Novel canonical and non-canonical viral antigens extend current targets for immunotherapy of HPV-driven cervical cancer

Summary Current immunotherapeutic approaches for human papillomavirus (HPV)-driven cervical cancer target the viral oncogenes E6 and E7. We report viral canonical and alternative reading frame (ARF)-derived sequences presented on cervical tumor cells, including antigens encoded by the conserved viral gene E1. We confirm immunogenicity of the identified viral peptides in HPV-positive women, and women with cervical intraepithelial neoplasia. We observe consistent transcription of the E1, E6, and E7 genes in 10 primary cervical tumor resections from the four most common high-risk HPV subtypes (HPV16, 18, 31, and 45), suggesting the suitability of E1 as therapeutic target. We finally confirm HLA presentation of canonical peptides derived from E6 and E7, and ARF-derived viral peptides from a reverse-strand transcript spanning the HPV E1 and E2 genes in primary human cervical tumor tissue. Our results extend currently known viral immunotherapeutic targets in cervical cancer and highlight E1 as an important cervical cancer antigen.


INTRODUCTION
Human papillomaviruses (HPVs) are small DNA viruses, of which there are more than 170 types, including 16 types that are designated ''high risk'' or oncogenic. 1,2 High-risk types are present in more than 99% of cervical cancers, which is the fourth most common type of female cancer worldwide. [3][4][5][6] Among the types of high-risk HPV, HPV16, HPV18, HPV45, and HPV31 are the most common, with 59%, 12%, 4.8%, and 3.7% prevalence, respectively. 7 Over 600,000 women were diagnosed with cervical cancer, and 341,831 deaths were recorded worldwide in 2020. 3 HPVs rely on the host cell replication proteins to mediate viral DNA and protein synthesis. The small HPV genome, approximately 8 kb in size, can be divided into six early genes (E1, E2, E4, E5, E6, and E7), which have regulatory function and play a role during the viral life cycle, and two late genes (L1 and L2) that form the viral capsid. 8 Most HPV infections are subclinical and are cleared by the human immune system within 2 years of acquisition; however, persistent infection with HPV can lead to cell transformation and accumulation of mutations over time, which can ultimately induce the development of a cancerous lesion. 9 The HPV oncogenic proteins E6 and E7 are crucial in the induction and maintenance of cellular transformation and are expressed in most HPV-associated cervical cancers. High-risk E6 proteins prevent cell growth inhibition by targeting the tumor suppressor p53 for degradation via the ubiquitin pathway, 10,11 which prevents apoptosis and extends cell life. Similarly, the HPV E7 protein binds to retinoblastoma protein family members and targets them for degradation, which drives the E2F-dependent induction of the cell life cycle and induces S phase and cell proliferation. [11][12][13] Both E6 and E7 can interact with c-Myc and support cell longevity through telomerase reverse transcriptase promotor activation. 14,15 As a result, among other interactions, these oncoproteins perturb the control of cell cycle progression, leading to increased cellular proliferation and survival, which contributes to the development and maintenance of HPV-associated cancers.
T cell immunity is important for control of HPV-driven cancers as shown by the increased prevalence of HPVassociated tumors in immunocompromised individuals. 16 It has further been shown that tumor-infiltrating lymphocytes correlate with improved outcomes in patients with cervical cancer. [17][18][19] Since it is established that HPV E6/E7 are expressed in HPV-driven malignancies to maintain continuous cell survival and proliferation, several therapeutic approaches have been developed and evaluated targeting these viral oncoproteins.
Adoptive T cell therapy selected for E6 and E7 antigens has been shown to be able to induce complete tumor regression in selected treated patients, 20 and adoptive T cell therapy showed a dominant T cell clonotype targeting HLA-A*02:01-restricted E6 [29][30][31][32][33][34][35][36] (TIHDIILECV) in a patient with metastatic anal cancer, who had been disease free 22 months after successful treatment. 21 Furthermore, treatment using T cell receptor (TCR)-engineered T cells targeting E7 for patients with metastatic HPV-associated epithelial cancers showed a number of partial responders and robust tumor regression in 6 of 12 patients including 4 of 8 patients with anti-PD-1 refractory disease. 22 DNA vaccines encoding both E6/E7 fusion genes are more potent than those containing E7 alone in the TC-1 cell line. 23 In addition, therapeutic vaccination targeting CD40 with HPV E6 and E7 showed protection against TC-1 tumors in human CD40 transgenic mice, and a significant increase of HPV16-specific CD8 + CTLs in the tumors. 24 Results of three proof-of-concept T cell vaccines currently in clinical trials for the treatment of cervical intraepithelial neoplasia (CIN) stage 2 or 3 have been published to date. 25-27 Inovio's DNA vaccine (VGX-3100) 25 is a mixture of two DNA plasmids encoding optimized HPV16/18 E6/E7 consensus sequences. GX-188E (Genexine Inc.) is a DNA vaccine consisting of a tissue plasminogen activator signal sequence, an Fms-like tyrosine kinase 3 (Flt3)-ligand, and an HPV16/18 E6/E7 fusion antigen. 26,28 The Tipapkinogen sovacivec (TS) vaccine is a modified vaccinia Ankara vectored vaccine encoding human cytokine interleukin (IL)-2, and non-oncogenic forms of HPV16 E6 and E7 proteins. 27 However, the observed clinical benefits are modest, possibly reflecting poor immunogenicity of the vaccine regimens and limited antigenic breadth of the immunogens.
It has been shown in HPV-driven squamous cell carcinoma that immune recognition of other viral antigens is highly relevant. Previous studies have demonstrated high mRNA levels and high titers of serum antibodies against E1, E2, E4, and E5, in addition to E6 and E7 in HPV-driven head and neck cancer. [29][30][31] Two recent studies indicate that HPV proteins other than E6 and E7 are relevant in T cell immune recognition of HPV-positive oropharyngeal cancer (OPC) and head and neck squamous cell carcinoma (HNSCC). Bhatt et al. profiled both CD4 + and CD8 + T cells that recognize E1, E2, E4, E5, and L1 proteins in patients with OPC, with E1 responses present in most of the patients interrogated in the study. Eberhardt et al. showed presence of HPV-specific T cells in HNSCC, and defined specificity to E2, E5, and E6 in HNSCC. 32 Since E1-specific T cell responses have been recently shown to be present in patients with cervical cancer, 33 we here set out to perform an unbiased analysis of the HPV antigenic landscape in cervical cancer tumor cell lines using immunopeptidomics to understand which viral proteins are processed and presented on the tumor surface to T cells.
Our results expand the known landscape of HPV-derived antigens presented on cervical tumor cells to cryptic, out-of-frame translated viral antigens and next to the known viral antigens E6 and E7, extending the source of viral antigens to the genes encoding E1 and E2.

RESULTS
Immunopeptidomic analysis yields highly specific HLA-associated peptide data for a panel of five cervical tumor cell lines   Figure 1A). More than 94% of the peptides had the typical length of HLA-I peptides (8-14 amino acids), with the majority of peptides (58%) being 9 amino acids long ( Figure 1B). We observed that the higher the HLA-I surface expression levels as measured by flow cytometry (median fluorescence intensity [MFI]), the higher the number of peptides identified in the experiment ( Figures 1C and  S1). Although the linear correlation between the MFI and peptide number was strong (R 2 = 0.65), it did not reach statistical significance (p = 0.1).
We also employed advanced neural-network-based algorithms (NetMHCpan v4.1) 38,39 to predict HLA binding of the identified peptide sequences. A range of 69%-88% of all peptides detected were predicted to bind to the HLA-A, -B, and -C alleles present in the respective cell lines (median 78%; Figure 1D). The obtained peptide sequences showed a low degree of overlap between the different cell lines, with more than 85% of the identified peptides being unique to one of the five cell lines, which is expected for samples with distinct HLA haplotypes ( Figure 1E). We observed a high degree of overlap of sequences for cells expressing the same HLA allele, such as DoTc2 and CaSki cells (HLA-A*03:01, 2,927 peptides in common), C33 and CaSki cells (HLA-A*02:01, B*07:02, C*07:02, 761 peptides in common), and SiHa and DoTc2 To confirm HPV status and establish expression levels of HPV-derived transcripts, we performed RNA sequencing across the cell line panel. All HPV16 genes with conventionally established expression (E1, E2, E4, E5, E6, E7, L1, and L2) were detected at the transcriptional level in DoTc2 and CaSki, whereas only transcripts from HPV16 E1, E2, E6, and E7 were detected in SiHa ( Figure 2A). Expression of HPV18 transcripts were detected exclusively in HeLa cells (E1, E2, E6, E7, and L1; Figure 2A), as expected. We observed evidence for transcription of E1 and E2 across all cell lines except in C33, which was negative for HPV transcripts as expected ( Figure 2A).

HPV18 E1 protein is expressed in addition to E6 and E7 in HeLa cells
The HeLa cell line is a commonly used proteomics standard for the interrogation of human proteome sequencing depth, 40 and, as such, there are many deep proteomics datasets available in the public domain. A dataset by the Olsen laboratory (PXD004452), 41 in which several enzymes were used for digestion to increase sequence coverage and depth, was selected and re-interrogated to validate expression of HPV proteins in HeLa cells. Our analysis confirmed protein expression of HPV18 E1, E6, and E7 proteins in HeLa cells, which had not been reported in the original publication ( Figure 2B), in line with the mRNA expression levels characterized in our RNA sequencing datasets.
E1 and E7 canonical and alternative reading frame (ARF)-derived viral peptide antigens are presented by HLA in cervical cancer cells To refine our search for HPV-derived HLA peptides, we reconstructed viral transcript sequences by de novo assembly of RNA reads using Trinity. 42 Following a first pass of mapping to the human reference genome, unmapped reads are used as input to Trinity for assembly. After filtration of the output transcript sequences by searching for partial overlaps with the sample-matched HPV genome using BLASTn, we performed sixframe translation of assembled transcripts and concatenated the resulting protein sequences over 8 amino acids in length to the SwissProt reviewed human reference proteome. Using this approach, we identified five peptides from the assembled viral transcripts that mapped to the HPV genome ( Figure 2C, Table 1).

Identified HPV peptides are recognized by T cells from women with persistent HPV infection or cervical intraepithelial neoplasia
To validate the identified sequences derived from viral transcripts for their potential to elicit immune responses in infected individuals, we generated short-term HPV-specific T cell lines (STCLs) from PBMC from seven women with CIN (grades 1-3) or colposcopic evidence of HPV changes without CIN (five HPV16-positive, and two HPV18-positive on vaginal sampling for HPV DNA). These STCLs were then tested for recognition of the cognate peptide sequences using cultured interferon (IFN)-g ELISpot assays. We detected a T cell response to each peptide sequence in at least 1 of the 7 patients, confirming that the mass spectrometry (MS)-identified peptides were immunogenic in the context of natural infection ( Figure 3). As a negative control, T cells from all seven women were simultaneously expanded using a pool of irrelevant peptides (FEC peptide pool sourced from influenza, Epstein-Barr virus, and

HPV E1 is consistently expressed in primary cervical tumor tissue
To confirm the clinical relevance of our finding that HPV E1 can be presented on cervical tumor cells, we performed RNA sequencing on 10 primary tumor tissue samples. We first determined the HPV genotype by mapping viral RNA reads to reference genomes reported in PAVE. 2 Reads were mapped to HPV16 in 7 tumor samples, and to HPV18, 31, and 45 in one sample each ( Figure 4A). We then analyzed the relative transcript expression levels for all HPV viral genes and found that only E1, E6, and E7 were consistently transcribed across all 10 primary tissue samples ( Figures 4A and 4B), whereas other early proteins and the structural proteins had more varying expression across the analyzed tissue cohort. L1 and L2 expression was lowest across the tissues, whereas E5 expression was only present in 6 of 10 tumor tissues, rendering it the least consistently expressed viral protein in this study. These data confirm the high relevance of E1 as a potential viral antigen target in primary HPV tumors, as it is highly expressed in cancer tissue from all four most frequent high-risk HPV subtypes, HPV 16, 18, 31, and 45.
HPV E1 and E2 genes are sources for non-canonical HLA-presented antigens in human cervical tumor tissue Finally, we performed immunopeptidomic profiling on the obtained cervical tumor tissues. We obtained between 8,439 and 24,298 peptides with a length of 8-14 amino acids per tissue ( Figures 4C and 4D), and between 56% and 86% of peptides were predicted to bind to the regarding HLA alleles present in   iScience Article each patient ( Figure 4D, patient HLA types listed in Data S2). We again performed Trinity de novo assembly to reconstruct the patients' individual viral transcript profile and to characterize the potential viral sequence pool. Within those peptide sequences that were predicted to be derived from HLA presentation (Figure 4E), we sequenced four peptides originating from the assembled viral transcripts ( Table 2). We were able to sequence a novel canonical peptide derived from HPV45-E6, EVPAVPETI, which was predicted to bind to HLA-A*25:01 (patient 3, Table 2). We also again observed the known HLA-A*02:01 epitope derived from HPV16-E7, YMLDLQPET (patient 5) (Table 2, Figure 4F). We further observed two additional peptides originating from E1 (SAEFPLQY, HLA-A*01:01) and E2 (AYVSIKLHFH, HLA-A*24:02) viral genes in patient 3, both peptides derived from a reverse-strand transcript spanning the E1 and E2 gene region ( Figures 4F and 4G, Table 2).
These data confirm the relevance of non-canonical viral antigens in cervical cancer, and underline the importance of their consideration in the development of therapeutic approaches to cervical cancer.

DISCUSSION
The first HLA class I-restricted HPV T cell epitopes published in the literature were identified through T cell immunogenicity screening in HLA-A*02:01-positive healthy donors of predicted peptide sequences from E6 and E7. 44 The list was expanded by B*18:01-bound sequences from E6 and E7, 45 and several HLA class I-bound E7 peptides were mapped in CaSki cells using a targeted liquid chromatography (LC)-MS 3 approach by screening a library of HPV16 E7-derived peptides that were predicted to bind to HLA-A2. 46,47 Such targeted MS approaches are highly sensitive but are limited by the need to establish a list of peptide candidates through prediction, and the need to synthesize these molecules for inclusion in the synthetic peptide library required for MS acquisition. The characterized E7 [11][12][13][14][15][16][17][18][19] (YMLDLQPET) in the latter study, which was also detected in our study, was later shown to be present in primary cervical cancer biopsies using the same targeted MS approach. 48 It was reported previously that cervical cancer cell lines exhibited antigen-processing defects for the E6 29-38 (TIHDIILECV) peptide; however, peptide-specific CD8 + T cells were able to kill CaSki indicating presentation of this peptide in CaSki cells. 49 This peptide was not identified in our study, likely due to the presence of a cysteine residue in the peptide. Cysteine is a highly reactive amino acid that has a high likelihood of undergoing modifications during sample preparation. It is well-documented that cysteine-containing peptides are often underrepresented in immunopeptidomic studies. [50][51][52] While profiling viral RNA expression levels in our cervical cell line panel, we detected expression of all main HPV proteins in DoTc2 cells, in contrast to earlier reports and current literature that suggest the cells to be HPV-negative. 53,54 We would like to suggest here to carefully re-test the cell line if used as an experimental model, and to regard the biological safety level as unknown until a full characterization is completed.
With our unbiased proteogenomics discovery approach combining next-generation RNA sequencing and immunopeptidomics, we report for the first time the direct discovery of HPV16 and HPV18-derived peptides presented by HeLa and CaSki cells. While confirming presentation of the known HLA-A*02:01 peptide YMLDLQPET derived from HPV16 E7 in CaSki cells, we further identified a peptide derived from an HPV16 ARF in the E1 gene transcript (MLYQMTRTK). In HeLa cells, we also detected two peptides derived from HPV18 E1, KQGAMLAVF and DTPEWIQRL, predicted to bind to B*15:03 and A*68:02, respectively, together with a peptide derived from an alternative reading frame of E7 (MKFRLTFY), which was mapped to B*15:03. It is to be noted that A*68:02 is an HLA-A2 supertype, suggesting that the ARF peptide could be iScience Article presented by A2-positive patients. We also confirmed the presence of HPV18 E1 protein in deep proteome data of HeLa cells, as previously shown by western blot. 55 These results demonstrate the importance of the E1 gene as a source for antigens presented in cervical cancer beyond E6 and E7. We confirmed that peptides are immunogenic in HPV-positive women, demonstrating the relevance of E1 T cell recognition in HPV infection and early disease. The observed HPV-specific T cell responses in PBMC from these individuals are consistent with their known HPV exposure and are thus likely to be mediated by effector and or memory T cells rather than naive T cells that have undergone differentiation in vitro. The generation of antigen-specific T cells from naïve precursors by in vitro priming typically requires stimulation by antigen-loaded mature dendritic cells, followed by in vitro expansion for a longer culture period than that employed here. These findings are also in close alignment with a study that investigated the frequency of E1-specific responses in peripheral blood in patients with cervical squamous cell carcinoma. 33 E1 is highly conserved across papillomavirus subtypes, making it an attractive therapeutic target. The protein is the only viral protein with a defined enzymatic activity and exhibits an ATP-dependent helicase function that recruits components of the cellular replication machinery for viral genome replication. 56,57 E1 can also directly downregulate the expression of interferons (IFNs), which are key mediators of the cellular antiviral immune response, and further promote the expression of IFN-stimulated genes that encode molecules with a multitude of antiviral effector functions. 58,59 To confirm the clinical relevance of E1 peptide presentation in primary human cervical tumors, we proceeded to profile RNA expression of viral transcripts and confirmed that E1 transcription is maintained in 10 of 10 primary human cervical tumor resections, including all four most frequent high-risk HPV subtypes (HPV16, HPV18, HPV31, and HPV45). Other early proteins were less consistently expressed across the analyzed cohort, suggesting a central role of E1 in HPV-driven tumors, and highlighting E1 as the most suitable additional viral protein target in addition to E6 and E7.
Finally, the relevance of ARF-derived viral peptides was confirmed in primary cervical tissue with the identification of two peptides derived from a reverse-sense transcript spanning the E1 and E2 gene region, suggesting that non-canonical transcription and translation events contribute to cervical tumor antigenicity in patients.
Together, we conclude that non-canonical antigens derived from HPV, most importantly from the viral gene E1, are important additional targets to be considered for future immunotherapeutic approaches to cervical cancer.
Limitations of the study LC-MS 2 remains a discovery technology, and the actual extent of viral antigens presented on cervical cancer cells and tissue remains to be determined. Although our study extends the known targetable viral iScience Article antigen space to the viral gene E1, and further highlights the presentation of viral antigens from ARFs and non-canonical antisense transcripts, the mechanism and contribution of such non-canonical viral transcription and translation events to viral antigen presentation in HPV-transformed cells and their suitability as targets for immunotherapy has yet to be fully characterized.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

DECLARATION OF INTERESTS
N.T. is or has been scientific advisor to Enara Bio, Grey Wolf Therapeutics, Roche Pharma, Infinitopes Ltd., and T-Cypher Bio. 17