heterochromatic bodies tend to be in the nu-
clear interior in Pvalb inhibitory neurons (Fig.
5C), unlike the more distributed positioning in
other cell types. Consequently, chromosome 17
tends to be radially positioned near the nu-
clear interior in Pvalb neurons (fig. S17, A to
C). Similarly, because chromosomes 11 and
19 have many H3K27me3 proximal points in
astrocytes (fig. S16E) and H3K27me3 globules
tend to be in the interior of the astrocyte nuclei
compared with neurons (fig. S17, A and C),
chromosomes 11 and 19 tend to be observed
near the interior and interact with other chro-
mosomes in astrocytes, but less so in neurons
(Fig. 5, A to C).
In addition, we observed a cell type depen-
dence in the average interchromosome spatial
distances (Fig. 5B, top, and fig. S17H). Notably,
those pairwise interchromosome distance maps
agree with averaged radial positioning of pairs
of chromosomes, such that chromosomes in
the nuclear interior tend to be spatially close
to each other (Fig. 5, B, bottom, and C). Con-
sequently, pairs of interchromosomal loci as-
signed as proximal points at nuclear bodies
tend to be spatially closer to each other than
the other nonproximal point pairs in neurons
and astrocytes (Fig. 5D). Transcriptionally
up-regulated loci between cell types tend to
show closer interchromosomal distances and
more interior radial positioning in the nu-
cleus at the population level (fig. S18). How-
ever, in single cells, these loci do not show
strong radial gradients (fig. S19), reflecting
the variable nuclear speckle organization in
individual nuclei and directly supporting a
refined model of radial nuclear organization
( 10 ). These results show that the complexity
in the global organization of chromosomes
can be captured in the different nuclear body
arrangements and proximal point chromatin
profiles.
Domain boundaries are variable in single cells
Bulk-averaged chromosome conformation
capture data revealed that chromosomes are
organized into topologically associating do-
mains (TADs) with clear insulation boundaries
( 19 , 27 , 28 ). More-recent imaging and sequenc-
ing experiments found TAD-like domain struc-
tures in single cells ( 7 , 12 , 15 , 16 , 29 ), whose
boundaries are variable and shift stochas-
tically across single cells upon depletion of
cohesin ( 12 ).
To systematically investigate such single-cell
chromosome domain structures across differ-
ent chromosomal regions, we determined
whether there are subpopulations of chromo-
somes with distinct configurations that differ
from the bulk averaged configurations in ex-
citatory neurons. We clustered the single chro-
mosome pairwise spatial distance data with
25-kb resolution using principal components
analysis ( 30 ) (fig. S20). Whereas the bulk data
displayed three domains at the chromosome
3 region (7.7 to 9.3 Mb) (Fig. 6A, top), analysis
of the single-cell data showed multiple con-
figurations with different pairwise spatial
associations and domain boundaries (Fig. 6, A,
bottom, and B, and fig. S20). Chromosomes
in single cells appeared to stochastically form
domains preferentially from a subset of CTCF-
and cohesin-binding sites (Fig. 6, A and B),
consistent with a previous observation ( 12 ).
We also observed single-cell domain structures
at chromosomal regions without clear en-
semble domain structures ( 12 ) (fig. S21), with
characteristic ensemble high-order domain
structures ( 9 ) (fig. S22), and with heteroge-
neous nucleolar associations (fig. S23), con-
firming the prevalence of single-cell domain
structures.Active and inactive X chromosome
organization in single cells
We further examined the differences between
the active X chromosome (Xa) and the inactive
X chromosome (Xi) ( 22 ) in the female mouse
brain cortex. The imaging-based genomics data
can straightforwardly distinguish the Xa and
Xi on the basis of their mutually exclusive
associations with Xist RNA, a long noncoding
RNA that is specifically expressed from and
associates with the Xi ( 22 ), and characterize
distinct epigenetic states between the Xa and
Xi (fig. S24, A and B).
We observed that the Xa and Xi have dis-
tinct median distances between pairs of loci at
different genomic length scales and in differ-
ent cell types (fig. S24, C to H). Although the Xi
is more compact than the Xa globally at the
larger scale of tens of megabases (fig. S24, C
and G) ( 11 , 31 ), we found, using the population-
averaged spatial distance quantification, that
theXaappearstobemorestructuredand
compact at megabase or smaller length scales
(fig. S24, C, D, G, and H).
Both the Xa and Xi have heterogeneous
domain structures in individual cells (Fig. 6,
C and D, and fig. S25). Even the region of the
Xi that appears unstructured in the bulk data
from both DNA seqFISH+ (Fig. 6C) and allele-
specific Hi-C studies ( 32 , 33 ) appeared to adopt
discrete domains in subsets of cells (Fig. 6D).
We found that similar domain subclusters
canbeusedinboththeXaandXi,butwith
different relative frequencies for specific chro-
mosome conformation (Fig. 6E), which leads
to different average conformations for the Xa
and Xi (Fig. 6C). A recent observation of het-
erogeneous single-cell domain structures in
the Xa and Xi even under epigenetic pertur-
bations ( 34 ) further supports our finding in
the Xa and Xi. Taken together, although the
Xa and Xi show very different epigenetic states
and ensemble-averaged chromosome confor-
mation, they can share similar underlying
single-cell domain structures with differentrelative conformational preferences at the sub-
megabase scale.Discussion
The integrated spatial measurements of RNA,
DNA, and chromatin marks in the same cell
in the mouse brain enable the observation of
proximal points, DNA loci that are present
at the exterior of nuclear bodies and protein
globules, across diverse cell types in tissues.
Because each chromosome contains a distinc-
tive pattern of proximal points associated with
nuclear bodies and chromatin marks, these
proximal points form a scaffold of the chro-
mosomes on the exterior of the nuclear bodies
in single cells. In addition, as the proximal
points profiles and nuclear body organization
are cell type dependent, these proximal points
lead to distinct chromosomal positioning and
interchromosomal relationships in the nu-
cleus in each cell type. These findings extend
our previous observations in mouse ES cells
( 17 ) and suggest a general principle of nuclear
organization.
The integrated measurements also allow
us to correlate the nuclear organization with
transcriptional states in different cell types.
We found that physical proximity to nuclear
speckles and H3K27me3 are related to tran-
scriptional changes in the tissues, extending
recent work in cell cultures ( 25 ). In contrast,
we observed that the submegabase domain
structures in the Xa and Xi can be similar in
single cells despite their large epigenetic and
transcriptional differences, suggesting a po-
tential decoupling between single-cell domain
structures and underlying epigenetic states for
the X chromosome.
The robust demonstration of integrated
spatial genomics in tissues indicates that the
same approach can be applied to a diverse
range of biological systems to further explore
the diversity and invariants in single-cell nu-
clear architecture. This integrated technology
can be further scaled up using signal ampli-
fication methods ( 35 – 37 ) to allow larger tissue
sections to be imaged. In addition, super-
resolution imaging of epigenetic markers over-
laid on the super-resolved DNA seqFISH+
data would provide finer resolution in single-
cell chromatin profiles ( 38 ). We anticipate
addressing many biological questions in
future studies using genome-scale chromo-
some imaging together with transcriptome-
scale profiling ( 39 , 40 ) and super-resolution
protein imaging ( 38 ).REFERENCESANDNOTES- T. Stuart, R. Satija,Nat.Rev.Genet. 20 , 257–272 (2019).
- C. Zhu, S. Preissl, B. Ren,Nat.Methods 17 , 11–14 (2020).
- T. Misteli,Cell 183 , 28–45 (2020).
- J. Dekkeretal.,Nature 549 , 219–226 (2017).
- R. Kempfer, A. Pombo,Nat.Rev.Genet. 21 , 207– 226
(2020). - A. M. Cardozo Gizzietal.,Mol.Cell 74 , 212–222.e5 (2019).
- L. J. Mateoetal.,Nature 568 , 49–54 (2019).
SCIENCEscience.org 29 OCTOBER 2021¥VOL 374 ISSUE 6567 593
RESEARCH | RESEARCH ARTICLES