Article

Ageing hallmarks exhibit organ-specific

temporal signatures

https://doi.org/10.1038/s41586-020-2499-y Nicholas Schaum1,2,38, Benoit Lehallier2,38, Oliver Hahn2,38, Róbert Pálovics2,

Shayan Hosseinzadeh3, Song E. Lee2, Rene Sit3, Davis P. Lee2,4, Patricia Morán Losada2,
Received: 5 June 2019
Macy E. Zardeneta2,4, Tobias Fehlmann5, James T. Webber3, Aaron McGeever3,
Accepted: 7 May 2020 Kruti Calcuttawala2, Hui Zhang2, Daniela Berdnik2, Vidhu Mathur2, Weilun Tan3,
Alexander Zee3, Michelle Tan3, The Tabula Muris Consortium*, Angela Oliveira Pisco3,
Published online: xx xx xxxx
Jim Karkanias3, Norma F. Neff3, Andreas Keller2,5, Spyros Darmanis3, Stephen R. Quake3,6 ✉ &
Check for updates Tony Wyss-Coray2,4,7,8 ✉

Ageing is the single greatest cause of disease and death worldwide, and
understanding the associated processes could vastly improve quality of life.
Although major categories of ageing damage have been identified—such as altered
intercellular communication, loss of proteostasis and eroded mitochondrial
function1—these deleterious processes interact with extraordinary complexity
within and between organs, and a comprehensive, whole-organism analysis of ageing
dynamics has been lacking. Here we performed bulk RNA sequencing of 17 organs
and plasma proteomics at 10 ages across the lifespan of Mus musculus, and integrated
these findings with data from the accompanying Tabula Muris Senis2—or ‘Mouse
Ageing Cell Atlas’—which follows on from the original Tabula Muris3. We reveal linear
and nonlinear shifts in gene expression during ageing, with the associated genes
clustered in consistent trajectory groups with coherent biological functions—
including extracellular matrix regulation, unfolded protein binding, mitochondrial
function, and inflammatory and immune response. Notably, these gene sets show
similar expression across tissues, differing only in the amplitude and the age of onset
of expression. Widespread activation of immune cells is especially pronounced, and
is first detectable in white adipose depots during middle age. Single-cell RNA
sequencing confirms the accumulation of T cells and B cells in adipose tissue—
including plasma cells that express immunoglobulin J—which also accrue
concurrently across diverse organs. Finally, we show how gene expression shifts in
distinct tissues are highly correlated with corresponding protein levels in plasma,
thus potentially contributing to the ageing of the systemic circulation. Together,
these data demonstrate a similar yet asynchronous inter- and intra-organ
progression of ageing, providing a foundation from which to track systemic sources
of declining health at old age.

To uncover ageing dynamics across the whole organism, we meas-
ured plasma proteins and sequenced RNA from 17 organ types isolated
from C57BL/6JN male mice (n = 4; aged 1, 3, 6, 9, 12, 15, 18, 21, 24 and
27 months; equivalent to humans aged 13, 20, 30, 36, 43, 50, 56, 63, 69
and 75 years, respectively) and female mice (n = 2; aged 1, 3, 6, 9, 12, 15, 18
and 21 months) (Fig. 1a, b). This encompasses development at 1 month
of age to maturity at 3–6 months, as well as ageing through adulthood
to the median lifespan of 27 months. We isolated all 17 organs from
each mouse, including bone (femurs and tibiae), brain (hemibrain),
brown adipose tissue (BAT, interscapular depot), gonadal adipose
tissue (GAT, inguinal depot), heart, kidney, limb muscle (tibialis ante-
rior), liver, lung, bone marrow, mesenteric adipose tissue (MAT), pan-
creas, skin, small intestine (duodenum), spleen, subcutaneous adipose
tissue (SCAT, posterior depot), and white blood cells (buffy coat). Raw
data are available from the Gene Expression Omnibus (under acces-
sion code GSE132040), and an interactive data browser is available at
https://twc-stanford.shinyapps.io/maca/. Concurrently, we performed
single-cell RNA sequencing on 529,823 cells from 20 organs across the
lifespan of the mouse, generating a Tabula Muris Senis that is presented
in the accompanying Article2.

Institute for Stem Cell Biology and Regenerative Medicine, Stanford University School of Medicine, Stanford, CA, USA. 2Department of Neurology and Neurological Sciences, Stanford
1

University School of Medicine, Stanford, CA, USA. 3Chan Zuckerberg Biohub, San Francisco, CA, USA. 4Veterans Administration Palo Alto Healthcare System, Palo Alto, CA, USA. 5Chair for
Clinical Bioinformatics, Saarland University, Saarbrücken, Germany. 6Department of Bioengineering, Stanford University, Stanford, CA, USA. 7Paul F. Glenn Center for the Biology of Aging,
Stanford University School of Medicine, Stanford, CA, USA. 8Wu Tsai Neurosciences Institute, Stanford University School of Medicine, Stanford, CA, USA. 38These authors contributed equally:
Nicholas Schaum, Benoit Lehallier, Oliver Hahn. *A list of members and their affiliations appears at the end of the paper. ✉e-mail: steve@quake-lab.org; twc@stanford.edu

Nature | www.nature.com | 1
Article
a =4 =2
b c d Age (months) e
GAT

Age (months)
BAT Heart 1 Downregulated Upregulated

12
15
18
21
24
27
18

1
3
6
9
1 4× MAT
Bone Intestine 9 Intestine
27 2× Skin
27 m 1m 18 SCAT

Dim. 2
Brain Kidney GAT
9 27 Kidney
24 m 3m GAT Liver 2,500 Lung
Dim. 1 Spleen Muscle
GAT SCAT
Lung Pancreas 1,250 Liver
21 m 6m Liver Lung BAT

DEGs
Marrow SCAT 9 27 Bone
0 Pancreas
MAT 18 Marrow

t-SNE 2
Skin
18 m 9m 27 −1,250 Heart
1 9 Brain
Muscle Spleen

Dim. 2
Dim. 2
Intestine WBC
15 m 12 m 18 −2,500 Spleen
WBC 1

6
9
12
15
18
21
24
27
Dim. 1 Dim. 1

15

24

15

24
6

6
t-SNE 1 Comparison
f g (relative to 3 months)
Age (months)
DEGs relative to 3 months
Gpnmb
3 Bhlhe41 1,000 1,000
specificity (AU)

1
Specificity for age (AU ×104)

Tissue/age

+3 Arntl
Apobec3 h i j k
Change in expression

2 Npas2 Igj (marrow) Igj (spleen) Igj (GAT) Igj

Per3

Normalized counts (×103)

Normalized counts (×103)

(×103)
10 2.5
Bhlhe40 20 3 2
Ciart 2.0

Normalized counts
1 15 1
Ephx1 1.5 2

z-score
Cd209f 10 0
Igj 1.0
1 −1
0 H2−Q6 0.5 5
0 1 2 3 –3 Ccl8
0 0 0 −2
Specificity for tissue (AU ×104) Cygb

1
6
12
18
24

1
6
12
18
24

1
6
12
18
24

1
6
12
18
24
Laptm5
Age (months) Age (months) Age (months) Age (months)
Tissue

Fig. 1 | Pairwise differential expression across organs. a, Experiment outline. of the data in d. f, Scatterplot displaying gene-wise enrichment scores for
17 organ types were collected from male (n = 4, 1–27 months old) and female tissue, age and tissue/age (see Methods section ‘Specificity of gene expression
(n = 2, 1–21 months old) mice. Images of organs were generated by Freepik, from for tissues and ages’). g, Tissue-wise expression changes with age (within each
www.flaticon.com. b, t-distributed stochastic neighbour embedding (t-SNE) column, from left to right) for the top 15 genes exhibiting shifts in most tissues.
visualization of all samples, based on the first 50 principal components. h–j, Igj expression in marrow (h), spleen (i) and GAT ( j). n = 51, 54 and 52
c, Diffusion maps of GAT, lung, and liver, coloured by age. n = 38 (GAT), n = 37 independent samples, respectively. The black lines indicate locally estimated
(liver), n = 38 (lung) independent samples. Dim., dimension. d, Smoothed line scatterplot smoothing (LOESS) regression. Data are mean ± s.e.m.
plot displaying the number of DEGs for pairwise comparisons, referenced k, z-transformed, smoothed gene expression trajectory of Ig j, coloured by
to data at 3 months. Positive (negative) values represent upregulated tissue. n = 53 (BAT), 54 (bone), 52 (GAT), 50 (intestine), 52 (kidney), 50 (liver),
(downregulated) genes, grey lines represent non-labelled tissues. e, Heat map 54 (lung), 51 (marrow), 54 (MAT), 52 (muscle), 54 (spleen) independent samples.

Sex also influences organ function, leading to divergent ageing and

Pairwise differential gene expression with age disease outcomes in humans6. We observed prominent sex effects in
Ageing instigates functional decline across organs, disrupting intricate GAT, SCAT, liver and kidney, which are possibly connected with known
crosstalk that is essential for maintaining healthy organismal processes. differences in fat storage, sex hormone regulation and renal haemody-
Although individual organ types transcriptionally segregate by age namics7–9 (Extended Data Figs. 1c, d, 3b). Differential gene expression
(Fig. 1c), we lack a basic comparative understanding of ageing between analysis between the sexes at each age revealed that these four tissues
these organs, including differences in the onset and rate of ageing. consistently displayed the most DEGs across the lifespan (Extended
We therefore performed pairwise differential expression to deter- Data Fig. 4a, b), and the large overlap between young (3-month) and
mine when differentially expressed genes (DEGs) arise and whether old (18-month) DEGs between the sexes suggests that these differences
they persist with advancing age. Although a few DEGs were observed are established early and are maintained throughout life (Extended
between organs at neighbouring ages, in most organs the number of Data Fig. 4c, d). However, the biological pathways that comprise these
DEGs increases markedly in older mice relative to 3-month-old adults, sex DEGs largely differ from those comprising ageing DEGs, therefore
suggesting progressive, gradual changes in gene expression that are providing no evidence that sex differences influence the transcriptional
detectable only after sufficient time (Fig. 1d, e). Some organs—such ageing profiles we observe here (Extended Data Fig. 4e–h).
as the pancreas and marrow— seem to be relatively refractory to gene We next asked whether early-life DEGs persist with advancing age,
expression changes upon ageing; this might be in part explained by or whether they give way to new DEGs at older ages. In bone, for exam-
the relatively small proportion of global variance due to ageing in the ple, the early-life expression of ossification genes decreases as bone
dataset (Extended Data Fig. 1a, b). Whereas the core profiles of age- formation is completed, and these genes are highly correlated with
ing mice are maintained relative to those of 6-month-old mice, the late-life DEGs for which expression is increased, typifying age-related
number of DEGs increases greatly when compared with organs from bone loss (Supplementary Table 1). Overall, most organs show cor-
1-month-old mice that are still undergoing development (Extended relation between early and late DEGs, which is exemplified by nearly
Data Fig. 2a–c). The clear outlier is the spleen, which displays large every pairwise comparison in GAT, liver, kidney and heart (Supple-
quantities of DEGs related to the cell cycle and blood cell development mentary Table 1). Few genes are unique to any individual age, with
even into middle age (data not shown). Notably, DEGs arise in SCAT and organ specificity outweighing age specificity (Fig. 1f). Differential
GAT in mid-life, before DEGs appear in other organs (Fig. 1d, Extended expression common between ageing organs is of particular interest,
Data Fig. 2d). These may be connected to known changes in adipose because ubiquitous ageing pathways may present new therapeutic
composition with age, especially regarding immune cell infiltration4. opportunities. When we isolated genes that are most frequently dif-
Some organs, such as mesenteric fat and the small intestine to which ferentially expressed across organs, we found strong enrichment for
it is attached, undergo acute and marked changes in gene expression those relating to immune response pathways (Fig. 1g, Extended Data
only late in life. To independently confirm our observations, we gener- Fig. 2e–h, Supplementary Table 2). Notably, the expression of Jchain
ated self-organizing maps for each organ to enable the visualization (also known as Igj)—which encodes the plasma B cell marker immu-
of correlated gene nodes5 (Extended Data Fig. 3a). Not only do global noglobulin J—is persistently increased throughout life in 11 out of
ageing nodes emerge, but white adipose tissues again exhibit strong 17 organs (Fig. 1g–k). The circadian clock genes Bhlhe40, Bhlhe41,
ageing profiles, and those of visceral GAT and MAT are highly similar. Arntl, Npas2, Per3, Ciart and Dbp also feature among the top DEGs

2 | Nature | www.nature.com
 a Cluster 8 Cluster 3 Cluster 7 Cluster 5 Cluster 1 Cluster 2 Cluster 4 Cluster 6 d Gene expression in GAT
Extra. matrix Mitochondrion Protein folding Stress response NA Signalling Immune resp. Immune resp.
0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4
z-score

+1
0 0 0 0 0 0 0 0

z-score
–0.4 n = 105 –0.4 n = 2,665 –0.4 n = 210 –0.4 n = 968 –0.4 n = 4,571 –0.4 n = 2,448 –0.4 n = 330 –0.4 n = 101
0 15 30 0 15 30 0 15 30 0 15 30 0 15 30 0 15 30 0 15 30 0 15 30
–1
b Age (months)
1.4 c
1.2 Cluster 4 Cluster 6 Cluster 7 Cluster 8
1.2 1.2 1 3 6 9 12 15 18 21 24 27
Cluster 6 GAT GAT 1.2 GAT MAT 1.8 MAT GAT
Cluster 8 SCAT MAT SCAT Age (months)
1.0 g

Amplitude (AU)
SCAT
Amplitude (AU)

BAT
MAT KEGG: Olfactory transduction
0.8 0.6 BAT BAT
Cluster 7 0.6 0.6 BAT 0.9 GO: Olfactory receptor activity
SCAT GO: Sensory perception
0.6 Cluster 4
GO: Sensory perception of smell
Cluster 3 GO: Sensory perception of chemical stimulus
0.4
Amplitude

Cluster 5 0 0 0 0 GO: Reg. of nucleic acid-templated transcription

Cluster 2 Variability GO: Reg. of transcription, DNA-templated
0.2 0 6.5 1.3 0 1.25 2.5 0 1.0 2.0 0 1.25 2.5
Cluster 1 GO: Regulation of RNA biosynthesis
Variability (AU)
GO: Regulation of RNA metabolism
0
BAT Brain Bone GAT Heart Intestine Kidney Liver Lung GO: Reg. of nucleobase-contng. comp. metab.
0 0.5 1.0 1.5 2.0 GO: Mitochondrial inner membrane
Variability (AU) Marrow MAT Muscle Pancreas SCAT Skin Spleen WBC GO: Organelle inner membrane
GO: Inflammatory response
e f 2.0 Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 GO: Response to external stimulus
2.0 2.0 2.0 2.0 GO: Immune system process
GO: Immune response
15 GO: Defense response
z-score

GO: Mitochondrion
10 0 0 0 GO: Mitochondrial part
Height

0 0
GO: Intracell. membrane-bounded organelle
5 GO: Cytoplasmic part
GO: Membrane-bound organelle
–2.0 n = 2,450 –2.0 n = 2,018 –2.0 n = 6,127 –2.0 n = 2,722 –2.0 n = 1,683 1 2 3 4 5 –log10(Padj)
0
Genes 0 15 30 0 15 30 0 15 30 Cluster
0 15 30 0 15 30 0 20
Age (months)

Fig. 2 | Ageing gene expression dynamics across organs. a, Whole-organism organs. For each cluster in a, an amplitude index and a variability index were
gene expression trajectory clustering. The trajectory for each gene was calculated. c, The amplitude and variability indices for the four clusters
averaged across all 17 organs, and those average trajectories were grouped into in b that showed the greatest trajectory change with age are shown, and
8 clusters. The number of genes and the top functionally enriched pathway for adipose tissues are indicated. d, Unsupervised hierarchical clustering was used
each cluster are reported. Within each cluster, the average trajectory for each to group genes with similar trajectories in GAT (n = 15,000 most highly
individual organ is overlaid. Cluster trajectories ± s.d. (n = 17 tissue trajectories) expressed genes). e, Clustering dendrogram and cut-off used to define five
are indicated in black and grey. Enrichment was tested using Fisher’s exact test independent clusters in GAT. f, Gene trajectories of the five clusters in e are
(Gene Ontology, GO) and the hypergeometric test (Reactome and Kyoto shown by grey lines. Purple lines outlined with white represent the average
Encyclopedia of Genes and Genomes, KEGG). q values were estimated with the trajectory for each cluster ± s.d. g, The top five pathways for each cluster in e.
Benjamini–Hochberg correction for each database separately, and for GO The number of genes is as in e, with the 15,000 most highly expressed genes as
classes (molecular function, cellular component, biological process) background. Enrichment and q values are as in a.
independently. b, Identification of stable and variable clusters between

(Extended Data Fig. 2h). Age-related circadian disruption is well known, Each cluster contains genes with similar global trajectories, but
but these data perhaps highlight the underappreciated, organism-wide organ-specific differences in the phase and magnitude of these tra-
role of circadian rhythms in declining health. Indeed, a malfunctioning jectories suggest that similar processes have unique dynamics . For
circadian clock is thought to contribute to metabolic and inflammatory each cluster we assigned an amplitude index (absolute change in
disorders, and shortened lifespan10. z-score of the mean trajectory between 1 month and 30 months) and
a variability index (a measure of the spread from the mean trajectory)
(Fig. 2b). This revealed that clusters with the largest amplitudes also
Gene expression dynamics with age show the strongest organ-specific behaviour, with adipose tissues
Pairwise comparisons are inherently limited, and our data enable the featuring prominently in clusters 4 and 6 (immune response), clus-
examination of gene expression dynamics with high temporal resolu- ter 7 (protein folding) and cluster 8 (extracellular matrix) (Fig. 2c).
tion across the lifespan. To reveal organism-wide processes, we first The lifelong increase in expression of genes relating to immune
searched for gene expression trajectories across the lifespan with response pathways is especially evident when organs are analysed
common behaviour between organs. We calculated the average tra- independently, specifically for adipose tissues such as GAT (Fig. 2d–g,
jectory for each gene across all 17 organs and clustered these averaged Extended Data Figs. 5, 6, Supplementary Table 5). Indeed, genes of
trajectories, revealing functional enrichment for hallmarks of ageing cytokine-mediated inflammatory pathways (such as GO:0019221)
such as increased inflammation, mitochondrial dysfunction, and loss show a pronounced increase in expression in GAT that begins at 18
of proteostasis (Fig. 2a, Supplementary Tables 3, 4). Notably, these months (Extended Data Fig. 7a, Supplementary Table 9). This includes
hallmarks have distinct dynamic patterns. For example, cluster 3 Ccl8, which shows very high correlation in the majority of tissue types
declines linearly across the lifespan and is strongly enriched for mito- (Extended Data Fig. 7b). Notably, however, even though many down-
chondrial genes, whereas cluster 7—encoding heat shock proteins stream processes are shared across organs (Fig. 2a), there is little
that are important for protein folding—demonstrates a sharp decline overlap of transcription factor regulatory networks (Supplementary
that begins only at 12 months of age. This is in contrast to cluster 8, Tables 6, 9). Additionally, although tissue-specific trajectory clusters
which contains extracellular matrix genes that decline rapidly until (Extended Data Fig. 5) show that several transcription factor genes
6 months, after which a more gradual decline prevails. Genes encod- are common between tissues—for example Irf1 in the kidney, lung
ing immune response pathways feature in clusters 4 and 6; the and heart—the clusters often differ in behaviour between tissues,
expression of cluster 4 genes such as beta-2 microglobulin (B2m) or are enriched for genes of disparate biological pathways (Sup-
and Igj increases steadily throughout life. Conversely, the expres- plementary Table 7). It therefore seems that, even though common
sion of immune genes of cluster 6—such as Cd74 and complement biological pathways emerge between tissues, these are not—for the
C1qa—exhibits a nonlinear increase, with a plateau between 9 and most part—driven by an underlying change to common transcription
15 months. factor regulatory networks. It is of course possible that more distal

Nature | www.nature.com | 3
Article
a 8 Aco1 (kidney) b Aco1 (kidney) c GAT Kidney Others j 6-month marrow k Kidney
1.4
U = –0.88 105 0.25
0.007%
P = 0.011

Cd138high cells (%)

Aco1 0.20

Disperse
4 104

Disperse

Single-cell dispersion score

(kidney)

t-SNE 2
Percentage 0.15
Normalized counts (x102)

B220-APC
of cells
103 0.10
25
0 0
t-SNE 1 0.9 50 0.05
1
6
12
18
24

–103 Cd138-PE
75 0
2 Ms4a7 (kidney) Ms4a7 (kidney) –103 0 103 –104 –105 3 m 24–29 m
U = 0.816 27-month marrow Marrow

Specific
105 0.4 P = 0.008
1 Igj (GAT) 0.18%

Cd138high cells (%)

Specific
t-SNE 2

0.4 0.3
Ms4a7 (kidney) 104

B220-APC
–1.0 –0.6 0.6 1.0 0.2
0
Correlation with age 103
1
6
12
18
24

t-SNE 1 Bulk RNA-seq

Age (months) 0 0.1

Expression (AU)
–103 Cd138-PE
d GAT
e B cell T cell
f Cd79a Igj 2.5 0
–103 0 103 –104 –105 3–6 m 18–29 m
25
P = 0.029 P = 0.029
Percentage of cells

MSCs Endothelial cell

0
l Igj Igj/DAPI

T cell
Myeloid cell

3-month kidney
t-SNE 2

t-SNE 2
3m
B cell FACS scRNA-seq
24 m
0
t-SNE 1 FACS scRNA-seq t-SNE 1 (n = 1,962)
m
m

m
m

(n = 1,962)
3
24

3
24

g h –ln(P value) i
Igj Xbp1 0 12 16 1–3 m
Expression (AU)

18–30 m

Igjhigh B cells (%)

ER stress 31
Cd79a exrpression

3
Retrograde protein transport 12

24-month kidney
ER to cytosol transport 12
8
0 ER-UPR 16
Antigen proc. & presentation 9
t-SNE 2

Actin filament polymerization 11

Droplet scRNA-seq 0
(n = 23,796) en
Ki Fat

H y
L t
M iver
e
Sp ow

t-SNE 1 –4 0 4

r
e

cl
ea
dn
le

us
r

log2fold change (Igj high/Igj low)

ar
M

Fig. 3 | Integration of bulk and single-cell transcriptomic data identifies present in the Tabula Muris Senis dataset (17 tissues), coloured by the plasma
cross-tissue infiltration of Igjhigh plasma B cells. a, Expression of Aco1 and B cell markers Ig j and Xbp1. h, GO terms enriched among the top 300 marker
Ms4a7 mRNA in the kidney. The black line indicates LOESS regression, and the genes of Ig jhigh (n = 1,198 cells) compared with B cells (n = 22,598 cells), with
Spearman’s rank correlation coefficient ρ is indicated. Data are mean ± s.e.m. 1,886 genes passing filtering as background. q-values are estimated using the
b, t-SNE plots of genes with ‘disperse’ (Aco1) and ‘specific’ (Ms4a7) single-cell Benjamini–Hochberg correction for each database separately, and for GO
expression patterns in the kidney. n = 1,108 cells. c, Single-cell dispersion classes (molecular function, cellular component, biological process)
scores (scRNA-seq) plotted against Spearman’s rank correlation coefficient independently. i, Distribution of IgJhigh cells, shown as percentages of
(≥0.6; bulk RNA-seq) for a given tissue. The colours represent different organ Cd79a-expressing cells for each tissue type. j, Representative FACS
types and the size of the dots corresponds to the percentage of cells per tissue scatterplots from 2 independent experiments, showing increased plasma
that express a given gene. d, t-SNE visualization of scRNA-seq data (FACS) cell abundance in aged marrow. Cd138, plasma cell marker; B220, B cell marker.
from GAT, coloured by age. A cluster of B cells that is present only in aged k, FACS quantification of plasma cells in kidney and marrow. n = 4 independent
GAT is circled. e, GAT B and T cells as a percentage of all analysed cells. n = 4 mice. Significance was assessed using Student’s t-test, data are mean ± s.e.m.
independent mice. Significance was assessed using Student’s t-test, data are l, Representative images from 2 independent experiments showing the
mean ± s.e.m. f, Expression of B cell marker Cd79a and plasma B cell marker Ig j. visualization of Igj in the kidneys of 3-month-old and 24-month-old mice using
g, t-SNE visualization of scRNA-seq data (droplet) of all Cd79a-expressing cells RNAscope. Almost no Igj signal was present in young kidneys. Scale bars, 100 μm.

regulatory elements—including enhancers or super enhancers—may in kidney macrophages (Fig. 3a–c). This is characteristic of many
provide a missing link, and it is anticipated that such regulatory other upregulated inflammatory genes that are specific to these
elements could be analysed in future studies. cells (Extended Data Fig. 8b). In addition to the dispersion score, we
used deconvolution software with cell-type-specific gene expression
profiles from the two datasets of the Tabula Muris Senis separately—
RNA sequencing confirms plasma B cell infiltration fluorescence activated cell sorting (FACS) coupled with Smart-seq2
A fundamental question emerging from transcriptomics of whole data (denoted ‘FACS’) and microfluidic-droplet-based scRNA-seq data
organs is whether the observed shifts in gene expression are driven (denoted ‘droplet’)—in order to estimate changes in abundance of cell
by cell-intrinsic changes with age or by changes in cell composi- types in each tissue with age11 (Extended Data Fig. 8h). Although the
tion. Using the Tabula Muris Senis2, a database of the ageing mouse cell types and profiles captured with these two methods do not always
obtained by single-cell RNA sequencing (scRNA-seq), we first asked overlap, in the nine cases for which both methods found the same
whether the top genes correlated with age from bulk RNA-seq (Sup- cell type to significantly change in abundance with age with an effect
plementary Table 8) were specific to an individual cell type or were size greater than 0.5, the correlations are highly concordant. In GAT
broadly expressed in multiple cell types. For each gene in each tis- and in the liver, a large increase in the number of B cells with age was
sue, we assigned a dispersion score based on the distribution of cells observed, providing further evidence that accumulating immune cells
expressing that gene in the single-cell data (Fig. 3a–c). For example, are a driver of the whole-organ inflammatory signal. Furthermore, cell
in the kidney, the aconitase genes Aco1 and Aco2—as well as citrate fractions as profiled by the four methods (FACS scRNA-seq, droplet
synthase (Cs)—are negatively correlated with age and are expressed scRNA-seq, FACS bulk deconvolution and droplet bulk deconvolu-
across multiple cell types, demonstrating an organ-wide decline of tion) show strong agreement, indicating that the results are highly
mitochondrial function (Extended Data Fig. 8a–f). Other genes, such stable (Extended Data Fig. 8g). Finally, as demonstrated by the Tabula
as Ms4a7, are positively correlated with age but are expressed only Muris Senis, a combination of changes in the abundance of different

4 | Nature | www.nature.com
 a 1.0 Vcam1*

correlated proteins
Fgf10* 6

Number of highly
BAT Kidney Pancreas
Postn* 4
Spearman’s correlation coefficient

Bone Liver SCAT

0.9
Cxcl12* 2 Brain Lung Skin
Thbs4* Postn* GAT Marrow Spleen
Vcam1* Ccdc80 0
Mmp2* Heart MAT WBC
Cd36 Sfn*

M le

SC ey
G T
Ki AT

Bo T
ne
0.8 Ccdc80

c
A

A
Intestine Muscle

dn
us
Mrc1*

M
Mmp2
Ide
Mrc1 Sema6a
Gpc3
High
expression rank
Average mRNA

0.7 Ptn
Nxph1
Matn2 Epb41 c d e f
Krt18 Stat3 VCAM1 (plasma) Vcam1 (kidney) Vcam1 (heart) Vcam1

Normalized counts (×102)

Normalized counts (×102)

Nr1d1 Ckb 3.7 8 8 2

log10(abundance, AU)
Ide H2afz
Gfap 3.5 6 6 1
Low
Bmp1

z-score
0.6 Ccl5
3.3 4 4 0
b 2 2 −1 Heart
3.1
GAT Kidney
Spleen 0 0 −2
Lung

1
6
12
18
24

1
6
12
18
24
1

6
12

18

24

6
12

18

24
BAT
1 Age (months) Age (months) Age (months) Age (months)
correlation coefficient
Muscle
MAT
Kidney g h i j
POSTN (plasma) Postn (BAT) Postn (GAT) Postn
Spearman

Heart

Normalized counts (×103)

Normalized counts (×103)

1.5 8 2

log10(abundance, AU)
Intestine 0 2.2
Liver BAT Lung
SCAT 2.1 1 GAT Muscle
Skin 1.0 MAT SCAT

z-score
WBC 2.0
Pancreas –1 4 0
Marrow 1.9 0.5
Bone −1
Brain 1.8
IDE
CKB
H2AFZ
KRT18
NXPH1
SEMA6A
MRC1
NR1D1
PTN
GPC3
GFAP
VCAM1
CCL5
MATN2
BMP1
EPB41
MMP2
CCDC80
POSTN
THBS4
FGF10
CXCL12
SFN
STAT3
CD36

0 0 −2
1

6
12

18

24

6
12

18

24
1
6
12
18
24

1
6
12
18
24
Age (months) Age (months) Age (months) Age (months)

Fig. 4 | Correlation of plasma proteins with organ-specific gene expression. (n = 52) (d) and in the heart (n = 52) (e). The black line indicates LOESS
a, Spearman correlation coefficient (≥0.6) between plasma protein abundance regression. Data are mean ± s.e.m. f, z-transformed, smoothed gene
and corresponding organ-specific gene expression. *q < 0.05; Benjamini– expression trajectory of Vcam1 in the kidney (n = 52) and in the heart (n = 52).
Hochberg correction per tissue. The size of the dots corresponds to the g, log-transformed abundance of the protein POSTN in plasma. n = 77
average gene expression across tissues. Top right, the number of proteins independent samples. In the box plots, the centre line represents the mean
correlated with gene expression in the top six organs. For details of the and the first and third quartiles show the minimum and maximum values.
analysis, see Methods. b, Heat map showing correlation coefficients for the top h, i, Expression of Postn mRNA in BAT (n = 53) (h) and in GAT (n = 51). (i). The
25 plasma proteins in a across all organs. c, log-transformed abundance of the black line indicates LOESS regression. Data are mean ± s.e.m. j, z-transformed,
protein VCAM1 in plasma. n = 77 independent samples. In the box plots, the smoothed gene expression trajectory of Postn in BAT (n = 53), GAT (n = 51), MAT
centre line represents the mean and the first and third quartiles show the (n = 54), lung (n = 54), muscle (n = 52) and SCAT (n = 55). Data are mean ± s.e.m.
minimum and maximum values. d, e, Expression of Vcam1 mRNA in the kidney

cell types and changes in cell-intrinsic gene expression is present in We observed an initial increase of Igj in marrow, bone and spleen—
most tissues; however, overall, cell-type composition seems to be organs responsible for producing adaptive immune cells (Extended
more important. Data Fig. 9i). Notably, the expression of Igj and Derl3 subsequently
Changes of cell-type composition with age, and in particular the increases in GAT and in the kidney at around 12 months of age, pre-
accumulation of immune cells in tissues such as visceral fat, is well ceding the increase of Igj in BAT, heart and lung. Igj expression also
established. However, temporal and cell-specific resolution is lacking. increases in human visceral and subcutaneous fat (Genotype-Tissue
Given that expansion of visceral fat predicts morbidity and mortality12, Expression (GTEx) data; Extended Data Fig. 9j). It is possible that the
we aimed to discover the origin of the age-related adipose inflamma- changes in expression of genes encoding chemokine signalling and
tory signature by using single-cell transcriptomic data from the Tabula cell-surface receptors and ligands that we observe here could explain
Muris Senis. We identified increasing numbers of T cells and B cells in this differential accumulation. We then reconstructed the B cell recep-
the gonadal adipose tissue with age, including a unique cluster of Cd79+ tor locus using scRNA-seq data, and found that Igjhigh plasma cells are
B cells that were present only in old mice (Fig. 3d, e); this is consistent predominantly of the IgM class (Extended Data Fig. 9k). Notably, we
with increased expression of adaptive immune response genes in whole detected several clones present across diverse tissues, suggesting that
organs (Fig. 2a, g). Unbiased screening of genes enriched in this popula- these cells are trafficked to tissues from a common origin (Extended
tion revealed high Igj expression (Fig. 3f, Supplementary Tables 11, 13). Data Fig. 9l). The role of these cells—or the specificity of the antibodies
Considering that we observed differential expression of Igj in 11 out they produce—is currently unknown, but it is tempting to speculate
of 17 whole organs (Fig. 1g), we analysed thousands of Cd79+ B cells that they may contribute to the global increase in autoantibodies that
across organs, revealing a unique cluster of Igjhigh cells in both the has been reported to occur upon ageing14.
FACS scRNA-seq and droplet datasets, concordant with high expres-
sion of plasma B cell markers Xbp1 and Derl3 (Fig. 3g, Extended Data
Fig. 9a, b, f, g). Relative to Igjlow cells, these cells show increased expression Correlation between plasma proteins and organ mRNA
of genes relating to the unfolded protein response and endoplasmic Investigating organs individually can reveal detailed ageing processes,
reticulum stress pathways, characteristic of highly secretory plasma and even common phenotypes that are potentially susceptible to inter-
B cells13 (Fig. 3h, Extended Data Fig. 9c, Supplementary Tables 12, 14). vention. However, ageing occurs systemically, with the decline of one
Notably, these plasma B cells originate almost entirely from aged organ possibly inciting or accelerating dysfunction throughout the
mice, and accumulate across diverse organs (Fig. 3i–l, Extended Data body. In part, this may be due to alterations in blood-borne factors that
Fig. 9d, e, h, Supplementary Fig. 1). Taking advantage of the high mediate intercellular and organ–organ communication. Encouraged
temporal resolution available with the whole-organ dataset, we also by heterochronic parabiosis experiments demonstrating rejuvena-
traced these cells across the lifespan through the expression of Igj. tion15,16, we and others have identified plasma proteins with detrimental

Nature | www.nature.com | 5
Article
or rejuvenating functions in ageing brain, muscle, pancreas, bone and
other organs17, as well as hundreds more proteins that are correlated
with human ageing and are associated with traits such as cognition and
grip strength18. However, the origins of these factors remain largely
unknown.
Here we attempted to determine which organs contribute to
age-related changes in the plasma proteome by correlating plasma
protein age trajectories with their corresponding gene expression
trajectories in each organ (Fig. 4a, b). This analysis revealed 25 plasma
proteins that are correlated (Spearman’s correlation coefficient
ρ > 0.6) with gene expression in at least one organ, totalling 35 unique
plasma protein/organ pairs. We discovered high correlation for several,
such as vascular cell adhesion molecule 1 (VCAM1) in the kidney and
fibroblast growth factor 10 (FGF10) in the spleen, and other notable
pairs such as glial fibrillary acidic protein (GFAP) in the brain. Of par-
ticular interest are VCAM1 and periostin (POSTN), which both show
exceptional correlation across several organs (Fig. 4c–j). VCAM1 was
recently identified as a critical mediator of brain ageing by old plasma19,
and the loss of POSTN in adipose tissues contributes to impaired lipid
metabolism20. Furthermore, both are implicated in extracellular matrix
regulation and fibrotic diseases, perhaps indicating age-related fibro-
sis. Throbospondin-4 (THBS4), levels of which decreased with age and
were highly correlated with gene expression in muscle in this study, has
recently been shown to promote synapse formation and be enriched
in young blood21. Notably, white adipose tissues emerge from this
analysis as well, with five plasma proteins highly correlated with gene
expression in visceral MAT and GAT, and three in SCAT (Fig. 4a). In limb
muscle, which has a modest number of DEGs, seven plasma proteins
are correlated with gene expression across the lifespan, including
POSTN, bone morphogenic protein-1 (BMP1), matrix metalloprotease-2
(MMP2) and other proteins associated with the extracellular matrix
(Extended Data Fig. 10a). Such extracellular-matrix-associated pro-
teins constitute a majority of the 25 plasma proteins we identified
(Extended Data Fig. 10b). Although these findings are intriguing, the
abundance of plasma proteins may change independently of differ-
ences in gene expression. For example, a tissue may preferentially
release greater amounts of certain proteins with age. Furthermore,
some tissues lack age correlation but show high expression of genes
encoding plasma proteins (Supplementary Table 10). Future research
will determine whether these changes in gene expression contribute to
functionally relevant age-related differences in the plasma proteome,
or whether they mediate other processes such as immune cell adhe-
sion and infiltration.
Discussion
Our dataset provides temporal resolution of the transcriptome and
the plasma proteome for all major organs across the entire lifespan
of the mouse, and can serve as a fundamental resource for biologists
across many disciplines. We discovered gene expression trajectories
consistent with previously identified deleterious processes such as
mitochondria dysfunction, impaired protein folding and inflammag-
ing. Furthermore, DEGs emerging in middle age are highly correlated
with those in late life, suggesting that these harmful processes emerge
early across diverse organs. With rejuvenation strategies such as
senescent cell ablation (senolytics), nutrient sensing manipulation
(by treatment with rapamycin and metformin) and plasma proteome
alteration progressing rapidly, we need an improved understanding of
where and when to apply these therapies22. Indeed, the common age-
ing patterns observed here may help to explain the profound health-
span benefits of such interventions. In summary, this organism-wide
characterization of ageing dynamics may accelerate the development
of therapeutics, and the insights into circadian rhythm disruption,
plasma cell accumulation and adipose decline suggest avenues for
renewed focus.
6 | Nature | www.nature.com
Online content
Any methods, additional references, Nature Research reporting sum-
maries, source data, extended data, supplementary information,
acknowledgements, peer review information; details of author con-
tributions and competing interests; and statements of data and code
availability are available at https://doi.org/10.1038/s41586-020-2499-y.
1. López-Otín, C., Blasco, M. A., Partridge, L., Serrano, M. & Kroemer, G. The hallmarks of
aging. Cell 153, 1194–1217 (2013).
2. The Tabula Muris Consortium. A single-cell transcriptomic atlas characterizes ageing
tissues in the mouse. Nature https://doi.org/10.1038/s41586-020-2496-1 (2020).
3. The Tabula Muris Consortium. Single-cell transcriptomics of 20 mouse organs creates a
Tabula Muris. Nature 562, 367–372 (2018).
4. Palmer, A. K. & Kirkland, J. L. Aging and adipose tissue: potential interventions for
diabetes and regenerative medicine. Exp. Gerontol. 86, 97–105 (2016).
5. Kohonen, T. The self-organizing map. Proc. IEEE 78, 1464–1480 (1990).
6. Austad, S. N. & Fischer, K. E. Sex differences in lifespan. Cell Metab. 23, 1022–1033 (2016).
7. Fuente-Martín, E., Argente-Arizón, P., Ros, P., Argente, J. & Chowen, J. A. Sex differences in
adipose tissue. Adipocyte 2, 128–134 (2013).
8. Zhang, Y. et al. Transcriptional profiling of human liver identifies sex-biased genes
associated with polygenic dyslipidemia and coronary artery disease. PLoS ONE 6,
e23506 (2011).
9. Sabolić, I. et al. Gender differences in kidney function. Pflugers Arch. 455, 397–429
(2007).
10. Hood, S. & Amir, S. The aging clock: circadian rhythms and later life. J. Clin. Invest. 127,
437–446 (2017).
11. Newman, A. M. et al. Determining cell type abundance and expression from bulk tissues
with digital cytometry. Nat. Biotechnol. 37, 773–782 (2019).
12. Kuk, J. L. et al. Visceral fat is an independent predictor of all-cause mortality in men.
Obesity (Silver Spring) 14, 336–341 (2006).
13. Shaffer, A. L. et al. XBP1, downstream of Blimp-1, expands the secretory apparatus and
other organelles, and increases protein synthesis in plasma cell differentiation. Immunity
21, 81–93 (2004).
14. Vadasz, Z., Haj, T., Kessel, A. & Toubi, E. Age-related autoimmunity. BMC Med. 11, 94
(2013).
15. Conboy, I. M. et al. Rejuvenation of aged progenitor cells by exposure to a young
systemic environment. Nature 433, 760–764 (2005).
16. Villeda, S. A. et al. Young blood reverses age-related impairments in cognitive function
and synaptic plasticity in mice. Nat. Med. 20, 659–663 (2014).
17. Castellano, J. M. Blood-based therapies to combat aging. Gerontology 65, 84–89 (2019).
18. Lehallier, B. et al. Undulating changes in human plasma proteome profiles across the
lifespan. Nat. Med. 25, 1843–1850 (2019).
19. Yousef, H. et al. Aged blood impairs hippocampal neural precursor activity and activates
microglia via brain endothelial cell VCAM1. Nat. Med. 25, 988–1000 (2019).
20. Graja, A. et al. Loss of periostin occurs in aging adipose tissue of mice and its genetic
ablation impairs adipose tissue lipid metabolism. Aging Cell 17, e12810 (2018).
21. Gan, K. J. & Südhof, T. C. Specific factors in blood from young but not old mice directly
promote synapse formation and NMDA-receptor recruitment. Proc. Natl Acad. Sci. USA
116, 12524–12533 (2019).
22. Mahmoudi, S., Xu, L. & Brunet, A. Turning back time with emerging rejuvenation
strategies. Nat. Cell Biol. 21, 32–43 (2019).
© The Author(s), under exclusive licence to Springer Nature Limited 2020
The Tabula Muris Consortium
Nicole Almanzar9, Jane Antony1, Ankit S. Baghel1, Isaac Bakerman1,10,11, Ishita Bansal1, Ben A.
Barres12, Philip A. Beachy1,15,22, Daniela Berdnik2, Biter Bilen2, Douglas Brownfield13, Corey
Cain16, Charles K. F. Chan17, Michelle B. Chen6, Michael F. Clarke1, Stephanie D. Conley3,
Spyros Darmanis3, Aaron Demers3, Kubilay Demir1,14, Antoine de Morree2, Tessa Divita3,
Haley du Bois4, Hamid Ebadi3, F. Hernán Espinoza13, Matt Fish1,14,15, Qiang Gan2, Benson M.
George1, Astrid Gillich13, Rafael Gòmez-Sjöberg3, Foad Green3, Geraldine Genetiano3,
Xueying Gu15, Gunsagar S. Gulati1, Oliver Hahn2, Michael Seamus Haney2, Yan Hang15,
Lincoln Harris3, Mu He18, Shayan Hosseinzadeh3, Albin Huang2, Kerwyn Casey Huang3,6,19, Tal
Iram2, Taichi Isobe1, Feather Ives3, Robert Jones3, Kevin S. Kao1, Jim Karkanias3, Guruswamy
Karnam20, Andreas Keller2,5, Aaron M. Kershner1, Nathalie Khoury2, Seung K. Kim15,21,
Bernhard M. Kiss1,22, William Kong1, Mark A. Krasnow13,14, Maya E. Kumar23,24, Christin S.
Kuo9,13,14, Jonathan Y. Lam15, Davis P. Lee2,4, Song E. Lee2, Benoit Lehallier2, Olivia Leventhal2,
Guang Li25,38, Qingyun Li12, Ling Liu2, Annie Lo3, Wan-Jin Lu1, Maria F. Lugo-Fagundo2, Anoop
Manjunath1, Andrew P. May3, Ashley Maynard3, Aaron McGeever3, Marina McKay3, M. Windy
McNerney26,27, Bryan Merrill19, Ross J. Metzger28,29, Marco Mignardi6, Dullei Min9, Ahmad N.
Nabhan13, Norma F. Neff3, Katharine M. Ng13, Patricia K. Nguyen1,39, Joseph Noh1, Roel
Nusse13,14,15, Róbert Pálovics2, Rasika Patkar20, Weng Chuan Peng15,40, Lolita Penland3, Angela
Oliveira Pisco3, Katherine Pollard30, Robert Puccinelli3, Zhen Qi3, Stephen R. Quake3,6,
Thomas A. Rando2,4,7, Eric J. Rulifson15, Nicholas Schaum1,2, Joe M. Segal20, Shaheen S.
Sikandar1, Rahul Sinha1,31,32,33, Rene V. Sit3, Justin Sonnenburg3,19, Daniel Staehli2, Krzysztof
Szade1,34, Michelle Tan3, Weilun Tan3, Cristina Tato3, Krissie Tellez15, Laughing Bear Torrez
Dulgeroff1, Kyle J. Travaglini13, Carolina Tropini19,41,42,43, Margaret Tsui20, Lucas Waldburger3,
Bruce M. Wang20, Linda J. van Weele1, Kenneth Weinberg9, Irving L. Weissman1,31,32,33, Michael
N. Wosczyna2, Sean M. Wu1,25, Tony Wyss-Coray2,4,7,8, Jinyi Xiang1, Soso Xue6, Kevin A.
Yamauchi3, Andrew C. Yang6, Lakshmi P. Yerra2, Justin Youngyunpipatkul3, Brian Yu3, Fabio
Zanini6, Macy E. Zardeneta2,4, Alexander Zee3, Chunyu Zhao3, Fan Zhang28,29, Hui Zhang2,
Martin Jinye Zhang35,36, Lu Zhou12 & James Zou3,35,37
9
Department of Pediatrics, Pulmonary Medicine, Stanford University School of Medicine,
Stanford, CA, USA. 10Stanford Cardiovascular Institute, Stanford University School of
Medicine, Stanford, CA, USA. 11Department of Medicine, Division of Cardiology, Stanford
University School of Medicine, Stanford, CA, USA. 12Department of Neurobiology,
Stanford University School of Medicine, Stanford, CA, USA. 13Department of Biochemistry,
Stanford University School of Medicine, Stanford, CA, USA. 14Howard Hughes Medical
Institute, Chevy Chase, MD, USA. 15Department of Developmental Biology, Stanford
University School of Medicine, Stanford, CA, USA. 16Flow Cytometry Core, Veterans
Administration Palo Alto Healthcare System, Palo Alto, CA, USA. 17Department of Surgery,
Division of Plastic and Reconstructive Surgery, Stanford University, Stanford, CA, USA.
18
Department of Physiology, University of California, San Francisco, CA, USA.
19
Department of Microbiology & Immunology, Stanford University School of Medicine,
Stanford, CA, USA. 20Department of Medicine and Liver Center, University of California
San Francisco, San Francisco, CA, USA. 21Department of Medicine and Stanford Diabetes
Research Center, Stanford University, Stanford, CA, USA. 22Department of Urology,
Stanford University School of Medicine, Stanford, CA, USA. 23Sean N. Parker Center for
Asthma and Allergy Research, Stanford University School of Medicine, Stanford, CA, USA.
24
Department of Medicine, Division of Pulmonary and Critical Care, Stanford University
School of Medicine, Stanford, CA, USA. 25Division of Cardiovascular Medicine, Department
of Medicine, Stanford University, Stanford, CA, USA. 26Mental Illness Research Education
and Clinical Center, Veterans Administration Palo Alto Healthcare System, Palo Alto, CA,
USA. 27Department of Psychiatry, Stanford University School of Medicine, Stanford, CA,
USA. 28Vera Moulton Wall Center for Pulmonary and Vascular Disease, Stanford University
School of Medicine, Stanford, CA, USA. 29Department of Pediatrics, Division of Cardiology,
Stanford University School of Medicine, Stanford, CA, USA. 30Department of Epidemiology
and Biostatistics, University of California, San Francisco, CA, USA. 31Department of
Pathology, Stanford University School of Medicine, Stanford, CA, USA. 32Ludwig Center for
Cancer Stem Cell Research and Medicine, Stanford University School of Medicine,
Stanford, CA, USA. 33Stanford Cancer Institute, Stanford University School of Medicine,
Stanford, CA, USA. 34Department of Medical Biotechnology, Faculty of Biochemistry,
Biophysics and Biotechnology, Jagiellonian University, Krakow, Poland. 35Department of
Electrical Engineering, Stanford University, Palo Alto, CA, USA. 36Department of
Epidemiology, Harvard T.H. Chan School of Public Health, Boston, MA, USA. 37Department
of Biomedical Data Science, Stanford University, Palo Alto, CA, USA. 38Department of
Developmental Biology, University of Pittsburgh School of Medicine, Pittsburgh, PA,
USA. 39Cardiology Section, Veterans Administration Palo Alto Healthcare System, Palo Alto,
CA, USA. 40Princess Máxima Center for Pediatric Oncology, Utrecht, The Netherlands.
41
School of Biomedical Engineering, University of British Columbia, Vancouver,
British Columbia, Canada. 42Department of Microbiology and Immunology, University
of British Columbia, Vancouver, British Columbia, Canada. 43Humans and the
Microbiome Program, Canadian Institute for Advanced Research, Toronto, Ontario,
Canada.
Nature | www.nature.com | 7
Article
Methods
Mice and organ collection
Male and virgin female C57BL/6JN mice were shipped from the National
Institute on Ageing colony at Charles River (housed at 19–23 °C) to the
Veterinary Medical Unit (VMU; housed at 20–24 °C)) at the VA Palo Alto
(VA). At both locations, mice were housed on a 12-h light/dark cycle,
and provided food and water ad libitum. The diet at Charles River was
NIH-31, and Teklad 2918 at the VA VMU. Littermates were not recorded
or tracked, and mice were housed at the VA VMU for no longer than
2 weeks before euthanasia, with the exception of mice older than
18 months, which were housed at the VA VMU beginning at 18 months
of age. After anaesthetization with 2.5% v/v Avertin, mice were weighed,
shaved, and blood was drawn via cardiac puncture before transcardial
perfusion with 20 ml PBS. Whole organs were then dissected in the fol-
lowing order: pancreas, spleen, brain, heart, lung, kidney, mesenteric
adipose tissue, intestine (duodenum), gonadal adipose tissue, muscle
(tibialis anterior), skin (dorsal), subcutaneous adipose tissue (inguinal
pad), brown adipose tissue (interscapular pad), bone and bone marrow
(femurs and tibiae). Mice were randomized and organs collected from
8:30–16:00 over several days. Organs were immediately snap-frozen on
dry ice. All animal care and procedures were carried out in accordance
with institutional guidelines approved by the VA Palo Alto Committee
on Animal Research.
Sample size, randomization and blinding
No sample size choice was performed before the study. Randomiza-
tion was performed in the case of mouse dissection order and during
the preparation of 96-well plates for cDNA creation. Blinding was not
performed: the authors were aware of all data and metadata-related
variables during the entire course of the study.
RNA isolation and preparation
Snap-frozen bone and skin was crushed on liquid nitrogen with a mortar
and pestle. Snap-frozen whole organs, or crushed bone or skin, were
placed in TRIzol and immediately homogenized with a TissueRuptor
in 50 ml conical flasks (see Supplementary Table 15 for organ-specific
details). Debris from the homogenate was pelleted in 1.5-ml tubes at
12,000g for 5 min at 4 °C. The supernatant was then transferred to a
new 1.5-ml tube to which chloroform was added. After vortexing on
maximum speed for 10 s, samples were transferred to 1.5-ml or 2-ml
Phase Lock Gel tubes, to which water was added before spinning at
12,000g for 5 min at 4 °C. The aqueous phase was then transferred to
a new tube, and after adding isopropanol, mixtures were vortexed at
max speed for 10 s. Solutions were then run through RNeasy columns
according to the manufacturer’s instructions, and eluted with the indi-
cated volume of water. RNA was then quantified with a nanodrop, and
frozen at −80 °C.
cDNA synthesis, library preparation, sequencing and data
processing
Methods including cDNA synthesis using the Smart-seq2 protocol23,
library preparation using an in-house version of Tn524,25, library pool-
ing, quality control, sequencing and data processing are provided at
https://doi.org/10.17504/protocols.io.2uvgew6.
Quality control and differential expression analysis
Samples with fewer than 4 million uniquely mapped reads were cen-
sored to exclude low-coverage samples. Data visualization and analy-
sis were performed using custom Rstudio scripts and the following
Bioconductor packages: Rtsne, Deseq226, topGO, destiny and org.
Mm.eg.db. To assess the quality of our dataset, the raw count matrix
was normalized using DEseq2 before conducting the built-in variance
stabilizing transformation. Principal component analysis revealed that
samples clustered mostly by tissue, with the exception of the white
fat tissues. We also plotted our data after running t-SNE, using 500
iterations and retaining either 50 or 6 principal components—that is,
with most principal components or only the ones explaining the most
variance, respectively. We further complemented these analyses with
hierarchical clustering using Ward’s clustering algorithm. In order to
detect whether samples within a given tissue would show profound
clustering by age, we finally calculated diffusion maps using the R pack-
age destiny with default parameters.
To identify significant differential expression changes with age, we
used the raw count matrix as recommended for the DEseq2 standard
analysis pipeline. Factors and dispersion estimates were calculated for
each tissue separately. We conducted differential expression analysis
comparing samples from 3-month-old mice to each consecutive time
point, using age and sex as covariates. P values were adjusted for mul-
tiple testing, and genes with an adjusted P value of less than 0.05 were
determined to be statistically significant. In addition, we ran similar
analyses using 1-month-old or 6-month-old mice as reference.
To rank genes on the basis of their regulation across tissues, we
summarized in how many tissues a given gene would be called as sig-
nificantly regulated in at least one comparison between samples from
3-month-old mice and any following sampling time point.
Gene expression trajectory analysis
To estimate gene trajectories during ageing, normalized counts
from DEseq2 were z-scored and LOESS regression was fitted for
each gene using the median expression per age group in each tissue.
Whole-organism trajectory per gene was estimated using the average
trajectory across the 17 tissues. Organism-wide analysis focused on
11,403 genes expressed in all tissues (that is, genes among the 15,000
most expressed genes in each tissue).
The distance matrix between whole-organism gene trajectories was
computed using the Euclidian distance and hierarchical clustering
was performed using the complete method. We identified 10 clusters
of genes changing with age, ranging from 1 to 4,571 genes. Clusters
9 and 10 were excluded from further analysis as they included fewer
than 10 genes.
To identify clusters that changed the most between tissues, we
computed an amplitude and variability index. The amplitude index
corresponds to the change in the z-score (absolute value) of the aver-
age trajectory between 1 month and 27 months. The variability index,
which measures the spread of organ trajectories, corresponds to the
average Euclidian distance between each organ-specific trajectory and
the organism-wide trajectory.
Reactome, KEGG and GO databases were queried to understand the
biological functions of each cluster. We used the R TopGO27 package
for GO analysis and the R clusterprofiler28 package for KEGG and Reac-
tome analyses. The 11,403 genes expressed in all tissues served as the
background set of genes against which to test for over-representation.
Because clusterprofiler requires EntrezID as input, we mapped gene
symbols to EntrezID using the org.Mm.eg.db29 package. When Gene
Symbols were mapped to multiple EntrezID, only the first EntrezID was
used. q values were estimated using the Benjamini–Hochberg approach
for the different databases taken separately. In addition, for GO analysis,
q values were calculated for the three GO classes (molecular function,
cellular component, biological process) independently.
Organ-specific clustering was performed using the 15,000 most
expressed genes per tissue. For each tissue, five clusters were con-
sidered for further analysis and the pathways analysis used the
corresponding background set of genes against which to test for
over-representation.
Single-cell RNA-sequencing analysis
Pre-processed and annotated scRNA-seq data (FACS followed by
Smart-seq2 protocol) from gonadal adipose tissue (3 and 24 months
old) were obtained from the Tabula Muris Senis consortium. Given
the lack of data from aged female mice, we focused our analyses on
samples derived from male mice. Additionally, cells with fewer than
200 or more than 6,500 genes were excluded. This yielded 1,962
high-quality cell transcriptomes derived from four young and four old
biological replicates. Data visualization and analysis were performed
using custom Rstudio scripts and the following Bioconductor pack-
ages: Seurat (v.3)30 and topGO. Data normalization and scaling was
performed using Seurat’s built-in SCTransform31 function with default
parameters. A shared-nearest-neighbours graph was constructed using
the first 30 principal components before clustering cells using Seu-
rat’s built-in FindClusters function with a resolution of 0.8 and default
parameters. Annotations for B cells and T cells were adopted from
the Tabula Muris Senis2. Cell numbers were normalized to the total
number of detected cells and compared using standard t-tests. Igjhigh
B cells formed a separate cluster and were identified using Seurat’s
FindMarkers function (parameters: only.pos = T min.pct = 0.15 thresh.
use = 0.25, test = ’MAST’).
To profile Igjhigh B cells organism-wide, we obtained the complete
and pre-processed scRNA-seq dataset (FACS followed by Smart-seq2
protocol) from the Tabula Muris Senis2—encompassing cells isolated
from all major tissues. Focusing on data from samples from male mice
only, we filtered for cells showing detectable expression of the Cd79a
gene (alpha chain of the B cell receptor). Cell transcriptomes were
thus derived from four young and four old biological replicates. The
resulting 10,867 cells were analysed using Seurat, as described above.
A shared-nearest-neighbours graph was constructed using the first
10 PC dimension before clustering cells using Seurat’s built-in Find-
Clusters function with a resolution of 0.4 and default parameters. Igjhigh
B cells formed a separate cluster (cluster 11; 129 cells). To character-
ize Igjhigh-specific expression profile, Seurat’s FindMarkers function
(parameters: only.pos = F min.pct = 0.15 thresh.use = 0.25, test = ’MAST’)
was run comparing the cluster of Igjhigh B cells against all other Cd79a
cells in the dataset. Functional enrichment analysis of the top 300
differentially expressed genes (sorted by adjusted P value) were com-
pared to all 1,051 genes passing the filtering parameters for the test.
Top-ranked GO terms were selected and visualized using the CellPlot
package (https://github.com/dieterich-lab/CellPlot). The full-length
GO terms were shortened to fit into the figure format; the complete
table of significantly enriched GO terms and associated genes can be
found in Supplementary Table 12.
Single-cell RNA-sequencing analysis (microfluidic droplet)
The Tabula Muris Senis encompasses scRNA-seq data generated with
microfluidic droplets, enabling the profiling of more cells without prior
selection of surface markers32. To profile Igjhigh B cells organism-wide,
we obtained the complete and pre-processed droplet dataset. Focus-
ing on data from samples from male mice only, we filtered for cells
showing detectable expression of the Cd79a gene. Cell transcriptomes
were derived from the following sampling time points: 2× 1 month,
2× 3 months, 2× 18 months, 4× 24 months and 3× 30 months. The result-
ing 23,796 cells were analysed using Seurat, as described above. A
shared-nearest-neighbours graph was constructed using the first 10 PC
dimension before clustering cells using Seurat’s built-in FindClus-
ters function with a resolution of 0.4 and default parameters. Igjhigh
B cells formed a separate cluster (cluster 5; 1,198 cells). To character-
ize Igjhigh-specific expression profile, Seurat’s FindMarkers function
(parameters: only.pos = F min.pct = 0.15 thresh.use = 0.25, test = ’MAST’)
was run comparing the cluster of Igjhigh B cells against all other Cd79a
cells in the dataset. Functional enrichment analysis of the top 300
differentially expressed genes (sorted by adjusted P value) were com-
pared to all 1,886 genes passing the filtering parameters for the test.
Top-ranked GO terms were selected and visualized using the CellPlot
package. The full-length terms were shortened to fit into the figure
format; the complete table of significantly enriched GO terms and
associated genes can be found in Supplementary Table 14.
Plasma proteomic analysis
All animal care and procedures were carried out in accordance with
institutional guidelines approved by the VA Palo Alto Committee on
Animal Research. Sixty-five male (n = 5–6 per age group; 1, 3, 6, 9, 12,
15, 18, 21, 24, 27 and 30 months old) and 16 virgin female (n = 4 per age
group; 3, 12, 18 and 21 months old) C57BL/6JN mice were shipped from
the National Institute on Ageing colony at Charles River (housed at
19–23 °C) to the Veterinary Medical Unit (VMU; housed at 20–24 °C))
at the VA Palo Alto (VA). Mice were provided food (NIH-31 at Charles
River, and Teklad 2918 at the VA VMU) and water ad libitum. Mice were
housed on a 12 h/12 h light/dark cycle at both places. Mice older than
18 months were housed at the VA VMU until they reached the experi-
mental age. Mice younger than 18 months were housed for less than
2 weeks at the VA VMU. After anaesthetization with 2.5% v/v Avertin,
blood was drawn via cardiac puncture. EDTA-plasma was isolated by
centrifugation at 1,000g for 10 min at 4 °C. Samples were aliquoted,
stored at −80 °C and sent on dry ice to SomaLogic.
The SomaLogic platform is primarily designed to detect and meas-
ure human proteins. To reduce the influence of cross-species effects
on our analysis, we first determined proteins in our dataset with high
evolutionary conservation between mice and humans. To this end,
we downloaded the plain text file containing all homologies between
mouse and human along with sequence identifiers for each species
(HOM_MouseHumanSequence.rpt) from MGI (http://www.informat-
ics.jax.org/). Next, reference protein sequences for human and mouse
were downloaded from UniProt (https://www.uniprot.org/). Using the
R “Biostrings” library a global pairwise sequence alignment was car-
ried out between the human and mouse sequences. Only sequences
with identity of 80% across the whole alignment were included in the
downstream analyses.
To determine the effect of age on the plasma proteome, relative
fluorescent units (RFUs) provided by Somalogic were log10-transformed
and linear models adjusted for age and sex were used. Type II sum of
squares (SS) were calculated using the R package car33 and q values were
estimated using the Benjamini–Hochberg approach34.
Raw protein abundance data, as measured by the Somalogic plat-
form, were scaled (z-transformed) before calculating the median
across all replicates for each sampled age time point to obtain an
average trajectory. Normalized RNA-seq counts after pre-processing
with DESeq2 were transformed for each tissue alike. To compare
plasma protein changes with shifts in mRNA expression in any tis-
sue, we calculated pairwise Spearman’s rank correlation coefficients
between a given protein trajectory with the expression trajectory
for its corresponding gene in each tissue separately. Thus, a given
protein could be correlated with expression changes in multiple tis-
sues. To limit our analysis to mRNA /protein pairs reflecting robust
changes, we filtered the resulting mRNA/protein correlations as fol-
lows: (1) The protein had to exhibit a sequence homology between
human and mouse of at least 75% (770 proteins). (2) The protein had
to change significantly with age according to the linear modelling
analysis (q < 0.05; 115 proteins). (3) The corresponding gene had to
be differentially expressed in the given tissue in at least one pairwise
comparison between 3 months old mice and any consecutive time
point (q < 0.05; 95 proteins). (4) The mRNA /protein profiles had to
exhibit a Spearman’s rank correlation of at least 0.6 (25 proteins; 35
protein/gene pairs). Benjamini–Hochberg correction per tissue was
applied to assess significance of proteins /mRNA profiles correlations.
Given that genes can be expressed at differing levels across tissues,
we additionally calculated average mRNA expression ranks for each
gene. To this end, we ranked for a given gene each tissue on its average
mRNA expression, based on baseMean in DESeq2.
To investigate connectivity networks between top proteins corre-
lated with organ gene expressed, we used String v.11.0, available at
https://string-db.org/35.
Article
Correlation analysis of gene expression and ageing using
self-organizing maps
For every gene a tissue-wise Spearman’s rank correlation coefficient
was computed on the basis of expression and age. Similarly, we com-
puted Spearman’s rank correlation coefficients for expression and
sex. The resulting correlation matrix was then filtered, such that only
genes that were significantly correlated with either age or sex in any
tissue (P < 0.01 after false discovery rate (FDR) adjustment with the
Benjamini–Hochberg procedure34) were considered. This matrix was
then used to create a self-organizing map (SOM)5 with the Kohonen R
package (v.3.0.8)36. In addition to Spearman’s rank correlation coef-
ficients for comparing gene expression and sex we also tested other
measures, such as log transformed P values from one-sided Wilcoxon
rank sum test (testing both—that is, ‘greater’ and ‘less’ alternatives and
flipping the sign if the ‘greater’ alternative yielded a lower P value) and
Somers’ D. Because the different approaches gave similar results we
continued to use Spearman’s correlation for gender and age for better
comparability.
Specificity of gene expression for tissues and ages
To identify how specific a gene is expressed at a certain time point or
in a certain tissue, we used an approach known from gene set analysis,
so-called gene set enrichment analysis (GSEA). Instead of computing
how significantly a biological category is enriched in a sorted list of
genes, we computed how enriched a certain tissue, age, or pair of tis-
sue/age is in each gene. For each gene we computed 10 enrichment
scores for the 10 ages, 17 enrichment scores for the 17 tissues and 170
enrichment scores for their combinations. The specificity is defined as
the difference between the maximal enrichment and second maximal
enrichment for all tissues, time points and combinations. The higher
the difference the more specific the gene either for a tissue, time point
or the combination thereof. Notably, the maximal enrichment score
can be translated into P values—for example, by random sampling. For
better comparability, however, we present the running sum directly
and do not translate them to P values. Here, the approach is referred
to as sample set enrichment analysis (SSEA).
Estimating the variance of the data depending on metadata
To estimate the variance in the data depending on age, tissue or gen-
der we made use of principal variance component analysis (PVCA) as
implemented in the Bioconductor Package pvca. PVCA combines the
strength of principal component analysis and variance components
analysis (VCA). Originally it was applied to quantify batch effects in
microarray data. In our case, however, we do not provide experimental
batches but rather groups of meta data as input.
Human transcriptomic data: human gene expression data were
obtained from the GTEx portal website (http://www.gtexportal.org/).
Read counts from the GTEx analysis v7 (GTEx_Analysis_2016-01-15_v7_
RNASeQCv1.1.8_gene_reads.gct) were normalized using yarn37.
Single-cell dispersion score
We first selected FACS cells from Tabula Muris Senis with more than
500 distinct genes and 50k reads. Owing to the lower number of rep-
licates of female mice, we considered only four 3-month-old and four
24-month-old male mice. We then created log (1 + counts per million
(cpm)) transformed single-cell count matrices per tissue, calculated
a 16-dimensional PCA from this data for each tissue, and embedded
each cell in a 16-dimensional latent space using the PCA components.
Note that the dimensionality of the PCA was chosen on the basis of the
explained variance per component. For each gene in each tissue, we
then selected those cells expressing it (log (1 + cpm) > 0), and calculated
their weighted centre in the single-cell latent space, where we consid-
ered normalized gene expression values as weights. We then defined the
‘single-cell dispersion’ as the weighted mean distance of the cells from
their centre, normalizing within each tissue to enable cross-tissue com-
parisons. Finally, we defined methodology to systematically analyse
gene expression: we computed for each gene per tissue the Spearman
correlation of its bulk DEseq2 normalized gene expression with ageing.
We then plotted the single-cell dispersion against the Spearman’s rank
correlation coefficient for each bulk ageing DEG.
Cell intrinsic gene expression versus cell abundance
We first selected FACS cells from the Tabula Muris Senis with more
than 500 distinct genes and 50k reads. Owing to the lower number of
replicates of female mice, we considered only data from male mice.
We then used the log (1 + cpm) transformed data, and considering
each tissue separately, we binned these cells into young (≤3 months;
Y) and old (>3 months; O). For each gene, we then calculated the log2
fold-change of cell counts and read counts between Y and O, where
cell count is defined as the fraction of cells within the tissue expressing
the gene (log (1 + cpm) > 0), and read count is defined as the mean read
count of the gene in the cells that express it. Next, we analysed each
gene in the bulk data, first computing the Spearman’s correlation of
DESeq2 normalized gene expression with age. We then binned genes
into those increasing with age (Spearman >0.7; I) and decreasing with
age (Spearman <0.7; D). Finally, we compare the single-cell log2 Y versus
O cell and read counts with the bulk correlations I and D by running
the Wilcoxon–Mann–Whitney test, to determine whether single-cell
data based cell or read count changes separate the bulk groups I and
D. We then plot the resultant read and cell count U statistics against
the corresponding P value for each tissue for the droplet and FACS
single-cell data separately.
Deconvolution with CIBERSORTx
We first used the signature matrix creation feature of CIBERSORTx11,
which detects cell-type-specific signature genes using annotated
single-cell data. As input, we used male cells from Tabula Muris Senis
of every age to create tissue-specific signature matrices from the FACS
and droplet data, separately. Consistent with cell selection criteria
used through the manuscript, we selected droplet cells with more than
500 distinct genes and 5k reads, and more than 500 distinct genes
and 50k reads from FACS cells. We input single-cell normalized cpm
data without log transformation, as well as cpm bulk tissue data on
which to perform deconvolution. Deconvolution was performed in
S-mode owing to the possibility of high technical variance, with all
other parameters set to default, and the resultant inferred cell type
fractions were correlated with age using Spearman’s rank correlation.
Igj and plasma cell validation experiments
RNAscope Multiplex Fluorescent Reagent Kit v2 was used on fresh
frozen kidney sections 20-μm thick from 3-month-old and 24-month-old
C57BL/6JN male mice, with Igj-C1 probes and Opal 690 Reagent Pack
(Akoya Biosciences FP1497001KT), with no modifications to the kit
instructions. Images were acquired at 20× on a Keyence BZ-X710 fluo-
rescence microscope.
Cell suspensions for FACS of plasma cells were generated from GAT
and kidney. After cardiac perfusion with PBS, tissues were immediately
dissected and minced to a paste with scissors. Samples were resus-
pended in 40 ml buffer (GAT, 10% horse serum in F12; kidney, 2% FBS
in RPMI) and passed through a 100-μm filter into a 50 ml conical tube,
grinding with a syringe plunger to further dissociate clumps. Filters
were washed with 5 ml buffer, and all 45 ml was then passed through a
35-μm filter, washing with another 5 ml buffer. Filtered cells were then
pelleted (500g, 5 min, 4 °C) and treated with 10 ml ACK lysis buffer for
5 min at room temperature. After pelleting, washing with 5 ml buffer
and pelleting again, cells were resuspended in 1 ml FACS buffer (2% FBS
in PBS), and aliquoted through 35-μm strainer caps into FACS tubes.
Cells were pelleted and resuspended in the antibody cocktail: Cd138-PE
(1:400; BioLegend142504), Cd19-BV421 (1:200; BioLegend 115538),
B220-APC (1:100; BioLegend 103212) for 30 min on ice. After pelleting
and resuspending, cells were stained with 1 μl 1:1,000 SYTOX Green
(Thermo Fisher S7020) immediately before sorting on a BD FACS Aria
III. Gates were set to capture live singlets and reduce debris according
to standard procedures, and Cd138high cells were quantified.
Reporting summary
Further information on research design is available in the Nature
Research Reporting Summary linked to this paper.
Data availability
Raw data are available from the Gene Expression Omnibus under acces-
sion code GSE132040. Source data are provided with this paper.
23. Picelli, S. et al. Full-length RNA-seq from single cells using Smart-seq2. Nat. Protoc. 9,
171–181 (2014).
24. Picelli, S. et al. Tn5 transposase and tagmentation procedures for massively scaled
sequencing projects. Genome Res. 24, 2033–2040 (2014).
25. Hennig, B. P. et al. Large-scale low-cost NGS library preparation using a robust Tn5
purification and tagmentation protocol. G3 (Bethesda) 8, 79–89 (2018).
26. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion
for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).
27. Alexa, A. & Rahnenfuhrer, J. topGO: enrichment analysis for gene ontology
(Bioconductor, 2016).
28. Yu, G., Wang, L. G., Han, Y. & He, Q. Y. clusterProfiler: an R package for comparing
biological themes among gene clusters. OMICS 16, 284–287 (2012).
29. Carlson, M. org.Mm.eg.db: Genome wide annotation for Mouse (Bioconductor, 2017).
30. Butler, A., Hoffman, P., Smibert, P., Papalexi, E. & Satija, R. Integrating single-cell
transcriptomic data across different conditions, technologies, and species. Nat.
Biotechnol. 36, 411–420 (2018).
31. Hafemeister, C. & Satija, R. Normalization and variance stabilization of single-cell RNA-seq
data using regularized negative binomial regression. Genome Biol. 20, 296 (2019).
32. Zheng, G. X. Y. et al. Massively parallel digital transcriptional profiling of single cells.
Nat. Commun. 8, 14049 (2017).
33. Fox, J. & Weisberg, S. An R Companion to Applied Regression (Sage, 2011).
34. Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a practical and powerful
approach to multiple testing. J. R. Stat. Soc. B 57, 289–300 (1995).
35. Szklarczyk, D. et al. STRING v11: protein-protein association networks with increased
coverage, supporting functional discovery in genome-wide experimental datasets.
Nucleic Acids Res. 47, D607–D613 (2019).
36. Wehrens, R. & Lutgarde, M. C. Buydens. “Self-and super-organizing maps in R: the
Kohonen package. J. Stat. Softw. 21, 1–19 (2007).
37. Paulson, J. N. et al. Tissue-aware RNA-seq processing and normalization for
heterogeneous and sparse data. BMC Bioinformatics 18, 437 (2017).
Acknowledgements We thank members of the Wyss-Coray laboratory for volunteering help
during organ collection and processing: E. Berber, K. Brewer, B. Chang, J. Marschallinger, L.
Bonanno, J. Sun, M. F. Lugo-Fagundo, A. Yang and M. W. McNerny. We also thank the members
of the Wyss-Coray laboratory and the Chan-Zuckerberg Biohub for feedback and support, and
H. Zhang and K. Dickey for laboratory management. This work was funded by the Department
of Veterans Affairs (BX004599 to T.W.-C.), the National Institute on Aging (R01-AG045034 and
DP1-AG053015 to T.W.-C.), the NOMIS Foundation (T.W.-C.), The Glenn Foundation for Medical
Research (T.W.-C.) and the Wu Tsai Neurosciences Institute (T.W.-C.).
Author contributions N.S., B.L., S.R.Q. and T.W.-C. conceptualized the study. N.S., O.H., B.L.,
R.P. and T.W.-C. conceptualized the analysis. O.H., B.L. and R.P., with contributions from A.K.
and T.F., conducted the transcriptomic analyses. N.S., S.E.L., D.P.L., M.E.Z., H.Z. and D.B.
collected samples and extracted RNA. S.H., A.Z. and W.T. conducted cDNA and library
preparation. P.M.L. created the Shiny web interface. R.S. and M.T. performed sequencing and
library quality control. S.H., A.O.P., J.T.W. and A.M. processed raw sequencing data. The Tabula
Muris Consortium generated the single-cell sequencing database. N.S. and K.C. performed
and analysed FACS and RNAscope experiments. N.S., B.L., O.H., T.W.-C. and S.R.Q. wrote and
edited the manuscript. T.W.-C., S.R.Q., S.D., N.F.N. and J.K. supervised the work. A full list of
author contributions for The Tabula Muris Consortium can be found in the Supplementary
Information.
Competing interests The authors declare no competing interests.
Additional information
Supplementary information is available for this paper at https://doi.org/10.1038/s41586-020-
2499-y.
Correspondence and requests for materials should be addressed to S.R.Q. or T.W.-C.
Peer review information Nature thanks Fan Zhang and the other, anonymous, reviewer(s) for
their contribution to the peer review of this work.
Reprints and permissions information is available at http://www.nature.com/reprints.
Article
Extended Data Fig. 1 | Gene expression variance analysis. a, Visualization of
the principal variance component analysis, displaying the gene expression
variance explained by residuals (that is, biological and technical noise) or
experimental factors such as tissue, age, sex and respective combinations.
n = 904 total samples. b, c, t-SNE visualization of all samples, based on the first
six principal components coloured by age (b) and sex (c). d, Hierarchical
clustering of all samples using Ward’s algorithm. Samples are annotated by
tissue, sex and age. Highlighted are samples clustering by sex in selected
tissues. Non-specific clustering of samples derived from white adipose tissues
is further highlighted.
Extended Data Fig. 2 | See next page for caption.
Article
Extended Data Fig. 2 | Validation of differential gene expression analysis.
a, Heat map displaying the number of DEGs per tissue for pairwise analysis on
adjacent time points. b, Heat map displaying the number of DEGs per tissue for
pairwise comparisons, referenced to tissues from 1-month-old mice. c, Heat
map displaying the number of DEGs per tissue for pairwise comparisons,
referenced to 6-month tissues. d, Box plot displaying the number of DEGs per
tissue (n = 17 tissues) for pairwise comparisons with a 3-month reference. In box
plots, the centre line denotes the mean, box boundaries denote the first and
third quartiles. Outliers show tissues undergoing exceptionally strong
expression shifts at a given age. e, Enrichment for functional categories in the
top 100 genes that are differentially expressed in the most tissues (ranked
using pairwise comparisons with a 3-month reference). Pathway enrichment
with GO, Reactome and KEGG databases. Enrichment was tested using Fisher’s
exact test (GO) and the hypergeometric test (Reactome and KEGG). To estimate
the contribution of each tissue, we used the number of genes per pathway in
the top 100 DEGs and estimated the percentage of significant genes per tissue.
q-values were estimated using the Benjamini–Hochberg procedure for each
database separately, and for GO classes (molecular function, cellular
component, biological process) independently. Values of n are as in d.
f, Cumulative sum of DEGs per tissue in the ranked top 100 genes. g, Number of
DEGs per tissue in the top 100 genes. n = 54 (MAT), 52 (kidney), 52 (GAT),
54 (spleen), 50 (liver), 54 (lung), 50 (intestine), 55 (SCAT), 51 (skin), 53 (BAT),
52 (heart), 52 (muscle), 53 (brain), 52 (WBC), 54 (bone), 51 (marrow),
46 (pancreas). q-values are as in e. h, STRING analysis of the top 30 genes in Fig. 1g.
Extended Data Fig. 3 | SOMs of gene correlation with age and sex. a, b, SOMs age (a) and sex (b). Genes with similar correlation are mapped to the same cell,
were generated from transcriptome-wide gene expression correlation and cells are grouped by similarity. The SOM cell layout is common across
(Spearman’s rank correlation coefficient) of each gene (n = 12,462 genes) with organs, with the average across all organs at the bottom.
Article
Extended Data Fig. 4 | Sex-specific expression changes across organs.
a, Smoothed line plot displaying the number of DEGs between female and
male mice at each age. Positive (negative) values represent upregulated
(downregulated) genes. Grey lines represent all other tissues. b, Heat map
representation of a. c, Expression of Apoe mRNA in GAT and Axin2 mRNA in
spleen. The black line indicates LOESS regression. n = 45 (GAT) and n = 47
(spleen) independent samples. d, Venn diagrams depicting the overlap of DEGs
in female and male mice detected at 3 months and 18 months of age in GAT,
SCAT, liver and kidney. One-sided Fisher’s exact test, ***P < 0.0001. e–h, Top 10
GO terms enriched among the DEGs between female and male mice at 18
months of age in GAT (e), SCAT (f), liver (g) and kidney (h). Data are mean ± s.e.m.
n = 2 (female) and n = 4 (male) independent mice for each organ. q values were
estimated using the Benjamini–Hochberg procedure for each database
separately, and for GO classes (molecular function, cellular component,
biological process) independently.
Extended Data Fig. 5 | Organ-specific gene expression dynamics. For each (columns 2–6) for further analysis. Average trajectories for each cluster ± s.d.
of the 17 organs (rows), the average trajectory of the 15,000 most highly are represented.
expressed genes is represented in column 1. Five clusters were used
Article
Extended Data Fig. 6 | Pathway enrichment analysis of organ-specific
clusters. Clusters from Extended Data Fig. 5 show enrichment for genes in
functional categories. Pathway enrichment was tested using GO, Reactome
and KEGG databases. Enrichment was tested using Fisher’s exact test (GO) and
the hypergeometric test (Reactome and KEGG). The top five pathways for each
cluster are shown. q values were estimated with the Benjamini–Hochberg
procedure for each database separately, and for GO classes (molecular
function, cellular component, biological process) independently. Sample size
per cluster or tissue is indicated in Extended Data Fig. 5.
Extended Data Fig. 7 | Analysis of cytokines and transcription factors. transcription factors (TRANSFAC database; n = 334 genes). Thicker lines
a, Age-related changes in inflammatory cytokine and chemokines outlined with white represent the average trajectory for each cluster ± s.d.
(cytokine-mediated signalling pathways GO:0019221; n = 501 genes), and b, c, Spearman correlation coefficient for ageing genes in a.
Article

Extended Data Fig. 8 | See next page for caption.

Extended Data Fig. 8 | Integration of bulk and single-cell transcriptomic
data. a, b, Representative GO terms enriched among the genes with highly
disperse (a) and cell-specific (b) expression patterns. n = 1,108 cells. q values
were estimated using the Benjamini–Hochberg procedure for each database
separately, and for GO classes (molecular function, cellular component,
biological process) independently. c, Expression of Aco2 mRNA in the kidney.
Black line shows LOESS regression; ρ, Spearman’s rank correlation coefficient.
n = 52 independent samples. Data are mean ± s.e.m. d, e, t-SNE visualization of
scRNA-seq data (FACS) from the kidney, coloured by expression of Aco2 (d) and
Cs (e). n = 1,108 cells. f, Violin plot representing expression of Aco1 and Aco2
across all profiled cell types in the kidney. Points indicate cell-wise expression
levels and the violin indicates average distribution split by age. Significance
was assessed using a Student’s t-test. n = 325 cells (3 months) and 783 cells
(24 months). g, Spearman’s rank correlation for cell type fractions significantly
(P < 0.05) changing with age, based on deconvolution with FACS or droplet
scRNA-seq expression signatures. n = 38 (FACS, BAT), n = 37 (droplet, GAT),
n = 37 (FACS, GAT), n = 34 (droplet, kidney), n = 35 (FACS, kidney), n = 35 (droplet,
liver), n = 35 (FACS, liver), n = 37 (droplet, lung), n = 37 (FACS, lung), n = 38
(droplet, marrow), n = 36 (FACS, marrow), n = 38 (droplet, MAT), n = 39 (FACS,
MAT), n = 34 (droplet, pancreas), n = 32 (FACS, pancreas), n = 37 (droplet, SCAT),
n = 38 (FACS, SCAT), n = 35 (droplet, skin), n = 33 (FACS, skin), n = 36 (droplet,
spleen), n = 37 (FACS, spleen) independent samples. h, Pairwise comparisons
cell fractions between scRNA-seq (FACS), scRNA-seq (droplet), FACS-based
bulk RNA-seq deconvolution and droplet-based bulk RNA-seq deconvolution.
Each point represents an individual cell type in an individual tissue type.
Article

Extended Data Fig. 9 | See next page for caption.

Extended Data Fig. 9 | Identifying Ig jhigh B cells with FACS and droplet
scRNA-seq. a, t-SNE visualization of all Cd79a-expressing cells present in the
Tabula Muris Senis FACS dataset (17 tissues). Coloured clusters as identified
with the Seurat software toolkit. Ig jhigh B cell cluster 11 is highlighted. n = 10,867
cells. b, t-SNE in a coloured by the Ig jhigh B cell markers Ig j, Xbp1 and Derl3. c, GO
terms enriched among the top 300 marker genes of Ig jhigh (n = 129 cells) versus
Ig jlow(n = 10,738 cells) (FACS). q values were estimated using the Benjamini–
Hochberg procedure for each database separately, and for GO classes
(molecular function, cellular component, biological process) independently.
d, Distribution of Ig jhigh as percentages of Cd79a-expressing cells per tissue.
e, Percentage of Ig jhigh B cells of all Cd79a-expressing cells across all tissues.
n = 5 (3 months) and n = 4 (24 months) independent mice. Significance was
assessed using a Student’s t-test. Data are mean ± s.e.m. f, t-SNE visualization of
all Cd79a-expressing cells present in the Tabula Muris Senis droplet dataset
(17 tissues). Coloured clusters are as identified with the Seurat software toolkit.
IgJhigh B cell cluster 5 is highlighted. n = 23,796 cells. g, t-SNE in f coloured by the
B cell marker Cd79a and Ig jhigh B cell marker Derl3. h, Percentage of Ig jhigh B cells
of all Cd79a-expressing cells across all tissues. i, Heat map of the z-transformed
Ig j expression trajectories across bone (n = 54), marrow (n = 51), spleen (n = 54),
liver (n = 50), GAT (n = 52), kidney (n = 52), heart (n = 52) and muscle (n = 52).
j, Change in Ig j mRNA expression in human visceral fat (aged 20–29, n = 25; aged
50–59, n = 124; aged 70–79, n = 12) and subcutaneous fat (aged 20–29, n = 32;
aged 50–59, n = 149; aged 70–79, n = 13) (data from GTEx consortium). In the
box plot, the median is represented by the centre line and the box boundaries
represent the first and third quartiles. k, Number of Ig jhigh B cells with a
successfully assembled B cell receptor locus, split by mouse and
immunoglobulin class. l, Clonally amplified Ig jhigh B cells as detected in
mouse 1 and mouse 3, grouped by tissue of origin (colour) and immunoglobulin
class (shape).
Article

Extended Data Fig. 10 | STRING analysis of top correlating plasma proteins. a, The top seven plasma proteins correlated with gene expression in muscle,
coloured by pathway. b, The top 25 plasma proteins correlated with gene expression in any organ.
 nature research | reporting summary
Tony Wyss-Coray, Stephen R. Quake, Andreas
Corresponding author(s): Keller

Reporting Summary
Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Research policies, see Authors & Referees and the Editorial Policy Checklist.

Statistical parameters
When statistical analyses are reported, confirm that the following items are present in the relevant location (e.g. figure legend, table legend, main
text, or Methods section).
n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
An indication of whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.

A description of all covariates tested

A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistics including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND
variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)

For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.

For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated

Clearly defined error bars

State explicitly what error bars represent (e.g. SD, SE, CI)

Our web collection on statistics for biologists may be useful.

Software and code

Policy information about availability of computer code
Data collection FACS data was collected with BD FACS Diva version 8. Sequencing data was collected with the NovaSeq Control Software version 1.6.0.

Data analysis FACS data was analyzed with FlowJo version 10.6.1

Sequences from the NovaSeq were de-multiplexed using bcl2fastq version 2.20. Reads were aligned to the GRCm39.p6 (mus musculus)
genome with Gencode v.M19 annotations using using STAR version 2.5.2b with parameters TK. Gene counts were produced using HTSEQ
version 0.6.1p1 with default parameters, except “stranded” was set to “false”, and “mode” was set to “intersection-nonempty”.

Seurat version 3
String version 11.0
Kohonen R package (version 3.0.8)
April 2018

Rtsne version 0.15

topGO version 2.36
destiny version 2.14
org.Mm.eg.db version 3.8.2
clusterProfiler version 3.12
car version 3.0-4
DESeq2 version 1.26.0
CellPlot version 1.0

1
 pvca version 1.18.0
biostrings version 2.42.1

nature research | reporting summary

Yarn version 1.10
CiberSortX web-interface accessed 2019/11/19
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors/reviewers
upon request. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

Data
Policy information about availability of data
All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
- A list of figures that have associated raw data
- A description of any restrictions on data availability
Raw data are available on GEO (GSE132040).

Field-specific reporting
Please select the best fit for your research. If you are not sure, read the appropriate sections before making your selection.
Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/authors/policies/ReportingSummary-flat.pdf

Life sciences study design

All studies must disclose on these points even when the disclosure is negative.
Sample size No statistical methods were used to pre-determine sample size. The number of mice was restricted due to costs and practical constraints of
collecting and processing organs. However, the study is designed such that there are many closely related experimental groups (up to 10
different ages in close proximity across the lifespan), allowing for pooling of closely related ages, and for analyzes across the lifespan that
include >50 samples. Additionally, as this is the largest study of its kind, so far, we believe the sample size is sufficient.

Data exclusions Exclusion criteria were pre-establish. Bulk: samples with fewer than 4 million uniquely mapped reads were excluded.

Preprocessed data from Tabula Muris Senis was filtered as follows: We removed genes not expressed in at least 3 cells and then cells that did
not have at least 250 detected genes. For FACS we removed cells with less than 5000 counts and for droplet cells with less than 2500 UMIs.

FACS single-cell adipose data was filtered as follows: cells with less than 200 or more than 6,500 genes were excluded.

Replication We have not replicated the RNA-sequencing findings in this study, except in the case of IgJ-high plasma cells, which were verified with RNA-
scope and FACS. All attempts (that were technically successful) at RNAscope and FACS for plasma cells successfully replicated the age-related
increase.

Randomization Randomization was performed in the case of mouse dissection order and during the preparation of 96-well plates for cDNA creation.

Blinding Blinding was not applicable for the transcriptomic data analysis. Blinding was performed for the RNAscope and FACS analysis.

Reporting for specific materials, systems and methods

Materials & experimental systems Methods

n/a Involved in the study n/a Involved in the study
Unique biological materials ChIP-seq
April 2018

Antibodies Flow cytometry

Eukaryotic cell lines MRI-based neuroimaging
Palaeontology
Animals and other organisms
Human research participants

2
Antibodies

nature research | reporting summary

Antibodies used Cd138-PE (1:400; Biolegend142504, Clone 281-2, Lot B246403), Cd19-BV421 (1:200; Biolegend 115538, Clone 6D5, Lot
B263601), B220-APC (1:100; Biolegend 103212, Clone RA3-6B2, Lot B271008)

Validation Antibodies were validated for murine flow cytometry:

CD138-PE and CD19-BV421

Pracht, K. et al. A new staining protocol for detection of murine antibody-secreting plasma cell subsets by flow cytometry. Eur. J.
Immunol. 47, 1389–1392 (2017).

B220-APC
Wilmore, J. R., Jones, D. D. & Allman, D. Protocol for improved resolution of plasma cell subpopulations by flow cytometry. Eur. J.
Immunol. 47, 1386–1388 (2017).

Additional sources include:

Biolegend142504:
Jalkanen M, et al. 1985. J. Cell. Biol. 101:976. (FC)
Miettinen H, et al. 1994. J. Cell. Sci. 107:1571. (IF)
Li Q, et al. 2002. Cell 111:635. (IF, IHC)
McCarthy BA, et al. 2012. BMC Cancer. 12:203. (IHC)

Biolegend 115538
Shoham T, et al. 2003. J. Immunol. 171:4062. (FC)
Goodyear CS, et al. 2004. J. Immunol. 172:2870. (FC)
Kamimura D, et al. 2006. J. Immunol. 177:306. (FC)
Andoniou CE, et al. 2005. Nat. Immunol. 6:1011. (FC)
Lawson BR, et al. 2007. J. Immunol. 178:5366. (FC)
Phan TG, et al. 2007. Nat. Immunol. 8:992. (FC)
Hayashida K, et al. 2008. J. Biol. Chem. 283:19895. (IF) PubMed
Charles N, et al. 2010. Nat. Med. 16:701. (FC) PubMed
Bankoti J, et al. 2010. Toxicol. Sci. 115:422. (FC) PubMed
Stadnisky MD, et al. 2011. Blood. 117:5133. (FC) PubMed
Perlot T, et al. 2012. J. Immunol. 188:1201. (FC) PubMed
Olive V, et al. 2013. Elife. 2:822. PubMed

Biolegend 103212
Coffman RL. 1982. Immunol. Rev. 69:5. (IP)
George A, et al. 1994. J. Immunol. 152:1014. (Activ)
Asensi V, et al. 1989. Immunology 68:204. (Activ)
Domiati-Saad R, et al. 1993. J. Immunol. 151:5936. (Activ)
Hata H, et al. 2004. J. Clin. Invest. 114:582. (IHC)
Monteith CE, et al. 1996. Can. J. Vet. Res. 60:193. (IHC)
Shih FF, et al. 2006. J. Immunol. 176:3438. (FC)
Chang C L-T, et al. 2007. J. Immunol. 178:6984.
Fazilleau N, et al. 2007. Nature Immunol. 8:753.
Lang GL, et al. 2008. Blood 111:2158. PubMed
Charles N, et al. 2010. Nat. Med. 16:701. (FC) PubMed
del Rio ML, et al. 2011. Transpl. Int. 24:501. (FC) PubMed

Animals and other organisms

Policy information about studies involving animals; ARRIVE guidelines recommended for reporting animal research
Laboratory animals C57BL/6JN males (n=4, aged 1, 3, 6, 9, 12, 15, 18, 21, 24, 27 months; and females (n=2, ages 1, 3, 6, 9, 12, 15, 18, 21 months)

Wild animals This study did not involve wild animals

Field-collected samples This study did not involve field-collected samples

April 2018

3
Flow Cytometry

nature research | reporting summary

Plots
Confirm that:
The axis labels state the marker and fluorochrome used (e.g. CD4-FITC).
The axis scales are clearly visible. Include numbers along axes only for bottom left plot of group (a 'group' is an analysis of identical markers).
All plots are contour plots with outliers or pseudocolor plots.
A numerical value for number of cells or percentage (with statistics) is provided.

Methodology
Sample preparation Cell suspensions for fluorescence activated cell sorting (FACS) of plasma cells were generated from gonadal adipose tissue (GAT)
and kidney. Following cardiac perfusion with PBS, tissues were immediately dissected and minced to a paste with scissors.
Samples were resuspended in 40ml buffer (GAT: 10% horse serum in F12; kidney 2% FBS in RPMI) and passed through a 100um
filter into a 50ml conical tube, grinding with a syringe plunger to further dissociate clumps. Filters were washed with 5ml buffer,
and all 45ml was then passed through a 35um filter, washing with another 5ml buffer. Filtered cells were then pelleted (500 x g,
5 minutes, 4°C) and treated with 10ml ACK lysis buffer for 5 minutes at room temperature. After pelleting, washing with 5ml
buffer, and pelleting again, cells were resuspended in 1ml FACS buffer (2% FBS in PBS), and aliquoted through 35um strainer
caps into FACS tubes. Cells were pelleted and resuspended in the antibody cocktail: Cd138-PE (1:400; Biolegend142504), Cd19-
BV421 (1:200; Biolegend 115538), B220-APC (1:100; Biolegend 103212) for 30min on ice. After pelleting and resuspending, cells
were stained with 1ul 1:1000 Sytox Green (Thermo S7020) immediately before sorting on a BD FACS Aria III. Gates were set to
capture live singlets and reduce debris according to standard procedures, and Cd138high cells were quantified.

Instrument BD FACS Aria III

Software FACS data was collected with BD FACS Diva version 8

FACS data was analyzed with FlowJo version 10.6.1

Cell population abundance Plasma cell abundance in young marrow: 0.007-0.023%; old marrow: 0.1-0.35%.
Plasma cell abundance in young kidney: 0-0.012%; old kidney: 0.041-0.19%

Gating strategy FSC-A vs. SSC-A - gated on cells

FSC-A vs. FSC-H - gated for singlets
SytoxGreen-FITC vs. SSC-A - gated for live cells on SytoxGreen-FITC
CD138-PE vs. B220-APC - gated positive on CD138-PE > 10^4

Tick this box to confirm that a figure exemplifying the gating strategy is provided in the Supplementary Information.

April 2018