1

Chronic myelogenous leukemia cells remodel the bone marrow niche 1

via exosome-mediated transfer of miR-320 2

Xiaotong Gao1#, Zhuo Wan1#, Mengying Wei2,3#, Yan Dong1, Yingxin Zhao1, Xutao Chen4, 3

Zhelong Li5, Weiwei Qin1*, Guodong Yang2,3*, Li Liu1,* 4

1Department of Hematology, Tangdu Hospital, Fourth Military Medical University, Xi’an, 5

710038, People’s Republic of China. 6

2State Key Laboratory of Cancer Biology, Fourth Military Medical University, Xi’an, 710032, 7

People’s Republic of China. 8

3Department of Biochemistry and Molecular Biology, Fourth Military Medical University, 9

Xi’an, 710032, People’s Republic of China. 10

4Department of Implantation, School of Stomatology, Fourth Military Medical University, 11

Xi’an, 710032, People’s Republic of China. 12

5Department of Ultrasound Diagnostics, Tangdu Hospital, Fourth Military Medical 13

University, Xi’an, 710038, People’s Republic of China. 14

# These authors contributed equally to this work. 15

*Correspondence and requests for materials should be addressed to Li Liu (email: 16

heamatol@fmmu.edu.cn). Correspondence may also be addressed to Guodong Yang (email: 17

yanggd@fmmu.edu.cn) and Weiwei Qin (email: vivianq1126@126.com) 18

All the authors declare no potential conflicts of interest. 19

2

Abstract 20

Rationale: Reciprocal interactions between leukemic cells and bone marrow mesenchymal 21

stromal cells (BMMSC) remodel the normal niche into a malignant niche, leading to leukemia 22

progression. Exosomes have emerged as an essential mediator of cell-cell communication. 23

Whether leukemic exosomes involved in bone marrow niche remodeling remains unknown. 24

Methods: We investigated the role of leukemic exosomes in molecular and functional changes 25

of BMMSC in vitro and in vivo. RNA sequencing and bioinformatics were employed to screen 26

for miRNAs that are selectively sorted into leukemic exosomes and the corresponding RNA 27

binding proteins. 28

Results: We demonstrated that leukemia cells significantly inhibited osteogenesis by BMMSC 29

both in vivo and in vitro. Some tumor suppressive miRNAs, especially miR-320, were enriched 30

in exosomes and thus secreted by leukemic cells, resulting in increased proliferation of the 31

donor cells. In turn, the secreted exosomes were significantly endocytosed by adjacent 32

BMMSC and thus inhibited osteogenesis at least partially via β-catenin inhibition. 33

Mechanistically, miR-320 and some other miRNAs were sorted out into the exosomes by RNA 34

binding protein heterogeneous nuclear ribonucleoprotein A1 (HNRNPA1), as these miRNAs 35

harbor the recognition site for HNRNPA1. 36

Conclusion: HNRNPA1-mediated exosomal transfer of miR-320 from leukemia cells to 37

BMMSC is an important mediator of leukemia progression and is a potential therapeutic target 38

for CML. 39

Key words: chronic myelogenous leukemia, exosomes, bone marrow mesenchymal stromal 40

cells, miR-320, HNRNPA1 41

3

Graphical Abstract 42

HNRNPA1-mediated exosomal transfer of miR-320 from leukemia cells to BMMSC is an 43

important mediator of leukemia progression and niche remodeling. 44

45

4

Introduction 46

Chronic myelogenous leukemia (CML) results from the transformation of normal 47

hematopoietic stem cells (HSC) by the BCR-ABL oncogene. Disruption of normal 48

hematopoiesis occurs even before outgrowth of leukemic cells, causing anemia, 49

thrombocytopenia, neutropenia and osteoporosis, which are the major causes of morbidity and 50

mortality of CML. The bone marrow hematopoietic microenvironment (“niche”) is a unique 51

composition of mesenchymal and non-mesenchymal cells, which supports the self-renewal of 52

normal and malignant hematopoietic cells [1]. Reciprocal interactions between leukemic cells 53

and the niche are essential for the leukemic progression and resistance to drug treatment [1]. In 54

addition, emerging evidence has shown that myeloid malignancies induce molecular and 55

functional changes in the niche [2-5], switching the normal hematopoietic cell favouring niche 56

to leukemic cell favouring one [6-8]. However, the mechanisms by which leukemic cells 57

remodel the bone marrow niche have not been adequately defined. 58

Exosomes are recognized as pivotal mediators of cell-cell communication, through 59

transmission of intracellular cargo such as miRNAs [9, 10]. Exosomes loaded with miRNAs 60

are secreted by cancer cells and internalized by neighbouring cells in the malignant niche, 61

affecting the phenotype of niche cells [11-15]. Specific miRNAs are selectively sorted into 62

exosomes, through mechanisms that are still poorly understood [16]. It has been reported that 63

RNA-binding proteins such as heterogeneous nuclear ribonucleoproteins A2B1 (HNRNPA2B1) 64

and Q (HNRNPQ) are involved in exosomal RNA export by recognizing and binding specific 65

motifs in miRNA sequences [17, 18]. Notably, HNRNPA1 is induced by BCR/ABL tyrosine 66

kinase in a dose- and kinase-dependent manner [19]. In addition, HNRNPA1 controls the 67

nuclear export of mRNAs that are indispensable for the leukemia phenotype of CML [20]. All 68

of these suggest that leukemic cells might selectively sort certain miRNAs into the exosomes 69

during the progression of CML. 70

Here, by using leukemic cell lines, a murine leukemic model, and patient samples, we 71

found that leukemic cells selectively transport growth-suppressing miRNAs into exosomes via 72

heterogeneous nuclear ribonucleoprotein A1 (HNRNPA1), resulted in promoted CML cell 73

5

growth. In addition, the leukemia cell-derived exosomes could be endocytosed by the 74

surrounding BMMSC, leading to inhibited osteogenesis of the recipient cells and thus 75

remodelled bone marrow niche favourable for CML progression. 76

Material and methods 77

Animals 78

Double-transgenic SCL-tTa X TER-BCR/ABL mice [21] were backcrossed for at least eight 79

generations into the C57/BL6 background and were maintained in the presence of 0.5 g/L 80

doxycycline in the drinking water. BCR/ABL expression was induced in 5-week-old double-81

transgenic mice (BA) by doxycycline withdrawal for at least 4 weeks. All mice were 82

maintained under specific pathogen free conditions and the experimental procedures were 83

performed in accordance with guidelines and protocols approved by the Institutional Animal 84

Experiment Administration Committee. 85

Cell line and human samples 86

Human CML cell lines K562 and LAMA84 were purchased from ATCC and cultured in RPMI 87

1640 medium (Hyclone, Loga, UT, USA) with 10% FBS and penicillin–streptomycin 88

(Hyclone). 89

Human bone marrow and peripheral blood specimens were obtained from 41 patients with 90

newly diagnosed CML; 24 patients were in chronic phase (CML-CP) and 17 patients in blast 91

crisis (CML-BC). Specimens from 11 age-matched healthy donors were used as control. All 92

subjects signed an informed consent form. All specimen acquisition was approved by the Ethics 93

Committee of Fourth Military Medical University. 94

Isolation and expansion of bone marrow mesenchymal stromal cells (BMMSC) 95

BMMSC were isolated from the bone marrow of CML-CP patients and age-matched healthy 96

donors after informed written consent as previously described [22]. Briefly, the mononuclear 97

cell fraction was isolated by density gradient centrifugation using Human Lymphocyte 98

Separation Medium (Dakewe Biotech, Shenzhen, China), re-suspended in alpha-MEM (Gibco, 99

Carlsbad, CA, USA) containing 5% human platelet lysate (HELIOS, Atlanta, GA, USA), L-100

6

glutamine, and penicillin–streptomycin (Hyclone), and plated at an initial density of 1×106 101

cells/cm2. Three days later, non-adherent cells were removed and the remaining monolayers of 102

adherent cells were cultured in fresh medium until they reached confluence. The cells were 103

harvested by trypsinization and sub-cultured at densities of 5000–6000 cells/cm2. BMMSC 104

were identified as CD34-/CD45-/HLA-DR-/CD29+/CD90+/CD105+ using flow cytometry. 105

Multipotent differentiation capacity was demonstrated by osteogenic, adipogenic and 106

chondrogenic inductions as described below. For osteogenesis analysis, BMMSC were used 107

during third to fifth passages for experiments. For miR-320a expression analysis in the 108

BMMSCs, the primary culture was used. 109

Exosome purification and characterization 110

For exosome isolation from K562 and LAMA84 cells, RPMI 1640 medium with 10% 111

exosome-depleted FBS (ultracentrifugation at 120,000g for 16 h) was used. 112

For exosome isolation from mice or patients’ hematopoietic cells, bone marrow 113

mononuclear cells were collected from either mice femurs or patients’ bone marrow samples 114

and then cultured using StemSpan SFEM II medium (Stemcell Technologies, Vancouver, 115

Canada) without FBS. Exosomes were isolated by differential centrifugation. In detail, 116

supernatant fractions collected from 48-72 hr cell cultures were centrifuged at 800g for 5 min 117

and 2000g for 10 min. The supernatant was then filtered using 0.22 μm filters, followed by a 118

final centrifugation at 10,0000g for 2 hr. The supernatant was aspirated and the pellet was re-119

suspended in PBS. The concentration of exosomes was determined by total exosomal proteins 120

using the Pierce BCA Protein Assay Kit (Thermo Scientific, Waltham, MA, USA), according 121

to the manufacturer’s instruction. For exosome isolation from serum, serum was collected from 122

all blood samples using a 2-step centrifugation protocol (2000g for 10 min, 10000g for 10 min) 123

to remove cellular debris. Serum was then incubated with ExoQuickTC exosome precipitation 124

solution (System Biosciences, Johnstown, PA, USA) according to manufacturer’s instructions. 125

Exosomes were stored at 4°C and analyzed or used for experimental procedures within one 126

week after harvest. 127

NanoSight and transmission electron microscopy were used to determine exosome size 128

7

distribution, concentration and morphology. For biomarker analysis of exosomes, the exosome 129

pellet was dissolved in lysis buffer for western blot (see below). 130

Fluorescently labeling of exosomes for in vitro and in vivo tracing 131

For in vitro analysis of the uptake of leukemic exosomes, exosomes in suspension were labeled 132

with 10 μM DiI (Invitrogen, Carlsbad, CA, USA) in a 37 oC water bath for 20 minutes and 133

repelleted by ultracentrifugation at 100,000g for 2 hr. DiI-labeled exosomes were added into 134

culture medium of normal BMMSC, which were plated on round cover slips (Electron 135

Microscopy Sciences, Hatfield, PA, USA) at 2×103/cm2, and incubated for another 24 hr. 136

Images of exosome uptake were acquired with a confocal laser scanning microscope. 137

For in vivo tracking of exosomes, exosomes prepared from cell lines K562, HeLa (human 138

cervical carcinoma), and A549 (human non-small cell lung cancer) were stained with 10 μM 139

DiR (Invitrogen, Carlsbad, CA, USA) similar as the DiI method mentioned above. About 200 140

μg of DiR-labeled K562, HeLa or A549-derived exosomes were intravenously injected in 141

C57BL/6 mice for 24 hours. The localization of the exosomes in organs was detected by 142

imaging using the IVIS® Lumina II in vivo imaging system (PerkinElmer, Waltham, MA, USA). 143

Flow cytometry was used to analyze the exosome distribution in different cell populations 144

in the bone marrow. Briefly, 200 μg of DiO-labeled exosomes were intravenously injected in 145

C57BL/6 mice, and bone marrow cells were harvested 24 hours later. DiO positive cells in 146

bone marrow mesenchymal stromal cells (CD45- Ter119- CD31-), endothelial cells (CD45- 147

Ter119- CD31+), granulocyte (Mac1+ Gr1+), B cells (B220+), and T cells (CD3+) were 148

analyzed by flow cytometry. 149

Colony forming unit-fibroblast assay 150

The colony forming unit-fibroblast (CFU-F) assay was performed as previous described [23]. 151

Briefly, normal or BA mice were sacrificed by CO2 and their femurs were cleaned from muscle 152

and crushed into pieces to release bone marrow cells. The marrow-free bone pieces were further 153

digested with 3% collagenase (Type I, Sigma-Aldrich, Saint Louis, MO USA) at 37 oC for 40 154

min. 2 x 104 cells were plated into a 60 mm dish in MesenCultTM Expansion Medium (Stemcell 155

Technologies). The medium was removed after 14 days culture and cells were fixed with 4% 156

8

PFA for 5 min at room temperature and then stained with crystal violet for 15 min. After 157

washing, colonies with more than 50 cells per colony were counted. 158

Flow cytometric analysis 159

Single cell suspensions from bone marrow, spleen, and peripheral blood were analyzed by flow 160

cytometry using monoclonal antibodies listed in the supplemental materials (Table S2). All 161

samples were first stained for 20 min at 4 oC with CD16/32 Fc blocking antibody (Biolegend, 162

San Diego, CA, USA) unless indicated otherwise. Stained cells were collected using Beckman 163

FC500 or BD FACSCalibur and analyzed using FlowJo software (FlowJo, Ashland, OR, USA). 164

Quantitative real-time PCR 165

Total RNA from cells or exosomes was extracted with Tripure Isolation Reagent (Roche, Basel 166

Switzerland) following the manufacturer’s instructions. RNA purity and concentration were 167

assessed with a Nanodrop-1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, 168

USA). miRNA expression was measured by quantitative real-time PCR (qPCR) using miRcute 169

miRNA First-Strand cDNA Synthesis Kit and miRcute miRNA SYBR Green qPCR Detection 170

Kit (Tiangen, Beijing, China). U6 was used as an internal control. For analysis of mRNA 171

expression, 1 μg total RNA was reverse transcribed using Transcriptor Reverse Transcriptase 172

Kit (Roche) and qPCR was performed using FastStart Essential DNA Green Master Kit 173

(Roche). GAPDH was used for normalization. All kits were used per manufacturers’ 174

instructions. The primer sequences are listed in Table S1. qPCR was run using the Roche 175

LightCycler 96 qPCR system (Roche) with each reaction run in triplicate. The change in 176

miRNA or mRNA expression was calculated by normalizing to the control sample as described 177

previously [24]. 178

Western blot 179

Cells or exosomes were homogenized in lysis buffer (20 mM Tris [pH7.5], 150 mM NaCl, 1% 180

Triton X-100 2 mM sodium pyrophosphate，25 mM β-glycerophosphate) supplemented with 181

Protease Inhibitor Cocktail (Roche). Protein extracts were separated in 10% or 12% SDS-182

PAGE and transferred to nitrocellulose membrane. The nitrocellulose membrane was blocked 183

with 5% nonfat milk for 1 hr and then incubated overnight with primary antibodies at 4 oC. 184

9

The membrane was then incubated 1 hr with peroxidase-conjugated secondary antibodies at 185

room temperature and visualized using the ECL Prime Western Blotting Detection Reagent 186

(GE Healthcare，Buckinghamshire UK). All the details of antibodies are listed in Table S2. 187

Co-culture assay 188

K562 cells pre-transfected with 50 nM C. elegans specific miRNA (cel-miR-39) were placed 189

on the insert of a trans-well plate (0.4 μm polycarbonate filter, Corning, Beijing China), while 190

the normal BMMSC was placed on the lower chamber. The medium was additionally treated 191

with or without GW4869 (10 μM, Sigma-Aldrich) for 24 hr during co-culture with BMMSC. 192

At the end of the experiment, BMMSC were washed with PBS and cel-miR-39 expression in 193

BMMSC were detected by qPCR. 194

Small RNA (sRNA) sequencing and analysis 195

Total RNA was extracted from K562 cells and exosomes. The preparation of sRNA libraries 196

and deep sequencing were performed as follows. Briefly, about 500 ng purified RNA from each 197

group was used to prepare the sRNA libraries with Small RNA Sample Pre Kit (Illumina, San 198

Diego, CA, USA). Then miRNA expression was analyzed by small RNA sequence using HiSeq 199

2500 (Illumina). The differential miRNAs expression was shown by heatmap performed using 200

Cluster3.0 and Java Treeview. Multiple alignment of the mature miRNA sequence of cells and 201

exosomes was performed with ClustalX2 and FigTree software. 202

An unbiased search to identify over-represented motifs on exosome-enriched miRNAs 203

was performed with Improbizer as previous described [25]. Zero-or-one occurrence model was 204

used to find over-represented 4-8 nucleotide-long motifs. The remaining miRNAs annotated in 205

miRBase V22 were used as background. A Markov model of order 0 was assumed for the 206

background sequences. The cladogram was generated with the Weblogo software 207

(http://weblogo.threeplusone.com/). 208

Plasmid construction and lentivirus production 209

For knockdown of HNRNPA1 in K562 cell line, lentivirus expression vector pLKO.1 was used 210

to express shRNAs for human HNRNPA1 (TRCN0000006583, TRCN0000006586, 211

10

TRCN0000368907) or scrambled shRNA as control. For overexpression of HNRNPA1, human 212

HNRNPA1 coding sequence was amplified from cDNA of K562 cell line and inserted into the 213

lentivirus expression vector pCDH-CMV-MCS-EF1a-PURO with BamHI and EcoRI sites. The 214

same vector expressing eGFP was used as control. 215

For overexpression and knockdown of miR-320a in K562 cell line, we used the lentivirus 216

expression vector pLKO.1 to express the mature sequence of miR-320a or a sponge sequence 217

that contain 6 repeats of miR-320a binding sites. Detailed sequences used to generate 218

constructs were listed in Table S3. 219

Lentivirus was produced by transfecting HEK293T cells with lentivirus expression vector 220

along with psPAX2 and pMD2.G packaging plasmids. Supernatants containing lentivirus 221

particles were collected 48 hr after transfection, filtered and stored. For virus infection, virus 222

containing medium were added to cells, and after 6 hr incubation, medium was changed to 223

RPMI 1640 complete medium for another 48 h. The lentivirus infected cells were selected 224

using 3μg/ml puromycin for 2-4 weeks. 225

Cell cycle analysis 226

For evaluation of K562 cells’ cell cycle distribution, cells were fixed with 70% ice-cold ethanol 227

for 1 h, treated with RNase A (100 μg/ml) for 30 min at 37oC, and then stained with 20 μg/ml 228

propidium iodide for 30 min at 37oC. Cells were analyzed with BD FACSCalibur flow 229

cytometer. 230

EdU incorporation assay 231

EdU was added to the culture medium for a final concentration of 20 μM and incubated at 37 232

oC for 30 min. Cells were fixed and stained with EdU In Vitro Imaging Kit (Sangon, Shanghai, 233

China), according to manufacturer’s instruction. The percentage of EdU positive cells was 234

determined using ImageJ software. 235

Cell Counting Kit-8 assay 236

Cell proliferation were also analyzed using the Cell Counting Kit-8 (CCK-8) (Dojindo, Mashiki, 237

Kumamoto Japan). Briefly, K562 cells with different treatments were plated in 96-well plates 238

11

at 400/well. The cells were normally cultured for 1 to 7 days before 10 μl CCK-8 reagent was 239

added and incubated for another 2 hours. Absorbance was measured at 450 nm using a 240

SpectraMax 96-well plate reader (Molecular Devices, Sunnyvale, CA, USA). 241

In vitro osteogenic differentiation assay 242

Osteogenic differentiation was assayed as previously described [26]. Briefly, pre-confluent 243

BMMSC were supplemented with osteogenic medium, α-MEM containing 100 nM 244

dexamethasone (Sigma-Aldrich), 50 µg/mL L-ascorbic acid 2-phosphate, and 10 mM β-245

glycerophosphate (Sigma-Aldrich), plus either 5% human platelet lysate (HELIOS) for human 246

BMMSC or 10% FBS for mouse BMMSC, for 28 days. After differentiation, cells were fixed 247

with 4% PFA and stained with Alizarin red S (Sigma-Aldrich). For mineralization 248

quantification, destaining was conducted by adding a 10% cetylpyridinium chloride (Sigma-249

Aldrich) solution. Absorbance was measured in a 96-well plate reader at 570 nm. For alkaline 250

phosphatase (ALP) staining, cells were induced with osteogenic medium for 14 days and ALP 251

activity was measured using BCIP/NBT Alkaline Phosphatase Color Development Kit (Sangon, 252

Shanghai, China) according to manufacturer’s instructions. 253

Transmission electron microscopy 254

Exosomes were fixed with 2% paraformaldehyde, loaded on 200-mesh Formvar-coated grids, 255

and then counterstained and embedded as previously described [27]. Grids were dried and 256

visualized on a transmission electron microscope at 80 kV. 257

Micro-computed tomography (microCT) analysis of bone density 258

Distal femoral metaphases were scanned by microCT (eXplore Locus SP, GE Healthcare, 259

Milwaukee, WI, USA) under the following specifications: voltage, 80kV; current, 80 mA; 260

exposure time, 3000 ms; total rotation, 360o; rotation angle increment, 0.4o; resolution rate, 21 261

μm. Bone morphometric indices were assessed using 3D image reconstructions, included 262

cortical wall thickness (mm), bone volume relative to tissue volume (BV/TV, %), trabecular 263

thickness (mm), trabecular number (1/mm) and trabecular separation (mm). 264

RNA immunoprecipitation 265

12

RNA immunoprecipitation was performed with cytoplasmic extracts. Briefly, K562 cells or 266

freshly isolated bone marrow cells derived from BCR/ABL+ mice were lysed in ice cold lysis 267

buffer (20 mM Tris [pH7.5], 150 mM NaCl, 1% Triton X-100 2 mM sodium pyrophosphate, 268

25 mM β-glycerophosphate) supplemented with Protease Inhibitor Cocktail (Roche) and 40 269

U/ml Protector RNase Inhibitors (Roche) for 20 min. The lysates were then centrifuged for 15 270

min at 12,000g and supernatant was collected. About 1 mg protein extract was incubated with 271

10 μg mouse anti-HNRNPA1 monoclonal antibody (9H10, Abcam) or 10 μg mouse 272

immunoglobulin G [IgG] (Abcam) for 12 hr at 4 oC at a vertical shaking table. After that, 30 273

μl protein A sepharose (Abcam) was added for another 2 hr, followed by three washes with ice 274

cold lysis buffer. Co-immunoprecipitated RNA was extracted using Tripure Isolation Reagent 275

(Roche). miRNA qPCR analysis was performed with SYBR green mix (Roche) after reverse 276

transcription using miRcute kit (Tiangen). The miRNA fold enrichment in immunoprecipitated 277

samples was expressed as percentage against input and compared to IgG control. 278

Biotinylated miRNA Pull-Down 279

Biotinylated miRNA pull-down was performed with cytoplasmic extracts. Briefly, 20 nmol 280

biotinylated single stand miRNA oligonucleotides (GenePharma, Shanghai, China) were re-281

suspended in 100 μL lysis buffer supplemented with Protease Inhibitor Cocktail (Roche) and 282

40 U/ml Protector RNase Inhibitors (Roche) and incubated with 400 μL protein extract ( about 283

1mg) for 4 hr at 4 oC on a vertical shaking table. Streptavidin sepharose beads (Cell Signaling 284

Technology), pre-washed three times with lysis buffer, were added to the mixture and incubated 285

for an additional 2 hr at 4 oC. The beads were then washed three times with 1 mL lysis buffer 286

each. Beads were mixed with protein loading buffer and heated at 95 oC for 10 min to allow 287

collection of bound proteins for western blot or mass spectrometry (MS) analysis. 288

Statistical analysis 289

Data were analyzed with GraphPad Prism7 software. Unpaired two-tailed t test was used to 290

compare data for two groups. One-way ANOVA test was used to compare data for more than 291

two groups. Data are presented as the mean±SEM. Differences were considered significant 292

when the p value was < 0.05 (* p<0.05, ** p<0.01, *** p<0.001). 293

294

13

Results 295

Chronic myelogenous leukemia inhibits osteogenesis of bone marrow mesenchymal 296

stromal cells (BMMSC) 297

The SCL-tTA x TRE-BCR/ABL double transgenic (BA) mice have been widely used as a 298

mouse model of chronic phase CML [8, 21, 28]. As expected, BCR/ABL expression following 299

doxycycline withdrawal results in accumulation of myeloid progenitors and mature 300

granulocytes in the bone marrow, spleen and peripheral blood (Figure S1A and 1B) and 301

splenomegaly (Figure S1C), within in one month. With the occurrence of the CML, micro-302

computed tomography (micro-CT) revealed abnormal bone formation in BA mice after one 303

month induction of BCR/ABL expression. Cortical wall thickness, bone volume relative to 304

tissue volume (BV/TV) and trabecular thickness were significantly decreased, together with an 305

increase of trabecular separation (Figure 1A). Consistent with the observed osteoporosis, 306

BMMSC from induced BA mice exhibited a significant decline of total CFU-F activity per 307

femur as compared to controls (Figure 1B). When cultured under osteoinductive condition, 308

BMMSC from BA mice showed reduced capacity for mineralized nodule formation (Figure 309

1C). Similarly, BMMSC derived from CML patients also showed reduced capacity for 310

osteogenesis (Figure 1D). These data indicate impairment of osteogenic differentiation and 311

proliferation capacity of BMMSC in both CML mouse model and CML patients. 312

Exosomes secreted by leukemia cells are efficiently taken up by BMMSC 313

Exosomes derived from cancer cells have been reported to mediate the remodeling of cancer 314

microenvironments, suggesting possible involvement of leukemic exosomes in tuning 315

BMMSC. Exosomes isolated from K562 culture medium by sequential ultracentrifugation 316

exhibited typical cup-shaped morphology by transmission electron microscopy (Figure 2A), 317

with a size distribution of 50 to 150 nm in diameter analyzed by nanoparticle tracking analysis 318

(Figure 2B). Leukemic exosomes were positive for specific exosomal marker TSG101 but 319

negative for GM103, a marker for Golgi complex (Figure 2C). Coomassie blue staining of total 320

protein showed different protein electrophoresis patterns between leukemia cells and leukemic 321

exosomes (Figure 2D). As revealed by confocal fluorescence microscopy, the labeled 322

14

exosomes were efficiently taken up by BMMSC (Figure 2E). Consistently, ex vivo imaging 323

also revealed that these CML-derived exosomes preferentially migrated into the bone marrow, 324

compared with exosomes from other sources, such as cell lines (HeLa and A549) of human 325

solid tumors (Figure 2F). Consistently, intravenous injection of exosomes loaded with a C. 326

elegans specific miRNA (cel-miR-39) also revealed abundant expression of cel-miR-39 in 327

femurs (Figure 2G). To further determine whether leukemic exosomes indeed endocytosed into 328

BMMSC, we transfected K562 cells with cel-miR-39, followed by co-culture with BMMSC in 329

a 0.4 µm transwell plate. The appearance of cel-miR-39 in BMMSC demonstrates that the cel-330

miR-39 was delivered from the K562 cells in the upper well to the BMMSC in the lower well 331

(Figure 2H). In addition, the exosome secretion inhibitor GW4869 significantly lowered the 332

delivery of cel-miR-39 to BMMSC (Figure 2H), further indicating that K562-derived 333

exosomes are efficiently endocytosed into BMMSC. Next, we examined whether BMMSC are 334

the main recipient cells of the leukemic exosomes in the bone marrow (BM) cells. DiO labeled 335

K562 exosomes were intravenously injected into normal C57BL/6 mice. Internalization of the 336

labeled exosomes into mouse BM cells was demonstrated after 24 hr by FACS. Bone marrow 337

mesenchymal stromal cells (CD45- Ter119- CD31-) exhibited the highest fluorescence 338

intensity than other BM cells, such as endothelial cells (CD45- Ter119- CD31+), granulocyte 339

(Mac1+ Gr1+), B cells (B220+), and T cells (CD3+) (Figure S2A), confirming that leukemic 340

exosomes were more efficiently taken up by BMMSC. 341

Exosomes from leukemia cells inhibit BMMSC osteogenesis 342

We next explored whether leukemic exosomes can modulate osteogenesis of human BMMSC 343

in vitro. BMMSC treated with leukemic exosomes exhibited decreased capacity of osteogenic 344

differentiation as shown by alkaline phosphatase (ALP) and Alizarin Red staining, compared 345

to healthy human serum-derived exosomes or PBS-treated BMMSC (Figure 3A). In addition, 346

the expression of RUNX2, a major osteogenic marker, was downregulated in a dose-dependent 347

manner by treatment with leukemic exosomes (Figure 3B). Furthermore, genes associated with 348

osteogenesis (OPN and COL1A1) or hematopoiesis (CXCL12 and KITL) were also 349

downregulated in BMMSC treated with leukemic exosomes in a dose-dependent manner 350

(Figure 3C and Figure S3A). 351

15

Next, we assumed that leukemic exosomes might also inhibit BMMSC functions in vivo. 352

C57BL/6 mice (6 to 8 weeks old) were intravenously injected with 50 µg of either leukemic 353

exosomes or control exosomes, derived from BA mice or control mice, twice per week for four 354

consecutive weeks. Mice that received leukemic exosomes exhibited significant loss of 355

trabecular bone and whole bone volume in the femurs (Figure 3D and Figure S3B), similar to 356

the phenotype of transgenic BA mice. Moreover, there was a significant decrease of CFU-F 357

frequencies of BMMSC from leukemic exosome-treated mice, compared to normal controls 358

(Figure 3E). Taken together, these results demonstrate that exosomes secreted by leukemia cells 359

can significantly inhibit the function of BMMSC in vitro and in vivo. 360

Specific repertoires of miRNAs are sorted into leukemia exosomes 361

MiRNAs have been reported to be key exosome cargo that can be actively secreted into 362

exosomes and delivered to target cells, resulting in direct modulation of their mRNA targets. 363

In order to clarify whether miRNAs are the key functional mediators in the observed function 364

of leukemic exosomes, we thus performed small RNA sequencing and analyzed the 365

intracellular and exosomal profile of the miRNome in K562 cells (Figure 4A and Figure S4A). 366

In particular, 148 miRNAs were enriched and 165 miRNAs were under represented in leukemic 367

exosomes compared to the cellular fraction. These miRNAs were further analyzed based on 368

fold changes and overall TPM (transcripts per million reads) (at least two data >1000). Forty 369

seven of the miRNAs met these criteria and were selected for further analysis (Figure S4B, 370

Table S5). A heat map of the selected miRNAs demonstrated the profound difference between 371

leukemic cells and exosomes (Figure 4B). The expression profile wase further confirmed by 372

qPCR analysis of selected miRNAs in both K562 cells (Figure 4C) and LAMA84 cells (Figure 373

S4C). These results showed that distinct subsets of miRNAs were either enriched or diminished 374

in leukemic exosomes. 375

ClustalW was used to find the sequence characteristics of the enriched/deminished 376

miRNAs. As shown in Figure 4D, a group of miRNAs that were enriched in exosomes harbored 377

a similar mature sequence, including miR-320a, miR-320b, miR-320c, miR-320d, miR-3180-378

3p, miR-128-3p and miR-423-5p. An unbiased search for over-represented sequence motifs 379

using Improbizer [25] detected a conserved minimal AGAGGG motif (Figure 4E). About 50% 380

16

of exosome-enriched miRNAs contain at least 4 nucleotide of the motif but only 9.7% of cell-381

enriched miRNAs (Table S5). These data indicated that specific repertoires of miRNAs with 382

certain motifs were sorted into leukemia exosomes 383

Exosomal miR-320 plays an essential role in the observed effects on both leukemia cells 384

and BMMSC 385

Among these miRNAs that were enriched in leukemic exosomes, we focused on the miR-320 386

family for three reasons: (1) family members such as miR-320a, miR-320b, miR-320c and 387

miR-320d were specifically loaded into leukemic exosomes (Figure 4B, 4D, S4B); (2) miR-388

320 family contains the AGAGGG motif which may control their exosomal sorting; and (3) 389

previous research has shown that miR-320 can directly target the BCR/ABL oncogene [29], 390

and β-catenin [30, 31] which is a key regulator of osteogenesis. We then investigated the 391

functional effect of miR-320 on both leukemic cells and BMMSC. The 3’-UTR of BCR/ABL 392

contains a binding site that can be recognized by all four miR-320 as predicted from a miRNA 393

target database (Figure 5A). As expected, miR-320a mimics transfection significantly 394

decreased the endogenous expression of BCR/ABL in K562 cells (Figure 5B). Forced 395

expression of miR-320a has significantly inhibited K562 cell proliferation (Figure 5C). Cell 396

cycle assessment suggested that miR-320a can cause G1 phase arrest in the cell cycle (Figure 397

5D), which was further confirmed by EdU incorporation assay (Figure 5E). 398

Notably, the 3’-UTR of β-catenin also contains a binding site that can be recognized by 399

all four miR-320 (Figure 6A). β-catenin is the transcriptional regulator of canonical Wnt 400

signaling, which can promote the maturation of osteoblastic precursor cells into mature 401

osteoblasts. Knocking down of β-catenin in BMMSC by siRNAs inhibited the osteogensis as 402

measured by ALP staining (Figure S5A), and reduced the expression of osteogenesis (OPN, 403

COLIA1) or hematopoiesis (CXCL12, KitL) related genes (Figure S5B). K562 derived 404

exosomes significantly decreased the expression of β-catenin in BMMSC, which was rescued 405

by prior miR-320 knockdown in K562 cells (Figure 6B and 6C). Accordingly, the inhibited 406

osteogenic differentiation capacity of BMMSC by exosomes derived from K562 cells was 407

compromised when miR-320a was knocked down in the parental cells (Figure 6D). To further 408

confirm the role of aberrant differentiation of BMMSC induced by miR-320, we have knocked 409

17

down miR-320 (miR-320 KD) in K562 cells, and then injected 50 µg of the derived exosomes 410

into the C57BL/6 mice, twice per week for four consecutive weeks. Compared with the control 411

groups, exosomes from miR-320 KD K562 cells had minimal effects on the bone abnormalities 412

(Figure 6E). These data indicate that miR-320 is essential for the observed effects of K562 413

exosomes on both K562 cells and BMMSC. 414

HNRNPA1 mediates exosomal sorting of miR-320 415

To unravel the molecular mechanisms that control the active sorting into exosomes of specific 416

miRNAs, especially the miR-320 family, K562 cell lysates were incubated with streptavidin 417

beads coated with biotinylated miR-320a. The bound proteins were subjected to high-418

throughput identification by mass spectrometry (MS). Non-coated beads and beads with 419

biotinylated miR-374b-5p (which was a cell-enriched miRNA) served as negative control. 420

Among the enriched proteins precipitated with biotinylated miRNA-320a (Table S4), we 421

focused on HNRNPA1, a ubiquitously expressed RNA-binding protein recognizing the motif 422

UAGGG(A/U) (Figure S6A). 423

Consistently, HNRNPA1 was abundantly expressed in both leukemic cells and their 424

exosomes, as confirmed by both immunofluorescence confocal microscopy (Figure 7A) and 425

western blot (Figure 7B). Specific binding of HNRNPA1 to miR-320a was verified with 426

miRNA pull-down followed by western blot analysis of HNRNPA1 (Figure 7C). HNRNPA1 427

can be efficiently pulled down by biotinylated miR-320a, in comparison to biotinylated miR-428

374b-5p, demonstrating specific binding of HNRNPA1 and miR-320a in cells. The result was 429

further confirmed with immunoprecipitation of HNRNPA1 from cytoplasmic lysates of both 430

K562 cells (Figure 7D) and bone marrow mononuclear cells from BA mice (Figure S6B). The 431

effects of HNRNPA1 knockdown and overexpression (Figure S6C) on miR-320 sorting into 432

exosomes were further assessed. After HNRNPA1 knockdown, miR-320 was significantly 433

increased in the cellular components but decreased in the exosomes (Figure 7E). In contrast, 434

overexpression of HNRNPA1 increased the exosomal enrichment but decreased the 435

intracellular abundance of miR-320 (Figure 7F). No significant change of cellular and 436

exosomal expression of cell-enriched miR-374b-5p was found upon manipulating HNRNPA1 437

expression. Consistent with the role of HNRNPA1 in exosomal sorting of miR-320, knockdown 438

18

of HNRNPA1 significantly decreased the growth rate of K562 cells, which can be rescued by 439

miR-320 inhibitor (Figure 7G). Notably, inhibition of BCR/ABL by Imatinib decreased both 440

the cellular and exosomal HNRNPA1 protein level (Figure S6D), together with increased miR-441

320 in the cells while decreased level in the exosomes (Figure S6E). Collectively, these data 442

demonstrate that HNRNPA1 functions importantly in the sorting of miR-320 into leukemic 443

exosomes, and Imatinib might inhibit leukemia cell growth partially via interfering the pathway. 444

Abnormal expression of HNRNPA1 and exosomal miR-320 in clinical CML samples 445

Finally, we asked whether our findings in the leukemic cell line and murine model had clinical 446

relevance. HNRNPA1 mRNA and protein levels were markedly increased in CP (chronic 447

phase)-CML cells and became even higher in BC (blast crisis) -CML cells (Figure 8A and 8B), 448

compared to normal BM cells. Moreover, BC-CML patients had a much lower miR-320 449

expression in intracellular levels but higher in plasma exosomal levels, compared to CP-CML 450

patients (Figure 8C), which might be explained by the higher expression of HNRNPA1 in BC-451

CML cells. In addition, some other differentially miRNAs identified in the RNA-seq data were 452

also found expressed in a similar pattern in the CML cells and plasma exosomes (Figure S7A). 453

Consistent with the assumption that exosomal miR-320 might be transported from leukemic 454

cells to BMMSC, there were higher intracellular level of miR-320 in primary BMMSC derived 455

from CML patients than from healthy donors (Figure 8E). 456

Discussion 457

In this study, we have found that the exosomes secreted from leukemia cells reduced the tumor 458

suppressive miR-320 in the donor cells, resulting promoted cell growth. Moreover, we found 459

that leukemic exosomes can decrease the osteogenic differentiation potential of BMMSC both 460

in vitro and in mouse model, resulting in reduced trabecular bone volume. The study 461

demonstrate that the exosome derived from leukemia cells is an important contributor of niche 462

remodeling 463

The functional importance of bone marrow niche remodeling for failure of normal 464

hematopoiesis in myeloproliferative neoplasms has become increasingly evident [5, 7, 8, 32]. 465

During the progress of myeloproliferative neoplasms, the functions of bone marrow 466

19

mesenchymal stromal cells (BMMSC) are profoundly impaired. AML-derived BMMSCs are 467

molecularly and functionally altered with decreased proliferation and osteogenic differentiation, 468

which contribute to hematopoietic insufficiency [2, 3, 33-35]. The mesenchymal niche of CML 469

is also found aggressively modified into a self-reinforcing leukemic niche that impairs normal 470

hematopoiesis [5, 7, 32]. Previous studies have suggested an essential role for tumor cell-471

derived exosomes in reprogramming the tumor microenvironment. Kumar et al [36] found that 472

acute myeloid leukemia cell-derived exosomes can transform the bone marrow niche into a 473

leukemia-permissive microenvironment by impairing the normal functions of BMMSC. Here, 474

we further revealed that CML secreted exosomes could simultaneously promote the CML cell 475

growth and remodel the bone marrow niche toward leukemia permissive microenvironment via 476

inhibiting osteogenesis. 477

Exosomes derived from CML cells can shuttle miRNAs, amphiregulin and even bcr/abl 478

mRNAs to nearby stromal cells [37-39], resulting in reprogramming of niche cell functions. In 479

our study, deep sequencing of small RNAs in leukemic cells and their exosomes shows a 480

profound difference between intracellular and exosomal miRNAs. Among the exosome-481

enriched miRNAs, a higher proportion has previously been identified as tumor suppressive 482

miRNAs [29, 30, 40, 41], including miR-320 family, miR-3180-3p, miR-128-3p and miR-423-483

5p, compared to cell-enriched miRNAs. Based on our observation, leukemia cells are more 484

likely to selectively export these suppressive miRNAs to promote tumor cell growth. Although 485

we here focused on the role of tumor suppressor miR-320, the function of other miRNAs should 486

not be excluded. In fact, all of the encapsulated miRNAs might work together to remodel the 487

niche, while secretion of these miRNAs via exosomes also promotes the donor cell 488

proliferation. 489

To date, different pathways and molecules have been described to involved in the sorting 490

of specific miRNAs into exosomes. For example, heterogeneous nuclear ribonucleoprotein 491

A2B1 (HNRNPA2B1) and heterogeneous nuclear ribonucleoprotein Q (HNRNPQ) can bind to 492

miRNAs containing a certain sequence (called EXO motif) and guide their inclusion into 493

exosomes [17, 18]. In addition, major vault protein (MVP) is also reported to mediate the 494

loading of miRNAs into exosomes [42]. We here found that HNRNPA1 at least selectively 495

20

sorted out miR-320 into exosomes via the AGAGGG motif in the miRNAs. HNRNPA1 is a 496

highly abundant RNA-binding protein that has been implicated in diverse cellular functions 497

related to RNA metabolism including nucleo-cytoplasmic shuttling [43], alternative splicing 498

regulation [44], mRNA stability and miRNA processing [45]. Indeed, some leukemia-related 499

factors such as BCL-XL and SET are processed by HNRNPA1 [20]. HNRNPA1 expression is 500

also induced by BCR/ABL protein in a dose and kinase-dependent manner [20]. Consistent 501

with previous research, HNRNPA1 levels were higher in CP-CML cells, and even higher in 502

BC-CML cells, compared to normal bone marrow cells. Together, our findings suggest a novel 503

mechanism of how HNRNPA1 promotes leukemia progression: by selectively sorting tumor 504

suppressor miRNAs into exosomes. As leukemia stem cells (LSC) are the origin of CML, it is 505

thus worthwhile to confirm whether the HNRNPA1 mediated miR-320 sorting into exosomes 506

occurs in LSCs. It is also important to compare the differences between HSC and LSC derived 507

exosomes, while the study is limited by the technical difficulty of in vitro culture of HSCs and 508

LSCs without feeder cells. Single cell sequencing might be the solution. 509

In summary, this study has revealed that HNRNPA1 promotes cell growth via exosomal 510

sorting of tumor suppressive miRNAs (mainly miR-320) in CML, while the secreted exosomes 511

in turn function to remodel the niche to a leukemic favorable microenvironment via inhibiting 512

osteogenesis of BMMSC (Figure 8E). Thus, HNRNPA1 and the exosomal miRNAs provide 513

novel therapeutic targets to restore the normal function of the bone marrow niche. 514

Acknowledgements 515

We thank Dr F. Lu for her assistance with the Nikon microscope system, J. Kang for her help 516

with the electron microscopy. We are grateful to the technical help of Q. Chen and J. Zhang 517

from the Department of Ultrasound Medicine in Tangdu Hospital, Fourth Military Medical 518

University. This work was supported by the Nature Science Foundation of China 519

(NSFC31573244 to L Liu, NSFC31771507 to G Yang), Key Projects of Shaanxi Province 520

(2018ZDXM-SF-063 to L Liu) and State Key Laboratory of Cancer Biology Projects 521

(CBSKL2017Z10 to G. Yang). 522

Author contributions 523

21

X.G, Z.W, and M.W conducted the experiments. Y. D, Y. Z, X. C, Z. L, and W. Q performed 524

the gene expression analysis, exosome purification, intravenous injection and tissue dissection. 525

G. Y and L. L designed the experiments, analyzed the data, and wrote the manuscript. 526

References 527

1. Schepers K, Campbell TB, Passegue E. Normal and leukemic stem cell niches: insights and 528

therapeutic opportunities. Cell Stem Cell. 2015; 16: 254-67. 529

2. Geyh S, Rodriguez-Paredes M, Jager P, Khandanpour C, Cadeddu RP, Gutekunst J, et al. 530

Functional inhibition of mesenchymal stromal cells in acute myeloid leukemia. Leukemia. 531

2016; 30: 683-91. 532

3. von der Heide EK, Neumann M, Vosberg S, James AR, Schroeder MP, Ortiz-Tanchez J, et 533

al. Molecular alterations in bone marrow mesenchymal stromal cells derived from acute 534

myeloid leukemia patients. Leukemia. 2017; 31: 1069-78. 535

4. Kim JA, Shim JS, Lee GY, Yim HW, Kim TM, Kim M, et al. Microenvironmental 536

remodeling as a parameter and prognostic factor of heterogeneous leukemogenesis in acute 537

myelogenous leukemia. Cancer Res. 2015; 75: 2222-31. 538

5. Schepers K, Pietras EM, Reynaud D, Flach J, Binnewies M, Garg T, et al. Myeloproliferative 539

neoplasia remodels the endosteal bone marrow niche into a self-reinforcing leukemic niche. 540

Cell Stem Cell. 2013; 13: 285-99. 541

6. Colmone A, Amorim M, Pontier AL, Wang S, Jablonski E, Sipkins DA. Leukemic cells 542

create bone marrow niches that disrupt the behavior of normal hematopoietic progenitor cells. 543

Science. 2008; 322: 1861-5. 544

7. Zhang B, Ho YW, Huang Q, Maeda T, Lin A, Lee SU, et al. Altered microenvironmental 545

regulation of leukemic and normal stem cells in chronic myelogenous leukemia. Cancer Cell. 546

2012; 21: 577-92. 547

8. Welner RS, Amabile G, Bararia D, Czibere A, Yang H, Zhang H, et al. Treatment of chronic 548

myelogenous leukemia by blocking cytokine alterations found in normal stem and 549

progenitor cells. Cancer Cell. 2015; 27: 671-81. 550

9. Tkach M, Thery C. Communication by Extracellular Vesicles: Where we are and where we 551

need to go. Cell. 2016; 164: 1226-32. 552

22

10. van Niel G, D'Angelo G, Raposo G. Shedding light on the cell biology of extracellular 553

vesicles. Nat Rev Mol Cell Biol. 2018; 19: 213-28. 554

11. Zhou W, Fong MY, Min Y, Somlo G, Liu L, Palomares MR, et al. Cancer-secreted miR-555

105 destroys vascular endothelial barriers to promote metastasis. Cancer Cell. 2014; 25: 501-556

15. 557

12. Bruns H, Bottcher M, Qorraj M, Fabri M, Jitschin S, Dindorf J, et al. CLL-cell-mediated 558

MDSC induction by exosomal miR-155 transfer is disrupted by vitamin D. Leukemia. 2017; 559

31: 985-8. 560

13. Zhang L, Zhang S, Yao J, Lowery FJ, Zhang Q, Huang WC, et al. Microenvironment-561

induced PTEN loss by exosomal microRNA primes brain metastasis outgrowth. Nature. 562

2015; 527: 100-4. 563

14. Hornick NI, Doron B, Abdelhamed S, Huan J, Harrington CA, Shen R, et al. AML 564

suppresses hematopoiesis by releasing exosomes that contain microRNAs targeting c-MYB. 565

Sci Signal. 2016; 9: ra88. 566

15. Boyiadzis M, Whiteside TL. The emerging roles of tumor-derived exosomes in 567

hematological malignancies. Leukemia. 2017; 31: 1259-68. 568

16. Zietzer A, Werner N, Jansen F. Regulatory mechanisms of microRNA sorting into 569

extracellular vesicles. Acta Physiol (Oxf). 2018; 222: e13018. 570

17. Villarroya-Beltri C, Gutierrez-Vazquez C, Sanchez-Cabo F, Perez-Hernandez D, Vazquez 571

J, Martin-Cofreces N, et al. Sumoylated hnRNPA2B1 controls the sorting of miRNAs into 572

exosomes through binding to specific motifs. Nat Commun. 2013; 4: 2980. 573

18. Santangelo L, Giurato G, Cicchini C, Montaldo C, Mancone C, Tarallo R, et al. The RNA-574

binding protein SYNCRIP is a component of the hepatocyte exosomal machinery controlling 575

microRNA sorting. Cell Rep. 2016; 17: 799-808. 576

19. Perrotti D, Neviani P. From mRNA metabolism to cancer therapy: chronic myelogenous 577

leukemia shows the way. Clin Cancer Res. 2007; 13: 1638-42. 578

20. Iervolino A, Santilli G, Trotta R, Guerzoni C, Cesi V, Bergamaschi A, et al. hnRNP A1 579

nucleocytoplasmic shuttling activity is required for normal myelopoiesis and BCR/ABL 580

leukemogenesis. Mol Cell Biol. 2002; 22: 2255-66. 581

21. Koschmieder S, Gottgens B, Zhang P, Iwasaki-Arai J, Akashi K, Kutok JL, et al. Inducible 582

23

chronic phase of myeloid leukemia with expansion of hematopoietic stem cells in a 583

transgenic model of BCR-ABL leukemogenesis. Blood. 2005; 105: 324-34. 584

22. Studeny M, Marini FC, Dembinski JL, Zompetta C, Cabreira-Hansen M, Bekele BN, et al. 585

Mesenchymal stem cells: potential precursors for tumor stroma and targeted-delivery 586

vehicles for anticancer agents. J Natl Cancer Inst. 2004; 96: 1593-603. 587

23. Abbuehl JP, Tatarova Z, Held W, Huelsken J. Long-term engraftment of primary bone 588

marrow stromal cells repairs niche damage and improves hematopoietic stem cell 589

transplantation. Cell Stem Cell. 2017; 21: 241-55.e6. 590

24. Schmittgen TD, Livak KJ. Analyzing real-time PCR data by the comparative CT method. 591

Nat Protoc. 2008; 3: 1101-8. 592

25. Ao W, Gaudet J, Kent WJ, Muttumu S, Mango SE. Environmentally induced foregut 593

remodeling by PHA-4/FoxA and DAF-12/NHR. Science. 2004; 305: 1743-6. 594

26. Ambrosi TH, Scialdone A, Graja A, Gohlke S, Jank AM, Bocian C, et al. Adipocyte 595

accumulation in the bone marrow during obesity and aging impairs stem cell-based 596

hematopoietic and bone regeneration. Cell Stem Cell. 2017; 20: 771-84.e6. 597

27. Wang J, Hendrix A, Hernot S, Lemaire M, De Bruyne E, Van Valckenborgh E, et al. Bone 598

marrow stromal cell-derived exosomes as communicators in drug resistance in multiple 599

myeloma cells. Blood. 2014; 124: 555-66. 600

28. Reynaud D, Pietras E, Barry-Holson K, Mir A, Binnewies M, Jeanne M, et al. IL-6 controls 601

leukemic multipotent progenitor cell fate and contributes to chronic myelogenous leukemia 602

development. Cancer Cell. 2011; 20: 661-73. 603

29. Xishan Z, Ziying L, Jing D, Gang L. MicroRNA-320a acts as a tumor suppressor by 604

targeting BCR/ABL oncogene in chronic myeloid leukemia. Sci Rep. 2015; 5: 12460. 605

30. Yuan SX, Wang J, Yang F, Tao QF, Zhang J, Wang LL, et al. Long noncoding RNA 606

DANCR increases stemness features of hepatocellular carcinoma by derepression of 607

CTNNB1. Hepatology. 2016; 63: 499-511. 608

31. Kang DW, Noh YN, Hwang WC, Choi KY, Min do S. Rebamipide attenuates Helicobacter 609

pylori CagA-induced self-renewal capacity via modulation of beta-catenin signaling axis in 610

gastric cancer-initiating cells. Biochem Pharmacol. 2016; 113: 36-44. 611

32. Arranz L, Sanchez-Aguilera A, Martin-Perez D, Isern J, Langa X, Tzankov A, et al. 612

24

Neuropathy of haematopoietic stem cell niche is essential for myeloproliferative neoplasms. 613

Nature. 2014; 512: 78-81. 614

33. Huang JC, Basu SK, Zhao X, Chien S, Fang M, Oehler VG, et al. Mesenchymal stromal 615

cells derived from acute myeloid leukemia bone marrow exhibit aberrant cytogenetics and 616

cytokine elaboration. Blood Cancer J. 2015; 5: e302. 617

34. Jacamo R, Davis RE, Ling X, Sonnylal S, Wang Z, Ma W, et al. Tumor Trp53 status and 618

genotype affect the bone marrow microenvironment in acute myeloid leukemia. Oncotarget. 619

2017; 8: 83354-69. 620

35. Kornblau SM, Ruvolo PP, Wang RY, Battula VL, Shpall EJ, Ruvolo VR, et al. Distinct 621

protein signatures of acute myeloid leukemia bone marrow-derived stromal cells are 622

prognostic for patient survival. Haematologica. 2018; 103: 810-21. 623

36. Kumar B, Garcia M, Weng L, Jung X, Murakami JL, Hu X, et al. Acute myeloid leukemia 624

transforms the bone marrow niche into a leukemia-permissive microenvironment through 625

exosome secretion. Leukemia. 2018; 32: 575-87. 626

37. Ramos TL, Sanchez-Abarca LI, Lopez-Ruano G, Muntion S, Preciado S, Hernandez-627

Ruano M, et al. Do endothelial cells belong to the primitive stem leukemic clone in CML? 628

Role of extracellular vesicles. Leuk Res. 2015; 39: 921-4. 629

38. Corrado C, Saieva L, Raimondo S, Santoro A, De Leo G, Alessandro R. Chronic 630

myelogenous leukaemia exosomes modulate bone marrow microenvironment through 631

activation of epidermal growth factor receptor. J Cell Mol Med. 2016; 20: 1829-39. 632

39. Taverna S, Amodeo V, Saieva L, Russo A, Giallombardo M, De Leo G, et al. Exosomal 633

shuttling of miR-126 in endothelial cells modulates adhesive and migratory abilities of 634

chronic myelogenous leukemia cells. Mol Cancer. 2014; 13: 169. 635

40. Aslan D, Garde C, Nygaard MK, Helbo AS, Dimopoulos K, Hansen JW, et al. Tumor 636

suppressor microRNAs are downregulated in myelodysplastic syndrome with spliceosome 637

mutations. Oncotarget. 2016; 7: 9951-63. 638

41. Yu WW, Jiang H, Zhang CT, Peng Y. The SNAIL/miR-128 axis regulated growth, invasion, 639

metastasis, and epithelial-to-mesenchymal transition of gastric cancer. Oncotarget. 2017; 8: 640

39280-95. 641

42. Teng Y, Ren Y, Hu X, Mu J, Samykutty A, Zhuang X, et al. MVP-mediated exosomal 642

25

sorting of miR-193a promotes colon cancer progression. Nat Commun. 2017; 8: 14448. 643

43. Pinol-Roma S, Dreyfuss G. Shuttling of pre-mRNA binding proteins between nucleus and 644

cytoplasm. Nature. 1992; 355: 730-2. 645

44. Tavanez JP, Madl T, Kooshapur H, Sattler M, Valcarcel J. hnRNP A1 proofreads 3' splice 646

site recognition by U2AF. Mol Cell. 2012; 45: 314-29. 647

45. Kooshapur H, Choudhury NR, Simon B, Muhlbauer M, Jussupow A, Fernandez N, et al. 648

Structural basis for terminal loop recognition and stimulation of pri-miRNA-18a processing 649

by hnRNP A1. Nat Commun. 2018; 9: 2479. 650

26

Figure legends 651

652

Figure 1. Functional inhibition of bone marrow mesenchymal stromal cells in chronic 653

myelogenous leukemia. (A) Representative micro-CT cross-section images of femurs from 654

BCR/ABL+ (BA) mice and control (Ctrl) mice. Bar graphs demonstrate average cortical wall 655

thickness in compact bone regions, relative bone volume (BV/TV), number of trabeculae and 656

space between trabeculae in trabecular bone regions of control and BCR/ABL+ mice. Data 657

are expressed as mean ± SEM (n=6 mice per group). (B) Representative images show 658

decreased frequency of CFU-Fs in BMMSC derived from BA mice vs. control mice. Data are 659

expressed as mean ± SEM (n=9 mice per group). (C) Decreased osteogenic differentiation 660

potential of BMMSC from BA mice vs. control mice. Osteogenesis was indicated by alizarin 661

red staining on day 28. Data are expressed as mean ± SEM of three independent experiments. 662

(D) Alizarin red staining photomicrographs of decreased osteogenic differentiation potential 663

of BMMSC from CML patient (CML-BMMSC) vs. age-matched healthy control (Normal-664

BMMSC). Data presented are representative of 3 normal control and CML patients 665

respectively. ** p<0.01 by t test. 666

27

667

Figure 2. Leukemia cell-secreted exosomes are efficiently taken up by BMMSC. (A) 668

Transmission electron microscopy images of exosomes derived from K562 cells. (B) 669

Nanoparticle tracking analysis shows size distribution of K562-derived exosomes. (C) Western 670

blot analysis of exosome-specific marker TSG101 and Golgi-specific marker GM130 in lysates 671

from whole cells (Cell) or exosomes (EXO). GAPDH serves as internal control. (D) Coomassie 672

blue-stained sulfate polyacrylamide gel after electrophoretic separation of total lysates from 673

K562 cells (Cell) and derived exosomes (EXO). (E) Confocal fluorescence image of the 674

endocytosis of DiI-labeled exosomes by primary human BMMSC. (F) Ex vivo fluorescence 675

imaging of excised femurs from C57BL/6 mice injected with DiR-labeled exosomes from 676

different cell lines: K562, CML cell line; Hela, human cervical cancer cell line; and A549, 677

human adenocarcinoma cell line. Data are expressed as mean ± SEM (n=3 mice per group). (G) 678

28

qPCR analysis of cel-miR-39 level in femurs of C57BL/6 mice. Femurs were harvested 24 679

hours after tail vein injection of exosomes derived from indicated cell lines loaded with cel-680

miR-39. NC: negative control. Data are expressed as mean ± SEM of three independent 681

experiments. (H) Schematic of co-culture system in Transwell (membrane pore diameter, 0.4 682

μm) to measure exosome transfer (Left panel). BMMSC were co-cultured with K562 cells 683

transfected with cel-miR-39. The culture medium was added with or without exosome secretion 684

inhibitor GW4869. Cel-miR-39 in BMMSC was quantified by qPCR (right panel). Data are 685

expressed as mean ± SEM of three independent experiments. * p<0.05, ** p<0.01 by one way 686

ANOVA. 687

29

688

Figure 3. Leukemia cell-secreted exosomes inhibit BMMSC function in vitro and in vivo. 689

(A) ALP and Alizarin red staining analysis of osteogenesis. BMMSC cultured in osteogenic 690

condition were additionally treated with PBS, or exosomes derived from either healthy human 691

serum (Serum-EXO) or cultured K562 cells (K562-EXO) every 3 days, followed by ALP 692

staining after 14 days or by alizarin red staining after 28 days. Images are representative of 693

three independent experiments. (B) Western blot analysis of RUNX2 expression in human 694

BMMSC treated as indicated. Cells were cultured in osteogenic condition for 3 days. Images 695

are representative of three independent experiments. (C) qPCR analysis of mRNA expression 696

of indicated genes in BMMSC treated same as above. Data are presented as the mean ± SEM 697

30

of three independent experiments. (D) microCT analysis of the femurs from C57BL/6 mice 698

treated as indicated. Mice were intravenously injected with 50 μg of indicated exosomes twice 699

per week for four consecutive weeks (Nor-EXO: exosomes derived from bone marrow 700

mononuclear cells (BM-MNC) of control normal mice; CML-EXO: exosomes derived from 701

the BM-MNC of BA mice). Mice treated with PBS served as control. Micro-CT was used to 702

illustrate the effect on cross-sections of femurs in trabecular bone regions and to quantify 703

relative bone volume (BV/TV) and number of trabeculae. Data are expressed as the mean ± 704

SEM (n=6 mice per group). (E) Representative images show the CFU-Fs of the BMMSC 705

derived from mice with indicated treatments (left panel). Data are presented as the mean ± 706

SEM of three different experiments from 3 mice per group. * p<0.05, ** p<0.01 by one way 707

ANOVA. 708

31

709

Figure 4. Specific repertoires of miRNAs are sorted into leukemia exosomes. (A) Deep 710

sequence data visualization by scatter plot comparing miRNAs in K562 cells and derived 711

exosomes. (B) Heat map depicting differentially expressed miRNAs between cells (Cell) and 712

exosomes (EXO) from human K562 cells. These miRNAs had overall TPM (transcripts per 713

million reads) >1000 in at least one group and fold change difference >2. (C) Verification of 714

32

selected miRNAs using qPCR. Data are presented as the mean ± SEM of three independent 715

experiments. ** p<0.01 by t test. (D) Cladogram showing the multiple alignment of mature 716

sequences of 14 leukemic exosome-enriched miRNAs (red) and 31 cell-enriched miRNAs 717

(black). (E) Over-represented motifs on exosome-enriched miRNAs. The core AGAGGG 718

shows high information content. 719

33

720

Figure 5. Exosomal miR-320 plays an essential role in leukemia cell growth. (A) Schematic 721

diagram of the putative binding sites of the miR-320 family in the 3’ untranslated region (UTR) 722

of bcr/abl. (B) BCR/ABL (P210) protein level, measured by western blot, in K562 cells 723

transfected with different dose of miR-320a mimics or scramble (NC) for 48 hr. Images are 724

representative of three independent experiments. (C) Proliferation of K562 cells with miR-725

320a (miR-320 OE) or scrambled miRNA (miR-scramble) overexpression, as analyzed by 726

CCK-8 assay. (D) Cell cycle phase analysis of K562 cells transfected with either miR-320a 727

(miR-320 OE) or scrambled control (miR-scramble). The percentage of cells in the G1, S, and 728

G2 phases is shown in the bar graph. Data are presented as the mean ± SEM of three 729

independent experiments. (E) EdU incorporation assay of K562 transfected with either miR-730

320a (miR-320 OE) or scrambled miRNA (miR-scramble). Images are representative of three 731

independent experiments. * p<0.05, ** p<0.01 by t test or two way ANOVA. 732

34

733

Figure 6. Exosomal miR-320 plays an essential role in osteogenesis of BMMSC. (A) 734

Schematic diagram of the putative binding sites of the miR-320 family in the 3’UTR of β-735

catenin. (B) Expression of β-catenin in BMMSC with indicated treatments. Images are 736

representative of three independent experiments. (C) Expression of β-catenin in BMMSC with 737

indicated exosome treatments. Serum-EXO: healthy human serum exosomes; miR-scramble-738

EXO and miR-320 KD-EXO: exosomes from control or miR-320 knocked down K562 cells 739

(infected with control or miR-320 sponge); EXO and EXO+miR-320: exosomes from 740

HEK293T cells loaded without or with miR-320. Images are representative of three 741

independent experiments. (D) Alizarin red staining showing the osteogenic differentiation of 742

BMMSC treated with indicated exosomes. BMMSC treated with healthy human serum 743

exosomes (Serum-EXO) served as control. Images are representative of 3 independent 744

experiments. (E) Micro-CT analysis of the femurs of mice treated as indicated. C57BL/6 mice 745

35

were tail vein injected with 50 μg of indicated exosomes twice per week for four consecutive 746

weeks. Mice treated with PBS served as control. Relative bone volume (BV/TV), trabecular 747

thickness, trabecular separation, and trabecular numbers were further analyzed. Data are 748

presented as the mean ± SEM (n=6 mice per group). * p<0.05, ** p<0.01, *** p<0.001 by one 749

way ANOVA. 750

36

751

Figure 7. HNRNPA1 mediates exosomal sorting of miR-320. (A) Confocal microscopy 752

image shows the co-localization of HNRNPA1 (green) and specific exosome marker TSG101 753

(red) in K562 cells. (B) Western blot showing HNRNPA1 and CD9 expression in both cellular 754

(Cell) and exosomal (EXO) lysates. GAPDH serves as internal control. (C) Western blot 755

showing HNRNPA1 pulled down by biotinylated miR-320 in whole cellular extracts (WCE) 756

of K562 cells, with biotinylated miR-374b-5p serving as control. Images are representative of 757

three independent experiments. (D) RNA-IP analysis of miR-320 bound by HNRNPA1. K562 758

cytoplasmic extracts were incubated with IgG or antiHNRNPA1 and the immunoprecipates 759

were isolated for qRT-PCR analysis. miR-374b-5p serves as a negative control. Data are means 760

± SEM of three independent experiments. (E) qRT-PCR analysis of intracellular (left) and 761

exosomal (right) levels of selected miRNAs in HNRNPA1 knock down (shHNRNPA1) or 762

control K562 cells (shControl). The values were normalized to U6 levels in cells or exosomes 763

and shown as mean ± SEM of three independent experiments. (F) qRT-PCR analysis of 764

37

intracellular (left) and exosomal (right) levels of selected miRNAs in control (GFP) or 765

HNRNPA1 overexpression K562 cells. Data are means ± SEM of three independent 766

experiments. (G) Proliferation of K562 cells treated as indicated were further analyzed by 767

CCK-8 assay. Data are expressed as mean ± SEM of 3 different experiments. * p<0.05, ** 768

p<0.01, *** p<0.001 by t test. 769

770

38

771

Figure 8. Abnormal expression of HNRNPA1 and miR-320 in chronic myelogenous 772

leukemia patients. mRNA (A) and protein (B) expression of HNRNPA1 in primary human 773

bone marrow mononuclear cells (BM-MNC) of normal control, chronic phase (CML-CP) or 774

blast crisis CML (CML-BC). (C) BM-MNC cellular and plasma exosomal expression of miR-775

320 in CML-CP or CML-BC patients. (D) Intracellular expression level of miR-320 in 776

BMMSC derived from healthy control and CML patients. (E) Graphic illustration of the study: 777

39

HNRNPA1 mediated exosomal transfer of miR-320 from leukemia cell to mesenchymal 778

stromal cell simultaneously promotes leukemia cell growth while inhibits mesenchymal 779

stromal cell osteogenesis. * p<0.05, ** p<0.01 by t test or one way ANOVA. 780