The Drosophila homeodomain transcription factor, Vnd,
associates with a variety of co-factors, is extensively
phosphorylated and forms multiple complexes in embryos
Huanqing Zhang
1,
*, Li-Jyun Syu
1,
, Vicky Modica
1
, Zhongxin Yu
1
, Tonia Von Ohlen
2
and
Dervla M. Mellerick
1
1 MSRBIII, Ann Arbor, MI, USA
2 Division of Biology, Kansas State University, Manhattan, KS, USA
The Drosophila homeodomain transcription factor,
Vnd, is the founding member of the Nk class of home-
odomain proteins, which are both necessary and suffi-
cient to specify ventral central nervous system (CNS)
cell fates in Drosophila embryos [1,2]. Drosophila neu-
roblasts are arranged in three columns, ventral, inter-
mediate and lateral, with the ventral cells juxtaposed
at either side of the ventral midline. Ventral column
identity is transformed to intermediate column identity
in vnd mutant embryos, whereas ectopic Vnd expres-
sion ventralizes the neuroectoderm [1,2]. Both mutant
analyses [1,3–5] and biochemical data [6,7] indicate
that Vnd functions as a dual regulator, capable of both
activating and repressing target gene expression. Vnd
interacts with the co-repressor, Groucho, to repress
target gene expression, whereas Vnd’s interaction with
the high mobility group (HMG) protein, Dichaete,
facilitates the activation of target genes [6,8]. However,
co-transfection analyses indicate that additional
unidentified co-factors are required to recapitulate the
robust regulatory effects of Vnd evident in embryos.
In this report, we further explore the biochemical
basis for Vnd’s capacity to function as a dual regula-
tor. We identify Drosophila Kc167 cells as an endoge-
nous source of Vnd and show that the co-repressor,
Groucho, physically interacts with Vnd in these cells.
Keywords
co-factor; complex; regulation; transcription;
Vnd
Correspondence
D. M. Mellerick, Med Sci I, M5240, Ann
Arbor, MI 48109-0602, USA
Fax: +1 734 369 3874
Tel: +1 734 709 3222
E-mail: dervlamellerick@yahoo.com
Present address
*5051 BSRB, Ann Arbor, MI, USA
1500 E. Medical Center Drive, Ann Arbor,
MI, USA
(Received 28 April 2008, revised 23 July
2008, accepted 12 August 2008)
doi:10.1111/j.1742-4658.2008.06639.x
Vnd is a dual transcriptional regulator that is essential for Drosophila dor-
sal–ventral patterning. Yet, our understanding of the biochemical basis for
its regulatory activity is limited. Consistent with Vnd’s ability to repress
target expression in embryos, endogenously expressed Vnd physically asso-
ciates with the co-repressor, Groucho, in Drosophila Kc167 cells. Vnd exists
as a single complex in Kc167 cells, in contrast with embryonic Vnd, which
forms multiple high-molecular-weight complexes. Unlike its vertebrate
homolog, Nkx2.2, full-length Vnd can bind its target in electrophoretic
mobility shift assay, suggesting that co-factor availability may influence
Vnd’s weak regulatory activity in transient transfections. We identify the
high mobility group 1-type protein, D1, and the novel helix–loop–helix
protein, Olig, as novel Vnd-interacting proteins using co-immunoprecipita-
tion assays. Furthermore, we demonstrate that both D1 and Olig are
co-expressed with Vnd during Drosophila embryogenesis, consistent with a
biological basis for this interaction. We also suggest that the phosphoryla-
tion state of Vnd influences its ability to interact with co-factors, because
Vnd is extensively phosphorylated in embryos and can be phosphorylated
by activated mitogen-activated protein kinase in vitro. These results high-
light the complexities of Vnd-mediated regulation.
Abbreviations
CNS, central nervous system; EMSA, electrophoretic mobility shift assay; Ets, erythroblast transformation-specific; HLH, helix–loop–helix;
HMG, high mobility group; MAP, mitogen-activated protein; UAS, upstream activating sequence.
5062 FEBS Journal 275 (2008) 5062–5073 ª2008 The Authors Journal compilation ª2008 FEBS
We show that Vnd exists as a complex in both Kc167
cells and in multiple high-molecular-weight complexes
in embryos. We eliminate target DNA binding as the
bottleneck in Vnd’s incapacity to exert robust regula-
tory effects in tissue culture cells, as full-length Vnd
from either transient transfections or bacteria can bind
its target in electrophoretic mobility shift assays (EM-
SAs). These observations suggest that the phosphoryla-
tion state of Vnd may impinge on Vnd-mediated
regulation. We demonstrate that Vnd exists as multiple
isoforms in embryos, and that phosphorylation con-
tributes to this diversity. We also confirm that the
AT-hook protein, D1, physically interacts with Vnd,
and show that D1 is co-expressed with Vnd in
embryos. In addition, we show that, like its vertebrate
homolog, Nkx2.2, Vnd physically interacts with the
unique Drosophila Olig homolog, and that Vnd and
Olig are co-expressed in at least one neuroblast and
many neuroblast progeny. The significance of these
findings is discussed.
Results
This study was prompted by previous findings indicat-
ing that transiently expressed Vnd is relatively inert in
both Drosophila S2 and vertebrate Hek 293 cells [6,7],
despite using enhancers that are regulated by Vnd
activity in embryos [9,10] to monitor regulatory activ-
ity. Co-expression of the co-regulators, Groucho and
Dichaete, which both physically interact with Vnd,
also generates weak readouts [6], despite the fact that
both of these co-regulators modulate Vnd activity in
Drosophila embryos [3,11]. Thus, unidentified co-fac-
tors, additional to the co-regulators so far identified,
appear to be necessary for Vnd to exert the robust reg-
ulatory effects evident in Drosophila CNS develop-
ment. Here, we explore the nature of these co-factors.
Drosophila Kc167 cells co-express Vnd
and Groucho
In an attempt to identify an easily accessible source of
Vnd for biochemical analyses, we screened a number
of different cell lines for endogenous Vnd expression
using immunoprecipitation analyses (data not shown).
Using our Vnd antibody for immunoprecipitations, we
identified Drosophila Kc167 cells as an endogenous
source of Vnd. Transfection of Kc167 cells with a
luciferase reporter, driven by either the 3¢ind enhancer
through which Vnd represses ind expression in
embryos [10] or the 5¢vnd enhancer through which
Vnd mediates auto-activation [12], resulted in the
repression of both reporters (data not shown), suggest-
ing that these cells probably co-express factors that are
required for Vnd-mediated repression but not activa-
tion. The association of Vnd with the co-repressor,
Groucho, mediates Vnd’s repressor capacity in
embryos [8]. Thus, we questioned whether Vnd physi-
cally interacts with this protein in Kc167 cells by deter-
mining whether endogenous Vnd can pull down
Groucho from these cells. Figure 1 shows that immuno-
precipitated Vnd pulls down endogenous Groucho,
suggesting that Vnd in Kc167 cells may be a good
source of Vnd for further biochemical analyses.
Vnd forms high-molecular-weight complexes in
Kc cells and in embryos
To explore this possibility further, we compared the
physical state of Vnd in Kc167 cells with that in
embryos by determining how both Vnd sources frac-
tionate on sucrose gradients. To determine the physical
distribution of Vnd, nuclear fractions from Kc167 cell
and embryonic lysates were separated by SDS-PAGE,
western blotted and incubated with a Vnd-specific anti-
body [13], following centrifugation on 0–30% sucrose
gradients. Recombinant full-length Vnd with a His tag
fractionates at approximately 77 kDa, based on the
migration of labeled molecular weight markers loaded
on parallel sucrose gradients, whereas Vnd from
Kc167 cells fractionates at approximately 230 kDa
(Fig. 2); this indicates that Vnd exists as a low-molecu-
lar-weight complex in these cells.
To examine how Vnd physically partitions from
embryos, Vnd levels were enriched by over-expression
using the upstream activating sequence (UAS)–Gal4
system [14] under the control of the Scabrous enha-
ncer. This driver directs over-expression in all neuro-
Fig. 1. Endogenous Vnd in Drosophila Kc167 cells associates with
the co-repressor Groucho. Immunoprecipitations, performed using
an anti-Vnd antibody (lanes 2 and 4) or preserum (lanes 1 and 3),
were western blotted and incubated with an anti-Vnd antibody
(lanes 1 and 2) or an anti-Groucho (Gro) antibody (lanes 3 and 4).
Immunoprecipitated Vnd (lane 2) pulls down Gro (lane 4). Thus,
endogenous Vnd physically associates with this co-repressor in
Kc167 cells.
H. Zhang et al. Biochemical studies on the Vnd transcription factor
FEBS Journal 275 (2008) 5062–5073 ª2008 The Authors Journal compilation ª2008 FEBS 5063
ectodermal cells from stage 10 onwards. Embryonic
Vnd generates a number of independent peaks of high
molecular weight (Fig. 2) on sucrose gradients, indicat-
ing that Vnd is present in a variety of complexes in
embryos, coincident with Vnd’s diverse roles in CNS
development [2,13]. Thus, although Kc167 cells
co-express Vnd and Groucho, they only partially
mimic the microenvironment of Vnd in embryos. Thus,
we did not consider them to be an ideal source for
further biochemical dissection of Vnd.
Vnd is extensively phosphorylated in embryos,
and can be phosphorylated in vitro by activated
mitogen-activated protein (MAP) kinase
Unpublished transient transfection competition experi-
ments to determine the basis for the selective interac-
tion of Vnd with either Dichaete or Groucho [6]
indicate that, when Vnd is bound to Dichaete, the
addition of Groucho fails to displace Dichaete, and
vice versa (data not shown). As Vnd’s activity is mod-
ulated by epidermal growth factor signaling [11,15]
and activated MAP kinase co-localizes with Vnd in
embryos [15,16], the phosphorylation of Vnd is likely
to be important in its role as a dual regulator. How-
ever, to date, the question of whether Vnd is modified
post-translationally or is a target of activated MAP
kinase has not been addressed directly. Vnd has five
candidate MAP kinase phosphorylation sites,
32
PASP,
43
PSSPATP,
157
PWSP and
419
PASP, that conform to
the consensus phosphorylation site PXS ⁄TP [17].
Moreover, the essential Pro and Ser ⁄Thr are conserved
between Drosophila melanogaster and D. virilis
(Fig. 3B). Thus, post-translational modification of Vnd
may modulate the differential affinity of this transcrip-
tion factor for embryonic co-factors that are simulta-
neously co-expressed with Vnd, and may modulate
Vnd’s capacity to function as either an activator or
repressor.
To address directly whether Vnd is post-transla-
tionally modified in embryos, we immunoprecipitated
this transcription factor from Scabrous-Gal4 ·UAS-
vnd embryos and subjected the immunoprecipitate to
two-dimensional gel electrophoresis and western blot-
ting analyses. Our Vnd antibody identified multiple
isoforms of Vnd (Fig. 3A), which are at least par-
tially a result of the different phosphorylation states
of the protein, as phosphatase treatment of the
immunoprecipitate generated many fewer isoforms of
Vnd (Fig. 3A). We also addressed the question of
whether Vnd can be phosphorylated in vitro by MAP
kinase. A His-tagged recombinant Vnd peptide
encompassing the five candidate Map kinase target
sites (Fig. 3B) was purified from bacteria and assayed
as phosphorylation substrate for activated MAP
kinase. For these experiments, we utilized the verte-
brate homolog of Rolled, the Drosophila MAP
kinase. These two kinases have identical active sites,
and therefore should phosphorylate targets with the
same specificity. Our results show that recombinant
Vnd is phosphorylated by activated MAP kinase
in vitro (Fig. 3C). Thus, phosphorylation of Vnd
probably plays a role in Vnd’s ability to function as
a dual regulator, potentially by affecting its selective
interaction with co-factors that mediate its opposing
regulatory activities.
Fig. 2. Vnd is present in Kc167 cells and in
embryos as high-molecular-weight com-
plexes. Top: western blots of sucrose gradi-
ent fractions prepared from nuclear extracts
of Drosophila Kc167 cells (top panel) or
UAS-Vnd ·Kruppel Gal4 embryos (bottom
panel) incubated with a Vnd antibody. Vnd
from Kc167 cells fractionates in a 0–30%
sucrose gradient at approximately 230 kDa,
whereas Vnd from embryos generates at
least four high-molecular-weight complexes
of 250, 300, 600 and more than 800 kDa.
Arrows indicate the location of recombinant
Vnd (77 kDa), lactose dehydrogenase
(140 kDa), catalase (232 kDa) and ferritin
(440 kDa), which were fractionated on paral-
lel gradients simultaneously. Bottom: densi-
tometer readings of the blots shown above.
Biochemical studies on the Vnd transcription factor H. Zhang et al.
5064 FEBS Journal 275 (2008) 5062–5073 ª2008 The Authors Journal compilation ª2008 FEBS
Full-length Vnd can bind its target in EMSAs
Watada et al. [18] reported previously that Nkx2.2, the
mouse homolog of Vnd, has a carboxyl terminal DNA
binding interference domain, which must be removed
(or possibly inactivated in vivo) for efficient target rec-
ognition [18]. If Vnd’s capacity to bind target DNA is
similarly regulated, this might explain why this tran-
scription factor is such an inefficient regulator in S2
cells and Hek 293 cells [6,7]. We have shown previ-
ously that the recombinant Vnd homeodomain binds
three Vnd target sites in the ind enhancer, and the Vnd
binding sites were further localized by footprinting
[10]. However, as yet, we have not determined whether
full-length Vnd can bind target DNA, and whether
post-translational modification of Vnd affects its
capacity to bind its target.
To address this question, we performed EMSAs
using either the recombinant full-length Vnd or Vnd
transiently expressed in Hek 293 cells, and thus poten-
tially post-translationally modified. We used an oligo-
nucleotide (oligo) that corresponds to the binding sites
in the 3¢ind enhancer [10] (through which Vnd
represses ind expression in embryos) as target
(Fig. 4A). When we compared the binding of Vnd
from transiently transfected Hek 293 cell lysates with
that of purified recombinant full-length Vnd, we found
that Vnd from both sources caused mobility shifts,
paralleling our previous finding using the Vnd homeo-
domain [10]. Thus, binding of Vnd to its target is
unlikely to be the bottleneck in Vnd’s inability to
robustly regulate target expression in the tissue culture
cells tested thus far. The availability of the essential
co-factors required for Vnd regulation is potentially
the factor limiting Vnd’s capacity to regulate target
gene expression in tissue culture cells.
Vnd physically associates with the AT-hook
protein, D1, and the helix–loop–helix (HLH)
protein, Olig
As unidentified co-factors in addition to Dichaete and
Groucho are likely to be required for Vnd’s capacity
to either activate or repress target gene expression, we
next explored other co-factors that may impinge on
Vnd’s regulatory activity. We selected two proteins,
D1 and Olig, as candidate Vnd co-regulators. D1 was
identified in a genome-wide two-yeast hybrid interac-
tion screen as a novel Vnd-interacting protein [19].
However, this reported interaction has not been con-
firmed using more rigorous interaction assays. Olig
was considered as a second probable Vnd-interacting
protein, as the Vnd homolog, Nkx2.2, both physically
and genetically interacts with the vertebrate HLH pro-
tein, Olig2, to influence neural ⁄glial cell subtype speci-
fication [20–22]. Thus, we hypothesized that either
Drosophila D1 and ⁄or Olig may bind Vnd and impinge
on its activities as a dual regulator.
To determine whether Vnd could immunoprecipitate
either of these transcription factors, we transiently
expressed Vnd with an amino terminal Gal4 domain in
Hek 293 cells (which were selected because of their
high transfection efficiency), and immunoprecipitated
Fig. 3. Vnd is extensively phosphorylated in embryos, and can be
phosphorylated by activated MAP kinase in vitro. (A) Western blots
of two-dimensional gels of Vnd immunoprecipitated from Drosoph-
ila embryos detected using a Vnd antibody. The sample was
divided: that on the left was untreated, whereas that on the right
was treated with alkaline phosphatase. The presence of multiple
Vnd reactive spots indicates that Vnd is extensively post-translation-
ally modified in embryos (left panel). The phosphatase-treated sam-
ple on the right shows significantly fewer bands, indicating that
phosphorylation contributes to the multiple isoforms of Vnd. Molec-
ular weight markers are as in Fig. 2. The band at approximately
50 kDa is the IgG heavy chain isoform. (B) Schematic diagram of
D. melanogaster Vnd showing the location of the five candidate
MAP kinase phosphorylation sites (in green, labeled 1–5) in the full-
length protein, and the N-terminal Vnd peptide used in (C). The
positions of the Eh domain (black), the homeodomain (red) and the
Nk-2 box (yellow) are highlighted. The candidate MAP kinase target
sites of Vnd from D. melanogaster and D. virilis are enlarged.Note
that the essential proline (P) and serine (S) or threonine (T) residues
are conserved. (C) Top: SDS-PAGE of Vnd (lanes 1–5) used as
in vitro phosphorylation substrate, as well as negative (lane 6) and
positive (lane 7) controls. Bottom: autoradiograph of in vitro phos-
phorylation time course using activated MAP kinase, showing
increasing incorporation of
32
P into Vnd with increasing time. The
negative control is not phosphorylated, whereas the positive con-
trol, PHAS-1, incorporates significant amounts of
32
P after 30 min.
H. Zhang et al. Biochemical studies on the Vnd transcription factor
FEBS Journal 275 (2008) 5062–5073 ª2008 The Authors Journal compilation ª2008 FEBS 5065
it using the Gal4 antibody. D1 and Olig were also
independently expressed, and cell lysates containing
either factor were incubated with the Vnd immunopre-
cipitate. To express D1 and Olig, both candidate
co-factors were cloned into an expression vector carry-
ing an amino terminal Flag tag (see Materials and
methods for details). As positive controls, Dichaete
expressed with a Flag tag and Groucho with a Myc
tag were also tested for Vnd interaction, as we have
demonstrated previously that Vnd can immunoprecipi-
tate both of these co-regulators [6].
The Vnd immunoprecipitate was split and incubated
with lysates from either nontransfected cells or cells
that transiently expressed Flag-tagged D1, Olig or
Dichaete, or Myc-tagged Groucho. The co-precipitates
were then size separated by SDS-PAGE, and western
blotted. To identify which proteins could be pulled
down by Vnd, blots were incubated with either an
anti-Flag or an anti-Myc tag antibody. It was shown
that Vnd physically associates not only with Myc-
tagged Groucho and Flag-tagged Dichaete, but also
with Flag-tagged D1 and Flag-tagged Olig (Fig. 5).
Thus, these co-immunoprecipitation analyses verify
that Vnd physically interacts with the AT-hook pro-
tein, D1, and identify the unique HLH protein, Olig,
Fig. 4. Full-length Vnd can bind its in vivo target. (A) Oligo used in
(B), corresponding to the Vnd target in the ind enhancer, which
contains four Vnd binding sites (underlined). (B) Lanes 1–3: EMSA
using recombinant full-length Vnd. Lanes 4–6: EMSA using
Hek 293 cell extracts. Lane 4 is the cell extract containing the
empty vector. Lane 5 is the Vnd-containing cell extract. Lane 6 cor-
responds to lane 5 with excess unlabelled probe. The negative con-
trol in lane 1 corresponds to the probe without protein. Incubation
of recombinant Vnd with labeled oligo generates a mobility shift
(lane 3) when large levels are used. Note that lane 2 contains five-
fold less recombinant protein than lane 3. Incubation of Vnd in cell
extract also generates a mobility shift (lane 5). Incubation of the cell
extract with excess unlabelled oligo interferes with binding to
labeled oligo (lane 6).
Fig. 5. Two new Vnd-interacting proteins: D1 and Olig. Western
blots of immunoprecipitates incubated with an anti-Gal4 antibody,
which detects Vnd with a Gal4 DNA binding domain tag or Gal4
DNA binding domain (DBD) alone (A), anti-Flag, which detects Flag-
tagged D1 and Dichaete, or anti-Myc, which detects Myc-tagged
Groucho (B). Bands of interest are highlighted with arrows. Lane 1:
Gal4-DBD+ Olig. Lane 2: Gal4-DBD-Vnd+ Olig. Lane 3: Gal4-DBD+
D1. Lane 4: Gal4-DBD-Vnd+ D1. Lane 5: Gal4-DBD+ Dichaete.
Lane 6: Gal4-DBD-Vnd+ Dichaete. Lane 7: Gal4-DBD+ Groucho.
Lane 8: Gal4-DBD-Vnd+ Groucho. (B) Gal4 DBD does not interact
with either antibody, indicating that there is no nonspecific binding.
In contrast, the presence of Vnd leads to Olig (lane 2) and D1 (lane
4) being pulled down. As reported previously, Dichaete (lane 6) and
Groucho (lane 8) are also pulled down by Vnd.
Biochemical studies on the Vnd transcription factor H. Zhang et al.
5066 FEBS Journal 275 (2008) 5062–5073 ª2008 The Authors Journal compilation ª2008 FEBS