A Novel PDZ Domain Interaction Mediates the Binding between Human Papillomavirus 16 L2 and Sorting Nexin 27 and Modulates Virion Trafficking

Plasmid constructs.

SNX27a and SNX27b, which express the two isoforms of SNX27 (Fig. 1A), and the SNX27b ΔPDZ and SNX27b ΔRA/FERM domains, all of which express myc-tagged SNX27 and fragments thereof (see Fig. 7A), were cloned into plasmid pCI-neo. The plasmid construct expressing SNX17-Flag has been described previously (23). Plasmid pCA16L2, expressing wild-type (WT) HPV-16 L2 (16L2) with amino-terminal Flag and hemagglutinin (HA) tags, has been described previously (8), and the 16L2 truncation mutants used in this study were produced by the introduction of stop codons into the open reading frame of pCA16L2 using the GeneArt site-directed mutagenesis system (Invitrogen), as were the SNX17 and SNX27 C-terminal mutations and the SNX27 Δ67–77 and Δ96–107 mutations.

Cell lines.

HEK293 cells (human embryonic kidney cells), H1299 cells, HaCaT cells (spontaneously immortalized human keratinocytes), and HEK293TT cells (24) were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS), penicillin-streptomycin (100 U/ml), and glutamine (300 μg/ml).

Pseudovirion production.

Pseudovirions with a packaged luciferase reporter were generated by transfecting HEK293TT cells with constructs expressing the various papillomavirus L1 and L2 open reading frames in codon-optimized bicistronic plasmid constructs along with the luciferase reporter pGL3Luci, as described previously (24).

Infectivity assays.

HaCaT cells were seeded in a 12-well plate at a density of 3 × 10⁴/well. After adherence, small interfering RNAs (siRNAs) against SNX17 and SNX27, and siSTABLE nontargeting siRNA 1 (referred to below as siScr [scrambled siRNA]), as a control (Dharmacon), were added to cells in pairs of wells by using Lipofectamine RNAiMax (Invitrogen) according to the manufacturer's instructions. siRNAs were dissolved and were used at a concentration of 20 μM, and because we observed differences in knockdown efficiency, their optimal inputs were determined empirically by pilot experiments. All siRNA inputs were balanced with the control siScr to bring the total siRNA concentration to the same amount per transfection. Briefly, for each two-well transfection, for the series of control, SNX17, SNX27, and double (SNX17-plus-SNX27) knockdowns, we used 4 μl siScr, 1 μl siSNX17 plus 3 μl siScr, 3 μl siSNX27 plus 1 μl SiScr, and 3 μl siSNX27 plus 1 μl siSNX17, respectively. After 48 h, HPV-16 pseudovirions were added at a concentration of approximately 12 ng/ml. Infection was monitored 48 h later by luminometric analysis of firefly luciferase activity in the cell lysates using a luciferase assay system kit (Promega), according to the manufacturer's instructions. The efficiency of siRNA knockdown was confirmed by SDS-PAGE and Western blotting of cell lysates from the luminometry assays. Blots were serially probed with anti-SNX17 antibodies (Santa Cruz), anti-SNX27 antibodies (Abcam), and anti-α-actinin antibodies (Sigma-Aldrich) and were visualized using the Amersham ECL Western blotting detection kit (GE Healthcare).

In vivo expression for coimmunoprecipitation or GST-binding assays.

Proteins were in vitro transcribed, translated, and radiolabeled using a TnT rabbit reticulocyte lysate kit (Promega) labeling with either [³⁵S]methionine or [³⁵S]cysteine. The radiolabeled proteins were incubated with glutathione S-transferase (GST) or GST-fused recombinant proteins in phosphate-buffered saline (PBS) with 1% Triton X-100. After a wash in the same buffer, bound proteins were eluted and separated by SDS-PAGE, and the gel was stained with Coomassie blue, dried, and exposed to X-ray film.

For the binding of GST fusion proteins to proteins expressed in mammalian cells, HEK293 cells were first transfected with plasmids for in vivo expression by the calcium phosphate precipitation method (25) and then lysed in a buffer containing 50 mM HEPES (pH 7.0), 250 mM NaCl, 0.1% NP-40, and 20 mg ml⁻¹ aprotinin. GST or GST-fused recombinant proteins bound to glutathione agarose (Sigma) were incubated with the cell lysates for 1 to 2 h at room temperature; then, after several washes in the lysis buffer, the bound proteins were eluted from the resin in PAGE sample buffer, separated by SDS-PAGE, transferred to nitrocellulose membranes, and probed with appropriate antibodies. Bound proteins were visualized using the Amersham ECL Western blotting detection kit (GE Healthcare).

For detection of the interaction between endogenous SNX27 and 16L2, transfection of the construct expressing Flag-tagged 16L2 (pCA16L2) was carried out as described above, but after 24 h, the cells were lysed in a buffer containing 20 mM Tris (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, and 1% Triton X-100. Immunoprecipitations were carried out using 1 μl of anti-SNX27 antibodies (Abcam) or 1 μl of rabbit anti-Flag antibodies (Sigma-Aldrich) for 2 h at 4°C. Then immune complexes were collected using protein A Sepharose that had previously been washed in lysis buffer. After SDS-PAGE separation of the immunoprecipitated proteins and transfer to nitrocellulose membranes, the Western blots were probed with either anti-SNX27 or anti-Flag antibodies to visualize the coimmunoprecipitated proteins. The precipitated proteins were detected using a SuperSignal West Femto Maximum Sensitivity Substrate kit (Pierce).

Immunofluorescence.

H1299 cells were plated out on coverslips at approximately 10⁵ per well. After 24 h, plasmids were transfected by calcium phosphate-mediated transfer (25). After a further 24 h, cells were washed once with PBS, fixed with 3.7% paraformaldehyde for 20 min at room temperature, washed once with PBS, and then permeabilized with 0.5% Triton X-100 in PBS for 5 min at room temperature. After two washes with PBS and one with 0.1 M glycine in PBS, coverslips were incubated with primary antibodies appropriately diluted in PBS for 2 h at 37°C. After three washes with PBS, the coverslips were incubated with fluorescein-conjugated secondary antibodies (Life Technologies) diluted in PBS for 1 h at 37°C. After a further three washes in PBS, and a final wash in distilled water (dH₂O), the coverslips were mounted on glass slides for UV microscopy.

Both SNX17 and SNX27 bind to HPV-16 L2.

To examine whether SNX27 can interact with HPV-16 L2 similarly to SNX17, we incubated in vitro-translated, radiolabeled SNX17 and SNX27 with GST alone or GST-fused HPV-16 L2, both bound to glutathione agarose. After several washes, bound proteins were separated by SDS-PAGE, and the gel was stained with Coomassie blue, dried, and exposed to X-ray film. The results (Fig. 1B) show that SNX27, like SNX17, can interact with HPV-16 L2.

HPV-16 L2 binds to endogenous SNX27 in HEK293 cells.

To verify that the HPV-16 L2–SNX27 interaction also occurs in vivo, we transfected HEK293 cells with either an empty plasmid or a plasmid expressing Flag-tagged HPV-16 L2. The whole-cell lysates were divided in two, retaining 10% as an input control, and were immunoprecipitated using anti-Flag or anti-SNX27 antibodies. The immunoprecipitates were analyzed by SDS-PAGE and Western blotting. The membrane from the anti-Flag immunoprecipitation was probed with anti-SNX27 antibodies to detect any SNX27 binding to HPV-16 L2, and the membrane from the anti-SNX27 immunoprecipitation was probed with anti-Flag antibodies to detect any bound HPV-16 L2. The results (Fig. 2) show that HPV-16 L2 can bind to endogenous SNX27 within HEK293 cells, although the levels of interaction between the two proteins appear to be low.

The SNX27-L2 interaction is stronger with HPV-16 than with the other papillomavirus types examined.

Previous studies have demonstrated that the interaction between L2 and SNX17 is highly conserved in a number of papillomavirus types other than HPV-16 (8). To examine the abilities of other papillomavirus L2 proteins to interact with SNX27, we expressed the L2 proteins from HPV-16, HPV-11, HPV-18, HPV-5, bovine papillomavirus 1 (BPV-1), and Sylvilagus floridanus papillomavirus (SfPV) as GST fusion proteins, bound to glutathione agarose, and incubated them with lysates from HEK293 cells in which either SNX17 or SNX27 was overexpressed. After washing, bound proteins were eluted in sample buffer, separated by SDS-PAGE, and analyzed by Western blotting. Bound proteins were visualized by probing with anti-Flag antibodies (SNX17) or anti-myc antibodies (SNX27). The results (Fig. 3) show that SNX27 binds most strongly to HPV-16 L2, followed by weaker interactions with HPV-11 L2, BPV-1 L2, and SfPV L2, while levels of interaction with HPV-5 L2 and HPV-18 L2 are lower still. This is in contrast to the interactions with SNX17, which are equally strong for all the papillomavirus types tested.

The interactions between HPV-16 L2 and both SNX17 and SNX27 are direct.

In order to investigate whether the interactions between HPV-16 L2 and the two sorting nexins are direct, we purified recombinant His-tagged HPV-16 L2 and incubated it with purified GST SNX17 and GST SNX27, bound to glutathione agarose. After several washes, we eluted the bound proteins in sample buffer, separated them by SDS-PAGE, transferred them to nitrocellulose membranes, and probed for recombinant L2 with anti-His antibodies. The results (Fig. 4) show that both SNX17 and SNX27 interact directly with HPV-16 L2.

HPV-16 L2 binds to SNX17 and SNX27 via different domains.

The interaction between papillomavirus L2 proteins and SNX17 has been shown to depend on highly conserved NPxY motifs in the L2 proteins (8). Sequences in the carboxy termini of both SNX17 and SNX27 have been shown to interact with conserved NPxY motifs in some substrates, such as P-selectin (11, 26), and we were interested in determining whether L2 was bound in a similar manner. First, we coexpressed Flag-tagged wild-type HPV-16 L2 or the SNX17 binding-defective mutant of HPV-16 L2 in which NPAY is replaced with AAAA (8) with either myc-tagged SNX27 or Flag-tagged SNX17, and then we immunoprecipitated the cell lysates with either anti-myc antibodies (for SNX27 immunoprecipitations) or anti-SNX17 antibodies (for SNX17 immunoprecipitations). Immunoprecipitated proteins were separated by SDS-PAGE, transferred to nitrocellulose membranes, and probed with anti-Flag antibodies to assess the levels of wild-type and mutant HPV-16 L2 binding to either sorting nexin. After stripping, the membranes were reprobed with anti-myc antibodies to visualize the levels of myc-tagged SNX27. As can be seen in Fig. 5, the NPAY-to-AAAA mutant of L2 binds to SNX27 as efficiently as wild type L2, suggesting that the NPAY binding motif on L2 is not involved in the interaction with SNX27.

To confirm that the carboxy-terminal NPxY motif-binding region of SNX27 is not required for its interaction with L2, we made mutations in both SNX17 and SNX27 (Fig. 6A) that had been shown previously to abolish P-selectin binding (26), expressed the mutants in HEK293 cells, and incubated the cell lysates with GST or GST-L2 bound to glutathione resin. After washing, separation of bound proteins by SDS-PAGE, and transfer to membranes, bound proteins were visualized by probing with anti-myc (for SNX27) or anti-Flag (for SNX17) antibodies. The results, shown in Fig. 6B, indicate clearly that while the mutation in the C terminus of SNX17 abolishes its interaction with L2, the corresponding mutation in SNX27 does not.

Taken together, these data show conclusively that the residues required for the L2–SNX27 interaction are completely different from those required for the interaction with SNX17.

HPV-16 L2 interacts with the PDZ domain of SNX27.

To analyze which region of SNX27 might be involved in the interaction with L2, we incubated GST or GST-L2 bound to glutathione resin with lysates from HEK293 cells overexpressing either the wild-type SNX27a or SNX27b isoform (shown schematically in Fig. 1A) (SNX27a has a C-terminal stretch of 15 amino acids that replace 2 residues in isoform b [14]) or mutants of SNX27b lacking the PDZ domain (SNX27b ΔPDZ) or the FERM domain (SNX27b ΔFERM) (shown schematically in Fig. 7A). After washing, bound proteins were eluted in sample buffer and were separated by SDS-PAGE. After transfer to nitrocellulose membranes, interacting proteins were visualized by probing with anti-myc antibodies. The results (Fig. 7B) show that the interaction between HPV-16 L2 and SNX27 requires an intact PDZ domain. We note that the ΔFERM mutant runs considerably faster than would be predicted considering the number of residues removed, but this is accordance with previous observations (Martin P. Playford, personal observations). To confirm this finding, we used peptides derived from the NPAY region of HPV-16 L2 and a PDZ-binding motif (PBM) derived from the carboxyl terminus of HPV-18 E6 (Fig. 8A) to attempt to block the interaction between HPV-16 L2 and either SNX17 or SNX27. We preincubated GST or GST-L2 bound to glutathione agarose with the peptides and then added in vitro-translated, radiolabeled SNX17 or SNX27. The results (Fig. 8B) demonstrate that the PBM-derived peptide inhibits the interaction between HPV-16 L2 and SNX27, but not that between HPV-16 L2 and SNX17. Similarly, the NPAY motif-derived peptide inhibits the interaction between L2 and SNX17, but not that between L2 and SNX27. These results strongly suggest that the interaction of L2 with SNX27 is mediated by the SNX27 PDZ domain. To confirm this, we purified the isolated PDZ domain of SNX27 as a GST fusion protein, bound to glutathione agarose, and incubated it with a cell extract in which 16L2 was expressed. The results (Fig. 8C) show that this domain alone is capable of interacting with L2.

Since papillomavirus L2 proteins do not contain a typical carboxy-terminal PBM, we asked how L2 might be able to interact with the PDZ domain of SNX27. Based on X-ray crystallographic and nuclear magnetic resonance (NMR) structure predictions of the SNX27 PDZ domain (27), we made three large deletions in the SNX27 PDZ domain to investigate which residues might interact with 16L2. One unique feature of the SNX27 PDZ domain is a stretch of amino acid residues upstream of the core PDZ pocket that mediate binding between SNX27 and the VPS26 subunit of the retromer complex, thus allowing SNX27 to link to the complex while still permitting cargos with a PBM to bind to the SNX27 PDZ domain (28) and also increasing the affinity of the PBM-PDZ interaction (27). We speculated that the same stretch of amino acids might be responsible for the SNX27–L2 interaction and made a deletion mutant, Δ67–77, that lacks this region and that has been shown to be unable to interact with VPS26 (28). However, when we incubated 16L2 as a GST fusion protein with a cell extract in which this mutant protein was expressed, we observed binding at wild-type levels (Fig. 9). In contrast, two deletion mutants that remove residues deemed to be part of the PDZ domain based on structural analysis (27)—the Δ97–106 mutant, which perturbs a β-sheet, and the Δ113–121 mutant, which perturbs an α-helix—show greatly reduced ability to bind to GST L2 (Fig. 9), suggesting that it is residues within the PDZ domain itself that are involved in the interaction with L2.

Identification of the HPV-16 L2 amino acid residues involved in the interaction with SNX27.

The finding that HPV-16 L2 interacts with the PDZ domain of SNX27 was unexpected, since L2 lacks a classical carboxy-terminal class 1 PBM. Although internal PBMs have been reported, and some residues within L2 might be considered to fall into this consensus, it is not clear whether such putative PBMs lie in regions of L2 where the molecular conformation would position them optimally for interaction with a PDZ domain. However, we note that sequence analysis of several papillomavirus L2 proteins by a Eukaryotic Linear Motif (ELM) predictive program suggests that the C termini of some (though not all) L2 proteins constitute class 2 PBMs. To search for the region of L2 involved in the interaction with SNX27, we made a series of L2 truncation mutants by inserting stop codons into the L2 open reading frame to generate proteins that terminate after amino acid 192, 292, or 347 (Fig. 10A). Wild-type 16L2 and each of these mutants were first in vitro translated and radiolabeled and then incubated with either GST alone or GST-SNX27, both bound to glutathione agarose. After washing, bound proteins were eluted and analyzed by SDS-PAGE and autoradiography. The results are shown in Fig. 10B, where it can be seen that the L2 truncation construct terminating at amino acid 192 was unable to bind to SNX27 but that the remaining L2 truncation mutants bound as well as full-length L2. This also shows that sequences in the carboxy terminus of L2 are not directly involved in the interaction with SNX27.

Both siRNA knockdown of SNX17 and siRNA knockdown of SNX27 affect papillomavirus pseudovirion infection.

To study the role played by SNX27 in papillomavirus infection, we used a well-established pseudovirion infection system (29). We selectively downregulated SNX27 or SNX17 expression in HaCaT cells by siRNA treatment. Forty-eight hours later, we infected the cells with pseudovirions in which a luciferase reporter gene had been packaged and then monitored the infection efficiency by detection of luciferase activity in cell lysates 48 h postinfection. We used SfPV, HPV-31, HPV-18, HPV-16, HPV-5, and BPV-1 pseudovirions, with siSNX17 as a positive control and a scrambled siRNA as a negative control. The results are shown in Fig. 11, where it can be seen clearly that siRNA knockdown of SNX17 reduces infection efficiency, as demonstrated previously (8). In contrast, siRNA knockdown of SNX27 leads to only a marginal reduction in infection efficiency for most of the papillomavirus types studied, but siRNA knockdown of both SNX17 and SNX27 results in a further reduction of infection efficiency, greater than that observed for SNX17 knockdown alone.

To examine the statistical significance of this finding, we performed two-way analysis of variance (ANOVA) evaluating the effects of different siSNX treatments on the various PV types. The analysis shows that infection efficiency, as reflected by luciferase expression, is highly dependent on the siSNX treatment (P < 0.0001), while the differences observed between different PV types are not significant (P, 0.327 by two-way ANOVA). When we further compared different siRNA treatments, we observed significantly lower infectivity with siSNX17 (P < 0.0001) or the double knockdown (siSNX27 and siSNX17) (P < 0.0001)—but, as expected, not with siSNX27—than with siScr. Also as expected, comparisons between different siSNX treatments confirmed significantly lower infectivity for siSNX17-treated cells than for siSNX27-treated cells (P < 0.0001). More interestingly, double silencing of SNX27 and SNX17 is significantly more effective at eliminating general PV infection than the single siSNX27 (P < 0.0001) or even siSNX17 (P, <0.001 by two-way ANOVA and Tukey's multiple-comparison test) silencing. Although we typically observe an increase in infection efficiency when using the control scrambled siRNA over that obtained using no siRNA, this increase is considered to be nonspecific, and we observed no further changes in infection efficiency when excess amounts of siScr were added to the siRNA against either of the sorting nexins (data not shown). Consequently, for all siRNA transfections, the control siRNA, siScr, was used to balance inputs, as described in Materials and Methods. siRNA knockdown efficiency was verified by Western blot analysis of the cell lysates, probing sequentially for SNX27, SNX17, and α-actinin (used as a control). These results suggest the involvement of both SNX17 and SNX27 in controlling the entry of L2 and associated viral DNA into the nucleus.

16L2, when expressed ectopically together with SNX17, has a different cellular localization than when it is expressed by itself or together with SNX27.

Although the result presented above suggests a level of functional redundancy between SNX17 and SNX27 with regard to pseudovirion infection, we wanted to determine any effects of ectopic expression of 16L2 with either SNX17 or SNX27 on its cellular localization. Red fluorescent 16L2 was transfected into H1299 cells on coverslips together with either myc-tagged SNX27 or Flag-tagged SNX17. After 24 h, cells were fixed and permeabilized, and the relevant proteins were detected by incubation with either mouse anti-SNX27 antibodies or rabbit anti-Flag antibodies (for SNX17), followed by the relevant fluorescein-labeled secondary antibodies. The results (Fig. 12) show that after 24 h, L2 expressed alone is localized to the nucleus. Ectopic expression of SNX27 together with L2 does not change the localization of L2 at this time point, although some colocalization of L2 and SNX27 can be observed adjacent to the nucleus in regions reminiscent of the trans-Golgi network. In contrast, L2 is localized mostly to the cytoplasm in the presence of ectopically expressed SNX17.