Functionally specialized human CD4+ T-cell subsets express physicochemically distinct TCRs

  1. Sofya A Kasatskaya
  2. Kristin Ladell
  3. Evgeniy S Egorov
  4. Kelly L Miners
  5. Alexey N Davydov
  6. Maria Metsger
  7. Dmitry B Staroverov
  8. Elena K Matveyshina
  9. Irina A Shagina
  10. Ilgar Z Mamedov
  11. Mark Izraelson
  12. Pavel V Shelyakin
  13. Olga V Britanova
  14. David A Price
  15. Dmitriy M Chudakov  Is a corresponding author
  1. Center of Life Sciences, Skolkovo Institute of Science and Technology, Russian Federation
  2. Genomics of Adaptive Immunity Department, Shemyakin-Ovchinnikov Institute of Bioorganic Chemistry, Russian Academy of Sciences, Russian Federation
  3. Division of Infection and Immunity, Cardiff University School of Medicine, United Kingdom
  4. Adaptive Immunity Group, Central European Institute of Technology, Czech Republic
  5. Institute of Translational Medicine, Center for Precision Genome Editing and Genetic Technologies for Biomedicine, Pirogov Russian National Research Medical University, Russian Federation
  6. Faculty of Bioengineering and Bioinformatics, Lomonosov Moscow State University, Russian Federation
  7. Systems Immunity Research Institute, Cardiff University School of Medicine, United Kingdom

Peer review process

This article was accepted for publication as part of eLife's original publishing model.

History

  1. Version of Record published
  2. Accepted Manuscript published
  3. Accepted
  4. Received

Decision letter

  1. Satyajit Rath
    Senior Editor; Indian Institute of Science Education and Research (IISER), India
  2. Armita Nourmohammad
    Reviewing Editor; University of Washington, United States
  3. Benjamin Chain
    Reviewer; University College London, United Kingdom
  4. Rob J de Boer
    Reviewer; Utrecht University, Netherlands

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This study uses T cell receptor sequencing to probe the structure of the functional T cell pool in healthy human blood. The key findings are (1) distinctive physicochemical, recombinational, and clonality characteristics for many of the repertoires, and (2) conserved and stereotyped patterns of clonal sharing between the subsets within donors and of public amino acid chains between donors. The results shed significant new light on how T cell dependent immunity is organized.

Decision letter after peer review:

Thank you for submitting your article "Origin and plasticity of human CD4+ T cell subsets tracked via TCRs" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Satyajit Rath as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Benjamin Chain (Reviewer #2); Rob J de Boer (Reviewer #5).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, we are asking editors to accept without delay manuscripts, like yours, that they judge can stand as eLife papers without additional data, even if they feel that they would make the manuscript stronger. Thus the revisions requested below only address clarity and presentation.

This study uses T cell receptor sequencing to probe the structure of the functional T cell pool in healthy human blood. In particular, the manuscript reports a characterization of TCR α and β repertoire features for eight effector/memory CD4+ T cell subsets defined by flow cytometry across 5 donors, and of several naive CD4 subsets across 12 donors. The key findings are (1) distinctive physicochemical, recombinational, and clonality characteristics for many of the repertoires, and (2) conserved and stereotyped patterns of clonal sharing between the subsets within the donors and of public amino acid chains between donors. All reviewers agree that the results shed significant new light on how T cell dependent immunity is organized. However, there are still some points that we would like to see addressed in the manuscript.

1) It would be beneficial to the readers if the authors could rewrite parts of the manuscript to make the story more coherent. Currently, the manuscript reads as very descriptive, and there is little by the way of hypothesis (except perhaps that Tregs are antigen-selected in the thymus more than are Teff). As a result, the study reads a bit like a series of disconnected observations, and it is hard to build up a clear unified message. Perhaps the authors could also add a table or cartoon of the novel classification suggested by the paper, in which the TCR properties and the repertoire overlaps are matched as much as possible. Currently, the first and the second part of the paper are disconnected, and if both parts are true, these results should be related and suggest a similar classification.

2) There is a lack of clarity about the statistical analysis of the differences between the populations which weakens the impact of the conclusions. For example, in Figure 1, it is difficult to get an indication of the extent of the variation that exists, and the biological amplitude of the effect. It is not clear if the parametric ANOVA is the right test here. It would be interesting to do a non-parametric test, based on shuffling of the repertoire labels, for example, and see the extent of variation observed by chance. It would be interesting also to see the results using unique sequences only (not weighted for frequency), perhaps a supplemental data. The magnitude of the effects is quite small – in the order of half a nucleotide length, for example. It would be useful to get a much better feel for the real variation in the population. Similar points apply to other figures.

3) The strikingly lower diversity of TH22 and Th2a in Figure 2 seems interesting. Could the authors provide a bit more detail on what is driving these changes? A few very large clones? A different clonal distribution? Or, fewer singlets?

4) With regard to the sequence characteristics that differ between subsets: it would be good to confirm that these are not due to differences in V/J gene usage, since the J gene in particular can contribute substantially to the CDR3 and the location of the 5 residue “central” window will overlap to varying degrees with this germline-derived sequence depending on the CDR3 length. It would also be good to rule out the possibility that there are recurrent, semi-invariant amino acid motifs present in the sequences, as opposed to generic sequence biases arising from physicochemical differences.

5) With regard to differential clonal dynamics: How can the authors rule out the possibility that the apparent differences in clonality arise from differences in mRNA expression of the TCR chains, leading to varying numbers of cDNA templates per cell?

6) The sharing data (e.g. Figure 3) is a central point of the paper, and is everywhere interpreted as evidence for plasticity, which is not necessarily true. Alternatively, the results could also mean that populations which share more sequences are derived form a common progenitor – in other words a lineage tree effect. It is not clear how the authors distinguish between these two possibilities, and a more detailed discussion would be helpful with this regards.

7) With regard to sharing, could the author discuss possibility for convergent recombination to lead to identical nucleotide sequences and hence apparent clone sharing, particularly for sequences that are close to germline. Also, are V/J genes included in the definition of "nucleotide clonotypes" or just the CDR3 sequence?

8) Figure 4 is very pretty but unfortunately, very opaque. The legend does not really explain at all what is being shown, or what it purports to show.

9) A general question: Are the TCR differences between the phenotypes large enough to classify a cell based upon its TCR into a phenotype? Probably not, which would mean that the results are (very interesting) trends.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Origin and plasticity of human CD4+ T cell subsets tracked via TCRs" for further consideration by eLife. Your revised article has been evaluated by Satyajit Rath (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

First, the evidence for plasticity is not convincing enough, and it is only one possible hypothesis, for which the manuscript does not provide a decisive and a strong support. Please change the title of the manuscript to reflect this.

Reviewer #2:

I raised a number a number of concerns in my first review, most of which I feel have not been addressed. In particular, I remain unconvinced that this paper has anything to do with plasticity. I still struggle to understand Figure 5. or what it adds to the narrative.

The title remains misleading – I don't think the paper says anything about either the origin or the plasticity of the cells.

However, I agree with the reviewer 5 comment "The fact that different CD4 T cell phenotypes tend to use different TCRs is a very innovative result ", and of general interest. This is an important result, even though the effect size is small, and worthy of publication. But it is not at all clear from the title that this is what the paper is about.

Reviewer #3:

My concerns have been addressed.

Reviewer #5:

Reading the revised manuscript convinced me again that this a truly interesting paper with several surprising results.

My only major recommendation would be to add a sentence saying that V and J usage are not different between the subsets.

https://doi.org/10.7554/eLife.57063.sa1

Author response

This study uses T cell receptor sequencing to probe the structure of the functional T cell pool in healthy human blood. In particular, the manuscript reports a characterization of TCR α and β repertoire features for eight effector/memory CD4+ T cell subsets defined by flow cytometry across 5 donors, and of several naive CD4 subsets across 12 donors. The key findings are (1) distinctive physicochemical, recombinational, and clonality characteristics for many of the repertoires, and (2) conserved and stereotyped patterns of clonal sharing between the subsets within the donors and of public amino acid chains between donors. All reviewers agree that the results shed significant new light on how T cell dependent immunity is organized. However, there are still some points that we would like to see addressed in the manuscript.

1) It would be beneficial to the readers if the authors could rewrite parts of the manuscript to make the story more coherent. Currently, the manuscript reads as very descriptive, and there is little by the way of hypothesis (except perhaps that Tregs are antigen-selected in the thymus more than are Teff). As a result, the study reads a bit like a series of disconnected observations, and it is hard to build up a clear unified message. Perhaps the authors could also add a table or cartoon of the novel classification suggested by the paper, in which the TCR properties and the repertoire overlaps are matched as much as possible. Currently, the first and the second part of the paper are disconnected, and if both parts are true, these results should be related and suggest a similar classification.

Thank you for the deep comment.

We worked through the whole manuscript to add our considerations where appropriate, and to some extent to better link the effector and naive parts of the manuscript.

We have also included a new section on experimental logic and workflow in the Results, which includes a graphical summary (new Figure 1).

However, it looks like we cannot currently build exact correlations between naive and effector subsets beyond Tregs and, to some extent, CXCR3-positive subsets. We also do not think that we could provide better classification and visualization compared with the current Figures 2, 4, and 5.

2) There is a lack of clarity about the statistical analysis of the differences between the populations which weakens the impact of the conclusions. For example, in Figure 1, it is difficult to get an indication of the extent of the variation that exists, and the biological amplitude of the effect. It is not clear if the parametric ANOVA is the right test here. It would be interesting to do a non-parametric test, based on shuffling of the repertoire labels, for example, and see the extent of variation observed by chance. It would be interesting also to see the results using unique sequences only (not weighted for frequency), perhaps a supplemental data. The magnitude of the effects is quite small – in the order of half a nucleotide length, for example. It would be useful to get a much better feel for the real variation in the population. Similar points apply to other figures.

Effects are relatively small, that’s correct, and so it is important that we observe the very same differences in unrelated healthy donors. To describe the level of deviation from the normal distribution, we built QQ plots for chosen physicochemical CDR3α/β characteristics, Author response image 1. Here we used the comparison to all samples as a post-hoc test instead of multiple pairwise comparisons among the subset groups. This analysis shows that the distribution is normal for most parameters, and thus ANOVA should be the right test.

Author response image 1

Non-parametric post-hoc test had little difference from the pairwise assessment of groups with a parametric post-hoc test.

Author response table 1
propertycolumn usedcompare tosubsetp adjusted (BH method)psigniftestp adjusted (BH method)psigniftest
cdr3_lengthzscore.all.Th25.1e-010.43258nsT-test0.6000.4481nsWilcoxon
cdr3_lengthzscore.all.Tfh2.4e-010.14003nsT-test0.3700.2333nsWilcoxon
cdr3_lengthzscore.all.Th2a7.4e-010.72099nsT-test0.7700.7726nsWilcoxon
cdr3_lengthzscore.all.Th14.3e-010.34963nsT-test0.7700.7451nsWilcoxon
cdr3_lengthzscore.all.Th1-179.2e-020.03070*T-test0.2500.0829nsWilcoxon
cdr3_lengthzscore.all.TREG4.3e-010.34053nsT-test0.6000.4481nsWilcoxon
cdr3_lengthzscore.all.Th177.5e-010.75446nsT-test0.6700.5633nsWilcoxon
cdr3_lengthzscore.all.Th221.1e-010.04145*T-test0.1200.0171*Wilcoxon
added_nucleotideszscore.all.Th23.1e-010.21194nsT-test0.3300.1935nsWilcoxon
added_nucleotideszscore.all.Tfh1.4e-010.07180nsT-test0.3300.1699nsWilcoxon
added_nucleotideszscore.all.Th2a3.7e-010.26285nsT-test0.3300.1935nsWilcoxon
added_nucleotideszscore.all.Th12.6e-010.16583nsT-test0.5400.3860nsWilcoxon
added_nucleotideszscore.all.Th1-172.2e-020.00365**T-test0.1200.0251*Wilcoxon
added_nucleotideszscore.all.TREG1.1e-010.04517*T-test0.3100.1485nsWilcoxon
added_nucleotideszscore.all.Th177.4e-010.70804nsT-test0.6700.5633nsWilcoxon
added_nucleotideszscore.all.Th227.6e-020.01900*T-test0.0730.0060**Wilcoxon
strengthzscore.all.Th21.1e-020.00135**T-test0.1600.0362*Wilcoxon
strengthzscore.all.Tfh1.8e-067.3e-08****T-test0.0490.0021**Wilcoxon
strengthzscore.all.Th2a4.3e-010.34108nsT-test0.4700.3118nsWilcoxon
strengthzscore.all.Th13.9e-010.28682nsT-test0.6100.4700nsWilcoxon
strengthzscore.all.Th1-172.6e-010.16136nsT-test0.3700.2333nsWilcoxon
strengthzscore.all.TREG1.5e-020.00219**T-test0.3100.1386nsWilcoxon
strengthzscore.all.Th171.0e-020.00104**T-test0.3100.1292nsWilcoxon
strengthzscore.all.Th221.0e-010.03680*T-test0.1900.0556nsWilcoxon
surfacezscore.all.Th29.2e-020.03016*T-test0.1200.0251*Wilcoxon
surfacezscore.all.Tfh2.0e-061.3e-07****T-test0.0490.0019**Wilcoxon
surfacezscore.all.Th2a2.4e-010.14789nsT-test0.3300.1935nsWilcoxon
surfacezscore.all.Th11.2e-010.05902nsT-test0.6700.5392nsWilcoxon
surfacezscore.all.Th1-175.5e-020.01270*T-test0.3100.1485nsWilcoxon
surfacezscore.all.TREG3.1e-010.21607nsT-test0.4300.2785nsWilcoxon
surfacezscore.all.Th171.1e-010.04722*T-test0.3300.1814nsWilcoxon
surfacezscore.all.Th221.6e-010.08850nsT-test0.1900.0603nsWilcoxon
volumezscore.all.Th25.5e-020.01168*T-test0.1200.0208*Wilcoxon
volumezscore.all.Tfh4.6e-030.00039***T-test0.1200.0229*Wilcoxon
volumezscore.all.Th2a7.8e-020.02114*T-test0.3100.1485nsWilcoxon
volumezscore.all.Th11.4e-010.06982nsT-test0.3100.1120nsWilcoxon
volumezscore.all.Th1-171.2e-010.05435nsT-test0.1900.0603nsWilcoxon
volumezscore.all.TREG5.3e-010.46185nsT-test0.6700.6131nsWilcoxon
volumezscore.all.Th172.3e-010.13201nsT-test0.4800.3294nsWilcoxon
volumezscore.all.Th223.8e-020.00721**T-test0.1200.0140*Wilcoxon
kf4zscore.all.Th26.3e-010.56570nsT-test0.7700.7726nsWilcoxon
kf4zscore.all.Tfh3.5e-077.3e-09****T-test0.0490.0031**Wilcoxon
kf4zscore.all.Th2a6.3e-010.57609nsT-test0.6800.6387nsWilcoxon
kf4zscore.all.Th15.0e-010.41381nsT-test0.6700.6131nsWilcoxon
kf4zscore.all.Th1-174.3e-010.32935nsT-test0.6700.5879nsWilcoxon
kf4zscore.all.TREG9.2e-020.02848*T-test0.1700.0431*Wilcoxon
kf4zscore.all.Th171.2e-010.05305nsT-test0.3100.1386nsWilcoxon
kf4zscore.all.Th227.1e-010.66248nsT-test0.6700.5879nsWilcoxon

3) The strikingly lower diversity of TH22 and Th2a in Figure 2 seems interesting. Could the authors provide a bit more detail on what is driving these changes? A few very large clones? A different clonal distribution? Or, fewer singlets?

As indicated in the text: “Prominent clonal expansions, reflected by low normalized Shannon-Wiener indices, were apparent in the Th22 and Th2a subsets, indicating focused antigen-specific proliferation.”

4) With regard to the sequence characteristics that differ between subsets: it would be good to confirm that these are not due to differences in V/J gene usage, since the J gene in particular can contribute substantially to the CDR3 and the location of the 5 residue “central” window will overlap to varying degrees with this germline-derived sequence depending on the CDR3 length. It would also be good to rule out the possibility that there are recurrent, semi-invariant amino acid motifs present in the sequences, as opposed to generic sequence biases arising from physicochemical differences.

Several approaches could be utilized to estimate the contribution of recurrent or semi-invariant sequences/motifs. For one, the prevalence of semi-invariant sequences could be reflected in the analysis of CDR3 length and N insertions, as shorter sequences appear in repertoires with higher frequencies. The distribution of N inserted nucleotides is shown in Figure 2A. Another approach could be to search for functional invariant sequences. We performed a search for classical TRA CDR3 described previously for human iNKT and MAIT cells:

iNKT:

CVVIDRGSTLGRLYF, CVVSDRGSTLGRLYF

MAIT:

CAVKDSNYQLIF, CAGMDSNYQLIF, CASIDSNYQLIF, CAAMDSNYQLIF, CAAEDSNYQLIF, CAVVDSNYQLIF, CAVRDSNYQLIF, CAVMDSSYKLIF, CAVMDSSYKLIF, CAVMDSSYKLIF, CAVRDGDYKLSF, CAVSDSNYQLIF, CAVMDSNYQLIF, CAFMDSNYQLIF

Cumulative frequencies of such invariant TRA CDR3s in repertoires were 0.17% for MAIT and 0.14% for the iNKT cells. There were no significant differences in frequencies among subsets. Therefore, this analysis did not reveal any substantial bias in invariant TCRs distribution between functional subsets.

Also, we have initially spent substantial time trying to find any dependencies in V and J usage between the subsets, and have not found any. In general, V and J usages are distributed randomly across the subsets, so this should not be a prominent contributing factor:

Author response image 2

5) With regard to differential clonal dynamics: How can the authors rule out the possibility that the apparent differences in clonality arise from differences in mRNA expression of the TCR chains, leading to varying numbers of cDNA templates per cell?

Thank you for the comment. We cannot formally completely exclude this possibility, but it seems unlikely, given that each subset was sorted rigorously on defined phenotypic parameters.

Hypothetically, distinct TCR mRNA expression levels between clones could to some extent influence the apparent differences in observed clonality, which remains beyond the scopes of the current work. However, in reality, all our experience with UMI-based TCR profiling shows that such influence should be negligible.

We would prefer not to overload the manuscript with these considerations, but ready to add the appropriate comment to the manuscript if Editor considers it appropriate.

6) The sharing data (e.g. Figure 3) is a central point of the paper, and is everywhere interpreted as evidence for plasticity, which is not necessarily true. Alternatively, the results could also mean that populations which share more sequences are derived form a common progenitor – in other words a lineage tree effect. It is not clear how the authors distinguish between these two possibilities, and a more detailed discussion would be helpful with this regards.

Formally, we cannot distinguish between these alternative scenarios in our analysis. However, we believe that sharing between the top-2000 most frequent clonotypes is much more likely explained by the current plasticity. Long-term evolution from a common progenitor would most probably result in prominent clonal expansions observed in a particular subset, so that we would not sample the same clone among dominating in two distinct subsets.

As well as with previous question, we are ready to add the appropriate comment to the manuscript if Editor considers it appropriate.

7) With regard to sharing, could the author discuss possibility for convergent recombination to lead to identical nucleotide sequences and hence apparent clone sharing, particularly for sequences that are close to germline. Also, are V/J genes included in the definition of "nucleotide clonotypes" or just the CDR3 sequence?

V-genes were included in the analysis. We have added a note to the legend of Figure 4. We cannot completely exclude the possibility of convergent recombination here. However, its input should be negligible, especially in the top-2000 clonotypes. Please note that increased plasticity is observed predominantly between the subsets with long CDR3s and high count of added N nucleotides (Th22-Th2-Th2a-Th17). i.e. this is not about convergence.

8) Figure 4 is very pretty but unfortunately, very opaque. The legend does not really explain at all what is being shown, or what it purports to show.

Thank you. We do have the high resolution figures and will definitely take care of the quality during the final stages of manuscript preparation together with the Editors. We have also worked on the figure legend to make it more informative.

9) A general question: Are the TCR differences between the phenotypes large enough to classify a cell based upon its TCR into a phenotype? Probably not, which would mean that the results are (very interesting) trends.

No, of course not, not only because the differences are small, but also because it is a clone-specific story. This only works as average characteristics of the subset-specific repertoires, likely reflecting other factors that determine fate decisions, including the context of antigen presentation, as highlighted in the Introduction.

As we indicate now in the text (first chapter of Results):

“It should be noted that the above characteristics were observed for the averaged, cumulative portrait of many TCR variants representing each subset. Each particular T cell with low number of N-added nucleotides and strongly interacting aminoacids in the middle of short CDR3 may belong to any Th subset, but more probably to Tfh, Th1, or Th1-17.”

https://doi.org/10.7554/eLife.57063.sa2

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Sofya A Kasatskaya
  2. Kristin Ladell
  3. Evgeniy S Egorov
  4. Kelly L Miners
  5. Alexey N Davydov
  6. Maria Metsger
  7. Dmitry B Staroverov
  8. Elena K Matveyshina
  9. Irina A Shagina
  10. Ilgar Z Mamedov
  11. Mark Izraelson
  12. Pavel V Shelyakin
  13. Olga V Britanova
  14. David A Price
  15. Dmitriy M Chudakov
(2020)
Functionally specialized human CD4+ T-cell subsets express physicochemically distinct TCRs
eLife 9:e57063.
https://doi.org/10.7554/eLife.57063

Share this article

https://doi.org/10.7554/eLife.57063