bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1 Human induced pluripotent stem cell-derived neuroectodermal epithelial cells 2 mistaken for blood-brain barrier-forming endothelial cells 3 4 Tyler M. Lu1,2, David Redmond3, Tarig Magdeldin4, Duc-Huy T. Nguyen1, Amanda 5,6 5,6 7 1 4 2 5 Snead , Andrew Sproul , Jenny Xiang , Koji Shido , Howard A. Fine , Zev Rosenwaks , 6 Shahin Rafii1, Dritan Agalliu5,8* & Raphaël Lis1,2* 7 8 9 1 Ansary Stem Cell Institute, Division of Regenerative Medicine, Department of Medicine, 10 Weill Cornell Medicine, New York, NY 10065, USA 11 2 Ronald O. Perelman and Claudia Cohen Center for Reproductive Medicine, Weill 12 Cornell Medicine, New York, NY 10065, USA 13 3 Parker Institute for Cancer Immunotherapy, Memorial Sloan Kettering, New York, NY 14 10065, USA 15 4 Department of Neurology and The Sandra and Edward Meyer Cancer Center, Weill 16 Cornell Medicine-New York Presbyterian Hospital, New York, NY 10065, USA 17 5 Department of Pathology & Cell Biology, Columbia University Irving Medical Center, 18 New York, NY 10032, USA 19 6 The Taub Institute for Research on Alzheimer’s Disease and the Aging Brain, Columbia 20 University Irving Medical Center, New York, NY 10032, USA 21 7 Genomics Resources Core Facility, Weill Cornell Medicine, New York, New York 10065, 22 USA 23 8 Department of Neurology, Columbia University Irving Medical Center, New York, NY 24 10032, USA 25 26 * e-mail: Raphaël Lis [email protected] 27 Dritan Agalliu [email protected] 28 29 30 31 32 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

33 Abstract 34 Brain microvascular endothelial cells (BMECs) possess unique properties

35 underlying the blood-brain-barrier (BBB), that are crucial for homeostatic brain functions

36 and interactions with the immune system. Modulation of BBB function is essential for

37 treatment of neurological diseases and effective tumor targeting. Studies to-date have

38 been hampered by the lack of physiological models using cultivated human BMECs that

39 sustain BBB properties. Recently, differentiation of induced pluripotent stem cells (iPSCs)

40 into cells with BBB-like properties has been reported, providing a robust in vitro model for

41 drug screening and mechanistic understanding of neurological diseases. However, the

42 precise identity of these iBMECs remains unclear. Employing single-cell RNA

43 sequencing, bioinformatic analysis and immunofluorescence for several pathways,

44 transcription factors (TFs), and surface markers, we examined the molecular and

45 functional properties of iBMECs differentiated either in the absence or presence of

46 retinoic acid. We found that iBMECs lack both endothelial-lineage genes and ETS TFs

47 that are essential for the establishment and maintenance of EC identity. Moreover,

48 iBMECs fail to respond to angiogenic stimuli and form lumenized vessels in vivo. We

49 demonstrate that human iBMECs are not barrier-forming ECs but rather EpCAM+

50 neuroectodermal epithelial cells (NE-EpiCs) that form tight junctions resembling those

51 present in BBB-forming BMECs. Finally, overexpression of ETS TFs (ETV2, FLI1, and

52 ERG) reprograms NE-EpiCs to become more like the BBB-forming ECs. Thus, although

53 directed differentiation of human iBMECs primarily gives rise to epithelial cells,

54 overexpression of several ETS TFs can divert them toward a vascular BBB in vitro.

55 56 57

2 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

58 Introduction 59 Brain microvascular endothelial cells (BMECs) which line the vascular network of

60 the central nervous system (CNS) form a specialized barrier termed the blood-brain

61 barrier (BBB) that regulates the dynamic transfer of select molecules into and out of the

62 CNS. The BBB also restricts the entry of immune cell and is essential for healthy brain

63 activity. These properties are achieved through the presence of specialized tight junctions

64 exhibiting a high transendothelial electrical resistance (TEER) in vivo, reduced caveolar-

65 mediated transport and the presence of selective transporters1-3. Loss of BBB integrity is

66 a hallmark of many neurological diseases such as ischemic stroke and multiple sclerosis

67 and contributes to tumor metastasis and pathogenesis of glioblastoma multiforme

68 (GBM)1,4. The presence of a neurovascular barrier, on the other hand, impedes

69 successful delivery of therapeutics into the CNS in many neurological and psychiatric

70 diseases. The development of in vitro models has accelerated mechanical studies on the

71 BBB as well as large scale screening of drugs with potential to penetrate the brain, with

72 some limitations5-9. Human primary BMECs are difficult to procure in sufficient numbers

73 for experimental purposes, and their barrier properties cannot be maintained in vitro.

74 Many studies have used immortalized human BMECs, however these cells lose their BBB

75 attributes and produce a sub-physiologic TEER, rendering them ineffective for functional

76 studies8.

77 Recently, the generation of a pure population of putative BMECs from pluripotent

78 stem cell sources (iBMECs) has been described to meet the need for a reliable and

79 reproducible in vitro human BBB model10-22. Human pluripotent stem cells, embryonic or

80 induced, can differentiate into large quantities of specialized cells in order to study

81 development and model disease23-27. iBMECs are generated primarily through

3 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

82 differentiation of pluripotent stem cells into neural and endothelial progenitors, followed by

83 selective purification (referred to as neuroendothelial differentiation)10,28,29. Adding retinoic

84 acid or inhibiting GSK3 during the differentiation process reportedly enhances the yield

85 and fidelity of these putative iBMECs10,30. These iBMECs display endothelial-like

86 properties, such as tube formation and low-density lipoprotein uptake, high

87 transendothelial electrical resistance (TEER) and barrier-like efflux transporter

88 activities10,29.

89 We set out to compare iBMECs obtained by neuroendothelial differentiation to

90 endothelial cells generated by transition through a mesodermal stage (iECs), referred to

91 as mesoendothelial differentiation31. Using a combination of single cell RNA sequencing,

92 bioinformatics and immunofluorescence, we find that iBMECs obtained by

93 neuroendothelial differentiation lack canonical endothelial cell markers (CDH5, PECAM1,

94 VEGFR2, APLNR, eNOS) as well as the ETS transcription factors (ETS1, ETV6, FLI1)

95 crucial for establishing the endothelial identity, when compared to their mesoendothelial

96 counterparts. Despite having some features of endothelial cells (e.g. high TEER), iBMECs

97 do not form lumenized vessels in immunocompromised mice (NOD-scid IL2Rgnull-SGM3).

98 Moreover, at a molecular level, iBMECs express high levels of EpCAM and exhibit a gene

99 signature typical of neuroectodermal epithelial cells. Analysis by both global and single-

100 cell RNA sequencing combined with immunofluorescence reveals multiple

101 neuroectodermal epithelial related genes32,33 such as TTR, TRP3, NKCC and AQP1

102 present in EpCAM+ NE-EpiCs cells.

103 Finally, NE-EpiCs can be directed towards a BBB-like EC fate by overexpressing

104 three key endothelial ETS transcription factors ETV2, FLI1, and ERG34. Thus, similar to

105 hematopoietic stem and progenitor cells35,36, directed differentiation of pluripotent stem

4 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

106 cells into human BMECs with appropriate neurovascular barrier properties requires

107 introduction of key EC TFs to develop a robust human BBB model in vitro that could

108 ultimately be utilized for functional studies and drug discovery.

109 Results

110 Neuroendothelial differentiation derived iBMECs are not phenotypically endothelial

111 cells.

112 Several groups have developed strategies that simultaneously differentiate human

113 pluripotent stem cells (hPSC) to both neural and EC lineages, rationalizing that this co-

114 differentiation yields hPSC-derived ECs possessing BBB attributes (Fig. 1a, referred to as

115 neuroendothelial iBMECs)10-22. To assess the properties of these cells, we compared this

116 protocol to a well-established method in which ECs are differentiated from pluripotent

117 sources through a mesodermal lineage (referred to as mesoendothelial iECs; Fig 1a)31,

118 using two IPS cell lines (IMR90-4 and H6) and one ES cell line (H1). The mesoendothelial

119 protocol drives hPSC differentiation towards mesodermal lineages using a chemically

120 defined, serum-free method that enhances vascular differentiation. Heterogenous

121 embryoid body cultures of hPSCs are stimulated with Bone Morphogenetic Protein 4

122 (BMP-4), Activin A, Fibroblast Growth Factor 2 (FGF-2) and VEGF-A, with the addition of

123 the Transforming Growth Factor beta (TGF-β) inhibitor SB43154231 (Fig. 1a). This

124 approach generates iECs that express VE-Cadherin (VECAD; CDH5) and PECAM1

125 (CD31) but lack EpCAM (CD326) and haematopoietic markers

126 (VECAD+PECAM1+EpCAM-CD45- cells; Fig. 1c, d).

127 Phase-contrast microscopy reveals that iBMECs exhibit a characteristic EC

128 morphology (Fig. 1b, Extended Data Fig. 1a; n = 5 biological replicates), however they

129 lack the canonical EC markers VECAD and PECAM1 by day 16 (Fig. 1c, Extended Data

5 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

130 Fig. 1b; n = 5 biological replicates). Few iBMECs derived from hPSCs express low levels

131 of VECAD early during differentiation; however, they lose VECAD expression by day 16

132 (Fig. 1c, d; Extended Data Fig. 1b). Moreover, the VECAD is not homogenously

133 distributed throughout the cell membrane in iBMECs. In contrast, the mesoendothelial-

134 derived iECs, primary Human Umbilical Vein Endothelial Cells (HUVECs) and Brain

135 Microvascular Endothelial Cells (BMECs) express high levels of junctional VECAD and

136 PECAM1 (Fig 1c-d, Extended Data Fig. 1b). Quantification of ECs by fluorescence

137 activated cell sorting (FACS) shows that the neuroendothelial differentiation protocol

138 yields on average 0.03±0.02%, 0.01±0.01%, 0.01±0.01%, VECAD+PECAM1+ marked

139 cells at day 11 and 0.13±0.07%, 0.02±0.01%, 0.03±0.015% VECAD+PECAM1+ marked

140 cells at day 16 (for H1, H6 and IMR90-4, respectively; Fig. 1d, e). We tested the in vivo

141 angiogenic potential of iBMECs using matrigel plugs implanted in immunocompromised

142 animals (NSG-SGM3, n=10 animals per cell line tested). Day 16 iBMECs derived from

143 H1, H6 or IMR90-4 fail to establish a vascular network compared to the bona fide

144 endothelial cells (HUVECs) (Extended Data Fig.1c-e). The poor yield of

145 VECAD+PECAM1+ in iBMECs is not due to the lack of endothelial “competence” from the

146 source cell lines (H1, H6 and IMR90-4), because they produce 3.23±2.41%, 4.43±1.47%,

147 8.70±2.78% VECAD+PECAM1+ iECs when differentiated with the mesoendothelial

148 protocol (H1, H6 and IMR90-4, respectively; Fig. 1d, f). Upon purification by FACS on day

149 12, these iECs can be readily expanded (data not shown). These findings were compared

150 to primary human umbilical vein ECs (HUVECs) and brain microvascular ECs (BMECs),

151 which contain 97.17±0.98% and 83.37±4.028% VECAD+PECAM1+ cells (Fig. 1d-f). Thus,

6 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

152 neuroendothelial differentiation conditions do not generate iBMEC cells with an

153 endothelial phenotype, raising questions regarding their identity.

154

155

156 The neurorendothelial iBMECs transcriptome is not congruent with an EC

157 signature

158 We tested the putative iBMEC EC identity using single-cell RNA sequencing (sc-

159 RNAseq). We sequenced and retained after quality control 2812, 2423 and 4980 iBMECs

160 (from H1, H6 and IMR90-4, respectively) plus 4833 iECs (all cell lines included) and 3219

161 primary ECs (HUVECs/BMECs) (Extended Data Fig. 2b). Single cells transcriptomes

162 were sequenced with an average depth of 7310 nUMI and 2684 genes per cell (Extended

163 Data Fig. 2b). To detect relationships among cells from various differentiation protocols

164 and sources, we visualized all cells with t-SNE and grouped them using unbiased

165 hierarchical clustering (Fig. 2a). Because most cells from a similar differentiation protocol

166 cluster together across cell source and biological replicates, this indicates that batch

167 effects are not a source of variance in the dataset (Fig. 2a). To define cell identities, we

168 investigated how the most highly variable genes contribute to cell-type identity, by

169 clustering averaged gene-expression profiles for each cell type using only the top 400

170 highly variable genes from the dataset (Fig. 2a, cluster a to i’). This shows that manual

171 annotation (Fig. 2a, sample name) of cell types is consistent with unbiased transcriptomic

172 clustering (Fig. 2a, cluster a to i’), suggesting that the effect of cell type, differentiation

173 protocol and cell source on measured gene expression is stronger than batch effects.

174 The resulting dendrogram (Fig. 2b) segregates into two main branches: node #37

175 and #38 (Fig. 2b). Node #37 accounts for gene clusters a-n and #38 for gene clusters o-i’.

176 Manual annotation revealed that node #37 is comprised of all primary ECs (HUVECs and

177 BMECs) as well as mesoendothelial iEC samples, thereby defining an endothelial core

7 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

178 population, while node #38 contains neuroendothelial iBMECs exclusively (Fig. 2a, b).

179 ECs are defined not only by their ability to form blood vessels, but also by expression of

180 unique transcription factors (TFs), signaling pathways and growth (angiocrine) factors37-40.

181 Using our unsupervised single cell RNA-seq analysis, we curated a list of essential genes

182 associated with EC identity that are highly expressed by cell clusters a-e from Fig. 2a.

183 This supervised analysis revealed that iBMECs, independently of the cell line and the day

184 of differentiation tested, lack expression of EC transcription factors (ETS1, ETS2, ELK3,

185 ELF1, GATA2, FLI1, SOX17, SOX18) compared to mesoendothelial iECs and bona fide

186 adult endothelial cell controls (Fig. 3). iBMECs also lack general EC cell signaling

187 machinery, including VEGFR2 (VEGF pathway), ICAM2, Notch ligands (JAG1, JAG2,

188 NOTCH4, EGFL7) and NOS3 (Fig. 3, Extended data Fig. 4,7). Moreover, iBMECs lack

189 secreted or membrane-bound angiocrine factors inherent to EC identity, such as

190 ANGPT2, APLN, BMP4, BMP6, FGF2, IGFBP4 and IGFBP7 (Fig. 3). CNS ECs are

191 characterized by high levels of WNT/β-catenin signaling which is required to induce and

192 maintain BBB properties41-44. We analyzed expression of WNT/β-catenin signaling

193 components in iBMECs and found that they express low levels of key receptors (FZD4,

194 LRP5/6) but lack downstream targets that are indicative of pathway activation (LEF1,

195 NKD1/2, AXIN2, SOX17) (Extended Data Fig. 5). Therefore, the neuroendothelial

196 differentiation protocol yields iBMECs cells that lack both phenotypic and molecular

197 signatures of both general EC lineage and the CNS endothelium.

198

199 iBMECs are EpCAM+ cells displaying neuroectodermal epithelial-like properties

200 Upon further interrogation of our single-cell RNAseq dataset comprising 20,069

201 cells (Fig. 2a), we identified several genes differentially expressed at high levels between

202 cell populations stemming from nodes 37 and 38 (Fig. 2b). Using these highly

203 differentially expressed genes, we performed a supervised analysis and found that

8 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

204 iBMECs express high levels of genes not generally associated with EC in addition to a

205 unique CLAUDIN repertoire. These included EPCAM, KRT7/8/19, SLC12A2, FREM2, and

206 SPP1 (Fig. 4a). Confocal microscopy and FACS analysis both confirm EPCAM

207 expression in a large subset of iBMECs (Fig. 4b). Expression of these genes suggest that

208 iBMECs have an epithelial phenotype in contrast to an endothelial one as previously

209 suggested10-22.

210 Neurovascular ECs form the BBB through expression of several tight junction

211 proteins (CLAUDIN-5, OCCLUDIN, ZO-1/2)3. To assess the BBB-like signature of

212 iBMECs, we mined our dataset using the KEGG tight junction pathway (map04530). Our

213 analysis reveals only minor gene expression differences among iBMECs,

214 mesoendothelial iECs and EC controls (HUVEC, BMEC) with the exception of the

215 CLAUDIN family repertoire (Extended Data Fig. 6, 7). All iECs and primary EC controls

216 express CLAUDIN-5 and -11, while iBMECs express a distinct repertoire (CLAUDIN-4, 6

217 and 7), which is characteristic of epithelial cells45 (Extended Data Fig. 6). However, in

218 contrast to epithelial cells or BBB-forming ECs, iBMECs express low levels OCCLUDIN,

219 JAM2 and JAM3 at the RNA level (Extended Data Fig. 6). We performed

220 immunofluorescence for CLAUDIN-5, OCCLUDIN and ZO-1 proteins in these cells using

221 the published antibodies10,29. We found that antibodies for CLAUDIN-5 and OCCLUDIN

222 recognize proteins at cell-cell junctions that co-localize with ZO-1 (Extended Data Fig. 8a-

223 f). We then measured TEER over time and found that these cells form high TEER

224 (Extended Data Fig. 8g,h) consistent with previous studies10,29. However, given that

225 CLAUDIN-5 and OCCLUDIN are not present at the RNA level in these cells (Extended

226 Data Fig. 6), we conclude that these antibodies may cross-react with epithelial CLAUDIN

227 proteins. Therefore, the neuroendothelial differentiation protocol generates iBMEC cells

228 that do not exhibit the prototypical tight junction protein repertoire normally found in either

229 peripheral or CNS ECs.

9 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

230 Within the CNS environment, the choroid plexus contains epithelial cells that form

231 tight junctions and prevent transcytosis of unwanted substances from the blood into the

232 cerebrospinal fluid (CSF), to form the blood-CSF barrier20,46-48. We investigated whether

233 iBMECs resemble these neuroectodermal epithelial cells (NE-EpiCs). Single-cell RNAseq

234 analysis shows that iBMECs express some of the genes associated with the ependymal

235 cell lineage of the choroid plexus such as EPCAM, KRT7/8/18/19, SLC12A2, FREM2 and

236 SPP1 (Fig. 4a,b,d). Additionally, iBMECs express low levels of some choroid plexus

237 epithelial genes such as TTR, AQP1, KRT17, or CLDN1 (Fig. 4a, Extended data Fig. 6).

238 Significantly, none of these epithelial genes were present in either iECs generated from

239 the same hPSC cell lines or control endothelium (Fig. 4a). We also validated the

240 expression of transthyretin (TTR), a terminal differentiation marker of choroid plexus

241 epithelial cells, by immunofluorescence. We found that some EpCAM+ cells express TTR,

242 suggesting an ependymal/choroid plexus-like identity (Fig. 4c).

243 244 The neuroendothelial differentiation protocol reproducibly generates non-neural

245 epithelial cells lacking endothelial identity

246 To substantiate the validity of our analysis, global gene expression profiles

247 obtained by bulk RNA sequencing of previously published iBMECs18 were compared to

248 our iBMECs and bona fide ECs (HUVECs and BMECs plus previously published iBMECs

249 18,34; Fig. 5a). Hierarchical clustering based on the top 500 highly variable genes across

250 the dataset reveals clear segregation between iBMECs and bona fide ECs (Fig. 5a).

251 Notably, a previously published primary BMEC control (Qian et al.18) clusters with our

252 endothelial cell control, strongly suggesting that segregation between iBMECs and ECs is

253 not due to either library preparation or sequencing bias. We then supervised this

254 clustering using the genes that define our epithelial/endothelial groups from single-cell

10 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

255 RNA-seq analysis. iBMECs derived from neuroendothelial co-differentiation show a

256 distinct separation from all adult ECs, regardless of the source of the laboratory that

257 generated them10,18,29 (Fig. 5b). Previously published lines10,18,29 as well as iBMECs

258 generated by a chemically defined differentiation protocol18 lack key endothelial surface

259 markers (CDH5, PECAM1, CLDN5, ICAM2, EGFL7), critical angiocrine factors (NOS3,

260 BMP6, BMP2, KDR) as well as ETS transcription factors (FLI1, ERG, SOX17, TAL1) that

261 are fundamental for endothelial identity (Fig. 5b). Taken together, these data strongly

262 suggest that the custom two-dimensional hPSC differentiation strategy promoting neural

263 and “endothelial” co-differentiation described in10,18,29 reproducibly generates pure

264 populations of neuroectodermal epithelial cells lacking an endothelial identity.

265

266 ETS transcription factors drive neuroectodermal epithelial cells towards an

267 endothelial fate

268 We previously showed that lineage-committed epithelial cells (EpCAM+Tra1-81−c-

269 Kit−) from amniotic fluid are amenable to transcription factor-mediated reprogramming into

270 vascular ECs34. Transient ETV2 along with constitutive FLI1 and ERG1 expression not

271 only stably induces expression of EC genes in committed epithelial cells, but also

272 suppresses expression of non-vascular genes34. We therefore transduced IMR90-4 iPS

273 cells with the ETS transcription factors ETV2, ERG and FLI1 (EEF) at day 6 of the

274 neuroendothelial co-differentiation protocol when epithelial fate is induced (Fig. 6a). We

275 allowed these TFs to be expressed for the duration of differentiation (Fig. 6a). Under

276 these conditions, EEF-iBMECs express both PECAM1 and VECAD by day 11 (Fig. 6b).

277 These cells can be sorted and, with the addition of a TGF-β signaling inhibitor

278 (SB431542), expanded as stable, phenotypically marked ECs defined by 11 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

279 VECAD+PECAM1+EpCAM- expression by day 27 (Fig. 6b, lower panel). Moreover, EEF-

280 enforced expression followed by TGF-β inhibition restores VECAD and PECAM1 proper

281 localization to cell junctions (Extended data Fig. 3b). We inspected the quality of our

282 reprogramming strategy using single-cell RNA sequencing. Compared to untransduced

283 controls, day 11 EEF-iBMECs show tight association with iECs, HUVECs and primary

284 BMECs (Fig. 6c-d). A significant number of vascular genes (VECAD, PECAM1, CLND5

285 and SOX17) are upregulated in EEF-iBMECs compared to untransduced control iBMECs

286 (Fig. 6c). Expression levels of these EC genes are comparable to those from iECs,

287 HUVEC and primary BMECs cultured in vitro, supporting the notion that EEF-iBMECs

288 have an endothelial identity. Moreover, angiocrine factors that regulate EC function40

289 such as BMPs and Notch ligands are all expressed in EEF-iBMECs (Fig. 6c). In

290 conclusion, genome-wide, single-cell RNA sequencing analyses demonstrate that forced

291 expression of ETV2, FLI1 and ERG1 in conjunction with TGF-β inhibition reprograms

292 neuroectodermal epithelial cells into endothelial cells while silencing non-vascular genes,

293 strengthening our assertion that the chemically defined neuroendothelial differentiation

294 protocol10,18,29 fails to produce CNS-specific endothelium with BBB properties.

295

296 Discussion

297 Contrary to several prior reports10-22, here we demonstrate that iBMECs generated

298 by neuroendothelial co-differentiation from H1, H6 and IMR90-4 hPSCs are not proper

299 BBB-forming ECs, but rather EPCAM+PECAM1-VECAD- neuroectodermal epithelial cells

300 with similarities to the choroid plexus cells, rendering their use as an in vitro model of the

301 BBB questionable.

12 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

302 Our analysis also illustrates that iBMECs are fundamentally different from iECs

303 arising from the same cell lines derived through mesodermal differentiation, as well as

304 primary HUVEC and BMECs. iECs derived from a mesodermal differentiation protocol are

305 EPCAM-PECAM1+VECAD+ cells that can expand in vitro and express bona fide ECs

306 characteristics. Therefore, there is no inherent restriction of human iPSC or ESC cells to

307 acquire EC properties. An unbiased, graph-based clustering of our single-cell RNAseq

308 data shows that mesodermal-derived iECs closely resemble primary human ECs while

309 neuroendothelial iBMECs do not, further underscoring the vast transcriptional and

310 phenotypic differences that exist between iBMECs and CNS ECs. Upon manual

311 annotation of these unbiased cell clusters with their corresponding sample identities, we

312 conclude that batch effect and starting cell line are not the main source of variance in

313 these results. Moreover, bulk RNA sequencing, which has a greater depth of sequencing,

314 confirms the identity of the iBMECs obtained from our single cell RNA sequencing data. In

315 an unbiased analysis using bulk RNA seq data, our iBMECs were found to have nearly

316 identical global gene expression profiles as cells generated from the neuro-endothelial co-

317 differentiation in several previous studies10,18,29. These findings substantiate our

318 conclusions that the published neuroendothelial protocol is working in our hands. The

319 same analysis also shows a primary BMEC control from the same previously published

320 dataset which closely clusters with our primary EC controls, allowing us to conclude that

321 sequencing bias and batch effect do not account for any major differences in the assigned

322 cell identities. Despite the presence of “BBB-like” properties such as the presence of tight

323 junction proteins, ability to form a high TEER and expression of several transporters due

324 to their epithelial nature, iBMECs cannot be used as an in vitro model of the human BBB

325 for future studies.

13 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

326 We have not shown, however, that iBMECs are definitive hPSC-derived

327 neuroectodermal epithelial-like cells (NE-EpiCs), since they harbor many, but not all

328 attributes of an epithelial lineage present in the CNS. We demonstrate that constitutive

329 overexpression of ETS factors (FLI1, ERG, ETV2) using lentiviral vectors during the

330 specification phase, reprograms NE-EpiCs towards an EC identity as demonstrated by an

331 EPCAM-PECAM1+VECAD+ phenotype. Although these cells may not represent definitive

332 BBB-forming CNS ECs, the EEF-iBMECs have a more valid EC identity than the iBMECs

333 currently in use as an in vitro BBB model. These reprogrammed cells have both the

334 capacity to be passaged and expand in vitro, while retaining their PECAM1+VECAD+ EC

335 phenotype. We believe that this is a crucial step towards de novo generation of true BBB-

336 forming CNS ECs for suitable in vitro modeling of physiological and pharmaceutical

337 studies which remains a major issue with the current iBMECs. Therefore, while

338 neuroendothelial differentiation of human iPCS to BMECs is rare, overexpression of ETS

339 transcription factors may promote conversion into vascular cells that are capable to form

340 barrier properties.

14 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

341 Methods

342 Maintenance of hPSCs and neuroendothelial co-differentiation to iBMECs

343 These protocols were precisely followed as described previously10,28,29. All

344 reagents were procured based on these protocols, and each step and duration for

345 differentiation was followed rigorously. IMR90-4 and H6 iPSCs and H1 hESCs were

346 maintained between passages 20–42 on Matrigel (BD Biosciences, 354277) in mTeSR1

347 or TeSR-E8 medium (STEMCELL Technologies, 85850, 05990). For differentiation, cells

348 were passaged onto Matrigel in mTeSR1 or StemFlex (Gibco, A3349401) medium and

349 allowed to expand for 3 days. Cultures were then switched to unconditioned medium

350 (UM) lacking bFGF for 6 days. EC medium consisting of human endothelial serum-free

351 medium (hESFM; Gibco, 11111044) supplemented with 20 ng/mL bFGF (Peprotech, 100-

352 18B), 1% platelet-poor plasma-derived bovine serum (Alfa Aesar, J64483AE), and 10 μM

353 all-trans retinoic acid (Sigma, R2625) was then added for an additional 2 days. Cells were

354 then dissociated with Accutase (Invitrogen, 00-4555-56) and plated onto 6-well

355 polystyrene plates or 1.12 cm2 Transwell-Clear permeable inserts (0.4 mm pore size) in

356 EC media. Culture plates were coated with a mixture of Collagen IV (400 mg/mL; Sigma,

357 C6745) and Fibronectin (100 mg/ mL; Sigma, F1141) in H2O for at least 30 min at 37°C,

358 whereas inserts were incubated for a minimum of 4 h at 37°C. The resulting purified

359 hPSC-derived iBMECs were then grown in EC medium for 24 h, after which RA was

360 removed and cells were cultured until the indicated experimental timepoints.

361

362 PSC mesoendothelial differentiation to iECs

363 hPSC were maintained exactly as described above. Embryoid bodies were

364 generated and cultured in base hPSC medium consisting of StemPro-34 (Gibco,

15 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

365 10639011), nonessential amino acids (Gibco, 11140050), glutamine (Gibco, 35050061),

366 penicillin/streptomycin/amphotericin B (Gibco, 15240096), β-mercaptoethanol

367 (MilliporeSigma, M3148) for 24 hours before the start of differentiation and the protocol

368 was carried out as previously described31. Medium is changed every 2 days for the

369 duration of the differentiation. Addition of 20 ng/ml BMP-4 (R&D Systems, 314-BP-05/C)

370 on day 0 (removed on day 7); on day 1, medium is supplemented with 10 ng/ml activinA

371 (STEMCELL technologies, 78001; removed at day 4); on day 2, medium is further

372 supplemented with 8 ng/ml FGF-2 (Peprotech, 100-18B; remained for the duration of

373 culture); on day 4, embryoid bodies are transferred to adherent conditions on Matrigel-

374 coated plates and medium was supplemented with 25 ng/ml VEGF-A (Peprotech, 100-20;

375 remained for the duration of culture) to specify vascular progenitors; on day 7, SB431542

376 (R&D Systems, 1614; TGFβ signaling inhibitor) is added at 10 μM concentration and

377 remained for indicated duration to expand terminally differentiated endothelial cells (iECs)

378

379 Directing hPSC neuroendothelial co-differentiation towards EEF-iBMECs

380 IMR90-4, H6, and H1 hPSCs were differentiated according to the Neuro-

381 Endothelial co-differentiation protocol as described previously34. Following the 6 days of

382 differentiation in UM, cells were transduced via lentiviral vectors for ETS transcription

383 factors FLI1, ERG, ETV2 and cultured in EC medium, exactly as mentioned above. No

384 further changes were made to the protocol because the cells were terminally cultured.

385

386 Single-cell RNAseq digital droplet sequencing (ddSeq)

387 A single-cell suspension was loaded into the Bio-Rad ddSEQ Single-Cell Isolator

388 on which cells were isolated, lysed and barcoded in droplets. Droplets were then

16 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

389 disrupted and cDNA was pooled for second strand synthesis. Libraries were generated

390 with direct tagmentation followed by 3’ enrichment and sample indexing using Illumina

391 Nextera library prep kit. Pooled libraries were sequenced on the Illumina NextSeq500

392 sequencer. Sequencing data were primarily analyzed using the SureCell RNA Single-Cell

393 App in Illumina BaseSpace Sequence Hub.

394 All single cell analyses were performed using the Seurat package in R (version

395 2.3.4). UMI counts files that had been knee-filtered were downloaded from the Illumina

396 BaseSpace Sequence Hub. After initial quality control, cells that were included in the

397 analysis were required to have a minimum 900 genes expressed and a maximum 5500

398 genes expressed, in addition to a minimum 1100 UMIs. This resulted in a total 20,069

399 cells passing quality filters across the 18 samples as seen in Extended Data Figure 2b.

400 Following best practices in the package suggestions UMI counts were log-normalized and

401 after the most highly variable genes selected the data matrices were scaled using a linear

402 model with variation arising from UMI counts mitigated for. Principal component analysis

403 was subsequently performed on this matrix and after reviewing principal component

404 heatmaps and jackstraw plots t-distributed stochastic neighbor embedding analysis

405 (tSNE) visualization were performed on the top 40 components and clustering resolution

406 was set at 1.0 for visualizations. Differential gene expression for gene marker discovery

407 across the clusters were performed using the Wilcoxon rank sum test as used in the

408 Seurat package.

409

410 Immunofluorescence

411 Fixed Samples: Cells were fixed with 4% paraformaldehyde for 15 minutes prior to

412 staining. To prevent non-specific binding of the primary antibody, samples were blocked

17 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

413 with 5% horse serum in PBS solution for 60 minutes. Cells were washed 3 times in PBS

414 and dilute primary antibody in 5% horse serum was added according to manufacturer’s

415 recommended concentration and left overnight at 4ºC. If the protein was intracellular,

416 permeabilization using 0.2% Triton X-100 was necessary during block and staining with

417 primary antibody. After staining with primary antibody, cells were incubated with

418 secondary antibody in 5% horse serum for 1 hour at room temperature on a shaker.

419 Following a thorough wash to remove unbound secondary antibody, DAPI was used at 1

420 µg/ml to stain nuclei.

421 Live Samples: Human IgG antibody was used at 1:50 in PBS with 0.5% BSA and

422 2mM EDTA to block Fc receptors prior to staining. The blocked cells were stained for 30

423 minutes at 4°C with fluorochrome-conjugated antibodies according to the

424 manufacturer’s recommendations. Stained cells were washed in PBS to prevent

425 saturation of the fluorescent dye during imaging. All imaging was performed using a

426 Zeiss 710 META confocal microscope.

427

428 Flow cytometry

429 Before staining, Fc receptors were blocked with human IgG antibody at 1:50

430 (Biolegend) in PBS (pH 7.2) containing 0.5% BSA (Fraction V) and 2mM EDTA for

431 10min at 4°C. For cell surface staining, samples were stained for 30min at 4°C

432 with fluorochrome-conjugated antibodies according to the manufacturer’s

433 recommendation. Stained cells were washed in blocking buffer and fixed in 1% PFA in

434 PBS (pH 7.2) with 2mM EDTA for flow analysis or resuspended in PBS (pH 7.2) with

435 2mM EDTA and 1μg/ml DAPI (Biolegend) for sorting. Samples were analysed using

436 a FACS ARIA II SORP (BD Biosciences) Data were collected and analysed using FACs

18 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

437 DIVA 8.0.1 software (BD Biosciences). All gating was determined using unstained

438 controls and fluorescence minus one strategies.

439

440 Bulk RNA sequencing

441 Total RNA, greater than 100ng, from cultured cells was isolated in TRIzol L and

442 purified using Qiagen RNeasy Mini Kit per manufacturer’s protocols. Agilent Technologies

443 2100 Bioanalyzer was used to assess the RNA quality. RNA libraries were prepared and

444 multiplexed using Illumina TruSeq RNA Library Preparation Kit v2 (non-stranded and

445 poly-A selection) and 10 nM of cDNA was used as the input for high-throughput

446 sequencing via Illumina’s HiSeq 2500 platform, producing 50 bp paired end reads. Paired

447 end sequencing reads were then quality-checked and trimmed where appropriate using

448 Trimmomatic 0.3649 and the resultant filtered reads reads were mapped to the human

449 reference genome GRCh38 using STAR aligner50. Aligned bam files were subsequently

450 sorted and raw transcript counts were quantified using HTSeq51 to produce counts

451 matrices based off the human reference transcriptome for GRCh38. Previously published

452 RNAseq data for hPSC derived iBMECs and primary BMECs were downloaded from the

453 Gene Expression Omnibus (GEO, GSE97575)18 along with previously published data for

454 HUVECs and LSECs (GEO, GSE40291)52 and were processed in the same way as for

455 the bulk RNAseq libraries prepared for the experiment. Raw counts were processed in R

456 package edgeR53 and a Log2 counts per million matrix was calculated after filtering and

457 quality control. A heatmap displaying the 500 most variable genes was subsequently

458 formulated with the distance between both rows and columns being calculated via

459 Euclidean distance and with dendrograms computed and reordered based on row and

460 column means.

19 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

461

462 In vivo tubulogenesis assay

463 H1, H6, and IMR90-4 derived neuroendothelial iBMECs at day 16 of differentiation

464 and primary HUVECs were resuspended in Matrigel (Corning, 254277) at

465 500,000cells/300μL and injected subcutaneously into immunocompromised mice (NSG-

466 SGM3; Jax, 013062). Mice were also injected with an empty Matrigel control. After 5

467 days, mice were sacrificed following institutional guidelines and solidified Matrigel plugs

468 were excised and fixed using 4% paraformaldehyde (Alfa Aesar, AA43368-9M). Plugs

469 were cryosectioned, stained using CD144 (BV9, Biolegend), Agglutinin 1 (UEA1, Vector

470 Labs), and EPCAM (9C4, Biolegend). Slides were fixed using Fluoroshield with DAPI

471 (Sigma-Aldrich, F6057) and imaging was performed using a Zeiss 710 META confocal

472 microscope. Quantification analysis of images were conducted using Angiotool54.

473

474 Virus production

475 Human ETV2, FLI1, and ERG lentiviral particles were generated and tittered as

476 described in Ginsberg et al., with the following modifications. 293T Lenti-X cells (passage

477 8–10; subconfluent) were used to produce lentiviral particles using Lenti-X packaging

478 single-shot system, following the manufacturer’s protocol. Filter through a 0.45μm PES

479 filter attached to a 30ml syringe and use Lenti- 358 X concentrator following

480 manufacturer’s protocol. Resuspend the concentrate from a 100mm plate of 293T cells in

481 100 μl of lentivirus storage buffer and store at −80°C for up to 1 year.

482

483 Trans-Endothelial Electrical Resistance Measurement

20 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

484 Cells were plated on poly-D-lysine and Collagen IV-coated gold electroarray 96-

485 well plates (Applied Biophysics) and grown in endothelial cell media supplemented with

486 growth factors and 10% FBS media (Cell Biologics) to maximum resistance. Cells were

487 switched to 1% FBS and the resistance recorded every 30 min for 48 hr using the ECIS

488 Z-θ system (Applied Biophysics). Resistance curves were generated using GraphPad

489 Prism software. Areas under the curve were quantified for the final 48 hr after cells

490 reached maximum resistance using GraphPad Prism software.

491

492 Animal experiments

493 12-15 weeks old immunodeficient mice, specifically NSG-SGM3, Jax, 013062,

494 were used in this study. All animal experiments were performed under the approval of

495 Weill Cornell Medicine Institutional Animal Care and Use Committee (#2009-0061).

496 497 498 Author contributions

499 TL, DA, SR and RL designed and conducted various aspects of the experiments

500 wrote and edited the manuscript. DR interpreted the single cell and bulk RNA seq. AS,

501 AnS, TM, DHN, JX provided technical assistance. KS, HAF, ZR helped to formulate the

502 hypothesis, interpret the results, and edited the manuscript. SR, DA & RL directed and

503 supervised all aspects of the study.

504

505 Acknowledgements

506 We thank Raul Chavez for the technical assistance and the staff in the Genetic

507 Resources Core Facility in Weill Cornell Medicine. We thank Tyler Cutforth at Columbia

508 University Irving Medical Center for his editorial input in the manuscript. D.A. is supported

21 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

509 by the National Institutes of Health (#R01MH112849 #R01NS107344), the Leducq

510 Foundation #15CVD-02 and unrestricted gifts from John Castle, Newport Equity LLC and

511 PANDAS Network to the Department of Neurology, Division of Cerebrovascular Diseases

512 and Stroke at CUIMC (D.A.). H.A.F is supported by the National Institute of Health

513 (NIH DP1CA228040), AnS is supported by funding from the Henry and Marylin Taub

514 Foundation. T.L., K.S., S.R. and R.L. are supported by the Ansary Stem Cell Institute,

515 Starr Foundation TRI-Institution Stem Cell Initiative grants (Stem Cell Derivation Core,

516 #2013-032, #2014-023, #2016-013), Qatar National Priorities Research Program (NPRP

517 8-1898-3-392), the Empire State Stem Cell Board and New York State Department of

518 Health grants (NYSTEM: C026438, C026878, C029156) and by grants from the NIH R01

519 (DK095039, HL119872, HL115128, RC2DK114777, HL128158, HL139056-01A1,

520 U01AI138329).

521 522

22 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

523 Figure Legends 524 525 Figure 1. Phenotypic comparison of hPSC-derived cells to bona fide endothelium

526 a. Diagrams for differentiation of hPSCs to neuroendothelial iBMECs and

527 mesoendothelial iECs.

528 b. Phase contrast microscopy of primary endothelial cells (HUVEC and BMEC) and

529 neuro-endothelial iBMECs derived from H1, H6, and IMR90-4 cell lines at day 11 of

530 differentiation. Representative pictures of n=5 biological replicates, scale bars, 200µm.

531 c. Confocal microscopy of HUVECs, BMECs, and NE-EpiCs; H1, H6, and IMR90-4-

532 derived iBMECs at day 16 of differentiation; H1, H6, and IMR90-4 derived iECs at day 12

533 of differentiation; DAPI (Blue), PECAM1 (Green), and VECAD (Red). Representative

534 pictures of n=5 biological replicates, scale bars, 200µm.

535 d. Representative flow cytometry plot assaying VECAD+ PECAM1+ phenotypically

536 marked cells. n=5 biological replicates.

537 e. Quantification of VECAD+PECAM1+ populations based on FACS analysis.

538

539 Figure 2. The neuroendothelial iBMECs transcriptome is not endothelial

540 a. scRNA expression profiles displayed on a t-distributed stochastic neighbor embedding

541 (t-SNE) plot illustrates multiple unique clusters (a-i’) of cells separated based on top 400

542 highest differentially expressed genes.

543 b. Dendrogram generated using hierarchical clustering of unique cell clusters represented

544 in Fig. 2a showing a clear distinction between iBMECs and all other endothelial cell

545 clusters.

546

23 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

547 Figure 3. Neuroendothelial iBMECs lack angiocrine factors

548 Dot plot emphasizing large difference in expression of vascular genes between

549 Endothelial cell population and hPSC-derived iBMECs. For this and all subsequent dot

550 plots, cell clusters and sample numbers are taken from Fig. 2a. Dot size indicates

551 percentage of cells in populations expressing the gene whereas intensity of color dictates

552 expression level.

553

554 Figure 4. Neuroendothelial iBMECs express genes characteristic of epithelial cells

555 a. Dot plot showing a clear bias in the expression of epithelial-associate genes towards

556 iBMEC samples compared to all endothelial control samples and mesodermal

557 differentiation derived iECs. Highlighted genes correspond to subset of epithelial genes

558 highly expressed in iBMEC cell clusters compared to human ECs.

559 b. Confocal microscopy of HUVECs, BMECs, and NE-EpiCs; H1, H6, and IMR90-4

560 derived iBMECs at day 16 of differentiation; H1, H6, and IMR90-4 derived iECs at day 12

561 of differentiation. DAPI (Blue) and EpCAM (Purple). Representative pictures of n=5

562 biological replicates, scale bars, 200µm.

563 c. Confocal microscopy showing NE-EpiCs and H1 and IMR90-4 derived iBMECs. DAPI

564 (Blue), EpCAM (Purple), TTR (Green). n=5 biological replications, scale bars, 200 µm.

565 d. Representative flow cytometry plot assaying VECAD+EpCAM+ phenotypically marked

566 cells. n=5 biological replicates.

567

568 Figure 5. Global RNA sequencing heat maps

569 a. Heat map of top 500 most variable genes between all cell samples both previously

570 published or acquired in house.

24 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

571 b. Heat map of endothelial and epithelial genes derived from the single cell dataset

572 expressed by all cell samples both previously published or acquired in house.

573

574 Figure 6. Generating phenotypic ECs from iBMECs using ETV2, ERG and FLI1

575 a. Schematic for the generation of EEF-iBMECs highlighting the induction of ETV2, ERG,

576 FLI1 at day 6 of the neuro-endothelial codifferentiation.

577 b. Representative flow cytometry plot assaying VECAD+PECAM1+ phenotypically marked

578 cells as well as EpCAM expression. Top panels; Untransduced control iBMEC sample,

579 Middle panels; iBMEC transduced with EEF factors at day 6 of neuro-endothelial

580 codifferentiation, Bottom panels; EEF-iBMECs at day 27 (post day 11 sorting). n=5

581 biological replicates.

582 c. Dot plot showing expression levels and prevalence of vascular genes between primary

583 ECs (HUVECs, BMECs), iECs, and IMR90-4 derived cells (iBMECs, EEF-iBMECs).

584 d. Left; scRNA expression profiles displayed on a t-distributed stochastic neighbor

585 embedding (t-SNE) plot illustrates multiple unique clusters (a-d’) of cells separated based

586 on top 500 highest differentially expressed genes. Right; Dendrogram generated using

587 hierarchical clustering of unique cell clusters.

588

589 Extended Data Figure 1. Phenotypic and functional comparative assay of iBMECs

590 a. Phase contrast microscopy of primary endothelial cells (HUVEC and BMEC) and

591 neuroendothelial iBMECs derived from H1, H6, and IMR90-4 cell lines at day 11 of

592 differentiation. Confocal microscopy of H1, H6, and IMR90-4 derived iBMECs at day 11 of

593 differentiation; DAPI (Blue), PECAM1 (Green), VECAD (Red). n=5 biological replicates,

594 scale bars, 200µm.

25 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

595 b. Confocal microscopy of H1, H6, and IMR90-4 derived iBMECs at day 11 and day 16 of

596 differentiation; H1, H6, and IMR90-4 derived iECs at day 12 of differentiation; HUVECs,

597 BMECs, and NE-EpiCs. DAPI (Blue), PECAM1 (Green), and VECAD (Red).

598 Representative pictures of n=5 biological replicates, scale bars, 50µm.

599 c. Schematic of in vivo vessel formation assay in NSG-SGM3 mice. 500,000 iBMEC or

600 HUVEC in Matrigel were injected subcutaneously into 10 mice per hPSC cell line and

601 excised 5 days later for microscopy analysis.

602 d. In vivo tubulogenesis assay of HUVEC and H1, H6, and IMR90-4 derived iBMECs at

603 day 16 of differentiation in Matrigel plugs. DAPI (Blue), EpCAM (Purple), Agglutinin

604 (Green), and VECAD (Red). White arrows in iBMEC panels identifying groups of iBMECs

605 forming EpCAM+ cell clusters.

606 e. Angiotool quantifications of in vivo tubulogenesis assay, Top; total number of vessel

607 junctions in each sample, Bottom; total length of all vessels in each sample. n=3

608 biological replicates, 10 mice per cell line tested.

609

610 Extended Data Figure 2. Single-cell RNA sequencing samples and quality control

611 a. scRNA expression profiles displayed on a t-distributed stochastic neighbor embedding

612 (t-SNE) plot illustrates 18 distinct cell samples used in primary analysis of endothelial

613 cells (HUVEC and BMEC), meso-endothelial iECs at day 12, and neuro-endothelial

614 iBMECs at days 11 and 16.

615 b. Violin plot describing total number of cells per sample that passed quality control and

616 therefore were used in the study.

617

618

26 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

619 Extended Data Figure 3. Relative identity of IMR90-4-derived EEF-iBMECs

620 a. scRNA expression profiles displayed on a t-distributed stochastic neighbor embedding

621 (t-SNE) plot illustrates endothelial cells (HUVEC and BMEC), IMR90-4 derived iBMECs at

622 days 11 and 16, and IMR90-4 derived EEF-iBMECs at day 11.

623 b. Confocal microscopy of EEF-iBMECs at days 11 and 27 (post day 11 FACS) showing

624 EpCAM (purple), PECAM1 (Green), and VECAD (red). White arrows showing regions of

625 junctional VECAD being restored in EEF-iBMECs. n=5 biological replicates, scale bars,

626 200µm.

627

628 Extended Data Figure 4. Expression of VEGF and NOTCH pathway genes across all

629 samples

630 a. Dot plot comparing expression of Notch pathway genes (KEGG, map04330) across all

631 single cell RNA sequencing derived cell clusters.

632 b. Dot plot comparing expression of VEGF pathway genes (KEGG, map04370) across all

633 single cell RNA sequencing derived cell clusters.

634

635 Extended Data Figure 5. Expression of WNT pathway genes across all samples

636 Dot plot comparing expression of WNT pathway genes (KEGG, map04310) across all

637 single cell RNA sequencing derived cell clusters.

638

639

640

641

27 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

642 Extended Data Figure 6. Expression of tight junction pathway genes across all

643 samples

644 Dot plot comparing expression of tight junction pathway genes (KEGG, map04530)

645 across all single cell RNA sequencing derived cell clusters. Genes highlighted show

646 difference in Claudin expression profiles of iBMECs compared to iECs and primary ECs.

647

648 Extended Data Figure 7. Distribution of endothelial and epithelial gene expression

649 a. scRNA expression profiles displayed on a t-distributed stochastic neighbor embedding

650 (t-SNE) plots showing distribution of cells expression epithelial genes.

651 b. scRNA expression profiles displayed on a t-distributed stochastic neighbor embedding

652 (t-SNE) plots showing distribution of cells expression endothelial genes.

653

654 Extended Data Figure 8. Immunofluorescence staining for tight junction proteins

655 and TEER analysis of iBMECs generated with the neuroendothelial protocol.

656 a-f. Immunofluorescence of CLAUDIN-5 (Red), OCCLUDIN (Red) and ZO-1 (Green) in

657 IMR90-4 derived iBMECs using the published antibodies and DAPI (Blue). These

658 antibodies recognize proteins at cell-cell junctions. Scale bars, 40µm.

659 g, h. Measurements of Trans Endothelial Electrical Resistance and Capacitance of

660 iBMECs over time. The cells were plated in ECIS plates after 8 days of neuroendothelial

661 differentiation in a Collagen IV and Fibronectin substrate.

662

28 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

663 References 664 665 1 Liebner, S. et al. Functional morphology of the blood-brain barrier in health and disease. 666 Acta Neuropathol 135, 311-336, doi:10.1007/s00401-018-1815-1 (2018). 667 2 Daneman, R. & Engelhardt, B. Brain barriers in health and disease. Neurobiol Dis 107, 1-3, 668 doi:10.1016/j.nbd.2017.05.008 (2017). 669 3 Chow, B. W. & Gu, C. The molecular constituents of the blood-brain barrier. Trends 670 Neurosci 38, 598-608, doi:10.1016/j.tins.2015.08.003 (2015). 671 4 Zhao, Z., Nelson, A. R., Betsholtz, C. & Zlokovic, B. V. Establishment and Dysfunction of the 672 Blood-Brain Barrier. Cell 163, 1064-1078, doi:10.1016/j.cell.2015.10.067 (2015). 673 5 Deli, M. A., Ábrahám, C. S., Kataoka, Y. & Niwa, M. Permeability Studies on In Vitro Blood– 674 Brain Barrier Models:Physiology, Pathology, and Pharmacology. Cellular and Molecular 675 Neurobiology 25, 59-127, doi:10.1007/s10571-004-1377-8 (2005). 676 6 Pardridge, W. M. The Blood-Brain Barrier: Bottleneck in Brain Drug Development. NeuroRx 677 2, 3-14 (2005). 678 7 Weksler, B. B. et al. Blood-brain barrier-specific properties of a human adult brain 679 endothelial cell line. The FASEB Journal 19, 1872-1874, doi:10.1096/fj.04-3458fje (2005). 680 8 Cecchelli, R. et al. Modelling of the blood–brain barrier in drug discovery and 681 development. Nature Reviews Drug Discovery 6, 650, doi:10.1038/nrd2368 (2007). 682 9 Zlokovic, B. V. The Blood-Brain Barrier in Health and Chronic Neurodegenerative 683 Disorders. Neuron 57, 178-201, doi:10.1016/j.neuron.2008.01.003 (2008). 684 10 Lippmann, E. S., Al-Ahmad, A., Azarin, S. M., Palecek, S. P. & Shusta, E. V. A retinoic acid- 685 enhanced, multicellular human blood-brain barrier model derived from stem cell sources. 686 Scientific Reports 4, 4160, doi:10.1038/srep04160 (2014). 687 11 Clark, P. A. et al. Analysis of Cancer-targeting Alkylphosphocholine Analog Permeability 688 Characteristics Using a Human Induced Pluripotent Stem Cell Blood-Brain Barrier Model. 689 Molecular pharmaceutics 13, 3341-3349, doi:10.1021/acs.molpharmaceut.6b00441 690 (2016). 691 12 Helms, H. C. et al. In vitro models of the blood–brain barrier: An overview of commonly 692 used brain endothelial cell culture models and guidelines for their use. Journal of Cerebral 693 Blood Flow & Metabolism 36, 862-890, doi:10.1177/0271678X16630991 (2016). 694 13 Stebbins, M. J. et al. Differentiation and Characterization of Human Pluripotent Stem Cell- 695 Derived Brain Microvascular Endothelial Cells. Methods (San Diego, Calif.) 101, 93-102, 696 doi:10.1016/j.ymeth.2015.10.016 (2016). 697 14 Canfield, S. G. et al. An Isogenic Blood-Brain Barrier Model Comprising Brain Endothelial 698 Cells, Astrocytes and Neurons Derived from Human Induced Pluripotent Stem Cells. 699 Journal of neurochemistry 140, 874-888, doi:10.1111/jnc.13923 (2017). 700 15 Hollmann, E. K. et al. Accelerated differentiation of human induced pluripotent stem cells 701 to blood–brain barrier endothelial cells. Fluids and Barriers of the CNS 14, 9, 702 doi:10.1186/s12987-017-0059-0 (2017). 703 16 Kim, B. J. et al. Modeling Group B Streptococcus and Blood-Brain Barrier Interaction by 704 Using Induced Pluripotent Stem Cell-Derived Brain Endothelial Cells. mSphere 2, e00398- 705 00317, doi:10.1128/mSphere.00398-17 (2017).

29 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

706 17 Lim, R. G. et al. Huntington’s Disease iPSC-Derived Brain Microvascular Endothelial Cells 707 Reveal WNT-Mediated Angiogenic and Blood-Brain Barrier Deficits. Cell reports 19, 1365- 708 1377, doi:10.1016/j.celrep.2017.04.021 (2017). 709 18 Qian, T. et al. Directed differentiation of human pluripotent stem cells to blood-brain 710 barrier endothelial cells. Science Advances 3, e1701679, doi:10.1126/sciadv.1701679 711 (2017). 712 19 Stebbins Matthew, J. et al. Activation of RARα, RARγ, or RXRα Increases Barrier Tightness 713 in Human Induced Pluripotent Stem Cell-Derived Brain Endothelial Cells. Biotechnology 714 Journal 13, 1700093, doi:10.1002/biot.201700093 (2017). 715 20 Vatine, G. D. et al. Modeling Psychomotor Retardation using iPSCs from MCT8-Deficient 716 Patients Indicates a Prominent Role for the Blood-Brain Barrier. Cell Stem Cell 20, 831- 717 843.e835, doi:10.1016/j.stem.2017.04.002 (2017). 718 21 Al-Ahmad, A. J., Patel, R., Palecek, S. P. & Shusta, E. V. Hyaluronan impairs the barrier 719 integrity of brain microvascular endothelial cells through a CD44-dependent pathway. 720 Journal of Cerebral Blood Flow & Metabolism, 0271678X18767748, 721 doi:10.1177/0271678X18767748 (2018). 722 22 Sances, S. et al. Human iPSC-Derived Endothelial Cells and Microengineered Organ-Chip 723 Enhance Neuronal Development. Stem Cell Reports 10, 1222-1236, 724 doi:10.1016/j.stemcr.2018.02.012 (2018). 725 23 Levenberg, S., Golub, J. S., Amit, M., Itskovitz-Eldor, J. & Langer, R. Endothelial cells 726 derived from human embryonic stem cells. Proceedings of the National Academy of 727 Sciences 99, 4391 (2002). 728 24 Ying, Q.-L., Stavridis, M., Griffiths, D., Li, M. & Smith, A. Conversion of embryonic stem 729 cells into neuroectodermal precursors in adherent monoculture. Nature Biotechnology 21, 730 183, doi:10.1038/nbt780 731 https://www.nature.com/articles/nbt780#supplementary-information (2003). 732 25 Vodyanik, M. A., Bork, J. A., Thomson, J. A. & Slukvin, I. I. Human embryonic stem cell– 733 derived CD34⁺ cells: efficient production in the coculture with 734 OP9 stromal cells and analysis of lymphohematopoietic potential. Blood 105, 617 (2005). 735 26 Takahashi, K. et al. Induction of Pluripotent Stem Cells from Adult Human Fibroblasts by 736 Defined Factors. Cell 131, 861-872, doi:https://doi.org/10.1016/j.cell.2007.11.019 (2007). 737 27 Yu, J. et al. Induced Pluripotent Stem Cell Lines Derived from Human Somatic Cells. 738 Science 318, 1917-1920, doi:10.1126/science.1151526 (2007). 739 28 Lippmann, E. S., Al-Ahmad, A., Palecek, S. P. & Shusta, E. V. Modeling the blood–brain 740 barrier using stem cell sources. Fluids and Barriers of the CNS 10, 2-2, doi:10.1186/2045- 741 8118-10-2 (2013). 742 29 Lippmann, E. S. et al. Derivation of blood-brain barrier endothelial cells from human 743 pluripotent stem cells. Nature Biotechnology 30, 783, doi:10.1038/nbt.2247 744 https://www.nature.com/articles/nbt.2247#supplementary-information (2012). 745 30 Mizee, M. R. et al. Retinoic Acid Induces Blood–Brain Barrier Development. The Journal of 746 Neuroscience 33, 1660 (2013). 747 31 James, D. et al. Expansion and maintenance of human embryonic stem cell–derived 748 endothelial cells by TGFβ inhibition is Id1 dependent. Nature Biotechnology 28, 161, 749 doi:10.1038/nbt.1605 750 https://www.nature.com/articles/nbt.1605#supplementary-information (2010).

30 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

751 32 Lun, M. P. et al. Spatially Heterogeneous Choroid Plexus Transcriptomes Encode Positional 752 Identity and Contribute to Regional CSF Production. The Journal of Neuroscience 35, 4903, 753 doi:10.1523/JNEUROSCI.3081-14.2015 (2015). 754 33 Janssen, S. F. et al. Gene Expression and Functional Annotation of the Human and Mouse 755 Choroid Plexus Epithelium. PLOS ONE 8, e83345, doi:10.1371/journal.pone.0083345 756 (2014). 757 34 Ginsberg, M., Schachterle, W., Shido, K. & Rafii, S. Direct conversion of human amniotic 758 cells into endothelial cells without transitioning through a pluripotent state. Nature 759 Protocols 10, 1975, doi:10.1038/nprot.2015.126 760 https://www.nature.com/articles/nprot.2015.126#supplementary-information (2015). 761 35 Sugimura, R. et al. Haematopoietic stem and progenitor cells from human pluripotent 762 stem cells. Nature 545, 432 (2017). 763 36 Lis, R. et al. Conversion of adult endothelium to immunocompetent haematopoietic stem 764 cells. Nature 545, 439 (2017). 765 37 Butler, J. M. et al. Development of a vascular niche platform for expansion of repopulating 766 human cord blood stem and progenitor cells. Blood 120, 1344, doi:10.1182/blood-2011- 767 12-398115 (2012). 768 38 Butler, J. M. et al. Endothelial Cells Are Essential for the Self-Renewal and Repopulation of 769 Notch-Dependent Hematopoietic Stem Cells. Cell Stem Cell 6, 251-264, 770 doi:https://doi.org/10.1016/j.stem.2010.02.001 (2010). 771 39 Seandel, M. et al. Generation of a functional and durable vascular niche by the adenoviral 772 E4ORF1 gene.(CELL BIOLOGY)(Author abstract). Proceedings of the National Academy of 773 Sciences of the United States 105, 19288, doi:10.1073/pnas.0805980105 (2008). 774 40 Rafii, S. et al. Human bone marrow microvascular endothelial cells support long-term 775 proliferation and differentiation of myeloid and megakaryocytic progenitors. Blood 86, 776 3353 (1995). 777 41 Stenman, J. M. et al. Canonical Wnt signaling regulates organ-specific assembly and 778 differentiation of CNS vasculature. Science 322, 1247-1250, doi:10.1126/science.1164594 779 (2008). 780 42 Daneman, R. et al. Wnt/beta-catenin signaling is required for CNS, but not non-CNS, 781 angiogenesis. Proc Natl Acad Sci U S A 106, 641-646, doi:10.1073/pnas.0805165106 782 (2009). 783 43 Liebner, S. et al. Wnt/beta-catenin signaling controls development of the blood-brain 784 barrier. J Cell Biol 183, 409-417, doi:10.1083/jcb.200806024 (2008). 785 44 Zhou, Y. et al. Canonical WNT signaling components in vascular development and barrier 786 formation. J Clin Invest 124, 3825-3846, doi:10.1172/JCI76431 (2014). 787 45 Günzel, D. & Yu, A. S. L. Claudins and the modulation of tight junction permeability. 788 Physiological reviews 93, 525-569, doi:10.1152/physrev.00019.2012 (2013). 789 46 Praetorius, J. & Damkier, H. H. Transport across the choroid plexus epithelium. American 790 Journal of Physiology-Cell Physiology 312, C673-C686, doi:10.1152/ajpcell.00041.2017 791 (2017). 792 47 Strazielle, N., Creidy, R., Malcus, C., Boucraut, J. & Ghersi-Egea, J.-F. T-Lymphocytes Traffic 793 into the Brain across the Blood-CSF Barrier: Evidence Using a Reconstituted Choroid 794 Plexus Epithelium. PLoS ONE 11, 1-17, doi:10.1371/journal.pone.0150945 (2016).

31 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

795 48 Liddelow, S. A. et al. Cellular Specificity of the Blood–CSF Barrier for Albumin Transfer 796 across the Choroid Plexus Epithelium. PLOS ONE 9, e106592, 797 doi:10.1371/journal.pone.0106592 (2014). 798 49 Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina 799 sequence data. Bioinformatics (Oxford, England) 30, 2114-2120, 800 doi:10.1093/bioinformatics/btu170 (2014). 801 50 Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics (Oxford, England) 802 29, 15-21, doi:10.1093/bioinformatics/bts635 (2013). 803 51 Anders, S., Pyl, P. T. & Huber, W. HTSeq--a Python framework to work with high- 804 throughput sequencing data. Bioinformatics (Oxford, England) 31, 166-169, 805 doi:10.1093/bioinformatics/btu638 (2015). 806 52 Ginsberg, M. et al. Efficient direct reprogramming of mature amniotic cells into 807 endothelial cells by ETS factors and TGFβ suppression. Cell 151, 559-575, 808 doi:10.1016/j.cell.2012.09.032 (2012). 809 53 Robinson, M. D., McCarthy, D. J. & Smyth, G. K. edgeR: a Bioconductor package for 810 differential expression analysis of digital gene expression data. Bioinformatics (Oxford, 811 England) 26, 139-140, doi:10.1093/bioinformatics/btp616 (2010). 812 54 Zudaire, E., Gambardella, L., Kurcz, C. & Vermeren, S. A Computational Tool for 813 Quantitative Analysis of Vascular Networks. PLOS ONE 6, e27385, 814 doi:10.1371/journal.pone.0027385 (2011). 815

32 Figure 1 H1 H1 H6 H6 iEC iBMEC IMR90-4 a IMR90-4 LN4 Day 8 Day 4 Day 12 + + Day -3 0 6 Subculture 11 Day 16 Day -1 0 1 2 Adherent Culture 7 Sort on VECad PECAM1

Unconditioned BMP4 mTeSR Endothelial Cell Media Media Activin A bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprintFGF-2 (which was not certifiedMatrigel by peer review) is theCollagen author/funder. IV / Fibronectin All rights reserved. No reuse allowed without permission. VEGF-A SB431542 b H1 iBMEC c HUVEC BMEC NE-EpiC Day 16 Control

HUVEC

H6 iBMEC VEcad H1 iBMEC H6 iBMEC IMR90-4 iBMEC Day 16 Day 16 Day 16 Day 16 Endothelial Neuro

BMEC PECAM1 IMR90-4 iBMEC H1 iEC H6 iEC IMR90-4 iEC Day 16 Day 12 Day 12 Day 12 Endothelial Meso DAPI

d HUVEC BMEC NE-EpiC

H1 iBMEC H6 iBMEC IMR90-4 iBMEC H1 iEC H6 iEC IMR90-4 iEC Day 16 DayDay 16 16 Day 16 Day 12 Day 12 Day 12 PECAM1- BV421 VEcad-AF647 e p<0.05 f p<0.05 100 100 + + 75 50 75 VE ca d VE ca d + 25 + 50 1.0 15

PE CAM 1 PE CAM 1 10 0.5

%of %of %of 5 0.0 0

H1 H6 H1 H6 H1 H6 BMEC CP-Epi BMEC CP-Epi HUVEC IMR90-4 IMR90-4 HUVEC IMR90-4 iBMEC Day 11 iBMEC Day 16 iEC Day 12 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Figure 2 a b Hierarchical clustering

a 6 HUVEC 49 50 b c 1 BMEC 42 d 54 i 50 f g e 6 HUVEC h f 55 18 LN4 iEC Day 21 62 37 g 67 h 1 BMEC b 17 a i LN4 iEC Day 17 25 j k j 16 LN4 iEC Day 27 56 43 k 7 IMR90 iEC Day 12 l 48 36 c m n m 16 LN4 iEC Day 12 57 66 n 7 IMR90 iEC Day 12 l o 1 BMEC d p p 5 H6 iBMEC Day 16 e’ q 14 40 u q c’ IMR90 iBMEC Day 16 38 0 h’ r 10 IMR90 iBMEC Day 16 tSNE_2 s 8 IMR90 iBMEC Day 11 44 b’ g’ t 53 39 t u 9 IMR90 iBMEC Day 11 63 a’ i’ f’ s v 3 H1 iBMEC Day 16 64 w 41 x 46 11 IMR90 iBMEC Day 16 65 z v y z −25 2 H1 iBMEC Day 11 60 a’ 45 51 y x w b’ 8 IMR90 iBMEC Day 11 r c’ 5 H6 iBMEC Day 16 61 14 IMR90 iBMEC Day 16 69 d’ 47 e’ 14 f’ IMR90 iBMEC Day 16 58 o 5 H6 iBMEC Day 16 68 g’ 52 −50 h’ 12 IMR90 iBMEC Day 11 i’ 4 H6 iBMEC Day 11 59 −50 −25 0 25 Gene Sample 5 4 3 2 1 tSNE_1 Cluster Name Distance (AU) bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Figure 3 Sample identity 1 BMEC TAL1 2 SOX18 H1 iBMEC PROCR 3 SOX17 4 PLVAP H6 iBMEC ROBO4 5 PECAM1 6 HUVEC RASIP1 NOTCH4 7 IMR90 iEC PTPRB 8 NOS3 MFNG 9 MEF2C 10 ETS2 11 KIT IMR90 ETS1 12 KDR iBMEC ESAM 13 JAG2 14 ERG JAG1 15 ENG 16 IGFBP7 ELK3 17 LN4 iEC IGFBP4 18 ELF1 ID3 EGFL7 pct.exp ICAM2 DLL4 0.00 GJA4 CLDN5 0.25 GATA2 0.50 CDH5 FLT4 0.75 CD34 1.00 FLT1 BMP6 FLI1 avg.exp. BMP4 FGF2 scale BMP2 ETV6 2 APLNR APLN 1 ANGPT2 0 a b c d e f g h i j k l m n o p q r s t u v w x y z a’ b’ c’ d’ e’ f’ g’ h’ i’ −1 Gene Cluster #Sample 6 1 6 18 1 17 16 7 16 7 1 5 14 10 8 13 9 3 11 2 5 14 8 5 14 4 12 15 Figure 4 a SPP1 Sample identity FREM2 1 BMEC 2 LIX1 3 H1 iBMEC 4 CXCL14 H6 iBMEC 5 bioRxiv preprintTAGLN doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not 6 HUVEC certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 7 IMR90 iEC SCD 8 9 NPPB 10 SERPINF1 11 IMR90 12 CLDN8 13 iBMEC 14 E2F5 15 SLC12A2 16 17 LN4 iEC KCNK1 18 AQP1 pct.exp TTR 0.00 KRT19 0.25 0.50 KRT18 0.75 KRT17 1.00 KRT8 avg.exp. KRT7 scale EPCAM 2 Gene a b c d e f g h i j k l m n o p q r s t u v w x y z a’ b’ c’ d’ e’ f’ g’ h’ i’ 1 Cluster 0 #Sample 6 1 6 18 1 17 16 7 16 7 1 5 14 10 8 13 9 3 11 2 5 14 8 5 14 4 12 −1 15

b HUVEC BMEC NE-EpiC c Control NE-EpiC TTR H1 iBMEC H6 iBMEC IMR90-4 iBMEC Day 16 Day 16 Day 16 Endothelial Neuro

EPCAM Neuro-Endothelial H1 iBMEC IMR90-4 iBMEC EPCAM Day 11 Day 11 H1 iEC H6 iEC IMR90-4 iEC Endothelial DAPI Day 12 Day 12 Day 12 Meso DAPI

d H1 iEC H6 iEC IMR90-4 iEC Day 12 Day 12 Day 12 HUVEC BMEC NE-EpiC

H1 iBMEC H6 iBMEC IMR90-4 iBMEC Day 16 Day 16 Day 16 AF647 VEcad- EPCAM-AF488 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Color Key Figure 5 and Histogram

a Top 500 most variable genes b scRNAseq curated genes 400 600 Count 200 0

−2 −1 0 1 2 Row Z−Score

Neuro-Endothelial adult Endothelial cells adult Endothelial cells Neuro-Endothelial ELK3 ETS1 JAG1 ID3 ELF1 SLC12A2 ETS2 ETV6 IGFBP4 IGFBP7 KRT7 KRT19 Cluster “p-i’ ” KRT8 SCD KRT18 enriched genes TAGLN SERPINF1 SPP1 KCNK1 NPPB EPCAM FREM2 CXCL14 FLI1 PTPRB FLT1 MEF2C BMP2 KDR PROCR EGFL7 ENG CDH5 BMP4 GATA2 JAG2 FGF2 E2F5 KIT ANGPT2 BMP6 NOS3 Cluster “a-e” NOTCH4 FLT4 RASIP1 enriched genes ESAM MFNG ERG AQP1 DLL4 GJA4 PECAM1 ROBO4 TAL1 SOX18 ICAM2 CLDN5 SOX17 CD34 PLVAP iBMEC iBMEC iBMEC iBMEC iBMEC LSEC 11 LSEC 11 HUVEC HUVEC HUVEC HUVEC from Lu et al. H6 day 11 H1 day 11 H6 day 16 H1 day 16 primary BMEC primary BMEC H6 day 11 H1 day 11 H6 day 16 H1 day 16 primary BMEC primary BMEC from Ginsberg et al. primary NE-EpiC IMR90 day IMR90 day 16 primary NE-EpiC IMR90 day IMR90 day 16 from Qian et al. Figure 6

ANGPT2 NOTCH4PECAM1 a IGFBP4IGFBP7 PROCR RASIP1ROBO4 APLNR CLDN5 EGFL7 MEF2C PTPRB ESAM GATA2 ICAM2 MFNG PLVAP SOX17SOX18 IMR90-4 iBMECs c APLN BMP2BMP4BMP6CD34CDH5 DLL4 ELK3 ETS1ETS2ETV6FGF2 GJA4 JAG1JAG2 NOS3 ELF1 TAL1 ENGERG FLI1FLT1FLT4 KDR ID3 KIT +ETV2, ERG, FLI1 10 EEF-iBMECs 11 Day -3 Day 0 Day 6 Day 11 Day 16

bioRxivUnconditionned preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not mTeSR Endothelial Cell Media Media certified by peer review) is the author/funder. All rights2 reserved. No reuse allowed without permission.

Matrigel Collagen IV / Fibronectin b 1 Day 11 Untransduced iBMEC

3 pct.exp 12 0.00 9 0.25 0.50 8 0.75 1.00 Day 11 13 EEF-iBMEC avg.exp. 8 scale 7 2 8 1 0 7 −1 8

7 Day 27 EEF-iBMEC 6

BV421 4

5 10 PECAM1- 11 VEcad-AF647 EPCAM-AF488 1

d 10 IMR90-4 EFF-iBMECs D11 a 11 IMR90-4 EEF-iBMECs D11 rerun b d 50 c 40 2 Primary HUVEC 57 d c e 37 b Lorem ipsum f 49 g 41 1 Primary BMEC 50 33 e h 31 i 25 u j 38 3 IMR90-4 iECs 39 q k l 12 IMR90-4 iBMEC D16 v a g m 9 IMR90-4 iBMEC D6 ECM 43 x s n 8 IMR90-4 iBMEC D11 48 r o 13 IMR90-4 iBMEC D6 UM 56 i l 32 8 IMR90-4 iBMEC D11 tSNE_2 b’ p 0 q 7 IMR90-4 iBMEC D16 51 t p 58 35 z r 8 IMR90-4 iBMEC D11 n s 7 IMR90-4 iBMEC D16 45 f 53 j y t 8 IMR90-4 iBMEC D11 a’ u 52 7 IMR90-4 iBMEC D16 44 d’ v 54 k h w 6 IMR90-4 iBMEC D16 34 −25 w x 46 4 IMR90-4 iBMEC D11 y 47 o z 55 c’ 5 IMR90-4 IBMEC D11 a’ 59 10 IMR90-4 EFF-iBMECs D11 m b’ 11 IMR90-4 EEF-iBMECs D11 rerun c’ 36 1 Primary BMEC 42 −50 d’ −30 0 30 tSNE_1 Extended Data Figure 1 a HUVEC Neuro-Endothelial

bioRxiv preprint doi: https://doi.org/10.1101/699173H1 ;iBMEC this version posted July 11, 2019.H6 The iBMEC copyright holder for this preprintIMR90-4 (which was iBMEC not certified by peer review)Day is the author/funder.11 All rights reserved.Day No reuse11 allowed without permission.Day 11

BMEC Day 11 Day 11 Day 11 VEcad PECAM1 DAPI b HUVEC BMEC NE-EpiC Neuro-Endothelial H1 iBMEC H6 iBMEC IMR90-4 iBMEC Day 11 Day 11 Day 11 VEcad

Meso-Endothelial H1 iEC H6 iEC IMR90-4 iEC Day 16 Day 16 Day 16 Day 12 Day 12 Day 12 PECAM1 DAPI

iBMEC day 16 iBMEC c HUVEC (H1, H6, IMR90-4) d Merge e 70 60 50

H1 40 30 EPCAM

NSG-SGM3 20 (n=10 per line) HUVEC

Merge of Junctions Merge Number Total 10 0 HUVEC IMR90-4 iBMEC H1 iBMEC H6 iBMEC H6 5 days 12000 Agglutinin 10000

8000

DAPI Merge 6000

4000

of Vessels 2000 Total Length Total

Microscopy VEcad

& IMR90-4 0 HUVEC IMR90-4 iBMEC H1 iBMEC H6 iBMEC Quantification Extended Data Figure 2 a Sample identity 50 1 BMEC 2 H1 iBMEC Day 11 bioRxiv preprint doi: https://doi.org/10.1101/69917317 18 ; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is1 the author/funder. All rights3 reserved.H1 iBMEC No reuse allowed Day without 16 permission. 4 H6 iBMEC Day 11 6 6 25 6 5 H6 iBMEC Day 16 16 6 HUVEC 1 7 7 IMR90 iEC Day 12 1 8 IMR90 iBMEC Day 11 9 IMR90 iBMEC Day 11 0 9 14 5 15 10 IMR90 iBMEC Day 16

tSNE_2 4 8 13 11 IMR90 iBMEC Day 16 12 12 IMR90 iBMEC Day 11 2 13 IMR90 iBMEC Day 6 in ECM −25 3 14 IMR90 iBMEC Day 16 10 11 15 IMR90 iBMEC Day 6 in UM 16 LN4 iEC Day 12 1 17 LN4 iEC Day 17 −50 18 LN4 iEC Day 21 −50 −25 0 25 tSNE_1 b

1 1525 2 1424 3 1388 4 1027 5 1396 6 1694 7 1068 8 706 9 1130 20069 10 820 11 1165 Cells `1 12 1077 13 339 14 1159 15 386 16 2198 17 796 18 771 1000 2000 3000 4000 5000 Number of cells passing QC Number of genes per cell Extended Data Figure 3 aa Sample identity 50 1 Primary BMEC

2 Primary HUVEC bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 2 3 IMR90-4 iEC 25 4 IMR90-4 iBMEC D11

10 7 5 IMR90-4 IBMEC D11 11 1 6 IMR90-4 iBMEC D16 12 0 4 7 IMR90-4 iBMEC D16 8 IMR90-4 iBMEC D11 tSNE_2 8 5 9 IMR90-4 iBMEC D6 ECM 3 10 IMR90-4 EFF-iBMECs D11 −25 6 11 IMR90-4 EEF-iBMECs D11 rerun 13 12 IMR90-4 iBMEC D16 9 13 IMR90-4 iBMEC D6 UM −50 −30 0 30 tSNE_1 b IMR90-4 derived EEF-iBMEC

PECAM1 VEcad EPCAM VEcad

Day 11

Sort PECAM1+ VEcad+ PECAM1 Purified IMR90-4 EEF-iBMEC EPCAM

Day 27 Sample identity Extended Data Figure 4 1 BMEC ADAM17 APH1A 2 CIR1 H1 iBMEC a CREBBP 3 CTBP1 4 CTBP2 H6 iBMEC DLL1 5 DLL3 DLL4 6 HUVEC DTX1 7 DTX2 IMR90 iEC DTX3 8 DTX3L DTX4 9 bioRxiv preprint doi:DVL1 https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not10 certifiedDVL2 by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. DVL3 11 EP300 IMR90 HDAC1 12 HDAC2 13 iBMEC HES1 HES5 14 JAG1 JAG2 15 KAT2A KAT2B 16 LFNG 17 LN4 iEC MAML1 MAML2 18 MAML3 MFNG NCOR2 pct.exp NCSTN NOTCH1 0.00 NOTCH2 0.25 NOTCH3 NOTCH4 0.50 NUMB 0.75 NUMBL PSEN1 1.00 PSEN2 PSENEN PTCRA avg.exp. RBPJ scale

Notch pathway (map04330) RFNG SNW1 2 a b c d e f g h i j k l m n o p q r s t u v w x y z a’ b’ c’ d’ e’ f’ g’ h’ i’ 1 Gene Cluster 0 #Sample 6 1 6 18 1 17 16 7 16 7 1 5 14 10 8 13 9 3 11 2 5 14 8 5 14 4 12 −1 15

b AKT1 AKT2 AKT3 BAD CASP9 CDC42 CHP2 HRAS HSPB1 KDR KRAS MAP2K1 MAP2K2 MAPK1 MAPK11 MAPK12 MAPK13 MAPK14 MAPK3 MAPKAPK2 MAPKAPK3 NFAT5 NFATC1 NFATC2 NFATC3 NFATC4 NOS3 NRAS PIK3CA PIK3CB PIK3CD PIK3CG Sample identity PIK3R1 1 BMEC PIK3R2 2 PIK3R3 3 H1 iBMEC PIK3R5 4 PLA2G10 H6 iBMEC PLA2G12A 5 PLA2G2A 6 HUVEC PLA2G2F 7 PLA2G3 IMR90 iEC PLA2G4A 8 PLA2G5 9 PLA2G6 10 PLCG1 11 PLCG2 IMR90 12 PPP3CA iBMEC PPP3CB 13 PPP3CC 14 PPP3R1 15 PPP3R2 16 PRKCA PRKCB 17 LN4 iEC PRKCG 18 PTGS2 PTK2 pct.exp PXN 0.00 RAC1 RAC2 0.25 RAC3 0.50 KEGG VEGF pathway (map04370) RAF1 0.75 SH2D2A 1.00 SHC2 SPHK1 avg.exp. SPHK2 scale SRC VEGFA 2 a b c d e f g h i j k l m n o p q r s t u v w x y z a’ b’ c’ d’ e’ f’ g’ h’ i’ 1 Gene Cluster 0 −1 #Sample 6 1 6 18 1 17 16 7 16 7 1 5 14 10 8 13 9 3 11 2 5 14 8 5 14 4 12 15 Extended Data Figure 5 APC APC2 AXIN1 AXIN2 BTRC CACYBP CAMK2A CAMK2B CAMK2D CAMK2G CCND1 CCND2 CCND3 CHD8 CHP2 bioRxiv preprintCREBBP doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not CSNK1A1 certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. CSNK1A1L CSNK2A1 CSNK2A2 CSNK2B CTBP1 CTBP2 CTNNB1 CTNNBIP1 CUL1 CXXC4 DAAM1 DAAM2 DKK1 DKK2 DVL1 DVL2 DVL3 EP300 FBXW11 FOSL1 FRAT1 FRAT2 FZD1 FZD10 FZD2 FZD3 FZD4 FZD5 FZD6 FZD7 FZD8 FZD9 GSK3B JUN LEF1 LRP5 LRP6 MAP3K7 MAPK10 MAPK8 MAPK9 MMP7 MYC NFAT5 NFATC1 NFATC2 NFATC3 NFATC4 NKD1 NKD2 NLK PLCB1 PLCB2 PLCB3 PLCB4 PORCN PPARD PPP2CA PPP2CB PPP2R1A PPP2R1B PPP2R5A PPP2R5B PPP2R5C PPP2R5D PPP2R5E PPP3CA PPP3CB PPP3CC PPP3R1 PPP3R2 PRICKLE1 PRICKLE2 PRKACA Sample identity PRKACB PRKCA 1 BMEC PRKCB 2 PRKCG H1 iBMEC PRKX 3 PSEN1 4 RAC1 H6 iBMEC RAC2 5 RAC3 RBX1 6 HUVEC 7

KEGG WNT pathway (map04310) RHOA ROCK1 IMR90 iEC ROCK2 8 RUVBL1 9 SENP2 SFRP1 10 SFRP2 11 SFRP4 IMR90 SFRP5 12 SIAH1 iBMEC SKP1 13 SMAD2 14 SMAD3 SMAD4 15 SOX17 16 TBL1X TBL1XR1 17 LN4 iEC TBL1Y TCF7 18 TCF7L1 TCF7L2 TP53 pct.exp VANGL1 VANGL2 0.00 WIF1 WNT10A 0.25 WNT10B 0.50 WNT11 WNT16 0.75 WNT2 WNT2B 1.00 WNT3 WNT3A avg.exp. WNT4 WNT5A scale WNT5B WNT6 WNT7A 2 WNT7B 1 WNT9B 0 a b c d e f g h i j k l m n o p q r s t u v w x y z a’ b’ c’ d’ e’ f’ g’ h’ i’ −1 Gene Cluster #Sample 6 1 6 18 1 17 16 7 16 7 1 5 14 10 8 13 9 3 11 2 5 14 8 5 14 4 12 15 Extended Data Figure 6

ACTB ACTG1 ACTN1 ACTN2 ACTN3 ACTN4 AKT1 AKT2 AKT3 bioRxivAMOTL1 preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not ASH1L CASK certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. CDC42 CDK4 CGN CLDN1 CLDN10 CLDN11 CLDN14 CLDN15 CLDN16 CLDN19 CLDN22 CLDN23 CLDN3 CLDN4 CLDN5 CLDN6 CLDN7 CLDN8 CLDN9 CRB3 CSNK2A1 CSNK2A2 CSNK2B CTNNA1 CTNNA2 CTNNA3 CTNNB1 CTTN EPB41 EPB41L1 EPB41L2 EPB41L3 EXOC3 EXOC4 F11R GNAI1 GNAI2 GNAI3 HCLS1 HRAS IGSF5 INADL JAM2 JAM3 KRAS LLGL1 LLGL2 MAGI1 MAGI2 MAGI3 MLLT4 Sample identity MPDZ 1 MPP5 BMEC MRAS 2 MYH10 MYH11 3 H1 iBMEC MYH14 MYH15 4 MYH3 H6 iBMEC MYH6 5 MYH7 MYH7B 6 MYH9 HUVEC MYL12A 7 MYL12B IMR90 iEC MYL2 8 MYL5 MYL7 9 MYL9 MYLPF 10 NRAS OCLN 11 PARD3 IMR90 PARD6A 12 PARD6B PARD6G 13 iBMEC PPP2CA PPP2CB 14 PPP2R1A PPP2R1B 15 PPP2R2A PPP2R2B 16

KEGG tight junction pathway (map04530) PPP2R2C PPP2R2D PRKCA 17 LN4 iEC PRKCB PRKCD 18 PRKCE PRKCG PRKCH PRKCI pct.exp PRKCQ PRKCZ 0.00 PTEN 0.25 RAB13 0.50 RAB3B RHOA 0.75 RRAS 1.00 RRAS2 SPTAN1 SRC avg.exp. SYMPK scale TJAP1 TJP1 TJP2 2 TJP3 1 VAPA YES1 0 ZAK −1 a b c d e f g h i j k l m n o p q r s t u v w x y z a’ b’ c’ d’ e’ f’ g’ h’ i’ Gene Cluster #Sample 6 1 6 18 1 17 16 7 16 7 1 5 14 10 8 13 9 3 11 2 5 14 8 5 14 4 12 15 Extended Data Figure 7 EPCAM CDH1 a Sample identity 50 1 BMEC 2 18 H1 iBMEC bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019.17 The copyright holder for this 3preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed1 without permission. 4 H6 iBMEC 6 6 5 CLDN4 CLDN6 25 6 6 HUVEC 16 1 7 7 IMR90 iEC 8 1 9 0 9 14 5 15 10 tSNE_2 4 CLDN7 PDPN 11 8 13 IMR90 iBMEC 12 12 2 13 −25 3 14 10 11 15 16 KRT19 KRT8 1 17 LN4 iEC −50 18 −50 −25 0 25 Max tSNE_1 avg.exp. scale Min. b PECAM1 CDH5 CLDN5 ESAM ICAM2

SOX17 SOX18 FLI1 ERG ENG

EGFL7 PROCR PTPRB FLT1 KDR

VWF CD93 RASIP1 ROBO4 MMRN2 bioRxiv preprint doi: https://doi.org/10.1101/699173; this version posted July 11, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Extended Data Figure 8

Neuro-Endothelial IMR90-4 iBMEC g

) 2 ZO-1 Claudin-5 ZO-1Claudin-5 DAPI a b c Ω x cm Resistance (

Time (hours post day 8 replate)

ZO-1 Occludin ZO-1Occludin DAPI

h ) d e f 2 Capacitance (nF x cm

Time (hours post day 8 replate)