RESEARCH ARTICLE

Human Bone Marrow Organoids for Disease

Modeling, Discovery, and Validation of
Therapeutic Targets in Hematologic
Malignancies
Abdullah O. Khan1,2, Antonio Rodriguez-Romera2, Jasmeet S. Reyat1, Aude-Anais Olijnik2,
Michela Colombo2, Guanlin Wang2,3, Wei Xiong Wen2,3, Nikolaos Sousos2,4, Lauren C. Murphy2,
Beata Grygielska1, Gina Perrella1, Christopher B. Mahony5, Rebecca E. Ling6, Natalina E. Elliott6,
Christina Simoglou Karali2, Andrew P. Stone7, Samuel Kemble5, Emily A. Cutler8, Adele K. Fielding8,

Downloaded from [Link] by guest on 13 February 2023

Adam P. Croft5, David Bassett9, Gowsihan Poologasundarampillai10, Anindita Roy6, Sarah Gooding2,4,
Julie Rayes1, Kellie R. Machlus7, and Bethan Psaila2,4

Illustration by Bianca Dunn

 ABSTRACT A lack of models that recapitulate the complexity of human bone marrow has
hampered mechanistic studies of normal and malignant hematopoiesis and the
validation of novel therapies. Here, we describe a step-wise, directed-differentiation protocol in which
organoids are generated from induced pluripotent stem cells committed to mesenchymal, endothelial,
and hematopoietic lineages. These 3D structures capture key features of human bone marrow—
stroma, lumen-forming sinusoids, and myeloid cells including proplatelet-forming megakaryocytes.
The organoids supported the engraftment and survival of cells from patients with blood malignancies,
including cancer types notoriously difficult to maintain ex vivo. Fibrosis of the organoid occurred
following TGFβ stimulation and engraftment with myelofibrosis but not healthy donor–derived cells,
validating this platform as a powerful tool for studies of malignant cells and their interactions within
a human bone marrow–like milieu. This enabling technology is likely to accelerate the discovery and
prioritization of novel targets for bone marrow disorders and blood cancers.

SIGNIFICANCE: We present a human bone marrow organoid that supports the growth of primary cells

Downloaded from [Link] by guest on 13 February 2023

from patients with myeloid and lymphoid blood cancers. This model allows for mechanistic studies of
blood cancers in the context of their microenvironment and provides a much-needed ex vivo tool for
the prioritization of new therapeutics.
See related commentary by Derecka and Crispino, p. 263.

Advances have been made in modeling certain marrow

INTRODUCTION components on “biochips” (12–16), but the lack of spe-
The specialized bone marrow microenvironment main- cialized stroma, vascularization, and active blood cell gen-
tains and regulates hematopoiesis, enabling an adequate eration remains a limitation with these methods. Coculture
supply of blood cells to meet changing physiologic require- with bone marrow mesenchymal stromal cells (MSC) can
ments throughout life. Perturbations in the bone marrow support hematopoietic cell growth, either in 2D or in 3D
hematopoietic niche contribute to the initiation and propa- with the addition of extracellular matrix support (17–19).
gation of hematologic malignancies. In addition, the stromal Although beneficial, these approaches are limited in terms
remodeling that occurs as a consequence of blood cancers of the elements of the bone marrow that are incorporated,
contributes to bone marrow failure (1–4). Modeling bone as well as their scalability and reproducibility due to the
marrow dysfunction is challenging, particularly in the con- limited availability and interdonor variability of primary
text of human diseases. In vitro studies are often limited to bone marrow MSCs. Improved in vitro systems are there-
2D systems and simple cocultures, in which the relevant cell fore required to enable more detailed mechanistic studies
types are absent, and many human diseases are inadequately of human hematopoiesis, and to allow for the functional
reproduced by mouse models. Patient-derived xenografts interrogation of the pathways and cross-talk that drive bone
have been used to model disease and validate targets in vivo, marrow malignancies.
but some malignancies and hematologic cell subtypes do The development and application of organoids—self-
not engraft well, even when humanized murine models are organizing, 3D, living multilineage structures—have the
used (5–11). potential to facilitate translational research by enabling

1
Institute of Cardiovascular Sciences, College of Medical and Dental of Dentistry, Institute of Clinical Sciences, University of Birmingham,
Sciences, University of Birmingham, Vincent Drive, Birmingham, United Birmingham, United Kingdom.
Kingdom. 2MRC Weatherall Institute of Molecular Medicine, Radcliffe Corresponding Authors: Bethan Psaila, Department of Haematology,
Department of Medicine and National Institute of Health Research (NIHR) MRC Weatherall Institute of Molecular Medicine, University of Oxford,
Oxford Biomedical Research Centre, University of Oxford, Oxford, United Oxford, UK. Phone: 44(0)1865 222340; E-mail: [Link]@[Link];
Kingdom. 3Centre for Computational Biology, MRC Weatherall Institute Abdullah O. Khan, Institute of Cardiovascular Sciences, College of Medical
of Molecular Medicine, University of Oxford, Oxford, United Kingdom. and Dental Sciences, University of Birmingham, Birmingham, UK. Phone: 44
4
Cancer and Haematology Centre, Churchill Hospital, Oxford University (0)7852 311295; E-mail: [Link].4@[Link]; and Kellie R. Machlus,
Hospitals NHS Foundation Trust, Oxford, United Kingdom. 5Rheumatology Vascular Biology Program, Boston Children’s Hospital, Department of
Research Group, Institute of Inflammation and Ageing, College of Surgery, Harvard Medical School, Boston, MA. Phone: 609-217-6618;
Medical and Dental Sciences, University of Birmingham, Birmingham, E-mail: [Link]@[Link]
United Kingdom. 6MRC Weatherall Institute of Molecular Medicine,
Department of Paediatrics and National Institute of Health Research Cancer Discov 2023;13:364–85
(NIHR) Oxford Biomedical Research Centre, University of Oxford, Oxford, doi: 10.1158/[Link]-22-0199
United Kingdom. 7Vascular Biology Program, Boston Children’s Hospital, This open access article is distributed under the Creative Commons Attribution
Department of Surgery, Harvard Medical School, Boston, Massachusetts. 4.0 International (CC BY 4.0) license.
8
University College London Cancer Institute, London, United Kingdom.
9
Healthcare Technologies Institute, School of Chemical Engineering, ©2022 The Authors; Published by the American Association for Cancer Research
University of Birmingham, Birmingham, United Kingdom. 10School

FEBRUARY 2023 CANCER DISCOVERY | 365

RESEARCH ARTICLE Khan et al.

genetic screens and pharmacologic modulation of disease and progenitor cell (HSPC; CD34+ Lin−), myelomonocytic
pathobiology (20). Our goal was to generate a vascular- (CD34− CD11b+ Lin+), megakaryocyte (CD34− Lin− CD41+
ized human bone marrow–like organoid that contains key CD42b+), endothelial (CD31+ CD144+), erythroid (CD34−
hematopoietic niche elements and supports active endog- Lin− CD71+ CD235a+), and MSC (CD31− CD140b+ VCAM1+
enous hematopoiesis, as well as the growth and survival of LepR+) populations were detected when organoids were
hematopoietic cells from adult donors, including malig- digested at day 18 of differentiation (Fig. 1C–E). Organoids
nant cell types that are difficult to grow and study ex vivo. developed in collagen type I–only hydrogels contained the
Such a system would offer a scalable and highly manip- highest fraction of HSPCs but a low proportion of myelo-
ulable human model for mechanistic studies and drug monocytic cells and megakaryocytes, and no MSCs, whereas
development and, importantly, may reduce dependence on collagen type IV and collagen type I  +  IV hydrogel–derived
animal models. organoids yielded significantly larger myeloid and MSC
To achieve this, we optimized a protocol in which human populations, indicating that the addition of collagen IV cre-
induced pluripotent stem cells (iPSC) generate mesenchymal ated more favorable conditions for multilineage differentia-
elements, myeloid cells, and “sinusoidal-like” vasculature in tion (Fig. 1D and E; Supplementary Fig. S1A).
a format that resembles the cellular, molecular, and spatial The hematopoietic niche contains a dense network of sinu-
architecture of myelopoietic bone marrow. We confirmed soidal vessels, and specialized sinusoidal endothelium regulates
the homology of these organoids to human bone mar- stem cell self-renewal and differentiation, progenitor matura-

Downloaded from [Link] by guest on 13 February 2023

row using multimodal imaging approaches and single-cell tion, and platelet generation (24, 25). To assess vasculariza-
RNA sequencing (scRNA-seq). Crucially, in addition to mod- tion of the organoids, we measured the degree of endothelial
eling physiologic hematopoietic cell–niche interactions, we sprouting. Vascular sprout radii were significantly smaller in
showed that the organoids supported the engraftment and collagen type IV–only hydrogels compared with collagen types
survival of healthy and malignant hematopoietic cells from I–only and I  +  IV (average radius of 203.6  ±  14.61 μm vs.
human donors, and enabled the screening of inhibitors 345.9 ± 32.11 μm vs. 476.3 ± 28.82 μm, respectively; Fig. 1F),
of bone marrow fibrosis, a complication that occurs in and the density of CD34 and CD144 [Vascular Endothelial
patients with certain blood cancers and is associated with (VE)-Cadherin]–positive sprouts was significantly lower in col-
poor prognosis. lagen IV–only hydrogels (Fig.  1G). To assess the vasculariza-
This platform addresses a long-standing need for 3D tion and cellular architecture of the whole organoid in 3D,
human bone marrow models for translational research in organoids were mounted in agarose and cleared with Ethyl
which both niche and hematopoietic components are spe- Cinnamate for confocal Z-stack imaging. This revealed an
cies and cell context specific and creates a dynamic platform elaborate network of UAE1+ CD144+ vessels throughout the
for high-throughput drug screening and studies aiming to collagen I  +  IV–derived organoids (Fig.  1H–J). In addition to
understand disease pathways. CD34+ HSPCs (Fig. 1H), CD41+ megakaryocytes (Fig. 1I) and
CD71+ erythroid cells (Fig. 1J) were distributed throughout the
RESULTS organoid volume and closely associated with the endothelium,
as occurs in the native bone marrow.
Mixed-Matrix Hydrogels Containing Matrigel
and Type I and IV Collagens Are Optimal for Addition of VEGFC Induces Specialization of
the Production of Vascularized, Myelopoietic Organoid Vasculature to a Bone Marrow
Organoids Sinusoid-like Phenotype
To mimic the central bone marrow space (Fig.  1A), we Having determined that collagen I + IV Matrigel hydrogels
devised a four-stage workflow to generate mesenchymal, vas- enabled multilineage differentiation and the generation of
cular, and myelopoietic marrow components. Human iPSCs a 3D vascular network, we sought to determine the optimal
were allowed to form nonadherent mesodermal aggregates balance of endothelial growth factor support to generate
(phase I, days 0–3; Fig.  1B) before commitment to vascular organoid vasculature that resembles bone marrow sinusoids
and hematopoietic lineages (phase II, days 3–5; Fig.  1B). (24, 25). VEGFA is a key regulator of blood vessel forma-
The resulting cell aggregates were then embedded in mixed tion in health and disease, acting via the VEGF receptors
collagen–Matrigel hydrogels to induce vascular sprouting VEGFR1 and 2 (26). Bone marrow sinusoidal endothelial cells
(phase III, days 5–12; Fig.  1B). At day 12, sprouts were col- express VEGFR1 and 2 as well as VEGFR3 (27), and VEGFC—
lected individually and cultured to form bone marrow orga- the main ligand for VEGFR3—was recently demonstrated to
noids in 96-well ultralow attachment (ULA) plates (phase IV, maintain the perivascular hematopoietic niche in murine
day 12 onward; Fig. 1B). bone marrow (28). We therefore tested the effect of adding
Hydrogels comprising Matrigel plus type I collagen have VEGFC to the vascular sprouting phase (day 5; Fig. 2A). The
previously been used to support the formation of iPSC- addition of VEGFC significantly increased the expression of
derived blood vessel organoids (21). However, type IV col- FLT4 (encoding VEGFR3) in the organoids, as well as HSPC
lagen is more permissive than type I collagen for myeloid adhesion molecules (VCAM1 and ITGA4) and HSPC-support-
and megakaryocyte maturation in standard 2D in vitro ing growth factors and chemotactic cytokines (CXCR4 and
culture systems (22, 23). We therefore compared hydrogels FGF4; Fig.  2B). Supplementation with VEGFC in addition
containing type I and/or type IV collagen plus Matrigel to VEGFA also induced retention of CD34 expression on
for the generation of stromal, endothelial, and myeloid organoid vessels, similar to native adult bone marrow vessels,
lineages. Distinct immunophenotypic hematopoietic stem whereas organoids stimulated with VEGFA alone expressed

366 | CANCER DISCOVERY FEBRUARY 2023 [Link]

Human Bone Marrow Organoids for Modeling Hematopoiesis RESEARCH ARTICLE

A
(i) (ii)

Neutrophil
Erythroid
progenitor
Dendritic
cell (iii) Central bone marrow space
Monocyte -Highly vascularized space
-Specialized sinusoidal endothelium
Adipocyte -Hematopoiesis
-Transmigration through sinusoids
HSC/HSPC -Specialized stroma

Endosteum

Downloaded from [Link] by guest on 13 February 2023

MSC

Fibroblast

Megakaryocyte

Arterial
endothelium
Sinusoidal
endothelium

B
d3 Mesoderm agg. d-1: Generation of StemFlex + RevitaCell, ULA
iPSC aggregates plates

Phase I: BMP4, FGF2, VEGFA

Days 0–3: Mesodermal (5% oxygen)
induction

100 µm
Phase II:
Sprouted d12 collagen IV Days 3–5: Vascular and BMP4, FGF2, VEGFA,
hematopoietic SCF, FLT3
commitment

Phase III:
Days 5–12: Embedded in
mixed-matrix hydrogel
(sprouting) VEGFA, FGF2, BMP4,
SCF, FLT3, IL3, IL6,
cGSF, EPO, TPO

D18 mature organoids Collagen + Matrigel

hydrogels
Phase IV:
Day 12+:
Collection and
culture in ULA
plates

Figure 1.  Mixed-matrix hydrogels containing Matrigel and type I and IV collagens are optimal for production of vascularized, myelopoietic organoids.
A, (i) Central bone marrow is a complex tissue including MSC, endothelial, hematopoietic stem and progenitor cell (HSC/HSPCs), and myeloid and lymphoid
subsets. (ii) Hematoxylin and eosin–stained section and (iii) model of human bone marrow highlighting the diverse hematopoietic and stromal cell types
(created using [Link]). B, Differentiation workflow, in which iPSC aggregates undergo mesodermal induction (days 0–3) and commitment to
hematopoietic and vascular lineages (days 3–5). Cell aggregates are then embedded in mixed-matrix hydrogels comprised of Matrigel and collagen I, col-
lagen IV, or collagen I + IV mix at a 40:60 ratio to support vascular sprouting. Key media components are listed for each phase. (continued on next page)

FEBRUARY 2023 CANCER DISCOVERY | 367

RESEARCH ARTICLE Khan et al.

C HSPC Myelomonocytic MK Erythroid Endothelial MSC

21.6% 17.6% 18.3% 11.9%

CD235
CD11b

CD144
CD41

LepR
CD34

11.9% 24.9%

Lineage Lineage CD42a CD71 CD31 VCAM1

D * * * *
50
* * * *
40
% Live cells

30

Downloaded from [Link] by guest on 13 February 2023

20

10

0
HSPC Myelomonocytic Megakaryocyte Erythroid Endothelial MSC
CD34+Lin– CD11b+Lin+ CD41+CD42+ CD71+CD235+ CD31+CD144+ PDGFRb+LepR+
Lin– Lin– VCAM1+

Collagen type I Collagen type IV Collagen type I + IV

E F ***
*** ****
100 600
MSC
Sprout radius (µm2)

Endothelial
% Population

400
Erythroid
50 MK
Myelomonocytic 200

HSPC

0 0
I

IV

IV

IV

IV
pe

n
ge
pe

I+

I+
ty

ge
la
ty

pe

n
n

la
ol

ge
ge

ol
ty

C
ge

la
la

C
n
la

ol
ol

ge
ol

C
C

la
C

ol
C

Figure 1. (Continued) Gating strategy (C) and quantification (D) of stromal and hematopoietic cell types in day 18 organoids supported by Matrigel + collagen
type I–only, collagen IV–only, and collagen I + IV hydrogels. MK, megakaryocyte. E, Distribution of lineages as fractions of the whole organoid population.
F, Radius of endothelial sprouts. (continued on following page)

CD34 during vessel sprouting (days 5–12; Fig.  1G) but lost vessels (Fig.  2F). A hallmark of the bone marrow perivascu-
CD34 expression by day 18 (Figs. 1H and 2C). lar niche in vivo is the close association of megakaryocytes
Organoid vessels formed clear lumens, containing extrava- with sinusoidal endothelium, where they extend long, beaded
sating hematopoietic cells (Fig.  2D and E), confirmed by proplatelet extensions into the vessel lumen (Fig. 2G). These
multiplexed immunostaining to identify CD45+ and CD71+ extensions subsequently generate platelet buds under shear
hematopoietic cells within the lumen of UEA1+ sinusoidal forces (Fig.  2G; ref.  29). Cryosections of the bone marrow

368 | CANCER DISCOVERY FEBRUARY 2023 [Link]

Human Bone Marrow Organoids for Modeling Hematopoiesis RESEARCH ARTICLE

G DAPI CD34 CD144 CD34 H

DAPI CD34 UEA1

Collagen I
Collagen IV

50 µm

Downloaded from [Link] by guest on 13 February 2023

Collagen I + IV

100 µm

50 µm 10 µm 10 µm

I DAPI CD41 CD144 J DAPI UEA1 CD71

50 µm 100 µm

Figure 1. (Continued) G, Sprouting day 12 organoids immunostained for nuclei (DAPI), CD34, and CD144 (VE-cadherin). H–J, Whole organoid Z-stack
imaging acquired at day 18 showing CD34+ HSPCs and UEA1+ vessels that are negative for CD34 (H), CD41+ megakaryocytes (I), and CD71+ erythroid
cells dispersed throughout the organoids and associating with CD144+/UEA1+ vasculature (J). *, P < 0.05; ***, P < 0.001 for one-way ANOVA with multiple
comparisons [Fisher least significant difference (LSD)]; n = 3 for endothelial sprout radius measurements. Two-way ANOVA with multiple comparisons
(Fisher LSD); n = 3 (3 independent differentiations, 15 pooled organoids each) for flow cytometry analysis. Representative images are shown. See also
Supplementary Fig. S1.

organoids stained for CD41+ MKs, CD144+ endothelium, extensive endothelial network invested with perivascular
and CD140b+ MSCs/fibroblasts demonstrated classic pro- CD140b+ fibroblasts/MSCs. A higher number of megakar-
platelet protrusions in association with the vessels (Fig.  2H yocytes were observed in close proximity (5 μm) to vessels
and I), with remarkable similarity to previously published in VEGFA  +  C–stimulated organoids than in VEGFA-only
in vivo images of thrombopoiesis occurring in calvarial bone organoids (Supplementary Fig.  S1B), consistent with the
marrow (29). Volumetric imaging of whole-mount organoids increased expression of chemotactic factors and adhesion
demonstrated that these cells were organized in 3D in an proteins in VEGFA  +  C organoids (Fig.  2B). Together, these

FEBRUARY 2023 CANCER DISCOVERY | 369

RESEARCH ARTICLE Khan et al.

A B FLT4 VCAM1 FGF4 ITGA4 CXCR4

**
** ** **
erm 0.8 ** 8 0.8 4,000 * 500
od
es tion
***
M c * *
d0 ndu 0.6 6 0.6 3,000
400
i
nd 300

∆∆CT
ra 0.4 4 0.4 2,000
c ula ietic 200
s o
Va op ent
d3 mat itm 0.2 2 0.2 1,000
he mm 100
co
l 0.0 0 0.0 0 0
ge
y dro ng
H ti VEGFA VEGFC VEGFA + C
d5 prou
s
Phase III:
C DAPI CD34 CD140b UEA1 D
CD34 IHC
VEGFA 50 ng/mL d12 d18
VEGFC 50 ng/mL
VEGFA + C 25 ng/mL each
on
VEGFA

cti
lle re in

Downloaded from [Link] by guest on 13 February 2023

o
2 C ltu s
d1 d cu late
an LA p
U

a m
8+ tre
d1 wns
Do says
200 µm
as
VEGFA + C

20 µm

100 µm

E F G H
UEA1 Perivascular DAPI CD41 CD144 CD140b
H&E CD45 MSC
DAPI MK

15 µm 5 µm

UEA1 Proplatelet
CD71 extension
DAPI

Sinusoidal
endothelium
10 µm 200 µm 100 µm
15 µm 5 µm

I
DAPI CD41 UEA1 CD140b CD41 MKs UEA1 endothelial network CD41 MKs UEA1 endothelial network CD140b
800

700

600

500

400

300

20 µm 200

100

–100

100
–200 200
–200 –100 0 100 200 300 400 500 600 700 800
100 µm

370 | CANCER DISCOVERY FEBRUARY 2023 [Link]

Human Bone Marrow Organoids for Modeling Hematopoiesis RESEARCH ARTICLE

data indicate that the addition of VEGFC improved vasculari- To determine the transcriptional similarity of the bone
zation of the organoids and hematopoietic support. marrow organoid to native human hematopoietic tissues,
organoid scRNA-seq data were projected onto published
scRNA-seq Confirmed that Hematopoietic single-cell datasets of cells isolated from adult human bone
and Stromal Cell Lineages within Organoids marrow (31, 32) and from fetal liver and fetal bone mar-
Have Transcriptional Homology to Human row (33) using the Symphony package (34). This revealed
Hematopoietic Tissues an extensive overlap of organoid-derived cells with HSPCs,
To compare the cell types and molecular profiles of the myeloid subsets, fibroblasts/MSCs, and endothelial cell types,
organoids to human hematopoietic tissues, scRNA-seq was with the predicted cell types matching the cluster annota-
performed on a total of 26,648 cells from 3 independent tions (Fig. 3F; Supplementary Fig. S3A–S3C).
organoid differentiations using VEGFA-only and VEGFA + C To explore the transcriptional similarity between the stro-
protocols. After quality control (see Methods), 19,506 cells mal support elements of the organoids with that of native
(10,205 from VEGFA only and 9,301 from VEGFA + C) were human bone marrow in more detail, we extracted the stro-
included in downstream analyses. Distinct populations mal cell populations from a recently published study of fetal
of key hematopoietic and stromal cell subtypes were iden- bone marrow. This provided the first comprehensive anno-
tified, including HSPC, erythroid, neutrophil, monocyte, tation of stromal cell subsets in human bone marrow (35),
megakaryocyte, eosinophil/basophil/mast (EBM), fibroblast, including endothelial, MSC, fibroblast, and osteochondral

Downloaded from [Link] by guest on 13 February 2023

endothelial cell, and MSC (Fig. 3A; Supplementary Table S1), lineage subsets. Integration and unsupervised clustering
annotated using gene set enrichment analysis with a curated showed a close approximation of organoid MSCs/fibro-
list of 64 published gene sets (Fig.  3B; Supplementary blasts (n = 687) and endothelial cells (n = 766) with the rel-
Table S2; refs. 30, 31), as well as the expression of canonical evant cell clusters extracted from the human bone marrow
marker genes (Fig.  3C). Erythroid, megakaryocytic, mono- dataset (endothelial, n = 766 and MSCs/fibroblasts, n = 687
cytic/neutrophil, and EBM populations demonstrated expres- cells; Supplementary Fig. S4A–S4C). As expected, osteoline-
sion of GYPA and KLF1, PF4 and PPBP, CD14 and RUNX1, and age cells, chondrocytes, and smooth muscle and Schwann
TPSB2 and KIT, respectively (Fig. 3C). cells present in fetal bone marrow were absent from the
Within stromal cell subsets, expression of COL3A1, platelet- organoids, and the distinct populations of sinusoidal and
derived growth factors (PDGFRA/B), and the key HSPC niche nonsinusoidal “tip” endothelial cells were also not detected.
factor CXCL12 (SDF-1) was observed in fibroblasts and MSCs, However, expression of adhesion proteins, cytokines, and
with particularly high expression of CXCL12 in the MSCs, sug- hematopoietic support factors was very similar between
gesting a key role for MSCs in homing and maintenance of organoid and bone marrow cells, including CD34, PECAM,
HSPCs to perivascular regions of the organoids (Fig. 3C; Sup- KITLG, FLT3LG, and ANGPT2 for endothelial cells (Fig. 4A)
plementary Fig.  S2A and S2B). High expression of PECAM1, and KITLG, PDGF, and VEGF family members for MSCs/
CDH5, and ENG was detected in endothelial cells, confirming fibroblasts (Fig. 4B), confirming that the iPSC bone marrow
vascular specification (Fig.  3C; Supplementary Fig.  S2A and organoid–derived stromal cells are highly homologous to
S2B). The relative frequency of cell types in VEGFA-only orga- their native bone marrow counterparts.
noids was broadly similar to that of VEGFA  +  C organoids,
with a relative increase in abundance of endothelial and eryth- Bone Marrow Organoids Recapitulate Cellular and
roid cells captured from organoids grown in the VEGFA + C Molecular Cross-talk between Hematopoietic,
condition (Supplementary Fig. S2C). Endothelial, and Stromal Cells
Trajectory analysis using a force-directed graph (FDG) To investigate the cellular and molecular interactions
showed that the organoids contained cells recapitulating the between hematopoietic, endothelial, and stromal cell subtypes
three main routes of hematopoietic myeloid differentiation within the organoids, we mapped the expression of interact-
(Fig. 3D; Supplementary Fig. S2D), similar to observations in ing receptor and ligand pairs across clusters. Complex com-
human bone marrow (31). FDG analysis of the stromal cell munication networks were detected, both within and between
populations showed independent routes of differentiation hematopoietic and stromal cell compartments (Fig.  4C).
for endothelium and MSCs/fibroblasts as expected (Fig. 3E; Strong autocrine and paracrine interactions were predicted
Supplementary Fig. S2E). between MSCs, fibroblasts, endothelial cells, and monocytes,

Figure 2.  The addition of VEGFC induces specialization of organoid vasculature to a bone marrow sinusoid-like phenotype. A, In the sprouting phase
of differentiation (D5) in hydrogels, organoids were supplemented with either VEGFA or VEGFC, or both VEGFA and VEGFC. B, mRNA expression of
canonical cell-surface receptors, growth factors, and adhesion markers of bone marrow sinusoidal endothelium in VEGFA-, VEGFC-, and VEGFA + C–
treated samples. ΔΔCt values relative to housekeeping (GAPDH) and undifferentiated iPSCs shown. Each datapoint represents 15 organoids; 3 independ-
ent differentiations shown. *, P < 0.05; **, P < 0.01; ***, P < 0.001 for one-way ANOVA with multiple comparisons (Fisher least significant difference). C,
CD34+ sprouting vessels at day 12 in both VEGFA and VEGFA + C conditions. At day 18, vessels were CD34 positive in VEGFA + C organoids but negative
in VEGFA-only organoids. D and E, IHC staining for CD34 (D) and hematoxylin and eosin (H&E) staining of formalin-fixed, paraffin-embedded VEGFA + C
organoid sections (E), with inset showing lumen-forming vessels containing hematopoietic cells (blue arrows). F, Immunofluorescence staining of paraffin-
embedded sections of VEGFA + C organoids showing CD45+ hematopoietic (white arrow) and CD71+ erythroid cells (yellow arrows) migrating into the
UAE1+ vessel lumen. G, Schematic demonstrating the process of proplatelet formation by megakaryocytes (MK; image created using [Link]). H,
Whole organoid image showing CD140b+ MSCs surrounding CD144+ vessels with CD41+ megakaryocytes. Insets show megakaryocytes extending proplate-
let protrusions into vessel lumen (red arrows). Top inset shows CD41+ platelet-like particles within vessel lumen. I, Confocal imaging and 3D render of whole-
mount VEGFA + C organoids showing CD41+ megakaryocytes (red arrow) closely associating with the UEA1+ vessel network that is invested with CD140b+
fibroblast/MSCs (blue arrow; left and center image). Inset (right) shows 3D rendered megakaryocytes displaying proplatelet formation (red arrow).

FEBRUARY 2023 CANCER DISCOVERY | 371

RESEARCH ARTICLE Khan et al.

A B 4

–Log10 P
3
Hematopoietic Stromal 2
1
EBM Megakaryocyte 1 0
Princeton monocyte
progenitors Princeton neutrophil
Velten et al. Monocyte/dendritic donor 1
Monocyte/neutrophil Megakaryocyte 2 MSC 1 Velten et al. Neutrophil progenitor 1 donor 1
progenitors Velten et al. Neutrophil progenitor 3 donor 1
MSC 2 Hallmark TNF signaling via NFKb
Princeton macrophage
Popescu et al. MK
Monocyte Early erythroid 2 Chen et al. MK
Fibroblast 1 Velten et al. MK progenitor donor 1
HSPC Eppert et al. Progenitor
Early erythroid 1 Fibroblast 3 Velten et al. Monocyte/dendritic progenitor donor 2
Endothelial Popescu et al. MEMP
WP Hematopoeitic stem cell differentiation
Velten et al. MkE progenitor 2 donor
Mid-erythroid Hay et al. Bone marrow CD34 HSC
+

Popescu et al. HSC/MPP

Fibroblast 2
UMAP1

Velten et al. Myeloid progenitor 1 donor 1

Late erythroid 1 Zheng et al. Cord blood C6 HSC multipotent progenitor
Late erythroid 2 Hay et al. Bone marrow CD34+ LMPP
Drissen et al. HSC MPP
Eppert et al. HSC
UMAP2 Popescu et al. DC precursor
Chen et al. HSC
Popescu et al. pDC precursor
C Canonical Markers
Drissen et al. LMPP

Downloaded from [Link] by guest on 13 February 2023

Popescu et al. NK
Stromal Hematopoietic Velten et al. Myeloid progenitor 2 donor 1
Velten et al. MkE progenitor 1 donor
PDGFRB TRIM6 0.8 Velten et al. Myeloid progenitor donor 2
HSPC

PDGFRA MYB Curated HSPC

MSC/fibroblast

NES HSPC MPP curated

4.0 CD34 Hallmark quiescence
KITLG
ITGA4 TPSB2 2.5 Popescu et al. Mast
Myeloid

PTPRC Popescu et al. Monocyte

EMCN KEGG hematopoietic cell lineage
CXCL12 PRSS57 Popescu et al. Monocyte/macrophage
CSPG4 ITGAM Popescu et al. DC1
COL3A1 CD14 Popescu et al. Kupffer
PPBP 4.0 Popescu et al. DC2
PECAM1 Popescu et al. Monocyte precursor
Erythroid MK
Endothelial

MCAM PF4
2.0 GP9
Drissen et al. GMP
FLT4 Popescu et al. Neutrophil myeloid progenitor
Velten et al. Neutrophil progenitor 2 donor 1
ENG KLF1 2.5 Popescu et al. Early erythroid
CDH5 GYPA Popescu et al. Mid erythroid
CD34 ALAS2 Velten et al. Erythroid progenitor 2 donor 1
Drissen et al. MEP
Expressing
Endothelial
MSC 1
MSC 2
Fibroblast 1
Fibroblast 2
Fibroblast 3

Monocyte/neutrophil prog.

Eosinophil/nasophil/mast prog.
HSPC

Monocytes

Megakaryocyte 1
Megakaryocyte 2
Early erythroid 1
Early erythroid 2
Mid erythroid
Late erythroid 1
Late erythroid 2

Popescu et al. VCAM EI macrophage

cells (%) Popescu et al. Late erythroid
0 Velten et al. Erythroid progenitor 1 donor 1
25
Monocyte/neutrophil prog.
EBM prog.
Monocytes

Mid-erythroid
Late erythroid 1
Late erythroid 2
Megakaryocyte 1
Megakaryocyte 2
HSPC
Early erythroid 1
Early erythroid 2
50
75
100

D E Name
F Predicted cell type: Popescu et al.
MSC Fibroblast
HSPC:Erythroid Fibroblast
Late erythroid
Endothelial
HSPC:Mono-Neut Endothelial cell

Mid-erythroid
Neutrophil-myeloid
Megakaryocyte progenitor
Mono-Mac
Name Monocyte
MEMP precursor
HSPC:MK Erythroid
HSPC DC2
Monocyte/neutrophil HSC_MPP
UMAP1

Megakaryocyte
FDG1

FDG1

Monocyte
Early erythroid Mast cell DC1
DC precursor

FDG2 FDG2 UMAP2

Figure 3.  scRNA-seq confirmed that hematopoietic and stromal cell lineages within organoids showed transcriptional similarity to human hematopoi-
etic tissues. A, Uniform manifold approximation and projection (UMAP) plot showing annotated cell clusters. B, Gene set enrichment analysis of differ-
entially expressed genes for each cluster using a curated set of 64 hematopoietic lineage gene sets. DC, dendritic cell; GMP, granulocyte-macrophage
progenitor; HSC, hematopoietic stem cell; KEGG, Kyoto Encyclopedia of Genes and Genomes; LMPP, lymphoid-primed multipotent progenitor; MEMP,
megakaryocyte-erythroid-mast cell progenitor; MK, megakaryocyte; MkE, megakaryocyte-erythroid; MPP, multipotent progenitor; NK, natural killer cell;
pDC, plasmacytoid DC; prog., progenitor; VCAM EI, VCAM erythroblastic island; WP, WikiPathways. C, Expression of canonical stromal and hematopoietic
cell genes for each of the annotated clusters. The color scale represents the average level of expression, and the circle size shows the percentage of cells
within each cluster in which expression was detected for each gene. D and E, FDG showing differentiation trajectories for hematopoietic (D) and stromal (E)
compartments, superimposed with expression scores of lineage signature gene sets. F, Organoid cells projected onto a published dataset of human hemat-
opoietic and stromal cells using the Symphony package (33). Mono-Mac, monocyte-macrophage. See also Supplementary Figs. 2 and 3 and Supplementary
Tables S1 and S2.

372 | CANCER DISCOVERY FEBRUARY 2023 [Link]

Human Bone Marrow Organoids for Modeling Hematopoiesis RESEARCH ARTICLE

A C D
CDH5 Erythroid DE genes in ECs from VEGFA + C– vs.
3.0 CD34 Fibroblast VEGFA-derived BM organoids
PECAM1
Adhesion EBM prog.
SELE
40
VCAM1
ICAM2
THBS1 HSPC
% Cells MCAM
30
CCL14 Endothelial DLK1
25
CXCL1
50 DLL4 HBEGFANGPT2

–Log10 (FDR)
Growth factors/chemokines

CXCL2
75 COL4A1
CSF1 MCAM
100 MK 20
CSF2
ITGA2 KITLG
CSF3
MSC CD34 JAG1
KITLG
SEMA3F
FLT3LG
10 ITGA6
ANGPT2 Monocyte JAM3
HBEGF FLT4 JAG2
PDGFA
Monocyte/neutrophil prog.
PDGFB
VEFGFA CDC42EP1
Number of interactions
VEGFB –6 –4 –2 0 2 4
250

Downloaded from [Link] by guest on 13 February 2023

VEGFC
200 Log2 (FC)
Receptors

NOTCH4
DLL4 100
JAG2 25
FLT4
FB .
M
rg
O

Endothelial

B E
1.0 IL2
VEGFA
IL4
VEGFA + C
(normalized UMI counts +1)

IL6
IL7
Growth factors/chemokines

IL11
CXCL1
4
CXCL2
Log2

CXCL10
2
CSF1
CSF2
CSF3 0
KITLG
FLT4
***

CD34
****

MCAM
****

ANGPT2
****

COL4A1
****

KITLG
****

HBEGF
****

DLK1
****

ITGA2
****

NOTCH4
****

DLL4
****
FLT3LG
PDGFA
PDGFB
VEGFA
VEGFB
VEGFC
FB .
M
rg
O

MSC/fibroblast

Figure 4.  Endothelial cells (EC), fibroblasts, and MSCs from organoid stroma support hematopoiesis with increased hematopoietic support from
VEGFA + C–stimulated vasculature. A, Comparison of expression of key receptors, adhesion proteins, and growth factors and chemokines in ECs from
bone marrow organoids (Org.) and human fetal bone marrow (FBM). B, Comparison of expression of growth factors and chemokines in MSCs from orga-
noids and FBM. Size of dots represents percentage of expressing cells, and color density indicates level of expression. C, Total predicted ligand–receptor
interactions across clusters as predicted by CellPhoneDB (V2.0), showing extensive autocrine and paracrine interactions across BM organoid. MK, mega-
karyocyte; prog., progenitor. D, Volcano plot showing significantly differentially expressed (DE) genes in ECs from VEGFA + C versus VEGFA-only orga-
noids (801 significantly upregulated and 700 significantly downregulated genes, P < 0.05, log2FC > 0.5 or −0.5). E, Violin plots showing key hematopoietic
support factors and markers of bone marrow sinusoidal endothelium in ECs of VEGFA + C and VEGFA-only organoids. P values are indicated below x-axis
labels, and mean value is indicated on violin plots. ***, P < 0.001; ****, P < 0.0001 for pairwise comparison Wilcox test applied (FDR). UMI, unique molecular
identifier. (continued on next page)

HSPCs, and megakaryocytes, whereas erythroid cells showed S6A), and NOTCH1–DLL4, KIT–KITLG, FGF2–CD44, SELP–
weak interactions (Fig. 4C; Supplementary Fig. S5A–S5C). CD34, and VEGFRA–FLT1 for megakaryocytes:endothelial cells
Numerous interacting receptor–ligand partners were detected (Supplementary Fig. S6A). Interactions between megakaryocytes
between megakaryocytes and endothelial cells (Supplementary and MSCs were dominated by growth factors produced by
Figs. S5B and S6A), and megakaryocytes with MSCs (Sup- megakaryocytes, including TGFB, PDGF, FGF, and VEGF family
plementary Fig.  S6B), indicating bidirectional regulatory members and their cognate receptors (Supplementary Fig. S6B).
interactions between these cell types. These included NOTCH1– Similarly, monocytes and endothelial cells demonstrated
JAG1/2, FLT4–PDGFC, ANGPT2–TEK, and FLT1–VEGFB for abundant interacting partners indicative of regulatory
endothelium:megakaryocytes (Supplementary Figs. S5B and interactions, including TNF–NOTCH1, JAG1/2–NOTCH,

FEBRUARY 2023 CANCER DISCOVERY | 373

RESEARCH ARTICLE Khan et al.

F Downregulated in VEGFA + C I
Upregulated in VEGFA + C
50 Adhesion MSC/
molecules fibroblast
TGFβ receptor 3 HSPC
Endo Endo
Receptor–
40 ligand pairings Endothelial
Fibro
Growth factors
Fibro
30 HSPC
TGFβ receptor 1
TGFb1

TGFβ HSPC
MK
20 MK
0.6 ESAM ESAM 4.0
Mono Mono CD44 SELE
TGFβ receptor 2 NRP2 SEMA3F

Adhesion
10
MSC
MSC ICAM3 ITGB2
EGF receptor
HLA-DRB1 OGN
0
Ery HLA-DPA1 GAL
HLA-DPB1 TNFSF13B % Cells
Ligand cluster Ligand Receptor Receptor cluster
PLXNC1 SEMA7A

GF/chem.
25
G 20 FGFR1 FGF7 50
Endo Endo
KIT KITLG 75
TNFRSF1B GRN

Downloaded from [Link] by guest on 13 February 2023

100
TNFRSR10C TNFSF10
ACKR3
Fibro NOTCH1 JAG2
Fibro NOTCH2 DLL4
CXCL12

Receptors
CD74 APP
HSPC
10 CXCL12 PGRMC2 CCL4L2
HSPC MK MERTK GAS6
MK SELL CD34
MSC PLAUR ITGB1
CXCR4
CD46 JAG1
MSC
LAMP1 FAM3C
Mono

.
FB .

FB .
M

M
rg
rg

rg
0

O
O

O
Ligand cluster Ligand Receptor Receptor cluster
HSPC Endo MSC/
fibro
H 50
Endo
FGF2 Endo
Fibro
40
Fibro
LGALS9
HSPC
CD44 HSPC
30
HGF MK
CD44

MK

20
Mono
CD44
Mono
10
HBEGF
SPP1 MSC
MSC
0 Ery

Ligand cluster Ligand Receptor Receptor cluster

J 10,000
25
7,500

20 5,000
200 2,000
200–7,000 pg/mL
20–200 pg/mL
1–20 pg/mL

15
1,500
pg/mL

pg/mL

pg/mL

10 1,000
100

5 500

0 0 0
G F-α

L
VE A

C CL1
2
3

C G
C 2
L3
X3 4
L1
6
8
11
M L10
M -β

XC 2
4
PD TN F

FLSF
C IL7
F

40

T3
C CL
IL
IL

C CL

IL
IL
A

IL
S

TL
G
G NF

C
C

IL
F-
-C

-C
D

C
X

KI
C
T

F/
SC

Figure 4. (Continued) F–H, Sankey plots comparing TGFβ1-mediated (F), CXCL12-mediated (G), and CD44-mediated (H) interactions in VEGFA + C
versus VEGFA-stimulated organoids. Endo, endothelial; Ery, erythroid; Fibro, fibroblast; Mono, monocyte. I, Expression of interacting receptor–ligand
pairs between organoid HSPCs and cognate partner in organoid or FBM ECs (endo) and MSCs/fibroblasts with percentage of expressing cells and level
of expression shown. J, Hematopoietic cytokines/growth factors produced by bone marrow organoids, measured by Luminex assay. Each datapoint repre-
sents supernatant pooled from 16 separately generated organoids. See also Supplementary Figs. 4–7.

374 | CANCER DISCOVERY FEBRUARY 2023 [Link]

Human Bone Marrow Organoids for Modeling Hematopoiesis RESEARCH ARTICLE

SIRPA–CD47, and LGALS9, ICAM, VEGF family members (Fig. 4J; Supplementary Fig. S7B), confirming that the orga-
with cognate receptors (Supplementary Figs. S5C and S6C). noid stroma expresses key growth factors that might endog-
Significantly interacting partners between monocytes and enously support hematopoiesis.
MSCs/fibroblasts included CXCL12–CXCR4, ICAM1–aXb2,
SPP1–CD44, IL1 and IL16, and hepatocyte and fibroblast Bone Marrow Organoids Model the TGFβ-Induced
growth factors with their respective binding partners (Sup- Bone Marrow Fibrosis That Occurs in Hematologic
plementary Fig. S6D). Cancers and Provide an Ex Vivo Platform for
Although a high number of regulatory interactions between Inhibitor Screening
hematopoietic and stromal compartments were detected, Pathologic hematopoietic niche remodeling occurs in the
interactions between the stromal cell subsets (endothelial majority of hematologic malignancies. In certain cancers,
cells:MSCs:fibroblasts) were particularly strong (Fig. 4C). Sig- particularly myeloproliferative neoplasms, myelodysplasia,
nificantly interacting partners identified included key regula- acute leukemia, and mast cell neoplasms, bone marrow fibro-
tory molecules such as JAG–NOTCH, VEGF, DLL4–NOTCH3, sis is a major cause of bone marrow failure and morbidity and
PDGF–PDGFR, ANPT1/2–TEK, IL33–IL33R, FGF, and TGFB is associated with a poor prognosis (40). Fibrosis results from
(Supplementary Fig. S6E; refs. 36–38). the excess production and release of profibrotic cytokines
To explore the impact of the addition of VEGFC to the by hematopoietic cells—in particular TGFβ—leading to the
differentiation protocol on the phenotype of organoid vas- deposition of reticulin and collagen fibers by marrow stroma

Downloaded from [Link] by guest on 13 February 2023

culature, we compared the transcriptomes of endothelial (40–42). To investigate whether the organoids could model
cells captured from VEGFA- and VEGFA  +  C–derived bone pathologic bone marrow fibrosis, we treated organoids with
marrow organoids by scRNA-seq. A total of 1,501 genes were varying doses of TGFβ (10 ng/mL, 25 ng/mL, and 50 ng/mL),
significantly differentially expressed between endothelial cells which resulted in a dose-dependent increase in the expres-
generated with VEGFA only versus VEGFA  +  C, including sion of hallmarks of fibrosis, including alpha smooth muscle
801 upregulated and 700 downregulated genes (P  <  0.05, actin (αSMA [ACTA2]) and collagen 1 (COL1A1; Fig.  5A),
log2FC  >  0.5 or  −0.5; Fig.  4D). Canonical markers of bone both canonical markers of fibroblast activation (42). A sig-
marrow sinusoidal endothelium were more highly expressed nificant increase in soluble IL11 was also observed (Fig.  5B;
in endothelial cells from VEGFA + C organoids than VEGFA- ref. 43). Collagen deposition within the organoids was mark-
only organoids, including FLT4 (VEGFR4), CD34, MCAM, edly increased following TGFβ treatment (Fig.  5C), with
ANGPT2, COL4A1, COL4A2, ITGA2, CDC42EP1, and the notch pronounced reticulin fibrosis (Fig. 5D and E), recapitulating
ligand DLL4 (25, 35), whereas DLK1—a negative regulator of changes seen in the bone marrow of patients with myelofibro-
hematopoiesis (39)—was significantly lower in VEGFA  +  C– sis. The induction of organoid fibrosis was accompanied by
stimulated organoids than in VEGFA-only (Fig. 4D and E). reduced vascularization, suggesting multilineage remodeling
In addition to improved hematopoietic support from as a consequence of TGFβ stimulation, as occurs in adult
endothelial cells, key regulatory axes were also upregulated bone marrow (Fig. 5F and G).
across other cellular subsets in VEGFA  +  C organoids com- We then explored the utility of this system to test potential
pared with VEGFA organoids (Fig.  4F–H). TGFβ1 signaling inhibitors of fibrosis, selecting two compounds that inhibit
primarily from MSCs, megakaryocytes, and fibroblasts was pathways currently under investigation in clinical trials for
increased overall in VEGFA +  C organoids across the differ- myeloid malignancies—SB431542, an inhibitor of the TGFβ
ent TGFβ receptors (Fig.  4F). Similarly, CXCL12 signaling superfamily type I activin receptors, and the BET bromodo-
via CXCR4 and ACKR3 receptors across both stromal and main inhibitor JQ1 (44). TGFβ-induced expression of soluble
hematopoietic cell types was also elevated (Fig. 4G), whereas IL11 was completely inhibited by both treatments (Fig. 5H),
the impact of VEGFC on signaling between CD44 and its whereas ACTA2 and COL1A1 overexpression was normalized
binding partners was more mixed (Fig. 4H). by JQ1 and reduced by 1.5-fold and 2.5-fold, respectively, by
The receptor–ligand communication networks predicted SB431542 at the transcript level (Fig. 5I), and this reduction
between hematopoietic cells and niche components in the was also evident at the protein level by immunofluorescence
organoid stroma mirrored many of the communication imaging (Fig.  5J). Together, these data confirm that bone
networks that have been reported in native bone marrow marrow organoids provide an efficient model of malignant
(35), including CD44–SELE, CD74–APP, KIT–KITLG, ICAM3– bone marrow fibrosis and enable screening for the efficacy of
ITGB2, CD46–JAG1, and NOTCH2–DLL4 (Fig. 4I). potential pharmacologic modulators.
Finally, we explored whether the expression of hematopoi-
etic support factors detected in the bone marrow organoid Organoid “Niche Remodeling” and Fibrosis
niche cells by scRNA-seq (Supplementary Fig.  S7A) could Occurs following Engraftment with Cells
be confirmed at protein level. Organoids generated with from Patients with Myelofibrosis but Not
VEGFA  +  C were harvested at day 18, washed, replated Healthy Donors
in media supplemented with L-Glutamine but no added Having confirmed substantial homology to native bone
cytokines or growth factors, and cultured for a further marrow, we hypothesized that the organoids may support
12 days without any added supplements with 50:50 media the engraftment of primary cells from patients with hema-
changes every 72 hours. In the absence of exogenously sup- tologic malignancies, enabling the modeling of cancer–
plied cytokines, the organoids secreted multiple hematopoi- stroma interactions and the possibility of patient-specific
etic factors including SCF/KITLG, CCL2–4, interleukins (IL2, cytotoxic screens. Given the fibrosis observed following
3, 4, 6, 7, 8, 11), PDGF, FLT3L, M-CSF, GM-CSF, and VEGF treatment with TGFβ, and the current lack of adequate

FEBRUARY 2023 CANCER DISCOVERY | 375

 RESEARCH ARTICLE Khan et al.

A B C
ACTA2 COL1A1 IL11
Untreated β
50 ng/mL TGFβ
**
**** * **

DAPI αSMA COL1A1

* * **
15,000 15,000 400
10,000 10,000

IL11 (pg/mL)
300
5,000 5,000
∆∆CT

∆∆CT
1,000 150 200
∆∆

750 ∆∆ 100 100

500
250 50 100 µm
0 0 0
10 ted

ng L
50 /mL

10 ted

ng L
50 /mL

10 ted

ng L
50 /mL

L
25 /m

/m

25 /m

/m

25 /m

/m
a

a
ng

ng

ng

ng

ng

ng
re

re

re
nt

nt

nt
U

U
β
TGFβ β
TGFβ β
TGFβ

IL11
D E H *
Untreated β
50 ng/mL TGFβ **** ****
40

Downloaded from [Link] by guest on 13 February 2023

**** 1,500 **** ****

Total % reticulin area

30

IL11 (pg/mL)
1,000
Reticulin

20
500
10
0
0 DMSO – + – – –
100 µm /m Fβ TGFβ 25 ng/mL – – + + +
SB431542 20 µmol/L
ed

– – – + –
TG

L
at

JQ1 0.5 µmol/L – – – – +

re

ng
nt
U

50

F Untreated β
50 ng/mL TGFβ G I ACTA2 COL1A1

**** *** ***

Total % vascular area (CD34)

15 **** ***
*** * *
Fold change in ∆∆CT

Fold change in ∆∆CT

10 6 **** ****
∆∆

∆∆
8
CD34 IHC

10
6 4

4
5 2
2
100 µm 0 0
0 DMSO – + – – – – + – – –
TGFβ 25 ng/mL – – + + + – – + + +
/m β
ed

ng GF

SB431542 20 µmol/L – – – + – – – – + –
L
at

T
re

JQ1 0.5 µmol/L – – – – + – – – – +

nt
U

50

J
DAPI αSMA COL1A1 DAPI αSMA COL1A1
Untreated

SB431542
β
25 ng/mL TGFβ

JQ1

100 µm

376 | CANCER DISCOVERY FEBRUARY 2023 [Link]

Human Bone Marrow Organoids for Modeling Hematopoiesis RESEARCH ARTICLE

in vitro and in vivo systems for modeling cancer-induced Patient Cell–Engrafted Organoids Allow for
bone marrow fibrosis, we first seeded the organoids with Testing of Potential Inhibitors of Fibrosis
cells from healthy donors and patients with myelofibrosis, We next assessed whether organoids engrafted with cells
and tested the impact of engraftment on the remodeling of from patients with myelofibrosis enabled screening of poten-
the bone marrow organoid “niche.” tial inhibitors of fibrosis, to explore the utility of this platform
Organoids were seeded with CD34+ HSPCs from healthy for precision medicine approaches. Organoids seeded with
donors (n = 7) and patients with myelofibrosis (n = 10; Sup- cells from patients with myelofibrosis were cultured for 7 days
plementary Table  S3). Donor cells were labeled with the and then treated with the TGFβ inhibitor SB431542, the BET
plekstrin homology domain dye CellVue Claret, and 5,000 inhibitor JQ1, or ruxolitinib (Fig.  6K). Only JQ1 treatment
donor cells were seeded into each well of a 96-well ULA restored COL1A1 expression to the level seen in nonengrafted
plate containing individual organoids (Fig.  6A). CellVue control organoids, and JQ1 also significantly reduced αSMA
labeling enabled the identification and tracking of donor (Fig.  6L and M). Expression of COL1A1 was lower follow-
cells within the organoid milieu. Confocal Z-stack imag- ing ruxolitinib treatment, with only a minimal reduction
ing confirmed that labeled cells from healthy donors and in αSMA (Fig.  6L and M). Although SB431542 significantly
patients with myelofibrosis efficiently engrafted and were inhibited the induction of COL1A1 and αSMA in organoids
distributed throughout the organoid architecture (Fig.  6B; in response to TGFβ treatment (Fig. 5I), no significant reversal
Supplementary Fig. S8A).

Downloaded from [Link] by guest on 13 February 2023

was observed on hallmarks of fibrosis induced by engraftment
After 14 days, engrafted organoids were assessed for fibro- of patient cells, suggesting that additional profibrotic signals
sis. Soluble TGFβ levels were significantly elevated in the derive from the myelofibrosis clone beyond TGFβ or an inabil-
culture media of organoids engrafted with myelofibrosis ity to reverse fibrosis once established.
cells when compared with healthy donor–engrafted samples
(Fig. 6C). Immunofluorescence imaging showed a significant Bone Marrow Organoids Support the Engraftment,
increase in collagen 1 and αSMA in organoids engrafted Survival, and Proliferation of Primary Cells from a
with cells from patients with myelofibrosis at protein level Range of Hematologic Malignancies
(Fig.  6D and E; Supplementary Fig.  S8B) as well as gene Finally, we investigated whether primary human cells of
expression (Fig.  6F), with a concomitant decrease in expres- other blood cancer types would also successfully engraft
sion of endothelial cell–associated genes CDH5 and TIE2 and the bone marrow organoids. We focused on hematologic
vascularity (Fig. 6G–I). malignancies that are particularly challenging to maintain ex
Patient-derived CD34+ HSPCs cultured within the orga- vivo and/or model in vivo—multiple myeloma, acute lymph-
noids underwent lineage differentiation. After 14 days, a oblastic leukemia (ALL), and myeloid leukemias—explored
quarter (22%) of the total cells in the organoids were posi- whether the organoids could improve the survival of cells
tive for the fluorescent label indicating adult donor origin, ex vivo, thereby enabling mechanistic studies and target
and these cells had undergone myeloid differentiation with screening for these cancer types.
erythroid (CD34− CD235+ CD71+), myelomonocytic (CD34− To test this, cryopreserved cells were thawed and fluores-
CD45+ CD11+ CD14+), and megakaryocytic cell (CD34− cently labeled prior to seeding into 96-well plates contain-
CD41+ CD42a+) immunophenotypes evident (Supplementary ing organoids. Cells from patients with multiple myeloma
Fig.  S9A). As expected, given the absence of lymphopoietic (n = 5; Supplementary Table S4), ALL (n = 6, Supplementary
cytokines, no B or T cells were detected (Supplementary Table S4), and chronic myeloid leukemia (CML; n = 2, Sup-
Fig.  S9A and S9B). A small population of the label-positive plementary Table  S3), a human AML cell line (THP-1), and
cells retained CD34 expression even 14 days after seeding, leukemic cells from a human fetal liver–derived infant ALL
and these cells showed lower rates of cell division, suggesting xenograft model (Xeno iALL; n = 3, Supplementary Fig. S10)
maintenance of a population of quiescent stem/progenitor rapidly engrafted and were observed throughout the orga-
cells in the organoids (Supplementary Fig. S9C) in addition noid volume (Fig.  7A and B). Distinct CellTrace-positive
to myeloid differentiation. populations were detectable in the organoids for all 17 donor
Crucially, the clonal architecture of the malignancies could samples, confirming successful engraftment and survival of
be tracked following culture in the organoids. All muta- primary human and xenograft-derived cells over a 12-day
tions present in the HSPCs prior to seeding were detected in time course (Supplementary Fig.  S11A). Primary multiple
labeled cells sorted from the organoids 12 days after seeding, myeloma cells were costained for CD38, confirming that cells
and at almost identical variant allele frequencies as the origi- derived from the malignant plasma cell clone had success-
nal sample (Fig. 6J). fully engrafted (Fig. 7B).

Figure 5.  Bone marrow organoids model TGFβ-induced bone marrow fibrosis and enable inhibitor screening. A, Organoids were treated with 10, 25,
or 50 ng/mL recombinant TGFβ and evaluated by qRT-PCR for expression of ACTA2 (αSMA) and COL1A1, indicators of fibrosis. B, Soluble IL11 detected
in organoid media following treatment of organoids with TGFβ. C, Confocal Z-stack images of whole, untreated, and TGFβ (50 ng/mL)-treated organoids
stained for αSMA and COL1A1. D, Reticulin staining of formalin-fixed, paraffin-embedded sections of TGFβ-treated organoids versus control. E, Meas-
urement of total reticulin stained area in untreated and TGFβ (50 ng/mL)-treated organoids. F and G, CD34 immunostaining of organoid vessels (F) and
quantification of total vascular area of organoids with/without TGFβ treatment (G). H and I, Effect of two potential inhibitors of TGFβ-induced fibrosis
(SB431542 and JQ1) on IL11 secretion (H) and ACTA2 and COL1A1 expression (I). J, αSMA and COL1A1 expression in TGFβ-treated organoids with/
without indicated inhibitors. Representative images are shown. *, P < 0.01; **, P < 0.05; ***, P < 0.001; ****, P < 0.0001 for one-way ANOVA with multiple
comparisons (Fisher least significant difference). T tests performed for image analysis of paraffin-embedded sections (reticulin and CD34). n = 3 with
each repeat comprising 15 organoids pooled from 3 independent differentiations and treatments.

FEBRUARY 2023 CANCER DISCOVERY | 377

RESEARCH ARTICLE Khan et al.

A Seed organoids with 5,000

Peripheral blood or bone marrow Fluorescent labeling of
cells isolated from adult donors adult donor–derived cells per
donor cells
well of 96-well plates

Fluorescently labeled
donor cells enabling
discrimination by
FACS or imaging
Patient-engrafted
organoids for
downstream assays

Downloaded from [Link] by guest on 13 February 2023

B C D
DAPI UEA1 CellVue TGFb ELISA COL1A1 aSMA
** **
10,000 *** 8,000 4,000
Healthy donor
TGFβ (pg/mL)

Myelofibrosis 8,000 6,000 3,000

6,000

MFI
MFI
4,000 2,000
4,000
50 mm 2,000 2,000 1,000
0 0 0

100 mm COL1A1 ACTA2 CDH5 TEK

F *** G * **
6 3 ** 1.5 0.04
E DAPI aSMA 0.03
COL1A1 aSMA COL1A1 4 2 1.0
∆∆CT

∆∆CT

∆∆CT

∆∆CT
0.02
Myelofibrosis Healthy donor

2 1 0.5
0.01
0 0 0.0 0.00

H Nonengrafted MF engrafted I

% Area vascular (CD34+)

60 ** 3 **
Reticulin

% Area reticulin

10 mm 40 2

J 20 1
MF1 MF8 MF10
CD34 (IHC)

100 100 100

JAK2 p.V617F 0 0
75 75 75
% VAF

TET2 p.S1612Lfs*4 ASXL1 c.1537-1G>C

50 50
U2AF1 p.Q157P
50 Nonengrafted control
DNMT3A p.A376T JAK2 p.V617F MF engrafted
25 25 ASXL1 p.E635Rfs*15
25
JAK2 p.V617F

0 0 ASXL1 p.R693* 0
D0 D12 D0 D12 D0 D12

Figure 6.  Engraftment of cells from patients with myelofibrosis, but not healthy donors, results in organoid “niche remodeling” and fibrosis. A, Cryo-
preserved peripheral blood or bone marrow cells from healthy donors and patients with blood cancers were fluorescently labeled, and 5,000 donor cells
were seeded into each well of a 96-well plate containing individual organoids. Schematic created using [Link]. B, Maximum-intensity projection of
confocal Z-stack of a whole engrafted organoid 72 hours after seeding of the wells with donor cells, indicating donor cells engrafted throughout the volume
of the organoids. C, Soluble TGFβ in organoids engrafted by cells from patients with myelofibrosis and controls. D–G, Comparison of organoids engrafted
with healthy donor and myelofibrosis cells for collagen 1 (COL1A1) and αSMA immunofluorescence (D) and representative images (E), Col1A1 and ACTA2
gene expression (F), and CDH5 and TIE2 expression (G). n = 4 healthy donors; n = 7 myelofibrosis samples for qRT-PCR; n = 4 healthy donors; n = 6 myelofi-
brosis samples for imaging and quantification of cryosections. MFI, mean fluorescence intensity. H and I, Increased reticulin deposition with a concomitant
reduction in vascular area in organoids engrafted with myelofibrosis (MF) cells versus nonengrafted control organoids with paired t tests; each datapoint
corresponds to a single organoid engrafted with cells from 3 donors. J, Variant allele frequencies (VAF) of mutations detected by next-generation sequenc-
ing of cells from patients with myelofibrosis before seeding in organoids (day 0) compared with cells isolated by flow cytometry 12 days after culture in
organoids, indicating maintenance of clonal architecture. (continued on following page)

378 | CANCER DISCOVERY FEBRUARY 2023 [Link]

Human Bone Marrow Organoids for Modeling Hematopoiesis RESEARCH ARTICLE

K d0–14
L aSMA
Organoid generation COL1A1
*
d14 + 0 Engraftment with MF patient *
cells
10,000 10,000 * *
* ***
Nonengrafted control
8,000 8,000
DMSO
Drug treatment SB431542 20 µmol/L 6,000 6,000

MFI
d14 + 7
(SB431542, JQ1, JQ1 0.5 µmol/L
ruxolitinib) 4,000 4,000
Ruxolitinib 0.1 µmol/L
Ruxolitinib 1 µmol/L 2,000 2,000
d14 +14 Endpoint assays (paraffin 0 0
embedding and imaging)

M
aSMA COL1A1 aSMA COL1A1

Downloaded from [Link] by guest on 13 February 2023

MF + JQ1 0.5 mmol/L
Nonengrafted

MF + Rux 0.1 mmol/L

MF + DMSO

MF + Rux 1 mmol/L
MF + SB431542

100 mm

Figure 6. (Continued) K, Workflow for organoid generation, engraftment with cells from patients with myelofibrosis, and treatment with inhibitors.
L, αSMA and collagen 1 expression in nonengrafted organoids, and organoids engrafted with myelofibrosis cells treated with DMSO (control), SB431542,
JQ1, and ruxolitinib. Each data point corresponds to total measurements per organoid within a block (n = 3 donors). One-way ANOVA with multiple com-
parisons (Fisher least significant difference). M, Representative images from L. Rux, ruxolitinib. *, P < 0.01; **, P < 0.05; ***, P < 0.001 for Mann–Whitney
test. See also Supplementary Fig. S8.

We compared the survival and proliferation of primary engraftment into organoids for all 5 donors tested (Fig.  7C
multiple myeloma, ALL, and Xeno iALL cells engrafted in and D). Similarly, the survival of primary ALL cells was sig-
the organoids with cells seeded in wells with media alone nificantly improved in the organoids compared with liquid
or into a single-lineage 3D coculture system containing pri- culture (Fig. 7C). The cell viability for ALL and Xeno iALL was
mary human bone marrow MSCs in a Matrigel  +  collagen also higher in the organoids than in 3D BM-MSC cocultures
I hydrogel (3D BM-MSC). Whereas multiple myeloma cells (Fig.  7C), with higher proliferation rates (Fig.  7D and E).
were less than 20% viable only 48 hours after seeding in Plasma cells from patients with myeloma showed minimal cell
wells with media alone, in stark contrast, the myeloma cells division in either model (Fig. 7D), but the cells retained their
expanded and remained >90% viable more than 12 days after original immunophenotype (CD38+, CD319+, and CD56+)

FEBRUARY 2023 CANCER DISCOVERY | 379

RESEARCH ARTICLE Khan et al.

CD38 UEA1 CellVue CellVue CD38

A Control Xeno iALL B
Control
(no seeding) (no seeding)
DAPI UEA1 CellVue

CML THP-1 Multiple myeloma

100 mm 100 mm

Downloaded from [Link] by guest on 13 February 2023

C Multiple myeloma ALL Xeno iALL

*** *** *** *** ns * * * 100 ns *** *** ***

% CellTrace+ live cells

% CellTrace+ live cells

% CellTrace+ live cells

100 100
80
80 80
Viability

60 60 60

40 40 40
20 20 20
0 0 0
0 5 10 15 0 5 10 15 0 5 10 15
Days Days Days

D ns * * * ns * * * ns ns ns *
CellTrace relative to D2

2−2
CellTrace relative to D2

2−5
CellTrace relative to D2
2−3
2−4
Proliferation

2−3 2−2
2−1 2−2
2−1 2−1
20
20 21 20
0 5 10 15 0 5 10 15 0 5 10 15
Days Days Days
BM organoid 3D BM-MSC coculture Liquid culture

E 3D BM-MSC Organoid F
D2 Multiple myeloma BM organoid 3D BM-MSC coculture
myeloma
Multiple

D5 100
Patient sample D12 BM organoid
D7
%CD319+

D12 90
80
70
Xeno iALL

CD319

60
D2 D5 D7 D12
FSC
100 105
90 100
%CD38+

%CD56+

95
80
Count ALL

90
CD56

70 85
60 80
FSC D7 D12 D7 D12
CellTrace

380 | CANCER DISCOVERY FEBRUARY 2023 [Link]

Human Bone Marrow Organoids for Modeling Hematopoiesis RESEARCH ARTICLE

more consistently in the organoids than in the 3D BM-MSC certain cellular subsets (e.g., stroma) from wild-type iPSCs,
(Fig. 7F). Similarly, ALL and Xeno iALL cells showed improved and then assembling these with hematopoietic cells gener-
maintenance of CD19 expression in the organoids as com- ated from a genetically modified parent iPSC line.
pared with 3D BM-MSC (Supplementary Fig. S12). Together, A key limitation of current human ex vivo bone marrow
these data confirm that the organoids provide a supportive models has been a lack of sinusoidal-like endothelium, with
niche for the survival and growth of primary blood cancer cells many systems reliant on human umbilical vein endothelial
from patients, including for cancer types that are otherwise cells (HUVEC; ref.  13). We show here that the addition of
poorly viable ex vivo and after cryopreservation. VEGFC, recently shown to support the bone marrow perivas-
cular niche (28), drives the generation of vasculature and
supporting stroma that are specialized for hematopoietic
DISCUSSION support and phenocopy bone marrow sinusoidal endothe-
Here we describe the development of a protocol gener- lium. The resulting bone marrow organoids thereby present
ating vascularized bone marrow organoids that faithfully a unique opportunity to study the hematopoietic-niche cross-
model key cellular, molecular, and architectural features of talk that underpins healthy hematopoiesis, and how pertur-
myelopoietic bone marrow including stromal cells, lumen- bations to these regulatory interactions are permissive for the
forming vasculature, and myeloid cell types. We demonstrate emergence and progression of cancers.
the utility of these organoids for modeling cancer-induced Although this system offers a substantial advance in the

Downloaded from [Link] by guest on 13 February 2023

perturbations to the bone marrow niche and myelofibrosis. field, in its current iteration, no osteoid lineage, lymphoid
Treatment of organoids with TGFβ, the primary cytokine cells, smooth muscle cells, or adipocytes are generated.
driving myelofibrosis, induced organoid fibrosis, enabling Similarly, although we show a high degree of homology of
target prioritization and screening of potential inhibitors. the organoid vasculature to sinusoidal endothelium from
Fibrosis also occurred following the engraftment of orga- human bone marrow, distinct arteriolar, and sinusoidal
noids with HSPCs from patients with myelofibrosis, but not endothelial subtypes are not present in the organoids. The
healthy donors. The ability to reliably model bone marrow current differentiation protocol was optimized primarily
fibrosis is an important advance, as the lack of adequate to study myeloid malignancies and cancer-associated bone
in vitro and in vivo models currently hampers the efficient pre- marrow fibrosis, and refinement of the growth factor sup-
clinical validation of strategies aiming to reduce or prevent plements may improve the maintenance of lymphoid malig-
fibrosis, which is a huge unmet need for patients with myelo- nancies. In addition, the introduction of recirculating flow
proliferative neoplasms and other blood cancers (45). As the (46) may allow for the generation of organoids that mimic
organoids are highly reproducible and feasible to generate at native bone marrow physiology more comprehensively.
scale in 96- or 384-well plate formats, this system presents Despite these limitations, maintenance of cells from B-cell
an ideal platform for high-throughput target screens using leukemias and plasma cell malignancies as well as myeloid
pharmacologic or genetic modulation. cancers provides proof of principle that bone marrow orga-
We also show that cells from patients with both myeloid noids can be used to support a range of bone marrow
and lymphoid malignancies readily engraft and survive within cancers and cell types, paving the way for customization to
the organoids, including cancer cell types that are notoriously support other relevant studies.
difficult to maintain ex vivo. Remarkably, malignant cells A protocol for the ectopic implantation of human bone
from patients with multiple myeloma that had been cryopre- marrow MSC-derived ossicles has previously been shown to
served and thawed prior to use were sustained by the orga- support the engraftment of adult HSCs in vivo (47). Although
noids for 12 days, while rapidly losing viability when plated the iPSC-derived organoids are not an in vivo system, they are
in vitro without stromal support. Maintenance of primary substantially more efficient to generate than in vivo ossicles
myeloma cells ex vivo and a method to study their interactions (weeks vs. months) and do not require human bone mar-
within a multicellular hematopoietic niche environment will row or platelet lysates, which are hard to source and may
enable preclinical pharmacogenomic screens and the study of induce experimental variability. The survival and continuous
disease mechanisms using primary cells from patients, which production of hematopoietic support factors in the absence
is currently a significant obstacle to translational research of exogenous cytokine supplementation suggest that longer-
in this disease. Targeted genetic modification of organoids term cultures may be possible. The implantation of orga-
could be performed either by using lineage-specific promot- noids into mice has not yet been tested but may allow for
ers or by modulating the differentiation protocol to generate longer-term studies in an in vivo setting.

Figure 7.  Bone marrow organoids support the engraftment, survival, and proliferation of cells from patients with myeloid and lymphoid hemato-
logic malignancies. A, Organoids engrafted with CellVue-labeled model infant ALL cells from xenografts (Xeno iALL), primary cells from a patient with
untreated CML, and THP-1 cells, an acute myeloid leukemia cell line. CellVue+ cells are visible throughout the volume of organoids. B, Organoids seeded
with CD138+ cells isolated from bone marrow aspirates of patients with multiple myeloma show CellVue+ CD38+ plasma cell engraftment. C–E, Viability
and proliferation of cells from 4 donors with multiple myeloma, 6 donors with ALL, and 3 Xeno iALL samples seeded simultaneously in the organoids, a
3D coculture with primary human BM-MSC (3D BM-MSC), and where possible, liquid culture. E, Serial dilution of CellTrace label, indicating cell prolifera-
tion, for multiple myeloma, Xeno iALL, and ALL cells in 3D BM-MSC and organoids on days 2, 5, 7, and 12 following thawing and plating. F, Engrafted
multiple myeloma cells retained their immunophenotype at day 12 with more consistent maintenance of CD319 and CD38 in organoids than 3D BM-MSC.
Representative images are shown. *, P < 0.01; ***, P < 0.001; ns, not significant. n = 4 for multiple myeloma, n = 3 for Xeno iALL, and n = 3 for ALL, with
each repeat comprised of a separate donor two-way ANOVA with repeated measures and multiple comparisons (organoid cultures vs. 3D BM-MSC; Fisher
least significant difference) for ALL and multiple myeloma and multiple unpaired t test for Xeno iALL data. BM, bone marrow; FSC, forward scatter.

FEBRUARY 2023 CANCER DISCOVERY | 381

RESEARCH ARTICLE Khan et al.

The development of organoids has been transformative in Samples were collected by gravitation in a 15-mL Falcon tube and
other disease settings—for example, cerebral, lung, and kidney washed 2× in PBS and then resuspended in collagenase type II. For
disease modeling (48). This platform may similarly be an ena- dissociation, samples were incubated at 37°C for 5 minutes before
bling technology for the interrogation of disease mechanisms trituration and a further 5-minute incubation. The dissociation reac-
tion was stopped through the addition of PBS supplemented with
in hematologic cancers as well as the development and test-
FBS. Ten organoids were dissociated per flow cytometry experiment.
ing of novel therapies using human cells in a tissue-relevant Analysis was performed using either a cyan flow cytometer (Beckman
system. Importantly, this platform is likely to reduce reliance Coulter) or an Attune NxT. Single color stained controls and flu-
on animal models. Target identification and screening using orescence-minus-one controls were used for all experiments using
a species-specific, clinically relevant ex vivo model that can antibodies as listed in Supplementary Table S5.
incorporate primary cells from patients may accelerate and
increase the success rate of clinical translation. TGFβ Treatment to Induce Organoid Fibrosis
Organoids were treated with TGFβ (PeproTech, cat. #100-21) at 10,
25, or 50 ng/mL for 72 hours. At 72 hours, all samples were approxi-
METHODS mately 90% viable after collagenase digestion. Samples were then
iPSC Culture and Differentiation collected for either whole-mount microscopy, paraffin/OCT embed-
A Gibco Human Episomal iPSC (Thermo Fisher Scientific, cat. ding and sectioning, or qRT-PCR. Thirty-two organoids were treated
#A18945) line was maintained in StemFlex medium (Thermo Fisher per replicate; of these, 16 were spun down for RNA extraction for

Downloaded from [Link] by guest on 13 February 2023

Scientific, cat. #A3349401) and on Geltrex (Thermo Fisher Scientific, subsequent qRT-PCR. The remaining 16 organoids were fixed, with
cat. #A1569601)-coated 6-well plates. The iPSC line was karyotyped 8 taken for whole volume imaging and 8 per repeat taken for embed-
prior to use (23), and potency markers were assessed upon expansion ding and sectioning. For drug treatment, samples were treated with
and freezing. A full, detailed description of passaging and differentia- DMSO or TGFβ at a concentration of 25 ng/mL, and supplemented
tion protocols is included in Supplementary Materials and Methods. with an inhibitor as described (either JQ1 at 0.5 μmol/L or SB431542
In brief, for differentiations, iPSCs were dissociated using EDTA at 20 μmol/L).
when colonies were approximately 100 μm in diameter. The resulting
iPSC aggregates were incubated overnight in StemFlex supplemented qRT-PCR
with RevitaCell in 6-well Co-star ULA plates (Corning, cat. #3471; Whole organoids were processed using either the Micro RNeasy
day  −1). After overnight incubation, cells were collected and resus- Kit (Qiagen, cat. #74004) or Qiagen Mini RNA isolation kit (Qiagen,
pended in phase I medium (Supplementary Methods) and cultured cat. #74104) as per kit instructions. cDNA was prepared using the
at 5% O2 for 3 days (days 0–3). Aggregated cells were then collected High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems,
again in phase II medium (Supplementary Methods). On day 5, cells cat. #4368814) or EvoScript Universal cDNA Master (Roche, cat.
were collected by gravitation for hydrogel embedding. Hydrogels #07912374001), and RT-PCR was performed using PowerUp SYBR
were composed of 60% collagen (either type I, type IV, or an equal Green Master Mix reagent (Applied Biosystems, cat. #A25742) or
parts type I + IV mix) and 40% Matrigel as detailed in Supplementary TaqMan Universal PCR Master Mix (Applied Biosystems; see Sup-
Methods. Fully polymerized gels with cell aggregates were then sup- plementary Table S6 for a list of primers).
plemented with phase III media (Supplementary Methods). Media
were replenished every 72 hours. Seeding of Organoids and 3D BM-MSC Cultures with
Hematopoietic Cells
Immunofluorescence Staining Peripheral blood and bone marrow samples were collected from
Sections were blocked using 2% goat serum (Thermo Fisher Sci- healthy donors and patients with hematologic malignancies follow-
entific, cat. #31872) and 1% bovine serum albumin (BSA; Sigma, cat. ing the provision of written informed consent in accordance with the
#A9418) prior to primary antibody labeling with antibody diluted Declaration of Helsinki. All participants donated human tissue for
in 1% BSA, sequential PBS washes, and secondary labeling with research without receiving monetary compensation, and the stud-
AlexaFluor conjugates. The whole organoid blocking solution was ies were approved by an institutional review board [myelofibrosis,
supplemented with Triton X100, Tween, and sodium deoxycholate CML, and G-CSF mobilized healthy apheresis donors—INForMed
as described by Wimmer and colleagues (21). Antibodies are listed Study, University of Oxford (IRAS: 199833; REC 16/LO/1376);
in Supplementary Table  S5, and additional details are given in multiple myeloma—Oxford Radcliffe Biobank (Oxford Clini-
Supplementary Methods. cal Research Ethics Committee 09/H0606/5/5, project 16/A185);
ALL samples: REC 16/LO/2055′ (IRAS 179685)]. Written informed
scRNA-seq, Data Processing, and Analysis consent was received from all the participants for the donation of
human tissue.
Cryopreserved cells pooled from 15 organoids from 3 differentia-
For engraftments with CD34+ HSPCs, cryopreserved mononuclear
tions from both VEGFA and VEGFA  +  C protocols were processed
cells were thawed and CD34+ viable cells were FACS isolated using a
for single-cell sequencing as described in Supplementary Methods,
Becton Dickinson FACSAria Fusion Cell Sorter with 100-nm noz-
and processed using the Chromium Single-Cell 3′  library and Gel
zles into 1.5  mL Eppendorf tubes prior to seeding in organoids.
Bead Kits v3.1 (10X Genomics) as per kit instructions. Demultiplexed
For inhibitor experiments, organoids were engrafted with CD34+
FASTQ files were aligned to the human reference genome (GRCh38/
myelofibrosis cells and cultured for 7 days prior to the addition of
hg38) using standard CellRanger (version 6.0.1) “cellranger count”
inhibitors. Myeloma cells were selected from total bone marrow
pipeline (10X Genomics). SingCellaR (ref.  31; [Link]
mononuclear cells using anti-CD138 magnetic bead enrichment
[Link]/[Link]/) was used for the downstream analysis.
(STEMCELL Technologies, cat. #17887) prior to cryopreservation.
Xeno iALL cells were derived from a recently published xenograft
Flow Cytometry model of infant ALL in which the t(4;11)/MLL–AF4 translocation
Organoids were dissociated for flow cytometry analysis using col- was introduced into primary human fetal liver hematopoietic cells
lagenase type II (Sigma-Aldrich, cat. #C6885) resuspended in HEPES by CRISPR–Cas9 gene editing prior to transplantation into immuno-
buffer (Sigma-Aldrich, cat. #H0887) at a concentration of 20 mg/mL. deficient mice (49). All experiments were performed under a project

382 | CANCER DISCOVERY FEBRUARY 2023 [Link]

Human Bone Marrow Organoids for Modeling Hematopoiesis RESEARCH ARTICLE

license approved by the UK Home Office under the Animal (Scientific investigation. J.S. Reyat: Investigation, methodology, writing–review
Procedures) Act 1986 after approval by the Oxford Clinical Medicine and editing. A.-A. Olijnik: Investigation, writing–review and editing.
Animal Welfare and Ethical Review Body and in accordance with the M. Colombo: Formal analysis, investigation, writing–review and edit-
principles of 3Rs (replacement, reduction, and refinement) in animal ing. G. Wang: Data curation, software, formal analysis, investigation,
research. Cells were harvested from the bone marrow of leukemic visualization, writing–review and editing. W.X. Wen: Software, inves-
mice at 17 to 18 weeks, and total bone marrow cells were cryopre- tigation, methodology, writing–review and editing. N. Sousos: For-
served. Following thawing and prior to seeding in the organoids, mal analysis, investigation, writing–review and editing. L.C. Murphy:
human CD45+ were selected using magnetic microbeads (Miltenyi Investigation. B. Grygielska: Investigation. G. Perrella: Formal
Biotec, cat. #130-045-801). Over 90% of human CD45+ cells were analysis, investigation. C.B. Mahony: Software, investigation,
CD19+ CD34+ lymphoblasts with a predominantly CD34+ pro-B phe- writing–review and editing. R.E. Ling: Formal analysis, investigation,
notype (CD10-CD20-IgM/IgD−; Supplementary Fig. S10). Informed writing–review and editing. N.E. Elliott: Formal analysis, investiga-
consent was provided by all participants for the donation of human tion. C. Simoglou Karali: Investigation. A.P. Stone: Investigation.
tissue, and this study was approved by an institutional review board S. Kemble: Software, investigation. E.A. Cutler: Resources, validation,
(REC: 18/NE/0290 and 18/LO/0822). The replicates included in this writing–original draft. A.K. Fielding: Conceptualization, resources,
study were from 4 mice transplanted with MLL–AF4-edited cells writing–review and editing. A.P. Croft: Supervision. D. Bassett:
from one human fetal liver sample. Details of donor cell labeling are Supervision, funding acquisition. G. Poologasundarampillai:
included in Supplementary Materials and Methods. The composi- Funding acquisition, methodology. A. Roy: Resources, investigation,
tion of the engrafted organoids was analyzed by flow cytometry using writing–review and editing. S. Gooding: Resources, investigation,
writing–review and editing. J. Rayes: Formal analysis, investigation,

Downloaded from [Link] by guest on 13 February 2023

either an LSRFortessa X50 (BD Biosciences) or an Attune NxT or a
next-generation sequencing panel (see Supplementary Methods and writing–review and editing. K.R. Machlus: Resources, supervision,
Supplementary Table S7). funding acquisition, writing–review and editing. B. Psaila: Concep-
For the 3D BM-MSC cocultures, 24-well plates were prepared tualization, resources, formal analysis, supervision, funding acquisi-
containing 300 μL of a 70:30 mix of collagen I:Matrigel per well and tion, methodology, writing–original draft, project administration,
incubated at 37°C for 90 minutes. Primary bone marrow MSCs resus- writing–review and editing.
pended at 10,000 cells/mL in StemPro-34 supplemented medium
were added to the wells (500  μL/well) and incubated overnight prior Acknowledgments
to seeding with primary hematopoietic cells. A.O. Khan is funded by a Sir Henry Wellcome fellowship
(218649/Z/19/Z). K.R. Machlus is supported by grants from the
Data Analysis Software NIH, the National Institute of Diabetes and Digestive and Kidney
scRNA-seq analyses were performed in R Studio (version 1.4.1106). Diseases (R03DK124746), and the National Heart, Lung, and Blood
Other statistical analyses were performed using GraphPad PRISM Institute (R01HL151494) and is an American Society of Hema-
7 with statistical tests as described in relevant figure legends. tology Scholar. G. Wang was supported by an Oxford Centre for
P values are defined as *, P  <  0.01; **, P  <  0.05; ***, P  <  0.001; Haematology Pump Priming Award and a Medical Science Division
****, P < 0.0001. Pump Priming Award (009800) from the Nuffield Benefaction for
Medicine and the Wellcome Institutional Strategic Support Fund
Data Availability (ISSF). A.K. Fielding is supported by the Cancer Research UK award
scRNA-seq data are available at the Gene Expression Omnibus C27995/A21019 (also awarded to Anthony V. Moorman), which
(accession GSE196684). Scripts used for analysis are available at allowed for biobanking and specimen collection from UKALL14. G.
[Link] and [Link] Perrella is supported by funding from the Engineering and Physical
supatt-lab/SingCellaR. Further data are available on request. Sciences Research Council (EP/V051342/1). S. Gooding receives
funding from a Cancer Research UK Clinician Scientist Fellow-
Authors’ Disclosures ship (RCCCSF-Nov21\100004) and unit funding from the Medical
Research Council (awarded to the MRC Molecular Haematology
A.O. Khan and B. Psaila are coinventors on a pending patent
Unit, MRC Weatherall Institute of Molecular Medicine). B. Psaila
(GB2202025.9 and GB2216647.4) relating to data in this article.
receives funding from a Cancer Research UK Advanced Clinician
A.-A. Olijnik reports grants from Cancer Research UK during the con-
Scientist Fellowship (C67633/A29034), a British Research Council
duct of the study. N. Sousos reports personal fees from Bristol Myers
(BRC) Senior Research Fellowship, the Haematology and Stem
Squibb, and grants from Bristol Myers Squibb and AOP Orphan out-
Cells Theme of the Oxford BRC, a Kay Kendall Leukemia Fund
side the submitted work. C. Simoglou Karali reports grants and per-
Project Grant, and unit funding from the Medical Research Council
sonal fees from Bristol Myers Squibb outside the submitted work. D.
(awarded to the MRC Molecular Haematology Unit, MRC Weather-
Bassett reports other support from CLEXBio outside the submitted
all Institute of Molecular Medicine). A CC BY or equivalent license
work. S. Gooding reports grants from Bristol Myers Squibb, Innovate
is applied to the author accepted manuscript arising from this sub-
UK (UKRI), and Cancer Research UK outside the submitted work.
mission in accordance with the funders’ open access conditions. We
B. Psaila reports grants from Cancer Research UK Advanced Clinician
thank all the patients who kindly donated samples and Nawshad
Scientist Fellowship and British Research Council Senior Research
Hayder and Sophie Reed who helped with sample banking; the Uni-
Fellowship during the conduct of the study, as well as grants, per-
versity of Birmingham TechHub Imaging Core Facilities and staff;
sonal fees, and other support from Alethiomics, personal fees from
the MRC WIMM Flow Cytometry Facility; the MRC WIMM Single
Novartis and Blueprint Medicines, and grants from Galecto and Con-
Cell Facility; Daniel Royston for input and advice on bone marrow
stellation Therapeutics outside the submitted work. No disclosures
architecture and histology; Zewen Kelvin Tuong for assistance with
were reported by the other authors.
KT Plots Scripting; and Arturs Habirovs and Ashleigh Danks for
Authors’ Contributions computational advice.
A.O. Khan: Conceptualization, resources, data curation, formal The publication costs of this article were defrayed in part by the
analysis, funding acquisition, validation, investigation, visualization, payment of publication fees. Therefore, and solely to indicate this
methodology, writing–original draft, project administration, writing– fact, this article is hereby marked “advertisement” in accordance with
review and editing. A. Rodriguez-Romera: Formal analysis, 18 USC section 1734.

FEBRUARY 2023 CANCER DISCOVERY | 383

RESEARCH ARTICLE Khan et al.

Note 17. Bessy T, Itkin T, Passaro D. Bioengineering the bone marrow vascular
niche. Frontiers Cell Dev Biology 2021;9:645496.
Supplementary data for this article are available at Cancer Discovery 18. Bourgine PE, Klein T, Paczulla AM, Shimizu T, Kunz L, Kokkaliaris KD,
Online ([Link] et  al. In vitro biomimetic engineering of a human hematopoietic
Received February 21, 2022; revised October 4, 2022; accepted niche with functional properties. Proc National Acad Sci 2018;115:
November 7, 2022; published first November 9, 2022. E5688–95.
19. Adamo A, Delfino P, Gatti A, Bonato A, Kamga PT, Bazzoni R, et al.
HS-5 and HS-27A stromal cell lines to study bone marrow mesenchy-
mal stromal cell-mediated support to cancer development. Frontiers
REFERENCES Cell Dev Biology 2020;8:584232.
1. Baryawno N, Przybylski D, Kowalczyk MS, Kfoury Y, Severe N, 20. Veninga V, Voest EE. Tumor organoids: Opportunities and challenges
Gustafsson K, et al. A cellular taxonomy of the bone marrow stroma to guide precision medicine. Cancer Cell 2021;39:1190–201.
in homeostasis and leukemia. Cell 2019;177:1915–32. 21. Wimmer RA, Leopoldi A, Aichinger M, Wick N, Hantusch B,
2. Méndez-Ferrer S, Bonnet D, Steensma DP, Hasserjian RP, Ghobrial IM, Novatchkova M, et al. Human blood vessel organoids as a model of
Gribben JG, et al. Bone marrow niches in haematological malignan- diabetic vasculopathy. Nature 2019;565:505–10.
cies. Nat Rev Cancer 2020;20:285–98. 22. Feng Q, Shabrani N, Thon JN, Huo H, Thiel A, Machlus KR, et  al.
3. Tikhonova AN, Dolgalev I, Hu H, Sivaraj KK, Hoxha E, Cuesta- Scalable generation of universal platelets from human induced pluri-
Domínguez Á, et al. The bone marrow microenvironment at single- potent stem cells. Stem Cell Rep 2014;3:817–31.
cell resolution. Nature 2019;569:222–8. 23. Khan AO, Slater A, Maclachlan A, Nicolson PLR, Pike JA, Reyat JS,
et  al. Post-translational polymodification of β1-tubulin regulates

Downloaded from [Link] by guest on 13 February 2023

4. Leimkühler NB, Gleitz HFE, Ronghui L, Snoeren IAM, Fuchs SNR,
Nagai JS, et  al. Heterogeneous bone-marrow stromal progenitors motor protein localisation in platelet production and function.
drive myelofibrosis via a druggable alarmin axis. Cell Stem Cell Haematologica 2020;107:243–60.
2021;28:637–52. 24. Itkin T, Gur-Cohen S, Spencer JA, Schajnovitz A, Ramasamy SK,
5. Celik H, Krug E, Zhang CR, Han W, Issa N, Koh WK, et  al. Kusumbe AP, et al. Distinct bone marrow blood vessels differentially
A humanized animal model predicts clonal evolution and therapeu- regulate hematopoiesis. Nature 2016;532:323–8.
tic vulnerabilities in myeloproliferative neoplasms. Cancer Discov 25. Kopp HG, Avecilla ST, Hooper AT, Rafii S. The bone marrow vascular
2021;11:3126–41. niche: home of HSC differentiation and mobilization. Physiology
6. Lysenko V, Wildner-Verhey van Wijk N, Zimmermann K, Weller MC, 2005;20:349–56.
Bühler M, Wildschut MHE, et  al. Enhanced engraftment of human 26. Apte RS, Chen DS, Ferrara N. VEGF in signaling and disease: beyond
myelofibrosis stem and progenitor cells in MISTRG mice. Blood Adv discovery and development. Cell 2019;176:1248–64.
2020;4:2477–88. 27. Durán JGB, Lu T, Houghton S, Geng F, Schreiner R, Xiang J, et  al.
7. Nicolini FE, Cashman JD, Hogge DE, Humphries RK, Eaves CJ. Endothelial Jak3 expression enhances pro-hematopoietic angiocrine
NOD/SCID mice engineered to express human IL-3, GM-CSF and function in mice. Commun Biology 2021;4:406.
Steel factor constitutively mobilize engrafted human progenitors 28. Fang S, Chen S, Nurmi H, Leppänen VM, Jeltsch M, Scadden D, et al.
and compromise human stem cell regeneration. Leukemia 2004;18: VEGF-C protects the integrity of the bone marrow perivascular niche
341–7. in mice. Blood 2020;136:1871–83.
8. Rongvaux A, Willinger T, Martinek J, Strowig T, Gearty SV, 29. Junt T, Schulze H, Chen Z, Massberg S, Goerge T, Krueger A,
Teichmann LL, et  al. Development and function of human innate et al. Dynamic visualization of thrombopoiesis within bone marrow.
immune cells in a humanized mouse model. Nat Biotechnol Science 2007;317:1767–70.
2014;32:364–72. 30. Psaila B, Wang G, Rodriguez-Meira A, Li R, Heuston EF, Murphy L,
9. Willinger T, Rongvaux A, Takizawa H, Yancopoulos GD, Valenzuela DM, et al. Single-cell analyses reveal megakaryocyte-biased hematopoiesis
Murphy AJ, et  al. Human IL-3/GM-CSF knock-in mice support in myelofibrosis and identify mutant clone-specific targets. Mol Cell
human alveolar macrophage development and human immune 2020;78:477–92.
responses in the lung. Proc National Acad Sci 2011;108:2390–5. 31. Roy A, Wang G, Iskander D, O’Byrne S, Elliott N, O’Sullivan J, et al.
10. Wunderlich M, Chou FS, Link KA, Mizukawa B, Perry RL, Carroll M, Transitions in lineage specification and gene regulatory networks in
et  al. AML xenograft efficiency is significantly improved in NOD/ hematopoietic stem/progenitor cells over human development. Cell
SCID-IL2RG mice constitutively expressing human SCF, GM-CSF Rep 2021;36:109698.
and IL-3. Leukemia 2010;24:1785–8. 32. JMME de K Z, Papazian N, Tahri S, BDH op, Hoogenboezem R, et al.
11. Das R, Strowig T, Verma R, Koduru S, Hafemann A, Hopf S, et  al. The multiple myeloma microenvironment is defined by an inflamma-
Microenvironment-dependent growth of preneoplastic and malig- tory stromal cell landscape. Nat Immunol 2021;22:769–80.
nant plasma cells in humanized mice. Nat Med 2016;22:1351–7. 33. Popescu DM, Botting RA, Stephenson E, Green K, Webb S,
12. Chou DB, Frismantas V, Milton Y, David R, Pop-Damkov P, Jardine L, et al. Decoding human fetal liver haematopoiesis. Nature
Ferguson D, et  al. On-chip recapitulation of clinical bone-marrow 2019;574:365–71.
toxicities and patient-specific pathophysiology. Nat Biomed Eng 34. Kang JB, Nathan A, Weinand K, Zhang F, Millard N, Rumker L, et al.
2020;4:394–406. Efficient and precise single-cell reference atlas mapping with Sym-
13. Janagama D, Hui SK. 3-D cell culture systems in bone marrow tis- phony. Nat Commun 2021;12:5890.
sue and organoid engineering, and BM phantoms as in vitro mod- 35. Jardine L, Webb S, Goh I, Londoño MQ, Reynolds G, Mather M, et al.
els of hematological cancer therapeutics—a review. Materials 2020; Blood and immune development in human fetal bone marrow and
13:5609. Down syndrome. Nature 2021;598:327–31.
14. Raic A, Naolou T, Mohra A, Chatterjee C, Lee-Thedieck C. 3D mod- 36. Sweeney M, Foldes G. It takes two: endothelial-perivascular cell
els of the bone marrow in health and disease: yesterday, today, and cross-talk in vascular development and disease. Frontiers Cardiovasc
tomorrow. Mrs Commun 2019;9:37–52. Medicine 2018;5:154.
15. Buduo CAD, Wray LS, Tozzi L, Malara A, Chen Y, Ghezzi CE, 37. Hirano K, Suganami A, Tamura Y, Yagita H, Habu S, Kitagawa M,
et al. Programmable 3D silk bone marrow niche for platelet genera- et  al. Delta-like 1 and Delta-like 4 differently require their extracel-
tion ex vivo and modeling of megakaryopoiesis pathologies. Blood lular domains for triggering Notch signaling in mice. Elife 2020;9:
2015;125:2254–64. e50979.
16. Tozzi L, Laurent P-A, Buduo CAD, Mu X, Massaro A, Bretherton R, 38. Ding W, Knox TR, Tschumper RC, Wu W, Schwager SM, Boysen JC,
et  al. Multi-channel silk sponge mimicking bone marrow vascular et  al. Platelet-derived growth factor (PDGF)–PDGF receptor inter-
niche for platelet production. Biomaterials 2018;178:122–33. action activates bone marrow–derived mesenchymal stromal cells

384 | CANCER DISCOVERY FEBRUARY 2023 [Link]

Human Bone Marrow Organoids for Modeling Hematopoiesis RESEARCH ARTICLE

derived from chronic lymphocytic leukemia: implications for an 44. Morsia E, Gangat N. Myelofibrosis: challenges for preclinical models
angiogenic switch. Blood 2010;116:2984–93. and emerging therapeutic targets. Expert Opin Ther Tar 2021;25:1–12.
39. Mirshekar-Syahkal B, Haak E, Kimber GM, LK van, Harvey K, 45. Waksal JA, Harrison CN, Mascarenhas JO. Novel therapeutics and
O’Rourke J, et  al. Dlk1 is a negative regulator of emerging hemat- targets in myelofibrosis. Leuk Lymphoma 2022;63:1020–33.
opoietic stem and progenitor cells. Haematologica 2013;98:163–71. 46. Homan KA, Gupta N, Kroll KT, Kolesky DB, Skylar-Scott M,
40. Kuter DJ, Bain B, Mufti G, Bagg A, Hasserjian RP. Bone marrow Miyoshi T, et  al. Flow-enhanced vascularization and maturation of
fibrosis: pathophysiology and clinical significance of increased bone kidney organoids in vitro. Nat Methods 2019;16:255–62.
marrow stromal fibres. Brit J Haematol 2007;139:351–62. 47. Reinisch A, Thomas D, Corces MR, Zhang X, Gratzinger D, Hong WJ,
41. Biernacka A, Dobaczewski M, Frangogiannis NG. TGF-β signaling in et al. A humanized bone marrow ossicle xenotransplantation model
fibrosis. Growth Factors 2011;29:196–202. enables improved engraftment of healthy and leukemic human
42. Kramann R, Schneider RK. The identification of fibrosis-driving hematopoietic cells. Nat Med 2016;22:812–21.
myofibroblast precursors reveals new therapeutic avenues in myelofi- 48. Bock C, Boutros M, Camp JG, Clarke L, Clevers H, Knoblich JA, et al.
brosis. Blood 2018;131:2111–9. The organoid cell atlas. Nat Biotechnol 2021;39:13–7.
43. Cominal JG, Cacemiro M da C, Berzoti-Coelho MG, Pereira IEG, 49. Rice S, Jackson T, Crump NT, Fordham N, Elliott N, O’Byrne S, et al.
Frantz FG, Souto EX, et al. Bone marrow soluble mediator signatures A human fetal liver-derived infant MLL-AF4 acute lymphoblastic
of patients with philadelphia chromosome-negative myeloprolifera- leukemia model reveals a distinct fetal gene expression program. Nat
tive neoplasms. Frontiers Oncol 2021;11:665037. Commun 2021;12:6905.

Downloaded from [Link] by guest on 13 February 2023

FEBRUARY 2023 CANCER DISCOVERY | 385