Supplementary Figure 1. Validation of antigen detection by CyTOF conjugates. a,
Representative antibody validation experiments for the specificity of epithelial markers (Sox2) in
LUSC and LUAD cell lines (left), cell death markers (cleaved caspase 3) in a LUSC cell line
treated with BH3 mimetics (A1331852: BCL-XL inhibitor; S63845: MCL1 inhibitor) to induce
apoptosis91 (middle) and T cell co-stimulatory markers (OX40) in human CD4+ T cells from bulk
PBMCs stimulated for 48h with anti-CD3/anti-CD28. b, Quantification of percentage of CD45+
cells in non-malignant lung and tumour tissue from ES and NS patients by CyTOF (left panel) and
immunohistochemistry (right panel, representative patients). CyTOF CD45hi defined as >80%
CD45+ cells; CyTOF CD45lo defined as <20% CD45+ cells. n = 10 ES, n = 8 NS. Paired t-test
between lung and tumour, unpaired t-test between non-malignant lung/tumour. *p<0.05. c,
Correlation between PDL1 expression in pan-keratin+ tumour cells detected by CyTOF (y-axis)
and PDL1 expression detection by clinical IHC, expressed as tumour proportional score. n = 9
early-stage primary tumours with matched CyTOF and clinical data available. d, Heatmap of the
median cell type marker expression (scaled, arcsinh-transformed) across all samples and
aggregated by clusters identified in Fig 1a. e, Quantification of individual cell cluster in nonmalignant
lung tissue and tumour tissue determined by FlowSOM clustering, expressed as a
percentage of CD45+ cells. Paired Wilcoxon ranked tests between lung and tumour, unpaired t-test
between samples from patients with different smoking statuses, p-value not corrected for multiple
testing. *p<0.05, **p<0.01.
Supplementary Figure 2. Phenotype and activity of tissue resident memory
T cells within non-malignant lung tissue of ES and NS lung cancer patients. a, Heatmap of the
proportion of co-stimulatory and co-inhibitory receptor expressing cells within
CD3+CD69+CD45RO+CCR7- resident Treg (left) and CD8+ TRM cells (right) in non-malignant lung
tissue analysed by CyTOF. Data is scaled to mean expression ± standard deviation per marker. n
= 9 ES and n = 10 NS patients. Unpaired t-test, not significant. Greyed out bars indicate patients
with too few cells (<50) to reliably assess cell phenotypes.
Supplementary Figure 3. Proportions of T cells infiltrating NS and ES tumours are variable.
a-d, Proportion of CD69+CD45O+CCR7- TRM-like cells in CD3+ T cells. Total TRM-like cells (a)
CD4+ TRM-like cells in CD4+ T cells (b), resident-like activated Treg in CD4+ T cells (c), and CD8+
TRM-like cells in CD8+ T cells (d). Paired t-test between lung and tumour, unpaired t-test between
NS and ES. p-values not corrected for multiple testing. *p<0.05.
Supplementary Figure 4. Relationship between tumour mutational burden, neoantigen
numbers, CD8+ T cell infiltration and smoking status in lung adenocarcinoma. a, Association
between TMB and neoantigen numbers in the TCGA LUAD cohort, coloured according to
smoking status. Linear regression analysis, p value as indicated. b, TMB in the TCGA LUAD
cohort related to patient smoking status. Unpaired t-test with Welch’s correction, ***p<0.001. c,
Number of neoantigens in TCGA LUAD cohort related to patient smoking status. Unpaired t-test
with Welch’s correction, ***p<0.001. d, TMB in the TRACERx cohort of LUAD, NSCLC-NOS
and LUSC related to patient smoking status. Unpaired t-test with Welch’s correction, ***p<0.001.
Data represents multiple samples per indicated numbers of patients. e, Number of neoantigens in
the TRACERx cohort of LUAD, NSCLC-NOS and LUSC related to patient smoking status.
Unpaired t-test with Welch’s correction, ***p<0.001. Data represents multiple samples per
patient. f, Scatter plot of TMB and CD8+ T cell score in the TCGA LUAD cohort. TMB calculated
from whole exome sequencing data84 and CIBERSORT CD8+ T cell score estimated from RNA
sequencing data using TIMER2.0 online platform85. Data represents the power transformed square
root of values to account for zero values not possible using log transformation. Linear regression
analysis, p value as indicated. g, Scatter plot of neoantigen numbers and CD8+ T cell score within
the TCGA LUAD cohort. Neoantigens calculated from whole exome sequencing data as published
in Rooney et al50. Linear regression analysis, p value as indicated. h, Volcano plot depicting fold
change in TRM accessory molecules between NS and ES late-stage tumours. 30 parameters
analysed by CyTOF, co-stimulatory molecules in green, CTLA4 in yellow and co-inhibitory
markers in purple. Dotted line indicates q value = 0.05. Data represents average marker fold change
within each cell type (CD4+ TRM-like, resident-like Treg, and CD8+ TRM-like), n = 6 ES and n =
4 NS late-stage (Stage IIIb/IV) tumours.
Supplementary Figure 5. Clonal immune escape events are only detected in ever-smoker
patients. a-c, Proportion of patients with loss of heterozygosity of one or more HLA alleles in
some tumour sites (subclonal LoH; green) or all sites (clonal LoH; blue) in the TRACERx cohort
of LUAD, NSCLC-NOS and LUSC related to smoking status (a), TMB (b) or immune infiltration
estimated from RNA sequencing data (c). Fisher’s exact test, *p<0.05. d, HLA LoH in
adenocarcinoma in situ and minimally invasive adenocarcinoma, the pre-invasive forms of LUAD
related to smoking status, data called from Chen et al51. Clonality of LoH cannot be determined
from single site data. Fisher’s exact test, not significant. e, TMB and the number of neoantigens
calculated in samples from the TRACERx and LxG cohort were subject to robust linear modelling
and residuals and weights of each data point are plotted. Samples with a weight < 0.9 and a negative
residual are indicated in red as ‘immunoedited’ or neoantigen depleted. f, Proportion of patients
with clonal immunoediting (present in all sites) or subclonal immunoediting (present in some sites)
in the TRACERx cohort of LUAD, NSCLC-NOS and LUSC, related to TMB, immune infiltration,
HLA LoH or smoking status. Fisher’s exact test, *p<0.05. g, Proportion of neoantigens shared in
all tumour sites (clonal neoantigens) in the TRACERx cohort of LUAD, NSCLC-NOS and LUSC,
related to TMB, immune infiltration, HLA LoH events and immunoediting. Unpaired t-test,
**p<0.01.
Supplementary Figure 6. Targeting anti-ICOS as an alternative immunotherapy in T cell
quiescent lung tumours. a, Quantification of co-stimulatory molecule OX40 and 4.1BB
expression in CD69+CD45RO+CCR7- CD4+ T cells in early-stage lung cancers, as analysed by
CyTOF. n = 10 ES, n = 8 NS. Paired t-test between lung and tumour, unpaired t-test between nonmalignant
lung/tumour. P-values not corrected for multiple testing, *p<0.05, **p<0.01. b,
Schematic of in vivo treatment of Ad5-CMV-Cre intranasally infected mice bearing
KrasG12D/+;p53fl/fl tumours with either rat IgG2a and hamster IgG control; single agent anti-PD1
and hamster IgG, single agent anti-ICOS agonist antibody and rat IgG2a or combination anti-PD1
and anti-ICOS. c, Expression of ICOS in CD4+ T cells in the spleen and left lung lobe of
KrasG12D/+;p53fl/fl mice, tumour-free (Ad-CMV-Cre -) or in tumour-bearing lungs (intranasal
infections with Ad-CMV-Cre +) measured by flow cytometry. d, Representative images of spatial
quantification of Ki67+(magenta substrate chromogen, MSC) CD8+ (DAB) cells in tumour areas
using QuPath software. Scale bar indicates 20 μm. e, Density of Ki67+CD8+ T cells per mm2 of
tumour in each treatment group. f, Number of ICOS+CD62L- effector Treg cells92 in the Treg
compartment in the left lung lobe as detected by flow cytometry in each treatment group. Unpaired
t-test, *p<0.05. g, Proportion of PD1+ cells in the effector Treg population in the left lung lobe as
detected by flow cytometry in each treatment group. Unpaired t-test, *p<0.05, **p<0.01. h,
Percent of Ki67+ cells in CD8+ T cells detected by flow cytometry 48h after anti-CD3/anti-CD28
stimulation and treatment with the indicated therapies. Control (ctrl) is human IgG4 and hamster
IgG treatment. n = 4 ES patients, n = 5 NS patients. Unpaired t-test, not significant.