Self-organization of human dorsal-ventral forebrain structures by light induced SHH

Engineering a collection of optogenetic ePiggyBac vectors

In order to provide light modulation of gene expression to human developmental studies, we re-engineered our original transposon ePiggyBac vector^30,31 to conditionally express a light-inducible Cre-recombinase enzyme that takes advantage of the Magnets dimerization system (Magnet-CRE)²⁰ (Fig. 1A). This allows for a stable integration of a blue-light dependent CRE enzyme in the genome. To avoid culturing cells in the dark, minimize leakage and to gain better control of light-sensitivity, we controlled the light-CRE enzyme using a Dox-inducible promoter and a second T2A peptide to improve the separation of its components (Fig. 1A, left panel). We paired this vector with a receiver ePiggyBac that carries two sequential ORFs (Red and Green modules) to be regulated by LoxP recombination in a mutually exclusive manner (Fig. 1A, right panel). Both vectors were stably transfected into our female XX hESC line, RUES2 (NIHhESC-09-0013) and stimulated with DOX and blue-light through a light-blocking photomask (Fig. 1B, Supplementary Fig. 1A–C). Dox- and Blue-light induced hESCs showed robust expression of the Green module in patterns imposed by the shape of the photomask (Fig. 1B, C, Supplementary Fig. 1B, C). The Green module expressing cells showed a highly reproducible pattern of expression that tightly correlate with the photomask that slightly increase over time, probably due to cell proliferation (Supplementary Fig. 1D). The efficiency of light conversion by single-cell fluorescence measurement shows that 24 h of blue-light stimulation converts 78.3% of the total cells, while controls kept in the dark, or in absence of Dox, show less than 1% of Green positive cells (Fig. 1D). The activation of the green module depends on the duration of blue-light stimulation. Pulsed blue-light for 600 cycles (24 h) is the most efficient treatment without inducing cell death, as measured by CASP3 activation (Fig. 1D, Supplementary Fig. 1E). The blue light-dependent induction of the Green module was also validated by measuring RNA levels by qRT-PCR in whole-illuminated samples (Fig. 1E). LoxP genomic recombination upon Dox and light treatment is shown by amplicon Sanger sequencing of the selected genomic region (Fig. 1F). The light-induced pattern of gene expression is consistently and stably maintained over six days in the absence of continuous light stimulation (Supplementary Fig. 1F). Light induced gene expression modulation was not confined to a single hESC line as another of our hESC line (RUES1, genetic background, male XY, NIHhESC-09-0012) responded in the same manner (Supplementary Fig. 1G). Collectively, these experiments demonstrate that pairing a light inducible CRE enzyme with a stable and drug inducible ePiggyBac vector allows for a rapid and efficient spatiotemporal control of exogenous gene expression in hESCs.

A Schematization of the collection of optogenetic ePiggyBac vectors to conditionally express a photoactivatable Cre-loxP recombinase vector. Left panel. Puromycin selectable and Dox inducible piggyBac transposon carrying the CRE-MAGNETs system. Dox treatment induces the CRE-MAGNETs system which reconstitutes a fully active CRE in the presence of blue light. Right panel. Blasticidin selectable LoxP exchangeable dual colors piggyBac transposon. The first ORF (Red module) is constitutive expresses while the second ORF (Green module) depends on CRE-LoxP recombination. Gray Box is the possible cargo protein taking advantage of the T2A peptide. B Photomask induced light patterns showing the expression of the Red module (dsRed), Green module (NG) and merge composite channel. Dark control and spatially localized activation in presence and absence of DOX with different spatial features are shown (1000 μm, 500 μm, 250 μm, 125 μm). Scale bar: 100 μm. C Cumulative fluorescence intensity analysis (line profile) over the x-axis. x-axis displays the linear distance in μm. y-axis reports the cumulative fluorescent intensity profile in arbitrary units. Line profile shows the average (line) and SD (area) for the Green fluorescent channel. Quantification after 600 cycles (24 h) of pulsed Blue light (n = 3 biologically independent samples). D Single-cell fluorescent intensity quantification. Conditions: presence or absence of Dox with concomitant Dark or Light stimulation. We titrated the time of light stimulation using different pulsed light conditions. 1 cycle of pulsed light is equal to 20 s Light-ON and 120 s Light-OFF. The light stimulation intervals are 1 cycle, 25 cycles (1 h), and 600 cycles (24 h), data are displayed as a boxplot where center lines show the medians, box limits indicate the 25th and 75th percentiles and whiskers extend to minimum and maximum values (n = 5 biologically independent samples). E qRT-PCR showing mRNA induction of the MAGNETs system and light-induced expression of the Green module, data are displayed as mean and SD (n = 3 biologically independent samples). F Sanger sequencing of the genomic region flanking the LoxP site showing the recombination of the Red module and the generation of the Green module (Red box: Pcag-promoter, Blue box: LoxP, Green box: NG).

Light induced dorso-ventral hESC-derived neural tissue

In order to study the M-L and D-V aspects of neural patterning in human models, we tested the ability of light stimulation to generate a localized SHH organizing center. We devised a LoxP inducible Green module to be co-expressed with SHH at the mRNA level, while producing two separate proteins using a T2A peptide. This setup can be regulated by blue-light stimuli through LoxP recombination (Fig. 2A). hESCs (RUES2) were differentiated using dual SMADs inhibition (SB431542 and LDN193189) to induce an anterior forebrain fate. The application of DOX at day 0 confers light sensitivity for the first two days of differentiation (Fig. 2A). Neural induction using dual SMADs inhibition has been shown to generate progenitors representative of the embryonic neuroepithelium^32,33,34. At this stage, a set of defined markers can be used to decipher patterning at various distances from the SHH source: PAX6 marks the most dorsal population, FOXA2 the floor plate and NKX2.1 the ventral neural progenitors in the basal telencephalon and hypothalamic primordia at E12.5 (Supplementary Fig. 2A)^16,32.

A Left panel. Schematization of Green module T2A human SHH. This setup allows conditional expression of SHH and fluorescent visualization of the light-induced cells. Right panel. Schematization of Blue-light stimulation and neural induction reporting analyzed time points. Neural induction is induced by inhibiting TGFβ signaling by dual SMADs inhibition that is maintained during the entire course of differentiation (14 days). Dox treatment starts at day “0” and it is washed-out at day “2”. B FOXA2 immunostaining in light induced SHH cells and NG-CNTRL during differentiation at day 2. DAPI-Gray, NG-Green, FOXA2-Yellow, Merge-Composite. Dashed cyan line indicates the border of SHH producing cells. Scale bar: 100μm. C Day 2 single-cell fluorescence quantification displayed as a scatterplot, reporting FOXA2 and NeonGreen (NG) intensity for NG-CNTRL and NG-T2A-SHH. D Immunostaining time-course analysis of the dorsal and ventral fates, revealing NG-Green (Organizer), FOXA2-Yellow (Ventral floor plate), NKX2-1-Magenta (Ventral neural progenitors), PAX6-Red (Dorsal neural progenitors), Merge-Composite at day 7 and day 14 during neural differentiation in response to ligh patterned SHH. Dashed cyan line indicates the border of SHH producing cells. Scale bar: 100 μm. E Single-cell fluorescence quantification of NG-T2A-SHH induced cells displayed as combined scatterplot and density histogram at day 7 (top) and day 14 (bottom). x-axis report FOXA2 while the y-axis the NKX2-1 fluorescence intensity profile. Each dot represents an individual cell that has been color coded according to its green module status. The green dots represent the NG-T2A-SHH positive cells, while the blue dots are NG-T2A-SHH negative cells. F Cumulative fluorescence intensity analysis (line profile) over the x-axis. x-axis displays the linear distance in μm. y-axis reports the cumulative fluorescent intensity profile in arbitrary units for each channel. Line profile shows the average (line) and SD (area) for each channel. The line profile is color-coded as the immunofluorescent channels, NG-Green, FOXA2-Yellow, NKX2-1-Magenta, PAX6-Red. Top panel. Day 7 quantification (n = 4 biologically independent samples). Bottom panel. Day 14 quantification (n = 4 biologically independent samples).

Blue light was shone on day 1 during neural induction through a 1 mm rectangular mask for 24 h to produce SHH and a green fluorescent protein (NG-T2A-SHH) or a control fluorescent protein (NG-CNTRL) in a spatially restricted domain. Light stimulation induced the expression of the Green module in both NG-CNTRL and NG-T2A-SHH lines (Fig. 2B). At day 2, FOXA2 positive cells were specifically induced by the NG-T2A-SHH but not in the NG-CNTRL line (Fig. 2B, C, Supplementary Fig. 2B). FOXA2 positive cells were detected in cells secreting SHH as well as in cells just next to the SHH secreting domain, providing functional evidence of autocrine as well as paracrine SHH activity (Fig. 2B, C, Supplementary Fig. 2B, C). Examination of the NG-T2A-SHH light-induced cells after 7 and 14 days display FOXA2⁺, NKX2-1⁺ and PAX6⁺ cells that are organized in discrete domains while NG-CNTRL cells acquire PAX6⁺ default neural fate, in absence of ventral cell types (Fig. 2D–F; Supplementary Fig. 3A–C). The expression NG-T2A-SHH is stably maintained in absence of light over the course of the differentiation (Supplementary Fig. 3D, E). SHH light induction unveiled M-L self-organization of the neural populations under the influence of an organizing center. Interestingly, at day 7, a population of cells co-expressing FOXA-2 and NKX2-1 was detected within and near by the light-induced organizer (Fig. 2D–F, Supplementary Fig. 3A). These FOXA2⁺/NKX2-1⁺ double-positive cells have been suggested to be human specific, as have not been detected in the mouse, while they are present in the ventral forebrain in human fetal samples at PCW5.5¹⁶. PAX6, NKX2-1 and FOXA2 domains gradually segregate over time inside and outside the SHH induced domain, with PAX6⁺ cells localized the farthest from the SHH source (Fig. 2D–F). Single-cell quantification shows that at day 7, the organizer induces population of cells double positive for NKX2-1⁺/FOXA2⁺, both cell autonomously and non-cell autonomously (Fig. 2E, upper panel). A fraction of light converted cells, express high levels of FOXA2 but not NKX2-1 (Fig. 2D, E, upper panel). At day 14, the ventral cellular populations induced by SHH differentiate into a NKX2-1⁺/FOXA2⁻ population, that is located both laterally and inside the light induced SHH organizer (Fig. 2D, E, lower panel). Also, there is a population of cells NG-T2A-SHH⁺/FOXA2⁺ /NKX2-1⁻ (Fig. 2E, F, Supplementary Fig. 4A). NKX2-1 domain juxtaposed to the SHH organizer is induced independently from the size of the SHH domain (Supplementary Fig. 4B). The RNA expression of the NG, the exogenous SHH and its downstream target GLI1 correlate with the expression of the NG-T2A-SHH module, validating our co-expression strategy (Supplementary Fig. 4C). Therefore, our analysis revealed a proximal distal pattern of ventral cell fates from the SHH organizer during neural induction. Spatiotemporal control of SHH induces ventral neural fates that are organized in a 2D space in vitro, resembling M-L and D-V neural populations (Supplementary Fig. 2D). Moreover, it validates the functionality of our optogenetic tool for its ability to induce and self-organize discrete fates in hESCs with a simple blue light stimulation.

Molecular characterization of light-induced neural fates

In order to precisely and unbiasedly identify the cell types present in our light-induced, self-organizing neural tissue, we characterized their transcriptome using scRNA-seq from two independent biological replica (12097 and 6207 cells) (Fig. 3A, B, Supplementary Fig. 5A–C). Cells were differentiated and stimulated with blue light as previously described in Fig. 2A and harvested at day 14 for single-cell RNA-sequencing analysis. Differentially expressed genes based on leiden clustering, RNA-velocity trajectories and cell-cycle predictions were used to classify 14 distinct cell identities: (i) FOXA2⁺/ARX Floor Plate; (ii) NKX2-1⁺/RAX⁺/SIX6⁺ Ventral tuberal hypothalamic progenitors; (iii) NKX2-1⁺/FOXG1⁺ Ventral telencephalic progenitors, (iv) NKX2-1⁺/NHLH2⁺/OTP⁺ Ventral hypothalamic neurons; (v) TFAP2A⁺/KRT19⁺ Superficial ectoderm; (vi) SOX10⁺/PLP1⁺ Neural Crest, (vii) PAX6⁺/EMX2⁺/OTX2⁺ Dorsal forebrain progenitors; (viii) IRX3⁺/OLIG3⁺ Dorsal thalamic progenitors; (ix) TBR1⁺/LHX1⁺ Dorsal Neurons_1; (x) HES6⁺/DLL3⁺ Dorsal Neurons_2; dorsal and ventral proliferating progenitors, (xi) Dorsal, (xii) Dorsal thalamic, (xiii) Ventral and (xiv) an unidentified population (UnId) (Fig. 3B, Supplementary Fig 5C-D, Source Data file). This analysis revealed the presence of multiple cell types that demarcate different domains along the embryonic D-V and A-P axes in agreement with what was previously shown by asymmetric SHH stimulation in 3D organoids¹⁸. No endoderm, mesoderm or extra-embryonic markers were detected, locating cells in the ectodermal compartment.

A Schematization of the scRNA-seq profiling strategy. B Left panel: UMAP plot labeled with the identified cell identities and RNA velocity vectors: (i) FOXA2⁺/ARX Floor Plate, (ii) NKX2-1⁺/RAX⁺/SIX6⁺ Ventral tuberal hypothalamic progenitors, (iii) NKX2-1⁺/FOXG1⁺ Ventral telencephalic progenitors, (iv) NKX2-1⁺/NHLH2⁺/OTP⁺ Ventral hypothalamic neurons, (v) TFAP2A⁺/KRT19⁺ Superficial ectoderm, (vi) SOX10⁺/PLP1⁺ Neural Crest, (vii) PAX6⁺/EMX2⁺/OTX2⁺ Dorsal forebrain progenitors, (viii) IRX3⁺/OLIG3⁺ Dorsal thalamic progenitors, (ix) TBR1⁺/LHX1⁺ Dorsal Neurons_1, (x) HES6⁺/DLL3⁺ Dorsal Neurons_2, dorsal and ventral proliferating progenitors, (xi) Dorsal, (xii) Dorsal thalamic, (xiii) Ventral and (xiv) an UnIdentified population (UnId). Right panel: z-score scaled heatmap of marker genes used for cell identities classification. C UMAP plot displaying SHH responsive and unresponsive status, and GLI3 and PTCH1 expression level. SHH responsive and unresponsive status have been computed imposing a threshold based on the normalized distribution of GLI3, GAS1 and PTCH1 counts, where GLI3-GAS1^low and PTCH1^high represent SHH responsive while GLI3-GAS1^high, PTCH1^low unresponsive. D qRT-PCR analysis of SHH responsive genes (GLI1 and GLI3) for NG-T2A-SHH and NG-CNTRL at day 14 of differentiation. The histogram displays the average and SD of differentiated cells exposed to light stimulation or dark control (n = 2 biologically independent samples). .

To correlate the timing of our self-organizing in vitro tissues with in-vivo events, we integrated our dataset with a scRNA-seq collection of mouse brain samples at different time points, E8.5, E10, E12, E12.5, E13, and E15³⁵. scRNA-seq transcriptomics of the human light-induced cells, grouped as neural precours (NPCs), floor plate, superficial ectoderm, neurons and UnId cells, integrate with the in-vivo mouse brain developmental atlas (Supplementary Fig. 6A). The human NPCs display high correlation with the Radial glia population at E10, while the in vitro derived neurons display high correlation with the mouse neuronal category at E12.5-E15 (Supplementary Fig. 6B, C), suggesting a temporal match with human development at PCW4-5 (https://embryology.med.unsw.edu.au/embryology/index.php/Carnegie_Stage_Comparison).

The expression of GLI3, GAS1 and PTCH was used to identify SHH receiving cells. In agreement with literature, we identified a GLI3-GAS1^low/PTCH1^high population as SHH stimulated while GLI3-GAS1^high/PTCH^low cells as unstimulated (Fig. 3C, Supplementary Fig. 7A). We also validated the specific induction of SHH signaling in the NG-T2A-SHH line compared to a NG-CNTRL line by testing GLI1 and GLI3 expression at day 14 using qRT-PCR (Fig. 3D). Among the SHH induced populations, scRNA-seq analysis confirmed the presence of a FOXA2⁺/ARX⁺, floor plate population and revealed the identity of four distinct NKX2-1⁺ populations (Fig. 3B, C, E). The first is representative of the tuberal hypothalamic neural progenitors positive for NKX2-1, SIX6, SIX3, RAX (Fig. 3C, Fig. 4A–C, Supplementary Fig. 7C)^36,37,38, the second is the NKX2-1⁺ FOXG1⁺ population representative of the ventral telencephalic population (Fig. 4A, B)^16,18, the third is a ventral population of proliferating progenitors and the fourth is a small population of ventral neurons that we classified as hypothalamic neurons positive for NKX2-1, OTP, NHLH2 (Fig. 3B, C, Source Data file)^37,38. Among the SHH unstimulated cells, we identified dorsal populations that consist of forebrain progenitors (PAX6⁺/EMX2⁺/OTX2⁺), thalamic progenitors (IRX3⁺/OLIG3⁺), two neuronal populations (TBR1⁺/LHX1⁺ and HES6⁺/DLL3⁺), and non-neural ectoderm derivatives such as superficial ectoderm and neural crest (Fig. 3B, Supplementary Fig. 5C, Source Data file). The non-neural ectoderm population derive mostly from the plate edge independently from the light organizer (Supplementary Fig. 7B).

A UMAP plot showing a selected population of cell NKX2-1⁺. Expression of markers that distinguish hypothalamic and telencephalic populations (NKX2-1, SIX6 and FOXG1). B Immunostaining shows the spatial segregation of ventral population arising from a light-induced SHH source at day 14. (Green NG, Magenta NKX2.1, Cyan SIX6, Red FOXG1, Gray DAPI), Scale bar = 100 μm. C Cumulative fluorescence intensity analysis (line profile) over the x-axis. x-axis displays the linear distance in μm. y-axis shows the cumulative fluorescent intensity profile in arbitrary units for each channel. Line profile shows the average (line) and SD (area) for each channel. The line profile is color-coded as the immunofluorescent channels, NG-Green, NKX2-1-Magenta, SIX6-Cyan, FOXG1-Red at day 14 quantification (n = 3 biologically independent samples). D Immunostaining shows OTP positive cells induced in proximity of the NG-T2A-SHH organizer but not in the NG-CNTRL. Light induced SHH drive the self-organization of both neural progenitors and neurons (Green NG, Magenta OTP, Gray DAPI), Scale bar = 100 μm.

Immunostaining for specific markers, SIX6, RAX, NKX2-2 and FOXG1, revealed the spatial segregation of telencephalic and hypothalamic territory (Fig. 4B, C, Supplementary Fig. 7C, D, E). We further showed that light-modulation of SHH not only self-organizes telencephalic and hypothalamic progenitors, but also neurons, since hypothalamic OTP⁺ neurons are preferentially located proximal to the light-induced organizer (Fig. 4D). While the differentiation of hypothalamic cells has been previously observed in traditional cell culture or 3D organoids^39,40,41, confinement of a SHH source in 2D instructs the self-organization of a ventral telencephalic-hypothalamic structures that are spatially organized in monolayer.

Finally, the hypothalamic marker genes used in this study were in-vivo validated for their specific expression in the human fetal hypothalamus at PCW10 (Supplementary Fig. 8A)⁴². In order to capture the gene expression modules that are shared between our in vitro dataset and the fetal hypothalamus, we computed the gene regulatory networks (regulons) in each dataset using pySCENIC⁴³. Among the 427 active regulons identified in the human fetal hypothalamus dataset, the 72.5% (310 regulons) are shared with our in vitro dataset (Supplementary Fig. 8B, Source Data file). Based on RNA velocity analysis, we identified in the light-induced scRNA-seq dataset a ventral differentiation trajectory that starts from the ventral proliferating progenitors and ends at the ventral hypothalamic neurons (Fig. 3B). Performing gene ontology analysis of genes that are differentially expressed along this trajectory, we identified waves of gene expression linked to cell-cycle regulation, neural progenitor expansion and neuronal maturation (Supplementary Fig. 8C). We explored whether some important gene expression patterns recently described in the context of human fetal hypothalamic development were recapitulated in our model, TTYH1, HMGA2 and MYBL2 show the same progenitor-neuronal trend observed in the human fetal hypothalamus at PCW10 (Supplementary Fig. 8D).

Collectively, our experiments demonstrate the ability to generate organizing centers by a simple blue light-induction of a morphogen, which specify a morphogenic source that patterns discrete cell types along a proximal distal axis in space, and lineage trajectory in time.