Abstract

Results and Discussion

While expression of either Gli1 or Gli2 in the epidermis of transgenic mice induces BCCs (Fig. 1a), we have observed a considerable delay in the appearance of Gli-dependent tumors. Analysis of transgenic mice expressing Gli2 revealed an average latency of 7 mo before tumor appearance (Fig. 1b). We ruled out changes in transcription of the transgene with age as a cause of the tumors, as similar levels of RNA are seen in both age groups as measured by quantitative PCR (Fig. 1c). This suggested the existence in keratinocytes of additional processes, whose loss or dysregulation is required to permit Gli activity and direct tumor formation. Our previous studies indicated that differential accumulation of Gli protein plays an important role in restricting Shh target gene induction in interfollicular epithelium (Oro and Higgins 2003). Indeed, we detected no transgenic Gli protein in normal skin, whereas we found high levels in the BCC tumors (Fig. 1d). Cultured explants of primary keratinocytes from normal skin also contained little detectable Gli protein (Fig. 1e). However, treatment of these cells with the proteasome inhibitor MG132 caused full-length Gli2 protein to accumulate many fold within 3 h, confirming the presence of an active Gli2 protein destruction mechanism. These data support the conclusion that proteasome-dependent Gli protein destruction underlies the latency in Shh target gene response.

Open in a separate window

Figure 1.

Onset of BCC formation correlates with Gli protein accumulation. (a) Clinical appearance of focal BCCs induced in transgenic animals expressing Gli2 in the skin epithelium with the keratin 5 promoter. (b) Bar graph showing representative onset of tumors in K5-Gli2 mice. (c) Quantitative PCR of Gli2 RNA levels from skin of wild-type or Gli2 transgenic animals of indicated age. Gli2 RNA levels in each sample were normalized to those of GAPDH. Error bars are standard error of the mean (SEM). (d) Immunofluorescence with anti-HA (red) antibody showing protein accumulation only in BCC tumor (arrowhead), not in interfollicular epidermis (arrow). (Green) Anti-laminin 5; (blue) Hoechst. (e) Western blot of lysates from explanted K5-Gli2 keratinocytes demonstrating the rapid accumulation of Gli2 protein with the addition of the proteasome inhibitor MG132, but not with DMSO. (f) Autoradiogram of ³⁵S-labeled Gli1 mixed with Xenopus oocyte extract. Gli1 is degraded in a proteasome-dependent manner with a half-life of ~40 min. (g) Western blot of HA-Gli1 in NIH 3T3 cells showing rapid, proteasome-dependent destruction. The nonspecific band demonstrates equal protein loading. (h) HA-Gli1 protein is rapidly degraded (C) via a process inhibited by proteasome inhibitors (M), but not cathepsin or lysosomal inhibitors E64 (E) or chloroquine (Q), respectively. The efficacy of the E64 and chloroquine used in this experiment was confirmed by their ability to inhibit ligand-dependent lysosomal destruction of EGFR in primary human keratinocytes.

To study the molecular mechanisms that govern Gli protein degradation, we chose to focus our initial studies on Gli1, which, unlike Gli2 or Gli3, is primarily a transcriptional activator and is not processed to a repressor form (Dai et al. 1999; von Mering and Basler 1999). In this way, Gli protein function and degradation could be examined independently of proteolytic processing and transcriptional repressor regulation. We tested Gli1 stability in a variety of in vitro settings and found that Gli1 is degraded by the proteasome. In Xenopus egg extracts, a system where the ubiquitin-proteasome system (UPS) is known to be active to control β-catenin and IκB stability (Winston et al. 1999; Margottin-Goguet et al. 2003), ³⁵S-labeled Gli1 protein is destroyed in a proteasome-dependent manner, with a half-life of 40 min (Fig. 1f). Similar kinetics are seen in a variety of cultured normal and cancer cells, including the Shh-responsive NIH 3T3 cells (Fig. 1g; Taipale et al. 2000). We ruled out degradation of Gli1 by other mechanisms such as lysosomal degradation (Dai et al. 2003), as cathepsin and lysosome inhibitors (E64 and chloroquine, respectively) had no effect on Gli levels (Fig. 1h). The efficacy of these inhibitors was confirmed in primary human keratinocytes where they inhibit the EGF-dependent lysosomal destruction of EGFR (Fig. 1h). These data provide strong support for destruction of vertebrate Gli1 by the UPS.

To identify signals that allow Gli1 to interact with the UPS, we were guided by the previous finding in Drosophila that the βTrCP locus is required for Ci processing (Jiang and Struhl 1998). The degron DSGXXS, recognized by βTrCP, is present in vertebrate regulatory proteins β-catenin, IκBα/β, and Emi1 (Spencer et al. 1999), although it is absent from Ci. In spite of this, we have identified a C-terminal motif, DSGVEM, that is conserved in chordate Gli homologs and vertebrate Gli1 and Gli2 proteins (Fig. 2a). To determine if the DSGVEM motif of human Gli1 mediates association with βTrCP1, reciprocal immunoprecipitations were performed from NIH 3T3 cells transfected with myc-tagged βTrCP1 and HA-tagged Gli1 (Fig. 2b). Gli1 protein lacking the DSGVEM motif (Gli1ΔD_C) failed to associate with βTrCP and exhibited delayed degradation kinetics (Fig. 2c). Levels of βTrCP appeared to be limiting for Gli1 degradation, as increasing the levels of βTrCP protein significantly decreased steady-state levels of Gli1 protein (Fig. 2d). Consistent with its role as an E3 ligase, βTrCP association with Gli1 facilitated ubiquitination. In ubiquitin coimmunoprecipitation assays, ubiquitinylated Gli1 (ΔN398), but not Gli1ΔD_C (ΔN398), could be detected in the presence of overexpressed βTrCP1 (Fig. 2e). Previous studies have shown that Protein kinase A (PKA) can enhance βTrCP-dependent Ci cleavage in Drosophila (Wang et al. 1999; Jia et al. 2004). We saw similar effects on Gli1, as inhibition of PKA impeded destruction (Supplementary Fig. 2a), and Gli1 constructs lacking consensus PKA sites in the C terminus failed to bind βTrCP and exhibited delayed destruction kinetics (Supplementary Fig. 2a,b). These data demonstrate that degron D_C mediates Gli destruction via the βTrCP-ubiquitin ligase complex.

Open in a separate window

Figure 2.

Degron D_C (DSGVEM) mediates Gli1 destruction via the βTrCP destruction complex. (a, left) Alignment of chordate Gli sequences showing conserved D_C sequence. The box details key residues that bind βTrCP. (Right) βTrCP-binding sequences from other vertebrate proteins. (b) Reciprocal coimmunoprecipitation of HA-Gli1 or HA-Gli1ΔD_C with myc-βTrCP. Note the lack of βTrCP binding in the mutant. The characteristic mobility shift of immunoprecipitated Gli1 is not an artifact of βTrCP overexpression as this shift is observed even in its absence. (c) Degradation of transfected Gli1 in NIH 3T3 cells. Note the small but significant delay in destruction kinetics of the mutant versus wild-type protein. The densitometry of both assays is shown to the right and is representative of three independent experiments. Equal sample loading and transfer was confirmed by post-staining of the experimental membranes with Coomassie blue. The 150-250-kDa region of the membranes is shown. (d) Western blot of Gli1 or Gli1ΔD_C and increasing amounts of transfected βTrCP. Note the decreased steady-state levels of wild-type, but not mutant Gli1. The difference is quantified below and is representative of three independent experiments. Error bars are standard error of the mean (SEM). (e) Coimmunoprecipitation assay of 6X-His-tagged ubiquitin and HA-Gli1ΔN398 containing degron D_C or HA-Gli1ΔN398ΔD_C mutant. Ubiquitinylated Gli (top panel) is detected in the wild-type Gli1 C terminus, but not the ΔD_C mutant, in the presence of βTrCP (bottom panel).

While βTrCP-dependent degradation clearly plays a role in Gli1 destruction, the Gli1ΔD_C mutation only partially altered the destruction kinetics of Gli1 protein in cultured cells. At 3 h after cycloheximide addition, destruction of Gli1ΔD_C was decreased by only ~25% relative to wild-type Gli1 (47.1% ± 6% vs. 21% ± 5%, Avg. ± SEM) (Fig. 3b). This argued that additional signals control Gli1 degradation. Through focused mutagenesis, we found that a small deletion of the N terminus further stabilized Gli1. As with the D_C degron, degron D_N mutations (GliΔN1-116, referred to as Gli1ΔD_N) alone had modest effects on Gli destruction kinetics in vitro (3 h: 40.1% ± 6% vs. 21% ± 5%, Avg. ± SEM) (Fig. 3b). However, Gli1 lacking both degrons (double mutant; Gli1ΔD_NΔD_C became stable, possessing destruction kinetics similar to that with addition of proteasome inhibitor (Figs. ​(Figs.3b,3b, ​,1g).1g). This argued for an additional degron in the N terminus. Further mutagenesis narrowed the region containing the degradation signal to residues 51-116 (Supplementary Fig. 3). This region contains a stretch of highly conserved residues present in all vertebrate Gli genes and in Drosophila Ci (Fig. 3a), suggesting that the destruction signal may be found in many Gli proteins.

Open in a separate window

Figure 3.

Degron D_N mediates Gli destruction independent of D_C or Sufu function. (a) Alignment showing the conserved N-terminal region containing degron D_N. A solid line indicates the most highly conserved region that is deleted in the D_N mutant, while the boxed area shows the Sufu-binding site, which is retained in the D_N mutant. (b) Destruction assays of HA-Gli1 in NIH 3T3 cells showing the effects of single D_C, D_N, and double mutants in comparison to wild-type (WT) Gli1 in the presence and absence of MG132. The densitometery of blots is shown to the right and is based on three independent experiments. Note that results are plotted on the base 2 logarithmic scale. Error bars are standard error of the mean (SEM). (c) Coimmunoprecipitation of wild-type and mutant Gli with βTrCP. Note that the D_N mutation does not affect the binding of βTrCP to degron D_C. (d) Changes in levels of green fluorescent protein (EGFP) fused to Gli1 N-terminal residues (top), or EGFP (bottom), in the presence of cycloheximide (left) or MG132 (right). The amount of fusion protein is identical at t = 0, but the exposure time for the left and right panels differs to avoid signal saturation. The densitometry is shown to the right with results plotted on a linear scale. The results are representative of three independent experiments. Error bars are SEM. (e) Coprecipitation assays with GST-Sufu and lysates from cells containing wild-type or mutant Gli proteins. Note that the Gli1ΔD_N mutation leaves Sufu binding intact. (f) Immunofluorescence of Gli1 shows similar subcellular localization of wild-type and mutant Gli1 proteins in the absence (left) or presence (right) of the Crm1-inhibitor leptomycin B. (g, left) Luciferase transcription assays of wild-type and double-mutant Gli1 protein with increasing amounts of transfected moles of plasmid. Error bars are SEM. Densitometry (middle) of Western blots (right) showing the amount of steady-state protein accumulation corresponding to the increase in luciferase activity. (h, left) Luciferase transcription assays of wild-type, single, and double-mutant Gli1 proteins. Error bars are SEM. Western blot (right) of levels of Gli1 protein in luciferase assay and quantitation (middle) of protein levels normalized for loading and transfer efficiency determined by immunoblot for nuclear pore complex (NPC).

We next determined whether degron D_N functioned independently of degron D_C. We tested whether βTrCP binding depends on D_N. Consistent with the notion of distinct signals, coimmunoprecipitation studies showed that βTrCP bound equally well to both wild-type Gli1 and the Gli1ΔD_N mutant (Fig. 3c). Moreover, we tested whether degron D_N could confer instability to a heterologous protein. Green Fluorescent Protein (GFP) is a stable protein with a long half-life. Addition of amino acids 1-208, a region that encompasses degron D_N sequences, destabilized GFP in a proteasome-dependent fashion, giving it a half-life of 180 min (Fig. 3d). Together, these data suggest the two destruction signals function independently.

Degron D_N is immediately adjacent to the binding site for Sufu (Fig. 3a), a powerful negative regulator of the Shh pathway, suggesting that the degron might work in conjunction with Sufu. Consequently, we tested whether D_N mutations affected the known Sufu functions of transcriptional corepression and Gli sequestration in the cytosol (Ding et al. 1999; Kogerman et al. 1999; Cheng and Bishop 2002). Gli1ΔD_N bound to Sufu as well as wild-type Gli1 in GST pull-down (Fig. 3e) assays. Also, Gli1 mutants had a similar subcellular distribution to wild-type Gli1 and accumulated in the nucleus with equal efficiency in the presence of leptomycin (Fig. 3f). These data argue that the N-terminal degron regulates Gli1 stability via a unique Sufu- and degron D_C-independent pathway.

Gli1 is known to activate transcription of Shh target genes via a transactivation domain in its C terminus (Yoon et al. 1998). To determine the functional significance of stabilizing Gli, we assessed the transcriptional activity of the mutants on Gli-responsive promoters (Taipale et al. 2000). Gli1ΔD_C and Gli1ΔD_N displayed modest increases in transcription when the same molar amount of plasmid was transfected into cells, with the double mutant displaying threefold higher target gene induction compared with wild-type Gli1 (Fig. 3g,h). This increase could be due to increased transactivation ability or increased protein levels. Analysis of protein levels relative to transcriptional output demonstrated a clear linear relationship between the amount of Gli1 protein for each of the mutants and reporter gene output (Fig. 3g,h). These data argue that the greater transcriptional activity of the mutants is due to increased protein stability rather than transactivation ability.

Tumor induction in Gli transgenic animals correlates with Gli protein accumulation. If the degrons identified in our studies are responsible for restricting Gli1 protein accumulation in vivo, then expressing Gli1 without these signals should shorten the latency to tumor induction. We assayed skin phenotypes of several lines of transgenic animals expressing different mutants of Gli1 in the basal layer of stratified epithelia. As expected, transgenic animals expressing wild-type Gli1 were born normally with no detectable transgenic Gli protein (Fig. 4b,l), and developed the predicted tumor phenotype at 6-8 wk after birth (Oro and Higgins 2003; data not shown). In contrast, animals expressing double-mutant Gli1 (GliΔD_NΔD_C) exhibited Gli protein accumulation at the time of birth in tumor and nontumor epithelium (Fig. 4c; Supplementary Fig. 4). The GliΔD_NΔD_C-expressing animals died at birth with shallow skin ulcers clinically similar to BCCs throughout the body. The tumors demonstrated characteristic features of BCCs (Fig. 4c,m; Supplementary Fig. 4; Oro et al. 1997; Callahan et al. 2004), including the up-regulation of ptch1 (Fig. 4r). Moreover, the tumors were rapidly dividing as evidenced by the significant increase in Ki67 staining and displayed the BCC marker keratin 17 (Supplementary Fig. 4). This demonstrates that altered protein accumulation can directly accelerate tumor induction.

Open in a separate window

Figure 4.

Removal of two destruction signals rapidly accelerates tumor induction. (a) Diagram of Gli1 wild-type (b,g,l,q), Gli1ΔD_CΔD_N (c,h,m,r), Gli1ΔD_C (d,i,n,s), Gli1ΔD_N (e,j,o,t), or nontransgenic (f,k,p,u) mice analyzed in this study. The number of independent founders is shown in parentheses. (b-f) Representative H&E sections from each founder line. Note BCC-like lesions from interfollicular epithelium in c and d, and BCC-like tumors from hair follicle in e. Bar, 50 μm. (g-k) In situ hybridization of transgene expression using transgene-specific gli1 probe. Bar, 25 μm. (l-p) Immunohistochemistry with anti-HA antibody for Gli1 protein. Note the absence of Gli1 protein in wild-type Gli1 transgenics and nuclear and cytoplasmic distribution in mutant Gli animals. Bar, 10 μm. (q-u) In situ hybridization with ptch1 probe showing Shh target gene induction in tumors. Bar, 10 μm. (v) Table of representative features of each group of Gli1 transgenic mice.

In cultured cells, both degrons were highly active in restricting Gli1 levels. However, depending on the specificity and/or capacity of the operative degradation pathway, one of the degrons may play a more active role in a given in vivo context. We determined the relative contribution of each degron to Gli1 destruction by comparing the phenotype of single Gli mutant transgenic mice to those expressing GliΔD_NΔD_C. In the skin, Gli1 mutants lacking degron D_C displayed a much stronger phenotype than those lacking degron D_N (Fig. 4v). While both GliΔD_C and GliΔD_N transgenic mice were viable and lacked the ulcerating lesions seen in the double mutant, GliΔD_C mutants demonstrated BCC-like lesions at birth more comparable to those expressing GliΔD_NΔD_C (depth of invasion, 111 μM vs. 140 μM, respectively) (Fig. 4d). GliΔD_N transgenic animals had small BCC-like proliferations that developed slightly after birth and appeared to come directly off the hair follicle (Fig. 4e). Also, many GliΔD_N mutant lesions were benign hair follicle tumors, indicative of lower Shh target gene induction (Callahan and Oro 2001; Grachtchouk et al. 2003). In each of the Gli mutants, the distribution of Gli protein was both nuclear and cytoplasmic, providing further evidence that the degron sequences do not play a role in nucleocytoplasmic shuttling of Gli1 (Fig. 4l-p). The phenotypic differences within each group could not be attributed to transgene expression differences, as only steady-state protein levels by IHC, not transgene copy number or RNA expression level, correlated with the phenotype (Fig. 4; Supplementary Fig. 5). These data demonstrate the combinatorial action of both D_C and D_N degrons in preventing ectopic Shh target gene induction and provide in vivo support for the role of Gli destruction in controlling tumor formation.

Here we have shown that Gli1 protein contains two destruction signals that regulate protein stability and tumor formation (Supplementary Fig. 1). As with other key regulatory proteins such as myc, p53, IκB, and β-catenin, there appears to be a finely balanced control of Gli1 protein levels to allow for proper target gene induction while preventing epithelial tumor formation. Our data suggest that the BCC tumors observed in the K5Gli2 transgenic mice likely arise as a result of secondary changes that lead to Gli2 stabilization rather than as a result of gradual saturation of the destruction machinery. Arguing against saturation is the lack of increased protein in adjacent normal tissue or in the explanted cells from older animals. Furthermore, with the addition of proteasome inhibitors, we see rapid accumulation of Gli2 protein. This suggests that halting the destruction of Gli proteins is an early step in the tumor process and that cellular changes that allow Gli1 protein accumulation may contribute to human carcinogenesis (Kinzler et al. 1988). Similarly, targeted therapies that delay the onset of Gli accumulation may have potent antitumor properties.

Our study illustrates how two destruction signals cooperate to prevent Gli protein accumulation, target gene induction, and subsequent tumor formation. While a role for βTrCP has been implicated in Ci processing, the present study is the first to demonstrate that it acts by directly binding Gli to facilitate ubiquitinylation and destruction. Interestingly, while Ci and Gli1 are both directed by PKA and βTrCP to interact with the proteasome, the end result differs in that Gli1 is degraded but not cleaved. This could be due to either the particular amino acid sequence of the degron or to surrounding amino acids that influence βTrCP/UPS function. The identified Gli degron differs significantly from that of β-catenin, Emi1, and IkB in that it lacks a second serine shown to be important for sequential phosphorylation and contains a phosphomimetic glutamic acid residue (Amit et al. 2002; Moshe et al. 2004). Future studies will focus on whether these sequence differences are sufficient to account for the different final disposition of the protein. This study further identifies a novel degron, D_N, that shares little identity with other known degradation signals. The conserved sequences in this degron are found in both Gli2 and Gli3, and removal of the region containing them has been associated with activation of Gli2 (Sasaki et al. 1999; Mill et al. 2003). Our data suggest that a portion of this activation may be due to Gli2 protein stabilization via degron D_N rather than simply loss of transcriptional repressor activity.

Material and methods

Acknowledgments

Notes

Supplemental material is available at http://www.genesdev.org.

Article published online ahead of print. Article and publication date are at http://www.genesdev.org/cgi/doi/10.1101/gad.1380906.

References