A Novel Family of Cell Surface Receptors with Tyrosine Kinase-like Domain*

Human cDNA clones encoding two novel proteins with a region strongly homologous to the tyrosine kinase domain of growth factor receptors, in particular of the Trk family, were obtained by a polymerase chain reaction-based approach. These proteins, Ror 1 and Ror2, share 58% overall amino acid identity and a structure indicative of cell surface molecules. A secretion signal sequence and a transmembrane domain delimit the extracellular portion, which contains immunoglobulin-like, cysteine-rich, and kringle domains. The cytoplasmic portion contains the tyrosine kinaselike domain which (in RorZ) appears to be associated with protein kinase  activity in vitro, followed by serinelthreonineand proline-rich motifs. Partial nucleotide sequences of the rat genes reveal striking evolutionary conservation of the proteins between human and rat. The level of expression of the rat genes is high in the head and body of early embryo and decreases dramatically after embryonic day 16. Based on these data, Rorl and Ror2 appear to define a new developmentally regulated family of cell surface receptors for unidentified ligands.


From Regeneron Pharmaceuticals, Tarrytown, New York 10591-6707
Human cDNA clones encoding two novel proteins with a region strongly homologous to the tyrosine kinase domain of growth factor receptors, in particular of the Trk family, were obtained by a polymerase chain reaction-based approach. These proteins, Ror 1 and Ror2, share 58% overall amino acid identity and a structure indicative of cell surface molecules. A secretion signal sequence and a transmembrane domain delimit the extracellular portion, which contains immunoglobulin-like, cysteine-rich, and kringle domains. The cytoplasmic portion contains the tyrosine kinaselike domain which (in RorZ) appears to be associated with protein kinase activity in vitro, followed by serinelthreonine-and proline-rich motifs. Partial nucleotide sequences of the rat genes reveal striking evolutionary conservation of the proteins between human and rat. The level of expression of the rat genes is high in the head and body of early embryo and decreases dramatically after embryonic day 16. Based on these data, Rorl and Ror2 appear to define a new developmentally regulated family of cell surface receptors for unidentified ligands.
Many intercellular signals which control eukaryotic cell growth and differentiation are mediated through polypeptide growth factors. Transmembrane protein tyrosine kinases which serve as receptors to these ligands constitute a large class of proteins sharing a characteristic modular design (1). The extracellular, ligand-binding portion of the kinase receptors contains an assortment of immunoglobulin-like, fibronectin-like, and cysteine-rich domains. The interactions of this region with soluble ligands has been extensively studied, but there is also growing evidence that membrane-bound growth factors (2), or large transmembrane proteins (3) may interact with, and activate, receptor-type tyrosine kinases.
The cytoplasmic portion of the receptors contains the catalytic kinase domain, positioned to directly interact with intracellular components. Interactions of this domain with several cytoplasmic proteins have been characterized at the molecular level (reviewed in Refs. 4,5). The phosphotyrosine residues on the receptor, products of the ligand-induced au-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in thispaper has been submitted to the GenBankTM/EMBL Data Bunk with accession number(s) M97639andM97675.
$ To whom correspondence should be addressed Regeneron Pharmaceuticals, 777 Old Saw Mill River Rd., Tarrytown, NY 10591-6707. Tel.: 914-345-7412;Fax: 914-347-5045. tophosphorylation, are recognized by the SH2l domains of phospholipase C -7 , p2lras GTPase-activating protein, and Src-like tyrosine kinases. Proteins which do not contain SH2 domains, such as Raf serinelthreonine kinase and phosphatidylinositol 3'-kinase, may interact with the receptors through the SH2-containing "adaptor proteins"; p85 subunit of phosphatidylinositol 3'-kinase appears to be an example of such a molecule. Phosphorylation of the target molecules triggers second messenger cascades which eventually reach the nucleus to modify the cell's genetic program. Moreover, recent evidence for receptor-mediated phosphorylation of the uau proto-oncogene product, a transcription factor-like molecule with the SH2 domain, suggested direct links between receptor kinase activation and transcriptional regulation (6,

7).
The emerging understanding of the basic mechanisms of signal transduction by kinase receptors is derived from work in relatively few well defined model systems, including the receptors for epidermal growth factor, platelet-derived growth factor, insulin, and a few others. However, the exponential rate of novel receptor discovery (8) suggests the existence of many still unknown receptors, and variations of signaling mechanisms. For example, recent data demonstrated that cell surface receptors for some polypeptide growth factors may be associated with serine/threonine, rather than tyrosine kinase activity (9)(10)(11)(12). Thus, the full appreciation of the role of these molecules in normal eukaryotic development and in malignancy will require identification and characterization of diverse receptors expressed in different tissues at different developmental stages.
Recent demonstration that members of the Trk family of transmembrane tyrosine kinases are receptors for nerve growth factor and related neurotrophic factors (13-1.8) reemphasized the role of receptor kinases in the control of development of the nervous system. Our interest in factors which regulate the development and regeneration of the nervous system prompted us to search for additional receptors whose similarity to Trk proteins in the tyrosine kinase domain may indicate their interactions with a similar range of cellular targets. Eventually, such molecules may allow identification of additional ligands and mechanisms controlling neuronal development.
We describe here two putative cell surface receptors which contain in their cytoplasmic portion a domain with strong amino acid homology to the tyrosine kinase domain of growth factor receptors. The overall similarity within this region identifies Rorl and Ror2 as close relatives of the Trk family. However, Rorl and Ror2 share several unique features, suggesting that they are members of a separate family of recep-AG)ATGGT(TCAG)TA(TC)(TC)T and CATGTCTAGAGGCATC-CA(TCAG)C(GT)(AGT)AT(TCAG)GG, respectively, with parentheses denoting degeneracy), were used to amplify fragments of RNA sequences with the GeneAmp RNA PCR kit (Perkin-Elmer Cetus). The mRNA from the human neuroblastoma cell line SH-SY5Y was a generous gift of Dr. Sam Davis (Regeneron). The PCR products were phenol/chloroform extracted and treated with restriction enzymes (Boehringer Mannheim) which had recognition sites in the corresponding fragments of trk and trkB genes. DNA resistant to digestion was cut from 1% NuSieve, 1% SeaPlaque agarose (FMC BioProducts) gels and reamplified by PCR after melting the agarose at 65 'C. The final PCR products were digested with XhoI and XbaI (whose sites were contained in 10 nucleotides at the 5' ends of TRK-9 and TRK-lor, respectively) and cloned into pBluescript I1 KS(-) phagemids (Stratagene). Bacterial colonies were screened by PCR using vector-specific primers. Direct sequencing of the resulting DNA fragments with FASTaq kit (IBI), and their labeling for library screening were performed as previously described (19). The SH-SY5Y cDNA library was constructed in the EcoRI site of Lambda ZAP I1 vector, using Stratagene kits. Plaques identified by hybridization with radioactively labeled PCR fragments were processed as recommended (Stratagene), to obtain the cDNA insert-containing phagemids. The longest cDNA inserts were sequenced on both strands with specific oligodeoxyribonucleotide primers, by a combination of automated sequencing using model 373A Sequencing System (Applied Biosystems) and manual sequencing using Sequenase Version 2.0 (U. S. Biochemicals, Cleveland, OH). Rat sequences were obtained by PCRamplifying fragments of RNA from rat embryos with GeneAmp RNA PCR kit and the human-specific primers used for sequencing. The resulting cDNA fragments were sequenced both manually with FAS-Taq and automatically.
Expression of Myc-tagged Kinase Domains in COS Cells-The recombinant genes, encoding the tyrosine kinase-like domains of Rors with 10 amino acids from the human c-myc proto-oncogene added at the C terminus (rorl-TK-myc and ror2-TK-myc), were prepared by "patch P C R as described earlier (21). The 5'-PCR primers (sense strand) consisted of tails containing XhoI site, and 18 nucleotides encoding the N termini of the recombinant proteins, corresponding to Met453 through Ala458 of Rorl or Met452 through Ile457 of Ror2. The patch PCR primers (antisense strand) encoded the C-terminal Ror sequences, corresponding to Val778 through AmTs3 of either Rorl or Ror2, a Gly-Gly bridge, c-Myc amino acid sequence EQKLISEEDL, and a termination codon. Finally, the 3'-PCR primer consisted of 19 nucleotides identical to the 5' end of the patch PCR primers, and a tail containing NotI site. The 3'-PCR primer (5"GCTAATGC-GGCCGCTACAGATCCTCCTCAGAA-37, and the Rorl-specific 5' and patch PCR primers (5'-CATTGCCTCGAGGATGGCCATGA-CAATGCTGAATGCA-3' and 5'-CTACAGATCCTCCTCAGAAA-TCAGCTTTTGCTCACCACCGTTACTGAGATTACTCAC-3', respectively) or Ror2-specific primers (5"CATAGCCTCGAGGAT-GGCCATGGAAATGCCCCTCATT-3' and 5'-CTACAGATCCTC-CTCAC-3') were used, in combination with the appropriate ror cDNA as a template, to amplify the DNA fragments encoding the synthetic rorl-TK-myc and ror2-TK-myc genes. The PCR products were di-
Immune Complex Kinase Assay-COS cells expressing Rorl-TK-Myc and Ror2-TK-Myc proteins were rinsed with cold phosphatebuffered saline containing 1 mM phenylmethylsulfonyl fluoride, 0.1 mM orthovanadate, and 0.14 unit/ml aprotinin, and scraped in 2.5 ml of this buffer/l4-cm plate (about 2.5 X lo6 cells). Cells were homogenized in Dounce tissue grinder and cell debris removed by centrifugation for 30 min at 11,000 X g. Aliquots of the supernatant (0.5 ml) were stored frozen at -80 "C. Proteins were immunoprecipitated with 9E10 antibody and 50 pl of GammaBind G Sepharose (Pharmacia LKB Biotechnology Inc.) per sample. Immunoprecipitates were washed three times with each of the following buffers: phosphatebuffered saline, high salt wash (50 mM Tris-HC1, pH 7.5,0.5 M NaCl, 5 mM EDTA, and 0.1% Nonidet P-40), and kinase buffer (20 mM HEPES, pH 7.5, 10 mM MnC12, and 10 p~ orthovanadate), and resuspended in 10 pl of kinase buffer containing 10 pCi of [r-"P] ATP (6000 Ci/mmol, Du Pont-New England Nuclear). The reaction was allowed to proceed for 30 min at room temperature and was terminated by several washes with kinase buffer containing 0.1% Nonidet P-40, and boiling in 2 X SDS sample buffer (with 2-mercaptoethanol). The products were analyzed by polyacrylamide gel electrophoresis and electroblotting, followed by autoradiography and immunodetection.
RNA Analysis-Total RNA samples were prepared (23) from heads and bodies (or, at later stages of development, livers) of rat embryos.
formaldehyde, blotted to nylon filters (MSI), and hybridized with lo6 10-pg aliquots were separated on 1% agarose gels in the presence of dpm/ml of 32P-labeled probes (24), followed by autoradiography.

RESULTS
Cloning of rorl and ror2 cDNAs-To amplify by PCR fragments of trk-related genes, we designed degenerate oligodeoxyribonucleotide primers corresponding to protein regions surrounding amino acids conserved in tyrosine kinases. We chose regions which contained amino acids shared between Trk and TrkB proteins, but not with other tyrosine kinases. In particular, TRK-9 and TRK-lor primers included all possible codon variations corresponding to amino acid sequences Ala-Gly-Met-Val-Tyr-Leu (in the sense orientation), and Pro-Ile-Arg-Trp-Met-Pro (in the antisense orientation), respectively. These sequences are identical in Trk and TrkB, but each contains a t least 2 amino acid substitutions even in insulin receptor, which shows the highest homology to Trks in the tyrosine kinase domain (25,26).
Products of PCRs were treated with restriction enzymes expected to cut trk and trkB sequences. Enzyme-resistant fragments were reamplified and cloned using the restriction sites built into the primers. The resulting bacterial colonies were screened by PCR followed by direct sequencing. Two DNA sequences (termed rorl and ror2), amplifiedusing cDNA from human neuroblastoma cell line SH-SY5Y as a template and TRK-9 and TRK-10r primers, and selected by their resistance to digestion with EcoR V, showed similarity among themselves and did not correspond to any published sequences. T o clone cDNAs corresponding to these PCR fragments, we prepared a cDNA library from SH-SY5Y mRNA and screened it with radioactively labeled PCR fragments.
Screening of 1.5 X lo6 clones yielded three different positive clones for rorl and two clones for ror2. The longest cDNA inserts were then sequenced. The rorl insert is 3358 bp long (excluding the poly(A) tail), and the ror2 insert is 4094 bp long (Fig. 1). These cDNA sequences contain open reading frames encoding proteins of 937 (Rorl) and 943 (Ror2) amino acids. ror2 cDNA has a 1066-bp 3"untranslated region with a polyadenylation signal AATAAA located 15 bp before the poly(A) tail, while rorl has only 172-bp 3"untranslated region without a polyadenylation signal. The 5"untranslated region    Ser Pro V a l Ser A s n V a l Ser A s n A l a A r g T y r V a l G l y Pro L y s G l n L y s A l a Pro Pro Phe Pro G l n Pro G l n

TTC ATC CCC ATG AAG GGC CAG ATC AGA CCC ATG GTG CCC CCG CCG CAG CTC TAC GTC CCC GTC AAC GGC TAC CAG 2674
Phe Ile Pro M e t L y s G l y G l n I l e A r g Pro M e t V a l Pro Pro Pro G l n L e u T y r V a l Pro V a l A s n G l y T y r G l n

CCG GTG CCG GCC TAT GGG GCC TAC CTG CCC AAC TTC TAC CCG GTG CAG ATC CCA ATG CAG ATG GCC CCG CAG CAG 2749
Pro V a l Pro A l a T y r G l y A l a T y r L e u Pro A s n Phe T y r Pro V a l G l n I l e Pro M e t G l n M e t A l a Pro G l n G l n

ACG GCC CCC TCC AAC ACA TCC ATG GCA GAC AGG GCA GCC CTG CTC TCA GAG GGC GCT GAT GAC ACA CAG AAC GCC 2899
T h r A l a Pro Ser A s n T h r Ser M e t A l a A s p A r g A l a A l a L e u L e u Ser G l u G l y A l a A s p A s p T h r G l n A s n A l a

900
CCA GAA GAT GGG GCC CAG AGC ACC GTG CAG GAA GCA GAG GAG GAG GAG GAA GGC TCT GTC CCA GAG ACT GAG CTG 2974 Pro G l u A s p G l y A l a G l n Ser T h r V a l G l n G l u A l a G l u G l u G l u G l u G l u G l y Ser V a l Pro G l u T h r G l u L e u

925
CTG GGG GAC TGT GAC ACT CTG CAG GTG GAC GAG GCC CAA GTC CAG CTG GAA GCT TGAGTGGCACCAGGGCCCGGGGTTCGGG 3056 L e u G l y A s p C y s A s p T h r L e u G l n V a l A s p G l u A l a G l n V a l G l n L e u G l u A l a

FIG. 2. Expression of Rorl and RorZ proteins in COS cells.
Samples of cell lysates from mock-transfected COS cells (01, or COS cells transfected with rorl (I) or ror2 ( 2 ) cDNAs were separated by gel electrophoresis and immunoblotted, as described under "Experimental Procedures." The blot on the left was incubated with anti-Rorl antibody RG81, and the blot on the right with anti-Ror2 antibody RG73. The molecular mass of markers (M), in kilodaltons, is indicated.
is 375-bp long in rorl and 199 bp in ror2 cDNA; in rorl this region contains two termination codons in the same frame as the putative protein. The putative initiation codon is preceded in each clone by a GC-rich region (80% G-C), which is typical of many growth factors and their receptors, and implies an additional level of post-transcriptional control (27). Experimental support for our assignment of the reading frames was obtained by specific detection of proteins in lysates of COS cells transfected with rorl and ror2 cDNAs, using anti-peptide antibodies (Fig. 2). The observed molecular mass of these proteins, 120-130 kDa, is larger then the mass calculated from the amino acid composition of the predicted mature peptides (102 kDa) and may indicate post-transcriptional modification at the potential N-glycosylation sites (see below). Unique Structure of Rorl and Ror2 Proteins-The deduced Rorl and Ror2 proteins, aligned in Fig. 3, are highly homologous and overall show 58% amino acid identity. Immediately following the putative NHz-terminal methionine they both have a potential secretion signal sequence (30); the aminoterminal charged region (31) of the signal sequence contains 5 arginines in Rorl and 3 in Ror2. In the center of each amino acid sequence there is a hydrophobic region which presumably comprises the transmembrane domain. In the extracellular portion of Rors flanked by these two regions, we have identified an immunoglobulin-like domain similar to that present in the FGF receptors (28), a cysteine-rich region, and a kringle domain. This last motif has not been reported for receptor proteins but is present in extracellular proteins involved in blood coagulation (32) and in HGF (29). Furthermore, Rorl has in its extracellular portion four potential N-glycosylation sites, three of which are shared with Ror2. The cytoplasmic part of Rors contains a region very similar to the catalytic domain of tyrosine kinase receptors. Amino acid sequence homology within this domain has a predictive value in the determination of the properties and function of novel protein kinases, since catalytic domains from functionally related kinases cluster together within the phylogenetic tree (25). The extent of similarity between Rors and other protein kinases, schematically represented in Fig. 4, suggests that Rorl and Ror2 constitute a separate family of tyrosine kinases, most closely related to the Trk family of receptors. When GenBank (release 72.0) and EMBL (release 31.0) databases were searched with Rors using TFASTA algorithm (34), the highest similarity scores were assigned to Trk and TrkB sequences (data not shown). For example, kinase-like domains of Rorl and Trk share 47% identical amino acids. This number is significantly lower than the identity between Rorl and Ror2 (68%) or Trk and TrkB (76%), and is similar to the amino acid identity between Trk and insulin receptor (45%). In addition, both Rors share with Trks the YXXDYY amino acid sequence motif, corresponding to the autophosphorylation site of insulin receptor (35). Despite these similarities, more detailed inspection distinguishes Rors from other tyrosine kinases. Most of the 40 consensus tyrosine kinase residues identified by Hanks and Quinn (26) are conserved but Rorl and Ror2 have 7 and 5 amino acid changes, respectively, and hence are among the most divergent known members of the family. (In the Hanks-Quinn alignment only one sequence, Ltk, has five substitutions and none has more.) More importantly, the known changes are limited to 19 out of the 40 consensus residues. In Rors, three changes occur at positions conserved in all tyrosine kinases (Fig. 2). Both Rors share an Arg to Lys6I4 and a Phe to substitutions within the conserved HRD and DFG motifs, respectively. A replacement of the middle Gly in the glycine-rich loop of the ATPbinding site by Cys4'* in Rorl and Asp"* in Ror2 is particularly radical: this position is one of the six positions with absolute amino acid conservation among all tyrosine and serine/threonine protein kinases (26).
Following the kinase domain, Rorl and Ror2 proteins have a tail of about 200 amino acids containing two regions which are more than 50% serine/threonine-rich, separated by an over 25% proline-rich region with 4 conserved tyrosines. The two proteins show little similarity in COOH-terminal stretch of 60 amino acids whose overall charge is close to neutral in Rorl (4 Arg/Lys, 5 Asp/Glu), and is strongly negative in RorZ with 19 acidic amino acids and only 1 basic residue.
Protein Kinase Activity of Rors-In view of the unusual amino acid substitutions in the tyrosine kinase-like domain of Rorl and RorZ, it is particularly important to experimentally verify the enzymatic activity of these proteins. In vitro autophosphorylation of the immunoprecipitated kinase has FGF receptors, insulin (imR) and insulin-like growth factor I (ZGFR) receptors, and trk and trkB neurotrophin receptors are transmembrane tyrosine kinases; src is a cytoplasmic tyrosine kinase, and CAMP-dependent protein kinase a-subunit (cAPKa) is a serine/ threonine kinase. The amino acid sequences were downloaded from Hanks-Quinn database (26). The dendrogram was generated by the PILEUP program in the GCG package (33). been frequently used for such purposes. Since our anti-Ror antibodies do not efficiently precipitate the proteins, we have used a modification of this approach, in which we first expressed in COS cells truncated versions of the proteins, Rorl-TK-Myc and Ror2-TK-Myc, consisting of the tyrosine kinase-like domain with the Myc epitope added at the COOH terminus. These proteins could be easily immunoprecipitated with anti-Myc monoclonal antibody 9E10 (22). When immunoprecipitates from cells expressing Ror2-TK-Myc were incubated in the presence of [y3*P]ATP, specific phosphorylation of the recombinant protein was evident on short exposures of film (Fig. 5, middle panel). In addition, several other phosphoproteins, presumably representing phosphorylated immunoglobulins and some coprecipitating contaminants, were present in this sample. All the latter bands were also present in the control sample, but their intensity, as judged from varying the exposure times of film, was approximately 10 times lower. At long exposures of film, a weak band corresponding to Rorl-TK-Myc protein could be seen in samples from cells transfected with rorl-TK-myc gene. Similar results were obtained using two different variants of COS cells for transfection (COS-M5 and COS-7). While the residual contaminating kinase activity makes the interpretation of the weak labeling of the Rorl-TK-Myc band difficult, these data strongly indicate that a protein kinase activity is specifically associated with the tyrosine kinase-like domain of Ror2.
Conservation of Rorl and Ror2 Proteins Among Species-Conservation of many unique features in proteins encoded by two different human genes suggests important physiological functions of these proteins. We sought further support for this notion by detecting rorl and ror2 sequences in other species. Initially, we performed genomic Southern blot analysis using human rorl and ror2 probes. Unique bands could be detected with both probes in human, rat, and chicken DNA digested with various restriction enzymes (data not shown), indicating that our sequences correspond to single copy genes strongly conserved among different species.
Subsequently, we have PCR-amplified DNA fragments of rors from embryonic rat RNA. Sequencing of these fragments revealed that rorl and ror2 from rat are highly homologous to their human counterparts, with most nucleotide changes conserving the amino acid sequence. In the region of the tyrosine kinase-like domain shown in Fig. 6, rat Rorl shows 97% amino acid, and 91% nucleotide sequence identity with its human counterpart. Rat and human Ror2 proteins are 93% identical in this region, with 87% identity at the nucleic acid level. The striking replacement of the middle Gly in the glycine-rich loop of the tyrosine kinase domain is preserved rat Rorl has Cys, and rat Rod has Asp, instead of Gly, exactly as the corresponding human proteins. This evolutionary conservation argues that rors play some critical roles in the cell.
Expression of rorl and ror2 Genes-To begin characterization of the expression of ror genes, we have performed Northern blot analysis with rat RNA samples. Rorl probe detected a single predominant band about 6 kb in size, and ror2 a 4.5kb band (Fig. 7); a similar pattern was observed in several human cell lines (data not shown). Both rors are expressed at high levels during early embryonic development, in both the head and body of the embryo, with maximal expression at embryonic day 12, the earliest analyzed time point. The expression levels drop strongly around embryonic day 16. Interestingly, this preferential expression of rors in the rapidly growing embryo is in contrast to the increased expression of trkB at the late stages of embryonic development. In a survey of adult rat tissues, we could detect only very low quantities of rorl and r0r2 transcripts (data not shown).

DISCUSSION
Using a PCR-based approach, we have identified two novel human genes, rorl and ror2, encoding potential cell surface receptors. The deduced proteins have in their cytoplasmic portion a region homologous to the tyrosine kinase domain of growth factor receptors, especially those in the Trk family. At the same time, the shared structural features set Rors apart from Trks and other known receptors, as prototype members of a new family. The occurrence of complex ligandreceptor systems, in which several structurally related ligands are recognized with more or less specificity by a family of similar receptors appears to be a general feature of eukaryotic development. Individual members of such receptor families share the general design and show very high amino acid similarity. In Fig. 8, we have plotted the local amino acid identity as a function of amino acid position for Rors, and two other pairs of related tyrosine kinase receptors. The extent of similarity among Rors is comparable to that among members of the Trk family of neurotrophin receptors and the FGF receptor family, but the distribution of similarity appears somewhat different. In Rors, the structural motifs in the extracellular domain show higher sequence similarity than the kinase domain, while the opposite is true for FGF receptors and Trks. Interestingly, the amino acid sequence of the hydrophobic transmembrane domain is highly conserved in   Rors, but not in the other pairs of proteins analyzed in Fig.   8. The finding of two members of a new class of receptors in a limited sample of our initial PCR products is most likely the result of our modified PCR amplification strategy. PCR amplification of DNA sequences of novel tyrosine kinases using degenerate oligodeoxyribonucleotide primers has been described by several laboratories (38-40). All these authors have designed their primers around the two largest continuous blocks of amino acids, HRDL and SDVW (see the consensus in Fig. 3A), strongly conserved among all known tyrosine kinases. We, on the other hand, have concentrated on Trkspecific clusters of amino acids. This approach tended to yield much more restricted populations of products. However, by coupling it with efficient removal of the trk sequences by restriction digest and rapid scanning of PCR products by subcloning and sequencing, we could test multiple pairs of primers and clone sequences that would almost certainly be missed by the previous approaches. Both Rors have amino acid substitutions in the HRDL and SDVW consensus motifs (Fig. 3A), and thus would be strongly underrepresented in the population of PCR products obtained with primers derived from these regions.

A l a C y s G l u G l y G l u L e u P r o Leu S e r A l a V a l A r g P h e Met G l u G l u L e u
The similarity of the putative kinase domain of Rors to the corresponding region in the Trk family of receptors suggested that Rors may have tyrosine kinase activity despite the three amino acid substitutions at positions conserved in all tyrosine kinases (26). The Arg residue substituted by Lys61' in Rorl and Ror2, and Phe substituted by Leum are conserved in most, but not all, serine/threonine kinases demonstrating that they are not indispensable for catalysis. While the 3 Gly residues in the glycine-rich loop are conserved in all tyrosine kinases, only the middle one, replaced by Cyse2 in Rorl and Aspa2 in Ror2, is conserved without exceptions in all tyrosine and serinelthreonine kinases. The 5 other invariant amino acid residues, clustered in the crystal structure of CAMPdependent protein kinase (41) around the catalytic center are conserved in Rors (Fig. 3A) and correspond to Aspa15 and in the catalytic loop, and LysMWJIm7, G~u~~ and Asp*3. This conservation, including residues directly involved in catalysis and anchored to defined parts of the secondary structure, may be more significant than the loss of a single Gly residue in the flexible glycine-rich loop, whose main function is to position the y-phosphate of ATP for transfer (41,42). Other features of the glycine-rich loop, including the remaining 2 Gly residues, and a highly conserved Phe residue, are present in both Rors. The occurrence of amino acid substitutions at the same three positions in both Rorl and Ror2 could indicate either compensatory mutations, or modified catalytic properties.
The results of our in vitro kinase assay, using the Myctagged tyrosine kinase-like domains of Rors expressed in COS cells, suggest the protein kinase activity associated with Ror2 and imply a role for Rors in signal transduction pathways. However, more work will be needed to better characterize this activity. We are now beginning affinity labeling and in vitro mutagenesis studies to rigorously exclude the possibility that another tightly associated molecule is coprecipitating with the Ror2-TK-Myc protein. In addition, phosphoamino acids analysis will be needed to prove that the Ror2-associated activity phosphorylates tyrosines. Finally, we do not know whether the difference between Rorl and Ror2 observed in our assay is a reflection of the situation in uiuo, or is merely the effect of our experimental protocol. For example, the sequences were scanned by calculating percentage of identity within a 26-amino acid window, which was moved from NH2 to COOH terminus, 1 residue at a time. The obtained values were then plotted as a function of the position of the middle amino acid in each window.
T h e position of the thin horizontal line represents the overall similarity of a given pair. The thick horizontal bars mark the extent of the identified structural domains, and their uertical position represents the percent of amino acid identity within a domain. ZG, immunoglobulin-like domain; CR, cysteine-rich domain; KR, kringle domain; TM, transmembrane domain; TK, tyrosine kinase domain; ST, serine/threonine-rich domain; PR, proline-rich domain.
removal of the COOH-terminal tail, which in many receptor kinases contains multiple autophosphorylation sites (l), could have different effects on the two proteins. The Myc epitope tagging technology has been used earlier t o distinguish experimentally introduced and endogenous proteins (43, 44) and to identify and clone CNTF receptor (20, 21). Our results indicate that the usefulness of this approach could be extended to the functional dissection of individual domains of large modular proteins, obviating the need of extensive purification or specific immunoprecipitating antibodies. In addition, this approach may remove possible interference by other parts of the protein. Such inhibitory effects, possibly neutralized in vivo by ligand binding, might explain some of the apparent lack of activity in the immune complex assay of novel tyrosine kinase-like proteins, such as Tie (45), JAKl (46) and Klg (47). While the Klg protein lacks the Asp residue within the consensus DFG triplet and was postulated to play a role in signal attenuation (47), the two other proteins have well preserved kinase consensus sequences. In fact, data suggesting autophosphorylation of Tie (45) in uiuo, and enzymatic activity of the tyrosine kinase-like domain of JAKl (46) expressed in bacteria as a fusion protein have been presented. Our model system may allow us to directly study the effect of other parts of molecule, and of different amino acid substitutions on catalysis. In addition, other functional aspects of Rors can be studied with this approach. For example, in preliminary experiments the recombinant Rorl protein in which the transmembrane and cytoplasmic domains have been replaced by the Myc epitope has been efficiently secreted from the COS cells, supporting the proposal that rors are cell surface receptors (data not shown).
Since the most important insight into the physiological role of Rors will come from the definition of their ligands, we are presently evaluating various strategies to clone the corresponding genes. Expression cloning strategies based on cell survival and proliferation assay (48) would be the method of choice to identify secreted ligands, either novel or known factors with unidentified receptors. An example of the latter could be the Wnt family of secreted peptides expressed in fetal development (49) at the time when expression of rors is the highest. Alternatively, Rorl and Ror2 could be involved in cell-cell interactions, suggesting cloning strategies more suited for membrane-bound ligands (50). The kringle domain held in a distinctive conformation by the three disulfide bonds is thought to participate in protein-protein interactions (32). Kringle domains occur in many proteins involved in blood clotting, but more recent data implicate this structural motif in the control of cell growth and differentiation. Thus, HGF which is a potent mitogen for hepatocytes and a ligand for the transmembrane tyrosine kinase product of the c-met proto-oncogene (51) has four kringle domains (29), but a single kringle domain may account for most of the receptor binding (52,53). Thrombin (two kringle domains) is a potent mitogen for several cell types, acting through a member of the seven transmembrane domain receptor family (54); in neuronal cells, this interaction causes neurite retraction mediated through protein kinases (55). The kringle domain of Rors might interact with receptors for HGF, thrombin, or other related proteins. Singer (56) has recently discussed the developmental importance of mechanisms integrating cell-cell adhesion and intercellular signaling in binary cell-cell interactions. Possible examples of tyrosine kinase receptors involved in such interactions include sevenless (3), Dtrk, a Trk homologue from Drosophila (57), and the recently described Tie receptor (45).
The unique structural features in both the extracellular and cytoplasmic domains of Rors suggest that this new family of receptors may be involved in novel interactions with extracellular ligands and cytoplasmic components. The availability of rorl and ror2 cDNAs allows us to express the whole proteins or their various fragments in order to study these interactions.

Two Novel
Receptors with Tyrosine Kinase-like Domain and John Rudge for discussions, advice, and materials. P.M. thanks Lena Hofer for helpful discussions.