Identification of sequence determinants that direct different intracellular folding pathways for aquaporin-1 and aquaporin-4

Homologous aquaporin water channels utilize different folding pathways to acquire their transmembrane (TM) topology in the endoplasmic reticulum (ER). AQP4 acquires each of its six TM segments via cotranslational translocation events, whereas AQP1 is initially synthesized with four TM segments and subsequently converted into a six membrane-spanning topology. To identify sequence determinants responsible for these pathways, peptide segments from AQP1 and AQP4 were systematically exchanged. Chimeric proteins were then truncated, fused to a C-terminal translocation reporter, and topology was analyzed by protease accessibility. In each chimeric context, TM1 initiated ER targeting and translocation. However, AQP4-TM2 cotranslationally terminated translocation, while AQP1-TM2 failed to terminate translocation and passed into the ER lumen. This difference in stop transfer activity was due to two residues that altered both the length and hydrophobicity of TM2 (Asn(49) and Lys(51) in AQP1 versus Met(48) and Leu(50) in AQP4). A second peptide region was identified within the TM3-4 peptide loop that enabled AQP4-TM3 but not AQP1-TM3 to reinitiate translocation and cotranslationally span the membrane. Based on these findings, it was possible to convert AQP1 into a cotranslational biogenesis mode similar to that of AQP4 by substituting just two peptide regions at the N terminus of TM2 and the C terminus of TM3. Interestingly, each of these substitutions disrupted water channel activity. These data thus establish the structural basis for different AQP folding pathways and provide evidence that variations in cotranslational folding enable polytopic proteins to acquire and/or maintain primary sequence determinants necessary for function.     

Evidence that the transmembrane biogenesis of aquaporin 1 is cotranslational in intact mammalian cells

Most polytopic membrane proteins are believed to integrate into the membrane of the endoplasmic reticulum (ER) cotranslationally. However, recent studies with Xenopus oocytes and dog pancreatic microsomes have suggested that this is not the case for human aquaporin 1 (AQP1). These experiments indicate that membrane-spanning segments (MSSs) 2 and 4 of AQP1 do not integrate into the membrane cotranslationally so that this protein initially adopts a four MSS topology. A later maturation event involving a 180-degree rotation of MSS 3 from an N(lum)/C(cyt) to an N(cyt)/C(lum) orientation and the concomitant integration of MSSs 2 and 4 into the membrane results in the final six MSS topology. Here we examine the biogenesis of AQP1 in the human embryonic kidney cell line HEK-293T. To do this, we constructed an expression vector for a fusion protein consisting of the enhanced green fluorescent protein followed by an insertion site for AQP1 sequences and a C-terminal glycosylation tag. We then transiently transfected HEK-293T cells with this vector containing the AQP1 sequence truncated after each MSS. Glycosylation of the C-terminal tag was used to monitor its location relative to the ER lumen and consequently the membrane integration and orientation of successive MSSs. In contrast to previous studies our results indicate that AQP1 integrates into the ER membrane cotranslationally in intact HEK-293T cells.     

Reorientation of aquaporin-1 topology during maturation in the endoplasmic reticulum

The topology of most eukaryotic polytopic membrane proteins is established cotranslationally in the endoplasmic reticulum (ER) through a series of coordinated translocation and membrane integration events. For the human aquaporin water channel AQP1, however, the initial four-segment-spanning topology at the ER membrane differs from the mature six-segment-spanning topology at the plasma membrane. Here we use epitope-tagged AQP1 constructs to follow the transmembrane (TM) orientation of key internal peptide loops in Xenopus oocyte and cell-free systems. This analysis revealed that AQP1 maturation in the ER involves a novel topological reorientation of three internal TM segments and two peptide loops. After the synthesis of TMs 4-6, TM3 underwent a 180-degree rotation in which TM3 C-terminal flanking residues were translocated from their initial cytosolic location into the ER lumen and N-terminal flanking residues underwent retrograde translocation from the ER lumen to the cytosol. These events convert TM3 from a type I to a type II topology and reposition TM2 and TM4 into transmembrane conformations consistent with the predicted six-segment-spanning AQP1 topology. AQP1 topological reorientation was also associated with maturation from a protease-sensitive conformation to a protease-resistant structure with water channel function. These studies demonstrate that initial protein topology established via cotranslational translocation events in the ER is dynamic and may be modified by subsequent steps of folding and/or maturation.     

Analysis of double knockout mice lacking aquaporin-1 and urea transporter UT-B. Evidence for UT-B-facilitated water transport in erythrocytes

We reported increased water permeability and a low urea reflection coefficient in Xenopus oocytes expressing urea transporter UT-B (former name UT3), suggesting that water and urea share a common aqueous pathway (Yang, B., and Verkman, A. S. (1998) J. Biol. Chem. 273, 9369-9372). Although increased water permeability was confirmed in the Xenopus oocyte expression system, it has been argued (Sidoux-Walter, F., Lucien, N., Olives, B., Gobin, R., Rousselet, G., Kamsteeg, E. J., Ripoche, P., Deen, P. M., Cartron, J. P., and Bailly, P. (1999) J. Biol. Chem. 274, 30228-30235) that UT-B does not transport water when expressed at normal levels in mammalian cells such as erythrocytes. To quantify UT-B-mediated water transport, we generated double knockout mice lacking UT-B and the major erythrocyte water channel, aquaporin-1 (AQP1). The mice had reduced survival, retarded growth, and defective urinary concentrating ability. However, erythrocyte size and morphology were not affected. Stopped-flow light scattering measurements indicated erythrocyte osmotic water permeabilities (in cm/s x 0.01, 10 degrees C): 2.1 +/- 0.2 (wild-type mice), 2.1 +/- 0.05 (UT-B null), 0.19 +/- 0.02 (AQP1 null), and 0.045 +/- 0.009 (AQP1/UT-B null). The low water permeability found in AQP1/UT-B null erythrocytes was also seen after HgCl(2) treatment of UT-B null erythrocytes or phloretin treatment of AQP1 null erythrocytes. The apparent activation energy for UT-B-mediated water transport was low, <2 kcal/mol. Estimating 14,000 UT-B molecules per mouse erythrocyte, the UT-B-dependent P(f) of 0.15 x 10(-4) cm/s indicated a substantial single channel water permeability of UT-B of 7.5 x 10(-14) cm(3)/s, similar to that of AQP1. These results provide direct functional evidence for UT-B-facilitated water transport in erythrocytes and suggest that urea traverses an aqueous pore in the UT-B protein.     

Topogenesis of cystic fibrosis transmembrane conductance regulator (CFTR): regulation by the amino terminal transmembrane sequences

Cystic fibrosis transmembrane conductance regulator (CFTR) is a member of the ATP-binding cassette (ABC) transport superfamily. CFTR folding and assembly appear to involve several events occurred in the cytosol and ER. Misfolding of CFTR causes cystic fibrosis, and thus, understanding the folding mechanism of CFTR is extremely important. Recently, detailed study of membrane insertion process suggests that the first two transmembrane (TM) segments of CFTR have two distinct but independent mechanisms to ensure the correct membrane folding of its amino terminal end [Lu, Y., Xiong, X., Helm, A., Kimani, K., Bragin, A., Skach, W. R. (1998) J. Biol. Chem. 273, 568-576]. To understand how other TM segments are ensured to insert into membranes correctly, we investigated the topogenesis of TM3 and TM4 of CFTR in a cell-free expression system. We found that the correct membrane insertion of TM3 and TM4 of CFTR was ensured by their flanking amino acid sequences and controlled by the correct membrane insertion of their preceding TM1 and TM2. Thus, correct membrane insertion and folding of TM1 and TM2 play an essential role in the membrane insertion and folding of the subsequent TM segments of CFTR.     

Transmembrane segment (TMS) VIII of the Na(+)/Citrate transporter CitS requires downstream TMS IX for insertion in the Escherichia coli membrane.

The amino acid sequence of the sodium ion-dependent citrate transporter CitS of K. pneumoniae contains 12 hydrophobic stretches that could form membrane-spanning segments. A previous analysis of the membrane topology in Escherichia coli using the PhoA gene fusion technique indicated that only nine of these hydrophobic segments span the membrane, while three segments, Vb, VIII and IX, were predicted to have a periplasmic location (Van Geest, M., and Lolkema, J. S. (1996) J. Biol. Chem. 271, 25582-25589). A topology study of C-terminally truncated CitS molecules in dog pancreas microsomes revealed that the protein traverses the endoplasmic reticulum membrane 11 times. In agreement with the PhoA fusion data, segment Vb was predicted to have a periplasmic location, but, in contrast, segments VIII and IX were found to be membrane-spanning (Van Geest, M., Nilsson, I., von Heijne, G., and Lolkema, J. S. (1999) J. Biol. Chem. 274, 2816-2823). In the present study, using site-directed Cys labeling, the topology of segments VIII and IX in the full-length CitS protein was determined in the E. coli membrane. Engineered cysteine residues in the loop between the two segments were accessible to a membrane-impermeable thiol reagent exclusively from the cytoplasmic side of the membrane, demonstrating that transmembrane segments (TMSs) VIII and IX are both membrane-spanning. It follows that the folding of CitS in the E. coli and endoplasmic reticulum membrane is the same. Cysteine accessibility studies of CitS-PhoA fusion molecules demonstrated that in the E. coli membrane segment VIII is exported to the periplasm in the absence of the C-terminal CitS sequences, thus explaining why the PhoA fusions do not correctly predict the topology. An engineered cysteine residue downstream of TMS VIII moved from a periplasmic to a cytoplasmic location when the fusion protein containing TMSs I-VIII was extended with segment IX. Thus, downstream segment IX is both essential and sufficient for the insertion of segment VIII of CitS in the E. coli membrane.     

Stepwise insertion and inversion of a type II signal anchor sequence in the ribosome-Sec61 translocon complex

In eukaryotic cells, the ribosome-Sec61 translocon complex (RTC) establishes membrane protein topology by cotranslationally partitioning nascent polypeptides into the cytosol, ER lumen, and lipid bilayer. Using photocrosslinking, collisional quenching, cysteine accessibility, and protease protection, we show that a canonical type II signal anchor (SA) acquires its topology through four tightly coupled and mechanistically distinct steps: (1) head-first insertion into Sec61α, (2) nascent chain accumulation within the RTC, (3) inversion from type I to type II topology, and (4) stable translocation of C-terminal flanking residues. Progression through each stage is induced by incremental increases in chain length and involves abrupt changes in the molecular environment of the SA. Importantly, type II SA inversion deviates from a type I SA at an unstable intermediate whose topology is controlled by dynamic interactions between the ribosome and translocon. Thus, the RTC coordinates SA topogenesis within a protected environment via sequential energetic transitions of the TM segment.     

The TGN38 glycoprotein contains two non-overlapping signals that mediate localization to the trans-Golgi network

The membrane-spanning and cytoplasmic domains of CD4 and CD8 were replaced by those of TGN38. After transient expression in HeLa cells, the location of the hybrid proteins was determined using immunofluorescence and quantitative immuno-electron microscopy, FACS analysis and metabolic labeling. The membrane-spanning domain was found to contain a signal that localized hybrid proteins to the TGN. This was in addition to the signal previously identified in the cytoplasmic domain (Bos, K., C. Wraight, and K. Stanley. 1993. EMBO (Eur. Mol. Biol. Organ) J. 12:2219-2228. Humphrey, J. S., P. J. Peters, L. C. Yuan, and J. S. Bonifacino. 1993. J. Cell Biol. 120:1123-1135. Wong, S. H., and W. Hong. 1993. J. Biol. Chem. 268:22853-22862). The different properties of these two signals suggest that each operates by a different mechanism.     

Proteolytic cleavages of proalbumin and complement Pro-C3 in vitro by a truncated soluble form of furin, a mammalian homologue of the yeast Kex2 protease.

We have recently purified and characterized a truncated soluble form of furin from which the predicted transmembrane domain and cytoplasmic tail were deleted (Hatsuzawa, K., Nagahama, M., Takahashi, S., Takada, K., Murakami, K., and Nakayama, K. (1992) J. Biol. Chem. 267, 16094-16099). Our results showed that furin resembles the yeast Kex2 protease with respect to both its enzymic properties and substrate specificity. Here we demonstrate that the soluble form of furin is capable of converting the precursors of albumin and the third component of complement (proalbumin and pro-C3, respectively) in vitro to mature proteins. Thus furin mimics the Ca(2+)-dependent proalbumin and pro-C3 convertases found in the Golgi membranes (Brennan, S. O., and Peach, R. J. (1988) FEBS Lett. 229, 167-170; Oda, K. (1992) J. Biol. Chem. 267, 17465-17471). Furthermore we show that the variant alpha 1-antitrypsin Pittsburgh, which is a specific inhibitor of the Golgi proalbumin convertase, inhibits not only the Golgi pro-C3 convertase, but also the soluble furin. These results suggest a role for furin in the cleavage of proproteins transported via the constitutive pathway.     

The nucleotide sequence of the gene for malF protein, an inner membrane component of the maltose transport system of Escherichia coli. Repeated DNA sequences are found in the malE-malF intercistronic region

The malF gene product is an inner membrane component of the maltose transport system in Escherichia coli. Some gene fusions between malF and lacZ (encoding the normally cytoplasmic enzyme beta-galactosidase) produce hybrid proteins which are membrane-bound while other fusions produce hybrid proteins which are cytoplasmic (Silhavy, T. J., Casadaban, M. J., Shuman, H. A., and Beckwith, J. R. (1976) Proc. Natl. Acad. Sci. U. S. A. 73, 3423-3427). To further analyze the localization properties of the different classes of fusion proteins and of the intact MalF protein, we have obtained the DNA sequence of 5 malF-lacZ fusions and the wild type malF gene. From the predicted amino acid sequence, MalF protein contains 514 amino acids and has a molecular weight of 56,947. Analysis of the hydropathic character of MalF using the Kyte-Doolittle assignments (Kyte, J., and Doolittle, R. F. (1982) J. Mol. Biol. 157, 105-132), indicates that the protein may have 2 or 3 amino-terminal membrane-spanning segments and 4 or 5 carboxy-terminal membrane-spanning segments separated by a region of 181 hydrophilic residues. Localization properties of the different fusion proteins correspond with degree of hydrophobicity. By sequencing upstream from malF, the malE-malF intercistronic region was found to be 153 base pairs in length and to contain inverted repeats, homologous to intercistronic repeats of many other operons. Further analysis of this region may help in understanding the observed step-down in synthesis of the MalF protein.     

Selectively reduced glycerol in skin of aquaporin-3-deficient mice may account for impaired skin hydration,elasticity, and barrier recovery

Deletion of the epidermal water/glycerol transporter aquaporin-3 (AQP3) in mice reduced superficial skin conductance by approximately 2-fold (Ma, T., Hara, M., Sougrat, R., Verbavatz, J. M., and Verkman, A. S. (2002) J. Biol. Chem. 277, 17147-17153), suggesting defective stratum corneum (SC) hydration. Here, we demonstrate significant impairment of skin hydration, elasticity, barrier recovery, and wound healing in AQP3 null mice in a hairless (SKH1) genetic background and investigate the cause of the functional defects by analysis of SC morphology and composition. Utilizing a novel (3)H(2)O distribution method, SC water content was reduced by approximately 50% in AQP3 null mice. Skin elasticity measured by cutometry was significantly reduced in AQP3 null mice with approximately 50% reductions in elasticity parameters Uf, Ue, and Ur. Although basal skin barrier function was not impaired, AQP3 deletion produced an approximately 2-fold delay in recovery of barrier function as measured by transepidermal water loss after tape stripping. Another biosynthetic skin function, wound healing, was also approximately 2-fold delayed by AQP3 deletion. By electron microscopy AQP3 deletion did not affect the structure of the unperturbed SC. The SC content of ions (Na(+), K(+), Ca(2+), Mg(2+)) and small solutes (urea, lactic acid, glucose) was not affected by AQP3 deletion nor was the absolute amount or profile of lipids and free amino acids. However, AQP3 deletion produced significant reductions in glycerol content in SC and epidermis (in nmol/microg protein: 5.5 +/- 0.4 versus 2.3 +/- 0.7 in SC; 0.037 +/- 0.007 versus 0.022 +/- 0.005 in epidermis) but not in dermis or blood. These results establish hydration, mechanical, and biosynthetic defects in skin of AQP3-deficient mice. The selective reduction in epidermal and SC glycerol content in AQP3 null mice may account for these defects, providing the first functional evidence for physiologically important glycerol transport by an aquaporin.     

Biogenesis and topology of the secretory Na+-K+-2Cl- cotransporter (NKCC1) studied in intact mammalian cells

The "secretory" Na+-K+-2Cl- cotransporter (NKCC1) is a member of a small gene family with nine homologues in vertebrates. Of these, seven are known to be electroneutral chloride transporters. These transporters play a number of important physiological roles related to salt and water homeostasis and the control of intracellular chloride levels. Hydropathy analyses suggest that all of these transporters have a similar transmembrane topology consisting of relatively large intracellular N and C termini and a central hydrophobic domain containing 12 membrane-spanning segments (MSSs). In recent experiments from our laboratory [Gerelsaikhan, T., and Turner, R. J. (2000) J. Biol. Chem. 275, 40471-40477], we employed an in vitro translation system to confirm that each of the putative MSSs of NKCC1 was capable of membrane integration in a manner consistent with a 12 MSS model. Here, we extend that work to the study of the biogenesis of NKCC1 in intact cells. We employ a truncation mutant approach that allows us to monitor this process quantitatively as successive MSSs are synthesized. While the results presented here confirm the 12 MSS model, they also indicate that the integration of NKCC1 into the membrane does not occur via a simple cotranslational process. In particular, we demonstrate that two MSSs, the second and sixth, require the presence of downstream sequence to efficiently integrate into the membrane.     

A nine-transmembrane domain topology for presenilin 1

Presenilin (PS) provides the catalytic core of the gamma-secretase complex. Gamma-secretase activity leads to generation of the amyloid beta-peptide, a key event implicated in the pathogenesis of Alzheimer disease. PS has ten hydrophobic regions, which can all theoretically form membrane-spanning domains. Various topology models have been proposed, and the prevalent view holds that PS has an eight-transmembrane (TM) domain organization; however, the precise topology has not been unequivocally determined. Previous topological studies are based on non-functional truncated variants of PS proteins fused to reporter domains, or immunocytochemical staining. In this study, we used a more subtle N-linked glycosylation scanning approach, which allowed us to assess the topology of functional PS1 molecules. Glycosylation acceptor sequences were introduced into full-length human PS1, and the results showed that the first hydrophilic loop is oriented toward the lumen of the endoplasmic reticulum, whereas the N terminus and large hydrophilic loop are in the cytosol. Although this is in accordance with most current models, our data unexpectedly revealed that the C terminus localized to the luminal side of the endoplasmic reticulum. Additional studies on the glycosylation pattern after TM domain deletions, combined with computer-based TM protein topology predictions and biotinylation assays of different PS1 mutants, led us to conclude that PS1 has nine TM domains and that the C terminus locates to the lumen/extracellular space.     

Lipids and topological rules of membrane protein assembly: balance between long and short range lipid-protein interactions

The N-terminal six-transmembrane domain (TM) bundle of lactose permease of Escherichia coli is uniformly inverted when assembled in membranes lacking phosphatidylethanolamine (PE). Inversion is dependent on the net charge of cytoplasmically exposed protein domains containing positive and negative residues, net charge of the membrane surface, and low hydrophobicity of TM VII acting as a molecular hinge between the two halves of lactose permease (Bogdanov, M., Xie, J., Heacock, P., and Dowhan, W. (2008) J. Cell Biol. 182, 925-935). Net neutral lipids suppress the membrane translocation potential of negatively charged amino acids, thus increasing the cytoplasmic retention potential of positively charged amino acids. Herein, TM organization of sucrose permease (CscB) and phenylalanine permease (PheP) as a function of membrane lipid composition was investigated to extend these principles to other proteins. For CscB, topological dependence on PE only becomes evident after a significant increase in the net negative charge of the cytoplasmic surface of the N-terminal TM bundle. High negative charge is required to overcome the thermodynamic block to inversion due to the high hydrophobicity of TM VII. Increasing the positive charge of the cytoplasmic surface of the N-terminal TM hairpin of PheP, which is misoriented in PE-lacking cells, favors native orientation in the absence of PE. PheP and CscB also display co-existing dual topologies dependent on changes in the charge balance between protein domains and the membrane lipids. Therefore, the topology of both permeases is dependent on PE. However, CscB topology is governed by thermodynamic balance between opposing lipid-dependent electrostatic and hydrophobic interactions.     

TM domain swapping of murine leukemia virus and human T-cell leukemia virus envelopes confers different infectious abilities despite similar incorporation into virions.

We investigated the influence of transmembrane protein (TM) domains on incorporation of retroviral envelopes into virions and on infectivity. We introduced complete, truncated, or chimeric Friend murine leukemia virus (F-MuLV) and human T-cell leukemia virus type 1 (HTLV-1) envelopes into an MuLV particle-producing complementation cell line. As shown previously for HTLV-1 envelopes containing extracellular domains of F-MuLV TM (C. Denesvre, P. Sonigo, A. Corbin, H. Ellerbrok, and M. Sitbon, J. Virol. 69:4149-4157, 1995), reverse chimeric F-MuLV envelopes containing the extracellular domain of HTLV-1 TM were not processed. In contrast, a chimeric MuLV envelope containing the entire HTLV membrane-spanning and cytoplasmic domains (FHTMi) was efficiently processed, fusogenic as tested in a cell-to-cell assay, and efficiently incorporated into MuLV particles. However, these MuLV particles bearing FHTMi envelope proteins could not infect mouse or rat cells which are susceptible to wild-type F-MuLV. Therefore, envelopes which are readily fusogenic in cell-to-cell assays and also efficiently incorporated into virions may not necessarily confer virus-to-cell fusogenicity. HTLV envelopes, whether parental, chimeric (containing the MuLV cytoplasmic tail) or with a truncated cytoplasmic domain, were incorporated into MuLV particles with equal efficiencies, indicating that the cytoplasmic tails of these envelopes did not determine their incorporation into virions. In contrast to FHTMi envelope, HTLV-1 envelopes with F-MuLV membrane-spanning and cytoplasmic domains, as well as wild-type HTLV-1 envelopes, conferred virion infectivity. These results help to define requirements for envelope incorporation into retroviral particles and their cell-free infectivity.     

Misfolding of mutant aquaporin-2 water channels in nephrogenic diabetes insipidus

We reported that several aquaporin-2 (AQP2) point mutants that cause nephrogenic diabetes insipidus (NDI) are retained in the endoplasmic reticulum (ER) of transfected mammalian cells and degraded but can be rescued by chemical chaperones to function as plasma membrane water channels (Tamarappoo, B. K., and Verkman, A. S. (1998) J. Clin. Invest. 101, 2257-2267). To test whether mutant AQP2 proteins are misfolded, AQP2 folding was assessed by comparative detergent extractability and limited proteolysis, and AQP2 degradation kinetics was measured by label-pulse-chase and immunoprecipitation. In ER membranes from transfected CHO cells containing [(35)S]methionine-labeled AQP2, mutants T126M and A147T were remarkably detergent-resistant; for example wild-type AQP2 was >95% solubilized by 0.5% CHAPS whereas T126M was <10% solubilized. E258K, an NDI-causing AQP2 mutant which is retained in the Golgi, is highly detergent soluble like wild-type AQP2. The mutants and wild-type AQP2 were equally susceptible to digestion by trypsin, thermolysin, and proteinase K. Stopped-flow light scattering measurements indicated that T126M AQP2 at the ER was fully functional as a water channel. Pulse-chase studies indicated that the increased degradation rates for T126M (t((1)/(2)) 2.5 h) and A147T (2 h) compared with wild-type AQP2 (4 h) involve a brefeldin A-resistant, ER-dependent degradation mechanism. After growth of cells for 48 h in the chemical chaperone glycerol, AQP2 mutants T126M and A147T became properly targeted and relatively detergent-soluble. These results provide evidence that NDI-causing mutant AQP2 proteins are misfolded, but functional, and that chemical chaperones both correct the trafficking and folding defects. Strategies to facilitate protein folding might thus have therapeutic efficacy in NDI.     

AQP1 is not only a water channel

AQPs are water channel proteins. In particular, AQP1 was demonstrated to be involved in cell migration. According to the model proposed by Verkman and collaborators, AQP drives water influx, facilitating lamellipodia extension and cell migration. Investigating the possible connection between AQP1 and cytoskeleton, our group showed that such a water channel through Lin7/beta catenin affects the organization of the cytoskeleton and proposed a model. All together, these data appear particularly intriguing since the use of AQP1 as target might be useful to modulate angiogenesis/vasculogenic mimicry     

Characterization of the adenovirus E3 protein that down-regulates the epidermal growth factor receptor. Evidence for intermolecular disulfide bonding and plasma membrane localization.

We have characterized the biosynthesis and processing of a 91 amino acid hydrophobic integral membrane protein encoded by human group C adenoviruses which down-regulates the EGF receptor (Carlin, C. R., Tollefson, A. E., Brady, H. A., Hoffman, B. L., and Wold, W. S. M. (1989) Cell 57, 135-144). Previous studies have shown that two immunologically related proteins are produced in vivo, a 13.7-kDa protein encoded by E3 message f and a 11.3-kDa protein derived from 13.7 kDa by proteolysis (Hoffman, B. L., Ullrich, A., Wold, W. S. M., and Carlin, C. R. (1990) Mol. Cell. Biol. 10, 5521-5524; Tollefson, A. E., Krajcsi, P., Yei, S., Carlin, C. R., and Wold, W. S. M. (1990) J. Virol. 64, 794-801). We report here that the 13.7- and 11.3-kDa proteins form intermolecular disulfide bonds cotranslationally at Cys-31 and tend to migrate as high molecular weight aggregates under nonreducing conditions. Both proteins are also present at the cell surface, as evidenced by specific immunoprecipitation from intact monolayers enzymatically labeled with 125I. Moreover, an antiserum specific for a putative extracellular epitope recognizes the same viral proteins as antibodies directed against a C-terminal synthetic 15-mer. The 13.7- and 11.3-kDa proteins are detected at early time points during pulse-chase radiolabeling of infected cells, do not undergo any further changes in molecular weight, and focus at their predicted isoelectric points (7.4 and 7.2, respectively). Identical results are obtained in stable transfectants constitutively expressing only 13.7 and 11.3 kDa, suggesting that biosynthesis and processing is not dependent on other viral proteins. These results have been incorporated into a computer-based model to predict the orientation of 13.7 and 11.3 kDa in the lipid bilayer. This model provides a basis for testing predictions regarding the topology of the viral proteins, as well as putative interactions with heterologous proteins in the microenvironment of the plasma membrane that cause down-regulation of the epidermal growth factor receptor.