www.sciencemag.org/cgi/content/full/science.aaf1490/DC1

Supplementary Material for

Clonal neoantigens elicit T cell immunoreactivity and sensitivity to immune checkpoint blockade

Nicholas McGranahan, Andrew J. S. Furness, Rachel Rosenthal, Sofie Ramskov,

Rikke Lyngaa, Sunil Kumar Saini, Mariam Jamal-Hanjani, Gareth A. Wilson, Nicolai J. Birkbak, Crispin T. Hiley, Thomas B. K. Watkins, Seema Shafi,

Nirupa Murugaesu, Richard Mitter, Ayse U. Akarca, Joseph Linares, Teresa Marafioti, Jake Y. Henry, Eliezer M. Van Allen, Diana Miao, Bastian Schilling, Dirk Schadendorf,

Levi A. Garraway, Vladimir Makarov, Naiyer A. Rizvi, Alexandra Snyder, Matthew D. Hellmann, Taha Merghoub, Jedd D. Wolchok, Sachet A. Shukla,

Catherine J. Wu, Karl S. Peggs, Timothy A. Chan, Sine R. Hadrup, Sergio A. Quezada,* Charles Swanton*

*Corresponding author. E-mail: s.quezada@ucl.ac.uk (S.A.Q.); charles.swanton@crick.ac.uk (C.S.)

Published 3 March 2016 on Science Express

DOI: 10.1126/science.aaf1490

This PDF file includes:

Materials and Methods Figs. S1 to S7 Tables S1 to S5 Full Reference List

Other Supplementary Material for this manuscript includes the following: (available at www.sciencemag.org/content/science.aaf1490/DC1)

Tables S3 to S5 as separate Excel files

Materials and Methods Description of Patient Cohorts

Samples for sequencing (L001, L002, L003, L004, L008, L011 and L012) were obtained from patients diagnosed with non-small cell lung cancer (NSCLC) who underwent definitive surgical resection prior to receiving any form of adjuvant therapy, such as chemotherapy or radiotherapy. Informed consent allowing for genome sequencing had been obtained. All samples were collected from University College London Hospital, London (UCLHRTB 10/H1306/42) and were subjected to pathology review to establish the histological subtype: five tumors were classified with CK7+/TTF1+ adenocarcinoma (L001, L003, L008 and L011) or poorly-differentiated CK7+ carcinoma (L004) histology (LUAD), one tumor (L012) with squamous cell carcinoma histology (LUSC) and one tumor (L002) with adenosquamous histology (LUAD/LUSC). Detailed clinical characteristics are provided in table S1. Additional details regarding these tumours, including tumor processing, can be found in the original publications (8, 9)

Samples obtained from (2) reflected a patient cohort of stage IV NSCLC, and a detailed description of this patient cohort, including tumor processing, can be found in supplementary material of (2). Detailed clinical characteristics of this cohort are provided in Table S5.

Samples obtained from (4) reflected a patient cohort of late stage melanoma, and a detailed description of this patient cohort, including tumor processing, can be found in supplementary material of (4).

Finally, samples obtained from (17) reflected an additional cohort of late stage melanoma. A detailed description of this patient cohort, including tumor processing, can be found in the supplementary material of (17).

Clinical efficacy analysis For each sample cohort, clinical efficacy analysis was kept consistent with the

original publication. In brief, for (2) cohort, clinical efficacy analysis was performed as in (2). In

brief, objective response to pembrolizumab was assessed by investigator-assessed immune-related response criteria (irRC) by a study radiologist. As outlined in protocol, CT scans were performed every nine weeks. Partial and complete responses were confirmed by repeat imaging occurring a minimum of 4 weeks after the initial identification of response; unconfirmed responses were considered stable or progressive disease dependent on results of the second CT scan. Durable clinical benefit (DCB) was defined as stable disease or partial response lasting longer than 6 months (week 27, the time of third protocol-scheduled response assessment). No durable benefit (NDB) was defined as progression of disease ≤ 6 months following commencement of therapy. For patients with ongoing response to study therapy, progression-free survival was censored at the date of the most recent imaging evaluation. For ‘alive’ patients, overall survival was censored at the date of last known contact. Details regarding response for each patient can be found in table S5.

For (4) cohort, long-term clinical benefit was defined by radiographic evidence of freedom from disease or decreased volume of disease for > 6 months. Conversely, lack of long-term benefit was defined by tumor growth on every computed tomographic scan after the initial treatment (no benefit) or a clinical benefit lasting 6 months or less (minimal benefit).

1

Finally, for (17) cohort, clinical benefit was defined as complete response or partial response to ipilimumab by RECIST criteria or overall survival greater than 1 year with stable-disease by RECIST classification. A third group of patients who had no clinical benefit but prolonged overall survival were evaluated as a third clinical cohort.

TCGA exome data sets

Tumor samples, with mutation calls and HLA typing described below, were obtained from the Cancer Genome Atlas (TCGA) for a cohort of lung adenocarcinoma (LUAD, n = 150) and lung squamous cell carcinoma (LUSC, n = 124). SNV data was obtained from TumorPortal (10) for the LUAD and LUSC TCGA cohorts (http://www.tumorportal.org/tumor_types?ttype=LUAD | LUSC). One LUAD patient, TCGA-05-4396, was excluded for having over 7000 low quality mutations called, mostly in a C[C>G]G context. A LUSC patient, TCGA-18-3409, was excluded for bearing a strong UV signature, uncharacteristic of a LUSC tumor. Multi-region Whole-Exome Sequencing and variant calling

L001, L002, L003, L004, L008, L011 and L-12 The sequencing and analysis of the germline, and primary tumor regions have previously been described in (8, 9). Variant calling from previously published cohorts

BAM files representing both the germline and tumor regions from each cohort wereas obtained and converted to FASTQ format using picard tools (1.107) SamToFastq .

Raw paired end reads (100bp) in FastQ format were aligned to the full hg19

genomic assembly (including unknown contigs) obtained from GATK bundle 2.8 (22), using bwa mem (bwa-0.7.7) (23). Picard tools v1.107 was used to clean, sort and merge files from the same patient region and to remove duplicate reads (http://broadinstitute.github.io/picard). Quality control metrics were obtained using a combination of picard tools (1.107), GATK (2.8.1) and FastQC (0.10.1) (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/).

SAMtools mpileup (0.1.16) (24) was used to locate non-reference positions in

tumour and germline samples. Bases with a phred score of <20 or reads with a mapping-quality <20 were skipped. BAQ computation was disabled and the coefficient for downgrading mapping quality was set to 50. Somatic variants between tumour and matched germline were determined using VarScan2 somatic (v2.3.6) (25) utilizing the output from SAMtools mpileup. Default parameters were used with the exception of minimum coverage for the germline sample that was set to 10, minimum variant frequency was changed to 0.01 and tumour purity was set to 0.5. VarScan2 processSomatic was used to extract the somatic variants. The resulting SNV calls were filtered for false positives using Varscan2's associated fpfilter.pl script, having first run the data through bam-readcount (0.5.1). Only INDEL calls classed as ‘high confidence‘ by VarScan2 processSomatic were kept for further analysis.

. Clonal architecture analysis

For samples subject to multi-region sequencing, clonal status of each mutation was estimated based on multi-region sequencing calls. In brief, each mutation was

2

classified as clonal if identified and present in each and every tumor region sequenced within the tumor. Conversely, any mutations not ubiquitously present in every tumor region were classified as subclonal.

For single sample sequenced tumors, a modified version of PyClone was used.

This modification was implemented to allow clustering on mutations occurring in regions of subclonal copy number. In brief, the cancer cell fraction of each mutation was estimated by integrating the local copy number (obtained from ASCAT, see below), tumor purity (also obtained from ASCAT), and variant allele frequency. For a given mutation we first calculated the observed mutation copy number, nmut, describing the fraction of tumor cells carrying a given mutation multiplied by the number of chromosomal copies at that locus using the following formula:

where VAF corresponds to the variant allele frequency at the mutated base, and

p, CNt , CNn are respectively the tumor purity, the tumor locus specific copy number,

and the normal locus specific copy number. We then calculated the expected mutation copy number, nchr, using the VAF and assigning a mutation to one of the possible copy numbers using maximum likelihood. Notably, mutations could reside on subclonal copy numbers. Specifically, if the observed variant allele frequency was significantly different from that expected (P<0.01, using prop.test in R), we determined whether a subclonal copy number event could result in a non-significant (P>0.01) difference between observed and expected VAFs.

All mutations were then clustered using the PyClone Dirichlet process clustering (26). Given that copy number and purity had already been corrected, we set integer copy numbers to 1 and purity to 1; allowing clustering to simply group clonal and subclonal mutations based on their cancer cell fraction estimates. We ran PyClone with 10,000 iterations and a burn-in of 1000, and default parameters.

Notably, for assessing mutation clonal status, mutations were first further filtered to ensure reliable clustering. In brief, only mutations with a read depth of at least 10 in both germline and tumour were used, a Varscan2 somatic p-value threshold of 0.01 and that passed Varscan2 filtering. A minimum of 5 alternate reads was required for each variant, as well as a minimum tumor variant allele frequency of 1%. Mutations were also filtered such that a maximum of 2 germline reads, and 2% germline variant allele frequency was permitted.

To ensure accurate subclonal reconstruction, only samples with a tumor coverage of at least 50X, and a normal coverage of 30X were utilized. In addition, for a number of tumors reliable copy number, mutation and purity estimations could not be extracted, rendering clonal architecture analysis intractable and these tumors were omitted from the analysis. The following samples were excluded due to either sequencing depth or lack of accurate copy number/clonality estimation: Pat02, Pat06, Pat100, Pat101, Pat103, Pat106, Pat110, Pat113, Pat131, Pat132, Pat135, Pat138, Pat139, Pat140, Pat148, Pat159, Pat160, Pat163, Pat165, Pat166, Pat170, Pat171, Pat174, Pat175, Pat24, Pat36, Pat38, Pat73, Pat77, Pat78, Pat79, Pat92, ZA6965, GR0134, VA1330, CR4880, NR9341, CRNR4941, LSD4691, LSDNR3086, LSDNR9298, NR2137, NR9445, PR03803 and PR4092.

Copy Number Analysis

3

For data obtained from (2, 4, 17) processed sample exome SNP and copy number data from paired tumor-normal was generated using VarScan2 (v2.3.6). Varscan2 copy number was run using default parameters with the exception of min-coverage (27) and data-ratio. The data-ratio was calculated on a per-sample basis as described in (25). The output from Varscan was processed using ASCAT v2.3 (28) to provide segmented copy number data and cellularity and ploidy estimates for all samples based on the exome sequence data. The following setting was altered from its default value: Threshold for setting ACF to 1 was adjusted from 0.2 to 0.15 and the package was run with gamma setting of 1.

For TCGA samples, SNP6.0 data was processed to yield copy number information, as described in (29)

Mutational Signature Analysis

Mutational signature analysis was performed using the deconstructSigs package in R (30), that selects which combination of known mutational signatures (18) can account for the observed mutational profile in each sample. For NSCLC, the associated signatures considered were Signature 1A, Signature 2, Signature 4 and Signature 5. For melanoma samples, the signatures considered were Signature 1A, Signature 5, Signature 7 and Signature 11. Given that Signature 11 corresponds to an alkylating agent signature, this signature was only considered either if the patient was treated with DTIC or temolozomide or if pre-treatment status was unknown. To avoid overcalling Signature 11, given its overlapping context with Signature 7, in samples where pre-treatment was unknown, only samples with >50% of all mutations or >75% of clonal or subclonal mutations corresponding to the Signature were classified as harboring Signature 11. Notably, allowing any sample to harbor Signature 11 did not change the results.

HLA Typing of Patient Samples For all TCGA patients, the 4-digit HLA type was determined using POLYSOLVER (POLYmorphic loci reSOLVER)(12). Patients L011 and L012 were serotyped and simultaneously genotyped using Optitype (31), which produced concordant results. Patients L001, L002, L004, L008 and all patients from (2, 4, 17) were HLA-typed using Optitype (31) Identification of Putative Neoantigens

Identified non-silent mutations were used to generate a comprehensive list of peptides 9-11 amino acids in length with the mutated amino acid represented in each possible position. The binding affinity of every mutant peptide and its corresponding wild-type peptide to the patient’s germline HLA alleles was predicted using netMHCpan-2.8 (32, 33). Candidate neo-antigens were identified as those with a predicted binding strength of < 500 nM. TCGA Survival Analysis

Clinical data for the TCGA patients was accessed through the TCGA data portal and downloaded from https://tcga-data.nci.nih.gov/tcgafiles/ftp_auth/distro_ftpusers/anonymous/tumor/CANCER.TYPE/bcr/biotab/clin/. Patients were first grouped according to quartile of the variable being considered. Survival analyses were then performed in R using the survival package. Complete survival data was available for 139/150 LUAD patients and 122/124 LUSC patients.

4

Differential Gene Expression Analysis

RNA-sequencing data was downloaded from the TCGA data portal. For each LUAD and LUSC sample, all available ‘Level_3’ gene-level data was obtained. The raw read counts were used as input into the R package DESeq2 for analysis. Transcriptome-wide differential gene expression analyses were performed between all TCGA LUAD and LUSC tumors and between high and low clonal neoantigen TCGA LUAD groups. Significantly differentially expressed immune-related genes (listed in Table S3 and Table S4, respectively) were identified. These genes were clustered on their co-expression using the metric 1-r2.

Isolation of tumor-infiltrating lymphocytes (TILs) for L011 and L012

Tumors were taken directly from the operating theatre to the department of pathology where the sample was divided into regions. Samples were subsequently minced under sterile conditions followed by enzymatic digestion (RPMI-1640 (Sigma) with Liberase TL research grade (Roche) and DNAse I (Roche)) at 37°C for 30 minutes before mechanical dissociation using gentleMACS (Miltenyi Biotech). Resulting single cell suspensions were filtered and enriched for leukocytes by passage through a Ficoll-paque (GE Healthcare) gradient. Live cells were counted and frozen in human AB serum (Sigma) with 10% dimethyl sulfoxide at -80°C before transfer to liquid nitrogen.

In-vitro expansion of tumor-infiltrating lymphocytes for L011 and L012

TILs were expanded using a rapid expansion protocol (REP) in T25 flasks containing EX-VIVO media (Lonza) supplemented with 10% human AB serum (Sigma), soluble anti-CD3 (OKT3, BioXCell), 6000IU/mL recombinant human (rhIL-2, PeproTech) and 2x107 irradiated PBMCs (30Gy) pooled from 3 allogeneic healthy donors. Fresh media containing rhIL-2 at 6000IU/mL was added every three days as required. Following 2 weeks of expansion, TILs were counted, phenotyped by flow cytometry and frozen in human AB serum (Sigma) at -80°C before use in relevant assays or long-term storage in liquid nitrogen. MHC multimer generation and combinatorial encoding-flow cytometry analysis

MHC-multimers holding the predicted neoepitopes were produced in-house (Technical University of Denmark, laboratory of SRH). Synthetic peptides were purchased at Pepscan Presto, NL. HLA molecules matching the HLA-expression of L011 (HLA-A1101, A2402, and B3501) and L012 (HLA-A1101, A2402, and B0702) were refolded with a UV-sensitive peptide, and exchanged to peptides of interest following UV exposure (34-37). Briefly, HLA complexes loaded with UV-sensitive peptide were subjected to 366-nm UV light (CAMAG) for one hour at 4°C in the presence of candidate neoantigen peptide in a 384-well plate. Peptide-MHC multimers were generated using a total of 9 different fluorescent streptavidin (SA) conjugates: PE, APC, PE-Cy7, PE-CF594, Brilliant Violet (BV)421, BV510, BV605, BV650, Brilliant Ultraviolet (BUV)395 (BioLegend). MHC-multimers were generated with two different streptavidin-conjugates for each peptide-specificity to allow a combinatorial encoding of each antigen responsive T cell, enabling analyzes for reactivity against up to 36 different peptides in parallel (15, 38).

5

Identification of neoantigen-reactive CD8+ T cells MHC-multimer analysis was performed on in-vitro expanded CD8+ T

lymphocytes isolated from region-specific lung cancer samples and adjacent normal lung tissue. 288 and 354 candidate mutant peptides (with predicted HLA binding affinity <500nM, including multiple potential peptide variations from the same missense mutation) were synthesized and used to screen expanded L011 and L012 TILs respectively. Simultaneously, TIL responses to HLA-matched viral peptides were assessed, demonstrating functionality of the employed MHC-multimer technology. Viral peptides for L011 included A11 EBV-EBNA4 (AVFDRKSDAK), A11 HCMV pp65 (GPISGHVLK), A24 EBV LMP-2 419-427 (TYGPVMCL), A3 EBV EBNA 3A RLR (RLRAEAQVK), A24 HCMV 248-256 (AYAQKIFKIL), B35 Flu Matrix (ASCMGLIY), B35 ENV EBNA 3B (AVLLHEESM), EBV EBNA-3 114-121 (RYSIFFDY) and EBV BZLF1 (APENAYQAY). For L012, these consisted of A11 EBV EBNA4 (AVFDRKSDAK), A11 HCMV pp65 (GPISGHVLK), A24 EBV EBNA-3 114-121 (RYSIFFDY), A24 EBV LMP-2 419-427 (TYGPVFMCL), A24 EBV RTA 28-37 (DYCNVLNKEF), A24 HCMV 248-256 (AYAQKIFKIL), B7 CMV pp65 RPH-L (RPHERNGFTV), B7 CMV pp65 TPR (TPRVTGGGAM) and B7 EBV EBNA RPP (RPPIFIRLL). Finally, reactivity of healthy donor CD8+ PBMC’s against the same peptides was assessed, demonstrating a lack of background/non-specific staining. Response of HD PBMCs was not performed for HLA B35 restricted peptides.

For staining of expanded CD8+ T lymphocytes, samples were thawed, treated

with DNAse for 10 minutes, washed and stained with MHC multimer panels for 15 minutes at 37°C. Subsequently, cells were stained with LIVE/DEAD® Fixable Near-IR Dead Cell Stain Kit for 633 or 635 nm excitation (Invitrogen, Life Technologies), CD8-PerCP (Invitrogen, Life Technologies) and FITC coupled antibodies to a panel of CD4, CD14, CD16, CD19 (all from BD Pharmingen) and CD40 (AbD Serotec) for an additional 20 minutes at 4°C. Data acquisition was performed on an LSR II flow cytometer (Becton Dickinson) with FACSDiva 6 software. Cutoff values for the definition of positive responses were ≥0.005% of total CD8+ cells and ≥10 events.

For patient L011, HLA-B3501 MTFR2D326Y -derived multimers were found to

bind the mutated sequence FAFQEYDSF (netMHC binding score: 22) but not the wild type sequence FAFQEDDSF (netMHC binding score: 10) (Fig 2 and Fig S5). No responses were found against overlapping peptides AFQEYDSFEK and KFAFQEYDSF. For patient L012, HLA-A1101 CHTF18L769V -derived multimers bound the mutated sequence LLLDIVAPK (netMHC binding score: 37) but not the wild type sequence: LLLDILAPK (netMHC binding score: 41) (Fig 2 and Fig S5). No responses were found against overlapping peptides CLLLDIVAPK and IVAPKLRPV. Finally, HLA-B0702 MYADMR30W-derived multimers bound the mutated sequence SPMIVGSPW (netMHC binding score: 15) as well as the wild type sequence SPMIVGSPR (netMHC binding score: 1329). No responses were found against overlapping peptides SPMIVGSPWA, SPMIVGSPWAL, SPWALTQPLGL and SPWALTQPL. MHC-multimer analysis and multi-parametric flow cytometric phenotyping of non-expanded tumor samples for L011 and L012

Tumor samples were thawed, washed and first stained with custom-made MHC-multimers for 10-15 minutes at 37°C in the dark. Cells were thereafter transferred

6

onto wet ice and stained for 30 minutes, in the dark, with a panel of surface antibodies used at the manufacturer’s recommended dilution: CD8-V500, SK1 clone (BD Biosciences), PD-1-BV605, EH12.2H7 clone (Biolegend), CD3-BV785, OKT3 clone (Biolegend), LAG-3-PE, 3DS223H clone (eBioscience). A fixable viability dye (eFlour780, eBioscience) was included the surface mastermix. Cells were permeablized for 20 minutes with use of an intracellular fixation and permeabilization buffer set from eBioscience. An intracellular staining panel was applied for 30 minutes, on ice, in the dark, and consisted of the following antibodies used at the manufacturer’s recommended dilution: granzyme B-V450, GB11 clone (BD Biosciences), FoxP3-PerCP-Cy5.5, PCH101 clone (eBioscience), Ki67-FITC, clone B56 (BD Biosciences) and CTLA-4 – APC, L3D10 clone (Biolegend). Data acquisition was performed on a BD FACSAria III flow cytometer (BD Biosciences) and analyzed in Flowjo version 10.0.8 (Tree Star Inc.).

7

Table S1 Clinical characteristics of multi-region NSCLC Patient ID

Age (years)

Gender Histology Lymph node(s)/ location

Stage (I-IV)

Regions sequenced

Smoking status (pack-years*)

L003 84 F LUAD 2/Station 4 IIIB R2 (RLL), R4 (RUL), LN

never- smoker

L008 75 M LUAD 2/Hilar IIIA R1 (RUL), R3 (RML), LN

ex- smoker (25)

L001† 59 F LUAD 3/Hilar IIA R1-R5, LN ex- smoker (10)

L004‡ 73 M Undiff. NSCLC

none IIB R1-R4 current smoker (50)

L011 49 F LUAD none IB R1-R3 current smoker (45)

L002 78 M LUAD/ LUSC

2/Station 5 IIIA R1-R4 current smoker (>50)

L012 69 F LUSC none IB R1-R3 current smoker (40)

Abbreviations: LLL, lower lobe; LUL, left upper lobe; RLL, right lower lobe; RUL, right upper lobe; RML, right middle lobe; R, region; LN, lymph node; Undiff, undifferentiated. * a pack-year is defined as the number of packs of cigarettes smoked per day multiplied by the number of years the person has smoked. †L001 presented a synchronous MEN1 syndrome-associated tumour, classified as separate tumour based on histological morphology, biochemical profile and octreotide scan imaging. ‡L004 presented a synchronous oesophageal adenocarcinoma, classified as separate primary tumours based on histological morphology and immunohistochemistry marker profile.

8

Table S2 Multivariate survival analysis

Neoantigens (without ITH threshold)

HR 95% CI lower

95% CI upper

p-value

Number neoantigens 0.996 0.992 1.000 0.025 Gender 0.675 0.372 1.226 0.200 Early stage (vs. late) 0.259 0.141 0.476 0.000 Age 1.015 0.985 1.045 0.330 ITH threshold =0 HR 95% CI

lower 95% CI

upper p-value

High neo and ITH<=0 0.291 0.069 1.237 0.095 Gender 0.660 0.362 1.204 0.180 Early stage (vs. late) 0.313 0.171 0.572 0.000 Age 1.017 0.989 1.047 0.230 ITH threshold =0.01 HR 95% CI

lower 95% CI

upper p-value

High neo and ITH<=0.01 0.262 0.103 0.667 0.005 Gender 0.619 0.339 1.132 0.120 Early stage (vs. late) 0.236 0.126 0.442 0.000 Age 1.017 0.989 1.047 0.230 ITH threshold =0.05 HR 95% CI

lower 95% CI

upper p-value

High neo and ITH<=0.05 0.366 0.172 0.781 0.009 Gender 0.660 0.363 1.200 0.174 Early stage (vs. late) 0.241 0.130 0.445 0.000 Age 1.012 0.982 1.042 0.432

9

Table S3 A) Differentially expressed immune genes between LUAD and LUSC tumor samples B) Expression of HLA genes between LUAD and LUSC tumor samples by neoantigen quartile C) Differentially expressed immune genes between LUSC normal and tumor samples Table is provided in Other Supplementary Material as an Excel file.

10

Table S4 A) Differentially expressed immune genes between tumors with a high neoantigen burden and low neoantigen ITH and remaining tumors B) Differentially expressed immune genes between tumors with a high clonal neoantigen burden (>upper quartile clonal neoantigens) and low clonal neoantigen burden (<lower quartile clonal neoantigens). Table is provided in Other Supplementary Material as an Excel file.

11

Table S5 Detailed clinical characteristics of patients from (2) Table is provided in Other Supplementary Material as an Excel file.

12

Fig. S1 TCGA LUAD tumors A) Relationship between overall survival and neoantigen burden using different peptide-MHC binding strengths. Log-rank p-values and hazard ratios are shown on plot. Notably, focusing only on weak-binders does not result in a significant relationship (log-rank, P=0.099). B) Overall survival curves for patients with tumors exhibiting high neoantigen ITH (≥ 1%; n = 86) compared to low neoantigen ITH (<1%, n = 53) (log-rank P = 0.061). C) Relationship between neoantigen ITH and total neoantigen burden.

13

Fig. S2 LUSC cohort summary. (A) Total putative neoantigen burden of TCGA LUSC patients. Columns colored to show proportion of neoantigens arising from clonal (blue) or subclonal (red) mutations. (B) Overall survival curves of patients with high neoantigen burden (n = 30) compared to those with a low neoantigen burden (n = 92) (log-rank P = 0.81), (C) high ITH (n = 65) compared to those with a low ITH (n = 57) (log-rank P = 0.76), and (D) Overall survival curves showing different ITH thresholds. For a schematic of ITH thresholds see Figure 1.

14

Fig. S3 Differential Gene Expression Analysis between LUAD and LUSC. A) Significantly differentially expressed immune genes between LUAD and LUSC tumor samples. Heatmap shows co-occurence matrix. B) Differential expression of HLA-A at different neoantigen burden quartiles. Notably, LUSC samples have significantly lower HLA-A expression within each quartile.

15

Fig. S4 Differential expression of immune-related genes.

A) Significantly differentially expressed immune-related genes between the LUAD tumors with high clonal neoantigen burden (blue) and those with low clonal neoantigen burden (red), defined as the bottom quartile of the cohort. The normalized

16

expression level of each gene in each tumor sample is displayed from -1 (purple) to +1 (green). B) Significantly differentially expressed genes in the high clonal neoantigen tumors (blue) as compared to the low clonal neoantigen tumors (red) are clustered by their level of co-expression using a metric of 1-r2, with the most highly correlated genes colored more lightly.

17

Fig. S5 Screening of expanded tumor-infiltrating CD8+ T cells and characterisation of non-expanded tumor-infiltrating neoantigen-reactive CD8+ T cells in patient L012 A) MHC-multimer screening of expanded, region-specific tumor-infiltrating CD8+ T lymphocytes identifies MTFR2D326Y -reactive CD8+ T cells in tumor regions 1-3 and at low frequency in normal lung tissue in patient L011. B) MHC-multimer screening of expanded, region-specific tumor-infiltrating CD8+ T lymphocytes identifies

18

CHTF18L769V and MYADMR30W-reactive CD8+ T cells at variable frequency in tumor regions 1-3 and at low frequency in normal lung tissue in patient L012. C) Binding of expanded tumor-infiltrating CD8+ T lymphocytes to MHC-multimers bearing mutant MTFR2 peptide for L011 and mutant CHTF18L769V /MYADMR30W peptide for L012, versus wild type peptide. Frequency of CD8+ peptide-reactive T cells out of total CD3+CD8+ TILs is displayed for A-C. D) Multi-parametric flow cytometric analysis of tumor-infiltrating T lymphocyte subsets isolated from L012 region 2. Phenotypic data is representative of all tumor regions. Relative expression of iCTLA-4 (intracellular CTLA-4), surface PD-1 and surface LAG-3 by CD4+FoxP3+ (regulatory T cell), CD4+FoxP3- (CD4 helper T cell), CD8+ multimer negative and CD8+ multimer-reactive (CD8+ CHTF18L769V + and CD8+ MYADMR30W+) T cells is displayed, plotted against iKi67 (intracellular Ki67). E) Co-expression of PD-1 and iGzmB (intracellular granzyme B) by tumor-infiltrating T lymphocyte subsets isolated from L012 region 2.

19

Fig. S6 Relationship of PD-L1 expression to clonal neoantigen burden and ITH in (2). PD-L1 exhibits significantly stronger expression in tumors harboring a high clonal neo-antigen burden (>= 70) and a low neoantigen ITH (<=5%) compared to tumors harboring a low clonal neoantigen burden or high subclonal neoantigen fraction.

20

Fig. S7 Neoantigen clonal architecture, clinical response and patient characteristics of (17) A)Samples are grouped according to response. Bar plot depicts clonal neoantigens in blue and subclonal neoantigens in red. Mutational signatures identified within each sample, subtype, expression of CTLA-4, LDH levels and whether patient received BRAF and/or DTIC therapy prior to anti-CTLA4 treatment is displayed below. Notably all patients with a strong Signature 11 received DTIC. Further clinical characteristics of each patient can be found in supplementary material of (17). B) Overall survival curves with and without an ITH threshold. .

21

References and Notes 1. H. Matsushita, M. D. Vesely, D. C. Koboldt, C. G. Rickert, R. Uppaluri, V. J. Magrini, C. D.

Arthur, J. M. White, Y. S. Chen, L. K. Shea, J. Hundal, M. C. Wendl, R. Demeter, T. Wylie, J. P. Allison, M. J. Smyth, L. J. Old, E. R. Mardis, R. D. Schreiber, Cancer exome analysis reveals a T-cell-dependent mechanism of cancer immunoediting. Nature 482, 400–404 (2012). Medline doi:10.1038/nature10755

2. N. A. Rizvi, M. D. Hellmann, A. Snyder, P. Kvistborg, V. Makarov, J. J. Havel, W. Lee, J. Yuan, P. Wong, T. S. Ho, M. L. Miller, N. Rekhtman, A. L. Moreira, F. Ibrahim, C. Bruggeman, B. Gasmi, R. Zappasodi, Y. Maeda, C. Sander, E. B. Garon, T. Merghoub, J. D. Wolchok, T. N. Schumacher, T. A. Chan, Mutational landscape determines sensitivity to PD-1 blockade in non-small cell lung cancer. Science 348, 124–128 (2015). Medline doi:10.1126/science.aaa1348

3. J. C. Castle, S. Kreiter, J. Diekmann, M. Löwer, N. van de Roemer, J. de Graaf, A. Selmi, M. Diken, S. Boegel, C. Paret, M. Koslowski, A. N. Kuhn, C. M. Britten, C. Huber, O. Türeci, U. Sahin, Exploiting the mutanome for tumor vaccination. Cancer Res. 72, 1081–1091 (2012). Medline doi:10.1158/0008-5472.CAN-11-3722

4. A. Snyder, V. Makarov, T. Merghoub, J. Yuan, J. M. Zaretsky, A. Desrichard, L. A. Walsh, M. A. Postow, P. Wong, T. S. Ho, T. J. Hollmann, C. Bruggeman, K. Kannan, Y. Li, C. Elipenahli, C. Liu, C. T. Harbison, L. Wang, A. Ribas, J. D. Wolchok, T. A. Chan, Genetic basis for clinical response to CTLA-4 blockade in melanoma. N. Engl. J. Med. 371, 2189–2199 (2014). Medline doi:10.1056/NEJMoa1406498

5. P. F. Robbins, Y. C. Lu, M. El-Gamil, Y. F. Li, C. Gross, J. Gartner, J. C. Lin, J. K. Teer, P. Cliften, E. Tycksen, Y. Samuels, S. A. Rosenberg, Mining exomic sequencing data to identify mutated antigens recognized by adoptively transferred tumor-reactive T cells. Nat. Med. 19, 747–752 (2013). Medline doi:10.1038/nm.3161

6. T. N. Schumacher, R. D. Schreiber, Neoantigens in cancer immunotherapy. Science 348, 69–74 (2015). Medline doi:10.1126/science.aaa4971

7. M. Greaves, Evolutionary determinants of cancer. Cancer Discovery 5, 806–820 (2015). Medline doi:10.1158/2159-8290.CD-15-0439

8. E. C. de Bruin, N. McGranahan, R. Mitter, M. Salm, D. C. Wedge, L. Yates, M. Jamal-Hanjani, S. Shafi, N. Murugaesu, A. J. Rowan, E. Grönroos, M. A. Muhammad, S. Horswell, M. Gerlinger, I. Varela, D. Jones, J. Marshall, T. Voet, P. Van Loo, D. M. Rassl, R. C. Rintoul, S. M. Janes, S. M. Lee, M. Forster, T. Ahmad, D. Lawrence, M. Falzon, A. Capitanio, T. T. Harkins, C. C. Lee, W. Tom, E. Teefe, S. C. Chen, S. Begum, A. Rabinowitz, B. Phillimore, B. Spencer-Dene, G. Stamp, Z. Szallasi, N. Matthews, A. Stewart, P. Campbell, C. Swanton, Spatial and temporal diversity in genomic instability processes defines lung cancer evolution. Science 346, 251–256 (2014). Medline doi:10.1126/science.1253462

9. M. Jamal-Hanjani, G. A. Wilson, S. Horswell, R. Mitter, O. Sakarya, T. Constantin, R. Salari, E. Kirkizlar, S. Sigurjonsson, R. Pelham, S. Kareht, B. Zimmermann, C. Swanton, Detection of ubiquitous and heterogeneous mutations in cell-free dna from patients with

early-stage non-small-cell lung cancer. Ann. Oncol. mdw037 (2016). 10.1093/annonc/mdw037 Medline doi:10.1093/annonc/mdw037

10. M. S. Lawrence, P. Stojanov, C. H. Mermel, J. T. Robinson, L. A. Garraway, T. R. Golub, M. Meyerson, S. B. Gabriel, E. S. Lander, G. Getz, Discovery and saturation analysis of cancer genes across 21 tumour types. Nature 505, 495–501 (2014). Medline doi:10.1038/nature12912

11. M. S. Rooney, S. A. Shukla, C. J. Wu, G. Getz, N. Hacohen, Molecular and genetic properties of tumors associated with local immune cytolytic activity. Cell 160, 48–61 (2015). Medline doi:10.1016/j.cell.2014.12.033

12. S. A. Shukla, M. S. Rooney, M. Rajasagi, G. Tiao, P. M. Dixon, M. S. Lawrence, J. Stevens, W. J. Lane, J. L. Dellagatta, S. Steelman, C. Sougnez, K. Cibulskis, A. Kiezun, N. Hacohen, V. Brusic, C. J. Wu, G. Getz, Comprehensive analysis of cancer-associated somatic mutations in class I HLA genes. Nat. Biotechnol. 33, 1152–1158 (2015). Medline doi:10.1038/nbt.3344

13. Materials and methods are available as supplementary materials on Science Online.

14. L. T. Nguyen, P. S. Ohashi, Clinical blockade of PD1 and LAG3—potential mechanisms of action. Nat. Rev. Immunol. 15, 45–56 (2015). Medline doi:10.1038/nri3790

15. S. R. Hadrup, A. H. Bakker, C. J. Shu, R. S. Andersen, J. van Veluw, P. Hombrink, E. Castermans, P. Thor Straten, C. Blank, J. B. Haanen, M. H. Heemskerk, T. N. Schumacher, Parallel detection of antigen-specific T-cell responses by multidimensional encoding of MHC multimers. Nat. Methods 6, 520–526 (2009). Medline doi:10.1038/nmeth.1345

16. S. Read, V. Malmström, F. Powrie, Cytotoxic T lymphocyte-associated antigen 4 plays an essential role in the function of CD25(+)CD4(+) regulatory cells that control intestinal inflammation. J. Exp. Med. 192, 295–302 (2000). Medline

17. E. M. Van Allen, D. Miao, B. Schilling, S. A. Shukla, C. Blank, L. Zimmer, A. Sucker, U. Hillen, M. H. Geukes Foppen, S. M. Goldinger, J. Utikal, J. C. Hassel, B. Weide, K. C. Kaehler, C. Loquai, P. Mohr, R. Gutzmer, R. Dummer, S. Gabriel, C. J. Wu, D. Schadendorf, L. A. Garraway, Genomic correlates of response to CTLA-4 blockade in metastatic melanoma. Science 350, 207–211 (2015). Medline doi:10.1126/science.aad0095

18. L. B. Alexandrov, S. Nik-Zainal, D. C. Wedge, S. A. Aparicio, S. Behjati, A. V. Biankin, G. R. Bignell, N. Bolli, A. Borg, A. L. Børresen-Dale, S. Boyault, B. Burkhardt, A. P. Butler, C. Caldas, H. R. Davies, C. Desmedt, R. Eils, J. E. Eyfjörd, J. A. Foekens, M. Greaves, F. Hosoda, B. Hutter, T. Ilicic, S. Imbeaud, M. Imielinski, N. Jäger, D. T. Jones, D. Jones, S. Knappskog, M. Kool, S. R. Lakhani, C. López-Otín, S. Martin, N. C. Munshi, H. Nakamura, P. A. Northcott, M. Pajic, E. Papaemmanuil, A. Paradiso, J. V. Pearson, X. S. Puente, K. Raine, M. Ramakrishna, A. L. Richardson, J. Richter, P. Rosenstiel, M. Schlesner, T. N. Schumacher, P. N. Span, J. W. Teague, Y. Totoki, A. N. Tutt, R. Valdés-Mas, M. M. van Buuren, L. van ’t Veer, A. Vincent-Salomon, N. Waddell, L. R. Yates, J. Zucman-Rossi, P. A. Futreal, U. McDermott, P. Lichter, M. Meyerson, S. M. Grimmond, R. Siebert, E. Campo, T. Shibata, S. M. Pfister, P. J. Campbell, M. R. Stratton, Australian Pancreatic Cancer Genome Initiative, ICGC Breast

Cancer Consortium, ICGC MMML-Seq Consortium, ICGC PedBrain, Signatures of mutational processes in human cancer. Nature 500, 415–421 (2013). Medline doi:10.1038/nature12477

19. B. E. Johnson, T. Mazor, C. Hong, M. Barnes, K. Aihara, C. Y. McLean, S. D. Fouse, S. Yamamoto, H. Ueda, K. Tatsuno, S. Asthana, L. E. Jalbert, S. J. Nelson, A. W. Bollen, W. C. Gustafson, E. Charron, W. A. Weiss, I. V. Smirnov, J. S. Song, A. B. Olshen, S. Cha, Y. Zhao, R. A. Moore, A. J. Mungall, S. J. Jones, M. Hirst, M. A. Marra, N. Saito, H. Aburatani, A. Mukasa, M. S. Berger, S. M. Chang, B. S. Taylor, J. F. Costello, Mutational analysis reveals the origin and therapy-driven evolution of recurrent glioma. Science 343, 189–193 (2014). Medline doi:10.1126/science.1239947

20. T. A. Yap, M. Gerlinger, P. A. Futreal, L. Pusztai, C. Swanton, Intratumor heterogeneity: Seeing the wood for the trees. Sci. Transl. Med. 4, 127ps10 (2012). Medline doi:10.1126/scitranslmed.3003854

21. J. Zhang, J. Fujimoto, J. Zhang, D. C. Wedge, X. Song, J. Zhang, S. Seth, C. W. Chow, Y. Cao, C. Gumbs, K. A. Gold, N. Kalhor, L. Little, H. Mahadeshwar, C. Moran, A. Protopopov, H. Sun, J. Tang, X. Wu, Y. Ye, W. N. William, J. J. Lee, J. V. Heymach, W. K. Hong, S. Swisher, I. I. Wistuba, P. A. Futreal, Intratumor heterogeneity in localized lung adenocarcinomas delineated by multiregion sequencing. Science 346, 256–259 (2014). Medline doi:10.1126/science.1256930

22. A. McKenna, M. Hanna, E. Banks, A. Sivachenko, K. Cibulskis, A. Kernytsky, K. Garimella, D. Altshuler, S. Gabriel, M. Daly, M. A. DePristo, The Genome Analysis Toolkit: A MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res. 20, 1297–1303 (2010). Medline doi:10.1101/gr.107524.110

23. H. Li, R. Durbin, Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754–1760 (2009). Medline doi:10.1093/bioinformatics/btp324

24. H. Li, B. Handsaker, A. Wysoker, T. Fennell, J. Ruan, N. Homer, G. Marth, G. Abecasis, R. Durbin; 1000 Genome Project Data Processing Subgroup, The Sequence Alignment/Map format and SAMtools. Bioinformatics 25, 2078–2079 (2009). Medline doi:10.1093/bioinformatics/btp352

25. D. C. Koboldt, Q. Zhang, D. E. Larson, D. Shen, M. D. McLellan, L. Lin, C. A. Miller, E. R. Mardis, L. Ding, R. K. Wilson, VarScan 2: Somatic mutation and copy number alteration discovery in cancer by exome sequencing. Genome Res. 22, 568–576 (2012). Medline doi:10.1101/gr.129684.111

26. A. Roth, J. Khattra, D. Yap, A. Wan, E. Laks, J. Biele, G. Ha, S. Aparicio, A. Bouchard-Côté, S. P. Shah, PyClone: Statistical inference of clonal population structure in cancer. Nat. Methods 11, 396–398 (2014). Medline doi:10.1038/nmeth.2883

27. J. T. Robinson, H. Thorvaldsdóttir, W. Winckler, M. Guttman, E. S. Lander, G. Getz, J. P. Mesirov, Integrative genomics viewer. Nat. Biotechnol. 29, 24–26 (2011). Medline doi:10.1038/nbt.1754

28. P. Van Loo, S. H. Nordgard, O. C. Lingjærde, H. G. Russnes, I. H. Rye, W. Sun, V. J. Weigman, P. Marynen, A. Zetterberg, B. Naume, C. M. Perou, A. L. Børresen-Dale, V.

N. Kristensen, Allele-specific copy number analysis of tumors. Proc. Natl. Acad. Sci. U.S.A. 107, 16910–16915 (2010). Medline doi:10.1073/pnas.1009843107

29. N. McGranahan, F. Favero, E. C. de Bruin, N. J. Birkbak, Z. Szallasi, C. Swanton, Clonal status of actionable driver events and the timing of mutational processes in cancer evolution. Sci. Transl. Med. 7, 283ra54 (2015). Medline doi:10.1126/scitranslmed.aaa1408

30. R. Rosenthal, deconstructSigs: Identifies Signatures Present in a Tumor Sample. R package version 1.6.0. https://github.com/raerose01/deconstructSigs (2016)

31. A. Szolek, B. Schubert, C. Mohr, M. Sturm, M. Feldhahn, O. Kohlbacher, OptiType: Precision HLA typing from next-generation sequencing data. Bioinformatics 30, 3310–3316 (2014). Medline doi:10.1093/bioinformatics/btu548

32. I. Hoof, B. Peters, J. Sidney, L. E. Pedersen, A. Sette, O. Lund, S. Buus, M. Nielsen, NetMHCpan, a method for MHC class I binding prediction beyond humans. Immunogenetics 61, 1–13 (2009). Medline doi:10.1007/s00251-008-0341-z

33. M. Nielsen, C. Lundegaard, T. Blicher, K. Lamberth, M. Harndahl, S. Justesen, G. Røder, B. Peters, A. Sette, O. Lund, S. Buus, NetMHCpan, a method for quantitative predictions of peptide binding to any HLA-A and -B locus protein of known sequence. PLOS ONE 2, e796 (2007). Medline doi:10.1371/journal.pone.0000796

34. M. Toebes, M. Coccoris, A. Bins, B. Rodenko, R. Gomez, N. J. Nieuwkoop, W. van de Kasteele, G. F. Rimmelzwaan, J. B. Haanen, H. Ovaa, T. N. Schumacher, Design and use of conditional MHC class I ligands. Nat. Med. 12, 246–251 (2006). Medline doi:10.1038/nm1360

35. A. H. Bakker, R. Hoppes, C. Linnemann, M. Toebes, B. Rodenko, C. R. Berkers, S. R. Hadrup, W. J. van Esch, M. H. Heemskerk, H. Ovaa, T. N. Schumacher, Conditional MHC class I ligands and peptide exchange technology for the human MHC gene products HLA-A1, -A3, -A11, and -B7. Proc. Natl. Acad. Sci. U.S.A. 105, 3825–3830 (2008). Medline doi:10.1073/pnas.0709717105

36. T. M. Frøsig, J. Yap, T. Seremet, R. Lyngaa, I. M. Svane, P. Thor Straten, M. H. Heemskerk, G. M. Grotenbreg, S. R. Hadrup, Design and validation of conditional ligands for HLA-B*08:01, HLA-B*15:01, HLA-B*35:01, and HLA-B*44:05. Cytometry A 87, 967–975 (2015). Medline doi:10.1002/cyto.a.22689

37. C. X. Chang, A. T. Tan, M. Y. Or, K. Y. Toh, P. Y. Lim, A. S. Chia, T. M. Froesig, K. D. Nadua, H. L. Oh, H. N. Leong, S. R. Hadrup, A. J. Gehring, Y. J. Tan, A. Bertoletti, G. M. Grotenbreg, Conditional ligands for Asian HLA variants facilitate the definition of CD8+ T-cell responses in acute and chronic viral diseases. Eur. J. Immunol. 43, 1109–1120 (2013). Medline doi:10.1002/eji.201243088

38. R. S. Andersen, P. Kvistborg, T. M. Frøsig, N. W. Pedersen, R. Lyngaa, A. H. Bakker, C. J. Shu, P. Straten, T. N. Schumacher, S. R. Hadrup, Parallel detection of antigen-specific T cell responses by combinatorial encoding of MHC multimers. Nat. Protoc. 7, 891–902 (2012). Medline doi:10.1038/nprot.2012.037