All 4 di-leucine motifs in the first hydrophobic domain of the E5 oncoprotein of human papillomavirus type 16 are essential for surface MHC class I downregulation activity and E5 endomembrane localization
Marc S. Cortese,
Division of Pathological Sciences, Institute of Comparative Medicine, University of Glasgow, Glasgow, Scotland, United Kingdom
The E5 oncoprotein of human papillomavirus type 16 downregulates surface MHC Class I and interacts with the heavy chain of the MHC complex via the first hydrophobic domain, believed to form the first helical transmembrane region (TM1) of E5. TM1 contains 4 equally spaced di-leucine (LL1-LL4) motifs. Di-leucine motifs have been implicated in protein–protein interactions and as localization signals. To see if any of the 4 di-leucine motifs of TM1 are involved in MHC downregulation by E5, we mutated each LL pair into valine pairs (VV1-VV4), as mutation of leucine to valine is not expected to cause major structural alterations in E5. We found that all 4 mutations disrupted the intracellular location of E5 and abrogated its MHC I downregulating activity; however VV2 and VV4 mutants were still able to interact physically with the MHC I heavy chain (HC) in vitro, while VV1 and VV3 mutants had lost this activity. We conclude that LL1 and LL3 are necessary for the interaction with HC, but LL2 and LL4 are not. However all 4 LL motifs are responsible for the proper localization of E5 in the Golgi/ER, and the displacement of E5 from this location contributes to the abrogation of MHC I downregulation. LL1 and LL3 motifs are expected to be on one face of the TM1 helix and LL2 and LL4 on the opposite face. We propose that E5 interacts with HC via LL1 and LL3 and that all 4 di-leucine motifs act as a targeting signal.
Viruses must be able to overcome the host immune response to replicate themselves and give rise to new infectious progeny. To this end they have evolved a number of immune escape strategies, from rapid antigenic variation to a direct fight with immune molecules.1 Papillomaviruses (PVs) are small oncogenic DNA viruses that infect epithelia and cause benign hyperproliferative lesions. In most cases, infection is cleared following activation of a host immune response against the virus, but clearance of infection and regression of lesions can take months or years. Occasionally the lesions do not regress and persistence of PV infection and failure of virus clearance can lead to onset of malignancy and progression to cancer. This is the case for human papillomavirus Type 16 (HPV-16), which is involved in the majority of cases of cervical cancer.2
One of the potential effectors of HPV-16 escape from host immunosurveillance is the viral oncoprotein E5.3, 4 HPV-16 E5 is a hydrophobic membrane protein of 83 amino acids, possessing 3 well-defined hydrophobic regions, localized mainly in the membranes of the endoplasmic reticulum (ER) and Golgi apparatus (GA). While E6 and E7, the main transforming proteins of HPV, are expressed throughout the course of the disease and are necessary for the maintenance of a transformed phenotype, E5 plays a lesser role in transformation. E5 is expressed during the early stages of infection and its expression is often, but not always, extinguished as the lesion progresses toward malignancy. E5 has been implicated in facilitating amplification of the viral genomes in differentiating suprabasal keratinocytes.5, 6 These characteristics point to a role of E5 in establishment of PV infection and the initiation of cell transformation,7 as shown by the induction of cell proliferation in vitro8 and neoplasia in vivo in transgenic mice,9via the activation of kinase cascades, primarily activation of the epidermal growth factor receptor (EGF-R).
We have previously shown that HPV-16 E5 downregulates the expression of surface major histocompatibility complex Class I (MHC I/HLA I) by retaining the complex in the GA3 and by physically interacting with the heavy chain (HC) component of the complex, and that both interaction with HC and downregulation of surface HLA I are mediated by the first hydrophobic domain of E5.4 These N-terminal 30 residues are necessary and sufficient not only for downregulation of surface MHC I but also for the characteristic perinuclear localization of E5; the remaining residues 31–83 have no effect on E5 localization and are also dispensable for surface MHC I downregulation activity. Downregulation of surface MHC I appears to be a characteristic common to all E5 proteins.10, 11
The literature consensus12 and predictive modeling techniques13 are in agreement that residues 8–30 form the first of 3 transmembrane helices (TMs) in E5 (referred to herein as TM1). There are 4 di-leucine motifs within TM1. While many E5 proteins such as those of BPV-4, HPV-11 and HPV-83 contain di-leucines, only those of HPV-16 E5 occur in pairs that, in turn, are particularly spaced. While inclusion of leucines is a common feature of TM sequences, the pairing and spacing of those in HPV-16 E5 suggest that they may form a structural motif and/or be functionally important. The di-leucine motifs could contribute to surface MHC I downregulation. First, they could serve as molecular recognition elements that mediate binding of E5 to endogenous proteins that participate in or can affect the MHC I maturation pathway. Alternatively, the di-leucines could serve as targeting signals that could direct E5, along with the bound HC, to a location that would prevent MHC I assembly and/or maturation.
Here we analyze the contribution of each of the 4 di-leucine motifs of TM1 to MHC I downregulation, HC binding and cellular localization of E5. We show that all 4 di-leucine motifs are important for MHC I downregulation and E5 cellular localization, and that the first and third di-leucine motifs interact physically with HC in vitro.
ER: endoplasmic reticulum; GA: Golgi apparatus; HC: heavy chain; HLA: human leukocyte antigen; HPV-16: human papillomavirus type 16; MHC I: major histocompatibility complex Class I; TM: transmembrane
Material and Methods
Immortalized human keratinocytes (HaCaT) were grown in monolayers in high glucose Dulbecco's modified Eagle medium (DMEM, Invitrogen), supplemented with 10% foetal calf serum and 2 mM glutamine in 5% CO2 at 37°C. Cells harboring pCI- or pEGFP-based constructs were supplemented with 500 μg ml−1 G418. Transfections were performed using TransIT-LT1 (Mirus) according to manufacture's directions. Stable transfected lines of pCI-16E5 mutants were selected with G418 for 4 weeks before individual clones were isolated for characterization. Cells transfected with pEGFP-based constructs were characterized 2 days after transfection.
Construction of di-leucine mutations in E5 expression vectors
HPV-16 E5 expression constructs in pCI (Promega) and pEGFP (Clontech) have been previously described.3, 4 The di-leucine to di-valine point mutations (Fig. 1a) were constructed by site directed mutagenesis using these constructs as templates.
Total RNA was isolated from HaCaT cells using the RNeasy Mini Kit (Qiagen) with QIAshredder spin columns (Qiagen), DNA was removed by treatment with DNase I (Invitrogen), first strand cDNA was synthesized using SuperScript III First-Strand Synthesis System (Invitrogen), residual RNA was removed with ribonuclease H (Invitrogen) and real-time relative quantification PCR was performed with β-actin as the endogenous control using Brilliant II SYBR Green QPCR Master Mix (Stratagene). Primers used were: 5′-ATGTACCCATACGATGTTCCAGA-3′ and 5′-TGCATGTGTATGTATTAAAAATAATGGTATATAA-3′ for E5 forward and reverse, respectively; and 5′-CCTCGCCTTTGCCGA-3′ and 5′-GGATCTTCATGAGGTAGTCAGTC-3′ for human β-actin forward and reverse, respectively.14 Reactions were performed using an ABI Prism 7500 Sequence Detection System. E5 mRNA expression levels were normalized relative to β-actin mRNA expression levels and converted to fold expression relative to the wild-type E5 sequence.
Surface expression of MHC I by flow cytometry was performed essentially as previously described.3 Briefly, HaCaT cells were trypsinised, incubated in 5% CO2 at 37°C for 30 min to allow surface antigens to equilibrate, blocked with 1% Albumin Bovine Serum Fraction V (Sigma) in PBS, then incubated with anti-human HLA Class I monoclonal antibody diluted 1:100 (MCA81, Serotec) at 4°C for 30 min. After washing, the cells were incubated with anti-mouse IgG-FITC diluted 1:100 (F2653, Sigma) at 4°C for 30 min in the dark and washed. The cells were analyzed immediately by flow cytometry using a Beckman Coulter EPICS Elite analyzer.
In vitro transcription/translation and coimmunoprecipitation
HLA-A2 heavy chain (HC) was expressed from a cDNA clone as previously described.4 Individual 50 μl in vitro transcription-translation reactions were run for each construct with 2 μg of plasmid using the TNT T7 Quick Coupled Transcription/Translation System (Promega) in the presence of NEG-709A EasyTag L-[35S] Methionine (Perkin Elmer) and Canine Pancreatic Microsomal Membranes (Promega) following the manufacturers instructions as previously described.15 For each coimmunoprecipitation, 22 μl of an E5 reaction was combined with 6-μl aliquot of a master HC reaction. The mixture was then precleared by incubating on rollers at 4°C for 1 hr with 20 μl Protein-A Sepharose (Sigma) that had been previously blocked by incubation in 1% BSA (Sigma) for 1 hr. The supernatant was transferred to a clean tube and 5 μl of a 1:3.6 dilution of antibody EP1395Y (Abcam) was added. After incubation overnight, 20 μl of preblocked Protein A-Sepharose bead suspension (Sigma) was added and the incubation continued for an additional hour. The sepharose and bound proteins were spun down and washed twice in high salt buffer (50 mM Tris HCl, pH 7.5, 500 mM NaCl, 1% Nonidet P40, 0.5% sodium deoxycholate, 0.1% sodium dodecylsulfate), once in low salt buffer (50 mM Tris HCl, pH 7.5, 1% Nonidet P40, 0.5% sodium deoxycholate, 0.1% sodium dodecylsulfate) and transferred to a clean tube. Control immunoprecipitations were conducted as above except either the HC or E5 reaction was omitted. Proteins were eluted from the Protein A-sepharose by incubating in 1× LDS loading buffer (Invitrogen) at 70°C for 10 min and electrophoresed in 10% NuPAGE gels (Invitrogen). Gels were fixed with glacial acetic acid and methanol, incubated for 30 min in Amplify Fluorographic Reagent (GE Healthcare), air dried and exposed to Hyperfilm ECL (GE Healthcare) for autoradiography at −70°C.
For immunofluorescence studies, stable or transiently transfected cells were grown on cover slips for 24 hr then treated with fixing solution (4% formaldehyde, 2% sucrose in PBS) for 10 min at room temperature. Cells were then washed with PBS and permeabilised by incubating in 0.25% NP40, 2% sucrose in PBS for 5 min at room temperature. Cells were blocked with ImageIT Enhancer (Invitrogen) according to manufactures instructions. After washing with PBS, cells were incubated with anti-MHC I antibody MCA81 (AbD Serotec) at 1:100 and Bap31 antibody ab15044 (Abcam) at 1:200 for 1 hr at room temperature. After washing with PBS, the cells were incubated with Alexa Fluor 647 anti-mouse IgG and Alexa Fluor 555 anti-rat secondary antibodies (Invitrogen), both at 1:250 dilutions, for 1 hr. The coverslips were then washed in PBS, dried and mounted with ProLong Gold (Invitrogen). Images where collected with a Leica TCS SP2 confocal microscope and processed with ImageJ software (National Institutes of Health, US).
Expression of E5 in cell lines
To investigate the role of the 4 di-leucine pairs of the first hydrophobic region for their contribution to E5 function, we chose to treat each pair as a separate motif and to mutate each pair individually. The designations assigned to these 4 pairs of leucines were: LL1 comprised the leucines at positions 11 and 12 and similarly the di-leucine pairs at positions 16/17, 22/23 and 27/28 were designated LL2, LL3 and LL4, respectively (Fig. 1a). We chose to make conservative mutations as a relevant structure could be disrupted if radical changes in amino acid properties were introduced and our intent was to study targeting and binding motifs rather than gross structural changes. The leucines were mutated pair-wise to valines and the mutants designated VV1, VV2, VV3 and VV4 (Fig 1a).
To screen individual clones of stably transfected HaCaT cells for expression of wild-type and mutant E5 mRNA, we determined the transcript levels relative to β-actin. The relative mRNA expression levels for 2 sets of clones are shown in Figure 1b. All mutant clones expressed transcripts at levels similar to wild-type E5, with the exception of the VV1#3 clone which expressed considerably more mRNA. We then assayed all clones for the expression of surface MHC I.
Effects of mutations on the surface MHC I downregulating activity of E5
Flow cytometry revealed that cells stably expressing wild type E5 exhibited the characteristic reduction of surface MHC I3, 4 compared with the vector only control (Fig. 1c). Mutations of any of the 4 di-leucine pairs abrogated E5 downregulation of surface MHC I expression to approximately the same extent (Fig. 1c); even the VV1#3 clone that expressed 12 times more mRNA than wild-type E5, showed similar levels of surface MHC I. This indicated that, at least for the VV1 mutation, dosage effect was not sufficient to counter the detrimental effect of the mutation.
VV1 and VV3 do not interact with HLA-A2 HC
As E5 has been shown to bind HLA HC through TM1,4 we asked whether any of the di-leucines contributed to this activity. We tested each mutant E5 sequence for its ability to be coimmunoprecipitated in vitro with an antibody specific for HLA-A2 HC. For controls, we verified that the E5 wild-type sequence coprecipitated with HLA-A2 HC and that there were no interfering bands when HLA-A2 HC was precipitated alone (Fig. 2, Lanes 1 and 6, respectively). Additionally, we verified that wild-type and mutant E5 sequences were all equally translated (Fig. 2, E5 input row) and not precipitated by the anti-HLA antibody (data not shown). When the 4 mutants were tested, only VV2 and VV4 coprecipitated with HLA-A2 HC (Fig. 2, Lanes 2–5). The inability of VV1 and VV3 to bind HC suggests that LL1 and LL3 are specific determinants for the HC binding ability of E5. On the contrary, LL2 and LL4 do not appear to be involved in HC binding as their mutation to VV did not affect their binding to HC in vitro.
All 4 mutants disrupt cellular localization of E5
Next we asked if localization of E5 was changed by any of the mutations since the first 30 residues of E5 have been shown to localize to the endomembrane compartment like wild type E5.4 As visualization of E5 in stable cells is difficult, due to the very low levels of expression,4 we looked at cells transiently transfected with a construct expressing the various E5 proteins fused to EGFP. The cells were also stained for Bap31, a chaperone of MHC I,16 used here as a marker for the ER/Golgi system. EGFP expressed from the empty vector was distributed uniformly throughout the cell (Fig. 3a, pEGFP). The wild-type EGFP-E5 fusion showed characteristic perinuclear localization (Fig. 3a, E5) while EGFP-E5 in each of the di-leucine mutations had a more diffuse localization (Fig. 3a, VV1, VV2, VV3, VV4). Bap31 also showed perinuclear localization in all cell lines (Fig. 3a, red). Wild type E5 colocalized with Bap31 (Fig. 3a, E5) while less colocalization was seen in cells expressing any of the di-leucine mutants, particularly VV2 and VV4 (Legend to Fig. 3a). In cells with wild-type E5, some Bap31 could be seen outside the region of E5/Bap31 colocalization but no EGFP-E5 was detectable outside the region demarked by Bap31 (i.e., the ER/Golgi complex). Intriguingly, Bap31 distribution appeared more restricted in cells expressing wild-type E5 than in cells expressing the di-leucine mutants, however this was not observed in cells stably expressing E5 where there was no difference in Bap31 distribution between cells with wild type E5 and the LL mutants (Fig. 3b).
None of the dileucine mutants exhibited the ability to restrict MHC I distribution
Since we observed a differential ability among the di-leucine mutants to bind HLA-A2, an investigation into how each mutant affected the cellular distribution of MHC I was warranted. For this assay we used the set of stably transfected HaCaT cells that expressed levels of E5 mRNA closest to wild-type E5 cells (Fig. 1b) and again used Bap31 as a marker for the ER/Golgi system. Cells expressing the vector only (pCI) showed perinuclear localization of Bap31 (red) and MHC I (green) with some of the MHC I colocalizing with Bap31 and some outside the region demarked by Bap31 (Figs. 3b and 3c, pCI). This represents the typical localization pattern for MHC I with a pool of the complex within the endomembrane system undergoing peptide loading and quality control and another pool on the cell surface. In the E5 expressing cells, the Bap31 localization was similar to control cells but the MHC I distribution had been altered so that much less was visible outside the region demarked by Bap31 (Figs. 3b and 3c, E5). This was consistent with the reduced surface expression of MHC I observed by flow cytometry in cells expressing wild-type E5 (Fig. 1c). In all the di-leucine mutant-expressing cell lines, the MHC I distribution was similar to the control cells with a pool of MHC I within the endomembrane system and another pool outside the region demarked by Bap31 (Figs. 3b and 3c, VV1-4). These results again are consistent with the flow cytometry results (Fig. 1c) which showed that none of the mutants retained the surface MHC I down-regulation activity of wild-type E5.
We conclude that (i) all the LL motifs are necessary for the proper localization of E5 and hence downregulation of surface MHC I, and (ii) LL1 and LL3 mediate E5 binding to HC whereas LL2 and LL4 are not involved.
The first hydrophobic TM domain of E5 encodes multiple functions. Besides being responsible for surface MHC I downregulation,4 it has been shown to confer HaCaT cells with the ability to grow in an anchorage independent manner18 and to mediate the ability of HaCaT cells to invade the matrix in organotypic raft assays.19 The importance of this domain for E5 functioning and HPV persistence is exemplified by the observation that mutations in E5 TM1 occur at a much lower frequency than in the rest of the protein in HPV-16 sequences from public database and clinical samples.19–22 The paucity of recorded mutations in this region suggests that either there are critical structural aspects that contribute to its function or that it contains multiple functional motifs (e.g., interaction, targeting, modification or inhibitory) such that there is no sequence wherein mutations could be tolerated.
Here we have studied the importance of TM1 of E5 in the downregulation of MHC I by mutating its 4 LL pairs to valines. The substitution of valine for leucine is not expected to lead to structural differences which may affect interaction of E5 with other proteins as valine and leucine have similar propensities for helix formation and, because valine has a smaller volume than leucine, it is unlikely to block specific interactions by steric interference. Mutants of individual di-leucine pairs have been studied previously, however in these cases the intent of the chosen mutagenesis strategies was to disrupt the helical structure of TM1,24, 25 whereas our object was to preserve it, precisely to eliminate gross structural deformations which would obscure the functions of individual LL pairs. By this analysis we show conclusively that all di-leucine motifs in TM1 are important and necessary for the downregulation of surface MHC I. This conclusion is based on the fact that mutation of each of the LL motifs abrogates the ability of E5 to downregulate MHC I and to retain the complex in the Golgi/ER. The reasons for the disruption of this activity can be various: failure of E5 mutants to bind HC, mis-localization of the mutants or inability to bind to other cellular partners.
Indeed, the mutations of LL to VV impact on E5-HC interaction but in a differential way: VV1 and VV3 mutants fail to bind HC in vitro while VV2 and VV4 mutants still do so. Thus LL1 and LL3 are specific determinants for the HC binding ability of E5, while, on the contrary, LL2 and LL4 do not appear to be involved in HC binding. It follows that the reason for the lack of surface MHC downregulation by VV2 and VV4 E5 mutants has to be ascribed to the different function of LL2 and LL4.
This conclusion is confirmed by the mis-localization of all 4 mutants away from the endomembrane compartments in which E5 is usually found to a more diffuse cytoplasmic location, particularly in the case of VV2 and VV4 (see Legend to Fig. 3a). These findings lead to the hypothesis that, despite their ability to interact with HC at least in vitro where the proteins are per force expressed in the same membranes, the failure of the VV2 and VV4 mutants to downregulate MHC I is due, at least in part, to their presence outside compartments inhabited by the MHC I complex. Di-leucines are known components of targeting motifs, either on their own26 or, more commonly, combined with acidic residues.27 However, to our knowledge, no instance of a di-leucine located within a TM of a protein has been shown to act as a targeting signal.
To reconcile the finding that only the first and third di-leucine pairs are required for binding to HLA HC with the fact that all 4 pairs abrogate surface MHC I downregulating activity and disrupt the characteristic localization pattern of E5, it is important to note that the helical structure of TM1 would place di-leucines 1 and 3 on 1 side of the helix with di-leucines 2 and 4 appearing on the opposite face (Fig. 4a). Given this configuration, it is improbable that both sides of the TM1 of E5 could interact with the single TM of MHC I, and therefore only the LL1/LL3 face likely take part in this interaction, leaving the helical face containing LL2/LL4 free to interact with other endogenous proteins.
The pathway for progressive maturation and quality control of the peptide-loaded MHC I complex would provide multiple opportunities for further interactions between E5 and components of the MHC Class I peptide loading complex (PLC). For instance, in addition to binding the HC of multiple HLA haplotypes (A1, A2, A3 and B8),4 E5 has also been shown to colocalize and interact with the chaperone calnexin via the first hydrophobic region of E518, 25 and this interaction appears to be required for E5-mediated downregulation of surface MHC I.24 It is not unreasonable that E5 could interact with HC via LL1 and LL3, and with calnexin via LL2 and LL4, either in turn or simultaneously, at some stage of the MHC I maturation process.
A second logical target for E5 to achieve the same ends is the quality control/chaperone function mediated by Bap31. Bap31 involvement in the transport of MHC I from the ER/Golgi to the cell surface has been demonstrated.16 Additionally, Bap31 has been identified as an E5 partner in a 2 hybrid screen and this association was confirmed by in vivo coprecipitation.28 Further, here we show that mutation of any di-leucine in TM1 of E5 disrupts the colocalization between E5 and Bap31. Basic informatics analyses show that a motif of 10 identical residues exists between the first TM of E5 and the third TM of Bap31 (Fig. 4b). The possibility that this identity is relevant is strengthened by the fact that the polarity (luminal N-terminus) and displacement within the membrane (spanning the luminal half) of both these regions are the same (Fig. 4c). Also note that a helical wheel representation of the third TM of Bap31 would place the identical residues in the same spatial orientation as they would be found in E5 (Fig. 4a). The existence of this shared motif suggests that E5 may take advantage of some or all the determinants that mediate Bap31 binding to MHC I. The fact that 2 of the leucine pairs involved in the surface MHC I downregulation activity of E5 are among the identities shared between Bap31 and E5 makes this motif a tempting target for further investigation of interactions among Bap31, MHC I and E5.
Knowing which face of the E5 TM1 interacts with HLA HC and the fact that the TM region of the HC contains a GXXXG motif (Fig. 4c) that has been shown to define a highly preferred helical face for TM-TM interactions,29 provide ample clues that could lead to the identity of the complete set of determinants that mediate binding between E5 and HC. With more knowledge of the specifics of this interaction, work toward inhibitors that could serve as preventative and/or curative drug intervention strategies becomes possible.
The authors thank Prof. Iain Morgan and Dr. Tina Rich for critical reading of the manuscript. MSC is a Cancer Research UK Fellow. This work was supported by the Association for International Cancer Research.