ABSTRACT
The early transcription unit 3 (E3) of human adenoviruses (Ads) encodes proteins with various immunomodulatory functions. Ads from different subgenera differ considerably in their E3 coding capacity, suggesting that distinct sets of immunomodulatory E3 proteins may influence the disease pattern associated with different Ad subgenera. Interestingly, the E3 region of Ads classified in subgenus D, which are often isolated from AIDS patients and have the propensity to cause eye infections, contains a unique gene, named E3/49K, that may encode a protein with a calculated molecular weight of 48,984 that might be implicated in diseases caused by this subgenus. The 49K sequence predicts a highly glycosylated type I transmembrane protein with a short cytoplasmic tail containing two motifs, YXXΦ and LL, potentially involved in targeting the protein to endosomal or lysosomal compartments. Remarkably, the 49K protein is predicted to contain an unusual immunoglobulin-like fold. Here we have characterized the E3/49K protein of Ad type 19a, an Ad of subgenus D which causes epidemic keratoconjunctivitis. E3/49K was synthesized as an 80- to 100-kDa protein, which is unusually large for an E3 protein. In contrast to another early protein, E3/19K, the expression of E3/49K started early but continued throughout the infection cycle. Analysis of the 49K glycosylation revealed that the majority of 49K molecules contained only 12 of the predicted 14 N-glycans. Furthermore, we provide evidence that 49K is O-glycosylated. At steady state, E3/49K was localized in the Golgi-trans-Golgi network and in early endosomes. Interestingly, the 49K protein has a rather short half-life and seems to be proteolytically cleaved. A processing pattern similar to that in the early stages of infection is seen in transfected cells, constitutively expressing 49K in the absence of other Ad proteins. Together, our data provide the first biochemical and cell biological characterization of an unique E3 protein of subgenus D Ads.
Adenoviruses (Ads) are nonenveloped viruses with a double-stranded DNA genome of approximately 36 kb. So far, 51 different human serotypes have been described, which are classified into 6 subgenera, designated A to F (17, 63). Ads can establish acute as well as persistent infections (22, 32). Diseases induced by Ads are in general mild. Infections might remain even subclinical, but they also can be severe or fatal, especially in immunocompromised patients. The pattern of Ad-associated diseases differs depending on the subgenus. Ads of subgenus A and F cause gastrointestinal diseases, while subgenus B and C Ads are mainly associated with acute respiratory diseases. Subgenus B Ads also can induce infections of the urinary tract. Ad type 4 (Ad4), the only member of subgenus E, has been identified as a predominant causative agent of acute respiratory disease epidemics among military recruits. The large number of subgenus D Ads commonly exhibit a tropism for the eye, and three serotypes, Ad8, Ad19a, and Ad37, cause epidemic keratoconjunctivitis (EKC), a highly contagious and severe eye disease with typical subepithelial corneal infiltrates (37, 43).
The molecular basis for the rather distinct pathogenesis of different Ad serotypes affecting different tissues is poorly understood. However, it is unlikely to originate solely from differential receptor usage. Ads of all subgenera with the exception of subgenus B can bind to host cells via the coxsackievirus Ad receptor (CAR), which is expressed in most tissues in vivo (8, 9, 57, 58, 68). In addition, the α2 domain of major histocompatibility complex (MHC) class I molecules, expressed in all nucleated cells, was proposed as a low-affinity receptor for Ad5 (15, 30). Interestingly, while the subgenus D Ads Ad9, Ad15, and Ad19p were shown to attach to CAR (57), the EKC-causing members of subgenus D, Ad8, Ad19a, and Ad37, seem to utilize α(2→3)-linked sialic acid but not CAR (3, 4). Although striking, this common feature of EKC-causing Ads cannot explain their distinct pathogenesis in the eye, since sialic acid is expressed on virtually all cell types of the body. Thus, the basis for the distinct disease patterns most likely has to be sought at the postattachment level (4, 11, 52).
In this regard, the early transcription unit 3 (E3) may play an important role, since it is not required for Ad replication in vitro, but is present in all human Ads. Previous studies clearly demonstrated that the E3 region encodes proteins with immunomodulatory functions, which might be the basis for immune evasion and the establishment of persistent infections (11, 14, 24, 46, 70). Studies on E3 proteins of Ad2 and Ad5 of subgenus C revealed several molecular mechanisms as to how these viruses might escape recognition and elimination by the host immune system. For example, E3/19K retains MHC class I molecules in the endoplasmic reticulum (ER) and thereby prevents the presentation of viral peptides to cytotoxic T lymphocytes and the subsequent lysis of infected cells (2, 10, 12). The E3/14.7K protein interacts with several cellular proteins involved in the tumor necrosis factor alpha (TNF-α)-induced signaling cascade and can protect infected cells from TNF-α-induced cell death (31, 46, 70). A complex of the E3 proteins 10.4K and 14.5K, also called RID (receptor internalization and degradation), down-regulates the epidermal growth factor receptor from the cell surface and has also been implicated in conferring resistance against TNF-α cytolysis (11, 70). In addition, these proteins dramatically reduce the expression of the apoptosis receptor Fas on the cell surface of infected and transfected cells (64) by inducing its internalization and degradation in lysosomes (21, 66). Recently, data have been presented suggesting that 10.4K and 14.5K act in concert with 6.7K to down-regulate TRAIL receptor 1 and TRAIL receptor 2, two other death receptors possibly involved in the apoptosis induction of infected cells (7).
Interestingly, the size and composition of the E3 regions of Ads exhibit substantial subgenus-specific variations (Fig. 1A). Whereas in subgenus F (Ad40/41) about 3 kb encodes five open reading frames (ORFs), in subgenus D Ads (e.g., Ad19a) approximately 5,200 bp encodes eight ORFs. Some genes (e.g., 10.4K, 14.5K, and 14.7K) are found in all or in the majority of subgenera (e.g., 19K in subgenera B to E), while others seem to be unique to a particular subgenus (Fig. 1A), e.g., 6.7K and 11.6K of subgenus C (11, 14). So far, no functions have been assigned to E3 proteins of Ads other than those of subgenus C. Some of them have been characterized biochemically and were shown to be transmembrane glycoproteins, e.g., the 16K (26) and 20.5K proteins of subgenus B (27, 28) and the 30K protein of Ad4 (subgenus E [44]).
(A) Coding capacity of the E3 region of representative members of subgenus C (Ad2), B (Ad3), D (Ad19a), and E (Ad4), adapted from reference 11. The line on the top denotes the size in base pairs. ORFs are indicated as bars and drawn to scale. The size or name of common ORFs is only given once. The shading code is depicted below the figure. Significant overall homology (similarity > 25%) is indicated by identical shading. Homology to a portion of a protein was neglected. pVIII is not an E3 protein, but part of its sequence overlaps with the E3 promoter. (B) Amino acid sequence of the Ad19a E3/49K protein. The putative signal sequence, the transmembrane region, the predicted N-glycosylation sites (★), O-glycosylation sites (▪) (25), and cysteine residues possibly involved in disulfide bonds (↓) are indicated. (C) Structural model of the Ad19a E3/49K protein. The luminal part is suggested to contain three domains with one disulfide bond each. The previously noted repeat structures R1 to R3 with the respective amino acid positions in parentheses are marked as black bars (18). Predicted N- and O-glycosylation sites are indicated as grey and black circles, respectively. Proposed disulfide-bonded loops are shown as open circles, and the YXXΦ and dileucine motifs are indicated.
We have previously identified a novel gene in the E3 region of Ad19a, designated E3/49K. This gene, which encodes a putative protein with a calculated molecular weight of 48,984, was not present in Ads of other subgenera (18). We recently showed that the corresponding E3/49K protein is expressed by all subgenus D Ads tested (Blusch et al., submitted for publication). Thus, E3/49K may be implicated in diseases characteristically caused by subgenus D Ads. The 49K sequence predicts a type I transmembrane protein with an N-terminal signal sequence (Fig. 1B and C). The luminal domain contains three stretches of ∼80 amino acids with internal homology (R1 to R3 in Fig. 1C) that exhibit limited similarity to the 20.1K and 20.5K proteins of subgenus B (18). Each repeat region contains two cysteine residues, possibly involved in disulfide bonding (Fig. 1B and C). The short cytoplasmic tail displays two motifs, YXXΦ and LL (Fig. 1B and C), implicated in endosomal or lysosomal targeting (29, 38, 48). Remarkably, the E3/49K sequence contains 14 potential sites for N- and three for O-glycosylation (25).
As a first attempt to gain insight into its function, we now report on the biochemical characterization of the Ad19a E3/49K protein. We show that indeed not all of the N-glycosylation sites are utilized and provide evidence for the presence of O-linked carbohydrates. E3/49K is abundantly synthesized in the early phase of infection but is also produced in the late phase of infection. The steady-state localization of E3/49K in the Golgi-trans-Golgi network (TGN) and early endosomal vesicles and the appearance of proteolytically cleaved forms suggest a novel processing pathway for E3 proteins. This processing pathway is fully reproduced upon stable expression of E3/49K in lung epithelial cells. Therefore, the established transfection system can be used to study the protein in further detail without the disturbances induced by the ongoing virus infection. In light of these new data potential functional activities are discussed.
MATERIALS AND METHODS
Cloning of the Ad19a E3/49K gene.Ad19a DNA was isolated as described (18). The E3/49K ORF was amplified by PCR using primers complementary to the 5′ (TTTCGCAGAACCATGAATACAGTG) and 3′ (GACTGAGTCTTAGTAAGAGAAGCTGAG) part of the gene, respectively. The PCR product was initially introduced by TA cloning into pCR3.1 (Invitrogen, Groningen, The Netherlands). The resulting plasmid was cut with EcoRI, and the E3/49K-containing fragment was isolated and cloned into the expression vector pSG5 (Stratagene, Amsterdam, The Netherlands) to obtain pSG5-E3/49K. Molecular biology techniques were performed as described by Sambrook et al. (59). The construct was verified by sequencing of both strands of the E3/49K ORF with the ABIPRISM sequencer (Perkin-Elmer, Weiterstadt, Germany).
Viruses, infection, transfection, and cell culture.The lung epithelial carcinoma cell line A549 (ATCC CCL-185) was used for growing Ad19a stocks, for plaque assays (49), and for most infection experiments. In some infection experiments human foreskin fibroblasts (SeBu) were used (21). Both cell types were maintained in Dulbecco’s modified Eagle’s medium supplemented with penicillin (100 U/ml), streptomycin (0.1 mg/ml; Gibco BRL, Karlsruhe, Germany) and 10% fetal calf serum (FCS) (Roche Diagnostics, Mannheim, Germany). In general, infection experiments were performed by applying 5 PFU of Ad19a per cell for 1 h. Then the virus was removed, and this time point was defined as the start of infection.
To create transfectants stably expressing the E3/49K protein, A549 cells were transfected with the pSG5-E3/49K construct together with the pSV2-neor plasmid conferring G418 resistance by using the calcium phosphate method as described (41). G418-resistant clones were isolated and maintained in medium containing 1 mg of G418 per ml.
Metabolic labeling, immunoprecipitation, and SDS-PAGE.Labeling of A549 cells with [35S]methionine alone or in combination with [35S]cysteine (Promix; Amersham Pharmacia Biotech, Freiburg, Germany), immunoprecipitation, and sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) were performed as described previously (11, 12). Dried gels were exposed to BioMaxMR films (Kodak, Rochester, N.Y.) at −70°C. As molecular weight markers 14C-methylated proteins were used (Amersham Pharmacia Biotech). Radioactive bands were quantified using a Storm 860 Molecular Imager (Molecular Dynamics, Sunnyvale, Calif.).
Glycosidase treatments.For endoglycosidase H (endo H) treatment, the protein A-Sepharose pellet with the immunoprecipitated protein was resuspended in 50 μl of reaction buffer (0.1 M sodium citrate, pH 5.5) and incubated for 24 h at 37°C with 5 mU of endo H from Streptomyces plicatus (Roche Diagnostics). The mock-treated samples were incubated under the same conditions without endo H. In the partial digestion experiment only 2 mU of endo H was used for each reaction, and the incubation time was shortened to the times indicated. Treatment with neuraminidase, O-glycosidase, and peptide-N-glycosidase F (PNGase F) was carried out sequentially. The protein A-Sepharose beads containing the immunoprecipitated protein were resuspended in 50 μl of reaction buffer (50 mM sodium phosphate, pH 7.2) and were first incubated for 6 h at 37°C with 20 mU of neuraminidase from Arthrobacter ureafaciens (Roche Diagnostics, Mannheim, Germany) and/or 2 mU of O-glycosidase from Diplococcus pneumoniae (Roche Diagnostics). Subsequently, the pellets were washed, resuspended in 12.5 μl of 0.1 M β-mercaptoethanol-0.5% SDS, and incubated at 95°C for 5 min. Reaction buffer (12.5 μl) including 1 U of PNGase F of Flavobacterium meningosepticum (Roche Diagnostics) was added to give final concentrations of 0.2 M Tris (pH 8), 0.02 M EDTA, and 2% IGEPAL (NP-40 replacement; Sigma, Taufkirchen, Germany), and the sample was incubated for 20 h at 37°C. Mock-treated samples were incubated under the same conditions without PNGase F.
Immunofluorescence.Subconfluent layers of A549 or SeBu cells were grown on glass coverslips. Cells were rinsed with phosphate-buffered saline (PBS) and fixed with 3% (wt/vol) paraformaldehyde in PBS for 20 min at room temperature. After quenching aldehyde groups with 50 mM NH4Cl and 20 mM glycine in PBS, cells were permeabilized with 0.2% saponin in PBS with 5% FCS to block nonspecific binding. The cells were incubated with the primary antibody diluted in 0.2% saponin-5% FCS in PBS for 1 h at room temperature, washed four times in the above buffer without FCS, and incubated with the secondary antibody (fluorescein- or rhodamine-conjugated goat or donkey anti-mouse, anti-rabbit, or anti-sheep immunoglobulin G, respectively; Dianova, Hamburg, Germany) for 1 h. After four further washing steps the coverslips were mounted on glass slides with Histogel (Linaris, Wertheim-Bettingen, Germany). The mounted cells were analyzed with a laser scanning confocal microscope (Leitz DM IRB; scanner, Leica TCS NT).
MAbs and antisera.The following antibodies were used in this study: polyclonal rabbit antisera R25050 and R25044, raised against the C-terminal 15 amino acids of E3/49K (Blusch et al., submitted); rabbit anti-Ad19a E3/19K (18); anti-TGN46 sheep serum (Serotec-Biozol, Munich, Germany) (and kindly provided by S. Ponnambalam, University of Dundee, Dundee, Scotland); rabbit serum 931-A against lysosomal membrane protein 1 (lamp-1) (kindly provided by S. Carlson, University of Umea, Umea, Sweden); and the monoclonal antibodies (MAbs) GTL2 against human β(1→4)-galactosyltransferase (36), 2Hx-2 (ATCC HB-8117) against the hexon of Ad2, 2D5 against lamp-2 (19), 6C4 against lysobisphosphatidic acid (40), L01.1 (Becton Dickinson, Heidelberg, Germany) and 5E9C11 (ATCC HB-21) against the transferrin receptor, J4-48 against CD46 (Dianova), 100/3 against AP1 (Sigma), W6/32 (ATCC HB95) against HLA-ABC, and clone 35 against GM130 and clone 14 against EEA1 (Transduction Laboratories, Lexington, Ky.).
RESULTS
Ad19a E3/49K synthesis starts early during infection and continues throughout the infection cycle.We previously identified a novel gene within the E3 region of the EKC-causing Ad strain Ad19a, designated E3/49K (18), and recently showed that it is specific for subgenus D Ads (Blusch et al., submitted). As E3/49K is encoded by E3, it is predicted to be expressed in the early phase. However, expression of some E3 proteins is greatly amplified at late times (67). Therefore, we monitored the expression level of E3/49K during the course of infection in comparison to E3/19K, another early protein, and the late protein hexon (Fig. 2). A549 cells were labeled for 1 h with [35S]methionine at different time points postinfection (p.i.), and 49K was immunoprecipitated using the peptide antiserum directed to the putative cytoplasmic tail of the protein. After SDS-PAGE analysis several bands could be visualized, presumably representing differentially glycosylated forms of E3/49K: three defined bands (b to d) representing proteins with apparent molecular masses ranging from 77 to 83 kDa and a diffuse band (a) representing proteins of about 87 to 100 kDa (Fig. 2A). Expression of both E3/49K and E3/19K was first detected at 3 h p.i. and reached a maximum at 6 h p.i. In contrast to the E3/19K expression (Fig. 2C), which declined rapidly, E3/49K was expressed in significant amounts also in the late phase of infection. Interestingly, the banding and glycosylation pattern of E3/49K seemed to change during the course of infection. The intensity of the diffuse band (a) decreased, and concomitantly the intensity of one of the lower bands (b) increased in the late phase of infection (>18 h p.i.). To investigate whether these changes are due to a differential carbohydrate processing, we performed a pulse-chase analysis early (6 h p.i.) (Fig. 3A) and late (30 h p.i.) (Fig. 3B) during infection. In the early phase of infection, the E3/49K species defined by bands b to d (Fig. 3A, lane 1) were converted into higher-molecular-weight forms (band a in Fig. 3A, lanes 2 to 7), presumably representing terminally glycosylated 49K products. In contrast, in the late stages of infection this type of processing was severely impaired (Fig. 3B). We next examined whether the processing defect observed for E3/49K is a more common phenotype by monitoring the processing of two cellular proteins, lamp-1 and the transferrin receptor. Interestingly, while lamp-1 processing was also affected during the late phase of infection, that of the transferrin receptor was normal (data not shown). Therefore, the altered processing was not a general phenomenon but rather seemed to be restricted to a selective set of glycoproteins. Remarkably, not only the glycosylation pattern but also the half-life of Ad19a E3/49K changed from ∼2 h early during infection (6 h p.i.) to ∼4 h late during infection (30 h p.i.).
Time course of Ad19a E3/49K synthesis during infection. A549 cells were labeled with [35S]methionine at different time points p.i. The start of infection was defined as the end of the 1-h adsorption period. The indicated times given on top of the figure correspond to the start of the 1-h labeling period. Lysates with equal amounts of radioactivity were used for immunoprecipitation. E3/49K (A); E3/19K (C), another early protein; and hexon (B), a late protein, were immunoprecipitated sequentially from the same lysates using the antibodies given in Materials and Methods. E3/49K is visualized by three defined bands with apparent molecular masses of 77 to 83 kDa, designated b, c, and d, and a diffuse band of 87 to 100 kDa, designated a. E3/19K is represented by six bands differing in the number of N-glycans attached (18).
Differential processing of E3/49K in the early and late phase of infection as revealed by pulse-chase analysis carried out at 6 h p.i. (early) (A) or at 30 h p.i. (late) (B). A549 cells were metabolically labeled for 20 min with [35S]methionine/cysteine (200 μCi/ml) and subsequently chased with medium containing nonradioactive methionine/cysteine for the indicated periods of time. Ad19a E3/49K protein was immunoprecipitated and analyzed by SDS-PAGE as in Fig. 2. Bands a to d represent different forms of E3/49K. The arrowhead indicates the position of the abundant hexon protein that coprecipitates unspecifically with any antiserum.
Fully processed E3/49K protein contains high-mannose and/or hybrid and complex N-glycans.The difference between the calculated molecular mass of the protein backbone of 46,915 kDa (without signal sequence) and the apparent molecular mass of ∼80 to 100 kDa in SDS-PAGE suggested that E3/49K is highly glycosylated. The early appearance of the three homogeneous E3/49K species b to d in the pulse-labeled sample (Fig. 3A, lane 1) indicated that they represent the high-mannose forms of the protein, which are processed to more heterogeneous protein species (represented by the diffuse upper band a), supposedly containing complex glycans (Fig. 3A).
To verify this assumption, the conversion of N glycans was directly assessed in a pulse-chase experiment using endo H (Fig. 4). Endo H cleaves high-mannose and hybrid but not complex carbohydrates. A549 cells were labeled at 6 h p.i., and immunoprecipitated E3/49K was treated with endo H. Treatment of the pulsed sample resulted in a decrease of the apparent molecular mass and in the appearance of two prominent protein species of about 50 and 53 kDa, respectively (Fig. 4, compare lane 2 with lane 10, bands f and g), demonstrating that only high-mannose and/or hybrid N-glycans were present. The 53-kDa species arises as a result of incomplete endo H digestion and was not observed after endo H treatment under denaturing conditions or PNGase F treatment (Fig. 6, lane 11). Therefore, it seems to represent E3/49K containing one uncleaved N-linked glycan. Parallel to the appearance of the diffuse upper band in the untreated samples (20- to 240-min chase, Fig. 4, lanes 3 to 8), E3/49K forms of 67 to 92 kDa were visualized in the endo H-treated samples (band e in Fig. 4, lanes 11 to 16). As these forms migrated faster than the heterogeneous species of 87 to 100 kDa in the untreated samples, the processed forms of E3/49K were not completely resistant to endo H. Therefore, processed E3/49K contained obviously apart from complex N-linked oligosaccharides and/or O-glycans, not cleavable by endo H, also high-mannose and/or hybrid N-linked sugars that were cleaved by endo H. Based on the apparent molecular mass difference between the mature endo H-treated and untreated forms (8 to 14 kDa), we estimate that three to five endo H-sensitive glycans have been removed.
Mature E3/49K molecules remain partially endo H sensitive. A549 cells were infected with Ad19a. At 6 h p.i. cells were labeled for 20 min with [35S]-methionine and then chased for the time periods indicated. Two culture dishes were mock infected (lanes 1 and 9). For each time point, two culture dishes were lysed, and the lysates were pooled and then split into two aliquots for E3/49K immunoprecipitation: one sample was treated with endo H for 24 h (lanes 9 to 16), the other one was mock treated (lanes 1 to 8). Band f corresponds to 49K molecules containing a residual oligosaccharide chain (53 kDa) which is obviously not efficiently removed under the nondenaturing conditions used. When 49K immunoprecipitates were denatured prior to endo H treatment or incubated with endo F (data not shown), only the 50-kDa form (g) and the heterogeneous species of 67 to 92 kDa, designated e, were visualized.
Interestingly, during the chase small fragments of 12 to 13 kDa (band h) were precipitated in increasing amounts, whereas the 87- to 100-kDa band of E3/49K gradually disappeared. Western blot analysis with the 49K-specific serum demonstrated that these fragments were derived from the C terminus of E3/49K and did not represent coprecipitated proteins (data not shown).
Only 12 of the 14 predicted N-glycosylation sites of E3/49K are predominantly utilized.To investigate how many of the 14 predicted N-glycosylation sites are actually utilized, we immunoprecipitated E3/49K from Ad19a infected A549 cells after metabolic labeling for 20 min. During this short labeling period exclusively high-mannose N-glycans should be present, which are fully sensitive to endo H. Upon incubation with endo H for different time periods and subsequent analysis of the samples by SDS-PAGE, partially deglycosylated forms of E3/49K could be distinguished that differed by one glycan each (Fig. 5). By counting the various bands, it is obvious that a fraction of E3/49K migrating with the highest apparent molecular mass (∼17%) contained 13 N-linked glycans, but the great majority of E3/49K proteins (65%) had only 12 carbohydrates attached. Therefore, not all of the 14 predicted N-glycosylation sites were utilized.
Not all of the predicted N-glycosylation sites are utilized. E3/49K was immunoprecipitated after 20 min of metabolic labeling of Ad19a-infected A549 cells and incubated with endo H for the times indicated at the top of the figure. Digested material was separated by SDS-PAGE, resulting in a series of 49K molecules differing by one oligosaccharide chain. The number of attached N-glycans for each 49K species is indicated on the right.
Evidence for O-glycosylation of E3/49K.The pulse-chase analysis of E3/49K (Fig. 4) demonstrated that the protein became partially endo H resistant, indicating the presence of complex N-glycans and/or O-glycans. According to the NetOGlyc program (25), E3/49K is predicted to contain three O-glycosylation sites (Fig. 1B and C). To experimentally substantiate this prediction, we pulse-labeled cells for 1 h and chased for an additional hour to potentially allow acquisition of O-linked glycans in the Golgi complex. Subsequently, the protein was immunoprecipitated and treated with various glycosidases, including O-glycosidase from D. pneumoniae (Fig. 6). This enzyme can release the disaccharide core Galβ(1→3)GalNAc of O-glycans only if no other groups, e.g., sialic acid or fucose, are attached. While neuraminidase reduced the apparent molecular mass of 49K, no shift was observed upon O-glycosidase treatment (Fig. 6, lanes 3 to 6). To better assess small molecular weight changes associated with the glycosidase treatments, all N-linked sugars were removed by treating immunoprecipitated E3/49K with PNGase F (Fig. 6, lanes 7 to 11). By cleaving N-linked glycans, this enzyme converts the respective asparagine residues into aspartic acid residues, thereby changing the migration behavior of the protein due to the added negative charges. As a reference for potential modifications occurring during the chase, E3/49K was also immunoprecipitated from pulse-labeled cells (20 min) and treated with PNGase F. This generated an E3/49K molecule (k) without any glycans, but containing aspartates instead of asparagines (Fig. 6, lane 11). Its apparent molecular mass of 50 kDa differed from that of unglycosylated 49K after tunicamycin treatment, which migrates at 48 kDa (Fig. 7, and data not shown). In comparison to this PNGase F-treated reference (Fig. 6, lane 11, band k), two E3/49K species (band i) could be differentiated after PNGase F treatment of the chased protein that migrated more slowly (53 and 54 kDa) than the pulsed PNGase F treated form (band k) (Fig. 6, compare lane 7 with lane 11). This indicated the presence of a posttranslational modification in addition to N-glycosylation. The reduction of the apparent molecular mass of the PNGase F-treated form (band i) upon neuraminidase treatment (band j, lane 9) clearly demonstrated the presence of residual glycan structures containing sialic acid, indicating O glycosylation. Again, the utilized O-glycosidase had no effect, either when used alone or in combination with neuraminidase (Fig. 6, lanes 8 and 10), although it did decrease the apparent molecular mass of CD46 (45), a highly O-glycosylated protein (data not shown). The finding that neuraminidase treatment caused a shift after all N-glycans had been removed can only be explained by the presence of O-glycans containing sialic acids. Thus, the lack of sensitivity for the O-glycosidase was either due to the presence of O-glycan core structures containing additional modifications or due to the inaccessibility of the disaccharide core Galβ(1→3)GalNAc, e.g., due to steric hindrance by the N-glycans.
Evidence for O-glycosylation of E3/49K. Ad19a-infected A549 cells (lanes 2 to 10) were labeled for 1 h and chased for an additional hour with medium containing an excess of unlabeled methionine/cysteine. One culture was mock infected (lane 1). To obtain a source of unprocessed 49K, Ad19a-infected A549 cells were in addition labeled for 20 min with [35S]methionine/cysteine and lysed (lane 11). Immunoprecipitated E3/49K was incubated with different glycosidases as indicated on top of the figure and analyzed by SDS-PAGE. Endo H cleaves only high-mannose and hybrid N-linked glycan structures. PNGase F removes all N-linked glycans. The O-glycosidase from D. pneumoniae cleaves only the unmodified disaccharide Gal(β1→3)GalNAc, one of the core structures of O-glycans bound to Ser/Thr. After PNGase F treatment, the pulsed protein is predominantly converted to a protein species migrating with an apparent molecular mass of about 50 kDa (lane 11, designated k), whereas the processed form of E3/49K shifts to two species with ∼53 to 54 kDa (band i, lane 7). Additional treatment with neuraminidase decreases their apparent molecular mass to ∼52 kDa (band j, lane 9).
E3/49K contains intramolecular disulfide bonds. Ad19a-infected A549 cells were preincubated in methionine/cysteine-free medium containing (lanes 1 and 2) or not containing tunicamycin (10 μg/ml) (lanes 3 and 4). After 1 h of metabolic labeling with [35S]methionine/cysteine in the presence (lanes 1 and 2) or absence (lanes 3 and 4) of tunicamycin (10 μg/ml), E3/49K was immunoprecipitated. Immunoprecipitates were heated for 5 min to 95°C in the presence (lanes 2 and 4) or absence (lanes 1 and 3) of the reducing agent dithiothreitol (DTT). SDS-PAGE analysis was done as described (12).
E3/49K contains intramolecular disulfide bonds.We further investigated whether the six cysteine residues in the luminal domain of E3/49K are involved in disulfide bond formation (Fig. 1B and C). E3/49K was immunoprecipitated and analyzed by SDS-PAGE in the presence or absence of the reducing agent dithiothreitol (Fig. 7). To evaluate the potential apparent molecular mass changes more easily, the experiment was also performed in the presence of tunicamycin, an inhibitor of N-glycosylation (Fig. 7, lanes 1 and 2). In both cases, no higher-molecular-weight products indicative for disulfide-linked E3/49K multimers were detected in the nonreduced samples. Rather, E3/49K migrated faster without dithiothreitol treatment, demonstrating the presence of intramolecular disulfide bonds. The difference in the migration behavior corresponded to ∼1 kDa. Compared to the altered migration, corresponding to about 1.2 kDa, seen upon reduction of the MHC class I heavy chain containing two known disulfide loops of ∼53 and 62 amino acids (references 13 and 62 and data not shown), the relatively small shift suggests that no large disulfide bonded loops are formed (e.g., between Cys 1/2 and Cys 5/6; Fig. 1B and C). Thus, it is likely that the six cysteine residues in the luminal domain form three disulfide bonds linking neighboring cysteines and creating loops of 13 and 9 amino acids, respectively (Fig. 1C). This interpretation is consistent with the proposed division of the luminal part of E3/49K into three domains (R1 to R3, Fig. 1C) (18).
The processing of E3/49K seen in the early phase of infection is reproduced in transfected cells.To examine the influence of viral infection on the processing of E3/49K, an A549 cell line was established stably expressing the protein. Immunoprecipitation and SDS-PAGE analysis revealed a similar processing pattern of 49K as in the early phase of Ad19a infection (compare bands a to d in Fig. 8 with those in Fig. 3A and 4). Likewise, the N- and O-glycosylation pattern of E3/49K in the transfected cell line was identical to that seen in the early phase of viral infection (data not shown). Moreover, as in infected cells (Fig. 4), low-molecular-mass fragments of 12 to 13 kDa (h) appeared during the chase of the transfected cell line (Fig. 8). Thus, neither the early processing pattern nor the generation of cleaved C-terminal 49K fragments depends on other Ad-encoded products. Therefore, they represent intrinsic properties of the E3/49K protein.
The processing of E3/49K synthesized from a transfected A549 cell line is identical to that seen during early times of infection. A cell line stably expressing the 49K protein under the control of the simian virus 40 promoter was labeled for 30 min, and E3/49K was immunoprecipitated (lanes 2 to 7) as described in the legend to Fig. 4. The bands representing the various E3/49K forms are marked as before (Fig. 3 and 4). As a control, lysates of untransfected A549 cells were reacted with 49K-specific antibodies (lane 1).
E3/49K localizes to the Golgi-TGN and early endosomal vesicles.As a first step to gain insight into the potential function of E3/49K, we sought to determine the intracellular localization of E3/49K by performing immunofluorescence with the polyclonal rabbit serum directed to the cytoplasmic tail of E3/49K. We chose infected primary fibroblasts for this analysis, since intracellular compartments can be better distinguished in such cells. Ad19a-infected SeBu cells were processed for immunofluorescence at 12 to 15 h p.i., corresponding to the early phase of infection in these cells. The protein was found in vesicular and tubular perinuclear structures and numerous vesicles in the periphery (Fig. 9). The identity of these intracellular compartments was revealed by colocalization with various cellular marker proteins. Minimal colocalization was observed with the ER marker calnexin (Fig. 9A). The perinuclear structure could be identified as the Golgi-TGN by colocalization with GM130, a marker for the cis-Golgi cisternae (51); β(1→4)-galactosyltransferase, a marker for trans-Golgi cisternae and the TGN (53); and AP-1 (data not shown) and TGN46 (Fig. 9B), both markers for the TGN (1, 54). To distinguish between the Golgi complex and the TGN, we used brefeldin A, a drug reported to have differential effects on proteins of the Golgi complex and the TGN (20, 39). Whereas proteins of the Golgi complex redistribute to the ER, the TGN collapses onto the microtubule-organizing center (42, 55). Upon treatment of Ad19a-infected SeBu cells with brefeldin A, E3/49K staining of the perinuclear region disappeared almost completely and a slightly increased ER staining was detected. However, similar to TGN46, E3/49K was also detected in a perinuclear spot, representing the microtubule-organizing center (data not shown). This suggests that E3/49K at steady-state was present in the TGN as well as in the Golgi complex.
E3/49K localizes to the Golgi-TGN and early endosomes. Primary fibroblasts (SeBu) were infected with Ad19a and processed for confocal laser microscopy 12 to 15 h p.i. Intracellular localization of E3/49K (green) was compared to that of marker proteins (red) for different cellular compartments: calnexin (ER) (A), TGN46 (TGN) (B), EEA-1 (early endosomes) (C), LBPA (late endosomes) (D), and lamp-2 (late endosomes and lysosomes) (E). The antibodies used are given in Materials and Methods.
We also attempted to determine the identity of the E3/49K+ peripheral vesicles. Colocalization was observed neither with the late endosomal marker lysobisphosphatidic acid (LBPA) (Fig. 9D) nor with the lysosomal marker lamp-2 (Fig. 9E), whereas a substantial number of E3/49K+ vesicles did colocalize with EEA1 (Fig. 9C), an early endosomal marker (50). These data suggest that E3/49K is targeted to early endosomes but not to late endosomal or lysosomal compartments under these conditions. Essentially the same intracellular localization of E3/49K was found in A549 cells transiently or stably expressing E3/49K upon transfection. Surprisingly, at later stages of infection colocalization of E3/49K was also observed with lamp-2 (data not shown). At around 18 h p.i., E3/49K was found in some peripheral vesicles that were only weakly lamp-2 positive, while the major fraction of the lamp-2 positive vesicles in the perinuclear area remained E3/49K negative. At later time points (24 to 48 h p.i. of SeBu cells) also this perinuclear lysosomal fraction exhibited a profound E3/49K staining. Thus, either E3/49K accumulates in this compartment with time or trafficking is altered at late stages of infection.
DISCUSSION
In this study we have characterized the E3/49K protein encoded by the E3 region of Ad19a, an Ad of subgenus D associated with EKC. The sequence predicts a type I transmembrane protein with a short cytoplasmic tail containing two motifs, YXXΦ and LL, that have been implicated in endosomal and lysosomal targeting (29, 38). Remarkably, the E3/49K contains 14 potential N-glycosylation sites and three potential O-glycosylation sites (Fig. 1B and C). We have demonstrated that E3/49K is indeed highly N-glycosylated and provide evidence for the presence of O-glycans. The abundant glycosylation resulted in an apparent molecular mass of ∼80 to 100 kDa (Fig. 2 to 6), about twofold that of the calculated molecular weight. The E3/49K protein was initially synthesized with endo H cleavable high-mannose oligosaccharides, migrating with a molecular mass of 77 to 83 kDa, that were processed to the 87- to 100-kDa protein species (Fig. 3 and 4). Endo H treatment shifted these processed E3/49K species to 67 to 92 kDa (Fig. 4, band e), demonstrating the presence of a mixture of high-mannose and/or hybrid and complex N- and/or O-glycans. Limiting endo H treatment of pulsed E3/49K revealed that the majority of the E3/49K proteins (∼65%) contained 12 N-linked oligosaccharides and a small fraction (∼17% each) contained 13 and 11 N-linked oligosaccharides (Fig. 5). Thus, in the largest fraction of E3/49K proteins one or two N-glycosylation sites are not utilized. Three rules are known to govern the usage of potential N-glycosylation sites. Firstly, NX(S/T)X sites are rarely glycosylated if X is a proline residue (23). This does not explain the incomplete glycosylation of E3/49K, since none of the potential N-glycosylation sites of E3/49K contains proline residues in these positions (Fig. 1B). Secondly, steric hindrance may prevent glycosylation if the potential acceptor sites are too close to each other. There are indeed two locations where potential glycosylation sites are in close proximity (N25 and N31, and N312 and N317), which might cross-inhibit the glycosylation of the adjacent acceptor sites. However, it has been shown that two NX(S/T) acceptor sites separated by a single amino acid (distance between the Ns 3 aa) can still be glycosylated at both sites (23). Thirdly, unglycosylated acceptor sites tend to be found more frequently towards the C termini of glycoproteins (23). Thus, it is possible that the site closest to the transmembrane segment (N343) may not be utilized. This is supported by the observation that the C-terminally derived fragments (band h in Fig. 4 and 8) were not glycosylated (data not shown). Considering their apparent molecular mass of 12 to 13 kDa, the cleavage site is predicted to be N terminal of the putative glycosylation site at position 343. Thus, it is likely that the E3/49K form with 13 oligosaccharides lacks a glycan at position 343 and that with 12 glycans might in addition be devoid of carbohydrates at position 312 or 317 (Fig. 1B and C). It is noteworthy that incomplete glycosylation was also observed for E3/19K molecules of Ad19a (18).
Although we could not directly demonstrate the presence of O-glycans with the O-glycosidase from D. pneumoniae that hydrolyzes only the unmodified disaccharide core Galβ(1→3)GalNAc from serine or threonine, the presence of O-glycans was indicated by a shift in the apparent molecular mass upon neuraminidase treatment of E3/49K lacking N-glycans (Fig. 6). Presumably other O-glycan core structures not cleavable by the used O-glycosidase were present. The apparent molecular mass difference between the i and k species accounts for ∼3.5 kDa (Fig. 6, compare lanes 7 and 11). This is consistent with the proposed three O-glycosylation sites, assuming core structures of four to five sugar groups each. Interestingly, two bands appeared after PNGase F treatment (band i in Fig. 6, lane 7), suggesting that E3/49K forms exist that differ either in the number of the utilized O-glycosylation sites or in the sugar side chains attached to these core units. In summary, the processed form of E3/49K contains a few high-mannose and/or hybrid but mostly complex N-glycans as well as O-linked oligosaccharides.
Unexpectedly, the glycosylation pattern changed in the late phase of infection, in which E3/49K forms with high-mannose carbohydrates represented by the defined bands of 77 to 83 kDa were the major species (bands b to d in Fig. 2A). The pulse-chase experiment (Fig. 3) demonstrated that late in infection the processing of the high-mannose to the complex glycans was severely impaired. Presumably, the exit of 49K out of the ER is prevented, and this prolonged residence time in the ER favors the generation of 49K molecules with 13 N-linked glycans. This was not a generalized processing defect caused by Ad infection, e.g., associated with the host shutoff late during infection, since the processing of the transferrin receptor was normal whereas lamp-1 processing was also affected (data not shown). This result suggests that only a selective set of glycoproteins seems to be affected. Lamp-1 and E3/49K are both heavily N-glycosylated with 18 (69) and 11 to 13 (Fig. 5) N-glycans, respectively, whereas the transferrin receptor contains only three N-glycosylation sites (60). The reason for the differential sensitivity of the glycoproteins is unclear, but they may utilize different transport pathways (35), which might be differentially affected late during Ad infection. Interestingly, the Golgi-TGN appeared more vesicular in infected compared to uninfected cells (data not shown). At present, we do not know whether these two phenotypic changes—the altered processing late and the altered morphology of the Golgi, which is visible already in the early phase—are related. However, it is unlikely that E3/49K plays a direct role in this respect. Firstly, transfectants stably expressing E3/49K exhibited a similar or the same processing as observed in the early phase during infection (Fig. 8). Secondly, the processing deficiency is established only late during infection, when E3/49K synthesis is already significantly decreased. Notably, a slower carbohydrate processing was also observed for the 20.5K protein of Ad3 (subgenus B) in the late phase of infection (27), suggesting that this phenomenon is not restricted to subgenus D Ads.
As expected for an E3 protein, expression of E3/49K was initiated in the early phase of infection and reached a maximum at 6 h p.i. (Fig. 2). Unlike E3/19K, E3/49K was also synthesized in significant amounts in the late phase of infection (30 h p.i.). Whereas E3/19K synthesis decreased dramatically to ∼1 to 2% of the maximal value, E3/49K was still produced at a level of ∼25 to 30%. Similar to E3/49K, the synthesis rate of 20.5K of Ad3 (subgenus B) was also shown to be high early during infection and continued at lower levels throughout the infection cycle (27). Remarkably, the synthesis of E3/49K seemed to reach a minimum at 18 h p.i. and increased again slightly at later time points. Some E3 proteins such as 11.6K of Ad2 (subgenus C) and 30K of Ad4 (subgenus E) were shown to be synthesized at higher rates late during infection. In contrast to 49K, 11.6K is scarcely synthesized during the first 12 h of infection and is greatly amplified (∼400-fold) in the late phase (>24 h p.i.) (67). Wold and colleagues showed that at late times the 11.6K mRNA is spliced to the y leader located within the E3 region and the major late tripartite leader (34, 70). Little is known about splicing of E3 transcripts of Ads other than subgenus C. However, it is clear that splicing of the E3 RNA of Ad35 and Ad3 (subgenus B) differs substantially (6, 65). As subgenus D Ads encode a 23K ORF in the equivalent position of the y leader, it is unlikely that this element is utilized in this subgenus. At present, we do not know whether the high expression of E3/49K relative to E3/19K in the late phase is functionally significant. We also do not know whether it is due to differential splicing regulation (early versus late) within the E3 transcription unit or involves splicing to the tripartite leader, a common feature of transcripts originating from the major late promoter. Nevertheless, the presence of a typical splice acceptor site in front of the 49K gene strongly suggests that 49K can be synthesized independently of E3/19K (18).
Another remarkable result from the pulse-chase experiments was the appearance of E3/49K derived fragments with apparent molecular masses of about 12 to 13 kDa starting at about 40 min of chase (band h in Fig. 4). Similar fragments were also observed in the cell line stably expressing Ad19a E3/49K (band h in Fig. 8). This suggests that E3/49K is proteolytically cleaved and that this cleavage does not depend on other viral products. We are presently addressing the question whether the large N-terminal domain is degraded or secreted by raising antibodies against the N-terminal part of E3/49K.
Several structural features of E3/49K, such as the abundant N-glycosylation, the presence of O-linked glycans and intramolecular disulfide bonds, and the short cytoplasmic tail with endosomal or lysosomal targeting signals, are reminiscent of the lysosome-associated membrane proteins (33). Despite these striking similarities the intracellular localization of E3/49K and lamp-2 differed considerably in primary foreskin fibroblasts in the early phase of infection (12 to 15 h p.i.). E3/49K was localized primarily in a perinuclear compartment, which was identified as the Golgi-TGN (Fig. 9B). In addition, numerous E3/49K+ vesicles were detected in the perinuclear region and in the periphery. A significant proportion of these vesicles were shown to be early endosomes (Fig. 9C). Surprisingly, in the early phase no colocalization with the late endosomal marker LBPA (Fig. 9D) and the late endosomal and lysosomal marker lamp-2 (Fig. 9E) was found. This changed in the late phase of infection when colocalization of E3/49K and lamp-2 was evident (data not shown). One explanation might be that rapid degradation of the protein in late endosomes and lysosomes in the early phase might prevent its detection in these compartments and this may be inhibited in the late phase due to unspecific effects of the infection (e.g., host shutoff) on the cellular degradation machinery. However, treatment of infected A549 cells with various acidotropic and lysosomotropic agents (16) and the serine and cysteine protease inhibitor leupeptin stabilized the 49K protein only slightly (data not shown). This suggests that lysosomal degradation of E3/49K may play only a minor role and that accumulation of E3/49K or its proteolytic fragments during the course of infection may be responsible for the observed colocalization with lamp-2 late in infection. As the applied antibody directed against the cytoplasmic tail of E3/49K does not distinguish between the full-length protein and the small E3/49K-derived fragments that exhibit a longer half-life (5 to 6 h) than does full-length E3/49K (∼2 h; data not shown), it is conceivable that these fragments accumulate during the course of infection.
In summary, at steady state E3/49K localizes to the Golgi-TGN and early endosomes. Some unidentified vesicular structures might represent secretory vesicles. In infected cells expressing E3/49K at high levels we also observed a significant staining of the cell surface, indicating that E3/49K may reach the plasma membrane. Cell surface expression was recently confirmed by fluorescence-activated cell sorter analysis, using a newly developed antibody directed against the N-terminal portion of the 49K protein (data not shown). Thus, E3/49K is transported to the cell surface, presumably following the secretory pathway. This indicates that 49K is targeted to early endosomes by internalization from the plasma membrane, possibly involving the YXXΦ and/or the LL motif in the cytoplasmic tail, rather than by direct transport from the Golgi-TGN. We are presently investigating the role of these putative transport motifs for the intracellular localization and the processing of the E3/49K. From the early endosomes it might be recycled to the plasma membrane or to the TGN, which would explain its steady-state localization. A similar trafficking pathway was shown to exist for TGN46, a protein that also localizes to the TGN at steady state. TGN46 also contains YXXΦ and LL motifs and is continuously recycled from the cell surface via early endosomes (47, 54).
Several potential functions can be envisaged based on the presented data. As shown for other Ad E3 proteins, like 10.4 and 14.5K (21, 66) and proteins of other viruses, e.g., murine cytomegalovirus m6 and gp40 (56, 71), E3/49K might target proteins with immunological importance in the Golgi-TGN and/or at the cell surface, inducing their rerouting and/or degradation. Interestingly, a SMART search for homologies to common domain families (61) revealed a similarity of repeat 3 (Fig. 1C) with the consensus sequence for immunoglobulin-like domains. Strikingly, about 54% of all leukocyte membrane proteins contain at least one domain with an immunoglobulin fold, which is mostly involved in protein interactions (5). This finding supports the hypothesis that E3/49K might modulate immunologically relevant proteins and the observed cell surface expression is compatible with the notion that E3/49K might bind to cell surface protein(s) of leukocytes, which was recently confirmed (M. Windheim, C. Falk, E. Kremmer and H.-G. Burgert, unpublished data). Considering the processing and the production of small E3/49K-derived fragments of 12 to 13 kDa, an alternative possibility arises in that the N-terminal domain might be secreted. This would represent a completely novel phenomenon for adenovirus proteins. Work is in progress to examine this interesting question.
ACKNOWLEDGMENTS
We thank K. Siedler and J. Blusch for generously providing the pSG5-E3/49K plasmid and A. Osterlehner for her excellent technical assistance. We also thank W. Muranyi for his expert introduction to the confocal laser microscope. We are grateful to S. Carlson, S. Ponnambalam, A. Hasilik, T. Suganuma, and J. Gruenberg for kindly providing antibodies. For critical reading of the manuscript, we thank A. Hilgendorf, Z. Ruzsics, and M. Sester.
This work was supported by a grant (BU 642/1) from the Deutsche Forschungsgemeinschaft to H.-G.B.
FOOTNOTES
- Received 25 July 2001.
- Accepted 8 October 2001.
- Copyright © 2002 American Society for Microbiology
