Hox homeodomain proteins exhibit selective complex stabilities with Pbx and DNA
Hox homeodomain proteins exhibit selective complex stabilities with Pbx and DNA
Wei-Fang
Shen
,
Ching-Pin
Chang
,
Sofia
Rozenfeld
,
Guy
Sauvageau
2
,
R. Keith Humphries
2
,
Ming
Lu
3
,
H. Jeffrey
Lawrence
,
Michael L.
Cleary
1
and
Corey
Largman*
Department of Medicine San Francisco Veterans Affairs Medical Center and
University of California,
San Francisco
, CA 94121,
USA
,
1
Department of Pathology, Stanford University Medical Center,
Stanford
, CA 94305,
USA
,
2
Terry Fox Laboratory, British Columbia Cancer Agency and the Department of
Medicine, University of British Columbia,
Vancouver
,
Canada
and
3
Cancer Center and Department of Medicine, University of California,
San Diego
, CA,
USA
Received November 1, 1995;
Revised and Accepted January 7, 1996
Abstract
Eight of the nine homeobox genes of the
Hoxb
locus encode proteins which contain a conserved hexapeptide motif upstream from
the homeodomain. All eight proteins (Hoxb-1-Hoxb-8) bind to a target oligonucleotide in the presence of Pbx1a
under conditions where minimal or no binding is detected for the Hox or Pbx1a
proteins alone. The stabilities of the Hox-Pbx1a-DNA complexes vary >100-fold, with the proteins from the middle of the locus (Hoxb-5 and Hoxb-6) forming very stable complexes, while Hoxb-4, Hoxb-7 and Hoxb-8 form complexes of
intermediate stability and proteins at the 3
'
-side of the locus (Hoxb-1-Hoxb-3) form complexes which are very unstable. Although Hox-b proteins containing longer linker sequences
between the hexapeptide and homeodomains formed unstable complexes, shortening the linker did not confer complex stability. Homeodomain swapping experiments revealed that this motif does not independently determine complex stability. Naturally occurring variations within the hexapeptides of specific Hox proteins also do not explain complex stability differences. However, two core amino acids (tryptophan and methionine) which are
absolutely conserved within the hexapeptide domains appear to be required for complex formation. Removal of N- and C-terminal flanking regions did not influence complex stability and the members of paralog group 4 (Hoxa-4, b-4, c-4 and d-4), which share highly conserved
hexapeptides, linkers and homeodomains but different flanking regions, form complexes of similar stability. These data suggest that the
structural features of Hox proteins which determine Hox-Pbx1a-DNA complex stability reside within the precise structural relationships between the homeodomain, hexapeptide and linker regions.
INTRODUCTION
The
Hox
homeobox-containing genes appear to regulate regional specification along the A-P axis and in the limb bud of the developing embryo (
1
,
2
). The distinctive morphogenetic role of these genes has been most clearly
demonstrated by the different phenotypic changes observed in mice carrying a
range of targeted disruptions of different
Hox
genes (
3
-
9
). The homeodomain has been shown to be a DNA binding motif (reviewed in
10
) and Hox proteins are thought to function as transcription factors (
11
,
12
). However, despite markedly divergent biological effects
in vivo
, the homeodomains of many different Hox
proteins can bind
in vitro
to similar or identical DNA sequences containing a TAAT core recognition sequence with similar affinities (
13
,
14
). Indeed, the highest degree of sequence conservation between different Hox
proteins is in helix three of the homeodomain (
15
), which appears to bind in the major groove of the target DNA and to contribute
to binding specificity (
16
-
18
). These observations raise the question of how Hox proteins mediate selective
gene regulation.
The 38 human or murine Hox
homeodomain proteins, arrayed within four genetic loci, can be aligned in so-called paralog groups based on the degree of amino acid homology within
the homeodomain (
15
,
19
). Outside the homeodomain there is variable sequence conservation among paralog
members and relatively little homology between proteins across a single locus. An exception to this general rule is the presence of a hexapeptide motif,
containing a relatively conserved YPWM core sequence and variably located 5-53 amino acids N-terminal of the homeodomain, which is shared by all of the Hox
proteins in paralog groups 1-8 and is absent in groups 9-13 (
15
). Although X-ray crystallographic studies which have defined the binding of the
homeodomain to DNA have not used proteins containing the hexapeptide (
16
,
17
), NMR studies have shown that the hexapeptide and linker sequences are unstructured and do not
appear to make direct DNA binding contacts (
18
). Much of the data on DNA binding by Hox proteins has been obtained using only
the 61 amino acid homeodomain (
13
,
14
). These experiments have shown that the highly variable N-terminal portion of the homeodomain can contribute incrementally to DNA binding selectivity and affinity. However, there is little data on
contributions to DNA binding by the flanking regions of the full-length Hox proteins. In one experiment the full-length Ubx-Ib protein and the corresponding homeodomain bound the same
consensus DNA binding site, suggesting that under the conditions studied the
hexapeptide motif did not directly influence DNA binding (
13
).
One proposed mechanism for generating binding specificity and increasing
affinity is the formation of multi-protein-DNA complexes. The best example of homeodomain protein
cooperativity is the interaction of yeast mating type proteins [alpha]2 and a1, for which both genetic and biochemical data have been obtained
for cooperative binding to the haploid-specific gene operator (
20
,
21
). Following genetic data which indicated that the
Drosophila
Ubx homeodomain protein acts in concert with the Exd homeoprotein, several
laboratories have demonstrated that these proteins form cooperative complexes
with DNA
in vitro
(
22
,
23
).
Pbx1
is a non-Hox homeobox-containing proto-oncogene which was initially discovered at the site of t(1,13)
chromosomal translocations in pre-B cell leukemias (
24
,
25
). The Pbx proteins share a high degree of sequence homology with the Exd and
Caenorhabditis elegans
ceh-20 proteins (
26
,
27
). We (
28
) and others (
29
,
30
) have recently shown that Hox homeodomain proteins cooperatively bind with Pbx
proteins to DNA targets. In our studies three homeodomain proteins from the
human
Hoxb
locus, Hoxb-4, b-6 and b-7, cooperatively bound with Pbx1a protein to an
oligonucleotide containing Hox and Pbx DNA recognition sites under conditions
where there was minimal DNA binding by either the full-length Hox or Pbx1a proteins alone. These studies also demonstrated that
the conserved hexapeptide motif and the homeodomain were required for
cooperative binding with Pbx1a. Introduction of the Hoxb-7 hexapeptide together with the 13 amino acid linker preceding the
homeodomain into Hoxa-10 protein, which lacks this motif, substantially increased cooperative
binding to Pbx1a/DNA.
In the current study the stabilities of complexes formed by the eight Hoxb
proteins which contain a hexapeptide motif with Pbx1a and an oligonucleotide
target have been determined. There is a >100-fold difference in dissociation rates across the locus, with the Hoxb-5 and Hoxb-6 proteins forming the most stable complexes with Pbx1a on an
oligonucleotide containing consensus binding sites for both proteins. Data from
mutagenesis and domain swapping studies suggest that within each Hox protein
the hexapeptide, the linker sequence and the homeodomain function as a unit to mediate cooperative DNA binding.
MATERIALS AND METHODS
Protein expression
cDNAs encoding individual full-length Hox proteins, tcl-3 and Pbx1 were subcloned into either a derivative of sp65 (Promega,
Madison, WI) which was engineered to contain a FLAG sequence (MDYKDDDDK) (Hoxb-2, b-3, b-4, b-6, b-7, tcl-3 and Pbx 1a), pSG5, which contains a T7
promoter (
19
) (Hoxb-1), Bluescript (Stratagene, La Jolla, CA), which contains a T7 promoter
(Hoxb-5) or into a pET vector (Novagen, Madison, WI) containing the T7 promoter
and His sequences (MGSSHHHHHH-SSGLVPRGSHMASMTGGQQMGR) (Hoxa-4, b-8, c-4 and d-4). Full-length Hox
proteins fused to the respective peptides were synthesized using the TNT coupled
in vitro
transcription/translation system (Promega, Madison, WI), in both the presence
and absence of [
35
S]methionine. The labeled proteins were electrophoresed separately to
demonstrate synthesis of the appropriate full-length product and to ensure that similar amounts of proteins were used
for complex formation. For each synthesis reaction a control containing
reticulocyte lysate and the appropriate viral polymerase was used to detect
possible DNA binding by endogenous factors. As shown in the figures, lysate
controls show variable intensity gel shift bands with the DNA target. These
bands varied with both the lysate batch and the batch of poly(dI[middot]dC) used as non-competitive inhibitor. Human Hoxb-6 and b-7 have been cloned previously (
39
,
40
). Variant forms of Hoxb-3 were cloned from either a human CD34
+
bone marrow library or by PCR amplification of mRNA from the human
erythroleukemia cell line K562. The other human full-length cDNA clones were kindly provided as follows: Hoxb-1 (
15
), b-2 (
15
), b-4 (
41
) and c-4 (
42
), Dr E. Boncinelli; Hoxb-5 (
43
), Dr C. Hauser. Full-length murine clones were provided as follows: Hoxa-4 (
44
), Dr D. Wolgemuth; Hoxb-8 (
45
), Dr S. Cory; Hoxd-4 (
46
), Dr G. Firestone.
DNA constructs
Specific amino acids within the YPWM sequence of Hoxb-4 were altered by insertion of synthetic double-stranded oligonucleotides containing the desired change and flanked
by
Acc
I and
Pml
I sites, which were ligated into the
Acc
I (463) and
Pml
I (487) sites in the original clone. The distance between the YPWM motif and the
homeodomain in Hoxb3 was reduced from 53 to either 17 amino acids, consisting
of the first 12 and last five residues in the linker region (Table
2
), or to a linker consisting of the first 12 residues by PCR amplification and
conventional cloning methods. Replacement of the Hoxb-3 homeodomain with that of Hoxb-6 and the Hoxb-6 homeodomain with that of Hoxb-1 were accomplished using PCR to generate specific
restriction sites allowing precise replacement of each homeodomain using
conventional cloning methods. Each mutation was checked by DNA sequence
analysis and mutant proteins were shown to have the correct size by gel
electrophoresis of the
in vitro
translation products generated in the presence of [
35
S]methionine.
Electrophoretic mobility shift assays
Complementary oligonucleotides which contained consensus binding sites for Hox
proteins (
18
) and Pbx1 (
47
) were synthesized (Operon Technologies, Alameda, CA). The orientation of the
Hox and Pbx sites to each other within the oligonucleotide sequence was
selected on the basis of the orientation of sites in the yeast haploid-specific operator (
48
). This DNA fragment was previously used to demonstrate cooperative binding of
Hoxb-4, b-6 and b-7 to DNA in the presence of Pbx1a (
28
). The standard conditions used were similar to those previously described.
Briefly, double-stranded end-labeled DNA (50 000 c.p.m./binding reaction, ~1 ng) was incubated with 2 [mu]l test Hox protein (<0.5 ng) either in the presence of Pbx1a (2 [mu]l) or with 2 [mu]l lysate control, in 75 mM NaCl, 1 mM EDTA, 1 mM
dithiothreitol, 10 mM Tris-HCl, pH 7.5, 6% glycerol, 2 [mu]g bovine serum albumin and 2 [mu]g poly(dI[middot]dC) as non-specific competitor, in a final reaction volume of
15 [mu]l. In reactions performed under reduced stringency conditions the
concentration of poly(dI[middot]dC) non-specific inhibitor was reduced to 16 ng/15 [mu]l reaction and the concentration of NaCl was decreased to 50
mM. Reaction mixtures were incubated on ice for 30 min and either applied
directly to a 6% polyacrylamide gel (zero time sample) or mixed with a 100-fold excess of unlabeled oligonucleotide followed by incubation for fixed
times (1-35 min) prior to application to the running polyacrylamide gel. Gel electrophoresis was performed in 0.25* TBE buffer as described previously. For the Hoxb-6/b-1 box complex, which migrates to a position very close
to non-specific gel shift bands attributed to lysate proteins, a supershift using
antiserum to the FLAG-tag protein (IBI, New Haven, CT) was performed to confirm the identity of
the complex (Fig.
4
, lane 15).
Calculation of complex half-lives
Gels from electrophoretic mobility shift assays were dried and autoradiographed
to permit densitometric estimation of complex band intensities by scanning
autoradiographs into a MacIntosh Quadra computer, followed by quantitation of
individual bands using the NIH Image software program. Each gel was
autoradiographed for various times to ensure that the densities measured were
within the linear range of the scanner and software program. A dissociation rate was calculated for each Hox-Pbx1a-DNA complex by plotting the log of the complex band intensities versus
time. The dissociation rate was estimated as the slope of the regression line
generated from this plot. For each dissociation experiment the correlation coefficient for the line was >0.96. For each complex
the half-life was calculated using the equation
T
½
= -log(0.5)/
k
d
.
RESULTS
Hoxb
locus proteins containing the hexapeptide motif form cooperative DNA binding
complexes with Pbx1a
We have previously shown cooperative binding by several Hoxb proteins with Pbx1a
on an oligonucleotide (5'-CGAAT
TGATTGAT
GCA
CTAATTG
GAG-3') containing consensus binding sites for both proteins (
28
). We and others have shown that both the Hox and Pbx proteins bind to the
TGATTGAT sequence within this or similar DNA fragments (
28
,
29
,
31
). The same oligonucleotide was now used to study cooperative binding of each of the full-length Hoxb proteins which contain a conserved hexapeptide motif. As shown
in Figure
1
, Hoxb proteins representing paralog groups 1-8 were capable of forming a cooperative DNA binding complex with Pbx1a.
With the exception of weak complexes formed by Hoxb-4 (Fig.
1
) and moderately strong complexes by Hoxb-3 (Fig.
4
, lane 6), there was no evidence of interaction of the full-length Hox proteins with the DNA in the absence of Pbx1a under the
conditions of our standard electrophoretic mobility shift assays (EMSA), which
employ moderately high levels of non-specific competitor DNA as well as low concentrations of
in vitro
synthesized Hox proteins. In this regard, we have observed that in each case
tested (Hoxb-4, b-6 and b-7) truncated proteins containing only the homeodomain were
able to form complexes with the oligonucleotide under conditions where DNA
binding by the full-length Hox proteins was not detected (results not shown).
Hoxb protein-Pbx1a protein-DNA complexes show varying stability
A non-Hox homeodomain protein containing a YPWM motif
binds cooperatively to DNA with Pbx1a
Although the majority of homeodomain-containing proteins which contain a conserved hexapeptide sequence are
found within the Antennapedia or class I homeoproteins, there are several
proteins with divergent homeodomains which contain a hexapeptide motif. The tcl-3 (Hox11) protein, which was isolated as a break-point mutation in T cell leukemias (
33
,
34
), contains a TFPWME sequence which is located 22 amino acids upstream from the
homeodomain (Tables
1
and
1
). tcl-3 protein bound cooperatively to the target oligonucleotide with Pbx1a,
but exhibited no capacity to bind in the absence of Pbx1a (Fig.
5
B). The tcl-3-Pbx-DNA complex was so unstable that a half-life could not be measured (data not shown).
DISCUSSION
Although all of the Hoxb proteins containing a recognizable hexapeptide sequence
(Hoxb-1-Hoxb-8) display cooperative binding with Pbx1a, the stability of
the complexes formed with a specific oligonucleotide sequence vary over 100-fold. We have examined the contributions of three structural features of
Hox proteins to complex stability: (i) variability in the linker length between
the hexapeptide and homeodomain; (ii) differences in homeodomain sequences; (iii) variability of amino acids within the hexapeptide. Our findings suggest
that these determinants function in restricted combinations to modulate complex
stabilities.
Within the proteins of the
Hoxb
locus the conserved hexapeptide motif is separated from the homeodomain by a
variable linker. Computer analysis of the linkers for the Hoxb proteins
revealed no conserved amino acids or apparent structural motifs. Most
Hox
genes contain an intron which interrupts the linker region, presenting the
possibility of multiple splicing variants which would place the hexapeptide and
homeodomains at different distances from each other, such as are found in the
protein variants of the
Drosophila
Ubx
gene (
35
). The
Hoxb-3
cDNA clones isolated from normal bone marrow CD34
+
cells and a leukemic cell line contain linker regions of 53 and 35 amino acids
respectively. However, in this case both variants formed unstable complexes and
it was not possible to detect differences between them. There appears to be a
rough inverse correlation between linker length and complex stability for the
Hoxb proteins, as well as for tcl-3, which contains a long linker arm between the two domains. However,
mutant Hoxb-3 proteins in which the linker was reduced to 17 or 12 residues showed
either no change (Hoxb-3-17) or decreased complex stability (Hoxb-3-12), suggesting that linker length by itself is not
sufficient to determine complex stability. Although stability data were not
reported, Ubx proteins containing different linkers show differences in their
abilities to interact with Exd in a yeast two hybrid system, suggesting that
linker length and composition may, under other conditions, contribute to
overall DNA-protein complex stability (
32
). Since many
Hox
genes express multiple cDNAs, most of which have not yet been characterized, it
remains to be seen if differences between linker distances represents a
mechanism for altering Hox protein function. In addition, individual linkers
may be involved in additional protein-protein interactions.
In some respects the Hox proteins resemble other transcription factors which
appear to have distinct modular structures which confer DNA binding or
transcriptional activation (
36
). Thus the homeodomains of Hox proteins can bind to DNA in the absence of
flanking N- or C-terminal regions (
13
,
14
). We have obtained contradictory data concerning the modularity of Hox proteins
with regard to cooperative DNA binding with Pbx1a. Replacement of either the
Hoxb-3 homeodomain with that of Hoxb-6 or the Hoxb-6 homeodomain with that of Hoxb-1 yields mutant proteins which form barely detectable
complexes with Pbx1a and DNA. These results support the concept that the unique
flanking regions of each Hox protein interact with its homeodomain in a manner
which is sensitive to subtle changes induced by domain swapping experiments and
that the capacity to bind cooperatively to DNA with Pbx1a reflects specific
protein-protein contact residues within each particular Hox protein. However, we
have previously demonstrated substantial enhancement of the cooperative DNA
binding of Hoxa-10 with Pbx1a by mutational insertion of the Hoxb-7 hexapeptide and linker domains into the Hoxa-10 structure (
28
). Taken together, these results suggest that the capacity to form a cooperative
DNA binding complex with Pbx1a appears to result from the combined interactions
of the hexapeptide and homeodomain motifs which, at least in some cases, do not
appear to function as autonomous modules. Furthermore, the structural
information required for stable complex formation appears to reside largely
within the hexapeptide, linker and homeodomain, since removal of amino acids
outside these regions did not affect the stability of complexes formed with
Hoxb-7. To exemplify this point further, the four proteins from the fourth
paralog, which have almost identical hexapeptide, linker and homeodomain
sequences, were shown to form Pbx1a-DNA complexes with very similar half-lives.
In addition to our previous report demonstrating the importance of the
hexapeptide motif for cooperative Hox protein binding with Pbx1a and DNA (
28
), several other groups have also recently identified the YPWM sequence as being
important for mediation of complex formation (
30
,
32
,
37
). While this work was under submission, Neuteboom
et al
. have reported that within the Hoxb-8 hexapeptide the phenylalanine, tryptophan and methinone at positions 2,
4 and 5 respectively are required for interaction with Pbx proteins (
37
). These results agree completely with those which we have observed for Hoxb-4 interactions with Pbx1a. However, other data suggest that a conserved
YPWM or similar motif may not be required for cooperative binding. The
Engrailed protein, which lacks a recognizable hexapeptide motif, cooperatively
binds to DNA in the presence of Exd (
22
). At much higher protein concentrations a Ubx construct lacking the N-terminal region, including the hexapeptide, forms a complex with Exd in
the presence of DNA (
23
). Ubx appears to react with Exd through the C-terminal portion of the homeodomain itself, suggesting that the analogous
region of the mammalian homeodomains may interact with Pbx and DNA. We have
previously demonstrated that the Hoxa-10 homeodomain protein, which lacks a YPWM motif, can weakly cooperate
with Pbx1 to bind DNA, but that cooperativity is greatly enhanced by insertion
of a YPWM domain into the Hoxa-10 protein (
28
). These data suggest that although elements of the hexapeptide appear to be
required for cooperative binding of certain Hox proteins to DNA through Pbx1a,
there are additional mechanisms which stabilize interactions between Hox and
Pbx proteins.
In addition to effects of protein-protein interactions, a likely explanation for the observed differences
in complex stabilities is that the various Pbx-Hox complexes might have different optimal DNA sequence preferences. Thus
there may exist endogenous sites in which either the sequence and/or the orientation between
Hox and Pbx1a recognition sites favors formation of strong complexes between the proteins which form weak complexes on the particular synthetic site we studied. In this regard, the Engrailed protein cooperatively
binds with Exd to a DNA sequence which differs from that to which Ubx and Exd
cooperatively bind (
22
). Furthermore, a different consensus DNA binding site for Hoxb-1 with Pbx/Exd has recently been identified (
38
). Thus, there may be a `code' of particular Hox proteins and Pbx proteins which
will bind to specific DNA sites which result in differing arrays of complex
stabilities across the
Hox
loci. Although we have initiated studies to identify optimal synthetic and
biological binding sites for each of the different Hox proteins with Pbx, these
experiments will require substantial efforts beyond the scope of the present
investigation.
ACKNOWLEDGEMENTS
This work was supported by the Research Service of the Department of Veterans
Affairs (CL and HJL), by grants from the National Institutes of Health (N44DK-3-2219, CL and CA42971, MLC) and by the National Cancer Institute of
Canada with funds from the Canadian Cancer Society and the Medical Research Council of Canada (GS
and RKH). HJL is the recipient of a Department of Veterans Affairs Career
Development Award. MLC is a fellow of the Leukemia Society of America.
