Apical secretion of hepatitis B surface antigen from transfected Madin-Darby canine kidney cells.

Hepatitis B surface antigen (HBsAg), the major envelope component of human hepatitis B virus, during infection drives the assembly and basolateral secretion from hepatocytes of both virions and subviral lipoprotein particles into the bloodstream. We studied the sorting behavior of HBsAg in the heterologous epithelial Madin-Darby canine kidney cells permanently transformed with the hepatitis B virus S gene. These cells, forming tightly packed monolayers in permeable supports, secreted HBsAg apically through a mechanism not involving transcytosis. This suggests that molecular features acting as apical addressing information, seemingly unfunctional or less efficiently used by the exocytic machinery of hepatocytes, could be contained in short hydrophilic regions of HBsAg. Lipids also could play a role in this asymmetric sorting because HBsAg is known to be secreted by forming macromolecular lipoprotein complexes rather than as a soluble protein. Together with available data, our results would imply not only the existence of tissue-specific variations in handling constitutively secreted proteins but also that these variations are strikingly dependent on the kind of protein examined. On the other hand, pulse-chase experiments with tunicamycin showed that the expression of apical information in HBsAg particles does not require N-linked glycosylation, contrasting with the known gp80 Madin-Darby canine kidney-endogenous apical secretory marker. This is the first experimental evidence that carbohydrate moieties in secretory proteins do not hold domain-specific sorting signals, a fact previously shown exclusively for membrane proteins. Thus, HBsAg provides a novel model system for the analysis of the molecular mechanisms of constitutive apical secretion.

Hepatitis B surface antigen (HBsAg), the major envelope component of human hepatitis B virus, during infection drives the assembly and basolateral secretion from hepatocytes of both virions and subviral lipoprotein particles into the bloodstream. We studied the sorting behavior of HBsAg in the heterologous epithelial Madin-Darby canine kidney cells permanently transformed with the hepatitis B virus S gene. These cells, forming tightly packed monolayers in permeable supports, secreted HBsAg apically through a mechanism not involving transcytosis. This suggests that molecular features acting as apical addressing information, seemingly unfunctional or less efficiently used by the exocytic machinery of hepatocytes, could be contained in short hydrophilic regions of HBsAg. Lipids also could play a role in this asymmetric sorting because HBsAg is known to be secreted by forming macromolecular lipoprotein complexes rather than as a soluble protein. Together with available data, our results would imply not only the existence of tissuespecific variations in handling constitutively secreted proteins but also that these variations are strikingly dependent on the kind of protein examined. On the other hand, pulse-chase experiments with tunicamycin showed that the expression of apical information in HBsAg particles does not require N-linked glycosylation, contrasting with the known gp80 Madin-Darby canine kidney-endogenous apical secretory marker. This is the first experimental evidence that carbohydrate moieties in secretory proteins do not hold domain-specific sorting signals, a fact previously shown exclusively for membrane proteins. Thus, HBsAg provides a novel model system for the analysis of the molecular mechanisms of constitutive apical secretion.
Polarized epithelial cells form barriers between external milieu compartments and the internal environment of the organism, functionally connecting these compartments through the vectorial processes of transport, absorption, and secretion. To accomplish their physiological role, these cells exhibit a plasma membrane divided by tight junctions in apical outward facing and basolateral inward facing domains ~~ * This work was supported in part by FONDECYT Grant 721-91.
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. (1). Special protein-sorting processes should operate to furnish each plasma membrane domain with distinct peptide compositions as well as to achieve the asymmetric secretion of specific proteins toward external or internal compartments (2, 3). Sorting signal recognition systems and Golgi-arising vesicular transporting systems are believed to operate concurrently in polarized protein delivery (3), displaying divergencies that deserve further scrutiny in functionally different epithelial cells (4-9). Secretion through domain-oriented exocytic pathways has been far less explored in these aspects than the addressing mechanisms of plasma membrane proteins. Packaging into secretory granules in exocrine cells would ensure an apical secretion regulated by external stimuli (10). However, processes of polarized constitutive (nonregulated) secretion used in general by distinct epithelial cells to release specific proteins at either their apical or basolateral domains (5,6,(11)(12)(13) remain poorly understood. Furthermore, molecular features entailing information for preferential segregation into a selected polarized pathway have not been identified in any secreted protein.
Madin-Darby canine kidney (MDCK)' cells in culture develop features of transporting polarized epithelia (14). These cells possess two optional exocytic pathways emerging from the trans-Golgi region to target either the apical or the basolateral plasmalemma (2, 3,15,16). In these cells, endogenous proteins such as the 80-81-kDa protein (5, 11,12) and a 20-kDa osteopontine-like protein (13) are secreted predominantly apical, and laminin and sulfate proteoglycans are secreted basolaterally (17), whereas several exogenous proteins expressed during transfection experiments showed unpolarized secretion (5, 11). Thus, domain-specific sorting signals should mediate asymmetric delivery of secretory proteins to each cell surface domain in these cells (3).
In contrast, hepatocytes, as cells specialized in the constitutive secretion of many blood protein components such as albumin and lipoproteins, address most newly synthesized exocytic products to the sinusoidal plasma membrane (18-20), the equivalent of the basolateral surface of other epithelial cells. Whether special sorting information would be required for polarized secretion in hepatocytes remains speculative, the prevalent idea being that these cells lack an apical exocytic pathway and, therefore, their basolateral secretion would rather function through a bulk-phase mechanism (4). HOWever, recent evidence is congruent with a direct albumin secretion into bile (18), suggesting instead that hepatocytes do possess an apical (canalicular) secretory pathway of sizable transporting capacity.
Our understanding of the addressing mechanisms used by different epithelial cells can be improved by analyzing the The abbreviations used are: MDCK, Madin-Darby canine kidney; HBsAg, hepatitis B surface antigen; HBV, hepatitis B virus; rGH, rat growth hormone.

Apical Secretory Information in HBsAg
Particles 6663 sorting behavior ofthe same protein in heterologous epithelial cell systems. The only hepatic secretory protein thus far expressed by transfection experiments in MDCK cells, namely cy2,,-globulin, was found secreted in a nonpolarized fashion as expected for a protein that either lacks or bears tissue-specific sorting information (5). Other secretory proteins basolaterally addressed by hepatocytes could contain specific targeting information of, as yet, unexpected sorting properties. It is interesting, for instance, that the gp80 apical secretory marker of MDCK cells, recently cloned and sequenced (21), is homologous to the human plasma SP-40,40 and apolipoprotein-J, which are instead basolaterally secreted by hepatocytes (21-23). The opposite polarity expressed by these proteins in hepatocytes and MDCK cells could be caused by minor sequence differences or could be reflecting special characteristics of lipid-associated secretory proteins. Additional examples that can be revealed by assay in MDCK cells seem necessary for allowing a more general and conclusive conviction.
The hepatitis B surface antigen (HBsAg) encoded by the S gene of the human hepatitis B virus (HBV) (24-26) parallels the sorting behavior of the gp80-like proteins (21). During HBV infection, hepatocytes basolaterally release to the bloodstream enormous quantities of subviral particles (22 nm) lacking nucleic acids and mainly composed of both unglycosylated (22-25 kDa) and glycosylated (28-30 kDa) forms of HBsAg associated with host-derived lipids (27,28). HBsAg is a highly hydrophobic 226-amino acid protein synthesized as a transmembrane intermediary, unique in its capacity to selfassemble with lipids forming empty envelope particles, independent of additional viral components (29, 30). A buddinglike step, presumably into the lumen of the endoplasmic reticulum (30) or into a post-endoplasmic reticulum, pre-Golgi compartment in which extensive intermolecular disulfide cross-linking of HBsAg dimers would occur as the later step during the assembly of HBsAg lipoprotein particles (31), would be followed by rapid transport of this lipoprotein complex along the exocytic route for secretion (29,30,(32)(33)(34).
HBsAg also plays a crucial role in the assembly and secretion of viral particles (35). However, its elaborate properties as substrate of constitutive exocytic pathways have hitherto only been explored in nonpolarized cells.
We transfected the HBV S gene into MDCK cells and analyzed the polarity of HBsAg secretion in permanent transformants of these cells growing over permeable supports. We found, unexpectedly, that this protein was secreted almost exclusively apical from these cells through a mechanism not involving transcytosis. With the exception of the apical secretion of several truncated soluble versions of plasma membrane proteins (36-40), a nonpolarized secretion has been reported for all the natural exogenous secretory proteins so far expressed in MDCK cells (5). As substrate of the sorting apparatus, HBsAg would be more closely related to gp80 (21) than to az,-globulin (5). HBsAg adds to the gp80 protein in providing the only two presently available model systems to analyze apical sorting information in constitutively secreted proteins through a genetic approach and for the exploration of apical pathways in other epithelial cells.

MATERIALS AND METHODS
Expression Plasmid Encoding HBsAg-Standard recombinant DNA techniques were followed (41). A PuuII/BamHI, 18.50-kilobase fragment carrying the HBsAg coding region (S gene) flanked by the early promoter and polyadenylation site of SV40 was isolated from plasmid LSV-HBsAg (32) kindly provided by Dr. P. Valenzuela (Chiron Co., Berkeley, CA). This fragment was inserted between the PuuIIIBamHI sites of a pSp64 vector and then removed with HindIII/ BamHI for its final ligation into the corresponding sites of a psv2derived mammalian expression vector that contains the bacterial neomycin-resistance gene under control of the herpes simplex virus thymidine kinase promoter (kindly provided by Drs. D. Sabatini and T. Gottlieb Transient Expression of rGH-The previously obtained neomycinresistant MDCK cells were additionally transfected in 75-cmZ flasks with an expression plasmid bearing the rGH under the control of the SV40 early promoter using previously described procedures (5). One day after transfection, lo6 cells were plated into Millicell chambers, treated with sodium butyrate for 12 h, and 48 h after plating, the polarity of HBsAg and rGH secretion was assessed as described in the previous section.
Pulse-Chose Experiments-Pulse-chase experiments were made in 35-mm plastic dishes by preincubating the cells for 1 h in methloninefree medium and then for 30 min with 1 mCi/ml [35S]methionine (Du Pont-New England Nuclear). After each chasing period, in the presence of 100-fold excess methionine (3 mg/ml), HBsAg was immunoprecipitated from cell extracts and medium using a rabbit polyclonal antibody raised against HBsAg particles produced in recombinant yeast (44).
Pulse-chase experiments, performed to assess the effect of tunicamycin treatment on cells forming monolayers over nitrocellulose filters, were done under the conditions previously described (12).
Briefly, the cells were labeled for 30 min with 150 pCi of [%I methionine/l50 pl of minimum Eagle's medium supplemented with 10 mM Hepes (pH 7.4) and 0.2% bovine serum albumin applied to the basolateral side of the inverted filter, and then chased in the presence of 3 mg/ml methionine. Tunicamycin was added at a final concentration of 12 pg/ml during a 60-min preincubation and all through the labeling and chase periods.

RESULTS
To study the sorting behavior of the HBsAg in MDCK cells, we followed a standard approach (5, 42). The S gene of the HBV was first inserted into an expression vector (depicted in Fig. 11, transfected into MDCK cells, and then permanent transformants that express and secrete HBsAg were obtained by selection in G418. Incubation with sodium butyrate was needed to achieve detectable levels of expression, as already Hindm Apical Secretory Information in HBsAg Particles

FIG. 2. Expression and secretion of HBsAg in permanent transformants of MDCK cells. MDCK cells stably transformed
with the pSV2-HBsAg-TKneo plasmid and grown in plastic dishes were labeled for 8 h with 125 pCi/ml [35S]methionine in methioninefree medium, either untreated ( l a n e I ) or pretreated ( l a n e 2) for 12 h with 10 mM sodium butyrate to enhance the expression. HBsAg was immunoprecipitated from the medium and analyzed by SDS-polyacrylamide gel electrophoresis and fluorography. Two bands representing unglycosylated ( p 2 4 ) and glycosylated (gp28) HBsAg were detectable only after pretreating the cells with sodium butyrate.
observed for other cDNAs expressed under the control of the SV40 early promoter in these cells (42, 43). Immunoprecipitation from the medium of metabolically labeled cells pretreated with butyrate showed secretion of the characteristic unglycosylated (p24) and glycosylated (gp28) forms of HBsAg (Fig. 2). Pulse-chase experiments revealed that about 90-95% of newly synthesized HBsAg is secreted in 6-8 h (Fig. 3), showing efficient secretion, although at a lower rate than that reported for the 81-kDa endogenous protein (5).
T o assess the polarity of HBsAg secretion, the cells were grown over permeable filter supports in Millicell chambers that allow a separate analysis of products released either apically or basolaterally. T o our surprise, a predominant apical release of HBsAg was found ( Fig. 4) with no more than 10-15%, and most frequently less than 5%, of the HBsAg being basolaterally detected in different experiments. The incubation period of 8 h was long enough for HBsAg to cross the protein-saturated filter, as judged by the relatively rapid equilibrium observed between both chamber compartments after opening the tight junctions by EDTA during the last hour of labeling (Fig. 4, lanes 3 and 4 ) . Similar   Hours of Chase  FIG. 3. HBsAg is efficiently secreted from MDCK cells. Cells grown to confluency in 35-mm plastic dishes were pulse-labeled for 30 min with 1 mCi/ml [36S]methionine and chased for the time periods indicated in medium containing cold methionine in excess. HBsAg was immunoprecipitated from cell extracts and medium. The corresponding bands in the fluorogram were quantitated by scanning densitometry, and the amount found in the medium was expressed as percent of the total.  4. Apical secretion of newly synthesized HJ3sAg from permanent transformants of MDCK cells. Monolayers of cells grown on nitrocellulose filters (0.45-pm pore size) of Millicell chambers and expressing transepithelial electrical resistance (300-400 ohms/cm2) were found to release HBsAg almost exclusively to the apical (Ap) chamber (lanes I and 2). In contrast, similar amounts of HBsAg were found in each compartment after opening the tight junctions with 5 mM EDTA added to the medium during the last hour of the labeling period (lanes 3 and 4 ) , demonstrating the HBsAg capability for crossing the transcellular and filter barriers. When prelabeled HBsAg was placed in the basolateral (Bl) compartment, none of this protein could be found apically after 8 h of incubation (lanes 5 and 6). Thus, there was no detectable transcytosis, and significant degradation of the protein was not observed as judged by the recovery. account for the apical appearance of HBsAg were discarded. In fact, previously labeled HBsAg placed in the basolateral compartment was almost entirely recovered by immunoprecipitation after 8 h of incubation and was undetectable in the apical compartment (Fig. 4, lanes 5 and 6).
It has been shown that rGH is secreted in a nonpolarized fashion from transfected MDCK cells during transient expression experiments (5). T o estimate the fluid phase transit of the apical and basolateral pathways and to determine whether our permanent transformants of MDCK cells still maintain the capability of basolateral secretion, we performed similar experiments. The described vector bearing rGH cDNA (5) was overtransfected into HBsAg transformant MDCK cells, and the simultaneous secretion of both rGH and HBsAg was assessed by immunoprecipitating each protein in tandem. A clear difference in the secretion pattern of these proteins was observed with a relatively nonpolarized secretion of rGH. As seen in Fig. 5, less than 5% of HBsAg uersus approximately 40% of the rGH was found basolaterally.  (lanes 1 and 2 ) and rGH (lanes 3 and 4 ) secretion was assessed by immunoprecipitating each protein in tandem, separately from the apical and the basolateral medium after 8 h of metabolic labeling. Although HBsAg secretion was predominantly (95%) apical, that of rGH was found 60% apical and 40% basolateral. Although glycosylated HBsAg accounted for only 10-20% of the total HBsAg secreted by MDCK cells and glycosylation is not necessary for secreting HBsAg particles in nonpolarized cells (34), it is still possible that the polarized release of HBsAg particles composed of both the glycosylated and unglycosylated forms of the protein could be dependent upon Nglycosylation. When filter-grown MDCK cells were preincubated, pulse-labeled, and then chased for 90 min in the presence of 12 pg/ml tunicamycin (45) following the same procedures previously used in studies of the apical endogenous gp80 secreted protein (12), only nonglycosylated HBsAg was secreted, however, maintaining its apical polarity as untreated cells (Fig. 6). This indicates that N-glycosylation is not an important factor in the asymmetric segregation of HBsAg, contrasting with the gp80 protein that lost polarity under similar conditions (12).

DISCUSSION
Previous studies have shown that HBsAg possesses structural features to drive the assembly and secretion of both virions and subviral lipoprotein particles through a process involving a transmembrane intermediate and a budding-like step into the endoplasmic reticulum lumen (29,30,(33)(34)(35). The results presented in this paper suggest that HBsAg also contains special sorting information that, depending on the kind of polarized epithelial cells, would determine its specific addressing into apical exocytic pathways.
We expressed the HBsAg in polarized monolayers of stable transformant MDCK cells growing over permeable supports and found its secretion predominantly apical. Although we show only results obtained in nitrocellulose supports of Mil-lice11 chambers, similar observations were made in polycarbonate filters of Transwell chambers. Control experiments excluded the possibility of transcytosis and selective basolat-era1 degradation. By contrast, the secretion of rGH was practically unpolarized when these permanent transformant cells were overtransfected with a rGH expression plasmid (5), excluding the possibility that our transformant cells had lost the ability for basolateral secretion.
MDCK cells have two co-existing secretory pathways, one directed to the apical and the other to the basolateral plasmalemma (5,11,17). A number of exogenous secretory proteins, including rGH, expressed after transfecting these cells with appropriate plasmids (5,12), are secreted in roughly equal amounts through both cell surface domains. This should occur with proteins that either do not express any sorting information for polarity or possess tissue-specific addressing information nonfunctional in MDCK cells, thus being secreted through both cell surfaces at random, in amounts proportional to their default partition (not signal-mediated) within the bulk phase of each pathway. Instead, the apical secretion of HBsAg most probably reflects that specific sorting information, remaining to be identified, is being intracellularly deciphered, presumably at the trans-Golgi network (31, and actively conducts this protein into a selected secretory pathway through a receptor(s)-mediated process, as has been suggested for the endogenous MDCK apical secretory marker (5).
Previous examples of exogenous secretory products apically released in transfected MDCK cells had been restricted to the ectoplasmic domains of several plasma membrane proteins. In fact, soluble anchor-minus versions of the transmembrane proteins aminopeptidase (38), polymeric immunoglobulin receptor (37), and neutral endopeptidase (36) and of the glycosyl phosphatidylinositol-linked proteins Thy-1 (39) and placental alkaline phosphatase (40) are all secreted preferentially apical, though in some cases with less than 80% polarity (37, 40) when expressed in MDCK cells. This, together with observations made using chimeric proteins (46, 47), suggests that the ectodomain of apical membrane proteins expresses targeting information, and, therefore, secretory and plasma membrane proteins destined to the apical cell surface could share sorting events taking place within a luminal compartment of the exocytic route, probably at the trans-Golgi network and maybe involving the same receptor. For transmembrane proteins, though, sorting events can also be directed from the cytosolic compartment as suggested by the influence of cytosolic-specific sequences in basolateral sorting of several plasma membrane proteins (48-52). Some transmembrane apical proteins could also require, in some yet unclear way, their anchoring and/or cytosolic segments for proper sorting. For instance, a truncated soluble form of the apical marker, influenza hemagglutinin, is secreted unpolarized in transfected MDCK cells (42), although this observation disagrees with that of other investigators (46). There is usually uncertainty about the characteristics that could be acquired by secretory products derived through genetically engineering the cDNA of plasma membrane proteins. Also, it is still unclear whether plasma membrane and secretory proteins follow the same exocytic pathways (4,53). The biogenesis of HBsAg particles includes a transmembrane precursor in the endoplasmic reticulum (29, 30); however, this is a natural process that more likely should lead to a product conserving the native characteristics of endogenous substrates for constitutive secretory pathways, especially those forming lipoprotein particles.
The opposite polarity by which HBsAg was secreted from transfected MDCK cells compared with the basolateral release known to occur from infected hepatocytes to the bloodstream (25) was not foreseen. Former evidence led to the belief that a direct apical exocytic pathway does not exist in hepatocytes and that bile-secreted proteins would be mainly transported by transcytosis after bulk phase or receptormediated basolateral endocytosis from the plasma or through a paracellular route (4,19,20). In challenging this view, Saucan and Palade (18) have recently presented evidence suggesting that the major source of bile albumin is newly synthesized albumin directly secreted from hepatocytes. A direct apical secretory pathway seemingly of considerable transporting capacity, as judged by the biliary albumin concentration, would, therefore, be functional in these cells. However, only very small amounts of HBsAg have been detected by radioimmunoassay in bile during acute HBV infection and in chronic carriers (54).
A prominent basolateral exocytic pathway would not require special addressing information to produce asymmetric secretion as has been shown experimentally by expressing rGH (6) and the K light chain (55) in the polarized cell line, Caco-2. Similarly, in hepatocytes, as a current assumption sustains (4), a default mechanism of protein segregation could determine the predominant sinusoidal secretion of newly synthesized secretory proteins that do not hold specific sorting information. This notion is compatible with the unpolarized secretion of the hepatic form of cYa,-globulin previously observed in MDCK cells (5). However, the apical secretion of HBsAg in MDCK cells suggests that distinctly secreted proteins, while sharing similar polarity in hepatocytes, can otherwise be handled in completely different fashions by other epithelial cells and that some of the endogenous hepatic proteins could hold addressing information by becoming functional in the heterologous cellular systems. Strikingly, HBsAg seems to mimic the sorting behavior of the kidney and hepatic forms of the gp80 protein (21), all of them bearing special molecular features to interact with lipids. In fact, the gp80 apical secretory protein of MDCK cells has as homologous proteins SP-40,40 and apolipoprotein-J of human plasma, produced and basolaterally secreted by hepatocytes (21-23). Apolipoprotein-J is a component of a subclass of high density lipoproteins bearing apoA-I, and the SP-40,40 protein was identified forming complexes with complement components, presumably through hydrophobic interactions (22, 23). It could be that elements featuring apical information in this kind of protein are not functional in hepatocytes, or in these cells, are less efficiently recognized or overridden by interactions with other basolaterally addressed proteins. Perhaps, in hepatocytes, intracellular associations with apolipoproteins, such as apoA-I, might command the basolateral release of apoJ and HBsAg. It would be of great interest to test the secretion of HBsAg in Caco-2 cells because these cells produce lipoproteins and are thought to display similar sorting characteristics with hepatocytes, as none of their endogenous secretory proteins have been found to be secreted preferentially to the apical pole (6).
The functional meaning of the HBsAg sorting characteristics for the HBV life cycle remains to be elucidated. Envelope glycoproteins of several viruses emulate endogenous substrates of constitutive (nonregulated) polarized exocytic routes as part of the viral strategy for exiting from infected epithelial cells toward external or internal body fluids (43,56, 57). Hepadna viruses are not strictly hepatotropic. Evidence exists of epithelial cell infection by these viruses in pancreas and kidney (58-60), and bile duct epithelium (61). In human carriers, low amounts of HBsAg and/or viral DNA have been detected in apically derived secretions such as milk (62), pancreatic juice (54), and semen and saliva (63), the last two seemingly important in non-parenteral HBV transmission. The presence of HBsAg in these fluids has been generally attributed to either transepithelial transport from the blood (64) or blood leakage from the intravascular space (63). However, the structural features driving HBsAg into specific exocytic pathways also may play a role in extrahepatic tissues as part of the mechanism of viral exit from the organism.
The only moieties known so far to function as apical targeting information are glycosyl phosphatidylinositol stalks anchoring a special class of membrane proteins to the bilayer (65). The presence of such a signal in HBsAg can be practically excluded because glycolipids were not detected in a detailed analysis of highly purified subviral particles (28). We found that inhibition of N-glycosylation by tunicamycin does not affect the HBsAg-polarized secretion. Because tunicamycin treatment produces a loss of polarity in the previously known apical constitutive secretory marker, the gp80 protein (12), this result constitutes the first formal evidence that oligosaccharides do not hold sorting information in secretory proteins, a fact previously shown for plasma membrane proteins only (2,66,67). This observation would also argue against the possibility that HBsAg was apically addressed by interacting with gp80 in MDCK cells.
The apical sorting determinants of HBsAg and the gp80 could differ in their N-linked carbohydrate dependence for functional expression. For instance, the active information could reside in a particular peptide conformation stabilized upon glycosylation in the gp80 protein but not in the HBsAg. It seems less likely that two different apical signals could exist in these two proteins and both be recognized by the same cell. It is possible that lipid association properties, for instance prompting the interaction with the membrane of apical exocytic vesicles that are likely to be enriched in glycosphingolipids (3), could be the important sorting determinant in both gp80 and HBsAg proteins. On the other hand, if the recognition would occur in an aqueous environment, a small hydrophilic region of 30-40 amino acids between two hydrophobic regions (68) presumably facing the surface of subviral HBsAg particles (28,69) could be suitable for holding a sorting signal. Mutants of the protein should enable a test of this possibility. Finally, because HBsAg particles are biotechnologically produced in large amounts (44), this viral glycoprotein could also provide a convenient ligand system for attempting the detection of hypothetical intracellular receptors mediating apical protein segregation. 19.