Characterization of sorCS1, an Alternatively Spliced Receptor with Completely Different Cytoplasmic Domains That Mediate Different Trafficking in Cells*

We previously isolated and sequenced murine sorCS1, a type 1 receptor containing a Vps10p-domain and a leucine-rich domain. We now show that human sorCS1 has three isoforms, sorCS1a–c, with completely different cytoplasmic tails and differential expression in tissues. The b tail shows high identity with that of murine sorCS1b, whereas the a and c tails have no reported counterparts. Like the Vps10p-domain receptor family members sortilin and sorLA, sorCS1 is synthesized as a proreceptor that is converted in late Golgi compartments by furin-mediated cleavage. Mature sorCS1 bound its own propeptide with low affinity but none of the ligands previously shown to interact with sortilin and sorLA. In transfected cells, about 10% of sorCS1a was expressed on the cell surface and proved capable of rapid endocytosis in complex with specific antibody, whereas sorCS1b presented a high cell surface expression but essentially no endocytosis, and sorCS1c was intermediate. This is an unusual example of an alternatively spliced single transmembrane receptor with completely different cytoplasmic domains that mediate different trafficking in cells.

SorCS1 is the first identified member of a subgroup of the mammalian Vps10p-domain (Vps10p-D) 1 receptor family that comprises an N-terminal Vps10p-D (named after the yeast vacuolar protein sorting 10 protein), a leucine-rich domain, a single transmembrane domain, and a short cytoplasmic domain (cd) (1). Two isoforms of sorCS1, with different cds arising from differential splicing, have been identified in the mouse (2). The two other known members of the subgroup are sorCS2 (3,4) and a highly homologous receptor tentatively designated sorCS3 (5). The mammalian Vps10p-D receptors also comprise the previously characterized sortilin, whose lumenal part consists of a Vps10p-D only (6), and the mosaic receptor sorLA/ LR11, which, in addition to an N-terminal Vps10p-D, contains elements also found in the low density lipoprotein receptor family as well as a cluster of fibronection type III repeats (7,8). The N termini of all the Vps10p-D family receptors contain sequences that conform to the consensus sequence for cleavage by furin (RX(R/K)R), and recent results (9,10) have shown that furin cleaves the precursor forms of sortilin and sorLA and that removal of the propeptides conditions these receptors for binding of ligands.
Sortilin and sorLA are mainly found in the trans-Golgi network, and only a few percent of the receptors are on the cell surface (10,11). The lumenal domains of the fully processed receptors bind certain neuropeptides (e.g. neurotensin) (10,12), the endoplasmic reticulum-resident receptor associated protein (RAP) (9,10), as well as lipoprotein lipase (10,11), apolipoprotein E (8,10), and notably their own propeptides, which block binding of all known ligands to the Vps10p-D (9,10). Both receptors provide internalization of cell surface-bound ligand (10,13), and the sortilin-cd has been shown to convey transport of cargo from Golgi to late endosomes and to bind GGAs (Golgilocalized ␥-adaptin ear containing ADP-ribosylation factorbinding proteins) (13), which are adaptor proteins believed to be involved in this type of trafficking (reviewed in Refs. 14 and 15). Considering that sorLA is mainly intracellular (10) and also binds GGAs (16), it seemed possible that Vps10p-D receptors other than sortilin might be targeted by similar sorting mechanisms.
SorCS1 is expressed in the murine central nervous system like sortilin and sorLA, and transcripts have also been observed in kidney, liver, and heart (1). The receptor prevails in neurons, and during embryonic development sorCS1 is expressed in a transient and dynamic pattern in areas of the nervous system where precursors proliferate as well as in regions where cells differentiate (17).
The purpose of the present work was to begin elucidating sorCS1 function. We cloned human sorCS1b, whose cd is highly similar to that of the murine orthologue, and identified two new isoforms with distinct cds and different distributions in tissues. The common lumenal domain was cleaved by furin in the late synthetic pathway demonstrating that sorCS1 is synthesized as a proprotein. Analysis of cells stably transfected with wild type and chimeric receptors showed that the three cds convey different distributions of sorCS1 in cells and have different capabilities for internalization. Neither the mature lumenal domain nor any of the cds bound any of the ligands previously shown to interact with sortilin and sorLA, demonstrating that sorCS1 is functionally different from the previously characterized Vps10p-D family receptors.

EXPERIMENTAL PROCEDURES
Cloning of Human sorCS1 Isoforms-Murine sorCS1b cDNA (Gen-Bank TM accession number AF195056) was radiolabeled and used for screening a human brain cDNA library in the lambda ZAP vector (Stratagene, La Jolla, CA). Five positive clones were purified and rescued into the pBK-CMV vector, and sequencing revealed three different overlapping clones representing bp 462-3504 of human sorCS1b cDNA in addition to a 3Ј untranslated region. The 5Ј part of the open reading frame was obtained by reverse transcription PCR. The first strand cDNA was synthesized from 1 g human fetal brain RNA (Clontech, Palo Alto, CA) using the primer 5Ј-AGGTGAAGGTGTAGTGAGCAAT-AGGG representing bp 1549 -1574. RNA was denatured in the presence of 40 pmol primer for 10 min at 94°C and transferred to prewarmed reaction mixture (50 mM Tris, pH 8.3, 75 mM KCl, 3 mM MgCl 2 , 0.5 mM deoxynucleoside triphosphates, 200 mM dithiothreitol). Following addition of reverse transcriptase (SuperScript II, Invitrogen), first strand reaction was performed for 55 min at 42°C. The reaction was terminated by 15 min heating to 70°C followed by 5 min on ice and incubation for 20 min at 37°C with RNase-H (AP Biotech, Little Chalfont, UK). A two-step PCR reaction was performed using Advantage-GC polymerase (Clontech). The same reverse primer was used as for the first strand reaction, and the forward primer (5Ј-CTCCCGCGATGG-GAAAAGTTGGC) was obtained from a sequence in the human chromosome 10 derived bacterial artificial chromosome RP11-557K21 (Gen-Bank™ accession number AL356439) homologous to the 5Ј end of murine sorCS1. The 1.5-kb product was subcloned into the pGem-T Easy Vector (Promega, Madison, WI) and sequenced. The full-length cDNA was obtained by ligation of overlapping fragments using the PstI (bp 1522), SapI (bp 1753), and BamHI (bp 3128) sites and transferred into pBluescript (Stratagene).
Chromosomal localization and organization were determined using the program tblastn (18) at the NCBI (National Institutes of Health, Bethesda, MD). Exonic sequences were determined by aligning cDNA and genomic sequences, and exon/intron boundaries were in agreement with consensus splice sites (19). Primers corresponding to the 3Ј ends of putative terminal exons of sorCS1a and sorCS1c (5Ј-TTAATAGAAAC-CATCACTGCTATG and 5Ј-TGGTGGCTACTGGGAATCATTTAC) were used for first strand synthesis on 1 g of total RNA from human fetal brain followed by two-step PCR reactions using the forward primer for generation of full-length sorCS1b.
Analysis of sorCS1 Transcripts-A multiple tissue blot (AP Biotech) was hybridized with a [ 32 P]dCTP randomly labeled probe of sorCS1 (bp 463-1924) and washed under high stringency conditions followed by autoradiography. For identification of sorCS1 isoforms, reverse transcription PCR reactions were performed on 1 g of total human RNA from fetal liver, fetal brain, adult cerebellum, and adult total brain (Clontech) using the reverse primer 5Ј-GCCTGTAGCCTTT-GGGGGTTTTCC for sorCS1b as well as the reverse primers designed for isolation of sorCS1a and -c cDNA. The forward primer (5Ј-TAC-CCACCACTGCTGAACTCTTTG) common to all isoforms was derived from exon 23. The PCR products were verified by sequencing.
Expression of sorCS1 Constructs-For generation of wild type receptor constructs, the sorCS1 cDNA was cut out of pBluescript using the KpnI site of the polylinker preceding the 5Ј end of the cDNA and the natural BamHI site at bp 3128. This was combined either with a BamHI/ApaI 3Ј fragment and cloned into pcDNA4/Myc-HisA (Invitrogen) for expression of the lumenal part of sorCS1 (L-sorCS1) or with BamHI/NotI 3Ј fragments and cloned into pcDNA3.1/Zeo(ϩ) (Invitrogen) for expression of full-length receptor isoforms. ApaI and NotI sites were introduced by PCR to the 3Ј ends using modified primers corresponding to nucleotides 2991-3308 or to segments following the stop codons. Alternatively, sorCS1a-c cDNA was transferred via NheI and NotI restriction sites from pcDNA3.1/Zeo(ϩ) to pcDNA3.1/Hygro(Ϫ) (Invitrogen) to generate doubly transfected cells. To produce L-sorCS1 mutated in the furin cleavage site ( 74 RRRR to 74 GRGR) we performed a two-step PCR using Advantage-GC polymerase. Overlapping 5Ј and 3Ј fragments were amplified from the original expression construct using the primers 5Ј-ATTAATACGACTCACTATAGGGAG, 5Ј-GATCCGCTC-CGCTCCGTCCCCTCCCGC, 5Ј-CCGGCGGGAGGGGACGGAGCGG-AG, and 5Ј-TTGTTCCATAATCGGTTGACCTCC. The resulting PCR fragment was digested with KpnI and SacI and used to replace the 5Ј end of L-sorCS1 cDNA in the pcDNA4/Myc-HisA vector. To produce chimeric constructs covering the lumenal and transmembrane parts of the interleukin-2 receptor-␣ (IL2R, Tac/CD25) and the sorCS1 cytoplas-mic tails (chi-a, -b, and -c), cDNA encoding the three tails was amplified by standard PCR technique using primers generating a 5Ј HindIII site and a 3Ј XhoI site. The primers for the a, b, and c tail constructs were: 5Ј-TCGTAAGCTTAAGTTTAAAAGGTGCG and 5Ј-CCTCTCGAGTTA-ATAGAAACCATCACTGCTATG, 5Ј-CGTCAAGCTTAAGTTTAAAAG-GAGAGTAGCTTTACCC and 5Ј-GGGGCTCGAGTTAAATTGCATACT-GTGCCCCAGCAGATCC, and 5Ј-CAAGCTTAAGTTTAAAAGGAA GATC and 5Ј-TACCTCGAGTCATTTACCTATGAGC. The HindIII/XhoI fragments were ligated into pcDNA3.1/Zeo(ϩ) together with a NheI/ HindIII fragment representing the lumenal and transmembrane parts of IL2R cut out of pCMV-IL2R/CD25/Tac (20).
Expression of sorCS1 Propeptide and Part of the Leucine-rich Domain-The propeptide sequence was amplified from sorCS1 cDNA using a 5Ј primer (5Ј-TCTGGATCCGGCGGCTCCTGCTGC) introducing a BamHI site to the propeptide N-terminal sequence and a 3Ј primer (5Ј-ATCCTCGAGTCACCGTCTCCTCCG) introducing a stop codon and a XhoI site after the furin cleavage site 72 RRRR. Following two-step PCR, the product was cloned via BamHI/XhoI into pGEX-4T-1 (AP Biotech). For expression of the leucine-rich domain (Glu 942 -Ile 1023 ) the native EcoRI and EcoRV sites were used to generate a construct in pGEX-4T-1. Both constructs were expressed in the bacterial strain BL21 (DE3), and the resulting GST-tagged proteins were purified using glutathion-Sepharose beads.
Yeast Two-hybrid Analysis-EcoRI and XhoI restriction sites were introduced into cDNA of the three sorCS1 cytoplasmic tails via PCR followed by insertion into the pLexA vector to generate bait strains. A Matchmaker LexA two-hybrid system (Clontech) was used as described before (13).
Surface Plasmon Resonance Analysis-Measurements were performed on a BIAcore 2000 instrument using CM5 sensor chips activated as described (9). L-sorCS1 and the lumenal domains of sortilin and sorLA were immobilized to an estimated density of ϳ60 fmol/mm 2 , and samples for binding (40 l, 25°C) were injected at 5 l/min in 10 mM Hepes, 150 mM NaCl, 1.5 mM CaCl 2 , 1 mM EGTA, 0.005% Tween 20, pH 7.4. Binding was expressed in units as the response obtained with immobilized receptor minus the response with an activated but uncoupled chip. The chips were regenerated as described, and kinetic parameters were determined using BIAevaluation 3.0 software as described (9).
Quantification of Cell Surface Expression-Cells were surface-labeled with the impermeable reagent sulfo-N-hydroxysuccinimidobiotin (Pierce), washed and lysed as described previously (10,11), and biotinylated proteins were precipitated with streptavidin-coupled Sepharose (Zymed Laboratories Inc., San Francisco, CA). The fractions of streptavidin-bound and unbound IL2R/sorCS1 chimera or sorCS1 isoforms in cell lysates were detected by Western blotting and quantified using a FUJIFILM LAS-1000 luminescence image analyzer.
Immunocytochemistry and Assay for Internalization-Transfected or control cells were washed in 10 mM phosphate, 150 mM NaCl, pH 7.3, fixed in the same buffer with 4% paraformaldehyde, and finally washed in buffer containing 0.05% Triton X-100 followed by incubation with primary (anti-Tac or anti-L-sorCS1) and secondary antibodies (fluorescein isothiocyanate-conjugated rabbit anti-mouse, DAKO, Alexa 488conjugated goat anti-rabbit or Alexa 568-conjugated goat anti-mouse, Molecular Probes, Leiden, The Netherlands). To visualize internalization, cells were surface-labeled with primary antibody at 4°C for 2 h followed by incubation at 37°C for various times. Fluorescence microscopy was performed using a laser scanning confocal unit (LSM510, Zeiss). Alternatively, cells transfected with chimeras were incubated at 4°C with 125 I-labeled monoclonal anti-Tac (3 ϫ 10 4 cpm/ml) for 2 h at 4°C, washed, and reincubated at 37°C for 0 -60 min. Incubations were stopped by the addition of ice cold acetic acid, 150 mM NaCl, pH 2.5. After 5 min, the supernatant was recovered, the cells lysed in 1 M NaOH, and radioactivity determined in the two fractions was defined as surface-associated and internalized antibody, respectively.

RESULTS
Identification of sorCS1 Isoforms-The human sorCS1b cDNA (GenBank TM accession number AF284756) encodes a 33 amino acid signal peptide followed by a 1135 amino acid type 1 receptor (Fig. 1A) with 92% sequence identity to the murine sorCS1b protein. The N-terminal part contains the sequence ( 74 RRRR 77 ) corresponding to the optimal multibasic motif (RXR/KR) for cleavage by furin and the sequence ( 93 RSPR 96 ) corresponding to the minimal requirements (RXXR) for cleavage. The human sorCS1b gene maps at 10q23.3, and analysis of genomic sequences indicated the presence of 26 exons. As two isoforms (sorCS1a and -b) differing only in their cds have been identified in the mouse (2), and only the sorCS1b variant was found by human cDNA library screening, we searched in human genomic databases for sequences corresponding to the 3Ј end of murine sorCS1a. However, no such sequence was found, and genomic sequences between exon 25 (encoding the transmembrane domain and four residues of the cytoplasmic tail) and exon 26 (encoding the remaining sorCS1b tail) were therefore analyzed to identify possible splice variants with little or no homology to murine sorCS1a. Putative exon sequences identified as open reading frames followed by polyadenylation signals were analyzed using primers 3Ј to potential stop codons and a 5Ј primer corresponding to exon 23 encoding part of the leucine-rich domain (cf. Fig. 1B). This approach revealed two new splice variants in human fetal brain RNA, and cDNA clones encoding the complete proteins were identified using a primer corresponding to the 5Ј untranslated region and the primers corresponding to sequences 3Ј to the stop codons. Fig. 1B shows the generation of human sorCS1 isoforms. Like in the mouse, human sorCS1a (GenBank™ accession number AY099453) is generated by using exon 25 as the terminal exon (stop codon a), but the amino acid sequence encoded by the part of exon 25 that is specific for the sorCS1a-cd is completely different from that of the murine receptor ( Fig. 1C  and Ref. 2). When the 5Ј splice site within exon 25 is active, sorCS1b is generated by skipping the 3Ј part of this exon and using exon 26 as the terminal exon. Comparison of the amino acid encoded by exon 26 of human sorCS1b showed 89% sequence identity with the corresponding segment of the mouse sorCS1b. Finally, human sorCS1c (GenBank™ accession number AY099452) is generated by skipping the middle part of exon 25 by using both splice sites within the exon. Because a murine counterpart was expected, we subsequently cloned mouse sorCS1c (GenBank™ accession number AF284755), 2 which revealed a cd with 87% amino acid sequence identity to the human sorCS1c-cd. Searches in databases indicated that the cds of human sorCS1a and -b show little similarity to other receptor cds, whereas the sorCS1c-cd exhibits about 50% identity to the cds of sorCS2 and -3.
Tissue-specific Expression-Northern blotting of human tissues using a probe common to all three isoforms showed high levels of transcripts in adult kidney and comparatively moderate levels in brain, heart, and small intestine ( Fig. 2A). To identify the sorCS1 protein, we performed immunoblots on two human tissue homogenates and on extracts of CHO cells mocktransfected or stably transfected with sorCS1a cDNA. A band of ϳ130 kDa was detected in the sorCS1a-transfected CHO cells and in kidney, but not in non-transfected CHO cells and spleen (Fig. 2B), in agreement with the Northern blots. We next examined expression of the three splice variants in human adult brain, adult cerebellum, fetal brain, and fetal liver by reverse transcription PCR. As shown in Fig. 2C, all three splice variants were detected in the brain samples (lanes 2-4), 2 G. Hermey, unpublished result. whereas only sorCS1c was found in human fetal liver (lane 1), demonstrating that sorCS1 isoforms are differentially expressed among tissues. SorCS1 Propeptide Is Removed by Furin-mediated Cleavage-To determine whether sorCS1 processing includes propeptide cleavage, CHO cells were stably transfected with the His 6 -tagged lumenal part (amino acid 1-1067) of sorCS1 (L-sorCS1). The transfectants were biolabeled, and L-sorCS1 secreted into the medium or present in cell lysates was recovered on Talon beads. Fig. 3A shows that labeled L-sorCS1 in the medium (lane 1) has a higher apparent molecular size than that obtained from lysates (lane 2), whereas, after deglycosylation with PNGase-F, the cellular form (lane 4) is larger than the secreted form (lane 3), demonstrating cleavage of L-sorCS1. This occurred within cells because labeled L-sorCS1 isolated from cell lysates was unchanged by incubation in conditioned CHO medium (not shown).  3 and 4). The results establish that sorCS1 is synthesized as a proprotein, which is converted to a mature form by cleavage in the furin-containing late synthetic pathway.
Characterization of Purified L-sorCS1-His 6 -tagged L-sorCS1 was purified from the medium of transfected CHO cells using a Talon column, and sequencing analysis yielded the N-terminal sequence 78 SGADQ, demonstrating cleavage at the optimal multi-basic motif ( 74 RRRR 77 ). In addition to this motif, a minimal basic site for potential cleavage by furin ( 93 RSPR 96 ) is present in the N terminus of human sorCS1, and optimal basic motifs are present at the same positions in the murine receptor (1), suggesting that some additional cleavage might occur after Arg 96 . To confirm the involvement of the optimal site for cleavage by furin and to analyze whether the minimal basic motif might be operational, CHO cells were transfected with L-sorCS1 mutated in the optimal motif ( 74 RRRR 77 to 74 GRGR 77 ). Fig. 4A shows that wild type and mutated L-sorCS1 were secreted to a similar extend (lanes 1 and 2), demonstrating that cleavage of the propeptide is not necessary for secretion of the receptor as previously shown for sortilin and sorLA (9,10). In contrast, propeptide cleavage is a prerequisite for transport of furin (21) and for most other proprotein convertases (Ref. 22 and references herein) out of the endoplasmic reticulum. The antibody against propeptide reacted with mutated sorCS1 (lane 3), and no propeptide immunoreactivity was detected in medium of cells secreting wild type L-sorCS1 (lane 4) even though GST-sorCS1 propeptide was readily detected following incubation in conditioned medium (not shown), suggesting that the propeptide is degraded within the cells. Because the presence of propeptide immunoreactivity in the medium of cells transfected with mutated L-sorCS1 did not exclude some cleavage after Arg 96 , cells were biolabeled, and mutated and wild type sorCS1 were recovered on Talon beads followed by treatment with furin. As shown in Fig. 4B, only the wild type receptor exhibited a smaller size after furin treatment, demonstrating little or no cleavage after Arg 96 . We conclude that L-sorCS1 secreted by CHO cells is cleaved only at the optimal site for cleavage by furin, i.e. after Arg 77 .
Binding of sorCS1 propeptide to mature L-sorCS1-(78 -1067) was measured because the propeptides of sortilin and sorLA bind to the Vps10p-Ds of their respective receptors with high affinities. Fig. 5A shows that GST-sorCS1 propeptide, but not GST alone, binds to purified L-sorCS1 (inset) with a K d estimated at about 0.7 M (range 0.5-1 M in 3 experiments). This low affinity binding was abolished by 20 mM EDTA and reduced to half at pH 6.0 (not shown). We then tested ligands previously shown to interact with sortilin and sorLA, including RAP (Fig.  5A), neurotensin, apolipoprotein E3, and lipoprotein lipase (not shown), and none of them bound to L-sorCS1 (K d values Ͼ 2 M).
Binding of the sorCS1 propeptide to mature sortilin was measured because the propeptides of sortilin and sorLA crossbind to the respective receptors (10). Surprisingly, GST-sorCS1 propeptide bound to sortilin (Fig. 5B) with a K d estimated at about 20 nM (range 17-30 nM in 3 experiments), which is comparable with the affinity for binding of sortilin to its own propeptide (9). The binding was abolished by 5 M RAP and markedly reduced by 20 M neurotensin (Fig. 5B) as previously shown for binding of the sortilin propeptide (9). In addition, the sorCS1 propeptide bound to mature sorLA and was inhibited by the 54-residue sorLA propeptide (not shown). Thus, sorCS1 is different from the previously characterized Vps10p-D receptors sortilin and sorLA in the sense that it does not bind ligands common to the two receptors and that it binds its own propeptide with low affinity. On the other hand, the sorCS1 and Sample origin: 1, kidney (1 day exposed); 2, liver; 3, small intestine; 4, brain; 5, heart; 6, skeletal muscle; 7, colon; 8, thymus; 9, spleen (all 5 days exposed). B, immunoblotting. Extracts of cells (50 g protein/lane) and human tissues (75 g protein/lane) were analyzed using anti L-sorCS1 antibody. Sample origin: 1, CHO cells stably transfected with sorCS1a; 2, mock-transfected CHO cells; 3, adult kidney; 4, adult spleen. C, reverse transcription PCR analysis. Reactions were carried out using total human RNA from: 1, fetal liver; 2, fetal brain; 3, adult cerebellum; 4, adult total brain. Splice variantspecific primers were used as indicated in Fig. 1B. The size of the detected fragments corresponds to the calculated base pairs: a, 404; b, 450; c, 487. The additional band of 592 base pairs generated with the primers used for variant c represents sorCS1a cDNA as the untranslated region of sorCS1a includes the sequence encoding sorCS1c. sortilin propeptides bind with similar affinities and apparently to the same or overlapping sites on sortilin.
Subcellular Distribution of sorCS1a-c-To determine whether the three alternative cds might mediate different subcellular distributions of the receptor, we analyzed transfectants expressing chimeras of the lumenal and transmembrane parts of IL2R (Tac/CD25) and the sorCS1a-c cds. Fig. 6 shows that the IL2R/sorCS1a-cd chimera (chi-a) is predominantly intracellular, whereas the chimeras containing the sorCS1b and -c tails (chi-b and chi-c) exhibit distinct cell surface expression in addition to a comparatively low expression in paranuclear compartments. We also performed surface biotinylation of the transfectants followed by lysis, recovery of biotinylated proteins on streptavidin-Sepharose beads, and scanning densitometry of Western blots of biotinylated and non-biotinylated chimeras. The estimated cell surface expressions were 10% for chi-a, 46% for chi-b, and 30% for chi-c (mean of two experiments). Parallel experiments were performed with cells transfected with wild type sorCS1a-c using polyclonal anti-L-sorCS1, and the estimated surface expressions were 11% for sorCS1a, 34% for sorCS1b, and 24% for sorCS1c (mean of three experiments). Thus, the a tail mediates a much lower cell surface expression of sorCS1 than the b tail, whereas the c tail appears to be intermediate.
Internalization of sorCS1a-c-As the different subcellular distributions could result from differences in capabilities of the cds to mediate internalization, the chi-a, -b, and -c transfectants were incubated with 125 I-anti-Tac for 2 h at 4°C followed by removal of unbound antibody and continued incubation at 37°C. Internalized antibody was then determined as the amount of cell-associated radioactivity not removed by a following incubation at 4°C, pH 2.5. The transfectants bound 10 -20% of the added tracer after the incubation at 4°C, whereas no binding was observed in mock-transfected cells. Fig. 7A shows that about 65% of the initially surface-bound 125 I-anti-Tac had been internalized by chi-a after 60 min, whereas internalization was much less pronounced in cells transfected with chi-b and chi-c. No degradation of cell-bound 125 I-anti-Tac was observed after 2 h (not shown), suggesting that complexes of anti-Tac and the chimeric receptors remained stable within the cells. We therefore followed internal- FIG. 3. Cleavage of L-sorCS1. A-C, CHO cells transfected with His 6 -tagged L-sorCS1-(1-1067) were biolabeled and receptor secreted into medium (M) or extracted from cell lysate (L) was recovered on Talon beads, treated with enzymes as indicated, and analyzed by reducing SDS-PAGE followed by autoradiography. D, Western blot analysis of lysate and medium using anti-sorCS1 propeptide (␣-propep) or antibody against the leucine-rich domain of sorCS1 (␣-leu).

FIG. 4. SorCS1 cleavage in CHO-cells depends on the intact consensus sequence 74 RRRR.
A, medium of CHO cells expressing L-sorCS1 (WT) or L-sorCS1 mutated in the consensus cleavage site (FM) was analyzed by Western blotting using antibody against the leucinerich domain (␣-leu) or the propeptide (␣-propep). B, cells were biolabeled, and L-sorCS1 and mutated L-sorCS1 recovered from lysates on Talon beads were treated with furin followed by SDS-PAGE and autoradiography of the 2,5diphenyloxazole-treated gel. ization at 37°C of anti-Tac bound to cell surfaces at 4°C by confocal microscopy. As shown in Fig. 7B, surface-bound anti-Tac was rapidly internalized to paranuclear compartments by the chi-a-transfected cells, whereas the chi-b or chi-c chimeras gave rise to little or no paranuclear staining.
To ensure that the cd is the major determinant for internalization, i.e. that wild type and chimeric receptors behave similarly, we generated double transfectants co-expressing wild type receptors and chimeras in different combinations. Fig. 8 presents examples of internalization of wild type sorCS1 isoforms (green fluorescence) as compared with chimeras (red fluorescence). The upper panels show that both sorCS1b and chi-b remain largely on the cell surface. The more patchy appearance of sorCS1b might suggest that it coalesces more easily than the chimera or might be related to the polyclonal antibody used. The middle panels demonstrate that sorCS1a is efficiently internalized in contrast to chi-b, and the reverse experiment (not shown) confirmed inefficient internalization of sorCS1b as compared with chi-a. Finally, the lower panels show that the c tail is capable of mediating some internalization although much less efficiently than the a tail. Thus, as compared with the b and c tails, the cytoplasmic tail of sorCS1a mediates a small fractional surface expression and an efficient internalization of surface-bound ligand.
Because sorCS1a exhibited a subcellular distribution and internalization to paranuclear compartments similar to sortilin and sorLA, whose cds bind GGA1 and -2 (13,16), we tested the sorCS1 cds for interaction with the GGAs by yeast two-hybrid analysis. However, none of the sorCS1 cds displayed any binding, suggesting that sorCS1 is not implicated in GGA-mediated functions in contrast to the two previously characterized members of the Vps10p-D family.

DISCUSSION
This study demonstrates alternative splicing of the Vps10p-D receptor sorCS1 and characterizes three isoforms that are differentially expressed in human tissues. The isoforms have identical lumenal and transmembrane domains but different cytoplasmic tails. The cds of both human and murine sorCS1a contain putative internalization signals (Fig. 1C and Ref. 2) and might therefore be functionally equivalent, although their amino acid sequences are totally different. In contrast, the cds of human sorCS1b and sorCS1c are highly similar to their murine orthologues, and the cd of sorCS1c exhibits about 50% sequence similarity to the cds of sorCS2 and -3.
We show that sorCS1 is synthesized as a proprotein, which is converted to the mature receptor in late Golgi compartments by cleavage of the RRRR2S 77 sequence. Furin is present in this compartment and is likely to be responsible for the cleavage, although proconvertases of the subtilisin/Kex-2-like family with overlapping specificities may also participate. Analysis of receptor constructs in which the typical consensus site was inactivated showed no signs of alternative cleavage after RSPR 96 within CHO cells or after treatment of the precipitated protein with furin under conditions that readily cleaved after Arg 77 . This is surprising because RSPR is thought to be a recognition motif for members of the proprotein convertase family of the general sequence (R/K)X n (R/K), where n ϭ 0, 2, 4, or 6 (23), and because furin is known to cleave the similar sequence RLPR2D in the mannose 6-phosphate uncovering enzyme (24). In addition, the pro-form of BACE (beta site FIG. 8. Internalization of sorCS1a-c. Cells doubly transfected with sorCS1b and chi-b, sorCS1a and chi-b, or sorCS1a and chi-c were incubated for 2 h with rabbit anti-sorCS1 and mouse anti-Tac, washed, re-incubated for 2 h in 37°C warm medium, stained after fixation (Alexa 488-conjugated anti-rabbit and Alexa 568-conjugated antimouse), and analyzed by confocal microscopy .   FIG. 7. Internalization of IL2R/ sorCS1-cd chimeras. A, CHO cells stably transfected with chi-a (circles), chi-b (squares), or chi-c (triangles) were incubated with 125 I-labeled anti-Tac Ig for 2 h at 4°C, unbound tracer was removed by washings, incubations were continued at 37°C and stopped by the addition of icecold acid buffer at the times indicated. Internalization was determined as the percent cell-associated radioactivity not released at pH 2.5. The values are averages of 3 experiments Ϯ 1 S.D. B, confocal microscopy of CHO cells transfected with IL2R/sorCS1a-c chimeras (chi-a, -b, and -c). Cells were incubated for 2 h at 4°C with mouse anti-Tac Ig, washed, and reincubated in 37°C warm medium for 0 -60 min, fixed and stained with fluorescein isothiocyanate-conjugated antimouse Ig. amyloid precursor protein-cleaving enzyme) is cleaved at a RLPR2E sequence (25). However, the presence of an Arg at position Ϫ6 in uncovering enzyme may greatly facilitate its cleavage (24), and it could not be completely excluded that a proprotein convertase other than furin accounted for the cleavage of BACE (25). Alternatively, the presence of a hydrophobic residue at position Ϫ3 (Leu) or an acidic residue (Asp or Glu) at position ϩ1 as in uncovering enzyme and BACE might facilitate cleavage by furin. Future studies should establish whether other members of the proprotein convertase family may cleave sorCS1 after Arg 96 .
The propeptides of sortilin and sorLA protect the receptors against premature ligand binding in the synthetic pathway (9,10) and might in addition function as intramolecular chaperones. This is in partial analogy to the propeptide of furin, which remains bound to the endoprotease after cleavage and permits its exit from the endoplasmic reticulum (26). The function of sorCS1 propeptide may be similar even though the affinity for binding to its own receptor is comparatively low. Surprisingly, the sorCS1 propeptide showed much higher affinity to mature sortilin and sorLA, raising the possibility that it might interact with these members of the Vps10p-D receptor family in the synthetic pathway.
We suspected that mature sorCS1 might bind one or more of the ligands previously shown to interact with the Vps10p-Ds of both sortilin and sorLA, i.e. RAP, neurotensin, and lipoprotein lipase (9 -12). However, this was not the case, indicating that the lumenal domain of sorCS1 has a ligand-binding profile different from those of previously characterized Vps10p-D receptors. In addition, none of the sorCS1 cds bound GGAs in contrast to those of sortilin and sorLA (13,16). This is in accordance with the lack of the previously defined requirements for GGA binding (13,16,(27)(28)(29), although the sorCS1c-cd contains a reminiscent sequence near the C terminus. It is therefore unlikely that sorCS1 is implicated in GGAmediated functions.
Irrespective of the mechanisms involved, we demonstrate that the three sorCS1 cds generated by alternative splicing mediate different localization and trafficking of the receptor, as sorCS1a is about 90% intracellular and sorCS1b exhibits a large cell surface expression. Because all three cds are expressed in the brain, different localizations of receptor isoforms might explain that sorCS1 immunoreactivity is seen as punctate cytoplasmic staining in some neurons, whereas other neurons exhibit significant cell surface expression (30). The efficient internalization of sorCS1a is in accordance with the presence of a YXXØ motif and a dileucine motif in the a tail, whereas the almost non-internalizing b tail lacks such motifs (Fig. 1C). Putative internalization signals are also present in the c tail, and it remains to be clarified why this cd only mediates internalization to a moderate extent. In addition, the presence of a single SH3 domain-binding motif (PXXP) in the a and c tails, and of multiple partially overlapping motifs in the b tail, suggest that the cds may interact with scaffold or adaptor proteins that may contribute to signal transduction events (31).
Alternative splicing is not uncommon among receptors and is an important mechanism for modulating functions, including those of cytoplasmic tails. The use of composite internal/terminal exons can lead to deletions or different lengths of receptor tails (32), and optional exons to insertion of additional amino acid sequences as demonstrated for the apolipoprotein E receptor-2 (33). On the other hand, the creation of completely different cytoplasmic domains is unusual. Interestingly, analysis of EST databases predicted an exchange of a transmembrane and cytoplasmic domain of an FC receptor ␤-chain homologue, which might modulate signal transduction activity (34). However, the present work is to our knowledge the first demonstration of a receptor with completely different cytoplasmic tails that mediate different trafficking in cells.
In conclusion, we show that sorCS1 is synthesized as a proprotein that is cleaved to mature forms in the trans-Golgi network and expressed in three isoforms with different cytoplasmic domains capable of mediating different trafficking of the receptor.