We developed a micropatterning method used for human pluripotent stem cell (hPSC) culture and ectoderm induction (Figure 1A). Briefly, the micropatterning method allows the generation of cell-adhesive areas of defined size and shape via the use of photomasks (Figure 1A). First, we confirmed that we were able to generate hPSC colonies of 1 mm diameter on a glass coverslip (12 mm diameter) glass coverslip (Figure 1B) corresponding to the photo-preserved areas.

To find the best experimental conditions to generate micropatterns, we kept the size and shape of the micropatterns constant (round shapes of 1 mm diameter) and tested two different influencing factors: 1) Matrigel growth factor reduced (MRF) coating concentration to optimize cell adhesion, and 2) the time of ROCK (Rho-associated protein kinase) inhibitor (Ri) exposure to enhance cell viability after seeding. We tested three different concentrations of MRF (0.5, 1.0, and 5.0% v/v) (Figure 1C) and three different times of Ri withdrawal after seeding (30, 60, and 240 min) (Figure 1D) and counted the number of well-organized colonies over time (i.e., a colony of 1 mm diameter with intact borders). From the results of these tests, we selected an MRF concentration of 1% and Ri treatment timing of 240 min. Next, we evaluated the stability of the micropatterns over time, since the integrity of the micropattern’s geometry is a requirement to increase the reproducibility of cell patterning. We longitudinally analyzed the appearance of the colonies and found that we were able to maintain at least 10 colonies showing an intact round geometry for up to 5 days (Figure 1E).

During development, the ectoderm is patterned by a combination of BMP and WNT signaling. We hypothesized that upon specific and time-regulated stimulation with BMP, TGFβ, and WNT inhibitors coupled with the controlled administration of exogenous BMP, the micropatterned hPSCs could self-organize into different cells fates within the ectoderm cell populations. First, we confirmed the ability of dual SMAD inhibition in promoting the pluripotency exit and acquisition of homogeneous neuroectoderm fate in standard culture conditions (Supplementary Figure S1A). We found that the neural ectodermal markers PAX6 and SOX1 started to be expressed significantly from day 3 and reached the peak between days 4 and 5, while the pluripotency marker OCT4 was highly expressed on day 1, decreased during the following days, and completely disappeared between days 4 and 5 of the culture (Supplementary Figures S1B,C). SOX2, instead, was stably expressed during the whole 5 days since it is a shared marker between the pluripotency and neural ectoderm (Feng and Wen, 2015) (Supplementary Figures S1B,C).

Then we applied TGFβ, WNT, and BMP inhibition to micropatterned hPSCs using a combination of small molecules named hereafter APD (i.e., Α83-01, PNU74654, and Dorsomorphin). As shown in Figures 2A–D, upon ectoderm induction on micropatterns of 1 mm diameter, we obtained radial segregation of the neural ectoderm in the outer region of the pattern (Figure 2A) with a peak of fluorescence intensity associated with PAX6 at 370 μm from the center (Figure 2C). In addition, we observed the appearance of a non-CNS ectoderm population (defined by AP2α positivity) (Tchieu et al., 2017) in the center of the pattern (Figure 2B) with a peak fluorescence intensity associated with AP2α at 200 μm from the center (Figure 2C). Thus, instead of a homogeneous layer of neural ectoderm as observed in standard culture conditions (Supplementary Figures S2B,C), the geometric confinement resulted in the generation and segregation of two different cell populations: a neural ectoderm in the outer region and a non-CNS ectoderm in the inner part of the pattern (Figure 2D).

Then we tested whether different combinations of WNT and BMP signals provided to geometrically confined hPSC colonies in a time-regulated fashion affect the appearance and spatial segregation of multiple ectodermal fates (Figures 2E–H). We tested six different conditions based on previously published protocols for human ectoderm patterning (Tchieu et al., 2017; Xue et al., 2018; Britton et al., 2019). As expected, changing the stimuli had a substantial impact on the micropattern formation and multiple ectoderm fates’ acquisition: continuous TGFβ and BMP inhibition (#1) led to the formation of an external ring of the neural ectoderm (PAX6 positive cells); continuous TGFβ, BMP, and WNT inhibition (#2) led to the segregation of two distinct populations (neural ectoderm, PAX6 positive, and non-neural ectoderm AP2α positive) with a central core of the non-neural ectoderm and an outer ring of the neural ectoderm; two-step induction with continuous TGFβ inhibition and inhibition of BMP from day 3 in DMEM-based medium (#3) abolished the formation of both populations, whereas the same combination of small molecules in an N2/B27-based medium (#5) led to the formation of a homogeneous neural ectoderm pattern; two-step induction with continuous TGFβ inhibition and the addition of BMP4 from day 3 in the DMEM-based medium (#4) led to the formation of a central core of the non-neural ectoderm within the micropatterns, whereas the same combination of small molecules in the N2/B27-based medium (#6) led to the segregation of the two distinct ectoderm populations but with inverted localization compared to condition #2, with the vast majority of the pattern covered by the neural ectoderm and a thin outer ring of the non-neural ectoderm (Figures 2F,H).

Thus, not only the temporally controlled perturbation of BMP and WNT signaling via specific inhibitors but also the stimuli contained within the basal media, among which BMP (secreted by MEFs or present in the KSR of DMEM) (Xu et al., 2005) could play a major role in defining the presence and spatial distribution of different fates within the ectoderm lineage.

Then we investigated whether this micropatterning technology would be a suitable platform to create multilayer systems and to investigate the interaction between cells belonging to different germinal layers in terms of selective cell sorting (Minn et al., 2020). In particular, we tested whether the hPSC-derived meso-endoderm can undergo cell sorting in vitro on pre-existing patterned ectoderm structures as previously observed for human gastruloid-derived cells (Minn et al., 2020). To distinguish the two cell populations, the meso-endoderm was derived from a GFP-expressing hPSC line.

We used a published protocol to generate a homogeneous meso-endoderm layer of cells in 48 h (Supplementary Figure S1A) (Giobbe et al., 2015) and confirmed the prevalence of the mesoderm marker T/Brachyury at 24 h of differentiation and the endoderm marker SOX17 at 48 h of differentiation (Supplementary Figures S1B,C).

Based on these data, we harvested meso-endoderm cells at 24 h of differentiation and plated them onto 5-day-old micropatterned ectoderm structures. We tested three of the protocols described in Figure 2F, which gave the most extreme results (i.e., #2 with an outer ring of PAX6 and an inner core of AP2α; #4 with only an inner core of AP2α; and #6 an inner core of PAX6 and an outer ring of AP2α) and co-cultured with meso-endodermal cells for an additional 3 days (Figure 1A).

To monitor the distribution of GFP-positive meso-endoderm cells under the three different micropattern conditions, we followed the co-cultures over time (Figures 3B–D). Immediately after seeding (45 min and 2 h), GFP-positive cells were evenly distributed in all three conditions (Figures 3B–D). Already at 1 day after seeding, we observed major differences which were maintained until day 3. In condition #2 (Figure 1B), the GFP-positive meso-endoderm was localized in the middle and at the outer border of the pattern (Figure 3B). In condition #4 (Figure 3C), the GFP-positive meso-endoderm spread all over the pattern. In condition #6 (Figure 3D), the GFP-positive meso-endoderm created an outer ring. Although the spatial distribution of the meso-endoderm differed among the three conditions, they all shared the common tendency of meso-endoderm cells to avoid regions occupied by the neural ectoderm (PAX6-positive) and accumulate in regions occupied by the non-CNS ectoderm (AP2a) (Figures 1B–D and Figures 3B–D).

During in vivo development, the ectoderm and endoderm are spatially separated by the presence of the mesoderm (Solnica-Krezel and Sepich, 2012). Human gastruloids dissociated and reseeded on micropatterns showed the ability to self-organize in segregated cell clusters with ectodermal cells more associated with the mesoderm and completely segregated from the endoderm, probably through selective cadherin expression (Minn et al., 2020). Based on this evidence, we tested whether the mesoderm (T/Brachyury+) and the endoderm (SOX17+) populations show a differential distribution on ectoderm patterns (condition n2) after 3 days of co-culture.

We found that both GFP + SOX17 + endoderm and GFP + T/Brachyury + cells preferentially clustered at the edges of the areas of micropatterns occupied by the non-CNS ectoderm (Figure 4A), and GFP + T/Brachyury + cells partially overlapped with regions covered by the neural ectoderm as previously observed (Fagotto, 2014) (Figure 4B).

A 12-mm diameter glass coverslip was pre-treated with a plasma cleaner machine (Harrick Plasma) for 3 min (3 × 10⁻¹ mbar) to oxidize the surface. Then 3–4 drops of solution A (10 μl of 3-(trimethoxysilyl) propyl methacrylate, 950 μl ethanol, and 50 μl acetic acid) were added onto the plasma-treated surface. After 3 min of incubation at room temperature, the coverslip was washed with ethanol two times and then let dry. A photo-patterning solution (solution B) was prepared right before use: 100 μl 20mg/100 μl IRGACURE 2959 (Ciba Specialty Chemicals) in methanol was diluted in 900 μl 8% (w/v) acrylamide (Sigma, 50 mM HEPES) and degassed for 20 min. Ten μl solution B was applied onto a 12-mm glass coverslip covered by a photo-mask and exposed for 40 s to UV irradiation with a distance of 5.5 cm between the glass coverslip and UV source. A functionalized glass coverslip was then washed with deionized water two times and stored at room temperature until use.

In the case of micropatterning within microfluidic devices, instead of using a 12-mm glass coverslip, a thick standard slide for microscopy was used and micropatterns were generated in correspondence with the areas that have been then covered by the microchannels.

Before cell seeding, the micropatterned glass coverslips were pre-coated with 50 μg/m Poly-L-lysin (Sigma) for 2 h at room temperature and 1% Matrigel overnight at 4°C. A washing procedure with MilliQ water and DPBS was performed after each coating. The glass coverslips were stored in the last wash solution at 4°C. After coating, the glass coverslips should be used within 2 weeks.

All experiments were performed with the H9 ESCs cell line. For routine culture maintenance, H9 ESCs were cultured in a StemMACSTM iPS-Brew XF pluripotent stem cell medium and passaged 1:5 to 1:10 every 3–5 days via EDTA dissociation. Culture plates were coated with 0.5% Matrigel (Corning) and incubated at room temperature for at least 2 h. Coated plates were stored at 4°C and pre-warmed at 37°C immediately before use.

Human PSCs were pre-adapted to MEF-CM containing 20 ng/ml Fgf2 for one passage. Cells were detached as single cells and suspended in the MEF-CM medium supplemented with 10 μM Rock-inhibitor Y27632 (Miltenyi Biotech). The MEF-CM differentiation basal medium comprised DMEM high glucose (GIBCO), 20% KnockOut Serum Replacement (KSR, GIBCO), 1x GlutaMAX™ (GIBCO), 1x MEM NEAA (GIBCO), 100 μM β-mercaptoethanol (GIBCO), 2% B27-without vitamin A (GIBCO), and 1 mM pyruvate (Sigma) and conditioned overnight with inactivated MEFs. PSCs were seeded at 0.5 × 10 cells per well (Minn et al., 2020) onto the functionalized glass coverslips within 24-well plates. After 3 h, the micropatterns were washed twice with pre-warmed PBS to remove cells attached outside the micropatterns and the medium was replaced with an ectoderm induction medium containing a different combination of small molecules and was changed daily.

Three different basal media were used to induce ectoderm differentiation, which are as follows:

1) The DMEM differentiation basal medium containing DMEM/F12 (GIBCO), 15% KnockOut Serum Replacement (KSR, GIBCO), 1 × GlutaMAX™ (GIBCO), 1 × MEM NEAA (GIBCO), and 50 μM β-mercaptoethanol (GIBCO).

2) The MEF-CM differentiation basal medium.

3) N2B27 medium basal medium containing 25 ml DMEM/F12 medium, 25 ml neurobasal medium (GIBCO), 0.25 ml N2, 0.5 ml B27 without vitamin A, 0.5 ml GlutaMAX™, and 50 μM β-mercaptoethanol.

To induce ectoderm fate, the basal media were supplemented with different small molecules’ cocktail:

(1) 2 μM A83-01 (Tocris), 2 μM PNU-74654 (Tocris), and 2 μM dorsomorphin (Sigma) named APD.

(2) 10 μM SB431542 (Stemgent) and 100 nM LDN193189 (Sigma) named SB + LDN.

H9 hESCs were seeded at high density as a single-cell onto Matrigel-coated Petri dishes. Upon attaining 90% confluence (usually the day after seeding), the medium was switched to MEF-CM supplemented with 1) 2 μM A83-01 (Tocris), 2 μM PNU-74654 (Tocris), and 2 μM dorsomorphin (Sigma) or 2) 10 μM SB431542 (Stemgent) and 100 nM LDN193189 (Sigma). Cells were fed daily and maintained till day 5 to evaluate the efficiency of neural induction.

Meso-endoderm differentiation was induced in a 6-well Petri dish. H9 hESCs were seeded (1:5 passage) onto a 2.5% Matrigel-coated plate as clumps. When the cells reached 30–50% confluence (usually the day after split), the meso-endoderm induction was initiated by replacing the expansion medium with an induction medium comprising RPMI 1640 (GIBCO), 2% B27-ins (GIBCO), and 1 × MEM NEAA (GIBCO) supplemented with 100 ng/ml Activin A (R&D), 10 ng/ml BMP4 (R&D), 20 ng/ml bFGF (Peprotech), and 3 μM CHIR99021 (Miltenyi Biotech).

Micropatterned neuroectoderm cell fate was induced as described previously after 5 days of cell differentiation. Meso-endoderm cell induction was initiated on day 3 of neuroectoderm differentiation to keep in step. Meso-endoderm cells were cultured in a 6-well Petri dish with the RPMI medium, through single cell splitting, 0.25 × 10⁶ cells (in 0.5 ml MEF-CM medium) were seeded on top of the micropatterned neuroectoderm cells. During the ectoderm and meso-endoderm co-culture procedure, patterns were grown in the MEF-CM medium without any small molecules. The medium was changed daily for days.

Cells were rinsed once with PBS, fixed in 4% paraformaldehyde for 30 min, and rinsed three times with PBS at room temperature. A blocking solution was made with 0.1% Triton-X and 5% normal donkey serum (Jackson ImmunoResearch, West Grove, PA) in PBS, and cells were incubated for 1 hour at room temperature before staining with primary antibodies in a blocking solution at 4°C overnight. The cells were then washed three times in PBS 0, 1% triton before being incubated for 2 h at room temperature with secondary antibodies, and 0.1 mg/ml DAPI (Invitrogen, Carlsbad, CA) in a blocking solution for 1 h. Finally, cells were washed three times in PBS 0, 1% triton and mounted using Vectashield Antifade Mounting Medium (Vector Laboratories, Burlingame, CA). All primary and secondary antibodies used are listed in Table 1.

Fluorescence images were acquired using a confocal TCS SP5 microscope (Leica Microsystems). Image processing was performed with ImageJ software (NIH).

All micropatterning experiments and differentiation in standard culture conditions were performed at least three times. The data and analyses in each micropattern induction figure belong to one representative experiment. The sample size was not pre-determined and no statistical tests were used to determine the significance of results on micropatterned colonies. Circular colonies with a non-radially symmetric cell density pattern at the end of the treatment were excluded from analyses. The average intensity of a marker was calculated for each cell as the average of the immunofluorescence intensity in that cell normalized to the intensity of DAPI staining in the same cell.