Membrane-bound Versus Secreted Forms of Human Asialoglycoprotein Receptor Subunits ROLE OF A JUXTAMEMBRANE PENTAPEPTIDE*

The H2a alternatively spliced variant of the human asialoglycoprotein receptor H2 subunit differs from the H2b variant by an extra pentapeptide, EGHRG, present in the ectodomain next to the membrane-span. This difference causes retention and degradation in the endo- plasmic reticulum (ER) of H2a when expressed without the H1 subunit in 3T3 cells (1). In contrast, a significant portion of singly expressed H2b is Golgi-processed and reaches the cell surface. Using a new specific anti-H2a antibody, we found that in HepG2 cells, H2a is rapidly cleaved to a 35-kDa fragment, comprising the entire ectodomain, most of which is secreted into the medium. The cleavage site for the secreted fragment was located at the lumenal end of the membrane span. No mem-brane-bound H2a exits the ER, indicating that the pen- tapeptide is a signal for ER retention and degradation of the membrane form but does not hinder secretion of the cleaved soluble form. H2a does not form a membrane receptor complex with H1 as H2b does. H2a is therefore not a subunit of the receptor but a precursor for a secreted form of the protein; signal peptidase is probably responsible for the cleavage to the soluble fragment (2). Therefore, the juxtamembrane sequence regulates the function of the transmembrane domain of

In recent years the concept of a unique pair consisting of a membrane receptor with its soluble ligand has been modified by the discovery of soluble forms of receptors and membranebound forms of ligands, i.e. soluble tumor necrosis factor receptor (3) and membrane-bound tumor necrosis factor (4). The formation of normal and aberrant soluble forms of a transmembrane protein is the basis for Alzheimer's disease, where the amyloid precursor protein is cleaved and releases a soluble peptide, which is secreted and may form ␤-amyloid deposits (5). In general little is known about the mechanism and regulation of the formation and function of the soluble forms of transmembrane receptors. The existence of a soluble form of a human asialoglycoprotein receptor (ASGPR) 1 subunit and its mechanism of formation from an alternatively spliced variant, which we report here, may help shed some light on these matters.
The human ASGPR is exclusively expressed in hepatocytes. It is constructed of two subunits of related amino acid sequence, H1 (46 kDa) and H2 (50 kDa) (6). The polypeptide chains span the membrane once in a type II orientation with a large carboxyl-terminal ectodomain and a small cytoplasmic tail. The transmembrane segment is an uncleaved signal anchor sequence (7). Both H1 and H2 undergo phosphorylation on serine or threonine residues and are N-glycosylated. H1 and H2 form a hetero-oligomeric complex. Expression of both H1 and H2 cDNAs is required to generate high affinity asialoglycoprotein binding sites (8), although each subunit has a binding pocket and can bind to a galactosylated matrix (9). An average subunit ratio of 3:1, H1:H2 is found on the surface of HepG2 cells (10).
We have described the different subcellular fates of two alternatively spliced variants of the H2 subunit of the ASGPR, H2a and H2b, when expressed in NIH 3T3 cells without H1. These variants differ only by the presence in H2a of an extra 5-amino acid insert in the ectodomain adjacent to the membrane spanning region (1). Without coexpression of the H1 subunit, H2a was inserted normally into the ER membrane but was retained in and degraded in the ER or another pre-Golgi compartment; none reached the cell surface (11,12). In contrast, a significant portion of the alternatively spliced form H2b, expressed without H1, was processed by Golgi enzymes and reached the cell surface. The sole difference determining the different subcellular localization of the two variants is the 5-amino acid insert in H2a (EGHRG) (1). Other alternatively spliced forms of H2a and H2b, lacking an 18-amino acid stretch in the cytoplasmic tail, have been described; their fate is similar to that of H2a and H2b, respectively, stressing the importance of the 5-amino acid membrane-adjacent segment in determining subcellular fate (13). Similar mechanisms have been described for other membrane proteins that undergo rapid degradation in the ER. In the case of unassembled ␣ subunit of the T-cell antigen receptor, specific charged residues in the membrane domain are determinants for ER retention and degradation (14). Similarly, the juxtamembrane charged residues in the extra pentapeptide of H2a could also be involved in ER retention and degradation of the singly expressed subunit.
In transfected 3T3 cells, H2a, after insertion into the ER membrane, is cleaved just exoplasmic to the membrane-spanning domain, in the region of the 5 extra amino acids, releasing a 35-kDa soluble fragment comprising the entire ectodomain. Our initial studies suggested that this cleavage might be an obligatory first step in ER degradation of the polypeptide (11,15). However, we recently showed that ER degradation can occur through at least two pathways, one of them not involving this cleavage in the juxtamembrane domain (2).
In HepG2 cells H2a represents about 8% of H2 mRNAs (1). In order to study the H2a and H2b polypeptides in HepG2 cells, we have developed specific anti-peptide antibodies against the region of the extra five amino acids in H2a that are unique to this variant. Using these antibodies we show here that in HepG2 cells the ectodomain of H2a, cleaved off as a 35-kDa fragment in the ER, is then efficiently secreted; the fragment is also partially secreted from transfected 3T3 cells, although in this case most of it undergoes ER degradation, as we described previously (1).
Secreted forms have been found recently for many membrane receptors (16). In some cases cleavage of the wild type receptor occurs at the cell surface (17). In others alternative splicing creates truncated mRNAs encoding proteins without the transmembrane domains that are secreted (18). In the case of the ASGPR H2a, alternative splicing creates a transcript containing an extra mini-exon, which plays a role in the cleavage and secretion of the ectodomain as we describe here. This result strengthens the theory that signal anchor domains on type II transmembrane proteins are functionally equivalent to signal peptides (19,20).
Antibodies-Antibodies specific for peptides corresponding to the carboxyl termini of H1 or H2 were the ones used in earlier studies (1,2,10). The "anti-H2a" antibody was raised in rabbits against a synthetic peptide including 18 residues from the juxtamembrane ectodomain of H2a: CVTGSQSEGHRGAQLQAE. For immunizing the rabbits, the peptide was coupled to the carrier protein keyhole limpet hemocyanin with sulfosuccinimidyl 4-(N-maleimidomethyl)cyclohexane-1-carboxylate (9).
Metabolic Labeling and Immunoprecipitation-Subconfluent (90%) monolayers of cells in 60-mm tissue culture dishes were rinsed and preincubated for 30 min at 37°C with cysteine-free DMEM plus 10% dialyzed calf serum. They were then pulse-labeled in the same medium containing [ 35 S]cysteine (0.3 mCi/ml) for different periods (usually 30 min). They were then rinsed and chased for different periods of time with normal DMEM plus 10% fetal calf serum as described (1,15). To label with [ 3 H]leucine, the same procedure was followed except 100-mm diameter culture dishes and leucine-free DMEM were used. The cells were rinsed with PBS and lysed in nondenaturing conditions except where indicated. For nondenaturing conditions, cells were solubilized with 1% Triton X-100 and 0.5% sodium deoxycholate in PBS in the presence of 2 mM PMSF. Denaturing conditions involved boiling in 1% SDS and 2 mM dithiothreitol in PBS, for 3 min and then adding 10 volumes of 1% Triton X-100 plus 0.5% sodium deoxycholate containing 2 mM oxidized glutathione and 2 mM PMSF. Immunoprecipitations from cell lysates and endo H treatments were performed as described before (1,11). Immunoprecipitations from cell supernatants (1.2 ml from 60-mm dishes) were done directly by addition of 2 mM PMSF, protein A-Sepharose, and the appropriate antiserum followed by incubation at 4°C with rotation for 4 -16 h. For treatment with N-glycanase, immunoprecipitates were washed and then boiled in 10 l of 0.5% SDS in 50 mM sodium citrate, pH 6.0. Then, 10 l of a solution containing 200 mM sodium phosphate, pH 8.0, 40 mM EDTA, pH 8.0, 3% N-octylglucoside was added, together with 40 milliunits of N-glycanase, and incubations were carried out overnight at 37°C. Twenty l of sample buffer were added and the samples boiled before loading for SDS-PAGE.
Cell Surface Biotinylation and Detection of Biotinylated Proteins-Subconfluent cell monolayers in 90 mm tissue culture dishes were rinsed twice in PBS-CM (PBS containing Ca ϩ2 and Mg ϩ2 ) and incubated in a fresh solution of 0.5 mg/ml sulfosuccinimidyl-6-(biotinamido)hexanoate in PBS-CM for 45 min at 4°C with rocking. The reaction was quenched by removing the biotin solution and incubating the cells with 50 mM NH 4 Cl in PBS-CM for 10 min at 4°C. The cells were then rinsed twice with cold PBS-CM and lysed in denaturing conditions, followed by immunoprecipitation and SDS-PAGE. The proteins were transferred to a nitrocellulose membrane. Blocking was done by incubating the membrane in TGG (Tris-buffered saline plus 10% glycerol, 1 M glucose and 0.5% Tween 20) containing 1% milk and 3% bovine serum albumin for 60 min at 4°C. Then the membrane was incubated in TGG containing 10 g/ml streptavidin conjugated to peroxidase for 60 min at room temperature. Three washes were done in Tris-buffered saline containing 0.5% Tween 20, and detection was performed by the ECL procedure as instructed in the kit from Amersham.
Amino-terminal Radiosequencing-The secreted H2a fragment was subjected to limited NH 2 -terminal protein sequencing as described before (2) using the technique of Matsudaira (21). Briefly, medium from cells metabolically labeled with [ 3 H]leucine was immunoprecipitated and the products run on SDS-PAGE and then electroblotted onto Immobilon-P paper. Protein bands were located on the paper by autoradiography and then excised for automated Edman degradation with an Applied Biosystem Inc. (Foster City, CA) model 470A protein sequencer. The products of each reaction cycle were quantitated in a liquid scintillation counter.

The H2a Ectodomain Is Secreted from HepG2 Cells and From
Transfected 3T3 Cells-In singly transfected 3T3 cells (without H1), H2a is completely retained in and degraded in the ER or a closely related pre-Golgi compartment (1). Both H2a and H2b produce 35-kDa fragments corresponding to their ectodomain (2,11). In a pulse-chase experiment with [ 35 S] cysteine ( Fig. 1), cells expressing H2a produce more of the fragment (immunoprecipitates with anti-H2 carboxyl-terminal antibody) than cells expressing H2b. At its peak level (1 h of chase), there is 2.5-fold more H2a fragment than H2b fragment as compared to the amount of pulse-labeled precursors. The intracellular fragments then disappear, and we had presumed that they were entirely degraded. Nevertheless, the transient higher abundance of the H2a 35-kDa fragment prompted us to study its fate in HepG2 cells.
In HepG2 cells H2a represents about 8% of total H2 RNAs. In order to study the H2a and H2b polypeptides in these cells, we developed specific anti-peptide antibodies against the region containing the extra 5 amino acids in H2a (as detailed under "Antibodies"). Fig. 2 shows a pulse-chase experiment followed by immunoprecipitation from cell lysates or from cell supernatants. HepG2 cells were compared to 3T3 cells express-ing H2a (cell line [2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18] or H2b (cell line 2C). Even though the anti-H2a antibodies were prepared against a peptide (H2a peptide) containing the extra 5 amino acids in H2a plus 5 on each side flanking them, (the latter shared by H2a and H2b), they reacted specifically against H2a and did not recognize H2b ( Fig As can be seen in Fig. 2 as well as in other figures, there is a considerable heterogeneity in the SDS-PAGE profiles of both 35-kDa fragment (bands from 28 kDa to 35 kDa), and of the membrane-attached precursor (bands around 42 kDa). This is due to variable degree of occupation of the three possible Nglycosylation sites, as all of these species are converted to two tight bands at 37 kDa (precursor) and 28 kDa (fragment) after treatment with endo H or N-glycanase (Fig. 2, lanes 8 and 9). The two deglycosylated precursor and two fragment bands present in HepG2 cells that are immunoreactive with the anti-H2-COOH antibody (B, lane 3) may represent the respective H2a and H2b species.
In HepG2 cells the H2a polypeptide was found mainly in the form of the cleaved 35-kDa fragment. After a 30-min pulse and 1 h of chase ( Fig. 2A, lane 1), no full-length precursor H2a was found in HepG2 cells compared to its presence in transfected 3T3 cells (2-18 cells) ( Fig. 2A, lane 7). Even after short pulses, only traces of precursor could be seen in HepG2 cells (data not shown), suggesting a cotranslational or very fast postranslational cleavage. After 6 h of chase, labeled H2a fragment was secreted into the medium, both from HepG2 cells and from 2-18 cells ( Fig. 2A, lanes 4 and 10). While the intracellular H2a fragment was sensitive to endo H, indicating a pre-Golgi localization ( Fig. 2A, lanes 2 and 8), the secreted fragment was completely resistant to endo H ( Fig. 2A, lanes 5 and 11) but sensitive to N-glycanase ( Fig. 2A, lanes 6 and 12), indicating that it had undergone Golgi processing to attain complex type oligosaccharides. The Golgi processed fragment could never be detected within the cells, suggesting rapid secretion after exit from the ER.
The Carboxyl Terminus of the H2a Fragment Is Modified before Secretion-Only a tiny fraction of the secreted H2a fragment can be immunoprecipitated by the anti-H2 COOH-terminal antibody (Fig. 2B, lanes 4 and 10), which would explain why we had not identified the fragment in the culture medium previously (1,2,11). It was only after N-glycanase treatment that a clear compact band of the secreted fragment could be observed in immunoprecipitates of the medium using the anti-H2-COOH antibody (Fig. 2B, lane 12). Recognition of the H2a fragment by both the anti-H2-COOH antibody and the anti-H2a antibody would indicate that it comprises the whole ectodomain. However, the weak recognition by the anti-H2-COOH antibody would suggest that a carboxyl-terminal modification occurs to the fragment before secretion, as the intracellular form of the fragment is well recognized by this antibody. The intracellular fragment formed in 3T3 cells expressing H2b (Fig. 2B, lanes 14 and 15) could not be detected in the medium (Fig. 2B, lane 18). Although a similar carboxylterminal modification could be occurring in the case of H2b, leading to poor immunoprecipitation by the H2-COOH antibody, no secreted form was detected in repeated pulse-chase experiments even after overexposing the autoradiograms (data not shown).
To determine if this carboxyl-terminal modification was occurring early in the secretory pathway, we pulse-labeled HepG2 cells for 30 min. Cell lysates were first immunoprecipitated with the anti-H2-COOH antibody. The supernatant was then re-immunoprecipitated with the anti-H2-COOH antibody. The supernatant from this second step was then immunoprecipitated with anti-H2a antibody. A strong band is seen at around 35 kDa (Fig. 4A, lanes 3 and 6), which shows that much of the ER (endo H-sensitive) fragment is not recognized by the anti-carboxyl-terminal antibody. In the case of 2-18 cells, a portion of the precursor species at 42 kDa is not recognized by the anti-carboxyl-terminal antibody, but it is immunoprecipitated by the anti-H2a antibody. This could indicate that in these cells, some carboxyl-terminal modification occurs even before the cleavage of precursor H2a to the 35-kDa fragment.
On the other hand, significant fractions of H2a 42-kDa precursor and 35-kDa fragment are recognized by both antibodies, as seen by immunoprecipitating with anti-H2a, boiling the pellet, and then re-immunoprecipitating with anti-H2-COOH (Fig. 4B). This indicates that, at least for some of the H2a polypeptides, the carboxyl-terminal modification occurs after the cleavage to the 35-kDa fragment.
The Cleavage Site to Form the Secreted H2a 35-kDa Frag- ment Is at the Boundary of the Transmembrane Domain-The recognition of the secreted H2a fragment by the anti-H2a antibody indicated that it contains the 5-amino acid H2a pentapeptide and therefore that the cleavage site must be near the 3 or 4 amino acids located between the transmembrane domain and the pentapeptide. In order to determine the amino terminus of the H2a fragment secreted from HepG2 cells, the cells were labeled with [ 3 H]leucine for 30 min and chased with unlabeled medium for 4 h. The medium was immunoprecipitated with the anti-H2a antibody, treated with N-glycanase, run on SDS-PAGE, and blotted onto Immobilon-P paper. The band corresponding to the deglycosylated H2a fragment (28 kDa) was excised and subjected to microsequencing (Fig. 5). The appearance of radiolabel after cycles 11 and 15 indicates  5 and 10) of the 18-amino acid peptide that was used to produce the anti-peptide antibodies and includes the extra 5 amino acids in H2a (H2a peptide), or alternatively in the presence of 1 g (lanes 2 and 7) or 10 g (lanes 3 and 8) of a 13-amino acid peptide that includes the same boundaries but does not have the extra 5 amino acids in H2a (H2b peptide). The immunoprecipitates were analyzed by SDS-PAGE followed by fluorography. On the right are molecular masses of protein standards in kilodaltons. the presence of leucine residues at positions 11 and 15 from the amino terminus. This would correspond to a cleavage between Gly-78 and Ser-79, at the boundary of the transmembrane domain. This is one of the two cleavage sites that we had identified for formation of intracellular H2a 35-kDa fragments in transfected 3T3 cells (2). The second cleavage site (after Ser 79) that also gives rise to intracellular 35-kDa fragments could also be present in some of the secreted species but might not be detected due to its low abundance.
H2a Is Rapidly Cleaved to Form the 35-kDa Fragment in HepG2 Cells-To quantify the secretion of the H2a 35-kDa fragment from HepG2 cells and from the transfected 2-18 cells, we labeled them for 30 min with [ 35 S]cysteine followed by different times of chase, immunoprecipitation, treatment with N-glycanase, and SDS-PAGE. The autoradiograms of the gels were quantified by densitometry (Fig. 6). The maximum amount of secreted fragment from HepG2 cells (after an 8-h chase) corresponded to 68% of the maximum fragment found intracellularly (after a 1-h chase). The efficiency of H2a secretion is similar to the efficiency of H1 maturation to the cell surface (8). In transfected 3T3 cells, only 40% of the maximum amount of fragment found intracellularly (after a 1-h chase) was secreted (after a 6-h chase). If we compare the maximum amount of secreted H2a fragment (after a 6-h chase) to the amount of intracellular 35-kDa fragment plus 42-kDa precursor present after the pulse, the efficiency of secretion would be 26%.
Since no 42-kDa precursor of H2a was observed in HepG2 cells, we assumed that a rapid co-translational cleavage was taking place to form the 35-kDa fragment. This difference in the rate of cleavage of H2a in HepG2 cells as compared to single expression in 3T3 cells could be due to the effect of a putative initial assembly of H1 and H2a when the former is present. To test this hypothesis, we prepared a doubly transfected 3T3 cell line expressing both H1 and H2a. As seen in Fig. 7, coexpression of H1 in 3T3 cells did not alter the fate of H2a; H2a precursor is seen after the pulse (Fig. 7A, lanes 6 and 7) and disappears at a similar rate to the singly expressed polypeptide (Fig. 7A, lanes 7-10 compared to 13-14, and Fig. 1). Coexpression of H1 does not hinder immunoprecipitation of H2a precursor by the anti-H2a antibody, as similar amounts of H2a are recovered in denaturing or nondenaturing conditions (Fig. 7A,  lanes 6 and 7). There might be coprecipitation of only a minor fraction of H1 with H2a (Fig. 7A, lane 6), although the band at ϳ40 kDa could also be an underglycosylated form of H2a. A protein of ϳ50 kDa also coprecipitates with H2a in nondenaturing conditions (Fig. 7A, lane 6, and Fig. 1, lane 2). The results were identical when the experiment was performed  , lanes 1, 4, and 7). The supernatants were re-immunoprecipitated with anti-H2-COOH antibodies (treatment B, lanes 2, 5, and 8), and the supernatants from this second immunoprecipitation were immunoprecipitated with anti-H2a antibodies (treatment C, lanes 3, 6, and 9). All antibodies were cross-linked to protein A-Sepharose as described under "Experimental Procedures." The immunoprecipitates were analyzed by SDS-PAGE followed by fluorography. On the right are molecular masses of protein standards in kilodaltons. B, some of the H2a molecules can be recognized by both anti-H2a and anti-carboxylterminal antibodies. The same cell lines as in A were metabolically labeled with [ 35 S]cysteine for 30 min. The cells were then lysed and immunoprecipitated with anti-H2a antibodies. The immunoprecipitates were then boiled with 1% SDS and 2 mM dithiothreitol in PBS, after which 10 volumes of 1% Triton X-100 plus 0.5% sodium deoxycholate containing 2 mM oxidized glutathione were added and the samples re-immunoprecipitated with anti-H2-COOH antibodies (lanes 1-3). The same procedure was followed but with anti-H2-COOH antibodies first and then anti-H2a antibodies (lanes 4 -6). The immunoprecipitates were analyzed by SDS-PAGE followed by fluorography. On the right are molecular masses of protein standards in kilodaltons. Proteins were electroblotted onto Immobilon-P paper and the 28-kDa band was located by autoradiography. The band on the paper was then cut out for NH 2 -terminal protein sequencing and the radioactivity from each cycle of the Edman degradation was determined by liquid scintillation counting. On top of the figure, the deduced cleavage site in the sequence of H2a has been depicted, outlining the radioactive leucine residues. using anti-H2 carboxyl-terminal antibody instead of anti-H2a (data not shown). As can be seen in panel B, secretion of H2a is also unaffected by coexpression of H1 in 3T3 cells as a similar amount of fragment (relative to precursor after the pulse) can be seen secreted from cells expressing H1 and H2a or H2a alone (Fig. 7B, lanes 5 and 6 compared to lanes 9 and 10). No secreted product is detected with the anti-H1 carboxyl-terminal antibody (Fig. 7B, lanes 7 and 8).
H2a Is Expressed Neither on the Surface of HepG2 cells Nor on That of Transfected 3T3 Cells-No H2a had been found at the surface of singly transfected 3T3 cells (1). To determine if membrane-bound H2a could be found at the surface of HepG2 cells or if any H2a fragment could form part of the cell surface receptor complex, we biotinylated surface proteins of HepG2 cells or the transfected cell lines with the membrane-impermeant reagent sulfosuccinimidyl-6-(biotinamido)hexanoate at 4°C. Cell lysates were then prepared in denaturing conditions and immunoprecipitated. The immunoprecipitates were run on SDS-PAGE and blotted onto nitrocellulose. The blot was reacted with streptavidin-peroxidase and revealed surface expression of H1 in HepG2 cells and 3T3 cells expressing H1 plus H2a (Fig. 8, lanes 1 and 8). H2b surface expression was also revealed in 3T3 cells expressing H2b and on HepG2 cells (Fig.  8, lanes 3-5). In contrast H2a surface expression was absent in all the cell lines (Fig. 8, lanes 2, 6, and 9). Intracellular H2a could be detected if the biotinylation was performed after the cell lysis (data not shown). In the denaturing conditions that were used to prepare the lysates, all H2b was recovered in HepG2 cells after immunoprecipitation of H1, indicating that no complexes remained (Fig. 8, compare lanes 3 and 4). To rule out the possibility that any trace amounts of H2a were on the surface, another experiment was performed in which we surface-labeled HepG2 cells at 4°C with the membrane-impermeable Bolton-Hunter reagent 125 I-sulfo-SHPP. Upon immunoprecipitation of cell lysates from these cells with anti-H2a antibodies, no cell surface H2a could be detected, although the same protocol readily detected cell surface H2b (data not shown). We conclude that no membrane-attached H2a reaches the cell surface, although the soluble ectodomain is freely secreted.

DISCUSSION
The ASGPR H2a subunit has been used as a model for studies on ER degradation, due to the fact that when the polypeptide was expressed in transfected 3T3 cells in the ab-sence of the H1 subunit it was retained in the ER, in contrast to its alternatively spliced variant H2b (1, 2, 11). We have now studied the fate of H2a in HepG2 cells with the help of a new antibody that discriminates between H2a and H2b. The amount of H2a polypeptide in HepG2 cells, primarily as the soluble 35-kDa fragment, is comparable to that of H2b (Fig. 4A,  lanes 1-3). In contrast, H2b mRNA is about 10 times more abundant than H2a mRNA (1). This would suggest different efficiencies in the translation of H2a and H2b mRNAs, assuming that the immunoprecipitation efficiency was similar for the two antibodies. In HepG2 cells the majority of the H2a polypeptides were recovered as 35-kDa endo H-sensitive fragments, which corresponds to the ectodomain of the polypeptide ( Fig.  2A, lanes 1-3; Fig. 9). Only trace amounts of the H2a precursor can be seen after pulse labeling in HepG2 cells, compared to the abundant amount of H2a 42-kDa precursor seen in transfected 3T3 cells and detected with the same antibody ( Figs. 2A and  7A). This suggests that in HepG2 cells the H2a 42-kDa precursor is cleaved cotranslationally or very soon after translation. This is consistent with efficient cleavage by signal peptidase (2). To our surprise we found that the cleaved 35-kDa ectodomain of H2a is efficiently secreted in the form of a 42-45-kDa Golgi processed polypeptide (Figs. 2 and 6). The fact that the fragment is secreted indicates that the endoproteolytic cleavage does not by itself determine further degradation in the ER. Assuming that the intracellular fragment seen is formed efficiently from its 42-kDa precursor (Fig. 4A), the secretion of H2a from HepG2 cells is as efficient as the extent of surface expression of newly made H1 or H2b: about 70% (Fig. 6). In contrast, the maximal extent of secretion of the 35-kDa H2a fragment from transfected fibroblasts (ϳ30%) is relatively inefficient, as is the surface expression of newly made H1 or H2b polypeptides synthesized in singly transfected 3T3 cells (40% and 30%, respectively; Refs. 1 and 8). This could suggest that hetero-oligomers of H1/H2a precursors initially form in the ER in HepG2 cells and that formation of these hetero-oligomers generates some protection from ER degradation. There are many examples of proteins that become targeted to ER degradation if they cannot assemble into normal oligomers. The immunoglobulin light chain is secreted only when the heavy chain is coexpressed (22). IgM is secreted only when its subunits are correctly assembled, while monomeric polypeptides are targeted for ER degradation (23,24). Nevertheless, coexpression of H1 did not lead to a faster cleavage or more efficient secretion of H2a fragment from 3T3 cells (Fig. 7). Metabolic labeling and immunoprecipitation with anti-H1 antibodies of cell lysates or supernatants showed no co-immunoprecipitation of H2a. Only a minor fraction of precursor H1 might co-immunoprecipitate with anti-H2a antibodies (Fig. 7A). Thus the differences in the efficiency of H2a cleavage and secretion between HepG2 cells and transfected 3T3 cells appear to be only related to the difference in cell type.
Our new specific anti-H2a antibody recognizes H2a and not H2b (Fig. 2). The recognition of the H2a secreted fragment by both the anti-H2-COOH antibody and the anti-H2a antibody (Fig. 2) and the identification of its amino terminus (Fig. 5) would indicate that it comprises the entire H2a ectodomain. However, the weak recognition by the anti-H2-COOH antibody suggests a carboxyl-terminal modification of the 35-kDa fragment before it is secreted (Figs. 4A and 9). This modification could involve conformational changes or proteolytic trimming, although there is no apparent change in the molecular mass of the deglycosylated fragment upon its secretion. This COOHterminal modification explains why we could not detect any secreted fragment before we had generated the "anti-H2a" antibody. It could be argued that ectodomains from H2b molecules are in fact secreted but cannot be recognized by the anti-carboxyl-terminal antibody because the same COOH-terminal modification has taken place. Due to the smaller amount of intracellular H2b fragment, even if it is secreted, the extent of secretion (relative to the amounts of 42-kDa precursors, Fig.  1) would be much less than for H2a.
H2a never leaves the ER as an intact membrane-bound protein, whether or not H1 is coexpressed. Similarly, H2a cannot be detected on the cell surface, whether or not H1 is coexpressed (Fig. 8). We conclude that H2a is not a subunit of the asialoglycoprotein receptor complex. The only function of the 42-kDa membrane-associated H2a precursor seems to be the generation of the soluble secreted fragment. The fact that membrane-bound H2a does not exit the ER, even in transfected 3T3 cells where substantial amounts of the 42-kDa precursor accumulate (Figs. 2 and 7), would indicate that the extra pentapeptide in H2a can function as a signal for ER retention and degradation when present in the membrane-bound 42-kDa form but does not hinder secretion when present on the cleaved 35-kDa fragment. The charged sequence, EGHRG, was not found in other membrane proteins. Nevertheless, it has been found for other proteins that charged residues in the membrane domain and its flanking residues can target them for the ER (14,25).
It was recently reported that a mixture of monoclonal anti-ASGPR antibodies revealed immunoreactive bands on a Western blot from samples of human serum (26). Although in that report it was not elucidated if the proteins were H1, H2a, or H2b, the molecular mass of ϳ40 kDa would be consistent with our finding for secreted H2a fragment ( fig. 2). Our own preliminary results show the existence of the H2a fragment and not of H1 in normal human serum. 2 Many other membrane receptors have been shown to have soluble, secreted forms (reviewed in Ref. 16). Some examples are receptors for: IgA (27), insulin-like growth factor II (mannose 6-phosphate) (28), interleukin-2 (29), epidermal growth factor (30), and interferon ␣ (18). Some of these are formed by alternatively spliced mRNAs that lead to the deletion of the transmembrane domain. In other cases the wild type membrane receptor is proteolitically cleaved to give a soluble fragment, as is the case for FC␥II receptor with a cleavage site next to the transmembrane domain (31). Signal peptidase was suggested to be involved in the production of soluble dopamine ␤-monooxygenase, which otherwise remains membrane-bound by an uncleaved signal-anchor peptide (32). In the case of the ASGPR H2a, it appears that the insertion of a mini-exon by alternative splicing plays a major role in the final fate of the protein as a soluble secreted form and not membrane-bound as its precursor. All the available evidence suggests that signal peptidase cleaves between Gly-78 and Ser-79 of H2a, next to the membrane, to form the ectodomain fragment which is then secreted (Fig. 5), (2). Shortening of the transmembrane domain (33) or deletion of the cytoplasmic tail (34) of other type II transmembrane proteins caused the protein to be cleaved in the juxtamembrane region. A deletion of the cytoplasmic tail of H1 rendered the polypeptide sensitive to cleavage by signal peptidase (35). In the case of other type II membrane receptors that generate soluble forms, signal peptidase does not seem to be involved. For example the low affinity IgE receptor (Fc⑀RII) gives rise by proteolysis to several secreted fragments of the ectodomain (36). The transferrin receptor has a secreted form comprising its ectodomain and the cleavage is between arginine and leucine, 11 residues carboxylterminal to the membrane span (37). This cleavage site is not predicted to be the target of signal peptidase. Nevertheless one cannot discard the possibility that it is the product of a second cleavage that might take place after the action of signal peptidase. In the case of ASGPR H2a, the NH 2 terminus of the secreted fragment (Fig. 5) is the same as that for the intracellular 35-kDa fragment (2). While in 3T3 cells expressing H2a alone or with H1, cleavage of the precursor to the 35-kDa fragment is post-translational and secretion not very efficient, the rapid cleavage in HepG2 cells suggests a cotranslational event, the transmembrane domain being recognized as an efficient signal peptide in these cells. In HepG2 cells the cleavage of H2a yields fragments predominantly bound for secretion and not targeted to ER degradation. Therefore in HepG2 cells H2a generates only a soluble, secreted proteolytic fragment corresponding to its ectodomain, while H2b, the variant lacking the charged, membrane adjacent pentapeptide associates with H1 to form the functional plasma membrane receptor (Fig. 9). The function of the secreted fragment, its state of oligomerization, and the regulation of its secretion are unknown. Recombinant truncated fragments of the subunit 1 of the rat ASGPR (which were similar or much shorter than the H2a secreted fragment) were able to bind to a galactosylated matrix (38). Therefore we could speculate that the secreted H2a ectodomain may function in binding asialoglycoproteins in the plasma and preventing their interaction with the hepatocyte cell surface receptor, thus prolonging their lifetime in the circulation. On the other hand, its function may be totally unrelated to the function of the membrane receptor, as has been shown for a soluble form of the low density lipoprotein receptor that triggers a response to viral infections (39).