table S2). The mutant marrow cells expressed
a stimulated inflammatory response relative
to controls, with increased expression of genes
such asfosabandchac1(fig. S7, A to C). The
source of inflammation appeared to be mutant
macrophages and neutrophils, as niche cells
were devoid of inflammatory cytokine ex-
pression (Fig. 3C and fig. S7, A and B). Elevated
inflammatory cytokine levels have been de-
scribed in clonal hematopoiesis, but their ef-
fect on clonal expansion is not known ( 13 , 14 ).
Acute inflammation enhances HSPC cycling
but can lead to exhaustion if it persists ( 15 ),
and chronic infection can potentiateDnmt3a
mutant hematopoiesis ( 16 ). We hypothesized
that mutant HSPCs moderate their response
to chronic inflammation maintained by their
mutant mature progeny, rendering them re-
sistant to replicative exhaustion. We identified
elevated gene expression of suppressors of in-
flammation, such asnr4a1andsocs3a,inmu-
tant HSPCs and progenitors (Fig. 3C and fig.
S7, A and B). We compared the gene expres-
sion of sorted marrow cells of dominant clones
from mutant zebrafish to that of total marrow
cells from control zebrafish and found simi-
larly dysregulated inflammatory gene signatures
(Fig. 3C and fig. S7D), suggesting that mutant
dominant clones express an anti-inflammatory
gene signature in response to inflammation.
Runx1deficiency has been associated with in-
flammation ( 17 ). A scRNA-seq comparison of
germline homozygousrunx1mutants with oli-
goclonal hematopoiesis to heterozygous mutants
with polyclonal hematopoiesis revealed a sim-
ilar pro-inflammatory/anti-inflammatory reg-
ulatory axis (fig. S8).
We reasoned that genetic abrogation of the
anti-inflammatory response in mutant clones
would reduce their fitness. We used TWISTR
to generate a cohort with gRNAs targeting
asxl1andnr4a1, a gene that encodes a nu-
clear receptor antagonizing immune activa-
tion and overexpressed in dominant mutant
clones ( 18 ) (Fig. 4A). We observed normal
hematopoiesis in these zebrafish (fig. S9, A
and B). We sorted 306 clusters from 20 zebra-
fish injected withasxl1-targeting gRNA and
58 zebrafish injected with gRNAs targeting
bothasxl1andnr4a1, and assessed indels (Fig.
4B). Of these, 186 harbored deleteriousasxl1
indels: 48 fromasxl1-injected and 138 from
asxl1/nr4a1-injected zebrafish. Comparing clus-
ter sizes, we observed that clusters with biallelic
deleterious indels ofnr4a1were significantly
smaller relative to those with nonr4a1indels,
in-frame/no indels, or heterozygous deleteri-
ous indels (Fig. 4C and fig. S9C).Asxl1-mutant
clones with biallelicnr4a1indels diminished
in peripheral blood over time, whereasasxl1-
mutant clones with intactnr4a1expanded
(fig. S9D). There was no difference between
wild-typeasxl1clusters acrossnr4a1genotypes,
which suggests thatnr4a1alone is not suffi-
cient to modulate clonal fitness (fig. S9E).
Within individual zebrafish, smallerasxl1-
mutant clusters were more likely than larger
asxl1-mutant clusters to harbor homozygous
deleteriousnr4a1indels (fig. S9F). These re-
sults suggest that a blunted anti-inflammatory
response inasxl1-mutant clones due tonr4a1
loss diminished their clonal fitness.
Our results provide evidence that progeny of
mutant clones generate an inflammatory envi-
ronment to which they are resistant (Fig. 4D).
Certain mutations in clonal hematopoiesis have
an association with specific cytokines, including
interleukin-6 withTET2mutations and tumor
necrosis factor–awithASXL1mutations ( 13 , 14 ).
It is plausible that there is mutation-dependent
cytokine expression in mutant mature cells, and
an up-regulation of mutation-associated anti-
inflammatory pathways in mutant progenitor
cells.Thus,clonalfitnessisachievedbyapro-
survival network engaged by mutations, which
offers an explanation for how gene mutations
with diverse functions confer similar clonal fit-
ness. Our study developed a genetic and live
clonal tracking technology to dissect the pro-
gression to clonal dominance, identifying new
SCIENCEscience.org 5 NOVEMBER 2021•VOL 374 ISSUE 6568 771
A
BCD
nr4a1
WT
-gRNA+gRNA
nr4a1
HET
0
20
40
60
80
Cluster size (%)
nr4a1
HOMO
*
*
*
nr4a1
WT
Cluster size (%)
0
10
20
30
40
50
60
70
frameshift
in-frame
Dominant clone
2bp D
9bp I
2bp D
asxl1 nr4a1
alleles wt
Small clone
alleles
5bp D
wt
2bp D
2bp D
asxl1 nr4a1
Grow to
adulthood
creER Clonal competition
drl T2
ubi
Color labeling
at 24 hpf
Sort clones
by FACS
DNA-seq
asxl1 gRNA
+/- nr4a1 gRNA
+ cas9 mRNA
Somatic mosaic mutants
3-5 months
asxl1
nr4a1
HSPCs
inflammatory
cytokines
mature progeny
asxl1 mutation
nr4a1
response
differentiation
Clonal fitness
Clusters with fs asxl1 indels nr4a1 mutation
Fig. 4. Biallelicnr4a1loss abrogates fitness ofasxl1-mutant clones.
(A) Testing the effect ofnr4a1with TWISTR. (B) Example DNA-seq analysis
of sorted clusters from a single zebrafish. D, deletion; I, insertion; bp, base pair.
(C) Size of sortedasxl1-mutant clusters in zebrafish injected with onlyasxl1
gRNA (nr4a1WT,n= 48) or withasxl1andnr4a1gRNAs that resulted in no
deleteriousnr4a1indels (nr4a1WT,n= 34), heterozygous deleterious indels
(nr4a1HET,n= 58), or homozygous deleterious indels (nr4a1HOMO,n= 46).
*P= 0.029asxl1-onlynr4a1WTversusnr4a1HOMO,P= 0.014nr4a1WTversusnr4a1HOMO,
P= 0.013nr4a1HETversusnr4a1HOMO(unpairedttest). (D) Model of inflammation-
driven clonal fitness with acquired mutations.
RESEARCH | REPORTS