TCR Sequencing Lack of Repertoire Overlap

Question

[bioinformatics.cancer]

Hi, I have analyzed a few T Cell Receptor (beta chain) repertoire sequencing data sets from cancer samples and observed the lack of overlap in Clones (AA sequences) within same treatment groups. This was true even for study with isogenic mice in which a group of mice received the same treatment yet there was no overlap seen in the expanded clones. Part of the reason could be that only the beta chain was sequenced, but it is still surprising not to see any overlap at all. Is this common and are there any studies/papers that discuss this in some details. Thank you, - Pankaj

[colonppg]

Actually this is not a surprise, and should be expected... majority of TCR sequence I have seen are very rare sequences

I suspect this is going to be true for two samples taking simutaneously from the same person... TCR repertoire is our wish, we can only see part of it by sampling

Answer 1

[Charles Plessy]

What do you mean by _a few_? TCR repertoires are very diverse, and if one analyses less than a hundred sequences, it would not be surprising that all the clones are observed only once.

[bioinformatics.cancer]

By few I meant a few different studies/experiments ("repertoire sequencing data sets"), not a few clones/sequences within repertoires. In none of them I find overlap in expanded clones among the repertoire of individual samples within the same treatment group. For example, in one study we have isogenic mice with tumor transplants. One set of mice are control (no treatment) and another set of mice in the treatment group were treated with PD1 inhibitors. When I look for overlapping (common) clones within the repertoire of mice in the same group, for example, the treatment group, even for mice who responded to treatment, I don't see common AA sequences. Sorry for the confusion, hope this clarifies the question.

[Charles Plessy]

Sorry that I misread! But still, a relatively low number of sequence reads per dataset could be the explanation. Or maybe I am misunderstanding you and by _"lack of overlap"_ you do not mean _"no overlap"_ ?

[bioinformatics.cancer]

Yes, in most cases no overlap at all. What I mean by that is you would think that the same clones (TCR CDR3 AA sequence) would expand in some (or all) the mice in the same treatment groups since they are isogenic and were treated the same way. So if you plotted the individual clones of two mice/samples in the same treatment group, one on the X axis and one on the Y axis, you would expect some clones to lie close to the 45 degree line, but all the expanded clones are close to either the X axis or the Y axis.

[Charles Plessy]

Since different clones may have the same target protein and therefore accomplish the same function, this is not necessarily surprising.

[bioinformatics.cancer]

Are you aware of any publication that discusses this issue? I fully agree with you and it is the same feedback I have received from others that I have talked to. I believe one of the motivations for doing paired sequencing (alpha and beta chains) is to identify the clones with more specificity, although if the beta chain sequence don't overlap and I am not sure why the combined sequence of alpha and beta would. One suggestion I have received is to cluster the AA sequences based on homology or some distance scoring method, but I have not pursued that yet. We are writing a paper on this analysis and wanted a citation to provide credibility to these observations.
Thanks.

[Charles Plessy]

No sorry, I have not been very active in the field recently and my own results are not yet published. Years ago, people would start by analysing the frequency of V-J combinations. At that level, you may have more chance to observe consistency within groups. While I have not used other tools than my own, I see that packages such as tcR seem to offer a framework for V-J analysis.

Answer 2

[mikhail.shugay]

Mice have a far less diverse repertoire than human, so some degree of overlap could be expected, while it strongly depends on the sampling depth (what is the number of clonotypes/cells analyzed?).

Note that there are two common pitfalls here:

• false-negatives, e.g. the following TCR beta sequences correspond to the same peptide in same MHC context

  • CASSVAVSAGTYEQYF and CASSAGTSGEYEQYF (the antigen is SGEGSFQPSQENP)
  • CASTYHGTGYF and CASSQIRETQYF (the antigen is GELIGILNAAKVPAD)

• false-positives, e.g. cross-sample contamination and public clonotypes (see discussion here https://bmcgenomics.biomedcentral.com/articles/10.1186/s12864-016-2799-7)

In this case you have to move to other ways of sample analysis. For example in this (http://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1004503) multiple sclerosis TCR repertoire study no public clonotypes were found, yet a number of remarkable differences was found for other repertoire characteristics.

Answer 3

[Obi Griffith]

Another formal possibility is that your mice are not as isogenic as you think. I have had several experiences where we were operating on the assumption that highly inbred lab mouse strains are "genetically identical" but in fact upon whole genome sequencing determine that substantial genetic diversity remains. A simple illustration of this is to use unmatched normals for somatic variant prediction (a common practice in early mouse sequencing studies based on the isogenic assumption). We have found several orders of magnitude more apparent somatic variants are predicted when comparing a mouse tumor genome to an unmatched mouse of the same strain than to a matched normal sample from the same mouse (dozens of coding mutations to thousands of coding mutations). I don't know much about mouse breeding but I imagine that the differences in source/batch/litter will effect how isogenic your mice really are. Given that TCRs produce sequence diversity by design, maybe even relatively small differences in genetic sequence could result in large differences in clonotype diversity/representation? I'm not an immunologist and so please take this speculation with a grain of salt.