Clonal dynamics of haematopoiesis across the human lifespan
Emily Mitchell
1,2,3
, Michael Spencer Chapman
1,#
, Nicholas Williams
1,#
, Kevin Dawson
1,#
,
Nicole Mende
2
, Emily F Calderbank
2
, Hyunchul Ju ng
1
, Thomas Mitchell
1
, Tim Coorens
1
, David
Spencer
4
, Heather Machado
1
, Henry Lee-Six
1
, Megan Davies
5
, Daniel Hayler
2
, Margarete
Fabre
1,2,3
, Krishnaa Mahbubani
6,7
, Fede Abascal
1
, Alex Cagan
1
, George Vassiliou
1,2,3
, Joanna
Baxter
3
, Inigo Martincorena
1
, Michael R Stratton
1
, David Kent
8
, Krishna Chatterjee
9
, Kourosh
Saeb Parsy
6,7
, Anthony R Green
2,3
, Jyoti Nangalia
1,2,3*
, Elisa Laurenti
2,3*
, Peter J Campbell
1,2*
.
#
These authors contributed equally.
*
These authors contributed equally.
Affiliations
(1)
Wellcome Sanger Institute, Hinxton, CB10 1SA, UK.
(2)
Wellcome-MRC Cambridge Stem Cell Institute, Cambridge Biomedical Campus,
Cambridge, CB2 0AW, UK.
(3)
Department of Haematology, University of Cambridge, Cambridge, CB2 2XY, UK.
(4)
Department of Medicine, McDonnell Genome Institute, Washington University, St. Louis,
MO, USA.
(5)
Cambridge Molecular Diagnostics, Milton Road, Cambridge, CB4 0FW, UK.
(6)
Department of Surgery, University of Cambridge, Cambridge, CB2 0QQ, UK.
(7)
Cambridge Biorepository for Translational Medicine, NIHR Cambridge Biomedical
Research Centre, University of Cambridge, Cambridge CB2 2XY, UK.
(8)
York Biomedical Research Institute, Department of Biology, University of York, York,
YO10 5DD, UK.
(9)
Wellcome Trust-MRC Institute of Metabolic Science, University of Cambridge, Cambridge,
CB2 0QQ, UK.
Address for correspondence:
Dr Peter Campbell,
Cancer, Ageing & Somatic Mutation Programme,
Wellcome Sanger Institute,
Hinxton CB10 1SA,
United Kingdom
e-mail: pc8@sanger.ac.uk
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprintthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.08.16.456475doi: bioRxiv preprint
Abstract
Age-related change in human haematopoiesis causes reduced regenerative capacity
1
,
cytopenias
2
, immune dysfunction
3
and increased risk of blood cancer. The cellular
alterations that underpin the abruptness of this functional decline after the age of 70 years
remain elusive. We sequenced 3579 genomes from single-cell-derived colonies of
haematopoietic stem cell/multipotent progenitors (HSC/MPPs) across 10 haematologically
normal subjects aged 0-81 years. HSC/MPPs accumulated 17 mutations/year after birth and
lost 30bp/year of telomere length. H aematopoiesis in adults aged <65 was massively
polyclonal, with high indices of c lona l diversity and a stable population of 20,000–200,000
HSC/MPPs contributing evenly to blood production. In contrast, haematopoiesis in
individuals aged >75 showed profoundly decreased clonal diversity. In each elderly su bject,
30-60% of haematopoiesis was accounted for by 12-18 independent clones, each
contributing 1-34% of blood production. Most clones had begun their expansion before age
40, but only 22% had known driver mutations. Genome-wide selection analysis estimated
that 1/34 to 1/12 non-synonymous mutations were drivers, occurring at a constant rate
throughout life, affecting a wider pool of genes than identified in blood cancers. Loss of Y
chromosome conferred selective benefits on HSC/MPPs in males. Simulations from a simpl e
model of haematopoiesis, with consta nt HSC population size and constant acquisition of
driver mutations conferring moderate fitness benefits, entirely explained the abrupt change
in clonal structure in the elderly. Rapidly decreasing clonal diversity is a universal feature of
haematopoiesis in aged humans, underpinned by pervasive positive selection acting on
many more genes than currently identified.
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprintthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.08.16.456475doi: bioRxiv preprint
Introduction
The age-related mortality curve for modern humans is an extreme outlier among species
across the tree of life, with an abrupt increase in standardised mortality rates after the
average lifespan
4
, leading to surprisingly low variance in age at death
5
. Behavioural, medical
and environmental interventions have driven down early and extrinsic causes of mortality,
thereby unmasking a substantial burden of intrinsic, age-associated disease. Studies of
ageing at the cellular level have demonstrated that accumulation of molecular damage
across the lifespan is gradual and lifelong, including telomere attrition
6–8
, somatic
mutation
9–11
, epigenetic change
12
and oxidative or replicative stress
13,14
. It remains
unresolved how such gradual accumulation of molecular damage can translate into an
abrupt increase in mortality after the age of 70 years.
The haematopoietic system is an interesting organ for studying human ageing. It manifests
several age-associated phenotypes including anaemia; loss of regenerative capacity,
especially in the face of insults such as infection, chemotherapy or blood loss; and increased
risk of blood cancer. One aspect of age-related change in human haematopoietic stem cells
(HSCs) that has been an area of intense study is ‘clonal haematopoiesis’
15
. This is defined by
single-cell-derived expansions in blood, most commonly associated with mutations in genes
recognised as causal for myeloid neoplasms, so-called driver mutations. Clonal
haematopoiesis increases with age, r eaching 10-20% prevalence
16–20
or even higher
21
after
70 years. Data are now emerging that many elderly individuals have evidence of clonal
expansions even in the absence of known driver mutations
22–24
. Cellular changes of ageing
have primarily been studied in mice
25
. While the numbers of phenotypic HSCs increase in
aged mice, aged HSCs produce fewer mature progeny both
in vitro
and in serial
transplantation compared to young HSCs
26–28
.
HSCs acc umulate somatic mutations linearly w ith age, with each cell acquiring a new
mutation every 2-3 weeks on average throughout life
10,29
. Whole-genome sequencing of
colonies grown from single cells enables comprehensive identification of these somatic
mutations and reconstruction of lineage relationships among cells
30
(
Fig. 1a
). We used this
approach to study the clonal dynamics of human haematopoiesis across the lifespan,
sequencing whole genomes of 3579 single-cell-derived colonies from 10 healthy individuals.
Whole genome sequencing of HSC-derived colonies
We obtained samples from 10 individuals, with no known haematological disease, aged
between 0 and 81 years (
Table S1
). One subject had inflammatory bowel disease treated
with azathioprine (KX002, 38-year male) and one had selenoprotein deficiency
31
, a genetic
disorder not known to impact HSC dynamics (SX001, 48 year male). The source of stem cells
was cord blood (CB) for the two neonates, and bone marrow and/or peripheral blood for
adult donors (
Fig. 1b
). Bone marrow samples were obtained peri-mortem, allowing
sampling of large volumes (50-80mL) from multiple vertebrae.
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprintthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.08.16.456475doi: bioRxiv preprint
For all individuals , single phenotypic haematopoietic stem cell/multipotent progenitors
(HSC/MPPs: Lin-, CD34+, CD38-, CD45RA-) were flow-sorted
32
and cultured (
Extended Fig.
1a
). Overall, 42-89% of sorted HSC/MPPs produced colonies, meaning that the colonies
were a representative sample of the HSC/MPP population in each individual (
Extended Fig.
1b
). For four individuals, haematopoietic progenitor cells (HPCs) (Lin-, CD34+, CD38+) were
also isolated for comparison of mutation burden between HSCs and HPCs.
We performed whole-genome sequencing at an average sequencing depth of 14x on 224-
453 colonies per individual (
Fig. 1b
). We excluded 17 colonies with low coverage, 34
technical duplicates and 7 colonies derived from more than a single cell (
Extended Fig. 2a-
c
). The final dataset comprised whole genomes from 3579 colonies, of which 3361 were
single HSC /MP P-derived and 218 were single HPC- derived. Raw mutation burdens were
corrected for sequencing depth using asymptotic regression (
Extended Fig. 2d
). Single base
substitution sp ectra were consistent with published results
29,30,33
(
Extended Fig. 2e-f
).
Phylogenetic trees for each adult individual were constructed from the patterns of shared
and unique somatic mutations (
Extended Fig. 3a-d
). Benchmarking of the phylogenies
included an assessment of their internal consistency, stability across phylogenetic inference
algorithms, and robustness to bootstrapping approaches (
Supplementary Fig. 3
). All code
for the variant filtering, signatures, phylogenetic reconstruction, benchmarking and
downstream analyses are available (
Supplementary Code
).
Mutation burden and telomere lengths
Consistent with published data
29,34
, point mutations accumulated linearly in HSC/MPPs
throughout life, at a rate of 16.8 substitutions/cell/year (CI
95%
=16.5-17.1;
Fig. 1c
) and 0.71
indels/cell/year (CI
95%
=0.65-0.77;
Fig. 1d
). We found no significant difference in mutation
burden between HSC/MPPs and HPCs (
Extended Fig. 4a
). Structural variants were rare, with
only 1-17 events observed in each individual, mostly deletions, c o rrelating with age (
Fig. 1e
;
Extended Fig. 4b
;
Table S3
). Autosomal copy number aberrations were rare at all ages and
comprised either copy-neutral loss-of-heterozygosity events or tetrasomies (
Fig. 1f
). In
contrast, Y chromosome copy number changes were frequent in males, increasing with age
as previously shown
35
(
Fig. 1g
).
We estimated telomere lengths for the 1505 HSC/MPP colonies from 7 individuals
sequenced on Hiseq X10
36
(
Fig. 1h
). As previously reported
8,37,38
, telomere lengths
decreased steadily with age, at an average attrition rate of 30.8 bp/ year in adult life
(CI
95%
=13.2-48.4), which is close to published estimates of 39 bp/year from bulk
granulocytes
8
. By sequencing single-cell-derived colonies, we can estimate the variance and
distribution in telomere lengths among cells with a resolution not possible for bulk
populations. In cord blood and adults aged <65, a small proportion of HSC/MPPs had
unexpectedly long telomeres, as assessed using several criteria for outliers, a proportion
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprintthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.08.16.456475doi: bioRxiv preprint
that reduced in frequency with age (
Extended Fig. 4c
). Given that telomeres shorten at cell
division, these ou tlier cells have presumably undergone fewer historic cell divisions. A rare
population of dormant HSCs dividing infrequently has been described in the mouse
39–41
and
our telomere data would be consistent with a n analogous population in humans, especially
children and young adults.
Qualitative change in HSC population structure after 70 years of age
The phylogenetic trees generated here depict the lineage relationships among ancestors of
the stem and progenitor cells sequenced. Given the consistent, linear rate of mutation
accumulation across the lifespan, we can scale the raw phylogeneti c trees (
Extended Fig.
3d
) to chronological time to study clonal dynamics of HSC/MPPs across the lifespan (
Figs 2-
3
). Branch-points in the tree, so-called ‘coalescences’, define historic phylogenetic
relationships between stem cells. The chance of c apturing coalescent events in a phylogeny
of randomly sampled stem cells depends on both the population size and time between
symmetric self-renewal divisions. In taxonomy, a ‘clade’ is defined as the group of organisms
descended from a single common ancestor – in the context of somatic cells, this represents
a clone, and its size can be estimated from the fraction of colonies derived from that
ancestor. Here, we define an ‘expanded clade’ as a post-natal ancestral lineage whose
descendants contributed >1% of colonies at the time of sampling.
Phylogenetic trees of the 4 adults aged <65 showed that healthy young and middle-aged
haematopoiesis is highly polyclonal (
Fig. 2
). Despite sequencing 361-408 colonies per
individual, we found at most a single expanded clade in each sample, and the two we did
observe contributed <2% of a ll haematopoiesis in those individuals. Known driver point
mutations were sparse, with 2 in
DNMT3A
and 1 possible oncogenic variant each in
CBL
and
JAK2
, none of which occurred in the two expanded clades.
Phylogenetic trees for the 4 adults aged >70 were qualitatively different (
Fig. 3
), with an
oligoclonal pattern of haematopoiesis. In each elderly individual, we found 12-18
independent clones established between birth and the age of 40 that each contributed
between 1% and 34% of colonies sequenced, most in the 1-3% range. Collectively, these
clones summed to a significant proportion of all blood production in ou r elderly research
subjects – between 32% and 61% of all colonies sequenced derived from the expanded
clades.
Most of the oligoclonal blood production in the elderly individuals could not be accounted
for by known driver mutations. Although mutations in
DNMT3A
,
TET2
and
CBL
did occur in
some expansions, mutations in the top 17 myeloid driver genes could only explain 10/58
expanded clades with a clonal fraction >1%. Only 3 additional mutations were identified if
we extended to a wider set of 92 genes implicated in myeloid neoplasms
42
(
Table S5
), still
leaving 45 clonal expansions unexplained.
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprintthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.08.16.456475doi: bioRxiv preprint