Exploring Functional Redundancy in the Immunoglobulin μ Heavy-Chain Gene Enhancer

Abstract

Immunoglobulin (Ig) μ heavy-chain gene enhancer activity is mediated by multiple DNA binding proteins. Mutations of several protein binding sites in the enhancer do not affect enhancer activity significantly. This feature, termed redundancy, is thought to be due to functional compensation of the mutated sites by other elements within the enhancer. In this study, we identified the elements that make the basic helix-loop-helix (bHLH) protein binding sites, μE2 and μE3, redundant. The major compensatory element is a binding site for interferon regulatory factors (IRFs) and not one of several other bHLH protein binding sites. These studies also provide the first evidence for a role of IRF proteins in Ig heavy-chain gene expression. In addition, we reconstituted the activity of a monomeric μ enhancer in nonlymphoid cells and defined the domains of the ETS gene required for function.

The immunoglobulin (Ig) μ heavy-chain gene enhancer activates transcription and V(D)J recombination at the Ig heavy-chain (IgH) locus in precursor B cells. Enhancer function is mediated by multiple DNA binding proteins that interact with the μ enhancer. Most of these factors are ubiquitously expressed in both B cells (where the enhancer is active) and non-B cells (where the enhancer is inactive), while a smaller subset of factors have more restricted tissue distribution (9, 19, 20). The basic helix-loop-helix (bHLH) family of proteins constitute a major portion of the ubiquitously expressed factors, whereas ETS and POU domain genes are examples of the second subset. However, no single factor that can account for the cell type specificity of IgH gene expression has been identified. It is likely, therefore, that activation of the enhancer at a precise developmental stage during B-cell ontogeny is governed by the combinatorial properties of several proteins.

In principle, identification of proteins that define a B-cell-specific transcriptional unit would allow further reconstruction of the functional multiprotein-DNA complex. However, mechanistic analysis of μ enhancer function is severely complicated by the multiplicity of protein binding sites within the enhancer. Several deletion and mutational studies have defined regions of the enhancer that activate transcription and the importance of specific protein binding sites within each enhancer fragment (13, 14). A feature of the enhancer revealed by these studies was that several binding site mutations affected enhancer activity minimally. For example, mutation of the μE3 element in the context of a 250- or 470-bp enhancer fragment decreased enhancer activity by 25 to 30%. Similarly, mutation of the μE1 or μE2 element individually did not affect μ enhancer activity significantly. These observations have been interpreted to mean that there is functional redundancy among various elements, so that absence of a particular element is functionally compensated for by other elements. The presence of functional redundancy further complicates enhancer analysis by making it difficult to discern which combination of factors is responsible for enhancer activity.

To systematically circumvent these problems, we first identified the smallest domain of the enhancer that conferred B-cell-specific transcriptional activity; such an enhancer should contain no redundant elements. Characterization of the tripartite 70-bp minimal enhancer (μ70) resulted from these studies (7, 8, 20–22, 24). The μA and μB sites within μ70 bind ETS domain proteins, and the intervening μE3 element binds various leucine zipper-containing bHLH proteins, as well as the core binding factor (CBF). The μ70 enhancer has several features that suggest it is an appropriate starting point for the analysis of the full enhancer. First, like the full enhancer, it is composed of elements that bind tissue-restricted proteins and one that binds a ubiquitous protein. Second, none of the three elements can individually activate transcription in B cells either as a monomer or as multimers, suggesting that transcription activation in B cells is a combinatorial property. Third, activity in B cells requires all three elements, which indicates that there is no redundancy in this enhancer. Presumably, the necessity of μE3 in this context is due to the absence of any other μE motif in the minimal enhancer which can functionally substitute for it.

We then incorporated a second μE element, μE2, into our analysis (4). Addition of a second μE element was expected to increase the activity of the tripartite μ70 enhancer. Furthermore, because the sequences of the μE2 and μE3 elements are very similar, it was possible that the μE3 element would be dispensable in the context of the four-part enhancer; that is, μE2 could functionally substitute for μE3. As expected, inclusion of μE2 substantially enhanced transcriptional activity. However, we found that the μE3 element was still absolutely essential for function. These observations demonstrated that the μE2 and μE3 elements were functionally not equivalent and showed that μE2 activity was dependent on the μE3 site, providing evidence for communication between μE elements. We proposed a plausible mechanism for μE2-μE3 synergy based on in vitro DNA binding analyses.

Despite the insights gained from the analysis of the three- and four-part enhancers, these studies did not address two important properties of the μ enhancer. First, both enhancers described above were only weakly active as monomers, and dimerization was necessary for robust transcriptional enhancement. Second, these studies did not provide insights into the basis for functional redundancy among the μE elements. Our goal in the present study was to characterize the smallest enhancer fragment that is active as a monomer and identify the motifs that compensate for the loss of elements such as μE2 and μE3.

We found that the smallest monomeric enhancer contains five elements, μA, μB, μE2, μE3, and μE5. All five elements were essential for enhancer activity, suggesting that this was a minimal monomeric enhancer with no functionally redundant motifs. The B-cell-specific transcriptional activity of this fragment could be reconstituted in COS cells by coexpression of the ETS proteins PU.1 and Ets-1 and the bHLH protein E47. The question of redundancy was explored by extending the enhancer fragment both 5′ and 3′. Surprisingly, a fragment containing a fourth μE element, μE1, still required μE3 for activity; that is, even in the presence of μE1 to μE3 and μE5, no other μE element could functionally substitute for μE3. However, an enhancer extending further 3′, but incorporating no additional μE elements, showed a reduced requirement for μE3. Deletions and point mutations were used to localize the putative element, which was shown to be a binding site for the family of interferon regulatory factors (IRFs) (15, 16, 26). These observations identify the basis for redundancy of μE elements and dispel the assumption that μE elements functionally substitute for each other. In addition, we provide the first evidence for a role of IRF family proteins in Ig gene expression.

MATERIALS AND METHODS

Plasmids.

The wild-type (WT), μA⁻, μE3⁻, and μB⁻ μ170 plasmids have been described previously (18, 21). The μE2⁻, μE2⁻ μE3⁻, C2⁻, C2⁻ μE3⁻, μ3′⁻, μ3′⁻ μE3⁻, and μ3′ μE2⁻ μ170 plasmids were generated by first subcloning the corresponding mutant HinfI-DdeI fragments (nucleotides 346 to 518 according to the numbering system of Ephrussi et al. [6]) into the EcoRV site of pSP72. The BglII-ClaI fragments isolated from these subclones were then blunted with Klenow fragment and cloned into the SalI site of reporter plasmid Δ56CAT.

The μ74 plasmids were made by cloning the Sau3A-BamHI fragments (nucleotides 359 to 432; the BamHI site was introduced by mutating the core 1 [C1] site [12]) of WT or mutant enhancer fragments into Δ56CAT, which was cut by SalI and treated with Klenow fragment.

The μ89 plasmids contain the HinfI-BamHI (nucleotides 346 to 432) enhancer fragment. Subclones of WT and μE2⁻ HinfI-DdeI fragments in the EcoRV site of pSP72 were digested with BglII and BamHI. Subclones of μA⁻ and μE3⁻ HinfI-DdeI fragments in the HincII site of pBluescript were digested with XhoI and BamHI. The resultant enhancer fragments were treated with Klenow fragment and cloned into the SalI site of Δ56CAT.

The μ128 (nucleotides 345 to 472) and μ152 (nucleotides 345 to 496) fragments were generated by PCR using μ170/pSP72 or μE3⁻μ170/pBluescript as the template and the following oligonucleotides as primers: 5′-GGGGTCGACGAGTCAAGATGGCCGATC-3′ (5′); 5′-GGGCTCGAGACTTCTTCAAACCACAGC-3′ (3′-μ128); and 5′-GGGCTCGAGCTGGACAGAGTGTTTC-3′ (3′-μ152). The PCR products were cut by SalI and XhoI and cloned into the SalI site of Δ56CAT.

The nucleotide sequences for mutations at μE5, μE2, μA, μE3, and μB sites have been described previously (4, 20, 22). For core 2 (C2) mutation, GTG (nucleotides 458 to 461) was changed to CGA; for μ3′ mutation, GAAA (nucleotides 507 to 510) was changed to CATG.

Mammalian expression vectors, all of which have been described before, were pEVRF-PU.1 and pEVRF-Ets-1 (20); Ets-1Δ167, Ets-1Δ231, Ets-1Δ286, and ETS(PU) (8); pRC.E47 (4); pAct-IRF-1 and pAct-IRF-2 (10); and Pip/CMV (5). The bacterial expression vector for glutathione S-transferase (GST)–Ets-1 was generated by cloning the BamHI fragment of pEVRF-Ets-1 into the BamHI site of pGEX.2T. Protein was purified as previously described (4).

Transfections.

Murine S194 and human DHL-9 cells were grown in RPMI medium supplemented with 5% fetal bovine serum, 5% calf serum, and 50 μg each of penicillin and streptomycin per ml. COS cells were grown in Dulbecco modified Eagle medium containing 10% newborn bovine serum and antibiotics in the amounts specified above. Transfections, cell extract preparation, and chloramphenicol acetyltransferase (CAT) activity measurements were carried out as previously described (4).

In vitro translation.

In vitro translation reactions were performed with 1 μg of plasmid HAPip1-380 (2) or pSP64-IRF1 and a TNT T7 quick coupled transcription-translation kit or TNT SP6 coupled transcription-translation kit (Promega), respectively.

Electrophoretic mobility shift assay (EMSA).

The μE3⁻ Pst-C2 fragment used in Ets-1 and CBF binding was generated by digesting with PstI the PCR product used for cloning plasmid μ128. The sequence for the λB probe used in the binding of HAPip1-380 is 5′-GAGAAATAAAAGGAAGTGAAACCAAG-3′. Binding conditions were as previously described (4).

RESULTS

Minimal monomeric μ enhancer.

A schematic of the μ enhancer region is shown in Fig. 1A. A central region of approximately 170 nucleotides is densely packed with protein binding sites that are indicated by assorted geometric shapes. Flanking the core enhancer are matrix attachment regions which have been proposed to increase the distance over which the central region can exert its influence (11). Our earlier studies have assayed a 70-bp enhancer (μ70) containing three elements and a 59-bp enhancer (μ59) that contains four elements. Both fragments needed to be dimerized to enhance transcription significantly. To identify a minimal monomeric enhancer, we extended μ59 at the 5′ end to incorporate the adjacent μE5 element (μ74) and tested its activity by transient transfection into S194 plasma cells. Whereas μ59 monomer was a very weak activator (approximately four- to fivefold above the background of the enhancerless reporter plasmid [data not shown]), addition of one more element made μ74 a strong enhancer (Fig. 1B). Mutations in any of the five elements significantly decreased enhancer activity, showing that there were no redundant elements in this fragment (Fig. 1B). Thus, μ74 is the smallest μ enhancer fragment that is active as a monomer.

FIG. 1.

(A) Schematic representation of μ enhancer derivatives used in this study. Squares represent CAXXTGG motifs that are binding sites for ubiquitous transcription factors. The bHLH protein, E47, binds μE5 and μE2 elements; several leucine zipper bHLH proteins, such as TFE3 and USF, bind to μE3; the hematopoietic cell transcription factor, CBF, also binds to μE3; the μE1 site binds the factor YY-1 (23). The μA and μB sites bind proteins belonging to the ETS family: PU.1, a macrophage- and B-cell-specific ETS protein, binds to μB; several proteins, such as Ets-1, Erg-3, and Fli-1, bind to μA. C1 to C3 designate three other CBF binding sites; in all enhancer derivatives used here, the C1 site has been mutated. The oval marked μ3′ represents a new element identified in this study. This motif functionally compensates for the loss of a μE3 or μE2 element and binds the proteins IRF-1 and IRF-2. Hf, HinfI. (B) The smallest μ enhancer fragment active as a monomer contains five essential motifs, as determined by transient transfection analysis of the WT and mutated μ74 enhancer derivatives in S194 plasma cells. Δ56 is the enhancerless reporter containing a c-fos gene promoter extending 56 nucleotides 5′ of the transcription initiation site. All enhancer derivatives were cloned upstream of the minimal promoter, and 5 μg of plasmid DNA was used for transfection. Results shown are averages from three experiments done in duplicate and normalized to the activity of the WT enhancer, which is assigned a value of 100.

Reconstruction of μ74 activity in nonlymphoid cells.

We have previously shown that the μ70 dimer is activated by coexpression of PU.1 and Ets-1 in nonlymphoid cells. Transactivation required all three elements of this enhancer and utilized an endogenous μE3 binding protein. We sought to reconstitute the activity of the μ74 enhancer in similar assays. Cotransfection of only the ETS genes activated this enhancer weakly (Fig. 2A, bar 4). Because endogenous μE3 binding proteins can work with the transfected ETS genes, the missing components were likely to be μE2 and μE5 binding proteins. Several earlier observations indicated the choice of E47 as a μE2/μE5 binding protein. E47 was originally cloned as a μE5 binding protein (3, 17), and the μE2 element falls within its consensus recognition site (25). E47 is known to be of importance in B-cell development (1, 27), and it can synergize with μE3 binding proteins to activate transcription (3, 4). Coexpression of E47 with either Ets-1 or PU.1 also activated the enhancer weakly (Fig. 2A, bars 2 and 3). However, inclusion of both ETS genes and E47 resulted in significant transcriptional activity (Fig. 2A, bar 5), which was dependent on all five sites of the μ74 enhancer (Fig. 2A, bars 6 to 10). The mutational analysis in COS cells was very similar to the pattern seen in B cells, suggesting that this “heterologous” assay reproduced several aspects of the transcriptional activity seen in B cells. Furthermore, the requirement for the μE3 site showed that a COS cell protein was recruited to this site, because none of the three exogenously expressed proteins bound significantly to the μE3 site.

FIG. 2.

Reconstitution of monomeric μ enhancer activity in nonlymphoid cells. (A) Reporter (2 μg) containing the μ74 enhancer (WT) was cotransfected into COS cells together with expression vectors for E47 (μE2/μE5 binding protein), Ets-1 (μA binding protein), and PU.1 (μB binding protein) in various combinations as noted below the graph. Total transfected DNA was kept constant at 5 μg by using pEVRF. Reporters (2 μg) containing mutated μ74 derivatives (μE5⁻, μE2⁻, μA⁻, μE3⁻, and μB⁻) were similarly assayed in the presence of all three transactivators (last five bars). Results shown were obtained by averaging three experiments done in duplicate and were normalized to the activity of the WT μ74 reporter in the presence of all three transactivators. (B) Domains of Ets-1 (440 amino acids) and PU.1 (272 amino acids) are shown at the top. The N-terminal Ets-1 deletion mutants used for the analysis are shown by the arrows marked Δ167, Δ231, and Δ286. The shaded box marked TD is a previously identified transcription activation domain, the box marked INH is an autoinhibitory domain for DNA binding, and the hatched box at the C terminus is the DNA binding ETS domain. A TD in PU.1 is shown as the cross-hatched box, and the DNA binding domain is shown as a hatched box. The deletion mutant of PU.1 contains the ETS domain plus C-terminal 14 amino acids. For cotransfection analysis of COS cells, the WT μ74 reporter was transfected along with E47, Ets-1, PU.1, or deletion mutants of the latter two, as indicated below the graph. Results shown are averages of three experiments carried out in duplicate, normalized to μ74 activity in the presence of all three full-length proteins.

Activation of the dimeric μ70 enhancer requires an N-terminal transactivation domain (TD) in Ets-1 but only the DNA binding ETS domain of PU.1 (8). We expressed PU.1 and Ets-1 deletion mutants to identify the domains of PU.1 and Ets-1 necessary to activate the μ74 monomer. The N-terminal TD of Ets-1 contributed significantly to enhancer activity (Fig. 2B; compare full-length Ets-1 with Δ167, Δ231, and Δ286), as shown previously with the μ70 enhancer. In this case, however, the ETS domain of Ets-1 also provided some transactivation potential because enhancer activity with EtsΔ286 was still significantly greater than that seen with PU.1 and E47. It is possible that the Ets-1Δ286 construct enhanced transcriptional synergy between the upstream μE elements (μE2 and μE5) and μE3. This aspect of Ets-1 function could not be assayed with the μ70 enhancer, which does not contain either μE2 or μE5. In contrast to the observations with Ets-1, the ETS domain of PU.1 was sufficient to transactivate μ74 in the presence of Ets-1 and E47 (Fig. 2B, rightmost bar). (The protein deletion mutants have been previously shown to be expressed at levels comparable to those of the full-length genes [8].) These observations strengthen the suggestion that a previously defined TD in PU.1 is not necessary to activate the μ enhancer.

μE3 redundancy.

Mutational analysis of the μ74 enhancer in B cells showed that μE3 was an essential component of this five-part enhancer. Yet, it is well established that mutation of μE3 in the context of longer enhancer fragments does not significantly reduce enhancer activity; that is, its function can be largely compensated for by other, presently undefined elements. To identify elements that make μE3 redundant, we assayed two other μ enhancer fragments and mutants thereof by transient transfection into S194 cells. μ87 contains one more μE element than μ74 (Fig. 1A) and was a strong enhancer (Fig. 3A, bar 2). Despite the presence of four μE elements in this fragment, mutation of μE3, μE2, or μE5 abolished enhancer function (Fig. 3A), showing that μE3 was still essential.

FIG. 3.

Elements that contribute to μE3 redundancy. (A) μE3 redundancy does not depend on other related μE elements. Activity of the μ87 enhancer (WT) that contains four μE elements and mutations thereof was analyzed by transient transfection into S194 plasma cells. Averaged CAT activity from three independent transfection experiments is shown normalized to the activity of the WT μ87 enhancer. (B) 3′ enhancer sequences that contain no μE-related motif compensate functionally for μE3, as determined by transient transfection analysis of μ170 enhancer and its mutated derivatives in S194 plasma cells. Results from three independent experiments are shown. (C) μ enhancer sequences that compensate for μE3. Deletion mutants of the μ170 enhancer shown schematically above the graph, or the μE3 mutated derivative of each, were transfected into S194 followed by CAT enzyme analysis. C2⁻μ170 refers to a C2-mutated μ170 enhancer. Data shown are averages of three independent transfections carried out in duplicate and are normalized to the activity of the WT μ170 enhancer.

However, the μE3 site was not essential in the μ170 enhancer that contained additional 3′ sequences (Fig. 3B). In this context, mutation of either μE3 or μE2 did not affect enhancer activity significantly, whereas the μE5, μA, or μB site was still essential for function (Fig. 3B). We conclude that μE3 function can be substituted by sequences present between the 3′ endpoints of μ87 and μ170 fragments. Interestingly, this region contains no known μE-like elements, suggesting that μE3 function was compensated for by other factors.

To further delineate sequences that substituted for μE3, we assayed several 3′ deletion mutants. The additional sequences present in μ170 compared to μ87 contain two core sites (C2 and C3; the C1 site is mutated in our enhancer fragment) and a PU.1 binding site between C2 and C3 (24). Taking into consideration these elements, we generated the three deletion mutants μ152, μ128, and μ87 (Fig. 3C). In each context, we assayed the activity of the WT and a μE3-mutated enhancer, with the objective of identifying the fragment where μE3 would no longer be redundant. As shown above, the μ170 enhancer mutated at μE3 (μE3⁻ μ170) was quite active in B cells (Fig. 3C, bars 1 to 3). The μ152 enhancer was less active than μ170; importantly, the μE3 mutation in this context significantly impaired enhancer function (Fig. 3C, bars 4 and 5). Specifically, mutation of μE3 in μ170 resulted in an enhancer that had 80% of WT activity, whereas μE3⁻ μ152 retained only about 20% of WT activity. These observations suggested that the 18 nucleotides missing in μ152 contained an element that was not essential for enhancer activity but which compensated for μE3 in a μE3-mutated enhancer. We also noted that μE3⁻ μ152 was not completely inactive, suggesting that there may be a second μE3-substituting element.

Removal of C3 in μ128 decreased enhancer activity compared to μ170, showing that C3 was a positive contributor. However, μE3 mutation in this context also decreased enhancer activity to about 25% of that seen with the WT μ128 (Fig. 3C, bars 6 and 7). The effect of mutating μE3 was therefore quantitatively similar to that seen with μ152. These results suggested that the 24 nucleotides between μ152 and μ128 contain positive regulatory sequences but no μE3-substituting elements. The next deletion (μ87), which removed an additional 42 nucleotides, was approximately as active as the μ128 enhancer; however, the μE3⁻ mutation decreased activity below that of μE3⁻ μ128. The more deleterious effect of the μE3 mutation in μ87 suggested that C2 contributed to μE3 redundancy, albeit less efficiently than the element between μ170 and μ152. We conclude that two elements contribute to μE3 redundancy. Comparison of the residual activities of μE3⁻ μ128 (or μ152) and μE3⁻ μ87 suggests that C2 provides only about two- to threefold compensatory activity for μE3. For example, μE3⁻ μ128 was 25 to 30% as active as the unmutated enhancer, whereas μE3⁻ μ87 had about 10% of its activity. This approximation was further strengthened by the analysis of C2-mutated enhancers. C2⁻ μ170 was as active as μ170, suggesting that C2 was not a strong positive activator in this context (Fig. 3C, bar 12). However, a C2⁻ μE3⁻ μ170 enhancer was about twofold less active than μE3⁻ μ170, indicating that C2 partially compensated for the loss of μE3 (see Fig. 6A).

FIG. 6.

Effects of μ3′ IRF site mutations on μ enhancer activity. (A and B) μ170 enhancer derivatives, as noted below the bars, were tested after cloning into the Δ56CAT reporter plasmid by transient transfection of S194 plasma cells followed by CAT enzyme analysis. Δ56 represents the enhancerless reporter plasmid, and WT is a reporter carrying an unmutated μ170 enhancer. The mutations tested (μE2⁻, μE3⁻, C2⁻, and μ3′⁻) were in the motifs shown in the schematic at the top; double mutations are indicated with two of these notations. Results shown are averages of three transfections carried out in duplicate. (C) Analysis of IRF site mutations in DHL-9 B-lymphoma cells. Reporter plasmids as indicated below the graph were transiently transfected into DHL-9 cells, which were then subjected to CAT enzyme analysis. Results shown are averages of two transfections carried out in duplicate.

Factors that regulate μE3 redundancy.

The mutational analyses described above showed that two regions of the μ enhancer could functionally substitute for μE3. The stronger element is located at the 3′ end of the μ170 fragment and is removed by the first 3′ deletion mutant, μ152. The residual μE3 redundancy appears to be conferred by the second core homology. A role for C2 was intriguing in light of our recent observation that the human Ig μ enhancer lacks a recognizable μE3 element but contains instead a functional core-like element between the μA and μB motifs (7). In that enhancer, therefore, a core binding protein can substitute for the lack of a μE3-like element. We surmised that C2 may behave similarly, although with reduced efficiency because of its distance from the essential μA-μB combination.

Because both μE3 binding proteins, TFE3 and CBF, enhance Ets-1 DNA binding, we tested the effects of Ets-1 and CBF binding to a μE3-mutated probe that extends from the μA site to the 3′ end of μ128 (Fig. 1A). EtsΔ231, which contains previously identified DNA binding inhibitory domains as well as the DNA binding domain of CBF, formed distinct nucleoprotein complexes on a 96-bp μ enhancer probe (Fig. 4, lanes 1 to 5). Coincubation of both factors resulted in each of the individual complexes, as well as a supershifted complex representing co-occupancy of both μA and C2 elements (Fig. 4, lanes 6 to 8). We did not detect cooperative binding between the two factors, probably because the sites are located too far apart (approximately 75 bp) on linear DNA. However, it is interesting that about 80 nucleotides are required to make a complete turn around a nucleosome, so that sequences such as μA and C2 will be juxtaposed in nucleosomal DNA, perhaps allowing protein-protein interactions.

FIG. 4.

Cobinding of Ets-1 and CBF to the μ enhancer. EtsΔ231 and the DNA binding domain of CBF (CBF.DBD) were purified from bacteria. In vitro binding assays were carried out with either protein alone or both together, as indicated above the lanes, using a 96-bp DNA probe derived from the enhancer (see Materials and Methods). The probe is mutated at μE3 and C1; thus, the only remaining CBF binding site is C2. Single protein-DNA complexes are indicated with arrows labeled Ets and CBF, and a two-protein–DNA complex is labeled Ets/CBF. Lane 1, no proteins added; lanes 2 to 4, binding reactions with increasing amounts of EtsΔ231; lane 5, CBF alone; lanes 6 to 8, constant amount of CBF as in lane 5 with increasing amounts of EtsΔ231 as in lanes 2 to 4.

An IRF binding site in the μ enhancer.

We noticed that the major μE3-substituting element at the 3′ end of the μ170 fragment contained a strong match to the consensus recognition site of IRFs (Fig. 5A). To test whether this region bound IRF proteins, IRF-1 was produced by in vitro translation and used in EMSA. An oligonucleotide probe spanning the 3′ end of the μ170 fragment generated a unique nucleoprotein complex (Fig. 5B, lane 2) which was efficiently competed away by inclusion of nonradioactive self-competitor DNA (lanes 3 and 4). Three other oligonucleotides containing clustered mutations in the μ enhancer IRF consensus sequences as well as the IRF binding site from the beta interferon (IFN-β) gene were also used in competition assays. Mutations (M1 and M3) that changed residues within the conserved region did not compete for IRF-1 binding (lanes 5, 6, 9, and 10), whereas the flanking mutation M2 and the IFN-β site competed efficiently (lanes 7, 8, 11, and 12). We also evaluated IRF-2 binding to the μ3′ site and found that the pattern was indistinguishable from that of IRF-1 (data not shown). In contrast, a third IRF family member, Pip-1, did not bind to the μ enhancer site (Fig. 5C) but bound well to the λB sequence from the Ig λ light-chain gene enhancer. We conclude that the μ3′ sequence binds a subset of IRF family proteins, and the location of this site coincides closely with the 3′ μE3 substituting element.

FIG. 5.

A binding site for IRF in the μ enhancer. (A) The 18 nucleotides deleted between μ170 and μ152 (line 2, μ3′) contain a sequence homologous to the consensus IRF binding site (top line). μ3′M1, μ3′M2, and μ3′M3 are mutated μ enhancer sequences within and outside the IRF consensus. The altered nucleotides are shown as outlined letters. The IRF binding site from the IFN-β gene is shown on the last line. (B) IRF-1 binding to the μ enhancer. IRF-1 protein was generated by in vitro transcription and translation for use in EMSA. The radioactive probe was the μ3′ oligonucleotide (A). Binding reactions were carried out in the absence of competitor DNA (lane 2) or in the presence of the μ3′ oligonucleotide (lanes 3 and 4), μ3′M1 (lanes 5 and 6), μ3′M2 (lanes 7 and 8), μ3′M3 (lanes 9 and 10), and IFN-β (lanes 11 and 12); 10 and 50 ng of competitor oligonucleotides were used. (C) The DNA binding domain of Pip produced by in vitro translation was used in EMSA with a μ3′ probe (lanes 1 to 5) or a probe containing the λB site (lanes 6 to 10) of the Ig λ light-chain gene enhancer. Lanes 1 and 6, probe alone; lanes 2 and 7, reticulocyte extract alone; lanes 3 to 5 and 8 to 10, increasing amounts of Pip containing in vitro translation reactions. The specific DNA-protein complex generated by Pip is marked with an arrow at the right.

To determine whether the IRF family of proteins contributed to μE3 redundancy in larger enhancer contexts, we assayed the effects of mutating the IRF binding site in the context of μ170. In S194 cells, the IRF site-mutated enhancer (μ3′⁻ μ170) was less active than the WT enhancer (Fig. 6A), which is similar to the observation that μ152, the deletion mutant that removed the IRF binding site, also had reduced activity. Mutation of μE3 in μ3′⁻ μ170 (μ3′⁻ μE3⁻ μ170) virtually abolished enhancer activity, indicating that μE3 was essential when the IRF binding site was missing. To rule out the possibility that mutation of any two sites would seriously impair enhancer activity, we checked two other double mutations. Double mutation of either μE3 plus μE2 or μE3 plus C2 reduced enhancer activity to about 40% of the WT activity (Fig. 6A). These observations showed that in the presence of an intact IRF site, substantial enhancer activity was retained even when several elements were mutated. We conclude that IRF proteins are positive regulators of the μ enhancer. When the IRF site is mutated, the μE3 site is necessary for enhancer function; in the presence of the IRF site, μE3 is not essential for enhancer activity.

Like μE3, the μE2 element is essential in the context of μ74 (five-part) and μ87 (six-part) enhancers but is redundant in the context of μ170 (Fig. 3). Our earlier studies have shown that μE2 and μE3 binding proteins activate transcription synergistically, indicating a close working relationship between these two elements (4). We therefore determined whether μ3′ was necessary to observe μE2 redundancy. In S194 plasma cells, μ3′⁻ μE2⁻ double mutation in μ170 significantly reduced its transcription activation potential compared to either of the single mutations (μE2⁻ μ170 or μ3′⁻ μ170 [Fig. 6B]). The residual activity of μ3′⁻ μE2⁻ μ170 was comparable to that of μ3′⁻ μE3⁻ μ170. To rule out the possibility that any double mutation involving μE2 would produce an inactive enhancer, we also compared the activities of the μ3′⁻ μE2⁻ and μE3⁻ μE2⁻ double mutations; μ3′⁻ μE2⁻ μ170 was significantly less active, suggesting that in the absence of the IRF binding site, μE2 is an essential element.

The studies described above were carried out with S194 plasma cells. To determine whether redundancy in the enhancer is unique to plasma cells, we repeated the transfection studies with DHL-9, a surface Ig-negative B-lymphoma cell line. Activity of a μE3-mutated μ170 enhancer was reduced to 80% of that of the unmutated enhancer, indicating that μE3 was not an essential element in these cells (Fig. 6C, bars 1 to 3). As found for S194 cells, mutation of the IRF site reduced enhancer activity to 50% and the double mutation (μ3′⁻ μE3⁻) was essentially inactive (Fig. 6C, bars 4 and 5). We conclude that the feature of μE3/μ3′ redundancy is common to DHL-9 and S194 cells.

The results presented above show that when μE2 or μE3 is mutated, the loss is compensated for primarily by the μ3′ motif. Our earlier studies with a four-part enhancer containing μE2, μE3, μA, and μB provided evidence for several interactions between bHLH and ETS proteins. In particular, the μA site was necessary for transcriptional synergy between μE2 and μE3 elements to be observed. Therefore, the simplest interpretation for the need for μ3′ when either μE2 or μE3 is mutated is that proteins bound to μ3′ interact with the μA/μB core. It is interesting that ETS-IRF interactions have been previously noted between PU.1 and Pip. However, unlike the composite PU-Pip binding sites in the Ig κ or λ gene enhancer, the μ3′ (IRF) site in the μ enhancer is located approximately 120 bp from μA and 100 bp from μB. Because μE2/μE3 binding proteins have been shown to interact with μA binding proteins, and μ3′ can substitute for either μE2 or μE3, we tested the binding of μ3′ and μA binding proteins. To study ETS and IRF protein binding to the μ enhancer, we carried out EMSAs using a 150-bp probe extending from μA to μ3′. IRF-1 was used for these assays because this protein provided the maximal transactivation in transfection assays (see below). In vitro-translated IRF-1 bound to the probe to generate complex 1 (Fig. 7, lane 6); the μA binding protein Ets-1 was expressed as a GST fusion protein and generated complex 2 (lanes 3 to 5). When both proteins were coincubated with DNA, we detected a new complex (complex 3 [lanes 8 and 9]) which likely represents the ternary Ets-1–IRF-1–DNA complex. Complex 3 was abolished when probes containing a mutated μ3′ element or a mutated μA element were used in binding assays (lanes 10 and 11). However, the presence of IRF-1 did not significantly influence Ets-1 binding under these conditions. (The apparent decrease in Ets-1 binding in lanes 7 to 9 is probably due to less probe being available in the presence of IRF-1.)

FIG. 7.

In vitro binding of Ets-1 and IRF-1 to the μ enhancer. Ets-1 was expressed as a GST fusion protein and purified from bacterial extracts. IRF-1 was obtained by in vitro translation in rabbit reticulocyte extracts. The DNA probe was obtained by PCR amplification and encompasses residues 376 to 519 (for WT) or 367 to 519 (for μ3′⁻ and μA⁻) of the μ enhancer (numbering as specified by Ephrussi et al. [6]); the μA element is located between nucleotides 386 and 396, and the μ3′ element is located between nucleotides 501 and 512. In vitro binding reaction mixtures contained the following: lane 1, no proteins; lane 2, 0.5 μl of reticulocyte extract; lanes 3 to 5, 0.5 μl of reticulocyte extracts plus 50, 100, and 200 ng of GST–Ets-1; lane 6, 0.5 μl of reticulocyte extracts containing in vitro-translated IRF-1; lanes 7 to 9, IRF-1 as in lane 6 plus GST–Ets-1 as in lanes 3 to 5; lanes 10 and 11, GST–Ets-1 plus IRF-1 as in lane 9 with mutated probes μ3′⁻ (lane 10) and μA⁻ (lane 11). Binary nucleoprotein complexes are labeled 1 and 2, and the ternary complex is labeled 3.

To directly test if IRF proteins could activate the μ enhancer, we carried out transfection studies with COS cells. The μ170 enhancer-containing reporter was weakly transactivated by the combination of PU.1 and E47 (Fig. 8, bar 2) or with any individual IRF family protein (bars 3, 6, and 9). However, coexpression of PU.1, E47, and IRF-1 resulted in dose-dependent activation of transcription (bars 4 and 5). Neither IRF-2 nor Pip (IRF-4) transactivated efficiently in this assay (bars 7, 8, 10, and 11); moreover, the weak transcriptional activation observed was not dose dependent. We conclude that IRF-1 can cooperate with other μ enhancer binding proteins to activate this enhancer.

FIG. 8.

Transactivation of the μ enhancer by IRF family members. COS cells were transfected with a μ170 enhancer-containing reporter plasmid together with expression vectors for IRF family members as well as the μ enhancer binding proteins as shown below the graph. Where indicated (+), 1 μg of PU.1 and 0.25 μg of E47 expression vectors were used. IRF expression plasmids were used at two different amounts (shown in micrograms). Results shown are averages of three transfections carried out in duplicate.

DISCUSSION

Our earlier studies did not address two key aspects of transcriptional regulation by the Ig μ heavy-chain gene enhancer. First, μ enhancer fragments that contained up to four protein binding sites needed to be dimerized to significantly activate transcription in B cells. Therefore, it was important to identify (i) the smallest enhancer fragment that activated transcription as a monomer and (ii) the factors that mediated this activity. Second, because our effort so far had focused on identifying enhancers in which every site was necessary for function, the basis for redundancy among enhancer elements had not been systematically addressed. In this report, we examined these questions.

We found that addition of μE5 to the previously studied four-part enhancer generated a B-cell-specific transcriptional enhancer that was active as a monomer. All five elements were necessary for activity, indicating that this five-part enhancer had no redundant elements. The reason for the jump in monomeric enhancer activity between the four- and five-part enhancers is unclear at present; however, it is unlikely to be a consequence simply of increasing the numbers of elements in the enhancer. For example, a dimer of a μE3⁻ four-part enhancer is inactive, despite retaining six functional factor binding sites. Similarly, when the μA and μB sites are moved apart in the context of μ170, enhancer activity is abolished even though all nine elements remain intact and capable of binding proteins. These observations underscore the importance of the organization of sites within the enhancer. In this regard it is interesting that the three μE elements in this enhancer fragment are aligned roughly on the same side of the DNA helix, whereas the alignment of the μA and μB sites with respect to the μE elements is shifted by approximately half a helical turn. Perhaps three appropriately positioned bHLH protein TDs can recruit a requisite coactivator significantly better than two such domains on a four-part enhancer.

The monomeric μ enhancer could be activated in nonlymphoid cells by the coexpression of PU.1, Ets-1, and E47 proteins, and as seen in B cells, all sites were necessary for enhancer function. Furthermore, we defined the domains of PU.1 and Ets-1 that were required to activate this enhancer and found them to be similar to those we had previously shown to be required to activate a dimerized tripartite enhancer. We envisage that the ETS domain of PU.1 participates in the nucleoprotein complex by bending the enhancer DNA and making direct contacts with Ets-1 bound at μA. Ets-1 at the μA site has a very different role. Unlike PU.1, N-terminal domains of Ets-1 (including a previously identified TD) serve at least two functions. First, the Ets-1 TD works together with the TFE3 TD bound at μE3 to accentuate the transcriptional potential of this pair of sites, perhaps by presenting a composite TD to the basal machinery. Second, the non-DNA binding N-terminal region of Ets-1 couples the activation potential of E47 and TFE3 bound to the μE2 and μE3 elements, respectively. In addition to providing a TD, TFE3 protein directly interacts with Ets-1 to stabilize its DNA binding as well as to alter Ets-1 conformation in a way that enhances E47 binding to the μE2 site (4). Thus, each protein has multiple jobs in the nucleoprotein complex that comprises the functional enhancer.

Redundancy among μ enhancer elements was functionally defined by the observation that mutation of certain elements such as μE2 and μE3 did not significantly diminish enhancer activity. The simplest interpretation of these results had been that another μE element functionally substituted for the loss of the mutated element. Unexpectedly, we found that the element that contributed most to making either the μE2 or μE3 element redundant was a novel IRF binding site, μ3′, that is located approximately 100 bp away from μE3. In the presence of the IRF site, a μE2 or μE3 mutation had proportionately less effect on enhancer activity, whereas in the absence of the IRF site, loss of either element crippled the enhancer significantly; that is, μE2 and μE3 are essential when the IRF site is missing. In addition, the second core homology (C2) also contributed to μE3 redundancy, though to a lesser extent. The μ3′ element was also active in DHL-9 B-lymphoma cells, indicating that redundancy was not a feature only of plasma cells such as S194. Finally, we showed that IRF-1, but not IRF-2 or Pip, activated the μ enhancer together with ETS and bHLH proteins in cotransfection assays. These observations suggest that IRF-1 is a likely candidate for being a functional μ enhancer binding protein but do not rule out the possibility that other IRF family members also activate this enhancer.

How do IRF proteins participate in μ enhancer activation? Two hypotheses are proposed below. First, proteins bound to μ3′ may directly interact with proteins bound at the μE2-A-E3-B region. From such a complex, if either a μE2 or μE3 binding protein is missing, its loss would be compensated by the IRF protein. It is interesting that the μE2⁻ μE3⁻ enhancer is approximately as active as the μ3′⁻ enhancer, which suggests that either IRF or the μE2-μE3 combination can activate the μA/μB core to similar levels. Alternatively, it is possible that IRFs interact with protein bound to C2, C3, and the intervening PU.1 binding site that are located within 30 bp of μ3′. Maybe this complex of proteins can activate the μA/μB core just as the μE2-μE3 complex does. In both models, the nonessential 3′ components are visualized as interacting with the essential μA/μB core. The main difference between the two models is that in the first model IRF is envisaged as working by itself, whereas in the second model it works along with other factors.

The importance of the newly identified IRF site was underscored by the observation that this site is conserved between the rodent and human μ enhancers. Furthermore, in their earliest in vivo methylation protection experiments, Ephrussi et al. (6) observed a protection over a guanosine residue that corresponds to the newly identified IRF binding site. These results suggest that IRF proteins interact with the μ enhancer in vivo. These characteristic hallmarks of functional significance of the IRF site raised the question as to why it is so important to ensure redundancy in the enhancer. We suggest that the property of the enhancer measured as a redundancy in transcription factor requirements reflects a more fundamental biologic characteristic, such as the need to modulate Ig expression during B-cell differentiation or activation. For example, there are several stages in the functional life of a B cell where Ig expression is known to be regulated. First, the transition from immature to mature B cells is accompanied by increased surface IgM expression; second, surface Ig expression is decreased in activated B cells present in germinal centers, presumably to select for high-affinity somatic mutants; lastly, IgH transcription is increased a few fold in terminally differentiated plasma cells. One way to achieve such quantitative differences is for the enhancer to contain more than the minimal number of protein binding sites; the activity of such an enhancer can then be up- or down-regulated by changing the nuclear concentration of one or more limiting transcription factors.