Science - 27.03.2020

GRAPHIC: KELLIE HOLOSKI/

SCIENCE

BASED ON SOTTORIVA

ET AL

., BIOCHIM. BIOPHYS. ACTA

1867

, 95 (2017) (CC BY-NC)

continuing to accrue additional mutations

that result in diversification, but the dynam-

ics of this process are poorly understood.

Using mathematical modeling, Watson et

al. show that the genetic diversity of cells

in the blood is predominantly determined

by positive selection, rather than neutral ge-

netic drift. Further, they quantified the fit-

ness benefit of specific mutations and dem-

onstrate that those that confer high fitness

are associated with risk of progression to

acute myeloid leukemia (AML)

(see the figure). In particular,

mutations in epigenetic modi-

fiers, including DNA methyl-

transferase 3a (DNMT3A), tet

methylcytosine dioxygenase 2

(TET2), additional sex combs–

like 1 (ASXL1), and the tumor

suppressor tumor protein 53

(TP53), were estimated to be

highly fit, and, consistent with

prior studies, the frequency of

such mutations in an individu-

al’s blood stratifies risk of AML

( 7 , 8 ). The authors also made the

surprising finding that a large

number of mutations (~2500)

in clonal hematopoiesis genes

conferred moderate to high fit-

ness, highlighting a broad and

rugged fitness landscape result-

ing from the potential for in-

teractions between mutations.

More generally, application of

the approach of Watson et al.

to broader-coverage sequenc-

ing data (that extend beyond

known hotspots) has the po-

ing tumor growth ( 11 ). Of note, selection was

found to be more stringent in premalignant

Barrett’s esophagus compared with matched

invasive esophageal carcinomas from the

same individual ( 9 ), consistent with the

strong selection for specific mutations that

occurs before overt hematologic malignancy,

as observed by Watson et al.

Premalignant tissues such as Barrett’s

esophagus, colorectal adenomas, and ductal

carcinoma in situ in the breast are under-

studied, but molecular characterization may

help to define the earliest alterations in these

tissues, features associated with progression

to invasion, and patients at risk. In recogni-

tion of the need for such studies, the U.S.

National Cancer Institute Human Tumor

Atlas Network will generate precancer atlases

for a variety of tissues—including breast, co-

lon, lung, and skin—with a focus on spatial

and temporal molecular profiling ( 12 ). Clonal

evolution in the blood is distinct from that

in tissues, both because it is not subject to

the same spatial constraints and because of

the large number of hematopoietic stem cell

progeny in the circulation, and this may po-

tentially explain the widespread clinical con-

sequences of clonal hematopoiesis ( 13 ).

Clonal hematopoiesis has additional im-

plications for response to therapy, bone

marrow transplantation, and noninvasive

detection of malignancy based on muta-

tional profiling of circulating cell-free DNA

with a reported prevalence of 10 to 20% of

individuals over 65 ( 13 ). However, this is a

low estimate owing to the constraints of ana-

lytic sensitivity. Error-corrected sequencing

methods capable of detecting single-nucle-

otide variants in the 0.01% frequency range

will enable refined estimates of the preva-

lence of clonal hematopoiesis, which may be

pervasive after middle age.

In myeloproliferative diseases and blood

cancers, a shift away from reliance on clini-

cal and morphological features

toward molecular classification

is under way ( 14 , 15 ). A major

objective of research is to de-

velop predictive models that

inform patient stratification on

the basis of specific mutations

and their frequency ( 7 , 8 ). By

demonstrating that the size of

clones harboring pathogenic

mutations and the distribution

of fitness effects influence the

pace of progression, Watson

et al. provide further ratio-

nale for such efforts. Critically,

however, numerous factors can

influence these dynamics—in-

cluding infection, inflamma-

tion, and treatment—making

long-term disease forecasting

more challenging. Future stud-

ies will likely leverage longitu-

dinal cohorts that follow sub-

jects over time. The integration

of clinical, demographic, and

phenotypic information with

detailed genomics will yield in-

sights into clonal dynamics and

mutation rates of healthy blood,

as well as molecular features as-

sociated with prognosis and ag-

ing. Such information should also provide

clues as to how to better detect and inter-

cept early malignancy, when therapeutic in-

tervention should be most effective. j

REFERENCES AND NOTES

1. R. L. Bowman et al., Cell Stem Cell 22 , 157 (2018).


2. C. J. Watson et al., Science 367 , 1449 (2020).


3. J. L. Tsao et al., Proc. Natl. Acad. Sci. U.S.A. 97 , 1236
   (2000).


4. I. Martincorena et al., Science 362 , 911 (2018).


5. H. Lee-Six et al., Nature 561 , 473 (2018).


6. K. Yizhak et al., Science 364 , eaaw0726 (2019).


7. S. Abelson et al., Nature 559 , 400 (2018).


8. P. Desai et al., Nat. Med. 24 , 1015 (2018).


9. M. J. Williams et al., Nat. Genet. 50 , 895 (2018).


10. M. Gruber et al., Nature 570 , 474 (2019).


11. R. Sun et al., Nat. Genet. 49 , 1015 (2017).


12. S. Srivastava et al., Trends Cancer 4 , 513 (2018).


13. D. P. Steensma, B. L. Ebert, Exp. Hematol. S0301-
    472X(19)31133-6 10.1016/j.exphem.2019.12.001
    (2019).


14. E. Papaemmanuil et al., N. Engl. J. Med. 374 , 2209
    (2016).


15. J. Grinfeld et al., N. Engl. J. Med. 379 , 1416 (2018).

10.1126/science.aba9891

Targeted DNA sequencing of blood

cells from ~50,000 individuals

Neutral drift

Positive selection

Mutation

prevalence

Selection

or drift?

Neutral

mutation

Advantageous

mutation, leading to

clonal hematopoiesis

Time (age)

0.25 0.50 0.75 1.00

100

75

50

25

0

tential to reveal additional 0.00^

preleukemic drivers and thus

has implications for predicting

progression to disease. Of note, different

mutations in the same gene are not neces-

sarily equivalent, and mutations may confer

fitness benefits through different mecha-

nisms or no fitness benefit at all. Functional

phenotypic studies will be needed to resolve

these differences.

The study of Watson et al. demonstrates the

value of moving beyond cataloging mutations

to modeling genomic data within an evolu-

tionary framework. Previous studies have

sought to quantify evolutionary dynamics

from genomic sequence data in established

cancers ( 9 – 11 ). For example, the relative fit-

ness of a clone has been inferred from variant

allele frequency distributions in solid tumors

( 9 ) and in chronic lymphocytic leukemia ( 10 ).

Given that tissue architecture is inherent to

solid tumors and influences clonal dynamics,

other studies have sought to explicitly model

tissue spatial constraints when inferring the

extent of selection versus neutral drift dur-

SCIENCE

Variant allele frequency

Number of mutations

27 MARCH 2020 • VOL 367 ISSUE 6485 1427

Age-related mutation accumulation in the blood

DNA sequencing of blood cells from ~50,000 individuals and mathematical

modeling revealed the accumulation of mutations with age and associated

fitness advantages. Positive selection for pathogenic mutations results in clonal

hematopoiesis and increased risk of acute myeloid leukemia.