Characterization of a human Rhomboid homolog, p100hRho/RHBDF1, which interacts with TGF-α family ligands

Authors

  • Takatoshi Nakagawa,

    1. Departments of Cell and Tissue Biology, and Anatomy, Programs in Cell Biology and Developmental Biology, University of California at San Francisco, San Francisco, California
    Current affiliation:
    1. Department of Glycotherapeutics, Osaka University Graduate School of Medicine, Yamadaoka 2-2, Suita-shi, Osaka 565-0082, Japan
    Search for more papers by this author
  • Annabel Guichard,

    1. Section of Cell and Developmental Biology, University of California, San Diego, 9500 Gilman Drive, La Jolla, California
    Search for more papers by this author
  • Carolina Perez Castro,

    1. Departments of Cell and Tissue Biology, and Anatomy, Programs in Cell Biology and Developmental Biology, University of California at San Francisco, San Francisco, California
    Search for more papers by this author
  • Yang Xiao,

    1. Departments of Cell and Tissue Biology, and Anatomy, Programs in Cell Biology and Developmental Biology, University of California at San Francisco, San Francisco, California
    Search for more papers by this author
  • Michael Rizen,

    1. Departments of Cell and Tissue Biology, and Anatomy, Programs in Cell Biology and Developmental Biology, University of California at San Francisco, San Francisco, California
    Search for more papers by this author
  • Hong-Zhong Zhang,

    1. Departments of Cell and Tissue Biology, and Anatomy, Programs in Cell Biology and Developmental Biology, University of California at San Francisco, San Francisco, California
    Search for more papers by this author
  • Diane Hu,

    1. Department of Orthopedic Surgery, University of California at San Francisco, San Francisco, California
    Search for more papers by this author
  • Anne Bang,

    1. Molecular Neurobiology Laboratory, The Salk Institute for Biological Studies, La Jolla, California
    Search for more papers by this author
  • Jill Helms,

    1. Molecular Neurobiology Laboratory, The Salk Institute for Biological Studies, La Jolla, California
    Current affiliation:
    1. Department of Surgery, Division of Plastic and Reconstructive Surgery, Stanford University Medical School, Stanford, California
    Search for more papers by this author
  • Ethan Bier,

    1. Section of Cell and Developmental Biology, University of California, San Diego, 9500 Gilman Drive, La Jolla, California
    Search for more papers by this author
  • Rik Derynck

    Corresponding author
    1. Departments of Cell and Tissue Biology, and Anatomy, Programs in Cell Biology and Developmental Biology, University of California at San Francisco, San Francisco, California
    • Department of Cell and Tissue Biology, University of California at San Francisco, San Francisco, CA 94143-0512
    Search for more papers by this author

Abstract

The activity of the TGF-α-like ligand Spitz in Drosophila depends on Rhomboid, a seven-transmembrane spanning protein that resides in the Golgi and acts as a serine protease to cleave Spitz, thereby releasing the soluble ligand. Several rhomboids in Drosophila have been implicated in the processing of TGF-α-like ligands, and consequent EGF receptor activation. The larger number of TGF-α-like ligands in vertebrates raises the possibility that they too might be subject to regulation by rhomboid-like proteins. We present the cDNA cloning and polypeptide sequence of an atypically long human rhomboid, which, based on the absence of critical residues for serine protease activity, is not predicted to act as a serine protease. We examined its tissue distribution, in comparison with TGF-α and the TGF-α-related protein HB-EGF, and the EGF/TGF-α receptor, in mouse embryo. This rhomboid, named p100hRho or RHBDF1, is a seven-transmembrane protein with a long N-terminal cytoplasmic extension that comprises half of the polypeptide sequence, and is found in the endoplasmic reticulum and Golgi, but not on the cell surface. It is expressed as two forms with different lengths, forms dimers and interacts with TGF-α ligands through a luminal interaction with the EGF core ectodomain. Finally, we evaluated the function of p100hRho/RHBDF1 in Drosophila, demonstrating that the short, but not the full-length form has functional activity. The characterization of this protein extends our understanding of the rhomboid family of regulatory proteins. Developmental Dynamics 233:1315–1331, 2005. © 2005 Wiley-Liss, Inc.

INTRODUCTION

Cell proliferation and differentiation are elaborated by growth factors that act through cell surface receptors to activate diverse signaling and gene expression programs. Among these, the EGF/TGF-α family, which acts through EGF receptor (EGFR) family tyrosine kinase receptors, regulates cell proliferation and differentiation of various cell types, including epithelial cells (Derynck,1992; Lee et al.,1995). In contrast to most growth and differentiation factors, EGF/TGF-α family members, with TGF-α as prototype, are transmembrane proteins that are subject to regulated proteolysis to release soluble ligands (Derynck,1992; Massagué and Pandiella,1993). In Drosophila, the three TGF-α-like ligands, Spitz (Rutledge et al.1992), Gurken (Neuman-Silberberg and Schupbach,1996), and Keren (Baonza et al.,2001; Reich and Shilo,2002), act through a single type of EGFR. In vertebrates, a larger number of ligands act through homo- or heterodimeric combinations of four different EGFR family members (Tzahar et al.,1996). In addition to their roles in normal development, increased signaling by TGF-α-like ligands and EGFR-like receptors drive malignant transformation of epithelial cells and progression of carcinomas (Khazaie et al.,1993).

Enhanced signaling is often achieved by increased expression of TGF-α ligands or EGFR or both. Their increased expression levels often correlate with the degree of malignancy and cancer progression (Sandgren et al.,1990; Jhappan et al.,1990; Wang et al.,1995; Amandudottir et al., 1996). Changes in ligand presentation also affect EGFR activation and act as important determinants of EGFR signaling. For example, ectodomain shedding of TGF-α and related ligands by membrane-associated proteases results in the release of diffusible ligands that can act at a distance, as soluble rather than transmembrane ligands. The role of ectodomain shedding in ligand presentation is illustrated by the lethal phenotype of mice deficient in TACE, a metalloprotease that mediates ectodomain shedding of TGF-α and other transmembrane proteins in response to various stimuli (Peschon et al.,1998). The similarities in phenotypes of these and the TGF-α-null mice strongly suggest a key role of ectodomain shedding in the presentation of TGF-α ligands, even though both the transmembrane and soluble TGF-α forms can activate the EGFR (Brachmann et al.,1989; Wong et al.,1989; Shi et al.,2000). While the importance of ligand presentation in EGFR family signaling is beginning to be appreciated, little is known about the mechanisms and effector proteins that regulate transport and presentation of TGF-α ligands, especially in vertebrates. Physical association of the tetraspanin CD9 with transmembrane TGF-α (Shi et al.,2000) or the TGF-α-like protein HB-EGF (Iwamoto et al.1994; Higashiyama et al.,1995) at the cell surface strongly enhances growth factor–induced activation of the EGFR, possibly through clustering of the transmembrane ligands into multiprotein signaling centers and conformational changes in TGF-α ligands. Also, the PDZ proteins syntenin (Fernandez-Larrea et al.,1999) and GRASP55 (Kuo et al.,2000) have been shown to interact with the C-terminal sequence of transmembrane TGF-α in the Golgi and have been implicated in the transport of TGF-α through the Golgi to the cell surface. While these and other findings strongly suggest a role of the growth factor's cytoplasmic domain and interacting proteins in transport, the signaling mechanisms that regulate the transit of TGF-α or related proteins through the Golgi to the cell surface remain to be defined.

Insights into the effector proteins and mechanisms of presentation have been provided by genetic analyses of Drosophila during development. Cornichon, a putative three-membrane spanner, may be involved in polarized localization of the TGF-α-like protein Gurken to the cell surface, presumably by guiding its transport through the endoplasmic reticulum (Roth et al.,1995). The genes for the type II transmembrane protein Star (Kolodkin et al.,1994) and the seven-transmembrane spanning Rhomboid (Rho; Bier et al.,1990) were identified as genes required for the signaling activity of the TGF-α-like protein Spitz. Star is required for trafficking of transmembrane Spitz from the endoplasmic reticulum into the Golgi (Lee et al.,2001; Tsruya et al.,2002), where Rho resides, and Rho then acts as a transmembrane serine protease that cleaves the transmembrane segment of Spitz, thus allowing its release as a soluble protein (Lee et al.,2001; Urban et al.,2001). Genomic analyses have revealed the existence of six additional genes for Rhomboid-like proteins in Drosophila (Wasserman et al.,2000). In addition to Rho (also known as Rho1), Rho2, 3, and 4 also act as serine proteases capable of cleaving not only Spitz, but also the two other transmembrane, TGF-α-like ligands Gurken and Keren (Urban et al.,2002a). The protease activity of these rhomboids is reflected in the conservation of critical residues in their transmembrane regions, including three amino acids that are thought to form the catalytic triad that effects the cleavage of the target transmembrane segment (Urban et al.,2001,2002a).

Rhomboid-like proteins with an evolutionary conservation of six, yet often containing seven putative transmembrane regions, have recently been identified as genes in nearly all genomes of archaea, bacteria, and eukaryotes (Koonin et al.,2003). Although their overall sequence similarity to Rho1 is low, several prokaryotic Rhomboid-like proteins have conserved the ability to cleave Spitz (Urban et al.,2002b), and one of them can functionally replace Rho1 in a wing vein development assay in Drosophila (Gallio et al.,2002). The existence of rhomboids in bacteria, which do not express TGF-α/EGFR signaling systems, implies their involvement in other processes, as shown for a Rhomboid-like protein in P. stuartii (Rather et al.,1999). Two rhomboids have been identified in yeast and one of these, Rbd1p, was shown to be involved in mitochondrial membrane remodeling, through its ability to cleave a mitochondrial transmembrane protein (McQuibban et al.,2003).

Little is as yet known about the role of Rhomboid-like proteins in vertebrate and mammalian cells. The role of Rho1 (Bang and Kintner,2000) in the presentation and intramembrane cleavage of TGF-α-like proteins in Drosophila suggests possible similar roles in the presentation of TGF-α-like proteins in vertebrates. However, Drosophila Rhomboid and the prokaryotic rhomboids that were shown to have serine protease activity toward Spitz are unable to cleave transmembrane TGF-α (Urban et al.,2002b; Urban and Freeman, 2003; Lohi et al.,2004). Currently, four vertebrate Rhomboid-like sequences have been reported in the literature, i.e., those for RHBDL (also called rrp1; Pascal and Brown,1998), RHBDL2 (Lee et al.,2001), Venthroid/RHBDL3 (Jaszai and Brand,2002; Lohi et al.,2004), and PARL (Pellegrini et al.,2001). RHBDL1 and 2 are unable to cleave TGF-α-like ligands, but RHBDL2, not RHBDL1 and 3, was found to cleave thrombomodulin (Lohi et al.,2004). RHBDL2 can act as a serine protease using Spitz as substrate, but RHBDL is unable to do so in spite of its conservation of critical amino acid residue for serine protease. PARL, the mammalian ortholog of the yeast Rbd1p, is able to rescue the phenotype rbd1-deficient yeast cells (McQuibban et al.,2003). Clearly, the characterization of the roles and substrates of the different vertebrate rhomboids will be a major undertaking.

We now present the cDNA cloning and polypeptide sequence of an atypically long human rhomboid, named p100hRho or RHBDF1, which, based on the absence of some critical residues for serine protease activity, is not predicted to act as a serine protease. We analyze its distribution in human tissues and compare its expression pattern with the EGFR and its ligands in mouse embryo. We also characterize its subcellular localization and transmembrane topology, and show that it is expressed as two forms with different lengths, forms dimers, and interacts with the TGF-α ligands through a luminal interaction with the EGF core ectodomain sequence. Finally, we evaluate the function of this rhomboid family member in Drosophila, demonstrating that the short form, but not the full-length form, has functional activity. The characterization of this protein extends our understanding of the rhomboid family of regulatory proteins.

MATERIALS AND METHODS

cDNA Cloning of p100hRho

To identify a human Rhomboid homolog, we searched databases using the Drosophila Rho1 (dRho1) sequence (Bier et al.,1990) and identified a short expressed sequence tag (GenBank accession no. D82416) from a human pancreatic islet cDNA library. An upstream primer 5′-CATAGCCATCATCTACCTGCTGAGTG- 3′ and a downstream primer 5′ CAAACCTGATGTAGGGCAAGAAGACG were used to amplify the expected 305-bp cDNA sequence from a human placental, λgt10-based cDNA library. This cDNA fragment was used as hybridization probe against the same cDNA library. This led to the identification of a recombinant phage with a 1.8-kbp cDNA insert that contained the 3′ half, including the stop codon, of the p100hRho coding sequence.

Continuous BLAST searches of databases identified a p100hRho cDNA of about 3.0 kbp, i.e., FLJ22357 (GenBank accession no. AK026010). This fragment encodes the full-size coding region of p100hRho (Fig. 1), with a start codon that matched the Kozak consensus sequence. This cDNA was obtained from Dr. S. Sugano (Institute of Medical Science, University of Tokyo, Laboratory of Genome Structure Analysis, Human Genome Center). The FLJ22357 sequence has 99% sequence identity with three other full-size sequences deposited since this work was finalized, i.e., GenBank accession numbers BC014425 and AK056708, and the NM_022450 sequence in the HUGO database. The latter has been designated the name RHBDF1, which we have adopted in this report.

Figure 1.

Polypeptide sequence and transmembrane topology of p100hRho. A: Alignment of p100hRho, RHBDL, RHBDL2, and Drosophila Rhomboid 1 (dRho). Identical residues are boxed in black, and similar amino acids are shaded. Predicted transmembrane domains are overlined. A predicted PEST sequence is italicized and underlined. Asterisks below the sequence indicate amino acids that are critical for the protease activity of dRho (Urban et al.,2001). Potential N-glycosylation sites in hRho are in bold and indicated with asterisks above the N. B: Schematic comparison of p100hRho and dRho. Transmembrane domains are boxed in black and asterisks mark the potential N-glycosylation sites. Note the very long N-terminal sequence prior to TM1 and the long sequence between TM1 and TM2.

Protein Sequence Analysis and Structural Predictions

Transmembrane domain prediction was performed using the TM-Pred program (Stoffel et al.1993). The program was accessed through the ISREC homepage (http://www.isrec.isb-sib.ch/software/software.html). Alternative predictions were also obtained using the program SOSUI (http://www.tuat.ac.jp/mitaku/adv_sosui). Protein sequence comparisons were performed using the GCG programs GAP and PILEUP (University of Wisconsin Genetics Computing Group). The PILEUP output was displayed by copying its output (in .msf format) into the BOXSHADE program available through the ISREC homepage. PEST domain prediction was performed using the PESTfind program available at the Genetic Data Environment WWW Server at IMB-Jena (http://genome.imb-jena.de/cgi-bin/GDEWWW/menu.cgi).

Northern Hybridization

A human 12-lane tissue Northern blot was purchased from Clontech (Palo Alto, CA). An Eco RI-NotI fragment encoding amino acids 1–653 was chosen as hybridization probe, labeled with 32P-αdCTP using a random primer labeling kit (Amersham, Arlington Heights, IL) and hybridized according to the manufacturer's directions (Clontech). Kodak MR film was used to expose the hybridized membrane.

In Situ Hybridization

Mouse cDNA fragments homologous to p100hRho, HB-EGF, TGF-α and the EGF receptor were used for in situ hybridization. A mouse HB-EGF cDNA was obtained from S. Higashiyama (Ehime University Medical School). Probes were prepared by PCR using primers described below. For TGF-α and EGF receptor cDNAs, oligonucleotide primers were designed based on the mouse cDNA sequences deposited in GenBank (accession numbers U65016, AF275367, respectively). For the mouse TGF-α sequence, the upstream and downstream primers were 5′-gtatcctgttagctgtgtgc and 5′-ttcagaccactgtctcagag, respectively. For the mouse EGF receptor sequence, the upstream and downstream primers were 5′-ctctgagtgcaactagcaac and 5′-actctgcattttcagctgtg, respectively. To obtain a mouse cDNA sequence, the p100hRho sequence was subjected to a BLAST search against a mouse EST database. A partial mouse cDNA sequence, identified in this way, was used as the basis for the oligonucleotide primer design. To avoid potential cross-reactivity with mRNAs for other Rhomboid-related proteins, we selected unique sequences corresponding to non-conserved protein sequences. The upstream and downstream primers were 5′-tgtgcatctatggcatagcg and 5′-catgaagtcacagtactccc, respectively. The cDNA segments obtained following PCR amplification from E13.5 whole embryo mRNA, were subjected to DNA sequencing to confirm their identity.

In situ hybridization was performed on 6.0-μm paraffin sections, as described (Hu and Helms,1999). Plasmids containing the cloned cDNA segments of TGF-α, EGF-R, HB-EGF, and p100hRho were linearized to transcribe 35S-labeled anti-sense or sense riboprobes, and hybridized to tissue sections overnight at 60°C. Slides were washed at high stringency (64°C, 2x SSC) and dipped in emulsion. Slides were developed after 2 weeks. The in situ hybridization signals were pseudo-colored in Photoshop and tissue architecture was determined using Hoechst nuclear stain (Sigma, St. Louis, MO). Images are shown as superimpositions of the in situ hybridization signal with a blue nuclear stain.

Expression Plasmids

The p100hRho cDNA, encoding the full-length coding sequence, was cloned using PCR-based approaches into EcoRI/SalI-linearized pRK5M, pRK5F (Feng et al.,1995) or pRK5HA (Feng et al.,1998), to allow expression of C-terminally Myc-, Flag-, or HA- tagged p100hRho versions, respectively. The 112-bp 5′ untranslated sequence, present in the p100hRho cDNA, obtained from Dr. Sugano, was replaced by the shorter sequence GAATTCAAACCATGAGTGAGGCCCGCAG (underlined EcoRI cloning site and bolded start codon) to increase expression efficiency. Full-size and mutant forms of p100hRho were also subcloned into pXF-myc, which has a Myc-tag sequence followed by the same multicloning sites as one of pRK5M, thus resulting in the addition of an N-terminal Myc tag following the start codon. MhRhoF, a full-size p100hRho with an N-terminal Myc tag and a C- terminal Flag tag, was generated by replacing a 3′ sequence of the N-Myc-tagged p100hRho with the corresponding sequence for C- Flag-tagged p100hRho.

Met399hRho, an N-terminally truncated p100hRho that initiates at the ATG encoding Met399, and Cd3 and Cd2, with stop codons introduced after Val441 and Leu658, respectively, were all generated using PCR-based cloning. The expression plasmids for p100hRho versions with Met399 or Met375 replaced with Leu were identical to the one encoding MhRhoF, except for the defined codon changes. The coding sequence for the Y371stop mutant was generated by replacing the Tyr371 codon in MhRhoF with a stop codon. The mutations were introduced using the Quick Change method (Stratagene, La Jolla, CA).

Three potential N-glycosylation sites were mutated individually or in combination by replacing the Asn codon with a Gln coding using Quick Change (Stratagene). In the NG1, NG2, and NG3 mutants, Asn131, Asn381, and Asn583, respectively, were substituted by Gln. The NG3 mutation was also introduced into the NG1 mutant to generate NG1/3.

We also made an expression plasmid for a short version of p100hRho, named shRho, which has an initiator Met followed by five amino acids connected in frame to Pro319 (Fig. 1), and was C-terminally Flag-tagged by subcloning into pRK5F (Feng et al.,1995).

The expression plasmids pRK7-TGF-α for wild-type TGF-α, pRK7-TGF-αΔC for cytoplasmically truncated TGF-α, and pRK-TGF-αΔE were described previously (Shum et al.,1994; Shi et al.,2000). The HB-EGF expression vector was also previously described (Shi et al.,2000). The coding sequences for Spitz and Star were subcloned from corresponding expression plasmids (Bang and Kintner,2000) into pRK5M and pRK5F (Feng et al.,1995), respectively, using PCR-based approaches. This resulted in expression of C-terminally Myc-tagged Spitz and C-terminally Flag-tagged Star.

To express p100hRho and shRho, described above, in Drosophila, the corresponding cDNA sequences were cloned into pUAST (Brand and Perrimon,1993). The plasmids pUAST-hRho and pUAST-sRho were generated by subcloning the EcoRI/KpnI cDNA fragment from pRK5-hRhoM or pRK5-shRhoF into EcoRI/KpnI-linearized pUAST.

Details of plasmid constructions will be provided upon request.

Antibodies

The monoclonal antibody α1 against TGF-α1 has been described (Bringman et al.1987). The anti-Flag monoclonal antibody M2 was from Sigma. The anti-Myc 9E10 and anti-HA monoclonal antibody HA.11 were purchased from Babco (Richmond, CA). Anti-TGF-α monoclonal antibody Ab-1 was purchased from Oncogene Sciences. Anti-Myc polyclonal antibody A-14 was from Santa Cruz (Santa Cruz, CA). Anti-calnexin polyclonal antibody was purchased from Stressgen (Victoria, BC Canada) and anti-GM130 monoclonal antibody was from Transduction Laboratories.

Transfection, Immunoprecipitation, and Western Blotting

COS-1 cells were grown in DMEM-H16, 3 g/l glucose containing 10% fetal bovine serum (Gibco-BRL), 100 U/ml penicillin, and 100 μg/ml streptomycin. Lipofectamine (Gibco BRL) was employed for transfections. Cells were plated on 100- or 60-mm dishes 1 day before transfection. Forty-eight hours after transfection, cells were collected, washed with PBS, and lysed with RIPA buffer (25 mM Tris-HCl, pH 7.4, 1%Triton, 0.5% deoxycholic acid, 0.1% SDS, 0.15M NaCl, 1 mM EDTA) with protein inhibitors including PMSF, aprotinin, benzamidine, and pepstatin. After centrifugation at 14,000 rpm at 4°C, supernatants were subjected to immunoprecipitation, followed by SDS-PAGE and Westen blotting with appropriate antibody as described (Kuo et al.,2000; Shi et al.,2000). For TGF-α detection, α1 antibody (Bringman et al.,1987) was used at ∼4 μg/ml for immunoprecipitation and a 1:1,000 dilution of anti-TGF-α Ab-1 (Calbiochem) was used for Western blotting. For visualization, ECL (Amersham Bioscience) was used.

Immunofluorescence Microscopy

Hela cells were grown in DMEM-H16, 3 g/l glucose containing 10% fetal bovine serum (Gibco-BRL), 100 U/ml penicillin, and 100 μg/ml streptomycin. Transiently transfected Hela cells were plated on cover slips 24 hr after transfection and allowed to reattach overnight. The cells were then washed with PBS and fixed with cold (-20°C) methanol/acetone for 3 min. After blocking with 5% bovine serum albumin in PBS, the cells were incubated with the antibody of interest in blocking solution for 1 hr at 37°C. Following four 5-min washes with PBS, cells were incubated in a 1:500 dilution of Rhodamine- or FITC-conjugated appropriate antibody (Jackson ImmunoResearch) in blocking solution at room temperature. Stained cells were visualized using a BioRad (Richmond, CA) MRC-1024 confocal microscope.

N-Glycosidase F Treatment

Each protein was prepared using an SP6 in vitro translation kit (Promega) in the presence of canine microsomes (Promega) and labeled with 35S-L-methionine (Perkin Elmer, Boston, MA). Translated proteins were centrifuged at 14,000 rpm for 30 min, and the pellets were denatured by boiling in 0.1% SDS and 1% β-mercaptoethanol. After 5-fold dilution with 50 mM phosphate buffer, 25 mM EDTA, 1% NP40 (pH 7.4), proteins were incubated with 1 U of N-glycosidase F (Roche) for 90 min. The reaction was stopped by adding SDS-sample buffer, followed by boiling for 5 min. Each protein was subjected to SDS-PAGE. The gel was dried and exposed onto Kodak MR Film. Immunoprecipitated p100hRho proteins, expressed in mammalian cells, were subjected to the same procedure described above.

Generation of hRho Transgenic Flies

We subcloned into the Drosophila pUAST expression vector (Brand and Perrimon,1993) the s-hrho cDNA that encodes the short version of hRho starting at Met399, the l-hrho cDNA that encodes full-size p100hRho, and the cDNAs for TGF-α or HB-EGF. These constructs were introduced into Drosophila using standard methods of P-element-mediated germline transformation (Rubin and Spradling,1982).

Chromosomal location of each insertion was determined, and independent balanced lines were established.

RESULTS

Identification of a Human Rhomboid Protein

Using the Drosophila Rhomboid 1(dRho) cDNA sequence (Bier et al.1990) and low-stringency hybridization conditions with PCR-generated probes, we screened human λgt10-based cDNA libraries for hybridizing sequences. This led to the identification of an incomplete cDNA sequence coding for a protein with sequence similarity to dRho. This cDNA sequence was extended through a combination of PCR-based cDNA cloning and computer-based homology searches of human sequence databases (see Materials and Methods section). The cDNA sequence showed identity with a partial sequence of the Dist1 or -89/91 gene, that was identified in the α-globin gene cluster but not recognized as encoding a Rhomboid-related protein (Vyas et al.,1992; Kielman et al.,1993,1996).

The human cDNA sequence encodes an 855–amino acid polypeptide (Fig. 1), which we named p100hRho because of its apparent molecular weight (see below). This sequence is much longer than the 355–amino acid dRho (Bier et al.,1990), and two previously reported human homologs, RHBDL (also called rrp1; Pascall and Brown,1998) and RHBDL2 (Urban et al.,2001). Its sequence corresponds to the RHBDF1 sequence deposited by the HUGO project (GenBank accession number NM_022450). Like dRho and other rhomboids, the p100hRho sequence has seven predicted transmembrane domains, named TM1 to TM7. The similarity of both proteins covers primarily the sequence from just before TM2 throughout TM6, corresponding to five of the six transmembrane helices conserved in all rhomboids (Koonin et al.,2003), and weakly TM7. The amino acids that are conserved between p100hRho and dRho are often also conserved in other rhomboids. In contrast, TM1 in p100hRho is located much further upstream from TM2 than the TM1 in dRho and many other rhomboids. Its location predicts a much larger loop between TM1 and TM2 than in dRho. The p100hRho TM1 is preceded by a hydrophilic sequence of 410 amino acids, i.e., about half of the entire polypeptide sequence, without detectable sequence similarity to other proteins. This long sequence contains a predicted PEST sequence that could be involved in rapid proteolytic degradation (Rechsteiner,1988). Additionally, the p100hRho sequence contains three potential N-glycosylation sites, two before TM1, and one between TM1 and TM2 (Fig. 1).

Drosophila Rhomboid and other rhomboids have been proposed to act as a protease that, in the case of dRho, can directly cleave Spitz in its transmembrane domain. This analysis predicted a role for six amino acids of dRho in the elaboration of its activity as a serine protease (Urban et al.,2001), with two critical residues, S217 and H281, exceptionally conserved among all rhomboids examined (Koonin et al.,2003). Three of these amino acids are conserved in the p100hRho sequence, i.e., R655, G718, and H772 (corresponding to H281 in dRho), while the others are replaced with another amino acid. Most notably, S217 in dRho, which is thought to represent the catalytic center for its function as serine protease, is replaced by A720 in p100hRho (Fig. 1), suggesting that p100hRho may not act as a serine protease.

Tissue Distribution of p100hRho Expression

In Drosophila, the expression of dRho is temporally and spatially controlled. We therefore evaluated the expression of p100hRho mRNA in human tissues using Northern hybridization (Fig. 2), and the expression of its mouse homolog during development using in situ hybridization (Fig. 3).

Figure 2.

Expression of p100hRho mRNA in human adult tissues. The membrane was hybridized with a 1.0-kbp p100hRho 5′ probe. p100hRho mRNA was expressed as two major mRNAs of 4.0 and 3.1 kb.

Figure 3.

Developmental expression of p100hRho in mouse embryos at embryonic day 14.5, in parallel with the EGF receptor, and two EGF receptor ligands. A–E: A sagital section of the entire embryo. F–J: The developing kidney (framed in A). K–O: The developing tooth bud (framed in A). In situ hybridization patterns of the expression of p100hRho (B, G, L), the EGF receptor (C, H, M), TGF-α (D, I, N), and HB-EGF (E, J, O) are shown.

Northern hybridization revealed the presence of two p100hRhomRNAs of 4.0 and 3.1 kb in most human tissues at variable levels, and a 7-kb transcript in heart and skeletal muscle (Fig. 2). These RNAs hybridized with a cDNA fragment encoding the non-conserved, N-terminal half of the p100hRho protein, strongly suggesting that they encode full-length p100hRho protein. A high level expression of p100hRho mRNA was observed in heart, skeletal muscle, lung, and placenta. p100hRho mRNA was expressed at lower levels in colon, kidney, and small intestine and was only barely detectable or absent in brain and peripheral blood cells. The relative expression of p100hRho mRNA was overall consistent with the reported expression of Dist1 mRNA in several tissues (Kielman et al.,1996).

To better review the distribution of p100hRho in tissues during development, we generated an antisense RNA probe for its mouse homolog and carried out in situ hybridizations of sagital sections of mouse embryos. We focused on embryonic day 14.5, when various organs are beginning to undergo epithelial-mesenchymal interactions that define their future tissue architecture. As shown in Figure 3, p100hRhomRNA was widely expressed in epithelial and, more predominantly, mesenchymal tissues. We compared the expression patterns of p100hRho with the EGF receptor using adjacent tissue sections. The EGF receptor and p100hRho showed extensive overlapping in their expression in numerous tissues, with the EGF receptor primarily expressed in epithelial tissues and in the developing perichondrium, e.g., around the vertebral bodies (Fig. 3B,C). We also assessed the mRNA expression of TGF-α and HB-EGF, two ligands for the EGF receptor that were used in the biochemical characterization of p100hRho (see below). TGF-α mRNA was predominantly restricted to the epithelium (Fig. 3D), while HB-EGF showed a wider pattern of mRNA distribution throughout the mesenchyme and epithelium (Fig. 3E).

More detailed comparisons of p100hRho, EGF receptor, TGF-α, and HB-EGF mRNAs were made in the developing kidney (Fig. 3F–J) and tooth bud (Fig. 3K–O). At this stage of development, the mesenchyme of both tissues is responsive to signals from the overlying epithelium. In the developing kidney, p100hRho and the EGF receptor exhibited overlapping expression in the metanephric mesenchyme (Fig. 3G,H), whereas TGF-α expression was primarily restricted to the epithelial metanephric tubule (Fig. 3I), thus complementing the surrounding EGF receptor expression. In the tooth bud (Fig. 3K–O), p100hRho and EGF receptor expression showed extensive similarity, in this case both being expressed in the epithelial component of the developing tooth that will give rise to enamel. Conversely, TGF-α was strongly expressed in the mesenchyme that eventually forms dentin (Fig. 3N). In both the developing kidney and tooth bud, HB-EGF showed a widespread distribution throughout the epithelium and mesenchyme (Fig. 3J,O).

Intracellular Localization of p100hRho

We next analyzed the subcellular localization of p100hRho by immunofluorescence of transfected Hela cells that expressed an N- or C-terminally HA- or Myc-tagged version of p100hRho. The same localization was seen using C- or N-terminally tagged versions, and at high or very low expression levels that resulted in barely detectable staining (data not shown). Since we were unable to generate sufficiently sensitive and specific antibodies, endogenous p100hRho could not be detected, similarly to the results with all other rhomboids studied so far. The cells showed a perinuclear staining and diffuse reticular cytoplasmic staining, suggesting localization of p100hRho in the endoplasmic reticulum (Fig. 4). A staining similar to that of p100hRho was apparent using an antibody for calnexin, a marker of the endoplasmic reticulum. Both staining patterns coincided largely. Since dRho has been localized in the Golgi apparatus, we also carried out immunofluorescence staining for GM130, a Golgi marker. However, the Golgi staining differed substantially from the p100hRho immunofluorecence with some partial overlapping (Fig. 4). Consistent with previous results (Kuo et al.,2000), TGF-α was primarily localized in the Golgi, with additional diffuse staining at the cell surface (Fig. 4, data not shown). The p100hRho staining overlapped largely with that of TGF-α, but no cell surface expression of p100hRho was detected on permeabilized or intact cells (data not shown). Together, these data suggest that p100hRho is primarily localized in the endoplasmic reticulum, extending, however, into the Golgi.

Figure 4.

Intracellular localization of p100hRho in transfected Hela cells. Cells were transfected with 200 ng total plasmid DNA expressing C-terminally HA-tagged p100Rho with or without TGF-α. A: Colocalization by immunofluorescence of p100hRho and the endoplasmic reticulum marker calnexin. B: p100hRho partially colocalized with the cis-Golgi marker GM130. As in A, only transfected cells expressed detectable p100hRho, whereas all cells expressed endogenous GM130. C: Co-expression of p100hRho and TGF-α. The patterns were partially overlapping. Note that cell surface TGF-α does not overlap with p100hRho. The same immunofluorescent staining patterns were apparent when cells were transfected with 50 ng plasmid DNA, or when N-terminally or C-terminally tagged p100hRho were expressed.

p100hRho Is Expressed as Two Isoforms

To characterize the p100hRho protein, we expressed N- and C-terminally tagged versions, and a version with an N-terminal Myc tag and a C-terminal Flag tag (named MhRhoF), in transfected COS cells. Double-tagged p100hRho and N-terminally Myc-tagged p100hRho were detected using a Myc antibody as a single band with an apparent molecular weight of about 100 kd, as assessed by SDS-PAGE (Fig. 5A, left; data not shown). This size is consistent with the predicted size of the 855–amino acid sequence. In contrast, double tagged p100hRho and C-terminally Flag-tagged p100hRho were detected using Flag antibody as two proteins (Fig 5A, right; data not shown). The larger proteins detected with either antibody had the same size and corresponded to full-size p100hRho protein. The smaller protein had an apparent molecular weight of 52 kd, which we called p52hRho. The size of p52hRho and its detection using the Flag antibody suggested that it corresponds to a segment that extends from the C-terminus to a site immediately upstream from the TM1 sequence. p52hRho was similar in size to an N-terminally truncated version of p100hRho that was initiated at the ATG encoding Met399 (Fig. 5B). The latter protein migrated as two bands for reasons unclear to us, but possibly due to proteolytic cleavage close to its N-terminus (see below).

Figure 5.

p100hRho is expressed as a large and small form. A: N-terminally Myc tagged and C-terminally Flag tagged p100hRho (MRhoF) were expressed in COS cells, and immunoprecipitated with anti-Myc or anti-Flag. Anti-Myc immunoprecipitation revealed one band corresponding to full-size p100hRho, while anti-Flag detected full-size p100hRho and a smaller form, p52hRho. B: C-terminally Myc-tagged p100hRho (hRhoM) was detected using anti-Myc antibody as two bands, full-size p100hRho and p52hRho (lane 2), while N-terminally Myc tagged p100hRho (MhRho) was detected as one band (lane 3). p52hRho migrated with a similar mobility as the doublet corresponding to C-terminally Myc-tagged Met399hRho, a p100hRho deletion mutant that starts at Met399. C: Expression of p100hRho and p52hRho from the wt MhRhoF coding sequence, the M399L and M375L mutants, and a coding sequence with the Y371 codon replaced by a premature stop codon. Asterisks identify IgG bands.

The size of p52hRho suggested that it could have been derived through internal translational initiation within the p100hRho coding region at the ATG corresponding to Met375 or Met399 (Fig. 1A). Both ATGs are located in a sequence context favorable for efficient translational initiation. To evaluate if the smaller p100hRho protein arises from initiation at Met375 or Met399, we mutated these codons of the double-tagged p100hRho to encode leucine. As shown in Figure 5C, both the M375L and M399L mutants expressed the smaller protein, similarly to wild-type p100hRho. The double mutant with both M375 and M399 changed to Leu, also expressed the same size of protein (data not shown). We also introduced an in-frame stop codon that replaced Tyr371, to evaluate if translation of full-size p100hRho is required for the generation of the smaller form. This mutation resulted in the expression of an N-terminally tagged protein of the predicted size, but not of p52hRho (Fig. 5C). These results argue against alternative splicing or internal initiation, and thus suggest that p52hRho derives from full-size p100hRho through proteolytic cleavage. We were unable to detect the expected N-tagged protein fragment that would be generated from full-size p100hRho, concomitantly with p52hRho (Fig. 5A). This could be explained through rapid degradation, presumably related to the presence of a PEST sequence (Fig. 1A).

p100hRho Is Expressed as a Dimer

Coimmunoprecipitation experiments using differentially tagged p100hRho versions revealed that p100hRho is expressed as a dimer or oligomer. Indeed, in cells co-transfected with plasmids for N- and C- terminally tagged p100hRho, immunoprecipitation of a C-terminally Flag-tagged p100hRho allowed for coprecipitation of N-terminally Myc-tagged p100hRho (Fig. 6A). To evaluate if the N-terminal half of p100hRho, upstream from TM1, is required for dimerization, we carried out similar coimmunoprecipitations using the short p100hRho version that starts at Met399 (Met399hRho) and thus resembles dRho in size. When C-terminally Flag-tagged or Myc-tagged versions were coexpressed, immunoprecipitation of the Myc-tagged version, resulted in coprecipitation of the Flag-tagged version (Fig. 6B). Met399hRho consistently migrated as a double band, as noted before (Fig. 5B), along with a somewhat diffuse band of twice the predicted size presumably corresponding to irreversible dimers. Together, these data demonstrate that p100hRho forms dimers or oligomers, without a requirement of its N-terminal half that precedes TM1.

Figure 6.

Dimerization of p100hRho. A: C-terminally Flag-tagged and N-terminally Myc-tagged p100hRho were expressed alone or together in COS cells (lanes 1–3). Double-tagged p100hRho was also expressed (lane 4). p100hRho was immunoprecipitated with anti-Flag antibody, followed by SDS-PAGE, and detected by Western blot using anti-Myc antibody. Immunoprecipitates using anti-Flag antibody from cells expressing both p100hRho cDNAs contained Myc-tagged p100hRho (lane 3, left) in spite of the lower protein levels compared to singly transfected cell (middle and right, lane 3 vs. lanes 1 and 2). B: A short form of p100hRho starting at Met399 forms dimers. C-terminally Flag-and Myc-tagged versions of Met399hRho were expressed alone (lanes 2–3) or together (lane 4) in COS cells (lane 4). Met399hRho was immunoprecipitated with anti-Myc antibody, followed by SDS-PAGE, and detection by Western blot using anti-Flag antibody. Immunoprecipitates of cells expressing both Met399hRho cDNAs using anti-Myc antibody contained Flag-tagged Met399hRho (lane 4, left). Met399hRho migrates as a double band. The open arrow marks the irreversible aggregates, presumably dimers, of Met399hRho. Asterisks identify IgG bands.

Transmembrane Topology of p100hRho

Drosophila Rhomboid has been proposed to be a seven-transmembrane protein, based on its hydropathy profile. Experimental evidence suggests that its N-terminal segment is located at the cytosolic side, thus resulting in a luminal orientation of its C-terminus (Lee et al.,2001). p100hRho has moderate sequence similarity with dRho in the sequence that covers TM2 to TM7 (amino acids 640–825; Fig. 1A) with similar spacing between transmembrane domains. The much longer sequence of p100hRho, including a long N-terminal extension preceding TM1 and the possibility of an additional transmembrane segment close to the N-terminus, led us to evaluate the transmembrane topology of p100hRho.

The amino acid sequence of p100hRho contains three predicted N-glycosylation sites. Two of these, Asn131 and Asn381, precede TM1, while a third site, Asn583, is located in the loop between TM1 and TM2 (Fig. 7A). Since N-glycosylation only occurs at the luminal side of transmembrane proteins, N-glycosidase F treatment allows assessment of not only the glycosylation status but also of the transmembrane topology. As shown in Figure 7B, N-glycosidase F treatment of immunoprecipitated p100hRho noticeably decreased the apparent molecular weight of both p100hRho and p52hRho, indicating the presence of N-glycosylation. Since p52hRho does not contain the first two potential N-glycosylation sites, this result suggests that Asn583 is likely to be glycosylated.

Figure 7.

Determination of glycosylation and transmembrane topology of p100hRho. A: Schematic diagram of full-size p100hRho with its seven proposed transmembrane segments and three candidate N-glycosylation sites (asterisks), and of p100hRho mutants with mutated N-glycosylation sites (marked with x; Asn131 or Asn583 replaced by Gln in NG1 or NG3, respectively). B: p100hRho (top) and p52hRho (bottom) are N-glycosylated. Cells were transfected with p100hRho expression plasmid, followed by immunoprecipitation and N-glycosidase F (+) or mock (-) treatment, and analyzed by SDS-PAGE and Western analysis. The increased mobility following N-glycosidase treatment, which also affects the IgG band, is indicative of N-glycosylation. Bullets mark the specific bands. C: Full-size p100hRho, either wild-type or with the NG1 and/or NG3 mutations, were in vitro transcribed and translated in the presence of microsomal membranes, and subjected to N-glycosidase F treatment. Wild-type and NG1 p100hRho showed a mobility shift following N-glycosidase F treatment, while the NG3 mutants were not detectably N-glycosylated.

Replacement of Asn131 by Gln (the NG1 mutation) did not affect the mobility of p100hRho in gel, and the mobility shift following N-glycosidase F treatment was similar in wild-type p100hRho and the NG1 mutant (Fig. 7C). Similar results were obtained using the NG2 mutant, in which Asn381 was replaced by Gln (data not shown). Mutation of Asn583 into Gln (the NG3 mutant) increased the mobility of this form to that of the deglycosylated form and abolished the effect of N-glycosidase F on the mobility of the protein in gel (Fig. 7C). We, therefore, conclude that Asn583 is N-glycosylated and located at the luminal side. The absence or presence of the NG1 mutation in the NG3 mutant did not affect the mobility of p100hRho. Instead, NG3 p100hRho and NG1/3 p100hRho ran with identical mobilities and were not affected by N-glycosidase F (Fig. 7C). Finally, an NG3 mutant with an introduced potential N-glycosylation site close to its N-terminus behaved similarly to the NG3 mutant (data not shown).

These results indicate that the segment between TM1 and TM2 is luminally oriented and glycosylated. The sequence preceding TM1 is not glycosylated, in spite of two putative N-glycosylation sites, and is therefore likely to be cytosolic. The presence of seven transmembrane regions consequently orients the C-terminus into the lumen, similarly to the transmembrane orientation of dRho (Lee et al.,2001).

Interactions of p100hRho With the Ectodomain of Transmembrane TGF-α Family Proteins

Since dRho has been shown to cleave Spitz within its transmembrane domain (Urban al.,2001,2002a,b), we evaluated whether p100hRho interacted with transmembrane TGF-α as well as the related transmembrane HB-EGF. As shown in Figure 8A, immunoprecipitation of p100hRho did coprecipitate transmembrane TGF-α. However, not all transmembrane TGF-α forms interacted with p100hRho. Only the two lower forms of transmembrane TGF-α that correspond to immature and unglycosylated TGF-α (Bringman et al.,1987) interacted with p100hRho, while the larger N-glycosylated TGF-α form associated only weakly if at all. p100hRho also associated with transmembrane HB-EGF, but interacted selectively with only one of the multiple HB-EGF forms (Fig. 8B). Both p100hRho and dRho also showed interaction with Spitz in transfected cells (Fig. 8C, data not shown), consistent with the ability of dRho to cleave the transmembrane segment of Spitz. p100hRho did not interact with the EGF receptor (data not shown).

Figure 8.

Interaction of p100hRho with TGF-α family members. A–C: Cells were transfected with expression plasmids for TGF-α (A), HB-EGF (B), or Spitz (C), with or without C-terminal Flag-tagged p100hRho plasmid (as marked). Left: The cell lysates were subjected to anti-Flag immunoprecipitation for p100hRho, followed by Western blotting with anti-TGF-α monoclonal antibody, Ab-1 (A), anti-HB-EGF (B), or anti-Myc to detect C-tagged Spitz (C). Specific bands corresponding to transmembrane TGF-α, HB-EGF, or Spitz coprecipitated with p100hRho. Right: TGF-α, HB-EGF, or Myc-tagged Spitz, detected in the cell lysates using direct Western blotting. D: The EGF core domain of transmembrane TGF-α is required for interaction with p100hRho. Cells were transfected with C-terminal Flag tagged p100hRho and full-length TGF-α, TGF-αΔC, or TGF-αΔE. Cell lysates were subjected to immunoprecipitations using anti-Flag (to precipitate p100hRho), followed by Western blot to detect associated TGF-α. TGF-α and TGF-αΔC were detected using Ab-1, while TGF-αΔE and Spitz were detected using anti-Myc. Middle: The expression of p100hRho, as assessed using anti-Flag immunoprecipitation. Bottom: The expression level of TGF-α and its derivatives, as assessed by Western blotting of cell lysates with corresponding antibodies. Asterisks identify non-specific bands.

We next evaluated which segment of transmembrane TGF-α interacted with p100hRho (Fig. 8D). Deletion of the cytoplasmic domain, as in TGF-αΔC, did not abolish the association of transmembrane TGF-α with p100hRho. In contrast, replacement of the 50–amino acid EGF-like core ectodomain with an unrelated Myc epitope tag sequence, in TGF-αΔE, abolished the interaction of transmembrane TGF-α with p100hRho (Fig. 8D). We, therefore, conclude that the EGF-like domain of transmembrane TGF-α, and presumably of other TGF-α-like proteins, plays a key role in the association with p100hRho. This also implies that ectodomain sequences of p100hRho are involved in this interaction.

Star Interacts With the Ectodomain of Transmembrane TGF-α Family Proteins

Star, a type II transmembrane protein with its N-terminus oriented into the cytosol, has been implicated in the role of dRho in Spitz processing (Urban et al.,2001; Tsruya et al.,2002). No mammalian Star homolog has as yet been reported. We evaluated the interactions of Drosophila Star with p100hRho and dRho, and with the EGF receptor ligands. Star was expressed with an apparent molecular weight of 75 kd and ran on gel as a doublet, presumably a result of glycosylation or proteolysis (Fig. 9A).

Figure 9.

Interaction of Star with transmembrane TGF-α ligands and colocalization with p100hRho. A: C-terminal Flag tagged Star is expressed in transfected COS cells and, following immunoprecipitation, runs as a 75-kd protein on gel. B–D: Interaction of Star with TGF-α, HB-EGF, or Spitz. Cells were transfected with expression plasmids for TGF-α (B), HB-EGF (C), or Spitz (D), with or without Star (as marked). Left: The cell lysates were subjected to anti-Flag immunoprecipitation for Star, followed by Western blotting with anti-TGF-α (B), anti-HB-EGF (C), or anti-Myc to detect C-tagged Spitz (D). Specific bands corresponding to transmembrane TGF-α, HB-EGF, or Spitz coprecipitated with Star. Right: TGF-α, HB-EGF, or Myc-tagged Spitz, detected in cell lysates using direct Western blotting. E: The EGF core domain of transmembrane TGF-α is required for interaction with p100hRho. Cells were transfected with C-terminal Flag tagged Star and full-length TGF-α, TGF-αΔC, or TGF-αΔE. Cell lysates were subjected to immunoprecipitations using anti-Flag antibody (to precipitate Star), followed by Western blot to detect associated TGF-α. TGF-α and TGF-αΔC were detected using anti-TGF-α, while TGF-αΔE and Spitz were detected using anti-Myc. Middle: The expression of Star, as assessed using anti-Flag immunoprecipitation. Bottom: The expression level of TGF-α and its derivatives, as assessed by Western blotting of cell lysates with corresponding antibodies. F: Immunofluorescence detection of p100hRho and Drosophila Star in transfected cells. Asterisks identify non-specific bands.

Using coimmunoprecipitation analyses, we found that Star physically interacted with transmembrane TGF-α (Fig. 9B) and HB-EGF (Fig. 9C), which is consistent with its ability to interact with Spitz (Hsiung et al.,2001; Fig. 9D). To define the interacting domains in transmembrane TGF-α, we analyzed the ability of Star to interact with different deletion mutants of transmembrane TGF-α. As with the interaction with p100hRho (Fig. 8D), the interaction of transmembrane TGF-α with Star required the EGF-like core ectodomain of TGF-α. Thus, Star interacted with TGF-αΔC, lacking the cytoplasmic domain of TGF-α, but did not interact with TGF-αΔE, which lacks this core ectodomain (Fig. 9E). We, therefore, conclude that Star interacts with TGF-α family members through ectodomain sequences of both Star and TGF-α-like proteins.

We also evaluated the physical interaction of Drosophila Star with dRho or p100hRho. No association was detected using coimmunoprecipitation analyses (data not shown). Nevertheless, a loose interaction may exist since both hRho and Star colocalized inside the cell (Fig. 9F). Indeed, the intracellular distribution of Star, as assessed by immunofluorescence, was identical to that of p100hRho. Thus, similarly to p100hRho, Star is intracellularly localized in the endoplasmic reticulum and Golgi.

Assaying the Function of p100hRho in Drosophila

To test the activity of p100hRho in the context of a developing organism, we examined the phenotypes resulting from expression of full-size p100hRho (l-hRho) or a short version of p100hRho (s-hRho), with most of its N-terminal half preceding TM1 deleted (see Materials and Methods section), in transgenic flies using the GAL4/UAS expression system. Both UAS transgenes were expressed under the control of wing-specific GAL4 driver (wingGAL4, also referred to as MS1096GAL4; Capdevila and Guerrero,1994) at high levels throughout the wing primordium during larval and pupal development. We observed that individuals expressing s-hrho had a strong wing phenotype consisting of small, blistered wings with thickened veins and occasional notches in the wing margin (Fig. 10C). In contrast, flies expressing l-hrho had no detectable defects (data not shown). The failure of l-hrho to induce a phenotype did not result from poor expression or protein instability, since histochemical staining revealed high protein expression from the UAS-l-hrho construct in the wing disc (data not shown).

Figure 10.

p100hRho is active in flies and synergizes with HB-EGF. A: Wild-type Drosophila wing. B: Wing from a fly ubiquitously expressing a full-length copy of the Drosophila rhomboid gene (d-rho) in the wing primordium during larval and pupal development, which has an all-vein phenotype. C: Wing from a fly ubiquitously expressing a strong insertion of the short version of the p100hRho gene (s-hrho) in the wing primordium. D: Wing from a fly ubiquitously expressing a truncated form of the Drosophila rhomboid gene, (d-rhoΔC), which lacks sequences encoding the C-terminal extracellular domain (the last 26 amino acids) of dRho, in the wing primordium. E: Wing from a fly expressing a weak insertion of s-hrho, which has thickened veins, but normal wing size. F: Wing from a fly expressing a weak insertion of d-rhoΔC, with thickened vein material restricted to the tips of L2 and L4. G: Wing from a fly co-expressing the same weak insertion of s-hrho as in E, with HB-EGF. The strong wing phenotype is similar to that caused by the strong s-hrho insertion in C. H: Wing from a fly expressing the same weak insertion of d-rhoΔC as in F, along with m-spi (the full-length version of spi). This strong phenotype is similar to that of the strong d-rhoΔC insertion shown in D. I: Wild-type expression of the Cut protein in cells along the wing margin in a third larval instar wing imaginal disc. J: Expression of the Cut protein is lost in cells along the wing margin of a third larval instar wing imaginal disc expressing the s-hrho gene throughout the wing primordium. A similar result is observed in discs expressing the rhoNeo allele (Guichard et al.,2002), or d-rhoΔC (data not shown).

The phenotype caused by s-hrho expression differed from that resulting from ectopic expression of the d-rho gene, which consists of a solid ectopic vein phenotype in an approximately normal size wing (Fig. 10B). Similar extra-vein phenotypes are observed in various situations in which EGF-R/MAPK signaling is ectopically activated (Sturtevant et al.,1993; Guichard et al.,1999). In contrast, the phenotype resulting from expression of the UAS-s-hrho, i.e., thickened veins and notched margin, is reminiscent of mutants in the Notch signaling pathway (Guichard et al.,2002; http://flybase.bio.indiana.edu). Interestingly, flies expressing a truncated dRho lacking the C-terminal portion after TM7 (encoded by the rhoΔC allele) or a truncated form lacking 6 of the 7 transmembrane domains and half of the first loop of Rho (referred to as the rhoNeo allele), have phenotypes that are very similar to that caused by expression of s-hrho (Fig. 10C–F; Guichard et al.,2002). Expression of s-hrho and rhoΔC indeed reduced the activity of the Notch pathway, as revealed by the loss of specific downstream target gene expression, such as cut (Fig. 10I,J, data not shown). Cumulatively, these data suggest that s-hrho and rhoΔC have similar activities and reduce Notch signaling.

As s-hrho did not induce ectopic veins typical of EGF-R activation, we reasoned that it may require the presence of missing vertebrate partners such as TGF-α-family ligands to reveal its normal activity. To address this possibility, we generated transgenic lines carrying UAS-TGF-α, or UAS-HB-EGF constructs. When these vertebrate ligands were expressed alone under the control of the wingGAL4 driver, they produced little if any phenotype (e.g., HB-EGF induced a weak phenotype consisting of slightly smaller and curved wing blades with ectopic vein speckles and occasional vein truncations, data not shown). Similarly, mis-expression of the Drosophila TGF-α-family members m-Spi or m-Grk have no detectable effect, unless co-expressed with star and rho (Pickup and Banerjee1999; Guichard et al.,1999,2000; Queenan et al.,1999), which facilitate their transit to the Golgi apparatus and cleavage into diffusible forms (Lee al.,2001; Urban et al.,2001,2002a). To test for similar interactions between h-rho and human TGF-α ligands, we co-expressed a weak UAS-s-hrho insertion (Fig. 10E) with TGF-α or HB-EGF. Although co-expression with TGF-α did not lead to a significant change in the weak s-hrho phenotype (data not shown), co-expression of HB-EGF with s-hrho greatly intensified the s-hrho phenotype (Fig. 10G, compare with Fig. 10E), to a level similar to that seen in wingGAL4>s-hrho(stg) flies, which consist of very small wings with thickened veins (Fig. 10G, compare with Fig. 10C). This potent enhancement of s-hrho activity by HB-EGF further supports the hypothesis of a functional connection between the two molecules. In parallel experiments, we co-expressed a weak d-rhoΔC insertion with m-spi or m-grk, and observed a strong synergistic enhancement of the d-rhoΔC phenotype (Fig. 10H, compare with Fig. 10F). These in vivo results strongly suggest functional cooperation of s-hrho and HB-EGF.

DISCUSSION

In Drosophila, Rhomboids have been shown to act as serine proteases for TGF-α-like ligands, and two of these been implicated in ligand-induced activation of the EGF receptor (Guichard et al.,2000; Lee et al.,2001; Urban et al.,2001,2002a). In yeast, the rhomboid-like protein Rbd1p has been shown to cleave Mgm1p, a mitochondrial GTPase, and to be required for mitochondrial membrane remodeling (McQuibban et al.,2003). Very little is known about vertebrate Rhomboid-like protein. Four mammalian rhomboid-like sequences have been published (Pascall and Brown,1998; Urban et al.,2001; Pellegrini et al.2001; Jaszai and Brand,2002), but the characterization of these proteins has been limited. We now present the identification of a novel human Rhomboid-like protein with an atypically long N-terminal, cytoplasmic sequence, and have explored its function and activity and biochemical characterization. This report provides some findings of relevance for other rhomboid family members.

With its 855 amino acids, p100hRho is more than twice the size of Drosophila Rhomboid and other rhomboids. Six of the seven predicted transmembrane domains (TM2 to TM7) align with the conserved six transmembrane helices that characterize the rhomboids, while the sequence from just before TM2 to TM6 shows conservation with dRho and RHBDL1 and 2. TM1 is located much further upstream from TM2 than the dRho TM1, thus creating a large loop between the two transmembrane segments, which, based on the transmembrane topology, is located at the luminal side. Analysis of the transmembrane topology also determined that the long hydrophilic sequence preceding TM1, which constitutes about half of the polypeptide, is oriented into the cytosol.

p100hRho does not contain all the critical amino acids that are required for intramembranous serine protease activity in dRho and are conserved in several other Drosophila Rhomboids (Urban et al.,2002a,b). Most importantly, S217 in dRho, which is thought to represent the catalytic center for its function as serine protease and is highly conserved among the many rhomboids (Koonin et al.2003), is replaced by an alanine in the p100hRho sequence. Accordingly, we did not detect protease activity using Spitz or TGF-α as substrates (data not shown). The lack of some critical residues, and consequently of protease activity, in p100hRho and some other rhomboid proteins (Koonin et al.,2003), strongly suggest that rhomboids have functions besides intramembranous proteases.

Remarkably, p100hRho was expressed as two forms that may arise from post-translational processing. The larger form is the full size protein, while the shorter one corresponds to the C-terminal half without the long cytoplasmic segment. p52hRho thus resembles in size and transmembrane topology other seven-transmembrane rhomboids, such as dRho and RHBDL1, 2, and 3. The presumed cleavage of p100hRho to generate p52hRho resembles the cleavage of PARL, which similarly to p100hRho is cleaved to give rise to a shorter form with seven-transmembrane helices (Sik et al.,2004). We were unable to detect the N-terminal half of p100hRho, possibly due to rapid degradation, resulting from the presence of PEST sequences in this segment. Nevertheless, this cleavage suggests a regulatory role and/or a separate function for this domain, as was postulated for the segment that is proteolytically cleaved from PARL (Sik et al.,2004).

Immunoprecipitations of differentially tagged p100hRho versions also revealed that p100hRho forms dimers or oligomers. Dimerization is likely since we detected irreversible complexes of a size compatible with dimer formation. This dimerization was mediated by the segment comprising the transmembrane domains, and not the N-terminal half of p100hRho, suggesting that dimerization may be an intrinsic property of the rhomboids, as supported by the dimerization of dRho (data not shown). Dimerization could also explain how wild-type PARL can supply in trans Pβ cleavage to a catalytically inactive PARL, which by itself can not be processed to release Pβ (Sik et al.,2004). Considering the existence of multiple rhomboids in Drosophila and vertebrates, it is conceivable that rhomboid homo- and heterodimerization provide an additional level of versatility.

We also demonstrated that p100hRho can interact with TGF-α family members and that this interaction requires the EGF core sequence in the ectodomain. In contrast, deletion of the cytoplasmic domain of TGF-α did not affect the interaction with p100hRho. Star can also interact with TGF-α family members and this interaction similarly required the EGF core, which is consistent with the interaction of the ectodomains of Spitz and Star (Hsiung et al.,2001; Tsruya et al.,2002). Although no mammalian homologs of Star have been found, the specific interaction with the EGF domain in mammalian TGF-α family proteins suggests the existence of Rhomboid-Star systems in mammalian cells, as in Drosophila (Guichard et al.,1999; Tsruya et al.,2002). Our data thus indicate that the EGF core of TGF-α proteins plays a critical dual role, i.e., in the interaction of transmembrane TGF-α-like proteins with Rhomboid, and, once externalized by the cell, in EGF receptor binding. It should also be noted that thrombomodulin, which can be cleaved by RHBDL2 (Lohi et al.2004), also contains EGF repeats in its extracellular domain, raising the possibility that these repeats, which can be found in a variety of transmembrane proteins, play a physiological role in the recognition by rhomboids. The presumed 1:1 stoichiometry of rhomboid interactions with TGF-α family members (or other transmembrane proteins) furthermore raises the possibility that rhomboid association results in dimerization of the transmembrane protein during transport.

We also noted that p100hRho interacts primarily with unglycosylated and immaturely glycosylated forms of transmembrane TGF-α and with a single form of HB-EGF, among the differentially glycosylated and processed forms. Thus, the glycosylation state of the pro-domain may be a determinant of the growth factor's ability to interact with rhomboids, and progression of glycosylation may play a role in the dissociation of this complex. The preferential interaction of p100hRho with immature transmembrane TGF-α proteins is consistent with its predominant localization in the endoplasmatic reticulum extending into the cis Golgi. This localization stands in contrast with the predominant localization of the Drosophila Rho1, 2, and 3 in the Golgi and Rho4 at the cell surface and in the Golgi (Urban et al.,2002a), and RHBDL1, 2, and 3 in the Golgi and/or at the cell surface of transfected cells (Lohi et al.,2004).

We also tested the activities of full-size p100hRho and its short form without its long N-terminal sequence, in transgenic Drosophila of various genetic backgrounds. The developing wing primordium provides a particularly sensitive assay to reveal ectopic EGF receptor activity, which results in the formation of extra wing veins (Sturtevant et al.,1993). The short version of p100hRho induced a strong wing phenotype, similarly to several Drosophila Rhomboids and a bacterial, rhomboid-like protein from P. stuartii (Gallio et al.,2002). In contrast, full-length p100hRho had no detectable activity, even though it was expressed at similar levels as the short version. This observation indicates that the N-terminal cytoplasmic extension inhibits the activity of p100hRho and suggests a regulatory function for this segment in its normal physiological context. It is intriguing that a short fragment of dRho is always detected in Drosophila (Guichard et al.,2002).

In contrast to dRho, expression of short p100hRho in the wing primordium did not induce ectopic veins as observed upon EGF receptor activation, but instead caused phenotypes typically observed when Notch signaling is impaired. This phenotype was very similar to the effect of a truncated form of dRho (d-RhoΔC) that lacks the C-terminal 24 residues and has reduced serine protease activity in cell culture (Lee et al.,2001). The “Notch-like” phenotypes induced by short p100hRho or dRhoΔC were greatly enhanced upon co-expression of their interacting transmembrane growth factors, i.e., HB-EGF or Spitz, respectively. Our results thus illustrate a functional in vivo interaction between p100hRho and HB-EGF, consistent with their similar expression patterns in mouse embryo and their physical association. Moreover, expression of short p100hRho did not rescue the lack of Drosophila rho expression in a rho[ve] mutant background (data not shown). The apparent inhibition of Notch signaling by short p100hRho and truncated forms of dRho may reflect a requirement for the C-terminal sequence in preventing interference with Notch signaling. In this view, the divergent C-terminus of p100hRho may functionally resemble the deletion of the corresponding sequence from endogenous dRho. These data raise the possibility that p100hRho, like truncated forms of dRho that are generated in vivo (Guichard et al.,2002), may also interfere with a component of the Notch signaling pathway in vertebrates.

To gain insight into the function of p100hRho, we examined its expression and tissue distribution in comparison with TGF-α, HB-EGF, and the EGF receptor. In Drosophila, dRho expression occurs in a pattern that resembles that of the EGF receptor activation site, while Spitz is ubiquitously expressed (Rutledge et al.,1992). In human tissues and in the E14.5 mouse embryo, p100hRho is expressed in a complex pattern that partially overlaps with that of the EGF receptor, suggesting a possible link with EGF receptor activation. No clear correlation with TGF-α or HB-EGF expression was apparent, although p100hRho was coexpressed with these ligands in many tissues. While informative in their own right, these analyses need to be extended to take into account the full complexity of the TGF-α family of ligands and the four EGF receptor family members.

In spite of extensive efforts, we currently do not know the endogenous partner transmembrane protein(s), with which p100hRho normally physically and functionally interacts. The interaction of dRho with Spitz, the abilities of different Drosophila Rhomboids to cleave the TGF-α-like proteins Spitz, Gurken, and Keren, and the role of the EGF core sequence in the physical interactions with TGF-α-like proteins all suggest that p100hRho may normally interact with one or several TGF-α-like ligands. However, EGF-like domains are also present in Notch family members and various other proteins including cell adhesion proteins; thus, the function of rhomboid family members may not be restricted to TGF-α-like proteins. Considering its subcellular localization and the lack of protease activity, p100hRho, similarly to a few other rhomboids that lack critical residues thought essential for protease activity (Koonin et al.,2003), may play a role in transport and maturation of selected transmembrane proteins with an EGF core sequence through different compartments to the cell surface and thus be involved in their presentation. Extensive studies will be required to address the normal physiological roles of the vertebrate rhomboids.

Acknowledgements

This research was supported by NIH R01 grant CA54826 to R.D., NSF IBN-0094634 to E.B., and NIH KO2 DE00421, NIH R29 DE12462, and NIH P60 DE 13058 to J.H.

Ancillary