Spatial Stem Cell Fate Engineering via Facile Morphogen Localization

Abstract

Spatiotemporally controlled presentation of morphogens and elaborate modulation of signaling pathways elicit pattern formation during development. Though this process is critical for proper organogenesis, unraveling mechanisms of human developmental biology have been restricted by practical and ethical challenges associated with studying human embryos. Human pluripotent stem cells (hPSCs) have been used to model human development in vitro, however difficulties in precise spatiotemporal control of the cell microenvironment have limited the utility of this model in exploring mechanisms of pattern formation. Here, we present a simple and versatile method to spatially pattern hPSC differentiation in 2D via localized adsorption of morphogens on substrates. Morphogens including BMP4, noggin, activin A, and Wnt3a are patterned to induce localized mesendoderm, endoderm, cardiomyocyte (CM), and epicardial cell (EpiC) differentiation from hPSCs and hPSC-derived progenitors. Patterned CM and EpiC co-differentiation allows investigation of interactions between these cells in a spatially controlled manner and demonstrated improved alignment of CMs, an important metric of maturity and coordinated function, in proximity to differentiating EpiCs. This simple approach provides a platform for the controlled, systematic study of spatiotemporal pattern formation during early development. Moreover, this study provides a facile approach to generate 2D patterned hPSC-derived tissue structures for modeling disease and drug interactions.

Graphical Abstract

A simple and versatile approach to generate patterned stem cell differentiation is suggested in this study. Deposition of morphogens on substrate is achieved by confining protein solution and it derives localized differentiation of stem cells which allows investigation of cellular interaction in a spatial context.

1. Introduction

During development, embryonic epiblast cells give rise to virtually all cell types in the body in a process governed by the elaborate, dynamic spatiotemporal control of physiochemical signaling exerted by the cellular microenvironment.^[1] Advances in imaging and computational analysis techniques have allowed researchers to investigate in vivo pattern formation induced by morphogens in animal models,^[2] but these in vivo models lack systematic, precise control of morphogens, limiting their use in understanding how these morphogen regulate cell fate and pattern formation.

The emergence of human pluripotent stem cells (hPSCs) has provided a powerful in vitro model for studying human development, evaluating therapeutics, modeling disease, and developing cell-based regenerative therapies. While modulation of developmental pathways at specific differentiation stages has led to efficient methods to differentiate hPSCs to many different somatic cell types, spatial pattern formation from hPSCs has been more challenging due to difficulties in localizing morphogens in conventional cell culture platforms.

Self-organization of PSCs has been used to demonstrate early developmental events such as germ layer specification,^[3] anteroposterior patterning,^[4] and ectodermal patterning.^[5] However, self-assembled patterning inherently lacks regulatory features with regards to spatial control of signaling cues. Alternatively, geometrically confining hPSC colonies into different shapes allowed investigation of effects of asymmetrically pre-patterned colonies on hPSC fate.^[6] Microfluidic devices offer a more complex model that can incorporate both spatial patterning and morphogen gradients, for example by mimicking the presence of proximal morphogen-secreting cells.^[7] Establishment of human neural tube patterning was also demonstrated in a microfluidic device via a Wnt-activating gradient.^[8] Substrate-mediated delivery of signaling factors, primarily through immobilization on a surface, is a common strategy to localize inductive factors that mimics natural protein presentation in the cellular microenvironment.^[9] In these methods surface functionalization is typically not spatially controlled and results in uniform substrate modification.^[10] Additionally, most approaches to generate patterns require complex strategies, including functionalizing factors to be immobilized^[11] and/or the substrate,^[12] or specialized techniques such as bioprinting.^[13]

Here, we report a simple method to pattern hPSC differentiation in 2D using a versatile approach that spatially regulates morphogen adsorption to a solid surface. Morphogens including BMP4 and activin A were patterned on culture plates to spatially regulate hPSC differentiation into mesendoderm and endoderm lineages. Wnt3a patterns permitted patterned control of hPSC-derived cardiac progenitor cell (CPC) differentiation to epicardial cells (EpiCs) and cardiomyocytes (CMs), permitting investigation of the effects of EpiCs on CMs during co-differentiation. We expect that this simple and versatile strategy for spatially controlled differentiation of hPSCs and progenitors will be a platform technology to study developmental cell fate patterning and intercellular communication between differentiating cells as well as facilitate construction of spatially-ordered structures for drug evaluation and tissue engineering applications.

2. Results

2.1. Localized protein patterns by confinement of protein solutions

We first verified that placing a protein solution on a substrate can generate localized protein adsorption. Bovine serum albumin conjugated with fluorescein isothiocyanate (BSA-FITC) solution was placed on a tissue culture plate as a droplet, drawn with a pipette tip, or confined with polydimethylsiloxane (PDMS) stencil masks. The BSA-FITC remained adsorbed after washing (Figure S1a). Importantly, more adsorption of BSA-FITC was observed on an untreated polystyrene (PS) plate when compared to a tissue culture (TC)-treated PS plate, suggesting that the adsorption likely involves hydrophobic interaction (Figure S1b). Next, the effect of incubation times on adsorption was assessed (Figure S1c). Even a 1 min incubation resulted in localized adsorption, consistent with the time scale of a hydrophobic interaction between a surfactant and a mineral surface.^[14]

To assess whether this patterned adsorption can be applied to protein morphogens, bone morphogenetic protein 4 (BMP4) solution was patterned as droplets, drawn letters, or stencil masks and then BMP4 was visualized by immunofluorescent microscopy. Like BSA-FITC, BMP4 retained its patterns after washing (Figure S2a) and more BMP4 was adsorbed on the untreated plate than the TC-treated plate (Figure S2b). These data demonstrate that a localized protein pattern can be created via a simple placement and confinement of a protein solution on a substrate.

2.2. Localized differentiation of hPSCs into mesendoderm by patterned BMP4

To explore whether spatial localization of adsorbed morphogens can mediate pattern formation during hPSC differentiation, different concentrations of BMP4 droplets were placed on either untreated PS or TC-treated PS plates followed by Matrigel coating and subsequent plating of hPSCs to generate a confluent monolayer. BMP4 in the presence of fibroblast growth factor 2 (FGF2) is known to induce primitive-streak like mesendoderm specification of hPSCs, characterized by expression of BRACHYURY (BRA).^[15] While none of the BMP4 droplets induced mesendoderm differentiation on untreated PS after two days of culture in mTeSR1, a range of BMP4 concentrations induced localized mesendoderm differentiation on TC-treated PS (Figure S3a, c). On the untreated PS surface, 0.1, 0.2, and 2.5 ng/μl BMP4 droplets inhibited cell attachment. Staining for laminin after placing a 2.5 ng/μl BMP4 droplet on untreated PS showed limited Matrigel coating on the patterned region (Figure S3c). Moving forward, we restricted our analysis to TC-treated PS surfaces since they permitted robust cell attachment atop morphogen-patterned surfaces.

Next, BMP4 solution was placed on TC-treated plates in various patterns, and hPSCs were seeded onto the substrate in mTeSR1. Immunostaining for BRA after 2 days of culture demonstrated that localized BMP4 patterns led to spatially confined mesendoderm differentiation (Figure 1a). Staining for SOX2 and NANOG showed that the cells outside the BMP4-patterned regions maintained their pluripotent state, while the BRA⁺ mesendoderm cells expressed these pluripotency markers at a lower level (Figure 1b). Some retained expression of pluripotency markers in mesendoderm cells is expected.^[15-16] The BRA expression profile at the pattern edge exhibited a sharp boundary (Figure S2b), suggesting that immobilized BMP4 acts directly on the hPSCs (Figure 1b). This primitive-streak like mesendoderm fate was further verified by the expression of MIXL1 in the cells on the BMP4 pattern (Figure 1c). Next, we antagonized BMP4-induced mesendoderm differentiation by noggin, a BMP inhibitor. A noggin solution was placed as a droplet on TC-treated PS, then hPSCs were seeded and cultured in mTeSR1 supplemented with 10 ng/ml BMP4 for 2 days. Only the cells outside of the noggin pattern expressed BRA, while the cells on the patterned area did not express BRA (Figure 1d), demonstrating that adsorbed noggin inhibited BMP4-induced mesendoderm differentiation. Small molecule inhibitors targeting actin-receptor like kinases (ALKs) including A83-01, SB431542, and LDN-212854 were tested to examine whether receptor-level inhibition can antagonize localized mesendoderm differentiation on BMP4 patterns (Figure S4a). When the ALK inhibitors were added for the initial 24 h of culture, they suppressed BRA expression in the cells on the BMP4-patterned area, suggesting that localized mesendoderm differentiation requires BMP4 binding and activation of the corresponding receptors. Localized mesendoderm differentiation was also achieved on different substrates (glass and ibidi polymer coverslip) (Figure S5a), different ECM coatings (Synthemax and vitronectin) (Figure S5b), and an additional hPSC cell line (Figure S6a, b). Collectively, these data show that this simple and robust strategy to localize morphogens such as BMP4 and noggin can spatially induce localized mesendoderm differentiation.

Figure 1. Localized differentiation of H9 hESCs into BRA⁺ mesendoderm by BMP4 patterning.

(a) Immunofluorescence images of BRACHYURY (BRA) at day 2 of differentiation on BMP4-patterned substrate and their quantification. 2.5 ng/μl BMP4 solution was placed as a droplet (Droplet), by pipette drawing (Drawing), or with PDMS stencil masks (Stencil) on TC-treated plates, incubated for 30 min at room temperature, and washed before Matrigel coating and hESC seeding. (b) Immunofluorescence images for the indicated fate markers. BMP4 was placed in a heart-shaped PDMS stencil before hESC seeding. Enlarged immunofluorescence images show the cells on the BMP4-pattern, outside the pattern, and at the pattern boundary. The normalized intensity profile was obtained from the average fluorescence intensity value of BRA along the y-axis divided by the average fluorescence intensity value of Hoechst along the y-axis of a representative image at the pattern boundary (c) Immunofluorescence images for BRA and MIXL1 at day 2 of differentiation on the BMP4-pattern and outside the pattern. 2.5 ng/μl BMP4 solution was placed as a droplet, incubated for 30 min at room temperature, and washed before Matrigel coating and hESC seeding. (d) Immunofluorescence images for the indicated fate marker after noggin patterning. 5 ng/μl noggin solution was placed as a droplet on a TC-treated plate, incubated for 30 min at room temperature, and washed before Matrigel coating and hESC seeding. Localized differentiation of H9 hESCs into BRA⁺ mesendoderm by BMP4 patterning was performed in at least ten independent differentiations with similar results. Scale bars=1 mm, unless otherwise indicated. The statistical comparison was performed using a two-tailed Student’s t-test (* P<0.05, n=3).

2.3. Localized differentiation of hPSCs into definitive endoderm by patterned activin A

Since activin A has been reported to induce differentiation of hPSCs into definitive endoderm,^[17] we patterned activin A to direct localized endoderm commitment of hPSCs. Activin A solution was placed on TC-treated plates in a multitude of patterns. hPSCs were seeded on activin A patterned substrates after coating with Matrigel, cultured for 3 days in RPMI1640 supplemented with fetal bovine serum, and immunostained for SOX17, a marker for definitive endoderm.^[17] Like BMP4-induced mesendoderm differentiation, adsorbed activin A induced localized endoderm differentiation indicated by the patterned expression of SOX17 (Figure 2a). Immunostaining for NANOG and SOX2 showed that cells outside the activin A patterned region retained SOX2 expression but lost NANOG expression, whereas the cells on the activin A-patterned region expressed both SOX2 and NANOG as well as SOX17 (Figure 2b). The retained expression of NANOG and SOX2 in SOX17⁺ endoderm population is consistent with the role of these pluripotency factors in directing early endoderm specification.^[18] The SOX17 expression profile at the pattern edge showed a sharp boundary, suggesting that activin A acted directly on the cells in contact with the patterns (Figure 2b). ALK receptor inhibition similarly abolished the patterned endoderm differentiation, showing that differentiation is induced via activin A activation of these receptors (Figure S4b). Localized patterning of definitive endoderm was also achieved in WTC11 iPSCs (Figure S6c, d). Since SOX17 also plays a role in the development of mesoderm lineages,^[19] we verified that the cells the activin A patterns expressed both SOX17 and FOXA2 at day 5 of differentiation, while virtually no cells outside of the activin A patterns expressed either of these markers (Figure 2c). Taken together, localized mesendoderm and endoderm differentiation can be accomplished by patterning BMP4 and activin A adsorption on a substrate, respectively.

Figure 2. Localized differentiation of H9 hESCs into SOX17⁺ definitive endoderm by activin A patterning.

(a) Immunofluorescence images of SOX17 at day 3 of the differentiation and their quantification. 25 ng/μl Activin A solution was placed as a droplet (Droplet), by pipette drawing (Drawing), or with PDMS stencil masks (Stencil) on TC-treated plates, incubated for 30 min at room temperature, and washed before Matrigel coating and hESC seeding. (b) Immunofluorescence images for the indicated fate markers. Activin A was patterned using a heart-shaped PDMS stencil before hESC seeding. Enlarged immunofluorescence images show the cells on the activin A patterns, outside the patterns, and at the pattern boundary. The normalized intensity profile was obtained from the average fluorescence intensity value of SOX17 along the y-axis divided by the average fluorescence intensity value of Hoechst along the y-axis of a representative image at the pattern. (c) Immunofluorescence images for SOX17 and FOXA2 at day 5 of the differentiation. 25 ng/μl Activin A solution was placed as a droplet on a TC-treated plate, incubated for 30 min at room temperature, and washed before Matrigel coating and hESC seeding. Localized differentiation of H9 hESCs into SOX17⁺ definitive endoderm by activin A patterning was performed in at least ten independent differentiations with similar results. Scale bars=1 mm, unless otherwise indicated. The statistical comparison was performed using a two-tailed Student’s t-test (n=3) (* P<0.05, n=3).

2.4. Localized multi-lineage differentiation of hPSCs by patterned BMP4 and activin A in a single well

We hypothesized that the versatility of spatially directing hPSC differentiation by patterning morphogens would permit spatial control of hPSC differentiation to multiple germ layers in a single culture. To examine this idea, BMP4 and activin A solutions were placed in PDMS stencils in different regions of single well, then hPSCs were seeded and cultured in these wells. At day 3 of differentiation, cells on BMP4-patterned areas expressed MESP1, a mesodermal transcription factor,^[20] and lost expression of SOX2 (Figure S7a). Some cells on BMP4 patterns expressed low levels of SOX17, which might suggest the presence of endoderm cells derived from the BMP4-induced mesendoderm population or mesoderm cells that transiently expressed SOX17.^[19a] Cells on activin A-patterned regions expressed SOX17 but not MESP1. At day 5 of differentiation, cells on BMP4-patterned regions were stained by TE7, an antibody that labels mesoderm-derived stromal cells,^[21] but did not express SOX2 (Figure S7b). Cells on the activin A patterns expressed SOX17 and SOX2, were TE7⁻. At day 7, most of the cells on the BMP4 patterns lost expression of SOX17 but stained positive with TE7, while the cells on the activin A patterns lost expression of SOX2 but retained expression of SOX17 (Figure 3a). The SOX2⁺ SOX17⁻ TE7⁻ cells outside the morphogen-patterned regions expressed the ectodermal transcription factor OTX2 (Figure 3b). A ~200 μm thick layer of SOX2⁺ SOX17⁻ TE7⁻ cells was observed at the BMP4 pattern boundary directly adjacent to the mesodermal derived cells (Figure 3c). A maximum intensity projection image of SOX2 z-slices showed a clear difference between this dense layer and the monolayer of ectodermal cells more distal to the morphogen patterns, suggesting that the differentiating mesodermal cells may affect proliferation and/or differentiation of the cells in the surrounding layer. Taken together, these data demonstrate spatially controlled co-differentiation of hPSC into the three germ layers in a single well and suggest the importance of crosstalk between developing germ layers (Figure 3a, b, c, d).

Figure 3. Localized multi-lineage differentiation of H9 hESCs into TE7⁺ mesoderm cells and SOX17⁺ endoderm cells on BMP4 and activin A patterns.

(a) Immunofluorescence images for the indicated fate markers at day 7 of differentiation and their quantification. 2.5 ng/μl BMP4 was patterned in a heart-shaped PDMS stencil while 25 ng/μl activin A was patterned in a star-shaped PDMS stencil on a TC-treated plate, incubated for 30 min at room temperature, and washed before Matrigel coating and hESC seeding. (b) Immunofluorescence images for ectodermal markers OTX2 and SOX2 on region outside of the BMP4 and activin A patterns at day 7 of differentiation. Immunofluorescence images at the boundary of BMP4 (c) and activin A (d) patterns. Mesoderm cells (Meso) were identified by TE7 staining on the BMP4 patterned area, endoderm cells (Endo) were identified by SOX17 staining on the activin A patterned area, and ectoderm cells (Ecto) were identified by SOX2 staining outside of the patterned morphogens. The normalized intensity profiles were obtained from the average fluorescence intensity values of SOX17, SOX2, and TE7 along the y-axis divided by the average fluorescence intensity value of Hoechst along the y-axis of a representative image at each boundary. The maximum intensity projection of multiple z-slices for SOX2 shows the distinct layer formation at the boundary of BMP4 patterns. Localized multi-lineage differentiation of H9 hESCs into TE7⁺ mesoderm cells and SOX17⁺ endoderm cells on BMP4 and activin A patterns was performed in at least three independent differentiations with similar results. Scale bars=1 mm, unless otherwise indicated. The statistical comparison was performed using a two-tailed Student’s t-test (* P<0.05 ** P<0.01, n=3).

2.5. Localized differentiation of hPSCs into cardiomyocytes by patterned BMP4

We next examined whether the mesodermal cells localized to BMP4 patterns could be further differentiated into cardiomyocytes. We used BMP4 patterns to direct hESCs to mesoderm and Wnt inhibition by addition of IWP2 to the culture medium at day 3 to specify cardiac mesoderm and achieve patterned CM differentiation (Figure 4a). At day 8 of differentiation, immunostaining of cardiac troponin T (cTnT) showed localization of CMs to the patterned area (Figure 4b) and H9-hTNNT2-GFP hESCs showed spontaneous contraction of the patterned CMs (Video S1-3). Cells outside the BMP4 patterns did not express cTnT but expressed SOX2 and OTX2 (Figure 4c, d), indicating an ectoderm fate. Nuclear expression of PAX6 in some of these cells is also consistent with the presence of neuroectoderm cells in this population (Figure 4d). These results demonstrate that the localized patterning of hPSCs into specific germ lineages can further be propagated toward terminally differentiated cell types while maintaining spatial patterns.

Figure 4. Localized differentiation of H9 and H9-hTNNT2-GFP hESCs into cTnT⁺ CMs on BMP4 patterns.

(a) Schematic of the CM differentiation protocol. BMP4 patterns drive mesoderm induction and Wnt inhibition directs cardiac mesoderm induction. (b) Immunofluorescence images for cTnT in CMs differentiated from H9 hESCs (2.5 ng/μl BMP4 patterned by a circle-shaped PDMS stencil on a TC-treated plate prior to washing, Matrigel coating, and hESC seeding) and H9-hTNNT2-GFP hESCs (2.5 ng/μl BMP4 pattern by a star-shaped PDMS stencil on a TC-treated plate prior to washing, Matrigel coating, and hESC seeding). (c) Immunofluorescence images for cTnT (CM marker) and SOX2 (ectoderm cell marker) in H9 hESC-derived CMs and their quantification (2.5 ng/μl BMP4 patterned by a heart-shaped PDMS stencil on a TC-treated plate prior to washing, Matrigel coating, and hESC seeding) and surrounding cells. Enlarged immunofluorescence images show cells on the BMP4-pattern, outside the pattern, and at the boundary of the pattern. The normalized intensity profiles were obtained from the average fluorescence intensity values of cTnT and SOX2 along the y-axis divided by the average fluorescence intensity value of Hoechst along the y-axis of a representative image at the pattern boundary. (d) Immunofluorescence images for OTX2 and SOX2 (ectoderm cell markers), and PAX6 (neuroectoderm marker) in the cells outside of the BMP4 pattern. Nuclear localization of PAX6 is indicated by white arrows. Localized differentiation of H9 and H9-hTNNT2-GFP hESCs into cTnT⁺ CMs on BMP4 patterns were performed in at least three independent differentiations with similar results. Scale bars=1 mm, unless otherwise indicated. The statistical comparison was performed using a two-tailed Student’s t-test (n=3) (* P<0.05 ** P<0.01, n=3).

2.6. Localized differentiation of cardiac progenitor cells into epicardial cells by patterned Wnt3a and BMP4

To mimic later developmental stage cell fate patterning, we postulated that patterned morphogens approach could be applied to hPSC-derived cardiac progenitors to generate patterns of cardiac cell types. To test this hypothesis, hPSC-derived cardiac progenitors were generated by the GSK inhibitor and Wnt inhibitor (GiWi) protocol.^[22] These cardiac progenitors will differentiate into CMs without any specific external factors, and can also differentiate into EpiCs upon activation of canonical Wnt signaling.^[22] Since previous studies showed that both BMP4 and Wnt3a play important roles in the generation of EpiCs.^[22-23] droplets containing BMP4, Wnt3a, or both BMP4 and Wnt3a were placed on a TC-treated PS plate and hPSC-derived cardiac progenitors were seeded on these patterns to assess localized EpiC differentiation. While BMP4 and Wnt3a individually induced limited EpiC differentiation, the pattern containing both BMP4 and Wnt3a induced localized EpiC differentiation (Figure 5a) indicated by the expression of WT1.

Figure 5. Localized differentiation of H9 hESC-derived cardiac progenitors into WT1⁺ EpiCs and cTnT⁺ CMs by patterning Wnt3a and BMP4 adsorption.

(a) Immunofluorescence images for the indicated fate markers 7d after cardiac progenitor seeding on morphogen patterns. Droplets containing 2 ng/μl BMP4, 40 ng/μl Wnt3a, and both BMP4 (2 ng/μl) and Wnt3a (40 ng/μl) were placed on a TC-treated plate, incubated for 30 min at room temperature, and washed before Matrigel coating and cardiac progenitor seeding. (b) Immunofluorescence images for cTnT (CM marker) and WT1 (EpiC marker) and their quantification. Wnt3a (40 ng/μl) + BMP4 (2 ng/μl) solution was placed in a circle-, heart-, or star-shaped PDMS stencils before Matrigel coating and cardiac progenitor seeding. Enlarged immunofluorescence image at the boundary of the pattern shows the two distinct cell fates (CM and EpiC) and its transition along the pattern edge. The normalized intensity profiles were obtained from the average fluorescence intensity values of cTnT and WT1 along the y-axis divided by the average fluorescence intensity value of Hoechst along the y-axis of a representative image at the pattern boundary. (c) Immunofluorescence images for cTnT (CM marker) and WT1 (EpiC marker). A heart-shaped PDMS mask was placed at the center of the well to exclude Wnt3a and BMP4 adsorption from the heart-shaped region upon application of Wnt3a (40 ng/μl) + BMP4 (2 ng/μl) solution prior to washing, Matrigel coating, and cardiac progenitor seeding. Localized differentiation of H9 hESC-derived cardiac progenitors into WT1⁺ EpiCs and cTnT⁺ CMs on Wnt3a and BMP4 patterns was performed in at least three independent differentiations with similar results. Scale bars=1 mm, unless otherwise indicated. The statistical comparison was performed using a two-tailed Student’s t-test (* P<0.05 ** P<0.01, n=3).

Using PDMS stencils, EpiC patterns were created in which EpiCs were generated on Wnt3a+BMP4 patterns while CMs differentiated outside of these patterned regions (Figure 5b). Using a PDMS fragment to exclude Wnt3a+BMP4 solution from contacting the substrate generated a CM population surrounded by EpiCs (Figure 5c). Time-lapse imaging of the cardiac progenitors generated from H9-hTNNT2-GFP hESCs showed that the CM pattern formation was observed beginning at day 4 after the cardiac progenitor seeding (Figure S8a). This cell line also exhibited localized EpiC-CM differentiation on different shaped patterns (Figure S8b, Videos S4-7). The EpiC fate was confirmed by the expression of TBX18 in the cells on the Wnt3a+BMP4 pattern (Figure S8c). Taken together, these data demonstrate that localized differentiation via morphogen patterning can be applied to more committed progenitor cells and with the choice of appropriate morphogens, cell types such as CMs and EpiCs can be spatially co-patterned on a substrate.

2.7. Effects of epicardial cells on cardiomyocytes in a spatially organized context

The epicardium and the myocardium interact during heart development. EpiCs or epicardial-derived cells affect the proliferation and maturation of primary and hPSC-derived CMs.^[24]

To examine if we could observe EpiC-CM crosstalk in a spatially controlled differentiation coculture, EpiC and CM co-patterned cells were stained for α-actinin and F-actin to enable analysis of sarcomere and myofilament structures in CMs. Epifluorescence images showed a stark difference in α-actinin and F-actin features between CMs in contact with EpiCs and CMs distal from the EpiC pattern. CMs located at the edge of the EpiC pattern had more aligned myofibers, suggestive of enhanced structural maturation (Figure 6a). Confocal microscopy of α-actinin with MLC2a and WT1 for the identification of CMs and EpiCs, respectively, showed more aligned sarcomeric α-actinin in CMs at the edge of the EpiC pattern compared to the CMs on the outside of the patterns and CMs in a monoculture control (Figure 6b). A maximum intensity projection image of z-slices at the pattern edge showed an EpiC monolayer on the patterned region and EpiCs under the volumetric CM layers. Quantification of the distance between z-lines and the angles between z-disc-crossing lines showed that CMs located adjacent to the EpiCs exhibited longer sarcomere length and increased myofibril alignment than CMs away from the boundary. Given that sarcomere length is an indicator of CM structural maturation and myofibril alignment of CMs is an important feature of CMs in vivo,^[25] these results suggest that the positive effects of EpiCs on the structural maturation of hPSC-derived CMs occurs over a short length scale. To verify this structural difference in CMs at the pattern boundary is induced by the adjacent EpiCs rather than a more general result of CM localization at the edge of a geometric pattern, we compared CMs adjacent to EpiCs with patterned CMs on BMP4 patterns surrounded by ectoderm cells as shown in Figure 4 for their sarcomere length and myofibril alignment. The CMs adjacent to ectodermal cells showed some myofibril alignment but were significantly less aligned compared to the CMs adjacent to EpiCs (Figure S9). CMs adjacent to ectodermal cells also exhibited shorter sarcomere length compared to the CMs at a boundary of EpiCs (Figure S9), suggesting that the structural maturation of hPSC-derived CMs was induced by the adjacent EpiCs. In summary, localized differentiation on morphogen patterns can be used to systematically derive spatially distinct cell populations and provides a tool to study cell-cell interactions in differentiating cell populations in a spatially-controlled context.

Figure 6. Effects of EpiCs on CM structure during localized codifferentiation on BMP4 and Wnt3a-patterned substrates.

Wnt3a (40 ng/μl) + BMP4 (2 ng/μl) solution was placed as a droplet on a ibidi 35 mm dish incubated for 20 min at room temperature, and washed before Matrigel coating and cardiac progenitor seeding. Cells were cultured 18 days prior to analysis. (a) Epifluorescence images of α-actinin and F-actin immunostaining showing the alignment of CM sarcomeres and myofibril filaments at the boundary of the BMP4 and Wnt3a pattern (indicated by arrow). Similar results were obtained in at least two independent differentiations. (b) Confocal fluorescence images for α-actinin, MLC2a (CM marker) and WT1 (EpiC marker) showing α-actinin staining at the edge of the patterns. The maximum intensity projection of multiple z-slices and the side view of these z-slices shows the spatially organized CMs and EpiCs (EpiCs indicated by arrows). (c) Quantification for the length and the alignment of sarcomeres. 25 images for each condition from three biological replicates were obtained and blinded for the test. 20 lines across 4 α-actinin⁺ z-lines per image were quantified for lengths and angles. Both lengths of 20 lines and angles between 20 lines were averaged to obtain a single value per image, thereby giving 25 lengths and angles for each condition. Means and standard deviations of these 25 lengths and angles for each condition were reported. The statistical comparison was performed using a one-way ANOVA with Tukey’s host hoc test (* P<0.05 *** P<0.0001, n=25 from three biological replicates). Scale bars are indicated in each image.

3. Discussion

In this study, we present a simple and versatile method to pattern hPSC differentiation in 2D via localized adsorption of morphogens. We showed that adsorption of morphogens, including BMP4, activin A, noggin, and Wnt3a, directed localized differentiation of hPSCs and hPSC-derived cardiac progenitors. These results suggest that patterning proteins which regulate developmental pathways can be used to spatially regulate cell fates. As a simple way to confine protein adsorption, we placed morphogen solutions as a droplet, drew patterns with a pipette, or confined protein solution using a PDMS stencil. These straightforward methods to create patterns are broadly accessible, but strategies such as a photolithography microfabrication to create more elaborate PDMS stencils have also been described ^[26] and can be applied to pattern protein adsorption on surfaces.

Patterning morphogens induced localized hPSC differentiation in a spatially controlled manner. Our approach permits systematic investigation of the effects of culture conditions such as cell density, protein concentration, and media composition, on early hPSC specification events in a spatial context. Furthermore, patterned substrates are easy to integrate into conventional cell culture platforms which allows for a simple transition of current hPSC differentiation protocols, as demonstrated in generating patterned CM populations using a protocol developed for differentiation in homogeneous monolayers.^[27] In the future, more stringent biochemical analysis of cells differentiated on adsorbed morphogens will be able to discern whether these cells are similar to the cells differentiated using standard differentiation protocols that supply morphogens in bulk medium. Sample preparation techniques such as laser capture microdissection or manual dissection could be used to isolate patterned cells.

In vitro hPSC models are an ideal candidate for studying crosstalk mechanisms between germ layers or developing tissues.^[28] One approach is to differentiate hPSCs into organoids that contain multiple cell types present in an organ. However, organoids are often heterogeneous in cell composition and/or spatial organization. Monolayer differentiation approaches have also been reported, such as codifferentiation of cardiomyocytes and endothelial cells,^[29] however these generally lack precise control of spatial resolution. Our approach offers an alternative codifferentiation patterning method. Here, we demonstrated the potential to analyze codifferentiation during early germ layer specification and later during tissue level specification. This approach enables robust and reproducible pattern generation in the context of surrounding cells allowing for study of crosstalk mechanisms during tissue development and specification. It is important to note that while the cell patterns have high fidelity to the adsorbed protein patterns, cell patterning is not perfect. For example, a small number SOX17-expressing endoderm cells were found in the mesoderm population differentiated on the BMP4 patterns in Figure 3. This is likely due to heterogeneities in differentiation, but may also result from migration of cells into the patterned region or cell detachment and reattachment.

In later stages of development, organ-specific progenitor cells receive cues to differentiate into terminal lineages. Here, we demonstrated codifferentiation of cardiac progenitor cells into EpiCs and CMs and investigated the effects of EpiCs on CM sarcomere structure. This proof-of-concept experiment suggests a role of EpiCs in CM development. Indeed, in vivo studies demonstrated that epicardial-derived signals such as FGF, insulin-like growth factor (IGF), and retinoid X receptor alpha (RXRα) signaling are required for proper myocardial growth and development.^[30] In the future, spatial gene profiling, such as RNA-fluorescence in situ hybridization (RNA-FISH) or spatial transcriptomics, and spatial proteomics can be used to study detailed molecular mechanisms of this crosstalk at the interface. Furthermore, patterned differentiation could be applied to study signaling cues that drive differentiation of progenitor cells in other organs and allow for spatially-controlled studies of different cocultured cells.

4. Conclusion

Overall, our versatile approach for localized differentiation of hPSCs and committed progenitors will provide opportunities to investigate aspects of developmental events and intercellular interaction in a spatially controlled manner in 2D. Unlike other strategies for patterning morphogens, our approach does not require complex functionalization of substrates or morphogen nor advanced instrumentation (e.g. bioprinter or microfluidic device), making it easily adoptable in biological laboratories. Additionally, our technology does not introduce hydrodynamic shear to cells, and thus allows studying localized paracrine effects in differentiating cells, a key aspect of development. Furthermore, elaborately controlled solution confinement will enable recapitulation of more complicated morphogenesis, thereby contributing to the construction of complex multicellular tissue structures. We anticipate that this patterned approach could be used as a way to fabricate in vitro models to study disease or to test drugs in a spatially controlled tissue context.

5. Experimental Section

Protein patterning:

Stock solution for each protein was prepared as follows: BSA-FITC (500 ng/μl) (Invitrogen, A23015) in sterile-filtered water (Sigma, W3500), human recombinant BMP4 (10 ng/μl ) (Invitrogen, PHC9534) in sterile-filtered HCl (4mM) containing human serum albumin (0.1%) (HSA) (Sigma, A9731), human recombinant activin A (100 ng/μl) (R&D Systems, 338-AC-010) in sterile-filtered HCl (4mM), human recombinant noggin (250 ng/μl) (R&D Systems, 6057-NG-025) in phosphate buffered saline (PBS) (Gibco, 14190-144) containing HAS (0.1%), human recombinant Wnt3a (200 ng/μl) (R&D Systems, 5036-WN-010) in PBS containing HSA (0.1%). Further dilution was done in water for BSA-FITC and PBS for other proteins to the desired final concentrations. Diluted protein solution was placed as a droplet (5μl), drawn with a pipette tip (10μl for each letter), or confined with PDMS stencil masks (20μl for each shape) on a TC-treated plate (Corning, 3513 (12-well plate) / 3526 (24-well plate)), untreated PS plate (Corning, 351143 (12-well plate)), 15mm round cover glass (Electron Microscopy Sciences, 71887-05) placed in a 24-well TC-treated plate, or 35mm μ-dish with a polymer coverslip bottom (ibidi, 81156). To create an inverse pattern, a heart-shaped PDMS mask was placed at the center of the well of a 24-well plate to exclude Wnt3a and BMP4 adsorption from the heart-shaped region upon application of Wnt3a+BMP4 solution in the well. To make PDMS stencil masks, a in house PDMS layer was prepared, cut, and punched with commercially available hole punches. Briefly, base elastomer and curing agent (Electron Microscopy Sciences, 24236-10) were mixed thoroughly with a weight ratio of 10:1, poured in a nonstick baking pan to cover the surface, de-gased in a desiccator until bubbles were removed and cured in an oven for an hour at 100°C and finally cooled for at least 24 hours at room temperature. The cured PDMS layer was detached from the pan and used to make stencil masks. The PDMS masks were firmly attached on substrates to ensure solution confinement. Protein solutions were incubated for 30 min at room temperature and aspirated. Protein solutions containing Wnt3a were incubated for 20 min at room temperature to minimize solution spreading. Immediately after aspiration, substrate was washed with PBS three times and further experiments were performed. All patterned substrates should be used immediately after patterning.

hPSC maintenance and CPC differentiation:

H9 hESC (WiCell), H9-hTNNT2-GFP hESC (H9-hTnnT2-pGZ-TD2, WiCell), and WTC11 (Allen Institute) were maintained in mTeSR1 basal medium (STEMCELL Technologies, 85851) supplemented with 5X supplement (STEMCELL Technologies, 85852). When cells reached 70-80% confluency, they were passaged onto a Matrigel (Corning, 354230)-coated 6-well TC-treated plate (Corning, 3516) with a split ratio of 1:9 using Versene (Gibco, 15040-066) (6 min incubation at 37°C). Matrigel-coated plates were prepared by diluting Matrigel in DMEM/F12 medium (Gibco, 11330-032) to a concentration of 0.08mg/ml, adding the Matrigel solution to TC-treated plates, and incubating the plates at least 2 hours at 37°C. CPCs were generated using the GiWi protocol.^[22] Briefly, hPSCs were singularized using Accutase (Innovative Cell Technologies, AT104) (10 min incubation at 37°C), resuspended in mTeSR1 supplemented with ROCK inhibitor Y-27632 (5 μM) (Selleckchem, S1049) and seeded at 250,000-400,000 cells/cm² on a Matrigel-coated 12-well TC-treated plate (Corning, 3513) (day −2). On the next day, the medium was changed with fresh mTeSR1. At day 0, cells were treated with CHIR99021 (9-12 μM) (Selleckchem, S1263) in RPMI1640 medium (Gibco, 11875-093) containing 1X B27 supplement minus insulin (A18956-01) (RPMI B27-) for 24 hours. At day 1, medium was changed with fresh RPMI B27-. At day 3, cells were treated with IWP2 (5 μM) (Tocris, 3533) in fresh RPMI B27-. At day 5, medium was changed with fresh RPMI B27-. At day 6, cells were detached using Accutase (10 min incubation at 37°C) and cryopreserved in a cryopreservation medium (60% RPMI1640 medium containing 1X B27 supplement (17504-044) (RPMI B27+), 30% FBS (Peak Serum, PS-FB1), 10% dimethyl sulfoxide (DMSO) (Sigma, D2650), 5 μM Y-27632)

Localized differentiation of hPSCs:

For localized mesendoderm differentiation, BMP4 solution (2-2.5 ng/μl) was placed with various patterns on different substrates. For glass surface, 1.5 ng/μl BMP4 solution was used. Noggin solution (5 ng/μl) was placed as a droplet to show inhibitory effect of noggin on BMP4-induced mesendoderm differentiation. After washing, substrate was coated with Matrigel as described previously for 2 hours. Glass was coated with Matrigel overnight. Synthemax (0.025 mg/ml) (Corning, 3535) diluted in water was added to a washed TC-treated plate and the plate was incubated for 2 hours at room temperature prior to cell seeding. Truncated recombinant human vitronectin (5 μg/ml) (Gibco, A14700) diluted in PBS was added to a washed TC-treated plate and the plate was incubated for 1 hour at room temperature prior to cell seeding. After ECM coating, the ECM solution was aspirated and hPSCs were singularized using Accutase (10 min incubation at 37°C), resuspended in mTeSR1 supplemented with Y-27632 (5 μM), and seeded at 700,000-800,000 cells/cm² on coated substrates (day 0). On the next day (day 1), medium was changed with fresh mTeSR1. When the noggin solution was placed as a droplet, mTeSR1 supplemented with BMP4 (10 ng/ml) was used as medium at day 0 and day 1. For the ALK inhibitor test, A83-01 (1 μM) (Tocris, 2939), SB-431542 (10 μM) (Tocris,1614), or LDN-212854 (2 μM) (Tocris, 6151) was added to mTeSR1 at day 0. At day 2, cells were immunostained for analysis.

For localized definitive endoderm differentiation, activin A solution (25 ng/μl) was placed with various patterns on a TC-treated plate. Matrigel coating and hPSC seeding was performed as the same way described previously. Medium was changed daily with RPMI1640 medium supplemented with FBS (2%). For the ALK inhibitor test, A83-01 (1 μM) (Tocris, 2939), SB-431542 (10 μM) (Tocris,1614), or LDN-212854 (2 μM) (Tocris, 6151) was added to mTeSR1 at day 0. At day 3, cells were immunostained for analysis.

For multi-lineage differentiation, BMP4 (2.5 ng/μl) and activin A (25 ng/μl) solution were placed in a heart-shaped and a star-shaped hole, respectively, of PDMS masks in a single well of a 12-well TC-treated plate. Matrigel coating and hPSC seeding was performed as the same way described previously. Medium was changed daily with RPMI1640 medium supplemented with FBS (2%). At day 3, 5, and 7, cells were immunostained for analysis.

For localized CM differentiation, BMP4 solution (2.5 ng/μl) was placed in PDMS stencil masks on a TC-treated plate. Matrigel coating and hPSC seeding was performed as the same way described previously. At day 1 and day 2, medium was changed with fresh RPMI B27-. Afterwards, CM differentiation was performed based on the GiWi protocol.^[27b] At day 3, cells were treated with IWP2 (5 μM) in fresh RPMI B27-. At day 5, medium was changed with fresh RPMI B27-. At day 7, medium was changed with fresh RPMI B27+. At day 8, beating phenotype of patterned CM derived from H9-hTNNT2-GFP hESC was recorded using a fluorescence microscope (Nikon, Eclipse Ti2-E) equipped with a heating stage top incubator (Tokai Hit, Tokai Hit STX Series Stage Top Incubator System). Cells were also immunostained for analysis at day 8.

For localized EpiC differentiation, BMP4 (2 ng/μl) + Wnt3a (40 ng/μl) solution was placed in PDMS stencil masks either on a TC-treated plate or an ibidi 35mm polymer dish. Matrigel coating was performed as the same way described previously. Cryopreserved CPCs were thawed and seeded at 500,000-550-000 cells/cm2 in LaSR medium (advanced DMEM/F12 (Gibco, 12634-010) supplemented with ascorbic acid (60 μg/ml) (Sigma, A8960) and Glutamax (2.5 mM) (Gibco, 35050-061)) supplemented with Y-27632 (5 μM). Medium was changed daily with fresh LaSR medium until analysis. Time lapse imaging of CPCs derived from H9-hTNNT2-GFP hESC during localized EpiC differentiation was performed using the fluorescence microscope equipped with the heating stage top incubator. 18 days after CPC seeding, cells cultured on a ibidi polymer dish were immunostained and imaged using a confocal microscope (Nikon, A1R-SI+) for the analysis of sarcomere structure.

Immunofluorescence:

For BMP4 staining, Alexa Fluor™ 488 Tyramide SuperBoost™ Kit (Invitrogen, B40941) was used according to the manufacturer’s instruction. Briefly, BMP-patterned plate was incubated in goat serum (10%) for 1 hour at room temperature and incubated in BMP4 antibody (OriGene, TA500014) diluted in 10% goat serum (1:100 dilution) overnight at 4°C. On next day, the surface was rinsed with PBS three times and incubated in poly-HRP-conjugated secondary antibody overnight at 4°C. Again, the surface was rinsed with PBS three times and incubated in the tyramide working solution for 8 min at room temperature. Then, the same volume of Reaction Stop Reagent was added to the surface followed by three times of PBS washing before observation under the fluorescence microscope.

For immunostaining of other protein markers, cells were fixed in paraformaldehyde (4%) (Electron Microscopy Sciences, 15710-S) in PBS for 15 min at room temperature, washed with PBS three times, and incubated in diluted primary antibodies in nonfat dry milk (5%) (Santa Cruz Biotechnology, sc-2324)/Triton X-100 (0.4%) (Fisher Scientific, BP151-500) PBS overnight at 4°C. Then, the cells were washed with PBS three times and incubated in diluted secondary antibody in nonfat dry milk (5%)/Triton X-100 (0.4%) PBS for 30 min at room temperature. After secondary antibody staining, the cells were washed with PBS three times, incubated in Hoechst 33342 (2μg/ml) diluted in PBS for 5 min at room temperature, washed with PBS two times, and analyzed under the fluorescence microscope.

Primary antibodies and corresponding dilution ratio were listed as follows: anti-BMP4 (OriGene, TA500014/1:100), anti-BRACHYURY (R&D Systems, AF2085/1:100), anti-MIXL1 (Proteintech, 22772-1-AP/1:100), anti-Laminin (Novus Biologicals, NB300-144/1:100), anti-SOX17 (R&D Systems, AF1924/1:100), anti-SOX2 (Invitrogen, 14-9811-82/1:100), anti-NANOG (Cell Signaling Technology, 4893S/1:1000), anti-FOXA2 (Invitrogen, 720061/1:500), anti-MESP1 (R&D Systems, MAB9219-100/1:100), anti-TE7 (EMD Millipore, CBL271/:100), anti-OTX2 (R&D Systems, AF1979/1:100), anti-cTnT (Invitrogen, MA5-12960/1:200), anti-PAX6 (EMD Millipore, AB2237/1:250), anti-WT1 (Invitrogen, MA5-32215/1:100), anti-TBX18 (Abcam, ab115262/1:200), anti-α-actinin (Sigma, A7811/1:1000), anti-MLC2a (Synaptic Systems, 311011/1:200), anti-F-actin (pre-conjugated to DyLight 594) (Invitrogen, 21836/1:50).

Secondary antibodies and corresponding dilution ratio were listed as follows: donkey anti-goat IgG Alexa Fluor 488 (Invitrogen, A11055/1:1000), goat anti-mouse IgG1 Alexa Fluor 488 (Invitrogen, A21121/1:1000), chicken anti-rabbit IgG Alexa Fluor 488 (Invitrogen, A21441/1:1000), donkey anti-rat IgG Alexa Fluor 555 (abcam, ab150154/1:1000), goat anti-mouse IgG2b Alexa Fluor 555 (Invitrogen, A21147/1:1000), donkey anti-mouse IgG1 Alexa Fluor 647 (Invitrogen, A21240/1:1000), donkey anti-rabbit IgG Alexa Fluor 647 (Invitrogen, A31573/1:1000).

Image analysis:

Fluorescence intensity of BMP4 was quantified using Fiji/ImageJ software. From gray scale images, mean gray value of BMP4-patterned region was subtracted by mean gray value of the background. Four independent droplets for each concentration were quantified. The statistical comparison was performed using two-tailed Student’s t test (***P<0.001, n=4) between untreated and TC-treated condition of each concentration. Intensity profiles at pattern boundaries were obtained using Fiji/ImageJ software. A representative image at each pattern boundary was transformed into gray scale images, then vertically averaged pixel intensities for all pixels along x-axis were obtained for each marker. Averaged intensities for each marker were divided by the corresponding averaged intensities of nuclei and plotted to get normalized intensity profiles. For quantification of fluorescence intensity, background subtraction was performed for fate marker and nucleus staining, then mean fluorescence intensity of each fate marker in cells either on morphogen patterns or outside patterns were divided by mean fluorescence intensities of nuclei. These intensities were normalized to the intensities of outside patterns. Maximum intensity projection images of z-stack images were obtained using NIS-Elements imaging software (NIS-Elements Advanced Research version 5.20.01). For analysis of length and alignment of sarcomeres, 25 images for each condition were obtained from two or three biological replicates. The images were blinded using an ImageJ plugin ‘Blind Analysis Tools’ generating encrypted file names with a mapping file. For each image, 20 lines across 4 α-actinin⁺ z-lines were quantified using NIS-Elements imaging software to get the lines’ lengths and positional information. The lengths of the 20 lines were averaged to get a representative value for length of a single image. The 25 lengths from 25 images of each condition were averaged and standard deviations were calculated to get the reported values. From the positional information of all 20 lines, angles between those lines were calculated. Specifically, a line was chosen as a reference line to obtain 19 different angles between the reference line and the other lines. These 19 values were averaged to get an angle difference value of a single line. The same calculation was performed for all 20 lines to get 20 different angles. These 20 angles were averaged again to get a representative value for angle between sarcomeres of a single image. The 25 angles from 25 images of each condition were averaged and standard deviations were calculated to get the reported values. The statistical comparison was performed with 25 lengths or 25 angles per each condition using a one-way ANOVA with Tukey’s host hoc test (* P<0.05 *** P<0.0001, n=25 from three biological replicates). The statistical comparison between CM-EpiC and CM-ectoderm cell pattern boundaries was performed using a two-tailed Student’s t-test (*** P<0.0001, n=25 from two or three biological replicates).

Statistical analysis:

All data is presented as mean ± standard deviation. The statistical comparison for fluorescence intensities of BMP4 was performed using two-tailed Student’s t test (*** P<0.001, n=4). The statistical comparison for fluorescence intensities of fate markers was performed using two-tailed Student’s t test (* P<0.05 ** P<0.01, n=3). The statistical comparison for length and alignment of sarcomeres was performed using a one-way ANOVA with Tukey’s host hoc test (* P<0.05 *** P<0.0001, n=25). The statistical comparison for length and alignment of sarcomeres of CM-EpiC and CM-ectoderm cell pattern boundaries was performed using two-tailed Student’s t test (*** P<0.0001, n=25). ORIGIN (one-way ANOVA) and EXCEL (Student’s t test) were used to perform statistical analysis.