Induction of an embryonic globin gene promoter by short-chain fatty acids

The 100-kb human β-globin gene cluster, typical of the vertebrate β-globin gene family, is organized in the same order as it is expressed temporally, 5′-embryonic ϵ-fetal ^Gγ-^Aγ-adult δ-β-3′. Abnormalities of or deficiencies in adult β-globin gene expression that cause human disease, like β-thalassemia or sickle cell (SS), are fully manifest only after completion of the normal fetal-to-adult globin switch during infancy. Patients with these diseases can have a milder clinical syndrome if their β-globin gene locus has an additional or complementary genetic abnormality, in which aberrantly high levels of fetal globin gene are expressed through adulthood, a so-called hereditary persistence of fetal hemoglobin [HPFH] syndrome.^1-3  These observations predict a parallel therapeutic effect from a pharmacologic up-regulation of fetal β-type globin gene expression in the majority of β-globin gene disorder patients who do not have an associated genetic HPFH syndrome.

Butyrate, propionate, and related short-chain fatty acids (SCFAs) increase expression from embryonic/fetal β-type globin genes during adult erythropoiesis in a range of models.⁴  Clinical trials of butyrate-like compounds in patients with sickle cell disease and β-thalassemia^5-9  have shown mixed results. Of interest, persistence of fetal globin gene expression from a normal human β-globin gene locus has been reported in association with abnormal elevations in SCFAs, as in infants of diabetic mothers and in children with inherited abnormalities of amino acid metabolism,^10-12  such as propionic acidemia.

Here, we focused on propionate since, when endogenously elevated in humans, alone among SCFAs, propionate sustains an up-regulation of fetal globin gene expression from a normal β-globin gene locus through at least late childhood.¹²  We analyzed promoter function from the murine embryonic globin gene ϵ^y during adult erythroid differentiation in murine erythroleukemia (MEL) cells. This cell line allows the study of the molecular mechanisms underlying SCFA-mediated inducibility of fetal/embryonic globin genes in an adult erythroid milieu. The 65-kb murine β-globin gene cluster is composed of 2 embryonic β-type genes, ϵ^y and β^H1, and 2 adult β-globin type genes.^13,14  The ϵ^y gene is not expressed during development beyond day 14 (d14) after coitus, nor is it typically expressed in Friend viruse-transformed MEL cells at basal state or following induction of adult erythroid differentiation by dimethyl sulfoxide (DMSO).^12-14  MEL cells have particular strength as a model for the SCFA-mediated reversed switch, since they up-regulate embryonic globin gene expression during differentiation to an adult erythroid phenotype only following induction by SCFAs.^12,15  In contrast, human K562 cells, the other commonly used model cell line, differentiate to an embryonic/fetal phenotype, which complicates assessments of differentiation versus induction. We attempt to minimize the limitations of MEL cells as a model of normal globin gene expression^16,17  by focusing on the function of a transfected murine embryonic globin gene promoter in stable MEL cell pools.

Additional globin regulatory sequences, relevant to this study and conserved between mouse and human, include the locus control region (LCR) and promoter CACCC sites. The murine LCR is a complex and remote 30-kb enhancer region, composed of multiple DNase I hypersensitivity sites (HSs) at more than 7 kb upstream of ϵ^y. The range of functions subsumed by the LCR remains controversial,^18,19  but it is clear that context within the β-globin locus or the proximity of genes to LCR elements is critical to experimental fidelity.²⁰ 

Human fetal (γ) and murine embryonic (ϵ^y) β-type promoters have identical distal core CACCC sites, at -145 and -114, respectively, relative to the transcription start site, that are similar to the functionally essential -90 adult β-type promoter proximal core CACCC site. The adult β-CACCC site binds the Krüppel-like transcription factor erythroid Krüppel-like factor (EKLF),²¹  which is essential for that promoter's function. EKLF-related transcription factors, such as basic KLF (BKLF) and embryonic/fetal 3-like globin gene-activating KLF (FKLF) 1 and 2,^22-24  play as-yet-undefined roles during normal development.

In this study, we asked what sites and cognate binding factors in an embryonic/fetal globin gene promoter could confer responsiveness to SCFAs. Our data lead us to conclude that one molecular mechanism through which SCFAs may act is by modification of “adult” and constitutive transcription factors, such as EKLF and specificity proteins 1 and 3 (SP1, SP3), in a way that allows transactivation of an embryonic/fetal globin gene promoter during adult erythroid differentiation.

Nested 5′ deletions of ϵ^y promoter subclones were derived by polymerase chain reaction (PCR) from a wild-type (wt) ϵ^y globin gene genomic template (Figure 1A), by means of the following KpnI-linked primers: gagcacaggaagccata (-1337), aatgtttctcaaggatattg (-960), gagaataagcaaaacaaaga (-239), aacacagtgtacagtt (-158), and acgggtcaggctgacca (-100). The downstream antisense KpnI-linked primer gatggcaagtctgggaggtt, common to all constructs, originated at the translation start site (ATG, +52). All numbering is relative to the transcription start site at +1. The ϵ^y promoter fragments were cloned 5′ to 3′ into Pgem 3zf- (Promega, Madison, WI), upstream of a promoterless 2300-bp neomycin resistance gene, neo. Some constructs included a 1200-bp EcoRI cassette, HS2, from the second hypersensitive site of the murine LCR,²⁵  cloned 5′ of ϵ^y promoter-neo. A mutated CACCC oligomer, CACCCmut (Figure 1B), in which DNA-protein interactions at the CACCC site are lost in gel-shift experiments (Figure 3B), was incorporated into wt ϵ^y promoter constructs by standard PCR-based mutagenesis, yielding HS2-158 ϵ^ymut neo and HS2 -1337 ϵ^ymut neo.

The minimal functional ϵ^y promoter, from -158 to +52 bp (ATG), with a wt or mutated CACCC site, was generated with the use of ϵ^y primers with XhoI and HindIII linkers, as described, and cloned into PGL3 Basic (Promega), with or without HS2. The resultant ϵ^y-luciferase clones, Pwt, HS2Pwt, and HS2Pmut, are shown in Figure 1A.

DNA-protein interactions at -100 to -158 bp in the ϵ^y promoter were investigated in gel-shift experiments with the use of radiolabeled 21-bp or 33-bp double-stranded oligomer probes that contained a wt or mutated CACCC site (Figure 1B). The 22-bp and 33-bp probes from the adult murine β-globin β-CACCC are also shown.

Plasmid cytomegalovirus β-HA-300 (pCMVβ-HA-p300), plasmid SG5/CREB-binding protein (pSG5/CBP), and plasmid CX/FLAG/p300/CBP-associated factor (pCX/FLAG/PCAF) expression vectors are as described,²⁶  as are the expression constructs for glutatione-S-transferase (GST)-EKLF,²⁷ wt EKLF, EKLF K288R, EKLF K302R, EKLF K288R/K302R,^28,29  BKLF,²²  CMV-SP1, CMV-SP3,³⁰  and CMV-secreted alkaline phosphatase (CMV-SEAP).³¹  Where noted, sequencing was per Sequenase protocol (USB, Cleveland, OH).

MEL cells were maintained, transfected, stably selected, and induced, as described previously,³²  except that propionate was used at 5 mM.

NIH/3T3 cells, grown in high-glucose Dulbecco minimum essential medium (DMEM)/10% fetal calf serum and penicillin/streptomycin, were cultured into 6-well plates and transfected, per protocol, with PolyFect (Qiagen, Valencia, CA). Endotoxin-free DNA preparations, at 0.25 μg ϵ^y promoter-luciferase reporter constructs (with or without HS2) and, as indicated, 0.5 μg transcription factor expression constructs, 0.5 μg histone/factor-acetyl transferase (H/FAT) expression constructs, and neutral plasmid to equalize DNA concentration, were transfected. Propionate, at a final concentration of 5 mM, was added to induced samples at transfection.

In vitro assays for coimmunoprecipitation were performed by transfection, as above, of 4 μg transcription factor or 3 μg transcription factor and 1 μg p300 onto 100-mm plates. Cells were harvested at 48 hours and were lysed in RIPA buffer, per protocol (Upstate Biotechnology, Lake Placid, NY). Extracts were precleared and incubated with 2 μg transcription factor-specific antibodies (Santa Cruz Biotechnology, Santa Cruz, CA) overnight. Incubation with Protein G PLUS-Agarose beads (Santa Cruz Biotechnology, Santa Cruz, CA) was followed by serial RIPA washes. Protein samples were resolved by 12% sodium dodecyl sulfate/polyacrylamide gel electrophoresis (SDS/PAGE), transferred onto nitrocellulose, probed with 1:1000 anti-acetyl-lysine antibody (Santa Cruz Biotechnology), and detected by chemiluminescence from a horseradish peroxidase-conjugated secondary antibody. Membranes were then stripped and reprobed with antibodies specific for the transfected transcription factor.

Neomycin mRNA expression from each ϵ^yneo construct in pools of induced MEL cells was analyzed by hybridization to in vitro radiolabeled RNA probes for neomycin, endogenous ϵ^y, and endogenous triose phosphoisomerase (TPI) mRNA, followed by RNase protection, as previously described.^12,32  Each sample was normalized to TPI content. Inducibility of ϵ^y promoter constructs was calculated by contrasting corrected neomycin mRNA levels between the DMSO- or SCFA-induced cells and the uninduced cells (set equal to 1).

Luciferase activity in 25 μL NIH/3T3 cell lysate, harvested per protocol (Enhanced Luciferase Assay kit; BD Pharmingen, San Diego, CA) at 36 to 48 hours after transfection, was measured in a monolight 2010 Luminometer (Analytical Luminescence Laboratory, San Diego, CA). HS2Pwt was analyzed in each experiment. To correct for variation in scale between experiments, all expression was normalized to basal HS2Pwt expression (Figure 5A), at 0.245 × 10⁷. Expression levels in the tabular material (Figures 5A,7A) are stated relative to the basal luciferase expression from HS2Pwt alone (set equal to 1) in those experiments.

In a subset of experiments, transfection efficiency was assessed by measuring secreted alkaline phosphatase levels from media of cells that had been cotransfected with CMV-SEAP.³¹  No difference in transfection efficiency was noted between SCFA-treated and untreated cells (not shown); however, CMV-SEAP was itself up-regulated by H/FAT coexpression, as was a second potential control plasmid tk-pHRL (Promega), as has been described by others.³³  Thereafter, all experiments were standardized to protein loading (DC protein assay kit; Bio-Rad, Hercules, CA).

Oligomers of interest were end-labeled with [P³²] γ-adenosine triphosphate ([P³²]γ-ATP) and polynucleotide kinase (Gibco BRL, Gaithersburg, MD), annealed to a 3-fold excess of unlabeled complementary oligomer, purified across a Sephadex G-50 column (Promega), and confirmed as having a specific activity greater than 10⁸ cpm/μg. Then, 10 fmol radiolabeled probe was incubated for 30 minutes in binding buffer (25 mM HEPES [N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid], pH 7.5; 16 mM KCl; 50 mM NaCl; 2 mM ZnCl₂; 8% glycerol; 0.6 mM β-mercaptoethanol; and proteinase inhibitors), 1 μL poly-deoxy-inosinic-deoxy-cytidylic acid (poly-dIdC; Roche, Indianapolis, IN), and 10 μg MEL cell nuclear extracts, induced or uninduced, or 20 ng GST-EKLF. Where indicated, 1 μL antibodies raised against the transcription factors SP1, SP3, EKLF (the latter a cross-reacting antibody raised against gut-enriched KLF [GKLF]) (Santa Cruz Biotechnology) or BKLF (Merlin Crossley) were preincubated with nuclear extracts for 1 hour at room temperature, after which radiolabeled probes were added, as described. There is 90% amino acid identity between GKLF and EKLF at the epitope against which the anti-GKLF antibody was raised, and this antibody also detects EKLF (Santa Cruz Biotechnology). Unlabeled specific competitor was added during incubation, where indicated, at 100- to 200-fold molar excess relative to radiolabeled probe. Reactions were run on 4% or 6% nondenaturing PAGE gels × 1 to 2 hours, at 4°C.

Nuclear extracts were isolated from MEL cells that had been cultured for 6 days with or without DMSO or SCFAs, as described previously.³⁴ 

GST-EKLF and the backbone plasmid pGEX TK (Amersham, Arlington Heights, IL), transfected into BL-21 cells (Stratagene, La Jolla, CA), were grown to log phase, induced for 2 hours with 1 mM isopropyl-beta-d-thiogalactopyranoside (IPTG), and lysed by 30 seconds of sonication. After incubation with GST beads and protease inhibitors for 20 minutes at 4°C, the proteins were eluted in 10 mM reduced glutathione/50 mM Tris (tris(hydroxymethyl) aminomethane), pH 8, at room temperature for 10 minutes, quantitated at optical density (OD) 280, and stored at -80°C in 10% glycerol. Unlabeled specific probe as competitor was added during incubation, where indicated, at 30- to 50-fold.

We first examined what sequences in a stably transfected ϵ^y promoter were required for its up-regulation in response to SCFAs. Figure 1A illustrates the constructs tested. Expression from ϵ^y promoter-neo constructs that contained at least 158 bp ϵ^y promoter alone was only minimally up-regulated in MEL cells following induction of adult erythroid differentiation by SCFAs (Figure 2). However, the inclusion of an HS2 cassette increased SCFA-mediated up-regulation of these constructs, relative to uninduced cells, 4- to 16-fold.

The shortest promoter construct tested, HS2 -100 ϵ^yneo, showed no detectable neo mRNA expression as a stable transfect, regardless of induction; sequencing confirmed intact basal promoter motifs (not shown). HS2 -100 ϵ^yneo transfects, therefore, expressed neo mRNA at a level sufficient to confer resistance to G418, but insufficient for detection by RNase protection assay.

These results suggest that the 58 bp between -100 and -158 bp 5′ of the cap site are essential to ϵ^y promoter function in uninduced MEL cells. This 58-bp sequence contains a number of potential regulatory sites, including the ϵ^y-CACCC site at -114 bp, whose 9 bp core is identical to that found in the human γ-globin gene promoters.

We tested the function of the core ϵ^y-CACCC site by creating a disruptive 4-bp CACCC mutation, CACCCmut (Figure 1A), within the ϵ^y promoter. The CACCCmut constructs, in spite of the presence of HS2, showed a diminution in, but not an abrogation of, SCFA-mediated induction of expression in MEL cells (-1337 and -158 promoters in “HS2 promoter” and “HS2 promoter CACCC-mut”; Figure 2). Indeed, the approximately 4-fold SCFA-mediated induction of HS2-1337 ϵ^ymut neo in MEL cells suggests that additional sequences between -158 and -1337 are capable of mediating modest induction by SCFAs in the absence of an intact CACCC site, whereas the lack of SCFA-mediated induction by the enhancerless -1337 ϵ^yneo suggests that the enhancer HS2 makes a major contribution to inducibility.

Expression from HS2ϵ^y promoter-neo was up-regulated, albeit much less so, during DMSO-mediated induction of adult erythroid differentiation in MEL cells, although endogenous ϵ^y globin gene expression was not up-regulated under these conditions, as described previously.¹²  We have evidence that the sequences necessary for full silencing during DMSO-mediated induction of the endogenous locus are absent from the constructs tested here (not shown).

The embryonic CACCC site was identified as functionally important to the SCFA-mediated induction of the ϵ^y promoter during adult erythroid differentiation in MEL cells, perhaps as a site through which the LCR can act. We tested for binding of candidate transcription factors in vitro.

Gel-shift experiments in which the radiolabeled 22-bp ϵ^y-CACCCwt probe was incubated with uninduced or induced MEL cell nuclear extracts showed 3 prominent gel-shift bands (Figure 3A; as labeled). Preincubation with SP1 or SP3 antibodies disrupted bands 1 and 2 or band 3, respectively. All gel-shift bands were lost when the CACCC site was mutated (Figure 3B). We conclude that SP1 and SP3 can form complexes with the ϵ^y-CACCC box in vitro.

Longer electrophoresis runs revealed additional bands from incubations of the 22-bp ϵ^y-CACCCwt probe and SCFA-induced MEL cell extracts (Figure 4A, bands 4 and 5). Gel-shift reactions with antibodies against BKLF or EKLF (Santa Cruz Biotechnology)²²  confirmed the presence of the former and suggested the presence of the latter. Comparison of gel-shift experiments with the adult β-CACCC oligomer as probe, under these same conditions, showed binding and disruption by the EKLF antibody, similar to those seen with the embryonic ϵ^y-CACCCwt oligomer (Figure 4A).

To resolve whether EKLF could interact at the ϵ^y-CACCC site in vitro, we tested the binding of purified murine GST-EKLF to the 33-bp adult β-CACCC and ϵ^y-CACCCwt oligomers (Figure 4B).^27,29  The resultant band was disrupted by 30- to 50-fold unlabeled competitor and did not form when the 33-bp ϵ^y-CACCCmut oligomer was used as probe.

The 4 factors EKLF, BKLF, SP1, and SP3 that bind the ϵ^y-CACCC site in vitro were tested for their ability to transactivate the HS2 ϵ^y promoter, with or without SCFAs, and with or without p300. Three luciferase constructs were tested (Figure 1B): Pwt, HS2Pwt, and HS2Pmut, in which the CACCC site is mutated. EKLF and SP1 were also cotransfected with PCAF or CBP, which have been reported to augment Krüppel-factor transactivation in vitro^26,28,35  and have been implicated in normal HS2 function.³⁶  EKLF was also cotransfected with constructs that lacked the HS2, Pwt. These transient transfections were performed in nonerythroid NIH/3T3 cells, to avoid confounding effects from endogenous globin transcription factors.

SP1, SP3, and EKLF, but not BKLF, show transactivation of the 158-bp ϵ^y promoter with enhancer, HS2Pwt, in NIH/3T3 cells. In the absence of SCFAs, only EKLF, without H/FATs, increases expression from HS2Pwt, relative to HS2Pwt alone, at 4 times baseline; coexpression of p300, without SCFAs, further augments expression, at 23, 7, and 6 times baseline, for EKLF, SP1, and SP3 respectively (Figure 5A; black bars and tabular material). Expression from HS2Pwt with transcription factors was augmented by H/FATs and further by SCFAs (Figure 5A; black and striped bars), with induction by SCFAs of the already elevated expression from HS2Pwt with transcription factors and coactivators alone, at 1.8- to 4-fold with EKLF, at 3.3- to 5.3-fold with SP1, and at 2.9-fold with SP3 (fold induction; Figure 5A). In the evaluation of expression relative to HS2Pwt alone, however, a marked augmentation in expression by SCFAs, of 42, 23 and 16 times baseline from, respectively, EKLF and p300, SP1 and CBP, and SP3 and p300, is seen (“Expression, relative to HS2Pwt,” Figure 5A). The degree of SCFA-mediated induction of HS2Pwt with transcription factors and H/FATs is variable, however (compare HS2Pwt with EKLF, with and without p300 and with and without SCFAs; Figures 5A,7A, and not shown). The activity of each H/FAT is also variable, with PCAF and p300 more active with EKLF, and CBP and p300 more active with SP1 (Figure 5A).

Expression from the ϵ^yluciferase constructs HS2Pwt and Pwt, with and without EKLF, suggests a role for HS2 in the SCFA-mediated induction of expression in vitro (Figure 5B; right with HS2, versus left without HS2). Without an enhancer and in the absence of SCFAs, Pwt shows low basal levels of expression, regardless of cotransfected factors (Figure 5B; left panel, black bars). In contrast, expression from HS2Pwt in the absence of SCFAs (Figure 5B; right panel, black bars) is affected by cotransfection with EKLF and p300, at up to 23-fold. Induction of expression by SCFAs from enhancerless Pwt was increased by EKLF, however, and was further increased in the presence of coactivators; maximal levels of luciferase expression from Pwt, at 2 × 10⁷ luciferase units (Figure 5B; left panel, hatched bars), are modest. HS2Pwt expression is up-regulated 32-fold, relative to uninduced HS2Pwt alone by EKLF and p300 in SCFA-containing media (Figure 5B; right panel, hatched bars), with maximal expression levels, at 10 × 10⁷ luciferase units, that are 5 times greater than those seen from Pwt.

The role of an intact CACCC site in inducibility was analyzed by coexpressing transcription factors with an HS2Pmut construct. High-level SCFA inducibility of this mutated CACCC construct by EKLF, SP1, SP3, or BKLF is lost, with or without p300 (Figure 5C).

SCFAs have been experimentally found to alter acetate and acetyl Coenzyme A (acetyl-CoA) flux in metabolically active cells.³⁷  Acetate in vivo can modestly induce fetal hemoglobin gene expression in treated baboons³⁸  and in vitro can acetylate EKLF in tissue culture.²⁶  We speculated that some of the effects noted in transient transfections could reflect stimulation of endogenous or cotransfected coactivators by SCFAs, with resultant acetylation of transcription factors. We next tested whether the acetylation status of any of the in vitro-identified active transcription factors was altered following exposure to SCFAs.

Unlike acetate, H³-labeled propionate was not commercially available, so direct measurement of propionate-derived H³-labeled acetylation of transcription factors was not feasible. Therefore, an immunoprecipitation strategy was undertaken in which transcription factors were transfected into NIH/3T3 cells with or without p300 and SCFAs. Each factor was “pulled down” by its appropriate antibody and then hybridized to a pan-acetyl antibody in a standard Western blot assay. SP1 showed increased acetylation associated with H/FAT coexpression and SCFAs (Figure 6). Immunoglobulin interference and weak antibody interactions precluded analysis of EKLF and SP3 by this method.

We also investigated the function of acetylation-deficient EKLF mutants^26,28  in the SCFA-mediated induction of expression from an embryonic globin gene promoter. EKLF expression constructs, in which functionally important conserved lysine residues, K288 in the transactivation domain and/or K302 in the second zinc finger of the DNA-binding domain, had been mutated to preclude acetylation,^26,28  were tested for their ability to transactivate the ϵ^yluciferase construct HS2Pwt in NIH/3T3 cells.

EKLFwt or EKLF mutants induce the same basal level of expression from HS2Pwt in uninduced media, at 2- to 4-fold (Figure 7; black bars and “basal” row in tabular material). However, HS2Pwt with EKLF acetylation-deficient mutants showed a smaller increase than was seen with EKLFwt, in the presence of coexpressed coactivators PCAF or p300 in uninduced media, at maximum levels of 13 versus 9 and 15 versus 11, respectively (Figure 7; compare numeric data from first with second to fourth black bars in p300 and PCAF cohorts). Overall expression and induction were less in the presence of CBP.

SCFA-mediated transactivation (SCFA/basal), calculated between members of each paired cohort under each cotransfection condition, with EKLFwt alone, and with PCAF or p300, at 7-, 2.7-, and 1.8-fold respectively, was mildly diminished in EKLF mutants; especially EKLF K288R, which alone, and with PCAF or p300, showed induction at 4-, 1.8-, and 1.6-fold respectively (Figure 7; “Fold induction by SCFAs,” striped and black bars). SCFA-mediated induction of expression from HS2Pwt showed that EKLF mutants K288R and K288R/K302R were unable to induce maximal expression as was seen from EKLFwt: without H/FATS, at 16 to 20 from EKLFmut versus 24 from EKLFwt, with PCAF, at 12 to 19 from EKLFmut versus 34 from EKLFwt, and, with p300, at 12 to 13 from EKLFmut versus 28 from EKLFwt (Figure 7; “Expression, relative to HS2Pwt”).

The detailed molecular mechanisms underlying SCFA responsiveness by embryonic/fetal globin genes during adult erythropoiesis are incompletely understood. Multiple putative butyrate and SCFA response elements (SCFA-REs) in embryonic and fetal globin genes have been described by a number of investigators, across a range of tissue-culture and animal models. Promoter sequences have been implicated,^32,39-42  as have cell-signaling pathways.^43,44 

The proximal γ-globin gene, +36 to -201, which includes the CACCC site, and a more distal sequence, at -822 to -893, were SCFA responsive in transgenic mice and MEL cells.^39,42  A consistent finding has been a diminution of basal or SCFA-induced expression when the fetal CACCC site is altered. While the CACCC site in the adult β-globin gene promoter is essential to that gene's expression, functional studies of the embryonic/fetal CACCC sites and its interactions with the LCR have been less conclusive. Experimental observations cumulatively suggest a functionally important but partially redundant role for the CACCC site of an isolated fetal/embryonic globin gene promoter. LCR sequences,⁴⁵  such as an HS2 E-box to which the butyrate-inducible factor inhibitor of DNA binding 2 (Id2) can bind in human embryonic/fetal K562 cell lines,⁴⁶  have been described by others as important to the SCFA-mediated up-regulation of embryonic/fetal globin genes. Human HS2 is more than 70% homologous with the core murine HS2 cassette used in our experiments, which includes the nuclear factor-erythroid 2 (NF-E2) and E-box-like binding motifs.²⁵ 

Using a rigorous adult erythroid model, we found that the murine embryonic globin gene promoter with enhancer is up-regulated, at 4- to 16-fold, during SCFA-mediated induction of adult erythroid differentiation in MEL cells (Figure 2). Deletional and mutational analyses of the stably transfected HS2 ϵ^y promoter in MEL cells suggested that SCFA-mediated induction of expression is lost without HS2 and is diminished without an intact CACCC site (Figure 2; contrast -1337 ϵ^y promoter, with and without HS2 and with and without an intact CACCC site). Transgenic mice have been described in which a 5′ deletion of the human γ^A globin gene beyond the γ-CACCC site abrogates SCFA-mediated induction during adult erythropoiesis,³⁹  and in which deletions of fetal CACCC sites, rather than mutations, are more deleterious.^39,47  This is consistent with our finding that expression from HS2 -100 ϵ^yneo in MEL cells in which the CACCC site is deleted is undetectable while expression from HS2 -158 ϵ^ymut neo, with a mutated CACCC site, is markedly diminished but not absent (Figure 2).

The proximal γ-globin gene promoter of patients who had been treated with butyrate-like compounds in vivo⁴¹  show novel footprints at 4 sites that have also been implicated in the molecular pathogenesis of HPFH syndromes.⁴⁸  These sites did not include the CACCC site. Investigations of the SCFA inducibility of the γ-globin gene in K562 cells have implicated the proximal promoter duplicated CAAT motif.^40,45,49  It is possible that, analogous to the acetylation of EKLF and SP1 that we describe, SCFA-mediated modification of other well-defined transcription factors may allow up-regulation of target genes at a number of promoter and enhancer sites, which could explain the lack of a well-characterized unique SCFA-response element in the literature to date.^{32,39,41,42,50,51}  Activity of SCFA-related factor modifications at additional non-CACCC promoter sites could likewise explain the persistent, albeit reduced, SCFA-mediated inducibility seen by us in the long promoter construct HS2-1337ϵ^ymut neo, in which the CACCC site is disrupted.

In our model, an enhancer and an intact promoter CACCC site are critical to maximal SCFA-mediated up-regulation of a stably transfected murine embryonic globin gene promoter during adult erythroid differentiation in MEL cells. The CACCC-binding factors SP1, SP3, EKLF, and BKLF are present in MEL cell extracts, can bind the ϵ^y CACCC site in vitro, and, except for BKLF, will augment transactivation of the murine embryonic globin gene promoter with enhancer in the presence of SCFAs in NIH/3T3 cells (Figure 5A). Binding by GST-EKLF to the ϵ^y-CACCC site probe in vitro was tested (Figure 4B), owing to weak activity of EKLF in MEL cell extracts (Figure 4A; Crossley et al²² ), even with an adult β-CACCC probe. The nonerythroid NIH/3T3 cells lack erythroid cofactors, the absence of which could explain the lower range of SCFA responsiveness, at 2- to 9-fold, seen from HS2Pwt in these cells relative to MEL cells. Similarly, the SCFA inducibility of enhancerless Pwt with EKLF and p300, at 22-fold (Figure 5B), in NIH/3T3 cells, far greater than the negligible inducibility seen from promoter-only constructs in MEL cells (Figure 2, left panel), may reflect activity of nonerythroid transcription factors at the ϵ^y promoter.

The loss of SCFA-mediated induction of the CACCC-mutant construct HS2Pmut, with EKLF and p300, suggests that EKLF binds primarily to the CACCC site rather than the HS2 (Figure 5C), although HS2 function is critically important, as Pwt lower overall inducibility suggests. Of note, HS2Pwt, with EKLF or p300, was unresponsive to SCFAs in NIH/3T3 cells when coexpressed with the viral oncoprotein E1A (not shown), which blocks coactivator-mediated interactions with globin locus-enhancer regions.³⁶ 

Of factors that can bind to, and transactivate through, the embryonic/fetal CACCC site in this model in vitro, EKLF has been reported to transactivate the human embryonic globin gene ϵ.⁵²  Knockout of the EKLF gene in murine models showed homozygous lethality in utero with an absence of adult β-globin gene expression,^53-55  but an associated modest up-regulation in embryonic and transgenic fetal globin gene expression^55,56  suggests a repressive function for EKLF at embryonic/fetal globin gene promoters in vivo. SP1 and SP3 are ubiquitous and essential. Transgenic mice in whom SP1 was knocked out developed poorly and died midgestation,⁵⁷  while SP-3 knockout mice died as newborns from lung disease⁵⁸  but have recently been shown to have defects in definitive erythropoiesis as well.⁵⁹ 

The SCFAs propionate and butyrate both increase cytoplasmic pools of acetate and acetyl-CoA.³⁷  We speculated that acetylation of transcription factors could result from the changes acetate and acetyl-CoA flux that have been described in metabolically active cells following exposure to SCFAs,³⁷  and could play a role in the transcription activation seen in vitro and in vivo from SCFAs. Histone modifications, primarily acetylation by the H/FAT coactivators CBP/p300 and PCAF and deacetylation by the HDAC proteins, are recognized powerful regulatory mechanisms, activating or repressing eukaryotic gene expression respectively.^60,61  Transcription factor modification by H/FAT coactivators also plays an important role in a range of hematopoetic pathways.⁶²  Acetylation of critical erythroid transcription factors, such as EKLF and FKLF 2, by these coactivators has been described.^26,28,35 

The functional effects of coactivators and transcription factors on luciferase expression from HS2Pwt, with and without SCFAs, was tested in NIH/3T3 cells. We found that SCFAs increased expression from HS2Pwt alone and in cells that had been cotransfected with EKLF, SP1, or SP3. SCFA-mediated induction of expression was often, but not always, increased by H/FATs, which may reflect modification by SCFAs of endogenous transcription factors and endogenous H/FATs in NIH/3T3 cells, thereby minimizing the difference between cohorts with and without coactivators. In our model, p300 and PCAF are strong activators of EKLF, while CBP and p300 are stronger activators of SP1. In earlier studies, CBP and p300 played the major role in EKLF acetylation and transactivation of HS2-adult β-globin gene expression.^26,36  These differences may reflect a change in cell lines between studies or differential developmental and pharmacologic recruitment of specific coactivators.

We wanted to know whether SCFAs could cause net acetylation of the transcription factors that were implicated in our model. Testing this directly is not straightforward in the absence of appropriate commercially available labeled SCFAs. Coimmunoprecipitation, with transcription factor-specific antibodies and an antibody raised against pan-acetylated proteins, showed that SCFA-associated acetylation of SP1 was detectable in vitro. Acetylation of SP1 has not been described previously, but extrapolation from other SP1-like/Krüppel factor family members would suggest that this is an activating modification.^63,64  Technical difficulties precluded investigation of other factors by this method. Constitutive acetylation of SP3 has been described.⁶⁵  However, SP3 function is highly context dependent,^63,64  and its role here is unclear.

Acetylation of EKLF, with resultant transcription activation and recruitment of chromatin-remodeling complexes in vitro, has been well characterized at the adult β-globin gene locus.^26,28  The SCFA-mediated activation of maximal transcription from the 158-bp ϵ^y promoter by EKLFwt in our transient transfection assay was impaired by 17% without H/FATs (24 versus 20) and up to 54% with H/FATs (28 versus 13) in EKLF K288R or K288R/K302R (Figure 7; “Expression, relative to HS2Pwt”). However, SCFA-mediated up-regulation of the ϵ^y promoter by the acetylation-deficient EKLF mutants was only diminished, and the residual induction by SCFAs (Figure 7; hatched bars) could reflect the proportion of induction that is due to acetylation of nonmutated lysine residues or to other, nonacetylation-related, effects. SCFAs, such as butyrate and propionate, have been shown to have an HDAC function in vitro.^66-68  EKLF, SP1, and SP3 have been reported to recruit HDACs,^64,69,70  and HDAC inhibitors can further transactivate EKLF and H/FATs at the adult β-globin gene promoter.²⁸  The induction of fetal/embryonic globin gene expression by HDAC inhibitors⁴⁰  suggests an etiologic role for the release of HDAC-mediated repression in the net up-regulation of embryonic globin genes. SCFA-mediated inhibition of HDAC recruitment to EKLF,⁶⁹  irrespective of lysine status, could explain residual transactivation activity seen in acetylation-deficient EKLF mutants (Figure 7).

We suggest that acetylated transcription factors, such as SP1, EKLF, and perhaps SP3, play a role in the SCFA-mediated augmentation in transactivation from an embryonic globin gene during adult erythropoiesis. We are working to clarify the role played by these factors in vivo; nonetheless, transcription factor acetylation by SCFAs is novel and may be one of the ways in which SCFAs, as a class of compounds, act to up-regulate embryonic globin gene expression during adult erythroid differentiation.

Note added at proofs: Experiments performed since this manuscript was submitted, in which the CBP expression construct was driven by a more powerful promoter (CMV-CBP, kindness of Dr. Jenny Ting), show a modest up-regulation of EKLF-driven expression of the embryonic murine globin gene promoter by CBP, as previously reported at the adult β-globin gene.²⁶  As with p300 in the current manuscript, SCFAs further augment CBP with EKLF-driven expression from a murine embryonic globin gene promoter, at up to 2- to 5-fold.

Generous support from the Masonic Cancer Fund and the Minnesota Medical Foundation, the Division of Hematology, Oncology and Transplantation at the University of Minnesota, Dr Gordon D. Ginder's laboratory (currently at the Medical College of Virginia), and Dr Ann Dean's laboratory (LCDB/NIDDK, NIH) are gratefully acknowledged. We are indebted to Drs James Bieker, Merlin Crossley, Tim Ley, Guntram Suske, and Jennifer Westendorf for the gift of plasmid constructs; Dr Crossley also supplied anti-BKLF antibodies. Drs Ann Dean and Howard Towle critically reviewed the manuscript. Tammy Dutter, Tuy Vien Mai, and Melissa O'Donnell supplied invaluable technical assistance. Gautam Siram helped prepare the manuscript.