Soluble HLA-associated peptide from PSF1 has a cancer vaccine potency

Partner of sld five 1 (PSF1) is an evolutionary conserved DNA replication factor involved in DNA replication in lower species, which is strongly expressed in normal stem cell populations and progenitor cell populations. Recently, we have investigated PSF1 functions in cancer cells and found that PSF1 plays a significant role in tumour growth. These findings provide initial evidence for the potential of PSF1 as a therapeutic target. Here, we reveal that PSF1 contains an immunogenic epitope suitable for an antitumour vaccine. We analysed PSF1 peptides eluted from affinity-purified human leukocyte antigen (HLA) by mass spectrometry and identified PSF179–87 peptide (YLYDRLLRI) that has the highest prediction score using an in silico algorithm. PSF179–87 peptide induced PSF1-specific cytotoxic T lymphocyte responses such as the production of interferon-γ and cytotoxicity. Because PSF1 is expressed in cancer cell populations and highly expressed in cancer stem cell populations, these data suggest that vaccination with PSF179–87 peptide may be a novel therapeutic strategy for cancer treatment.

HLA peptidome by mass spectrometry (MS). MS is a methodology to comprehensively analyse the repertoire of the HLA peptidome presented naturally in vivo. Each having inherent advantages and disadvantages, both in silico-based prediction and direct MS-based approaches led to successful identification of the HLA peptidome for the development of cancer immunotherapy.
The goal of the present study was to identify PSF1-derived peptides presented by HLA using two methodologies, in silico-based prediction and an MS-based approach. In addition, we examined whether the identified peptide could induce CTL responses from peripheral blood mononuclear cells (PBMCs) of healthy donors. Our results suggest that PSF1 peptide generated in vitro shows the characteristics of an effective peptide vaccine against CSCs.

Results
Strategy for identification of major histocompatibility complex (MHC) class I-presented epitopes from PSF1. We identified processed and presented PSF1-specific, MHC class I-restricted epitopes using an immunoproteomics approach (Fig. 1). Briefly, we established the MDA-MB-231 cell line expressing soluble HLA-A*02:01 (sHLA) molecule (hereafter MDA-MB-231-sHLA cells) as reported previously 21 . PSF1 protein expression was observed in the original MDA-MB-231 cells; however, to identify PSF1-derived peptide captured on HLA effectively, we overexpressed the PSF1 gene in MDA-MB-231-sHLA cells by transient transfection using the MaxCyte electroporation transfection system. To determine the optimal concentration of plasmid for transfection, cells were electroporated with varying amounts of plasmid. The protein expression level of PSF1 was increased several times (Supplementary Figure 1). Cell damage was observed in a dose-dependent manner (data not shown) and the optimal concentration was 125 µg/ml. Using this cell line, peptide/MHC complexes secreted in 3.6 L of culture medium were purified by an anti-19 amino acid collagen type II (CII) tag antibody 22 column. To identify human PSF1-specific epitopes that formed a complex with MHC class I, transient expression of human PSF1 was induced in MDA-MB-231-sHLA cells.
Subsequently, peptides bound to MHC class I molecules were isolated from the conditioned medium, and we identified YLYDRLLRI peptide (PSF1 79-87 peptide) as an epitope derived from PSF1 by MS analysis (Fig. 2A). We then confirmed the sequence identity of the epitope using synthetic analogues. The MS/MS spectra of the endogenous peptide was in strong agreement with that of the synthetic peptide (Fig. 2B). In addition, we predicted the HLA-binding affinities of PSF1 peptides using three types of in silico-based software: BIMAS (http://www-bimas. cit.nih.gov/molbio/hla_bind/), SYFPEITHI (http://www.syfpeithi.de/), and NetMHC (http://www.cbs.dtu.dk/ services/NetMHC/). As shown in Table 1, PSF1 79-87 peptide showed the highest score in all applied software (BIMAS and SYFPEITHI calculate binding scores, and NetMHC calculates IC 50 values). We thought that it is of special value in PSF1 79-87 peptide identified by two different methodologies: an MS-based experimental approach Figure 1. Immunoproteomics method work flow for identification and characterisation of PSF1-specific T cell epitopes. Strategy of antigen discovery by immunoproteomics for cancer immunotherapy. After immunoprecipitation of peptides by an anti-CII tag antibody column, sHLA-presented peptides were identified by tandem mass spectrometry. Adequacy of the identified peptides was further calculated by in silico prediction and validated by in vitro and in vivo immunogenic methods. and in silico-based prediction. Additionally, PSF1 79-87 peptide is a unique peptide compared with the other GINS component molecules (SLD5, PSF2, and PSF3).
Vaccination induces PSF1-specific CD8 + T cell responses. To assess whether the experimentally identified peptide stimulates splenocytes in vivo, we used CB6F1-Tg (HLA-A*02:01/H2-Kb) (hereafter HLA-A*02:01 transgenic) mice. The HLA-A*02:01 transgenic mice were immunised with PSF1 79-87 peptide in Montanide ISA 51VG to induce CTL activity. Two weeks after immunisation, splenocytes were harvested and co-cultured with T2 cells . Spectral differences in endogenous and synthetic peptide spectra may be due to co-eluting peptides in the highly complex sample. Although present, not all fragment ions are labelled because of overlapping peak labels. pulsed with PSF1 79-87 peptide or Mart-1 26-35 peptide as a control in an interferon (IFN)-γ enzyme-linked immunospot (ELISPOT) assay. We found that splenocytes produced a significant amount of IFN-γ in a peptide-specific manner only when co-cultured with T2 cells pulsed with PSF1 79-87 peptide ( Fig. 3A-E). At an effector cell/target cell (E/T) ratio of 20:1, these splenocytes showed 36 spot counts/well in response to T2 cells pulsed with PSF1 79-87 peptide, whereas they showed 2 spot counts/well in the presence of T2 cells with negative-control peptide loading ( Fig. 3E, P < 0.01). The trend was also observed significantly at an E/T ratio of 40:1. These observations strongly demonstrate the potential of PSF1 79-87 peptide to be presented by HLA-A*02:01.

Rank
Cytotoxic activity of PSF1 79-87 peptide-reactive CTLs from PBMCs of HLA-A2-positive healthy donors. As a next step to validate the potency of PSF1 79-87 peptide in the human immune system, we evaluated the peptide-specific immune responses of human CTLs. CTLs were generated by peptide stimulation of PBMCs isolated from healthy donors who were positive for HLA-A*02:01, and CD8 + T cells purified from PBMCs were co-cultured with monocyte-derived dendritic cells (MoDCs) pulsed with the peptide. To detect HLA-A*02:01-restricted and PSF1 79-87 -specific CTLs, we used a HLA-A*02:01-PSF1 79-87 peptide tetramer and observed the population of HLA-A*02:01/PSF1 79-87 peptide-tetramer + /CD8 + T cells (Fig. 4A).
To determine the functional cytotoxic activity of the peptide-specific CTLs, we tested their ability to kill T2 cells pulsed with PSF1 79-87 peptide. The peptide-specific CTL line killed 30.5% of PSF1 79-87 peptide-pulsed T2 cells and 12.9% of non-pulsed T2 cells ( Fig. 4B-E). The cytotoxicity of CTLs was induced in a dose-dependent manner of the E/T cell ratio (Fig. 4F). These results suggest that vaccination with PSF1 79-87 peptide could induce immune responses in the clinical setting.

Discussion
CSCs mediate tumour metastasis by their relative resistance to current therapies and contribute to relapse following treatment 9,10 . Many strategies that are not selective enough against CSCs can be toxic to healthy tissues, and patients are usually at risk of recurrence and metastasis because of the lack of CSC elimination 23 . Additionally, CSC populations are more resistant to conventional cancer therapies than non-CSC populations [24][25][26] . It was recently reported that CSCs might be immunogenic in certain settings. Thus, they have the potential to stimulate specific immune responses that can target and remove tumour CSCs in cancer patients 27 . We believe that novel therapeutic strategies by vaccination selectively targeting CSCs may be of the greatest therapeutic benefit. Our previous study suggested that PSF1 is highly expressed by malignant cancer cells, which are defined as CSCs, and that this molecule has important roles in the survival and proliferation of cancer cells 16 . Here, we identified PSF1-derived peptides presented by HLA by direct (MS-based) and indirect (in silico-based) methods (Fig. 1).
Considering that the class I HLA molecule is a cell surface protein with intracellular peptide epitopes, many indirect methods have been used to discover HLA-associated tumour rejection antigens. A small number of tumour antigens were revealed by purification of HLA-associated peptides from cell lysates. Thus far, identification of HLA-binding peptides from cell lysates has an innate limitation associated with detergent in the MS-based approach. To overcome the problem, we used secreted class I HLA molecules for peptide identification without detergent lysis 28 . In this study, we purified peptides associated with sHLA from conditioned medium of the MDA-MB-231 cell line expressing sHLA and human PSF1, and identified PSF1 79-87 peptide by liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS). We identified other proteins, such as histone, cytoskeletal proteins (actin and tubulin), and cytosolic chaperones (heat shock proteins). However we found no other cancer vaccine target proteins other than PSF1.
Comprehensive lists of peptide sequences predicted to associate with HLA can be obtained by in silico algorithms. It is one of the most powerful tools to reveal the peptidome that represents HLA-binding peptides. However, because of the intrinsic and unavoidable high number of false positives among the predicted epitopes, and the fact that these algorithms predict binding affinity and not cellular processing or immunogenicity 29 , supporting in vitro experiments are required to screen for the actual presented and immunogenic peptides. We carried out a preliminary analysis to compare CTL activity with binding affinity to HLA in silico in a mouse model, and found that it is difficult to predict the relationship between them. In the present study, we used a bioinformatics approach based on BIMAS, SYFPEITHI, and NetMHC algorithms to predict HLA-binding peptides derived from PSF1 (Table 1). Remarkably, the results obtained by MS-based analysis were confirmed by all in silico analyses. Currently, MS-based methods continue to drive significant improvements in sensitivity and throughput for discovery of HLA-binding peptides. We believe that, as larger data sets become available, predictive algorithms will be used to ensure the potential immunological significance of directly identified HLA ligands for further study, as shown in this report.
Several studies have reported the use of molecules expressed in CSCs as sources of antigens for vaccine development [30][31][32][33] . However, there are still few peptide vaccine targets in CSCs. For example, survivin, which is the smallest member of the inhibitor of apoptosis family of proteins, is a known marker of CSCs and important target for cancer vaccines and therapeutics 34 . Therapeutic outcomes using survivin as a vaccine have not been clarified in terms of killing CSCs clinically. In this study, we determined whether injection of PSF1 79-87 peptide would cause immunological responses in vivo. We used HLA-A*02:01 transgenic mice to induce CTL activity. HLA-A*02:01 transgenic mice, which express a chimeric gene consisting of the α1 and α2 domains of human HLA-A*02:01 and the α3 domain of murine H-2K b , have been reported to be a versatile animal model for preclinical evaluation of peptide-based immunotherapy [35][36][37] . In our analysis, we verified immunological responses induced by PSF1 79-87 peptide using this mouse model. Moreover, we confirmed that the CTLs induced by peptide-loaded MoDCs exerted a significant cytotoxic effect against PSF1 79-87 peptide-expressing cancer cells. Although the population was not very large, these CTLs significantly killed T2 cells pulsed with PSF1 79-87 peptide. Therefore, we believe that these CTL responses were effective and specific to PSF1 79-87 peptide, and it is possible that PSF1 79-87 peptide can target highly PSF1-expressing malignant cancer cell populations, i.e., CSCs. PSF1 is expressed in immature cell populations and stem cell populations [e.g., epiblasts during embryogenesis, bone marrow hematopoietic stem cell populations, sperm stem cells (spermatogonia)] in mice 11,12 . However, the PSF1 expression levels of these normal cells are extremely lower compared with cancer cells. Moreover, it is not known whether the PSF1 79-87 peptide binds to HLA expressed on normal stem cells. It will be necessary to compare captured PSF1 79-87 peptide levels in normal stem cells and cancer stem cells in vivo in tumours in future research.
In summary, we identified a candidate cancer vaccine peptide for CSCs, PSF1 79-87 peptide, derived from PSF1. We demonstrated that PSF1 79-87 peptide induces CTL responses in vitro and in vivo in an HLA-A*02:01-restricted manner. This report is the first to reveal presented PSF1-specific T cell epitopes identified by direct (MS) and indirect (in silico) methods. Vaccination with the peptide identified in this study may provide a new cancer immunotherapy for targeting CSCs. Prediction of HLA-A*02:01-presented immunogenic 9-mer peptides. The PSF1 sequence was subjected to analysis by computerised HLA-binding prediction based on freely accessible online databases, BIMAS, SYFPEITHI, and NetMHC. All programs provide peptide sequences that are likely to be presented by the selected HLA molecules together with a ranking, score, or affinity. The top five peptides in each analysis were chosen.
Purification and identification of PSF1 epitope peptides. Peptide-HLA-A*02:01 complexes were purified according to a previous report 39 with minor modifications. A total of 3.6 L of collected culture supernatant (1.2 × 10 9 cells) was pre-cleared with inactivated AminoLink Plus Coupling Resin (Thermo Fisher Scientific) to remove nonspecific proteins, followed by immunoprecipitation with anti-CII mouse monoclonal antibody (6G4)-immobilised AminoLink Plus Coupling Resin (Thermo Fisher Scientific). After a series of wash steps, the interacted epitope peptides were dissociated from HLA molecules using 10% acetic acid at 90 °C for 5 min and then passed through a 10 kDa molecular mass cutoff filter (Millipore Corp, Bedford, MA). The isolated peptides were fractionated using a MonoSpin SCX spin column (GL Sciences Inc., Tokyo, Japan) according the manufacturer's instructions. The eluate from 350 mM NaCl was desalted using a MonoSpin C18 spin column (GL Sciences Inc.) according to the manufacturer's instructions and dried using a savant SPD2010 SpeedVac system (Thermo Fisher Scientific).

MS data analysis.
All volunteers provided informed consent that was approved by the Ethics Committee of Shionogi & Co., Ltd. All procedures were approved by the Committee and all experiments were performed in accordance with the Committee's guidelines. MoDCs were generated by in vitro culture as described previously 45,46 . Briefly, PBMCs isolated from healthy volunteers positive for HLA-A*02:01 using Ficoll-Plaque solution (GE Healthcare UK, Buckinghamshire, UK). CD8 + or CD14 + cell populations were purified from PBMCs with microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany). CD14 + cells were subjected to differentiation induction to MoDCs, whereas CD8 + cells were frozen for storage. CD14 + cell population was cultured in the presence of 100 ng/ml granulocyte-macrophage colony-stimulating factor (R&D Systems Inc., Minneapolis, MN) and 10 ng/ml interleukin (IL)-4 (R&D Systems) in AIM-V culture medium containing 2% heat-inactivated autologous plasma in two RepCell plates (CellSeed Inc., Tokyo, Japan). After 5 days of culture, 0.1 KE/ml OK-432 (Chugai Pharmaceutical Co., Ltd., Tokyo, Japan) was added to one plate and pulsed with 20 µg/ml PSF1 79-87 peptide. On day 6, these peptide-pulsed MoDCs were irradiated with X rays (30 Gy) and mixed with thawed CD8 + T cells. These cultures were prepared in 96-well plates with each well containing 1 × 10 4 peptide-pulsed MoDCs, 1 × 10 5 CD8 + T cells, 20 U/ml IL-2 (Shionogi & Co., Ltd., Osaka, Japan) and 10 ng/ml IL-7 (R&D Systems) with 2% autologous plasma. On day 12, the remaining CD8 + T cells were stimulated with peptide-pulsed autologous MoDCs. We prepared MoDCs each time in the same manner as described above. On day 18, CD8 + T cells were treated with 100 ng/ml IL-15 (Miltenyi Biotec). The antigen-specific responses of the CTLs were investigated using a cytotoxicity assay on day 26 47 .
Staining of HLA-A*02:01-restricted PSF1 79-87 CTLs. Cells were incubated with a phycoerythrin (PE)-labelled HLA-A*02:01 PSF1 79-87 peptide-binding HLA tetramer (PSF1 tetramer) (Medical & Biological Laboratories, Nagoya, Japan) and then stained with an allophycocyanin (APC)-labelled anti-human CD8 monoclonal antibody (BD Biosciences). The cells were analysed by a MACSQuant device (Miltenyi Biotec). HLA-A*02:01-restricted PSF1 79-87 CTLs were defined as CD8 + /PSF1-tetramer + T cells 48 . Cytotoxicity assay. CTLs derived from PBMCs were co-cultured with peptide-pulsed/unpulsed T2 cells at several E/T ratios. In brief, T2 cells were pulsed with 20 µg/ml PSF1 79-87 peptide overnight at 37 °C in AIM-V culture medium. As target cells for use in the cytotoxicity assay, 1 µM CellTracker ® Green 5-chloromethylfluorescein diacetate (CMFDA) (Invitrogen) was added to T2 cells, and then these cells were allowed to react with each other at 37 °C for 15 min. The labelled target cells were rinsed three times, seeded at 1 × 10 4 cells per well in a 96-well half area plate (Corning, Tokyo, Japan), and cultured with PSF1 79-87 CTLs in a final volume of 200 µl. After 5 h of incubation, the cells were subjected to an Annexin V Apoptosis Detection Kit APC (eBioscience, San Diego, USA) at room temperature for 15 min in the dark and then immediately analysed by the MACSQuant. Cellstain DAPI solution (0.1 µg/ml; Dojindo Laboratories, Kumamoto, Japan) was added to distinguish live or dead cells. To quantify apoptotic target cells, the percentages of CMFDA-positive and Annexin V/DAPI double-positive cells were calculated 49 .