Single cell transcriptomic analysis of HPV16-infected epithelium identifies a keratinocyte subpopulation implicated in cancer

Bedard, Mary C.; Chihanga, Tafadzwa; Carlile, Adrean; Jackson, Robert; Brusadelli, Marion G.; Lee, Denis; VonHandorf, Andrew; Rochman, Mark; Dexheimer, Phillip J.; Chalmers, Jeffrey; Nuovo, Gerard; Lehn, Maria; Williams, David E. J.; Kulkarni, Aditi; Carey, Molly; Jackson, Amanda; Billingsley, Caroline; Tang, Alice; Zender, Chad; Patil, Yash; Wise-Draper, Trisha M.; Herzog, Thomas J.; Ferris, Robert L.; Kendler, Ady; Aronow, Bruce J.; Kofron, Matthew; Rothenberg, Marc E.; Weirauch, Matthew T.; Van Doorslaer, Koenraad; Wikenheiser-Brokamp, Kathryn A.; Lambert, Paul F.; Adam, Mike; Steven Potter, S.; Wells, Susanne I.

doi:10.1038/s41467-023-37377-0

Download PDF

Article
Open access
Published: 08 April 2023

Single cell transcriptomic analysis of HPV16-infected epithelium identifies a keratinocyte subpopulation implicated in cancer

Nature Communications volume 14, Article number: 1975 (2023) Cite this article

8434 Accesses
4 Citations
41 Altmetric
Metrics details

Subjects

Abstract

Persistent HPV16 infection is a major cause of the global cancer burden. The viral life cycle is dependent on the differentiation program of stratified squamous epithelium, but the landscape of keratinocyte subpopulations which support distinct phases of the viral life cycle has yet to be elucidated. Here, single cell RNA sequencing of HPV16 infected compared to uninfected organoids identifies twelve distinct keratinocyte populations, with a subset mapped to reconstruct their respective 3D geography in stratified squamous epithelium. Instead of conventional terminally differentiated cells, an HPV-reprogrammed keratinocyte subpopulation (HIDDEN cells) forms the surface compartment and requires overexpression of the ELF3/ESE-1 transcription factor. HIDDEN cells are detected throughout stages of human carcinogenesis including primary human cervical intraepithelial neoplasias and HPV positive head and neck cancers, and a possible role in promoting viral carcinogenesis is supported by TCGA analyses. Single cell transcriptome information on HPV-infected versus uninfected epithelium will enable broader studies of the role of individual keratinocyte subpopulations in tumor virus infection and cancer evolution.

Single-cell RNA-sequencing of herpes simplex virus 1-infected cells connects NRF2 activation to an antiviral program

Article Open access 25 October 2019

The landscape of viral associations in human cancers

Article Open access 05 February 2020

Mutations in the HPV16 genome induced by APOBEC3 are associated with viral clearance

Article Open access 14 February 2020

Introduction

Infection with high-risk human papillomavirus (HPV) types causes 5% of all cancers worldwide, including cervical and oropharyngeal cancers¹. HPV16 accounts for the majority, with persistent keratinocyte infection being a primary risk factor^2,3,4,5,6. The role of the viral oncogenes in human carcinogenesis^{7,8,9,10,11,12} has been intensely investigated for decades, and the importance of differentiation for the viral life cycle is known^{13,14,15,16,17,18}. However, the transcriptomic landscape of heterogeneous stratified squamous epithelial subpopulations which support different phases of the viral life cycle has yet to be described, as bulk sequencing approaches^19,20 preclude a granular understanding of the distinct keratinocyte cell types comprising the host epithelium.

In this work, we report the transcriptomic landscape of isogenic HPV16-infected versus uninfected 3D stratified squamous epithelium²¹ and demonstrate its utility in discovering a keratinocyte subpopulation that is greatly amplified and reprogrammed by HPV, termed HPV-induced differentiation-dissonant epithelial nonconventional (HIDDEN) cells. This subpopulation forms an unexpected distinct superficial cellular compartment in HPV-infected epithelial models; the same compartment is not detected in their uninfected counterparts or in primary human oropharyngeal or cervical mucosa. HIDDEN cells persist through multiple stages of human carcinogenesis, including p16+ cervical intraepithelial neoplasia (CIN) lesions and HPV+ head and neck squamous cell cancers (HNSCC) (visual summary, Fig. 1a). TCGA analyses identify a signature of co-occurring biomarkers shared between HPV-infected and HPV-transformed tissues, and correlate high expression of these biomarkers with worse clinical outcomes. Using the HIDDEN cell transcriptome to identify candidate master regulators, we demonstrate that depletion of ELF3/ESE-1 in HPV+ epithelium greatly reduces HIDDEN cell biomarker expression and compartment formation. Altogether, our studies support an important role for this keratinocyte compartment in HPV-driven disease and for ELF3 as a potential therapeutic target in the fight against HPV persistence. The single-cell transcriptome data will serve as a key resource in the field of oncogenic viruses, wherein keratinocyte subpopulations can now be individually investigated for their roles in supporting viral life cycles and multistep cancer evolution.

**Fig. 1: Generation of isogenic HPV16+ versus HPV16- squamous epithelia.**

Results

scRNAseq reveals enrichment of distinct keratinocyte subpopulations in HPV16+ versus HPV16− stratified squamous epithelial rafts

For single-cell RNA sequencing (scRNAseq) experiments, we selected a well-established 3D organotypic epithelial raft model which recapitulates the stratified squamous epithelium required for differentiation-dependent HPV viral gene expression and replication (Fig. 1b). Normal immortalized keratinocytes (NIKS)²² and derivative, isogenic HPV16+ NIKS were used to generate HPV16− and HPV16+ epithelial rafts, respectively (Fig. 1c). Histopathological features of HPV16 infection were confirmed to be present by a board-certified pathologist in HPV16+ but not HPV16− rafts (n = 4 independent sets of rafts), including atypical epithelial cells with koilocytic features and mitotic figures within suprabasal epithelial cells. Genomic HPV DNA was uniformly present in HPV16+ but not HPV16− rafts by in situ hybridization (DNA-ISH, Fig. 1d).

Very few cells stained positive for viral episomes in terminally differentiated layers, demonstrating the rare occcurrence of the productive phase of the HPV viral life cycle²². This was further supported by RT-qPCR analysis, where early viral E1, E6, and E7 gene expression was more robust than late viral L1 gene expression (Fig. 1e), and by Southern blot analysis, which showed a modest increase in episomal HPV DNA in 3D rafts when compared to monolayer 2D cultures (Fig. 1f). Taken together, these data demonstrate persistent maintenance of infection in HPV+ rafts with rare cells supporting productive viral replication. Additionally, HPV16 viral activities were evident by increased expression of two established clinical markers of HPV+ tissues, exportin 5²³ (Fig. 1g) and p16²⁴ (Fig. 1h), known to be induced by HPV infection.

One matched pair of isogenic HPV16+ and HPV16− rafts was harvested on day 14 after keratinocytes were plated, separately dissociated into single-cell suspensions, and subjected independently to 10x Genomics Chromium RNA sequencing (Fig. 2a). Five additional sets of matched rafts were dissociated into single-cell suspensions and snap-frozen for validation studies. Following batch correction, the combined UMAP showed a high degree of overlap between HPV16+ and HPV16− cells (Fig. 2b), reflecting the known maintenance of the overall stratified squamous epithelial morphology upon HPV infection. HPV16 early (E1, E2, E5, E6, and E7) and late (L1 and L2) gene expression was confined to cells originating from HPV16+ rafts with a preponderance of cells expressing early viral genes (Fig. 2c), in alignment with previous RT-qPCR results (Fig. 1e). Further analysis identified twelve transcriptomically distinct clusters of keratinocyte subpopulations that were designated clusters 0–11 (C0–C11), with most clusters containing cells originating from both HPV16− and HPV16+ rafts (Fig. 2d and Supplementary Data 2). Interestingly, C9 was uniquely separated from the remaining 11 clusters (Fig. 2e) and was highly enriched in HPV16+ cells (72% of C9 cells, Fig. 2f). In contrast, four clusters (C0–C3) had relatively similar numbers of HPV16+ versus HPV16− cells, three clusters (C4, C5, and C7) were enriched in HPV16− cells, and two clusters (C10 and C11) had too few cells to characterize with certainty. In addition to C9, subpopulation C6 was also enriched in HPV16+ rafts (67% of C6 cells) and subpopulation C8 was exclusive to HPV16+ rafts (>95% of C8 cells).

**Fig. 2: scRNAseq of isogenic HPV16+ and HPV16– epithelium defines 12 distinct keratinocyte subpopulations.**

Keratinocyte subpopulations are primarily defined by the differentiation program of stratified squamous epithelium

Pseudotemporal trajectory analysis orders cells along a spectrum describing progression through a biological process²⁵. All clustered keratinocyte subpopulations were found to align along a single continuum of the squamous epithelial differentiation program (Fig. 3a), with the exception of C9. Although C9 was initially included in pseudotime analysis (trajectory Supplementary Fig. 1a, heatmap Supplementary Fig. 1b, cluster plot Supplementary Fig. 1c, and gene expression Supplementary Fig. 1d), its drastically divergent transcriptome results in non-contiguous positioning in the UMAP and thus it could not be included in robust pseudotime analyses. The exclusion from the conventional epithelial differentiation program represented by the pseudotime trajectory resulted in C9 cells being termed differentiation dissonant. This description was reinforced by low C9 expression of conventional markers of the layers of the epithelium (Supplementary Fig. 1d).

**Fig. 3: Pseudotime analysis aligns cells along the differentiation program of 3D stratified epithelium.**

In the final pseudotime analysis, gene expression signatures along the pseudotime trajectory progressed from basal (CO17A1+ and ITGB4+) to suprabasal (CARD18+ and KLK7+) cell states as evidenced by characteristic markers established in the literature (Fig. 3b). When clusters were plotted by pseudotime value (Fig. 3c), they grouped into populations with either more basal cell (C10, C7, C6, C4, and C5) or differentiated cell characteristics (C1, C8, C0, C11, C2, and C3).

Layer-specific epidermal markers routinely used at the protein level, namely COL17A and ITGB4 for the stratum basale, SBSN and DMKN for the stratum spinosum, CARD18 and DSG1 for the stratum granulosum, and KLK5 and KLK7 for the stratum corneum, were queried to define the cell clusters based upon RNA expression. Expression levels across pseudotime (Fig. 3d and Supplementary Fig. 1e) and in the UMAP (Fig. 3e and Supplementary Fig. 1f) confirmed that these known protein-level layer markers were captured in the scRNAseq analysis and that mRNA expression across rafts mirrored their protein-level progression from basal to terminally differentiated cell states. Proper laminar differentiation and recapitulation of stratified squamous layers in rafts was examined by IHC staining for protein expression of layer-specific markers in HPV16– and HPV16+ rafts (Fig. 3f). The protein staining pattern in HPV16− rafts mirrored normal human epidermis, demonstrating that 3D rafts recapitulate the cellular processes and architecture of the native in vivo tissue (Fig. 3g). Laminar cellular differentiation was also present in the HPV16+ rafts by protein expression analysis, consistent with the retained epithelial stratification demonstrated by histological analysis (Fig. 1c)^22,26. Of note, however, HPV16+ epithelium had less intense staining of differentiation markers (CARD18 and KLK5), demonstrating that HPV reprogramming suppresses the normal differentiation program^26,27,28,29, particularly in the uppermost layers of the epithelium.

Spatial mapping of suprabasal cell clusters reveals a yet undescribed epithelial compartment in HPV16+ rafts

By definition, scRNAseq data does not provide geographical localization of subpopulations in 3D tissues. We used RNA-ISH to spatially locate select keratinocyte subpopulations identified by scRNAseq within the 3D geography of HPV16+ and HPV16− rafts. Given the differential protein staining observed in suprabasal HPV+ vs HPV− rafts (Fig. 3f), the uppermost differentiated subpopulations were of particular interest. We hypothesized that differences in geographical distribution could be reflective of transcriptional reprogramming by HPV and also underlie protein-level changes observed in Fig. 3f. To this end, C3 was identified as the most terminally differentiated cell subpopulation (Fig. 4a) and selected for validation and spatial localization. While each scRNAseq cluster is defined cumulatively by a distinct transcriptomic gene list, we sought to identify “cluster-defining genes” that could serve as biomarkers to independently represent clusters without the need for more complex combinatorial approaches. Candidate cluster-defining genes were identified using three criteria: (1) Expression in at least 80% of cells in the cluster of interest. (2) Robust expression in rafts by RT-qPCR (CT <35) that was limited to the cluster of interest in UMAP feature plot visualization. (3) Availability of commercial RNA-ISH probes and potential for combinatorial detection based on channel availability.

**Fig. 4: Spatial localization of the terminally differentiated KLK7+ C3 keratinocyte subpopulation in stratified squamous epithelium.**

Using this strategy, candidate cluster-defining markers were identified for proof of concept in C2 (C10orf99 and DSG1) and C3 (KLK7, SPRR1A, RDH12, and POF1B) (Fig. 4a). C2 expressed known markers of the stratum spinosum (DSG1) and C3 expressed markers of the stratum granulosum and stratum corneum (POF1B and KLK7, respectively). Importantly, little overlap was observed between cells expressing DSG1 (C2) versus KLK7 (C3) by dual expression plots (Fig. 4b), suggesting that these clusters define distinct subpopulations in rafts. Similar expression levels of C2 and C3 cluster-defining markers were observed in HPV16+ versus HPV16− rafts (Fig. 4c), consistent with the comparable number of cells from these subpopulations in the single cell capture (Fig. 2f). Next, we determined dependence on differentiation of C2 and C3 by comparing the level of gene expression in differentiated specimens (overconfluent cells and rafts) relative to undifferentiated subconfluent cells (Fig. 4d). The data confirmed that, as expected, the terminally differentiated C2 and C3 keratinocyte subpopulations are largely absent in subconfluent monolayer cultures and the epithelial differentiation program is required for their formation. Altogether, we selected KLK7 as a cluster-defining marker for C3 (expressed in 90.6% HPV− C3 cells and 93.2% of HPV+ C3 cells) and termed the C3 population as the terminally differentiated KLK7+ epithelial subpopulation.

RNA-ISH spatial mapping and the C3 biomarker KLK7 were used to locate this differentiated subpopulation in rafts. We discovered an unexpected differential pattern of KLK7 localization in HPV+ versus HPV− rafts, indicative of de-regulation of suprabasal clusters in HPV16+ rafts. While KLK7+ C3 cells were localized to the uppermost layers of HPV16- rafts, HPV16+ rafts harbored an additional compartment above the KLK7+ cells (red bracket, Fig. 4e). These data are consistent with the decreased protein staining previously noted for CARD18 and KLK5 (Fig. 2f) in the superficial epithelial layer and suggest that HPV reprograms the differentiated epithelium to drive the emergence of an HPV-reprogrammed keratinocyte subpopulation. We sought to quantify the localization and thickness of the raft portion above KLK7+ cells in order to demonstrate robust upregulation of this compartment, specifically in HPV+ rafts. To this end, we measured raft thickness from the top of the raft to the top of the KLK7+ cell layer and normalized these values to total raft thickness in each case (Fig. 4f). Three biologically independent rafts were quantified. The KLK7+ cell layer largely extended to the uppermost surface of HPV− rafts (an average of only 5.9% of the raft remaining above), but not in HPV+ rafts where the top third of HPV+ rafts was composed of a distinct superficial compartment (an average of 28.5% of the raft above KLK7+ cells) (Fig. 4g). Thus, these experiments identified an epithelial compartment-specific to HPV16+ epithelium.

The C9 transcriptome defines a differentiation-dissonant keratinocyte subpopulation that forms the uppermost compartment of HPV16+ rafts

The C9 subpopulation was greatly enriched in HPV16+ rafts (Fig. 2f) and was uniquely distinct from the remaining 11 clusters by UMAP visualization (Fig. 2e) and in the initial pseudotemporal linear trajectory (Supplementary Fig. 1a), leading to its exclusion. Using our three-step strategy described above, we next identified and visualized candidate cluster-defining C9 biomarkers, which include ELF3, GPR110, and LCN2 (Fig. 5a and Supplementary Fig. 2a). For example, ELF3 was expressed in 100 and 96.6% of HPV16+ and HPV16− C9 cells (pct.1), respectively, and only 21.9 and 6.5% of remaining HPV16+ and HPV16− cells (pct.2), respectively. UMAP visualization confirmed high expressing ELF3 cells to be confined to C9. Relative gene expression by RT-qPCR showed increased expression of C9 markers in HPV16+ versus HPV16− rafts, in line with HPV16-driven enrichment of this population (Fig. 5b and Supplementary Fig. 2b). Additionally, expression of C9 cell markers was increased in models of differentiation (overconfluent cells and rafts), relative to undifferentiated (subconfluent) cells (Supplementary Fig. 2c). These Supplementary Data suggest that although the C9 subpopulation lacks conventional differentiation markers, it is nonetheless dependent upon the differentiation program and must therefore be studied in models of stratified epithelium. RNA-ISH for five C9 cluster-defining markers localized this population to the uppermost suprabasal compartment with robust staining in many cells of HPV16+ rafts; in contrast, HPV16– rafts exhibited weak staining in few cells (Fig. 5c). Quantification of total ELF3 and GPR110 RNA puncta showed increased expression of C9 markers in HPV+ vs HPV− rafts (Fig. 5d). The spatial localization of C9 biomarkers within the stratified squamous epithelium was determined by subdividing rafts into quintiles, from the most basal (Q5) to the most superficial (Q1) layer of the rafts. The distribution of ELF3 (Supplementary Fig. 2d) and GPR110 (Supplementary Fig. 2e) RNA expression was determined across quintiles for three independent rafts. ELF3 (Fig. 5e) and GPR110 (Fig. 5f) expression consistently increased in the suprabasal compartments of HPV+ rafts, most strikingly in the uppermost 20–40% of rafts. Protein levels were examined by IF to determine whether gene expression of C9 markers translates into protein expression in HPV16+ rafts. Consistent with the RNA-ISH results, ELF3 and LCN2 proteins localized to the uppermost compartment of HPV16+ rafts, whereas they were largely absent in HPV16− rafts (Fig. 5g). Altogether, we find that the uppermost suprabasal compartment of HPV16+ rafts lacks gene expression of conventional markers of differentiated stratified epithelia (Fig. 3f), in line with the previously observed suppression of the corresponding proteins (e.g., CARD18 and KLK5; Fig. 3f). Instead, this layer is populated by epithelial cells harboring the unique C9 transcriptome.

**Fig. 5: C9 cells occupy the superficial cell layers of the HPV16+ epithelium.**

The C9 transcriptome was further interrogated to identify candidate targets using pathway analysis and transcription factor enrichment tools. Reactome and KEGG pathway analysis revealed ontology hits, including viral infection, carcinogenesis, keratinization, and immune signaling (Supplementary Fig. 3a, b). Differential expression of HPV16+ versus HPV16− C9 cells showed upregulation of viral and bacterial pathogenesis and intercellular junctions (Supplementary Fig. 3c, d) and downregulation of the formation of the cornified envelope (Supplementary Fig. 3e). Together, these ontologies in HPV+ C9 cells indicate the importance for viral pathogenesis and squamous differentiation, albeit not classical differentiation, for the C9 compartment. To capture these features of the C9 transcriptome, we termed the C9 subpopulation HPV-induced differentiation-dissonant epithelial nonconventional (HIDDEN) cells.

HIDDEN cells are upregulated in HPV+ cervical rafts

To address the possibility that HIDDEN cell detection was an artifact of the NIKS model, we next used organotypic epithelial rafts engineered from primary cervical keratinocytes that were infected with high-risk HPV18 quasivirus versus mock-infected controls (Fig. 6a). HPV+ (versus HPV−) epithelium harbored HPV genomes by DNA-ISH (Fig. 6b) and many intensely stained ELF3+ suprabasal cells by RNA-ISH (Fig. 6c) and by IF (Fig. 6d). This demonstrates that (1) few ELF3+ cells are observed at baseline in the healthy cervical epithelium, (2) HIDDEN cells are highly upregulated by HPV in the cervical epithelium and (3) other high-risk HPV types such as HPV18 also induce the HIDDEN cell compartment.

**Fig. 6: HIDDEN cells are detected in HPV+ cervical rafts and in premalignant cervical lesions.**

HIDDEN cells are detected at various stages of HPV-driven cervical carcinogenesis

We next sought to evaluate the extent to which our findings in patient-derived cervical rafts translate to primary patient tissues at distinct stages of cervical intraepithelial neoplasia (CIN). CIN lesions, precursors to cervical squamous cell carcinoma (SCC), are ideal for studying HPV infection and progression toward cervical SCCs. Generally, CIN1 lesions harbor viral episomes, whereas CIN2 and CIN3 harbor a mix of episomal and integrated viral genomes with increasing dysplasia and progression to cervical SCC. However, the overall structure of basal and suprabasal epithelial components are retained in CIN1/2 lesions, but not in CIN3 lesions³⁰.

Cervical biopsies from patients suspected to have CINs were collected, processed, and screened for the high-risk HPV marker p16. H&E stained sections were blindly reviewed for HPV features and CIN lesions by a board-certified pathologist and selected specimens stained by IF for the HIDDEN cell biomarker ELF3. ELF3 protein expression is reported to be low to undetectable in normal ectocervical epithelial tissue according to the Protein Atlas³¹ and was correspondingly expressed at low levels in cervical tissue appearing morphologically normal and without evidence of HPV infection (Fig. 6e). Consistent with our in vitro models, we found that tissue suprajacent to p16+ basal cells demonstrated strong staining for the HIDDEN cell biomarker ELF3 (Fig. 6f). This included tissues with HPV features and CIN lesions regardless of grade. Typically, ELF3 expression appeared strongest in the uppermost tissue layer, mirroring the pattern observed in rafts. In some areas of tissues with CIN lesions, ELF3 staining was observed to be more intense in the middle compartment with less intense staining in more superficial cells (representative example Fig. 6f, CIN1). Nonetheless, in all p16+ tissues, HIDDEN cells were greatly amplified. Notably, while established HPV biomarkers (e.g., p16) reflect basal or proliferating cells which must be procured invasively, the ELF3+ HIDDEN cells instead demarcate superficial epithelial cells with the potential of the noninvasive collection.

HIDDEN cell biomarkers are elevated in HPV+ SCC tumors

Since HIDDEN cells were found to persist through all stages of premalignant cervical tissues (CIN1-3), we proposed that HIDDEN cells further persist from premalignancy into malignant squamous cell carcinomas. In order to evaluate this hypothesis, we analyzed the PanCancer TCGA to determine the expression of HIDDEN cell biomarkers in cancer. We initially focused on HNSCC patient cohorts to directly compare HPV+ versus HPV− tumors; the high prevalence of HPV in cervical SCCs precludes the ability to robustly compare patients by HPV status. Of the 488 HNSCC patients available for analysis, 9% had significantly elevated ELF3 expression (z ≥ 2.0) compared to the remaining patients. Importantly, this patient sub-group had statistically significant differences in HPV status, International Classification of Diseases (ICD)−10 classification (indicating anatomic location), chromosome 1q abnormalities, and hypoxia scores (Fig. 7a). ELF3 expression was higher in HPV+ vs. HPV− HNSCC (Fig. 7b, p < 0.0001), with ELF3 mRNA levels elevated in 24% of HPV+ HNSCCs (z-score ≥2.0) and above average in 75% (Fig. 7c). Conversely, in HPV− HNSCC patients, ELF3 mRNA levels were elevated in only 7% of cases and below average in 67% (Fig. 7c). We next examined ELF3 mRNA levels in HNSCCs grouped by ICD-10 (Fig. 7d). This analysis revealed that, although average ELF3 expression varied across anatomical locations (p = 0.015, it was consistently higher in HPV+ vs HPV− samples (p < 0.0001). As expected, the distribution of HNSCCs across subsites varied by HPV status: HPV+ HNSCCs predominated in subsites within the oropharynx (e.g., tonsils and base of tongue) while HPV− HNSCCs were more likely to occur in the larynx, lip/oral cavity/pharynx, or tongue. These associations build on our findings in rafts and patient tissue, supporting the presence of an ELF3-driven differentiated squamous subpopulation in HPV+ disease processes, including cancer.

**Fig. 7: The ELF3-dependent C9 cell population is associated with HPV+ head and neck and cervical SCC.**

The remaining significant clinical attributes (1q status and hypoxia) that varied by ELF3 expression were also evaluated. ELF3 mRNA levels correlated negatively with hypoxia scores (Supplementary Fig. 4a). This correlation was unexpected; ELF3 is a known HIF1A target³² shown to be increased with hypoxia in vitro³³. While hypoxia is a negative prognostic factor in HNSCC^34,35, its contribution to HPV+ vs HPV− HNSCC is complex. Tumor hypoxia does not appear to vary significantly by HPV status, yet HPV+ patients have improved therapy responses and survival^34,35. Thus, the observed correlation between ELF3 and hypoxia may be incidental, confounded by other variables, or maybe more complex in vivo than is currently understood. Lastly, since 1q status differed in HNSCCs with high (vs normal) ELF3 mRNA, we assessed whether differences in ELF3 mRNA expression could be accounted for by genome-level abnormalities and/or mutations. As expected^36,37, the overall mutational burden was lower for HPV+ HNSCC than for HPV− HNSCC (Supplementary Fig. 4b). Thus, differences in ELF3 mRNA expression between HPV− and HPV+ HNSCC are unlikely to be explained by mutations in ELF3. ELF3 mRNA levels were also higher in HNSCCs with a gain or amplification of ELF3 copy number, and lower in HNSCCs with a shallow loss (heterozygous) (Supplementary Fig. 4c). Within each of these categories, including normal diploid copy number of ELF3, the expression of ELF3 was higher in HPV+ than in HPV− HNSCC (Supplementary Fig. 4c, p < 0.001). Other HIDDEN cell biomarkers, such as MACC1 and GPR110, also had higher levels of expression in HPV+ HNSCCs (Supplementary Fig. 4d). Taken together, we conclude that higher ELF3 expression in HPV+ HNSCC cannot be explained by genome-level differences, but rather reflects HPV-driven mechanisms.

TCGA data refines the HIDDEN cell signature to core biomarkers shared between HPV-infected and transformed cells

Large HNSCC and cervical SCC patient cohorts in the TCGA dataset enabled us to refine the HIDDEN cell transcriptome to identify robustly co-occurring biomarkers for the detection of an analogous population in SCCs. We identified 652 genes in HNSCC and 175 genes in cervical SCC whose expression correlated with ELF3, using a Pearson cutoff of ≥|0.4|(Supplementary Data 5). Overlay of these two gene sets with the HIDDEN cell transcriptome yielded 42 shared genes, including ELF3 (Fig. 7e and Supplementary Data 5). A mutual exclusivity test determined that 17 of the 42 genes co-occurred with ELF3 in patients (Supplementary Table 1). Correlation of these 18 (including ELF3) co-occurring HIDDEN cell biomarkers with patient survival could not be analyzed in the HNSCC patient cohort due to the high variance in ELF3 expression across anatomical sites (Fig. 7d), with each anatomical site having either predominately HPV+ or HPV− tumors.

However, in cervical SCC patients, the high mRNA expression of the 18 co-occurring HIDDEN cell biomarkers was associated with significantly worse patient outcomes by Kaplan–Meier survival curves (Fig. 7f, p = 0.013). Taken together, these results from large patient cohorts further substantiate our findings in rafts and patient tissue samples, and provide strong evidence for the existence of a clinically significant epithelial subpopulation that promotes HPV+ carcinogenesis and/or therapy resistance.

The HPV+ HIDDEN cell signature is detected in published scRNAseq data of HNSCCs

In order to validate the above signature in an independent dataset, we analyzed published scRNAseq data of nine HPV− and six HPV+ HNSCC tumors³⁸. This report was exclusively focused on diverse immune cell populations and did not analyze epithelial cells. As described in Materials and Methods, all cells were jointly analyzed, followed by isolation and reclustering of epithelial-derived cells. Cells primarily clustered by tumor specimen as previously reported³⁸ (Supplementary Fig. 5). To address patient variability underlying this analysis, the batch correction was applied. The resulting clusters of all cells (UMAP by HNSCC HPV status, Fig. 8a, and by clusters, Fig. 8b) demonstrated clustering by cell type, rather than by specimen. Analysis of batch-corrected epithelial cells showed overlap between HPV+ and HPV− HNSCC tumors (Fig. 8c). Using the 18 gene core signature, cells were colored by the number of HIDDEN cell biomarkers expressed (Fig. 8d). Overall, epithelial cells from HPV+ HNSCC demonstrated greater expression than HPV− HNSCC tumors, with the highest number of co-occuring biomarkers concentrated in the C7 subset of cells (circle, Fig. 8d, e). To confirm that C7 was not derived from a subset of tumors, the proportion of C7 cells from each tumor was determined (Fig. 8f). We found that all but one (HPV− tumor HN5) contributed to C7 and furthermore that C7 was composed of a greater proportion of cells derived from HPV+ vs HPV− specimens. C7 shares approximately one-quarter of genes (89 of 336) with the original C9 gene list (Supplementary Data 2). Altogether, this analysis confirms the increased presence of HIDDEN cells using a published dataset of HPV+ versus HPV− HNSCCs.

**Fig. 8: HIDDEN cells are detected in HPV+ vs HPV− HNSCC scRNAseq data.**

HIDDEN cells are enriched in HPV+ tonsillar rafts and in HPV+ HNSCC patient biopsies

We next tested the presence of HIDDEN cells in HPV-infected and transformed head and neck tissue. The primary tonsillar epithelium was utilized as a model of HPV infection of the oropharynx, the most common site of HPV+ HNSCC development. Tonsil tissue (representative H&E stained section, Fig. 9a) harbored few ELF3+ cells at baseline (Fig. 9b). Tonsillar keratinocytes were derived from two separate donors, nucleofected with HPV16 episomes or maintained as negative controls, prior to generation of HPV+ and HPV− organotypic epithelial rafts (Fig. 9c and Supplementary Fig. 6a). DNA-ISH for high-risk HPV confirmed the expected HPV status (Fig. 9d). RNA-ISH for ELF3 and GPR110 (Fig. 9e and Supplementary Fig. 6b) and IF for ELF3 (Fig. 9f and Supplementary Fig. 6c) validated significant upregulation of HIDDEN cells with HPV infection in tonsillar rafts derived from both donors.

**Fig. 9: HIDDEN cells are detected in HPV+ tonsil-derived epithelium and in HNSCC patient biopsies.**

Next, HNSCC patient biopsies were probed for gene and protein-level expression of HIDDEN cell biomarkers. Many cells with multiple puncta were detected in two cases of p16+ HNSCCs by RNA-ISH but were largely absent in two cases of p16-negative HNSCCs (Fig. 9g). Similarly, ELF3 protein was robustly detected in HPV+, but minimally in HPV−, HNSCCs with HPV status confirmed by p16 staining (Fig. 9h). These Supplementary Data Suggest HIDDEN cells persist through stages of HPV-driven carcinogenesis into SCCs. Altogether, we show enrichment of HIDDEN cell biomarkers in patient-derived 3D in vitro models of cutaneous, cervical, and tonsillar epithelium, as well as in HPV-driven premalignant and malignant patient biopsies that reflect stages of cancer evolution.

ELF3 is a potential master regulator of HIDDEN cells

The C9 gene list (Supplementary Data 2) was next utilized to identify candidate master regulators of HIDDEN cells. The HIDDEN cell transcriptome was initially analyzed by Transfac for putative shared transcription factor binding sites (Supplementary Fig. 7). Each candidate transcription factor motif was plotted based on statistical significance, with multiple motifs identified for some transcription factors (Supplementary Fig. 7, genes above dotted line). These potential master regulators included HPV gene targets (e.g., the E2F family as a target of E7), challenging candidates for inhibition studies due to ubiquitous expression in many tissues (e.g., MAZ³⁹ and SP1⁴⁰), and genes important for basal keratinocyte functions (e.g., KLF family)⁴¹ that were not expressed exclusively in HIDDEN cells. Such characteristics suggested a lack of specificity and, ultimately, safety for the targeting of these candidates to suppress the HIDDEN cells and eliminated consideration for further mechanistic studies. ELF5, another member of the ETS transcription factor family with an identical binding site to ELF3^42,43, was identified as a potential master regulator in this analysis. ELF3 is a known regulator of at least one other verified HIDDEN biomarker, LCN2⁴⁴, and, importantly, its expression was limited to C9 by UMAP visualization (Fig. 5a). We, therefore, selected ELF3 for further analysis. The list of transcription factors was additionally evaluated for binding sites, specifically within ELF3, representing potential upstream regulators. These hits, indicated by red arrows (Supplementary Fig. 7), included NF-kB, a known regulator of ELF3⁴⁵.

Advanced analysis of promoter regions of the C9 gene list were next performed using two complementary methods: HOMER⁴⁶ to assess motif enrichment and RELI ChIP-seq peak enrichment⁴⁷ to estimate the significance of transcription factor binding intersections in these regions against a library of 11,054 functional genomics datasets. These analyses yielded a ranked list of potential transcription factors that may drive HIDDEN cell formation. The top 25 results were manually reviewed (Supplementary Data 6). Importantly, ELF3 was in the top 97th percentile of hits (#22 of 752, Fig. 10a summary, Fig. 10b ELF3 motif). Thus, ELF3 was independently identified as a top candidate transcriptional regulator of the HIDDEN cells.

**Fig. 10: The HIDDEN cell compartment requires ELF3 expression.**

ELF3 knockdown suppresses the HIDDEN cells in HPV16+ epithelial rafts

To test whether ELF3 is required to sustain HIDDEN cells, ELF3 was knocked down in HPV16+ NIKS and used to generate rafts. H&E staining of vector control and ELF3-depleted rafts showed a more uniform presence of the superficial granulosa and keratinized corneal cell layers that was variable across rafts, being most prominent in the ELF3sh-2 knockdown rafts (Fig. 10c). ELF3 knockdown was confirmed by IF staining for ELF3 (Fig. 10d) and by western blot analysis (Fig. 10e). Total ELF3 protein levels were elevated in HPV+ versus HPV− rafts and greatly reduced in rafts generated from knockdown cells (Fig. 10d, e). Notably, ELF3 knockdown almost completely eliminated ELF3+ HIDDEN cells in the uppermost compartment of HPV+ rafts, suprabasal to the K10+ layer (Fig. 10f). Consistent with the loss of the ELF3 cellular compartment, there was reduced expression of genes co-expressed with ELF3 in the HIDDEN epithelial subpopulation including GPR110, CEACAM6, MACC1, LCN2, and PSCA (Fig. 10g). These data demonstrate that the ELF3 transcription factor is required for the formation and/or maintenance of the suprabasal HIDDEN cell compartment in stratified HPV16-infected epithelium.

Discussion

In this study, we used scRNAseq to define keratinocyte subpopulations in HPV16+ versus HPV16− epithelium, and mapped select epithelial subpopulations within the geography of 3D stratified squamous epithelium. We chose the well-established HPV16-isogenic NIKS organotypic epithelial raft model^48,49,50 that mimics squamous epithelial morphology with basal and successively more differentiated suprabasal epithelial layers present in human squamous tissues⁵¹. NIKS rafts have been extensively used to uncover mechanisms of squamous differentiation^52,53, HPV maintenance and replication^48,54,55, viral gene expression^49,56,57, and HPV-induced carcinogenesis^58,59,60. This manipulatable isogenic model system captures the effects of HPV infection on epithelial biology in the absence of confounding variables such as host-dependent immune and environmental exposures to discover cell-autonomous mechanisms critical for viral persistence.

Although widely used for studies of HPV, important limitations of this system include the immortalized nature of the NIKS, a duplication of the long arm of chromosome 8, the relative absence of productive viral amplification, which was reflected in our scRNAseq data, and a male origin from human foreskin. While high-risk HPV-associated penile cancer can originate in the foreskin, the incidence of penile cancer is rare when compared to cervical or head and neck cancer. We, therefore, also included HPV+/- isogenic primary tonsillar and cervical organotypic epithelial rafts to validate key findings, thereby extending our data to primary mucosal rafts that are relevant to cervical and head and neck SCC.

The scRNAseq analysis described here identifies 12 transcriptomically distinct keratinocyte clusters, which now enables studies to precisely define the molecular reprogramming of HPV+ compared to HPV− naïve epithelium. Batch correction aligned analogous keratinocyte subpopulations to highlight differences in relative abundance and presence of specific subpopulations and to enable direct differential expression analysis between HPV-infected and uninfected cells per epithelial cell subtype. HPV reprogramming of each epithelial subpopulation forming the conventional basal and suprabasal epithelial layers could thus be assessed.

Using the scRNAseq atlas, we identified differences in the composition of the suprabasal compartment of HPV+ epithelium. Namely, we found that HIDDEN cells typically form the surface layers of HPV16-infected rafts and are marked by the expression of a broad array of genes implicated in stemness (e.g., KRT15^61,62 and KLF5⁴¹), terminal differentiation (e.g., TJP3⁶³), and carcinogenesis (e.g., CEACAM6)^64,65,66. These findings extended to primary human models of the cervical epithelium (+/-HPV18) and tonsillar epithelium (±HPV16), indicating the importance of HIDDEN cells in high-risk HPV-driven diseases at multiple epithelial subsites. Although HIDDEN cells were differentiation dissonant in the pseudotime trajectory (Supplementary Fig. 1a), their localization within the superficial layers of stratified squamous epithelium (Fig. 5e, f) and formation in differentiated but not undifferentiated models (Supplementary Fig. 2c) support dependence of this compartment on a differentiated environment as a fundamental part of the HPV life cycle. We further identified the epithelium-specific ESE (ETS) transcription factor ELF3 as an HPV effector that is required to form or sustain the HIDDEN compartment. Elevated ELF3 activity induced by HPV16 may be reflective of possible links to cancer development or progression.

Despite a substantial body of research on HPV infection, there is a paucity of studies that specifically analyze epithelial cells in HPV+ vs HPV− HNSCCs which would have identified the C9/HIDDEN cell signature. The existing literature is limited by the frequent exclusion of epithelial cells prior to scRNAseq, by unknown HPV status, and by the overwhelming focus on immune-related processes. Identification of the epithelial HIDDEN cell transcriptome will now enable mechanistic and functional studies in the field.

Detection of HIDDEN cells upregulated in p16+ CINs and HPV+ HNSCCs implies that these cells are maintained and continuously upregulated by HPV throughout the progression from HPV-infected epithelium to dysplastic preneoplasia and cancer. TCGA analyses demonstrated that high ELF3 expression levels were associated with HPV+ head and neck cancers regardless of the subsite, and co-occurring HIDDEN cell biomarkers were preliminarily associated with worse outcomes in cervical cancer. Outcomes by ELF3 expression could not be similarly determined in HNSCC TCGA data due to the paucity of HPV+ and HPV– specimens occurring at the same anatomical site. Thus, the association between HIDDEN cell upregulation and clinical outcomes requires additional study. Nonetheless, while a functional role for HIDDEN cells in human SCC development is not yet clear, enrichment of this subpopulation in high-risk HPV-infected organotypic rafts, primary tissues, and published scRNAseq experiments supports pro-carcinogenic activities.

Multiple scenarios could be at play for a hypothesized role for HIDDEN cells in worse outcomes for HPV+ cancers. HIDDEN cell effects on basal cells might occur via paracrine activities that create a pro-tumor microenvironment, and particularly in severely dysplastic lesions where ELF3+ HIDDEN cells appear to expand downwards into the basal layer (Fig. 6f). Such activities may act on aspiring malignant cells to promote proliferation or survival, or may reprogram the tumor microenvironment to promote angiogenesis or suppress immune surveillance. Alternatively, the HIDDEN keratinocyte signature harbored nonconventional differentiation markers, including select mucins and enzymes involved in o-linked glycosylation (Supplementary Fig. 3a). HIDDEN cells might therefore provide a physical barrier against exogenous anti-viral or anti-tumor responses. Lastly, HIDDEN cells featured gene markers characteristic of primitive stem/progenitor cells, such as KLF5 and K15. It is, therefore, also possible that these cells harbor quiescent stem-like features which might convert to a proliferative state during tumor evolution. Such potential roles remain to be identified.

Altogether, our studies (1) report the transcriptomic landscape of high-risk HPV-infected stratified epithelium, (2) discover epithelial HIDDEN cells that form a compartment and have the potential to mark the surface of HPV-infected tissue, (3) identify ELF3 as a required HPV-induced transcriptional driver of the HIDDEN compartment, and (4) detect and associate HIDDEN cell enrichment in HPV+ SCCs with poor prognosis.

Methods

All studies comply with all relevant ethical regulations. All research protocols involving human subjects were subjected to review by Institutional Review Boards at the University of Cincinnati, Cincinnati Children’s Hospital Medical Center, or the University of Arizona, as detailed in the relevant study methods below, and received approval for the study protocols.

Monolayer cell culture

HPV+ and HPV− near diploid immortalized keratinocytes that form a skin (NIKS) were provided by the Lambert laboratory, confirmed with STR profiling, and cultured on irradiated murine fibroblasts (J2-3T3)⁵¹. These cells were either used to generate organotypic rafts or were harvested at subconfluency (40–50% confluency) or overconfluency (48 h past confluency)⁶⁷. Mycoplasma testing was performed regularly and cell lines were confirmed to be negative. Cells were provided by the Lambert lab and confirmed by STR profiling.

Generation of 3D organotypic rafts from NIKS

3D organotypic epithelial rafts were generated using validated protocols⁶⁸. Briefly, 1 × 10⁶ NIKS were plated on porcine collagen matrix harboring embedded murine J2-3T3 fibroblasts. After 4 days, rafts were transitioned to an air-liquid interface and high calcium medium to generate stratified epithelium. The medium was then changed every other day for 14 days, at which point rafts were harvested, as appropriate, for subsequent analysis. For embedding, 3D rafts were fixed in 4% PFA for 30–60 min at room temperature, dehydrated, embedded in paraffin, and sectioned at 5 µm thickness onto acid-treated SuperFrost Plus glass slides (Cardinal Health, Dublin, OH, USA M6146-PLUS). After physical removal of the underlying collagen layer, rafts were rapidly dissociated using 0.25% trypsin and mechanical disruption, quenched with medium, filtered through 35 µM cell strainers (Corning, Corning, NY, USA 352235) to remove debris, and confirmed to be a single cell suspension with viability above 70% as assessed by Trypan Blue staining.

Immunohistochemistry (IHC) and IF

Antibody information, including dilutions and validations, are detailed in Supplementary Data 1. Tissue sections were deparaffinized and antigen retrieval was performed in 1X RNAscope Target Retrieval Reagent (Advanced Cell Diagnostics, Inc., Newark, CA, USA 322000) in a Decloaking chamber (Biocare Medical, Pacheco, CA, USA DV2004MX) for 15 min at 110 °C and 6 psi. IF staining was performed⁶⁷. For IHC, the Vectorstain Immunodetection kit (Vector Laboratories, Burlingame, CA, and HOH-3000) was used with modifications^69,70,71. In brief, following antigen retrieval, slides were treated in 0.3% hydrogen peroxide for 30 min to eliminate endogenous peroxidase activity. Non-specific antibody binding was inhibited by incubating sections with a serum-free protein-blocking solution. Tissue sections were incubated with primary antibody for 1 h at room temperature or overnight at 4 °C (p16 only). A secondary biotinylated multilink antibody was applied for 1 h, followed by streptavidin-peroxidase incubation for 30 min. Using freshly prepared 3.3’-diaminobenzidine (Vector Laboratories, Burlingame, CA, SK-4100) and DAKO Liquid DAB Substrate-Chromogen solution as a chromogen, the enzymatic reaction was visualized. Slides were counterstained with hematoxylin.

RNA-ISH

mRNA expression was detected using the RNAscope® 2.0 Assay (Advanced Cell Diagnostics, Inc., Hayward, CA, USA)⁷². Selected probes for human tissue are detailed in Supplementary Data 1.

RNA-ISH quantifications

For quantification, Z-stacks were converted from. nd2 to. ims format using the Imaris converter software and opened in Imaris. To define the organotypic raft height, the Imaris “surface creation tool” was used to manually demarcate the top of the organotypic raft and spot creation tool was used to indicate the bottom of the organotypic raft, the average height of the raft was measured using the shortest distance from the surface (top of the raft) to the bottom of the raft. The spots tool was used to outline the region above (cells <5 KLK7 puncta) the KLK7 expressing cells. The average distance was measured using spot statistics for the “shortest distance to the surface” (top of the raft) to the top of the KLK7 expressing cells. This distance was plotted as a percentage of the total raft height.

To quantify individual ELF3 or GPR110 RNA-ISH signals, the height of the raft was determined as described above. The raft height was divided into five equal quintiles, where Q5 was the bottom of the organotypic raft and Q1 was the top. The Imaris “spots” detection tool was then used with an estimated “XY diameter” for spot detection of 0.5 μm to enable the measurement of puncta. “Spots” were created for both ELF3 and GPR110 channels. The number of spots in each quintile was quantified per frame, with ten frames representing each raft. The total number of spots was summed per frame and reported, as well as the breakdown of spots per quintile to indicate spatial localization.

Quantitative RT-PCR

RNA was isolated using the RNAeasy Mini Kit (Qiagen, Valencia, CA, USA) in parallel sets of HPV16- and HPV16+ NIKS cell pellets from 2D cultures, cells from dissociated rafts, and intact rafts after removal of the underlying collagen. For intact rafts, tissue was consecutively disrupted manually using an 18 and then 21-gauge needle (Thermo Fisher Scientific, Waltham, MA, USA, 305167, 305196). The quality of extracted RNA was confirmed by NanoDrop to have 260/280 and 260/230 ratios between 1.8–2.0. A 1 µg aliquot of each sample was reverse transcribed into cDNA using the QuantiTect Reverse Transcription kit (Qiagen, Valencia, CA, USA, 74106) according to the manufacturer’s instructions. Primers specific to each gene (Supplementary Data 1) were designed using Primer-BLAST⁷³ and purchased from IDT(IDT, Coralville, IA). RT-qPCR was performed on an Applied Biosystems Step One Plus real-time PCR Stem with StepOne software v2.2 (Applied Biosystems, Carlsbad, CA, USA). Assays were performed in accordance with the manufacturer’s instructions. RT-qPCR data were analyzed by calculating the ΔΔCt of genes of interest relative to the housekeeping gene β-actin.

Southern blot analysis

Total genomic DNA was extracted from parallel sets of HPV16− and HPV16+ NIKS cell pellets from 2D cultures and 3D rafts using Qiagen’s DNeasy Blood and Tissue kit. A 5 µg DNA aliquot was digested with Bam HI to linearize the HPV16 genome for the purpose of determining viral genome load. Digested genome sequences were electrophoresed on a 0.8% agarose gel along with an HPV16-containing plasmid that had been digested with Bam HI to release the viral genome as a standard. After electrophoresis, DNAs were transferred to the Hybond N+ membrane (Amersham). The membrane was then probed with a set of 20 oligos complementary and specific to HPV16 that had been labeled with γ−32P-ATP. To visualize HPV16 DNA, the washed membrane was exposed to a PhosphorImager screen that was then scanned using a Typhoon laser-scanning platform (GE Healthcare).

DNA-ISH for HPV

HPV DNA was detected by ISH by the CCHMC Pathology Research Core⁷⁴. In brief, a chromogenic ISH test for high-risk HPV genomes, including HPV16, was performed (Ventana Medical Systems). Sections were incubated with a fluorescein-tagged DNA probe (Enzo Life Sciences, Farmingdale, NY, ENZGEN1146000) and counterstained using the fully automated Ventana BenchMark instrument.

Single-cell RNA sequencing

For 3′ scRNAseq, matched HPV16− and HPV16+ rafts generated from isogenic NIKS lines were harvested on day 14 post-plating, dissociated into single cells separately but in parallel, and subjected to 10x Genomics Chromium genomic sequencing. A total of 3833 HPV16− and 4275 HPV16+ dissociated cells were included in the analysis. There was an average of 20,619 reads per cell from HPV− rafts and 29,537 reads per cell from HPV+ rafts. Five sets of biological replicates were dissociated in parallel, snap-frozen, and stored at −80 °C for subsequent validation. HPV16− and HPV16+ raft transcriptome datasets were then combined for unsupervised clustering using the Seurat R package^75,76. Fastq files were processed through the 10x pipeline, CellRanger v3.0.2, to obtain a gene expression matrix using a custom-generated human genome, hg19, that contained annotated sequences for HPV16. To maximize cell quality, cells expressing <1000 unique human genes and/or >20% mitochondrial genes were excluded from subsequent analysis. The gene expression matrix was normalized to 10,000 molecules per cell and log-transformed. The batch correction was performed by integrating the Supplementary DataSet based on anchors, and shared cell types between the two Supplementary DataSets. This was done by using the FindIntegrationAnchors and IntegrateData functions using the first 20 dimensions⁷⁷. The most highly variable expressed genes were used for principal component analysis. The most significant principal components were used for UMAP dimension reduction. Cell clusters were determined by shared nearest-neighbor using the Louvain algorithm. Unbiased clustering was performed separately for each sample and for the combined samples to facilitate comparisons of subpopulations in HPV16+ and HPV16− rafts. Cluster gene lists (Supplementary Data 2) were determined by Seurat’s FindAllMarkers functions using the Wilcoxon rank-sum test⁷⁷. Genes that were uniquely expressed in each cluster were considered cluster-defining—these genes were expressed in the majority of cells in the cluster and minimally expressed (<30% of cells) or absent in all other clusters. A subset of these genes were selected for further analysis using quantitative RT-PCR and RNA-ISH. The scRNAseq data were deposited at the Gene Expression Omnibus (GEO) under accession number GSE189670.

Pseudotime, including slingshot plot and individual gene pseudotimes

All cells were subjected to diffusion map dimension reduction and were labeled with the previously defined clusters from the UMAP plot. Monocle v2.22.0^25,78,79 was used to order cells in pseudotime. The original cell clusters from the Seurat pipeline were preserved to perform differential expression and subsequently order the cells. Analysis was performed with C9 cells included and excluded prior to ordering the cells. Cell cycle normalization was applied. The top 100 most significantly variable genes were plotted on a heatmap in pseudotime using scaled expression by row.

Strategy for identifying cluster-defining genes

Detection of individual genes by RNA-ISH is a time- and cost-intensive assay. We, therefore, developed a strategy to narrow down candidate cluster-defining markers prior to RNA-ISH validation. Genes for RNA-ISH validation were selected based on (1) Expression in at least 80% of cells in the cluster of interest (percentage 1, pct.1) with minimal expression in remaining cells. (2) Robust cluster-specific expression in UMAP feature plots and in raft samples by RT-qPCR. (3) Availability of commercial RNA-ISH probes, and potential for combinatorial detection based on technical considerations, including channel availability. To this end, the gene list for each cluster was filtered by pct.1 ≥ 80% and sorted in increasing order by expression in the percent of remaining cells (percentage 2, pct.2) (Supplementary Data 3). The top 20 genes per cluster were considered; feature plots were visualized to confirm robust cluster-specific expression for each gene, RT-qPCR primers were designed for candidate genes, and expression was validated (CT <35) using RNA from monolayer cell extracts and subsequently raft extracts (Supplementary Data 3). Candidate cluster-defining genes were also evaluated by RT-qPCR for relative expression in HPV16− versus HPV16+ rafts and compared to expected relative ratios based on the scRNAseq data. Additionally, RNA expression was determined in subconfluent cells and overconfluent cells in 2D culture as well as in 3D rafts in order to determine the effect of differentiation and stratification. RNA-ISH validation was then performed in HPV+ and HPV− raft tissues for select cluster-defining genes in order to spatially localize clusters in 3D tissues. This validation of RNA-ISH probes for cluster mapping was performed in a minimum of three independently generated sets of HPV16+ and HPV16− rafts to ensure reproducibility.

Analysis of published scRNAseq supplementary datasets

All individual samples from the published dataset³⁸ were combined and normalized individually by SCTransform v2⁸⁰. Raw data are available on NCBI Sequence Read Archive: accession ID SRP301444. Processed gene barcodes are available on the Gene Expression Omnibus database: accession ID GSE164690. Samples were integrated based on anchor genes and identified by reciprocal PCA using the first 30 dimensions. Dimension reduction and clustering on the integrated Supplementary DataSet was performed as described above. Epithelial cells were further isolated by only using the samples that were from the CD45 negative samples and removing non-epithelial cell types, T-cells, endothelial, stromal, macrophage, and B-cells. Reclustering was performed as above.

Ontology analysis

Functional enrichment analysis and conversions of gene lists was performed using the gProfiler web server (https://biit.cs.ut.ee/gprofiler/gost)⁸¹. The server includes Gene Ontology (GO) terms, curated by the Ensembl database⁸², and pathways from KEGG (https://www.genome.jp/kegg/), WikiPathways (https://www.wikipathways.org/index.php/WikiPathways), and Reactome (https://reactome.org/). The Gene Ontology Statistics (GOSt) function was used to perform GO enrichment analysis of the input gene lists (cluster-defining genes/differentially expressed genes (DEGs)). GOSt uses multiple testing corrections by comparing the number of GO terms, KEGG pathways, etc., against the given input query. In addition, Benjamini–Hochberg FDR values for the GO terms were also included; FDR values less than 0.05 were considered significantly enriched (Supplementary Data 4).

Cytoscape visualization

To visualize complex enriched gene sets, analysis was performed using Cytoscape, a software platform containing Apps for visualizing complex networks. To functionally characterize these complex Supplementary DataSets, the Enrichment Map⁸³ App within Cytoscape was used. In brief, enrichment files and gene sets outputs from g:Profiler were processed using Enrichment Map and terms with a false discovery rate less than 0.01 were considered significantly enriched. Nodes represent pathways, and they are connected by edges (lines) between nodes with genes in common. Node size is dependent on the enrichment score associated with that pathway, while edges thickness is relative to the number of genes shared between nodes.

Transcription factor enrichment analyses

To identify transcription factors (TFs) that might play a role in the observed gene expression patterns, we employed complementary methods: TF binding site motif enrichment and ChIP-seq peak enrichment. For TF motif enrichment analyses, we used the HOMER software package⁴⁶ to calculate the enrichment of each motif in the promoter sequences (−200 bases to +50 bases relative to the gene TSS) of gene clusters with expression patterns of interest. HOMER was modified to use the large library of human position weight matrix (PWM) binding site models contained in build 2.0 of the CisBP database⁸⁴ and a log base 2 log likelihood scoring system. For ChIP-seq peak enrichment analysis, we used the RELI⁴⁷ algorithm. In brief, RELI calculates the enrichment of ChIP-seq peaks using a permutation-based method. RELI was run on the promoter sequences of gene sets of interest, using the “refGene_TSS_−1k_500” null model and a custom ChIP-seq library containing 11,504 human ChIP-seq datasets created by running MACS2⁸⁵ with default parameter settings on datasets obtained from the GEO database⁸⁶.

A final ranked list of TFs (Supplementary Data 6) was generated by combining the TF motif enrichment and ChIP-seq peak enrichment results. Prior to summarizing the HOMER data, results were pre-processed by expanding the TF components of each motif name into additional rows in order to account for multiple TFs recognizing a single motif. Unique TF entries were then summarized for HOMER and RELI datasets independently by selecting the record containing the lowest P value and assigning a percentile score based on the total number of unique records in each dataset (HOMER n = 1200 TFs; RELI n = 1487 TFs). These ranked summaries were then merged into a common dataset using the name of the TF and subsequently filtered for records present in both HOMER and RELI (n = 752 TFs). Each TF was then assigned a combined rank based on the average of the two percentile scores that were calculated in the respective summary datasets.

Collection of HNSCC patient samples

Head and neck cancer specimens were collected at the time of planned surgical resection under an IRB-approved study protocol UCCI-UMB-14-01 (IRB #2014-4755), a general specimen collection protocol (IRB #2017-2137), and a CCHMC-approved protocol (IRB# 2009-2700). Samples were de-identified prior to transport to the laboratory. These studies were approved by the Institutional Review Board at the University of Cincinnati and Cincinnati Children’s Hospital and were conducted in accordance with Good Clinical Practice guidelines and the Declaration of Helsinki. Written informed consent was received from all participating patients prior to enrollment. After surgical resection, tissue was placed in cold HypoThermosol FRS Preservation solution (Sigma-Aldrich, St Louis, MO, H4416) and transported directly to the laboratory on ice.

Collection of cervical tissue with potential CIN lesions

De-identified cervical tissues were obtained from women who were undergoing loop electrosurgical procedures (LEEPs) or cold cone biopsies as part of clinical care based on the Risk-Based Management Consensus Guidelines released by the American Society for Colposcopy and Cervical Pathology. Cervical cones were transferred in conical tubes with media on ice to the University of Cincinnati Department of Pathology & Laboratory Medicine, where they were grossed per routine procedure. During processing, a small slice from each quadrant was removed, de-identified, and assigned a study ID before being transferred to research. Specimens were fixed, embedded, sectioned, H&E stained, and screened by certified pathologist review for areas of ectocervical tissue with HPV features or CIN lesions. This research activity was determined to be exempt from continuing IRB review by the University of Cincinnati (2021-0194) and Cincinnati Children’s Hospital Medical Center (2021-0296).

Generation of 3D organotypic rafts from the patient-derived cervical and tonsillar epithelium

Primary cervical keratinocytes from hysterectomy-derived ectocervix were kindly provided by Dr. Aloysius Klingelhutz⁸⁷. Primary oral keratinocytes were isolated from tonsillectomy-derived palatine tonsils collected from Banner University Medical Center Tucson as approved by The University of Arizona Institutional Review Board⁸⁸. Mycoplasma testing was performed regularly and cell lines were confirmed to be negative. Prior to rafting, primary cervical and tonsillar cells were maintained as monolayer cultures (as described above for NIKS) but with conditional immortalization using the Rho-kinase inhibitor Y-27632 (Chemdea, Ridgewood, NJ, CD0141) at 10 μM^89,90. HPV18+ cervical cells were generated by removing Y-27632 from media and infecting with whole-genome HPV18 quasivirus at a multiplicity of infection of 100 viral genome equivalents per cell. Mock-infected cervical cells received an equivalent volume of only viral storage buffer (25 mM HEPES pH 7.5, 500 mM NaCl, 1 mM MgCl₂). HPV16+ tonsillar clones harboring episomal viral DNA were generated⁹¹. Cervical and tonsillar 3D organotypic rafts were both grown⁸⁸. Briefly, rafts were constructed by creating a dermal equivalent, where hTERT-immortalized neonatal human foreskin fibroblasts (HFF-hTERT cells) were embedded (8 × 10⁴ per raft) into a solidified rat tail type I collagen matrix (Corning, Corning, NY, 354236). On the next day, HPV18 ± cervical or HPV16 ± tonsillar keratinocytes were seeded (2.5 × 10⁵ per raft) atop the dermal equivalent to create confluent basal layers. Finally, on the third day, cultures were lifted onto a hydrophilic polytetrafluoroethylene membrane insert (0.4-μm pore size; MilliporeSigma, Burlington, MA, PICM0RG50) for an air-liquid interface and fed every 2 days (seven total media changes) with differentiation media (1.88 mM Ca²⁺ and no epidermal growth factor, EGF) to stratify until fully differentiated raft cultures were harvested on day 15 post-lift.

TCGA analysis

The TCGA PanCancer Atlas was analyzed using cBioPortal^92,93. In the Head and Neck Squamous Cell Carcinoma Supplementary DataSet, 488 patients had complete data (including mRNA expression data), and were included in the analysis; 85% (415 patients) were classified as HPV- HNSCC, 15% (72 patients) as HPV+ HNSCC, and one patient was unknown. In the Cervical Squamous Cell Carcinoma Supplementary DataSet, 275 patients had complete data and were included in the analysis. ELF3 mRNA expression relative to all diploid samples with z ≥ 2.0 was used to define the high ELF3 expression sub-group (10% of the HNSCC dataset and 9% of the cervical SCC dataset). A Pearson cutoff of ≥|0.4| was applied to select genes correlating with ELF3 expression. Overlap between gene lists was evaluated using BioVenn⁹⁴ and confirmed manually.

ELF3 knockdown experiments

Nontargeting and ELF3-specific short hairpin RNA (shRNA)-expressing lentiviral vectors were obtained through the Sigma MISSION shRNA program (Sigma-Aldrich, Inc., St Louis, MO, 68178). HPV16+ NIKS were transduced with NTsh or ELF3sh vectors (ELF3sh-1 TRCN0000013864; ELF3-sh2 TRCN0000013865, Sigma-Aldrich, Inc., St Louis, MO, 61878) and selected and maintained in media with 1 µg/mL puromycin.

Western blot analysis

Antibody information, including dilutions and validations, are detailed in Supplementary Data 1. Organotypic raft lysates generated from ELF3 NT and ELF3 knockdown cells were prepared in RIPA buffer (10%SDS, 1 M Tris, 5 M NaCl, 0.25 M EDTA, Triton, sodium deoxycholate) containing HALT protease and phosphate inhibitor cocktail (Thermo Fisher Scientific, Waltham, MA,78443) and 200 μM Na₃O₄V. About 30 μg of protein was loaded per lane on 4–20% Bis-Tris gels (Bio-Rad, Hercules, CA, 5678094), and protein was transferred to Transblot turbo midi-size membrane (Bio-Rad, Hercules, CA, 1704275) at 200 mA for 2 h. Membranes were blocked in 10% dry milk reconstituted in TNET and incubated with ELF3 antibody (Sigma St Louis MO, HPA003479) at 1:1000 (Sigma St Louis MO, HPA003479-100ul) and Rhodamine Anti-actin (Bio-Rad, Hercules, CA, 12004163) at 1:5000 in TNET overnight at 4 °C. Membranes were incubated in ECL anti-Rabbit IgG secondary antibody (Thermo Fisher Scientific, Waltham, MA, NA934V) at 1:5000 in dry milk reconstituted.

Statistics and reproducibility

scRNAseq cluster gene lists were determined by Seurat’s FindAllMarkers functions using the Wilcoxon rank-sum test⁷⁷ as detailed above. Statistical analyses were performed using GraphPad Prism. Significance was set at p < 0.05 for all experiments, with the exception of the mutual exclusivity test on TCGA data, where significance was set at p < 0.10. For data with one independent variable, Mann–Whitney test with two-tailed distribution, unpaired t-test, or one-way ANOVA was used. For data with two independent variables, two-way ANOVA was used. The most appropriate tests for multiple comparisons following two-way ANOVA analyses was selected according to GraphPad recommendations and is specified per figure. Correlations were evaluated by Spearman’s rank correlation test. Differences in Kaplan–Meier survival curves were tested by the log-rank test. Statistical values reported by cBioPortal on TCGA PanCancer Atlas data were noted. The number of replicates, statistical tests, and p values are noted for each analysis in the figure legends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

The RT-qPCR data generated in this study are provided in the Source Data file. The scRNAseq data generated in this study have been deposited in the GEO database under accession code GSE189670. The analyzed scRNAseq data were available in Supplementary Information files. The scRNAseq publicly available data used in this study are available on the Gene Expression Omnibus database: accession ID GSE164690³⁸. The TCGA data were accessed and are available through cBioPortal^92,93. The remaining data were provided with this paper within the Article, Supplementary Information or Source Data file. Source data are provided with this paper.

References

zur Hausen, H. The search for infectious causes of human cancers: where and why (Nobel lecture). Angew. Chem. Int. Ed. Engl. 48, 5798–5808 (2009).
Article CAS PubMed Google Scholar
McBride, A. A. Human papillomaviruses: diversity, infection and host interactions. Nat. Rev. Microbiol. 20, 95–108 (2022).
Article CAS PubMed Google Scholar
Della Fera, A. N. et al. Persistent human papillomavirus infection. Viruses 13, 321 (2021).
Article PubMed PubMed Central Google Scholar
Bosch, F. X. et al. The causal relation between human papillomavirus and cervical cancer. J. Clin. Pathol. 55, 244–265 (2002).
Article CAS PubMed PubMed Central Google Scholar
Serrano, B. et al. Epidemiology and burden of HPV-related disease. Best Pract. Res. Clin. Obstet. Gynaecol. 47, 14–26 (2018).
Article PubMed Google Scholar
Doorbar, J. et al. Principles of epithelial homeostasis control during persistent human papillomavirus infection and its deregulation at the cervical transformation zone. Curr. Opin. Virol. 51, 96–105 (2021).
Article CAS PubMed Google Scholar
Scheffner, M. et al. The E6 oncoprotein encoded by human papillomavirus types 16 and 18 promotes the degradation of p53. Cell 63, 1129–1136 (1990).
Article CAS PubMed Google Scholar
Münger, K. et al. Complex formation of human papillomavirus E7 proteins with the retinoblastoma tumor suppressor gene product. EMBO J. 8, 4099–4105 (1989).
Article PubMed PubMed Central Google Scholar
Basukala, O. & Banks, L. The not-so-good, the bad and the ugly: HPV E5, E6 and E7 oncoproteins in the orchestration of carcinogenesis. Viruses 13, 1892 (2021).
Article CAS PubMed PubMed Central Google Scholar
Phelps, W. C. et al. The human papillomavirus type 16 E7 gene encodes transactivation and transformation functions similar to those of adenovirus E1A. Cell 53, 539–547 (1988).
Article CAS PubMed Google Scholar
Schiffman, M. et al. Carcinogenic human papillomavirus infection. Nat. Rev. Dis. Primers 2, 16086 (2016).
Article PubMed Google Scholar
Moody, C. A. & Laimins, L. A. Human papillomavirus oncoproteins: pathways to transformation. Nat. Rev. Cancer 10, 550–560 (2010).
Article CAS PubMed Google Scholar
Meyers, C. et al. Biosynthesis of human papillomavirus from a continuous cell line upon epithelial differentiation. Science 257, 971–973 (1992).
Article ADS CAS PubMed Google Scholar
Chow, L. T. Model systems to study the life cycle of human papillomaviruses and HPV-associated cancers. Virol. Sin. 30, 92–100 (2015).
Article CAS PubMed PubMed Central Google Scholar
Doorbar, J. Model systems of human papillomavirus-associated disease. J. Pathol. 238, 166–179 (2016).
Article PubMed Google Scholar
Frattini, M. G., Lim, H. B. & Laimins, L. A. In vitro synthesis of oncogenic human papillomaviruses requires episomal genomes for differentiation-dependent late expression. Proc. Natl Acad. Sci. USA 93, 3062–3067 (1996).
Article ADS CAS PubMed PubMed Central Google Scholar
Doorbar, J. et al. The biology and life-cycle of human papillomaviruses. Vaccine 30, F55–F70 (2012).
Article CAS PubMed Google Scholar
Lorenz et al. Inactivation of p53 rescues the maintenance of high risk HPV DNA genomes deficient in expression of E6. PLoS Pathog. 9, e1003717 (2013).
Article PubMed PubMed Central Google Scholar
Yang, R. et al. Combined transcriptome and proteome analysis of immortalized human keratinocytes expressing human papillomavirus 16 (HPV16) oncogenes reveals novel key factors and networks in HPV-induced carcinogenesis. mSphere 4, e00129–19 (2019).
Article CAS PubMed PubMed Central Google Scholar
Chatterjee, S. et al. Tissue-specific gene expression during productive human papillomavirus 16 infection of cervical, foreskin, and tonsil epithelium. J. Virol. 93, e00915–e00919 (2019).
Article CAS PubMed PubMed Central Google Scholar
Lambert, P. F. et al. in Human Papillomaviruses: Methods and Protocols, (eds Davy, C. & J. Doorbar, J.) Ch. 12 (Humana Press, 2006)
Flores, E. R. et al. Establishment of the human papillomavirus type 16 (HPV-16) life cycle in an immortalized human foreskin keratinocyte cell line. Virology 262, 344–354 (1999).
Article CAS PubMed Google Scholar
Nuovo, G. J. et al. Importin-β and exportin-5 are indicators of acute viral infection: correlation of their detection with commercially available detection kits. Ann. Diagn. Pathol. 34, 36–41 (2018).
Article PubMed Google Scholar
Geißler, C. et al. The role of p16 expression as a predictive marker in HPV-positive oral SCCHN–a retrospective single-center study. Anticancer Res. 33, 913–916 (2013).
PubMed Google Scholar
Trapnell, C. et al. The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells. Nat. Biotechnol. 32, 381–386 (2014).
Article CAS PubMed PubMed Central Google Scholar
White, E. A. Manipulation of epithelial differentiation by HPV oncoproteins. Viruses 11, 369 (2019).
Article CAS PubMed PubMed Central Google Scholar
McCance, D. J. et al. Human papillomavirus type 16 alters human epithelial cell differentiation in vitro. Proc. Natl Acad. Sci. USA 85, 7169–7173 (1988).
Article ADS CAS PubMed PubMed Central Google Scholar
Kranjec, C. et al. Modulation of basal cell fate during productive and transforming HPV-16 infection is mediated by progressive E6-driven depletion of Notch. J. Pathol. 242, 448–462 (2017).
Article CAS PubMed PubMed Central Google Scholar
Bergner, S. et al. Individual and complementary effects of human papillomavirus oncogenes on epithelial cell proliferation and differentiation. Cells Tissues Organs 201, 97–108 (2016).
Article PubMed Google Scholar
Smedts, F. et al. Keratin expression in cervical cancer. Am. J. Pathol. 141, 497–511 (1992).
CAS PubMed PubMed Central Google Scholar
Uhlén, M. et al. Proteomics. Tissue-based map of the human proteome. Science 347, 1260419 (2015).
Andrysik, Z. et al. Multi-omics analysis reveals contextual tumor suppressive and oncogenic gene modules within the acute hypoxic response. Nat. Commun. 12, 1375 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Fu, L. et al. HIF-1α-induced histone demethylase JMJD2B contributes to the malignant phenotype of colorectal cancer cells via an epigenetic mechanism. Carcinogenesis 33, 1664–1673 (2012).
Article CAS PubMed Google Scholar
Bredell, M. G. et al. Current relevance of hypoxia in head and neck cancer. Oncotarget 7, 50781–50804 (2016).
Article PubMed PubMed Central Google Scholar
Göttgens, E. L. et al. HPV, hypoxia and radiation response in head and neck cancer. Br. J. Radiol. 92, 20180047 (2019).
PubMed Google Scholar
Ghasemi, F. et al. Mutational analysis of head and neck squamous cell carcinoma stratified by smoking status. JCI Insight 4, e123443 (2019).
Article PubMed PubMed Central Google Scholar
Stransky, N. et al. The mutational landscape of head and neck squamous cell carcinoma. Science 333, 1157–1160 (2011).
Article ADS CAS PubMed PubMed Central Google Scholar
Kürten, C. H. L. et al. Investigating immune and non-immune cell interactions in head and neck tumors by single-cell RNA sequencing. Nat. Commun. 12, 7338 (2021).
Article ADS PubMed PubMed Central Google Scholar
Song, J. et al. Genomic organization and expression of a human gene for Myc-associated zinc finger protein (MAZ). J. Biol. Chem. 273, 20603–20614 (1998).
Article CAS PubMed Google Scholar
O’Connor, L., Gilmour, J. & Bonifer, C. The role of the ubiquitously expressed transcription factor Sp1 in tissue-specific transcriptional regulation and in disease. Yale J. Biol. Med. 89, 513–525 (2016).
PubMed PubMed Central Google Scholar
McConnell, B. B. et al. The diverse functions of Krüppel-like factors 4 and 5 in epithelial biology and pathobiology. Bioessays 29, 549–557 (2007).
Article CAS PubMed PubMed Central Google Scholar
Oliver, J. R., Kushwah, R. & Hu, J. Multiple roles of the epithelium-specific ETS transcription factor, ESE-1, in development and disease. Lab Invest. 92, 320–330 (2012).
Article CAS PubMed Google Scholar
Luk, I. Y., Reehorst, C. M. & Mariadason, J. M. ELF3, ELF5, EHF and SPDEF transcription factors in tissue homeostasis and cancer. Molecules 23, 2191 (2018).
Article PubMed PubMed Central Google Scholar
Conde, J. et al. E74-like factor 3 and nuclear factor-κB regulate lipocalin-2 expression in chondrocytes. J. Physiol. 594, 6133–6146 (2016).
Article CAS PubMed PubMed Central Google Scholar
Longoni, N. et al. ETS transcription factor ESE1/ELF3 orchestrates a positive feedback loop that constitutively activates NF-κB and drives prostate cancer progression. Cancer Res. 73, 4533–4547 (2013).
Article CAS PubMed Google Scholar
Heinz, S. et al. Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol. Cell 38, 576–589 (2010).
Article CAS PubMed PubMed Central Google Scholar
Harley, J. B. et al. Transcription factors operate across disease loci, with EBNA2 implicated in autoimmunity. Nat. Genet. 50, 699–707 (2018).
Article CAS PubMed PubMed Central Google Scholar
Nakahara, T. et al. Human papillomavirus type 16 E1circumflexE4 contributes to multiple facets of the papillomavirus life cycle. J. Virol. 79, 13150–13165 (2005).
Article CAS PubMed PubMed Central Google Scholar
Egawa, N. et al. HPV16 and 18 genome amplification show different E4-dependence, with 16E4 enhancing E1 nuclear accumulation and replicative efficiency via its cell cycle arrest and kinase activation functions. PLoS Pathog. 13, e1006282 (2017).
Article PubMed PubMed Central Google Scholar
Klymenko, T. et al. RNA-seq analysis of differentiated keratinocytes reveals a massive response to late events during human papillomavirus 16 infection, including loss of epithelial barrier function. J. Virol. 91, e01001–e01017 (2017).
Article CAS PubMed PubMed Central Google Scholar
Lambert, P. F. et al. Using an immortalized cell line to study the HPV life cycle in organotypic “raft” cultures. Methods Mol. Med. 119, 141–155 (2005).
CAS PubMed Google Scholar
Allen-Hoffmann, B. L. et al. Normal growth and differentiation in a spontaneously immortalized near-diploid human keratinocyte cell line, NIKS. J. Invest. Dermatol. 114, 444–455 (2000).
Article CAS PubMed Google Scholar
Roy, M. et al. Role of Notch1 in NIKS cellular differentiation. J. Am. Coll. Surg. 213, S102 (2011).
Article Google Scholar
Isaacson Wechsler, E. et al. Reconstruction of human papillomavirus type 16-mediated early-stage neoplasia implicates E6/E7 deregulation and the loss of contact inhibition in neoplastic progression. J. Virol. 86, 6358–6364 (2012).
Article PubMed PubMed Central Google Scholar
Hoffmann, R. et al. Different modes of human papillomavirus DNA replication during maintenance. J. Virol. 80, 4431–4439 (2006).
Article CAS PubMed PubMed Central Google Scholar
Flores, E. R. et al. The human papillomavirus type 16 E7 oncogene is required for the productive stage of the viral life cycle. J. Virol. 74, 6622–6631 (2000).
Article CAS PubMed PubMed Central Google Scholar
Egawa, N. et al. The E1 protein of human papillomavirus type 16 is dispensable for maintenance replication of the viral genome. J. Virol. 86, 3276–3283 (2012).
Article CAS PubMed PubMed Central Google Scholar
Jackson, R. et al. Tumourigenesis driven by the human papillomavirus type 16 Asian-American E6 variant in a three-dimensional keratinocyte model. PLoS ONE 9, e101540 (2014).
Article ADS PubMed PubMed Central Google Scholar
Klymenko, T. et al. RNA-Seq analysis of differentiated keratinocytes reveals a massive response to late events during human papillomavirus 16 infection, including loss of epithelial barrier function. J. Virol. 91, e01001–e01017 (2017).
Article CAS PubMed PubMed Central Google Scholar
Zehbe, I. et al. Human papillomavirus 16 E6 variants differ in their dysregulation of human keratinocyte differentiation and apoptosis. Virology 383, 69–77 (2009).
Article CAS PubMed Google Scholar
Bose, A. et al. Keratin k15 as a biomarker of epidermal stem cells. Int. J. Mol. Sci. 14, 19385–19398 (2013).
Article CAS PubMed PubMed Central Google Scholar
Liu, Y. et al. Keratin 15 promoter targets putative epithelial stem cells in the hair follicle bulge. J. Invest. Dermatol. 121, 963–968 (2003).
Article CAS PubMed Google Scholar
Lee, J., Lee, P. & Wu, X. Molecular and cytoskeletal regulations in epidermal development. Semin. Cell Dev. Biol. 69, 18–25 (2017).
Article CAS PubMed PubMed Central Google Scholar
Duxbury, M. S. et al. CEACAM6 is a novel biomarker in pancreatic adenocarcinoma and PanIN lesions. Ann. Surg. 241, 491–496 (2005).
Article PubMed PubMed Central Google Scholar
Ru, G. Q. et al. CEACAM6 is a prognostic biomarker and potential therapeutic target for gastric carcinoma. Oncotarget 8, 83673–83683 (2017).
Article PubMed PubMed Central Google Scholar
Pandey, R. et al. Carcinoembryonic antigen cell adhesion molecule 6 (CEACAM6) in Pancreatic Ductal Adenocarcinoma (PDA): An integrative analysis of a novel therapeutic target. Sci. Rep. 9, 18347 (2019).
Article ADS CAS PubMed PubMed Central Google Scholar
Bedard, M. C. et al. Patient-derived organotypic epithelial rafts model phenotypes in juvenile-onset recurrent respiratory papillomatosis. Viruses 13, 68 (2021).
Article CAS PubMed PubMed Central Google Scholar
Lee, D. et al. Using organotypic epithelial tissue culture to study the human papillomavirus life cycle. Curr. Protoc. Microbiol. 41, 14b.8.1–14b.8.19 (2016).
Article PubMed Google Scholar
Nuovo, G. False-positive results in diagnostic immunohistochemistry are related to horseradish peroxidase conjugates in commercially available assays. Ann. Diagn. Pathol. 25, 54–59 (2016).
Article PubMed Google Scholar
Nuovo, G. et al. Importin-β and exportin-5 are strong biomarkers of productive reoviral infection of cancer cells. Ann. Diagn. Pathol. 32, 28–34 (2018).
Article PubMed Google Scholar
Schlecht, N. F. et al. A comparison of clinically utilized human papillomavirus detection methods in head and neck cancer. Mod. Pathol. 24, 1295–1305 (2011).
Article CAS PubMed PubMed Central Google Scholar
Bedard, M. C. et al. HPV strain predicts severity of juvenile-onset recurrent respiratory papillomatosis with implications for disease screening. Cancers 13, 2556 (2021).
Article CAS PubMed PubMed Central Google Scholar
Ye, J. et al. Primer-BLAST: a tool to design target-specific primers for polymerase chain reaction. BMC Bioinformatics 13, 134 (2012).
Article CAS PubMed PubMed Central Google Scholar
Bolick, D. R. et al. Laboratory implementation of human papillomavirus testing. Arch. Pathol. Lab Med. 127, 984–990 (2003).
Article PubMed Google Scholar
Macosko, E. Z. et al. Highly parallel genome-wide expression profiling of individual cells using nanoliter droplets. Cell 161, 1202–1214 (2015).
Article CAS PubMed PubMed Central Google Scholar
Adam, M., Potter, A. S. & Potter, S. S. Psychrophilic proteases dramatically reduce single-cell RNA-seq artifacts: a molecular atlas of kidney development. Development 144, 3625–3632 (2017).
CAS PubMed PubMed Central Google Scholar
Stuart, T. et al. Comprehensive integration of single-cell data. Cell 177, 1888–1902.e21 (2019).
Article CAS PubMed PubMed Central Google Scholar
Qiu, X. et al. Single-cell mRNA quantification and differential analysis with Census. Nat. Methods 14, 309–315 (2017).
Article CAS PubMed PubMed Central Google Scholar
Qiu, X. et al. Reversed graph embedding resolves complex single-cell developmental trajectories. Nat. Methods 14, 979–982 (2017)
Choudhary, S. & Satija, R. Comparison and evaluation of statistical error models for scRNA-seq. Genome Biol. 23, 27 (2022).
Article CAS PubMed PubMed Central Google Scholar
Raudvere, U. et al. g:Profiler: a web server for functional enrichment analysis and conversions of gene lists (2019 update). Nucleic Acids Res. 47, W191–W198 (2019).
Article CAS PubMed PubMed Central Google Scholar
Cunningham, F. et al. Ensembl 2019. Nucleic Acids Res. 47, D745–D751 (2019).
Article CAS PubMed Google Scholar
Merico, D. et al. Enrichment map: a network-based method for gene-set enrichment visualization and interpretation. PLoS ONE 5, e13984 (2010).
Article ADS PubMed PubMed Central Google Scholar
Lambert, S. A. et al. Similarity regression predicts evolution of transcription factor sequence specificity. Nat. Genet. 51, 981–989 (2019).
Article ADS CAS PubMed Google Scholar
Zhang, Y. et al. Model-based analysis of ChIP-Seq (MACS). Genome Biol. 9, R137 (2008).
Article PubMed PubMed Central Google Scholar
Barrett, T. et al. NCBI GEO: archive for functional genomics data sets–update. Nucleic Acids Res. 41(Database issue), D991-D995 (2013).
Sprague, D. L. et al. Telomerase activation in cervical keratinocytes containing stably replicating human papillomavirus type 16 episomes. Virology 301, 247–254 (2002).
Article CAS PubMed Google Scholar
Jackson, R. et al. 3D oral and cervical tissue models for studying papillomavirus host-pathogen interactions. Curr. Protoc. Microbiol. 59, e129 (2020).
Article CAS PubMed Google Scholar
Chapman, S. et al. Human keratinocytes are efficiently immortalized by a Rho kinase inhibitor. J. Clin. Invest. 120, 2619–2626 (2010).
Article CAS PubMed PubMed Central Google Scholar
Chapman, S. et al. The effect of Rho kinase inhibition on long-term keratinocyte proliferation is rapid and conditional. Stem Cell Res. Ther. 5, 60 (2014).
Article PubMed PubMed Central Google Scholar
Coursey, T. L. & McBride, A. A. Development of keratinocyte cell lines containing extrachromosomal human papillomavirus genomes. Curr. Protoc. 1, e235 (2021).
Article CAS PubMed PubMed Central Google Scholar
Gao, J. et al. Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal. Sci. Signal 6, pl1 (2013).
Article PubMed PubMed Central Google Scholar
Cerami, E. et al. The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data. Cancer Discov. 2, 401–404 (2012).
Article PubMed Google Scholar
Hulsen, T., de Vlieg, J. & Alkema, W. BioVenn – a web application for the comparison and visualization of biological lists using area-proportional Venn diagrams. BMC Genomics 9, 488 (2008).
Article PubMed PubMed Central Google Scholar

Download references

Acknowledgements

For the receipt of cervical samples, we thank the OB/GYN residents at the University of Cincinnati. We are thankful to Dr. Aloysius Klingelhutz for primary cervical keratinocytes, Dr. Deepta Bhattacharya and Dr. Lucas D’Souza for palatine tonsils, and Dr. Felicia Goodrum for HFF-hTERT cells to the Doorslaer laboratory. For assistance with the processing of clinical samples, we thank the pathology assistants, particularly Mojun Zhao, in the Department of Pathology at the University of Cincinnati. For reagents and discussions, we thank all members of the Wells laboratory. We thank Sarah McLeod and Evan Meyer in the CCHMC Confocal Imaging Core for hands-on support. This study was supported by services from the Pathology Research Core shared facility in the Cincinnati Children’s Research Foundation with excellent technical expertize provided by Meredith Taylor and Lynn Duncan. Histology services were additionally provided by the University of Arizona Cancer Center’s Tissue Acquisition and Cellular/Molecular Analysis (TACMASR) shared resource core which is supported by the National Cancer Institute of the National Institutes of Health (NIH) under award number P30 CA023074. This research was funded by the following NIH grants: T32ES007250 (M.C.B.), F30AI157229 (M.C.B.), CA228113 (S.I.W.), U01AI130830 (M.T.W.), R01HG010730 (M.T.W.), R03DE030211 (K.V.D.), as well as support from CancerFreeKids (M.C.B.), the State of Arizona Improving Health TRIF (K.V.D), the BI05 Institute (R.J.), and the Natural Sciences and Engineering Research Council of Canada PDF-546182-2020 (R.J.)

Author information

These authors contributed equally: Mary C. Bedard, Tafadzwa Chihanga.

Authors and Affiliations

Division of Oncology, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH, 45229, USA
Mary C. Bedard, Tafadzwa Chihanga, Adrean Carlile & Susanne I. Wells
School of Animal and Comparative Biomedical Sciences, University of Arizona, Tucson, AZ, 85721, USA
Robert Jackson & Koenraad Van Doorslaer
Division of Clinical Operations, Medpace, Cincinnati, OH, 45227, USA
Marion G. Brusadelli
McArdle Laboratory for Cancer Research, School of Medicine and Public Health, University of Wisconsin, Madison, WI, 53705, USA
Denis Lee & Paul F. Lambert
Center for Autoimmune Genomics and Etiology (CAGE), Cincinnati Children’s Hospital Medical Center, Cincinnati, OH, 45229, USA
Andrew VonHandorf & Matthew T. Weirauch
Division of Allergy and Immunology, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH, 45229, USA
Mark Rochman & Marc E. Rothenberg
Division of Biomedical Informatics, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH, 45229, USA
Phillip J. Dexheimer & Bruce J. Aronow
William G. Lowrie Department of Chemical and Biomolecular Engineering, Ohio State University, 151 W. Woodruff Ave, Columbus, OH, 43210, USA
Jeffrey Chalmers
Department of Pathology, Ohio State University Medical Center, Columbus, OH, 43210, USA
Gerard Nuovo
Division of Hematology/Oncology, Department of Internal Medicine, University of Cincinnati College of Medicine, Cincinnati, OH, 45267, USA
Maria Lehn & Trisha M. Wise-Draper
Cancer Biology Graduate Interdisciplinary Program, University of Arizona, Tucson, AZ, 85721, USA
David E. J. Williams & Koenraad Van Doorslaer
Medical Scientist Training M.D.-Ph.D. Program (MSTP), College of Medicine-Tucson, University of Arizona, Tucson, AZ, USA
David E. J. Williams
UPMC Hillman Cancer Center, University of Pittsburgh, Pittsburgh, PA, 15232, USA
Aditi Kulkarni & Robert L. Ferris
Tumor Microenvironment Center, UPMC Hillman Cancer Center, Pittsburgh, PA, 15232, USA
Aditi Kulkarni & Robert L. Ferris
Department of Obstetrics and Gynecology, University of Cincinnati Medical Center, Cincinnati, OH, 45267, USA
Molly Carey, Amanda Jackson, Caroline Billingsley & Thomas J. Herzog
Department of Otolaryngology, University of Cincinnati, Cincinnati, OH, 45267, USA
Alice Tang, Chad Zender & Yash Patil
Department of Otolaryngology, University of Pittsburgh, Pittsburgh, PA, USA
Robert L. Ferris
Department of Immunology, University of Pittsburgh, Pittsburgh, PA, 15232, USA
Robert L. Ferris
Department of Pathology & Laboratory Medicine, University of Cincinnati College of Medicine, Cincinnati, OH, 45267, USA
Ady Kendler & Kathryn A. Wikenheiser-Brokamp
Division of Developmental Biology, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH, 45229, USA
Bruce J. Aronow, Matthew Kofron, Mike Adam & S. Steven Potter
Department of Pediatrics, University of Cincinnati College of Medicine, Cincinnati, OH, 45267, USA
Bruce J. Aronow, Matthew Kofron, Marc E. Rothenberg, Matthew T. Weirauch & Susanne I. Wells
Divisions of Human Genetics, Biomedical Informatics and Developmental Biology, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH, 45229, USA
Matthew T. Weirauch
The BIO5 Institute, University of Arizona, Tucson, AZ, 85721, USA
Koenraad Van Doorslaer
Department of Immunobiology, University of Arizona, Tucson, AZ, 85721, USA
Koenraad Van Doorslaer
UA Cancer Center, University of Arizona, Tucson, AZ, 85721, USA
Koenraad Van Doorslaer
Division of Pathology & Laboratory Medicine and The Perinatal Institute Division of Pulmonary Biology, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH, 45229, USA
Kathryn A. Wikenheiser-Brokamp

Authors

Mary C. Bedard
View author publications
You can also search for this author in PubMed Google Scholar
Tafadzwa Chihanga
View author publications
You can also search for this author in PubMed Google Scholar
Adrean Carlile
View author publications
You can also search for this author in PubMed Google Scholar
Robert Jackson
View author publications
You can also search for this author in PubMed Google Scholar
Marion G. Brusadelli
View author publications
You can also search for this author in PubMed Google Scholar
Denis Lee
View author publications
You can also search for this author in PubMed Google Scholar
Andrew VonHandorf
View author publications
You can also search for this author in PubMed Google Scholar
Mark Rochman
View author publications
You can also search for this author in PubMed Google Scholar
Phillip J. Dexheimer
View author publications
You can also search for this author in PubMed Google Scholar
Jeffrey Chalmers
View author publications
You can also search for this author in PubMed Google Scholar
Gerard Nuovo
View author publications
You can also search for this author in PubMed Google Scholar
Maria Lehn
View author publications
You can also search for this author in PubMed Google Scholar
David E. J. Williams
View author publications
You can also search for this author in PubMed Google Scholar
Aditi Kulkarni
View author publications
You can also search for this author in PubMed Google Scholar
Molly Carey
View author publications
You can also search for this author in PubMed Google Scholar
Amanda Jackson
View author publications
You can also search for this author in PubMed Google Scholar
Caroline Billingsley
View author publications
You can also search for this author in PubMed Google Scholar
Alice Tang
View author publications
You can also search for this author in PubMed Google Scholar
Chad Zender
View author publications
You can also search for this author in PubMed Google Scholar
Yash Patil
View author publications
You can also search for this author in PubMed Google Scholar
Trisha M. Wise-Draper
View author publications
You can also search for this author in PubMed Google Scholar
Thomas J. Herzog
View author publications
You can also search for this author in PubMed Google Scholar
Robert L. Ferris
View author publications
You can also search for this author in PubMed Google Scholar
Ady Kendler
View author publications
You can also search for this author in PubMed Google Scholar
Bruce J. Aronow
View author publications
You can also search for this author in PubMed Google Scholar
Matthew Kofron
View author publications
You can also search for this author in PubMed Google Scholar
Marc E. Rothenberg
View author publications
You can also search for this author in PubMed Google Scholar
Matthew T. Weirauch
View author publications
You can also search for this author in PubMed Google Scholar
Koenraad Van Doorslaer
View author publications
You can also search for this author in PubMed Google Scholar
Kathryn A. Wikenheiser-Brokamp
View author publications
You can also search for this author in PubMed Google Scholar
Paul F. Lambert
View author publications
You can also search for this author in PubMed Google Scholar
Mike Adam
View author publications
You can also search for this author in PubMed Google Scholar
S. Steven Potter
View author publications
You can also search for this author in PubMed Google Scholar
Susanne I. Wells
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

M.C.B., M.G.B., M.A., S.S.P., and S.I.W. conceptualized and designed the project. M.C.B, T.C., A.C., M.G.B., D.L., J.C., G.N., R.J., and D.E.J.W. conducted experiments and analyzed the data. M.C.B., A.V., M.A., M.T.W., B.A., and P.D. performed bioinformatic analyses. M.R., M.K., M.E.R., A.K., R.L.F., and K.V.D. provided experimental training and conceptual input. M.L., T.M.W.-D., P.F.L., R.J., D.E.J.W., and K.V.D. provided key reagents, samples, and/or protocols. M.C.B. coordinated cervical tissue collections supervised by M.C., A.J., C.B., and T.J.H., and HNSCC tissue collections supervised by A.T., C.Z., Y.P., M.L., and T.M.W.-D. A.C. assisted in the transport and processing of tissues. G.N., A.K., and K.A.W.-B. provided board-certified pathology analyses. M.T.W., P.F.L., M.A., S.S.P., K.V.D., and S.I.W. supervised the project. M.C.B. and T.C. wrote the manuscript, with edits and revisions by M.C.B., K.A.W-B., P.F.L., M.A., and S.I.W. All authors provided important intellectual contributions and critical feedback and approved the manuscript.

Corresponding authors

Correspondence to Mike Adam, S. Steven Potter or Susanne I. Wells.

Ethics declarations

Competing interests

R.L.F: Adagene Incorporated: Consulting; Aduro Biotech, Inc: Consulting, AstraZeneca/MedImmune: Clinical Trial, Research Funding; Bicara Therapeutics, Inc: Consultant; Bristol-Myers Squibb: Advisory Board, Clinical Trial, Research Funding; Brooklyn Immunotherapeutics LLC: Consultant; Catenion: Consultant; Coherus BioSciences, Inc.: Advisory Board; Eisai Europe Limited: Advisory Board; EMD Serono: Consultant; Everest Clinical Research Corporation: Consultant; F. Hoffmann-La Roche Ltd: Consultant; Federation Bio, Inc: Consultant; Genocea Biosciences, Inc: Consultant; Genmab: Advisory Board; Hookipa Biotech GmbH: Advisory Board; Instil Bio, Inc: Advisory Board; Kowa Research Institute, Inc.: Consultant; Lifescience Dynamics Limited: Advisory Board; MacroGenics, Inc.: Advisory Board; MeiraGTx, LLC: Advisory Board; Merck: Advisory Board, Clinical Trial; Mirati Therapeutics, Inc: Consultant; Mirror Biologics Inc: Supplementary Data Safety Monitoring Board; Nanobiotix: Consultant; Novartis Pharmaceutical Corporation: Consulting; Novasenta: Consulting, Stock, Research Funding; Numab Therapeutics AG: Advisory Board; OncoCyte Corporation: Advisory Board; Pfizer: Advisory Board; PPD Development, L.P.: Consultant; Rakuten Medical, Inc: Advisory Board; Sanofi: Consultant; Seagen, Inc: Advisory Board; SIRPant Immunotherapeutics, Inc: Advisory Board; Tesaro: Research Funding; Vir Biotechnology, Inc: Advisory Board; Zymeworks, Inc.: Consultant. T.J.H.: AstraZeneca, Caris, Clovis, Eisai, Genentech, GlaxoSmithKline, Immunogen, Johnson & Johnson, Merck, Mersana and Seagen (Advisory Boards). The remaining authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. Peer reviewer reports are available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Peer Review File

Description of Additional Supplementary Files

Supplementary Data 1

Supplementary Data 2

Supplementary Data 3

Supplementary Data 4

Supplementary Data 5

Supplementary Data 6

Reporting Summary

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Bedard, M.C., Chihanga, T., Carlile, A. et al. Single cell transcriptomic analysis of HPV16-infected epithelium identifies a keratinocyte subpopulation implicated in cancer. Nat Commun 14, 1975 (2023). https://doi.org/10.1038/s41467-023-37377-0

Download citation

Received: 21 March 2022
Accepted: 15 March 2023
Published: 08 April 2023
DOI: https://doi.org/10.1038/s41467-023-37377-0

This article is cited by

Multiomic analysis of cervical squamous cell carcinoma identifies cellular ecosystems with biological and clinical relevance
- Junpeng Fan
- Funian Lu
- Chaoyang Sun
Nature Genetics (2023)

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.