When the dissociation constant doesn’t tell you what you want to know

Drug chemists spend a lot of time getting their drugs to bind tightly to their chosen target.  Kd’s (dissociation constants) are measured with care –https://en.wikipedia.org/wiki/Dissociation_constant.  But Kd’s are only  a marker for the biologic effects that are the real reason for the drug.  That’s why it was shocking to find that Kd’s don’t seem to matter in a very important and very well studied system.

It’s not the small molecule ligand protein receptor most drug chemists deal with, it’s the goings on at the immunologic synapse between antigen presenting cell and T lymphocyte (a much larger ligand target interface — 1,000 – 2,000 Angstroms^2 — than the usual site of drug/protein binding).   A peptide fragment lies down in a groove on the Major Histocompatibility Complex (pMHC) where it is presented to the T lymphoCyte Receptor (TCR) — another protein complex.  The hope is that an immune response to the parent protein of the peptide fragment will occur.

 

However, the Kd’s (affinities)of strong (e.g. producing an immune response) peptide agonist ligands and those producing not much (e.g. weak) are similar and at times overlapping.  High affinity yet nonStimulatory interactions occur with high frequency in the human T cell repertoire [ Cell vol. 174 pp. 672 – 687 ’18 ].  The authors  determined the structure of both weak and strong ligands bound to the TCR.  One particular TCR had virtually the same structure when bound to strong and weak agonist ligands. When studied in two dimensional membranes, the dwell time of ligand with receptor didn’t distinguish strong from weak antigens (surprising).

In general the Kds  pMHC/TCR  are quite low — not in the nanoMolar range beloved by drug chemists (and found in antigen/antibody binding), but 1000 times weaker in the micromolar range.  So [ Proc. Natl. Acad. Sci. vol. 115 pp. E7369 – E7378 ’18 ] cleverly added an extra few amino acids which they call molecular velcro, to boost the affinity x 10 (actually this decreases Kd tenfold).

One rationale for the weak binding is that it facilitates scanning by the TCR of  the pMHC  repertoire allowing the TCR to choose the best.  So they added the velcro, expecting the repertoire to be less diverse (since the binding was tighter).  It was just the same. Again the Kd didn’t seem to matter.

https://en.wikipedia.org/wiki/Catch_bond

Even more interesting, the first paper noted that productive TCR/pMHC bonds had catch bonds — e.g. bonds which get stronger the more you pull on them. The authors were actually able to measure the phenomenon. Catch bonds been shown to exist in a variety of systems (white cells sticking to blood vessel lining, bacterial adhesion), but their actual mechanism is still under debate.  The great thing about this paper (p. 682) is molecular dynamics simulation showed the conformational changes which occurred during catch bond formation in one case..   They even have videos.  Impressive.

This sort of thing is totally foreign to all solution chemistry, as there is no way to pull on a bond in solution.  Optical tweezers allow you to pull and stretch molecules (if you can attach them to large styrofoam balls).

The proteasome branches out

The surface of a protein is not at all like a ball of yarn, even though they are both one long string. This has profound implications for the immune system. Look at any solved protein structure. The backbone bobs and weaves taking water hating (hydrophobic) amino acids into the center of the protein, and putting water loving (hydrophilic) amino acids on the surface. So even though the peptide backbone is continuous, only discontinuous patches of it are displayed on the protein surface.

Which is a big problem for the immune system which wants to recognize the surface of the protein (which is all it first gets to see with an invading bug). Now we know that foreign proteins are ingested by the cell, chopped up by the proteasome, and fragments loaded on to immune molecules (class I Major Histocompatibility Complex antigens) and displayed on the cell surface so the immune system can learn what it looks like and react to it. The peptides aren’t very long — under 11 or so amino acids, but they are continuous.

What if the really distinct part of the protein surface (e.g. the immunogen)  is made of two distinct patches from the backbone? A fascinating paper shows how the immune system might still recognize it. Chop the protein up into fragments by the proteasome, and then have the fragments from adjacent patches put back together. You know that any enzyme can be run in reverse, so if the proteasome can split peptide bonds apart it can also join them together.

This is exactly what was found in a recent paper — Science vol. 354 pp. 354 – 358 ’16. The small peptides (containing at most 11 amino acids) finding their way to the cell surface were analyzed in a technical tour de force. In aggregate they go by the fancy name of immunopeptidome. They found that the proteasome IS actually splicing peptide fragments together. This is called Proteasome Catalyzed Peptide Splicing (PCPS). The present work shows that it accounts for 1/3 of the class I immunopeptidome in terms of diversity and 1/4 in terms of abundance. One-third of self antigens are represented on the cell surface of the immune cell line they studied (GR-LCL the GR-lymphoblastoid cell line) ONLY by spliced peptides. The ordering of the spliced peptide was the same as the parent protein in only half. There was no preference for the length of the protein skipped by the splice.

The work has huge implications for immunology, not least autoimmune disease.

So today I wrote the author the following

Dr. Mishto

Terrific paper ! Do you have any evidence for the spliced peptides being spatially contiguous on the surface of the parent protein. Have you looked?

This makes a lot of sense, because the immune system should ‘want’ to recognize protein conformations as they exist in the living cell, rather than stretches of amino acid sequence in the parent protein. Also, with few exceptions the surface of a given protein in vivo is a collection of discontinuous peptide sequences of the parent protein. I’ve always wondered how the immune system did this, and perhaps your paper explains things.

Luysii

and got this back almost immediately

Dear Luysii

Interesting idea. We shall have a look for few examples where the crystallography structure or the parental protein is disclosed already.

regards

Michele

It doesn’t get any better than this. Tomorrow I will be exactly 78 years and 6 months old. It shows I can still think (on occasion).

Addendum 17 Nov ’16;  It looks as though proteins are fed into the central cavity of the proteasome as a completely denatured single strand.  See figure 5 of PNAS 113 pp 12991 -m12996 ’16.  The channel to get in appears quite narrow.