Considerable progress has been made in the last decade to induce tumor-specific immunity. One bottleneck in this progress is the identification of tumor-specific antigens. The ideal target antigen should be highly immunogenic and abundant in the tumor, but scarce or non-existent in normal tissues. In addition, target antigens must be processed by tumor cells into peptides that have a high affinity for MHC-I molecules. These peptides are usually between 8–10 amino acids residues. Different studies reported that several CTL raised against high-affinity binding peptides did not recognize tumor cells expressing the targeted antigen, mainly because their protein processing machinery did not display the peptides.28–30
Furthermore, TCRs that have a high avidity for their peptides are often eliminated. Thus, there is an increasing interest in generating a low avidity TCR repertoire for improved tumor-specific immunity. A vector that permits in vivo and/or ex vivo evaluation of different TSA or TAA should: (1) deliver a broad variety of antigen candidates with different CD4+
epitopes to APCs; (2) stimulate the innate immune system; and (3) be versatile and easy to produce. The live bacterial vector, CHA-OST, was engineered to: (1) reduce its toxicity while maintaining its ability to stimulate DCs8
and (2) allow the easy and rapid expression of different proteins.11
Furthermore, the production of the vector is very simple, since 1 ml of bacteria culture is sufficient to vaccinate up to 200 mice. This approach allows the rapid construction and production of numerous vaccine strains with different antigen/epitope compositions for rapid antigen/epitope screening. We investigated the capacity of this vector to produce various antigens of different length, immunize mice before subcutaneous glioma challenge, elicit TRP-2-specific CD8+
T cell responses, and identify the portion of the TRP-2 antigen that induces efficient tumor-specific immunity.
We have shown that different antigen fragments from TRP-2, gp100 and MUC18 can be efficiently produced and transported through the TTSS of our live bacterial vector. Variations in the level of secreted protein were most likely due to the amino acids sequence and/or the tertiary structure of the proteins. The structure of the needle complex, isolated from Salmonella typhimurium
, suggests that it serves as a hollow tube through which the exported proteins are actively transferred to the target cell.31
Additionally, the TTSS includes components that may be involved in proteins trafficking to the channel. Little is known about the actual export mechanism, but recognition of a signal sequence on the mRNA or on the exported protein itself may be involved. In addition, specific chaperones may be directly involved in protein export or in stabilizing mature proteins prior to export.32–34
Whether the exported proteins travel folded or unfolded through the needle complex has not yet been determined; however, given the size of the channel, it is likely they travel at least partially unfolded. A broad variety of proteins were previously expressed as fusion proteins with the N-terminal part of ExoS and secreted through the TTSS.11
From our experience, proteins secreted at the level observed in the SDS gel in , are also translocated in cells. Here we demonstrate the injection of active β-lactamase (36 kDa) in vivo in DCs after subcutaneous injection of bacteria. Nevertheless, some small and non-hydrophobic proteins, such as survivin, can interfere with the TTSS channel and may not be secreted and/or injected. Our future investigations seek to predict the effectiveness of TTSS mediated transport to enhance antigen screening procedures.
Prins et al. showed that the melanoma associated antigens, (MAA) TRP-2 and gp100, are expressed in murine glioma and presented in a MHC class-I context.35
TRP-2 is widely used as a tumor associated antigen in different mouse and human tumor models. These studies typically used the TRP-2180–188
(SVYDFFVWL) peptide for MHC class I-presentation on CD8+
We produced and evaluated three constructs of this MAA. TRP2 and TRP2L contain the TRP-2180–188
epitope. The construct TRP2S and TRP2L contain epitopes that include H2-Kb restricted TRP-2218–220
(TWHRYHLL) and H2-Db restricted TRP-2363–372
(SQVMNLHNL) and are predicted to have strong binding capacity. Survival analysis showed protection after vaccination with CHA-OST expressing TRP2L but not TRP2 or TRP2S. Neither the TCR repertoire diversity nor the CTL responses showed a significant difference between the three constructs. Although there was a significant change in CTLs the ability of TRP2S vaccine to protect against tumor development was low and comparable to the TRP2 construct. It has been demonstrated that the C-terminal flanking residues determine the efficiency of epitope liberation by the proteasome. 36 More recently, Textoris-Taube et al. showed that the N-terminal flanking region of TRP–2 determines whether the proteasome activator, PA28, will process the epitope.37
Therefore one explanation for the lower efficiency of TRP2S, compared to TRP2L, could be the length and/or the N- or C-terminal flanking region of the construct. On the other hand, different CD8+
epitopes, including TRP-2180–188
(SVYDFFVWL) are not contained in TRP2S. With the addition of the widely-used MHC class II-presented CD4+
T-helper epitope, PADRE, we have shown that CD8+
epitopes on the TRP2S construct elicit a potent tumor-specific CTL response. Therefore, we hypothesize that TRP2L may contain a strong tumor-specific CD4+
epitope that stimulates Th CD4+
cells. Indeed, Robbins et al. identified a human leukocyte antigen (HLA)-DR15 restricted epitope on human TRP-2 (80% homology to mouse TRP-2) at the position 241–250 (ALPYWNFATG).38
Our TRP2L contains the same sequence and is also predicted to bind to I-Ab molecules. This T helper epitope on TRP2L could, in part, explain its superior anti-tumor protection. The high homology between human and mouse TRP-2 sequences allows a comparison of our results to human applications. The two predicted epitopes TRP-2218–220
(TWHRYHLL) and TRP-2363–372
(SQVMNLHNL) are very similar to two identified HLA-I restricted immunodominant epitopes TRP-2197–205
Therefore, we believe that optimizing the peptide length and including different immunodominant and cryptic CD8+
epitopes could improve the efficiency of anti-tumor vaccination, especially for new antigen targets where specific epitopes have not been identified.
In this work, we evaluated two other MAA, including MUC18 and gp100, as well as the universal TAA, survivin. MUC18 is known as a melanoma cell adhesion molecule and is expressed by advanced and metastatic melanomas but not in normal melanocytes.41
Recently Leslie et al. showed efficient CD8+
T-cell immune responses against MUC18 expressing melanomas in response to vaccination with an a-virus based DNA plasmid.18
In our experiment, the candidate vaccines did not effectively protect mice from challenge with glioma. However, the appearance of tumors for MUC18N was slightly delayed compared to the control (p = 0.065), indicating that optimizing the amino acid sequences and/or length could improve protection against glioma cells. For the gp100 and hgp100 vaccines, only three amino acids from the original gp100 sequence were included at the N-terminus. As previously described for TRP-2, the epitope sequence environment for our gp100 and hgp100 vaccines may not be optimized for epitope processing and presentation. These results highlight the optimization required to develop a long peptide or small protein vaccine. Since the TTSS mediated in vitro secretion of survivin was very low, the delivery of the antigen in vivo was probably not sufficient to elicit CD8+
T cell immunity against survivin-expressing tumor cells.
Immunization with single or multiple epitopes from one tumor-specific antigen often results in the outgrowth of antigen loss variants.42
It is not clear how the same antigen can stimulate CD4+
helper cells and/or regulatory T (Treg) cells, which mediate immune suppression. It is likely that future approaches to immunotherapy will benefit from immunizing with long peptides or even small protein vaccines that contain multiple HLA-class I epitopes from a variety of target antigens and different tumor-specific MHC II epitopes. Therefore the live bacterial vaccine presented in this study is an ideal vector for in vivo and ex vivo optimization and evaluation of each candidate antigen. The development of humanized mouse models will allow antigen screening and optimization to be translated into relevant human therapies.