Much of the graft versus leukemia (GvL) effect observed following allogeneic stem cell transplant (HSCT) is a result of minor histocompatibility mismatches between the donor and the patient; however, only a handful of these useful immune targets have been discovered by traditional biochemical or cell cloning methods over the past few decades. New approaches to antigen discovery are therefore needed. By combining the large patient sample availability at M.D. Anderson Cancer Center with high throughput genotyping, statistical analysis and computational modeling, we have developed a method to efficiently screen > 250,000 potential minor antigens (mHAgs) before immunological confirmation. This approach allows us to synthesize a relatively small number of candidate peptides and peptide/HLA tetramers that can be used to validate predictions using post-transplant patient samples. Additionally, because the statistical tests used to rank candidate antigens is based upon donor/patient non-synonymous, single nucleotide polymorphism (cSNP) disparities, minor antigens with high minor allele frequencies will be ranked higher which should lead to the discovery of minor antigens that could be clinically useful. Using this novel method, we report the discovery of a new mHAg derived from the tripartite motif protein TRIM42. To validate this approach, genomic DNA samples from 50 HLA-A2+ myeloid leukemia patients and their matched related donors underwent genotyping at 13,937 well-characterized, non-synonymous cSNPs using Illumina NS-12 microarrays. The frequency of cSNP disparities predicted to yield minor histocompatibility mismatches (homozygous donor transplanted into either a heterozygous patient or a patient who was homozygous for the other allele) was correlated with the presence of an immune response (IR) defined as graft versus host disease (GvHD) and/or a complete remission at 2 years post-transplant. Of the 13,937 cSNPs investigated, 14 were identified by Fisher’s exact test to have disparities highly correlated with IR. Of the 266 potential nonameric and decameric peptides containing the identified cSNPs, 23 were predicted to bind to HLA-A2 according to the SYFPEITHI (binding score > 14) and IEDP (binding score < 500) algorithms. Seventeen of these 23 peptides could be synthesized, and 5 of those were confirmed to bind HLA-A2 using a T2 cell based binding assay. To determine whether these peptides were potential targets of CD8+ T lymphocytes, we synthesized peptide/HLA-A2 tetramers for all 5 peptides that bound HLA-A2 and analyzed post-transplant patient samples. A population of CD8+ T lymphocytes that bound to a HLA-A2 tetramer loaded with the peptide T4A (GLYTYWSAGA) was identified in post-transplant samples from 2 of 3 patients predicted to have a minor histocompatibility mismatch while no tetramer positive lymphocytes were observed in post-transplant samples in 2 out of 2 HLA-A2+ patients not predicted to have this mismatch. The TRIM42-derived peptide T4A contains a cSNP, rs9876490, that encodes either a cytosine (HapMap allele frequency = 0.575), which results in the translation of the c-terminal alanine (GLYTYWSAGA) in T4A or adenine (HapMap allele frequency = 0.425), which yields a c-terminal glutamic acid (GLYTYWSAGE). TRIM42 is a member of the tripartite motif family of proteins and likely associates with the microtubule cytoskeleton; however, its actual function is unknown. The protein is normally expressed in brain, testis, and peripheral blood mononuclear cells (PBMC); however, it has also been detected both in leukemia samples and leukemic cell lines (www.proteinatlas.org). Our data demonstrate that this novel approach predicts and biologically validates new mHAgs. The major use for this method is to show which of many potential mHAgs will be the dominant CD8+ lymphocyte targets for GvL and/or GvHD. The method and the antigens discovered by this approach might also help to clarify the relationship between GvL and GvHD and how they are linked.
Disclosures: No relevant conflicts of interest to declare.