Sarcoma Eradication by Doxorubicin and Targeted TNF Relies upon CD8⁺ T-cell Recognition of a Retroviral Antigen

Soft tissue sarcomas (STS) are a heterogeneous group of more than 50 malignancies, which account for 1% of all adult and 15% of pediatric cancers. These tumors form in nonepithelial extraskeletal tissues, including muscle, fat, and fibrous supporting structures, arising mainly from embryonic mesoderm (1). The primary tumor is typically treated by surgical excision plus radiotherapy. The overall 5-year survival rate for STS patients is 50% to 60% in adults and 75% in children, depending on the tumor size, depth, site, grade and histologic subtype (2). However, if the tumor develops into a metastatic disease, the prognosis for the patient is dismal, with an expected median overall survival of only 8 to 12 months. The majority of STSs are treated with doxorubicin-based regimens as first-line chemotherapy. The reported response rates for this type of intervention ranges between 12% and 24% (3). Combination therapies involving doxorubicin, ifosfamide, and dacarbazine were developed to increase response rates and improve survival of patients. However, the combinations have not been proven to substantially increase response rates, progression-free survival or overall survival, while increasing the toxicity burden (2, 4). Thus, the improvement of treatment modalities for metastatic STSs has remained an unmet medical need.

More than 100 years ago, William Coley observed a spontaneous regression of an established STS in a patient who had experienced a streptococcal infection of the skin (erysipelas infection) and went on to treat patients with a mixture of heat-inactivated mycobacteria (hence named Coley toxin). A comparative evaluation, published in Nature in 1992, revealed that more than 50% of sarcoma patients, who had been treated with Coley toxin, enjoyed durable complete remissions (CR) from the disease (5), while CRs are virtually never observed with modern chemotherapy (2, 3). The author concluded that “in the light of the predominantly disappointing results with chemotherapy in the treatment of such advanced stages of cancer,” an approach based on Coley toxin or on related immunostimulatory strategies “is certainly a reasonable place to concentrate our efforts.”

The endotoxins in Coley vaccine stimulated the release of high concentrations of TNF, among other proinflammatory cytokines. The sensitivity of tumors of mesodermal origin to TNF has prompted numerous investigations. Carswell and colleagues (6) used a sarcoma in the initial discovery of TNF, whereas Berendt and colleagues (7) used STS to describe the essential importance of tumor immunogenicity and a corresponding T-cell immune response to the curative effects of endotoxin therapy. The systemic use of recombinant TNF was not successful in the clinic. However, the use of TNF in isolated limb perfusion procedures in combination with melphalan for the treatment of inoperable STSs was found to be potently active even for the eradication of large tumor masses and has received marketing authorization in Europe (8).

We have previously reported that the therapeutic index of murine TNF can be dramatically enhanced by fusion to suitable antibody fragments capable of selective localization to the tumor environment. In particular, a strong activity in mouse models of sarcoma has been observed for TNF fusions to the F8 or the L19 antibody, specific to the alternatively spliced EDA and EDB domains of fibronectin, respectively (9, 10). These splice isoforms of fibronectin are virtually undetectable in normal adult tissues (exception made for placenta, endometrium, and some vessels in the ovaries; ref. 11), but are abundantly found around the tumor blood vessels in most malignancies (11, 12). In two immunocompetent mouse models of STS, doxorubicin did not exhibit any detectable inhibition of tumor growth, while its combination with F8-TNF was curative (9). Similarly, potent therapeutic activity in sarcoma has been reported for L19-TNF in combination with melphalan (10). The fully human version of L19-TNF has been shown to be clinically active in isolated limb perfusion procedures (13) and for the intralesional administration to patients with stage III melanoma (14). The systemic administration of L19-TNF has been found to be safe for up to 1 mg/patient. A clinical trial featuring a combination with doxorubicin in STS patients is currently ongoing in Italy and Germany (Eudra-CT no. 2012-000950-75).

Here, we present a detailed analysis of how the antibody-based delivery of TNF to sarcoma potently synergizes with doxorubicin and confers a protective immunity against homologous and heterologous tumors. The combination of T-cell receptor and exome sequencing, as well as the analysis of MHC class I–bound peptides, led to the identification of the retroviral AH1 peptide (SPSYVYHQF) as a contributor to the tumor rejection process.

All tumor cell lines were obtained from the ATCC with the exception of C51 colon carcinoma and F1F fibrosarcoma (both kindly provided by M.P. Colombo, Istituto Nazionale Tumori, Milan, Italy). Cell lines were received between 2010 and 2017, expanded, and stored as cryopreserved aliquots in liquid nitrogen. Cells were grown according the supplier's protocol and kept in culture for no longer than 2 months. Authentication of the cell lines also including check of postfreeze viability, growth properties, and morphology, test for mycoplasma contamination, isoenzyme assay, and sterility test were performed by the cell bank before shipment. Eight-week-old female BALB/c mice were purchased from Charles River Laboratories (Germany). All animal experiments were performed under a project license granted by the Veterinäramt des Kantons Zürich, Zürich, Switzerland (42/2012, 27/2015) in agreement with Swiss regulations.

The F8-TNF immunocytokine was produced as described previously (9). Doxorubicin was purchased in the commercially available form of 10 mg/5 mL solution for injection (Sandoz Pharmaceuticals AG). Rat anti-CD4 (GK1.5, BioXCell), rat anti-CD8 (YTS169.4, Bio X Cell) and rabbit anti-Asialo GM1 (Wako Chemicals) antibodies were used for in vivo depletion.

Exponentially growing WEHI-164 cells were harvested, repeatedly washed, and resuspended in saline prior to injection. Tumor cells were implanted subcutaneously in the right flank of BALB/c mice using 3 × 10⁶ cells per mouse. Tumor volume was calculated as follows: [length (mm) × width (mm) × width (mm)]/2. Mice were randomly divided into two groups and injected into the lateral tail vein. The treatment group (n = 37) received an initial injection of doxorubicin (5 mg/kg) followed by three injections of 2 μg F8-TNF every 48 hours starting on the same day. The negative control group was treated with saline (n = 5). For the in vivo depletion of natural killer (NK), CD4⁺, and CD8⁺ cells, WEHI-164 tumor–bearing mice (n = 5/group) were repeatedly injected intraperitoneally with 30 μL anti-Asialo GM1, 250 μg anti-CD4, or 250 μg anti-CD8 antibodies on days 2, 5, 8, and 11 after tumor implantation. An additional group (n = 5) was injected with 250 μg anti-CD4 antibodies on days −1, 2, 4, and 8. A saline group (n = 5) and a treatment group without depletion (n = 5) were included as controls. Animals were euthanized when tumor volumes reached a maximum of 2,000 mm³ or weight loss exceeded 15%.

Cured mice were injected subcutaneously with 3 × 10⁶ WEHI-164, 1 × 10⁶ C51, 1 × 10⁶ CT26, or 1 × 10⁶ F1F cells 21 days after completion of the therapy study. In addition, tumor challenge of cured mice with WEHI-164 cells was also tested in NK cell-, CD4⁺ T cell-, and CD8⁺ T-cell–depleted mice (n = 5 per group). Depletion antibodies were injected intraperitoneally every third day starting on day 2 before tumor implantation. As controls, naïve BALB/c mice of the same age were also injected with the tumor cells (n = 3 per group) to monitor tumor growth and cell viability after injection.

APC-conjugated H-2L^d tetramers were produced as described by Toebes and colleagues (15) with minor modifications. Plasmids of the H-2L^d heavy chain and of human β2-microglobulin were a kind gift from A. Oxenius (ETH Zürich, Zürich, Switzerland). AH1 (SPSYVYHQF) and p29 (YPNVNIHNF) peptides were ordered from Biomatik.

Tumors were excised and immediately embedded in frozen section medium (Thermo Fisher Scienfic) 48 hours after the first injection. Staining was performed on 10-μm cryosections fixed in ice-cold acetone. For immunofluorescence analysis, antibodies against the following antigens were used: CD11c (N418, BioLegend), F4/80 (BM8, eBiosciences), Asialo GM1 (Wako Chemicals), CD4 (GK1.5, BioLegend), CD8 (53-6.7, BioLegend), Foxp3 (FJK-16s, eBiosciences), EDA (F8), and CD31/PECAM-1 (390, Invitrogen; M-185, Santa Cruz Biotechnology). Primary antibodies were detected either with Alexa Fluor 488- or Alexa Fluor 594–coupled secondary antibodies (Invitrogen). For immunofluorescence staining with H2-L^d tetramers, freshly frozen tissue sections were generated from F8-TNF/doxorubicin–treated and saline-treated mouse spleens from day 18 of the therapy, from tumors 48 hours after the first injection and from spleens of healthy BALB/c mice. Immunofluorescence experiments with H2-L^d tetramers were performed according to established protocols (16). Stained slides were analyzed with an Axioskop2 mot plus microscope (with 10×/0.30 and 20×/0.50 objective lenses, Zeiss) and documented with an AxioCam color or an AxioCam MRm (Zeiss) camera, using the AxioVision software (4.7.2 Release, Zeiss). Images were processed in Photoshop CS6 for Macintosh (Adobe), and the number of cells was counted manually using ImageJ.

F8-TNF/doxorubicin–treated and saline control mouse spleens were excised at day 18 of the therapy. CD8⁺ T cells were isolated from single-cell suspensions through depletion of CD8⁻ leukocytes using the Dynabeads Untouched Mouse CD8 Cells Kit (Thermo Fisher Scientific). RNA was isolated using TRIzol reagent (Thermo Fisher Scientific) according to the manufacturer's protocol. Reverse transcription was performed using an OneStep RT-PCR Kit (Qiagen) with iRepertoire mouse T-cell receptor beta primers (MTBIvc, iRepertoire Inc.) under the conditions specified by iRepertoire. PCR products were purified using SPRIselect magnetic beads (Beckman Coulter) and subjected to a second PCR using a Multiplex PCR Kit (Qiagen) with iRepertoire primers containing communal Illumina sequencing adaptors. Quality and quantity of the libraries were determined using Qubit quantitation (Thermo Fisher Scientific) and the D1000 ScreenTape Assay (Agilent). Final libraries were pooled according to the manufacturer's instructions and sequenced on an Ilumina MiSeq platform.

Sequencing analysis, including filtering and sequence alignment, was performed by iRepertoire Inc. The Lorenz curve and the Gini coefficient are commonly used to measure the inequality among values in economic distributions, such as the levels of income. However, in recent studies, they have also been applied to characterize the frequency distribution of T-cell clones (17). The Lorenz curve was constructed by plotting the cumulative percentage of the whole CD8⁺ T-cell repertoire sequences on the y-axis relative to the cumulative proportion of unique T-cell receptor β (TCRβ) clonotype sequences on the x-axis. The Gini coefficient corresponds to the ratio of the area under the Lorenz curve and the uniform distribution line to the total area under the uniform distribution line. It ranges between 0 and 1, a value of 1 indicating a monoclonal sample and 0 meaning equal distribution of all clones. The Gini coefficient can be approximated with the trapezoidal rule as follows:

Unique clonotype sequences were ranked from least frequent (k = 1) to most frequent (k = n). X_k is the cumulative proportion of unique clonotype sequences 1 to k and Y_k is the cumulative percentage of clonotypes 1 to k.

DNA from the tumor cell line and from liver cells of a healthy 8-week-old female BALB/c mouse was extracted using the QIAamp DNA Mini Kit (Qiagen). Exome capture was performed with the use of the SureSelectXT Mouse All-Exon Kit (Agilent Technologies). Exome libraries were enriched using the SureSelectXT Target Enrichment System for Illumina Paired-End Multiplexed Sequencing Library protocol for 3 μg DNA (Version B4, August 2015, Agilent Technologies) and sequenced on the HiSeq 2500 v4 platform (Illumina). Sequence reads were aligned to the mm10/GRCm38 genome using Bowtie 2 version 2.2.6 (options: no-mixed, no-discordant, very-sensitive; ref. 18). The average on-target coverage was 159× with approximately 93% of the target sequence being covered by at least 30 reads. GATK v3.5 (19) was used for base quality recalibration (known sites from dbSNP v137), variant calling/genotyping (HaploTypeCaller, GATK), and variant filtering. We used a hard-filtering strategy with the following criteria: DP > 20, FS < 30, QD > 2. The resulting variants were annotated with snpEff version 4.11 (20), and only variants with impact levels “moderate” and “high” were kept. Finally, variants that were also detected in wild-type BALB/c were removed from the candidate list.

Affinity purification of H-2 class I complexes was performed as previously described for human HLA class I molecules (21). Lysis was performed at a cell density of approximately 2.5 × 10⁷ cells/mL, and H-2 class I complexes were purified from lysates by incubation with M1/42 (Bio X Cell) antibody-coupled resin.

Analysis of H-2 class I peptides by LC/MS was carried out as previously described for human HLA class I peptides (22). MS/MS data were searched against all murine proteins (84,716 entries) of the UniProt database, downloaded on June 5, 2016. To this protein database, additional sequences were added corresponding to nonsynonymous mutations identified by exome sequencing. Both wild-type and mutated sequences were added as 39mers, with the mutation present at position 20. A total of 1,735 mutated and 1,727 corresponding wild type sequences were added. The following analysis settings were used with SEQUEST: (i) no-enzyme (unspecific); (ii) precursor mass tolerance 4 ppm; (iii) fragment mass tolerance 0.02 Da; and (iv) two variable modifications (oxidation of methionine, and phosphorylation of serine, threonine, or tyrosine). FDRs were calculated with the Percolator plug-in. MaxQuant parameters were set as described previously (23).

Bioinformatics processing using the DeepQuanTR (24) software suite was performed as described previously (22). The six replicate analyses of the WEHI-164 MHC class I peptidome were aligned, resulting in one dataset. Sequence identifications was removed by DeepQuanTR if ambiguity remained after alignment, a process that improves data quality and is necessary when combining results from two or more search engines (22). Peptide to protein annotation was also performed with the DeepQuanTR software (24).

Unique peptide identifications with a length between 8 and 11 amino acids were subjected to MHC class I binding prediction analysis using NetMHCpan 3.0 net (25). Each peptide was assigned the minimal rank for the three BALB/c H-2 alleles D^d, K^d, and L^d. Peptides were annotated as being predicted to bind if the rank calculated by NetMHCpan was below 2% for at least one of the three alleles.

All 9mers of the WEHI-164 MHC class I peptidome were subjected to GibbsCluster-1.1 Server (26) analysis using the default settings without alignment (1–5 clusters, and “use trash cluster to remove outliers” enabled). On the basis of the resulting clusters, 9mers were annotated to D^d, K^d, and L^d alleles according to the motifs presented by the SYFPEITHI database (27).

Spleens were excised at day 18 of the therapy or at day 3 after rechallenge, mashed in Red Blood Cell Lysis Buffer (Roche), and passed through a 40-μm cell strainer (EASYstrainer, Greiner Bio-One). Single-cell suspensions were either used directly for FACS staining or cultured for 2 days in RPMI1640 (Thermo Fisher Scientific, supplemented with 10% FBS, 2 mmol/L glutamine, 50 μmol/L 2-mercaptoethanol, and antibiotic-antimycotic solution) with 10 μg/mL AH1 or p29 peptide. Afterwards, 100 U/mL IL2 was added, and cells were cultured for another 3 days. For flow cytometry analysis, 1 × 10⁶ splenocytes were incubated for 1 hour at 4°C with H-2L^d tetramer, LIVE/DEAD Fixable Near-IR Dead Cell Stain (Thermo Fisher Scientific), and fluorochrome-conjugated antibodies against CD8 (FITC, KT15, Thermo Fisher Scientific), CD4 (PerCP, GK1.5, BioLegend), and B220 (PerCP, RA3-6B2, BioLegend) in PBS containing 2% FBS and 2 mmol/L EDTA.

For intracellular cytokine staining, splenocyte cultures were stimulated with the AH1 peptide and GolgiStop for 4 hours according to the manufacturer's instructions (BD Cytofix/Cytoperm Plus Fixation/Permeabilization Kit, BD Biosciences). Cells were surface stained as described above. Following fixation and permeabilization, cells were stained against mouse IFNγ (PE, XMG1.2, BioLegend) for 1 hour at 4 °C. Cells were analyzed on a CytoFLEX cytometer (Beckman Coulter), and data were processed using FlowJo (v.10, Tree Star).

In vitro cytotoxicity assays were performed using the CytoTox96 Non-Radioactive Cytotoxicity Assay Kit (Promega) according to the manufacturer's protocol. Briefly, splenocytes of cured mice were cultured as described above and used as effector cells. Tumor cells (5 × 10³) were cocultured with the effector cells at a ratio of 1:0.1, 1:1, 1:5, 1:25, and 1:50 in 96-well plates for 4 hours. Each ratio was measured in quadruplicates. Absorbance (A) values were measured at 492 nm. The percentage of specific lysis for each ratio was calculated as follows:

RNA of cell lines and tissues of an 8-week-old female BALB/c mouse was prepared using TRIzol (Thermo Fisher Scientific). Two microgram RNA was reverse transcribed with GoScript reverse transcriptase (Promega) using Oligo(dT)₂₃ primer (Sigma Aldrich). The gene expression levels were analyzed in a SYBR Green real-time RT-PCR reaction with the AB7900 HT Fast RT-PCR system (Life Technologies) in three different cell isolates in triplicates. Rplp0 was used as reference gene for normalization. Primers were as follows:

• murine Rplp0 (forward) 5′-AGATTCGGGATATGCTGTTGG-3′,

• murine Rplp0 (reverse), 5′-TCGGGTCCTAGACCAGTGTTC-3′;

• human Rplp0 (forward) 5′-CAGATTGGCTACCCAACTGTT-3′,

• human Rplp0 (reverse), 5′-GGGAAGGTGTAATCCGTCTCC-3′;

• gp70 (forward), 5′-CACCAATTTGAAAGACGAGCC-3′,

• gp70 (reverse) 5′-CAATTCCGCCCATAGTGAGTC-3′.

Data were analyzed using Prism 6.0 (GraphPad Software, Inc.). Statistical significance of in vivo experiments, the in vitro cytotoxicity assay, the usage of T-cell receptor beta joining (TRBJ) and T-cell receptor beta variable (TRBV) gene segments, the number of trimmed TRBJ and TRBV nucleotides, and the CDR3 length distribution were determined with a regular two-way ANOVA test with the Bonferroni posttest. A Student t test was used to assess differences of the Gini coefficients between the F8-TNF/doxorubicin and the saline treatment group. Data represent means ± SEM. P < 0.05 was considered statistically significant (* = P < 0.05, ** = P < 0.01, *** P = < 0.001, **** = P < 0.0001).

BALB/c mice, bearing subcutaneous WEHI-164 fibrosarcoma, were treated with doxorubicin (5 mg/kg) and F8-TNF (three injections of 2 μg) as described previously (9). Figure 1A shows that 29 of 37 mice were cured in the treatment group, while tumors grew in the saline control group. The cancer cure proceeded with the conversion of the tumor mass into a black scab, which eventually fell off without regrowing (Fig. 1B).

Execution of the therapy procedure after selective antibody-based leukocyte depletion revealed a requirement for CD8⁺ T cells and NK cells. CD4⁺ T cells were dispensable if depleted after tumor implantation. However, depletion of CD4⁺ T lymphocytes prior to tumor growth had an impact on the therapeutic outcome (Fig. 1C).

Cured mice were rechallenged with WEHI-164 or with heterologous C51 or CT26 tumor cells. In all cases, the tumors did not grow in the cured mice, whereas they grew in naïve control mice. Depletion experiments indicated that the tumor protection was mediated by CD8⁺ T lymphocytes (Table 1).

A microscopic analysis of tumor sections, performed 48 hours after the first intravenous injection of mice with the F8-TNF/doxorubicin combination or with saline, confirmed that the majority of the tumor had been converted into necrotic tissue, most probably by a direct cytotoxic activity of doxorubicin and the TNF moiety (Fig. 2A). Analysis of the leukocyte infiltrate, using CD11c, F4/80, asialo GM1, CD4, CD8, and Foxp3 antibodies, revealed an increased infiltration of macrophages, dendritic cells, NK cells, and CD8⁺ T cells into the tumor mass, while CD4⁺ T cells and regulatory T cells were essentially not affected (Fig. 2B and G). Fibronectin molecules, containing the alternatively spliced EDA domain that is targeted by the F8 antibody, could still be found in the dying tissue (Fig. 2H).

Exome sequencing of WEHI-164 was performed to identify mutations in protein-coding regions of the DNA of the tumor cell line. The analysis revealed the presence of 1,648 missense mutations, 85 nonsense variants, and 94 short insertions and deletions in the WEHI-164 exome compared with wild-type BALB/c (Supplementary Table S1). The high mutation rate observed in WEHI-164 cells is in line with similar recent analyses of mouse tumor exomes (28, 29).

Using a combination of immunocapture and mass spectrometry, we have recently reported the confident identification of thousands of peptide sequences bound to HLA class I on human tumor cells (22). For this work, the methodology was adapted to allow the study of the murine MHC class I peptidome, using the M1/42 antibody for immunocapture. WEHI-164 cell lysates were analyzed in six independent experiments, which led to the identification of 4,639 unique peptides with high confidence (i.e., less than 1% FDR). The majority of the identified peptides (4,151, 90%) were 8-11 amino acid long, which represents the typical length of MHC class I–bound peptides (Fig. 3; Supplementary Table S2). Binding prediction analysis revealed that 84% of the 8-11mers were predicted to bind to the cognate MHC class I molecule. The identified 9mers were clustered into the three alleles H-2K^d, H-2D^d, and H-2L^d leading to MHC-specific motifs in agreement with the literature (Fig. 3; ref. 27). Even though mass spectrometric data were interrogated against a database containing not only wild-type sequences, but also the catalog of somatic mutations identified by whole-exome sequencing, no neoepitope was found with high confidence. Seventeen phosphorylated 8-11mers were observed (Supplementary Table S2), which is in line with previous reports on the incidence of this class of epitopes in cancer cell peptidomes (30, 31).

As CT26 and C51 tumors were rejected after the initial cure from WEHI-164 sarcomas, we hypothesized that the three malignancies would share at least one common tumor rejection antigen. The AH1 (SPSYVYHQF) peptide, which is derived from the gp70 envelope protein of the murine leukemia virus, had been previously reported as the immunodominant antigen of CT26 tumor cells (32). Interestingly, we could also identify the AH1 peptide on WEHI-164 cells by MHC I peptidome analysis (Supplementary Table S2). Presentation of the retroviral peptide on C51 and CT26 cells was verified by peptidome analysis (data not shown), and a high expression of the gp70 gene was detected in all three cell lines (Fig. 4).

Tetramer analysis was performed with recombinant H-2L^d molecules loaded with AH1 (SPSYVYHQF) or with p29 (YPNVNIHNF), an irrelevant peptide (33) serving as negative control. Spleen sections from F8-TNF/doxorubicin treated, saline treated, and from healthy BALB/c mice were compared in immunofluorescence analysis. A specific staining of AH1-binding T cells was observed in sections of mice, which had been cured with F8-TNF and doxorubicin. In parts of the analyzed sections, more than 10% of the CD8⁺ T cells were found to recognize the retroviral AH1 peptide (Fig. 5A). Flow cytometry analysis revealed an expansion of AH1-specific CD8⁺ T cells in F8-TNF/doxorubicin–treated mice compared with saline-treated mice. Highest levels of AH1-specific CD8⁺ T cells were found in cured mice that were rechallenged in vivo with WEHI-164 cells (Fig. 5B). Samples from healthy mice or from tumor-bearing mice, which had received saline injections, were largely negative in these analyses (Fig. 5; Supplementary Figs. S1 and S2).

We performed a rechallenge experiment with syngeneic F1F fibrosarcoma cells to demonstrate the importance of the anti-AH1 response in cross-protection of WEHI-164–cured mice against other tumors. F1F had been previously reported to be gp70 negative (34). Analysis of gp70 expression by qPCR revealed an extremely low expression level of the retroviral protein compared with other BALB/c cell lines (Fig. 4). Interestingly, cured mice did not reject F1F tumor cells (Table 1). In addition, an in vitro cytotoxicity assay revealed a specific lysis of CT26 and WEHI-164 cells by AH1-stimulated splenocytes, whereas no specific lysis of F1F cells could be detected (Fig. 5C).

To learn about the dynamics of the TCR sequences as a result of the therapy intervention, CD8⁺ T cells were purified from spleens of mice treated with F8-TNF/doxorubicin or saline, 18 days after the start of therapy. TCRβ sequencing libraries of the isolated T cells were constructed and analyzed on the Illumina MiSeq platform. In total, 41,200 ± 2,571 and 46,007 ± 25,379 unique TCRβ sequences were determined in the two study groups. The average number of productive TCRβ sequence reads was 262,318 ± 36,756 for the treatment group and 1,012,046 ± 198,544 for the saline-treated group samples. A significant increase in sequence diversity was observed for lymphocytes from the F8-TNF/doxorubicin group, as evidenced by using the Gini coefficient (Fig. 6A and B). Analysis of the usage of V and J segments in the TCR β-chain revealed small but significant changes in the usage of the TRBV 5 and TRBV 19 gene segments, the frequency of trimmed TRBJ and TRBV nucleotides, and the CDR3 sequence length distribution (Fig. 6C–E).

Ten sequences of the beta chain of T cells specific to the AH1 peptide have recently been reported on the basis of anticancer peptide vaccination experiments (35). Nine of these sequences were found in our analysis (Supplementary Table S3), thus providing additional evidence for the specific recognition of AH1 in mice with WEHI-164 sarcomas.

In this article, we showed that immunocompetent mice, bearing syngeneic WEHI-164 fibrosarcoma, could be cured using a combination of the F8-TNF immunocytokine and doxorubicin. In contrast, the use of doxorubicin as a single agent showed no detectable tumor growth inhibition in sarcoma-bearing mice, confirming the essential contribution of targeted TNF (9). Anthracyclines are commonly used for the first-line treatment of metastatic STSs in humans but are rarely associated with objective responses and lead to median progression-free survival of approximately 4 months (36, 37).

Cancer cures in mice crucially relied on the action of CD8⁺ T cells and of NK cells, as evidenced by depletion experiments. Cured mice could reject subsequent challenges with WEHI-164, C51, or CT26 cells, in a process that was due to the action of CD8⁺ T cells. The bulk of the neoplastic mass is rapidly killed by the combined biocidal activity of F8-TNF and doxorubicin. However, CD8⁺ T cells are crucially important for the selective elimination of the last residual tumor cells and for the development of protective immunity.

It is often assumed that immune rejection of solid tumors relies on the recognition of mutated peptides presented on MHC class I by CD8⁺ T cells (28, 38). However, in this work and in other recent reports, mass spectrometric analysis of MHC class I complexes has shown that mutated peptides are rarely found (21, 39, 40). For example, only one mutated peptide could be observed among over 10,000 peptide sequences presented on HLA class I in human melanoma cells (21). In our work, over 1,700 nonsynonymous mutations were observed in exome sequencing of WEHI-164 cells compared with wild-type BALB/c, but no mutated epitope could be found to be associated with MHC class I. However, the analysis of the MHC class I peptidome of WEHI-164 allowed the identification of potential tumor-associated antigens and in particular of the AH1 peptide (SPSYVYHQF, derived from the gp70 envelope protein of the murine leukemia virus), which was also presented on C51 and CT26 tumor cells. The SPSYVYHQF sequence was recognized by cognate T cells, which had been expanded as a result of the F8-TNF/doxorubicin pharmacologic intervention. Flow cytometry analysis and the staining of spleen sections revealed a specific detection of SPSYVYHQF by CD8⁺ T cells, which had increased in number after cancer cure, while being largely undetectable before pharmacological intervention and in suitable negative controls. An in vitro cytotoxicity assay confirmed the cytotoxic potential of AH1-specific T cells against gp70⁺ tumor cells. F1F, a syngeneic BALB/c fibrosarcoma cell line that showed extremely low levels of gp70 expression, was not rejected by WEHI-164–cured mice and F1F cells were resistant to lysis by AH1-specific T cells in vitro. These findings provide additional evidence in support of the contribution of an anti-AH1 response in the treatment of WEHI-164 tumors and in providing cross-protective immunity.

Immune responses to endogenous retroelements have been a matter of intense investigations. Retroviral sequences have been found to be abundant in the genome of all vertebrate species that have been studied. Many of the endogenous retroviral elements that have been integrated in the host germline have retained the capacity to replicate (41). As for self-peptides, thymic presentation of these retroviral products leads to immunologic tolerance due to deletion of the respective TCR or BCR specificities (41). However, adaptive immunity to endogenous retroviral antigens can be triggered if the retroelements are transcriptionally induced after the establishment of the adaptive immune cell repertoire. Induction of retroviral protein expression has been associated with chronic infection and autoimmune diseases both in mice and humans (42–44). For example, human endogenous retrovirus (HERV) antigens were proposed as putative autoantigens for the development of systemic lupus erythematosus (45). In addition, adaptive immunity to retroelements is particularly relevant in cancer, as transformed cells typically exhibit major epigenetic alterations compared with healthy cells. Consequently, recognition of endogenous retroviral antigens, which are no longer transcriptionally repressed, has been frequently detected in cancer (46). Like other cancer testis antigens, the expression of these retroviral proteins is often restricted to germ cells and testis. The expression of the gp70 envelope protein of the endogenous murine leukemia virus had previously been shown to be restricted to mouse cancer cell lines (47), and products of the HERV-K family have shown abundant expression in various human cancers (48). In addition, retroviral antigens have been shown to elicit a potent T-cell activity against murine tumors, as well as human cancers (49).

It is thus conceivable that peptides similar to SPSYVYHQF may be upregulated in human malignancies and displayed on tumor cells, thereby contributing to the tumor surveillance process by T cells. This issue may be particularly relevant for STSs, as these tumors tend to be highly sensitive to the action of TNF (5). Encouraged by promising results in exploratory clinical trials in sarcoma patients, L19-TNF (a fully human fusion protein specific to the EDB domain of fibronectin) in combination with doxorubicin is about to enter phase III clinical trials as a first-line treatment for different subtypes of STSs.

D. Neri is a co-founder and shareholder at Philogen SpA. No potential conflicts of interest were disclosed by the other authors.

Conception and design: P. Probst, D. Neri

Development of methodology: P. Probst, D. Neri

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P. Probst, J. Kopp, D. Ritz, T. Fugmann, D. Neri

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P. Probst, D. Ritz, T. Fugmann

Writing, review, and/or revision of the manuscript: P. Probst, D. Neri

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Oxenius, M.P. Colombo, D. Neri

Study supervision: D. Neri

We would like to thank Dr. J. Kühn-Georgijevic (Functional Genomics Center Zurich) for the help in preparing the exome sequencing libraries and Dr. L. Opitz (Functional Genomics Center Zurich) for the analysis of the NGS data.

D. Neri received financial support from the ETH Zürich, the Swiss National Science Foundation (grant nr. 310030B_163479/1), the ERC Advanced Grant “ZAUBERKUGEL,” and the Federal Commission for Technology and Innovation (KTI, grant nr. 12803.1 VOUCH-LS).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.