Transactivation properties of c-Myb are critically dependent
on two SUMO-1 acceptor sites that are conjugated
in a PIASy enhanced manner
Øyvind Dahle
1
, Tor Ø. Andersen
1
, Oddmund Nordga
˚rd
1
, Vilborg Matre
1
, Giannino Del Sal
2,3
and Odd S. Gabrielsen
1
1
Department of Biochemistry, University of Oslo, Norway;
2
Laboratorio Nazionale CIB, Area Science Park, Trieste, Italy;
3
Dipartimento di Biochimica, Biofisica e Chimica delle Macromolecole, Universita
`degli Studi di Trieste, Italy
The transcription factor v-Myb is a potent inducer of
myeloid leukemias, and its cellular homologue c-Myb
plays a crucial role in the regulation of hematopoiesis.
Recently, Bies and coworkers (Bies, J., Markus, J. &
Wolff, L. (2002) J. Biol. Chem,277, 8999–9009) presented
evidence that murine c-Myb can be sumoylated under
overexpression conditions in COS7 cells when cotrans-
fected with FLAG-tagged SUMO-1. Here we provide
independent evidence that human c-Myb is also subject to
SUMO-1 conjugation under more physiological condi-
tions as revealed by coimmunoprecipitation analysis of
Jurkat cells and transfected CV-1 cells. Analysis in an
in vitro conjugation system showed that modification of
the two sites K503 and K527 is interdependent. A two-
hybrid screening revealed that the SUMO-1 conjugase
Ubc9 is one of a few major Myb-interacting proteins. The
moderate basal level of sumoylation was greatly enhanced
by cotransfection of PIASy, an E3 ligase for SUMO-1.
The functional consequence of abolishing sumoylation
was enhanced activation both of a transiently transfected
reporter gene and of a resident Myb-target gene. When
single and double mutants were compared, we found a
clear correlation between reduction in sumoylation and
increase in transcriptional activation. Enhancing sumoy-
lation by contransfection of PIASy had a negative effect
on both Myb-induced and basal level reporter activation.
Furthermore, PIASy caused a shift in nuclear distribution
of c-Myb towards the insoluble matrix fraction. We
propose that the negative influence on transactivation
properties by the negative regulatory domain region of
c-Myb depends on the sumoylation sites located here.
Keywords: c-Myb; transcription; SUMO-1; Ubc9; PIASy.
The c-Myb transcription factor plays a central role in the
regulation of cell growth and differentiation, in particular in
hematopoietic progenitor cells (reviewed in [1]). Homozy-
gous null c-Myb/Rag1 chimerical mice are blocked in early
T-cell development, while mice with a c-myb
null
mutation
display severe hematopoietic defects leading to in utero
death at E15 [2,3]. The c-Myb protein consists of an
N-terminal DNA-binding domain (DBD), a central trans-
activation domain (TAD) and a C-terminal negative
regulatory domain (NRD). The DBD of c-Myb is com-
prised of the three imperfect repeats: R
1
,R
2
and R
3
,each
related to the helix-turn-helix motif [4–7].
Oncogenic alterations, as found in AMV v-Myb, include
both N- and C-terminal deletions as well as point mutations
[8]. AMV v-myb is a potent and cell-type specific oncogene
that transforms target cells in the macrophage lineage and
induces monocytic leukemia [8,9]. Several studies have
attempted to define oncogenic determinants of v-myb.
N- and C-terminal deletions remove several sites of protein
modification, including an N-terminal CK2 phosphoryla-
tion site (S11 and S12) [10], and a putative MAPK-site
(S528) [11–13] as well as acetylation sites [14,15] located in
the deleted portion of the C-terminal NRD. In addition,
specific point mutations in v-Myb abolish protein–protein
interactions [5], as well as phosphorylation as in the case of
V117D [16]. c-Myb has recently also been reported to be
subjected to SUMO-1 (small ubiquitin-related modifier)
conjugation [17].
The SUMO-1 protein is related to ubiquitin, but its
function, although presently unclear, seem to be other
than proteasomal degradation (reviewed in [18,19]). The
sequence homology between ubiquitin and SUMO-1 is
low, but the structures are highly similar [20], and they
use related conjugation mechanisms [21], including the
use of E3-like factors, which was recently identified for
sumoylation as the PIAS proteins (protein inhibitor of
activated STATs) [22–25]. The sequence YKXE has
been proposed as a consensus sequence for SUMO-1
conjugation [26]. The process of sumoylation is conserved
from yeast to man and is a dynamic and reversible
Correspondence to O. S. Gabrielsen, Department of Biochemistry,
University of Oslo, PO Box 1041 Blindern, N-0316 Oslo, Norway.
Fax:+4722854443,Tel.:+4722857346,
E-mail: o.s.gabrielsen@biokjemi.uio.no
Abbreviations: DBD, DNA-binding domain; MRE, Myb recognition
element; NRD, negative regulatory domain; PIAS, protein inhibitior
of activated STATs; SUMO-1, small ubiquitin-related modifier;
TAD, transactivation domain; Ubc9, ubiquitin conjugation enzyme 9.
(Received 5 December 2002, revised 31 January 2003,
accepted 6 February 2003)
Eur. J. Biochem. 270, 1338–1348 (2003) FEBS 2003 doi:10.1046/j.1432-1033.2003.03504.x
process. Both sumoylation and desumoylation are needed
for viability in yeast [27].
Because several different classes of proteins are targets for
SUMO-1 conjugation, it is rather unlikely that a single
explanation for the biological role of sumoylation will be
found. A more general proposition is that sumoylation plays
a role in the stabilization of higher order protein complexes
and modification of protein–protein interactions [18]. This is
consistent with the role of sumoylation in PML nuclear
bodies where it is important for PML nuclear body dynamics
and for recruiting other nuclear body components [28].
In the present study we have extended the findings of Bies
et al. [17] by providing several lines of independent evidence
for this novel modification of c-Myb. We show that c-Myb
interacts strongly with Ubc9 causing sumoylation at two
specific sites in the NRD region of the protein, K527 being a
dominant site and K503 being secondary. When sumoyla-
tion was blocked by mutation of the two modification sites,
this caused a large increase in transcriptional activity of
c-Myb, both when assayed with a transiently transfected
reporter gene and when measuring a resident Myb-target
gene. SUMO-1 conjugation was significantly enhanced by
cotransfection with PIASy, which is the E3 like factor
reported to enhance sumoylation of LEF1 [29]. Further-
more, PIASy seems to increase the fraction of Myb species
in the insoluble part after subnuclear fractionation, which
indicates that sumoylation might be involved in modulating
the protein–protein interactions of c-Myb.
Materials and methods
Plasmids
The yeast bait plasmid pDBT-hcM encoding full-length
human c-Myb fused to the Gal4p DBD was generated
from a cDNA clone [30] and the vector pDBT [31]. The
mammalian expression plasmid pCIneo-hcM contains full-
length human c-Myb cDNA with an optimized ATG
context. A c-Myb-HA fusion cDNA was generated by
cloning oligos encoding the C-terminal part of c-Myb in
fusion with an HA-tag, between PshAI and SalIin
pCIneo-hcM, to give the plasmid pCIneo-hcM-HA. The
cDNAs encoding c-Myb mutants K503R, K527R and
K503/527R (abbreviated 2KR) were generated using the
Quick Change Site-Directed Mutagenesis Kit (Stratagene)
on a subfragment of human c-MYB. Plasmids expressing
full-length SUMO-1 with an HA epitope (HA-SUMO-1)
or in fusion with GFP (GFP-SUMO-1) have been
described [32]. The expression plasmid pGEX-UBC9 was
constructed from a human UBC9 cDNA (isolated in the
two-hybrid screening), and cloned in-frame into pGEX-
6P-2 between SalIandNotI. All cloned fragments
generated by PCR were verified by sequencing. The
c-Myb-responsive luciferase reporter construct pGL2/tk/
3xGG contains multimerized Myb response elements and
its construction is described in [33].
Yeast two-hybrid screen
The yeast two-hybrid screen was performed in the yeast
strain PJ69-4a [34,35] with pDBT-hcM as bait and using
two Matchmaker cDNA libraries (Clontech): from human
bone marrow (HL4053AH) and from the erythroleukemia
cell line K562 (HL4032AH).
Cell culture, transfections and luciferase assays
CV-1 and HD11 cells were grown as described [33,36].
Transient transfections were performed by lipofection
(Lipofectamine-Plus, Gibco Life Technologies) or using
Fugene (Roche Diagnostics). Luciferase assays were per-
formed in triplicate using the Luciferase Assay Reagent
(Promega). Data from three independent transfection
experiments were normalized for protein concentration in
the samples. Equal transfection efficiency was verified by
Western analysis of the transfected species.
In vitro
conjugation assay
The various forms of human c-Myb were generated in
the TNT rabbit reticulocyte lysate system (Promega) in the
presence of [
35
S]methionine. Templates used were either the
appropriate plasmid (pCIneo-hcM) or a PCR product with
T7 promoter added during amplification (ÔTpC-fragmentÕ:
amino acids 410–639). GST-SUMO-1 [37] and GST-UBC9
were expressed and affinity-purified using standard methods
(Amersham Pharmacia Biotech). SUMO-activating enzyme
(E1 fraction) was prepared from CV1 cells as described [37].
SUMO-1 conjugation assays were performed as described
in [32] with purified GST-UBC9 included and incubation
for two hours at 30 C. Reaction mixtures were analysed on
10% polyacrylamide gels revealed by fluorography.
Antibodies
For Myb detection, we used the polyclonal antibody H141
(Santa Cruz) and the monoclonal antibody 5e11 [38].
SUMO-1 was detected with monoclonal antibodies from
Zymed. PIASy-T7 was detected using anti-T7 Ig (Novagen).
Immunoprecipitation and Western blot
CV-1 cells were transfected with the indicated plasmids to
analyse sumoylation of c-Myb. After transfection, cells were
lysed and subjected to coimmunoprecipitation as described
[32] using standard methods.
RNA isolation and real time PCR
Total RNA was extracted from transfected HD11 cells
using Absolutely RNATM RT-PCR Miniprep kit (Strata-
gene). RNA (1–2 lg) was reverse transcribed with Super-
script II reverse transcriptase (Life Technologies). The
cDNA was diluted fivefold prior to PCR amplification
using primers specific for chicken mim-1 and chicken
GAPDH, respectively. Real-time PCR was performed on
a LightCycler rapid thermal cycler system (Roche Diag-
nostics) using the LightCycler FastStart DNA Master
SYBR Green I mix for amplification (Roche Diagnostics).
Reactions were performed in 20 lL with 0.5 l
M
primers
and 3 m
M
MgCl
2
. The amplification specificity of the PCR
products was confirmed by using melting curve analysis and
gel electrophoresis. We calculated the relative level of mim-1
mRNA as 100/E
(CP1–CP2)
, where CP1 and CP2 are crossing
FEBS 2003 Sumoylation of c-Myb (Eur. J. Biochem. 270) 1339
points for mim-1 and GAPDH mRNAs, respectively, and E
is the average efficiency of amplification obtained with the
same primer sets on a positive control template. The mRNA
levels of GAPDH were thus set at 100%.
Nuclear matrix preparation
CV-1 cells seeded out in 10 cm Petri dishes were transfected
with the indicated plasmids. Cells were harvested 24 h after
transfection in NaCl/P
i
and 30% were lysed directly in
loading buffer as a control for transfection. Nuclear matrix
samples and soluble fractions were prepared essentially as
described in [39].
Results
Bies et al. [17] have shown that murine c-Myb can be
sumoylated under overexpression conditions in COS7 cells
when cotransfected with FLAG-tagged SUMO-1. The
conjugation sites were mapped to the NRD region of the
protein. This work raised several questions that we have
addressed in a parallel study focusing on human c-Myb.
Several lines of independent evidence for sumoylation
of c-Myb
Our first interest was to find independent evidence for this
novel type of post-translational modification of c-Myb to
better establish its physiological relevance. In particular, we
were concerned by the overexpression conditions exclusively
used in the previous work on sumoylation of c-Myb [17].
We therefore initially performed a cotransfection experi-
ment similar to those reported by Bies et al.[17]but
replacing the COS cells with CV-1 cells, known to cause less
amplification of transfected plasmids than COS cells [40].
When CV-1 cells were cotransfected with constructs
expressing human c-Myb and GFP-tagged SUMO-1, two
retarded doublet bands were observed (Fig. 1A, lane 3).
These totally disappeared when the two putative sumoyla-
tion sites (in human c-Myb K503 and K527) were both
mutated (Ô2KR-MybÕ, Fig. 1A, lane 9). Single mutations
K503R and K527R had intermediary effects, with a strong
reduction with K527R and less effect with K503R where
only the upper doublet disappeared (Fig. 1A, lanes 7 and 5).
The same doublets of bands were seen in the control lanes 2
and 4 due to endogenous SUMO-1. Sumoylation at two
sites in c-Myb would be expected to generate two retarded
simple bands. Hence, the doublets probably represent
c-Myb with one and two conjugated SUMO-1 moieties,
respectively, combined with or without a second type of
modification (such as phosphorylation) affecting migration.
This confirms the observations of Bies et al. [17] under more
moderate conditions of overexpression.
To increase the stringency further we performed a similar
experiment in the absence of transfected SUMO-1 to see
whether endogenous levels of the peptide and its conjuga-
tion enzymes were sufficient to cause sumoylation. This
experiment was similar to what is shown in the control lanes
2, 4 and 6 in Fig. 1A but the use of a higher exposure allows
the effects of the mutants to be more evident. Again shifted
Myb-bands were observed in addition to the main 75 kDa
band (Fig. 1B, lane 2), although the mobility shifts now
were more modest, consistent with conjugation of untagged
SUMO-1. The K503R mutant caused the upper doublet to
disappear (Fig. 1B, lane 3). The K527R mutant caused a
much more important reduction in intensity of the slower
migrating forms (Fig. 1B, lane 4). In this mutant, only a
single additional band is seen, probably due to a less efficient
sumoylation of the remaining K503 site. Again, the 2KR
mutant showed no retarded bands (Fig. 1B, lane 5). To
Fig. 1. Human c-Myb is sumoylated in residues 503 and 527. (A) CV-1
cells transfected with the Myb-expressing plasmids as indicated, and in
addition with (+) or without (–) pGFP-SUMO-1. The Myb proteins
expressed were full-length human c-Myb (hcM) and c-Myb mutated in
lysine 503 (K503R) or 527 (K527R) or both (2KR). Cells were lysed
directly in loading buffer before separation on SDS/PAGE and
immunoblotting revealed by a monoclonal anti-Myb Ig (5E11). (B)
CV-1 cells transfected with empty pCIneo vector (v) or plasmids
expressing indicated Myb proteins as in (A). Cell lysates were subjected
to direct immunoblot with monoclonal anti-(c-Myb) Ig. (C) CV-1 cells
were transfected as in (B). Immunoprecipitation was performed with
monoclonal anti-SUMO-1 Ig (upper panel) and polyclonal anti-
(c-Myb) Ig (lower panel). After SDS/PAGE the blot was revealed by
mAb 5E11. (D) Cell lysates from Jurkat cells expressing endogenous
c-Myb, was subjected to immunoprecipitation with polyclonal
anti-HA Ig, polyclonal anti-(c-Myb) Ig and polyclonal anti-(SUMO-1)
Ig. After SDS/PAGE of the immunoprecipitates, immunoblot analysis
was performed using monoclonal anti-(c-Myb) Ig. The arrow indicates
the migration of unmodified c-Myb.
1340 Ø. Dahle et al.(Eur. J. Biochem. 270)FEBS 2003
verify that the observed modifications were indeed due to
SUMO-1 conjugation we performed at coimmunoprecipi-
tation experiment. While the lysate from CV-1 cells trans-
fected with wild type c-Myb contained modified Myb-forms
that became immunoprecipitated with the anti-SUMO-1 Ig
(Fig. 1C, lane 2), this was not the case with the 2KR mutant
(Fig. 1C, lane 3). This supports that wild type c-Myb
becomes conjugated with SUMO-1, and that this modifi-
cation is abolished in the 2KR mutant. Both variants of
c-Myb were equally expressed (Fig. 1C).
Having shown that c-Myb is sumoylated in CV-1 cells by
endogenous levels of the conjugation machinery, we finally
addressed whether the same was true for endogenous c-Myb
proteins in myb-positive cells. Therefore we carried out a
similar analysis in Jurkat cells. Immunoprecipitation of
c-Myb with polyclonal anti-(c-Myb) Ig, and detection with
monoclonal anti(-c-Myb) Ig revealed the main c-Myb band
at 75 kDa and several c-Myb species with higher molecular
mass (Fig. 1D, lane 2). Immunoprecipitation of sumoylated
proteins with polyclonal anti-(SUMO-1) Ig in the same
experiment revealed that at least one of these bands are
sumoylated c-Myb. This was also confirmed by immuno-
precipitation of c-Myb with polyclonal anti-(c-Myb) Ig and
detection with monoclonal anti-(SUMO-1) Ig, which
revealed one band with a size corresponding to c-Myb
conjugated with one SUMO-1 molecule (results not shown).
We conclude that the SUMO-1 conjugation c-Myb
observed by Bies et al. [17] under overexpression conditions
seems to be a robust phenomenon that also occurs under
more physiological conditions.
A second line of experiments further supported that
c-Myb is a good substrate for SUMO-1 conjugation. An
in vitro system for sumoylation was set up to investigate
sumoylation of c-Myb (Fig. 2). When in vitro translated
human c-Myb was incubated with an E1 fraction, GST-
UBC9 and GST-SUMO-1, two more slowly migrating
forms were generated with sizes corresponding to the
addition of one or two moieties of GST-SUMO-1, respect-
ively (+39 kDa and +78 kDa) (Fig. 2, lane 4). These
modified forms disappeared when either GST-UBC9 or
GST-SUMO-1 was omitted from the reaction mixture
(Fig. 2, lanes 3 and 5), strongly suggesting that they
correspond to c-Myb conjugated to SUMO-1 peptides.
Both retarded bands observed with the wild type protein
disappeared when the double mutant (2KR) was subjected
to in vitro sumoylation, demonstrating their function as
conjugation sites (Fig. 2, lanes 7 and 9). Consistent with the
location of K503 and K527 in a region that is deleted in
AMV v-Myb, an AMV v-Myb protein did not generate
retarded modified forms in this system (results not shown).
The two single mutants, K503R and K527R, and the 2KR
mutant were also subjected to in vitro sumoylation in the
context of a c-Myb fragment (amino acids 410–566, more
efficiently translated in vitro). When the conjugated forms of
the single mutants were compared, it was evident that the
two sites were not equivalent. While the K527R mutation
caused a sharp drop in sumoylation efficiency, requiring a
high input of UBC9 to become sumoylated on the
remaining site, the K503R protein was still efficiently
sumoylated at low inputs of UBC9 similar to wild type. This
strongly suggests that K527 is a much more efficiently
conjugated site than K503. It is also noteworthy that
bis-sumoylated wild type protein (modified in K503 and
K527) is formed as efficiently as mono-sumoylated (pre-
sumably mainly modified in K527), while mono-sumoylated
K527R protein (presumably modified in K503) is formed
with low efficiency. This suggests that K527-conjugation
enhances the efficiency of sumoylation at the other site.
A third line of independent evidence for sumoylation of
c–Myb is the interaction between c-Myb and Ubc9, the
latter acting as an E2-type SUMO-1 conjugase. Assuming
such an interaction, Bies et al. [17] performed a direct two-
hybrid test for this interaction between Ubc9 fused to
Gal4p-DBD and c-Myb domains fused to Gal4p-TAD.
Both fusion proteins were expressed from high-copy yeast
vectors. In an independent series of experiments we set up a
two-hybrid screen using full-length human c-Myb as bait
fused to Gal4p-DBD, but in our case expressed from a low-
copy CEN vector. Screening of 4 ·10
6
transformants from
two mixed cDNA Matchmaker libraries (human bone
marrow and human erythroleukemia K562 cell line)
resulted in the isolation of 23 triple-positive independent
clones. Three of these were identical to mRNA for
human ubiquitin-conjugating enzyme UBC9 (Accession
no. AJ002385). Retransformation and growth on reporter-
selective media (not shown) verified the Myb–UBC9
interaction, and by determination of reporter activation
using both a 5-bromo-4-chlorindol-3-yl b-
D
-galactoside
overlay and a liquid b-galactosidase assay (Fig. 3). Similar
analysis of several subdomains of c-Myb revealed strongest
subdomain interaction with the EVES-domain in the NRD-
region of c-Myb, suggesting that this region might be
involved in the UBC9 interaction (results not shown). These
two-hybrid results show that Ubc9 is amongst the strongest
interaction partners of c-Myb as judged by a low-copy bait
screening in a cDNA library containing 2 million inde-
pendent clones, lending further support to the importance
of the c-Myb–Ubc9 interaction.
We conclude that SUMO-1 conjugation of c-Myb is not
only a phenomenon induced under favourable conditions of
overexpression of c-Myb and SUMO-1, but a robust
modification caused by a strong interaction between
c-Myb and Ubc9. This leads to modification at two residues
in the NRD part of the protein with K527 being the major
sumoylation site. The conjugation of SUMO-1 to c-Myb
raises the question of the role of this modification with
respect to the transcriptional activity of c-Myb.
Disruption of the SUMO-1 acceptor sites in c-Myb
causes a superactivation phenotype
Bies et al. [17] observed that c-Myb mutated in one of the
sumoylation sites was more active than wild type Myb in an
effector-reporter assay under overexpression conditions in
COS7 cells. To confirm this observation in CV-1 cells and
to extend the analysis to clarify the relative functional
importance of the two conjugation sites, we compared
reporter activation induced by the individual mutants (K503
and K527), the double mutant (2KR) and wild type c-Myb
using a reporter with multimerized Myb response elements
(Fig. 4A). While full-length c-Myb caused a modest level of
reporter activation (1.3-fold relative to empty effector), the
K503R mutant was slightly more active (3.6-fold), the
K527R mutant significantly more active (9.6-fold) and
FEBS 2003 Sumoylation of c-Myb (Eur. J. Biochem. 270) 1341
finally the double mutant c-Myb-2KR gave rise to a 23-fold
increase in reporter activity, which is 17-fold higher than
wild type c-Myb. A Western blot confirmed that all Myb
variants were equally expressed (Fig. 4A). It is noteworthy
that a clear correlation seems to exist between the increase in
transcriptional activity of the individual mutants (Fig. 4A)
and the reduction in their degree of sumoylation (Fig.
1A,B).
Because effector-reporter assays in transfected cells is a
method with recognized limitations, we wanted to see
whether the conjugation sites influenced transcriptional
activity in a more physiological setting and therefore tested
activation of the resident mim-1 gene using the HD11 cell
line, an established Myb-model [36]. The mim-1 target gene
is only activated by c-Myb, not by v-Myb, when residing in
its chromosomal locus because v-Myb has lost the ability to
Fig. 2. Human c-Myb is sumoylated by UBC9 in vitro. (A) In vitro translated
35
S-labelled full-length c-Myb was incubated in the presence (+) or in
the absence (–) of the indicated components described in Materials and methods (lanes 1–5). Single sumoylated (1·Sumo) and bis-sumoylated (2·
Sumo) Myb, respectively. In lanes 6–9 full-length c-Myb (hcM) and c-Myb mutated at K503 and K527 (2KR) were compared in the presence (+)
or absence (–) of the full set of sumoylation components. (B) Wild type or mutant (K503R and K527R) subdomains of human c-Myb (TpC
fragments, amino acids 410–639) were
35
S-labelled in vitro and subjected to sumoylation as in 2A, but with variable limiting amounts of GST-UBC9
as indicated (given as ng of UBC9 only). The amount of sumoylated Myb species were quantified by the
NIH IMAGE
1.62 software (upper panel) and
the different sumoylated Myb species measured are shown in the lower panel. ÔWt bis-SÕand Ôwt mono-SÕrepresent double and single sumoylated
wild type TpC c-Myb, respectively; ÔK503R mono-SÕ, single sumoylated K503R TpC c-Myb; ÔK527R mono-SÕ, single sumoylated K527R TpC
c-Myb.
1342 Ø. Dahle et al.(Eur. J. Biochem. 270)FEBS 2003