A couple of weeks ago, talking about tumor vaccination, I said this:

It has been suggested that cancer antigens tend to be poor MHC binders. That would mean that the peptide falls out of the MHC complex relatively rapidly and becomes invisible to T cells, so that to keep a certain level of target on the cell surface, you’d need to start with much more; in other words, the peptide would be less immunogenic. (This has been explicitly shown for a number of tumor epitopes, but as far as I know has not been globally demonstrated. It occurs to me that the Immune Epitope Database [IEDB] may have enough information to at least make a start at that analysis; maybe I’ll take a run at it, in whatever of my free time isn’t taken up by playing baseball with my fanatic son.)

I took a quick pass at the question last night and came up with a firm “Maybe“.

The question is whether tumor epitopes — that is, peptides that cytotoxic T lymphocytes recognize when they specifically attack tumors — tend to bind with lower affinity to the MHC class I molecules, compared to regular old epitopes from, say, viruses or bacteria.  Using the Immune Epitope Database I came up with 8620 MHC class I epitopes whose affinities were recorded. 1   I exported these using a custom report.2   I compared these to the 750-odd peptides listed in the Cancer Immunity Tumor Antigen database, finding 145 hits — i.e. 145 tumor antigens whose affinity has been recorded.  I took the affinity (in nM) from these and from the remaining 8475 epitopes.

<1 0.69 7.89
1-2 2.07 5.26
2-4 8.28 5.71
4-16 25.52 19.14
16-256 55.17 45.92
256-65536 8.28 16.01
>65536 0 0.06

Because MHC class I binding affinity spreads over a huge range, averages are pretty meaningless.   I broke the samples down into ranges of affinity and calculated the percentage of epitopes that fell into each (see the table to the right). Results are summarized in the chart below (click for a larger version).   In brief, tumor epitopes are less likely to have very high affinity (about 3% of tumor epitopes, vs. 13% of the remainder), and more likely to have moderate affinity (89% vs. 70%).   On the other hand the tumor epitopes were less likely to have very low affinity.   I suspect this is because the tumor epitopes are more likely to have been detected clinically, which biases toward at least moderate affinity or you can’t find them.

Tumor epitopes, IC50 nM

This is just a quick and dirty check. I didn’t try to validate any of the peptides or the affinities.   I pooled together affinities that were derived using different methods.   There are only a handful of tumor epitopes with affinities known, so this is a small sample size.   Still, there does seem to be a trend there, which I didn’t expect to see — I thought I was going to throw cold water on the hypothesis, which was proposed based on only four or five epitopes. Kind of interesting, I thought.

The scripts etc that I used to work these out are available on request, though I suspect anyone able to tweak them for their own use could do a better job writing them than me.

  1. Summary of the process: Select class I binders
    Limit to IC50 info: Purified MHC – Radioactivity-Competition (or equilibrium binding)-IC50 nM
    More than 10,000 so split into three parts for export:
    <10 nM: 4,716 items found
    10-100 nM: 6,855 items found
    >100 nM:  6,463 items found

    In addition: Limit to Cell bound MHC – Fluorescence-Competition (or equilibrium binding)-IC50 nM; Cell bound MHC – Fluorescence-Association (or direct binding)-EC50 nM; Cell bound MHC – Radioactivity-Competition (or equilibrium binding)-KD nM; Cell bound MHC – Radioactivity-Competition (or equilibrium binding)-IC50 nM; Cell bound MHC – T cell response-Competition (or equilibrium binding)-IC50 nM; Lysate – Radioactivity-Association (or direct binding)-EC50 nM; Purified MHC – Fluorescence-Competition (or equilibrium binding)-IC50 nM; Purified MHC – Fluorescence-Association (or direct binding)-EC50 nM; Purified MHC – Fluorescence-Competition (or equilibrium binding)-KD nM
    <100000 1,822 items found []

  2. Article PubMed ID
    1 Epitope Linear Sequence
    2 Epitope Source Accession Number
    3 Assay Type Category
    4 Assay Type
    5 Qualitative Measurement
    6 Quantitative Measurement
    7 Measurement Inequality
    8 Units
    9 MHC Allele
    10 MHC Allele Class []