Engineering the haemogenic niche mitigates endogenous inhibitory signals and controls pluripotent stem cell-derived blood emergence

Efforts to recapitulate haematopoiesis, a process guided by spatial and temporal inductive signals, to generate haematopoietic progenitors from human pluripotent stem cells (hPSCs) have focused primarily on exogenous signalling pathway activation or inhibition. Here we show haemogenic niches can be engineered using microfabrication strategies by micropatterning hPSC-derived haemogenic endothelial (HE) cells into spatially-organized, size-controlled colonies. CD34+VECAD+ HE cells were generated with multi-lineage potential in serum-free conditions and cultured as size-specific haemogenic niches that displayed enhanced blood cell induction over non-micropatterned cultures. Intra-colony analysis revealed radial organization of CD34 and VECAD expression levels, with CD45+ blood cells emerging primarily from the colony centroid area. We identify the induced interferon gamma protein (IP-10)/p-38 MAPK signalling pathway as the mechanism for haematopoietic inhibition in our culture system. Our results highlight the role of spatial organization in hPSC-derived blood generation, and provide a quantitative platform for interrogating molecular pathways that regulate human haematopoiesis.

H uman pluripotent stem cells (hPSCs) facilitate strategies to model human development and disease 1 , and may serve as a renewable source of cells in a variety of cell therapy applications 2 . However, current protocols for the generation of hPSC-derived cells often fail to deliver mature, adult-like cells 3 . For example, the generation of definitive blood progenitors that behave similar to blood stem cells isolated from somatic tissue remains a challenge. Blood progenitor cells arise from haemogenic endothelium (HE) through a process termed endothelial-to-haematopoietic transition (EHT) 4,5 . This process occurs at specific locations during development and is spatially and temporally regulated by a balance of activating and inhibiting signals that is not completely understood. A well-studied region for mammalian blood cell emergence is the aorta-gonad-mesonephros (AGM). The first human definitive haematopoietic stem cells (HSCs) are spatially restricted to the ventral floor of the dorsal aorta 6 . In the AGM, paracrine signals from tissues ventral to the dorsal aorta (such as the mesenchyme, primitive gut and sympathetic nervous system) promote haematopoiesis, while tissues dorsal to the dorsal aorta (such as the neural tube and notochord) suppress blood formation [7][8][9][10] . These observations support our hypothesis that exerting spatial control over the local microenvironment of hPSC-derived HE will modulate blood cell yields and provide a platform to reveal organizing principles for this difficult-to-access developmental event.
Micropatterning has been used to spatially organize cells, allowing investigation of endogenous autocrine and paracrine signalling 11,12 . Using hPSCs, we have previously demonstrated that spatial control of endogenous BMP2 and GDF3 signalling can directly modulate pluripotency 11 . In addition, we have shown mouse embryonic stem cell colony size manipulation can control JAK-STAT activation, enabling subsequent transition towards epiblast stem cells 13,14 . Similarly, others have shown that geometric confinement of hPSCs can be used to recapitulate germ layer patterning. Herein we extend this approach to the micropatterning of hPSC-derived HE, and use this platform to control microenvironmental signals and spatial gradients during blood progenitor cell development.
We specifically report the serum-free generation of hPSCderived HE cells capable of producing both myeloid and lymphoid progenitors. We explore key engineered niche parameters for their impact on blood cell development and identify conditions that enhance blood cell generation. We subsequently identify interferon gamma-induced protein (IP)-10 as an endogenous inhibitory factor for hPSC-and cord bloodderived blood induction. Furthermore, we use live cell imaging to visualize location-dependent human CD45 þ cell emergence. Our results demonstrate the use of in vitro engineered cell niches to enhance PSC-derived blood cell development and provide a quantitative platform for interrogating molecular pathways that regulate this process.

Results
Serum-free generation of blood progenitor cells. To study the role of spatial control on signalling and developmental dynamics during in vitro blood differentiation we developed a protocol for efficient production of HE cells from which to generate and isolate appropriate progenitor populations (Fig. 1a). Using RUNX1C-GFP HES3 cells, we used Aggrewell plates to generate 1,000 cell hPSC aggregates by forced aggregation 15 , a size previously reported to be optimal for mesoderm differentiation 16,17 , and cultured them under hypoxic conditions (5% O 2 ). Differentiating aggregates were transferred from Aggrewell plates to low-attachment six-well plates after 5 days ( Supplementary Fig. 1b-left column) and placed on a shaker for the remaining 20 days of hypoxic culture. Using this differentiation platform, we screened three serum-free differentiation protocols [18][19][20] for production of kinase insert domain receptor/tyrosine protein kinase kit/epithelial cadherin (KDR þ CKIT À ECAD À )-expressing cells ( Supplementary  Fig. 1a), an early population that has been postulated to contain HE cells 21 . The protocol based on Ng et al. 22 yielded significant levels of KDR þ CKIT À ECAD À cells in our hands and was used to generate hPSC-derived HE cells in all subsequent studies. The haematopoietic differentiation profile was phenotypically assessed by tracking protein expression of VECAD, CD34, CD43 and CD45 by flow cytometry. Expression of CD34 and VECAD peaked at day 8, whereas committed haematopoietic progenitors marked by CD43 and CD45 were detectable by day 10 and continued increasing until day 25 (Fig. 1b). Colony-forming cell (CFC) generation paralleled CD43 and CD45 expression and was significantly higher at day 25 (Pr0.05, n ¼ 3, one-way analysis of variance (ANOVA) with post hoc Tukey test; Fig. 1c) compared to day 13 (see Supplementary Fig. 1b for representative colonies of myeloid lineage from hPSC-derived blood progenitor cells). Moreover, early CD43 expression marks differentiating cells capable of giving rise to erythroid progenitors (Ery-P) indicating these cells represent the primitive wave of embryonic haematopoiesis ( Supplementary Fig. 2a) 23,24 . This protocol for haematopoietic differentiation was robustly transferable to H1 ( Supplementary Fig. 1b), HES2 and H9 ( Supplementary Fig. 2) cell lines. Our results demonstrate that hPSCs can faithfully recapitulate embryonic haematopoiesis in a growth factor-defined, serum-free differentiation protocol (see Fig. 1d for representative flow cytometry plots). We next sought to isolate an intermediate population that possess characteristics of HE cells to investigate spatial control of signalling during blood cell differentiation.
Characterization of an hPSC-derived HE cell population. To characterize HE populations capable of blood cell generation we used CD34 þ VECAD þ expression 6 to isolate cells from day 8 differentiating hPSC aggregates (see Fig. 2a for representative flow cytometry plots). We also assessed CD184 and CD73 expression 21,25 (Supplementary Fig. 2b), which have previously been shown to identify arterial and venous endothelium, respectively. To maximize the yield of day 8 CD34 þ VECAD þ cells, we performed an aggregate size screen (100 to 1,000 cells per aggregate) and determined the highest yield of CD34 þ VECAD þ cells per input hPSC, 0.32±0.015 (mean ± s.e.m.), was obtained in cultures initiated with 500 cell aggregates (Pr0.05, n ¼ 3, one-way ANOVA with post hoc Tukey test; Supplementary Fig. 1c). The CD34 þ VECAD þ HE cells generated using our protocol were composed of 34.5%±3.5% (mean ± s.e.m.) CD73 À CD184 À cells, a substantially higher frequency than has been previously reported 25,26 . Lack of both markers can be used to distinguish HE-enriched for multi-lineage progenitors that can efficiently give rise to blood cells of myeloid and lymphoid lineages. Moreover, our serum-free differentiation protocol was robustly transferable across multiple hPSC lines, including the HES2 and H9 lines ( Supplementary Fig. 2c). To functionally validate the HE phenotype of day 8 hPSC-derived cells, cell sorting was performed based on CD34 and VECAD expression (Fig. 2a). Positive fractions (CD34 þ VECAD þ ) were confirmed using immunocytochemistry and were capable of metabolizing DiI-acetylated-low-density lipoprotein (DiI-Ac-LDL) and stained exclusively for von Willebrand factor (vWF; Fig. 2b), consistent with characteristics of HE cells 5 . For comparative purposes, the impact of SB-431542 and CHIR99012 on HE induction in our serum-free differentiation protocol was measured 27 . The addition of SB-431542 led to a twofold increase in the frequency of CD34 þ cells on day 6 compared to untreated control conditions, while the addition of CHIR99012 did not significantly enhance CD34 expression compared to control treatments ( Supplementary Fig. 2d). The CD34 þ expression frequency observed in SB-431542-treated Aggrewell cultures was substantially higher than previously reported (B60% versus 20%) 27 , a result attributed to the generation of uniform aggregates of optimal size. Furthermore, CD235a (a marker indicating progenitor cells primed for primitive haematopoiesis) was minimally expressed in our differentiation culture compared to SB-43152-and CHIR99012supplemented treatments (Supplementary Fig. 3A). Collectively,   our data indicate that day 8 CD34 þ VECAD þ cells are enriched for HE and can be used to investigate EHT transition and blood development. We next tested lymphoid differentiation of our HE population to assess multi-lineage potential.
HE cells display multi-lineage differentiation. Sorted CD34þ cells were seeded on an OP9-Delta-Like 4 (OP9-DL4) stromal feeder layer, and cultured as described previously 28 (Fig. 2c). T-cell generation from hPSC-derived cells has previously been used to demonstrate the generation of haematopoietic progenitors during the second wave of embryonic haematopoiesis 29 . By day 15 of OP9-DL4 culture, CD34 þ expression was downregulated and CD45 þ cells, a subset of which expressed CD5 þ CD7 þ indicative of pro-T lymphoid cells 30 , emerged (Fig. 2c). Further T-cell maturation was demonstrated by the appearance of CD4 þ CD7 þ cells by day 25 (Fig. 2c). As a final characterization, cells from day 8 of our serum-free differentiation culture were sorted into HE-positive (CD34 þ VECAD þ CD43 À CD45 À ) and -negative (CD34 À VECAD À CD43 À CD45 À ) fractions, to test for short-term in vivo haematopoietic engraftment in mice. Positive and negative fractions from five differentiation cultures were pooled and frozen ( Supplementary  Fig. 4a). After thawing, purity analyses of pre-transplanted positive and negative fractions were conducted by flow cytometry (Supplementary Fig. 4b). These fractions were subsequently injected intravenously into sublethally irradiated mice. Ten weeks after transplantation, mice were killed and human-derived blood cells were identified using antibodies against human leukocyte and CD45 antigens. Human CD45 þ reconstitution in the peripheral blood was significantly higher in mice injected with the positive fraction (0.77%) compared to mice injected with the negative fraction (0.20%) (Pr0.05, n ¼ 3, Student's t-test; Fig. 2d). Furthermore, frequencies of both CD41 þ (megakaryocytes) and CD3 þ (T cells) cells were highest in the peripheral blood of mice injected with the positive fraction compared to those injected with the negative fraction (summarized in Supplementary  Fig. 4c; flow cytometry plots in Supplementary Fig. 5). These results demonstrate that day 8 hPSC-derived HE cells are capable of giving rise to T-cell-restricted lineage cells in vitro, and haematopoietic blood cells in vivo (albeit at low levels), indicative of their potential for definitive blood differentiation. We next set out to use this population of cells to explore conditions that modulate blood cell emergence.
HE colony size and pitch impact blood progenitor yield.
Micropatterning of extracellular matrix (ECM) on cell culture surfaces enables the direct control of cell colony size and configuration. To control and study cellular organization during EHT, day 8 hPSC-derived HE cells were seeded on ultraviolettreated ECM micropatterned 96-well plate surfaces, and CD45 þ blood cell emergence was tracked over 5 days (schematic depicted in Fig. 3a). These micropatterned surfaces allowed us to control the local microenvironment by manipulating colony size, colony pitch (distance between colony centres) and colony clustering (the density for fixed number of colonies). Colony size was varied, independent of colony pitch, by increasing colony diameter at a fixed pitch (representative images in Figs 3b,c). Colony pitch was manipulated by maintaining colony diameter and varying distance between colony centres. Colony clustering was controlled by configuring colony size and spacing to keep the total number of cells within a well constant (constant global cell density). Representative images and average colonies per manipulation are summarized in Supplementary Fig. 6a and b, respectively. To ensure that patterning conditions were not selecting for CD34 þ VECAD þ cells, analysis was performed at 6 h and 5 days post seeding. Similar patterned cell densities were observed across all treatments before the onset of CD45 þ blood generation ( Supplementary Fig. 7). Intra-colony densities in increasing colony size and increasing pitch (Supplementary Fig. 7e) showed no significant changes throughout the assay (t ¼ 6 h to 5 days post seeding).
To investigate a wide range of colony pitches, small colonies with 200 mm diameters were generated at varying pitch lengths of 400, 500 and 800 mm. CD45 þ blood cell induction was enhanced with increasing pitch with colonies of 200-800 generating 2.5-fold more CD45 þ blood cells compared to 200-400 colonies (Pr0.05, n ¼ 3, one-way ANOVA with post hoc Tukey test; Fig. 3d). CD34 þ VECAD þ expression did not appear to deviate with changes in colony pitch ( Supplementary Fig. 7d). These results demonstrate that colonies of the same size but with greater pitch enhance CD45 þ blood generation compared to colonies with smaller pitch. To test the effect of clustering on blood induction, colony size was varied while maintaining constant colony coverage (calculated at B12.5% of total area within well). CD45 þ blood cell induction was similar between all treatments (Fig. 3d). These results indicate that at constant coverage, colony clustering (irrespective of colony size) does not enhance haematopoietic induction.
We next tested gene expression levels of haematopoieticassociated factors in blood cells generated from 150-500 and 400-500 micropatterned colonies. We assayed TAL1, MYB, NR2F2, EPHRINB2, NOTCH1, KDR, DLL4 and JAG1 expression via quantitative real time-PCR on day 0, 2 and 5 post patterning. It was found that MYB and TAL1 gene expression was similar between blood cells generated from 150-500 and 400-500 colonies. Definitive haematopoietic cells have been postulated to arise primarily via blood vessel specification with the Notch and VEGF pathway being instrumental in this process. Day 5 blood cells generated from 150-500 micropatterns displayed higher compared to 400-500 micropatterns (Fig. 3e). In addition to gene expression, we utilized the RUNX1C-GFP HES3 cell line to track haematopoietic differentiation. The RUNX1C isoform has been shown to be an important marker to identify definitive blood cells in vivo 31,32 , and cells expressing this marker give rise to multipotent blood progenitor cells in vitro 33 . We patterned 150 and 400 mm diameter HE colonies in EHT cultures and assessed for CD45 þ CD34 þ RUNX1C þ blood expression by flow cytometry. Our results indicate that blood cells generated from 150 mm colonies had seven fold higher CD45 þ CD34 þ RUNX1C þ expression compared to blood cells from 400 mm colonies (representative flow cytometry plots are displayed in Supplementary Fig. 8). These results demonstrate that geometric restriction of HE cells affects blood cell progenitor yield and their respective gene expression patterns. Overall, our results demonstrate that blood cell generation is enhanced in smaller colonies that are farther apart, suggesting endogenous inhibitory factors influence haematopoietic differentiation from HE cells.
Colony size controls IP-10 secretion. Recent literature has indicated inflammatory molecules provide cues for the generation of blood progenitor stem cells 34 . To test whether the observed increase in CD45 þ blood generation in lower-coverage micropatterns was the result of a reduction in inhibitory factors produced over 5 days, we used an inflammatory human cytokine magnetic bead panel to analyse secreted factor concentrations in the media of micropatterned and non-micropatterned cultures. This analysis revealed that IP-10 was significantly elevated (4 Â ) in non-micropatterned treatments (0.017±0.0070 pg ml À 1 per cell) compared to micropatterned conditions (0.0040 ± 0.0012 pg ml À 1 per cell; Pr0.05, n ¼ 3, Student's t-test; Fig. 3f). We next tested whether varying IP-10 concentration in differentiating HE cells affects blood generation. Supplementation of exogenous IP-10 to both 150-500 and 400-500 mm treatments leads to significant reduction (65%) in blood generation compared to respective control treatments (Fig. 3g). To   circumvent IP-10 inhibition, anti-CXCR3 (receptor to IP-10)neutralizing antibodies were added to micropatterned differentiating HE cultures over 5 days (Fig. 3g). This treatment moderately increased haematopoietic induction in high-coverage 400-500 mm treatments compared to respective controls. Previously, IP-10 has been reported to signal through the p38 mitogen-activated protein kinase (p38 MAPK) pathway 35 and thus we explored activation of phosphorylated p38 in HE-optimized 150-500 conditions on day 5 across control, IP-10, anti-CXCR3 and VX-702 (p38 MAPK inhibitor) treatments ( Supplementary Fig. 9a). We observed a moderate, but not significant, increase in VECAD þ p-p38 þ -expressing cells in the presence of IP-10 (1.1 ± 0.06-fold change over control treatments; Supplementary Fig. 9b). However, anti-CXCR3 and VX-702 treatments significantly reduced p-p38-expressing cell numbers compared to control conditions (0.69±0.07 and 0.04 ± 0.008-fold change, respectively, Pr0.05, n ¼ 3, one-way ANOVA with post hoc Tukey test; Supplementary Fig. 9b). Importantly, although IP-10 treatment moderately increased VECAD þ p-p38 þ cell levels, CD45 þ cell induction was significantly reduced compared to control treatments ( Supplementary Fig. 9b). When IP-10 was supplemented with anti-CXCR3 antibody and VX-702, CD45 þ cell induction was rescued (0.83±0.02-and 1.13±0.09-fold change to control treatments, respectively; Supplementary Fig. 9b). VX-702-treated conditions were also tested using CFC assays ( Supplementary  Fig. 9), demonstrating a similar rescue from IP-10 treatment as the CD45 þ output. Our results indicate that in the presence of IP-10, CD45 þ induction can be rescued by decreasing p-p38 levels. Interestingly, although anti-CXCR3 treatment reduced p38 phosphorylation, enhancement in CD45 þ blood generation (compared to control) was not significantly enhanced, perhaps due to effects of other endogenous inhibitors. To investigate our hypothesis, we cultured HE cells as colonies with 400 mm diameter and 2,000 mm pitch (400-2,000) to further decrease overall cell density (theoretical well coverage ¼ 3%) compared to 150-500 (theoretical well coverage ¼ 7%) and 400-500 (theoretical well coverage ¼ 50%). Our results show that CD45 þ haematopoietic induction from 400-2,000 colonies was significantly better than 150-500 (1.4 Â ; Fig. 3g) and 400-500 (3.2 Â ; Fig. 3g) colonies inferring that endogenous inhibitors were further mitigated by culturing HE cells as lower-coverage micropatterns. Together these results identify IP-10 as a molecule that inhibits hPSC-derived blood induction and whose secreted concentration is colony size-controlled.
To further assess the biological relevance of IP-10 during human haematopoiesis, we cultured magnetically sorted CD34 þ cells from umbilical cord blood cells for 7 days. We used flow cytometry to assess HSC (CD34 þ CD45RA À CD38 À CD90 þ and CD34 þ CD45RA À CD38 À CD90 þ CD49f þ ), progenitor cell (CD34 þ CD45RA þ CD90 þ , CD34 þ CD45RA þ CD90 À and CD34 þ CD45RA À CD90 À ) and differentiated cell (CD34 À ) yields. Our results indicate that IP-10 supplementation to CD34 þ cord blood (CB) cells significantly decreases the HSC phenotype (Pr0.05, n ¼ 3, Student's t-test) compared to control treatments ( Supplementary Fig. 10a). Similar to our hPSC study, addition of anti-CXCR3 antibody with IP-10 rescued HSC yields to levels comparable to control treatments. Moreover, since IP-10 signals through the p38 MAPK signalling pathway, we hypothesized that addition of VX-702 (p38 MAPK kinase inhibitor) could potentially enhance the HSC phenotype yield. The addition of VX-702, either alone or in conjunction with anti-CXCR3 antibody, rescued the HSC phenotype to levels that were comparable to control treatments. Together, our results extend our data to adult definitive haematopoiesis and demonstrate that IP-10 is a potent inhibitor of HSC-enriched cells in cord blood and its effect can be mitigated by using anti-CXCR3 antibody and VX-702 treatment.
Colony size affects HE organization and CD45 þ emergence. Our observation that endogenous signalling may impact EHT, along with recent observations from our group and others of the role of spatial gradients in cell fate patterning 11,36 , led us to examine the role of HE spatial organization on the emergence of hPSC-derived human blood progenitors. Post-imaging analysis was used to interrogate spatial CD34 and VECAD expression in HE colonies at various colony sizes (schematic depicted in Fig. 4a). Pronounced radial organization of CD34 þ expression was observed in larger colonies (200 and 400 mm diameter), wherein protein expression was higher in colony centres compared to peripheral cells (Fig. 4b). Moreover, radial organization was weaker as colony size decreased to 150 mm, consistent with smaller-sized colonies having smaller endogenous signalling gradients 11,13 . Importantly, no gradients were detected in colonies cultured in base medium devoid of haematopoietic cytokines (Fig. 4b) implying that radial organization only occurs during haematopoietic induction. To eliminate false-positive signals from variable intra-colony cell densities, we assessed for b-actin expression and found similar signals across 150-500 and 400-500 micropatterns on day 5 ( Supplementary Fig. 7a). To quantify this intra-colony spatial organization, CD34 þ and VECAD þ intensity values were binned into six groups corresponding to distances from respective colony centres and then normalized to the bin closest to the centre. It was found that CD34 þ and VECAD þ expression levels were uniform throughout each colony in the 150-500 condition, whereas in the 200-500 and 400-500 conditions, protein expression levels were significantly greater in colony centres compared to colony edges (Pr0.05, n ¼ 3, Student's t-test; Fig. 4c). These results indicate that manipulating HE colony size spatially regulates CD34 and VECAD expression.
To assess whether colony size impacts spatial control of haematopoietic cell emergence, we tracked blood cell-budding frequencies (emerging CD45 þ cells) in inner and outer areas (inner: o50% colony radius; outer: 450% colony radius) of 150-500 and 400-500 colonies using live imaging microscopy (Supplementary Movies 1 and 2, representative image in Supplementary Fig. 7b). Blood cell-budding frequencies were 4.5-fold higher in colony centroids than in colony edges in 400-500 colonies, and comparable in colony centroids and edges in 150-500 mm colonies (Fig. 4d). These data indicate that geometric restriction of HE cells can spatially impact intra-colony marker expression and blood cell emergence.

Discussion
Embryonic haematopoiesis is a dynamic process guided by temporal and spatial cues. The dorsal aorta is the first site of definitive blood cell emergence during embryogenesis. Interestingly, HE cells that give rise to these blood cells are spatially restricted to the ventral floor and are surrounded by stromal cells containing a mix of endothelial and mesenchymal cells 8 . Moreover, HE cells-situated in a region of densely packed cells 37 , ECM proteins 37 and polarized cytokine gradients 38receive signals from surrounding tissues for intra-arterial blood cluster formation 9,39 . Mimicking individual niche components not only elucidates how blood emergence occurs but also provides information that can improve in vitro differentiation protocols to generate target cell types from hPSCs. Here we present a robust protocol for hPSC-derived HE cells using size-controlled aggregates and chemically defined medium, which consistently gives rise to myeloid and pro-T cells in vitro and display short-term reconstitution in vivo. Further, we provide evidence that micropatterning size-restricted hPSC-derived HE colonies significantly increases CD45 þ blood generation, with higher blood cell emergence frequencies in smaller and more separated colonies. These configurations mitigate the inhibitory effects of endogenous molecules such as IP-10 on blood development. Mechanistically, we link the p38 MAPK pathway to the decrease of CD45 þ blood generation via IP-10 supplementation, and extend this data set to cord bloodderived cells. Finally, we show that gradients of blood cell emergence occur in spatially restricted HE colonies with larger diameters (4200 mm). To develop an in vitro model of EHT, which would allow us to study the role of spatial organization and endogenous signalling during definitive blood development, we designed and optimized a protocol to derive HE cells from hPSC. In this protocol, we control hPSC aggregate size in serum-free differentiation conditions to efficiently generate HE cells. The optimized aggregate size for induction to definitive HE phenotypes (500 cells per aggregate) yielded 19-fold higher expression of the definitive HE phenotype compared to current protocols 25 . Our HE cells possessed multi-lineage potential consistently giving rise to myeloid and lymphoid lineages (CD5 þ CD7 þ pro-T cells), cardinal features of definitive haematopoiesis. Moreover, our candidate HE population (sorted CD34 þ VECAD þ CD43 À CD45 À ) was capable of short-term reconstitution in non-obese diabetic/ SCID/IL-2Rgc-null (NSG) mice and harboured the majority of blood-inducing cells, as this fraction showed significantly higher engraftment than the negative fraction. Although low levels of mean CD45 þ engraftment (0.77%) were observed, these levels fall within the upper range of previously reported engraftment levels by Tian et al.  42 have reported CD45 þ engraftment levels close to 16%, however these hPSC-derived cells were cultured on AGM-derived stroma, thereby negating clinical use. Higher CD45 þ reconstitution levels in primary recipients have been recently reported but in these cases hPSCderived cells were genetically modified 43 . Low engraftment of hPSC-derived cells may be due to functional differences between embryonic blood progenitors and mature definitive HSCs. For example, the absence of CXCR4 expression on hPSC-derived cells is a requirement for proper homing to the haematopoietic organ. To circumvent this issue, studies have injected hPSC-derived cells intra-femorally into immunocompromised mice as the femur is a more conducive environment for haematopoietic reconstitution 43 . Interestingly, our studies showed undetectable reconstitution levels in bone marrow samples compared to levels observed in peripheral blood indicating that embryonic-like blood progenitors lack homing capabilities or may home to secondary haematopoietic organs in adult recipient mice. Although engraftment rates of the negative fraction are 3.5-fold lower compared to the positive fraction, the ability of the negative fraction to minimally reconstitute immunocompromised mice could be due to maturation to the CD34 þ VECAD þ phenotype at time of injection or due to in vivo microenvironments conferring haematopoietic fates on injected hPSC-derived cells as has been previously reported 44 . By controlling culture parameters such as aggregate size and oxygen tension, we present a scalable, serum-free differentiation platform that robustly generates hPSCderived HE with multi-lineage potential.
Determining the signalling pathways, which are regulated by colony size and spacing, that impact blood cell emergence was of particular interest. We have previously shown that micropatterning influences autocrine and paracrine signalling during PSC maintenance and differentiation 11,13,14 , and speculate that endogenous control over signalling pathways plays an important role in the haematopoietic differentiation system described here as well. Consistent with our data set, smaller islands of micropatterned hPSCs have been reported to be efficient in endothelial differentiation compared to larger islands 45 -perhaps due to lower levels of inhibitory factors. We observed that IP-10 endogenously inhibited hPSC-derived haematopoietic induction in non-micropatterned conditions, and its concentration can be attenuated by configuring HE cells into smaller colonies, rescuing haematopoietic differentiation. A similar effect was observed when we supplemented CD34 þ sorted cord blood-derived cells with IP-10. It has been shown that IP-10 inhibits human endothelial cell motility and tube formation 46 , which could lead to improper patterning of HE intermediates thereby resulting in decreased blood cell formation. We observed that IP-10-mediated haematopoietic inhibition was controlled via the p38 MAPK signalling pathway, which has been reported to play a crucial role in promoting ex vivo expansion of human cord blood HSCs 47 . Our results are the first to identify the role of p38 MAPK signalling pathway in generating haematopoietic cells from PSCderived sources, and motivate future studies on the role of this pathway in de novo blood development. Interestingly, IP-10 knockout mice are not known to exhibit haematopoietic defects 48 . To explore a potential role for IP-10 during murine embryonic haematopoiesis, we mined literature for IP-10 and CXCR3 gene expression in haemogenic cells during EHT in the AGM. RNAsequencing analysis of single-cell and 10-cell sorted haemogenic populations during murine AGM haematopoiesis 49 revealed the highest IP-10 and CXCR3 expression was present in mature HSC subtypes (defined as T2 pre-HSCs; Supplementary Fig. 10b). Supporting this data set, microarray analysis from E 11.5 AGM tissues 34 and RNA-sequencing analysis from E 10.5 AGM tissues 50 indicate that CXCR3 expression is enhanced in the HSC compartment compared to endothelial subtypes. Our data support a hypothesis that IP-10 can inhibit haematopoietic development either autonomously or non-autonomously by interacting with CXCR3 receptors present on HSCs. Inflammatory signals such as interferon gamma (IFNg) and tumour necrosis factor-a (TNF-a) are greatly affected in IP-10deficient mice, drastically impacting lymphoid cell circulation 51 . As inflammatory signals are important for murine embryonic haematopoiesis, IP-10 may play a role in modulating these signals. For example, in IP-10 À / À mice, there was an increase in IFNg-producing CD8 þ T cells compared to wild-type mice 51 . These results suggest IP-10 can modulate IFNg expression in a feedback mechanism. During embryonic haematopoiesis, IP-10 could attenuate pro-inflammatory signals since chronic exposure to IFNg has been reported to be detrimental for HSC function 52 . In our studies, blood cells generated from smaller colonies displayed higher expression levels of definitive haematopoietic factors such as KDR, NOTCH1 and JAG1 compared to highercoverage micropatterns. This is in line with previous reports, which have indicated that endogenous signalling can skew haematopoietic differentiation. Temporal VEGFA inhibition during mouse embryonic stem cell haematopoietic differentiation is crucial for promoting CFC generation at the expense of endothelial differentiation 53 . Inhibition of transforming growth factor-b and hedgehog signalling can enhance blood progenitor output, suggesting that negative feedback from differentiating cells plays a key role in haematopoiesis 54 . Studies in murine and zebrafish models have also revealed roles for endogenous factors such as bone morphogenetic factor-4 (ref. 38), IFNg (ref. 34) and phenylephrine 39 in AGM haematopoiesis. By engineering the cellular environment using micropatterning technology, we can further investigate endogenous factors that are crucial for hPSC-derived HE differentiation to definitive blood cells.
Collectively, the findings presented here demonstrate that engineering the haematopoietic developmental niche can impact blood cell yields and reveal parameters and molecules that control blood cell emergence. Our system provides a promising platform to investigate signalling pathways that inhibit blood cell emergence from differentiating hPSCs. Additional niche parameters such as immobilized ligands 8,55 and synergistic cell types 56 that are important for embryonic haematopoiesis can also be integrated into our system to closely mimic in vivo niches. Notably, our current results and the techniques we have described here represent crucial steps in unravelling the underlying mechanisms that drive the generation of hPSC-derived HE cells to HSCs.
Ultraviolet-lithography micropatterning of culture plates. To control the size and density of hPSC-derived HE colonies, ECM was micropatterned at spatially controlled geometries. Glass slides (Schott, 110 mm Â 74 mm glass slide) were activated in a plasma cleaner and then coated with poly L-lysine-grafted-polyethylene glycol (SUSOS, 0.1 mg ml À 1 ). The glass slide was incubated at 37°C in a humidified environment for 24 h. To create micropatterned islands, a photomask with specific geometries was activated in a plasma cleaner. The photomask was sandwiched with polyethylene glycol-treated glass slides and placed in an ultraviolet ozone (UVO) chamber (Jetlight) and activated for 10 min to create patterned surfaces. After deep ultraviolet etching, the photomask-glass slide superstructure was gently washed with ddH 2 O and dried gently with N 2 gas. To construct 96-well patterned plates, 96-well bottomless plates (VWR) were coated with Loctite epoxy adhesive and placed on the activated glass slide to create a sealed patterned plate. The following colony geometries were generated using this process (diameter (mm)-pitch (mm)): 150-500; 200-500; 400-500; 200-400; 200-500; 200-800; 150-375; 200-500; and 400-1,000). For immobilization of proteins, ddH 2 O was dispensed into the 96-well micropatterned plate and incubated for 30 min. N-(3-dimethylaminopropyl)-N 0 -ethylcarbodiimide hydrochloride (15 mg ml À 1 , Sigma) was dissolved in ddH 2 O and aliquoted at 100 ml per well for 20 min. The wells were washed with ddH 2 O twice. Next, 100 ml of an ECM mixture comprising of fibronectin-gelatin (0.00125% fibronectin/0.002% gelatin) in ddH2O was added to each well and the plate was incubated for 3 h to allow patterned ECM islands to form. The ECM mixture used was previously published as an appropriate substrate for screening pluripotency regulators on micropatterned hPSCs 57,58 . After the incubation, the wells were washed twice with PBS and stored at room temperature until initiating cell culture.
Endothelial to haematopoietic transition assay. Day 8 hPSC-derived HE cells were dissociated as single cells using collagenase (Invitrogen) and TrypLE Express as described above. Cells were strained through a 40 mm cell strainer, centrifuged and resuspended in StemPro-34 base medium with supplement and RI (1:1,000). Cells were seeded at 3.0 Â 10 4 cells per micropatterned well and incubated at 37°C for 6-12 h, gently washed twice with PBS to remove unattached cells, then incubated in haematopoietic inducing medium for 5 days to investigate blood generation from HE colonies. Haematopoietic induction from HE colonies was decreased by IP-10 supplementation (10 ng ml À 1 , R&D). To mitigate endogenous IP-10 secretion from HE colonies, anti-CXCR3 (0.2 mg ml À 1 , R&D) was added every 24 h for 5 days as previously described 35 .
Haematopoietic CFC assay. To assess blood progenitor potential, hPSC-derived cells or CD45 þ blood cells generated from micropatterns were seeded in duplicate at 1.0 Â 10 5 or 1.5 Â 10 5 cells per 35 mm Greiner dish containing Methocult H4435 enriched medium (StemCell Technologies). Samples were scored based on morphology 14 days after the cells were plated according to the manufacturer's protocol.
Analysis of NSG mouse engraftment. For engraftment studies, 8-week-old female mice, housed in a barrier facility under stringent pathogen-free conditions in filter-top cages, were used. Animals were handled in sterile cross flow hoods and given access to sterile food and water. All animal studies were performed according to the approved University Health Network research ethics board protocols. Female NSG mice were sublethally irradiated (250 rad) 24 h before transplantation. Day 8 hPSC-derived HE cells were magnetically sorted for CD34 þ (HE-positive) or FACS-sorted for CD34 þ VECAD þ CD43 À CD45 À (HE-positive) and CD34 À VECAD À CD43 À CD45 À (HE-negative) fractions and frozen down in cryopreservation vials until day of transplantation. Transplantation was carried out by injecting 2.3 Â 10 5 pooled cells of each fraction into irradiated NSG mice via tail vein based on previously published literature [41][42][43] . On a random basis, a total of three mice were injected per fraction. Mice were killed 10 weeks post transplantation. Peripheral blood, bone marrow and spleen from each mouse were collected and prepared for flow cytometry analysis. Red blood cells were lysed with 0.8% ammonium chloride solution, and the remaining cells were washed in PBS containing 5% FBS. The following markers were evaluated using antibodies purchased from BD Biosciences unless otherwise stated: OP9-DL4 co-culture for T-cell lineage differentiation. CD34 þ enriched progenitors were cultured on OP9-DL4 cells 59 at a density of 1.0-3.5 Â 10 5 cells per well of a six-well plate, and grown in alpha-MEM supplemented with 20% FBS (Gibco), 50 mg ml À 1 phospho-ascorbic acid (Sigma-Aldrich), 5 ng ml À 1 rhIL-7, 5 ng ml À 1 rhFLT3L and 10 ng ml À 1 rhSCF (Miltenyi Biotec). Cells were fed every 2-3 days, and transferred to fresh OP9-DL4 stroma every 5 days.
Live imaging and analysis. To visualize and track emerging blood cells, day 8 hPSC-derived cells were dissociated as single cells, seeded as micropatterned colonies in 96-well plates and allowed to adhere for 6-12 h. Non-adherent cells were washed away and placed in 200 ml haematopoietic-inducing medium with PE-CD45 (1:400) for 5 days. Cells were cultured in a humidified 5% (v/v) CO 2 air environment at 37°C housed in an on-stage incubation system with an inverted microscope (Zeiss). Phase-contrast images (acquired every 20 min) and fluorescent images (acquired every 2 h) were taken of representative colonies at Â 10 objective using Zeiss AxioVision software. The X-cite 120 LED lamp (Lumen dynamics) was used for fluorescence illumination at 60% intensity. Images were analysed and manually tracked using ZEN blue software (Zeiss). For 400-500 mm patterns, quarter or half images were used as samples for calculating budding frequencies.
Only cells with clear identity and behaviour were used for analysis.
Quantitative real time-PCR. Blood cells were generated from respective micropatterned colony treatments and pooled from three independent experiments (B450 colonies per time point). Total RNA was isolated and purified using PureLink RNA minikit (Invitrogen) according to the manufacturer's instructions. Normalized RNA across treatments were used to generated cDNA using Superscript III reverse transcriptase (Invitrogen) according to the manufacturer's instructions. Generated cDNA was mixed with primers and SYBR green mix (Roche) and run on the Applied Biosystems Quantostudio 6 Flex (Life Technologies). Relative expressions of genes were calculated using the delta-delta cycle threshold (C T ) method with the expression of GAPDH as an internal reference. Primer sequences are listed in Supplementary Table 1.
Human conditioned media analysis. Secreted levels of chemokines/cytokines were analysed in duplicate from conditioned media samples using an inflammatory panel against the human cytokine/chemokine 29-plex magnetic bead panel (Millipore, Billerica, MA, USA) with a Luminex 200 system (Luminex Co., Austin, TX, USA). Conditioned media samples were stored at À 80°C until use.
Statistical analysis and data representation. Statistics were performed using OriginPro to identify significant trends between/among groups. A minimum sample replicate size of n ¼ 3 was used for all data analysis, unless otherwise stated, for appropriate statistical testing. Comparisons were tested for normality before performing a one-way ANOVA with Tukey's post hoc analysis or Student's t-test. Treatments were significant using an alpha level of 0.05 with at least three independent replicates. Data are reported as mean±s.e.m.
Data availability. The authors declare that the data supporting all the findings of this study are available within this paper and its Supplementary Information Files.