US2, a Human Cytomegalovirus-encoded Type I Membrane Protein, Contains a Non-cleavable Amino-terminal Signal Peptide* 210

The human cytomegalovirus US2 gene product targets major histocompatibility class I molecules for degradation in a proteasome-dependent fashion. Degradation requires interaction between the endoplasmic reticulum (ER) lumenal domains of US2 and class I. While ER insertion of US2 is essential for US2 function, US2 lacks a cleavable signal peptide. Radiosequence analysis of glycosylated US2 confirms the presence of the NH2 terminus predicted on the basis of the amino acid sequence, with no evidence for processing by signal peptidase. Despite the absence of cleavage, the US2 NH2-terminal segment constitutes its signal peptide and is sufficient to drive ER translocation of chimeric reporter proteins, again without further cleavage. The putative US2 signal peptide c-region is responsible for the absence of cleavage, despite the presence of a suitable −3,−1 amino acid motif for signal peptidase recognition. In addition, the US2 signal peptide affects the early processing events of the nascent polypeptide, altering the efficiency of ER insertion and subsequentN-linked glycosylation. To our knowledge, US2 is the first example of a membrane protein that does not contain a cleavable signal peptide, yet otherwise behaves like a type I membrane glycoprotein.

The human cytomegalovirus US2 gene product targets major histocompatibility class I molecules for degradation in a proteasome-dependent fashion. Degradation requires interaction between the endoplasmic reticulum (ER) lumenal domains of US2 and class I. While ER insertion of US2 is essential for US2 function, US2 lacks a cleavable signal peptide. Radiosequence analysis of glycosylated US2 confirms the presence of the NH 2 terminus predicted on the basis of the amino acid sequence, with no evidence for processing by signal peptidase. Despite the absence of cleavage, the US2 NH 2terminal segment constitutes its signal peptide and is sufficient to drive ER translocation of chimeric reporter proteins, again without further cleavage. The putative US2 signal peptide c-region is responsible for the absence of cleavage, despite the presence of a suitable ؊3,؊1 amino acid motif for signal peptidase recognition. In addition, the US2 signal peptide affects the early processing events of the nascent polypeptide, altering the efficiency of ER insertion and subsequent N-linked glycosylation. To our knowledge, US2 is the first example of a membrane protein that does not contain a cleavable signal peptide, yet otherwise behaves like a type I membrane glycoprotein.
Signal peptides dictate ER insertion of integral membrane proteins and proteins destined for the secretory pathway in either co-or post-translational fashion (1)(2)(3). Signal peptides are recognized by the signal recognition particle, which directs the nascent chain and ribosome to the signal recognition particle receptor embedded within ER 1 membrane (4 -9). Upon docking with the ER membrane, many proteins are co-translationally inserted into the ER lumen via a ribosome-translocon channel that includes the heterotrimeric Sec61p complex and the translocating chain-associating membrane protein (10,11). Once the translocon has been engaged, NH 2 -terminal signal peptides are cleaved from the nascent chain by signal pepti-dase, a serine endopeptidase present near the translocon in the ER membrane (12). Signal-anchor sequences also interact transiently with the ER translocation complex, but are not cleaved. Instead, signal-anchor sequences move laterally out of the translocon to become permanent membrane anchors (13,14).
Signal peptides are composed of three distinct regions, n-, h-, and c-regions. The n-region consists of polar residues, often with a net positive charge, at the NH 2 terminus (15). The h-region is the central hydrophobic 7-15 residue helical core that can insert into the ER membrane (13,16). Finally, the carboxyl-terminal c-region has more polar character and contains the signal peptidase cleavage site (15). Signal peptidase recognizes a pattern that includes amino acids with small side chains in the Ϫ1 and Ϫ3 positions relative to the cleavage site, present in the c-region in extended conformation near the head groups on the inner leaflet of the ER membrane (2,14,16,17).
The human cytomegalovirus (HCMV) encodes two ER-resident membrane glycoproteins, US2 and US11, that each destabilize major histocompatibility class I molecules (18). The 199residue US2 glycoprotein contains an ER-lumenal portion, a predicted single transmembrane domain, and a short cytoplasmic tail (Fig. 1). US2 recognizes class I molecules via an immunoglobulin-like fold that attaches to the class I ER-lumenal domain (19) and US2 subsequently targets class I heavy chains for dislocation from the ER to the cytosol, where they are rapidly degraded by the proteasome (20). The mechanism of ER dislocation and degradation is not clearly understood.
HCMV US2 and US11 exhibit unusual characteristics concerning their ER processing events. The US11 signal peptide is cleaved from the nascent chain in delayed fashion (21). Both the signal peptide n-region and the COOH-terminal membrane anchor influence processing of the US11 signal peptide (21). Here we report several highly unusual properties of US2 ER translocation. We show that US2 behaves like a type I membrane protein that contains a non-cleavable signal sequence at its NH 2 terminus.
Metabolic Labeling/Pulse-Chase Analysis and Immunoprecipitation-Cells were released from tissue culture flasks by trypsinization and incubated in cysteine/methionine-free Dulbecco's modified Eagle's medium for 45 min at 37°C. Cells were metabolically labeled with 500 Ci of [ 35 S]cysteine/methionine (1200 Ci/mmol; PerkinElmer Life Sciences, Boston, MA) at 37 o C for the indicated times. Pulse-chase experiments and immunoprecipitation were performed as previously described (21).
Endoglycosidase H Digestion and Gel Electrophoresis-Digestion with endoglycosidase H (EndoH) (New England Biolabs, Beverly, MA) was performed on immunoprecipitated complexes according to the manufacturer's instructions. Proteins were separated by SDS-PAGE and [ 35 S]methionine-labeled proteins were visualized by fluorography/autoradiography (25).
In Vitro Transcription and Translation-In vitro transcription and translation of US2  were performed as described (26). Subcellular Fractionation, Na 2 CO 3 and Urea Treatment-The subcellular fractionation and Na 2 CO 3 treatment were performed as previously described, respectively (21). Urea treatment of homogenates prepared from HEK-293 transfectants was performed in a similar manner as Na 2 CO 3 treatment, except 4.5 M urea was used.
NH 2 -terminal Sequence Analysis-Following SDS-gel electrophoresis, US2 was transferred from the polyacrylamide gel to a Sequi-Blot polyvinylidene difluoride membrane (0.2 mm) (Bio-Rad, Hercules, CA) using a semi-dry blotting apparatus (Labconco, Seattle, WA). The glycosylated US2 polypeptide was excised from the polyvinylidene difluoride membrane and subjected to automated Edman degradation. An Applied Biosystem Protein Sequencer, Model 477, using ATZ chemistry, at the Biopolymers Laboratory at MIT, Center for Cancer Research was utilized. The fractions from each degradation cycle were collected and counted by liquid scintillation spectrometry.
N-Linked glycosylation was examined to further assess the topology of US2  . It contains consensus N-linked glycan attachment sequons at residues 68 -70, 172-174, and 188 -190 ( Fig. 1); the latter two are within the predicted transmembrane anchor and cytoplasmic tail, respectively. To determine the number of glycosylation sites utilized in vivo, US2 immunoprecipitates were treated with varying amounts of EndoH. Independently of enzyme concentration, EndoH treatment produces a single novel polypeptide, whose mass is reduced by ϳ4 kDa, consistent with the presence of a single N-linked glycan on US2 (Fig. 2B). Furthermore, US2 deletion mutants that lack the two COOH-terminal glycosylation sequons are glycosylated in a similar manner in transfectants (Fig. 5). We conclude that a single N-linked glycan is attached at asparagine 68, present in a loop between the B and C ␤-strands of the US2 Ig-like fold ( Fig. 6A) (19).
US2 Is Inefficiently Inserted into the ER-US2 polypeptides that differ by the presence of the single N-linked glycan are recovered from lysates of metabolically labeled US2 1-199 cells (Fig. 2B). To determine whether the non-glycosylated species arises as a deglycosylated degradation intermediate produced in the course of a dislocation reaction (20), pulse-chase experiments were performed on US2  cells in the presence and absence of the proteasome inhibitor, ZL 3 VS (27). US2 molecules were recovered from cell lysates and analyzed by SDS-PAGE (Fig. 3A). The recovery of glycosylated and non-glycosylated US2 polypeptides decreases at the later chase times both in the presence or absence of ZL 3 VS (Fig. 3A), but as expected, the degradation rate of non-glycosylated US2  is slower in the presence of proteasome inhibition (Fig. 3, compare lanes 1-4 and 5-8). We observed no evidence for a precursor-product relationship for glycosylated and non-glycosylated US2 polypeptides. We infer that the non-glycosylated US2 polypeptide does not arise as a degradation intermediate.
Non-glycosylated US2 could result either from incomplete ER insertion or inefficient glycosylation. To distinguish between these possibilities, the localization of non-glycosylated US2 was determined by subcellular fractionation. Metabolically labeled US2 1-199 cells were homogenized by mechanical disruption, followed by differential centrifugation (Fig. 3B). The 100,000 ϫ g supernatant (S) contains cytosolic proteins, while the 100,000 ϫ g pellet (P) contains the microsomal fraction. Glycosylated US2 is recovered exclusively from the 100,000 ϫ g pellet (P) (Fig. 3B, lane 4), indicative of its membrane disposition. However, the bulk of the non-glycosylated US2 polypeptide is recovered from the cytosolic fraction, with only a minor fraction co-sedimenting with the microsomal fraction (Fig. 3B, lanes 3 and 4). The cytosolic form of US2 is probably material that did not insert into the ER. The interaction of US2 with a membrane protein, possibly through its signal peptide, may account for the association of non-glycosylated US2 with the microsomal fraction. A subpopulation of non-glycosylated US2 continues to co-sediment with the microsomal fraction upon Na 2 CO 3 extraction. 2 These results fail to resolve completely the source of non-glycosylated US2. However, in the absence of a precursor-product relationship for glycosylated and non-glycosylated US2 in the presence and absence of proteasome inhibition, we favor the notion that membrane insertion, and hence N-linked glycosylation, did not occur for non-glycosylated US2.
US2 Lacks a Cleavable Signal Sequence-The SignalP com-puter algorithm (www.cbs.dtu.dk/services/SignalP/index.html) predicts a US2 NH 2 -terminal signal peptide with probable cleavage site between residues 20 and 21. However, most signal peptides exhibit greater hydrophobicity than that present at the US2 NH 2 terminus ( Fig. 1) (14). We therefore examined whether the US2 NH 2 terminus corresponds to a bona fide signal peptide. An NH 2 -terminal truncation mutant that lacks the NH 2 -terminal 20 residues (US2 20 -199 ) was stably transfected into U373 cells and examined for ER insertion by the acquisition of the N-linked glycan. US2 molecules were recovered from SDS-treated lysates of metabolically labeled US2  and US2 20 -199 cells using anti-US2 serum. Half of the US2 immunoprecipitates were treated with EndoH and analyzed by SDS-PAGE (Fig. 4A). In contrast to US2  , the US2 20 -199 polypeptides recovered from cell lysates were not glycosylated, as judged from their resistance to EndoH, and were of a size consistent with that of the predicted non-glycosylated polypeptide (Fig. 4A, lanes 3 and 4). Subcellular fractionation likewise indicated that US2 20 -199 molecules localize to the cytosol (Fig. 4B). Nearly all of the US2 20 -199 molecules were recovered from the 100,000 ϫ g supernatant (S) fraction (Fig. 4B, lane 3). In contrast, a major population of the US2 chimera that contains the signal peptide of murine H-2K b as the replacement for residues 1-20 of US2 (K b 1-21 /US2  ) was inserted into the ER (Fig. 4A, lanes 5 and 6). Not only does the 21-residue signal sequence of murine H-2K b class I heavy chain drive the ER insertion of US2 20 -199    cells were treated with 100 mM Na 2 CO 3 , followed by centrifugation at 150,000 ϫ g (150Kg). US2   (lanes 1-3) and ␤ 2 m (lanes 4 -6) were recovered directly from cell lysates (Ϫ) ( lanes  1 and 4), from the 150,000 ϫ g supernatant (S) (lanes 2 and 5) and from the 150,000 ϫ g pellet (P) (lanes 3 and 6) using anti-US2 and anti-␤ 2 m sera and analyzed by SDS-PAGE (15%). B, US2 1-199 immunoprecipitates were treated with increasing concentrations of EndoH to reveal the number of N-linked glycans attached to US2. Glycosylated (US2(ϩ)CHO) and non-glycosylated (US2(Ϫ)CHO) US2 polypeptides are indicated. CHO, carbohydrate. lanes 5 and 6)), the glycosylated K b 1-21 /US2 21-199 species (Fig.  4A, lane 5, upper band) also has a faster electrophoretic mobility than glycosylated US2   (Fig. 4A, compare lanes 1 and 5). This difference in mobility between US2 1-199 and K b 1-21 / US2  is also observed when the N-linked glycans are removed by EndoH treatment (Fig. 4A, compare lanes 2 and 6). This result was altogether unexpected and suggests that the US2 1-199 molecule fully retains its NH 2 -terminal signal peptide. Consistent with this possibility, US2 translated in vitro in the absence of microsomal membranes (Fig. 4C, lane 1) comigrates with EndoH-treated US2 molecules translated in the presence of microsomes (Fig. 4C, lane 3).
To confirm the absence of signal peptide cleavage for US2, NH 2 -terminal sequencing was carried out on glycosylated US2 molecules recovered from metabolically labeled cellular transfectants (Fig. 4D). The major peak of radioactivity is observed at position one, with a smaller amount of radioactivity released at position 15, consistent with methionines present in the US2 sequence at residues 1 and 15 (Fig. 1). We conclude that US2 fully retains its NH 2 terminus following ER insertion.
The US2 Transmembrane Domain Is Not Required for ER Insertion-A type I membrane protein that lacks an NH 2terminal signal sequence is without precedent. Internal hydrophobic domains can serve as internal signal peptides as well as stop transfer sequences. Such hydrophobic domains are observed in most proteins that lack a cleavable signal sequence. Is it possible that the putative transmembrane domain (Fig. 1, aa 161-185) of US2  can act as a signal peptide? We constructed a US2 COOH-terminal deletion mutant (US2 1-160 ) and examined its ability to insert into ER. We used glycosylation of US2 1-160 as a marker for ER insertion. U373 cells that stably express US2 1-160 still show efficient glycosylation of US2 1-160 (Fig. 5). The elimination of the putative transmembrane segment does not interfere with ER insertion of US2 and therefore the US2 contains an NH 2 -terminal signal peptide. US2 1-160 Is a Soluble Molecule-A Kyte-Doolittle hydropathy plot of US2  demonstrates hydrophobic segments at the 3), and the 100,000 ϫ g (100Kg) pellet (P) (lane 4) of fractionated US2 20 -199 cells with anti-US2 sera, followed by SDS-PAGE (12.5%). C, US2  was immunoprecipitated from an in vitro translation of US2  mRNA in the presence (lane 3 and 4) and absence (lane 1) of dog pancreatic microsomes or metabolically labeled US2

cells (lane 4)
using anti-US2 sera. The US2  immunoprecipitate was digested with EndoH as indicated. D, US2 recovered from [ 35 S]methionine-labeled cells was subjected to 20 cycles of Edman degradation. The radioactivity recovered from each fraction of Edman degradation was plotted against the cycle number. The NH 2 -terminal residues of US2 are represented using the single letter code. CHO, carbohydrate.  4. US2 lacks a cleavable signal sequence. A, US2 was recovered from stable U373 astrocytoma transfectants expressing full-length US2 (US2 1-199 ) (lanes 1 and 2), a NH 2 -terminal deletion mutant that lacks the first 19 residues (US2 20 -199 ) (lanes 3 and 4), or a US2 chimera that contains the H2-K b cleavable signal sequence in place of the NH 2 -terminal 20 US2 residues (K b 1-21 /US2  ) (lanes 5 and 6). Lysates from metabolically labeled transfectants were immunoprecipitated with anti-US2 sera and analyzed by SDS-PAGE (12.5%). Half of the immunoprecipitates were digested with EndoH, as indicated. Glycosylated (US2(ϩ)CHO) and non-glycosylated (US2(Ϫ)CHO) US2 polypeptides are indicated. B, US2 polypeptides were immunoprecipitated from non-fractionated US2 20 -199  NH 2 terminus (residues 1-20) and between residues 110 and 130, in addition to the putative transmembrane domain (Fig.  1). Can these hydrophobic domains act as additional membrane anchors? The crystal structure of the US2/class I complex shows that residues 110 -130 comprise the F and G ␤-strands of the US2 Ig-like fold, indeed forming an important part of the class I binding surface (Fig. 6A) (19). Therefore, this hydrophobic segment resides within the ER lumen. Consistent with this data, an additional N-linked glycosylation sequon introduced at position 149 is efficiently utilized (supplemental Fig. 1). Since this US2 mutant acquires N-linked glycans at positions 68 and 149, the intervening segment must also be present in the ER. Together, these results demonstrate that US2 residues 110 -130 do not traverse the membrane bilayer and are consistent with a type I topology for US2.
Sublocalization of US2 Signal Peptide Properties-The Sig-nalP algorithm was used to predict the n-, h-, and c-regions of both the putative US2 and H-2K b signal peptides. To determine which region of US2's signal peptide is responsible for the observed lack of signal peptide cleavage, we replaced the predicted US2 n, n ϩ h, or n ϩ h ϩ c regions with the corresponding regions of the H-2K b signal peptide (Fig. 7A). US2 chimeras that contain the K b n-region in place of the putative US2 n-region (K b 1-5 /US2  ) are poorly inserted in the ER within these transfectants (Fig. 7B, lanes 3 and 4). Chimeras that contain the putative n ϩ h-region of H-2K b (K b 1-16 /US2  ) are efficiently glycosylated (Fig. 7B, lanes 5 and 6). Despite efficient ER translocation and glycosylation, the hybrid K b 1-16 / US2  signal peptide apparently remains uncleaved following ER insertion (Fig. 7B, lanes 5 and 6). In contrast, chimeras that contain the entire H-2K b signal sequence in place of the US2 20 NH 2 -terminal residues (K b 1-21 /US2  ) are efficiently inserted into the ER with concomitant signal peptide cleavage (Fig. 7B, lanes 7 and 8). For the K b 1-21 /US2 21-199 chimera, its signal peptide appears to be removed, since this product migrates faster than US2  and other H-2K b /US2 chimeras upon SDS-PAGE separation. Thus, elements within the putative US2 signal peptide c-region are responsible for the lack of signal peptide cleavage.
The US2 NH 2 Terminus Serves as a Signal Peptide-To determine whether the US2 NH 2 terminus can direct an exogenous membrane protein to the ER, we constructed chimeras comprised of US2 NH 2 -terminal peptides fused to a reporter type I membrane protein, the major histocompatibility class I (HLA-A2) heavy chain (HC), lacking its own signal peptide (Fig. 8). Since the length of the US2 signal peptide is not precisely known, we generated a series of chimeric proteins that contain either the NH 2 -terminal 20, 25, 30, 35, or 40 residues of US2 fused to the NH 2 -terminal end of the signal sequence-less class I heavy chain. US2/HC chimeras were transiently expressed in HEK-293 cells and ER translocation was assayed by the acquisition of the single N-linked glycan at position 86 of the class I heavy chain. Transfectants were FIG. 6. US2 1-160 is a soluble protein. A, the hydrophobic stretch between residues 110 -130 of US2 is located in the ER lumen; the US2⅐HLA-A2⅐Tax complex structure (19): class I heavy chain (yellow), ␤ 2 -microglobin (gray), US2 (blue, residues 43-109 and 131-137; red, residues 110 -130), and Tax peptide (green). B and C, homogenates from metabolically labeled HEK-293 cells transfected with US2 1-199 , US2 1-160 , and ␤ 2 m were treated with 100 mM Na 2 CO 3 (B) or 4.5 M (C) urea followed by high speed centrifugation (150,000 ϫ g) (150Kg). US2   (lanes 1-3), US2 1-160 (lanes 4 -6), and ␤ 2 m (lanes 7-9) were recovered directly from cell lysates (Ϫ) (lanes 1, 4, and 7), from the 100 mM Na 2 CO 3 or 4.5 M 150,000 ϫ g supernatant (S) (lanes 2, 5, and 8) and from the 100 mM Na 2 CO 3 or 4.5 M urea 150,000 ϫ K pellet (P) (lanes 3,  6, and 9), respectively, using the respective sera and analyzed by SDS-PAGE (15%). CHO, carbohydrate. metabolically labeled with [ 35 S]methionine and the US2/HC chimeric molecules were recovered from SDS-treated cell lysates with anti-heavy chain serum. Half of the immunoprecipitates were treated with EndoH and analyzed by SDS-PAGE (Fig. 8). A heavy chain construct lacking its signal peptide (HC 25-365 ) does not acquire an N-linked glycan (Fig. 8, lanes 3  and 4), when compared with wild type heavy chains HC 1-365 (Fig. 8, lanes 1 and 2) and endogenously expressed HEK-293 heavy chains (lanes 3-14). An US2/HC chimera that contains only the 20 NH 2 -terminal residues of US2 (US2 1-20 /HC  ) was resistant to EndoH treatment, similar to heavy chain molecules that lack a signal sequence (Fig. 8, lanes 3-6). In contrast, about half of the chimeric molecules that contain the NH 2 -terminal 25 residues of US2 (US2 1-25 /HC  ) acquire an N-linked glycan (Fig. 8, lanes 7 and 8). The NH 2 -terminal 30 US2 residues (US2 1-30 /HC 25-365 ) were required for efficient glycosylation of chimeric proteins, yielding a majority of glycosylated molecules (Fig. 8, lanes 9 and 10). Interestingly, although the US2 NH 2 terminus directs heavy chain translocation into the ER, the US2 signal peptide still fails to be cleaved from the US2/HC fusion proteins (Fig. 8). US2/HC chimeric proteins migrate more slowly on SDS-PAGE than either fulllength or signal sequence-less heavy chain molecules (Fig. 8,  lanes 1-4). These results demonstrate that US2 contains a bona fide NH 2 -terminal signal peptide, and that it remains uncleaved even when transposed onto a different membrane glycoprotein. DISCUSSION US2 behaves like a type I membrane protein whose NH 2terminal signal peptide is not removed upon translocation into the ER. The NH 2 -terminal 30 US2 residues function as a noncleavable signal peptide, even when appended onto an unrelated type I membrane protein. Despite the hydrophobic nature of the non-cleavable US2 signal peptide, it does not appear to act as membrane anchor. A carboxyl-terminal truncation mutant that lacks its transmembrane domain, but retains its signal peptide (US2 1-160 ) remains in a soluble form upon urea extraction of microsomal membranes (Fig. 6C). Furthermore, the crystal structure of the US2-class I complex (19) makes it difficult to envision a US2-class I complex in which both the NH 2 -and COOH-terminal ends of US2 are anchored to the membrane bilayer (Fig. 6A). In addition, replacement of the US2 signal peptide with the cleavable murine class I H-2K b signal peptide (K b 1-21 /US2 21-199 ) (Fig. 4) does not affect the function of US2 (Supplemental Material Fig. 2). If the US2 signal peptide were tethered to the membrane, presumably the overall structures of US2 1-199 and K b 1-21 /US2  would be different and would likely preclude US2-induced destruction of class I molecules for the latter chimera.
There are several examples of secreted proteins whose signal peptides are not cleaved: fibroblast growth factor (FGF)-9, ovalbumin, plasminogen activator inhibitor-2 (plasminogen activator inhibitor-2), and the carp retinol-binding protein (RBP). Reminiscent of US2, (FGF)-9 has a relatively weak hydrophobic NH 2 terminus and the NH 2 -terminal 28 residues can direct ER translocation of a heterologous fusion protein (30). While the US2 NH 2 -terminal deletion mutant (US2 20 -199 ) fails to translocate into the ER, FGF-9 deletion mutants that lack 22 NH 2 -terminal residues can still enter the ER (30). Furthermore, an internal sequence of considerable hydrophobicity is required for FGF-9 translocation (31). The FGF-9 internal hydrophobic region can function as a cleavable signal sequence when placed artificially at the NH 2 terminus, and FGF-9 mutants that lack the NH 2 -terminal 90 residues are capable of ER insertion when translated in vitro in the presence of microsomes (31). In contrast, disruption of this central hydrophobic region prevents FGF-9 translocation. Taken together, FGF-9 appears to contain two separate regions that cooperate to allow ER insertion.
Similar to FGF-9, the SERPINs plasminogen activator inhibitor-2 and ovalbumin contain internal hydrophobic regions to direct ER insertion (32)(33)(34). The mildly hydrophobic NH 2 termini of these serpins are not cleaved upon ER insertion. Plasminogen activator inhibitor-2 relies upon a bipartite signal sequence, composed of two internal hydrophobic domains near its NH 2 terminus (32, 35). While signal recognition particle recognizes the second internal hydrophobic region (32), both hydrophobic regions of plasminogen activator inhibitor-2 are required for ER insertion. The first weakly hydrophobic region may allow recognition of the central hydrophobic region by the translocation machinery (32). In the case of RBP, the 17 NH 2terminal residues can direct ER insertion of a heterologous fusion protein, yet it is not known whether additional regions of RBP participate in ER insertion (36).
Elements within the US2 signal peptide itself are responsible for the lack of signal sequence cleavage. Chimeric molecules that possess portions of the murine H-2K b class I heavy chain signal peptide in place of the corresponding US2 signal peptide regions highlight several properties of US2's signal peptide (Fig. 3A). First, the chimeric proteins that possess the H-2K b n-region instead of that of US2 (K b 1-5 /US2  ) remain mostly cytosolic. These results could be explained by incorrect assignment of the US2 n-region by the SignalP algorithm. This possibility appears improbable. The role of the n-region in ER insertion is not fully understood, but the h-region has been implicated in directing the nascent chain to the ER membrane (1). Is it possible that the n-and h-regions work together to efficiently insert the nascent chain into the ER? In the case of the chimeric molecule, K b 1-5 /US2  , the hydrophobic nature of the US2 h-region was not sufficient to drive ER translocation of the chimera in the presence of the H-2K b n-region. Therefore, K b 1-5 /US2  would poorly insert into the ER. This contrasts with the results obtained for similar US11 hybrid mutants (21). The US11 glycoprotein exhibits delayed cleavage of its signal peptide, a property determined by its signal peptide n-region and at considerable distance, its transmembrane domain. Substitution of the US11 n-region with that of H-2K b restored normal signal sequence properties (21). These results suggest that the function of the n-region in ER insertion and signal peptide cleavage have yet to be fully explored.
The molecular basis for the proposed lack of US2 signal peptide cleavage by its c-region remains unknown. The c-region is usually a polar region that contains prolines and glycines with small uncharged residues at the Ϫ1 and Ϫ3 position of the signal peptide cleavage site (2). Residues following the cleavage site can also similarly modulate signal peptidase proteolysis (37)(38)(39). The absence of the "helix-breaking" prolines and glycines residues can alter the kinetics of signal peptide cleavage (40,41). This is not the case for the US2 signal peptide: there is a glycine and proline in its predicted c-region (Fig. 1). Signal peptides that lack a suitable Ϫ3, Ϫ1 pair are refractory to cleavage by signal peptidase (42). The SignalP algorithm predicts the cleavage site of the US2 signal peptide at position 21. This would put an isoleucine at position Ϫ3 and a leucine at position Ϫ1, two residues that are not consistent with the cleavage site consensus sequence. However, a classical Ϫ3,Ϫ1 cleavage site maps to residues Thr 26 and Ala 28 of the apparent US2 c-region. Signal peptide cleavage efficiency decreases as c-regions extend beyond 9 residues, and signal peptides with c-regions of 13 residues fail to be cleaved (43). The putative Ϫ3,Ϫ1 US2 pair would allow for a cleavage site 11 residues into the c-region. Thus, the US2 signal peptide should allow signal peptide cleavage, albeit with reduced efficiency. It remains possible that the US2 c-region cleavage site is not accessible to the active site of signal peptidase. For instance, the US2 COOH terminus may associate with a factor within the ER membrane such as a translocon component, or perhaps the c-region possesses a secondary structure that impairs signal peptidase activity.
The US2 NH 2 terminus dictates the efficacy of US2 insertion and its glycosylation. Subcellular fractionation shows that nonglycosylated US2 is present both in cytosolic and microsomal fractions derived from US2 transfectants (Fig. 3). Non-glycosylated US2 is present in the cytosol prior to the chase, which suggests inefficient ER insertion for at least a fraction of US2. Alternatively, this subpopulation could represent US2 mole-cules that have docked with the ER membrane, but failed to translocate. Since US2 chimeras that contain the H-2K b signal peptide are translocated and glycosylated to a greater extent (Figs. 4 and 7), the US2 signal peptide appears to influence ER insertion and N-linked glycosylation.
Both US2 and US11, the HCMV proteins that target class I heavy chains for dislocation, contain signal sequences with unusual properties. The closely related US3 glycoprotein retains class I molecules within the ER and contains a signal peptide with the usual cleavage properties. 2 It is intriguing to propose a role for the US2 and US11 signal peptides in ER dislocation of major histocompatibility class I. Do US2 and US11 signal peptides act to keep US2/US11 in close proximity of the translocon? Alternatively, do the US2/US11 signal peptides affect the dislocation machinery to increase the efficiency of class I dislocation? We have sequenced the US2 gene from a number of clinical HCMV isolates, and find that the NH 2 terminus of US2 is highly conserved, identical to the sequence shown in Fig. 1. 3 Thus, the behavior of US2, as reported here, is representative of US2 as found in HCMV in the human population. In contrast, significant NH 2 -terminal sequence variation has been reported for the HCMV UL144 gene product among clinical isolates (44).
The role of the NH 2 -terminal signal peptide during processing events is generally believed to be restricted to directing the nascent polypeptide chains to the ER membrane. However, signal peptides have also been implicated in stabilizing the ribosome-translocon junction (45), and may affect the functional properties of the translocon, such as dislocation. The signal peptides of neither US2 nor US11 are required for dislocation of class I molecules. Chimeric mutants of US2 and US11, whose signal peptides have been replaced with that of H-2K b , still target class I molecules for degradation (21). 2 Importantly, these experiments were carried out in astrocytoma transfectants that constitutively express high levels of both class I and US2/US11. The US2 and US11 signal peptides may play a role in class I dislocation in vivo, in the setting of a virus-infected cell. It is not known, for example, if and to what extent the structural and functional properties of the ER are modified by infection with HCMV. In view of the lack of sequence conservation between US2 and US11, the fact that both proteins are involved in dislocation and both show anomalies in the processing of their signal sequences is remarkable. FIG. 8. The US2 NH 2 terminus contains a functional signal peptide. HEK-293 cells were transiently transfected with constructs that express the full-length HLA-A2 heavy chain (HC 1-365 , lanes 1 and 2), an HLA-A2 heavy chain lacking a signal peptide (HC  , lanes 3 and 4), and US2/HLA-A2 chimeras: US2 1-20 / HC   (lanes 5 and 6); US2 1-25 /HC   (lanes 7 and 8); US2 1-30 /HC 25-365 (lanes 9 and 10); US2 1-35 /HC   (lanes 11 and  12); US2 1-40 /HC   (lanes 13 and 14)). Lysates from metabolically labeled cells were immunoprecipitated with antiheavy chain sera and analyzed by SDS-PAGE (12.5%). Half of the immunoprecipitates were treated with EndoH. The endogenously expressed class I heavy chains in HEK-293 cells migrate at a similar mobility as HC   (lanes 3-14).