The cellular endosomal protein stannin inhibits intracellular trafficking of human papillomavirus during virus entry

Abstract

Human papillomaviruses (HPVs) are the most common sexually transmitted viruses and one of the most important infectious causes of cancers worldwide. While prophylactic vaccines are effective against certain strains of HPV, established infections still cause deadly cancers in both men and women. HPV traffics to the nucleus via the retrograde transport pathway, but the mechanism of intracellular transport of non-enveloped viruses such as HPV is incompletely understood. Using an overexpression screen, we identify several genes that control HPV16 entry. We focused on the mechanism by which one of the screen hits, stannin, blocks HPV16 infection. Stannin has not been previously implicated in virus entry. Overexpression of stannin specifically inhibits infection by several HPV types, but not other viruses tested. Stannin is constitutively expressed in human keratinocytes, and its basal levels limit entry by HPV16. Stannin is localized to the endolysosomal compartment and does not affect HPV16 binding to cells, virus uptake, or virus uncoating, but inhibits the entry of HPV into the trans-Golgi network (TGN) and stimulates HPV degradation. We further show that stannin interacts with L1 major capsid protein and impairs the interaction of the L2 minor capsid protein with retromer, which is required for virus trafficking to the TGN. Our findings shed light on a novel cellular protein that interferes with HPV entry and highlight the role of retrograde transport in HPV entry.

Introduction

Human papillomaviruses (HPVs) are small DNA tumour viruses responsible for 5 % of all cancers worldwide [1]. Among them, HPV16 is linked to 50 % of all cervical cancers and a large fraction of anogenital and oropharyngeal malignancies [2]. HPVs are non-enveloped viruses consisting of a double-stranded DNA viral genome encapsidated by the major capsid protein L1, which forms the outer rigid shell of the virion, and the minor capsid protein L2, which plays an essential role in viral intracellular trafficking events during virus entry. HPVs exclusively infect basal keratinocytes of the skin and mucosal epithelium. Entry is initiated by virus binding to plasma membrane heparan sulfate proteoglycans (HSPGs) [3]. This triggers conformational changes in the HPV capsid, followed by proteolytic cleavage of L1 and L2 by kallikrein-8 and furin, respectively [4, 5], transfer to an elusive secondary uptake receptor [6], and endocytosis via a poorly understood mechanism [7]. Capsid disassembly begins in the endosome followed by vesicular retrograde transport of the viral L1/L2/DNA subcomplex to the trans-Golgi network (TGN), Golgi apparatus and endoplasmic reticulum (ER) [8–11]. A role for the multi-vesicular endosomes in HPV16 trafficking has also been reported [7, 12]. Productive HPV16 infection requires cell cycle progression and nuclear envelope breakdown during mitosis to allow viral DNA to exit the vesicular compartment and enter the nucleus [13–17]. Viral trafficking can be a rapid process if the target cells are infected shortly prior to mitosis [18]. However, since mitotic entry of most cells in culture is not synchronized, HPV is typically internalized over a period of many hours with endosomal, TGN and nuclear localization most apparent after approximately 8, 16 and 24 h, respectively [10, 17]. Viral gene expression is tightly linked to differentiation of host epithelial cells, and viral assembly does not take place until shortly prior to keratinocyte shedding from the uppermost layer of the infected epithelium [19].

Viral cell entry and intracellular trafficking depends on the coordinated activities of numerous cellular proteins [20]. Therefore, identification and characterization of cellular factors that facilitate or inhibit infection provide mechanistic insight into virus entry and the cell biology underlying this complex process. Most studies of the initial stages of HPV infection have utilized HPV pseudoviruses (PsV), which contain L1 and L2 and are structurally similar to authentic virions. HPV PsV do not incorporate the authentic HPV genome [21], but instead contain a reporter gene, which allows ready assessment of infection. Use of HPV PsV also allows the construction and analysis of viral mutants.

Several cellular retrograde trafficking factors and other cellular proteins required for entry have been shown to associate with the L2 protein before the viral DNA reaches the nucleus [22–25]. An important entry factor is the retromer, a cytoplasmic protein complex that typically binds to the cytoplasmic tail of cellular transmembrane protein cargoes at the endosomal membrane and recruits them into transport vesicles that later fuse with the membranes of the TGN [8, 26]. The retromer binds directly to a highly conserved sequence motif in the C-terminus of the L2 protein, and retromer knock-down or mutation of the retromer-binding sites in L2 results in accumulation of the virus in the endosome and inhibits transport to the TGN [22]. Interestingly, the retromer and several other L2-binding partners (e.g. SNX17) are located in the cytoplasm, whereas the incoming virus itself is in the lumen of the endosome. These results suggest that a portion of the L2 protein protrudes through the endosomal membrane into the cytoplasm where it can bind the retromer and other cytoplasmic factors [8, 23, 27]. Recent antibody and protease sensitivity studies showed that much of L2 is in fact accessible at the cytoplasmic face of the endosomal membrane prior to transport to the TGN, apparently anchored into the membrane by a putative N-terminal transmembrane domain in L2 [27, 28]. This membrane penetration event may be mediated by a membrane-destabilizing sequence in the extreme C-terminus of the L2 protein near the retromer-binding sites [29].

To identify additional cellular factors involved in HPV infection, we conducted an overexpression genetic screen using a library of IFN-stimulated genes (ISGs) [30]. IFNs are potent antiviral cytokines that block virus infection and replication by inducing hundreds of ISGs with direct antiviral activity [31]. Treatment of human keratinocytes with exogenous IFNs inhibits HPV16 infection [32–34]. Our screen identified a unique set of genes that affect HPV16 PsV infection, most of which have not been previously implicated in infection by viruses. One of the top hits in the screen, stannin, inhibits infection by several HPV PsV types but does not appear to be IFN-inducible in our system. We discovered that stannin binds to the HPV capsid during infection and inhibits proper intracellular trafficking of HPV16 to the TGN. Our report identifies a cellular protein that can inhibit infection by HPVs and provides additional insight into the molecular machinery and processes that control retrograde trafficking of HPV during virus entry.

Results

Genetic overexpression screen identifies a unique set of genes involved in inhibition of HPV16 infection

To identify host genes that control HPV16 entry, we used HeLa cervical carcinoma cells, which are amenable to lentiviral transduction and are susceptible to HPV16 infection. The cells were screened for HPV16 PsV reporter gene expression as described previously using a high-throughput flow cytometry-based platform [30]. HeLa cells were seeded into 96-well plates and transduced in a one gene per well format with an arrayed lentiviral cDNA library expressing 387 ISGs. Since the lentiviral vector co-expresses a bicistronic ISG-IRES-TagTagRFP cassette, efficiency of transduction was monitored by RFP expression. Transduced cells were infected with HPV16 PsV incorporating a GFP reporter plasmid (HPV16-GFP), produced as described previously [21]. Two days after infection, the cells were assayed by flow cytometry for RFP and GFP fluorescence (Fig. 1a). Our screens achieved an average lentivirus transduction efficiency of >50 %. Poor transduction (<10 % RFP + cells) was observed for 40 genes and only four ISGs were found to be cytotoxic (resulting in <50 % of the control cell number) (Table S1, available in the online Supplementary Material). The majority of the ISGs had a minor effect on HPV16 PsV entry as assessed by GFP expression (−1<Z<1). However, a handful of ISGs were found to strongly inhibit (Z<–2) or potentiate (Z>2) PsV infection (Fig. 1b). There was excellent reproducibility between two independent screen replicates for both the top inhibitors and the top enhancers of HPV16 PsV infection (Fig. 1b).

Fig. 1.

ISG screen identifies human genes that restrict HPV16 entry. (a) Schematic diagram of the flow cytometry-based overexpression screen. Top two panels show 96-well plates of HeLa cells sequentially infected with ISG lentiviruses expressing RFP and HPV16 pseudovirus expressing GFP. Hypothetical flow cytometry plots of ISGs that inhibit or enhance HPV16-GFP PsV infection are shown at the bottom. GFP⁺TagRFP⁺ cells are depicted as grey. (b) Distribution of Z score values for the two replicates of the ISG screen. Each spot represents a single ISG. Top hits are highlighted with red (inhibitors) or green (potentiators). Table lists top hits with Z score for both replicates.

Top screen hits specifically control infection of multiple HPV types

None of the top hits identified in the genetic screen were previously implicated in HPV16 infection. We validated a subset of genes with the strongest inhibitory activity, including SNN (stannin), THBD (thrombomodulin), SERPINE1 (serpin E1) and VMP1 (vacuole membrane protein 1). Other than SERPINE1, previously reported to have activity against influenza virus [35], none of the other top hits have known antiviral activities. To ensure that the screen findings were not due to a cell line-specific artifact, we validated the top inhibitory hits in HaCaT cells, a human skin keratinocyte cell line. Lentiviral transduction of HaCaT cells led to target gene overexpression at the mRNA level (Fig. S1). HPV16-GFP infection was inhibited by overexpression of SNN, THBD and VMP1, but not SERPINE1 in HaCaT cells (Fig. 2a). In contrast, infection with adenovirus, an unrelated small, non-enveloped double-stranded DNA virus, was not significantly affected by any of the tested genes (Fig. 2b). To evaluate the ability of SNN, THBD and VMP1 to inhibit entry by other oncogenic HPV types, we infected the corresponding overexpressing cell lines with HPV5 and HPV18 PsVs. HPV5 and HPV18 are linked to skin and cervical cancer, respectively [1, 36]. SNN, THBD and VMP1 overexpression inhibited HPV5-GFP and HPV18-GFP infection to a similar extent as HPV16-GFP (Fig. 2b). In addition, SNN overexpression did not inhibit infection by JC polyomavirus, herpes simplex virus type 1 (HSV1), or adeno-associated virus type 2 (AAV2) (Fig. 2c). Inhibition of infection by multiple HPV types, but not other tested viruses, suggests that these genes specifically inhibit HPV infection, as opposed to disrupting essential cellular processes or acting as pan-antiviral factors.

Fig. 2.

Validation of top inhibitory ISG screen hits in HaCaT cells. (a, b) Unmodified HaCaT cells (no ISG) or cells stably overexpressing Luciferase control, SERPINE1, SNN, THBD or VMP1 were infected with HPV5-GFP [3×10⁴ viral genome equivalents (vge) cell⁻¹], HPV16-GFP (0.5−1×10³ vge cell⁻¹), HPV18-GFP (3×10³ vge cell⁻¹) or adenovirus5-GFP (2×10² vge cell⁻¹). GFP expression was assayed by flow cytometry 48 h (HPV) or 36 h (adenovirus) later. (a) Representative flow cytometry plots of cells infected with HPV16-GFP. (b) Infection efficiency of adenovirus, HPV5-GFP, HPV16-GFP and HPV18-GFP in cells transduced with the indicated ISG, normalized to infection in control cells expressing luciferase (luc). (c) HaCaT cells stably overexpressing Luciferase (black bars) or SNN (grey bars) were infected with HSV1-GFP (m.o.i.=0.25), JCV-GFP (5×10⁴ vge cell⁻¹) or AAV2-GFP (5×10⁴ vge cell⁻¹), and GFP expression assayed by flow cytometry after 24 h (HSV1 and AAV2) or 48 h (JC). (d) The SNN genomic locus in HeLa cells was edited with the CRISPR Cas9 system as described in Methods. Control unedited cells and three clones of SNN knockout cells were incubated with 50–100 vge cell⁻¹ of HPV16-GFP PsV, 50 vge cell⁻¹ of adenovirus5-GFP, or HSV1-GFP and assayed for SNN mRNA expression (blue bars) and for GFP expression by flow cytometry 48 h (HPV), 36 h (adenovirus), or 24 h (HSV1) later. For (b–d), results show the mean and sd from three independent experiments. Where indicated, statistical significance was determined by ANOVA (b) or an unpaired two-tailed t-test (c, d): *P<0.05; **P<0.01. ns, not significant.

Genetic deletion of stannin promotes HPV16 infection

Stannin, an 88-residue transmembrane protein encoded by the SNN gene, was one of the strongest inhibitors identified, leading to an approximately three and fivefold inhibition of HPV16-GFP infection in HaCaT and HeLa cells, respectively. We therefore focused on studying the role played by stannin during HPV16 entry. We first determined the effect of basal SNN expression on HPV16 infection efficiency by using CRISPR-Cas9 genome editing. HeLa cells were transduced with three lentiviral vectors each encoding Cas9 as well as a guide RNA specific for a unique sequence within the SNN genomic locus (see Methods). We were unable to confirm SNN mutagenesis by Western blotting because endogenous stannin protein levels could not be detected with the available antibodies. Therefore, we screened clonal cell lines for the presence of deletions within the SNN locus by PCR and identified three independent clonal cell lines lacking full-length SNN (Fig. S2). All three cell lines expressed reduced levels of SNN mRNA (Fig. 2d). Notably, HPV16-GFP PsV infection efficiency was increased by more than 50 % in all three SNN mutant cell lines (Fig. 2d). In contrast, infection by adenovirus and herpes simplex virus was not significantly affected by SNN knockout. In one of these cell lines, we showed by DNA sequencing that all SNN alleles contain a 69-base pair deletion that removed more than one-quarter of the stannin-coding region, combined in some cases with smaller deletions (Fig. S2). These deletions in SNN are predicted to produce short aberrant protein product(s) lacking much or all of the stannin linker region, which we show below is essential for antiviral activity. There results indicate that basal levels of stannin inhibit HPV16-GFP infection in HeLa cells.

Stannin is not IFN-inducible in HaCaT cells

Stannin is a highly evolutionarily conserved protein with no homology to known proteins and with poorly understood biological function [37]. Stannin expression in neurons is associated with susceptibility to neurotoxicity induced by trimethyltin [38]. Stannin mRNA and protein are expressed in a wide variety of human tissues, including the squamous epithelial cells of the cervical mucosa [39, 40] – a major in vivo target of HPV16 infection. We detected endogenous SNN mRNA in all cell types examined, including human primary keratinocytes and peripheral blood mononuclear cells (Fig. S3a). However, treatment of HaCaT cells with recombinant IFN-β, IFN-κ or IFN-γ did not upregulate SNN mRNA expression, despite inducing a robust increase in the mRNA levels for bona fide ISGs such as Mx1 (Fig. S3b, and data not shown). Thus, SNN does not appear to be an ISG in these cells. Therefore, SNN overexpression interferes with HPV entry in a manner independent of IFNs.

Stannin control of HPV16 entry does not rely on autophagy, cell cycle perturbation or inhibition of furin-mediated L2 cleavage

We examined whether stannin-mediated inhibition of HPV required autophagy, an antiviral defence mechanism reported to control HPV16 entry [41, 42]. To investigate this, we generated ATG5−/− HeLa cells using the CRISPR Cas9 system. ATG5 is required for the execution of autophagy [43]. Wild-type and ATG5−/− HeLa cells were stably transduced with lentiviruses encoding Luciferase or SNN cDNAs and then infected with HPV16-GFP. HPV16 PsV infection efficiency increased by 50 % in ATG5−/− cells compared to unmodified wild-type cells, as expected, providing evidence that ATG5 knockout impaired autophagy. Overexpression of SNN in wild-type and ATG5−/− cells led to similar inhibition of HPV16-GFP infection efficiency (Fig. S4a). These results indicated that autophagy is not required for the SNN-mediated inhibition of HPV16 PsV infection.

A previous study of stannin showed that it could regulate cell cycle progression [44]. Furthermore, HPV infection requires progression of the cell cycle [13]. We therefore investigated the possibility of a link between stannin and the cell cycle by examining the DNA content of cells stably overexpressing Luciferase or SNN and cultured in either complete or serum-free growth medium. When cells were incubated in complete growth medium, stained with the DNA-binding dye DAPI and analysed by flow cytometry, stannin did not affect the proportion of cells in the G1, S and G2 phases of the cell cycle nor generate a population of cells with sub-G1 DNA content, indicative of apoptosis (Fig. S4b). In addition, SNN overexpression did not prevent G1 growth arrest induced by incubation of cells in serum-free medium. Thus, the anti-HPV effect of stannin does not depend on stannin-mediated regulation of the cell cycle.

HPV16 infection requires furin, a cellular proprotein convertase that cleaves the N-terminus of the HPV16 L2 protein at the cell surface [5]. In the absence of furin action, HPV16 virions can enter cells and disassemble, but their escape from the endosomal compartment is impaired. To determine if SNN overexpression inhibited HPV16 L2 cleavage by furin, we used a recombinant HPV16 PsV tagged with the 70-residue Propionibacterium shermanii transcarboxylase protein domain (PSTCD) at the N-terminus of L2 (HPV16.L2-PSTCD) upstream of the furin cleavage site [45]. Cleavage of L2-PSTCD by furin at the consensus site releases the PSTCD, allowing the full-length and cleaved forms of L2 to be resolved by SDS-polyacrylamide gel electrophoresis. HeLa cells stably overexpressing Luciferase or SNN were infected with HPV16.L2-PSTCD PsV. Lysates prepared 8 h after infection were subjected to Western blotting with an antibody (K4_20-38) specific for an L2 epitope downstream of the furin cleavage site. In agreement with previous reports, the furin-cleaved form of L2 was shorter and less abundant compared to the full-length protein and was absent in infected cells pre-treated with a furin inhibitor (Fig. S4c). The quantity of cleaved L2 was similar in control and SNN-overexpressing cells, demonstrating that stannin did not inhibit furin-mediated cleavage of L2.

Stannin and viral DNA co-localize to the endolysosomal compartment

We next determined where stannin localized in the cell. In order to detect stannin, a FLAG epitope tag sequence was cloned in-frame into the SNN cDNA at either the 5′ or 3′ position of the SNN open reading frame. Stable expression of 5′FLAG-SNN or 3′FLAG-SNN in HeLa cells inhibited HPV16-GFP infection to a similar extent as untagged stannin, but had no impact on adenovirus infection (Fig. S5). HeLa cells stably overexpressing C-terminally tagged stannin were stained with a FLAG-specific antibody and visualized by confocal immunofluorescence microscopy. We observed significant co-localization between stannin and the early endosome marker EEA1, the late endosome marker Rab9, the endolysosomal marker LAMP1 and the TGN marker TGN46, but little co-localization with the mitochondrial marker COX IV or the ER marker BiP (Fig. 3a, b). HPV16 PsV infection did not affect stannin co-localization with these markers. These results indicate that stannin is broadly distributed in a variety of intracellular organelles in the retrograde transport pathway whether or not the cells were infected with HPV16 PsV.

Fig. 3.

Stannin co-localizes with the endolysosomal compartment and incoming HPV16 genomes. (a) HeLa cells stably overexpressing FLAG-tagged stannin were seeded onto coverslips, fixed, and immunostained with antibodies against FLAG (stannin, green) and EEA1 (red). Nuclei were stained with DAPI (blue). Co-localization of FLAG and EEA1 appears yellow in the bottom panels. The boxed area in the merged image is displayed separately at a higher magnification at the lower right. The same confocal slice is shown in each panel. Scale bar: 10 µm. (b) HeLa cells stably overexpressing FLAG-tagged stannin were seeded onto coverslips and either remained untreated (control, black bars) or incubated with HPV16 PsV (3×10⁴ vge cell^–1, grey bars). Twenty four h later cells were fixed and immunostained with antibodies against FLAG (to detect stannin) and against EEA1, Rab9, LAMP1, TGN46, BiP or COX IV. The bar graph shows the extent of overlap between FLAG staining and that of the various subcellular markers in control and infected cells, analysed with the JACoP plugin for ImageJ. For each analysis, the mean and sd were calculated from more than 100 cells and are representative of two independent experiments. (c) HeLa cells stably overexpressing FLAG-tagged stannin were seeded onto coverslips and incubated with 5-ethynyl-2′-deoxyuridine (EdU)-labelled HPV16 PsV (3×10⁴ vge cell⁻¹). After 8 h the cells were fixed, processed with the Click-iT imaging kit to stain the viral pseudogenomes (EdU, red), and then immunostained with a FLAG antibody (stannin, green). Nuclei were stained with DAPI (blue). Co-localization of FLAG and EdU appears yellow in the bottom panels. The boxed area in the merged image is displayed separately at a higher magnification at the lower right. The same confocal slice is shown in each panel. Scale bar: 10 µm. (d) HeLa cells infected and processed as described in (c) 8 or 24 h after infection. The bar graph shows the extent of co-localization between EdU and stannin staining analysed with the JACoP plugin for ImageJ. For each analysis, the mean and sd were calculated from more than 100 cells and are representative of three independent experiments. For (a) and (c), images are representative of three or more independent experiments.

To examine whether HPV16 trafficking involves passage through cellular compartments containing stannin, we generated HPV16 PsV incorporating plasmid DNA labelled with the nucleotide analogue 5-ethynyl-2′-deoxyuridine (EdU) [46]. Unlabelled HeLa cells stably overexpressing FLAG-tagged stannin were infected with EdU-labelled HPV16 PsV for 8 or 24 h and stained with the Click-iT EdU Imaging reaction cocktail to visualize the pseudovirus DNA and with anti-FLAG to detect stannin. Confocal immunofluorescence microscopy revealed significant overlap between pseudovirus DNA and stannin (Fig. 3c, d). Collectively, these results indicate that stannin co-localizes with HPV16 PsV DNA in a range of retrograde compartments during intracellular trafficking during entry.

Identification of stannin domains required for inhibition of HPV16 entry

Stannin is comprised of N-terminal transmembrane, linker, and C-terminal domains. We performed deletion analysis of stannin to examine the contribution of these domains to its anti-HPV function (Fig. 4a). For the stannin(1–59) mutant, we deleted amino acids 60–88 comprising the cytoplasmic C-terminal domain. The stannin(1–36) mutant contained a larger deletion (removing amino acids 37–88) spanning both the linker and the C-terminal domains. Rather than deleting the N-terminal transmembrane domain of stannin, which would most likely abolish the membrane localization of the protein, we instead replaced it with the transmembrane domain of TLR9 [stannin(TLR9TM) mutant], which targets TLR9 to the ER [47]. Each mutant was further engineered to contain a HA tag at the N-terminus to facilitate intracellular detection. The HA tag did not impair the ability of full-length stannin to inhibit HPV16 PsV infection (Fig. 4b).

Fig. 4.

Mapping domains of stannin required for antiviral activity. (a) Schematic diagrams of domain composition of full-length (FL) and mutant stannin. Stannin(1–36) and stannin(1–59) mutants comprise the first 36 and 59 amino acids of stannin, respectively. The stannin(TLR9TM) mutant is a chimera of FL stannin and 20 amino acids from the TLR9 transmembrane (TLR9TM) domain. (b–d) HeLa cells were transduced with lentiviruses encoding Luciferase (control), HA-tagged full-length (FL) stannin, or the different HA-tagged stannin mutants. (b) Cells transduced with the indicated gene were infected with HPV16-GFP PsV (100–200 vge cell⁻¹) for 48 h, fixed and assayed by flow cytometry for GFP expression. Infection efficiency was calculated as described in Methods. Results show the mean and sd from three independent experiments, normalized to cells lacking transduced stannin. (c) Transduced cells were seeded onto glass coverslips, fixed and immunostained with antibodies against HA (stannin, green) and EEA1 or LAMP1 (both red). Nuclei were stained with DAPI (blue). Co-localization of stannin and EEA1 appears yellow in the right column in each set. The same confocal slice is shown in each row of each set. Scale bar: 10 µm. (d) Transduced cells were processed for immunofluorescence microscopy as described in (c) and images analysed for co-localization between stannin and either EEA1 (black bars) or LAMP1 (grey bars) with the JACoP plugin for ImageJ. Results show the mean and sd from two independent experiments with at least 100 cells analysed for each co-localization, normalized to cells expressing full-length stannin. Where indicated, statistical significance was determined by an unpaired two-tailed t-test: **P<0.01. ns, not significant.

Following overexpression in HeLa cells, we found that the stannin(1-59) mutant, like full-length (FL) stannin, localized to both EEA1- and LAMP1-positive compartments (Fig. 4c, d), and retained the antiviral activity of the wild-type stannin (Fig. 4b). In contrast, stannin(1-36) co-localized almost exclusively with LAMP1-positive structures and failed to inhibit HPV16-GFP PsV infection. The chimeric stannin(TLR9TM) protein was not found in endosomes, and instead formed a reticular network suggestive of the ER, as expected (Fig. 4b). This chimera also failed to inhibit HPV16 entry. Taken together, these data suggest that the C-terminal domain of stannin is required neither for stannin localization to the early endosome nor for inhibition of HPV16 infection. In contrast, deletion of both the C-terminal and the linker domains mislocalized stannin(1–36) to LAMP1-positive compartments and rendered the protein incapable of blocking viral entry. Moreover, the intact linker and C-terminus of stannin, when ectopically localized to the ER, did not inhibit HPV. Thus, the presence of the N-terminal transmembrane domain and linker domain is required for proper SNN localization and anti-HPV activity.

Stannin overexpression inhibits HPV16 transport to the TGN

The localization of stannin in the endosomal compartment suggested that SNN overexpression might inhibit HPV16 entry by perturbing virus uptake into, uncoating in or escape from the endosome. To examine HPV16 attachment to the plasma membrane, HaCaT cells stably overexpressing Luciferase or SNN were incubated with HPV16 PsV at 4 °C to allow virus binding but not internalization. After 1 h, unbound virus was washed away, and cells were harvested for analysis by Western blotting with an antibody recognizing HPV16 L1. The amount of cell surface-bound L1 was similar for the two cell lines (Fig. 5a), indicating that stannin does not inhibit viral attachment to target cells.

Fig. 5.

Stannin overexpression does not affect virus binding or internalization but inhibits HPV16 transport to the TGN. (a) HaCaT cells stably overexpressing Luciferase (luc) or SNN were mock-infected or incubated with HPV16 PsV (1×10⁴ vge cell⁻¹) for 1 h at 4 °C. Unbound virus was washed away, cells were harvested in NP40 lysis buffer and lysates were analysed by SDS-PAGE and immunoblotting with an actin or an L1 antibody. Non-specific bands are indicated with an asterisk. (b) HaCaT cells stably overexpressing Luciferase (control) or SNN were incubated with HPV16 PsV (1×10³ vge cell⁻¹) for 8 h at 37 °C or remained untreated. Cells were trypsinized to remove surface-bound PsV, immunostained with the 33L1-7 antibody, and assayed by flow cytometry. Uninfected wild-type HaCaT cells (no ISG) were included in the experiment to serve as a TagRFP-negative gating control. The percentage of 33L1-7⁺ cells is indicated in the upper right quadrant. (c) HeLa cells stably overexpressing Luciferase (Luc) or SNN were incubated with EdU-labelled HPV16 PsV (3×10⁴ vge cell⁻¹). After 20 h the cells were fixed, processed with the Click-iT imaging kit to visualize viral pseudogenomes (EdU, green) and then immunostained with a TGN antibody (p230, red). Nuclei were stained with DAPI (blue). Co-localization of p230 and EdU appears yellow in the panels labelled merged. Boxed areas in the merged images are displayed separately in the right-most panels at a higher magnification. The same confocal slice is shown in each row. Scale bar: 10 µm. (d) HeLa cells were infected and processed as described in (c). Co-localization between EdU and p230 was analysed with the JACoP plugin for ImageJ. The mean and sd were calculated from three independent experiments with at least 100 cells counted per experiment. Statistical significance was determined by an unpaired two-tailed t-test: *P<0.05. (e) HeLa cells stably overexpressing Luciferase were mock-infected and cells overexpressing Luciferase or SNN were infected HPV16.L2F PsV (3×10⁴ vge cell⁻¹) for 16 h. Cells were processed for proximity ligation assay (PLA) with antibodies recognizing the FLAG epitope on L2 and TGN46. Nuclei were stained with DAPI (blue). Top panels show PLA signal (green); bottom panels show merged DAPI and PLA signals. Images are representative of three independent experiments. (f) HeLa cells stably overexpressing Luciferase or SNN were infected and processed as described in (e). The bar graph shows the mean and sd of the total PLA signal from three independent experiments, normalized to infected control cells. Statistical significance was determined by an unpaired two-tailed t-test: **P<0.01.

To study virus uptake and the initiation of uncoating, HaCaT cells were incubated with HPV16 PsV for 8 h at 37 °C, trypsinized to remove cell surface-bound virions and stained with an antibody (33L1-7) that recognizes a linear epitope within L1 that is inaccessible in intact virions but exposed after internalization and uncoating [48]. 33L1-7 antibody binding as assessed by flow cytometry was similar in the presence and absence of overexpressed stannin (Fig. 5b), indicating that virus internalization occurs in SNN-overexpressing cells and that the virus initiated uncoating in these cells.

To address whether stannin might inhibit HPV16 trafficking to the TGN, we used EdU-labelled HPV16 PsV to track pseudovirus DNA within infected cells. HeLa cells stably overexpressing Luciferase or SNN were incubated with EdU-labelled HPV16 PsV, and cells were fixed and stained at various times post-infection (p.i.). We observed an approximately threefold reduction in co-localization between EdU and p230 (a TGN marker) in cells overexpressing SNN compared to the Luciferase control (Fig. 5c, d), suggesting that stannin inhibited HPV trafficking to the TGN.

We used the in situ proximity ligation assay (PLA) to confirm that stannin affected the arrival of L2 at the TGN. PLA is a highly sensitive method to quantitate protein co-localization that utilizes two antibodies conjugated to short oligonucleotides [49]. If the proteins of interest are within 40 nm of each other, the oligonucleotides can be ligated, the ligation product amplified and the amplified DNA labelled with fluorescent probes [50]. For this experiment, we infected control and SNN-overexpressing HeLa cells with HPV16 PsV containing FLAG-tagged L2 (HPV16.L2F), and after 16 h performed PLA using anti-FLAG antibody (to recognize L2) and an antibody recognizing TGN46, a marker of the TGN. We used HPV16 PsV incorporating a 3xFLAG at the C-terminus of L2 (HPV16.L2F). The 3xFLAG tag is constitutively exposed on the surface of the HPV16.L2F capsids and does not require virus uncoating for detection [10]. As shown in Fig. 5(e, f), there is an approximately twofold reduction in the PLA signal in the SNN cells, providing further evidence that SNN overexpression in the endolysosomal compartment impairs the delivery of incoming virions into the TGN.

Stannin associates with the HPV16 capsid during infection and inhibits retromer recruitment

To test whether stannin physically interacts with incoming HPV16 pseudovirions, we transduced HeLa cells with SNN encoding a C-terminal HA tag (3′HA-SNN). Cells stably overexpressing Luciferase or 3′HA-SNN were infected for 8 h with HPV16.L2F PsV (containing L1 and FLAG-tagged L2) or with HPV16.L1 PsV, which contains only the L1 protein without L2. We lysed the infected cells with a high (5 mg ml⁻¹) concentration of digitonin, a non-ionic detergent which efficiently solubilizes membrane protein complexes [51]. Lysates were immunoprecipitated with a HA antibody and assayed by Western blotting with antibodies recognizing L1 and the FLAG tag (to detect L2). No immunoprecipitation of L1 or L2 occurred from cells lacking HA-SNN expression. In contrast, both L1 and L2 reproducibly co-precipitated with stannin early after infection with HPV16.L2F (Fig. 6a), demonstrating complex formation between the PsV particle and stannin. We also observed co-immunoprecipitation between stannin and L1 and L2 in extracts prepared in buffers containing the non-ionic detergent Nonidet-P40 and the chemical crosslinker dithiobis(succinimidyl propionate) (data not shown). Notably, L1 was also precipitated, albeit at a lower level, from cells infected with PsV containing only L1 (Fig. 6a), showing that L1 was sufficient for stannin binding. We conclude that L2 as well as L1 is co-immunoprecipitated with PsV particles containing both viral proteins because they are tightly associated in the complete PsV particle. These results indicate that stannin stably associates with the incoming viral particle via an interaction with L1, although we have not ruled out the possibility that the interaction between L1 and stannin is indirect. The reduced association observed with the L1-only particles suggested that L2 modulates this interaction.

Fig. 6.

HPV16 associates with stannin. (a) HeLa cells stably overexpressing Luciferase or HA-tagged stannin were mock-infected or infected with HPV16.L2F (L1L2) or with HPV16 L1-only (L1) PsV (1×10⁴ vge cell⁻¹). Eight hours later, cells were trypsinized to remove surface-bound virions and lysed in buffer containing 0.5 % digitonin. Lysates were precipitated with a HA antibody-bead conjugate, and immunoprecipitates were analysed by SDS-PAGE and Western blotting with antibodies specific for actin, HA (stannin), HPV16 L1 and FLAG (to visualize L2). Input lanes show 5 % of each lysate prior to immunoprecipitation (IP). For each of the indicated antibodies, images for the input and IP sets of samples were derived from the same blot. (b) HeLa cells stably overexpressing Luciferase or SNN were mock-infected or infected with HPV16.L2F (6×10³ vge cell⁻¹). Eight hours later, cells were fixed, reacted with antibodies against FLAG (L2) and VPS35 (retromer) and processed for the proximity ligation assay (green). Nuclei were stained with DAPI (blue). A single confocal slice is shown in each panel. Images are representative of three or more independent experiments. (c) HeLa cells stably overexpressing Luciferase or SNN were infected and processed as described in (b). The bar graph shows the mean and sd of the total PLA signal from three independent experiments, normalized to infected control cells. Statistical significance was determined by an unpaired two-tailed t-test: **P<0.01.

Retromer is a protein complex that binds directly to the L2 protein and plays a key role in HPV16 escape from the endosome to the Golgi [8, 22]. We hypothesized that decreased trafficking of HPV16 to the TGN in cells stably overexpressing SNN might be associated with decreased recruitment of L2 to the retromer. We used PLA to assess if L2 is in close proximity to the retromer subunit VPS35 following infection with HPV16.L2F [22]. Notably, SNN overexpression led to an approximately fourfold decrease in the PLA signal between L2 and VPS35 (Fig. 6b, c). The dramatic reduction in signal indicates that stannin overexpression inhibited the interaction between L2 and the retromer.

Stannin overexpression does not disrupt endolysosomal morphology or retrograde transport of other cargo proteins

Stannin overexpression could block HPV trafficking by changing endolysosomal maturation or retrograde transport in general. First, we examined the morphology and intracellular distribution of endolysosomes in SNN-overexpressing cells. Fig. S6(a) shows that the morphology and subcellular localization of EEA1- and LAMP1-positive compartments was not noticeably altered by SNN overexpression in HeLa cells. Moreover, SNN overexpression did not change the extent of co-localization between EEA1 and LAMP1 (Fig. S6b). Similarly, SNN overexpression had no noticeable effect on the size, morphology or intracellular distribution of the Golgi apparatus or TGN, as assessed by staining for the markers GM130 and p230 (Fig. 7a, b).

Fig. 7.

SNN overexpression does not inhibit Golgi trafficking of cholera toxin and CD8-CIMPR. (a) HeLa cells stably overexpressing Luciferase (Luc) or SNN were seeded onto coverslips and incubated for 1 h with AF488-conjugated cholera toxin B subunit (CTXB, green). Cells were then fixed and immunostained with a p230 antibody (red). Nuclei were stained with DAPI (blue). Co-localization of CTXB and p230 appears yellow in the panels labelled merged. The boxed area in the merged image is displayed separately at a higher magnification in the far-right panels. The same confocal slice is shown in each row. Scale bar: 10 µm. Images are representative of three independent experiments. (b) Co-localization analysis of CTXB and p230 staining calculated with the JACoP plugin for ImageJ. The mean and sd were calculated from more than 100 cells and are representative of three independent experiments. The differences were not statistically significant (ns) as determined by an unpaired two-tailed t-test. (c) HeLa cells stably overexpressing Luciferase (Luc) or SNN were seeded onto coverslips and transfected with a vector encoding the CD8-CIMPR chimeric protein. Forty eight hours later, the cells were incubated for 3 h with anti-CD8 antibody (green) at 37 °C, fixed and immunostained with a GM130 antibody (red). Nuclei were stained with DAPI (blue). Co-localization of CD8 and GM130 appears yellow in panels labelled merged. The boxed area in the merged image is displayed separately at a higher magnification in the far-right panels. The same confocal slice is shown in each panel. Scale bar: 10 µm. Images are representative of three independent experiments. (d) Co-localization analysis of CD8 and GM130 staining calculated with the JACoP plugin for ImageJ. The mean and sd were calculated from more than 100 cells and are representative of three independent experiments. The differences were not statistically significant (ns) as determined by an unpaired two-tailed t-test.

Further, we examined intracellular transport of two well-established model cargo proteins that reach the TGN through the retrograde pathway, the cholera toxin B subunit (CTXB) [52] and cation-independent mannose phosphate receptor (CIMPR) [53]. Stannin overexpression had no impact on the co-localization of these cargo proteins and TGN markers (Fig. 7). Taken together, these results indicate that stannin overexpression does not perturb cellular endosomal trafficking pathways of these two retromer cargoes tested.

Stannin promotes HPV16 L1 and L2 degradation

HPV that cannot escape the endosomal compartment may ultimately be sorted to the lysosome where it is degraded [23]. We therefore reasoned that a block in endosome-to-TGN viral trafficking imposed by SNN overexpression might lead to increased degradation of incoming HPV16. To test this, we infected control and SNN-overexpressing HeLa cells with HPV16 PsV at 37 °C for 16 or 24 h. Cells were then trypsinized to remove cell surface-bound virions, and cell lysates were analysed by Western blotting for L1 and L2. At 16 and 24 h p.i., SNN overexpression led to an increase in the intensity of a low molecular weight band corresponding to an L1 cleavage product (Fig. 8a, b). The appearance of this L1 cleavage product was abolished by treatment of infected cells with chloroquine, an inhibitor of lysosome acidification (Fig. S7). SNN overexpression also led to an approximately twofold decrease in the levels of full-length L1 and L2 proteins at 16 and 24 h p.i. (Figs 8a, b and S7). To determine whether the observed reduction in L1 and L2 protein levels was accompanied by degradation of pseudovirus reporter plasmid DNA, we measured the quantity of plasmid DNA in infected cells 24 h p.i. Total DNA was extracted from infected cells and assayed by qPCR with primers specific for both the cellular and pseudoviral genomes. SNN overexpression led to an approximately twofold reduction in the number of HPV16 pseudogenomes over a range of virus doses (Fig. 8c).

Fig. 8.

Stannin promotes HPV16 capsid protein and genome degradation. (a) HeLa cells stably overexpressing Luciferase or untagged stannin were mock-infected or infected with HPV16.L2F PsV (1×10⁴ vge cell⁻¹) for 16 or 24 h. Cells were trypsinized to remove surface-bound virions, lysed and analysed by Western blotting with antibodies against actin, L1 and FLAG (L2). Non-specific bands are indicated with an asterisk. Full-length L2 and L1 are shown in the top two panels. The major small L1 cleavage product is shown with an arrow. (b) Quantitation of chemiluminescence signal intensities from the blot shown in (a) corresponding to samples prepared from cells infected for 24 h. ‘Cleaved L1’ corresponds to the smallest cleavage fragment of L1 indicated with an arrow in (a). Results show the mean and sd from three (cleaved L1) or two (full-length L1 and L2) independent experiments. Statistical significance was determined by an unpaired two-tailed t-test: **P≤0.01, *, P≤0.05. (c) HeLa cells stably overexpressing Luciferase or SNN were infected with 1×10³, 3×10³ or 1×10⁴ vge cell⁻¹ of HPV16 PsV as indicated. Twenty four h later, the cells were trypsinized to remove surface-bound virions. Extracted DNA was assayed by qPCR for neomycin gene copy number (pseudoviral genomes) and values were normalized to GAPDH gene copy number (cellular DNA). Results show the mean and sd from three independent experiments. Statistical significance was determined by an unpaired two-tailed t-test: **P≤0.01. (d) HeLa cells stably overexpressing Luciferase or SNN were mock-infected or infected with HPV16.L2F (6×10³ vge cell⁻¹). Eight hours later, all cells were fixed, reacted with antibodies against FLAG (L2) and LAMP1, and processed for the proximity ligation assay (green). Nuclei were stained with DAPI (blue). A single confocal slice is shown in each panel. Top panels show PLA signal (green); bottom panels show merged DAPI and PLA signals. Images are representative of three or more independent experiments. (e) HeLa cells stably overexpressing Luciferase or SNN were infected and processed as described in (d). The bar graph shows the mean and sd of the total PLA signal from three independent experiments, normalized to infected control cells. Statistical significance was determined by an unpaired two-tailed t-test: **P<0.01.

The increased abundance of a chloroquine-sensitive L1 cleavage product in SNN-overexpressing cells, together with the reduced amounts of pseudovirus DNA and full-length L1 and L2 in these cells at late times after infection, suggested that incoming virus was directed to the lysosome for degradation. To test this hypothesis, we used PLA to determine whether there were increased levels of HPV16 pseudovirions in LAMP1-positive subcellular compartments in SNN-overexpressing cells. SNN overexpression led to an almost twofold increase in the PLA signal between the FLAG-tagged L2 protein and the lysosome marker LAMP1 late after infection (Fig. 8d, e). Taken together, these data indicate that SNN overexpression causes incoming HPV to be degraded in the lysosome.

Discussion

Here we report an ISG overexpression screen to identify genes that regulate HPV16 entry into human epithelial cells. ISG overexpression studies for other viruses typically identified common sets of ISGs with broad antiviral activity, such as IRF1 and MyD88, which encode transcription factors that activate IFN-stimulated response element-mediated gene expression and TLR signalling, and cGAS, which encodes a cytosolic DNA sensor involved in antiviral resistance [54]. However, overexpression of IRF1, MyD88 and cGAS did not inhibit HPV infection in our screen, suggesting that HPV16 may use a unique strategy to evade intracellular immune surveillance. Rather, we identified a unique set of genes, including SNN, thrombomodulin (THBD) and vacuole membrane protein 1 (VMP1), which are not widely known for antiviral activity. Several of the top hits conferred protection against several HPV types (but not other viruses) and were active in both HeLa and HaCaT cells. Some of these hits participate in cellular processes known to play a role in HPV16 entry. For example, heparanase is an enzyme that cleaves HSPGs, which are critical for HPV16 binding to the cell surface [3], and VMP1 was reported to initiate autophagy, an antiviral mechanism that limits HPV16 infection [41, 55].

The atypical ISG signature revealed by our screen may reflect the retrograde cell entry pathway utilized by HPVs to traffic to the nucleus [56]. Consistent with this idea, the screen hit ankyrin repeat and FYVE domain containing 1 regulates the subcellular localization of the retromer complex [57], a key factor for HPV16 trafficking to the TGN. Other hits, such as stannin, were not previously known to affect processes involved in HPV infection. Because the retrograde pathway comprises a number of membrane-bound organelles, including the endosome, the Golgi and the ER, use of this entry pathway by HPV may sequester the virus away from the typical antiviral effector ISG products found in the cytosol.

Here we focused on the role of stannin in inhibiting HPV16 cell entry. SNN is an evolutionary conserved gene with poorly understood biological function. We show that overexpression of SNN inhibits HPV infection, and knockout of SNN enhances HPV16 infection. Overexpressed stannin inhibited infection by HPV16 and HPV18, which are responsible for ~70 % of cervical cancer and the vast majority of other HPV-associated mucosal cancer, as well as by HPV5, which infects skin cells. The N-terminal transmembrane domain and the linker region of stannin, but not the cytoplasmic C-terminal helix are required for its proper localization and HPV inhibitory activity. Although stannin was present in the ISG library, it is not induced by IFNs in HaCaT cells and is therefore not an ISG in this system. Our experiments also revealed that SNN is constitutively expressed in human epithelial cells including keratinocytes. As shown by analysis of SNN-knockout cells constructed by CRISPR-mediated mutagenesis, basal levels of the stannin protein are sufficient to inhibit infection of HeLa cells by HPV16 PsV. In this regard, stannin is similar to the restriction factor APOBEC3A whose endogenous levels significantly impair HPV16 cell entry even in the absence of IFN treatment [32].

Our results showed that stannin co-localized with internalized HPV16 PsV genomes in retrograde compartments, suggesting that stannin might directly interact with HPV virions. Indeed, co-immunoprecipitation studies showed that stannin binds to the capsid. This binding is mediated by L1, since L1-only PsV bind stannin, but binding may be indirect and appears to be modulated by L2.

We performed a series of experiments to determine the step in HPV16 infection regulated by stannin. SNN overexpression did not affect attachment, uptake or uncoating of HPV16 PsV in infected cells. In contrast, SNN overexpression decreased the abundance of HPV16 PsV in the TGN, suggesting that stannin disrupts proper HPV trafficking. Indeed, SNN overexpression disrupted the retromer-L2 interaction, as shown by the reduced PLA signal between L2 and the retromer subunit VPS35. However, further analysis revealed that the levels of full-length viral proteins and viral DNA are reduced by 16 h p.i. in SNN-overexpressing cells, complicating the interpretation of the reduced co-localization of viral DNA and L2 protein in the TGN at late times of infection.

The increased degradation of viral capsid proteins and pseudovirus DNA appeared to be mediated by lysosomal degradation because appearance of the major L1 cleavage product required acidification of the lysosome and increased levels of L2 were detected in LAMP1-positive compartments. Based on these results, we propose that endocytosed HPV virions enter the endosome and L1 interacts with endogenously expressed stannin (Fig. S8). This interaction directly or indirectly prevents the virus from engaging the retromer. In the direct model, the inability of the retromer to engage L2 blocks HPV trafficking to the TGN, and the virus is instead routed to the lysosome where degradation occurs. Alternatively, binding of stannin to L1 may directly target the virion to the lysosome, in which case is it no longer available to engage the retromer. The association of L1 with L2 and the viral genome throughout HPV entry suggests that L1 and perhaps L1/stannin binding plays a role in endosome escape [58]. Although the mechanism by which stannin promotes degradation of HPV and affects retromer association is unknown, this process resembles a recently described pathway of lysosome targeting of HPV by α-Defensin HD5 [59].

Our study reveals that stannin can bind to incoming HPV16 capsids and act as a selective inhibitor of entry by several HPV types. Although endogenous levels of stannin inhibit HPV16 infection, its role during natural infection is unclear. Stannin expression is not induced by IFN in skin keratinocytes, SNN knockout causes no more than a twofold increase of infection in HeLa cells, and most of our studies were performed in cells overexpressing stannin. It is possible that in some physiological conditions, stannin is an important regulator of HPV entry. For example, stannin expression may contribute to the low level of infectivity of HPV in cultured human primary keratinocytes [60]. Alternatively, the binding of L1 to stannin raises the possibility that this interaction may also affect other aspects of the virus life cycle.

Methods

Additional detailed Methods are shown in Supplementary Material.

Cells

HeLa cells and Vero cells were purchased from the American Type Culture Collection (ATCC, #CRM-CCL-2 and #CCL-81, respectively). HeLa-sen2 cells, a clone of HeLa cells, were reported previously [61]. ATG5−/− HeLa cells were produced by CRISPR Cas9-mediated mutation of the ATG5 gene. 293TT and HaCaT cells were obtained from Christopher Buck (NIH) and G. Paolo Dotto (Massachusetts General Hospital), respectively. Primary neonatal human foreskin keratinocytes were purchased from Lonza (#192906). Primary oral mucosal keratinocytes were a generous gift of Natalia Isaeva (Yale University). All primary keratinocytes were cultured in keratinocyte serum-free basal medium supplemented with bovine pituitary extract and human recombinant epidermal growth factor from Thermo Fisher Scientific (Rockford, IL). All other cell lines were cultured in Dulbecco’s modified Eagle's medium supplemented with 10 % FBS.

Viruses

HPV type 5, 16 and 18 PsV were produced as described previously by transfection of 293TT cells with plasmids encoding the viral capsid proteins and a reporter plasmid [21]. EdU-labelled HPV16 PsV were generated by incubation of transfected 293TT cells with 30 µM EdU 8 h post-transfection. All PsV were purified through Optiprep density gradient medium by ultracentrifugation. Purified PsV stocks were titred by proteinase K digestion and qPCR to determine the number of viral genome equivalents (vge). Replication-deficient recombinant human adenovirus type 5 (Ad-GFP, #1060) and adeno-associated virus type 2 (AAV-GFP, #7004) were purchased from Vector Biolabs (Malvern, PA). JC polyomavirus PsV was generated in 293TT cells and HSV1 encoding GFP was produced in Vero cells, as previously described [62].

ISG screening and confirmation assays

Lentiviral ISG-expressing libraries were generated as previously described [30]. For the screen, HeLa cells were seeded into 96-well plates at a density of 6000 cells per well and transduced with lentiviral pseudoparticles by spinoculation at 800 g for 45 min at 37 °C in cell culture medium containing 3 % FBS, 20 mM Hepes and 4 µg ml⁻¹ polybrene. Cells were infected 48 h post-transduction with HPV16-GFP PsV and harvested 48 h p.i. using the Accumax cell dissociation medium (Sigma). Cells were fixed in 1 % paraformaldehyde (PFA) and stored in PBS containing 3 % FBS until analysis by flow cytometry. Samples were analysed on a S1000 flow cytometer (Stratedigm) and analysed in FlowJo Cell Analysis software (TreeStar). To calculate the Z score, we first determined HPV16 infectivity for each ISG as a percentage of GFP-positive cells in the RFP-positive population. Z score was then computed with the STANDARDIZE function in Microsoft Excel according to the formula Z=(X-µ)/σ where X is infectivity, µ is the mean infectivity for all ISGs and σ is sd of the mean infectivity µ. Z scores for both screen replicates are plotted on a linear scale in Fig. 1b.

Immunofluorescence (IF) microscopy

We used HeLa and HeLa-sen2 cells for all IF experiments since they require 10 times less virus than HaCaT cells for efficient infection and because they provide better resolution of intracellular structures and virions. 2×10⁴ cells were seeded onto glass coverslips (#12-545-80 from Fisher Scientific) at low density (10–20 % confluency), incubated overnight and the coverslips transferred to a fresh well on the next day. Cells were fixed at room temperature in 2 % paraformaldehyde (PFA) in PBS for 30 min. Fixed cells were simultaneously permeabilized and blocked for 30 min at room temperature in PBS containing 3 % bovine serum albumin (BSA) and 0.5 % Triton X-100. Primary and secondary antibody incubations were carried out sequentially for 1 h at room temperature in PBS containing 3 % BSA. Edu-labelled HPV16 PsV genomes were detected with the Click-iT EdU Alexa Fluor 488 Imaging Kit (#C10337) and coverslips were mounted using Prolong Diamond Antifade medium with DAPI (#P36962) from Thermo Fisher Scientific (Rockford, IL). Images were visualized with a Leica SP8 confocal microscope and processed with the ImageJ program (NIH). More than 100 cells from at least five images were counted per experiment. The extent of co-localization was measured with the JaCoP plugin of ImageJ using Mander’s coefficient M1. Thresholds were set manually and kept constant for each set of images.

Immunoblotting

Cells were harvested in NP40 lysis buffer (1 % NP40 and 5 % glycerol in PBS) supplemented with a protease inhibitor cocktail and incubated for 30 min on a rotator at 4 °C. Lysates were then centrifuged at 15 000 r.p.m. for 10 min to sediment nuclei and cellular debris. Supernatants were transferred to new tubes, mixed with Laemmli buffer containing 5 % beta mercaptoethanol and boiled for 10 min at 95 °C. Proteins in each sample were separated by SDS-PAGE, transferred to a 0.2 µm PVDF membrane and incubated for 30 min in blocking buffer (Tris-buffered saline with 0.05 % Tween 20 and 5 % nonfat milk). Blots were incubated with primary antibodies diluted in blocking buffer for at least 12 h at 4 °C and then with secondary antibodies in blocking buffer for 1 h at 4 °C. Proteins were visualized by incubating the blots with the SuperSignal West Pico or West Femto chemiluminescent substrates from Thermo Fisher Scientific (Rockford, IL) and signals captured with the ChemiDoc XRS+System equipped with the Image Lab image acquisition and analysis software from Bio-Rad.

Immunoprecipitation to detect complex formation between stannin and HPV16

2×10⁵ HeLa cells expressing Luciferase or HA-stannin were seeded into 60 mm dishes and incubated overnight. On the next day, some of the cells were infected with 1×10⁴ vge cell⁻¹ of HPV16.L2F while control cells remained uninfected. Eight hours later, cells were harvested in 0.5 ml of ice-cold digitonin lysis buffer (0.5 % digitonin, 5 % glycerol, protease inhibitor cocktail, in PBS). Lysis was allowed to proceed for 1 h on ice with shaking once every 10–15 min to promote protein extraction. Insoluble cellular material was removed by centrifugation at maximum speed for 10 min in a tabletop centrifuge (Eppendorf). Supernatants were incubated with 10 µL of magnetic anti-HA beads overnight at 4 °C on a rotator. Beads were washed twice in PBS, bound proteins eluted in 2x Laemmli buffer and eluates analysed by SDS-PAGE and Western blotting.