REvIEWS

Approaches to treat immune hot,

altered and cold tumours with
combination immunotherapies
Jérôme Galon   * and Daniela Bruni
Abstract | Immunotherapies are the most rapidly growing drug class and have a major impact in
oncology and on human health. It is increasingly clear that the effectiveness of immunomodulatory
strategies depends on the presence of a baseline immune response and on unleashing
of pre-​existing immunity. Therefore, a general consensus emerged on the central part
played by effector T cells in the antitumour responses. Recent technological, analytical and
mechanistic advances in immunology have enabled the identification of patients who are more
likely to respond to immunotherapy. In this Review , we focus on defining hot, altered and cold
tumours, the complexity of the tumour microenvironment, the Immunoscore and immune
contexture of tumours, and we describe approaches to treat such tumours with combination
immunotherapies, including checkpoint inhibitors. In the upcoming era of combination
immunotherapy , it is becoming critical to understand the mechanisms responsible for hot,
altered or cold immune tumours in order to boost a weak antitumour immunity. The impact of
combination therapy on the immune response to convert an immune cold into a hot tumour will
be discussed.

TNM system
The tumour-​node-metastasis
(TNM) staging system is
a globally recognized
classification of tumours based
on their anatomical extent.
T refers to the size and extent
of the primary tumour, N refers
to the involvement of regional
lymph nodes and M describes
the presence of distant
metastases.
INSERM, Laboratory of
Integrative Cancer
Immunology, Sorbonne
Université, Sorbonne Paris
Cité, Université Paris
Descartes, Université Paris
Diderot, Centre de Recherche
des Cordeliers, Paris, France.
*e-​mail: jerome.galon@
crc.jussieu.fr
https://doi.org/10.1038/
s41573-018-0007-y
The remarkable results achieved in the past few years
with the advent of cancer immunotherapies and check-
point inhibitors have revolutionized the field of oncology
by putting the host immune response under the spotlight
as a target for anticancer therapeutic interventions. The
founding principles of the immunotherapy of cancer
are threefold: first, the demonstration of immuno-​
surveillance using immune-​deficient mouse models1,2;
second, the demonstration of the major importance of
pre-​existing immunity and natural intratumoural T cells
in humans3; and third, the unleashing of pre-​existing
immunity via inhibition of checkpoint receptors on
T cells in human cancers4–8. Insight knowledge of the
basic mechanisms responsible for the establishment and
development of tumours, on the basis of their interac-
tion with and control of the host immune system, has
enabled us to draw a more comprehensive picture of the
possible points of intervention and has provided us with
reasons that might account for therapeutic failure. An
update in the current guidelines for tumour classifica-
tion and subsequent treatment is therefore becoming
a pressing necessity. The recent yet unofficial classifi-
cation of tumours into two categories, ‘hot’ and ‘cold’,
has been increasingly advocated. In this Review, we aim
to suggest a more comprehensive main four-​category
classification of tumours — hot, altered-​excluded,
altered-​immunosuppressed and cold — and to provide
an overview of both current and potential therapeutic
strategies to best target these four categories of tumour,
which in most cases involve combinatorial immunother-
apy strategies. We believe that a rational, standardized
and harmonized approach embracing the central role
of the immune system must be adopted to guide ther-
apeutic decisions. This adoption will require a general
consensus and a large collective effort, as anything of
great value.
Definition of hot and cold tumours
Current knowledge of the tumour–immune system
interaction has already set the foundations for a
rationally guided stratification of patients and ther-
apeutic strategies. A powerful concept for patient
stratification came with the observation that the
type, density and location of immune cells within the
tumour site could predict survival in colorectal cancer
(CRC) more accurately than the classical TNM system
for the first time in any type of cancer9. This concept
led to the development and implementation of the
Immunoscore3,10–13 — a robust, consensus, standard-
ized scoring system based on the quantification of
two lymphocyte populations (CD3 and CD8) both
at the tumour centre and the invasive margin 14,15.

Nature Reviews | Drug Discovery

Reviews
Pathologic T (pT) stage
The staging assigned post-​
surgery to guide treatment
stratification, patient selection
for clinical trials and prognosis
prediction (as opposed to
clinical staging that relies
on physical exams and
imaging tests).
The Immunoscore ranges from Immunoscore 0 (I0, for
low densities, such as absence of both cell types in
both regions) to I4 (high immune cell densities
in both locations). By classifying cancers according
to their immune infiltration, the system proposed for
the first time an immune-​based, rather than a cancer-​
based, classification of tumours3, de facto introducing
the notion of ‘hot’ (highly infiltrated, Immunoscore I4)
and ‘cold’ (non-​infiltrated, Immunoscore I0) tumours
(Fig.  1) . Tumour progression (T stage) and inva-
sion (N stage) were dependent on the pre-​e xisting
adaptive intratumour immunity 3,9. The consensus
Immunoscore has been validated globally in colon
cancer and has a greater relative prognostic value
than pathologic T (pT) stage, pN stage, lymphovascular
invasion, tumour differentiation and microsatellite
instability (MSI) status16.
On the basis of these findings, the novel concept of
‘immune contexture’ was proposed13 and adopted to
refer to the combination of immune variables associat-
ing the nature, density, immune functional orientation
and distribution of immune cells within the tumour13.
These immune contexture parameters are associated
with long-​term survival and prediction of response to
treatments10. In 2009, Camus et al. first described three
major immune coordination profiles (hot, altered and
cold) observed within primary CRCs, which enabled
classification according to the balance between tumour
escape and immune coordination17. The 2-year risk of
relapse for these three types of tumour was 10%, 50%
and 80%, respectively. The altered phenotype was fur-
ther divided into two distinct patterns — ‘excluded’ and
‘immunosuppressed’17. In some cases, T cells are found
at the edge of tumour sites (invasive margin) without
being able to infiltrate them. This ‘excluded’ phenotype
reflects the intrinsic ability of the host immune system
to effectively mount a T cell-​mediated immune response
and the ability of the tumour to escape such response by
physically hindering T cell infiltration (Fig. 1a). In other
cases, tumour sites display a low degree of immune
infiltration (Fig.  1a) , which suggests the absence of
physical barriers and the presence of an immuno­
suppressive environment that limits further recruitment17
and expansion18; this can be defined as an ‘immuno­
suppressed’ phenotype17. By more suitably simplifying
the complexity of the tumour phenotype spectrum, these
four characteristic subgroups can potentially represent a
practical tool to direct therapeutic intervention (Box 1;
Fig. 1b). Immunoscore proved to be a better prognostic
tool for patients with CRC than MSI19, which is currently
tested to predict the response of these patients to anti-​
programmed cell death protein 1 (PD-1) therapy20. A
worldwide consensus Immunoscore study validated the
prediction of risk of recurrence and survival on the basis
of the three main subtypes of tumour — the immune
hot, altered and cold tumours15,16.
Currently, the terms ‘hot’ and ‘cold’ are routinely used
to refer to T cell-​infiltrated, inflamed but non-​infiltrated,
and non-​inflamed tumours, reflecting well the higher
(I4) and lower (I0) Immunoscore categories3 (Fig. 1). This
immune classification has been validated in other can-
cer types such as melanoma21. Apart from the presence
of tumour-​infiltrating lymphocytes (TILs), additional
features such as the expression of anti-​programmed
death-​ligand 1 (PD-​L1) on tumour-​associated immune
cells, possible genomic instability and the presence of
a pre-​existing antitumour immune response have been
described as characteristics of hot tumours22. Conversely,
apart from being poorly infiltrated, cold tumours have
also been described to be immunologically ignorant
(scarcely expressing PD-​L1) and characterized by high
proliferation with low mutational burden (low expres-
sion of neoantigens) and low expression of antigen
presentation machinery markers such as major histo-
compatibility complex class I (MHC I)22. In an attempt to
propose a more simplistic yet comprehensive classifica-
tion, the four proposed types of tumour (hot, excluded,
immunosuppressed and cold), based on Immunoscore,
could be the first routine immune parameters to eval-
uate at the time of diagnosis. Recently, a novel theory
of cancer evolution at the metastatic stage was demon-
strated and highlighted a model of tumour evolution in
which a parallel immune selection exists, with a major
role of Immunoscore and T cells in this process23. The
fact that T cells are currently widely recognized as
the key fighters in the antitumour battle makes the use
of the Immunoscore an attractive option to help in guid-
ing treatment selection. Of course, this option does not
exclude the possible use of additional parameters, which
in fact are required to gain further insights into the spe-
cifics of each case, but its routine incorporation into clin-
ical practice could definitely relieve the existing, urging
need for therapeutic guidance. The limitations associ-
ated with the current techniques for immune monitoring
are discussed in Box 2.
Beyond hot, altered and cold tumours
The distinction between hot, altered (excluded and
immunosuppressed) and cold tumours is based on the
cytotoxic T cell landscape within a tumour. This power-
ful simplification reflects the outcome of a tremendously
complex interplay between the tumour and the immune
system. For instance, apart from T cells, high expression
of markers of B and follicular helper T (TFH) cells was
also correlated with a significantly prolonged disease-​
free survival time in patients with CRCs24. By secreting
CXC-​chemokine ligand 13 (CXCL13), TFH cells activate
a positive loop that involves increased densities of B, TFH,
T helper 1 (TH1), cytotoxic and memory T cells in breast
cancer25 and CRC24. Acquisition of a memory phenotype
by these adaptive immune cells long term can extend the
survival of patients24.
The immune functional orientation associated with
the immune contexture was first described in CRC and
showed the major importance of T cells (TH1 and cyto-
toxic T cells) and associated factors (interferon-​γ (IFNγ),
granulysin (GNLY), perforin (PRF1) and granzymes
(GZMs)3). These immune signatures were associated
with prolonged survival and validated in other cancer
types26–31. A similar T cell-​inflamed gene expression pro-
file exhibited predictive utility in identifying responders
to treatment with an anti-​PD-1 antibody32. These obser-
vations support the notion that these signatures are both
prognostic and predictive10.
www.nature.com/nrd
 Reviews

a
Optimal: high Immunoscore (inflamed, hot) Absent: low Immunoscore (non-inflamed, cold)

Altered: intermediate Immunoscore Altered: intermediate Immunoscore

Excluded Immunosuppressed

b
Absent Altered Optimal
Low Immunoscore Intermediate Immunoscore High Immunoscore

Cold Excluded Immunosuppressed Hot

Non-inflamed CT-Lo, Hi-IM Inflamed

Response to T cell checkpoint inhibition

Fig. 1 | Defining ‘hot’, ‘altered’ and ‘cold’ immune tumours — immunoscore as a new approach for the classification
of cancer. a | Illustrative example of hot, altered and cold immune tumours. Brown (3,3ʹ-diaminobenzidine (DAB)) staining
represents CD3+ T cells and blue (alkaline phosphatase) counterstaining provides homogeneous tissue background
staining. The level and spatial distribution of CD3+ and CD8+ T cell infiltration differentiates four distinct solid tumour
phenotypes: hot (or inflamed); altered, which can be excluded or immunosuppressed; and cold (or non-​inflamed). These
tumour phenotypes are characterized by high, intermediate and low Immunoscore, respectively. b | Schematic
representation of the four subtypes of immune tumour. Of note, in altered-​excluded tumours, CD3+ and CD8+ T cell
infiltrates are low at the tumour centre and high at the invasive margin, resulting overall in an intermediate Immunoscore.
Altered-​immunosuppressed tumours display instead a more uniform pattern of (low) CD3+ and CD8+ T cell infiltration.
CT, centre of tumour ; Hi, high; IM, invasive margin; Lo, low.

Activated lymphocytes, such as cytotoxic T cells, are
among the main sources of IFNγ. By increasing MHC I
and immunoproteasome expression in tumour cells, IFNγ
enhances antigen presentation and subsequent immuno-
surveillance33. IFNγ also attenuates cancer cell growth,
and it may exert pro-​apoptotic and anti-​proliferative
effects via induction of several interferon-​stimulated
genes (ISGs)33,34, although these effects seem to be dose-​
dependent35. It should be noted that IFNγ can exert
both antitumour and pro-​tumour functions, depending

Nature Reviews | Drug Discovery

Reviews

Box 1 | T cell-​based classification of tumours excluded or immuno­suppressed) tumour. In turn, the

local adaptive immunosuppression is triggered by tumour-​
characteristics of hot immune tumours specific T cells that are infiltrated in combinations being
• High degree of T cell and cytotoxic T cell infiltration (high Immunoscore) tested in cold immune tumou tumours. By producing
• Checkpoint activation (programmed cell death protein 1 (PD-1), cytotoxic T IFNγ, TILs can induce immune checkpoint molecules
lymphocyte-​associated antigen 4 (CTLA4), T cell immunoglobulin mucin receptor 3 (such as PD-L1) or immunosuppressive factors such as
(TIM3) and lymphocyte activation gene 3 (LAG3)) or otherwise impaired T cell indoleamine-​pyrrole 2,3-dioxygenase 1 (IDO1). Thus,
functions (for example, extracellular potassium-​driven T cell suppression) local adaptive immuno­suppression can generate hot or
Characteristics of altered-​immunosuppressed immune tumours immunosuppressed tumours, depending on the power of
• Poor, albeit not absent, T cell and cytotoxic T cell infiltration (intermediate Immunoscore) the driving mechanisms.
• Presence of soluble inhibitory mediators (transforming growth factor-​β (TGFβ), Great diversity in the TME composition exists across
interleukin 10 (IL-10) and vascular endothelial growth factor (VEGF)) different cancer types but also among patients with the
• Presence of immune suppressive cells (myeloid-​derived suppressor cells and same cancer and even in different tumour sites within
regulatory T cells) the same patient50,51. This diversity results from the com-
• Presence of T cell checkpoints (PD-1, CTLA4, TIM3 and LAG3) bination of a multitude of factors, including the occur-
rence of specific driver mutations and deregulation
Characteristics of altered-​excluded immune tumours
of oncogenes in cancer cells44,52; the load and quality of
• No T cell infiltration inside the tumour bed; accumulation of T cells at tumour borders passenger mutations in cancer cells19,53; the presence
(invasive margin) (intermediate Immunoscore)
of immunosuppressive components in the TME, whether
• Activation of oncogenic pathways soluble (such as transforming growth factor-​β (TGFβ)54)
• Epigenetic regulation and reprogramming of the tumour microenvironment or cell-​associated (such as PD-1 and/or PD-​L1 (ref.55));
• Aberrant tumour vasculature and/or stroma the presence of factors directing immune attraction24,56,57;
• Hypoxia and the presence of factors (such as interleukin 15
characteristics of cold immune tumours (IL-15)) that mediate the expansion and proliferation
• Absence of T cells within the tumour and at the tumour edges (low Immunoscore)
of cytotoxic CD8+ T cells18. This complex scenario is
further complicated by the fact that the TME evolves
• Failed T cell priming (low tumour mutational burden, poor antigen presentation and
intrinsic insensitivity to T cell killing)
with disease progression and recurrence24. Future stud-
ies will have to reveal advanced biomarkers to groups of
patients who display similar features, thereby creating
on the cellular, microenvironmental and/or molecu- specific models that can guide the choice of therapeutic
lar context35–37. Furthermore, prolonged IFNγ signal- regimen as well as therapeutic development. Some of
ling and enduring antigen exposure may also have an the challenges associated with the approaches used to
Immunoproteasome
Proteasome isoform
immuno­suppressive role38,39, leading to T cell exhaustion characterize the TME are discussed in Box 2.
constitutively expressed in and resistance to immune checkpoint blockade39.
haematopoietic cells and CD4+ T cells also contribute to antitumour immu- Unleashing the pre-​existing immunity
induced in non-​immune cells nity. By expressing the transcription factor T-​bet (also Effectiveness of immunomodulatory strategies depends
following exposure to
interferon-​γ (IFNγ) and other
known as T-​box transcription factor TBX21) and pro- on the presence of a baseline antitumour immune
pro-​inflammatory cytokines ducing IFNγ, TH1 cells hinder neo-​angiogenesis by response, involving either tumour-​associated6,7,10,58–62
(type I interferons and tumour inhibiting vascular endothelial growth factor (VEGF)- or circulating immune components8. Clinical response
necrosis factor (TNF)). It is producing tumour-​associated macrophages40 and pro- of human melanoma following treatment with mono-
involved in antigen processing
mote the recruitment of CD8+ T cells41. A fine regulation clonal antibody (mAb) anti-​PD-1 was dependent on
and in the expansion,
maintenance and differentiation
of the genetic and epigenetic networks controlling TH1 immune reinvigoration of circulating exhausted CD8+
of T cell populations during responses is required for optimal T cell functioning42. T cells (implying the pre-​existence of an antitumour
an immune response. A series of immunosuppressive mechanisms also immune response) and pretreatment tumour burden8.
occur in the tumour microenvironment (TME), which This finding shows that even when in peripheral blood,
T cell receptor (TCR)
repertoire
hinder not only the natural host immune responses pre-​existing antitumour T cells can predict response.
The variety of the TCR diversity, but also the efficacy of cancer immunotherapies. Two Immune checkpoint blockade inhibitors (ICIs) may
as generated by the somatic types of immunosuppression mechanism operate in trigger changes in the T cell receptor (TCR) repertoire and
recombination of the germ line the TME — a tumour-​intrinsic and a local adaptive the expansion of specific clones of tumour-​reactive
V, D and J gene segments and
immuno­suppression43. The former may be induced by T cells63. In patients with melanoma who responded to
the deletion and insertion of
nucleotides at the V(D)J
genetic alterations of the tumour and involves the acti- anti-​PD-1 treatment, the comparison of the TCR reper-
junctions. Such variety is vation of various oncogenic pathways, including the toire before and after treatment revealed the emergence
required to recognize a wide WNT–β-catenin44,45, mitogen-​activated protein kinase of TCR Vß subfamilies — which specifically recognize the
spectrum of antigens. (MAPK)46,47, Janus kinase (JAK)–signal transducer melanoma antigen recognized by T cells 1 (MART1;
TCR Vß subfamilies
and activator of transcription 3 (STAT3)48 and nuclear also known as Melan-​A) — that were undetectable
Human TCR ß locus is on factor-κB (NF-​κB)49 signalling pathways. The engagement before treatment64. This emergence could be due to an
chromosome 7, comprising of these pathways results in the expression of cytokines insufficient sensitivity to detect lowly expressed clones
nine multimember V and chemokines that ultimately mediate the exclusion of or to a de novo second wave of immune activation,
subfamilies plus additional
T cells from the TME43, or, alternatively, the repression of with subsequent emergence of mutant neo-​epitope-tar-
elements on chromosome 9.
The presence of multiple
factors that facilitate T cell recruitment44. Depending on geting T cells65. Preclinical studies demonstrated that
subfamilies is due to the specifics of the case, the tumour-​intrinsic immuno­ PD-1 blockade cannot unleash antitumour T  cell
evolutionary duplication events. suppression can result in a cold or an altered (either responses in the absence of fully primed and committed

www.nature.com/nrd

Box 2 | challenges of immune monitoring
A characterization of hot, altered and cold tumours should be ideally carried out on
resected tumours (primary or metastatic). However, often only biopsy specimens are
available, thereby limiting the accuracy of the measurement. Although biopsy samples
have been extremely valuable in providing information on the disease, they have
several limitations, including being invasive, not always feasible and potentially not
representative of the whole tumoural landscape. Major intermetastasis genomic and
immune heterogeneity exists so that each metastasis could be considered as a separate
tumour50. Furthermore, immune parameters change in the course of the disease24.
Enormous effort has been put into the development of less invasive and equally
informative strategies, such as liquid biopsy270. However, blood evaluation is a poor
surrogate of what happens in the tumour microenvironment (TME), where ultimately
the main informative and prognostic features lie. A viable alternative to biopsy has noma (RCC), inhibition of PD-1 alone could not rescue
yet to be determined. Less invasive diagnostic procedures, such as immuno-​
positron-emission tomography (PET) imaging detecting intratumoural CD8 T cells271,
are quite promising. Apart from immunohistochemistry-​based assays, current methods
to characterize the immune landscape of the TME rely mostly on bulk gene expression
profiling272–274. Bindea et al. first demonstrated the possibility to infer immune cell
infiltration from gene expression profiles from a complex mixture of cells, revealing the
immune landscape in human tumours24. Subsequently, multiple additional tools, such as
CIBERSORT (which infers the relative fractions of immune subsets in the total leukocyte
population)273,275, xCell (which predicts the abundance of immune cells in the overall
TME)276, TIMER (which generates enrichment scores on the basis of proportions among
64 immune and stromal cell types)277 and integrated immunogenomics methods (using
a CIBERSORT-​based approach, which, of note, identified six immune subtypes of
cancer)278 were developed to estimate the abundance of intratumoural immune
infiltrates by using deconvolution of bulk gene expression data. Evident limitations of
these techniques are sample variability, inconsistency in the RNA extraction step, the
impossibility of univocally assigning transcripts to specific cell types and differences
between the immune phenotypes from blood and TMEs belonging to distinct cancer
types. Novel techniques based on single-​cell approaches279 or in situ barcode
sequencing280 are costly and far from large-​scale diagnostic use.
antigen-​specific T cells66, and the presence of function-
ally impaired PD-1high tumour antigen-​specific CD8+
T cells infiltrating the tumour has been described58.
Furthermore, TCR β chain (TCRβ) deep sequencing
revealed that clonally expanded tumour-​reactive CD8+
TILs express PD-1 (ref.59). Altogether, this evidence
strongly supports the existence of an in situ and/or
peripheral antitumour immunity that confers clinical
efficacy to subsequent checkpoint blockade.
There seems to be a general consensus on the cen-
tral part played by effector T cells in the antitumour
responses67. Upon recognition of antigenic peptides
presented on the surface of cancer cells by MHC I–
β2-microglobulin (β2m) complexes, CD8+ T cells kill
target cells mainly by releasing cytotoxic factors such
as PRF1, GNLY and GZM68. TBX21, STAT1, STAT4
and interferon regulatory factor 1 (IRF1) are among the
main signalling molecules and transcription factors that
regulate the production of these mediators that result
ultimately in tumour rejection68,69. High T cell infil-
trates in primary CRCs are associated with decreased
metastatic invasion and increased overall survival70,71,
and T cells are crucial for the clinical benefit of current
immunotherapies. For example, the presence of CD8+
T cells at the tumour invasive margin is a prerequisite for
the therapeutic success of PD-1 blockade in metastatic
melanoma6, and proliferation of said CD8+ T cells in
responding patients directly correlates with reduction
in tumour size6. Furthermore, a high mutational burden in
the TME correlated with the response to anti-​PD-1 and
Nature Reviews | Drug Discovery
Reviews
anti-​PD-L1 in melanoma, lung, MSI-​positive CRC7,72
and several other cancer types32. By breaking tolerance,
ICIs allow unleashing a pre-​existing immune response to
reject the tumour. However, they fail to reject the tumour
in the absence of such a pre-​existing response (for
example, in excluded or cold tumours)6,73,74. Of note,
the ‘quality’ of this pre-​existing immunity also affects the
response to ICI. For instance, the upregulation of alter-
native immune checkpoints (such as hepatitis A virus
cellular receptor 2, HAVcr-2, also known as TIM3) fol-
lowing anti-​PD-1 treatment confers adaptive resistance
to therapy75,76. Indeed, in patients with renal cell carci-
the functionality of CD8+ T cells that co-​expressed PD-1
and TIM3 (ref.77).
As single agents, ICIs have response rates in the range
10–35%78. Indeed, at the time of diagnosis, most stage IV
solid cancers are poorly infiltrated in primary tumours,
if not non-​infiltrated by T cells, possibly explaining
the limited response to ICIs73. When used as adjuvant
therapy for high-​risk stage III melanoma, the anti-PD-
1 mAb pembrolizumab resulted in significantly longer
recurrence-​f ree survival79. In an attempt to achieve
increased clinical benefit of ICIs, an impressive amount
of studies and clinical trials testing combinations of vari­
ous immunotherapy agents, as well as combinations of
immunotherapy agents with standard-​of-care treatments,
are currently under evaluation67 (Fig. 2). These have dif-
ferent likelihood of response based on pre-​existing hot,
altered or cold immune tumours (Fig. 1b).
Treating hot tumours with immunotherapy
T cell-​targeting immunotherapies. By displaying a
high degree of T cell infiltration, hot tumours represent
a fertile ground for effective ICI-​based monotherapy or
combination therapy (Box 3; Fig. 3). Exhausted or dys-
functional TILs express a number of inhibitory recep-
tors, most notably cytotoxic T lymphocyte-​associated
antigen 4 (CTLA4) and PD-1. CTLA4 inhibits T cells’
early activation and differentiation (typically in the
lymph nodes) whereas PD-1 modulates their effector
functions (mostly within tumours), which can lead to
T cell exhaustion78,80. Strategies used to target CTLA4
and PD-1 and/or PD-​L1 have now been approved by
the US Food and Drug Administration (FDA) for the
treatment of multiple cancers (Box 3). Hot (infiltrated)
TME, TH1 immune signature and PD-​L1 expression are
features associated with increased response to anti-​PD-1
or anti-​PD-L1 monotherapy6,81,82. The non-​redundant
nature of CTLA4 and PD-1 makes them good targets
for dual checkpoint blockade; indeed, anti-​CTLA4–
PD-1 dual therapy has been successful in the treatment
of advanced-​stage melanoma83, RCC84 and non-​small-
cell lung cancer (NSCLC)85, resulting in regulatory
approval. It can be postulated that such combinations
would be effective only in the context of hot or immuno­
suppressed tumours as they rely on a certain degree of
T cell infiltration.
Another promising target to be considered in associ-
ation with anti-​PD-1 and PD-​L1 strategies is lymphocyte
activation gene 3 (LAG3), a co-​inhibitory receptor on
T cells that, among other functions, enhances activity of
Reviews

a Number of trials assessing combinations with

anti-PD-1 and/or PD-L1 therapies b Number of IO agents
Multi-combo Preclinical
Radiotherapy plus
chemotherapy Phase I

Radiotherapy Phase I/II

Chemotherapy Phase II

Targeted Phase III

IO Approved

0 200 400 10 100 1,000

Number of trials Number of agents

c Number of IO drugs d Major treatment combination approaches

T cell-targeted T cell-targeted
immunomodulator immunomodulator
Immune Immune
microenvironment microenvironment
Cancer vaccine Cancer vaccine

Cell therapy Cell therapy

Oncolytic virus Oncolytic virus

CD3-targeted CD3-targeted
bispecific antibody bispecific antibody

0 200 400 600 800 1 2 3 4 5 6 7 8 9

Number of drugs Approaches

IO drugs to treat hot ( ), altered ( )

and cold ( ) immune tumours

Fig. 2 | overview of more than 2,000 immuno-​oncology agents currently tested or in use. a | Numbers of ongoing trials
assessing therapies in combination with anti-​programmed cell death protein 1 (PD-1) and/or PD-1 ligand (PD-​L1). With
‘multi-​combo’, we designate triple combinations, specifically , PD-1 and/or PD-​L1 combined with radiochemotherapy ,
chemotherapy and targeted therapy , chemotherapy and immunotherapy , targeted therapy and immunotherapy or
radiotherapy and immunotherapy (source: https://clinicaltrials.gov/). b | Number of immune-​oncology (IO) agents currently
in development, from preclinical phase to regulatory approval. c | Number of IO drugs in clinical trials sorted by mechanism
of action and identifying six categories of IO agent. The ‘immune microenvironment agents’ category includes any
immune modulator that does not belong to any other indicated category (for example, innate immune cell modulators).
d | IO agents efficient in hot immune tumours (red circle), IO agent combinations likely efficient in altered immune
tumours (yellow circle) and IO agent combinations being tested in cold immune tumours (blue circle). Nine (1–9) single or
combinatorial approaches are illustrated to possibly treat hot, altered and cold tumours. Combo, combinational therapy.

regulatory T (Treg) cells and regulates T cell proliferation,
differentiation and effector function78. Whereas LAG3
blockade yields only modest efficacy as a monotherapy,
its combination with anti-​PD-1 has synergistic potential
in preclinical models78. Other likely targets to combine
with existing ICIs include additional co-​inhibitory recep-
tors such as TIM3 (a marker for exhausted T cells86);
T cell immunoglobulin and ITIM domain (TIGIT, which
counterbalances the co-​stimulatory function of CD226,
that is rapidly induced following T cell activation87);
B and T lymphocyte attenuator (BTLA; also known as
CD272), which is expressed by T cells and synergizes
with herpesvirus entry mediator (HVEM; also known
as TNFRSF14), expressed on antigen-​presenting cells
(APCs)88; V-​domain Ig suppressor of T cell activation
(VISTA, which mediates a compensatory inhibitory
pathway following anti-​CTLA4 therapy in prostate
cancer89); and sialic acid-​binding Ig-​like lectin 9 (SIGLEC9),
which is upregulated in TILs and possibly determines a
subclass of tumour-​specific CD8+ TILs90.
Alternatively, ICIs could be combined with co-​
stimulatory checkpoint molecules, such as OX40 anti-
gen (also known as TNFRSF4 (or CD134)), TNFRSF7
(also known as CD27), CD28, TNFRSF9 (also known as
4-1BB ligand receptor or CD137) and glucocorticoid-​
induced TNFR-​related protein (GITR), all of which
enhance T cell expansion and effector functions while
controlling Treg cell suppressive functions78,91. The notion
that CD28 conveys the second signal required to complete
T cell activation upon TCR engagement has been known
for over three decades92. Recently, CD28, and not the TCR,
has been shown to be the target of PD-1 signalling93 and
thus to be required for efficient PD-1 therapies94. Apart
from enhancing effector T function, CD28 blocks the sup-
pressive function of Treg cells92, which suggests its potential
as an anticancer therapy95. The first-​in-human clinical trial
testing a CD28 super-​agonist was associated with a life-​
threatening cytokine release syndrome in all six subjects
receiving the drug96, illustrating the need to fine-​tune the
dose and scheduling of these powerful immunotherapies.

www.nature.com/nrd
 Reviews

Box 3 | Predicting the response to anticancer therapy responding patients showed an upregulation of PD-​L1,
which suggested that the ‘right’ microbiota could help
Monoclonal antibodies targeting programmed cell death protein 1 (PD-1) and in the development of a hot TME101. Another study on
PD-1 ligand (PD-​L1) have been approved by the US Food and Drug Administration metastatic melanoma demonstrated the significant asso-
(FDA) for the treatment of multiple cancers (melanoma; non-​small-cell lung cancer; ciation between commensal microbial composition and
kidney, bladder, head and neck, hepatocellular, gastric and cervical cancer; Merkel cell
clinical response to anti-​PD-1-based immunotherapy102.
carcinoma; Hodgkin lymphoma; and tumours with microsatellite instability (MSI)).
A considerable effort is being put into establishing valid predictors of response to cancer It is unlikely that the modulation of the microbiota
immunotherapies. Multiple tumour-​related or immune-​related predictive biomarkers alone would work in an altered or cold tumour; none-
have been proposed, including the expression of immunosuppressive molecules theless, the presented evidence suggests that fairly simple
(such as PD-​L1) by tumour cells281,282; the molecular profiling of the tumour lifestyle changes and/or oral supplementation of good
microenvironment, which encompasses the expression of inflammatory genes10,283; bacteria could shape a favourable stage for subsequent
the assessment of the mutational landscape and neoantigen load5; mismatch-​repair immune-​based therapies, which could be effective in the
deficiency and MSI20; tumour aneuploidy284; immune infiltration; and Immunoscore19. context of hot tumours or possibly in combination with
A striking observation is that all of these are ultimately immune-​related biomarkers, other agents in altered or cold tumours.
including classical tumour-​related features such as somatic copy number alterations
(SCNAs). High levels of SCNA are associated with a poor response to anti-​cytotoxic
Treatment of immune-​altered tumours
T lymphocyte-​associated antigen 4 (CTLA4) therapy and with reduced expression of
cytotoxic immune infiltration in patients with melanoma284. The immune contexture T cell trafficking modulators. In excluded tumours, an
parameters act as prognostic (associated with survival), predictive (associated with accumulation of CD8+ T cells occurs at the tumour bor-
response to treatment) and mechanistic (increased after treatment in responding ders. This phenomenon indicates the ability of the host
patients) biomarkers. to mount a T cell-​mediated immune response and the
Attempts to rationalize and set instrumental guidelines for personalized therapies have physical inability of the T cells to reach the tumour bed
been made. Blank et al. introduced a dynamic model (the ‘cancer immunogram’), which (Fig. 1). Many explanations for this phenotype can be pro-
required the assessment of a combination of biomarkers as a tool to guide treatment posed. One reason for T cell exclusion could be the lack
options for individual patients62. The model relies on published data from patients who of T cell-​recruiting signals, such as chemokines directing
responded to anti-​PD-1 and/or anti-​PD-L1 treatment62. An initial framework of seven T cell trafficking, including CXCL9, CXCL10, CXCL11,
parameters has been established: tumour foreignness; general immune status; immune
CXCL13, CX3C-​c hemokine ligand 1 (CX3CL1),
cell infiltration; absence of checkpoints; absence of soluble inhibitors; absence of
inhibitory tumour metabolism; and tumour sensitivity to immune effectors62. The CC-​chemokine ligand 2 (CCL2) and CCL5 (refs57,103). This
evaluation of these factors can be achieved by a combination of tumour genomics, shortage of chemokines may result from the modulation
Immunoscore assay, immunohistochemistry, standard blood assays and immune gene of the oncogenic, genetic and epigenetic pathways that
signature, both pre-​therapy and post-​therapy, and could be helpful in designing control their expression104. Histone modification and
possibly the most efficient therapeutic intervention for individual patients62. DNA methylation can repress the expression of TH1-
derived CXCL9 and CXCL10 in ovarian105 and colon
Another co-​stimulatory immune checkpoint molecule, cancer106. Therapeutic epigenetic modulation promoted
inducible T cell co-​stimulator (ICOS), could also be a tumour infiltration of effector T cells, slowed tumour pro-
candidate owing to its expression on activated T cells. gression and improved the efficacy of PD-​L1 blockade in
However, despite showing a promising profile in tumour preclinical models106,107. In metastatic melanoma, constitu-
mouse models97, its concomitant expression on Treg cells98 tive activation of the β-​catenin pathway resulted in defec-
might limit its clinical impact. On the basis of preclini- tive recruitment of CD103+ dendritic cells (DCs) into the
cal evidence, the requirement for combinatorial immune TME and subsequent absence of CD103+ DC-​derived
checkpoint blockade could be bypassed by inhibiting the CXCL9 and CXCL10 in hot tumours44. Of note, the cited
chronic interferon response, shown to mediate resist- studies rely on the more simplistic distinction of T cell-​
ance to ICIs39. Altogether, albeit promising, the use of inflamed (hot) versus non-​T cell-​inflamed (cold) tumours.
co-​stimulatory molecules could be limited clinically by Therefore, it can be only postulated that between these two
systemic toxicity due to important off-​tumour effects. types, there could be cases of excluded tumours; however,
it is tempting to speculate that these represent the cases in
Microbiome modulation. Antibiotics can inhibit the which the association of T cell tracking modulators with
clinical benefit of ICIs in patients with advanced-​stage immunotherapy yields clinical benefit.
cancer, and a correlation between clinical responses to Activation of tumour-​oncogenic pathways correlat-
ICIs and the relative abundance of the Gram-​negative ing with T cell exclusion has been described in muscle-​
commensal bacterium Akkermansia muciniphila has invasive urothelial bladder cancer (MIUBC)108, usually
been demonstrated99. Therefore, selective modulation characterized by poor outcome. Non-​surprisingly, genes
of gut microbiome composition might overcome resist- encoding PD-​L1, IDO, FOXP3 (forkhead box P3; the
ance to ICIs99,100. Abundance of ‘good’ bacteria, including master regulator of Treg cells), TIM3 and LAG3 were
those belonging to the Faecalibacterium genus101, corre- upregulated in T  cell-​inflamed MIUBCs, whereas
lated with a higher number of effector CD4+ and CD8+ an activation of β-​catenin, peroxisome proliferator-​
T cells in peripheral blood and with response to anti-​ activated receptor-​γ (PPARγ) and fibroblast growth
PD-1 mAbs in patients with melanoma101. Conversely, factor receptor 3 (FGFR3) pathways was found in the
non-​responders had a higher abundance of members of most common non-​T cell-​inflamed MIUBCs108.
the Bacteroidales order, which correlated with higher A ‘sufficient’ T cell infiltration in mice tumour sites,
frequencies of Treg cells and myeloid-​derived suppres- rather than a differential PD-​L1 expression, is critical
sor cells (MDSCs) in the systemic circulation101. Mice for the response to anti-​PD-L1 therapy109. On the basis
that received faecal microbiota transplantation from of this assumption, it can be envisaged that strategies

Nature Reviews | Drug Discovery

Reviews

Activators TGFβi* Anti-PD-1#

of NK cells$ Anti-ECM* Anti-PD-L1#
Anti-CTLA4#
Radiotherapy#
CD3, CD8 Anti-TIM3*
Oncolytic peptides* TFH, TH1
ECM Memory
Collagen exhausted
Oncolytic virus# PD-1 Anti-LAG3*
EMT/MET T cells PD-L1
TKI #

Microbiome CTLA4
Barrier Anti-
modulators$ molecules TIM3 CTLA4/PD-1#
ICD
Vaccine* Calreticulin Mesenchymal Tolerance LAG3
Neo-epitope barrier Combo
Stress T cell Anti-PD-1/
vaccine* checkpoint
No/low Immun PD-L1*
Anti-CTLA4 e Anti-LAG3*
combo* adjuvancity Type D
Mutations ens Anti-TIM3*
ity C Anti-BTLA$
Instability MSI Absent Optimal Lo
No/low ca Anti-VISTA$

on
inducer$ CIN t

te
immuno- Anti-SIGLEC9$

xtu
io
genicity CTLA4

n
Low Immunoscore High Immunoscore + other ICP*

re
Fu
CAR T cell# PD-1/PD-L1

nct
HLA TIM3, Anti-OX40*
TLRa*

ion
β2M No/low Hot Combination LAG3 Anti-ICOS*
Proteasome presentation Cold checkpoints TIGIT, BTLA Anti-CD137*
DDR agent* Anti-GITR*
Other ICP
Immunogenicity Anti-CD40*
Anti- OX40 CD40 Anti-CD28*
ADORA2A* Hypoxia Adjuvanticity Co-
HIF1α ICOS CD28 Anti-CD27*
Anti-CD73* Adenosine stimulation CD137 CD27
Anti-CD39$ GITR Cyto- IL-2 #
, IL-7*,
Angiogenesis
HIF1αi * immune kines IL-15*, IL-21*
modulation Adhesion MAdCAM1 GM-CSF*
VEGFα IL-12*
Anti-VEGF* Excluded Immunosuppressed ICAM1
VCAM1 IFNα#
Anti- Epigenetic
reprograming Inhibitory CD103 ICAM1
angiogenesis*
mediators VCAM1$
Anti-HVEM$
HDAC IDO
Oncogenic MDSCs
HDACi* activation Treg cells TDO
HMA* WNT–
Immuno- NOS1 Combo IDOi*
BETi* β-Cat T-cell
trafficking Apoptotic suppression Arginase Combo TDOi*
MEK
PPARγ recycling Presentation CSF1R
TKI#
FGFR3 and priming FOXP3
WNTi# Cyclophosphamide
PI3Ki# CXCL9/10/11 Chemotherapy#
CXCL1/13 IL-6 TGFβ
MEKi# Survivin BATF3 PI3Kγi*
METi# CCL2/5 IL-10 Tasquinimod*
IAP XCR1/XCL1
mTOR STING Anti-CSF1R*
Chemokines* MCL1 IFNα MDSC
Anti-LIGHT* depletion*
Anti-CCR5* TGFβi*
Anti-IL-6#
PI3Ki#
Survivini# STINGα$
mTORi#
MCL1i#

Tumour Pathways #
FDA approved
Intermediate Targets *Ongoing trials
Immunoscore Drugs $
Preclinical evidence
Immunity

that facilitate the recruitment of T cells to excluded
tumours could overcome tumour resistance to check-
point blockade109. TNFSF14 (also known as LIGHT,
or HVEM ligand) is an activator of lymphotoxin
β-​receptor signalling, which triggers the production of
T cell-​targeting chemokines, thereby creating a T cell-​
inflamed microenvironment109. Accordingly, the use
of an antibody-​guided LIGHT modulated the recruit-
ment of T cells to the tumour site and overcame tumour
resistance to ICIs109.
In summary, the epigenetic modulation of T H1-
derived chemokines, as well as a blockade of β-​catenin
signalling, could turn excluded tumours into hot
tumours, thus increasing the likelihood of success of
concomitant immunotherapy.
Physical barrier breakers. Another explanation for
the inability of T cells to penetrate tumour sites could
be the presence of physical barriers. The development
of abnormal structural features is a hallmark of cancer

www.nature.com/nrd

◀ Fig. 3 | The tumour–immune classification cycle as a tool to direct anticancer
therapy. Tumours can be classified into four main subtypes (hot, altered-​excluded,
altered-​immunosuppressed and cold) according to their associated T cell (CD3+ and
CD8+) presence and distribution. Hot tumours are defined by the simultaneous presence
of immune contexture parameters: the cell type (CD3+, CD8+, follicular helper T (TFH),
T helper 1 (TH1), memory and exhausted T cells); the location (invasive margin, tumour
core and tertiary lymphoid structures); the density (immune density and quantity); and
the functional immune orientation (chemokines, cytokines, cytotoxic factors, adhesion,
attraction and TH1)10. The main components, pathways and features (in green trapezoids)
of the immunogram have been identified and may represent potential targets (in blue).
One or more of these features can be present at once in a specific tumour.
Immunogenicity and adjuvanticity are tumour-​intrinsic core characteristics that
contribute to the shaping of the tumour-​associated T cell landscape; we postulate that
they are strikingly low or absent in cold tumours, although they may contribute to a
certain extent to other subtypes. By representing a tool for the classification of a cancer
into one of the four categories together with the identification of a specific defect or
deregulated pathway , the spin chart could prove instrumental for building the most
successful therapeutic scheme, and, conversely , narrow down the list of possible
therapies by excluding potentially inefficient ones. ADORA2A , A2A adenosine receptor ;
β2m, β2-microglobulin; BET, bromodomain and extra-​terminal motif proteins; BTL A ,
B and T lymphocyte attenuator ; CAR T cell, chimeric antigen receptor T cell; CCR ,
CC-​chemokine receptor ; CIN, chromosomal instability , CSF1R , colony-​stimulating factor 1
receptor ; CTL A4, cytotoxic T lymphocyte-​associated antigen; CXCL , CXC-​chemokine
ligand; DDR , DNA damage response; ECM, extracellular matrix; EMT, epithelial–
mesenchymal transition; FDA , US Food and Drug Administration; FGFR3, fibroblast
growth factor receptor 3; FOXP3, forkhead box P3; GITR , glucocorticoid-​induced TNFR-​
related protein; GM-​CSF, granulocyte–macrophage colony-​stimulating factor ; HDAC,
histone deacetylase; HIF1α, hypoxia-​inducible factor 1-α; HL A , human leukocyte
antigen; HMA , hypomethylating agent; IAP, inhibitors of apoptosis family (also known
as XIAP); ICAM1, intercellular adhesion molecule 1; ICD, immunogenic cell death;
ICOS, inducible T cell co-​stimulator ; ICP, immune checkpoint; IDO, indoleamine
2,3-dioxygenase; IFN, interferon; IL , interleukin; L AG3, lymphocyte activation gene 3;
LIGHT, tumour necrosis factor superfamily member 14; MAdCAM1, mucosal addressin
cell adhesion molecule 1; MCL1, induced myeloid leukaemia cell differentiation protein
Mcl1; MDSCs, myeloid-​derived suppressor cells; MEK , mitogen-​activated protein kinase
kinase; MET, mesenchymal–epithelial transition; MSI, microsatellite instability ; NK ,
natural killer ; NOS1, nitric oxide synthase 1; PD-1, programmed cell death protein 1;
PD-​L1, PD-1 ligand; PI3Kγ, phosphoinositide 3-kinase-​γ; PPARγ, peroxisome proliferator-​
activated receptor-​γ; SIGLEC9, sialic acid-​binding Ig-​like lectin 9; STING, stimulator of
interferon genes; TDO, tryptophan 2,3-dioxygenase; TGFβ, transforming growth factor-​β;
TIGIT, T cell immunoglobulin and ITIM domain; TIM3, T cell immunoglobulin and mucin
domain-​containing 3; TKI, tyrosine kinase inhibitor ; TLR , Toll-​like receptor ; Treg cells,
regulatory T cells; VCAM1, vascular cell adhesion molecule 1; VEGF, vascular endothelial
growth factor ; VISTA , V-​domain Ig suppressor of T cell activation; XCL1, lymphotactin;
XCR1, chemokine XC receptor 1. Lower case i following any acronym or abbreviation
indicates inhibitor ; lower case a following any acronym or abbreviation indicates agonist.
progression. Many extracellular matrix proteins are sig-
nificantly deregulated during the progression of cancer,
causing both biochemical and biomechanical changes
possibly inhibiting immunity and promoting the met-
astatic cascade110. Changes in the tumour-​associated
vasculature (both blood and lymphatic) have been
extensively described, as well as the resulting hypoxic
milieu111,112. Tumour vasculature acts as an impor-
tant barrier to T cells via the deregulation of adhesion
molecules (such as intercellular adhesion molecule 1
(ICAM1), vascular cell adhesion protein 1 (VCAM1)
and mucosal addressin cell adhesion molecule 1
(MAdCAM1)) required for T cell extravasation113,114.
Hypoxia favours the establishment of an immunosup-
pressed TME, and ultimately cancer progression and
treatment resistance, mostly acting through the hypoxia-​
inducible factor (HIF) family of transcription factors115.
Among these, HIF1α not only can drive the expression
of PD-​L1 but also can increase adenosine generation
Nature Reviews | Drug Discovery
Reviews
and signalling116. The enzymatic activity of two ecto-​
nucleotidases, CD39 (also known as NTPDase I) and
CD73 (also known as 5ʹ-nucleotidase), is responsible
for the conversion of extracellular ATP to adenosine117.
Adenosine accumulation in the TME exerts a plethora of
pro-​cancer effects117,118. CD73 blockade in several in vivo
models significantly reduced tumour growth and meta-
static spread119, and combination of CD73 blockade with
ICIs has been proposed as an exploitable therapeutic
strategy120. In turn, CD39 blockade prolonged survival
in a lethal metastatic patient-​derived sarcoma model121.
However, deletion of CD39 resulted in autochthonous
liver cancer in mice122, suggesting caution when using
CD39 blocking agents. Nonetheless, agents that tar-
get CD39 are currently in the preclinical stage, whereas
CD73 inhibitors are already being assessed in clinical
trials123. Blocking the A2A adenosine receptor ADORA2A
can restore immune competence and T cell proliferation
in chronic lymphocytic leukaemia116. Therefore, HIF
(mostly HIF1α and HIF2α) as well as CD73, CD39 and
adenosine receptor inhibitors and/or mAbs have been
developed as possible anticancer therapeutics, although
thus far, no agent has reached regulatory approval123–125.
Hypoxia also exerts systemic effects via the secre-
tion of growth factors and cytokines altering immune
cell proliferation, differentiation and function111. One
of these factors is the pro-​angiogenic cytokine VEGF,
which has been targeted by various pharmacological
and immune-​based antitumour strategies in the past
decade112,126. VEGF-​dependent effects extend beyond its
angiogenic capabilities. For example, high expression of
VEGFA inhibits the maturation of DCs, modulates TCR
signalling and counteracts the beneficial effects of TH1
cells and cytotoxic T cells by suppressing the expression
of IRF1 and GNLY, respectively127. Nonetheless, anti-​
VEGF and other angiogenesis inhibitors have not been
successful as single agents, mostly owing to the devel-
opment of a compensatory mechanism128. Furthermore,
these inhibitors resulted in some cases in a somewhat
counterintuitive increase in metastatic spread129,130.
Despite these setbacks, evidence exists in support of the
complementary therapeutic value of the normalization
of the vascular abnormalities112, or vessel normaliza-
tion. Vessel normalization is characterized by increased
pericyte coverage, improved tumour vessel perfusion,
reduced vascular permeability and subsequent reduced
hypoxia131. Infiltration and activity of TH1 lymphocytes
correlate with vessel normalization. In addition, TH1
activation by ICIs increased vessel normalization in
various mouse models of breast cancer132, suggesting
that a synergistic effect can be achieved by combining
anti-​angiogenic therapies with ICIs.
Soluble factor inhibitors. In the case of immuno­
suppressed tumours, tumour sites display a modest, insuf­
ficient degree of immune infiltration (Fig. 1), suggesting
that the presence of an immunosuppressive environ-
ment, rather than that of physical barriers, limits further
T cell recruitment17 and expansion18. IL-10 and TGFβ
are among the best-​characterized tumour-​derived solu-
ble factors impairing the development of an antitumour
immune response133,134. As a matter of fact, IL-10 and
Reviews
TGFβ display both pro-​tumour and antitumour abilities.
This apparent contradiction can be explained by their
numerous, yet not fully characterized, functions as well
as by their different local versus systemic effects135,136.
Therefore, they would not seem to constitute optimal
molecular targets. Nevertheless, the recent evaluation of
the combined inhibition of TGFβ and PD-​L1 in multi-
ple tumour types showed clinical benefit137. One of the
characterized functions of IL-10 and TGFβ is their abil-
ity to disrupt DC differentiation, migration and antigen
presentation, some of the crucial mechanisms required
to mount an effective antitumour T cell response138,139. In
the presence of this type of immunosuppression, CD40-
activated B cells were suggested as alternative APCs
owing to their resistance to IL-10, TGFβ and VEGF138.
A new therapeutic strategy aiming at improving the
effector functions of T cells is based on the ionic repro-
gramming of tumour-​specific T cells140. Tissue necrosis,
common in solid tumours, releases potassium (usually
intracellular) into the extracellular milieu, suppressing
T cell effector functions141. This effect was mediated by
a subsequent increase in intracellular potassium levels,
inhibiting the TCR-​driven AKT–mTOR pathway in a
protein phosphatase 2-dependent manner140. By low-
ering intracellular potassium, the overexpression of the
potassium channel Kv1.3 restored T cell effector func-
tions, tumour clearance and survival in a melanoma
mouse model140. These results show the potential of this
alternative type of intervention as a further promising
anticancer strategy.
Cellular modulators of local adaptive immunity.
Two key cellular mediators of local adaptive immune
suppression in the TME are MDSCs and Treg cells67. As
both favour tumour progression, it is not surprising that
strategies aiming at reducing their number are currently
being explored. This seems to be somewhat feasible for
MDSCs and has indeed been achieved by blocking their
main suppressive pathways (IDO142, arginine, trypto-
phan and nitric-​oxide-related pathways143–145), by regu-
lating myelopoiesis or by preventing their trafficking67.
Box 4 | cancer stem cells: key mysterious players in cancer progression
Cancer stem cells (CSCs), which include mesenchymal stem cells, are a vast population
of cells involved in cancer initiation and progression285. CSCs are characterized by self-​
renewal and differentiation capacity, similarly to their stem cell counterparts.
Importantly, their undifferentiated state makes them resistant to immune recognition,
owing to the lack of expression of human leukocyte antigen (HLA) class I, as well as to
chemotherapy and radiotherapy285. The epithelial–mesenchymal transition (EMT) and
the mesenchymal–epithelial transition (MET) are processes of phenotypic transition
between epithelial and mesenchymal states286. It is widely accepted that EMT in cancer
is at least associated with, if not necessary for, the metastatic process287. The elucidation
of the exact involvement of EMT and MET and CSCs in tumorigenesis and tumour
aggressiveness is still an object of debate, hampered by contrasting data285.
Nevertheless, targeting the EMT has been proposed as a strategy to at least partly
overcome resistance to immune checkpoint blockade288. EMT might contribute to
immune escape via multiple routes, including the shaping of the tumour
microenvironment (through the production of soluble factors such as interleukin 10
(IL-10) and transforming growth factor-​β (TGFβ)) and decreased susceptibility to
immune effector cells288. CSCs express natural killer (NK) receptor ligands, implying
that the exploitation of NK cells could provide a valuable strategy to counteract cells is represented by cancer stem cells, which are
CSC-​mediated immunosuppression285.
Nonetheless, the recent failure of the phase III, rand-
omized, double-​blind study involving the combination
of an IDO inhibitor plus a PD-1 blocking agent to yield
greater clinical benefit than anti-​PD-1 alone in unre-
sectable or metastatic melanoma raises questions on the
efficacy of these strategies146.
The selective targeting of the γ-​isoform of phospho-
inositide 3-kinase (PI3Kγ), highly expressed in myeloid
cells, successfully reshaped the TME and promoted
cytotoxic-​T cell-​mediated tumour regression in preclini-
cal mouse models for several cancers147. The combination
of this pharmacological approach with PD-1 blockade is
currently under investigation in clinical trials147.
Blockade of colony-​stimulating factor 1 receptor
(CSF1R) could deplete the pro-​tumoural macrophage
population, in particular the M2 subtype, thus favouring
the induction of a cytotoxic antitumour T cell response
following PD-​L1 blockade148. Clinical trials are ongo-
ing to test the clinical activity of a combined treatment
associating antibodies against CSF1R andanti-​PD-L1 in
patients with metastatic cancers148. Another example of
modulation of tumour-​associated myeloid cells can be
found in the study by Nakhlé et al. on bladder tumours.
In mouse models, targeting S100A9 — a zinc and cal-
cium protein with a prominent role in the regulation
of inflammatory processes and immune response —
with tasquinimod, a regulator of MDSC accumulation,
resulted in a re-​education of tumour-​infiltrating mye-
loid cells from a pro-​tumoural M2 phenotype towards
an antitumoural M1 phenotype149. The parallel increase
in PD-​L1 expression could explain the lack of tumour
growth inhibition following treatment with tasquini-
mod as single therapy149 while simultaneously providing
a solution to overcome this obstacle. Indeed, the com-
bination of tasquinimod with an anti-​PD-L1 antibody
enhanced the antitumour immune response in preclini-
cal bladder tumour models149. Along the same lines, the
depletion of MDSCs induced a CD8+ T cell-​dependent
tumour rejection in mouse models of head and neck
squamous cell carcinoma (HNSCC) when combined
with anti-​CTLA4 therapy150. The same study revealed
the existence of an MDSC-​rich gene expression profile
with a T cell-​inflamed phenotype in > 60% of patients
in an HNSCC cancer cohort150. This observation reit-
erates the idea that a comprehensive characterization
of the TME and its key components could prove to be
tremendously useful in pinpointing the most promising
therapeutic strategy.
The targeting and reduction of tumour-​associated
Treg cells remain quite challenging to achieve. Impor­
tantly, Treg cells express high levels of PD-1, which acts
as a stimulatory receptor, rather than inhibitory, in these
cells67, which may therefore reduce the benefits of an
eventual anti–PD-1 antibody-​based therapy67. Clinical
approaches for Treg cell depletion have also not been
very successful, having the downside of also reducing
the tumour-​suppressive Treg cells, critical for preventing
autoimmunity. Crucially, Treg cells in the TME are not
dysfunctional, in contrast to other TILs67.
A third, less characterized population of pro-​cancer
discussed in Box 4.
www.nature.com/nrd

Tumour-​specific antigen
(TSA). Antigen not encoded in
the normal genome, expressed
exclusively by tumour cells.
Antigenicity
Presence of tumour-​associated
antigens (TAAs) capable of
engaging with T cell receptors
or antibodies (B cell receptors),
thereby driving adaptive
immunity.
Nature Reviews | Drug Discovery
Innate immune response modulators. The lack of ade-
quate innate immune response could constitute a lim-
iting factor restraining the development of an effective
adaptive, antitumour response, therefore originating an
immunosuppressed tumour or contributing to the estab-
lishment of a cold tumour. One of the first indications of
the activation of innate immune pathways in tumour set-
tings came with the correlation between the expression
of ISGs and T cell-​associated transcripts in metastases
from melanoma151. Indeed, a type I interferon signature
predicted favourable clinical outcome in breast carci-
noma following treatment with cancer vaccines152 and
classic chemotherapy153. Type I interferons in tumours
derive mostly from the activation of the stimulator of
interferon genes (STING) pathway by cytosolic tumour-​
derived DNA within conventional, basic leucine zipper
ATF-​like transcription factor 3 (BATF3)-expressing,
tumour-​a ssociated DCs. The pattern-​recognition
receptor cyclic GMP-​AMP synthase (cGAS) recognizes
cytosolic DNA, thereby stimulating the formation of the
STING ligand cyclic-​GMP-AMP154. These events culmi-
nate in the promotion of tumour-​specific antigen (TSA)
cross-​presentation and T cell cross-​priming by these
cells155–157. Regression of established tumours and sys-
temic antitumour responses were observed in preclinical
models following intratumoural injection of a STING
agonist158. Therefore, STING agonists are currently being
tested in clinical trials159.
The modulation of the innate immune landscape
within the TME represents a possible point of inter-
vention in the case of T cell-​infiltrated tumours that
do not respond to ICI (such as immunosuppressed
tumours). The local injection of a STING agonist in
the tumour site, followed by treatment with an anti-​
PD-L1 antibody, was able to control, or completely
reject, T cell-​inflamed tumours in a mouse model of
HNSCC149. Again, as the study refers to only hot and
cold tumours, we can only assume that this case may
fall into the category of an immunosuppressed tumour
rather than that of a hot tumour. STING activation trig-
gered the production of type I and II interferons and
the accumulation of DCs in tumour-​draining lymph
nodes, boosting the T cell-​mediated response; this
came with a concomitant increase in the expression of
PD-​L1 in the TME149. The subsequent addition of an
anti-​PD-L1 mAb successfully removed the immuno-
suppressive status, thereby allowing the establishment
of a local, as well as systemic, T cell-​mediated anti­
tumour activity149. Of note, the STING-agonist-mediated
induction of type I interferon and accumulation of DCs
in draining lymph nodes were not enough to mount an
adaptive immune response in a parallel model of non-​
T cell-​inflamed (cold) tumour149. Poor antigenicity or
an intrinsic insensitivity to T cell killing in this model
were proposed as possible explanations for these find-
ings149. The potential of STING as an immune adjuvant
that promotes the priming of tumour antigen-​specific
CD8 T cells has also been shown in murine tolerogenic
HER2+ breast tumours160. In this study, the injection of
a STING agonist in combination with PD-​L1 blockade
and OX40 activation resulted in tumour regression by
providing priming and overcoming antigen-​enforced
Reviews
immune tolerance 160. It should be noted that also
persistent type I interferons, just as IFNγ, can confer
resistance to immune checkpoint blockade39, thereby
suggesting possible setbacks of this therapeutic
strategy.
Apart from the DNA sensor cGAS, various other
pattern-​recognition receptors, such as Toll-​like recep-
tors (TLRs), RNA helicase RIG-​I-like receptors and
NOD-​like receptors have a role in endogenous stress
signal recognition occurring in tumours161,162. Therefore,
they are being investigated for their anticancer poten-
tial. TLRs are highly expressed by tumour-​infiltrating
immune cells, notably APCs, leading to their activation
upon ligand stimulation. The upregulation of MHC II,
CD80 and CD86 by TLR-​stimulated APCs and their
conversion from tolerogenic to immunogenic have
been demonstrated in murine and human TMEs163,164.
TLRs can also be expressed by tumour cells, in which they
can exert direct cytotoxic effects upon stimulation165.
Intratumoural injection of a TLR9 agonist reverted
resistance to PD-1 blockade, resulting in durable and
systemic tumour rejection in mouse models166. Similarly,
intratumoural injection of TLR7 and TLR9 agonists, in
combination with PD-1 blockade, suppressed primary
tumour growth and prevented metastatic spread in
HNSCC models167. Finally, intratumoural injection of
a TLR9 ligand, combined with administration of an
anti-​OX40 antibody, successfully mediated regression
even of a variety of histological tumour types, includ-
ing that of spontaneous breast cancers168. The efficacy of
TLR agonists has been investigated in multiple clinical
trials169–173. When used as monotherapies, these agents
yielded limited success174; however, combination strate-
gies (for example, with anticancer vaccines) seem to be
quite promising175–177.
Modulation of APC activation can be achieved by
using agonist anti-​CD40 mAbs. CD40 is a TNFRSF
member mostly expressed on APCs. By interacting with
its ligand on activated TH cells, it triggers APC activa-
tion and subsequent induction of adaptive immunity178.
Agonistic anti-​CD40 antibodies performed remark-
ably well in preclinical models of B cell lymphomas179
and bladder tumours180. There are multiple ongoing
clinical trials involving CD40 agonists, and potential
combinations with ICIs are being evaluated181.
The chemokine XC receptor 1 (XCR1) and its ligand
lymphotactin (XCL1) regulate migration and function
of CD103+CD11b− DCs182–184. Intratumoural natural
killer (NK) cells produce CCL5 and XCL1, thereby pro-
moting DC recruitment in multiple human cancers and
affecting patient survival184. Future anticancer strate-
gies exploiting these newly discovered axes could prove
successful.
The treatment of immune cold tumours
Cold tumours, characterized by low Immunoscore,
are the most challenging to eradicate and are invar-
iably associated with poor prognosis. A proposed
approach to overcome the lack of a pre-​existing immune
response — and ultimately to turn cold tumours into
hot tumours — is to combine a priming therapy that
enhances T cell responses (such as vaccines, adoptive
Reviews
Abscopal effect
Phenomenon characterized by
the regression of metastases
outside the field of radiation
after irradiation of one tumour
site. Although rarely detected,
it is well documented in
patients with more
immunogenic tumours.
Adjuvanticity
Presence of damage-​
associated molecular patterns
(DAMPs) and stress signals
driving the innate immunity.
Genotoxic chemotherapies
Chemical agents that cause
DNA damage, such as single-​
strand and double-​strand
breaks, loss of excision repair,
crosslinking, alkali-​labile sites,
point mutations and structural
and numerical chromosomal
aberrations.
T cell transfer (ACT) or strategies that turn the tumour
into a vaccine) with the removal of co-​inhibitory signals
(through approaches such as ICIs or MDSC depletion)
and/or the supply of co-​stimulatory signals (such as
anti-​OX40 or anti-​GITR)67. The concern with this type
of approach would be the concurrent increase in unde-
sired side effects67, which is the case of most combina-
torial therapies, surely requiring careful evaluation. In
principle, it would make sense that a priming therapy
would be beneficial in the case of non-​inflamed, cold
tumours, whereas inflamed, hot tumours would benefit
more from immune interventions that counteract the
tumour-​induced T cell dysfunctions22,67. However, this
black and white distinction of hot versus cold tumour is
an oversimplification of an incredibly complex scenario.
The introduction of the altered categories (excluded and
immunosuppressed) may more suitably represent inter-
mediate phenotypes. Hence, it is possible that some of
the proposed approaches could in fact be effective only
in the case of altered tumours. It is likely that the paral-
lel occurrence of multiple pro-​tumour mechanisms ulti-
mately leads to the establishment of a cold tumour and
combinatorial approaches are likely needed to achieve
clinical benefit.
Radiotherapy. A promising priming therapy to be
associated with subsequent immunotherapy is ioniz-
ing radiotherapy67,73. The currently achievable precise
delivery of radiotherapy and the resulting induction
of immunogenic cell death (ICD) pathways can poten-
tially convert the tumour into an in situ vaccine67. The
consequences of this approach not only involve the irra-
diated tumour site itself but can possibly contribute
to the achievement of systemic tumour control through
the so-​called ‘abscopal effect’67. The DNA released fol-
lowing the radiation-​induced cell damage might trig-
ger a STING-​mediated type I interferon production,
possibly by tumour-​infiltrating CD8α+CD103+ DCs in
mice (or their human counterpart CD141+ DCs185), a
DC subset specialized in antigen cross-​presentation186;
this will then fuel a T cell-​mediated antitumour immune
response186,187. This process relies on a certain degree of
infiltration of CD103+ DCs, which could be a limiting
factor in cold tumours but may be effective in excluded
tumours. A recently developed mouse melanoma model
that lacks intratumoural CD8α+CD103+ DCs44 could
shed light onto the ability of radiotherapy to prime T cell
responses in non-​infiltrated tumours. The combination
of radiotherapy with further anti-​immunosuppressive
strategies could potentially increase the frequency of the
abscopal effect. Indeed, combining radiotherapy with
anti-​CTLA4 mAbs in melanoma and NSCLC (which
normally does not respond to anti-​CTLA4) significantly
improved the therapeutic outcome67. Additionally, maxi-
mal anti-​CTLA4 therapy-​driven abscopal responses in a
mouse melanoma model was dependent on STING sig-
nalling188, further supporting the key role of STING and
type I interferons in this context. The same study
revealed a possible reason for the observed time frame
(days rather than minutes) in which the radiation-​
induced inflammatory responses occur. In a classi-
cal model, the cGAS recognizes cytosolic DNA, then
stimulates the formation of the STING ligand cyclic-​
GMP-AMP154. The revised model adds an extra level of
complexity by showing that pattern recognitions occur
within cGAS-​c ontaining micronuclei, which form
and accumulate only following cell cycle progression
through mitosis188. This dependence on actively cycling
cells could possibly explain the observed delayed onset
of inflammatory signalling upon radiotherapy and other
DNA double-​stranded-breaking therapies188.
By being readily available and free from patent rights,
radiotherapy presents the risk of being used as a sim-
ple add-​on to any immunotherapy, without a rationale
defining dose, fractionation, sequencing and timing. A
deeper knowledge on the molecular mechanisms trig-
gered by different regimens of radiotherapy within the
TME needs to be gained to carefully design efficient
therapeutic schemes73. There are evident limitations in
studying the effect of radiotherapy in mouse models,
including the intrinsic radiosensitivity and immuno-
genicity differences and the choice of tumour implanta-
tion site73. Nonetheless, abscopal effects were observed
in patients treated with anti-​CTLA4 mAbs and under-
going radiation regimens similar to those regimes used
in mice, highlighting the translational potential of mice
studies in this context73. As tumours can develop differ-
ent immune-​evasion strategies, the choice of radiation
regimen should be made according to the specifics of
the case. For example, as a low dose of radiation pro-
moted vascular normalization, such an approach could
be useful in excluded tumours73.
A promising strategy to promote T cell infiltration
into a cold tumour and convert it into a hot one came
from a study by Zheng et al. on a pancreatic cancer
mouse model189. Patients with CD8+Tlo PD-​L1hi pan-
creatic cancer respond poorly to treatment with ICIs,
vaccine or their combination, as well as to radiother-
apy. The development of a murine model that mim-
icked CD8+ Tlo PD-​L1hi pancreatic tumours enabled the
demonstration that the sequential combination of radio­
therapy, vaccination (with live cells expressing a model
immunodominant antigen) and checkpoint inhibition
(anti-​PD-L1 mAb) resulted in tumour regression and
improved survival compared with individual treatments
or radiotherapy plus vaccination189. It is likely that the
radiation enabled the recruitment of vaccine-​primed
T cells, which could exert their antitumour effects in a
no-​longer immunosuppressive microenvironment.
Despite the growing amount of information, the lack
of sufficient preclinical data impedes a guided design of
clinical trials, the results of which are often inconclusive
in attributing a therapeutic benefit to radiotherapy. It
would be challenging, yet crucial, to demonstrate une-
quivocally the contribution of radiation to immuno-
therapy response. An optimal integration of radiation
biology with tumour immunology could give rise to
potentially great clinical benefits190.
Chemotherapy. Anticancer agents can augment the
immunogenicity of tumour cells through two main
routes — the antigenicity191 and the adjuvanticity191,192
(Fig. 4). Genotoxic chemotherapies can induce mutations
leading to the generation of neo-​epitopes (therefore
www.nature.com/nrd
 Reviews

a b
Lymph vessel Immune priming Modulating
immune
response
Dendritic
cell

Oncolytic
virus Follicular
dendritic cell
MDSC
h
Adjuvanticity
TFH cell
Treg cell

B cell
c
TGFβ Expansion
TLS
Blood vessel
NK cell d
Recruitment
g CTL
e
Immunological Breaking
cell death tolerance
TAA PD-1
Invasive margin

f Chemotherapy
Immunogenicity Radiotherapy

Fig. 4 | schematic representation of treatments of immune cold tumours. The key factors or processes to be tackled to
achieve clinical benefit are the following: priming the immune response (for example, with a neo-​epitope cancer vaccine)
(a); modulating the immune response (for example, with transforming growth factor-​β (TGFβ) inhibitors) (b); expanding
cytotoxic T cell proliferation (for example, with interleukin-15 (IL-15)) (c); inducing recruitment of cytotoxic immune cells
(for example, by epigenetic modulation or the use of dendritic or T cell-​recruiting chemokines) (d); breaking tolerance
(using agents such as anti-​programmed cell death protein 1 (PD-1) and/or PD-1 ligand (PD-​L1) therapy) (e); inducing
immunogenicity (for example, with an instability inducer, such as an inhibitor of MutL homologue 1 (MLH1)) (f); inducing
immunological cell death (for example, by means of chemotherapy) (g); and inducing adjuvanticity (with an oncolytic virus)
(h). CTL , cytotoxic T lymphocyte; MDSC, myeloid-​derived suppressor cell; NK , natural killer ; TAA , tumour-​associated
antigen; TFH, follicular helper T cell; TLS, tertiary lymphoid structure; Treg cell, regulatory T cell.

Damage-​associated
molecular patterns
(DAMPs). Intracellular
molecules that are hidden from
immune recognition under
physiological conditions. These
molecules are secreted,
exposed or released upon
cellular stress or tissue injury
and recognized by pattern-​
recognition receptors expressed
on innate immune cells.
Tumour-​associated antigens
(TAAs). Antigens that are
preferentially expressed by
tumour cells but they can also
be found in normal tissues
(except for the TSAs, which are
exclusively expressed by
tumour cells). They can be
broadly categorized into
aberrantly expressed self-​
antigens, mutated self-​antigens
and TSAs.
increasing the antigenicity). However, such neoantigens
may be lowly expressed on tumour cells, thus having a
modest impact on the immune response193. Nonetheless,
chemotherapies that trigger an ICD — such as anthra-
cyclines, cyclophosphamide, oxaliplatin and taxanes —
can concomitantly increase the adjuvanticity by releasing
damage-​a ssociated molecular patterns (DAMPs) 194,195
and activating apoptotic or necroptotic pathways196.
The calreticulin exposure at the membrane provides
an ‘eat-​me’ signal that favours the transfer of tumour-​
associated antigens (TAAs) to DCs197. Dying tumour
cells can also stimulate a TLR3-dependent, cancer-cell-
autonomous type I interferon response, which induces
the local production of CXCL10, attracting T cells and
memory T cells to the tumour bed56,153,198. Even if these
ICD pathways have been elucidated only in mouse mod-
els, they might be clinically relevant. Indeed, chemother-
apy modified the local immune microenvironment in
patients with breast cancer199. Following neoadjuvant
chemotherapy (NAC), these patients had an increased
ratio of intratumoural CD8+ T cells to FOXP3+ cells200.
This event was accompanied by a clonal expansion of
antitumour T cells that correlated with response to NAC,
followed by a complete pathological response200. After
chemotherapy, the absence of autophagy-​related protein
LC3B (LC3B+) puncta and a low CD8+ to FOXP3+ cells
ratio were associated with a bad prognosis in patients
with breast cancer 201,202. Low calreticulin levels on
tumour cells correlated with a weaker immunosurveil-
lance in ovarian cancer204 and NSCLC203,204. Low levels
of antigen-​specific circulating T cells were associated
with poor clinical outcome in patients with acute mye-
loid leukaemia205,206. In CRC, neoadjuvant chemotherapy
increased the adaptive immune response, and there was
a significantly higher frequency of high Immunoscore
metastases in patients achieving pathological and
radiological responses50. Chemotherapy might also
stimulate the activation of immune effectors through
off-​target effects, resulting in general immune stimu-
lation207. Agents such as 5-fluorouracil deplete intra­
tumour MDSCs208, whereas cyclophosphamide depletes
Treg cells209 and triggers the translocation of immuno­
stimulatory bacteria from the gut lumen to the damaged
epithelium210. Effector T cells abrogate stroma-​mediated
chemoresistance in ovarian cancer211. IFNγ-​producing
CD8+ T cells may block cysteine and glutathione (both

Nature Reviews | Drug Discovery

Reviews
(LC3B+) puncta
LC3 is a protein involved in the
formation of autophagosomes
and autolysosomes. Punctate
(as opposed to diffused) LC3
staining indicates autophagy,
as determined by fluorescence
microscopy.
of which confer resistance to platinum-​based chemo- treatments may sensitize patients to further immuno-
therapy) synthesis by fibroblast-​associated tumour therapies. This approach might be a well-​suited option
cells211. Altogether, these observations support the ever-​in the case of immunosuppressed tumours, which
growingly endorsed hypothesis that chemotherapy is show potential for infiltration (no physical barriers)
not simply tumour suppressive but is also involved in but insufficient T cell response. However, it may be not
the positive modulation of the immune system. It could the most suitable choice in the case of cold tumours.
be postulated that the beneficial effect of chemotherapy In support of this hypothesis, a study by Spranger et al. in
depends on the presence of an adaptive pre-​existing patients with melanoma demonstrated that no dif-
immunity. In such a case, hot and altered immune ference exists between inflamed and non-​inflamed
tumours as measured by Immunoscore may be more tumours in terms of antigenic load and/or mutational
susceptible to chemotherapy (as well as radiotherapy) burden108. In this particular study, the lack of BATF3-
than cold tumours212. Furthermore, this would constitute lineage DCs may have been the cause for the ‘cold’
the rationale for coupling chemotherapy with ICIs. The TME. An assessment of the mutational load within the
success of the combination of anti-​PD-1 therapy plus tumour could provide a rational basis for the use of
chemotherapy in metastatic NSCLC213 demonstrates the antigenicity-​enhancing therapies.
strength of this dual approach. The proteasomal machinery, which leads to the pro-
cessing of peptides and their presentation on human leu-
Targeted therapies. An insufficient TAA load could kocyte antigen (HLA) class I (MHC I) molecules, has
in principle impair the mounting of an efficient T cell-​ been pinpointed as a novel exploitable point of inter-
mediated immune response. Therefore, therapies that vention to increase antigenicity224. The set of epitopes
increase the antigenicity could prove beneficial in pro- presented for CD8 recognition are collectively termed
moting the recruitment of T cells to tumour sites and the ‘immunopeptidome’. However, compelling evidence
the subsequent elimination of tumour cells (Fig.  4). indicates that the proteasome hosts a more complex
Tumour antigenicity can be enhanced by therapies that mechanism of epitope splicing, defined as proteasome-​
favour the re-​expression of TAAs, for example, DNA-​ catalysed peptide splicing, by which the fusion of two
demethylating agents (such as 5-azacytidine (AZA)) non-​contiguous peptidic segments generates novel
or epidermal growth factor receptor (EGFR) and MEK epitopes224. The proteasome-​generated spliced peptide
inhibitors214. AZA is a cytosine analogue and a potent pool accounts for one-​third of the entire HLA class I
DNA methyltransferase inhibitor that has been used for immunopeptidome in terms of diversity and one-​fourth
many years to treat myelodysplastic syndromes. The fact in terms of abundance224. Intriguingly, this newly iden-
that it induced a late clinical response in some patients tified pool represents a unique set of antigens that bear
suggested the possible implication of the immune sys- distinguishing immunological characteristics224, which
tem in its mode of action215. In fact, AZA upregulated makes it an attractive target for future therapeutic
MHC I, β2m and cancer testis antigen genes, as well as manipulations.
genes involved in IFNγ signalling216. In ovarian cancer,
AZA also induces a cytosolic double-​stranded RNA-​ DNA-​repair-based therapy. High mutational and
dependent type I interferon response by increasing the putative neoantigen load correlate with clinical bene­
expression of DNA hypermethylated endogenous retro- fit from immune checkpoint blockade therapy in
viruses (ERVs)217. AZA-​induced ERV transcripts were lung cancer and melanoma225. Therefore, strategies
found in melanoma218 and endometrial cancer cells219. that increase the burden of neoantigens in tumour
Patients can be stratified according to their basal ERV cells could in principle be used in combination with
levels and antiviral gene. A high antiviral gene signa- subsequent checkpoint inhibition. This notion is sup-
ture was significantly associated with durable clini- ported by a recent mouse study by Germano et al. that
cal responses in patients with melanoma treated with involved the genetic inactivation of DNA mismatch
anti-​CTLA4 therapy217. repair (MMR) protein MutL homologue 1 (MLH1) in
Inhibitors of EGFR, RET kinase and MEK are colorectal, breast and pancreatic cancer cells, ultimately
broadly used in the current clinical routine, despite the inactivating the cellular DNA MMR mechanisms, thus
lack of knowledge on their detailed molecular mech- inducing genomic instability and triggering immune
anisms of action. Their role as negative regulators of surveillance226. Apart from highlighting the importance
MHC I expression and antigen presentation machinery of neoantigen burden, this study points out the poten-
in multiple cancer types was revealed by a pooled short tial impact of strategies inhibiting the DNA damage
hairpin RNA interference-​based analysis of the human response (DDR) in tumour cells226. The inhibition of
kinome214. In vivo studies demonstrated that activated the DDR has been previously proposed to sensitize
MAPK signalling inhibited components of MHC I and cervical cancer cells to subsequent chemotherapy
the antigen presentation machinery214. The use of MAPK and/or radiotherapy, thereby contributing to the suc-
inhibitors indeed enhanced the T cell-​mediated killing cess of such treatments227. Numerous agents block-
of tumour cells214. ing DDR components, such as ATR serine/threonine
Proteins that prevent tumour cell apoptosis, includ- kinase, ATM (ataxia telangiectasia mutated), check-
ing members of the inhibitor of apoptosis protein point kinase 1 (CHK1), DNA-​d ependent protein
family220, MCL1 (ref.221) and mTOR222 (as well as IL-6 kinase (DNA-​PK), p38 MAPK and MAPK-activated
(ref. 223) ), represent further possible targets. Overall, protein kinase 2 (MK2), are currently being evaluated
given their ability to increase tumour antigenicity, these preclinically and clinically228.
www.nature.com/nrd

Adoptive cell therapy. Engineered T cells that express
artificial chimeric antigen receptors (CARs) targeting
a tumour cell surface molecule were first produced in
1993 (ref.229), but only since 2010 have they revealed their
true potential as anticancer agents230. CAR-​T cell-​based
therapies rely on the isolation, ex vivo manipulation
and expansion of antigen-​specific T cells, which are
subsequently transferred to the same patient (through
adoptive cell therapy)231. This method has shown sev-
eral potential advantages over conventional therapies,
including specificity, rapidity, high success rate and long-​
lasting effects232. The two presently approved therapies,
as well as most of the current clinical trials based on this
technology, are directed against CD19, a classical B cell
malignancy-​associated antigen.
Typically, CARs are transduced into T cells from the
patient using randomly integrating vectors233, which is
a limitation of this system as it may lead to undesired
effects such as oncogenic transformation, variable
expression levels and transcriptional silencing234,235. To
circumvent these limitations, Eyquem et al. directed
a CD19-specific CAR to the TCR α constant (TRAC)
locus by using CRISPR–Cas9-mediated genome editing,
yielding a uniform CAR expression in human peripheral
blood T cells236. By enhancing tumour rejection, such
optimization of the CAR-​T cell-​based technology is an
attractive option to adopt for future therapeutic designs.
A further optimization strategy to enhance the
efficacy of adoptive T cell therapy was suggested by
Hervas-​Stubbs et al., and it relies on priming of naive
CD8 + T cells with type I interferon 237. Specifically,
IFNα-​primed CD8+ T cells show enhanced ability to
persist and to mount a robust recall response compared
with naive CD8+ T cells while preserving a low differ-
entiation profile; in addition, they display heightened
responsiveness to IL-15 and IL-7, which mediate T cell
expansion and activation237. It could be envisaged that
combing IFNα-​primed ACT with subsequent cytokine
(IL-15 and/or IL-7) administration might constitute a
successful combinatorial therapy. Despite most studies
on CAR T cells focusing on targeting CD19 (that is,
B cell malignancies), a significant effort is being directed
at identifying alternative candidates given the more lim-
ited success against solid tumours. TSAs would consti-
tute ideal targets because their cognate T cells, unlike
T cells targeting TAAs, would not trigger undesirable
autoimmune reactions 238. Epitopes carrying driver
somatic mutations would be the best candidates as
they are TSAs and are critically involved in the process
of malignant transformation238. However, their limited
mutation frequency and the restrictions offered by all
the steps of the antigen presentation pathway make
these tumour-​specific mutated epitopes hard to target.
The enlarged immunopeptidome spectrum recently
revealed with the existence of spliced epitopes might
therefore have a profound impact on the adoptive T cell
therapy field238.
In a study by Verdegaal et al. that included two
patients with stage IV melanoma, tumour-​specific
T cells were expanded by repeated stimulation with cell
lines established from the resected lesion239. Albeit labo-
rious, ACT proved to be effective, as both patients were
Nature Reviews | Drug Discovery
Reviews
long-​term survivors. It should be noted that melanoma
represents the solid cancer type with the best response
to ACT thus far43. The ACT was also useful to analyse
the stability of neoantigen-​specific T cell responses
and the antigens they are directed against239. The study
revealed a highly dynamic environment, in which the
T cell-​recognized neoantigens selectively disappeared
from the tumour cell population, with a concomitant
development of neoantigen-​specific T cells among the
TILs239. Importantly, the authors suggested the occur-
rence of T cell-​mediated neoantigen immunoediting;
therefore, a broad neoantigen-​specific T cell response
should be sought to avoid tumour resistance239. In fact,
this is a concept that should be considered in all strat-
egies aiming at inducing neoantigen-​specific immune
responses — the broader, the better.
The crucial question at this point is whether ACT
therapies will be able to overcome failed spontaneous
T cell priming and convert cold into hot tumours.
This question was addressed in a non-​T cell-​inflamed
β-​c atenin-expressing murine melanoma model by
Spranger et al.240. The study revealed failed trafficking
of tumour-​specific effector T cells that had been adop-
tively transferred into tumours owing to the absence of
the T cell-​recruiting chemokines CXCL9 and CXCL10
(ref.240). The same question was also addressed in patients
with non-​Hodgkin lymphoma undergoing anti-​CD19
CAR-​T cell therapy in a multicentre trial (ZUMA-1) and
revealed that a stronger immune contexture predicted
increased likelihood of response, supporting the notion
that CAR-​T cell therapy may not be enough by itself to
treat patients with cold tumours241.
Taken together, these results further reiterate the
idea that a deeper analysis of the TME and a consequent
tumour classification are required before and/or during
any therapeutic intervention.
Oncolytic therapy. Despite being known for nearly a
century, the ability of viruses to kill tumour cells, and
hence their therapeutic benefit in cancer patients, has
been documented only recently in several clinical tri-
als242. Oncolytic viruses are native or genetically mod-
ified viruses that selectively infect and replicate within
tumour cells, eventually leading to tumour cell lysis242.
Alongside this direct and local antitumour activity,
oncolytic viruses can also induce a potent, systemic and
potentially durable antitumour immunity242. The dying
tumour cells release TAAs and additional DAMPs,
thereby eliciting an efficient antitumour immune
response242. In fact, the power of this approach lies in
the ultimate engagement of systemic immunity, which
results in therapeutic responses not only at the site of
injection but also at distant tumour sites242.
Cancer cells represent a fertile environment for viral
replication owing to their intrinsic abnormalities in the
signalling pathways involved in cell stress and homeo-
stasis and, possibly, in the antiviral machinery. The lat-
ter should be carefully considered in individual patients
as disease-​induced deregulation of the host-​antiviral
mechanisms can influence the therapeutic activity of the
oncolytic viruses. For example, protein kinase R (PKR),
which helps in the clearing of intracellular viruses, may
Reviews
Differentiation antigens
Antigens derived from proteins
that are expressed in a given
type of tumour and the
corresponding healthy tissue,
often in lower amounts.
be differentially expressed and/or activated in different
cancer types242,243.
In order to be used therapeutically, oncolytic viruses
have to be devoid of virulence factors yet retain their
immunostimulatory abilities. Many viruses have been
engineered accordingly and assessed in clinical trials,
including adenoviruses, poxviruses, herpes simplex virus
type 1 (HSV-1), coxsackieviruses, poliovirus, measles
virus, Newcastle disease virus (NDV) and reovirus242.
Many of the currently evaluated oncolytic viruses have
a natural tropism for cell surface proteins overexpressed
by cancer cells. Two examples are CD46 and HVEM,
which are cell entry receptors for the Edmonton strain
of measles virus and HSV-1, respectively242.
Talimogene laherparepvec (T-​VEC) represents the
first FDA-​approved virotherapeutic approach for
the localized treatment of patients with unresectable
melanoma. As mentioned above, a potentially thera-
peutic strategy to ‘heat up’ cold tumours could lie in the
use of an oncolytic virus as priming therapy, combined
with the removal of co-​inhibitory signals (Fig. 4). Indeed,
T-​VEC administration in combination with ipilimumab
(an anti–CTLA4 antibody) in patients with unresecta-
ble stage IIIB–IV melanoma yielded greater antitumour
activity than ipilimumab alone244. A phase Ib study using
T-​VEC followed by the anti-​PD-1 antibody pembroli-
zumab in patients with advanced melanoma resulted
in an outstanding response rate of 62%245. Patients who
responded to treatment with T-​VEC displayed increased
CD8+ T cells, high mutational burden and high expres-
sion of PD-​L1 protein and IFNG in several cell subsets
in tumours245, suggesting the need for a pre-​existing
tumour-​specific T cell pool for the anti-​PD-1 therapy
to be effective 246. Importantly, this improved anti­
tumour activity did not come at the expense of the safety
profile244,245. Despite being highly encouraging, the per-
centage of responders to the combination treatment
shows still room for improvement and further points
of interventions should be envisaged. A possible rea-
son for therapeutic failure could lie in the possible poor
antigenicity (low TAAs) of cold tumours; in this case,
virion-​mediated lysis of cancer cells may not release
enough TAAs to prime antitumour T cell responses.
Vaccine-​based therapy (discussed in the next section)
may represent a more suitable option in this context.
Another intratumour therapeutic approach to achieve
an abscopal effect is offered by the use of oncolytic pep-
tides. The nine amino acid residue (9-mer) oncolytic
peptide LTX-315 acts on both drug-​resistant and drug-​
sensitive cancer cells, rather than healthy cells, thereby
causing the lysis of their plasma and organelle mem-
branes, which act as danger signals initiating an innate
and subsequent systemic adaptive immune response247.
The intratumour injection of LTX-315 resulted in a
complete rejection of fibrosarcomas established in a rat
model. This tumour rejection relied on T cell infiltration
in both injected and distal tumour sites247.
Physical barriers (including necrosis, calcification,
hypoxia, acidosis, increased proteolytic activity, high
interstitial pressure, poor vascularization and/or dense
extracellular matrix) may reduce the spread of oncolytic
viruses, limiting their biodistribution and consequent
therapeutic efficacy242. Although this limitation can
be overcome in physically accessible tumours by intra­
tumoural injections, the same factors could limit T cell
trafficking and the establishment of a successful immune
response. In other words, it is reasonable to hypothesize
the existence of cases in which such priming therapy
only ‘promotes’ a tumour to an altered (excluded in this
example) phenotype; this will then additionally require
not only an ICI but also an additional strategy (such as
anti-​VEGF) to become hot.
Vaccine-​based therapy. After an initial identification
of neoantigens from tumour cells, putative antigen or
antigenic epitope or epitopes can be presented through a
variety of platforms, such as whole tumour cell prepara-
tions, through MHC-​specific peptides, whole or partial
proteins encoded by RNA or DNA, or in recombinant
viral or bacterial vectors expressed in DCs248. The use
of additional vaccine adjuvants, such as TLR agonists,
could boost the immune response against TAAs. Despite
seemingly valuable, anticancer vaccines turned out to be
quite disappointing, as proved by the low overall objec-
tive response rate obtained in numerous clinical trials248
(Fig. 2c). However, this lack of success translated into a
quest for the reasons underlying this ineffectiveness. A
key aspect to consider is that anticancer vaccines were
tested in patients with established cancers, in which
immunosuppressive mechanisms were already in place.
Therefore, the increase in ICIs brought along a renewed
interest in these therapeutic approaches248. Indeed, stud-
ies on ICIs highlighted the positive correlation of the
somatic mutation burden and consequent emergence
of neoantigens with clinical benefit225,249, providing a
rationale for combination therapies involving ICIs and
T cell priming anticancer vaccines (Fig. 4). ICIs would de
facto play the part of vaccine adjuvants. It is tempting
to speculate that future studies involving the combina-
tion of T cell boosting tumour vaccines with the T cell
suppression-​preventing ICIs may translate into clinical
benefit in patients with cold tumours.
A challenging aspect of anticancer vaccination is
finding the optimal antigen for vaccination. Among the
possible candidates are overexpressed self-​proteins (such
as prostate-​specific antigen (PSA)), differentiation antigens
(such as protein Melan-​A or gp100 (ref.250)) and mutated
antigens (neoantigens)248. Thus far, there is only one
FDA-​approved vaccine, sipuleucel-​T, for the treatment
of asymptomatic or minimally symptomatic hormone-​
refractory prostate cancer. Sipuleucel-​T consists of
autologous DCs loaded with recombinant human fusion
protein encoding the prostatic acid phosphatase (PAP)
antigen (the expression of which increases with pros-
tate cancer progression) and granulocyte–macrophage
colony-​stimulating factor (GM-​CSF) to sustain DC
maturation.
The selective detection of cytomegalovirus (CMV)
antigens in certain types of tumour (such as glioblas-
toma (GBM)) but not in normal tissue suggested the
opportunity to use immunological interventions that
target these viral proteins. Accordingly, vaccination
with CMV pp65 RNA-​pulsed DCs was developed to
treat GBM251. Unfortunately, this approach did not
www.nature.com/nrd

Neoantigen fitness
The likelihood of a peptide to
be immunogenic, as measured
by its binding affinity to major
histocompatibility complex
(MHC) and subsequent
recognition by T cells.
Nature Reviews | Drug Discovery
translate into higher clinical benefit than the stand-
ard of care251,252. However, a clear clinical benefit was
observed when preconditioning the vaccine site with
tetanus–diphtheria (Td) toxoid; such a potent recall anti-
gen significantly improved lymph node homing and the
efficacy of tumour antigen-​specific DCs251. Interestingly,
the blinded interim data of the overall patient popula-
tion enrolled in a phase III randomized, double-​blind,
placebo-​controlled clinical trial of an autologous tumour
lysate-​pulsed DC vaccine (DCVax-​L) for newly diag-
nosed GBM revealed extended survival253. Saying that
tumour vaccines could per se provide sufficient basis to
convert a cold into a hot tumour is an overstatement, as
this example clearly shows. Notwithstanding, it shows
once again the power of a multifactorial approach and
opens the intriguing perspective of exploiting past
vaccinations against infectious diseases.
Other vaccines targeting defined tumours or TAAs
are currently being evaluated as they present obvious
advantages, including the possibility of mass production,
easy monitoring of immune responses and a generally
tolerable safety profile248.
The broad range of neoantigens and their positive
association with improved clinical responses suggested
their possible use for vaccination. Furthermore, as new
epitopes appear continuously during tumour progres-
sion, immune editing and T cell suppression seem
unlikely as they require time248. A neoantigen fitness
model, which measured the likelihood of neoantigen
presentation by the MHC and subsequent recognition
by T cells, predicted survival in patients with melanoma
treated with anti-​CTLA4 therapy and patients with lung
cancer treated with PD-1 blockade254. Newly identified
low-​fitness neoantigens could constitute ideal targets
for developing novel cancer vaccines254. A difficulty
of this approach is its intrinsic personalized nature —
hence the bench-​to-bedside time frame. Nonetheless,
the development and optimization of high-​throughput
screening techniques and epitope-​predicting algorithms
may allow for such strategy. The recently reported
RNA-​b ased poly-​n eo-epitope approach used in
13 patients with melanoma made the concept of indi-
vidualized mutanome vaccines a promising reality255.
All patients completed treatment with a maximum of
20 neo-​epitope vaccine doses, which were well tolerated,
and developed T cell responses against at least three
mutations, resulting in sustained progression-​free sur-
vival255. This strategy identified a patient in which the
combination of the vaccine and PD-1 blockade yielded
a complete response. Another study demonstrated the
feasibility, safety and immunogenicity of a vaccine that
targets up to 20 predicted personal tumour neoantigens.
Of six vaccinated patients, four had no recurrence at
25 months after vaccination, whereas two with recur-
rent disease were subsequently treated with anti-​PD-1
treatment and experienced complete tumour regression,
with expansion of the repertoire of neoantigen-​specific
T cells256. Apart from the undoubted merit of paving
the way to personalized vaccine-​based immunother-
apy, this study supplied evidence of the potential of
the combinatorial therapeutic approach with current
immunotherapies.
Reviews
T cell immunomodulators. Several cytokines (such as
IL-2, IL-7, IL-15, IL-21, IL-12, GM-​CSF and IFNα) are
known to modulate T cell expansion, survival and/or
function and were or are being tested in clinical trials as
potential anticancer agents, in monotherapy or in combi-
nation257. IL-2 was the first cytokine to be characterized
as a ‘T cell growth factor’ (TCGF) in 1976 (ref.258). IL-2
became an early candidate for cancer immunotherapy
and showed promising results as a single agent in met-
astatic RCC and melanoma258. However, its pleiotropic
effects on the immune system (notably, the targeting
of Treg cells) as well as the severe associated side effects
greatly reduced its importance as a single anticancer
agent258. Combinations with traditional anticancer ther-
apies as well as immunotherapies, including ACTs, have
since been investigated but displayed similar limitations
to the monotherapy. A renewed interest in IL-2 came
with the development of a strategy to redirect the speci-
ficity of IL-2 towards adoptively transferred T cells259. By
using orthogonal IL-2 cytokine–receptor pairs (mutant
IL-2 and IL-2 receptor (IL-2R)), this approach enables
selective expansion of desired T cells, with negligible
off-​target effects and toxicity259.
The IL-2-related cytokine IL-15 has lately attracted
attention in the cancer immunotherapy field as it does
not target Treg cells but NK cells and lacks other unde-
sired features of IL-2 (ref.260). Reduced IL-15 expression
due to chromosomal instability impaired the intratu-
moural T and B cell proliferation and correlated with
higher risk of tumour recurrence and reduced survival
of patients with CRC18, providing a clear indication of
its key anticancer role. Current efforts aim at ensuring
sufficient bioavailability, which is low at present owing
to rapid renal clearance260. ALT-803 is a super-​agonist
of the IL-15–IL-15Rα complex that successfully pro-
longed the half-​life of IL-15 and promoted enhanced
immune activation in vivo261. ALT-803 was well tolerated
and resulted in clinical responses in patients with hae-
matological malignancies who relapsed after allogeneic
haematopoietic cell transplantation261.
IL-7 is yet another potent growth, activation and
survival factor for CD8+ T cells that could improve anti-
tumour CD8+ T cell responses262. Owing to its features,
it is not surprising that recombinant IL-7 (rIL-7) was
considered as a tool to improve ACT263. Indeed, an ACT
mouse model demonstrated the efficacy of rIL-7 in
inducing CD8+ T cell proliferation and differentiation,
and tumour rejection, in an IL-7R-​dependent manner263.
Combining T cell-​modulating cytokines with other
approaches could improve the therapeutic success. This
possibility was demonstrated by an in vivo study that
featured a modified NDV that co-​expressed IL15 and
IL7 (ref.264). Such a tumour vaccine displayed improved
antitumour activity compared with a non-​modified
vaccine264.
IL-21 is a promoter of T cell responses265 and could
be exploited for its potential antitumour activity. IL-21
counteracts the in vitro TGFβ1-induced FOXP3 expres-
sion in purified naive CD4 T cells, confirming its known
additional ability to alleviate Treg-​mediated immuno-
suppression in several cancers266. Apart from targeting
T cells, IL-21 also regulates TFH (ref.267) and B cells265.
Reviews

Together with CXCL13, IL-21 is a crucial element reg-
ulating the TFH–B cell axis within the TME in CRC,
and its presence correlated with increased patient
survival24. In addition, Lewis et al. demonstrated the
synergistic antitumour activity of IL-21 and anti-​
CTLA4 and/or anti-​PD-1 therapy in various tumour
models268. The decreased percentages of intratumour
CD4 +CD25 +FOXP3 + T  cells were accompanied by
increased CD8+ T cell infiltrates, CD8+ T cell prolifer-
ation, increased levels of effector memory T cells and
overall enhanced antitumour activity268. Combining the
adoptive transfer of IL-21-primed, melanoma-​reactive
CD8+ T cells with anti-​CTLA4 therapy controlled refrac-
tory metastatic melanoma in a patient269, showing the
potential of IL-21 in combinatorial anticancer therapies.
Concluding remarks
The concept of personalized cancer immunotherapies
has been ever-​growingly advocated in recent years.
Possibly the biggest challenge impeding the achieve-
ment of such an ambitious approach lies in the lack of
a comprehensive knowledge of the cancer–immune
interaction parameters; even when this knowledge
is available, standardized methods of measuring
most parameters are lacking. Accurate measurement
of parameters is quite crucial, as their relative ‘weight’
varies considerably among individual patients62. It is
clear that the identification of key features, immune-​
related or tumour-​related, at the moment of diagnosis is
needed to build a solid classification strategy supporting
subsequent therapy. Here, we provide a panel of thera-
peutic strategies to be deployed or developed (if not yet
available) against hot, altered and cold tumours. It is
reasonable to assume that the colder the tumour is, the
more approaches are needed. Interestingly, all proposed
therapeutic designs ultimately involve combinations
with immunotherapies to achieve maximal efficiency.
Given the pivotal role of T cells against cancer, the care-
ful assessment of the pre-​existing T cell landscape and
of the immune microenvironment should be routinely
used to differentiate specific cases, not only in the clin-
ical setting but also at the preclinical stage. This infor-
mation could help in the translation of study findings
in clinical indications.
Published online xx xx xxxx

1. Shankaran, V. et al. IFNgamma and lymphocytes
prevent primary tumour development and shape
tumour immunogenicity. Nature 410, 1107–1111
(2001).
2. Schreiber, R. D., Old, L. J. & Smyth, M. J. Cancer
immunoediting: integrating immunity’s roles in cancer
suppression and promotion. Science 331,
1565–1570 (2011).
3. Galon, J. et al. Type, density, and location of immune
cells within human colorectal tumors predict clinical
outcome. Science 313, 1960–1964 (2006).
This article is the first demonstration of the
dependence of tumour progression and invasion on
the intratumoural adaptive immunity. T cell
infiltrates and IFNγ signatures have predictive
value superior to TNM with respect to the natural
history of primary cancers.
4. Hodi, F. S. et al. Improved survival with ipilimumab in
patients with metastatic melanoma. N. Engl. J. Med.
363, 711–723 (2010).
5. Topalian, S. L. et al. Safety, activity, and immune
correlates of anti-​PD-1 antibody in cancer. N. Engl.
J. Med. 366, 2443–2454 (2012).
6. Tumeh, P. C. et al. PD-1 blockade induces responses
by inhibiting adaptive immune resistance. Nature
515, 568–571 (2014).
This article shows that the baseline density and
location at the invasive margin of T cells in
metastatic melanomas predicts the treatment
outcome of patients receiving PD-1-targeting
therapies.
7. Van Allen, E. M. et al. Genomic correlates of
response to CTLA-4 blockade in metastatic
melanoma. Science 350, 207–211 (2015).
This article is the first to show that patients
with metastatic melanoma with high
mutational burden, neoantigen load and
expression of cytolytic markers in their tumours
are more likely to respond to anti-​CTLA4
immunotherapy.
8. Huang, A. C. et al. T cell invigoration to tumour
burden ratio associated with anti-​PD-1 response.
Nature 545, 60–65 (2017).
9. Mlecnik, B. et al. Histopathologic-​based prognostic
factors of colorectal cancers are associated with the
state of the local immune reaction. J. Clin. Oncol. 29,
610–618 (2011).
10. Galon, J., Angell, H. K., Bedognetti, D.
& Marincola, F. M. The continuum of cancer
immunosurveillance: prognostic, predictive, and
mechanistic signatures. Immunity 39, 11–26
(2013).
11. Angell, H. & Galon, J. From the immune contexture to
the Immunoscore: the role of prognostic and
predictive immune markers in cancer. Curr. Opin.
Immunol. 25, 261–267 (2013).
12. Galluzzi, L. et al. Trial watch: dendritic cell-​based
interventions for cancer therapy. Oncoimmunology 1,
1111–1134 (2012).
13. Galon, J., Fridman, W. H. & Pages, F. The adaptive
immunologic microenvironment in colorectal cancer:
a novel perspective. Cancer Res. 67, 1883–1886
(2007).
14. Pages, F. et al. In situ cytotoxic and memory T cells
predict outcome in patients with early-​stage colorectal
cancer. J. Clin. Oncol. 27, 5944–5951 (2009).
15. Galon, J. et al. Towards the introduction of the
‘Immunoscore’ in the classification of malignant
tumours. J. Pathol. 232, 199–209 (2014).
16. Pages, F. et al. International validation of the
consensus Immunoscore for the classification of colon
cancer: a prognostic and accuracy study. Lancet 391,
2128–2139 (2018).
This article validates Immunoscore as a consensus
and standardized cytotoxic T cell assay that defines
immune hot, altered and cold tumours and has a
greater prognostic value in CRC than T stage,
N stage, lymphovascular invasion, tumour
differentiation and MSI status.
17. Camus, M. et al. Coordination of intratumoral immune
reaction and human colorectal cancer recurrence.
Cancer Res. 69, 2685–2693 (2009).
This paper presents the first description of the
immune hot (optimal), altered-​excluded, altered-​
immunosuppressed and cold (absent) tumours.
18. Mlecnik, B. et al. Functional network pipeline reveals
genetic determinants associated with in situ
lymphocyte proliferation and survival of cancer
patients. Sci. Transl Med. 6, 228ra37 (2014).
19. Mlecnik, B. et al. Integrative analyses of colorectal
cancer show Immunoscore is a stronger predictor of
patient survival than microsatellite instability.
Immunity 44, 698–711 (2016).
20. Boland, C. R. & Goel, A. Microsatellite instability in
colorectal cancer. Gastroenterology 138, 2073–2087
(2010).
21. Gajewski, T. F. et al. Cancer immunotherapy targets
based on understanding the T cell-​inflamed versus
non-​T cell-​inflamed tumor microenvironment. Adv.
Exp. Med. Biol. 1036, 19–31 (2017).
22. Hegde, P. S., Karanikas, V. & Evers, S. The where, the
when, and the how of immune monitoring for cancer
immunotherapies in the era of checkpoint inhibition.
Clin. Cancer Res. 22, 1865–1874 (2016).
23. Angelova, M. et al. Evolution of metastases in space and
time under immune selection. Cell 175, 601 (2018).
24. Bindea, G. et al. Spatiotemporal dynamics of
intratumoral immune cells reveal the immune
landscape in human cancer. Immunity 39, 782–795
(2013).
This article is the first to describe the immunome
from immune signatures of purified immune cell
subpopulations applied to human tumours.
Immune infiltrate composition changes at each
tumour stage and T, B and TFH cells have a major
impact on survival.
25. Gu-​Trantien, C. et al. CXCL13-producing TFH
cells link immune suppression and adaptive
memory in human breast cancer. JCI Insight 2, 91487
(2017).
26. Chew, V. et al. Chemokine-​driven lymphocyte
infiltration: an early intratumoural event determining
long-​term survival in resectable hepatocellular
carcinoma. Gut 61, 427–438 (2012).
27. Gajewski, T. F., Schreiber, H. & Fu, Y. X. Innate and
adaptive immune cells in the tumor microenvironment.
Nat. Immunol. 14, 1014–1022 (2013).
28. Goc, J. et al. Dendritic cells in tumor-​associated
tertiary lymphoid structures signal a Th1 cytotoxic
immune contexture and license the positive prognostic
value of infiltrating CD8+T cells. Cancer Res. 74,
705–715 (2014).
29. Ingels, A. et al. T-​helper 1 immunoreaction
influences survival in muscle-​invasive bladder cancer:
proof of concept. Ecancermedicalscience 8, 486
(2014).
30. Mulligan, A. M., Pinnaduwage, D., Tchatchou, S.,
Bull, S. B. & Andrulis, I. L. Validation of intratumoral
T-​bet+lymphoid cells as predictors of disease-​free
survival in breast cancer. Cancer Immunol. Res. 4,
41–48 (2016).
31. Mulligan, A. M. et al. Tumoral lymphocytic infiltration
and expression of the chemokine CXCL10 in
breast cancers from the Ontario Familial Breast
Cancer Registry. Clin. Cancer Res. 19, 336–346
(2013).
32. Cristescu, R. et al. Pan-​tumor genomic biomarkers for
PD-1 checkpoint blockade-​based immunotherapy.
Science 362, eaar3593 (2018).
33. Cheon, H., Borden, E. C. & Stark, G. R. Interferons
and their stimulated genes in the tumor
microenvironment. Semin. Oncol. 41, 156–173
(2014).
34. Lin, C. F. et al. Escape from IFN-​gamma-dependent
immunosurveillance in tumorigenesis. J. Biomed. Sci.
24, 10 (2017).
35. Zaidi, M. R. & Merlino, G. The two faces of interferon-​
gamma in cancer. Clin. Cancer Res. 17, 6118–6124
(2011).
36. Mandai, M. et al. Dual faces of IFNgamma in cancer
progression: a role of PD-​L1 induction in the
determination of pro- and antitumor immunity. Clin.
Cancer Res. 22, 2329–2334 (2016).
37. Snell, L. M., McGaha, T. L. & Brooks, D. G. Type I
interferon in chronic virus infection and cancer. Trends
Immunol. 38, 542–557 (2017).
38. Minn, A. J. & Wherry, E. J. Combination cancer
therapies with immune checkpoint blockade:

www.nature.com/nrd

convergence on interferon signaling. Cell 165,
272–275 (2016).
39. Benci, J. L. et al. Tumor interferon signaling regulates
a multigenic resistance program to immune
checkpoint blockade. Cell 167, 1540–1554 (2016).
40. Sun, T. et al. Inhibition of tumor angiogenesis by
interferon-​gamma by suppression of tumor-​associated
macrophage differentiation. Oncol. Res. 21, 227–235
(2014).
41. Kim, H. J. & Cantor, H. CD4 T cell subsets and tumor
immunity: the helpful and the not-​so-helpful. Cancer
Immunol. Res. 2, 91–98 (2014).
42. Placek, K., Coffre, M., Maiella, S., Bianchi, E. &
Rogge, L. Genetic and epigenetic networks controlling
T helper 1 cell differentiation. Immunology 127,
155–162 (2009).
43. Kato, D. et al. Prospects for personalized combination
immunotherapy for solid tumors based on adoptive
cell therapies and immune checkpoint blockade
therapies. Nihon Rinsho Meneki Gakkai Kaishi 40,
68–77 (2017).
44. Spranger, S., Bao, R. & Gajewski, T. F. Melanoma-​
intrinsic beta-​catenin signalling prevents anti-​tumour
immunity. Nature 523, 231–235 (2015).
This article shows that a melanoma cell-​intrinsic
oncogenic pathway (active β-​catenin signalling)
contributes to a lack of T cell infiltration in tumour
sites and resistance to anti-​PD-L1 and/or
anti-​CTLA4 mAb therapy.
45. Yaguchi, T. et al. Immune suppression and resistance
mediated by constitutive activation of Wnt/beta-​catenin
signaling in human melanoma cells. J. Immunol. 189,
2110–2117 (2012).
46. Sumimoto, H., Imabayashi, F., Iwata, T. & Kawakami, Y.
The BRAF-​MAPK signaling pathway is essential for
cancer-​immune evasion in human melanoma cells.
J. Exp. Med. 203, 1651–1656 (2006).
47. Sumimoto, H. et al. Inhibition of growth and invasive
ability of melanoma by inactivation of mutated BRAF
with lentivirus-​mediated RNA interference. Oncogene
23, 6031–6039 (2004).
48. Iwata-​Kajihara, T. et al. Enhanced cancer
immunotherapy using STAT3-depleted dendritic cells
with high Th1-inducing ability and resistance to cancer
cell-​derived inhibitory factors. J. Immunol. 187,
27–36 (2011).
49. Nishio, H. et al. Immunosuppression through
constitutively activated NF-​kappaB signalling in human
ovarian cancer and its reversal by an NF-​kappaB
inhibitor. Br. J. Cancer 110, 2965–2974 (2014).
50. Mlecnik, B. et al. Comprehensive intrametastatic
immune quantification and major impact of
immunoscore on survival. J. Natl Cancer Inst. 110,
97–108 (2018).
51. Zhang, A. W. et al. Interfaces of malignant and
immunologic clonal dynamics in ovarian cancer.
Cell 173, 1755–1769 (2018).
52. Yoshida, M. et al. Modification of the tumor
microenvironment in KRAS or c-​MYC-induced ovarian
cancer-​associated peritonitis. PLOS ONE 11,
e0160330 (2016).
53. McFarland, C. D. et al. The damaging effect of
passenger mutations on cancer progression. Cancer
Res. 77, 4763–4772 (2017).
54. Tauriello, D. V. F. et al. TGFbeta drives immune evasion
in genetically reconstituted colon cancer metastasis.
Nature 554, 538–543 (2018).
55. Zitvogel, L. & Kroemer, G. Targeting PD-1/PD-​L1
interactions for cancer immunotherapy.
Oncoimmunology 1, 1223–1225 (2012).
56. Mlecnik, B. et al. Biomolecular network reconstruction
identifies T cell homing factors associated with survival
in colorectal cancer. Gastroenterology 138,
1429–1440 (2010).
57. van der Woude, L. L., Gorris, M. A. J., Halilovic, A.,
Figdor, C. G. & de Vries, I. J. M. Migrating into the
tumor: a roadmap for T cells. Trends Cancer 3,
797–808 (2017).
58. Ahmadzadeh, M. et al. Tumor antigen-​specific CD8
T cells infiltrating the tumor express high levels of
PD-1 and are functionally impaired. Blood 114,
1537–1544 (2009).
59. Gros, A. et al. PD-1 identifies the patient-​specific
CD8(+) tumor-​reactive repertoire infiltrating human
tumors. J. Clin. Invest. 124, 2246–2259 (2014).
60. Ji, R. R. et al. An immune-​active tumor
microenvironment favors clinical response to
ipilimumab. Cancer Immunol. Immunother. 61,
1019–1031 (2012).
61. Eroglu, Z. et al. High response rate to PD-1 blockade
in desmoplastic melanomas. Nature 553, 347–350
(2018).
Nature Reviews | Drug Discovery
62. Blank, C. U., Haanen, J. B., Ribas, A.
& Schumacher, T. N. The “cancer immunogram”.
Science 352, 658–660 (2016).
63. Wieland, A. et al. T cell receptor sequencing of
activated CD8 T cells in the blood identifies tumor-​
infiltrating clones that expand after PD-1 therapy and
radiation in a melanoma patient. Cancer Immunol.
Immunother. 67, 1767–1776 (2018).
64. Simon, S. et al. Emergence of high-​avidity Melan-​
A-specific clonotypes as a reflection of anti-​PD-1
clinical efficacy. Cancer Res. 77, 7083–7093 (2017).
65. Riaz, N. et al. Tumor and microenvironment evolution
during immunotherapy with nivolumab. Cell 171,
934–949 (2017).
66. Fife, B. T. et al. Insulin-​induced remission in new-​onset
NOD mice is maintained by the PD-1–PD-​L1 pathway.
J. Exp. Med. 203, 2737–2747 (2006).
67. Whiteside, T. L., Demaria, S., Rodriguez-​Ruiz, M. E.,
Zarour, H. M. & Melero, I. Emerging opportunities and
challenges in cancer immunotherapy. Clin. Cancer Res.
22, 1845–1855 (2016).
68. Weng, N. P., Araki, Y. & Subedi, K. The molecular basis
of the memory T cell response: differential gene
expression and its epigenetic regulation. Nat. Rev.
Immunol. 12, 306–315 (2012).
69. Durgeau, A., Virk, Y., Corgnac, S. & Mami-​Chouaib, F.
Recent advances in targeting CD8 T-​Cell immunity for
more effective cancer immunotherapy. Front. Immunol.
9, 14 (2018).
70. Mlecnik, B. et al. The tumor microenvironment and
Immunoscore are critical determinants of
dissemination to distant metastasis. Sci. Transl Med.
8, 327ra26 (2016).
71. Pages, F. et al. Effector memory T cells, early
metastasis, and survival in colorectal cancer. N. Engl.
J. Med. 353, 2654–2666 (2005).
72. Church, S. E. & Galon, J. Tumor microenvironment and
immunotherapy: the whole picture is better than a
glimpse. Immunity 43, 631–633 (2015).
73. Demaria, S., Coleman, C. N. & Formenti, S. C.
Radiotherapy: changing the game in immunotherapy.
Trends Cancer 2, 286–294 (2016).
74. Sharma, P. & Allison, J. P. The future of immune
checkpoint therapy. Science 348, 56–61 (2015).
75. Koyama, S. et al. Adaptive resistance to therapeutic
PD-1 blockade is associated with upregulation of
alternative immune checkpoints. Nat. Commun. 7,
10501 (2016).
76. Shayan, G. et al. Adaptive resistance to anti-​PD1
therapy by Tim-3 upregulation is mediated by the
PI3K-​Akt pathway in head and neck cancer.
Oncoimmunology 6, e1261779 (2017).
77. Granier, C. et al. Tim-3 expression on tumor-​
infiltrating PD-1(+)CD8(+) T cells correlates with poor
clinical outcome in renal cell carcinoma. Cancer Res.
77, 1075–1082 (2017).
78. Hellmann, M. D., Friedman, C. F. & Wolchok, J. D.
Combinatorial cancer immunotherapies. Adv.
Immunol. 130, 251–277 (2016).
79. Eggermont, A. M. M. et al. Adjuvant pembrolizumab
versus placebo in resected stage III melanoma.
N. Engl. J. Med. 378, 1789–1801 (2018).
80. Buchbinder, E. I. & Desai, A. CTLA-4 and PD-1
pathways: similarities, differences, and implications of
their inhibition. Am. J. Clin. Oncol. 39, 98–106
(2016).
81. Taube, J. M. Unleashing the immune system: PD-1 and
PD-​Ls in the pre-​treatment tumor microenvironment
and correlation with response to PD-1/PD-​L1
blockade. Oncoimmunology 3, e963413 (2014).
82. Taube, J. M. et al. Implications of the tumor immune
microenvironment for staging and therapeutics. Mod.
Pathol. 31, 214–234 (2018).
83. Wolchok, J. D. et al. Overall survival with combined
nivolumab and ipilimumab in advanced melanoma.
N. Engl. J. Med. 377, 1345–1356 (2017).
84. Motzer, R. J. et al. Nivolumab plus ipilimumab versus
sunitinib in advanced renal-​cell carcinoma. N. Engl.
J. Med. 378, 1277–1290 (2018).
85. Hellmann, M. D. et al. Nivolumab plus ipilimumab in
lung cancer with a high tumor mutational burden.
N. Engl. J. Med. 378, 2093–2104 (2018).
86. Du, W. et al. TIM-3 as a target for cancer
immunotherapy and mechanisms of action. Int. J. Mol.
Sci. 18, E645 (2017).
87. Manieri, N. A., Chiang, E. Y. & Grogan, J. L. TIGIT:
a key inhibitor of the cancer immunity cycle. Trends
Immunol. 38, 20–28 (2017).
88. Sedy, J. R. et al. B and T lymphocyte attenuator
regulates T cell activation through interaction with
herpesvirus entry mediator. Nat. Immunol. 6, 90–98
(2005).
Reviews
89. Gao, J. et al. VISTA is an inhibitory immune checkpoint
that is increased after ipilimumab therapy in patients
with prostate cancer. Nat. Med. 23, 551–555 (2017).
90. Stanczak, M. A. et al. Self-​associated molecular
patterns mediate cancer immune evasion by engaging
Siglecs on T cells. J. Clin. Invest. https://doi.
org/10.1172/JCI120612 (2018).
91. Buchan, S. L., Rogel, A. & Al-​Shamkhani, A. The
immunobiology of CD27 and OX40 and their potential
as targets for cancer immunotherapy. Blood 131,
39–48 (2018).
92. Esensten, J. H., Helou, Y. A., Chopra, G., Weiss, A.
& Bluestone, J. A. CD28 costimulation: from mechanism
to therapy. Immunity 44, 973–988 (2016).
93. Hui, E. et al. T cell costimulatory receptor CD28 is a
primary target for PD-1-mediated inhibition. Science
355, 1428–1433 (2017).
94. Kamphorst, A. O. et al. Rescue of exhausted CD8
T cells by PD-1-targeted therapies is CD28-dependent.
Science 355, 1423–1427 (2017).
95. Sanmamed, M. F. et al. Agonists of co-​stimulation in
cancer immunotherapy directed against CD137,
OX40, GITR, CD27, CD28, and ICOS. Semin. Oncol.
42, 640–655 (2015).
96. Hunig, T. The storm has cleared: lessons from the
CD28 superagonist TGN1412 trial. Nat. Rev.
Immunol. 12, 317–318 (2012).
97. Elpek, K. et al. Efficacy of anti-​ICOS agonist
monoclonal antibodies in preclinical tumor models
provides a rationale for clinical development as cancer
immunotherapeutics. Cancer Immunol. Res. 4, A059
(2016).
98. Chaudhary, B. & Elkord, E. Regulatory T cells in the
tumor microenvironment and cancer progression: role
and therapeutic targeting. Vaccines 4, E28 (2016).
99. Routy, B. et al. Gut microbiome influences efficacy of
PD-1-based immunotherapy against epithelial tumors.
Science 359, 91–97 (2017).
100. Snyder, A., Pamer, E. & Wolchok, J. Could microbial
therapy boost cancer immunotherapy? Science 350,
1031–1032 (2015).
101. Gopalakrishnan, V. et al. Gut microbiome modulates
response to anti-​PD-1 immunotherapy in melanoma
patients. Science 359, 97–103 (2017).
102. Matson, V. et al. The commensal microbiome is
associated with anti-​PD-1 efficacy in metastatic
melanoma patients. Science 359, 104–108 (2018).
103. Nolz, J. C. Molecular mechanisms of CD8(+) T cell
trafficking and localization. Cell. Mol. Life Sci. 72,
2461–2473 (2015).
104. Nagarsheth, N., Wicha, M. S. & Zou, W. Chemokines
in the cancer microenvironment and their relevance in
cancer immunotherapy. Nat. Rev. Immunol. 17,
559–572 (2017).
105. Peng, D. et al. Epigenetic silencing of TH1-type
chemokines shapes tumour immunity and
immunotherapy. Nature 527, 249–253 (2015).
106. Nagarsheth, N. et al. PRC2 epigenetically silences
Th1-type chemokines to suppress effector T-​cell
trafficking in colon cancer. Cancer Res. 76, 275–282
(2016).
107. Huang, Y. et al. CD4+and CD8+T cells have opposing
roles in breast cancer progression and outcome.
Oncotarget 6, 17462–17478 (2015).
108. Sweis, R. F. et al. Molecular drivers of the non-​T
cell-​inflamed tumor microenvironment in urothelial
bladder cancer. Cancer Immunol. Res. 4, 563–568
(2016).
109. Tang, H. et al. Facilitating T cell infiltration in tumor
microenvironment overcomes resistance to PD-​L1
blockade. Cancer Cell 29, 285–296 (2016).
110. Venning, F. A., Wullkopf, L. & Erler, J. T. Targeting ECM
disrupts cancer progression. Front. Oncol. 5, 224
(2015).
111. Huang, Y., Goel, S., Duda, D. G., Fukumura, D. &
Jain, R. K. Vascular normalization as an emerging
strategy to enhance cancer immunotherapy.
Cancer Res. 73, 2943–2948 (2013).
112. Carmeliet, P. & Jain, R. K. Principles and mechanisms
of vessel normalization for cancer and other
angiogenic diseases. Nat. Rev. Drug Discov. 10,
417–427 (2011).
113. Lanitis, E., Irving, M. & Coukos, G. Targeting the tumor
vasculature to enhance T cell activity. Curr. Opin.
Immunol. 33, 55–63 (2015).
114. Tan, L. Y. et al. Control of immune cell entry through
the tumour vasculature: a missing link in optimising
melanoma immunotherapy? Clin. Transl Immunol. 6,
e134 (2017).
115. Wigerup, C., Pahlman, S. & Bexell, D. Therapeutic
targeting of hypoxia and hypoxia-​inducible factors in
cancer. Pharmacol. Ther. 164, 152–169 (2016).
Reviews
116. Serra, S. et al. Adenosine signaling mediates hypoxic
responses in the chronic lymphocytic leukemia
microenvironment. Blood Adv. 1, 47–61 (2016).
117. Stagg, J. & Smyth, M. J. Extracellular adenosine
triphosphate and adenosine in cancer. Oncogene 29,
5346–5358 (2010).
118. Antonioli, L., Blandizzi, C., Pacher, P. & Hasko, G.
Immunity, inflammation and cancer: a leading role for
adenosine. Nat. Rev. Cancer 13, 842–857 (2013).
119. Stagg, J. et al. CD73-deficient mice have increased
antitumor immunity and are resistant to experimental
metastasis. Cancer Res. 71, 2892–2900 (2011).
120. Antonioli, L., Yegutkin, G. G., Pacher, P., Blandizzi, C.
& Hasko, G. Anti-​CD73 in cancer immunotherapy:
awakening new opportunities. Trends Cancer 2,
95–109 (2016).
121. Hayes, G. M. et al. CD39 is a promising therapeutic
antibody target for the treatment of soft tissue
sarcoma. Am. J. Transl Res. 7, 1181–1188 (2015).
122. Sun, X. et al. Disordered purinergic signaling and
abnormal cellular metabolism are associated with
development of liver cancer in Cd39/ENTPD1 null
mice. Hepatology 57, 205–216 (2013).
123. Leone, R. D. & Emens, L. A. Targeting adenosine for
cancer immunotherapy. J. Immunother. Cancer 6, 57
(2018).
124. Yu, T., Tang, B. & Sun, X. Development of inhibitors
targeting hypoxia-​inducible factor 1 and 2 for cancer
therapy. Yonsei Med. J. 58, 489–496 (2017).
125. Leone, R. D., Lo, Y. C. & Powell, J. D. A2aR
antagonists: next generation checkpoint blockade for
cancer immunotherapy. Comput. Struct. Biotechnol. J.
13, 265–272 (2015).
126. Cardones, A. R. & Banez, L. L. VEGF inhibitors in
cancer therapy. Curr. Pharm. Des. 12, 387–394
(2006).
127. Fridman, W. H., Pages, F., Sautes-​Fridman, C. &
Galon, J. The immune contexture in human tumours:
impact on clinical outcome. Nat. Rev. Cancer 12,
298–306 (2012).
128. Rajabi, M. & Mousa, S. A. The role of angiogenesis in
cancer treatment. Biomedicines 5, 34 (2017).
129. Ebos, J. M. et al. Accelerated metastasis after
short-​term treatment with a potent inhibitor of tumor
angiogenesis. Cancer Cell 15, 232–239 (2009).
130. Paez-​Ribes, M. et al. Antiangiogenic therapy elicits
malignant progression of tumors to increased local
invasion and distant metastasis. Cancer Cell 15,
220–231 (2009).
131. Goel, S., Wong, A. H. & Jain, R. K. Vascular
normalization as a therapeutic strategy for malignant
and nonmalignant disease. Cold Spring Harb.
Perspect. Med. 2, a006486 (2012).
132. Tian, L. et al. Mutual regulation of tumour vessel
normalization and immunostimulatory
reprogramming. Nature 544, 250–254 (2017).
133. Sabat, R. et al. Biology of interleukin-10. Cytokine
Growth Factor Rev. 21, 331–344 (2010).
134. Yang, L. TGFbeta a potent regulator of tumor
microenvironment and host immune response,
implication for therapy. Curr. Mol. Med. 10, 374–380
(2010).
135. Zhang, H., Wang, Y., Hwang, E. S. & He, Y. W.
Interleukin-10: an immune-​activating cytokine in
cancer immunotherapy. J. Clin. Oncol. 34,
3576–3578 (2016).
136. Neuzillet, C. et al. Targeting the TGFbeta pathway for
cancer therapy. Pharmacol. Ther. 147, 22–31
(2015).
137. Lan, Y. et al. Enhanced preclinical antitumor activity of
M7824, a bifunctional fusion protein simultaneously
targeting PD-​L1 and TGF-​beta. Sci. Transl Med. 10,
eaan5488 (2018).
138. Shimabukuro-​Vornhagen, A. et al. The
immunosuppressive factors IL-10, TGF-​beta, and VEGF
do not affect the antigen-​presenting function of CD40-
activated B cells. J. Exp. Clin. Cancer Res. 31, 47
(2012).
139. Gabrilovich, D. Mechanisms and functional
significance of tumour-​induced dendritic-​cell defects.
Nat. Rev. Immunol. 4, 941–952 (2004).
140. Eil, R. et al. Ionic immune suppression within the
tumour microenvironment limits T cell effector
function. Nature 537, 539–543 (2016).
141. Richards, C. H., Mohammed, Z., Qayyum, T.,
Horgan, P. G. & McMillan, D. C. The prognostic value
of histological tumor necrosis in solid organ malignant
disease: a systematic review. Future Oncol. 7,
1223–1235 (2011).
142. Holmgaard, R. B. et al. Tumor-​expressed IDO recruits
and activates MDSCs in a Treg-​dependent manner.
Cell Rep. 13, 412–424 (2015).
143. Timosenko, E., Hadjinicolaou, A. V. & Cerundolo, V.
Modulation of cancer-​specific immune responses by
amino acid degrading enzymes. Immunotherapy 9,
83–97 (2017).
144. Kumar, V., Patel, S., Tcyganov, E. & Gabrilovich, D. I.
The nature of myeloid-​derived suppressor cells in the
tumor microenvironment. Trends Immunol. 37,
208–220 (2016).
145. Raber, P. L. et al. Subpopulations of myeloid-​derived
suppressor cells impair T cell responses through
independent nitric oxide-​related pathways. Int. J.
Cancer 134, 2853–2864 (2014).
146. Long, G. V. et al. Epacadostat (E) plus pembrolizumab
(P) versus pembrolizumab alone in patients (pts) with
unresectable or metastatic melanoma: results of the
phase 3 ECHO-301/KEYNOTE-252 study. J. Clin.
Oncol. 36, 108 (2018).
147. De Henau, O. et al. Overcoming resistance to
checkpoint blockade therapy by targeting PI3Kgamma
in myeloid cells. Nature 539, 443–447 (2016).
148. Cannarile, M. A. et al. Colony-​stimulating factor 1
receptor (CSF1R) inhibitors in cancer therapy.
J. Immunother. Cancer 5, 53 (2017).
149. Moore, E. et al. Established T cell-​inflamed tumors
rejected after adaptive resistance was reversed by
combination STING activation and PD-1 pathway
blockade. Cancer Immunol. Res. 4, 1061–1071
(2016).
150. Clavijo, P. E. et al. Resistance to CTLA-4 checkpoint
inhibition reversed through selective elimination of
granulocytic myeloid cells. Oncotarget 8,
55804–55820 (2017).
151. Harlin, H. et al. Chemokine expression in melanoma
metastases associated with CD8+T cell recruitment.
Cancer Res. 69, 3077–3085 (2009).
152. Ulloa-​Montoya, F. et al. Predictive gene signature in
MAGE-​A3 antigen-​specific cancer immunotherapy.
J. Clin. Oncol. 31, 2388–2395 (2013).
153. Sistigu, A. et al. Cancer cell-​autonomous contribution
of type I interferon signaling to the efficacy of
chemotherapy. Nat. Med. 20, 1301–1309 (2014).
154. Cai, X., Chiu, Y. H. & Chen, Z. J. The cGAS-​cGAMP-
STING pathway of cytosolic DNA sensing and
signaling. Mol. Cell 54, 289–296 (2014).
155. Corrales, L., McWhirter, S. M., Dubensky, T. W. Jr
& Gajewski, T. F. The host STING pathway at the
interface of cancer and immunity. J. Clin. Invest. 126,
2404–2411 (2016).
156. Sanchez-​Paulete, A. R. et al. Antigen cross-​
presentation and T cell cross-​priming in cancer
immunology and immunotherapy. Ann. Oncol. 28,
xii44–xii55 (2017).
157. Zitvogel, L., Galluzzi, L., Kepp, O., Smyth, M. J.
& Kroemer, G. Type I interferons in anticancer
immunity. Nat. Rev. Immunol. 15, 405–414 (2015).
158. Corrales, L. et al. Direct activation of STING in the
tumor microenvironment leads to potent and systemic
tumor regression and immunity. Cell Rep. 11,
1018–1030 (2015).
159. Huck, B. R., Kotzner, L. & Urbahns, K. Small molecules
drive big improvements in immuno-​oncology
therapies. Angew. Chem. Int. Ed. 57, 4412–4428
(2018).
160. Foote, J. B. et al. A STING agonist given with OX40
receptor and PD-​L1 modulators primes immunity and
reduces tumor growth in tolerized mice. Cancer
Immunol. Res. 5, 468–479 (2017).
161. Shekarian, T. et al. Pattern recognition receptors:
immune targets to enhance cancer immunotherapy.
Ann. Oncol. 28, 1756–1766 (2017).
162. Elion, D. L. & Cook, R. S. Harnessing RIG-​I and
intrinsic immunity in the tumor microenvironment for
therapeutic cancer treatment. Oncotarget 9,
29007–29017 (2018).
163. Le Mercier, I. et al. Tumor promotion by intratumoral
plasmacytoid dendritic cells is reversed by TLR7
ligand treatment. Cancer Res. 73, 4629–4640
(2013).
164. Kim, Y. H. et al. In situ vaccination against mycosis
fungoides by intratumoral injection of a TLR9 agonist
combined with radiation: a phase 1/2 study. Blood
119, 355–363 (2012).
165. Li, J. et al. Lymphoma immunotherapy with CpG
oligodeoxynucleotides requires TLR9 either in the host
or in the tumor itself. J. Immunol. 179, 2493–2500
(2007).
166. Wang, S. et al. Intratumoral injection of a CpG
oligonucleotide reverts resistance to PD-1 blockade by
expanding multifunctional CD8+ T cells. Proc. Natl
Acad. Sci. USA 113, E7240–E7249 (2016).
167. Sato-​Kaneko, F. et al. Combination immunotherapy
with TLR agonists and checkpoint inhibitors
suppresses head and neck cancer. JCI Insight 2,
93397 (2017).
168. Sagiv-​Barfi, I. et al. Eradication of spontaneous
malignancy by local immunotherapy. Sci. Transl Med.
10, eaan4488 (2018).
169. Marin-​Acevedo, J. A., Soyano, A. E., Dholaria, B.,
Knutson, K. L. & Lou, Y. Cancer immunotherapy
beyond immune checkpoint inhibitors. J. Hematol.
Oncol. 11, 8 (2018).
170. Wilkinson, R. W. & Leishman, A. J. Further advances
in cancer immunotherapy: going beyond checkpoint
blockade. Front. Immunol. 9, 1082 (2018).
171. Frank, M. J. et al. In situ vaccination with a TLR 9
agonist and local low dose radiation induces systemic
responses in untreated indolent lymphoma. Cancer
Discov. https://doi.org/10.1158/2159-8290.CD-18-
0743 (2018).
172. Chow, L. Q. M. et al. Phase Ib trial of the toll-​like
receptor 8 agonist, motolimod (VTX-2337), combined
with cetuximab in patients with recurrent or
metastatic SCCHN. Clin. Cancer Res. 23, 2442–2450
(2017).
173. Marabelle, A., Kohrt, H., Caux, C. & Levy, R.
Intratumoral immunization: a new paradigm for cancer
therapy. Clin. Cancer Res. 20, 1747–1756 (2014).
174. Ridnour, L. A. et al. Molecular pathways: toll-​like
receptors in the tumor microenvironment—poor
prognosis or new therapeutic opportunity. Clin. Cancer
Res. 19, 1340–1346 (2013).
175. Huang, L., Xu, H. & Peng, G. TLR-​mediated metabolic
reprogramming in the tumor microenvironment:
potential novel strategies for cancer immunotherapy.
Cell. Mol. Immunol. 15, 428–437 (2018).
176. Li, J. K., Balic, J. J., Yu, L. & Jenkins, B. TLR agonists
as adjuvants for cancer vaccines. Adv. Exp. Med. Biol.
1024, 195–212 (2017).
177. Iribarren, K. et al. Trial watch: immunostimulation with
toll-​like receptor agonists in cancer therapy.
Oncoimmunology 5, e1088631 (2016).
178. Vonderheide, R. H. & Glennie, M. J. Agonistic CD40
antibodies and cancer therapy. Clin. Cancer Res. 19,
1035–1043 (2013).
179. Tutt, A. L. et al. T cell immunity to lymphoma following
treatment with anti-​CD40 monoclonal antibody.
J. Immunol. 168, 2720–2728 (2002).
180. Mangsbo, S. M. et al. The human agonistic CD40
antibody ADC-1013 eradicates bladder tumors and
generates T cell-​dependent tumor immunity.
Clin. Cancer Res. 21, 1115–1126 (2015).
181. Beatty, G. L., Li, Y. & Long, K. B. Cancer
immunotherapy: activating innate and adaptive
immunity through CD40 agonists. Expert Rev.
Anticancer Ther. 17, 175–186 (2017).
182. Ohta, T. et al. Crucial roles of XCR1-expressing
dendritic cells and the XCR1-XCL1 chemokine axis in
intestinal immune homeostasis. Sci. Rep. 6, 23505
(2016).
183. Krummel, M. F., Bartumeus, F. & Gerard, A. T cell
migration, search strategies and mechanisms. Nat.
Rev. Immunol. 16, 193–201 (2016).
184. Bottcher, J. P. et al. NK cells stimulate recruitment of
cDC1 into the tumor microenvironment promoting
cancer immune control. Cell 172, 1022–1037 (2018).
185. Villadangos, J. A. & Shortman, K. Found in translation:
the human equivalent of mouse CD8+dendritic cells.
J. Exp. Med. 207, 1131–1134 (2010).
186. Woo, S. R. et al. STING-​dependent cytosolic DNA
sensing mediates innate immune recognition of
immunogenic tumors. Immunity 41, 830–842 (2014).
This article shows that innate immune sensing of
cancer occurs via the host STING pathway and
subsequent type I interferon production.
Spontaneous CD8+ T cell priming against tumours
depends on STING.
187. Deng, L. et al. STING-​dependent cytosolic DNA
sensing promotes radiation-​induced type I interferon-​
dependent antitumor immunity in immunogenic
tumors. Immunity 41, 843–852 (2014).
188. Harding, S. M. et al. Mitotic progression following
DNA damage enables pattern recognition within
micronuclei. Nature 548, 466–470 (2017).
189. Zheng, W. et al. Combination of radiotherapy and
vaccination overcomes checkpoint blockade
resistance. Oncotarget 7, 43039–43051 (2016).
190. Bonvalot, S. et al. First-​in-human study testing a new
radioenhancer using nanoparticles (NBTXR3)
activated by radiation therapy in patients with locally
advanced soft tissue sarcomas. Clin. Cancer Res. 23,
908–917 (2017).
191. Beatty, G. L. & Gladney, W. L. Immune escape
mechanisms as a guide for cancer immunotherapy.
Clin. Cancer Res. 21, 687–692 (2015).
www.nature.com/nrd

192. Zitvogel, L., Kepp, O. & Kroemer, G. Decoding cell
death signals in inflammation and immunity. Cell 140,
798–804 (2010).
193. McGranahan, N. et al. Clonal neoantigens elicit T cell
immunoreactivity and sensitivity to immune
checkpoint blockade. Science 351, 1463–1469
(2016).
This article demonstrates that a relationship exists
between clonal neoantigen burden and overall
survival in primary lung adenocarcinomas.
Sensitivity to PD-1 and CTLA4 blockade in patients
with advanced NSCLC and melanoma was enhanced
in tumours enriched for clonal neoantigens.
194. Krysko, D. V. et al. Immunogenic cell death and
DAMPs in cancer therapy. Nat. Rev. Cancer 12,
860–875 (2012).
195. Kroemer, G., Galluzzi, L., Kepp, O. & Zitvogel, L.
Immunogenic cell death in cancer therapy. Annu. Rev.
Immunol. 31, 51–72 (2013).
196. Galluzzi, L., Buque, A., Kepp, O., Zitvogel, L.
& Kroemer, G. Immunogenic cell death in cancer and
infectious disease. Nat. Rev. Immunol. 17, 97–111
(2017).
197. Obeid, M. et al. Calreticulin exposure dictates the
immunogenicity of cancer cell death. Nat. Med. 13,
54–61 (2007).
198. Vacchelli, E. et al. Chemotherapy-​induced antitumor
immunity requires formyl peptide receptor 1. Science
350, 972–978 (2015).
199. Ruffell, B. et al. Leukocyte composition of human
breast cancer. Proc. Natl Acad. Sci. USA 109,
2796–2801 (2012).
200. Park, J. H. et al. Clonal expansion of antitumor T cells
in breast cancer correlates with response to
neoadjuvant chemotherapy. Int. J. Oncol. 49,
471–478 (2016).
201. Ladoire, S. et al. Combined evaluation of LC3B puncta
and HMGB1 expression predicts residual risk of
relapse after adjuvant chemotherapy in breast cancer.
Autophagy 11, 1878–1890 (2015).
202. Ladoire, S. et al. The presence of LC3B puncta and
HMGB1 expression in malignant cells correlate with
the immune infiltrate in breast cancer. Autophagy 12,
864–875 (2016).
203. Fucikova, J. et al. Calreticulin expression in human
non-​small cell lung cancers correlates with increased
accumulation of antitumor immune cells and favorable
prognosis. Cancer Res. 76, 1746–1756 (2016).
204. Stoll, G. et al. Calreticulin expression: interaction with
the immune infiltrate and impact on survival in
patients with ovarian and non-​small cell lung cancer.
Oncoimmunology 5, e1177692 (2016).
205. Wemeau, M. et al. Calreticulin exposure on malignant
blasts predicts a cellular anticancer immune response
in patients with acute myeloid leukemia. Cell Death
Dis. 1, e104 (2010).
206. Fucikova, J. et al. Calreticulin exposure by malignant
blasts correlates with robust anticancer immunity and
improved clinical outcome in AML patients. Blood
128, 3113–3124 (2016).
207. Galluzzi, L., Buque, A., Kepp, O., Zitvogel, L.
& Kroemer, G. Immunological effects of conventional
chemotherapy and targeted anticancer agents. Cancer
Cell 28, 690–714 (2015).
208. Vincent, J. et al. 5-Fluorouracil selectively kills tumor-​
associated myeloid-​derived suppressor cells resulting
in enhanced T cell-​dependent antitumor immunity.
Cancer Res. 70, 3052–3061 (2010).
209. Ghiringhelli, F. et al. Metronomic cyclophosphamide
regimen selectively depletes CD4+CD25+ regulatory
T cells and restores T and NK effector functions in end
stage cancer patients. Cancer Immunol. Immunother.
56, 641–648 (2007).
210. Viaud, S. et al. The intestinal microbiota modulates
the anticancer immune effects of cyclophosphamide.
Science 342, 971–976 (2013).
211. Wang, W. et al. Effector T cells abrogate stroma-​
mediated chemoresistance in ovarian cancer. Cell 165,
1092–1105 (2016).
212. Anitei, M. G. et al. Prognostic and predictive values of
the immunoscore in patients with rectal cancer. Clin.
Cancer Res. 20, 1891–1899 (2014).
213. Gandhi, L. et al. Pembrolizumab plus chemotherapy in
metastatic non-​small-cell lung cancer. N. Engl. J. Med.
378, 2078–2092 (2018).
214. Brea, E. J. et al. Kinase regulation of human MHC
class I molecule expression on cancer cells. Cancer
Immunol. Res. 4, 936–947 (2016).
215. Gang, A. O. et al. 5-Azacytidine treatment sensitizes
tumor cells to T cell mediated cytotoxicity and
modulates NK cells in patients with myeloid
malignancies. Blood Cancer J. 4, e197 (2014).
Nature Reviews | Drug Discovery
216. de Charette, M., Marabelle, A. & Houot, R. Turning
tumour cells into antigen presenting cells: the next
step to improve cancer immunotherapy? Eur. J.
Cancer 68, 134–147 (2016).
217. Chiappinelli, K. B. et al. Inhibiting DNA methylation
causes an interferon response in cancer via dsRNA
including endogenous retroviruses. Cell 162,
974–986 (2015).
218. Stengel, S., Fiebig, U., Kurth, R. & Denner, J.
Regulation of human endogenous retrovirus-​K
expression in melanomas by CpG methylation. Genes
Chromosomes Cancer 49, 401–411 (2010).
219. Strissel, P. L. et al. Reactivation of codogenic
endogenous retroviral (ERV) envelope genes in human
endometrial carcinoma and prestages: emergence of
new molecular targets. Oncotarget 3, 1204–1219
(2012).
220. Sharma, S., Kaufmann, T. & Biswas, S. Impact of
inhibitor of apoptosis proteins on immune modulation
and inflammation. Immunol. Cell Biol. 95, 236–243
(2017).
221. Kotschy, A. et al. The MCL1 inhibitor S63845 is
tolerable and effective in diverse cancer models.
Nature 538, 477–482 (2016).
222. Yea, S. S. & Fruman, D. A. Achieving cancer cell death
with PI3K/mTOR-​targeted therapies. Ann. NY Acad.
Sci. 1280, 15–18 (2013).
223. Kumari, N., Dwarakanath, B. S., Das, A. & Bhatt, A. N.
Role of interleukin-6 in cancer progression and
therapeutic resistance. Tumour Biol. 37,
11553–11572 (2016).
224. Liepe, J. et al. A large fraction of HLA class I ligands
are proteasome-​generated spliced peptides. Science
354, 354–358 (2016).
225. Lauss, M. et al. Mutational and putative neoantigen
load predict clinical benefit of adoptive T cell
therapy in melanoma. Nat. Commun. 8, 1738
(2017).
226. Germano, G. et al. Inactivation of DNA repair triggers
neoantigen generation and impairs tumour growth.
Nature 552, 116–120 (2017).
This article shows that inactivation of DNA MMR
mechanisms increased mutational load, promoted
continuous renewal of neoantigens in human CRCs
and triggered immune surveillance in mouse
models.
227. Wieringa, H. W., van der Zee, A. G., de Vries, E. G. &
van Vugt, M. A. Breaking the DNA damage response
to improve cervical cancer treatment. Cancer Treat.
Rev. 42, 30–40 (2016).
228. Hemann, M. T. From breaking bad to worse: exploiting
homologous DNA repair deficiency in cancer. Cancer
Discov. 4, 516–518 (2014).
229. Yang, Y. Cancer immunotherapy: harnessing the
immune system to battle cancer. J. Clin. Invest. 125,
3335–3337 (2015).
230. Rosenbaum, L. Tragedy, perseverance, and chance —
the story of CAR-​T therapy. N. Engl. J. Med. 377,
1313–1315 (2017).
231. Jensen, M. C. & Riddell, S. R. Designing chimeric
antigen receptors to effectively and safely target
tumors. Curr. Opin. Immunol. 33, 9–15 (2015).
232. Gomes-​Silva, D. & Ramos, C. A. Cancer
immunotherapy using CAR-​T cells: from the research
bench to the assembly line. Biotechnol. J. 13,
1700097 (2018).
233. Wang, X. & Riviere, I. Clinical manufacturing of CAR
T cells: foundation of a promising therapy. Mol. Ther.
Oncolyt. 3, 16015 (2016).
234. von Kalle, C., Deichmann, A. & Schmidt, M. Vector
integration and tumorigenesis. Hum. Gene Ther. 25,
475–481 (2014).
235. Wright, A. V., Nunez, J. K. & Doudna, J. A. Biology and
applications of CRISPR systems: harnessing nature’s
toolbox for genome engineering. Cell 164, 29–44
(2016).
236. Eyquem, J. et al. Targeting a CAR to the TRAC locus
with CRISPR/Cas9 enhances tumour rejection. Nature
543, 113–117 (2017).
237. Hervas-​Stubbs, S. et al. CD8 T cell priming in the
presence of IFN-​alpha renders CTLs with improved
responsiveness to homeostatic cytokines and recall
antigens: important traits for adoptive T cell therapy.
J. Immunol. 189, 3299–3310 (2012).
238. Mishto, M. & Liepe, J. Post-​translational peptide
splicing and T cell responses. Trends Immunol. 38,
904–915 (2017).
239. Verdegaal, E. M. et al. Neoantigen landscape
dynamics during human melanoma-​T cell interactions.
Nature 536, 91–95 (2016).
240. Spranger, S., Dai, D., Horton, B. & Gajewski, T. F.
Tumor-​residing Batf3 dendritic cells are required for
Reviews
effector T cell trafficking and adoptive T cell therapy.
Cancer Cell 31, 711–723 (2017).
241. Galon, J. et al. Characterization of anti-​CD19 chimeric
antigen receptor (CAR) T cell-​mediated tumor
microenvironment immune gene profile in a
multicenter trial (ZUMA-1) with axicabtagene
ciloleucel (axi-​cel, KTE-​C19). J. Clin. Oncol. 35, 3025
(2017).
242. Kaufman, H. L., Kohlhapp, F. J. & Zloza, A. Oncolytic
viruses: a new class of immunotherapy drugs. Nat.
Rev. Drug Discov. 14, 642–662 (2015).
243. Pataer, A., Swisher, S. G., Roth, J. A., Logothetis, C. J.
& Corn, P. G. Inhibition of RNA-​dependent protein
kinase (PKR) leads to cancer cell death and increases
chemosensitivity. Cancer Biol. Ther. 8, 245–252
(2009).
244. Chesney, J. et al. Randomized, open-​label phase II
study evaluating the efficacy and safety of talimogene
laherparepvec in combination with ipilimumab
versus ipilimumab alone in patients with advanced,
unresectable melanoma. J. Clin. Oncol. 36, 1658–1667
(2018).
This paper presents the first randomized trial
evaluating the combination of an oncolytic virus
with a checkpoint inhibitor and showing a
significantly higher objective response rate for
T-​VEC plus ipilimumab versus ipilimumab alone.
245. Ribas, A. et al. Oncolytic virotherapy promotes
intratumoral T cell infiltration and improves anti-​PD-1
immunotherapy. Cell 170, 1109–1119 (2017).
This article shows that oncolytic virotherapy with
T-​VEC in patients with advanced melanoma
increased the cytotoxic T cell infiltration and
therapeutic efficacy of an anti-​PD-1 antibody.
246. Haanen, J. Converting cold into hot tumors by
combining immunotherapies. Cell 170, 1055–1056
(2017).
247. Nestvold, J. et al. Oncolytic peptide LTX-315 induces
an immune-​mediated abscopal effect in a rat
sarcoma model. Oncoimmunology 6, e1338236
(2017).
248. Patel, A., Kaufman, H. L. & Disis, M. L. Next
generation approaches for tumor vaccination. Chin.
Clin. Oncol. 6, 19 (2017).
249. Gibney, G. T., Weiner, L. M. & Atkins, M. B. Predictive
biomarkers for checkpoint inhibitor-​based
immunotherapy. Lancet Oncol. 17, e542–e551
(2016).
250. Vigneron, N. Human tumor antigens and cancer
immunotherapy. Biomed. Res. Int. 2015, 948501
(2015).
251. Mitchell, D. A. et al. Tetanus toxoid and CCL3 improve
dendritic cell vaccines in mice and glioblastoma
patients. Nature 519, 366–369 (2015).
252. Stupp, R. et al. Radiotherapy plus concomitant and
adjuvant temozolomide for glioblastoma. N. Engl.
J. Med. 352, 987–996 (2005).
253. Liau, L. M. et al. First results on survival from a large
phase 3 clinical trial of an autologous dendritic cell
vaccine in newly diagnosed glioblastoma. J. Transl
Med. 16, 142 (2018).
254. Luksza, M. et al. A neoantigen fitness model predicts
tumour response to checkpoint blockade
immunotherapy. Nature 551, 517–520 (2017).
255. Sahin, U. et al. Personalized RNA mutanome vaccines
mobilize poly-​specific therapeutic immunity against
cancer. Nature 547, 222–226 (2017).
This paper presents the development and
successful application of a personalized vaccine-​
based immunotherapy exploiting the concept of an
individualized mutanome and computational
prediction of neo-​epitopes.
256. Ott, P. A. et al. An immunogenic personal neoantigen
vaccine for patients with melanoma. Nature 547,
217–221 (2017).
This article provides a strong rationale for further
development of neo-​epitope vaccines, alone and in
combination with checkpoint blockade or other
immunotherapies.
257. Garcia-​Martinez, E. et al. Trial watch:
immunostimulation with recombinant cytokines for
cancer therapy. Oncoimmunology 7, e1433982
(2018).
258. Jiang, T., Zhou, C. & Ren, S. Role of IL-2 in cancer
immunotherapy. Oncoimmunology 5, e1163462
(2016).
259. Sockolosky, J. T. et al. Selective targeting of
engineered T cells using orthogonal IL-2 cytokine-​
receptor complexes. Science 359, 1037–1042
(2018).
260. Ochoa, M. C. et al. Interleukin-15 in gene therapy of
cancer. Curr. Gene Ther. 13, 15–30 (2013).
Reviews
261. Romee, R. et al. First-​in-human phase 1 clinical study
of the IL-15 superagonist complex ALT-803 to treat
relapse after transplantation. Blood 131, 2515–2527
(2018).
262. Mazzucchelli, R. & Durum, S. K. Interleukin-7 receptor
expression: intelligent design. Nat. Rev. Immunol. 7,
144–154 (2007).
263. Deiser, K., Stoycheva, D., Bank, U., Blankenstein, T.
& Schuler, T. Interleukin-7 modulates anti-​tumor
CD8+T cell responses via its action on host cells. PLOS
ONE 11, e0159690 (2016).
264. Xu, X., Sun, Q., Mei, Y., Liu, Y. & Zhao, L. Newcastle
disease virus co-​expressing interleukin 7 and
interleukin 15 modified tumor cells as a vaccine for
cancer immunotherapy. Cancer Sci. 109, 279–288
(2018).
265. Al-​Chami, E., Tormo, A., Khodayarian, F. & Rafei, M.
Therapeutic utility of the newly discovered properties
of interleukin-21. Cytokine 82, 33–37 (2016).
266. Kannappan, V. et al. Interleukin 21 inhibits cancer-​
mediated FOXP3 induction in naive human CD4
T cells. Cancer Immunol. Immunother. 66, 637–645
(2017).
267. Spolski, R. & Leonard, W. J. IL-21 and T follicular
helper cells. Int. Immunol. 22, 7–12 (2010).
268. Lewis, K. E. et al. Interleukin-21 combined with PD-1
or CTLA-4 blockade enhances antitumor immunity in
mouse tumor models. Oncoimmunology 7, e1377873
(2017).
269. Chapuis, A. G. et al. Combined IL-21-primed polyclonal
CTL plus CTLA4 blockade controls refractory
metastatic melanoma in a patient. J. Exp. Med. 213,
1133–1139 (2016).
270. Kaiser, J. ‘Liquid biopsy’ for cancer promises early
detection. Science 359, 259 (2018).
271. Tavare, R. et al. An effective immuno-​PET imaging
method to monitor CD8-dependent responses to
immunotherapy. Cancer Res. 76, 73–82 (2016).
272. Ali, H. R., Chlon, L., Pharoah, P. D., Markowetz, F.
& Caldas, C. Patterns of immune infiltration in breast
cancer and their clinical implications: a gene-​
expression-based retrospective study. PLOS Med. 13,
e1002194 (2016).
273. Newman, A. M. et al. Robust enumeration of cell
subsets from tissue expression profiles. Nat. Methods
12, 453–457 (2015).
274. Rooney, M. S., Shukla, S. A., Wu, C. J., Getz, G.
& Hacohen, N. Molecular and genetic properties of
tumors associated with local immune cytolytic activity.
Cell 160, 48–61 (2015).
275. Gentles, A. J. et al. The prognostic landscape of genes
and infiltrating immune cells across human cancers.
Nat. Med. 21, 938–945 (2015).
276. Aran, D., Hu, Z. & Butte, A. J. xCell: digitally
portraying the tissue cellular heterogeneity landscape.
Genome Biol. 18, 220 (2017).
277. Li, T. et al. TIMER: a web server for comprehensive
analysis of tumor-​infiltrating immune cells. Cancer
Res. 77, e108–e110 (2017).
278. Thorsson, V. et al. The immune landscape of cancer.
Immunity 48, 812–830 (2018).
279. Zappia, L., Phipson, B. & Oshlack, A. Exploring the
single-​cell RNA-​seq analysis landscape with the
scRNA-​tools database. PLOS Comput. Biol. 14,
e1006245 (2018).
280. Chen, X., Sun, Y. C., Church, G. M., Lee, J. H. &
Zador, A. M. Efficient in situ barcode sequencing using
padlock probe-​based BaristaSeq. Nucleic Acids Res.
46, e22 (2018).
281. Herbst, R. S. et al. Predictive correlates of response to
the anti-​PD-L1 antibody MPDL3280A in cancer
patients. Nature 515, 563–567 (2014).
282. Topalian, S. L., Drake, C. G. & Pardoll, D. M. Targeting
the PD-1/B7-H1(PD-​L1) pathway to activate anti-​tumor
immunity. Curr. Opin. Immunol. 24, 207–212 (2012).
283. Ayers, M. et al. IFN-​gamma-related mRNA profile
predicts clinical response to PD-1 blockade. J. Clin.
Invest. 127, 2930–2940 (2017).
284. Davoli, T., Uno, H., Wooten, E. C. & Elledge, S. J.
Tumor aneuploidy correlates with markers of immune
evasion and with reduced response to immunotherapy.
Science 355, eaaf8399 (2017).
285. Papaccio, F. et al. Concise review: cancer cells, cancer
stem cells, and mesenchymal stem cells: influence in
cancer development. Stem Cells Transl Med. 6,
2115–2125 (2017).
286. Jolly, M. K., Ware, K. E., Gilja, S., Somarelli, J. A.
& Levine, H. EMT and MET: necessary or permissive
for metastasis? Mol. Oncol. 11, 755–769
(2017).
287. Moustakas, A. & de Herreros, A. G. Epithelial-​
mesenchymal transition in cancer. Mol. Oncol. 11,
715–717 (2017).
288. Terry, S. et al. New insights into the role of EMT in
tumor immune escape. Mol. Oncol. 11, 824–846
(2017).
Acknowledgements
The authors thank the following institutions for their financial
support: the National Cancer Institute of France (INCa), the
Plan Cancer, the Canceropole Ile de France, INSERM, Cancer
Research for Personalized Medicine (CARPEM), the Paris
Alliance of Cancer Research Institutes (PACRI), H2020 PHC-
32-2014 APERIM grant number EEAA15006DDA,
MedImmune (grant number RVE15004DSA) and LabEx
Immuno-​oncology.
Author contributions
The authors contributed equally to all aspects of the article.
Competing interests
Immunoscore is a registered trademark from INSERM. J.G.
is co-​founder and chairman of the scientific advisory board
of HalioDx. J.G. has patents associated with an ‘in vitro
method for the prognosis of progression of a cancer’ (PCT/
IB2006/003168 and PCT/EP2013/062405). J.G. estab-
lished Collaborative Research Agreement (grants) with
Perkin-​E lmer, IO Biotech, MedImmune, Astra Zeneca,
Janssen, Imcheck Therapeutics. J.G. participated to
Scientific Advisory Boards of BMS, MedImmune, Astra
Zeneca, Novartis, Definiens, Merck Serono, IO Biotech,
ImmunID, Nanostring, Illumina, Northwest Biotherapeutics,
Actelion, Amgen, Kite Pharma and Merck MSD. J.G. was a
consultant for BMS, Roche, GSK, Compugen, Mologen
and Sanofi.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
www.nature.com/nrd