Large-scale production of megakaryocytes from human pluripotent stem cells by chemically defined forward programming

Abstract

The production of megakaryocytes (MKs)--the precursors of blood platelets--from human pluripotent stem cells (hPSCs) offers exciting clinical opportunities for transfusion medicine. Here we describe an original approach for the large-scale generation of MKs in chemically defined conditions using a forward programming strategy relying on the concurrent exogenous expression of three transcription factors: GATA1, FLI1 and TAL1. The forward programmed MKs proliferate and differentiate in culture for several months with MK purity over 90% reaching up to 2 × 10⁵ mature MKs per input hPSC. Functional platelets are generated throughout the culture allowing the prospective collection of several transfusion units from as few as 1 million starting hPSCs. The high cell purity and yield achieved by MK forward programming, combined with efficient cryopreservation and good manufacturing practice (GMP)-compatible culture, make this approach eminently suitable to both in vitro production of platelets for transfusion and basic research in MK and platelet biology.

ARTICLE

Received 23 Jun 2015 | Accepted 2 Mar 2016 | Published 7 Apr 2016

Thomas Moreau1,2,3, Amanda L. Evans1,3,*, Louella Vasquez4,*, Marloes R. Tijssen1, Ying Yan4,Matthew W. Trotter2,w, Daniel Howard1,3, Maria Colzani1,3, Meera Arumugam1,3, Wing Han Wu1,3, Amanda Dalby1,3, Riina Lampela3, Guenaelle Bouet1,3, Catherine M. Hobbs1,3, Dean C. Pask1,3, Holly Payne5, Tatyana Ponomaryov5, Alexander Brill5, Nicole Soranzo4, Willem H. Ouwehand1, Roger A. Pedersen2,3,**& Cedric Ghevaert1,3,**

The production of megakaryocytes (MKs)the precursors of blood plateletsfrom human pluripotent stem cells (hPSCs) offers exciting clinical opportunities for transfusion medicine. Here we describe an original approach for the large-scale generation of MKs in chemically dened conditions using a forward programming strategy relying on the concurrent exogenous expression of three transcription factors: GATA1, FLI1 and TAL1. The forward programmed MKs proliferate and differentiate in culture for several months with MK purity over 90% reaching up to 2 105 mature MKs per input hPSC. Functional platelets are

generated throughout the culture allowing the prospective collection of several transfusion units from as few as 1 million starting hPSCs. The high cell purity and yield achieved by MK forward programming, combined with efcient cryopreservation and good manufacturing practice (GMP)-compatible culture, make this approach eminently suitable to both in vitro production of platelets for transfusion and basic research in MK and platelet biology.

1 Department of Haematology, University of Cambridge and NHS Blood and Transplant, Long Road, Cambridge CB2 0PT, UK. 2 The Anne McLaren Laboratory, Wellcome Trust-Medical Research Council Cambridge Stem Cell Institute and Department of Surgery, University of Cambridge, West Forvie Site, Robinson Way, Cambridge CB2 0SZ, UK. 3 Wellcome Trust-Medical Research Council Cambridge Stem Cell Institute, Tennis Court Road, Cambridge CB2 1QR, UK.

4 Human Genetics, Wellcome Trust Sanger Institute, Genome Campus, Hinxton CB10 1RQ, UK. 5 Institute of Cardiovascular Sciences, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK. w Present address: Celgene Institute for Translational Research Europe (CITRE), Parque Cientco y

Tecnolgico Cartuja 93, Centro de Empresas Pabelln de Italia, Isaac Newton, 4, Seville E-41092, Spain. * These authors contributed equally to this work. ** These authors jointly supervised this work. Correspondence and requests for materials should be addressed to R.A.P. (email: mailto:[email protected]

Web End [email protected] )or to C.G. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 7:11208 | DOI: 10.1038/ncomms11208 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

DOI: 10.1038/ncomms11208 OPEN

Large-scale production of megakaryocytes from human pluripotent stem cells by chemically dened forward programming

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11208

Megakaryocytes (MKs) generate blood platelets whose primary role is to stop haemorrhages via localized clot formation at the site of vessel injury1,2. MKs are

polyploid cells derived from haematopoietic stem cells residing in the bone marrow where they represent only 0.01% of the total nucleated blood cells. By extension of cytoplasmic protrusions through bone marrow sinusoids, they release daily B1 1011

platelets into the blood stream to sustain the count of short-lived (710 days) circulating platelets between 150450 109 per litre

of blood3,4. A decrease in platelet number, or thrombocytopenia, may occur following bone marrow failure (inherited or acquired, such as post-cancer treatment) or severe peripheral bleeding after trauma or surgery, and potentially leads to life-threatening haemorrhages. Currently, prophylactic and therapeutic treatment essentially relies on transfusion of ABO and Rhesus-D-matched platelet concentratesat 42.4 1011 platelets per unitfrom

voluntary donations5,6. Recently, the increase in high-dose cancer therapy, advanced surgical procedures and the ageing population has led to a rising demand for platelets with over 4.5 million platelet units transfused per year in Europe and the United States7. In addition, platelet transfusion refractoriness in HLA class I alloimmunized chronically transfused patients and multiparous women necessitates the special provision of matched platelet units sourced from a small pool of genotyped recallable donors8. Altogether, the dependence on donations combined with the limited shelf life of platelet concentrates (57 days) represents a logistical, nancial and biosafety challenge for health organizations worldwide.

Human pluripotent stem cells (hPSCs)including embryonic stem cells (hESCs) derived from embryos and induced PSCs (hiPSCs) generated from post-natal somatic cellscan be maintained in vitro for prolonged periods while retaining the capacity to differentiate towards virtually any cell type upon adequate stimulation911. Therefore, they offer huge opportunities for basic research and clinical applications12. The production of platelets in vitro from genetically dened hPSC lines could revolutionize transfusion medicine by providing a controllable source of platelets13. Moreover, platelets are anucleate and do not proliferate which means they can be irradiated before transfusion. This provides a marked safety advantage over other hPSC-derived therapeutic cells which can potentially retain oncogenic cell fractions14. However, in vitro systems for the production of large amounts of MKs and subsequent platelet release to match the needs for making transfusion units still require considerable optimization.

Our work describes a novel approach for generating large quantities of functional MKs from hPSCs with unique advantages for clinical development. Existing protocols have so far relied on external signals provided by cytokines or stromal cells to mimic embryonic development in vitro and thus direct sequential differentiation of hPSCs into MKs, a process designated as directed differentiation1520. While mature MKs showing functional platelet release are produced, this strategy has been limited by the relatively low number of MKs generated or by the complex genetic modications and clonal selection required to immortalize MKs post differentiation. Urged by the recent discoveries on the plasticity of cell identities controlled by limited sets of transcription factors (TFs)21, we adopted a radically different approach for the generation of MKs by exploring the potential of exogenous TFs to drive the differentiation process from hPSCs, a strategy called forward programming (FOP). Proceeding from a methodically curated list of candidate genes, we discovered that the combination of GATA1FLI1TAL1 uniquely promoted highly efcient MKFOP from an array of hPSC lines in chemically dened conditions. Critically, the forward programmed MKs (fopMKs)

matured into platelet-producing cells that could be cryopreserved, maintained and amplied in vitro for over 90 days showing an average yield of 200,000 MKs per input hPSC. This unprecedented efciency combined with minimal cell manipulation and low cytokine requirements makes MK-FOP a promising platform for basic research as well as future clinical applications in the eld of transfusion medicine.

ResultsGATA1FLI1TAL1 induces hPSC to MK commitment. Based on the knowledge that cooperative binding of TFs to target sites can activate gene expression in repressed chromatin22, we hypothesized that MK-specic TFs forming complexes able to interact with chromatin remodelers would be able to rewire the hPSC gene regulatory network to induce MK-FOP. We compared TF expression in the H9 hESC line and cord blood-derived megakaryocytes (cbMKs)23 and narrowed down to a list of 46 TF candidates subsequently ranked based on differential gene expression level and reported protein interactions between candidates and epigenome modiers using VisANT24 (see Supplementary Fig.1 and Methods for details). Nine TFs from the top 20 candidates were individually cloned into a lentiviral vector backbone under the control of an EF1a promoter to assess their MK-FOP potential (Fig. 1a).

H9 cells were transduced concurrently with equal amounts of each lentiviral vector and subsequently maintained in hPSC medium (FGF2 Activin-A) for 2 days followed by MK medium

(thrombopoietin (TPO) SCF (stem cell factor)) for a further 5

days. The transduction efciency measured with a green uorescent protein (GFP) reporter vector was 60.35.6% in these conditions (n 4). Transduction with nine TFs generated a

well-dened population of cells expressing CD41a (integrin alpha-IIb: ITGA2B) on day 7 (1.10.6%; Fig. 1b). In an attempt to identify the critical TFs responsible for the generation of this population expressing the marker typically associated with MK lineage commitment, CD41a cells were sorted and transgene

expression measured by RT-qPCR (quantitative PCR with reverse transcription). Intriguingly, we detected a striking bias for GATA1, FLI1 and TAL1 (hereafter 3-TFs) transgene expression in the CD41a population compared with its negative

counterpart which suggested their instrumental role in the acquisition of the CD41a phenotype (Fig. 1b). Accordingly, the 3-TF combination tested in two hiPSC lines generated more CD41a cells compared with all nine TFs together demonstrat

ing its superior efciency (2.50.04% versus 0.20.04% respectively; Fig. 1c). We assessed all permutations of the 3-TFs and showed that the maximum CD41a cell yield was achieved

upon concurrent transduction of the 3-TFs which also correlated with a higher clonogenic potential (Fig. 1d; Supplementary Fig. 2a). Importantly, no CD41a cells were detected after

transduction with a GFP control vector demonstrating that culture conditions were not inductive per se (Fig. 1d). Finally, we combined the 3-TFs using a single lentiviral vector carrying a polycistronic expression cassette encoding for the 3-TFs and a GFP reporter. Virtually all GFP-positive cells differentiated into CD41a cells by day 7, conrming the efcacy of the 3-TFs in

inducing MK-FOP (Supplementary Fig. 2b).Additional analyses conrmed that the emerging CD41a

population at day 7 truly represented early MK lineage commitment. We rst showed that the expression of key MK genes including MPL (coding for the TPO receptor), the TFs ZFPM1, RUNX1, NFE2 as well as the endogenous expression of GATA1, FLI1 and TAL1 were specically induced in the CD41a cell population (Fig. 1e). Moreover, the MK clonogenic

potential was exclusively retained in the CD41a cell population

which further developed to mature MK colonies expressing

2 NATURE COMMUNICATIONS | 7:11208 | DOI: 10.1038/ncomms11208 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11208 ARTICLE

a b

PSC CB-MK TF nodes CH nodes

Rank Gene

1 GATA12 IKZF13 SPI14 HOXA55 MEIS16 FLI17 PBX18 ZBTB169 IKZF310 TAL111 NFATC112 TSC22D313 KLF114 ZFPM215 STAT416 NFE217 ZNF46718 ZFPM119 RELA20 MEF2C

CD41a +ve CD41a ve

TAL1 GATA1

FLI1 MEIS1

7 5 2 6 2 5 2 1 2 1 2 0 2 0 1 9 1 4 1 3 1 2 1 2 1 1 1 1 1 1 1 1 1 0 1 0 0 10 0 6

H2AFY STAT4

ZFPM1

NFATC1

10

GFP log

10

ZNF467

+

SPI1

10

FL1

1.1 0.6%

ZFPM2 MEIS1

TAL1 PBX1

ZBTB16

TCSD

+

+

IKZF1

KLF1

ZNFN1A3 TSC22D3

RELA

SPI1 ZBTB16

PBX1 MEF2C

NFE2

GATA1

10

HOXA5

+

FLI1

10

10 10 10 10 10

CD41a-PE

NFE2 MEF2C

0 0

100

200

100

100,000

Transgene expression (rel. HMBS)

4 Row Z-score

+4

c d

12 **

1.4

CD41a+ cells fold increase

(rel. 9-TFs)

10

1.2

*

CD41a+ cells fold increase

(rel. 3-TFs)

1

8

0.8

6

0.6

4

0.4

2

0.2

0 NT 9-TFs

0 3TFs G+F

G+T F+T G F T GFP

e f g

CD41a +ve CD41a ve

0 1 2 3 0 1 2 3

35

CD41a +ve CFU

CD41a

MPL ZFPM1 RUNX1

NFE2 GATA1

FLI1 TAL1 PBX1 MEF2C

MEIS1

MK colonies per 5,000 cells

30

25

20

15

10

5

CD42b CD41a

0 CD41a +ve CD41a ve

Endogene expression (rel. HMBS)

3-TFs

Figure 1 | TF candidate screening for MK forward programming. (a) TF expression in hPSCs versus cbMKs plotted as a row normalized heatmap with indications of recorded TF internal and chromatin modiers node numbers and protein interactions among the top 20 TF candidates reported by VisANT. The nine experimentally tested TFs are highlighted in red. (b) The H9 hESC line was concurrently transduced with the 9-TFs and maintained in pluripotency medium (FGF2 Activin-A) for 2 days followed by MK medium (TPO SCF) for a further 5 days. CD41a cells generated 7 days after lentiviral

transduction (dot plot, mean% s.e.m., n 5; FL1: 530/40 nm channel) were sorted by ow cytometry and transgene expression levels quantied by

RT-qPCR (n 1). (c) The percentage of CD41a cells was monitored by ow cytometry 7 days after transduction of the hiPSC lines #1 and #2 with the

9-TFs or 3-TFs combination. Bar graphs represent the fold increase of CD41a cell count relative to the 9-TFs combination (mean s.e.m., n 4;

**Po0.01 by two-tail t-test). NT: non-transduced cells. (d) The hiPSC#1 line was transduced with all permutations of the 3-TFs and percentages of CD41a cells measured by ow cytometry at day 7. Bar graphs represent the fold increase of CD41a cell count relative to the 3-TFs combination (mean

s.d., n 2; *P 0.06 and Po0.01 versus G F and other combination respectively by two-tail t-test). G: GATA1, F: FLI1, T: TAL1, GFP: control vector.

(e) The endogenous expression of key MK genes was monitored by RT-qPCR from CD41a ow sorted cells 7 days after transduction of the hiPSC lines #1 and #2 (mean s.d., n 2). (f) The hiPSC lines #1 and #2 were transduced with the 3-TFs and sorted by ow cytometry for expression of CD41a at day 7.

The clonogenic potential of sorted cells was tested in methylcellulose semi-solid medium supplemented with TPO and SCF. The number of colonies per 5,000 sown cells was determined after 10 days from duplicate wells (means.e.m., n 4). (g) A representative MK colony obtained from a CD41a cell

co-expressing CD41a and CD42b as detected by immunouorescence is shown (scale bar, 50 mm).

CD42b (glycoprotein-Ib: GPIBA) functionally demonstrating MK commitment (Fig. 1f,g). Altogether, we thereby identied GATA1, FLI1 and TAL1 as a minimal and sufcient combination of TFs to induce the formation of MK precursors from hPSCs.

Chemically dened MK-FOP generates high purity MK cultures. To facilitate transfer to the clinic, we developed a chemically dened xeno-free 3-TF MK-FOP protocol. Induction of

mesoderm commitment from hiPSCs by a 2-day exposure to FGF2, BMP4 and LY-294002 (ref. 25) signicantly increased the percentage of CD41a cells at day 7 compared with

pluripotency maintenance conditions (5.41.6-fold increase; Supplementary Fig. 3a). Next, we assessed the benet of using forced aggregation embryoid body formation instead of the base two-dimensional (2D) culture. When coupled with mesoderm-inducing conditions, embryoid body culture further improved the

NATURE COMMUNICATIONS | 7:11208 | DOI: 10.1038/ncomms11208 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11208

published results for in vitro-derived neonate and hPSC MKs17,28, the cell ploidy remained low on average and was comparable for both fopMKs and cbMKs (o2% 8N cells;

Supplementary Fig. 3g). Analysis of the fopMK ultrastructure by electron microscopy showed lobulated nuclei, developing demarcation membrane system (cytoplasmic cell membrane supply for platelet release), cytoplasmic multi-vesicular bodies (precursors to platelet granules) and mature granules (Fig. 2g).We conrmed by confocal microscopy that major alpha-granule proteins (P-selectin, thrombospondin, brinogen and vWF) were indeed correctly expressed and patterned in fopMKs (Fig. 2h).Importantly, the total expression level of the 3-TFs in fopMKs as sum of transgenic and endogenous expressionwas not signicantly different from cbMKs (Supplementary Fig. 3h).Collectively, these data demonstrate the efcient production and maturation of MKs derived by forward programming in xeno-free chemically dened conditions.

MK-FOP produces expandable and cryobankable mature MKs.

We found that fopMKs could be maintained in culture and kept expanding for an extended period of up to 60 days in the TPO IL1b condition described above (Supplementary Fig. 4ac).

The expression of KIT (the receptor for SCF, an haematopoietic progenitor pleiotropic cytokine) on a fraction of the cells suggested the persistence of a progenitor population in MK-FOP cultures (Supplementary Fig. 4d). This led us to test the continuous supplementation with SCF (50 ng ml 1)instead of

IL1b originally used through the second step of culturein an attempt to improve further long-term maintenance. In addition, we reasoned that the high TPO level (100 ng ml 1) may not be required in the MK-FOP context where differentiation was sustained internally by expression of the 3-TFs, and indeed high TPO may be responsible for precocious exhaustion of MK-FOP cultures by over stimulation of differentiation. Consequently, we tested a lower TPO concentration (20 ng ml 1) in combination with SCF through the second step of MK-FOP (Fig. 3a). In these conditions, we were able to maintain fopMKs in culture with steady proliferation for at least 90 days, achieving close to an average 200,000 MK yield per input hiPSC (1.941.59 105,

n 7 for hiPSC#1 and #3 cumulatively; Fig. 3b). This was in

striking contrast with the maximum 1,300 MK fold increase, earlier loss of CD42a expression and cell viability in long-term culture using the original high TPO and IL1b condition (Supplementary Fig. 4ac). The MKs harvested from optimized long-term cultures (430 days, dened thereafter as LT-fopMKs)

maintained a purity of over 90% CD41a cells with levels of

CD42a expression 460% (Fig. 3ce) and an increase in late

Figure 2 | Generation of mature MKs by forward programming using chemically dened conditions. (a) Schematic representation of the optimized MK-FOP protocol. Viral transduction at day 0 concurrent with embryoid body generation and mesoderm induction for 2 days was followed by a period of culture in an MK induction medium (TPO SCF) for 8 days. Embryoid bodies showing cystic structures and actively growing cell aggregates were

dissociated to single cells at day 10 and further differentiated to mature MKs (TPO IL1b) until day 20 post transduction. (b) Time course of fopMK

differentiation showing MK lineage commitment (%CD41a cells) and MK maturation (%CD42a cells) from whole culture (means.e.m. from hiPSC

lines #1 and #2; n 2 (day010); n 6 (day1422)). Representative ow cytometry dot plots for CD41a and CD42a expression are shown below. (c) The

MK fold increase at day 20 relative to the day 0 hiPSC input is shown on a logarithmic scale for the hiPSC lines #14 differentiated by forward programming (fopMK: means.e.m.; n 14, 7, 3, 5 respectively) or directed differentiation (hiPSCs#1 and #2; ddMK: means.e.m.; n 3,2 respectively). **Po0.01 and

*Po0.05 by two-tail t-test. (d) Representative histograms of the expression of major platelet receptors detected by ow cytometry are shown for day 20 fopMKs (red line) and day 10 cbMKs (blue line) against isotype control (grey shade). (e) The morphology of day 20 fopMKs and day 10 cbMKs was analysed by modied Romanowsky staining on xed cells. Arrowheads point to multinucleated cells. Scale bars, 25 mm. (f) Cell size distribution from fopMK and cbMK cultures is shown as box plots: centre lines show the medians; box limits indicate the 25th and 75th percentiles; whiskers extend 1.5 times the interquartile range from the 25th and 75th percentiles, outliers are represented by dots (n 50, 74 respectively; **Po0.01 by two-tail t-test).

(g) Cell ultrastructure of fopMKs and cbMKs was visualized by transmission electron microscopy. dms, demarcation membrane system; gr, granules; mvb, multi-vesicular body; mit, mitochondria. Scale bars, 2 mm. (h) Representative confocal pictures of fopMKs immunostained for major alpha-granule proteins (thrombospondin, brinogen, P-selectin and Von Willebrand Factor; iPSC#3 fopMKs, day 40). Scale bars, 25 mm.

yield of CD41a cells at day 7 (5.20.9-fold increase;

Supplementary Fig. 3b). The embryoid body culture had the additional advantage of providing a standardized quantity of input cells thereby conferring reproducibility. The nal protocol combined lentiviral transduction with embryoid body formation over the rst 24 h with mesoderm induction for the rst 2 days. This protocol allowed consistent transduction efciencies amongst hiPSC lines (68.24.1% using a single GFP control vector, n 11 using three hiPSC lines), resulting in 22% co

transduction efciency as estimated using reporter vectors (Supplementary Fig. 3c). As MK maturation was not optimally sustained in the low-adherence embryoid body culture setting, we included a single-cell dissociation step which gave an optimal MK yield when performed at day 10 post transduction (Supplementary Fig. 3d) followed by a further 10 day culture in medium containing TPO and IL1b routinely used for cbMK differentiation26. The optimized MK-FOP protocol using xeno-free GMP-grade basal medium and recombinant cytokines is depicted in Fig. 2a.

We observed a gradual increase in CD41a cells from day 4

post transduction followed by the acquisition of the mature MK marker CD42a (glycoprotein IX:GP9, component of the Von Willebrand platelet receptor complex) from day 6 onwards mimicking cbMK differentiation (Fig. 2b; Supplementary Fig. 3e). Critically, using two different hiPSC lines, MK-FOP consistently achieved MK lineage purity (495% CD41a cells) by day 15

post transduction with 450% CD42a mature MKs by day 20

(Fig. 2b). A robust cell expansion during the single-cell culture step led to the generation of large quantities of MKs with up to28.47.8-fold increase relative to the hiPSC input at day 20 (iPSC#1, n 14; Fig. 2c). Interestingly, MK-FOP achieved

signicantly higher cell yields compared with a standard MK-directed differentiation approach27, producing in our hands on average 26.3 and 11.7 times more mature MKs from the hiPSC lines #1 and #2, respectively (Fig. 2c). In addition, the MK purity obtained by MK-FOP was signicantly higher than MK-directed differentiation (97.70.8% versus 21.32.9%, respectively; Supplementary Fig. 3f).

To further assess the MK identity of FOP cells, day 20 fopMKs were compared with day 10 cbMKs as a benchmark. At this stage, both cultures show similar MK maturity (5060% CD42a cells;

Fig. 2b; Supplementary Fig. 3e). The key platelet surface receptors for brinogen (aIIb-b3), Von Willebrand factor (gpIb-V-IX) and collagen (gpVI) were readily detected in both cultures (Fig. 2d). Morphologically, fopMK cultures displayed a typical mix of megakaryoblasts with large peripheral nuclei and mature MKs with increased cytoplasmic volume and frequent polyploid cells similarly to cbMK cultures (Fig. 2e,f). In accordance with

4 NATURE COMMUNICATIONS | 7:11208 | DOI: 10.1038/ncomms11208 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11208 ARTICLE

maturation markers like GPVI (Supplementary Fig. 4e). The LT-fopMK cultures contained a mixed population of small megakaryoblasts growing in loose clusters representing the actively proliferating cell fraction together with larger polyploid MKs (Fig. 3fh). Critically, LT-fopMKs were successfully frozen

over a day 2170 timeframe and subsequently recovered for further culture and expansion allowing cryobanking of fopMK batches (Supplementary Fig. 4f).

We found that LT-fopMK cultures were not immortalized but nite with the longest iPSC#1 and iPSC#3 cultures kept for 120

a iPSC

Embryoid body culture Single cell culture

Embryoid bodies Megakaryocytes

GATA1 FLI1 TAL1

Day 0

2 10 20

Mesoderm [FGF2+BMP4]

MK progenitor [TPO+SCF]

MK maturation [TPO+IL1b]

b

c

MK -CD41a+

Mature MK -CD42a+

**

100

50

MK-CD41a+

90

100.0

Mature MK-CD42a+

%positive cells

80

28.4

70

15.8

*

60

MK fold increase

(rel. iPSC input)

7.7

40

10.0

30

3.5 2.8 3.0

3.7 3.2

1.2

20

10 0 0 5 10 15 20 25

1.0

0.6

0.5

Forward programming day

0.3

Day0 Day10 Day15 Day20

0.1

CD41a

iPSC#1 iPSC#2 iPSC#3 iPSC#4 iPSC#1 iPSC#2

fopMK ddMK

CD42a

d

e

Receptors

fopMKs cbMKs

400 300 200

f

CollagenVon willebrand factor

Counts Counts Counts

250 187 125

62

0

250 187 125

62 0

433 324 216 108

0

cbMKsfopMKs

Counts

CD61 (gpIIIa)

CD41a (gpIIb)

cbMKs(74)

fopMKs(50)

Counts Counts

100

0

500 375 250 125

0

Fibrinogen

0

10 20 30 40

Cell size (m)

**

612 459 306 153

0

RFGP56 (gpIIb/IIIa)

CD42a (gpIX)

CD42b (gpIb)

GPVI

g

h

vwf Fibrinogen

p-selectin Thrombospondin

dms

gr

fopMKs

mvb

mit

Alpha-tubulin

cbMKs

mit mvb

gr

dms

NATURE COMMUNICATIONS | 7:11208 | DOI: 10.1038/ncomms11208 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11208

a e

b

Day0 10

2 >90

Embryoid body culture Single cell culture

Day20 Day30 Day90

Day40

74% 89% 75% 68%

CD41a

1E+06

Mature MK fold increase

(rel. iPSC input)

1E+05

CD42a

1E+04

f g

1E+03

1E+02

Average iPSC#1 iPSC#3

1E+01

1E+00

0 25 50 75 100

Forward programming day

Forward programming day

c

d

120%

100%

h

100%

Percentage MK

(CD41a+)

80%

60%

60%

Percentage mature MK

(CD42a+)

80%

fopMKs(50)

40%

40%

LT-fopMKs(65)

20%

Average iPSC#1 iPSC#3

20%

Average iPSC#1 iPSC#3

0%

0%

10 20 30 40 50

0 25 50 75 100

0 25 50 75 100

Cell size (um)

Forward programming day

i

GATA1

FLI1

TAL1

180

0 10 20 30 40 50 60 70

160

140

Gene expression

(rel. HMBS)

120

100

80

60

40

20

10 20 30 40 50 60 70

Forward programming day

Endogene + transgene Transgene

Endogene

CB-MKs

10 20 30 40 50 60 70

Figure 3 | Long-term expansion of fopMKs. (a) Schematic representation of the culture conditions for long-term MK-FOP. The combination of TPO (20 ng ml 1) and SCF (50 ng ml 1) allowed mature MK expansion for 90 days and beyond. (b) The cumulative mature MK (CD41a/CD42a double positive cells) fold increase relative to day 0 hiPSC cell input is shown for the hiPSC lines #1 (n 4) and #3 (n 5) over 90 days in culture (means.e.m.).

(c,d) The corresponding percentages of CD41a and CD42a cells monitored by ow cytometry from whole cultures are shown over the 90 day period

(means.e.m.). (e) Representative ow cytometry dot plots for CD41a and CD42a expression from the hiPSC#3 line. (f,g) Representative phase contrast picture of a day 70 hiPSC#3 fopMK long-term culture and associated Romanowsky staining of xed cells (scale bars 50 mm). White arrowheads: clumps of actively growing small cells; red arrowheads: single big cells in suspension culture and polyploid cells identied by Romanowsky staining. (h) Cell size distribution from fopMK and LT-fopMK cultures is shown as box plots (n 50, 65 respectively). (i) Endogenous and transgenic expression levels of the 3-

TFs were independently monitored by RT-qPCR throughout MK-FOP long-term cultures (means.e.m. from hiPSC lines #1 and #3; n 3). The average

range of expression levels in cbMKs (means.e.m.; n 5) is shown as a benchmark (in red).

and 132 days, culminating respectively in 17 million and 800,000 MK fold increases (n 4/5 biological replicate respectively;

Supplementary Fig. 4g). Beyond day 90, most LT-fopMK cultures showed a drastic loss of cell viability (Supplementary Fig. 4h) associated with a steady decrease of clonogenic potential and of CD34 haematopoietic progenitor content (Supplementary

Fig. 4i,j). At the population level, the average expression of the 3-TFs (sum of endogenous and transgenic expression) in LT-fopMK cultures was maintained in a close range to day 10 cbMKs (Fig. 3i). Together, we showed the 3-TF-driven MK-FOP generated a highly proliferative albeit exhaustible progenitor pool sustaining the expansion of mature MKs in vitro for over 3 months.

Genome-wide analysis conrms long-term MK identity. We analysed the whole-genome microarray expression data of MKs derived in vitro by different protocols and from different stem cell sources: by FOP or directed differentiation from hiPSCs (fopMKs, LT-fopMKs and ddMKs) and from cbMKs, all samples with purity greater than 95% CD41a and 80%

CD42a cells.

First, enrichment analysis using hyperG test (false discovery rate (FDR) r5%) of gene ontology terms from upregulated genes versus undifferentiated hiPSC conrmed that MK-FOP efciently induced the MK phenotype with top biological processes in fopMKs and LT-fopMKs related to haemostasis/platelet gene ontology terms as was the case for ddMKs and cbMKs (Fig. 4a).

6 NATURE COMMUNICATIONS | 7:11208 | DOI: 10.1038/ncomms11208 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11208 ARTICLE

a

fopMK

LT-fopMK

cbMK

e

ddMK Regulation bodyfluid levels

Response to wounding

Response to biotic stimulus

Platelet activation

Intracellular signal transduction

Cell morphogenesis

Neuron differentiation

Neurogenesis

Reg. Cell. Component organization

Cell development

Regulation body fluid levels

Response to wounding Response to wounding Response to wounding

Platelet activation Platelet activation Platelet activation

Platelet degranulation

Vesicle-mediated transport

Regulation body fluid levels

Regulation body fluid levels

Vesicle-mediated transport

Neuron differentiation

Cell. component biogenesis

Cell projection organization

Neurogenesis

Neuron proj. morphogenesis

Platelet degranulation

Regulation response to stress

Multicellular org. process

Neurogenesis

Cell part morphogenesis

Neuron projection morphogenesis

Neuron differentiation

Positive DE Negative DE

0.15

0.10

+ve reg. response to stimulus

Cell. component biogenesis

Cell. component organization

Epithelial tube formation

Morphogenesis of embryonic epithelium

ncRNA metabolic process

log10P

20 15 10 50

0

0.05

log10P

6

4

2

0

LT-fopMK#1_d51

LT-fopMK#3_d51

LT-fopMK#3_d40

LT-fopMK#3_d40

LT-fopMK#1_d51

LT-fopMK#1_d69

LT-fopMK#3_d51

LT-fopMK#1_d51

LT-fopMK#1_d69

LT-fopMK#1_d30

LT-fopMK#3_d30

LT-fopMK#1_d30

LT-fopMK#3_d30

LT-fopMK#1_d30

iPSC#1

iPSC#1

cbMK#1

cbMK#2

cbMK#3

cbMK#4

fopMK#1_d20

fopMK#1_d20

fopMK#1_d20

fopMK#2_d20

fopMK#1_d20

fopMK#2_d20

ddMK#1_d25

ddMK#1_d25

ddMK#1_d25

ddMK#1_d25

GO term enrichment (vs iPSC)

iPSC

cbMK LT-fopMK

fopMK ddMK

b

Positive DE genes (vs. iPSC)

r =0.880 (DE=3,354) r =0.963 (DE=1,106) r =0.978 (DE=598) r =0.998 (DE=0)

fopMK#1 (n=4)

f

ddMK NES=1.19 FDR=16%

MK

fopMK NES=1.25 FDR=21%

LT-fopMK NES=1.28 FDR=33%

cbMK NES=1.24 FDR=14%

CBMK

FOPMK

LT-FOPMK

Blood

Blood

Blood

Blood

MK

MK

MK

DDMK

hPSC MKs

8

6

c

8

6

2

Gene rank (0 - 40,225)

4

PC3 (12%)

4

2

2

PC3 (12%)

2

14

14

14

14

0

12

12

12

12

2

PC2 (19%)

6

PC2 (19%)

10

10

10

10

0

8

8

8

6 10 5 0 5

8

5

0

6 4 2 0 2 4 6

PC1 (39%)

5 10

PC1 (39%)

8

10

12

14

8 10 12 14 8 10 12 14 8 10 12 14

hiPSC#1 (n=2) cbMK (n=4) ddMK#1 (n=4) fopMK#2 (n=2)

Expression value (Log2: 715)

Similarity

d

DE genes (vs CBMK)

Common (503)

DDMK

FOPMK

fopMK only (505)

ddMK only (277)

LT-fopMK only (824)

Type I interferon-mediated signaling pathway

Response to type I interferon

Defense response to virus

Cellular response to chemical stimulus

Cellular response to cytokine stimulus

87

277

186

505

Blood vessel development

Cardiovascular system development

Angiogenesis

Cell motility

Response to external stimulus

Response to stress

Intracellular protein kinase cascade

Macrophage differentiation

Regulation of catalytic activity

Cholesterol biosynthetic process

Response to stress

Intracellular protein kinase cascade

Macrophage differentiation

Regulation of catalytic activity

Cholesterol biosynthetic process

503

824

log10P

8 6 4 2 0

log10P

3 2 1 0

1.51.00.5 0

1.51.00.5 0

LT-FOPMK

285

Figure 4 | Transcriptional landscape of forward programmed MKs. The undifferentiated hiPSC#1 line (n 2), day 20 fopMKs (hiPSC#12, n 4,2), day

25 ddMKs (hiPSC#1, n 4), day 10 cbMKs (n 4; all three groups CD42b sorted 495%) and day 3069 LT-fopMKs (hiPSC#1 and #3, n 8 and 6;

480% CD42b ), were analysed for gene expression using Illumina Human HT-12 v4 BeadArrays. (a) Top ve enriched gene ontology biological

processes for differentially expressed (DE) genes in all MK samples compared to hiPSCs. Log10 P values are shown as colour scale. (b) Gene set

enrichment analyses for DE genes from the different MK samples (versus hiPSCs; grey circles) against a MK versus other blood types gene expression data set (from Haematlas23). NES, normalized enrichment score; FDR, false discovery rate. (c) Gene expression correlation scatter plots on the whole-gene set using pairwise comparisons of different MK groups. R2 Pearson correlation value and differentially expressed gene numbers are indicated. The differential expression threshold (two-fold-change; FDR 5%) is shown as dotted red lines. (d) Venn diagrams recording DE genes (|Log2 fold-change| 41;

FDR5%) in hiPSC-derived MKs compared with cbMKs. The number of DE genes is indicated for each intersection and the top ve enriched gene ontology term biological processes from the Venn Diagram intersections are shown. (e) Hierarchical clustering using the average agglomerative method on whole-gene data set. (f) Three-dimensional plots of the principal component analysis of MK populations. The rst 3 PC are shown with respective percentages of variance indicated in brackets. The two principal component analysis boxes are snapshots of a rotation along the PC3 axis with MK sample groups highlighted.

This was mirrored by a depletion in differentiation/morphogenesis gene ontology term-associated genes indicating appropriate down-regulation of pluripotency features (Fig. 4a). Moreover, the gene set enrichment analysis29 against a panel of blood cells (cbMKs versus other blood cells from Haematlas23; n 4 and 46, respectively)

showed specic enrichment of MK-specic genes in the list of

genes upregulated in fopMKs further demonstrating the acquisition of a genuine MK phenotype (normalized enrichment score (NES) 1.25, FDR 21%; Fig. 4b). The highly similar

expression proles obtained from two distinct hiPSC lines demonstrates the inter-line qualitative reproducibility of MKFOP (r2 0.998 for iPSC#1 and #2; Fig. 4c).

NATURE COMMUNICATIONS | 7:11208 | DOI: 10.1038/ncomms11208 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11208

Differential expression analysis comparing hiPSC-derived MKs to cbMKs revealed a number of differential expression genes that were (in accordance with the MK phenotype acquired by fopMKs) not related to haemostasis/platelet-related biological processes (503 common, 2,667 total; Z2-fold-change and

FDRr5% cutoff; Fig. 4d). Hierarchical clustering distinguished the four MK groups while showing a separate cluster encompassing the hiPSC-derived MKs (i.e., fopMKs, LT-fopMKs and ddMKs) distinct from cbMKs indicative of an intrinsic difference resulting from their hPSC provenance, as previously described30 (Fig. 4e). Interestingly, while conrming a shared hiPSC-MK identity distinct from cbMKs, the principal component analysis indicated that LT-fopMKs clustered closer to cbMKs than

fopMKs and ddMKs (Fig. 4f)30. Altogether, the whole-genome expression analysis validated the acquisition by MK-FOP of a genuine MK phenotype that was effectively maintained throughout long-term culture.

In vitro production of functional platelets by fopMKs. Mature MKs produce platelets by a process of proplatelet formation whereby MKs extend cytoplasmic protrusions into the bone marrow blood stream31. These protrusions contain multiple branching points and bulbous ends with active accumulation of granules into the end processes that represent the nascent platelets, which then mature further in the circulation. In culture,

b

c

a SELP TUBA DAPI

fopMKBlood

MVB

AG

d

e

cbMK fopMK

LT-fopMK

Forward programming day

**

Blood fopMK cbMK

6

CD41a

Average platelets /MK

4

cbMK

SS

2

0 cbMK

20 28 50 90

h

Donor platelets

fopMK platelets

Exp. trend

Exp. trend

FS

CD42a

1.4

25

g

Platelet half-life

(hours)

20

1.2

f

Immature platelet fraction

(IPF, percentages)

16 14 12 108

Human platelet count

(relative to 30min)

15 19.7

1

10

5

0.8

7.1

Mean platelet volume

(MPV, femtolitres)

12 10

8 6 4 2 0

0.6

0 Donor (n=5)

fopMK

(n=4)

9.2 8.6

Donor fopMK

0.4

0.2

BOB (7.9fL)

FFDK (9.2fL)

Blood (7.6fL)

64 20

4.0%

11.6%

0 0 500 1000 1500

40fL fopMK

Donor

Time post-transfusion (minutes)

Figure 5 | Platelet release in vitro from fopMKs throughout long-term culture. (a) Phase contrast picture of spontaneous proplatelet-forming fopMKs (arrowheads) in suspension culture (hiPSC#3, day26; scale bar, 50 mm). (b) Proplatelet-forming fopMK on brinogen-coated slide immunostained for alpha-tubulin (TUBA) and P-selectin (SELP). Arrowheads indicate nascent platelet tips containing SELP-positive granules (hiPSC#4, day21; scale bar,50 mm). (c) Transmission electron microscopy pictures of blood and in vitro-produced platelets showing typical ultrastructure. AG, alpha-granules; MVB, multi-vesicular bodies (scale bars, 1 mm). (d) Representative ow cytometry dot plots of platelet analysis. Platelets are dened within the human platelet size gate as CD41a/CD42a double positive events. (e) cbMK and fopMK from different culture time points were sown on C3H10T1/2 feeder cells for 48 h and the number of platelets released in the supernatant quantied by ow cytometry. Data represent the means.e.m. of platelet number per MK sown (n 7, 4, 10, 6, 2 for cbMK and fopMK (hiPSC#14 pool) at day 2090 respectively; **Po0.01 versus fopMK day 20 by two-tail t-test). (f,g) Mean platelet

volume (MPV) s.d. and immature platelet fraction (IPF) s.d. of washed donor and fopMK platelets as measured on a Sysmex whole-blood analyzer (n 4/2 respectively; hiPSC#1 and #5). (h) Human platelet survival in NSG mice circulation over 24 h measured by ow cytometry following the systemic

injection of 20 million washed platelets. The exponential trend of human platelet absolute count decrease over time was used for half-life calculation (means.e.m.; n 5/4 for donor and fopMK (hiPSC#1 and #5) platelets respectively).

8 NATURE COMMUNICATIONS | 7:11208 | DOI: 10.1038/ncomms11208 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11208 ARTICLE

fopMKs formed proplatelets containing P-selectin-positive a-granules (Fig. 5a,b; Supplementary Movies 1 and 2). For further analyses, in vitro platelet production was maximized using a static co-culture system with the murine C3H10T1/2 feeder cell line as previously described27. Electron microscopy showed that platelets produced in vitro from fopMKs and cbMKs, while heterogeneous in quality reecting the current limitations of 2D static cultures, showed the typical platelet ultrastructure notably including high alpha-granule content (Fig. 5c). We further used ow cytometry for the quantitative measurement of in vitro platelet production from fopMKs, strictly dening platelets as CD41a /CD42a particles of human platelet size (Fig. 5d). A

signicant increase in platelet production from LT-fopMKs versus day 20 fopMKs was observed matching platelet release from cbMKs (50.2 versus 0.80.2 platelets per MK at day 90 and day 20, respectively; Fig. 5e). The platelet production rate was similar amongst fopMKs derived from different hiPSC lines (hiPSC#14; Supplementary Fig. 5a). In vitro platelets showed surface expression of the main thrombocyte receptors including the GPIIb/IIIa complex (brinogen receptor), GPIb and GPIX (Von Willebrand factor receptor), GPIIa and GPVI (collagen receptors), some of them with a decreased intensity compared with donor platelets which has been previously described and likely originated from the static 37 C in vitro culture conditions used for production (Supplementary Fig. 5b)20,32. The fopMK platelets showed a normal mean volume on a clinical Sysmex blood analyzer (8.60.7 , n 2, iPSC#1 and #5; Fig. 5f),

and interestingly an increased immature platelet fraction compared with normal circulating blood (11.63% versus 41%, respectively; Fig. 5g). Eventually, we assessed fopMK platelet survival in vivo in immunodecient NOD scid gamma (NSG) mice with further splenic macrophage depletion to allow human platelet maintenance in the circulation33. The fopMK platelets were readily detected in the circulation for several hours while showing a shorter half-life than primary donor platelets(7.10.8 h versus 19.72.2 h, respectively; Fig. 5h), a result which was very similar to previously published data (7.5 and18.3 h, respectively)19 and probably biased by the limitations of the current static in vitro production systems for the generation of homogenous populations of genuine resting platelets endowed with longer circulation half-life. To further distinguish functional platelets from the heterogeneous in vitro-produced pool, we used Calcein-AM as a marker of platelet viability and membrane integrity34. The proportion of Calcein-AM-positive platelets within fopMK- and cbMK-derived platelet harvest was32.40.8% (n 3; iPSC#1, #5) and 40.13.6%, respectively

(n 3) and dened a phenotypically more homogenous

CD41a /CD42a platelet population similar to control blood

(Supplementary g. 5c). Hereafter, in vitro generated platelets were identied using Calcein-AM staining to compare their function with donor-derived platelets.

To full their haemostatic role, platelets must be able to sequentially adhere to damaged vessels (using collagen and other extra-cellular matrix receptors), increase their surface by spreading (by active remodelling of their cytoskeleton), build-up the thrombus by aggregation of other platelets (through brinogen binding) and eventually amplify the haemostatic response through degranulation (granule content release to the surface)35. We rst compared adhesion with brinogen using a quantitative ow cytometry assay36. Similarly to platelets generated in vitro from cbMKs, an increased adhesion to brinogen was observed for fopMK-derived platelets as compared with blood platelets (Fig. 6a). This distinction may be inherent to the platelet harvest methodology or originate from intrinsic developmental biological differences of hiPSC and neonates platelets compared with adult blood platelets37. We further showed that fopMK-derived platelets spread efciently

upon contact with brinogen presenting typical tubulin cytoskeletal reorganization (Fig. 6b). We then used a ow cytometry approach to quantitatively measure agonist-induced platelet aggregation38 (Fig. 6c). We showed that in response to agonist stimulation (a combination of ADP and TRAP), fopMK-derived platelets efciently aggregated with fresh blood platelets as well as between themselves with no signicant difference in aggregation potential as compared with blood platelets (Fig. 6d,e). Eventually, we performed thrombus formation in vitro under physiological shear stress as a high-level assay of integrated platelet functions. To mimic a transfusion setting, we introduced a dened amount of Calcein-AM-labelled in vitro-produced platelets or blood platelets from a transfusion unit into normal or thrombocytopenic fresh human blood (platelet count o50 109 per litre; spiked platelets at

10 109 per litre; Supplementary Movies 3 and 4). When exposed

to collagen under arterial shear rates (1,600 s 1), fopMK-derived platelets took part in clot formation similarly to platelets from a concentrate unit (Fig. 6f,g). We further observed an increase in the contribution of fopMK platelets to the thrombi in the context of thrombocytopenic blood (Fig. 6fg; Supplementary Fig. 5d). Furthermore, P-selectin surface expression was observed in the platelets participating in the thrombi, thereby proving adequate degranulation of activated fopMK platelets (Supplementary Fig. 5e). The validation of fopMK platelet function was ultimately addressed in vivo using intravital microscopy and laser injury-induced thrombus formation in the cremaster vasculature of NSG mice39. The transfusion of 5E 7 Calcein-AM-labelled platelets into mouse

circulation allowed the visualization of human platelet integration into thrombi with comparable efciencies between donor and fopMK platelets (1.50.2 and 1.40.2 platelets per 100 mm2 thrombus area respectively, n 16/12; Fig. 6h,i ). Strikingly,

in vitro-produced fopMK platelets showed active participation in clot formation with visible sequential rolling, binding, thrombus surface surveying and spreading (Supplementary Movies 5 and 6).

While current systems for in vitro platelet production from cultured MKs can be further optimized beyond the 2D static approach used here, platelets produced in vitro from fopMKs were nevertheless functionally similar to donor-derived platelets across a range of assays, thereby afrming their suitability and promise for a host of regenerative medicine research and therapeutic applications.

DiscussionWe describe here forward programming, a novel approach for the chemically dened large-scale production of MKs from hiPSCs. We drive MK development by the combined ectopic expression of three TFs: GATA1, FLI1 and TAL1. These three TFs have well-documented roles in haematopoiesis, especially in the maintenance of early haematopoietic progenitors, red blood cell and megakaryocyte differentiation4044. Forced expression of GATA1 or TAL1 alone in haematopoietic progenitors has been shown to bias differentiation towards MK and erythroid fates45,46, while FLI1 cooperates with GATA1 to enable MK maturation47. In addition, TAL1 and FLI1 play an earlier role in the specication of the haematopoietic programme during vertebrate embryonic development48,49. Accordingly, exogenous expression of TAL1 in hESCs has been reported to promote haematopoiesis and megakaryopoiesis50,51. More recently, the combinatorial expression of TAL1 with GATA2 was found to induce an hemogenic endothelial phenotype biased towards erythromegakaryocytic differentiation from hPSCs52. Nevertheless, the MK-FOP approach described here involving co-expression of the 3-TFs GATA1, FLI1 and TAL1 is unprecedented in regard to its efciency in rapidly imposing MK progenitor identity to hPSCs in very stringent MK-specic culture conditions. The high

NATURE COMMUNICATIONS | 7:11208 | DOI: 10.1038/ncomms11208 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 9

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11208

hierarchical status of the 3-TFs within the MK gene regulatory network, already supported by previously published gene interaction network and ChIP-sequencing data in primary human MKs26,53 (Supplementary Fig. 2c), is now functionally conrmed in human cells by the efciency of the MK-FOP approach.

It remains to be demonstrated how closely the 3-TF programming recapitulates normal haematopoietic development from hPSCs. Current data indicate that mesoderm commitment is strongly benecial to MK-FOP, consistent with the normal ontogeny of blood cells in the embryo. Intriguingly, we observed an early expression of hemogenic endothelium markers (FLK1,

Adhesion Spreading

fopMK

F-actin Tubulin DAPI

a b

100 BSA Fibrinogen

**

Blood

% Adhesion (ADP+TRAP)

80

**

60

40

20

0 Blood fopMK cbMK

Aggregation

c d e

No agonist

ADP+TRAP

fopMK plt

CD31-APC

103

102

101 2 0

103

102

101

2 0 2

10

No agonist

ADP+TRAP

7

P=0.99

6

8

5

6

0.5

00.5

% Aggregation

100

101

102

0.5

Delta-aggregation

2

103

102

101

2 0 2

103

00.5

100

101

102

103

4

4

3

P=0.20

2

Blood plt

CD31-APC

103

102

101

2 0 2

2

1

0 Blood x Blood

fopMK x Blood

fopMK x fopMK

0 Blood x Blood

fopMK x Blood

fopMK x fopMK

0.5

00.5

100

101

102

103

0.5

00.5

100

101

102

103

Blood plt CD31-V450

In vitro thrombus

f g

Spiked calceinAM platelets (1E+7/ml)

fopMK#1 Blood

fopMK#5

bloodNormal blood

1.5

Normal blood Thrombocytopenic blood

Platelets (CAM+) per 100m2 thrombus

P=0.04

P=0.08

1.0

P=0.05

Thrombocytopenic

0.5

P=0.03

0.0

Blood fopMK#1 fopMK#5

In vivo thrombus

h i

fopMK#5

Blood

2.0

1.5

1.0

0.5

0

00:00:25.439

00:04:11.042

Human platelets (CAM+)

per 100m2 thrombus

P=0.30

P=0.76

10 um 10 um

Blood fopMK#1 fopMK#5

10 NATURE COMMUNICATIONS | 7:11208 | DOI: 10.1038/ncomms11208 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11208 ARTICLE

Recently, another approach has been published that allows the expansion of MK progenitors derived from PSCs55. In contrast with MK-FOP, that system is based on the articially controlled downregulation of GATA1 in haematopoietic progenitors allowing MK progenitor expansion and its restoration to enable subsequent MK maturation. The approach, which was developed with mouse ESCs, has yet to be demonstrated to work in human cells. In addition, the authors did not show efcient platelet production from these MKs in vitro, instead choosing an adoptive transfer of MKs into recipient mice. The injection of nucleated cells derived from human pluripotent stem cells would raise crucial issues of potential tumorigenicity that could constrain future clinical use.

The quantitative functional platelet data collected here proved that platelets derived from fopMKs in vitro are endowed with major platelet functions allowing efcient thrombus formation as shown in previous studies19,20. Although abnormalities in key TFs for the MK lineage can lead to abnormal platelet function43, the preservation of fopMK platelet function is consistent with the fact that the overall expression levels of the 3-TFs is controlled throughout the programming progression and similar to cbMKs.The remaining bottleneck for application to transfusion medicine is the optimization of platelet generation in vitro: this presently remains on average 1,000-fold lower than the in vivo platelet yield per MK and is accompanied by issues regarding the purity of the nal product where functional platelets represent only a fraction of the platelet size particles in the whole harvest19,20,56. In this study, which focuses on the production of the MKs themselves, we have used a previously published platelet production system based on co-culture with a mouse stromal cell line27,57. Our ndings clearly confer an impetus to achieve further progress using newly developed three-dimensional laminar ow systems and bone marrow mimicking scaffolds5860 by providing ample quantities of functional MKs from hPSCs using a simplied chemically dened protocol amenable to the generation of a clinical product.

In conclusion, our study demonstrates the feasibility of a forward programming approach to generate mature functional MKs from human PSCs that signicantly transcends available directed differentiation protocols through a unique combination of key achievements. First, the methodology results in a very high cell yield and MK purity using fully chemically dened xeno-free culture conditions. To put this in the clinical context, the long-term culture expansion allows a cumulative production of 2 1011 MKs releasing 1 1012 plateletsthe equivalent of

3 transfusion unitsstarting from only one million hiPSCs.

Figure 6 | Functional assessment of fopMK in vitro platelets. (a) Adhesion to brinogen of fopMK and cbMK in vitro platelets upon combined TRAP ADP

stimulation compared to blood platelets using a ow cytometry bead based assay. Percentages of adhesion on BSA or brinogen-coated beads are shown (means.e.m., n 4, 4, 2 for blood, fopMK and cbMK respectively; **Po0.01 versus blood by two-tail t-test). (b) Representative pictures from in vitro

spreading assay. Washed platelets were sown on brinogen-coated slides, incubated for 45 min at 37 C and immunostained for alpha-tubulin (TUBA) and F-actin (scale bars, 10 mm). (c) Aggregation of fopMK platelets upon agonists stimulation was tested both with and compared with blood platelets using a ow cytometry-based assay: representative dot plots are shown. (d) Percentages of aggregation from Calcein-AM live platelets upon stimulation are shown for

blood and fopMK platelets reactions (2 107 platelets per ml; means.e.m., n 7 for each reaction group; no statistical difference versus blood at Po0.01 by

two-tail t-test). (e) Associated delta-aggregation dened as ((% ADP TRAP aggregation)(% no agonist aggregation)) for the different reaction groups

(mean s.e.m., n 7; P values by two-tail t-test versus blood indicated). (f) Thrombus formation in vitro under arterial shear stress. The participation of

Calcein-AM live platelets spiked into human blood (at 1 107 per ml) is shown per 100 mm2 thrombus area. Normal or thrombocytopenic blood (4150 109

and o50 109 l 1 respectively) was used as recipient. Spiked platelets were sourced from day-8 concentrate unit (blood) or from fopMK platelets

(iPSC#1 and #5; n430 analysed thrombi per group; P values by two-tail t-test versus blood indicated). (g) Representative pictures from in vitro thrombus formation assays. Thrombi identied using bright eld images are delineated and Calcein-AM platelets uorescing in green; in vitro platelets Calcein-AM labelling is intrinsically dimmer than donor-derived platelets (Supplementary Fig. 5b). Scale bar, 50 mm. (h) Thrombus formation in vivo by laser injury of an arteriole in the cremaster muscle of NSG mice and intravital confocal microscopy. The incorporation of human Calcein-AM-labelled platelets (50 million transfused per mouse) to mouse thrombi is shown per 100 mm2 thrombus area at T_max (thrombus maximum size). Mean valuess.e.m. and P values by two-tail t-test versus donor platelets are shown (n 16/4/8 thrombi analysed for blood, fopMK#1 and #5 platelets respectively). (i) Representative snapshots

of Calcein-AM human platelets incorporated to mouse thrombus (scale bar 10 mm).

CD34 and VE-Cadherin) from day 2 of programming which was intermingled with expression of the megakaryocyte commitment marker CD41a and quickly followed by a marker of blood commitment with the increasing concurrent CD43 expression from day 3 onwards (Supplementary Fig. 6). The progressive decrease of endothelial populations associated with a steady TPO-dependent growth of CD41a cells from day 4, as well as the

absence of monocyte/granulocyte colonies in clonogenic assays, indicates an early commitment to MK fate (Supplementary Fig. 6). Taken together, these observations suggest that over-expression of the 3-TFs in hPSCs may at least partially recapitulate key steps of the blood ontogeny with a contracted hemogenic endothelium phase and early enforcement of MK identity. Controllable and traceable 3-TF expression systems combined with molecular analysis at the single-cell level will be necessary to further decipher the precise molecular mechanisms governing MK-FOP.

There is high biotechnological signicance in our ability to maintain the forward programmed cells in long-term cultures and expand them over time whilst preserving MK purity, markers of maturity and platelet production. This TPO- and SCF-dependent cellular proliferation persisted for an extended but not indenite timeframe (90134 days) and is presumably sustained by non-transformed early MK-biased progenitors (o1% by CFU assays) able to proliferate and differentiate further towards mature MKs. In this respect the LT-fopMK culture appears different in nature to the recently described expandable MK cell lines obtained from hPSCs by sequential ectopic introduction of MYC, BMI1 and BCL-XL into MK-committed cells20. The generation of these MK cell lines notably requires the precise control of MYC expression levels and subsequent silencing of those three overexpressed genes to achieve full MK maturation. The complex genetic modications involved and the limited success rate of immortalized MK line derivation involving manual screening and clonal selectionalong with the initial generation of blood progenitors in undened conditions are limitations of that protocol for both research and clinical development. MK-FOP did not produce an immortalized MK progenitor and did not require manipulation of gene expression levels to produce fully mature MKs. Interestingly, MK lineage biased self-renewing progenitors have been recently identied in the haematopoietic stem cell phenotypic compartment of the bone marrow54. Further investigations are needed to understand the full nature and the 3-TF-driven molecular mechanisms underpinning the generation of this long-term expanding MK progenitor.

NATURE COMMUNICATIONS | 7:11208 | DOI: 10.1038/ncomms11208 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 11

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11208

Moreover, the minimal cell handling and cytokine requirements, the cryopreservation of fopMKs enabling cell banking and future stock management and the successful differentiation of an array of hPSC lines with qualitative reproducibility are additional critical strengths of the MK-FOP approach for future manufacturing of platelets for human therapeutic applications. The effectiveness and versatility of MK-FOP opens new avenues for future basic research and functional studies on novel MK and platelet genes as well as disease modelling using hiPSCs53,61,62.

Methods

Human pluripotent stem cell culture. The H9 hESC line (WiCell; passages 7595) and hiPSC lines (iPSC#15; A1ATD1, BBHX8, A1ATD1-c, S4-SF5 and FFDK1, respectively, p3050) were cultivated as clumps in a chemically dened medium (CDM) containing recombinant human FGF2 and Activin-A (15 ng ml 1 each, internal) on feeder-free gelatin or vitronectin coated wells as previously described63. All hiPSC lines were obtained from the Cambridge Biomedical Research Centre iPSC Core Facility and have been derived from adult dermal broblasts using integrative murine retroviral vectors (iPSC#12), cytoplasmic Sendai viral vectors (iPSC#34) or episomal vectors (iPSC#5) expressing the human OCT4, SOX2, KLF4 and MYC reprogramming factors.

Selection of transcription factor candidates. We performed a differential gene expression analysis focused on DNA binding protein coding genes (PANTHER Classication System) from whole-genome expression data generated using the H9 hESC line (internal data, Illumina HumanWG-6 v3) and human cord blood-derived MKs23. The list of 116 MK-specic genes generated was further rened by removal of 21 histone-coding genes and addition of 6 candidates based on previous knowledge of their role in megakaryopoiesis (Supplementary Fig. 1a). Using the VisANT web-based software24, the resulting 101 candidate geneswere subsequently ranked based on number of (1) internal protein interactions,(2) interaction with epigenetic modiers (HAC, HDAC, DNMT, list in Supplementary Fig. 1c) and (3) differential expression levels (Supplementary Fig. 1b). Genes with low differential expression (Log2(MK-ESC)o1 or o2 with no reported interactions) were excluded from the candidate list.

Recombinant lentiviral vectors. Transcription factor cloning. The human coding sequences of the nine candidate genes (variants 1 from NCBI Reference Sequence Database) including the 50 Kozak consensus sequence were generated by PCR using cbMK cDNA, individually cloned into the pWPT lentiviral vector backbone(Dr Trono, Addgene #12255) downstream of the human EF1-alpha ubiquitous promoter and checked for sequence integrity.

Viral particle production. Replication decient lentiviral vector particles (LVPs) were produced by transient co-transfection of HEK 293T/17 cells (ATCC CRL-11268) with pWPT constructs along with the psPAX2 and pMD2.G helper plasmids (Addgene #12260, #12259) using TransIT-LT1 transfection reagent (MirusBIO). Crude supernatants containing LVPs were concentrated by PEG-based precipitation (LentiX-concentrator, Clontech) and functional titres determined by qPCR measurement of provirus copy number in genomicDNA of transduced HCT116 cells (ATCC CCL-247).

Human pluripotent stem cell transduction. hPSC lines were routinely transduced by 1824 h single exposure to LVPs using multiplicity of infection of 20 in presence of 10 mg ml 1 Protamine Sulfate (Sigma) in routine culture medium.

Megakaryocyte forward programming. Optimized embryoid body based protocol.

On transduction day (day 0), sub-conuent (5080%) hPSC cultures were dissociated to single cells using TrypLE (Life Technologies) Fig. 2a. Embryoid body formation was initiated with 612E 5 viable cells per well of an Aggrewell400

plates (Stem Cell Technologies) leading to 5001,000 cells per embryoid body following spin aggregation. Lentiviral transduction was performed concomitantly to the aggregation step in CDM supplemented with Y-27632 (10 mM, Sigma),

BMP4 (10 ng ml 1, R&D) and protamine sulfate. After 24 h, transduced embryoid bodies were collected and sown in ultralow adherent cell culture plates (Corning)

at 1,200 embryoid bodies per 10 cm2 dish in CDM with BMP4 and FGF2(5 ng ml 1). Twenty-four hours later, embryoid bodies were washed and sown in ultralow adherent plates at 600 embryoid bodies per 10 cm2 in CellGroSCGM medium (CellGenix) supplemented with TPO (100 ng ml 1, Cellgenix) and SCF (25 ng ml 1, Gibco). At day 10, embryoid bodies were dissociated to single cells using Collagenase-IV and Dispase-II (1 mg ml 1, Gibco) followed by enzyme free cell dissociation buffer (Gibco) treatment. Single cells were cultivated at 2E 5 per

ml on tissue culture plates (Corning) for an additional 10 days in CellgroSCGM with TPO (100 ng ml 1) and IL1-b (10 ng ml 1, Miltenyi Biotec). Half of culture media was renewed every 3 days.

Long-term MK-FOP cultures. From day 10 after dissociation to single cells, the cultures were maintained in CellGroSCGM with low TPO concentration(20 ng ml 1) and SCF (50 ng ml 1). Culture medium was half-renewed every 3

days and cells split every 710 days when reaching concentration of 11.5E 6

cells per ml.FOPMK freezing. Cells from day 3070 cultures were collected and frozen at0.52E 6 cells per ml in IMDM 20% fetal bovine serum (Gibco) and 5% DMSO

(Sigma).

Adherent cell protocol. Small cell clumps were generated from sub-conuent hPSC cultures using a Collagenase-IV and Dispase-II mix and sown on human bronectin coated (50 mg ml 1, Millipore) wells in CDM-containing FGF2 and

Activin-A (15 ng ml 1 each) at an approximate density of 25E 5 cells per

10 cm2. Cells were transduced with LVPs the next day and kept for 2 days in FGF2 Activin-A (pluripotency) or FGF2 LY-294002 BMP4 (ref. 25)

(mesoderm) depending on experiment settings. From day 2, cells were maintained in CellgroSCGM with TPO and SCF as described above.

MK-directed differentiation. hPSC lines were differentiated as described27 using batch tested serum and stromal cells from Prof Koji Eto Laboratory.

Cord blood-derived megakaryocytes. Cord blood was obtained after informed consent under a protocol approved by the Cambridgeshire 4 Research Ethics Committee (07/MRE05/44). CD34-positive cells (Z98%) isolated by magnetic cell sorting (Myltenyi Biotec) were seeded at 1E 5 cells per ml in CellgroSCGM

containing TPO (100 ng ml 1) and IL1-b (10 ng ml 1) and cultivated for 10 days. We routinely obtained 7090% CD41a and 2060% CD42a cells by the end of

the culture.

Flow cytometry analysis. Flow cytometry experiments were performed on a CyAn ADP (Beckman Coulter). Single-cell suspensions were generated using a Collagenase-IV/Dispase-II mix and/or enzyme free dissociation buffer. Cells were stained for 20 min at room temperature (RT) in PBS 0.5%BSA 2 mM EDTA using combinations of FITC, PE and APC-conjugated antibodies (Supplementary Table 1). Background uorescence was set against matched isotype control antibodies and compensation matrix dened using single-colour-stained cells. Flow count uorospheres (Beckman Coulter) and DAPI (1 mg ml 1) were used to determine viable cell count in samples.

Cell morphology and phenotype analysis. Cell morphology analysis. Cells were spun onto a glass slide using cytofunnels at 400g for 5 min, methanol xed and stained using the Romanowsky method.

Megakaryocyte colony forming assay. Around 5,000 cells per chamber were used in MethoCult methylcellulose assays (#4230, StemCell Technologies) containing screened fetal bovine serum and supplemented with TPO and SCF (100 and50 ng ml 1 respectively); MegaCult collagen cultures (StemCell Technologies)

were dehydrated following manufacturers instructions for colony immunostaining.

Immunouorescence analysis. Megakaryocytes were cultivated on human brinogen (50 mg ml 1, Millipore) coated glass cover slips for 48 h to foster adhesion and proplatelet formation. Cells were xed in 2% formaldehyde, permeabilized with 0.1% Saponin/0.2% Gelatin and incubated 2 h at RT with selected primary antibodies (Supplementary Table 1) then with uorochrome conjugated secondary antibodies for 45 min atRT. Cell nuclei were stained with DAPI. Images were acquired on a uorescent microscope Axiovert 40 (Zeiss) or a SP5 confocal microscope for granule imaging (Leica).

Ploidy analysis. Cells were xed using 4% formaldehyde for 10 min at RT, immunostained for CD41a and CD42a expression and subsequently incubated in PBS 0.1% Tween with DAPI at 1 mg ml 1 for 15 min at RT before ow cytometry analysis.

Transmission electron microscopy. Megakaryocytes were xed in 2% glutaraldehyde/formaldehyde followed by post xation, resin embedding and staining as previously described17,64, and eventually analysed on a FEI Tecnai G2 microscope.

Gene expression analysis by RT-qPCR. Total RNA was extracted using RNeasy kits (Qiagen) according to the manufacturers instructions including DNase treatment. cDNA was prepared from 250500 ng total RNA using Maxima First Strand cDNA Synthesis Kit (random hexamers and oligo(dT)18 mix for priming;

Fermentas). Two-step qPCR reactions were performed in duplicates using SYBR green chemistry on ABI 7500HT or Mx3000P instruments (Applied Biosystems; Agilent Technologies). Relative gene expression was calculated by the 2 Dct method using HMBS as housekeeping gene for normalization. qPCR primer pairs (Supplementary Table 2) designed to amplify only cDNA, to detect all known isoforms, and to have no reported off-target matches searching the human NCBI RefSeq database were tested within 80120% PCR efciencies with single dissociation curves. We used UTR targeting (absent from transgenes) to monitor endogene expression while transgene specic primer pairs used a common reverse primer specic to the viral vector RNA.

Whole-genome expression microarray analysis. DNA-free total RNA was extracted as above from sorted CD42b cell fractions using the EasySep system

(Stemcell Technologies; 495% purity, for cbMK, ddMK and fopMK samples) or

12 NATURE COMMUNICATIONS | 7:11208 | DOI: 10.1038/ncomms11208 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11208 ARTICLE

unsorted 480% CD42b cells (LT-fopMK samples) and 500 ng hybridized to

Illumina Human HT-12 v4 BeadArrays.

Microarray gene expression. Microarray signal intensities from Genome Studio version1.9 were variance stabilization transformed and robust spline normalized (RSN). For differential analysis, we removed low signal probes with detectionP value40.01 in all samples, leaving only informative probes. These procedures were carried out using R package Lumi. For each pairwise differential analysis, we applied surrogate variable analysis (SVA) to correct for un-modelled factors that may bring about batch effects. We then tested for differential expression using limma where statistical signicance was set to 2-fold change and 5% false discovery rate or Benjamini-Hochberg adjusted P value.

Hierarchical clustering. Hierarchical clustering was performed using R package pvclust. For dissimilarity or distance measure, we use 1-correlation and average as agglomerative method.

Gene ontology enrichment analysis. We performed enrichment analysis using GOstat 2.24 of over-represented gene ontology terms in the set of differentially expressed genes. We used standard hypergeometric test using only informative probes in the chip as the universe.

Gene set enrichment analysis. We used gsea22.0.13 with datasets for megakaryocytes (n 4) and other blood cells (n 46) from the Haematlas

study23 (E-TABM-633; Illumina Human-6v2 array).

Principal component analysis. Classication of samples in multiple dimensional factor spaces was applied by calling the function cmdscale.

Gene expression heatmaps. Heatmap builder v1.1 (Dr Ashley lab, Stanford) was used for dataset normalized representations.

In vitro platelet analysis. Co-culture on feeder cells. To promote platelet production, day 10 cbMKs and day 2090 fopMKs were further cultivated for 48 h in CellgroSCGM plus Heparin (25 U ml 1) without further cytokine addition at 1E 5 cells per cm2 on gamma-irradiated C3H10T1/2 feeder cells (Riken Institute;

1E 4 cells per cm2).

Platelet ow analysis. Crude supernatant containing the platelets was analysed by ow cytometry after addition of 1:9 volume of acid citrate dextrose (ACD, Sigma) and cell removal by centrifugation 150g at 10 min. Antibodies against human platelet receptors were added directly to the media (1:50 dilution, see Supplementary Table 1) for 30 min at RT and unwashed platelets analysed using uorospheres for quantication. Human platelets from fresh whole-blood diluted in PBS/ACD or from day 810 platelet concentrate units (NHSBT, UK) were analysed in parallel.

Washed platelet preparation. Human platelet-rich plasma and in vitro generated platelets collected as above were washed following the optimized protocolfrom Cazenave et al.65 to prevent activation and maximize platelet function preservation.

Platelet size analysis. Washed platelets were run on a Sysmex XE-2100 Automated

Hematology System.

Electron microscopy. Washed platelets were xed and stained as described for MKs.

Spreading assay. Washed platelets in Tyrode-HEPES buffer complemented with CaCl2 1 mM were sown on bronectin coated glass cover slips and incubated45 min at 37 C. They were subsequently xed using 2% formaldehyde, immunostained and analysed as described for MKs.

Platelet survival in vivo. Washed platelets were injected to macrophage depleted (by clodronate liposome injection at day 3) 812-week-old male NOD scid

gamma (NOD.Cg-Prkdc scid Il2rg tm1Wjl /SzJ; NSG) mice through the tail vein as single dose of 2E 7 CD41a/42a platelets. The human versus mouse platelet

content was monitored by ow cytometry from whole-blood samples at 1, 30, 120 min and 24 h after transfusion using antibodies specic for human and murine CD41a. The absolute human platelet count was determined using ow count uorosphere (Beckman Coulter) and platelet survival calculated by the multiple hits method relative to the 30 min equilibrium time point66,67. All mice were kept in specic pathogen-free conditions, and all procedures were performed according to the United Kingdom Home Ofce regulations and approved by the University of Cambridge Animal Welfare Ethical Review Body.

In vitro thrombus formation in laminar ow. Human blood or in vitro platelets were stained with Calcein-AM (100 nM; Life Technologies) for 10 min at 37 C before being washed and dened amount were mixed with 1 ml of human blood collected in 3.2% Citrate. Thrombocytopenic blood was articially prepared by depleting the platelets from plasma by performing 2 sequential centrifugation steps at 2,200g for 10 min and washing the red blood cell fraction with Tyrode-HEPES buffer before blood reconstitution. The procedure for clot formation under ow was modied from de Witt et al.68 Briey, glass slides were locally coated with Horm collagen spots (50 mg ml 1) and mounted into a ow chamber placed under a uorescent microscope (EVOS system, Advanced Microscopy Group). The blood was then perfused through the chamber at 1,600 s 1 (7.2 ml h 1) for 3 min allowing thrombi formation on collagen spots and imaging subsequently performed. Image analysis was carried on using ImageJ recording thrombus area and calcein-AM events per clot.

Thrombus formation intravital imaging. Washed platelets were labelled with Calcein-AM and injected to macrophage depleted 812-week-old male NSG mice through the carotid artery after cannulation as single dose of 5E 7 CD41a/42a/

Calcein-AM platelets. Thrombus formation was induced from 15 min

onwards after platelet transfusion by micropoint laser injury of the cremaster vessel walls as previously described39. Fluorescent and bright elds were simultaneously recorded by confocal microscopy and analysed using SlideBook 6 (Intelligent Imaging Innovations). The number of Calcein-AM human platelets

incorporated in mouse thrombi until maximal size was reached was recorded and reported to thrombus area. All animal experiments have been approved by the United Kingdom Home Ofce and the Birmingham University Animal Welfare Ethical Review Body.

Flow cytometry aggregation assay. Modied from original protocol38 to include the livedead particle discriminator Calcein-AM, it allows quantication of platelet aggregation using small numbers of platelets (2E 6 per reaction). Duplicate

platelet samples were stained with Calcein-AM and anti-CD31 (clone WM59) conjugated with APC or V450 in HEPES buffer supplemented with 20 mM PPACK dihydrochloride (Calbiochem), washed then mixed in the presence or absence of the agonists thrombin receptor-activating peptide (TRAP) and ADP (10 mM each).

Percentage aggregation was determined by ow cytometry for Calcein-AM-positive double CD31-positive events.

Bead platelet adhesion ow cytometry assay. Modied from original protocol36, it quanties adhesion of platelets to single 20mM polystyrene beads (Sigma) coated with either BSA or Fibrinogen. 3 105 platelets stained with 100 nM Calcein-AM

were mixed in basal culture media with either 20 ml BSA or 20 ml brinogen-coated beads, agonists (TRAP and ADP as above), incubated for 10/37 C and subsequently stained with anti-CD41a-APC antibody. DAPI negative, Calcein-AM and CD41a positive single beads bound by platelets are quantied by ow cytometry.

Statistics. Results are presented as meanstandard error of the mean (s.e.m.) with n representing the number of biological replicates unless otherwise stated. Statistical P values were calculated by two-tail Students t-test unless otherwise stated.

References

1. Kaushansky, K. Historical review: megakaryopoiesis and thrombopoiesis. Blood 111, 981986 (2008).

2. Michelson, A. D. Platelets 3rd Edn. (Academic Press, 2013).3. Machlus, K. R. & Italiano, Jr J. E. The incredible journey: from megakaryocyte development to platelet formation. J. Cell Biol. 201, 785796 (2013).

4. Junt, T. et al. Dynamic visualization of thrombopoiesis within bone marrow. Science 317, 17671770 (2007).

5. Estcourt, L. J. et al. Platelet transfusions in haematology patients: are we using them appropriately? Vox Sang. 103, 284293 (2012).

6. Williamson, L. M. & Devine, D. V. Challenges in the management of the blood supply. Lancet 381, 18661875 (2013).

7. Stroncek, D. F. & Rebulla, P. Platelet transfusions. Lancet 370, 427438 (2007).8. Brown, C. J. & Navarrete, C. V. Clinical relevance of the HLA system in blood transfusion. Vox Sang. 101, 93105 (2011).

9. Thomson, J. A. et al. Embryonic stem cell lines derived from human blastocysts. Science 282, 11451147 (1998).

10. Takahashi, K. et al. Induction of pluripotent stem cells from adult human broblasts by dened factors. Cell 131, 861872 (2007).

11. Murry, C. E. & Keller, G. Differentiation of embryonic stem cells to clinically relevant populations: lessons from embryonic development. Cell 132, 661680 (2008).

12. Wu, S. M. & Hochedlinger, K. Harnessing the potential of induced pluripotent stem cells for regenerative medicine. Nat. Cell Biol. 13, 497505 (2011).

13. Takayama, N. & Eto, K. Pluripotent stem cells reveal the developmental biology of human megakaryocytes and provide a source of platelets for clinical application. Cell. Mol. Life. Sci. 69, 34193428 (2012).

14. Corash, L. & Lin, L. Novel processes for inactivation of leukocytes to prevent transfusion-associated graft-versus-host disease. Bone Marrow Transplant. 33, 17 (2004).

15. Gaur, M. et al. Megakaryocytes derived from human embryonic stem cells: a genetically tractable system to study megakaryocytopoiesis and integrin function. J. Thromb. Haemost. 4, 436442 (2006).

16. Klimchenko, O. et al. A common bipotent progenitor generates the erythroid and megakaryocyte lineages in embryonic stem cell-derived primitive hematopoiesis. Blood 114, 15061517 (2009).

17. Takayama, N. et al. Transient activation of c-MYC expression is critical for efcient platelet generation from human induced pluripotent stem cells. J. Exp. Med. 207, 28172830 (2010).

18. Pick, M., Azzola, L., Osborne, E., Stanley, E. G. & Elefanty, A. G. Generation of megakaryocytic progenitors from human embryonic stem cells in a feeder- and serum-free medium. PLoS ONE 8, e55530 (2013).

19. Feng, Q. et al. Scalable generation of universal platelets from human induced pluripotent stem cells. Stem Cell Rep. 3, 817831 (2014).

NATURE COMMUNICATIONS | 7:11208 | DOI: 10.1038/ncomms11208 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 13

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11208

20. Nakamura, S. et al. Expandable megakaryocyte cell lines enable clinically applicable generation of platelets from human induced pluripotent stem cells. Cell Stem Cell 14, 535548 (2014).

21. Cherry, A. B. & Daley, G. Q. Reprogramming cellular identity for regenerative medicine. Cell 148, 11101122 (2012).

22. Zaret, K. S. & Carroll, J. S. Pioneer transcription factors: establishing competence for gene expression. Genes Dev. 25, 22272241 (2011).

23. Watkins, N. A. et al. A HaemAtlas: characterizing gene expression in differentiated human blood cells. Blood 113, e1e9 (2009).

24. Hu, Z. et al. VisANT 4.0: integrative network platform to connect genes, drugs, diseases and therapies. Nucleic Acids Res. 41, W225W231 (2013).25. Bernardo, A. S. et al. BRACHYURY and CDX2 mediate BMP-induced differentiation of human and mouse pluripotent stem cells into embryonic and extraembryonic lineages. Cell Stem Cell 9, 144155 (2011).

26. Tijssen, M. R. et al. Genome-wide analysis of simultaneous GATA1/2, RUNX1, FLI1, and SCL binding in megakaryocytes identies hematopoietic regulators. Dev. Cell 20, 597609 (2011).

27. Takayama, N. et al. Generation of functional platelets from human embryonic stem cells in vitro via ES-sacs, VEGF-promoted structures that concentrate hematopoietic progenitors. Blood 111, 52985306 (2008).

28. Liu, Z. J. & Sola-Visner, M. Neonatal and adult megakaryopoiesis. Curr. Opin. Hematol. 18, 330337 (2011).

29. Mootha, V. K. et al. PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nat. Genet. 34, 267273 (2003).

30. Bluteau, O. et al. Developmental changes in human megakaryopoiesis.J. Thromb. Haemost. 11, 17301741 (2013).31. Thon, J. N. et al. Cytoskeletal mechanics of proplatelet maturation and platelet release. J. Cell Biol. 191, 861874 (2010).

32. Gardiner, E. E. & Andrews, R. K. Platelet receptor expression and shedding: glycoprotein Ib-IX-V and glycoprotein VI. Transfus. Med. Rev. 28, 5660 (2014).

33. Hu, Z. & Yang, Y. G. Full reconstitution of human platelets in humanized mice after macrophage depletion. Blood 120, 17131716 (2012).

34. Hartley, P. S., Savill, J. & Brown, S. B. The death of human platelets during incubation in citrated plasma involves shedding of CD42b and aggregation of dead platelets. Thromb. Haemost. 95, 100106 (2006).

35. Versteeg, H. H., Heemskerk, J. W., Levi, M. & Reitsma, P. H. New fundamentals in hemostasis. Physiol. Rev. 93, 327358 (2013).

36. Tynngard, N., Wallstedt, M., Sodergren, A. L., Faxalv, L. & Ramstrom, S. Platelet adhesion changes during storage studied with a novel method using ow cytometry and protein-coated beads. Platelets 26, 177185 (2015).

37. Strauss, T., Sidlik-Muskatel, R. & Kenet, G. Developmental hemostasis: primary hemostasis and evaluation of platelet function in neonates. Semin. Fetal Neonatal Med. 16, 301304 (2011).

38. De Cuyper, I. M. et al. A novel ow cytometry-based platelet aggregation assay. Blood 121, e70e80 (2013).

39. Darbousset, R. et al. P2X1 expressed on polymorphonuclear neutrophils and platelets is required for thrombosis in mice. Blood 124, 25752585 (2014).40. Dore, L. C. & Crispino, J. D. Transcription factor networks in erythroid cell and megakaryocyte development. Blood 118, 231239 (2011).

41. Pimanda, J. E. et al. Gata2, Fli1, and Scl form a recursively wired gene-regulatory circuit during early hematopoietic development. Proc. Natl. Acad. Sci. USA 104, 1769217697 (2007).

42. Fujiwara, Y., Chang, A. N., Williams, A. M. & Orkin, S. H. Functional overlap of GATA-1 and GATA-2 in primitive hematopoietic development. Blood 103, 583585 (2004).

43. Tijssen, M. R. & Ghevaert, C. Transcription factors in late megakaryopoiesis and related platelet disorders. J. Thromb. Haemost. 11, 593604 (2013).

44. Heffner, G. C., Clutter, M. R., Nolan, G. P. & Weissman, I. L. Novel hematopoietic progenitor populations revealed by direct assessment of GATA1 protein expression and cMPL signaling events. Stem Cells 29, 17741782 (2011).

45. Iwasaki, H. et al. GATA-1 converts lymphoid and myelomonocytic progenitors into the megakaryocyte/erythrocyte lineages. Immunity 19, 451462 (2003).

46. Valtieri, M. et al. Enforced TAL-1 expression stimulates primitive, erythroid and megakaryocytic progenitors but blocks the granulopoietic differentiation program. Cancer Res. 58, 562569 (1998).

47. Moussa, O. et al. Thrombocytopenia in mice lacking the carboxy-terminal regulatory domain of the Ets transcription factor Fli1. Mol. Cell. Biol. 30, 51945206 (2010).

48. Liu, F., Walmsley, M., Rodaway, A. & Patient, R. Fli1 acts at the top of the transcriptional network driving blood and endothelial development. Curr. Biol. 18, 12341240 (2008).

49. Ismailoglu, I., Yeamans, G., Daley, G. Q., Perlingeiro, R. C. & Kyba, M. Mesodermal patterning activity of SCL. Exp. Hematol. 36, 15931603 (2008).

50. Real, P. J. et al. SCL/TAL1 regulates hematopoietic specication from human embryonic stem cells. Mol. Ther. 20, 14431453 (2012).

51. Toscano, M. G. et al. SCL/TAL1-mediated transcriptional network enhances megakaryocytic specication of human embryonic stem cells. Mol. Ther. 23, 158170 (2015).

52. Elcheva, I. et al. Direct induction of haematoendothelial programs in human pluripotent stem cells by transcriptional regulators. Nat. Commun. 5, 4372 (2014).

53. Gieger, C. et al. New gene functions in megakaryopoiesis and platelet formation. Nature 480, 201208 (2011).

54. Yamamoto, R. et al. Clonal analysis unveils self-renewing lineage-restricted progenitors generated directly from hematopoietic stem cells. Cell 154, 11121126 (2013).

55. Noh, J. Y. et al. Inducible Gata1 suppression expands megakaryocyte-erythroid progenitors from embryonic stem cells. J. Clin. Invest. 125, 23692374 (2015).

56. Wang, Y. et al. Comparative analysis of human ex vivo-generated platelets vs. megakaryocyte-generated platelets in mice: a cautionary tale. Blood 125, 36273636 (2015).

57. Lu, S. J. et al. Platelets generated from human embryonic stem cells are functional in vitro and in the microcirculation of living mice. Cell Res. 21, 530545 (2011).

58. Di Buduo, C. A. et al. Programmable 3D silk bone marrow niche for platelet generation ex vivo and modeling of megakaryopoiesis pathologies. Blood 125, 22542264 (2015).

59. Thon, J. N. et al. Platelet bioreactor-on-a-chip. Blood 124, 18571867 (2014).60. Nakagawa, Y. et al. Two differential ows in a bioreactor promoted platelet generation from human pluripotent stem cell-derived megakaryocytes. Exp. Hematol. 41, 742748 (2013).

61. Albers, C. A. et al. Exome sequencing identies NBEAL2 as the causative gene for gray platelet syndrome. Nat. Genet. 43, 735737 (2011).

62. Albers, C. A. et al. Compound inheritance of a low-frequency regulatory SNP and a rare null mutation in exon-junction complex subunit RBM8A causes TAR syndrome. Nat. Genet. 44, 435439 (2012).

63. Vallier, L. & Pedersen, R. Differentiation of human embryonic stem cells in adherent and in chemically dened culture conditions. Curr. Protoc. Stem Cell Biol.. Chapter 1, Unit 1D 4 11D 4 7 (2008).

64. Skepper, J. N. Immunocytochemical strategies for electron microscopy: choice or compromise. J. Microsc. 199, 136 (2000).

65. Cazenave, J. P. et al. Preparation of washed platelet suspensions from human and rodent blood. Methods Mol. Biol. 272, 1328 (2004).

66. Newman, P. J., Aster, R. & Boylan, B. Human platelets circulating in mice: applications for interrogating platelet function and survival, the efcacy of antiplatelet therapeutics, and the molecular basis of platelet immunological disorders. J. Thromb. Haemost. 5, 305309 (2007).

67. Ghevaert, C. et al. Recombinant HPA-1a antibody therapy for treatment of fetomaternal alloimmune thrombocytopenia: proof of principle in human volunteers. Blood 122, 313320 (2013).

68. de Witt, S. M. et al. Identication of platelet function defects by multi-parameter assessment of thrombus formation. Nat. Commun. 5, 4257 (2014).

Acknowledgements

We thank Ludovic Vallier, Tamir Rashid, Foad Rouhani, Filipa Soares, the BRC hiPSC and tissue culture core facilities for providing the human-induced pluripotent stem cell lines and reagents used in this study. We thank Katrin Voss for providing nucleic acid material for programming factor cloning as well as Francoise Pumio for the TAL1 lentiviral vector plasmid. We thank Jose Guerrero and Peter Smethurst for their intellectual and practical support in platelet function analyses, Stuart Meacham for the MK network data integration and visual, Nicholas Pugh for help in confocal imaging. We are grateful to Paquita Nurden and Jeremy Skepper for their advice on electron microscopy analyses. We thank Koji Eto and his lab for teaching us the megakaryocyte directed differentiation protocol in Tokyo. Finally we thank members of the Anne McLaren Laboratory, the NHSBT Cambridge Centre and University of Cambridge for stimulating discussions and technical support. This work was supported by the Leukemia and Lymphoma Society grant, the UK Medical Research Council (Roger Pedersen), the National Institute for Health Research (NIHR; RP-PG-0310-1002; Willem Ouwehand and Cedric Ghevaert) and a core support grant from the Wellcome Trust and MRC to the Wellcome TrustMedical Research Council Cambridge Stem Cell Institute. Cedric Ghevaert is supported by the British Heart Foundation (FS/09/039); Marloes Tijssen is supported by the European Hematology Association (Research fellowship) and the British Heart Foundation (PG/13/77/30375). Catherine Hobbs was supported by the National Health Service Blood and Transplant. Matthew Trotter was supported by a Medical Research Council Centre grant (MRC Centre for Stem Cell Biology and Regenerative Medicine); since participation in the work described, Matthew Trotter has become an employee of Celgene Research SLU, part of Celgene Corporation. Nicole Soranzos research and Sanger Institute afliates are supported by the WellcomeTrust (WT098051 and WT091310), the EU FP7 (Epigenesys 257082 and Blueprint HEALTH-F5-2011-282510). The Cambridge Biomedical Centre (BRC) hIPSCs core facility is funded by the NIHR.

14 NATURE COMMUNICATIONS | 7:11208 | DOI: 10.1038/ncomms11208 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11208 ARTICLE

Author contributions

T.M. designed, analysed, performed and interpreted most experiments and wrote the paper. A.L.E. performed and analysed experiments. L.V. and Y.Y. performed and interpreted bioinformatics analyses pertaining to the genomic characterization of the MKs. M.R.T. contributed to megakaryocyte characterization and wrote the paper. M.W.T. performed and interpreted bioinformatics analyses pertaining to the forward programming concept. M.C. and M.A. contributed to megakaryocyte characterization. D.H., W.H.W., C.M.H., A.D., R.L., G.B. and D.C.P. performed experiments. H.P., T.P. and A.B. performed intravital microscopy experiments. N.S. interpreted bioinformatics analyses pertaining to the genomic characterization of the MKs. W.H.O. contributed platelet expert input to the forward programming concept. R.A.P. conceived the forward programming approach, designed, analysed and interpreted experiments and wrote the paper. C.G. drove the platelet biology, designed, analysed and interpreted experiments and wrote the paper.

Additional information

Accession codes: Gene expression microarray data are available from the GEO repository under the accession number GSE54822.

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/ http://www.nature.com/naturecommunications

Web End =naturecommunications

Competing nancial interests: The authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Moreau, T. et al. Large-scale production of megakaryocytes from human pluripotent stem cells by chemically dened forward programming. Nat. Commun. 7:11208 doi: 10.1038/ncomms11208 (2016).

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

NATURE COMMUNICATIONS | 7:11208 | DOI: 10.1038/ncomms11208 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 15