Antibody-mediated delivery of a viral MHC-I epitope into the cytosol of target tumor cells repurposes virus-specific CD8+ T cells for cancer immunotherapy

Redirecting pre-existing virus-specific cytotoxic CD8+ T lymphocytes (CTLs) to tumors by simulating a viral infection of the tumor cells has great potential for cancer immunotherapy. However, this strategy is limited by lack of amenable method for viral antigen delivery into the cytosol of target tumors. Here, we addressed the limit by developing a CD8+T cell epitope-delivering antibody, termed a TEDbody, which was engineered to deliver a viral MHC-I epitope peptide into the cytosol of target tumor cells by fusion with a tumor-specific cytosol-penetrating antibody. To direct human cytomegalovirus (CMV)-specific CTLs against tumors, we designed a series of TEDbodies carrying various CMV pp65 antigen-derived peptides. CMV-specific CTLs from blood of CMV-seropositive healthy donors were expanded for use in in vitro and in vivo experiments. Comprehensive cellular assays were performed to determine the presentation mechanism of TEDbody-mediated CMV peptide-MHC-I complex (CMV-pMHCI) on the surface of target tumor cells and the recognition and lysis by CMV-specific CTLs. In vivo CMV-pMHCI presentation and antitumor efficacy of TEDbody were evaluated in immunodeficient mice bearing human tumors. TEDbody delivered the fused epitope peptides into target tumor cells to be intracellularly processed and surface displayed in the form of CMV-pMHCI, leading to disguise target tumor cells as virally infected cells for recognition and lysis by CMV-specific CTLs. When systemically injected into tumor-bearing immunodeficient mice, TEDbody efficiently marked tumor cells with CMV-pMHCI to augment the proliferation and cytotoxic property of tumor-infiltrated CMV-specific CTLs, resulting in significant inhibition of the in vivo tumor growth by redirecting adoptively transferred CMV-specific CTLs. Further, combination of TEDbody with anti-OX40 agonistic antibody substantially enhanced the in vivo antitumor activity. Our study offers an effective technology for MHC-I antigen cytosolic delivery. TEDbody may thus have utility as a therapeutic cancer vaccine to redirect pre-existing anti-viral CTLs arising from previously exposed viral infections to attack tumors.


Background
Cancer immunotherapy based on cytotoxic CD8 + T lymphocyte (CTL)-mediated tumor recognition and elimination has shown remarkable anticancer efficacy [1]. Nonetheless, the suppression or absence of antitumor CTLs in the tumor microenvironment (TME) and tumor immune escape (evasion of an antitumor CTL response) limit the clinical efficacy in many patients with solid tumors [1,2]. CTLs recognize cancer cells through the T cell receptor (TCR) specific to tumor-derived antigenic peptides presented by major histocompatibility complex class I (MHC-I; HLA class I in humans) [2]. Personalized therapeutic cancer vaccines are designed to deliver tumor-specific T cell epitopes, so-called neoantigens, in various forms (synthetic long peptides, mRNA, and DNA) with the intent of inducing de novo tumor-specific CTL responses and/or amplifying the endogenous tumor-specific CTL responses [1][2][3]. This approach has shown promising antitumor activity in early clinical trials for melanoma and other cancers [2,4]. However, some of the challenging aspects of neoantigen-based cancer vaccines include the difficulty in identifying and selecting immunogenic neoantigens, the downregulation or loss of the MHC-I alleles on tumor cells, the poor homing of neoantigen-specific CTLs into TME, and their personalized nature, which may not be suitable for all patients [1][2][3]. Remarkably, antiviral CTLs specific to human viruses (that have previously infected the host), such as human cytomegalovirus (CMV), Epstein-Barr virus, and influenza virus, have been found to infiltrate various solid tumors abundantly [5,6]. The tumor-infiltrating antiviral CTLs cannot attack tumor cells because of the lack of specific recognition of the tumor through the TCR. Nevertheless, repurposing of some antiviral bystander CTLs arising from common human viral infections to attack tumors holds great potential for cancer immunotherapy due to their abundance, high potency, and specificity to common viruses [7].
To utilize pre-existing antiviral CTLs to attack tumors, tumor cells should first be marked with a viral peptide-MHC-I complex (pMHCI) on the cell surface, to be recognized by the antiviral CTLs. To this end, some approaches, such as extracellular surface loading [6,[8][9][10] and endosomal loading [11] of a viral CTL epitope peptide onto MHC-I, have been explored. All the existing strategies deviate from the conventional MHC-I antigen presentation pathway [12], wherein cytosolic localization of viral antigens is critical for intracellular processing intended to form mature MHC-I epitopes (8)(9)(10)(11) amino acid residues in length) for efficient binding to MHC-I in the endoplasmic reticulum (ER), followed by surface presentation in the form of pMHCI. Few studies have addressed the delivery of MHC-I-restricted CTL epitopes into the cytosol of tumor cells, mainly due to the dearth of efficient cytosolic delivery tools. To overcome this limitation, we envisioned that a tumor-specific cytosolpenetrating antibody (Ab) that we recently documented, dubbed the inCT cytotransmab [13][14][15], can serve as a carrier for cytosolic delivery of MHC-I-specific antigenic peptides into target tumor cells. In human immunoglobulin G1 (IgG1/κ) format, inCT can access the cytosol of target tumor cells after endocytotic internalization via a tumor cell-associated receptor (integrin αvβ3 or αvβ5; mainly integrin αvβ5 on epithelial cancer cells), followed by endosomal escape into the cytosol [13][14][15].
In this study, we sought to direct CMV-specific CTLs against tumors because CMV infection is very common among healthy adults (60-90% of the population, with higher infection rates with increasing age) [16] and is characterized by accumulation and maintenance of CMV-specific CTLs with a majority of effector memory phenotypes: a phenomenon termed memory inflation [17,18]. In CMV-seropositive hosts, the 65 kDa phosphoprotein (pp65) antigen-derived 9-mer peptide 495 NLVP-MVATV 503 (residues 495-503, hereafter referred to as the "CMVp 495-503 peptide") is the most immunogenic CTL epitope among CMV antigens [19] and is predominantly displayed on HLA-A*02:01 [20,21], which is the most prevalent MHC-I variant in the human population [22]. CTLs specific to the CMVp 495-503 -HLA-A*02:01 complex (hereafter referred to as "CMV-pMHCI") are present in the blood with high prevalence and functional competence among both HLA-A*02:01 + healthy donors and cancer patients [9,23,24]. CMV-pMHCI-specific CTLs (hereafter referred to as "CMVp-CTLs") efficiently kill tumor cells displaying CMV-pMHCI on the surface [25], and therefore, are an attractive tool for cancer immunotherapy.
Here, to deliver a viral MHC-I epitope peptide into the cytosol of target tumor cells for converting them into virally infected cells, we engineered a CD8 + T cell epitope-delivering Ab, termed a TEDbody, by genetic fusion of the viral MHC-I antigen peptide to the inCT cytotransmab. The TEDbody efficiently delivered the payload of a CMVp 495-503 -encompassing peptide into the target tumor cells, such that it was intracellularly Keywords: MHC-I epitope cytosolic delivery, Cytosol-penetrating antibody, Peptide-MHC-I complex, Anti-viral cytotoxic T lymphocytes, Cytomegalovirus therapeutic cancer vaccine processed for surface presentation by the cognate MHC-I, thereby rendering the marked tumor cells recognizable and killable by the corresponding antiviral CMVp-CTLs. Furthermore, we showed that systemic injection of a TEDbody carrying the CMVp 495-503 -encompassing peptide into immunodeficient mice bearing preestablished human tumor xenografts recruits adoptively transferred CMVp-CTLs into the tumor, thus substantially inhibiting tumor growth. The potency of this approach was further augmented when combined with a CTL-stimulating anti-OX40 agonistic Ab.

Cell lines
The human cell lines [breast adenocarcinoma (MDA-MB-231 cells), colorectal carcinoma (HCT116 and LoVo cells), and small cell lung cancer (NCI-H889 cells)] were purchased from the Korean Cell Line Bank (Korea) and maintained in RPMI 1640 (HyClone, Korea). HEK293FT cells purchased from Invitrogen were maintained in Dulbecco's modified Eagle's medium (DMEM, Hyclone). All cells were cultured in a growth medium that was supplemented with 10% heat-inactivated fetal bovine serum (FBS; HyClone), penicillin (100 U/ml), streptomycin (100 μg/ml; Welgene), and amphotericin B (0.25 μg/ml; HyClone). All cell lines were authenticated by DNA short tandem repeat profiling (ABION CRO, Korea) and used within 10 passages. All the cell lines were maintained at 37 °C in a humidified 5% CO 2 incubator and were routinely screened for Mycoplasma contamination (CellSafe, Korea).

Reagents and abs
The peptides used in this study were synthesized (> 95% purity) by AnyGen (Gwangju, Korea) and are listed in Table S1 with the sequence information. Chemicals, protein reagents, and Abs used as reagents in this study are listed in Table S2.
For the expression of TEDbodies and Abs, the plasmids encoding HC and LC were transiently cotransfected in pairs at an equimolar ratio into cultured HEK293F cells in the FreeStyle 293F medium (Invitrogen) following the standard protocol [14]. Culture supernatants were collected after 6 to 7 d by centrifugation and filtration (0.22 μm, polyethersulfone; Corning, CL S43118). TEDbodies and Abs were purified from the culture supernatants on a protein A-agarose chromatographic column (GE Healthcare) and extensively dialyzed to switch the solution to histidine buffer (25 mM histidine, pH 6.5, 150 mM NaCl) [14,29].
Before cell treatment, all the purified TEDbodies, Abs, and proteins were sterilized using a cellulose acetate membrane filter (0.22 μm; Corning) and Mustang Q membrane filter (0.8 μm; Pall, MSTG25Q6). Protein concentration was determined with the Bicinchoninic Acid (BCA) Kit (Thermo Fisher Scientific) and by measuring the absorbance at 280 nm using the molar extinction coefficient calculated from the primary sequence.

Peripheral blood mononuclear cell (PBMC) preparation and ex vivo expansion of CMVp-CTLs
PBMCs from healthy donors were acquired using protocols approved by the Institutional Review Board of Ajou University (approval ID: 201602-HM-001-01). All donors provided written informed consent before blood collection into a BD Vacutainer (BD Biosciences, 367,874). PBMCs were isolated using Ficoll-Paque Plus (GE Healthcare, 17-5442-03) density gradient centrifugation [29]. For long-term storage, PBMCs were resuspended with 10% FBS in DMSO and stored in liquid nitrogen at 1-5 × 10 6 cells/ml [9]. For in vitro expansion of CMVp-CTLs, 4 × 10 6 PBMCs (2 × 10 6 cells/ml) were stimulated for 3 d at 37 °C with 5 μg/ml CMVp 495-503 peptide in medium A [X-vivo medium supplemented with 2% of heat-inactivated human serum (Sigma)] in a 14-ml U-bottom tube (SPL). Then, the PBMCs were cultured for 3 d at 37 °C in medium B [medium A plus 200 IU/ml IL-2 (Peprotech)]. Next, the PBMCs were cultured every 2 to 3 d in medium C (medium B plus 0.5 nM IL-15/IL-15Rα-Fc protein) for up to 2 weeks. To evaluate the prevalence and phenotype of CMVp-CTLs, PBMCs before and after the ex vivo expansion were monitored by flow cytometric analysis involving double staining with a monoclonal Ab specific to CD8α (HIT8a) and a phycoerythrin (PE)-conjugated CMVp 495-503 -HLA-A*02:01 pentamer. At least two hundred thousand events were collected using a FACSCalibur flow cytometer (Becton Dickinson). Only PBMCs containing > 50% of CMVp-CTLs among all ex vivo-expanded PBMCs were used in in vitro cytotoxicity assays and in vivo adoptive transfers. The E:T ratio was calculated based on the prevalence of ex vivo-expanded CMVp-CTLs among the PBMCs.

Flow cytometry
To detect CMV-pMHCI complex formed on the cell surface at 4 °C, the indicated cells (4 × 10 5 ) were incubated at 4 °C for 3 h with a synthetic peptide, TEDbody, or control Ab at the indicated concentrations. For the detection of CMV-pMHCI on the cell surface at 37 °C, the indicated cells (1.5 × 10 5 ) were seeded in a 12-well plate in the medium containing 10% FBS and cultured until they were fully attached to the bottom of the plates (~ 12 to 15 h). Next, the cells were treated for 18 h with a synthetic peptide, TEDbody, or control Ab at the indicated concentrations, washed with PBS, and stained with the CMV-pMHCI-specific C1-17 Ab (10 nM) [27] in a blocking solution [PBS (pH 7.4) and 2% FBS] for 1 h at 4 °C. After a wash with the ice-cold blocking solution, the cells were stained with an Alexa Fluor 647-labeled rabbit antimouse IgG Ab (Thermo Fisher Scientific) for 30 min at 4 °C. After a wash with 1 ml of ice-cold PBS, twenty thousand events were collected using the FACSCalibur flow cytometer. Flow cytometric data were analyzed using the FlowJo v10 software (Tree Star) to calculate mean fluorescence intensity (MFI) and geometric MFI (gMFI). Fold changes in the gMFI of CMV-pMHCI presentation on the cell surface were calculated by normalization to that of the cells stained solely with the secondary Ab.
To identify the memory phenotype of CMVp-CTLs and determine the cell surface expression of PD1, OX40, LAG-3, TIGIT, and CXCR3 on CMVp-CTLs, the cells were stained with fluorescently labeled primary Abs specific to CD45RA, CCR7, PD1, OX40, LAG-3, TIGIT, CXCR3, or the respective isotype control Abs [31]. At least two hundred thousand events were collected using the FACSCalibur flow cytometer. The cell surface expression of HLA-A*02, integrin αvβ3, integrin αvβ5, and PD-L1 on human cancer cells, as well as the binding ability of pembrolizumab (Keytruda ® ) and anti-OX40 1166/1167 Abs toward HEK293FT-PD1 and HEK293FT-OX40, were determined using the respective primary Ab or isotype control Abs and then an appropriate secondary Ab, according to the standard protocol [29,31,32]. For intracellular staining of IFN-γ, cells were activated with phorbol 12-myristate 13-acetate (100 ng/mL) plus ionomycin (500 ng/mL) in a humidified incubator with 5% CO 2 at 37 °C for 10 h and further incubated for 6 h with brefeldin A (BD Biosciences, 1:1000) to prevent protein transport from the ER to the Golgi apparatus. All intracellular staining was performed using the BD Cytofix/Cytoperm Kit (BD Biosciences, cat. # 554714). Twenty thousand events were collected using the FACS-Calibur flow cytometer. All flow cytometry experiments were performed at least three times independently, the data were analyzed using the FlowJo v10 software, and representative data are shown, unless otherwise stated.

In vitro tumor cell lysis by ex vivo-expanded CMVp-CTLs
To evaluate the lysis of TEDbody-treated target cells by ex vivo-expanded CMVp-CTLs, cancer cells (5 × 10 3 ) were seeded in a 96-well plate in the medium containing 10% of FBS and cultivated at 37 °C and 5% CO 2 until they were fully attached to the bottom of the plates. After 12 to 15 h, the cells were treated with a synthetic peptide, TEDbody, or control Ab at the indicated concentrations for 12 h, washed with the medium, and cocultured for 18 h with ex vivo-expanded CMVp-CTLs at an E:T ratio of 5:1 (unless specified otherwise). In experiments with inhibitors, the cells were pretreated for 1 h with MG132 (Thermo Fisher Scientific) or cotreated for 8 h with either ERAP1-IN-1 (Chem Scene) or brefeldin A and with a synthetic peptide, TEDbody, or control Ab. Supernatants were employed to assess target cell lysis by lactate dehydrogenase (LDH) measurement and to evaluate T cell activation by an IFN-γ secretion assay. The LDH release was measured using the Cyto96 Non-Radio Cytotoxicity Assay (Promega) [29]. Absorbance was read at 492 nm using a Cytation 3 imaging multimode reader (Biotek). The maximum LDH release was determined by lysing the target cells with 1% Triton X-100 (Promega). The percentage of tumor cell lysis was calculated according to the following formula [29,32]: tumor cell lysis (%) = 100 × [(LDH release with peptide or TEDbody or control Ab treatment minus spontaneous LDH release of target cells)/(maximum LDH release minus spontaneous LDH release of target cells)]. IFN-γ secretion into the supernatant was determined using the ELISA Ready-SET-GO Kit (Thermo Fisher Scientific). Absorbance was read at 450 nm on the Cytation 3 imaging multimode reader.

Real-time cell lysis assays
Real-time kinetics of cell lysis were examined under a Lionheart FX automated microscope (BioTek Instruments) equipped with full temperature and CO 2 control to maintain 37 °C and 5% CO 2 . Ex vivo-expanded CMVp-CTLs were stained with red fluorescent dye PKH26 (Sigma-Aldrich). MDA-MB-231-EGFP cells (5 × 10 3 ) were seeded in a 96-well black clear-bottom plate (Greiner) in the medium containing 10% of FBS and cultured at 37 °C and 5% CO 2 until they were fully attached to the bottom of the plates. After 15 h, the cells were treated for 12 h with 1 μM synthetic peptide, TEDbody, or control Ab, washed with the medium, and cocultured with PKH26-labeled CMVp-CTLs at an E:T ratio of 3:1 inside the Lionheart FX automated microscopy system. Images were captured every 1 h for up to 14 h in triplicate via a 10× objective. All cells were photographed in the bright-field channel, MDA-MB-231-EGFP cells in the GFP channel, and PKH26-labeled CMVp-CTLs were photographed in the TRITC channel. Nine photos per well were taken and stitched to cover the center of the well [23]. Quantification of fluorescence intensity from the total area of MDA-MB-231-EGFP cells was performed in the Gen5 software (BioTek). To adjust the data for differences in the initial cell number across the wells, fluorescence intensity from the total cancer cell area at each time point was normalized to that at the initial time point (0 h). This normalized fluorescence intensity from the total cancer cell area is referred to as the cell index in this article.

Confocal immunofluorescence microscopy
Intracellular CMV-pMHCI induced by a TEDbody was visualized by confocal microscopy with a CMV-pMHCIspecific C1-17 Ab that was conjugated with DyLight 550 using the DyLight 550 Ab Labeling Kit (Thermo Fisher Scientific) [14]. Briefly, MDA-MB-231 cells (5 × 10 4 ) grown on cell culture slides (SPL) were treated with a TEDbody or a control Ab for 18 h. After two washes with PBS, the cells were fixed with 2% paraformaldehyde in PBS for 10 min at 25 °C, permeabilized with 0.1% Triton X-100 in PBS for 10 min at 25 °C, blocked with 2% bovine serum albumin in PBS for 1 h at 25 °C, and then incubated with the DyLight 550-labeled C1-17 Ab (20 nM) for 1.5 h at 25 °C to stain intracellular CMV-pMHCI. Early endosomes and the Golgi apparatus were also stained with a mouse anti-early endosome antigen 1 (EEA1) Ab and a mouse anti-FTCD (58 K-9) Golgi protein Ab, respectively, for 1.5 h at 25 °C [33]. Next, the cells were washed and further incubated with a goat anti-mouse Ab conjugated with Alexa Flour 488 (secondary Ab; Thermo Fisher Scientific) for 1.5 h at 25 °C. Nuclei were stained with Hoechst 33342 in PBS for 5 min at 25 °C. After mounting of the coverslips onto glass slides with the Fluorescence Mounting Medium (Dako), center-focused single z-section images were captured using a Zeiss LSM 710 system with the ZEN software (Carl Zeiss).

Mice
All animal experiments were approved by the Animal and Ethics Review Committee of Woojung Bio Inc. (Suwon, Korea) and performed in accordance with the guidelines established by the Institutional Animal Care and Use Committee. The approval ID for using the animals was IACUC2003-004 at the Animal Facility of Woojung Bio. Immunodeficient NSG mice (NOD.Cg-Prkdc scid IL2rg tm1Wjl/ SzJ) were originally obtained from the Jackson Laboratory and bred and maintained at the Animal Facility of Woojung Bio. Female C57BL/6 mice were purchased from Orient Bio (Korea) and allowed to reach 5-6 weeks of age before tumor inoculation.

In vivo CMV-pMHCI presentation and activation of tumor-infiltrating CMVp-CTLs
Female NSG mice (4-6 weeks old) received an orthotopic injection of MDA-MB-231 cells (5 × 10 6 per mouse) in 150 μl of a 1:1 mixture of PBS and Matrigel (BD Biosciences) into the mammary fat pad. When the mean tumor volume reached approximately 100-120 mm 3 , the mice were randomly assigned to a treatment group and intraperitoneally (i.p.) injected with the TEDbody or a control Ab. If necessary, at 6 h after the TEDbody or control Ab injection, all mice were peritumorally injected with 5 × 10 6 ex vivo-expanded CMVp-CTLs. After 24 h, the tumors were excised for immunohistochemistry (IHC) staining and an analysis of tumor-infiltrating CMVp-CTLs. IHC analysis of the tumor tissues excised from the mice was performed using the Zeiss LSM 710 system, as described before [14,15]. The CMV-pMHCI complex was detected using DyLight 550-conjugated C1-17 Ab. Nuclei were stained with Hoechst 33342 for 5 min at 25 °C. After the tissue sections were washed three times with 0.1% Triton X-100 in PBS and mounted on slides with the Perma Fluor aqueous mounting medium, center-focused single z-section images were obtained on the Zeiss LSM 710 system. Quantitative analysis of the images was performed using ImageJ software (National Institutes of Health) [14].
For the analysis of tumor-infiltrating CMVp-CTLs, single-cell suspensions were prepared by mechanical dissociation of the tumors through a 70-mm wire-mesh screen. To determine the number of tumor-infiltrating CMVp-CTLs, the cells were counted using a hemocytometer, and we analyzed the prevalence of CMVp-CTLs labeled with a monoclonal Ab specific for CD8α and the PE-conjugated CMVp 495-503 -HLA-A*02:01 pentamer by flow cytometry. To identify the functional phenotype of CMVp-CTLs, the cells labeled with the monoclonal Ab specific for CD8α and the PE-conjugated CMVp 495-503 peptide-HLA-A*02:01 pentamer were analyzed for T cell activation markers (CD69 and IFN-γ), as well as CD107a and granzyme B, for assessment of T cell cytotoxic function via flow cytometry. At least two hundred thousand events were collected using the FACSCalibur flow cytometer.

In vivo antitumor experiments
For the MDA-MB-231 orthotopic xenograft tumor model, 4-to 6-week-old female NSG mice received an orthotopic injection of MDA-MB-231 cells (5 × 10 6 cells per mouse) into the mammary fat pad. To set up the HCT116 xenograft tumor model, 4-to 6-week-old male NSG mice were inoculated subcutaneously, in the right thigh, with HCT116 cells (5 × 10 6 per mouse). All cancer cells were injected in 150 μl of a 1:1 mixture of PBS and Matrigel (BD Biosciences). The appropriate number of mice per group (sample size) for comparison of multiple groups by one-way analysis of variance (ANOVA) was calculated according to the resource equation method using the following formula: n = (DF/k) + 1, where n = number of mice per group, DF = degrees of freedom with acceptable range between 10 and 20, and k = number of groups [34]. When the mean tumor volume reached approximately 100-120 mm 3 , the mice were randomly assigned to treatment groups, and the TEDbody or a control Ab, in addition to the IL-15/IL-15Rα-Fc protein (15 μg), was i.p. injected every 3 d in a dose/weightmatched manner (20 mpk). In a combination experiment, inCT †CMVp 480-516 (20 mpk) combined with either the anti-PD1 Ab (5 mpk) or the anti-OX40 Ab (5 mpk) was i.p. injected, in addition to the IL-15/IL-15Rα-Fc protein (15 μg), every 3 d in a dose/weight-matched manner. At 6 h after the injection of the TEDbody, control Ab, or their combination, each mouse was injected in the tail vein with 10 7 CMVp-CTLs (ex vivo-expanded cells derived from human PBMCs), every 6 d for the indicated period. Tumor volume (V) was evaluated using digital calipers and was estimated using the formula V = L × W 2 /2, where L and W are the long and short dimensions of a tumor, respectively [15]. Tumor growth inhibition (TGI) caused by the TEDbody, in comparison to TGI in the inCT group, was determined on day 3 after the last treatment according to the formula TGI where V i is the initial mean tumor volume in the TEDbody group or inCT treatment group and V f is the final mean tumor volume in the TEDbody or inCT treatment group, as indicated by the superscript text [15]. If necessary, the tumors were excised on day 3, after the last treatment, for analyzing the CMV-pMHCI presentation (by IHC staining) and activation of tumor-infiltrating CMVp-CTLs, as described above. The mice were euthanized by CO 2 asphyxiation, and some tumors were excised for histological analysis, as described in a previous study.

Statistical analysis
Data are presented as a representative image for imaging experiments, mean ± SEM for pooled data, or mean ± SD for representative assays involving at least three independent experiments, unless specified otherwise. Differences between experimental groups and controls were analyzed for statistical significance by unpaired two-tailed Student's t-test. One-way ANOVA with the Newman-Keuls multiple-comparison post hoc test was performed to determine the significance of in vivo tumor growth data using GraphPad Prism software (GraphPad, Inc.). No corrections were implemented in the statistical tests. P value < 0.05 was considered to denote statistical significance.

Design and preparation of TEDbodies carrying various CMVp 495-503 -encompassing peptides
For cytosolic delivery of viral CTL epitope peptides specifically into integrin αvβ5-expressing tumor cells, we used a tumor-specific cytotransmab, inCT, previously engineered to have two functional parts: (i) the light chain N-terminus-fused cyclic peptide (in4) specific to αvβ5 for receptor-mediated endocytosis, and (ii) VH and VL with an endosomal escape ability for relocation to the cytosol from an endosome [13,14] (Fig. 1A). For the proof-of-concept experiment with a TEDbody, we chose to deliver the HLA-A*02:01-restricted CTL epitope of CMV pp65-derived CMVp 495-503 into the cytosol of target tumor cells. To identify optimal fusion peptides for the TEDbody in terms of CMV-pMHCI presentation efficiency after cytosolic localization, we screened a panel of precursor peptides encompassing 9-mer mature epitope CMVp 495-503 with N-terminally or N-and C(N/C)-terminally extended sequences by fusion with the C-terminus of the heavy chain of inCT via an uncleavable 5-mer G 4 S linker (Fig. 1B). To exclude nonspecific cellular uptake of the TEDbody by Fcγ receptorexpressing antigen-presenting cells, inCT with a silenced Fc domain carrying LALAPG mutations (L234A/L235A/ P329G) was employed to abrogate the interactions with Fcγ receptors [35] (Fig. 1A). Hereafter, such TEDbody clones (i.e., CD8 + T cell epitope peptide-fused inCT) are referred to as "inCT †[peptide name]" (Fig. 1B). For example, the TEDbody carrying CMVp 495-503 was named inCT †CMVp 495-503 . The designed TEDbodies were found to be well expressed in the correctly assembled form in a standard HEK293F transient expression system (Fig. S1).
For cells treated with a TEDbody, we observed that inCT †CMVp 480-503 -induced CMV-pMHCI formation on the surface of HLA-A*02:01 + cells at both 4 °C and 37 °C, but not on HLA-A*02:01 − cells, suggesting that CMVp 480-503 in the Ab-fused form can be extracellularly loaded onto HLA-A*02:01 (Fig. 1C, D and Fig. S3). In contrast, inCT †CMVp 495-503 , carrying the mature 9-mer CMVp 495-503 epitope with the N-terminal G 4 S linker  (Fig. 1C, D and Fig. S3). For integrin αvβ5 − HLA-A*02:01 + NCI-H889 cells (Fig. S2), the CMV-pMHCI presentation was detectable only with extracellular surface loading-capable inCT †CMVp 480-503 but not with N/C-extended peptide-carrying TEDbodies (Fig. 1C). The above findings meant that the N/C-extended peptides fused to inCT cannot be extracellularly loaded onto HLA-A*02:01, as seen with the N/C-extended peptides themselves. Nonetheless, they can be displayed in the form of CMV-pMHCI on the cell surface by TEDbodymediated cellular delivery and processing if they have a long enough C-terminal flanking sequence (i.e., at least beyond the residue number 510, as is the case with inCT †CMVp 480-510 ) to be recognized and cleaved by cytosolic proteasomes [39]. Overall, the above data indicated that a TEDbody can intracellularly deliver N/Cextended CMVp 495-503 -encompassing peptides to be displayed in the pMHCI form of CMV-pMHCI on the surface of target cells expressing both integrin αvβ5 and HLA-A*02:01.

CMVp-CTLs recognize and kill tumor cells presenting CMV-pMHCI via a TEDbody
To test whether TEDbody-mediated CMV-pMHCI presentation renders the marked tumor cells susceptible to lysis by CMVp-CTLs, TEDbodies were incubated with tumor cells for 12 h to ensure internalization and cellular processing for CMV-pMHCI presentation, followed by coculturing for another 18 h with ex vivo-expanded CMVp-CTLs, at an E:T ratio of 5:1. Compared with the control inCT, TEDbodies that had shown the ability to implement CMV-pMHCI presentation at 37 °C substantially induced the lysis of HLA-A*02:01 + MDA-MB-231 cells, but not HLA-A*02 − LoVo cells, in a concentrationdependent manner (Fig. 1E). This evidence suggested that TEDbody-mediated CMV-pMHCI presentation can activate CMVp-CTLs among PBMCs to elicit immune responses. The cell lysis potency of the TEDbodies roughly correlated with the magnitude of CMV-pMHCI surface display at 37 °C. In contrast, the TEDbodies carrying either off-target HPV E11-19 or a CMV-pMHCI display-incapable peptide only negligibly triggered cell lysis. Notably, the TEDbody carrying a N/C-extended CMVp 480-516 or CMVp 480-519 peptide that requires cellular uptake to present CMV-pMHCI manifested the strongest ability to induce tumor cell lysis; the lysis magnitude was comparable to that induced by CMVp 495-503 peptide pulsing (Fig. 1E). In integrin αvβ5 − HLA-A*02:01 + NCI-H889 cells, the lysis was caused by extracellular surface loading-competent inCT †CMVp 480-503 , but not by the N/C-extended peptide-bearing TEDbodies requiring cellular uptake for CMV-pMHCI presentation (Fig. 1E). These results demonstrated the target cell specificity of the cellular internalization-requiring TEDbody, specifically, its ability to drive CMV-pMHCI presentation on the surface of cells expressing both integrin αvβ5 and HLA-A*02:01 to be recognized and lysed by CMVp-CTLs.

TEDbody-mediated CMV-pMHCI presentation proceeds via the conventional MHC-I antigen-processing pathway
To further dissect the TEDbody-mediated CMV-pMHCI presentation, we generated peptide-fused Ab controls via fusion of the N-extended CMVp 480-503 or N/C-extended CMVp 480-516 to endosomal escape-incapable Abs [anti-EGFR therapeutic Ab called necitumumab (Portrazza ™ ) or inCT (AAA) [13,14]] in the same manner as with a TEDbody. Although the anti-EGFR Ab and endosomal escape motif-deficient inCT (AAA) get internalized into the cells through a specific receptor of EGFR or integrin αvβ5, respectively, they are not expected to deliver the fused peptide into the cytosol of target cells because of the absence of the endosomal escape ability. Both inCT †CMVp 480-503 -and inCT (AAA) †CMVp 480-503 -induced CMV-pMHCI presentation at both 4 °C and 37 °C and elicited cytolysis by CMVp-CTLs at 37 °C in an HLA-A*02:01-restricted manner ( Fig. 2A,B), thereby confirming the extracellular surface loading capability of the CMVp 480-503 peptide even in the Ab-fused form. On the contrary, CMVp 480-516 -fused necitumumab and inCT (AAA) failed to implement CMV-pMHCI presentation at 4 °C and 37 °C and did not induce cytolysis by CMVp-CTLs, in contrast to the efficient CMV-pMHCI presentation and cytolysis of inCT †CMVp 480-516 -treated HLA-A*02:01 + cells in proportion to the concentration ( Fig. 2A,B). These findings indicate that TEDbody-mediated cytosolic delivery of N/C-extended CMVp 480-516 is essential for surface presentation of CMV-pMHCI.
To determine the involvement of intracellular processing in TEDbody-induced CMV-pMHCI presentation, we examined effects of a proteasome inhibitor (MG132), an ER-resident aminopeptidase 1 inhibitor (ERAP1-IN-1) [41], and an inhibitor of vesicle-mediated transport from the ER to the Golgi apparatus (brefeldin A) [42] on the activation of CMVp-CTLs mediated by a TEDbody through CMV-pMHCI display. CMVp-CTL activation was detected by quantifying IFN-γ released into the supernatant to exclude any possible cytotoxic impact of the inhibitor on tumor cells. The three inhibitors did not significantly affect the CMVp-CTL activation induced by pulsing with either the CMVp 495-503 peptide or inCT (AAA) †CMVp 480-503 (Fig. 2C), confirming their extracellular surface loading ability without the need for further intracellular processing. On the other hand, the activation of CMVp-CTLs by inCT †CMVp 480-503 was significantly inhibited (but was still substantial) by treatment with each of the three inhibitors (Fig. 2C). These findings indicated that inCT †CMVp 480-503 -mediated CMV-pMHCI formation proceeds via two pathways: i) extracellular surface loading onto HLA-A*02:01 at the cell surface, as observed at 4 °C ( Fig. 2A), and ii) conventional class I antigen processing after cytosolic access. In contrast, inCT †CMVp 480-516 -mediated activation of CMVp-CTLs was almost completely abrogated by the presence of each of the three inhibitors (Fig. 2C). One study elucidated that cytosolically generated CMVp 495-503 -encompassing peptides are efficiently transported into the ER by TAP [39]. A knockout of TAP1 in MDA-MB-231 cells (Fig. S6A, B) substantially reduced the magnitude of the CMV-pMHCI surface presentation caused by inCT †CMVp 480-516 but only slightly by inCT †CMVp 480-503 or inCT (AAA) †CMVp 480-503 (Fig. 2D), indicating that TAP1 is involved in the transport of cytosolically processed CMVp 480-516 -derived epitope precursor peptides from the cytosol into the ER.
Next, we visualized the intracellular trafficking of CMV-pMHCI by its costaining with EEA1 (specific for early endosomes) or with 58 K Golgi (specific for the Golgi apparatus) [33]. In the controls of endosomal escape motif-deficient necitumumab and inCT (AAA) carrying CMVp 480-516 , CMV-pMHCI was not detectable in any cellular compartment (Fig. 2E), which was consistent with the inability to cause CMV-pMHCI display. However, in cells treated with inCT †CMVp 480-516 , CMV-pMHCI was present inside the cells and colocalized with 58 K Golgi but not with EEA1 (Fig. 2E), thus pointing to the trafficking of CMV-pMHCI through the ER-Golgi pathway for surface presentation. Collectively, the above data suggested that inCT †CMVp 480-516 -induced CMV-pMHCI presentation proceeds entirely through the class I antigen-processing pathway [12], as follows: i) cytosolic delivery of CMVp 480-516 and its cleavage by a cytosolic proteasome to generate precursor peptides with the correct C-terminus; ii) transport of the precursors into the ER by TAP and trimming of the residues of the N-extended sequence (if any) by ERAPs in the ER to generate the mature epitope (CMVp 495-503 ) for loading onto HLA-A*02:01 to form CMV-pMHCI; and iii) exocytosis of CMV-pMHCI through the ER-Golgi transport pathway for surface presentation, as in the natural processing of the CMV pp65 antigen [39].

TEDbody-induced tumor cell lysis by CMVp-CTLs is proportional to the number of effector cells and incubation time
By means of inCT †CMVp 480-516 requiring cytosolic processing for CMV-pMHCI presentation, we further characterized the lysis of target tumor cells by CMVp-CTLs. The magnitude of tumor cell lysis driven by inCT †CMVp 480-516 increased in proportion to the E:T ratio only for HLA-A*02:01 + tumor cells but not for HLA-A*02:01 − tumor cells (Fig. 2F). For a real-time cell lysis assay, we treated MDA-MB-231-EGFP cells stably expressing EGFP (Fig. S6C) with inCT †CMVp 480-516 for 12 h and then cocultured them with ex vivo-expanded CMVp-CTLs labeled with red fluorescent dye PKH26, while monitoring a decrease in fluorescence intensity of EGFP as an indicator of MDA-MB-231-EGFP cell lysis. In kinetic experiments, the treatment of inCT †CMVp 480-516 led to rapid lysis of MDA-MB-231-EGFP cells (Fig. 2G, H), i.e., caused more than 70% lysis after 14 h, similar to that of CMVp 495-503 peptide. In contrast, cytosolic access-incapable inCT (AAA) †CMVp 480-516 and necitumumab †CMVp 480-516 as well as the extracellular surface loading-incapable N/C-extended CMVp 480-516 peptide triggered negligible cell-killing activity (Fig. 2G,H). These results clearly indicated that TEDbody-mediated cytosolic delivery of CMVp 480-516 resulted in efficient presentation of CMV-pMHCI on the surface of target tumor cells for their lysis by CMVp-CTLs.

TEDbody-mediated CMV-pMHCI presentation activates CMVp-CTLs in vivo
To determine whether a TEDbody can label tumor cells with CMV-pMHCI in vivo, a test TEDbody at 20 mg/kg (mpk) was i.p. injected once into immunodeficient NSG mice bearing preestablished orthotopic MDA-MB-231 breast cancer cell-derived xenografts having an average tumor volume of 100-120 mm 3 . At 24 h postinjection, the tumors were excised and subjected to IHC analysis of the surface expression of CMV-pMHCI. Notably, treatment with either inCT †CMVp 480-503 or inCT †CMVp 480-516 yielded CMV-pMHCI presentation of similar magnitudes near the plasma membrane and inside the tumor cells ( Fig. 3A and Fig. S7A), thereby pointing to in vivo capacity for inducing CMV-pMHCI presentation on tumor cells after systemic administration. In contrast, neither inCT (AAA) †CMVp 480-503 nor necitumumab †CMVp 480-516 had this effect. Although inCT (AAA) †CMVp 480-503 caused CMV-pMHCI display in vitro via extracellular surface loading, it failed in vivo, in line with some reports that pMHCI formation through extracellular surface loading rarely occurs in vivo [11,36]. Consequently, the in vivo presentation of CMV-pMHCI with the help of inCT †CMVp 480-503 seemed to be mediated mainly by the intracellular antigen-processing pathway, as was the case for inCT †CMVp  .

In vivo antitumor efficacy of the TEDbody in immunodeficient mice bearing human tumors
To assess the in vivo antitumor activity of the TEDbody in conjunction with adoptive transfer of ex vivo-expanded CMVp-CTLs, we treated NSG mice bearing preestablished orthotopic MDA-MB-231 xenografts or subcutaneous HCT116 colorectal cancer cell-derived xenografts having an average tumor volume of 100-120 mm 3 with i.p. injection of the TEDbody or one of control substances (20 mpk) plus IL-15/IL-15Rα-Fc (15 μg); 6 h later, we intravenously injected ex vivo-expanded CMVp-CTLs (10 7 cells; Fig. 3C). The TEDbody was injected every 3 d for a total of six doses, and CMVp-CTLs were administered every 6 d for a total of three doses (Fig. 3C). The IL-15/IL-15Rα-Fc protein, known as P22339 [30], was concomitantly injected to increase the survival of the transferred CMVp-CTLs. Compared with the inCT-treated control, cytosolic access-incapable inCT (AAA) †CMVp 480-503 and necitumumab †CMVp 480-516 failed to inhibit tumor growth (Fig. 3D), which was consistent with their in vivo inability to cause CMV-pMHCI presentation (Fig. 3A). By contrast, inCT †CMVp 480-516 and inCT †CMVp 480-503 markedly slowed the tumor growth, manifesting an in vivo antitumor activity via the transferred CMVp-CTLs ( Fig. 3D and Fig. S8). Compared to the inCT-treated control, they showed similar antitumor potency levels, with TGI (at the end of treatment) of 46% (inCT †CMVp 480-503 ) and 58% (inCT †CMVp 480-516 ) for MDA-MB-231 xenografts and 61% (inCT †CMVp 480-503 ) and 68% (inCT †CMVp 480-516 ) for HCT116 xenografts. When the tumors were excised on day 3 after the last treatment and analyzed via The arrows indicate each time point for the treatment or assay. D Tumor growth, measured as the average tumor volume, in response to the indicated treatment, as described in (C). Error bars: ±SEM (n = 9 to 14 per group for HCT116 tumors, n = 8 to 13 per group for MDA-MB-231 tumors). Data were pooled from two independent experiments with at least four mice per group. E and F IHC detection of CMV-pMHCI (red) on tumor tissues (E) and the number of tumor-infiltrating CMVp-CTLs per gram of a tumor (F) excised from mice on day 3 after the last treatment, as described in (C). In (E), nuclei were stained with Hoechst 33342 (blue), and images are representative of three independent experiments; additional images are shown in Fig. S7B. Scale bar: 20 μm. The right panel shows the quantification of red fluorescence intensity, obtained by ImageJ software. Error bars, ±SD of 2 fields per tumor (n = 3 per group). In (F), bar graphs present the mean ± SEM (n ≥ 3 different tumors). In (B), (D), and (F), **P < 0.01 and ***P < 0.001 denote a significant difference between the indicated groups (B and F) or a significant difference from the inCT group (B, D, and F), as determined by one-way ANOVA with the Newman-Keuls post hoc test; ns: not significant IHC staining, CMV-pMHCI presentation was detectable in the tumors treated with inCT †CMVp 480-516 or inCT †CMVp 480-503 but not in those treated with the controls ( Fig. 3E and Fig. S7B). Moreover, compared to the control groups, treatment with CMV-pMHCI-presenting TEDbody increased the number of tumor-infiltrating CMVp-CTLs by ~ 15-fold (Fig. 3F), indicating that TEDbody-mediated CMV-pMHCI presentation induced efficient infiltration of the transferred CMVp-CTLs into tumor tissue and/or stimulated their proliferation in the TME. Collectively, the above results explained the in vivo antitumor mechanisms of action of TEDbody, namely, the marking of target tumor cells with CMV-pMHCI for recognition and killing by transferred CMVp-CTLs.

Combination of the TEDbody with an anti-OX40 agonistic ab enhances the antitumor activity
In studies on adoptive transfer of CTLs for cancer immunotherapy, a T cell costimulatory agonist or inhibitory antagonist has been added to enhance the antitumor effects [9,11]. Because a T cell costimulatory molecule called OX40 was found to be upregulated in the ex vivo-expanded effector memory CMVp-CTLs (Fig.  S5B, C), we conducted an in vivo assay by combining inCT †CMVp 480-516 with an anti-OX40 1166/1167 agonistic Ab (anti-OX40; Fig. S9) [28] in NSG mice with preestablished MDA-MB-231 xenografted tumors (Fig. 4A). Compared with the inCT-treated control, the combined treatment with inCT †CMVp 480-516 and anti-OX40 markedly enhanced the in vivo antitumor activity, with a TGI of 83% which was notably higher than that of the monotherapy with either inCT †CMVp 480-516 (56%) or anti-OX40 (0%; Fig. 4B,C and Fig. S10). Compared with each monotherapy, the combined treatment further increased the number of tumor-infiltrating CMVp-CTLs and potentiated the cytotoxic effector function, as evidenced by expression analysis of granzyme B (Fig. 4D). A portion of CMVp-CTLs expressed inhibitory receptors, such as PD1, LAG-3, and/or TIGIT (Fig. S5B, C). To explore any blocking effects of PD1, the representative inhibitory receptor for CTL exhaustion, on the antitumor activity of CMVp-CTLs, we combined inCT †CMVp 480-516 with an anti-PD1 antagonistic Ab, pembrolizumab (Fig. S9). Compared to treatment with inCT †CMVp 480-516 alone, the combined treatment did not provide any additional antitumor effects (Fig. 4E,F) nor did it increase the number and cytotoxicity of tumor-infiltrating CMVp-CTLs (Fig. 4G), even though PD1 and its ligand PD-L1 were expressed on the surface of CMV-CTLs (Fig. S5C) and MDA-MB-231 cells (Fig. S2), respectively. This suggested that a combination of a TEDbody with an anti-OX40 agonist, but not an anti-PD1 antagonist, is a viable approach to enhancing the proliferation and cytotoxic effector function of transferred CMVp-CTLs in the TME.

Discussion
Cytosolic delivery of an MHC-I antigen into target tumor cells that lack tumor antigens amenable to immunotherapy remains a serious limitation but holds great potential as a gateway to the development of a therapeutic cancer vaccine, if achievable. Our study revealed an ability of a TEDbody to deliver an MHC-I-restricted viral CTL epitope peptide into the cytosol of integrin αvβ5-expressing target tumor cells for pMHCI presentation on the cell surface, which is a prerequisite step for priming the target cells for recognition and lysis by pre-existing antiviral CTLs arising from common viral infections in cancer patients (Fig. 5). To simulate tumor cells infected by human CMV, we generated TEDbodies carrying various CMVp 495-503 -encompassing peptides for surface expression of CMV-pMHCI via the conventional MHC-I antigen-processing pathway. The TEDbody-mediated CMV-pMHCI presentation redirected CMVp-CTLs of CMV-seropositive donors to recognize and kill target tumor cells in vitro and suppressed tumor growth in immunodeficient mouse models. Therefore, the TEDbody is a useful technology for cytosolic delivery of MHC-I-restricted viral peptides mimicking the natural presentation pathway of an MHC-I viral antigen, and therefore, may lead to possible therapeutic cancer vaccines directly targeting tumor cells rather than antigenpresenting cells.
Several approaches have been explored to deliver viral CTL epitope peptides to tumor cells with the aim of simulating cancer cells that are virally infected, with consequent redirection of the corresponding antiviral CTLs to the tumor. The previous approaches can be classified into two: (i) extracellular surface loading without cellular uptake [6,[8][9][10], and (ii) endosomal loading after cellular internalization [11]. As for the first approach of extracellular surface loading, intratumoral injection of various synthetic viral epitope peptides, including CMVp 495-503 peptide itself, has been investigated but has resulted in very weak in vivo antitumor activity as monotherapy [6]. Another extracellular surface loading strategy is to take advantage of a tumor-targeting Ab-peptide epitope conjugate [9,10] in which a mature peptide epitope is conjugated with an Ab via a linker cleavable by a tumor-associated protease such that the peptide epitope is proteolytically released onto the tumor cell surface and then extracellularly loaded onto MHC-I to form pMHCI without cellular uptake. The Ab-peptide epitope conjugates that were designed to release CMVp 495-503 coated tumor cells with CMV-pMHCI for recognition by CMVp-CTLs, resulting in delayed tumor growth in mouse models, particularly when combined with immune checkpoint-blocking Abs [9]. However, in our study, we noted extracellular surface loading of N-extended CMVp 480-503 (even in the form of fusion to an IgG Ab via an uncleavable G 4 S linker) onto HLA-A*02:01 with the formation of functional CMV-pMHCI reactive with CMVp-CTLs, suggesting that proteolytic cleavage of Ab-fused CTL epitopes at the cell surface is not a prerequisite for MHC-I loading. As for the second approach involving endosomal loading after cellular internalization, Ab-targeted pathogen-derived peptides have been generated by conjugation of an Epstein-Barr virus-derived CTL epitope peptide to a In (B) to (G), *P < 0.05, **P < 0.01, and ***P < 0.001 indicate a significant difference between the indicated groups or a significant difference from the inCT group (B and E), determined by one-way ANOVA with the Newman-Keuls post hoc test; ns: not significant tumor-targeting Ab via a cleavable (reducible) disulfide bond in the endosomal environment after cellular endocytosis [11]. However, the necessity of the presence of a Cys residue in the CTL epitope sequence may limit the applicability of Ab-targeted pathogen-derived peptide technology. In contrast to TEDbody, the previous two approaches do not require cytosolic access and processing for cell surface presentation of a CTL epitope peptide on MHC-I. As demonstrated in studies with a pharmacological inhibitor and TAP1 knockout MDA-MB-231 cells treated with inCT †CMVp 480-516 , a TEDbody delivers MHC-I-restricted peptides into the cytosol of target tumor cells for pMHCI presentation through the endogenous MHC-I antigen presentation pathway [12,39]. Although the surface loading of inCT (AAA) †CMVp 480-503 onto HLA-A*02:01 to form CMV-pMHCI gave rise to significant cytolytic activity of CMVp-CTLs in vitro in our study, the in vivo CMV-pMHCI presentation and antitumor activity were negligible, suggesting that extracellular surface loading does not take place efficiently in vivo, as demonstrated in previous studies [6,11,36].
For the formation of the pMHCI complex, MHC-I molecules bind short peptides that are typically 8-11 amino acid residues in length (with a 9-mer preferred) with both ends tucked inside the binding groove of MHC-I [36]. In the current study, following the size restriction of MHC-I, the N/C-extended peptides and their fused form in a TEDbody failed to induce CMV-pMHCI via extracellular pulsing. In contrast, extracellular pulsing of N-extended peptide CMVp 480-503 with the correct C-terminus in either the free or Ab-fused form in a TEDbody caused extracellular loading onto HLA-A*02:01 at 4 °C, thereby generating CMV-pMHCI. Moreover, the N-extended peptide-loaded CMV-pMHCI can also be recognized by CMVp-CTLs, as evidenced by the cytolytic activity at 37 °C. There are some documented cases of unconventional presentation of either N-or C-terminally extended peptides (but never peptides that are N/C-extended at the same time) on a cognate MHC-I molecule, including

Anti-viral CTL
Pre-existing anti-viral CTLs Fig. 5 A schematic diagram of the proposed mode of action of a TEDbody in cancer immunotherapy. A TEDbody delivers a fused MHC-I-restricted viral CTL epitope peptide into the cytosol of integrin αvβ5-expressing cancer cells to be processed for surface presentation by cognate MHC-I, thereby rendering the marked cancer cells recognizable and killable by pre-existing antiviral CTLs arising from common human viral infections in cancer patients as follows: 1) binding to the tumor-associated receptor, integrin αvβ5, for cellular internalization; 2) cytosolic localization through endosomal escape; 3) proteasomal cleavage and degradation generating CTL epitope precursor peptides; 4) ER uptake by TAP and N-terminal trimming of the precursor peptides to generate the mature CTL epitope, followed by its binding to cognate MHC-I; 5) cell surface presentation of the pMHCI through the ER-to-Golgi pathway; 6) recognition and lysis of pMHCI-presenting cells by pre-existing pMHCI-specific antiviral CTLs HLA-A*02:01, via structural changes either within the MHC-I that opens the confined binding groove ("protrusion") or within the peptide itself ("bulging") [43]. Nonetheless, unlike inCT †CMVp 480-503 , inCT †CMVp [495][496][497][498][499][500][501][502][503] with the N-terminal artificial G 4 S linker sequence failed to be loaded onto HLA-A*02:01 in our study, indicating that N-terminal endogenous sequences flanking the matured epitope strongly affect the loading and/or proteolytic generation of an MHC-I-presented mature epitope peptide [44]. Collectively, our findings illustrate the flexibility of the HLA-A*02:01 molecule for the accommodation of a CMVp 495-503 peptide with an N-extended endogenous sequence, even as fusion to an IgG Ab; this is not the case for N/C-extended peptides. Considering the size of a TEDbody, protrusion, rather than bulging, is a possible extension mechanism [45], although details of the mechanism remain to be determined via further structural studies. For practical in vivo applications, a TEDbody was generated based on the full-length human IgG1/κ form but with an effector function-silenced Fc and a tumor tissuehoming ability by targeting of tumor-associated integrin αvβ3/αvβ5. Integrin αvβ3/αvβ5 are overexpressed on the surface of many types of epithelial tumor cells and tumorassociated blood vessels compared to normal cells and tissues [46]. However, any on-target/off-tumor adverse effects of TEDbody should be evaluated during clinical trials. Cytosolic processing-requiring TEDbody-mediated CMV-pMHCI presentation proved to be specific to tumor cells expressing both the tumor-associated membrane receptor (integrin αvβ5, for cytosolic access) and the HLA-A*02:01 allele (for epitope presentation), consequently confirming target cell specificity of the TEDbody. Moreover, systemic injection of inCT †CMVp 480-516 into human tumor-bearing immunodeficient mice in our study marked a substantial proportion of the target tumor cells with CMV-pMHCI, thereby demonstrating the feasibility of converting tumor cells in vivo into CMV-infected cells, which renders the tumor susceptible to lysis by transferred CMVp-CTLs for attaining substantial antitumor effects. Furthermore, the in vivo TEDbody-mediated CMV-pMHCI presentation on target tumors seemed to serve as persistent antigenic stimulation of tumor-infiltrating CMVp-CTLs by augmenting their tumor infiltration and cytotoxicity in the TME.
Here, a combination therapy consisting of a TEDbody with an anti-OX40 agonistic Ab, but not with an anti-PD1 antagonistic Ab, yielded more potent TGI (than TEDbody monotherapy did) by enhancing the expansion and cytotoxic effector function of adoptively transferred CMVp-CTLs in the TME of immunodeficient mouse models. This evidence is in agreement with some reports on immune responses of murine CTLs against mouse CMV infection, wherein the proliferation and effector function of memory inflation-associated CTLs generated by low-level persistent CMV infection were promoted by OX40 costimulation [47,48] but were retained independently of PD1 expression [49]. Our results suggest that antigenic stimulation by TEDbody-mediated CMV-pMHCI presentation maintains the effector memory phenotype of CMVp-CTLs, whereas the additional OX40 costimulation by the anti-OX40 agonistic Ab reinforces the expansion and cytotoxic effector function in the TME, thereby potentiating the antitumor action in comparison with TEDbody monotherapy. Our results also suggest that the PD1-PD-L1 immune checkpoint axis is not strongly associated with dysfunction of effector memory CMVp-CTLs [49]. On the other hand, simultaneous blockade of PD1 with other inhibitory receptors might be required to elicit strong antitumor responses of CMVp-CTLs. Overall, our data suggest that a combination with an anti-OX40 agonistic Ab, rather than PD1 blockade, is a potentially suitable strategy for improving clinical responses when CMVp-CTLs are repurposed against tumors.
Although CMV is not considered as an oncogenic virus, CMV antigens and nucleic acids have been detected prevalently in patients with a variety of cancer types, particularly glioblastoma [50]. This provides a rationale for adoptive transfer of ex vivo-expanded autologous CMV-specific CTLs to patients with CMVinfected tumors as cancer immunotherapy, which prolonged progression-free survival of some patients with recurrent glioblastoma [24,50]. Mimicking a viral infection, specifically in tumor cells, with the aim of harnessing pre-existing virus-specific CTLs to attack the tumor, may offer an alternative to cancer immunotherapies involving neoantigen-based cancer vaccines [7]. Most neoantigens are tumor type-specific and/or patient-specific and require a personalized vaccine approach [1,2], but antigens from viruses that commonly infect humans are likely to be shared among individuals, thus enabling the development of universal off-the-shelf cancer vaccines. Clinical relevance of the proposed TEDbody lies in the delivery of the immunodominant CTL epitope (CMVp 495-503 ) for presentation on the most prevalent MHC-I molecule-HLA-A*02:01 (30-50% prevalence in the human population, depending on ethnicity)-according to the following facts: i) the widespread infection of adults by CMV (60-90% of the population), and ii) the unique properties of pre-existing CMVp-CTLs such as the high functional competence (mainly T EM and T EMRA ) and abundance (up to ~ 11% of CD8 + T cells) [9,23]. Accordingly, the CMVp 495-503 -armed TEDbody is applicable to approximately 18-45% of cancer patients (in our study, 37.3% CMVp-CTL-positive PBMCs were detected in 28 out of 75 donors). In addition, given that viral infections are estimated to account for up to 20% of all cancer cases worldwide [51], the TEDbody technology may gain high popularity if it is extended to various cancer types. Nonetheless, there are major challenges for practical uses of TEDbodies against tumors, such as the following: presence of thousands of MHC-I allelic variants within the human population, and MHC-I loss or downregulation in tumor cells [52]. These challenges might be addressed by i) the development of a TEDbody carrying multiple epitope peptides to broaden the range of cancer patients that can be treated with this modality, and ii) combination therapy with either a cytokine (e.g., IFNγ) or an appropriate chemical agent to restore MHC-I expression on tumor cells [52,53].

Conclusion
In conclusion, our study offers an effective technology for MHC-I antigen cytosolic delivery called TEDbody, which may help to develop a therapeutic cancer vaccine that delivers a viral CTL epitope directly into a target tumor, thereby making the tumor recognizable and killable by pre-existing antiviral CTLs in patients. Our approach requires information on immunogenic MHC-I-restricted viral epitopes and antiviral CTL immunophenotypes. The TEDbody technology can be utilized for cytosolic delivery of other (nonviral) MHC-I antigens, such as tumor neoantigens, into tumors lacking tumor antigens suitable for immunotherapy and has good potential for expanding the current arsenal of cancer immunotherapies.
Additional file 1: Figure S1. SDS-PAGE analysis of the TEDbodies carrying various CMVp 495-503 -encompassing peptides. Figure S2. Cell surface expression levels of HLA-A*02, integrin αvβ3, integrin αvβ5, and PD-L1 in human cancer cell lines, as analyzed by flow cytometry. Figure S3. Flow cytometric determination of cell surface CMV-pMHCI presentation induced by a synthetic peptide (A) or TEDbody and control Ab (B). Figure  S4. Prevalence of CMVp-CTLs among PBMCs before and after ex vivo expansion with the CMVp 495-503 peptide. Figure S5. Immunophenotyping of CMVp-CTLs among PBMCs before and after ex vivo expansion with the CMVp 495-503 peptide. Figure S6. Construction of TAP1 knockout MDA-MB-231 cells and MDA-MB-231-EGFP cells. Figure S7. Additional images of IHC detection of CMV-pMHCI (red) on MDA-MB-231 tumor tissues, the representative of which is shown in Fig. 3A (A) and Fig. 3E (B). Figure S8. The TEDbody suppresses in vivo growth of human tumor xenografts in immunodeficient NSG mice, as described in Fig. 3D. Figure  S9. Binding specificity of the anti-OX40 agonistic 1166/1167 Ab and anti-PD1 antagonistic Ab (pembrolizumab), constructed and used in this study, respectively, to the surface-expressed antigen. Figure S10. In vivo antitumor efficacy of the inCT †CMVp 480-516 TEDbody, combined with either the anti-OX40 1166/1167 Ab or anti-PD1 pembrolizumab, in NSG mice harboring preestablished MDA-MB-231 orthotopic tumor xenografts, as described in Fig. 4A. Table S1. List of synthesized peptides, used in this study. Table S2. List of resources (antibodies, recombinant proteins, and chemicals), used in this study. Table S3. Prevalence rates of CMVp-CTLs among PBMCs from HLA-A*02-positive healthy donors. Table S4. Characterization of CMVp-CTLs before and after ex vivo expansion with the CMVp 495-503 peptide.