276 | Nature | Vol 579 | 12 March 2020
Article
TCR-independent mechanisms of activation^11. In addition, subclus-
ters 8.3a to 8.3c showed increasing prevalence in tumour versus NAT
(Extended Data Fig. 4f ), supporting a relationship between PD1 expres-
sion, terminal exhaustion, and increased exposure to cognate antigens
within the tumour microenvironment^12.
In addition, our deep sequencing analysis revealed several clusters
of T cells not matching published reference gene signatures (Extended
Data Fig. 4a). We characterized these clusters using gene set enrich-
ment analysis, which showed active histone modification and mitosis
in clusters 8.4 and 8.5, respectively (Extended Data Fig. 4g). Chromatin
regulation in 8.4 was further evidenced by high expression of the long
noncoding RNA MALAT1^13 and the chromatin remodelling enzyme
CHD1^14 (Extended Data Fig. 4b). Cluster 3.1 expressed abundant mito-
chondrial genes, suggestive of apoptosis, and we set this cluster aside as
generic CD3+ cells. Clusters 8.4-Chrom and 8.5-Mitosis had similarities
to Trm cells, with cells expressing a range of ITGAE levels (Extended Data
Fig. 4b) and the highest mean expression of PD1 (Extended Data Fig. 4d),
generally considered a marker of exhaustion. Moreover, 8.5-Mitosis
also cross-labelled most closely with clusters that have been reported
to represent exhausted T cells, specifically, the CD8_ex cluster observed
by Yost et al.^5 and the CD8_C6-LAYN cluster observed by Guo et al.^3
(Extended Data Fig. 4a). However, neither 8.4-Chrom nor 8.5-Mitosis
scored highly on the published signature for Trm cells, which are thought
to be the main source of exhausted T cells (Extended Data Fig. 4c).
Combining scTCR-seq and scRNA-seq data yielded insights into the
clonal expansion behaviour of clones and T cells. One challenge in such
combined analysis is that whereas expansion behaviour is attributable
to clones, clusters are assigned to individual cells based on their tran-
scriptional profiles. Therefore, a given clone may comprise a mixture
of T cell phenotypes with distinct transcriptional profiles and hence
different clusters. Accordingly, clones in our dataset exhibited a range
of transcriptional diversity: some clones were homogeneous for a sin-
gle cluster, whereas others had heterogeneous cellular mixtures of
several clusters (Extended Data Fig. 4h, i). Nevertheless, most clones
had a primary cluster that predominated and served as an approximate
phenotype for subsequent analysis.
As expected, clones partitioned into CD8 or CD4 types, which exhib-
ited distinct tissue expansion behaviours (Extended Data Fig. 5a, b).
Clones of primarily 8.1, 8.2 and 8.3 T cells were largely dual-expanded
(Fig. 2b), whereas clones of CD4+ T cells were generally singletons,
with some exceptions: (1) low-level dual expansion by 4.4-FOS clones;
(2) tumour and NAT multiplet expansion by clones of 4.1-Trm cells;
and (3) tumour multiplet expansion by 4.5-IL6ST and 4.6-Treg clones
(Extended Data Fig. 5c). Clusters 8.4, 8.5 and 8.6 showed very little
clonal expansion in tissue. Our analysis of external datasets^3 ,^4 yielded
similar results, with the analogous CX3CR1 and GZMK clusters showing
similar patterns of clonal expansion to our 8.1-Teff and 8.2-Tem clusters,
respectively (Extended Data Fig. 5d, e).
When we mapped primary clusters onto clones from our four patients
with blood samples (shown previously in Fig. 1a), we observed that
many clones exhibiting parallel dual expansion were 8.1-Teff clones
(Extended Data Fig. 6a). To investigate this association, we re-grouped
clones from the patients according to their blood expansion patterns
and primary cluster (Fig. 2c), and found that blood-expanded clones
with a primary cluster of 8.1-Teff had a significant (P = 9 × 10−14) correla-
tion between tumour and NAT clone sizes. In addition, 8.3b-Trm clones,
possibly containing T cells with TCR-independent activation, showed
lower levels of parallel dual expansion. Nevertheless, 8.1-Teff clones were
more numerous and were the only ones to show a statistically significant
(P = 1.4 × 10−5 and P = 3.0 × 10−5) relationship between blood expan-
sion and dual-expanded clone size (Fig. 2d, Extended Data Fig. 6b),
indicating their predominant role in peripheral clonal expansion and
tumour infiltration.
Blood-expanded 8.1-Teff clones may have a role not only in tumour
infiltration but also in differentiation into other T cell phenotypes^15.
Although blood-expanded 8.1-Teff clones constituted only a small frac-
tion of all clones (Extended Data Fig. 6c, d), they had a disproportion-
ate influence on T cell composition in blood and tumours. By linking
individual T cell phenotypes with the blood expansion patterns of
their parent clones, we observed that 8.2-Tem and 8.3-Trm cells in blood
(Extended Data Fig. 6e, f ) and tumours (Extended Data Fig. 6g, h) shared
clonotypes extensively with blood-expanded clonal lineages of 8.1-Teff
cells. Thus, differentiation of blood-expanded 8.1-Teff clones may yield
a diverse array of T cell subtypes in blood and tumours. By contrast,
most 8.2-Tem and 8.3-Trm cells in NAT belonged to blood-independent
clones (Extended Data Fig. 6i, j), consistent with mechanisms of local
expansion.
Our data capture only a single snapshot of clone sizes across tumour,
NAT and blood, representing the accumulation over time of various
dynamic processes. To understand these dynamics, we analysed an
external dataset^5 of pre- and post-treatment T cells from basal cell
carcinomas, with the understanding that the two sets of measurement
differ both by the effects of immunotherapy and by the passage of time.8.3b-T
rm8.3a-T
8.3c-rm
Trm8.2-T
8.1- em
TeffRPL324.2-
FOS4.4-
4.1-T
rm4.6a-T
regIL6ST4.5-3.1-MT8.6-KLRB1TCF74.3-Mitosis8.5- 4.6b-
TregChr8.4-om
1 500tTDNn Ind1 500Non1500Exp010500 010500 010 500 010 500nr = 0.7 = 26
P = 8 × 10–5010 500010500 010500 010 500 010 500 010 500010500nr = 0.76 = 68
P = 9 × 10–14
010500 010 500 010 500nr = 0.82 = 12
P = 0.0011
010 500Clone size, NATClone size,tumour 0 ∞1∞(^01)
8.1-Teff 8.2-Tem 8.3a-Trm 8.3b-Trm 8.3c-Trm
Blood-independent
Blood non-expanded
Blood-expanded
scRNA-seq (n = 141,623) scTCR-seq (n = 89,319) 8.1-Teff
Clone size, NAT Clone size, blood: 0 1 2 5 10 20 50 100
Clone size,tumour
Clone size, tumour + NAT
P = 3.0 × 10–5P = 1.4 × 10–5
n = 16 22 68
c
abd
0
20
500
0
20
500
0
20
500
0
20
500
0
20
500
0
20
500
0
20
500
0
20
500
0
20
500
0
20
500
0
20
500
0
20
500
0
20
500
0
20
500
0
20
500
Fig. 2 | T cell clusters and clonal expansion. a, Cluster analysis. T cells from all
14 patients are plotted by uniform manifold approximation and projection
(UMAP) dimensionality reduction of the scRNA-seq data, and coloured by
cluster assignment, with cluster regions approximated by labelled ovals. The
curved line highlights the separation of CD8 (left) and CD4 (right) T cells.
b, Tissue expansion patterns by cluster. T cells in tumour and NAT with a
clonotype are plotted by UMAP coordinates from a and coloured by tissue
expansion pattern of their parent clone using a two-dimensional palette.
c, Clonal expansion by cluster and blood expansion pattern. Clones from Fig. 1a
are plotted by blood independence, non-expansion and expansion (rows) and
by primary cluster (columns), using coordinates for NAT and tumour clone size,
and dot size for blood clone size. Grey lines separate NAT and tumour clone
sizes of zero and non-zero. Two-sided P values are shown when P < 0.01 from a
Pearson’s correlation coefficient r on log-transformed clone sizes for dual-
expanded clones. d, Analysis of 8.1-Teff clones. Scatter plots from c f o r 8. 1 -Teff for
blood-independent, non-expanded and expanded clones are shown as
distributions of clone sizes in tumour plus NAT for each tissue expansion
pattern, abbreviated as in Fig. 1b. Vertical jitter distinguishes overlapping
points. Two-sided P values are shown for Bonferroni-adjusted P < 0.01 from a
t-test on log-transformed clone sizes for dual-expanded clones.