HMGA1 co-activates transcription in B cells through indirect association with DNA.

The immunoglobulin heavy chain enhancer, or mu enhancer, is required for B cell development. Only the appropriate combination of transcription factors results in B cell-specific enhancer activation. HMGA1 (formerly (HMG-I(Y)) is a proposed co-activator of the ETS transcription factors required for mu enhancer activity. HMGA1 associates with the ETS factor PU.1, resulting in changes in PU.1 structure, and enhanced transcriptional synergy with Ets-1 on the mu enhancer in nonlymphoid cells. New data show HMGA1 directly interacts with Ets-1 in addition to PU.1. In vitro HMGA1/Ets-1 interaction facilitates Ets-1/mu enhancer binding in the absence of an HMGA1.Ets-1.DNA complex. To address whether HMGA1 is present in the transcriptionally active mu nucleoprotein complex, we completed DNA pull-down assays to detect protein tethering in the context of protein/DNA interaction. Results show that HMGA1 is not tightly associated with mu enhancer DNA through PU.1 or Ets-1, despite strong associations between these proteins in solution. However, chromatin immunoprecipitation assays show HMGA1 associates with the endogenous enhancer in B cells. Furthermore, antisense HMGA1 substantially decreases mu enhancer activity in B cells. Taken together, these data suggest that HMGA1 functions as a transcriptional mu enhancer co-activator in B cells through indirect association with DNA.

B lymphocyte development is critically dependent on the immunoglobulin heavy chain intronic enhancer, or enhancer. The enhancer is required for V-to (D)J rearrangement of the immunoglobulin heavy chain (1,2) and subsequent tissue-specific transcription of the properly rearranged locus (3,4). enhancer transcriptional activity has been extensively analyzed in the context of a three-element minimal enhancer, containing the A, E3, and B sites necessary for activity in B cells (5). Functional analysis of the minimal enhancer has demonstrated that transcriptional activation requires combinatorial activity of transcription factors binding the three elements in a precise stereochemistry (6,7).
Two proteins that are required for synergistically activating the enhancer, PU.1 and Ets-1, are members of the ets family of transcription factors. PU.1 and Ets-1 activate transcription by binding to the B or A sites, respectively. The third minimal enhancer activator is the basic helix-loop-helix leucine zipper protein TFE3 that binds to the E3 site of the enhancer. Arguably, the transcription factor CBF may instead occupy an overlapping site and activate the enhancer in lieu of TFE3 (8).
Of these critical enhancer activators, only PU.1 is restricted to the hematopoetic lineage (9,10). The importance of PU.1 in B cell development is highlighted in PU.1 null mice, which lack B cells (11,12). In the absence of Ets-1 (13,14) or TFE3 (15), B cell development is apparently normal. However, B cell activation and plasma cell development are perturbed in these mice (13)(14)(15), indicating that each of these proteins is key for terminal B lineage differentiation after exit from the bone marrow. It has been previously demonstrated that the transcriptional co-activator HMGA1 (formerly designated HMG-I(Y)) forms a protein-protein interaction with PU.1 in solution (16,17). HMGA1a (formerly HMG-I) and HMGA1b (formerly HMG-Y) are splice variants of a single gene. Both proteins contain peptide sequences designated A-T hooks, which are involved in DNA binding and have multiple surfaces capable of proteinprotein interactions (18). In many cases HMGA1 co-activates transcription by aiding assembly of a large nucleoprotein complex. Perhaps the best-characterized transcriptional nucleoprotein complex, or enhanceosome, is dependent on sequencespecific HMGA1 binding to the inducible interferon ␤ (IFN-␤) 1 enhancer (19,20). HMGA1 binding alters IFN-␤ DNA structure, decreasing the activation energy required for the transcription factors NF-B and ATF/c-jun to bind (19). HMGA1 apparently serves a similar function in many other promoters and enhancers. For example, HMGA1 is necessary for increased binding of c-rel (21) and the AP-1 complex (22) to the interleukin-2 promoter. HMGA1 also facilitates formation of a transcriptional complex at the interleukin-2R␣ promoter/enhancer through both DNA and protein-protein interactions (23,24). In contrast, HMGA1 does not directly bind enhancer DNA (16); therefore, the mechanism of action at the aforementioned promoters/enhancers cannot be applicable to demonstrated co-activation of the enhancer by HMGA1 (16).
Recent work (16) has begun to uncover the novel mechanisms by which HMGA1 co-activates the enhancer. HMGA1 first interacts with the critical enhancer regulator PU.1 in solution (16,17). This interaction results in increased PU.1/ enhancer binding likely through an HMGA1-induced change in PU.1 structure. This increased binding manifests itself in the ability of HMGA1 to potentiate PU.1/Ets-1 functional synergy on the enhancer. Overall, the data suggest that HMGA1 activates the enhancer through an indirect mechanism. Because these assays were completed in non-lymphoid cells, we questioned whether HMGA1 is a co-activator of the enhancer in B cells. The current analysis begins to address the role HMGA1 plays in activating the enhancer in B cells.
To differentiate between HMGA1 playing a role in assembling the enhanceosome versus activating the enhancer through other mechanisms, we have extended our initial analysis to detail how HMGA1 interacts with other minimal enhancer activators, TFE3 and Ets-1. HMGA1 increases TFE3/ E3 binding (25) reminiscent of the effect HMGA1 has on PU.1. New data demonstrate that HMGA1 also forms protein-protein interactions with Ets-1 and facilitates Ets-1/ enhancer binding. Although recombinant HMGA1 does not associate with enhancer DNA through a PU.1/Ets-1 tether in vitro, chromatin immunoprecipitation assays show HMGA1 associates with the enhancer in B cells. Furthermore, antisense HMGA1 substantially decreases enhancer activity in B cells. Our model for HMGA1 functioning through indirect association with the enhancer may explain how HMGA1 functions on a wide variety of promoters and enhancers independent of direct HMGA1/DNA binding.

MATERIALS AND METHODS
Cell Lines and Transfections-The mature plamacytoma Ag8.653 was maintained in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal calf serum and 0.05% penicillin/streptomycin. 2017 pro-T or 38B9 pre-B cells were grown in RPMI supplemented with 10% heat-inactivated fetal calf serum, 0.05% penicillin/ streptomycin, and 10 Ϫ5 M ␤-mercaptoethanol. BAL-17 mature B cells were grown similarly, except using 5% serum. For enhancer activity assays 5-8 ϫ 10 5 cells were transfected with 4 g CAT or luciferase reporter plasmid plus 4 g antisense HMGA1 (21) or control plasmid using Superfect (Qiagen). Cells were harvested 48 h after transfection, and equivalent amounts of whole cell extracts were assayed for the presence of the CAT reporter by enzyme-linked immunosorbent assay (Roche Applied Science). Luciferase activity was measured using luciferase assay reagents from Promega. Note that in functional assays, different activities of HMGA1a and HMGA1b cannot be distinguished because the antisense HMGA1 depletes both proteins. Resulting data is therefore interpreted based on effects of HMGA1.
DNA and Oligonucleotides-The control Moloney murine leukemia virus (Mo-MLV) enhancer reporter and the minimal enhancer reporter construct, (70) 2 CAT, have been previously described (5,26). The dominant negative Ets-1 protein Ets-1 ⌬286 is a 5Ј deletion of Ets-1 to amino acid 286 cloned into the EVRF mammalian expression vector as previously described (27). Ets-1 ⌬286 binds enhancer DNA in an electrophoretic mobility shift assay (EMSA) but does not transactivate the enhancer in combination with PU.1 in non-lymphoid cells. All plasmid DNAs were verified for identity by restriction endonuclease digestion and, in all cases, more than one plasmid preparation was used for analysis. A biotinylated T7 oligonucleotide and a non-biotinylated T3 oligonucleotide (Integrated DNA Technologies) was used to PCR amplify enhancer DNA from a pBluescript vector containing either the wild type or B Ϫ enhancer DNA (5) for biotinylated DNA pulldowns. All amplified DNA was purified by low melt agarose gel electrophoresis before use. For the IFN-␤ enhancer HMGA1 binding site, we annealed the following oligonucleotides containing the adjacent PRDII and NRDI binding sites (underlined) (20,28): 5Ј-TTGGGAAAT-TCCTCTGAATAGAGAGA-3Ј biotinylated top strand, 5Ј-GATCCTCTC-TCTATTCAGAGGAATTTCCC-3Ј non-biotinylated bottom strand.
Recombinant Proteins and Protein Analysis-The His-HMGA1a (formerly HMG-I) construct was provided by Dimitris Thanos; His-Ets-1 and His-PU.1 have been previously described (6,29). All recombinant proteins were prepared by standard methods (6,16). Proteins were purified on a HisTrap nickel column or a GSTrap (Pharmacia), using a BioLogic HR Work station (BioRad). Eluted fractions were analyzed by Western blot and DNA binding assays to confirm protein identity. High expressing fractions were pooled and dialyzed against buffer D (20 mM Hepes, pH 7.9, 100 mM KCl, 0.2 mM EDTA, 0.5 mM dithiothreitol, 20% glycerol), aliquoted, flash frozen, and stored at Ϫ80°C. In all cases, multiple protein preparations were analyzed in duplicate experiments.
GST pull-down experiments using recombinant proteins have been previously described (16). Alternatively, GST-HMGA1a was used to test association with PU.1 and Ets-1 from 200 -1000 g pre-cleared Ag8 or 38B9 nuclear extracts in 1000 l total volume using an identical protocol. Eluted proteins were separated on 10 -12% SDS-polyacrylamide gels, transferred to polyvinylidene difluoride membranes, and probed with a rabbit ␣-Ets-1 or goat ␣-PU.1 antibody (Santa Cruz Biotechnology). Rabbit antibodies raised against the His 6 tag or PU.1 were likewise purchased. All primary antibodies were used at a 1:200 -1:1000 dilution for Western blot analyses. Horseradish peroxidase-conjugated donkey ␣-rabbit IgG (Amersham Biosciences) or donkey ␣-goat IgG (Jackson Labs) were used as secondary antibodies at a dilution of 1:3000. Immune complexes were visualized using the Renaissance Reagent kit (PerkinElmer Life Science) in all cases.
Immunoprecipitation of PU.1 from 38B9 pre-B cell nuclear extracts was done from 1000 g protein using 20 l rabbit-␣PU.1 antibody (Santa Cruz) and standard protocols. Precipitated PU.1 was detected on Western blots using goat ␣-PU.1 antibody as detailed above.
Biotinylated DNA Pull-downs-Approximately 7 ϫ 10 6 streptavidin conjugated Dynabeads (Dynal) were washed with PBS-BSA (PBS, pH 7.4, 0.1% BSA) for each sample. Biotinylated enhancer DNA (500 -1000 ng) was incubated with the streptavidin beads for Ͼ4 h at 4°C with rotation. Dynabead⅐DNA complexes were extensively washed with PBS-BSA to remove unbound DNA. Beads were resuspended in pulldown buffer (PBS, pH 7.4, 0.1% BSA, 1 g/ml poly(dI⅐dC)⅐(dI⅐dC) before addition of 10 g BSA as a nonspecific protein competitor. We then added recombinant His-PU.1 (200 -300 ng) alone, recombinant His-Ets-1 (200 -300 ng) alone, or recombinant His-HMGA1a (200 -300 ng) alone or in pair-wise or three-way combinations. Samples were incubated for 4 h at 4°C with rotation. Dynabead⅐DNA⅐protein complexes were spun down and washed three times with ice cold PBS-BSA. Samples were transferred to fresh microfuge tubes prior to final wash to avoid eluting plastic bound proteins. Dynabead⅐DNA⅐protein complexes were eluted in SDS-reducing sample buffer by heating at 95°C. Quadruplicate samples were pooled and equal volumes loaded onto either 10% (for analysis of PU.1 and Ets-1) or 16% (for analysis of HMGA1) polyacrylamide gels for SDS-PAGE. Samples were transferred to polyvinylidene difluoride membranes and subjected to Western blot analysis as described above.
Chromatin Immunoprecipitations-For each immunoprecipitation 1 ϫ 10 7 2017 or 38B9 cells were fixed by adding formaldehyde to a final concentration of 1% directly to growth media at 37°C for 10 min. The reaction was quenched with glycine at a final concentration of 0.125 M. Cells were washed once in ice-cold PBS containing protease inhibitor mini-mixture (Roche) and 1 mM phenylmethylsulfonyl fluoride. Nuclei were pelleted at low speed, lysed in SDS lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl, pH 8) for 10 min on ice and diluted to 1 ml in dilution buffer (0.01%SDS, 1.1%Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-HCl, pH 8, 16.7 mM NaCl, 1 mM phenylmethylsulfonyl fluoride, protease inhibitor mini-mixture). Chromatin was sonicated to ϳ500-bp fragments and centrifuged at 13,000 rpm for 10 min at 4°C to remove debris. A portion (5%) of each sample was set aside to measure input DNA, and the remainder was diluted to 2 ml in dilution buffer and split into two 1 ml aliquots. Nonspecific background was pre-cleared with 30 l of salmon sperm DNA/protein A agarose beads (Upstate) and 1 l of normal mouse IgG (Upstate) for 1 h at 4°C with rotation. Supernatants were incubated with either 5 l of ␣-acetylated histone H3 (Upstate), ␣-HMGA1 (MR19), or normal mouse IgG overnight at 4°C with rotation before immune complexes were collected with 30 l salmon sperm DNA/protein A agarose for 1 h at 4°C. Beads were washed once each with low salt wash buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris, pH 8, 150 mM NaCl), high salt wash buffer (same as low salt but with 500 mM NaCl), and LiCl wash buffer (0.25 M LiCl, 1% Nonidet P-40, 1% sodium deoxycholate, 1 mM EDTA, 10 mM Tris, pH 8.0) then twice in TE (10 mM Tris 1 mM EDTA) (pH 8.0). Pellets were resuspended in 150 l of chromatin immunoprecipitation (ChIP) assay elution buffer (1% SDS, 0.1 M NaHCO 3 ) and rotated at room temperature for 15 min. Samples were centrifuged and eluates removed. Elution was repeated one time and eluates were combined. Input DNA (5% of total sample previously set aside) were diluted to 300 l in elution buffer, and cross-links were reversed in all samples with the addition of NaCl to a concentration of 0.3 M and 20 g of RNase A for 5-6 h at 65°C. Proteins were removed from samples with 10 mM EDTA, 53 mM Tris-HCl, pH 6.5, and 50 g proteinase K overnight at 37°C. Samples were extracted once with phenol, once with 1:1 phenol/chloroform, and once with chloroform and precipitated in EtOH with 30 l of 5 M NaCl and 20 g glycogen at Ϫ80°C for one hour. DNA was quantified using Picogreen quantification reagent (Molecular Probes). 500 pg of each sample was then analyzed with quantitative real-time PCR.
Real-Time PCR and Analysis-Duplicates or triplicates of each sample were analyzed in a quantitative PCR reaction using the ABI Prism 7000 sequence detector and QPCR Mastermix Plus SYBR Green kit (VWR Scientific Products). Data was analyzed with a threshold set in the linear range of amplification, which for most experiments was 0.05. The cycle number that any particular sample crossed that threshold (Ct) was then used to determine fold difference (enrichment). Fold difference was calculated as 2 (Ct(input)ϪCt(ChIP)) (31). Melt curves of each amplified sample indicated formation of a single product in all cases.

HMGA1 Associates with ETS Proteins in Solution-HMGA1
has been shown to interact with two enhancer activators in vitro, PU.1 and TFE3 (16,17,25). Because activation of the enhancer also requires Ets-1 binding (5), we questioned whether HMGA1 interacts with this third protein of the minimal enhanceosome. Using a GST pull-down assay, we tested whether a GST-tagged HMGA1a forms a stable protein-protein interaction with Ets-1 in solution. Ets-1 interacts with HMGA1a as shown in Fig. 1A (lanes 3 and 5, representing independent duplicate samples). The precipitated Ets-1 co-migrated with recombinant Ets-1 loaded directly onto the gel as a control (lane 9). In the presence of the GST tag alone, Ets-1 did not precipitate (lanes 4 and 6), indicating that Ets-1 interacts with HMGA1a rather than the GST tag. The data clearly demonstrate a stable protein-protein interaction between HMGA1a and Ets-1 in vitro.
Although recombinant HMGA1a interacts with recombinant Ets-1 or PU.1 in a GST pull-down assay, this interaction must occur in the context of a complex mix of nuclear proteins to be physiologically relevant. Fig. 1B demonstrates that GSTtagged HMGA1a forms a stable interaction with PU.1 (lane 3) in the context of 38B9 pre-B cell nuclear extracts. Recombinant PU.1 (lane 1) loaded directly onto the gel co-migrated with the HMGA1a-associated PU.1, suggesting identity of the immunoreactive bands. The GST tag alone does not interact with the PU.1 naturally present in 38B9 nuclear extracts (lane 2). PU.1 identity was further confirmed by co-migration of the HMGA1a-associated protein with PU.1 directly immunoprecipitated from 38B9 nuclear extracts using an ␣-PU.1 antibody (lane 4). In contrast, ␣-GST did not immunoprecipitate PU.1 from 38B9 nuclear extracts (lane 5).
GST pull-down analyses of HMGA1a-interacting proteins in Ag8 plasmacytoma nuclear extracts demonstrated that a likely Ets-1 proteolytic fragment bound HMGA1a but not GST. The putative nuclear Ets-1 fragment associated with HMGA1a comigrated with an HMGA1a-associated recombinant Ets-1 fragment (data not shown). Full-length Ets-1 could not be recovered from the HMGA1a-associated protein pool, likely due to the relatively high protease level in Ag8 cells in combination with the relative lability of Ets-1 to proteases. Taken together these data demonstrate PU.1 and Ets-1 form protein-protein interactions with HMGA1 in both simple and more physiological contexts.
HMGA1 Increases Ets-1/ Enhancer Binding-HMGA1a facilitates PU.1/B binding through protein-protein interaction (16). Our demonstration that HMGA1a interacts with Ets-1 in solution raises the possibility that HMGA1a augments Ets-1/ enhancer binding in a similar fashion. Ets-1 specifically binds to the A site of the enhancer, and mutation of this site destroys Ets-1/DNA interaction and Ets-1-mediated enhancer activation (5). EMSA analysis with full-length Ets-1 is difficult to interpret as the inhibitory domains cause weak binding to the A site under EMSA conditions (32,33). To simplify analysis of how the demonstrated Ets-1/HMGA1a interaction affects Ets-1 function, we used a truncated form of Ets-1, ETS(Ets-1), containing only the ETS DNA binding domain. At relatively high amounts of protein (64 ng), ETS(Ets-1) complexed with enhancer DNA (Fig. 2A, lane 2). At relatively low levels of ETS(Ets-1) (6 ng), no detectable complex was formed with the enhancer (lane 3). To determine whether HMGA1a enhances ETS(Ets-1)⅐ enhancer complex formation, we added HMGA1a protein to the EMSA reaction. HMGA1a increased ETS(Ets-1)/ enhancer binding at low levels of Ets-1 (compare lanes 4 -6 to lane 3). Very low levels of HMGA1a had no effect on complex formation (lanes 7 and 8). In the presence of equivalent amounts of BSA (instead of HMGA1a), the ETS(Ets-1)⅐ enhancer complex was undetectable using 6 ng ETS(Ets-1) (data not shown). To further demonstrate the specificity of the enhanced binding complex, we repeated the analysis using a A Ϫ mutated probe. Lanes 11-15 demonstrate that the A binding site was necessary for ETS(Ets-1)⅐DNA complex formation in the presence of HMGA1a. Similar results were obtained for full-length Ets-1 protein despite the generally weak enhancer binding activity as previously published (data not shown and Refs. 26 and 29). Overall, the data suggest that enhancer co-activation by HMGA1 stems from its ability to increase DNA binding by Ets-1 ( Fig. 2A), TFE3 (25), and PU.1 (16).
Taken together, our GST pull-down and EMSA assays suggest HMGA1 directly interacts with the ETS domain of Ets-1 to facilitate DNA binding. To determine whether HMGA1 function in this context is more generally applicable to other ETS domains, we extended our binding analysis to the DNA binding ETS domain of PU.1, ETS(PU.1). In the presence of BSA, low levels of ETS(PU.1) do not form a complex with enhancer DNA (Fig. 2B, lanes 1-3). In contrast, HMGA1a facilitates ETS(PU.1)/ enhancer interaction (lanes 4 -5). This complex co-migrates with the ETS(PU.1)⅐ enhancer complex formed using relatively high levels of ETS(PU.1) (data not shown). These data demonstrate that HMGA1a facilitates nucleoprotein complex formation between the enhancer and the DNA binding domains of both PU.1 and Ets-1. Interestingly, the protein⅐DNA complex migrates identically in the presence or absence of HMGA1a in all cases (for example compare Fig. 2A,  lane 2 with lanes 4 -6).
HMGA1 Is Not Tightly Associated with an in Vitro Enhancer Nucleoprotein Complex-Because the ETS(Ets-1) DNA complex induced by the addition of HMGA1a co-migrated with the nucleoprotein complex formed by high amounts of ETS(Ets-1), the data suggest that HMGA1a is absent from the ETS(Ets-1)⅐DNA complex. Similarly, the EMSA mobility of the HMGA1a-induced PU.1/ enhancer complexes are not altered when compared with samples using high protein levels that bind the enhancer in the absence of HMGA1a (16,25). However, for HMGA1a to supershift a protein⅐DNA complex, HMGA1a would have to remain in the EMSA complex for the duration of the gel run (Ͼ2 h). The absence of a trimolecular complex in EMSA is therefore not a reliable indication of the presence or absence of HMGA1a in the ETS protein⅐DNA complex. To more rigorously challenge the apparent absence of HMGA1a in the PU.1⅐Ets-1⅐ enhancer complex in a more physiological setting, we used biotinylated DNA pull-downs. In this analysis, enhancer DNA will precipitate associated proteins forming either protein-DNA or protein-protein interactions. Because the final washes in this protocol are completed within 5-10 min, loosely bound proteins are much more likely to be retained in the nucleoprotein complex as compared with an EMSA complex. To test the validity of this strategy, we precipitated PU.1 and Ets-1 with either wild type or B Ϫ DNA. The B Ϫ mutation is a 3 base pair mutation that efficiently destroys both PU.1/B binding and PU.1-mediated functional synergy with Ets-1 (5). Because PU.1 and Ets-1 interact in a yeast 2-hybrid assay (34), we predicted Ets-1 could tether PU.1 to a B Ϫ enhancer if our pull-down assay was a valid measure of protein-protein interactions on DNA. Biotinylated wild type or B Ϫ enhancer DNA was conjugated to streptavidin beads and subsequently incubated with PU.1 and Ets-1, alone or in combination. Bound proteins were detected by Western blot. Fig. 3A shows PU.1 is precipitated by wild type biotinylated DNA alone or in the presence of Ets-1 as expected (lanes 1 and  3). B Ϫ DNA does not bind PU.1 (lane 7) consistent with previous results (5). However, in the presence of Ets-1, B Ϫ DNA precipitates PU.1 likely through protein-protein interactions between PU.1 and Ets-1 (lane 5). By independently verifying PU.1/Ets-1 interaction using biotinylated DNA pulldowns, we conclude that this method detects DNA-bound protein complexes and protein-protein tethers that have been independently validated.
To determine whether HMGA1a is stably associated with the ETS protein⅐DNA nucleoprotein complex, we used the pulldown protocol to precipitate proteins stably associated with the nucleoprotein complex. Recombinant proteins alone or in combination were incubated with biotinylated wild type enhancer DNA as indicated at the top of One trivial explanation for the apparent lack of an ets protein/HMGA1a tether in the biotinylated pull-down assay is that the assay conditions preclude HMGA1 binding. To test this possibility, we took advantage of the fact that HMGA1 interacts with both DNA and proteins through an A-T hook motif, suggesting that the basic requirements for HMGA1 interacting with both species are likely to be similar. To test the compatibility of the pull-down assay conditions with general HMGA1 A-T hook function, we took advantage of the demonstration that HMGA1 binds to the IFN-␤ enhancer via direct protein/DNA contact in both EMSA and in vitro footprint assays (20). We biotinylated an HMGA1a-binding fragment of the IFN-␤ enhancer and demonstrated that this fragment could precipitate HMGA1a under conditions identical to conditions in Fig. 3B. Although HMGA1a is not precipitated in the presence of -bound PU.1 and Ets-1, the protein can be precipitated by IFN-␤ DNA (Fig. 3C, lane 1) and co-migrates with recombinant HMGA1a loaded directly onto the gel (lane 2). PU.1 was not precipitated by IFN-␤ DNA and was included as a negative control (data not shown). Based upon these analyses, we conclude that, although HMGA1a facilitates the binding of PU.1 and Ets-1 to the enhancer, HMGA1a may not remain stably in the DNA-bound transcriptionally active minimal enhancer (A⅐E3⅐B) complex.
HMGA1 Is Associated with Enhancer in B Cells-Although protein-protein interactions measured in EMSA and DNA pulldown assays may accurately reflect similar interactions in the cell, these assays cannot recapitulate the more complicated protein⅐DNA complex formed at the cellular enhancer. The lack of detectable interactions between ETS domain proteins and the enhancer in the pull-down assay suggests that a PU.1/Ets-1 combination cannot tether HMGA1 to DNA. However, additional transcription factors such as TFE3, CBF, IRF-1, and E47 activate the enhancer by various measures (8,29,35,36) and could participate in retaining HMGA1 in the cellular enhanceosome. To test whether HMGA1 is associated with the enhancer in B cells, we performed ChIPs using ␣-HMGA1 antibody to precipitate chromatin complexes from To confirm that HMGA1 is associated with only active loci, we asked whether the 38B9 enhancer is packaged into hyperacetylated histones. Previous analyses have demonstrated that active loci are packaged by hyperacetylated histones,  1-4) or B Ϫ (lanes [5][6][7][8] enhancer DNA was bound to streptavidin beads and then incubated with BSA (500 ng, lanes 4 and 8), PU.1 (200 ng, lanes 3 and 7), Ets-1 (200 ng, lanes 2 and 6), or PU.1 plus Ets-1 (lanes 1 and 5). Nucleoprotein complexes were separated on 8% reducing SDS-PAGE gels. Shown is Western detection of DNA-associated proteins using an ␣-PU. whereas inactive loci are associated with hypoacetylated histones (37). We completed ChIP assays using an antibody specific for acetylated histone H3 in 38B9 pre-B and 2017 pro-T cells, which have active or inactive loci, respectively. Fig. 4C (leftmost bars) demonstrates that, as expected, acetylated H3 associated with the enhancer in 38B9 pre-B cells 3.6-fold more than in 2017 pro-T cells. This level of increase is consistent with analyses on primary bone marrow B cells (38). The ␤-globin locus associates with hypoacetylated histone H3 in both cell types (middle bars), and the nonspecific IgG control did not precipitate DNA (rightmost bars). The data confirm HMGA1 associates only with an active enhancer in B cells and not with an inactive T cell enhancer, strongly supporting the contention that HMGA1 co-activates the enhancer in vivo.
HMGA1 Determines Enhancer Activity in B Cells-The ChIP analyses (Fig. 4), in combination with the demonstration that HMGA1 synergizes with an Ets-1/PU.1 combination to activate the enhancer in non-B cells (16), are consistent with the interpretation that HMGA1 plays a role in tissue-specific enhancer activation in B cells. Transient transfection assays in multiple B cell lines reproducibly demonstrated HMGA1 overexpression increased enhancer activity only 50 -90% (data not shown), to a level difficult to interpret using this method. This low level increase is consistent with our previous demonstration that HMGA1/ETS protein stoichiometry is critical for activating the enhancer (16). We speculate that B cells contain the appropriate concentrations of factors for optimal enhancer activation, such that overexpression of any given factor will not hyperactivate the endogenous enhancer substantially.
To more definitively test whether HMGA1 functions in combination with endogenous B cell proteins to activate the enhancer, we completed transient transfection assays with an antisense HMGA1 construct, which decreases HMGA1 protein levels to ϳ5% of normal lymphoid cell levels (21). For this analysis, antisense HMGA1 and the enhancer reporter (70) 2 CAT were transiently co-transfected into the plasmacytoma line Ag8. Alternatively, we transfected in the antisense parent vector RcCMV in combination with (70) 2 CAT. As shown in Fig. 5A, endogenous B cell proteins, likely PU.1 and Ets-1, drive CAT reporter protein expression in the presence of the parent RcCMV vector by activating the enhancer (left bar). Co-transfection of antisense HMGA1 decreases enhancer activity ϳ71% (right bar) as compared with activity in the presence of RcCMV. This data demonstrates that HMGA1 expression is necessary for robust activation of a relatively accessible enhancer by endogenous B cell proteins.
One possible explanation for decreased enhancer activity in the presence of antisense HMGA1 is that antisense HMGA1 decreases transcription nonspecifically. To test the specificity of the antisense construct, we measured the effect of antisense HMGA1 on a CAT expression vector driven by the Mo-MLV enhancer. This enhancer contains binding sites for several transcription factors including Ets-2, Fli-1, GA-binding protein, and core binding factor, and is active in B cells (26,39). Co-transfection of either the parent RcCMV vector or the antisense HMGA1 resulted in approximately equivalent activation of the Mo-MLV enhancer (Fig. 5B). This data suggests that the antisense HMGA1 is not interfering with transcription in general, but is specifically inhibiting enhancer activity.
To get a sense of the significance of a Ͼ70% decrease in enhancer activity, we compared the effect of antisense HMGA1 to the effect of an Ets-1 truncation mutation on enhancer activation in B cells. For this analysis we used N-terminally truncated Ets-1 ⌬286, which is incapable of activating the enhancer in combination with PU.1 in non-lymphoid cells, yet retains wild type DNA binding properties (27). We transfected Ag8 plasmacytoma cells with the (70) 2 CAT reporter in the presence or absence of Ets-1 ⌬286 (Fig. 5C). Ets-1 ⌬286 decreased enhancer activity 88.2% (left bar) as compared with co-transfection of an empty expression vector (right bar). In contrast, Ets-1 ⌬286 had no significant effect on activity of an unrelated SV40 promoter-driven luciferase gene lacking an Ets-1 binding site in Ag8 cells (Fig. 5D). These data indicate that Ets-1 ⌬286 neutralizes enhancer activation mediated by endogenous B cell proteins likely through competitive binding, and hence is a dominant negative enhancer regulatory protein. Furthermore, antisense HMGA1 is approximately as effective as Ets-1 ⌬286 in decreasing enhancer activity. Taken together, the functional data is consistent with ChIP data in suggesting that endogenous HMGA1 is important for enhancer activity in B cells in the presence of normal transcription factor levels. DISCUSSION Although the identities of DNA binding proteins and transcriptional co-activators regulating the enhancer have been the subject of intense investigation for over a decade, the detailed mechanism of enhancer activation remains unclear. The data strongly suggest HMGA1 plays an important role in activating the enhancer in B cells through modulating function of the ets family transcription factors Ets-1 and PU.1. Three findings bolster this conclusion. First, multiple biochemical assays show functional importance of an HMGA1/ETS protein interaction in vitro with respect to DNA binding activity. Second, we have directly demonstrated that HMGA1 specifically associates with the enhancer in B cells but not in pro-T cells. Third, decreasing HMGA1 levels substantially decreases enhancer activity in B cells but does not affect transcription in general. We hypothesize that in B cells, which express relatively low levels of PU.1 (7,40), PU.1 cannot efficiently bind and activate the enhancer in the absence of HMGA1 or an analogous co-activating factor. Extension of this hypothesis to include Ets-1 is likely upon more rigorous characterization of interactions described in our work.
A second level of interpretation addresses the biochemical mechanism of HMGA1 activity on the enhancer: HMGA1 interacts with the enhancer indirectly, likely through weak protein-protein interactions. Although ChIP assays show HMGA1 associating with the enhancer in the context of cellular chromatin, multiple assays, including EMSA, DNase I footprinting (16), methylation interference, 2 and DNA pulldowns (this study) show no specific enhancer/HMGA1 binding under a variety of experimental conditions. We interpret this apparent conundrum as follows. Although the sole interpretation of the ChIP data is that HMGA1 is preferentially associated with the enhancer through either protein-protein or protein-DNA interactions in B cells, there are multiple alternative explanations for our inability to detect HMGA1/ETS protein association in the context of ETS protein/DNA binding in vitro. First, HMGA1 may associate with the enhancer through an ets protein-independent mechanism, because multiple transcription factors occupy the enhancer in in vivo footprint analyses (30). Preliminary pull-down assays using a TFE3/ enhancer combination show TFE3 does not tether HMGA1 to the enhancer. One possibility is that CBF, a transcription factor that can activate the enhancer via a site overlapping the TFE3 site (8), tethers HMGA1 to the enhancer. This possibility is currently under investigation. Because HMGA1 functions on an enhancer fragment containing only the two ets sites and the TFE3/CBF site (16), it is unlikely that interactions outside this core enhancer are necessary to explain the data. A second possibility explaining the apparent contradictory data is that the combination of three proteins (PU.1, Ets-1, and TFE3) forms an HMGA1-associating platform. We deem this possibility unlikely due to the small size of HMGA1 which could limit its availability for interaction with multiple proteins as further discussed below. A third scenario is that protein post-translational modification absent from the recombinant proteins used in the various binding assays is required 2 B. S. Nikolajczyk, unpublished observation. for HMGA1 association with the enhancer. HMGA1 can be modified by phosphorylation, ubiquitination, or acetylation in vivo. Biological significance of these modifications has been demonstrated in multiple analyses (18,(41)(42)(43). Such modifications could increase affinity of an ETS protein⅐HMGA1 complex in cells, countering the apparent absence of a stable association in vitro. Alternatively, phosphorylation sites within PU.1 have a proposed role in regulating multiple immunologically important enhancers (44,45). A role for HMGA1 has not been tested in these contexts. One additional possibility is that formaldehyde cross-linking in the ChIP assays locks HMGA1 onto the DNA, to avoid potential dissociation during the course of the assay. However, we maintain that HMGA1 does not bind enhancer DNA directly, as described above, despite extensive evidence that HMGA1 associates with bona fide target elements in the IFN-␤ enhancer (Ref . 20 and this study, Fig. 3C). Our preliminary analysis of in vitro acetylated or phosphorylated recombinant HMGA1 show these modifications do not lead to enhancer binding activity (data not shown), further discounting mechanisms based on direct HMGA1/DNA interaction. Further analyses will refine our current interpretation, that HMGA1 associates with DNA-bound ets proteins to indirectly associate with enhancer DNA in B cells and thereby regulate enhancer activity.
We previously demonstrated that the stoichiometry between ETS proteins and HMGA1 determined levels of enhancer activity (16). Ectopic HMGA1 expression modestly increases enhancer activity by endogenous proteins in B cells (50 -90% increase, data not shown). Because optimal stoichiometry between limited concentrations of enhancer activators and HMGA1 is required for transcriptional co-activation (16), the inability to artificially optimize this stoichiometry in a B cell is not surprising. The new antisense HMGA1 results strongly suggest the HMGA1/ETS protein stoichiometry is naturally optimal for co-activation in B cells. Now that we know HMGA1 can increase protein binding at three sites in the minimal enhancer, we envision several possible models of the in vivo stoichiometry between HMGA1 and enhancer activators. First, one HMGA1 molecule may interact with all three members of the minimal -activating complex simultaneously. A single HMGA1 molecule can interact with multiple proteins because HMGA1 has several demonstrated protein-protein binding surfaces (20,46). However, we consider this possibility unlikely as the ultimate positioning of PU.1, Ets-1, and TFE3 on the enhancer shows PU.1 and Ets-1 on one face of the DNA helix while TFE3 binds 120°away around the circumference of the helix (6). HMGA1a and b, at 10 and 11 kDa, are likely too small to interact with 3 proteins to simultaneously escort them to the DNA in the stated configuration. Second, HMGA1 may be associated with and affect each activator of the minimal enhancer individually. One piece of data perhaps inconsistent with this possibility is our demonstration that antisense HMGA1 has no effect on the Mo-MLV enhancer, a transcriptional element that requires an Ets-1/2 site for activity. If a bimolecular interaction between Ets-1 (or the Ͼ90% identical Ets-2) (47, 48) and HMGA1 augmented Mo-MLV enhancer activity, antisense HMGA1 would affect reporter expression in our control experiment. Because Ets-1 may function through unrelated mechanisms on the Mo-MLV versus the enhancer, further experimentation is required to test the 1:1 HMGA1/ETS protein model. Finally, one HMGA1 molecule may interact with both ETS family members, and a second HMGA1 molecule may interact with TFE3, a possibility that is currently untested. Interactions between a single HMGA1 molecule and multiple transcription factors are consistent with a role for HMGA1 in coordinating transcription factor binding to the enhancer much like a protein chaperone.
It remains to be seen whether, in the context of chromatin, the association of HMGA1 with transcription factors facilitates chromatin remodeling, in part, through alteration of ETS protein structure and concomitant recruitment/release of chromatin structural complexes. HMGA1 itself may determine chromatin structure upon directly binding target sequences (23). However, the demonstrated lack of direct HMGA1/ enhancer interaction precludes a direct role for HMGA1 in determining chromatin structure. One partner of HMGA1, PU.1, is a demonstrated chromatin accessibility factor for the locus, although its ability to increase accessibility of a closed chromatin structure is weak (49). Knowing that HMGA1 is associated with the enhancer in B cells but not T cells is important for determining how these various molecules cooperate in the context of chromatin, either re-enforcing or neutralizing activities attributed to single proteins. Taken together, our findings that HMGA1 functions on the enhancer in B cells spurs our interest in understanding precisely how this transcriptional co-activator regulates an important tissue-specific protein in vivo.