Vesicular trafficking permits evasion of cGAS/STING surveillance during initial human papillomavirus infection

Oncogenic human papillomaviruses (HPVs) replicate in differentiating epithelium, causing 5% of cancers worldwide. Like most other DNA viruses, HPV infection initiates after trafficking viral genome (vDNA) to host cell nuclei. Cells possess innate surveillance pathways to detect microbial components or physiological stresses often associated with microbial infections. One of these pathways, cGAS/STING, induces IRF3-dependent antiviral interferon (IFN) responses upon detection of cytosolic DNA. Virion-associated vDNA can activate cGAS/STING during initial viral entry and uncoating/trafficking, and thus cGAS/STING is an obstacle to many DNA viruses. HPV has a unique vesicular trafficking pathway compared to many other DNA viruses. As the capsid uncoats within acidic endosomal compartments, minor capsid protein L2 protrudes across vesicular membranes to facilitate transport of vDNA to the Golgi. L2/vDNA resides within the Golgi lumen until G2/M, whereupon vesicular L2/vDNA traffics along spindle microtubules, tethering to chromosomes to access daughter cell nuclei. L2/vDNA-containing vesicles likely remain intact until G1, following nuclear envelope reformation. We hypothesize that this unique vesicular trafficking protects HPV from cGAS/STING surveillance. Here, we investigate cGAS/STING responses to HPV infection. DNA transfection resulted in acute cGAS/STING activation and downstream IFN responses. In contrast, HPV infection elicited minimal cGAS/STING and IFN responses. To determine the role of vesicular trafficking in cGAS/STING evasion, we forced premature viral penetration of vesicular membranes with membrane-perturbing cationic lipids. Such treatment renders a non-infectious trafficking-defective mutant HPV infectious, yet susceptible to cGAS/STING detection. Overall, HPV evades cGAS/STING by its unique subcellular trafficking, a property that may contribute to establishment of infection.


Introduction
Human papillomaviruses (HPVs) are circular, double-stranded DNA viruses that infect and replicate in differentiating epithelium. HPV is the most commonly transmitted sexual infection, and oncogenic HPVs cause nearly 5% of all cancers worldwide and essentially all cervical cancers in women [1]. The HPV lifecycle and viral gene expression is dependent on differentiating epithelium [2][3][4]. To infect, HPV must traffic its circular dsDNA viral genome (vDNA) to the nucleus of basal keratinocytes, the only actively dividing cells within differentiating epithelium. Infected basal keratinocytes maintain vDNA episomes and serve as a reservoir for infected suprabasal cells that support the productive amplification of vDNA and assembly of progeny virions in upper epithelial layers [5,6]. Thus, persistently infected basal keratinocytes support sustained production of progeny virions to ensure efficient transmission within hosts. Notably, persistence of high risk HPV types is considered the risk factor for HPV-associated malignancies [7]. Unraveling the mechanisms that underlie HPV persistence is therefore important to understand both HPV lifecycle and cancer.
Structurally, HPV is a simple virus, yet it has evolved a unique mitosis-dependent mode of subcellular trafficking and delivery of vDNA to the host cell nucleus [8][9][10]. HPV virions bind to the extracellular matrix and cell surface heparan sulfate proteoglycans of basal keratinocytes, where upon proteolysis of the L1 and L2 capsid [11][12][13] and capsid conformational changes, the virion associates with entry receptor complexes and is endocytosed by the host cell [14][15][16][17][18][19][20]. As internalized virions travel through endosomal compartments, most of the L1 capsid is disassembled and degraded en route to the lysosome [9,21,22]. In complex with the vDNA, minor capsid protein L2 facilitates retrograde trafficking of the vesicular vDNA away from the degradative endo/lysosomal compartment to the lumen of the Golgi, where it resides during interphase [23][24][25][26][27]. L2 is an inducible transmembrane protein that can extend into the cytosol while remaining in complex with luminal vDNA by utilizing a transmembrane-like domain to protrude across local vesicular membranes to physically recruit retromer and other cytosolic sorting factors [8,[28][29][30][31][32].
Upon entry into mitosis, vesicular L2/vDNA traffics away from the fragmenting Golgi and accumulates on metaphase chromosomes via a chromatin-binding domain within the central portion of L2 [33]. This L2-dependent chromosomal tethering of vDNA-containing vesicles ensures that the vDNA will be partitioned to the daughter cell nuclei after nuclear envelope reassembly and mitotic exit into G1. Our prior work using an L2-BirA fusion virus to report on L2/vDNA compartmentalization within vesicular membranes suggested that L2/vDNA may penetrate the Golgi-derived limiting membranes upon chromosome binding during Perturbance of intracellular vesicular membranes during infection results in antiviral responses to HPV infection, suggesting that HPV's distinctive vesicular trafficking underlies its stealthy abilities.

HPV16 pseudovirus evades cGAS/STING responses during initial infection
We initially investigated cellular cGAS/STING responses to HPV16 pseudovirus (herein referred to as HPV) in HaCaT cells, spontaneously immortalized keratinocytes [82] that have a functional DNA-responsive cGAS/STING pathway [83][84][85]. HaCaT cells were either transfected with 250 ng of endotoxin free dsDNA plasmid (pGL3) for 90 min or infected with the HPV virion equivalent to 250 ng DNA (approximately equivalent to 950 ng L1 capsid, encapsidating pGL3 and non-specifically packaged cellular DNA). We wanted to ensure that cells were exposed to equivalent amounts of dsDNA during transfection and infection. Importantly, infection with increasing amounts of HPV results in a linear increase in luciferase activity (S1 Fig), suggesting that cells do not become saturated, thus allowing for the delivery of equivalent amounts of DNA by transfection and infection. H 2 O or viral storage buffer (VSB) were used as controls for pGL3 or HPV in these experiments. Cellular responses to DNA transfection were followed over the next 12 hr by western blotting. Since HPV infection is slow and asynchronous [14] with different DNA delivery kinetics compared to transfection, HaCaT cells were infected with HPV and cGAS/STING activity was assessed over an extended 24 hr time course. In these experiments, virus was present until the sample was collected for analysis. An acute and robust IRF3 phosphorylation of serine 396 (pIRF3) was observed in response to DNA transfection ( Fig 1A, left panel). In contrast, robust phosphorylation of IRF3 was not evident at any time point post-infection with HPV, indicating that the cGAS/STING pathway was not able to sense and respond to encapsidated vDNA (Fig 1A and 1B).
These HPV pseudovirions package pGL3, which contains a firefly luciferase expression cassette driven by an SV40 promoter. To gauge the relative pGL3 delivery between plasmid transfection and HPV infection, luciferase assays were performed 24 hr post-treatment. Since papillomaviruses lack specific genome packaging signals [86], HPV pseudoviruses generated in the 293TT system will package both pGL3 reporter plasmid and cellular dsDNA � 8kb in size. The majority of virions will contain 293TT-derived DNA, with a smaller fraction containing pGL3 [87]. For simplicity, the encapsidated DNA will be referred to as "vDNA" herein, even though it is not the authentic viral genome. To account for partial packaging of pGL3 and host DNA, the luciferase numbers for the HPV16 infections were corrected by factoring in the measured virion:pGL3 ratio of 11.16 (See Materials and Methods). Despite a complete lack of pIRF3 responses, HPV infection resulted in equivalent luciferase activity compared to pGL3 transfected cells (Fig 1C), arguing that differences in cellular DNA delivery do not explain the apparent evasion of cGAS/STING surveillance by HPV.
While immortalized HaCaT cells represent a good model for the basal keratinocytes that HPV has tropism for in vivo, we sought to investigate cGAS/STING responses in cultured primary human foreskin keratinocytes (HFKs), a better model for the cells HPV naturally infects. Primary HFKs were transfected with 500 ng pGL3 or infected with the viral DNA equivalent. As with HaCaT cells, transfection of cells with H 2 O or addition of VSB to cells were used as controls. pGL3 transfection resulted in robust activation of the cGAS/STING pathway, as seen by the detection of phosphorylated STING (pSTING) and pIRF3 (Fig 2). As before, HPV infection did not activate the pathway as detected by phosphorylation of STING or IRF3 during infection (Fig 2A). We observed a delay in activation between HaCaTs (peak near 90 min) and HFKs (4 hr). This may be due to differences in transfection efficiency between primary HFKs

PLOS PATHOGENS
HPV evades cGAS/STING surveillance and HaCaTs, or the kinetics of cellular activation may be slower in primary cells. To ensure comparable delivery of the 500 ng dsDNA via HPV virions compared to plasmid transfection we performed a luciferase assay 24 hr post-transfection/infection. As with HaCaT cells, HPV Statistics calculated by two-way ANOVA (P interaction = 0.0138) followed by Sidak's multiple comparison test ( ��� P < 0.005). (C) Cells were either transfected for 90 min with 250 ng pGL3 or infected for the duration of the experiment with an equivalent amount of HPV virions. Luciferase activity was measured 24 hr posttreatment, normalized to GAPDH, and corrected for pGL3 encapsidation as described in Materials and Methods. Differences were not significant (n.s.) by unpaired t-test, n = 3 biological replicates.

PLOS PATHOGENS
HPV evades cGAS/STING surveillance infection again yielded robust luciferase activity as compared to plasmid transfection of HFKs ( Fig 2B). These data further support a stealthy mode of HPV infection whereby cGAS/STING responses are avoided, despite effective delivery of encapsidated pGL3 DNA.

Transcriptional responses to DNA transfection and HPV16 infection of primary HFKs
We performed RNA-seq analysis to profile cellular transcriptional responses to dsDNA delivered via cationic liposomes or HPV pseudovirions (Fig 3, S1 Data, S2 Data). As above, HFKs were either transfected with 500 ng pGL3 or infected with the 500 ng DNA equivalent of HPV virions. Total RNA was extracted at the indicated timepoints. At 4 hr post-transfection, 142 genes (95 up and 47 down) were differentially expressed ( Fig 3A, S3 Data) in HFKs relative to mock (H 2 O complexed with transfection reagent) transfection (n = 2, P-adj < 0.05). The upregulated genes (Fig 3B), representing expected transcriptional responses to cytosolic DNA, were functionally enriched (see S4 Data for detailed results) for 82 Gene Ontology (GO) Biological Processes, including "defense response to virus" (GO:0051607, P-adj = 2.26 x 10 −15 ) as the top hit among innate immune-related terms. Of the 10 enriched Reactome Pathways, the
At 8 hr post-transfection, 469 genes (372 up and 97 down) were differentially expressed ( Fig 3A, S3 Data) in HFKs (n = 2, P-adj < 0.05). These up-regulated genes were also functionally enriched for IFN signaling, with the top five most significantly up-regulated genes ( Fig 3B) at 8 hr post-transfection including IFIT2, RSAD2, OASL, IFIT3, and HERC5. The genes with the highest fold change at 8 hr, ranging from~200 to 500 times above baseline, were CXCL10, CXCL11, IFNL1, GBP4, and IFIT2. Investigation of the heatmap (Fig 3C) identifies 3 main functional groups of genes ( Fig 3D). Genes that are upregulated acutely, at 4 hr (I; 'Immediate'/'Primary'), and either go back down ('Immediate') or remain up at 8 hr ('Primary'), genes that are upregulated by 4 hr and continue to increase throughout the experiment (II; 'Intermediate'), and a group of genes that are upregulated by 8 hr but were not significantly affected at the 4 hr timepoint (III; 'Secondary'). Complete gene sets for each cluster are provided in supplementary S5 Data.
While 500 ng of pGL3 delivered via cationic liposomes triggered a robust IRF3-based transcriptional response in HFKs, with the strongest overall response at 8 hr post-transfection, the equivalent amount of DNA delivered via HPV pseudovirus infection yielded no response ( Fig  3B, right panel). At 8 hr post-infection, there were no genes differentially expressed in HFKs relative to mock infection with VSB (n = 3). Importantly, preliminary analysis of the later 24 hr time-point did not show activation of IFN or ISG related transcription compared to VSB treatment (S2 Fig). Overall, compared to tranfection, HPV infection is unable to induce a classic innate immune response signature in primary HFKs (Fig 3B and 3D).

Bypassing HPV's natural trafficking pathway activates cGAS/STING
cGAS/STING is capable of sensing the incoming vDNA from a number of viruses including adenoviruses, poxviruses, herpesviruses, and the reverse-transcribed cDNA products of lentiviruses like HIV [62,63,88,89]. These viruses either breach (non-enveloped viruses) or fuse with (enveloped viruses) intracellular limiting membranes to provide vDNA-capsid complexes access to the cytosol and eventually reach the nucleus, or remain in the cytoplasm in the case of pox. In contrast, HPV uses the minor capsid protein L2 to transport vDNA within vesicular membranes en route to the nucleus during mitosis-dependent subcellular trafficking [8][9][10]. We hypothesize this unique vesicular trafficking enables evasion of cellular cGAS/STING surveillance.
To test this hypothesis, we stimulated premature membrane penetration of HPV vDNA to determine if cGAS/STING would then be capable of detecting HPV infection. As no HPV capsid protein mutations are known to cause leakage or transfer of vDNA across vesicular

PLOS PATHOGENS
HPV evades cGAS/STING surveillance membranes, we instead used cationic lipids which are known to perturb intracellular limiting membranes [90][91][92]. Indeed, our prior work using a sensitive enzyme-based L2 membrane penetration assay showed that even low amounts of cationic lipid (Lipofectamine RNAiMax) during HPV infection are capable of disrupting vesicular membranes that normally limit HPV from cytosolic exposure (see supplemental figure S2 in ref [27]).
HPV pseudovirions encapsidating pGL3 were mixed to the cationic lipid Lipofectamine 2000, as described in Materials and Methods. Transmission electron microscopy of this mixture revealed that intact HPV16 virions were in the close vicinity of submicron cationic lipid complexes ( Fig 5A). We next assessed the ability of wildtype HPV (wildtype L2 protein) and the post-Golgi trafficking defective L2 R302/5A mutant HPV to deliver vDNA to the nucleus in the presence of cationic lipids compared to control media. The R302/5A mutation in L2 renders particles non-infectious ( Fig 5B). Particles that contain R302/5A mutant L2 undergo normal early trafficking but fail to penetrate Golgi-derived vesicular compartments due to defective binding between the mutant L2 and host mitotic chromosomes [27,33,93]. R302/5A vDNA therefore remains cloaked within vesicular membranes, failing to reach the nucleus. Addition of cationic lipids caused a~60% decrease in wildtype HPV infectivity, as measured by luciferase assay for the delivery of virion-packaged pGL3 to the nucleus ( Fig 5B). In stark contrast, cationic lipids rescued the infectivity of R302/5A mutant virus to levels comparable to that of wildtype HPV in the presence of cationic lipids ( Fig 5B). The natural subcellular retrograde trafficking of vDNA is dependent on endosomal acidification, furin cleavage of L2, and γ-secretase activity. Small molecule compounds that perturb these processes are potent inhibitors of HPV16 infection [8]. The cationic lipid-dependent infectivity of R302/5A was insensitive to endosomal acidification inhibitor BafA, furin inhibitor dRVKR, or γ-secretase inhibitor XXI ( Fig 5C). Interestingly, the HPV16-specific monoclonal antibody H16.V5 but not the BPV1-specific B1.A1 was able to completely neutralize cationic lipid-dependent

PLOS PATHOGENS
HPV evades cGAS/STING surveillance infection of R302/5A. Collectively, these data indicate that although initial entry of intact particles into the endosomal compartment is still L1-mediated, cationic lipids mediate vesicular escape of HPV16 to rescue infectivity of R302/5A via an alternative route, independent of the classical HPV16 vesicular trafficking pathway.
Next we utilized our previously published L2-BirA system [27,94] to assess membrane penetration of virions in the presence of cationic lipids. This system works based on fusion of the biotin protein ligase BirA to the L2 capsid protein. BirA catalyses covalent addition of biotin to proteins tagged with a small peptide substrate-the "biotin acceptor peptide" (BAP). The system works as a compartmentalization assay to detect virus reaching the cytosol. Vesicular membranes prevent lumenal L2-BirA virus from contacting the cytosolic GFP-BAP substrate until the virus penetrates or crosses limiting vesicular membranes. Once the L2-BirA virus While naturally non-infectious, addition of cationic lipids restored infectivity of the R302/5A mutant to levels nearly comparable to those of wildtype HPV. Statistics calculated by one-way ANOVA (P < 0.001) followed by Tukey's multiple comparisons test ( �� P < 0.01, ��� P < 0.005), n = 3 biological replicates (C) R302/5A infection in the presence of cationic lipids is insensitive to biochemical inhibitors of endosomal acidification (BafA), furin (dRVKR), γ-secretase (XXI), and BPV1-specific Nab B1.A1, but is sensitive to HPV16-specific H16.V5. With exception of H16.V5 ( ��� P < 0.005), differences were not significant by one-way ANOVA followed by Dunnett's multiple comparisons test, n = 3 biological replicates.

PLOS PATHOGENS
HPV evades cGAS/STING surveillance penetrates, the BirA will be in contact with GFP-BAP substrate and GFP-biotin can be detected (Fig 5D). Our prior work showed that R302/5A mutant L2-BirA viruses were incapable of penetrating limiting membranes in this system due to the mutations in L2 [27]. Furthermore, penetration by wt L2-BirA virus is mitosis-dependent, and occurs only after proper Golgi trafficking and progression of cells through G2/M. Therefore, in a bulk infection of an asynchronous cell population, the earliest time GFP-biotin can be detected is 8-10 hr post infection [27].
We performed a brief timecourse infection with R302/5A mutant L2-BirA viruses, complexed with or without cationic lipid Lipofectamine 2000 as described above. Infection with R302/5A L2-BirA failed to generate any GFP-biotin within the first 6 hr, as expected. In contrast, the presence of Lipofectamine 2000 resulted in robust GFP-biotin signal by 4 hr and 6 hr (Fig 5E), indicative of premature membrane penetration caused by the cationic lipid Lipofectamine 2000.
Since the use of cationic lipids to bypass the natural subcellular trafficking pathway of HPV rescued infectivity of R302/5A mutant virions by perturbing intracellular membranes, we examined whether premature disruption of vesicular membranes by cationic lipids would allow for activation of the cGAS/STING pathway by HPV16. HaCaT cells were infected with R302/5A virions +/-cationic lipids and cGAS/STING activity was assessed by pIRF3 blot as before. Similar to wildtype HPV infection (Fig 1), R302/5A mutant infection alone did not activate the cGAS/STING pathway. However, premature vesicle disruption allowed for robust sensing of DNA delivered by R302/5A infection as measured by IRF3 phosphorylation, most prominently at 4 hr and 8 hr post-infection (Fig 5F and 5G). Similar results were seen in experiments using wt HPV16 particles (Fig 5H and 5I). These results indicate that when cationic lipids allow HPV16 virions to breach vesicular compartments, the cGAS/STING pathway is activated, generating a pIRF3 response.

Discussion
Cells are equipped with numerous PAMP-and DAMP-sensing PRRs, designed to detect the first signs of cellular stress and respond in an IRF3 and NFκB transcription factor-dependent manner to induce the expression of IFN, antimicrobial ISGs, and proinflammatory cytokine genes [95]. Geared towards recognizing dsDNA in the cytoplasm of a cell, the recently characterized cGAS/STING pathway has proven to be a critical arm of innate immunity and an important cellular antiviral defense system that is targeted for evasion and/or antagonism by many different viruses [51,52,96].
The HPV early proteins have been shown to block cGAS/STING and downstream IRF3-dependent IFN responses, as well as NFκB-dependent cytokine responses through a variety of mechanisms [70,71,[79][80][81]97,98]. The early viral proteins that counteract the cellular antiviral responses are not packaged within incoming virions. Thus, passive evasion mechanisms to limit early detection by the innate immune system during early infection would likely benefit later stages of the viral lifecycle. Viral blunting of these cell intrinsic and extrinsic responses likely promotes viral persistence by enabling maintenance of episomal vDNA and reducing inflammatory responses that cause activation of antigen-presenting cells and promote adaptive antiviral immune responses [99].
Here we investigate cGAS/STING responses to initial HPV infection. Although some prior work has reported minimal induction of IFN and downstream ISGs in response to HPV and canine PV infection [100,101], no studies have directly addressed cGAS/STING responses to incoming HPV virions. Importantly, other than the viral L1 and L2 proteins that are part of the incoming viral capsid, the pGL3 reporter plasmid does not express any papillomavirus

PLOS PATHOGENS
HPV evades cGAS/STING surveillance proteins. We find that while keratinocytes mount acute pIRF3 and downstream IFN/ISG responses to dsDNA plasmid transfection, dsDNA delivered through HPV infection proceeds undetected by cGAS/STING. To allow for direct comparison, we infected cells with the virion equivalent of 500 ng DNA, corresponding to a multiplicity of infection (MOI) of 400,000 virions/cell. These super-physiological conditions strengthen the hypothesis that a physiological infection is unlikely to be detected. The transcriptional response to dsDNA included type-I IFNB1, type-III IFNL1 and IFNL2, and many IFN-dependent ISGs and chemokines (Figs 3  and 4). Many of these ISGs, like IFIT1/2/3, ISG15 and CXCL10/11, can be upregulated directly through IRF3-mediated transcription, in addition to secondary IFN-dependent signaling (pSTAT1/pSTAT2/IRF9) mechanisms in certain cellular contexts [102][103][104].
Not surprisingly, many of these proteins have been implicated in the viral lifecycle. IFIT1 (p56/ISG56) has the potential to directly restrict persistent HPV infection via inhibition of E1-dependent episomal vDNA maintenance [77]. ISG15 encodes a ubiquitin-like molecule with broad antiviral activity. HERC5 is an E3 ligase that catalyzes the addition of ISG15 to nascent viral proteins leading to the restriction of HPV replication [105]. CXCL10, CXCL11, and related CXCL-family chemokines attract activated T cells via CXCR3 [106,107]. These chemokine responses may also be detrimental to HPV persistence, as evidenced by a recent mouse papillomavirus study where MmuPV1 was observed to specifically downregulate stress keratin-induced expression of CXCR3 ligands including CXCL10 [108]. IFI16 is a DNA-binding inflammasome component and can restrict HPV replication by epigenetically silencing viral gene expression reducing vDNA copy number [109,110]. Several of the genes upregulated by dsDNA transfection are counteracted by papillomavirus gene products. IRF1 and IRF7 gene products augment IFN/ISG responses. IRF1 activity is counteracted by the E7 protein from high risk HPVs [111,112]. Similarly, RSAD2 (viperin) has broad antiviral activity and is downregulated by cutaneous HPV2 E7 [113].
Double-stranded DNA transfection also induced cell survival and proliferation modulators like ATF3, CDKN2C, and TNFSF10. ATF3 interferes with E6's ability to degrade p53 by preventing E6AP from binding to p53, thus interfering with HPV immortalization [114]. Interestingly, ATF3 expression is downregulated in cervical cancer [114]. CDKN2C encodes p18INK4C, which restricts cellular proliferation through inhibition of CDK4 and CDK6 [115]. Expression of CDKN2C is upregulated by oncogenic E6 proteins [116], suggesting that the virus must be able to replicate in the presence of this anti-proliferative signal. TNFSF10 encodes TRAIL, a member of the TNF family of ligands that can initiate apoptosis [117]. The viral E5 protein counteracts TRAIL signaling to block apoptosis and promote HPV infection [118,119]. If activated, the induction of cellular effectors would likely restrict initial HPV replication following infection. Thus, initial evasion of cGAS/STING and these downstream cellular IRF3/IFN/ISG responses would likely be beneficial to long-term HPV infection. Collectively, our data suggest that the ability of the viral proteins to antagonize so many of these effectors is not to counteract sensing of the viral DNA upon initial infection. Rather, counteraction of these antiviral/antitumor responses is important for later steps in the viral lifecycle.
Interestingly, HPV16 infection is sensed when vesicular membranes are perturbed by the inclusion of cationic lipids during infection. These findings also indicate that the chromatinized encapsidated vDNA is insufficient to (completely) mask the viral genome from cGAS surveillance, as recent reports have shown that the nucleosomal components of mitotic chromatin prevents efficient cGAS activation by dsDNA [85,120,121]. We conclude that HPV's unique L2-dependent vesicular membrane trafficking effectively shields the vDNA PAMP from cytosolic cGAS/STING surveillance during nuclear transit, likely contributing towards viral persistence. A recent report describes a DNA-PK mediated DNA-sensing mechanism that appears to be specific to linear DNA [122]. Since the HPV genome is circular, it is unlikely that this pathway would be involved in the sensing of HPV infection. Furthermore, during production, HPV pseudovirus particles package linear 293TT DNA. During infection with wildtype L2 virions, we did not detect activation of STING of IRF3, suggesting that linear DNA is also shielded.
Our data suggest that papillomaviruses use subcellular trafficking to evade innate immune detection. Variation in subcellular trafficking routes has been shown to influence innate immune responses to other viruses. Different trafficking mechanisms during HAdV entry of several cell types have been shown to affect the stimulation of certain PRRs, influencing distinct immune responses that are dependent on particular virus-and cell-type combinations [123]. For example, different HAdV serotypes naturally vary in their receptor usage as well as downstream early vs. late endosomal trafficking pathways. These inherent differences in HAdV serotypes have been shown to affect innate TLR-dependent sensing of the virus [124]. For example, dynamin-2 (DNM2) is a cellular GTPase involved in microtubule-dependent transport and scission of endosomes. Modulation of DNM2 affects the trafficking of HAdV, resulting in an altered cellular cytokine response to infection [125]. Reovirus T1D and T3L strains elicit differing magnitudes of IRF3-dependent IFN responses in a manner dependent on viral uncoating during the late steps of viral entry and trafficking [126]. In the case of HIV, the CD4 receptor and the cellular dynein adapter BICD2 play roles in viral trafficking, influencing innate sensing and downstream IFN/ISG responses in infected cells [127,128]. Likewise, host phospholipase D affects innate sensing of the influenza A virus by modulation of entry [129].
Thus, while differing modes of trafficking can affect cellular responses to many viruses, HPV appears to have evolved an extremely covert means of evading cellular IRF3-dependent responses by hiding incoming vDNA inside the vesicular trafficking network. Other viruses may have evolved similar trafficking-dependent immunoevasion strategies. Adeno-associated viruses (AAVs) are parvoviruses that have recently been shown to retrograde traffic to the Golgi [130]. Although cellular pIRF3 responses to initial infection have yet to be investigated, it is tempting to speculate that they too may hide the genome from cytoplasmic DNA sensors. The polyomaviruses (PyVs) are dsDNA viruses that are structurally similar to papillomaviruses. PyVs also undergo subcellular retrograde transport after cellular uptake, but unlike HPVs, SV40 and the related BKPyV and JCPyV bypass the Golgi and traffic directly from endosomes to the ER, where redox-dependent chaperones loosen the VP1 capsid to expose membrane-interacting minor capsid proteins VP2/3 [131][132][133]. The post-ER fate of the PyV vDNA is unclear, with data suggesting cell-and virus-type specific differences in transport of vDNA from ER to cytosol/nucleus [134]. A recent study found a striking lack of pIRF3 and downstream IFN/ISG responses to BKPyV infection [135], similar to what we observe with papillomaviruses in the present manuscript.
Of note, these smaller viruses do not package tegument or core proteins that may be able to counteract cytoplasmic sensors or their downstream effector proteins. Indeed, these viruses, like papillomaviruses, need to avoid triggering these cytoplasmic sensors. This raises the possibility that multiple viral families may have converged on a trafficking-dependent immunoevasive strategy.

Tissue culture
HaCaT cells [82] and HaCaT-GFP-BAP cells that stably express cytosolic GFP-BAP [27] were grown in high glucose DMEM supplemented with 10% FBS and Ab/Am. An additional 200

PLOS PATHOGENS
HPV evades cGAS/STING surveillance ng/mL puromycin (EMD Millipore 508838) was used for maintaining stable expression in HaCaT-GFP-BAP cells. Cells were maintained at 37˚C with 5% CO 2 and passaged every 2-3 days. Murine J2 fibroblasts were grown in high glucose DMEM supplemented with 10% NCS, 1% penicillin/streptomycin and 1% L-glutamine. Cells were maintained at 37˚C with 5% CO 2 and passaged every 3-4 days. To create a feeder layer to support the growth of primary human foreskin keratinocytes, J2 fibroblasts were irradiated every 3-4 days. Cells were removed from continuing culture flasks, resuspended in media in a 15 mL conical, and irradiated with 6000 rads using a Gammacell 40 cesium-137 source as described [136]. Irradiated cells were plated at 1 million cells per 10 cm plate to later be used for support of primary keratinocytes. Primary human foreskin keratinocytes (HFKs) were derived from foreskin samples as previously described [137]. For continuing culture only, primary HFKs were plated on top of the irradiated J2 fibroblast layer. After trypsinizing or thawing, primary HFKs were grown in high glucose DMEM and F12 media supplemented with 10% FBS, 0.4 μg/mL hydrocortisone, 5 μg/mL insulin, 8.4 ng/mL cholera toxin, 24 μg/mL adenine, and 1% L-glutamine. One day after plating, the media was changed to the same as above but with 5% FBS, 10 ng/mL EGF, and 1% penicillin/streptomycin. The media was supplemented with 10 μM Y-27632 dihydrochloride (Chemdea CD0141) to prevent cellular differentiation and senescence [138]. Cells were maintained at 37˚C with 10% CO 2 and passaged every 3-4 days. Y-27632 dihydrochloride was omitted when HFKs were plated for use in experiments.

Nucleic acid transfections
HaCaTs were plated at 60,000 cells per well or HFKs were plated at 90,000 cells per well in a 24-well plate in 500 μL complete media. Cells were transfected with 250 or 500 ng endotoxinfree pGL3 using Lipofectamine 2000 (ThermoFisher 11668) in OptiMEM (Life Technologies). Complexes were made in the following manner: 50 μL OptiMEM was combined with pGL3 (either 250 or 500 ng) or water, and a separate 50 μL of OptiMEM was combined with 2 μL Lipofectamine 2000. Each solution was vortexed, combined, vortexed, and incubated for 15 min at room temperature, prior to dropwise addition of 100 μL of complex to the cells. At various timepoints post-transfection, keratinocytes were washed once with PBS and lysed in 1x RIPA lysis buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 1% NP40, 0.5% sodium deoxycholate, 0.1% SDS), supplemented with 1x reducing SDS-PAGE loading buffer, 1x protease inhibitor cocktail (Sigma P1860), 1mM PMSF and 1x PhosSTOP phosphatase inhibitor cocktail (Roche 04906845001). Samples were then boiled for 5 minutes at 95˚C and stored at -80˚C until gel electrophoresis.

HPV16 production
The R302/5A mutant was generated by site-directed mutagenesis of pXULL-based constructs using the QuikChange XL-II kit and verified by Sanger sequencing [27,139]. Luciferase expressing wt and R302/5A mutant HPV16 pseudovirions were generated as previously described [139]. Briefly, 293TT cells were co-transfected with a pXULL-derived plasmid expressing both L1 and L2. Transfected cells were harvested at 48 hr post-transfection by trypsinization and pelleted/resuspended in PBS + 9.5 mM MgCl 2 at 100 μl/10 cm plate, followed by the addition of Brij58 to 0.35%, ammonium sulfate (pH = 9.0) to 25 mM, Benzonase nuclease (Sigma E1014) to 0.3%, and 20U/mL exonuclease V (Epicentre E3105K) with overnight incubation at 37˚C to promote disulfide formation and maturation of capsids. After maturation, 0.17 volumes of 5M NaCl was added and lysates were 1x freeze/thawed to further break apart cellular structures. Lysates were cleared by centrifugation at 3000x g and supernatants were loaded onto discontinuous CsCl gradients made from 4 ml light (1.25 g/ml) CsCl

PLOS PATHOGENS
HPV evades cGAS/STING surveillance underlaid with 4 ml heavy (1.4 g/ml) CsCl, dissolved in virus storage buffer (VSB, 25 mM HEPES pH 7.5, 0.5 M NaCl, 1 mM MgCl 2 ). Virions were purified by 18 hr ultracentrifugation at 20,000 rpm at 4˚C in Beckman SW41 Ti rotor/buckets. Viral bands were visible slightly above the gradient interface and were collected by side puncture with a 1.5" 18 gauge needle and 5 ml syringe. CsCl was removed by washing and concentrating 3x in VSB using 100,000 MWCO centrifugation filter units (Sartorius VS04T42). Purified and concentrated virus was stored at -80˚C. Encapsidated pGL3 content (viral genome equivalent, vge) was determined by SYBR green qPCR (Thermo Fisher K0252) against a standard curve dilution series using primers qLuc2-A (ACGATTTTGTGCCAGAGTCC) and qLuc2-B (TATGAGGCAGAGCGA-CACC), specific for the luciferase gene in the pGL3 plasmid. DNA concentration of the purified virus was determined by measurement of OD260 on a nanodrop spectrophotometer. From this, the capsid content was calculated and the capsid:pGL3 ratios were determined. HPVs do not utilize specific vDNA packaging signals, and the 293TT method used to generate pGL3-containing HPV16 virions results in the promiscuous packaging of chromatinized cellular DNA, in addition to chromatinized pGL3 reporter plasmid [87,140]. The calculated packaging ratio of 11.16 was therefore used to normalize luciferase data in experiments comparing pGL3 delivery via cationic lipid transfection to wildtype HPV infection (Figs 1 and 2).

HPV16 infections
HaCaTs or HFKs were plated at 60,000 or 90,000 cells per well, respectively, in a 24-well plate in 500 μL complete media. Cells were infected the following day with wildtype or R302/5A mutant HPV16 virions at 250 ng or 500 ng vDNA equivalents per well for pIRF3 blotting experiments. Infections were performed at 2e8 vge/well for luciferase experiments. For infection experiments with cationic lipids, HPV16 virions were diluted into 50 μL OptiMEM, and a separate 50 μL of OptiMEM was combined with 3 μL Lipofectamine 2000. Solutions were vortexed, combined, vortexed again and incubated for 15 min at RT prior to addition of the 100 μL onto subconfluent HaCaT or HFK cells plated in 500 μL of the appropriate complete media. Nanopure H 2 O was substituted for Lipofectamine 2000 as a control in these experiments. For biochemical inhibition experiments the endosomal acidification and H + -ATPase inhibitor Bafilomycin A (BafA, Millipore 196000) was used at 100 nM, furin inhibitor decanoyl-RVKR-cmk (dRVKR, Millipore 344930) was used at 25 μM, and γ-secretase inhibitor XXI (XXI, Millipore 565790) was used at 200 nM, as previously described [27]. Neutralizing antibodies H16.V5 specific for HPV16, and B1.A1 specific for BPV1 [141], were provided in cell culture supernatants and used at 1:500 dilutions. Luciferase assays were performed 24 hr postinfection.

Luciferase assay
HPV infected or pGL3 transfected cells were washed once with PBS and lysed in 100 μL reporter lysis buffer (Promega E3971). Luciferase activity was measured on a DTX-800 multimode plate reader (Beckman Coulter) using Luciferase Assay Reagent (Promega E4550). A fraction of each lysate was blotted for GAPDH to normalize luciferase activity.

L2-BirA membrane penetration experiments
HaCaT-GFP-BAP cells were seeded in 24-well plates at 60,000 cells per well. The following day, cells were infected with 250ng L1/well of R302/5A L2-BirA virus [27], complexed +/-Lipofectamine 2000 cationic lipids or media alone, as described above. At indicated times post-infection, samples were processed for reducing SDS-PAGE followed by western blot to detect total and biotinylated GFP.

RNA-seq analysis
RNA was isolated from transfected and infected HFKs using Qiagen's RNeasy Mini Kit (Qiagen 74104). RNA was eluted in 60 μL nuclease-free water, further purified with a TURBO DNA-free Kit (Life Technologies AM1907), then shipped on dry ice to Novogene Corporation (Sacramento, CA). High-quality eukaryotic mRNA libraries were prepared and sequenced on an Illumina platform, yielding >20 million paired-end reads (2x150 bp length) per sample. High-throughput sequencing data were initially processed using the CyVerse Discovery Environment (www.cyverse.org), where FastQC was used for quality screening, HISAT2 [142] for mapping sequences to the human reference genome GRCh38, and featureCounts [143] for generating gene-level read counts (S1 Data). Differential expression analysis of read counts, including library-size correction and statistical testing (accounting for batch effects with the DESeq model, as specified in S2 Data), was performed using the DESeq2 Bioconductor package [144] implemented in R [145] via RStudio's integrated development environment [146]. Lists of differentially expressed genes (S3 Data), with an adjusted P-value < 0.05, were analyzed for functional enrichment (S4 Data) using g:Profiler [147]. Heatmaps and volcano plots where constructed in R using the pheatmap [148] and EnhancedVolcano [149] packages. Microsoft Office Suite and Adobe Illustrator software were used to create and compile additional figures.

Transmission electron microscopy
HPV16 samples were prepared with cationic lipids as described above. Each sample was applied to an ultra-thin carbon film over Lacey Carbon Support Film on 400-mesh copper grids (Ted Pella, Inc.) that were glow discharged for 1 min. Excess solution was blotted with Whatman #1 filter paper and the grid was rapidly stained with 2% uranyl acetate. The uranyl acetate was immediately blotted with Whatman #1 filter paper and rapidly stained with 2% methylamine tungstate. The second stain was immediately blotted away with Whatman #1 filter paper and allowed to air dry. Data were collected on a JEOL 3200FS microscope operated

PLOS PATHOGENS
HPV evades cGAS/STING surveillance at 300 kV. Images were acquired at a magnification of 87,000x and 160,000x, and at 1.0 μm underfocus using a Gatan UltraScan 4000 charge-coupled device (CCD) camera.

Statistics
Statistical analyses were performed using Prism 6 (GraphPad Software). Significant differences were determined by one-way ANOVA followed by Dunnett's or Tukey's multiple comparisons test or two-way ANOVA followed by Sidak's multiple comparisons test. An alpha of 0.05 was used to determine the statistical significance of the null-hypothesis.