C57BL/6 mice were obtained from Charles River Laboratories (L’Arbresle, France) or Jackson Laboratory (Sacramento, California, USA). Cd70^−/− 27 and Cd80/86^−/− 28 and Ly5.1 (CD45.1) mice were bred in house. At the start of the experiments, mice were 6 to 8 week-old. Animals were housed in individually-ventilated cages under specific-pathogen free conditions at the animal facility at the Leiden University Medical Center (LUMC). All animal experiments were approved by the Animal Experiments Committee of LUMC and performed according to the recommendations and guidelines set by the LUMC and Dutch Experiments on Animals Act.

E1/E3-deleted replication-deficient adenoviral vectors (type 5) expressing full-length E7 (Ad-E7) or the E7 epitope (Ad-R9F) transgenes under control of the human CMV promoter were produced at the Jenner Institute Viral Vector Core Facility (University of Oxford), as previously described.19 Expression cassettes were flanked by an initiator methionine, stop codon and a SV40 polyA tail. Peptide sequence full-length E7: MHGDTPTLHE YMLDLQPETT DLYCYEQLND SSEEEDEIDG PAGQAEPDRA HYNIVTFCCK CDSTLRLCVQ STHVDIRTLE DLLMGTLGIV CPICSQKP. Peptide sequence E7 epitope (R9F): RAHYNIVTF.

HEK293A T-Rex cells were transfected using linearized plasmid and the presence of the antigen gene was confirmed by PCR. The integrity of the antigenic DNA sequence and absence of contaminating adenovirus was confirmed by Flank-Flank PCR. The virus was titred to obtain infectious units (IU) per mL, and assayed by spectrophotometry to quantify the number of virus particles per mL. The P:I ratios of virus particles to infectivity were 25 (Ad-E7) and 40 (Ad-R9F). The sterility of the virus was also confirmed by inoculation of TSB broth with 10 µL of purified virus and incubation for 3 days at 35°C. Adenoviral vectors were stored at −80°C and injected intravenously via the retro-orbital route at a dose of 10⁷, 10⁸ or 10⁹ IU per mouse.

Mice were perfused with 20 mL PBS containing 2 mM EDTA to remove circulating blood cells from the organs of interest. The liver, lungs and tumors were cut into small pieces using surgical knives. The liver was resuspended in 3.5 mL IMDM containing 250 U/mL collagenase type 1-A (C2674, Sigma) and 20 µg/mL DNase I (D5025, Sigma). Lung tissue was incubated with 1 mL IMDM and 250 U/mL collagenase and 20 µg/mL DNase. For both liver and lungs, collagenase and DNase were incubated for 25 min at 37°C. Tumors were incubated for 15 min at 37°C with IMDM-containing Liberase (#05401020001, Roche). Single cell suspensions were made using 70 µm cell strainers and subsequently lymphocytes were isolated using a Percoll gradient.

For intraepithelial lymphocyte (IEL) isolation, fat tissue and Peyer’s patches were removed from the small intestine after which it was cut into 1 cm pieces. After washing, the intestine was incubated with 20 mL IMDM containing 10% FCS, 5 mM EDTA and 1 mM DTT (D0632, Sigma) for 30 min at 37°C. Supernatant containing the IEL fraction was isolated by filtering over a 70 µm cell strainer and lymphocyte fractions were isolated using a Percoll gradient. For skin T cell isolation, first remaining hair was removed using Veet hair removal cream after which 1 cm by 1.5 cm skin tissue (located at the subcutaneous tumor site) was dissected. Skin tissue was incubated in 1.5 mL of 2.5 mg/mL Dispase II (D4693, Sigma) at 37°C for 1.5 hours. The skin tissue was cut into small pieces using surgical knives after which it was incubated in 1.5 mL 3 mg/mL collagenase type IV (C5138 Sigma) and 5 µg/mL DNAse I (D5025, Sigma) for 30 min at 37°C. Single cell suspensions were made using 70 µm cell strainers.

The TC-1 tumor cell line (a kind gift from T.C. Wu, John Hopkins University, Baltimore, MD) was developed by retroviral transduction of C57BL/6 lung epithelial cells with the HPV16 E6/E7 and c-H-ras oncogenes.29 The tumor cell line C3 was generated by transfection of mouse embryonic cells with the HPV16 genome and an activated-ras oncogene.30 Mice were inoculated subcutaneously in the flank with 1×10⁵ TC-1 tumor cells or 5×10⁵ C3 tumor cells. Tumor size (mm³) was measured two times a week using a caliper and calculated as (L×W× H)×0.52 (L: for length, W: width, H: height). Mice were euthanized when tumor size reached 500–1000 mm³ in volume. Mycoplasma tests were frequently performed by PCR and were negative for all cell lines. Cell lines were authenticated with a microsatellite PCR.

Luciferase-expressing TC-1 tumor cells were generated by transducing the TC-1 cells with a lentiviral vector expressing IRIS-GFP and the luciferase gene luc2. Before performing surgery, mice received 0.1 mg/kg buprenorphine (Temgesic) subcutaneously as analgesia and isoflurane for anesthesia. After opening the peritoneum, the tip of the spleen was lifted to inject 1×10⁵ TC-1-luc2 tumor cells in the spleen. To visualize tumor outgrowth by bioluminescence imaging, mice were injected intraperitoneally (IP) with 100 mg/kg D-luciferin (Synchem, Germany) and imaged after 10 min with the IVIS Spectrum Imager. Bioluminescence signals were measured twice a week starting from day 2 after tumor challenge.

The cytotoxicity of CD8⁺ T cells was assessed by transferring target cells (splenocytes from Ly5.1 (CD45.1) mice) that were prior differentially labeled with CFSE and peptide. Target cells were either CFSE high labeled (5 µM) and then loaded with specific peptide (RAHYNIVTF) or were CFSE low labeled (0.5 µM), and then loaded with a-specific peptide. Subsequently, the two target cell populations were mixed in a 1:1 ratio and injected intravenously into recipient mice. Recipient mice (wild-type C57BL/6 mice, Cd80/86 ^-/- mice or mice containing CD70 blocking antibodies (clone FR70)) were either naïve or were previously vaccinated with 10⁹ IU Ad-R9F for 50 days at the time of transfer. At 16 hours after transfer, the target cell killing in the spleens of recipient mice was determined by flow cytometry.

Spleens from Ly5.1 (CD45.1) mice were isolated and subsequently CD8⁺ effector-memory T (T_EM, CD44⁺CD62L^-CD69^-)) and central-memory T (T_CM, CD44⁺CD62L⁺CD69^-)) cells were purified on a BD FACSAria cell sorter (BD Bioscience). T_EM and T_CM cells were adoptively transferred via retro-orbital injection into naïve Ly5.2 mice, TC-1 tumor-bearing mice or into mice previously vaccinated with 10⁸ IU Ad-R9F.

CD8⁺ T cell depleting monoclonal antibodies (clone 2.43) were administered twice per week, starting 4 days before tumor challenge, at 150 µg per mouse by IP injection. For low-dose anti-CD8 treatment, CD8⁺ T cell depleting monoclonal antibodies (clone 2.43, Bio X Cell, West Lebanon, New Hampshire, USA) were administered twice per week, starting 4 days before tumor challenge, at 20 µg per mouse by IP injection. To assess CD8⁺ T cell infiltration, the CD8⁺ T cells in the circulation were labeled by injecting mice retro-orbitally with 3 µg CD8a APC (clone 53.6–7, BioLegend). After 3 min mice were sacrificed and subsequently tumor, spleen and liver were analyzed.

CTLA-4 (clone 9H10) monoclonal antibodies (Bio X Cell) were administered IP twice per week for 2 weeks starting at the day of infection at 200 µg per mouse. CXCR3 (clone CXCR3-173, Bio X Cell) monoclonal antibodies were administered IP twice per week starting 4 days before tumor inoculation at 200 µg per mouse. CD80 (clone RM80) and CD86 (clone P03.1) blocking antibodies31 were administered IP two times per week at 300 µg per mouse for 2 weeks starting the day of infection. CD70 (clone FR70) blocking antibody32 was administered IP twice a week at 150 µg per mouse starting at the beginning of the experiment and continued for 3 weeks.

FTY720 (SML0700, SIGMA) was dissolved in 0,9% NaCl and administered IP twice per week at 50 µg starting 1 week before tumor challenge.

Fluorescently labeled monoclonal antibodies against the following mouse antigens were used: CD3 (clone 145–2 C11, BD Biosciences), CD4 (clone RM4-5, BioLegend), CD8a (clone 53–6.7, BioLegend), CD8b (clone YTS156.7.7, BioLegend), CD11a (clone M17/4, eBioscience), CD127 (clone A7R34, Thermo Fisher), KLRG-1 (clone 2F1, Thermo Fisher), CD44 (clone IM7, BioLegend), CD49a (clone Ha31/8, BD Biosciences), CD62L (clone MEL-14, BioLegend), CD69 (clone H1.2F3, BD Biosciences), CD38 (clone 90, Thermo Fisher) and CD103 (clone 2E7, Thermo Fisher). Cells were stained according to our previously published protocol.33 7-AAD (A1310, Invitrogen) staining was used to exclude dead cells. E7-specific CD8⁺ T cells were quantified using MHC class I tetramers for the RAHYNIVTF epitope. Flow cytometric acquisition was performed on a BD Fortessa flow cytometer (BD Biosciences).

Statistical analyzes were performed using GraphPad Prism (La Jolla, CA, Unites States). The One-way analysis of variance or Student’s t-test was used for statistical analysis of unpaired data following a Gaussian (normal) distribution test. Difference in survival was tested with the log-rank (Mantel-Cox) survival test. All P values were two sided, and P<0.05 was considered statistically significant.

Mice were immunized with different dosages of adenoviral vectors encoding full-length HPV E7 protein (Ad-E7) or the immunodominant epitope RAHYNIVTF from E7 (Ad-R9F). Vaccination with 10⁸ and 10⁹ IU of Ad-E7 or Ad-R9F elicited high frequencies of circulating E7-specific CD8⁺ T cell populations in the blood, that is, 4%–12% of the total CD8⁺ T cell population, which were maintained for months (figure 1A). In contrast, immunization with 10⁷ IU elicited E7-specific responses <2%. The higher vaccination dose correlated with an enhancement of circulating E7-specific CD8⁺ T cells with an effector-memory phenotype (ie, CD44⁺CD62L^-CD127^-KLRG1⁺) (figure 1B and online supplemental figure 1A).

Next, we interrogated the impact of adenoviral vaccination on the development of E7-specific CD8⁺ T cells in multiple organs. E7-specific CD8⁺ T cells with a T_RM-like phenotype (ie, CD69⁺ or CD69⁺CD103⁺) were present in liver, lungs, intestine and skin (figure 1C and online supplemental figure 1B). Especially, in the liver large quantities of E7-specific CD69⁺ CD8⁺ T cells were induced, which were stably maintained for more than 200 days after vaccination (figure 1D). The E7-specific CD69⁺ CD8⁺ T cells in the liver also express other markers associated with CD8⁺ T cell tissue-residency such as CD38, CD11a and CD49a (figure 1E).34–36 The adenovirus-vectored vaccines also elicited circulating effector-memory cells (CD44⁺CD62L^-KLRG1⁺CD69^-) in these organs. Remarkably, T_RM cells increased particularly in the liver compared with the non-resident CD8⁺ T cells on high dose (10⁹ IU) vaccination. The increase in the total E7-specific CD8⁺ T cell population was mainly due to the rise in E7-specific CD8⁺ T_RM cell numbers (figure 1F). In the lungs, mainly non-resident CD69^- E7-specific CD8⁺ T cells are increased in numbers after high dose vaccination (figure 1G). Thus, high-dosed adenoviral vaccination elicits memory T cell inflation and specifically enhanced T_RM cell formation in liver.

To assess whether the inflationary CD8⁺ T cells induced by adenoviral vectors have the capacity to migrate to the liver and differentiate into T_RM cells, we performed adoptive transfers of the circulating CD8⁺ effector-memory (T_EM; CD44⁺CD62L^-CD69^-) or central-memory (T_CM; CD44⁺CD62L⁺CD69^-) T cells (online supplemental figure 2A) from Ad-R9F vaccinated mice into naïve recipient mice or into recipient mice that where prior challenged with TC-1 tumor cells or prior vaccinated with Ad-R9F. When transferred into naïve mice, the T_EM and T_CM subsets largely retain their phenotype in the liver after 14 days of transfer (figure 2A). However, in mice-bearing E7-expressing TC-1 tumors a substantial fraction of the transferred T_EM cells differentiated within 10 days into liver CD69⁺ T_RM cells (figure 2B and online supplemental figure 2B). This conversion was also evident in mice previously vaccinated with the same adenoviral vector (figure 2C and online supplemental figure 2C). A portion of the E7-specific CD8⁺ T_CM cells converted also converted into T_RM cells, although this conversion was to a lower extent as compared with their effector-memory counterparts. Thus, adenoviral vaccine-induced E7-specific CD8⁺ T cells that circulate in the body can convert rapidly into cells with a T_RM phenotype in an antigen-dependent manner.

To investigate the potential of Ad-E7 and Ad-R9F vaccine vectors to provide tumor protection, mice received a subcutaneous TC-1 tumor challenge 50 days after vaccination. The strength of the vaccine-induced CD8⁺ T cell responses correlated with long-term protection against subcutaneously transplanted TC-1 tumors (figure 3A,B). To confirm the importance of the vaccine-induced CD8⁺ T cells in tumor protection, these cells were specifically depleted by anti-CD8 antibodies. On CD8⁺ T cell depletion, none of the vaccinated mice were protected (figure 3A,B). Subsequently, we validated in a second model the tumor protection efficacy of adenoviral vaccines. For this, we used the HPV16-transformed C3 tumor model. In this model, Ad-R9F vaccination resulted in substantial protection against C3 tumors in a dose-dependent manner, and this protection also required CD8⁺ T cells (figure 3C,D). Thus, adenoviral vaccines elicit E7-specific CD8⁺ T cells that protect against subcutaneous tumors.

To study the importance of circulating and tissue-resident CD8⁺ T cells elicited by adenoviral vaccines for tumor protection, we trapped the circulating cells in the lymph nodes by the sphingosine 1-phosphate receptor modulator FTY72037 or depleted either the circulating CD8⁺ T cells by low dose anti-CD8 antibody treatment or tissue-resident CD8⁺ T cells by CXCR3 antibodies (figure 4A,B). Compared with untreated vaccinated mice, FTY720 treatment resulted in a substantial reduction in tumor protection. Moreover, on low-dose anti-CD8 antibody treatment, none of the vaccinated mice were protected against tumor outgrowth (figure 4C,D). Both FTY720 treatment and low dose anti-CD8 affected the total amount of CD8⁺ T cells in the circulation (figure 4A), and also those present in the tumor (online supplemental figure S3). Nevertheless, the few CD8⁺ T cells that were present in the tumor still converted into T_RM cells, indicating that T_RM formation is not blocked but severely limited and is likely due to limited infiltration from the circulation (online supplemental figure S3). Thus, diminishing the number of circulating CD8⁺ T cells prevent robust accumulation of T_RM cells in the tumor, indicating that the circulating CD8⁺ T cells are an important source for sufficient formation of tumor-resident CD8⁺ T cells.

To directly study the role of T_RM cells in tumor control, we tested whether CXCR3 antibodies, known to specifically deplete T_RM cells in the liver,38 also deplete T_RM cells in subcutaneous tumors. Treatment with CXCR3 antibodies resulted in a nearly complete depletion of liver T_RM cells and also of T_RM cells in the tumor(figure 4B), whereas CD8⁺ T cells in the circulation were not depleted (figure 4A). CXCR3 treatment did however not affect skin T_RM (online supplemental figure S4), which corroborates the notion that circulation CD8⁺ T cells are vital for the formation of T_RM cells in the tumor. To examine possible effects of CXCR3 antibody treatment on the migration of CD8⁺ T cells, we labeled the circulating CD8⁺ T cells by injecting APC-labeled CD8 antibodies 3 min before sacrifice. CXCR3 treatment did not alter the migration properties of CD8⁺ T cells in the tumor, spleen or liver (online supplemental figure S5). Taken together, CXCR3 antibody treatment does not affect T cell migration into tissues or the number of skin T_RM cells but results in specific depletion of T_RM cells in the tumor. Importantly, on the specific T_RM cell depletion in the tumor by CXCR3 antibodies, Ad-R9F vaccination did not show any efficacy and all mice succumbed on tumor challenge (figure 4C,D).

To corroborate the importance of T_RM cells for tumor protection, we setup a model to assess the efficacy of the E7-specific CD8⁺ T cell-eliciting adenoviral vectors against tumor development in the liver. Previous studies showed that on intrasplenic injection of tumor cells, the tumor cells are initially trapped in the liver vasculature followed by invasion of the surrounding tissue.39 40 By performing this procedure with luciferase-expressing TC-1 tumor cells, we were able to track the metastatic tumor development in the liver by in vivo bioluminescence (figure 4E,F). In contrast to unvaccinated mice, 10⁹ IU Ad-R9F vaccinated mice, with intact numbers of T_RM cells, were protected against liver tumor outgrowth. Strikingly, T_RM depletion in vaccinated mice resulted in considerable outgrowth of the tumor cells (figure 4G). Thus, T_RM cells mediate critical immunity against tumors growing in the skin and liver.

Costimulatory molecules regulate T cell activation and differentiation, and can serve as therapeutic targets for modulating T cell responses. At day 3 after adenoviral vector vaccination, the costimulatory molecules CD80 and CD86 were upregulated on CD11b⁺CD11c⁺ cells in the liver (figure 5A). To directly dissect the role of CD80/CD86-mediated costimulation in the induction of memory inflation and the formation of T_RM cells after adenovirus-vector vaccination, we evaluated mice deficient in these costimulatory molecules and in addition abrogated CD80/CD86 interactions at different time points post 10⁹ IU Ad-R9F vaccination using blocking antibodies. Cd80/86 ^-/- mice were not able to generate inflationary E7-specific CD8⁺ T cell responses in the blood after Ad-R9F vaccination (figure 5B). Consistently, in adenovirus-vector vaccinated wild-type mice, but not in vaccinated Cd80/86 ^-/- mice, a functional cytolytic E7-specific CD8⁺ T cell response was generated that killed E7⁺ target cells in vivo (figure 5C,D). Blockade of CD80/CD86-mediated costimulation at different time points after adenovirus-vector vaccination indicated that predominantly the initial CD80/CD86 interactions are required for the initiation of inflationary E7-specific CD8⁺ T cell responses in the blood (figure 5B). The deprived E7-specific CD8⁺ T cell response in mice with deficient CD80/CD86-mediated costimulation, related to a complete lack of E7-specific CD8⁺ T_RM cell formation in liver and lungs of these mice (figure 5E). In contrast to CD80/CD86, CD70 was not upregulated on vaccination and consistently the abrogation of CD70 did not affect the generation or cytolytic function of E7-specific CD8⁺ T cells in the circulation or in the tissues (figure 5A–E). Thus, in particular CD80/CD86-mediated costimulation is essential for the inflationary E7-specific CD8⁺ T cell response and development of T_RM cell formation on adenoviral vaccination.

Next, we evaluated whether enhancing CD80/86-mediated costimulation by blocking the inhibitory receptor CTLA-4, which competes with CD28 for CD80/86, could positively impact the formation of adenoviral vaccine-elicited CD8⁺ T_RM cells, and hence the protection against subcutaneous tumors. To create a therapeutic window, mice were vaccinated with 10⁷ IU Ad-R9F and subsequently treated with CTLA-4 blocking antibodies during the first 2 weeks after vaccination. Blockade of CTLA-4 delayed tumor outgrowth as compared with control treated vaccinated mice, and this delay of tumor outgrowth was dependent on T_RM cells (figure 6A,B). Although CTLA-4 blockade resulted in similar frequencies of circulating E7-specific CD8⁺ T cells, a large portion of these cells displayed an effector-memory phenotype (figure 6C). Moreover, the formation of CD8⁺ T_RM cells in the liver was enhanced by CTLA-4 blockade (figure 6D). Thus, temporal improving CD80/CD86-mediated costimulation increases CD8⁺ T_RM cell formation leading to delayed tumor outgrowth.