For personalized therapies based on our prediction of target antigen, we need a customizable chassis to present this antigen to the immune system. We went through a selection process for the chassis presenting the best features to deliver the patient tumor antigen to the immune dendritic cells (DCs). We chose to target DCs because they are considered the best candidate for T-cells activation against cancer (8). The chassis must be immunogenic to serve as an adjuvant for the antigen, because immature DCs presenting a tumor antigen without danger signal will induce T cell tolerance. Micro-organisms are natural adjuvants bearing the danger signal as well as the antigen targeted. In addition, they can be standardized and target DCs in vivo, thereby reducing production costs. However, skeptic shock or cytokine storm must be absolutely prevented.

Bacteria are well-adapted vector for in vivo therapy with Dendritic Cells (DCs). First, they naturally increase tumor immunogenicity because their toll-like receptors induce an inflammatory cytokine response that attract DCs. Second, they are able to deliver tumor antigens to DCs through variant mechanisms. However, they raise safety concerns for virulent reversion in patient. P. aeruginosa, a human pathogen, has been designed to inject directly the tumor antigen with its type III secretion system (T3SS) in DCs cytosol (9). In a curative assay on mice with melanoma, injection of the vaccine vector after tumor implantation led to a complete cure in five of six animals (10). This virulent bacterium is attenuated with the killed but metabolically active (KBMA) method. While this system has shown to attenuate toxicity both in vitro and in murine model, it also reduces antigen presentation by 70 % in comparison with the live vector, raising both safety and efficiency problem  (11).

Listeria monocytogenes is a natural choice because this intracellular pathogen has the ability to enter DCs and to activate MHC I Pathway. Once inside the phagosome, it releases a lysosomal enzyme that is active at acidic pH to degrade the phagolysosome, the Listeriolysine O (LLO). This therapy performed well in Phase I clinical trial against invasive carcinoma of the cervix (12). However, Listeria has to be genetically modified to suppress genes that could cause virulence reversion and raises again safety concerns. In addition, Listeria immunotherapy is not as adapted as E. coli for metabolic engineering to overproduce antigens and co-stimulatory molecules within DCs for a strong immune response.

E. coli has therefore been proposed to inject the tumor antigen in DCs with the heterologous LLO from Listeria. Immunization of mice by direct injection of E. coli LLO/OVA provided a potent anti-tumor response, resulting in complete protection in 75% of mice (13). The drawback is LLO toxicity for the cell, with the absence of regulation by E. coli on the contrary of Listeria. In addition, even if E. coli is less pathogenic than Listeria, a live vector that replicates inside a patient raises safety issues. (13). 

To address both safety and efficiency, we selected S. cerevisiae for the following advantages:

A. S. cerevisiae is non pathogenic : phase I clinical trial with subcutaneous injection of heat-killed yeasts S. cerevisiae against hepatitis C showed no dose-related toxicity (14), demonstrating the safety of this vector for future human applications in cancer immunotherapy.

B. S. cerevisiae has a strong adjuvant effect, explaining why zymosan, an extract of S. cerevisiae cell wall, has been used to stimulate inflammation for 50 years (15). In particular, the mannose stimulate pro-inflammatory cytokines production in monocytes and dendritic cells, making the vector appropriate for APC targeting (16,17). 

C. S. cerevisiae already proved its anti-tumor capacity : the first demonstration of recombinant yeast to induce adaptative immunity was shown in 2001 by Stubbs et al against ovalbumin tumor cells (18). Yeast expressing the mutated RAS protein inside their cytosol induced reduction of lung tumors in mice and a quarter of the tumors were eradicated (19). Tumor antigen MART-1 was also expressed in yeast cytosol and induced both CD4+ and CD8+ in mice. (20)

D. S. cerevisiae prior immunization with the wild type do not create a response neutralizing the antigen expressing yeast, allowing repeated injections of the vector. (21)

Surface display of tumor antigen for CD8+ cross-priming

We chose to express our antigen on the membranes of S. cerevisiae because surface displayed antigen is cross-presented much more efficiently than yeast cytosol antigen (22). This is due to a particular kinetics inside the early phagosome, allowing the external antigen to escape from the phagosome. Cross-presentation can be further enhanced by inserting linkers susceptible to Cathepsin S cleavage between the antigen and Aga2p, supporting the evidence that early antigen release is important for cross-presentation (22).

Enhancing cross-priming with the antibody anti-DEC205

Our surface display antigen for ovalbumin was fused to DEC205 scFv. DEC205 is a lectin receptor expressed by some DCs subsets, including mouse spleen DC (23). It was shown that antibody targeting DEC-205, fused to tumor antigen, can induce T cell stimulation if administered with an additional stimulus triggering DC maturation, like anti-CD40 agonistic antibody (24). In the same way, immunization with DNA vectors encoding antigens fused to a DEC-205 scFv elicits a strong specific CD8+ responses in vivo (25).

The scFv of DEC205 was fused to our ovalbumin tumor antigen and surface displayed in order to get the yeast internalized in a DC endosome through DEC205 receptor, favoring CD8+ cross-presentation. We used the scFv instead of the whole antibody for the possibility to perform repeated immunisations without inducing deleterious host responses against the Fc part of the immunoglobulin chains.

Advantages of yeast expressing DEC205 over DEC205 protein vaccines

Pure protein vaccines with DEC205 are far less immunogenic than vaccine with micro-organisms mimicking pathogens and request an additional adjuvant. Moreover, the protein needs a prior step of antigen purification (26), leading us to develop this yeast surface display of DEC205 scFv fused to the antigen. The advantages of our system include better concentration of the yeast due to less diffusion than the protein DEC205 alone, the codelivery of both antigen and adjuvant, the possibility to target multiple DCs compartments at the same time (MHCI and MHC II) and the absence of purification step.

Surface display design

Several surface display systems exist for the yeast S. cerevisiae. In the context of cancer immunotherapy, whole yeast cells has been coated with several layers of cancer-testis antigen NY-ESO-1 with a chemical conjugation (27) and this system was able to cross-prime naive CD8+ T cells in vitro. Antigen was also linked chemically to the surface of a capsular yeast shell instead of the whole yeast (28). The advantage of chemical conjugation is the ability to reach a high antigen loading. However, this technique is limited to soluble antigens and most antigens are not soluble, leading us to reject this solution in order to broaden our system to any tumor antigen. In addition, chemical conjugation requires a purified antigen, increasing therapeutic application costs.

We selected the surface display system based on the mating adhesion receptor Aga2p and Aga1p. This system is widely used for antibody affinity studies and was used to anchor the antibody ScFv DEC205 fused to the ovalbumin tumor antigen to the yeast surface. Aga1p was expressed separately and aga2p fused in C-terminal to our displayed protein.

To establish a proof of concept, our system was tested in vivo on C57BL/6 mice injected with the melanoma cell line B16-OVA expressing the ovalbumin antigen. We also tested the system in vitro on hybridoma B3Z T-cells specific for SIINFEKL. The tumor antigen cloned in our vector was OVA1 corresponding to the sequence QLESIINFEKLTEW, class I (Kb)-restricted peptide epitope of ovalbumin (OVA) plus 3 amino acids around the epitope to allow better digestion by the proteasome. It is presented by the class I MHC molecule H-2Kb (29).

Figure 1: Yeast surface display expressing troll antigen to carry out immunotherapy via MHC-I

(1) Surface Display of tumor antigen OVA1 fused to DEC205 scFv

(2) Yeast internalization in cross-presenting endosomes specific for DEC205

(3) CD8+ T cell cross-priming with tumor antigen OVA1

(4) Cancer cell lysis with antigen OVA1 targetin

Cloning results

We transformed the yeast to express OVA1, DEC-205 or OVA1-DEC205 on surface. We used AGA1P co-expression and AGA2P C-terminal fusion to the protein in order to get membrane presentation.

Figure 2: Plasmids with the constructions : A) AGA1P (B) AGA2P-OVA1-DEC205 (C) AGA2P-OVA1 (D) AGA2P-DEC205

Surface display results

Before displaying our tumor antigen fused to the scFv DEC205, we first cloned the fusion protein AGA2P-GFP in order to observe surface display of GFP with our without the coexpression AGA1P. We obtained a GFP signal located around the membrane only in presence of AGA1P coexpression.

Figure 3: GFP fused to AGA2P observed in fluorescence microscopy. A) without AGA1P (B) with AGA1P (C) with AGA1P with Z-scale on one single yeast.

Conclusion

We have transformed a yeast producing high levels of GFP and surface displaying the protein through AGA1P. Instead of GFP, this chassis is ready to produce a high antigen loading to induce the strongest immune response.

References