Characterization of the human Suppressor of fused, a negative regulator of the zinc-finger transcription factor Gli

The Hedgehogs (HH) constitute a family of secreted proteins which play important roles in tissue patterning during early embryogenesis in vertebrates and invertebrates (Hammerschmidt et al., 1997; Ingham, 1995). Drosophila Hedgehog (Hh) is required for proper segmentation of the larvae, and for growth and organization of the wing and other appendages in the adult fly. The mammalian HH proteins, which include Sonic hedgehog (Shh), Indian hedgehog (Ihh) and Desert hedgehog (Dhh), are expressed in a tissue-specific manner. They control the specification of ventral cell types in the central nervous system, left-right asymmetry, growth and patterning of the somites and limbs, cartilage differentiation, organogenesis and spermatogenesis. Mutations in genes that mediate the HH signal have been linked to human cancer and developmental disorders (reviewed in Johnson and Scott, 1998), thus establishing an important role for this pathway in normal cell growth control.

The mechanism of Hh signal transduction is not fully understood. However, genetic studies in Drosophila have identified a diverse array of transmembrane and intracellular proteins which serve as specific components in the Hh signaling pathway (reviewed in Ingham, 1998; Johnson and Scott, 1998; Tabin and McMahon, 1997). The pathway culminates in the activation of Cubitus interruptus (Ci) (Alexandre et al., 1996; Dominguez et al., 1996), a zinc finger transcription factor homologous to vertebrate Gli proteins (Orenic et al., 1990). Conservation of Hh signal transduction mechanisms is suggested by the ability of ectopically expressed Xenopus or human Gli (hGli) to mimic Shh in the induction of floor-plate specific markers and ventral neuronal cell types, both in frog (Lee et al., 1997) and mouse (Hynes et al., 1997). Likewise, the cyclic AMP-dependent protein kinase (PKA) exerts a common negative regulatory effect on HH signaling in both flies (Jiang and Struhl, 1995; Li et al., 1995) and rodents (Epstein et al., 1996; Fan et al., 1995; Hammerschmidt et al., 1996; Hynes et al., 1995). Moreover, vertebrate homologues have been identified for Drosophila Patched (Goodrich et al., 1996), a multi-pass transmembrane protein which, by genetic analysis, functions downstream of Hh to inhibit signaling (Alcedo et al., 1996; Hooper and Scott, 1989), and Drosophila Smoothened (Smo; Stone et al., 1996), a seven-pass transmembrane protein absolutely required for transduction of the Hh signal (Alcedo et al., 1996; Hooper, 1994). A combination of biochemical data obtained in vertebrate systems and genetic analyses in Drosophila predict Patched to be the ligand-binding component and Smo the signaling component, in a multi-subunit receptor complex for HH proteins (Chen and Struhl, 1998; Marigo et al., 1996; Murone et al., 1999; Stone et al., 1996). Taken together, this evidence for evolutionary conservation suggests that other identified members of the Drosophila signaling pathway may likely have vertebrate counterparts.

Drosophila Suppressor of fused (dSu(fu)) is a novel cytoplasmic PEST-containing protein (Pham et al., 1995) which, when mutated in a wild-type background, confers a mild phenotype suggestive of constitutive Hh signaling (Ohlmeyer and Kalderon, 1998). Moreover, the same mutation can fully suppress both embryonic and adult phenotypes resulting from mutations in Fused (Fu) (Preat, 1992), a serine-threonine kinase required for Hh signaling (Mariol et al., 1987). DSu(fu) interacts physically with Fu and Ci (Monnier et al., 1998), and the latter interaction has been hypothesized to maintain Ci in an inactive state by sequestering it in the cytoplasm and/or by preventing its processing to an active form (Ohlmeyer and Kalderon, 1998). In the absence of Hh signaling, full-length Ci is proteolytically cleaved to produce an amino-terminal 75-kDa transcriptional repressor form (Aza-Blanc et al., 1997), presumably through targeting of PKA-phosphorylated Ci (Chen et al., 1998) to the ubiquitin-proteasome pathway by the F-box containing protein, Slimb (Jiang and Struhl, 1998). Reception of the Hh signal is predicted to activate Fu, inactivate dSu(fu) and trigger downstream events culminating in the conversion of Ci into a transactivator of Hh target genes.

To gain further biochemical and functional insight into the role of Su(fu) in HH signaling, we have cloned two isoforms of a human homologue of this protein and examined the expression pattern of Su(fu) during development and in the adult. Additionally, we studied physical interactions between hSu(fu) and other signaling components in the HH pathway, including members of the vertebrate Gli protein family and a vertebrate Slimb homologue, and analyzed the functional implications of these interactions.

Human fetal lung and human fetal testis libraries prepared in a CMV-based mammalian expression vector (pRK) were probed with an oligonucleotide derived from a mouse EST (accession no. AA061391; 5′GAGCACTGGCACTACATCAGCTTTGGCCTGAG-TGATCTCT3′). Several positive clones were selected and analyzed by restriction digest; two clones which exhibited different restriction patterns were sequenced on both strands by standard protocols.

Two different hSu(fu) cDNAs were identified, hSu(fu)433 and hSu(fu)484, which differed in sequence at the 3′ end of the open reading frame (ORF). By PCR, the hSu(fu)433 cDNA was epitope-tagged with the flag peptide at its carboxy terminus, to produce pRK-hSu(fu)433. pRK-hSu(fu)484 was tagged in a similar manner, and the 5′ end (to the XhoI site at base 946 of hSu(fu)433) was replaced with the 5′ end of hSu(fu)433, since the hSu(fu)484 cDNA was missing 458-bp of 5′ sequence. The human Gli cDNA (provided by Dr Ken Kinzler) was cloned into the same expression vector, and a 9E10 c-myc epitope was introduced at the amino terminus (immediately after the first ATG), to produce pRK-hGli. Human Gli3 (provided by Dr Mike Ruppert) was also cloned into pRK, to produce pRK-hGli3. The coding region of mouse Gli2 was obtained by PCR with Takara LA polymerase (Takara Shuzo Co., Ltd.) using Marathon Ready mouse E11 cDNA (Clontech) as template, and was cloned into pRK, yielding pRK-mGli2. A partial mouse Slimb cDNA was obtained from Genome Systems (IMAGE clone #1068742) and extended by 5′ RACE. Several different 5′ RACE products were recovered, suggesting that the gene is subject to alternative splicing at its 5′ end (see also Theodosiou et al., 1998; Margottin et al., 1998). The sequence most closely matching the amino terminus of human Slimb (Theodosiou et al., 1998) was isolated and an HA tag was introduced at the amino terminus to produce pRK-Slimb. Mouse Slimb differed from human Slimb at 9 out of 572 amino acids. pRK-SlimbΔF was generated by deleting amino acids 1-207, which encompassed the F-box motif, and adding an HA tag at the amino terminus. The glutathione-S-transferase (GST)-hSu(fu) expression construct (pGEX-hSu(fu)) was made by fusing the hSu(fu)433 cDNA in-frame to the carboxy terminus of GST in a pGEX vector (Pharmacia).

Human metaphase chromosome spreads from cultured blood lymphocytes were prepared by standard procedures, and subject to fluorescence in situ hybridization (FISH) as described (Heng et al., 1992; Heng and Tsui, 1993), using the entire hSu(fu)433 cDNA as probe. FISH signals and DAPI banding patterns of each chromosomal spread were recorded separately, then superimposed to assign the hSu(fu) mapping position.

Human multiple tissue northern blots (Clontech) were hybridized with a ³²P-labelled SmaI-XhoI 853-bp fragment of hSu(fu)433 cDNA according to the manufacturer’s protocol, and washed to a stringency of 0.2⨯SSC, 0.1% SDS at 65°C, prior to exposure to X-ray film. Semi-quantitative PCR was performed using 10 pmol each of primers: P1, 5′-CCAATCAACCCTCAGCGGCAGAATG-3′; P2, 5′-CGAGGC-CAGCAGCTCGTTC-3′; and P3, 5′-GTAGGTGAGAAAGAG-

GGCTGTC-3′, in a standard 25 μl PCR reaction containing 100 μM dNTPs, 0.25 μl [³³P]dATP (10 mCi/ml), and 50 ng plasmid DNA from human tissue cDNA expression libraries as template. DNA was amplified by Taq DNA Polymerase for 30 cycles, then 10 μl was run on a 4% TBE acrylamide gel, which was dried and exposed to X-ray film.

Whole-mount in situ hybridization to embryonic day 8.5 (E8.5) mouse embryos was performed as described (Shimamura and Rubenstein, 1997). The probe was a digoxigenin-labeled RNA, synthesized with T7 RNA polymerase and a mouse Su(fu) cDNA PCR template, corresponding to nucleotides 116-390 (nucleotide 1=A in the initiator ATG) of the human sequence. For in situ hybridization to tissue sections, rat E11.5 and E15.5 whole embryos, and postnatal day 1 (P1) rat brains were immersion-fixed overnight at 4°C in 4% paraformaldehyde, cryoprotected overnight in 15% sucrose, embedded in OTC (VWR Scientific), and frozen on liquid nitrogen. Adult rat brains were fresh frozen with powdered dry ice. Adult rat spinal cord and mouse testis were embedded in OTC and frozen on liquid nitrogen. Sections were cut at 16 μm, and processed for in situ hybridization as described previously (Phillips et al., 1990). [³³P]UTP-labeled RNA probes were generated as described (Melton et al., 1984). Sense and antisense probes were synthesized with T7 RNA polymerase from a hSu(fu)433 cDNA PCR fragment encompassing nucleotides 97-424 of the human sequence.

C3H10T1/2 or COS-7 cells were maintained in DMEM with 10% fetal bovine serum. Subconfluent cultures in ProNectin F (Stratagene)-coated glass chamber slides were transiently cotransfected with pRK-hGli, pRK-hSu(fu)484 or pRK-hSu(fu)433 using lipofectamine (Gibco BRL). 24 hours later, cells were fixed in 4% paraformaldehyde for 10 minutes, with or without prior detergent extraction (0.1% NP-40 in stabilization buffer: 4% PEG 8000, 1 mM EGTA, 100 mM Pipes, pH 6.9) for 10 minutes at 37°C. For tubulin staining, proteins were cross-linked with 1 mM DSP (Pierce) in stabilization buffer for 10 minutes at 37°C prior to fixation. Cells were permeabilized in 0.1% Triton-X 100 for 5 minutes, blocked in block buffer (5% goat serum in PBS) for 30 minutes, and reacted with primary antibodies consisting of rabbit anti-hSu(fu) (3 μg/ml; see below), anti-c-myc monoclonal (Genentech; 3 μg/ml in block buffer for 1 hour), or monoclonal anti-β-tubulin (Sigma, clone TUB 2.1; 1:250 dilution). Cells were washed and double-labeled with cy3-anti-rabbit IgG (1:350) and cy2-anti-mouse IgG (1:100; Jackson ImmunoResearch) for 1 hour in block buffer. Slides were washed and coverslips attached with Fluoromount-G (Southern Biotechnology Assoc., Inc.). The hSu(fu) polyclonal antibody was produced by immunization of rabbits with purified GST-hSu(fu) fusion protein. Resultant antibodies were purified by affinity chromatography on a Protein A column.

293 cells were grown in DMEM containing 10% fetal bovine serum, to 70% confluence in 10-cm tissue culture dishes. Cells were transfected with lipofectamine according to the manufacturer’s protocol (Gibco BRL) using a total of 8 μg of DNA/dish; for cotransfections, 4 μg of each were used. 24 hours later, cells were washed in PBS (4°C) and lysed directly in 1 ml ice-cold lysis buffer (containing 20 mM Hepes, pH 8.0, 150 mM NaCl, 1% NP-40, 5 μg/ml each leupeptin and aprotinin, 1 mM PMSF and 250 μM orthovanadate). Lysate was rotated at 4°C for 20 minutes, centrifuged at 14000 rpm for 20 minutes, and the supernatant subjected to immunoprecipitation with either 2 μl anti-flag M2 monoclonal antibody (Kodak IBI) or 2 μl anti-myc monoclonal antibody (9E10; Genentech) overnight (4°C). In some cases, ethidium bromide was added to the lysates prior to immunoprecipitation to preclude the possibility of DNA-dependent protein associations (Lai et al., 1992). Protein A sepharose (Pharmacia) was added (25 μl) for 1 hour at 4°C, the beads were washed 3⨯ with lysis buffer and 2⨯ with 1 M NaCl, and samples were heated in SDS-loading buffer to 70°C for 10 minutes. Samples were electrophoresed on 4-12% NuPAGE denaturing SDS gels (Novex), and proteins detected by blotting to nitrocellulose and probing with antibodies to flag or myc epitopes, using the ECL detection system (Amersham).

pGEX-hSu(fu) was transformed into DH12S bacterial cells (Gibco BRL), and a 500-ml overnight culture was processed for purification of GST-hSu(fu) fusion protein according to the manufacturer’s protocol (Pharmacia). Fusion protein was eluted from the beads with excess reduced glutathione, and eluted protein was quantified by OD₂₈₀ measurement and visualization on denaturing SDS-polyacrylamide gel (data not shown). Glutathione-sepharose beads were loaded with 4 μg fusion protein or GST (Sigma) for 2 hours at 4°C, then washed 3⨯ with binding buffer. Beads (25 μl of a 50:50 slurry) were incubated with 2-8 μl of ³⁵S-labeled in vitro-translated hGli, mGli2, hGli3, mSlimb or hSu(fu)433 in 50 μl binding buffer for 2 hours at 4°C. The beads were washed 3⨯ with lysis buffer, and processed for SDS-PAGE. Gels were fixed, amplified in EN³HANCE (Dupont NEN), dried and exposed to Kodak X-AR film. Binding buffer was 50 mM Tris-HCl, pH 8.0, 150 mM NaCl and protease inhibitors (as above). pRK-hGli, pRK-mGli2, pRK-hGli3, pRK-mSlimb, pRK-hSu(fu)433 and SP6-Luciferase control plasmid were transcribed and translated in vitro using the TNT-coupled reticulocyte lysate system (Promega), with 20 μCi [³⁵S]methionine (Amersham) and SP6 RNA polymerase in a 50 μl reaction volume. 1 μl of each reaction was subjected to denaturing SDS-PAGE for approximate protein quantitation. Equivalent amounts of each protein were used in binding assays.

The reporter assay was performed in C3H10T1/2 cells as described (Murone et al., 1999), using a Dual-Luciferase Reporter Assay System (Promega, Inc). Differences in transfection efficiency were corrected by normalizing the activity of the firefly Luciferase reporter to the activity of a cotransfected Renilla Luciferase internal control.

A BLAST search of the GenBank database revealed a mouse EST whose conceptual translation matched 55/91 amino acids of the dSu(fu) protein. To obtain the human Su(fu) cDNA, we screened human fetal lung and testis libraries with a [γ-³²P]dCTP-labeled oligonucleotide probe based on the mouse EST sequence. The longest clone contained 1839 base pairs (bp), comprising 146 nucleotides of upstream sequence, a 1299-bp ORF, and a 394-nucleotide 3′UTR, in addition to an extensive poly(A) tail. Conceptual translation predicted an approximate 48-kDa protein containing 433 amino acids (referred to as hSu(fu)433). A second clone, 1720 bp in length, was identical to hSu(fu)433 from bases 313-1296 of the ORF (A in intiator ATG = base 1), then diverged completely at the 3′ end. This clone, which was missing the 5′ UTR and starting methionine, comprised a partial 1139-bp ORF, in addition to a 581-bp 3′-UTR and a poly(A) tail. Assuming identity to hSu(fu)433 at the 5′ end, the second cDNA predicted a protein identical to hSu(fu)433 with an additional 52 amino acids at the carboxy terminus (referred to as hSu(fu)484). Alignment of hSu(fu) with dSu(fu) revealed a 37% identity at the amino acid level (Fig. 1), which increased to 63% allowing conservative amino acid substitutions. During preparation of this manuscript, a mouse Su(fu) cDNA was deposited in GenBank (accession no. AJ131692), and is included in the sequence alignment. The longer splice variant of human Su(fu) exhibits 97% sequence identity to the mouse Su(fu). A search of hSu(fu) against the Prosite database revealed 15 potential phosphorylation sites (16 in hSu(fu)484), several of which were conserved between Drosophila and vertebrate Su(fu) (indicated in Fig. 1). Three candidate PKA phosphorylation sites were identified in hSu(fu) and none in dSu(fu). However, by including in our search strategy several less active potential PKA phophorylation site motifs, we identified two additional sites in hSu(fu) and five such sites in dSu(fu) (Fig. 1). The PEST algorithm (Rechsteiner and Rogers, 1996) identified a high scoring PEST sequence (score=26.29), which spanned amino acids 280-289, and was conserved between human and mouse. The hSu(fu) gene was mapped to chromosome 10, region q24-q25 by FISH analysis (Fig. 2).

High stringency hybridization to human poly(A)+ RNA revealed hSu(fu) mRNA to be expressed widely in both embryonic and adult tissues (Fig. 3A). A major transcript of approx. 5.5 kb was detected in all tissues examined, predicting a relatively long 5′ UTR for the hSu(fu) message that was absent in our isolated clones. In several tissues, a 2.5 kb transcript could also be seen, and in testis two smaller transcripts of approx. 2.0 and 1.0 kb were also detectable. Semi-quantitative PCR was performed on cDNA derived from a number of human tissues, to determine the relative expression levels of the two alternatively spliced hSu(fu) mRNA transcripts. Both transcripts could be detected in all tissues examined; however, hSu(fu)484 was always the more abundant isoform (Fig. 3B). Adult testis contained the highest relative level of the shorter isoform, in which the ratio of hSu(fu)433:hSu(fu)484 was approximately 1:3 (Fig. 3B).

Whole-mount in situ hybridization revealed Su(fu) mRNA to be widely expressed in E8.5 mouse (Fig. 4A,B), the earliest developmental time point examined. Labeling was uniformly intense throughout the developing neural plate. Only the anlage of the heart was not specifically labeled at this stage (Fig. 4B). In E11.5 rat, Su(fu) message remained widespread throughout the central nervous system, spinal cord and somites (Fig. 4C). In E11.5 and E15.5 rat spinal cord, a prominent signal was seen within the developing neuroepithelium of the ventricular zone (Fig. 4E,F), a region of active cellular proliferation. Tissues throughout the E15.5 embryo, including the brain, spinal cord, gut, lung and testis were labeled for Su(fu); the liver displayed only a very low signal (Fig. 4D). In the P1 rat brain, Su(fu) mRNA was widely expressed, with prominent signals overlying the neuroepithelium, subventricular zone and hippocampal neuronal cell fields (Fig. 4G). Message was profoundly downregulated in adult brain yet still weakly detectable throughout; relatively high expression was observed in the hippocampus, cerebellar granule and Purkinje cell layers, and olfactory bulb (Fig. 4H-J).

In a cross section of adult mouse testis, Su(fu) mRNA was intensely expressed in a subset of seminiferous tubules, suggesting that its transcription may be regulated according to the stages of germinal cell differentiation (Fig. 5A). Su(fu) message was observed as a ring of silver grains over the region of developing spermatocytes (Fig. 5C); in many sites within the tubule, highest expression was concentrated in the center, where the latest stages of germinal cell differentiation occur (Fig. 5D). Hybridization of a sense strand control probe to an adjacent tissue section showed no signal above background (Fig. 5B).

Since previous studies demonstrated binding of dSu(fu) to Ci (Monnier et al., 1998), the Drosophila Gli homologue, we asked whether a similar interaction might exist between their human protein counterparts. We used immunocytochemistry to visualize the subcellular localization of transiently overexpressed hSu(Fu) and hGli in transfected C3H10T1/2 cells. When individually expressed, hSu(fu) and hGli were extensively distributed throughout the cytoplasm; hGli was sometimes detected in the nucleus (data not shown). Double labeling of cotransfected C3H10T1/2 cells revealed an extensive overlap in the two staining patterns (Fig. 6, row A). Moreover, hGli frequently appeared to be depleted from the cytoplasm and instead concentrated in widely dispersed punctate, densely stained regions, which also labeled strongly for hSu(fu) (Fig. 6, row A). These densely stained regions, which were not seen in cells overexpressing hSu(fu) alone, and always stained for both proteins, might represent cytoplasmic sequestration of hGli by hSu(fu). Double labeling of cotransfected cells with an anti-β-tubulin antibody showed hSu(fu) to be partially colocalized with tubulin, predominantly in perinuclear regions (Fig. 6, row B), although Su(fu) was also seen in areas devoid of tubulin staining. Interestingly, treatment of cotransfected cells with nonionic detergent prior to fixation resulted in frequent nuclear localization of hSu(fu) antigen (data not shown). In COS-7 cells, hGli was predominantly nuclear when expressed either alone or with hSu(fu)433, yet relocalized to the cytoplasm when coexpressed with hSu(fu)484 (Fig. 6C).

To determine whether hSu(fu) and hGli formed a physical complex, we looked for coimmunoprecipitation of the two proteins from transiently transfected 293 cells. We found that hSu(fu) and hGli could be readily coimmunoprecipitated from cells expressing both proteins (Fig. 7A). This interaction appeared not to involve indirect protein-DNA associations, since inclusion of ethidium bromide in the immunoprecipitation reaction (Lai et al., 1992) did not inhibit complex formation (Fig. 7A). The hSu(fu)-hGli interaction was confirmed using an in vitro binding assay. For this purpose, bacterially produced GST-hSu(fu) protein was loaded on glutathione-sepharose beads and examined for its ability to retain in vitro-translated ³⁵S-labeled hGli. We found that hGli was specifically retained on GST-hSu(fu) glutathione-sepharose beads, but not on beads loaded with GST alone (Fig. 7B).

In further experiments, we examined the ability of the Gli homologues Gli2 and Gli3 to interact with hSu(fu). Both ³⁵S-labeled mGli2 and hGli3 but not Luciferase, a negative control, were specifically retained by GST-hSu(fu)-conjugated beads (Fig. 7B). A version of hGli3 in which the 9E10 c-myc epitope was fused to the extreme carboxy terminus of the protein did not bind to hSu(fu) in this assay (data not shown), indicating that the carboxy-terminal region of Gli3 is important for the interaction. Interestingly, ³⁵S-labeled Su(fu) also bound to GST-hSu(fu) (Fig. 7B), and hSu(fu) was found to bind to itself in a 2-hybrid assay (data not shown), indicating that hSu(fu) may function as a dimer.

Our binding data suggested that the activity of vertebrate Glis may be regulated by interaction with Su(fu). Thus, we examined directly whether hSu(fu) could modulate the activity of hGli in a functional Gli reporter assay. To this end, nine copies of a Gli binding site (Sasaki et al., 1997) were linked to a Herpes simplex virus Thymidine kinase minimal promoter, which directs the transcription of a reporter firefly Luciferase gene. Expression of Luciferase from this construct was shown to be specifically regulated by Gli and by components of the Shh receptor (Murone et al., 1999). As previously demonstrated (Murone et al., 1999), cotransfection of C3H10T1/2 cells with the Luciferase reporter construct and either empty vector or an expression plasmid encoding an irrelevant protein (pRK-GFP) resulted in very low levels of Luciferase activity (Figs 7C, 8). In contrast, cotransfection of the reporter gene with an hGli expression plasmid resulted in an approximately 40-to 70-fold increase in the level of Luciferase activity (Figs 7C, 8). Consistent with the notion that dSu(fu) is a negative regulator of Ci, hGli-mediated reporter activation was significantly suppressed in the presence of coexpressed hSu(fu)433 or hSu(fu)484, but not an irrelevant protein (Fig. 7C). Thus, our findings suggest that the physical interaction of hSu(fu) with hGli leads to inhibition of its transactivator function.

Su(fu) might repress Gli activity in part by altering its processing or degradation, or by influencing its DNA-binding activity. To begin to examine these possibilities, we looked for interactions between hSu(fu) and a vertebrate homologue of Slimb/βTrCP, an F-box containing protein implicated in targeting of Ci and other proteins to the Ubiquitin-proteasomal degradation pathway (Jiang and Struhl, 1998; Margottin et al., 1998; Yaron et al., 1998). We found that in vitro-translated ³⁵S-labeled Slimb indeed specifically bound to Su(fu) in the GST pull-down assay (Fig. 7B). To investigate a potential functional role for this interaction, we tested the effect of Slimb on Gli transcriptional activity. In our functional Gli reporter assay, Slimb alone did not alter Gli-induced reporter expression; however, when cotransfected with hSu(fu), Slimb significantly potentiated the inhibitory effect of Su(fu) on Gli activity (Fig. 8). This potentiation was not seen with substitution of SlimbΔF, a truncated version of Slimb lacking the F-box domain (Fig. 8).

We have identified two isoforms of a human protein exhibiting 63% similarity to dSu(fu), and a developmental expression profile consistent with a role in vertebrate HH signaling. The identification of multiple hSu(fu) isoforms suggests that the gene is subject to alternative splicing, and that the divergent carboxy terminus of the protein may serve an important regulatory function. While minor differences were observed in the extent of Gli repression mediated by the two proteins in a Gli reporter assay, further studies are required to determine the functional role of alternative splicing in the regulation of HH signaling in vivo. In preliminary studies, we found that the ratio of the two mRNA transcripts differed between tissues, with hSu(fu)484 predominating.

Immediately prior to submission of this manuscript, a mouse Su(fu) cDNA was deposited in GenBank. Alignment of the longer human splice variant with mouse Su(fu), revealed a 97% sequence identy at the amino acid level. Several potential phosphorylation sites conserved between hSu(fu) and dSu(fu) were identified, two of which are candidate PKA phophorylation sites; the only other protein motif identified was a high-scoring PEST domain (Rechsteiner and Rogers, 1996) in the carboxy terminal half. By FISH analysis, we mapped the hSu(fu) gene to chromosome 10q24-25. Interestingly, two loci for tumor suppressor genes have been proposed within the interval 10q.23-qter, based on loss of heterozygosity (LOH) analysis in a number of tumors, including glioblastoma multiforme, prostate cancer, malignant melanoma and endometrial cancer (Albarosa et al., 1996; Gray et al., 1995; Peiffer-Schneider et al., 1998; Rasheed et al., 1995). Two other candidate tumor suppressor genes found mutated in a number of cancers have also recently been mapped to this region: MMAC1/PTEN at 10q23.3 (Li et al., 1997; Steck et al., 1997) and DMBT1 (deleted in malignant brain tumors) at 10q25.3-26.1 (Mollenhauer et al., 1998).

By northern analysis, hSu(fu) was found to be widely expressed in both fetal and adult tissues. The presence of hSu(fu) mRNA in many adult tissues suggests that they may continue to posess a functional HH signaling pathway, and that hSu(fu) might be required therein to suppress Gli activity. Conceivably, loss of hSu(fu) function in these tissues could provide an additional route to activation of HH signaling, and consequent oncogenesis. In light of this possibility, the mapping of hSu(fu) to a known tumor suppressor locus is particularly intriguing.

Examination of the developmental expression of rodent Su(fu) by in situ hybridization revealed that many HH-responsive tissues prominantly expressed Su(fu) mRNA (see Fig. 4D), including Shh-responsive embryonic neural folds and neural tube (Hammerschmidt et al., 1997; Ingham, 1995), presomitic mesoderm and somites (Fan et al., 1995), and embryonic foregut, esophagus and lung (Litingtung et al., 1998; Motoyama et al., 1998), Ihh-responsive cartilage (Vortkamp et al., 1996), and Dhh-responsive testis (Bitgood et al., 1996). Expression was maintained in a subset of cells within the adult brain, including hippocampal pyramidal and granule cells, cerebellar granule and Purkinje cells, and olfactory bulb granule cells, suggesting that regions which remain mitotically active or retain the capacity for such activity may require the continued expression of Su(fu). In adult rat cerebellum, expression of Su(fu) overlapped with that of Shh, Smo and Patched mRNA in Purkinje cells (Fig. 4J) (Traiffort et al., 1998). In humans, Shh signaling is known to play an important role in cerebellar development (Wechsler-Reya and Scott, 1999). This is evidenced by the involvement of Patched mutations in the etiology of Naevoid basal cell carcinoma (Gorlin) syndrome (Hahn et al., 1996; Johnson et al., 1996), which is associated with an increased frequency of cerebellar medulloblastoma. Also, mice heterozygous for a targeted disruption of the Patched gene develop medulloblastomas (Goodrich et al., 1997). The presence of Shh signaling components in adult cerebellum may indicate a continued role for this pathway in cerebellar maintenance.

In testis, expression of Su(fu) mRNA in developing germ cells overlapped with that of Patched2, a second vertebrate HH-binding protein with homology to Patched, and a vertebrate Fu homologue (Carpenter, 1998). Additionally, both Gli and Gli3 are expressed in developing spermatagonia (Persengiev et al., 1997); together, the data support the hypothesis that Sertoli cell-derived Dhh (Bitgood et al., 1996) may signal developing germ cells through a Smo-Patched2 receptor complex. Su(fu) was not observed in the interstitial Leydig cells, the site of Patched gene expression in adult testis (Bitgood et al., 1996; Carpenter, 1998). The presence of Su(fu) mRNA in tissues responsive to Shh, Ihh and Dhh suggests that the same signaling components and mechanisms may be used by all mammalian HH family members.

Consistent with a role for Su(fu) in vertebrate HH signaling, immunocytochemical localization of coexpressed hSu(fu) and hGli in cultured cells revealed the two proteins to be colocalized (Fig. 6A,C). While the staining pattern observed for either of the splice variants of hSu(fu) appeared similar when expressed alone, the shorter isoform was more prominently localized to punctate cytoplasmic densities upon coexpression with hGli (data not shown). Interestingly, selective extraction of cells with nonionic detergent prior to fixation revealed nuclear localization of hSu(fu) labeling (data not shown). This effect might be due to an unmasking of hidden antigenic sites, or an enhancement of the relative nuclear labeling intensity due to elimination of background cytoplasmic staining. Nuclear hSu(fu) localization was not altered by pretreatment of cells with recombinant soluble N-Shh, and was often, although not always, coincident with nuclear hGli labeling (see Fig. 6C). Although the hSu(fu) sequence does not contain a conventional nuclear localization signal, Reinhardt’s method for nuclear/cytoplasmic discrimination (Reinhardt and Hubbard, 1998) predicts hSu(fu) to be predominantly nuclear.

In Drosophila, Ci is found in complex with Su(fu), Fu (Monnier et al., 1998), and Costal2 (Robbins et al., 1997; Sisson et al., 1997), a kinesin-related microtubule-associated protein thought to repress Hh signal transduction by tethering the signaling machinery to the cytoskeleton. We thus examined whether hSu(fu) and hGli colocalized with microtubules in mammalian cells. Labeling of transfected cells with an anti-β-tubulin antibody revealed a partial overlap in the staining patterns of β-tubulin and either hSu(fu) or hGli (Fig. 6B and data not shown), allowing the possibility that, as in fly, the activity of HH signaling components might be regulated by association with cytoskeletal elements.

A physical association between hGli and hSu(fu) was demonstrated in two different assay systems. First, either splice variant of hSu(fu) could be co-immunoprecipitated with hGli from cotransfected 293 cells, using antibodies to epitope tags on hSu(fu) and hGli (Fig. 7A). Second, ³⁵S-labeled in vitro-translated hGli bound specifically to a GST-hSu(fu) fusion protein in an in vitro binding assay (Fig. 7B). Our results complement those of Monnier and collegues (1998) who demonstrated an analogous interaction between Drosophila Ci and dSu(fu). These authors found that Su(fu) could act as a link between Fu and Ci in the formation of a trimolecular complex (Monnier et al., 1998). They predicted that activation of Fu might trigger the dissociation of dSu(fu) and Ci, conceivably through phosphorylation of the PEST sequence in dSu(fu) and consequent dSu(fu) degradation. Consistent with this hypothesis, we have observed a decrease in the amount of hSu(fu) co-immunoprecipitated with hGli, upon exposure of cotransfected C3H10T1/2 cells to soluble N-Shh (M. Murone and F. J. de Sauvage, unpublished).

We further examined the ability of Gli2 and Gli3, two additional members of the Gli family of zinc finger transcription factors, to interact with hSu(fu). The three Gli proteins appear to subserve both specific and redundant functions in HH-mediated developmental processes. This is indicated by their differential expression patterns (Hui et al., 1994) and by the observed phenotypes of mice harboring targeted disruptions in Gli2 and Gli3 genes (Ding et al., 1998; Matise et al., 1998; Mo et al., 1997; Motoyama, 1998). Both mGli2 and hGli3 were found to bind specifically to GST-hSu(fu) protein in our in vitro binding assay (Fig. 7B). Our data thus support a role for hSu(fu) in regulating the activity of all members of the Gli protein family. Previous genetic studies have suggested that the interaction of dSu(fu) with Ci may inhibit Ci by preventing its maturation into a transcriptionally active form (Monnier et al., 1998; Ohlmeyer and Kalderon, 1998). This inhibition, is thought to be relieved by reception of the Hh signal. We directly examined whether hSu(fu) could influence the activity of hGli in a Gli transcriptional activation reporter assay (Murone et al., 1999). We found that hGli could activate reporter expression up to 70-fold, and this activation was dramatically suppressed by coexpression of hSu(fu) (Fig. 7C). The finding that Su(fu) did not completely abolish Gli-mediated activation might reflect differences in relative protein levels due to unequal transfection or expression efficiencies. In this regard, evidence derived from Drosophila genetics (Ohlmeyer and Kalderon, 1998) demonstrates the critical role of the stoichiometric ratio between Ci and dSu(fu) in determining cellular response.

In addition to its interaction with Gli family members, hSu(fu) associated with itself and with a vertebrate homologue of Slimb (Fig. 7B). The former interaction was independently identified in a yeast 2-hybrid screen, in which Su(fu) from a human testis library was isolated as an interacting partner with full-length hSu(fu). Homo- or heterodimers of hSu(fu) might function to bring together other effector proteins, thus allowing for differential regulation of Gli activity. The association of Su(fu) with Slimb prompted us to investigate a potential role for this interaction in Gli regulation. Genetic data from Drosophila implicate Slimb as a negative regulator of Hh signaling (Jiang and Struhl, 1998; Theodosiou et al., 1998), by enhancing Ci degradation. Slimb contains an F-box and several WD-40 repeat domains which function, respectively, as a binding site for components of the E2 ubiquitin-conjugating protein degradation complex (Bai et al., 1996; Margottin et al., 1998; Skowyra et al., 1997) and as protein-protein interaction regions (Neer et al., 1994). When tested in a functional Gli activity assay, Slimb alone had no effect on Gli-mediated transactivation, yet it potentated the inhibitory effect of Su(fu) on Gli activity (Fig. 8). The Slimb-enhanced suppression of Gli required the presence of the F-box domain, consistent with involvement of the multimeric SCF E3 ubiquitin ligase (Skowyra et al., 1997). In related experiments, we found that Slimb and hSu(fu) (either isoform), as well as Slimb and hGli, could be co-immunoprecipitated from 293 cells overexpressing the two proteins (data not shown). The data support a model in which Su(fu) may act as an adapter protein to link Gli to the Slimb-SCF-dependent degradation pathway (Skowyra et al., 1997). A similar function was demonstrated for the HIV-1 protein Vpu in the association between βTrCP, an alternatively spliced isoform of human Slimb (Theodosiou et al., 1998), and CD4 (Margottin et al., 1998). However, the precise role of Slimb in this context remains to be defined. Unlike Ci, Gli appears not to undergo proteasome-dependent cleavage into a truncated repressor form (Yoon et al., 1998; Dai et al., 1999). Furthermore, recent studies in Drosophila cl-8 cells suggest that the formation of Ubiquitin-Ci conjugates – presumably mediated by Slimb – is not under Hh regulation (Chen et al., 1999). Thus, we cannot exclude the possibility that the Slimb/Gli/Su(fu) interaction functions, in part, to maintain the steady state level of Gli and is not directly involved in HH signaling.

Our results suggest that Gli may be regulated by Su(fu) at multiple levels. Alterations in Gli protein stability by recruitment of the Ubiquitin degradation machinery, represents one potential regulatory step. However, preliminary studies suggest that, although both forms of Su(fu) inhibit Gli activity, they exert different effects on Gli stability/solubility and subcellular localization. Thus, by analogy to Drosophila systems (Ohlmeyer and Kalderon, 1998; Chen et al., 1999), the predominant role of Su(fu) may be to sequester Gli, and, through this association, prevent its post-translational modification into a transcriptionally active form.

In sum, the biochemical interactions which we have demonstrated between hSu(fu) and Gli family members, in conjunction with results from a Gli transactivation assay, complement genetic studies in Drosophila and support the idea that Su(fu) is a direct negative regulator of Gli. This regulation may occur at multiple levels, affecting the stability, processing and cellular localization of Gli. Our data further emphasize the importance of the relative intracellular concentrations of different signaling components in determining the cellular response to HH family members, and extend the conservation (with some notable differences) of Hh signaling components and mechanisms from Drosophila to human.

We have recently identified an ∼5 kb hSu(fu) cDNA which contains a 2715-bp extension of the hSu(fu)484 cDNA in the 3′ UTR. This cDNA correlates in size with the most abundant mRNA transcript on northern blots, and likely represents use of an alternative polyadenylation signal. Additionally, we have identified a third alternative splice variant of hSu(fu) (referred to as hSu(fu)481) encoding a protein identical to hSu(fu)433 up to residue 431, and including the following additional 50 amino acids at the carboxy terminus:

HVRWPFFFSLLPFIDFLAHPSSSPLAALDGTPSWGAGHE-CLMDSGPGACV.

While this paper was in press, three groups have independently described the cloning of mammalian Su(fu) and its ability to bind Gli family members (Ding et al. (1999) Curr. Biol. 9, 1119-1122; Kogerman et al. (1999) Nature Cell Biol. 1, 312-319; Pearse et al. (1999) Dev. Biol. 212, 323-336). In agreement with two of these reports, we find that hSu(fu)484 sequesters Gli in the cytoplasm of cotransfected COS-7, NIH-3T3, or C3H10T1/2 cells. However, neither splice variant hSu(fu)433 or hSu(fu)481 similarly alters Gli subcellular distribution, although both can interact with, and suppress the activity of Gli.

The GenBank accession numbers for hSu(fu) are AF144231 and AF159447.

We thank Kenneth Kinzler for the Gli cDNA, Mike Ruppert for the Gli3 cDNA, the Genentech DNA synthesis group for oligonucleotide synthesis and purification, and Wei-Hsein Ho for epitope-tagged hSu(fu). We are grateful to SeeDNA Biotech (Toronto, Ontario, Canada) for performing the FISH analysis, to Greg Bennett for help with antibody production, to Christa Gray and Alan Zhong for DNA sequencing, and to Wayne Anstine and Evelyn Berry for assistance with graphics and manuscript preparation.