The HoxC4 Homeodomain Protein Mediates Activation of the Immunoglobulin Heavy Chain 3 (cid:1) hs1,2 Enhancer in Human B Cells TO

The immunoglobulin heavy chain (IgH) 3 (cid:1) regulatory region modulates IgH locus transcription, upon induction by specific trans -acting factors, and plays a significant role in class switch DNA recombination (CSR) and, perhaps, somatic hypermutation (SHM). CSR and SHM are central to the maturation of the antibody response. In contrast to the single 5 (cid:1) -hs3a-hs1,2-hs3b-hs4-3 (cid:1) mouse IgH 3 (cid:1) regulatory region, the human IgH 3 (cid:1) regulatory region exists as a 5 (cid:1) -hs3-hs1,2-hs4-3 (cid:1) cluster duplicated 3 (cid:1) of C (cid:1) 1 and C (cid:1) 2. We show here that the human hs1,2 element is the strongest enhancer of transcription, as directed by a V H 1 or the ECS-I (cid:2) 3 promoter, thereby sug- gesting a dominant role for hs1,2 over hs3 and hs4 in the overall activity of the 3 (cid:1) regulatory region. Within hs1,2, we identified three regions (1, 2, and 3) that are all necessary, but individually not sufficient, for enhancement of transcription. In region 2, a HoxC4 site and a HoxC4/embedded

The immunoglobulin heavy chain (IgH) 3 regulatory region modulates IgH locus transcription, upon induction by specific trans-acting factors, and plays a significant role in class switch DNA recombination (CSR) and, perhaps, somatic hypermutation (SHM). CSR and SHM are central to the maturation of the antibody response. In contrast to the single 5-hs3a-hs1,2-hs3b-hs4-3 mouse IgH 3 regulatory region, the human IgH 3 regulatory region exists as a 5-hs3-hs1,2-hs4-3 cluster duplicated 3 of C␣1 and C␣2. We show here that the human hs1,2 element is the strongest enhancer of transcription, as directed by a V H 1 or the ECS-I␥3 promoter, thereby suggesting a dominant role for hs1,2 over hs3 and hs4 in the overall activity of the 3 regulatory region. Within hs1,2, we identified three regions (1, 2, and 3) that are all necessary, but individually not sufficient, for enhancement of transcription. In region 2, a HoxC4 site and a HoxC4/embedded octamer (HoxC4/Oct) site are conserved across human, mouse, rat, and rabbit. These two sites recruit HoxC4 and Oct-1/Oct-2, which act synergistically with the Oca-B coactivator to effect the full hs1,2enhancing activity. HoxC4, Oct-1/Oct-2, and Oca-B recruitment is negligible in pro-B cells, moderate in pre-B cells, and maximal in germinal center B cells and plasma cells, where HoxC4, Oct-2, and Oca-B expression correlates with hs1,2 activation and ongoing CSR. The hs1,2mediated enhancement of V H and C H promoter-driven transcription as induced by HoxC4 and Oct-1/Oct-2 suggests an important role of these homeodomain proteins in the overall regulation of the IgH locus expression.
Gene transcription of the Ig heavy (H) and light (L) chain locus proceeds in a lymphoid-restricted and developmental stage-specific fashion, leading to Ig V(D)J recombination and the emergence of mature B cells expressing unique receptors for antigen. Upon encounter with antigen, B cells undergo somatic hypermutation (SHM) 1 and class switch DNA recom-bination (CSR) to produce affinity-mature, isotype-switched antibodies. Like V(D)J recombination, SHM and CSR are critically dependent on transcription, as driven by three main cis-regulatory regions: the promoter upstream of each V gene (V), the IgH evolutionarily conserved sequence (ECS)-intervening (I) region promoter, which is located upstream of each constant region (C H ) exon cluster, and the IgH intronic enhancer (iE) (1)(2)(3)(4). The identification of an additional cisregulatory region was suspected after mouse cell lines lacking the iE enhancer were found to still effectively transcribe IgH genes, and a mouse cell line containing an intact iE enhancer but showing a large deletion of sequence downstream of the C␣ gene showed decreased IgH transcription (5,6).
Indeed, a second B cell-specific regulatory region was identified ϳ25 kb downstream of the rat C␣ gene and 16 kb downstream of the mouse C␣ gene, with 82% sequence identity (7)(8)(9). The murine 3Ј regulatory region consists of five B cellspecific DNase I hypersensitivity sites, each characterized as an IgH 3Ј enhancer (3ЈE H ): hs3a, hs1,2, hs3b, and hs4, with hs1,2 lying at the center of a region of symmetry flanked by inverted repeat sequences (6). hs1-4 collectively function as a locus control region (LCR) (5,6), as suggested by the positionindependent and copy number-dependent deregulation of c-MYC expression in plasmacytoma cells transfected with a hs1,2-hs3b-hs4-linked c-MYC construct (10). Additional sequences may be required to allow the murine hs1-4 enhancers to act as a classical LCR, as in transgenic mice harboring a V H promoter-␤-globin reporter gene linked to the Ig 3ЈE H regulatory region, transgene expression was strictly confined to B cells, and reporter gene activity was integration-independent but not copy number-dependent (11).
A role for the mouse hs1,2 and hs3a enhancers in CSR to IgG 2b and IgE has been suggested by Cre/loxP gene targeting experiments (12). These extended previous findings obtained by replacement of DNA encompassing the mouse hs1,2 element with the neomycin (neo) gene (13). Further experiments involving hs3b and hs4 and Cre/loxP knockouts showed severe impairment of germ-line I H -C H transcription and CSR to IgG 2a , IgG 2b , IgG 3 , IgE, and IgA, indicating that the distal portion of the regulatory region is required for CSR to most isotypes (14).
The role of the 3Ј regulatory region in SHM awaits better definition. The use of a transgenic construct containing murine hs1,2 suggested that this element does not play a role in SHM, even when coupled with the iE enhancer (15). In contrast, the use of a transgene construct containing the distal hs3b and hs4 elements, but not hs1,2, resulted in an increase in the SHM level of transgenes driven by a V H promoter, pointing to a role for hs3b and hs4 in SHM (15,16). However, more recent experiments have indicated that hs3b and hs4 are dispensable for SHM and V H DJ H gene assembly (17).
The human IgH 3Ј regulatory region comprises the B cellspecific DNase I hypersensitivity sites hs1,2, hs3, and hs4, which are arranged in the 5Ј hs3-hs1,2-hs4 3Ј sequence and duplicated as discrete enhancer clusters 3Ј of C␣1 and 3Ј of C␣2 (supplementary Fig. S1), with the C␣1 and C␣2 hs1,2 sequences inverted with respect to each other (18,19). Like their murine counterparts, the human 3ЈE H hs1,2 and hs4 elements are induced by the Oct-2 transcription factor (trans-factor) and its interacting coactivator Oca-B (6). In contrast to the mouse, negative regulatory mechanisms would not be conserved, as binding sites for B cell-specific activator protein (BSAP), a repressor of the mouse 3ЈE H hs1,2, are lacking in human hs1,2 (6). Human 3ЈE H elements have been shown to enhance transcription as driven by a mouse V or V promoter or a human ECS-I␣ or ECS-I␥3 promoter (18 -21). Like its murine counterpart, the human 3ЈE H regulatory region likely functions as an LCR and may play a role in V H gene rearrangement, CSR, and SHM (2,6).
To identify the elements involved in the induction of the human IgH 3Ј regulatory region, we created hs1,2, hs3, and hs4 enhancer-DNA constructs and 5Ј, 3Ј, and internal deletion hs1,2 mutants, and inserted them into luciferase (luc) reporter gene vectors driven by a human V H or ECS-I H promoter. By bearing the luc gene downstream of the promoter and upstream of hs1,2, hs3, or hs4, these vectors mimicked the physiological promoter-gene-enhancer structure found in the IgH locus. In addition, we generated hs1,2 enhancer constructs containing mutations of selected cis-elements and enforced expression of respective trans-factors. Finally, to determine the stages of B cell ontogeny at which the human hs1,2 is activated, we transfected pro-B cells, pre-B cells, early and late germinal center B cells, and plasma cells with a reporter gene vector containing a fully inducible hs1,2 enhancer construct, as driven by the human ␤-globin promoter. We then measured the enhancement of such a promoter activity and analyzed it with the levels of endogenous germ-line I H -C H and mature V H DJ H -C H transcripts, the expression of trans-factors targeting hs1,2, and the formation of specific nucleoprotein complexes.

EXPERIMENTAL PROCEDURES
Human IgH 3Ј Regulatory Region-The human hs1,2, hs3, and hs4 DNAs were from the 3ЈE H cluster lying 3Ј of the C␣2 exon (19). The 1079-bp hs1,2 DNA spans residues 322-1400 according to the numeration of this genomic clone (GenBank TM accession nos. AF013724 and U84574); the 695-bp hs3 DNA spans residues 526 -1221 (GenBank TM accession nos. AF013719 and Y14406); and the 426-bp hs4 DNA spans residues 1-426 (GenBank TM accession no. AF013726). The hs1,2 region 1 G-rich repeats were identified by Pustell DNA matrix analysis (22), enabling the search for regions of high similarity between two nucleic acid sequences using a dot matrix plot. The human, mouse, rat, and rabbit hs1,2 region 2 sequences were compared using the ClustalW algorithm (23) allowing for multiple alignments of nucleotide sequences. All sequence comparisons were implemented by MacVector, version 6.5.3 (Accelrys Inc., San Diego, CA). The identification of putative cis-regulatory binding sites was performed using Matlnspector (www.genomatix.de/cgibin/matinspector/matinspector.pl), which utilizes the TRANSFAC library of matrices (www.gene-regulation.de/) to locate consensus matches in nucleotide sequences (24).
Vectors-The pGL3-Basic luc-reporter gene vector (Promega Corp., Madison, WI) was modified by inserting three different promoter se-quences between the SacI and BglII restriction sites: Ϫ449/ϩ265 ECS-I␥3 promoter DNA (GenBank TM accession no. S79588) (25), V H 1 promoter DNA isolated from the IgG 1 mAb57-producing cell line (26,27), or the human ␤-globin promoter DNA (GenBank TM accession no. U01317). The pGL3 vectors containing the V H 1, ECS-I␥3 or ␤-globin promoter were further modified by inserting between the BamHI/SalI restriction sites the hs3, hs1,2, and hs4 elements, resulting in the positioning of each 3ЈE H ϳ2 kb downstream of the respective promoter and immediately flanking the firefly luc gene. To generate sequential 5Ј-and 3Ј-end truncation mutants, hs1,2 was PCR-amplified using appropriate primers with BamHI/SalI overhangs. Internal hs1,2 deletion mutants were generated by first amplifying 5Ј-and 3Ј-halves, minus the targeted sequence, ligating the two fragments, and re-amplifying the "complete" DNA. Ligation of the 5Ј and 3Ј PCR fragments flanking the targeted motifs generated site-targeted mutations of the 5Ј HoxC4 and 3Ј HoxC4/ Oct-binding sites. The 5Ј HoxC4 site (ATTT, residues 715-718) was replaced by cggg. The HoxC4/Oct site (ATTTGCAT, residues 773-780) was replaced with a KpnI restriction site. The double mutant hs1,2 DNA was generated using the single 5Ј HoxC4 and 3Ј HoxC4/Oct mutants as templates. All digested PCR products were gel-purified and subcloned into the pGL3 vectors driven by the V H 1 or ECS-I␥3 promoter.
For enforced expression studies, cDNA encoding human HoxC4, Oct-1, Oct-2, or Oca-B was subcloned into pcDNA3.1 vectors using the pcDNA3.1/V5-His TOPO TA expression kit (Invitrogen, Carlsbad, CA). The pcDNA3.1 expression vector contains a cytomegalovirus promoter for high level expression and a T7 promoter for in vitro translation using the TNT Quick Coupled Transcription/Translation System (Promega Corp.). The expression vector encoding the dominant negative HoxC4 lacking the homeodomain was described (28). The glutathione S-transferase (GST) fusion proteins were generated by subcloning human HoxC4, Oct-1, Oct-2, and Oca-B cDNAs into the pGEX-6P-1 vector as described previously (28). GST fusion proteins were expressed in BL21 bacteria, purified using GSH-agarose beads according to the manufacturer's protocol (Sigma-Aldrich) and analyzed for homogeneity by SDS-PAGE and silver staining. Proteins were eluted in 15 mM reduced glutathione in 50 mM Tris buffer, pH 8.0. Restriction enzyme mapping and DNA sequencing were used to verify all of the correct plasmids.

RESULTS
hs1,2 Is the Strongest Enhancer in the Human C␣2 3ЈRegulatory Region-To address the function of the human IgH 3Ј regulatory region, we analyzed the ability of the C␣2 hs1,2, hs3 and hs4 elements to enhance transcription of a luciferase (luc)reporter gene vector, as driven by a human V H 1 promoter, the human ECS-I␥3 promoter, or the Ig-irrelevant human ␤-globin promoter, upon transfection of human 4B6 B cells. These cells are IgM ϩ IgD ϩ and spontaneously switch to IgG, IgA, and IgE. They are the neoplastic equivalent of centroblasts/centrocytes (28), a B cell differentiation stage at which the 3Ј regulatory region has been suggested to be active (6). In 4B6 cells, the hs1,2 element was the strongest enhancer of transcription, yielding a 172-fold (V H 1 promoter), 100-fold (ECS-I␥3 promoter), and 77-fold (␤-globin promoter) increase in Luc activity; hs4 was second, yielding a 46-fold (V H 1 promoter), 10-fold (ECS-I␥3 promoter), and 15-fold (␤-globin promoter) increase in Luc activity, whereas hs3 was the weakest of the three elements, yielding 6-fold (V H 1 promoter), 2-fold (ECS-I␥3 promoter), and 2-fold (␤-globin promoter) increase in Luc activity (Fig. 1).
hs1,2 Contains Three Regions with Multiple cis-Elements-Because of its dominant activity, we analyzed hs1,2 for putative trans-factor-binding motifs and segregated it into three regions (1, 2, and 3) based on the identified cis-elements (Fig.  2). Region 1 encompasses residues 452-669 and comprises four highly G-rich ϳ53-bp tandem arrayed segments (G-rich repeats 1-4 (GRR)), which have been suggested to display some polymorphism (19,37). In each GRR, we identified a putative B (GGGGYNNCCY consensus) site. Region 2, originally classified as an ECS core (19), encompasses residues 684 -818. In this region, we identified two Hox-binding sites (ATTT, 5Ј at residues 715-718 and 3Ј at residues 773-780), in addition to the previously identified E-Box (characterized as a E5 site, CANNTG, residues 687-698)-, Oct (ATTTGCAT, residues 773-780)-, and AP-1 (TGAGTCA, residues 809 -815)-binding sites (19). We found that these ATTT sites recruit HoxC4 (not shown). We determined that the 3Ј HoxC4-binding site is embedded within the 5Ј portion of the Oct site and therefore designated it as HoxC4/Oct site. Region 3 encompasses residues 819 -1133 and contains eight putative cis-elements: a GATA (WGATAR consensus) site, an E-Box site, two Myb (YAACKG consensus) sites, two Ikaros (GGGAA consensus), Note the identity of the ATTT and ATTTGTAC sites across human, mouse, rat, and rabbit. Also, note the lack of BSAP sites in the human. and two AP sites (an AP-4 (CAGCTG consensus) and an AP-2 (GSSWGSCC consensus)). Regions 1 and 3 are only 47 and 39% identical, respectively, in human, mouse, rat, and rabbit (Fig.  3). In contrast, region 2 is 87% identical overall and 100% identical in the identified HoxC4 and HoxC4/Oct elements across human, mouse, rat, and rabbit.
Regions 1, 2, and 3 Are All Necessary for C␣2 hs1,2 Activation-To determine the minimal sequence responsible for mediating hs1,2 activation, we generated 5Ј and 3Ј truncation and internal deletion mutants and inserted them in luc-reporter gene vectors, which were then used to transfect 4B6 cells and to measure the enhancement of transcription, as driven by a V H 1 or the ECS-I␥3 promoter. The transcription-enhancing activity mediated by the full-length C␣2 hs1,2 construct (322-1400) was referred to as 100% and used as the term of comparison for all other constructs (Fig. 4). Analysis of the 5Ј and 3Ј truncation mutants showed that the sequences 322-451 (compare C6 with C2) and 1133-1400 (compare C13 with C2) do not make a significant contribution to the overall enhancement of transcription, as mediated by V H 1 or ECS-I␥3 promoter. This was confirmed by the analysis of the 5Ј and 3Ј truncated construct C17, which displayed 97% of the activity of C2. Analysis of sequential 5Ј deletion mutants showed that each of the four GRRs critically contributes to the overall enhancing activity of region 1 (C6 -C11), whereas 3Ј deletion mutants suggested a significant role for GATA, E-Box, Ikaros, AP-2, AP-4, and/or Myb in the overall enhancing activity of region 3 (C14 and C15). Furthermore, deletion of region 1 (⌬452-669, C3), 2 (⌬684 -818, C4), or 3 (⌬819 -1133, C5) DNA resulted in 84, 96, and 93% loss, respectively, of such an activity, regardless of the promoter used. Conversely, when alone, region 1, 2, or 3 failed to significantly enhance transcription, as driven by the V H 1 or ECS-I␥3 promoter (C12, C16, C20, and C21). Regions 1 and 2 together displayed only partial enhancing activity when disjoined from region 3 (C18), whereas regions 2 and 3 together displayed only partial activity when disjoined from region 1 (C19). Thus, regions 1, 2, and 3 are all necessary, but none is sufficient per se to effect full C␣2 hs1,2 activation.
The HoxC4 and Oct-1/Oct-2 Homeodomain Proteins Are Specifically Recruited to C␣2 hs1,2-To identify the trans-factors that are recruited to the HoxC4 and HoxC4/Oct sites of hs1,2 region 2, we used wild-type (wt) and mutated (mt) oligonucleotide probes encompassing residues 704 -729 (HoxC4 site) and 764 -789 (HoxC4/Oct site) (Fig. 6) in EMSAs involving 4B6 nuclear cell extracts. These gave rise to a distinct nucleoprotein complex when incubated with the HoxC4 probe (Fig. 7A, top  panel). Such a complex was specific, as its formation was in- hibited by cold wild-type but not mutant hs1,2 HoxC4 (HoxC4 mt4) oligonucleotides or oligonucleotides containing an Ig NF-B or BSAP site. The involvement of HoxC4 in the nucleoprotein complex was demonstrated by the inhibition of complex formation by a mouse mAb to HoxC4 and the direct in vitro interaction of a GST-HoxC4 fusion protein with the radiolabeled DNA containing the wild-type but not the mutant hs1,2 HoxC4-binding site (Fig. 7A, bottom panel).
Incubation of nuclear extracts from 4B6 cells with the HoxC4/Oct probe gave rise to the two major nucleoprotein complexes, A and B (Fig. 7B). These were specific, as shown by the failure of mutated HoxC4/Oct oligonucleotides (HoxC4/Oct mt1-4) or Ig NF-B and BSAP oligonucleotides to inhibit the formation of both complexes. They included the HoxC4 and Oct-1 or Oct-2 homeodomain proteins as well as the Oca-B coactivator, as inferred from their supershifting or inhibition by specific antibodies to HoxC4, Oct-1, Oct-2, and Oca-B (Fig.  7C) and the binding of GST-HoxC4, GST-Oct-1, and/or GST-Oct-2 to the HoxC4/Oct oligonucleotide probe (Fig. 7D, lanes  1-6). Such a binding was specific, as mutation of the whole HoxC4/Oct-binding site (mt5) or its 5Ј-end containing the HoxC4-binding motif (mt6 and mt7) abolished binding of HoxC4, Oct-1, and Oct-2 (Fig. 7D, lanes 7-24), whereas mutation of the HoxC4/Oct 3Ј portion (mt8) abrogated the binding of Oct-1 and Oct-2 but not HoxC4 (lanes 25-30). Finally, the Hox/Oct nucleoprotein complexes did not include homeodomain-interacting Ku70/Ku86 proteins, as Ku-specific antibodies did not supershift either complex A or B (Fig. 7C, lanes  6 -8).
That HoxC4, Oct-1, Oct-2, and Oca-B are bound in vivo to hs1,2 region 2 was demonstrated by ChIP experiments in which the hs1,2 HoxC4-HoxC4/Oct sequence was specified in DNA precipitated from HS Sultan B cells using antibodies to HoxC4, Oct-1, Oct-2, or Oca-B but not Ku70⅐Ku86 or Stat-1 (Fig. 7E). In addition, the binding of in vitro translated 35 S-Oct-1, 35 S-Oct-2, and 35 Oca-B to immobilized GST-HoxC4 indicated that Oct-1, Oct-2, and Oca-B can be recruited to hs1,2 by DNA-bound HoxC4 through direct protein-protein interaction (Fig. 7F). Thus, the hs1,2 5Ј HoxC4 and 3Ј HoxC4/Oct ciselements effectively recruit HoxC4, Oct-1/Oct-2 and Oca-B as  6. hs1,2 HoxC4 and HoxC4/Oct oligonucleotides. HoxC4-and HoxC4/Oct-binding site oligonucleotides used as radiolabeled probes in EMSA. For the HoxC4 site, the ATTT motif was mutated to cggg (HoxC4 mt4). For the HoxC4/Oct site, the ATTTGCAT motif was mutated as follows: HoxC4/Oct mt5, complete element replaced with a KpnI sequence; HoxC4/Oct mt6, 5Ј HoxC4 portion (ATTT) mutated to cggg; HoxC4/Oct mt7, AT of 5Ј HoxC4 portion mutated to gg; and HoxC4/Oct mt8, AT of 3Ј Oct portion mutated to cg. FIG. 7. Recruitment of HoxC4, Oct-1/Oct-2, and Oca-B to the 3E H hs1,2. A, nuclear proteins from spontaneously switching 4B6 cells specifically bind an oligonucleotide probe containing the HoxC4-binding site of hs1,2 region 2 (top panel). Efficient competition was achieved by 50-fold molar excess of wild-type (HoxC4 wt) but not mutant (HoxC4 mt4), Ig B, or BSAP cold oligonucleotides. The formation of the DNA-binding complex was inhibited by a specific mAb to HoxC4. Mouse IgG served as a negative control. The specificity of the HoxC4 oligonucleotide probe was further verified by binding of recombinant GST-HoxC4 protein (250 ng) to wild-type HoxC4 but not to mutated HoxC4 (mt4) oligonucleotide probe (bottom panel). NE, nuclear extract. B, nuclear proteins from 4B6 cells bind specifically to radiolabeled hs1,2 HoxC4/Oct wt probe. Efficient competition was achieved by 50-fold molar excess of wt (HoxC4/Oct wt) but not mt (HoxC4/Oct mt5-8) or nonspecific (Ig B and BSAP) cold oligonucleotides. C, identity of the nuclear protein complexes as assessed by supershift EMSA. 4B6 nuclear extracts were preincubated with the indicated antibodies prior to the addition of the HoxC4/Oct oligonucleotide probe and EMSA. Arrowheads indicate complexes A and B as well as the supershifted complex AЈ. D, direct binding of HoxC4, Oct-1, and Oct-2 to hs1,2 DNA. Recombinant GST fusion HoxC4, Oct-1, Oct-2, HoxC4, and Oct-1 or HoxC4 and Oct-2 protein (250 ng) were analyzed for direct binding to HoxC4/Oct probes with GST serving as a negative control. Sequences of HoxC4/Oct mt5, mt6, mt7, and mt8 were as described in the legend to Fig. 6. Note that mutation of the 3Ј-end of HoxC4/Oct motif (mt8) allowed for binding of HoxC4 only, as predicted. E, in vivo binding of HoxC4, Oct-1, and Oct-2 proteins to hs1,2 region 2 as determined by ChIP. Cross-linked chromatin from HS Sultan B cells was precipitated by mAb to human HoxC4 or antibodies specific for human Oct-1, Oct-2, or Oca-B. The precipitated DNA was specified by PCR using primers (black arrows) and detected by Southern blotting using a specific probe (gray bar) as listed under "Experimental Procedures." mAbs that detect the interface between Ku70 and Ku86 and a rabbit antibody to Stat-1 were used as controls, along with mouse and rabbit IgG. F, GST fusion protein pull-down assays. In vitro 35 S-labeled translated HoxC4, Oct-1, Oct-2, or Oca-B proteins were mixed with GST or GST-HoxC4 immobilized on glutathione-agarose resin, subjected to 12% SDS-PAGE, and exposed for autoradiography. discrete components of a nucleoprotein complex that can be detected in vivo in germinal center B cells.
Activation of C␣2 hs1,2 Is Maximal in Germinal Center B Cells and Plasma Cells-To confirm the relevance of our findings to B cell ontogeny, we transfected human cell lines corresponding to sequential stages of B cell differentiation with the C␣2 hs1,2 (452-1133) enhancer-luc reporter gene vector, as driven by the ␤-globin promoter. We then measured Luc activity, the levels of endogenous germ-line I H -C H and mature V H -DJ H -C H transcripts, and the expression of HoxC4, Oct-1/Oct-2, and Oca-B, and we monitored the formation of HoxC4⅐Oct-1⅐Oct-2⅐ Oca-B nucleoprotein complexes. C␣2 hs1,2 was not activated in pro-B cells (RS4;11), was moderately activated (4-fold) in pre-B cells (Nalm-6), and was significantly activated in early (4B6) and late (HS Sultan) germinal center B cells (12-50-fold) and plasma cells (U266) (10-fold) (Fig. 9A). Increased hs1,2-enhancing activity was associated with the appearance of germ-line I H -C H and/or mature V(D)J-C H transcripts, expression of HoxC4, Oct-1/Oct-2, and Oca-B transcripts and proteins, and the formation of related nucleoprotein complexes involving the ATTT and ATTTGCAT DNA motifs (Fig. 9, B-D). Thus, hs1,2 activation and consequent enhancement of IgH locus transcription is B cell stage-specific and occurs concomitantly with the formation of HoxC4, HoxC4/Oct-1, and HoxC4/Oct-2 nucleoprotein complexes. DISCUSSION Together with our previous studies on the human I␥ and I⑀ promoters (25,28), these findings point to HoxC4 as an important regulator of transcription in the human IgH locus and outline a critical role for this homeodomain protein in B cell differentiation. We show here that HoxC4 mediates activation of the 3ЈE H hs1,2 enhancer in human B cells by binding to newly identified conserved ATTT and ATTTGCAT motifs and through synergy with two other homeodomain proteins, Oct-1/ Oct-2, and the Oca-B coactivator. By showing that hs1,2 is dominant over hs4 and hs3 in enhancing transcription, as driven by a human V H , ECS-I H or the ␤-globin promoter, our experiments extend previous findings (18 -21). Further, they show that hs1,2 can be segregated into three regions (1, 2, and 3), which are all necessary to enhance germ-line I H -C H and mature V H DJ H -C H transcription. Finally, they define the minimal requirements for human hs1,2-mediated transcriptional enhancement and determine a B cell stage specificity in HoxC4dependent activation of hs1,2 (25,28).
The enhancement of V H 1 or ECS-I␥3 promoter-driven transcription suggests a primary role for hs1,2 in the overall function of the IgH 3Ј regulatory region, as V H 1 and ECS-I␥3 are IgH locus promoters. The V H promoter is required for steadystate V H gene transcription, V H DJ H gene rearrangement, and IgH SHM; the ECS-I H promoter is required for germ-line I H -C H transcription and CSR. The enhancing activity displayed by hs1,2 in conjunction with the Ig-irrelevant ␤-globin promoter emphasizes the strength of hs1,2 as a bona fide enhancer (1,5,6,38,39). This dominant transcription-enhancing activity of human hs1,2 would be reflected in the SHM-enhancing activity displayed by this element, but not hs3 or hs4, when inserted downstream of a SHM "inducible" human DNA V H DJ H -iE-S-S␥-C␥1 construct 2 (40).
Our analysis of the human C␣2 hs1,2 identified two HoxC4binding sites, which are both critical for full hs1,2 activation. The 3Ј cis-element is a HoxC4/Oct-binding site and was previously recognized as a mere Oct-binding site in the reports originally detailing the structure of the human IgH 3Ј regulatory region (18 -21). That Oct site was included together with Ets/AP-1 and E-Box sites in the ϳ0.3 kb PstI/PstI fragment P300. P300 comprised region 2, perhaps together with region 1 or 3, and was found to yield about 55% of the overall hs1,2enhancing activity, to which neither region 1 nor region 3 was identified as a further contributor (19). In another study (18), a sequence corresponding to our region 2, but also containing one copy of a ϳ53-bp motif (equal to region 1 GRR4), was identified as an ECS lacking CTGCAGCTGCAGGT, which includes the E-Box site, and resulting in an overall structure/function of hs1,2 significantly different from that reported here. In that study, the cis-binding sites downstream of the ECS were not identified, nor was a "region 3" defined and shown to be required for full hs1,2 activation. An analysis of the rat hs1,2 element resulted in the dissection of the enhancer into three domains, designated as A, B, and C, which were shown to effect transcriptional enhancer activity (41). As in the human region 2, such enhancer activity was contributed mainly by the second hs1,2 DNA region, domain B. But unlike our human region 2, the main rat hs1,2 domain B cis-elements are a group of Etslike binding sites. The maximal activation of hs1,2 in human germinal center B cells originally went unrecognized, as IgG ϩ HS Sultan B cells were classified as a myeloma rather than a germinal center B cell line (18,19,34). It has been further confirmed here by the demonstration that hs1,2 is highly activated in 4B6 cells, which effectively express I H -C H transcripts and undergo CSR to IgG, IgA and IgE (28). The maximal activation of hs4 in Ramos B cells (42) further indicates that the 3Ј regulatory region plays an important role in the IgH locus transcriptional regulation in germinal center B cells. However, significant hs1,2 activation was seen in murine Ig-secreting plasma cells but not in germinal center B cells (8 -10, 43, 44), perhaps reflecting the presence of BSAP-binding motifs in this murine IgH enhancer element (5,6,39). By binding to hs1,2 and hs4, BSAP represses the 3ЈE H activity in murine germinal center B cells, and BSAP down-regulation is likely central to the full activation of the IgH 3Ј regulatory region observed in plasma cells (45,46). Because of the putative lack of BSAP-binding sites in the hs1,2 and hs4 sequences (2,6,19), human germinal center B cells would likely evade BSAP-mediated repression. This and the demonstration that, like hs3a, the murine hs1,2 is dispensable for germ-line I H -C H transcription and CSR (14) underscore significant differences between the mouse and human IgH 3Ј regulatory regions.
Hox proteins are phylogenetically conserved helix-loop-helix homeodomain proteins that recognize the ATTT/A consensus (47,48). They regulate embryonic pattern formation, axis specification and organogenesis, selective hematopoietic differentiation, and stem cell renewal (49). Genes belonging to the C cluster are preferentially expressed in developing and differentiated lymphoid lineages. HoxC4 is expressed in activated and/or proliferating T, B, and NK cells. No data, even from targeted deletions in the mouse (50,51), have been available on HoxC4 function in the lymphoid system. Its early expression and nuclear localization suggest an involvement of HoxC4 in the regulation of genes controlling lymphocyte activation and/or proliferation (52). By defining critical new roles in the regulation of the IgH locus expression, our present and previous (28) findings point to HoxC4 as an important element in human lymphocyte differentiation.
HoxC4 synergizes with Oct-1/Oct-2 and the Oca-B coactivator, which it recruits to induce the human hs1,2 enhancer. Indeed, Oct-1/Oct-2 and Oca-B are components of the newly identified nucleoprotein complexes A and B, which assemble on the hs1,2 ATTTGCAT cis-element through recruitment of HoxC4 to the 5Ј-ATTT end (Fig. 7C). Although HoxC4/Oct-1 heterodimers can form and bind hs1,2 we propose that it is HoxC4/Oct-2 that is recruited preferentially, as suggested by the ChIP assays and the specific DNA binding by increased HoxC4 and Oct-2 proteins in germinal B cells (Figs. 7E and 9, C and D). This heterodimer assembly and recruitment further involves Oca-B. The ATTT cis-element would not recruit Oct-1/Oct-2 and Oca-B directly but through DNA-bound HoxC4, as suggested by protein-protein interaction experiments (Fig. 7F). The trans-factors recruited at the HoxC4 site would synergize with the HoxC4⅐Oct-1/Oct-2/Oca-B complex recruited at the HoxC4⅐Oct site to potentiate hs1,2 activation. This paradigm of transcriptional regulation through protein-protein cooperation has been effectively shown for other Hox homeodomain proteins in which the transcriptional activation function appears to be dependent on the nature of the target DNA sequence, implicating the importance of partner(s) or cofactor(s) and the relative properties of this interaction in mediating specific transcriptional regulation.
The HoxC4 homeodomain is important for hs1,2 activation, as expression of a mutant HoxC4 lacking this domain abolished hs1,2 activity. Overexpression of Oct-1, Oct-2, or Oca-B could not overcome this inactivation, presumably because the HoxC4 homeodomain mutant behaved as a potent "dominant negative" regulator of Oct-mediated hs1,2 activation. Oct-1 and Oct-2 are members of the POU family, a group of homeodomain-containing trans-factors that contain the DNA-binding POU domain. This comprises the "POU-homeodomain" and "POU-specific" subdomains (53). Oct-1 and Oct-2 regulate both general and cell type-specific genes (54), including V H and C H (53). Although Oct-1 is ubiquitous, Oct-2 is preferentially expressed in B cells. In the human IgH 3Ј regulatory region, Oct-2 is required for not only hs1,2 (our data), but also hs4 activation (42). Accordingly, in Oct-2-deficient mice, B cell development to surface IgM expression is normal, but germinal center formation is impaired, and IgG 1 and IgG 3 levels are severely decreased (55,56), indicating that Oct-2 is required for germinal center formation, CSR to secondary isotypes, and a high level of Ig transcription.
Coexpression of Oca-B and HoxC4 yielded the highest level of hs1,2 activation, presumably through interactions with the endogenous pool of Oct-1/Oct-2 proteins (Fig. 8). Oca-B (Oct coactivator from B cells, or Oct-binding factor-1 (OBF-1)) functions as an important transcriptional coactivator in B cells. It increases the binding affinity of Oct-1 and Oct-2 for DNA by clamping the POU H and POU S subdomains, which can further stimulate Oct-dependent gene transcription (57). Our demonstration that Oca-B plays an important role in HoxC4/Oct-2mediated activation of human hs1,2 further emphasizes the critical role of this coactivator in B cell differentiation. Accordingly, Oca-B interacts with Oct-2 in modulating the activity of the 3ЈE H and IgH transcription in murine B cells (58,59), and Oca-B Ϫ/Ϫ B cells stimulated with anti-CD40 and interleukin-4 fail to activate a luc-reporter gene construct bearing the regulatory hs3a-hs1,2-hs3b-hs4 cluster (44). Further, mice lacking Oca-B are viable and have normal serum IgM levels but lack GCs and show a significant impairment in CSR to IgG and serum IgG levels (60 -62). Finally, Oct-2/Oca-B double deficient mice display a similar but more pronounced phenotype with impairment of T cell-dependent antibody responses (59).
The transcription-enhancing activity induced by the binding of HoxC4 to the human hs1,2 ATTT and ATTTGCAT motifs contrasts with the repression of germ-line I␥-C␥ and I⑀-C⑀ transcription that, as we showed (28), is dependent on recruitment of HoxC4 to ATTT sites embedded in the human ECS-I␥ and ECS-I⑀ promoters. ATTT motifs exist as multiple copies in the human ECS-I␥ and ECS-I⑀ promoters (28). Because of the lack of ATTT sites in the ECS-I␣1/I␣2 promoters, the C␣1/C␣2 loci can undergo CSR to IgA even when HoxC4 expression is up-regulated (28). The repression exerted by HoxC4 on the ECS-I␥ and ECS-I⑀ promoters, germ-line I␥-C␥ and I⑀-C⑀ transcription, and CSR to IgG and IgE is dependent on the recruitment of the Ku70/Ku86 heterodimer, as a mutant Ku70 lacking the homeodomain interaction motif relieved all HoxC4-mediated inhibition (28). As we show here (Fig. 7), instead of recruiting Ku70/Ku86, HoxC4 bound to human hs1,2 recruits Oct-1/Oct-2 and Oca-B, consistent with the notion that Hox proteins are multifunctional transcriptional regulators that interact with different cofactors and/or components of the transcriptional machinery depending on the broader structure of their target regulatory elements (63).
The combined recruitment of HoxC4 and Oct-2 to the HoxC4 and HoxC4/Oct sites, as complemented by Oca-B, would represent a paradigm of gene regulation by homeodomain transcription factors (Fig. 10). Once the HoxC4⅐Oct-1⅐Oct-2⅐Oca-B complex is bound to hs1,2, long-range interactions with V H and FIG. 10. IgH 3 regulatory region in B cell differentiation. A, induction of the IgH 3Ј hs1,2 enhancer at sequential stages of (human) B cell differentiation. B, schematic depiction of the structure and putative long-range activity of IgH 3Ј regulatory region. Once the HoxC4⅐Oct-1/Oct-2/Oca-B complex is bound to hs1,2, long-range interactions with V H and ECS-I H promoters, presumably by looping of the 3Ј regulatory region, would confer greater IgH locus accessibility.
ECS-I H promoters, presumably by looping of the 3Ј regulatory region, would confer greater IgH locus accessibility. This would result, perhaps in the context of a promoter competition mechanism as proposed in the mouse (12,13,64), in markedly differential enhancement levels of transcription, as seen with human ECS-I␣1/I␣2 and ECS-I␥3 promoters (21). Further studies are needed to address such possible mechanisms and the role of hs1,2, hs3 and hs4 in the human IgH 3Ј regulatory region as a LCR. Such studies would require the generation of constructs containing hs1,2, hs3 and/or hs4 together with the appropriate promoters and rearrangeable, switchable, or hypermutable Ig DNA for in vitro and in vivo expression, CSR, and SHM.