Physical and Functional Interactions of the Cytomegalovirus US6 Glycoprotein with the Transporter Associated with Antigen Processing*

The endoplasmic reticulum-resident human cytomegalovirus glycoprotein US6 (gpUS6) inhibits peptide translocation by the transporter associated with antigen processing (TAP) to prevent loading of major histocompatibility complex class I molecules and antigen presentation to CD8+ T cells. TAP is formed by two subunits, TAP1 and TAP2, each containing one multispanning transmembrane domain (TMD) and a cytosolic nucleotide binding domain. Here we reported that the blockade of TAP by gpUS6 is species-restricted, i.e. gpUS6 inhibits human TAP but not rat TAP. Co-expression of human and rat subunits of TAP demonstrates independent binding of gpUS6 to human TAP1 and TAP2, whereas gpUS6 does not bind to rat TAP subunits. gpUS6 associates with preformed TAP1/2 heterodimers but not with unassembled TAP subunits. To locate domains of TAP required for gpUS6 binding and function, we took advantage of reciprocal human/rat intrachain TAP chimeras. Each TAP subunit forms two contact sites within its TMD interacting with gpUS6. The dominant gpUS6-binding site on TAP2 maps to an N-terminal loop, whereas inhibition of peptide transport is mediated by a C-terminal loop of the TMD. For TAP1, two gpUS6 binding domains are formed by loops of the C-terminal TMD. The domain required for TAP inactivation is built by a distal loop of the C-terminal TMD, indicating a topology of TAP1 comprising 10 endoplasmic reticulum transmembrane segments. By forming multimeric complexes, gpUS6 reaches the distant target domains to arrest peptide transport. The data revealed a nonanalogous multipolar bridging of the TAP TMDs by gpUS6.

The endoplasmic reticulum-resident human cytomegalovirus glycoprotein US6 (gpUS6) inhibits peptide translocation by the transporter associated with antigen processing (TAP) to prevent loading of major histocompatibility complex class I molecules and antigen presentation to CD8؉ T cells. TAP is formed by two subunits, TAP1 and TAP2, each containing one multispanning transmembrane domain (TMD) and a cytosolic nucleotide binding domain. Here we reported that the blockade of TAP by gpUS6 is species-restricted, i.e. gpUS6 inhibits human TAP but not rat TAP. Co-expression of human and rat subunits of TAP demonstrates independent binding of gpUS6 to human TAP1 and TAP2, whereas gpUS6 does not bind to rat TAP subunits. gpUS6 associates with preformed TAP1/2 heterodimers but not with unassembled TAP subunits. To locate domains of TAP required for gpUS6 binding and function, we took advantage of reciprocal human/rat intrachain TAP chimeras. Each TAP subunit forms two contact sites within its TMD interacting with gpUS6. The dominant gpUS6-binding site on TAP2 maps to an N-terminal loop, whereas inhibition of peptide transport is mediated by a C-terminal loop of the TMD. For TAP1, two gpUS6 binding domains are formed by loops of the C-terminal TMD. The domain required for TAP inactivation is built by a distal loop of the C-terminal TMD, indicating a topology of TAP1 comprising 10 endoplasmic reticulum transmembrane segments. By forming multimeric complexes, gpUS6 reaches the distant target domains to arrest peptide transport. The data revealed a nonanalogous multipolar bridging of the TAP TMDs by gpUS6.
The transporter associated with antigen processing (TAP) 3 represents a bottleneck in the MHC class I antigen-presenting pathway. TAP translocates 8 -16-mer peptides generated in the cytosol across the ER membrane into the ER lumen for loading onto MHC class I molecules (1,2). Upon anchoring of the peptide, the MHC class I heavy chain and ␤ 2 -microglobulin form a stable ternary complex that is released from the peptide loading complex consisting of TAP, tapasin, calreticulin, and the thioreductase ERp57 (3,4). MHC class I molecules carry the peptide to the cell surface for monitoring by CD8ϩ cytotoxic T lymphocytes. TAP-deficient cells show drastically reduced levels of MHC class I surface expression indicating that TAP represents the principal source of peptide ligands.
TAP is a member of the superfamily of ATP-binding cassette (ABC) transporters, which utilize ATP energy for translocation of their substrates across membrane barriers. TAP is built by two subunits, TAP1 and TAP2, each consisting of one transmembrane domain (TMD) and one nucleotide binding domain (NBD). Peptide binding to TAP is dependent on the formation of heterodimers (5). It was shown that the TMDs, and not the NBDs, of TAP1 and TAP2 encompass the peptide binding domain (6), which has been mapped to the C-terminal cytosolic loops on both subunits (7). The NBDs harbor conserved sequence motifs required for ATP binding and hydrolysis (8). According to a recently proposed model for the peptide translocation cycle (9), binding of ATP to TAP2 is needed for peptide binding to the transporter. This step is followed by ATP hydrolysis at TAP2, which allows ATP binding to TAP1 and conformational changes of TAP. The opening of the TAP pore is probably a consequence of ADP release at TAP2 and ATP hydrolysis at TAP1, which lowers the affinity for the peptide ligand and permits membrane translocation. Finally, according to this model, release of ADP from TAP1 resets the transport cycle.
The lack of crystallographic data, hydrophobicity analysis, and sequence alignments with other ABC transporters have given the basis for a hypothetical model of the ER membrane topology predicting a core TMD with six transmembrane segments (TMSs) and an additional four and three N-terminal TMSs for TAP1 and TAP2, respectively (2). Truncated TAP subunits containing only the putative core TMD were found recently to be functional in terms of peptide translocation (10). In another study, however, analysis of glycosylation sites added to C-terminally truncated nonfunctional TAP mutants suggested the formation of only 8 TMSs for TAP1 and 7 TMSs for TAP2, whereas no evidence for the most C-terminal pair of TMSs was obtained (11).
Human cytomegalovirus (HCMV) is equipped with various gene functions interfering with the MHC class I pathway of antigen presentation (12). One of them is the US6-encoded 21-kDa glycoprotein (gpUS6) that inhibits the translocation of peptides across the ER membrane (13)(14)(15)(16). The ER resident gpUS6 is a type I transmembrane protein, with a short cytoplasmic tail, a transmembrane segment (TMS), and a bulky luminal part. It has been shown that a truncated gpUS6 mutant, lacking the TMS, is still able to bind and inactivate TAP (16 -18). In contrast to gpUS6, the herpes simplex virus (HSV)-encoded protein ICP47 binds to TAP at the cytosolic side. ICP47 is a small soluble protein described to act as a high affinity competitive inhibitor of the peptide-binding site on TAP (19 -21).
In this study we took advantage of the species-restricted specificity of gpUS6 for TAP; gpUS6 interacts with human TAP but not rat TAP. By constructing intrachain human/rat TAP chimeras, we define minimal gpUS6 binding domains on TAP and assess their relative contribution to the inactivation of the transporter. The data provide a model in which gpUS6 binds to distant, conformation-dependent ER-luminal loops of TAP1 and TAP2. gpUS6 forms oligomers revealing that gpUS6 forms a bridge between TAP1/2 subunits. Moreover, our study provides evidence for a TAP1 TMD structure consisting of 10 TMSs.
Human TAP1A (35) was subcloned into the XbaI site of pBluescript II KSϩ. Rat TAP1a was present as clone 510-10 in pBluescript II KSϩ (36). The conserved restriction site KpnI in human TAP1A cDNA (1137 in X57522) 4 and rat TAP1a (905 in X57523) was used to generate the chimeric TAP1 r6hT1 and h6rT1 cDNAs. The restriction site BsiWI present in rat TAP1a (1106 in X57523) was introduced into human TAP1A by PCR mutagenesis (1338 in X57522). The BsiWI site was used to generate chimeric TAP1 r8hT1 and h8rT1 cDNAs. The conserved restriction site SanDI in hTAP1a (1822 in X57522) and rTAP1a (1590 in X57523) was used for the construction of the chimeric TAP1 cDNA rxhT1. Human TAP2E (37) was subcloned between HincII (5Ј) and NotI (3Ј) of pBluescript II KSϩ. Rat TAP2u (38) was ligated into the EcoRI site of pBluescript II KSϩ. The conserved sites BsaBI in hTAP2E (608 in Z22936) and rTAP2 u (631 in X75307) were used to generate chimeric TAP2 cDNAs h4rT2 and r4hT2. The restriction site SalI present in rTAP2u (1184 in X75307) was introduced by PCR mutagenesis into hTAP2E (1161 in Z22936) (39) and used to generate the chimeric rTAP2 cDNAs h7rT2 and r7hT2. All TAP1 and TAP2 constructs were cleaved from pBluescript II KSϩ and subcloned into p7.5k131a using appropriate restriction sites. For the VR, amino acid substitution in the r8hT1 chimera mutations was inserted using the primers 5Ј-GGAG-GCTGTGGCCTATGCAG-3Ј and 5Ј-CTGCTTACAGCCCCTCT-GACCACCAGCTGCCCACC-3Ј (mutations underlined), and as a template r8hT1-p7.5k131A was used. The amplified product was used in a second PCR as a primer together with 5Ј-GCTGTGATTTCCTC-CATAGTTGGC-3Ј and again with r8hT1-p7.5k131A as template. The second PCR product was digested by the restriction enzymes SpeI and DraIII and ligated to r8hT1-p7.5k131A digested by the same enzymes.
Generation of Recombinant Vaccinia Viruses-Construction of rVV was performed as described (40). Genes of interest inserted in p7.5k131A were transferred through homologous recombination into vaccinia virus genome (strain Copenhagen) by interrupting the thymidine kinase gene. rVV were selected by bromodeoxyuridine (100 g/ml) using HU143tk Ϫ cells.
Immunoprecipitation and Western Blot-For binding analysis of gpUS6 to TAP, 10 6 CMT64. Lysates were cleared from membrane debris at 13,000 rpm for 30 min at 4°C. A 50-l aliquot was stored for control of protein expression. Lysates were incubated with antibodies at 4°C over night before immunocomplexes were retrieved by adding protein A-Sepharose (Amersham Biosciences). Sepharose pellets were washed three times, and immune complexes were dissociated by addition of sample buffer (80 mM Tris (pH 6.8), 5 mM EDTA, 34% sucrose, 3.2% SDS, and 40 mM DTT) and incubation at 95°C for 5 min. Under nonreducing conditions DTT was omitted. Proteins were separated by SDS-PAGE, blotted onto a Protran membrane, and detected by specific antibodies followed by peroxidase-conjugated secondary antibodies and detection using the enhanced chemiluminescence substrate ECL Plus (Amersham Biosciences). For immunoprecipitation of metabolically labeled proteins, cells were labeled for 1 h with 35 S-labeled Redivue TM Pro-Mix TM (Amersham Biosciences) prior to lysis, performed as described above. 35 S incorporation into proteins was quantitated in each experiment by liquid scintillation counting. Lysates were adjusted to ensure comparability in quantitative terms. Incubation with antibodies was reduced to 1 h. Immune complexes were collected with protein A-Sepharose and dissociated with sample buffer. Proteins were separated on a 10 -13% gradient SDS-PAGE. Dried gels were exposed to BioMax MR films at Ϫ70°C for 1-7 days.

Inactivation of TAP by gpUS6 Is Species-restricted-Herpesviruses
are pathogens with a long history of co-speciation with their host leading to a close adaptation of molecular functions. As a consequence, some of the herpesviral inhibitors of the MHC class I pathway, including the HSV-encoded inhibitor of TAP, ICP47 (21), exhibit a remarkable specificity regarding their target recognition (42). To test whether the HCMV-encoded TAP inhibitor gpUS6 is able to discriminate between TAP homologs of different species, cell lines derived from several mammalian species (human, African green monkey, rabbit, rat, and mouse) were infected either by an rVV expressing US6 (rVV-US6) or with WT VV. To assess translocation efficiency, peptides were selected that are know as good substrates for human/monkey TAP as well as rodent TAP (39,43,44). Although the peptide translocation efficiency generally varied between cell lines, VV infection did not affect TAP function. In human HeLa cells and monkey CV1 cells gpUS6 inhibited transport of peptide by more than 85% in comparison to WT VV-infected cells (Fig.  1). In rabbit Rab-9 cells peptide translocation was also sensitive to gpUS6. In contrast, in mouse and rat cells peptide translocation was not impaired by gpUS6. The same pattern of TAP sensitivity to gpUS6 was observed when a further peptide was used in the translocation assay (data not shown). The data suggest that gpUS6 inhibits TAP-mediated peptide translocation in a species-dependent manner and that rat and mouse TAP resist functional inactivation by gpUS6.

gpUS6 Associates Only with Preformed TAP Heterodimers and Both TAP1 and TAP2 Provide Independent Binding Domains-Previous
work has demonstrated that the luminal domain of gpUS6 physically interacts with the TAP-tapasin-MHC class I peptide loading complex (16 -18), but the molecular basis for this interaction is still unknown. To analyze gpUS6 binding to TAP subunits, we expressed human and rat TAP1 and TAP2 subunits (hT1 and hT2; rT1 and rT2) with rVV. TAPdeficient mouse CMT64.5 cells were infected with rVV expressing gpUS6 and different combinations of human and rat TAP subunits. Binding of gpUS6 to TAP was assessed by co-immunoprecipitation using anti-gpUS6 antibodies. Single expressed human TAP subunits did not bind to gpUS6 (Fig. 2, lanes 2 and 3). Only co-expression of human TAP1 with human TAP2 established co-precipitation of both TAP subunits (Fig. 2, lane 4). In contrast, co-expressed rat TAP subunits were not precipitated by anti-gpUS6 antibodies (Fig. 2, lane 6). However, combinations of one rat subunit with a human counterpart (hT1 with rT2 or rT1 with hT2) were sufficient to restore co-precipitation of the TAP heterodimer (Fig. 2, lanes 8 and 9). We conclude that both subunits of human TAP each contain independent recognition sites for gpUS6. gpUS6 binding to a single TAP chain is sufficient for precipitating the preformed heterodimeric TAP complex, the formation of which depends on TAP subunit assembly with a complementary TAP molecule.
gpUS6 Binding to Both TAP Subunits Is Required for Efficient Inactivation of Peptide Transport-Next, we asked whether gpUS6 binding to a single TAP subunit is sufficient to inhibit peptide translocation. Peptide transport was assessed in TAP-deficient CMT64.5 cells stably transfected with US6 or an empty vector control after reconstitution of human and rat TAP subunits by rVV infection. As an index peptide, TNKTRIDGQY was chosen, which is efficiently transported by both human and rat TAP (43). This peptide was equally well transported by hT1/hT2 and hT1/rT2, and transport by rT1/rT2 was only slightly better than translocation by rT1/hT2 (supplemental Fig. 9). In this experimental setting, peptide translocation was inhibited by 60% in US6transfected cells after reconstitution with human TAP1/2 when compared with nontransfected CMT64.5 cells (Fig. 3). As expected,  One aliquot of the lysate gpUS6 was immunoprecipitated using rabbit anti-gpUS6 antiserum, and the other aliquot was immunoprecipitated with anti-TAP antibodies 1p4 and 2p3 to control TAP expression. Proteins were separated on a 10 -13% SDS-PAGE. gpUS6-associated TAP subunits are indicated on the right. Films were exposed for 42 h (left panels) and 72 h (right panels), respectively. Co-IP, co-immunoprecipitation.
peptide translocation by rat TAP was not inhibited in CMT64.5-US6 cells. Expression of combinations of human/rat transporter subunits, i.e. hT1/rT2 or rT1/hT2, allowed gpUS6 to block TAP transport to a rate of about half of what was determined for hT1/hT2 (Fig. 3). Thus, simultaneous gpUS6 binding to both TAP1 and TAP2 is a prerequisite for efficient functional inactivation of peptide translocation.
The TAP1 C-terminal TMD Contains a Binding Site for gpUS6-To locate binding domains sufficient for gpUS6 interaction with TAP subunits, a set of intrachain human/rat TAP chimeras was constructed. Because the gpUS6 ectodomain is sufficient for TAP inhibition (16 -18), the TMD of TAP was selected for sequence substitution. Conserved regions in the human and rat nucleotide sequence were used for construction of the chimeras. TAP1 chimeras constructed and analyzed are depicted in Fig. 4A and Table 1. All TAP1 hybrids were able to translocate peptides in an ATP-dependent manner when combined with rat TAP2, albeit with different efficiencies (supplemental Fig. 9). To test the sensitivity to gpUS6 inhibition, the TAP1 chimeras were combined with either hTAP2 or rTAP2, and the efficiency of peptide translocation was assessed. The chimeras r6hT1 and r8hT1 behaved like hT1, i.e. when combined with hT2 gpUS6 blocked peptide translocation by 50 -60% (Fig. 4B, white bars) and when combined with rTAP2 the gpUS6-mediated inhibition reached 25-30% (Fig. 4B, black bars). In contrast, the chimeras h6rT1 and h8rT1 showed a similar type of result as rTAP1. When combined with rTAP2, the gpUS6-mediated inhibition was not restored (Fig. 4B, black bars), whereas the combination with hTAP2 yielded an inhibition of about 20% (Fig. 4B, white bars).
To determine whether the functional interference of gpUS6 with TAP1 chimeras translates into detectable binding, the TAP1 chimeras were co-expressed with rat or human TAP2 in CMT64.5-US6 cells and detected after co-immunoprecipitation with anti-gpUS6 antibodies in Western blot. As expected, TAP1 subunits were present in gpUS6 precipitates after co-expression with human TAP2 (data shown for h6rT1 and h8rT1, Fig. 4C, lanes 5 and 7). To define domains of TAP1 sufficient for gpUS6 binding, TAP1 chimeras were combined with the rat TAP2 subunit (Fig. 4C). Expression of the r6hT1 chimera resulted in co-precipitation of rTAP2 (Fig. 4C, lane 10), which was not the case for the reciprocal h6rT1 chimera (Fig. 4C, lane 11). Both the r8hT1 and the h8rT1 chimeras bound to gpUS6 when combined with rTAP2. This result implies that human TAP1 forms two independent domains in the C-terminal TMD for gpUS6 binding, one of each present in the r8hT1 and the h8rT1 chimera. From these results we conclude that the C-ter-minal part of the TAP1 TMD, including the predicted TMSs 9 and 10 of human TAP1, is necessary to mediate binding and inhibition by gpUS6. In contrast, the second domain of TAP1 interacting with gpUS6, identified by the chimera h8rT1, does not affect peptide translocation.  gpUS6-mediated inhibition of TAP1 chimeras co-expressed with hT2 is shown as open bars and with rT2 as filled bars. % gpUS6-inhibition was calculated by the following formula: (1 Ϫ transport in US6-transfected cells/transport in nontransfected cells) ϫ 100. Standard deviations are based on two independent experiments. C, co-precipitation of gpUS6-associated TAP. TAP subunits were expressed by rVV in US6-transfected CMT64.5 cells. At 15 h post-infection digitonin lysates were prepared, and gpUS6 was immunoprecipitated from 600 l of the lysate by rabbit anti-gpUS6 antiserum. Proteins were separated by 10% SDS-PAGE before transfer to a nitrocellulose membrane. TAP subunits and gpUS6 were detected by Western blot. An aliquot of 15 l from the lysate was analyzed by Western blot for expression control. Antibodies used for protein detection in Western blots are shown on left of panels and rat and human TAP expression are indicated by r and h, respectively. Co-IP, co-immunoprecipitation.  10 TMSs-A C-terminal truncated gpUS6 mutant, lacking the TMS and the cytosolic tail, maintains the ability to inactivate TAP (16 -18), implying a strictly luminal TAP-gpUS6 interaction. We therefore hypothesized that human sequences of the r8hT1 chimera must be able to traverse the ER membrane and form the gpUS6-binding site in the ER lumen, indicating that TAP1 indeed forms a fifth pair of TMSs. To this end we constructed a further TAP1 chimera lacking all human sequences that might be a part of the TMD (rxhT1, see Fig. 4A). When gpUS6 binding to the r8hT1 and the rxhT1 chimeras was compared after co-expression with rT2, a profound difference was observed (Fig.  5A). This finding indicates that the rxhT1 mutant is lacking a site for stable gpUS6 binding, also affirmed by the peptide translocation assay, in which rxhT1 behaved rat TAP1-like in the presence of gpUS6 (Fig.  5B). To ensure an exclusively luminal interaction between gpUS6 and r8hT1, a soluble truncated mutant of gpUS6 (aa 1-139) (rVV-US6sol) was constructed. gpUS6sol resides in ER, as indicated by complete sensitivity to endo-␤-N-acetylglucosaminidase H digestion (data not shown). Again, gpUS6sol selectively recognized r8hT1 but not rxhT1, confirming that the ER lumen is the site of gpUS6-TAP1 interaction (see Fig. 5A). TAP1 recognition by gpUS6 thus requires the presence of human sequences in the C-terminal part of the TMD between aa 377 and 537, which become exposed to the ER lumen, indicating that the controversial TMS9 -10 in the core TMD do exist.
The experimental data collected here confirmed a previously proposed model of TAP membrane topology based on sequence alignment with the ABC transporter P-glycoprotein (2). Accordingly, the r8hT1 chimera exposes a short loop within the ER lumen, forming the fifth C-terminal luminal loop of TAP1 between TMS9 and TMS10. This section is shared between human and rat TAP1 except at positions 435 and 436 where threonine and serine, respectively, are found, corresponding to valine and arginine at positions 412 and 413, respectively, in the rat sequence (Fig. 6A). Next we substituted TS for VR in the r8hT1 chimera (r8hVRT1) to further assess the role of loop 5 for gpUS6 recognition. As demonstrated in Fig. 6, B and C, gpUS6 bound to both r8hT1 and r8hVRT1 and inhibited peptide transport by the chimeras with similar efficiency. We conclude that gpUS6 recognition of the hTAP1 aa 431-443 constituting the predicted fifth ER-luminal loop depends on the additional cis-acting effects of neighboring TAP1 sequences.
The gpUS6 Binding Domain on TAP2 Localizes to the N-terminal TMD-Next, we analyzed a set of reciprocal human/rat TAP2 chimeras (Fig. 7A) for gpUS6 binding after metabolic labeling by co-immunoprecipitation using anti-gpUS6 antibodies (Fig. 7B and Table 2). The heterodimeric chimeras h7rT2 and h4rT2 when combined with rT1 were recognized by gpUS6 (Fig. 7B, lanes 5 and 9), whereas r7hT2 and r4hT2 were not (Fig. 7B, lanes 3 and 7). The same binding pattern of TAP2 FIGURE 5. Luminal interaction of TAP1 with gpUS6 but not with the predicted luminal loop 5 of TAP1. A, TAP subunits were expressed and co-precipitated by anti-gpUS6 antibodies as described in Fig. 4C. TAP subunits and gpUS6 were detected by Western blot. Precipitated proteins were diluted in three 2-fold dilution steps for quantification of TAP-gpUS6 interaction before loading onto the gel. B, gpUS6-mediated inhibition of peptide translocation by rVV-expressed TAP1 chimeras was measured as in Fig. 4B. Co-IP, co-immunoprecipitation.  Fig. 4C. TAP subunits and gpUS6 were detected by Western blot. C, gpUS6-mediated inhibition of peptide translocation by rVV-expressed TAP1 chimeras was measured as in Fig. 4B. hybrids was found after expression of gpUS6sol (data not shown). This result infers that in contrast to TAP1 the TAP2-binding site for gpUS6 maps to the N-terminal TMD between aa 1 and 177.
The TAP2 TMD Contains Further Sites for gpUS6-mediated Inactivation of Peptide Transport-Like for TAP1, we evaluated the functional consequences of gpUS6 binding to TAP2 by testing the peptide transport efficiency of human/rat TAP2 hybrids. Remarkably, transport function of TAP2 chimeras h4rT2 and h7rT2, which exhibited strong gpUS6 binding, was almost not affected by gpUS6 binding (Fig. 7C). In contrast, the reciprocal constructs r4hT2 and r7hT2 exhibiting no gpUS6 binding in immunoprecipitation assays were sensitive to gpUS6 in their translocation capacity (Fig. 7C). These findings indicate different and independent modes of gpUS6 interaction with TAP2. Although gpUS6 recognition of a luminal determinant encompassing aa 1 and 177 does not suffice for a clear inactivation of peptide transport, an undetectable interaction with the C-terminal part of the TAP2 TMD accounts for the blockade of peptide translocation.
Oligomerization of gpUS6-Based on the findings above, at least four independent contact sites for gpUS6 on TAP1 and TAP2 must exist. This raises the question how the ectodomain of gpUS6 might be able to access all the distant interaction sites at the same time, which is required to achieve a complete inactivation of TAP. Noticing the fact that the sequence of gpUS6 contains multiple cysteines, we surmised that gpUS6 might form homodimers to bridge the recognition sites between the TAP heterodimers. To test this possibility, gpUS6 was expressed in CMT64.5 cells by rVV, and cell lysates were separated by SDS-PAGE in parallel under reducing or nonreducing conditions followed by Western blotting using gpUS6-specific antibodies. Under nonreducing conditions, a prominent band fitting with the molecular mass of 42 kDa for the gpUS6 homodimer was observed (Fig. 8). Moreover, multiple further bands exhibiting a molecular mass of n ϫ 21 kDa were detected under nonreducing conditions, which contracted to a single 21-kDa band under reducing conditions corresponding to gpUS6 monomers. More importantly, newly synthesized gpUS6 dimers and trimers were also observed in the presence of iodoacetamide (see supplemental Fig.  10, A and B), excluding the formation of gpUS6 multimers during in vitro incubations. This result indicates that the majority of gpUS6 molecules assembles into homodimers and multimers. Such gpUS6 oligomers are able to reach the distant binding sites on the TAP TMDs. Light gray and dark gray boxes indicate predicted TMSs (2), numbered from 1 to 9, and the controversial TMS8 and -9 are striped. B, analysis of gpUS6 binding to TAP2 chimeras. Combinations of TAP2 chimeras and hT1 or rT1 were co-expressed in US6-transfected CMT64.5 cells as indicated. Binding of gpUS6 to TAP was assessed in metabolically labeled cells by co-immunoprecipitation (Co-IP) using anti-gpUS6 antibodies as described in Fig. 2. Expression of TAP subunits was controlled by anti-TAP antibodies. TAP subunits are indicated by lines at right, beginning from the top: h4rT2, h7rT2, rT2, hT1, rT1a, r7hT2, r4hT2, and rT1b. C, gpUS6-mediated inhibition of TAP2 chimeras. TAP2 chimeras were expressed by rVV together with hT1 or rT1 in US6-transfected of nontransfected CMT64.5 cells. At 10 h post-infection, gpUS6-mediated inhibition of peptide translocation was determined as described in Fig. 4B. Standard deviations are based on three independent experiments.

DISCUSSION
To date biological inhibition of ABC transporters is limited to very few herpesviral proteins like gpUS6 and ICP47, representing elaborated inhibitors designed to interrupt the catalytic transport cycle of TAP at different stages (45). To elucidate molecular aspects of TAP structure and function, we used HCMV gpUS6 and probed its luminal interaction with TAP. We demonstrate that the gpUS6-mediated inactivation of TAP is species-restricted, i.e. gpUS6 affects peptide translocation by human TAP but not by rat TAP. gpUS6 recognition of both transporter subunits is essential to control the transport function. Based on rathuman intrachain TAP1 and TAP2 hybrids, nonanalogous recognition and inhibition of TAP1 and TAP2 by gpUS6 was found (see Tables 1 and  2). By forming oligomers, gpUS6 is able to reach contact sites on the N-terminal TMD of TAP2 and the last C-terminal luminal loop of TAP1. gpUS6 binding at this site of TAP1 provides experimental evidence for a TAP1 topology with 10 TMSs (see Table 1).
Species Specificity of HCMV gpUS6-Herpesvirus family members like HCMV have co-evolved over millions of years with a distinct natural host. As a result of selection pressure, immuno-evasive proteins adapted to a very high target specificity reaching a species-restricted phenotype. The affinity of the HSV ICP47 protein to the cytosolic face of human TAP has been determined to be 50 nM, whereas the binding affinity to mouse TAP is at least 50 -100-fold lower (21,42). At first glance gpUS6 resembles the unrelated ICP47 protein by inhibiting TAP in human and monkey cells but fails to block rat and mouse TAP. Human and rat TAP exhibit a high degree of sequence relatedness reaching about 80% homology. Their conserved structure is highlighted by a replaceable function, i.e. both subunits of rat TAP assemble with the complementary human subunit to form a transport-competent interspecies hybrid transporter. Thus, it is remarkable that both ICP47 and gpUS6 distinguish between human and rat TAP. This feature of gpUS6 was our starting point to elucidate minimal gpUS6 binding domains on TAP by constructing functional rat-human intrachain TAP hybrids.
The Membrane Topology of TAP as Determined by gpUS6-By inhibiting the TAP heterodimer exclusively via the ER luminal surface, gpUS6 qualifies as a unique sensor recognizing distinct loops surfacing at the ER membrane. Assembly of TAP subunits into a preformed heterodimer is absolutely required for gpUS6 recognition of both TAP1 and TAP2, because neither TAP1 nor TAP2 monomers could bind to gpUS6, as observed previously with TAP1 (17). This finding indicates profound conformational rearrangements of the TAP1 and TAP2 TMD upon heterodimer association. To dissect the sequence requirements of TAP1 and TAP2 subunits for gpUS6 binding, full-length human/rat intrachain TAP chimeras were analyzed. Pursuing this approach, two contact sites of gpUS6 on TAP1 TMD could be mapped, which exhibit a different and hierarchical impact upon TAP function. The major target sequence for gpUS6 inhibition was identified at the fifth predicted ER-luminal loop, although gpUS6 binding was also observed to the fourth ER loop. The evidence calling for the fifth ER luminal loop is in contrast with earlier results suggesting that these sequences may locate to the cytoplasmic peptide binding domain (11). Based on the introduction of N-linked glycosylation consensus sites into nonfunctional C-terminal truncated TAP mutants, it was concluded that this part of TAP would be exposed to the cytosol without reaching the ER luminal surface, meaning that TAP1 and TAP2 subunits would form only 8 and 7 TMSs, respectively (11,46). Lacking a stabilizing effect of the C-terminal NBD, the TMD topology of truncated TAPs may, however, not have formed regularly. This could explain why no evidence for membrane integration of the isolated TMS9 was obtained previously (11). Based on the fact that gpUS6 does not prevent peptide binding to TAP (14, 16), we conclude that exposure of this loop between TMS9 and TMS10 into the ER represents a conformational state of TAP1 with high substrate affinity.
Hydrophobicity analysis and sequence alignments with the monomeric multiple drug resistance transporter led to an alternative model of TAP topology that fits with our results, predicting a core TMD with 6 TMSs and additional 4 and 3 N-terminal TMSs for TAP1 and TAP2, respectively (2). The last C-terminal loop of TAP1, predicted to be formed by only 9 or less aa between TMS9 and -10, differs at two positions between human and rat. Therefore, it was tempting to hypothesize that exchanging the human threonine and serine residues for the rat valine and arginine would suffice to disrupt gpUS6 recognition. At least two explanations are possible for why gpUS6 binding was maintained. The hydrophobicity of TMS9 and TMS10 is relatively moderate compared with that of TMS1-8, raising the possibility that neighboring sequences of the TMD are relatively mobile within the ER membrane and become exposed to the ER lumen and accessible for gpUS6 at a certain step during the TAP transport cycle. In the flanking sequences of the last loop, further sequence heterogeneity between human and rat TAP1 is found, which may account for the observed species-restricted binding of gpUS6. Alternatively, distantly located cis-or trans-acting rat sequences may prevent the gpUS6 target sequence from being formed by the rat TAP1 TMD. Notably, the induction of this gpUS6 target structure in human TAP1 critically depends on ATP-binding sequences present in the Walker A motif of TAP2, 5 suggesting that this loop is not constitutively present in TAP1 but rather coupled to signaling from the cytosolic NBD. Our data indicating the presence of TMS9 and TMS10 are also concordant with the lateral diffusion of TAP that was measured using a fluorescence recovery after photobleaching technique (46). The diffusion of integral membrane proteins is not dependent on their mass in the luminal part but on the size and number of TMSs (47). Depletion of ATP led to fast diffusion of TAP, which is consistent with a model of only 8 constitutive TMSs on TAP1 or at least incomplete incorporation of TMS9 and -10. gpUS6 binding to TAP reduced the rate of TAP diffusion (46), consistent with an arrested TAP1 structure with 10 TMSs.
Previous studies have demonstrated that gpUS6 binding to TAP prevents ATP binding to TAP1 (17,18). We make the tacit assumption that the gpUS6-mediated stabilization of ER-exposed sequences in the TMD of TAP1 accounts for this inhibition of ATP binding to the NBD of TAP1.
gpUS6 Interference with TAP2-At first sight the binding and inhibition results of gpUS6 on TAP2 are puzzling. Our data delimit binding of gpUS6 to TAP2 to the luminal N-terminal TMD (human aa 1-177). This part of TAP2 has two predicted parts in the ER lumen, the N-terminal tail (aa 1-7) and the first loop, predicted to encompass aa 76 -97. It appears unlikely that gpUS6 recognizes TAP2 at its small N-terminal tail, and the first luminal loop is a more expedient target. This site is part of the TMD that was thought to form the pore domain that is build in a head-head/tail-tail orientation (46,48). Recently, Koch et al. (10) demonstrated this part of TAP2 being dispensable for peptide transport in insect cells but required for tapasin association. Tapasin is, however, not required for the inhibitory function of gpUS6 (14). Therefore, it is surprising that gpUS6 interacts more strongly with the N-terminal TMD of TAP2 than with its C-terminal portion, given that the former interaction does not suffice to mediate TAP inhibition. rTAP1 combined with TAP2 chimeras containing a rat N-terminal and a human C-terminal TMD were not precipitated by gpUS6 but were blocked by gpUS6. At least two scenarios are possible to reconcile these findings. Although dispensable for peptide transport, the N-terminal TMD of hTAP2 gains a regulatory function onto peptide transport when directly complexed with gpUS6. Hence, this regulation is secondary and not manifest in the absence of the gpUS6 interaction with the C-terminal TMD. We were not able to demonstrate direct binding between gpUS6 and the C-terminal TMD of TAP2, which implies that only a transient or slight interaction takes place. Consequentially, a stabilizing binding to TAP2 is plausible. Alternatively, the gpUS6-binding site on the N-terminal of TAP2 may not be directly involved in the gpUS6-mediated TAP inhibition but in the stabilization of the peptide loading complex (PLC). We have observed a stabilization of the PLC in the presence of gpUS6 (14). The N-terminal gpUS6 binding to TAP2 proximal to the tapasin-binding site might prevent components of the PLC from leaving the complex. This way gpUS6 would not only inhibit peptide translocation but also hinder PLC components from being recruited to other PLCs still not arrested by gpUS6. In this model, inhibition of peptide translocation would mainly be operated from TAP1, whereas interaction to TAP2 would only be transient but still required for efficient inhibition.
Altogether, our data demonstrate a nonanalogous interaction with the transporter subunits. The findings appear consistent with a model of TAP function that assumes coordinated but nonsynchronous conformational rearrangements of TAP subunits in which TAP2 is always one step ahead (9).
Structural Requirements of gpUS6-Our findings suggest that at least four independent gpUS6 contact sites must exist on TAP, two located in the TAP1 subunit and two mapping to TAP2. The formation of the interfaces toward TAP1 and TAP2 is required for efficient TAP inactivation. Our data propose a model in which gpUS6 forms larger oligomeric complexes that could bridge between distant contact sites on the TMDs of TAP1 and TAP2. A similar oligomerization pattern was found also for the gpUS6-related HCMV glycoprotein US3 (49), which prevents MHC class I molecules from leaving the ER. The authors propose a model in which interacting gpUS3 molecules enhance their ability to retain MHC class I molecules in the ER. Oligomerization of gpUS3 was described not to involve disulfide bonds, whereas we found gpUS6 clusters to be sensitive to reducing agents. The active luminal domain of gpUS6 contains eight cysteines and thus is likely to form a complex intra-as well as intermolecular network of disulfide bridges. Indeed, expression of this active soluble domain in Escherichia coli yielded functional gpUS6 monomers and dimers (18). Based on our findings, one may envisage gpUS6 mutants selectively lacking the interaction with either TAP1 or TAP2. By studying such gpUS6 mutants, segregation of the molecular inhibition upon TAP subunits will be possible and will complement our findings based on rat human intrachain chimeras.