MMD Masterclass, April 18th 2018


Learning by example:

How T cells learn to discriminate "self" from "foreign" during negative selection


Inge Wortel
inge.wortel@radboudumc.nl
Department of Tumor Immunology, Radboudumc

Back to the basics: T cells

Adapted from National Cancer Institute (NIH)



  • Adaptive immune system
  • Specific responses against pathogen-infected cells
  • T cell receptor (TCR) recognizes short peptides on MHC
  • Each cell in your T cell repertoire recognizes something else

How many different pathogenic peptides might your T cells encounter?

T cell diversity: a problem of numbers

The immune system's T cell repertoire has to:

  • recognize >1015 foreign peptides ... 1
  • ... while tolerating hundreds of thousands of self peptides


1 Sewell (2012). Nature Reviews Immunology.

T cell diversity: a problem of numbers

The immune system's T cell repertoire has to:

  • recognize >1015 foreign peptides ... 1
  • ... while tolerating hundreds of thousands of self peptides


Diversity: random TCR

Tolerance: negative selection

1 Sewell (2012). Nature Reviews Immunology.

How "imperfect" is negative selection?

Does negative selection achieve tolerance?

Healthy humans have many autoreactive T cells! 1

$\rightarrow$ Negative selection is incomplete!

1 Yu et al. (2015). Immunity.

What is the purpose of negative selection
if it is incomplete?

Approach: Artificial Immune System (AIS)

AISs:

  • Artificial intelligence
  • Inspired by the immune system

Use AIS to investigate the purpose of incomplete negative selection:

  • Make repertoires with millions of different TCRs
  • Simulate negative selection in silico
  • General classification algorithm

A classic classification problem: languages

Can you see the difference between English and Xhosa?


ceived   kuqale   erness
kubang   lwaban   rkness


ceived   kuqale   erness
kubang   lwaban   rkness

A simple model of TCR-sequence recognition

Three ingredients needed:

1 Sequence
2 TCR
3 Affinity

A simple model of TCR-sequence recognition

Three ingredients needed:

1 Sequence 6-letter strings of text in a certain language
2 TCR
3 Affinity

A simple model of TCR-sequence recognition

Three ingredients needed:

1 Sequence 6-letter strings of text in a certain language
2 TCR binding motif
3 Affinity

A simple model of TCR-sequence recognition

Three ingredients needed:

1 Sequence 6-letter strings of text in a certain language
2 TCR binding motif
3 Affinity Longest stretch of adjacent "matches"
$\rightarrow$ binding if affinity $\geq$ threshold t

Simulating negative selection in silico

Example: Xhosa recognition after negative selection on English

Negative selection allows discrimination

Negative selection on 500 English strings (t = 3),
Compare recognition of different English and Xhosa strings

Before:After:Best-recognized:

$\rightarrow$ TCRs distinguish strings they have not seen!

Why discrimination on unseen strings?

English and Xhosa are rarely recognized by the same TCRs:


Nodes:

  • English
  • Xhosa

Edge if >10,000 TCRs in common


Concordance (same-language neighbors): 81%



Discuss (5 min): Why does this explain discrimination after selection?

Why discrimination on unseen strings?

This only works if strings are truly different:


Nodes:

  • English
  • More English

Edge if >10,000 TCRs in common


Concordance (same-language neighbors): 50%

Two key requirements for self-foreign discrimination

  1. Appropriate TCR specificity/cross-reactivity
  2. Sufficient self-foreign dissimilarity

Two key requirements for self-foreign discrimination

  1. Appropriate TCR specificity/cross-reactivity
  2. Sufficient self-foreign dissimilarity
Complete specificity (t = 6)Low specificity (t = 1)Intermediate specificity (t = 3)

Two key requirements for self-foreign discrimination

  1. Appropriate TCR specificity/cross-reactivity
  2. Sufficient self-foreign dissimilarity

Summary: what have we learned from languages?


Negative selection allows self-foreign discrimination:

  • Even on "unseen" sequences
  • Even when tolerance is incomplete



This works well when:

  1. TCRs are moderately cross-reactive
  2. Self and foreign are sufficiently dissimilar

Back to self versus foreign peptides

Back to self versus foreign peptides

Use the same model for peptides instead of strings:




TCRs are moderately cross-reactive

Model at t = 4 closely matches experimental estimates


Model (t = 4) Literature
Cross-reactivity: 1 : 55,000 peptides 1 : 30,000 peptides 1
Typical peptide
recognition:
0 - 20 TCRs / million
0 - 100 TCRs / million 2-4



1 Ishizuka et al (2009). J Immunol.
2 Legoux et al (2010). J Immunol.   3 Blattman et al (2002). J Exp Med.   4 Alanio et al (2010). Blood.

How similar are peptides from "self" and "foreign" proteomes?

How similar are peptides from "self" and "foreign" proteomes?

Peptides from "self" and "foreign" proteomes are similar

Peptides from "self" and "foreign" proteomes are similar


 HIV peptides are embedded in large
 clusters of self peptides.









  Xhosa/English: separate clusters.

$\rightarrow$ self-foreign discrimination will be difficult!

Self-foreign discrimination is difficult

... but possible if TCRs are specific enough:

Can we make self-foreign discrimination easier?

  • self peptides in the thymus are probably not random
  • What can an "optimal" combination of self peptides achieve?

How could the thymus be "optimal"?

Computed "optimal" set is enriched in rare AAs, depleted of common AAs

$\rightarrow$ Could enrichment of rare AAs help self-foreign discrimination?

How could the thymus be "optimal"?

Choose "training" self peptides with a bias for peptides with rare AAs:

Simple bias already improves self-foreign discrimination.

The same is true for other pathogens



  • Poor self-foreign discrimination after selection on random peptides
  • Improved self-foreign discrimination after selection on biased peptides

Summary: self-foreign discrimination in the immune system

Negative selection remains useful...

  • Even incomplete negative selection can allow T cells
    to discriminate "self" from "foreign".
  • This reconciles textbook theory with
    surprising experimental data.

...but self-foreign discrimination is hard:

  • high self-foreign similarity hampers discrimination after selection
  • Prediction: we can overcome this with training peptides biased for rare AAs
    $\rightarrow$ because of AA bias in peptide presentation pathway?

Discussion



  • Systems immunology?
  • How can T cells learn?
  • Other questions?

Acknowledgements

Tumor Immunology, Nijmegen

  • Johannes Textor
  • Jolanda de Vries
  • Carl Figdor
  • Human DLM





Funding

  • Radboudumc PhD grant

Theoretical Biology, Utrecht

  • Rob de Boer
  • Can Kesmir


Physiology, McGill University Montreal

  • Judith Mandl


Theory: choosing self peptides for negative selection

When "self" and "foreign" are similar: