RESEARCH ARTICLE SUMMARY
◥CANCER
Stepwise-edited, human melanoma models reveal
mutationsÕeffect on tumor and microenvironment
Eran Hodis, Elena Torlai Triglia, John Y. H. Kwon, Tommaso Biancalani, Labib R. Zakka,
Saurabh Parkar, Jan-Christian Hütter, Lorenzo Buffoni, Toni M. Delorey, Devan Phillips,
Danielle Dionne, Lan T. Nguyen, Denis Schapiro, Zoltan Maliga, Connor A. Jacobson,
Ayal Hendel, Orit Rozenblatt-Rosen, Martin C. Mihm Jr., Levi A. Garraway, Aviv Regev
INTRODUCTION:Cancer develops from cells that
become malignant because of mutations in
multiple genes—often accumulated over long
time periods—that produce a phenotypic diver-
sity across patient tumors. Specific genetic al-
terations in particular cancer types have been
linked to prognosis, to response or resistance
to therapies (especially targeted therapeutics),
and to a tumor’s propensity to acquire further
mutations, among other phenotypes. However,
genotype-phenotype connections are challeng-
ing to infer in patients, as any two tumors differ
genetically in too many ways to isolate the
effect of one or several mutations. The ability
to systematically connect cancer-associated
mutations or combinations thereof with their
phenotypic consequences would advance our
understanding of the mechanisms of cancer
pathogenesis and genetically linked disease
features.RATIONALE:We reasoned that genome editing
and the fitness advantage of cancer-associated
mutations could be leveraged to generate hu-
man cellular models of tumor development.
Such genome-edited models would replicate
the precise genetics, lineage relationships, and
stepwise progression of cancer and allow us
to establish genotype-to-phenotype links in a
controlled experimental design. While similar
models have been realized for tumors origi-
nating from self-renewing stem cells as cells of
origin, specifically in colorectal cancer, no com-
parable models exist for tumor types that arisefrom nonstem differentiated cells. We present
an approach that starts from the nonstem cell
of origin of melanoma, the healthy human
melanocyte; this approach then generates a
series of cells with precise genome editing of
mutations in key cancer genes, thus expand-
ing the horizon of possible cellular models of
cancer development.RESULTS:We generated a progressive series
of genome-edited human models of melanoma.
We started from healthy human melanocytes
and introduced, in a stepwise fashion, muta-
tions in up to five genes spanning six pathways
commonly dysregulated in melanoma:CDKN2A
(part of the RB pathway),BRAF(MAPK),TERT
(telomerase),PTEN(PI3K/AKT),TP53(p53),
andAPC(Wnt), for a total of nine genetically
distinct cellular models. We characterized these
models during growth in vitro and after intra-
dermal injection through mouse xenografts,
using physiological assessment, histopathol-
ogy, and single-cell RNA sequencing (scRNA-
Seq), leveraging computational methods and
machine learning algorithms. Through these
models, we connected melanocyte genotypes
to phenotypes such as gene expression programs,
replicative immortality, malignancy, rapid tu-
mor growth, tumor pigmentation, metastasis,
and histopathological features. In vitro, con-
secutive mutations produced an ordered pro-
gression through expression space. In vivo,
mutations in malignant cells also affected the
cell-type composition and expression states of
tumor-infiltrating microenvironment cells. Our
melanoma models shared genotype-associated
expression programs with patient melanomas
and partially recapitulated patient melanoma
genotype-associated histopathological features.CONCLUSION:The genotype-phenotype con-
nections we identify highlight how the impact
of mutations often depends on genetic context.
Such genetic epistasis makes understanding
the phenotypic consequences of the mutational
landscape of human cancers a combinatorial
problem whose study requires modeling strat-
egies that can scale to multiple mutations, such
as the one presented in this study. Genome-
edited human models of cancer enable the
identification of causal relationships be-
tween defined sets of genetic alterations
and disease-relevant phenotypes, further-
ing understanding of how cancer mutations
help give rise to the diverse and varied pheno-
types of human malignancy.▪RESEARCH
474 29 APRIL 2022•VOL 376 ISSUE 6592 science.orgSCIENCE
The list of author affiliations is available in the full article online.
*Corresponding author. Email: [email protected]
(E.H.); [email protected] (A.R.)
Cite this article as E. Hodiset al.,Science 376 , eabi8175
(2022). DOI: 10.1126/science.abi8175READ THE FULL ARTICLE AT
https://doi.org/10.1126/science.abi8175CBTP
CBT
CBCBT3CCBTPA
WT CBTP3Cancer genotype engineering
Genotype to phenotype characterization
CDKN2A-/- CDKN2A-/-
BRAFV600ECDKN2A-/-
BRAFV600E
TERT -124C>T+ TP53-/-+ PTEN-/-+ APC-/-+ TP53-/-+ APC-/-- RB (CDKN2A)
- MAPK (BRAF)
- Telomerase (TERT)
- PI3K/Akt (PTEN)
- p53 (TP53)
- Wnt (APC)
Melanoma pathways
(edited gene)Building human melanoma models from melanocytes
Genetically modified melanocytesHealthy
melanocytesMicroenvironment changes Similarity to patients’ programs Histopathological appearance
C CB CBT
CBTP3CBTPACBT3CBTPC BTAMalignancy and
immortality
Genotypes with
C, B and T mutations CBTPA
C BTAMITF-Low OxPhos CBTP3 CBTPAWnt-driven metastasisC BTA CBTPNeutrophilsCBT3 CBTPCBTP3Ordered expression continuum Genetic epistasisCBTAGenome-engineered human cell models connect melanoma genotypes to phenotypes.Sequential,
precise gene editing of human melanocytes produced a series of melanoma models. The editing approach we
developed—without selection markers or single-cell cloning—is applicable to other cell types, expanding
the toolkit of available cancer models. Through in-depth phenotypic characterization of in vitro cells and
in vivo tumors from xenografts in immunodeficient mice, multimutant melanoma genotypes were causally
linked to specific phenotypes.