Residues of BCL9 and B9L required for the binding to β-catenin
The homology domain 2 (HD2) of BCL9 proteins mediates their binding to Armadillo/β-catenin in vitro
and in vivo
]. To identify relevant residues within HD2 of human BCL9, we mutagenized amino acids conserved between BCL9, BCL9-2/B9L and Lgs (Fig. ), and tested the ability of these mutants to bind the Armadillo repeat domain (ARD) of β-catenin. Additionally, we generated two mutants (L363F, L366K) that mimic loss-of-function alleles of Drosophila lgs
; Fig. ). Using pull-down assays, we found that each of these mutants reduced the binding of HD2 to GST-ARD either partially, or completely (Fig. ). In particular, binding was eliminated by L366K, consistent with the results for Lgs [20
], but only reduced by L363F (Fig. ). Binding was not affected by mutations in residues flanking HD2 (P348G, E377Q; Fig. ). We conclude that the three leucine residues 351, 366, 373 and the HRE cluster 358–360 are critical for the binding between BCL9 and β-catenin, consistent with the recent structural analysis [40
]. For the following experiments, we decided to use L363F (L>F) as a partial, and L366K (L>K) as a complete loss-of-function mutant.
Figure 1 Mutants of BCL9 proteins that cannot bind to β-catenin. A, Sequence alignments of the HD2 domains of Lgs, BCL9 and BCL9-2/B9L (residue numbers in the latter two refer to mouse or human HD2 sequences, which are identical). Individual conserved (more ...)
To test the binding of these mutants to β-catenin in vivo, we introduced them into full-length FLAG-tagged BCL9, and also generated the corresponding mutants in BCL9-2 (L408F, L411K; Fig. ), and co-expressed these with HA-tagged β-catenin in HEK 293T cells. As expected, HA-β-catenin readily co-immunoprecipitated with wt BCL9, and also with L363F, but not with L366K (Fig. ). Likewise, HA-β-catenin co-immunoprecipitated with BCL9-2 and L408F, but not with L411K (Fig. ). The converse experiments (immunoprecipitations of HA-β-catenin) showed the same results (not shown). This is consistent with our in vitro binding data (Fig. ) and confirms that the L366 and L411 residues in BCL9 and BCL9-2, respectively, are critical for their binding to β-catenin (though the L>F substitutions at L363 and L408 are compatible with normal β-catenin binding, at least at high expression levels). Importantly, these results indicate that HD2 is the only domain in BCL9 and BCL9-2 that binds to β-catenin in vivo.
Different activities of overexpressed BCL9 and BCL9-2 in colorectal cancer cell lines
To see whether the L>F and L>K mutations affected the subcellular distributions of BCL9 or BCL9-2, we expressed these proteins in SW480 colorectal cancer cells whose Wnt pathway activity is high, due to mutation of APC [41
]. As expected from previous studies in other mammalian cells [24
], BCL9 was distributed throughout the cytoplasm and nucleus (Fig. ) while BCL9-2 was strictly nuclear (Fig. ; the subcellular distribution of B9L was indistinguishable from that of BCL9-2; not shown). Both proteins showed a tendency to form puncta, however, in neither case could we detect significant differences in the subcellular distributions between wt and mutants (not shown). We also used photobleaching experiments (essentially as described [34
]) to confirm that GFP-tagged BCL9(HD1+2) is a highly dynamic nuclear-cytoplasmic shuttling protein, like Lgs [34
], but we were unable to detect any significant differences in the shuttling rates between wt and mutant proteins (not shown).
Figure 2 Subcellular distribution and transactivation potential of wt and mutant BCL9 and B9L. A and B, SW480 cells overexpressing FLAG-BCL9 or FLAG-BCL9-2, fixed and stained with antibodies against β-catenin and FLAG, as indicated in panels (the staining (more ...)
Given the adaptor role of Lgs/BCL9 [20
], full activity would require both an intact HD2 and Pygo-binding domain (called homology domain 1, or HD1). To test this, we overexpressed the L366K and L411K mutants, and BCL9 and BCL9-2 mutants with small internal HD1 deletions (ΔHD1), as well as ΔHD1 L366K and ΔHD1 L411K double-mutants in colorectal cancer cells, and we examined their effects on the Wnt pathway activity of these cells by using TCF reporter activity (TOPFLASH) as a specific and quantitative read-out [39
]. We expected one of three outcomes for these ligand-binding mutants: namely, a mutant could be (1) less active than its wt counterpart, implying that the corresponding ligand is required for the function of the wt protein; (2) dominant-negative (DN), suggesting that the mutant protein sequesters a ligand required for the function of the wt protein; or (3) as active as the wt, indicating that the corresponding ligand is not required for the function of the wt protein, at least after overexpression. In particular, we were interested in the behaviour of the double-mutants which cannot bind to either of their known ligands, β-catenin or Pygo: if these were DN, then this would suggest an additional functional ligand of BCL9 proteins.
In SW480 cells, we found that overexpression of wt BCL9 did not affect TOPFLASH activity (Fig. ), consistent with previous results of BCL9 overexpression in Wnt-stimulated HEK 293 cells [24
]. Likewise, we could not detect any DN effects with any of the BCL9 mutants (Fig. ), which were expressed at similarly high levels as the wt (not shown), but this is most likely due to the relatively low transfection efficiency of these cells (see also below, for the behaviour of these mutants in HEK 293 cells whose transfection efficiency is higher). In contrast, overexpressed wt BCL9-2 potentiated TCF reporter activity approximately two-fold in SW480 cells (Fig. ), consistent with previous results with B9L [24
]. However, none of the BCL9-2 mutants displayed any significant stimulatory activity (Fig. ). This indicates that the function of BCL9-2 in TCF-mediated transcription of these cells depends on its ability to bind to β-catenin and Pygo proteins.
We also tested wt and mutant BCL9 proteins in HEK 293 cells transiently stimulated with Wnt3A, since this provides a more sensitive assay (partly due to the higher transfection efficiency of these cells). We optimized this assay, minimizing the Wnt exposure time as much as possible, which allowed us to focus on the primary transcriptional Wnt response and to avoid secondary knock-on effects. We thus found that culturing 293 cells in the presence of Wnt3A for 6 hours resulted in a 2× increase in TOPFLASH activity (Fig. ). Like in SW480 cells, overexpression of wt BCL9 did not further increase this activity (Fig. ). However, overexpression of L366K, but not of ΔHD1, showed consistently a slight DN effect in reducing TOPFLASH activity compared to the control (Fig. ). This DN effect did not appear to be due to the sequestration of Pygo, since it was still detectable with the overexpressed double-mutant L366K ΔHD1 (Fig. ). This suggests that this double-mutant BCL9 may sequester a functionally important ligand other than β-catenin and Pygo (see below).
As in SW480 cells, overexpression of BCL9-2 in Wnt3A-stimulated HEK 293 cells resulted in a 2× increase of TOPFLASH activity (Fig. ). Its ΔHD1 mutant version also stimulated the TOPFLASH reporter to some extent (by ~50%), whereas L411K showed a slight DN effect, like its L366K counterpart (Fig. ), although the ΔHD1 L411K double-mutant appeared neutral and inactive (Fig. ). It thus seemed that the latter, unlike its BCL9 counterpart, failed to behave as a DN, but this may have been due to technical reasons (e.g. the expression levels of ΔHD1 L411K were significantly lower than those of L366K ΔHD1; not shown; see also Fig. ). Taken together, these results indicated that both BCL9 and BCL9-2 rely on their binding to β-catenin and Pygo, and possibly on an additional unknown ligand (at least in the case of BCL9), for their function in mediating Wnt-dependent transcription.
BCL9 and B9L are Wnt target genes
We also tested the effects of overexpressed wt and mutant BCL9 proteins on the Wnt inducibility of endogenous TCF-mediated transcription. Two of the best-established TCF target genes in intestinal epithelial and colorectal cancer cells are c-myc
]. We stimulated HEK 293 cells for 6 hours with either Wnt3A-conditioned medium, or with LiCl (an inhibitor of GSK3β), and we monitored the transcript levels of these genes by RT-qPCR. As internal controls, we measured the expression levels of TATA-box binding protein (TBP), β-actin and the house-keeping gene HPRT
. As expected, we observed a 2–2.5× upregulation of c-myc
transcripts in response to both Wnt3A and LiCl stimulation (Fig. ), while the transcript levels of the three internal control genes did not change after these treatments. Interestingly, we also discovered that the transcript levels of both BCL9
were upregulated to a similar degree (~2× and ~2.5×, respectively) after Wnt stimulation (Fig. ). Therefore, BCL9
are themselves Wnt-responsive, and are thus likely to represent TCF target genes.
Figure 3 BCL9 and B9L are Wnt-inducible genes. A, Transcript levels of BCL9 and B9L in comparison to AXIN2, as measured by RT-qPCR, after induction of Wnt pathway activity in HEK 293 cells by addition of Wnt3A-conditioned medium, or 20 mM LiCl, for 6 hours. Statistical (more ...)
Next, we tested the effects of overexpressed wt and mutant BCL9 proteins on the Wnt-induced expression levels of c-myc and AXIN2. As expected from the TOPFLASH assays (Fig. ), overexpressed wt BCL9 did not stimulate the Wnt-induced expression of these genes further (Fig. ), however overexpressed BCL9-2 synergized with Wnt3A to increase the transcript levels of AXIN2 and c-myc by ~30% (Fig. ). Likewise, the BCL9 mutants behaved the same as in the TOPFLASH assays (Fig. ), with L366K and L366K ΔHD1 displaying clear DN effects (Fig. ). In the case of BCL9-2, each mutant was inactive (i.e. no synergy with Wnt stimulation was observed), and we detected slight albeit statistically significant DN effects of ΔHD1 and L411K ΔHD1 for one target gene each (Fig. ). Once again, the DN effect of the double-mutant suggested that BCL9-2, like BCL9, may have an additional functionally important ligand. Taken together, these results fully confirmed those from the TOPFLASH assays, and suggested that the functions of BCL9 proteins regarding the transcriptional Wnt response of endogenous target genes are mediated by Pygo and β-catenin, and possibly also involve a third ligand (see below).
Given the Wnt-responsiveness of BCL9
in HEK 293 cells, we wondered whether these genes might be hyperexpressed in colorectal cancer cell lines. We thus used RT-qPCR to monitor their transcript levels in SW480 cells, and also in HCT116 cells whose elevated Wnt pathway activity is due to an activating mutation in β-catenin [41
]. We found that the BCL9
transcript levels were comparable in both cell lines to those in Wnt-stimulated HEK 293 cells, whereas the B9L
transcript levels were considerably higher in the colorectal cell lines (35× and 50× higher in SW480 and HCT116 cells, respectively, compared to unstimulated HEK 293 cells; Fig. ). These elevated B9L
transcript levels are similar to those of BCL9
in the same cell lines (Fig. ). As a comparison, we also measured the expression levels of AXIN2
, and of CD44
whose expression in the intestinal epithelium is controlled by TCF and APC [15
]. As expected, all three genes are highly expressed in SW480 cells (25×, 6× and ~100× higher, respectively, than in unstimulated HEK 293 cells; Fig. ), although in HCT116 cells, only CD44
is overexpressed, while the transcript levels of AXIN2
are lower than in Wnt-stimulated 293 cells (Fig. ), in the case of AXIN2
due to epigenetic silencing by DNA methylation which has been observed in colorectal cancer cell lines with microsatellite instability, such as HCT116 [43
]. Thus, B9L
and to some extent BCL9
are hyperexpressed in colorectal cancer cells compared to unstimulated HEK 293 cells – in the case of B9L
, more than an order of magnitude, much like some of the other TCF target genes.
Both BCL9 and B9L are required for Wnt pathway activity
BCL9-2/B9L has previously been shown to be a positive regulator of Wnt signaling in Wnt-stimulated mammalian cells, and in SW480 colorectal cancer cells [24
]. This is consistent with the DN behaviour of its mutant version, L411K, which cannot bind to β-catenin (Fig. ). Given that the corresponding mutant of BCL9 (L366K) showed similar DN effects (at least in Wnt-stimulated HEK 293 cells), we asked whether BCL9 is also required for efficient Wnt pathway activity in SW480 cells.
We thus used siRNA-mediated depletion to test this, and also to assess the relative contributions of BCL9 and B9L to TCF-mediated transcription in SW480 cells. This allowed us to reduce the transcript levels of BCL9 to ~50%, and those of B9L to ~30% of their normal levels in these cells (Fig. ); under the same conditions, we were able to deplete the levels of β-catenin transcripts to ~25% (not shown). We found that the expression levels of c-myc and AXIN2 were reduced after depletion of BCL9 to a similar degree as after depletion of B9L, or after depletion of β-catenin transcripts (Fig. ). B9L expression was also reduced after BCL9 depletion (which thus mimics a double knock-down of both paralogs, albeit a partial one) although the converse was not true, possibly because this gene is only mildly Wnt-inducible, and is less hyperactive than B9L in colorectal cancer cell lines (see Fig. ). These results indicate that BCL9 is required for efficient TCF-mediated transcription in these colorectal cancer cells, similarly to B9L and β-catenin.
Figure 4 BCL9 is required for the transcription of endogenous Wnt target genes. Transcript levels of BCL9, B9L, c-MYC and AXIN2, as measured by RT-qPCR, in (A) SW480 cells or (B) Wnt-induced HEK 293 cells (as in Fig. 3), treated with siRNA to deplete β-catenin, (more ...)
We also depleted BCL9 transcripts in Wnt-stimulated HEK 293 cells, to test the role of BCL9 in TCF-mediated transcription under transient Wnt signaling conditions. Again, we found marked reductions of c-myc, AXIN2 and B9L after BCL9 depletion, similarly as after depletion of B9L (Fig. ). Indeed, the effects on Wnt target gene expression appeared stronger in these cells, most likely due to their higher transfection efficiency (i.e. a higher fraction of cells experienced RNAi-mediated depletion). We conclude that BCL9, like B9L, is a positive regulator of Wnt-induced transcription in HEK 293 cells.
The C-termini of BCL9 proteins are required for their function in Wnt signaling
Previous work revealed a DN effect of a C-terminal deletion of B9L in colorectal cancer cells, suggesting a role of the C-terminus of B9L in its Wnt response [24
]. Likewise, our result that the BCL9 double-mutant behaved as a DN, despite being defective in Pygo and β-catenin binding (see above), suggested that this protein may bind to an additional ligand required for its function in Wnt signaling. We thus generated a C-terminal deletion (ΔC) of BCL9, to test whether this would behave as a DN in Wnt-stimulated HEK 293 cells. Furthermore, we introduced the L366K and ΔHD1 single and double mutations into this C-terminal truncation, and also into the ΔC truncation of B9L [24
], to ask whether the DN effect(s) would be abolished if β-catenin and/or Pygo binding was eliminated.
Indeed, the ΔC mutant of BCL9 showed a mild DN effect on the TOPFLASH activity of Wnt-stimulated HEK 293 cells (Fig. ). By comparison, B9L ΔC behaved as a more potent DN [24
], reducing the TOPFLASH values almost to those measured in unstimulated cells (Fig. ), possibly because it could compete more effectively with endogenous B9L (whose expression level is far lower than that of BCL9; Fig. ) and is targeted to the nucleus more efficiently than BCL9 (Fig. ). Interestingly though, in both cases, the DN effects remained detectable in the ΔC ΔHD1 double-mutants, but was eliminated in the ΔC L366K/L411K double-mutants, and in the triple-mutants (Fig. ), all of which are defective in β-catenin binding. These results indicated that the C-terminus of BCL9, like that of B9L, harbours a function in TCF-mediated transcription, thus explaining why the ΔHD1 L366K and ΔHD1 L411K double-mutants behave as DNs (Fig. ). Furthermore, the results from all three types of double-mutants imply that the putative sequestration of β-catenin (by ΔC ΔHD1) or C-terminal ligand (by ΔHD1 L366K and ΔHD1 L411K) reduced Wnt-mediated transcription, whereas sequestration of Pygo (by ΔC L366K and ΔC L411K) did not affect this process.
Figure 5 Requirement of the C-terminus of BCL9 for Wnt-induced transcription. Transcript levels of c-MYC and AXIN2, as measured by RT-qPCR, in Wnt-induced HEK 293 cells (as in Fig. 3) after overexpression of wt or mutant (A) FLAG-BCL9 or (B) FLAG-BCL9-2. Statistical (more ...)