Hox genes control cell fates and specify regional identities in vertebrate development. Hox proteins show a relaxed DNA‐binding selectivity in vitro, suggesting that functional specificity is achieved in vivo through the action of transcriptional co‐factors. Pbx proteins are good candidates for such a role, on the basis of both genetic and biochemical evidence. We report that the human Pbx1 and HOXB1 proteins can cooperatively activate transcription through a genetically characterized Hox target, i.e. an autoregulatory element directing spatially restricted expression of the murine Hoxb‐1 gene (b1‐ARE) in the developing hindbrain. On the b1‐ARE, only a restricted subset of HOX proteins (HOXA1, HOXB1, HOXA2) are able to bind cooperatively with Pbx1 and activate transcription. Selective recognition of the b1‐ARE is mediated by the N‐terminal region of the HOX homeodomain. The DNA‐binding and protein–protein interaction functions of HOXB1 and Pbx1 are all necessary for the assembly of a transcriptionally active complex on the b1‐ARE. Functional dissection of the complex allowed the localization of the main activation domain in the HOXB1 N‐terminal region, and of an additional one in the C‐terminal region of Pbx1 contained in the Pbx1a but not in the alternatively spliced Pbx1b isoform. Our results indicate that Pbx1 acts as a transcriptional co‐factor of Hox proteins, allowing selective recognition and cooperative activation of regulatory target sequences.
Genes encoding the Hox family of homeodomain‐containing transcription factors regulate cell fates and developmental pathways in metazoans, eventually leading to the generation of morphological differences along body axes (reviewed in Krumlauf, 1994). Despite the high degree of functional specificity observed in vivo, a large number of Hox proteins (up to 38 in mammalian species) appear to bind in vitro to a restricted set of very similar sites with similar affinities. For this reason, the specificity of action of Hox genes in vivo has been postulated to rely not only on their DNA‐binding properties, but also on the activity of transcriptional (co‐)factors (reviewed in Manak and Scott, 1993). Good candidates for such a role are the homeodomain‐containing products of the extradenticle (exd)/Pbx gene family. exd was originally identified as a mutation causing patterning defects in Drosophila without affecting homeotic gene expression (Peifer and Wieschaus, 1990). exd function is not dependent on Hox gene activity (Wieschaus and Noell, 1986), while it is required for the regulation of Hox‐controlled target genes (Chan et al., 1994; Rauskolb and Wieschaus, 1994), consistent with the proposed role of exd as a Hox co‐factor.
Three mammalian Exd homologues, termed Pbx1, 2 and 3, have been isolated independently due to the involvement of Pbx1 in a chromosomal translocation associated with pre‐B cell human leukaemia (Kamps et al., 1990; Nourse et al., 1990). The high degree of homology (∼70%) between exd, the mammalian Pbx1, 2 and 3, and the Caenorhabditis elegans ceh‐20 genes (Burglin and Ruvkun, 1992) parallels that observed for the Hox genes, and suggests the conservation throughout evolution of co‐factor functions. In agreement with the proposed role as Hox co‐factors, exd/Pbx proteins have been found to cooperatively stimulate sequence‐specific DNA binding of several Hox proteins (Chan et al., 1994; Van Dijk and Murre, 1994; Chang et al., 1995; Knoepfler and Kamps, 1995; Lu et al., 1995; Pöpperl et al., 1995), and to select different consensus binding sites in association with different Hox proteins in vitro (Chan and Mann, 1996; Chang et al., 1996). The formation of Hox–Pbx complexes has been shown to rely on protein–protein contacts mediated by the Hox homeodomain and a short conserved motif (consensus sequence: YPWM) located upstream of the homeodomain in a subset of Hox proteins (reviewed in Mann and Chan, 1996). The ‘YPWM’ motif appears to be required (Chang et al., 1995; Knoepfler and Kamps, 1995; Phelan et al., 1995), although it is not sufficient (Chan et al., 1994; Johnson et al., 1995), to stabilize the Hox–Pbx complex.
The function of exd/Pbx proteins as Hox co‐factors has been studied extensively in vivo, in Drosophila and in transgenic mice, as well as in vitro (reviewed in Mann, 1996; Mann and Chan, 1996). However, few studies have directly addressed the role of these proteins as modulators of Hox function at the level of transcriptional regulation (Lu et al., 1995; Peers et al., 1995; Phelan et al., 1995; Chang et al., 1996; Lu and Kamps, 1996a). We have analysed the transcriptional activity of Hox–Pbx complexes on a well characterized Hox target, the Hoxb‐1 autoregulatory element (Pöpperl et al., 1995). The expression of Hoxb‐1 in the developing mouse hindbrain is confined to a segmentally organized cellular compartment referred to as rhombomere 4 (reviewed in Keynes and Krumlauf, 1994), as the result of a positive autoregulatory feedback loop, mediated by a conserved 148 bp genomic region (Pöpperl et al., 1995). The activity of the Hoxb‐1 autoregulatory element (b1‐ARE) was shown in transgenic flies to be dependent on the functions of both exd and the Drosophila Hoxb‐1 homologue labial (Pöpperl et al., 1995). The b1‐ARE contains three related sequence motifs (repeats 1, 2 and 3) which are nearly identical to the in vitro defined Pbx consensus binding sequence ATCAATCAA (Van Dijk et al., 1993; Lu et al., 1995). These repeats are necessary and sufficient to drive expression of a reporter gene in rhombomere 4 of transgenic mice. In particular, repeat 3 (R3) proved to be crucial for rhombomere‐restricted expression, and to mediate cooperative binding in vitro of hoxb‐1 and exd proteins, indicating that the b1‐ARE may indeed represent a natural target for positive regulation by a Hoxb‐1–Pbx complex (Pöpperl et al., 1995).
Here we show that the human HOXB1 and Pbx1 proteins cooperatively activate transcription through the b1‐ARE in transfected cells. The functional specificity of various HOX–Pbx1 complexes on the b1‐ARE was analysed by co‐expressing with Pbx1 a set of HOX proteins, belonging to different clusters and/or paralogy groups, and assaying for transcriptional activation. We show that transcription from the b1‐ARE is activated selectively by members of HOX paralogy groups 1 and 2 in combination with Pbx1. Domain‐swap experiments showed that the homeodomain N‐terminal region of HOXB1 is sufficient to transfer the ability to bind to the R3 site in vitro, and to activate transcription from the b1‐ARE in transfected cells, to the otherwise inactive HOXB3 protein. The transcriptional activity of the HOXB1–Pbx1 complex on the b1‐ARE is dependent essentially on the activation domain of HOXB1, located in the N‐terminal portion of the protein. An additional activation domain was mapped to the C‐terminus of the Pbx1 protein, within a region present in the Pbx1a but not in the alternatively spliced Pbx1b isoform.
HOXB1 and Pbx1 cooperatively activate transcription through the Hoxb‐1 gene autoregulatory element
The ability of the HOXB1–Pbx1 complex to regulate transcription was assayed in transient co‐transfection experiments, using the b1‐ARE (Pöpperl et al., 1995) as a responsive element. The pAdMLARE reporter construct contains a 148 bp fragment from the b1‐ARE upstream of the adenovirus major late (AdML) minimal promoter in the pAdMLluc luciferase reporter gene (Figure 1A). Transfection of COS7 cells with the pAdMLARE reporter (ARE in Figure 1B), together with a construct expressing either HOXB1 or Pbx1 under the control of the SV40 promoter (pSGHOXB1 and pSGPbx1 respectively), did not significantly stimulate the reporter basal activity (Figure 1B). Conversely, co‐transfection of both the Pbx1 and the HOXB1 expressors caused a 9‐fold enhancement of the pAdMLARE activity (ARE in Figure 1B), but not of the enhancer‐less pAdMLluc reporter transfected as a control (C in Figure 1B).
Among the three related sequence motifs present in the b1‐ARE, R3 was shown to mediate most of the activity of the b1‐ARE in transgenic mice (Pöpperl et al., 1995). To study the relative role of this motif in mediating transcriptional activation from the HOXB1–Pbx1 complex, we generated a knockout mutation of R3 within the b1‐ARE reporter (pAdMLAREmR3, Figure 1A). Co‐transfection of pSGHOXB1 and pSGPbx1 expressors with the pAdMLAREmR3 reporter (AREmR3 in Figure 1B) led to a relative stimulation of the reporter basal activity of ∼40% of that observed with the wild‐type pAdMLARE, indicating that R3 is required for maximal activity of the ARE element. To test whether R3 alone was sufficient for mediating transcriptional activation by HOXB1 and Pbx1, we generated a reporter construct containing a 3mer of R3 (pAdMLR3, Figure 1A). As shown in Figure 1B, HOXB1 and Pbx1 were able to activate the pAdMLR3 reporter (R3) at a level comparable with that observed with the complete b1‐ARE. Transfection of either protein alone had no effect on the reporter activity (Figure 1B). Identical results were obtained with a reporter in which the R3 3mer was cloned upstream of a different minimal promoter, such as the herpes simplex virus thymidine kinase (TK) −81 promoter (Figure 1A and C). Cooperative action of HOXB1 and Pbx1 in activating transcription from the b1‐ARE was also observed by co‐transfection of a different cell type, i.e. the murine embryonal carcinoma P19 cell line (Figure 1D).
These data show that a Pbx1‐dependent activation of the b1‐ARE by the HOXB1 protein can be reproduced in cultured cells. In this context, Pbx1 behaves as a transcriptional co‐factor of HOXB1, allowing cooperative activation of the b1‐ARE in a promoter‐ and cell context‐independent fashion. Within the b1‐ARE, the R3 site appears to be sufficient for transactivation by HOXB1 and Pbx1.
The b1‐ARE selectively mediates Pbx1‐dependent binding and transcriptional activation by members of the HOX paralogy groups 1 and 2
Next we tested whether other HOX gene products could transactivate the b1‐ARE in combination with Pbx1. Twelve HOX genes belonging to different clusters and/or paralogy groups were co‐expressed with Pbx1 in transient transfections of COS7 cells, and tested for cooperative transcriptional activation of the pAdMLARE reporter. As shown in Figure 2A, none of the tested HOX proteins alone were able to transactivate the pAdMLARE reporter. Expression of the same set of HOX proteins in combination with Pbx1 led to significant (i.e. 4‐ to 6‐fold) transactivation of the reporter activity only in the case of HOXA1, HOXB1 and HOXB2, whereas very weak (<2‐fold) or no transactivation was observed for the other HOX proteins (Figure 2B). Identical results were obtained by co‐transfecting the pAdMLR3, containing the R3 trimer (data not shown). Production of HOX proteins by the transfected plasmid was quantitatively comparable, as tested by an electrophoretic mobility shift assay (EMSA) of the transfected cell nuclear extracts with an oligonucleotide containing a Hox consensus binding site (not shown).
A representative subset of the same HOX proteins was tested for cooperative DNA binding with Pbx1 to the b1‐ARE R3 site in EMSA. As shown in Figure 2C, only the HOXA1, HOXB1 (lanes 3 and 4) and HOXB2 (not shown) proteins formed a detectable retarded complex with a labelled R3 oligonucleotide in the presence of Pbx1. No binding was detected with Pbx1 alone (lane 2), or with either of the HOX proteins alone (not shown). Polyclonal antibodies against the full‐length Pbx1 protein completely abolished the formation of the HOXA1–Pbx1 and HOXB1–Pbx1 complexes (Figure 2C, lanes 12 and 13).
These results show that the b1‐ARE, and the R3 site within it, selectively mediate transcriptional activation by HOX–Pbx1 complexes made of members of the HOX paralogy groups 1 and 2. DNA binding in vitro experiments suggest that this selectivity relies on the differential recognition of the R3 site by HOX proteins, allowing only members of paralogy groups 1 and 2 to bind this sequence cooperatively with Pbx1.
Specific target recognition by the HOXB1–Pbx1 complex is encoded in the Hox homeodomain N‐terminal arm, and can be transferred to HOXB3 by domain swapping
To identify the HOX protein domain responsible for the differential activation of the R3 site in vivo, we carried out a domain‐swap experiment between HOXB1 and a HOX protein unable to activate transcription through the b1‐ARE in combination with Pbx1, i.e. HOXB3. Previous studies had indicated that the specificity of action of HOX genes in vivo resides within the homeodomain and its flanking regions (reviewed in Krumlauf, 1994). In particular, the flexible N‐terminal arm of the homeodomain was shown to be crucial for HOX specificity both in vivo (Lin and McGinnis, 1992; Furukubo‐Tokunaga et al., 1993; Zeng et al., 1993) and in transfected cells (Zappavigna et al., 1994). The conserved ‘YPWM’ motif was identified as an important interaction surface between Hox and Pbx proteins (reviewed in Mann, 1996). Therefore, we decided to replace a region encompassing the ‘YPWM’ motif (FDWM in HOXB1 and FPWM in HOXB3) and the homeodomain N‐terminus in the HOXB3 protein (amino acids 130–201) with the corresponding region of HOXB1 (175–211) (Figure 3A), and to assay the resulting chimeric protein (HOXB3/B1HN in Figure 3A) for transcriptional activation of the pAdMLARE reporter in combination with Pbx1. As shown in Figure 3B, while the wild‐type HOXB3 was unable to transactivate the pAdMLARE reporter, either alone or in combination with Pbx1, the HOXB3/B1HN chimera could activate the reporter together with Pbx1 to a level comparable with that obtained with HOXB1. Accordingly, the HOXB3/B1HN chimera could cooperatively bind to the R3 site in EMSA with Pbx1 (Figure 3C).
To narrow down further the region responsible for the differential activity of HOXB1 and HOXB3 on the R3 site, we constructed an additional chimeric mutant (HOXB3/B1N in Figure 3A), in which the N‐terminal portion of the HOXB3 homeodomain (amino acids 184–201) was replaced by the corresponding region of the HOXB1 homeodomain (195–211). As shown in Figure 3B, the substitution of the 17‐most N‐terminal amino acids of the HOXB3 homeodomain with the corresponding amino acids from HOXB1 is sufficient to allow the HOXB3/B1N chimera to activate transcription in combination with Pbx1 at a level comparable with that of HOXB1. Consistently, the HOXB3/B1N chimera was able to bind cooperatively with Pbx1 the R3 site in EMSA (Figure 3C)
These results indicate that the homeodomain N‐terminus region of HOXB1 is sufficient to confer to HOXB3 the ability to bind cooperatively with Pbx1 at the R3 site and to activate transcription through the b1‐ARE in transfected cells. These data also indicate that the absence of activation by the HOXB3 protein in combination with Pbx1 on the b1‐ARE element is not due to the lack of a functional activation domain on the HOXB3 protein, but rather to the inability of a HOXB3–Pbx1 complex to form on the b1‐ARE R3 site.
The transcriptional activation function of the HOXB1–Pbx1 complex resides mainly in the HOXB1 activation domain
We took advantage of the activity of HOXB1 and Pbx1 on the b1‐ARE to dissect the functional properties of the HOXB1–Pbx1 complex in transcriptional activation. To identify the protein domain(s) responsible for transcriptional activation, we generated a series of deletion mutants in the HOXB1 and Pbx1 proteins (Figure 4A), and tested them for the ability to activate the pAdMLR3 reporter in transfected COS7 cells. Deletion of the N‐terminal region (amino acids 1–155) of HOXB1 (HOXB1HD in Figure 4A) caused a reduction in the activity of the HOXB1–Pbx1 complex of ∼70% with respect to the complex containing the wild‐type HOXB1 (Figure 4B). The HOXB1HD mutant was still able to bind cooperatively with Pbx1 on the b1‐ARE R3 site (Figure 5, lane 8). Conversely, two N‐terminal deletions of Pbx1 (Pbx1aΔ1–140 and Pbx1aΔ1–230 in Figure 4A), lacking one or both of the conserved PBC‐A and PBC‐B domains previously identified in exd/Pbx family members (Burglin and Ruvkun, 1992), led to cooperative activation of the pAdMLR3 reporter activity at levels comparable with, or higher than, those observed with the wild‐type Pbx1 (Figure 4B). Interestingly, the Pbx1aΔ1–230 mutant was able to activate the pAdMLR3 reporter even in the absence of HOXB1 (Figure 4B). These data indicate that the transcriptional activating function of the HOXB1–Pbx1 complex on the b1‐ARE element resides essentially in the HOXB1 N‐terminal region. This region contains a bona fide activation domain, since it was active also in the context of a fusion with the yeast GAL4 DNA‐binding domain (amino acids 1–147) on a GAL4‐responsive reporter in transfected COS7 cells (results not shown).
To test whether binding of either HOXB1 or Pbx1 to DNA is necessary for activation of the pAdMLR3 reporter, we generated and tested DNA‐binding‐defective mutants in both proteins. A HOXB1 mutant lacking helix 3 of the homeodomain (B1Δ236–274 in Figure 4A) did not stimulate the pAdMLR3 reporter basal activity when expressed in combination with Pbx1 (Figure 4C). Similarly, a Pbx1 deletion mutant lacking three amino acids within helix 3 of the homeodomain (Pbx1aΔ283–285, in Figure 4A) showed no activity when co‐expressed with HOXB1 (Figure 4C). When tested in EMSA, neither B1Δ236–274 (not shown) nor Pbx1aΔ283–285 (Figure 5, lane 12) were able to bind cooperatively to the R3 site in association with the wild‐type Pbx1 or HOXB1 respectively. Thus, mutations affecting DNA binding of either HOXB1 or Pbx1 disrupt the formation of a complex in vitro and suppress transcriptional activation in transfected cells.
Cooperative DNA binding by Hox and exd/Pbx proteins has been shown to require the Hox ‘YPWM’ motif, in particular the highly conserved tryptophan residue (reviewed in Mann and Chan, 1996), and the Pbx1 C‐terminal region (Chang et al., 1995; Lu and Kamps, 1996a; Mann and Chan, 1996), which are supposed to mediate protein–protein contacts between the two proteins. To test the role of the ‘YPWM’ motif in the function of the HOXB1–Pbx1 complex, we generated two HOXB1 mutant derivatives, HOXB1ΔWM (Figure 4A) in which the conserved tryptophan and methionine residues were deleted, and HOXB1W177A in which the single tryptophan residue was replaced by an alanine residue. As shown in Figure 4C, both mutations completely abolished the ability of HOXB1 to activate the pAdMLR3 reporter in combination with Pbx1. Consistently, the B1ΔWM and the B1W177A mutants were unable to bind cooperatively with Pbx1 to the b1‐ARE R3 site in EMSA (Figure 5, lanes 10 and 11). To test the role of the Pbx1 C‐terminus, we generated a mutant carrying a deletion of the entire region downstream of the homeodomain (Pbx1aΔ296–430 in Figure 4A). The Pbx1aΔ296–430 mutant was unable to stimulate transcription from the pAdMLR3 reporter in combination with HOXB1 (Figure 4C). Consistently, this mutant did not form a complex with HOXB1 on the R3 site in EMSA (Figure 5, lanes 12 and 13). Production and nuclear localization of the mutants was tested by Western blot analysis of nuclear extracts from transfected cells (not shown).
Taken together, these results indicate that the DNA‐binding domains of both HOXB1 and Pbx1, and the putative protein–protein interaction domains on both proteins, are all individually necessary for the assembly of a cooperative DNA‐binding and transcriptionally active complex.
The C‐terminus of the Pbx1a isoform contains an activation domain which is absent in the Pbx1b splicing variant
As reported above, the deletion of the N‐terminal region of HOXB1 caused a significant but not complete reduction in the transcriptional activity of the HOXB1–Pbx1 complex (HOXB1HD mutant in Figure 4B). This observation led us to assume that the Pbx1 protein could contribute in part to the transcriptional activity of the HOX–Pbx complex, and therefore be responsible for the residual activity of the B1HD–Pbx1 complex on the b1‐ARE element. To identify a potential activation domain within the Pbx1 protein, we focused on the C‐terminal region, since deletions within the N‐terminus did not reduce the activity of the HOXB1–Pbx1 complex (Δ1–140 and Δ1–230 mutants in Figure 4B). In particular, we tested the activity of the Pbx1b isoform, a naturally occurring, alternative splicing variant of the full‐length Pbx1 protein (or Pbx1a), which is 83 amino acids shorter at the C–terminus (Monica et al., 1991). As shown in Figure 6A, co‐expression of Pbx1b and HOXB1 led to activation of the pAdMLR3 reporter at levels of ∼50% of those obtained with Pbx1a, while co‐expression of Pbx1b and the HOXB1 mutant carrying a deletion of the N‐terminal, major activation domain (HOXB1HD) showed no residual activity. These data indicate that the C‐terminal 83 amino acids of the Pbx1a isoform carry an activation domain which is not contained in Pbx1b, and which contributes to the overall transcriptional activity of the HOXB1–Pbx1 complex.
As described above, a Pbx1a deletion mutant lacking amino acids 1–230 (Pbx1aΔ1–230) was able to activate the pAdMLR3 reporter even in the absence of HOXB1 (Figure 4B). To test whether this activity relied on the activation domain located within the C‐terminal region of Pbx1a, we generated a Δ1–230 mutant also of the Pbx1b isoform (Pbx1bΔ1–230), and compared its activity with that of Pbx1aΔ1–230. As shown in Figure 6A, Pbx1bΔ1–230 alone had no significant activity on the pAdMLR3 reporter, and showed an activity consistently lower than that of Pbx1aΔ1–230 in combination with HOXB1. The Pbx1b variant showed the same efficiency of Pbx1a in cooperative binding to the b1‐ARE R3 site in vitro, in combination with either HOXB1 or its N‐terminal deletion HOXB1HD (Figure 5, lanes 6 and 9).
Finally, to prove the existence of an activation domain in the Pbx1a C‐terminus, we fused the 83 amino acid region to the DNA‐binding domain (amino acids 1–147) of the yeast GAL4 transcription factor to generate the GAL4Pb×1aCT chimeric protein. This protein was able to activate transcription from a reporter containing five GAL4‐binding sites (UAS in Figure 6B).
The specificity of action of Hox gene products in vivo is probably achieved through the activity of still ill‐defined co‐factors modulating their transcriptional functions. Requirement for such factors in vivo would account for the relatively relaxed target specificity displayed by Hox proteins in vitro. While Hox gene products are apparently able to regulate transcription in a co‐factor‐independent manner (Zappavigna et al., 1991; Arcioni et al., 1992; Jones et al., 1992, 1993; Pöpperl and Featherstone, 1992), regulation of some target genes in vivo was indeed shown to require the presence of additional factors (Chan et al., 1994; Pöpperl et al., 1995; Gross and McGinnis, 1996). The products of the exd/Pbx genes have been proposed as Hox co‐factors on the basis of genetic analysis and of their ability to modulate DNA binding of Hox proteins in vitro (reviewed in Mann and Chan, 1996). A large amount of data has been gathered on site‐selective and cooperative DNA binding by Pbx and Hox proteins on artificial sites in vitro, and a few studies on the E2A–Pbx oncogene fusion provided evidence for functional cooperativity between the two families of proteins (Lu et al., 1995; Peers et al., 1995; Phelan et al., 1995; Chang et al., 1996; Lu and Kamps, 1996a). Nevertheless, the complex formed by wild‐type Pbx and Hox proteins had never been functionally characterized in terms of its ability to activate transcription cooperatively from known target sequence.
We have shown here that Pbx1 can modulate Hox protein function at the level of transcription, by acting as a co‐factor in the activation of a target element by a restricted class of human HOX proteins. We used as a model a genetically characterized, Pbx‐dependent Hox target, the 148 bp autoregulatory element (b1‐ARE) driving Hoxb‐1 gene expression in the fourth rhombomere of the mouse developing hindbrain (Pöpperl et al., 1995). We show that the activity of a reporter construct containing the b1‐ARE can be stimulated by the HOXB1 protein in transfected cells only upon co‐expression of Pbx1. The b1‐ARE contains three repeated sequences closely related to an in vitro‐selected Pbx1 consensus binding site (Van Dijk et al., 1993; Pöpperl et al., 1995), one of which (R3) was shown previously to bind in vitro a Hoxb‐1–exd complex, and to be crucial for the in vivo activity of the b1‐ARE (Pöpperl et al., 1995). Consistent with these findings, transactivation of the b1‐ARE by the HOXB1–Pbx1 complex is significantly reduced by a mutation within the R3 site, while a multimerized R3 site is sufficient to mediate cooperative activation in a promoter‐ and cell context‐independent fashion. Thus, HOXB1 requires Pbx1 as a transcriptional co‐factor to regulate the activity of the b1‐ARE through the R3 site.
Co‐expression of Pbx1 with a representative variety of HOX proteins showed that the b1‐ARE, or the multimerized R3 site, is transactivated only by members of the paralogy groups 1 and 2, namely HOXA1, HOXB1 and HOXB2. The fact that the b1‐ARE is activated by other proteins besides HOXB1 may reflect the existence of cross‐regulation of this element in vivo by a subset of Hox genes. It is noteworthy, in this respect, that Hoxa‐1 and Hoxb‐2 are both expressed in rhombomere 4 during development (reviewed in Keynes and Krumlauf, 1994), and that ectopic expression of Hoxa‐1 was shown to cause activation of a Hoxb‐1–lacZ reporter construct in the hindbrain of transgenic mice (Zhang et al., 1994). This activation may, however, reflect functional redundancy among paralogous group 1 Hox proteins rather than a requirement for Hoxa‐1 in the onset and maintenance of Hoxb‐1 expression, since Hoxb‐1 is expressed normally in rhombomere 4 in Hoxa‐1 null mouse mutants (Carpenter et al., 1993; Mark et al., 1993). Nevertheless, expression of a b1‐ARE–lacZ reporter in Drosophila embryos is not entirely dependent on the function of the Drosophila gene labial (Pöpperl et al., 1995), suggesting that other Hox genes could indeed participate in the regulation the b1‐ARE in a co‐factor‐dependent manner.
The selectivity of the b1‐ARE, and particularly of the R3 site, appears to depend on its ability to allow the assembly of a Hox–Pbx1–DNA ternary complex only with group 1 and 2 Hox proteins, as indicated by DNA‐binding analysis in vitro. This co‐factor‐mediated, differential site recognition appears to be specified by the extended N‐terminal region of the Hox homeodomain. In fact, the ability to complex with Pbx1 on the R3 site, and to activate transcription through the b1‐ARE, can be transferred to the HOXB3 protein by swapping a region containing only the homeodomain N‐terminal arm. This result is in agreement with previous studies showing that the flexible homeodomain N‐terminus is crucial for determining the specificity of action of Hox proteins in vivo (Lin and McGinnis, 1992; Furukubo‐Tokunaga et al., 1993; Zeng et al., 1993; Chan et al., 1994; Zappavigna et al., 1994), and with previous models derived from in vitro studies on PCR‐selected targets suggesting that the ‘YPWM’/homeodomain N‐terminal region is an important determinant in the site selectivity of different Hox–Pbx complexes in vitro (Chan et al., 1996; Chang et al., 1996; Lu and Kamps, 1996b). In the context of the HOXB1–Pbx1 complex, the ‘YPWM’ motif therefore provides only a neutral, protein–protein interaction function, as previously suggested by yeast two‐hybrid experiments (Johnson et al., 1995), without necessarily contributing to the DNA‐binding specificity. Indeed, mutations in the domains involved in protein–protein contacts, either in HOXB1 (deletion or mutation of the critical tryptophan in the FDWM region) or in Pbx1 [deletion of the C‐terminus downstream from the homeodomain (see Chang et al., 1996; Lu and Kamps, 1996b)] disrupt the formation of the ternary complex in vitro and abolish transcriptional activation altogether. Furthermore, we show that transcriptional activation by the HOXB1–Pbx1 complex requires intact DNA binding functions by both HOXB1 and Pbx1 proteins. Thus, unlike the MATa1–MATα2 homeodomain protein complex, where a MATα2 mutation impairing DNA binding is still able to form a functional complex with MATa1 on DNA in vivo (Vershon et al., 1995), in the case of the HOXB1–Pbx1 complex, tethering of either protein onto DNA through protein–protein interactions is not sufficient for activity.
The transcriptional activity of the HOXB1–Pbx1 complex on the b1‐ARE was exploited to characterize functional domains in both proteins. The main transcriptional activation domain within the complex was mapped to the N‐terminal region of the HOXB1 protein. In fact, deletion of the 155 N‐terminal amino acids of the HOXB1 protein considerably reduced the transcriptional activation strength without affecting cooperative DNA binding. The N‐terminus of the HOXB1 protein activates transcription also when fused to a heterologous DNA‐binding domain. However, deletion of the HOXB1 N‐terminal domain did not abolish transcriptional activity completely, leading us to speculate that Pbx1 could also contribute to the activation strength of the complex. A transcriptional activation domain was indeed mapped within the 83 most C‐terminal amino acids of the Pbx1 protein. This region is absent in the Pbx1b alternative splicing variant, which reproducibly displayed lower levels of activation when compared with the full‐length protein (Pbx1a) in combination with HOXB1. The C‐terminal region of Pbx1a is rich in Ser/Thr residues, and is able to activate transcription when fused to a heterologous DNA‐binding domain. Interestingly, a Ser/Thr‐rich region located within the C‐terminus of the LFB1 (HNF1) homeodomain protein has been reported previously to be implicated in transcriptional activation, and to account for 50% of the LFB1 activity (Nicosia et al., 1990).
The C‐terminal region of Pbx1a is apparently not functional as an activation domain in the uncomplexed Pbx1a protein, which is unable to activate transcription even through a PCR‐derived optimal consensus binding site (Van Dijk et al., 1993; Mann and Chan, 1996). Deletion of the N‐terminal 230 amino acids, containing the two conserved PBC‐A and PBC‐B regions (Burglin and Ruvkun, 1992), uncovers the function of the Pbx1 C‐terminal activation domain, since the Pbx1Δ1–230 mutant is able to stimulate the b1‐ARE R3 reporter activity significantly, even in the absence of HOXB1. A deletion of the N‐terminal 232 amino acids of Pbx1 has been reported previously to bind to a Pbx consensus sequence with higher affinity with respect to its wild‐type counterpart in the absence of Hox proteins (Lu and Kamps, 1996b), suggesting that the activity of the Δ1–230 mutant on the R3 reporter could be due to an increased binding affinity for the R3 site. Transcriptional activation by the Pbx1aΔ1–230 mutant is sustained by the C‐terminal activation domain, since the same N‐terminal deletion in the shorter Pbx1b variant, Pbx1bΔ1–230, does not significantly activate transcription. The Pbx1b alternatively spliced variant may, therefore, represent a less active form of Pbx1, capable of forming a DNA‐bound complex with HOX proteins like the longer Pbx1a variant, but leading to lower levels of transcriptional activation. Pbx1b could antagonize and/or substitute for the longer and more active Pbx1a isoform in different tissues or body regions, allowing a fine tuning of the activity of Hox–Pbx complexes on their targets.
Materials and methods
Protein expression and reporter plasmids
All expression constructs are derivatives of the SV40 promoter‐driven expression vector pSG5 (Green et al., 1988). Hox expression vectors were described previously (Zappavigna et al., 1991, 1994; Arcioni et al., 1992; Guazzi et al., 1994), with the exception of pSGA1 and pSGC6. pSGA1 was generated by ligating a BamHI–XbaI Klenow‐filled Hoxa–1 cDNA coding sequence obtained from the pHoxa‐1(HD+) plasmid (Phelan et al., 1995) into the BamHI–BglII Klenow‐filled sites of pSG5. pSGC6 was obtained by cloning a PCR‐amplified BamHI insert containing the full‐length HOXC6‐coding sequence from the pCT‐H3C plasmid (Arcioni et al., 1992) into the BamHI site of pSG5. pSGPbx1a was generated by ligating a HindIII‐filled–EcoRI‐filled cDNA fragment encompassing the complete coding sequence of Pbx1a (Van Dijk et al., 1993) into the filled BamHI site of pSG5. pSGPbx1b was reconstructed by substituting the E1A N‐terminal region of E1APbx1b (kind gift of F.Blasi) at a unique NcoI site with the N‐terminal region of Pbx1a. Pbx1a deletion mutants (pSGpbx1aΔ1–140, Δ1–230, Δ296–430 and Δ283–285) were generated by ligating BamHI PCR‐amplified and deleted cDNA fragments in the BamHI site of pSG5. To generate pSGPbx1aΔ1–140 and Δ1–230, an ATG start codon was introduced in‐frame upstream of amino acids 141 and 231, respectively. To generate pSGPbx1aΔ296–430, a TGA stop codon was added downstream of amino acid 295. pSGpbx1bΔ1–230 was generated using the same PCR primer as for Pbx1aΔ1–230 and cloning into the EcoRI–BamHI sites of pSG5. pSGB1ΔWM and W177A were obtained by PCR mutagenesis, covering the full‐length HOXB1, as BamHI inserts into pSG5. In pSGB1ΔWM, amino acids W177 and M178 of the HOXB1 ‘YPWM’ motif were deleted and in pSGB1W177A amino acid W177 was replaced by an alanine. pSGB1Δ236–274 was obtained by removing the SacI–PvuII fragment from pSGB1 and re‐ligating the vector after blunting of the SacI site with T4 DNA polymerase. This generates a deletion in the HOXB1 cDNA extending from the C‐terminal region of the homeodomain (helix 3/4) to part of the adjacent C‐terminus of the protein, without any amino acid substitution. pSGHOXB3/B1HN was generated by replacing the ‘YPWM’/homeodomain N‐terminal region of HOXB3 (amino acids 130–201) with the analogous region of HOXB1 (amino acids 175–211). The mutated cDNA was obtained by PCR amplification and cloned as a BamHI insert into pSG5. pSGHOXB3/B1N was generated by replacing the homeodomain N‐terminal region of HOXB3 (amino acids 184–201) with the analogous region of HOXB1 (amino acids 195–211). The mutated cDNA was obtained by PCR amplification and cloned as a BamHI insert into pSG5. pSGB1HD was generated by PCR amplification of a region comprising amino acids 155–286 and introducing a methionine residue in position 154. pSGGAL1–147 was obtained by cloning a BamHI–BglII insert containing the DNA‐binding domain of yeast GAL4 (amino acids 1–147) into pSG5. pGAL1–147pb×1CT was generated by PCR amplifying a region comprising amino acids 348–430 and cloning in‐frame with the GAL4 1–147 protein at the EcoRI site of the pGAL1–147 vector. The correctness of all cloned PCR products was verified by DNA sequencing, and the expression of all proteins was tested in a rabbit reticulocyte system.
The luciferase reporter construct pML is described elsewhere (Pöpperl and Featherstone, 1992). pAdMLARE (generous gift of Mark S.Featherstone) contains the AvaI–HaeII R4 enhancer of Hoxb‐1 (Pöpperl et al., 1995) cloned as a PCR‐amplified HindIII–XhoI fragment into pML. In pAdMLAREmR3, the Pbx consensus site of R3 in the R4 enhancer TGATGGAT was changed to TGTCGACT. pAdMLR3 contains a trimer of R3 of the b1‐ARE enhancer cloned as a BamHI–HindIII fragment into pML. The same trimer of R3 was cloned into the BamHI site of the pT81luc luciferase reporter vector. The sequence of the 30 bp oligo used to generate the trimer is 5′‐GATCCGGGGGGTGATGGATGGGCGCTGGGA‐3′. The pTUAS luc GAL‐4 reporter construct was described in Zappavigna et al. (1996).
Protein production and DNA‐binding assays
Pbx and HOX proteins were produced in vitro from the corresponding pSG5‐derived expression vectors using a T7 polymerase‐based coupled transcription–translation reticulocyte lysate system (Promega, Madison, WI) according to recommended conditions. HOX and Pbx proteins were translated separately in the presence of [35S]methionine and quantitated after SDS–PAGE using a PhosphorImager (Molecular Dynamics). The amounts of proteins were normalized for the methionine content of each protein.
Gel retardation analysis was performed by pre‐incubating the in vitro synthesized proteins for 30 min on ice in 15 μl of binding buffer (75 mM NaCl, 6% glycerol, 10 mM Tris–HCl pH 7.6, 1 mM EDTA), together with 2 μl (0.5 ng, 5×104 c.p.m.) of 32P‐labelled oligonucleotide R3 probe 5′‐GATCCGGGGGGTGATGGATGGGCGCTGGGA‐3′. The incubation mixture was resolved by electrophoresis on a 6% polyacrylamide gel in 0.25× TBE at 10 V/cm. Gels were dried and exposed to a Kodak X‐AR film at −70°C. The amount of reticulocyte lysate added to each incubation mixture was adjusted in order to normalize for the translated protein content.
Cell culture and transfection
COS7 cells were maintained in Dulbecco‘s modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (Gibco), 100 IU/ml of penicillin and 100 μg/ml streptomycin. Cells were transfected by calcium phosphate precipitation (Di Nocera and Dawid, 1983). In a typical transfection experiment, 8 μg of reporter plasmid, 4–8 μg of expression construct and 0.2 μg of pCMV‐β‐gal as an internal control were used per 9 cm dish. Cells were harvested 48–60 h after transfection, lysed and assayed for luciferase and β‐galactosidase expression as described previously (Zappavigna et al., 1994).
Thanks are due to F.Blasi and E.Boncinelli for critically reading the manuscript, to C.Murre for providing the full‐length cDNA of Pbx1, and to M.S.Featherstone for the pHoxa‐1 and the pAdMLARE plasmids. The expert assistance of V.Ganazzoli in typing the manuscript is hereby gratefully acknowledged. This work was supported by grants from the Italian Association for Cancer Research (AIRC) and the CNR (Progetto Finalizzato ACRO).
- Copyright © 1997 European Molecular Biology Organization