Overlap of the cancer genome atlas and the immune epitope database

Introduction

The development of cancer vaccines has become a high priority in the field of cancer treatment. However, there are numerous parameters that affect the success of an immune response against an antigen, including the binding of a T-cell receptor to a major histocompatibility complex (MHC)-bound antigen and antigen processing. Empirical evaluations of vaccine efficacy parameters are costly and time-consuming. Thus, bioinformatic approaches may provide a useful alternative.

The Immune Epitope Database (IEDB) includes >35,000 human peptides known to either bind to human leukocyte antigen (HLA) class I or II, or to have other immune receptor binding properties (1,2). Knowledge of the capacity of a peptide to bind to antigen-presenting molecules could potentially improve the selection of cancer vaccine candidates that are based on mutant peptides, whether these result from cancer drivers or passenger mutations. In addition, sufficient database development may allow for a better understanding of any presumed selection against the binding of cancer peptide neoantigens to MHC molecules as an aspect of cancer development.

The present study focused on searching for overlaps of The Cancer Genome Atlas (TCGA) mutant peptides (3,4) and peptides in the IEDB, in order to discover cancer-related peptides that have the demonstrable capability to bind to MHC molecules.

Materials and methods

The overview of the approach is provided in Fig. 1. Supporting online material (SOM) representing each stage of the approach were also used in the present study (http://www.universityseminarassociates.com/Supporting_online_material_for_scholarly_pubs.php) (5). Briefly, human epitopes were downloaded from the IEDB (www.iedb.org) using the following search terms: Epitope, linear epitope; antigen, Homo sapiens (human) (ID: 9606, Homo sapiens); host, humans; assay, all assays; MHC restriction, MHCI and MHCII; disease, any disease. The results comprised ~35,000 epitopes and were downloaded as an Excel file. Epitopes that ranged between 14–18 amino acids (AAs) in size were used to determine mismatches with the human genome version 19 (hg19) reference genome at genome.ucsc.edu. The nucleotide spans of the mismatches were obtained by using the BLAT search (http://genome.ucsc.edu/cgi-bin/hgBlat?command=start) to search a local database, which was created from the TCGA download portal and consisted of a collection of all TCGA cancer mutations (https://tcga-data.nci.nih.gov/tcga/tcgaDownload.jsp). The recovered nucleotide positions were extended with hg19 nucleotides using The Extract Genomic DNA tool at https://usegalaxy.org/. The extended regions were then translated into all possible reading frames (including forward and reverse) using the European Molecular Biology Laboratory-European Bioinformatics Institute to generate a database for screening the IEDB 14–18 AA set, in order to verify the IEDB matches and to detect mismatches at the location of the TCGA mutation. All HLA candidates were removed due to overly extensive sequence variation. Gene family members that were originally inaccurately regarded as hg19 mismatches, may be found in the SOM files by Sait et al (5) (Table I).

Figure 1. Overview of the procedure used to determine whether any IEDB peptides, which did not match the hg19, matched putative TCGA mutant peptides. The file numbers (1–4) in the figure refer to the supporting online material files by Sait et al (http://www.universityseminarassociates.com/Supporting_online_material_for_scholarly_pubs.php), which include results and details on each step of the procedure. IEDB, Immune Epitope Database; TCGA, The Cancer Genome Atlas; hg19, human genome version 19; EMBL-EBI, European Molecular Biology Laboratory-European Bioinformatics Institute.

Table I. Identification of IEDB peptides that overlap the position of a mutant amino acid in the TCGA database.

Table I.

Identification of IEDB peptides that overlap the position of a mutant amino acid in the TCGA database.

Hugo gene symbol	Normal peptide sequence (hg19)	IEDB ID number	IEDB peptide (hg19 mismatched amino acid in large type, bold)	hg19 translation of TCGA mutation on either side of mutation position	TCGA mutation chromosome number	TCGA cancer dataset	TCGA nucleotide number	MHC
ATP1A2	NPREAKACVVHGSDLK	103447	NPRDAKACVVHGSDLK	VVHGSDL	1	LUAD/LIHC	160104960	HLA-DQB
				ACVVHG	1	LIHC	160104952
				CVVHGS	1	LIHC	160104955
				VVHGSDL	1	LUAD/LIHC	160104960
COL2A1	GEPGIAGFKGEQGPKG	107398	GKPGIAGFKGEQGPKG	IAGFKG	12	READ	48380651	HLA-DRB1
TROVE2	LQEMPLTALLRNLGKM	118499	LQEMPTLALLRNLGKM	ALLRNL	1	SKCM	193045674	NA

[i] Data were obtained from the supporting online material files by Sait et al (5) (http://www.universityseminarassociates.com/Supporting_online_material_for_scholarly_pubs.php) using the procedure described in Fig. 1. IEDB, Immune Epitope Database; TCGA, The Cancer Genome Atlas; hg19, human genome version 19; ID, identification; MHC, major histocompatibility complex; ATP1A2, ATPase Na⁺/K⁺ transporting subunit α 2; LUAD, lung adenocarcinoma; LIHC, liver hepatocellular carcinoma; READ, rectum adenocarcinoma; SKCM, skin cutaneous melanoma; COL2A1, collagen type II α 1; TROVE2, Telomerase, Ro and Vault domain family member 2; LUAD, Lung Adenocarcinoma; HLA-DQB, major histocompatibility complex, class II, DO β; HLA-DRB1, major histocompatibility complex, class II, DR β 1; NA, not applicable.

Results and Discussion

The present study was required to determine whether detecting an IEDB peptide that had a mismatch at the exact position of a TCGA mutant AA was possible. Therefore, a search was performed among the 8,890 IEDB human peptides consisting of 14–18 AAs, with translated AAs on either side of all TCGA point mutations, to check for overlap with an IEDB epitope that had a mismatch with the hg19 version of the reference genome. Since the translations represented exact matches with the hg19 translations, the 8,890 epitopes consisting of 14–18 AA were searched, allowing for one mismatch with the translations used, in order to ‘surround’ the location of the TCGA mutation. According to this protocol, while the TCGA point mutation-referenced translations overlapped the position of the TCGA mutation, these translations matched hg19 exactly, thus requiring the single mismatch standard for searching the aforementioned 8,890 IEDB epitopes for an exact match.

Numerous IEDB epitopes were identified using this method; however, following the exclusion of IEDB epitopes that did not match the gene of the TCGA mutation, only one IEDB peptide had a non-hg19 AA in the position of the TCGA mutant AA. This IEDB epitope mapped to integrin subunit β 3 (ITGB3), which is a known ITGB3 single nucleotide polymorphism. The data supporting this finding is presented in SOM file no. 5 of Sait et al (5).

To determine whether the TCGA mutant AA positions overlapped IEDB peptides that contained a mismatch with the hg19 AA sequence, without the TCGA position equaling the precise location of the IEDB mismatched AAs, the protocol indicated in Fig. 1 was followed. The results are provided in Table I. This protocol indicated that, following the removal of mismatches attributable to closely associated family members or mismatches detected anomalously due to repeats within a protein, 3 IEDB peptides, which were a mismatch to hg19, also overlapped the position of the TCGA mutant AA. For details of the results that were obtained by pursuing this approach, including the discounted IEDB peptides that were anomalously recovered using the Fig. 1 approach, please see SOM file no. 6 in Sait et al (5). Overall, these results indicate that mutant peptides in human cancer overlap apparent mutant peptides in the IEDB, suggesting that the AAs surrounding TCGA mutants are not fundamentally a hindrance to MHC binding. Notably, two of the proteins represented by the overlap of TCGA mutations and IEDB non-hg19 peptides represent the extracellular matrix, ITGB3 and collagen type II α 1, an emerging topic in the field of cancer research (4,6,7).

However, the general paucity of the overlap of the two databases strongly indicates that, from a bioinformatic perspective, there is very little information available for determining which cancer drivers or passenger mutations have the potential of significant MHC binding. This conclusion is even more striking considering the extensive MHC polymorphism and protease activities that could impact binding affinities of cancer peptides (8).

In conclusion, there is a strong case to be made for the development of a more comprehensive human immuno-peptidome project, with the particular aim of determining whether cancer peptides are selected for the reduced likelihood of MHC occupancy.

References