Background

Over the last decade, CD3 bispecific antibody (CD3 bsAb) or “T-cell engager” therapy has become an established treatment modality for the treatment of certain B-cell malignancies. Treatment with CD3 bsAbs achieved its breakthrough in the clinic with blinatumomab (CD3xCD19), which demonstrated response rates up to 90% in relapsed or refractory acute lymphoblastic B-cell leukemia.1 Following this initial success, six other CD3 bsAbs targeting B-cell antigens, one targeting delta-like ligand 3 for small cell lung cancer and one CD3 bispecific molecule targeting a peptide/major histocompatibility complex expressed in uveal melanoma have been Food and Drug Administration (FDA)-approved.2–8 As already suggested by the difference in the number of FDA-approved bsAbs, clinical development of the CD3 bsAb is lagging behind in solid tumor indications, where safety and efficacy are major obstacles.9

BsAbs act by crosslinking a tumor-associated antigen expressed on the surface of tumor cells to CD3ε on the surface of T cells, resulting in the formation of an immunological synapse, followed by T-cell activation and tumor cell lysis.9 As binding to CD3ε is independent of T-cell specificity, an important advantage of CD3 bsAbs is the ability to engage all T cells in the tumor microenvironment (TME). Although these therapeutics activate intratumoral T cells, it is uncertain if this therapy invigorates tumor-specific T cells or leads to the induction of de novo antitumor T cells via antigen spread. According to the cancer-immunity cycle paradigm, initial cancer cell kill might lead to the presentation of tumor antigens by professional antigen-presenting cells (APCs) followed by ignition of tumor-specific T-cell responses.10 The presence of systemic tumor-specific T cells is essential to fend off recurrences of tumor cells or (dormant) metastases that were not destroyed during the therapeutic intervention. Additionally, the induction of such responses could help to overcome tumor heterogeneity or resistance due to loss of surface expression of the targeted epitope, which are frequently reported causes for relapse after CD3 bsAb therapy in the clinic for hematological malignancies,11–16 as will likely also be the case for solid tumors once CD3 bsAb therapy for these cancer indications is successfully applied in the clinic.

Thus far, most studies involving CD3 bsAbs focus on specificity and therapeutic efficacy, but little is known about the induction, functionality and persistence of tumor-specific memory responses. To our knowledge, only one clinical study has described the induction of tumor-specific T-cell responses after catumaxomab (CD3xEpCAM) therapy, which was mediated via engagement of Fc-gamma receptors (FcγRs) on myeloid cells by its active Fc tail.17 Since antigen loss is described as the major escape mechanism for CD3 bsAb treatment of hematological tumors in the clinic, this suggests that insufficient tumor-specific responses were raised to eradicate antigen-negative tumor cells and prevent relapse. In murine tumor models, contradicting results have been reported. One study showed induction of systemic tumor-specific responses following CD3 bsAb therapy in a syngeneic mouse model overexpressing a human tumor antigen, enabling rejection of a second tumor challenge in the absence of further treatment.18 In contrast, we previously showed that protective memory was not installed after CD3 bsAb treatment in the fully syngeneic B16F10 murine tumor model targeting the endogenous tyrosinase-related protein 1 (TRP1) antigen.19 These data suggest that the immunogenicity of the involved tumor antigen or the degree of spontaneous T-cell infiltration in the TME may determine the outcome. This warrants a thorough investigation on the induction of protective tumor-specific immunity during CD3 bsAb therapy in the context of solid tumors.

In this manuscript, we examine multiple combination strategies to improve the therapeutic efficacy against the primary tumor and concomitantly augment the formation of tumor-specific memory responses that protect against rechallenge. We test combinations of CD3xTRP1 with tumor-localized costimulation, Fc-active tumor-targeted monospecific antibodies and cancer vaccination approaches in the immunologically “cold” B16F10 melanoma model and in the immunologically “hot” MC38 colon carcinoma model. Collectively, our findings advocate for a combination of tumor-specific vaccination with CD3 bsAb therapy to improve therapeutic efficacy against primary tumors and to install long-term protective memory in immunologically “hot” as well as “cold” tumors.

ResultsCD3xTRP1 bsAb therapy induces weak and short-lived systemic CD8 T-cell immunity

First, mice harboring immunologically “cold” syngeneic B16F10 tumors, endogenously expressing the tumor antigen TRP1, were treated with the Fc-inert mouse CD3xTRP1 bsAb at days 6 and 9 post-tumor challenge (figure 1A). In concordance with our previous results,19 this early treatment delayed tumor outgrowth in all mice and completely cleared tumors in approximately 40% of the mice (figure 1B and online supplemental S1A). However, no protective immunity was installed after this treatment, as rechallenge of the complete responder mice with a second injection with B16F10 tumor cells at the contralateral flank at day 80 did not result in delayed tumor outgrowth compared with age-matched naïve control mice (figure 1A and C and online supplemental S1A). Analysis of T cells in peripheral blood revealed that CD3 bsAb treatment increased central memory (CD44+/CD62L+) CD8, but not CD4, T-cell frequencies at the expense of their naïve (CD44−/CD62L+) counterparts, suggesting therapy-induced T-cell differentiation (figure 1D). Next, we examined the presence of tumor-specific CD8 T-cell responses in peripheral blood by measuring ex vivo cytokine production after a brief co-culture with B16F10 tumor cells. Tumor necrosis factor alpha (TNFα), but not interferon-gamma (IFNγ) production, by CD8 T cells was specifically observed in the blood of CD3xTRP1-treated mice 12 days after treatment initiation (figure 1E), thereby illustrating induction of systemic CD8 T-cell immunity. The failure to produce IFNγ suggested a low activation status of these tumor-specific T cells20 and these responses waned before day 39 (figure 1E), corroborating the lack of long-term protective memory.

Figure 1

CD3xTRP1 monotherapy induces tumor-specific responses, but does not protect from rechallenge. (A) Treatment schedule for mice inoculated with B16F10 tumors and treated with CD3xTRP1 bsAb. (B–C) Kaplan-Meier survival graphs for primary tumor challenge (B) or rechallenge (C). Numbers indicate surviving mice. Age-matched naïve mice were used as untreated controls for rechallenge. (D) Differentiation of T cells in blood 9 days after the final CD3xTRP1 administration. (E) Tumor-specific responses were measured by determining cytokine production in CD8 T cells from blood taken on days 18 or 39 co-cultured with medium or IFNγ prestimulated B16F10 tumor cells. Data represented as mean±SEM for n=10–24 (D) or n=2–11 (E). Statistics were calculated using Mantel-Cox log-rank tests (B–C), unpaired two-sided t-tests to compare treatments (D–E), or paired two-sided t-tests to compare different stimulations within the same treatment (E). bsAb, bispecific antibody; IFNγ, interferon-gamma; Tcm, central memory T cell; Tem, effector memory T cell; Tn, naïve T cell; TNFα, tumor necrosis factor alpha; TRP1, tyrosinase-related protein 1.

As we did observe some tumor-specific T-cell responses after CD3xTRP1 monotherapy, we aimed to increase the persistence and quality of these T cells. Previous reports indicated that the combination of CD3 bsAb with 4-1BB costimulation enhanced these traits, resulting in superior antitumor activity.18 To this end, a murine Fc-inert 4-1BBxTRP1 bsAb was coadministered together with CD3xTRP1, leading to targeted costimulation at the tumor site (online supplemental figure S1B), as agonistic 4-1BB monospecific antibodies come with the risk of adverse events.21 While this combination treatment improved antitumor activity against the primary tumor compared with CD3xTRP1 monotherapy, the complete responder mice did not display protection against rechallenge compared with age-matched naïve mice (online supplemental figure S1C-G). Together, these results imply that other combinations should be pursued to induce protective systemic immunity in immunologically “cold” tumors.

Combination with a tumor-opsonizing antibody improves primary survival and activates intratumoral APCs, but fails to install protective memory

Next, we examined the induction of tumor-specific T-cell responses by combining CD3 bsAb with tumor-opsonizing antibodies, reasoning that these may promote antigen uptake and cross-presentation by engaging APCs.22 23 To increase the priming of tumor-specific T cells, we first injected the Fc-inert CD3xTRP1, resulting in the lysis of tumor cells and release of tumor debris, followed by administration of the monospecific Fc-active TA99 antibody targeting the same TRP1 epitope, which could engage FcγRs on APCs (figure 2A). This combination treatment of CD3xTRP1 with TA99 enhanced the survival compared with CD3xTRP1 monotherapy, while TA99 alone had no impact on tumor outgrowth (figure 2B and online supplemental figure S2A-D). Surprisingly, a similar tendency for survival benefit was observed when CD3xTRP1 was combined with an Fc-silenced TA99 antibody, indicating that at least part of this antitumor effect was independent of FcγR engagement. Spectral flow cytometry profiling of the TME revealed no differences in the immune cell frequencies for both combinations of bsAb with TA99 compared with CD3xTRP1 monotherapy (online supplemental figure S3A-B and S4A-B). However, phenotypic analysis demonstrated that tumors treated with a combination of CD3 bsAb with Fc-active TA99 contained higher frequencies of CD86-positive macrophages, conventional dendritic cell 1 (cDC1s) and cDC2s, and higher expression of inducible nitric oxide synthase (iNOS) on macrophages, in addition to minor changes in the activation status of Natural Killer cells (figure 2C–D and online supplemental figure S4C-G). These data pointed to the engagement of FcγRs on innate immune cells by the Fc-active TA99 antibody, but did not explain the therapeutic benefit of Fc-inactive TA99 in combination with bsAb. As previous studies have reported a role for TRP1 in tumor cell proliferation,24 25 in addition to its commonly known role in melanin production, we decided to investigate the potential direct effects of TA99 on tumor cells. Indeed, significant inhibition of B16F10 tumor cell growth was detected in vitro in the presence of monospecific TA99 antibodies, irrespective of their capability to bind FcγRs (online supplemental figure S4H). The effect of the bispecific CD3xTRP1 on tumor growth was much smaller, most likely due to the difference in avidity. These results suggested that monospecific (Fc-silenced) TA99 improved the efficacy of CD3 bsAb treatment through direct impairment of tumor cell growth. Furthermore, the TA99 antibody with an active Fc tail enhanced the engagement of the myeloid cell compartment, which might aid the development of tumor-specific T-cell responses.

Figure 2

Combination of CD3xTRP1 with Fc-active TA99 augments primary survival but does not improve tumor-specific T-cell responses. (A) Treatment schedule for mice inoculated with B16F10 tumor cells and treated with a combination of CD3xTRP1 with TA99. (B) Kaplan-Meier survival graphs for indicated groups, numbers indicate surviving mice. (C–D) Expression of phenotypical markers on intratumoral macrophages (C), or cDC1s and cDC2s (D) treated according to the treatment scheme in online supplemental figure S4A. Numbers indicate average expression. Statistical significance is only shown for comparisons against the bsAb treatment group. (E) Tumor-specific responses on days 25 and 39 were measured by determining cytokine production in CD8 T cells from blood co-cultured with medium or IFNγ prestimulated B16F10 tumor cells. (F) Kaplan-Meier survival graphs for indicated groups, numbers indicate surviving mice. Data represented as mean on a heatmap for n=4–6 (C–D, two untreated mice were excluded from the analysis based on tumor size<27 mm3), or mean±SEM for n=3–14 (E). Statistics were calculated using Mantel-Cox log-rank tests (B and F), one-way ANOVA followed by Tukey’s post hoc tests to compare all treatments (C–E), or paired two-sided t-tests to compare different stimulations within the same treatment (E). ANOVA, analysis of variance; bsAb, bispecific antibody; cDC1s, conventional dendritic cells 1; cDC2s, conventional dendritic cells 2; IFNγ, interferon-gamma; iNOS, inducible nitric oxide synthase; TNFα, tumor necrosis factor alpha; TRP1, tyrosinase-related protein 1.

We then investigated tumor-specific T-cell responses in the blood of these mice. Lower levels of TNFα and IFNγ were observed in CD8 T cells at day 25 for either of the combination therapies compared with CD3 bsAb monotherapy and again these responses were not detectable anymore on day 39 (figure 2E). When mice that rejected the primary tumor were rechallenged at day 80, no efficient protection against the second B16F10 tumors was observed, with the exception of a marginal delay in tumor outgrowth for the combination of CD3xTRP1 and (Fc-active) TA99 (figure 2F and online supplemental S2A and C). Altogether, these results demonstrate that a monospecific Fc-active tumor-opsonizing antibody could improve the activity of CD3 bsAb therapy against the primary tumor and enhance the activation status of intratumoral APCs, but fails to install long-term protective memory.

Tumor-non-specific vaccines promote therapy response to primary but not secondary tumors

We recently reported that therapeutic synthetic long peptide (SLP) vaccines, containing tumor-unrelated CD8 T-cell antigens and adjuvants, facilitated the accumulation of intratumoral T cells, promoted a proinflammatory TH1-like TME and empowered the therapeutic activity of CD3 bsAbs in solid tumors.26 We wondered whether this combination strategy would also induce tumor-specific T-cell responses and protection from rechallenge. Since multiple studies have demonstrated the importance of CD4 T-cell help in memory formation of T-cell responses,27 28 a CD4 epitope was included in the SLP vaccine in addition to the CD8 epitope. Tumor-non-specific ovalbumin (OVA) SLP vaccines were administered together with TLR9-ligand CpG (Cytosin-phosphatidyc-Guanin) as an adjuvant, followed by CD3xTRP1 treatment. Administration of CD3xTRP1 bsAb was applied on days 12 and 15 (contrasting earlier experiments where days 6 and 9 were used) to first allow vaccine-induced T-cell infiltration into the tumor (figure 3A).26

Figure 3

Combination of CD3xTRP1 with tumor-non-specific vaccination improves primary survival and tumor-specific responses, but does not install protective memory. (A) Treatment schedule for mice inoculated with B16F10 tumor cells and treated with a combination of CD3xTRP1 and SLP vaccination containing OVA CD8 and/or CD4 epitopes and CpG as adjuvant. (B) Kaplan-Meier survival graphs for indicated groups, numbers indicate surviving mice. (C) Frequency of OVA Tm+ CD8 T cells in blood on days 25 and 45. (D) Tumor-specific responses were measured by determining cytokine production in CD8 T cells from blood co-cultured with medium or IFNγ prestimulated B16F10 tumor cells on day 25. (E) Kaplan-Meier survival graphs for indicated groups, numbers indicate surviving mice. Data represented as mean±SEM for n=1–14 (C) or n=4–14 (D). Statistics were calculated using Mantel-Cox log-rank tests (B and E), one-way ANOVA followed by Tukey’s post hoc tests comparing all groups (C), or comparing all treatments for CD8 T cells stimulated with B16F10 tumor cells (D), or paired two-sided t-tests to compare different stimulations within the same treatment (D). ANOVA, analysis of variance; bsAb, bispecific antibody; CpG, Cytosin-phosphatidyc-Guanin; IFNγ, interferon-gamma; OVA, ovalbumin; SLP, synthetic long peptide; TNFα, tumor necrosis factor alpha; TRP1, tyrosinase-related protein 1.

In line with our previous results,26 OVA CD8 SLP vaccination enhanced the therapeutic efficacy of CD3xTRP1 (figure 3B and online supplemental S5A). Also, the combination of CD3xTRP1 with OVA CD4 or OVA CD8+CD4 SLP vaccine improved antitumor activity, with the latter combination yielding the highest fraction of surviving mice. At day 25, OVA-specific tetramer-positive (Tm+) CD8 T cells were detected in the blood of all treatment groups that received CD8 SLP vaccination (figure 3C). Interestingly, these tumor-unrelated OVA-specific CD8 T cells waned by day 45 in the vaccinated only group, whereas mice receiving combinations of CD8 SLP vaccination with CD3 bsAb still harbored detectable T-cell frequencies, suggesting that the CD3 bsAb promoted T-cell longevity. In all CD8 SLP vaccinated mice, OVA Tm+ CD8 T cells displayed markers indicative of an effector memory phenotype (CD44+/CD62L−) (online supplemental figure S5B). Importantly, higher B16F10-specific responses of peripheral CD8 T cells were observed at day 25 in all groups receiving combination treatment of vaccination with CD3 bsAb compared with single component controls (figure 3D). These findings implied that the vaccination, as well as the CD3 bsAb component, is essential for strong and systemic induction of tumor-specific CD8 T cells, most likely due to tumor cell kill in the presence of immune-stimulating vaccine adjuvant, as we previously unraveled.26 However, tumor-specific T-cell responses were hardly detectable at day 45 in blood samples (online supplemental figure S5C). Accordingly, no protective immunity was observed when the surviving complete responder mice were rechallenged at day 80 (figure 3E and online supplemental S5A).

We concluded that, despite clear induction of tumor-specific T cells, our previously defined combination strategy of T-cell stimulating tumor-non-specific vaccines with CD3 bsAb therapy failed to install protective memory in the immunologically “cold” B16F10 model.

Combination of CD3 bsAb with tumor-non-specific vaccines installs protective memory in the immunologically “hot” MC38 tumor model

As the B16F10 tumor model is notoriously known for its “cold” TME,29 reflected by the low activity of immunotherapeutic treatments such as checkpoint inhibition, we wondered if protective memory would be induced in a more immunogenic tumor model. We selected the MC38 colon carcinoma model, as this is known for its high mutational load and susceptibility to immunotherapeutic interventions. The MC38 cell line was transfected with the TRP1 tumor antigen (MC38.TRP1) to allow targeting by the CD3xTRP1 bsAb. MC38 tumors indeed displayed a denser spontaneous immune cell infiltrate in vivo than the B16F10 tumors, including higher frequencies of CD8 T cells displaying an activated phenotype (online supplemental figure S6). MC38.TRP1 tumors had a similar immune cell infiltrate as wild-type (WT) MC38 tumors.

BsAb monotherapy and combinations with a vaccine comprizing the tumor-non-specific OVA CD8 SLP or the tumor-specific Rpl18 antigen (ribosomal protein L18) CD8 SLP, a strong endogenous neoantigen in this tumor model,30 were tested in the MC38.TRP1 model (figure 4A). CD3xTRP1 monotherapy only slightly delayed tumor outgrowth, whereas combinations with tumor-non-specific OVA or tumor-specific Rpl18 vaccination controlled nearly all tumors (figure 4B and online supplemental S7A-B),26 demonstrating the sensitivity of MC38.TRP1 tumors to this combination therapy compared with B16F10. Interestingly, all outgrowing tumors from CD3 bsAb monotherapy-treated mice lost TRP1 expression, in contrast to those from untreated mice (figure 4C–D and online supplemental S7C), illustrating selective immune pressure by the CD3 bsAb. Importantly, these results implied that the addition of a vaccine prevented escape through antigen loss and effectively eradicated residual TRP1-negative tumor cells. Strikingly, the combination of CD3xTRP1 with the vaccine adjuvant alone (TLR9 agonist CpG) demonstrated a similar survival benefit compared with the combinations with OVA or Rpl18 SLP (figure 4B and online supplemental S7B), suggesting that peptide antigen was redundant in this model. Of note, the adjuvant CpG was injected at the vaccination site at the tail base, and not intratumorally, but was still able to boost the antitumor activity of the tumor-targeting CD3 bsAb.

Figure 4

In the immunologically “hot” MC38.TRP1 model long-term systemic immunity in installed by CD3 bsAb in combination with a vaccine adjuvant. (A) Treatment schedule for mice inoculated with MC38.TRP1 tumor cells and treated with a combination of CD3xTRP1 and vaccination with tumor-non-specific OVA or tumor-specific Rpl18 CD8 SLP epitopes and CpG as an adjuvant. (B) Kaplan-Meier survival graphs for indicated groups, numbers indicate surviving mice. (C–D) TRP1 expression of end-stage MC38.TRP1 tumors are visualized as exemplary histograms (C) and bar graphs (D). (E–F) Analysis of Rpl18 tetramer CD8 T-cell responses and phenotype in the blood on day 25 (E) or day 39/42 (F). Phenotypical data is only shown for mice with>0.2% of Rpl18 Tm+ from CD8 T cells (E). The dotted line indicates the mean Rpl18 Tm+ frequency + two SD of the naïve mice. (G–H) Tumor-specific responses were measured by determining cytokine production in CD8 T cells from blood co-cultured with medium, or IFNγ prestimulated MC38 or MC38.TRP1 tumor cells on day 25 (G) or day 39/42 (H). (I–J) Kaplan-Meier survival graphs for mice receiving sequential rechallenges with MC38.TRP1 (I) and MC38 tumors (J). (K) Kaplan-Meier survival graphs for mice receiving a rechallenge with MC38 in the presence of CD8-depleting antibodies. Numbers indicate surviving mice. Data represented as mean±SEM for n=3–7 (D), n=7–23 (E, G), or n=3–24 (F, H). Statistics were calculated using Mantel-Cox log-rank tests (B and I–K), using one-way ANOVA followed by Tukey’s post hoc tests comparing all groups (D–F), comparing all treatments for cells stimulated with MC38.TRP1 tumor cells (G–H), or using repeated measures ANOVA followed by Dunnett’s post hoc tests compared with medium to compare different stimulations within the same treatment (G–H). ANOVA, analysis of variance; bsAb, bispecific antibody; CpG, Cytosin-phosphatidyc-Guanin; FMO, Fluorescence Minus One; IFNγ, interferon-gamma; MFI, Mean Fluorescence Intensity; OVA, ovalbumin; Rpl18, ribosomal protein L18; SLP, synthetic long peptide; TNFα, tumor necrosis factor alpha; TRP1, tyrosinase-related protein 1.

We then interrogated antitumor T-cell responses and first measured Rpl18 Tm+ T-cell frequencies in blood. As expected, Rpl18 Tm+ frequencies were high in mice after Rpl18 vaccination, while the other groups showed very low percentages which tended to be higher than in naïve mice (figure 4E–F). The Rpl18 Tm+ CD8 T cells from all treated mice demonstrated higher expression of proliferation marker ki67 compared with untreated mice (figure 4E and online supplemental S7D-F), suggesting the presence of the MC38.TRP1 tumor already induced low levels of Rpl18-specific CD8 T cells, while treatment promoted the proliferation of these cells. Analysis of immune cells from end-stage tumors confirmed this notion, as clearly detectable frequencies of Rpl18 Tm+ CD8 T cells were present in mice that did not receive Rpl18 vaccination (online supplemental figure S7G). Next, the antitumor activity of CD8 T cells in blood was assessed following co-culture with WT or TRP1-transfected MC38 tumor cells. On day 25, the strongest MC38-directed T-cell responses were found in mice treated with tumor-specific Rpl18 vaccination alone or together with CD3 bsAb (figure 4G and online supplemental S7H). As expected, T-cell responses to WT MC38 tumor cells were much stronger after (combination) treatment with the Rpl18 neoantigen vaccine than for CD3xTRP1 monotherapy or its combination with OVA vaccine, or adjuvant alone (figure 4G and online supplemental S7H). Moreover, CD3xTRP1 treatment stimulated the induction of T-cell responses towards the MC38.TRP1 tumor, which was most pronounced in combination with Rpl18 or OVA SLP vaccine. Finally, except for the Rpl18-vaccinated mice, WT MC38-specific and MC38.TRP1-specific T-cell responses waned around day 40 (figure 4H). These data demonstrated the induction of some Rpl18-specific CD8 T cells following MC38.TRP1 tumor inoculation, a strong and lasting tumor-directed response after Rpl18 vaccination and a waning response for all other treatments.

Finally, we rechallenged mice that rejected the primary tumor with MC38.TRP1 at the opposite flank and observed full protection in nearly all mice (figure 4I and online supplemental S7A). To test whether the protection was directed towards the transfected TRP1 antigen or endogenous MC38 antigens, we rechallenged the surviving mice again, but now with WT MC38 tumor cells. Again, almost all mice fended off this tumor rechallenge (figure 4J and online supplemental S7A), illustrating the successful induction of protective immunity towards endogenous MC38 antigens. Importantly, protection from tumor rechallenge was completely ablated on CD8 T-cell depletion, delineating the role of tumor-specific CD8 T-cell responses in protective memory (figure 4K and online supplemental S7I).

Together, these findings show that T-cell inducing vaccines, and even their adjuvants, improve the efficacy of CD3 bsAb therapy and simultaneously induce protective tumor-specific immune responses in the “hot” MC38 tumor model.

Tumor-specific vaccines with CD3 bsAb install protective memory in the “cold” B16F10 model

We next examined whether the combination of a tumor-specific vaccine with CD3xTRP1 antibody would result in protective memory in a “cold” tumor model. For that, we used B16F10 cells expressing OVA as a tumor-specific non-self antigen. No significant differences in the frequency or phenotype of intratumoral CD8 T cells were observed between untreated B16F10 and B16F10.OVA tumors, despite the introduction of the immunogenic OVA antigen (online supplemental figure S8A). To ensure high frequencies of tumor-specific T cells, we first included adoptively transferred transgenic OT-1 CD8 T cells (recognizing the OVA antigen), followed by vaccination with OVA CD8 SLP and CD3xTRP1 (online supplemental figure S8B). All mice receiving this triple combination therapy cleared the primary tumors (online supplemental figure S8C) and retained high frequencies of tumor-specific OT-1 T cells in the blood at day 79, indicative of long-term survival of these cells (online supplemental figure S8D). Importantly, all mice in the triple combination group were also protected from rechallenge with B16F10.OVA cells, whereas approximately 70% of the mice receiving OT-1 transfer and vaccination were protected (online supplemental figure S8E). This demonstrated that protective memory can be induced in the context of CD3 bsAb treatment in this melanoma model.

Next, OT-1 T-cell transfer was omitted from the treatment regimen (figure 5A) and the antitumor activity of the combination treatment of CD3 bsAb with OVA CD8 and/or CD4 SLP vaccination was tested in the B16F10.OVA tumor model. The combination of CD3 bsAb with OVA CD8+CD4 SLP vaccination displayed the strongest survival benefit, while the individual OVA CD8 and CD4 SLP vaccination also improved CD3 bsAb therapeutic efficacy (figure 5B and online supplemental figure S8F). Monotherapy vaccination with OVA CD8+CD4 SLP (but without CD3 bsAb) also delayed tumor growth, supporting the notion that OVA is a strong non-self tumor antigen. Analysis of blood samples on day 24 and day 39 revealed that the groups receiving OVA CD8 vaccination showed strong and lasting induction of OVA-specific Tm+ CD8 T cells with predominantly an effector memory phenotype (figure 5C and online supplemental figure S8G). When comparing the OVA Tm+ CD8 T cells in blood between the different experiments containing tumor-non-specific (B16F10) and tumor-specific (B16F10.OVA) vaccination, we found no differences at the early time point (day 24/25), suggesting that these responses were mainly vaccine-mediated (online supplemental figure S8G). However, at the late time point (day 39/45), mice bearing B16F10.OVA tumors displayed higher frequencies of OVA Tm+ CD8 T cells with a central memory phenotype (online supplemental figure S8G). Analysis of cytokine production by peripheral blood T cells co-cultured with tumor cells revealed strong and durable CD8 T-cell responses towards B16F10.OVA after OVA CD8 SLP vaccination, whereas responses towards WT B16F10 tumor cells were less pronounced and lost over time, in line with our previous results (figure 5D–E). Importantly, the combination of CD4+CD8 SLP vaccination with CD3 bsAb treatment induced long-term protection to rechallenge with B16F10.OVA tumor cells on day 80 (figure 5F and online supplemental S8F). Of note, the two complete responder mice after the OVA CD8 SLP+bsAb treatment both rejected the rechallenge as well, in contrast to the two complete responder mice in the CD4 SLP+bsAb group, speculating that the CD8 T cells are essential for protection against rechallenge (online supplemental figure S8F). When the mice that completely rejected the first rechallenge with B16F10.OVA was again rechallenged with WT B16F10 tumor cells, no protection was observed, indicating that despite the clearance of two OVA-positive tumors, no protective immune responses were induced against endogenous antigens (figure 5F and online supplemental figure S8F).

Figure 5

In the B16F10.OVA tumor model, combination of CD3xTRP1 with a tumor-specific vaccination improves primary survival and installs protective memory responses. (A) Treatment schedule for mice inoculated with B16F10.OVA tumor cells and treated with a combination of CD3xTRP1 and SLP vaccination with OVA CD8 and CD4 epitopes. (B) Kaplan-Meier survival graphs for indicated groups, numbers indicate surviving mice. (C) Frequency of OVA Tm+ CD8 T cells in blood on day 24. (D–E) Tumor-specific responses were measured by determining cytokine production in CD8 T cells from blood co-cultured with medium, or IFNγ prestimulated B16F10 or B16F10.OVA tumor cells on day 24 (D) or day 39 (E). (F) Individual tumor outgrowth curves and Kaplan-Meier survival graphs for tumor rechallenges of indicated groups. Arrows indicate tumor cell inoculations, numbers indicate surviving mice. Data represented as mean±SEM for n=6–14 (C–D), or n=1–13 (E). Statistics were calculated using Mantel-Cox log-rank tests (B, F), one-way ANOVA followed by Tukey’s post hoc tests comparing all groups (C), or comparing all treatments for cells stimulated with B16F10.OVA tumor cells (D–E), or repeated measures ANOVA followed by Dunnett’s post hoc tests compared with medium to compare different stimulations within the same treatment (D–E). ANOVA, analysis of variance; bsAb, bispecific antibody; IFNγ, interferon-gamma; n.s., not significant; OVA, ovalbumin; SLP, synthetic long peptide; TNFα, tumor necrosis factor alpha; TRP1, tyrosinase-related protein 1.

Then, we corroborated the finding that tumor-specific vaccination before CD3 bsAb therapy installs protective immunity by exploiting the endogenous tumor-associated Gp100 antigen instead of the immunogenic non-self OVA antigen in the WT B16F10 tumor model. We used a previously established protocol in our lab using a Gp100 CD8 SLP vaccine, adjuvanted with TLR7/8 agonist imiquimod and interleukin (IL)-2 (figure 6A).26 31 As expected, the addition of Gp100 vaccination to CD3xTRP1 therapy improved survival percentages (figure 6B and online supplemental figure S8H). This combination treatment also increased the frequency of cytokine-producing CD8 T cells in the blood after brief co-culture with B16F10 tumor cells on day 25, which was low but still detectable on day 39 (figure 6C). Importantly, rechallenging the complete responder mice with B16F10 tumor cells resulted in significantly delayed tumor outgrowth when compared with naïve control mice (figure 6D and online supplemental figure S8H).