cell-associated episialin is a complex containing two proteins

THE JOURNAL OF BIOLOGICAL CHEMISTRY 0 1992 by The American Society for Biochemistry and Molecular Biology, Inc.

Vol. 267, No. 9, Issue of March 25, pp. 61714177,1992 Printed in U.S.A.

Cell-associated Episialin Is a Complex Containing Two Proteins Derived from a Common Precursor*

(Received for publication, August 7, 1991)

Marjolijn J. L. LigtenbergS, Lars Kruijshaar, Femke Buijs, Marja van MeijerO, Sergey V. Litvinovy, and John Hilkensll From the Division of Tumor Biology, The Netherlands Cancer Institute (Antoni van Leeuwenhoekhuis), Plesmanlaan 121,1066 CX Amsterdam, The Netherlands

cDNA for the epithelial sialomucin episialin encodes a transmembrane molecule with a large extracellular domain, which mainly consists of repeats of 20 amino acids. Here we confirm the existence of a previously proposed proteolytic cleavage of episialin that occurs in the endoplasmic reticulum (Hilkens, J., and Buijs, F. (1988) J. Biol. Chem. 263, 4216-4222) and show that a similar cleavage takes place in in vitro translation systems. Using in vitro translation of truncated mRNAs, we map the cleavage site to a region located between 71 and 63 amino acids upstream of the transmembrane domain. Analysis of a mutant, in which this region has been deleted, indicates that the cleavage sites used in vitro and in vivo are identical or in close proximity. Both cleavage products remain associated although they are not linked through disulfide bonds. Therefore, the subunit derived from the N terminus, which represents the actual mucin-like domain, remains indirectly anchored to the cell membrane as a result of its interaction with the C-terminal subunit.

Mucins are large molecules containing many 0-linked glycans. The structure of the epithelial sialomucin episialin has been studied extensively. The availability of monoclonal antibodies directed against the protein backbone of episialin has made it possible to clone its cDNA and to study its biosynthesis. The cDNA sequence predicts that episialin is a type I membrane molecule with a large extracellular domain, which varies in length between approximately 1000 and 2200 amino acids, and a cytoplasmic domain of 69 amino acids (Ligtenberg et al., 1990; Wreschner et al., 1990; Gendler et al., 1990; Lan et al., 1990). The major part of the extracellular domain consists of repeats of 20 amino acids that have a high content of serine and threonine residues, many of which are potential 0-linked glycosylation sites. As a result of a genetic polymorphism, the number of these repeats may vary between 30 and 90, causing the variation in length of the extracellular domain. An additional polymorphism, which affects only a single nucleotide, results in the use of either one of two splice

* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

4 Supported by a grant from the Dutch Cancer Society. f Present address: Dept. of Molecular Biology, Central Laboratory

of the Netherlands Red Cross Blood Transfusion Service, Plesman- laan 125,1066 CX Amsterdam, The Netherlands.

W Supported by a fellowship from the International Union for Cancer Research.

205122018; Fax: 31-206172625. I1 TO whom correspondence should be addressed. Tel.: 31-

acceptor sites for the second exon (Ligtenberg et al., 1991). These splice acceptor sites are located 27 base pairs apart. As a consequence, the splice variants differ by only nine amino acids at the protein level, variant A being nine amino acids longer than variant B.

The biosynthesis of episialin has been described by several groups (Hilkens and Buijs, 1988; Linsley et al., 1988; Abe and Kufe, 1989). Both Hilkens and Buijs (1988) and Linsley et al. (1988) have observed that the molecular mass of the first detectable precursor is reduced by 20 kDa within 4 min. Hilkens and Buijs (1988) have proposed that this shift is the result of a proteolytic cleavage. This putative cleavage should take place in the endoplasmic reticulum, since it occurs shortly after translation. The mobility on SDS’-polyacrylamide gels of the subsequent intermediate is drastically reduced, because a large number of 0-linked sugars is added to the molecule. During the last step of the processing, the mobility of the episialin molecules is slightly altered by the addition of sialic acids. However, as a result of the high molecular weight of the molecules, an additional proteolytic cleavage at this stage could not be excluded (Hilkens and Buijs, 1988).

The presumed proteolytic cleavage in the endoplasmic reticulum should remove 20 kDa from either the N or the C terminus of the molecule. Based on the length of the different domains predicted by the cDNA sequence, cleavage in the C- terminal part of the molecule is expected to separate the actual mucin-like domain containing the repetitive region from the transmembrane and the cytoplasmic domain. This would require an alternative mechanism for the membrane association of this mucin-like domain. On the other hand, this cleavage could explain why the mucin-like molecule is released from carcinoma cells and found in serum of patients with breast cancer, in which it is demonstrated to be an important marker to monitor breast cancer therapy.

In this study, we have analyzed the early proteolytic cleavage step in detail by comparing the processing of episialin synthesized in vitro and in vivo. We demonstrate that the cleavage indeed occurs in the C-terminal part of the molecule and that the resulting cleavage products remain associated, which explains the membrane anchorage of the mucin-like domain.

MATERIALS AND METHODS

cDNA Constructs-The generation of full-length cDNAs encoding episialin has been described previously (Ligtenberg et al., 1992). cDNA constructs containing a single repeat were generated from these full-length cDNA constructs of the A and B variant by exchang- ing a BsmI-EcoNI fragment of the cDNA constructs with a BsmI-

The abbreviations used are: SDS, sodium dodecyl sulfate; LRP, lipoprotein receptor-related protein; ASGP, ascites sialoglycoprotein.

6171

6172 Episialin Complex Formation EcoNI fragment of a genomic clone that had lost all hut one repeat during replication in a Rec' hacterial strain. This recomhination event had not affected the reading frame downstream of the repeat region.

The coding domains of these cDNAs were isolated hy digestion with RamHI (located in the polylinker just 5' of the cDNA and 18 nucleotides upstream of the polyadenylation signal) and inserted into the RamHI site of pGEM.77, (Promega Corp.). Plasmids containing the insert in hoth orientations were isolated. RNA could thus he made using either SP6 or T7 RNA polymerase.

In Vitro Transcription and Translation-In vitro transcription and translation reactions were performed using SP6 or T7 polymerase and rahhit reticulocyte lysates according to the instructions of the manufacturer (Promega Corp.). When indicated, in vitro translation was stopped by addition of cycloheximide (100 pM). In vitro translation products were analyzed on standard SDS-polyacrylamide (Laem- mli, 1970) or tricine-SDS-polyacrylamide (Schiigger and von Jagow, 1987) gels, as indicated in the figure legends.

The synthetic oligopeptide THGRYVPPSSTDRSPYE, which rep- Generation of Polyclonal Antiserum and Immunoprecipitation-

resents a region in the cytoplasmic domain, was coupled to keyhole limpet hemocyanin and used to immunize a rahhit. Antibodies directed against the C-terminal domain were purified from the polyclonal antiserum on an affinity column carrying a fusion protein consisting of the major part of &galactosidase and part of the cytoplasmic domain of episialin.

Pulse chase experiments and immunoprecipitations were performed essentially as descrihed previously (Hilkens and Ruijs, 1988). Cellulose heads coated with donkey anti-mouse antihodies or agarose beads coated with protein A were used as immunoadsorhents for immunoprecipitations with murine or rahhit antihodies, respectively. Unless otherwise stated, precipitations were performed in 25 mM Tris/HCI, pH 7.4, 1% Nonidet P-40,4 mM EDTA and 100 mM sodium chloride.

Trratment of Cell Lysates-The human breast carcinoma cell line ZR-75-1 was laheled with ['H]glucosarnine for 6 h and lysed in PRS containing 0.5% Nonidet P-40. The cell lysate was treated with 6 M urea, 5% [j-mercaptoethanol, or 1% SDS for 20 min a t room temper- ature. The huffers containing urea and P-mercaptoethanol were replaced by PHS, 0.5% Nonidet P-40 using a Sephadex G-25 column. The sample containing 1% SDS was diluted 10 times in PRS, 0.5% Nonidet P-40 to ohtain conditions suitahle for immunoprecipitation with monoclonal antihody 199H2 and with the polyclonal antiserum. The treated cell lysates were split just hefore immunoprecipitation wit,h these antihodies.

Generation of Mutant cDNA-A mutant cDNA clone was generated using the polymerase chain reaction. We took advantage of the fact that the region that was to he deleted was located hetween the only two KpnI sites within the cDNA, just upstream of the most 3' KpnI site. The oligonucleotide 5' GCCAGGGCTACCACAAC 3' homolo- gous with the sequence located 129 to 113 nucleotides upstream of the 5' KpnI site WBR used as the upstream primer in the polymerase chain reaction. The downstream primer was directed to a sequence located I00 to 85 nucleotides upstream of the 3' KpnI site. A KpnI site was introduced directly 5' of this sequence, skipping the region of the intended deletion, which resulted in a primer with the sequence 5' GCCGCCGGTACCATAAATCTGCAAAAAC 3', in which the Kpnl site is underlined. The polymerase chain reaction was performed as described previously (Ligtenherg et al., 1991), the product was digested with KpnI, and the resulting fragment was used to replace t.he wild-type Kpnl fragment of the cDNA construct. In this way, a region of 84 nucleotides, encoding 28 amino acids, was deleted immediately upstream of the 3' KpnI site, without affecting the reading frame (Fig. 5A ).

Transfmtion of HRL-100 Cells-The SV40-transformed mammary epithelial cell line HRL-100 was transfected with either the wild-type episialin cDNA or the deletion mutant in which the region containing the proteolytic cleavage site had heen eliminated. Wild-type and mutant cDNAs hoth containing approximately 40 repeats were cloned under the control of t,he cytomegalovirus immediate early promoter in the expression vector pCMVIK-AK1-DHFR (Whang et a[., 1987). of which the RamHl fragment containing the region encoding the dihydrofolate reductase (DHFR) gene had heen removed. This vector contains a neomycin resistance gene. Transfections were performed using the calcium-phosphate co-precipitation method. Stahle transfectants were select.ed hy culturing the cells in the presence of the neomycin analogue G418.

Sandwich Rodioimmunoassay-The sandwich assay was performed

as described previously (Hilkens el ai.. 1986). Either monoclonal antibody 115D8 (Hilkens et ai.. 1984) or the purified polvclonal antihodies directed against part of the cytoplasmic domain were used as catcher. In hoth cases, monoclonal antihodv 1l:iI)H laheled with ""I was used as a tracer. Cell lysates were prepared in I'HS rontaining 0.5% Nonidet P-40. .Just hefore the assay. Triton X-100, sodium deoxycholate, and sodium dodecyl sulfate were added to the Ivsates to a final concentration of 1, 0.5, and 0.15, respertivelv. Samples were added in triplicate to the wells either directly or after incuhation at 100 "C for 10 min, followed hy rapid cooling on ice.

RESULTS

Episialin Is Proteolytically Cleawd in Vitro-To analyze the early proteolytic cleavage, in vitro synthesized RNA e n - coded by episialin cDNA conta in ing on ly one repea t was translated for 60 min in rahhit reticulocyte lysates. After labeling with (""S]methionine, the major translation product had a length of ahout 41 kDa. However, after laheling with [""S]cysteine, these products were not detected (resul ts not shown). The full-length in vitro t ranslat ion product was expected to be laheled with both methionine and cysteine, since the open reading f rame conta ins 4 widely spread methionine residues (1 in the signal peptide, 2 in the remainder o f the extracellular domain, and 1 in the cytoplasmic domain) and 3 cysteine residues that are all present in the t r ansmemhrane domain. The absence of a detectable product after labeling with [:'%]cysteine suggested therefore, that the t ranslat ion product did not contain the t ransmemhrane domain . This might be caused either by removal of the C-terminal par t of episialin, including this domain af ter a full-length translation product has been synthesized, or by a strong t ranslat ional stop signal in the mRNA, which can not be overcome by the translation machinery. To investigate whether the ohserved translat ion products were formed after a proteolytic cleavage. the t rans la t ion was arrested after 22 min using cycloheximide, whereas the incuhation of the t rans la t ion mix ture was continued. A t var ious t ime points , samples of the in oitro t ranslat ion react ion were taken and analyzed on a tricine-SDS-polyacrvl- amide gel (Fig. 1). After only 14 min of t ranslat ion in both

N- t

C-

-43

-31

I " 1 2 3 4 5 6 7 8 9 I O 1 1 1 2 1 3 1 4

FIG. 1. Cleavage of in vitro translated episialin. Autoradi- ogram of in oitro translated episialin rnrodrd hy splire variant R separated on a tricine-SDS-polvacwlamide gel (10';' sparer and 16.5'7 separating gel). The in uifro translation was performed in the presence of I."'S]methionine (lanes 1-6) or ['.%]cysteine (lnnrs 7-12). After 22 min, cycloheximide (100 p M ) was added. and samples were taken at different time points (14, 30. 45, 90. 150. and X U ) min. as indicated in the figure). Imnrs 1.7 and 14 demonstrate two separate in ~ i t r o translation reactions, to which no cycloheximide was added. performed in the presence of [ '~'SJcysteine for 7 0 min. In the react ion presented in lane 13. RNA derived from cDNA truncated at the I'ruI site was added; in lanr 14, RNA derived from full-length cDNA was added as in lanes 1-12. F, the full-length translation products; N a n d C, the N- and C-terminal cleavage products. respectivcbly. (The hand of ahout 30 klla has heen shown to represent a protein that is encoded hy an open reading frame of 245 amino acids that starts downcitrearn of the in oioo start codon in an alternative reading frame.)

Episialin Complex Formation 61 73

the methionine- and cysteine-labeled samples, a major product with a length of about 58 kDa was detected (lanes 1 and 7). After inhibition of the translation and continued incubation, this band disappeared, and in the methionine-labeled samples, a band of about 41 kDa, which we had detected before, and a doublet of bands of about 20 kDa were observed (lanes 2-6), whereas in the cysteine labeled samples only the doublet could be detected (lanes 8-12). This indicates that the 41-kDa translation product is indeed generated as a result of a proteolytic cleavage step, which removes the transmembrane region containing the cysteine residues and the cytoplasmic domain. The doublet of bands of 20 kDa, which should contain about a third of the methionine residues present in the larger N-terminal cleavage product, is relatively poorly visualized. This is probably the result of further degradation of the cleavage products, since a general loss of cysteine labeled products is observed on short exposures of the auto- radiogram. Moreover, distortion of the electropherogram due to the presence of large amounts of unlabeled globin cannot be excluded.

To confirm that the doublet of bands of about 20 kDa represents the C-terminal cleavage product, RNA that was truncated at the PuuI site, which is located within the region encoding the cytoplasmic domain, was translated in the presence of [“S]cysteine for 70 min. The resulting translation products were compared with those of a similar reaction in which nontruncated RNA was used (compare lanes 13 and 14). Both the noncleaved precursor and the C-terminal cleavage product derived from the truncated RNA were expected to be about 3 kDa smaller than those derived from the full- length RNA. This shift in molecular mass was indeed observed for the noncleaved precursor. Upon truncation, the doublet of bands of about 20 kDa was absent, which supports the assumption that this doublet represents the C-terminal cleavage product. No alternative C-terminal fragment derived from the truncated molecule was observed, probably because it comi- grates with the large amount of globin present in reticulocyte lysates. The appearance of the C-terminal cleavage product as a doublet might be the result of the use of two closely spaced cleavage sites or of two consecutive cleavage events.

Determination of the Site of Proteolytic Cleavage in Vitro- To localize the amino acid sequence that is the target of the observed proteolytic cleavage, RNA was synthesized from the cDNA constructs of both variants A and B, which were truncated at different restriction sites, as indicated in Fig. 2A. Translation reactions were performed for 35 min to obtain a mixture of precursor and proteolytically cleaved translation products (Fig. 2R). All products were compared with the full- length translation products of either the A or the B variant (about 59 and 58 kDa, respectively) and their proteolytically cleaved products (about 42 and 41 kDa, respectively) (lanes 6 and 7). The translation products encoded by the cDNAs truncated at or upstream of the Ssp1 sites are smaller than these proteolytic cleavage products (lanes 1-3 and 10-13), whereas the precursors encoded by the cDNAs truncated at the 3‘ KpnI and the PvuI sites are larger than these cleavage products (lanes 4 , 5, 8, and 9). This pinpoints the cleavage site of both variant A and B to a stretch of 18 amino acids encoded by the region of the cDNA located between the Ssp1 and the 3’ KpnI site. This area is located in the extracellular domain, 71 to 53 amino acids upstream of the transmembrane domain. The length of the resulting C-terminal cleavage product is approximately 160 amino acids, which is in good agree- ment with its estimated molecular mass of about 20 kDa on the tricine-SDS-polyacrylamide gel (Fig. 1). Since the proteolytic cleavage site maps within the same region for both

0 varlant A

-06

-43

1 2 3 4 5 0 7 8 9 1 0 1 1 1 2 1 3

FIG. 2. AnalyRis of the cleavage Rite used in vitro. A , schematic representation of the different domains in the translation product and of several restriction sites present in the cl>NA. The black b x indicates a single repeat unit. The arrowhmd indicates the site where the nine-amino acid difference hetween variants A and H is located. 7”, extra, and intra, the transmemhrane, extracellular, and intracellular domains. respectively. I ( . variant A Ilnnm 1-6) and variant H (lanes 7-13) RNAs were translated in rnhhit reticuloc.yte lysates in the presence of [ ‘”Slmethionine for :IS min. ‘The translation products were analyzed on a 10? SI)S-polyncrylamide gel. The restriction sites that were used to truncate the episialin c D N A s are indicated in the figure. The translation products prrsent in lnnm 6 and 7 are derived from cDNAs that were not tn~ncated. Kpnl was used to make both a complete (Ianm 1 and 1 9 ) and a partial digestion (Inncs 4 and 9). In the latter lanes. translation products of the cDNA truncated at the two different Kpnl sites. as well as the full-length episialin molecules, and their proteolytically rleaved products nre present.

splice variants, in all further experiments only splice variant A was used.

The issue of cotranslational membrane insertion as a factor that could affect the proteolytic cleavage(s) observed was addressed by performing similar in u i tm translation experiments but now in a system supplemented with microsomal membranes. The translation products were analvzed after N- glycanase treatment to remove the N-linked glvcans. The size of the larger proteolytic cleavage product was again slightlv smaller than that observed for the translation product derived from the cDNA truncated at the 3’ KpnI site (results not shown). This suggests that the cleavage occurs at the same site independent of the presence of microsomal membranes. The structural elements recognized hv the proteol-ytic machinery are therefore affected neither hv membrane insertion nor by N-linked glycosylation.

Proteolytic Cleavage of Epkialin in Viuo-To check whether the proteolytic cleavage ohserved in vitro is indeed similar to the cleavage that has been reported to occur in the endoplasmic reticulum, a polyclonal antiserum was raised against a synthetic peptide with a sequence identical with a stretch of 17 amino acids located in the cytoplasmic tail of episialin. If a similar cleavage would occur in vivo. the c-ytoplasmic domain would be separated from the extracellular domain containing the repeat region, and it would thus be expected that the latter domain would not be immunoprecipitated hv the polyclonal antiserum. To test this hypothesis, the human breast carcinoma cell line T4SD was labeled with [ ‘Hlthreo- nine for 20 min followed by chase periods of 0, 20, 40, 60. 90 or 180 min. Cell lysates were immunoprecipitated with monoclonal antibody 139H2 directed against the repeat, region and with the polyclonal antiserum directed against part of the

6174 Episialin Complex Formation

cytoplasmic domain. Both the monoclonal and polyclonal antiserum immunoprecipitated the precursor and mature forms of both the large and small alleles (Fig. 3), suggesting that the observed decrease in molecular mass of about 20 kDa in the endoplasmic reticulum was not the result of a proteolytic cleavage at the site we had determined i n vitro. However, an alternative explanation would be that the two proteolytic cleavage products form a complex that remains intact during the immunoprecipitation but dissociates during the prepara- tion of the samples for SDS-polyacrylamide gel electropho- resis. The C-terminal cleavage product might easily have remained undetected because of its small size and its low content of threonine residues. To investigate whether such a complex indeed existed, the human breast carcinoma cell line ZR-75-1 was labeled with [3H]glucosamine, after which the cells were lysed in PBS containing 0.5% Nonidet P-40. To disrupt putative protein complexes prior to immunoprecipitation, the cell lysate was treated with either 6 M urea, 5% p- mercaptoethanol, or 1% SDS, as described under “Materials and Methods.” Immunoprecipitations were performed using either a monoclonal antibody directed against the repeats (139H2) or the polyclonal antiserum that recognizes the cytoplasmic domain (Fig. 4A). Comparable amounts of episialin could be immunoprecipitated with 139H2, irrespective of the treatment of the cell lysate. Using the anti-C-terminal antibody, episialin was immunoprecipitated after treatment with /3-mercaptoethanol and urea, but not after treatment with SDS. Apparently, a complex is formed between the extracellular domain and the remainder of the molecule containing the cytoplasmic domain, which is dissociated in the presence of 1% SDS but not in the presence of 6 M urea or p-mercaptoethanol. (The latter treatment was not expected to dissociate the complex, since the only 3 cysteine residues are located closely together in the transmembrane domain, mak- ing disulfide bonds between two possible cleavage products highly unlikely.) Unfortunately, in contrast to the in vitro translation experiments, it turned out to be difficult to effi- ciently label episialin in vivo with a radioactive amino acid, which should enable us to visualize the putative C-terminal cleavage product on a polyacrylamide gel. However, an inde-

139H2 anti4 min.0 20 40 60 90 180 0 20 40 60 90 180 180 180

mS- PL- I -200

Ps- i 1 2 3 4 5 6 7 8 9 10 11 12 1314

FIG. 3. Immunoprecipitation of episialin f rom T47D cells. T47D breast carcinoma cells were labeled with [’Hlthreonine for 20 min and chased for the times indicated. Immunoprecipitations were performed with monoclonal antibody 139H2 (lunes 1-6), the anti-C- terminal antiserum (lanes 7-12), normal mouse serum (lane 13), or normal rabbit serum (lane 14) . Immunoprecipitates were analyzed on a 4-10% SDS-polyacrylamide gradient gel. mL and mS indicate the 0-linked glycosylated (mature) products of the large and small alleles of T47D cells, respectively. p L and pS indicate the non-0-linked glycosylated precursors of these alleles. After a 0-min chase, this is a doublet of bands that differ by 20 kDa. The upper band of this doublet is converted to the lower band after prolonged incubation (compare e.g. lanes I and 7 with lanes 3 and 8) .

A 139H2 anti-C

C B S U C B S U

1 2 3 4 5 6 7 8

B 30 min 60 min

H B H* B* H B H* B* kDa

F-

N-

C-

-66

-43

-31

-22

1 2 3 4 5 6 7 8

FIG. 4. Complex formation of episialin in vivo a n d in vitro. A, ZR-75-1 breast carcinoma cells were labeled with [3H]glucosamine for 6 h. Cell lysates were prepared in PBS containing 0.5% Nonidet P-40. Immunoprecipitations were performed with either monoclonal antibody 139H2 (lanes 1-4) or the anti-C-terminal antiserum (lanes 5-8). Prior to immunoprecipitation, the samples were treated with 5% P-mercaptoethanol (0) (lanes 2 and 6) , 1% SDS ( S ) (lunes 3 and 7) , or 6 M urea (U) (lanes 4 and 8). The buffers containing urea and P-mercaptoethanol were replaced by PBS containing 0.5% Nonidet P-40, using a Sephadex G-25 column, and the sample containing SDS was diluted in this buffer to obtain a similar precipitation buffer containing 0.1% SDS. The treated cell lysates were split just before immunoprecipitation. Untreated cell lysates were used as a control (lanes I and 5 ) . The immunoprecipitates were analyzed on a 5% SDS-polyacrylamide gel. L and S , the mature forms of the large and small alleles, respectively. B, in oitro translation of full-length RNA (If) and RNAs derived from BulI-digested cDNA ( B ) were performed in the presence of [‘%]methionine for either 30 (lunes 1-4) or 60 (lanes 5-8) min. The products were analyzed on a 10-15% SDS- polyacrylamide gel either before (lunes 1, 2, 5, and 6 ) or after (lunes 3, 4 , 7, and 8) immunoprecipitation with the anti-C-terminal antiserum. The latter lanes are indicated with an asterisk. F, the full- length translation products; N a n d C indicate the N- and C-terminal cleavage products, respectively.

pendent confirmation for the presence of a complex in vivo is presented below.

To obtain additional evidence for the formation of a complex, we examined whether a similar complex could be formed in vitro. Therefore, we tested whether the N-terminal proteolytic cleavage product of the in vitro translated material could be immunoprecipitated by the anti-C-terminal antiserum. As demonstrated in Fig. 4B, not only the noncleaved translation product and the doublet of bands representing the C terminus but also the N-terminal cleavage product could be immunoprecipitated with the polyclonal antiserum directed against the cytoplasmic domain (lane 3 ) . This was neither caused by cleavage of the precursor after the immunoprecipitation was completed nor by association of the N-terminal cleavage product with the noncleaved precursor, since the N- terminal cleavage product was also immunoprecipitated from a sample that was almost completely cleaved at the start of the immunoprecipitation (lunes 5 and 7). To exclude that the N-terminal cleavage product was immunoprecipitated as a

Episialin Complex Formation 6175

result of cross-reactivity of the antiserum, an in vitro translation product that is encoded by a cDNA truncated at the BalI site and thus contains the entire extracellular domain but lacks the cytoplasmic domain (see Fig. 2 A ) , was made. As expected, this translation product could not be immunoprecipitated using the polyclonal antiserum directed against the cytoplasmic tail (lanes 4 and 8) . These results prove that the N-terminal cleavage product is immunoprecipitated by the antibody recognizing the cytoplasmic domain as a result of the formation of a complex between both cleavage products. Since this stable association is found both in vitro and in vivo, and since the interaction is expected to be dependent on the structure of the proteins near the cleavage site, it is likely that episialin is cleaved similarly in both systems.

The Proteolytic Cleavage Site Can Be Deleted from Episi- alin-To support our assumption that a similar cleavage occurs in vivo and in uitro, a region of 28 amino acids com- prising the cleavage site used in vitro was deleted in both the construct used for the in vitro translations containing only one repeat and in episialin cDNA containing the entire repeat region, which was used for transfection of HBL-100 cells (see Fig. 5 A ) . In contrast to the wild-type molecule, the in vitro translation product derived from the deletion mutant was not converted to a discrete smaller product, even after prolonged

A Tm

extra. ~ , . . .- . ~ I ., " ,. ',, , " . , . , ...t 1.: I1 intra

I t SSP( KPP!

WBd typ.: M F L ( 1 I Y K O G Q F L Q L S N I K E S ~ ~ ~ ~ Q l l N V

Mumnt 1: MFLW GTlNV

r " 7

0 C min. 20 40 60 -" W M

W M W M W M min. 0 15 0 15 C "

kDa " ~ . . ~.

F- -66

F. N'

N- -43

1 2 3 4 5 6 1 2 3 4 5

FIG. 5. Analysis of a deletion mutant lacking the putative cleavage site. A, schematic representation of the protein backbone of episialin containing only one repeat (black box). The region in which the proteolytic cleavage takes place is indicated by the Ssp1 and 3' KpnI sites. The amino acid sequence in the vicinity of this region of both the wild-type and the mutant molecules is shown. The arrows indicate the sites where the protein products terminate after cleavage of the cDNA at the Ssp1 and at the KpnI sites. The putative recognition sequences of the kallikreins are underlined. TM, extra, and intra, the transmembrane, extracellular, and intracellular domains, respectively. R, RNA derived from wild-type ( W ) or mutant (M) cDNA encoding variant A were translated in the presence of ['"SJmethionine. Translations were stopped with 100 p~ cycloheximide after 19.5 min. Samples of both the wild-type (lanes 1, 3, and .5) and the mutant (lanes 2,4, and 6 ) were analyzed after 20 (lanes 1 and 2), 40 (lanes 3 and 4 ) , and 60 (lanes 5 and 6 ) min on a 10% SDS- polyacrylamide gel. F, the full-length product. N , the N-terminal cleavage product. C, HBL-100 cells, transfected with wild-type or mutant episialin, were labeled for 5 min with ["Hlthreonine and lysed either directly (lanes 1 and 3) or after a 15-min chase period (lanes 2 and 4). Episialin was immunoprecipitated with monoclonal antibody 13982. As a control mutant, HC1.28 was labeled for 15 min, and precipitation was performed with normal mouse serum (C) (lane 5). The precipitates were analyzed on a 4-10% polyacrylamide gradient gel. Lanes 1 and 2, wild-type HCA1.3 ( W ) ; lanes 3 and 4, mutant HC1.28 (M). F, the noncleaved precursor. N , the proteolytically cleaved precursor.

incubation, as was to be expected following the removal of the specific cleavage site (compare Fig. 5R, lanes I, 3, and 5 and lanes 2, 4, and 6, respectively). The in vitro translated mutant molecule seems to be instable, since its amount is markedly reduced during the incubation period.

For the in vivo experiments, both wild-type and mutant episialin cDNAs containing the entire repeat region were cloned downstream of the cytomegalovirus immediate-early promoter in the vector pCMVIE-AK1-DHFR (Whang et al., 1987), of which the region containing the dihydrofolate reductase gene had been removed. Both constructs were transfected into the SV40-transformed human mammary epithelial cell line HBL-100. G418-resistant clones were analyzed for the expression of episialin by membrane immunofluorescence, and the presence of the deletion was confirmed by Southern blotting (results not shown). The biosynthesis of both wild- type and mutant episialin was studied in these transfectants. Cells were labeled with ["]threonine for 5 min and lysed directly or chased for an additional 15 min (Fig. 5C). The molecular mass of the wild-type molecule decreased with approximately 20 kDa during the 15-min chase period (compare lane 1 with lane 2). During this chase period, no shift in the molecular mass was observed for the mutant molecule (compare lane 3 with lane 4 ) . Both after the 0- and 15-min chase periods, a band of a relatively low intensity was observed with a molecular weight similar to the proteolytically cleaved wild-type molecule. However, this band varied in intensity among different experiments, and in contrast to what was observed for the wild-type molecule, the intensity of this band did not increase after the 15-min chase period. Therefore, it probably is the result of proteolytic degradation during cell lysis or immunoprecipitation, as has been observed before (Hilkens and Buijs, 1988). This indicates that in contrast to wild-type episialin, the mutant molecules are not proteolytically cleaved in the endoplasmic reticulum. The similar be- havior of the mutant molecules in vivo and in vitro implies that the proteolytic cleavage site in vivo is located at the same site or in close vicinity to the site used in vitro.

Evidence for the Association of the Cleavage Products in Vivo-To confirm that the wild-type molecules are present as complexes, whereas the entire extracellular domain is intact in the mutant molecules and is thus covalently linked to the cytoplasmic tail, a sandwich radioimmunoassay was performed. Either monoclonal antibody 115D8, directed against an epitope in the repetitive region, or the polyclonal antiserum recognizing the cytoplasmic domain were used as a catcher, and '2sI-labeled 115D8 was used as a tracer. Cell lysates of five different wild-type and mutant transfectants and of the parental cell line HBL-100 were tested either before or after boiling of the samples to dissociate existing complexes. The results of the assay are shown in Table I. Boiling of the antigen does not affect the epitope recognized by monoclonal antibody 115D8, since the binding of the antibody was similar before and after boiling of the antigen. The same holds true for the polyclonal antiserum, since the amount of mutant episialin detected before and after boiling of the samples is essentially the same. However, after boiling of the cell lysates containing wild-type episialin, hardly any molecules could be detected by the tracer 115D8 when the polyclonal antiserum was used as a catcher. We conclude that upon heat treatment the repeat region is detached from the cytoplasmic domain in wild-type episialin, whereas these regions remain associated in the mutant molecules that were shown to be resistant to the proteolytic cleavage. This confirms that the N- and C- terminal cleavage products of wild-type episialin, which are

6176 Episialin Complex Formation TABLE I

Amount of episialin in HBL-100 cells and their transfectants as determined with different sandwich radioimmunoassavs

115D8" Anti-C terminus"

Native Boiled b/nb Native Boiled b/nb countslmin' % counts/min' %

HBL-100 1,020 f 23 952 f 37 93 759 f 32 374 f 103 49 Wild-typed

HCA1.3 26,382 * 858 26,242 f 1,066 99 11,896 f 304 1,675 f 808 14 HCA1.5 37,599 f 746 38,093 f 221 101 22,289 f 1,007 1,040 f 49 5 HCA1.ll 4,703 -+ 690 4,174 f 273 89 3,215 f 290 421 f 181 13 HCA1.28 9,756 f 175 9,408 k 201 96 4,864 f 111 533 f 54 11 HCA2.7 14,440 f 431 14,389 f 809 100 6,380 f 263 508 f 104 8

HC1.6 9,096 f 327 8,576 f 375 94 5,679 f 241 5,316 f 380 94 HC1.9 11,362 f 322 10,658 f 437 93 5,620 f 141 5,562 f 110 99 HC1.18 31,169 f 1,190 26,715 f 927 86 8,496 f 716 9,467 f 1,100 112 HC1.27 37,776 f 1,612 31,226 f 116 83 15,168 f 1,078 14,416 k 766 95 HC1.28 37,629 f 847 30,411 f 530 81 14,020 f 426 15,758 f 535 112

Mutant

Either monoclonal antibody 115D8 or the anti-C-terminal antiserum was used to catch episialin from lysates. In both cases, 1261-labeled monoclonal antibody 115D8 was used as a tracer.

* b/n, boiled/native, ratio of episialin detected after and before boiling of the samples. ' n = 3 .

HBL-100 cells transfected with either wild-type or mutant episialin.

generated in the endoplasmic reticulum, are present as a complex.

DISCUSSION

The results of the in uitro translation assays and the analysis of the mutant episialin molecules demonstrate that episialin is synthesized as a transmembrane molecule, which is cleaved in the endoplasmic reticulum to generate two subunits that remain associated through what appear to be noncovalent interactions. The N-terminal subunit comprises the major part of the extracellular domain, including the entire repetitive region. The C-terminal domain is much smaller and consists of about 65 amino acids of the extracellular domain and the entire transmembrane and cytoplasmic domain. Since even the mature mucin-like domain is co-precipitated with the polyclonal antiserum directed against the C-terminal subunit (Fig. 3), the formation of the complex explains the association of this domain with the cell surface.

The in vitro translation product of episialin could be cleaved in rabbit reticulocyte lysates regardless of the addition of canine pancreatic microsomal vesicles (results not shown). Analysis of the translation products has revealed that the cleavage sites are identical or in close proximity in both translation systems. Therefore, the region which is affected by the cleavage should be vulnerable for proteases irrespective of structural differences, which are the result of N-linked glycosylation and membrane anchorage. Using truncated RNAs, we have mapped the proteolytic cleavage site to a region of less than 18 amino acids. Apparently, the cleavage site used in cell lines is also the same or in close proximity to the site at which the molecule is cleaved in the in vitro systems, since deletion of a region of 28 amino acids, including the in vitro cleavage site, abolishes the cleavage of the mutant molecule in HBL-100 cells.

The protease involved is not likely to be the same in all systems tested, since this would require a protease that is present both at the luminal and cytoplasmic sides of the membrane. Therefore, the processing should take place at a site that is recognized by various proteases or it should be the result of an autoproteolytic process. Proteolytic processing of proteins often occurs directly downstream of two consecutive basic amino acids (Sossin et al., 1989; Barr, 1991) or after the

second glycine in the sequence Gly-Gly-X, where X is often an amino acid residue with a hydrophobic side chain (L6pez- 0th et al., 1989). No such sequences are present in the region to which the cleavage has been mapped in vitro, but the latter motif is found close to this region upstream of the position, which corresponds to the Ssp1 site in the cDNA. This putative recognition site has been deleted in the mutant that has proven to be resistant to the cleavage in uiuo. Therefore, we can not formally exclude that this site is used in uiuo, although this would imply that the in vitro and in uivo cleavage sites are not identical. In the region to which the cleavage has been mapped in both systems, two Phe-Arg dipeptides are present, which are putative substrates for a family of serine proteases, the kallikreins (Fiedler et al., 1987). Cleavage at these sites could explain the presence of a doublet of bands representing the C-terminal cleavage product in the in uitro translation experiments. Amino acid sequencing of the N terminus of the C-terminal subunit should elucidate the exact cleavage site(s) and could give us a key to the identification of the protease@) responsible for the cleavage(s).

The stable association of both cleavage products implies that an additional step is needed to release episialin from the cell surface, as is observed both in vitro and in viuo. This can either be the result of a second proteolytic cleavage of cell surface episialin or of the dissociation of the complex. In the latter case, transfectants expressing the mutant that is not cleaved would be expected to release hardly any episialin. However, these transfectants were found to release at least as much episialin as those expressing the wild-type form (results not shown). This might reflect the reduced stability of the mutant molecule, which we have observed in the in uitro translation system and which is probably caused by an alter- ation of the three-dimensional structure of the mutant. It is also conceivable that a second cleavage is involved in the release of the molecule, since we have evidence that cell surface-associated episialin can be internalized and cycled back to the cell surface.' During this recycling process, the molecule is likely to pass compartments that are rich in proteases. These proteases might affect the region close to the transmembrane domain, which contains only a few car- bohydrate side chains. The remainder of the extracellular

S. V. Litvinov and J. Hilkens, submitted for publication.

Episialin Complex Formation 6177

domain is expected to be less sensitive to proteases, since its protein backbone is protected by closely packed glycans. Upon arrival at the cell surface, this part of the molecule might be released, because it is no longer anchored to the membrane.

For many other molecules, it is known that two subunits remain associated upon proteolytic cleavage of a common precursor. However, most of these molecules, such as some of the a-chains of the integrins (for review, see Hynes (1987)), several proteases and blood clotting factors, and the receptors for insulin (Deutsch et al., 1983; Hedo et al., 1983; Ronnett et al., 1984) and insulin-like growth factor-I (Jacobs et al., 1983), are linked by disulfide bonds. Like episialin, low density lipoprotein receptor-related protein (LRP) (Herz et aZ., 1990) and the ascites sialoglycoprotein (ASGP) complex (Sheng et aZ., 1990) consist of two subunits that are derived from a common precursor and that are associated through noncovalent interactions. The cleavage of LRP was shown to occur after the molecule had reached the Golgi complex. Sequencing of the N terminus of its C-terminal subunit has established that the cleavage takes place immediately downstream of 2 consecutive arginine residues. This suggests that the proteases involved in the cleavage of LRP and episialin are different. Ascites sialoglycoprotein-1, the N-terminal part of the ASGP- precursor, has many biochemical properties in common with episialin. The molecule, which was characterized in 13762 rat ascites mammary adenocarcinoma cells, has a high molecular mass (about 600 kDa) as a result of extensive 0-linked glycosylation and sialylation. The molecule forms a complex with ASGP-2, a membrane-spanning molecule of 120 kDa carrying many N-linked sugars. As has been established for episialin (Hilkens and Buijs, 1988), both the common precursor of ASGP-1 and -2 and the earliest detectable form of ASGP-2 carry N-linked sugars of the high mannose type. This suggests that the ASGP precursor is cleaved before it reaches the medial Golgi, where conversion of high mannose to complex oligosaccharides might occur. Why LRP, ASGP, and episialin are proteolytically processed is not known. Both ASGP-1 and episialin are eventually released from the cell surface, which might be caused by dissociation of the complex. Identification of the function of the released episialin molecules might reveal the logic behind the formation of this unusual complex.

Acknowledgments-We thank Drs. H. L. Vos, G. F. M. Verheijden, and H. L. Ploegh for helpful discussions and critical reading of the manuscript.

REFERENCES Abe, M., and Kufe, D. (1989) Cancer Res. 49, 2834-2839 Barr, P. J. (1991) Cell 66, 1-3 Deutsch, P. J., Wan, C. F., Rosen, 0. M., and Rubin, C. S. (1983)

Fiedler, F. (1987) Eur. J. Biochem. 163, 303-312 Gendler, S. J., Lancaster, C. A., Taylor-Papadimitriou, J., Duhig, T.,

D. (1990) J . Bid. Chem. 265, 15286-15293 Peat, N., Burchell, J., Pemberton, L., Lalani, E.-N., and Wilson,

Hedo, J. A., Kahn, C. R., Hayashi, M., Yamada, K. M., and Kasuga, M. (1983) J. Biol. Chem. 258 , 10020-10026

Herz, J., Kowal, R. C., Goldstein, J. L., and Brown, M. S. (1990)

Hilkens, J., and Buijs, F. (1988) J. Biol. Chem. 263, 4215-4222 Hilkens, J., Buijs, F., Hilgers, J., Hageman, P., Calafat, J., Sonnen-

berg, A., and van der Valk, M. (1984) Znt. J . Cancer 34, 197-206 Hilkens, J., Kroezen, V., Bonfrer, J. M. G., De Jong-Bakker, M., and

Bruning, P. F. (1986) Cancer Res. 46, 2582-2587 Hynes, R. 0. (1987) Cell 48,549-554 Jacobs, S., Kull, F. C., and Cuatrecasas, P. (1983) Proc. Natl. Acad.

Laemmli, U. K. (1970) Nature 227,680-685 Lan, M. S., Batra, S. K., Qi, W.-N., Metzgar, R. S., and Hollingsworth,

Ligtenberg, M. J. L., Vos, H. L., Gennissen, A. M. C., and Hilkens,

Ligtenberg, M. J. L., Gennissen, A. M. C., Vos, H. L., and Hilkens,

Ligtenberg, M. J. L., Buijs, F., Vos, H. L., and Hilkens, J . (1992)

Linsley, P. S., Kallestad, J . C, and Horn, D. (1988) J. Biol. Chem.

Lbpez-Otin, C., Simbn-Mateo, C., Martinez, L., and Viiiuela, E.

Ronnett, G. V., Knutson, V. P., Kohanski, R. A., Simpson, T. L., and

Schagger, H., and von Jagow, G. (1987) Anal. Biochem. 166, 368-

Proc. Natl. Acad. Sci. U. S. A. 80, 133-136

EMBO J. 9 , 1769-1776

Sci. U. S. A. 80,1228-1231

M. A. (1990) J. Biol. Chem. 265, 15294-15299

J. (1990) J. Bi01. Chem. 265, 5573-5578

J. (1991) Nucleic Acids Res. 19, 297-301

Cancer Res., in press

263,8390-8397

(1989) J. Biol. Chem. 264,9107-9110

Lane, M. D. (1984) J. Biol. Chem. 259, 4566-4575

379 Sheng, Z., Hull, S. R., and Carraway, K. L. (1990) J. Bid. Chem.

265,8505-8510 Sossin, W. S., Fisher, J. M., and Scheller, R. H. (1989) Neuron 2,

Whang, Y., Silberklang, M., Morgan, A., Munshi, S., Lenny, A. B., Ellis, R. W., and Kieff, E. (1987) J. Virol. 61, 1796-1807

Wreschner, D. H., Hareuveni, M., Tsarfaty, I., Smorodinsky, N., Horev, J., Zaretsky, J., Kotkes, P., Weiss, M., Lathe, R., Dion, A., and Keydar, I. (1990) Eur. J . Biochem. 189.463-473

1407-1417

cell-associated episialin is a complex containing two proteins

Documents