In situ cryo-ET redefines HPV disassembly and transport paradigms

Abstract

Human papillomavirus (HPV) causes 4.5% of all human cancers, although a complete cellular basis of infection remains unclear. The mechanisms of virus disassembly and transport are particularly enigmatic. Here, we use in situ cryo-electron tomography (cryo-ET) to provide direct visualization of high-risk HPV16 during its infection cycle. We demonstrate that, when HPV reaches the lysosome, it remains intact and infection competent, contrary to the prevailing view of viral inactivation within lysosomes. We challenge the current model of HPV trafficking, which predicts progressive disassembly of the L1 capsid during retrograde transport, by showing that HPV remains intact when it reaches the Golgi en route to the nucleus for infection. Finally, we provide snapshots of the HPV containing transport vesicles budding from the Golgi lumen in a COPI coat-dependent fashion and show that HPV is connected to the overlying membrane by proteinaceous structural bridges, possibly formed by the L2 capsid, to guide virus trafficking. Our results redefine the dominant models of HPV entry, revealing the virus capsid’s resilience to disassembly during retrograde trafficking, a coat-dependent budding mechanism for transport, and demonstrate that the lysosome serves as a reservoir for infectious HPV.

Introduction

Human papillomaviruses (HPV) are a group of >200 related viruses—over 40 of which infect the anogenital and/or oropharyngeal mucosa—and are sexually transmitted (1–3). Infection is highly prevalent; however, 90% of infections clear spontaneously within ~2 years without symptoms (4, 5), but persistent infection with oncogenic (high-risk) types (e.g., HPV16/18) can lead to precancerous lesions and cancer (3, 6). Globally, HPV accounts for ~4.5% of all cancers (7, 8). It is the leading cause of cervical cancer and contributes significantly to anogenital and oropharyngeal cancers (9–11). HPV also causes a wide range of warts, including common warts, plantar warts, and anogenital warts (12, 13). Although effective prophylactic vaccines against multiple high-risk HPV types are available, there are currently no antiviral drugs, and the global vaccine coverage remains less than 15% (14, 15).

Structurally, HPV is a non-enveloped DNA tumor virus composed of 72 pentamers of the major capsid protein L1, which forms the outer shell of the capsid (16–22), and up to 72 copies of the minor capsid L2 protein buried within the capsid shell (20–22). When fully assembled, the capsid is approximately 55 nm in diameter (16–22) and encapsulates an 8 kb viral genome (23). To initiate cell entry, the L1 capsid binds to heparan sulfate proteoglycans (HSPGs) on the host plasma membrane, enabling furin-mediated cleavage of L2 at the cell surface (21, 24–28). Following cleavage, the virus is transferred to the cell surface receptor(s) and internalized through actin-mediated endocytosis (29, 30). The acidic environment of the endosome is thought to trigger partial disassembly of the HPV capsid, exposing the C-terminus of the L2 protein (31–44). The cell-penetrating peptide (CPP) motif at the extreme C-terminus of L2 (45–47), in concert with the host protein γ-secretase, assists in the insertion of L2 into and across the endosome membrane (43, 48, 49). This exposes a portion of L2 to the cytosol, enabling L2 to recruit cellular trafficking factors—including retromer, retriever, and the dynein-BICD2 motor complex—that target HPV to the trans-Golgi network (TGN) (50–58).

Upon reaching the TGN (59), the cytosol-exposed segment of L2 binds directly to the coat protein I (COPI) coatamer complex (60). This coat protein complex is postulated to induce the formation of membrane-bound vesicles that function to ferry HPV from the TGN to and across the Golgi stacks. When HPV reaches the cis-Golgi compartment, importin receptors, including IPO7, help deliver HPV to the nuclear membrane (61). Nuclear envelope breakdown at the onset of mitosis enables the virus to enter the nucleus and access the genome to cause infection (62–64). Therefore, retrograde trafficking along the endosome-TGN/Golgi-nucleus axis represents the productive infection pathway. By contrast, because endosome-localized HPV can also be delivered to the lysosome, where it is thought to be degraded, endosome-to-lysosome transport of HPV has been considered a non-productive route (30, 32–34, 38, 65, 66).

Despite its clinical importance, the processes of HPV entry, intracellular trafficking, and uncoating remain incompletely understood. A central question is the extent of HPV disassembly during retrograde trafficking, as biochemical studies suggest a progressive, stepwise disassembly (51, 67–70); however, visual evidence to support this model is lacking. The identity of transport vesicles mediating retrograde HPV movement is also unclear. Disruption of host factors involved in membrane tubulation and vesicle formation (including retromer (47, 51, 53, 56, 57), sorting nexins (50, 54, 55), and COPI (60)) blocks infection, highlighting their importance in virus transport. However, due to the asynchronous nature of HPV entry and the transient presence of HPV in these compartments, direct visualization of the virus in membrane-bound vesicles has not been achieved (71).

Here, using the high-risk HPV16 model system, we redefine the prevailing model of HPV retrograde trafficking by providing unprecedented, high-resolution in situ snapshots of key entry events. Integrating cryogenic correlative light and electron microscopy (cryo-CLEM) with in situ cryo-electron tomography (cryo-ET), pharmacological perturbations, confocal microscopy, and biochemical assays, we show that: (i) virions within the lysosomal compartment remain structurally intact and infectious; (ii) intact HPV is retrograde trafficked to the Golgi; (iii) intact HPV is a cargo of COPI-coated vesicles; and (iv) the vesicular carriers that ferry HPV, and persist through mitosis, exhibit defined architectural features.

Together, these findings reposition the endolysosomal system as a reservoir and staging site for infectious HPV and establish a revised framework for HPV entry and retrograde transport to the Golgi.

Results

Characterizing fluorescently labeled HPV pseudovirus.

To precisely localize HPV in the cell during infection, we first optimized a method to fluorescently label the well-established model system for studying papillomavirus infection mechanism, the HPV16 pseudovirus (PsV) (29, 72). These PsV are comprised of high-risk HPV16 L1 and L2 capsid proteins (where L2 is fused with a C-terminal 3x-FLAG tag) that encapsulate a GFP-S reporter plasmid in place of the native HPV16 viral genome. In this system, GFP-S (hereafter referred to as GFP) expression serves as an indicator of the successful delivery of the PsV DNA to the nucleus, confirming productive entry into the cell. Importantly, infection by HPV16 PsV closely mimics the entry pathway of authentic HPV isolated from stratified keratinocyte raft cultures; thus, this system has been widely adopted and used as a standard tool for investigating HPV entry mechanisms (73–77).

To generate a fluorescently labeled virus, purified HPV16 PsV (hereafter, HPV) was covalently labeled with Alexa Fluor 647 (Figure S1A: AF647-HPV). Successful fluorescent labeling was confirmed by SDS-PAGE, which revealed distinct fluorescent bands corresponding to the L1 and L2-FLAG proteins (Figure S1B). We verified that the AF647-HPV retained its structural integrity by negative staining electron microscopy (nsEM). AF647-HPV had an average diameter of 56.71 ± 2.133 nm (Figure S1C), consistent with the size of unlabeled HPV PsV (16–22). We next verified whether labeling the virus affected its infection competency. Using the widely studied model for HPV, the immortalized cervical cancer cell line HeLa (78), we observed no significant difference in GFP expression between cells infected with labeled or unlabeled pseudovirus (Figure S1D; relative infection is quantified in bottom graph). Infection by AF647-HPV was also sensitive to pharmacological treatment that disrupt critical HPV infection steps: XXI, a γ-secretase inhibitor that blocks L2 insertion across the endosome membrane required for trafficking (48, 79–81); NAV-2729, an Arf1 inhibitor that interferes with COPI recruitment to the Golgi necessary for retrograde transport across the Golgi (60, 82, 83); and RO-3306, a Cdk1 inhibitor, which prevents nuclear envelope breakdown essential for HPV nuclear entry (Figure S1E; relative infection is quantified in bottom graph) (84, 85). Additionally, at 22 hours post-infection (h.p.i), when HPV begins to reach the Golgi, AF647-HPV colocalizes strongly with the Golgi marker HA-COPγ1-mCherry (Figure S1F), with quantitative analysis demonstrating a high Pearson correlation coefficient between AF647-HPV, L1, and HA-COPγ1-mCherry (Figure S1G–I). Together, these findings demonstrate that AF647-HPV uses the same entry pathway as the unlabeled virus to promote infection, supporting its use as a reliable tool for detailed imaging and mechanistic studies of virus trafficking.

Lysosomes act as reservoirs for infection competent HPV

The canonical model of HPV trafficking hypothesizes that upon reaching the endosome from the cell surface, HPV is targeted either along the TGN/Golgi axis for productive infection or to the lysosome for degradation, thereby being rendered ineffective for infection. To examine this paradigm, we first determined whether HPV is present within the lysosomes by confocal fluorescence microscopy. We found that HPV L1 colocalizes strongly with the lysosome marker LAMP1 as early as 8 h.p.i (Figure S2A, 1^st row; quantified in Figure S2B), consistent with previous reports (30, 32–34, 38, 65, 66). The extent of L1-LAMP colocalization increases at 16 h.p.i (Figure S2A, 2^nd row; quantified in Figure S2B) and remains at this high level even at 72 h.p.i (Figure S2A, 3^rd-5^th rows; quantified in Figure S2B). These colocalization studies suggest that, in contrast to the accepted model, HPV is present in the lysosome for an extended time and is unlikely to be completely degraded by the lysosomal proteases.

Given our initial finding of viral retention within lysosomes, we questioned whether HPV remained intact within the acidic environment. To enable the high-resolution visualization in situ, we employed a cryogenic-correlative light and electron microscopy workflow (cryo-CLEM, Figure S3), where HeLa cells were infected with AF647-HPV for 16 hours, treated with Lysotracker for 3–5 minutes for precise localization of lysosomes, and plunge-frozen in liquid ethane. Cryo-confocal microscopy of these cells revealed a near-perfect fluorescence colocalization between HPV and lysosomes, enabling precise cryo-focused ion beam (cryo-FIB) milling and cryo-electron tomography (cryo-ET) tilt series data collection (Figures 1A and 1B, Figure S3). We observed that lysosomes can contain as few as a single virus particle or be densely packed with visually discernible HPV capsids (Figures 1C and 1D, Figure S2D; Movie S1). Strikingly, these lysosomal HPV were indistinguishable from purified HPV PsV, retaining the characteristic appearance of the L1 pentamer (Figure 1Ci–v, yellow arrowheads), and contained density at the center, likely of the plasmid DNA (Figure 1Ci–v, green arrowheads). Moreover, quantification of the diameter of lysosomal HPV particles revealed that it was similar to that of purified HPV PsV (56.92 ± 1.318 nm) (Figure 1E).

Figure 1. Human papillomavirus retains its morphological integrity within host lysosomes.

(A) Cryogenic-correlative light and electron tomography (cryo-CLEM) analysis of HeLa cells infected with human papillomavirus (AF647-HPV). Overlay of the cryo-transmission electron microscopy (cryo-TEM, 6500×) and maximum intensity projection (MIP) cryo-confocal fluorescence image of a lamella through a HeLa cell, 16 hours post-infection (h.p.i.). Cryo-TEM (gray), and fluorescence signals from Lysotracker DND-26, and HPV, pseudocolored in green and red, respectively, are shown. The black dashed box highlights the region selected for tilt series acquisition. Scale bar: 5 μm. (B) Slice through a tomogram of the area highlighted in (A), overlaid with MIP cryo-confocal fluorescence signals: AF647-HPV (red) and LysoTracker (green). Scale bar: 100 nm. (C) Slice through a tomogram of the area highlighted in (A), highlighting intact cellular interiors of an HPV-infected HeLa cell. Cellular components such as ribosomes (Ribo), lysosomes (Lyso), microtubules (MT), and the virus (HPV) are easily discernible and pointed out by white arrows. Zoom-in views (5×, i-v) highlight five HPV particles within the lysosomes. The distinct crown-like shape of the HPV L1 pentamer and the density of plasmid DNA are marked by yellow and green arrowheads, respectively. Scale bar: 25 nm, inset: 100 nm. (D) 3D Segmentation of tomogram in (C) highlights cellular features rendered as: HPV (yellow), lysosome (dark blue), endoplasmic reticulum (ER, olive green), ribosome (salmon), microtubule (dark green). Scale bar: 100 nm. (E) Quantification of HPV capsid diameters. Diameter is defined as the maximum inter-pentamer (L1-L1) distance on the opposite side of the HPV capsid. Lysosomal HPV (n=112, 56.92 ± 1.318 nm) is similar in size to HPV purified in vitro (n=47, 56.71 ± 2.133 nm). (mean ± SD, Mann-Whitney t-test).

LysoTracker is a weak base that becomes protonated and fluoresces in acidic environments (86). Our ability to precisely detect lysosomes in HPV-infected HeLa cells in situ suggests that these lysosomes retain their acidic pH, and yet the HPV capsids remain relatively intact. Intrigued by this finding, we next assessed whether lysosome-localized HPV contains both capsid proteins, L1 and L2, and is infection-competent. Notably, a prior study has suggested that the redox state of specific amino acids within L2 is detrimental to infection (87). We immunopurified the lysosomes (lyso-IP) from HPV-infected HeLa cells that contained an HA tag at the C-terminus of an integral lysosomal membrane protein, TMEM192 (88–90) (Figure 2A). We first verified that our lyso-IP method enriched only lysosomes. Immunoblot analysis revealed the presence of HPV capsid proteins L1 and L2, as well as lysosomal proteins LAMP1 and TMEM192, and the absence of other organelles, including the endoplasmic reticulum (ER), Golgi apparatus, nucleus, and early endosomes (EE) (Figure 2B). Next, we sonicated the lysosomes to release the lysosome-resident HPV in an acidic buffer pH 4.5 and evaluated its infectivity. Remarkably, using GFP expression to assess infection (Figure S1D), we found that the lysosome-derived HPV displayed similar infectivity to purified HPV PsV (Figure 2C). We reasoned that because the lysosome-localized HPV is infectious, artificially permeabilizing the lysosomal membrane with L-leucyl-L-leucine O-methyl ester (LLOME) (91–94) might allow the virus to escape the lysosome and reach the cytosol, providing the virus an opportunity to enter the nucleus and cause infection. Indeed, LLOME treatment of HPV-infected cells increased infection significantly (Figure 2D), further supporting the idea that HPV in the lysosome is infectious. Furthermore, when cell cycle was arrested with the Cdk1 inhibitor RO-3306, LLOME treatment failed to restore infection (Figure S2C). These results indicate that the enhanced HPV infection resulting from the artificial release of the virus from the lysosome still requires cell cycle progression.

Figure 2. Lysosome-derived HPV is structurally intact and infection competent.

(A) Schematic of lysosomal immunoprecipitation (lyso-IP) workflow (see detailed explanation in Materials and Methods). Briefly, 1) HeLa cells are transfected with the TMEM192-HA plasmid. 2) 24 hours post-transfection (h.p.i.), cells are infected with unlabeled HPV pseudovirus (MOI=50). 3) 16 hours later, cells are lysed and the lysate incubated with anti-HA beads for 5 minutes at room temperature. 4) Beads are incubated in a hypotonic pH 4.5 buffer for 20 minutes on ice and then briefly sonicated to rupture the lysosomes. (B) Representative immunoblot showing whole-cell lysate (WCL), immunoprecipitated beads (Lyso-IP) against antibodies recognizing HPV capsid proteins (L1 and L2-FLAG) as well as different organelle markers, namely: LAMP1 and TMEM192-HA (lysosome), BIP (ER), EEA1 (early endosome), β-actin (cytosol), GM130 (Golgi), and Histone H3 (nucleus). (C) Representative immunoblot (bottom) and quantification (top) of reporter GFP expression in HPV-infected HeLa cells, prepared using standard procedures (in vitro) or extracted from lysosomes (Lyso. HPV). The amount of added virus was standardized based on L1 protein levels determined by immunoblotting (see Materials and Methods for details). At 48 h.p.i., cells were harvested, lysed, and the resulting whole-cell lysate was subjected to SDS-PAGE, followed by immunoblotting with antibodies recognizing GFP and β-actin as a loading control. GFP band intensities are normalized to β-actin in each sample. Total GFP band intensities per condition are normalized against the in vitro sample to give normalized infectivity (Normalized inf., %) (Welch’s t-test) (N = 3). (D) Immunoblots and quantification of GFP expression during HPV infection in the presence of LLOME were performed as in (C). Cells were either treated with Dimethyl sulfoxide (DMSO, mock control) or 1 mM of L-leucyl-L-leucine O-methyl ester (LLOME). For quantification, DMSO was used as the normalization standard. (Welch’s t-test, ** p=0.0054) (N=3). (E) Single particle cryo-electron microscopy analysis of lysosomal-extracted HPV. 6.3Å Coulomb potential map (center, yellow) showing the characteristic icosahedral architecture of HPV. L1 capsomer forms both hexavalent (inset right, top) and pentavalent (inset right, bottom) interactions with other L1 capsids. Pentavalent L1 capsomer is highlighted in magenta. A rigid-body fit of the L1 structure (PDB: 7KZF) into the L1 capsomer of the lysosomal HPV map (inset left top) is shown with a view direction from within the virus to the outside. Additionally, a close-up view of the selected regions of the capsomer density is shown (inset left, bottom).

We next assessed whether lysosomal HPV retains the characteristic icosahedral architecture. We employed single-particle cryo-electron microscopy (cryo-EM) to determine the structure of the lysosomal HPV. Cryo-EM analysis revealed that the lysosomal HPV consists of 72 capsomers, where each capsomer is formed by 5 L1 proteins, akin to the previously determined structure of HPV (21, 22). We found that 12 capsomers form pentavalent interactions with neighboring capsomers (magenta, Figure 2E), while 60 L1 capsomers form hexavalent interactions. The previously determined structure of the capsomer (PDB: 7KZF) fits very well in our Coulomb potential map (Figure 2E, insets to the left). Our map aligns perfectly well with the previously determined cryo-EM structure of the isolated HPV capsid (Figure S4D, EMD-23081). Therefore, at 6.3 Å, our reconstruction of lysosomal HPV provided sufficient architectural definition for this study, and we chose not to pursue additional data collection and resolution improvement.

Collectively, we show that lysosome-localized HPV is intact and infectious. Our results demonstrate that HPV virions in the lysosome maintain their structural integrity and do not undergo significant disassembly or degradation as previously proposed (31–44). Remarkably, the structurally intact virus remains infectious, suggesting that the lysosome may act as a reservoir for infectious HPV.

Intact HPV traffics to the Golgi, where it buds into COPI-coated transport vesicles

The prevailing paradigm in HPV trafficking and infection posits that L1 capsid disassembles progressively as the virus undergoes retrograde transport to reach the TGN/Golgi (51, 67–70). L1 capsid disassembly is thought to expose the hidden L2 capsid, allowing it to insert and translocate into the host membrane, which is necessary to guide the virus for productive infection (46, 48, 95, 96). Intrigued by our discovery that HPV is resilient to lysosomal degradation, we asked whether intact HPV capsids are trafficked to the TGN/Golgi. To visualize the architecture of HPV during retrograde trafficking, we collected and analyzed cellular tomograms of HeLa cells infected with HPV at 22 h.p.i., specifically at sites of AF647-HPV and HA-COPγ1-mCherry fluorescence colocalization. To our surprise, we observed intact HPV capsids in the Golgi lumen (Figures 3A and 3B; Movie S2). Detailed examination of various tomographic Z-slices of the HPV in the process of budding reveals the formation of the COPI coat at the initiation of budding from the Golgi membrane (Figure 3C–E; a schematic of the Golgi membrane, COPI coat, and HPV is shown on the right). We also observed a fully budded vesicle harboring HPV that is surrounded by a partial COPI coat, localized adjacent to the discernible Golgi stacks (Figures 3F and 3G). Finally, across several tomograms, we captured snapshots of HPV in the cellular milieu, thereby allowing us to reconstruct the entire budding process. We categorized the process into four distinct steps: HPV capsids initially reside in the Golgi lumen, followed by the recruitment of the COPI coat and membrane budding, which leads to the formation of a COPI-coated vesicle containing the virus (Figure 3H). Eventually, the COPI coat disassembles from the budded vesicles (Figure 3H, right).

Figure 3. In situ structural analysis of HPV at the Golgi.

(A) Slice through a reconstructed and filtered tomogram collected at the site of COPγ1 and HPV colocalization (see Figure S3), revealing the cellular milieu at the Golgi. Two HPV particles are present within the Golgi lumen. Cellular components such as the endoplasmic reticulum (ER), Golgi, microtubules (MT), ribosomes (Ribo.), and HPV are depicted by white arrows. Scale bar: 100 nm. (B) 3D segmentation of tomogram in (A) with key cellular features rendered as: HPV (yellow), Golgi (maroon), ER (olive green), vesicle (purple), multivesicular bodies (MVB, light blue), ribosome (salmon pink), and microtubules (dark green). Scale bar: 100 nm. (C-E) Left: close-up views of the boxed region in (A), showing three different tomographic slices at 184 (C), 154 (D), and 134 (E), showing the formation of the COPI coat (blue arrowheads) around the HPV particles (yellow arrowheads) ready to exit the Golgi lumen. Right: Cartoon schematic of the COPI coat (cyan), Golgi (maroon), and HPV (yellow) corresponding to the slices on the left. Scale bar: 50 nm. (F) Slice through a tomogram showing a characteristic Golgi stack and the HPV particles post exit from the Golgi, within vesicles. Scale bar: 50 nm. (G) 3D segmentation of the tomogram in (F), highlighting the Golgi (maroon) and the vesicles (lilac) harboring HPV particles. Scale bar: 50 nm. (H) Top: representative snapshots depicting the process of HPV exit from the Golgi. Slices through different tomograms, from left to right, show formation of the COPI coat, expansion of the coat, a fully mature vesicle about to be released from the Golgi, and HPV in a vesicle derived from the Golgi. Scale bar: 50 nm. Bottom: Cartoon schematic of the various stages of HPV exit from the Golgi. COPI (cyan), Golgi (maroon), and HPV (yellow) are shown. (I) Quantification of the capsid diameters of purified HPV (left, black circles), lysosomal HPV (middle, yellow circles), and HPV within the Golgi or coated vesicles (yellow circles, right). Diameter is defined as the maximum inter-pentamer (L1-L1) distance on the opposite side of the HPV capsid. HPV within the Golgi and vesicles (n=63, 56.91 ± 1.584 nm) shows no change in diameter, in comparison to purified in vitro (n=47, 56.71 ± 2.133 nm) or lysosomal HPV (n=112, 56.92 ± 1.318). (mean ± SD, Mann-Whitney t-test) (J) Quantification of COPI vesicle diameters, measured as the maximum distance between the centers of opposing membrane leaflets, with HPV (n=22, 81.05 ± 4.325 nm) and without HPV (n=104, 56.4 ± 5.762 nm) (mean ± SD, Mann-Whitney t-test,**** p<0.0001) (K) Bar plot illustrating the number of observed instances of vesicles with a full COPI coat (2/22), or partial coat (20/22) in cryo-ET data.

Throughout all stages of this process, the HPV capsids remain morphologically assembled as defined by the observed crowns of the L1 pentamers in the tomographic slices. We measured the diameters of the HPV particles within the Golgi compartment and COPI-coated vesicles. We found no difference between the mean capsid diameter in the Golgi/vesicle populations (56.91 ± 1.584 nm), lysosomal HPV (56.92 ± 1.318 nm), and purified HPV particles (56.71 ± 2.133 nm) (Figure 3I), revealing that the virus capsids remain structurally intact and do not undergo extensive disassembly during trafficking to the Golgi.

We next assessed the diameter of COPI-coated vesicles with and without the HPV cargo. Vesicles lacking virus cargo had a mean diameter of 56.4 ± 5.762 nm (Figure 3J), consistent with previous reports (97–99). By contrast, COPI-coated vesicles containing HPV were significantly larger, with a mean diameter of 81.05 ± 4.325 nm. This substantial size increase demonstrates that the COPI vesicles can adapt in size to accommodate an unusually large cargo. Finally, we also examined the state of COPI-coated vesicles containing HPV. We found that 20 out of 22 vesicles exhibited partial disassembly of the COPI coat, while only two remained fully coated (Figure 3K). This finding is consistent with the established maturation mechanism of COPI-coated vesicles in which the COPI coat dissociates from the membrane following vesicle formation (100–103).

Taken together, we show unequivocally that intact HPV capsids bud out of the Golgi in COPI-coated vesicles.

Proteinaceous structural bridges link HPV to host membranes during retrograde trafficking

We consistently observed distinct protein densities spanning the gap between HPV and the Golgi lumen [Figure 4A; red arrowheads (top panel) and red segmented densities (middle and bottom panel) in representative examples of three independent vesicles containing HPV]. Each instance showed multiple discrete densities bridging the virus capsid and the overlaying membrane surface. Similarly, we also found connector densities between HPV and the encapsulating membranes in fully budded vesicles (Figure 4B). Again, the tomograms revealed the presence of comparable proteinaceous densities at the interface between the virus capsid and vesicular membrane, with several apparent contact points. We did not detect these densities in vesicles lacking HPV (Figure S3H), suggesting that these structures may be derived from the virus rather than the host. We quantified both the length and width of these bridging densities, in which the length is defined as the distance from the surface of the virus to the inner leaflet of the vesicular membrane, and the width as the span from one end of the proteinaceous density to the other. On average, the densities measured 7.268 ± 1.214 nm in length and 2.326 ± 0.511 nm in width (Figure 4C). The remarkably consistent length of the connector density would imply that HPV within Golgi-derived vesicles of a substantially larger diameter would not be centrally positioned within. To verify this hypothesis, we performed quantitative centroid analysis of HPV and its enclosing vesicles in our tomograms, where we defined virus offset as the distance between the center of the virus and the center of the vesicle, as illustrated in the schematic (Figure 4D). Our analysis revealed that the center of the virus is systematically offset from the center of the vesicle, indicating an intrinsic asymmetry in virus–vesicle interactions that skews the virus towards one side of the vesicle (Figure 4D).

Figure 4. Protein densities bridge the viral capsid to Golgi and COPI-coated vesicle membranes.

(A, B) Top: Representative slices through reconstructed and denoised tomograms showing HPV within the Golgi lumen (A) and inside a COPI-coated vesicle (B), each exhibiting multiple proteinaceous densities (red arrowheads) bridging the capsid and the membrane. Middle and bottom row: The corresponding 3D segmentation, with and without membrane, highlights cellular features rendered as: HPV capsid (yellow), Golgi (maroon), proteinaceous density (red). Bright green arrowheads indicate extension of the bridging densities. Scale bars: 50 nm. (C) Schematic and quantification of the length (7.268 ± 1.214) and the width (2.326 ± 0.511) of the bridging densities, defined as the length as the distance from the surface of the virus to the inner leaflet of the vesicular membrane, and as the distance spanning from one end of the proteinaceous density to the other, respectively. (mean ± SD) (n=45). (D) Scatter plot and schematic illustrating virus offset relative to its diameter within the vesicle, with each point representing an individual HPV-containing vesicle (n = 22). (E) Representative negative stain electron micrographs of immunogold (anti-FLAG) labeled vesicles containing HPV, isolated via FLAG-IP. Yellow arrowheads indicate HPV, and red arrowheads indicate gold beads recognizing the C-terminus of HPV L2 protein and decorating the surface of the vesicle. Scale bar: 50 nm. (F) Schematic with proposed model for L2 conformational changes. In isolated (free) HPV, L2 is tucked under the HPV L1 capsid pentamer, as reported by Buck et al. (20). However, when HPV reaches the endosome during entry, cellular triggers in this organelle expose L2. L2 then inserts into the endosome membrane, revealing most of it (including the CSR, which contains a COPI-binding site;(60)) to the cytosol. After reaching the TGN/Golgi, L2 engages cytosolic COPI, which delivers HPV across the Golgi stacks. When HPV finally exits the cis-Golgi in a vesicle, COPI disassembles from the vesicle. This allows L2 to undergo “backsliding” (see Discussion), enabling the CSR to withdraw back into the vesicular membrane, while the L2 C-terminus remains on the exterior of the vesicular membrane

While it remains debated, the current understanding is that during retrograde trafficking, the HPV L2 capsid is exposed from the HPV L1 capsid, enabling it to insert and translocate across host membranes to recruit cytosolic cellular factors that guide virus transport (45, 46, 48, 95). In some slices of our tomograms, we observe the extension of the connector density across the vesicle membrane bilayer (Figure 4B, light green arrowheads). Additionally, it has been estimated that up to 72 copies of L2 are present per HPV capsid (20–22). In our in situ tomograms, we never observed more than 8 virus-vesicle connector densities. Moreover, the structural prediction of L2 revealed a central structured region (CSR), located between the disordered N- and C-termini, with approximate dimensions of 6.2 nm and 2.8–4 nm along the long and short axes, respectively (Figure S5). These dimensions align reasonably with those of the proteinaceous bridges (Figure 4C). We therefore surmised that the CSR of the L2 capsid protein may form this structural bridge between the virus and the encapsulating vesicle and would orient in a way that the distal L2 C-terminus would be accessible on the exterior of the vesicle.

To verify that the C-terminus of L2 is exposed on the exterior of the vesicle, we took advantage of the 3X-FLAG epitope engineered at the C-terminus of L2 in our HPV PsV. We used FLAG immunoprecipitation to isolate intact HPV containing vesicles rapidly. In line with our hypothesis that the C-terminus is exposed to the exterior of the vesicle and accessible, we were able to purify the vesicles. Furthermore, we also verified that the L2 C-terminus is accessible on the vesicle surface using gold-labeled anti-FLAG M2 antibody. In several cases, we were clearly able to identify the virus within the vesicles (Figure 4E, yellow arrowheads). We also observed gold beads decorating the exterior of isolated vesicles (Figure 4E, red arrowheads). These results demonstrate that the C-terminus of L2 is exposed on the exterior of the HPV-containing vesicles.

Previous structural studies of isolated HPV particles demonstrated that the density for L2 is adjacent to or below the L1 pentamer in the Coulombe potential maps (20, 22). However, L2 has not been distinctly resolved, owing to its large intrinsically disordered regions (104). We used AlphaFold3 (105) to predict the co-complex structure of HPV L1 pentamer with full-length L2 (Figure S5). While the overall robustness is moderate (ipTM = 0.7, pTM = 0.73), in the prediction, L2 CSR is adjacent to the L1 pentamer, and the C-terminus of L2 is buried within the capsid shell. Based on previous structural studies and AlphaFold predictions, we propose a model where, in isolated HPV, L2 is concealed beneath the L1 pentamers in the native capsid, and the C-terminus of HPV L2 can interact with viral DNA (106, 107). However, when HPV reaches the endosome during entry, cellular cues in this compartment induce dramatic conformational changes to the virus that expose L2. The exposed L2 then inserts into the endosome membrane, revealing a majority of L2 [including the CSR, which harbors a COPI-binding motif; (60)] to the cytosol. Upon reaching the TGN/Golgi, HPV L2 recruits cytosolic COPI, which transports the virus across the Golgi stacks. When HPV finally exits the cis-Golgi in a vesicle, COPI disassembles from the vesicle. This allows L2 to partially “backslide” (see Discussion), enabling the CSR to retract into the vesicular membrane—a configuration we directly visualized in our in situ cryo-tomograms—while the C-terminus of L2 remains on the exterior of the vesicular membrane (Figure 4F).

Collectively, these findings highlight a direct physical connection between HPV and host vesicular compartments, indicating that the CSR of L2 serves as a key mediator of this crucial membrane engagement.

Discussion

Globally, HPV is responsible for 4.5% of all human cancers and remains a significant public health threat (7, 8). While highly effective prophylactic HPV vaccines exist, the great majority of eligible individuals worldwide have not been vaccinated; current estimates suggest that 13 million new infections are reported annually in the U.S. alone, and the average lifetime probability of acquiring HPV among those with at least one opposite-sex partner is 84.6% for women and 91.3% for men (14, 15, 108, 109). Several reports also suggest that older people are more susceptible to HPV reactivation (110–117). Clarifying the cellular infection routes of HPV will provide better insights into the design of new anti-HPV strategies. While biochemical, cell-based, and conventional microscopy studies have advanced our understanding of HPV cell entry and trafficking, key assumptions about the mechanism of capsid disassembly and vesicular transport—decisive events in infection—have not been rigorously examined. Our in situ cryo-ET study provides key molecular insights into the cellular trafficking of HPV within the host cytoplasm.

After receptor-mediated endocytosis, HPV reaches the endosome, where the L1 capsids are thought to undergo partial disassembly (31–44). Specifically, host factors, including specific Rab GTPases, the ESCRT machinery (i.e., Tsg101 and Vps4), and the V-ATPase, have all been implicated in promoting virus disassembly in the endosome (36, 40, 37). However, it remains unclear whether their role in HPV disassembly is direct or whether it is instead related to maintaining general endosomal homeostasis. Reports also suggest that other endosome-associated host factors, such as tetraspanin CD63, syntenin-1, ALIX, and the A2t complex, participate in the disassembly of HPV (35, 39, 41). Similarly, another hypothesis posits that the low pH environment of the endosome triggers capsid “loosening” to initiate disassembly. This idea is primarily based on in vitro studies demonstrating that the low pH enables DNase I to access and degrade the internal viral genome (32).

Regardless of the precise mechanism, disassembly is thought to expose the L2 capsid protein hidden within the L1 capsids. The exposed L2 protein inserts across the endosome membrane and becomes accessible to the cytosol (45, 46, 48, 95). The cytosol-exposed region of L2, in turn, recruits cytosolic trafficking factors that guide the virus to the TGN, committing the virus to the productive infection route (50–58). We previously demonstrated that at the TGN, HPV L2 binds directly to the COPI coat complex, enabling the virus to undergo retrograde transport to and across the Golgi stacks before reaching the nucleus to cause infection (60). However, how COPI promotes retrograde trafficking of HPV is not entirely clear. Importantly, in this scenario, HPV is expected to be at least partially disassembled upon arrival at the Golgi.

HPV at the lysosome

As an alternative destination to the TGN/Golgi, a pool of HPV in the endosome can also be delivered to the lysosome, where it is thought to be degraded after entry (30, 32–34, 38, 65, 66). We demonstrate conclusively that lysosome-localized HPV remains morphologically intact, with an average diameter indistinguishable from that of native or Golgi-localized virus (Figures 1B and C). Moreover, by isolating lysosome-localized HPV in buffer conditions mimicking lysosomal pH (4.5–5.0) (118), and determining the moderate-resolution single-particle cryo-EM structure, we unequivocally demonstrated that the capsid structure can withstand a lysosomal acidic environment (Figure 2A, E). This finding is in agreement with our confocal microscopy data, which reveal a high level of colocalization between L1 and the lysosomal marker LAMP1, even at 72 hours post-infection (Figure S2A and B). In parallel studies, in vitro single-particle cryo-EM of HPV incubated in buffers spanning near-neutral to moderately acidic conditions (pH 7.11, 6.08, 5.45) showed no detectable change in capsid diameter, with pronounced capsid degradation only occurring at pH 2 (119). Finally, we also showed that lysosome-localized HPV is equally potent in its infection capability as the purified pseudovirus (Figure 2B–D). Our findings strongly suggest that the lysosome is unlikely to degrade all HPV and may serve as a reservoir of infectious HPV.

What might be the advantage of storing HPV in the lysosome? One possibility is that, should the endosome be overcrowded with HPV, which prevents its efficient delivery to the TGN/Golgi for productive infection, endosome-localized HPV might instead be directed to the lysosome for temporary storage and only re-routed back to the endosome when a less crowded condition prevails. In this manner, there would be a sustained supply of HPV particles that can be used for infection. This proposed lysosome-endosome-TGN/Golgi transport pathway has precedence. For instance, in yeast, the Atg27 membrane protein is recycled from the vacuole (lysosome equivalent in yeast) to the endosome and then targeted to the Golgi by the retromer (120). The possible involvement of the lysosome during HPV entry is not exclusive to this virus. For instance, coronaviruses, including SARS-CoV-2, utilize the lysosome for unconventional egress by modifying the harsh conditions of this compartment (121). In contrast, the viral glycoprotein in the Ebola virus is cleaved by lysosomal proteases, allowing the virus and lysosome membranes to fuse and deliver its genome into the cytosol (122).

HPV at the Golgi

Our study demonstrates that the HPV capsid retains its structural integrity in the Golgi lumen, on the COPI-coated Golgi membrane, during budding initiation, and in Golgi-derived budded vesicles (Figure 3). In these different steps, quantitative measurements reveal no significant differences in the HPV capsid diameters when compared to the native virions. These results indicate that the capsid architecture remains unchanged mainly throughout retrograde transport from the cell surface to the Golgi. Whether HPV is disassembled after exiting the Golgi en route to the nucleus is unclear and deserves future investigation. In this context, a previous study using standard TEM showed that HPV particles within a membrane-bound vesicle can be found in the nucleus (123). However, the lower resolution of this approach cannot determine if the virus is fully intact.

Serial snapshots of intact HPV in the Golgi lumen, COPI-coated pits in the Golgi membrane, and fully-budded vesicles (Figure 3H) strongly suggest that the COPI budding machinery is responsible for generating membrane-bound vesicles harboring HPV that are used for retrograde trafficking. Our in situ cryo-ET data not only provide visual evidence but also validate our previous work, suggesting that HPV L2 binds to COPI during retrograde trafficking across the Golgi, thereby facilitating infection (60). In situ structural analysis of COPI coats in Chlamydomonas reinhardtii indicated the presence of cargo bound beneath the B’ subunit of COPI (98); nevertheless, the precise nature of cargo remained unknown. Here, we provide the first direct in situ evidence of a virus exploiting the mammalian COPI machinery for retrograde transport (Figure 3). The only previous study demonstrating bona fide cargo with coated vesicles, by in situ cryo-ET, was in COPII-coats; however, this was performed using a reconstituted approach with purified proteins and cell-derived membranes, rather than in the native cellular condition (124).

Our cryo-ET analysis revealed that HPV-containing budded vesicles were significantly larger than those without viral cargo. While vesicles devoid of HPV are approximately 56 nm in diameter and are consistent with previous reports (97–99), HPV-containing vesicles expand to approximately 81 nm in diameter (Figure 3J). This is the first report demonstrating that mammalian COPI-coated vesicles can adapt in size, growing to accommodate the cargo. The ability to adjust likely reflects the inherent structural flexibility of the COPI coatomer complex, which exists in the cytosol as a dynamic, highly flexible assembled protein complex that is recruited to membranes as a heteroheptameric unit during vesicle formation (97, 98, 125–129). The proposed flexibility in COPI-derived vesicles is reminiscent of COPII-dependent vesicles, which can also expand in size to accommodate large cargos such as procollagen (130–132).

Furthermore, we also demonstrated that distinct proteinaceous bridging densities in the Golgi lumen and budded vesicles bridge HPV to the overlaying host membrane (Figure 4). Although the dimensions of these densities correspond closely to the CSR of the L2 protein, additional high-resolution structural studies are required for definitive identification. If the CSR of L2 forms part of the proteinaceous bridge, this would indicate that the CSR is not exposed to the cytosolic side of the Golgi membrane or the budded vesicle. How then can host cytosolic trafficking factors, such as the COPI coatamer, bind to a sequence motif in the CSR to promote trafficking? In the case of the budded vesicle where the COPI coat has disassembled from the membrane, it is possible that without COPI, the membrane-inserted form of L2—in which the CSR is exposed to the cytosol—destabilizes, undergoing “backsliding” so that the CSR is now topologically localized to the luminal space of the vesicle. Such backsliding has indeed been shown for L2 insertion across the endosome membrane. In this case, where the cytosolic retromer binds to membrane-inserted L2 to promote trafficking, depletion of the retromer triggered the retraction of L2 into the endosome so that L2 is no longer inserted into the membrane (57). Another alternative explanation is that, because there are potentially 72 copies of the L2 protein in a single assembled virion, not all copies of the L2 protein participate in similar interactions. In this scenario, some L2 proteins on a single HPV particle can be fully inserted to expose the CSR to the cytosol and recruit the COPI coat, while other L2 proteins on the same viral particle are in an alternative conformation that allows CSR to bridge HPV with the vesicle membrane, exposing only the C-terminus of the protein to the cytosolic side. This is clearly evident from our vesicle analysis, where we do not see a symmetric distribution of connecting densities (Figures 4D, F).

In sum, we report the first high-resolution molecular snapshots of HPV trafficking in its native cellular environment. In contrast to the accepted model, our results reveal the resilience of the virus capsid to disassembly during retrograde trafficking to the Golgi, suggest a coat-dependent budding mechanism for the generation of virus transport vesicles, and demonstrate that the lysosome does not degrade HPV but instead functions as a potential storage compartment for infectious HPV. These new insights should guide future efforts to identify novel targets for anti-HPV intervention.

Funding Statement

This work was supported by American Heart Association Grant 25PRE1361891 (R.S.); National Institutes of Health 5R01AI150897-05 (B.T.); National Institutes of Health 1DP2GM150019-01, and Klatskin Sutker Discovery Fund (S.M.); National Institutes of Health S10OD030275 and the Arnold and Mabel Beckman Foundation award to the University of Michigan Cryo-EM facility. The funders had no role in the study design, data collection, analysis, or the content and publication of this manuscript.

Data and materials availability:

Single particle cryo-EM map of HPV from infected host cell lysosomes has been deposited at the Electron Microscopy Data Bank (EMDB) with the accession code: EMD-72538. All materials used in the study are available upon request.