Factor binding to the human γ-globin gene distal CCAAT site: candidates for repression of the normal gene or activation of HPFH mutants


Dr Patient Developmental Biology Research Centre, The Randall Institute, Division of Biomedical Sciences, King's College London, 26–29 Drury Lane, London WC2B 5RL.


We have examined factor binding to the distal human γ-globin CCAAT site and three naturally occurring hereditary persistence of fetal haemoglobin (HPFH) mutations of this site. Factor binding was examined using nuclear extracts from the erythroleukaemic cell lines K562 and MEL, and from A4 cells, a non-transformed mouse bone marrow stem cell line, using the electrophoretic mobility shift assay. Under standard binding conditions, in addition to the previously reported binding by a CCAAT factor (CP1) and GATA-1, the wild-type (wt) sequence bound high mobility factors which appeared to be GATA-2 isoforms. However, when the non-specific competitor conditions were varied, the binding profile with K562, but not MEL nuclear extract, was substantially altered. CP1 and GATA-1 were absent, and two new factors were detected, one of which bound preferentially to the Greek and Japanese non-deletion HPFH mutants. However, binding by the GATA-2 isoforms to the wt sequence was maintained with both cell types, as it was using the A4 cell line. With modified binding conditions, in A4 cells the two non-deletion and the Black deletion HPFH mutants each had a different protein binding profile which was lost on erythroid induction of the cells. We discuss the possiblity that the GATA-2 isoforms bound to the wt sequence may function to suppress wt γ gene expression in the bone marrow. Additionally, those factors which bind preferentially either to the deletion or non-deletion HPFH mutants may play positive roles in establishing an active chromatin structure.

In humans, switching from fetal (HbF) to the adult type (HbA) haemoglobin occurs predominantly at around the time of birth and involves a change from the fetal liver transcription of the Gγ- and Aγ-globin genes to transcription of the adult β-globin gene in the bone marrow. The amount of HbF changes from being the predominant species (60–80%) to about 3% after 16–20 weeks of postnatal development, and throughout adult life the amount of HbF remains at about 1% (reviewed by Stamatoyannopoulos & Nienhuis, 1987).

Hereditary persistence of fetal haemoglobin (HPFH) is a benign condition characterized by the production in adults of >1% HbF, in the absence of haemopoietic stress. Molecular analysis of several HPFH types has led to the classification of this condition into two groups, the deletion and non-deletion types. As shown in Fig 1, the Greek type of non-deletion HPFH, for example, is characterized by continued high-level expression after birth of the Aγ-globin gene and is associated with a single G to A mutation two nucleotides upstream of the distal CCAAT element at nucleotide −117, one of two CCAAT elements in the Aγ-globin gene promoter ( Collins et al, 1985 ; Gelinas et al, 1985 ). The Japanese non-deletion HPFH results from a C to T transition, at nucleotide −114, within the distal CCAAT element of the Gγ-globin gene ( Fucharoen et al, 1990 ).

Figure 1.

4 are indicated; the sequence deleted in the Black HPFH is also shown.

A comparison of factor binding to the wt distal CCAAT box and to the mutants has been the subject of several reports (e.g. Superti-Furga et al, 1988 ; Gumucio et al, 1988 ; Fucharoen et al, 1990 ; Berry et al, 1992 ; Ronchi et al, 1995 , 1996; Katsube & Fukumaki, 1995). For these studies the in vitro binding of factors in nuclear extracts of either K562 or MEL cells was investigated mainly by means of the electrophoretic mobility shift assay (EMSA). The human K562 cell line is a bipotential cell line that expresses erythroid markers and can be further differentiated in vitro along the erythroid pathway, with expression of embryonic and fetal globins (see Rowley et al, 1981 ), or towards the myelomonocytic lineage ( Lozzio & Lozzio, 1975). MEL cells are erythroblasts from mice infected with the Friend virus and express adult globins on induction ( Marks & Rifkind, 1978; Nudel et al, 1977 ). Using extracts from these cell types the only consensus finding of the EMSA studies to date has been that a CCAAT factor (CP1) binds to this site. However, although some workers have reported that the −117 mutation slightly strengthened binding of CP1 ( Superti-Furga et al, 1988 ; Gumucio et al, 1988 ), others have not found this ( Berry et al, 1992 ; Katsube & Fukumaki, 1995). GATA-1 has also been reported to bind to this site ( Superti-Furga et al, 1988 ; Fucharoen et al, 1990 ; Berry et al, 1992 ; Ronchi et al, 1995 , 1996) and the observation that GATA-1 binding was considerably weaker to the −117 mutant led to the suggestion that, in this context, GATA-1 may have a repressor function during development ( Berry et al, 1992 ). Factors for which there has been no consensus findings include CDP, a vertebrate homologue of the sea urchin CCAAT displacement protein ( Superti-Furga et al, 1988 ) and NFE6, which was found to bind to the wt sequence but not the −117 mutant ( Berry et al, 1992 ). Recently it was reported that CP1, GATA-1 and NFE6 from MEL cells bind to the normal distal CCAAT box, whereas CP1 and NFE3 bind with K562 extract ( Ronchi et al, 1996 ). In this study it was reported that NFE3 bound with higher affinity to the wt than to the −117 mutant. A comparison of binding to the normal Gγ-gene distal CCAAT element and the −114 mutant of the Gγ-gene revealed that CP1 binding was abolished by the mutation ( Fucharoen et al, 1990 ). It has also been reported that a 13 nucleotide deletion of the distal CCAAT element, known as the Black HPFH, abolished binding of CP1, GATA-1 and NFE3 ( Gilman et al, 1988 ).

Transgenic mice have been utilized as a model system to study the control of the fetal to adult globin switch (e.g. Berry et al, 1992 ; Katsube et al, 1993 ; Peterson et al, 1995 ). Although mice have only embyronic and adult globins ( Whitelaw et al, 1990 ), correct developmentally regulated expression of the β-like globins has been reported in this system, including high-level transcription of the −117 and −114 non-deletion HPFH genes in adult animals ( Berry et al, 1992 ; Peterson et al, 1995 ; Katsube et al, 1993 ). In a recent study ( Ronchi et al, 1996 ), where both the −117 and the −114 phenotypes were reproduced, it was shown that the −117 HPFH phenotype was dependent on the presence of a functional proximal CCAAT box. It was also reported that inactivating both the distal and proximal CCAAT sites had little effect on the level of expression of the γ-gene early in development. Furthermore mutations of the distal CCAAT box, which specifically abolish binding of either NFE3 or GATA-1, did not result in an HPFH phenotype. These authors concluded that there were substantial differences in the proteins complexed to give the active structure of the normal γ-promoter at the fetal stage in comparison to those bound to the active HPFH promoter in adult cells. From the above it follows that the studies to date have not revealed the factors which account for the altered expression of the HPFH mutants.

We considered the possibility that the relevant factor(s) may not have been detected previously by EMSA because their binding was prevented by the non-specific competitor poly(dI-dC).poly(dI-dC), as it has been reported that this competitor can occlude factor binding ( Kristie & Roizman, 1986; SivaRaman & Thimmappaya, 1987). We showed by EMSA using either modified binding conditions (where factor binding was allowed to occur prior to the addition of the non-specific competitor), or with genomic DNA as a competitor, that two previously undetected factors bound to the wt, −117, −114 HPFHs, one of which was probably erythroid specific and showed increased affinity for the non-deletion HPFHs. The other did not show enhanced affinity for the mutants, and furthermore appeared not to be erythroid specific. The results obtained under standard binding conditions were similar to those reported previously by Berry et al (1992 ). Furthermore, under normal and modified conditions of binding, three high-mobility proteins (band D), which we showed were likely to be truncated forms of GATA-2, bound preferentially to the wt sequence. The Black deletion sequence was found to bind only one factor specifically, and this was distinct from those bound to the −117 and −114 HPFH mutants, indicating that there was no common factor present in K562 or MEL cells that binds to the deletion and non-deletion HPFH types.

K562 and MEL cells have been used extensively to study the HPFH syndromes, nevertheless the most appropriate source to establish whether there are differences in factor binding between the wt and mutant HPFH γ-gene probes is more probably progenitor haemopoietic cells where the activation of the globin genes occurs. For this purpose we used the FDCP-mix A4 (A4) cells, which are a multipotent, non-immortalized, non-transformed haemopoietic stem cell line derived from adult mouse bone marrow ( Spooncer et al, 1984 ) which can be induced to erythroid differentiation ( Heyworth et al, 1995 ). It has been postulated that prior to lineage specification these cells exist in a state of multilocus activation ( Hu et al, 1997 ) which includes partial activation of the erythroid programme (hypersensitive site formation in the β-globin Locus Control Region) ( Jimenez et al, 1992 ). We found that under normal binding conditions only CP1-like factors bound. However, using the modified binding conditions, each of the γ probes was observed to bind a distinct set of factors, with stronger binding to the HPFH mutants. The exception being the high-mobility GATA-2 isoforms which still bound more tightly to the wt sequence. On erythroid induction most of the factor binding was lost, but not the high mobility GATA-2 isoforms bound to the wt sequence, nor the factor bound to the Black deletion HPFH. The data favours a role for the GATA-2 isoforms in suppressing the expression of the γ-gene in adult bone marrow. Additionally, the factors which bind preferentially to the HPFH mutants may play positive roles in establishing an active chromatin structure.


Cell culture

MEL cells were cultured in Dulbecco's modified Eagles' medium (GIBCO) with 10% fetal calf serum (GIBCO) and K562 cells were cultured in RPMI 1640 medium (GIBCO) with 10% fetal calf serum in a 5% CO2 atmosphere. The K562 cells were erythroid induced, either with 0.05 m M haemin ( Rutherford et al, 1979 ) or 2 m M sodium butyrate ( Rowley et al, 1981 ). The cells were incubated with the inducers for 3 d prior to nuclear extract preparation.

Nuclear extract preparation

Nuclear extracts were prepared from MEL and K562 cells as previously described for the preparation of Xenopus and chicken red blood cell nuclear extracts ( Partington et al, 1997 ), except that PMSF was replaced in all buffers by 1 m M AEBSF [4-(2-aminoethyl)benzenesulphonylfluoride, HCl] (Calbiochem).

The protein concentration in nuclear extracts was determined using the BCA protein assay reagent (Pierce).

Electrophoretic mobility shift assay (EMSA)

The assay for DNA binding activity in extract preparations was as described ( Perkins et al, 1989 ; Partington et al, 1997 ). For experiments where the poly(dI-dC).poly(dI-dC) was added after the probe, incubations were for 30 min at 4°C prior to the addition of the non-specific competitor and for 30 min at 4°C subsequent to addition. Extracts were used at a final concentration of 50–100 mg/ml and the assays in each experiment were carried out with equal protein loadings.


The double-stranded γ-globin distal CCAAT site oligonucleotides used as probes are shown below ( Berry et al, 1992 ). The single base changes of the Greek (−117, G to A) and the Japanese (−114, C to T) non-deletion HPFHs are shown underlined.




The Black HPFH deletion probe sequence is shown below.



The proximal γ-CCAAT oligonucleotide probe sequence is shown below.



The αG2 and αG8 GATA-1 binding oligonucleotides ( Plumb et al, 1989 ), the βCCAAT and αCCAAT probes ( deBoer et al, 1988 ), the NF1 ( Rupp et al, 1988 ) and the SCL oligonucleotides ( Hsu et al, 1994 ) were as described.


Factor binding to the wt Aγ and −117 HPFH CCAAT elements under alternative non-specific competitor conditions

To test the possibility that poly(dI-dC).poly(dI-dC) could compete with the distal γ-gene CCAAT site for factors, EMSA was used to compare binding to the wt and the −117 mutant when poly(dI-dC).poly(dI-dC) was added to K562 nuclear extract either before (standard binding conditions) or after the probe (modified binding conditions). Under standard binding conditions a weak doublet (band A) was evident, bound to the wt and the −117 probes (Fig 2, lanes 3 and 4), but with a slightly higher affinity for the −117 mutant. Both bands were competed by the human α- or β-globin CCAAT site oligonucleotides, known to bind CP1 ( deBoer et al, 1988 ; Kim et al, 1988 , 1990; and data not shown). In comparison to GATA-1 binding to the canonical GATA-1 binding oligonucleotide αG2 (lane 1), there was no protein of similar mobility bound to either of the γ probes. Further comparison of binding to the wt and −117 mutant probes revealed the major difference to be the presence of three high-mobility bands (collectively designated D, Fig 2) bound to the wt probe, whereas there was only one factor of different mobility bound to the −117 probe in this region. The band D proteins bound to the wt probe migrated in the region of factors bound to the GATC probe, which was bound by GATA-2 and -3 but not GATA-1 (lane 2) ( Ko & Engel, 1993). However, full-length GATA-2 had a mobility just less than that of GATA-1 (Fig 4, lane 2; Zon et al, 1991 ; Towatari et al, 1995 ).

Figure 2.

) using standard binding conditions are also shown.

Figure 4.

γ probes with K562 extract from uninduced (lanes 1–5), haemin-induced (lanes 6–8) and butyrate-induced (lanes 9–11) cells, using modified binding conditions. 32P-labelled αG2 and GATC-labelled probes (20 fmol/assay) were used under standard binding conditions.

The factor binding profile to the wt γ and −117 probes when the non-specific competitor was added after the probe is shown in Fig 2 (lanes 5 and 11). A considerably different pattern of binding was seen: the most prominent feature was the presence of two new bands (B and C), whereas CP1 binding was absent. The lower mobility band (B) was bound with higher affinity to the −117 probe than to the wt probe, but the higher mobility band (C), was only slightly more intense with the −117 probe and was just visible under standard binding conditions (lanes 3 and 4). In contrast, binding of factors in the band D region for both the wt and −117 probes, although weaker, was identical to that seen under normal binding conditions. To determine which of the factors was specifically bound to the wt and mutant probes, each was competed with an excess of several unlabelled oligonucleotides. Binding of bands B and C to the wt probe and the −117 mutant was competed by either an excess of unlabelled wt (lanes 6 and 12) or −117 (lanes 7 and 13) oligonucleotide. Strong competition of band B, and to a lesser extent band C, was also evident with the GATA factor site αG2 (lanes 9 and 15). The band D factors bound to the wt γ-probe were not competed by αG2, nor was the single factor bound to the −117 probe in this region. With the GATC oligonucleotide as competitor (lanes 10 and 16), bands B and C were reduced, as was the binding of all factors to both probes in the band D-region. The proximal γ-CCAAT oligonucleotide did not affect binding of either band B or C (lanes 8 and 14) but did compete for the binding of the D-region proteins to both probes.

In summary, we observed previously unreported differences in factor binding to the wt and the −117 mutant probes under the modified EMSA conditions. The finding that factors B and C were detected only under the modified binding conditions indicated that they had a low on-rate, but once bound have in addition a low off-rate, and were then not susceptible to competition by poly(dI-dC).poly(dI-dC). Although band C was competed by an excess of several competing oligonucleotides, a protein of similar mobility was observed with extract from HeLa cells, whereas band B was absent (data not shown). Furthermore, a band of similar mobility to band C but not band B was observed to bind to the αG2 and the human α- and β-CCAAT oligonucleotides when these probes were assayed with K562 extract under the modified binding conditions (data not shown). Therefore we consider it to be more probable that of bands B and C, only band B protein, which had a higher binding affinity for the HPFH mutant site and may be an erythroid-specific factor, represented a potentially significant difference between the wt and the mutant. The other observed difference in factor binding to the wt and −117 mutant was band D, the set of three specific factors which bound under both conditions to the wt γ, whereas with the −117 probe the pattern was altered in this region. These bands have a similar mobility to the NFE6 factor previously reported to bind selectively to the wt γ-probe ( Berry et al, 1992 ).

Competition of factor binding by salmon sperm DNA

To determine which factor binding profile would be observed when another non-specific binding competitor was used, EMSA was carried out under normal binding conditions using salmon sperm DNA ( Kristie & Roizman, 1986; SivaRaman & Thimmappaya, 1987). For this purpose the −114 HPFH mutant, which has a binding profile of factors B and C similar to the −117 mutant (see Fig 4, lane 5), was competed over a concentration range (20 ng to 2 mg per assay). The majority of band B binding was competed by 250 ng of salmon sperm DNA, whereas band C was only slightly affected by the DNA competitor over the entire concentration range (Fig 3, lanes 1–8). GATA-1 was bound at low concentrations of competitor DNA (lanes 1–3) but was competed out at higher concentrations (lanes 4–8). This result shows that with competition by genomic DNA under standard binding conditions the binding profile resembles that seen with poly(dI-dC).poly(dI-dC) under the modified conditions. Moreover band B and GATA-1 binding were competed by the genomic DNA which would be expected to have a low level of competing binding sites. However, band C was competed poorly, suggesting that it has a much higher binding preference for poly(dI-dC).poly(dI-dC) than for salmon sperm DNA. When poly(dI-dC).poly(dI-dC) was used as the nonspecific competitor under normal binding conditions, band B binding was stable when the heteropolymer was at a concentration about a third of that routinely used in this laboratory and the overall factor binding profile was the same as that seen with salmon sperm DNA (data not shown). Therefore these results further strengthen the notion that the in vitro binding of factors B and C more closely reflects the in vivo situation.

Figure 3.

2P-labelled probe used was −114 γ (20 fmol/assay).

Changes to the protein binding profile seen on induction of K562 cells with haemin or butyrate

K562 cells constitutively express embryonic and fetal globin genes and a weak stimulation of globin synthesis is observed on exposure to haemin ( Rutherford et al, 1979 ), as a consequence of enhanced transcription ( Dean et al, 1983 ). The cells can also be induced with sodium butyrate ( Rowley et al, 1981 ). Consequently we decided to monitor any changes which might occur in the binding of either the band B or band D factors during erythroid induction to illuminate their respective roles as transcriptional activators or repressors.

A comparison was made by EMSA (Fig 4) under modified binding conditions to wt, −117 and −114 HPFH mutant probes with nuclear extracts prepared from uninduced (lanes 1–5) and haemin (lanes 6–8) or butyrate (lanes 9–11) induced K562 cells. The pattern of binding to the wt and −117 probes (lanes 3 and 4) with uninduced K562 extract was identical to that shown in Fig 1. In addition, the −114 mutant probe (lane 5) showed the strongest binding of band B and a similar intensity of binding with band C as the −117 mutant, and no factor binding in the band D region. After haemin induction, binding of the band B factor was reduced with all three probes, and two additional bands were seen in this region with the −117 and −114 probes (lanes 7 and 8). Furthermore, with all three probes an additional band was now evident in the band D region. Binding of the band D factors to the wt probe after haemin induction decreased (cf. lane 3 to lane 6). In contrast to haemin induction, extract from butyrate-induced cells showed decreased binding of both bands B and C to all three probes, but no other changes in comparison to uninduced cells (lanes 9–11). In particular, band D binding to the wt probe was not reduced.

In summary, more pronounced changes in factor binding occurred with haemin than with butyrate induction. Although both induced a reduction of band B, band D binding to the wt γ probe was decreased only by haemin induction, whereas butyrate had little effect. Thus there was no clear correlation between binding of factors B and D and induction of γ-globin gene transcription in the K562 cell system.

Factor binding under modified binding conditions with MEL cell extract shows major differences from K562 cells

To determine if MEL cells, which express an adult globin gene profile ( Marks & Rifkind, 1978; Nudel et al, 1977 ), exhibited similar protein binding to that seen with the embryonic/fetal globin expressing K562 cells under the modified binding conditions, an EMSA of MEL nuclear extract was carried out (Fig 5). For comparison, binding under normal assay conditions is shown where CP1 binding (band A) was seen with the wt and −117 probes (lanes 3 and 4) but not with the −114 probe (lane 5), as previously reported ( Fucharoen et al, 1990 ). In comparison to the GATA-1 binding seen with the αG2 probe (lane 1), GATA-1 binding was strongest to the −114 probe (lane 5) and equally weak to the wt (lane 3) and −117 mutant (lane 4). The identity of this band, with all probes, was confirmed to be GATA-1 by supershifting with a GATA-1 monoclonal antibody (data not shown). As with K562 extract, the wt probe (lane 3) bound the D band proteins which comigrated with those bound to the GATC probe (lane 2). When the poly(dI-dC).poly(dI-dC) was added after the probe (lanes 6–8), as for K562 cells, the binding of both the CCAAT factor and GATA-1 to all three γ probes was reduced, but in contrast to K562 cells there were no new bands. Comparison of the normal binding conditions with the αG2 (lane 1) and GATC probes (lane 2) to those seen with the modified conditions (lanes 9 and 10) revealed only minor changes in the main bands. The binding of the band D factors to the wt probe was reduced slightly under the modified conditions.

Figure 5.

) or after the probe (lanes 6–10); αG2 probe (lanes 1 and 9); GATC probe (lanes 2 and 10).

In summary, this result showed that K562 and MEL cells both had the band D factors but did not share the band B and C factors. It furthermore showed that CP1 and GATA-1 would bind under the modified conditions, albeit more weakly, when there was no competition by factors B and C. Overall it is possible that the B factor found in K562 cells, an embyronic/fetal globin expressing cell line ( Rutherford et al, 1979 ), may be a stage-specific factor not present in the proerythroblastoid, adult globin expressing MEL cells.

GATC affinity column purified factors contain GATA-2 epitopes and bind to the wt γ-CCAAT probe

The band D factors, which have a higher affinity for the wt γ-probe than for the mutant probes, are expressed in the adult environment and undergo a decrease in abundance on haemin induction of K562 cells, may have a repressor function. To further characterize their relationship with the proteins of similar mobility which bound to the GATC probe, MEL nuclear extract was fractionated on a GATC affinity column and one of the peak bound fractions was assayed for binding to the wt γ-probe. The results are shown in Fig 6. The binding of affinity purified protein to the GATC probe (lane 1) was inhibited by the addition of monoclonal GATA-2 antibody 27 (lane 2), whereas inclusion of GATA-2 monoclonal antibody 96 resulted in supershifting of the factor (lane 3), consistent with the fraction containing GATA-2 epitopes. This same purified fraction bound to the wt γ-probe (lane 4) and was competed by either self competition (lane 5) or by an excess of unlabelled GATC (lane 6). The mobilities of the complexes bound to the wt γ-probe were not as discrete as those bound to the GATC probe, for reasons which were unclear. These results indicated that the high mobility bands bound to the GATC probe in MEL extracts contained GATA-2 isoforms and bound to the wt γ-probe.

Figure 6.

is present in lane 3. Unlabelled competitor oligonucleotides (1 pmol/assay) were added as indicated.

The Black deletion HPFH

Another naturally occurring mutation of the distal Aγ-globin CCAAT site is a 13 bp deletion known as the Black deletion HPFH (Fig 1) ( Gilman et al, 1988 ). An initial EMSA analysis of factor binding to this probe under standard and modified binding conditions revealed that the same factors were bound under both conditions, although there were some differences in the relative band intensities (data not shown). An EMSA, under standard binding conditions using K562 extract, is shown in Fig 7. Of the three bands present (lane 3), only the highest mobility band (designated with an arrow) was competed when an excess of self (lane 4), wt γ (lane 5), −117 mutant (lane 6), proximal γ-CCAAT (lane 7), GATC (lane 10) or β-CCAAT (lane 12) unlabelled oligonucleotides were included. The αG2, αG8, SCL, αCCAAT and NF1 oligonucleotides (lanes 8, 9, 11, 13 and 14) did not compete. These results showed that only one factor bound specifically to this sequence. Furthermore, in addition to being self-competed, it was also competed by both the wt and the −117 mutant CCAAT oligonucleotides, showing that the same factor will bind to the non-deleted sequences. In addition to being competed by GATC, this factor had a similar mobility to the main GATC-bound factor (cf. lanes 2 and 3). One of the slower mobility bands, which were not competed by any of the unlabelled oligonucleotides, had a similar mobility to band C. Band B was absent. Therefore there was no specific factor displaying preferential binding to HPFH mutant CCAAT boxes that was common to both this deletion HPFH and the two non-deletion mutants.

Figure 7.

Fig 7. EMSA of Black deletion HPFH probe with K562 nuclear extract and competition by unlabelled oligonucleotides (1 pmol/assay) as indicated. 32P-labelled αG2 or GATC were also used as probes as indicated.

The pattern of protein binding under modified conditions with uninduced A4 cells has similarities to that of K562 cells but most of the factors are lost upon erythroid induction

As discussed in the introduction, it is most probable that in vivo the distinction between the wt and HPFH mutant promoters is made in adult bone marrow progenitor cells. We therefore investigated the profile of factor binding with nuclear extract prepared from uninduced and erythroid induced A4 cells. Under standard binding conditions we found only CP1-like factors (data not shown), hence the repertoire of factors detected in the A4 progenitor cell extracts was more limited than that found in the erythroleukaemia cell lines. We then compared factor binding under the modified binding conditions to the wt γ, −117, −114 and Black deletion mutants with A4 uninduced, and 3 d and 4 d erythroid induced extracts. The results using K562 extract, shown for comparison (Fig 8, lanes 1–4), are as seen previously (see Figs 1 and 3). Using uninduced A4 extract the wt and −117 mutant were similar in the band B region, each having two bands (cf. lanes 5 and 6), although binding to the mutant was stronger. In this region the −114 (lane 7) and the deletion mutant (lane 8) had only one band, but these had different mobilities. In the band C region, each of the HPFH mutant probes showed increased factor binding in comparison to the wt sequence, although each appeared to bind a different factor or factors. Factor binding in the band D region was evident with the wt probe but there was no specific factor binding in this region to the −114 or the −117 probe. On erythroid induction factor binding to the wt and non-deletion γ-probes in the B and C regions decreased (cf. lanes 5, 6 and 7 to lanes 9, 10 and 11, and lanes 13, 14 and 15) but band D binding to the wt sequence remained (cf. lane 5 to lanes 9 and 13). With the deletion mutant, binding in the band B region was lost on erythroid induction, and binding of the factor in the band C region became weaker, whereas the weak binding in the D region was significantly strengthened on erythroid induction (cf. lane 8 to lanes 12 and 16).

Figure 8.

), 3 d erythroid induced (lanes 9–12), 4 d erythroid induced (lanes 13–16). Uninduced K562 nuclear extract is also shown (lanes 1–4).

In summary, the factor binding we detected in A4 cells, unlike that in either K562 or MEL cells, individually distinguished the wt sequence and the non-deletion mutants. However, as with K562 and MEL cells, we found that the wt γ-probe strongly bound the D band factors in uninduced A4 cells, and furthermore this binding was maintained on erythroid induction of the cells.


It has been widely recognized for some time that the pharmacological reactivation of the human γ-globin genes represents an attractive strategy for alleviating the symptoms of β thalassaemia and sickle-cell disease ( McCaffrey et al, 1997 ; Ronchi et al, 1996 ; Wood & Weatherall, 1983; Karlsson & Neinhuis, 1985). As a first step in determining how this programme could be modified by drug intervention it would be useful to ascertain whether it occurs by silencing of the wt γ-genes, achieved by binding of repressor proteins, or whether alternatively the HPFH mutants bind activator proteins.

In this paper we have shown that under both standard and modified binding conditions the normal γ-gene, but not the HPFH mutants, bound high mobility factors which were candidate repressors. They contain GATA-2 epitopes and thus appear to be GATA-2 isoforms. In relation to this observation, GATA-2 truncated forms have been found previously, but were attributed to degradation of the full-length protein ( Wilson et al, 1990 ; Kamesaki et al, 1996 ). In our case, the integrity of extracts was established by monitoring other transcription factors, including GATA-1. That GATA-2 variants may have a role during cell differentiation was indicated by the observation that mast cells have a splice variant of GATA-2, lacking the N-terminal zinc finger ( Paisley & Huff, 1996). Furthermore it has been reported recently that GATA-5 has two isoforms, one of which lacks the N-terminal finger ( MacNeill et al, 1997 ). The single-finger isoform was observed to have a similar efficiency of binding to DNA, but to be less efficient than the full-length protein at transactivating a reporter construct. Also, alternative splicing of a GATA transcript has been shown to play a role in Bombyx egg development ( Drevet et al, 1995 ).

We observed the GATA-2 isoforms (band D) in all of the adult-derived erythroid cell lines examined, including K562 cells ( Lozzio & Lozzio, 1975; Andersson et al, 1979 ). The one source of erythroid cells that lacked band D was embryonic mouse liver (data not shown), where expression of the human γ-gene occurs in transgenic mice. It is probable that it is the presence of these GATA-2 isoforms in the adult progenitor cells which is critical, as it is here that they may act as repressors of the wt γ-gene. Furthermore, it is possible that both the distal and proximal CCAAT sites are targets for stage-specific repression, as it has been shown that several proteins bind to the proximal CCAAT region of the embryonically expressed galago γ-gene, but not to fetally expressed γ-genes, to repress the activity of the embryonic genes at the fetal stage ( Gumucio et al, 1994 ).

Under standard EMSA binding conditions the results presented here are in general agreement with data previously published by Berry et al (1992 ). However, under modified binding conditions, EMSA was characterized by binding of two new factors (B and C) in nuclear extracts of K562 but not MEL cells. The lower mobility factor (band B) was more strongly bound to the non-deletion HPFH oligonucleotides than to the wt probe. Competition data indicated that binding by both of these factors was specific. However, as the affinity of C for the wt and non-deletion probes was very similar it is unlikely that it plays a selective role in the transcriptional control of the HPFH mutants. In contrast, factor B is a candidate activator of the HPFH mutants in erythroid cells. Although we have not identified factor B in this study, competition studies indicate that it has affinity for GATA binding sites.

In formulating a model for the control of γ-gene expression via the distal CCAAT site, it is necessary to take into consideration the finding that the distal CCAAT box is not required for expression of the γ-gene at the fetal stage of development ( Ronchi et al, 1996 ). Furthermore, in transient expression assays in fetal/embryonic globin expressing K562 cells, which contain both the putative wild-type gene repressor (D) and the putative HPFH activator (B), the −117, the −114 and the Black deletion HPFH mutants were not up-regulated in comparison to the wt promoter ( Katsube & Fukumaki, 1995; Ulrich & Ley, 1990; Mantovani et al, 1989 ). Thus, factors B and D were unable to act in these cells. Furthermore, although D was present in adult globin expressing MEL cells, B was not. However, it has previously been suggested that the developmental regulation of globin gene expression involves an initial exposure to trans-acting factors at an early stage of erythroid differentiation, subsequently allowing a differential accessibility of the genes to transcription factors ( Stanworth et al, 1995 ). These workers concluded that the transcription factor complement of adult globin expressing erythroblasts, such as MEL cells, was not sufficient for selective globin gene control. Consistent with the proposal that stem cells have a distinct set of factors, in addition to selective binding of factor D to the wild type probe, we observed selective factor binding to the non-deletion HPFH probes in uninduced but not erythroid induced A4 cells. The band B seen in nuclear extracts from K562 cells which, although expressing fetal/embryonic globins, were derived from an adult ( Lozzio & Lozzio, 1975), may be one of the factors seen in the A4 cells. Furthermore a change in the factor binding profile to the Black deletion HPFH probe was also seen after erythroid induction of the A4 cells, where after induction the most prominent band was a high mobility band, with the same mobility as that bound using K562 extract.

Overall we propose that the factors binding preferentially to the HPFH mutants may have an activating role, whereas the GATA-2 isoforms (band D), which bind preferentially to the wt gene, may act as repressors.


We gratefully acknowledge Claire Heyworth and Stella Pearson of the Patterson Institute (Manchester) for providing uninduced and erythroid induced A4 cells, Lyn Healy of the Leukaemia Research Fund Centre (Institute of Cancer Research) for supplying uninduced A4 cells, and Elisabeth Ehler (DBRC) for mouse embyro liver. We also thank Doug Engel for supplying the GATA-2 monoclonal antibodies. This research was supported by a grant from the Wellcome Trust.