HPV18 Persistence Impairs Basal and DNA Ligand–Mediated IFN-β and IFN-λ₁ Production through Transcriptional Repression of Multiple Downstream Effectors of Pattern Recognition Receptor Signaling

This article is featured in In This Issue, p.2076

Human papillomaviruses (HPVs) are a large family of sexually transmitted DNA viruses that can cause benign and malignant lesions in humans [(1–4), https://pave.niaid.nih.gov/]. Although it has been known for quite some time that HPVs are able to evade the innate immune response and persist in the host, the molecular mechanisms regulating these critical events have only recently begun to emerge and still remain largely uncharacterized (5–7). Thus, gaining mechanistic insights into the immune escape by HPVs would allow us to better understand how these viruses can favor cancer progression.

Among the HPV family members, high-risk HPV (hrHPV) genotypes, especially HPV16 and HPV18, selectively infect human keratinocytes (KCs) in stratified epithelia of mucosa, leading to epithelial hyperplasia that can subsequently progress to cancer at different anatomical sites, such as the anogenital tract and oropharynx (8, 9). Because undifferentiated KCs express several pattern recognition receptors (PRRs), which are able to sense viral pathogens and promote the innate immune response (10–12), it is highly likely that hrHPVs have developed effective strategies to evade innate immunity by inhibiting PRR downstream signaling (13–16). In support of this hypothesis, a recent study by Lau et al. (17) has shown that E6 and E7 deregulation in transformed KCs antagonizes the cyclic GMP-AMP synthase (cGAS)–stimulator of IFN genes (STING) pathway. In particular, E7 was found to directly bind STING, thereby acting as a specific antagonist of the DNA-activated antiviral response. In addition, other investigators (18–22) have shown that E6 or E7 protein from hrHPV genotypes inhibits the transcriptional activity of IFN regulatory factor (IRF) family members. However, there are several caveats affecting the interpretation and generalizability of the aforementioned findings, such as the heterogeneity of the cell models used (e.g., cells of different origin, such as epithelial cells and fibroblasts; cells overexpressing only E6 and E7; cells transfected with episomal viral genomes; or transformed cells harboring multiple copies of the integrated viral genome) and the multiplicity of stimuli used to test host innate immunity. Thus, although it is clear that hrHPVs can selectively infect basal KCs and persist in the host, it still remains to be unequivocally determined how, in HPV-infected KCs, the physical status of the virus, different cell type–specific microenvironments, or different stimuli may affect the host innate antiviral response. Furthermore, whether HPV interferes with the expression of PRRs in HPV episome–containing KCs remains an open question. Thus, there is a significant gap in our knowledge of the pathogenic mechanisms of hrHPV in human KCs.

The IFN system constitutes the first line of defense against viruses in mammals. IFNs are categorized into three groups: type I (IFN-α/β), type II (IFN-γ), and type III (IFN-λ). Among them, IFN-λ is the most recently described group of small helical cytokines capable of inducing an antiviral state in responsive cells (23–26). Although type I IFNs act in most mammalian cell types, type III IFNs appear to primarily target mucosal surfaces, particularly epithelial cells of the intestine, liver, lung, and presumably skin (27–29).

In this study, we have deliberately chosen as our cell model the near-diploid, spontaneously immortalized human KC cell line (NIKS), which retains a normal response to contact inhibition, supports the full productive HPV life cycle, and provides an isogenic cell background on which to study virus–host interactions (30, 31). Using this cell model, which recapitulates the full viral life cycle of HPV, we show that NIKS cells harboring multiple copies of episomal HPV18 genomes fail to produce types I and III IFNs not only under differentiating conditions but also following exposure to salmon sperm (SS) DNA or poly(deoxyadenylic-deoxythymidylic) acid [poly(dA:dT)], two potent inducers of PRR signaling (32–35). Lastly, we report the existence of multiple evasion mechanisms relying on HPV18-mediated transcriptional inhibition of key components of the cGAS–STING and retinoic acid–inducible gene I (RIG-I) DNA–sensing pathways.

Overall, our findings provide novel insights into HPV18 immune-escape mechanisms in human KCs, with possible implications in cervical carcinogenesis.

NIKS cells (Stratatech) were cultured in the presence of J2 3T3 fibroblast feeders, as previously described (36). HeLa cells were grown in DMEM supplemented with 10% FBS (both from Sigma-Aldrich). HPV18 minicircle genome was produced as previously described (37). Briefly, for construction of minicircle viral genomes, a BglII site was introduced into the HPV18 genome after nucleotide 7473 and the minicircle vector pMC.BESPX were subcloned into this site. For the production of minicircles, Escherichia coli strain ZYCY10P3S2T was transformed and grown in Terrific Broth medium until an OD₆₀₀ of four to five was reached. An equal volume of induction mix (0.04 N NaOH and 0.02% l-arabinose in Lysogeny Broth) was added to induce recombination, and the culture was incubated for an additional 5 h at 32°C. Subsequently, plasmid DNA was extracted from bacteria and gel purified to obtain only the covalently closed circular DNA form of the viral genome.

NIKSmcHPV18 cells were obtained by nucleofection of NIKS cells with an Amaxa Nucleofector II with 2 μg of HPV18 minicircles, according to the manufacturer’s instructions, grown as pooled cells, and used from passages 20 to 30. Organotypic raft cultures were generated as previously described by Wilson and Laimins (38). Briefly, organotypic cultures were grown in specialized culture chambers on a collagen base, formed by mixing normal human neonatal fibroblasts with Collagen I Rat Tail (Sigma) in Ham’s F-12 medium containing 10% FBS and penicillin/streptomycin. NIKS cells were plated on the collagen base and after 15 d, raft cultures were harvested and fixed in 10% buffered formalin, embedded in paraffin, and cut into 5-μm sections for immunostaining analysis.

Poly(dA:dT) (1.25 μg/ml) and sheared SS DNA (1.25 μg/ml) (InvivoGen) were transfected into cells using Lipofectamine 3000, according to the manufacturer’s instructions (Thermo Fisher Scientific). Cells were treated with 10 μM carbonyl cyanide m-chlorophenylhydrazone (CCCP; Sigma-Aldrich) or DMSO for 30 min and then transfected for 24 h with poly(dA:dT). MG132 (Sigma-Aldrich) was used at 30 μM concentration for 8 h.

Whole-cell protein extracts were prepared and subjected to immunoblot analysis as previously described (39). The following Abs were used: rabbit polyclonal Abs anti-cGAS (product number HPA031700; Sigma-Aldrich, diluted 1:500), RIG-I (catalog number 06-1040; Merck Millipore, diluted 1:10,000), IFN-γ–inducible protein 16 (IFI16; C-terminal, diluted 1:1,000), IRF7 (catalog number sc-9083; Santa Cruz Biotechnology, diluted 1:200), and IRF3 (catalog number sc-9082; Santa Cruz Biotechnology, diluted 1:500); rabbit mAbs anti-IRF1 (product number 8478; Cell Signaling Technology, diluted 1:250) and p-STAT1 (product number 9167; Cell Signaling Technology, diluted 1:1,000); and mouse mAbs anti-STING (catalog number MAB7169; R&D Systems, 1:1,500), mitochondrial antiviral signaling protein (MAVS; catalog number sc-166583; Santa Cruz Biotechnology, diluted 1:200), and STAT1 (catalog number 610186; BD Biosciences, diluted 1:1,000). mAb against α-tubulin (catalog number 39527; Active Motif, diluted 1:4000) was used as a control for protein loading. Immunocomplexes were detected using sheep anti-mouse or donkey anti-rabbit Ig Abs conjugated to HRP (GE Healthcare Europe) and visualized by ECL (Super Signal West Pico; Thermo Fisher Scientific). Native PAGE was performed using Ready Gels (7.5%; Bio-Rad), as described previously (40). In brief, the gel was prerun with 25 mM Tris base and 192 mM glycine (pH 8.4) with 1% deoxycholate in the cathode chamber for 30 min at 40 mA. Samples in native sample buffer (20 μg protein, 62.5 mM Tris-HCl [pH 6.8], 10% glycerol, and 1% deoxycholate) were size fractionated by electrophoresis for 60 min at 25 mA and transferred to nitrocellulose membranes for Western blot analysis. Images were acquired, and densitometry of the bands was performed using Quantity One software (version 4.6.9; Bio-Rad). Densitometry values were normalized using the corresponding loading controls.

Quantitative real-time PCR (qRT-PCR) analysis was performed on a CFX96 Real-Time PCR Detection System (Bio-Rad). Total RNA was extracted using TRI Reagent (Sigma-Aldrich), and 1 μg was retrotranscribed using an iScript cDNA Synthesis Kit (Bio-Rad). Reverse-transcribed cDNAs were amplified in duplicate using SsoAdvanced Universal SYBR Green Supermix (Bio-Rad) for viral genes, as well as cellular genes. The glucuronidase β (GUSB) housekeeping gene was used to normalize for variations in cDNA levels.

Total cellular DNA was isolated with a QIAamp DNA Mini Kit (QIAGEN). A 600-ng DNA sample was digested with DpnI to remove the unreplicated input DNA. After digestion, 40 ng was analyzed by quantitative PCR (qPCR) using 500 nM primers and SsoAdvanced Universal SYBR Green Supermix (Bio-Rad). The reaction conditions consisted of a 30-s 95°C enzyme-activation cycle, 40 cycles (10 s) of denaturation at 95°C, and 10 s of annealing at 60°C. Copy number analysis was completed by comparing the unknown samples with standard curves of linearized HPV18 DNA. The GAPDH DNA copy number was used as an endogenous control. The specificity of the L2 primers was tested in nontransfected cells in which no amplification occurred. The primer sequences are detailed in Supplemental Table I.

The cytokines secreted in culture supernatants were analyzed using single analyte human ELISA kits for IFN-β (catalog number 41410; VeriKine-HS Human Interferon Beta ELISA KIT; PBL Assay Science) and IFN-λ₁ (catalog number DY7246; Human IL-29/IFN-lambda 1 DuoSet ELISA; R&D Systems), according to the manufacturers’ instructions. All absorbance readings were measured at 450 nM using a Victor X4 Multilabel Plate Reader (PerkinElmer).

Nuclear extracts were prepared using NE-PER Nuclear and Cytoplasmic Extraction Reagents (catalog number 78883; Thermo Fisher Scientific), according to the manufacturer’s instructions. IRF binding activity to IFN-β and λ₁ enhancers was measured using a Universal Transcription Factor Assay Colorimetric kit (catalog number 70501; Merck Millipore), according to the manufacturer’s instructions. In brief, 200 ng of biotin-labeled oligonucleotides containing the consensus sequence for the specific transcription factor under study were mixed with nuclear extract into each well of a streptavidin-coated microtiter plate. The bound transcription factor was detected with a specific primary Ab: anti-IRF1 (catalog number sc-497 X; diluted 1:400), anti-IRF3 (catalog number sc-9082 X; diluted 1:400), and anti-IRF7 (catalog number sc-9083 X; diluted 1:200; all from Santa Cruz Biotechnology). An HRP-conjugated Ab was then used for detection with TMB substrate. The intensity of the reaction was measured at 450 nM using a Victor X4 Multilabel Plate Reader. The following biotinylated oligonucleotides were used: IFN-β enhancer probe sense 5′-biotin ATGACATAGGAAAACTGAAAGGGAGAAGTGAAAG-TGGGAAATCCTCTG-3′ and IFN-β enhancer probe antisense 5′-CAGAGGAATTTCCCACTTTCACTTCTCCCTTTCAGTTTTCCTATGTCAT-3′, IFN-λ₁ enhancer probe sense 5′-biotin AGGGAGTTCTAAGGATTTCAGTTTCTCTTTCCTTCTTGATGCAGCTCCCA-3′ and IFN-λ₁ enhancer probe antisense 5′-TGGGAGCTGCATCAAGAAGGAAAGAGAAACTGAAATCCTTAGAACTCCCT-3′.

Southern blot analysis was performed as described previously (36). In brief, genomic DNA (10 μg) was digested with DpnI to remove any residual input DNA and with HindIII, which has no restriction site in HPV18, or with EcoRI, which has two restriction sites in HPV18 minicircles. The digested DNA was then separated on a 0.8% agarose gel, blotted, and hybridized with an HPV18 genome sequence-specific probe labeled with (α-[³²P])dCTP using Ready-to-go DNA Labeling Beads. The results were quantitated using a Personal Molecular Imager (PMI) System (Bio-Rad) equipped with Quantity One software.

Five-micrometer sections obtained from NIKSmcHPV18 organotypic raft cultures were processed for immunofluorescent analysis and DNA–fluorescent in situ hybridization (FISH), as previously described (41). The following Abs were used: anti- minichromosome maintenance-7 (MCM7; CDC47, MS-862-P; Neo Markers, diluted 1:200), anti-p16 (clone E6H4; Ventana Medical Systems), and anti-HPV α-genus E4 protein (PanHPVE4). FISH probe was generated using a Biotin Nick Translation Mix (Roche Diagnostics), according to the manufacturer’s protocol, with the HPV18 minicircle genome as a template. Images were acquired using a digital scanner (Pannoramic MIDI; 3DHISTECH). For the assessment of histological features, the slides analyzed by HPV18 E4, MCM7, or HPV18 DNA were disassembled and stained with H&E.

Chromatin immunoprecipitation (ChIP) assays were performed as previously described (36). Immunoprecipitation was performed with 3 μg of unmodified histone H3 (06-755), dimethyl-histone H3 (Lys4; 07-030), and dimethyl-histone H3 (Lys9; 07-441) Abs (all from Merck Millipore). Threshold cycle (CT) values for the samples were equated to input CT values to give percentages of input for comparison; these were normalized to the enrichment level of unmodified histone H3 for each cell line. The primers used to amplify STING, cGAS, and RIG-I promoters are detailed in Supplemental Table I.

All statistical tests were performed using Prism Windows 5.00 (GraphPad Software). The data are presented as mean ± SD. For comparisons between two groups, means were compared using a two-tailed Student t test; for comparisons among three groups, means were compared using one-way or two-way ANOVA with the Bonferroni posttest. Differences were considered statistically significant at a p value < 0.05.

First, we asked whether HPV18 replication per se would induce an antiviral response in KCs. For this purpose, we generated a human KC cell line (NIKSmcHPV18) stably harboring a high viral load of HPV18 episomal genomes (37, 42). These cells were cultured as pooled cells and used throughout the study between passages 20 and 30. NIKSmcHPV18 cells maintained episomal HPV18, as assessed by the representative Southern blot analysis shown in Fig. 1A. The slower migrating bands, seen in the DNA sample digested with the noncutter restriction enzyme Hind III, indicate the presence of concatemers (Fig. 1A). As expected for episomal-harboring cells, NIKSmcHPV18 formed low squamous intraepithelial lesions in organotypic raft cultures, as determined by enhanced expression levels of the cellular proliferation marker MCM7 in the suprabasal layers where E4 expression was also evident (Fig. 1B) (43). Viral load was measured by qPCR of total genomic DNA at various passages, along with the quantification of viral transcription from total RNA extracts. We measured mRNA expression levels of E6, E7, and E2 oncogenes, the last being a specific marker of episome-derived transcription. As shown in Fig. 1C, the viral load ranged from 200 to 60 copies per cell. Furthermore, cells cultured from passage 20 expressed much higher levels of E6 and E7 mRNA compared with E2 mRNA (Fig. 1D). Consistent with this viral mRNA expression pattern, cells were negative for p53 protein expression (data not shown). Despite the variations observed in viral load and viral mRNA expression levels, the Southern blotting pattern and the low squamous intraepithelial lesion phenotype remained unchanged between passages 20 and 30, and the results described hereafter were obtained at different passages with reproducible results.

Next, we measured mRNA expression levels of types I and III IFN genes, along with those of some proinflammatory cytokines, in NIKSmcHPV18 and NIKS cells. All IFN mRNAs, with the exception of IFN-α, were significantly downregulated in NIKSmcHPV18 cells compared with parental cells (Fig. 2A). Intriguingly, we observed a significant upregulation of the IL-6 gene product, whereas the other cytokines were only marginally affected.

To obtain a cell model that would more closely recapitulate the natural replication of HPV, we generated organotypic raft cultures using both NIKS and NIKSmcHPV18 cells and measured the mRNA expression levels of the same panel of genes described above. As shown in Fig. 2B, types I and III IFN mRNA levels were significantly downregulated in NIKSmcHPV18 cells compared with those of parental cells, indicating that HPV-mediated escape from the immune response correlates with inhibition of IFN gene expression. IL-18 mRNA expression levels were also significantly downregulated, whereas those of IL-6 and IL-8 were significantly upregulated.

We next asked whether NIKSmcHPV18 cells were still able to react to exogenous DNA ligands in terms of types I and III IFN production. Because the physical status of the virus (episomal or integrated) may generate variability in the innate immune response of epithelial cells, we included HeLa cells harboring an integrated HPV18 DNA in our analysis (44, 45). Cells cultured between passages 20 and 30 were transfected with SS DNA or the viral dsDNA analog poly(dA:dT), and total RNAs were isolated from cells at the 12-h time point, whereas their supernatants were collected after 24 h of treatment to allow enough time for lymphokines to accumulate in the medium. Transfection of NIKS with poly(dA:dT) or SS DNA increased the mRNA expression levels of all IFNs tested, with IFN-β and IFN-λ₁ being the most highly induced genes (Fig. 3A); in all cases, SS DNA was a less potent inducer than poly(dA:dT). Remarkably, DNA ligand–mediated induction of all IFN genes tested was dramatically reduced in NIKSmcHPV18 and HeLa cells compared with NIKS cells (Fig. 3A).

Next, we assessed the extent of IFN-β and IFN-λ₁ production at the protein level by ELISA. Consistent with the mRNA levels, IFN-β production in poly(dA:dT)- or SSDNA-transfected NIKSmcHPV18 cells was markedly downregulated compared with control cells (i.e., 80 and 83% reduction, respectively), whereas it was barely detectable in HeLa cells (Fig. 3B). Likewise, DNA ligand–mediated IFN-λ₁ production was significantly inhibited in both cell types compared with control cells [e.g., 68 and 46% reduction in NIKSmcHPV18 cells, and 89 and 98% reduction in HeLa cells transfected with poly(dA:dT) and SS DNA, respectively].

Lastly, in good agreement with previous findings showing hyperactivation of NF-κB transcriptional activity in KCs overexpressing hrHPVE6 and E7 (46–48), we found that NIKSmcHPV18 and HeLa cells transfected with DNA ligands displayed increased expression of IL-6, an NF-κB downstream target gene, at the mRNA and protein levels (Fig. 3A, 3C, respectively) compared with control cells.

Altogether, these findings clearly indicate that episomal HPV18 does not induce an antiviral innate immune response; KCs carrying episomal HPV18, as well as HeLa cells, respond poorly to exogenous DNA ligands in terms of types I and III IFN production compared with parental cells; and the HPV18 inhibitory activity does not seem to affect NF-κB function.

To gain more insights into the molecular mechanisms of HPV18-mediated regulation of types I and III IFN expression levels in NIKSmcHPV18 cells and HeLa cells, we used Western blot to measure the protein levels of various PRRs (i.e., cGAS, RIG-I, and IFI16), the adaptor molecules STING and MAVS, as well as the transcription factors IRF1, IRF3, and IRF7. Fig. 4A shows a schematic representation of the pathways involved in the innate immune response to exogenous DNA. cGAS, STING, RIG-I, and IFI16 were all very low in untreated NIKSmcHPV18 and HeLa cells compared with NIKS cells, whereas MAVS did not vary significantly (Fig. 4B, 4C). When we transfected these cells with poly(dA:dT) for 24 h, we observed a slight increase in cGAS expression in NIKS cells but not in NIKSmcHPV18 and HeLa cells. IFI16 displayed a dual expression pattern. Although it was downregulated in poly(dA:dT)-transfected NIKS cells, it was significantly upregulated in similarly treated NIKSmcHPV18 and HeLa cells. RIG-I induction by poly(dA:dT) was observed in KCs carrying episomal HPV18 and NIKS. As expected, STING expression levels were reduced after poly(dA:dT) transfection in NIKS cells, whereas they remained barely detectable in HPV-infected cells (49).

A recent report has shown that E7 is a potent and specific inhibitor of the cGAS–STING pathway, thereby hampering type I IFN production by DNA ligands in HeLa cells (17). Although not reported in that study, in this study we found that STING and cGAS are barely detectable or absent in NIKSmcHPV18 and HeLa cells (Fig. 4B, 4C), suggesting that one of the possible mechanisms by which HPV18 keeps antiviral factors in check is through downregulation of STING expression.

When we measured the expression of IRF1, IRF3, and IRF7 proteins (Fig. 4D–G), we made the following observations: IRF1, which was barely detectable in all untreated cells, was strongly induced in poly(dA:dT)-transfected NIKS and HeLa cells but not in NIKSmcHPV18 cells, reaching a peak in both cells at the 24-h time point; and IRF7 expression, which was very low in untreated NIKS cells, was strongly induced upon poly(dA:dT) transfection, with a peak at the 12-h time point. In contrast, IRF7 induction by poly(dA:dT) was completely ablated in HeLa cells and strongly delayed in NIKSmcHPV18, where it became evident only at the 24-h time point; and IRF3 protein expression, which was readily detectable in all untreated cells, did not vary following poly(dA:dT) transfection (Fig. 4F, 4G). We also observed IRF3 dimerization after poly(dA:dT) transfection in all cell lines, although dimer formation was slower in KCs harboring episomal HPV18 compared with NIKS cells (Fig. 4H, 4I).

Thus, it seems that the defects in types I and III IFN production observed in NIKSmcHPV18 cells after poly(dA:dT) transfection may be ascribed to multiple abnormalities in antiviral innate signaling pathways. In particular, the reduced availability of cGAS, STING, RIG-I, and IFI16 in HPV-infected cells, together with the lack of induction of IRF1 and IRF7, might provide the rationale for HPV18 immune evasion after DNA ligand stimulation.

Because poly(dA:dT) transfection was able to induce IFN-β, even in cells devoid of cGAS and STING (50), we asked whether downregulation of the polymerase III–RIG-I–MAVS signaling pathway activity by HPV could partly explain our observation that this stimulus failed to induce types I and III IFN production in NIKSmcHPV18 and HeLa cells but not NIKS cells. To this end, we first looked at RIG-I mRNA expression levels in mock- or poly(dA:dT)-transfected cells at different time points. Consistent with our previous data (Fig. 4B, 4C), basal RIG-I mRNA levels were reduced in NIKSmcHPV18 and HeLa cells compared with NIKS cells (20 and 98%, respectively) (Fig. 5A). Upon poly(dA:dT) transfection, RIG-I mRNA was quickly induced in NIKS cells at 3 h, reaching a peak at 12 h, whereas RIG-I started to increase in HPV⁺ cells only at the 6-h time point, and to a lesser extent throughout the time course. The same delayed kinetics was observed at the protein level: the protein became more evident after 6 h in HPV⁺ cells, whereas it was induced at the 3-h time point in parental cells (Fig. 5B). This delay in RIG-I induction in HPV⁺ cells might also explain the delayed formation of IRF3 homodimers in these cells after poly(dA:dT) stimulation (Fig. 4H, 4I). When we measured IFNs in the supernatants, we found that were rapidly released in NIKS cells, IFN-β at 6 h and IFN-λ₁ at 12 h, whereas they were induced at 12 and 24 h, respectively, in HPV⁺ cells (Fig. 5C).

Next, we asked whether the RIG-I–MAVS pathway mediated IFN induction in response to poly(dA:dT) transfection. To this end, we inhibited MAVS with the protonophore CCCP, which is capable of ablating RIG-I–like receptor (RLR) signaling through disruption of mitochondrial integrity (51). Consistent with a previous report (52), CCCP-treated NIKS cells remained viable and metabolically active throughout the 2-d-long experiment (data not shown). As expected, poly(dA:dT)-induced IFN-β and IFN-λ₁ production was markedly reduced in both NIKSmcHPV18 and HeLa cells compared with NIKS cells (Fig. 5C, 5D). CCCP treatment of NIKS cells led to a 2.5-fold decrease in IFN-β and IFN-λ₁ production, which nevertheless remained much higher than in DMSO-treated NIKSmcHPV18 and HeLa cells (Fig. 5D). Likewise, CCCP treatment of poly(dA:dT)-transfected NIKSmcHPV18 downregulated IFN-β and IFN-λ₁ by ∼1.5- and 1.8-fold, respectively, compared with DMSO-treated NIKSmcHPV18. IFN-λ₁ production was also reduced in HeLa cells by ∼2.6-fold compared with DMSO-treated cells, whereas levels of IFN-β remained consistently low. Thus, in HPV⁺ cells, where the STING pathway is apparently turned off, the amount of IFN produced upon poly(dA:dT) treatment seems to be primarily mediated by the RIG-I–MAVS pathway. Furthermore, the delayed kinetics of IFN production in these cells might be due to the unavailability of RIG-I under basal conditions. Lastly, reduced RIG-I mRNA levels in untreated HPV⁺ cells suggest that HPV acts as a RIG-I transcriptional repressor that is able to dampen the innate antiviral response during persistent infection. Likewise, basal mRNA levels of cGAS and STING were significantly lower than those seen in parental cells (i.e., 60 and 90% for cGAS and STING in NIKSmcHPV18 cells, and 64 and 65% in HeLa cells, respectively) (Supplemental Fig. 1).

Recent evidence indicates that STING transcriptional regulation is mediated by STAT1 binding (53, 54). In addition, STAT1 transcriptional activity was markedly inhibited in hrHPV-infected KCs (55), providing a possible mechanistic framework through which HPVs downregulate STING in host cells. Therefore, we assessed STAT1 expression at the mRNA and protein levels at baseline and after poly(dA:dT) transfection, as described above. In good agreement with previous findings, basal levels of STAT1 mRNA in HPV⁺ cells were reduced by 52% in NIKSmcHPV18 cells and by 92% in HeLa cells compared with NIKS cells. Furthermore, STAT1 expression was induced in all cell lines following poly(dA:dT) transfection, albeit to a much lesser extent in HPV⁺ cells (∼80% less than parental cells at the 12-h time point) (Fig. 6A). Consistent with the mRNA induction kinetics, total and phosphorylated STAT1 protein levels increased upon poly(dA:dT) transfection in HPV⁺ cells, with a delay of 6 and 12 h in NIKSmcHPV18 and HeLa cells, respectively (Fig. 6B).

To confirm that the downregulation of these proteins by HPV18 was occurring at the transcriptional level, we treated cells with the proteasome inhibitor MG132 for 8 h and assessed the protein levels of various PRRs and adaptor molecules, as described above. Even though the drug induced accumulation of ubiquitylated proteins, it did not promote any accumulation of the proteins analyzed (Fig. 6C).

Thus, our results indicate that viral immune escape in HPV⁺ cells is due to constitutive downregulation of at least two important cytoplasmic PRRs, cGAS and RIG-I, and the adaptor protein STING. Furthermore, low levels of STAT1 may explain why the basal expression of STING is reduced in HPV⁺ cells.

Broad regulation of the transcriptional competence of host cell chromatin has been reported in HPV-infected cells (56–59). To verify whether the transcriptional inhibition of the cGAS, STING, and RIG-I genes observed in HPV⁺ cells could also reflect changes in chromatin structure, we examined histone associations with the promoter region of the above-mentioned genes in HPV⁺ versus parental cells by ChIP assay. For this experiment, we chose dimethylation of histone H3 lysine 4 (H3K4me2) as a mark of actively transcribing genes and dimethylation of histone H3 lysine 9 (H3K9me2) as a mark of heterochromatin. We then performed ChIP assays using lysates from formaldehyde-fixed NIKSmcHPV18, HeLa, and NIKS cells, and two sets of PCR primers that could specifically amplify the promoter regions of the STING, cGAS, and RIG-I genes. The first primer set encompassed the promoter region where the putative STAT1 binding site is located (segment 1), whereas the second set was directed to a flanking region always within the promoter that included the transcription start site, which, in the case of cGAS, also included the putative IFN-sensitive response element binding site (segment 2) (Fig. 7A). As shown in Fig. 7B, HPV18 had little or no effect on the association of H3K4me2 (active chromatin) with segment 1 or 2 in all three promoters. In contrast, we observed a significant increase in H3K9me2 (repressive chromatin) bound to the two segments in all promoters from lysates of HPV⁺ cells versus NIKS cells. Interestingly, the binding levels of dimethylated H3K9me2 to the three promoter regions in NIKSmcHPV18 were 5–15-fold higher than those seen in NIKS cells for both segments. In HeLa cells, we detected even higher levels of H3K9me2 binding to segments 1 and 2 of the same promoter regions compared with those observed in NIKSmcHPV18 cells. The levels of H3K9me2 and H3K4me2 bound to gene segments located far away from the promoter region were comparable in all three cell lines, as well as in the GAPDH promoter region (Supplemental Fig. 2). Thus, HPV18 represses STING, cGAS, and RIG-I gene expression by promoting heterochromatin association with their promoter regions.

Because IRF family members displayed different temporal protein profiles in HPV⁺ cells versus parental cells upon DNA ligand stimulation (Fig. 4D, 4E), and we detected the concomitant induction of IRF3 homodimer formation (Fig. 4H, 4I), although it was significantly delayed in HPV-expressing cells, we sought to determine whether IRF species were transcriptionally active in these cells. For this purpose, we performed a sensitive quantitative ELISA-based assay using a biotin-labeled probe that spanned the tandem IRF binding sites in the IFN-β or IFN-λ₁ enhancer (Fig. 8A). Because identical results were obtained with both probes, only the set of panels for IFN-β is shown in Fig. 8B. IRF1, IRF3, and IRF7 all bound very efficiently to the immobilized probes in poly(dA:dT)-transfected NIKS, and their binding kinetics mirrored the changes in protein expression. Specifically, IRF1 binding activity was readily induced in poly(dA:dT)-transfected NIKS and HeLa cells (Fig. 8B), in a fashion consistent with the changes in protein expression (Fig. 4D, 4E). In contrast, NIKSmcHPV18 cells displayed low basal IRF1 binding activity, which remained basically unchanged throughout the entire time course following poly(dA:dT) transfection, in good agreement with the protein-expression kinetics shown in Fig. 4D and 4E. In contrast, induction of IRF3 binding activity by poly(dA:dT) was readily detectable at the 3-h time point and did not differ among cell lines (Fig. 8B). Lastly, induction of IRF7 binding activity was observed only after 12 h of poly(dA:dT) transfection of NIKS cells, whereas it was inhibited in similarly treated NIKSmcHPV18 and HeLa cells, mirroring the protein-expression kinetics shown in Fig. 4D and 4E. Thus, it appears that HPV can interfere with IRF DNA binding activity following DNA ligand stimulation in a cell type–specific fashion, thereby hampering the innate immune response in these cells (Fig. 9).

Escape from innate immune surveillance appears to be the hallmark of HPV infections (6, 7, 60). Although some mechanisms of immune evasion by HPVs, especially HPV16, have been characterized, they were mostly based on results obtained from KCs overexpressing only E6 and E7 or nonepithelial cells, thereby hampering data interpretation (17–21).

In this study, to better recapitulate the impact of HPV on its natural target cells (i.e., KCs), we have assessed the innate immune response in NIKSmcHPV18 cells, which are KCs carrying high numbers of episomal viral genome copies. These cells were used between passages 20 and 30, when the E6 and E7 transcripts were higher than those of E2, an expression pattern that is typical of persistent HPV infection. For comparison, we also included HeLa cells, which are cervical carcinoma–derived transformed cells harboring integrated HPV18 genomic DNA characterized by deregulated overexpression of E6 and E7 oncogenes (44, 45). We then used these cells to determine how persistent infection with HPV would affect their response to exogenous DNA.

Our findings demonstrate that KCs can maintain a high copy number of episomal viral DNA without triggering an antiviral response, because multiple points of the molecular pathways involved in the induction of types I and III IFNs are being inhibited. In this regard, we failed to detect any IFN production in KCs grown in monolayers or under differentiating conditions using organotypic raft cultures. Consistent with other reports, the NF-κB–dependent gene IL-6 was upregulated at higher levels in HPV⁺ cells compared with parental cells, indicating that the NF-κB pathway was functionally active in KCs carrying episomal HPV18, as well as in HeLa cells (46–48).

When we stimulated NIKSmcHPV18 cells with DNA ligands, we found that induction of IFN-β and IFN-λ₁ was significantly reduced compared with parental cells. Remarkably, cGAS, STING, RIG-I, and IFI16 proteins were all poorly expressed or almost undetectable in NIKSmcHPV18 cells compared with parental NIKS cells. Their suppression primarily occurred at the mRNA level rather than at the protein level. The observed increase in repressive heterochromatin markers at the promoter region of STING, cGAS, and RIG-I genes argues in favor of epigenetic silencing of these genes as a mechanism to stably repress key components of the innate antiviral response against DNA viruses.

Thus, altogether, our findings support a model whereby reduced expression of PRRs in HPV⁺ cells, along with that of the adaptor protein STING (61), which bridges most DNA receptors to downstream signaling events, creates an unreactive cellular milieu suitable for viral persistence, replication, and tumorigenesis (Fig. 9). In support of this model, human osteosarcoma U2OS cells, which are highly permissive to HPV replication, display a series of defects in innate immunity, including the absence of cGAS and STING proteins (37, 42, 62) (S. Albertini, I. Lo Cigno, F. Calati, and M. Gariglio, unpublished observations). Because all of these proteins are considered IFN-stimulated genes (ISGs), our findings are consistent with previous reports demonstrating that hrHPV genotypes inhibit a number of ISGs at the transcriptional level (22, 58, 63–65). However, these mechanisms differ from the evasion strategies reported for many other viruses that usually target PRRs and downstream molecules through posttranslational modifications, leading to increased protein degradation and temporary shutdown of the signaling cascade (66–69). These events usually take place at the early stages of infection. This discrepancy can be easily explained by the fact that we are dealing with a virus that displays an unusual life cycle, because it does not cause lytic infection; rather, it has evolved strategies to remain inside the cells for a very long time, can replicate without being recognized by innate sensors, and eventually promotes tumorigenesis (70). Thus, in our model of viral persistence, it is not unexpected that we found alternative strategies used by these viruses to keep the guardians in a prolonged inactive state. This inhibitory activity seems to be irreversible in the case of the cGAS–STING pathway, because we did not find any recovery of these proteins, even after treatment with exogenous DNA, whereas RIG-I protein expression was induced in response to poly(dA:dT) transfection and likely mediated the residual IFN production observed in both HPV⁺ cells (Fig. 9). Indeed, when we exposed poly(dA:dT)-transfected HPV⁺ cells to the protonophore CCCP, a known disruptor of RIG-I–MAVS signaling (51, 52), the levels of types I and III IFN in the supernatants were dramatically reduced. Furthermore, RIG-I upregulation was delayed in HPV⁺ cells and was accompanied by the induction of IFNs, indicating that this pathway could be restored and was responsible for the delayed antiviral response. Intriguingly, we found the same pattern of PRR inhibition and epigenetic modifications in NIKSmcHPV18 and HeLa cells, indicating that the evasion strategies are put in place at early stages of cancer progression and are maintained over time, even when the virus is fully integrated into the human genome, as in the case of HeLa cells.

Frequent suppression of cGAS and STING expression has been observed in many types of human cancer, suggesting that this pathway may play a major role in suppressing tumorigenesis and that its selective inhibition may occur frequently in viral-induced cancers (71, 72). In this regard, the cGAS–STING pathway is crucial in triggering a potent downstream IFN response against cytosolic DNA that is often present in cancer cells (73). Thus, inhibition of this signaling pathway by HPV18 is consistent with a model whereby infected cells escape the attention of the immune surveillance system, acquire further genetic mutations, and eventually become transformed. The observed inhibition of cGAS–STING signaling may also help to clarify why cells harboring hrHPV infection do replicate, despite activation of the DNA damage response, which ordinarily arrests cellular replication, also through activation of the innate response (74).

Although RIG-I was originally identified as a crucial cytoplasmic PRR for the recognition of many negative-strand RNA viruses, mounting evidence indicates that it also plays a role in detecting several DNA viruses (e.g., EBV, Kaposi sarcoma–associated herpesvirus, HSV-1, and adenoviruses), and in some cases it can recognize RNA species generated by RNA polymerase III, thus explaining the observed inhibition in HPV-infected cells reported in this article (32, 33, 75, 76).

In recent years, several intracellular DNA sensor candidates have been identified. Most of them appear to function through the essential adaptor protein, STING (11, 14, 66, 67). Although the functional relevance of some of these DNA sensors still needs to be fully established, cGAS and IFI16 have been identified as bona fide intracellular viral DNA receptors (77). In this article, we demonstrate that, in KCs stably maintaining episomal viral DNA, cGAS and STING expression levels are very low and are not induced by poly(dA:dT). Thus, the lack of the universal adaptor protein STING per se is sufficient to explain the absence of IFN induction during HPV infection, even though it still remains to be defined which DNA sensor is engaged by HPV. The IFI16 protein is a viral DNA sensor that could be a potential candidate for binding and recognition of the HPV DNA (78). Unfortunately, and despite many efforts, we have failed to demonstrate any IFI16–HPV DNA interaction (36) (S. Albertini, I. Lo Cigno, F. Calati, and M. Gariglio, unpublished observations). However, we found that IFI16 is downregulated under basal conditions in HPV⁺ cells as a mechanism to attenuate its activity as a DNA sensor or restriction factor (79).

When we turned our attention to the downstream transcriptional factors activated by the cGAS–STING and RIG-I–MAVS pathways, we made a series of interesting observations that helped us to further elucidate the complex modulation of these pathways during HPV infection. Although IRF3 protein levels were only marginally decreased by the presence of HPV18, IRF1 and, albeit to a lesser extent, IRF7 were reduced in NIKSmcHPV18 cells. Consistent with the reduced availability of the IRF1 and IRF7 proteins, their binding activities to the consensus binding sites present in β and λ₁ IFN enhancers were significantly reduced. In HeLa cells, IRF7 protein expression was almost undetectable, and it was not induced by DNA ligands, whereas IRF1 protein expression levels were less affected. According to the notion that IRF7 is a crucial factor for IFN-β production, the induction of IRF7 by DNA ligand was robust in NIKS cells, whereas it was strongly reduced and delayed in NIKSmcHPV18 cells and completely ablated in HeLa cells (80–82).

A partial limitation of our study is that we are not providing definitive mechanistic details underlying the defects in the innate antiviral system observed in HPV⁺ cells. Therefore, further studies are clearly needed to clarify, for example, how the oncoproteins E6 and E7 or other early genes contribute to this inhibition. Despite this limitation, one of the strengths of this study is represented by the establishment of a reliable cell model in which KCs stably harbor the entire viral genome, thereby closely recapitulating persistent HPV infection. Of note, similar results were obtained with HeLa cells, which are known to contain integrated HPV18 DNA. In addition, our findings provide valuable information about innate immunity in KCs, which is a process that is still poorly characterized, despite the fundamental role played by these cells in providing a physical barrier against infection and environmental insults, as well as in sensing viral pathogens, thereby initiating and shaping local immune responses.

Overall, our findings provide compelling evidence that HPV persistence in KCs leads to the inhibition of not only type I IFN but also type III IFN production in response to DNA ligands and that this effect is mainly due to the suppression of cGAS–STING signaling. As stated above, deregulation of STING signaling in cells with persistent hrHPV infection can hamper DNA damage response, enabling infected cells to evade host immunosurveillance and eventually become tumorigenic.

The fact that we used HPV-infected cells harboring the entire genome could explain some inconsistencies between our results and those of other investigators. In this regard, Lau et al. (17) have recently shown that E7 binds and degrades STING, thereby antagonizing the cGAS–STING DNA-sensing pathway. Our data imply that inhibition of STING activity occurs mainly at the transcriptional level rather than the posttranscriptional level. However, based on our data, we cannot rule out that both mechanisms might be involved.

In summary, a series of reports dating back to the first decade of the 2000s clearly documented that hrHPV can inhibit several ISG transcripts, mainly through E6 and E7 (22, 58, 63–65). In this study, we provide new evidence that the inhibitory action of HPV18 also affects some ISGs that are crucial for the innate antiviral response, such as PRRs and IRFs. In addition, production of IFN-β and IFN-λ₁ in response to poly(dA:dT) transfection was also impaired in CaSki cells harboring integrated HPV16 (data not shown) through inhibition of cGAS–STING signaling. Thus, our findings indicate that hrHPV genotypes have evolved broad-spectrum mechanisms that allow simultaneous depletion of multiple effectors of the innate immunity network rather than single downstream effectors.

These novel mechanistic insights into HPV immune evasion are critical for understanding how HPV can persistently infect steadily unreactive cells and promote cancer.

We thank Mart Ustav (Estonian Biocenter-Tartu) for providing the minicircle system for HPV18 genome generation and Marcello Arsura for critically reviewing the manuscript.

The authors have no financial conflicts of interest.