The landscape of T cell infiltration in human cancer and its association with antigen presenting gene expression

TLDR: A novel mRNA-based T cell infiltration score (TIS) and profile infiltration levels in 19 tumor types are defined and it is found that clear cell renal cell carcinoma (ccRCC) is the high for TIS and among the highest for the correlation between TISand APM expression, despite a modest mutation burden.

Abstract: Infiltrating T cells in the tumor microenvironment have crucial roles in the competing processes of pro-tumor and anti-tumor immune response. However, the infiltration level of distinct T cell subsets and the signals that draw them into a tumor, such as the expression of antigen presenting machinery (APM) genes, remain poorly characterized across human cancers. Here, we define a novel mRNA-based T cell infiltration score (TIS) and profile infiltration levels in 19 tumor types. We find that clear cell renal cell carcinoma (ccRCC) is the highest for TIS and among the highest for the correlation between TIS and APM expression, despite a modest mutation burden. This finding is contrary to the expectation that immune infiltration and mutation burden are linked. To further characterize the immune infiltration in ccRCC, we use RNA-seq data to computationally infer the infiltration levels of 24 immune cell types in a discovery cohort of 415 ccRCC patients and validate our findings in an independent cohort of 101 ccRCC patients. We find three clusters of tumors that are primarily separated by levels of T cell infiltration and APM gene expression. In ccRCC, the levels of Th17 cells and the ratio of CD8+ T/Treg levels are associated with improved survival whereas the levels of Th2 cells and Tregs are associated with negative clinical outcome. Our analysis illustrates the utility of computational immune cell decomposition for solid tumors, and the potential of this method to guide clinical decision-making.

Figures

Fig. 3. Pan-cancer analysis of T cell infiltration and its association with antigen presenting machinery (APM) gene expression (a) T cell infiltration scores (TIS) and the corresponding mutation load in 19 tumor types. TIS is an aggregate score obtained as the average of nine distinct T cell subset scores (CD8+ T, Th1, Th2, Th17, Treg, T effector memory, T central memory, T helper, and T cells). Each circle in the top panel shows the TIS for a

Fig. 4. Characterization of immune infiltration clusters in ccRCC (a) Unsupervised clustering of 415 ccRCC patients from the TCGA cohort using ssGSEA scores from 24 immune cell types, 3 immunotherapy targets (PD-1, PDL1, CTLA-4), and angiogenesis. Hierarchical clustering was performed with Euclidean distance and Ward linkage. We discover three distinct immune infiltration clusters, here termed 1) non-infiltrated, 2) heterogeneously infiltrated, and 3) T cell enriched. The T cell enriched cluster is characterized by tumors with high APM scores and high granzyme B and interferon gamma mRNA expression levels. (b) Differential expression analysis with Mann-Whitney tests for all genes in the TCGA RNA-seq dataset excluding signature genes. Only genes that are significantly overexpressed in one cluster at a q-value cutoff of 5x10-5 are shown. Pathway analysis using DAVID(38) reveals that the genes overexpressed in the three clusters (N=1110, 181, and 277 respectively) are enriched in 1) adaptive

Fig. 2. Validation of the immune cell scoring method (a) Principal component (PC) analysis on immune cell types using transcript levels of signature genes. Microarray gene expression data were generated from immune cell types sorted by magnetic or fluorescent activated cell sorting(28-30) and used in Bindea et al.(14) to derive the signatures. (b) Immunofluorescence validation in MSKCC cohort. ssGSEA scores were computed using RNA-seq data to assess the infiltration levels of NK cells, CD8+ T cells, and Treg cells in 10 ccRCC tumors (top left panel). The infiltration levels of the same cell types were probed by immunofluorescence (IF) using the CD56, CD8 and FOXP3 antibodies respectively. IF staining is shown for two samples that are at opposite ends of the unsupervised hiearchy (bottom left panel). The association of the immune infiltrate levels inferred by these two orthogonal methods (ssGSEA and IF) is shown in the right panel where each dot represents a sample. Spearman correlations are given in the scatter plots. The IF staining level for a given sample was determined as the average across three representative regions on the slide. (c) Validation of ssGSEA scores with transcript levels of immune cell markers.

Fig. 6. Prognostic significance of ccRCC immune infiltration classes and distinct T cell subsets (a) Kaplan-Meier curves for cancer-specific survival in ccRCC immune infiltration classes. The T cell enriched class has the poorest survival whereas the non-

Fig. 5. Subclustering within the T cell enriched cohort demonstrates gene expression and survival differences (a) Hierarchical clustering within the T cell enriched cohort revealed two distinct subclusters, here termed TCa and TCb, that had differences in immune cell

Fig. 1. Workflow and the investigated immune cell types. (a) Workflow for tumor immune-infiltrate profiling. Gene signatures for 24 immune cell types were

Full text

!
"!
The landscape of T cell infiltration in human cancer and
its association with antigen presenting gene expression
Yasin Şenbabaoğlu
1†
*, Andrew G. Winer
2†
, Ron S. Gejman
3,11†
, Ming Liu
4
,
Augustin Luna
1
, Irina Ostrovnaya
5
, Nils Weinhold
1
, William Lee
1,6
, Samuel D.
Kaffenberger
2
, Ying Bei Chen
7
, Martin H. Voss
8
, Jonathan A. Coleman
2
, Paul
Russo
2
, Victor E. Reuter
7
, Timothy A. Chan
6,9,11
, Emily H. Cheng
7,9
, David A.
Scheinberg
3,10,11
, Ming O. Li
4
, James J. Hsieh
8,9
‡
, Chris Sander
1
*
‡
, A. Ari
Hakimi
1,2
*
‡
† co-first authors
‡ co-last authors
* corresponding authors
Memorial Sloan Kettering Cancer Center
1
Computational Biology Center
2
Urology Service, Department of Surgery
3
Molecular Pharmacology and Chemistry Program
4
Immunology Program
5
Department of Epidemiology and Biostatistics
6
Department of Radiation Oncology
7
Department of Pathology
8
Genitourinary Oncology, Department of Medicine
9
Human Oncology & Pathogenesis Program
10
Department of Medicine
11
Weill Cornell Medical College
Correspondence to: msk.immune.decomposition@gmail.com
One sentence summary: In silico decomposition of the immune
microenvironment among common tumor types identified clear cell renal cell
carcinoma as the most highly infiltrated by T-cells and further analysis of this
tumor type revealed three distinct and clinically relevant clusters which were
validated in an independent cohort.
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted September 1, 2015. ; https://doi.org/10.1101/025908doi: bioRxiv preprint

!
#!
Abstract
Infiltrating T cells in the tumor microenvironment have crucial roles in the
competing processes of pro-tumor and anti-tumor immune response. However,
the infiltration level of distinct T cell subsets and the signals that draw them into a
tumor, such as the expression of antigen presenting machinery (APM) genes,
remain poorly characterized across human cancers. Here, we define a novel
mRNA-based T cell infiltration score (TIS) and profile infiltration levels in 19
tumor types. We find that clear cell renal cell carcinoma (ccRCC) is the highest
for TIS and among the highest for the correlation between TIS and APM
expression, despite a modest mutation burden. This finding is contrary to the
expectation that immune infiltration and mutation burden are linked. To further
characterize the immune infiltration in ccRCC, we use RNA-seq data to
computationally infer the infiltration levels of 24 immune cell types in a discovery
cohort of 415 ccRCC patients and validate our findings in an independent cohort
of 101 ccRCC patients. We find three clusters of tumors that are primarily
separated by levels of T cell infiltration and APM gene expression. In ccRCC, the
levels of Th17 cells and the ratio of CD8
+
T/Treg levels are associated with
improved survival whereas the levels of Th2 cells and Tregs are associated with
negative clinical outcome. Our analysis illustrates the utility of computational
immune cell decomposition for solid tumors, and the potential of this method to
guide clinical decision-making.
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted September 1, 2015. ; https://doi.org/10.1101/025908doi: bioRxiv preprint

!
$!
Introduction
Tumors are complex environments, composed of transformed cells as well as
stromal and immune infiltrates. Tumor-infiltrating cells can demonstrate either
tumor-suppressive or tumor-promoting effects, depending on the cancer type or
the tumor model. For instance, regulatory T cells (Tregs) and tumor associated
macrophages (TAMs) have been associated with pro-tumor functions(1-3),
whereas CD8
+
T cells have been associated with improved clinical outcomes and
response to immunotherapy(4-8). Antitumor activity of antigen-specific CD8
+
T
cells may underlie the efficacy of immune checkpoint blockade therapy(9-11) as
such CD8
+
T cells have been shown to increase in quantity and activity after
treatment with these drugs.
CD8
+
T cells are activated by peptide antigens presented on major
histocompatibility class I (MHC-I) molecules. A CD8
+
T cell can proliferate when
its T cell receptor (TCR) recognizes antigens presented by MHC-I on a target
cell, leading to an antigen-specific immune response that kills antigen bearing
cells(12). All nucleated cells express antigen presenting machinery (APM) genes
that code for MHC-I subunits and proteins necessary to process antigens and
load them onto MHC-I. The APM genes can be upregulated by type II interferon
(IFNγ), which is secreted by activated CD8
+
T cells and other immune infiltrates.
Upregulation of APM genes can lead to a cytotoxic feed-forward loop: more
antigen presentation increases the number of T cells that find their cognate
antigens, which in turn increases IFNγ release, antigen presentation and
cytotoxicity. Yet, identification of CD8
+
T cells alone is not sufficient to
characterize the cytotoxic potential of the complex tumor microenvironment. The
net inflammatory nature of the tumor can better be understood by quantifying the
infiltration levels of diverse immune cell types.
Tumor immune infiltrates have largely been characterized by tissue-based
approaches such as immunohistochemistry (IHC) and flow cytometry. These
approaches are limited by a number of factors including the number of cell types
that can be assayed simultaneously and the amount of tissue required.
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted September 1, 2015. ; https://doi.org/10.1101/025908doi: bioRxiv preprint

!
%!
Computational techniques applied to gene expression profiles of bulk tumors can
rapidly provide a broader perspective on the intratumoral immune landscape (13,
14).
ccRCC has been shown to be a highly immune-infiltrated tumor in multiple
clinical and genomic studies(15, 16).!A recent study found that transcript levels of
two genes expressed by cytolytic cells (GZMA and PRF1) were highest in clear
cell renal cell carcinoma (ccRCC) when compared to 17 other human cancers
(13). The spontaneous regression seen in up to 1% of ccRCC cases is also
thought to be largely immune-mediated(17). Additionally, ccRCC was historically
one of the first malignancies to respond to immunotherapy, and continues to be
among the most responsive (18-21). However, the mechanisms underlying high
immune infiltration, spontaneous remissions and response to immunotherapy in
this malignancy remain poorly understood.
The success of immune checkpoint blockade in melanoma and non-small cell
lung cancer has largely been attributed to the high mutation burden in these
tumors(10, 11). A higher number of tumor mutations is expected to result in
greater numbers of MHC binding neo-antigens that have been proposed to drive
tumor immune-infiltration and response to immunotherapy (10, 13, 22, 23).
However, the modest mutation load of ccRCC compared with other
immunotherapy-responsive tumor types(24) challenges the notion that neo-
antigens alone can drive immune infiltration and response to immunotherapy in
tumors.
As depicted in the workflow in Fig. 1a, we employed 24 immune cell type-specific
gene signatures from Bindea et al.(14) (Fig. 1b) to computationally infer the
infiltration levels in tumor samples (Step 1). We validated the gene signatures
and our inference methodology using a ccRCC cohort from our institution (Step
2). We then defined a T cell infiltration score (TIS), an overall immune infiltration
score (IIS) and an APM score to highlight the immune response differences
between ccRCC(25) and 18 other tumor types profiled by The Cancer Genome
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted September 1, 2015. ; https://doi.org/10.1101/025908doi: bioRxiv preprint

!
&!
Atlas (TCGA) research network (Step 3). Next, we characterized the immune-
infiltration patterns in ccRCC patients by using the levels of 24 immune cells,
angiogenesis, and expression of immunotherapeutic targets such as PD-1, PD-
L1 and CTLA-4 (Step 4). We then investigated a suite of mechanisms that could
potentially drive tumor immune-infiltration and explain the observed infiltration
patterns in ccRCC. Finally, we validated our findings in an independent multi-
platform ccRCC dataset(26) (Step 5). This integrative study utilizing rich whole-
exome, whole-transcriptome, proteomic, and clinical data substantially improves
our understanding of the tumor microenvironment in ccRCC and establishes an
approach that can easily be extended to other human cancers.
Fig. 1. Workflow and the investigated immune cell types. (a) Workflow for
tumor immune-infiltrate profiling. Gene signatures for 24 immune cell types were
1"
Immune cells sorted with flow cytometry
Gene expression profiling
Gene signatures for immune cell types
Bindea et al. (Immunity, 2013)
(Chtanova et al. 2005, Hyrcza et al. 2007, Wendt et al. 2006)
TCGA ccRCC discovery cohort
• Discovery and characterization of immune
infiltration clusters with distinct molecular and
prognostic characteristics
TCGA PANCAN19 analysis
• Computation of per patient immune cell scores in
19 cancer types using RNA-seq data
• Definition of and profiling patients for immune
infiltration score (IIS) and T cell infiltration score
(TIS)
In vitro validation by immunofluorescence
Sato et al. ccRCC validation cohort
• Validation of molecular and prognostic features of
ccRCC immune infiltration clusters
2"
3"
4"
5"
T cells (19)
T helper cells (24)
Th1
(27)
Th2
(26)
Th17
(3)
Treg
(1)
T follicular
helper (32)
CD3/TCR
T  cells (6)
CD3/TCR
CD3/TCR
CD4
+
CD8
+
Effector cells
(cytotoxic)
CD3/TCR
CD8
+
T cells (37)
Memory cells
Myeloid progenitor
Megakaryocytes
Lymphoid progenitor
Erythrocytes Mast cells (30) Myeloblasts
Basophils
Neutrophils (31) Eosinophils (31) Monocytes
Macrophages (33) Dendritic cells (7)
Immature
DCs (31)
Activated
DCs (5)
Plasmacytoid
DCs (1)
Natural killer
cells (35)
Small
lymphocytes
NK
CD56dim
cells (15)
NK
CD56bright
cells (11)
B cells
(33)
T cells
(19)
T cell
subsets
Memory cells
T central
memory (39)*
T central
memory (39)*
T effector
memory (15)*
T effector
memory (15)*
CD3
+
* T central and effector memory cells were not gated for CD4 or CD8.
Innate immunity cells Adaptive immunity cells Other
Angiogenesis (41)
Cytotoxic cells (17)**
** The cytotoxic cell signature includes genes overexpressed in CD8
+
T cells, T  cells, and NK cells.
Figure 1
a. Workflow for tumor immune-infiltrate profiling b. Investigated immune cell types and gene signatures
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted September 1, 2015. ; https://doi.org/10.1101/025908doi: bioRxiv preprint

Frequently asked questions

Q1. What contributions have the authors mentioned in the paper "The landscape of t cell infiltration in human cancer and its association with antigen presenting gene expression" ?

Memorial Sloan Kettering Cancer Center 1 Computational Biology Center 2 Urology Service, Department of Surgery 3 Molecular Pharmacology and Chemistry Program 4 Immunology Program 5 Department of Epidemiology and Biostatistics 6 Department of Radiation Oncology 7 Department of Pathology 8 Genitourinary oncology, Department Of Medicine 9 Human Oncolonization & Pathogenesis Program 10 Department of Medicine this paper