Cancer Immune Therapy Edited by G. Stuhler and P. Walden ...

36 3 Processing and Presentation of Tumor-associated Antigens
antigen differentially seems indicative of this assumption. Since the DC lineage also
expresses PA28 at high levels, the role of PA28 in the processing of self-protein-derived
epitopes may be a great interest. We recently analyzed the processing and presentation
of two CTL epitopes derived from a melanoma differentiation antigen, i. e.
TRP2. In these studies we found that several of our melanoma cell lines failed to present
one of the two antigenic peptides. This deficiency of epitope presentation correlated
with the absence of PA28 in these cells. In support of this we found that restoration
of PA28ab expression rescued the presentation of this CTL epitope. These
data could be further supported by in vitro experiments demonstrating that only in
the presence of PA28 was the proteasome able to liberate detectable amounts of the
TRP2 epitope. Therefore the effects of PA28 on tumor epitope processing are not
fundamentally different from those on viral epitope processing and that, apparently,
the CD8 T cell repertoire available for tumor antigen recognition is not biased towards
epitopes that are processed in a PA28-independent manner.
3.5
Exploiting Proteasome Knowledge
With regard to the development of epitope-based vaccines, a detailed knowledge of
proteasomal cleavage properties is of great practical relevance as it would allow the
identification of CTL epitope candidates within protein stretches. Kessler et al. first
exploited such a ªreverse immunologyº method to characterize HLA-A2-presented
CTL epitopes in PRAME, a tumor-associated antigen that is expressed in a wide variety
of tumors [50]. The PRAME protein sequence was first screened for the presence
of nonamer and decamer peptides with the HLA-A*0201-binding motif. Identified
peptides were synthesized and tested for actual HLA binding. Putative epitope candidates
found to bind with high affinity were synthesized as part of larger polypeptides,
encompassing the potential antigenic sequences and their natural flanking residues,
and offered as substrates to purified 20S immuno-proteasomes. The analysis
of the digestion products led to the identification of four epitope candidates that
were generated by the proteasome. Each of the PRAME-derived peptides induced
specific CTL responses in vitro and CTL clones established by this method specifically
lysed PRAME-expressing HLA*0201 + tumor cells.
To simplify CTL epitope identification strategies, computer prediction programs that
predict cleavage site usage within proteins have been established [51±53]. So far two
prediction programs based on different parameters have been published. One program
was trained using 20S cleavage data of the enolase-1 protein; the other program
is based on polypeptide cleavage data generated in a large number of studies.
Both programs predict proteasomal cleavages with high fidelity, whereby the latter
program not only identifies proteasome cleavage sites but also predicts the probability
with which specific fragments are generated. Indeed, the application of this program
in combination with a program that predicts MHC class I binding affinity led
to the identification of eight CTL epitopes derived from the entire deduced proteome
of Chlamydia trachomatis that are presented by HLA-B27 molecules [54].