NEWS & ANALYSIS

Nature Reviews Drug Discovery | Published online 9 Mar 2018; doi:10.1038/nrd.2018.6

Tissue-agnostic cancer drug pipeline

grows, despite doubts
The FDA could soon approve the first drug developed explicitly for a tissue-agnostic cancer indication,
but biological complexity may limit this approach.

Ken Garber
Loxo Oncology’s recent FDA filing
for its TRK inhibitor larotrectinib
in any advanced solid tumour with
an NTRK fusion is another step
forward for the new, purely molecular
approach to cancer therapy. The
supplemental approval last May
for Merck & Co.’s PD1 blocker
pembrolizumab for all microsatellite
instability-high (MSI-H) tumours
was the first tissue-agnostic approval,
but larotrectinib would be the first
drug developed with a tissue-agnostic
indication in mind.
Other tissue-agnostic contenders
are in the pipeline (TABLE 1). And
big pharma is now investing big
sums to get into the game. Roche
acquired Ignyta for US$1.7 billion in
December for access to a promising
multikinase inhibitor, entrectinib.
In November, Bayer partnered with
Loxo on larotrectinib, committing up
to $1.55 billion in upfront payments
and milestones to collaborate on
tissue-agnostic drug programmes.
Other pharma companies, including
Roche’s Genentech and Eli Lilly, are
exploring tissue-agnostic indications
within their oncology pipelines.
“This is something that has a chance
to happen again and again,” says Levi
Garraway, senior vice president of
Oncology Global Development and
Medical Affairs at Lilly.
But experts sharply disagree on
how far this approach can be taken.
The pembrolizumab approval “really
opens the door for other drugs,” says
oncologist Razelle Kurzrock, director
of the Center for Personalized
Cancer Therapy at the University
of California, San Diego. A driver
gene alteration is fundamental to
the malignant phenotype, she says,
whereas the organ of origin is much
less important. But cancer geneticist
Barb Weber, CEO of oncology
start-up Tango Therapeutics and
former global head of Oncology
Translational Medicine at Novartis,
sees many obstacles ahead.
Tissue-agnostic approvals “are going
to be the exception rather than the
rule,” she says.
Tissue-agnostic logic
Molecular oncologists have envisioned
a molecular tumour classification
system since the oncogene hypothesis
of cancer gained acceptance in the
early 1980s. Universal, molecularly
targeted treatments, they expected,
would follow. But this hope faded
as it gradually became clear that
differences between and even
within tumour types — based on
developmental lineage and not on
cancer genetics — can shape the
malignant transformations that
occur, and the resulting sensitivity
to targeted drugs.
With the advent of
immunotherapies, and the discovery
of tumour neoantigens that are
displayed prolifically by cancer
[Tissue-agnostic
approvals] are going to
be the exception rather
than the rule
cells, tissue-agnostic cancer drug
development has finally become
a reality. In 2012 a group at Johns
Hopkins University hypothesized
that because tumours with DNA
mismatch repair defects produce large
numbers of neoantigens that can elicit
immune responses, these cancers
should be particularly sensitive
to immune checkpoint inhibitors
that unleash the power of T cells.
A trial of pembrolizumab that used
MSI-H as a hypermutation marker
yielded spectacular outcomes, and
the FDA’s supplemental approval
of pembrolizumab in this setting
followed. Kinase inhibitors that target
rare gene fusions, biotech companies
have recently shown, can also work
uniformly across tissue types, further
validating the tissue-agnostic strategy.
And yet, skeptics of tissue-agnostic
cancer drugs invariably bring up the
case of mutant BRAF kinase inhibitors
as an example of the limitation of this
approach. “It’s everyone’s negative
story,” says Kurzrock. Activating point
mutations in the BRAF oncogene are
present in roughly 50% of melanomas
and 10% of colorectal cancers.
But whereas melanomas respond
dramatically to BRAF inhibitors,
BRAF-mutated colorectal cancers
do not. Subsequent work showed
that in colorectal cancer BRAF
inhibition triggers activation of EGFR
signalling, driving continued cancer
cell proliferation. Melanomas express
little EGFR, so this feedback loop is
not active and the drug works in this
disease. This fundamental difference
in biology and drug response argues
against the agnostic approach for most
tumours, say skeptics.

NATURE REVIEWS | DRUG DISCOVERY ADVANCE ONLINE PUBLICATION | 1

©
2
0
1
8
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
N E W S & A N A LY S I S

Table 1 | Selected tissue-agnostic drugs in clinical development

Agent Company Target Indication Status
Pembrolizumab Merck & Co. PD1 MSI-H (MMR-deficient) solid tumours Approved
Larotrectinib Loxo Oncology, Bayer TRK Solid tumours with NTRK fusions NDA
Entrectinib Ignyta, Roche TRK, ALK, ROS1 Solid tumours with NTRK fusions Phase II
Merestinib Eli Lilly MET, TRK Solid tumours with NTRK rearrangements Phase II
Atezolizumab Genentech/Roche PDL1 Solid tumours with MSI-H, high mutation burden or alterations in Phase II
DNA proofreading genes

TPX-0005 TP Therapeutics TRK, ALK, ROS1 Solid tumours with NTRK, ALK and ROS1 rearrangements Phase I/II
LOXO-195 Loxo Oncology TRK Solid tumours with NTRK fusions, including those resistant to Phase I/II
larotrectinib
LOXO-292 Loxo Oncology RET Solid tumours with RET rearrangements Phase I
RXDX-105 Ignyta, Roche RET Solid tumours with RET fusions Phase I
LY3300054 Eli Lilly PDL1 Monotherapy in MSI-H solid tumours; various combination criteria Phase I
PLX8394 Plexxikon/Daiichi Mutant BRAF and Solid tumours with BRAF mutation Phase I/IIa
Sankyo wild-type CRAF
PLX9486 Plexxikon KIT Solid tumours with KIT mutations Phase I/II
Tissue-agnostic indications contingent on trial data. MMR, mismatch repair; MSI-H, microsatellite instability-high; NDA, new drug application; PD1, programmed
cell death protein 1; PDL1, PD1 ligand 1.

But recent data show that BRAF inhibitors
do work in colorectal cancer when combined
with an EGFR inhibitor, points out Kurzrock.
Smart use of combinations may therefore be
able to compensate for biological differences.
“I don’t actually think colorectal cancer is the
exception,” says Kurzrock. “It’s further along
the [agnostic] spectrum where a single agent
isn’t enough.”
“There are many more examples where
oncogenes do respond to therapy across
multiple different tumour types than there
are where they don’t,” agrees University of
Colorado oncologist Bob Doebele. “If [you
are blocking] the dominant driver that’s
preventing apoptosis and driving proliferation,
cancer cells are going to respond to therapy.”
But genetic complexity makes the precise
outcome unpredictable. Weber points out
that tumour types differ in the ways they
adapt to genetic instability, and this affects
drug sensitivity. And they retain the features
of their particular developmental lineage, the
growth pattern adopted during embryonic
development upon exposure to specific
transcription factor combinations. Different
lineages can confer different signalling
dependencies, Weber notes, independent of
cancer genetics. When she was at Novartis, the
company tested novel drug combinations in
tumour type-specific patient cohorts.
“I don’t recall a single one of those, and
we did a bunch — EGFR inhibitors, PI3K
inhibitors and so on — where the responses
were the same across histologies,” Weber says.
“That’s what keeps you from doing those
histology-agnostic registration trials.”
Given these biological uncertainties,
it fell to academics and small biotechs to
define and validate the first tissue-agnostic
indications. Researchers at Johns Hopkins
University thought up and secured
non-industry funding for the clinical trial
that led to Merck’s supplemental approval of
pembrolizumab for MSI-H tumours. Loxo
and Ignyta, meanwhile, have trailblazed
the development of TRK inhibitors for first
approvals in a tissue-agnostic indication.
Drug companies developed TRK
inhibitors beginning in the 1990s, but
mostly for pain indications because the
three members of the TRK family are the
receptors for nerve growth factor and other
neurotrophins. (Larotrectinib was originally
developed as an arthritis drug.) Research
beginning in the 1980s showed that fusions
involving the three encoding genes, NTRK1,
NRTK2 or NTRK3, can result in persistent
oncogenic signalling, but cancer drug
development languished until ALK and ROS1
inhibitors proved effective and profitable in
the late 2000s for treating rare lung cancer
fusions. Although NTRK fusions are probably
present in under 1% of all solid tumours, a
business model now exists. And in clinical
trials, the TRK inhibitors larotrectinib and
we’re already on an arc of
progress
entrectinib efficiently shrank tumours with
NTRK fusions, regardless of tumour type.
This may be because, apart from the fusion,
“the rest of the tumour is relatively genetically
bland,” suspects Weber.
The same may be true for tumours
harbouring fusions of the kinases ALK, ROS1
or RET, which also seem to drive tumour
growth regardless of histology. All are rare,
however — an analysis of 7,000 tumours
found that only 3% harboured a probably
oncogenic, recurrent kinase fusion. “We’ve
probably found a lot of these simple ones,”
says Doebele, whose lab first identified NTRK
fusions in lung cancer, jump-starting TRK
inhibitor drug development. “You’re not likely
to find tonnes more of these that cross many
different tumour types.”
Immunotherapy may therefore be
the most fruitful direction for future
tissue-agnostic development. “The immune
system doesn’t really care what the histology
is, or what the tissue is; the immune system
only cares about what the target is, and
what the biology is,” says Johns Hopkins
immunologist Jonathan Schneck. To this end,
Genentech and Lilly are both testing drugs
that act on the PD1–PDL1 axis for possible
tissue-agnostic indications beyond the MSI-H
setting. “We are particularly interested in
some of the combinations, [taking] at least a
partially lineage-agnostic development path,”
says Garraway.
New definitions of disease needed
Two big future challenges for everyone
working in this space will be finding the

2 | ADVANCE ONLINE PUBLICATION www.nature.com/nrd

©
2
0
1
8
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 N E W S & A N A LY S I S

rare patients who qualify for treatment,
and developing biomarkers that pass FDA
muster.
In the case of MSI-H and NTRK fusions,
biomarker development was a relatively low
hurdle. MSI-H assays have been in wide use
for almost two decades to screen colorectal
cancer patients for Lynch syndrome. And
Foundation Medicine’s recently approved
Foundation One CDx comprehensive
gene-profiling test includes an NTRK fusion
assay. Similar tests are in development.
The development of other biomarkers may
prove more complicated. Several companies
are looking at high tumour mutational burden
(TMB) as a marker of checkpoint inhibitor
sensitivity. One large-scale study showed
almost 10% of all tumours have a high TMB,
more than double the estimated prevalence of
the MSI-H subset. But there are many possible
ways to measure TMB, and results can depend
on the size of the genomic region analysed, the
depth of coverage and the type of sequencing
that is used. TMB scores can also vary with
the purity of the tumour sample, the presence
of artefact mutations in formalin-fixed
paraffin-embedded samples and even the way
mutations are defined. (Genentech is using the
Foundation One CDx diagnostic in one trial
of its PDL1 inhibitor atezolizumab to qualify
patients with high TMB, querying a panel of
315 cancer-related genes to assess mutational
load.)
Given all these variables, and a new
oncology reality in which a biomarker can
define the disease, the FDA recently called
for the community to work together on the
development of next-generation biomarkers.
“The establishment of some new disease
indication-defining biomarkers … may
require a more collaborative approach
than conventional drug development. If a
biomarker will, in essence, define the disease
indication, then it should be developed
through the collaboration of multiple
stakeholders including commercial sponsors,
device manufacturers, academia and patients,”
they wrote in the New England Journal of
Medicine.
Other tissue-agnostic biomarkers are also
in the works. Whereas PDL1 overexpression
is only roughly predictive for response to
checkpoint inhibitors, PDL1 amplifications
that increase copy number may prove highly
specific. “The vast majority of our patients
[with PDL1 amplifications] we’ve treated
have responded, regardless of their histology,”
says Kurzrock. PDL1 amplification is a
very rare event, she adds, similar to NTRK
fusions.
Kurzrock’s group is also looking at
alterations in APOBEC, an mRNA-editing
enzyme that is upregulated in response to
viral infections, as a possible tissue-agnostic
predictive biomarker for checkpoint
blockade. APOBEC activity creates a
localized hypermutation signature called
kataegis, and these mutation clusters
are found in many tumours and across
histologies. Mutations in the DNA
proofreading gene POLE, which lead to
mutations across the genome, may also
be predictive.
All these potential tissue-agnostic
biomarkers remain to be defined and
validated. But the detection technologies
exist at least, says Garraway. “It’s really
just a matter of whether or not they can
be, for lack of a better word, ‘platformized’
effectively,” he says. Ultimately, he predicts,
a single combination test will emerge.
“It’s early days, because we need prospective
studies,” he says. “But we’re already on an
arc of progress.”

NATURE REVIEWS | DRUG DISCOVERY ADVANCE ONLINE PUBLICATION | 3

©
2
0
1
8
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.