Biosynthesis, post-translation modification, and functional characterization of Drm/Gremlin.

Down-regulated by mos (Drm)/Gremlin is a highly conserved protein whose properties and expression pattern suggest a role in early development, tissue-specific differentiation, and cell transformation. We have investigated the biosynthesis and processing of Drm expressed endogenously in rat fibroblasts or overexpressed following transient or stable transfection. Analysis of metabolically labeled cells revealed that Drm exists in secreted and cell-associated forms that exhibit similar mobilities in SDS-polyacrylamide gel electrophoresis. Protein analysis indicated that Drm is present in two major species: a slow migrating glycosylated form and a nonglycosylated form. Both forms of Drm are able to undergo phosphorylation. Drm is released into the media within 30 min of synthesis and is detectable for up to 4-5 h, whereas the cell-associated form has a half-life of about 1 h. Confocal immunofluorescent microscopy indicates that Drm is present both on the external surface of expressing cells, as well as within the endoplasmic reticulum and the Golgi. Both glycosylated and nonglycosylated forms of Drm exhibit identical distributions and are able to antagonize bone morphogenetic protein signaling. Like the soluble form, the cell-associated forms are capable of binding (125)I-bone morphogenetic protein-4. These properties are consistent with a role for Drm in interfering with signaling and indicate that Drm may act at the cell surface during tissue development and transformation.

The down-regulated by Mos (drm) 1 gene was originally iso-lated during a differential screen of a transformation-resistant revertant of v-mos-transformed rat fibroblasts (1). Drm was identified on the basis of its high level of expression in nontransformed fibroblasts and the loss of expression following transformation by a variety of viral oncogenes, including v-mos, v-raf, and v-ras. Whereas a number of genes have been identified whose expression is suppressed by oncogene-mediated transformation (2,3), analysis of the cDNA sequence of drm revealed that it represented a novel gene containing a cysteinerich repeat region (1). This motif is also found in DAN, a previously identified tumor suppressor (4), the Xenopus gene cerberus (5), as well as the human Muc2 (6). A portion of this conserved repeat structure has been termed a cystine knot (7)(8)(9) and is also shared by members of the tumor growth factor-␤ family, platelet-derived growth factor, nerve growth factor, and other secreted proteins (10).
Interestingly, the Xenopus homolog of drm, designated Gremlin, was isolated during a screen for proteins, which would act on Xenopus explants as dorsalizing factors (11), and was shown, along with cerberus and DAN, to antagonize bone morphogenetic protein (BMP) function. Both XeGremlin/drm and DAN appeared to bind BMP and act in a fashion similar to the previously identified pattern-inducing genes noggin and chordin (12,13). It was thus suggested that Gremlin/drm, DAN, and cerberus made up a family of related secreted proteins whose members functioned during differentiation to interfere with the interactions of specific tumor growth factor-␤ ligands with their receptors (9,11). The peptide sequences of Xenopus, rat, human, mouse, and chicken drm revealed a high degree of identity across species (11), 2 suggesting that they play an important role in conserved cell structures or functions. Consistent with this hypothesis, Drm/Gremlin has recently been reported to function in limb bud development in both mouse (27) and chicken (28). However, in addition to its possible role as a regulator of early development, our previous results indicated that Drm expression is highly regulated in various adult rat tissues and is particularly expressed in terminally differentiated cells (1). Taken together, these results suggest drm is likely to play an important role in regulating multiple cell functions both during early development and in adult tissues.
These properties and potential functional significance led us to undertake the biochemical characterization of the Drm gene product. Analysis of the predicted amino acid sequence indicated the presence of several significant features, including potential nuclear localization signals near the C terminus, potential N-linked glycosylation sites, and multiple potential sites for phosphorylation. Our results indicate that many of these features are active in regulating the processing and localization of the Drm protein. Specifically, we show that Drm is a glycosylated, phosphorylated, secreted protein present both on the external cell surface and within the ER-Golgi compartments. We demonstrated that Drm is secreted via the constitutive secretory pathway in both glycosylated and nonglycosylated forms and that both forms of Drm can antagonize BMP-4 in C2C12 differentiation assays. In addition, we show that cell surface-associated Drm is capable of binding specifically to 125 I-BMP-4, indicating that it also has the potential to interfere with signaling. This report describes the first detailed characterization of the processing and localization of a member of the Drm/Gremlin family and suggests properties of potential functional significance.

EXPERIMENTAL PROCEDURES
Preparation of GST-Drm Polyclonal Antibodies-The entire coding region of rat drm cDNA was cloned in frame fused to GST sequences of the pGEX-4T1 bacterial vector (Amersham Pharmacia Biotech) by polymerase chain reaction, according to the Expand long template polymerase chain reaction system (Roche Molecular Biochemicals). The drm cDNA (1) was used as template. The region encoding the Drm protein was amplified using as 5Ј-and 3Ј-primers P1 (5Ј-CCGGAATTCATCA-ATGGCACGGCATA-3Ј) and P2 (5Ј-GTCGACTTAATCCAAGTCGATG-GA-3Ј), respectively. Amplified DNA was sequenced, digested with EcoRI/SalI, and inserted into pGEXT1 between the EcoRI and SalI sites. Recombinant plasmids were transfected into Escherichia coli DH5a and production of fusion protein was induced by 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside (14). The fusion protein was purified as described (15), and antiserum was prepared in rabbits by Agro-Bio (France) as described previously (16).
Plasmids-Constructs used in this study are shown schematically in Fig. 1, and significant regions of the Drm protein are indicated (see legend and "Results"). pMEX and pMEX-DRM have been previously described (1). To construct pHA-DRM-21N, pHA-drm was digested with XhoI and BipI, generating a large fragment containing the region encoding amino acids 40 -184 of drm, the HA tag, vector sequences, and a small fragment containing the region encoding amino acids 1-39. The large fragment was purified and ligated to the product formed by annealing the complimentary oligonucleotides C34 and C35. The sequences of those oligonucleotides are shown below with the C34 sense strand, which encodes amino acids 1 and 23-39 of rat drm (underlined). Ligation links this sequence in frame at the BipI site to the fragment encoding the remainder of the gene, generating a construct that encodes an N terminally deleted Drm protein lacking amino acids 2-22 of the native form. C34: 5Ј-TCGAGGTGACAGAATGGAAGGGAAAAAGAAA-GGGTCCCAAGGAGCCATCCCACCTCCTGACAAGGC-3Ј; C35: 5Ј-CC-ACTGTCTTACCTTCCCTTTTTCTTTCCCAGGGTTCCTCGGTAGGG-TGGAGGACTGTTCCGAGT-3Ј.
The glycosylation defective mutant pHA-GDRM was generated utilizing a polymerase chain reaction primer designed to alter the presumptive glycosylation signal 42 NDS 45 E to 42 IEA 45 E. The primers C54, 5Ј-TCCCTCGAGGTGACHGAATGAAT-3Ј (5Ј) and C55, 5Ј-TGGGGAC-TGGGTCTGCTCGGCTTCAATGTGCT-3Ј (3Ј) were used to amplify a 166-base pair fragment encoding amino acids 1-50 of rat drm, using pSVL DRM (1) as a template. The modified codons are underlined. The polymerase chain reaction product was digested with XhoI and ThIIII and used to replace the identical fragment in pSVL DRM. Single colonies were picked, and the presence of the mutation was verified by sequencing.
Cell Culture and Transfection-COS7 and HT1080 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) with high glucose supplemented with 10% fetal calf serum (FCS) (Life Technologies, Inc.). Primary rat fibroblasts were maintained in DMEM with low glucose supplemented with 10% FCS. Chinese hamster ovary cells were maintained in F12 medium supplemented with 10% FCS. Transient transfections were performed using LipofectAMINE PLUS as specified by the manufacturer (Life Technologies, Inc.).
Radiolabeling and in Vitro Translation-For metabolic labeling, cells were incubated in L-cysteine-free DMEM for 30 min. The medium was replaced with L-cysteine-free DMEM containing 2% dialyzed FCS and 200 -500 Ci/ml [ 35 S]L-cysteine (ICN, Irvin, CA), and cells were incubated for the time indicated. Asynchronously growing rat primary fibroblasts were labeled with 33 P (as 33 PO 4 ) for 5 h in phosphate-free medium containing 0.2 mCi/ml 33 P. Cells were then lysed and analyzed directly (pulse) or after incubation for different periods of time in complete culture medium in excess of cold cysteine (chase). The protein was subsequently immunoprecipitated and separated on 4 -20% SDS-PAGE (Novex). For analysis of secreted proteins, Drm-specific serum was added directly into supernatants. Following SDS-PAGE and autoradiography, bands were analyzed densitometrically. Unless otherwise indicated, the protein analysis was done in the Novex system with the use of MultiMark MW standard (Novex).
For in vitro translation assay, the EcoRI and BamHI/XhoI fragments containing the coding regions of wild type (wt) and the glycosylation mutant of rat drm, respectively, were inserted into pBKS. Plasmid DNAs were transcribed and translated using the TNT/T7 reticulocyte lysate system (Promega, Inc.) with [ 35 S]L-cysteine (Amersham Pharmacia Biotech) according to the manufacturer's instructions. To address processing in vitro, 1.5 l of canine pancreatic microsomal membranes (Promega, Inc.) were included in the reaction mixtures. Translation products were separated by SDS-PAGE and processed for fluorography.
Western Blot Analysis-Cells were lysed in boiling 2ϫ SDS sample buffer. Cellular lysates were normalized and electrophoresed on Trisglycine SDS-PAGE (Novex) and transferred to Hybond ECL nitrocellulose membrane (Amersham Pharmacia Biotech). The membranes were processed as described previously (1).
In Vitro Phosphatase Treatment- 35 S-labeled immunoprecipitates were washed three times in radioimmune precipitation buffer, resuspended in dephosphorylation buffer (50 mM Tri-HCl, pH 8.5, 0.1 mM EDTA) with 40 -60 units of calf intestinal alkaline phosphatase (Roche Molecular Biochemicals), and incubated at 37°C for 60 min. After the addition of 2ϫ SDS-loading buffer, samples were boiled and analyzed by SDS-PAGE (Novex).
Acidic Treatment-Cells grown on 100-mm dishes were washed with PBS and treated (two times with 2 ml for 5 min each) with an acidic buffer (50 mM glycine, pH 3.0, 100 mM NaCl) (25,26). Viability of the cells was controlled by staining with trypan blue. Where indicated, 4 ml of acidic wash were clarified by centrifugation 2000 ϫ g for 10 min at 4°C and precipitated by the addition of 10 volumes of ice-cold acetone.
Immunofluorescence-Cells cultured on glass coverslips were fixed for 15 min in buffered 3% formaldehyde solution and washed for 30 min in PBS. The fixed cells were permeabilized in cold acetone-methanol (1:1) for 10 min. Live cells were washed with warm serum-free medium, incubated with primary antibodies for 30 min at 37°C, then washed twice with warm medium, and incubated with secondary antibodies for 30 min at 37°C. After washing with warm medium, the cells were fixed with 4% formaldehyde in PBS for 10 min and mounted.
Immunofluorescence microscopy was performed using a Zeiss Laser Scanning Confocal Microscope (LSM 310) configured with a 25 milliwatt argon, UV coherent, and internal HeNe laser with the appropriate lines (488, 347, 543) for fluorescein isothiocyanate, Dapi, and rhodamine excitation. Confocal images were assembled as montages using Adobe Photoshop 4.0 and printed using either a Sony color video printer UP5200 MD Mavigraph, Focus Graphics 4700 with 35-mm camera film back (100 ASA Ektachrome) or a Codonics NP1600 color printer.

125
I-BMP-4 Binding, Affinity Cross-linking, and Isolation of the Cross-linked Complexes-Recombinant human BMP-4 (R & D Systems) was iodinated according to the chloramine-T method as described (18). Cells were incubated on ice for 2-3 h with 0.2-0.5 nM 125 I-BMP-4 in the presence or absence of unlabeled ligands in binding buffer (PBS containing 0.9 mM CaCl 2 , 0.49 mM MgCl 2 , and 1 mg/ml bovine serum albumin) and cross-linked using 1 mM bis[sulfosuccinimidyl] suberate and disuccinimidyl suberate (Pierce) according to the protocol described by Nishitoh et al. (19). Cross-linked material was immunoprecipitated with specific antisera and analyzed by PAGE.
Osteogenic Differentiation Assay-C2C12 (ATCC) cells were cultured in DMEM (H-G) supplemented with 20% FCS (Life Technologies, Inc.). The osteogenic differentiation assay was performed according to Katagiri et al. (20). C2C12 was plated in 24-well plates (Corning) at 60% confluence. After a 24-h incubation, the medium was changed to DMEM or supernatant from the cells with 5% FCS containing the indicated concentrations of BMP-4. After 24 h of incubation, alkaline phosphatase activity of the cells was assayed using a colorimetric kit (Sigma). Cells were washed in PBS and lysed by sonication in 300 l of 2 mM MgCl 2 , 0.2% Nonidet P-40. Substrate solution (200 l) was added and incubated at 37°C for 30 min. After stopping with 0.1 N NaOH, absorbance at 405 nM was measured and compared with a standard curve of p-nitrophenol (20).

Drm Undergoes
Post-translational Modification-To characterize the Drm gene product, we immunized rabbits against purified GST-Drm fusion protein and assessed the ability of immune sera to specifically recognize Drm proteins. In Western blots the sera detected two bands with apparent mobilities of 22 and 26 kDa in COS cells transfected with pMEX-DRM ( Fig. 2A, lane 2). HA-tagged Drm migrated with a slightly reduced mobility (Fig. 2, lane 3), consistent with the presence of the tag, and these bands were also detected using anti-HA monoclonal antibody (data not shown). Major bands migrating at about 22 and 26 kDa, and an intermediate band migrating slightly above the 22-kDa band (Fig. 2B, lane 1), were also detected in normal primary rat fibroblasts that express Drm mRNA (1). However, we failed to detect Drm proteins in v-mostransformed DTM cells, in agreement with our results that Drm mRNA is down-regulated in these cells (1).
The presence of multiple protein bands in cells expressing a drm cDNA suggested that Drm undergoes post-translational modifications, in agreement with the presence of a signal peptide cleavage site (21), a site for N-glycosylation, and several putative phosphorylation sites in the expected product. To confirm that these signals were functional both in vitro and in vivo, we initially analyzed the in vitro translation products of wt drm and a mutant construct pHA-GDRM (Fig. 1), in which the presumptive 42-NDSE-45 glycosylation site was changed to 42-IEAE-45 (Fig. 1). As seen in Fig. 2C, both translation products migrated with an apparent molecular mass of 23 kDa. To examine post-translational modifications, we also performed a second translation reaction that included a preparation of canine pancreatic microsomal membranes, which allows signal peptide cleavage and core glycosylation to take place. Under these conditions, translation of wt drm yielded two additional products, which migrated with apparent molecular masses of 24 and 20 kDa (Fig. 2C, lane 2). In contrast, the 24-kDa product was no longer detectable following translation of the pKA-GDRM mutant (Fig. 2C, lane 4). The size of the lower 20-kDa form is compatible with the product generated by removal of the predicted signal peptide sequence, whereas the 24-kDa band is likely to arise through glycosylation.
To verify that glycosylation also takes place in vivo, we transfected Cos cells with wt drm or the pKA-GDRM mutant and analyzed protein expression in control cells and cells treated with tunicamycin, which prevents cotranslational Nglycosylation (22). Following tunicamycin treatment the slower migrating band was no longer detectable in cells transfected with wt drm (Fig. 3, A, lanes 3 and 4; and B, lanes 1 and 2). Similarly, this band was not visible in normal rat fibroblasts treated with tunicamycin (Fig. 3B, lanes 3 and 4). A similar effect was seen (Fig. 3A, lanes 1 and 2) if cell extracts were treated with the deglycosylating enzyme peptide N-glycosidase F (17). These results indicated that the more slowly migrating form of the protein (26 kDa) observed in vivo represented the glycosylated form of Drm and that the faster migrating (22 kDa) form was generated by signal peptide cleavage. This possibility was verified by analyzing Cos cells transfected with the pHA-DRM-21N mutant, which encodes a protein lacking 21 of the presumptive 24-amino acid signal peptide. These cells exhibited only the faster migrating band (Fig. 3C, lane 3), and the mobility of this band was not affected by treatment with tunicamycin (Fig. 3C, lane 4). Finally, we could not detect the 26-kDa band in cells transfected with the pHA-GDRM mutant lacking the glycosylation site (Fig. 3D, lane 2), in agreement with results of in vitro translation. Analysis of the predicted Drm sequence also showed the presence of potential phosphorylation sites by various kinases, such as protein kinase C, cAMP-and GMP-dependent kinases, and casein kinase II (Fig.  1). We examined in vivo phosphorylation by metabolically labeling primary rat cells with [ 35 S]L-cysteine, immunoprecipitating the Drm protein, and treating the immunocomplexes with calf intestine alkaline phosphatase before electrophoretic resolution. We observed an overall increased mobility in phosphatase-treated samples (Fig. 3E, lanes 1 and 2), consistent with the removal of phosphate residues from all Drm forms. Direct confirmation of in vivo phosphorylation was obtained by metabolic labeling with [ 33 P]orthophosphate and immunoprecipitation of Drm. All detectable forms of Drm appeared to be labeled (Fig. 3F), consistent with the results observed following phosphatase treatment. Thus, our analysis indicates that the multiple forms of Drm protein detected in vivo probably result from a combination of signal peptide cleavage, glycosylation, and phosphorylation.
Drm Secretion-The presence of a putative N-terminal secretory signal sequence (1, 11), together with previous reports (11), strongly suggested that Drm was a secreted protein.
Therefore, to further analyze the distribution of inter-and extracellular Drm, we performed indirect immunofluorescence microscopy, as shown in Fig. 4. In Fig. 4, A and G, fixed, permeabilized Cos-7 cells transiently transfected with HA-DRM exhibited a diffuse, fiber-like network of staining suggestive of a localization in the endoplasmic reticulum-Golgi complex. Some cells also exhibited a distinct perinuclear staining, which may indicate that this is the site of Drm synthesis (Fig.  4, A and G). This pattern is analogous to that observed for proteins known to be located in the ER, such as calnexin (23), and we confirmed this localization using monoclonal antibodies directed against the Golgi-specific p58K protein (Fig. 4F), a marker of the ER-Golgi intermediate compartment (24). Our results showed that both Drm (Fig. 4G) and p58K (Fig. 4F) colocalized in the Golgi stacks (Fig. 4H).
In contrast to the permeabilized cells, nonpermeabilized cells showed a clumped, punctate pattern that appeared to surround the outer surface of the cell membrane (Fig. 4B), indicating the presence of Drm on the external cell surface. Analysis of live, unfixed cells showed a similar pattern (Fig. 4D). We observed a similar subcellular distribution of Drm in COS cells by using anti-HA antibodies and in rat cells expressing the endogenous protein (data not shown).
These data indicate that Drm is accumulated in the ER and after release from the ER transverses the Golgi apparatus to the outer surface of the cell. To confirm that the hydrophobic region was necessary for the entrance of Drm into the secretory pathway, we transfected Cos-7 cells with pHA-DRM-21N, which lacked the signal sequence. The truncated protein was exclusively intranuclear (Fig. 4C), in agreement with the fact that the protein contains two nuclear localization signals (amino acids 147-150 and 168 -171). As expected, we did not observe surface staining in these live or nonpermeabilized cells (data not shown).
Bands corresponding to glycosylated and unglycosylated Drm, as well as minor bands consistent with phosphorylated forms, could be detected in immunoprecipitates from the supernatants of 35 S-labeled Drm expressing cells using Drmspecific antibodies (Fig. 5A, lane 1), confirming that DRM could be secreted into the medium. Preimmune serum failed to detect DRM-specific bands (Fig. 5A, lanes 2 and 4). The mobilities of the secreted and cell-associated forms (Fig. 5A, lanes 1 and 3) appeared to be identical, suggesting all forms of Drm were  1, 5, and 6) and cellular lysates (lane 3) were incubated with Drmspecific antibodies, and the precipitates were analyzed in 4 -20% SDS-PAGE. Preimmune serum was used as a control (lanes 2 and 4). B, turnover of secreted and intracellular forms of Drm. HT1080 cells stably transfected with pMEX-DRM were pulsed with [ 35 S]L-cysteine for 30 min (lane 1) and then chased for the indicated times (lanes 2-8). Drm protein was immunoprecipitated with anti-Drm antibodies from supernatant and cellular lysates and analyzed in 4 -20% SDS-PAGE. released into the supernatant. Interestingly, cells expressing Drm containing a mutation that destroyed the glycosylation site still secreted the protein. The supernatant form exhibited a mobility consistent with a cleaved nonglycosylated form of tagged Drm (Fig. 5A, lane 6 compared with lane 5), indicating that glycosylation was not necessary for the processing and release of Drm from expressing cells.
To elucidate the kinetics of Drm synthesis, release, and degradation, we performed pulse-chase experiments using HT1080 cells transfected with a drm expression construct and stably expressing high levels of exogeneous Drm. As shown in Fig. 5B, lane 1, both the glycosylated and nonglycosylated forms could be detected in the cell lysates after a 30-min pulse. The intensity of these cell-associated bands decreased during the first 30-min chase period and could not be detected after 6 h. Densitometer analysis of this and other similar experiments suggested that the half-life of cell-associated Drm was between 45-60 min. Both glycosylated and nonglycosylated forms were lost at equivalent rates, indicating that glycosylation did not influence protein stability.
We also observed a consistent mobility shift of all Drm bands during the chase (Fig. 5B, compare lane 1 with lanes 2 and 3), suggesting that phosphorylation might be involved in degradation. To confirm that the shift involved phosphorylation, we treated cell extracts after a 30-min pulse and after a 2.5-h chase period with alkaline phosphatase. All shifted Drm bands exhibited increased mobility following alkaline phosphatase treatment, indicating that the shift seen during the chase period was likely to be because of phosphorylation (data not shown).
In contrast to the rapid appearance of cell-associated Drm, we were able to detect Drm in the medium only after a 30-min chase (Fig. 5B, lane 2). The level in the supernatant increased during the first 2 h of the chase and then rapidly decreased over the next 4 h (Fig. 5B, lanes 3-7). In the experiment shown in Fig. 5B, the Drm band represents the immunoprecipitate of the total labeled cell extract and supernatant from a single 60-mm dish. Comparison of the amount of secreted Drm to the amount found in cells suggests that only a small amount is actually released in a soluble form and that the majority of the protein remains cell-associated.
Cell-associated Drm Binds BMP-Soluble secreted, Myctagged Gremlin/Drm, overexpressed in COS cells, was shown to block the activity of purified BMP-2 (11). Because our data indicated that Drm was present both on the external cell surface and in the culture supernatant, we examined the function of both forms in a biological assay, measuring the ability of Drm to interfere with the osteogenic differentiation of C2C12 mouse myoblastic cells induced by BMP-4 (20). In our assays C2C12 cells cultured in the presence of BMP-4 expressed high levels of alkaline phosphatase (AP) activity within 24 h, and this induction was not affected when BMP-4 was preincubated with culture medium from control HT1080 cells or of cells transfected with vector (Fig. 6A, vector alone). In contrast, preincubation with culture supernatants from cells expressing DRM, HA-DRM, or HA-GDRM proteins blocked AP induction in C2C12 cells over a range of BMP from 1 to 5 nM (Fig. 6A). The supernatant from cells expressing Drm was still active when diluted 4 -10-fold (Fig. 6B), consistent with the fact that a 5-fold increase in BMP concentration only restored about 50% of the maximal activity in the absence of Drm (Fig. 6A). Culture supernatants from rat primary fibroblasts expressing endogenous Drm could also antagonize BMP-4, demonstrating that endogenous Drm is also released from fibroblasts in a biologically active form. However, as shown in Fig. 6A, the inhibitory activity of this supernatant was much weaker than that of supernatants from Drm expressing HT1080 cells, which may be related to the total level of Drm protein detected in these cells (Fig. 6A, inset). Cocultivation of C2C12 cells with cells expressing different forms of Drm also interferes with the osteogenic differentiation of mouse myoblastic cells (data not shown). The results shown in Fig. 6A also demonstrate that glycosylation has no apparent effect on the ability of Drm to block BMP-4 activity. Thus, neither secretion nor BMP antagonism appears to be affected by the failure of the mutant GDRM protein to be glycosylated.
Although soluble forms of Drm interact with BMP and interfere with its activity (Fig. 6 and Ref. 11), a large fraction of secreted Drm appears to remain cell associated (see above), and this form, if biologically active, could be of functional significance. We therefore investigated whether the cell-bound Drm was able to bind BMP directly. We incubated 125 I-BMP-4 with HT1080 cells expressing Drm or vector alone, then extensively washed the cells, lysed them, and treated the lysates with anti-Drm serum. We found that 125 I-BMP was present in the anti-Drm immunoprecipitate (Fig. 7, lanes 1, 2, and 4) but not when preimmune serum was used (Fig. 7A, lane 3) or when vector-transfected cells were analyzed (Fig. 7A, lane 7). Binding of labeled BMP could be competed with an excess of cold BMP (Fig. 7B), indicating that the Drm-BMP interaction was specific. Furthermore, when cells exposed to 125 I-BMP were treated with a bifunctional cross-linker (bis[sulfosuccinimidyl] suberate) before lysis, Drm precipitates contained two labeled bands, one of about 20 kDa consistent with the size of free BMP and a second 44-kDa band of the size expected for a Drm-BMP complex (Fig. 7A, lane 1). The relative intensity of the 44-kDa band was reduced when cold BMP was added to the cells prior to washing and cross-linking (Fig. 7A, lane 2), further supporting the hypothesis that it represented a cross-linked BMP-Drm complex. These binding studies indicate that cell surface-associated as well as secreted Drm (11) are able to specifically bind to BMP-4. DISCUSSION We previously described the isolation of the rat drm gene (1), also known as Gremlin (11), and its expression pattern in tissues and cell lines. Drm/Gremlin is a member of a family of related BMP antagonists (1,11,26) whose critical functions in embryo patterning (11) and limb development (27,28) are becoming increasingly clear. Here we have reported the first biochemical characterization of the Drm/Gremlin gene product, showing that it is a glycosylated, phosphorylated, and secreted protein. Following secretion, we show that Drm protein is detected at the cell surface and in the extracellular medium and that both forms bind BMP-4. Our data also demonstrate that glycosylated and nonglycosylated forms of Drm can antagonize its ability to induce osteogenic differentiation of mouse C2C12 cells.
The drm gene was initially identified in phenotypically normal fibroblasts and was found to be repressed when these cells were transformed by oncogenes, which constitutively activate the extracellular signal-regulated kinase pathway (1). Downregulation of Drm expression is seen in a wide range of human cancer cell lines, 3 and we also have reported expression of Drm in differentiated cells from various adult tissues (1). 3 These observations underlined the importance of Drm expression in adult tissues and raised the possibility that loss of Drm expression may be involved in regulating the emergence or maintenance of the transformed phenotype.
The Drm/Gremlin amino acid sequence contains multiple sites for potential post-translational modifications, and consistent with these structures, we have shown that the two major forms of Drm/Gremlin detected in primary rat fibroblasts and cells overexpressing drm cDNA are generated through protein cleavage and glycosylation. Tunicamycin and peptide N-glycosidase F treatments demonstrate that the two bands represent glycosylated and nonglycosylated protein, respectively, and this is further supported by the absence of the higher molecular weight band in cells transfected with a drm construct containing a mutant glycosylation site. Moreover, the smaller form of the protein comigrates with the single band of Drm protein detected in cells overexpressing the HA-DRM-21N mutant (which lacks the first 21 amino acids of the presumptive signal peptide), indicating that all Drm forms detected in vitro are cleaved. Similar forms have been seen in cells from various species and tissues. 4 In agreement with previous reports (11), this study shows that secreted Drm antagonizes the ability of BMP-4 to induce osteogenic differentiation of C2C12 mouse myoblasts, placing drm within the family of BMP antagonists that includes DAN, cerberus, noggin, and chordin. These proteins comprise an evolutionarily conserved, functionally related gene family (26), and although, as in the developing limb bud, they are frequently expressed in the same tissues and developing structures, their expression pattern suggests that they act in a FIG. 7. Cell surface-associated Drm binds directly to BMP-4. A, HT1080 cells expressing Drm or vector alone were incubated with 125 I-BMP-4 in the presence (lanes 2 and 6) or absence (lanes 1, 3,  4, 5, and 7) of cold BMP, washed, and cross-linked before immunoprecipitation (lanes 1, 2, 3, 5, and 6) or left without cross-linking (lanes 4 and 7). Cellular lysates were immunoprecipitated with Drm-specific antibodies (lanes 1, 2, 4 complementary rather than a redundant fashion (27,28). Thus, it is likely that the various family members exert specific functions and might be expected to express unique properties.
Consistent with this, we observed that Drm/Gremlin exhibited several potentially significant differences in comparison to other related BMP antagonists. We found that, in contrast to DAN and Cerberus, two other secreted members of this protein family, only a small fraction of Drm is glycosylated. Preventing glycosylation did not alter its secretion or its ability to antagonize BMP signaling. Thus, glycosylation does not appear to be required for the interaction of Drm with BMP-4, although the functional role of Drm glycosylation remains unclear.
We also show that Drm is a phosphoprotein by in vivo labeling with [ 33 P]orthophosphate, as well as by demonstrating changes in protein electrophoretic mobility after treatment with phosphatase. Phosphorylation may be unique to Drm, because none of the other members of the Drm/DAN family are known to be phosphorylated. Interestingly, besides potential sites for phosphorylation by kinases such as casein kinase II, cAMP-dependent, and protein kinase C, Drm/Gremlin also contains two potential targets for serine kinases that phosphorylate secreted proteins (29). The Ser x Glu-Ser motif is repeated twice in Drm at positions Ser 77 GlnGlu 79 and Ser 140 CysSer 142 but are not conserved within the cysteine repeat domains of the other family members.
Immunofluorescent microscopy indicated that intracellular Drm is localized in the ER-Golgi intermediate compartment. It appears to enter the ER upon synthesis, where it is posttranslationally modified and moved through the secretory pathway. The half-life of cell-associated Drm is relatively short (less than 1 h), and the lysosomal and proteosomal pathways appear to be involved in Drm degradation. 5 The phosphorylation of Drm that we have detected could indicate its involvement in targeting the protein for ubiquitination and subsequent degradation via the proteosomal pathway (30).
Consistent with previous reports (11), we found that Drm is released into the extracellular medium, although a significant fraction of secreted Drm appeared to remain noncovalently bound to the outer cell surface. Besides immunofluorescent data, this localization of the protein was supported by experiments showing treatment of Drm-expressing cells with trypsin or acidic buffer (25) significantly reduced the amount of cellassociated Drm. 6 This association with the external cell surface could reflect an intermediate stage before release in a fully soluble form, perhaps in response to appropriate external or internal signal. Alternatively, the fact that only a small amount of the protein is detectable outside the cells could suggest that cell association, which would allow Drm to regulate signaling at the cell surface in cis, is functionally important. Based on co-immunoprecipitation and cross-linking experiments, our results indicate that Drm can bind BMP when exposed on the cell surface. This raises the possibility that Drm/Gremlin could exert antagonistic effects on BMP-mediated signaling both in cis and in trans. Cocultivation experiments suggested that cell surface-associated forms of DRM could also antagonize BMP-4, although we could not rule out the possibility that this inhibition is because of a combination of biological activities of the secreted and cell surface-associated forms.
Understanding the structure and mechanisms of action of BMP antagonists such as Drm/Gremlin is important, because these molecules have been shown to play critical roles in several BMP-mediated functions. In addition to its initially re-ported role in early embryonic patterning (11), Drm/Gremlin has been reported to control the Sonic Hedgehog/FGF4 feedback loop in murine limb buds (27). Similarly, it has been shown to perform multiple functions in avian limb bud development, including the inhibition of BMP-induced apoptosis during limb outgrowth (28). However, BMP molecules are also known to be involved in an increasingly broad range of tissuespecific functions, including those in the brain (31)(32)(33) and ovaries (34) and to effect the properties of neuroectodermal tumors (35). Because expression of Drm is high in the brain and other differentiated adult tissues (1), 3 the function of this protein is likely to be important outside embryonic development. We have reported that Drm is down-regulated in transformed cells (1), 3 and, interestingly, the related DAN gene has been characterized as a tumor suppressor (4). Furthermore, a recent report showed that Drm is up-regulated by the homeobox gsh-1 gene (36), consistent with an involvement of Drm in the negative control of cell proliferation.
Drm/Gremlin shares structural and functional properties with other members of the Drm/Gremlin/DAN family of BMP antagonists that are consistent with its important, defined role as a regulator of BMP action in embryonic development. Our results describe the basic mechanisms of synthesis and processing of Drm/Gremlin, adding to our understanding of this important class of proteins. In addition, our results show that Drm/Gremlin has unique features that could suggest specific roles and functional mechanisms for this protein in embryonic as well as adult tissues.