Changes in tumor-specific antigenicity overtime in melanoma patients undergoing adoptive T cell therapy

Most tumors, and most likely "all" tumors, express one or more non-synonymous somatic mutations that generate sufficient number of novel HLA/peptides that potentially could be targeted by T cells. However, similar to infections, when under immune pressure tumor can also develop escape variants by modulating its HLA/peptide landscape. 

This is what new study published in journal Nature found in tumor patients. For this study, the authors (who are associated with AIMM Therapeutics) sequenced patients tumor mRNA at different time points (before or after adoptive cell therapy) and found that some of the tumor-specific HLA/peptides had disappeared [overtime] and some new tumor-specific HLA/peptides became enriched instead. 

Of note, within each patient almost all of T cell reactivity to tumors were patient-specific (i.e. private HLA/peptide) and did not share sequence similarities with publicly known shared tumor-associated epitopes. For example, analysis of patient's CD8+ T cell specificity against a panel of >200 MHC-multimers containing known shared tumor-associated epitopes demonstrated that only 1.24% of the CD8+ T cells responded to three different "shared" gp100 epitopes. The rest were patient's specific. 

For this study two melanoma patients had their tumors mRNA sequenced and underwent autologous adoptive tumor-specific T cell therapy. In one patient, initial tumor mRNA sequencing revealed  501 non-synonymous mutated genes. Most of CD8+ T cell reactivity were directed to two HLA/peptides complexes: KIAA0020 p.P451L (KIA P>L) and ribosomal protein RPL28 p.S76F (RPL28 S>F).

Around 1 year later, however, patient's CD8+ T cell reactivity was only observed against the KIAP>L neo-antigen, and not against RPL28S>F and tumor mRNA sequencing confirmed absence of mutant allele encoding the RPL28S>F neo-antigen within tumors harvested at a later time.

Similarly, in an another patient undergoing similar procedures, the authors found that initially melanoma patient's CD8+ T cells were specific for the neo-antigens echinoderm microtubule associated protein like 1 p.R64W (EML1R>W), Septin-2 p.R300C (SEPT2R>C), and CAD protein p.R1854Q (CADR>Q).  

However, later analysis revealed no significant T cell reactivity against EML1R>W, CADR>Q neo-antigen and SEPT2R>C neo-antigens. Instead, it showed T cell response towards the Programmed Cell Death Protein 10 p.P28S (PDCD10P>S) neo-antigen. Indeed, tumor mRNA sequencing confirmed that while mutant allele encoding the SEPT2R>C neo-antigen that was present in the original tumor, it was selectively lost in the tumor samples harvested later, and that RNA for novel PDCD10 neo-antigen had increased > 40-fold in later time points instead.

These data points to two important considerations: first, tumor undergoes changes in their mutational landscape and can evade T cell detection and second, most tumor mutations are patient's specific. This means that adoptive immunotherapy using single T cell specificity would be less efficient overtime and that tumor vaccines using "generic-shared" tumor epitopes would provide no benefits in most patients. 

David Usharauli