Acute myeloid leukemia does not deplete normalhematopoietic stem cells but induces cytopeniasby impeding their differentiationFarideh Miraki-Mouda, Fernando Anjos-Afonsob, Katharine A. Hodbya, Emmanuel Griessingerb, Guglielmo Rosignolic,Debra Lillingtond, Li Jiaa, Jeff K. Daviesa, Jamie Cavenagha, Matthew Smitha, Heather Oakerveea, Samir Agrawala,John G. Gribbena, Dominique Bonnetb,1,2, and David C. Taussiga,1,2

aDepartment of Haemato-Oncology, Barts Cancer Institute, Queen Mary University of London, London EC1M 6BQ, United Kingdom; bHaematopoietic StemCell Laboratory, Cancer Research UK London Research Institute, London WC2A 3LY, United Kingdom; cFlow Cytometry Core Facility, Barts and the LondonSchool of Medicine and Dentistry, Queen Mary University of London, London EC1M 6BQ, United Kingdom; and dCytogenetics Department, Royal LondonHospital, London E1 2ES, United Kingdom

Edited by Stuart H. Orkin, Children’s Hospital and the Dana Farber Cancer Institute, Howard Hughes Medical Institute and Harvard Medical School, Boston,MA, and approved July 9, 2013 (received for review January 29, 2013)

Acute myeloid leukemia (AML) induces bone marrow (BM) failurein patients, predisposing them to life-threatening infections andbleeding. The mechanism by which AML mediates this complica-tion is unknown but one widely accepted explanation is that AMLdepletes the BM of hematopoietic stem cells (HSCs) throughdisplacement. We sought to investigate how AML affects hema-topoiesis by quantifying residual normal hematopoietic subpopu-lations in the BM of immunodeficient mice transplanted withhuman AML cells with a range of genetic lesions. The numbers ofnormal mouse HSCs were preserved whereas normal progenitorsand other downstream hematopoietic cells were reduced follow-ing transplantation of primary AMLs, findings consistent witha differentiation block at the HSC–progenitor transition, ratherthan displacement. Once removed from the leukemic environ-ment, residual normal hematopoietic cells differentiated normallyand outcompeted steady-state hematopoietic cells, indicating thatthis effect is reversible. We confirmed the clinical significance ofthis by ex vivo analysis of normal hematopoietic subpopulationsfrom BM of 16 patients with AML. This analysis demonstrated thatthe numbers of normal CD34+CD38− stem-progenitor cells weresimilar in the BM of AML patients and controls, whereas normalCD34+CD38+ progenitors were reduced. Residual normal CD34+

cells from patients with AML were enriched in long-term culture,initiating cells and repopulating cells compared with controls. Inconclusion the data do not support the idea that BM failure in AMLis due to HSC depletion. Rather, AML inhibits production of down-stream hematopoietic cells by impeding differentiation at theHSC–progenitor transition.

xenotransplant | anemia | thrombocytopenia | neutropenia

Hematopoiesis is tightly regulated under normal circum-stances to ensure adequate production of mature blood

cells. At steady state hematopoietic stem cells (HSCs) are rela-tively quiescent and the majority of proliferation occurs down-stream of HSCs. Hematopoietic stresses such as bleeding orbone marrow (BM) damage from chemotherapy induce HSCs toenter the cell cycle to replenish mature blood cells (1–3).BM failure (reduced production of neutrophils, red cells, and

platelets) is almost universal at diagnosis of acute myeloid leu-kemia (AML) and contributes significantly to morbidity andmortality by inducing severe infections and bleeding. Thesecomplications often compromise the delivery of intensive che-motherapy and lead to a high frequency of induction death (4).A common assumption is that marrow failure occurs due to

displacement of normal hematopoietic cells from the marrow byAML cells, resulting in depletion of normal hematopoietic cells.However, AML has a more profound impact on BM functionthan many other types of hematologic malignancy (e.g., chronic

lymphocytic leukemia, follicular lymphoma) even where thereis a similar degree of diffuse infiltration of BM by leukemia/lymphoma cells.To clarify how AML suppresses normal hematopoiesis, we

investigated the impact of AML on residual normal hemato-poietic subpopulations, using primary patient samples and a xe-nograft model of AML.

ResultsEngraftment of Human AML Does Not Reduce Numbers of MurineHSCs in a Xenograft Model of AML. Immunodeficient mice trans-planted with human AML develop marrow failure as evidencedby peripheral cytopenias (5, 6) and pallor of the BM in associ-ation with reduced BM erythrocyte numbers (Fig. S1 A and B).We used nonobese diabetic/severe combined immunodeficiency/interleukin-2 receptor γ-chain null (NSG) mice to investigatethe effect of human AML on normal murine hematopoieticpopulations.Ten human AML samples with a range of genetic abnormal-

ities (Table S1) were transplanted into 111 unirradiated NSGmice to determine the impact of AML on mouse CD45+ cells,hematopoietic progenitors (CD45+Lineage−c-kit+), and HSCs(CD45+ Lineage−c-kit+CD150+CD48−) (7) (Fig. 1 A and B). Amedian of 3 AML-transplanted mice and 3 control mice werekilled at different time points posttransplantation to allowchanges in mouse hematopoietic populations to be studied asAML proliferated.The percentage of AML in the BM increased with time, al-

though the growth rates of AML varied from sample to sample.The growth rate of AML was not related to the cytogenetic riskgroup (Fig. S1C). We observed changes in mouse HSCs andprogenitors in three discrete phases, which we designated early,mid-, and late phases. In the early phase there was no significantdifference in mouse progenitor (P = 0.4) or HSC (P = 0.4)numbers in mice transplanted with AML compared with con-trols. In the midphase HSC numbers were preserved whereas

Author contributions: F.M.-M., F.A.-A., E.G., L.J., J.G.G., D.B., and D.C.T. designed research;F.M.-M., F.A.-A., E.G., G.R., D.L., L.J., D.B., and D.C.T. performed research; J.C., M.S., H.O.,S.A., D.B., and D.C.T. contributed new reagents/analytic tools; F.M.-M., F.A.-A., K.A.H., J.G.G.,D.B., and D.C.T. analyzed data; and F.M.-M., F.A.-A., K.A.H., J.K.D., J.C., M.S., H.O., S.A.,J.G.G., D.B., and D.C.T. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1D.C.T. and D.B. contributed equally to this work.2To whom correspondence may be addressed. E-mail: d.taussig@qmul.ac.uk or dominique.bonnet@cancer.org.uk.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1301891110/-/DCSupplemental.

13576–13581 | PNAS | August 13, 2013 | vol. 110 | no. 33 www.pnas.org/cgi/doi/10.1073/pnas.1301891110

mouse progenitors were significantly reduced (P < 0.0001). Inthe late phase both mouse progenitors (P = 0.008) and HSCs(P = 0.009) were significantly reduced. Two representativeexamples are shown in Fig. 1C and the full data are presented inTable S2. We have expressed the data for all 10 AML samplesas a percentage of the values in control mice (to normalize thedata) and present the collated data in Fig. 1D. Reductions inmouse CD45+ cells paralleled changes in mouse progenitors(Fig. 1D).We observed that the early phase was present where AML

infiltration was less than 15% of BM CD45+ cells (n = 19/111mice transplanted with AML) (Fig. 1 D and E). The midphasewas seen in the majority of mice (n = 69/111 mice) across a widerange of AML infiltration (22–84%) (Fig. 1 D and E) and oc-curred following transplant of all 10 AML samples, indicating itis unrelated to the cytogenetic subtype of the AML (Table S1).The late phase was seen only at very high levels of AML in-filtration in 5 of the 10 AML samples (n = 23/111 mice) (Fig. 1C–E and Table S2). Late phase may have occurred in allexperiments if the experiments were continued for longer asthere was movement of HSC numbers in a downward directionat the last time point in some experiments in which late phasewas not reached (Fig. S1D). Paraplegia due to AML infiltratingthe spinal cord developed in midphase in some experiments,preventing longer follow-up.Mouse HSCs (and progenitors) were expressed as a percent-

age of total BM CD45+ cells (human CD45+ AML cells plusmouse CD45+ cells) as the best available estimate of total HSC

numbers (Fig. 1 A and B). Total HSC numbers are a product ofthe percentage of HSCs and the overall cellularity of the BM.We examined BM sections from mice in midphase to exclude thepossibility that overall BM cellularity was reduced in micetransplanted with AML. As seen in the majority of human casesof AML, cellularity was at least as high (if not indeed increased)in mice with AML in midphase compared with controls (Fig. 1F).We conclude that the observed maintenance of HSC percentagerepresents a true preservation of absolute HSC numbers. Nota-bly, we detected an actual increase in HSC numbers in midphasedespite a reduction in progenitors in some experiments (Fig. 1Dand Table S2). Therefore, BM failure in the mouse model inmidphase is not due to depletion of HSCs.To control for the effect of transplantation of cells we injected

mice with AML cells from three AML samples that do notproliferate well in NSG mice (grafting 15% or less of BM cells).At 14 wk there was no difference in HSC (P = 0.2) or progenitornumbers (P = 0.7, Fig. S2). These data in combination with thosefrom the early phase indicate that transplantation of cells per sedoes not induce changes in HSC or progenitor numbers.In midphase there were significantly fewer progenitors per

HSC in mice transplanted with AML (54 ± 10 progenitors perHSC) compared with controls (199 ± 23 progenitors per HSC)(P = 0.0002). This is consistent with the hypothesis that AMLinduces BM failure by impeding differentiation at the HSC–progenitor transition, leading to a failure of progenitor pro-duction. By contrast, there were significantly more mouse CD45+

cells per progenitor in mice transplanted with AML (43 ± 6 mouse

Mouse Progenitor gate

Mouse HSC gate

B

D

C

FE

Mea

n A

ML

% in

bon

e m

arro

w a

t eac

h tim

e po

int

0

20

40

60

80

100

Early Mid

Late

Phase: Early Mid Late mHSCs: N N mProgenitors: N

AM

L %

in

bone

mar

row

Mou

se H

SC

% in

bo

ne m

arro

w

††

†

Days post transplant

0 50 1000

20406080

100

0.000.010.020.030.040.05

0

2

4

6

Mou

se p

roge

nito

r %

in b

one

mar

row

0 50 1000

20406080

100

0.00

0.01

0.02

0.03

Phase: Early Mid mHSCs: N NmProgenitors: N

012345

† †

Days post transplant

ControlAML

ControlAML

ControlAML

ControlAML

AM

L %

in

bone

mar

row

Mou

se H

SC

% in

bo

ne m

arro

wM

ouse

pro

geni

tor

% in

bon

e m

arro

w

0 50 100 0 50 100

0 50 100 0 50 100

†

50

CA

ContAML

50

0 5

0 5

†

†

100

ControlML

rol

100

C

100

† †

ol

ol100

0 50

ControAML

ControAML

0 50

Control AML 4 AML 6 AML 7

Mou

se h

emat

opoi

etic

po

pula

tions

exp

ress

ed a

s %

of

valu

es in

con

trol m

ice

Phase: Early Mid Late

‡0

50

100

150 ‡

‡‡

mHSC

Control

mProgenitorsmCD45+ cells

‡‡

83%0.4%*

0.02%*

0%3.2%*

g

0.02%*H

uman

CD

45

c-K

itLineage stain

c-K

itCD150

CD

48C

D48

Hum

an C

D45

Mouse CD45

A

Lineage stain CD150Mouse CD45

Fig. 1. Effect of AML on mouse hematopoietic populations in a xenograft model. (A and B) The gating strategy for identifying mouse hematopoieticprogenitors and HSCs in BM is shown for a control mouse (A) and a mouse transplanted with AML (B). (C) The early (blue bars), mid- (pink bars), and late (graybars) phases are seen following transplant of AML sample 1 (Left) whereas only early and midphases are seen following transplant of sample 3 (Right). (D)Summary of data from all experiments showing numbers of mouse CD45+ cells, progenitors, and HSCs in mice with AML as a percentage of values in controlsin the three phases. The gray line represents the controls. (E) The mean AML percentage is shown for each of the time points and its relation to the phase. (F)The cellularity of BM in the ilium is not reduced in midphase following transplant of AML samples (three examples, Right) compared with control (Left) (20×objective magnification). *The HSCs and progenitors are expressed as a percentage of total CD45+ cells (AML plus mouse). †P < 0.05 and ‡P < 0.05, comparingmean percentages in mice with AML to control values at each time-point prenormalization, using a paired T test.

Miraki-Moud et al. PNAS | August 13, 2013 | vol. 110 | no. 33 | 13577

MED

ICALSC

IENCE

S

CD45+ cells per progenitor) compared with controls (31 ± 4mouse CD45+ cells per progenitor) (P = 0.0008), suggestingdownstream differentiation is not adversely affected by AML.Although there was a modest increase in mouse CD45+ cells perprogenitor cell in mice with AML, total mouse CD45+ cellnumbers were dramatically depressed (Fig. 1D) because there isa much greater fall in progenitors per HSC.If BM failure is due to decreased production of downstream

hematopoietic progenitor cells by HSCs, one would expect to seereduced HSC cycling. Therefore, we tested HSC cycling in micetransplanted with AML in midphase and in controls, using thebromodeoxyuridine (BrdU) assay. In two independent experi-ments the percentage of cycling HSCs was reduced (P = 0.007and P = 0.04) in mice transplanted with AML (Fig. 2A), con-sistent with our hypothesis.We felt it important to discount other potential explanations

for the midphase pattern (reduced progenitors despite preservedHSC numbers). To discriminate between a block in differentia-tion at the HSC–progenitor transition and an AML-inducedapoptosis in mouse hematopoietic cells downstream of HSCs wequantified apoptosis in mouse hematopoietic cells. The per-centage of apoptotic mouse CD45+ cells (Fig. 2B) and mouseprogenitors (Fig. S3 A–C) was not increased in mice transplantedwith AML in midphase compared with controls. The data sup-port the hypothesis that AML has its effects by preventing pro-duction of downstream hematopoietic cells rather than byinducing apoptosis in these cells.

Residual Mouse CD45+ Cells from Midphase Are Enriched in Long-Term Repopulating Cells. The immunophenotyping data indicatethat HSCs were preserved in midphase despite reductions indownstream hematopoietic cells. To provide independentquantification of primitive hematopoietic cells, we subjectedmouse CD45+ cells from the BM of mice transplanted with AMLin midphase to colony-forming and repopulation assays. Therewas no significant difference in the number of colony-formingcells per CD45+ cell on the initial plating in methylcellulose (P >0.4 for all, Fig. 2C) or type of colonies (P = 0.5, Fig. S3D).However, when the cells were replated, more colonies were seen

from mouse CD45+ cells derived from mice transplanted withAML (P < 0.0007 for all, Fig. 2C). This suggests the mouseCD45+ cells from mice transplanted with AML are relativelyenriched in more primitive hematopoietic cells.The ability to repopulate BM with hematopoietic cells fol-

lowing transplantation is considered a fundamental property ofHSCs. To quantify HSCs in mice transplanted with AML inmidphase we tested the ability of residual mouse CD45+ cells (orcontrols) to repopulate the BM of 88 recipient mice in two in-dependent experiments (Fig. S4). Between four and eight timesmore repopulating cells were present per mouse CD45+ cellfrom the mice transplanted with AML sample 10 compared withcontrol mice at each time point (P < 0.02 for each time point,Fig. 2D and Table S3). Between three and nine times morerepopulating cells were present per mouse cell from micetransplanted with AML sample 3 at each time point (P < 0.04 foreach time point, Fig. S5). The functional analyses confirm thatHSCs make up a greater proportion of mouse hematopoieticcells in mice transplanted with AML than in controls and supportthe findings from our immunophenotyping experiments. Theseexperiments also demonstrate that the mouse CD45+ cells pro-liferate and differentiate at least as well as steady-state mouseCD45+ cells when removed from the leukemic environment.Thus, the differentiation block induced in normal HSCs by AMLis reversible.

Primitive Normal Hematopoietic Cells Are Preserved in the BM ofPatients with AML at Diagnosis. To assess the clinical relevanceof our findings in the xenograft model, we next determinedwhether HSC numbers are preserved (as seen in midphase) ordepleted (as seen in late phase) in humans at diagnosis of AML.We quantified the numbers of hematopoietic stem-progenitorcells (HSPCs) and progenitors in the BM of patients with AMLat diagnosis by studying a specific subtype of AML with very lowCD34 expression. Up to 40% of nucleophosmin (NPM) mutantAMLs and a small minority of NPM wild-type AMLs havea specific phenotype with very low expression of CD34 that wehave previously termed subtype A (Fig. 3A) (8). Critically,a distinct CD34+ population persists in these subtype A samplesthat lacks leukemic genetic lesions, gives rise to normal coloniesin methylcellulose, and contains normal SCID-repopulatingcells, indicating that the CD34+ fraction contains normal HSPCs(8, 9). Therefore, these subtype A AML samples provide aunique opportunity to quantify the residual normal progenitorand HSC populations.Phenotypically defined HSPCs (CD34+CD38− cells) and pro-

genitors (CD34+CD38+) were quantified in fresh BM aspiratesof 16 patients with subtype A phenotype AML at diagnosis and42 age-matched controls (Fig. 3 A and B). We observed thatnormal progenitor numbers were significantly reduced in AMLBM (P < 0.0001). Progenitor subtypes were also all reducedsignificantly (10) (P < 0.02, Fig. S6A). By contrast, normalHSPCs were not significantly different from controls (P = 0.6,Fig. 3B). This pattern is the same as was seen in midphase in thexenograft model (Fig. 1D). The number of progenitors per HSPCwas 10-fold lower in AML (3.5 ± 1 progenitors per HSPC) thanin controls (41 ± 5 progenitors per HSPC) (P < 0.0001), con-sistent with a differentiation block at the HSC–progenitortransition.A recent publication has established improved markers for the

identification of human HSCs (11). Using a similar approach weshowed that the proportion of CD34+CD38− cells from subtypeA AML that has the HSC phenotype (CD49f+CD45RA−Rho-damine123

lo) is greater than in control BM (P = 0.003, Fig. 3C).

Normal HSPCs from BM of Patients with AML Show Reduced Cycling.As in the xenograft model, we found no evidence to support thenotion that the depletion of downstream hematopoietic cells in

A

DC

Weeks post transplant0 10 20 30

10-610-510-410-3

AMLControl

Freq

uenc

y of

repo

pula

ting

cells

per

mou

se C

D45

+ ce

ll (lo

g sc

ale,

i.e.

10-

4 is

1

in 1

0,00

0 ce

lls)

** **

ReplatingPrimary culture

ControlAML

020406080

100

Col

onie

s pe

r 20,

000

mou

se C

D45

+ ce

lls

AML: 3 4 7 3 4 7

* **

ControlAML

1 3 4 5 6 7 8 10 AML :02468

B

% o

f apo

ptot

ic

mou

se C

D45

+ ce

lls

** *

020406080

100

% o

f Brd

U+

mou

se H

SC

s

ControlAML

AML : 2 5

**

Fig. 2. Mouse CD45+ cells from mice with AML in midphase are enriched inHSCs. (A) HSCs from mice transplanted with AML show reduced cycling asassessed by BrdU incorporation. (B) The percentage of apoptotic mouseCD45+ cells was not increased in mice transplanted with AML in midphase.(C) Similar numbers of colonies were obtained from mouse CD45+ cells de-rived from mice with AML and controls on initial culture (P > 0.4) but onreplating more colonies were derived from mouse CD45+ cells from micetransplanted with AML (P < 0.0007). (D) The frequency of repopulating cellswithin mouse CD45+ cells from mice transplanted with AML sample 10 inmidphase was significantly higher than in controls at each time point. Errorbars indicate 95% confidence intervals. *P < 0.05.

13578 | www.pnas.org/cgi/doi/10.1073/pnas.1301891110 Miraki-Moud et al.

AML is due to increased apoptosis in the downstream hemato-poietic progenitors (P = 0.5, Fig. S6B).To look for evidence that the reduced numbers of progenitors

are due to reduced production by HSPCs we determined the cellcycle profile of phenotypically defined normal HSPCs in the BMof patients with subtype A phenotype AML, using Ki67 staining.A lower percentage of HSPCs from AML BM were in cell cyclecompared with controls (P = 0.002), indicating that HSPCs aremore quiescent in AML (Fig. 3D). The results support theconclusion that HSPCs are failing to produce sufficient progen-itors in the context of AML.

Normal CD34+ Hematopoietic Cells from Patients with AML AreEnriched in Long-Term Culture, Initiating Cells and RepopulatingCells. To provide independent quantification of residual normalprimitive hematopoietic cells, we tested CD34+ cells from theBM of subtype A phenotype AMLs in colony-forming unit, long-term culture (LTC), and repopulating assays. We confirmed thenormal nature of the progeny of these CD34+ cells throughgenotyping consistent with earlier publications (Table 1) (8, 9).Similar numbers of normal myelo-erythroid colonies were pro-duced by sorted CD34+ cells in the initial plating from BM fromAML or controls (P > 0.1 for all, Fig. 3E and Table 1). Onreplating, significantly more normal colonies were seen fromnormal CD34+ cells from AML marrow (P < 0.03 for all, Fig.3E), suggesting that they are enriched in primitive hematopoieticcells. The results are comparable to those from analogousexperiments from midphase in the xenograft model (Fig. 2C).Similar results were also seen following coculture of AML andnormal hematopoietic cells in vitro (Fig. S6 C and D). Consistentwith these data, we detected more LTC initiating cells (LTC-ICs)per CD34+ cell (P < 0.0001) and per CD34+CD38− cell (P <0.002) from subtype A AML than from control BM (Fig. 3 F andG and Table 1), indicating that the normal residual hemato-poietic stem-progenitor compartment is skewed toward the mostprimitive hematopoietic cells in patients with AML.To test the repopulating potential of residual normal hema-

topoietic cells in AML, CD34+ cells from BM from two subtypeA AML samples were transplanted into NSG mice. There weremore human CD45+ cells in the mice transplanted with CD34+

cells from one AML sample than from its control (P < 0.001) andalthough there was no significant difference between the secondAML sample and its control (P = 0.3) (Fig. 3H), more humanCD45+ cells were detected in secondary recipients from bothsubtype A AML samples (P < 0.02 for both) (Fig. 3I). Thesegrafts from the AML samples lacked the leukemia-specific ge-netic lesions (Table 1). These experiments indicate that thenormal CD34+ compartment in subtype A AML is enriched inlong-term repopulating cells compared with control BM.The primary and secondary grafts from AML BM constituted

a predominant B lymphoid population (B lymphocytes were88% ± 5% and 91% ± 2% of human cells from control andAML, respectively, P = 0.5) and a smaller population of CD33+

cells (Fig. 3J), similar to that seen after transplant of normal BM,providing evidence of normal differentiation of CD34+ cells onceremoved from the leukemic marrow.

Normal CD34+ Cells Are Not Displaced from the ParatrabecularRegion by AML. The above experiments demonstrated that he-matopoietic stem-progenitor cells are not reduced in number at

Normal HSPCs

CD34

CD

38

CD34

CD

38

050

100150200

0246810

Progenitors HSPCs

Control AML

ControlAML

0103050

% o

f CD

34+

C38

- cel

ls

havi

ng R

ho12

3lo ,

CD

49f+

CD

45R

A - p

heno

type

0.00.10.20.30.40.5

0

2

4

6

% o

f hum

an C

D45

+ ce

lls in

BM

of

seco

ndar

y re

cipi

ents

Normal progenitors

Normal progenitors

Experiment: i ii

Source of normal CD34+ cells:

ControlAML

Human CD45 Human CD33

Hum

an C

D19

Sid

e sc

atte

r r

6.2%

19

84.3%

6.8%

Sid

e sc

atte

1.35%

Hum

an C

D

2.2%

89.8%

*

Primary Replatingculture

Col

onie

s pe

r 100

0 C

D34

+ ce

lls

Col

onie

s pe

r 10,

000

cells

repl

atedControl

AML

50

100

010203040

0Experiment:i ii iii i ii iii

* **

*

ControlAML

Num

ber o

f LTC

-ICs

per C

D34

+ C

D38

- cel

l

0.0

0.2

0.4

0.6

Experiment: i ii iii

** *

0.00

0.05

0.10

0.15

ControlAML

Num

ber o

f LTC

-ICs

per C

D34

+ ce

ll

Experiment: i ii

**

Experiment: i ii % o

f hum

an C

D45

+ ce

lls in

BM

of

prim

ary

reci

pien

ts

0

20

40

60

Source of normal CD34+ cells:

ControlAML

*

* *

A B

C

Control AML

% o

f cyc

ling

(Ki6

7+)

norm

al H

SP

Cs

02468 *D

E F

G H

I J

Normal HSPCs

Fig. 3. Assessment of normal residual hematopoietic populations inhumans with AML. (A) The gating strategy displayed was used to definenormal HSPCs and progenitors in the BM of a control (Upper) and a patientwith subtype A phenotype AML (Lower). (B) Numbers of phenotypicallydefined HSPCs are preserved (P = 0.6) in the BM of patients with subtype AAML (n = 16) whereas progenitors are reduced compared with controls(n = 42) (P < 0.0001). (C) A significantly greater proportion of BM CD34+

CD38− cells from patients with subtype A AML (n = 7) were CD49f+ CD45RA−

Rhodamine123low compared with controls (n = 7) (P = 0.003). (D) The per-centage of HSPCs in cell cycle was lower in subtype A AML (n = 7) BM thancontrols (n = 9) (P = 0.002). (E) Similar numbers of colonies were observedafter culture of CD34+ cells in methylcellulose from three subtype A AMLBMs and three controls. On replating, increased numbers of colonies wereseen from CD34+ cells derived from AML marrow. (F) More LTC-ICs werepresent in the CD34+ cells from two subtype A AML BMs than in controls.Error bars indicate 95% confidence intervals. (G) More LTC-ICs were presentin the CD34+CD38− cells from three subtype A AML BMs than in controls.Error bars indicate 95% confidence intervals. (H) More human CD45+ cellswere detected in mice transplanted with CD34+ from one subtype A AMLsample (n = 3) than in controls (n = 4) whereas similar numbers were seenfollowing transplant of another sample (n = 4 for each arm). (I) More CD45+

cells were seen in the BM of secondary recipients (n = 3 per arm)

transplanted with 4 million human CD45+ cells derived from subtype A AMLthan in controls in both experiments. (J) The plots shows human grafts withpredominant B lineage populations and smaller myeloid populations insecondary recipients injected with human CD45+ cells derived from CD34+

BM cells from a subtype A AML sample (Upper) and from control CD34+ BMcells (Lower). *P < 0.05.

Miraki-Moud et al. PNAS | August 13, 2013 | vol. 110 | no. 33 | 13579

MED

ICALSC

IENCE

S

diagnosis of AML. However, AML might dislocate them fromtheir usual position within the BM, impairing their function.Human stem-progenitor cells are found at increased numbersnear the trabecular surface of the bone in BM (12). We thereforedetermined whether subtype A AML displaces normal stem-progenitor cells from their usual paratrabecular position, usingimmunostaining of BM biopsies from patients with AML orcontrols. The distribution pattern of dual-positive (CD45+

CD34+) normal stem-progenitor cells was not different in sub-type A AML compared with controls (P > 0.1, Fig. 4), suggestingthat there is no gross displacement.Together the immunophenotyping and functional studies on

patient BM support the conclusion that the numbers of HSPCsare preserved in the BM at diagnosis (and progenitors are depleted)and that these cells differentiate effectively once removed fromtheir leukemic environment. The results closely parallel the datafrom the midphase of the xenograft model. The cumulative datalead us to propose a model in which BM failure occurs due toa failure of production of progenitors by HSCs through a differ-entiation block at the HSC–progenitor transition.

DiscussionBM failure is a prominent clinical feature of AML but little isknown about how AML affects the residual normal hematopoi-etic subpopulations. One potential explanation is that BM failureis due to HSC depletion caused by AML displacing normal he-matopoietic cells from the BM. In our xenograft experimentsmouse HSC numbers were not reduced whereas hematopoieticprogenitors and other downstream hematopoietic cells weresignificantly depressed, indicating that BM failure in this modelis not simply due to HSC depletion. We were careful to includefunctional assays, including the gold standard long-term repo-pulating assay, to produce robust data on HSC quantification.These experiments demonstrate that BM failure is not simplydue to HSC depletion.A potential problem with the xenograft model is that the in-

teraction of human AML with mouse hematopoietic cells in animmunodeficient setting might not reflect the interaction ofAML cells with human hematopoietic cells in patients. However,our studies that quantified normal HSCs and progenitors in bonemarrow from patients at diagnosis of AML suggest that thismodel recapitulates what occurs in humans. Future plannedexperiments include investigation in syngeneic mouse models.We studied AML samples with a range of cytogenetic abnor-

malities (including those from good, intermediate, and poor riskcategories) and phenotypes (French, American, and Britishclassification types, CD34 positive and CD34 negative) and allhad a similar effect on residual hematopoiesis at a given levelof AML infiltration (Fig. 1E). Thus, there appears to be a

common mechanism unrelated to the genetic lesions that drivethe disease.The mechanism of BM failure involves a block in differenti-

ation at the HSC–progenitor transition as the number of pro-genitors per HSC was significantly reduced in humans and in the

Table 1. Leukemia-specific genetic lesions are not present in CD34+ cells or their progeny from subtype A AML samples

Figure in which resultsare displayed and experiment

Genetic lesion in primaryAML sample Fraction

Postsort frequencyof leukemic mutation

Post-CFC/LTC/xenograft frequencyof leukemic mutation Method

Fig. 3 E, i None identified CD34+ NA NA NAFig. 3 E, ii NPM mutant CD34+ ND <0.1% NPM PCRFig. 3 E, iii NPM mutant CD34+ 0% 0% NPM PCRFig. 3 F, i NPM mutant CD34+ ND 0% NPM PCRFig. 3 F, ii NPM mutant CD34+ ND 0% NPM PCRFig. 3 G, i t(11;17) CD34+CD38− 0/100 cells 0/100 cells MLL FISHFig. 3 G, ii NPM mutant CD34+CD38− 0% 0% NPM PCRFig. 3 G, iii NPM mutant CD34+CD38− <0.1% <0.1% NPM PCRFig. 3 H, i t(11;17) CD34+ ND 0/100 cells MLL FISHFig. 3 H, ii NPM mutant CD34+ ND 0% NPM PCR

CFC, colony forming cells; FISH, fluorescence in situ hybridization; LTC, Long-term culture; MLL, mixed-lineage leukemia; NA, not applicable; ND, notdetermined; NPM, nucleophosmin.

0

10

20

30

40

50

0 10 20 30 40 50 60 70 80 90 100+

Per

cent

age

of

CD

45+

CD

34+

cells

Distance of CD45+ CD34+ cellsfrom trabecular bone (microns)

ControlAML

C

10 microns

A B

DAPI

CD34

CD45

Merge

10 microns

I=56.33µm, 81.8º

I=54.33µm, 81.4º I=9.83µm, 80.5º

I=15.98µm, 69.3ºI=21.09µm, 73.3º

I=65.60µm, 78.3º

Fig. 4. Normal CD34+ cells are not displaced from the paratrabecular region byAML. (A) Normal stem-progenitor cells, expressing dual CD45 and CD34, wereidentified in BM biopsy sections from patients with subtype A AML (n = 4) andcontrols (n = 8). CD45+ and CD34+ cells were visualized using Fluorescein andAlexaFluor 546, respectively. An example from a patientwith AML is shown. TwoCD45+CD34+ cells are indicated by white arrowheads. (B) The distance of normalstem-progenitor cells (CD45+CD34+ cells) from the nearest trabecular bone wasmeasured. Six stem-progenitor cells are seen with distance from the trabecularbone indicated in microns. (C) There was no significant difference in the distri-butionof stem-progenitor cells betweenAMLandcontrols. Errorbars indicateSD.

13580 | www.pnas.org/cgi/doi/10.1073/pnas.1301891110 Miraki-Moud et al.

xenograft model. There was no evidence for a block in differ-entiation downstream of progenitors and therefore the mech-anism appears to be specific for differentiation at the HSC–progenitor transition.Hematopoietic stresses normally induce HSCs to enter the cell

cycle to replenish mature blood cells (1–3). Therefore, the he-matopoietic stress of BM failure induced by AML might beexpected to induce HSC cycling to maintain blood cell numbers.The opposite (i.e., reduced HSC cycling) occurred in the xeno-graft model and in patient BM. This observation supports thenotion that AML is impairing the function of HSCs rather thanjust depleting HSC numbers. HSCs might show reduced cyclingif they are displaced from their niche but we could find no evi-dence of this, using our methodology.The reduced cycling of HSCs may seem at odds with data from

some xenograft experiments in which HSCs were actually in-creased in midphase (Fig. 1D and Table S2). However, this mightreflect a relative increase in the number of daughter cells retainingan HSC phenotype after HSC division as a consequence of thedifferentiation block at the HSC–progenitor transition. The rela-tively small reduction in HSC cycling implies that BM failure isprincipally due to failure of differentiation with a lesser contri-bution related to reduced HSC cycling.Our data are consistent with clinical observations. Recovery of

normal hematopoiesis is usually rapid after successful blastclearance by induction chemotherapy (∼20 d) (13). Time to re-generation is similar to that seen after conventional allogeneicBM transplantation (14), consistent with the notion that HSCsare preserved. If HSCs were severely depleted at diagnosis, onemight expect greater delays in regeneration of hematopoiesis.Occasionally patients do, however, experience prolonged aplasiadespite leukemia cell clearance by chemotherapy. These patientsmay have HSC depletion as observed in late phase in the xe-nograft experiments and may reflect delayed presentation tomedical services.In some patients with AML HSCs exist, termed preleukemic

HSCs, that are functionally normal but contain some but not allof the mutations seen in the AML clone (15). Although we couldnot detect key driver mutations in the CD34+ cells from subtypeA AML, we cannot exclude the possibility that some of thesecells contain other mutations and might be preleukemic HSCs.Preleukemic HSCs may have a growth advantage over unmu-tated HSCs and this might have explained the superior func-tional ability of CD34+ cells from patients with subtype A AML(Fig. 3). Against this, similar data were seen in the xenograftmodel presented here and colony formation was similar from

preleukemic HSCs and control HSCs in the work from StanfordUniversity (15).Laboratory studies from the 1970s suggested that AML

inhibited normal hematopoiesis (16, 17) but the tools to identifyhematopoietic subpopulations had not yet been developed. Ourwork has elucidated the defects in the normal hematopoieticpopulations induced by AML but further work is needed toclarify the mechanism by which HSC differentiation is blocked. Ifthe differentiation block can be reversed, the residual HSCs maybe used to ameliorate BM failure. This may be of benefit for olderpatients who tolerate intensive chemotherapy poorly as well as foryounger patients with refractory disease.In conclusion we demonstrate that BM failure in AML is not

due to depletion of HSC numbers but rather that HSCs fromleukemic BM fail to produce sufficient progenitors as a result ofa differentiation block at the HSC–progenitor transition. Thesefindings may be exploited to provide treatments for one of thekey complications of AML.

Materials and MethodsQuantification of Hematopoietic Cells in Patient BM. A standard operatingprocedure was used to ensure that the BM samples were obtained from thefirst pull of the aspirate and in a small volume (less than 0.5 mL) to reducehemodilution for quantification of normal progenitors and HSPCs. Stem-progenitor cells were quantified in fresh BM, using surface marker stainingcombined with Countbright Absolute Counting Beads (Invitrogen), accordingto manufacturer’s instructions.

Quantification of Mouse Repopulating Cells. B2Mmice were transplanted withAML cells. After 2 mo mice were killed and the residual mouse CD45+ cellswere purified and transplanted into irradiated (375 cGy) NSG mice. Differentdoses of CD45+ cells were transplanted to allow calculation of repopulatingcell frequency by limiting dilution analysis (LDA) (Fig. S4). Mice were bled at4, 8, 16, and 24 wk and the number of mice with a B2M graft was de-termined. Antibody against mouse histocompatibility complex (MHC) class Iantigens was used to distinguish donor B2M cells (lacking MHC class I) fromrecipient NSG cells (expressing MHC class I) in the blood.

Statistical Analysis. Data are presented as means ± SE of mean (SEM) unlessotherwise stated. Error bars indicate SEM unless otherwise stated.

Further details of materials and methods are given in SI Materialsand Methods.

ACKNOWLEDGMENTS. We thank Hal Broxmeyer and Jude Fitzgibbon forcritical comments and Sameena Iqbal, the Haemato-Oncology Tissue Bankteam, and patients for providing samples. This work was supported bya Medical Research Council Clinician Scientist Fellowship (to D.C.T.) and byCancer Research UK (D.B.).

1. Wilson A, et al. (2008) Hematopoietic stem cells reversibly switch from dormancy toself-renewal during homeostasis and repair. Cell 135(6):1118–1129.

2. Trumpp A, Essers M, Wilson A (2010) Awakening dormant haematopoietic stem cells.Nat Rev Immunol 10(3):201–209.

3. Cheshier SH, Prohaska SS, Weissman IL (2007) The effect of bleeding on hematopoi-etic stem cell cycling and self-renewal. Stem Cells Dev 16(5):707–717.

4. Kantarjian H, et al. (2006) Results of intensive chemotherapy in 998 patients age 65years or older with acute myeloid leukemia or high-risk myelodysplastic syndrome:Predictive prognostic models for outcome. Cancer 106(5):1090–1098.

5. Ishikawa F, et al. (2007) Chemotherapy-resistant human AML stem cells home to andengraft within the bone-marrow endosteal region. Nat Biotechnol 25(11):1315–1321.

6. Malaisé M, et al. (2011) Stable and reproducible engraftment of primary adult andpediatric acute myeloid leukemia in NSG mice. Leukemia 25(10):1635–1639.

7. Kiel MJ, et al. (2005) SLAM family receptors distinguish hematopoietic stem andprogenitor cells and reveal endothelial niches for stem cells. Cell 121(7):1109–1121.

8. Taussig DC, et al. (2010) Leukemia-initiating cells from some acute myeloid leukemiapatients with mutated nucleophosmin reside in the CD34(-) fraction. Blood 115(10):1976–1984.

9. van der Pol MA, et al. (2003) Assessment of the normal or leukemic nature of CD34+cells in acute myeloid leukemia with low percentages of CD34 cells. Haematologica88(9):983–993.

10. Manz MG, Miyamoto T, Akashi K, Weissman IL (2002) Prospective isolation of humanclonogenic common myeloid progenitors. Proc Natl Acad Sci USA 99(18):11872–11877.

11. Notta F, et al. (2011) Isolation of single human hematopoietic stem cells capable oflong-term multilineage engraftment. Science 333(6039):218–221.

12. Bourke VA, et al. (2009) Spatial gradients of blood vessels and hematopoietic stemand progenitor cells within the marrow cavities of the human skeleton. Blood 114(19):4077–4080.

13. Goldstone AH, et al.; Medical Research Council Adult Leukemia Working Party (2001)Attempts to improve treatment outcomes in acute myeloid leukemia (AML) in olderpatients: The results of the United Kingdom Medical Research Council AML11 trial.Blood 98(5):1302–1311.

14. Gisselbrecht C, et al. (1994) Placebo-controlled phase III trial of lenograstim in bone-marrow transplantation. Lancet 343(8899):696–700.

15. Jan M, et al. (2012) Clonal evolution of preleukemic hematopoietic stem cells pre-cedes human acute myeloid leukemia. Sci Transl Med 4(149):149ra118.

16. Broxmeyer HE, Jacobsen N, Kurland J, Mendelsohn N, Moore AS (1978) In vitro sup-pression of normal granulocytic stem cells by inhibitory activity derived from humanleukemia cells. J Natl Cancer Inst 60(3):497–511.

17. Miller AM, Marmor JB, Page PL, Russell JL, Robinson SH (1976) Unregulated growth ofmurine leukemic cells and suppression of normal granulocyte growth in diffusionchamber cultures. Blood 47(5):737–745.

Miraki-Moud et al. PNAS | August 13, 2013 | vol. 110 | no. 33 | 13581

MED

ICALSC

IENCE

S