Despite the resounding clinical success in cancer treatment of antibodies that block the interaction of PD1 with its ligand PDL11, the mechanisms involved remain unknown. A major limitation to understanding the origin and fate of T cells in tumour immunity is the lack of quantitative information on the distribution of individual clonotypes of T cells in patients with cancer. Here, by performing deep single-cell sequencing of RNA and T cell receptors in patients with different types of cancer, we survey the profiles of various populations of T cells and T cell receptors in tumours, normal adjacent tissue, and peripheral blood. We find clear evidence of clonotypic expansion of effector-like T cells not only within the tumour but also in normal adjacent tissue. Patients with gene signatures of such clonotypic expansion respond best to anti-PDL1 therapy. Notably, expanded clonotypes found in the tumour and normal adjacent tissue can also typically be detected in peripheral blood, which suggests a convenient approach to patient identification. Analyses of our data together with several external datasets suggest that intratumoural T cells, especially in responsive patients, are replenished with fresh, non-exhausted replacement cells from sites outside the tumour, suggesting continued activity of the cancer immunity cycle in these patients, the acceleration of which may be associated with clinical response.
This is a preview of subscription content, access via your institution
Open Access articles citing this article.
Association of peripheral basophils with tumor M2 macrophage infiltration and outcomes of the anti-PD-1 inhibitor plus chemotherapy combination in advanced gastric cancer
Journal of Translational Medicine Open Access 04 September 2022
Signal Transduction and Targeted Therapy Open Access 03 May 2022
Single-cell characterization of leukemic and non-leukemic immune repertoires in CD8+ T-cell large granular lymphocytic leukemia
Nature Communications Open Access 11 April 2022
Subscribe to Nature+
Get immediate online access to Nature and 55 other Nature journal
Subscribe to Journal
Get full journal access for 1 year
only $3.90 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Tax calculation will be finalised during checkout.
Get time limited or full article access on ReadCube.
All prices are NET prices.
FASTQ files containing raw reads from the scRNA-seq and scTCR-seq analyses have been deposited with the European Genome-phenome Archive (EGA) under studies EGAS00001003993 and EGAS00001003994, and datasets EGAD00001005464 and EGAD00001005465. These files are available under controlled access upon request to the Data Access Committee, with contact information provided at EGA (https://www.ebi.ac.uk/ega/home). Processed output files from Cell Ranger, integrated assay results from Seurat, and metadata with UMAP coordinates, cluster assignments, and clonotypes are available from the NCBI GEO under accession GSE139555.
Computer code used to generate the analyses and figures in this paper are provided as a as a Supplementary File to the NCBI Gene Expression Omnibus (GEO) accession GSE139555.
Mellman, I., Coukos, G. & Dranoff, G. Cancer immunotherapy comes of age. Nature 480, 480–489 (2011).
Shulman, Z. et al. Transendothelial migration of lymphocytes mediated by intraendothelial vesicle stores rather than by extracellular chemokine depots. Nat. Immunol. 13, 67–76 (2012).
Guo, X. et al. Global characterization of T cells in non-small-cell lung cancer by single-cell sequencing. Nat. Med. 24, 978–985 (2018).
Zhang, L. et al. Lineage tracking reveals dynamic relationships of T cells in colorectal cancer. Nature 564, 268–272 (2018).
Yost, K. E. et al. Clonal replacement of tumor-specific T cells following PD-1 blockade. Nat. Med. 25, 1251–1259 (2019).
Schenkel, J. M. & Masopust, D. Tissue-resident memory T cells. Immunity 41, 886–897 (2014).
Mackay, L. K. et al. Hobit and Blimp1 instruct a universal transcriptional program of tissue residency in lymphocytes. Science 352, 459–463 (2016).
Kumar, B. V. et al. Human tissue-resident memory T cells are defined by core transcriptional and functional signatures in lymphoid and mucosal sites. Cell Rep. 20, 2921–2934 (2017).
Im, S. J. et al. Defining CD8+ T cells that provide the proliferative burst after PD-1 therapy. Nature 537, 417–421 (2016).
Miller, B. C. et al. Subsets of exhausted CD8+ T cells differentially mediate tumor control and respond to checkpoint blockade. Nat. Immunol. 20, 326–336 (2019).
Balin, S. J. et al. Human antimicrobial cytotoxic T lymphocytes, defined by NK receptors and antimicrobial proteins, kill intracellular bacteria. Sci. Immunol. 3, eaat7668 (2018).
Thommen, D. S. et al. A transcriptionally and functionally distinct PD-1+ CD8+ T cell pool with predictive potential in non-small-cell lung cancer treated with PD-1 blockade. Nat. Med. 24, 994–1004 (2018).
Sun, Q., Hao, Q. & Prasanth, K. V. Nuclear long noncoding RNAs: key regulators of gene expression. Trends Genet. 34, 142–157 (2018).
Delmas, V., Stokes, D. G. & Perry, R. P. A mammalian DNA-binding protein that contains a chromodomain and an SNF2/SWI2-like helicase domain. Proc. Natl Acad. Sci. USA 90, 2414–2418 (1993).
Gaide, O. et al. Common clonal origin of central and resident memory T cells following skin immunization. Nat. Med. 21, 647–653 (2015).
Simoni, Y. et al. Bystander CD8+ T cells are abundant and phenotypically distinct in human tumour infiltrates. Nature 557, 575–579 (2018).
Scheper, W. et al. Low and variable tumor reactivity of the intratumoral TCR repertoire in human cancers. Nat. Med. 25, 89–94 (2019).
Mariathasan, S. et al. TGFβ attenuates tumour response to PD-L1 blockade by contributing to exclusion of T cells. Nature 554, 544–548 (2018).
Fehrenbacher, L. et al. Atezolizumab versus docetaxel for patients with previously treated non-small-cell lung cancer (POPLAR): a multicentre, open-label, phase 2 randomised controlled trial. Lancet 387, 1837–1846 (2016).
McDermott, D. F. et al. Clinical activity and molecular correlates of response to atezolizumab alone or in combination with bevacizumab versus sunitinib in renal cell carcinoma. Nat. Med. 24, 749–757 (2018).
Tumeh, P. C. et al. PD-1 blockade induces responses by inhibiting adaptive immune resistance. Nature 515, 568–571 (2014).
Araujo, J. M. et al. Effect of CCL5 expression in the recruitment of immune cells in triple negative breast cancer. Sci. Rep. 8, 4899 (2018).
Chen, D. S. & Mellman, I. Elements of cancer immunity and the cancer-immune set point. Nature 541, 321–330 (2017).
Wherry, E. J. & Kurachi, M. Molecular and cellular insights into T cell exhaustion. Nat. Rev. Immunol. 15, 486–499 (2015).
Topalian, S. L., Drake, C. G. & Pardoll, D. M. Immune checkpoint blockade: a common denominator approach to cancer therapy. Cancer Cell 27, 450–461 (2015).
Khan, O. et al. TOX transcriptionally and epigenetically programs CD8+ T cell exhaustion. Nature 571, 211–218 (2019).
Scott, A. C. et al. TOX is a critical regulator of tumour-specific T cell differentiation. Nature 571, 270–274 (2019).
Sade-Feldman, M. et al. Defining T cell states associated with response to checkpoint immunotherapy in melanoma. Cell 175, 998–1013 (2018).
Yan, Y. et al. CX3CR1 identifies PD-1 therapy-responsive CD8+ T cells that withstand chemotherapy during cancer chemoimmunotherapy. JCI Insight 3, e97828 (2018).
Schumacher, T. N. & Scheper, W. A liquid biopsy for cancer immunotherapy. Nat. Med. 22, 340–341 (2016).
Hogan, S. A. et al. Peripheral blood TCR repertoire profiling may facilitate patient stratification for immunotherapy against melanoma. Cancer Immunol. Res. 7, 77–85 (2019).
Zheng, G. X. Y. et al. Massively parallel digital transcriptional profiling of single cells. Nat. Commun. 8, 14049 (2017).
Stuart, T. et al. Comprehensive integration of single-cell data. Cell 177, 1888–1902 (2019).
Aran, D. et al. Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage. Nat. Immunol. 20, 163–172 (2019).
Del Carratore, F. et al. RankProd 2.0: a refactored bioconductor package for detecting differentially expressed features in molecular profiling datasets. Bioinformatics 33, 2774–2775 (2017).
Ritchie, M. E. et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 43, e47 (2015).
Diedenhofen, B. & Musch, J. cocor: a comprehensive solution for the statistical comparison of correlations. PLoS One 10, e0121945 (2015).
Chao, A. et al. Rarefaction and extrapolation with Hill numbers: a framework for sampling and estimation in species diversity studies. Ecol. Monogr. 84, 45–67 (2014).
Therneau, T. M. & Grambsch, P. M. Modeling Survival Data: Extending the Cox Model (Springer, 2000).
McInnes, L. & Healy, J. UMAP: uniform manifold approximation and projection for dimension reduction. Preprint at https://arXiv.org/abs/1802.03426 (2018).
Becht, E. et al. Dimensionality reduction for visualizing single-cell data using UMAP. Nat. Biotechnol. 37, 38–44 (2019).
Zeileis, A. et al. colorspace: a toolbox for manipulating and assessing colors and palettes. Preprint at https://arXiv.org/abs/1903.06490 (2019).
Barter, R. L. & Yu, B. Superheat: an R package for creating beautiful and extendable heatmaps for visualizing complex data. J. Comput. Graph. Stat. 27, 910–922 (2018).
Morgan, M. T. & Davis, S. R. GenomicDataCommons: a bioconductor interface to the NCI Genomic Data Commons. Preprint at https://www.bioRxiv.org/content/10.1101/117200v1 (2017).
Davis, S. & Meltzer, P. S. GEOquery: a bridge between the Gene Expression Omnibus (GEO) and BioConductor. Bioinformatics 23, 1846–1847 (2007).
Shimizu, Y., Meunier, L. & Hendershot, L. M. pERp1 is significantly up-regulated during plasma cell differentiation and contributes to the oxidative folding of immunoglobulin. Proc. Natl Acad. Sci. USA 106, 17013–17018 (2009).
Andreani, V. et al. Cochaperone Mzb1 is a key effector of Blimp1 in plasma cell differentiation and β1-integrin function. Proc. Natl Acad. Sci. USA 115, E9630–E9639 (2018).
Beham, A. W. et al. A TNF-regulated recombinatorial macrophage immune receptor implicated in granuloma formation in tuberculosis. PLoS Pathog. 7, e1002375 (2011).
Fuchs, T. et al. Expression of combinatorial immunoglobulins in macrophages in the tumor microenvironment. PLoS One 13, e0204108 (2018).
Thul, P. J. et al. A subcellular map of the human proteome. Science 356, 820 (2017).
Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl Acad. Sci. USA 102, 15545–15550 (2005).
Liberzon, A. et al. Molecular signatures database (MSigDB) 3.0. Bioinformatics 27, 1739–1740 (2011).
Irizarry, R. A., Wang, C., Zhou, Y. & Speed, T. P. Gene set enrichment analysis made simple. Stat. Methods Med. Res. 18, 565–575 (2009).
Shugay, M. et al. VDJdb: a curated database of T-cell receptor sequences with known antigen specificity. Nucleic Acids Res. 46, D419–D427 (2018).
Li, B. et al. Landscape of tumor-infiltrating T cell repertoire of human cancers. Nat. Genet. 48, 725–732 (2016).
Shan, G. & Gerstenberger, S. Fisher’s exact approach for post hoc analysis of a chi-squared test. PLoS One 12, e0188709 (2017).
We thank the Genentech FACS Core for supporting the prompt sorting of T cells, and S. Kummerfeld for advice and assistance with initial data analyses. We also thank A. Lun for discussions regarding reference gene signatures and labelling of single cells.
All authors are employees of Genentech, which develops and markets drugs for profit.
Peer review information Nature thanks Xiang Chen, Xiaole Shirley Liu and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data figures and tables
a, Criteria for separation. Clusters from the overall cluster analysis of all immune cells assayed by scRNA-seq from the 14 patients in this study. Clusters are plotted according to the fraction of cells with a TCR clonotype and the mean expression of the T cell marker CD3E from scRNA-seq. Clusters with high values on both metrics were considered to represent T cells, whereas those with low values on both represent non-T cells. Intermediate values warranted further consideration and are annotated with the assigned cluster label for further reference. The eventual division of clusters into T and non-T cells is represented by the colours red and blue, respectively. b, Isolation of T cells. Immune cells are plotted as dots, positioned by the UMAP dimensionality reduction of gene expression. Cells from non-T clusters (blue in a) are in black, and cells from T clusters (red in a) are coloured by their subsequent cluster shown in d. Numbers in brackets indicate T cell clusters annotated in a. c, Isolation of non-T cells. The UMAP plot of b for immune cells is coloured for cells from non-T clusters based on their subsequent cluster, as shown in e and g. d, Origin of T cells. Heat map shows the cross-labelling of T cells by their original cluster assignment in the combined analysis of a (rows) and their subsequent cluster analysis of T cells separately (columns). Intensities indicate the normalized frequency by column. The subsequent cluster is assigned a distinct colour, shown in the row labelled ‘new’, matching the schema in Fig. 2a. e, Origin of non-T cells. Cross-labelling as in d but for non-T cells. Colours in the row labelled ‘new’ match the schema in h. f, Mixing of T cells across patients. T cells are mapped by the UMAP algorithm in a subsequent analysis of the T cell division, and are coloured by the patient of origin. Patients are observed to be well mixed across the map, indicating adequate integration of the individual samples. g, Clonotype fractions for T cell clusters. Bar plot shows the fraction of cells with a TCR clonotype for each T cell cluster, coloured according to the schema in d. h, UMAP plot of non-T cells. Non-T cells from CD45-selected samples are mapped by the UMAP algorithm in a subsequent analysis of the non-T cell division, and are coloured by their cluster assigned by cluster analysis, using the schema in e. i, Sample statistics. Statistics are provided for the cells in our dataset after separation into T cell and non-T cell categories. Patients are labelled by their cancer type: non-small-cell lung adenocarcinoma (lung), endometrial adenocarcinoma (endo), colorectal adenocarcinoma (colon), and renal clear cell carcinoma (renal). Each patient is annotated by whether the cells were selected by CD3 or CD45, and statistics are given separately for tumour (T), normal adjacent tissue (N), and peripheral blood (B) samples. Numbers indicate the total counts of transcripts and cells from scRNA-seq, as well as the count of cells with clonotypes from scTCR-seq in the column labelled ‘typed’. Cells were grouped into distinct clonotypes, and the count of distinct clonotypes is shown for each patient in the column labelled ‘clones’.
a, TCR sharing across compartments. Venn diagram for each patient shows sharing of TCR clonotypes across compartments. Values indicate the numbers of distinct clonotypes unique to each compartment or shared among compartments in the overlapping oval regions. b, Distribution of clones by tissue expansion patterns. Bar plot for each patient shows the fraction of clones having each tissue expansion pattern. c, Distribution of cells by tissue expansion pattern. Bar plots for each patient show the distribution of cells in the NAT (left) and tumour (right) compartments according to the tissue expansion pattern of their parent clone. d, Clonal expansion in tissue. Scatter plot for each patient shows each distinct clonotype as a dot, with coordinates indicating normalized clone size, or cell fraction, in the NAT and tumour compartments. Dots are coloured by a two-dimensional palette in the bottom right, in which blue shades intensify with increasing NAT clone size, pink shades intensify with increasing tumour clone size, and purple shades intensify with increasing clone sizes in both compartments. NAT and tumour singletons are indicated by yellow and orange, respectively. Vertical and horizontal grey lines in each scatter plot indicate divisions between absence (clone size of 0 cells) and presence (clone size of 1 or more cells). Diagonal grey lines indicate equal cell fractions in the two compartments. Numerical values in each title indicate the count of distinct clonotypes in each patient. Two-sided P values are from a Pearson’s correlation coefficient r on log-transformed clone sizes from the dual-expanded clones (nD). NA indicates that statistics could not be computed for two or fewer clones. Patients are ordered by decreasing values of r. e, TCR sharing by T cells across compartments. Each patient in a is represented by a Venn diagram, following the schema in Fig. 1b. Numbers within the Venn diagram regions represent counts of T cells by the tissue and blood expansion patterns of their parent clone. Numbers to the right of each diagram indicate the total number of cells from blood non-expanded and expanded clones, used for computing peripheral clonal expansion in Fig. 1c.
The 14 patients with non-small cell lung adenocarcinoma in the dataset from Guo et al.3 are each depicted by a set of plots, as in Fig. 1a, c–e. Scatter plots of distinct clonotypes are shown, plotted by cell fractions in NAT and tumour, with random jitter added to distinguish points. Clone size in blood is indicated by dot size, and clones are coloured by the two-dimensional palette for tissue expansion pattern. Vertical and horizontal lines separate the absence and presence of clones within compartments. Diagonal lines indicate equal cell fractions in tumour and NAT. Numerical values denote the extent of parallel dual expansion, measured by a Pearson’s correlation coefficient, weighted (rw) by (1 + blood clone size), on the dual-expanded clones (nD). Underneath the scatter plots, bar plots for the corresponding patient show the extent of peripheral clonal expansion (top), used to order patients, as well as infiltration into tissue expansion patterns by blood-independent, non-expanded and expanded clones (middle). P values are shown from a chi-square test on counts of cells from tumour or NAT (tissue-resident). Additional bar plots (bottom) show the fractions of tissue-resident cells with clonotypes observed in a blood-expanded clone for each tissue expansion pattern. Two patients (single asterisk) had no cells collected from NAT in the original dataset. In addition, patients P0616P and P0616A (double asterisks) each had only a single dual-expanded clone, so a correlation coefficient could not be computed. The remaining ten patients are summarized in Fig. 1i to show the relationship between peripheral clonal expansion and parallel dual expansion in tissue.
a, Characterizing clusters of T cell clusters with reference gene signatures. Heat maps show cross-labelling of T cell clusters (columns) to reference gene signatures (rows), taken from the analyses in Guo et al.3, Zhang et al.4 and Yost et al.5, with intensities indicating normalized frequency. CD8 and CD4 clusters from Guo et al.3 and Zhang et al.4 are separated by an extra space to aid visualization. b, Expression of selected genes. Box plots show distributions of gene expression on all T cells in the dataset, with cells grouped by their clusters, coloured as in Fig. 2a. Tops and bottoms of boxes indicate interquartile ranges, and lines within boxes indicate medians. Whiskers extend an additional 1.5 × the interquartile range from the median. c–f, Characterization of T cell clusters. Bar plots show mean values across clusters of various measures on T cells. ‘PD1 expr’ denotes expression of PD1 (d); ‘Term ex’ denotes a published signature of terminal versus stem-like exhaustion9 (e); ‘Trm sig’ denotes a published signature of Trm cells8 (c); and ‘Tumour pct’ denotes the fraction of cells sampled from tumour versus NAT (f). Horizontal lines in d and f indicate mean values over all cells. g, Gene set enrichment analysis for selected clusters and gene sets. The expression of selected gene sets (columns) is shown for clusters (rows) by plotting each gene in the gene set that was assayed in the integrated dataset as a dot according to its t-statistic from a logistic regression analysis to identify biomarkers for each cluster. Gene Ontology gene sets shown are: histone demethylase activity (HDM); histone methyltransferase activity (HMT); mitotic cell cycle (mitosis); and mitochondrial chromosome genes (MT). A predominance of dots to the right of the vertical line (t = 0) indicates overexpression of the gene set relative to the expected zero mean. Statistically significant cases of overexpression are shown in red with the associated genes when a one-sided P < 0.001 from a one-sample z-test on the t-statistics. h, Transcriptional heterogeneity of T cell clones. Each pie chart represents one of the 20 largest clonotypes in this study, as measured by total clone size across tumour, NAT and blood. Each clone represents a set of cells, indicating its total clone size, used to order clonotypes. The area of each pie is proportional to the clone size. Regions of each pie chart indicate the fractions of cells in the given clone assigned to each cluster. i, Composition of clones by T cell cluster and compartment. Heat map shows the unit-normalized cellular composition of 770 clones with a tumour + NAT clone size ≥ 10 (columns) across T cell clusters and tumour or NAT compartment (rows). Clones are integrated from all patients and grouped by their primary cluster. Within each primary cluster, clones are ordered to show a gradation of cell fraction from tumour to NAT. Each clone is further characterized by its clone size (top) and tissue expansion pattern (coloured bars above the heat map).
a, Tissue expansion patterns of clonotypes by T cell cluster. Bar plot shows the distribution of tissue-associated clones—having at least one cell in tumour or NAT—and a primary T cell cluster assigned, grouped by primary cluster. Clones in each primary cluster are further divided by their tissue residency pattern. b, Tissue expansion patterns of cells by T cell cluster. Bar plots show distributions of T cells in NAT and tumour compartments, grouped by their assigned cluster. The counts in each row, corresponding to a cluster in a, comprise all tissue-resident cells—from tumour or NAT—assigned to that cluster. Cell counts are further distinguished by the tissue expansion pattern of their parent clone, with dual expansion shown on the right pair of bar plots, and singletons and multiplets shown on the left. P value is from a chi-square test on counts of tissue-resident T cells. Asterisks indicate statistically significant over-representation of the given T cell cluster and tissue expansion pattern, with a one-sided P value from a post hoc Fisher exact test on the same counts of tissue-resident T cells as the chi-square test, shown when a Bonferroni-adjusted P < 0.01. c, Clonal expansion patterns for T cell clusters. Scatter plot for each T cell cluster shows tissue-associated clones with the corresponding primary cluster, integrated from all 14 patients in this study and plotted by their clone sizes in NAT and tumour on logarithmic scales. Dots are coloured by their tissue expansion pattern, as per the two-dimensional palette, except that blood singleton and multiplet clones were not plotted because only four patients had blood samples. d, e, Analysis of external datasets. The same methodology of c was applied to datasets from Guo et al.3 on 14 patients with non-small cell lung carcinoma (c) and Zhang et al.4 on 12 patients with colorectal adenocarcinoma (d). Clones were grouped according to their primary cluster from the original analyses, and coloured by the two-dimensional palette for tissue expansion pattern at the bottom right of e. Blood clone sizes are indicated by dot size, as in e.
a, Clonal expansion scatter plots by patient. Data from Fig. 1a are shown, except clones are coloured by their primary cluster (see legend). b, Clonal expansion by cluster and tissue and blood expansion pattern. The 15 scatter plots in Fig. 2c are represented as vertical sets of strip charts, with each chart showing the clone sizes in tumour plus NAT for clones in each tissue expansion pattern in the scatter plot (abbreviated as n, N, D, T and t, as in Fig. 1b). Strip charts are organized by primary cluster and blood expansion pattern: blood-independent, blood non-expanded and blood-expanded. These one-dimensional representations facilitate the comparison of clone sizes and depiction of statistical results. P values are shown from a chi-square test of counts of clones. For each strip chart, the observed/expected ratio and one-sided P values are shown in red when a Bonferroni-adjusted P < 0.01 from post hoc Fisher exact tests of the same counts of clones as the chi-square test. Additional statistical tests were performed to compare mean clone sizes between the blood-independent, blood non-expanded and blood-expanded categories for each tissue expansion pattern in each cluster. Only two tests had Bonferroni-adjusted P < 0.01, shown as bars in the 8.1-Teff dual-expanded category, with two-sided P values from a t-test on log-transformed clone sizes. c, Blood expansion patterns by T cell cluster. Bar plots show the numbers of clones in each of the four patients with a blood sample, with clones grouped by their primary cluster and further divided by their blood clone size as being blood-independent, blood non-expanded or blood-expanded. d, Blood-associated expansion by T cell cluster. As in c, except blood-independent clones are excluded, and only blood-associated clones are tabulated. e, Distribution of T cells in blood by blood expansion pattern. Bar plots show numbers of T cells found in blood, grouped according to their cluster and further divided by the blood expansion or non-expansion pattern of their parent clone. Because parent clones are guaranteed to have the given cell in blood, blood independence is not possible. f, Two-dimensional map of cells in blood by peripheral clonal expansion. Cells from blood with a clonotype are plotted onto the UMAP coordinates from Fig. 2a and coloured green if non-expanded in blood (blood clone size = 1) or a shade of brown for increasing expansion in blood. Ovals from Fig. 2a are added for reference. g, i, Distribution of T cell clusters in tumours (g) and NAT (i) by blood expansion pattern. As in e, except for cells in tumour (g) and NAT (i) from the four patients with blood samples. P values are from a chi-square test of counts of cells over T cell clusters in blood versus counts over T cell clusters in tumour (g) or NAT (i). h, j, Two-dimensional maps of cells in tumour (h) and NAT (j) by blood expansion pattern. As in f, except for cells in tumour (h) and NAT (j) from the four patients with blood samples.
a, Matching bulk TCR-seq and scTCR-seq clonotypes. An example from the Yost et al.5 dataset is shown to illustrate issues in matching clonotypes across bulk and single-cell technologies. Bulk TCR-seq from Adaptive Biotechnologies immunoSEQ technology yields single 87-base-pair segments of individual β-chains, whereas scTCR-seq from 10x Genomics potentially yields combinations of α- and β-chain CDR3 sequences per clonotype, indicated here by four clonotype IDs and associated sequences in grey boxes. The immunoSEQ output also provides a CDR3 amino acid sequence (bulk-CDR3-aa, rectangle) for productive β-chains, which we used to facilitate matching. We considered clonotypes to match if either β-chain CDR3 from scRNA-seq aligned exactly to the bulk TCR-seq sequence at the nucleotide level, at the position consistent with bulk-CDR3-aa. α-chain CDR3 sequences were disregarded in this process. All four clonotypes shown were therefore considered matches to the bulk TCR-seq sequence. For the purpose of counting T cells, a sum was taken over all matching scRNA-seq clonotypes. Further considerations are provided in Methods. b, Correlation of tumour and blood clone sizes in novel CD8 clones. Scatter plots are shown for each patient in Yost et al.5 that had both single-cell RNA-seq and TCR-seq of tumour-infiltrating lymphocytes in pre- and post-treatment tumours as well as bulk TCR-seq of T cells in blood. Dots represent novel CD8 clones based on the primary post-treatment cluster from the original analysis. Novel clones are plotted by the count of transcripts in pre-treatment blood (resorting to post-treatment blood for bcc.su002, which lacked a pre-treatment blood sample), used as a proxy for blood clone size, and clone size in post-treatment tumour. Vertical bar separates novel clones matching a clonotype in blood (blood-associated, right) from those that did not (blood-independent, left). Two-sided P values are shown for a Pearson’s correlation coefficient r on blood-associated novel clones. Patients are ordered by their total (blood-associated plus blood-independent) number of novel CD8 clones. Two-sided P values are shown from a Fisher’s z-test for the comparison of the correlation coefficient of CD8 novel clones and the correlation coefficient of the CD4 novel clones. c, Correlation of tumour and blood clone sizes in novel CD4 clones. Scatter plots are shown for the novel CD4 clones from patients in a, in corresponding order, as in b. d, Clonal diversity in blood. Scatter plots are shown for the patients in b and c, in corresponding order. Dots represent distinct TCR β-chain rearrangements as provided in the original immunoSEQ analysis, plotted by the numbers of templates reported in pre- and post-treatment blood. For patient bcc.su002, which lacked a pre-treatment blood sample, a one-dimensional strip chart shows the post-treatment TCR repertoire with horizontal jitter added to display points more clearly. Increasing clonal diversity can be observed qualitatively as the increasing presence of clones along the main diagonal, and is quantified using Shannon entropy. e, Completeness curves for blood TCR-seq samples. Each plot shows a sample completeness curve for a bulk TCR-seq sample in d based on a rarefaction and extrapolation analysis38, with pre- and post-treatment samples coloured as shown. Each curve indicates the estimated coverage of the total set of TCR β-chain rearrangements as a function of the total number of transcripts sampled. Dot indicates the actual number of transcripts sampled, solid lines indicate the interpolated completeness curve, and dashed lines indicate an extrapolation of the completeness curve.
Extended Data Fig. 8 Matching clonotypes against databases of known and putative virally reactive TCRs.
a, TCR repertoire of clonotypes matching VDJdb. A set of rug plots is shown for each patient, with each plot representing the repertoire of clonotypes matching TCRs listed as reacting against common viral antigens from the VDJdb database54. Viral antigens shown are from cytomegalovirus (p65 antigen), Epstein–Barr virus (BMLF1, EBNA3) and influenza A (M1). Other antigens from these viruses are listed as ‘other’. Each rug plot depicts each distinct clonotype as a region, coloured by its primary cluster, with the height of each region indicating its total clone size in tumour plus NAT. Clonotypes are stacked on top of one another in random order. In situations in which adjacent clones share the same colour, black lines were used to separate them, when both clones had a clone count greater than 5, indicating a need to resolve them visually. Plots show that most matching clonotypes were singletons, but that patients often had a few virally reactive clonotypes that had expanded greatly. b, Association of viral reactivity with clonal expansion patterns. Clonotypes matching VDJdb and multi-cancer TCRs computed55 from The Cancer Genome Atlas (TCGA), suggesting reactivity to a viral antigen, were grouped according to their clonal expansion pattern. Bar plot shows frequencies of matches for each database. P values are from a chi-square test on counts of distinct clonotypes. One-sided P values are shown next to bars when Bonferroni-adjusted P < 0.05 from post hoc Fisher tests performed over the same counts of clonotypes as the chi-square test. c, Association of viral reactivity with primary cell clusters. Clonotypes matching VDJdb (left) and multi-cancer TCRs from TCGA (right) were grouped according to their primary cluster. Bar plots show frequencies of matches for each cluster, with bars coloured as in a. Vertical bars show the mean fraction of clonotype matches across all clonotypes. P values are from a chi-square test on counts of distinct clonotypes with a primary cluster assigned. One-sided P values are shown next to bars when Bonferroni-adjusted P < 0.05 from post hoc Fisher tests performed over the same counts of clonotypes as the chi-square test.
Extended Data Fig. 9 Gene signatures of tissue expansion patterns and relationship to CD8A expression.
a, Consistency of gene signatures in bulk tumour RNA-seq. The 30 highest-ranking genes for each tissue expression pattern involving a tumour sample are shown in a heat map of correlated gene expression from bulk tumour RNA-seq data from all patients in the three clinical trials analysed in this study. Intensities of each cell represent the Pearson’s correlation coefficient on the gene expression values from patients between each pair of genes, and dendrograms indicate the hierarchical clustering of genes. The first division of each dendrogram is used to eliminate genes that are inconsistent with other genes in the signature, possibly due to expression by non-T cells. b, Expansion signatures in scRNA-seq data. Heat map shows the relative gene expression of signatures from a in the scRNA-seq data of this study, used for the initial selection of gene signatures. Intensities indicate the mean log2-transformed fold change of each gene (rows) across patients (columns), in which the fold change was computed within each patient for the T cells from the tumour compartment of a given cluster against all other T cells from the tumour compartment from that patient. ‘Consistent’ indicates genes that passed the filtering step in a by black cells, which are used for subsequent analysis. Genes common to more than one signature are marked by an asterisk. c, Correlation of expansion signatures with CD8A expression. Scatter plots show the correlation of CD8A expression with expansion signature scores across patient bulk RNA-seq samples from three clinical trials. Expression of CCL5 is also included as a marker of expansion, ranking highly in both the tumour multiplet and dual expansion signatures from the scRNA-seq analysis. Each dot represents a pre-treatment bulk tumour RNA-seq sample, coloured by the clinical trial. d, Survival analysis of CD8A expression. Kaplan–Meier plots of PFS are shown for each arm in a clinical trial, with patients dichotomized by their expression of the CD8A gene in bulk tumour above (CD8Ahigh) or below (CD8Alow) the median expression among all patients in the corresponding clinical trial. CD8A expression is used as a marker for the prevalence of intratumoural CD8+ T cells, which is known to be a predictor of response to cancer immunotherapy. Censored observations are indicated by a plus symbol. Hazard ratios (HR) and two-sided P values from a Cox proportional-hazards model on patients in both groups are shown, highlighted in red with the associated survival curve when P < 0.05. Six patients in IMvigor210 were omitted owing to missing values for PFS. e, As in d, except for gene expression of CCL5.
a, Survival based on clonal expansion patterns. Kaplan–Meier survival curves for PFS are shown for three clonal expansion signatures (rows) in each arm of a clinical trial (columns), in which patients are dichotomized by scores above (dotted lines) and below (solid lines) the median in each clinical trial. Plus symbols indicate censoring events. Hazard ratios (HR) and two-sided P values from a Cox proportional-hazards model on patients in both groups are shown, highlighted in red with the corresponding survival curve when P < 0.05. b, Survival based on expansion signatures in the context of CD8A expression. Kaplan–Meier plots for PFS are shown using both CD8A expression and expansion signature scores (rows) in each arm of a clinical trial (columns). Patients in each clinical trial were divided into four groups based on CD8A expression above (CD8Ahigh) or below (CD8Alow) the median and on expansion signature score above (signaturehigh) or below (signaturelow) the median among all patients in the corresponding clinical trial. Patients with low CD8A expression and low expansion signature score were used as a control for each of the other three groups. Hazard ratios and two-sided P values are from a Cox proportional-hazards model on patients in each group, highlighted in red with the corresponding survival curve when P < 0.05. c, Survival based on dual-expanded and tumour multiplet signatures. Kaplan–Meier survival curves are plotted as in b, except that patients were divided into four groups based on dual-expanded clone signature above (Dhigh) or below (Dlow) the median and on tumour multiplet signature above (Thigh) or below (Tlow) the median among all patients in the corresponding clinical trial.
Supplementary Methods: Example of gating strategy for one of the samples in the study.
Supplementary Table 1: Patient information. Demographic and clinical information for the patients in our study.
Supplementary Table 2: Markers of immune cell clusters. Results of the FindMarkers procedure to find genes over-expressed in each immune cell cluster relative to all other immune cells.
Supplementary Table 3: Shared clonotypes across patients. A list of clonotypes that were identical in two or more patients in this study, based on matches between all available alpha- and beta-chain CDR3 nucleotide sequences from scTCR-seq.
Supplementary Table 4: Markers of T cell clusters. Results of two procedures to find genes over-expressed in each T cell cluster relative to all other T cells.
Supplementary Table 5: Gene signatures for tissue expansion patterns. Lists of genes included in the signatures for the tumour singleton, tumour multiplet, and dual-expanded patterns.
About this article
Cite this article
Wu, T.D., Madireddi, S., de Almeida, P.E. et al. Peripheral T cell expansion predicts tumour infiltration and clinical response. Nature 579, 274–278 (2020). https://doi.org/10.1038/s41586-020-2056-8
This article is cited by
Association of peripheral basophils with tumor M2 macrophage infiltration and outcomes of the anti-PD-1 inhibitor plus chemotherapy combination in advanced gastric cancer
Journal of Translational Medicine (2022)
Identification of a ZC3H12D-regulated competing endogenous RNA network for prognosis of lung adenocarcinoma at single-cell level
BMC Cancer (2022)
Nature Reviews Immunology (2022)
Integrating T cell receptor sequences and transcriptional profiles by clonotype neighbor graph analysis (CoNGA)
Nature Biotechnology (2022)
Single-cell characterization of leukemic and non-leukemic immune repertoires in CD8+ T-cell large granular lymphocytic leukemia
Nature Communications (2022)