ABSTRACT
Heterodimers between the Pbx/Exd and Hox/HOM-C classes of homeodomain proteins bind regulatory elements in tissue-specific and developmentally regulated genes. In this work, we
characterize the half-site bound by both Pbx1 and Hox proteins on a prototypic element
(TGATTAAT) and determine how the orientation of the Hox protein contributes to
the DNA binding specificity of Pbx-Hox heterodimers. We demonstrate that the Hox protein binds the 3
'
TAAT sequence as its recognition core and exhibits sequence-specific binding at positions 3
'
to the TAAT core. Unfavored sequences at this position, such as two cytosines,
abrogate binding to the element. The upstream Pbx1 core sequence, TGAT, must immediately juxtapose the Hox core.
This geometry maintains the preference of Hox/HOM-C proteins for a T base at position -1, as T represents the fourth position of the Pbx1 core, and
suggests that this T base is bound by both Pbx1 and Hox proteins, Pbx1 binding
in the major grove and the Hox protein binding in the minor grove. Pbx1 also exhibits base selectivity 5'
to its TGAT recognition sequence.
Homeobox genes regulate cell fate and segmental patterning (
1
-
4
) and encode a conserved DNA binding motif known as the homeodomain (HD;
4
). HD proteins act with great biological specificity in development and their
functional differences are largely mediated by their HDs (
5
-
8
). The vertebrate Hox proteins and
Drosophila
HOM-C proteins are a subset of HD proteins that specify structural development
along the anterior-posterior (A-P) axis (
1
,
3
,
9
) through specific recognition of DNA motifs in target genes (
10
-
13
). Hox and HOM-C proteins share structural and functional features, including the primary
sequence of their HDs, the chromosomal organization of their respective genes
and their spatial and temporal expression patterns along the A-P axis (
1
,
3
,
9
). Mutations in these genes induce specific segments to develop morphological
structures of another.
Despite the large number of different biological functions of homeoproteins,
most HDs bind DNA motifs containing a TAAT core (
12
,
14
-
16
). Within the three [alpha]-helices and flexible N-terminal arm that comprise the HD (
17
), basic residues at positions 3 and 5 bind the 5' TA, and Asp51 and a hydrophobic residue at position 47, both of which
lie in helix 3, bind the 3' A and T bases of the core, respectively (
18
,
19
). Residues in helix 3 make sequence-specific contacts 3' of the TAAT core that permit different HD proteins to discriminate
among related binding sites. When residue 50 is Lys, two consecutive cytosines
are strongly favored immediately 3' to the TAAT core (
20
,
21
), but when residue 50 is Gln two guanines are favored. Residue 50 independently
designates this specificity, because interconversion of Gln and Lys at position
50 in Bicoid or in Hox/HOM-C proteins interconverts DNA binding specificity between TAATGG and TAATCC
(
20
,
21
).
While all Hox/HOM-C proteins encode Gln at residue 50, they maintain different intrinsic
specificities for sequences positioned up to four bases 3' to the TAAT core. For example, Ubx binds TAAT
Members of the Pbx/Exd family of HD proteins cooperate with Hox proteins to
regulate target gene expression and bind DNA cooperatively with many Hox and
HOM-C proteins (
24
-
28
). The Pbx/Exd family includes the human proto-oncoprotein Pbx1 (
29
-
30
), its mammalian homologs, Pbx2 and Pbx3 (
31
),
Drosophila
Extradenticle (Exd;
32
) and
Caenorhabditis elegans
ceh-20 (
33
). The importance of Pbx1 in differentiation is also implicated by the fact that
E2A-Pbx1, an oncogenic and transcriptionally activated derivative of Pbx1
found in human pre-B cell leukemia (
29
,
30
), can block normal hematopoietic differentiation (
34
-
36
). Cooperative binding of Pbx and Hox proteins was first described on TTGATTGAT
(
26
,
37
), which was derived from selection experiments using degenerate
oligonucleotides and recombinant Pbx proteins. Heterodimer formation requires
only the HD and adjacent C-terminal 17 residues of Pbx1 (
38
) and only the HD and N-terminal sequences sufficient to include an essential pentapeptide
sequence (YPWMR) of Hox proteins (
25
,
39
,
40
). Because transfer of the Hox pentapeptide sequence onto the HD of Pbx1 causes
two molecules of Pbx1 to exhibit cooperative binding to TTGATTGAT, this element
was likely initially isolated due to binding of Pbx1 as a dimer (
38
). Thirty two genes of the Hox loci that are either known or proposed to exist
(A1-A8, B1-B8, C1-C8 and D1-D8) encode HD proteins containing variants of this
pentapeptide sequence and thus are predicted to be able to heterodimerize with
Pbx1 on appropriate DNA elements. In addition, the HD protein STF-1, which is not encoded by the Hox loci and is expressed in the pancreas
and small intestine, also contains a pentapeptide sequence (
41
) and binds TTGATTGAT cooperatively with Pbx1 (
42
).
Elements similar to TTGATTGAT are bound by Hox/Pbx heterodimers
in vivo
and are required for expression of tissue-specific genes. The Hox B1 promoter contains TGATGGAT and AGATTGAT
elements that are essential for stage-specific and tissue-specific expression and bind Hox B1 and Exd (Table
1
;
27
). In the somatostatin promoter a Pbx-STF-1 heterodimer binds TGATTAAT (
42
), a site important for expression of somatostatin in pancreatic islet-cells (
41
). This element can also be activated synergistically by STF-1 and E2A-Pbx1 (
42
). Cooperative DNA binding by Pbx and Hox proteins raises the question of how
each heterodimer partner is oriented on its half-site, whether these half-sites are separable and whether heterodimer formation alters the DNA
binding specificity of Hox proteins. In this work, we address these issues.
Table 1
We demonstrate that the Hox protein binds to these elements by utilizing the 3' TGAT or TAAT sequence as its recognition core, that its HD assumes the
same orientation on DNA as that of Hom-C proteins and that it exhibits sequence-specific contacts at positions 3' to sequences currently considered to comprise such elements.
The upstream Pbx1 core sequence, TGAT, must be immediately juxtaposed to the
Hox core. This geometry maintains the preference of Hox/HOM-C proteins for a T residue at position -1 (
14
) because T represents the fourth position of the Pbx1 core. Pbx1 also exhibits
additional 5' base selectivity. We suggest that an internal portion of the Pbx-Hox element is bound by both partners, Pbx1 approaching from the
major grove and the Hox protein binding in the minor grove.
All cDNAs were cloned in vectors pGEM 3Z F
-
, pGEM 4Z or pGEM 3Z (Promega).
Residue 50 of the HD of Hox A5 and Hox B8 cDNAs was changed to Lys using the
Muta-gene Phagemid
In Vitro
Mutagenesis kit (BioRad) according to the manufacturer's protocol. All
mutations were verified by DNA sequence analysis.
In vitro
transcription/translation was performed using the Promega TNT Coupled
Reticulocyte Lysate System in accordance with the manufacturer's protocol.
Double-stranded oligonucleotides were labeled with [
32
P]ATP to the same specific activities by phosphorylating a common reverse
oligonucleotide that was annealed to the oligonucleotides shown in Table
1
and made double-stranded by addition of dNTPs and the Klenow fragment of DNA polymerase I.
Bound and free probe were separated by electrophoresis in native 5% acrylamide
gels formed in 0.5* TBE (27 mM Tris, 27 mM boric acid, 0.6 mM EDTA) and run in the same
buffer. Binding reactions contained 20 000 c.p.m. probe, 4 [mu]l
in vitro
translated proteins, 1 [mu]g poly(dI[middot]dC), 10 mM Tris, pH 7.5, 1 mM EDTA, 1 mM dithiothreitol, 0.1%
Nonidet P-40 and 5% glycerol. Reactions were incubated for 30 min at 23oC. For Hox A5 and Hox B8 proteins containing point mutations at
residue 50 of the HD, equivalent amounts of wild-type and mutant proteins were used in binding reactions, as determined by
labeling during transcription/translation reactions using [
35
S]methionine.
The half-sites bound by Pbx1 and Hox proteins on a single element have only been
partially characterized. DNase I protection analysis suggests that Pbx1 is
bound 5' to Hox proteins (
43
). Mutation of the 5' G in TGATTGATT to A abrogates cooperative DNA binding, while mutation of
the 3' G to A maintains cooperativity, consistent with the proposal that Pbx1
binds a 5' half-site containing TGAT and that the Hox protein binds a 3' half site consisting of either TGAT or TAAT. On each
proposed half-site these core nucleotides are numerically designated 2-5, and for the combined Pbx/Hox site that we propose they are
designated 2-5 and 6-9 for Pbx1 and Hox proteins, respectively (Table
1
). Hox proteins exhibit negative selection for CC at the two positions 3' to their TAAT core and positive selection for G or T at the first 3' base (6 in the monomeric site) and either G or A at the second 3' base (7 in the monomeric site;
14
,
15
,
23
). Therefore, if the Hox protein binds the 3' TAAT or TGAT core in the Pbx1-Hox heterodimer site, positions 10 and 11 should prove crucial for
strong binding of this element by either Hox monomers or Hox/Pbx heterodimers.
To develop a model system in which we could observe the effect of sequence
alterations on the binding of both monomeric Hox proteins as well as Pbx1-Hox heterodimers, we selected the Pbx/STF-1 binding site contained in the somatostatin promoter (TGATTAAT),
which contains an optimal TAAT core. Another TAAT core sequence is encoded by
the opposite strand, however, this sequence is followed by a CA dinucleotide
which creates an unfavorable Hox monomer binding site. If changes at positions
10 and 11 of the combined Pbx1/Hox binding site altered both Hox and Hox-Pbx1 binding, it would be very unlikely that the Hox protein was bound to
the reverse TAAT core because nucleotides located 3-4 bases 5' of a TAAT site have little to no effect on Hox binding. Due to
the co-migration of the Pbx-Hox-DNA complex with a non-specific complex produced by the binding of a factor
in reticulocyte lysate, we used the larger E2A-Pbx1 fusion form of Pbx1 for these experiments. E2A-Pbx1 contains all the determinants required for cooperative DNA
binding with Hox proteins and, in a previous study, E2A-Pbx1 and Pbx1 were found to exhibit indistinguishable behavior in their
cooperative binding to DNA with Hox B7, Hox B8, Hox A5 and STF-1 (
43
). DNA probes containing GG, TA or CC at positions 10 and 11 exhibited dramatic
differences in their abilities to bind monomers of Hox B7 (lanes 5-7), Hox B8 (lanes 8-10), Hox A5 (lanes 11-13) and STF-1 (lanes 14-16) or heterodimers of E2A-Pbx1 and each Hox protein (Fig.
1
). Like Pbx1, E2A-Pbx1 failed to bind probe in the absence of a Hox protein (lane 1) and
each of the Hox proteins alone failed to form the upper complex in the absence
of E2A-Pbx1 (lanes 2-4). This suggested that the Hox protein bound the 3' TAAT sequence (positions 6-9) in a manner consistent with the sequence-specific DNA binding characteristics of Hox
proteins.
In the crystal structures of Eng and Ant, N51 contacts the A at position 4 and
residue 50 makes base-specific contacts at positions 6 and 7 (
17
,
18
). In Bicoid, residue 50 is Lys and specifies CC at these positions, while HD
proteins encoded by the HOM-C/Hox gene contain Gln at this residue (
20
). In order to prove that the Hox protein in the E2A-Pbx1-Hox complex was bound in the hypothesized orientation, Gln50 of Hox
A5 was converted to Lys in order to provide biochemical evidence that this
would alter binding and favor a TAATCC motif (Fig.
3
). Indeed, the K50 mutation in Hox A5 eliminated monomeric Hox A5 and E2A-Pbx1-Hox A5 binding to TGATTAAT
Figure
Pbx1 is strongly implicated to bind DNA in the same orientation as the Hox
protein binds its 3' half-site, because fusion of the Hox pentapeptide N-terminal to Pbx1 causes it to bind the internally repeated
sequence TGATTGATT as a strong and cooperative homodimer (
38
). Thus, Pbx1 is likely to bind the upstream half-site core TGAT (bases 2-5), employing N51 to bind the invariant A at position 4
(designations in Table
1
). Because HD proteins such as Ftz, Ubx and Dfd prefer C or T located just 5' of their TAAT core site (
15
), we examined whether Pbx1 exhibited sequence specificity at this location
(position 1 in the Pbx1/Hox site) and whether it was similar or different to
that predicted for Hox proteins (Fig.
4
). Changes at position 1 had a dramatic effect on the degree of cooperative
binding with Hox proteins, A (lanes 2 and 6) being favored over T or G (lanes
1, 4, 5 and 8) and C permitting very little binding (lanes 3 and 7). By
contrast, changes at position -1 had little effect on formation of the complex containing E2A-Pbx1 and Hox B7 or Hox A5, consistent with the observation that the
-1 position has little effect on sequence-specific binding of HD proteins (lanes 9-16).
Figure
The cooperative binding of Pbx and Hox proteins requires an essential
pentapeptide sequence that is positioned variably from the Hox HD (
25
,
39
,
40
). In Hox B8 it resides five amino acids N-terminal to the HD, while in Hox A5 it is positioned 15 residues N-terminal to the HD. Thus, interaction of Pbx1 with this pentapeptide
sequence could be permitted for DNA elements containing spacing between the
Pbx1 and Hox half-sites. Alternatively, precise spacing geometry between Pbx and Hox half-sites may be required in order to achieve formation of the
heterodimer. To test directly whether these defined half-sites can be separated, a T, G or GG sequence was inserted between
residues 5 and 6. The T preserves the Hox preference for T at its position 1
and maintains a possible preference of Pbx1 for T at position 6. All insertions
completely abrogated formation of the E2A-Pbx1-Hox complexes, indicating that the Pbx1 and Hox half-sites cannot be separated (Fig.
5
). Failure to form heterodimers was not due to diminished binding of the Hox
protein, because insertion of a single T actually enhanced the binding of Hox
A5.
Figure
In this paper we define the binding orientation of Hox proteins in Pbx-Hox complexes and the sequence specificity of each partner at positions 5' and 3' to its half-site. The binding of E2A-Pbx1 and Hox proteins on DNA provides sequence-specific contacts in at least 11
positions (Table
1
and Fig.
6
A). Mutation of Q50 to K50 in Hox proteins altered specificity of positions 10
and 11 to CC and binding analysis on all 16 dinucleotide possibilities mirrored
the specificity of monomeric binding by Hox B7 and Hox B8 and the general
specificity of Hox and HOM-C proteins characterized by others (
14
,
44
-
46
). Thus, the Hox protein binds the TAAT core spanned by positions 6-9 using the same conformation as that used by Ant or Eng, based on
crystallographic studies (
17
,
18
). Dimer formation of Pbx with Hox A5, Hox B7 and Hox B8 on DNA thus permits the
Hox protein to exhibit its major sequence-specific contacts. In this conformation N51 would bind A at position 8 and the N-terminal arm of the HD would contact bases 5-7 (Fig.
6
B). Extension of Pbx1-Hox-DNA interactions to positions 10 and 11 is also consistent with
earlier methylation interference studies, in which a heterodimer of E2A-Pbx1 and Hox A5 exhibited partial protection of guanine 10 and 12 in the
DNA sequence TTGATTGAT
Figure
Pbx1 is proposed to bind TGAT (positions 2-5) as its core, with N51 recognizing A at position 4 (Fig.
6
B). This orientation is based on the fact that fusion of a Hox pentapeptide
sequence to the Pbx1 N-terminus produces a protein that exhibits strong cooperative binding to
the repetitive element TGATTGAT (
38
), suggesting that adjacent Pbx1 proteins bind in the same orientation and that
the Pbx1 pentapeptide fusion protein binds the 3' site in the same orientation as would a Hox protein. Having defined the
orientation of Hox protein binding, we propose that Pbx1 binds in this same orientation. In this orientation, the specificity of
Pbx1 would be very similar to that of yeast a1. When Pbx1 was first identified,
the sequence of its HD exhibited the highest degree of amino acid identity with
that of yeast a1 (19/30 residues in helices 2 and 3). Yeast a1 binds a very
similar sequence (TGATGT-A/G) and forms cooperative heterodimers with a second HD protein, [alpha]2 (
46
). The crystal structure of an al-[alpha]2 heterodimer bound to DNA has recently been solved (
47
) and there are several similarities in binding properties to those proposed in
our model for heterodimers of Pbx and Hox proteins. In the a1-[alpha]2 heterodimer, a1, like Pbx1, binds the 5' half-site containing a TGAT. During heterodimerization
with [alpha]2, residue N51 of a1 contacts the third base, A, of the 5' half-site, which is the same HD-base contact occurring in our model of Pbx binding
during heterodimerization with Hox proteins. Further, an unstructured C-terminal helix of [alpha]2 is responsible for making contact with the HD of a1, which
parallels the interaction of the unstructured pentapeptide motif N-terminal to the HD of Hox proteins with the HD of Pbx1 during cooperative
DNA binding. Finally, the crystal structure of al-[alpha]2 indicates that al and [alpha]2 bind in a head-tail/head-tail fashion, also consistent with the model of
Pbx-Hox heterodimerization that we have proposed. However, while the Pbx and
Hox half-sites are unspaced, those of a1 and [alpha]2 are separated by 6 bp (
46
) and a1 and [alpha]2 bind DNA on the same side of the double helix.
In contrast to the binding of a1 and [alpha]2, unspaced binding of Pbx1 and Hox to adjacent 4 bp cores would position
each protein approximately on opposite sides of the double helix, in a
conformation similar to that demonstrated by crystallographic studies of the
binary POU
(homeo)
and POU
(specific)
DNA binding domains that typify the POU family of HD proteins (prototyped by
the Brn-2, Oct-1 and Pit-1 transcription factors). The POU
(homeo)
domain has a structure almost identical to that of a Hox HD and the adjacent
POU
(specific)
domain contains a structure similar to the HD but which is stabilized by a
fourth helix. When Brn-2 binds the element GCATTAAT, the POU
(specific)
domain binds GCAT and the POU
(homeo)
domain binds TAAT (
48
), and both domains bind in the same orientation. Proteins in the POU family of
transcription factors retain high affinity binding even when the individual
half-sites are separated by up to 3 bp (
48
).
Unlike the opposing DNA binding domains of POU proteins, the Pbx1 and Hox half-sites could not be separated by 1 (T or G) or 2 (GG) bp without a dramatic
loss in cooperative binding. This was not due to a reduction in the binding of
the Hox protein, as evidenced by the fact that both Hox A5 and Hox B7 bound the
site containing the T insertion as effectively as they did the parental site.
The inseparability of the half-sites suggests that precise protein-protein interactions occur between each HD. Specific protein-protein interactions were also suggested by the behavior of
mutations at Glu28 of the Pbx1 HD (
38
). An E28R mutation abolished cooperative DNA binding, maintained monomeric
binding and maintained the ability of the HD to exhibit increased DNA binding
upon addition of a synthetic Hox A5 pentapeptide sequence, suggesting that this
mutation did not abrogate interaction with the pentapeptide, but eliminated a
second direct interaction between the Pbx1 and Hox A5 HDs. In this regard,
dimer formation between Pbx1 and Hox proteins is more similar to dimers of the
Drosophila
Paired HD, which binds cooperatively to a palindrome composed of two inverted
TAAT sites separated by 2 bp and in which case cooperative DNA binding is
drastically impaired by further separation of the TAAT sites (
49
).
The monomeric DNA binding specificity of Hox proteins has been examined
previously (
50
,
51
) and those studies suggest that an optimal site that binds a Hox monomer is
somewhat different to the optimal half-site that binds a Hox protein in the Pbx1-Hox heterodimer. Class one Hox proteins vary minimally in their
monomeric DNA binding specificity, preferring C/G-TAATTG, with TG being a more optimal 3' dinucleotide than GG (
51
). Although we find that a 3' TG dinucleotide creates a good Pbx1-Hox heterodimer site, a 3' GG nucleotide is clearly better for all Hox proteins
tested. This may suggest that heterodimer formation with Pbx1 alters, to some
degree, the specificity of DNA binding by helix 3 of the Hox HD at its contacts
just 3' to the core sequence. Similarly, the optimal nucleotides 5' to the TAAT Hox monomeric core also differ from what we observe
in an optimal Pbx1-Hox heterodimer. In the heterodimer the adjacent Pbx1 binding site
presumably dictates the AT dinucleotide 5' to the TAAT half-site. While this dinucleotide combination is invariantly retained
in optimal DNA motifs selected by combinations of Pbx1 plus Hox proteins (Lu
and Kamps, unpublished), the optimal dinuclotide at this position for Hox
monomers is AC, and AT is disfavored (
51
). Furthermore, as observed in Figure
5
, when the two half-sites of the Pbx1/Hox element were spaced by addition of a T, which
destroys formation of the Pbx1-Hox-DNA complex, the resulting TT dinucleotide 5' to the TAAT half-site strongly enhanced monomeric binding by Hox A5,
suggesting that Hox A5 prefers a TT dinucleotide 5' to its TAAT core when binding as a monomer. Thus, the site bound by Hox
proteins in optimal Pbx1-Hox heterodimers is different to the optimal site bound by monomeric Hox
proteins, and this may serve as one mechanism
in vivo
to prevent Hox proteins from binding as monomers to optimal Pbx1/Hox motifs.
Although Hox A5, Hox B7 and Hox B8 all display approximately the same preference
for binding dinucleotides 3' to the TAAT half-site during heterodimeric binding with Pbx1, there is now strong
evidence that different Pbx1-Hox heterodimers do indeed bind different DNA motifs, however, the
variation is not within the 3' dinucleotide, but within the second position of the proposed Hox core
(position 7 of Table
1
and Fig.
6
). We have found that heterodimers of Pbx1 plus Hox A1 favor G at this position
and bind much less tightly to probes containing an A or T at this position. By
comparison, Hox D4 and Hox D5 form the most abundant complex with Pbx1 on DNA
containing A in this position and somewhat less with those containing G or T,
while Hox B7, Hox B8 and Hox C8 strongly prefer a T (
52
). This same binding specificity was also recently reported by Chang
et al
(
53
). We demonstrated that as a monomer, Hox B8 binds most stabily to the Pbx1/Hox
DNA motif containing a TAAT Hox core, while as a heterodimer with Pbx1 it binds
more stabily to the DNA motif containing a TTAT-containing probe. Thus, we postulate that heterodimer formation with Pbx1 alters interaction
between the N-terminal arm of the Hox HD and nucleotide 7 (Fig.
6
) in the Pbx1/Hox motif, shifting the stability from A to T or G, depending on
the unique Hox sequences within the N-terminal arm.
The Pbx/Exd family of HD proteins is unique in its content of Gly at position
50. The close proximity of Pbx1 and Hox proteins in the Pbx1-Hox-DNA complex may indicate that Gly50, which lacks a side chain,
allows Pbx to exhibit a lower degree of sequence specificity for nucleotides 3' to its TGAT core, permitting the adjacent Hox protein to dictate most of
the sequence specificity associated with the heterodimer element. The HDs of
Pbx1, Pbx2 or Pbx3 select sequences containing TGATG and TGATT as monomer
binding motifs (
26
), indicating that Pbx proteins exhibit some sequence-specific selection for the T at position 6, but little specificity for
bases 3' to this position. Thus, in the Pbx1-Hox-DNA complex both HD proteins are proposed to contact
positions 5 and 6, the N-terminal arm of the Hox protein making contacts in the minor groove and
helix 3 of Pbx1 binding in the major groove. In this manner, Pbx proteins would
differ from yeast a1, which encodes Ile at position 50 and displays DNA-specific contacts three bases 3' of its core (TGAT
The half-sites we propose for Pbx and Hox proteins, as well as their inseparable
nature, is consistent with the sequences of known Pbx/Hox sites in cellular
promoters, as well as with sites known to bind Pbx1 and unidentified partners.
In addition to the Pbx1/Hox sites found in the Hox B1 and somatostatin
promoters, Kagawa
et al
. (
54
) have described a site in the bovine CYP17 (P450
17[alpha]
) gene that binds Pbx1 and a second factor and functions as a cAMP-responsive element (Table
1
). This element contains TTGAT at its 5'-half, consistent with the general positioning of Pbx1 on the 5' half-site and GGACAG on its 3' half-site. Likewise, we have used antibodies
to Pbx1 and nuclear extracts from pre-B cells to select the best heterodimer binding motif for Pbx1 and derived
the consensus TTGATTGAC-A/G-G, again finding that the Pbx1 half-site consensus is positioned at the 5' half-site (Knoepfler, unpublished). In addition, all
elements (Table
1
and the aforementioned elements) contain invariant A residues in the fourth and
eighth positions, consistent with the proposal that N51 of helix 3 contacts
this base and that spacing between the half-sites is not allowed and suggesting that the undefined partners may be HD
proteins as well. Positions 10 and 11 of all elements also contain dinucleotide
combinations favorable for the binding of Hox proteins. Interestingly, there is
no consistent requirement for the T and A residues at positions 6 and 7, which
is normally specified by the N-terminal arm of the Hox protein and which is present in almost all
naturally occurring monomeric Hox binding sites (
11
,
12
,
14
-
16
). This may indicate that altered DNA binding by the Hox protein occurs upon
heterodimer formation with Pbx1. Confirmation of this general model, as well as
elucidation of specific variations in binding induced at the heterodimer
interface, must await further experimentation and resolution of the crystal
structure of a Pbx-Hox-DNA complex.
This work was supported by Public Health Services grants 2 PO1 CA50528 and
CA56876. MPK is a PEW scholar in the Biomedical sciences. PSK is a fellow of
the Lucille P.Markey Charitable Trust.




REFERENCES
Return
