The leukaemia-associated transcription factors EVI-1 and MDS1/EVI1 repress transcription and interact with histone deacetylase


Rotraud Wieser, Institut für Medizinische Biologie der Universitaet Wien, Waehringerstr. 10, A-1090 Wien, Austria. E-mail:


EVI-1 and its variant form, MDS1/EVI1, have been reported to act in an antagonistic manner and be differentially regulated in samples from patients with acute myeloid leukaemia and rearrangements of the long arm of chromosome 3. Here, we show that both EVI-1 and MDS1/EVI1 can repress transcription from a reporter construct containing EVI-1 binding sites and interact with histone deacetylase in mammalian cells. This interaction can be recapitulated in vitro and is mediated by a previously characterized transcription repression domain, whose activity is alleviated by the histone deacetylase inhibitor trichostatin A.

EVI-1 (ecotropic viral integration site 1) was originally identified because its transcriptional activation through retroviral integration causes leukaemia in a murine model system (Morishita et al, 1988). A role of EVI-1 in human leukaemogenesis is supported by the fact that the hEVI-1 gene maps to a subregion of chromosome band 3q26 that is affected by several recurrent chromosome aberrations in myeloid leukaemia (Morishita et al, 1992a). Some of these lead to the formation of leukaemogenic fusion proteins involving EVI-1 (Nucifora et al, 1994; Peeters et al, 1997), others to the aberrant activation of its transcription (Fichelson et al, 1992; Morishita et al, 1992a; Suzukawa et al, 1994; Soderholm et al, 1997; Testoni et al, 1999). EVI-1 mRNA is also expressed in a subset of myeloid leukaemia samples that lack cytogenetically detectable 3q26 rearrangements (Russell et al, 1994; Dreyfus et al, 1995; Ogawa et al, 1996; Gerhardt et al, 1997; Xi et al, 1997), but is present at very low levels in normal peripheral blood and bone marrow.

The EVI-1 gene codes for a zinc finger protein that may activate transcription under certain circumstances (Morishita et al, 1995), but more generally is presumed to act as a transcriptional repressor (Kreider et al, 1993; Matsugi et al, 1995; Bartholomew et al, 1997; Kilbey & Bartholomew, 1998). A proline-rich repression domain (RD) has been mapped to the central portion of the protein, between two zinc finger domains. These comprise seven and three zinc fingers, respectively (Bartholomew et al, 1997), and have been shown to be able to bind to specific DNA sequences independently of each other (Funabiki et al, 1994; Matsugi et al, 1995; Perkins & Kim, 1996). The consensus recognition sequence for the first set of zinc fingers includes a binding site for the erythroid-specific transcription factor GATA-1, and EVI-1 antagonizes transcriptional activation of synthetic reporter constructs by GATA-1 (Kreider et al, 1993). Furthermore, EVI-1 may inhibit GATA-1-dependent erythroid differentiation (Kreider et al, 1993; Louz et al, 2000), as well as myeloid differentiation of 32Dc13 cells (Morishita et al, 1992b). However, contradictory results have also been presented (Khanna Gupta et al, 1996; Fontenay Roupie et al, 1997).

More recently, an alternative form of EVI-1, MDS1/EVI1, has been described (Fears et al, 1996; Wimmer et al, 1998). MDS1/EVI1 differs from EVI-1 only by the presence of an extended N-terminus, and has been reported to act as a transcriptional activator and antagonist of EVI-1. Only the EVI-1 isoform was induced in myeloid leukaemia samples with 3q rearrangements, whereas MDS1/EVI1 levels were similar to those in healthy controls (Soderholm et al, 1997).

To date, little is known about genes physiologically regulated by either EVI-1 or MDS1/EVI1, even though essential biological functions of these transcription factors were revealed by gene-targeting experiments in mice (Hoyt et al, 1997). Also relatively unexplored are the interactions of EVI-1 with other cellular proteins. EVI-1 has been shown to interfere with transforming growth factor β (TGF-β) signalling by binding to Smad proteins (Kurokawa et al, 1998), an effect that may be brought about by its interaction with the transcriptional co-repressor, CtBP (Turner & Crossley, 1998; Izutsu et al, 2001). MDS1/EVI1, on the other hand, enhanced TGF-β-dependent effects (Sood et al, 1999). Finally, a role of EVI-1 in addition to that in transcriptional regulation has been suggested by its ability to interact with and inhibit jun N-terminal kinase (JNK) (Kurokawa et al, 2000).

In recent years, it has become apparent that the action of many transcription factors involves the modification of chromatin structure. Transcriptional activators induce an open, transcriptionally competent chromatin configuration in their target promoters by recruiting histone acetyltransferases (HATs), which loosen nucleosome/DNA interactions by acetylating core histones. Conversely, many transcription repressors interact with histone deacetylases (HDACs), whose enzymatic activity contributes to tight nucleosomal packing of DNA (Jacobson & Pillus, 1999; Kouzarides, 1999; Redner et al, 1999). Genes encoding transcription factors are frequent targets of recurrent leukaemia-associated chromosome aberrations, which lead to their aberrant expression and/or the formation of oncogenic fusion proteins. A number of these leukaemogenic factors, e.g. PML-RARα, PLZF-RARα, AML-ETO, AML-TEL, or TAL1 have been shown to functionally associate with HDACs (Grignani et al, 1998; Lin et al, 1998; Wang et al, 1998, 1999; Chakrabarti & Nucifora, 1999; Fenrick et al, 1999; Huang & Brandt, 2000). Inhibitors of HDAC are currently being tested in clinical trials and have already proved to be of some therapeutic benefit (Warrell et al, 1998).

Here, we show that not only EVI-1, but also MDS1/EVI1, are able to act as repressors of transcription, and that both proteins interact with HDAC-1. The functional significance of this interaction is demonstrated by the ability of the histone deacetylase inhibitor, trichostatin A (TSA), to diminish the effects of the proline-rich repression domain of EVI-1.

Materials and methods

Expression and reporter plasmids EVI-1 and MDS1/EVI1 were cloned from human placenta (gift of Dr R. Hofbauer) and pancreas cDNA libraries (Stratagene, La Jolla, CA, USA) through filter screening and polymerase chain reaction (PCR) amplification. The sequence of the reconstituted full-length EVI-1 clone corresponded to that reported by Mitani et al (1994), which is slightly divergent from that reported by Morishita et al (1990). The MDS1 sequence was as reported previously (Fears et al, 1996). pCMV5 or pEFzeo (Kovarik et al, 2001) were used for expression in mammalian cells. To facilitate immunoprecipitation and detection on Western blots, two copies of an HA epitope tag were engineered into the very N-termini of both EVI-1 and MDS1/EVI1. This modification did not affect the biological properties of the proteins.

The GST-EVI-1 expression plasmid, pGEX-2T-EVI-1, as well as plasmids encoding the truncated fusion proteins GST-EVI-1(1–856), GST-EVI-1(1–731), GST-EVI-1(1–268), GST-EVI-1(ZF1) (amino acids 17–249), GST-EVI-1(RD) (amino acids 484–739), GST-EVI-1(ZF2) (amino acids 733–827) and GST-EVI-1(AR) (amino acids 875–949) were created by subcloning appropriate restriction or PCR fragments into pGEX-2T (Amersham Pharmacia). Similarly, full-length EVI-1 or the ZF1, RD, ZF2 or AR subdomains were cloned into the expression vector pM (Clontech, Palo Alto, CA, USA) to generate GAL4 DNA binding domain (GAL4 DBD) fusion constructs (pM-EVI-1 series).

The luciferase reporter plasmid pERE-luc was constructed by inserting two repeats of the recognition site for EVI-1 zinc fingers 1–7 (TGACAAGATAAGATAA; Perkins & Kim, 1996) and a TK minimal promoter into pGL3basic (Promega, Madison, WI, USA). Replacing the CAT part of pGAL4-TK-CAT (Lee et al, 1995) with the luciferase part of pGL3-basic yielded pGAL4-TK-luc.

Cell culture, transfections and protein assays 293T cells were grown in Dulbecco's modified Eagle's medium (DMEM) and COS-1 cells in Roswell Park Memorial Institute (RPMI)-1640 medium, both supplemented with 10% fetal calf serum (FCS; Gibco Life Technologies). Exponentially growing 293T cells were transfected in 24-well plates using DAC-30 (Eurogentec, Seraing, Belgium) according to the manufacturer's instructions. pM expression vectors containing GAL4 DBD fusions were transfected at 0·3 µg per well, along with 0·1 µg of pGAL4-TK-luc. pERE-luc (0·1 µg) was co-transfected with 0·3 µg of the GATA-1 expression plasmid pXM-GATA-1 (Martin & Orkin, 1990) and the indicated amounts of pEFzeo-EVI-1 or pEFzeo-MDS1/EVI1, filled up to 2 µg of total DNA with pEFzeo. The renilla luciferase vector pRL (0·01 µg/well) (Promega, Madison, WI, USA) was used to control for transfection efficiency. Cell extracts for luciferase assays were prepared 2 d after transfection. For some experiments, 50 ng/ml trichostatin A [TSA; Wako Pure Chemical Industries, Osaka, Japan; 1 µg/µl stock solution in dimethyl sulphate (DMSO)] or an equivalent amount of solvent were added for the last 15–18 h before cell lysis. Luciferase activity was measured in a Victor multilabel counter (Perkin Elmer) or a Mediators PhL counter (Mediators, Vienna, Austria), using the Luciferase Assay System or the Dual Luciferase Reporter Assay Kit (Promega).

COS-1 cells were transiently transfected by electroporation in ice-cold serum-free RPMI medium at 960 µF and 240 V using a BioRad (Hercules, CA, USA) Gene Pulser apparatus. pCMV5-HA-EVI-1 or pCMV5-HA-MDS1/EVI1 (7·5 µg), along with 7·5 µg of pClneoHDAC1myc (Bartl et al, 1997), were used per 150 mm plate. After 2 d, cell extracts were prepared and immunoprecipitations and Western blots were performed as described previously (Doetzlhofer et al, 1999).

GST pull-down assays HDAC-1 and various truncation mutants thereof were synthesized in vitro in the presence of 35S-methionine using the Quick Coupled Transcription/Translation System (Promega). GST-EVI-1 fusion proteins were expressed in and purified from Escherichia coli BL21 as described (Karlseder et al, 1996). Amounts of protein used in pull-down assays were normalized by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and Phosphoimager quantification of in vitro translation products; and by SDS-PAGE and Coomassie staining of purified GST fusion proteins. Glutathione agarose beads (Sigma-Aldrich, St. Louis, MO, USA) coated with GST fusion proteins were incubated with in vitro-translated HDAC-1. After repeated washes with incubation buffer (HUNT buffer, 20 mmol/l TrisHCl, pH 8·0, 100 mmol/l NaCl, 1 mmol/l EDTA, 0·5% IGEPAL), bound proteins were released by boiling in SDS-PAGE loading buffer and resolved using SDS-PAGE. Results were analysed using a Storm 840 Phosphoimager (Amersham Pharmacia).


Both EVI-1 and MDS1/EVI1 act as repressors of transcription

It has been reported that EVI-1 acts as a transcriptional repressor, and MDS1/EVI1 as a transcriptional activator and antagonist of EVI-1. Furthermore, only EVI-1, but not MDS1/EVI1, appeared to be induced in samples from patients with acute myeloid leukaemia (AML) and 3q26 aberrations. In contrast, we found that MDS1/EVI1 mRNA levels may also be elevated in some leukaemia samples with 3q26 rearrangements (unpublished observations). In order to determine the functional consequences of the observed differences in the relative levels of EVI-1 and MDS1/EVI1, we transfected 293T cells with an EVI-1 responsive reporter construct (pERE-luc), along with varying amounts of EVI-1 and MDS1/EVI1 expression plasmids and the transcriptional activator GATA-1. To our surprise, even when MDS1/EVI1 was transfected in a 10-fold excess over EVI-1, it did not activate transcription, but rather slightly enhanced repression by EVI-1 (Fig 1A). Furthermore, MDS1/EVI1 also repressed the luciferase reporter in the absence of co-transfected EVI-1, and was almost as potent as EVI-1 in this assay (Fig 1B). Similar observations were made when the GATA-1 plasmid was omitted from the transfections, probably as a result of the expression of endogenous GATA-1 in 293T cells, as determined by reverse transcription (RT)-PCR (data not shown). Neither EVI-1 nor MDS1/EVI1 had more than a marginal effect on the luciferase reporter lacking the ERE (not shown).

Figure 1.

 Both EVI-1 and MDS1/EVI1 repress transcription from a reporter construct containing EVI-1 binding sites. The indicated amounts (µg) of EVI-1 and MDS1/EVI1 expression plasmids were transiently transfected into 293T cells, along with a GATA-1 expression plasmid, a firefly luciferase reporter under the control of two EVI-1 binding sites, and a renilla luciferase plasmid to control for transfection efficiency. Empty expression vector was added so that equal amounts of DNA were present in each transfection. Indicated activity is firefly luciferase relative to renilla luciferase activity. (A) MDS1/EVI1 and EVI-1 repress transcription in an additive manner. (B) MDS1/EVI1 is almost as potent a repressor of transcription as EVI-1.

Both EVI-1 and MDS1/EVI1 interact with histone deacetylase

In order to elucidate the mechanism of repression by EVI-1 and MDS1/EVI1, we tested whether these proteins could interact with histone deacetylase in mammalian cells. HA epitope tags were engineered onto the amino termini of EVI-1 and MDS1/EVI1, which did not affect their functional properties, as determined by luciferase reporter assays (not shown). HA-EVI-1 or HA-MDS1/EVI1 expression plasmids were transfected into COS-1 cells, along with an expression plasmid for HDAC-1. After 2 d, cell lysates were prepared and HDAC-1 or HA-tagged proteins were immunoprecipitated with the respective antibodies, resolved in SDS polyacrylamide gels and blotted onto nitrocellulose membranes. Proteins were specifically detected using either HA or HDAC-1 antibodies (Fig 2). HDAC-1 was co-immunoprecipitated with both EVI-1 and MDS1/EVI1, and, conversely, these transcription factors were present in HDAC-1 immunoprecipitates. Thus, both EVI-1 and MDS1/EVI1 interacted with histone deacetylase in mammalian cells.

Figure 2.

 EVI-1 and MDS1/EVI1 interact with HDAC-1 in mammalian cells. COS-1 cells were transiently transfected with expression constructs for HA-tagged EVI-1 or HA-tagged MDS1/EVI1, along with an HDAC-1 expression vector. After 48 h, cells were lysed and immunoprecipitation (IP) with HA or HDAC-1 antibodies (α-HA and α-HDAC-1 respectively) was carried out. Immunoprecipitates as well as whole cell lysates (‘input’, corresponding to one twentieth of the lysate used for immunoprecipitation) were denatured, electrophoresed and blotted onto nitrocellulose membranes. Proteins were specifically detected using HA or HDAC-1 antibodies as indicated (Western).

Mapping of protein domains involved in EVI-1/HDAC−1 interactions

In order to determine whether the interaction between EVI-1 and HDAC-1 was direct, bacterially expressed, affinity-purified GST-EVI-1 fusion protein was incubated with in vitro-translated, 35S-methionine-labelled HDAC-1 protein. A weak but specific and reproducible interaction between GST-EVI-1 and HDAC-1 was observed (Fig 3A), indicating that additional mammalian cofactors are not stringently required for their association. Full-length MDS1/EVI1 could not be stably expressed in bacteria (unpublished observations) and therefore could not be tested in the pull-down assays.

Figure 3.

 Mapping of HDAC−1 interaction domains in EVI-1. EVI-1 or the indicated deletion derivatives were expressed as GST fusion proteins, affinity purified, and incubated with in vitro-translated, 35S-methionine-labelled HDAC-1. Specifically bound protein was resolved by SDS-PAGE and detected using a Phosphoimager. Coomassie staining of the gels confirmed that similar amounts of each GST fusion protein were used for the assays (not shown). (A) C-terminal truncation mutants of EVI-1. Extracts from untransformed bacteria (‘BL21’) and from bacteria transformed with pGEX-2T (‘GST’) were used as negative controls. Input, in vitro-translated HDAC-1 protein (one fifth of the amount used for binding reactions). (B) Analysis of specific subregions of EVI-1. ZF1, ZF2, zinc finger region 1 (aa 17–249) and 2 (aa 733–827) respectively; RD, repression domain (aa 484–739); AR, acidic region (aa 875–949).

Various deletion derivatives of EVI-1 were used in order to map the region(s) involved in HDAC-1 interactions. Removal of the C-terminal acidic region (AR) and zinc finger region 2 (ZF2) did not affect the interaction between EVI-1 and HDAC-1 (Fig 3A). However, further deletion of the proline-rich repression domain (RD) reduced the affinity between them. When individual regions of EVI-1 were expressed as GST fusions and used for pull-down assays, HDAC-1 interacted with the isolated ZF1, RD and ZF2 domains, but only very weakly with the acidic region (Fig 3B). In summary therefore, the ZF1 and RD domains of EVI-1 are sufficient to mediate interaction with HDAC-1, but the ZF2 domain may also contribute to it.

To determine which part(s) of HDAC-1 is(are) involved in the interaction with EVI-1, various HDAC-1 deletion constructs were used. Similar amounts of the in vitro-translated mutant proteins were incubated with full-length GST-EVI-1. Removal of the N-terminal 130 amino acids of HDAC-1 effectively abolished the interaction with GST-EVI1, while deletion of all but the N-terminal 130 amino acids actually enhanced it (Fig 4), possibly indicating partial masking of the interaction domain in the context of the full-length protein.

Figure 4.

 Mapping of the EVI−1 interaction domains in HDAC-1. Full-length HDAC-1 and various deletion derivatives thereof were translated in vitro in the presence of 35S-methionine. Similar amounts of each protein were incubated with glutathione agarose beads coated with full-length GST-EVI-1 fusion protein. Specifically retained proteins were resolved by SDS-PAGE and detected using a Phosphoimager. The C-terminal nuclear localization sequence was present in all HDAC-1 constructs, but did not affect EVI-1 binding.

The HDAC inhibitor, trichostatin A, attenuates the activity of the EVI-1 repression domain

Finally, we wished to establish whether the interaction with HDAC-1 was of functional importance to transcriptional repression by EVI-1. To this end, EVI-1 or its above described subdomains (Fig 3) were fused to the DNA binding domain (DBD) of the yeast transcription factor GAL4. The resulting expression constructs were transfected into 293T cells, along with a luciferase reporter driven by four copies of the GAL4 DNA binding sequence (pGAL4-TK-luc). Luciferase activity was determined after incubation in the presence or absence of 50 ng/ml trichostatin A (TSA). The previously characterized repression domain (Bartholomew et al, 1997) reduced luciferase activity almost to the same extent as full-length GAL4-EVI-1. The ZF1 domain exhibited some repressive properties, but the ZF2 and AR regions essentially did not affect transcription from the GAL4 promoter (Fig 5A). TSA did not reduce repression by full-length EVI-1 and had virtually no effect on the ZF1, ZF2 and AR regions (Fig 5B). In contrast, inhibition of HDAC partially alleviated repression by the RD, demonstrating the importance of histone deacetylase activity for the function of this domain.

Figure 5.

 TSA partially relieves transcriptional repression by the proline-rich repression domain. Expression constructs containing GAL4 DBD fusions of EVI-1 or the indicated subdomains (corresponding to those described in Fig 3) were transfected into 293T cells, along with a luciferase reporter construct driven by four GAL4 DNA binding sites. (A) Repression by GAL4 DBD fusion proteins compared with the GAL4 DBD alone. (B) De-repression in the presence of 50 ng/ml TSA, compared with the GAL4 DBD alone.


Contrasting roles have been ascribed to EVI-1 and MDS1/EVI1, both with respect to their aberrant expression in samples from patients with leukaemia and 3q26 rearrangements, and their effects on transcription from target promoters (Soderholm et al, 1997; Sood et al, 1999). Our own results, however, demonstrate that not only EVI-1, but also MDS1/EVI1 mRNA levels may be elevated in leukaemias with 3q26 aberrations (3/9 samples tested; data not shown). Furthermore, in our experimental system, MDS1/EVI1 and EVI-1 downregulated transcription in an additive manner, and MDS1/EVI1 by itself, acted as almost as potent a repressor of transcription as EVI-1. This discrepancy between our results and those previously reported may be as a result of differences between the EVI-1 response elements driving the respective reporter constructs: Soderholm et al (1997) used a short genomic fragment selected for in vitro binding to EVI-1, whereas our reporter contained two copies of an oligonucleotide that previously had been shown to specifically bind to both EVI-1 and MDS1/EVI1 in electrophoretic mobility shift assays. It conferred activation by GATA-1 and repression by EVI-1 (Kreider et al, 1993) but, notably, was not activated by MDS1/EVI1 (Soderholm et al, 1997). In our system, it revealed a previously unreported effect of MDS1/EVI1 as a transcriptional repressor. Parallel effects of EVI-1 and MDS1/EVI1 have been observed before in that both diminished transcription from the synthetic TGF-β responsive reporter, p3TP-lux, but this was probably through an indirect mechanism (Sood et al, 1999). Clearly, elucidation of the physiological function of the transcription factors EVI-1 and MDS1/EVI1 will require the identification of their target genes, some of which they may affect in parallel and others in opposite ways.

However, in line with a possible role for MDS1/EVI1 as a transcriptional repressor, we found that this protein, like EVI-1, interacts with histone deacetylase 1 in mammalian cells. This interaction could be recapitulated with purified proteins in vitro. Nevertheless, analogy with other transcription repressors suggests that, in mammalian cells, EVI-1 (and/or MDS1/EVI1) and HDAC-1 are probably present in multiprotein complexes that may include other histone deacetylases (Jacobson & Pillus, 1999; Kouzarides, 1999; Ahringer, 2000) as well as CtBP (Turner & Crossley, 1998; Izutsu et al, 2001). Furthermore, even though not addressed experimentally in this report, EVI-1 and MDS1/EVI1 may also directly bind to other histone deacetylases, especially of the type I class, whose members exhibit sequence similarity to the EVI-1 binding domain of HDAC-1 (amino acids 1–130, Fig 4).

The interaction of EVI-1 with HDAC-1 is mediated mainly through the first zinc finger domain and a previously characterized repression domain. The ZF1 domain of EVI-1 also binds to the TGF-β responsive transcription factor, Smad3 (Kurokawa et al, 1998), and the c-jun N-terminal kinase, JNK (Kurokawa et al, 2000). The RD has very recently been implicated in interaction with the co-repressor CtBP (Izutsu et al, 2001). Consistent with its biochemical activity, we report here that it also binds HDAC-1 and is sensitive to the histone deacetylase inhibitor TSA. In contrast, TSA has only a marginal effect on the full-length EVI-1 protein, possibly as a result of the presence of additional repression domains (Kilbey & Bartholomew, 1998; this report), which may act in a HDAC-1-independent manner. Similar unmasking of TSA-sensitive transcriptional repression in a protein that may downregulate transcription through HDAC-1-dependent and -independent mechanisms has been described for the Smad co-repressor TGIF (Wotton et al, 1999). Furthermore, the growth-repressing Rb and Stra13 proteins negatively regulate some of their target genes in a TSA-sensitive and others in a TSA-insensitive manner (Luo et al, 1998; Sun & Taneja, 2000). Determining the physiological targets of EVI-1 and MDS1/EVI1 will help to resolve the question of whether these proteins similarly reveal dependence on histone deacetylase for transcriptional repression of certain promoters but not of others, and whether the observed interaction between these leukaemia-associated transcription factors and HDAC can be exploited therapeutically.


We gratefully acknowledge the help of M. Nemethova, C. Holzhauser, and Drs E. Wintersberger, W. Strobl, and M. Hüttinger with reporter assays. Dr P. Kovarik generously provided pEFzeo, Dr R. Hofbauer the human placenta cDNA library, and Dr S. Orkin the GATA-1 expression plasmid, pXM-Gata1. We thank Dr W. Pinsker for helpful discussions.

This work was supported by the Fonds zur Förderung der wissenschaftlichen Forschung in Österreich (grant number P14101-GEN to R.W. and P13638-GEN to C.S.).