ll
Bridging the Gap
Somatic genomic mosaicism in the
brain: Identified mutations provide
challenges and opportunities for
the clinic
Jerold Chun1,*
Recent work by Bae et al.1 represents a major next-generation
sequencing effort to identify somatic genomic mosaicism in normal
and diseased human brains. Some samples displayed age-associ-
ated hypermutability, and the general possibility that somatic muta-
tions can drive brain disease has implications for new therapeutic
strategies, disease staging, biomarkers, and cohort selection for
clinical trials.
A single human brain is composed of hun-
dreds of billions of truly diverse cells that
are exquisitely organized both neuroana-
tomically and functionally to enable
human activities. Disruptions to this orga-
nization likely underlie all brain disor-
ders—covering the gamut of neurological
and neuropsychiatric diseases—and the
development of disease modifying thera-
pies (DMTs) represents an enormous un-
met medical need, as epitomized by their
absence for most brain diseases, including
sporadic Alzheimer’s disease, Parkinson
disease, and numerous neuropsychiatric
disorders. This deficiency reflects in part
our incomplete understanding of the hu-
man brain, which has historically included
a fundamental assumption that all cells
within it share an identical genome.
However, over the last two decades, it has
become clear that individual brain cells
can contain different genomic informa-
tion, producing brains that are a highly
complex genomic mosaic. Genomic
mosaicism arises somatically and is thus
not ‘‘genetic’’ in the formal sense of being
hereditary, although there are likely ge-
netic mechanisms that contribute to
mosaicism. The DNA-altering forms
of genomic mosaicism are many,
including (from large to small) aneu-
648 Med 3, 648–650, October 14, 2022 ª 2022 Elsevier Inc.
ploidies,2 smaller copy number variations
(CNVs),3,4,5 Line1 elements,6 genomic
cDNAs (gencDNAs),7,8 and single nucleo-
tide variations and/or mutations (hereon
mutations).9,7,10,11 Each of these forms
involve alterations to the actual genomic
DNA sequence within a cell, contrasting
with epigenetic modifications, while mul-
tiple DNA sequence forms can concur-
rently exist. Detection depends on the
techniques utilized, involving many
variables that are generally biased to
detect clonally expanded mutations—
thus arising during neurogenesis for
post-mitotic neurons or in mitotic cell
populations such as glia or non-brain cells
from blood—that amplify identical DNA
targets via clonal cell expansion to allow
more sequencing depth (and statistical
significance).
Two notable consequences of genomic
mosaicism are, first, that discovery ef-
forts for causal (vs. risk-factor) disease
genes from peripheral (non-brain)
samples—a mainstay of genome-wide
association studies (GWASs)—cannot
assess functional disease genes that
are mutated somatically within the
brain; by analogy to cancers, oncogene
mutation discovery typically identifies
genes within actual tumors (cancers)
rather than normal tissues. Second, a
disease gene can have different, if not
opposite, effects depending on the
cell type and genome within which it re-
sides; this point is underscored by gene
effects that are cell autonomous vs.
non-cell autonomous, highlighting a
need to understand an identified muta-
tion within the brain’s complex cellular
and neuroanatomical organization at
the level of single cells.
Recent work by Bae et al.1 represents
a major next-generation sequencing
(NGS) effort to assess somatic genomic
mosaicism in non-diseased (neurotypic)
vs. sporadic psychiatric (Tourette syn-
drome, schizophrenia, and autism)
diseased human brain, empowered by
an NIH-supported consortium involving
scores of scientists from some 15 organi-
zations. They identified 20–60 mutations
present in almost all examined normal
and diseased brains, which were likely
generated during neurodevelopment
and did not markedly change with age.
Even more notable, however, was the
presence of samples with ‘‘hypermutabili-
ty’’ that did track with age, comprising
6% of the brains—both normal and
diseased—containing hundreds (and
in one schizophrenia brain, thousands)
of mutations, along with examples of
aneuploidies, CNVs, and other structural
variations, which further showed neuroan-
atomical regional differences, consistent
with prior work on mosaicism.11,5 Around
2% of mutations occurred in coding se-
quences (directing protein synthesis)
for all brains, similar to the 2% of the
germline human genome that provides
coding sequence and suggesting a
somewhat uniform mutational effect
across the entire genome. However, a
few mutations were found to affect cancer
1Sanford Burnham Prebys Medical Discovery
Institute, 10901 North Torrey Pines Road, La Jolla,
CA, USA
*Correspondence: jchun@sbpdiscovery.org
https://doi.org/10.1016/j.medj.2022.09.009

Bridging the Gap
and neuropsychiatric genes, albeit in both
normal and diseased brains. Aneu-
ploidies associated with glioblastoma
(but in the absence of tumors) were
observed in the hippocampus of one
brain (schizophrenia sample), although
the affected cell identities and tumori-
genic potential remain obscure albeit
intriguing. Non-coding myeloid ectopic
viral integration site motif mutations
were also observed in autism brains at a
higher frequency (4:1) than normal.
In terms of methodology, all analyses
were from post-mortem samples, pro-
cessed from small bits of brain: ‘‘bulk’’
tissue of 1 cm3 cubes or sorted brain
cell nuclei (using fluorescence-activated
nuclei sorting [FANS]) from 131 non-
diseased vs. neuropsychiatric diseased
brains, along with the addition of 16 sin-
gle-cell samples used for a validation
experiment. Cerebral cortex, striatum,
and hippocampus (and blood) were
sampled by one or more investigator
sites. Similar but non-identical wet-lab
procedures were used to generate
sequencing libraries in varied ways,
followed by short-read sequencing
(150 bp) using Illumina platforms
(HiSeqX or NovaSeq) at a depth of
over 2003. Bioinformatic treatments
including variable filtering steps that
can affect final data outputs were used
to call mutations. Some unknowns that
could affect mutation detection include
histological data (allowing better loca-
tion and information on cellular compo-
sition, such as cortical layers) for origi-
nating brain samples; clinical data on
the specific causes of donor death;
other metadata (e.g., co-morbid condi-
tions); postmortem intervals; and
RNA integrity number (RIN: a proxy for
tissue quality). FANS may miss or skew
analyzed cell populations, whereas
bulk sequencing can reduce bias but at
a cost of increased sample complexity
and reduced detection sensitivity. All
samples represented a tiny part of the
intact human brain (out of practical ne-
cessity), thus leaveing a vast majority
of brain cells unassessed. Bioinformatic
filtering and use of algorithms originally
designed for germline analyses but
applied to the highly mosaic brain could
miss mutations, especially at the lowest
assessed variant allele frequencies
(VAFs) of a few percent (VAF of a gene
for a germline heterozygote is 50%, ho-
mozygote 100%). Brain hypervariability
may be even more prevalent, based on
the somewhat arbitrary cutoff values
used to define it. For example, a cutoff
of >101 mutations reduced to, say,
>50 mutations would increase hyper-
variable brains from 6% to 20%. More-
over, the number and percentage of
observed mutations are not static since
they will be affected by the dynamic
loss (and/or possibly gain) of cells that
occurs with age. Lower VAF mutations
below a few percent of cells were not as-
sessed and/or not reported in bulk or
sorted populations, yet post-mitotic
events that affect single post-mitotic
neurons or other non-mitotic cells
would fall into this category, which limits
reported mutations to those occurring
in expanded proliferative cell popula-
tions. Single-cell analyses can identify
mutations with greater sensitivity, how-
ever some mutations (e.g., CNVs) may
be missed because of required enor-
mous template amplification, while the
approach fundamentally lacks formal
experimental replication because the
single cell is destroyed by sequencing
that cannot then be repeated on the
same cell. Representative sampling is
also a challenge, as reflected by the 16
single cells reported compared to 131
brains that each consist of hundreds of
billions of cells. The exclusive use of
short-read Illumina sequencing may
have missed sequences that could be
revealed by long-read (Pacific Biosci-
ences [PacBio] or Oxford Nanopore)
sequencing. The functional significance
of all mutations remains unclear,
especially for non-coding mutations
(representing 97% of the reported
mutations and seen in both non-
diseased and diseased brains). Never-
theless, these caveats do not detract
from the positive mutation data and
ll
simply underscore potential limita-
tions, especially affecting those singular
examples of hypervariable brains.
Bae et al.1 adds substantively to the
growing literature on somatic genomic
mosaicism within the mammalian brain
by examining a relatively large collection
of non-diseased and selected examples
of (presumably sporadic) neuropsychiatric
diseases. The universal presence of so-
matic DNA sequence changes involving
multiple mutational forms even within
single brains underscores an enlarging
universe of somatic genomic differences
present in a single brain and among indi-
viduals. These mutations are currently
unaccounted for in many studies of the
human brain or its diseases, including
those utilizing GWASs, induced pluripo-
tent stem cells, and organoids. This
universe is very likely much larger than re-
ported, given the sensitivity cutoffs used
by Bae et al. that would not detect muta-
tions present below a few percent of cells,
which would likely include all mutations
arising in post-mitotic neurons or non-
mitotic brain cells (which do not undergo
clonal expansion). The fascinating obser-
vation of hypermutable brains could be
extended to—and indeed may reflect—
hypermutable sites within a single brain
and/or hypermutable cells of a defined
type that populate the brain, a specula-
tion consistent with regionally varying
brain DNA content (DNA content varia-
tion) or gencDNAs associated with so-
matic gene recombination reported in
Alzheimer’s disease neurons.8,5 Aneu-
ploidies are a normal form of genomic
mosaicism and can occur in both
neurons and non-neuronal cells, high-
lighting potential linkage to cancer-like
mechanisms, as noted for potential glio-
blastoma but in the absence of a tumor,
or having different effects in cells of the
normal and diseased brain. Although the
linkages between identified gene muta-
tions from Bae et al. and disease states
remain to be established, the general
possibility that somatic genomic muta-
tions (as observed in cancers) may drive
brain disease could, if true, enable novel
Med 3, 648–650, October 14, 2022 649
 ll
target discovery for both the affected
gene and the enzymology that gives rise
to the sequence changes, which could
be accessed by a gamut of therapeutic
modalities including small molecules, an-
tibodies, gene therapies, and CRISPR-
based strategies. These same mutational
elements could also potentially be used
for disease staging, biomarkers from ce-
rebrospinal fluid and blood, and cohort
selection to improve clinical trials. Cancer
parallels likely exist, wherein the success
of DMTs to treat cancers may augur the
development of DMTs leveraging knowl-
edge of somatic genomic mosaicism to
treat diseases of the mind and brain.
ACKNOWLEDGMENTS
This work was supported by the NIH
(R01AG065541, R01AG071465, R56AG0
73965), the DOD (W81XWH-21-10642),
the Shaffer Family Foundation, and the
Bruce Ford and Anne Smith Bundy
Foundation.
DECLARATION OF INTERESTS
The author declares no competing
interests.
650 Med 3, 648–650, October 14, 2022
REFERENCES
1. Bae, T., Fasching, L., Wang, Y., Shin, J.H.,
Suvakov, M., Jang, Y., Norton, S., Dias, C.,
Mariani, J., Jourdon, A., et al. (2022). Analysis
of somatic mutations in 131 human brains
reveals aging-associated hypermutability.
Science 377, 511–517. https://doi.org/10.
1126/science.abm6222.
2. Rehen, S.K., McConnell, M.J., Kaushal, D.,
Kingsbury, M.A., Yang, A.H., and Chun, J.
(2001). Chromosomal variation in neurons
of the developing and adult mammalian
nervous system. Proc. Natl. Acad. Sci. USA 98,
13361–13366. https://doi.org/10.1073/pnas.
231487398.
3. Gole, J., Gore, A., Richards, A., Chiu,
Y.J., Fung, H.L., Bushman, D., Chiang,
H.I., Chun, J., Lo, Y.H., and Zhang, K.
(2013). Massively parallel polymerase
cloning and genome sequencing of
single cells using nanoliter microwells.
Nat. Biotechnol. 31, 1126–1132. https://
doi.org/10.1038/nbt.2720.
4. McConnell, M.J., Lindberg, M.R., Brennand,
K.J., Piper, J.C., Voet, T., Cowing-Zitron, C.,
Shumilina, S., Lasken, R.S., Vermeesch, J.R.,
Hall, I.M., and Gage, F.H. (2013). Mosaic copy
number variation in human neurons. Science
342, 632–637. https://doi.org/10.1126/
science.1243472.
5. Westra, J.W., Rivera, R.R., Bushman, D.M.,
Yung, Y.C., Peterson, S.E., Barral, S., and
Chun, J. (2010). Neuronal DNA content
variation (DCV) with regional and
individual differences in the human brain.
J. Comp. Neurol. 518, 3981–4000. https://
doi.org/10.1002/cne.22436.
Bridging the Gap
6. Muotri, A.R., Chu, V.T., Marchetto, M.C.N.,
Deng, W., Moran, J.V., and Gage, F.H. (2005).
Somatic mosaicism in neuronal precursor
cells mediated by L1 retrotransposition.
Nature 435, 903–910. nature03663 [pii] 10.
1038/nature03663. https://doi.org/10.1038/
nature03663.
7. Kaeser, G., and Chun, J. (2020). Brain cell
somatic gene recombination and its
phylogenetic foundations. J. Biol. Chem. 295,
12786–12795. https://doi.org/10.1074/jbc.
REV120.009192.
8. Lee, M.H., Siddoway, B., Kaeser, G.E.,
Segota, I., Rivera, R., Romanow, W.J.,
Liu, C.S., Park, C., Kennedy, G., Long, T.,
and Chun, J. (2018). Somatic APP gene
recombination in Alzheimer’s disease
and normal neurons. Nature 563,
639–645. https://doi.org/10.1038/s41586-
018-0718-6.
9. Costantino, I., Nicodemus, J., and Chun, J.
(2021). Genomic mosaicism formed by
somatic variation in the aging and diseased
brain. Genes 12, 1071. https://doi.org/10.
3390/genes12071071.
10. Lodato, M.A., Rodin, R.E., Bohrson, C.L.,
Coulter, M.E., Barton, A.R., Kwon, M.,
Sherman, M.A., Vitzthum, C.M., Luquette,
L.J., Yandava, C.N., et al. (2018). Aging and
neurodegeneration are associated with
increased mutations in single human
neurons. Science 359, 555–559. https://doi.
org/10.1126/science.aao4426.
11. Rohrback, S., Siddoway, B., Liu, C.S., and
Chun, J. (2018). Genomic mosaicism in the
developing and adult brain. Dev Neurobiol
78, 1026–1048. https://doi.org/10.1002/dneu.
22626.