Enhancer Complexes Located Downstream of Both Human Immunoglobulin Cα Genes

To obtain DNA between Cα1 and the previously reported ψγ pseudogene which lacks associated Sγ sequences, we initially sought clones containing the pseudogene, which should hybridize to a Cγ probe but not to an Sγ probe. We screened a commercial library of partial MboI-digested human placental DNA in the λFixII phage vector (Stratagene, La Jolla, CA) with a Cγ probe: probe f (Fig. 2), a 7-kb HindIII fragment, was isolated from a pBR322 plasmid clone originally derived from λ phage CH·Ig·H·g-11(32) (Health Science Research Resources Bank, Osaka, Japan; e-mail: hsrrb@nihs.go.jp). Cγ+ plaques were replated and duplicate plaque-lift filters were hybridized with the Cγ probe and an Sγ probe (1,100-bp KpnI to PstI fragment containing human Sγ2; reference 33). Southern blots of BamHI-digested DNA from eight candidate clones (Cγ+, Sγ−) were hybridized with a ³²P-labeled oligonucleotide Pseudogam-1 (see Table 1 for sequences of oligonucleotides) specific for the ψγ hinge region at 48°C in hybridization buffer (1 M NaCl, 0.1 M sodium phosphate, pH 7.0, 10% dextran sulfate, 10 mM EDTA, 1% SDS), followed by washing at 48°C in 1× SSC, 10 mM EDTA, 0.1% SDS. This analysis identified three overlapping ψγ+ clones (ψγ-25, ψγ-2, and ψγ-38) extending over an ∼30-kb region containing the ψγ gene (Fig. 2).

To obtain clones for the expected homologous region downstream of the Cα2 region, two probes derived from the farthest upstream ψγ clone (ψγ-25) were used to rescreen two duplicate plaque lifts of the λFIXII library. Probe b (Fig. 2) was a 4.5-kb fragment extending from the 5′ end of ψγ-25 to the first internal EcoRI site. Probe e was a 1.3-kb PCR fragment which lies ∼2.5 kb upstream from ψγ and which was obtained by amplification from a ψγ-25 subclone containing a 10.5-kb EcoRI-XbaI fragment which extends from the farthest downstream EcoRI site in ψγ-25 to an artificial XbaI site at the 3′ end of the phage insert. Primers used for this amplification were a reverse sequencing primer (Table 1) and elkb2, a primer based on sequence data from a ψγ-25 subclone. This probe contains a sequence homologous to a gene encoding the transcription factor elk-1, which is known to lie on the X chromosome (34). Clones EG3-6 and EG3-5 were selected for the following properties: they hybridized to probe e but not to a Cγ probe (as expected for DNA downstream of Cα2), and they had restriction maps distinct from that of the α1 locus and that of the elk-1 gene on the X chromosome, as established by genomic Southern blots of DNA from a mouse/human somatic hybrid cell line (NIGMS repository No. GM06318B; Coriell Institute for Medical Research, Camden, NJ) carrying the human X chromosome as its only human DNA. Clone H3E-1 was isolated on the basis of hybridization with probe b but not with a Cγ probe. By restriction maps and Southern blots, clones H3E-1, EG3-6, and EG3-5 were found to overlap with each other, forming a contig of ∼30 kb.

From genomic Southern blots, the phage contigs downstream of the two Cα genes were found to lie ∼20 kb away from clones containing the respective Cα genes. To bridge these gaps, a bacterial artificial chromosome (BAC) library of human genomic DNA (Genome Systems, Inc., St. Louis, MO) was screened using both a Cα membrane exon probe (Fig. 2, probe a) made by amplification from the Cα1 plasmid GMA5 using primer 5′TM-A1 in combination with primer 3′TM-B (35) and a probe extending from the most 5′ HindIII site in ψγ-25 to a position ∼1,300 bp upstream (Fig. 2, probe b′). A single BAC clone was obtained, spanning ∼120 kb, from ∼20 kb upstream of Cγ2 to 35 kb downstream from the Cα2 membrane exon. From this clone, designated BAC 11771, two subclones were prepared: a 10.5-kb HindIII fragment (A2H10.5) and an overlapping 14-kb EcoRI fragment (A2E14). Fragments of these subclones were found to have sequence homology to murine HS12, HS3, and HS4, and enhancer activity in a transient transfection assay (see Results).

Because several library screens were unsuccessful in identifying phage or BAC clones covering the gap between Cα1 and our downstream phage contig, selected DNA segments in this region were amplified by PCR. Southern blots of human genomic DNA indicated that the HS12 regions near α1 and α2 lie in opposite orientation. This inversion made it possible to selectively amplify α1-derived sequence from genomic DNA using two primers which both corresponded to the sense strand of the α2 locus. Amplification with the upstream primer SA2.5-A2 and the downstream primer SA2.1-A2 was performed using Taq polymerase XL (PE Applied Biosystems, Foster City, CA) for 32 cycles (94°C for 1 min, 61°C for 2 min, and 72°C for 10 min), and yielded the 5.5-kb A1-HS3-12 PCR fragment (Fig. 2). From this fragment a 982-bp α1 HS-3 product was amplified using primers SA2.5A and SA2.6A; and 964 bp α1HS12T and 892 bp α1HS12B fragments were amplified using primers SA2.1A and SA2.2B. For comparisons between enhancer activities of corresponding α1 and α2 fragments, the homologous 1070-bp A2HS12 PCR product was generated with the same primer pair but using plasmid pEH1.3 as template. In experiments seeking the α1 homolog of HS4, amplifications exploiting the inversion were unsuccessful. As an alternative strategy, we prepared an α1-specific genomic template for PCR: a 23-kb HindIII fragment from the human myeloma HS Sultan (American Type Culture Collection, Rockville, MD) which Southern blots indicated extended from the α1 membrane exon to a HindIII site in our ψγ-25 clone, including DNA homologous to HS4. The corresponding regions from α2 fall in 12-kb and 14-kb HindIII fragments. A 23-kb preparative electrophoresis fraction of HindIII-digested DNA from HS Sultan was used as template for amplification with primers SA8.A and SA11.B based on sequence from the α2 locus. The resulting 4.2-kb fragment (A1-HS4-3′PCR; Fig. 2) was cloned into a λZAP Express phage vector (Stratagene), yielding the clone α1HS4-4.5; the cloned insert showed a restriction map consistent with the α1 locus and distinct from the α2 locus. From the 4.2 A1-HS4-3′PCR fragment a 468-bp segment containing the α1 HS4 region was amplified using primers SA.8A and SA.9B; this segment was used for sequence analysis and enhancer studies. With the same primer pair, the corresponding α2 fragment A2-HS4 was generated using plasmid pA2E14 as template in order to enable meaningful α1 versus α2 comparisons of HS4 enhancer activity.

Nuclei were prepared and digested with DNase I according to a previously described protocol (36). K562 and HS Sultan cells (both obtained from American Type Culture Collection) were grown to densities of 5–8 × 10⁵ cells/ml. For each experiment, 3–6 × 10⁸ cells were harvested, lysed by addition of NP-40, centrifuged through a 1.7 M sucrose cushion, and resuspended in 5 ml; 450-μl aliquots of suspended nuclei were treated with serially diluted DNase I (Boehringer-Mannheim, Indianapolis, IN) to give final DNase I concentrations of 0–8 μg/ml. Nuclei were digested with DNase I for 3 min at 25°C. For one experiment HS Sultan nuclei were digested with the restriction endonuclease SspI (New England Biolabs, Beverly, MA) for 15 min at 37°C. DNase I (or SspI) digestion was terminated by adding 50 μl 1% SDS, 100 mM EDTA. DNA samples were deproteinized for 5–48 h at 37°C using proteinase K (Boehringer-Mannheim) at a final concentration of 100 μg/ml. DNA was purified by phenol/chloroform extraction and ethanol precipitation, resuspended in 50–100 μl deionized water, and DNA concentrations were measured using a Fluorometer (TKO 100; Hoeffer, San Francisco, CA). 5-μg DNA samples were digested for 5–24 h in 50 μl of appropriate restriction enzyme buffer with BglII, EcoRI, or HindIII (New England Biolabs). To assess the SspI sensitivity of the β globin locus in HS Sultan nuclei, a 1,511-bp human β globin probe was amplified from total human genomic DNA using primers β5PR-A and β3PR-B. Restriction-digested samples, together with ³²P-labeled size markers, were electrophoresed, blotted, and hybridized with the probes indicated in the figure legends. After washing the membranes, radioactive images were obtained with a PhosphorImager (Molecular Dynamics, Sunnyvale, CA).

To analyze DNA fragments for enhancer activity, we used the luciferase reporter plasmid pGL3 (Promega Corp., Madison, WI), modified so that the SV40 promoter between BglII and HindIII sites was replaced by a 76-bp Vκ promoter (37) containing an octamer motif and TATA box. This plasmid, named pGL3-Vκ, served as an enhancerless, promoter-only control. Fragments to be assayed for enhancer activity were blunt-ended with Klenow fragments of DNA polymerase and ligated with MluI linkers, or amplified with primers incorporating an MluI (or KpnI) site; the fragments were then cloned into the MluI (or KpnI) site in the polylinker upstream of the promoter in pGL3-Vκ. Plasmid DNAs for transfection were twice purified on a CsCl gradient.

Approximately 5 × 10⁶ cells in mid-log phase were washed with PBS and electroporated with either 5 μg of promoter-only control construct or equimolar amounts of enhancer test constructs along with 5 μg of CMV-β–galactosidase plasmid (pCMVβ; Clontech, Palo Alto, CA) as an internal control. Each plasmid was electroporated in triplicate in 400 μl PBS using an Electro Cell Manipulator (model 600; BTX Inc., San Diego, CA) at 200 V, 900 μF capacitance, and 13 Ω resistance in a 0.2-cm electroporation cuvette. Immediately after transfection cells were returned to culture medium. 24 h after transfection, the cells were harvested, washed with PBS, and lysed in 50 μl reporter lysis buffer (Promega Corp.) at room temperature for 20 min. After centrifugation at 15,000 rpm for 5 min, 15 μl of supernatant extracts were assayed for β-galactosidase using Galacto-Light Plus Chemiluminescent Reporter Assay kit (Tropix Inc., Bedford, MA) according to the manufacturer's instructions. To assay for luciferase activity, 35 μl of cell extracts was incubated with 100 μl of luciferase substrate [470 μM luciferin in 20 mM tricine, 1.07 mM (MgCO₃) × Mg(OH)₂ × H₂O, 2.67 mM MgSO₄, 0.1 mM EDTA, 33.33 mM DTT, 270 μM coenzyme A, and 540 μM ATP]. For both β-galactosidase and luciferase assays light output was detected in a luminometer (Dynatech Instruments, Chantilly, VA).

A PCR-based methodology employing P³³-labeled ddNTPs was used for sequencing reactions (ThermoSequenase kit; Amersham Life Science, Arlington Heights, IL). All samples were amplified for 50 cycles (95°C for 30 s; 60°C for 30 s; 72°C for 60 s). Sequencing reactions were electrophoresed on 6% gels, dried, autoradiographed, and read manually. All reported sequences were read on both A and B strands, except for the sequences of the HS12 59-bp repeats, which, despite several attempts employing various strategies, could be read only on one strand.

Because attempts to clone the human 3′ Cα regions using strategies based on gene walking from Cα or cross-species hybridization had failed in other laboratories, we used an alternative strategy based on the fact that a Cγ-like pseudogene (ψγ) has been described downstream of human Cα1 (32, 38). We reasoned that we could clone DNA downstream of Cα1 by walking upstream from ψγ. The similarity in the restriction maps reported downstream of the two Cα genes based on Southern blotting (26) furthermore suggested that these regions would be highly similar to each other. Therefore, probes derived by walking upstream from ψγ should also hybridize to clones containing DNA from downstream of Cα2.

By screening a phage library for clones that hybridized to Cγ but not Sγ, and then screening those clones with a ψγ-specific oligonucleotide (see Materials and Methods), we obtained three overlapping clones spanning ∼30 kb (Fig. 2). To obtain clones covering the corresponding region from the α2 locus, the library was then rescreened with probes b and e (Fig. 2). Clones deriving from downstream of Cα2 were selected on the basis of restriction site differences between the α1 and α2 which had been established by analysis of genomic Southern blots and comparisons with our cloned DNA from the α1 duplication. Three clones deriving from the α2 locus and spanning ∼30 kb were obtained (Fig. 2).

To estimate the distance between our two phage contigs and the corresponding Cα genes, we performed genomic Southern blot experiments using a panel of restriction enzymes with known sites in the Cα loci and in our contigs. For example, co-migrating BglII bands of ∼35 kb were found to hybridize to both probe a (from the Cα membrane exon αm) and probe b (Fig. 2); these experiments suggested a gap of ∼20 kb between each contig and the corresponding membrane exon of Cα (data not shown). For the α2 locus, the gap was bridged by a BAC clone (see Materials and Methods). The corresponding regions from the α1 locus have resisted direct cloning and have been obtained by PCR using primers designed from the sequence of the α2 locus, as described in Materials and Methods.

To map potential regulatory sequences downstream of the Cα genes, we used fragments from the cloned DNA downstream of the Cα2 gene to search for DNase I hypersensitivity sites; in cells where such enhancers or promoters are active, they generally are hypersensitive to endonucleases, apparently because binding of transcription factors disrupts the nuclease protection afforded by nucleosomes. Intact nuclei from the human myeloma HS Sultan were incubated briefly with various concentrations of DNase I; DNA samples purified from these treated nuclei were then analyzed using several Southern blot strategies in order to localize the positions of DNase I cleavage. The analyses were complicated by the fact that all of the probes we used hybridized to both the α1 and α2 loci, but various restriction map differences between the two loci allowed us to position all the hypersensitivity sites with respect to restriction sites mapped from our clones.

Fig. 3 A demonstrates analyses of DNA from the promyelocyte K562 line, representative of a cell not expressing immunoglobulin genes, and the myeloma HS Sultan line. The DNase I–treated DNA was digested with BglII, which cuts ∼1 kb upstream of the membrane exon of both Cα genes; the blots were hybridized with a probe corresponding to this exon. Since the next downstream BglII site is >20 kb away, this strategy can display hypersensitivity sites over this wide distance from the α membrane exon. The blot demonstrates at least seven hypersensitivity sites, which were subsequently assigned to the α1 or α2 locus by other blotting experiments and were named according to sequence similarity to the homologous murine regions as described below. The sites we have tentatively designated α1X and α2X do not correspond to any reported murine enhancer sequence. None of these sites were visible in the DNA from K562, in which the enhancer region is expected to be inactive.

The BglII blot of HS Sultan DNA fails to resolve the two HS4 sites because they are too far away from the BglII sites. The two HS4 sites were resolved by an alternative Southern blot strategy employing EcoRI digests of the DNA from DNase I–digested nuclei (Fig 3,B). To determine which of the two resulting bands represented HS4 from the α1 versus α2 loci, we exploited the observation from our sequence analysis that recognition sites for the restriction enzyme SspI lie in each HS4 site. Since many regulatory regions accessible to DNase I are also accessible to restriction endonucleases, we digested HS Sultan nuclei with SspI and localized the cleavage sites by isolating the DNA, digesting with EcoRI, and hybridizing Southern blots with probe b′ as shown in Fig. 3,B. From the map positions of the α1 and α2 SspI sites we assigned the SspI hypersensitivity bands as shown in Fig. 3. A control experiment on the same DNA isolated from SspI-digested HS Sultan nuclei demonstrated that SspI recognition sites in the β globin locus were not cut, indicating that the SspI sites associated with HS4 were indeed hypersensitive in HS Sultan cells (data not shown; see Materials and Methods).

Fig. 3,C illustrates an experiment which allowed assignment of several HS sites to the α2 locus rather than α1. DNA samples from DNase I–treated HS Sultan nuclei were digested with HindIII and hybridized with a probe for the HS12 site derived from the α2 locus. Although this probe hybridizes to both the α1 and α2 loci, in this DNA the α1 band is ∼23 kb, so large that any fragments from this locus that were generated by HindIII and DNase I and which hybridized to the probe would be larger than the 12-kb band representing the α2 locus. Thus all HS bands <12 kb derive from α2. This blot therefore defines the position of the HS12, X, and HS3 sites in the α2 locus. By implication, the other HS sites in the BglII blot of Fig. 3 A must derive from α1. (It should be noted that a common allele in other DNA samples shows an additional polymorphic HindIII site which cuts within the 23 kb corresponding to the HS Sultan band; this allele would have confounded the above strategy, but was absent in HS Sultan.)

To analyze enhancer activity in the cloned DNA downstream of the Cα2 locus, fragments containing one or more DNase I hypersensitive sites were subcloned into a luciferase reporter gene driven by a Vκ promoter as described in Materials and Methods. Luciferase activity in each sample was normalized to the β-galactosidase activity of the promoter-only control plasmid, and expressed as fold-increase over luciferase activity of that control plasmid.

The luciferase assays (Fig. 4) revealed strong enhancer activity in the 5-kb SmaI-HindIII fragment (SM5) which was found to contain sequence homologous to murine HS12 (see below). Within this fragment, enhancer activity seemed to be confined to the 1.3-kb EcoRI-HindIII fragment (EH1.3) containing the HS12 site. This segment was further cut into an upstream 0.3-kb EcoRI-PstI fragment (EP300), a 0.3-kb PstI-PstI fragment (P300), and a 0.6-kb PstI-HindIII fragment (PH600). Of these, only the P300 fragment, which contained HS12, showed enhancer activity. However, it should be noted that the enhancer activity of P300 was less than that measured for a larger PCR-generated fragment A2HS12. Thus it is possible that additional elements that do not show intrinsic enhancer activity when isolated in constructs can nevertheless augment the activity of the core HS12 enhancer lying in the P300 fragment. Furthermore, the EH1.3 fragment, but not the slightly shorter A2HS12 fragment, showed significantly less enhancer activity in the A orientation, suggesting the possibility of inhibitory sequences located near one of the ends of EH1.3.

Two fragments containing the α2 HS3 site showed no significant enhancer activity in the luciferase construct, but a plasmid in which the HS3-containing 0.7-kb BamHI-SmaI fragment was dimerized showed consistent low activity (Fig. 4, SM0.7-X2). The homologous murine enhancer was reported to show greater activity with a c-myc promoter than with an immunoglobulin Vλ promoter (12), but we found no significant increase in enhancer activity when the Vκ promoter was replaced by a human c-myc promoter (data not shown; construct described in Materials and Methods).

Two restriction fragments, B5.5 and SpB0.9, containing the α2 HS4 site demonstrated significant stimulation of luciferase activity, but only when cloned in the B (inverted) orientation (Fig. 4 A). When these fragments were oriented so that the DNA strand continuous with the transcribed strand in the luciferase gene corresponded to the transcribed strand of the immunoglobulin gene in genomic DNA (the A orientation), luciferase activity was no greater than that from the promoter-only plasmid. Unexpectedly, a smaller fragment (the 468-bp PCR-generated fragment designated A2-HS4, designed to span the sequence showing the strongest homology to the murine HS4 site) was found to give a strong activation of luciferase activity. The stronger activation of luciferase conferred by the short A2-HS4 PCR fragment compared with that of the longer SpB0.9 fragment (in the same A orientation) suggests that the latter may contain an inhibitory sequence that is either position- or orientation-dependent. This possibility is currently being explored.

Because of the lack of genomic clones spanning the HS3, HS12, and HS4 regions of α1 locus, fragments corresponding to each of these sites were obtained by PCR and analyzed; and, to facilitate α1 versus α2 locus comparisons, the same primer pairs were used to generate corresponding fragments from the α2 locus. As shown in Fig. 4, the activity of each HS site fragment from α1 was similar to that from the α2 locus (compare SM0.7-X2 α2 versus HS3-X2 α1; A2HS12 α2 versus A1HS12T α1; and A2-HS4 α2 versus A1-HS4 α1). The PCR amplification of the HS12 site yielded two fragments, designated A1HS12T (top) and A1HS12B (bottom), which differ in size by <0.1 kb and may represent alleles (see below). Both fragments showed substantial enhancer activity (Fig. 4 B).

Since the mouse 3′α enhancers show different activation patterns during B cell differentiation, with HS4 being activated at the pre–B cell stage, while HS12 and HS3 are only active in mature B cells, we examined the activity of the human 3′ α2 enhancers in a range of human and mouse B cell lines (Table 2). The pattern of activation for the human HS12 enhancer is similar to that of mouse HS12; i.e., human HS12 is inactive in the human pro–B cell line FLEB-14 and the mouse pre–B cell line 18-81, but functions in the human mature B cell line Raji, as well as three plasmacytomas (HS Sultan, human; S194, mouse; and MOPC 315, mouse). HS3 is also inactive in mouse 18-81 pre–B cell line, but shows modest activity in most of the more mature lines tested. HS3 shows surprisingly strong activity in the mouse S194 myeloma, indicating that unknown factors varying between cell lines at similar stages of differentiation can modulate the activity of this enhancer. Finally, the HS4 enhancer shows strong activity in the human pro-B cell line FLEB14, and is also variably active in all of the more mature cell lines (except 18-81) in which this enhancer was assayed in the B orientation.

The nucleotide sequence of all DNase I hypersensitivity sites was determined, revealing ∼99% sequence identity between the human α1 and α2 elements, and similarity between the human and mouse enhancers ranging from 74 to 90%.

HS12, the strongest enhancer, showed 90% sequence identity to the homologous murine enhancer over a 135-bp core homology (Fig. 5,A). In the α2 locus, four tandem repeats with a 59-bp consensus sequence lie immediately upstream of the HS12 core. However, this sequence has been inverted in Fig. 5 A, to facilitate comparison with the homologous α1 sequence in opposite orientation. In the corresponding region of α1 (which, due to the inversion, lies downstream of the core homology region) a 115-bp deletion removes the second and third repeats; the α1HS12B region shows an additional deletion of 70 bp. The core homology region includes several of the functional motifs identified in the murine enhancer: the AP1-Ets site reported to confer responsiveness to B cell receptor cross-linking (22, 39); an exact octamer sequence (ATGCAAAT); and a μE5 site (except in the α1HS12B sequence, in which the μE5 is missing owing to the 70-bp deletion). The sequences of these three motifs from the human HS12 are identical to their murine homologs except for a single base change in the AP1 site which causes the human sequence to exactly match the consensus AP1 site, where the murine motif has one mismatch. The murine element designated μE1 (39), which has never been thoroughly documented even in the murine enhancer, is poorly conserved in the human homologs. Although the murine binding site for NF-κB lies outside the 135-bp region of strongest sequence similarity, a reasonable match to consensus for this element is found in a position roughly homologous to the murine κB site in the α1HS12T and α2HS12 sequences, but is part of the 70 bp deleted in α1HS12B. One of the mouse BSAP sites (BSAP2) is not conserved, but most residues in a second mouse BSAP binding site (BSAP1) are maintained in the human α2 HS12 enhancer. The murine αP site, which binds to an ETS-related transcription factor which augments enhancer activity (17), is not conserved in the human sequences.

The DNase I hypersensitive sites lying <3 kb downstream from the membrane exons of Cα1 and Cα2, roughly in the position of the weak enhancer reported by Matthias and Baltimore (11) and here designated HS3A, were found to contain sequences which are 74% identical to the murine HS3 over a 200 bp core segment. The two human HS3 segments are identical in the 326 bp shown (Fig. 5,B), and lie in the same orientation as HS3A. We have assumed that the correct orientation of the murine HS3A sequence is that described by Chauveau and Cogné (13). This orientation is opposite to that of the murine HS3B, which lies downstream from HS12 in the mouse, as described by Madisen and Groudine (12). Of the enhancer/HS sites downstream of murine Cα, HS3 is the least well investigated for functional motifs, in part because of its weak enhancer activity. Independent sequence analysis of the murine HS3A and HS3B regions detected several similarities to octamer motifs, AP1 sites, and consensus E box motifs (CANNTG). The AP1 site identified in Fig. 4 is a precise match to the consensus AP1 binding motif TGANTCA (40) in the two human and two mouse HS3 sequences, and the murine sequence has been shown to bind to c-jun and c-fos in vitro (Neurath, M., personal communication). Similarly, several of the E box consensus motifs in the murine sequence have been shown to bind in vitro to proteins of the HLH family (Neurath, M., personal communication); some of these motifs are conserved in the two human HS3 sequences. The significance of the conserved motifs remains uncertain in the absence of a functional analysis of HS3 sequences.

The DNase I hypersensitivity sites furthest downstream from human Cα1 and Cα2 in Fig. 3, which we have designated HS4, are 76% similar to the murine HS4 site over a core 145-bp sequence which spans the three functional motifs demonstrated in murine HS4 (23); see Fig. 5,C. The NF-κB motif and the downstream octamer motif in Fig. 5 C are both precisely conserved; in the murine HS4 these motifs both contribute to functional enhancer activity (23). In contrast, the BSAP site which upregulates murine HS4 enhancer activity in B cells but downregulates it in pre–B cells is completely absent from the human HS4 sequences.

Lying between HS3 and HS12 in both the α1 and α2 loci are DNase I hypersensitive sites which are not associated with any of the known enhancer elements, but that do map to the position of a 61-bp segment of 70% mouse- human homology. These conserved regions are provisionally designated X sites in part because of their unknown function, and in part because both the human and mouse segments contain XbaI restriction enzyme sites. In mouse this sequence is duplicated as part of the large inverted repeat centered on HS12, so that one copy lies between HS3A and HS12 while a second copy lies between HS12 and HS3B. A segment of (GA)_n repeats is found near the X site in the direction of HS12 at an interval of 80 bp for both mouse X sites (13) and an interval of ∼70 bp for the human (data not shown). Within the 61-bp conserved segment, the most highly conserved sequence is a consensus heat shock element (HSE; 41, 42). An HSE could potentially bind heat shock transcription factors (HSTFs), which are known to activate several heat shock response genes (HSP70, HSP90) in response to cellular stress such as heating (43–45). Fragments containing an X site do not appear to dramatically affect enhancer activity in HS Sultan, but may contribute to regulation through mechanisms not captured in our transient transfection assays.

The enhancers clustered 3′ of Cα in the mouse IgH locus can activate the upstream genes, functioning as an LCR. In human IgH locus, arrays of enhancers homologous to those 3′ of mouse Cα are located at two positions within the human IgH locus, 3′ of each Cα gene. On the basis of sequence homology and conservation of restriction sites between the two human enhancer arrays, it is apparent that these enhancers lie near the 3′ ends of the two duplication units which encompass the γ3-γ1-ψε-α1 and γ2-γ4-ε-α2 gene clusters (25–27), indicating that the 3′α enhancer arrays were present in an approximation of the human arrangement preceding the duplication event that gave rise to the present human IgH locus structure. Moreover, the arrangement downstream of both human Cα genes is 5′-HS3-HS12-HS4-3′, in contrast to the large palindromic structure downstream from mouse Cα that contains a 5′-HS3A-HS12-HS3B-HS4-3′ arrangement (13). Therefore, although an arrangement containing an HS3 enhancer proximal to the Cα membrane exon and an HS4 enhancer farther downstream would seem to have been present in the common ancestor of rodents and primates, the mouse HS3A-HS12-HS3B palindrome probably arose after the primate–rodent divergence. Finally, there is a major structural difference between the 3′α1 and 3′α2 enhancer arrays; namely, that a DNA segment containing HS12 is inverted between the two loci. Using probes containing the 135-bp human HS12 core, it should now be possible to examine DNA from a number of primates for inversion of 3′α2 HS12 relative to 3′α1 HS12; such data may indicate when in evolution the inversion event occurred, and which orientation was present initially in the locus. What caused this inversion? Interestingly, the single mouse HS12 lies in opposite orientation from the rat HS12, and both are flanked by inverted repeats (13) which are known to mediate inversions in other genomic contexts, e.g., in the iduronate-2-sulfatase gene causing Hunter syndrome (46) and in the factor VIII gene (47). Limited Southern blot experiments have not provided evidence for inverted repeats flanking the human HS12 sequences (data not shown). Some hints about the mechanism of the inversion may be found when the inversion breakpoints are identified and sequenced, work currently in progress in our laboratory.

The 135-bp HS12 core homology sequence is likely to contain essential motifs important for the strong, late, B cell–specific enhancer activity characteristic of HS12 in mice and humans. Although the function of transcription factor binding sites within the human HS12 core has not yet been demonstrated experimentally, this segment contains sequences nearly identical to the murine AP1, ETS, Oct, and, in α1HS12T and α2HS12, μE5 motifs, all of which are functional in the mouse HS12 enhancer. However, the high degree of sequence conservation in the HS12 core homology extends beyond the transcription factor–binding sites identified in the mouse enhancer, indicating that there may be additional conserved motifs that have not been characterized in either mice or humans.

Despite the fact that a number of other transcription factor motifs whose function has been demonstrated in the mouse HS12 lie outside the HS12 135-bp core homology and are absent in one or more of the human α1HS12T, α1HS12B, and α2HS12 enhancers that we have studied, these enhancers all show roughly equivalent activities. This result suggests that elements missing from these enhancers are not essential for enhancer function in HS Sultan. These inconsistently conserved elements include human sequences corresponding to μE5 and NF-κB sites (absent from α1HS12B) and the BSAP2 site (absent from both α1 alleles). On the other hand, a 1.3-kb α2 EcoRI–HindIII fragment containing the α2 HS12 core plus considerable flanking sequence shows a dependence of enhancer activity on orientation of the fragment (Table 2), suggesting that uncharacterized elements beyond the HS12 core may have some inhibitory function.

Outside the human HS12 core are GC-rich 59-bp repeat units which by themselves do not have enhancer activity in the HS Sultan myeloma (EP300, Fig. 4,A), and are not conserved between mice and humans (Fig. 5 A). However, it is possible that these repeats contribute to enhancer activity because they are present in the A2HS12 PCR-generated fragment, which gives significantly higher enhancer activity than we observe in the p300 fragment containing the core homology. Deletions of the 59-bp repeats have given rise to apparent allelic polymorphisms, as evidenced by α1HS12T (deletion of the second and third repeats found in α2HS12), α1HS12 (deletion extending from 28 bp 5′ of the first repeat through the third repeat), and other alleles (Harindranath, N., unpublished results).

The other two enhancer components of the mouse 3′α LCR, HS3 and HS4, are weaker enhancers than HS12, but nonetheless are essential for locus control activity (12). Although these elements are less well characterized than the HS12 enhancer, the existing data indicate general human–mouse similarity of the HS3 and HS4 elements, with some notable differences.

In the mouse system, the HS3A element assayed in CAT reporter gene constructs driven by c-fos or thymidine kinase promoters (11) showed weak enhancer activity, although the nearly identical HS3B enhancer showed substantial activity in certain constructs with other promoters tested by another laboratory (12). These disparate results resemble our data on the human α1 and α2 HS3 elements in that single copies and dimers of human HS3 generally gave very low enhancer activity, except in the mouse S194 myeloma in which the same constructs gave substantial enhancer activity comparable to that of HS12 (Table 1). Taken together, these data suggest that HS3, though typically the weakest of the 3′α enhancers, contains uncharacterized motifs that in some cells and/or in combination with certain promoters, can mediate a strong enhancer function.

HS4 is the most downstream 3′α enhancer in both mice and humans, and shows activity intermediate between that of HS3 and HS12. The HS4 enhancer data in the mouse (12, 48), as well as our data on the human α1 and α2 HS4 elements, demonstrate that HS4 is active from the early stages of the B cell lineage onward (Table 2), and thus is qualitatively different from HS3 and HS12. In mouse HS4, there is a binding site for the BSAP, which is expressed in the early B cell lineage. However, in the human α1 and α2 HS4 enhancers, the BSAP site is deleted, indicating that BSAP binding is not an essential feature for HS4 activation in human pre–B cells. The human HS4 is inactive in the 18-81 mouse pre–B cell line, which was reported to support the activity of mouse HS4 (12). The significance of this difference is not clear; it could be related to the BSAP site deletion in the human HS4, or perhaps to other differences between the mouse and human HS4 sequences.

The DNase I X sites may represent novel control elements that function together with the HS3, HS12, and HS4 enhancers to activate the IgH locus. Although the significance of the conserved HSE motif is unclear, binding of HSTF protein to an HSE has been shown to be critical for maintaining the DNase I hypersensitivity of the yeast HSC82 gene promoter (49). Heat shock activation of the Drosophila HSP70 gene promoter results from binding of HSTF to HSE sites after accessibility of the HSEs has been established by binding of the GAGA protein to adjacent (GA)_n repeats (50). This demonstrated interaction between HSEs and (GA)_n motifs suggests that the location of a (GA)_n repeat region 70–80 bp away from the X site HSE may be of some significance. Furthermore, in the context of IgH gene regulation, it is of interest to note that HSE motifs have been shown to respond to IL-2 and IL-4 (51).

In the mouse, it has been demonstrated that when HS3, HS12, and HS4 are linked together in a construct containing the c-myc gene and stably transfected into the Raji human B cell line, the c-myc gene is transcribed independent of integration site (12). This observation suggests that the HS1234 combination confers LCR activity, although LCRs have more typically been described based on position-independent transcription of mouse transgenes rather than genes transfected into a cell line. Because the regions 3′ of the human Cα1 and Cα2 genes contain similar HS3, HS12, and HS4 elements that function as enhancers, it is reasonable to hypothesize that these elements also function together in the human system as LCRs. The different arrangement of 3′α enhancers in mice and humans (HS3A-HS12-HS3B-HS4 versus HS3-HS12-HS4) may cause some functional differences in these control regions. Moreover the distance between 3′α enhancers also differs between mice and humans, with the mouse enhancer complex spanning a 30-kb region (13), whereas both the human 3′α1 and 3′α2 enhancers span ∼15 kb.

Our finding that arrays of enhancers homologous to those in the mouse 3′α LCR lie downstream of both human Cα genes raises the possibility that differences in the activation of each human γ–γ–ε–α duplication unit result from differences between the putative 3′α1 and 3′α2 LCRs. Even though sequence comparison shows that there is near identity between homologous enhancer elements in the α1 versus α2 locus (Fig. 5), transcription and expression of the upstream heavy chain duplication unit (γ3-γ1-ψε-α1) is greatly elevated relative to the downstream unit (γ2-γ4-ε-α2; reference 2). This difference could result from the fact that the 3′α2 HS12 element is inverted relative to the 3′α1 HS12, and is also at a greater distance from HS3 than in the α1 locus, possibly reducing synergistic interactions between HS3 and HS12. Alternatively, it may be that only the 3′α1 enhancers are activated early in B cell development, possibly falling under the influence of the upstream Eμ enhancer, which itself can function as an LCR (52, 53). In this model, the Eμ and 3′α1 enhancers together would activate a large domain encompassing the first duplication unit, whereas the second duplication unit and the 3′α2 enhancers would fall outside this combined domain. Thus activation of the second duplication unit would depend solely on the 3′α2 enhancers, and expression of genes in this unit might therefore be reduced. Clarification of the basis for this difference will have to await experiments that involve specific deletion of either the 3′α1 or 3′α2 enhancers, as well as studies identifying matrix attachment sites and chromatin insulator elements that define domains within the human IgH region (52–54).

In the work presented here, we have laid the foundation for experimental studies on the activation of the human IgH gene transcription, as well as the regulation of isotype switching. In addition, knowledge of the action of these enhancers on distant constant region genes should contribute to a general understanding of the mechanisms underlying activation of large gene clusters.

We thank Drs. Rose Mage, Louis Staudt, and Steven Bauer for discussions and useful comments on this manuscript. We also thank Dr. Markus Neurath for communicating unpublished results on functional elements in the mouse HS3 enhancers.