Female C57BL/6 and non-obese diabetic/severe combined immunodeficiency (NOD/SCID) mice aged 8 weeks were purchased from CER Janvier (Le Genest Saint Isle, France) and maintained under pathogen-free conditions in the animal facility of the University of Tours.

N. caninum (NC1 strain) and T. gondii (RH strain) tachyzoites were grown as described previously.25 For mouse injections, tachyzoites were collected by centrifugation at 600 g for 10 min and washed in phosphate-buffered saline (PBS). Heat-killed tachyzoites were obtained by heating freshly isolated tachyzoites 30 min at 60°C.

The N. caninum total antigenic extract (TE) was prepared following the protocol described for T. gondii antigenic extract.26

The murine thymoma cell line EG7 (EL4-OVA: EL-4 thymoma cells transfected with chicken albumin cDNA) was obtained from ATCC (CRL-2113). Tumor cells were cultured in RPMI (Pan Biotech) with 10% of fetal calf serum (FCS) (Dutscher), 0.05 mM of 2-mercaptoethanol (Pan Biotech), 50 IU/mL of penicillin/streptomycin (Pan Biotech) and 0.4 mg/mL geneticin (G418, Pan Biotech). The WaGa (RRID:CVCL_E998) cell line was used for Merkel cell carcinoma (MCC) studies, and was cultured in RPMI containing 10% FCS, 100 U/mL penicillin, streptomycin and amphotericin B.

EG7 (5×10⁵) or WaGa (10⁷), associated with 10% Matrigel (10 mg/mL, Corning) cells were subcutaneously inoculated in the flank of mice. Mice were then injected intratumorally or subcutaneously in the opposite flank with isolated tachyzoites in PBS, at the doses and times indicated in the figures.

Tumor volume was determined as π/6×width×length×height. Blood sera were collected via cheek bleeds from animals at the indicated time points.

Spleens and lymph nodes were mechanically disrupted, to obtain cell suspensions. The cells were then either plated in RPMI (5% FCS, 2% Hepes, 1% glutamine, 1% sodium-pyruvate, 50 mL/mL β-mercapto-ethanol, 100 U/mL penicillin and streptomycin) for in vitro restimulation or resuspended in PBS–1% BSA–0.2mM EDTA for flow cytometry analysis.

Tumors were harvested, weighed and measured after dissection. A first sample was collected and stored at −80°C for further quantitative PCR (qPCR) analysis. Other samples were collected respectively in 4% formol or 4% paraformaldehyde for histological analysis and immunofluorescence on frozen section, respectively. The rest of the tumor was dissociated using the Tumor Dissociation Kit, mouse (Miltenyi Biotech) and gentleMACS Dissociator (Miltenyi Biotech). Cells were resuspended (PBS, 2 mM EDTA, 1% fetal bovine serum) for staining and flow cytometry analysis.

Ultra-LEAF purified antimouse CD4 (GK1.5), antimouse CD8 (53–6.7), antimouse NK1.1 (PK136), antimouse Ly-6G (1A8) from BioLegend and clodronate liposomes from Liposoma were used for cells depletion. Ultra-LEAF purified Rat IgG2a, Rat IgG2b, mouse IgG2a isotype control antibodies from BioLegend, control liposomes from Liposoma and PBS were used like controls. Antibody (30 µg), liposomes (50 µL) or PBS were injected intratumorally 1 day after N. caninum treatment and then every 2–3 days until the sacrifice of mice.

Antibody reagents were obtained from Miltenyi Biotech: REA clones of antimouse CD3-APC-Vio770 (REA606), CD4-Vioblue (REA 604), CD8a-PE-Vio770 (REA 601), Foxp3-Vio515 (REA 788), NKp46-APC (REA 815), CD11c-PE (REA 754), CD11b-APC-Vio770 (REA 592), Ly6C-Vioblue (REA 796), Ly6G-PE-Vio770 (REA 526). Foxp3 staining was performed using Foxp3 Staining Buffer Set (Miltenyi Biotech).

Flow cytometry analysis was performed using a Miltenyi 8-color MACSQuant, and data were analyzed using Flowlogic (Miltenyi Biotech).

Cytokine production in splenocyte culture supernatants and/or tumor dissociation supernatants were evaluated by ELISA with specific kits: IL-1β, IL-6, IL-12p40, IL-18 and transforming growth factor (TGF)-β (eBioscience), CCL2 and VEGF (Invitrogen), IL-15 and PDL1 (R&D systems). For GM-CSF, interferon (IFN)-γ, IL-2, IL-4, IL-5, IL-10, IL-12p70, IL-17A, IL-23 and tumor necrosis factor (TNF)-α, MACSPlex cytokine 10 kit (Miltenyi Biotec) was used according to the manufacturer’s instructions.

All experiments were repeated at least two times with similar results between experiments. Statistical analysis was performed using GraphPad Prism software. P values <0.05 were considered to be statistically significant. Error bars show the SEM. Bar graphs were analyzed by one-way analysis of variance and scatter plots of tumor volume and weight were compared using Kruskal-Wallis non-parametric test.

More method details are provided in the online supplementary material and methods file.

Since humans are not natural hosts of N. caninum and the cross-talk between the protozoa and human innate immune cells has not yet been shown, it is important to study in vitro the relationship between Neospora and such cells in terms of infection and activation. Furthermore, those interactions should be also evaluated on mouse cells as a prerequisite to anticancer immunotherapy assay in mouse model.

We first studied the invasion and survival capacity of N. caninum tachyzoites in mouse bone marrow dendritic cells (BMDCs) and human blood dendritic cells (DCs), and their respective induced cytokine expression patterns.

We observed that N. caninum tachyzoites had entered, survived and multiplied within human DCs demonstrating for the first time that a strictly non-zoonotic protozoan could infect and proliferate in human DCs (figure 1A). We also demonstrated that human DCs secreted proinflammatory cytokines such as IL-12, TNF-α, IL-10 and IL-6 in response to stimulation with live N. caninum tachyzoites (figure 1B). Similar infectious capacity and cytokine induction profiles were observed in mouse cells including BMDCs (figure 1C,D), and in mouse macrophages and neutrophils (online supplemental figure 1). However, in contrast to mouse BMDCs, human DCs failed to secrete those important innate cytokines when exposed to dead protozoa. Finally, soluble tachyzoite antigens were not able to activate murine or human DCs.

In context of immunotherapeutic treatment with N. caninum, those findings convincingly suggest that the use of live parasites is a sine qua non condition for inducing robust immune responses.

To determine if N. caninum can induce a strong and early immune response without toxicity, mice were subcutaneously inoculated with several doses of N. caninum tachyzoites (2×10³ to 2×10⁷ per mouse). Significant humoral and cellular immune responses were observed via production of N. caninum-specific IgG and of IFN-γ in sera (online supplemental figure 2B-C). In parallel, survival of infected mice was recorded daily (online supplemental figure 2A). Except for the lowest dose of 2×10³, the amount of specific IgG induced was essentially the same for all other doses. Concerning the cellular immune response, the highest doses (2×10⁶ and 2×10⁷) of protozoa were the most immunogenic as we observed a secretion of IFN-γ (≅ 500 pg/mL) in sera from the second day of infection. For lower doses, an IFN-γ production was observed only after a week. As the dose of 2×10⁷ tachyzoites induced the death of all mice after 1 week, we decided to choose the dose of 2×10⁶ parasites with a survival percentage of 60% for immunotherapy assays in mice.

The antitumor effect of N. caninum was evaluated using the tumor-bearing mouse model wild type (WT) EG7 (PD‐L1^lo EG7; OVA‐expressing EL4 lymphoma). The EG7 cells were implanted in the right flank of C57BL/6 mice. To determine an optimal protocol and to assess whether the intratumorous injection of N. caninum could significantly inhibited tumor growth, 2×10⁶ tachyzoites were injected 4 days after implantation with or without a second injection 3 days after (figure 2A).

We observed that intratumorous injection of protozoa significantly reduced tumor growth (figure 2B). Regardless of the protocol, tumors stopped growing 1 week after treatment and then rapidly regressed. Tumors from mice treated twice were undetectable in at least 85% of cases within 22–25 days, leaving a scar that disappears in the following days (figure 2C and J). If a drastic regression of tumors was observed for mice treated only once, very small tumors still persisted at the end of the protocol (<400 mm³) suggesting the need for two injections. Moreover, no recurrence has occurred on treated mice >40 days after treatment (data not shown).

As N. caninum exhibited strong antitumor properties in a non-natural host, suggesting a potential efficacy in humans, we decided to compare its capacity with T. gondii, a naturally infectious parasite of rodents and humans, which had been shown to reverse tumor-associated immunosuppresion and to stimulate effective immune responses to eradicate established tumors.18 After two injections of tachyzoites at the site of the tumor, both groups showed a strong inhibition of tumor development, and total clearance of tumors in some cases within 25 days (figure 2D, E and K). Thus, both N. caninum and T. gondii showed similar therapeutic properties against the EG7 tumor regardless of the mice being a natural or non-natural host of the agent used.

We then tested if N. caninum could still be effective if injected as a distant site from the tumor. As well as intratumorous delivery, remote inoculation succeeded to drastically decrease distant tumors. However, very small tumors persisted and total eradication was not obtained suggesting that N. caninum administration in the tumor microenvironment (TME) enhances its efficacy (figure 2F, G and K). Dead protozoa (heat-killed) failed to induce such tumor regressions; which indicated the need of live protozoa, for a direct or indirect oncolytic activity of N. caninum (figure 2H, 1 and K). To test the long-term effect of tumor treatment, mice in which a complete tumor regression had occurred were re-implanted with EG7 cells (25 days after the first implantation, on the opposite flank). While 100% of the mice injected for the first time with EG7 develop tumors, only 16,7% of the previously treated mice (2 out of 12 mice) harbored small tumors (50 vs 350 mm³ in the control group), within 2 weeks. We observed that all remaining mice rejected the tumor rechallenge and remained tumor-free for 14 days. In naive mice, tumors can be observed as soon as 6 days postinjection of tumor cells and then reach an important volume of approximately 500 mm³. Therefore, an absence of tumors in 14 days allow us to conclude that Neospora induce a protection against a rechallenge with EG7 tumor in those mice, showing that N. caninum therapy also induces a long-term antitumor immunity (online supplemental figure 3A,B).

Finally, we explored the effects of pre-existing anti-Neospora immunity on therapeutic efficacy. We infected mice with 2×10⁶ tachyzoites of N. caninum and used the same treatment regimen 1 month after this first infection. We observed that primary infection did not affect the capacity of protozoa to induce a protective antitumor response as we obtained the same rate of tumor regression (online supplemental figure 3C).

These results demonstrate a durable efficacy of the antitumorous therapeutic treatments with N. caninum, preventing potential relapses without impact of a pre-existing N. caninum immunity, suggesting the possibility of repeated treatment without efficacy loss.

In order to decipher the antitumor potential of N. caninum, we investigated different possible mechanisms of protozoan-mediated tumor regression/destruction: 1) direct cytotoxic activity (oncolytic pathogens), 2) induction of antitumor immune response and 3) reprogramming the tumor microenvironment.

In vitro, transmission electron microscopy and immunofluorescence revealed that N. caninum tachyzoites were able to infect and multiply inside thymoma cells and lyze them (figure 3A). Four days after injection in mice, fluorescence microscopy revealed the presence of parasitophorous vacuoles (the replicative structure of Apicomplexa) in the tumor tissue, proof of the active replication of N. caninum in the TME (figure 3B, right panel). N. caninum was detected by qPCR in the tumor until 11 days post-treatment, but in quantity lower than the initial injected dose. However, beyond day 18 post-treatment, N. caninum was no longer detectable in the tumor or peripheral organs: spleen, brain and liver suggesting a natural clearance of the infectious agent by the mouse immune system (figure 3B). On the contrary, T. gondii was still detected in the tumor 25 days after the first injection, and in quantity far greater that the initial injected dose (200-fold), showing the high multiplication, and persistence abilities of T. gondii compared with N. caninum, despite similar antitumor efficacy. These results strengthen the interest, in efficacy and safety, of using a non-naturally infectious agent as a therapeutic, as N. caninum displays strong antitumor properties and is then naturally cleared from the organism.

We evaluated systemic immune cell activation induced by N. caninum at blood and spleen levels. Four days after intratumorous inoculation of N. caninum tachyzoites, we observed an increase level of circulating IFN-γ, high levels of IFN-γ and IL-12 (online supplemental figure 6A) and low levels of IL-5, IL-10 and IL-17 in spleen with recruitment of DCs, macrophages and natural killer (NK) cells (online supplemental figure 6B). Of note, EG-7 cells being OVA expressing cells, splenocytes were stimulated in vitro with OVA for 72 hours, but no difference in expression of the 10 tested cytokines (including IFN-γ) was observed compared with unstimulated cells, therefore suggesting that the protection mechanisms observed are not OVA dependent (data not shown).

We then investigated in vitro, the ability of N. caninum to alter tolerogenic cells. BMDCs were cultured in presence of vitamin D₃ to induce a tolerogenic phenotype, characterized by a strong secretion of IL-10 and a very low secretion of IL-12 compared with control BMDCs (figure 3C). We confirmed that N. caninum tachyzoites were able to infect tolerogenic DCs (figure 3C) and, 18 hours after infection, to drive the tolerogenic DCs in a non-tolerogenic pattern, once again secreting high levels of IL-12 and low level of IL-10 as compared with control uninfected cells.

Analysis of common immunosuppressive factors, ex vivo, revealed that N. caninum treatment induced a reduction of factors associated with poor prognosis: VEGF-A, PD-L1, IL-10 and TGF-β profile within the EG-7 TME (figure 3D). The favorable capacity seemed specific to N. caninum in contrast to T. gondii that induced an increased pattern of poor prognostic factors (figure 3D).

Together, those results show that N. caninum is able to reprogram tolerogenic cells and reverse the in vivo immunosuppressive environment into an immune-competent antitumor environment.

Generation of an efficient tumor-specific immune response is crucial to sustain tumor control. Cellular and molecular mechanisms underlying protective effects of N. caninum were also studied at tumor level and tumor infiltration was characterized by immune cell phenotyping and cytokine secretion. Hematein-phloxin-saffron stained tumors revealed that almost no necrosis was observed in untreated tumors (median necrotic area=3.1% of the tumor area) while large areas of necrotic tumor cells were observed in mice treated with live N. caninum (median necrotic area=35.8% in NC1-treated group) as shown in figure 4A,B. Necrotic areas within NC1-treated tumors were infiltrated with neutrophils with hypersegmented nucleus, a histological feature of ‘activation’ (N1) as compared with untreated tumors infiltrated with neutrophils of less segmented (ring-shaped) nucleus suggesting an ‘ immature’ (N2) phenotype (figure 4C).

Histological observation and confocal mosaic reconstitution of large section of the tumor parenchyma of untreated mice showed a dense network of EG7 cells (CD3⁺) without necrosis (figure 4A and D), with regularly scattered macrophages (CD68⁺) and few neutrophils (Ly6G⁺).

Following NC1 treatment, the parenchyma of the tumor showed a dismantled cellular structure (figure 4A and E) highly infiltrated with neutrophils and macrophages, with EG7 cells (CD3⁺) harboring a necrotic phenotype (altered DAPI staining) and highly infiltrated with clusters of neutrophils (Ly6G⁺) intermixed with macrophages (CD68⁺) (figure 4E). Higher magnification of tumor sections from NC1-treated mice clearly suggests the close interaction of neutrophils and macrophages within those inflammatory clusters as indicated by the presence of merged area (yellow) of Ly6G and CD68 staining (figure 4F). Three-dimensional reconstitution of Z stacks from confocal images of the tumor parenchyma of NC1-treated mice confirms those interactions within areas of dense infiltrates (figure 4H) on the contrary of untreated mice (figure 4G). Finally, the staining of N. caninum tachyzoites within the tumor 5 days after the first treatment (so before Neospora clearance) revealed that the tissue distribution of Neospora does not seem to follow a particular pattern as tachyzoites can be observed in all areas of the tumor as displayed on large section images (figure 4I).

Moreover, we can assume that Neospora can invade immune cells in vivo, as tachyzoites were colocalized with macrophages (figure 4I and J) within the TME.

We investigated by flow cytometry the immune cell infiltration in the TME. The total amount of myeloid cells was significantly higher in mice treated intratumorally with N. caninum. An 8-fold increase of neutrophils was observed in the tumor, associated with a 3.3-fold increase of macrophages (F4/80⁺ cells) (figure 5A). The number of lymphoid cells also increased in the treated groups, with high infiltration of NK cells (4.4-fold), CD8⁺ T cells (13-fold) and CD4⁺ T cells (2.4) suggesting the development of innate and adaptive immune responses (figure 5A). A somewhat similar profile was observed with the T. gondii treatment. Still, T. gondii seems to induce preferentially a recruitment of macrophages and B cells in the tumor, while N. caninum treatment induces preferentially the recruitment of neutrophils and CD8⁺ T cells.

A similar profile—to a lesser extent—was observed in tumors of mice with remote site treatment. Moreover, as expected, no variation in cell population was found in tumors of mice treated with heat-killed parasites (data not shown).

The increase of those cell populations was correlated with a strong increase of T helper (Th)1 profile cytokines in the TME. Indeed, N. caninum treatment induces a strong increase in the secretion of IL-12, IFN-γ, IL-2 and TNF-α in the TME (figure 5B). As expected, a similar profile was observed in the remotely treated group, but with a lower increase, correlating with the immune cell populations data.

As a matter of comparison, T. gondii strongly increased the secretion of a broad range of cytokines, and the effect on Th1 profile cytokines was similar to the one observed with N. caninum.

In order to assess the respective involvement of immune cells in the Neospora-induced antitumor effect, we administered anti-CD4, anti-CD8, anti-NK1.1, anti-Ly6G monoclonal antibodies (mAbs) or chlodronate liposomes (macrophages depletion) from D1 after Neospora injection, to deplete the respective cell populations. While injection of anti-NK1.1, anti-CD8 mAbs and chlodronate liposomes completely abolished the antitumor activity of Neospora, anti-Ly6G and anti-CD4 mAb did not have any effect on the tumor regression (figure 6). These data show the critical role of NK cells, CD8⁺ T cells and macrophages in the protection processes induced by N. caninum treatment.

In order to assess the effect of N. caninum on human tumors, we used a model of NOD/SCID mice bearing human MCC. After treatment, tumors of around 600 mm³ regressed to a measured volume of 200 mm³ (figure 7A). MCC that developed into established tumors (about 1000 mm³ at day 43 postinoculation) were injected with N. caninum tachyzoites (figure 7A). The volume of established tumors also regressed by threefold, 10 days after treatment.

As expected due to their extremely immunocompromised status, the NOD/SCID mice ended up dying of the N. caninum infection. Hence, qPCR analysis revealed an average of 2×10⁸ tachyzoites per gram of tumor (figure 7B), showing that without the immune pressure, the protozoan has multiplied extensively within the tumor. Thus, the total and natural clearance of the parasite previously observed in WT mice is most probably related to their immune status, and not to a defect in the parasite ability to replicate in host cells (figure 3B).

The ability of the protozoan to infect and multiply in human tumor Merkel cells was confirmed in vitro by scanning electron microscopy and immunofluorescence (figure 7C). These data indicate the potential of N. caninum to infect and multiply in human tumor cells.

Improving the protective effects of N. caninum seems essential in order to obtain a total protection in advanced or refractory tumors. To this aim, we engineered a N. caninum strain able to secrete the human IL-15 (cross-reactive with mouse cells), associated with its sushi domain (alpha-subunit of the IL-15 receptor), increasing its stability, binding and biological abilities, to strongly induce the expansion of Th1-associated lymphocyte subsets and prevent their apoptosis in vivo.27

We first assessed that the engineered clones were able to secrete a biologically active form of the cytokine (figure 7D). We showed that supernatant of NC1-IL15hRec cultures was able to induce IFN-γ secretion by mouse splenocytes in vitro, 4 to 5 times more than supernatant from WT Neospora (figure 7E). The NC1-IL15hRec strain was thus able to secrete a functional IL-15 in its environment.

We then tested the effect of the NC1-IL15hRec strain on human cells. Human PBMCs were infected with NC1 and NC1-IL15hRec. After 24 hours, levels of IL-15 in the supernatant were only detectable for cells treated with NC1-IL15hRec (figure 7F). Then we observed that, as the recombinant IL-15, the NC1-IL15hRec strain was able to induce proliferation of human NK cells as shown by the increase of Ki67 expression by human NK cells (figure 7G). Meanwhile, WT N. caninum displayed a much lower increase of proliferation by NK cells. Moreover, while recombinant IL-15 was only able to induce proliferation, but not IFN-γ secretion, by human PBMCs, NC1-IL15hRec induced a strong IFN-γ secretion by human PBMCs, much more than WT N. caninum (figure 7H). Finally, EG7 bearing mice were treated intratumorally with NC1 or NC1-IL15hRec tachyzoites. This experiment revealed, after two injections of tachyzoites at the site of the tumor, an added-value of the IL-15 secretion by Neospora, as the volume of treated tumors was significantly smaller in tumor treated with NC1-IL15hRec compared with NC1 treatment (figure 7I and J). The analysis of immune cells in the TME revealed that the improved protection was also correlated with an increase of NK cells, CD8T cells and CD4 T cells in the tumor microenvironment (online supplemental figure 4A). No added value was observed for the recruitment of DCs and macrophages. Of note, the level of expression of MHC-II on DCs or macrophages remained unchanged after Neospora treatment, regardless of the strain used, as shown in online supplemental figure 5B). IL-15 was detected in the TME at D25, but was not detected in the blood of treated animals, suggesting a local secretion of IL-15 within the tumor, after Neospora treatment (online supplemental figure 5C).

Those data clearly demonstrated the added-value potential of an IL-15-armed strain of N. caninum that might enhance the already potent immunomodulatory properties of the WT protozoan.