CD40 Agonist Targeted to Fibroblast Activation Protein α Synergizes with Radiotherapy in Murine HPV-Positive Head and Neck Tumors

With an incidence of about 890,000 new cases per year, head and neck cancer is the seventh most common cancer worldwide (1). Tobacco smoking, alcohol consumption, and human papillomavirus (HPV) infection represent the main risk factors in the development of head and neck squamous cell carcinoma (HNSCC). HPV-associated HNSCC (HPV⁺-HNSCC) represents a distinct subgroup of head and neck cancer with improved outcomes compared with HPV-negative HNSCC. This results from a combination of unique biologic features of HPV⁺-HNSCC and favorable patient characteristics such as younger age and a minor role for tobacco and alcohol abuse. These patients have shown enhanced and durable benefit from radiotherapy and dose deescalation trials are currently ongoing (2). However, further optimization of treatments by integrating novel combination strategies to enhance therapeutic index and limit toxicity are still required. Among these novel options, immunomodulation appears a promising target as tumor microenvironment (TME) characterization in HNSCC biopsies reveals that it is one of the 10 most immune-infiltrated cancers (3, 4). In this regard, the proven immunostimulatory properties of hypofractionated radiotherapy (RT) make this therapeutic modality a good combination partner for immunotherapy in this type of cancer (5).

The costimulatory receptor CD40 is a member of the TNF receptor family expressed on the surface of antigen-presenting cells (APC), such as B cells, dendritic cells (DC), and macrophages (6). It is also expressed on epithelial, stromal, and endothelial cells and platelets. Engagement by its natural ligand CD40L, present on recently activated T lymphocytes, promotes the maturation and antigen presentation capabilities of the APCs (7). Agonistic CD40 mAbs have shown effective preclinical results in different cancer types such as melanoma, lung adenocarcinoma, and pancreatic cancer (8–11). Both the adaptive and innate immune system drive the main mechanisms of action of the agonistic CD40 mAbs. Indeed, the activation of the CD40 signaling in DCs results in a more efficient priming of T lymphocytes together with robust immunologic memory generation. CD40 agonists have shown clinical efficacy, yet serious toxicity has been described to be associated with their systemic administration and has, unfortunately, limited their use in patients (7, 12). To prevent these adverse effects, intratumoral injections of agonists have been used in many preclinical settings (13, 14). However, the feasibility of this approach in the clinic is limited to tumor accessibility and the expression of Fc receptors (FcR) for antibody cross-linking at the tumor bed (15, 16). To overcome these limitations, a protein was engineered that simultaneously targets CD40 and fibroblast activation protein α (FAP), expressed on tumor stroma and, to a lesser degree, on lymph node fibroblastic reticular cells (FRC). As the Fc region of the engineered agonists contains mutations abrogating cross-linking by the Fcγ receptors, FAP-CD40 molecules can only induce CD40 activation when cross-linked via FAP-expressing cells ensuring the targeting to the stroma of solid tumors (17, 18).

In this study, using a clinically relevant mouse model of HPV-associated HNSCC (19), we investigated the safety and therapeutic efficacy of the bispecific antibody (bi-Ab) FAP-CD40 alone or in combination with hypofractionated radiotherapy, which is currently part of clinical practice for the treatment of solid tumors including head and neck. The model mimics human disease in two key aspects: invasive growth at the neck area and intratumoral accumulation of FAP-expressing stromal fibroblasts, which makes it suitable for FAP-targeted therapy and the hypofractionated regimen chosen that has been described as highly immunogenic, safe, and optimal for the study of combinatorial therapies (20).

Interestingly, we report that the combination of FAP-CD40 with local hypofractionated radiotherapy induced complete tumor regressions and durable control of the disease in more than 80% of the treated mice. Moreover, the remarkable antitumor efficacy was associated with absence of toxicity, a remodeling of the tumor immune landscape and a strong immunologic memory formation, which efficiently prevented the development of tumor relapses.

Seven- to 11-week-old female C57BL/6J mice were purchased from Janvier Labs. hCD40Tg mice were provided by Roche Innovation Center Zurich [accompanying paper (21)]. Batf3^−/− mice were kindly gifted by Hans Acha Orbea (University of Lausanne, Lausanne, Switzerland) and bred in our facilities. All animal procedures were carried out under the license VD3173j which was approved by the Veterinary Authority of the Swiss Canton of Vaud.

The mEERL95 cell line (19) was derived from mEERL cell line (obtained from John H. Lee, Sanford Research) and was cultured in DMEM/nutrient mixture F‐12 medium containing GlutaMAX (Thermo Fisher Scientific) and supplemented with 5% FBS (Life Technologies) and 1× HKGS (Thermo Fisher Scientific). T.C. Wu (Johns Hopkins University, Baltimore, MD) kindly provided TC-1 cells in 2009 and they were cultured in RPMI1640 medium (Life Technologies) supplemented with 10% FBS (Life Technologies), penicillin/streptomycin (Life Technologies), and 5 × 10^–5 M 2-mercaptoethanol (Life Technologies). Both cell lines were cultured in an incubator at 37°C and 5% CO₂ and routinely tested to dismiss Mycoplasma infection.

Mice were subcutaneously injected in the submental space as described previously (19, 22) with 1 × 10⁵ mEERL95 or TC-1 cells resuspended in 30 μL of Hank's Balanced Salt Solution (Thermo Fisher Scientific) and 20 μL of Matrigel (Corning). Tumors were irradiated with a hypofractionated regimen consisting in two consecutive doses of 6 Gy (2 × 6 Gy) at day 10 and 11 post-tumor engraftment. Doses were delivered with an Xrad-225CX-PXi Instrument using a 15-mm collimator that allows to locally irradiate the tumors. At day 11, mice were treated with a single intraperitoneal injection of anti-FAP-CD40 or the corresponding control antibody DP47-CD40 at 18.3 mg/kg in PBS. mEERL95 tumor-bearing huCD40Tg mice were irradiated with the 2 × 6 Gy regimen and treated with a single intraperitoneal dose of FAP-huCD40 at 13 mg/kg (21). Tumor size was followed up using a caliper twice per week and the volume was calculated with the following formula: V = (length × width²)/2.

For the depletion studies, 200 μg/dose of anti-CD8β mAb (clone H35-17-2), anti-CD4 (clone GK1.5), or a RatIgG2b used as Ig isotype control, were given to mEERL95 tumor-bearing mice on days −2 and 0 with respect to the day of treatment initiation. The administration of the antibodies was done every 3 days for 2 weeks. The efficiency of the depletion was checked by flow cytometry at day 18 on peripheral blood lymphocyte. For the in vivo experiments blocking IL12, mice were treated daily with 500 μg of anti-IL12 (clone C17.8, BioXCell) diluted in PBS for 1 week starting the same day of the therapeutic antibody administration. Same doses of a RatIgG2a (clone 2A3, BioXCell) were used as Ig isotype control.

To characterize the development of immunologic memory, 10⁶ mEERL95 or 10⁵ TC-1 tumor cells resuspended in PBS were subcutaneously injected in the flank of mice that had rejected the primary tumors following the therapy. Injections of tumor cell were performed approximately 40 days after developing complete response (CR). Tumor growth was followed up by measuring every 2–3 days with a caliper.

Tumors and regional lymph nodes were harvested 10, 14, 16, or 18 days post-tumor engraftment. Collected tumors and lymph nodes were incubated at 37°C with 1 mg/mL of Collagenase-D (Roche) and 40 μg/mL DNase-I (Sigma-Aldrich) for 45 and 20 minutes, respectively. Afterward, tissue samples were mechanically disaggregated and filtrated using a 70-μmol/L cell strainer (Falcon, BD Biosciences) to obtain single-cell suspensions. In addition, tumor-infiltrating lymphocytes (TIL) were isolated from stromal cells with a 35% Percoll gradient. When needed, red blood cells were lysed for 3 minutes at room temperature with RBC buffer (Qiagen) before flow cytometry staining.

Single-cell suspensions were first incubated with FcR-Block (anti-CD16/32 clone 2.4G2, homemade) for 15 minutes on ice in FACS buffer (2% FCS and 2 mmol/L EDTA in PBS). To characterize the immune populations, samples were surface stained with the following antibody panel in darkness for 20 minutes on ice: CD11c-BV421(clone N418), CD8-BV510 (clone 53-6.7), F4/80-BV605 (clone BM8), NK1.1-BV650 (clone PK136), CD11b-BV711 (clone M1/70), CD4-BV785 (clone RM4-5), CD19-BV785 (clone 6D5), Ly6G-FITC (clone 1A8), CD45-PerCP (clone 30-F11), Ly6C-AF700 (clone HK1.4), IA/IE-APC-Cy7(clone M5/114.15.2), and B220-APC(clone RA3-B2) from BioLegend; and CD80-Pe-Vio770 (clone 16-10A) from Miltenyi Biotec; CD103-PE (clone 2E7), CD3-PE-Cy5.5 (clone 145-2C11), and Foxp3-PeFluor-610 (clone FJK-165) from eBioscience. Intracellular staining of Foxp3 was performed with the Foxp3/Transcription Factor Staining Buffer Set (eBioscience) according to the manufacturer's instructions. LIVE/DEAD Fixable Blue Dead Cell Stain Kit (Thermo Fisher Scientific) was used as a viability marker.

For the ex vivo stimulation assays, lymphocytes from tumor and lymph nodes were cocultured with mEERL95 cells in a 10:1 ratio. The mEERL95 cells were previously incubated with IFNγ (200 ng/mL; ImmunoTools) for 24 hours. The coculture was done in presence of 1 μg/mL of anti-CD28 (clone 37.51) and 10 μg/mL of anti-PD1 (clone RMP1-14, BioXCell), and kept for 16 hours at 37°C in complete medium. Unstimulated cells were used as negative controls. GolgiPlug and GolgiStop (both 1:1,000; BD Biosciences) were added to the cells 4 hours before starting the staining. Upon surface staining with CD45-BV650 (clone 104, BioLegend), CD3-BV510 (clone 17A2, eBioscience), CD8-AF700 (clone 53-6.7, eBioscience), CD4-BV711 (clone RMA5, BioLegend), cells were fixed and permeabilized with a fixation buffer (BioLegend) for 20 minutes on ice. Afterward, cell suspensions were stained with the following intracellular antibodies diluted in Perm/wash buffer 1× (BioLegend): IFNγ-PerCp-Cy5.5 (clone XMG1.2, eBioscience), TNFα-PB (clone MP6-XT22, BioLegend), and Granzyme B-PE/Dazzle594 (clone QA16A02, BioLegend) for 30 minutes on ice. Unstimulated TILs were surface stained at room temperature for 30 minutes with the H-2Db HPV16 E7 Tetramer-RAHYNIVTF-PE. Samples were acquired either with LSRII-SORP and Fortessa flow cytometers (BD Biosciences). Data analyses were performed using FlowJo v10 (FlowJo LLC).

Differences in immune cell counts and frequencies, and tumor growth and mice survival were analyzed with GraphPad Prism (GraphPad Software). t test, ANOVA for multiple comparison, or the corresponding nonparametric tests (Mann–Whitney or Kruskal–Wallis) was used. Log-rank test was used for survival analysis. P <0.05 was considered significant.

FAP is a prolyl endopeptidase highly expressed in the microenvironment of several human solid tumors as compared with the corresponding normal tissue, including HNSCC (Supplementary Fig. S1A). Indeed, in this type of cancer, high levels of FAP are associated with a significant reduction of the overall survival of patients with HNSCC (Supplementary Fig. S1B).

To assess the therapeutic potential of the FAP-CD40 bi-Ab, we used a murine orthotopic head and neck tumor model (mEERL95; ref. 19) that is characterized by expression of FAP in α-SMA–positive cancer-associated fibroblasts (CAF; Fig. 1A), thus resembling stromal expression of FAP in human HNSCC (Supplementary Fig. S1C). First, we investigated the antitumor efficacy and safety of FAP-CD40 alone or in combination with local hypofractionated radiotherapy (consisting of two consecutive doses of 6 Gy) according to the scheme shown in Fig. 1B. FAP-CD40 bi-Ab led to complete tumor regression in 40% of treated mice (2/5) followed by tumor relapse in one mouse (Fig. 1C). On the contrary, the targeting of CD40 with the control antibody DP47-CD40 (in which DP47 replaces the anti-FAP moiety as a germline control) showed no antitumor effect indicating that FAP-CD40 requires cross-linking via FAP to be therapeutically efficient.

Local radiotherapy, given alone or together with the control DP47-CD40 antibody, resulted in tumor growth delay as compared with control groups (Fig. 1C). However, from day 20 after mEERL95 injection, tumors resumed growth, thus resulting in no objective CRs. Interestingly, all tumor-bearing mice responded to the combination of 2 × 6 Gy with FAP-CD40 antibody, which induced complete tumor regressions and durable control in 83% (5/6) of the animals (Fig. 1C). Of note, an additional dose of anti-FAP-CD40 did not improve the outcome of the combination (Supplementary Fig. S2A). Remarkably, all responder mice remained relapse free even after prolonged periods of observation (more than 90 days) showing that this radioimmunotherapy combination not only induces a high rate of tumor regression but also extends survival (Fig. 1D). The efficacy of the FAP-CD40 bi-Ab as a single agent or in combination with hypofractionated radiotherapy was accompanied with absence of systemic toxicity, a common feature observed in the therapy with CD40 agonistic mAbs (12, 23, 24). Tumor-bearing mice treated with the FAP-CD40 antibody did not present body weight loss whereas a single injection of a CD40 mAb, combined or not with radiotherapy, induced a significant loss of body weight with tumor growth control comparable to that mediated by bi-Ab therapy (Supplementary Fig. S2B and S2C).

To investigate further whether FAP-CD40–mediated antitumor activity required FAP-expressing stromal fibroblasts, we used mouse TC-1 lung cancer cells expressing mutant H-ras, HPV16 E6, and E7 proteins (25) but devoid of FAP expression (26). TC-1 tumor cells injected into the submental space resulted in tumors with peritumoral αSMA-expressing CAFs (not shown) and a few intratumoral areas of αSMA-positive, FAP-negative cells (Fig. 2A and B). In line with the results of immunostaining, mRNA expression level of Fap gene was significantly lower in the TC-1 tumors (Fig. 2B). Consistently, and unlike the outcome observed in the mEERL95 model, TC-1 tumors did not respond to FAP-CD40 alone and showed only 9% (1/11) of CRs upon the treatment with either radiotherapy alone or the combination of 2 × 6 Gy + anti-FAP-CD40 (Fig. 2C). To assess FAP-CD40 bi-Ab access to the TC-1 tumor, we intraperitoneally injected the FAP-CD40 Alexa647-labeled antibody into TC-1 and mEERL95 tumor-bearing mice that underwent local irradiation. As shown in Supplementary Fig. S3A, the antibody was detectable in both tumors 4 days after administration. No significant difference in the fluorescence intensity was detected in both TC-1 and mEERL95 tumors (Supplementary Fig. S3B) which present similar CD40 mRNA expression levels (Fig. 2D). These results indicate that the FAP-CD40 bi-Ab is indeed able to target the CD40-expressing cells present in the TC-1 TME but fails to promote protective immunity as no activation of CD40 signaling can occur in the absence of FAP-mediated cross-linking [accompanying article (21)].

As FAP is also present in the lymph node stromal compartment, especially in the FRCs (27) [and (21) accompanying article], we confirmed by IHC that FAP is expressed in cervical lymph nodes from treatment-naïve mice 10 days post-injection of both cell lines (Fig. 2E). The frequency and the expression of FAP on FRCs (gp38⁺CD31⁻) measured by flow cytometry was also similar in TC-1 and mEERL95 tumor-bearing mice (Fig. 2F). These results emphasize the requirement of FAP expression in the tumor stroma.

We sought to analyze whether the 2 × 6 Gy + FAP-CD40 treatment was able to induce tumor-specific long-term protective immunologic memory in the mEERL95 model. Thus, mice that rejected mEERL95 primary tumors upon the combination (long-term responders) were rechallenged with the homologous tumor in the right flank. All mice controlled tumor growth and eventually rejected them (Fig. 3A), indicating the existence of specific immune memory, presumably against the immunodominant epitope of HPV16 E7 protein. To test this hypothesis, cured mice were exposed to a second rechallenge with the TC-1 cell line, which also expresses the E7 protein, in the left flank (Fig. 3B). Only 3 of 8 mice were able to reject TC-1 cells, suggesting that E7 may not be the single immunodominant antigen in this context (Fig. 3B). In addition, around 40%–60% of long-term responder mice (which rejected the mEERL95 primary tumor upon treatment with either FAP-CD40 as single agent or combined with radiotherapy) presented E7-specific T lymphocytes upon in vitro expansion of peripheral blood mononuclear cells (Fig. 3C). In fact, the frequency of E7-specific CD8 T cells (detected by specific tetramer staining) decreased in the mEERL95 TME upon the combination as compared with untreated tumors (Fig. 3D and E). These data are consistent with other reports in which vaccination with E7 peptide, alone or in combination with other immunotherapies, augments E7-specific T-cell responses in tumor preclinical models (28, 29).

Prompted by the encouraging results and with the aim of promoting the translation of this combination therapy to the clinic, we sought to investigate the antitumor properties of a surrogate bi-Ab that is composed of the anti-mouse FAP moiety but targets the human CD40 receptor [FAP-huCD40; accompanying article (21)]. Therefore, we injected mEERL95 tumor cells in the submental space of transgenic mice expressing the human CD40 receptor (hCD40 Tg mice). Following the same schedule of treatment as in wild-type tumor-bearing mice, we locally irradiated the tumors with two consecutive doses of 6 Gy. Afterward, mice were administered an intraperitoneal injection of the FAP-huCD40 antibody coinciding with the second radiotherapy session (Supplementary Fig. S4A). With no CRs observed, FAP-huCD40 as a single agent did not differ in tumor growth or survival from untreated mice (Supplementary Fig. S4B–S4D). Mice that underwent 2 × 6 Gy showed slower tumor growth, tumor regression (4/10), and increased survival compared with FAP-CD40 and control mice; however, none of them completely rejected the tumors (Supplementary Fig. S4B–S4D). The combination of 2 × 6 Gy + FAP-huCD40 resulted in a synergistic therapeutic effect increasing the number of tumor regression (8/10), promoting a decrease in the tumor growth rate and significantly prolonging long-term survival in comparison with the rest of the groups (Supplementary Fig. S4B–S4D).

We next explored the cellular and molecular immune components associated with the antitumor effect of the combination. We first analyzed the intratumoral immune infiltrate 8 days following the first radiotherapy dose. Using a flow cytometry–based 16-color antibody panel (Supplementary Fig. S5A), we observed an overall increase in the number of immune cells (CD45⁺ cells) per mg of tumor in all treated groups compared with PBS and DP47-CD40 (Fig. 4A). Compared with PBS and DP47-CD40 control, the number of CD8 T cells increased significantly upon irradiation. This increase was less pronounced upon FAP-CD40 monotherapy and combination treatment (Fig. 4A; Supplementary Fig. S5B). The number of regulatory T cells (Treg) present in the tumor was only increased upon radiotherapy as compared with control mice, giving rise to an increase in the CD8/Tregs ratio in FAP-CD40, radiotherapy, and combined treatment groups (Fig. 4A). Interestingly, CD8 cells/macrophages ratio increased only in the 2 × 6 Gy + FAP-CD40 group supporting the idea that the combination may shift the balance in favor of CD8 T-cell infiltration in the tumor. In this context, although the frequencies of CD4 T cells, DCs, and natural killer (NK) cells decreased with the different treatments compared with the untreated mice (Supplementary Fig. S5B), the numbers per mg of tumor were comparable among all of the groups (Fig. 4A).

We observed an increase in the production of IFNγ by CD8 and CD4 TILs from 2 × 6 Gy + FAP-CD40–treated tumors, in ex vivo restimulation assays against mEERL95 cells (Fig. 4B and C). TNFα and Granzyme B were also augmented in both CD8 and CD4 T-cell subsets (Fig. 4B and C) indicating an effector and cytotoxic phenotype (30). In contrast, we did not observe cytokine production on T lymphocytes from the cervical lymph nodes (not shown). The proliferation of CD8 TILs, measured as Ki67-positive cells, was increased in all treated groups compared with the controls, whereas the expression of PD-1 remained lower in a FAP-CD40–dependent manner suggesting a less exhausted phenotype (Fig. 4D). The combination therapy promoted increased proportions of memory phenotype (CD62L⁺CD44⁺) CD8 T cells from the cervical lymph nodes (Fig. 4E).

The combination therapy significantly amplified the molecular remodeling of the TME by upregulating 399 and downregulating 39 immune-related genes as compared with untreated tumors (Supplementary Fig. S5C). This molecular imprint triggered by the combination therapy encompassed an induction in the expression of different pathways involved in adaptive immune response, antigen processing, IFN signaling, inflammation, chemokines, and cytokines or DC functions (Supplementary Fig. S5D). In this regard, we found a significant enrichment in a DC-specific inflammatory gene signature (31) in tumors that underwent 2 × 6 Gy + FAP-CD40 treatment as compared with the monotherapies that include Cd40, Ccl22, Il12b, Ccr7, Irf8, and Cd86 upregulation (Fig. 4F). Consistent with this, we also found an induction of the maturation markers CD80 and CD86 on the surface of DCs from mEERL95 infiltrate upon 2 × 6 Gy + FAP-CD40 (Fig. 4G). By multiplex cytokine/chemokine assay, we confirmed the upregulation of intratumoral levels of IFNγ and certain chemokines such as CXCL9 and CXCL10 that were triggered by the combination (Fig. 4H).

By immunofluorescence staining, we observed that the FAP area did not change in the TME after all treatments but there was a tendency to diminish following FAP-CD40 and 2 × 6 Gy + FAP-CD40 (Supplementary Fig. S6A and S6B) suggesting a reduction in protumoral stromal cells (32). Remarkably, we found an increase of CD11c⁺ cells in the Fap-positive (FAP⁺) compartment upon radiotherapy (also upon the combination but less prominent), suggesting that irradiation might be promoting the effect of the bi-Ab favoring the proximity of both target cell types (Supplementary Fig. S6A and S6B). Interestingly, we observed that CD8 T cells (measured by CD8⁺ area) were distributed within FAP⁺ cells in the TME upon 2 × 6 Gy + FAP-CD40 treatment (Supplementary Fig. S6C–S6E).

Altogether, these data demonstrate that the remarkable efficacy of the 2 × 6 Gy + FAP-CD40 therapy is associated with major remodeling of the tumors' immune landscape including reduced immunosuppression, increased CD8 T-cell infiltration and proliferation, cytokine production, and maturation of DCs.

By in vivo selective depletion, we examined the contribution of T lymphocytes to the antitumor effect of combination therapy. mEERL95 tumor-bearing mice received intraperitoneal injections of either anti-CD8β, anti-CD4, or the respective IgG control antibody every 3 days starting 48 hours before the therapeutic treatment. The depletion treatment of mice with CD8 T lymphocyte–targeted antibodies resulted in an abrogation of the curative effect observed upon the combination therapy (Fig. 5A and B). Mice depleted of CD4 T cells showed CRs by day 30 posttreatment (Fig. 5A). However, 5 of 7 (71%) mice from this group underwent tumor relapses as compared with the 2 of 7 (28%) mice from the IgG control group indicating reduced survival (Fig. 5A and B). These results highlight an essential role of CD4 T cells in the formation of durable antitumor responses.

The enrichment of DCs maturation transcriptional signature in the tumors upon combination therapy (Fig. 4F) suggested a role of activated Batf3-dependent DCs (33) and the necessity for CD8 T cells-cDC1 cross-talk in CD8 T cell–mediated responses. We therefore sought to test the combination therapy in the absence of such an immune population. As Fig. 5C shows, the therapeutic effect of the 2 × 6 Gy + FAP-CD40 treatment was abolished in Batf3^−/− mice (which lack cDC1; ref. 34). Although tumor growth kinetics decreased in some treated Batf3^−/− mice as compared with nontreated tumor-bearing Batf3^−/− mice, the antitumor effect of combination therapy was abrogated as opposed to treated tumor-bearing wild-type mice (Fig. 5C), indicating the requirement of cross-priming DCs for the FAP-CD40–mediated robust protective response.

Given that Il12b, which is known to be expressed by cross-priming DCs (31), is one of the top tumor upregulated genes upon the combination therapy, we sought to examine its functional relevance in the 2 × 6 Gy + FAP-CD40 therapeutic response. Thus, we administered an IL12 blocking antibody for 7 days starting one day before radiotherapy. As Fig. 5D shows, in all mice tumors regressed upon combination therapy and IL12 neutralization, but tumors resumed growth in 85% of mice (6/7). This supports a key role of IL12 in attaining the full strength of the antitumor effect mediated by combination therapy. Interestingly, the blocking of IL12 did not affect the response at early stages when antitumor activity is likely mainly due to radiotherapy. The loss of the therapeutic efficacy upon IL12 neutralization was accompanied by a reduction in the serum levels of certain chemokines produced by DCs such as CXCL9, CXCL10, CCL4, and CCL22 (Fig. 5E), which are associated with T-cell recruitment, migration, and priming (35–38).

In this study, we report a novel safe and efficacious strategy to enhance radiotherapy in a murine model of HPV⁺-HNSCC and achieve durable protective immune memory by adding a bispecific FAP-CD40 antibody to local radiotherapy. The key component of this combination is based on limiting the activity of a CD40 agonist to the TME through the targeting of FAP-expressing stromal cells. We identified a relevant treatment regimen and showed that the CD8 T cell-cDC1 axis drives the main cellular mechanism of action upon IL12 induction at the tumor bed (Supplementary Fig. S7). Together, the results provide a strong biological rationale for the translation of this therapeutic approach to the clinic. Moreover, the remarkable effects, mediated by human CD40 signaling, upon the combination of radiotherapy with a bispecific FAP-huCD40 antibody, supports the clinical applicability of this approach to treatment of not only patients with head and neck cancer, but also other cancer types characterized by stromal accumulation of FAP-expressing fibroblasts (39).

The FAP-CD40 agonist, whose structure and pharmacologic properties are reported in the accompanying article (21), only induces CD40 costimulation when cross-linked to FAP⁺ cells. Such an intrinsic characteristic of this bi-Ab, to target the CD40 therapeutic antibody to tumor areas where FAP- and CD40-expressing cells colocalize, is the key to avoiding the classical pattern of systemic toxicity described with the use of CD40 agonist (17, 19). FAP is an endopeptidase mainly expressed by activated fibroblasts from the tumor stroma (27), but also other stromal compartments, such as FRC from lymph nodes, demonstrate expression of FAP (40). Its targeting has been used in some preclinical studies with heterogeneous results (41, 42). Here, by treating the TC-1 tumor model in which minimal efficacy of the combination therapy is consistent with low intratumoral FAP expression, we found that FAP-mediated cross-linking at the tumor stroma was essential for the antitumor effects of the FAP-CD40 antibody. Although, the FAP-CD40 was detected in the tumor area by in vivo fluorescence imaging, we could not formally exclude its presence in the neighboring lymph nodes. However, the equivalent presence of FAP⁺ cells in the cervical lymph nodes of TC-1 and mEERL95 tumors along with their opposite response to the therapy provide indirect evidence that RT+FAP-CD40–mediated antitumor immune activation occurs in the tumor bed. It would be interesting to see how FAP-CD40 performs in immunologically “cold” tumors. In addition, it has been reported that the CD40 agonist–mediated antitumor effect is mainly driven at the tumor site, while the contribution of the draining lymph nodes remains controversial (43). Given the heterogeneity in the expression of FAP in human solid tumors (39, 44), further studies are required to elucidate which levels of expression are sufficient for cross-linking of the antibody and consequently, for achieving an efficient antitumor response in patients.

The magnitude of the efficacy reported in mEERL95 tumors and the rarity of relapses among responder mice to the combination therapy as compared with the tumor escape observed upon anti-FAP-CD40 monotherapy deserve further investigations at the biological level. Therefore, we are currently addressing the molecular and cellular events involved in such resistance by combining RNA sequencing and cellular barcoding. The long-term survival of the animals treated with RT+Fap-CD40 combination and the rejection of the homologous tumor in rechallenged cured mice supported an impact of the combination on long-term immunologic memory, which is consistent with the induction of a T-cell memory phenotype found in the local lymph nodes. While identification of specific antigens was beyond the scope of the study, we ruled out an exclusive and dominant role of E7-derived antigens.

The abrogation of the effects of the RT+FAP-CD40 combination in T cell–depleted mice confirmed that such population was crucial for the therapeutic outcome. While CD8 T cells were required for the antitumor effect of the RT+FAP-CD40, CD4 cells proved to be essential in the generation of durable responses and prevention of relapses followed by a robust immunologic CD8 T memory, as it has been demonstrated in vaccine and viral infection settings (45). Such findings are also supported by previous preclinical studies in which CD4 T cell–mediated immune responses are crucial in anti-CD40 therapy (11).

The therapeutic effect of the combination was accompanied by a profound remodeling of the tumor immune landscape. The decrease of Tregs, together with the reduction of FAP-expressing CAFs, which are known to be the main source of TGFβ in the TME, indicated that the combination prompted a less immunosuppressive context favoring the function of effector T lymphocytes. In this regard, we found an increase in the IFNγ intratumoral levels and low expression levels of PD-1 on restimulated TILs from RT+FAP-CD40–treated tumors. Indeed, IFNγ, which has been postulated as a key factor in the CD40-mediated antitumor responses in combination with IL12 (43), was induced upon the combination in a FAP-CD40–specific manner. Interestingly, IL12 neutralization of the response abrogated the therapeutic effect of RT+FAP-CD40 indicating that FAP-CD40 relies on such a cytokine for efficacy. Activated cross-priming DCs are known to be the main producers of IL12 in humans and mice (31, 33). In this sense, two noteworthy findings reveal that these cells are crucial in the mechanism of action of the combination. First, the enrichment in an activated cross-priming DC-specific gene signature upon RT+FAP-CD40 in the TME, and second, the loss of the antitumor response in Batf3^−/− mice, which lack cDC1 cells. Although it has recently been described that Batf3 deficiency plays a CD8 T-cell intrinsic role in the generation of memory cells upon viral infection, this might need to be closely examined in our tumor model, where the antitumor response measured for a period of around 25 days may be independent of Batf3 deficiency in T cells (34). Altogether, our observations are in agreement with previous studies that reported a dependency of anti-CD40 therapies on cross-priming DCs (31, 46). Despite the fact that macrophages have been described to play a role in the CD40-mediated responses (47), the decrease in their numbers and the key role of IL12 suggested that the radioimmunotherapeutic effect in our setting is mainly mediated by DCs. However, further analyses are required to fully elucidate their contribution in this context.

We believe that the remarkable results that we have obtained in terms of safety and antitumor response warrant the translation of this novel therapeutic approach to the clinic, which combines a gold standard-of-care, radiotherapy, and a novel and safe molecularly engineered CD40-targeted immunotherapy.

S. Labiano reports grants from Roche Glycart during the conduct of the study; in addition, S. Labiano has a patent for no. EP20207768 pending to F. Hoffmann-La Roche AG. V. Roh reports grants from Roche Glycart during the conduct of the study. C. Godfroid reports grants from Roche Glycart during the conduct of the study. A. Hiou-Feige reports grants from Roche Glycart during the conduct of the study. J. Romero reports grants from Roche Glycart during the conduct of the study. E. Sum reports a patent for WO 2018/185045 A1 pending to F. Hoffmann-La Roche AG, and has ownership of Roche stock (options). M. Rapp reports a patent for WO 2018/185045 A1 pending to F. Hoffmann-La Roche AG and a patent for WO 2020/070041 A1 pending to F. Hoffmann-La Roche AG; M. Rapp also has ownership of Roche stock (options). G. Boivin reports grants from Roche Glycart during the conduct of the study. C. Simon reports grants from Roche Glycart during the conduct of the study. C. Simon also reports grants from Roche and Intuitive, and personal fees from Pfizer, Merck, MSD, and Seattle Genetics outside the submitted work. P. Umaña reports a patent for WO 2018/185045 A1 pending to F. Hoffmann-La Roche AG, a patent for WO 2020/070041 A1 pending to F. Hoffmann-La Roche AG, and a patent for EP20207768.1 pending to F. Hoffmann-La Roche AG. P. Umaña also has ownership of Roche shares. C. Trumpfheller reports a patent for EP20207768.1 pending to F. Hoffmann-La Roche AG, a patent for WO 2018/185045 A1 pending to F. Hoffmann-La Roche AG, and a patent for WO 2020/070041 A1 pending to F. Hoffmann-La Roche AG. C. Trumpfheller also has ownership of Roche stock (options). G.V. Tolstonog reports grants from Roche Glycart during the conduct of the study; in addition, G.V. Tolstonog has a patent for European patent application no. EP20207768 filed on November 16, 2020 pending. M.-C. Vozenin reports grants from Roche during the conduct of the study; in addition, M.-C. Vozenin has a patent for European patent application no. EP20207768 filed on November 16, 2020 pending. P. Romero reports grants from Roche pRED during the conduct of the study, as well as personal fees from MaxiVax outside the submitted work; in addition, P. Romero has a patent for AD2838 EP BS pending and a patent for AC1689 PCT BS pending. P. Romero is editor-in-chief of Journal for Immunotherapy of Cancer. No disclosures were reported by the other authors.

S. Labiano: Conceptualization, data curation, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. V. Roh: Conceptualization, data curation, formal analysis, methodology. C. Godfroid: Data curation, validation, investigation, methodology. A. Hiou-Feige: Conceptualization, data curation, investigation, methodology. J. Romero: Conceptualization, data curation, formal analysis, investigation, methodology. E. Sum: Resources, methodology. M. Rapp: Resources, methodology. G. Boivin: Methodology. T. Wyss: Software, formal analysis. C. Simon: Conceptualization, funding acquisition. J. Bourhis: Conceptualization, funding acquisition. P. Umaña: Conceptualization, resources. C. Trumpfheller: Conceptualization, resources, supervision, writing–original draft, project administration, writing–review and editing. G.V. Tolstonog: Conceptualization, resources, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing. M.-C. Vozenin: Conceptualization, resources, supervision, writing–original draft, project administration, writing–review and editing. P. Romero: Conceptualization, resources, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

The authors would like to thank the Mouse Pathology Facility (Janine Horlbeck and Jean-Christophe Stehle), Cellular Imaging Facility (Florence Morgenthaler), Lausanne Genomic Technologies Facility (Corinne Peter), FACS Facility (Romain Bedel), and Animal Facility for the excellent assistance. We thank Yan Monnier for providing clinical tumor samples, Benoit Petit for technical support with the Xrad-225CX-PXi Instrument, and Alena Donda for fruitful discussions.

This work was supported by the Roche pRED (Pharma Research and Early Development) of Roche. P Romero is supported by grants from Swiss Cancer League KFS-4404-02-2018) and the Swiss National Science Foundation (310030_182735).

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.