TIM, 10-03-2020

Is T cell negative selection a
learning algorithm?

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

T cell diversity & negative selection: 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 & negative selection: 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.

Let's vote:

Does negative selection achieve tolerance?

Healthy humans have many self-reactive T cells! ¹

$\rightarrow$ Negative selection is incomplete!

¹ Yu et al. (2015). Immunity.

Tolerance and discrimination are not the same thing

Tolerance and discrimination are not the same thing

Which of these scenarios holds?

Can you see the difference between English and Xhosa?

Which of these scenarios holds?

Which of these scenarios holds?

Approach: Artificial Immune System (AIS)

A simple model of TCR-sequence recognition

Three ingredients needed:

A simple model of TCR-sequence recognition

Three ingredients needed:

A simple model of TCR-sequence recognition

Three ingredients needed:

A simple model of TCR-sequence recognition

Three ingredients needed:

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

Motifs per string, before vs after:	Most frequently recognized:

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

Why discrimination on unseen strings?

Motifs rarely react to both English and Xhosa:

Nodes:

• English
• Xhosa

Edge if >10,000 motifs in common

Concordance (same-language neighbors): 81%

Why discrimination on unseen strings?

This only works if strings are truly different:

Nodes:

• English
• More English

Edge if >10,000 motifs in common

Concordance (same-language neighbors): 50%

Two key requirements for self-foreign discrimination

Two key requirements for self-foreign discrimination

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

Summary: what have we learned from languages?

Proof of concept:
Negative selection can foster "learning by example".

• Self-foreign discrimination even on "unseen" sequences
• Even when tolerance is incomplete

But this works only if:

• self and foreign are sufficiently dissimilar
• cross-reactivity is intermediate

Back to self versus foreign peptides

Back to self versus foreign peptides

Use the same model for peptides instead of strings:

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

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

 HIV peptides are embedded in
 clusters of self peptides.

  Xhosa/English: separate clusters.

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

Self-foreign discrimination is difficult

... but remains possible:

What if thymic self peptides are non-random?

• random self peptides are often similar and will delete the same TCRs
  $\rightarrow$ this is inefficient
• some peptides are exchangeable, others are not
  $\rightarrow$ a training set with more non-exchangeable peptides might do better:

"Optimal" training peptides improve self-foreign discrimination

Removal of self-reactivity $\neq$ self-foreign discrimination! Why does this work?

Why better self-foreign discrimination?

Self peptides with low exchangeability less often resemble foreign peptides:

$\rightarrow$ non-exchangeable peptides efficiently remove self-reactive TCRs, but preserve foreign-reactive ones!

How could the thymus be "optimal"?

Computed "optimal" set is enriched in rare AAs, depleted of common AAs.
Peptides with rare AAs tend to be less exchangeable:

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

How could the thymus be "optimal"?

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

Self-foreign dissimilarity is no longer necessary

Randomly split the self peptides into "self" and "foreign":

$\rightarrow$ T cells can discriminate, no matter how much a virus resembles self!

Summary: self-foreign discrimination in the immune system

Negative selection allows "learning by example"

• Even incomplete negative selection can allow T cells
  to discriminate "self" from "foreign" in principle.
• This reconciles textbook theory with
  surprising experimental data.
• ...And it shows that our brain is not the only
  organ that can learn!

...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?

Acknowledgements

Theoretical Biology, Utrecht

• Rob de Boer
• Can Keşmir

Physiology, McGill University Montreal

• Judith Mandl