed to confirm that the appropriate modifications had occurred. For example, the DNA probe J H D (Fig. 1B, black rectangle downstream of the J H region) revealed a 6.6-kb BamHI band in the original (CHC) BAC, which was replaced with an 8.5-kb BamHI band in IgH⌬3a,lx3b BAC (Fig. 1B). Expected changes in the EcoRI and HindIII fragments detected with this probe further confirmed that the V H insertion had occurred as planned. The modifications involved in deleting hs3a and flanking hs3b with loxP sites and tag DNA were also confirmed by Southern blot analyses of the BACs (supplemental Fig. 1). A broader analysis of BAC DNA, using ethidium bromide-stained gels of enzymecleaved BACs, revealed no unexpected changes (data not shown). As described below, several regions along the BAC were further checked for integrity in transgenic mice.
Generation of IgH⌬3alx3b Transgenic Mice-Twenty-two founder mice with apparently intact IgH⌬3alx3b BAC transgenes were recovered, and seven of these (lines 24, 60, 69, 79, 81, 83, and 111) were maintained and mated to C57BL/6J for further study. As will be described below, all seven of the transgenic lines showed robust expression of the IgH⌬3alx3b BAC. Five of the lines (60, 69, 79, 81, and 83) were further analyzed to confirm intact BAC structure in regions mapping from the assembled V H gene through to the loxP-flanked hs3b (supplemental Figs. 2 and 3). Allelic differences between the BAC that carried Igh a sequences derived from the 129 mouse strain and the endogenous Igh b loci (C57BL/6-derived) aided in these analyses (3). Genomic Southern blots also provided an estimate of BAC copy number in each of the lines (60, 69, 79, and 83 had 1-3 copies of the BAC, whereas 81 had many) (supplemental Fig. 3 and data not shown).

Allelic Exclusion of the Endogenous (Igh b ) Loci in IgH⌬3alx3b
Transgenic Mice-Expression of the IgH⌬3alx3b BAC was first analyzed by flow cytometry in transgene-positive mice. Allotype-specific reagents demonstrated that all seven lines analyzed (lines 24, 60, 69, 79, 81, 83, and 111) expressed transgenic Ig a in the spleen but little or no Ig b ( Fig. 2A, and data not shown), suggesting that the functional IgH⌬3alx3b BAC was effectively inhibiting assembly and expression of IgH genes within the endogenous Igh b loci.
The level of surface IgM expression on splenic B cells closely matched that seen in normal mice expressing an Igh a allele. As shown in the histogram in Fig. 2A, for example, spleen cells in the transgenic line carrying one copy of the BAC (lx3b-69) expressed a at the same level as spleen cells from a 129P2/ Olahsd (129) mouse.
It should be pointed out that the site-independent expression of IgH⌬3alx3b was not dependent upon the introduction of heterologous insulator sequences. In studies by others, the chicken ␤-globin locus-derived insulator was inserted at the 5Ј end of an Igh BAC to assist in generating integration site-independent expression (4). The authors noted, however, that another BAC modification that may have explained better expression of this BAC relative to earlier versions was removal of an upstream D H element, preventing deletion of the assembled V H by D H -J H recombination. As described above, IgH⌬3alx3b was designed to preclude B1-8V H deletion by D H -J H rearrangement as well, suggesting that this, rather than the presence of the ␤-globin locus-derived insulator, improved expression in the prior study (4). Notably, CTCF-associated insulator sequences lie just downstream of hs4 (11).
Deletion of hs3b within Four Independent Transgenic Mouse Lines-Having confirmed that lines 60, 69, 79, and 83 carried one or more copies of the intact BAC transgene, we bred these lines (each assumed to carry the BAC at a unique integration site) to mice carrying the bacterial cre gene under control of the adenovirus EIIa promoter (expressed in early embryos) (26). In those progeny that inherit both the BAC and the cre gene, the hs3b element within the BAC transgene should undergo loxPmediated deletion (Fig. 1A, diagram). Progeny were screened for hs3b deletion by PCR (see "Experimental Procedures," data not shown). hs3b deletion within the BAC was further confirmed by Southern blot (Fig. 1C and supplemental Fig. 2B). As maps of this region show (Fig. 1C), a probe downstream of hs3b (3bD; black rectangle under maps) should anneal to the BAC both before and after hs3b deletion but revealing HindIII fragments of distinguishable sizes (3.8 and 4.2 kb, respectively). An example of such analyses for four independent transgenic lines (60, 69, 79, and 83), before and after hs3b deletion, is shown in Fig. 1C. Using the 8.1-kb HindIII fragment derived from the endogenous (C57BL/6-derived) Igh loci for normalization, we confirmed that BAC copy number was unchanged upon hs3b deletion. After hs3b deletion, the BAC lacks all the sequences shared by hs3a and hs3b (see "Experimental Procedures"). As shown in Fig. 2A, deletion of hs3b in combination with the absence of hs3a did not affect allelic exclusion (see contour plot for ⌬3a⌬3b-69), nor did it affect surface IgM levels (compare lx3b-69 and ⌬3b-69 in the histogram).
Class-switch Recombination in the Absence of Both hs3a and hs3b-Resting, splenic B cells were isolated from pairs of transgenic mice and assayed for class-switch recombination by flow cytometry ("Experimental Procedures"). As noted above, BAC copy number was not changed upon hs3b deletion, and in these and in all subsequent experiments, transgenic mice were kept hemizygous for the BAC transgene (carrying only one transgene allele) to ensure that matched pairs always carried the same BAC copy number.
Initially, the single-copy transgenic line 69 on the C57BL/6J background was assayed for CSR, and it was found that the percentage of switched cells did not change significantly upon hs3b deletion (representative results, supplemental Fig. 4). Consistent with these flow cytometry data, germline transcripts emanating from each of the BAC heavy chain constant region genes (a necessary precursor to CSR) were detectable in appropriately stimulated B cells, whether or not the BAC retained hs3b (Fig. 2B). Allelic differences permitted distinction between BAC-derived and wildtype Igh loci-derived transcripts (3).
We extended our flow cytometry analyses of CSR in the presence and absence of hs3b after back-crossing all four pairs of transgenic mouse lines to the Igh Jtm1Cgn /J mouse line (cited here as ⌬J H /⌬J H ). The ⌬J H chromosome lacks all of the J H genes and the intronic enhancer (E), rendering it incapable of V H gene assembly (27). In the BAC transgenic ⌬J H /⌬J H mice, therefore, the BAC transgene is the only source of Ig heavy chain and is responsible both for development of B cells and for surface Ig expression in stimulated B cell cultures. AUGUST 5, 2011 • VOLUME 286 • NUMBER 31

3RR Elements hs3a/hs3b Dispensable for CSR
Representative flow cytometry data for CSR-stimulated B cells from the lx3b-83/⌬3b-83 pair on this ⌬J H /⌬J H background are shown in Fig. 3. Class-switch recombination was cytokinedependent and robust in these animals, whether hs3a was the only 3ЈRR element missing (⌬3alx3b-83) or both hs3a and hs3b were absent (⌬3a⌬3b-83) (Figs. 3 and 4A and supplemental Fig.  5). As summarized in Fig. 4A, this proved true for all of the transgenic mouse lines; with the exception of switching to IgG2b, there was no significant difference in CSR between the matched pairs, regardless of the BAC integration site. In the case of switching to IgG2b, the difference was statistically significant (paired, two-tailed t test; p Ͻ 0.02) but small as the percentage of cells switching to this isotype was reduced by only ϳ20% when hs3b was removed (means ϭ 44 versus 36).
In all these experiments, B cells from wild-type animals (siblings of the BAC transgenic mice) or from mice carrying an endogenous locus knock-in of the same V H gene used in the BACs (V H E) (22,32) were used as controls to ensure that culture conditions were adequate to induce class-switch recombination. The percentage of cells switching in wild-type or V H knock-in animals was similar to that in the BAC transgenic mice (generally differing by no more than 2-fold), indicating that CSR on the BACs was very efficient (data not shown).
Ig Serum Levels Remain the Same, before and after hs3b Deletion-As another measure of class-switch recombination, we compared serum levels of the various Ig heavy chain classes in mice carrying either the IgH⌬3alx3b or the IgH⌬3a⌬3b BAC (on the ⌬J H /⌬J H background). As shown in Fig. 4B, the mean concentrations did not differ significantly between these two groups of mice. Paired comparisons of the mice from a single line (e.g. line 60 mice with or without hs3b) were also done, again with no significant difference detected (data not shown). The serum phenotypes of these mice, therefore, are in concordance with the in vitro CSR results; CSR proceeds equivalently in IgH⌬3alx3b (no hs3a) and IgH⌬a⌬3b (no hs3a or hs3b) transgenic mice.
Physical Interactions between I Region Promoters and 3ЈRR Persist in the Absence of hs3a and hs3b-Prior studies have shown that the 3ЈRR physically associates with the IgH transcription unit in resting B cells and shows a cytokine-enhanced association with the intronic promoters upstream of the various heavy chain constant region genes (19). Given that CSR continued unabated on the IgH⌬3a⌬3b BAC, we looked for evidence of physical association between this truncated 3ЈRR (retaining only hs1,2 and hs4) and upstream regions involved in CSR. Matched pairs from two independent mouse lines (79 and 83) were analyzed in two sets of experiments. As shown in Fig. 5, the 3ЈRR of the IgH⌬3alx3b BAC mimicked the behavior of a normal Igh locus: strong association prior to stimulation among the 3ЈRR elements (hs1.2 and hs4) and between the 3ЈRR and the IgH transcription unit (E-region) and cytokine-dependent increases in the association of the 3ЈRR with the intronic promoters of C␥1 and C␥3 in an isotype-specific manner (i.e. maximum cross-linking efficiency between I␥3 and hs4 when stimulated by LPS and between I␥1 and hs4 when stimulated by LPSϩIL-4; Fig. 5, B and C). When hs1,2 and hs4 were the only remaining elements in the 3ЈRR (IgH⌬3a⌬3b), all of these associations still took place. In fact, cross-linking efficiency among the elements was generally higher in nuclei isolated from these cells, even without cytokine. Interestingly, the associations showed no isotype-specific response to cytokine. As described above in the flow cytometry studies, however, the B cells of IgH⌬3a⌬3b mice continued to show an isotype-specific response to cytokine with respect to class-switch recombination.  Plots show the percentage of switched cells obtained in a total of 11 CSR experiments (two or more with each transgenic line). Mice were homozygous for J H region and E deletion on the endogenous Igh loci (⌬J H /⌬J H ), preventing IgH gene assembly within these loci. Horizontal lines denote mean; S.D. is provided above each plot. In every case, IgH⌬3alx3b (⌬3alx3b) and IgH⌬3a⌬3b (⌬3a⌬3b) mice were tested in parallel. Data were analyzed by paired, two-tailed t test (GraphPad Prism 5.0). p Ͻ 0.05 for IgG2b (20% reduction in the percentage of switched cells in IgH⌬3a⌬3b mice). B, graphs comparing serum Ig levels in mice expressing the IgH BAC with or without hs3b. Plots are of serum levels of IgH isotypes measured in 40 individual IgH⌬3alx3b mice (⌬3a) and 17 individual IgH⌬3a⌬3b mice (⌬3a⌬3b). As in A, BAC transgenes were carried on the ⌬J H /⌬J H genetic background. Data were analyzed by two-tailed t test (GraphPad Prism 5.0); no significant differences were found. Mean concentration, with S.D., is provided above each scattered dot plot.

DISCUSSION
In the present study, we tested the hypothesis that an essential role for the hs3a/hs3b elements in class-switch recombination was being masked by their sequence identity and underlying functional redundancy. Deletions were designed to fully encompass the regions of shared identity (95%) between hs3a and hs3b (nt 1192-2154 and 24019 -24989, GenBank accession number AF450245). The IgH⌬3alx3b BAC introduced into animals already carried the deletion of hs3a as earlier studies had shown that this deletion had no substantial effect on IgH gene expression or class-switch recombination (18). Consistent with those findings, the IgH⌬3alx3b transgene was expressed efficiently in all of the transgenic lines analyzed. When on the C57BL/6 background (Igh b ), the BAC (Igh a ) was expressed to the exclusion of the endogenous locus and at wild-type levels, as judged by surface IgM expression in splenic B cells ( Fig. 2A). When on the ⌬J H /⌬J H background, B cell development was normal, demonstrating that heavy chain gene expression from the hs3a-deficient BAC was sufficient to drive this process (data not shown).
Remarkably, these observations were duplicated in mice carrying the BAC that lacked both hs3a and hs3b. Class-switch recombination, as measured in vitro, and serum levels of the various Ig isotypes were also indistinguishable, whether B cells were dependent upon the IgH⌬3alx3b or the IgH⌬3a⌬3b transgene for Ig expression. Although it is possible that some of the IgH expressed by the B cells in these mice resulted from transallelic switching (V H on the BAC, switching to the C H genes within the ⌬J H /⌬J H or Ig b /Ig b endogenous loci), it has been demonstrated that such switching is rare between a BAC and an endogenous Igh chromosome when assayed in culture (4). Moreover, when the transgenes were on the C57BL/6 background (Ig b /Ig b ), germline transcripts from the BACs could be specifically examined. These BAC-derived GLTs were detected in both IgH⌬3alx3b and IgH⌬3a⌬3b B cell cultures, supporting the flow cytometry data (Fig. 2B).
Consistent with the ability of the IgH⌬3a⌬3b transgene to undergo CSR, its "mini"-3ЈRR (hs1.2 and hs4) could be seen to physically associate with both the IgH transcription unit and the C H region promoters. In contrast, earlier studies showed that deletion of both hs3b and hs4 (leaving only hs3a and hs1.2) resulted in a dramatic reduction in these associations, along with diminished CSR to all isotypes except IgG1 (2, 19) Similarly, a "technical knock-out" of the 3ЈRR in a cell line resulted in loss of its association with the IgH transcription unit and loss of IgH gene expression (33,34). The present studies, therefore, support a developing model of CSR in which long range interactions involving the 3ЈRR play an essential role.
Interestingly, the interaction between hs1.2 and hs4 increases upon loss of both hs3 elements, and this increased interaction extends to that between the 3ЈRR and the IgH transcription unit (E) and the I␥3 and I␥1 promoters. The increased interaction between the 3ЈRR and I␥3 and I␥1 promoters in IgH⌬3a⌬3b B cells matches the levels seen only after cytokine stimulation in the IgH⌬3alx3b B cells. Importantly, CSR itself remains cytokine-dependent, demonstrating that augmented 3ЈRR/intronic promoter interactions do not supplant the role of cytokine. This is most likely because the I␥3 and I␥1 promoters themselves require cytokine-induced activation of promoter-specific transcription factors. The finding that the hs1.2/hs4 interaction is increased upon loss of both hs3a and hs3b further emphasizes the dynamic interplay taking place among elements in the 3ЈRR region, allowing for the kind of functional flexibility that underlies the functional redundancy revealed by single and (in the present study) double element deletion.
To summarize the experimental findings to date on the role of the 3ЈRR in class-switch recombination, deletion of all the enhancer-like elements (deletions of hs3a through hs4) impedes CSR to all IgH classes (3)(4)(5). A much smaller deletion covering only hs3b and hs4 (4.6-kb deletion) impedes CSR to all IgH classes except ␥1 (2). Remarkably, the dramatic effects of this relatively small deletion can be reversed by leaving intact either the 2000 bp surrounding hs3b or the 1400 bp surround- FIGURE 5. Long range interactions between the 3RR, IgH transcription unit, and C H region promoters. A, maps of BAC transgenes before (⌬3alx3b) and after (⌬3a⌬3b) hs3b deletion. 5Ј primers for E, the I␥3 promoter, I␥1 promoter, and 3ЈRR element hs1.2 are indicated as arrows, as is the transgenespecific 3Ј primer near hs4 (anchor). VDJ, inserted V H region. 3C, chromatin conformation capture. B, PCR products generated by chromatin conformation capture (see "Experimental Procedures") with the indicated primer pairs (left of gel). Interactions between fragments within the CalR gene served as a positive control (top panel). ϩ ϭ positive control template (mixture of CalR fragments and the IgH⌬3alx3b BAC); Ϫ ϭ negative control template (mixture of CalR fragments and a BAC with sequences equivalent to the wild-type, endogenous Igh loci). Cell source of template is shown above the lanes (⌬3alx3b and ⌬3a⌬3b B cells). Lanes 1 and 4 are from B cells at time 0 (resting B cells); lanes 2 and 5 are from B cells 48 h after LPS stimulation, and lanes 3 and 6 are from B cells 48 h after LPSϩIL-4 stimulation. C, summary of cross-linking efficiency data. Cross-linking was measured between the hs4 region of the BAC transgene (hs4(Tg) and the indicated regions upstream (see "Experimental Procedures" for details). Two independent CSR experiments were performed using two different paired mouse lines, and PCRs were performed at least twice with each template. The first three bars in each set of six (for each primer pair) are the results with IgH⌬3alx3b B cells; the second three bars are with IgH⌬3a⌬3b B cells (see legend). * ϭ p Ͻ 0.05; ** ϭ p Ͻ 0.01. ing hs4 (20,21). The restored sequences (hs3b and hs4) do not share obvious sequence homology, and in fact, follow distinct time lines for chromatin modification during development (11), yet each can work in synergy with the remaining elements of the 3ЈRR to effect normal CSR activity.
These previous studies showed that more than one combination of three enhancer-like elements from the 3ЈRR could support normal CSR (hs3a, hs1.2, and hs4 versus hs3a, hs1.2, and hs3b). Notably, each set of three included one or more of the hs3 elements, suggesting a possible requirement for one or the other. The current study disproves that hypothesis, demonstrating that the hs3a and hs3b elements can be removed simultaneously without significant effect on CSR to the various heavy chain classes. More generally, this finding reveals that functional redundancy in this system is not based on sequence homology. These findings prompt an adjustment in approach to further analyses of this important regulatory region. It is now clear that there is more than one way to "build" a CSR regulatory region. The present study identifies the smallest number of building blocks to date (hs1.2 and hs4).
Given the multiple combinations capable of activity, however, it is clear that discerning the nature of the mechanistic cooperation among these elements is critical. Perhaps there is an overall structure required that can be achieved by more than one combination of elements (e.g. providing a required density of transcription factor-binding sites and associated activities; a tether to a nuclear subregion; the scaffold for interaction with the IgH transcription unit and the C H region promoters, etc.). Alternatively, it is possible that each of these elements preferentially supports a subset of CSR events, with each of these subsets partially overlapping the others. These overlapping preferences would be masked by the individual element deletion approach employed to date. To answer these questions of mechanism, future studies will need to aim toward identification of the CSR-related activities supported by each individual element of the 3ЈRR (knowing that none is absolutely essential for this process). Comparing the activity of individual elements with that displayed by distinct combinations of elements can reveal the combinatorial logic that apparently provides many mechanistic routes to complete CSR activity.